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 #define NFSDBG_FACILITY         NFSDBG_PROC
  75
  76 #define NFS4_BITMASK_SZ         3
  77
  78 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  79 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  80
  81 /* file attributes which can be mapped to nfs attributes */
  82 #define NFS4_VALID_ATTRS (ATTR_MODE \
  83         | ATTR_UID \
  84         | ATTR_GID \
  85         | ATTR_SIZE \
  86         | ATTR_ATIME \
  87         | ATTR_MTIME \
  88         | ATTR_CTIME \
  89         | ATTR_ATIME_SET \
  90         | ATTR_MTIME_SET)
  91
  92 struct nfs4_opendata;
  93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fattr *fattr, struct nfs4_label *label, struct inode *inode);
  97 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  98                             struct nfs_fattr *fattr, struct iattr *sattr,
  99                             struct nfs_open_context *ctx, struct nfs4_label *ilabel,
 100                             struct nfs4_label *olabel);
 101 #ifdef CONFIG_NFS_V4_1
 102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 103                 const struct cred *cred,
 104                 struct nfs4_slot *slot,
 105                 bool is_privileged);
 106 static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
 107                 const struct cred *);
 108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 109                 const struct cred *, bool);
 110 #endif
 111 static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode,
 112                 struct nfs_server *server,
 113                 struct nfs4_label *label);
 114
 115 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 116 static inline struct nfs4_label *
 117 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 118         struct iattr *sattr, struct nfs4_label *label)
 119 {
 120         int err;
 121
 122         if (label == NULL)
 123                 return NULL;
 124
 125         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 126                 return NULL;
 127
 128         err = security_dentry_init_security(dentry, sattr->ia_mode,
 129                                 &dentry->d_name, (void **)&label->label, &label->len);
 130         if (err == 0)
 131                 return label;
 132
 133         return NULL;
 134 }
 135 static inline void
 136 nfs4_label_release_security(struct nfs4_label *label)
 137 {
 138         if (label)
 139                 security_release_secctx(label->label, label->len);
 140 }
 141 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 142 {
 143         if (label)
 144                 return server->attr_bitmask;
 145
 146         return server->attr_bitmask_nl;
 147 }
 148 #else
 149 static inline struct nfs4_label *
 150 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 151         struct iattr *sattr, struct nfs4_label *l)
 152 { return NULL; }
 153 static inline void
 154 nfs4_label_release_security(struct nfs4_label *label)
 155 { return; }
 156 static inline u32 *
 157 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 158 { return server->attr_bitmask; }
 159 #endif
 160
 161 /* Prevent leaks of NFSv4 errors into userland */
 162 static int nfs4_map_errors(int err)
 163 {
 164         if (err >= -1000)
 165                 return err;
 166         switch (err) {
 167         case -NFS4ERR_RESOURCE:
 168         case -NFS4ERR_LAYOUTTRYLATER:
 169         case -NFS4ERR_RECALLCONFLICT:
 170                 return -EREMOTEIO;
 171         case -NFS4ERR_WRONGSEC:
 172         case -NFS4ERR_WRONG_CRED:
 173                 return -EPERM;
 174         case -NFS4ERR_BADOWNER:
 175         case -NFS4ERR_BADNAME:
 176                 return -EINVAL;
 177         case -NFS4ERR_SHARE_DENIED:
 178                 return -EACCES;
 179         case -NFS4ERR_MINOR_VERS_MISMATCH:
 180                 return -EPROTONOSUPPORT;
 181         case -NFS4ERR_FILE_OPEN:
 182                 return -EBUSY;
 183         case -NFS4ERR_NOT_SAME:
 184                 return -ENOTSYNC;
 185         default:
 186                 dprintk("%s could not handle NFSv4 error %d\n",
 187                                 __func__, -err);
 188                 break;
 189         }
 190         return -EIO;
 191 }
 192
 193 /*
 194  * This is our standard bitmap for GETATTR requests.
 195  */
 196 const u32 nfs4_fattr_bitmap[3] = {
 197         FATTR4_WORD0_TYPE
 198         | FATTR4_WORD0_CHANGE
 199         | FATTR4_WORD0_SIZE
 200         | FATTR4_WORD0_FSID
 201         | FATTR4_WORD0_FILEID,
 202         FATTR4_WORD1_MODE
 203         | FATTR4_WORD1_NUMLINKS
 204         | FATTR4_WORD1_OWNER
 205         | FATTR4_WORD1_OWNER_GROUP
 206         | FATTR4_WORD1_RAWDEV
 207         | FATTR4_WORD1_SPACE_USED
 208         | FATTR4_WORD1_TIME_ACCESS
 209         | FATTR4_WORD1_TIME_METADATA
 210         | FATTR4_WORD1_TIME_MODIFY
 211         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 212 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 213         FATTR4_WORD2_SECURITY_LABEL
 214 #endif
 215 };
 216
 217 static const u32 nfs4_pnfs_open_bitmap[3] = {
 218         FATTR4_WORD0_TYPE
 219         | FATTR4_WORD0_CHANGE
 220         | FATTR4_WORD0_SIZE
 221         | FATTR4_WORD0_FSID
 222         | FATTR4_WORD0_FILEID,
 223         FATTR4_WORD1_MODE
 224         | FATTR4_WORD1_NUMLINKS
 225         | FATTR4_WORD1_OWNER
 226         | FATTR4_WORD1_OWNER_GROUP
 227         | FATTR4_WORD1_RAWDEV
 228         | FATTR4_WORD1_SPACE_USED
 229         | FATTR4_WORD1_TIME_ACCESS
 230         | FATTR4_WORD1_TIME_METADATA
 231         | FATTR4_WORD1_TIME_MODIFY,
 232         FATTR4_WORD2_MDSTHRESHOLD
 233 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 234         | FATTR4_WORD2_SECURITY_LABEL
 235 #endif
 236 };
 237
 238 static const u32 nfs4_open_noattr_bitmap[3] = {
 239         FATTR4_WORD0_TYPE
 240         | FATTR4_WORD0_FILEID,
 241 };
 242
 243 const u32 nfs4_statfs_bitmap[3] = {
 244         FATTR4_WORD0_FILES_AVAIL
 245         | FATTR4_WORD0_FILES_FREE
 246         | FATTR4_WORD0_FILES_TOTAL,
 247         FATTR4_WORD1_SPACE_AVAIL
 248         | FATTR4_WORD1_SPACE_FREE
 249         | FATTR4_WORD1_SPACE_TOTAL
 250 };
 251
 252 const u32 nfs4_pathconf_bitmap[3] = {
 253         FATTR4_WORD0_MAXLINK
 254         | FATTR4_WORD0_MAXNAME,
 255         0
 256 };
 257
 258 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 259                         | FATTR4_WORD0_MAXREAD
 260                         | FATTR4_WORD0_MAXWRITE
 261                         | FATTR4_WORD0_LEASE_TIME,
 262                         FATTR4_WORD1_TIME_DELTA
 263                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 264                         FATTR4_WORD2_LAYOUT_BLKSIZE
 265                         | FATTR4_WORD2_CLONE_BLKSIZE
 266                         | FATTR4_WORD2_XATTR_SUPPORT
 267 };
 268
 269 const u32 nfs4_fs_locations_bitmap[3] = {
 270         FATTR4_WORD0_CHANGE
 271         | FATTR4_WORD0_SIZE
 272         | FATTR4_WORD0_FSID
 273         | FATTR4_WORD0_FILEID
 274         | FATTR4_WORD0_FS_LOCATIONS,
 275         FATTR4_WORD1_OWNER
 276         | FATTR4_WORD1_OWNER_GROUP
 277         | FATTR4_WORD1_RAWDEV
 278         | FATTR4_WORD1_SPACE_USED
 279         | FATTR4_WORD1_TIME_ACCESS
 280         | FATTR4_WORD1_TIME_METADATA
 281         | FATTR4_WORD1_TIME_MODIFY
 282         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 283 };
 284
 285 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 286                 struct inode *inode)
 287 {
 288         unsigned long cache_validity;
 289
 290         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 291         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 292                 return;
 293
 294         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
 295         if (!(cache_validity & NFS_INO_REVAL_FORCED))
 296                 cache_validity &= ~(NFS_INO_INVALID_CHANGE
 297                                 | NFS_INO_INVALID_SIZE);
 298
 299         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 300                 dst[0] &= ~FATTR4_WORD0_SIZE;
 301
 302         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 303                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 304 }
 305
 306 static void nfs4_bitmap_copy_adjust_setattr(__u32 *dst,
 307                 const __u32 *src, struct inode *inode)
 308 {
 309         nfs4_bitmap_copy_adjust(dst, src, inode);
 310 }
 311
 312 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 313                 struct nfs4_readdir_arg *readdir)
 314 {
 315         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 316         __be32 *start, *p;
 317
 318         if (cookie > 2) {
 319                 readdir->cookie = cookie;
 320                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 321                 return;
 322         }
 323
 324         readdir->cookie = 0;
 325         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 326         if (cookie == 2)
 327                 return;
 328
 329         /*
 330          * NFSv4 servers do not return entries for '.' and '..'
 331          * Therefore, we fake these entries here.  We let '.'
 332          * have cookie 0 and '..' have cookie 1.  Note that
 333          * when talking to the server, we always send cookie 0
 334          * instead of 1 or 2.
 335          */
 336         start = p = kmap_atomic(*readdir->pages);
 337
 338         if (cookie == 0) {
 339                 *p++ = xdr_one;                                  /* next */
 340                 *p++ = xdr_zero;                   /* cookie, first word */
 341                 *p++ = xdr_one;                   /* cookie, second word */
 342                 *p++ = xdr_one;                             /* entry len */
 343                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 344                 p++;
 345                 *p++ = xdr_one;                         /* bitmap length */
 346                 *p++ = htonl(attrs);                           /* bitmap */
 347                 *p++ = htonl(12);             /* attribute buffer length */
 348                 *p++ = htonl(NF4DIR);
 349                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 350         }
 351
 352         *p++ = xdr_one;                                  /* next */
 353         *p++ = xdr_zero;                   /* cookie, first word */
 354         *p++ = xdr_two;                   /* cookie, second word */
 355         *p++ = xdr_two;                             /* entry len */
 356         memcpy(p, "..\0\0", 4);                         /* entry */
 357         p++;
 358         *p++ = xdr_one;                         /* bitmap length */
 359         *p++ = htonl(attrs);                           /* bitmap */
 360         *p++ = htonl(12);             /* attribute buffer length */
 361         *p++ = htonl(NF4DIR);
 362         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 363
 364         readdir->pgbase = (char *)p - (char *)start;
 365         readdir->count -= readdir->pgbase;
 366         kunmap_atomic(start);
 367 }
 368
 369 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 370                 nfs4_stateid *stateid,
 371                 const struct cred *cred)
 372 {
 373         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 374
 375         ops->test_and_free_expired(server, stateid, cred);
 376 }
 377
 378 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 379                 nfs4_stateid *stateid,
 380                 const struct cred *cred)
 381 {
 382         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 383         nfs4_test_and_free_stateid(server, stateid, cred);
 384 }
 385
 386 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 387                 const nfs4_stateid *stateid,
 388                 const struct cred *cred)
 389 {
 390         nfs4_stateid tmp;
 391
 392         nfs4_stateid_copy(&tmp, stateid);
 393         __nfs4_free_revoked_stateid(server, &tmp, cred);
 394 }
 395
 396 static long nfs4_update_delay(long *timeout)
 397 {
 398         long ret;
 399         if (!timeout)
 400                 return NFS4_POLL_RETRY_MAX;
 401         if (*timeout <= 0)
 402                 *timeout = NFS4_POLL_RETRY_MIN;
 403         if (*timeout > NFS4_POLL_RETRY_MAX)
 404                 *timeout = NFS4_POLL_RETRY_MAX;
 405         ret = *timeout;
 406         *timeout <<= 1;
 407         return ret;
 408 }
 409
 410 static int nfs4_delay_killable(long *timeout)
 411 {
 412         might_sleep();
 413
 414         freezable_schedule_timeout_killable_unsafe(
 415                 nfs4_update_delay(timeout));
 416         if (!__fatal_signal_pending(current))
 417                 return 0;
 418         return -EINTR;
 419 }
 420
 421 static int nfs4_delay_interruptible(long *timeout)
 422 {
 423         might_sleep();
 424
 425         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
 426         if (!signal_pending(current))
 427                 return 0;
 428         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 429 }
 430
 431 static int nfs4_delay(long *timeout, bool interruptible)
 432 {
 433         if (interruptible)
 434                 return nfs4_delay_interruptible(timeout);
 435         return nfs4_delay_killable(timeout);
 436 }
 437
 438 static const nfs4_stateid *
 439 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 440 {
 441         if (!stateid)
 442                 return NULL;
 443         switch (stateid->type) {
 444         case NFS4_OPEN_STATEID_TYPE:
 445         case NFS4_LOCK_STATEID_TYPE:
 446         case NFS4_DELEGATION_STATEID_TYPE:
 447                 return stateid;
 448         default:
 449                 break;
 450         }
 451         return NULL;
 452 }
 453
 454 /* This is the error handling routine for processes that are allowed
 455  * to sleep.
 456  */
 457 static int nfs4_do_handle_exception(struct nfs_server *server,
 458                 int errorcode, struct nfs4_exception *exception)
 459 {
 460         struct nfs_client *clp = server->nfs_client;
 461         struct nfs4_state *state = exception->state;
 462         const nfs4_stateid *stateid;
 463         struct inode *inode = exception->inode;
 464         int ret = errorcode;
 465
 466         exception->delay = 0;
 467         exception->recovering = 0;
 468         exception->retry = 0;
 469
 470         stateid = nfs4_recoverable_stateid(exception->stateid);
 471         if (stateid == NULL && state != NULL)
 472                 stateid = nfs4_recoverable_stateid(&state->stateid);
 473
 474         switch(errorcode) {
 475                 case 0:
 476                         return 0;
 477                 case -NFS4ERR_BADHANDLE:
 478                 case -ESTALE:
 479                         if (inode != NULL && S_ISREG(inode->i_mode))
 480                                 pnfs_destroy_layout(NFS_I(inode));
 481                         break;
 482                 case -NFS4ERR_DELEG_REVOKED:
 483                 case -NFS4ERR_ADMIN_REVOKED:
 484                 case -NFS4ERR_EXPIRED:
 485                 case -NFS4ERR_BAD_STATEID:
 486                 case -NFS4ERR_PARTNER_NO_AUTH:
 487                         if (inode != NULL && stateid != NULL) {
 488                                 nfs_inode_find_state_and_recover(inode,
 489                                                 stateid);
 490                                 goto wait_on_recovery;
 491                         }
 492                         fallthrough;
 493                 case -NFS4ERR_OPENMODE:
 494                         if (inode) {
 495                                 int err;
 496
 497                                 err = nfs_async_inode_return_delegation(inode,
 498                                                 stateid);
 499                                 if (err == 0)
 500                                         goto wait_on_recovery;
 501                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 502                                         exception->retry = 1;
 503                                         break;
 504                                 }
 505                         }
 506                         if (state == NULL)
 507                                 break;
 508                         ret = nfs4_schedule_stateid_recovery(server, state);
 509                         if (ret < 0)
 510                                 break;
 511                         goto wait_on_recovery;
 512                 case -NFS4ERR_STALE_STATEID:
 513                 case -NFS4ERR_STALE_CLIENTID:
 514                         nfs4_schedule_lease_recovery(clp);
 515                         goto wait_on_recovery;
 516                 case -NFS4ERR_MOVED:
 517                         ret = nfs4_schedule_migration_recovery(server);
 518                         if (ret < 0)
 519                                 break;
 520                         goto wait_on_recovery;
 521                 case -NFS4ERR_LEASE_MOVED:
 522                         nfs4_schedule_lease_moved_recovery(clp);
 523                         goto wait_on_recovery;
 524 #if defined(CONFIG_NFS_V4_1)
 525                 case -NFS4ERR_BADSESSION:
 526                 case -NFS4ERR_BADSLOT:
 527                 case -NFS4ERR_BAD_HIGH_SLOT:
 528                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 529                 case -NFS4ERR_DEADSESSION:
 530                 case -NFS4ERR_SEQ_FALSE_RETRY:
 531                 case -NFS4ERR_SEQ_MISORDERED:
 532                         /* Handled in nfs41_sequence_process() */
 533                         goto wait_on_recovery;
 534 #endif /* defined(CONFIG_NFS_V4_1) */
 535                 case -NFS4ERR_FILE_OPEN:
 536                         if (exception->timeout > HZ) {
 537                                 /* We have retried a decent amount, time to
 538                                  * fail
 539                                  */
 540                                 ret = -EBUSY;
 541                                 break;
 542                         }
 543                         fallthrough;
 544                 case -NFS4ERR_DELAY:
 545                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 546                         fallthrough;
 547                 case -NFS4ERR_GRACE:
 548                 case -NFS4ERR_LAYOUTTRYLATER:
 549                 case -NFS4ERR_RECALLCONFLICT:
 550                         exception->delay = 1;
 551                         return 0;
 552
 553                 case -NFS4ERR_RETRY_UNCACHED_REP:
 554                 case -NFS4ERR_OLD_STATEID:
 555                         exception->retry = 1;
 556                         break;
 557                 case -NFS4ERR_BADOWNER:
 558                         /* The following works around a Linux server bug! */
 559                 case -NFS4ERR_BADNAME:
 560                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 561                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 562                                 exception->retry = 1;
 563                                 printk(KERN_WARNING "NFS: v4 server %s "
 564                                                 "does not accept raw "
 565                                                 "uid/gids. "
 566                                                 "Reenabling the idmapper.\n",
 567                                                 server->nfs_client->cl_hostname);
 568                         }
 569         }
 570         /* We failed to handle the error */
 571         return nfs4_map_errors(ret);
 572 wait_on_recovery:
 573         exception->recovering = 1;
 574         return 0;
 575 }
 576
 577 /* This is the error handling routine for processes that are allowed
 578  * to sleep.
 579  */
 580 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 581 {
 582         struct nfs_client *clp = server->nfs_client;
 583         int ret;
 584
 585         ret = nfs4_do_handle_exception(server, errorcode, exception);
 586         if (exception->delay) {
 587                 ret = nfs4_delay(&exception->timeout,
 588                                 exception->interruptible);
 589                 goto out_retry;
 590         }
 591         if (exception->recovering) {
 592                 ret = nfs4_wait_clnt_recover(clp);
 593                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 594                         return -EIO;
 595                 goto out_retry;
 596         }
 597         return ret;
 598 out_retry:
 599         if (ret == 0)
 600                 exception->retry = 1;
 601         return ret;
 602 }
 603
 604 static int
 605 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 606                 int errorcode, struct nfs4_exception *exception)
 607 {
 608         struct nfs_client *clp = server->nfs_client;
 609         int ret;
 610
 611         ret = nfs4_do_handle_exception(server, errorcode, exception);
 612         if (exception->delay) {
 613                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 614                 goto out_retry;
 615         }
 616         if (exception->recovering) {
 617                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 618                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 619                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 620                 goto out_retry;
 621         }
 622         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 623                 ret = -EIO;
 624         return ret;
 625 out_retry:
 626         if (ret == 0) {
 627                 exception->retry = 1;
 628                 /*
 629                  * For NFS4ERR_MOVED, the client transport will need to
 630                  * be recomputed after migration recovery has completed.
 631                  */
 632                 if (errorcode == -NFS4ERR_MOVED)
 633                         rpc_task_release_transport(task);
 634         }
 635         return ret;
 636 }
 637
 638 int
 639 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 640                         struct nfs4_state *state, long *timeout)
 641 {
 642         struct nfs4_exception exception = {
 643                 .state = state,
 644         };
 645
 646         if (task->tk_status >= 0)
 647                 return 0;
 648         if (timeout)
 649                 exception.timeout = *timeout;
 650         task->tk_status = nfs4_async_handle_exception(task, server,
 651                         task->tk_status,
 652                         &exception);
 653         if (exception.delay && timeout)
 654                 *timeout = exception.timeout;
 655         if (exception.retry)
 656                 return -EAGAIN;
 657         return 0;
 658 }
 659
 660 /*
 661  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 662  * or 'false' otherwise.
 663  */
 664 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 665 {
 666         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 667         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 668 }
 669
 670 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 671 {
 672         spin_lock(&clp->cl_lock);
 673         if (time_before(clp->cl_last_renewal,timestamp))
 674                 clp->cl_last_renewal = timestamp;
 675         spin_unlock(&clp->cl_lock);
 676 }
 677
 678 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 679 {
 680         struct nfs_client *clp = server->nfs_client;
 681
 682         if (!nfs4_has_session(clp))
 683                 do_renew_lease(clp, timestamp);
 684 }
 685
 686 struct nfs4_call_sync_data {
 687         const struct nfs_server *seq_server;
 688         struct nfs4_sequence_args *seq_args;
 689         struct nfs4_sequence_res *seq_res;
 690 };
 691
 692 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 693                         struct nfs4_sequence_res *res, int cache_reply,
 694                         int privileged)
 695 {
 696         args->sa_slot = NULL;
 697         args->sa_cache_this = cache_reply;
 698         args->sa_privileged = privileged;
 699
 700         res->sr_slot = NULL;
 701 }
 702
 703 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 704 {
 705         struct nfs4_slot *slot = res->sr_slot;
 706         struct nfs4_slot_table *tbl;
 707
 708         tbl = slot->table;
 709         spin_lock(&tbl->slot_tbl_lock);
 710         if (!nfs41_wake_and_assign_slot(tbl, slot))
 711                 nfs4_free_slot(tbl, slot);
 712         spin_unlock(&tbl->slot_tbl_lock);
 713
 714         res->sr_slot = NULL;
 715 }
 716
 717 static int nfs40_sequence_done(struct rpc_task *task,
 718                                struct nfs4_sequence_res *res)
 719 {
 720         if (res->sr_slot != NULL)
 721                 nfs40_sequence_free_slot(res);
 722         return 1;
 723 }
 724
 725 #if defined(CONFIG_NFS_V4_1)
 726
 727 static void nfs41_release_slot(struct nfs4_slot *slot)
 728 {
 729         struct nfs4_session *session;
 730         struct nfs4_slot_table *tbl;
 731         bool send_new_highest_used_slotid = false;
 732
 733         if (!slot)
 734                 return;
 735         tbl = slot->table;
 736         session = tbl->session;
 737
 738         /* Bump the slot sequence number */
 739         if (slot->seq_done)
 740                 slot->seq_nr++;
 741         slot->seq_done = 0;
 742
 743         spin_lock(&tbl->slot_tbl_lock);
 744         /* Be nice to the server: try to ensure that the last transmitted
 745          * value for highest_user_slotid <= target_highest_slotid
 746          */
 747         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 748                 send_new_highest_used_slotid = true;
 749
 750         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 751                 send_new_highest_used_slotid = false;
 752                 goto out_unlock;
 753         }
 754         nfs4_free_slot(tbl, slot);
 755
 756         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 757                 send_new_highest_used_slotid = false;
 758 out_unlock:
 759         spin_unlock(&tbl->slot_tbl_lock);
 760         if (send_new_highest_used_slotid)
 761                 nfs41_notify_server(session->clp);
 762         if (waitqueue_active(&tbl->slot_waitq))
 763                 wake_up_all(&tbl->slot_waitq);
 764 }
 765
 766 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 767 {
 768         nfs41_release_slot(res->sr_slot);
 769         res->sr_slot = NULL;
 770 }
 771
 772 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 773                 u32 seqnr)
 774 {
 775         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 776                 slot->seq_nr_highest_sent = seqnr;
 777 }
 778 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 779                 u32 seqnr)
 780 {
 781         slot->seq_nr_highest_sent = seqnr;
 782         slot->seq_nr_last_acked = seqnr;
 783 }
 784
 785 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 786                                 struct nfs4_slot *slot)
 787 {
 788         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 789         if (!IS_ERR(task))
 790                 rpc_put_task_async(task);
 791 }
 792
 793 static int nfs41_sequence_process(struct rpc_task *task,
 794                 struct nfs4_sequence_res *res)
 795 {
 796         struct nfs4_session *session;
 797         struct nfs4_slot *slot = res->sr_slot;
 798         struct nfs_client *clp;
 799         int status;
 800         int ret = 1;
 801
 802         if (slot == NULL)
 803                 goto out_noaction;
 804         /* don't increment the sequence number if the task wasn't sent */
 805         if (!RPC_WAS_SENT(task) || slot->seq_done)
 806                 goto out;
 807
 808         session = slot->table->session;
 809         clp = session->clp;
 810
 811         trace_nfs4_sequence_done(session, res);
 812
 813         status = res->sr_status;
 814         if (task->tk_status == -NFS4ERR_DEADSESSION)
 815                 status = -NFS4ERR_DEADSESSION;
 816
 817         /* Check the SEQUENCE operation status */
 818         switch (status) {
 819         case 0:
 820                 /* Mark this sequence number as having been acked */
 821                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 822                 /* Update the slot's sequence and clientid lease timer */
 823                 slot->seq_done = 1;
 824                 do_renew_lease(clp, res->sr_timestamp);
 825                 /* Check sequence flags */
 826                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 827                                 !!slot->privileged);
 828                 nfs41_update_target_slotid(slot->table, slot, res);
 829                 break;
 830         case 1:
 831                 /*
 832                  * sr_status remains 1 if an RPC level error occurred.
 833                  * The server may or may not have processed the sequence
 834                  * operation..
 835                  */
 836                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 837                 slot->seq_done = 1;
 838                 goto out;
 839         case -NFS4ERR_DELAY:
 840                 /* The server detected a resend of the RPC call and
 841                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 842                  * of RFC5661.
 843                  */
 844                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 845                         __func__,
 846                         slot->slot_nr,
 847                         slot->seq_nr);
 848                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 849                 goto out_retry;
 850         case -NFS4ERR_RETRY_UNCACHED_REP:
 851         case -NFS4ERR_SEQ_FALSE_RETRY:
 852                 /*
 853                  * The server thinks we tried to replay a request.
 854                  * Retry the call after bumping the sequence ID.
 855                  */
 856                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 857                 goto retry_new_seq;
 858         case -NFS4ERR_BADSLOT:
 859                 /*
 860                  * The slot id we used was probably retired. Try again
 861                  * using a different slot id.
 862                  */
 863                 if (slot->slot_nr < slot->table->target_highest_slotid)
 864                         goto session_recover;
 865                 goto retry_nowait;
 866         case -NFS4ERR_SEQ_MISORDERED:
 867                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 868                 /*
 869                  * Were one or more calls using this slot interrupted?
 870                  * If the server never received the request, then our
 871                  * transmitted slot sequence number may be too high. However,
 872                  * if the server did receive the request then it might
 873                  * accidentally give us a reply with a mismatched operation.
 874                  * We can sort this out by sending a lone sequence operation
 875                  * to the server on the same slot.
 876                  */
 877                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 878                         slot->seq_nr--;
 879                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 880                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 881                                 res->sr_slot = NULL;
 882                         }
 883                         goto retry_nowait;
 884                 }
 885                 /*
 886                  * RFC5661:
 887                  * A retry might be sent while the original request is
 888                  * still in progress on the replier. The replier SHOULD
 889                  * deal with the issue by returning NFS4ERR_DELAY as the
 890                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 891                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 892                  *
 893                  * Restart the search after a delay.
 894                  */
 895                 slot->seq_nr = slot->seq_nr_highest_sent;
 896                 goto out_retry;
 897         case -NFS4ERR_BADSESSION:
 898         case -NFS4ERR_DEADSESSION:
 899         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 900                 goto session_recover;
 901         default:
 902                 /* Just update the slot sequence no. */
 903                 slot->seq_done = 1;
 904         }
 905 out:
 906         /* The session may be reset by one of the error handlers. */
 907         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 908 out_noaction:
 909         return ret;
 910 session_recover:
 911         nfs4_schedule_session_recovery(session, status);
 912         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 913         nfs41_sequence_free_slot(res);
 914         goto out;
 915 retry_new_seq:
 916         ++slot->seq_nr;
 917 retry_nowait:
 918         if (rpc_restart_call_prepare(task)) {
 919                 nfs41_sequence_free_slot(res);
 920                 task->tk_status = 0;
 921                 ret = 0;
 922         }
 923         goto out;
 924 out_retry:
 925         if (!rpc_restart_call(task))
 926                 goto out;
 927         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 928         return 0;
 929 }
 930
 931 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 932 {
 933         if (!nfs41_sequence_process(task, res))
 934                 return 0;
 935         if (res->sr_slot != NULL)
 936                 nfs41_sequence_free_slot(res);
 937         return 1;
 938
 939 }
 940 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 941
 942 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 943 {
 944         if (res->sr_slot == NULL)
 945                 return 1;
 946         if (res->sr_slot->table->session != NULL)
 947                 return nfs41_sequence_process(task, res);
 948         return nfs40_sequence_done(task, res);
 949 }
 950
 951 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 952 {
 953         if (res->sr_slot != NULL) {
 954                 if (res->sr_slot->table->session != NULL)
 955                         nfs41_sequence_free_slot(res);
 956                 else
 957                         nfs40_sequence_free_slot(res);
 958         }
 959 }
 960
 961 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 962 {
 963         if (res->sr_slot == NULL)
 964                 return 1;
 965         if (!res->sr_slot->table->session)
 966                 return nfs40_sequence_done(task, res);
 967         return nfs41_sequence_done(task, res);
 968 }
 969 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 970
 971 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 972 {
 973         struct nfs4_call_sync_data *data = calldata;
 974
 975         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 976
 977         nfs4_setup_sequence(data->seq_server->nfs_client,
 978                             data->seq_args, data->seq_res, task);
 979 }
 980
 981 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 982 {
 983         struct nfs4_call_sync_data *data = calldata;
 984
 985         nfs41_sequence_done(task, data->seq_res);
 986 }
 987
 988 static const struct rpc_call_ops nfs41_call_sync_ops = {
 989         .rpc_call_prepare = nfs41_call_sync_prepare,
 990         .rpc_call_done = nfs41_call_sync_done,
 991 };
 992
 993 #else   /* !CONFIG_NFS_V4_1 */
 994
 995 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 996 {
 997         return nfs40_sequence_done(task, res);
 998 }
 999
1000 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1001 {
1002         if (res->sr_slot != NULL)
1003                 nfs40_sequence_free_slot(res);
1004 }
1005
1006 int nfs4_sequence_done(struct rpc_task *task,
1007                        struct nfs4_sequence_res *res)
1008 {
1009         return nfs40_sequence_done(task, res);
1010 }
1011 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1012
1013 #endif  /* !CONFIG_NFS_V4_1 */
1014
1015 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1016 {
1017         res->sr_timestamp = jiffies;
1018         res->sr_status_flags = 0;
1019         res->sr_status = 1;
1020 }
1021
1022 static
1023 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1024                 struct nfs4_sequence_res *res,
1025                 struct nfs4_slot *slot)
1026 {
1027         if (!slot)
1028                 return;
1029         slot->privileged = args->sa_privileged ? 1 : 0;
1030         args->sa_slot = slot;
1031
1032         res->sr_slot = slot;
1033 }
1034
1035 int nfs4_setup_sequence(struct nfs_client *client,
1036                         struct nfs4_sequence_args *args,
1037                         struct nfs4_sequence_res *res,
1038                         struct rpc_task *task)
1039 {
1040         struct nfs4_session *session = nfs4_get_session(client);
1041         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1042         struct nfs4_slot *slot;
1043
1044         /* slot already allocated? */
1045         if (res->sr_slot != NULL)
1046                 goto out_start;
1047
1048         if (session)
1049                 tbl = &session->fc_slot_table;
1050
1051         spin_lock(&tbl->slot_tbl_lock);
1052         /* The state manager will wait until the slot table is empty */
1053         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1054                 goto out_sleep;
1055
1056         slot = nfs4_alloc_slot(tbl);
1057         if (IS_ERR(slot)) {
1058                 if (slot == ERR_PTR(-ENOMEM))
1059                         goto out_sleep_timeout;
1060                 goto out_sleep;
1061         }
1062         spin_unlock(&tbl->slot_tbl_lock);
1063
1064         nfs4_sequence_attach_slot(args, res, slot);
1065
1066         trace_nfs4_setup_sequence(session, args);
1067 out_start:
1068         nfs41_sequence_res_init(res);
1069         rpc_call_start(task);
1070         return 0;
1071 out_sleep_timeout:
1072         /* Try again in 1/4 second */
1073         if (args->sa_privileged)
1074                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1075                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1076         else
1077                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1078                                 NULL, jiffies + (HZ >> 2));
1079         spin_unlock(&tbl->slot_tbl_lock);
1080         return -EAGAIN;
1081 out_sleep:
1082         if (args->sa_privileged)
1083                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1084                                 RPC_PRIORITY_PRIVILEGED);
1085         else
1086                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1087         spin_unlock(&tbl->slot_tbl_lock);
1088         return -EAGAIN;
1089 }
1090 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1091
1092 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1093 {
1094         struct nfs4_call_sync_data *data = calldata;
1095         nfs4_setup_sequence(data->seq_server->nfs_client,
1096                                 data->seq_args, data->seq_res, task);
1097 }
1098
1099 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1100 {
1101         struct nfs4_call_sync_data *data = calldata;
1102         nfs4_sequence_done(task, data->seq_res);
1103 }
1104
1105 static const struct rpc_call_ops nfs40_call_sync_ops = {
1106         .rpc_call_prepare = nfs40_call_sync_prepare,
1107         .rpc_call_done = nfs40_call_sync_done,
1108 };
1109
1110 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1111 {
1112         int ret;
1113         struct rpc_task *task;
1114
1115         task = rpc_run_task(task_setup);
1116         if (IS_ERR(task))
1117                 return PTR_ERR(task);
1118
1119         ret = task->tk_status;
1120         rpc_put_task(task);
1121         return ret;
1122 }
1123
1124 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1125                              struct nfs_server *server,
1126                              struct rpc_message *msg,
1127                              struct nfs4_sequence_args *args,
1128                              struct nfs4_sequence_res *res,
1129                              unsigned short task_flags)
1130 {
1131         struct nfs_client *clp = server->nfs_client;
1132         struct nfs4_call_sync_data data = {
1133                 .seq_server = server,
1134                 .seq_args = args,
1135                 .seq_res = res,
1136         };
1137         struct rpc_task_setup task_setup = {
1138                 .rpc_client = clnt,
1139                 .rpc_message = msg,
1140                 .callback_ops = clp->cl_mvops->call_sync_ops,
1141                 .callback_data = &data,
1142                 .flags = task_flags,
1143         };
1144
1145         return nfs4_call_sync_custom(&task_setup);
1146 }
1147
1148 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1149                                    struct nfs_server *server,
1150                                    struct rpc_message *msg,
1151                                    struct nfs4_sequence_args *args,
1152                                    struct nfs4_sequence_res *res)
1153 {
1154         return nfs4_do_call_sync(clnt, server, msg, args, res, 0);
1155 }
1156
1157
1158 int nfs4_call_sync(struct rpc_clnt *clnt,
1159                    struct nfs_server *server,
1160                    struct rpc_message *msg,
1161                    struct nfs4_sequence_args *args,
1162                    struct nfs4_sequence_res *res,
1163                    int cache_reply)
1164 {
1165         nfs4_init_sequence(args, res, cache_reply, 0);
1166         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1167 }
1168
1169 static void
1170 nfs4_inc_nlink_locked(struct inode *inode)
1171 {
1172         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1173         inc_nlink(inode);
1174 }
1175
1176 static void
1177 nfs4_dec_nlink_locked(struct inode *inode)
1178 {
1179         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_OTHER;
1180         drop_nlink(inode);
1181 }
1182
1183 static void
1184 nfs4_update_changeattr_locked(struct inode *inode,
1185                 struct nfs4_change_info *cinfo,
1186                 unsigned long timestamp, unsigned long cache_validity)
1187 {
1188         struct nfs_inode *nfsi = NFS_I(inode);
1189
1190         nfsi->cache_validity |= NFS_INO_INVALID_CTIME
1191                 | NFS_INO_INVALID_MTIME
1192                 | cache_validity;
1193
1194         if (cinfo->atomic && cinfo->before == inode_peek_iversion_raw(inode)) {
1195                 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1196                 nfsi->attrtimeo_timestamp = jiffies;
1197         } else {
1198                 if (S_ISDIR(inode->i_mode)) {
1199                         nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1200                         nfs_force_lookup_revalidate(inode);
1201                 } else {
1202                         if (!NFS_PROTO(inode)->have_delegation(inode,
1203                                                                FMODE_READ))
1204                                 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
1205                 }
1206
1207                 if (cinfo->before != inode_peek_iversion_raw(inode))
1208                         nfsi->cache_validity |= NFS_INO_INVALID_ACCESS |
1209                                                 NFS_INO_INVALID_ACL |
1210                                                 NFS_INO_INVALID_XATTR;
1211         }
1212         inode_set_iversion_raw(inode, cinfo->after);
1213         nfsi->read_cache_jiffies = timestamp;
1214         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1215         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1216
1217         if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
1218                 nfs_fscache_invalidate(inode);
1219 }
1220
1221 void
1222 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1223                 unsigned long timestamp, unsigned long cache_validity)
1224 {
1225         spin_lock(&dir->i_lock);
1226         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1227         spin_unlock(&dir->i_lock);
1228 }
1229
1230 struct nfs4_open_createattrs {
1231         struct nfs4_label *label;
1232         struct iattr *sattr;
1233         const __u32 verf[2];
1234 };
1235
1236 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1237                 int err, struct nfs4_exception *exception)
1238 {
1239         if (err != -EINVAL)
1240                 return false;
1241         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1242                 return false;
1243         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1244         exception->retry = 1;
1245         return true;
1246 }
1247
1248 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1249 {
1250          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1251 }
1252
1253 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1254 {
1255         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1256
1257         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1258 }
1259
1260 static u32
1261 nfs4_map_atomic_open_share(struct nfs_server *server,
1262                 fmode_t fmode, int openflags)
1263 {
1264         u32 res = 0;
1265
1266         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1267         case FMODE_READ:
1268                 res = NFS4_SHARE_ACCESS_READ;
1269                 break;
1270         case FMODE_WRITE:
1271                 res = NFS4_SHARE_ACCESS_WRITE;
1272                 break;
1273         case FMODE_READ|FMODE_WRITE:
1274                 res = NFS4_SHARE_ACCESS_BOTH;
1275         }
1276         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1277                 goto out;
1278         /* Want no delegation if we're using O_DIRECT */
1279         if (openflags & O_DIRECT)
1280                 res |= NFS4_SHARE_WANT_NO_DELEG;
1281 out:
1282         return res;
1283 }
1284
1285 static enum open_claim_type4
1286 nfs4_map_atomic_open_claim(struct nfs_server *server,
1287                 enum open_claim_type4 claim)
1288 {
1289         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1290                 return claim;
1291         switch (claim) {
1292         default:
1293                 return claim;
1294         case NFS4_OPEN_CLAIM_FH:
1295                 return NFS4_OPEN_CLAIM_NULL;
1296         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1297                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1298         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1299                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1300         }
1301 }
1302
1303 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1304 {
1305         p->o_res.f_attr = &p->f_attr;
1306         p->o_res.f_label = p->f_label;
1307         p->o_res.seqid = p->o_arg.seqid;
1308         p->c_res.seqid = p->c_arg.seqid;
1309         p->o_res.server = p->o_arg.server;
1310         p->o_res.access_request = p->o_arg.access;
1311         nfs_fattr_init(&p->f_attr);
1312         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1313 }
1314
1315 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1316                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1317                 const struct nfs4_open_createattrs *c,
1318                 enum open_claim_type4 claim,
1319                 gfp_t gfp_mask)
1320 {
1321         struct dentry *parent = dget_parent(dentry);
1322         struct inode *dir = d_inode(parent);
1323         struct nfs_server *server = NFS_SERVER(dir);
1324         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1325         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1326         struct nfs4_opendata *p;
1327
1328         p = kzalloc(sizeof(*p), gfp_mask);
1329         if (p == NULL)
1330                 goto err;
1331
1332         p->f_label = nfs4_label_alloc(server, gfp_mask);
1333         if (IS_ERR(p->f_label))
1334                 goto err_free_p;
1335
1336         p->a_label = nfs4_label_alloc(server, gfp_mask);
1337         if (IS_ERR(p->a_label))
1338                 goto err_free_f;
1339
1340         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1341         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1342         if (IS_ERR(p->o_arg.seqid))
1343                 goto err_free_label;
1344         nfs_sb_active(dentry->d_sb);
1345         p->dentry = dget(dentry);
1346         p->dir = parent;
1347         p->owner = sp;
1348         atomic_inc(&sp->so_count);
1349         p->o_arg.open_flags = flags;
1350         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1351         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1352         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1353                         fmode, flags);
1354         if (flags & O_CREAT) {
1355                 p->o_arg.umask = current_umask();
1356                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1357                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1358                         p->o_arg.u.attrs = &p->attrs;
1359                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1360
1361                         memcpy(p->o_arg.u.verifier.data, c->verf,
1362                                         sizeof(p->o_arg.u.verifier.data));
1363                 }
1364         }
1365         /* don't put an ACCESS op in OPEN compound if O_EXCL, because ACCESS
1366          * will return permission denied for all bits until close */
1367         if (!(flags & O_EXCL)) {
1368                 /* ask server to check for all possible rights as results
1369                  * are cached */
1370                 switch (p->o_arg.claim) {
1371                 default:
1372                         break;
1373                 case NFS4_OPEN_CLAIM_NULL:
1374                 case NFS4_OPEN_CLAIM_FH:
1375                         p->o_arg.access = NFS4_ACCESS_READ |
1376                                 NFS4_ACCESS_MODIFY |
1377                                 NFS4_ACCESS_EXTEND |
1378                                 NFS4_ACCESS_EXECUTE;
1379 #ifdef CONFIG_NFS_V4_2
1380                         if (server->caps & NFS_CAP_XATTR)
1381                                 p->o_arg.access |= NFS4_ACCESS_XAREAD |
1382                                     NFS4_ACCESS_XAWRITE |
1383                                     NFS4_ACCESS_XALIST;
1384 #endif
1385                 }
1386         }
1387         p->o_arg.clientid = server->nfs_client->cl_clientid;
1388         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1389         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1390         p->o_arg.name = &dentry->d_name;
1391         p->o_arg.server = server;
1392         p->o_arg.bitmask = nfs4_bitmask(server, label);
1393         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1394         switch (p->o_arg.claim) {
1395         case NFS4_OPEN_CLAIM_NULL:
1396         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1397         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1398                 p->o_arg.fh = NFS_FH(dir);
1399                 break;
1400         case NFS4_OPEN_CLAIM_PREVIOUS:
1401         case NFS4_OPEN_CLAIM_FH:
1402         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1403         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1404                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1405         }
1406         p->c_arg.fh = &p->o_res.fh;
1407         p->c_arg.stateid = &p->o_res.stateid;
1408         p->c_arg.seqid = p->o_arg.seqid;
1409         nfs4_init_opendata_res(p);
1410         kref_init(&p->kref);
1411         return p;
1412
1413 err_free_label:
1414         nfs4_label_free(p->a_label);
1415 err_free_f:
1416         nfs4_label_free(p->f_label);
1417 err_free_p:
1418         kfree(p);
1419 err:
1420         dput(parent);
1421         return NULL;
1422 }
1423
1424 static void nfs4_opendata_free(struct kref *kref)
1425 {
1426         struct nfs4_opendata *p = container_of(kref,
1427                         struct nfs4_opendata, kref);
1428         struct super_block *sb = p->dentry->d_sb;
1429
1430         nfs4_lgopen_release(p->lgp);
1431         nfs_free_seqid(p->o_arg.seqid);
1432         nfs4_sequence_free_slot(&p->o_res.seq_res);
1433         if (p->state != NULL)
1434                 nfs4_put_open_state(p->state);
1435         nfs4_put_state_owner(p->owner);
1436
1437         nfs4_label_free(p->a_label);
1438         nfs4_label_free(p->f_label);
1439
1440         dput(p->dir);
1441         dput(p->dentry);
1442         nfs_sb_deactive(sb);
1443         nfs_fattr_free_names(&p->f_attr);
1444         kfree(p->f_attr.mdsthreshold);
1445         kfree(p);
1446 }
1447
1448 static void nfs4_opendata_put(struct nfs4_opendata *p)
1449 {
1450         if (p != NULL)
1451                 kref_put(&p->kref, nfs4_opendata_free);
1452 }
1453
1454 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1455                 fmode_t fmode)
1456 {
1457         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1458         case FMODE_READ|FMODE_WRITE:
1459                 return state->n_rdwr != 0;
1460         case FMODE_WRITE:
1461                 return state->n_wronly != 0;
1462         case FMODE_READ:
1463                 return state->n_rdonly != 0;
1464         }
1465         WARN_ON_ONCE(1);
1466         return false;
1467 }
1468
1469 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1470                 int open_mode, enum open_claim_type4 claim)
1471 {
1472         int ret = 0;
1473
1474         if (open_mode & (O_EXCL|O_TRUNC))
1475                 goto out;
1476         switch (claim) {
1477         case NFS4_OPEN_CLAIM_NULL:
1478         case NFS4_OPEN_CLAIM_FH:
1479                 goto out;
1480         default:
1481                 break;
1482         }
1483         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1484                 case FMODE_READ:
1485                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1486                                 && state->n_rdonly != 0;
1487                         break;
1488                 case FMODE_WRITE:
1489                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1490                                 && state->n_wronly != 0;
1491                         break;
1492                 case FMODE_READ|FMODE_WRITE:
1493                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1494                                 && state->n_rdwr != 0;
1495         }
1496 out:
1497         return ret;
1498 }
1499
1500 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1501                 enum open_claim_type4 claim)
1502 {
1503         if (delegation == NULL)
1504                 return 0;
1505         if ((delegation->type & fmode) != fmode)
1506                 return 0;
1507         switch (claim) {
1508         case NFS4_OPEN_CLAIM_NULL:
1509         case NFS4_OPEN_CLAIM_FH:
1510                 break;
1511         case NFS4_OPEN_CLAIM_PREVIOUS:
1512                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1513                         break;
1514                 fallthrough;
1515         default:
1516                 return 0;
1517         }
1518         nfs_mark_delegation_referenced(delegation);
1519         return 1;
1520 }
1521
1522 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1523 {
1524         switch (fmode) {
1525                 case FMODE_WRITE:
1526                         state->n_wronly++;
1527                         break;
1528                 case FMODE_READ:
1529                         state->n_rdonly++;
1530                         break;
1531                 case FMODE_READ|FMODE_WRITE:
1532                         state->n_rdwr++;
1533         }
1534         nfs4_state_set_mode_locked(state, state->state | fmode);
1535 }
1536
1537 #ifdef CONFIG_NFS_V4_1
1538 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1539 {
1540         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1541                 return true;
1542         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1543                 return true;
1544         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1545                 return true;
1546         return false;
1547 }
1548 #endif /* CONFIG_NFS_V4_1 */
1549
1550 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1551 {
1552         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1553                 wake_up_all(&state->waitq);
1554 }
1555
1556 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1557 {
1558         struct nfs_client *clp = state->owner->so_server->nfs_client;
1559         bool need_recover = false;
1560
1561         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1562                 need_recover = true;
1563         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1564                 need_recover = true;
1565         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1566                 need_recover = true;
1567         if (need_recover)
1568                 nfs4_state_mark_reclaim_nograce(clp, state);
1569 }
1570
1571 /*
1572  * Check for whether or not the caller may update the open stateid
1573  * to the value passed in by stateid.
