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_adjust(__u32 *bitmask, struct inode *inode,
 116                 struct nfs_server *server,
 117                 struct nfs4_label *label);
 118
 119 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 120 static inline struct nfs4_label *
 121 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 122         struct iattr *sattr, struct nfs4_label *label)
 123 {
 124         int err;
 125
 126         if (label == NULL)
 127                 return NULL;
 128
 129         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 130                 return NULL;
 131
 132         err = security_dentry_init_security(dentry, sattr->ia_mode,
 133                                 &dentry->d_name, (void **)&label->label, &label->len);
 134         if (err == 0)
 135                 return label;
 136
 137         return NULL;
 138 }
 139 static inline void
 140 nfs4_label_release_security(struct nfs4_label *label)
 141 {
 142         if (label)
 143                 security_release_secctx(label->label, label->len);
 144 }
 145 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 146 {
 147         if (label)
 148                 return server->attr_bitmask;
 149
 150         return server->attr_bitmask_nl;
 151 }
 152 #else
 153 static inline struct nfs4_label *
 154 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 155         struct iattr *sattr, struct nfs4_label *l)
 156 { return NULL; }
 157 static inline void
 158 nfs4_label_release_security(struct nfs4_label *label)
 159 { return; }
 160 static inline u32 *
 161 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 162 { return server->attr_bitmask; }
 163 #endif
 164
 165 /* Prevent leaks of NFSv4 errors into userland */
 166 static int nfs4_map_errors(int err)
 167 {
 168         if (err >= -1000)
 169                 return err;
 170         switch (err) {
 171         case -NFS4ERR_RESOURCE:
 172         case -NFS4ERR_LAYOUTTRYLATER:
 173         case -NFS4ERR_RECALLCONFLICT:
 174                 return -EREMOTEIO;
 175         case -NFS4ERR_WRONGSEC:
 176         case -NFS4ERR_WRONG_CRED:
 177                 return -EPERM;
 178         case -NFS4ERR_BADOWNER:
 179         case -NFS4ERR_BADNAME:
 180                 return -EINVAL;
 181         case -NFS4ERR_SHARE_DENIED:
 182                 return -EACCES;
 183         case -NFS4ERR_MINOR_VERS_MISMATCH:
 184                 return -EPROTONOSUPPORT;
 185         case -NFS4ERR_FILE_OPEN:
 186                 return -EBUSY;
 187         case -NFS4ERR_NOT_SAME:
 188                 return -ENOTSYNC;
 189         default:
 190                 dprintk("%s could not handle NFSv4 error %d\n",
 191                                 __func__, -err);
 192                 break;
 193         }
 194         return -EIO;
 195 }
 196
 197 /*
 198  * This is our standard bitmap for GETATTR requests.
 199  */
 200 const u32 nfs4_fattr_bitmap[3] = {
 201         FATTR4_WORD0_TYPE
 202         | FATTR4_WORD0_CHANGE
 203         | FATTR4_WORD0_SIZE
 204         | FATTR4_WORD0_FSID
 205         | FATTR4_WORD0_FILEID,
 206         FATTR4_WORD1_MODE
 207         | FATTR4_WORD1_NUMLINKS
 208         | FATTR4_WORD1_OWNER
 209         | FATTR4_WORD1_OWNER_GROUP
 210         | FATTR4_WORD1_RAWDEV
 211         | FATTR4_WORD1_SPACE_USED
 212         | FATTR4_WORD1_TIME_ACCESS
 213         | FATTR4_WORD1_TIME_METADATA
 214         | FATTR4_WORD1_TIME_MODIFY
 215         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 216 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 217         FATTR4_WORD2_SECURITY_LABEL
 218 #endif
 219 };
 220
 221 static const u32 nfs4_pnfs_open_bitmap[3] = {
 222         FATTR4_WORD0_TYPE
 223         | FATTR4_WORD0_CHANGE
 224         | FATTR4_WORD0_SIZE
 225         | FATTR4_WORD0_FSID
 226         | FATTR4_WORD0_FILEID,
 227         FATTR4_WORD1_MODE
 228         | FATTR4_WORD1_NUMLINKS
 229         | FATTR4_WORD1_OWNER
 230         | FATTR4_WORD1_OWNER_GROUP
 231         | FATTR4_WORD1_RAWDEV
 232         | FATTR4_WORD1_SPACE_USED
 233         | FATTR4_WORD1_TIME_ACCESS
 234         | FATTR4_WORD1_TIME_METADATA
 235         | FATTR4_WORD1_TIME_MODIFY,
 236         FATTR4_WORD2_MDSTHRESHOLD
 237 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 238         | FATTR4_WORD2_SECURITY_LABEL
 239 #endif
 240 };
 241
 242 static const u32 nfs4_open_noattr_bitmap[3] = {
 243         FATTR4_WORD0_TYPE
 244         | FATTR4_WORD0_FILEID,
 245 };
 246
 247 const u32 nfs4_statfs_bitmap[3] = {
 248         FATTR4_WORD0_FILES_AVAIL
 249         | FATTR4_WORD0_FILES_FREE
 250         | FATTR4_WORD0_FILES_TOTAL,
 251         FATTR4_WORD1_SPACE_AVAIL
 252         | FATTR4_WORD1_SPACE_FREE
 253         | FATTR4_WORD1_SPACE_TOTAL
 254 };
 255
 256 const u32 nfs4_pathconf_bitmap[3] = {
 257         FATTR4_WORD0_MAXLINK
 258         | FATTR4_WORD0_MAXNAME,
 259         0
 260 };
 261
 262 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 263                         | FATTR4_WORD0_MAXREAD
 264                         | FATTR4_WORD0_MAXWRITE
 265                         | FATTR4_WORD0_LEASE_TIME,
 266                         FATTR4_WORD1_TIME_DELTA
 267                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 268                         FATTR4_WORD2_LAYOUT_BLKSIZE
 269                         | FATTR4_WORD2_CLONE_BLKSIZE
 270                         | FATTR4_WORD2_XATTR_SUPPORT
 271 };
 272
 273 const u32 nfs4_fs_locations_bitmap[3] = {
 274         FATTR4_WORD0_CHANGE
 275         | FATTR4_WORD0_SIZE
 276         | FATTR4_WORD0_FSID
 277         | FATTR4_WORD0_FILEID
 278         | FATTR4_WORD0_FS_LOCATIONS,
 279         FATTR4_WORD1_OWNER
 280         | FATTR4_WORD1_OWNER_GROUP
 281         | FATTR4_WORD1_RAWDEV
 282         | FATTR4_WORD1_SPACE_USED
 283         | FATTR4_WORD1_TIME_ACCESS
 284         | FATTR4_WORD1_TIME_METADATA
 285         | FATTR4_WORD1_TIME_MODIFY
 286         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 287 };
 288
 289 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 290                 struct inode *inode)
 291 {
 292         unsigned long cache_validity;
 293
 294         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 295         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 296                 return;
 297
 298         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 299         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 300                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 301                                 | NFS_INO_INVALID_SIZE);
 302
 303         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 304                 dst[0] &= ~FATTR4_WORD0_SIZE;
 305
 306         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 307                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 308 }
 309
 310 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 311                 const __u32 *src, struct inode *inode)
 312 {
 313         nfs4_bitmap_copy_adjust(dst, src, inode);
 314 }
 315
 316 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 317                 struct nfs4_readdir_arg *readdir)
 318 {
 319         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 320         __be32 *start, *p;
 321
 322         if (cookie > 2) {
 323                 readdir->cookie = cookie;
 324                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 325                 return;
 326         }
 327
 328         readdir->cookie = 0;
 329         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 330         if (cookie == 2)
 331                 return;
 332
 333         /*
 334          * NFSv4 servers do not return entries for '.' and '..'
 335          * Therefore, we fake these entries here.  We let '.'
 336          * have cookie 0 and '..' have cookie 1.  Note that
 337          * when talking to the server, we always send cookie 0
 338          * instead of 1 or 2.
 339          */
 340         start = p = kmap_atomic(*readdir->pages);
 341
 342         if (cookie == 0) {
 343                 *p++ = xdr_one;                                  /* next */
 344                 *p++ = xdr_zero;                   /* cookie, first word */
 345                 *p++ = xdr_one;                   /* cookie, second word */
 346                 *p++ = xdr_one;                             /* entry len */
 347                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 348                 p++;
 349                 *p++ = xdr_one;                         /* bitmap length */
 350                 *p++ = htonl(attrs);                           /* bitmap */
 351                 *p++ = htonl(12);             /* attribute buffer length */
 352                 *p++ = htonl(NF4DIR);
 353                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 354         }
 355
 356         *p++ = xdr_one;                                  /* next */
 357         *p++ = xdr_zero;                   /* cookie, first word */
 358         *p++ = xdr_two;                   /* cookie, second word */
 359         *p++ = xdr_two;                             /* entry len */
 360         memcpy(p, "..\0\0", 4);                         /* entry */
 361         p++;
 362         *p++ = xdr_one;                         /* bitmap length */
 363         *p++ = htonl(attrs);                           /* bitmap */
 364         *p++ = htonl(12);             /* attribute buffer length */
 365         *p++ = htonl(NF4DIR);
 366         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 367
 368         readdir->pgbase = (char *)p - (char *)start;
 369         readdir->count -= readdir->pgbase;
 370         kunmap_atomic(start);
 371 }
 372
 373 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 374                 nfs4_stateid *stateid,
 375                 const struct cred *cred)
 376 {
 377         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 378
 379         ops->test_and_free_expired(server, stateid, cred);
 380 }
 381
 382 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 383                 nfs4_stateid *stateid,
 384                 const struct cred *cred)
 385 {
 386         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 387         nfs4_test_and_free_stateid(server, stateid, cred);
 388 }
 389
 390 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 391                 const nfs4_stateid *stateid,
 392                 const struct cred *cred)
 393 {
 394         nfs4_stateid tmp;
 395
 396         nfs4_stateid_copy(&tmp, stateid);
 397         __nfs4_free_revoked_stateid(server, &tmp, cred);
 398 }
 399
 400 static long nfs4_update_delay(long *timeout)
 401 {
 402         long ret;
 403         if (!timeout)
 404                 return NFS4_POLL_RETRY_MAX;
 405         if (*timeout <= 0)
 406                 *timeout = NFS4_POLL_RETRY_MIN;
 407         if (*timeout > NFS4_POLL_RETRY_MAX)
 408                 *timeout = NFS4_POLL_RETRY_MAX;
 409         ret = *timeout;
 410         *timeout <<= 1;
 411         return ret;
 412 }
 413
 414 static int nfs4_delay_killable(long *timeout)
 415 {
 416         might_sleep();
 417
 418         freezable_schedule_timeout_killable_unsafe(
 419                 nfs4_update_delay(timeout));
 420         if (!__fatal_signal_pending(current))
 421                 return 0;
 422         return -EINTR;
 423 }
 424
 425 static int nfs4_delay_interruptible(long *timeout)
 426 {
 427         might_sleep();
 428
 429         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
 430         if (!signal_pending(current))
 431                 return 0;
 432         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 433 }
 434
 435 static int nfs4_delay(long *timeout, bool interruptible)
 436 {
 437         if (interruptible)
 438                 return nfs4_delay_interruptible(timeout);
 439         return nfs4_delay_killable(timeout);
 440 }
 441
 442 static const nfs4_stateid *
 443 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 444 {
 445         if (!stateid)
 446                 return NULL;
 447         switch (stateid->type) {
 448         case NFS4_OPEN_STATEID_TYPE:
 449         case NFS4_LOCK_STATEID_TYPE:
 450         case NFS4_DELEGATION_STATEID_TYPE:
 451                 return stateid;
 452         default:
 453                 break;
 454         }
 455         return NULL;
 456 }
 457
 458 /* This is the error handling routine for processes that are allowed
 459  * to sleep.
 460  */
 461 static int nfs4_do_handle_exception(struct nfs_server *server,
 462                 int errorcode, struct nfs4_exception *exception)
 463 {
 464         struct nfs_client *clp = server->nfs_client;
 465         struct nfs4_state *state = exception->state;
 466         const nfs4_stateid *stateid;
 467         struct inode *inode = exception->inode;
 468         int ret = errorcode;
 469
 470         exception->delay = 0;
 471         exception->recovering = 0;
 472         exception->retry = 0;
 473
 474         stateid = nfs4_recoverable_stateid(exception->stateid);
 475         if (stateid == NULL && state != NULL)
 476                 stateid = nfs4_recoverable_stateid(&state->stateid);
 477
 478         switch(errorcode) {
 479                 case 0:
 480                         return 0;
 481                 case -NFS4ERR_BADHANDLE:
 482                 case -ESTALE:
 483                         if (inode != NULL && S_ISREG(inode->i_mode))
 484                                 pnfs_destroy_layout(NFS_I(inode));
 485                         break;
 486                 case -NFS4ERR_DELEG_REVOKED:
 487                 case -NFS4ERR_ADMIN_REVOKED:
 488                 case -NFS4ERR_EXPIRED:
 489                 case -NFS4ERR_BAD_STATEID:
 490                 case -NFS4ERR_PARTNER_NO_AUTH:
 491                         if (inode != NULL && stateid != NULL) {
 492                                 nfs_inode_find_state_and_recover(inode,
 493                                                 stateid);
 494                                 goto wait_on_recovery;
 495                         }
 496                         fallthrough;
 497                 case -NFS4ERR_OPENMODE:
 498                         if (inode) {
 499                                 int err;
 500
 501                                 err = nfs_async_inode_return_delegation(inode,
 502                                                 stateid);
 503                                 if (err == 0)
 504                                         goto wait_on_recovery;
 505                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 506                                         exception->retry = 1;
 507                                         break;
 508                                 }
 509                         }
 510                         if (state == NULL)
 511                                 break;
 512                         ret = nfs4_schedule_stateid_recovery(server, state);
 513                         if (ret < 0)
 514                                 break;
 515                         goto wait_on_recovery;
 516                 case -NFS4ERR_STALE_STATEID:
 517                 case -NFS4ERR_STALE_CLIENTID:
 518                         nfs4_schedule_lease_recovery(clp);
 519                         goto wait_on_recovery;
 520                 case -NFS4ERR_MOVED:
 521                         ret = nfs4_schedule_migration_recovery(server);
 522                         if (ret < 0)
 523                                 break;
 524                         goto wait_on_recovery;
 525                 case -NFS4ERR_LEASE_MOVED:
 526                         nfs4_schedule_lease_moved_recovery(clp);
 527                         goto wait_on_recovery;
 528 #if defined(CONFIG_NFS_V4_1)
 529                 case -NFS4ERR_BADSESSION:
 530                 case -NFS4ERR_BADSLOT:
 531                 case -NFS4ERR_BAD_HIGH_SLOT:
 532                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 533                 case -NFS4ERR_DEADSESSION:
 534                 case -NFS4ERR_SEQ_FALSE_RETRY:
 535                 case -NFS4ERR_SEQ_MISORDERED:
 536                         /* Handled in nfs41_sequence_process() */
 537                         goto wait_on_recovery;
 538 #endif /* defined(CONFIG_NFS_V4_1) */
 539                 case -NFS4ERR_FILE_OPEN:
 540                         if (exception->timeout > HZ) {
 541                                 /* We have retried a decent amount, time to
 542                                  * fail
 543                                  */
 544                                 ret = -EBUSY;
 545                                 break;
 546                         }
 547                         fallthrough;
 548                 case -NFS4ERR_DELAY:
 549                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 550                         fallthrough;
 551                 case -NFS4ERR_GRACE:
 552                 case -NFS4ERR_LAYOUTTRYLATER:
 553                 case -NFS4ERR_RECALLCONFLICT:
 554                         exception->delay = 1;
 555                         return 0;
 556
 557                 case -NFS4ERR_RETRY_UNCACHED_REP:
 558                 case -NFS4ERR_OLD_STATEID:
 559                         exception->retry = 1;
 560                         break;
 561                 case -NFS4ERR_BADOWNER:
 562                         /* The following works around a Linux server bug! */
 563                 case -NFS4ERR_BADNAME:
 564                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 565                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 566                                 exception->retry = 1;
 567                                 printk(KERN_WARNING "NFS: v4 server %s "
 568                                                 "does not accept raw "
 569                                                 "uid/gids. "
 570                                                 "Reenabling the idmapper.\n",
 571                                                 server->nfs_client->cl_hostname);
 572                         }
 573         }
 574         /* We failed to handle the error */
 575         return nfs4_map_errors(ret);
 576 wait_on_recovery:
 577         exception->recovering = 1;
 578         return 0;
 579 }
 580
 581 /* This is the error handling routine for processes that are allowed
 582  * to sleep.
 583  */
 584 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 585 {
 586         struct nfs_client *clp = server->nfs_client;
 587         int ret;
 588
 589         ret = nfs4_do_handle_exception(server, errorcode, exception);
 590         if (exception->delay) {
 591                 ret = nfs4_delay(&exception->timeout,
 592                                 exception->interruptible);
 593                 goto out_retry;
 594         }
 595         if (exception->recovering) {
 596                 ret = nfs4_wait_clnt_recover(clp);
 597                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 598                         return -EIO;
 599                 goto out_retry;
 600         }
 601         return ret;
 602 out_retry:
 603         if (ret == 0)
 604                 exception->retry = 1;
 605         return ret;
 606 }
 607
 608 static int
 609 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 610                 int errorcode, struct nfs4_exception *exception)
 611 {
 612         struct nfs_client *clp = server->nfs_client;
 613         int ret;
 614
 615         ret = nfs4_do_handle_exception(server, errorcode, exception);
 616         if (exception->delay) {
 617                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 618                 goto out_retry;
 619         }
 620         if (exception->recovering) {
 621                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 622                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 623                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 624                 goto out_retry;
 625         }
 626         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 627                 ret = -EIO;
 628         return ret;
 629 out_retry:
 630         if (ret == 0) {
 631                 exception->retry = 1;
 632                 /*
 633                  * For NFS4ERR_MOVED, the client transport will need to
 634                  * be recomputed after migration recovery has completed.
 635                  */
 636                 if (errorcode == -NFS4ERR_MOVED)
 637                         rpc_task_release_transport(task);
 638         }
 639         return ret;
 640 }
 641
 642 int
 643 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 644                         struct nfs4_state *state, long *timeout)
 645 {
 646         struct nfs4_exception exception = {
 647                 .state = state,
 648         };
 649
 650         if (task->tk_status >= 0)
 651                 return 0;
 652         if (timeout)
 653                 exception.timeout = *timeout;
 654         task->tk_status = nfs4_async_handle_exception(task, server,
 655                         task->tk_status,
 656                         &exception);
 657         if (exception.delay && timeout)
 658                 *timeout = exception.timeout;
 659         if (exception.retry)
 660                 return -EAGAIN;
 661         return 0;
 662 }
 663
 664 /*
 665  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 666  * or 'false' otherwise.
 667  */
 668 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 669 {
 670         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 671         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 672 }
 673
 674 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 675 {
 676         spin_lock(&clp->cl_lock);
 677         if (time_before(clp->cl_last_renewal,timestamp))
 678                 clp->cl_last_renewal = timestamp;
 679         spin_unlock(&clp->cl_lock);
 680 }
 681
 682 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 683 {
 684         struct nfs_client *clp = server->nfs_client;
 685
 686         if (!nfs4_has_session(clp))
 687                 do_renew_lease(clp, timestamp);
 688 }
 689
 690 struct nfs4_call_sync_data {
 691         const struct nfs_server *seq_server;
 692         struct nfs4_sequence_args *seq_args;
 693         struct nfs4_sequence_res *seq_res;
 694 };
 695
 696 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 697                         struct nfs4_sequence_res *res, int cache_reply,
 698                         int privileged)
 699 {
 700         args->sa_slot = NULL;
 701         args->sa_cache_this = cache_reply;
 702         args->sa_privileged = privileged;
 703
 704         res->sr_slot = NULL;
 705 }
 706
 707 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 708 {
 709         struct nfs4_slot *slot = res->sr_slot;
 710         struct nfs4_slot_table *tbl;
 711
 712         tbl = slot->table;
 713         spin_lock(&tbl->slot_tbl_lock);
 714         if (!nfs41_wake_and_assign_slot(tbl, slot))
 715                 nfs4_free_slot(tbl, slot);
 716         spin_unlock(&tbl->slot_tbl_lock);
 717
 718         res->sr_slot = NULL;
 719 }
 720
 721 static int nfs40_sequence_done(struct rpc_task *task,
 722                                struct nfs4_sequence_res *res)
 723 {
 724         if (res->sr_slot != NULL)
 725                 nfs40_sequence_free_slot(res);
 726         return 1;
 727 }
 728
 729 #if defined(CONFIG_NFS_V4_1)
 730
 731 static void nfs41_release_slot(struct nfs4_slot *slot)
 732 {
 733         struct nfs4_session *session;
 734         struct nfs4_slot_table *tbl;
 735         bool send_new_highest_used_slotid = false;
 736
 737         if (!slot)
 738                 return;
 739         tbl = slot->table;
 740         session = tbl->session;
 741
 742         /* Bump the slot sequence number */
 743         if (slot->seq_done)
 744                 slot->seq_nr++;
 745         slot->seq_done = 0;
 746
 747         spin_lock(&tbl->slot_tbl_lock);
 748         /* Be nice to the server: try to ensure that the last transmitted
 749          * value for highest_user_slotid <= target_highest_slotid
 750          */
 751         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 752                 send_new_highest_used_slotid = true;
 753
 754         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 755                 send_new_highest_used_slotid = false;
 756                 goto out_unlock;
 757         }
 758         nfs4_free_slot(tbl, slot);
 759
 760         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 761                 send_new_highest_used_slotid = false;
 762 out_unlock:
 763         spin_unlock(&tbl->slot_tbl_lock);
 764         if (send_new_highest_used_slotid)
 765                 nfs41_notify_server(session->clp);
 766         if (waitqueue_active(&tbl->slot_waitq))
 767                 wake_up_all(&tbl->slot_waitq);
 768 }
 769
 770 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 771 {
 772         nfs41_release_slot(res->sr_slot);
 773         res->sr_slot = NULL;
 774 }
 775
 776 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 777                 u32 seqnr)
 778 {
 779         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 780                 slot->seq_nr_highest_sent = seqnr;
 781 }
 782 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 783                 u32 seqnr)
 784 {
 785         slot->seq_nr_highest_sent = seqnr;
 786         slot->seq_nr_last_acked = seqnr;
 787 }
 788
 789 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 790                                 struct nfs4_slot *slot)
 791 {
 792         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 793         if (!IS_ERR(task))
 794                 rpc_put_task_async(task);
 795 }
 796
 797 static int nfs41_sequence_process(struct rpc_task *task,
 798                 struct nfs4_sequence_res *res)
 799 {
 800         struct nfs4_session *session;
 801         struct nfs4_slot *slot = res->sr_slot;
 802         struct nfs_client *clp;
 803         int status;
 804         int ret = 1;
 805
 806         if (slot == NULL)
 807                 goto out_noaction;
 808         /* don't increment the sequence number if the task wasn't sent */
 809         if (!RPC_WAS_SENT(task) || slot->seq_done)
 810                 goto out;
 811
 812         session = slot->table->session;
 813         clp = session->clp;
 814
 815         trace_nfs4_sequence_done(session, res);
 816
 817         status = res->sr_status;
 818         if (task->tk_status == -NFS4ERR_DEADSESSION)
 819                 status = -NFS4ERR_DEADSESSION;
 820
 821         /* Check the SEQUENCE operation status */
 822         switch (status) {
 823         case 0:
 824                 /* Mark this sequence number as having been acked */
 825                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 826                 /* Update the slot's sequence and clientid lease timer */
 827                 slot->seq_done = 1;
 828                 do_renew_lease(clp, res->sr_timestamp);
 829                 /* Check sequence flags */
 830                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 831                                 !!slot->privileged);
 832                 nfs41_update_target_slotid(slot->table, slot, res);
 833                 break;
 834         case 1:
 835                 /*
 836                  * sr_status remains 1 if an RPC level error occurred.
 837                  * The server may or may not have processed the sequence
 838                  * operation..
 839                  */
 840                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 841                 slot->seq_done = 1;
 842                 goto out;
 843         case -NFS4ERR_DELAY:
 844                 /* The server detected a resend of the RPC call and
 845                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 846                  * of RFC5661.
 847                  */
 848                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 849                         __func__,
 850                         slot->slot_nr,
 851                         slot->seq_nr);
 852                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 853                 goto out_retry;
 854         case -NFS4ERR_RETRY_UNCACHED_REP:
 855         case -NFS4ERR_SEQ_FALSE_RETRY:
 856                 /*
 857                  * The server thinks we tried to replay a request.
 858                  * Retry the call after bumping the sequence ID.
 859                  */
 860                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 861                 goto retry_new_seq;
 862         case -NFS4ERR_BADSLOT:
 863                 /*
 864                  * The slot id we used was probably retired. Try again
 865                  * using a different slot id.
 866                  */
 867                 if (slot->slot_nr < slot->table->target_highest_slotid)
 868                         goto session_recover;
 869                 goto retry_nowait;
 870         case -NFS4ERR_SEQ_MISORDERED:
 871                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 872                 /*
 873                  * Were one or more calls using this slot interrupted?
 874                  * If the server never received the request, then our
 875                  * transmitted slot sequence number may be too high. However,
 876                  * if the server did receive the request then it might
 877                  * accidentally give us a reply with a mismatched operation.
 878                  * We can sort this out by sending a lone sequence operation
 879                  * to the server on the same slot.
 880                  */
 881                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 882                         slot->seq_nr--;
 883                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 884                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 885                                 res->sr_slot = NULL;
 886                         }
 887                         goto retry_nowait;
 888                 }
 889                 /*
 890                  * RFC5661:
 891                  * A retry might be sent while the original request is
 892                  * still in progress on the replier. The replier SHOULD
 893                  * deal with the issue by returning NFS4ERR_DELAY as the
 894                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 895                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 896                  *
 897                  * Restart the search after a delay.
 898                  */
 899                 slot->seq_nr = slot->seq_nr_highest_sent;
 900                 goto out_retry;
 901         case -NFS4ERR_BADSESSION:
 902         case -NFS4ERR_DEADSESSION:
 903         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 904                 goto session_recover;
 905         default:
 906                 /* Just update the slot sequence no. */
 907                 slot->seq_done = 1;
 908         }
 909 out:
 910         /* The session may be reset by one of the error handlers. */
 911         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 912 out_noaction:
 913         return ret;
 914 session_recover:
 915         nfs4_schedule_session_recovery(session, status);
 916         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 917         nfs41_sequence_free_slot(res);
 918         goto out;
 919 retry_new_seq:
 920         ++slot->seq_nr;
 921 retry_nowait:
 922         if (rpc_restart_call_prepare(task)) {
 923                 nfs41_sequence_free_slot(res);
 924                 task->tk_status = 0;
 925                 ret = 0;
 926         }
 927         goto out;
 928 out_retry:
 929         if (!rpc_restart_call(task))
 930                 goto out;
 931         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 932         return 0;
 933 }
 934
 935 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 936 {
 937         if (!nfs41_sequence_process(task, res))
 938                 return 0;
 939         if (res->sr_slot != NULL)
 940                 nfs41_sequence_free_slot(res);
 941         return 1;
 942
 943 }
 944 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 945
 946 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 947 {
 948         if (res->sr_slot == NULL)
 949                 return 1;
 950         if (res->sr_slot->table->session != NULL)
 951                 return nfs41_sequence_process(task, res);
 952         return nfs40_sequence_done(task, res);
 953 }
 954
 955 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 956 {
 957         if (res->sr_slot != NULL) {
 958                 if (res->sr_slot->table->session != NULL)
 959                         nfs41_sequence_free_slot(res);
 960                 else
 961                         nfs40_sequence_free_slot(res);
 962         }
 963 }
 964
 965 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 966 {
 967         if (res->sr_slot == NULL)
 968                 return 1;
 969         if (!res->sr_slot->table->session)
 970                 return nfs40_sequence_done(task, res);
 971         return nfs41_sequence_done(task, res);
 972 }
 973 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 974
 975 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 976 {
 977         struct nfs4_call_sync_data *data = calldata;
 978
 979         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 980
 981         nfs4_setup_sequence(data->seq_server->nfs_client,
 982                             data->seq_args, data->seq_res, task);
 983 }
 984
 985 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 986 {
 987         struct nfs4_call_sync_data *data = calldata;
 988
 989         nfs41_sequence_done(task, data->seq_res);
 990 }
 991
 992 static const struct rpc_call_ops nfs41_call_sync_ops = {
 993         .rpc_call_prepare = nfs41_call_sync_prepare,
 994         .rpc_call_done = nfs41_call_sync_done,
 995 };
 996
 997 #else   /* !CONFIG_NFS_V4_1 */
 998
 999 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1000 {
1001         return nfs40_sequence_done(task, res);
1002 }
1003
1004 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1005 {
1006         if (res->sr_slot != NULL)
1007                 nfs40_sequence_free_slot(res);
1008 }
1009
1010 int nfs4_sequence_done(struct rpc_task *task,
1011                        struct nfs4_sequence_res *res)
1012 {
1013         return nfs40_sequence_done(task, res);
1014 }
1015 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1016
1017 #endif  /* !CONFIG_NFS_V4_1 */
1018
1019 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1020 {
1021         res->sr_timestamp = jiffies;
1022         res->sr_status_flags = 0;
1023         res->sr_status = 1;
1024 }
1025
1026 static
1027 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1028                 struct nfs4_sequence_res *res,
1029                 struct nfs4_slot *slot)
1030 {
1031         if (!slot)
1032                 return;
1033         slot->privileged = args->sa_privileged ? 1 : 0;
1034         args->sa_slot = slot;
1035
1036         res->sr_slot = slot;
1037 }
1038
1039 int nfs4_setup_sequence(struct nfs_client *client,
1040                         struct nfs4_sequence_args *args,
1041                         struct nfs4_sequence_res *res,
1042                         struct rpc_task *task)
1043 {
1044         struct nfs4_session *session = nfs4_get_session(client);
1045         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1046         struct nfs4_slot *slot;
1047
1048         /* slot already allocated? */
1049         if (res->sr_slot != NULL)
1050                 goto out_start;
1051
1052         if (session)
1053                 tbl = &session->fc_slot_table;
1054
1055         spin_lock(&tbl->slot_tbl_lock);
1056         /* The state manager will wait until the slot table is empty */
1057         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1058                 goto out_sleep;
1059
1060         slot = nfs4_alloc_slot(tbl);
1061         if (IS_ERR(slot)) {
1062                 if (slot == ERR_PTR(-ENOMEM))
1063                         goto out_sleep_timeout;
1064                 goto out_sleep;
1065         }
1066         spin_unlock(&tbl->slot_tbl_lock);
1067
1068         nfs4_sequence_attach_slot(args, res, slot);
1069
1070         trace_nfs4_setup_sequence(session, args);
1071 out_start:
1072         nfs41_sequence_res_init(res);
1073         rpc_call_start(task);
1074         return 0;
1075 out_sleep_timeout:
1076         /* Try again in 1/4 second */
1077         if (args->sa_privileged)
1078                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1079                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1080         else
1081                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1082                                 NULL, jiffies + (HZ >> 2));
1083         spin_unlock(&tbl->slot_tbl_lock);
1084         return -EAGAIN;
1085 out_sleep:
1086         if (args->sa_privileged)
1087                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1088                                 RPC_PRIORITY_PRIVILEGED);
1089         else
1090                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1091         spin_unlock(&tbl->slot_tbl_lock);
1092         return -EAGAIN;
1093 }
1094 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1095
1096 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1097 {
1098         struct nfs4_call_sync_data *data = calldata;
1099         nfs4_setup_sequence(data->seq_server->nfs_client,
1100                                 data->seq_args, data->seq_res, task);
1101 }
1102
1103 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1104 {
1105         struct nfs4_call_sync_data *data = calldata;
1106         nfs4_sequence_done(task, data->seq_res);
1107 }
1108
1109 static const struct rpc_call_ops nfs40_call_sync_ops = {
1110         .rpc_call_prepare = nfs40_call_sync_prepare,
1111         .rpc_call_done = nfs40_call_sync_done,
1112 };
1113
1114 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1115 {
1116         int ret;
1117         struct rpc_task *task;
1118
1119         task = rpc_run_task(task_setup);
1120         if (IS_ERR(task))
1121                 return PTR_ERR(task);
1122
1123         ret = task->tk_status;
1124         rpc_put_task(task);
1125         return ret;
1126 }
1127
1128 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1129                              struct nfs_server *server,
1130                              struct rpc_message *msg,
1131                              struct nfs4_sequence_args *args,
1132                              struct nfs4_sequence_res *res,
1133                              unsigned short task_flags)
1134 {
1135         struct nfs_client *clp = server->nfs_client;
1136         struct nfs4_call_sync_data data = {
1137                 .seq_server = server,
1138                 .seq_args = args,
1139                 .seq_res = res,
1140         };
1141         struct rpc_task_setup task_setup = {
1142                 .rpc_client = clnt,
1143                 .rpc_message = msg,
1144                 .callback_ops = clp->cl_mvops->call_sync_ops,
1145                 .callback_data = &data,
1146                 .flags = task_flags,
1147         };
1148
1149         return nfs4_call_sync_custom(&task_setup);
1150 }
1151
1152 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1153                                    struct nfs_server *server,
1154                                    struct rpc_message *msg,
1155                                    struct nfs4_sequence_args *args,
1156                                    struct nfs4_sequence_res *res)
1157 {
1158         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1159 }
1160
1161
1162 int nfs4_call_sync(struct rpc_clnt *clnt,
1163                    struct nfs_server *server,
1164                    struct rpc_message *msg,
1165                    struct nfs4_sequence_args *args,
1166                    struct nfs4_sequence_res *res,
1167                    int cache_reply)
1168 {
1169         nfs4_init_sequence(args, res, cache_reply, 0);
1170         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1171 }
1172
1173 static void
1174 nfs4_inc_nlink_locked(struct inode *inode)
1175 {
1176         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1177         inc_nlink(inode);
1178 }
1179
1180 static void
1181 nfs4_dec_nlink_locked(struct inode *inode)
1182 {
1183         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1184         drop_nlink(inode);
1185 }
1186
1187 static void
1188 nfs4_update_changeattr_locked(struct inode *inode,
1189                 struct nfs4_change_info *cinfo,
1190                 unsigned long timestamp, unsigned long cache_validity)
1191 {
1192         struct nfs_inode *nfsi = NFS_I(inode);
1193
1194         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1195                 | NFS_INO_INVALID_MTIME
1196                 | cache_validity;
1197
1198         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(inode)) {
1199                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1200                 nfsi->attrtimeo_timestamp = jiffies;
1201         } else {
1202                 if (S_ISDIR(inode->i_mode)) {
1203                         nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1204                         nfs_force_lookup_revalidate(inode);
1205                 } else {
1206                         if (!NFS_PROTO(inode)->have_delegation(inode,
1207                                                                FMODE_READ))
1208                                 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
1209                 }
1210
1211                 if (cinfo->before != inode_peek_iversion_raw(inode))
1212                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1213                                                 NFS_INO_INVALID_ACL |
1214                                                 NFS_INO_INVALID_XATTR;
1215         }
1216         inode_set_iversion_raw(inode, cinfo->after);
1217         nfsi->read_cache_jiffies = timestamp;
1218         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1219         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1220
1221         if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
1222                 nfs_fscache_invalidate(inode);
1223 }
1224
1225 void
1226 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1227                 unsigned long timestamp, unsigned long cache_validity)
1228 {
1229         spin_lock(&dir->i_lock);
1230         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1231         spin_unlock(&dir->i_lock);
1232 }
1233
1234 struct nfs4_open_createattrs {
1235         struct nfs4_label *label;
1236         struct iattr *sattr;
1237         const __u32 verf[2];
1238 };
1239
1240 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1241                 int err, struct nfs4_exception *exception)
1242 {
1243         if (err != -EINVAL)
1244                 return false;
1245         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1246                 return false;
1247         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1248         exception->retry = 1;
1249         return true;
1250 }
1251
1252 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1253 {
1254          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1255 }
1256
1257 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1258 {
1259         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1260
1261         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1262 }
1263
1264 static u32
1265 nfs4_map_atomic_open_share(struct nfs_server *server,
1266                 fmode_t fmode, int openflags)
1267 {
1268         u32 res = 0;
1269
1270         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1271         case FMODE_READ:
1272                 res = NFS4_SHARE_ACCESS_READ;
1273                 break;
1274         case FMODE_WRITE:
1275                 res = NFS4_SHARE_ACCESS_WRITE;
1276                 break;
1277         case FMODE_READ|FMODE_WRITE:
1278                 res = NFS4_SHARE_ACCESS_BOTH;
1279         }
1280         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1281                 goto out;
1282         /* Want no delegation if we're using O_DIRECT */
1283         if (openflags & O_DIRECT)
1284                 res |= NFS4_SHARE_WANT_NO_DELEG;
1285 out:
1286         return res;
1287 }
1288
1289 static enum open_claim_type4
1290 nfs4_map_atomic_open_claim(struct nfs_server *server,
1291                 enum open_claim_type4 claim)
1292 {
1293         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1294                 return claim;
1295         switch (claim) {
1296         default:
1297                 return claim;
1298         case NFS4_OPEN_CLAIM_FH:
1299                 return NFS4_OPEN_CLAIM_NULL;
1300         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1301                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1302         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1303                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1304         }
1305 }
1306
1307 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1308 {
1309         p->o_res.f_attr = &p->f_attr;
1310         p->o_res.f_label = p->f_label;
1311         p->o_res.seqid = p->o_arg.seqid;
1312         p->c_res.seqid = p->c_arg.seqid;
1313         p->o_res.server = p->o_arg.server;
1314         p->o_res.access_request = p->o_arg.access;
1315         nfs_fattr_init(&p->f_attr);
1316         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1317 }
1318
1319 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1320                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1321                 const struct nfs4_open_createattrs *c,
1322                 enum open_claim_type4 claim,
1323                 gfp_t gfp_mask)
1324 {
1325         struct dentry *parent = dget_parent(dentry);
1326         struct inode *dir = d_inode(parent);
1327         struct nfs_server *server = NFS_SERVER(dir);
1328         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1329         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1330         struct nfs4_opendata *p;
1331
1332         p = kzalloc(sizeof(*p), gfp_mask);
1333         if (p == NULL)
1334                 goto err;
1335
1336         p->f_label = nfs4_label_alloc(server, gfp_mask);
1337         if (IS_ERR(p->f_label))
1338                 goto err_free_p;
1339
1340         p->a_label = nfs4_label_alloc(server, gfp_mask);
1341         if (IS_ERR(p->a_label))
1342                 goto err_free_f;
1343
1344         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1345         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1346         if (IS_ERR(p->o_arg.seqid))
1347                 goto err_free_label;
1348         nfs_sb_active(dentry->d_sb);
1349         p->dentry = dget(dentry);
1350         p->dir = parent;
1351         p->owner = sp;
1352         atomic_inc(&sp->so_count);
1353         p->o_arg.open_flags = flags;
1354         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1355         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1356         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1357                         fmode, flags);
1358         if (flags & O_CREAT) {
1359                 p->o_arg.umask = current_umask();
1360                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1361                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1362                         p->o_arg.u.attrs = &p->attrs;
1363                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1364
1365                         memcpy(p->o_arg.u.verifier.data, c->verf,
1366                                         sizeof(p->o_arg.u.verifier.data));
1367                 }
1368         }
1369         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1370          * will return permission denied for all bits until close */
1371         if (!(flags & O_EXCL)) {
1372                 /* ask server to check for all possible rights as results
1373                  * are cached */
1374                 switch (p->o_arg.claim) {
1375                 default:
1376                         break;
1377                 case NFS4_OPEN_CLAIM_NULL:
1378                 case NFS4_OPEN_CLAIM_FH:
1379                         p->o_arg.access = NFS4_ACCESS_READ |
1380                                 NFS4_ACCESS_MODIFY |
1381                                 NFS4_ACCESS_EXTEND |
1382                                 NFS4_ACCESS_EXECUTE;
1383 #ifdef CONFIG_NFS_V4_2
1384                         if (server->caps & NFS_CAP_XATTR)
1385                                 p->o_arg.access |= NFS4_ACCESS_XAREAD |
1386                                     NFS4_ACCESS_XAWRITE |
1387                                     NFS4_ACCESS_XALIST;
1388 #endif
1389                 }
1390         }
1391         p->o_arg.clientid = server->nfs_client->cl_clientid;
1392         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1393         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1394         p->o_arg.name = &dentry->d_name;
1395         p->o_arg.server = server;
1396         p->o_arg.bitmask = nfs4_bitmask(server, label);
1397         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1398         switch (p->o_arg.claim) {
1399         case NFS4_OPEN_CLAIM_NULL:
1400         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1401         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1402                 p->o_arg.fh = NFS_FH(dir);
1403                 break;
1404         case NFS4_OPEN_CLAIM_PREVIOUS:
1405         case NFS4_OPEN_CLAIM_FH:
1406         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1407         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1408                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1409         }
1410         p->c_arg.fh = &p->o_res.fh;
1411         p->c_arg.stateid = &p->o_res.stateid;
1412         p->c_arg.seqid = p->o_arg.seqid;
1413         nfs4_init_opendata_res(p);
1414         kref_init(&p->kref);
1415         return p;
1416
1417 err_free_label:
1418         nfs4_label_free(p->a_label);
1419 err_free_f:
1420         nfs4_label_free(p->f_label);
1421 err_free_p:
1422         kfree(p);
1423 err:
1424         dput(parent);
1425         return NULL;
1426 }
1427
1428 static void nfs4_opendata_free(struct kref *kref)
1429 {
1430         struct nfs4_opendata *p = container_of(kref,
1431                         struct nfs4_opendata, kref);
1432         struct super_block *sb = p->dentry->d_sb;
1433
1434         nfs4_lgopen_release(p->lgp);
1435         nfs_free_seqid(p->o_arg.seqid);
1436         nfs4_sequence_free_slot(&p->o_res.seq_res);
1437         if (p->state != NULL)
1438                 nfs4_put_open_state(p->state);
1439         nfs4_put_state_owner(p->owner);
1440
1441         nfs4_label_free(p->a_label);
1442         nfs4_label_free(p->f_label);
1443
1444         dput(p->dir);
1445         dput(p->dentry);
1446         nfs_sb_deactive(sb);
1447         nfs_fattr_free_names(&p->f_attr);
1448         kfree(p->f_attr.mdsthreshold);
1449         kfree(p);
1450 }
1451
1452 static void nfs4_opendata_put(struct nfs4_opendata *p)
1453 {
1454         if (p != NULL)
1455                 kref_put(&p->kref, nfs4_opendata_free);
1456 }
1457
1458 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1459                 fmode_t fmode)
1460 {
1461         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1462         case FMODE_READ|FMODE_WRITE:
1463                 return state->n_rdwr != 0;
1464         case FMODE_WRITE:
1465                 return state->n_wronly != 0;
1466         case FMODE_READ:
1467                 return state->n_rdonly != 0;
1468         }
1469         WARN_ON_ONCE(1);
1470         return false;
1471 }
1472
1473 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1474                 int open_mode, enum open_claim_type4 claim)
1475 {
1476         int ret = 0;
1477
1478         if (open_mode & (O_EXCL|O_TRUNC))
1479                 goto out;
1480         switch (claim) {
1481         case NFS4_OPEN_CLAIM_NULL:
1482         case NFS4_OPEN_CLAIM_FH:
1483                 goto out;
1484         default:
1485                 break;
1486         }
1487         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1488                 case FMODE_READ:
1489                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1490                                 && state->n_rdonly != 0;
1491                         break;
1492                 case FMODE_WRITE:
1493                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1494                                 && state->n_wronly != 0;
1495                         break;
1496                 case FMODE_READ|FMODE_WRITE:
1497                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1498                                 && state->n_rdwr != 0;
1499         }
1500 out:
1501         return ret;
1502 }
1503
1504 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1505                 enum open_claim_type4 claim)
1506 {
1507         if (delegation == NULL)
1508                 return 0;
1509         if ((delegation->type & fmode) != fmode)
1510                 return 0;
1511         switch (claim) {
1512         case NFS4_OPEN_CLAIM_NULL:
1513         case NFS4_OPEN_CLAIM_FH:
1514                 break;
1515         case NFS4_OPEN_CLAIM_PREVIOUS:
1516                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1517                         break;
1518                 fallthrough;
1519         default:
1520                 return 0;
1521         }
1522         nfs_mark_delegation_referenced(delegation);
1523         return 1;
1524 }
1525
1526 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1527 {
1528         switch (fmode) {
1529                 case FMODE_WRITE:
1530                         state->n_wronly++;
1531                         break;
1532                 case FMODE_READ:
1533                         state->n_rdonly++;
1534                         break;
1535                 case FMODE_READ|FMODE_WRITE:
1536                         state->n_rdwr++;
1537         }
1538         nfs4_state_set_mode_locked(state, state->state | fmode);
1539 }
1540
1541 #ifdef CONFIG_NFS_V4_1
1542 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1543 {
1544         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1545                 return true;
1546         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1547                 return true;
1548         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1549                 return true;
1550         return false;
1551 }
1552 #endif /* CONFIG_NFS_V4_1 */
1553
1554 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1555 {
1556         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1557                 wake_up_all(&state->waitq);
1558 }
1559
1560 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1561 {
1562         struct nfs_client *clp = state->owner->so_server->nfs_client;
1563         bool need_recover = false;
1564
1565         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1566                 need_recover = true;
1567         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1568                 need_recover = true;
1569         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1570                 need_recover = true;
1571         if (need_recover)
1572                 nfs4_state_mark_reclaim_nograce(clp, state);
1573 }
1574
1575 /*
1576  * Check for whether or not the caller may update the open stateid
1577  * to the value passed in by stateid.