1574  *
1575  * Note: This function relies heavily on the server implementing
1576  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1577  * correctly.
1578  * i.e. The stateid seqids have to be initialised to 1, and
1579  * are then incremented on every state transition.
1580  */
1581 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1582                 const nfs4_stateid *stateid)
1583 {
1584         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1585                 /* The common case - we're updating to a new sequence number */
1586                 if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1587                         nfs4_stateid_is_next(&state->open_stateid, stateid)) {
1588                         return true;
1589                 }
1590         } else {
1591                 /* This is the first OPEN in this generation */
1592                 if (stateid->seqid == cpu_to_be32(1))
1593                         return true;
1594         }
1595         return false;
1596 }
1597
1598 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1599 {
1600         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1601                 return;
1602         if (state->n_wronly)
1603                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1604         if (state->n_rdonly)
1605                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1606         if (state->n_rdwr)
1607                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1608         set_bit(NFS_OPEN_STATE, &state->flags);
1609 }
1610
1611 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1612                 nfs4_stateid *stateid, fmode_t fmode)
1613 {
1614         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1615         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1616         case FMODE_WRITE:
1617                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1618                 break;
1619         case FMODE_READ:
1620                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1621                 break;
1622         case 0:
1623                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1624                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1625                 clear_bit(NFS_OPEN_STATE, &state->flags);
1626         }
1627         if (stateid == NULL)
1628                 return;
1629         /* Handle OPEN+OPEN_DOWNGRADE races */
1630         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1631             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1632                 nfs_resync_open_stateid_locked(state);
1633                 goto out;
1634         }
1635         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1636                 nfs4_stateid_copy(&state->stateid, stateid);
1637         nfs4_stateid_copy(&state->open_stateid, stateid);
1638         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1639 out:
1640         nfs_state_log_update_open_stateid(state);
1641 }
1642
1643 static void nfs_clear_open_stateid(struct nfs4_state *state,
1644         nfs4_stateid *arg_stateid,
1645         nfs4_stateid *stateid, fmode_t fmode)
1646 {
1647         write_seqlock(&state->seqlock);
1648         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1649         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1650                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1651         write_sequnlock(&state->seqlock);
1652         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1653                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1654 }
1655
1656 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1657                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1658         __must_hold(&state->owner->so_lock)
1659         __must_hold(&state->seqlock)
1660         __must_hold(RCU)
1661
1662 {
1663         DEFINE_WAIT(wait);
1664         int status = 0;
1665         for (;;) {
1666
1667                 if (nfs_stateid_is_sequential(state, stateid))
1668                         break;
1669
1670                 if (status)
1671                         break;
1672                 /* Rely on seqids for serialisation with NFSv4.0 */
1673                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1674                         break;
1675
1676                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1677                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1678                 /*
1679                  * Ensure we process the state changes in the same order
1680                  * in which the server processed them by delaying the
1681                  * update of the stateid until we are in sequence.
1682                  */
1683                 write_sequnlock(&state->seqlock);
1684                 spin_unlock(&state->owner->so_lock);
1685                 rcu_read_unlock();
1686                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1687
1688                 if (!signal_pending(current)) {
1689                         if (schedule_timeout(5*HZ) == 0)
1690                                 status = -EAGAIN;
1691                         else
1692                                 status = 0;
1693                 } else
1694                         status = -EINTR;
1695                 finish_wait(&state->waitq, &wait);
1696                 rcu_read_lock();
1697                 spin_lock(&state->owner->so_lock);
1698                 write_seqlock(&state->seqlock);
1699         }
1700
1701         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1702             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1703                 nfs4_stateid_copy(freeme, &state->open_stateid);
1704                 nfs_test_and_clear_all_open_stateid(state);
1705         }
1706
1707         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1708                 nfs4_stateid_copy(&state->stateid, stateid);
1709         nfs4_stateid_copy(&state->open_stateid, stateid);
1710         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1711         nfs_state_log_update_open_stateid(state);
1712 }
1713
1714 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1715                 const nfs4_stateid *open_stateid,
1716                 fmode_t fmode,
1717                 nfs4_stateid *freeme)
1718 {
1719         /*
1720          * Protect the call to nfs4_state_set_mode_locked and
1721          * serialise the stateid update
1722          */
1723         write_seqlock(&state->seqlock);
1724         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1725         switch (fmode) {
1726         case FMODE_READ:
1727                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1728                 break;
1729         case FMODE_WRITE:
1730                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1731                 break;
1732         case FMODE_READ|FMODE_WRITE:
1733                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1734         }
1735         set_bit(NFS_OPEN_STATE, &state->flags);
1736         write_sequnlock(&state->seqlock);
1737 }
1738
1739 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1740 {
1741         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1742         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1743         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1744         clear_bit(NFS_OPEN_STATE, &state->flags);
1745 }
1746
1747 static void nfs_state_set_delegation(struct nfs4_state *state,
1748                 const nfs4_stateid *deleg_stateid,
1749                 fmode_t fmode)
1750 {
1751         /*
1752          * Protect the call to nfs4_state_set_mode_locked and
1753          * serialise the stateid update
1754          */
1755         write_seqlock(&state->seqlock);
1756         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1757         set_bit(NFS_DELEGATED_STATE, &state->flags);
1758         write_sequnlock(&state->seqlock);
1759 }
1760
1761 static void nfs_state_clear_delegation(struct nfs4_state *state)
1762 {
1763         write_seqlock(&state->seqlock);
1764         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1765         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1766         write_sequnlock(&state->seqlock);
1767 }
1768
1769 int update_open_stateid(struct nfs4_state *state,
1770                 const nfs4_stateid *open_stateid,
1771                 const nfs4_stateid *delegation,
1772                 fmode_t fmode)
1773 {
1774         struct nfs_server *server = NFS_SERVER(state->inode);
1775         struct nfs_client *clp = server->nfs_client;
1776         struct nfs_inode *nfsi = NFS_I(state->inode);
1777         struct nfs_delegation *deleg_cur;
1778         nfs4_stateid freeme = { };
1779         int ret = 0;
1780
1781         fmode &= (FMODE_READ|FMODE_WRITE);
1782
1783         rcu_read_lock();
1784         spin_lock(&state->owner->so_lock);
1785         if (open_stateid != NULL) {
1786                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1787                 ret = 1;
1788         }
1789
1790         deleg_cur = nfs4_get_valid_delegation(state->inode);
1791         if (deleg_cur == NULL)
1792                 goto no_delegation;
1793
1794         spin_lock(&deleg_cur->lock);
1795         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1796            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1797             (deleg_cur->type & fmode) != fmode)
1798                 goto no_delegation_unlock;
1799
1800         if (delegation == NULL)
1801                 delegation = &deleg_cur->stateid;
1802         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1803                 goto no_delegation_unlock;
1804
1805         nfs_mark_delegation_referenced(deleg_cur);
1806         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1807         ret = 1;
1808 no_delegation_unlock:
1809         spin_unlock(&deleg_cur->lock);
1810 no_delegation:
1811         if (ret)
1812                 update_open_stateflags(state, fmode);
1813         spin_unlock(&state->owner->so_lock);
1814         rcu_read_unlock();
1815
1816         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1817                 nfs4_schedule_state_manager(clp);
1818         if (freeme.type != 0)
1819                 nfs4_test_and_free_stateid(server, &freeme,
1820                                 state->owner->so_cred);
1821
1822         return ret;
1823 }
1824
1825 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1826                 const nfs4_stateid *stateid)
1827 {
1828         struct nfs4_state *state = lsp->ls_state;
1829         bool ret = false;
1830
1831         spin_lock(&state->state_lock);
1832         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1833                 goto out_noupdate;
1834         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1835                 goto out_noupdate;
1836         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1837         ret = true;
1838 out_noupdate:
1839         spin_unlock(&state->state_lock);
1840         return ret;
1841 }
1842
1843 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1844 {
1845         struct nfs_delegation *delegation;
1846
1847         fmode &= FMODE_READ|FMODE_WRITE;
1848         rcu_read_lock();
1849         delegation = nfs4_get_valid_delegation(inode);
1850         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1851                 rcu_read_unlock();
1852                 return;
1853         }
1854         rcu_read_unlock();
1855         nfs4_inode_return_delegation(inode);
1856 }
1857
1858 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1859 {
1860         struct nfs4_state *state = opendata->state;
1861         struct nfs_delegation *delegation;
1862         int open_mode = opendata->o_arg.open_flags;
1863         fmode_t fmode = opendata->o_arg.fmode;
1864         enum open_claim_type4 claim = opendata->o_arg.claim;
1865         nfs4_stateid stateid;
1866         int ret = -EAGAIN;
1867
1868         for (;;) {
1869                 spin_lock(&state->owner->so_lock);
1870                 if (can_open_cached(state, fmode, open_mode, claim)) {
1871                         update_open_stateflags(state, fmode);
1872                         spin_unlock(&state->owner->so_lock);
1873                         goto out_return_state;
1874                 }
1875                 spin_unlock(&state->owner->so_lock);
1876                 rcu_read_lock();
1877                 delegation = nfs4_get_valid_delegation(state->inode);
1878                 if (!can_open_delegated(delegation, fmode, claim)) {
1879                         rcu_read_unlock();
1880                         break;
1881                 }
1882                 /* Save the delegation */
1883                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1884                 rcu_read_unlock();
1885                 nfs_release_seqid(opendata->o_arg.seqid);
1886                 if (!opendata->is_recover) {
1887                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1888                         if (ret != 0)
1889                                 goto out;
1890                 }
1891                 ret = -EAGAIN;
1892
1893                 /* Try to update the stateid using the delegation */
1894                 if (update_open_stateid(state, NULL, &stateid, fmode))
1895                         goto out_return_state;
1896         }
1897 out:
1898         return ERR_PTR(ret);
1899 out_return_state:
1900         refcount_inc(&state->count);
1901         return state;
1902 }
1903
1904 static void
1905 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1906 {
1907         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1908         struct nfs_delegation *delegation;
1909         int delegation_flags = 0;
1910
1911         rcu_read_lock();
1912         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1913         if (delegation)
1914                 delegation_flags = delegation->flags;
1915         rcu_read_unlock();
1916         switch (data->o_arg.claim) {
1917         default:
1918                 break;
1919         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1920         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1921                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1922                                    "returning a delegation for "
1923                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1924                                    clp->cl_hostname);
1925                 return;
1926         }
1927         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1928                 nfs_inode_set_delegation(state->inode,
1929                                 data->owner->so_cred,
1930                                 data->o_res.delegation_type,
1931                                 &data->o_res.delegation,
1932                                 data->o_res.pagemod_limit);
1933         else
1934                 nfs_inode_reclaim_delegation(state->inode,
1935                                 data->owner->so_cred,
1936                                 data->o_res.delegation_type,
1937                                 &data->o_res.delegation,
1938                                 data->o_res.pagemod_limit);
1939
1940         if (data->o_res.do_recall)
1941                 nfs_async_inode_return_delegation(state->inode,
1942                                                   &data->o_res.delegation);
1943 }
1944
1945 /*
1946  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1947  * and update the nfs4_state.
1948  */
1949 static struct nfs4_state *
1950 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1951 {
1952         struct inode *inode = data->state->inode;
1953         struct nfs4_state *state = data->state;
1954         int ret;
1955
1956         if (!data->rpc_done) {
1957                 if (data->rpc_status)
1958                         return ERR_PTR(data->rpc_status);
1959                 /* cached opens have already been processed */
1960                 goto update;
1961         }
1962
1963         ret = nfs_refresh_inode(inode, &data->f_attr);
1964         if (ret)
1965                 return ERR_PTR(ret);
1966
1967         if (data->o_res.delegation_type != 0)
1968                 nfs4_opendata_check_deleg(data, state);
1969 update:
1970         if (!update_open_stateid(state, &data->o_res.stateid,
1971                                 NULL, data->o_arg.fmode))
1972                 return ERR_PTR(-EAGAIN);
1973         refcount_inc(&state->count);
1974
1975         return state;
1976 }
1977
1978 static struct inode *
1979 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1980 {
1981         struct inode *inode;
1982
1983         switch (data->o_arg.claim) {
1984         case NFS4_OPEN_CLAIM_NULL:
1985         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1986         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1987                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
1988                         return ERR_PTR(-EAGAIN);
1989                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
1990                                 &data->f_attr, data->f_label);
1991                 break;
1992         default:
1993                 inode = d_inode(data->dentry);
1994                 ihold(inode);
1995                 nfs_refresh_inode(inode, &data->f_attr);
1996         }
1997         return inode;
1998 }
1999
2000 static struct nfs4_state *
2001 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2002 {
2003         struct nfs4_state *state;
2004         struct inode *inode;
2005
2006         inode = nfs4_opendata_get_inode(data);
2007         if (IS_ERR(inode))
2008                 return ERR_CAST(inode);
2009         if (data->state != NULL && data->state->inode == inode) {
2010                 state = data->state;
2011                 refcount_inc(&state->count);
2012         } else
2013                 state = nfs4_get_open_state(inode, data->owner);
2014         iput(inode);
2015         if (state == NULL)
2016                 state = ERR_PTR(-ENOMEM);
2017         return state;
2018 }
2019
2020 static struct nfs4_state *
2021 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2022 {
2023         struct nfs4_state *state;
2024
2025         if (!data->rpc_done) {
2026                 state = nfs4_try_open_cached(data);
2027                 trace_nfs4_cached_open(data->state);
2028                 goto out;
2029         }
2030
2031         state = nfs4_opendata_find_nfs4_state(data);
2032         if (IS_ERR(state))
2033                 goto out;
2034
2035         if (data->o_res.delegation_type != 0)
2036                 nfs4_opendata_check_deleg(data, state);
2037         if (!update_open_stateid(state, &data->o_res.stateid,
2038                                 NULL, data->o_arg.fmode)) {
2039                 nfs4_put_open_state(state);
2040                 state = ERR_PTR(-EAGAIN);
2041         }
2042 out:
2043         nfs_release_seqid(data->o_arg.seqid);
2044         return state;
2045 }
2046
2047 static struct nfs4_state *
2048 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2049 {
2050         struct nfs4_state *ret;
2051
2052         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2053                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2054         else
2055                 ret = _nfs4_opendata_to_nfs4_state(data);
2056         nfs4_sequence_free_slot(&data->o_res.seq_res);
2057         return ret;
2058 }
2059
2060 static struct nfs_open_context *
2061 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2062 {
2063         struct nfs_inode *nfsi = NFS_I(state->inode);
2064         struct nfs_open_context *ctx;
2065
2066         rcu_read_lock();
2067         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2068                 if (ctx->state != state)
2069                         continue;
2070                 if ((ctx->mode & mode) != mode)
2071                         continue;
2072                 if (!get_nfs_open_context(ctx))
2073                         continue;
2074                 rcu_read_unlock();
2075                 return ctx;
2076         }
2077         rcu_read_unlock();
2078         return ERR_PTR(-ENOENT);
2079 }
2080
2081 static struct nfs_open_context *
2082 nfs4_state_find_open_context(struct nfs4_state *state)
2083 {
2084         struct nfs_open_context *ctx;
2085
2086         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2087         if (!IS_ERR(ctx))
2088                 return ctx;
2089         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2090         if (!IS_ERR(ctx))
2091                 return ctx;
2092         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2093 }
2094
2095 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2096                 struct nfs4_state *state, enum open_claim_type4 claim)
2097 {
2098         struct nfs4_opendata *opendata;
2099
2100         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2101                         NULL, claim, GFP_NOFS);
2102         if (opendata == NULL)
2103                 return ERR_PTR(-ENOMEM);
2104         opendata->state = state;
2105         refcount_inc(&state->count);
2106         return opendata;
2107 }
2108
2109 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2110                 fmode_t fmode)
2111 {
2112         struct nfs4_state *newstate;
2113         int ret;
2114
2115         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2116                 return 0;
2117         opendata->o_arg.open_flags = 0;
2118         opendata->o_arg.fmode = fmode;
2119         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2120                         NFS_SB(opendata->dentry->d_sb),
2121                         fmode, 0);
2122         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2123         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2124         nfs4_init_opendata_res(opendata);
2125         ret = _nfs4_recover_proc_open(opendata);
2126         if (ret != 0)
2127                 return ret;
2128         newstate = nfs4_opendata_to_nfs4_state(opendata);
2129         if (IS_ERR(newstate))
2130                 return PTR_ERR(newstate);
2131         if (newstate != opendata->state)
2132                 ret = -ESTALE;
2133         nfs4_close_state(newstate, fmode);
2134         return ret;
2135 }
2136
2137 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2138 {
2139         int ret;
2140
2141         /* memory barrier prior to reading state->n_* */
2142         smp_rmb();
2143         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2144         if (ret != 0)
2145                 return ret;
2146         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2147         if (ret != 0)
2148                 return ret;
2149         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2150         if (ret != 0)
2151                 return ret;
2152         /*
2153          * We may have performed cached opens for all three recoveries.
2154          * Check if we need to update the current stateid.
2155          */
2156         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2157             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2158                 write_seqlock(&state->seqlock);
2159                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2160                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2161                 write_sequnlock(&state->seqlock);
2162         }
2163         return 0;
2164 }
2165
2166 /*
2167  * OPEN_RECLAIM:
2168  *      reclaim state on the server after a reboot.
2169  */
2170 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2171 {
2172         struct nfs_delegation *delegation;
2173         struct nfs4_opendata *opendata;
2174         fmode_t delegation_type = 0;
2175         int status;
2176
2177         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2178                         NFS4_OPEN_CLAIM_PREVIOUS);
2179         if (IS_ERR(opendata))
2180                 return PTR_ERR(opendata);
2181         rcu_read_lock();
2182         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2183         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2184                 delegation_type = delegation->type;
2185         rcu_read_unlock();
2186         opendata->o_arg.u.delegation_type = delegation_type;
2187         status = nfs4_open_recover(opendata, state);
2188         nfs4_opendata_put(opendata);
2189         return status;
2190 }
2191
2192 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2193 {
2194         struct nfs_server *server = NFS_SERVER(state->inode);
2195         struct nfs4_exception exception = { };
2196         int err;
2197         do {
2198                 err = _nfs4_do_open_reclaim(ctx, state);
2199                 trace_nfs4_open_reclaim(ctx, 0, err);
2200                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2201                         continue;
2202                 if (err != -NFS4ERR_DELAY)
2203                         break;
2204                 nfs4_handle_exception(server, err, &exception);
2205         } while (exception.retry);
2206         return err;
2207 }
2208
2209 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2210 {
2211         struct nfs_open_context *ctx;
2212         int ret;
2213
2214         ctx = nfs4_state_find_open_context(state);
2215         if (IS_ERR(ctx))
2216                 return -EAGAIN;
2217         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2218         nfs_state_clear_open_state_flags(state);
2219         ret = nfs4_do_open_reclaim(ctx, state);
2220         put_nfs_open_context(ctx);
2221         return ret;
2222 }
2223
2224 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)
2225 {
2226         switch (err) {
2227                 default:
2228                         printk(KERN_ERR "NFS: %s: unhandled error "
2229                                         "%d.\n", __func__, err);
2230                         fallthrough;
2231                 case 0:
2232                 case -ENOENT:
2233                 case -EAGAIN:
2234                 case -ESTALE:
2235                 case -ETIMEDOUT:
2236                         break;
2237                 case -NFS4ERR_BADSESSION:
2238                 case -NFS4ERR_BADSLOT:
2239                 case -NFS4ERR_BAD_HIGH_SLOT:
2240                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2241                 case -NFS4ERR_DEADSESSION:
2242                         return -EAGAIN;
2243                 case -NFS4ERR_STALE_CLIENTID:
2244                 case -NFS4ERR_STALE_STATEID:
2245                         /* Don't recall a delegation if it was lost */
2246                         nfs4_schedule_lease_recovery(server->nfs_client);
2247                         return -EAGAIN;
2248                 case -NFS4ERR_MOVED:
2249                         nfs4_schedule_migration_recovery(server);
2250                         return -EAGAIN;
2251                 case -NFS4ERR_LEASE_MOVED:
2252                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2253                         return -EAGAIN;
2254                 case -NFS4ERR_DELEG_REVOKED:
2255                 case -NFS4ERR_ADMIN_REVOKED:
2256                 case -NFS4ERR_EXPIRED:
2257                 case -NFS4ERR_BAD_STATEID:
2258                 case -NFS4ERR_OPENMODE:
2259                         nfs_inode_find_state_and_recover(state->inode,
2260                                         stateid);
2261                         nfs4_schedule_stateid_recovery(server, state);
2262                         return -EAGAIN;
2263                 case -NFS4ERR_DELAY:
2264                 case -NFS4ERR_GRACE:
2265                         ssleep(1);
2266                         return -EAGAIN;
2267                 case -ENOMEM:
2268                 case -NFS4ERR_DENIED:
2269                         if (fl) {
2270                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2271                                 if (lsp)
2272                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2273                         }
2274                         return 0;
2275         }
2276         return err;
2277 }
2278
2279 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2280                 struct nfs4_state *state, const nfs4_stateid *stateid)
2281 {
2282         struct nfs_server *server = NFS_SERVER(state->inode);
2283         struct nfs4_opendata *opendata;
2284         int err = 0;
2285
2286         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2287                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2288         if (IS_ERR(opendata))
2289                 return PTR_ERR(opendata);
2290         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2291         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2292                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2293                 if (err)
2294                         goto out;
2295         }
2296         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2297                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2298                 if (err)
2299                         goto out;
2300         }
2301         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2302                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2303                 if (err)
2304                         goto out;
2305         }
2306         nfs_state_clear_delegation(state);
2307 out:
2308         nfs4_opendata_put(opendata);
2309         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2310 }
2311
2312 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2313 {
2314         struct nfs4_opendata *data = calldata;
2315
2316         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2317                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2318 }
2319
2320 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2321 {
2322         struct nfs4_opendata *data = calldata;
2323
2324         nfs40_sequence_done(task, &data->c_res.seq_res);
2325
2326         data->rpc_status = task->tk_status;
2327         if (data->rpc_status == 0) {
2328                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2329                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2330                 renew_lease(data->o_res.server, data->timestamp);
2331                 data->rpc_done = true;
2332         }
2333 }
2334
2335 static void nfs4_open_confirm_release(void *calldata)
2336 {
2337         struct nfs4_opendata *data = calldata;
2338         struct nfs4_state *state = NULL;
2339
2340         /* If this request hasn't been cancelled, do nothing */
2341         if (!data->cancelled)
2342                 goto out_free;
2343         /* In case of error, no cleanup! */
2344         if (!data->rpc_done)
2345                 goto out_free;
2346         state = nfs4_opendata_to_nfs4_state(data);
2347         if (!IS_ERR(state))
2348                 nfs4_close_state(state, data->o_arg.fmode);
2349 out_free:
2350         nfs4_opendata_put(data);
2351 }
2352
2353 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2354         .rpc_call_prepare = nfs4_open_confirm_prepare,
2355         .rpc_call_done = nfs4_open_confirm_done,
2356         .rpc_release = nfs4_open_confirm_release,
2357 };
2358
2359 /*
2360  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2361  */
2362 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2363 {
2364         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2365         struct rpc_task *task;
2366         struct  rpc_message msg = {
2367                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2368                 .rpc_argp = &data->c_arg,
2369                 .rpc_resp = &data->c_res,
2370                 .rpc_cred = data->owner->so_cred,
2371         };
2372         struct rpc_task_setup task_setup_data = {
2373                 .rpc_client = server->client,
2374                 .rpc_message = &msg,
2375                 .callback_ops = &nfs4_open_confirm_ops,
2376                 .callback_data = data,
2377                 .workqueue = nfsiod_workqueue,
2378                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2379         };
2380         int status;
2381
2382         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2383                                 data->is_recover);
2384         kref_get(&data->kref);
2385         data->rpc_done = false;
2386         data->rpc_status = 0;
2387         data->timestamp = jiffies;
2388         task = rpc_run_task(&task_setup_data);
2389         if (IS_ERR(task))
2390                 return PTR_ERR(task);
2391         status = rpc_wait_for_completion_task(task);
2392         if (status != 0) {
2393                 data->cancelled = true;
2394                 smp_wmb();
2395         } else
2396                 status = data->rpc_status;
2397         rpc_put_task(task);
2398         return status;
2399 }
2400
2401 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2402 {
2403         struct nfs4_opendata *data = calldata;
2404         struct nfs4_state_owner *sp = data->owner;
2405         struct nfs_client *clp = sp->so_server->nfs_client;
2406         enum open_claim_type4 claim = data->o_arg.claim;
2407
2408         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2409                 goto out_wait;
2410         /*
2411          * Check if we still need to send an OPEN call, or if we can use
2412          * a delegation instead.
2413          */
2414         if (data->state != NULL) {
2415                 struct nfs_delegation *delegation;
2416
2417                 if (can_open_cached(data->state, data->o_arg.fmode,
2418                                         data->o_arg.open_flags, claim))
2419                         goto out_no_action;
2420                 rcu_read_lock();
2421                 delegation = nfs4_get_valid_delegation(data->state->inode);
2422                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2423                         goto unlock_no_action;
2424                 rcu_read_unlock();
2425         }
2426         /* Update client id. */
2427         data->o_arg.clientid = clp->cl_clientid;
2428         switch (claim) {
2429         default:
2430                 break;
2431         case NFS4_OPEN_CLAIM_PREVIOUS:
2432         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2433         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2434                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2435                 fallthrough;
2436         case NFS4_OPEN_CLAIM_FH:
2437                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2438         }
2439         data->timestamp = jiffies;
2440         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2441                                 &data->o_arg.seq_args,
2442                                 &data->o_res.seq_res,
2443                                 task) != 0)
2444                 nfs_release_seqid(data->o_arg.seqid);
2445
2446         /* Set the create mode (note dependency on the session type) */
2447         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2448         if (data->o_arg.open_flags & O_EXCL) {
2449                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2450                 if (nfs4_has_persistent_session(clp))
2451                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2452                 else if (clp->cl_mvops->minor_version > 0)
2453                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2454         }
2455         return;
2456 unlock_no_action:
2457         trace_nfs4_cached_open(data->state);
2458         rcu_read_unlock();
2459 out_no_action:
2460         task->tk_action = NULL;
2461 out_wait:
2462         nfs4_sequence_done(task, &data->o_res.seq_res);
2463 }
2464
2465 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2466 {
2467         struct nfs4_opendata *data = calldata;
2468
2469         data->rpc_status = task->tk_status;
2470
2471         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2472                 return;
2473
2474         if (task->tk_status == 0) {
2475                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2476                         switch (data->o_res.f_attr->mode & S_IFMT) {
2477                         case S_IFREG:
2478                                 break;
2479                         case S_IFLNK:
2480                                 data->rpc_status = -ELOOP;
2481                                 break;
2482                         case S_IFDIR:
2483                                 data->rpc_status = -EISDIR;
2484                                 break;
2485                         default:
2486                                 data->rpc_status = -ENOTDIR;
2487                         }
2488                 }
2489                 renew_lease(data->o_res.server, data->timestamp);
2490                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2491                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2492         }
2493         data->rpc_done = true;
2494 }
2495
2496 static void nfs4_open_release(void *calldata)
2497 {
2498         struct nfs4_opendata *data = calldata;
2499         struct nfs4_state *state = NULL;
2500
2501         /* If this request hasn't been cancelled, do nothing */
2502         if (!data->cancelled)
2503                 goto out_free;
2504         /* In case of error, no cleanup! */
2505         if (data->rpc_status != 0 || !data->rpc_done)
2506                 goto out_free;
2507         /* In case we need an open_confirm, no cleanup! */
2508         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2509                 goto out_free;
2510         state = nfs4_opendata_to_nfs4_state(data);
2511         if (!IS_ERR(state))
2512                 nfs4_close_state(state, data->o_arg.fmode);
2513 out_free:
2514         nfs4_opendata_put(data);
2515 }
2516
2517 static const struct rpc_call_ops nfs4_open_ops = {
2518         .rpc_call_prepare = nfs4_open_prepare,
2519         .rpc_call_done = nfs4_open_done,
2520         .rpc_release = nfs4_open_release,
2521 };
2522
2523 static int nfs4_run_open_task(struct nfs4_opendata *data,
2524                               struct nfs_open_context *ctx)
2525 {
2526         struct inode *dir = d_inode(data->dir);
2527         struct nfs_server *server = NFS_SERVER(dir);
2528         struct nfs_openargs *o_arg = &data->o_arg;
2529         struct nfs_openres *o_res = &data->o_res;
2530         struct rpc_task *task;
2531         struct rpc_message msg = {
2532                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2533                 .rpc_argp = o_arg,
2534                 .rpc_resp = o_res,
2535                 .rpc_cred = data->owner->so_cred,
2536         };
2537         struct rpc_task_setup task_setup_data = {
2538                 .rpc_client = server->client,
2539                 .rpc_message = &msg,
2540                 .callback_ops = &nfs4_open_ops,
2541                 .callback_data = data,
2542                 .workqueue = nfsiod_workqueue,
2543                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2544         };
2545         int status;
2546
2547         kref_get(&data->kref);
2548         data->rpc_done = false;
2549         data->rpc_status = 0;
2550         data->cancelled = false;
2551         data->is_recover = false;
2552         if (!ctx) {
2553                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2554                 data->is_recover = true;
2555                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2556         } else {
2557                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2558                 pnfs_lgopen_prepare(data, ctx);
2559         }
2560         task = rpc_run_task(&task_setup_data);
2561         if (IS_ERR(task))
2562                 return PTR_ERR(task);
2563         status = rpc_wait_for_completion_task(task);
2564         if (status != 0) {
2565                 data->cancelled = true;
2566                 smp_wmb();
2567         } else
2568                 status = data->rpc_status;
2569         rpc_put_task(task);
2570
2571         return status;
2572 }
2573
2574 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2575 {
2576         struct inode *dir = d_inode(data->dir);
2577         struct nfs_openres *o_res = &data->o_res;
2578         int status;
2579
2580         status = nfs4_run_open_task(data, NULL);
2581         if (status != 0 || !data->rpc_done)
2582                 return status;
2583
2584         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2585
2586         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2587                 status = _nfs4_proc_open_confirm(data);
2588
2589         return status;
2590 }
2591
2592 /*
2593  * Additional permission checks in order to distinguish between an
2594  * open for read, and an open for execute. This works around the
2595  * fact that NFSv4 OPEN treats read and execute permissions as being
2596  * the same.