1578  *
1579  * Note: This function relies heavily on the server implementing
1580  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1581  * correctly.
1582  * i.e. The stateid seqids have to be initialised to 1, and
1583  * are then incremented on every state transition.
1584  */
1585 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1586                 const nfs4_stateid *stateid)
1587 {
1588         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1589                 /* The common case - we're updating to a new sequence number */
1590                 if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1591                         nfs4_stateid_is_next(&state->open_stateid, stateid)) {
1592                         return true;
1593                 }
1594         } else {
1595                 /* This is the first OPEN in this generation */
1596                 if (stateid->seqid == cpu_to_be32(1))
1597                         return true;
1598         }
1599         return false;
1600 }
1601
1602 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1603 {
1604         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1605                 return;
1606         if (state->n_wronly)
1607                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1608         if (state->n_rdonly)
1609                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1610         if (state->n_rdwr)
1611                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1612         set_bit(NFS_OPEN_STATE, &state->flags);
1613 }
1614
1615 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1616                 nfs4_stateid *stateid, fmode_t fmode)
1617 {
1618         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1619         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1620         case FMODE_WRITE:
1621                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1622                 break;
1623         case FMODE_READ:
1624                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1625                 break;
1626         case 0:
1627                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1628                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1629                 clear_bit(NFS_OPEN_STATE, &state->flags);
1630         }
1631         if (stateid == NULL)
1632                 return;
1633         /* Handle OPEN+OPEN_DOWNGRADE races */
1634         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1635             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1636                 nfs_resync_open_stateid_locked(state);
1637                 goto out;
1638         }
1639         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1640                 nfs4_stateid_copy(&state->stateid, stateid);
1641         nfs4_stateid_copy(&state->open_stateid, stateid);
1642         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1643 out:
1644         nfs_state_log_update_open_stateid(state);
1645 }
1646
1647 static void nfs_clear_open_stateid(struct nfs4_state *state,
1648         nfs4_stateid *arg_stateid,
1649         nfs4_stateid *stateid, fmode_t fmode)
1650 {
1651         write_seqlock(&state->seqlock);
1652         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1653         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1654                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1655         write_sequnlock(&state->seqlock);
1656         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1657                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1658 }
1659
1660 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1661                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1662         __must_hold(&state->owner->so_lock)
1663         __must_hold(&state->seqlock)
1664         __must_hold(RCU)
1665
1666 {
1667         DEFINE_WAIT(wait);
1668         int status = 0;
1669         for (;;) {
1670
1671                 if (nfs_stateid_is_sequential(state, stateid))
1672                         break;
1673
1674                 if (status)
1675                         break;
1676                 /* Rely on seqids for serialisation with NFSv4.0 */
1677                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1678                         break;
1679
1680                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1681                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1682                 /*
1683                  * Ensure we process the state changes in the same order
1684                  * in which the server processed them by delaying the
1685                  * update of the stateid until we are in sequence.
1686                  */
1687                 write_sequnlock(&state->seqlock);
1688                 spin_unlock(&state->owner->so_lock);
1689                 rcu_read_unlock();
1690                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1691
1692                 if (!signal_pending(current)) {
1693                         if (schedule_timeout(5*HZ) == 0)
1694                                 status = -EAGAIN;
1695                         else
1696                                 status = 0;
1697                 } else
1698                         status = -EINTR;
1699                 finish_wait(&state->waitq, &wait);
1700                 rcu_read_lock();
1701                 spin_lock(&state->owner->so_lock);
1702                 write_seqlock(&state->seqlock);
1703         }
1704
1705         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1706             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1707                 nfs4_stateid_copy(freeme, &state->open_stateid);
1708                 nfs_test_and_clear_all_open_stateid(state);
1709         }
1710
1711         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1712                 nfs4_stateid_copy(&state->stateid, stateid);
1713         nfs4_stateid_copy(&state->open_stateid, stateid);
1714         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1715         nfs_state_log_update_open_stateid(state);
1716 }
1717
1718 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1719                 const nfs4_stateid *open_stateid,
1720                 fmode_t fmode,
1721                 nfs4_stateid *freeme)
1722 {
1723         /*
1724          * Protect the call to nfs4_state_set_mode_locked and
1725          * serialise the stateid update
1726          */
1727         write_seqlock(&state->seqlock);
1728         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1729         switch (fmode) {
1730         case FMODE_READ:
1731                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1732                 break;
1733         case FMODE_WRITE:
1734                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1735                 break;
1736         case FMODE_READ|FMODE_WRITE:
1737                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1738         }
1739         set_bit(NFS_OPEN_STATE, &state->flags);
1740         write_sequnlock(&state->seqlock);
1741 }
1742
1743 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1744 {
1745         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1746         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1747         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1748         clear_bit(NFS_OPEN_STATE, &state->flags);
1749 }
1750
1751 static void nfs_state_set_delegation(struct nfs4_state *state,
1752                 const nfs4_stateid *deleg_stateid,
1753                 fmode_t fmode)
1754 {
1755         /*
1756          * Protect the call to nfs4_state_set_mode_locked and
1757          * serialise the stateid update
1758          */
1759         write_seqlock(&state->seqlock);
1760         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1761         set_bit(NFS_DELEGATED_STATE, &state->flags);
1762         write_sequnlock(&state->seqlock);
1763 }
1764
1765 static void nfs_state_clear_delegation(struct nfs4_state *state)
1766 {
1767         write_seqlock(&state->seqlock);
1768         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1769         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1770         write_sequnlock(&state->seqlock);
1771 }
1772
1773 int update_open_stateid(struct nfs4_state *state,
1774                 const nfs4_stateid *open_stateid,
1775                 const nfs4_stateid *delegation,
1776                 fmode_t fmode)
1777 {
1778         struct nfs_server *server = NFS_SERVER(state->inode);
1779         struct nfs_client *clp = server->nfs_client;
1780         struct nfs_inode *nfsi = NFS_I(state->inode);
1781         struct nfs_delegation *deleg_cur;
1782         nfs4_stateid freeme = { };
1783         int ret = 0;
1784
1785         fmode &= (FMODE_READ|FMODE_WRITE);
1786
1787         rcu_read_lock();
1788         spin_lock(&state->owner->so_lock);
1789         if (open_stateid != NULL) {
1790                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1791                 ret = 1;
1792         }
1793
1794         deleg_cur = nfs4_get_valid_delegation(state->inode);
1795         if (deleg_cur == NULL)
1796                 goto no_delegation;
1797
1798         spin_lock(&deleg_cur->lock);
1799         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1800            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1801             (deleg_cur->type & fmode) != fmode)
1802                 goto no_delegation_unlock;
1803
1804         if (delegation == NULL)
1805                 delegation = &deleg_cur->stateid;
1806         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1807                 goto no_delegation_unlock;
1808
1809         nfs_mark_delegation_referenced(deleg_cur);
1810         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1811         ret = 1;
1812 no_delegation_unlock:
1813         spin_unlock(&deleg_cur->lock);
1814 no_delegation:
1815         if (ret)
1816                 update_open_stateflags(state, fmode);
1817         spin_unlock(&state->owner->so_lock);
1818         rcu_read_unlock();
1819
1820         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1821                 nfs4_schedule_state_manager(clp);
1822         if (freeme.type != 0)
1823                 nfs4_test_and_free_stateid(server, &freeme,
1824                                 state->owner->so_cred);
1825
1826         return ret;
1827 }
1828
1829 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1830                 const nfs4_stateid *stateid)
1831 {
1832         struct nfs4_state *state = lsp->ls_state;
1833         bool ret = false;
1834
1835         spin_lock(&state->state_lock);
1836         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1837                 goto out_noupdate;
1838         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1839                 goto out_noupdate;
1840         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1841         ret = true;
1842 out_noupdate:
1843         spin_unlock(&state->state_lock);
1844         return ret;
1845 }
1846
1847 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1848 {
1849         struct nfs_delegation *delegation;
1850
1851         fmode &= FMODE_READ|FMODE_WRITE;
1852         rcu_read_lock();
1853         delegation = nfs4_get_valid_delegation(inode);
1854         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1855                 rcu_read_unlock();
1856                 return;
1857         }
1858         rcu_read_unlock();
1859         nfs4_inode_return_delegation(inode);
1860 }
1861
1862 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1863 {
1864         struct nfs4_state *state = opendata->state;
1865         struct nfs_delegation *delegation;
1866         int open_mode = opendata->o_arg.open_flags;
1867         fmode_t fmode = opendata->o_arg.fmode;
1868         enum open_claim_type4 claim = opendata->o_arg.claim;
1869         nfs4_stateid stateid;
1870         int ret = -EAGAIN;
1871
1872         for (;;) {
1873                 spin_lock(&state->owner->so_lock);
1874                 if (can_open_cached(state, fmode, open_mode, claim)) {
1875                         update_open_stateflags(state, fmode);
1876                         spin_unlock(&state->owner->so_lock);
1877                         goto out_return_state;
1878                 }
1879                 spin_unlock(&state->owner->so_lock);
1880                 rcu_read_lock();
1881                 delegation = nfs4_get_valid_delegation(state->inode);
1882                 if (!can_open_delegated(delegation, fmode, claim)) {
1883                         rcu_read_unlock();
1884                         break;
1885                 }
1886                 /* Save the delegation */
1887                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1888                 rcu_read_unlock();
1889                 nfs_release_seqid(opendata->o_arg.seqid);
1890                 if (!opendata->is_recover) {
1891                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1892                         if (ret != 0)
1893                                 goto out;
1894                 }
1895                 ret = -EAGAIN;
1896
1897                 /* Try to update the stateid using the delegation */
1898                 if (update_open_stateid(state, NULL, &stateid, fmode))
1899                         goto out_return_state;
1900         }
1901 out:
1902         return ERR_PTR(ret);
1903 out_return_state:
1904         refcount_inc(&state->count);
1905         return state;
1906 }
1907
1908 static void
1909 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1910 {
1911         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1912         struct nfs_delegation *delegation;
1913         int delegation_flags = 0;
1914
1915         rcu_read_lock();
1916         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1917         if (delegation)
1918                 delegation_flags = delegation->flags;
1919         rcu_read_unlock();
1920         switch (data->o_arg.claim) {
1921         default:
1922                 break;
1923         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1924         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1925                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1926                                    "returning a delegation for "
1927                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1928                                    clp->cl_hostname);
1929                 return;
1930         }
1931         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1932                 nfs_inode_set_delegation(state->inode,
1933                                 data->owner->so_cred,
1934                                 data->o_res.delegation_type,
1935                                 &data->o_res.delegation,
1936                                 data->o_res.pagemod_limit);
1937         else
1938                 nfs_inode_reclaim_delegation(state->inode,
1939                                 data->owner->so_cred,
1940                                 data->o_res.delegation_type,
1941                                 &data->o_res.delegation,
1942                                 data->o_res.pagemod_limit);
1943
1944         if (data->o_res.do_recall)
1945                 nfs_async_inode_return_delegation(state->inode,
1946                                                   &data->o_res.delegation);
1947 }
1948
1949 /*
1950  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1951  * and update the nfs4_state.
1952  */
1953 static struct nfs4_state *
1954 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1955 {
1956         struct inode *inode = data->state->inode;
1957         struct nfs4_state *state = data->state;
1958         int ret;
1959
1960         if (!data->rpc_done) {
1961                 if (data->rpc_status)
1962                         return ERR_PTR(data->rpc_status);
1963                 /* cached opens have already been processed */
1964                 goto update;
1965         }
1966
1967         ret = nfs_refresh_inode(inode, &data->f_attr);
1968         if (ret)
1969                 return ERR_PTR(ret);
1970
1971         if (data->o_res.delegation_type != 0)
1972                 nfs4_opendata_check_deleg(data, state);
1973 update:
1974         if (!update_open_stateid(state, &data->o_res.stateid,
1975                                 NULL, data->o_arg.fmode))
1976                 return ERR_PTR(-EAGAIN);
1977         refcount_inc(&state->count);
1978
1979         return state;
1980 }
1981
1982 static struct inode *
1983 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1984 {
1985         struct inode *inode;
1986
1987         switch (data->o_arg.claim) {
1988         case NFS4_OPEN_CLAIM_NULL:
1989         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1990         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1991                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1992                         return ERR_PTR(-EAGAIN);
1993                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1994                                 &data->f_attr, data->f_label);
1995                 break;
1996         default:
1997                 inode = d_inode(data->dentry);
1998                 ihold(inode);
1999                 nfs_refresh_inode(inode, &data->f_attr);
2000         }
2001         return inode;
2002 }
2003
2004 static struct nfs4_state *
2005 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2006 {
2007         struct nfs4_state *state;
2008         struct inode *inode;
2009
2010         inode = nfs4_opendata_get_inode(data);
2011         if (IS_ERR(inode))
2012                 return ERR_CAST(inode);
2013         if (data->state != NULL && data->state->inode == inode) {
2014                 state = data->state;
2015                 refcount_inc(&state->count);
2016         } else
2017                 state = nfs4_get_open_state(inode, data->owner);
2018         iput(inode);
2019         if (state == NULL)
2020                 state = ERR_PTR(-ENOMEM);
2021         return state;
2022 }
2023
2024 static struct nfs4_state *
2025 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2026 {
2027         struct nfs4_state *state;
2028
2029         if (!data->rpc_done) {
2030                 state = nfs4_try_open_cached(data);
2031                 trace_nfs4_cached_open(data->state);
2032                 goto out;
2033         }
2034
2035         state = nfs4_opendata_find_nfs4_state(data);
2036         if (IS_ERR(state))
2037                 goto out;
2038
2039         if (data->o_res.delegation_type != 0)
2040                 nfs4_opendata_check_deleg(data, state);
2041         if (!update_open_stateid(state, &data->o_res.stateid,
2042                                 NULL, data->o_arg.fmode)) {
2043                 nfs4_put_open_state(state);
2044                 state = ERR_PTR(-EAGAIN);
2045         }
2046 out:
2047         nfs_release_seqid(data->o_arg.seqid);
2048         return state;
2049 }
2050
2051 static struct nfs4_state *
2052 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2053 {
2054         struct nfs4_state *ret;
2055
2056         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2057                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2058         else
2059                 ret = _nfs4_opendata_to_nfs4_state(data);
2060         nfs4_sequence_free_slot(&data->o_res.seq_res);
2061         return ret;
2062 }
2063
2064 static struct nfs_open_context *
2065 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2066 {
2067         struct nfs_inode *nfsi = NFS_I(state->inode);
2068         struct nfs_open_context *ctx;
2069
2070         rcu_read_lock();
2071         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2072                 if (ctx->state != state)
2073                         continue;
2074                 if ((ctx->mode & mode) != mode)
2075                         continue;
2076                 if (!get_nfs_open_context(ctx))
2077                         continue;
2078                 rcu_read_unlock();
2079                 return ctx;
2080         }
2081         rcu_read_unlock();
2082         return ERR_PTR(-ENOENT);
2083 }
2084
2085 static struct nfs_open_context *
2086 nfs4_state_find_open_context(struct nfs4_state *state)
2087 {
2088         struct nfs_open_context *ctx;
2089
2090         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2091         if (!IS_ERR(ctx))
2092                 return ctx;
2093         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2094         if (!IS_ERR(ctx))
2095                 return ctx;
2096         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2097 }
2098
2099 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2100                 struct nfs4_state *state, enum open_claim_type4 claim)
2101 {
2102         struct nfs4_opendata *opendata;
2103
2104         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2105                         NULL, claim, GFP_NOFS);
2106         if (opendata == NULL)
2107                 return ERR_PTR(-ENOMEM);
2108         opendata->state = state;
2109         refcount_inc(&state->count);
2110         return opendata;
2111 }
2112
2113 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2114                 fmode_t fmode)
2115 {
2116         struct nfs4_state *newstate;
2117         int ret;
2118
2119         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2120                 return 0;
2121         opendata->o_arg.open_flags = 0;
2122         opendata->o_arg.fmode = fmode;
2123         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2124                         NFS_SB(opendata->dentry->d_sb),
2125                         fmode, 0);
2126         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2127         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2128         nfs4_init_opendata_res(opendata);
2129         ret = _nfs4_recover_proc_open(opendata);
2130         if (ret != 0)
2131                 return ret;
2132         newstate = nfs4_opendata_to_nfs4_state(opendata);
2133         if (IS_ERR(newstate))
2134                 return PTR_ERR(newstate);
2135         if (newstate != opendata->state)
2136                 ret = -ESTALE;
2137         nfs4_close_state(newstate, fmode);
2138         return ret;
2139 }
2140
2141 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2142 {
2143         int ret;
2144
2145         /* memory barrier prior to reading state->n_* */
2146         smp_rmb();
2147         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2148         if (ret != 0)
2149                 return ret;
2150         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2151         if (ret != 0)
2152                 return ret;
2153         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2154         if (ret != 0)
2155                 return ret;
2156         /*
2157          * We may have performed cached opens for all three recoveries.
2158          * Check if we need to update the current stateid.
2159          */
2160         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2161             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2162                 write_seqlock(&state->seqlock);
2163                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2164                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2165                 write_sequnlock(&state->seqlock);
2166         }
2167         return 0;
2168 }
2169
2170 /*
2171  * OPEN_RECLAIM:
2172  *      reclaim state on the server after a reboot.
2173  */
2174 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2175 {
2176         struct nfs_delegation *delegation;
2177         struct nfs4_opendata *opendata;
2178         fmode_t delegation_type = 0;
2179         int status;
2180
2181         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2182                         NFS4_OPEN_CLAIM_PREVIOUS);
2183         if (IS_ERR(opendata))
2184                 return PTR_ERR(opendata);
2185         rcu_read_lock();
2186         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2187         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2188                 delegation_type = delegation->type;
2189         rcu_read_unlock();
2190         opendata->o_arg.u.delegation_type = delegation_type;
2191         status = nfs4_open_recover(opendata, state);
2192         nfs4_opendata_put(opendata);
2193         return status;
2194 }
2195
2196 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2197 {
2198         struct nfs_server *server = NFS_SERVER(state->inode);
2199         struct nfs4_exception exception = { };
2200         int err;
2201         do {
2202                 err = _nfs4_do_open_reclaim(ctx, state);
2203                 trace_nfs4_open_reclaim(ctx, 0, err);
2204                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2205                         continue;
2206                 if (err != -NFS4ERR_DELAY)
2207                         break;
2208                 nfs4_handle_exception(server, err, &exception);
2209         } while (exception.retry);
2210         return err;
2211 }
2212
2213 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2214 {
2215         struct nfs_open_context *ctx;
2216         int ret;
2217
2218         ctx = nfs4_state_find_open_context(state);
2219         if (IS_ERR(ctx))
2220                 return -EAGAIN;
2221         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2222         nfs_state_clear_open_state_flags(state);
2223         ret = nfs4_do_open_reclaim(ctx, state);
2224         put_nfs_open_context(ctx);
2225         return ret;
2226 }
2227
2228 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)
2229 {
2230         switch (err) {
2231                 default:
2232                         printk(KERN_ERR "NFS: %s: unhandled error "
2233                                         "%d.\n", __func__, err);
2234                 case 0:
2235                 case -ENOENT:
2236                 case -EAGAIN:
2237                 case -ESTALE:
2238                 case -ETIMEDOUT:
2239                         break;
2240                 case -NFS4ERR_BADSESSION:
2241                 case -NFS4ERR_BADSLOT:
2242                 case -NFS4ERR_BAD_HIGH_SLOT:
2243                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2244                 case -NFS4ERR_DEADSESSION:
2245                         return -EAGAIN;
2246                 case -NFS4ERR_STALE_CLIENTID:
2247                 case -NFS4ERR_STALE_STATEID:
2248                         /* Don't recall a delegation if it was lost */
2249                         nfs4_schedule_lease_recovery(server->nfs_client);
2250                         return -EAGAIN;
2251                 case -NFS4ERR_MOVED:
2252                         nfs4_schedule_migration_recovery(server);
2253                         return -EAGAIN;
2254                 case -NFS4ERR_LEASE_MOVED:
2255                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2256                         return -EAGAIN;
2257                 case -NFS4ERR_DELEG_REVOKED:
2258                 case -NFS4ERR_ADMIN_REVOKED:
2259                 case -NFS4ERR_EXPIRED:
2260                 case -NFS4ERR_BAD_STATEID:
2261                 case -NFS4ERR_OPENMODE:
2262                         nfs_inode_find_state_and_recover(state->inode,
2263                                         stateid);
2264                         nfs4_schedule_stateid_recovery(server, state);
2265                         return -EAGAIN;
2266                 case -NFS4ERR_DELAY:
2267                 case -NFS4ERR_GRACE:
2268                         ssleep(1);
2269                         return -EAGAIN;
2270                 case -ENOMEM:
2271                 case -NFS4ERR_DENIED:
2272                         if (fl) {
2273                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2274                                 if (lsp)
2275                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2276                         }
2277                         return 0;
2278         }
2279         return err;
2280 }
2281
2282 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2283                 struct nfs4_state *state, const nfs4_stateid *stateid)
2284 {
2285         struct nfs_server *server = NFS_SERVER(state->inode);
2286         struct nfs4_opendata *opendata;
2287         int err = 0;
2288
2289         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2290                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2291         if (IS_ERR(opendata))
2292                 return PTR_ERR(opendata);
2293         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2294         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2295                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2296                 if (err)
2297                         goto out;
2298         }
2299         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2300                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2301                 if (err)
2302                         goto out;
2303         }
2304         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2305                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2306                 if (err)
2307                         goto out;
2308         }
2309         nfs_state_clear_delegation(state);
2310 out:
2311         nfs4_opendata_put(opendata);
2312         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2313 }
2314
2315 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2316 {
2317         struct nfs4_opendata *data = calldata;
2318
2319         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2320                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2321 }
2322
2323 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2324 {
2325         struct nfs4_opendata *data = calldata;
2326
2327         nfs40_sequence_done(task, &data->c_res.seq_res);
2328
2329         data->rpc_status = task->tk_status;
2330         if (data->rpc_status == 0) {
2331                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2332                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2333                 renew_lease(data->o_res.server, data->timestamp);
2334                 data->rpc_done = true;
2335         }
2336 }
2337
2338 static void nfs4_open_confirm_release(void *calldata)
2339 {
2340         struct nfs4_opendata *data = calldata;
2341         struct nfs4_state *state = NULL;
2342
2343         /* If this request hasn't been cancelled, do nothing */
2344         if (!data->cancelled)
2345                 goto out_free;
2346         /* In case of error, no cleanup! */
2347         if (!data->rpc_done)
2348                 goto out_free;
2349         state = nfs4_opendata_to_nfs4_state(data);
2350         if (!IS_ERR(state))
2351                 nfs4_close_state(state, data->o_arg.fmode);
2352 out_free:
2353         nfs4_opendata_put(data);
2354 }
2355
2356 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2357         .rpc_call_prepare = nfs4_open_confirm_prepare,
2358         .rpc_call_done = nfs4_open_confirm_done,
2359         .rpc_release = nfs4_open_confirm_release,
2360 };
2361
2362 /*
2363  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2364  */
2365 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2366 {
2367         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2368         struct rpc_task *task;
2369         struct  rpc_message msg = {
2370                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2371                 .rpc_argp = &data->c_arg,
2372                 .rpc_resp = &data->c_res,
2373                 .rpc_cred = data->owner->so_cred,
2374         };
2375         struct rpc_task_setup task_setup_data = {
2376                 .rpc_client = server->client,
2377                 .rpc_message = &msg,
2378                 .callback_ops = &nfs4_open_confirm_ops,
2379                 .callback_data = data,
2380                 .workqueue = nfsiod_workqueue,
2381                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2382         };
2383         int status;
2384
2385         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2386                                 data->is_recover);
2387         kref_get(&data->kref);
2388         data->rpc_done = false;
2389         data->rpc_status = 0;
2390         data->timestamp = jiffies;
2391         task = rpc_run_task(&task_setup_data);
2392         if (IS_ERR(task))
2393                 return PTR_ERR(task);
2394         status = rpc_wait_for_completion_task(task);
2395         if (status != 0) {
2396                 data->cancelled = true;
2397                 smp_wmb();
2398         } else
2399                 status = data->rpc_status;
2400         rpc_put_task(task);
2401         return status;
2402 }
2403
2404 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2405 {
2406         struct nfs4_opendata *data = calldata;
2407         struct nfs4_state_owner *sp = data->owner;
2408         struct nfs_client *clp = sp->so_server->nfs_client;
2409         enum open_claim_type4 claim = data->o_arg.claim;
2410
2411         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2412                 goto out_wait;
2413         /*
2414          * Check if we still need to send an OPEN call, or if we can use
2415          * a delegation instead.
2416          */
2417         if (data->state != NULL) {
2418                 struct nfs_delegation *delegation;
2419
2420                 if (can_open_cached(data->state, data->o_arg.fmode,
2421                                         data->o_arg.open_flags, claim))
2422                         goto out_no_action;
2423                 rcu_read_lock();
2424                 delegation = nfs4_get_valid_delegation(data->state->inode);
2425                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2426                         goto unlock_no_action;
2427                 rcu_read_unlock();
2428         }
2429         /* Update client id. */
2430         data->o_arg.clientid = clp->cl_clientid;
2431         switch (claim) {
2432         default:
2433                 break;
2434         case NFS4_OPEN_CLAIM_PREVIOUS:
2435         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2436         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2437                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2438                 fallthrough;
2439         case NFS4_OPEN_CLAIM_FH:
2440                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2441         }
2442         data->timestamp = jiffies;
2443         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2444                                 &data->o_arg.seq_args,
2445                                 &data->o_res.seq_res,
2446                                 task) != 0)
2447                 nfs_release_seqid(data->o_arg.seqid);
2448
2449         /* Set the create mode (note dependency on the session type) */
2450         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2451         if (data->o_arg.open_flags & O_EXCL) {
2452                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2453                 if (nfs4_has_persistent_session(clp))
2454                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2455                 else if (clp->cl_mvops->minor_version > 0)
2456                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2457         }
2458         return;
2459 unlock_no_action:
2460         trace_nfs4_cached_open(data->state);
2461         rcu_read_unlock();
2462 out_no_action:
2463         task->tk_action = NULL;
2464 out_wait:
2465         nfs4_sequence_done(task, &data->o_res.seq_res);
2466 }
2467
2468 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2469 {
2470         struct nfs4_opendata *data = calldata;
2471
2472         data->rpc_status = task->tk_status;
2473
2474         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2475                 return;
2476
2477         if (task->tk_status == 0) {
2478                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2479                         switch (data->o_res.f_attr->mode & S_IFMT) {
2480                         case S_IFREG:
2481                                 break;
2482                         case S_IFLNK:
2483                                 data->rpc_status = -ELOOP;
2484                                 break;
2485                         case S_IFDIR:
2486                                 data->rpc_status = -EISDIR;
2487                                 break;
2488                         default:
2489                                 data->rpc_status = -ENOTDIR;
2490                         }
2491                 }
2492                 renew_lease(data->o_res.server, data->timestamp);
2493                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2494                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2495         }
2496         data->rpc_done = true;
2497 }
2498
2499 static void nfs4_open_release(void *calldata)
2500 {
2501         struct nfs4_opendata *data = calldata;
2502         struct nfs4_state *state = NULL;
2503
2504         /* If this request hasn't been cancelled, do nothing */
2505         if (!data->cancelled)
2506                 goto out_free;
2507         /* In case of error, no cleanup! */
2508         if (data->rpc_status != 0 || !data->rpc_done)
2509                 goto out_free;
2510         /* In case we need an open_confirm, no cleanup! */
2511         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2512                 goto out_free;
2513         state = nfs4_opendata_to_nfs4_state(data);
2514         if (!IS_ERR(state))
2515                 nfs4_close_state(state, data->o_arg.fmode);
2516 out_free:
2517         nfs4_opendata_put(data);
2518 }
2519
2520 static const struct rpc_call_ops nfs4_open_ops = {
2521         .rpc_call_prepare = nfs4_open_prepare,
2522         .rpc_call_done = nfs4_open_done,
2523         .rpc_release = nfs4_open_release,
2524 };
2525
2526 static int nfs4_run_open_task(struct nfs4_opendata *data,
2527                               struct nfs_open_context *ctx)
2528 {
2529         struct inode *dir = d_inode(data->dir);
2530         struct nfs_server *server = NFS_SERVER(dir);
2531         struct nfs_openargs *o_arg = &data->o_arg;
2532         struct nfs_openres *o_res = &data->o_res;
2533         struct rpc_task *task;
2534         struct rpc_message msg = {
2535                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2536                 .rpc_argp = o_arg,
2537                 .rpc_resp = o_res,
2538                 .rpc_cred = data->owner->so_cred,
2539         };
2540         struct rpc_task_setup task_setup_data = {
2541                 .rpc_client = server->client,
2542                 .rpc_message = &msg,
2543                 .callback_ops = &nfs4_open_ops,
2544                 .callback_data = data,
2545                 .workqueue = nfsiod_workqueue,
2546                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2547         };
2548         int status;
2549
2550         kref_get(&data->kref);
2551         data->rpc_done = false;
2552         data->rpc_status = 0;
2553         data->cancelled = false;
2554         data->is_recover = false;
2555         if (!ctx) {
2556                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2557                 data->is_recover = true;
2558                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2559         } else {
2560                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2561                 pnfs_lgopen_prepare(data, ctx);
2562         }
2563         task = rpc_run_task(&task_setup_data);
2564         if (IS_ERR(task))
2565                 return PTR_ERR(task);
2566         status = rpc_wait_for_completion_task(task);
2567         if (status != 0) {
2568                 data->cancelled = true;
2569                 smp_wmb();
2570         } else
2571                 status = data->rpc_status;
2572         rpc_put_task(task);
2573
2574         return status;
2575 }
2576
2577 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2578 {
2579         struct inode *dir = d_inode(data->dir);
2580         struct nfs_openres *o_res = &data->o_res;
2581         int status;
2582
2583         status = nfs4_run_open_task(data, NULL);
2584         if (status != 0 || !data->rpc_done)
2585                 return status;
2586
2587         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2588
2589         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2590                 status = _nfs4_proc_open_confirm(data);
2591
2592         return status;
2593 }
2594
2595 /*
2596  * Additional permission checks in order to distinguish between an
2597  * open for read, and an open for execute. This works around the
2598  * fact that NFSv4 OPEN treats read and execute permissions as being
2599  * the same.
2600  * Note that in the non-execute case, we want to turn off permission
2601  * checking if we just created a new file (POSIX open() semantics).
2602  */
2603 static int nfs4_opendata_access(const struct cred *cred,
2604                                 struct nfs4_opendata *opendata,
2605                                 struct nfs4_state *state, fmode_t fmode,
2606                                 int openflags)
2607 {
2608         struct nfs_access_entry cache;
2609         u32 mask, flags;
2610
2611         /* access call failed or for some reason the server doesn't
2612          * support any access modes -- defer access call until later */
2613         if (opendata->o_res.access_supported == 0)
2614                 return 0;
2615
2616         mask = 0;
2617         /*
2618          * Use openflags to check for exec, because fmode won't
2619          * always have FMODE_EXEC set when file open for exec.