2597  * Note that in the non-execute case, we want to turn off permission
2598  * checking if we just created a new file (POSIX open() semantics).
2599  */
2600 static int nfs4_opendata_access(const struct cred *cred,
2601                                 struct nfs4_opendata *opendata,
2602                                 struct nfs4_state *state, fmode_t fmode,
2603                                 int openflags)
2604 {
2605         struct nfs_access_entry cache;
2606         u32 mask, flags;
2607
2608         /* access call failed or for some reason the server doesn't
2609          * support any access modes -- defer access call until later */
2610         if (opendata->o_res.access_supported == 0)
2611                 return 0;
2612
2613         mask = 0;
2614         /*
2615          * Use openflags to check for exec, because fmode won't
2616          * always have FMODE_EXEC set when file open for exec.
2617          */
2618         if (openflags & __FMODE_EXEC) {
2619                 /* ONLY check for exec rights */
2620                 if (S_ISDIR(state->inode->i_mode))
2621                         mask = NFS4_ACCESS_LOOKUP;
2622                 else
2623                         mask = NFS4_ACCESS_EXECUTE;
2624         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2625                 mask = NFS4_ACCESS_READ;
2626
2627         cache.cred = cred;
2628         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2629         nfs_access_add_cache(state->inode, &cache);
2630
2631         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2632         if ((mask & ~cache.mask & flags) == 0)
2633                 return 0;
2634
2635         return -EACCES;
2636 }
2637
2638 /*
2639  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2640  */
2641 static int _nfs4_proc_open(struct nfs4_opendata *data,
2642                            struct nfs_open_context *ctx)
2643 {
2644         struct inode *dir = d_inode(data->dir);
2645         struct nfs_server *server = NFS_SERVER(dir);
2646         struct nfs_openargs *o_arg = &data->o_arg;
2647         struct nfs_openres *o_res = &data->o_res;
2648         int status;
2649
2650         status = nfs4_run_open_task(data, ctx);
2651         if (!data->rpc_done)
2652                 return status;
2653         if (status != 0) {
2654                 if (status == -NFS4ERR_BADNAME &&
2655                                 !(o_arg->open_flags & O_CREAT))
2656                         return -ENOENT;
2657                 return status;
2658         }
2659
2660         nfs_fattr_map_and_free_names(server, &data->f_attr);
2661
2662         if (o_arg->open_flags & O_CREAT) {
2663                 if (o_arg->open_flags & O_EXCL)
2664                         data->file_created = true;
2665                 else if (o_res->cinfo.before != o_res->cinfo.after)
2666                         data->file_created = true;
2667                 if (data->file_created ||
2668                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2669                         nfs4_update_changeattr(dir, &o_res->cinfo,
2670                                         o_res->f_attr->time_start,
2671                                         NFS_INO_INVALID_DATA);
2672         }
2673         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2674                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2675         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2676                 status = _nfs4_proc_open_confirm(data);
2677                 if (status != 0)
2678                         return status;
2679         }
2680         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2681                 nfs4_sequence_free_slot(&o_res->seq_res);
2682                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr,
2683                                 o_res->f_label, NULL);
2684         }
2685         return 0;
2686 }
2687
2688 /*
2689  * OPEN_EXPIRED:
2690  *      reclaim state on the server after a network partition.
2691  *      Assumes caller holds the appropriate lock
2692  */
2693 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2694 {
2695         struct nfs4_opendata *opendata;
2696         int ret;
2697
2698         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2699                         NFS4_OPEN_CLAIM_FH);
2700         if (IS_ERR(opendata))
2701                 return PTR_ERR(opendata);
2702         ret = nfs4_open_recover(opendata, state);
2703         if (ret == -ESTALE)
2704                 d_drop(ctx->dentry);
2705         nfs4_opendata_put(opendata);
2706         return ret;
2707 }
2708
2709 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2710 {
2711         struct nfs_server *server = NFS_SERVER(state->inode);
2712         struct nfs4_exception exception = { };
2713         int err;
2714
2715         do {
2716                 err = _nfs4_open_expired(ctx, state);
2717                 trace_nfs4_open_expired(ctx, 0, err);
2718                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2719                         continue;
2720                 switch (err) {
2721                 default:
2722                         goto out;
2723                 case -NFS4ERR_GRACE:
2724                 case -NFS4ERR_DELAY:
2725                         nfs4_handle_exception(server, err, &exception);
2726                         err = 0;
2727                 }
2728         } while (exception.retry);
2729 out:
2730         return err;
2731 }
2732
2733 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2734 {
2735         struct nfs_open_context *ctx;
2736         int ret;
2737
2738         ctx = nfs4_state_find_open_context(state);
2739         if (IS_ERR(ctx))
2740                 return -EAGAIN;
2741         ret = nfs4_do_open_expired(ctx, state);
2742         put_nfs_open_context(ctx);
2743         return ret;
2744 }
2745
2746 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2747                 const nfs4_stateid *stateid)
2748 {
2749         nfs_remove_bad_delegation(state->inode, stateid);
2750         nfs_state_clear_delegation(state);
2751 }
2752
2753 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2754 {
2755         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2756                 nfs_finish_clear_delegation_stateid(state, NULL);
2757 }
2758
2759 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2760 {
2761         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2762         nfs40_clear_delegation_stateid(state);
2763         nfs_state_clear_open_state_flags(state);
2764         return nfs4_open_expired(sp, state);
2765 }
2766
2767 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2768                 nfs4_stateid *stateid,
2769                 const struct cred *cred)
2770 {
2771         return -NFS4ERR_BAD_STATEID;
2772 }
2773
2774 #if defined(CONFIG_NFS_V4_1)
2775 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2776                 nfs4_stateid *stateid,
2777                 const struct cred *cred)
2778 {
2779         int status;
2780
2781         switch (stateid->type) {
2782         default:
2783                 break;
2784         case NFS4_INVALID_STATEID_TYPE:
2785         case NFS4_SPECIAL_STATEID_TYPE:
2786                 return -NFS4ERR_BAD_STATEID;
2787         case NFS4_REVOKED_STATEID_TYPE:
2788                 goto out_free;
2789         }
2790
2791         status = nfs41_test_stateid(server, stateid, cred);
2792         switch (status) {
2793         case -NFS4ERR_EXPIRED:
2794         case -NFS4ERR_ADMIN_REVOKED:
2795         case -NFS4ERR_DELEG_REVOKED:
2796                 break;
2797         default:
2798                 return status;
2799         }
2800 out_free:
2801         /* Ack the revoked state to the server */
2802         nfs41_free_stateid(server, stateid, cred, true);
2803         return -NFS4ERR_EXPIRED;
2804 }
2805
2806 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2807 {
2808         struct nfs_server *server = NFS_SERVER(state->inode);
2809         nfs4_stateid stateid;
2810         struct nfs_delegation *delegation;
2811         const struct cred *cred = NULL;
2812         int status, ret = NFS_OK;
2813
2814         /* Get the delegation credential for use by test/free_stateid */
2815         rcu_read_lock();
2816         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2817         if (delegation == NULL) {
2818                 rcu_read_unlock();
2819                 nfs_state_clear_delegation(state);
2820                 return NFS_OK;
2821         }
2822
2823         spin_lock(&delegation->lock);
2824         nfs4_stateid_copy(&stateid, &delegation->stateid);
2825
2826         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2827                                 &delegation->flags)) {
2828                 spin_unlock(&delegation->lock);
2829                 rcu_read_unlock();
2830                 return NFS_OK;
2831         }
2832
2833         if (delegation->cred)
2834                 cred = get_cred(delegation->cred);
2835         spin_unlock(&delegation->lock);
2836         rcu_read_unlock();
2837         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2838         trace_nfs4_test_delegation_stateid(state, NULL, status);
2839         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2840                 nfs_finish_clear_delegation_stateid(state, &stateid);
2841         else
2842                 ret = status;
2843
2844         put_cred(cred);
2845         return ret;
2846 }
2847
2848 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2849 {
2850         nfs4_stateid tmp;
2851
2852         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2853             nfs4_copy_delegation_stateid(state->inode, state->state,
2854                                 &tmp, NULL) &&
2855             nfs4_stateid_match_other(&state->stateid, &tmp))
2856                 nfs_state_set_delegation(state, &tmp, state->state);
2857         else
2858                 nfs_state_clear_delegation(state);
2859 }
2860
2861 /**
2862  * nfs41_check_expired_locks - possibly free a lock stateid
2863  *
2864  * @state: NFSv4 state for an inode
2865  *
2866  * Returns NFS_OK if recovery for this stateid is now finished.
2867  * Otherwise a negative NFS4ERR value is returned.
2868  */
2869 static int nfs41_check_expired_locks(struct nfs4_state *state)
2870 {
2871         int status, ret = NFS_OK;
2872         struct nfs4_lock_state *lsp, *prev = NULL;
2873         struct nfs_server *server = NFS_SERVER(state->inode);
2874
2875         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2876                 goto out;
2877
2878         spin_lock(&state->state_lock);
2879         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2880                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2881                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2882
2883                         refcount_inc(&lsp->ls_count);
2884                         spin_unlock(&state->state_lock);
2885
2886                         nfs4_put_lock_state(prev);
2887                         prev = lsp;
2888
2889                         status = nfs41_test_and_free_expired_stateid(server,
2890                                         &lsp->ls_stateid,
2891                                         cred);
2892                         trace_nfs4_test_lock_stateid(state, lsp, status);
2893                         if (status == -NFS4ERR_EXPIRED ||
2894                             status == -NFS4ERR_BAD_STATEID) {
2895                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2896                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2897                                 if (!recover_lost_locks)
2898                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2899                         } else if (status != NFS_OK) {
2900                                 ret = status;
2901                                 nfs4_put_lock_state(prev);
2902                                 goto out;
2903                         }
2904                         spin_lock(&state->state_lock);
2905                 }
2906         }
2907         spin_unlock(&state->state_lock);
2908         nfs4_put_lock_state(prev);
2909 out:
2910         return ret;
2911 }
2912
2913 /**
2914  * nfs41_check_open_stateid - possibly free an open stateid
2915  *
2916  * @state: NFSv4 state for an inode
2917  *
2918  * Returns NFS_OK if recovery for this stateid is now finished.
2919  * Otherwise a negative NFS4ERR value is returned.
2920  */
2921 static int nfs41_check_open_stateid(struct nfs4_state *state)
2922 {
2923         struct nfs_server *server = NFS_SERVER(state->inode);
2924         nfs4_stateid *stateid = &state->open_stateid;
2925         const struct cred *cred = state->owner->so_cred;
2926         int status;
2927
2928         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2929                 return -NFS4ERR_BAD_STATEID;
2930         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2931         trace_nfs4_test_open_stateid(state, NULL, status);
2932         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2933                 nfs_state_clear_open_state_flags(state);
2934                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2935                 return status;
2936         }
2937         if (nfs_open_stateid_recover_openmode(state))
2938                 return -NFS4ERR_OPENMODE;
2939         return NFS_OK;
2940 }
2941
2942 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2943 {
2944         int status;
2945
2946         status = nfs41_check_delegation_stateid(state);
2947         if (status != NFS_OK)
2948                 return status;
2949         nfs41_delegation_recover_stateid(state);
2950
2951         status = nfs41_check_expired_locks(state);
2952         if (status != NFS_OK)
2953                 return status;
2954         status = nfs41_check_open_stateid(state);
2955         if (status != NFS_OK)
2956                 status = nfs4_open_expired(sp, state);
2957         return status;
2958 }
2959 #endif
2960
2961 /*
2962  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2963  * fields corresponding to attributes that were used to store the verifier.
2964  * Make sure we clobber those fields in the later setattr call
2965  */
2966 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2967                                 struct iattr *sattr, struct nfs4_label **label)
2968 {
2969         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2970         __u32 attrset[3];
2971         unsigned ret;
2972         unsigned i;
2973
2974         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2975                 attrset[i] = opendata->o_res.attrset[i];
2976                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2977                         attrset[i] &= ~bitmask[i];
2978         }
2979
2980         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
2981                 sattr->ia_valid : 0;
2982
2983         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
2984                 if (sattr->ia_valid & ATTR_ATIME_SET)
2985                         ret |= ATTR_ATIME_SET;
2986                 else
2987                         ret |= ATTR_ATIME;
2988         }
2989
2990         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
2991                 if (sattr->ia_valid & ATTR_MTIME_SET)
2992                         ret |= ATTR_MTIME_SET;
2993                 else
2994                         ret |= ATTR_MTIME;
2995         }
2996
2997         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
2998                 *label = NULL;
2999         return ret;
3000 }
3001
3002 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3003                 int flags, struct nfs_open_context *ctx)
3004 {
3005         struct nfs4_state_owner *sp = opendata->owner;
3006         struct nfs_server *server = sp->so_server;
3007         struct dentry *dentry;
3008         struct nfs4_state *state;
3009         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3010         struct inode *dir = d_inode(opendata->dir);
3011         unsigned long dir_verifier;
3012         unsigned int seq;
3013         int ret;
3014
3015         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3016         dir_verifier = nfs_save_change_attribute(dir);
3017
3018         ret = _nfs4_proc_open(opendata, ctx);
3019         if (ret != 0)
3020                 goto out;
3021
3022         state = _nfs4_opendata_to_nfs4_state(opendata);
3023         ret = PTR_ERR(state);
3024         if (IS_ERR(state))
3025                 goto out;
3026         ctx->state = state;
3027         if (server->caps & NFS_CAP_POSIX_LOCK)
3028                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3029         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3030                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3031
3032         dentry = opendata->dentry;
3033         if (d_really_is_negative(dentry)) {
3034                 struct dentry *alias;
3035                 d_drop(dentry);
3036                 alias = d_exact_alias(dentry, state->inode);
3037                 if (!alias)
3038                         alias = d_splice_alias(igrab(state->inode), dentry);
3039                 /* d_splice_alias() can't fail here - it's a non-directory */
3040                 if (alias) {
3041                         dput(ctx->dentry);
3042                         ctx->dentry = dentry = alias;
3043                 }
3044         }
3045
3046         switch(opendata->o_arg.claim) {
3047         default:
3048                 break;
3049         case NFS4_OPEN_CLAIM_NULL:
3050         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3051         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3052                 if (!opendata->rpc_done)
3053                         break;
3054                 if (opendata->o_res.delegation_type != 0)
3055                         dir_verifier = nfs_save_change_attribute(dir);
3056                 nfs_set_verifier(dentry, dir_verifier);
3057         }
3058
3059         /* Parse layoutget results before we check for access */
3060         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3061
3062         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3063                         acc_mode, flags);
3064         if (ret != 0)
3065                 goto out;
3066
3067         if (d_inode(dentry) == state->inode) {
3068                 nfs_inode_attach_open_context(ctx);
3069                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3070                         nfs4_schedule_stateid_recovery(server, state);
3071         }
3072
3073 out:
3074         if (!opendata->cancelled)
3075                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3076         return ret;
3077 }
3078
3079 /*
3080  * Returns a referenced nfs4_state
3081  */
3082 static int _nfs4_do_open(struct inode *dir,
3083                         struct nfs_open_context *ctx,
3084                         int flags,
3085                         const struct nfs4_open_createattrs *c,
3086                         int *opened)
3087 {
3088         struct nfs4_state_owner  *sp;
3089         struct nfs4_state     *state = NULL;
3090         struct nfs_server       *server = NFS_SERVER(dir);
3091         struct nfs4_opendata *opendata;
3092         struct dentry *dentry = ctx->dentry;
3093         const struct cred *cred = ctx->cred;
3094         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3095         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3096         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3097         struct iattr *sattr = c->sattr;
3098         struct nfs4_label *label = c->label;
3099         struct nfs4_label *olabel = NULL;
3100         int status;
3101
3102         /* Protect against reboot recovery conflicts */
3103         status = -ENOMEM;
3104         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3105         if (sp == NULL) {
3106                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3107                 goto out_err;
3108         }
3109         status = nfs4_client_recover_expired_lease(server->nfs_client);
3110         if (status != 0)
3111                 goto err_put_state_owner;
3112         if (d_really_is_positive(dentry))
3113                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3114         status = -ENOMEM;
3115         if (d_really_is_positive(dentry))
3116                 claim = NFS4_OPEN_CLAIM_FH;
3117         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3118                         c, claim, GFP_KERNEL);
3119         if (opendata == NULL)
3120                 goto err_put_state_owner;
3121
3122         if (label) {
3123                 olabel = nfs4_label_alloc(server, GFP_KERNEL);
3124                 if (IS_ERR(olabel)) {
3125                         status = PTR_ERR(olabel);
3126                         goto err_opendata_put;
3127                 }
3128         }
3129
3130         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3131                 if (!opendata->f_attr.mdsthreshold) {
3132                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3133                         if (!opendata->f_attr.mdsthreshold)
3134                                 goto err_free_label;
3135                 }
3136                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3137         }
3138         if (d_really_is_positive(dentry))
3139                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3140
3141         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3142         if (status != 0)
3143                 goto err_free_label;
3144         state = ctx->state;
3145
3146         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3147             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3148                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3149                 /*
3150                  * send create attributes which was not set by open
3151                  * with an extra setattr.
3152                  */
3153                 if (attrs || label) {
3154                         unsigned ia_old = sattr->ia_valid;
3155
3156                         sattr->ia_valid = attrs;
3157                         nfs_fattr_init(opendata->o_res.f_attr);
3158                         status = nfs4_do_setattr(state->inode, cred,
3159                                         opendata->o_res.f_attr, sattr,
3160                                         ctx, label, olabel);
3161                         if (status == 0) {
3162                                 nfs_setattr_update_inode(state->inode, sattr,
3163                                                 opendata->o_res.f_attr);
3164                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
3165                         }
3166                         sattr->ia_valid = ia_old;
3167                 }
3168         }
3169         if (opened && opendata->file_created)
3170                 *opened = 1;
3171
3172         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3173                 *ctx_th = opendata->f_attr.mdsthreshold;
3174                 opendata->f_attr.mdsthreshold = NULL;
3175         }
3176
3177         nfs4_label_free(olabel);
3178
3179         nfs4_opendata_put(opendata);
3180         nfs4_put_state_owner(sp);
3181         return 0;
3182 err_free_label:
3183         nfs4_label_free(olabel);
3184 err_opendata_put:
3185         nfs4_opendata_put(opendata);
3186 err_put_state_owner:
3187         nfs4_put_state_owner(sp);
3188 out_err:
3189         return status;
3190 }
3191
3192
3193 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3194                                         struct nfs_open_context *ctx,
3195                                         int flags,
3196                                         struct iattr *sattr,
3197                                         struct nfs4_label *label,
3198                                         int *opened)
3199 {
3200         struct nfs_server *server = NFS_SERVER(dir);
3201         struct nfs4_exception exception = {
3202                 .interruptible = true,
3203         };
3204         struct nfs4_state *res;
3205         struct nfs4_open_createattrs c = {
3206                 .label = label,
3207                 .sattr = sattr,
3208                 .verf = {
3209                         [0] = (__u32)jiffies,
3210                         [1] = (__u32)current->pid,
3211                 },
3212         };
3213         int status;
3214
3215         do {
3216                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3217                 res = ctx->state;
3218                 trace_nfs4_open_file(ctx, flags, status);
3219                 if (status == 0)
3220                         break;
3221                 /* NOTE: BAD_SEQID means the server and client disagree about the
3222                  * book-keeping w.r.t. state-changing operations
3223                  * (OPEN/CLOSE/LOCK/LOCKU...)
3224                  * It is actually a sign of a bug on the client or on the server.
3225                  *
3226                  * If we receive a BAD_SEQID error in the particular case of
3227                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3228                  * have unhashed the old state_owner for us, and that we can
3229                  * therefore safely retry using a new one. We should still warn
3230                  * the user though...
3231                  */
3232                 if (status == -NFS4ERR_BAD_SEQID) {
3233                         pr_warn_ratelimited("NFS: v4 server %s "
3234                                         " returned a bad sequence-id error!\n",
3235                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3236                         exception.retry = 1;
3237                         continue;
3238                 }
3239                 /*
3240                  * BAD_STATEID on OPEN means that the server cancelled our
3241                  * state before it received the OPEN_CONFIRM.
3242                  * Recover by retrying the request as per the discussion
3243                  * on Page 181 of RFC3530.
3244                  */
3245                 if (status == -NFS4ERR_BAD_STATEID) {
3246                         exception.retry = 1;
3247                         continue;
3248                 }
3249                 if (status == -NFS4ERR_EXPIRED) {
3250                         nfs4_schedule_lease_recovery(server->nfs_client);
3251                         exception.retry = 1;
3252                         continue;
3253                 }
3254                 if (status == -EAGAIN) {
3255                         /* We must have found a delegation */
3256                         exception.retry = 1;
3257                         continue;
3258                 }
3259                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3260                         continue;
3261                 res = ERR_PTR(nfs4_handle_exception(server,
3262                                         status, &exception));
3263         } while (exception.retry);
3264         return res;
3265 }
3266
3267 static int _nfs4_do_setattr(struct inode *inode,
3268                             struct nfs_setattrargs *arg,
3269                             struct nfs_setattrres *res,
3270                             const struct cred *cred,
3271                             struct nfs_open_context *ctx)
3272 {
3273         struct nfs_server *server = NFS_SERVER(inode);
3274         struct rpc_message msg = {
3275                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3276                 .rpc_argp       = arg,
3277                 .rpc_resp       = res,
3278                 .rpc_cred       = cred,
3279         };
3280         const struct cred *delegation_cred = NULL;
3281         unsigned long timestamp = jiffies;
3282         bool truncate;
3283         int status;
3284
3285         nfs_fattr_init(res->fattr);
3286
3287         /* Servers should only apply open mode checks for file size changes */
3288         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3289         if (!truncate) {
3290                 nfs4_inode_make_writeable(inode);
3291                 goto zero_stateid;
3292         }
3293
3294         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3295                 /* Use that stateid */
3296         } else if (ctx != NULL && ctx->state) {
3297                 struct nfs_lock_context *l_ctx;
3298                 if (!nfs4_valid_open_stateid(ctx->state))
3299                         return -EBADF;
3300                 l_ctx = nfs_get_lock_context(ctx);
3301                 if (IS_ERR(l_ctx))
3302                         return PTR_ERR(l_ctx);
3303                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3304                                                 &arg->stateid, &delegation_cred);
3305                 nfs_put_lock_context(l_ctx);
3306                 if (status == -EIO)
3307                         return -EBADF;
3308                 else if (status == -EAGAIN)
3309                         goto zero_stateid;
3310         } else {
3311 zero_stateid:
3312                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3313         }
3314         if (delegation_cred)
3315                 msg.rpc_cred = delegation_cred;
3316
3317         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3318
3319         put_cred(delegation_cred);
3320         if (status == 0 && ctx != NULL)
3321                 renew_lease(server, timestamp);
3322         trace_nfs4_setattr(inode, &arg->stateid, status);
3323         return status;
3324 }
3325
3326 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3327                            struct nfs_fattr *fattr, struct iattr *sattr,
3328                            struct nfs_open_context *ctx, struct nfs4_label *ilabel,
3329                            struct nfs4_label *olabel)
3330 {
3331         struct nfs_server *server = NFS_SERVER(inode);
3332         __u32 bitmask[NFS4_BITMASK_SZ];
3333         struct nfs4_state *state = ctx ? ctx->state : NULL;
3334         struct nfs_setattrargs  arg = {
3335                 .fh             = NFS_FH(inode),
3336                 .iap            = sattr,
3337                 .server         = server,
3338                 .bitmask = bitmask,
3339                 .label          = ilabel,
3340         };
3341         struct nfs_setattrres  res = {
3342                 .fattr          = fattr,
3343                 .label          = olabel,
3344                 .server         = server,
3345         };
3346         struct nfs4_exception exception = {
3347                 .state = state,
3348                 .inode = inode,
3349                 .stateid = &arg.stateid,
3350         };
3351         int err;
3352
3353         do {
3354                 nfs4_bitmap_copy_adjust_setattr(bitmask,
3355                                 nfs4_bitmask(server, olabel),
3356                                 inode);
3357
3358                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3359                 switch (err) {
3360                 case -NFS4ERR_OPENMODE:
3361                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3362                                 pr_warn_once("NFSv4: server %s is incorrectly "
3363                                                 "applying open mode checks to "
3364                                                 "a SETATTR that is not "
3365                                                 "changing file size.\n",
3366                                                 server->nfs_client->cl_hostname);
3367                         }
3368                         if (state && !(state->state & FMODE_WRITE)) {
3369                                 err = -EBADF;
3370                                 if (sattr->ia_valid & ATTR_OPEN)
3371                                         err = -EACCES;
3372                                 goto out;
3373                         }
3374                 }
3375                 err = nfs4_handle_exception(server, err, &exception);
3376         } while (exception.retry);
3377 out:
3378         return err;
3379 }
3380
3381 static bool
3382 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3383 {
3384         if (inode == NULL || !nfs_have_layout(inode))
3385                 return false;
3386
3387         return pnfs_wait_on_layoutreturn(inode, task);
3388 }
3389
3390 /*
3391  * Update the seqid of an open stateid
3392  */
3393 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3394                 struct nfs4_state *state)
3395 {
3396         __be32 seqid_open;
3397         u32 dst_seqid;
3398         int seq;
3399
3400         for (;;) {
3401                 if (!nfs4_valid_open_stateid(state))
3402                         break;
3403                 seq = read_seqbegin(&state->seqlock);
3404                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3405                         nfs4_stateid_copy(dst, &state->open_stateid);
3406                         if (read_seqretry(&state->seqlock, seq))
3407                                 continue;
3408                         break;
3409                 }
3410                 seqid_open = state->open_stateid.seqid;
3411                 if (read_seqretry(&state->seqlock, seq))
3412                         continue;
3413
3414                 dst_seqid = be32_to_cpu(dst->seqid);
3415                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3416                         dst->seqid = seqid_open;
3417                 break;
3418         }
3419 }
3420
3421 /*
3422  * Update the seqid of an open stateid after receiving
3423  * NFS4ERR_OLD_STATEID
3424  */
3425 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3426                 struct nfs4_state *state)
3427 {
3428         __be32 seqid_open;
3429         u32 dst_seqid;
3430         bool ret;
3431         int seq, status = -EAGAIN;
3432         DEFINE_WAIT(wait);
3433
3434         for (;;) {
3435                 ret = false;
3436                 if (!nfs4_valid_open_stateid(state))
3437                         break;
3438                 seq = read_seqbegin(&state->seqlock);
3439                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3440                         if (read_seqretry(&state->seqlock, seq))
3441                                 continue;
3442                         break;
3443                 }
3444
3445                 write_seqlock(&state->seqlock);
3446                 seqid_open = state->open_stateid.seqid;
3447
3448                 dst_seqid = be32_to_cpu(dst->seqid);
3449
3450                 /* Did another OPEN bump the state's seqid?  try again: */
3451                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3452                         dst->seqid = seqid_open;
3453                         write_sequnlock(&state->seqlock);
3454                         ret = true;
3455                         break;
3456                 }
3457
3458                 /* server says we're behind but we haven't seen the update yet */
3459                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3460                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3461                 write_sequnlock(&state->seqlock);
3462                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3463
3464                 if (signal_pending(current))
3465                         status = -EINTR;
3466                 else
3467                         if (schedule_timeout(5*HZ) != 0)
3468                                 status = 0;
3469
3470                 finish_wait(&state->waitq, &wait);
3471
3472                 if (!status)
3473                         continue;
3474                 if (status == -EINTR)
3475                         break;
3476
3477                 /* we slept the whole 5 seconds, we must have lost a seqid */
3478                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3479                 ret = true;
3480                 break;
3481         }
3482
3483         return ret;
3484 }
3485
3486 struct nfs4_closedata {
3487         struct inode *inode;
3488         struct nfs4_state *state;
3489         struct nfs_closeargs arg;
3490         struct nfs_closeres res;
3491         struct {
3492                 struct nfs4_layoutreturn_args arg;
3493                 struct nfs4_layoutreturn_res res;
3494                 struct nfs4_xdr_opaque_data ld_private;
3495                 u32 roc_barrier;
3496                 bool roc;
3497         } lr;
3498         struct nfs_fattr fattr;
3499         unsigned long timestamp;
3500 };
3501
3502 static void nfs4_free_closedata(void *data)
3503 {
3504         struct nfs4_closedata *calldata = data;
3505         struct nfs4_state_owner *sp = calldata->state->owner;
3506         struct super_block *sb = calldata->state->inode->i_sb;
3507
3508         if (calldata->lr.roc)
3509                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3510                                 calldata->res.lr_ret);
3511         nfs4_put_open_state(calldata->state);
3512         nfs_free_seqid(calldata->arg.seqid);
3513         nfs4_put_state_owner(sp);
3514         nfs_sb_deactive(sb);
3515         kfree(calldata);
3516 }
3517
3518 static void nfs4_close_done(struct rpc_task *task, void *data)
3519 {
3520         struct nfs4_closedata *calldata = data;
3521         struct nfs4_state *state = calldata->state;
3522         struct nfs_server *server = NFS_SERVER(calldata->inode);
3523         nfs4_stateid *res_stateid = NULL;
3524         struct nfs4_exception exception = {
3525                 .state = state,
3526                 .inode = calldata->inode,
3527                 .stateid = &calldata->arg.stateid,
3528         };
3529
3530         dprintk("%s: begin!\n", __func__);
3531         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3532                 return;
3533         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3534
3535         /* Handle Layoutreturn errors */
3536         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3537                           &calldata->res.lr_ret) == -EAGAIN)
3538                 goto out_restart;
3539
3540         /* hmm. we are done with the inode, and in the process of freeing
3541          * the state_owner. we keep this around to process errors
3542          */
3543         switch (task->tk_status) {
3544                 case 0:
3545                         res_stateid = &calldata->res.stateid;
3546                         renew_lease(server, calldata->timestamp);
3547                         break;
3548                 case -NFS4ERR_ACCESS:
3549                         if (calldata->arg.bitmask != NULL) {
3550                                 calldata->arg.bitmask = NULL;
3551                                 calldata->res.fattr = NULL;
3552                                 goto out_restart;
3553
3554                         }
3555                         break;
3556                 case -NFS4ERR_OLD_STATEID:
3557                         /* Did we race with OPEN? */
3558                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3559                                                 state))
3560                                 goto out_restart;
3561                         goto out_release;
3562                 case -NFS4ERR_ADMIN_REVOKED:
3563                 case -NFS4ERR_STALE_STATEID:
3564                 case -NFS4ERR_EXPIRED:
3565                         nfs4_free_revoked_stateid(server,
3566                                         &calldata->arg.stateid,
3567                                         task->tk_msg.rpc_cred);
3568                         fallthrough;
3569                 case -NFS4ERR_BAD_STATEID:
3570                         if (calldata->arg.fmode == 0)
3571                                 break;
3572                         fallthrough;
3573                 default:
3574                         task->tk_status = nfs4_async_handle_exception(task,
3575                                         server, task->tk_status, &exception);
3576                         if (exception.retry)
3577                                 goto out_restart;
3578         }
3579         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3580                         res_stateid, calldata->arg.fmode);
3581 out_release:
3582         task->tk_status = 0;
3583         nfs_release_seqid(calldata->arg.seqid);
3584         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3585         dprintk("%s: done, ret = %d!\n", __func__, task->tk_status);
3586         return;
3587 out_restart:
3588         task->tk_status = 0;
3589         rpc_restart_call_prepare(task);
3590         goto out_release;
3591 }
3592
3593 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3594 {
3595         struct nfs4_closedata *calldata = data;
3596         struct nfs4_state *state = calldata->state;
3597         struct inode *inode = calldata->inode;
3598         struct pnfs_layout_hdr *lo;
3599         bool is_rdonly, is_wronly, is_rdwr;
3600         int call_close = 0;
3601
3602         dprintk("%s: begin!\n", __func__);
3603         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3604                 goto out_wait;
3605
3606         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3607         spin_lock(&state->owner->so_lock);
3608         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3609         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3610         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3611         /* Calculate the change in open mode */
3612         calldata->arg.fmode = 0;
3613         if (state->n_rdwr == 0) {
3614                 if (state->n_rdonly == 0)
3615                         call_close |= is_rdonly;
3616                 else if (is_rdonly)
3617                         calldata->arg.fmode |= FMODE_READ;
3618                 if (state->n_wronly == 0)
3619                         call_close |= is_wronly;
3620                 else if (is_wronly)
3621                         calldata->arg.fmode |= FMODE_WRITE;
3622                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3623                         call_close |= is_rdwr;
3624         } else if (is_rdwr)
3625                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3626
3627         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3628         if (!nfs4_valid_open_stateid(state))
3629                 call_close = 0;
3630         spin_unlock(&state->owner->so_lock);
3631
3632         if (!call_close) {
3633                 /* Note: exit _without_ calling nfs4_close_done */
3634                 goto out_no_action;
3635         }
3636
3637         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3638                 nfs_release_seqid(calldata->arg.seqid);
3639                 goto out_wait;
3640         }
3641
3642         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3643         if (lo && !pnfs_layout_is_valid(lo)) {
3644                 calldata->arg.lr_args = NULL;
3645                 calldata->res.lr_res = NULL;
3646         }
3647
3648         if (calldata->arg.fmode == 0)
3649                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3650
3651         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3652                 /* Close-to-open cache consistency revalidation */
3653                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3654                         calldata->arg.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
3655                         nfs4_bitmask_adjust(calldata->arg.bitmask, inode, NFS_SERVER(inode), NULL);
3656                 } else
3657                         calldata->arg.bitmask = NULL;
3658         }
3659
3660         calldata->arg.share_access =
3661                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3662                                 calldata->arg.fmode, 0);
3663
3664         if (calldata->res.fattr == NULL)
3665                 calldata->arg.bitmask = NULL;
3666         else if (calldata->arg.bitmask == NULL)
3667                 calldata->res.fattr = NULL;
3668         calldata->timestamp = jiffies;
3669         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3670                                 &calldata->arg.seq_args,
3671                                 &calldata->res.seq_res,
3672                                 task) != 0)
3673                 nfs_release_seqid(calldata->arg.seqid);
3674         dprintk("%s: done!\n", __func__);
3675         return;
3676 out_no_action:
3677         task->tk_action = NULL;
3678 out_wait:
3679         nfs4_sequence_done(task, &calldata->res.seq_res);
3680 }
3681
3682 static const struct rpc_call_ops nfs4_close_ops = {
3683         .rpc_call_prepare = nfs4_close_prepare,
3684         .rpc_call_done = nfs4_close_done,
3685         .rpc_release = nfs4_free_closedata,
3686 };
3687
3688 /*
3689  * It is possible for data to be read/written from a mem-mapped file
3690  * after the sys_close call (which hits the vfs layer as a flush).
3691  * This means that we can't safely call nfsv4 close on a file until
3692  * the inode is cleared. This in turn means that we are not good
3693  * NFSv4 citizens - we do not indicate to the server to update the file's
3694  * share state even when we are done with one of the three share
3695  * stateid's in the inode.
3696  *
3697  * NOTE: Caller must be holding the sp->so_owner semaphore!