2620          */
2621         if (openflags & __FMODE_EXEC) {
2622                 /* ONLY check for exec rights */
2623                 if (S_ISDIR(state->inode->i_mode))
2624                         mask = NFS4_ACCESS_LOOKUP;
2625                 else
2626                         mask = NFS4_ACCESS_EXECUTE;
2627         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2628                 mask = NFS4_ACCESS_READ;
2629
2630         cache.cred = cred;
2631         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2632         nfs_access_add_cache(state->inode, &cache);
2633
2634         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2635         if ((mask & ~cache.mask & flags) == 0)
2636                 return 0;
2637
2638         return -EACCES;
2639 }
2640
2641 /*
2642  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2643  */
2644 static int _nfs4_proc_open(struct nfs4_opendata *data,
2645                            struct nfs_open_context *ctx)
2646 {
2647         struct inode *dir = d_inode(data->dir);
2648         struct nfs_server *server = NFS_SERVER(dir);
2649         struct nfs_openargs *o_arg = &data->o_arg;
2650         struct nfs_openres *o_res = &data->o_res;
2651         int status;
2652
2653         status = nfs4_run_open_task(data, ctx);
2654         if (!data->rpc_done)
2655                 return status;
2656         if (status != 0) {
2657                 if (status == -NFS4ERR_BADNAME &&
2658                                 !(o_arg->open_flags & O_CREAT))
2659                         return -ENOENT;
2660                 return status;
2661         }
2662
2663         nfs_fattr_map_and_free_names(server, &data->f_attr);
2664
2665         if (o_arg->open_flags & O_CREAT) {
2666                 if (o_arg->open_flags & O_EXCL)
2667                         data->file_created = true;
2668                 else if (o_res->cinfo.before != o_res->cinfo.after)
2669                         data->file_created = true;
2670                 if (data->file_created ||
2671                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2672                         nfs4_update_changeattr(dir, &o_res->cinfo,
2673                                         o_res->f_attr->time_start,
2674                                         NFS_INO_INVALID_DATA);
2675         }
2676         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2677                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2678         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2679                 status = _nfs4_proc_open_confirm(data);
2680                 if (status != 0)
2681                         return status;
2682         }
2683         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2684                 nfs4_sequence_free_slot(&o_res->seq_res);
2685                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2686                                 o_res->f_label, NULL);
2687         }
2688         return 0;
2689 }
2690
2691 /*
2692  * OPEN_EXPIRED:
2693  *      reclaim state on the server after a network partition.
2694  *      Assumes caller holds the appropriate lock
2695  */
2696 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2697 {
2698         struct nfs4_opendata *opendata;
2699         int ret;
2700
2701         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2702                         NFS4_OPEN_CLAIM_FH);
2703         if (IS_ERR(opendata))
2704                 return PTR_ERR(opendata);
2705         ret = nfs4_open_recover(opendata, state);
2706         if (ret == -ESTALE)
2707                 d_drop(ctx->dentry);
2708         nfs4_opendata_put(opendata);
2709         return ret;
2710 }
2711
2712 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2713 {
2714         struct nfs_server *server = NFS_SERVER(state->inode);
2715         struct nfs4_exception exception = { };
2716         int err;
2717
2718         do {
2719                 err = _nfs4_open_expired(ctx, state);
2720                 trace_nfs4_open_expired(ctx, 0, err);
2721                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2722                         continue;
2723                 switch (err) {
2724                 default:
2725                         goto out;
2726                 case -NFS4ERR_GRACE:
2727                 case -NFS4ERR_DELAY:
2728                         nfs4_handle_exception(server, err, &exception);
2729                         err = 0;
2730                 }
2731         } while (exception.retry);
2732 out:
2733         return err;
2734 }
2735
2736 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2737 {
2738         struct nfs_open_context *ctx;
2739         int ret;
2740
2741         ctx = nfs4_state_find_open_context(state);
2742         if (IS_ERR(ctx))
2743                 return -EAGAIN;
2744         ret = nfs4_do_open_expired(ctx, state);
2745         put_nfs_open_context(ctx);
2746         return ret;
2747 }
2748
2749 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2750                 const nfs4_stateid *stateid)
2751 {
2752         nfs_remove_bad_delegation(state->inode, stateid);
2753         nfs_state_clear_delegation(state);
2754 }
2755
2756 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2757 {
2758         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2759                 nfs_finish_clear_delegation_stateid(state, NULL);
2760 }
2761
2762 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2763 {
2764         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2765         nfs40_clear_delegation_stateid(state);
2766         nfs_state_clear_open_state_flags(state);
2767         return nfs4_open_expired(sp, state);
2768 }
2769
2770 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2771                 nfs4_stateid *stateid,
2772                 const struct cred *cred)
2773 {
2774         return -NFS4ERR_BAD_STATEID;
2775 }
2776
2777 #if defined(CONFIG_NFS_V4_1)
2778 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2779                 nfs4_stateid *stateid,
2780                 const struct cred *cred)
2781 {
2782         int status;
2783
2784         switch (stateid->type) {
2785         default:
2786                 break;
2787         case NFS4_INVALID_STATEID_TYPE:
2788         case NFS4_SPECIAL_STATEID_TYPE:
2789                 return -NFS4ERR_BAD_STATEID;
2790         case NFS4_REVOKED_STATEID_TYPE:
2791                 goto out_free;
2792         }
2793
2794         status = nfs41_test_stateid(server, stateid, cred);
2795         switch (status) {
2796         case -NFS4ERR_EXPIRED:
2797         case -NFS4ERR_ADMIN_REVOKED:
2798         case -NFS4ERR_DELEG_REVOKED:
2799                 break;
2800         default:
2801                 return status;
2802         }
2803 out_free:
2804         /* Ack the revoked state to the server */
2805         nfs41_free_stateid(server, stateid, cred, true);
2806         return -NFS4ERR_EXPIRED;
2807 }
2808
2809 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2810 {
2811         struct nfs_server *server = NFS_SERVER(state->inode);
2812         nfs4_stateid stateid;
2813         struct nfs_delegation *delegation;
2814         const struct cred *cred = NULL;
2815         int status, ret = NFS_OK;
2816
2817         /* Get the delegation credential for use by test/free_stateid */
2818         rcu_read_lock();
2819         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2820         if (delegation == NULL) {
2821                 rcu_read_unlock();
2822                 nfs_state_clear_delegation(state);
2823                 return NFS_OK;
2824         }
2825
2826         spin_lock(&delegation->lock);
2827         nfs4_stateid_copy(&stateid, &delegation->stateid);
2828
2829         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2830                                 &delegation->flags)) {
2831                 spin_unlock(&delegation->lock);
2832                 rcu_read_unlock();
2833                 return NFS_OK;
2834         }
2835
2836         if (delegation->cred)
2837                 cred = get_cred(delegation->cred);
2838         spin_unlock(&delegation->lock);
2839         rcu_read_unlock();
2840         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2841         trace_nfs4_test_delegation_stateid(state, NULL, status);
2842         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2843                 nfs_finish_clear_delegation_stateid(state, &stateid);
2844         else
2845                 ret = status;
2846
2847         put_cred(cred);
2848         return ret;
2849 }
2850
2851 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2852 {
2853         nfs4_stateid tmp;
2854
2855         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2856             nfs4_copy_delegation_stateid(state->inode, state->state,
2857                                 &tmp, NULL) &&
2858             nfs4_stateid_match_other(&state->stateid, &tmp))
2859                 nfs_state_set_delegation(state, &tmp, state->state);
2860         else
2861                 nfs_state_clear_delegation(state);
2862 }
2863
2864 /**
2865  * nfs41_check_expired_locks - possibly free a lock stateid
2866  *
2867  * @state: NFSv4 state for an inode
2868  *
2869  * Returns NFS_OK if recovery for this stateid is now finished.
2870  * Otherwise a negative NFS4ERR value is returned.
2871  */
2872 static int nfs41_check_expired_locks(struct nfs4_state *state)
2873 {
2874         int status, ret = NFS_OK;
2875         struct nfs4_lock_state *lsp, *prev = NULL;
2876         struct nfs_server *server = NFS_SERVER(state->inode);
2877
2878         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2879                 goto out;
2880
2881         spin_lock(&state->state_lock);
2882         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2883                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2884                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2885
2886                         refcount_inc(&lsp->ls_count);
2887                         spin_unlock(&state->state_lock);
2888
2889                         nfs4_put_lock_state(prev);
2890                         prev = lsp;
2891
2892                         status = nfs41_test_and_free_expired_stateid(server,
2893                                         &lsp->ls_stateid,
2894                                         cred);
2895                         trace_nfs4_test_lock_stateid(state, lsp, status);
2896                         if (status == -NFS4ERR_EXPIRED ||
2897                             status == -NFS4ERR_BAD_STATEID) {
2898                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2899                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2900                                 if (!recover_lost_locks)
2901                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2902                         } else if (status != NFS_OK) {
2903                                 ret = status;
2904                                 nfs4_put_lock_state(prev);
2905                                 goto out;
2906                         }
2907                         spin_lock(&state->state_lock);
2908                 }
2909         }
2910         spin_unlock(&state->state_lock);
2911         nfs4_put_lock_state(prev);
2912 out:
2913         return ret;
2914 }
2915
2916 /**
2917  * nfs41_check_open_stateid - possibly free an open stateid
2918  *
2919  * @state: NFSv4 state for an inode
2920  *
2921  * Returns NFS_OK if recovery for this stateid is now finished.
2922  * Otherwise a negative NFS4ERR value is returned.
2923  */
2924 static int nfs41_check_open_stateid(struct nfs4_state *state)
2925 {
2926         struct nfs_server *server = NFS_SERVER(state->inode);
2927         nfs4_stateid *stateid = &state->open_stateid;
2928         const struct cred *cred = state->owner->so_cred;
2929         int status;
2930
2931         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2932                 return -NFS4ERR_BAD_STATEID;
2933         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2934         trace_nfs4_test_open_stateid(state, NULL, status);
2935         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2936                 nfs_state_clear_open_state_flags(state);
2937                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2938                 return status;
2939         }
2940         if (nfs_open_stateid_recover_openmode(state))
2941                 return -NFS4ERR_OPENMODE;
2942         return NFS_OK;
2943 }
2944
2945 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2946 {
2947         int status;
2948
2949         status = nfs41_check_delegation_stateid(state);
2950         if (status != NFS_OK)
2951                 return status;
2952         nfs41_delegation_recover_stateid(state);
2953
2954         status = nfs41_check_expired_locks(state);
2955         if (status != NFS_OK)
2956                 return status;
2957         status = nfs41_check_open_stateid(state);
2958         if (status != NFS_OK)
2959                 status = nfs4_open_expired(sp, state);
2960         return status;
2961 }
2962 #endif
2963
2964 /*
2965  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2966  * fields corresponding to attributes that were used to store the verifier.
2967  * Make sure we clobber those fields in the later setattr call
2968  */
2969 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2970                                 struct iattr *sattr, struct nfs4_label **label)
2971 {
2972         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2973         __u32 attrset[3];
2974         unsigned ret;
2975         unsigned i;
2976
2977         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2978                 attrset[i] = opendata->o_res.attrset[i];
2979                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2980                         attrset[i] &= ~bitmask[i];
2981         }
2982
2983         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2984                 sattr->ia_valid : 0;
2985
2986         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2987                 if (sattr->ia_valid & ATTR_ATIME_SET)
2988                         ret |= ATTR_ATIME_SET;
2989                 else
2990                         ret |= ATTR_ATIME;
2991         }
2992
2993         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2994                 if (sattr->ia_valid & ATTR_MTIME_SET)
2995                         ret |= ATTR_MTIME_SET;
2996                 else
2997                         ret |= ATTR_MTIME;
2998         }
2999
3000         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3001                 *label = NULL;
3002         return ret;
3003 }
3004
3005 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3006                 int flags, struct nfs_open_context *ctx)
3007 {
3008         struct nfs4_state_owner *sp = opendata->owner;
3009         struct nfs_server *server = sp->so_server;
3010         struct dentry *dentry;
3011         struct nfs4_state *state;
3012         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3013         struct inode *dir = d_inode(opendata->dir);
3014         unsigned long dir_verifier;
3015         unsigned int seq;
3016         int ret;
3017
3018         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3019         dir_verifier = nfs_save_change_attribute(dir);
3020
3021         ret = _nfs4_proc_open(opendata, ctx);
3022         if (ret != 0)
3023                 goto out;
3024
3025         state = _nfs4_opendata_to_nfs4_state(opendata);
3026         ret = PTR_ERR(state);
3027         if (IS_ERR(state))
3028                 goto out;
3029         ctx->state = state;
3030         if (server->caps & NFS_CAP_POSIX_LOCK)
3031                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3032         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3033                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3034
3035         dentry = opendata->dentry;
3036         if (d_really_is_negative(dentry)) {
3037                 struct dentry *alias;
3038                 d_drop(dentry);
3039                 alias = d_exact_alias(dentry, state->inode);
3040                 if (!alias)
3041                         alias = d_splice_alias(igrab(state->inode), dentry);
3042                 /* d_splice_alias() can't fail here - it's a non-directory */
3043                 if (alias) {
3044                         dput(ctx->dentry);
3045                         ctx->dentry = dentry = alias;
3046                 }
3047         }
3048
3049         switch(opendata->o_arg.claim) {
3050         default:
3051                 break;
3052         case NFS4_OPEN_CLAIM_NULL:
3053         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3054         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3055                 if (!opendata->rpc_done)
3056                         break;
3057                 if (opendata->o_res.delegation_type != 0)
3058                         dir_verifier = nfs_save_change_attribute(dir);
3059                 nfs_set_verifier(dentry, dir_verifier);
3060         }
3061
3062         /* Parse layoutget results before we check for access */
3063         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3064
3065         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3066                         acc_mode, flags);
3067         if (ret != 0)
3068                 goto out;
3069
3070         if (d_inode(dentry) == state->inode) {
3071                 nfs_inode_attach_open_context(ctx);
3072                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3073                         nfs4_schedule_stateid_recovery(server, state);
3074         }
3075
3076 out:
3077         if (!opendata->cancelled)
3078                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3079         return ret;
3080 }
3081
3082 /*
3083  * Returns a referenced nfs4_state
3084  */
3085 static int _nfs4_do_open(struct inode *dir,
3086                         struct nfs_open_context *ctx,
3087                         int flags,
3088                         const struct nfs4_open_createattrs *c,
3089                         int *opened)
3090 {
3091         struct nfs4_state_owner  *sp;
3092         struct nfs4_state     *state = NULL;
3093         struct nfs_server       *server = NFS_SERVER(dir);
3094         struct nfs4_opendata *opendata;
3095         struct dentry *dentry = ctx->dentry;
3096         const struct cred *cred = ctx->cred;
3097         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3098         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3099         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3100         struct iattr *sattr = c->sattr;
3101         struct nfs4_label *label = c->label;
3102         struct nfs4_label *olabel = NULL;
3103         int status;
3104
3105         /* Protect against reboot recovery conflicts */
3106         status = -ENOMEM;
3107         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3108         if (sp == NULL) {
3109                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3110                 goto out_err;
3111         }
3112         status = nfs4_client_recover_expired_lease(server->nfs_client);
3113         if (status != 0)
3114                 goto err_put_state_owner;
3115         if (d_really_is_positive(dentry))
3116                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3117         status = -ENOMEM;
3118         if (d_really_is_positive(dentry))
3119                 claim = NFS4_OPEN_CLAIM_FH;
3120         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3121                         c, claim, GFP_KERNEL);
3122         if (opendata == NULL)
3123                 goto err_put_state_owner;
3124
3125         if (label) {
3126                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3127                 if (IS_ERR(olabel)) {
3128                         status = PTR_ERR(olabel);
3129                         goto err_opendata_put;
3130                 }
3131         }
3132
3133         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3134                 if (!opendata->f_attr.mdsthreshold) {
3135                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3136                         if (!opendata->f_attr.mdsthreshold)
3137                                 goto err_free_label;
3138                 }
3139                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3140         }
3141         if (d_really_is_positive(dentry))
3142                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3143
3144         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3145         if (status != 0)
3146                 goto err_free_label;
3147         state = ctx->state;
3148
3149         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3150             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3151                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3152                 /*
3153                  * send create attributes which was not set by open
3154                  * with an extra setattr.
3155                  */
3156                 if (attrs || label) {
3157                         unsigned ia_old = sattr->ia_valid;
3158
3159                         sattr->ia_valid = attrs;
3160                         nfs_fattr_init(opendata->o_res.f_attr);
3161                         status = nfs4_do_setattr(state->inode, cred,
3162                                         opendata->o_res.f_attr, sattr,
3163                                         ctx, label, olabel);
3164                         if (status == 0) {
3165                                 nfs_setattr_update_inode(state->inode, sattr,
3166                                                 opendata->o_res.f_attr);
3167                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3168                         }
3169                         sattr->ia_valid = ia_old;
3170                 }
3171         }
3172         if (opened && opendata->file_created)
3173                 *opened = 1;
3174
3175         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3176                 *ctx_th = opendata->f_attr.mdsthreshold;
3177                 opendata->f_attr.mdsthreshold = NULL;
3178         }
3179
3180         nfs4_label_free(olabel);
3181
3182         nfs4_opendata_put(opendata);
3183         nfs4_put_state_owner(sp);
3184         return 0;
3185 err_free_label:
3186         nfs4_label_free(olabel);
3187 err_opendata_put:
3188         nfs4_opendata_put(opendata);
3189 err_put_state_owner:
3190         nfs4_put_state_owner(sp);
3191 out_err:
3192         return status;
3193 }
3194
3195
3196 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3197                                         struct nfs_open_context *ctx,
3198                                         int flags,
3199                                         struct iattr *sattr,
3200                                         struct nfs4_label *label,
3201                                         int *opened)
3202 {
3203         struct nfs_server *server = NFS_SERVER(dir);
3204         struct nfs4_exception exception = {
3205                 .interruptible = true,
3206         };
3207         struct nfs4_state *res;
3208         struct nfs4_open_createattrs c = {
3209                 .label = label,
3210                 .sattr = sattr,
3211                 .verf = {
3212                         [0] = (__u32)jiffies,
3213                         [1] = (__u32)current->pid,
3214                 },
3215         };
3216         int status;
3217
3218         do {
3219                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3220                 res = ctx->state;
3221                 trace_nfs4_open_file(ctx, flags, status);
3222                 if (status == 0)
3223                         break;
3224                 /* NOTE: BAD_SEQID means the server and client disagree about the
3225                  * book-keeping w.r.t. state-changing operations
3226                  * (OPEN/CLOSE/LOCK/LOCKU...)
3227                  * It is actually a sign of a bug on the client or on the server.
3228                  *
3229                  * If we receive a BAD_SEQID error in the particular case of
3230                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3231                  * have unhashed the old state_owner for us, and that we can
3232                  * therefore safely retry using a new one. We should still warn
3233                  * the user though...
3234                  */
3235                 if (status == -NFS4ERR_BAD_SEQID) {
3236                         pr_warn_ratelimited("NFS: v4 server %s "
3237                                         " returned a bad sequence-id error!\n",
3238                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3239                         exception.retry = 1;
3240                         continue;
3241                 }
3242                 /*
3243                  * BAD_STATEID on OPEN means that the server cancelled our
3244                  * state before it received the OPEN_CONFIRM.
3245                  * Recover by retrying the request as per the discussion
3246                  * on Page 181 of RFC3530.
3247                  */
3248                 if (status == -NFS4ERR_BAD_STATEID) {
3249                         exception.retry = 1;
3250                         continue;
3251                 }
3252                 if (status == -NFS4ERR_EXPIRED) {
3253                         nfs4_schedule_lease_recovery(server->nfs_client);
3254                         exception.retry = 1;
3255                         continue;
3256                 }
3257                 if (status == -EAGAIN) {
3258                         /* We must have found a delegation */
3259                         exception.retry = 1;
3260                         continue;
3261                 }
3262                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3263                         continue;
3264                 res = ERR_PTR(nfs4_handle_exception(server,
3265                                         status, &exception));
3266         } while (exception.retry);
3267         return res;
3268 }
3269
3270 static int _nfs4_do_setattr(struct inode *inode,
3271                             struct nfs_setattrargs *arg,
3272                             struct nfs_setattrres *res,
3273                             const struct cred *cred,
3274                             struct nfs_open_context *ctx)
3275 {
3276         struct nfs_server *server = NFS_SERVER(inode);
3277         struct rpc_message msg = {
3278                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3279                 .rpc_argp       = arg,
3280                 .rpc_resp       = res,
3281                 .rpc_cred       = cred,
3282         };
3283         const struct cred *delegation_cred = NULL;
3284         unsigned long timestamp = jiffies;
3285         bool truncate;
3286         int status;
3287
3288         nfs_fattr_init(res->fattr);
3289
3290         /* Servers should only apply open mode checks for file size changes */
3291         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3292         if (!truncate) {
3293                 nfs4_inode_make_writeable(inode);
3294                 goto zero_stateid;
3295         }
3296
3297         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3298                 /* Use that stateid */
3299         } else if (ctx != NULL && ctx->state) {
3300                 struct nfs_lock_context *l_ctx;
3301                 if (!nfs4_valid_open_stateid(ctx->state))
3302                         return -EBADF;
3303                 l_ctx = nfs_get_lock_context(ctx);
3304                 if (IS_ERR(l_ctx))
3305                         return PTR_ERR(l_ctx);
3306                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3307                                                 &arg->stateid, &delegation_cred);
3308                 nfs_put_lock_context(l_ctx);
3309                 if (status == -EIO)
3310                         return -EBADF;
3311                 else if (status == -EAGAIN)
3312                         goto zero_stateid;
3313         } else {
3314 zero_stateid:
3315                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3316         }
3317         if (delegation_cred)
3318                 msg.rpc_cred = delegation_cred;
3319
3320         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3321
3322         put_cred(delegation_cred);
3323         if (status == 0 && ctx != NULL)
3324                 renew_lease(server, timestamp);
3325         trace_nfs4_setattr(inode, &arg->stateid, status);
3326         return status;
3327 }
3328
3329 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3330                            struct nfs_fattr *fattr, struct iattr *sattr,
3331                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3332                            struct nfs4_label *olabel)
3333 {
3334         struct nfs_server *server = NFS_SERVER(inode);
3335         __u32 bitmask[NFS4_BITMASK_SZ];
3336         struct nfs4_state *state = ctx ? ctx->state : NULL;
3337         struct nfs_setattrargs  arg = {
3338                 .fh             = NFS_FH(inode),
3339                 .iap            = sattr,
3340                 .server         = server,
3341                 .bitmask = bitmask,
3342                 .label          = ilabel,
3343         };
3344         struct nfs_setattrres  res = {
3345                 .fattr          = fattr,
3346                 .label          = olabel,
3347                 .server         = server,
3348         };
3349         struct nfs4_exception exception = {
3350                 .state = state,
3351                 .inode = inode,
3352                 .stateid = &arg.stateid,
3353         };
3354         int err;
3355
3356         do {
3357                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3358                                 nfs4_bitmask(server, olabel),
3359                                 inode);
3360
3361                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3362                 switch (err) {
3363                 case -NFS4ERR_OPENMODE:
3364                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3365                                 pr_warn_once("NFSv4: server %s is incorrectly "
3366                                                 "applying open mode checks to "
3367                                                 "a SETATTR that is not "
3368                                                 "changing file size.\n",
3369                                                 server->nfs_client->cl_hostname);
3370                         }
3371                         if (state && !(state->state & FMODE_WRITE)) {
3372                                 err = -EBADF;
3373                                 if (sattr->ia_valid & ATTR_OPEN)
3374                                         err = -EACCES;
3375                                 goto out;
3376                         }
3377                 }
3378                 err = nfs4_handle_exception(server, err, &exception);
3379         } while (exception.retry);
3380 out:
3381         return err;
3382 }
3383
3384 static bool
3385 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3386 {
3387         if (inode == NULL || !nfs_have_layout(inode))
3388                 return false;
3389
3390         return pnfs_wait_on_layoutreturn(inode, task);
3391 }
3392
3393 /*
3394  * Update the seqid of an open stateid
3395  */
3396 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3397                 struct nfs4_state *state)
3398 {
3399         __be32 seqid_open;
3400         u32 dst_seqid;
3401         int seq;
3402
3403         for (;;) {
3404                 if (!nfs4_valid_open_stateid(state))
3405                         break;
3406                 seq = read_seqbegin(&state->seqlock);
3407                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3408                         nfs4_stateid_copy(dst, &state->open_stateid);
3409                         if (read_seqretry(&state->seqlock, seq))
3410                                 continue;
3411                         break;
3412                 }
3413                 seqid_open = state->open_stateid.seqid;
3414                 if (read_seqretry(&state->seqlock, seq))
3415                         continue;
3416
3417                 dst_seqid = be32_to_cpu(dst->seqid);
3418                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3419                         dst->seqid = seqid_open;
3420                 break;
3421         }
3422 }
3423
3424 /*
3425  * Update the seqid of an open stateid after receiving
3426  * NFS4ERR_OLD_STATEID
3427  */
3428 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3429                 struct nfs4_state *state)
3430 {
3431         __be32 seqid_open;
3432         u32 dst_seqid;
3433         bool ret;
3434         int seq, status = -EAGAIN;
3435         DEFINE_WAIT(wait);
3436
3437         for (;;) {
3438                 ret = false;
3439                 if (!nfs4_valid_open_stateid(state))
3440                         break;
3441                 seq = read_seqbegin(&state->seqlock);
3442                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3443                         if (read_seqretry(&state->seqlock, seq))
3444                                 continue;
3445                         break;
3446                 }
3447
3448                 write_seqlock(&state->seqlock);
3449                 seqid_open = state->open_stateid.seqid;
3450
3451                 dst_seqid = be32_to_cpu(dst->seqid);
3452
3453                 /* Did another OPEN bump the state's seqid?  try again: */
3454                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3455                         dst->seqid = seqid_open;
3456                         write_sequnlock(&state->seqlock);
3457                         ret = true;
3458                         break;
3459                 }
3460
3461                 /* server says we're behind but we haven't seen the update yet */
3462                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3463                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3464                 write_sequnlock(&state->seqlock);
3465                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3466
3467                 if (signal_pending(current))
3468                         status = -EINTR;
3469                 else
3470                         if (schedule_timeout(5*HZ) != 0)
3471                                 status = 0;
3472
3473                 finish_wait(&state->waitq, &wait);
3474
3475                 if (!status)
3476                         continue;
3477                 if (status == -EINTR)
3478                         break;
3479
3480                 /* we slept the whole 5 seconds, we must have lost a seqid */
3481                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3482                 ret = true;
3483                 break;
3484         }
3485
3486         return ret;
3487 }
3488
3489 struct nfs4_closedata {
3490         struct inode *inode;
3491         struct nfs4_state *state;
3492         struct nfs_closeargs arg;
3493         struct nfs_closeres res;
3494         struct {
3495                 struct nfs4_layoutreturn_args arg;
3496                 struct nfs4_layoutreturn_res res;
3497                 struct nfs4_xdr_opaque_data ld_private;
3498                 u32 roc_barrier;
3499                 bool roc;
3500         } lr;
3501         struct nfs_fattr fattr;
3502         unsigned long timestamp;
3503 };
3504
3505 static void nfs4_free_closedata(void *data)
3506 {
3507         struct nfs4_closedata *calldata = data;
3508         struct nfs4_state_owner *sp = calldata->state->owner;
3509         struct super_block *sb = calldata->state->inode->i_sb;
3510
3511         if (calldata->lr.roc)
3512                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3513                                 calldata->res.lr_ret);
3514         nfs4_put_open_state(calldata->state);
3515         nfs_free_seqid(calldata->arg.seqid);
3516         nfs4_put_state_owner(sp);
3517         nfs_sb_deactive(sb);
3518         kfree(calldata);
3519 }
3520
3521 static void nfs4_close_done(struct rpc_task *task, void *data)
3522 {
3523         struct nfs4_closedata *calldata = data;
3524         struct nfs4_state *state = calldata->state;
3525         struct nfs_server *server = NFS_SERVER(calldata->inode);
3526         nfs4_stateid *res_stateid = NULL;
3527         struct nfs4_exception exception = {
3528                 .state = state,
3529                 .inode = calldata->inode,
3530                 .stateid = &calldata->arg.stateid,
3531         };
3532
3533         dprintk("%s: begin!\n", __func__);
3534         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3535                 return;
3536         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3537
3538         /* Handle Layoutreturn errors */
3539         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3540                           &calldata->res.lr_ret) == -EAGAIN)
3541                 goto out_restart;
3542
3543         /* hmm. we are done with the inode, and in the process of freeing
3544          * the state_owner. we keep this around to process errors
3545          */
3546         switch (task->tk_status) {
3547                 case 0:
3548                         res_stateid = &calldata->res.stateid;
3549                         renew_lease(server, calldata->timestamp);
3550                         break;
3551                 case -NFS4ERR_ACCESS:
3552                         if (calldata->arg.bitmask != NULL) {
3553                                 calldata->arg.bitmask = NULL;
3554                                 calldata->res.fattr = NULL;
3555                                 goto out_restart;
3556
3557                         }
3558                         break;
3559                 case -NFS4ERR_OLD_STATEID:
3560                         /* Did we race with OPEN? */
3561                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3562                                                 state))
3563                                 goto out_restart;
3564                         goto out_release;
3565                 case -NFS4ERR_ADMIN_REVOKED:
3566                 case -NFS4ERR_STALE_STATEID:
3567                 case -NFS4ERR_EXPIRED:
3568                         nfs4_free_revoked_stateid(server,
3569                                         &calldata->arg.stateid,
3570                                         task->tk_msg.rpc_cred);
3571                         fallthrough;
3572                 case -NFS4ERR_BAD_STATEID:
3573                         if (calldata->arg.fmode == 0)
3574                                 break;
3575                         fallthrough;
3576                 default:
3577                         task->tk_status = nfs4_async_handle_exception(task,
3578                                         server, task->tk_status, &exception);
3579                         if (exception.retry)
3580                                 goto out_restart;
3581         }
3582         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3583                         res_stateid, calldata->arg.fmode);
3584 out_release:
3585         task->tk_status = 0;
3586         nfs_release_seqid(calldata->arg.seqid);
3587         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3588         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3589         return;
3590 out_restart:
3591         task->tk_status = 0;
3592         rpc_restart_call_prepare(task);
3593         goto out_release;
3594 }
3595
3596 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3597 {
3598         struct nfs4_closedata *calldata = data;
3599         struct nfs4_state *state = calldata->state;
3600         struct inode *inode = calldata->inode;
3601         struct pnfs_layout_hdr *lo;
3602         bool is_rdonly, is_wronly, is_rdwr;
3603         int call_close = 0;
3604
3605         dprintk("%s: begin!\n", __func__);
3606         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3607                 goto out_wait;
3608
3609         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3610         spin_lock(&state->owner->so_lock);
3611         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3612         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3613         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3614         /* Calculate the change in open mode */
3615         calldata->arg.fmode = 0;
3616         if (state->n_rdwr == 0) {
3617                 if (state->n_rdonly == 0)
3618                         call_close |= is_rdonly;
3619                 else if (is_rdonly)
3620                         calldata->arg.fmode |= FMODE_READ;
3621                 if (state->n_wronly == 0)
3622                         call_close |= is_wronly;
3623                 else if (is_wronly)
3624                         calldata->arg.fmode |= FMODE_WRITE;
3625                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3626                         call_close |= is_rdwr;
3627         } else if (is_rdwr)
3628                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3629
3630         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3631         if (!nfs4_valid_open_stateid(state))
3632                 call_close = 0;
3633         spin_unlock(&state->owner->so_lock);
3634
3635         if (!call_close) {
3636                 /* Note: exit _without_ calling nfs4_close_done */
3637                 goto out_no_action;
3638         }
3639
3640         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3641                 nfs_release_seqid(calldata->arg.seqid);
3642                 goto out_wait;
3643         }
3644
3645         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3646         if (lo && !pnfs_layout_is_valid(lo)) {
3647                 calldata->arg.lr_args = NULL;
3648                 calldata->res.lr_res = NULL;
3649         }
3650
3651         if (calldata->arg.fmode == 0)
3652                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3653
3654         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3655                 /* Close-to-open cache consistency revalidation */
3656                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3657                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3658                         nfs4_bitmask_adjust(calldata->arg.bitmask, inode, NFS_SERVER(inode), NULL);
3659                 } else
3660                         calldata->arg.bitmask = NULL;
3661         }
3662
3663         calldata->arg.share_access =
3664                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3665                                 calldata->arg.fmode, 0);
3666
3667         if (calldata->res.fattr == NULL)
3668                 calldata->arg.bitmask = NULL;
3669         else if (calldata->arg.bitmask == NULL)
3670                 calldata->res.fattr = NULL;
3671         calldata->timestamp = jiffies;
3672         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3673                                 &calldata->arg.seq_args,
3674                                 &calldata->res.seq_res,
3675                                 task) != 0)
3676                 nfs_release_seqid(calldata->arg.seqid);
3677         dprintk("%s: done!\n", __func__);
3678         return;
3679 out_no_action:
3680         task->tk_action = NULL;
3681 out_wait:
3682         nfs4_sequence_done(task, &calldata->res.seq_res);
3683 }
3684
3685 static const struct rpc_call_ops nfs4_close_ops = {
3686         .rpc_call_prepare = nfs4_close_prepare,
3687         .rpc_call_done = nfs4_close_done,
3688         .rpc_release = nfs4_free_closedata,
3689 };
3690
3691 /*
3692  * It is possible for data to be read/written from a mem-mapped file
3693  * after the sys_close call (which hits the vfs layer as a flush).
3694  * This means that we can't safely call nfsv4 close on a file until
3695  * the inode is cleared. This in turn means that we are not good
3696  * NFSv4 citizens - we do not indicate to the server to update the file's
3697  * share state even when we are done with one of the three share
3698  * stateid's in the inode.
3699  *
3700  * NOTE: Caller must be holding the sp->so_owner semaphore!
3701  */
3702 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3703 {
3704         struct nfs_server *server = NFS_SERVER(state->inode);
3705         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3706         struct nfs4_closedata *calldata;
3707         struct nfs4_state_owner *sp = state->owner;
3708         struct rpc_task *task;
3709         struct rpc_message msg = {
3710                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3711                 .rpc_cred = state->owner->so_cred,
3712         };
3713         struct rpc_task_setup task_setup_data = {
3714                 .rpc_client = server->client,
3715                 .rpc_message = &msg,
3716                 .callback_ops = &nfs4_close_ops,
3717                 .workqueue = nfsiod_workqueue,
3718                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3719         };
3720         int status = -ENOMEM;
3721
3722         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3723                 &task_setup_data.rpc_client, &msg);
3724
3725         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3726         if (calldata == NULL)
3727                 goto out;
3728         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3729         calldata->inode = state->inode;
3730         calldata->state = state;
3731         calldata->arg.fh = NFS_FH(state->inode);
3732         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3733                 goto out_free_calldata;
3734         /* Serialization for the sequence id */
3735         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3736         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3737         if (IS_ERR(calldata->arg.seqid))
3738                 goto out_free_calldata;
3739         nfs_fattr_init(&calldata->fattr);
3740         calldata->arg.fmode = 0;
3741         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3742         calldata->res.fattr = &calldata->fattr;
3743         calldata->res.seqid = calldata->arg.seqid;
3744         calldata->res.server = server;
3745         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3746         calldata->lr.roc = pnfs_roc(state->inode,
3747                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3748         if (calldata->lr.roc) {
3749                 calldata->arg.lr_args = &calldata->lr.arg;
3750                 calldata->res.lr_res = &calldata->lr.res;
3751         }
3752         nfs_sb_active(calldata->inode->i_sb);
3753
3754         msg.rpc_argp = &calldata->arg;
3755         msg.rpc_resp = &calldata->res;
3756         task_setup_data.callback_data = calldata;
3757         task = rpc_run_task(&task_setup_data);
3758         if (IS_ERR(task))
3759                 return PTR_ERR(task);
3760         status = 0;
3761         if (wait)
3762                 status = rpc_wait_for_completion_task(task);
3763         rpc_put_task(task);
3764         return status;
3765 out_free_calldata:
3766         kfree(calldata);
3767 out:
3768         nfs4_put_open_state(state);
3769         nfs4_put_state_owner(sp);
3770         return status;
3771 }
3772
3773 static struct inode *
3774 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3775                 int open_flags, struct iattr *attr, int *opened)
3776 {
3777         struct nfs4_state *state;
3778         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3779
3780         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3781
3782         /* Protect against concurrent sillydeletes */
3783         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3784
3785         nfs4_label_release_security(label);
3786
3787         if (IS_ERR(state))
3788                 return ERR_CAST(state);
3789         return state->inode;
3790 }
3791
3792 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3793 {
3794         if (ctx->state == NULL)
3795                 return;
3796         if (is_sync)
3797                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3798         else
3799                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3800 }
3801
3802 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3803 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3804 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3805
3806 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3807 {
3808         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3809         struct nfs4_server_caps_arg args = {
3810                 .fhandle = fhandle,
3811                 .bitmask = bitmask,
3812         };
3813         struct nfs4_server_caps_res res = {};
3814         struct rpc_message msg = {
3815                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3816                 .rpc_argp = &args,
3817                 .rpc_resp = &res,
3818         };
3819         int status;
3820         int i;
3821
3822         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3823                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3824                      FATTR4_WORD0_LINK_SUPPORT |
3825                      FATTR4_WORD0_SYMLINK_SUPPORT |
3826                      FATTR4_WORD0_ACLSUPPORT;
3827         if (minorversion)
3828                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3829
3830         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3831         if (status == 0) {
3832                 /* Sanity check the server answers */
3833                 switch (minorversion) {
3834                 case 0:
3835                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3836                         res.attr_bitmask[2] = 0;
3837                         break;
3838                 case 1:
3839                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3840                         break;
3841                 case 2:
3842                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3843                 }
3844                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3845                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3846                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3847                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3848                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3849                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3850                                 NFS_CAP_SECURITY_LABEL);
3851                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3852                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3853                         server->caps |= NFS_CAP_ACLS;
3854                 if (res.has_links != 0)
3855                         server->caps |= NFS_CAP_HARDLINKS;
3856                 if (res.has_symlinks != 0)
3857                         server->caps |= NFS_CAP_SYMLINKS;
3858                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3859                         server->caps |= NFS_CAP_FILEID;
3860                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3861                         server->caps |= NFS_CAP_MODE;
3862                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3863                         server->caps |= NFS_CAP_NLINK;
3864                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3865                         server->caps |= NFS_CAP_OWNER;
3866                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3867                         server->caps |= NFS_CAP_OWNER_GROUP;
3868                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3869                         server->caps |= NFS_CAP_ATIME;
3870                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3871                         server->caps |= NFS_CAP_CTIME;
3872                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3873                         server->caps |= NFS_CAP_MTIME;
3874 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3875                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3876                         server->caps |= NFS_CAP_SECURITY_LABEL;
3877 #endif
3878                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3879                                 sizeof(server->attr_bitmask));
3880                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3881
3882                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3883                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3884                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3885                 server->cache_consistency_bitmask[2] = 0;
3886
3887                 /* Avoid a regression due to buggy server */
3888                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3889                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3890                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3891                         sizeof(server->exclcreat_bitmask));
3892
3893                 server->acl_bitmask = res.acl_bitmask;
3894                 server->fh_expire_type = res.fh_expire_type;
3895         }
3896
3897         return status;
3898 }
3899
3900 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3901 {
3902         struct nfs4_exception exception = {
3903                 .interruptible = true,
3904         };
3905         int err;
3906         do {
3907                 err = nfs4_handle_exception(server,
3908                                 _nfs4_server_capabilities(server, fhandle),
3909                                 &exception);
3910         } while (exception.retry);
3911         return err;
3912 }
3913
3914 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3915                 struct nfs_fsinfo *info)
3916 {
3917         u32 bitmask[3];
3918         struct nfs4_lookup_root_arg args = {
3919                 .bitmask = bitmask,
3920         };
3921         struct nfs4_lookup_res res = {
3922                 .server = server,
3923                 .fattr = info->fattr,
3924                 .fh = fhandle,
3925         };
3926         struct rpc_message msg = {
3927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3928                 .rpc_argp = &args,
3929                 .rpc_resp = &res,
3930         };
3931
3932         bitmask[0] = nfs4_fattr_bitmap[0];
3933         bitmask[1] = nfs4_fattr_bitmap[1];
3934         /*
3935          * Process the label in the upcoming getfattr
3936          */
3937         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3938
3939         nfs_fattr_init(info->fattr);
3940         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3941 }
3942
3943 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3944                 struct nfs_fsinfo *info)
3945 {
3946         struct nfs4_exception exception = {
3947                 .interruptible = true,
3948         };
3949         int err;
3950         do {
3951                 err = _nfs4_lookup_root(server, fhandle, info);
3952                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3953                 switch (err) {
3954                 case 0:
3955                 case -NFS4ERR_WRONGSEC:
3956                         goto out;
3957                 default:
3958                         err = nfs4_handle_exception(server, err, &exception);
3959                 }
3960         } while (exception.retry);
3961 out:
3962         return err;
3963 }
3964
3965 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3966                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3967 {
3968         struct rpc_auth_create_args auth_args = {
3969                 .pseudoflavor = flavor,
3970         };
3971         struct rpc_auth *auth;
3972
3973         auth = rpcauth_create(&auth_args, server->client);
3974         if (IS_ERR(auth))
3975                 return -EACCES;
3976         return nfs4_lookup_root(server, fhandle, info);
3977 }
3978
3979 /*
3980  * Retry pseudoroot lookup with various security flavors.  We do this when:
3981  *
3982  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3983  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3984  *
3985  * Returns zero on success, or a negative NFS4ERR value, or a
3986  * negative errno value.