3698  */
3699 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3700 {
3701         struct nfs_server *server = NFS_SERVER(state->inode);
3702         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3703         struct nfs4_closedata *calldata;
3704         struct nfs4_state_owner *sp = state->owner;
3705         struct rpc_task *task;
3706         struct rpc_message msg = {
3707                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3708                 .rpc_cred = state->owner->so_cred,
3709         };
3710         struct rpc_task_setup task_setup_data = {
3711                 .rpc_client = server->client,
3712                 .rpc_message = &msg,
3713                 .callback_ops = &nfs4_close_ops,
3714                 .workqueue = nfsiod_workqueue,
3715                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3716         };
3717         int status = -ENOMEM;
3718
3719         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3720                 &task_setup_data.rpc_client, &msg);
3721
3722         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3723         if (calldata == NULL)
3724                 goto out;
3725         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3726         calldata->inode = state->inode;
3727         calldata->state = state;
3728         calldata->arg.fh = NFS_FH(state->inode);
3729         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3730                 goto out_free_calldata;
3731         /* Serialization for the sequence id */
3732         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3733         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3734         if (IS_ERR(calldata->arg.seqid))
3735                 goto out_free_calldata;
3736         nfs_fattr_init(&calldata->fattr);
3737         calldata->arg.fmode = 0;
3738         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3739         calldata->res.fattr = &calldata->fattr;
3740         calldata->res.seqid = calldata->arg.seqid;
3741         calldata->res.server = server;
3742         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3743         calldata->lr.roc = pnfs_roc(state->inode,
3744                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3745         if (calldata->lr.roc) {
3746                 calldata->arg.lr_args = &calldata->lr.arg;
3747                 calldata->res.lr_res = &calldata->lr.res;
3748         }
3749         nfs_sb_active(calldata->inode->i_sb);
3750
3751         msg.rpc_argp = &calldata->arg;
3752         msg.rpc_resp = &calldata->res;
3753         task_setup_data.callback_data = calldata;
3754         task = rpc_run_task(&task_setup_data);
3755         if (IS_ERR(task))
3756                 return PTR_ERR(task);
3757         status = 0;
3758         if (wait)
3759                 status = rpc_wait_for_completion_task(task);
3760         rpc_put_task(task);
3761         return status;
3762 out_free_calldata:
3763         kfree(calldata);
3764 out:
3765         nfs4_put_open_state(state);
3766         nfs4_put_state_owner(sp);
3767         return status;
3768 }
3769
3770 static struct inode *
3771 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3772                 int open_flags, struct iattr *attr, int *opened)
3773 {
3774         struct nfs4_state *state;
3775         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3776
3777         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3778
3779         /* Protect against concurrent sillydeletes */
3780         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3781
3782         nfs4_label_release_security(label);
3783
3784         if (IS_ERR(state))
3785                 return ERR_CAST(state);
3786         return state->inode;
3787 }
3788
3789 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3790 {
3791         if (ctx->state == NULL)
3792                 return;
3793         if (is_sync)
3794                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3795         else
3796                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3797 }
3798
3799 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3800 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3801 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3802
3803 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3804 {
3805         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3806         struct nfs4_server_caps_arg args = {
3807                 .fhandle = fhandle,
3808                 .bitmask = bitmask,
3809         };
3810         struct nfs4_server_caps_res res = {};
3811         struct rpc_message msg = {
3812                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3813                 .rpc_argp = &args,
3814                 .rpc_resp = &res,
3815         };
3816         int status;
3817         int i;
3818
3819         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3820                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3821                      FATTR4_WORD0_LINK_SUPPORT |
3822                      FATTR4_WORD0_SYMLINK_SUPPORT |
3823                      FATTR4_WORD0_ACLSUPPORT;
3824         if (minorversion)
3825                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3826
3827         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3828         if (status == 0) {
3829                 /* Sanity check the server answers */
3830                 switch (minorversion) {
3831                 case 0:
3832                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3833                         res.attr_bitmask[2] = 0;
3834                         break;
3835                 case 1:
3836                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3837                         break;
3838                 case 2:
3839                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3840                 }
3841                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3842                 server->caps &= ~(NFS_CAP_ACLS|NFS_CAP_HARDLINKS|
3843                                 NFS_CAP_SYMLINKS|NFS_CAP_FILEID|
3844                                 NFS_CAP_MODE|NFS_CAP_NLINK|NFS_CAP_OWNER|
3845                                 NFS_CAP_OWNER_GROUP|NFS_CAP_ATIME|
3846                                 NFS_CAP_CTIME|NFS_CAP_MTIME|
3847                                 NFS_CAP_SECURITY_LABEL);
3848                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3849                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3850                         server->caps |= NFS_CAP_ACLS;
3851                 if (res.has_links != 0)
3852                         server->caps |= NFS_CAP_HARDLINKS;
3853                 if (res.has_symlinks != 0)
3854                         server->caps |= NFS_CAP_SYMLINKS;
3855                 if (res.attr_bitmask[0] & FATTR4_WORD0_FILEID)
3856                         server->caps |= NFS_CAP_FILEID;
3857                 if (res.attr_bitmask[1] & FATTR4_WORD1_MODE)
3858                         server->caps |= NFS_CAP_MODE;
3859                 if (res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS)
3860                         server->caps |= NFS_CAP_NLINK;
3861                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER)
3862                         server->caps |= NFS_CAP_OWNER;
3863                 if (res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP)
3864                         server->caps |= NFS_CAP_OWNER_GROUP;
3865                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS)
3866                         server->caps |= NFS_CAP_ATIME;
3867                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA)
3868                         server->caps |= NFS_CAP_CTIME;
3869                 if (res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY)
3870                         server->caps |= NFS_CAP_MTIME;
3871 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3872                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3873                         server->caps |= NFS_CAP_SECURITY_LABEL;
3874 #endif
3875                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3876                                 sizeof(server->attr_bitmask));
3877                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3878
3879                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3880                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3881                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3882                 server->cache_consistency_bitmask[2] = 0;
3883
3884                 /* Avoid a regression due to buggy server */
3885                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3886                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3887                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3888                         sizeof(server->exclcreat_bitmask));
3889
3890                 server->acl_bitmask = res.acl_bitmask;
3891                 server->fh_expire_type = res.fh_expire_type;
3892         }
3893
3894         return status;
3895 }
3896
3897 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3898 {
3899         struct nfs4_exception exception = {
3900                 .interruptible = true,
3901         };
3902         int err;
3903         do {
3904                 err = nfs4_handle_exception(server,
3905                                 _nfs4_server_capabilities(server, fhandle),
3906                                 &exception);
3907         } while (exception.retry);
3908         return err;
3909 }
3910
3911 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3912                 struct nfs_fsinfo *info)
3913 {
3914         u32 bitmask[3];
3915         struct nfs4_lookup_root_arg args = {
3916                 .bitmask = bitmask,
3917         };
3918         struct nfs4_lookup_res res = {
3919                 .server = server,
3920                 .fattr = info->fattr,
3921                 .fh = fhandle,
3922         };
3923         struct rpc_message msg = {
3924                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
3925                 .rpc_argp = &args,
3926                 .rpc_resp = &res,
3927         };
3928
3929         bitmask[0] = nfs4_fattr_bitmap[0];
3930         bitmask[1] = nfs4_fattr_bitmap[1];
3931         /*
3932          * Process the label in the upcoming getfattr
3933          */
3934         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
3935
3936         nfs_fattr_init(info->fattr);
3937         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3938 }
3939
3940 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
3941                 struct nfs_fsinfo *info)
3942 {
3943         struct nfs4_exception exception = {
3944                 .interruptible = true,
3945         };
3946         int err;
3947         do {
3948                 err = _nfs4_lookup_root(server, fhandle, info);
3949                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
3950                 switch (err) {
3951                 case 0:
3952                 case -NFS4ERR_WRONGSEC:
3953                         goto out;
3954                 default:
3955                         err = nfs4_handle_exception(server, err, &exception);
3956                 }
3957         } while (exception.retry);
3958 out:
3959         return err;
3960 }
3961
3962 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3963                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
3964 {
3965         struct rpc_auth_create_args auth_args = {
3966                 .pseudoflavor = flavor,
3967         };
3968         struct rpc_auth *auth;
3969
3970         auth = rpcauth_create(&auth_args, server->client);
3971         if (IS_ERR(auth))
3972                 return -EACCES;
3973         return nfs4_lookup_root(server, fhandle, info);
3974 }
3975
3976 /*
3977  * Retry pseudoroot lookup with various security flavors.  We do this when:
3978  *
3979  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
3980  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
3981  *
3982  * Returns zero on success, or a negative NFS4ERR value, or a
3983  * negative errno value.
3984  */
3985 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
3986                               struct nfs_fsinfo *info)
3987 {
3988         /* Per 3530bis 15.33.5 */
3989         static const rpc_authflavor_t flav_array[] = {
3990                 RPC_AUTH_GSS_KRB5P,
3991                 RPC_AUTH_GSS_KRB5I,
3992                 RPC_AUTH_GSS_KRB5,
3993                 RPC_AUTH_UNIX,                  /* courtesy */
3994                 RPC_AUTH_NULL,
3995         };
3996         int status = -EPERM;
3997         size_t i;
3998
3999         if (server->auth_info.flavor_len > 0) {
4000                 /* try each flavor specified by user */
4001                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4002                         status = nfs4_lookup_root_sec(server, fhandle, info,
4003                                                 server->auth_info.flavors[i]);
4004                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4005                                 continue;
4006                         break;
4007                 }
4008         } else {
4009                 /* no flavors specified by user, try default list */
4010                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4011                         status = nfs4_lookup_root_sec(server, fhandle, info,
4012                                                       flav_array[i]);
4013                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4014                                 continue;
4015                         break;
4016                 }
4017         }
4018
4019         /*
4020          * -EACCES could mean that the user doesn't have correct permissions
4021          * to access the mount.  It could also mean that we tried to mount
4022          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4023          * existing mount programs don't handle -EACCES very well so it should
4024          * be mapped to -EPERM instead.
4025          */
4026         if (status == -EACCES)
4027                 status = -EPERM;
4028         return status;
4029 }
4030
4031 /**
4032  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4033  * @server: initialized nfs_server handle
4034  * @fhandle: we fill in the pseudo-fs root file handle
4035  * @info: we fill in an FSINFO struct
4036  * @auth_probe: probe the auth flavours
4037  *
4038  * Returns zero on success, or a negative errno.
4039  */
4040 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4041                          struct nfs_fsinfo *info,
4042                          bool auth_probe)
4043 {
4044         int status = 0;
4045
4046         if (!auth_probe)
4047                 status = nfs4_lookup_root(server, fhandle, info);
4048
4049         if (auth_probe || status == NFS4ERR_WRONGSEC)
4050                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4051                                 fhandle, info);
4052
4053         if (status == 0)
4054                 status = nfs4_server_capabilities(server, fhandle);
4055         if (status == 0)
4056                 status = nfs4_do_fsinfo(server, fhandle, info);
4057
4058         return nfs4_map_errors(status);
4059 }
4060
4061 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4062                               struct nfs_fsinfo *info)
4063 {
4064         int error;
4065         struct nfs_fattr *fattr = info->fattr;
4066         struct nfs4_label *label = fattr->label;
4067
4068         error = nfs4_server_capabilities(server, mntfh);
4069         if (error < 0) {
4070                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4071                 return error;
4072         }
4073
4074         error = nfs4_proc_getattr(server, mntfh, fattr, label, NULL);
4075         if (error < 0) {
4076                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4077                 goto out;
4078         }
4079
4080         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4081             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4082                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4083
4084 out:
4085         return error;
4086 }
4087
4088 /*
4089  * Get locations and (maybe) other attributes of a referral.
4090  * Note that we'll actually follow the referral later when
4091  * we detect fsid mismatch in inode revalidation
4092  */
4093 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4094                              const struct qstr *name, struct nfs_fattr *fattr,
4095                              struct nfs_fh *fhandle)
4096 {
4097         int status = -ENOMEM;
4098         struct page *page = NULL;
4099         struct nfs4_fs_locations *locations = NULL;
4100
4101         page = alloc_page(GFP_KERNEL);
4102         if (page == NULL)
4103                 goto out;
4104         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4105         if (locations == NULL)
4106                 goto out;
4107
4108         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4109         if (status != 0)
4110                 goto out;
4111
4112         /*
4113          * If the fsid didn't change, this is a migration event, not a
4114          * referral.  Cause us to drop into the exception handler, which
4115          * will kick off migration recovery.
4116          */
4117         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4118                 dprintk("%s: server did not return a different fsid for"
4119                         " a referral at %s\n", __func__, name->name);
4120                 status = -NFS4ERR_MOVED;
4121                 goto out;
4122         }
4123         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4124         nfs_fixup_referral_attributes(&locations->fattr);
4125
4126         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4127         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4128         memset(fhandle, 0, sizeof(struct nfs_fh));
4129 out:
4130         if (page)
4131                 __free_page(page);
4132         kfree(locations);
4133         return status;
4134 }
4135
4136 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4137                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4138                                 struct inode *inode)
4139 {
4140         __u32 bitmask[NFS4_BITMASK_SZ];
4141         struct nfs4_getattr_arg args = {
4142                 .fh = fhandle,
4143                 .bitmask = bitmask,
4144         };
4145         struct nfs4_getattr_res res = {
4146                 .fattr = fattr,
4147                 .label = label,
4148                 .server = server,
4149         };
4150         struct rpc_message msg = {
4151                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4152                 .rpc_argp = &args,
4153                 .rpc_resp = &res,
4154         };
4155         unsigned short task_flags = 0;
4156
4157         /* Is this is an attribute revalidation, subject to softreval? */
4158         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4159                 task_flags |= RPC_TASK_TIMEOUT;
4160
4161         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
4162
4163         nfs_fattr_init(fattr);
4164         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4165         return nfs4_do_call_sync(server->client, server, &msg,
4166                         &args.seq_args, &res.seq_res, task_flags);
4167 }
4168
4169 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4170                                 struct nfs_fattr *fattr, struct nfs4_label *label,
4171                                 struct inode *inode)
4172 {
4173         struct nfs4_exception exception = {
4174                 .interruptible = true,
4175         };
4176         int err;
4177         do {
4178                 err = _nfs4_proc_getattr(server, fhandle, fattr, label, inode);
4179                 trace_nfs4_getattr(server, fhandle, fattr, err);
4180                 err = nfs4_handle_exception(server, err,
4181                                 &exception);
4182         } while (exception.retry);
4183         return err;
4184 }
4185
4186 /*
4187  * The file is not closed if it is opened due to the a request to change
4188  * the size of the file. The open call will not be needed once the
4189  * VFS layer lookup-intents are implemented.
4190  *
4191  * Close is called when the inode is destroyed.
4192  * If we haven't opened the file for O_WRONLY, we
4193  * need to in the size_change case to obtain a stateid.
4194  *
4195  * Got race?
4196  * Because OPEN is always done by name in nfsv4, it is
4197  * possible that we opened a different file by the same
4198  * name.  We can recognize this race condition, but we
4199  * can't do anything about it besides returning an error.
4200  *
4201  * This will be fixed with VFS changes (lookup-intent).
4202  */
4203 static int
4204 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4205                   struct iattr *sattr)
4206 {
4207         struct inode *inode = d_inode(dentry);
4208         const struct cred *cred = NULL;
4209         struct nfs_open_context *ctx = NULL;
4210         struct nfs4_label *label = NULL;
4211         int status;
4212
4213         if (pnfs_ld_layoutret_on_setattr(inode) &&
4214             sattr->ia_valid & ATTR_SIZE &&
4215             sattr->ia_size < i_size_read(inode))
4216                 pnfs_commit_and_return_layout(inode);
4217
4218         nfs_fattr_init(fattr);
4219
4220         /* Deal with open(O_TRUNC) */
4221         if (sattr->ia_valid & ATTR_OPEN)
4222                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4223
4224         /* Optimization: if the end result is no change, don't RPC */
4225         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4226                 return 0;
4227
4228         /* Search for an existing open(O_WRITE) file */
4229         if (sattr->ia_valid & ATTR_FILE) {
4230
4231                 ctx = nfs_file_open_context(sattr->ia_file);
4232                 if (ctx)
4233                         cred = ctx->cred;
4234         }
4235
4236         label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
4237         if (IS_ERR(label))
4238                 return PTR_ERR(label);
4239
4240         /* Return any delegations if we're going to change ACLs */
4241         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4242                 nfs4_inode_make_writeable(inode);
4243
4244         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL, label);
4245         if (status == 0) {
4246                 nfs_setattr_update_inode(inode, sattr, fattr);
4247                 nfs_setsecurity(inode, fattr, label);
4248         }
4249         nfs4_label_free(label);
4250         return status;
4251 }
4252
4253 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4254                 struct dentry *dentry, struct nfs_fh *fhandle,
4255                 struct nfs_fattr *fattr, struct nfs4_label *label)
4256 {
4257         struct nfs_server *server = NFS_SERVER(dir);
4258         int                    status;
4259         struct nfs4_lookup_arg args = {
4260                 .bitmask = server->attr_bitmask,
4261                 .dir_fh = NFS_FH(dir),
4262                 .name = &dentry->d_name,
4263         };
4264         struct nfs4_lookup_res res = {
4265                 .server = server,
4266                 .fattr = fattr,
4267                 .label = label,
4268                 .fh = fhandle,
4269         };
4270         struct rpc_message msg = {
4271                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4272                 .rpc_argp = &args,
4273                 .rpc_resp = &res,
4274         };
4275         unsigned short task_flags = 0;
4276
4277         /* Is this is an attribute revalidation, subject to softreval? */
4278         if (nfs_lookup_is_soft_revalidate(dentry))
4279                 task_flags |= RPC_TASK_TIMEOUT;
4280
4281         args.bitmask = nfs4_bitmask(server, label);
4282
4283         nfs_fattr_init(fattr);
4284
4285         dprintk("NFS call  lookup %pd2\n", dentry);
4286         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4287         status = nfs4_do_call_sync(clnt, server, &msg,
4288                         &args.seq_args, &res.seq_res, task_flags);
4289         dprintk("NFS reply lookup: %d\n", status);
4290         return status;
4291 }
4292
4293 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4294 {
4295         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4296                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4297         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4298         fattr->nlink = 2;
4299 }
4300
4301 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4302                                    struct dentry *dentry, struct nfs_fh *fhandle,
4303                                    struct nfs_fattr *fattr, struct nfs4_label *label)
4304 {
4305         struct nfs4_exception exception = {
4306                 .interruptible = true,
4307         };
4308         struct rpc_clnt *client = *clnt;
4309         const struct qstr *name = &dentry->d_name;
4310         int err;
4311         do {
4312                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr, label);
4313                 trace_nfs4_lookup(dir, name, err);
4314                 switch (err) {
4315                 case -NFS4ERR_BADNAME:
4316                         err = -ENOENT;
4317                         goto out;
4318                 case -NFS4ERR_MOVED:
4319                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4320                         if (err == -NFS4ERR_MOVED)
4321                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4322                         goto out;
4323                 case -NFS4ERR_WRONGSEC:
4324                         err = -EPERM;
4325                         if (client != *clnt)
4326                                 goto out;
4327                         client = nfs4_negotiate_security(client, dir, name);
4328                         if (IS_ERR(client))
4329                                 return PTR_ERR(client);
4330
4331                         exception.retry = 1;
4332                         break;
4333                 default:
4334                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4335                 }
4336         } while (exception.retry);
4337
4338 out:
4339         if (err == 0)
4340                 *clnt = client;
4341         else if (client != *clnt)
4342                 rpc_shutdown_client(client);
4343
4344         return err;
4345 }
4346
4347 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4348                             struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4349                             struct nfs4_label *label)
4350 {
4351         int status;
4352         struct rpc_clnt *client = NFS_CLIENT(dir);
4353
4354         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, label);
4355         if (client != NFS_CLIENT(dir)) {
4356                 rpc_shutdown_client(client);
4357                 nfs_fixup_secinfo_attributes(fattr);
4358         }
4359         return status;
4360 }
4361
4362 struct rpc_clnt *
4363 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4364                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4365 {
4366         struct rpc_clnt *client = NFS_CLIENT(dir);
4367         int status;
4368
4369         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr, NULL);
4370         if (status < 0)
4371                 return ERR_PTR(status);
4372         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4373 }
4374
4375 static int _nfs4_proc_lookupp(struct inode *inode,
4376                 struct nfs_fh *fhandle, struct nfs_fattr *fattr,
4377                 struct nfs4_label *label)
4378 {
4379         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4380         struct nfs_server *server = NFS_SERVER(inode);
4381         int                    status;
4382         struct nfs4_lookupp_arg args = {
4383                 .bitmask = server->attr_bitmask,
4384                 .fh = NFS_FH(inode),
4385         };
4386         struct nfs4_lookupp_res res = {
4387                 .server = server,
4388                 .fattr = fattr,
4389                 .label = label,
4390                 .fh = fhandle,
4391         };
4392         struct rpc_message msg = {
4393                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4394                 .rpc_argp = &args,
4395                 .rpc_resp = &res,
4396         };
4397         unsigned short task_flags = 0;
4398
4399         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4400                 task_flags |= RPC_TASK_TIMEOUT;
4401
4402         args.bitmask = nfs4_bitmask(server, label);
4403
4404         nfs_fattr_init(fattr);
4405
4406         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4407         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4408                                 &res.seq_res, task_flags);
4409         dprintk("NFS reply lookupp: %d\n", status);
4410         return status;
4411 }
4412
4413 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4414                              struct nfs_fattr *fattr, struct nfs4_label *label)
4415 {
4416         struct nfs4_exception exception = {
4417                 .interruptible = true,
4418         };
4419         int err;
4420         do {
4421                 err = _nfs4_proc_lookupp(inode, fhandle, fattr, label);
4422                 trace_nfs4_lookupp(inode, err);
4423                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4424                                 &exception);
4425         } while (exception.retry);
4426         return err;
4427 }
4428
4429 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4430 {
4431         struct nfs_server *server = NFS_SERVER(inode);
4432         struct nfs4_accessargs args = {
4433                 .fh = NFS_FH(inode),
4434                 .access = entry->mask,
4435         };
4436         struct nfs4_accessres res = {
4437                 .server = server,
4438         };
4439         struct rpc_message msg = {
4440                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4441                 .rpc_argp = &args,
4442                 .rpc_resp = &res,
4443                 .rpc_cred = entry->cred,
4444         };
4445         int status = 0;
4446
4447         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4448                 res.fattr = nfs_alloc_fattr();
4449                 if (res.fattr == NULL)
4450                         return -ENOMEM;
4451                 args.bitmask = server->cache_consistency_bitmask;
4452         }
4453         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4454         if (!status) {
4455                 nfs_access_set_mask(entry, res.access);
4456                 if (res.fattr)
4457                         nfs_refresh_inode(inode, res.fattr);
4458         }
4459         nfs_free_fattr(res.fattr);
4460         return status;
4461 }
4462
4463 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry)
4464 {
4465         struct nfs4_exception exception = {
4466                 .interruptible = true,
4467         };
4468         int err;
4469         do {
4470                 err = _nfs4_proc_access(inode, entry);
4471                 trace_nfs4_access(inode, err);
4472                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4473                                 &exception);
4474         } while (exception.retry);
4475         return err;
4476 }
4477
4478 /*
4479  * TODO: For the time being, we don't try to get any attributes
4480  * along with any of the zero-copy operations READ, READDIR,
4481  * READLINK, WRITE.
4482  *
4483  * In the case of the first three, we want to put the GETATTR
4484  * after the read-type operation -- this is because it is hard
4485  * to predict the length of a GETATTR response in v4, and thus
4486  * align the READ data correctly.  This means that the GETATTR
4487  * may end up partially falling into the page cache, and we should
4488  * shift it into the 'tail' of the xdr_buf before processing.
4489  * To do this efficiently, we need to know the total length
4490  * of data received, which doesn't seem to be available outside
4491  * of the RPC layer.
4492  *
4493  * In the case of WRITE, we also want to put the GETATTR after
4494  * the operation -- in this case because we want to make sure
4495  * we get the post-operation mtime and size.
4496  *
4497  * Both of these changes to the XDR layer would in fact be quite
4498  * minor, but I decided to leave them for a subsequent patch.
4499  */
4500 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4501                 unsigned int pgbase, unsigned int pglen)
4502 {
4503         struct nfs4_readlink args = {
4504                 .fh       = NFS_FH(inode),
4505                 .pgbase   = pgbase,
4506                 .pglen    = pglen,
4507                 .pages    = &page,
4508         };
4509         struct nfs4_readlink_res res;
4510         struct rpc_message msg = {
4511                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4512                 .rpc_argp = &args,
4513                 .rpc_resp = &res,
4514         };
4515
4516         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4517 }
4518
4519 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4520                 unsigned int pgbase, unsigned int pglen)
4521 {
4522         struct nfs4_exception exception = {
4523                 .interruptible = true,
4524         };
4525         int err;
4526         do {
4527                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4528                 trace_nfs4_readlink(inode, err);
4529                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4530                                 &exception);
4531         } while (exception.retry);
4532         return err;
4533 }
4534
4535 /*
4536  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4537  */
4538 static int
4539 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4540                  int flags)
4541 {
4542         struct nfs_server *server = NFS_SERVER(dir);
4543         struct nfs4_label l, *ilabel = NULL;
4544         struct nfs_open_context *ctx;
4545         struct nfs4_state *state;
4546         int status = 0;
4547
4548         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4549         if (IS_ERR(ctx))
4550                 return PTR_ERR(ctx);
4551
4552         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4553
4554         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4555                 sattr->ia_mode &= ~current_umask();
4556         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4557         if (IS_ERR(state)) {
4558                 status = PTR_ERR(state);
4559                 goto out;
4560         }
4561 out:
4562         nfs4_label_release_security(ilabel);
4563         put_nfs_open_context(ctx);
4564         return status;
4565 }
4566
4567 static int
4568 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4569 {
4570         struct nfs_server *server = NFS_SERVER(dir);
4571         struct nfs_removeargs args = {
4572                 .fh = NFS_FH(dir),
4573                 .name = *name,
4574         };
4575         struct nfs_removeres res = {
4576                 .server = server,
4577         };
4578         struct rpc_message msg = {
4579                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4580                 .rpc_argp = &args,
4581                 .rpc_resp = &res,
4582         };
4583         unsigned long timestamp = jiffies;
4584         int status;
4585
4586         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4587         if (status == 0) {
4588                 spin_lock(&dir->i_lock);
4589                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4590                                               NFS_INO_INVALID_DATA);
4591                 /* Removing a directory decrements nlink in the parent */
4592                 if (ftype == NF4DIR && dir->i_nlink > 2)
4593                         nfs4_dec_nlink_locked(dir);
4594                 spin_unlock(&dir->i_lock);
4595         }
4596         return status;
4597 }
4598
4599 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4600 {
4601         struct nfs4_exception exception = {
4602                 .interruptible = true,
4603         };
4604         struct inode *inode = d_inode(dentry);
4605         int err;
4606
4607         if (inode) {
4608                 if (inode->i_nlink == 1)
4609                         nfs4_inode_return_delegation(inode);
4610                 else
4611                         nfs4_inode_make_writeable(inode);
4612         }
4613         do {
4614                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4615                 trace_nfs4_remove(dir, &dentry->d_name, err);
4616                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4617                                 &exception);
4618         } while (exception.retry);
4619         return err;
4620 }
4621
4622 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4623 {
4624         struct nfs4_exception exception = {
4625                 .interruptible = true,
4626         };
4627         int err;
4628
4629         do {
4630                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4631                 trace_nfs4_remove(dir, name, err);
4632                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4633                                 &exception);
4634         } while (exception.retry);
4635         return err;
4636 }
4637
4638 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4639                 struct dentry *dentry,
4640                 struct inode *inode)
4641 {
4642         struct nfs_removeargs *args = msg->rpc_argp;
4643         struct nfs_removeres *res = msg->rpc_resp;
4644
4645         res->server = NFS_SB(dentry->d_sb);
4646         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4647         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4648
4649         nfs_fattr_init(res->dir_attr);
4650
4651         if (inode)
4652                 nfs4_inode_return_delegation(inode);
4653 }
4654
4655 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4656 {
4657         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4658                         &data->args.seq_args,
4659                         &data->res.seq_res,
4660                         task);
4661 }
4662
4663 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4664 {
4665         struct nfs_unlinkdata *data = task->tk_calldata;
4666         struct nfs_removeres *res = &data->res;
4667
4668         if (!nfs4_sequence_done(task, &res->seq_res))
4669                 return 0;
4670         if (nfs4_async_handle_error(task, res->server, NULL,
4671                                     &data->timeout) == -EAGAIN)
4672                 return 0;
4673         if (task->tk_status == 0)
4674                 nfs4_update_changeattr(dir, &res->cinfo,
4675                                 res->dir_attr->time_start,
4676                                 NFS_INO_INVALID_DATA);
4677         return 1;
4678 }
4679
4680 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4681                 struct dentry *old_dentry,
4682                 struct dentry *new_dentry)
4683 {
4684         struct nfs_renameargs *arg = msg->rpc_argp;
4685         struct nfs_renameres *res = msg->rpc_resp;
4686         struct inode *old_inode = d_inode(old_dentry);
4687         struct inode *new_inode = d_inode(new_dentry);
4688
4689         if (old_inode)
4690                 nfs4_inode_make_writeable(old_inode);
4691         if (new_inode)
4692                 nfs4_inode_return_delegation(new_inode);
4693         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4694         res->server = NFS_SB(old_dentry->d_sb);
4695         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4696 }
4697
4698 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4699 {
4700         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4701                         &data->args.seq_args,
4702                         &data->res.seq_res,
4703                         task);
4704 }
4705
4706 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4707                                  struct inode *new_dir)
4708 {
4709         struct nfs_renamedata *data = task->tk_calldata;
4710         struct nfs_renameres *res = &data->res;
4711
4712         if (!nfs4_sequence_done(task, &res->seq_res))
4713                 return 0;
4714         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4715                 return 0;
4716
4717         if (task->tk_status == 0) {
4718                 if (new_dir != old_dir) {
4719                         /* Note: If we moved a directory, nlink will change */
4720                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4721                                         res->old_fattr->time_start,
4722                                         NFS_INO_INVALID_OTHER |
4723                                             NFS_INO_INVALID_DATA);
4724                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4725                                         res->new_fattr->time_start,
4726                                         NFS_INO_INVALID_OTHER |
4727                                             NFS_INO_INVALID_DATA);
4728                 } else
4729                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4730                                         res->old_fattr->time_start,
4731                                         NFS_INO_INVALID_DATA);
4732         }
4733         return 1;
4734 }
4735
4736 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4737 {
4738         struct nfs_server *server = NFS_SERVER(inode);
4739         __u32 bitmask[NFS4_BITMASK_SZ];
4740         struct nfs4_link_arg arg = {
4741                 .fh     = NFS_FH(inode),
4742                 .dir_fh = NFS_FH(dir),
4743                 .name   = name,
4744                 .bitmask = bitmask,
4745         };
4746         struct nfs4_link_res res = {
4747                 .server = server,
4748                 .label = NULL,
4749         };
4750         struct rpc_message msg = {
4751                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4752                 .rpc_argp = &arg,
4753                 .rpc_resp = &res,
4754         };
4755         int status = -ENOMEM;
4756
4757         res.fattr = nfs_alloc_fattr();
4758         if (res.fattr == NULL)
4759                 goto out;
4760
4761         res.label = nfs4_label_alloc(server, GFP_KERNEL);
4762         if (IS_ERR(res.label)) {
4763                 status = PTR_ERR(res.label);
4764                 goto out;
4765         }
4766
4767         nfs4_inode_make_writeable(inode);
4768         nfs4_bitmap_copy_adjust_setattr(bitmask, nfs4_bitmask(server, res.label), inode);
4769
4770         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4771         if (!status) {
4772                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4773                                        NFS_INO_INVALID_DATA);
4774                 status = nfs_post_op_update_inode(inode, res.fattr);
4775                 if (!status)
4776                         nfs_setsecurity(inode, res.fattr, res.label);
4777         }
4778
4779
4780         nfs4_label_free(res.label);
4781
4782 out:
4783         nfs_free_fattr(res.fattr);
4784         return status;
4785 }
4786
4787 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4788 {
4789         struct nfs4_exception exception = {
4790                 .interruptible = true,
4791         };
4792         int err;
4793         do {
4794                 err = nfs4_handle_exception(NFS_SERVER(inode),
4795                                 _nfs4_proc_link(inode, dir, name),
4796                                 &exception);
4797         } while (exception.retry);
4798         return err;
4799 }
4800
4801 struct nfs4_createdata {
4802         struct rpc_message msg;
4803         struct nfs4_create_arg arg;
4804         struct nfs4_create_res res;
4805         struct nfs_fh fh;
4806         struct nfs_fattr fattr;
4807         struct nfs4_label *label;
4808 };
4809
4810 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4811                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4812 {
4813         struct nfs4_createdata *data;
4814
4815         data = kzalloc(sizeof(*data), GFP_KERNEL);
4816         if (data != NULL) {
4817                 struct nfs_server *server = NFS_SERVER(dir);
4818
4819                 data->label = nfs4_label_alloc(server, GFP_KERNEL);
4820                 if (IS_ERR(data->label))
4821                         goto out_free;
4822
4823                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4824                 data->msg.rpc_argp = &data->arg;
4825                 data->msg.rpc_resp = &data->res;
4826                 data->arg.dir_fh = NFS_FH(dir);
4827                 data->arg.server = server;
4828                 data->arg.name = name;
4829                 data->arg.attrs = sattr;
4830                 data->arg.ftype = ftype;
4831                 data->arg.bitmask = nfs4_bitmask(server, data->label);
4832                 data->arg.umask = current_umask();
4833                 data->res.server = server;
4834                 data->res.fh = &data->fh;
4835                 data->res.fattr = &data->fattr;
4836                 data->res.label = data->label;
4837                 nfs_fattr_init(data->res.fattr);
4838         }
4839         return data;
4840 out_free:
4841         kfree(data);
4842         return NULL;
4843 }
4844
4845 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4846 {
4847         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4848                                     &data->arg.seq_args, &data->res.seq_res, 1);
4849         if (status == 0) {
4850                 spin_lock(&dir->i_lock);
4851                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4852                                 data->res.fattr->time_start,
4853                                 NFS_INO_INVALID_DATA);
4854                 /* Creating a directory bumps nlink in the parent */
4855                 if (data->arg.ftype == NF4DIR)
4856                         nfs4_inc_nlink_locked(dir);
4857                 spin_unlock(&dir->i_lock);
4858                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr, data->res.label);
4859         }
4860         return status;
4861 }
4862
4863 static void nfs4_free_createdata(struct nfs4_createdata *data)
4864 {
4865         nfs4_label_free(data->label);
4866         kfree(data);
4867 }
4868
4869 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4870                 struct page *page, unsigned int len, struct iattr *sattr,
4871                 struct nfs4_label *label)
4872 {
4873         struct nfs4_createdata *data;
4874         int status = -ENAMETOOLONG;
4875
4876         if (len > NFS4_MAXPATHLEN)
4877                 goto out;
4878
4879         status = -ENOMEM;
4880         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
4881         if (data == NULL)
4882                 goto out;
4883
4884         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
4885         data->arg.u.symlink.pages = &page;
4886         data->arg.u.symlink.len = len;
4887         data->arg.label = label;
4888
4889         status = nfs4_do_create(dir, dentry, data);
4890
4891         nfs4_free_createdata(data);
4892 out:
4893         return status;
4894 }
4895
4896 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4897                 struct page *page, unsigned int len, struct iattr *sattr)
4898 {
4899         struct nfs4_exception exception = {
4900                 .interruptible = true,
4901         };
4902         struct nfs4_label l, *label = NULL;
4903         int err;
4904
4905         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4906
4907         do {
4908                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
4909                 trace_nfs4_symlink(dir, &dentry->d_name, err);
4910                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4911                                 &exception);
4912         } while (exception.retry);