3987  */
3988 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3989                               struct nfs_fsinfo *info)
3990 {
3991         /* Per 3530bis 15.33.5 */
3992         static const rpc_authflavor_t flav_array[] = {
3993                 RPC_AUTH_GSS_KRB5P,
3994                 RPC_AUTH_GSS_KRB5I,
3995                 RPC_AUTH_GSS_KRB5,
3996                 RPC_AUTH_UNIX,                  /* courtesy */
3997                 RPC_AUTH_NULL,
3998         };
3999         int status = -EPERM;
4000         size_t i;
4001
4002         if (server->auth_info.flavor_len > 0) {
4003                 /* try each flavor specified by user */
4004                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4005                         status = nfs4_lookup_root_sec(server, fhandle, info,
4006                                                 server->auth_info.flavors[i]);
4007                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4008                                 continue;
4009                         break;
4010                 }
4011         } else {
4012                 /* no flavors specified by user, try default list */
4013                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4014                         status = nfs4_lookup_root_sec(server, fhandle, info,
4015                                                       flav_array[i]);
4016                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4017                                 continue;
4018                         break;
4019                 }
4020         }
4021
4022         /*
4023          * -EACCES could mean that the user doesn't have correct permissions
4024          * to access the mount.  It could also mean that we tried to mount
4025          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4026          * existing mount programs don't handle -EACCES very well so it should
4027          * be mapped to -EPERM instead.
4028          */
4029         if (status == -EACCES)
4030                 status = -EPERM;
4031         return status;
4032 }
4033
4034 /**
4035  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4036  * @server: initialized nfs_server handle
4037  * @fhandle: we fill in the pseudo-fs root file handle
4038  * @info: we fill in an FSINFO struct
4039  * @auth_probe: probe the auth flavours
4040  *
4041  * Returns zero on success, or a negative errno.
4042  */
4043 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4044                          struct nfs_fsinfo *info,
4045                          bool auth_probe)
4046 {
4047         int status = 0;
4048
4049         if (!auth_probe)
4050                 status = nfs4_lookup_root(server, fhandle, info);
4051
4052         if (auth_probe || status == NFS4ERR_WRONGSEC)
4053                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4054                                 fhandle, info);
4055
4056         if (status == 0)
4057                 status = nfs4_server_capabilities(server, fhandle);
4058         if (status == 0)
4059                 status = nfs4_do_fsinfo(server, fhandle, info);
4060
4061         return nfs4_map_errors(status);
4062 }
4063
4064 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4065                               struct nfs_fsinfo *info)
4066 {
4067         int error;
4068         struct nfs_fattr *fattr = info->fattr;
4069         struct nfs4_label *label = fattr->label;
4070
4071         error = nfs4_server_capabilities(server, mntfh);
4072         if (error < 0) {
4073                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4074                 return error;
4075         }
4076
4077         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4078         if (error < 0) {
4079                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4080                 goto out;
4081         }
4082
4083         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4084             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4085                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4086
4087 out:
4088         return error;
4089 }
4090
4091 /*
4092  * Get locations and (maybe) other attributes of a referral.
4093  * Note that we'll actually follow the referral later when
4094  * we detect fsid mismatch in inode revalidation
4095  */
4096 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4097                              const struct qstr *name, struct nfs_fattr *fattr,
4098                              struct nfs_fh *fhandle)
4099 {
4100         int status = -ENOMEM;
4101         struct page *page = NULL;
4102         struct nfs4_fs_locations *locations = NULL;
4103
4104         page = alloc_page(GFP_KERNEL);
4105         if (page == NULL)
4106                 goto out;
4107         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4108         if (locations == NULL)
4109                 goto out;
4110
4111         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4112         if (status != 0)
4113                 goto out;
4114
4115         /*
4116          * If the fsid didn't change, this is a migration event, not a
4117          * referral.  Cause us to drop into the exception handler, which
4118          * will kick off migration recovery.
4119          */
4120         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4121                 dprintk("%s: server did not return a different fsid for"
4122                         " a referral at %s\n", __func__, name->name);
4123                 status = -NFS4ERR_MOVED;
4124                 goto out;
4125         }
4126         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4127         nfs_fixup_referral_attributes(&locations->fattr);
4128
4129         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4130         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4131         memset(fhandle, 0, sizeof(struct nfs_fh));
4132 out:
4133         if (page)
4134                 __free_page(page);
4135         kfree(locations);
4136         return status;
4137 }
4138
4139 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4140                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4141                                 struct inode *inode)
4142 {
4143         __u32 bitmask[NFS4_BITMASK_SZ];
4144         struct nfs4_getattr_arg args = {
4145                 .fh = fhandle,
4146                 .bitmask = bitmask,
4147         };
4148         struct nfs4_getattr_res res = {
4149                 .fattr = fattr,
4150                 .label = label,
4151                 .server = server,
4152         };
4153         struct rpc_message msg = {
4154                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4155                 .rpc_argp = &args,
4156                 .rpc_resp = &res,
4157         };
4158         unsigned short task_flags = 0;
4159
4160         /* Is this is an attribute revalidation, subject to softreval? */
4161         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4162                 task_flags |= RPC_TASK_TIMEOUT;
4163
4164         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4165
4166         nfs_fattr_init(fattr);
4167         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4168         return nfs4_do_call_sync(server->client, server, &msg,
4169                         &args.seq_args, &res.seq_res, task_flags);
4170 }
4171
4172 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4173                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4174                                 struct inode *inode)
4175 {
4176         struct nfs4_exception exception = {
4177                 .interruptible = true,
4178         };
4179         int err;
4180         do {
4181                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4182                 trace_nfs4_getattr(server, fhandle, fattr, err);
4183                 err = nfs4_handle_exception(server, err,
4184                                 &exception);
4185         } while (exception.retry);
4186         return err;
4187 }
4188
4189 /*
4190  * The file is not closed if it is opened due to the a request to change
4191  * the size of the file. The open call will not be needed once the
4192  * VFS layer lookup-intents are implemented.
4193  *
4194  * Close is called when the inode is destroyed.
4195  * If we haven't opened the file for O_WRONLY, we
4196  * need to in the size_change case to obtain a stateid.
4197  *
4198  * Got race?
4199  * Because OPEN is always done by name in nfsv4, it is
4200  * possible that we opened a different file by the same
4201  * name.  We can recognize this race condition, but we
4202  * can't do anything about it besides returning an error.
4203  *
4204  * This will be fixed with VFS changes (lookup-intent).
4205  */
4206 static int
4207 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4208                   struct iattr *sattr)
4209 {
4210         struct inode *inode = d_inode(dentry);
4211         const struct cred *cred = NULL;
4212         struct nfs_open_context *ctx = NULL;
4213         struct nfs4_label *label = NULL;
4214         int status;
4215
4216         if (pnfs_ld_layoutret_on_setattr(inode) &&
4217             sattr->ia_valid & ATTR_SIZE &&
4218             sattr->ia_size < i_size_read(inode))
4219                 pnfs_commit_and_return_layout(inode);
4220
4221         nfs_fattr_init(fattr);
4222
4223         /* Deal with open(O_TRUNC) */
4224         if (sattr->ia_valid & ATTR_OPEN)
4225                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4226
4227         /* Optimization: if the end result is no change, don't RPC */
4228         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4229                 return 0;
4230
4231         /* Search for an existing open(O_WRITE) file */
4232         if (sattr->ia_valid & ATTR_FILE) {
4233
4234                 ctx = nfs_file_open_context(sattr->ia_file);
4235                 if (ctx)
4236                         cred = ctx->cred;
4237         }
4238
4239         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4240         if (IS_ERR(label))
4241                 return PTR_ERR(label);
4242
4243         /* Return any delegations if we're going to change ACLs */
4244         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4245                 nfs4_inode_make_writeable(inode);
4246
4247         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4248         if (status == 0) {
4249                 nfs_setattr_update_inode(inode, sattr, fattr);
4250                 nfs_setsecurity(inode, fattr, label);
4251         }
4252         nfs4_label_free(label);
4253         return status;
4254 }
4255
4256 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4257                 struct dentry *dentry, struct nfs_fh *fhandle,
4258                 struct nfs_fattr *fattr, struct nfs4_label *label)
4259 {
4260         struct nfs_server *server = NFS_SERVER(dir);
4261         int                    status;
4262         struct nfs4_lookup_arg args = {
4263                 .bitmask = server->attr_bitmask,
4264                 .dir_fh = NFS_FH(dir),
4265                 .name = &dentry->d_name,
4266         };
4267         struct nfs4_lookup_res res = {
4268                 .server = server,
4269                 .fattr = fattr,
4270                 .label = label,
4271                 .fh = fhandle,
4272         };
4273         struct rpc_message msg = {
4274                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4275                 .rpc_argp = &args,
4276                 .rpc_resp = &res,
4277         };
4278         unsigned short task_flags = 0;
4279
4280         /* Is this is an attribute revalidation, subject to softreval? */
4281         if (nfs_lookup_is_soft_revalidate(dentry))
4282                 task_flags |= RPC_TASK_TIMEOUT;
4283
4284         args.bitmask = nfs4_bitmask(server, label);
4285
4286         nfs_fattr_init(fattr);
4287
4288         dprintk("NFS call  lookup %pd2\n", dentry);
4289         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4290         status = nfs4_do_call_sync(clnt, server, &msg,
4291                         &args.seq_args, &res.seq_res, task_flags);
4292         dprintk("NFS reply lookup: %d\n", status);
4293         return status;
4294 }
4295
4296 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4297 {
4298         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4299                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4300         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4301         fattr->nlink = 2;
4302 }
4303
4304 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4305                                    struct dentry *dentry, struct nfs_fh *fhandle,
4306                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4307 {
4308         struct nfs4_exception exception = {
4309                 .interruptible = true,
4310         };
4311         struct rpc_clnt *client = *clnt;
4312         const struct qstr *name = &dentry->d_name;
4313         int err;
4314         do {
4315                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4316                 trace_nfs4_lookup(dir, name, err);
4317                 switch (err) {
4318                 case -NFS4ERR_BADNAME:
4319                         err = -ENOENT;
4320                         goto out;
4321                 case -NFS4ERR_MOVED:
4322                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4323                         if (err == -NFS4ERR_MOVED)
4324                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4325                         goto out;
4326                 case -NFS4ERR_WRONGSEC:
4327                         err = -EPERM;
4328                         if (client != *clnt)
4329                                 goto out;
4330                         client = nfs4_negotiate_security(client, dir, name);
4331                         if (IS_ERR(client))
4332                                 return PTR_ERR(client);
4333
4334                         exception.retry = 1;
4335                         break;
4336                 default:
4337                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4338                 }
4339         } while (exception.retry);
4340
4341 out:
4342         if (err == 0)
4343                 *clnt = client;
4344         else if (client != *clnt)
4345                 rpc_shutdown_client(client);
4346
4347         return err;
4348 }
4349
4350 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4351                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4352                             struct nfs4_label *label)
4353 {
4354         int status;
4355         struct rpc_clnt *client = NFS_CLIENT(dir);
4356
4357         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4358         if (client != NFS_CLIENT(dir)) {
4359                 rpc_shutdown_client(client);
4360                 nfs_fixup_secinfo_attributes(fattr);
4361         }
4362         return status;
4363 }
4364
4365 struct rpc_clnt *
4366 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4367                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4368 {
4369         struct rpc_clnt *client = NFS_CLIENT(dir);
4370         int status;
4371
4372         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4373         if (status < 0)
4374                 return ERR_PTR(status);
4375         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4376 }
4377
4378 static int _nfs4_proc_lookupp(struct inode *inode,
4379                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4380                 struct nfs4_label *label)
4381 {
4382         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4383         struct nfs_server *server = NFS_SERVER(inode);
4384         int                    status;
4385         struct nfs4_lookupp_arg args = {
4386                 .bitmask = server->attr_bitmask,
4387                 .fh = NFS_FH(inode),
4388         };
4389         struct nfs4_lookupp_res res = {
4390                 .server = server,
4391                 .fattr = fattr,
4392                 .label = label,
4393                 .fh = fhandle,
4394         };
4395         struct rpc_message msg = {
4396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4397                 .rpc_argp = &args,
4398                 .rpc_resp = &res,
4399         };
4400         unsigned short task_flags = 0;
4401
4402         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4403                 task_flags |= RPC_TASK_TIMEOUT;
4404
4405         args.bitmask = nfs4_bitmask(server, label);
4406
4407         nfs_fattr_init(fattr);
4408
4409         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4410         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4411                                 &res.seq_res, task_flags);
4412         dprintk("NFS reply lookupp: %d\n", status);
4413         return status;
4414 }
4415
4416 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4417                              struct nfs_fattr *fattr, struct nfs4_label *label)
4418 {
4419         struct nfs4_exception exception = {
4420                 .interruptible = true,
4421         };
4422         int err;
4423         do {
4424                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4425                 trace_nfs4_lookupp(inode, err);
4426                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4427                                 &exception);
4428         } while (exception.retry);
4429         return err;
4430 }
4431
4432 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4433 {
4434         struct nfs_server *server = NFS_SERVER(inode);
4435         struct nfs4_accessargs args = {
4436                 .fh = NFS_FH(inode),
4437                 .access = entry->mask,
4438         };
4439         struct nfs4_accessres res = {
4440                 .server = server,
4441         };
4442         struct rpc_message msg = {
4443                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4444                 .rpc_argp = &args,
4445                 .rpc_resp = &res,
4446                 .rpc_cred = entry->cred,
4447         };
4448         int status = 0;
4449
4450         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4451                 res.fattr = nfs_alloc_fattr();
4452                 if (res.fattr == NULL)
4453                         return -ENOMEM;
4454                 args.bitmask = server->cache_consistency_bitmask;
4455         }
4456         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4457         if (!status) {
4458                 nfs_access_set_mask(entry, res.access);
4459                 if (res.fattr)
4460                         nfs_refresh_inode(inode, res.fattr);
4461         }
4462         nfs_free_fattr(res.fattr);
4463         return status;
4464 }
4465
4466 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4467 {
4468         struct nfs4_exception exception = {
4469                 .interruptible = true,
4470         };
4471         int err;
4472         do {
4473                 err = _nfs4_proc_access(inode, entry);
4474                 trace_nfs4_access(inode, err);
4475                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4476                                 &exception);
4477         } while (exception.retry);
4478         return err;
4479 }
4480
4481 /*
4482  * TODO: For the time being, we don't try to get any attributes
4483  * along with any of the zero-copy operations READ, READDIR,
4484  * READLINK, WRITE.
4485  *
4486  * In the case of the first three, we want to put the GETATTR
4487  * after the read-type operation -- this is because it is hard
4488  * to predict the length of a GETATTR response in v4, and thus
4489  * align the READ data correctly.  This means that the GETATTR
4490  * may end up partially falling into the page cache, and we should
4491  * shift it into the 'tail' of the xdr_buf before processing.
4492  * To do this efficiently, we need to know the total length
4493  * of data received, which doesn't seem to be available outside
4494  * of the RPC layer.
4495  *
4496  * In the case of WRITE, we also want to put the GETATTR after
4497  * the operation -- in this case because we want to make sure
4498  * we get the post-operation mtime and size.
4499  *
4500  * Both of these changes to the XDR layer would in fact be quite
4501  * minor, but I decided to leave them for a subsequent patch.
4502  */
4503 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4504                 unsigned int pgbase, unsigned int pglen)
4505 {
4506         struct nfs4_readlink args = {
4507                 .fh       = NFS_FH(inode),
4508                 .pgbase   = pgbase,
4509                 .pglen    = pglen,
4510                 .pages    = &page,
4511         };
4512         struct nfs4_readlink_res res;
4513         struct rpc_message msg = {
4514                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4515                 .rpc_argp = &args,
4516                 .rpc_resp = &res,
4517         };
4518
4519         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4520 }
4521
4522 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4523                 unsigned int pgbase, unsigned int pglen)
4524 {
4525         struct nfs4_exception exception = {
4526                 .interruptible = true,
4527         };
4528         int err;
4529         do {
4530                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4531                 trace_nfs4_readlink(inode, err);
4532                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4533                                 &exception);
4534         } while (exception.retry);
4535         return err;
4536 }
4537
4538 /*
4539  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4540  */
4541 static int
4542 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4543                  int flags)
4544 {
4545         struct nfs_server *server = NFS_SERVER(dir);
4546         struct nfs4_label l, *ilabel = NULL;
4547         struct nfs_open_context *ctx;
4548         struct nfs4_state *state;
4549         int status = 0;
4550
4551         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4552         if (IS_ERR(ctx))
4553                 return PTR_ERR(ctx);
4554
4555         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4556
4557         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4558                 sattr->ia_mode &= ~current_umask();
4559         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4560         if (IS_ERR(state)) {
4561                 status = PTR_ERR(state);
4562                 goto out;
4563         }
4564 out:
4565         nfs4_label_release_security(ilabel);
4566         put_nfs_open_context(ctx);
4567         return status;
4568 }
4569
4570 static int
4571 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4572 {
4573         struct nfs_server *server = NFS_SERVER(dir);
4574         struct nfs_removeargs args = {
4575                 .fh = NFS_FH(dir),
4576                 .name = *name,
4577         };
4578         struct nfs_removeres res = {
4579                 .server = server,
4580         };
4581         struct rpc_message msg = {
4582                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4583                 .rpc_argp = &args,
4584                 .rpc_resp = &res,
4585         };
4586         unsigned long timestamp = jiffies;
4587         int status;
4588
4589         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4590         if (status == 0) {
4591                 spin_lock(&dir->i_lock);
4592                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4593                                               NFS_INO_INVALID_DATA);
4594                 /* Removing a directory decrements nlink in the parent */
4595                 if (ftype == NF4DIR && dir->i_nlink > 2)
4596                         nfs4_dec_nlink_locked(dir);
4597                 spin_unlock(&dir->i_lock);
4598         }
4599         return status;
4600 }
4601
4602 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4603 {
4604         struct nfs4_exception exception = {
4605                 .interruptible = true,
4606         };
4607         struct inode *inode = d_inode(dentry);
4608         int err;
4609
4610         if (inode) {
4611                 if (inode->i_nlink == 1)
4612                         nfs4_inode_return_delegation(inode);
4613                 else
4614                         nfs4_inode_make_writeable(inode);
4615         }
4616         do {
4617                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4618                 trace_nfs4_remove(dir, &dentry->d_name, err);
4619                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4620                                 &exception);
4621         } while (exception.retry);
4622         return err;
4623 }
4624
4625 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4626 {
4627         struct nfs4_exception exception = {
4628                 .interruptible = true,
4629         };
4630         int err;
4631
4632         do {
4633                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4634                 trace_nfs4_remove(dir, name, err);
4635                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4636                                 &exception);
4637         } while (exception.retry);
4638         return err;
4639 }
4640
4641 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4642                 struct dentry *dentry,
4643                 struct inode *inode)
4644 {
4645         struct nfs_removeargs *args = msg->rpc_argp;
4646         struct nfs_removeres *res = msg->rpc_resp;
4647
4648         res->server = NFS_SB(dentry->d_sb);
4649         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4650         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4651
4652         nfs_fattr_init(res->dir_attr);
4653
4654         if (inode)
4655                 nfs4_inode_return_delegation(inode);
4656 }
4657
4658 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4659 {
4660         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4661                         &data->args.seq_args,
4662                         &data->res.seq_res,
4663                         task);
4664 }
4665
4666 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4667 {
4668         struct nfs_unlinkdata *data = task->tk_calldata;
4669         struct nfs_removeres *res = &data->res;
4670
4671         if (!nfs4_sequence_done(task, &res->seq_res))
4672                 return 0;
4673         if (nfs4_async_handle_error(task, res->server, NULL,
4674                                     &data->timeout) == -EAGAIN)
4675                 return 0;
4676         if (task->tk_status == 0)
4677                 nfs4_update_changeattr(dir, &res->cinfo,
4678                                 res->dir_attr->time_start,
4679                                 NFS_INO_INVALID_DATA);
4680         return 1;
4681 }
4682
4683 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4684                 struct dentry *old_dentry,
4685                 struct dentry *new_dentry)
4686 {
4687         struct nfs_renameargs *arg = msg->rpc_argp;
4688         struct nfs_renameres *res = msg->rpc_resp;
4689         struct inode *old_inode = d_inode(old_dentry);
4690         struct inode *new_inode = d_inode(new_dentry);
4691
4692         if (old_inode)
4693                 nfs4_inode_make_writeable(old_inode);
4694         if (new_inode)
4695                 nfs4_inode_return_delegation(new_inode);
4696         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4697         res->server = NFS_SB(old_dentry->d_sb);
4698         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4699 }
4700
4701 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4702 {
4703         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4704                         &data->args.seq_args,
4705                         &data->res.seq_res,
4706                         task);
4707 }
4708
4709 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4710                                  struct inode *new_dir)
4711 {
4712         struct nfs_renamedata *data = task->tk_calldata;
4713         struct nfs_renameres *res = &data->res;
4714
4715         if (!nfs4_sequence_done(task, &res->seq_res))
4716                 return 0;
4717         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4718                 return 0;
4719
4720         if (task->tk_status == 0) {
4721                 if (new_dir != old_dir) {
4722                         /* Note: If we moved a directory, nlink will change */
4723                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4724                                         res->old_fattr->time_start,
4725                                         NFS_INO_INVALID_OTHER |
4726                                             NFS_INO_INVALID_DATA);
4727                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4728                                         res->new_fattr->time_start,
4729                                         NFS_INO_INVALID_OTHER |
4730                                             NFS_INO_INVALID_DATA);
4731                 } else
4732                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4733                                         res->old_fattr->time_start,
4734                                         NFS_INO_INVALID_DATA);
4735         }
4736         return 1;
4737 }
4738
4739 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4740 {
4741         struct nfs_server *server = NFS_SERVER(inode);
4742         __u32 bitmask[NFS4_BITMASK_SZ];
4743         struct nfs4_link_arg arg = {
4744                 .fh     = NFS_FH(inode),
4745                 .dir_fh = NFS_FH(dir),
4746                 .name   = name,
4747                 .bitmask = bitmask,
4748         };
4749         struct nfs4_link_res res = {
4750                 .server = server,
4751                 .label = NULL,
4752         };
4753         struct rpc_message msg = {
4754                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4755                 .rpc_argp = &arg,
4756                 .rpc_resp = &res,
4757         };
4758         int status = -ENOMEM;
4759
4760         res.fattr = nfs_alloc_fattr();
4761         if (res.fattr == NULL)
4762                 goto out;
4763
4764         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4765         if (IS_ERR(res.label)) {
4766                 status = PTR_ERR(res.label);
4767                 goto out;
4768         }
4769
4770         nfs4_inode_make_writeable(inode);
4771         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4772
4773         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4774         if (!status) {
4775                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4776                                        NFS_INO_INVALID_DATA);
4777                 status = nfs_post_op_update_inode(inode, res.fattr);
4778                 if (!status)
4779                         nfs_setsecurity(inode, res.fattr, res.label);
4780         }
4781
4782
4783         nfs4_label_free(res.label);
4784
4785 out:
4786         nfs_free_fattr(res.fattr);
4787         return status;
4788 }
4789
4790 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4791 {
4792         struct nfs4_exception exception = {
4793                 .interruptible = true,
4794         };
4795         int err;
4796         do {
4797                 err = nfs4_handle_exception(NFS_SERVER(inode),
4798                                 _nfs4_proc_link(inode, dir, name),
4799                                 &exception);
4800         } while (exception.retry);
4801         return err;
4802 }
4803
4804 struct nfs4_createdata {
4805         struct rpc_message msg;
4806         struct nfs4_create_arg arg;
4807         struct nfs4_create_res res;
4808         struct nfs_fh fh;
4809         struct nfs_fattr fattr;
4810         struct nfs4_label *label;
4811 };
4812
4813 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4814                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4815 {
4816         struct nfs4_createdata *data;
4817
4818         data = kzalloc(sizeof(*data), GFP_KERNEL);
4819         if (data != NULL) {
4820                 struct nfs_server *server = NFS_SERVER(dir);
4821
4822                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4823                 if (IS_ERR(data->label))
4824                         goto out_free;
4825
4826                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4827                 data->msg.rpc_argp = &data->arg;
4828                 data->msg.rpc_resp = &data->res;
4829                 data->arg.dir_fh = NFS_FH(dir);
4830                 data->arg.server = server;
4831                 data->arg.name = name;
4832                 data->arg.attrs = sattr;
4833                 data->arg.ftype = ftype;
4834                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4835                 data->arg.umask = current_umask();
4836                 data->res.server = server;
4837                 data->res.fh = &data->fh;
4838                 data->res.fattr = &data->fattr;
4839                 data->res.label = data->label;
4840                 nfs_fattr_init(data->res.fattr);
4841         }
4842         return data;
4843 out_free:
4844         kfree(data);
4845         return NULL;
4846 }
4847
4848 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4849 {
4850         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4851                                     &data->arg.seq_args, &data->res.seq_res, 1);
4852         if (status == 0) {
4853                 spin_lock(&dir->i_lock);
4854                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4855                                 data->res.fattr->time_start,
4856                                 NFS_INO_INVALID_DATA);
4857                 /* Creating a directory bumps nlink in the parent */
4858                 if (data->arg.ftype == NF4DIR)
4859                         nfs4_inc_nlink_locked(dir);
4860                 spin_unlock(&dir->i_lock);
4861                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4862         }
4863         return status;
4864 }
4865
4866 static void nfs4_free_createdata(struct nfs4_createdata *data)
4867 {
4868         nfs4_label_free(data->label);
4869         kfree(data);
4870 }
4871
4872 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4873                 struct page *page, unsigned int len, struct iattr *sattr,
4874                 struct nfs4_label *label)
4875 {
4876         struct nfs4_createdata *data;
4877         int status = -ENAMETOOLONG;
4878
4879         if (len > NFS4_MAXPATHLEN)
4880                 goto out;
4881
4882         status = -ENOMEM;
4883         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4884         if (data == NULL)
4885                 goto out;
4886
4887         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4888         data->arg.u.symlink.pages = &page;
4889         data->arg.u.symlink.len = len;
4890         data->arg.label = label;
4891
4892         status = nfs4_do_create(dir, dentry, data);
4893
4894         nfs4_free_createdata(data);
4895 out:
4896         return status;
4897 }
4898
4899 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4900                 struct page *page, unsigned int len, struct iattr *sattr)
4901 {
4902         struct nfs4_exception exception = {
4903                 .interruptible = true,
4904         };
4905         struct nfs4_label l, *label = NULL;
4906         int err;
4907
4908         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4909
4910         do {
4911                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4912                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4913                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4914                                 &exception);
4915         } while (exception.retry);
4916
4917         nfs4_label_release_security(label);
4918         return err;
4919 }
4920
4921 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4922                 struct iattr *sattr, struct nfs4_label *label)
4923 {
4924         struct nfs4_createdata *data;
4925         int status = -ENOMEM;
4926
4927         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4928         if (data == NULL)
4929                 goto out;
4930
4931         data->arg.label = label;
4932         status = nfs4_do_create(dir, dentry, data);
4933
4934         nfs4_free_createdata(data);
4935 out:
4936         return status;
4937 }
4938
4939 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4940                 struct iattr *sattr)
4941 {
4942         struct nfs_server *server = NFS_SERVER(dir);
4943         struct nfs4_exception exception = {
4944                 .interruptible = true,
4945         };
4946         struct nfs4_label l, *label = NULL;
4947         int err;
4948
4949         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4950
4951         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4952                 sattr->ia_mode &= ~current_umask();
4953         do {
4954                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4955                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4956                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4957                                 &exception);
4958         } while (exception.retry);
4959         nfs4_label_release_security(label);
4960
4961         return err;
4962 }
4963
4964 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
4965                               struct nfs_readdir_res *nr_res)
4966 {
4967         struct inode            *dir = d_inode(nr_arg->dentry);
4968         struct nfs_server       *server = NFS_SERVER(dir);
4969         struct nfs4_readdir_arg args = {
4970                 .fh = NFS_FH(dir),
4971                 .pages = nr_arg->pages,
4972                 .pgbase = 0,
4973                 .count = nr_arg->page_len,
4974                 .plus = nr_arg->plus,
4975         };
4976         struct nfs4_readdir_res res;
4977         struct rpc_message msg = {
4978                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4979                 .rpc_argp = &args,
4980                 .rpc_resp = &res,
4981                 .rpc_cred = nr_arg->cred,
4982         };
4983         int                     status;
4984
4985         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
4986                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
4987         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
4988                 args.bitmask = server->attr_bitmask_nl;
4989         else
4990                 args.bitmask = server->attr_bitmask;
4991
4992         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
4993         res.pgbase = args.pgbase;
4994         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
4995                         &res.seq_res, 0);
4996         if (status >= 0) {
4997                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
4998                 status += args.pgbase;
4999         }
5000
5001         nfs_invalidate_atime(dir);
5002
5003         dprintk("%s: returns %d\n", __func__, status);
5004         return status;
5005 }
5006
5007 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5008                              struct nfs_readdir_res *res)
5009 {
5010         struct nfs4_exception exception = {
5011                 .interruptible = true,
5012         };
5013         int err;
5014         do {
5015                 err = _nfs4_proc_readdir(arg, res);
5016                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5017                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5018                                             err, &exception);
5019         } while (exception.retry);
5020         return err;
5021 }
5022
5023 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5024                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5025 {
5026         struct nfs4_createdata *data;
5027         int mode = sattr->ia_mode;
5028         int status = -ENOMEM;
5029
5030         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5031         if (data == NULL)
5032                 goto out;
5033
5034         if (S_ISFIFO(mode))
5035                 data->arg.ftype = NF4FIFO;
5036         else if (S_ISBLK(mode)) {
5037                 data->arg.ftype = NF4BLK;
5038                 data->arg.u.device.specdata1 = MAJOR(rdev);
5039                 data->arg.u.device.specdata2 = MINOR(rdev);
5040         }
5041         else if (S_ISCHR(mode)) {
5042                 data->arg.ftype = NF4CHR;
5043                 data->arg.u.device.specdata1 = MAJOR(rdev);
5044                 data->arg.u.device.specdata2 = MINOR(rdev);
5045         } else if (!S_ISSOCK(mode)) {
5046                 status = -EINVAL;
5047                 goto out_free;
5048         }
5049
5050         data->arg.label = label;
5051         status = nfs4_do_create(dir, dentry, data);
5052 out_free:
5053         nfs4_free_createdata(data);
5054 out:
5055         return status;
5056 }
5057
5058 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5059                 struct iattr *sattr, dev_t rdev)
5060 {
5061         struct nfs_server *server = NFS_SERVER(dir);
5062         struct nfs4_exception exception = {
5063                 .interruptible = true,
5064         };
5065         struct nfs4_label l, *label = NULL;
5066         int err;
5067
5068         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5069
5070         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5071                 sattr->ia_mode &= ~current_umask();
5072         do {
5073                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5074                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5075                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5076                                 &exception);
5077         } while (exception.retry);
5078
5079         nfs4_label_release_security(label);
5080
5081         return err;
5082 }
5083
5084 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5085                  struct nfs_fsstat *fsstat)
5086 {
5087         struct nfs4_statfs_arg args = {
5088                 .fh = fhandle,
5089                 .bitmask = server->attr_bitmask,
5090         };
5091         struct nfs4_statfs_res res = {
5092                 .fsstat = fsstat,
5093         };
5094         struct rpc_message msg = {
5095                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5096                 .rpc_argp = &args,
5097                 .rpc_resp = &res,
5098         };
5099
5100         nfs_fattr_init(fsstat->fattr);
5101         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5102 }
5103
5104 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5105 {
5106         struct nfs4_exception exception = {
5107                 .interruptible = true,
5108         };
5109         int err;
5110         do {
5111                 err = nfs4_handle_exception(server,
5112                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5113                                 &exception);
5114         } while (exception.retry);
5115         return err;
5116 }
5117
5118 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5119                 struct nfs_fsinfo *fsinfo)
5120 {
5121         struct nfs4_fsinfo_arg args = {
5122                 .fh = fhandle,
5123                 .bitmask = server->attr_bitmask,
5124         };
5125         struct nfs4_fsinfo_res res = {
5126                 .fsinfo = fsinfo,
5127         };
5128         struct rpc_message msg = {
5129                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5130                 .rpc_argp = &args,
5131                 .rpc_resp = &res,
5132         };
5133
5134         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5135 }
5136
5137 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5138 {
5139         struct nfs4_exception exception = {
5140                 .interruptible = true,
5141         };
5142         int err;
5143
5144         do {
5145                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5146                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5147                 if (err == 0) {
5148                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5149                         break;
5150                 }
5151                 err = nfs4_handle_exception(server, err, &exception);
5152         } while (exception.retry);
5153         return err;
5154 }
5155
5156 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5157 {
5158         int error;
5159
5160         nfs_fattr_init(fsinfo->fattr);
5161         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5162         if (error == 0) {
5163                 /* block layout checks this! */
5164                 server->pnfs_blksize = fsinfo->blksize;
5165                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5166         }
5167
5168         return error;
5169 }
5170
5171 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5172                 struct nfs_pathconf *pathconf)
5173 {
5174         struct nfs4_pathconf_arg args = {
5175                 .fh = fhandle,
5176                 .bitmask = server->attr_bitmask,
5177         };
5178         struct nfs4_pathconf_res res = {
5179                 .pathconf = pathconf,
5180         };
5181         struct rpc_message msg = {
5182                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5183                 .rpc_argp = &args,
5184                 .rpc_resp = &res,
5185         };
5186
5187         /* None of the pathconf attributes are mandatory to implement */
5188         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5189                 memset(pathconf, 0, sizeof(*pathconf));
5190                 return 0;
5191         }
5192
5193         nfs_fattr_init(pathconf->fattr);
5194         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5195 }
5196
5197 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5198                 struct nfs_pathconf *pathconf)
5199 {
5200         struct nfs4_exception exception = {
5201                 .interruptible = true,
5202         };
5203         int err;
5204
5205         do {
5206                 err = nfs4_handle_exception(server,
5207                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5208                                 &exception);
5209         } while (exception.retry);
5210         return err;
5211 }
5212
5213 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5214                 const struct nfs_open_context *ctx,
5215                 const struct nfs_lock_context *l_ctx,
5216                 fmode_t fmode)
5217 {
5218         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5219 }
5220 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5221
5222 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5223                 const struct nfs_open_context *ctx,
5224                 const struct nfs_lock_context *l_ctx,
5225                 fmode_t fmode)
5226 {
5227         nfs4_stateid _current_stateid;
5228
5229         /* If the current stateid represents a lost lock, then exit */
5230         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5231                 return true;
5232         return nfs4_stateid_match(stateid, &_current_stateid);
5233 }
5234
5235 static bool nfs4_error_stateid_expired(int err)
5236 {
5237         switch (err) {
5238         case -NFS4ERR_DELEG_REVOKED:
5239         case -NFS4ERR_ADMIN_REVOKED:
5240         case -NFS4ERR_BAD_STATEID:
5241         case -NFS4ERR_STALE_STATEID:
5242         case -NFS4ERR_OLD_STATEID:
5243         case -NFS4ERR_OPENMODE:
5244         case -NFS4ERR_EXPIRED:
5245                 return true;
5246         }
5247         return false;
5248 }
5249
5250 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5251 {
5252         struct nfs_server *server = NFS_SERVER(hdr->inode);
5253
5254         trace_nfs4_read(hdr, task->tk_status);
5255         if (task->tk_status < 0) {
5256                 struct nfs4_exception exception = {
5257                         .inode = hdr->inode,
5258                         .state = hdr->args.context->state,
5259                         .stateid = &hdr->args.stateid,
5260                 };
5261                 task->tk_status = nfs4_async_handle_exception(task,
5262                                 server, task->tk_status, &exception);
5263                 if (exception.retry) {
5264                         rpc_restart_call_prepare(task);
5265                         return -EAGAIN;
5266                 }
5267         }
5268
5269         if (task->tk_status > 0)
5270                 renew_lease(server, hdr->timestamp);
5271         return 0;
5272 }
5273
5274 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5275                 struct nfs_pgio_args *args)
5276 {
5277
5278         if (!nfs4_error_stateid_expired(task->tk_status) ||
5279                 nfs4_stateid_is_current(&args->stateid,
5280                                 args->context,
5281                                 args->lock_context,
5282                                 FMODE_READ))
5283                 return false;
5284         rpc_restart_call_prepare(task);
5285         return true;
5286 }
5287
5288 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5289                                          struct nfs_pgio_header *hdr)
5290 {
5291         struct nfs_server *server = NFS_SERVER(hdr->inode);
5292         struct rpc_message *msg = &task->tk_msg;
5293
5294         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5295             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5296                 server->caps &= ~NFS_CAP_READ_PLUS;
5297                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5298                 rpc_restart_call_prepare(task);
5299                 return true;
5300         }
5301         return false;
5302 }
5303
5304 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5305 {
5306         dprintk("--> %s\n", __func__);
5307
5308         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5309                 return -EAGAIN;
5310         if (nfs4_read_stateid_changed(task, &hdr->args))
5311                 return -EAGAIN;
5312         if (nfs4_read_plus_not_supported(task, hdr))
5313                 return -EAGAIN;
5314         if (task->tk_status > 0)
5315                 nfs_invalidate_atime(hdr->inode);
5316         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5317                                     nfs4_read_done_cb(task, hdr);
5318 }
5319
5320 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5321 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5322                                     struct rpc_message *msg)
5323 {
5324         /* Note: We don't use READ_PLUS with pNFS yet */
5325         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5326                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5327 }
5328 #else
5329 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5330                                     struct rpc_message *msg)
5331 {
5332 }
5333 #endif /* CONFIG_NFS_V4_2 */
5334
5335 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5336                                  struct rpc_message *msg)
5337 {
5338         hdr->timestamp   = jiffies;
5339         if (!hdr->pgio_done_cb)
5340                 hdr->pgio_done_cb = nfs4_read_done_cb;
5341         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5342         nfs42_read_plus_support(hdr, msg);
5343         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5344 }
5345
5346 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5347                                       struct nfs_pgio_header *hdr)
5348 {
5349         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5350                         &hdr->args.seq_args,
5351                         &hdr->res.seq_res,
5352                         task))
5353                 return 0;
5354         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5355                                 hdr->args.lock_context,
5356                                 hdr->rw_mode) == -EIO)
5357                 return -EIO;
5358         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5359                 return -EIO;
5360         return 0;
5361 }
5362
5363 static int nfs4_write_done_cb(struct rpc_task *task,
5364                               struct nfs_pgio_header *hdr)
5365 {
5366         struct inode *inode = hdr->inode;
5367
5368         trace_nfs4_write(hdr, task->tk_status);
5369         if (task->tk_status < 0) {
5370                 struct nfs4_exception exception = {
5371                         .inode = hdr->inode,
5372                         .state = hdr->args.context->state,
5373                         .stateid = &hdr->args.stateid,
5374                 };
5375                 task->tk_status = nfs4_async_handle_exception(task,
5376                                 NFS_SERVER(inode), task->tk_status,
5377                                 &exception);
5378                 if (exception.retry) {
5379                         rpc_restart_call_prepare(task);
5380                         return -EAGAIN;
5381                 }
5382         }
5383         if (task->tk_status >= 0) {
5384                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5385                 nfs_writeback_update_inode(hdr);
5386         }
5387         return 0;
5388 }
5389
5390 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5391                 struct nfs_pgio_args *args)
5392 {
5393
5394         if (!nfs4_error_stateid_expired(task->tk_status) ||
5395                 nfs4_stateid_is_current(&args->stateid,
5396                                 args->context,
5397                                 args->lock_context,
5398                                 FMODE_WRITE))
5399                 return false;
5400         rpc_restart_call_prepare(task);
5401         return true;
5402 }
5403
5404 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5405 {
5406         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5407                 return -EAGAIN;
5408         if (nfs4_write_stateid_changed(task, &hdr->args))
5409                 return -EAGAIN;
5410         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5411                 nfs4_write_done_cb(task, hdr);
5412 }
5413
5414 static
5415 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5416 {
5417         /* Don't request attributes for pNFS or O_DIRECT writes */
5418         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5419                 return false;
5420         /* Otherwise, request attributes if and only if we don't hold
5421          * a delegation
5422          */
5423         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5424 }
5425
5426 static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode,
5427                                 struct nfs_server *server,
5428                                 struct nfs4_label *label)
5429 {
5430
5431         unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5432
5433         if ((cache_validity & NFS_INO_INVALID_DATA) ||
5434                 (cache_validity & NFS_INO_REVAL_PAGECACHE) ||
5435                 (cache_validity & NFS_INO_REVAL_FORCED) ||
5436                 (cache_validity & NFS_INO_INVALID_OTHER))
5437                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
5438
5439         if (cache_validity & NFS_INO_INVALID_ATIME)
5440                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5441         if (cache_validity & NFS_INO_INVALID_ACCESS)
5442                 bitmask[0] |= FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER |
5443                                 FATTR4_WORD1_OWNER_GROUP;
5444         if (cache_validity & NFS_INO_INVALID_ACL)
5445                 bitmask[0] |= FATTR4_WORD0_ACL;
5446         if (cache_validity & NFS_INO_INVALID_LABEL)
5447                 bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5448         if (cache_validity & NFS_INO_INVALID_CTIME)
5449                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5450         if (cache_validity & NFS_INO_INVALID_MTIME)
5451                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5452         if (cache_validity & NFS_INO_INVALID_SIZE)
5453                 bitmask[0] |= FATTR4_WORD0_SIZE;
5454         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5455                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5456 }
5457
5458 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5459                                   struct rpc_message *msg,
5460                                   struct rpc_clnt **clnt)
5461 {
5462         struct nfs_server *server = NFS_SERVER(hdr->inode);
5463
5464         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5465                 hdr->args.bitmask = NULL;
5466                 hdr->res.fattr = NULL;
5467         } else {
5468                 hdr->args.bitmask = server->cache_consistency_bitmask;
5469                 nfs4_bitmask_adjust(hdr->args.bitmask, hdr->inode, server, NULL);
5470         }
5471
5472         if (!hdr->pgio_done_cb)
5473                 hdr->pgio_done_cb = nfs4_write_done_cb;
5474         hdr->res.server = server;
5475         hdr->timestamp   = jiffies;
5476
5477         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5478         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5479         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5480 }
5481
5482 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5483 {
5484         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5485                         &data->args.seq_args,
5486                         &data->res.seq_res,
5487                         task);
5488 }
5489
5490 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5491 {
5492         struct inode *inode = data->inode;
5493
5494         trace_nfs4_commit(data, task->tk_status);
5495         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5496                                     NULL, NULL) == -EAGAIN) {
5497                 rpc_restart_call_prepare(task);
5498                 return -EAGAIN;
5499         }
5500         return 0;
5501 }
5502
5503 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5504 {
5505         if (!nfs4_sequence_done(task, &data->res.seq_res))
5506                 return -EAGAIN;
5507         return data->commit_done_cb(task, data);
5508 }
5509
5510 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5511                                    struct rpc_clnt **clnt)
5512 {
5513         struct nfs_server *server = NFS_SERVER(data->inode);
5514
5515         if (data->commit_done_cb == NULL)
5516                 data->commit_done_cb = nfs4_commit_done_cb;
5517         data->res.server = server;
5518         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5519         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5520         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5521 }
5522
5523 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5524                                 struct nfs_commitres *res)
5525 {
5526         struct inode *dst_inode = file_inode(dst);
5527         struct nfs_server *server = NFS_SERVER(dst_inode);
5528         struct rpc_message msg = {
5529                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5530                 .rpc_argp = args,
5531                 .rpc_resp = res,
5532         };
5533
5534         args->fh = NFS_FH(dst_inode);
5535         return nfs4_call_sync(server->client, server, &msg,
5536                         &args->seq_args, &res->seq_res, 1);
5537 }
5538
5539 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5540 {
5541         struct nfs_commitargs args = {
5542                 .offset = offset,
5543                 .count = count,
5544         };
5545         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5546         struct nfs4_exception exception = { };
5547         int status;
5548
5549         do {
5550                 status = _nfs4_proc_commit(dst, &args, res);
5551                 status = nfs4_handle_exception(dst_server, status, &exception);
5552         } while (exception.retry);
5553
5554         return status;
5555 }
5556
5557 struct nfs4_renewdata {
5558         struct nfs_client       *client;
5559         unsigned long           timestamp;
5560 };
5561
5562 /*
5563  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5564  * standalone procedure for queueing an asynchronous RENEW.
5565  */
5566 static void nfs4_renew_release(void *calldata)
5567 {
5568         struct nfs4_renewdata *data = calldata;
5569         struct nfs_client *clp = data->client;
5570
5571         if (refcount_read(&clp->cl_count) > 1)
5572                 nfs4_schedule_state_renewal(clp);
5573         nfs_put_client(clp);
5574         kfree(data);
5575 }
5576
5577 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5578 {
5579         struct nfs4_renewdata *data = calldata;
5580         struct nfs_client *clp = data->client;
5581         unsigned long timestamp = data->timestamp;
5582
5583         trace_nfs4_renew_async(clp, task->tk_status);
5584         switch (task->tk_status) {
5585         case 0:
5586                 break;
5587         case -NFS4ERR_LEASE_MOVED:
5588                 nfs4_schedule_lease_moved_recovery(clp);
5589                 break;
5590         default:
5591                 /* Unless we're shutting down, schedule state recovery! */
5592                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5593                         return;
5594                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5595                         nfs4_schedule_lease_recovery(clp);
5596                         return;
5597                 }
5598                 nfs4_schedule_path_down_recovery(clp);
5599         }
5600         do_renew_lease(clp, timestamp);
5601 }
5602
5603 static const struct rpc_call_ops nfs4_renew_ops = {
5604         .rpc_call_done = nfs4_renew_done,
5605         .rpc_release = nfs4_renew_release,
5606 };
5607
5608 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5609 {
5610         struct rpc_message msg = {
5611                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5612                 .rpc_argp       = clp,
5613                 .rpc_cred       = cred,
5614         };
5615         struct nfs4_renewdata *data;
5616
5617         if (renew_flags == 0)
5618                 return 0;
5619         if (!refcount_inc_not_zero(&clp->cl_count))
5620                 return -EIO;
5621         data = kmalloc(sizeof(*data), GFP_NOFS);
5622         if (data == NULL) {
5623                 nfs_put_client(clp);
5624                 return -ENOMEM;
5625         }
5626         data->client = clp;
5627         data->timestamp = jiffies;
5628         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5629                         &nfs4_renew_ops, data);
5630 }
5631
5632 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5633 {
5634         struct rpc_message msg = {
5635                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5636                 .rpc_argp       = clp,
5637                 .rpc_cred       = cred,
5638         };
5639         unsigned long now = jiffies;
5640         int status;
5641
5642         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5643         if (status < 0)
5644                 return status;
5645         do_renew_lease(clp, now);
5646         return 0;
5647 }
5648
5649 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5650 {
5651         return server->caps & NFS_CAP_ACLS;
5652 }
5653
5654 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5655  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5656  * the stack.
5657  */
5658 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5659
5660 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5661                 struct page **pages)
5662 {
5663         struct page *newpage, **spages;
5664         int rc = 0;
5665         size_t len;
5666         spages = pages;
5667
5668         do {
5669                 len = min_t(size_t, PAGE_SIZE, buflen);
5670                 newpage = alloc_page(GFP_KERNEL);
5671
5672                 if (newpage == NULL)
5673                         goto unwind;
5674                 memcpy(page_address(newpage), buf, len);
5675                 buf += len;
5676                 buflen -= len;
5677                 *pages++ = newpage;
5678                 rc++;
5679         } while (buflen != 0);
5680
5681         return rc;
5682
5683 unwind:
5684         for(; rc > 0; rc--)
5685                 __free_page(spages[rc-1]);
5686         return -ENOMEM;
5687 }
5688
5689 struct nfs4_cached_acl {
5690         int cached;
5691         size_t len;
5692         char data[];
5693 };
5694
5695 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5696 {
5697         struct nfs_inode *nfsi = NFS_I(inode);
5698
5699         spin_lock(&inode->i_lock);
5700         kfree(nfsi->nfs4_acl);
5701         nfsi->nfs4_acl = acl;
5702         spin_unlock(&inode->i_lock);
5703 }
5704
5705 static void nfs4_zap_acl_attr(struct inode *inode)
5706 {
5707         nfs4_set_cached_acl(inode, NULL);
5708 }
5709
5710 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5711 {
5712         struct nfs_inode *nfsi = NFS_I(inode);
5713         struct nfs4_cached_acl *acl;
5714         int ret = -ENOENT;
5715
5716         spin_lock(&inode->i_lock);
5717         acl = nfsi->nfs4_acl;
5718         if (acl == NULL)
5719                 goto out;
5720         if (buf == NULL) /* user is just asking for length */
5721                 goto out_len;
5722         if (acl->cached == 0)
5723                 goto out;
5724         ret = -ERANGE; /* see getxattr(2) man page */
5725         if (acl->len > buflen)
5726                 goto out;
5727         memcpy(buf, acl->data, acl->len);
5728 out_len:
5729         ret = acl->len;
5730 out:
5731         spin_unlock(&inode->i_lock);
5732         return ret;
5733 }
5734
5735 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5736 {
5737         struct nfs4_cached_acl *acl;
5738         size_t buflen = sizeof(*acl) + acl_len;
5739
5740         if (buflen <= PAGE_SIZE) {
5741                 acl = kmalloc(buflen, GFP_KERNEL);
5742                 if (acl == NULL)
5743                         goto out;
5744                 acl->cached = 1;
5745                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5746         } else {
5747                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5748                 if (acl == NULL)
5749                         goto out;
5750                 acl->cached = 0;
5751         }
5752         acl->len = acl_len;
5753 out:
5754         nfs4_set_cached_acl(inode, acl);
5755 }
5756
5757 /*
5758  * The getxattr API returns the required buffer length when called with a
5759  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5760  * the required buf.  On a NULL buf, we send a page of data to the server
5761  * guessing that the ACL request can be serviced by a page. If so, we cache
5762  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5763  * the cache. If not so, we throw away the page, and cache the required
5764  * length. The next getxattr call will then produce another round trip to
5765  * the server, this time with the input buf of the required size.
5766  */
5767 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5768 {
5769         struct page **pages;
5770         struct nfs_getaclargs args = {
5771                 .fh = NFS_FH(inode),
5772                 .acl_len = buflen,
5773         };
5774         struct nfs_getaclres res = {
5775                 .acl_len = buflen,
5776         };
5777         struct rpc_message msg = {
5778                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5779                 .rpc_argp = &args,
5780                 .rpc_resp = &res,
5781         };
5782         unsigned int npages;
5783         int ret = -ENOMEM, i;
5784         struct nfs_server *server = NFS_SERVER(inode);
5785
5786         if (buflen == 0)
5787                 buflen = server->rsize;
5788
5789         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5790         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5791         if (!pages)
5792                 return -ENOMEM;
5793
5794         args.acl_pages = pages;
5795
5796         for (i = 0; i < npages; i++) {
5797                 pages[i] = alloc_page(GFP_KERNEL);
5798                 if (!pages[i])
5799                         goto out_free;
5800         }
5801
5802         /* for decoding across pages */
5803         res.acl_scratch = alloc_page(GFP_KERNEL);
5804         if (!res.acl_scratch)
5805                 goto out_free;
5806
5807         args.acl_len = npages * PAGE_SIZE;
5808
5809         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5810                 __func__, buf, buflen, npages, args.acl_len);
5811         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5812                              &msg, &args.seq_args, &res.seq_res, 0);
5813         if (ret)
5814                 goto out_free;
5815
5816         /* Handle the case where the passed-in buffer is too short */
5817         if (res.acl_flags & NFS4_ACL_TRUNC) {
5818                 /* Did the user only issue a request for the acl length? */
5819                 if (buf == NULL)
5820                         goto out_ok;
5821                 ret = -ERANGE;
5822                 goto out_free;
5823         }
5824         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5825         if (buf) {
5826                 if (res.acl_len > buflen) {
5827                         ret = -ERANGE;
5828                         goto out_free;
5829                 }
5830                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5831         }
5832 out_ok:
5833         ret = res.acl_len;
5834 out_free:
5835         for (i = 0; i < npages; i++)
5836                 if (pages[i])
5837                         __free_page(pages[i]);
5838         if (res.acl_scratch)
5839                 __free_page(res.acl_scratch);
5840         kfree(pages);
5841         return ret;
5842 }
5843
5844 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5845 {
5846         struct nfs4_exception exception = {
5847                 .interruptible = true,
5848         };
5849         ssize_t ret;
5850         do {
5851                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5852                 trace_nfs4_get_acl(inode, ret);
5853                 if (ret >= 0)
5854                         break;
5855                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5856         } while (exception.retry);
5857         return ret;
5858 }
5859
5860 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5861 {
5862         struct nfs_server *server = NFS_SERVER(inode);
5863         int ret;
5864
5865         if (!nfs4_server_supports_acls(server))
5866                 return -EOPNOTSUPP;
5867         ret = nfs_revalidate_inode(server, inode);
5868         if (ret < 0)
5869                 return ret;
5870         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5871                 nfs_zap_acl_cache(inode);
5872         ret = nfs4_read_cached_acl(inode, buf, buflen);
5873         if (ret != -ENOENT)
5874                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5875                  * but no cached acl data, just the acl length */
5876                 return ret;
5877         return nfs4_get_acl_uncached(inode, buf, buflen);
5878 }
5879
5880 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5881 {
5882         struct nfs_server *server = NFS_SERVER(inode);
5883         struct page *pages[NFS4ACL_MAXPAGES];
5884         struct nfs_setaclargs arg = {
5885                 .fh             = NFS_FH(inode),
5886                 .acl_pages      = pages,
5887                 .acl_len        = buflen,
5888         };
5889         struct nfs_setaclres res;
5890         struct rpc_message msg = {
5891                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5892                 .rpc_argp       = &arg,
5893                 .rpc_resp       = &res,
5894         };
5895         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5896         int ret, i;
5897
5898         if (!nfs4_server_supports_acls(server))
5899                 return -EOPNOTSUPP;
5900         if (npages > ARRAY_SIZE(pages))
5901                 return -ERANGE;
5902         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5903         if (i < 0)
5904                 return i;
5905         nfs4_inode_make_writeable(inode);
5906         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5907
5908         /*
5909          * Free each page after tx, so the only ref left is
5910          * held by the network stack
5911          */
5912         for (; i > 0; i--)
5913                 put_page(pages[i-1]);
5914
5915         /*
5916          * Acl update can result in inode attribute update.
5917          * so mark the attribute cache invalid.
5918          */
5919         spin_lock(&inode->i_lock);
5920         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5921                 | NFS_INO_INVALID_CTIME
5922                 | NFS_INO_REVAL_FORCED;
5923         spin_unlock(&inode->i_lock);
5924         nfs_access_zap_cache(inode);
5925         nfs_zap_acl_cache(inode);
5926         return ret;
5927 }
5928
5929 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5930 {
5931         struct nfs4_exception exception = { };
5932         int err;
5933         do {
5934                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5935                 trace_nfs4_set_acl(inode, err);
5936                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5937                                 &exception);
5938         } while (exception.retry);
5939         return err;
5940 }
5941
5942 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5943 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5944                                         size_t buflen)
5945 {
5946         struct nfs_server *server = NFS_SERVER(inode);
5947         struct nfs_fattr fattr;
5948         struct nfs4_label label = {0, 0, buflen, buf};
5949
5950         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5951         struct nfs4_getattr_arg arg = {
5952                 .fh             = NFS_FH(inode),
5953                 .bitmask        = bitmask,
5954         };
5955         struct nfs4_getattr_res res = {
5956                 .fattr          = &fattr,
5957                 .label          = &label,
5958                 .server         = server,
5959         };
5960         struct rpc_message msg = {
5961                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5962                 .rpc_argp       = &arg,
5963                 .rpc_resp       = &res,
5964         };
5965         int ret;
5966
5967         nfs_fattr_init(&fattr);
5968
5969         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5970         if (ret)
5971                 return ret;
5972         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5973                 return -ENOENT;
5974         return 0;
5975 }
5976
5977 static int nfs4_get_security_label(struct inode *inode, void *buf,
5978                                         size_t buflen)
5979 {
5980         struct nfs4_exception exception = {
5981                 .interruptible = true,
5982         };
5983         int err;
5984
5985         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5986                 return -EOPNOTSUPP;
5987
5988         do {
5989                 err = _nfs4_get_security_label(inode, buf, buflen);
5990                 trace_nfs4_get_security_label(inode, err);
5991                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5992                                 &exception);
5993         } while (exception.retry);
5994         return err;
5995 }
5996
5997 static int _nfs4_do_set_security_label(struct inode *inode,
5998                 struct nfs4_label *ilabel,
5999                 struct nfs_fattr *fattr,
6000                 struct nfs4_label *olabel)
6001 {
6002
6003         struct iattr sattr = {0};
6004         struct nfs_server *server = NFS_SERVER(inode);
6005         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6006         struct nfs_setattrargs arg = {
6007                 .fh             = NFS_FH(inode),
6008                 .iap            = &sattr,
6009                 .server         = server,
6010                 .bitmask        = bitmask,
6011                 .label          = ilabel,
6012         };
6013         struct nfs_setattrres res = {
6014                 .fattr          = fattr,
6015                 .label          = olabel,
6016                 .server         = server,
6017         };
6018         struct rpc_message msg = {
6019                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6020                 .rpc_argp       = &arg,
6021                 .rpc_resp       = &res,
6022         };
6023         int status;
6024
6025         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6026
6027         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6028         if (status)
6029                 dprintk("%s failed: %d\n", __func__, status);
6030
6031         return status;
6032 }
6033
6034 static int nfs4_do_set_security_label(struct inode *inode,
6035                 struct nfs4_label *ilabel,
6036                 struct nfs_fattr *fattr,
6037                 struct nfs4_label *olabel)
6038 {
6039         struct nfs4_exception exception = { };
6040         int err;
6041
6042         do {
6043                 err = _nfs4_do_set_security_label(inode, ilabel,
6044                                 fattr, olabel);
6045                 trace_nfs4_set_security_label(inode, err);
6046                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6047                                 &exception);
6048         } while (exception.retry);
6049         return err;
6050 }
6051
6052 static int
6053 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6054 {
6055         struct nfs4_label ilabel, *olabel = NULL;
6056         struct nfs_fattr fattr;
6057         int status;
6058
6059         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6060                 return -EOPNOTSUPP;
6061
6062         nfs_fattr_init(&fattr);
6063
6064         ilabel.pi = 0;
6065         ilabel.lfs = 0;
6066         ilabel.label = (char *)buf;
6067         ilabel.len = buflen;
6068
6069         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
6070         if (IS_ERR(olabel)) {
6071                 status = -PTR_ERR(olabel);
6072                 goto out;
6073         }
6074
6075         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
6076         if (status == 0)
6077                 nfs_setsecurity(inode, &fattr, olabel);
6078
6079         nfs4_label_free(olabel);
6080 out:
6081         return status;
6082 }
6083 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6084
6085
6086 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6087                                     nfs4_verifier *bootverf)
6088 {
6089         __be32 verf[2];
6090
6091         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6092                 /* An impossible timestamp guarantees this value
6093                  * will never match a generated boot time. */
6094                 verf[0] = cpu_to_be32(U32_MAX);
6095                 verf[1] = cpu_to_be32(U32_MAX);
6096         } else {
6097                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6098                 u64 ns = ktime_to_ns(nn->boot_time);
6099
6100                 verf[0] = cpu_to_be32(ns >> 32);
6101                 verf[1] = cpu_to_be32(ns);
6102         }
6103         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6104 }
6105
6106 static size_t
6107 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6108 {
6109         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6110         struct nfs_netns_client *nn_clp = nn->nfs_client;
6111         const char *id;
6112
6113         buf[0] = '\0';
6114
6115         if (nn_clp) {
6116                 rcu_read_lock();
6117                 id = rcu_dereference(nn_clp->identifier);
6118                 if (id)
6119                         strscpy(buf, id, buflen);
6120                 rcu_read_unlock();
6121         }
6122
6123         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6124                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6125
6126         return strlen(buf);
6127 }
6128
6129 static int
6130 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6131 {
6132         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6133         size_t buflen;
6134         size_t len;
6135         char *str;
6136
6137         if (clp->cl_owner_id != NULL)
6138                 return 0;
6139
6140         rcu_read_lock();
6141         len = 14 +
6142                 strlen(clp->cl_rpcclient->cl_nodename) +
6143                 1 +
6144                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6145                 1;
6146         rcu_read_unlock();
6147
6148         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6149         if (buflen)
6150                 len += buflen + 1;
6151
6152         if (len > NFS4_OPAQUE_LIMIT + 1)
6153                 return -EINVAL;
6154
6155         /*
6156          * Since this string is allocated at mount time, and held until the
6157          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6158          * about a memory-reclaim deadlock.
6159          */
6160         str = kmalloc(len, GFP_KERNEL);
6161         if (!str)
6162                 return -ENOMEM;
6163
6164         rcu_read_lock();
6165         if (buflen)
6166                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6167                           clp->cl_rpcclient->cl_nodename, buf,
6168                           rpc_peeraddr2str(clp->cl_rpcclient,
6169                                            RPC_DISPLAY_ADDR));
6170         else
6171                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6172                           clp->cl_rpcclient->cl_nodename,
6173                           rpc_peeraddr2str(clp->cl_rpcclient,
6174                                            RPC_DISPLAY_ADDR));
6175         rcu_read_unlock();
6176
6177         clp->cl_owner_id = str;
6178         return 0;
6179 }
6180
6181 static int
6182 nfs4_init_uniform_client_string(struct nfs_client *clp)
6183 {
6184         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6185         size_t buflen;
6186         size_t len;
6187         char *str;
6188
6189         if (clp->cl_owner_id != NULL)
6190                 return 0;
6191
6192         len = 10 + 10 + 1 + 10 + 1 +
6193                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6194
6195         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6196         if (buflen)
6197                 len += buflen + 1;
6198
6199         if (len > NFS4_OPAQUE_LIMIT + 1)
6200                 return -EINVAL;
6201
6202         /*
6203          * Since this string is allocated at mount time, and held until the
6204          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6205          * about a memory-reclaim deadlock.
6206          */
6207         str = kmalloc(len, GFP_KERNEL);
6208         if (!str)
6209                 return -ENOMEM;
6210
6211         if (buflen)
6212                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6213                           clp->rpc_ops->version, clp->cl_minorversion,
6214                           buf, clp->cl_rpcclient->cl_nodename);
6215         else
6216                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6217                           clp->rpc_ops->version, clp->cl_minorversion,
6218                           clp->cl_rpcclient->cl_nodename);
6219         clp->cl_owner_id = str;
6220         return 0;
6221 }
6222
6223 /*
6224  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6225  * services.  Advertise one based on the address family of the
6226  * clientaddr.
6227  */
6228 static unsigned int
6229 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6230 {
6231         if (strchr(clp->cl_ipaddr, ':') != NULL)
6232                 return scnprintf(buf, len, "tcp6");
6233         else
6234                 return scnprintf(buf, len, "tcp");
6235 }
6236
6237 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6238 {
6239         struct nfs4_setclientid *sc = calldata;
6240
6241         if (task->tk_status == 0)
6242                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6243 }
6244
6245 static const struct rpc_call_ops nfs4_setclientid_ops = {
6246         .rpc_call_done = nfs4_setclientid_done,
6247 };
6248
6249 /**
6250  * nfs4_proc_setclientid - Negotiate client ID
6251  * @clp: state data structure
6252  * @program: RPC program for NFSv4 callback service
6253  * @port: IP port number for NFS4 callback service
6254  * @cred: credential to use for this call
6255  * @res: where to place the result
6256  *
6257  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6258  */
6259 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6260                 unsigned short port, const struct cred *cred,
6261                 struct nfs4_setclientid_res *res)
6262 {
6263         nfs4_verifier sc_verifier;
6264         struct nfs4_setclientid setclientid = {
6265                 .sc_verifier = &sc_verifier,
6266                 .sc_prog = program,
6267                 .sc_clnt = clp,
6268         };
6269         struct rpc_message msg = {
6270                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6271                 .rpc_argp = &setclientid,
6272                 .rpc_resp = res,
6273                 .rpc_cred = cred,
6274         };
6275         struct rpc_task_setup task_setup_data = {
6276                 .rpc_client = clp->cl_rpcclient,
6277                 .rpc_message = &msg,
6278                 .callback_ops = &nfs4_setclientid_ops,
6279                 .callback_data = &setclientid,
6280                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6281         };
6282         unsigned long now = jiffies;
6283         int status;
6284
6285         /* nfs_client_id4 */
6286         nfs4_init_boot_verifier(clp, &sc_verifier);
6287
6288         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6289                 status = nfs4_init_uniform_client_string(clp);
6290         else
6291                 status = nfs4_init_nonuniform_client_string(clp);
6292
6293         if (status)
6294                 goto out;
6295
6296         /* cb_client4 */
6297         setclientid.sc_netid_len =
6298                                 nfs4_init_callback_netid(clp,
6299                                                 setclientid.sc_netid,
6300                                                 sizeof(setclientid.sc_netid));
6301         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6302                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6303                                 clp->cl_ipaddr, port >> 8, port & 255);
6304
6305         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6306                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6307                 clp->cl_owner_id);
6308
6309         status = nfs4_call_sync_custom(&task_setup_data);
6310         if (setclientid.sc_cred) {
6311                 kfree(clp->cl_acceptor);
6312                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6313                 put_rpccred(setclientid.sc_cred);
6314         }
6315
6316         if (status == 0)
6317                 do_renew_lease(clp, now);
6318 out:
6319         trace_nfs4_setclientid(clp, status);
6320         dprintk("NFS reply setclientid: %d\n", status);
6321         return status;
6322 }
6323
6324 /**
6325  * nfs4_proc_setclientid_confirm - Confirm client ID
6326  * @clp: state data structure
6327  * @arg: result of a previous SETCLIENTID
6328  * @cred: credential to use for this call
6329  *
6330  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6331  */
6332 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6333                 struct nfs4_setclientid_res *arg,
6334                 const struct cred *cred)
6335 {
6336         struct rpc_message msg = {
6337                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6338                 .rpc_argp = arg,
6339                 .rpc_cred = cred,
6340         };
6341         int status;
6342
6343         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6344                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6345                 clp->cl_clientid);
6346         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6347                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6348         trace_nfs4_setclientid_confirm(clp, status);
6349         dprintk("NFS reply setclientid_confirm: %d\n", status);
6350         return status;
6351 }
6352
6353 struct nfs4_delegreturndata {
6354         struct nfs4_delegreturnargs args;
6355         struct nfs4_delegreturnres res;
6356         struct nfs_fh fh;
6357         nfs4_stateid stateid;
6358         unsigned long timestamp;
6359         struct {
6360                 struct nfs4_layoutreturn_args arg;
6361                 struct nfs4_layoutreturn_res res;
6362                 struct nfs4_xdr_opaque_data ld_private;
6363                 u32 roc_barrier;
6364                 bool roc;
6365         } lr;
6366         struct nfs_fattr fattr;
6367         int rpc_status;
6368         struct inode *inode;
6369 };
6370
6371 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6372 {
6373         struct nfs4_delegreturndata *data = calldata;
6374         struct nfs4_exception exception = {
6375                 .inode = data->inode,
6376                 .stateid = &data->stateid,
6377         };
6378
6379         if (!nfs4_sequence_done(task, &data->res.seq_res))
6380                 return;
6381
6382         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6383
6384         /* Handle Layoutreturn errors */
6385         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6386                           &data->res.lr_ret) == -EAGAIN)
6387                 goto out_restart;
6388
6389         switch (task->tk_status) {
6390         case 0:
6391                 renew_lease(data->res.server, data->timestamp);
6392                 break;
6393         case -NFS4ERR_ADMIN_REVOKED:
6394         case -NFS4ERR_DELEG_REVOKED:
6395         case -NFS4ERR_EXPIRED:
6396                 nfs4_free_revoked_stateid(data->res.server,
6397                                 data->args.stateid,
6398                                 task->tk_msg.rpc_cred);
6399                 fallthrough;
6400         case -NFS4ERR_BAD_STATEID:
6401         case -NFS4ERR_STALE_STATEID:
6402         case -ETIMEDOUT:
6403                 task->tk_status = 0;
6404                 break;
6405         case -NFS4ERR_OLD_STATEID:
6406                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6407                         nfs4_stateid_seqid_inc(&data->stateid);
6408                 if (data->args.bitmask) {
6409                         data->args.bitmask = NULL;
6410                         data->res.fattr = NULL;
6411                 }
6412                 goto out_restart;
6413         case -NFS4ERR_ACCESS:
6414                 if (data->args.bitmask) {
6415                         data->args.bitmask = NULL;
6416                         data->res.fattr = NULL;
6417                         goto out_restart;
6418                 }
6419                 fallthrough;
6420         default:
6421                 task->tk_status = nfs4_async_handle_exception(task,
6422                                 data->res.server, task->tk_status,
6423                                 &exception);
6424                 if (exception.retry)
6425                         goto out_restart;
6426         }
6427         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6428         data->rpc_status = task->tk_status;
6429         return;
6430 out_restart:
6431         task->tk_status = 0;
6432         rpc_restart_call_prepare(task);
6433 }
6434
6435 static void nfs4_delegreturn_release(void *calldata)
6436 {
6437         struct nfs4_delegreturndata *data = calldata;
6438         struct inode *inode = data->inode;
6439
6440         if (data->lr.roc)
6441                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6442                                  data->res.lr_ret);
6443         if (inode) {
6444                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6445                 nfs_iput_and_deactive(inode);
6446         }
6447         kfree(calldata);
6448 }
6449
6450 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6451 {
6452         struct nfs4_delegreturndata *d_data;
6453         struct pnfs_layout_hdr *lo;
6454
6455         d_data = (struct nfs4_delegreturndata *)data;
6456
6457         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6458                 nfs4_sequence_done(task, &d_data->res.seq_res);
6459                 return;
6460         }
6461
6462         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6463         if (lo && !pnfs_layout_is_valid(lo)) {
6464                 d_data->args.lr_args = NULL;
6465                 d_data->res.lr_res = NULL;
6466         }
6467
6468         nfs4_setup_sequence(d_data->res.server->nfs_client,
6469                         &d_data->args.seq_args,
6470                         &d_data->res.seq_res,
6471                         task);
6472 }
6473
6474 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6475         .rpc_call_prepare = nfs4_delegreturn_prepare,
6476         .rpc_call_done = nfs4_delegreturn_done,
6477         .rpc_release = nfs4_delegreturn_release,
6478 };
6479
6480 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6481 {
6482         struct nfs4_delegreturndata *data;
6483         struct nfs_server *server = NFS_SERVER(inode);
6484         struct rpc_task *task;
6485         struct rpc_message msg = {
6486                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6487                 .rpc_cred = cred,
6488         };
6489         struct rpc_task_setup task_setup_data = {
6490                 .rpc_client = server->client,
6491                 .rpc_message = &msg,
6492                 .callback_ops = &nfs4_delegreturn_ops,
6493                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6494         };
6495         int status = 0;
6496
6497         data = kzalloc(sizeof(*data), GFP_NOFS);
6498         if (data == NULL)
6499                 return -ENOMEM;
6500         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6501
6502         nfs4_state_protect(server->nfs_client,
6503                         NFS_SP4_MACH_CRED_CLEANUP,
6504                         &task_setup_data.rpc_client, &msg);
6505
6506         data->args.fhandle = &data->fh;
6507         data->args.stateid = &data->stateid;
6508         data->args.bitmask = server->cache_consistency_bitmask;
6509         nfs4_bitmask_adjust(data->args.bitmask, inode, server, NULL);
6510         nfs_copy_fh(&data->fh, NFS_FH(inode));
6511         nfs4_stateid_copy(&data->stateid, stateid);
6512         data->res.fattr = &data->fattr;
6513         data->res.server = server;
6514         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6515         data->lr.arg.ld_private = &data->lr.ld_private;
6516         nfs_fattr_init(data->res.fattr);
6517         data->timestamp = jiffies;
6518         data->rpc_status = 0;
6519         data->inode = nfs_igrab_and_active(inode);
6520         if (data->inode || issync) {
6521                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6522                                         cred);
6523                 if (data->lr.roc) {
6524                         data->args.lr_args = &data->lr.arg;
6525                         data->res.lr_res = &data->lr.res;
6526                 }
6527         }
6528
6529         task_setup_data.callback_data = data;
6530         msg.rpc_argp = &data->args;
6531         msg.rpc_resp = &data->res;
6532         task = rpc_run_task(&task_setup_data);
6533         if (IS_ERR(task))
6534                 return PTR_ERR(task);
6535         if (!issync)
6536                 goto out;
6537         status = rpc_wait_for_completion_task(task);
6538         if (status != 0)
6539                 goto out;
6540         status = data->rpc_status;
6541 out:
6542         rpc_put_task(task);
6543         return status;
6544 }
6545
6546 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6547 {
6548         struct nfs_server *server = NFS_SERVER(inode);
6549         struct nfs4_exception exception = { };
6550         int err;
6551         do {
6552                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6553                 trace_nfs4_delegreturn(inode, stateid, err);
6554                 switch (err) {
6555                         case -NFS4ERR_STALE_STATEID:
6556                         case -NFS4ERR_EXPIRED:
6557                         case 0:
6558                                 return 0;
6559                 }
6560                 err = nfs4_handle_exception(server, err, &exception);
6561         } while (exception.retry);
6562         return err;
6563 }
6564
6565 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6566 {
6567         struct inode *inode = state->inode;
6568         struct nfs_server *server = NFS_SERVER(inode);
6569         struct nfs_client *clp = server->nfs_client;
6570         struct nfs_lockt_args arg = {
6571                 .fh = NFS_FH(inode),
6572                 .fl = request,
6573         };
6574         struct nfs_lockt_res res = {
6575                 .denied = request,
6576         };
6577         struct rpc_message msg = {
6578                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6579                 .rpc_argp       = &arg,
6580                 .rpc_resp       = &res,
6581                 .rpc_cred       = state->owner->so_cred,
6582         };
6583         struct nfs4_lock_state *lsp;
6584         int status;
6585
6586         arg.lock_owner.clientid = clp->cl_clientid;
6587         status = nfs4_set_lock_state(state, request);
6588         if (status != 0)
6589                 goto out;
6590         lsp = request->fl_u.nfs4_fl.owner;
6591         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6592         arg.lock_owner.s_dev = server->s_dev;
6593         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6594         switch (status) {
6595                 case 0:
6596                         request->fl_type = F_UNLCK;
6597                         break;
6598                 case -NFS4ERR_DENIED:
6599                         status = 0;
6600         }
6601         request->fl_ops->fl_release_private(request);
6602         request->fl_ops = NULL;
6603 out:
6604         return status;
6605 }
6606
6607 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6608 {
6609         struct nfs4_exception exception = {
6610                 .interruptible = true,
6611         };
6612         int err;
6613
6614         do {
6615                 err = _nfs4_proc_getlk(state, cmd, request);
6616                 trace_nfs4_get_lock(request, state, cmd, err);
6617                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6618                                 &exception);
6619         } while (exception.retry);
6620         return err;
6621 }
6622
6623 /*
6624  * Update the seqid of a lock stateid after receiving
6625  * NFS4ERR_OLD_STATEID
6626  */
6627 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6628                 struct nfs4_lock_state *lsp)
6629 {
6630         struct nfs4_state *state = lsp->ls_state;
6631         bool ret = false;
6632
6633         spin_lock(&state->state_lock);
6634         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6635                 goto out;
6636         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6637                 nfs4_stateid_seqid_inc(dst);
6638         else
6639                 dst->seqid = lsp->ls_stateid.seqid;
6640         ret = true;
6641 out:
6642         spin_unlock(&state->state_lock);
6643         return ret;
6644 }
6645
6646 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6647                 struct nfs4_lock_state *lsp)
6648 {
6649         struct nfs4_state *state = lsp->ls_state;
6650         bool ret;
6651
6652         spin_lock(&state->state_lock);
6653         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6654         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6655         spin_unlock(&state->state_lock);
6656         return ret;
6657 }
6658
6659 struct nfs4_unlockdata {
6660         struct nfs_locku_args arg;
6661         struct nfs_locku_res res;
6662         struct nfs4_lock_state *lsp;
6663         struct nfs_open_context *ctx;
6664         struct nfs_lock_context *l_ctx;
6665         struct file_lock fl;
6666         struct nfs_server *server;
6667         unsigned long timestamp;
6668 };
6669
6670 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6671                 struct nfs_open_context *ctx,
6672                 struct nfs4_lock_state *lsp,
6673                 struct nfs_seqid *seqid)
6674 {
6675         struct nfs4_unlockdata *p;
6676         struct nfs4_state *state = lsp->ls_state;
6677         struct inode *inode = state->inode;
6678
6679         p = kzalloc(sizeof(*p), GFP_NOFS);
6680         if (p == NULL)
6681                 return NULL;
6682         p->arg.fh = NFS_FH(inode);
6683         p->arg.fl = &p->fl;
6684         p->arg.seqid = seqid;
6685         p->res.seqid = seqid;
6686         p->lsp = lsp;
6687         /* Ensure we don't close file until we're done freeing locks! */
6688         p->ctx = get_nfs_open_context(ctx);
6689         p->l_ctx = nfs_get_lock_context(ctx);
6690         locks_init_lock(&p->fl);
6691         locks_copy_lock(&p->fl, fl);
6692         p->server = NFS_SERVER(inode);
6693         spin_lock(&state->state_lock);
6694         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6695         spin_unlock(&state->state_lock);
6696         return p;
6697 }
6698
6699 static void nfs4_locku_release_calldata(void *data)
6700 {
6701         struct nfs4_unlockdata *calldata = data;
6702         nfs_free_seqid(calldata->arg.seqid);
6703         nfs4_put_lock_state(calldata->lsp);
6704         nfs_put_lock_context(calldata->l_ctx);
6705         put_nfs_open_context(calldata->ctx);
6706         kfree(calldata);
6707 }
6708
6709 static void nfs4_locku_done(struct rpc_task *task, void *data)
6710 {
6711         struct nfs4_unlockdata *calldata = data;
6712         struct nfs4_exception exception = {
6713                 .inode = calldata->lsp->ls_state->inode,
6714                 .stateid = &calldata->arg.stateid,
6715         };
6716
6717         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6718                 return;
6719         switch (task->tk_status) {
6720                 case 0:
6721                         renew_lease(calldata->server, calldata->timestamp);
6722                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6723                         if (nfs4_update_lock_stateid(calldata->lsp,
6724                                         &calldata->res.stateid))
6725                                 break;
6726                         fallthrough;
6727                 case -NFS4ERR_ADMIN_REVOKED:
6728                 case -NFS4ERR_EXPIRED:
6729                         nfs4_free_revoked_stateid(calldata->server,
6730                                         &calldata->arg.stateid,
6731                                         task->tk_msg.rpc_cred);
6732                         fallthrough;
6733                 case -NFS4ERR_BAD_STATEID:
6734                 case -NFS4ERR_STALE_STATEID:
6735                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6736                                                 calldata->lsp))
6737                                 rpc_restart_call_prepare(task);
6738                         break;
6739                 case -NFS4ERR_OLD_STATEID:
6740                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6741                                                 calldata->lsp))
6742                                 rpc_restart_call_prepare(task);
6743                         break;
6744                 default:
6745                         task->tk_status = nfs4_async_handle_exception(task,
6746                                         calldata->server, task->tk_status,
6747                                         &exception);
6748                         if (exception.retry)
6749                                 rpc_restart_call_prepare(task);
6750         }
6751         nfs_release_seqid(calldata->arg.seqid);
6752 }
6753
6754 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6755 {
6756         struct nfs4_unlockdata *calldata = data;
6757
6758         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6759                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6760                 return;
6761
6762         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6763                 goto out_wait;
6764         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6765                 /* Note: exit _without_ running nfs4_locku_done */
6766                 goto out_no_action;
6767         }
6768         calldata->timestamp = jiffies;
6769         if (nfs4_setup_sequence(calldata->server->nfs_client,
6770                                 &calldata->arg.seq_args,
6771                                 &calldata->res.seq_res,
6772                                 task) != 0)
6773                 nfs_release_seqid(calldata->arg.seqid);
6774         return;
6775 out_no_action:
6776         task->tk_action = NULL;
6777 out_wait:
6778         nfs4_sequence_done(task, &calldata->res.seq_res);
6779 }
6780
6781 static const struct rpc_call_ops nfs4_locku_ops = {
6782         .rpc_call_prepare = nfs4_locku_prepare,
6783         .rpc_call_done = nfs4_locku_done,
6784         .rpc_release = nfs4_locku_release_calldata,
6785 };
6786
6787 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6788                 struct nfs_open_context *ctx,
6789                 struct nfs4_lock_state *lsp,
6790                 struct nfs_seqid *seqid)
6791 {
6792         struct nfs4_unlockdata *data;
6793         struct rpc_message msg = {
6794                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6795                 .rpc_cred = ctx->cred,
6796         };
6797         struct rpc_task_setup task_setup_data = {
6798                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6799                 .rpc_message = &msg,
6800                 .callback_ops = &nfs4_locku_ops,
6801                 .workqueue = nfsiod_workqueue,
6802                 .flags = RPC_TASK_ASYNC,
6803         };
6804
6805         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6806                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6807
6808         /* Ensure this is an unlock - when canceling a lock, the
6809          * canceled lock is passed in, and it won't be an unlock.