4913
4914         nfs4_label_release_security(label);
4915         return err;
4916 }
4917
4918 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4919                 struct iattr *sattr, struct nfs4_label *label)
4920 {
4921         struct nfs4_createdata *data;
4922         int status = -ENOMEM;
4923
4924         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
4925         if (data == NULL)
4926                 goto out;
4927
4928         data->arg.label = label;
4929         status = nfs4_do_create(dir, dentry, data);
4930
4931         nfs4_free_createdata(data);
4932 out:
4933         return status;
4934 }
4935
4936 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
4937                 struct iattr *sattr)
4938 {
4939         struct nfs_server *server = NFS_SERVER(dir);
4940         struct nfs4_exception exception = {
4941                 .interruptible = true,
4942         };
4943         struct nfs4_label l, *label = NULL;
4944         int err;
4945
4946         label = nfs4_label_init_security(dir, dentry, sattr, &l);
4947
4948         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4949                 sattr->ia_mode &= ~current_umask();
4950         do {
4951                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
4952                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
4953                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4954                                 &exception);
4955         } while (exception.retry);
4956         nfs4_label_release_security(label);
4957
4958         return err;
4959 }
4960
4961 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
4962                               struct nfs_readdir_res *nr_res)
4963 {
4964         struct inode            *dir = d_inode(nr_arg->dentry);
4965         struct nfs_server       *server = NFS_SERVER(dir);
4966         struct nfs4_readdir_arg args = {
4967                 .fh = NFS_FH(dir),
4968                 .pages = nr_arg->pages,
4969                 .pgbase = 0,
4970                 .count = nr_arg->page_len,
4971                 .plus = nr_arg->plus,
4972         };
4973         struct nfs4_readdir_res res;
4974         struct rpc_message msg = {
4975                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
4976                 .rpc_argp = &args,
4977                 .rpc_resp = &res,
4978                 .rpc_cred = nr_arg->cred,
4979         };
4980         int                     status;
4981
4982         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
4983                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
4984         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
4985                 args.bitmask = server->attr_bitmask_nl;
4986         else
4987                 args.bitmask = server->attr_bitmask;
4988
4989         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
4990         res.pgbase = args.pgbase;
4991         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
4992                         &res.seq_res, 0);
4993         if (status >= 0) {
4994                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
4995                 status += args.pgbase;
4996         }
4997
4998         nfs_invalidate_atime(dir);
4999
5000         dprintk("%s: returns %d\n", __func__, status);
5001         return status;
5002 }
5003
5004 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5005                              struct nfs_readdir_res *res)
5006 {
5007         struct nfs4_exception exception = {
5008                 .interruptible = true,
5009         };
5010         int err;
5011         do {
5012                 err = _nfs4_proc_readdir(arg, res);
5013                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5014                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5015                                             err, &exception);
5016         } while (exception.retry);
5017         return err;
5018 }
5019
5020 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5021                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5022 {
5023         struct nfs4_createdata *data;
5024         int mode = sattr->ia_mode;
5025         int status = -ENOMEM;
5026
5027         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5028         if (data == NULL)
5029                 goto out;
5030
5031         if (S_ISFIFO(mode))
5032                 data->arg.ftype = NF4FIFO;
5033         else if (S_ISBLK(mode)) {
5034                 data->arg.ftype = NF4BLK;
5035                 data->arg.u.device.specdata1 = MAJOR(rdev);
5036                 data->arg.u.device.specdata2 = MINOR(rdev);
5037         }
5038         else if (S_ISCHR(mode)) {
5039                 data->arg.ftype = NF4CHR;
5040                 data->arg.u.device.specdata1 = MAJOR(rdev);
5041                 data->arg.u.device.specdata2 = MINOR(rdev);
5042         } else if (!S_ISSOCK(mode)) {
5043                 status = -EINVAL;
5044                 goto out_free;
5045         }
5046
5047         data->arg.label = label;
5048         status = nfs4_do_create(dir, dentry, data);
5049 out_free:
5050         nfs4_free_createdata(data);
5051 out:
5052         return status;
5053 }
5054
5055 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5056                 struct iattr *sattr, dev_t rdev)
5057 {
5058         struct nfs_server *server = NFS_SERVER(dir);
5059         struct nfs4_exception exception = {
5060                 .interruptible = true,
5061         };
5062         struct nfs4_label l, *label = NULL;
5063         int err;
5064
5065         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5066
5067         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5068                 sattr->ia_mode &= ~current_umask();
5069         do {
5070                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5071                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5072                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5073                                 &exception);
5074         } while (exception.retry);
5075
5076         nfs4_label_release_security(label);
5077
5078         return err;
5079 }
5080
5081 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5082                  struct nfs_fsstat *fsstat)
5083 {
5084         struct nfs4_statfs_arg args = {
5085                 .fh = fhandle,
5086                 .bitmask = server->attr_bitmask,
5087         };
5088         struct nfs4_statfs_res res = {
5089                 .fsstat = fsstat,
5090         };
5091         struct rpc_message msg = {
5092                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5093                 .rpc_argp = &args,
5094                 .rpc_resp = &res,
5095         };
5096
5097         nfs_fattr_init(fsstat->fattr);
5098         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5099 }
5100
5101 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5102 {
5103         struct nfs4_exception exception = {
5104                 .interruptible = true,
5105         };
5106         int err;
5107         do {
5108                 err = nfs4_handle_exception(server,
5109                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5110                                 &exception);
5111         } while (exception.retry);
5112         return err;
5113 }
5114
5115 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5116                 struct nfs_fsinfo *fsinfo)
5117 {
5118         struct nfs4_fsinfo_arg args = {
5119                 .fh = fhandle,
5120                 .bitmask = server->attr_bitmask,
5121         };
5122         struct nfs4_fsinfo_res res = {
5123                 .fsinfo = fsinfo,
5124         };
5125         struct rpc_message msg = {
5126                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5127                 .rpc_argp = &args,
5128                 .rpc_resp = &res,
5129         };
5130
5131         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5132 }
5133
5134 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5135 {
5136         struct nfs4_exception exception = {
5137                 .interruptible = true,
5138         };
5139         int err;
5140
5141         do {
5142                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5143                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5144                 if (err == 0) {
5145                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5146                         break;
5147                 }
5148                 err = nfs4_handle_exception(server, err, &exception);
5149         } while (exception.retry);
5150         return err;
5151 }
5152
5153 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5154 {
5155         int error;
5156
5157         nfs_fattr_init(fsinfo->fattr);
5158         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5159         if (error == 0) {
5160                 /* block layout checks this! */
5161                 server->pnfs_blksize = fsinfo->blksize;
5162                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5163         }
5164
5165         return error;
5166 }
5167
5168 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5169                 struct nfs_pathconf *pathconf)
5170 {
5171         struct nfs4_pathconf_arg args = {
5172                 .fh = fhandle,
5173                 .bitmask = server->attr_bitmask,
5174         };
5175         struct nfs4_pathconf_res res = {
5176                 .pathconf = pathconf,
5177         };
5178         struct rpc_message msg = {
5179                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5180                 .rpc_argp = &args,
5181                 .rpc_resp = &res,
5182         };
5183
5184         /* None of the pathconf attributes are mandatory to implement */
5185         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5186                 memset(pathconf, 0, sizeof(*pathconf));
5187                 return 0;
5188         }
5189
5190         nfs_fattr_init(pathconf->fattr);
5191         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5192 }
5193
5194 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5195                 struct nfs_pathconf *pathconf)
5196 {
5197         struct nfs4_exception exception = {
5198                 .interruptible = true,
5199         };
5200         int err;
5201
5202         do {
5203                 err = nfs4_handle_exception(server,
5204                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5205                                 &exception);
5206         } while (exception.retry);
5207         return err;
5208 }
5209
5210 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5211                 const struct nfs_open_context *ctx,
5212                 const struct nfs_lock_context *l_ctx,
5213                 fmode_t fmode)
5214 {
5215         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5216 }
5217 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5218
5219 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5220                 const struct nfs_open_context *ctx,
5221                 const struct nfs_lock_context *l_ctx,
5222                 fmode_t fmode)
5223 {
5224         nfs4_stateid _current_stateid;
5225
5226         /* If the current stateid represents a lost lock, then exit */
5227         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5228                 return true;
5229         return nfs4_stateid_match(stateid, &_current_stateid);
5230 }
5231
5232 static bool nfs4_error_stateid_expired(int err)
5233 {
5234         switch (err) {
5235         case -NFS4ERR_DELEG_REVOKED:
5236         case -NFS4ERR_ADMIN_REVOKED:
5237         case -NFS4ERR_BAD_STATEID:
5238         case -NFS4ERR_STALE_STATEID:
5239         case -NFS4ERR_OLD_STATEID:
5240         case -NFS4ERR_OPENMODE:
5241         case -NFS4ERR_EXPIRED:
5242                 return true;
5243         }
5244         return false;
5245 }
5246
5247 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5248 {
5249         struct nfs_server *server = NFS_SERVER(hdr->inode);
5250
5251         trace_nfs4_read(hdr, task->tk_status);
5252         if (task->tk_status < 0) {
5253                 struct nfs4_exception exception = {
5254                         .inode = hdr->inode,
5255                         .state = hdr->args.context->state,
5256                         .stateid = &hdr->args.stateid,
5257                 };
5258                 task->tk_status = nfs4_async_handle_exception(task,
5259                                 server, task->tk_status, &exception);
5260                 if (exception.retry) {
5261                         rpc_restart_call_prepare(task);
5262                         return -EAGAIN;
5263                 }
5264         }
5265
5266         if (task->tk_status > 0)
5267                 renew_lease(server, hdr->timestamp);
5268         return 0;
5269 }
5270
5271 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5272                 struct nfs_pgio_args *args)
5273 {
5274
5275         if (!nfs4_error_stateid_expired(task->tk_status) ||
5276                 nfs4_stateid_is_current(&args->stateid,
5277                                 args->context,
5278                                 args->lock_context,
5279                                 FMODE_READ))
5280                 return false;
5281         rpc_restart_call_prepare(task);
5282         return true;
5283 }
5284
5285 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5286                                          struct nfs_pgio_header *hdr)
5287 {
5288         struct nfs_server *server = NFS_SERVER(hdr->inode);
5289         struct rpc_message *msg = &task->tk_msg;
5290
5291         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5292             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5293                 server->caps &= ~NFS_CAP_READ_PLUS;
5294                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5295                 rpc_restart_call_prepare(task);
5296                 return true;
5297         }
5298         return false;
5299 }
5300
5301 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5302 {
5303         dprintk("--> %s\n", __func__);
5304
5305         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5306                 return -EAGAIN;
5307         if (nfs4_read_stateid_changed(task, &hdr->args))
5308                 return -EAGAIN;
5309         if (nfs4_read_plus_not_supported(task, hdr))
5310                 return -EAGAIN;
5311         if (task->tk_status > 0)
5312                 nfs_invalidate_atime(hdr->inode);
5313         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5314                                     nfs4_read_done_cb(task, hdr);
5315 }
5316
5317 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5318 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5319                                     struct rpc_message *msg)
5320 {
5321         /* Note: We don't use READ_PLUS with pNFS yet */
5322         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5323                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5324 }
5325 #else
5326 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5327                                     struct rpc_message *msg)
5328 {
5329 }
5330 #endif /* CONFIG_NFS_V4_2 */
5331
5332 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5333                                  struct rpc_message *msg)
5334 {
5335         hdr->timestamp   = jiffies;
5336         if (!hdr->pgio_done_cb)
5337                 hdr->pgio_done_cb = nfs4_read_done_cb;
5338         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5339         nfs42_read_plus_support(hdr, msg);
5340         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5341 }
5342
5343 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5344                                       struct nfs_pgio_header *hdr)
5345 {
5346         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5347                         &hdr->args.seq_args,
5348                         &hdr->res.seq_res,
5349                         task))
5350                 return 0;
5351         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5352                                 hdr->args.lock_context,
5353                                 hdr->rw_mode) == -EIO)
5354                 return -EIO;
5355         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5356                 return -EIO;
5357         return 0;
5358 }
5359
5360 static int nfs4_write_done_cb(struct rpc_task *task,
5361                               struct nfs_pgio_header *hdr)
5362 {
5363         struct inode *inode = hdr->inode;
5364
5365         trace_nfs4_write(hdr, task->tk_status);
5366         if (task->tk_status < 0) {
5367                 struct nfs4_exception exception = {
5368                         .inode = hdr->inode,
5369                         .state = hdr->args.context->state,
5370                         .stateid = &hdr->args.stateid,
5371                 };
5372                 task->tk_status = nfs4_async_handle_exception(task,
5373                                 NFS_SERVER(inode), task->tk_status,
5374                                 &exception);
5375                 if (exception.retry) {
5376                         rpc_restart_call_prepare(task);
5377                         return -EAGAIN;
5378                 }
5379         }
5380         if (task->tk_status >= 0) {
5381                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5382                 nfs_writeback_update_inode(hdr);
5383         }
5384         return 0;
5385 }
5386
5387 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5388                 struct nfs_pgio_args *args)
5389 {
5390
5391         if (!nfs4_error_stateid_expired(task->tk_status) ||
5392                 nfs4_stateid_is_current(&args->stateid,
5393                                 args->context,
5394                                 args->lock_context,
5395                                 FMODE_WRITE))
5396                 return false;
5397         rpc_restart_call_prepare(task);
5398         return true;
5399 }
5400
5401 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5402 {
5403         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5404                 return -EAGAIN;
5405         if (nfs4_write_stateid_changed(task, &hdr->args))
5406                 return -EAGAIN;
5407         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5408                 nfs4_write_done_cb(task, hdr);
5409 }
5410
5411 static
5412 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5413 {
5414         /* Don't request attributes for pNFS or O_DIRECT writes */
5415         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5416                 return false;
5417         /* Otherwise, request attributes if and only if we don't hold
5418          * a delegation
5419          */
5420         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5421 }
5422
5423 static void nfs4_bitmask_adjust(__u32 *bitmask, struct inode *inode,
5424                                 struct nfs_server *server,
5425                                 struct nfs4_label *label)
5426 {
5427
5428         unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
5429
5430         if ((cache_validity & NFS_INO_INVALID_DATA) ||
5431                 (cache_validity & NFS_INO_REVAL_PAGECACHE) ||
5432                 (cache_validity & NFS_INO_REVAL_FORCED) ||
5433                 (cache_validity & NFS_INO_INVALID_OTHER))
5434                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, label), inode);
5435
5436         if (cache_validity & NFS_INO_INVALID_ATIME)
5437                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5438         if (cache_validity & NFS_INO_INVALID_OTHER)
5439                 bitmask[1] |= FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER |
5440                                 FATTR4_WORD1_OWNER_GROUP |
5441                                 FATTR4_WORD1_NUMLINKS;
5442         if (label && label->len && cache_validity & NFS_INO_INVALID_LABEL)
5443                 bitmask[2] |= FATTR4_WORD2_SECURITY_LABEL;
5444         if (cache_validity & NFS_INO_INVALID_CHANGE)
5445                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5446         if (cache_validity & NFS_INO_INVALID_CTIME)
5447                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5448         if (cache_validity & NFS_INO_INVALID_MTIME)
5449                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5450         if (cache_validity & NFS_INO_INVALID_SIZE)
5451                 bitmask[0] |= FATTR4_WORD0_SIZE;
5452         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5453                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5454 }
5455
5456 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5457                                   struct rpc_message *msg,
5458                                   struct rpc_clnt **clnt)
5459 {
5460         struct nfs_server *server = NFS_SERVER(hdr->inode);
5461
5462         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5463                 hdr->args.bitmask = NULL;
5464                 hdr->res.fattr = NULL;
5465         } else {
5466                 hdr->args.bitmask = server->cache_consistency_bitmask;
5467                 nfs4_bitmask_adjust(hdr->args.bitmask, hdr->inode, server, NULL);
5468         }
5469
5470         if (!hdr->pgio_done_cb)
5471                 hdr->pgio_done_cb = nfs4_write_done_cb;
5472         hdr->res.server = server;
5473         hdr->timestamp   = jiffies;
5474
5475         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5476         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5477         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5478 }
5479
5480 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5481 {
5482         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5483                         &data->args.seq_args,
5484                         &data->res.seq_res,
5485                         task);
5486 }
5487
5488 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5489 {
5490         struct inode *inode = data->inode;
5491
5492         trace_nfs4_commit(data, task->tk_status);
5493         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5494                                     NULL, NULL) == -EAGAIN) {
5495                 rpc_restart_call_prepare(task);
5496                 return -EAGAIN;
5497         }
5498         return 0;
5499 }
5500
5501 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5502 {
5503         if (!nfs4_sequence_done(task, &data->res.seq_res))
5504                 return -EAGAIN;
5505         return data->commit_done_cb(task, data);
5506 }
5507
5508 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5509                                    struct rpc_clnt **clnt)
5510 {
5511         struct nfs_server *server = NFS_SERVER(data->inode);
5512
5513         if (data->commit_done_cb == NULL)
5514                 data->commit_done_cb = nfs4_commit_done_cb;
5515         data->res.server = server;
5516         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5517         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5518         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5519 }
5520
5521 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5522                                 struct nfs_commitres *res)
5523 {
5524         struct inode *dst_inode = file_inode(dst);
5525         struct nfs_server *server = NFS_SERVER(dst_inode);
5526         struct rpc_message msg = {
5527                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5528                 .rpc_argp = args,
5529                 .rpc_resp = res,
5530         };
5531
5532         args->fh = NFS_FH(dst_inode);
5533         return nfs4_call_sync(server->client, server, &msg,
5534                         &args->seq_args, &res->seq_res, 1);
5535 }
5536
5537 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5538 {
5539         struct nfs_commitargs args = {
5540                 .offset = offset,
5541                 .count = count,
5542         };
5543         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5544         struct nfs4_exception exception = { };
5545         int status;
5546
5547         do {
5548                 status = _nfs4_proc_commit(dst, &args, res);
5549                 status = nfs4_handle_exception(dst_server, status, &exception);
5550         } while (exception.retry);
5551
5552         return status;
5553 }
5554
5555 struct nfs4_renewdata {
5556         struct nfs_client       *client;
5557         unsigned long           timestamp;
5558 };
5559
5560 /*
5561  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5562  * standalone procedure for queueing an asynchronous RENEW.
5563  */
5564 static void nfs4_renew_release(void *calldata)
5565 {
5566         struct nfs4_renewdata *data = calldata;
5567         struct nfs_client *clp = data->client;
5568
5569         if (refcount_read(&clp->cl_count) > 1)
5570                 nfs4_schedule_state_renewal(clp);
5571         nfs_put_client(clp);
5572         kfree(data);
5573 }
5574
5575 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5576 {
5577         struct nfs4_renewdata *data = calldata;
5578         struct nfs_client *clp = data->client;
5579         unsigned long timestamp = data->timestamp;
5580
5581         trace_nfs4_renew_async(clp, task->tk_status);
5582         switch (task->tk_status) {
5583         case 0:
5584                 break;
5585         case -NFS4ERR_LEASE_MOVED:
5586                 nfs4_schedule_lease_moved_recovery(clp);
5587                 break;
5588         default:
5589                 /* Unless we're shutting down, schedule state recovery! */
5590                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5591                         return;
5592                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5593                         nfs4_schedule_lease_recovery(clp);
5594                         return;
5595                 }
5596                 nfs4_schedule_path_down_recovery(clp);
5597         }
5598         do_renew_lease(clp, timestamp);
5599 }
5600
5601 static const struct rpc_call_ops nfs4_renew_ops = {
5602         .rpc_call_done = nfs4_renew_done,
5603         .rpc_release = nfs4_renew_release,
5604 };
5605
5606 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5607 {
5608         struct rpc_message msg = {
5609                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5610                 .rpc_argp       = clp,
5611                 .rpc_cred       = cred,
5612         };
5613         struct nfs4_renewdata *data;
5614
5615         if (renew_flags == 0)
5616                 return 0;
5617         if (!refcount_inc_not_zero(&clp->cl_count))
5618                 return -EIO;
5619         data = kmalloc(sizeof(*data), GFP_NOFS);
5620         if (data == NULL) {
5621                 nfs_put_client(clp);
5622                 return -ENOMEM;
5623         }
5624         data->client = clp;
5625         data->timestamp = jiffies;
5626         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5627                         &nfs4_renew_ops, data);
5628 }
5629
5630 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5631 {
5632         struct rpc_message msg = {
5633                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5634                 .rpc_argp       = clp,
5635                 .rpc_cred       = cred,
5636         };
5637         unsigned long now = jiffies;
5638         int status;
5639
5640         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5641         if (status < 0)
5642                 return status;
5643         do_renew_lease(clp, now);
5644         return 0;
5645 }
5646
5647 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5648 {
5649         return server->caps & NFS_CAP_ACLS;
5650 }
5651
5652 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5653  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5654  * the stack.
5655  */
5656 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5657
5658 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5659                 struct page **pages)
5660 {
5661         struct page *newpage, **spages;
5662         int rc = 0;
5663         size_t len;
5664         spages = pages;
5665
5666         do {
5667                 len = min_t(size_t, PAGE_SIZE, buflen);
5668                 newpage = alloc_page(GFP_KERNEL);
5669
5670                 if (newpage == NULL)
5671                         goto unwind;
5672                 memcpy(page_address(newpage), buf, len);
5673                 buf += len;
5674                 buflen -= len;
5675                 *pages++ = newpage;
5676                 rc++;
5677         } while (buflen != 0);
5678
5679         return rc;
5680
5681 unwind:
5682         for(; rc > 0; rc--)
5683                 __free_page(spages[rc-1]);
5684         return -ENOMEM;
5685 }
5686
5687 struct nfs4_cached_acl {
5688         int cached;
5689         size_t len;
5690         char data[];
5691 };
5692
5693 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5694 {
5695         struct nfs_inode *nfsi = NFS_I(inode);
5696
5697         spin_lock(&inode->i_lock);
5698         kfree(nfsi->nfs4_acl);
5699         nfsi->nfs4_acl = acl;
5700         spin_unlock(&inode->i_lock);
5701 }
5702
5703 static void nfs4_zap_acl_attr(struct inode *inode)
5704 {
5705         nfs4_set_cached_acl(inode, NULL);
5706 }
5707
5708 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5709 {
5710         struct nfs_inode *nfsi = NFS_I(inode);
5711         struct nfs4_cached_acl *acl;
5712         int ret = -ENOENT;
5713
5714         spin_lock(&inode->i_lock);
5715         acl = nfsi->nfs4_acl;
5716         if (acl == NULL)
5717                 goto out;
5718         if (buf == NULL) /* user is just asking for length */
5719                 goto out_len;
5720         if (acl->cached == 0)
5721                 goto out;
5722         ret = -ERANGE; /* see getxattr(2) man page */
5723         if (acl->len > buflen)
5724                 goto out;
5725         memcpy(buf, acl->data, acl->len);
5726 out_len:
5727         ret = acl->len;
5728 out:
5729         spin_unlock(&inode->i_lock);
5730         return ret;
5731 }
5732
5733 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5734 {
5735         struct nfs4_cached_acl *acl;
5736         size_t buflen = sizeof(*acl) + acl_len;
5737
5738         if (buflen <= PAGE_SIZE) {
5739                 acl = kmalloc(buflen, GFP_KERNEL);
5740                 if (acl == NULL)
5741                         goto out;
5742                 acl->cached = 1;
5743                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5744         } else {
5745                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5746                 if (acl == NULL)
5747                         goto out;
5748                 acl->cached = 0;
5749         }
5750         acl->len = acl_len;
5751 out:
5752         nfs4_set_cached_acl(inode, acl);
5753 }
5754
5755 /*
5756  * The getxattr API returns the required buffer length when called with a
5757  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5758  * the required buf.  On a NULL buf, we send a page of data to the server
5759  * guessing that the ACL request can be serviced by a page. If so, we cache
5760  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5761  * the cache. If not so, we throw away the page, and cache the required
5762  * length. The next getxattr call will then produce another round trip to
5763  * the server, this time with the input buf of the required size.
5764  */
5765 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5766 {
5767         struct page **pages;
5768         struct nfs_getaclargs args = {
5769                 .fh = NFS_FH(inode),
5770                 .acl_len = buflen,
5771         };
5772         struct nfs_getaclres res = {
5773                 .acl_len = buflen,
5774         };
5775         struct rpc_message msg = {
5776                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5777                 .rpc_argp = &args,
5778                 .rpc_resp = &res,
5779         };
5780         unsigned int npages;
5781         int ret = -ENOMEM, i;
5782         struct nfs_server *server = NFS_SERVER(inode);
5783
5784         if (buflen == 0)
5785                 buflen = server->rsize;
5786
5787         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5788         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5789         if (!pages)
5790                 return -ENOMEM;
5791
5792         args.acl_pages = pages;
5793
5794         for (i = 0; i < npages; i++) {
5795                 pages[i] = alloc_page(GFP_KERNEL);
5796                 if (!pages[i])
5797                         goto out_free;
5798         }
5799
5800         /* for decoding across pages */
5801         res.acl_scratch = alloc_page(GFP_KERNEL);
5802         if (!res.acl_scratch)
5803                 goto out_free;
5804
5805         args.acl_len = npages * PAGE_SIZE;
5806
5807         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5808                 __func__, buf, buflen, npages, args.acl_len);
5809         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5810                              &msg, &args.seq_args, &res.seq_res, 0);
5811         if (ret)
5812                 goto out_free;
5813
5814         /* Handle the case where the passed-in buffer is too short */
5815         if (res.acl_flags & NFS4_ACL_TRUNC) {
5816                 /* Did the user only issue a request for the acl length? */
5817                 if (buf == NULL)
5818                         goto out_ok;
5819                 ret = -ERANGE;
5820                 goto out_free;
5821         }
5822         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5823         if (buf) {
5824                 if (res.acl_len > buflen) {
5825                         ret = -ERANGE;
5826                         goto out_free;
5827                 }
5828                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5829         }
5830 out_ok:
5831         ret = res.acl_len;
5832 out_free:
5833         for (i = 0; i < npages; i++)
5834                 if (pages[i])
5835                         __free_page(pages[i]);
5836         if (res.acl_scratch)
5837                 __free_page(res.acl_scratch);
5838         kfree(pages);
5839         return ret;
5840 }
5841
5842 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5843 {
5844         struct nfs4_exception exception = {
5845                 .interruptible = true,
5846         };
5847         ssize_t ret;
5848         do {
5849                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5850                 trace_nfs4_get_acl(inode, ret);
5851                 if (ret >= 0)
5852                         break;
5853                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5854         } while (exception.retry);
5855         return ret;
5856 }
5857
5858 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5859 {
5860         struct nfs_server *server = NFS_SERVER(inode);
5861         int ret;
5862
5863         if (!nfs4_server_supports_acls(server))
5864                 return -EOPNOTSUPP;
5865         ret = nfs_revalidate_inode(server, inode);
5866         if (ret < 0)
5867                 return ret;
5868         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5869                 nfs_zap_acl_cache(inode);
5870         ret = nfs4_read_cached_acl(inode, buf, buflen);
5871         if (ret != -ENOENT)
5872                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5873                  * but no cached acl data, just the acl length */
5874                 return ret;
5875         return nfs4_get_acl_uncached(inode, buf, buflen);
5876 }
5877
5878 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5879 {
5880         struct nfs_server *server = NFS_SERVER(inode);
5881         struct page *pages[NFS4ACL_MAXPAGES];
5882         struct nfs_setaclargs arg = {
5883                 .fh             = NFS_FH(inode),
5884                 .acl_pages      = pages,
5885                 .acl_len        = buflen,
5886         };
5887         struct nfs_setaclres res;
5888         struct rpc_message msg = {
5889                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
5890                 .rpc_argp       = &arg,
5891                 .rpc_resp       = &res,
5892         };
5893         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
5894         int ret, i;
5895
5896         if (!nfs4_server_supports_acls(server))
5897                 return -EOPNOTSUPP;
5898         if (npages > ARRAY_SIZE(pages))
5899                 return -ERANGE;
5900         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
5901         if (i < 0)
5902                 return i;
5903         nfs4_inode_make_writeable(inode);
5904         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5905
5906         /*
5907          * Free each page after tx, so the only ref left is
5908          * held by the network stack
5909          */
5910         for (; i > 0; i--)
5911                 put_page(pages[i-1]);
5912
5913         /*
5914          * Acl update can result in inode attribute update.
5915          * so mark the attribute cache invalid.
5916          */
5917         spin_lock(&inode->i_lock);
5918         NFS_I(inode)->cache_validity |= NFS_INO_INVALID_CHANGE
5919                 | NFS_INO_INVALID_CTIME
5920                 | NFS_INO_REVAL_FORCED;
5921         spin_unlock(&inode->i_lock);
5922         nfs_access_zap_cache(inode);
5923         nfs_zap_acl_cache(inode);
5924         return ret;
5925 }
5926
5927 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5928 {
5929         struct nfs4_exception exception = { };
5930         int err;
5931         do {
5932                 err = __nfs4_proc_set_acl(inode, buf, buflen);
5933                 trace_nfs4_set_acl(inode, err);
5934                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5935                                 &exception);
5936         } while (exception.retry);
5937         return err;
5938 }
5939
5940 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
5941 static int _nfs4_get_security_label(struct inode *inode, void *buf,
5942                                         size_t buflen)
5943 {
5944         struct nfs_server *server = NFS_SERVER(inode);
5945         struct nfs_fattr fattr;
5946         struct nfs4_label label = {0, 0, buflen, buf};
5947
5948         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
5949         struct nfs4_getattr_arg arg = {
5950                 .fh             = NFS_FH(inode),
5951                 .bitmask        = bitmask,
5952         };
5953         struct nfs4_getattr_res res = {
5954                 .fattr          = &fattr,
5955                 .label          = &label,
5956                 .server         = server,
5957         };
5958         struct rpc_message msg = {
5959                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
5960                 .rpc_argp       = &arg,
5961                 .rpc_resp       = &res,
5962         };
5963         int ret;
5964
5965         nfs_fattr_init(&fattr);
5966
5967         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
5968         if (ret)
5969                 return ret;
5970         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
5971                 return -ENOENT;
5972         return 0;
5973 }
5974
5975 static int nfs4_get_security_label(struct inode *inode, void *buf,
5976                                         size_t buflen)
5977 {
5978         struct nfs4_exception exception = {
5979                 .interruptible = true,
5980         };
5981         int err;
5982
5983         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
5984                 return -EOPNOTSUPP;
5985
5986         do {
5987                 err = _nfs4_get_security_label(inode, buf, buflen);
5988                 trace_nfs4_get_security_label(inode, err);
5989                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
5990                                 &exception);
5991         } while (exception.retry);
5992         return err;
5993 }
5994
5995 static int _nfs4_do_set_security_label(struct inode *inode,
5996                 struct nfs4_label *ilabel,
5997                 struct nfs_fattr *fattr,
5998                 struct nfs4_label *olabel)
5999 {
6000
6001         struct iattr sattr = {0};
6002         struct nfs_server *server = NFS_SERVER(inode);
6003         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6004         struct nfs_setattrargs arg = {
6005                 .fh             = NFS_FH(inode),
6006                 .iap            = &sattr,
6007                 .server         = server,
6008                 .bitmask        = bitmask,
6009                 .label          = ilabel,
6010         };
6011         struct nfs_setattrres res = {
6012                 .fattr          = fattr,
6013                 .label          = olabel,
6014                 .server         = server,
6015         };
6016         struct rpc_message msg = {
6017                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6018                 .rpc_argp       = &arg,
6019                 .rpc_resp       = &res,
6020         };
6021         int status;
6022
6023         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6024
6025         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6026         if (status)
6027                 dprintk("%s failed: %d\n", __func__, status);
6028
6029         return status;
6030 }
6031
6032 static int nfs4_do_set_security_label(struct inode *inode,
6033                 struct nfs4_label *ilabel,
6034                 struct nfs_fattr *fattr,
6035                 struct nfs4_label *olabel)
6036 {
6037         struct nfs4_exception exception = { };
6038         int err;
6039
6040         do {
6041                 err = _nfs4_do_set_security_label(inode, ilabel,
6042                                 fattr, olabel);
6043                 trace_nfs4_set_security_label(inode, err);
6044                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6045                                 &exception);
6046         } while (exception.retry);
6047         return err;
6048 }
6049
6050 static int
6051 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6052 {
6053         struct nfs4_label ilabel, *olabel = NULL;
6054         struct nfs_fattr fattr;
6055         int status;
6056
6057         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6058                 return -EOPNOTSUPP;
6059
6060         nfs_fattr_init(&fattr);
6061
6062         ilabel.pi = 0;
6063         ilabel.lfs = 0;
6064         ilabel.label = (char *)buf;
6065         ilabel.len = buflen;
6066
6067         olabel = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
6068         if (IS_ERR(olabel)) {
6069                 status = -PTR_ERR(olabel);
6070                 goto out;
6071         }
6072
6073         status = nfs4_do_set_security_label(inode, &ilabel, &fattr, olabel);
6074         if (status == 0)
6075                 nfs_setsecurity(inode, &fattr, olabel);
6076
6077         nfs4_label_free(olabel);
6078 out:
6079         return status;
6080 }
6081 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6082
6083
6084 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6085                                     nfs4_verifier *bootverf)
6086 {
6087         __be32 verf[2];
6088
6089         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6090                 /* An impossible timestamp guarantees this value
6091                  * will never match a generated boot time. */
6092                 verf[0] = cpu_to_be32(U32_MAX);
6093                 verf[1] = cpu_to_be32(U32_MAX);
6094         } else {
6095                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6096                 u64 ns = ktime_to_ns(nn->boot_time);
6097
6098                 verf[0] = cpu_to_be32(ns >> 32);
6099                 verf[1] = cpu_to_be32(ns);
6100         }
6101         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6102 }
6103
6104 static size_t
6105 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6106 {
6107         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6108         struct nfs_netns_client *nn_clp = nn->nfs_client;
6109         const char *id;
6110
6111         buf[0] = '\0';
6112
6113         if (nn_clp) {
6114                 rcu_read_lock();
6115                 id = rcu_dereference(nn_clp->identifier);
6116                 if (id)
6117                         strscpy(buf, id, buflen);
6118                 rcu_read_unlock();
6119         }
6120
6121         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6122                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6123
6124         return strlen(buf);
6125 }
6126
6127 static int
6128 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6129 {
6130         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6131         size_t buflen;
6132         size_t len;
6133         char *str;
6134
6135         if (clp->cl_owner_id != NULL)
6136                 return 0;
6137
6138         rcu_read_lock();
6139         len = 14 +
6140                 strlen(clp->cl_rpcclient->cl_nodename) +
6141                 1 +
6142                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6143                 1;
6144         rcu_read_unlock();
6145
6146         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6147         if (buflen)
6148                 len += buflen + 1;
6149
6150         if (len > NFS4_OPAQUE_LIMIT + 1)
6151                 return -EINVAL;
6152
6153         /*
6154          * Since this string is allocated at mount time, and held until the
6155          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6156          * about a memory-reclaim deadlock.
6157          */
6158         str = kmalloc(len, GFP_KERNEL);
6159         if (!str)
6160                 return -ENOMEM;
6161
6162         rcu_read_lock();
6163         if (buflen)
6164                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6165                           clp->cl_rpcclient->cl_nodename, buf,
6166                           rpc_peeraddr2str(clp->cl_rpcclient,
6167                                            RPC_DISPLAY_ADDR));
6168         else
6169                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6170                           clp->cl_rpcclient->cl_nodename,
6171                           rpc_peeraddr2str(clp->cl_rpcclient,
6172                                            RPC_DISPLAY_ADDR));
6173         rcu_read_unlock();
6174
6175         clp->cl_owner_id = str;
6176         return 0;
6177 }
6178
6179 static int
6180 nfs4_init_uniform_client_string(struct nfs_client *clp)
6181 {
6182         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6183         size_t buflen;
6184         size_t len;
6185         char *str;
6186
6187         if (clp->cl_owner_id != NULL)
6188                 return 0;
6189
6190         len = 10 + 10 + 1 + 10 + 1 +
6191                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6192
6193         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6194         if (buflen)
6195                 len += buflen + 1;
6196
6197         if (len > NFS4_OPAQUE_LIMIT + 1)
6198                 return -EINVAL;
6199
6200         /*
6201          * Since this string is allocated at mount time, and held until the
6202          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6203          * about a memory-reclaim deadlock.
6204          */
6205         str = kmalloc(len, GFP_KERNEL);
6206         if (!str)
6207                 return -ENOMEM;
6208
6209         if (buflen)
6210                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6211                           clp->rpc_ops->version, clp->cl_minorversion,
6212                           buf, clp->cl_rpcclient->cl_nodename);
6213         else
6214                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6215                           clp->rpc_ops->version, clp->cl_minorversion,
6216                           clp->cl_rpcclient->cl_nodename);
6217         clp->cl_owner_id = str;
6218         return 0;
6219 }
6220
6221 /*
6222  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6223  * services.  Advertise one based on the address family of the
6224  * clientaddr.