6810          */
6811         fl->fl_type = F_UNLCK;
6812         if (fl->fl_flags & FL_CLOSE)
6813                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6814
6815         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6816         if (data == NULL) {
6817                 nfs_free_seqid(seqid);
6818                 return ERR_PTR(-ENOMEM);
6819         }
6820
6821         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6822         msg.rpc_argp = &data->arg;
6823         msg.rpc_resp = &data->res;
6824         task_setup_data.callback_data = data;
6825         return rpc_run_task(&task_setup_data);
6826 }
6827
6828 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6829 {
6830         struct inode *inode = state->inode;
6831         struct nfs4_state_owner *sp = state->owner;
6832         struct nfs_inode *nfsi = NFS_I(inode);
6833         struct nfs_seqid *seqid;
6834         struct nfs4_lock_state *lsp;
6835         struct rpc_task *task;
6836         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6837         int status = 0;
6838         unsigned char fl_flags = request->fl_flags;
6839
6840         status = nfs4_set_lock_state(state, request);
6841         /* Unlock _before_ we do the RPC call */
6842         request->fl_flags |= FL_EXISTS;
6843         /* Exclude nfs_delegation_claim_locks() */
6844         mutex_lock(&sp->so_delegreturn_mutex);
6845         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6846         down_read(&nfsi->rwsem);
6847         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6848                 up_read(&nfsi->rwsem);
6849                 mutex_unlock(&sp->so_delegreturn_mutex);
6850                 goto out;
6851         }
6852         up_read(&nfsi->rwsem);
6853         mutex_unlock(&sp->so_delegreturn_mutex);
6854         if (status != 0)
6855                 goto out;
6856         /* Is this a delegated lock? */
6857         lsp = request->fl_u.nfs4_fl.owner;
6858         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6859                 goto out;
6860         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6861         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6862         status = -ENOMEM;
6863         if (IS_ERR(seqid))
6864                 goto out;
6865         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6866         status = PTR_ERR(task);
6867         if (IS_ERR(task))
6868                 goto out;
6869         status = rpc_wait_for_completion_task(task);
6870         rpc_put_task(task);
6871 out:
6872         request->fl_flags = fl_flags;
6873         trace_nfs4_unlock(request, state, F_SETLK, status);
6874         return status;
6875 }
6876
6877 struct nfs4_lockdata {
6878         struct nfs_lock_args arg;
6879         struct nfs_lock_res res;
6880         struct nfs4_lock_state *lsp;
6881         struct nfs_open_context *ctx;
6882         struct file_lock fl;
6883         unsigned long timestamp;
6884         int rpc_status;
6885         int cancelled;
6886         struct nfs_server *server;
6887 };
6888
6889 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6890                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6891                 gfp_t gfp_mask)
6892 {
6893         struct nfs4_lockdata *p;
6894         struct inode *inode = lsp->ls_state->inode;
6895         struct nfs_server *server = NFS_SERVER(inode);
6896         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6897
6898         p = kzalloc(sizeof(*p), gfp_mask);
6899         if (p == NULL)
6900                 return NULL;
6901
6902         p->arg.fh = NFS_FH(inode);
6903         p->arg.fl = &p->fl;
6904         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6905         if (IS_ERR(p->arg.open_seqid))
6906                 goto out_free;
6907         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6908         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6909         if (IS_ERR(p->arg.lock_seqid))
6910                 goto out_free_seqid;
6911         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6912         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6913         p->arg.lock_owner.s_dev = server->s_dev;
6914         p->res.lock_seqid = p->arg.lock_seqid;
6915         p->lsp = lsp;
6916         p->server = server;
6917         p->ctx = get_nfs_open_context(ctx);
6918         locks_init_lock(&p->fl);
6919         locks_copy_lock(&p->fl, fl);
6920         return p;
6921 out_free_seqid:
6922         nfs_free_seqid(p->arg.open_seqid);
6923 out_free:
6924         kfree(p);
6925         return NULL;
6926 }
6927
6928 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6929 {
6930         struct nfs4_lockdata *data = calldata;
6931         struct nfs4_state *state = data->lsp->ls_state;
6932
6933         dprintk("%s: begin!\n", __func__);
6934         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6935                 goto out_wait;
6936         /* Do we need to do an open_to_lock_owner? */
6937         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6938                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6939                         goto out_release_lock_seqid;
6940                 }
6941                 nfs4_stateid_copy(&data->arg.open_stateid,
6942                                 &state->open_stateid);
6943                 data->arg.new_lock_owner = 1;
6944                 data->res.open_seqid = data->arg.open_seqid;
6945         } else {
6946                 data->arg.new_lock_owner = 0;
6947                 nfs4_stateid_copy(&data->arg.lock_stateid,
6948                                 &data->lsp->ls_stateid);
6949         }
6950         if (!nfs4_valid_open_stateid(state)) {
6951                 data->rpc_status = -EBADF;
6952                 task->tk_action = NULL;
6953                 goto out_release_open_seqid;
6954         }
6955         data->timestamp = jiffies;
6956         if (nfs4_setup_sequence(data->server->nfs_client,
6957                                 &data->arg.seq_args,
6958                                 &data->res.seq_res,
6959                                 task) == 0)
6960                 return;
6961 out_release_open_seqid:
6962         nfs_release_seqid(data->arg.open_seqid);
6963 out_release_lock_seqid:
6964         nfs_release_seqid(data->arg.lock_seqid);
6965 out_wait:
6966         nfs4_sequence_done(task, &data->res.seq_res);
6967         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6968 }
6969
6970 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6971 {
6972         struct nfs4_lockdata *data = calldata;
6973         struct nfs4_lock_state *lsp = data->lsp;
6974
6975         dprintk("%s: begin!\n", __func__);
6976
6977         if (!nfs4_sequence_done(task, &data->res.seq_res))
6978                 return;
6979
6980         data->rpc_status = task->tk_status;
6981         switch (task->tk_status) {
6982         case 0:
6983                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6984                                 data->timestamp);
6985                 if (data->arg.new_lock && !data->cancelled) {
6986                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6987                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6988                                 goto out_restart;
6989                 }
6990                 if (data->arg.new_lock_owner != 0) {
6991                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6992                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6993                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6994                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6995                         goto out_restart;
6996                 break;
6997         case -NFS4ERR_BAD_STATEID:
6998         case -NFS4ERR_OLD_STATEID:
6999         case -NFS4ERR_STALE_STATEID:
7000         case -NFS4ERR_EXPIRED:
7001                 if (data->arg.new_lock_owner != 0) {
7002                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7003                                                 &lsp->ls_state->open_stateid))
7004                                 goto out_restart;
7005                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7006                                                 &lsp->ls_stateid))
7007                                 goto out_restart;
7008         }
7009 out_done:
7010         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
7011         return;
7012 out_restart:
7013         if (!data->cancelled)
7014                 rpc_restart_call_prepare(task);
7015         goto out_done;
7016 }
7017
7018 static void nfs4_lock_release(void *calldata)
7019 {
7020         struct nfs4_lockdata *data = calldata;
7021
7022         dprintk("%s: begin!\n", __func__);
7023         nfs_free_seqid(data->arg.open_seqid);
7024         if (data->cancelled && data->rpc_status == 0) {
7025                 struct rpc_task *task;
7026                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7027                                 data->arg.lock_seqid);
7028                 if (!IS_ERR(task))
7029                         rpc_put_task_async(task);
7030                 dprintk("%s: cancelling lock!\n", __func__);
7031         } else
7032                 nfs_free_seqid(data->arg.lock_seqid);
7033         nfs4_put_lock_state(data->lsp);
7034         put_nfs_open_context(data->ctx);
7035         kfree(data);
7036         dprintk("%s: done!\n", __func__);
7037 }
7038
7039 static const struct rpc_call_ops nfs4_lock_ops = {
7040         .rpc_call_prepare = nfs4_lock_prepare,
7041         .rpc_call_done = nfs4_lock_done,
7042         .rpc_release = nfs4_lock_release,
7043 };
7044
7045 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7046 {
7047         switch (error) {
7048         case -NFS4ERR_ADMIN_REVOKED:
7049         case -NFS4ERR_EXPIRED:
7050         case -NFS4ERR_BAD_STATEID:
7051                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7052                 if (new_lock_owner != 0 ||
7053                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7054                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7055                 break;
7056         case -NFS4ERR_STALE_STATEID:
7057                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7058                 nfs4_schedule_lease_recovery(server->nfs_client);
7059         }
7060 }
7061
7062 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7063 {
7064         struct nfs4_lockdata *data;
7065         struct rpc_task *task;
7066         struct rpc_message msg = {
7067                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7068                 .rpc_cred = state->owner->so_cred,
7069         };
7070         struct rpc_task_setup task_setup_data = {
7071                 .rpc_client = NFS_CLIENT(state->inode),
7072                 .rpc_message = &msg,
7073                 .callback_ops = &nfs4_lock_ops,
7074                 .workqueue = nfsiod_workqueue,
7075                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7076         };
7077         int ret;
7078
7079         dprintk("%s: begin!\n", __func__);
7080         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7081                         fl->fl_u.nfs4_fl.owner,
7082                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7083         if (data == NULL)
7084                 return -ENOMEM;
7085         if (IS_SETLKW(cmd))
7086                 data->arg.block = 1;
7087         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7088                                 recovery_type > NFS_LOCK_NEW);
7089         msg.rpc_argp = &data->arg;
7090         msg.rpc_resp = &data->res;
7091         task_setup_data.callback_data = data;
7092         if (recovery_type > NFS_LOCK_NEW) {
7093                 if (recovery_type == NFS_LOCK_RECLAIM)
7094                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7095         } else
7096                 data->arg.new_lock = 1;
7097         task = rpc_run_task(&task_setup_data);
7098         if (IS_ERR(task))
7099                 return PTR_ERR(task);
7100         ret = rpc_wait_for_completion_task(task);
7101         if (ret == 0) {
7102                 ret = data->rpc_status;
7103                 if (ret)
7104                         nfs4_handle_setlk_error(data->server, data->lsp,
7105                                         data->arg.new_lock_owner, ret);
7106         } else
7107                 data->cancelled = true;
7108         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7109         rpc_put_task(task);
7110         dprintk("%s: done, ret = %d!\n", __func__, ret);
7111         return ret;
7112 }
7113
7114 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7115 {
7116         struct nfs_server *server = NFS_SERVER(state->inode);
7117         struct nfs4_exception exception = {
7118                 .inode = state->inode,
7119         };
7120         int err;
7121
7122         do {
7123                 /* Cache the lock if possible... */
7124                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7125                         return 0;
7126                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7127                 if (err != -NFS4ERR_DELAY)
7128                         break;
7129                 nfs4_handle_exception(server, err, &exception);
7130         } while (exception.retry);
7131         return err;
7132 }
7133
7134 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7135 {
7136         struct nfs_server *server = NFS_SERVER(state->inode);
7137         struct nfs4_exception exception = {
7138                 .inode = state->inode,
7139         };
7140         int err;
7141
7142         err = nfs4_set_lock_state(state, request);
7143         if (err != 0)
7144                 return err;
7145         if (!recover_lost_locks) {
7146                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7147                 return 0;
7148         }
7149         do {
7150                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7151                         return 0;
7152                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7153                 switch (err) {
7154                 default:
7155                         goto out;
7156                 case -NFS4ERR_GRACE:
7157                 case -NFS4ERR_DELAY:
7158                         nfs4_handle_exception(server, err, &exception);
7159                         err = 0;
7160                 }
7161         } while (exception.retry);
7162 out:
7163         return err;
7164 }
7165
7166 #if defined(CONFIG_NFS_V4_1)
7167 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7168 {
7169         struct nfs4_lock_state *lsp;
7170         int status;
7171
7172         status = nfs4_set_lock_state(state, request);
7173         if (status != 0)
7174                 return status;
7175         lsp = request->fl_u.nfs4_fl.owner;
7176         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7177             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7178                 return 0;
7179         return nfs4_lock_expired(state, request);
7180 }
7181 #endif
7182
7183 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7184 {
7185         struct nfs_inode *nfsi = NFS_I(state->inode);
7186         struct nfs4_state_owner *sp = state->owner;
7187         unsigned char fl_flags = request->fl_flags;
7188         int status;
7189
7190         request->fl_flags |= FL_ACCESS;
7191         status = locks_lock_inode_wait(state->inode, request);
7192         if (status < 0)
7193                 goto out;
7194         mutex_lock(&sp->so_delegreturn_mutex);
7195         down_read(&nfsi->rwsem);
7196         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7197                 /* Yes: cache locks! */
7198                 /* ...but avoid races with delegation recall... */
7199                 request->fl_flags = fl_flags & ~FL_SLEEP;
7200                 status = locks_lock_inode_wait(state->inode, request);
7201                 up_read(&nfsi->rwsem);
7202                 mutex_unlock(&sp->so_delegreturn_mutex);
7203                 goto out;
7204         }
7205         up_read(&nfsi->rwsem);
7206         mutex_unlock(&sp->so_delegreturn_mutex);
7207         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7208 out:
7209         request->fl_flags = fl_flags;
7210         return status;
7211 }
7212
7213 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7214 {
7215         struct nfs4_exception exception = {
7216                 .state = state,
7217                 .inode = state->inode,
7218                 .interruptible = true,
7219         };
7220         int err;
7221
7222         do {
7223                 err = _nfs4_proc_setlk(state, cmd, request);
7224                 if (err == -NFS4ERR_DENIED)
7225                         err = -EAGAIN;
7226                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7227                                 err, &exception);
7228         } while (exception.retry);
7229         return err;
7230 }
7231
7232 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7233 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7234
7235 static int
7236 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7237                         struct file_lock *request)
7238 {
7239         int             status = -ERESTARTSYS;
7240         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7241
7242         while(!signalled()) {
7243                 status = nfs4_proc_setlk(state, cmd, request);
7244                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7245                         break;
7246                 freezable_schedule_timeout_interruptible(timeout);
7247                 timeout *= 2;
7248                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7249                 status = -ERESTARTSYS;
7250         }
7251         return status;
7252 }
7253
7254 #ifdef CONFIG_NFS_V4_1
7255 struct nfs4_lock_waiter {
7256         struct task_struct      *task;
7257         struct inode            *inode;
7258         struct nfs_lowner       *owner;
7259 };
7260
7261 static int
7262 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7263 {
7264         int ret;
7265         struct nfs4_lock_waiter *waiter = wait->private;
7266
7267         /* NULL key means to wake up everyone */
7268         if (key) {
7269                 struct cb_notify_lock_args      *cbnl = key;
7270                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7271                                                 *wowner = waiter->owner;
7272
7273                 /* Only wake if the callback was for the same owner. */
7274                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7275                         return 0;
7276
7277                 /* Make sure it's for the right inode */
7278                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7279                         return 0;
7280         }
7281
7282         /* override "private" so we can use default_wake_function */
7283         wait->private = waiter->task;
7284         ret = woken_wake_function(wait, mode, flags, key);
7285         if (ret)
7286                 list_del_init(&wait->entry);
7287         wait->private = waiter;
7288         return ret;
7289 }
7290
7291 static int
7292 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7293 {
7294         int status = -ERESTARTSYS;
7295         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7296         struct nfs_server *server = NFS_SERVER(state->inode);
7297         struct nfs_client *clp = server->nfs_client;
7298         wait_queue_head_t *q = &clp->cl_lock_waitq;
7299         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7300                                     .id = lsp->ls_seqid.owner_id,
7301                                     .s_dev = server->s_dev };
7302         struct nfs4_lock_waiter waiter = { .task  = current,
7303                                            .inode = state->inode,
7304                                            .owner = &owner};
7305         wait_queue_entry_t wait;
7306
7307         /* Don't bother with waitqueue if we don't expect a callback */
7308         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7309                 return nfs4_retry_setlk_simple(state, cmd, request);
7310
7311         init_wait(&wait);
7312         wait.private = &waiter;
7313         wait.func = nfs4_wake_lock_waiter;
7314
7315         while(!signalled()) {
7316                 add_wait_queue(q, &wait);
7317                 status = nfs4_proc_setlk(state, cmd, request);
7318                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7319                         finish_wait(q, &wait);
7320                         break;
7321                 }
7322
7323                 status = -ERESTARTSYS;
7324                 freezer_do_not_count();
7325                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7326                 freezer_count();
7327                 finish_wait(q, &wait);
7328         }
7329
7330         return status;
7331 }
7332 #else /* !CONFIG_NFS_V4_1 */
7333 static inline int
7334 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7335 {
7336         return nfs4_retry_setlk_simple(state, cmd, request);
7337 }
7338 #endif
7339
7340 static int
7341 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7342 {
7343         struct nfs_open_context *ctx;
7344         struct nfs4_state *state;
7345         int status;
7346
7347         /* verify open state */
7348         ctx = nfs_file_open_context(filp);
7349         state = ctx->state;
7350
7351         if (IS_GETLK(cmd)) {
7352                 if (state != NULL)
7353                         return nfs4_proc_getlk(state, F_GETLK, request);
7354                 return 0;
7355         }
7356
7357         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7358                 return -EINVAL;
7359
7360         if (request->fl_type == F_UNLCK) {
7361                 if (state != NULL)
7362                         return nfs4_proc_unlck(state, cmd, request);
7363                 return 0;
7364         }
7365
7366         if (state == NULL)
7367                 return -ENOLCK;
7368
7369         if ((request->fl_flags & FL_POSIX) &&
7370             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7371                 return -ENOLCK;
7372
7373         /*
7374          * Don't rely on the VFS having checked the file open mode,
7375          * since it won't do this for flock() locks.
7376          */
7377         switch (request->fl_type) {
7378         case F_RDLCK:
7379                 if (!(filp->f_mode & FMODE_READ))
7380                         return -EBADF;
7381                 break;
7382         case F_WRLCK:
7383                 if (!(filp->f_mode & FMODE_WRITE))
7384                         return -EBADF;
7385         }
7386
7387         status = nfs4_set_lock_state(state, request);
7388         if (status != 0)
7389                 return status;
7390
7391         return nfs4_retry_setlk(state, cmd, request);
7392 }
7393
7394 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7395 {
7396         struct nfs_server *server = NFS_SERVER(state->inode);
7397         int err;
7398
7399         err = nfs4_set_lock_state(state, fl);
7400         if (err != 0)
7401                 return err;
7402         do {
7403                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7404                 if (err != -NFS4ERR_DELAY)
7405                         break;
7406                 ssleep(1);
7407         } while (err == -NFS4ERR_DELAY);
7408         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7409 }
7410
7411 struct nfs_release_lockowner_data {
7412         struct nfs4_lock_state *lsp;
7413         struct nfs_server *server;
7414         struct nfs_release_lockowner_args args;
7415         struct nfs_release_lockowner_res res;
7416         unsigned long timestamp;
7417 };
7418
7419 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7420 {
7421         struct nfs_release_lockowner_data *data = calldata;
7422         struct nfs_server *server = data->server;
7423         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7424                            &data->res.seq_res, task);
7425         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7426         data->timestamp = jiffies;
7427 }
7428
7429 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7430 {
7431         struct nfs_release_lockowner_data *data = calldata;
7432         struct nfs_server *server = data->server;
7433
7434         nfs40_sequence_done(task, &data->res.seq_res);
7435
7436         switch (task->tk_status) {
7437         case 0:
7438                 renew_lease(server, data->timestamp);
7439                 break;
7440         case -NFS4ERR_STALE_CLIENTID:
7441         case -NFS4ERR_EXPIRED:
7442                 nfs4_schedule_lease_recovery(server->nfs_client);
7443                 break;
7444         case -NFS4ERR_LEASE_MOVED:
7445         case -NFS4ERR_DELAY:
7446                 if (nfs4_async_handle_error(task, server,
7447                                             NULL, NULL) == -EAGAIN)
7448                         rpc_restart_call_prepare(task);
7449         }
7450 }
7451
7452 static void nfs4_release_lockowner_release(void *calldata)
7453 {
7454         struct nfs_release_lockowner_data *data = calldata;
7455         nfs4_free_lock_state(data->server, data->lsp);
7456         kfree(calldata);
7457 }
7458
7459 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7460         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7461         .rpc_call_done = nfs4_release_lockowner_done,
7462         .rpc_release = nfs4_release_lockowner_release,
7463 };
7464
7465 static void
7466 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7467 {
7468         struct nfs_release_lockowner_data *data;
7469         struct rpc_message msg = {
7470                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7471         };
7472
7473         if (server->nfs_client->cl_mvops->minor_version != 0)
7474                 return;
7475
7476         data = kmalloc(sizeof(*data), GFP_NOFS);
7477         if (!data)
7478                 return;
7479         data->lsp = lsp;
7480         data->server = server;
7481         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7482         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7483         data->args.lock_owner.s_dev = server->s_dev;
7484
7485         msg.rpc_argp = &data->args;
7486         msg.rpc_resp = &data->res;
7487         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7488         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7489 }
7490
7491 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7492
7493 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7494                                    struct dentry *unused, struct inode *inode,
7495                                    const char *key, const void *buf,
7496                                    size_t buflen, int flags)
7497 {
7498         return nfs4_proc_set_acl(inode, buf, buflen);
7499 }
7500
7501 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7502                                    struct dentry *unused, struct inode *inode,
7503                                    const char *key, void *buf, size_t buflen)
7504 {
7505         return nfs4_proc_get_acl(inode, buf, buflen);
7506 }
7507
7508 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7509 {
7510         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7511 }
7512
7513 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7514
7515 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7516                                      struct dentry *unused, struct inode *inode,
7517                                      const char *key, const void *buf,
7518                                      size_t buflen, int flags)
7519 {
7520         if (security_ismaclabel(key))
7521                 return nfs4_set_security_label(inode, buf, buflen);
7522
7523         return -EOPNOTSUPP;
7524 }
7525
7526 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7527                                      struct dentry *unused, struct inode *inode,
7528                                      const char *key, void *buf, size_t buflen)
7529 {
7530         if (security_ismaclabel(key))
7531                 return nfs4_get_security_label(inode, buf, buflen);
7532         return -EOPNOTSUPP;
7533 }
7534
7535 static ssize_t
7536 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7537 {
7538         int len = 0;
7539
7540         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7541                 len = security_inode_listsecurity(inode, list, list_len);
7542                 if (len >= 0 && list_len && len > list_len)
7543                         return -ERANGE;
7544         }
7545         return len;
7546 }
7547
7548 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7549         .prefix = XATTR_SECURITY_PREFIX,
7550         .get    = nfs4_xattr_get_nfs4_label,
7551         .set    = nfs4_xattr_set_nfs4_label,
7552 };
7553
7554 #else
7555
7556 static ssize_t
7557 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7558 {
7559         return 0;
7560 }
7561
7562 #endif
7563
7564 #ifdef CONFIG_NFS_V4_2
7565 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7566                                     struct dentry *unused, struct inode *inode,
7567                                     const char *key, const void *buf,
7568                                     size_t buflen, int flags)
7569 {
7570         struct nfs_access_entry cache;
7571         int ret;
7572
7573         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7574                 return -EOPNOTSUPP;
7575
7576         /*
7577          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7578          * flags right now. Handling of xattr operations use the normal
7579          * file read/write permissions.
7580          *
7581          * Just in case the server has other ideas (which RFC 8276 allows),
7582          * do a cached access check for the XA* flags to possibly avoid
7583          * doing an RPC and getting EACCES back.
7584          */
7585         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7586                 if (!(cache.mask & NFS_ACCESS_XAWRITE))
7587                         return -EACCES;
7588         }
7589
7590         if (buf == NULL) {
7591                 ret = nfs42_proc_removexattr(inode, key);
7592                 if (!ret)
7593                         nfs4_xattr_cache_remove(inode, key);
7594         } else {
7595                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7596                 if (!ret)
7597                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7598         }
7599
7600         return ret;
7601 }
7602
7603 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7604                                     struct dentry *unused, struct inode *inode,
7605                                     const char *key, void *buf, size_t buflen)
7606 {
7607         struct nfs_access_entry cache;
7608         ssize_t ret;
7609
7610         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7611                 return -EOPNOTSUPP;
7612
7613         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7614                 if (!(cache.mask & NFS_ACCESS_XAREAD))
7615                         return -EACCES;
7616         }
7617
7618         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7619         if (ret)
7620                 return ret;
7621
7622         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7623         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7624                 return ret;
7625
7626         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7627
7628         return ret;
7629 }
7630
7631 static ssize_t
7632 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7633 {
7634         u64 cookie;
7635         bool eof;
7636         ssize_t ret, size;
7637         char *buf;
7638         size_t buflen;
7639         struct nfs_access_entry cache;
7640
7641         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7642                 return 0;
7643
7644         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7645                 if (!(cache.mask & NFS_ACCESS_XALIST))
7646                         return 0;
7647         }
7648
7649         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7650         if (ret)
7651                 return ret;
7652
7653         ret = nfs4_xattr_cache_list(inode, list, list_len);
7654         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7655                 return ret;
7656
7657         cookie = 0;
7658         eof = false;
7659         buflen = list_len ? list_len : XATTR_LIST_MAX;
7660         buf = list_len ? list : NULL;
7661         size = 0;
7662
7663         while (!eof) {
7664                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7665                     &cookie, &eof);
7666                 if (ret < 0)
7667                         return ret;
7668
7669                 if (list_len) {
7670                         buf += ret;
7671                         buflen -= ret;
7672                 }
7673                 size += ret;
7674         }
7675
7676         if (list_len)
7677                 nfs4_xattr_cache_set_list(inode, list, size);
7678
7679         return size;
7680 }
7681
7682 #else
7683
7684 static ssize_t
7685 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7686 {
7687         return 0;
7688 }
7689 #endif /* CONFIG_NFS_V4_2 */
7690
7691 /*
7692  * nfs_fhget will use either the mounted_on_fileid or the fileid
7693  */
7694 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7695 {
7696         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7697                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7698               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7699               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7700                 return;
7701
7702         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7703                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7704         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7705         fattr->nlink = 2;
7706 }
7707
7708 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7709                                    const struct qstr *name,
7710                                    struct nfs4_fs_locations *fs_locations,
7711                                    struct page *page)
7712 {
7713         struct nfs_server *server = NFS_SERVER(dir);
7714         u32 bitmask[3];
7715         struct nfs4_fs_locations_arg args = {
7716                 .dir_fh = NFS_FH(dir),
7717                 .name = name,
7718                 .page = page,
7719                 .bitmask = bitmask,
7720         };
7721         struct nfs4_fs_locations_res res = {
7722                 .fs_locations = fs_locations,
7723         };
7724         struct rpc_message msg = {
7725                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7726                 .rpc_argp = &args,
7727                 .rpc_resp = &res,
7728         };
7729         int status;
7730
7731         dprintk("%s: start\n", __func__);
7732
7733         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7734         bitmask[1] = nfs4_fattr_bitmap[1];
7735
7736         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7737          * is not supported */
7738         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7739                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7740         else
7741                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7742
7743         nfs_fattr_init(&fs_locations->fattr);
7744         fs_locations->server = server;
7745         fs_locations->nlocations = 0;
7746         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7747         dprintk("%s: returned status = %d\n", __func__, status);
7748         return status;
7749 }
7750
7751 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7752                            const struct qstr *name,
7753                            struct nfs4_fs_locations *fs_locations,
7754                            struct page *page)
7755 {
7756         struct nfs4_exception exception = {
7757                 .interruptible = true,
7758         };
7759         int err;
7760         do {
7761                 err = _nfs4_proc_fs_locations(client, dir, name,
7762                                 fs_locations, page);
7763                 trace_nfs4_get_fs_locations(dir, name, err);
7764                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7765                                 &exception);
7766         } while (exception.retry);
7767         return err;
7768 }
7769
7770 /*
7771  * This operation also signals the server that this client is
7772  * performing migration recovery.  The server can stop returning
7773  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7774  * appended to this compound to identify the client ID which is
7775  * performing recovery.
7776  */
7777 static int _nfs40_proc_get_locations(struct inode *inode,
7778                                      struct nfs4_fs_locations *locations,
7779                                      struct page *page, const struct cred *cred)
7780 {
7781         struct nfs_server *server = NFS_SERVER(inode);
7782         struct rpc_clnt *clnt = server->client;
7783         u32 bitmask[2] = {
7784                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7785         };
7786         struct nfs4_fs_locations_arg args = {
7787                 .clientid       = server->nfs_client->cl_clientid,
7788                 .fh             = NFS_FH(inode),
7789                 .page           = page,
7790                 .bitmask        = bitmask,
7791                 .migration      = 1,            /* skip LOOKUP */
7792                 .renew          = 1,            /* append RENEW */
7793         };
7794         struct nfs4_fs_locations_res res = {
7795                 .fs_locations   = locations,
7796                 .migration      = 1,
7797                 .renew          = 1,
7798         };
7799         struct rpc_message msg = {
7800                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7801                 .rpc_argp       = &args,
7802                 .rpc_resp       = &res,
7803                 .rpc_cred       = cred,
7804         };
7805         unsigned long now = jiffies;
7806         int status;
7807
7808         nfs_fattr_init(&locations->fattr);
7809         locations->server = server;
7810         locations->nlocations = 0;
7811
7812         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7813         status = nfs4_call_sync_sequence(clnt, server, &msg,
7814                                         &args.seq_args, &res.seq_res);
7815         if (status)
7816                 return status;
7817
7818         renew_lease(server, now);
7819         return 0;
7820 }
7821
7822 #ifdef CONFIG_NFS_V4_1
7823
7824 /*
7825  * This operation also signals the server that this client is
7826  * performing migration recovery.  The server can stop asserting
7827  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7828  * performing this operation is identified in the SEQUENCE
7829  * operation in this compound.
7830  *
7831  * When the client supports GETATTR(fs_locations_info), it can
7832  * be plumbed in here.
7833  */
7834 static int _nfs41_proc_get_locations(struct inode *inode,
7835                                      struct nfs4_fs_locations *locations,
7836                                      struct page *page, const struct cred *cred)
7837 {
7838         struct nfs_server *server = NFS_SERVER(inode);
7839         struct rpc_clnt *clnt = server->client;
7840         u32 bitmask[2] = {
7841                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7842         };
7843         struct nfs4_fs_locations_arg args = {
7844                 .fh             = NFS_FH(inode),
7845                 .page           = page,
7846                 .bitmask        = bitmask,
7847                 .migration      = 1,            /* skip LOOKUP */
7848         };
7849         struct nfs4_fs_locations_res res = {
7850                 .fs_locations   = locations,
7851                 .migration      = 1,
7852         };
7853         struct rpc_message msg = {
7854                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7855                 .rpc_argp       = &args,
7856                 .rpc_resp       = &res,
7857                 .rpc_cred       = cred,
7858         };
7859         int status;
7860
7861         nfs_fattr_init(&locations->fattr);
7862         locations->server = server;
7863         locations->nlocations = 0;
7864
7865         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7866         status = nfs4_call_sync_sequence(clnt, server, &msg,
7867                                         &args.seq_args, &res.seq_res);
7868         if (status == NFS4_OK &&
7869             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7870                 status = -NFS4ERR_LEASE_MOVED;
7871         return status;
7872 }
7873
7874 #endif  /* CONFIG_NFS_V4_1 */
7875
7876 /**
7877  * nfs4_proc_get_locations - discover locations for a migrated FSID
7878  * @inode: inode on FSID that is migrating
7879  * @locations: result of query
7880  * @page: buffer
7881  * @cred: credential to use for this operation
7882  *
7883  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7884  * operation failed, or a negative errno if a local error occurred.