6225  */
6226 static unsigned int
6227 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6228 {
6229         if (strchr(clp->cl_ipaddr, ':') != NULL)
6230                 return scnprintf(buf, len, "tcp6");
6231         else
6232                 return scnprintf(buf, len, "tcp");
6233 }
6234
6235 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6236 {
6237         struct nfs4_setclientid *sc = calldata;
6238
6239         if (task->tk_status == 0)
6240                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6241 }
6242
6243 static const struct rpc_call_ops nfs4_setclientid_ops = {
6244         .rpc_call_done = nfs4_setclientid_done,
6245 };
6246
6247 /**
6248  * nfs4_proc_setclientid - Negotiate client ID
6249  * @clp: state data structure
6250  * @program: RPC program for NFSv4 callback service
6251  * @port: IP port number for NFS4 callback service
6252  * @cred: credential to use for this call
6253  * @res: where to place the result
6254  *
6255  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6256  */
6257 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6258                 unsigned short port, const struct cred *cred,
6259                 struct nfs4_setclientid_res *res)
6260 {
6261         nfs4_verifier sc_verifier;
6262         struct nfs4_setclientid setclientid = {
6263                 .sc_verifier = &sc_verifier,
6264                 .sc_prog = program,
6265                 .sc_clnt = clp,
6266         };
6267         struct rpc_message msg = {
6268                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6269                 .rpc_argp = &setclientid,
6270                 .rpc_resp = res,
6271                 .rpc_cred = cred,
6272         };
6273         struct rpc_task_setup task_setup_data = {
6274                 .rpc_client = clp->cl_rpcclient,
6275                 .rpc_message = &msg,
6276                 .callback_ops = &nfs4_setclientid_ops,
6277                 .callback_data = &setclientid,
6278                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6279         };
6280         unsigned long now = jiffies;
6281         int status;
6282
6283         /* nfs_client_id4 */
6284         nfs4_init_boot_verifier(clp, &sc_verifier);
6285
6286         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6287                 status = nfs4_init_uniform_client_string(clp);
6288         else
6289                 status = nfs4_init_nonuniform_client_string(clp);
6290
6291         if (status)
6292                 goto out;
6293
6294         /* cb_client4 */
6295         setclientid.sc_netid_len =
6296                                 nfs4_init_callback_netid(clp,
6297                                                 setclientid.sc_netid,
6298                                                 sizeof(setclientid.sc_netid));
6299         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6300                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6301                                 clp->cl_ipaddr, port >> 8, port & 255);
6302
6303         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6304                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6305                 clp->cl_owner_id);
6306
6307         status = nfs4_call_sync_custom(&task_setup_data);
6308         if (setclientid.sc_cred) {
6309                 kfree(clp->cl_acceptor);
6310                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6311                 put_rpccred(setclientid.sc_cred);
6312         }
6313
6314         if (status == 0)
6315                 do_renew_lease(clp, now);
6316 out:
6317         trace_nfs4_setclientid(clp, status);
6318         dprintk("NFS reply setclientid: %d\n", status);
6319         return status;
6320 }
6321
6322 /**
6323  * nfs4_proc_setclientid_confirm - Confirm client ID
6324  * @clp: state data structure
6325  * @arg: result of a previous SETCLIENTID
6326  * @cred: credential to use for this call
6327  *
6328  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6329  */
6330 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6331                 struct nfs4_setclientid_res *arg,
6332                 const struct cred *cred)
6333 {
6334         struct rpc_message msg = {
6335                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6336                 .rpc_argp = arg,
6337                 .rpc_cred = cred,
6338         };
6339         int status;
6340
6341         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6342                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6343                 clp->cl_clientid);
6344         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6345                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6346         trace_nfs4_setclientid_confirm(clp, status);
6347         dprintk("NFS reply setclientid_confirm: %d\n", status);
6348         return status;
6349 }
6350
6351 struct nfs4_delegreturndata {
6352         struct nfs4_delegreturnargs args;
6353         struct nfs4_delegreturnres res;
6354         struct nfs_fh fh;
6355         nfs4_stateid stateid;
6356         unsigned long timestamp;
6357         struct {
6358                 struct nfs4_layoutreturn_args arg;
6359                 struct nfs4_layoutreturn_res res;
6360                 struct nfs4_xdr_opaque_data ld_private;
6361                 u32 roc_barrier;
6362                 bool roc;
6363         } lr;
6364         struct nfs_fattr fattr;
6365         int rpc_status;
6366         struct inode *inode;
6367 };
6368
6369 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6370 {
6371         struct nfs4_delegreturndata *data = calldata;
6372         struct nfs4_exception exception = {
6373                 .inode = data->inode,
6374                 .stateid = &data->stateid,
6375         };
6376
6377         if (!nfs4_sequence_done(task, &data->res.seq_res))
6378                 return;
6379
6380         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6381
6382         /* Handle Layoutreturn errors */
6383         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6384                           &data->res.lr_ret) == -EAGAIN)
6385                 goto out_restart;
6386
6387         switch (task->tk_status) {
6388         case 0:
6389                 renew_lease(data->res.server, data->timestamp);
6390                 break;
6391         case -NFS4ERR_ADMIN_REVOKED:
6392         case -NFS4ERR_DELEG_REVOKED:
6393         case -NFS4ERR_EXPIRED:
6394                 nfs4_free_revoked_stateid(data->res.server,
6395                                 data->args.stateid,
6396                                 task->tk_msg.rpc_cred);
6397                 fallthrough;
6398         case -NFS4ERR_BAD_STATEID:
6399         case -NFS4ERR_STALE_STATEID:
6400         case -ETIMEDOUT:
6401                 task->tk_status = 0;
6402                 break;
6403         case -NFS4ERR_OLD_STATEID:
6404                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6405                         nfs4_stateid_seqid_inc(&data->stateid);
6406                 if (data->args.bitmask) {
6407                         data->args.bitmask = NULL;
6408                         data->res.fattr = NULL;
6409                 }
6410                 goto out_restart;
6411         case -NFS4ERR_ACCESS:
6412                 if (data->args.bitmask) {
6413                         data->args.bitmask = NULL;
6414                         data->res.fattr = NULL;
6415                         goto out_restart;
6416                 }
6417                 fallthrough;
6418         default:
6419                 task->tk_status = nfs4_async_handle_exception(task,
6420                                 data->res.server, task->tk_status,
6421                                 &exception);
6422                 if (exception.retry)
6423                         goto out_restart;
6424         }
6425         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6426         data->rpc_status = task->tk_status;
6427         return;
6428 out_restart:
6429         task->tk_status = 0;
6430         rpc_restart_call_prepare(task);
6431 }
6432
6433 static void nfs4_delegreturn_release(void *calldata)
6434 {
6435         struct nfs4_delegreturndata *data = calldata;
6436         struct inode *inode = data->inode;
6437
6438         if (data->lr.roc)
6439                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6440                                  data->res.lr_ret);
6441         if (inode) {
6442                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6443                 nfs_iput_and_deactive(inode);
6444         }
6445         kfree(calldata);
6446 }
6447
6448 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6449 {
6450         struct nfs4_delegreturndata *d_data;
6451         struct pnfs_layout_hdr *lo;
6452
6453         d_data = (struct nfs4_delegreturndata *)data;
6454
6455         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6456                 nfs4_sequence_done(task, &d_data->res.seq_res);
6457                 return;
6458         }
6459
6460         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6461         if (lo && !pnfs_layout_is_valid(lo)) {
6462                 d_data->args.lr_args = NULL;
6463                 d_data->res.lr_res = NULL;
6464         }
6465
6466         nfs4_setup_sequence(d_data->res.server->nfs_client,
6467                         &d_data->args.seq_args,
6468                         &d_data->res.seq_res,
6469                         task);
6470 }
6471
6472 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6473         .rpc_call_prepare = nfs4_delegreturn_prepare,
6474         .rpc_call_done = nfs4_delegreturn_done,
6475         .rpc_release = nfs4_delegreturn_release,
6476 };
6477
6478 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6479 {
6480         struct nfs4_delegreturndata *data;
6481         struct nfs_server *server = NFS_SERVER(inode);
6482         struct rpc_task *task;
6483         struct rpc_message msg = {
6484                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6485                 .rpc_cred = cred,
6486         };
6487         struct rpc_task_setup task_setup_data = {
6488                 .rpc_client = server->client,
6489                 .rpc_message = &msg,
6490                 .callback_ops = &nfs4_delegreturn_ops,
6491                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6492         };
6493         int status = 0;
6494
6495         data = kzalloc(sizeof(*data), GFP_NOFS);
6496         if (data == NULL)
6497                 return -ENOMEM;
6498         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
6499
6500         nfs4_state_protect(server->nfs_client,
6501                         NFS_SP4_MACH_CRED_CLEANUP,
6502                         &task_setup_data.rpc_client, &msg);
6503
6504         data->args.fhandle = &data->fh;
6505         data->args.stateid = &data->stateid;
6506         data->args.bitmask = server->cache_consistency_bitmask;
6507         nfs4_bitmask_adjust(data->args.bitmask, inode, server, NULL);
6508         nfs_copy_fh(&data->fh, NFS_FH(inode));
6509         nfs4_stateid_copy(&data->stateid, stateid);
6510         data->res.fattr = &data->fattr;
6511         data->res.server = server;
6512         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6513         data->lr.arg.ld_private = &data->lr.ld_private;
6514         nfs_fattr_init(data->res.fattr);
6515         data->timestamp = jiffies;
6516         data->rpc_status = 0;
6517         data->inode = nfs_igrab_and_active(inode);
6518         if (data->inode || issync) {
6519                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6520                                         cred);
6521                 if (data->lr.roc) {
6522                         data->args.lr_args = &data->lr.arg;
6523                         data->res.lr_res = &data->lr.res;
6524                 }
6525         }
6526
6527         task_setup_data.callback_data = data;
6528         msg.rpc_argp = &data->args;
6529         msg.rpc_resp = &data->res;
6530         task = rpc_run_task(&task_setup_data);
6531         if (IS_ERR(task))
6532                 return PTR_ERR(task);
6533         if (!issync)
6534                 goto out;
6535         status = rpc_wait_for_completion_task(task);
6536         if (status != 0)
6537                 goto out;
6538         status = data->rpc_status;
6539 out:
6540         rpc_put_task(task);
6541         return status;
6542 }
6543
6544 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6545 {
6546         struct nfs_server *server = NFS_SERVER(inode);
6547         struct nfs4_exception exception = { };
6548         int err;
6549         do {
6550                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6551                 trace_nfs4_delegreturn(inode, stateid, err);
6552                 switch (err) {
6553                         case -NFS4ERR_STALE_STATEID:
6554                         case -NFS4ERR_EXPIRED:
6555                         case 0:
6556                                 return 0;
6557                 }
6558                 err = nfs4_handle_exception(server, err, &exception);
6559         } while (exception.retry);
6560         return err;
6561 }
6562
6563 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6564 {
6565         struct inode *inode = state->inode;
6566         struct nfs_server *server = NFS_SERVER(inode);
6567         struct nfs_client *clp = server->nfs_client;
6568         struct nfs_lockt_args arg = {
6569                 .fh = NFS_FH(inode),
6570                 .fl = request,
6571         };
6572         struct nfs_lockt_res res = {
6573                 .denied = request,
6574         };
6575         struct rpc_message msg = {
6576                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6577                 .rpc_argp       = &arg,
6578                 .rpc_resp       = &res,
6579                 .rpc_cred       = state->owner->so_cred,
6580         };
6581         struct nfs4_lock_state *lsp;
6582         int status;
6583
6584         arg.lock_owner.clientid = clp->cl_clientid;
6585         status = nfs4_set_lock_state(state, request);
6586         if (status != 0)
6587                 goto out;
6588         lsp = request->fl_u.nfs4_fl.owner;
6589         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6590         arg.lock_owner.s_dev = server->s_dev;
6591         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6592         switch (status) {
6593                 case 0:
6594                         request->fl_type = F_UNLCK;
6595                         break;
6596                 case -NFS4ERR_DENIED:
6597                         status = 0;
6598         }
6599         request->fl_ops->fl_release_private(request);
6600         request->fl_ops = NULL;
6601 out:
6602         return status;
6603 }
6604
6605 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6606 {
6607         struct nfs4_exception exception = {
6608                 .interruptible = true,
6609         };
6610         int err;
6611
6612         do {
6613                 err = _nfs4_proc_getlk(state, cmd, request);
6614                 trace_nfs4_get_lock(request, state, cmd, err);
6615                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6616                                 &exception);
6617         } while (exception.retry);
6618         return err;
6619 }
6620
6621 /*
6622  * Update the seqid of a lock stateid after receiving
6623  * NFS4ERR_OLD_STATEID
6624  */
6625 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6626                 struct nfs4_lock_state *lsp)
6627 {
6628         struct nfs4_state *state = lsp->ls_state;
6629         bool ret = false;
6630
6631         spin_lock(&state->state_lock);
6632         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6633                 goto out;
6634         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6635                 nfs4_stateid_seqid_inc(dst);
6636         else
6637                 dst->seqid = lsp->ls_stateid.seqid;
6638         ret = true;
6639 out:
6640         spin_unlock(&state->state_lock);
6641         return ret;
6642 }
6643
6644 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6645                 struct nfs4_lock_state *lsp)
6646 {
6647         struct nfs4_state *state = lsp->ls_state;
6648         bool ret;
6649
6650         spin_lock(&state->state_lock);
6651         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6652         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6653         spin_unlock(&state->state_lock);
6654         return ret;
6655 }
6656
6657 struct nfs4_unlockdata {
6658         struct nfs_locku_args arg;
6659         struct nfs_locku_res res;
6660         struct nfs4_lock_state *lsp;
6661         struct nfs_open_context *ctx;
6662         struct nfs_lock_context *l_ctx;
6663         struct file_lock fl;
6664         struct nfs_server *server;
6665         unsigned long timestamp;
6666 };
6667
6668 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6669                 struct nfs_open_context *ctx,
6670                 struct nfs4_lock_state *lsp,
6671                 struct nfs_seqid *seqid)
6672 {
6673         struct nfs4_unlockdata *p;
6674         struct nfs4_state *state = lsp->ls_state;
6675         struct inode *inode = state->inode;
6676
6677         p = kzalloc(sizeof(*p), GFP_NOFS);
6678         if (p == NULL)
6679                 return NULL;
6680         p->arg.fh = NFS_FH(inode);
6681         p->arg.fl = &p->fl;
6682         p->arg.seqid = seqid;
6683         p->res.seqid = seqid;
6684         p->lsp = lsp;
6685         /* Ensure we don't close file until we're done freeing locks! */
6686         p->ctx = get_nfs_open_context(ctx);
6687         p->l_ctx = nfs_get_lock_context(ctx);
6688         locks_init_lock(&p->fl);
6689         locks_copy_lock(&p->fl, fl);
6690         p->server = NFS_SERVER(inode);
6691         spin_lock(&state->state_lock);
6692         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6693         spin_unlock(&state->state_lock);
6694         return p;
6695 }
6696
6697 static void nfs4_locku_release_calldata(void *data)
6698 {
6699         struct nfs4_unlockdata *calldata = data;
6700         nfs_free_seqid(calldata->arg.seqid);
6701         nfs4_put_lock_state(calldata->lsp);
6702         nfs_put_lock_context(calldata->l_ctx);
6703         put_nfs_open_context(calldata->ctx);
6704         kfree(calldata);
6705 }
6706
6707 static void nfs4_locku_done(struct rpc_task *task, void *data)
6708 {
6709         struct nfs4_unlockdata *calldata = data;
6710         struct nfs4_exception exception = {
6711                 .inode = calldata->lsp->ls_state->inode,
6712                 .stateid = &calldata->arg.stateid,
6713         };
6714
6715         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6716                 return;
6717         switch (task->tk_status) {
6718                 case 0:
6719                         renew_lease(calldata->server, calldata->timestamp);
6720                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6721                         if (nfs4_update_lock_stateid(calldata->lsp,
6722                                         &calldata->res.stateid))
6723                                 break;
6724                         fallthrough;
6725                 case -NFS4ERR_ADMIN_REVOKED:
6726                 case -NFS4ERR_EXPIRED:
6727                         nfs4_free_revoked_stateid(calldata->server,
6728                                         &calldata->arg.stateid,
6729                                         task->tk_msg.rpc_cred);
6730                         fallthrough;
6731                 case -NFS4ERR_BAD_STATEID:
6732                 case -NFS4ERR_STALE_STATEID:
6733                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6734                                                 calldata->lsp))
6735                                 rpc_restart_call_prepare(task);
6736                         break;
6737                 case -NFS4ERR_OLD_STATEID:
6738                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6739                                                 calldata->lsp))
6740                                 rpc_restart_call_prepare(task);
6741                         break;
6742                 default:
6743                         task->tk_status = nfs4_async_handle_exception(task,
6744                                         calldata->server, task->tk_status,
6745                                         &exception);
6746                         if (exception.retry)
6747                                 rpc_restart_call_prepare(task);
6748         }
6749         nfs_release_seqid(calldata->arg.seqid);
6750 }
6751
6752 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6753 {
6754         struct nfs4_unlockdata *calldata = data;
6755
6756         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6757                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6758                 return;
6759
6760         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6761                 goto out_wait;
6762         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6763                 /* Note: exit _without_ running nfs4_locku_done */
6764                 goto out_no_action;
6765         }
6766         calldata->timestamp = jiffies;
6767         if (nfs4_setup_sequence(calldata->server->nfs_client,
6768                                 &calldata->arg.seq_args,
6769                                 &calldata->res.seq_res,
6770                                 task) != 0)
6771                 nfs_release_seqid(calldata->arg.seqid);
6772         return;
6773 out_no_action:
6774         task->tk_action = NULL;
6775 out_wait:
6776         nfs4_sequence_done(task, &calldata->res.seq_res);
6777 }
6778
6779 static const struct rpc_call_ops nfs4_locku_ops = {
6780         .rpc_call_prepare = nfs4_locku_prepare,
6781         .rpc_call_done = nfs4_locku_done,
6782         .rpc_release = nfs4_locku_release_calldata,
6783 };
6784
6785 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6786                 struct nfs_open_context *ctx,
6787                 struct nfs4_lock_state *lsp,
6788                 struct nfs_seqid *seqid)
6789 {
6790         struct nfs4_unlockdata *data;
6791         struct rpc_message msg = {
6792                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6793                 .rpc_cred = ctx->cred,
6794         };
6795         struct rpc_task_setup task_setup_data = {
6796                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6797                 .rpc_message = &msg,
6798                 .callback_ops = &nfs4_locku_ops,
6799                 .workqueue = nfsiod_workqueue,
6800                 .flags = RPC_TASK_ASYNC,
6801         };
6802
6803         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6804                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6805
6806         /* Ensure this is an unlock - when canceling a lock, the
6807          * canceled lock is passed in, and it won't be an unlock.
6808          */
6809         fl->fl_type = F_UNLCK;
6810         if (fl->fl_flags & FL_CLOSE)
6811                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6812
6813         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6814         if (data == NULL) {
6815                 nfs_free_seqid(seqid);
6816                 return ERR_PTR(-ENOMEM);
6817         }
6818
6819         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6820         msg.rpc_argp = &data->arg;
6821         msg.rpc_resp = &data->res;
6822         task_setup_data.callback_data = data;
6823         return rpc_run_task(&task_setup_data);
6824 }
6825
6826 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6827 {
6828         struct inode *inode = state->inode;
6829         struct nfs4_state_owner *sp = state->owner;
6830         struct nfs_inode *nfsi = NFS_I(inode);
6831         struct nfs_seqid *seqid;
6832         struct nfs4_lock_state *lsp;
6833         struct rpc_task *task;
6834         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6835         int status = 0;
6836         unsigned char fl_flags = request->fl_flags;
6837
6838         status = nfs4_set_lock_state(state, request);
6839         /* Unlock _before_ we do the RPC call */
6840         request->fl_flags |= FL_EXISTS;
6841         /* Exclude nfs_delegation_claim_locks() */
6842         mutex_lock(&sp->so_delegreturn_mutex);
6843         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6844         down_read(&nfsi->rwsem);
6845         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6846                 up_read(&nfsi->rwsem);
6847                 mutex_unlock(&sp->so_delegreturn_mutex);
6848                 goto out;
6849         }
6850         up_read(&nfsi->rwsem);
6851         mutex_unlock(&sp->so_delegreturn_mutex);
6852         if (status != 0)
6853                 goto out;
6854         /* Is this a delegated lock? */
6855         lsp = request->fl_u.nfs4_fl.owner;
6856         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6857                 goto out;
6858         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6859         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6860         status = -ENOMEM;
6861         if (IS_ERR(seqid))
6862                 goto out;
6863         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6864         status = PTR_ERR(task);
6865         if (IS_ERR(task))
6866                 goto out;
6867         status = rpc_wait_for_completion_task(task);
6868         rpc_put_task(task);
6869 out:
6870         request->fl_flags = fl_flags;
6871         trace_nfs4_unlock(request, state, F_SETLK, status);
6872         return status;
6873 }
6874
6875 struct nfs4_lockdata {
6876         struct nfs_lock_args arg;
6877         struct nfs_lock_res res;
6878         struct nfs4_lock_state *lsp;
6879         struct nfs_open_context *ctx;
6880         struct file_lock fl;
6881         unsigned long timestamp;
6882         int rpc_status;
6883         int cancelled;
6884         struct nfs_server *server;
6885 };
6886
6887 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
6888                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
6889                 gfp_t gfp_mask)
6890 {
6891         struct nfs4_lockdata *p;
6892         struct inode *inode = lsp->ls_state->inode;
6893         struct nfs_server *server = NFS_SERVER(inode);
6894         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6895
6896         p = kzalloc(sizeof(*p), gfp_mask);
6897         if (p == NULL)
6898                 return NULL;
6899
6900         p->arg.fh = NFS_FH(inode);
6901         p->arg.fl = &p->fl;
6902         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
6903         if (IS_ERR(p->arg.open_seqid))
6904                 goto out_free;
6905         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
6906         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
6907         if (IS_ERR(p->arg.lock_seqid))
6908                 goto out_free_seqid;
6909         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
6910         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
6911         p->arg.lock_owner.s_dev = server->s_dev;
6912         p->res.lock_seqid = p->arg.lock_seqid;
6913         p->lsp = lsp;
6914         p->server = server;
6915         p->ctx = get_nfs_open_context(ctx);
6916         locks_init_lock(&p->fl);
6917         locks_copy_lock(&p->fl, fl);
6918         return p;
6919 out_free_seqid:
6920         nfs_free_seqid(p->arg.open_seqid);
6921 out_free:
6922         kfree(p);
6923         return NULL;
6924 }
6925
6926 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
6927 {
6928         struct nfs4_lockdata *data = calldata;
6929         struct nfs4_state *state = data->lsp->ls_state;
6930
6931         dprintk("%s: begin!\n", __func__);
6932         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
6933                 goto out_wait;
6934         /* Do we need to do an open_to_lock_owner? */
6935         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
6936                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
6937                         goto out_release_lock_seqid;
6938                 }
6939                 nfs4_stateid_copy(&data->arg.open_stateid,
6940                                 &state->open_stateid);
6941                 data->arg.new_lock_owner = 1;
6942                 data->res.open_seqid = data->arg.open_seqid;
6943         } else {
6944                 data->arg.new_lock_owner = 0;
6945                 nfs4_stateid_copy(&data->arg.lock_stateid,
6946                                 &data->lsp->ls_stateid);
6947         }
6948         if (!nfs4_valid_open_stateid(state)) {
6949                 data->rpc_status = -EBADF;
6950                 task->tk_action = NULL;
6951                 goto out_release_open_seqid;
6952         }
6953         data->timestamp = jiffies;
6954         if (nfs4_setup_sequence(data->server->nfs_client,
6955                                 &data->arg.seq_args,
6956                                 &data->res.seq_res,
6957                                 task) == 0)
6958                 return;
6959 out_release_open_seqid:
6960         nfs_release_seqid(data->arg.open_seqid);
6961 out_release_lock_seqid:
6962         nfs_release_seqid(data->arg.lock_seqid);
6963 out_wait:
6964         nfs4_sequence_done(task, &data->res.seq_res);
6965         dprintk("%s: done!, ret = %d\n", __func__, data->rpc_status);
6966 }
6967
6968 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
6969 {
6970         struct nfs4_lockdata *data = calldata;
6971         struct nfs4_lock_state *lsp = data->lsp;
6972
6973         dprintk("%s: begin!\n", __func__);
6974
6975         if (!nfs4_sequence_done(task, &data->res.seq_res))
6976                 return;
6977
6978         data->rpc_status = task->tk_status;
6979         switch (task->tk_status) {
6980         case 0:
6981                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
6982                                 data->timestamp);
6983                 if (data->arg.new_lock && !data->cancelled) {
6984                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
6985                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
6986                                 goto out_restart;
6987                 }
6988                 if (data->arg.new_lock_owner != 0) {
6989                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
6990                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
6991                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
6992                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
6993                         goto out_restart;
6994                 break;
6995         case -NFS4ERR_BAD_STATEID:
6996         case -NFS4ERR_OLD_STATEID:
6997         case -NFS4ERR_STALE_STATEID:
6998         case -NFS4ERR_EXPIRED:
6999                 if (data->arg.new_lock_owner != 0) {
7000                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7001                                                 &lsp->ls_state->open_stateid))
7002                                 goto out_restart;
7003                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7004                                                 &lsp->ls_stateid))
7005                                 goto out_restart;
7006         }
7007 out_done:
7008         dprintk("%s: done, ret = %d!\n", __func__, data->rpc_status);
7009         return;
7010 out_restart:
7011         if (!data->cancelled)
7012                 rpc_restart_call_prepare(task);
7013         goto out_done;
7014 }
7015
7016 static void nfs4_lock_release(void *calldata)
7017 {
7018         struct nfs4_lockdata *data = calldata;
7019
7020         dprintk("%s: begin!\n", __func__);
7021         nfs_free_seqid(data->arg.open_seqid);
7022         if (data->cancelled && data->rpc_status == 0) {
7023                 struct rpc_task *task;
7024                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7025                                 data->arg.lock_seqid);
7026                 if (!IS_ERR(task))
7027                         rpc_put_task_async(task);
7028                 dprintk("%s: cancelling lock!\n", __func__);
7029         } else
7030                 nfs_free_seqid(data->arg.lock_seqid);
7031         nfs4_put_lock_state(data->lsp);
7032         put_nfs_open_context(data->ctx);
7033         kfree(data);
7034         dprintk("%s: done!\n", __func__);
7035 }
7036
7037 static const struct rpc_call_ops nfs4_lock_ops = {
7038         .rpc_call_prepare = nfs4_lock_prepare,
7039         .rpc_call_done = nfs4_lock_done,
7040         .rpc_release = nfs4_lock_release,
7041 };
7042
7043 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7044 {
7045         switch (error) {
7046         case -NFS4ERR_ADMIN_REVOKED:
7047         case -NFS4ERR_EXPIRED:
7048         case -NFS4ERR_BAD_STATEID:
7049                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7050                 if (new_lock_owner != 0 ||
7051                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7052                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7053                 break;
7054         case -NFS4ERR_STALE_STATEID:
7055                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7056                 nfs4_schedule_lease_recovery(server->nfs_client);
7057         }
7058 }
7059
7060 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7061 {
7062         struct nfs4_lockdata *data;
7063         struct rpc_task *task;
7064         struct rpc_message msg = {
7065                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7066                 .rpc_cred = state->owner->so_cred,
7067         };
7068         struct rpc_task_setup task_setup_data = {
7069                 .rpc_client = NFS_CLIENT(state->inode),
7070                 .rpc_message = &msg,
7071                 .callback_ops = &nfs4_lock_ops,
7072                 .workqueue = nfsiod_workqueue,
7073                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7074         };
7075         int ret;
7076
7077         dprintk("%s: begin!\n", __func__);
7078         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7079                         fl->fl_u.nfs4_fl.owner,
7080                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7081         if (data == NULL)
7082                 return -ENOMEM;
7083         if (IS_SETLKW(cmd))
7084                 data->arg.block = 1;
7085         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7086                                 recovery_type > NFS_LOCK_NEW);
7087         msg.rpc_argp = &data->arg;
7088         msg.rpc_resp = &data->res;
7089         task_setup_data.callback_data = data;
7090         if (recovery_type > NFS_LOCK_NEW) {
7091                 if (recovery_type == NFS_LOCK_RECLAIM)
7092                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7093         } else
7094                 data->arg.new_lock = 1;
7095         task = rpc_run_task(&task_setup_data);
7096         if (IS_ERR(task))
7097                 return PTR_ERR(task);
7098         ret = rpc_wait_for_completion_task(task);
7099         if (ret == 0) {
7100                 ret = data->rpc_status;
7101                 if (ret)
7102                         nfs4_handle_setlk_error(data->server, data->lsp,
7103                                         data->arg.new_lock_owner, ret);
7104         } else
7105                 data->cancelled = true;
7106         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7107         rpc_put_task(task);
7108         dprintk("%s: done, ret = %d!\n", __func__, ret);
7109         return ret;
7110 }
7111
7112 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7113 {
7114         struct nfs_server *server = NFS_SERVER(state->inode);
7115         struct nfs4_exception exception = {
7116                 .inode = state->inode,
7117         };
7118         int err;
7119
7120         do {
7121                 /* Cache the lock if possible... */
7122                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7123                         return 0;
7124                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7125                 if (err != -NFS4ERR_DELAY)
7126                         break;
7127                 nfs4_handle_exception(server, err, &exception);
7128         } while (exception.retry);
7129         return err;
7130 }
7131
7132 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7133 {
7134         struct nfs_server *server = NFS_SERVER(state->inode);
7135         struct nfs4_exception exception = {
7136                 .inode = state->inode,
7137         };
7138         int err;
7139
7140         err = nfs4_set_lock_state(state, request);
7141         if (err != 0)
7142                 return err;
7143         if (!recover_lost_locks) {
7144                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7145                 return 0;
7146         }
7147         do {
7148                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7149                         return 0;
7150                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7151                 switch (err) {
7152                 default:
7153                         goto out;
7154                 case -NFS4ERR_GRACE:
7155                 case -NFS4ERR_DELAY:
7156                         nfs4_handle_exception(server, err, &exception);
7157                         err = 0;
7158                 }
7159         } while (exception.retry);
7160 out:
7161         return err;
7162 }
7163
7164 #if defined(CONFIG_NFS_V4_1)
7165 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7166 {
7167         struct nfs4_lock_state *lsp;
7168         int status;
7169
7170         status = nfs4_set_lock_state(state, request);
7171         if (status != 0)
7172                 return status;
7173         lsp = request->fl_u.nfs4_fl.owner;
7174         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7175             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7176                 return 0;
7177         return nfs4_lock_expired(state, request);
7178 }
7179 #endif
7180
7181 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7182 {
7183         struct nfs_inode *nfsi = NFS_I(state->inode);
7184         struct nfs4_state_owner *sp = state->owner;
7185         unsigned char fl_flags = request->fl_flags;
7186         int status;
7187
7188         request->fl_flags |= FL_ACCESS;
7189         status = locks_lock_inode_wait(state->inode, request);
7190         if (status < 0)
7191                 goto out;
7192         mutex_lock(&sp->so_delegreturn_mutex);
7193         down_read(&nfsi->rwsem);
7194         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7195                 /* Yes: cache locks! */
7196                 /* ...but avoid races with delegation recall... */
7197                 request->fl_flags = fl_flags & ~FL_SLEEP;
7198                 status = locks_lock_inode_wait(state->inode, request);
7199                 up_read(&nfsi->rwsem);
7200                 mutex_unlock(&sp->so_delegreturn_mutex);
7201                 goto out;
7202         }
7203         up_read(&nfsi->rwsem);
7204         mutex_unlock(&sp->so_delegreturn_mutex);
7205         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7206 out:
7207         request->fl_flags = fl_flags;
7208         return status;
7209 }
7210
7211 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7212 {
7213         struct nfs4_exception exception = {
7214                 .state = state,
7215                 .inode = state->inode,
7216                 .interruptible = true,
7217         };
7218         int err;
7219
7220         do {
7221                 err = _nfs4_proc_setlk(state, cmd, request);
7222                 if (err == -NFS4ERR_DENIED)
7223                         err = -EAGAIN;
7224                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7225                                 err, &exception);
7226         } while (exception.retry);
7227         return err;
7228 }
7229
7230 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7231 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7232
7233 static int
7234 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7235                         struct file_lock *request)
7236 {
7237         int             status = -ERESTARTSYS;
7238         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7239
7240         while(!signalled()) {
7241                 status = nfs4_proc_setlk(state, cmd, request);
7242                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7243                         break;
7244                 freezable_schedule_timeout_interruptible(timeout);
7245                 timeout *= 2;
7246                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7247                 status = -ERESTARTSYS;
7248         }
7249         return status;
7250 }
7251
7252 #ifdef CONFIG_NFS_V4_1
7253 struct nfs4_lock_waiter {
7254         struct task_struct      *task;
7255         struct inode            *inode;
7256         struct nfs_lowner       *owner;
7257 };
7258
7259 static int
7260 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7261 {
7262         int ret;
7263         struct nfs4_lock_waiter *waiter = wait->private;
7264
7265         /* NULL key means to wake up everyone */
7266         if (key) {
7267                 struct cb_notify_lock_args      *cbnl = key;
7268                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7269                                                 *wowner = waiter->owner;
7270
7271                 /* Only wake if the callback was for the same owner. */
7272                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7273                         return 0;
7274
7275                 /* Make sure it's for the right inode */
7276                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7277                         return 0;
7278         }
7279
7280         /* override "private" so we can use default_wake_function */
7281         wait->private = waiter->task;
7282         ret = woken_wake_function(wait, mode, flags, key);
7283         if (ret)
7284                 list_del_init(&wait->entry);
7285         wait->private = waiter;
7286         return ret;
7287 }
7288
7289 static int
7290 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7291 {
7292         int status = -ERESTARTSYS;
7293         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7294         struct nfs_server *server = NFS_SERVER(state->inode);
7295         struct nfs_client *clp = server->nfs_client;
7296         wait_queue_head_t *q = &clp->cl_lock_waitq;
7297         struct nfs_lowner owner = { .clientid = clp->cl_clientid,
7298                                     .id = lsp->ls_seqid.owner_id,
7299                                     .s_dev = server->s_dev };
7300         struct nfs4_lock_waiter waiter = { .task  = current,
7301                                            .inode = state->inode,
7302                                            .owner = &owner};
7303         wait_queue_entry_t wait;
7304
7305         /* Don't bother with waitqueue if we don't expect a callback */
7306         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7307                 return nfs4_retry_setlk_simple(state, cmd, request);
7308
7309         init_wait(&wait);
7310         wait.private = &waiter;
7311         wait.func = nfs4_wake_lock_waiter;
7312
7313         while(!signalled()) {
7314                 add_wait_queue(q, &wait);
7315                 status = nfs4_proc_setlk(state, cmd, request);
7316                 if ((status != -EAGAIN) || IS_SETLK(cmd)) {
7317                         finish_wait(q, &wait);
7318                         break;
7319                 }
7320
7321                 status = -ERESTARTSYS;
7322                 freezer_do_not_count();
7323                 wait_woken(&wait, TASK_INTERRUPTIBLE, NFS4_LOCK_MAXTIMEOUT);
7324                 freezer_count();
7325                 finish_wait(q, &wait);
7326         }
7327
7328         return status;
7329 }
7330 #else /* !CONFIG_NFS_V4_1 */
7331 static inline int
7332 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7333 {
7334         return nfs4_retry_setlk_simple(state, cmd, request);
7335 }
7336 #endif
7337
7338 static int
7339 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7340 {
7341         struct nfs_open_context *ctx;
7342         struct nfs4_state *state;
7343         int status;
7344
7345         /* verify open state */
7346         ctx = nfs_file_open_context(filp);
7347         state = ctx->state;
7348
7349         if (IS_GETLK(cmd)) {
7350                 if (state != NULL)
7351                         return nfs4_proc_getlk(state, F_GETLK, request);
7352                 return 0;
7353         }
7354
7355         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7356                 return -EINVAL;
7357
7358         if (request->fl_type == F_UNLCK) {
7359                 if (state != NULL)
7360                         return nfs4_proc_unlck(state, cmd, request);
7361                 return 0;
7362         }
7363
7364         if (state == NULL)
7365                 return -ENOLCK;
7366
7367         if ((request->fl_flags & FL_POSIX) &&
7368             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7369                 return -ENOLCK;
7370
7371         /*
7372          * Don't rely on the VFS having checked the file open mode,
7373          * since it won't do this for flock() locks.
7374          */
7375         switch (request->fl_type) {
7376         case F_RDLCK:
7377                 if (!(filp->f_mode & FMODE_READ))
7378                         return -EBADF;
7379                 break;
7380         case F_WRLCK:
7381                 if (!(filp->f_mode & FMODE_WRITE))
7382                         return -EBADF;
7383         }
7384
7385         status = nfs4_set_lock_state(state, request);
7386         if (status != 0)
7387                 return status;
7388
7389         return nfs4_retry_setlk(state, cmd, request);
7390 }
7391
7392 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7393 {
7394         struct nfs_server *server = NFS_SERVER(state->inode);
7395         int err;
7396
7397         err = nfs4_set_lock_state(state, fl);
7398         if (err != 0)
7399                 return err;
7400         do {
7401                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7402                 if (err != -NFS4ERR_DELAY)
7403                         break;
7404                 ssleep(1);
7405         } while (err == -NFS4ERR_DELAY);
7406         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7407 }
7408
7409 struct nfs_release_lockowner_data {
7410         struct nfs4_lock_state *lsp;
7411         struct nfs_server *server;
7412         struct nfs_release_lockowner_args args;
7413         struct nfs_release_lockowner_res res;
7414         unsigned long timestamp;
7415 };
7416
7417 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7418 {
7419         struct nfs_release_lockowner_data *data = calldata;
7420         struct nfs_server *server = data->server;
7421         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7422                            &data->res.seq_res, task);
7423         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7424         data->timestamp = jiffies;
7425 }
7426
7427 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7428 {
7429         struct nfs_release_lockowner_data *data = calldata;
7430         struct nfs_server *server = data->server;
7431
7432         nfs40_sequence_done(task, &data->res.seq_res);
7433
7434         switch (task->tk_status) {
7435         case 0:
7436                 renew_lease(server, data->timestamp);
7437                 break;
7438         case -NFS4ERR_STALE_CLIENTID:
7439         case -NFS4ERR_EXPIRED:
7440                 nfs4_schedule_lease_recovery(server->nfs_client);
7441                 break;
7442         case -NFS4ERR_LEASE_MOVED:
7443         case -NFS4ERR_DELAY:
7444                 if (nfs4_async_handle_error(task, server,
7445                                             NULL, NULL) == -EAGAIN)
7446                         rpc_restart_call_prepare(task);
7447         }
7448 }
7449
7450 static void nfs4_release_lockowner_release(void *calldata)
7451 {
7452         struct nfs_release_lockowner_data *data = calldata;
7453         nfs4_free_lock_state(data->server, data->lsp);
7454         kfree(calldata);
7455 }
7456
7457 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7458         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7459         .rpc_call_done = nfs4_release_lockowner_done,
7460         .rpc_release = nfs4_release_lockowner_release,
7461 };
7462
7463 static void
7464 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7465 {
7466         struct nfs_release_lockowner_data *data;
7467         struct rpc_message msg = {
7468                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7469         };
7470
7471         if (server->nfs_client->cl_mvops->minor_version != 0)
7472                 return;
7473
7474         data = kmalloc(sizeof(*data), GFP_NOFS);
7475         if (!data)
7476                 return;
7477         data->lsp = lsp;
7478         data->server = server;
7479         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7480         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7481         data->args.lock_owner.s_dev = server->s_dev;
7482
7483         msg.rpc_argp = &data->args;
7484         msg.rpc_resp = &data->res;
7485         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7486         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7487 }
7488
7489 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7490
7491 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7492                                    struct user_namespace *mnt_userns,
7493                                    struct dentry *unused, struct inode *inode,
7494                                    const char *key, const void *buf,
7495                                    size_t buflen, int flags)
7496 {
7497         return nfs4_proc_set_acl(inode, buf, buflen);
7498 }
7499
7500 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7501                                    struct dentry *unused, struct inode *inode,
7502                                    const char *key, void *buf, size_t buflen)
7503 {
7504         return nfs4_proc_get_acl(inode, buf, buflen);
7505 }
7506
7507 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7508 {
7509         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7510 }
7511
7512 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7513
7514 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7515                                      struct user_namespace *mnt_userns,
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 user_namespace *mnt_userns,
7567                                     struct dentry *unused, struct inode *inode,
7568                                     const char *key, const void *buf,
7569                                     size_t buflen, int flags)
7570 {
7571         struct nfs_access_entry cache;
7572         int ret;
7573
7574         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7575                 return -EOPNOTSUPP;
7576
7577         /*
7578          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7579          * flags right now. Handling of xattr operations use the normal
7580          * file read/write permissions.
7581          *
7582          * Just in case the server has other ideas (which RFC 8276 allows),
7583          * do a cached access check for the XA* flags to possibly avoid
7584          * doing an RPC and getting EACCES back.
7585          */
7586         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7587                 if (!(cache.mask & NFS_ACCESS_XAWRITE))
7588                         return -EACCES;
7589         }
7590
7591         if (buf == NULL) {
7592                 ret = nfs42_proc_removexattr(inode, key);
7593                 if (!ret)
7594                         nfs4_xattr_cache_remove(inode, key);
7595         } else {
7596                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7597                 if (!ret)
7598                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7599         }
7600
7601         return ret;
7602 }
7603
7604 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7605                                     struct dentry *unused, struct inode *inode,
7606                                     const char *key, void *buf, size_t buflen)
7607 {
7608         struct nfs_access_entry cache;
7609         ssize_t ret;
7610
7611         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7612                 return -EOPNOTSUPP;
7613
7614         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7615                 if (!(cache.mask & NFS_ACCESS_XAREAD))
7616                         return -EACCES;
7617         }
7618
7619         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7620         if (ret)
7621                 return ret;
7622
7623         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7624         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7625                 return ret;
7626
7627         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7628
7629         return ret;
7630 }
7631
7632 static ssize_t
7633 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7634 {
7635         u64 cookie;
7636         bool eof;
7637         ssize_t ret, size;
7638         char *buf;
7639         size_t buflen;
7640         struct nfs_access_entry cache;
7641
7642         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7643                 return 0;
7644
7645         if (!nfs_access_get_cached(inode, current_cred(), &cache, true)) {
7646                 if (!(cache.mask & NFS_ACCESS_XALIST))
7647                         return 0;
7648         }
7649
7650         ret = nfs_revalidate_inode(NFS_SERVER(inode), inode);
7651         if (ret)
7652                 return ret;
7653
7654         ret = nfs4_xattr_cache_list(inode, list, list_len);
7655         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7656                 return ret;
7657
7658         cookie = 0;
7659         eof = false;
7660         buflen = list_len ? list_len : XATTR_LIST_MAX;
7661         buf = list_len ? list : NULL;
7662         size = 0;
7663
7664         while (!eof) {
7665                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7666                     &cookie, &eof);
7667                 if (ret < 0)
7668                         return ret;
7669
7670                 if (list_len) {
7671                         buf += ret;
7672                         buflen -= ret;
7673                 }
7674                 size += ret;
7675         }
7676
7677         if (list_len)
7678                 nfs4_xattr_cache_set_list(inode, list, size);
7679
7680         return size;
7681 }
7682
7683 #else
7684
7685 static ssize_t
7686 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7687 {
7688         return 0;
7689 }
7690 #endif /* CONFIG_NFS_V4_2 */
7691
7692 /*
7693  * nfs_fhget will use either the mounted_on_fileid or the fileid
7694  */
7695 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7696 {
7697         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7698                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7699               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7700               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7701                 return;
7702
7703         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7704                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7705         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7706         fattr->nlink = 2;
7707 }
7708
7709 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7710                                    const struct qstr *name,
7711                                    struct nfs4_fs_locations *fs_locations,
7712                                    struct page *page)
7713 {
7714         struct nfs_server *server = NFS_SERVER(dir);
7715         u32 bitmask[3];
7716         struct nfs4_fs_locations_arg args = {
7717                 .dir_fh = NFS_FH(dir),
7718                 .name = name,
7719                 .page = page,
7720                 .bitmask = bitmask,
7721         };
7722         struct nfs4_fs_locations_res res = {
7723                 .fs_locations = fs_locations,
7724         };
7725         struct rpc_message msg = {
7726                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7727                 .rpc_argp = &args,
7728                 .rpc_resp = &res,
7729         };
7730         int status;
7731
7732         dprintk("%s: start\n", __func__);
7733
7734         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7735         bitmask[1] = nfs4_fattr_bitmap[1];
7736
7737         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7738          * is not supported */
7739         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7740                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7741         else
7742                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7743
7744         nfs_fattr_init(&fs_locations->fattr);
7745         fs_locations->server = server;
7746         fs_locations->nlocations = 0;
7747         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7748         dprintk("%s: returned status = %d\n", __func__, status);
7749         return status;
7750 }
7751
7752 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7753                            const struct qstr *name,
7754                            struct nfs4_fs_locations *fs_locations,
7755                            struct page *page)
7756 {
7757         struct nfs4_exception exception = {
7758                 .interruptible = true,
7759         };
7760         int err;
7761         do {
7762                 err = _nfs4_proc_fs_locations(client, dir, name,
7763                                 fs_locations, page);
7764                 trace_nfs4_get_fs_locations(dir, name, err);
7765                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7766                                 &exception);
7767         } while (exception.retry);
7768         return err;
7769 }
7770
7771 /*
7772  * This operation also signals the server that this client is
7773  * performing migration recovery.  The server can stop returning
7774  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7775  * appended to this compound to identify the client ID which is
7776  * performing recovery.
7777  */
7778 static int _nfs40_proc_get_locations(struct inode *inode,
7779                                      struct nfs4_fs_locations *locations,
7780                                      struct page *page, const struct cred *cred)
7781 {
7782         struct nfs_server *server = NFS_SERVER(inode);
7783         struct rpc_clnt *clnt = server->client;
7784         u32 bitmask[2] = {
7785                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7786         };
7787         struct nfs4_fs_locations_arg args = {
7788                 .clientid       = server->nfs_client->cl_clientid,
7789                 .fh             = NFS_FH(inode),
7790                 .page           = page,
7791                 .bitmask        = bitmask,
7792                 .migration      = 1,            /* skip LOOKUP */
7793                 .renew          = 1,            /* append RENEW */
7794         };
7795         struct nfs4_fs_locations_res res = {
7796                 .fs_locations   = locations,
7797                 .migration      = 1,
7798                 .renew          = 1,
7799         };
7800         struct rpc_message msg = {
7801                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7802                 .rpc_argp       = &args,
7803                 .rpc_resp       = &res,
7804                 .rpc_cred       = cred,
7805         };
7806         unsigned long now = jiffies;
7807         int status;
7808
7809         nfs_fattr_init(&locations->fattr);
7810         locations->server = server;
7811         locations->nlocations = 0;
7812
7813         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7814         status = nfs4_call_sync_sequence(clnt, server, &msg,
7815                                         &args.seq_args, &res.seq_res);
7816         if (status)
7817                 return status;
7818
7819         renew_lease(server, now);
7820         return 0;
7821 }
7822
7823 #ifdef CONFIG_NFS_V4_1
7824
7825 /*
7826  * This operation also signals the server that this client is
7827  * performing migration recovery.  The server can stop asserting
7828  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7829  * performing this operation is identified in the SEQUENCE
7830  * operation in this compound.
7831  *
7832  * When the client supports GETATTR(fs_locations_info), it can
7833  * be plumbed in here.
7834  */
7835 static int _nfs41_proc_get_locations(struct inode *inode,
7836                                      struct nfs4_fs_locations *locations,
7837                                      struct page *page, const struct cred *cred)
7838 {
7839         struct nfs_server *server = NFS_SERVER(inode);
7840         struct rpc_clnt *clnt = server->client;
7841         u32 bitmask[2] = {
7842                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7843         };
7844         struct nfs4_fs_locations_arg args = {
7845                 .fh             = NFS_FH(inode),
7846                 .page           = page,
7847                 .bitmask        = bitmask,
7848                 .migration      = 1,            /* skip LOOKUP */
7849         };
7850         struct nfs4_fs_locations_res res = {
7851                 .fs_locations   = locations,
7852                 .migration      = 1,
7853         };
7854         struct rpc_message msg = {
7855                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7856                 .rpc_argp       = &args,
7857                 .rpc_resp       = &res,
7858                 .rpc_cred       = cred,
7859         };
7860         int status;
7861
7862         nfs_fattr_init(&locations->fattr);
7863         locations->server = server;
7864         locations->nlocations = 0;
7865
7866         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7867         status = nfs4_call_sync_sequence(clnt, server, &msg,
7868                                         &args.seq_args, &res.seq_res);
7869         if (status == NFS4_OK &&
7870             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
7871                 status = -NFS4ERR_LEASE_MOVED;
7872         return status;
7873 }
7874
7875 #endif  /* CONFIG_NFS_V4_1 */
7876
7877 /**
7878  * nfs4_proc_get_locations - discover locations for a migrated FSID
7879  * @inode: inode on FSID that is migrating
7880  * @locations: result of query
7881  * @page: buffer
7882  * @cred: credential to use for this operation
7883  *
7884  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
7885  * operation failed, or a negative errno if a local error occurred.
7886  *
7887  * On success, "locations" is filled in, but if the server has
7888  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
7889  * asserted.
7890  *
7891  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
7892  * from this client that require migration recovery.
7893  */
7894 int nfs4_proc_get_locations(struct inode *inode,
7895                             struct nfs4_fs_locations *locations,
7896                             struct page *page, const struct cred *cred)
7897 {
7898         struct nfs_server *server = NFS_SERVER(inode);
7899         struct nfs_client *clp = server->nfs_client;
7900         const struct nfs4_mig_recovery_ops *ops =
7901                                         clp->cl_mvops->mig_recovery_ops;
7902         struct nfs4_exception exception = {
7903                 .interruptible = true,
7904         };
7905         int status;
7906
7907         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
7908                 (unsigned long long)server->fsid.major,
7909                 (unsigned long long)server->fsid.minor,
7910                 clp->cl_hostname);
7911         nfs_display_fhandle(NFS_FH(inode), __func__);
7912
7913         do {
7914                 status = ops->get_locations(inode, locations, page, cred);
7915                 if (status != -NFS4ERR_DELAY)
7916                         break;
7917                 nfs4_handle_exception(server, status, &exception);
7918         } while (exception.retry);
7919         return status;
7920 }
7921
7922 /*
7923  * This operation also signals the server that this client is
7924  * performing "lease moved" recovery.  The server can stop
7925  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
7926  * is appended to this compound to identify the client ID which is
7927  * performing recovery.
7928  */
7929 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
7930 {
7931         struct nfs_server *server = NFS_SERVER(inode);
7932         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
7933         struct rpc_clnt *clnt = server->client;
7934         struct nfs4_fsid_present_arg args = {
7935                 .fh             = NFS_FH(inode),
7936                 .clientid       = clp->cl_clientid,
7937                 .renew          = 1,            /* append RENEW */
7938         };
7939         struct nfs4_fsid_present_res res = {
7940                 .renew          = 1,
7941         };
7942         struct rpc_message msg = {
7943                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7944                 .rpc_argp       = &args,
7945                 .rpc_resp       = &res,
7946                 .rpc_cred       = cred,
7947         };
7948         unsigned long now = jiffies;
7949         int status;
7950
7951         res.fh = nfs_alloc_fhandle();
7952         if (res.fh == NULL)
7953                 return -ENOMEM;
7954
7955         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7956         status = nfs4_call_sync_sequence(clnt, server, &msg,
7957                                                 &args.seq_args, &res.seq_res);
7958         nfs_free_fhandle(res.fh);
7959         if (status)
7960                 return status;
7961
7962         do_renew_lease(clp, now);
7963         return 0;
7964 }
7965
7966 #ifdef CONFIG_NFS_V4_1
7967
7968 /*
7969  * This operation also signals the server that this client is
7970  * performing "lease moved" recovery.  The server can stop asserting
7971  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
7972  * this operation is identified in the SEQUENCE operation in this
7973  * compound.
7974  */
7975 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
7976 {
7977         struct nfs_server *server = NFS_SERVER(inode);
7978         struct rpc_clnt *clnt = server->client;
7979         struct nfs4_fsid_present_arg args = {
7980                 .fh             = NFS_FH(inode),
7981         };
7982         struct nfs4_fsid_present_res res = {
7983         };
7984         struct rpc_message msg = {
7985                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
7986                 .rpc_argp       = &args,
7987                 .rpc_resp       = &res,
7988                 .rpc_cred       = cred,
7989         };
7990         int status;
7991
7992         res.fh = nfs_alloc_fhandle();
7993         if (res.fh == NULL)
7994                 return -ENOMEM;
7995
7996         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7997         status = nfs4_call_sync_sequence(clnt, server, &msg,
7998                                                 &args.seq_args, &res.seq_res);
7999         nfs_free_fhandle(res.fh);
8000         if (status == NFS4_OK &&
8001             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8002                 status = -NFS4ERR_LEASE_MOVED;
8003         return status;
8004 }
8005
8006 #endif  /* CONFIG_NFS_V4_1 */
8007
8008 /**
8009  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8010  * @inode: inode on FSID to check
8011  * @cred: credential to use for this operation
8012  *
8013  * Server indicates whether the FSID is present, moved, or not
8014  * recognized.  This operation is necessary to clear a LEASE_MOVED
8015  * condition for this client ID.
8016  *
8017  * Returns NFS4_OK if the FSID is present on this server,
8018  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8019  *  NFS4ERR code if some error occurred on the server, or a
8020  *  negative errno if a local failure occurred.
8021  */
8022 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8023 {
8024         struct nfs_server *server = NFS_SERVER(inode);
8025         struct nfs_client *clp = server->nfs_client;
8026         const struct nfs4_mig_recovery_ops *ops =
8027                                         clp->cl_mvops->mig_recovery_ops;
8028         struct nfs4_exception exception = {
8029                 .interruptible = true,
8030         };
8031         int status;
8032
8033         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8034                 (unsigned long long)server->fsid.major,
8035                 (unsigned long long)server->fsid.minor,
8036                 clp->cl_hostname);
8037         nfs_display_fhandle(NFS_FH(inode), __func__);
8038
8039         do {
8040                 status = ops->fsid_present(inode, cred);
8041                 if (status != -NFS4ERR_DELAY)
8042                         break;
8043                 nfs4_handle_exception(server, status, &exception);
8044         } while (exception.retry);
8045         return status;
8046 }
8047
8048 /*
8049  * If 'use_integrity' is true and the state managment nfs_client
8050  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8051  * and the machine credential as per RFC3530bis and RFC5661 Security
8052  * Considerations sections. Otherwise, just use the user cred with the
8053  * filesystem's rpc_client.
8054  */
8055 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8056 {
8057         int status;
8058         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8059         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8060         struct nfs4_secinfo_arg args = {
8061                 .dir_fh = NFS_FH(dir),
8062                 .name   = name,
8063         };
8064         struct nfs4_secinfo_res res = {
8065                 .flavors     = flavors,
8066         };
8067         struct rpc_message msg = {
8068                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8069                 .rpc_argp = &args,
8070                 .rpc_resp = &res,
8071         };
8072         struct nfs4_call_sync_data data = {
8073                 .seq_server = NFS_SERVER(dir),
8074                 .seq_args = &args.seq_args,
8075                 .seq_res = &res.seq_res,
8076         };
8077         struct rpc_task_setup task_setup = {
8078                 .rpc_client = clnt,
8079                 .rpc_message = &msg,
8080                 .callback_ops = clp->cl_mvops->call_sync_ops,
8081                 .callback_data = &data,
8082                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8083         };
8084         const struct cred *cred = NULL;
8085
8086         if (use_integrity) {
8087                 clnt = clp->cl_rpcclient;
8088                 task_setup.rpc_client = clnt;
8089
8090                 cred = nfs4_get_clid_cred(clp);
8091                 msg.rpc_cred = cred;
8092         }
8093
8094         dprintk("NFS call  secinfo %s\n", name->name);
8095
8096         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8097         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8098         status = nfs4_call_sync_custom(&task_setup);
8099
8100         dprintk("NFS reply  secinfo: %d\n", status);
8101
8102         put_cred(cred);
8103         return status;
8104 }
8105
8106 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8107                       struct nfs4_secinfo_flavors *flavors)
8108 {
8109         struct nfs4_exception exception = {
8110                 .interruptible = true,
8111         };
8112         int err;
8113         do {
8114                 err = -NFS4ERR_WRONGSEC;
8115
8116                 /* try to use integrity protection with machine cred */
8117                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8118                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8119
8120                 /*
8121                  * if unable to use integrity protection, or SECINFO with
8122                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8123                  * disallowed by spec, but exists in deployed servers) use
8124                  * the current filesystem's rpc_client and the user cred.
8125                  */
8126                 if (err == -NFS4ERR_WRONGSEC)
8127                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8128
8129                 trace_nfs4_secinfo(dir, name, err);
8130                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8131                                 &exception);
8132         } while (exception.retry);
8133         return err;
8134 }
8135
8136 #ifdef CONFIG_NFS_V4_1
8137 /*
8138  * Check the exchange flags returned by the server for invalid flags, having
8139  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8140  * DS flags set.
8141  */
8142 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8143 {
8144         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8145                 goto out_inval;
8146         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8147                 goto out_inval;
8148         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8149             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8150                 goto out_inval;
8151         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8152                 goto out_inval;
8153         return NFS_OK;
8154 out_inval:
8155         return -NFS4ERR_INVAL;
8156 }
8157
8158 static bool
8159 nfs41_same_server_scope(struct nfs41_server_scope *a,
8160                         struct nfs41_server_scope *b)
8161 {
8162         if (a->server_scope_sz != b->server_scope_sz)
8163                 return false;
8164         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8165 }
8166
8167 static void
8168 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8169 {
8170         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8171         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8172         struct nfs_client *clp = args->client;
8173
8174         switch (task->tk_status) {
8175         case -NFS4ERR_BADSESSION:
8176         case -NFS4ERR_DEADSESSION:
8177                 nfs4_schedule_session_recovery(clp->cl_session,
8178                                 task->tk_status);
8179         }
8180         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8181                         res->dir != NFS4_CDFS4_BOTH) {
8182                 rpc_task_close_connection(task);
8183                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8184                         rpc_restart_call(task);
8185         }
8186 }
8187
8188 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8189         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8190 };
8191
8192 /*
8193  * nfs4_proc_bind_one_conn_to_session()
8194  *
8195  * The 4.1 client currently uses the same TCP connection for the
8196  * fore and backchannel.
8197  */
8198 static
8199 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8200                 struct rpc_xprt *xprt,
8201                 struct nfs_client *clp,
8202                 const struct cred *cred)
8203 {
8204         int status;
8205         struct nfs41_bind_conn_to_session_args args = {
8206                 .client = clp,
8207                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8208                 .retries = 0,
8209         };
8210         struct nfs41_bind_conn_to_session_res res;
8211         struct rpc_message msg = {
8212                 .rpc_proc =
8213                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8214                 .rpc_argp = &args,
8215                 .rpc_resp = &res,
8216                 .rpc_cred = cred,
8217         };
8218         struct rpc_task_setup task_setup_data = {
8219                 .rpc_client = clnt,
8220                 .rpc_xprt = xprt,
8221                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8222                 .rpc_message = &msg,
8223                 .flags = RPC_TASK_TIMEOUT,
8224         };
8225         struct rpc_task *task;
8226
8227         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8228         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8229                 args.dir = NFS4_CDFC4_FORE;
8230
8231         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8232         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8233                 args.dir = NFS4_CDFC4_FORE;
8234
8235         task = rpc_run_task(&task_setup_data);
8236         if (!IS_ERR(task)) {
8237                 status = task->tk_status;
8238                 rpc_put_task(task);
8239         } else
8240                 status = PTR_ERR(task);
8241         trace_nfs4_bind_conn_to_session(clp, status);
8242         if (status == 0) {
8243                 if (memcmp(res.sessionid.data,
8244                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8245                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8246                         return -EIO;
8247                 }
8248                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8249                         dprintk("NFS: %s: Unexpected direction from server\n",
8250                                 __func__);
8251                         return -EIO;
8252                 }
8253                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8254                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8255                                 __func__);
8256                         return -EIO;
8257                 }
8258         }
8259
8260         return status;
8261 }
8262
8263 struct rpc_bind_conn_calldata {
8264         struct nfs_client *clp;
8265         const struct cred *cred;
8266 };
8267
8268 static int
8269 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8270                 struct rpc_xprt *xprt,
8271                 void *calldata)
8272 {
8273         struct rpc_bind_conn_calldata *p = calldata;
8274
8275         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8276 }
8277
8278 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8279 {
8280         struct rpc_bind_conn_calldata data = {
8281                 .clp = clp,
8282                 .cred = cred,
8283         };
8284         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8285                         nfs4_proc_bind_conn_to_session_callback, &data);
8286 }
8287
8288 /*
8289  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8290  * and operations we'd like to see to enable certain features in the allow map
8291  */
8292 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8293         .how = SP4_MACH_CRED,
8294         .enforce.u.words = {
8295                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8296                       1 << (OP_EXCHANGE_ID - 32) |
8297                       1 << (OP_CREATE_SESSION - 32) |
8298                       1 << (OP_DESTROY_SESSION - 32) |
8299                       1 << (OP_DESTROY_CLIENTID - 32)
8300         },
8301         .allow.u.words = {
8302                 [0] = 1 << (OP_CLOSE) |
8303                       1 << (OP_OPEN_DOWNGRADE) |
8304                       1 << (OP_LOCKU) |
8305                       1 << (OP_DELEGRETURN) |
8306                       1 << (OP_COMMIT),
8307                 [1] = 1 << (OP_SECINFO - 32) |
8308                       1 << (OP_SECINFO_NO_NAME - 32) |
8309                       1 << (OP_LAYOUTRETURN - 32) |
8310                       1 << (OP_TEST_STATEID - 32) |
8311                       1 << (OP_FREE_STATEID - 32) |
8312                       1 << (OP_WRITE - 32)
8313         }
8314 };
8315
8316 /*
8317  * Select the state protection mode for client `clp' given the server results
8318  * from exchange_id in `sp'.
8319  *
8320  * Returns 0 on success, negative errno otherwise.
8321  */
8322 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8323                                  struct nfs41_state_protection *sp)
8324 {
8325         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8326                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8327                       1 << (OP_EXCHANGE_ID - 32) |
8328                       1 << (OP_CREATE_SESSION - 32) |
8329                       1 << (OP_DESTROY_SESSION - 32) |
8330                       1 << (OP_DESTROY_CLIENTID - 32)
8331         };
8332         unsigned long flags = 0;
8333         unsigned int i;
8334         int ret = 0;
8335
8336         if (sp->how == SP4_MACH_CRED) {
8337                 /* Print state protect result */
8338                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8339                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8340                         if (test_bit(i, sp->enforce.u.longs))
8341                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8342                         if (test_bit(i, sp->allow.u.longs))
8343                                 dfprintk(MOUNT, "  allow op %d\n", i);
8344                 }
8345
8346                 /* make sure nothing is on enforce list that isn't supported */
8347                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8348                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8349                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8350                                 ret = -EINVAL;
8351                                 goto out;
8352                         }
8353                 }
8354
8355                 /*
8356                  * Minimal mode - state operations are allowed to use machine
8357                  * credential.  Note this already happens by default, so the
8358                  * client doesn't have to do anything more than the negotiation.
8359                  *
8360                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8361                  *       we're already using the machine cred for exchange_id
8362                  *       and will never use a different cred.
8363                  */
8364                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8365                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8366                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8367                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8368                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8369                         dfprintk(MOUNT, "  minimal mode enabled\n");
8370                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8371                 } else {
8372                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8373                         ret = -EINVAL;
8374                         goto out;
8375                 }
8376
8377                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8378                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8379                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8380                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8381                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8382                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8383                 }
8384
8385                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8386                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8387                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8388                 }
8389
8390                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8391                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8392                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8393                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8394                 }
8395
8396                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8397                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8398                         dfprintk(MOUNT, "  stateid mode enabled\n");
8399                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8400                 }
8401
8402                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8403                         dfprintk(MOUNT, "  write mode enabled\n");
8404                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8405                 }
8406
8407                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8408                         dfprintk(MOUNT, "  commit mode enabled\n");
8409                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8410                 }
8411         }
8412 out:
8413         clp->cl_sp4_flags = flags;
8414         return ret;
8415 }
8416
8417 struct nfs41_exchange_id_data {
8418         struct nfs41_exchange_id_res res;
8419         struct nfs41_exchange_id_args args;
8420 };
8421
8422 static void nfs4_exchange_id_release(void *data)
8423 {
8424         struct nfs41_exchange_id_data *cdata =
8425                                         (struct nfs41_exchange_id_data *)data;
8426
8427         nfs_put_client(cdata->args.client);
8428         kfree(cdata->res.impl_id);
8429         kfree(cdata->res.server_scope);
8430         kfree(cdata->res.server_owner);
8431         kfree(cdata);
8432 }
8433
8434 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8435         .rpc_release = nfs4_exchange_id_release,
8436 };
8437
8438 /*
8439  * _nfs4_proc_exchange_id()
8440  *
8441  * Wrapper for EXCHANGE_ID operation.
8442  */
8443 static struct rpc_task *
8444 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8445                         u32 sp4_how, struct rpc_xprt *xprt)
8446 {
8447         struct rpc_message msg = {
8448                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8449                 .rpc_cred = cred,
8450         };
8451         struct rpc_task_setup task_setup_data = {
8452                 .rpc_client = clp->cl_rpcclient,
8453                 .callback_ops = &nfs4_exchange_id_call_ops,
8454                 .rpc_message = &msg,
8455                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8456         };
8457         struct nfs41_exchange_id_data *calldata;
8458         int status;
8459
8460         if (!refcount_inc_not_zero(&clp->cl_count))
8461                 return ERR_PTR(-EIO);
8462
8463         status = -ENOMEM;
8464         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8465         if (!calldata)
8466                 goto out;
8467
8468         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8469
8470         status = nfs4_init_uniform_client_string(clp);
8471         if (status)
8472                 goto out_calldata;
8473
8474         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8475                                                 GFP_NOFS);
8476         status = -ENOMEM;
8477         if (unlikely(calldata->res.server_owner == NULL))
8478                 goto out_calldata;
8479
8480         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8481                                         GFP_NOFS);
8482         if (unlikely(calldata->res.server_scope == NULL))
8483                 goto out_server_owner;
8484
8485         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8486         if (unlikely(calldata->res.impl_id == NULL))
8487                 goto out_server_scope;
8488
8489         switch (sp4_how) {
8490         case SP4_NONE:
8491                 calldata->args.state_protect.how = SP4_NONE;
8492                 break;
8493
8494         case SP4_MACH_CRED:
8495                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8496                 break;
8497
8498         default:
8499                 /* unsupported! */
8500                 WARN_ON_ONCE(1);
8501                 status = -EINVAL;
8502                 goto out_impl_id;
8503         }
8504         if (xprt) {
8505                 task_setup_data.rpc_xprt = xprt;
8506                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8507                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8508                                 sizeof(calldata->args.verifier.data));
8509         }
8510         calldata->args.client = clp;
8511         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8512         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8513 #ifdef CONFIG_NFS_V4_1_MIGRATION
8514         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8515 #endif
8516         msg.rpc_argp = &calldata->args;
8517         msg.rpc_resp = &calldata->res;
8518         task_setup_data.callback_data = calldata;
8519
8520         return rpc_run_task(&task_setup_data);
8521
8522 out_impl_id:
8523         kfree(calldata->res.impl_id);
8524 out_server_scope:
8525         kfree(calldata->res.server_scope);
8526 out_server_owner:
8527         kfree(calldata->res.server_owner);
8528 out_calldata:
8529         kfree(calldata);
8530 out:
8531         nfs_put_client(clp);
8532         return ERR_PTR(status);
8533 }
8534
8535 /*
8536  * _nfs4_proc_exchange_id()
8537  *
8538  * Wrapper for EXCHANGE_ID operation.
8539  */
8540 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8541                         u32 sp4_how)
8542 {
8543         struct rpc_task *task;
8544         struct nfs41_exchange_id_args *argp;
8545         struct nfs41_exchange_id_res *resp;
8546         unsigned long now = jiffies;
8547         int status;
8548
8549         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8550         if (IS_ERR(task))
8551                 return PTR_ERR(task);
8552
8553         argp = task->tk_msg.rpc_argp;
8554         resp = task->tk_msg.rpc_resp;
8555         status = task->tk_status;
8556         if (status  != 0)
8557                 goto out;
8558
8559         status = nfs4_check_cl_exchange_flags(resp->flags,
8560                         clp->cl_mvops->minor_version);
8561         if (status  != 0)
8562                 goto out;
8563
8564         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8565         if (status != 0)
8566                 goto out;
8567
8568         do_renew_lease(clp, now);
8569
8570         clp->cl_clientid = resp->clientid;
8571         clp->cl_exchange_flags = resp->flags;
8572         clp->cl_seqid = resp->seqid;
8573         /* Client ID is not confirmed */
8574         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8575                 clear_bit(NFS4_SESSION_ESTABLISHED,
8576                           &clp->cl_session->session_state);
8577
8578         if (clp->cl_serverscope != NULL &&
8579             !nfs41_same_server_scope(clp->cl_serverscope,
8580                                 resp->server_scope)) {
8581                 dprintk("%s: server_scope mismatch detected\n",
8582                         __func__);
8583                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8584         }
8585
8586         swap(clp->cl_serverowner, resp->server_owner);
8587         swap(clp->cl_serverscope, resp->server_scope);
8588         swap(clp->cl_implid, resp->impl_id);
8589
8590         /* Save the EXCHANGE_ID verifier session trunk tests */
8591         memcpy(clp->cl_confirm.data, argp->verifier.data,
8592                sizeof(clp->cl_confirm.data));
8593 out:
8594         trace_nfs4_exchange_id(clp, status);
8595         rpc_put_task(task);
8596         return status;
8597 }
8598
8599 /*
8600  * nfs4_proc_exchange_id()
8601  *
8602  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8603  *
8604  * Since the clientid has expired, all compounds using sessions
8605  * associated with the stale clientid will be returning
8606  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8607  * be in some phase of session reset.
8608  *
8609  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8610  */
8611 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8612 {
8613         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8614         int status;
8615
8616         /* try SP4_MACH_CRED if krb5i/p */
8617         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8618             authflavor == RPC_AUTH_GSS_KRB5P) {
8619                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8620                 if (!status)
8621                         return 0;
8622         }
8623
8624         /* try SP4_NONE */
8625         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8626 }
8627
8628 /**
8629  * nfs4_test_session_trunk
8630  *
8631  * This is an add_xprt_test() test function called from
8632  * rpc_clnt_setup_test_and_add_xprt.
8633  *
8634  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8635  * and is dereferrenced in nfs4_exchange_id_release
8636  *
8637  * Upon success, add the new transport to the rpc_clnt
8638  *
8639  * @clnt: struct rpc_clnt to get new transport
8640  * @xprt: the rpc_xprt to test
8641  * @data: call data for _nfs4_proc_exchange_id.
8642  */
8643 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8644                             void *data)
8645 {
8646         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8647         struct rpc_task *task;
8648         int status;
8649
8650         u32 sp4_how;
8651
8652         dprintk("--> %s try %s\n", __func__,
8653                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8654
8655         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8656
8657         /* Test connection for session trunking. Async exchange_id call */
8658         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8659         if (IS_ERR(task))
8660                 return;
8661
8662         status = task->tk_status;
8663         if (status == 0)
8664                 status = nfs4_detect_session_trunking(adata->clp,
8665                                 task->tk_msg.rpc_resp, xprt);
8666
8667         if (status == 0)
8668                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8669
8670         rpc_put_task(task);
8671 }
8672 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8673
8674 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8675                 const struct cred *cred)
8676 {
8677         struct rpc_message msg = {
8678                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8679                 .rpc_argp = clp,
8680                 .rpc_cred = cred,
8681         };
8682         int status;
8683
8684         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8685                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8686         trace_nfs4_destroy_clientid(clp, status);
8687         if (status)
8688                 dprintk("NFS: Got error %d from the server %s on "
8689                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8690         return status;
8691 }
8692
8693 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8694                 const struct cred *cred)
8695 {
8696         unsigned int loop;
8697         int ret;
8698
8699         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8700                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8701                 switch (ret) {
8702                 case -NFS4ERR_DELAY:
8703                 case -NFS4ERR_CLIENTID_BUSY:
8704                         ssleep(1);
8705                         break;
8706                 default:
8707                         return ret;
8708                 }
8709         }
8710         return 0;
8711 }
8712
8713 int nfs4_destroy_clientid(struct nfs_client *clp)
8714 {
8715         const struct cred *cred;
8716         int ret = 0;
8717
8718         if (clp->cl_mvops->minor_version < 1)
8719                 goto out;
8720         if (clp->cl_exchange_flags == 0)
8721                 goto out;
8722         if (clp->cl_preserve_clid)
8723                 goto out;
8724         cred = nfs4_get_clid_cred(clp);
8725         ret = nfs4_proc_destroy_clientid(clp, cred);
8726         put_cred(cred);
8727         switch (ret) {
8728         case 0:
8729         case -NFS4ERR_STALE_CLIENTID:
8730                 clp->cl_exchange_flags = 0;
8731         }
8732 out:
8733         return ret;
8734 }
8735
8736 #endif /* CONFIG_NFS_V4_1 */
8737
8738 struct nfs4_get_lease_time_data {
8739         struct nfs4_get_lease_time_args *args;
8740         struct nfs4_get_lease_time_res *res;
8741         struct nfs_client *clp;
8742 };
8743
8744 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8745                                         void *calldata)
8746 {
8747         struct nfs4_get_lease_time_data *data =
8748                         (struct nfs4_get_lease_time_data *)calldata;
8749
8750         dprintk("--> %s\n", __func__);
8751         /* just setup sequence, do not trigger session recovery
8752            since we're invoked within one */
8753         nfs4_setup_sequence(data->clp,
8754                         &data->args->la_seq_args,
8755                         &data->res->lr_seq_res,
8756                         task);
8757         dprintk("<-- %s\n", __func__);
8758 }
8759
8760 /*
8761  * Called from nfs4_state_manager thread for session setup, so don't recover
8762  * from sequence operation or clientid errors.
8763  */
8764 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8765 {
8766         struct nfs4_get_lease_time_data *data =
8767                         (struct nfs4_get_lease_time_data *)calldata;
8768
8769         dprintk("--> %s\n", __func__);
8770         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8771                 return;
8772         switch (task->tk_status) {
8773         case -NFS4ERR_DELAY:
8774         case -NFS4ERR_GRACE:
8775                 dprintk("%s Retry: tk_status %d\n", __func__, task->tk_status);
8776                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8777                 task->tk_status = 0;
8778                 fallthrough;
8779         case -NFS4ERR_RETRY_UNCACHED_REP:
8780                 rpc_restart_call_prepare(task);
8781                 return;
8782         }
8783         dprintk("<-- %s\n", __func__);
8784 }
8785
8786 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8787         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8788         .rpc_call_done = nfs4_get_lease_time_done,
8789 };
8790
8791 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8792 {
8793         struct nfs4_get_lease_time_args args;
8794         struct nfs4_get_lease_time_res res = {
8795                 .lr_fsinfo = fsinfo,
8796         };
8797         struct nfs4_get_lease_time_data data = {
8798                 .args = &args,
8799                 .res = &res,
8800                 .clp = clp,
8801         };
8802         struct rpc_message msg = {
8803                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8804                 .rpc_argp = &args,
8805                 .rpc_resp = &res,
8806         };
8807         struct rpc_task_setup task_setup = {
8808                 .rpc_client = clp->cl_rpcclient,
8809                 .rpc_message = &msg,
8810                 .callback_ops = &nfs4_get_lease_time_ops,
8811                 .callback_data = &data,
8812                 .flags = RPC_TASK_TIMEOUT,
8813         };
8814
8815         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8816         return nfs4_call_sync_custom(&task_setup);
8817 }
8818
8819 #ifdef CONFIG_NFS_V4_1
8820
8821 /*
8822  * Initialize the values to be used by the client in CREATE_SESSION
8823  * If nfs4_init_session set the fore channel request and response sizes,
8824  * use them.
8825  *
8826  * Set the back channel max_resp_sz_cached to zero to force the client to
8827  * always set csa_cachethis to FALSE because the current implementation
8828  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8829  */
8830 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8831                                     struct rpc_clnt *clnt)
8832 {
8833         unsigned int max_rqst_sz, max_resp_sz;
8834         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8835         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8836
8837         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8838         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8839
8840         /* Fore channel attributes */
8841         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8842         args->fc_attrs.max_resp_sz = max_resp_sz;
8843         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8844         args->fc_attrs.max_reqs = max_session_slots;
8845
8846         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8847                 "max_ops=%u max_reqs=%u\n",
8848                 __func__,
8849                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
8850                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
8851
8852         /* Back channel attributes */
8853         args->bc_attrs.max_rqst_sz = max_bc_payload;
8854         args->bc_attrs.max_resp_sz = max_bc_payload;
8855         args->bc_attrs.max_resp_sz_cached = 0;
8856         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
8857         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
8858         if (args->bc_attrs.max_reqs > max_bc_slots)
8859                 args->bc_attrs.max_reqs = max_bc_slots;
8860
8861         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
8862                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
8863                 __func__,
8864                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
8865                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
8866                 args->bc_attrs.max_reqs);
8867 }
8868
8869 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
8870                 struct nfs41_create_session_res *res)
8871 {
8872         struct nfs4_channel_attrs *sent = &args->fc_attrs;
8873         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
8874
8875         if (rcvd->max_resp_sz > sent->max_resp_sz)
8876                 return -EINVAL;
8877         /*
8878          * Our requested max_ops is the minimum we need; we're not
8879          * prepared to break up compounds into smaller pieces than that.
8880          * So, no point even trying to continue if the server won't
8881          * cooperate:
8882          */
8883         if (rcvd->max_ops < sent->max_ops)
8884                 return -EINVAL;
8885         if (rcvd->max_reqs == 0)
8886                 return -EINVAL;
8887         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
8888                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
8889         return 0;
8890 }
8891
8892 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
8893                 struct nfs41_create_session_res *res)
8894 {
8895         struct nfs4_channel_attrs *sent = &args->bc_attrs;
8896         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
8897
8898         if (!(res->flags & SESSION4_BACK_CHAN))
8899                 goto out;
8900         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
8901                 return -EINVAL;
8902         if (rcvd->max_resp_sz < sent->max_resp_sz)
8903                 return -EINVAL;
8904         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
8905                 return -EINVAL;
8906         if (rcvd->max_ops > sent->max_ops)
8907                 return -EINVAL;
8908         if (rcvd->max_reqs > sent->max_reqs)
8909                 return -EINVAL;
8910 out:
8911         return 0;
8912 }
8913
8914 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
8915                                      struct nfs41_create_session_res *res)
8916 {
8917         int ret;
8918
8919         ret = nfs4_verify_fore_channel_attrs(args, res);
8920         if (ret)
8921                 return ret;
8922         return nfs4_verify_back_channel_attrs(args, res);
8923 }
8924
8925 static void nfs4_update_session(struct nfs4_session *session,
8926                 struct nfs41_create_session_res *res)
8927 {
8928         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
8929         /* Mark client id and session as being confirmed */
8930         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
8931         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
8932         session->flags = res->flags;
8933         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
8934         if (res->flags & SESSION4_BACK_CHAN)
8935                 memcpy(&session->bc_attrs, &res->bc_attrs,
8936                                 sizeof(session->bc_attrs));
8937 }
8938
8939 static int _nfs4_proc_create_session(struct nfs_client *clp,
8940                 const struct cred *cred)
8941 {
8942         struct nfs4_session *session = clp->cl_session;
8943         struct nfs41_create_session_args args = {
8944                 .client = clp,
8945                 .clientid = clp->cl_clientid,
8946                 .seqid = clp->cl_seqid,
8947                 .cb_program = NFS4_CALLBACK,
8948         };
8949         struct nfs41_create_session_res res;
8950
8951         struct rpc_message msg = {
8952                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
8953                 .rpc_argp = &args,
8954                 .rpc_resp = &res,
8955                 .rpc_cred = cred,
8956         };
8957         int status;
8958
8959         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
8960         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
8961
8962         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
8963                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8964         trace_nfs4_create_session(clp, status);
8965
8966         switch (status) {
8967         case -NFS4ERR_STALE_CLIENTID:
8968         case -NFS4ERR_DELAY:
8969         case -ETIMEDOUT:
8970         case -EACCES:
8971         case -EAGAIN:
8972                 goto out;
8973         }
8974
8975         clp->cl_seqid++;
8976         if (!status) {
8977                 /* Verify the session's negotiated channel_attrs values */
8978                 status = nfs4_verify_channel_attrs(&args, &res);
8979                 /* Increment the clientid slot sequence id */
8980                 if (status)
8981                         goto out;
8982                 nfs4_update_session(session, &res);
8983         }
8984 out:
8985         return status;
8986 }
8987
8988 /*
8989  * Issues a CREATE_SESSION operation to the server.
8990  * It is the responsibility of the caller to verify the session is
8991  * expired before calling this routine.
8992  */
8993 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
8994 {
8995         int status;
8996         unsigned *ptr;
8997         struct nfs4_session *session = clp->cl_session;
8998
8999         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9000
9001         status = _nfs4_proc_create_session(clp, cred);
9002         if (status)
9003                 goto out;
9004
9005         /* Init or reset the session slot tables */
9006         status = nfs4_setup_session_slot_tables(session);
9007         dprintk("slot table setup returned %d\n", status);
9008         if (status)
9009                 goto out;
9010
9011         ptr = (unsigned *)&session->sess_id.data[0];
9012         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9013                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9014 out:
9015         dprintk("<-- %s\n", __func__);
9016         return status;
9017 }
9018
9019 /*
9020  * Issue the over-the-wire RPC DESTROY_SESSION.
9021  * The caller must serialize access to this routine.