7885  *
7886  * On success, "locations" is filled in, but if the server has
7887  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7888  * asserted.
7889  *
7890  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7891  * from this client that require migration recovery.
7892  */
7893 int nfs4_proc_get_locations(struct inode *inode,
7894                             struct nfs4_fs_locations *locations,
7895                             struct page *page, const struct cred *cred)
7896 {
7897         struct nfs_server *server = NFS_SERVER(inode);
7898         struct nfs_client *clp = server->nfs_client;
7899         const struct nfs4_mig_recovery_ops *ops =
7900                                         clp->cl_mvops->mig_recovery_ops;
7901         struct nfs4_exception exception = {
7902                 .interruptible = true,
7903         };
7904         int status;
7905
7906         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7907                 (unsigned long long)server->fsid.major,
7908                 (unsigned long long)server->fsid.minor,
7909                 clp->cl_hostname);
7910         nfs_display_fhandle(NFS_FH(inode), __func__);
7911
7912         do {
7913                 status = ops->get_locations(inode, locations, page, cred);
7914                 if (status != -NFS4ERR_DELAY)
7915                         break;
7916                 nfs4_handle_exception(server, status, &exception);
7917         } while (exception.retry);
7918         return status;
7919 }
7920
7921 /*
7922  * This operation also signals the server that this client is
7923  * performing "lease moved" recovery.  The server can stop
7924  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7925  * is appended to this compound to identify the client ID which is
7926  * performing recovery.
7927  */
7928 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7929 {
7930         struct nfs_server *server = NFS_SERVER(inode);
7931         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7932         struct rpc_clnt *clnt = server->client;
7933         struct nfs4_fsid_present_arg args = {
7934                 .fh             = NFS_FH(inode),
7935                 .clientid       = clp->cl_clientid,
7936                 .renew          = 1,            /* append RENEW */
7937         };
7938         struct nfs4_fsid_present_res res = {
7939                 .renew          = 1,
7940         };
7941         struct rpc_message msg = {
7942                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7943                 .rpc_argp       = &args,
7944                 .rpc_resp       = &res,
7945                 .rpc_cred       = cred,
7946         };
7947         unsigned long now = jiffies;
7948         int status;
7949
7950         res.fh = nfs_alloc_fhandle();
7951         if (res.fh == NULL)
7952                 return -ENOMEM;
7953
7954         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7955         status = nfs4_call_sync_sequence(clnt, server, &msg,
7956                                                 &args.seq_args, &res.seq_res);
7957         nfs_free_fhandle(res.fh);
7958         if (status)
7959                 return status;
7960
7961         do_renew_lease(clp, now);
7962         return 0;
7963 }
7964
7965 #ifdef CONFIG_NFS_V4_1
7966
7967 /*
7968  * This operation also signals the server that this client is
7969  * performing "lease moved" recovery.  The server can stop asserting
7970  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7971  * this operation is identified in the SEQUENCE operation in this
7972  * compound.
7973  */
7974 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7975 {
7976         struct nfs_server *server = NFS_SERVER(inode);
7977         struct rpc_clnt *clnt = server->client;
7978         struct nfs4_fsid_present_arg args = {
7979                 .fh             = NFS_FH(inode),
7980         };
7981         struct nfs4_fsid_present_res res = {
7982         };
7983         struct rpc_message msg = {
7984                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7985                 .rpc_argp       = &args,
7986                 .rpc_resp       = &res,
7987                 .rpc_cred       = cred,
7988         };
7989         int status;
7990
7991         res.fh = nfs_alloc_fhandle();
7992         if (res.fh == NULL)
7993                 return -ENOMEM;
7994
7995         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7996         status = nfs4_call_sync_sequence(clnt, server, &msg,
7997                                                 &args.seq_args, &res.seq_res);
7998         nfs_free_fhandle(res.fh);
7999         if (status == NFS4_OK &&
8000             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8001                 status = -NFS4ERR_LEASE_MOVED;
8002         return status;
8003 }
8004
8005 #endif  /* CONFIG_NFS_V4_1 */
8006
8007 /**
8008  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8009  * @inode: inode on FSID to check
8010  * @cred: credential to use for this operation
8011  *
8012  * Server indicates whether the FSID is present, moved, or not
8013  * recognized.  This operation is necessary to clear a LEASE_MOVED
8014  * condition for this client ID.
8015  *
8016  * Returns NFS4_OK if the FSID is present on this server,
8017  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8018  *  NFS4ERR code if some error occurred on the server, or a
8019  *  negative errno if a local failure occurred.
8020  */
8021 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8022 {
8023         struct nfs_server *server = NFS_SERVER(inode);
8024         struct nfs_client *clp = server->nfs_client;
8025         const struct nfs4_mig_recovery_ops *ops =
8026                                         clp->cl_mvops->mig_recovery_ops;
8027         struct nfs4_exception exception = {
8028                 .interruptible = true,
8029         };
8030         int status;
8031
8032         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8033                 (unsigned long long)server->fsid.major,
8034                 (unsigned long long)server->fsid.minor,
8035                 clp->cl_hostname);
8036         nfs_display_fhandle(NFS_FH(inode), __func__);
8037
8038         do {
8039                 status = ops->fsid_present(inode, cred);
8040                 if (status != -NFS4ERR_DELAY)
8041                         break;
8042                 nfs4_handle_exception(server, status, &exception);
8043         } while (exception.retry);
8044         return status;
8045 }
8046
8047 /*
8048  * If 'use_integrity' is true and the state managment nfs_client
8049  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8050  * and the machine credential as per RFC3530bis and RFC5661 Security
8051  * Considerations sections. Otherwise, just use the user cred with the
8052  * filesystem's rpc_client.
8053  */
8054 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8055 {
8056         int status;
8057         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8058         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8059         struct nfs4_secinfo_arg args = {
8060                 .dir_fh = NFS_FH(dir),
8061                 .name   = name,
8062         };
8063         struct nfs4_secinfo_res res = {
8064                 .flavors     = flavors,
8065         };
8066         struct rpc_message msg = {
8067                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8068                 .rpc_argp = &args,
8069                 .rpc_resp = &res,
8070         };
8071         struct nfs4_call_sync_data data = {
8072                 .seq_server = NFS_SERVER(dir),
8073                 .seq_args = &args.seq_args,
8074                 .seq_res = &res.seq_res,
8075         };
8076         struct rpc_task_setup task_setup = {
8077                 .rpc_client = clnt,
8078                 .rpc_message = &msg,
8079                 .callback_ops = clp->cl_mvops->call_sync_ops,
8080                 .callback_data = &data,
8081                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8082         };
8083         const struct cred *cred = NULL;
8084
8085         if (use_integrity) {
8086                 clnt = clp->cl_rpcclient;
8087                 task_setup.rpc_client = clnt;
8088
8089                 cred = nfs4_get_clid_cred(clp);
8090                 msg.rpc_cred = cred;
8091         }
8092
8093         dprintk("NFS call  secinfo %s\n", name->name);
8094
8095         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8096         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8097         status = nfs4_call_sync_custom(&task_setup);
8098
8099         dprintk("NFS reply  secinfo: %d\n", status);
8100
8101         put_cred(cred);
8102         return status;
8103 }
8104
8105 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8106                       struct nfs4_secinfo_flavors *flavors)
8107 {
8108         struct nfs4_exception exception = {
8109                 .interruptible = true,
8110         };
8111         int err;
8112         do {
8113                 err = -NFS4ERR_WRONGSEC;
8114
8115                 /* try to use integrity protection with machine cred */
8116                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8117                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8118
8119                 /*
8120                  * if unable to use integrity protection, or SECINFO with
8121                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8122                  * disallowed by spec, but exists in deployed servers) use
8123                  * the current filesystem's rpc_client and the user cred.
8124                  */
8125                 if (err == -NFS4ERR_WRONGSEC)
8126                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8127
8128                 trace_nfs4_secinfo(dir, name, err);
8129                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8130                                 &exception);
8131         } while (exception.retry);
8132         return err;
8133 }
8134
8135 #ifdef CONFIG_NFS_V4_1
8136 /*
8137  * Check the exchange flags returned by the server for invalid flags, having
8138  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8139  * DS flags set.
8140  */
8141 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8142 {
8143         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8144                 goto out_inval;
8145         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8146                 goto out_inval;
8147         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8148             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8149                 goto out_inval;
8150         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8151                 goto out_inval;
8152         return NFS_OK;
8153 out_inval:
8154         return -NFS4ERR_INVAL;
8155 }
8156
8157 static bool
8158 nfs41_same_server_scope(struct nfs41_server_scope *a,
8159                         struct nfs41_server_scope *b)
8160 {
8161         if (a->server_scope_sz != b->server_scope_sz)
8162                 return false;
8163         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8164 }
8165
8166 static void
8167 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8168 {
8169         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8170         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8171         struct nfs_client *clp = args->client;
8172
8173         switch (task->tk_status) {
8174         case -NFS4ERR_BADSESSION:
8175         case -NFS4ERR_DEADSESSION:
8176                 nfs4_schedule_session_recovery(clp->cl_session,
8177                                 task->tk_status);
8178         }
8179         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8180                         res->dir != NFS4_CDFS4_BOTH) {
8181                 rpc_task_close_connection(task);
8182                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8183                         rpc_restart_call(task);
8184         }
8185 }
8186
8187 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8188         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8189 };
8190
8191 /*
8192  * nfs4_proc_bind_one_conn_to_session()
8193  *
8194  * The 4.1 client currently uses the same TCP connection for the
8195  * fore and backchannel.
8196  */
8197 static
8198 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8199                 struct rpc_xprt *xprt,
8200                 struct nfs_client *clp,
8201                 const struct cred *cred)
8202 {
8203         int status;
8204         struct nfs41_bind_conn_to_session_args args = {
8205                 .client = clp,
8206                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8207                 .retries = 0,
8208         };
8209         struct nfs41_bind_conn_to_session_res res;
8210         struct rpc_message msg = {
8211                 .rpc_proc =
8212                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8213                 .rpc_argp = &args,
8214                 .rpc_resp = &res,
8215                 .rpc_cred = cred,
8216         };
8217         struct rpc_task_setup task_setup_data = {
8218                 .rpc_client = clnt,
8219                 .rpc_xprt = xprt,
8220                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8221                 .rpc_message = &msg,
8222                 .flags = RPC_TASK_TIMEOUT,
8223         };
8224         struct rpc_task *task;
8225
8226         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8227         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8228                 args.dir = NFS4_CDFC4_FORE;
8229
8230         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8231         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8232                 args.dir = NFS4_CDFC4_FORE;
8233
8234         task = rpc_run_task(&task_setup_data);
8235         if (!IS_ERR(task)) {
8236                 status = task->tk_status;
8237                 rpc_put_task(task);
8238         } else
8239                 status = PTR_ERR(task);
8240         trace_nfs4_bind_conn_to_session(clp, status);
8241         if (status == 0) {
8242                 if (memcmp(res.sessionid.data,
8243                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8244                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8245                         return -EIO;
8246                 }
8247                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8248                         dprintk("NFS: %s: Unexpected direction from server\n",
8249                                 __func__);
8250                         return -EIO;
8251                 }
8252                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8253                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8254                                 __func__);
8255                         return -EIO;
8256                 }
8257         }
8258
8259         return status;
8260 }
8261
8262 struct rpc_bind_conn_calldata {
8263         struct nfs_client *clp;
8264         const struct cred *cred;
8265 };
8266
8267 static int
8268 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8269                 struct rpc_xprt *xprt,
8270                 void *calldata)
8271 {
8272         struct rpc_bind_conn_calldata *p = calldata;
8273
8274         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8275 }
8276
8277 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8278 {
8279         struct rpc_bind_conn_calldata data = {
8280                 .clp = clp,
8281                 .cred = cred,
8282         };
8283         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8284                         nfs4_proc_bind_conn_to_session_callback, &data);
8285 }
8286
8287 /*
8288  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8289  * and operations we'd like to see to enable certain features in the allow map
8290  */
8291 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8292         .how = SP4_MACH_CRED,
8293         .enforce.u.words = {
8294                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8295                       1 << (OP_EXCHANGE_ID - 32) |
8296                       1 << (OP_CREATE_SESSION - 32) |
8297                       1 << (OP_DESTROY_SESSION - 32) |
8298                       1 << (OP_DESTROY_CLIENTID - 32)
8299         },
8300         .allow.u.words = {
8301                 [0] = 1 << (OP_CLOSE) |
8302                       1 << (OP_OPEN_DOWNGRADE) |
8303                       1 << (OP_LOCKU) |
8304                       1 << (OP_DELEGRETURN) |
8305                       1 << (OP_COMMIT),
8306                 [1] = 1 << (OP_SECINFO - 32) |
8307                       1 << (OP_SECINFO_NO_NAME - 32) |
8308                       1 << (OP_LAYOUTRETURN - 32) |
8309                       1 << (OP_TEST_STATEID - 32) |
8310                       1 << (OP_FREE_STATEID - 32) |
8311                       1 << (OP_WRITE - 32)
8312         }
8313 };
8314
8315 /*
8316  * Select the state protection mode for client `clp' given the server results
8317  * from exchange_id in `sp'.
8318  *
8319  * Returns 0 on success, negative errno otherwise.
8320  */
8321 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8322                                  struct nfs41_state_protection *sp)
8323 {
8324         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8325                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8326                       1 << (OP_EXCHANGE_ID - 32) |
8327                       1 << (OP_CREATE_SESSION - 32) |
8328                       1 << (OP_DESTROY_SESSION - 32) |
8329                       1 << (OP_DESTROY_CLIENTID - 32)
8330         };
8331         unsigned long flags = 0;
8332         unsigned int i;
8333         int ret = 0;
8334
8335         if (sp->how == SP4_MACH_CRED) {
8336                 /* Print state protect result */
8337                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8338                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8339                         if (test_bit(i, sp->enforce.u.longs))
8340                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8341                         if (test_bit(i, sp->allow.u.longs))
8342                                 dfprintk(MOUNT, "  allow op %d\n", i);
8343                 }
8344
8345                 /* make sure nothing is on enforce list that isn't supported */
8346                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8347                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8348                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8349                                 ret = -EINVAL;
8350                                 goto out;
8351                         }
8352                 }
8353
8354                 /*
8355                  * Minimal mode - state operations are allowed to use machine
8356                  * credential.  Note this already happens by default, so the
8357                  * client doesn't have to do anything more than the negotiation.
8358                  *
8359                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8360                  *       we're already using the machine cred for exchange_id
8361                  *       and will never use a different cred.
8362                  */
8363                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8364                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8365                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8366                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8367                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8368                         dfprintk(MOUNT, "  minimal mode enabled\n");
8369                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8370                 } else {
8371                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8372                         ret = -EINVAL;
8373                         goto out;
8374                 }
8375
8376                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8377                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8378                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8379                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8380                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8381                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8382                 }
8383
8384                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8385                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8386                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8387                 }
8388
8389                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8390                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8391                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8392                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8393                 }
8394
8395                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8396                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8397                         dfprintk(MOUNT, "  stateid mode enabled\n");
8398                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8399                 }
8400
8401                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8402                         dfprintk(MOUNT, "  write mode enabled\n");
8403                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8404                 }
8405
8406                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8407                         dfprintk(MOUNT, "  commit mode enabled\n");
8408                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8409                 }
8410         }
8411 out:
8412         clp->cl_sp4_flags = flags;
8413         return ret;
8414 }
8415
8416 struct nfs41_exchange_id_data {
8417         struct nfs41_exchange_id_res res;
8418         struct nfs41_exchange_id_args args;
8419 };
8420
8421 static void nfs4_exchange_id_release(void *data)
8422 {
8423         struct nfs41_exchange_id_data *cdata =
8424                                         (struct nfs41_exchange_id_data *)data;
8425
8426         nfs_put_client(cdata->args.client);
8427         kfree(cdata->res.impl_id);
8428         kfree(cdata->res.server_scope);
8429         kfree(cdata->res.server_owner);
8430         kfree(cdata);
8431 }
8432
8433 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8434         .rpc_release = nfs4_exchange_id_release,
8435 };
8436
8437 /*
8438  * _nfs4_proc_exchange_id()
8439  *
8440  * Wrapper for EXCHANGE_ID operation.
8441  */
8442 static struct rpc_task *
8443 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8444                         u32 sp4_how, struct rpc_xprt *xprt)
8445 {
8446         struct rpc_message msg = {
8447                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8448                 .rpc_cred = cred,
8449         };
8450         struct rpc_task_setup task_setup_data = {
8451                 .rpc_client = clp->cl_rpcclient,
8452                 .callback_ops = &nfs4_exchange_id_call_ops,
8453                 .rpc_message = &msg,
8454                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8455         };
8456         struct nfs41_exchange_id_data *calldata;
8457         int status;
8458
8459         if (!refcount_inc_not_zero(&clp->cl_count))
8460                 return ERR_PTR(-EIO);
8461
8462         status = -ENOMEM;
8463         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8464         if (!calldata)
8465                 goto out;
8466
8467         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8468
8469         status = nfs4_init_uniform_client_string(clp);
8470         if (status)
8471                 goto out_calldata;
8472
8473         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8474                                                 GFP_NOFS);
8475         status = -ENOMEM;
8476         if (unlikely(calldata->res.server_owner == NULL))
8477                 goto out_calldata;
8478
8479         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8480                                         GFP_NOFS);
8481         if (unlikely(calldata->res.server_scope == NULL))
8482                 goto out_server_owner;
8483
8484         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8485         if (unlikely(calldata->res.impl_id == NULL))
8486                 goto out_server_scope;
8487
8488         switch (sp4_how) {
8489         case SP4_NONE:
8490                 calldata->args.state_protect.how = SP4_NONE;
8491                 break;
8492
8493         case SP4_MACH_CRED:
8494                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8495                 break;
8496
8497         default:
8498                 /* unsupported! */
8499                 WARN_ON_ONCE(1);
8500                 status = -EINVAL;
8501                 goto out_impl_id;
8502         }
8503         if (xprt) {
8504                 task_setup_data.rpc_xprt = xprt;
8505                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8506                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8507                                 sizeof(calldata->args.verifier.data));
8508         }
8509         calldata->args.client = clp;
8510         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8511         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8512 #ifdef CONFIG_NFS_V4_1_MIGRATION
8513         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8514 #endif
8515         msg.rpc_argp = &calldata->args;
8516         msg.rpc_resp = &calldata->res;
8517         task_setup_data.callback_data = calldata;
8518
8519         return rpc_run_task(&task_setup_data);
8520
8521 out_impl_id:
8522         kfree(calldata->res.impl_id);
8523 out_server_scope:
8524         kfree(calldata->res.server_scope);
8525 out_server_owner:
8526         kfree(calldata->res.server_owner);
8527 out_calldata:
8528         kfree(calldata);
8529 out:
8530         nfs_put_client(clp);
8531         return ERR_PTR(status);
8532 }
8533
8534 /*
8535  * _nfs4_proc_exchange_id()
8536  *
8537  * Wrapper for EXCHANGE_ID operation.
8538  */
8539 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8540                         u32 sp4_how)
8541 {
8542         struct rpc_task *task;
8543         struct nfs41_exchange_id_args *argp;
8544         struct nfs41_exchange_id_res *resp;
8545         unsigned long now = jiffies;
8546         int status;
8547
8548         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8549         if (IS_ERR(task))
8550                 return PTR_ERR(task);
8551
8552         argp = task->tk_msg.rpc_argp;
8553         resp = task->tk_msg.rpc_resp;
8554         status = task->tk_status;
8555         if (status  != 0)
8556                 goto out;
8557
8558         status = nfs4_check_cl_exchange_flags(resp->flags,
8559                         clp->cl_mvops->minor_version);
8560         if (status  != 0)
8561                 goto out;
8562
8563         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8564         if (status != 0)
8565                 goto out;
8566
8567         do_renew_lease(clp, now);
8568
8569         clp->cl_clientid = resp->clientid;
8570         clp->cl_exchange_flags = resp->flags;
8571         clp->cl_seqid = resp->seqid;
8572         /* Client ID is not confirmed */
8573         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8574                 clear_bit(NFS4_SESSION_ESTABLISHED,
8575                           &clp->cl_session->session_state);
8576
8577         if (clp->cl_serverscope != NULL &&
8578             !nfs41_same_server_scope(clp->cl_serverscope,
8579                                 resp->server_scope)) {
8580                 dprintk("%s: server_scope mismatch detected\n",
8581                         __func__);
8582                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8583         }
8584
8585         swap(clp->cl_serverowner, resp->server_owner);
8586         swap(clp->cl_serverscope, resp->server_scope);
8587         swap(clp->cl_implid, resp->impl_id);
8588
8589         /* Save the EXCHANGE_ID verifier session trunk tests */
8590         memcpy(clp->cl_confirm.data, argp->verifier.data,
8591                sizeof(clp->cl_confirm.data));
8592 out:
8593         trace_nfs4_exchange_id(clp, status);
8594         rpc_put_task(task);
8595         return status;
8596 }
8597
8598 /*
8599  * nfs4_proc_exchange_id()
8600  *
8601  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8602  *
8603  * Since the clientid has expired, all compounds using sessions
8604  * associated with the stale clientid will be returning
8605  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8606  * be in some phase of session reset.
8607  *
8608  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8609  */
8610 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8611 {
8612         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8613         int status;
8614
8615         /* try SP4_MACH_CRED if krb5i/p */
8616         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8617             authflavor == RPC_AUTH_GSS_KRB5P) {
8618                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8619                 if (!status)
8620                         return 0;
8621         }
8622
8623         /* try SP4_NONE */
8624         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8625 }
8626
8627 /**
8628  * nfs4_test_session_trunk
8629  *
8630  * This is an add_xprt_test() test function called from
8631  * rpc_clnt_setup_test_and_add_xprt.
8632  *
8633  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8634  * and is dereferrenced in nfs4_exchange_id_release
8635  *
8636  * Upon success, add the new transport to the rpc_clnt
8637  *
8638  * @clnt: struct rpc_clnt to get new transport
8639  * @xprt: the rpc_xprt to test
8640  * @data: call data for _nfs4_proc_exchange_id.
8641  */
8642 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8643                             void *data)
8644 {
8645         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8646         struct rpc_task *task;
8647         int status;
8648
8649         u32 sp4_how;
8650
8651         dprintk("--> %s try %s\n", __func__,
8652                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8653
8654         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8655
8656         /* Test connection for session trunking. Async exchange_id call */
8657         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8658         if (IS_ERR(task))
8659                 return;
8660
8661         status = task->tk_status;
8662         if (status == 0)
8663                 status = nfs4_detect_session_trunking(adata->clp,
8664                                 task->tk_msg.rpc_resp, xprt);
8665
8666         if (status == 0)
8667                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8668
8669         rpc_put_task(task);
8670 }
8671 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8672
8673 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8674                 const struct cred *cred)
8675 {
8676         struct rpc_message msg = {
8677                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8678                 .rpc_argp = clp,
8679                 .rpc_cred = cred,
8680         };
8681         int status;
8682
8683         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8684                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8685         trace_nfs4_destroy_clientid(clp, status);
8686         if (status)
8687                 dprintk("NFS: Got error %d from the server %s on "
8688                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8689         return status;
8690 }
8691
8692 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8693                 const struct cred *cred)
8694 {
8695         unsigned int loop;
8696         int ret;
8697
8698         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8699                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8700                 switch (ret) {
8701                 case -NFS4ERR_DELAY:
8702                 case -NFS4ERR_CLIENTID_BUSY:
8703                         ssleep(1);
8704                         break;
8705                 default:
8706                         return ret;
8707                 }
8708         }
8709         return 0;
8710 }
8711
8712 int nfs4_destroy_clientid(struct nfs_client *clp)
8713 {
8714         const struct cred *cred;
8715         int ret = 0;
8716
8717         if (clp->cl_mvops->minor_version < 1)
8718                 goto out;
8719         if (clp->cl_exchange_flags == 0)
8720                 goto out;
8721         if (clp->cl_preserve_clid)
8722                 goto out;
8723         cred = nfs4_get_clid_cred(clp);
8724         ret = nfs4_proc_destroy_clientid(clp, cred);
8725         put_cred(cred);
8726         switch (ret) {
8727         case 0:
8728         case -NFS4ERR_STALE_CLIENTID:
8729                 clp->cl_exchange_flags = 0;
8730         }
8731 out:
8732         return ret;
8733 }
8734
8735 #endif /* CONFIG_NFS_V4_1 */
8736
8737 struct nfs4_get_lease_time_data {
8738         struct nfs4_get_lease_time_args *args;
8739         struct nfs4_get_lease_time_res *res;
8740         struct nfs_client *clp;
8741 };
8742
8743 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8744                                         void *calldata)
8745 {
8746         struct nfs4_get_lease_time_data *data =
8747                         (struct nfs4_get_lease_time_data *)calldata;
8748
8749         dprintk("--> %s\n", __func__);
8750         /* just setup sequence, do not trigger session recovery
8751            since we're invoked within one */
8752         nfs4_setup_sequence(data->clp,
8753                         &data->args->la_seq_args,
8754                         &data->res->lr_seq_res,
8755                         task);
8756         dprintk("<-- %s\n", __func__);
8757 }
8758
8759 /*
8760  * Called from nfs4_state_manager thread for session setup, so don't recover
8761  * from sequence operation or clientid errors.
8762  */
8763 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8764 {
8765         struct nfs4_get_lease_time_data *data =
8766                         (struct nfs4_get_lease_time_data *)calldata;
8767
8768         dprintk("--> %s\n", __func__);
8769         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8770                 return;
8771         switch (task->tk_status) {
8772         case -NFS4ERR_DELAY:
8773         case -NFS4ERR_GRACE:
8774                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8775                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8776                 task->tk_status = 0;
8777                 fallthrough;
8778         case -NFS4ERR_RETRY_UNCACHED_REP:
8779                 rpc_restart_call_prepare(task);
8780                 return;
8781         }
8782         dprintk("<-- %s\n", __func__);
8783 }
8784
8785 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8786         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8787         .rpc_call_done = nfs4_get_lease_time_done,
8788 };
8789
8790 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8791 {
8792         struct nfs4_get_lease_time_args args;
8793         struct nfs4_get_lease_time_res res = {
8794                 .lr_fsinfo = fsinfo,
8795         };
8796         struct nfs4_get_lease_time_data data = {
8797                 .args = &args,
8798                 .res = &res,
8799                 .clp = clp,
8800         };
8801         struct rpc_message msg = {
8802                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8803                 .rpc_argp = &args,
8804                 .rpc_resp = &res,
8805         };
8806         struct rpc_task_setup task_setup = {
8807                 .rpc_client = clp->cl_rpcclient,
8808                 .rpc_message = &msg,
8809                 .callback_ops = &nfs4_get_lease_time_ops,
8810                 .callback_data = &data,
8811                 .flags = RPC_TASK_TIMEOUT,
8812         };
8813
8814         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8815         return nfs4_call_sync_custom(&task_setup);
8816 }
8817
8818 #ifdef CONFIG_NFS_V4_1
8819
8820 /*
8821  * Initialize the values to be used by the client in CREATE_SESSION
8822  * If nfs4_init_session set the fore channel request and response sizes,
8823  * use them.
8824  *
8825  * Set the back channel max_resp_sz_cached to zero to force the client to
8826  * always set csa_cachethis to FALSE because the current implementation
8827  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8828  */
8829 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8830                                     struct rpc_clnt *clnt)
8831 {
8832         unsigned int max_rqst_sz, max_resp_sz;
8833         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8834         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8835
8836         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8837         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8838
8839         /* Fore channel attributes */
8840         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8841         args->fc_attrs.max_resp_sz = max_resp_sz;
8842         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8843         args->fc_attrs.max_reqs = max_session_slots;
8844
8845         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8846                 "max_ops=%u max_reqs=%u\n",
8847                 __func__,
8848                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8849                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8850
8851         /* Back channel attributes */
8852         args->bc_attrs.max_rqst_sz = max_bc_payload;
8853         args->bc_attrs.max_resp_sz = max_bc_payload;
8854         args->bc_attrs.max_resp_sz_cached = 0;
8855         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8856         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8857         if (args->bc_attrs.max_reqs > max_bc_slots)
8858                 args->bc_attrs.max_reqs = max_bc_slots;
8859
8860         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8861                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8862                 __func__,
8863                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8864                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8865                 args->bc_attrs.max_reqs);
8866 }
8867
8868 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8869                 struct nfs41_create_session_res *res)
8870 {
8871         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8872         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8873
8874         if (rcvd->max_resp_sz > sent->max_resp_sz)
8875                 return -EINVAL;
8876         /*
8877          * Our requested max_ops is the minimum we need; we're not
8878          * prepared to break up compounds into smaller pieces than that.
8879          * So, no point even trying to continue if the server won't
8880          * cooperate:
8881          */
8882         if (rcvd->max_ops < sent->max_ops)
8883                 return -EINVAL;
8884         if (rcvd->max_reqs == 0)
8885                 return -EINVAL;
8886         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8887                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8888         return 0;
8889 }
8890
8891 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8892                 struct nfs41_create_session_res *res)
8893 {
8894         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8895         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8896
8897         if (!(res->flags & SESSION4_BACK_CHAN))
8898                 goto out;
8899         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8900                 return -EINVAL;
8901         if (rcvd->max_resp_sz < sent->max_resp_sz)
8902                 return -EINVAL;
8903         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8904                 return -EINVAL;
8905         if (rcvd->max_ops > sent->max_ops)
8906                 return -EINVAL;
8907         if (rcvd->max_reqs > sent->max_reqs)
8908                 return -EINVAL;
8909 out:
8910         return 0;
8911 }
8912
8913 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8914                                      struct nfs41_create_session_res *res)
8915 {
8916         int ret;
8917
8918         ret = nfs4_verify_fore_channel_attrs(args, res);
8919         if (ret)
8920                 return ret;
8921         return nfs4_verify_back_channel_attrs(args, res);
8922 }
8923
8924 static void nfs4_update_session(struct nfs4_session *session,
8925                 struct nfs41_create_session_res *res)
8926 {
8927         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8928         /* Mark client id and session as being confirmed */
8929         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8930         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8931         session->flags = res->flags;
8932         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8933         if (res->flags & SESSION4_BACK_CHAN)
8934                 memcpy(&session->bc_attrs, &res->bc_attrs,
8935                                 sizeof(session->bc_attrs));
8936 }
8937
8938 static int _nfs4_proc_create_session(struct nfs_client *clp,
8939                 const struct cred *cred)
8940 {
8941         struct nfs4_session *session = clp->cl_session;
8942         struct nfs41_create_session_args args = {
8943                 .client = clp,
8944                 .clientid = clp->cl_clientid,
8945                 .seqid = clp->cl_seqid,
8946                 .cb_program = NFS4_CALLBACK,
8947         };
8948         struct nfs41_create_session_res res;
8949
8950         struct rpc_message msg = {
8951                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8952                 .rpc_argp = &args,
8953                 .rpc_resp = &res,
8954                 .rpc_cred = cred,
8955         };
8956         int status;
8957
8958         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8959         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8960
8961         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8962                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8963         trace_nfs4_create_session(clp, status);
8964
8965         switch (status) {
8966         case -NFS4ERR_STALE_CLIENTID:
8967         case -NFS4ERR_DELAY:
8968         case -ETIMEDOUT:
8969         case -EACCES:
8970         case -EAGAIN:
8971                 goto out;
8972         }
8973
8974         clp->cl_seqid++;
8975         if (!status) {
8976                 /* Verify the session's negotiated channel_attrs values */
8977                 status = nfs4_verify_channel_attrs(&args, &res);
8978                 /* Increment the clientid slot sequence id */
8979                 if (status)
8980                         goto out;
8981                 nfs4_update_session(session, &res);
8982         }
8983 out:
8984         return status;
8985 }
8986
8987 /*
8988  * Issues a CREATE_SESSION operation to the server.
8989  * It is the responsibility of the caller to verify the session is
8990  * expired before calling this routine.
8991  */
8992 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8993 {
8994         int status;
8995         unsigned *ptr;
8996         struct nfs4_session *session = clp->cl_session;
8997
8998         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
8999
9000         status = _nfs4_proc_create_session(clp, cred);
9001         if (status)
9002                 goto out;
9003
9004         /* Init or reset the session slot tables */
9005         status = nfs4_setup_session_slot_tables(session);
9006         dprintk("slot table setup returned %d\n", status);
9007         if (status)
9008                 goto out;
9009
9010         ptr = (unsigned *)&session->sess_id.data[0];
9011         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9012                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9013 out:
9014         dprintk("<-- %s\n", __func__);
9015         return status;
9016 }
9017
9018 /*
9019  * Issue the over-the-wire RPC DESTROY_SESSION.
9020  * The caller must serialize access to this routine.
9021  */
9022 int nfs4_proc_destroy_session(struct nfs4_session *session,
9023                 const struct cred *cred)
9024 {
9025         struct rpc_message msg = {
9026                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9027                 .rpc_argp = session,
9028                 .rpc_cred = cred,
9029         };
9030         int status = 0;
9031
9032         dprintk("--> nfs4_proc_destroy_session\n");
9033
9034         /* session is still being setup */
9035         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9036                 return 0;
9037
9038         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9039                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9040         trace_nfs4_destroy_session(session->clp, status);
9041
9042         if (status)
9043                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9044                         "Session has been destroyed regardless...\n", status);
9045
9046         dprintk("<-- nfs4_proc_destroy_session\n");
9047         return status;
9048 }
9049
9050 /*
9051  * Renew the cl_session lease.
9052  */
9053 struct nfs4_sequence_data {
9054         struct nfs_client *clp;
9055         struct nfs4_sequence_args args;
9056         struct nfs4_sequence_res res;
9057 };
9058
9059 static void nfs41_sequence_release(void *data)
9060 {
9061         struct nfs4_sequence_data *calldata = data;
9062         struct nfs_client *clp = calldata->clp;
9063
9064         if (refcount_read(&clp->cl_count) > 1)
9065                 nfs4_schedule_state_renewal(clp);
9066         nfs_put_client(clp);
9067         kfree(calldata);
9068 }
9069
9070 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9071 {
9072         switch(task->tk_status) {
9073         case -NFS4ERR_DELAY:
9074                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9075                 return -EAGAIN;
9076         default:
9077                 nfs4_schedule_lease_recovery(clp);
9078         }
9079         return 0;
9080 }
9081
9082 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9083 {
9084         struct nfs4_sequence_data *calldata = data;
9085         struct nfs_client *clp = calldata->clp;
9086
9087         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9088                 return;
9089
9090         trace_nfs4_sequence(clp, task->tk_status);
9091         if (task->tk_status < 0) {
9092                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9093                 if (refcount_read(&clp->cl_count) == 1)
9094                         goto out;
9095
9096                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9097                         rpc_restart_call_prepare(task);
9098                         return;
9099                 }
9100         }
9101         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9102 out:
9103         dprintk("<-- %s\n", __func__);
9104 }
9105
9106 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9107 {
9108         struct nfs4_sequence_data *calldata = data;
9109         struct nfs_client *clp = calldata->clp;
9110         struct nfs4_sequence_args *args;
9111         struct nfs4_sequence_res *res;
9112
9113         args = task->tk_msg.rpc_argp;
9114         res = task->tk_msg.rpc_resp;
9115
9116         nfs4_setup_sequence(clp, args, res, task);
9117 }
9118
9119 static const struct rpc_call_ops nfs41_sequence_ops = {
9120         .rpc_call_done = nfs41_sequence_call_done,
9121         .rpc_call_prepare = nfs41_sequence_prepare,
9122         .rpc_release = nfs41_sequence_release,
9123 };
9124
9125 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9126                 const struct cred *cred,
9127                 struct nfs4_slot *slot,
9128                 bool is_privileged)
9129 {
9130         struct nfs4_sequence_data *calldata;
9131         struct rpc_message msg = {
9132                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9133                 .rpc_cred = cred,
9134         };
9135         struct rpc_task_setup task_setup_data = {
9136                 .rpc_client = clp->cl_rpcclient,
9137                 .rpc_message = &msg,
9138                 .callback_ops = &nfs41_sequence_ops,
9139                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
9140         };
9141         struct rpc_task *ret;
9142
9143         ret = ERR_PTR(-EIO);
9144         if (!refcount_inc_not_zero(&clp->cl_count))
9145                 goto out_err;
9146
9147         ret = ERR_PTR(-ENOMEM);
9148         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9149         if (calldata == NULL)
9150                 goto out_put_clp;
9151         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9152         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9153         msg.rpc_argp = &calldata->args;
9154         msg.rpc_resp = &calldata->res;
9155         calldata->clp = clp;
9156         task_setup_data.callback_data = calldata;
9157
9158         ret = rpc_run_task(&task_setup_data);
9159         if (IS_ERR(ret))
9160                 goto out_err;
9161         return ret;
9162 out_put_clp:
9163         nfs_put_client(clp);
9164 out_err:
9165         nfs41_release_slot(slot);
9166         return ret;
9167 }
9168
9169 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9170 {
9171         struct rpc_task *task;
9172         int ret = 0;
9173
9174         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9175                 return -EAGAIN;
9176         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9177         if (IS_ERR(task))
9178                 ret = PTR_ERR(task);
9179         else
9180                 rpc_put_task_async(task);
9181         dprintk("<-- %s status=%d\n", __func__, ret);
9182         return ret;
9183 }
9184
9185 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9186 {
9187         struct rpc_task *task;
9188         int ret;
9189
9190         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9191         if (IS_ERR(task)) {
9192                 ret = PTR_ERR(task);
9193                 goto out;
9194         }
9195         ret = rpc_wait_for_completion_task(task);
9196         if (!ret)
9197                 ret = task->tk_status;
9198         rpc_put_task(task);
9199 out:
9200         dprintk("<-- %s status=%d\n", __func__, ret);
9201         return ret;
9202 }
9203
9204 struct nfs4_reclaim_complete_data {
9205         struct nfs_client *clp;
9206         struct nfs41_reclaim_complete_args arg;
9207         struct nfs41_reclaim_complete_res res;
9208 };
9209
9210 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9211 {
9212         struct nfs4_reclaim_complete_data *calldata = data;
9213
9214         nfs4_setup_sequence(calldata->clp,
9215                         &calldata->arg.seq_args,
9216                         &calldata->res.seq_res,
9217                         task);
9218 }
9219
9220 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9221 {
9222         switch(task->tk_status) {
9223         case 0:
9224                 wake_up_all(&clp->cl_lock_waitq);
9225                 fallthrough;
9226         case -NFS4ERR_COMPLETE_ALREADY:
9227         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9228                 break;
9229         case -NFS4ERR_DELAY:
9230                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9231                 fallthrough;
9232         case -NFS4ERR_RETRY_UNCACHED_REP:
9233                 return -EAGAIN;
9234         case -NFS4ERR_BADSESSION:
9235         case -NFS4ERR_DEADSESSION:
9236         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9237                 break;
9238         default:
9239                 nfs4_schedule_lease_recovery(clp);
9240         }
9241         return 0;
9242 }
9243
9244 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9245 {
9246         struct nfs4_reclaim_complete_data *calldata = data;
9247         struct nfs_client *clp = calldata->clp;
9248         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9249
9250         dprintk("--> %s\n", __func__);
9251         if (!nfs41_sequence_done(task, res))
9252                 return;
9253
9254         trace_nfs4_reclaim_complete(clp, task->tk_status);
9255         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9256                 rpc_restart_call_prepare(task);
9257                 return;
9258         }
9259         dprintk("<-- %s\n", __func__);
9260 }
9261
9262 static void nfs4_free_reclaim_complete_data(void *data)
9263 {
9264         struct nfs4_reclaim_complete_data *calldata = data;
9265
9266         kfree(calldata);
9267 }
9268
9269 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9270         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9271         .rpc_call_done = nfs4_reclaim_complete_done,
9272         .rpc_release = nfs4_free_reclaim_complete_data,
9273 };
9274
9275 /*
9276  * Issue a global reclaim complete.
9277  */
9278 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9279                 const struct cred *cred)
9280 {
9281         struct nfs4_reclaim_complete_data *calldata;
9282         struct rpc_message msg = {
9283                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9284                 .rpc_cred = cred,
9285         };
9286         struct rpc_task_setup task_setup_data = {
9287                 .rpc_client = clp->cl_rpcclient,
9288                 .rpc_message = &msg,
9289                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9290                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9291         };
9292         int status = -ENOMEM;
9293
9294         dprintk("--> %s\n", __func__);
9295         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9296         if (calldata == NULL)
9297                 goto out;
9298         calldata->clp = clp;
9299         calldata->arg.one_fs = 0;
9300
9301         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9302         msg.rpc_argp = &calldata->arg;
9303         msg.rpc_resp = &calldata->res;
9304         task_setup_data.callback_data = calldata;
9305         status = nfs4_call_sync_custom(&task_setup_data);
9306 out:
9307         dprintk("<-- %s status=%d\n", __func__, status);
9308         return status;
9309 }
9310
9311 static void
9312 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9313 {
9314         struct nfs4_layoutget *lgp = calldata;
9315         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9316
9317         dprintk("--> %s\n", __func__);
9318         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9319                                 &lgp->res.seq_res, task);
9320         dprintk("<-- %s\n", __func__);
9321 }
9322
9323 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9324 {
9325         struct nfs4_layoutget *lgp = calldata;
9326
9327         dprintk("--> %s\n", __func__);
9328         nfs41_sequence_process(task, &lgp->res.seq_res);
9329         dprintk("<-- %s\n", __func__);
9330 }
9331
9332 static int
9333 nfs4_layoutget_handle_exception(struct rpc_task *task,
9334                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9335 {
9336         struct inode *inode = lgp->args.inode;
9337         struct nfs_server *server = NFS_SERVER(inode);
9338         struct pnfs_layout_hdr *lo;
9339         int nfs4err = task->tk_status;
9340         int err, status = 0;
9341         LIST_HEAD(head);
9342
9343         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9344
9345         nfs4_sequence_free_slot(&lgp->res.seq_res);
9346
9347         switch (nfs4err) {
9348         case 0:
9349                 goto out;
9350
9351         /*
9352          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9353          * on the file. set tk_status to -ENODATA to tell upper layer to
9354          * retry go inband.
9355          */
9356         case -NFS4ERR_LAYOUTUNAVAILABLE:
9357                 status = -ENODATA;
9358                 goto out;
9359         /*
9360          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9361          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9362          */
9363         case -NFS4ERR_BADLAYOUT:
9364                 status = -EOVERFLOW;
9365                 goto out;
9366         /*
9367          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9368          * (or clients) writing to the same RAID stripe except when
9369          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9370          *
9371          * Treat it like we would RECALLCONFLICT -- we retry for a little
9372          * while, and then eventually give up.
9373          */
9374         case -NFS4ERR_LAYOUTTRYLATER:
9375                 if (lgp->args.minlength == 0) {
9376                         status = -EOVERFLOW;
9377                         goto out;
9378                 }
9379                 status = -EBUSY;
9380                 break;
9381         case -NFS4ERR_RECALLCONFLICT:
9382                 status = -ERECALLCONFLICT;
9383                 break;
9384         case -NFS4ERR_DELEG_REVOKED:
9385         case -NFS4ERR_ADMIN_REVOKED:
9386         case -NFS4ERR_EXPIRED:
9387         case -NFS4ERR_BAD_STATEID:
9388                 exception->timeout = 0;
9389                 spin_lock(&inode->i_lock);
9390                 lo = NFS_I(inode)->layout;
9391                 /* If the open stateid was bad, then recover it. */
9392                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9393                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9394                         spin_unlock(&inode->i_lock);
9395                         exception->state = lgp->args.ctx->state;
9396                         exception->stateid = &lgp->args.stateid;
9397                         break;
9398                 }
9399
9400                 /*
9401                  * Mark the bad layout state as invalid, then retry
9402                  */
9403                 pnfs_mark_layout_stateid_invalid(lo, &head);
9404                 spin_unlock(&inode->i_lock);
9405                 nfs_commit_inode(inode, 0);
9406                 pnfs_free_lseg_list(&head);
9407                 status = -EAGAIN;
9408                 goto out;
9409         }
9410
9411         err = nfs4_handle_exception(server, nfs4err, exception);
9412         if (!status) {
9413                 if (exception->retry)
9414                         status = -EAGAIN;
9415                 else
9416                         status = err;
9417         }
9418 out:
9419         dprintk("<-- %s\n", __func__);
9420         return status;
9421 }
9422
9423 size_t max_response_pages(struct nfs_server *server)
9424 {
9425         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9426         return nfs_page_array_len(0, max_resp_sz);
9427 }
9428
9429 static void nfs4_layoutget_release(void *calldata)
9430 {
9431         struct nfs4_layoutget *lgp = calldata;
9432
9433         dprintk("--> %s\n", __func__);
9434         nfs4_sequence_free_slot(&lgp->res.seq_res);
9435         pnfs_layoutget_free(lgp);
9436         dprintk("<-- %s\n", __func__);
9437 }
9438
9439 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9440         .rpc_call_prepare = nfs4_layoutget_prepare,
9441         .rpc_call_done = nfs4_layoutget_done,
9442         .rpc_release = nfs4_layoutget_release,
9443 };
9444
9445 struct pnfs_layout_segment *
9446 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9447 {
9448         struct inode *inode = lgp->args.inode;
9449         struct nfs_server *server = NFS_SERVER(inode);
9450         struct rpc_task *task;
9451         struct rpc_message msg = {
9452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9453                 .rpc_argp = &lgp->args,
9454                 .rpc_resp = &lgp->res,
9455                 .rpc_cred = lgp->cred,
9456         };
9457         struct rpc_task_setup task_setup_data = {
9458                 .rpc_client = server->client,
9459                 .rpc_message = &msg,
9460                 .callback_ops = &nfs4_layoutget_call_ops,
9461                 .callback_data = lgp,
9462                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9463         };
9464         struct pnfs_layout_segment *lseg = NULL;
9465         struct nfs4_exception exception = {
9466                 .inode = inode,
9467                 .timeout = *timeout,
9468         };
9469         int status = 0;
9470
9471         dprintk("--> %s\n", __func__);
9472
9473         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9474         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9475
9476         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9477
9478         task = rpc_run_task(&task_setup_data);
9479
9480         status = rpc_wait_for_completion_task(task);
9481         if (status != 0)
9482                 goto out;
9483
9484         if (task->tk_status < 0) {
9485                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9486                 *timeout = exception.timeout;
9487         } else if (lgp->res.layoutp->len == 0) {
9488                 status = -EAGAIN;
9489                 *timeout = nfs4_update_delay(&exception.timeout);
9490         } else
9491                 lseg = pnfs_layout_process(lgp);
9492 out:
9493         trace_nfs4_layoutget(lgp->args.ctx,
9494                         &lgp->args.range,
9495                         &lgp->res.range,
9496                         &lgp->res.stateid,
9497                         status);
9498
9499         rpc_put_task(task);
9500         dprintk("<-- %s status=%d\n", __func__, status);
9501         if (status)
9502                 return ERR_PTR(status);
9503         return lseg;
9504 }
9505
9506 static void
9507 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9508 {
9509         struct nfs4_layoutreturn *lrp = calldata;
9510
9511         dprintk("--> %s\n", __func__);
9512         nfs4_setup_sequence(lrp->clp,
9513                         &lrp->args.seq_args,
9514                         &lrp->res.seq_res,
9515                         task);
9516         if (!pnfs_layout_is_valid(lrp->args.layout))
9517                 rpc_exit(task, 0);
9518 }
9519
9520 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9521 {
9522         struct nfs4_layoutreturn *lrp = calldata;
9523         struct nfs_server *server;
9524
9525         dprintk("--> %s\n", __func__);
9526
9527         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9528                 return;
9529
9530         /*
9531          * Was there an RPC level error? Assume the call succeeded,
9532          * and that we need to release the layout
9533          */
9534         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9535                 lrp->res.lrs_present = 0;
9536                 return;
9537         }
9538
9539         server = NFS_SERVER(lrp->args.inode);
9540         switch (task->tk_status) {
9541         case -NFS4ERR_OLD_STATEID:
9542                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9543                                         &lrp->args.range,
9544                                         lrp->args.inode))
9545                         goto out_restart;
9546                 fallthrough;
9547         default:
9548                 task->tk_status = 0;
9549                 fallthrough;
9550         case 0:
9551                 break;
9552         case -NFS4ERR_DELAY:
9553                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9554                         break;
9555                 goto out_restart;
9556         }
9557         dprintk("<-- %s\n", __func__);
9558         return;
9559 out_restart:
9560         task->tk_status = 0;
9561         nfs4_sequence_free_slot(&lrp->res.seq_res);
9562         rpc_restart_call_prepare(task);
9563 }
9564
9565 static void nfs4_layoutreturn_release(void *calldata)
9566 {
9567         struct nfs4_layoutreturn *lrp = calldata;
9568         struct pnfs_layout_hdr *lo = lrp->args.layout;
9569
9570         dprintk("--> %s\n", __func__);
9571         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9572                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9573         nfs4_sequence_free_slot(&lrp->res.seq_res);
9574         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9575                 lrp->ld_private.ops->free(&lrp->ld_private);
9576         pnfs_put_layout_hdr(lrp->args.layout);
9577         nfs_iput_and_deactive(lrp->inode);
9578         put_cred(lrp->cred);
9579         kfree(calldata);
9580         dprintk("<-- %s\n", __func__);
9581 }
9582
9583 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9584         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9585         .rpc_call_done = nfs4_layoutreturn_done,
9586         .rpc_release = nfs4_layoutreturn_release,
9587 };
9588
9589 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9590 {
9591         struct rpc_task *task;
9592         struct rpc_message msg = {
9593                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9594                 .rpc_argp = &lrp->args,
9595                 .rpc_resp = &lrp->res,
9596                 .rpc_cred = lrp->cred,
9597         };
9598         struct rpc_task_setup task_setup_data = {
9599                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9600                 .rpc_message = &msg,
9601                 .callback_ops = &nfs4_layoutreturn_call_ops,
9602                 .callback_data = lrp,
9603         };
9604         int status = 0;
9605
9606         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9607                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9608                         &task_setup_data.rpc_client, &msg);
9609
9610         dprintk("--> %s\n", __func__);
9611         if (!sync) {
9612                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9613                 if (!lrp->inode) {
9614                         nfs4_layoutreturn_release(lrp);
9615                         return -EAGAIN;
9616                 }
9617                 task_setup_data.flags |= RPC_TASK_ASYNC;
9618         }
9619         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9620         task = rpc_run_task(&task_setup_data);
9621         if (IS_ERR(task))
9622                 return PTR_ERR(task);
9623         if (sync)
9624                 status = task->tk_status;
9625         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9626         dprintk("<-- %s status=%d\n", __func__, status);
9627         rpc_put_task(task);
9628         return status;
9629 }
9630
9631 static int
9632 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9633                 struct pnfs_device *pdev,
9634                 const struct cred *cred)
9635 {
9636         struct nfs4_getdeviceinfo_args args = {
9637                 .pdev = pdev,
9638                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9639                         NOTIFY_DEVICEID4_DELETE,
9640         };
9641         struct nfs4_getdeviceinfo_res res = {
9642                 .pdev = pdev,
9643         };
9644         struct rpc_message msg = {
9645                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9646                 .rpc_argp = &args,
9647                 .rpc_resp = &res,
9648                 .rpc_cred = cred,
9649         };
9650         int status;
9651
9652         dprintk("--> %s\n", __func__);
9653         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9654         if (res.notification & ~args.notify_types)
9655                 dprintk("%s: unsupported notification\n", __func__);
9656         if (res.notification != args.notify_types)
9657                 pdev->nocache = 1;
9658
9659         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9660
9661         dprintk("<-- %s status=%d\n", __func__, status);
9662
9663         return status;
9664 }
9665
9666 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9667                 struct pnfs_device *pdev,
9668                 const struct cred *cred)
9669 {
9670         struct nfs4_exception exception = { };
9671         int err;
9672
9673         do {
9674                 err = nfs4_handle_exception(server,
9675                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9676                                         &exception);
9677         } while (exception.retry);
9678         return err;
9679 }
9680 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9681
9682 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9683 {
9684         struct nfs4_layoutcommit_data *data = calldata;
9685         struct nfs_server *server = NFS_SERVER(data->args.inode);
9686
9687         nfs4_setup_sequence(server->nfs_client,
9688                         &data->args.seq_args,
9689                         &data->res.seq_res,
9690                         task);
9691 }
9692
9693 static void
9694 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9695 {
9696         struct nfs4_layoutcommit_data *data = calldata;
9697         struct nfs_server *server = NFS_SERVER(data->args.inode);
9698
9699         if (!nfs41_sequence_done(task, &data->res.seq_res))
9700                 return;
9701
9702         switch (task->tk_status) { /* Just ignore these failures */
9703         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9704         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9705         case -NFS4ERR_BADLAYOUT:     /* no layout */
9706         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9707                 task->tk_status = 0;
9708         case 0:
9709                 break;
9710         default:
9711                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9712                         rpc_restart_call_prepare(task);
9713                         return;
9714                 }
9715         }
9716 }
9717
9718 static void nfs4_layoutcommit_release(void *calldata)
9719 {
9720         struct nfs4_layoutcommit_data *data = calldata;
9721
9722         pnfs_cleanup_layoutcommit(data);
9723         nfs_post_op_update_inode_force_wcc(data->args.inode,
9724                                            data->res.fattr);
9725         put_cred(data->cred);
9726         nfs_iput_and_deactive(data->inode);
9727         kfree(data);
9728 }
9729
9730 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9731         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9732         .rpc_call_done = nfs4_layoutcommit_done,
9733         .rpc_release = nfs4_layoutcommit_release,
9734 };
9735
9736 int
9737 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9738 {
9739         struct rpc_message msg = {
9740                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9741                 .rpc_argp = &data->args,
9742                 .rpc_resp = &data->res,
9743                 .rpc_cred = data->cred,
9744         };
9745         struct rpc_task_setup task_setup_data = {
9746                 .task = &data->task,
9747                 .rpc_client = NFS_CLIENT(data->args.inode),
9748                 .rpc_message = &msg,
9749                 .callback_ops = &nfs4_layoutcommit_ops,
9750                 .callback_data = data,
9751         };
9752         struct rpc_task *task;
9753         int status = 0;
9754
9755         dprintk("NFS: initiating layoutcommit call. sync %d "
9756                 "lbw: %llu inode %lu\n", sync,
9757                 data->args.lastbytewritten,
9758                 data->args.inode->i_ino);
9759
9760         if (!sync) {
9761                 data->inode = nfs_igrab_and_active(data->args.inode);
9762                 if (data->inode == NULL) {
9763                         nfs4_layoutcommit_release(data);
9764                         return -EAGAIN;
9765                 }
9766                 task_setup_data.flags = RPC_TASK_ASYNC;
9767         }
9768         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9769         task = rpc_run_task(&task_setup_data);
9770         if (IS_ERR(task))
9771                 return PTR_ERR(task);
9772         if (sync)
9773                 status = task->tk_status;
9774         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9775         dprintk("%s: status %d\n", __func__, status);
9776         rpc_put_task(task);
9777         return status;
9778 }
9779
9780 /*
9781  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9782  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9783  */
9784 static int
9785 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9786                     struct nfs_fsinfo *info,
9787                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9788 {
9789         struct nfs41_secinfo_no_name_args args = {
9790                 .style = SECINFO_STYLE_CURRENT_FH,
9791         };
9792         struct nfs4_secinfo_res res = {
9793                 .flavors = flavors,
9794         };
9795         struct rpc_message msg = {
9796                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9797                 .rpc_argp = &args,
9798                 .rpc_resp = &res,
9799         };
9800         struct nfs4_call_sync_data data = {
9801                 .seq_server = server,
9802                 .seq_args = &args.seq_args,
9803                 .seq_res = &res.seq_res,
9804         };
9805         struct rpc_task_setup task_setup = {
9806                 .rpc_client = server->client,
9807                 .rpc_message = &msg,
9808                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9809                 .callback_data = &data,
9810                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9811         };
9812         const struct cred *cred = NULL;
9813         int status;
9814
9815         if (use_integrity) {
9816                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9817
9818                 cred = nfs4_get_clid_cred(server->nfs_client);
9819                 msg.rpc_cred = cred;
9820         }
9821
9822         dprintk("--> %s\n", __func__);
9823         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9824         status = nfs4_call_sync_custom(&task_setup);
9825         dprintk("<-- %s status=%d\n", __func__, status);
9826
9827         put_cred(cred);
9828
9829         return status;
9830 }
9831
9832 static int
9833 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9834                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9835 {
9836         struct nfs4_exception exception = {
9837                 .interruptible = true,
9838         };
9839         int err;
9840         do {
9841                 /* first try using integrity protection */
9842                 err = -NFS4ERR_WRONGSEC;
9843
9844                 /* try to use integrity protection with machine cred */
9845                 if (_nfs4_is_integrity_protected(server->nfs_client))
9846                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9847                                                           flavors, true);
9848
9849                 /*
9850                  * if unable to use integrity protection, or SECINFO with
9851                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9852                  * disallowed by spec, but exists in deployed servers) use
9853                  * the current filesystem's rpc_client and the user cred.
9854                  */
9855                 if (err == -NFS4ERR_WRONGSEC)
9856                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9857                                                           flavors, false);
9858
9859                 switch (err) {
9860                 case 0:
9861                 case -NFS4ERR_WRONGSEC:
9862                 case -ENOTSUPP:
9863                         goto out;
9864                 default:
9865                         err = nfs4_handle_exception(server, err, &exception);
9866                 }
9867         } while (exception.retry);
9868 out:
9869         return err;
9870 }
9871
9872 static int
9873 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9874                     struct nfs_fsinfo *info)
9875 {
9876         int err;
9877         struct page *page;
9878         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9879         struct nfs4_secinfo_flavors *flavors;
9880         struct nfs4_secinfo4 *secinfo;
9881         int i;
9882
9883         page = alloc_page(GFP_KERNEL);
9884         if (!page) {
9885                 err = -ENOMEM;
9886                 goto out;
9887         }
9888
9889         flavors = page_address(page);
9890         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9891
9892         /*
9893          * Fall back on "guess and check" method if
9894          * the server doesn't support SECINFO_NO_NAME
9895          */
9896         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9897                 err = nfs4_find_root_sec(server, fhandle, info);
9898                 goto out_freepage;
9899         }
9900         if (err)
9901                 goto out_freepage;
9902
9903         for (i = 0; i < flavors->num_flavors; i++) {
9904                 secinfo = &flavors->flavors[i];
9905
9906                 switch (secinfo->flavor) {
9907                 case RPC_AUTH_NULL:
9908                 case RPC_AUTH_UNIX:
9909                 case RPC_AUTH_GSS:
9910                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9911                                         &secinfo->flavor_info);
9912                         break;
9913                 default:
9914                         flavor = RPC_AUTH_MAXFLAVOR;
9915                         break;
9916                 }
9917
9918                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9919                         flavor = RPC_AUTH_MAXFLAVOR;
9920
9921                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9922                         err = nfs4_lookup_root_sec(server, fhandle,
9923                                                    info, flavor);
9924                         if (!err)
9925                                 break;
9926                 }
9927         }
9928
9929         if (flavor == RPC_AUTH_MAXFLAVOR)
9930                 err = -EPERM;
9931
9932 out_freepage:
9933         put_page(page);
9934         if (err == -EACCES)
9935                 return -EPERM;
9936 out:
9937         return err;
9938 }
9939
9940 static int _nfs41_test_stateid(struct nfs_server *server,
9941                 nfs4_stateid *stateid,
9942                 const struct cred *cred)
9943 {
9944         int status;
9945         struct nfs41_test_stateid_args args = {
9946                 .stateid = stateid,
9947         };
9948         struct nfs41_test_stateid_res res;
9949         struct rpc_message msg = {
9950                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9951                 .rpc_argp = &args,
9952                 .rpc_resp = &res,
9953                 .rpc_cred = cred,
9954         };
9955         struct rpc_clnt *rpc_client = server->client;
9956
9957         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9958                 &rpc_client, &msg);
9959
9960         dprintk("NFS call  test_stateid %p\n", stateid);
9961         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9962         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9963                         &args.seq_args, &res.seq_res);
9964         if (status != NFS_OK) {
9965                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9966                 return status;
9967         }
9968         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9969         return -res.status;
9970 }
9971
9972 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9973                 int err, struct nfs4_exception *exception)
9974 {
9975         exception->retry = 0;
9976         switch(err) {
9977         case -NFS4ERR_DELAY:
9978         case -NFS4ERR_RETRY_UNCACHED_REP:
9979                 nfs4_handle_exception(server, err, exception);
9980                 break;
9981         case -NFS4ERR_BADSESSION:
9982         case -NFS4ERR_BADSLOT:
9983         case -NFS4ERR_BAD_HIGH_SLOT:
9984         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9985         case -NFS4ERR_DEADSESSION:
9986                 nfs4_do_handle_exception(server, err, exception);
9987         }
9988 }
9989
9990 /**
9991  * nfs41_test_stateid - perform a TEST_STATEID operation
9992  *
9993  * @server: server / transport on which to perform the operation
9994  * @stateid: state ID to test
9995  * @cred: credential
9996  *
9997  * Returns NFS_OK if the server recognizes that "stateid" is valid.
9998  * Otherwise a negative NFS4ERR value is returned if the operation
9999  * failed or the state ID is not currently valid.
10000  */
10001 static int nfs41_test_stateid(struct nfs_server *server,
10002                 nfs4_stateid *stateid,
10003                 const struct cred *cred)
10004 {
10005         struct nfs4_exception exception = {
10006                 .interruptible = true,
10007         };
10008         int err;
10009         do {
10010                 err = _nfs41_test_stateid(server, stateid, cred);
10011                 nfs4_handle_delay_or_session_error(server, err, &exception);
10012         } while (exception.retry);
10013         return err;
10014 }
10015
10016 struct nfs_free_stateid_data {
10017         struct nfs_server *server;
10018         struct nfs41_free_stateid_args args;
10019         struct nfs41_free_stateid_res res;
10020 };
10021
10022 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10023 {
10024         struct nfs_free_stateid_data *data = calldata;
10025         nfs4_setup_sequence(data->server->nfs_client,
10026                         &data->args.seq_args,
10027                         &data->res.seq_res,
10028                         task);
10029 }
10030
10031 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10032 {
10033         struct nfs_free_stateid_data *data = calldata;
10034
10035         nfs41_sequence_done(task, &data->res.seq_res);
10036
10037         switch (task->tk_status) {
10038         case -NFS4ERR_DELAY:
10039                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10040                         rpc_restart_call_prepare(task);
10041         }
10042 }
10043
10044 static void nfs41_free_stateid_release(void *calldata)
10045 {
10046         kfree(calldata);
10047 }
10048
10049 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10050         .rpc_call_prepare = nfs41_free_stateid_prepare,
10051         .rpc_call_done = nfs41_free_stateid_done,
10052         .rpc_release = nfs41_free_stateid_release,
10053 };
10054
10055 /**
10056  * nfs41_free_stateid - perform a FREE_STATEID operation
10057  *
10058  * @server: server / transport on which to perform the operation
10059  * @stateid: state ID to release
10060  * @cred: credential
10061  * @privileged: set to true if this call needs to be privileged
10062  *
10063  * Note: this function is always asynchronous.
10064  */
10065 static int nfs41_free_stateid(struct nfs_server *server,
10066                 const nfs4_stateid *stateid,
10067                 const struct cred *cred,
10068                 bool privileged)
10069 {
10070         struct rpc_message msg = {
10071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10072                 .rpc_cred = cred,
10073         };
10074         struct rpc_task_setup task_setup = {
10075                 .rpc_client = server->client,
10076                 .rpc_message = &msg,
10077                 .callback_ops = &nfs41_free_stateid_ops,
10078                 .flags = RPC_TASK_ASYNC,
10079         };
10080         struct nfs_free_stateid_data *data;
10081         struct rpc_task *task;
10082
10083         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10084                 &task_setup.rpc_client, &msg);
10085
10086         dprintk("NFS call  free_stateid %p\n", stateid);
10087         data = kmalloc(sizeof(*data), GFP_NOFS);
10088         if (!data)
10089                 return -ENOMEM;
10090         data->server = server;
10091         nfs4_stateid_copy(&data->args.stateid, stateid);
10092
10093         task_setup.callback_data = data;
10094
10095         msg.rpc_argp = &data->args;
10096         msg.rpc_resp = &data->res;
10097         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10098         task = rpc_run_task(&task_setup);
10099         if (IS_ERR(task))
10100                 return PTR_ERR(task);
10101         rpc_put_task(task);
10102         return 0;
10103 }
10104
10105 static void
10106 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10107 {
10108         const struct cred *cred = lsp->ls_state->owner->so_cred;
10109
10110         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10111         nfs4_free_lock_state(server, lsp);
10112 }
10113
10114 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10115                 const nfs4_stateid *s2)
10116 {
10117         if (s1->type != s2->type)
10118                 return false;
10119
10120         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10121                 return false;
10122
10123         if (s1->seqid == s2->seqid)
10124                 return true;
10125
10126         return s1->seqid == 0 || s2->seqid == 0;
10127 }
10128
10129 #endif /* CONFIG_NFS_V4_1 */
10130
10131 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10132                 const nfs4_stateid *s2)
10133 {
10134         return nfs4_stateid_match(s1, s2);
10135 }
10136
10137
10138 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10139         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10140         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10141         .recover_open   = nfs4_open_reclaim,
10142         .recover_lock   = nfs4_lock_reclaim,
10143         .establish_clid = nfs4_init_clientid,
10144         .detect_trunking = nfs40_discover_server_trunking,
10145 };
10146
10147 #if defined(CONFIG_NFS_V4_1)
10148 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10149         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10150         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10151         .recover_open   = nfs4_open_reclaim,
10152         .recover_lock   = nfs4_lock_reclaim,
10153         .establish_clid = nfs41_init_clientid,
10154         .reclaim_complete = nfs41_proc_reclaim_complete,
10155         .detect_trunking = nfs41_discover_server_trunking,
10156 };
10157 #endif /* CONFIG_NFS_V4_1 */
10158
10159 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10160         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10161         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10162         .recover_open   = nfs40_open_expired,
10163         .recover_lock   = nfs4_lock_expired,
10164         .establish_clid = nfs4_init_clientid,
10165 };
10166
10167 #if defined(CONFIG_NFS_V4_1)
10168 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10169         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10170         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10171         .recover_open   = nfs41_open_expired,
10172         .recover_lock   = nfs41_lock_expired,
10173         .establish_clid = nfs41_init_clientid,
10174 };
10175 #endif /* CONFIG_NFS_V4_1 */
10176
10177 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10178         .sched_state_renewal = nfs4_proc_async_renew,
10179         .get_state_renewal_cred = nfs4_get_renew_cred,
10180         .renew_lease = nfs4_proc_renew,
10181 };
10182
10183 #if defined(CONFIG_NFS_V4_1)
10184 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10185         .sched_state_renewal = nfs41_proc_async_sequence,
10186         .get_state_renewal_cred = nfs4_get_machine_cred,
10187         .renew_lease = nfs4_proc_sequence,
10188 };
10189 #endif
10190
10191 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10192         .get_locations = _nfs40_proc_get_locations,
10193         .fsid_present = _nfs40_proc_fsid_present,
10194 };
10195
10196 #if defined(CONFIG_NFS_V4_1)
10197 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10198         .get_locations = _nfs41_proc_get_locations,
10199         .fsid_present = _nfs41_proc_fsid_present,
10200 };
10201 #endif  /* CONFIG_NFS_V4_1 */
10202
10203 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10204         .minor_version = 0,
10205         .init_caps = NFS_CAP_READDIRPLUS
10206                 | NFS_CAP_ATOMIC_OPEN
10207                 | NFS_CAP_POSIX_LOCK,
10208         .init_client = nfs40_init_client,
10209         .shutdown_client = nfs40_shutdown_client,
10210         .match_stateid = nfs4_match_stateid,
10211         .find_root_sec = nfs4_find_root_sec,
10212         .free_lock_state = nfs4_release_lockowner,
10213         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10214         .alloc_seqid = nfs_alloc_seqid,
10215         .call_sync_ops = &nfs40_call_sync_ops,
10216         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10217         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10218         .state_renewal_ops = &nfs40_state_renewal_ops,
10219         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10220 };
10221
10222 #if defined(CONFIG_NFS_V4_1)
10223 static struct nfs_seqid *
10224 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10225 {
10226         return NULL;
10227 }
10228
10229 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10230         .minor_version = 1,
10231         .init_caps = NFS_CAP_READDIRPLUS
10232                 | NFS_CAP_ATOMIC_OPEN
10233                 | NFS_CAP_POSIX_LOCK
10234                 | NFS_CAP_STATEID_NFSV41
10235                 | NFS_CAP_ATOMIC_OPEN_V1
10236                 | NFS_CAP_LGOPEN,
10237         .init_client = nfs41_init_client,
10238         .shutdown_client = nfs41_shutdown_client,
10239         .match_stateid = nfs41_match_stateid,
10240         .find_root_sec = nfs41_find_root_sec,
10241         .free_lock_state = nfs41_free_lock_state,
10242         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10243         .alloc_seqid = nfs_alloc_no_seqid,
10244         .session_trunk = nfs4_test_session_trunk,
10245         .call_sync_ops = &nfs41_call_sync_ops,
10246         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10247         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10248         .state_renewal_ops = &nfs41_state_renewal_ops,
10249         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10250 };
10251 #endif
10252
10253 #if defined(CONFIG_NFS_V4_2)
10254 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10255         .minor_version = 2,
10256         .init_caps = NFS_CAP_READDIRPLUS
10257                 | NFS_CAP_ATOMIC_OPEN
10258                 | NFS_CAP_POSIX_LOCK
10259                 | NFS_CAP_STATEID_NFSV41
10260                 | NFS_CAP_ATOMIC_OPEN_V1
10261                 | NFS_CAP_LGOPEN
10262                 | NFS_CAP_ALLOCATE
10263                 | NFS_CAP_COPY
10264                 | NFS_CAP_OFFLOAD_CANCEL
10265                 | NFS_CAP_COPY_NOTIFY
10266                 | NFS_CAP_DEALLOCATE
10267                 | NFS_CAP_SEEK
10268                 | NFS_CAP_LAYOUTSTATS
10269                 | NFS_CAP_CLONE
10270                 | NFS_CAP_LAYOUTERROR
10271                 | NFS_CAP_READ_PLUS,
10272         .init_client = nfs41_init_client,
10273         .shutdown_client = nfs41_shutdown_client,
10274         .match_stateid = nfs41_match_stateid,
10275         .find_root_sec = nfs41_find_root_sec,
10276         .free_lock_state = nfs41_free_lock_state,
10277         .call_sync_ops = &nfs41_call_sync_ops,
10278         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10279         .alloc_seqid = nfs_alloc_no_seqid,
10280         .session_trunk = nfs4_test_session_trunk,
10281         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10282         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10283         .state_renewal_ops = &nfs41_state_renewal_ops,
10284         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10285 };
10286 #endif
10287
10288 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10289         [0] = &nfs_v4_0_minor_ops,
10290 #if defined(CONFIG_NFS_V4_1)
10291         [1] = &nfs_v4_1_minor_ops,
10292 #endif
10293 #if defined(CONFIG_NFS_V4_2)
10294         [2] = &nfs_v4_2_minor_ops,
10295 #endif
10296 };
10297
10298 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10299 {
10300         ssize_t error, error2, error3;
10301
10302         error = generic_listxattr(dentry, list, size);
10303         if (error < 0)
10304                 return error;
10305         if (list) {
10306                 list += error;
10307                 size -= error;
10308         }
10309
10310         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10311         if (error2 < 0)
10312                 return error2;
10313
10314         if (list) {
10315                 list += error2;
10316                 size -= error2;
10317         }
10318
10319         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10320         if (error3 < 0)
10321                 return error3;
10322
10323         return error + error2 + error3;
10324 }
10325
10326 static const struct inode_operations nfs4_dir_inode_operations = {
10327         .create         = nfs_create,
10328         .lookup         = nfs_lookup,
10329         .atomic_open    = nfs_atomic_open,
10330         .link           = nfs_link,
10331         .unlink         = nfs_unlink,
10332         .symlink        = nfs_symlink,
10333         .mkdir          = nfs_mkdir,
10334         .rmdir          = nfs_rmdir,
10335         .mknod          = nfs_mknod,
10336         .rename         = nfs_rename,
10337         .permission     = nfs_permission,
10338         .getattr        = nfs_getattr,
10339         .setattr        = nfs_setattr,
10340         .listxattr      = nfs4_listxattr,
10341 };
10342
10343 static const struct inode_operations nfs4_file_inode_operations = {
10344         .permission     = nfs_permission,
10345         .getattr        = nfs_getattr,
10346         .setattr        = nfs_setattr,
10347         .listxattr      = nfs4_listxattr,
10348 };
10349
10350 const struct nfs_rpc_ops nfs_v4_clientops = {
10351         .version        = 4,                    /* protocol version */
10352         .dentry_ops     = &nfs4_dentry_operations,
10353         .dir_inode_ops  = &nfs4_dir_inode_operations,
10354         .file_inode_ops = &nfs4_file_inode_operations,
10355         .file_ops       = &nfs4_file_operations,
10356         .getroot        = nfs4_proc_get_root,
10357         .submount       = nfs4_submount,
10358         .try_get_tree   = nfs4_try_get_tree,
10359         .getattr        = nfs4_proc_getattr,
10360         .setattr        = nfs4_proc_setattr,
10361         .lookup         = nfs4_proc_lookup,
10362         .lookupp        = nfs4_proc_lookupp,
10363         .access         = nfs4_proc_access,
10364         .readlink       = nfs4_proc_readlink,
10365         .create         = nfs4_proc_create,
10366         .remove         = nfs4_proc_remove,
10367         .unlink_setup   = nfs4_proc_unlink_setup,
10368         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10369         .unlink_done    = nfs4_proc_unlink_done,
10370         .rename_setup   = nfs4_proc_rename_setup,
10371         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10372         .rename_done    = nfs4_proc_rename_done,
10373         .link           = nfs4_proc_link,
10374         .symlink        = nfs4_proc_symlink,
10375         .mkdir          = nfs4_proc_mkdir,
10376         .rmdir          = nfs4_proc_rmdir,
10377         .readdir        = nfs4_proc_readdir,
10378         .mknod          = nfs4_proc_mknod,
10379         .statfs         = nfs4_proc_statfs,
10380         .fsinfo         = nfs4_proc_fsinfo,
10381         .pathconf       = nfs4_proc_pathconf,
10382         .set_capabilities = nfs4_server_capabilities,
10383         .decode_dirent  = nfs4_decode_dirent,
10384         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10385         .read_setup     = nfs4_proc_read_setup,
10386         .read_done      = nfs4_read_done,
10387         .write_setup    = nfs4_proc_write_setup,
10388         .write_done     = nfs4_write_done,
10389         .commit_setup   = nfs4_proc_commit_setup,
10390         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10391         .commit_done    = nfs4_commit_done,
10392         .lock           = nfs4_proc_lock,
10393         .clear_acl_cache = nfs4_zap_acl_attr,
10394         .close_context  = nfs4_close_context,
10395         .open_context   = nfs4_atomic_open,
10396         .have_delegation = nfs4_have_delegation,
10397         .alloc_client   = nfs4_alloc_client,
10398         .init_client    = nfs4_init_client,
10399         .free_client    = nfs4_free_client,
10400         .create_server  = nfs4_create_server,
10401         .clone_server   = nfs_clone_server,
10402 };
10403
10404 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10405         .name   = XATTR_NAME_NFSV4_ACL,
10406         .list   = nfs4_xattr_list_nfs4_acl,
10407         .get    = nfs4_xattr_get_nfs4_acl,
10408         .set    = nfs4_xattr_set_nfs4_acl,
10409 };
10410
10411 #ifdef CONFIG_NFS_V4_2
10412 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10413         .prefix = XATTR_USER_PREFIX,
10414         .get    = nfs4_xattr_get_nfs4_user,
10415         .set    = nfs4_xattr_set_nfs4_user,
10416 };
10417 #endif
10418
10419 const struct xattr_handler *nfs4_xattr_handlers[] = {
10420         &nfs4_xattr_nfs4_acl_handler,
10421 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10422         &nfs4_xattr_nfs4_label_handler,
10423 #endif
10424 #ifdef CONFIG_NFS_V4_2
10425         &nfs4_xattr_nfs4_user_handler,
10426 #endif
10427         NULL
10428 };
10429
10430 /*
10431  * Local variables:
10432  *  c-basic-offset: 8
10433  * End:
10434  */