9022  */
9023 int nfs4_proc_destroy_session(struct nfs4_session *session,
9024                 const struct cred *cred)
9025 {
9026         struct rpc_message msg = {
9027                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9028                 .rpc_argp = session,
9029                 .rpc_cred = cred,
9030         };
9031         int status = 0;
9032
9033         dprintk("--> nfs4_proc_destroy_session\n");
9034
9035         /* session is still being setup */
9036         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9037                 return 0;
9038
9039         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9040                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9041         trace_nfs4_destroy_session(session->clp, status);
9042
9043         if (status)
9044                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9045                         "Session has been destroyed regardless...\n", status);
9046
9047         dprintk("<-- nfs4_proc_destroy_session\n");
9048         return status;
9049 }
9050
9051 /*
9052  * Renew the cl_session lease.
9053  */
9054 struct nfs4_sequence_data {
9055         struct nfs_client *clp;
9056         struct nfs4_sequence_args args;
9057         struct nfs4_sequence_res res;
9058 };
9059
9060 static void nfs41_sequence_release(void *data)
9061 {
9062         struct nfs4_sequence_data *calldata = data;
9063         struct nfs_client *clp = calldata->clp;
9064
9065         if (refcount_read(&clp->cl_count) > 1)
9066                 nfs4_schedule_state_renewal(clp);
9067         nfs_put_client(clp);
9068         kfree(calldata);
9069 }
9070
9071 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9072 {
9073         switch(task->tk_status) {
9074         case -NFS4ERR_DELAY:
9075                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9076                 return -EAGAIN;
9077         default:
9078                 nfs4_schedule_lease_recovery(clp);
9079         }
9080         return 0;
9081 }
9082
9083 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9084 {
9085         struct nfs4_sequence_data *calldata = data;
9086         struct nfs_client *clp = calldata->clp;
9087
9088         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9089                 return;
9090
9091         trace_nfs4_sequence(clp, task->tk_status);
9092         if (task->tk_status < 0) {
9093                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9094                 if (refcount_read(&clp->cl_count) == 1)
9095                         goto out;
9096
9097                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9098                         rpc_restart_call_prepare(task);
9099                         return;
9100                 }
9101         }
9102         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9103 out:
9104         dprintk("<-- %s\n", __func__);
9105 }
9106
9107 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9108 {
9109         struct nfs4_sequence_data *calldata = data;
9110         struct nfs_client *clp = calldata->clp;
9111         struct nfs4_sequence_args *args;
9112         struct nfs4_sequence_res *res;
9113
9114         args = task->tk_msg.rpc_argp;
9115         res = task->tk_msg.rpc_resp;
9116
9117         nfs4_setup_sequence(clp, args, res, task);
9118 }
9119
9120 static const struct rpc_call_ops nfs41_sequence_ops = {
9121         .rpc_call_done = nfs41_sequence_call_done,
9122         .rpc_call_prepare = nfs41_sequence_prepare,
9123         .rpc_release = nfs41_sequence_release,
9124 };
9125
9126 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9127                 const struct cred *cred,
9128                 struct nfs4_slot *slot,
9129                 bool is_privileged)
9130 {
9131         struct nfs4_sequence_data *calldata;
9132         struct rpc_message msg = {
9133                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9134                 .rpc_cred = cred,
9135         };
9136         struct rpc_task_setup task_setup_data = {
9137                 .rpc_client = clp->cl_rpcclient,
9138                 .rpc_message = &msg,
9139                 .callback_ops = &nfs41_sequence_ops,
9140                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
9141         };
9142         struct rpc_task *ret;
9143
9144         ret = ERR_PTR(-EIO);
9145         if (!refcount_inc_not_zero(&clp->cl_count))
9146                 goto out_err;
9147
9148         ret = ERR_PTR(-ENOMEM);
9149         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9150         if (calldata == NULL)
9151                 goto out_put_clp;
9152         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9153         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9154         msg.rpc_argp = &calldata->args;
9155         msg.rpc_resp = &calldata->res;
9156         calldata->clp = clp;
9157         task_setup_data.callback_data = calldata;
9158
9159         ret = rpc_run_task(&task_setup_data);
9160         if (IS_ERR(ret))
9161                 goto out_err;
9162         return ret;
9163 out_put_clp:
9164         nfs_put_client(clp);
9165 out_err:
9166         nfs41_release_slot(slot);
9167         return ret;
9168 }
9169
9170 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9171 {
9172         struct rpc_task *task;
9173         int ret = 0;
9174
9175         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9176                 return -EAGAIN;
9177         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9178         if (IS_ERR(task))
9179                 ret = PTR_ERR(task);
9180         else
9181                 rpc_put_task_async(task);
9182         dprintk("<-- %s status=%d\n", __func__, ret);
9183         return ret;
9184 }
9185
9186 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9187 {
9188         struct rpc_task *task;
9189         int ret;
9190
9191         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9192         if (IS_ERR(task)) {
9193                 ret = PTR_ERR(task);
9194                 goto out;
9195         }
9196         ret = rpc_wait_for_completion_task(task);
9197         if (!ret)
9198                 ret = task->tk_status;
9199         rpc_put_task(task);
9200 out:
9201         dprintk("<-- %s status=%d\n", __func__, ret);
9202         return ret;
9203 }
9204
9205 struct nfs4_reclaim_complete_data {
9206         struct nfs_client *clp;
9207         struct nfs41_reclaim_complete_args arg;
9208         struct nfs41_reclaim_complete_res res;
9209 };
9210
9211 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9212 {
9213         struct nfs4_reclaim_complete_data *calldata = data;
9214
9215         nfs4_setup_sequence(calldata->clp,
9216                         &calldata->arg.seq_args,
9217                         &calldata->res.seq_res,
9218                         task);
9219 }
9220
9221 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9222 {
9223         switch(task->tk_status) {
9224         case 0:
9225                 wake_up_all(&clp->cl_lock_waitq);
9226                 fallthrough;
9227         case -NFS4ERR_COMPLETE_ALREADY:
9228         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9229                 break;
9230         case -NFS4ERR_DELAY:
9231                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9232                 fallthrough;
9233         case -NFS4ERR_RETRY_UNCACHED_REP:
9234                 return -EAGAIN;
9235         case -NFS4ERR_BADSESSION:
9236         case -NFS4ERR_DEADSESSION:
9237         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9238                 break;
9239         default:
9240                 nfs4_schedule_lease_recovery(clp);
9241         }
9242         return 0;
9243 }
9244
9245 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9246 {
9247         struct nfs4_reclaim_complete_data *calldata = data;
9248         struct nfs_client *clp = calldata->clp;
9249         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9250
9251         dprintk("--> %s\n", __func__);
9252         if (!nfs41_sequence_done(task, res))
9253                 return;
9254
9255         trace_nfs4_reclaim_complete(clp, task->tk_status);
9256         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9257                 rpc_restart_call_prepare(task);
9258                 return;
9259         }
9260         dprintk("<-- %s\n", __func__);
9261 }
9262
9263 static void nfs4_free_reclaim_complete_data(void *data)
9264 {
9265         struct nfs4_reclaim_complete_data *calldata = data;
9266
9267         kfree(calldata);
9268 }
9269
9270 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9271         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9272         .rpc_call_done = nfs4_reclaim_complete_done,
9273         .rpc_release = nfs4_free_reclaim_complete_data,
9274 };
9275
9276 /*
9277  * Issue a global reclaim complete.
9278  */
9279 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9280                 const struct cred *cred)
9281 {
9282         struct nfs4_reclaim_complete_data *calldata;
9283         struct rpc_message msg = {
9284                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9285                 .rpc_cred = cred,
9286         };
9287         struct rpc_task_setup task_setup_data = {
9288                 .rpc_client = clp->cl_rpcclient,
9289                 .rpc_message = &msg,
9290                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9291                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9292         };
9293         int status = -ENOMEM;
9294
9295         dprintk("--> %s\n", __func__);
9296         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9297         if (calldata == NULL)
9298                 goto out;
9299         calldata->clp = clp;
9300         calldata->arg.one_fs = 0;
9301
9302         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9303         msg.rpc_argp = &calldata->arg;
9304         msg.rpc_resp = &calldata->res;
9305         task_setup_data.callback_data = calldata;
9306         status = nfs4_call_sync_custom(&task_setup_data);
9307 out:
9308         dprintk("<-- %s status=%d\n", __func__, status);
9309         return status;
9310 }
9311
9312 static void
9313 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9314 {
9315         struct nfs4_layoutget *lgp = calldata;
9316         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9317
9318         dprintk("--> %s\n", __func__);
9319         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9320                                 &lgp->res.seq_res, task);
9321         dprintk("<-- %s\n", __func__);
9322 }
9323
9324 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9325 {
9326         struct nfs4_layoutget *lgp = calldata;
9327
9328         dprintk("--> %s\n", __func__);
9329         nfs41_sequence_process(task, &lgp->res.seq_res);
9330         dprintk("<-- %s\n", __func__);
9331 }
9332
9333 static int
9334 nfs4_layoutget_handle_exception(struct rpc_task *task,
9335                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9336 {
9337         struct inode *inode = lgp->args.inode;
9338         struct nfs_server *server = NFS_SERVER(inode);
9339         struct pnfs_layout_hdr *lo;
9340         int nfs4err = task->tk_status;
9341         int err, status = 0;
9342         LIST_HEAD(head);
9343
9344         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9345
9346         nfs4_sequence_free_slot(&lgp->res.seq_res);
9347
9348         switch (nfs4err) {
9349         case 0:
9350                 goto out;
9351
9352         /*
9353          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9354          * on the file. set tk_status to -ENODATA to tell upper layer to
9355          * retry go inband.
9356          */
9357         case -NFS4ERR_LAYOUTUNAVAILABLE:
9358                 status = -ENODATA;
9359                 goto out;
9360         /*
9361          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9362          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9363          */
9364         case -NFS4ERR_BADLAYOUT:
9365                 status = -EOVERFLOW;
9366                 goto out;
9367         /*
9368          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9369          * (or clients) writing to the same RAID stripe except when
9370          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9371          *
9372          * Treat it like we would RECALLCONFLICT -- we retry for a little
9373          * while, and then eventually give up.
9374          */
9375         case -NFS4ERR_LAYOUTTRYLATER:
9376                 if (lgp->args.minlength == 0) {
9377                         status = -EOVERFLOW;
9378                         goto out;
9379                 }
9380                 status = -EBUSY;
9381                 break;
9382         case -NFS4ERR_RECALLCONFLICT:
9383                 status = -ERECALLCONFLICT;
9384                 break;
9385         case -NFS4ERR_DELEG_REVOKED:
9386         case -NFS4ERR_ADMIN_REVOKED:
9387         case -NFS4ERR_EXPIRED:
9388         case -NFS4ERR_BAD_STATEID:
9389                 exception->timeout = 0;
9390                 spin_lock(&inode->i_lock);
9391                 lo = NFS_I(inode)->layout;
9392                 /* If the open stateid was bad, then recover it. */
9393                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9394                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9395                         spin_unlock(&inode->i_lock);
9396                         exception->state = lgp->args.ctx->state;
9397                         exception->stateid = &lgp->args.stateid;
9398                         break;
9399                 }
9400
9401                 /*
9402                  * Mark the bad layout state as invalid, then retry
9403                  */
9404                 pnfs_mark_layout_stateid_invalid(lo, &head);
9405                 spin_unlock(&inode->i_lock);
9406                 nfs_commit_inode(inode, 0);
9407                 pnfs_free_lseg_list(&head);
9408                 status = -EAGAIN;
9409                 goto out;
9410         }
9411
9412         err = nfs4_handle_exception(server, nfs4err, exception);
9413         if (!status) {
9414                 if (exception->retry)
9415                         status = -EAGAIN;
9416                 else
9417                         status = err;
9418         }
9419 out:
9420         dprintk("<-- %s\n", __func__);
9421         return status;
9422 }
9423
9424 size_t max_response_pages(struct nfs_server *server)
9425 {
9426         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9427         return nfs_page_array_len(0, max_resp_sz);
9428 }
9429
9430 static void nfs4_layoutget_release(void *calldata)
9431 {
9432         struct nfs4_layoutget *lgp = calldata;
9433
9434         dprintk("--> %s\n", __func__);
9435         nfs4_sequence_free_slot(&lgp->res.seq_res);
9436         pnfs_layoutget_free(lgp);
9437         dprintk("<-- %s\n", __func__);
9438 }
9439
9440 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9441         .rpc_call_prepare = nfs4_layoutget_prepare,
9442         .rpc_call_done = nfs4_layoutget_done,
9443         .rpc_release = nfs4_layoutget_release,
9444 };
9445
9446 struct pnfs_layout_segment *
9447 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9448 {
9449         struct inode *inode = lgp->args.inode;
9450         struct nfs_server *server = NFS_SERVER(inode);
9451         struct rpc_task *task;
9452         struct rpc_message msg = {
9453                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9454                 .rpc_argp = &lgp->args,
9455                 .rpc_resp = &lgp->res,
9456                 .rpc_cred = lgp->cred,
9457         };
9458         struct rpc_task_setup task_setup_data = {
9459                 .rpc_client = server->client,
9460                 .rpc_message = &msg,
9461                 .callback_ops = &nfs4_layoutget_call_ops,
9462                 .callback_data = lgp,
9463                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
9464         };
9465         struct pnfs_layout_segment *lseg = NULL;
9466         struct nfs4_exception exception = {
9467                 .inode = inode,
9468                 .timeout = *timeout,
9469         };
9470         int status = 0;
9471
9472         dprintk("--> %s\n", __func__);
9473
9474         /* nfs4_layoutget_release calls pnfs_put_layout_hdr */
9475         pnfs_get_layout_hdr(NFS_I(inode)->layout);
9476
9477         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9478
9479         task = rpc_run_task(&task_setup_data);
9480
9481         status = rpc_wait_for_completion_task(task);
9482         if (status != 0)
9483                 goto out;
9484
9485         if (task->tk_status < 0) {
9486                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9487                 *timeout = exception.timeout;
9488         } else if (lgp->res.layoutp->len == 0) {
9489                 status = -EAGAIN;
9490                 *timeout = nfs4_update_delay(&exception.timeout);
9491         } else
9492                 lseg = pnfs_layout_process(lgp);
9493 out:
9494         trace_nfs4_layoutget(lgp->args.ctx,
9495                         &lgp->args.range,
9496                         &lgp->res.range,
9497                         &lgp->res.stateid,
9498                         status);
9499
9500         rpc_put_task(task);
9501         dprintk("<-- %s status=%d\n", __func__, status);
9502         if (status)
9503                 return ERR_PTR(status);
9504         return lseg;
9505 }
9506
9507 static void
9508 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9509 {
9510         struct nfs4_layoutreturn *lrp = calldata;
9511
9512         dprintk("--> %s\n", __func__);
9513         nfs4_setup_sequence(lrp->clp,
9514                         &lrp->args.seq_args,
9515                         &lrp->res.seq_res,
9516                         task);
9517         if (!pnfs_layout_is_valid(lrp->args.layout))
9518                 rpc_exit(task, 0);
9519 }
9520
9521 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9522 {
9523         struct nfs4_layoutreturn *lrp = calldata;
9524         struct nfs_server *server;
9525
9526         dprintk("--> %s\n", __func__);
9527
9528         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9529                 return;
9530
9531         /*
9532          * Was there an RPC level error? Assume the call succeeded,
9533          * and that we need to release the layout
9534          */
9535         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9536                 lrp->res.lrs_present = 0;
9537                 return;
9538         }
9539
9540         server = NFS_SERVER(lrp->args.inode);
9541         switch (task->tk_status) {
9542         case -NFS4ERR_OLD_STATEID:
9543                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9544                                         &lrp->args.range,
9545                                         lrp->args.inode))
9546                         goto out_restart;
9547                 fallthrough;
9548         default:
9549                 task->tk_status = 0;
9550                 fallthrough;
9551         case 0:
9552                 break;
9553         case -NFS4ERR_DELAY:
9554                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9555                         break;
9556                 goto out_restart;
9557         }
9558         dprintk("<-- %s\n", __func__);
9559         return;
9560 out_restart:
9561         task->tk_status = 0;
9562         nfs4_sequence_free_slot(&lrp->res.seq_res);
9563         rpc_restart_call_prepare(task);
9564 }
9565
9566 static void nfs4_layoutreturn_release(void *calldata)
9567 {
9568         struct nfs4_layoutreturn *lrp = calldata;
9569         struct pnfs_layout_hdr *lo = lrp->args.layout;
9570
9571         dprintk("--> %s\n", __func__);
9572         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9573                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9574         nfs4_sequence_free_slot(&lrp->res.seq_res);
9575         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9576                 lrp->ld_private.ops->free(&lrp->ld_private);
9577         pnfs_put_layout_hdr(lrp->args.layout);
9578         nfs_iput_and_deactive(lrp->inode);
9579         put_cred(lrp->cred);
9580         kfree(calldata);
9581         dprintk("<-- %s\n", __func__);
9582 }
9583
9584 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9585         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9586         .rpc_call_done = nfs4_layoutreturn_done,
9587         .rpc_release = nfs4_layoutreturn_release,
9588 };
9589
9590 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9591 {
9592         struct rpc_task *task;
9593         struct rpc_message msg = {
9594                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9595                 .rpc_argp = &lrp->args,
9596                 .rpc_resp = &lrp->res,
9597                 .rpc_cred = lrp->cred,
9598         };
9599         struct rpc_task_setup task_setup_data = {
9600                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9601                 .rpc_message = &msg,
9602                 .callback_ops = &nfs4_layoutreturn_call_ops,
9603                 .callback_data = lrp,
9604         };
9605         int status = 0;
9606
9607         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9608                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9609                         &task_setup_data.rpc_client, &msg);
9610
9611         dprintk("--> %s\n", __func__);
9612         if (!sync) {
9613                 lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9614                 if (!lrp->inode) {
9615                         nfs4_layoutreturn_release(lrp);
9616                         return -EAGAIN;
9617                 }
9618                 task_setup_data.flags |= RPC_TASK_ASYNC;
9619         }
9620         nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1, 0);
9621         task = rpc_run_task(&task_setup_data);
9622         if (IS_ERR(task))
9623                 return PTR_ERR(task);
9624         if (sync)
9625                 status = task->tk_status;
9626         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9627         dprintk("<-- %s status=%d\n", __func__, status);
9628         rpc_put_task(task);
9629         return status;
9630 }
9631
9632 static int
9633 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9634                 struct pnfs_device *pdev,
9635                 const struct cred *cred)
9636 {
9637         struct nfs4_getdeviceinfo_args args = {
9638                 .pdev = pdev,
9639                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9640                         NOTIFY_DEVICEID4_DELETE,
9641         };
9642         struct nfs4_getdeviceinfo_res res = {
9643                 .pdev = pdev,
9644         };
9645         struct rpc_message msg = {
9646                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9647                 .rpc_argp = &args,
9648                 .rpc_resp = &res,
9649                 .rpc_cred = cred,
9650         };
9651         int status;
9652
9653         dprintk("--> %s\n", __func__);
9654         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9655         if (res.notification & ~args.notify_types)
9656                 dprintk("%s: unsupported notification\n", __func__);
9657         if (res.notification != args.notify_types)
9658                 pdev->nocache = 1;
9659
9660         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9661
9662         dprintk("<-- %s status=%d\n", __func__, status);
9663
9664         return status;
9665 }
9666
9667 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9668                 struct pnfs_device *pdev,
9669                 const struct cred *cred)
9670 {
9671         struct nfs4_exception exception = { };
9672         int err;
9673
9674         do {
9675                 err = nfs4_handle_exception(server,
9676                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9677                                         &exception);
9678         } while (exception.retry);
9679         return err;
9680 }
9681 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9682
9683 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9684 {
9685         struct nfs4_layoutcommit_data *data = calldata;
9686         struct nfs_server *server = NFS_SERVER(data->args.inode);
9687
9688         nfs4_setup_sequence(server->nfs_client,
9689                         &data->args.seq_args,
9690                         &data->res.seq_res,
9691                         task);
9692 }
9693
9694 static void
9695 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9696 {
9697         struct nfs4_layoutcommit_data *data = calldata;
9698         struct nfs_server *server = NFS_SERVER(data->args.inode);
9699
9700         if (!nfs41_sequence_done(task, &data->res.seq_res))
9701                 return;
9702
9703         switch (task->tk_status) { /* Just ignore these failures */
9704         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9705         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9706         case -NFS4ERR_BADLAYOUT:     /* no layout */
9707         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9708                 task->tk_status = 0;
9709                 break;
9710         case 0:
9711                 break;
9712         default:
9713                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9714                         rpc_restart_call_prepare(task);
9715                         return;
9716                 }
9717         }
9718 }
9719
9720 static void nfs4_layoutcommit_release(void *calldata)
9721 {
9722         struct nfs4_layoutcommit_data *data = calldata;
9723
9724         pnfs_cleanup_layoutcommit(data);
9725         nfs_post_op_update_inode_force_wcc(data->args.inode,
9726                                            data->res.fattr);
9727         put_cred(data->cred);
9728         nfs_iput_and_deactive(data->inode);
9729         kfree(data);
9730 }
9731
9732 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9733         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9734         .rpc_call_done = nfs4_layoutcommit_done,
9735         .rpc_release = nfs4_layoutcommit_release,
9736 };
9737
9738 int
9739 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9740 {
9741         struct rpc_message msg = {
9742                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9743                 .rpc_argp = &data->args,
9744                 .rpc_resp = &data->res,
9745                 .rpc_cred = data->cred,
9746         };
9747         struct rpc_task_setup task_setup_data = {
9748                 .task = &data->task,
9749                 .rpc_client = NFS_CLIENT(data->args.inode),
9750                 .rpc_message = &msg,
9751                 .callback_ops = &nfs4_layoutcommit_ops,
9752                 .callback_data = data,
9753         };
9754         struct rpc_task *task;
9755         int status = 0;
9756
9757         dprintk("NFS: initiating layoutcommit call. sync %d "
9758                 "lbw: %llu inode %lu\n", sync,
9759                 data->args.lastbytewritten,
9760                 data->args.inode->i_ino);
9761
9762         if (!sync) {
9763                 data->inode = nfs_igrab_and_active(data->args.inode);
9764                 if (data->inode == NULL) {
9765                         nfs4_layoutcommit_release(data);
9766                         return -EAGAIN;
9767                 }
9768                 task_setup_data.flags = RPC_TASK_ASYNC;
9769         }
9770         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9771         task = rpc_run_task(&task_setup_data);
9772         if (IS_ERR(task))
9773                 return PTR_ERR(task);
9774         if (sync)
9775                 status = task->tk_status;
9776         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9777         dprintk("%s: status %d\n", __func__, status);
9778         rpc_put_task(task);
9779         return status;
9780 }
9781
9782 /*
9783  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9784  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9785  */
9786 static int
9787 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9788                     struct nfs_fsinfo *info,
9789                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9790 {
9791         struct nfs41_secinfo_no_name_args args = {
9792                 .style = SECINFO_STYLE_CURRENT_FH,
9793         };
9794         struct nfs4_secinfo_res res = {
9795                 .flavors = flavors,
9796         };
9797         struct rpc_message msg = {
9798                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9799                 .rpc_argp = &args,
9800                 .rpc_resp = &res,
9801         };
9802         struct nfs4_call_sync_data data = {
9803                 .seq_server = server,
9804                 .seq_args = &args.seq_args,
9805                 .seq_res = &res.seq_res,
9806         };
9807         struct rpc_task_setup task_setup = {
9808                 .rpc_client = server->client,
9809                 .rpc_message = &msg,
9810                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9811                 .callback_data = &data,
9812                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9813         };
9814         const struct cred *cred = NULL;
9815         int status;
9816
9817         if (use_integrity) {
9818                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9819
9820                 cred = nfs4_get_clid_cred(server->nfs_client);
9821                 msg.rpc_cred = cred;
9822         }
9823
9824         dprintk("--> %s\n", __func__);
9825         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9826         status = nfs4_call_sync_custom(&task_setup);
9827         dprintk("<-- %s status=%d\n", __func__, status);
9828
9829         put_cred(cred);
9830
9831         return status;
9832 }
9833
9834 static int
9835 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9836                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9837 {
9838         struct nfs4_exception exception = {
9839                 .interruptible = true,
9840         };
9841         int err;
9842         do {
9843                 /* first try using integrity protection */
9844                 err = -NFS4ERR_WRONGSEC;
9845
9846                 /* try to use integrity protection with machine cred */
9847                 if (_nfs4_is_integrity_protected(server->nfs_client))
9848                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9849                                                           flavors, true);
9850
9851                 /*
9852                  * if unable to use integrity protection, or SECINFO with
9853                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9854                  * disallowed by spec, but exists in deployed servers) use
9855                  * the current filesystem's rpc_client and the user cred.
9856                  */
9857                 if (err == -NFS4ERR_WRONGSEC)
9858                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9859                                                           flavors, false);
9860
9861                 switch (err) {
9862                 case 0:
9863                 case -NFS4ERR_WRONGSEC:
9864                 case -ENOTSUPP:
9865                         goto out;
9866                 default:
9867                         err = nfs4_handle_exception(server, err, &exception);
9868                 }
9869         } while (exception.retry);
9870 out:
9871         return err;
9872 }
9873
9874 static int
9875 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
9876                     struct nfs_fsinfo *info)
9877 {
9878         int err;
9879         struct page *page;
9880         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
9881         struct nfs4_secinfo_flavors *flavors;
9882         struct nfs4_secinfo4 *secinfo;
9883         int i;
9884
9885         page = alloc_page(GFP_KERNEL);
9886         if (!page) {
9887                 err = -ENOMEM;
9888                 goto out;
9889         }
9890
9891         flavors = page_address(page);
9892         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
9893
9894         /*
9895          * Fall back on "guess and check" method if
9896          * the server doesn't support SECINFO_NO_NAME
9897          */
9898         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
9899                 err = nfs4_find_root_sec(server, fhandle, info);
9900                 goto out_freepage;
9901         }
9902         if (err)
9903                 goto out_freepage;
9904
9905         for (i = 0; i < flavors->num_flavors; i++) {
9906                 secinfo = &flavors->flavors[i];
9907
9908                 switch (secinfo->flavor) {
9909                 case RPC_AUTH_NULL:
9910                 case RPC_AUTH_UNIX:
9911                 case RPC_AUTH_GSS:
9912                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
9913                                         &secinfo->flavor_info);
9914                         break;
9915                 default:
9916                         flavor = RPC_AUTH_MAXFLAVOR;
9917                         break;
9918                 }
9919
9920                 if (!nfs_auth_info_match(&server->auth_info, flavor))
9921                         flavor = RPC_AUTH_MAXFLAVOR;
9922
9923                 if (flavor != RPC_AUTH_MAXFLAVOR) {
9924                         err = nfs4_lookup_root_sec(server, fhandle,
9925                                                    info, flavor);
9926                         if (!err)
9927                                 break;
9928                 }
9929         }
9930
9931         if (flavor == RPC_AUTH_MAXFLAVOR)
9932                 err = -EPERM;
9933
9934 out_freepage:
9935         put_page(page);
9936         if (err == -EACCES)
9937                 return -EPERM;
9938 out:
9939         return err;
9940 }
9941
9942 static int _nfs41_test_stateid(struct nfs_server *server,
9943                 nfs4_stateid *stateid,
9944                 const struct cred *cred)
9945 {
9946         int status;
9947         struct nfs41_test_stateid_args args = {
9948                 .stateid = stateid,
9949         };
9950         struct nfs41_test_stateid_res res;
9951         struct rpc_message msg = {
9952                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
9953                 .rpc_argp = &args,
9954                 .rpc_resp = &res,
9955                 .rpc_cred = cred,
9956         };
9957         struct rpc_clnt *rpc_client = server->client;
9958
9959         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
9960                 &rpc_client, &msg);
9961
9962         dprintk("NFS call  test_stateid %p\n", stateid);
9963         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
9964         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
9965                         &args.seq_args, &res.seq_res);
9966         if (status != NFS_OK) {
9967                 dprintk("NFS reply test_stateid: failed, %d\n", status);
9968                 return status;
9969         }
9970         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
9971         return -res.status;
9972 }
9973
9974 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
9975                 int err, struct nfs4_exception *exception)
9976 {
9977         exception->retry = 0;
9978         switch(err) {
9979         case -NFS4ERR_DELAY:
9980         case -NFS4ERR_RETRY_UNCACHED_REP:
9981                 nfs4_handle_exception(server, err, exception);
9982                 break;
9983         case -NFS4ERR_BADSESSION:
9984         case -NFS4ERR_BADSLOT:
9985         case -NFS4ERR_BAD_HIGH_SLOT:
9986         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9987         case -NFS4ERR_DEADSESSION:
9988                 nfs4_do_handle_exception(server, err, exception);
9989         }
9990 }
9991
9992 /**
9993  * nfs41_test_stateid - perform a TEST_STATEID operation
9994  *
9995  * @server: server / transport on which to perform the operation
9996  * @stateid: state ID to test
9997  * @cred: credential
9998  *
9999  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10000  * Otherwise a negative NFS4ERR value is returned if the operation
10001  * failed or the state ID is not currently valid.
10002  */
10003 static int nfs41_test_stateid(struct nfs_server *server,
10004                 nfs4_stateid *stateid,
10005                 const struct cred *cred)
10006 {
10007         struct nfs4_exception exception = {
10008                 .interruptible = true,
10009         };
10010         int err;
10011         do {
10012                 err = _nfs41_test_stateid(server, stateid, cred);
10013                 nfs4_handle_delay_or_session_error(server, err, &exception);
10014         } while (exception.retry);
10015         return err;
10016 }
10017
10018 struct nfs_free_stateid_data {
10019         struct nfs_server *server;
10020         struct nfs41_free_stateid_args args;
10021         struct nfs41_free_stateid_res res;
10022 };
10023
10024 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10025 {
10026         struct nfs_free_stateid_data *data = calldata;
10027         nfs4_setup_sequence(data->server->nfs_client,
10028                         &data->args.seq_args,
10029                         &data->res.seq_res,
10030                         task);
10031 }
10032
10033 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10034 {
10035         struct nfs_free_stateid_data *data = calldata;
10036
10037         nfs41_sequence_done(task, &data->res.seq_res);
10038
10039         switch (task->tk_status) {
10040         case -NFS4ERR_DELAY:
10041                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10042                         rpc_restart_call_prepare(task);
10043         }
10044 }
10045
10046 static void nfs41_free_stateid_release(void *calldata)
10047 {
10048         kfree(calldata);
10049 }
10050
10051 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10052         .rpc_call_prepare = nfs41_free_stateid_prepare,
10053         .rpc_call_done = nfs41_free_stateid_done,
10054         .rpc_release = nfs41_free_stateid_release,
10055 };
10056
10057 /**
10058  * nfs41_free_stateid - perform a FREE_STATEID operation
10059  *
10060  * @server: server / transport on which to perform the operation
10061  * @stateid: state ID to release
10062  * @cred: credential
10063  * @privileged: set to true if this call needs to be privileged
10064  *
10065  * Note: this function is always asynchronous.
10066  */
10067 static int nfs41_free_stateid(struct nfs_server *server,
10068                 const nfs4_stateid *stateid,
10069                 const struct cred *cred,
10070                 bool privileged)
10071 {
10072         struct rpc_message msg = {
10073                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10074                 .rpc_cred = cred,
10075         };
10076         struct rpc_task_setup task_setup = {
10077                 .rpc_client = server->client,
10078                 .rpc_message = &msg,
10079                 .callback_ops = &nfs41_free_stateid_ops,
10080                 .flags = RPC_TASK_ASYNC,
10081         };
10082         struct nfs_free_stateid_data *data;
10083         struct rpc_task *task;
10084
10085         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10086                 &task_setup.rpc_client, &msg);
10087
10088         dprintk("NFS call  free_stateid %p\n", stateid);
10089         data = kmalloc(sizeof(*data), GFP_NOFS);
10090         if (!data)
10091                 return -ENOMEM;
10092         data->server = server;
10093         nfs4_stateid_copy(&data->args.stateid, stateid);
10094
10095         task_setup.callback_data = data;
10096
10097         msg.rpc_argp = &data->args;
10098         msg.rpc_resp = &data->res;
10099         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10100         task = rpc_run_task(&task_setup);
10101         if (IS_ERR(task))
10102                 return PTR_ERR(task);
10103         rpc_put_task(task);
10104         return 0;
10105 }
10106
10107 static void
10108 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10109 {
10110         const struct cred *cred = lsp->ls_state->owner->so_cred;
10111
10112         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10113         nfs4_free_lock_state(server, lsp);
10114 }
10115
10116 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10117                 const nfs4_stateid *s2)
10118 {
10119         if (s1->type != s2->type)
10120                 return false;
10121
10122         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10123                 return false;
10124
10125         if (s1->seqid == s2->seqid)
10126                 return true;
10127
10128         return s1->seqid == 0 || s2->seqid == 0;
10129 }
10130
10131 #endif /* CONFIG_NFS_V4_1 */
10132
10133 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10134                 const nfs4_stateid *s2)
10135 {
10136         return nfs4_stateid_match(s1, s2);
10137 }
10138
10139
10140 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10141         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10142         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10143         .recover_open   = nfs4_open_reclaim,
10144         .recover_lock   = nfs4_lock_reclaim,
10145         .establish_clid = nfs4_init_clientid,
10146         .detect_trunking = nfs40_discover_server_trunking,
10147 };
10148
10149 #if defined(CONFIG_NFS_V4_1)
10150 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10151         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10152         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10153         .recover_open   = nfs4_open_reclaim,
10154         .recover_lock   = nfs4_lock_reclaim,
10155         .establish_clid = nfs41_init_clientid,
10156         .reclaim_complete = nfs41_proc_reclaim_complete,
10157         .detect_trunking = nfs41_discover_server_trunking,
10158 };
10159 #endif /* CONFIG_NFS_V4_1 */
10160
10161 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10162         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10163         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10164         .recover_open   = nfs40_open_expired,
10165         .recover_lock   = nfs4_lock_expired,
10166         .establish_clid = nfs4_init_clientid,
10167 };
10168
10169 #if defined(CONFIG_NFS_V4_1)
10170 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10171         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10172         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10173         .recover_open   = nfs41_open_expired,
10174         .recover_lock   = nfs41_lock_expired,
10175         .establish_clid = nfs41_init_clientid,
10176 };
10177 #endif /* CONFIG_NFS_V4_1 */
10178
10179 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10180         .sched_state_renewal = nfs4_proc_async_renew,
10181         .get_state_renewal_cred = nfs4_get_renew_cred,
10182         .renew_lease = nfs4_proc_renew,
10183 };
10184
10185 #if defined(CONFIG_NFS_V4_1)
10186 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10187         .sched_state_renewal = nfs41_proc_async_sequence,
10188         .get_state_renewal_cred = nfs4_get_machine_cred,
10189         .renew_lease = nfs4_proc_sequence,
10190 };
10191 #endif
10192
10193 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10194         .get_locations = _nfs40_proc_get_locations,
10195         .fsid_present = _nfs40_proc_fsid_present,
10196 };
10197
10198 #if defined(CONFIG_NFS_V4_1)
10199 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10200         .get_locations = _nfs41_proc_get_locations,
10201         .fsid_present = _nfs41_proc_fsid_present,
10202 };
10203 #endif  /* CONFIG_NFS_V4_1 */
10204
10205 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10206         .minor_version = 0,
10207         .init_caps = NFS_CAP_READDIRPLUS
10208                 | NFS_CAP_ATOMIC_OPEN
10209                 | NFS_CAP_POSIX_LOCK,
10210         .init_client = nfs40_init_client,
10211         .shutdown_client = nfs40_shutdown_client,
10212         .match_stateid = nfs4_match_stateid,
10213         .find_root_sec = nfs4_find_root_sec,
10214         .free_lock_state = nfs4_release_lockowner,
10215         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10216         .alloc_seqid = nfs_alloc_seqid,
10217         .call_sync_ops = &nfs40_call_sync_ops,
10218         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10219         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10220         .state_renewal_ops = &nfs40_state_renewal_ops,
10221         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10222 };
10223
10224 #if defined(CONFIG_NFS_V4_1)
10225 static struct nfs_seqid *
10226 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10227 {
10228         return NULL;
10229 }
10230
10231 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10232         .minor_version = 1,
10233         .init_caps = NFS_CAP_READDIRPLUS
10234                 | NFS_CAP_ATOMIC_OPEN
10235                 | NFS_CAP_POSIX_LOCK
10236                 | NFS_CAP_STATEID_NFSV41
10237                 | NFS_CAP_ATOMIC_OPEN_V1
10238                 | NFS_CAP_LGOPEN,
10239         .init_client = nfs41_init_client,
10240         .shutdown_client = nfs41_shutdown_client,
10241         .match_stateid = nfs41_match_stateid,
10242         .find_root_sec = nfs41_find_root_sec,
10243         .free_lock_state = nfs41_free_lock_state,
10244         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10245         .alloc_seqid = nfs_alloc_no_seqid,
10246         .session_trunk = nfs4_test_session_trunk,
10247         .call_sync_ops = &nfs41_call_sync_ops,
10248         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10249         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10250         .state_renewal_ops = &nfs41_state_renewal_ops,
10251         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10252 };
10253 #endif
10254
10255 #if defined(CONFIG_NFS_V4_2)
10256 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10257         .minor_version = 2,
10258         .init_caps = NFS_CAP_READDIRPLUS
10259                 | NFS_CAP_ATOMIC_OPEN
10260                 | NFS_CAP_POSIX_LOCK
10261                 | NFS_CAP_STATEID_NFSV41
10262                 | NFS_CAP_ATOMIC_OPEN_V1
10263                 | NFS_CAP_LGOPEN
10264                 | NFS_CAP_ALLOCATE
10265                 | NFS_CAP_COPY
10266                 | NFS_CAP_OFFLOAD_CANCEL
10267                 | NFS_CAP_COPY_NOTIFY
10268                 | NFS_CAP_DEALLOCATE
10269                 | NFS_CAP_SEEK
10270                 | NFS_CAP_LAYOUTSTATS
10271                 | NFS_CAP_CLONE
10272                 | NFS_CAP_LAYOUTERROR
10273                 | NFS_CAP_READ_PLUS,
10274         .init_client = nfs41_init_client,
10275         .shutdown_client = nfs41_shutdown_client,
10276         .match_stateid = nfs41_match_stateid,
10277         .find_root_sec = nfs41_find_root_sec,
10278         .free_lock_state = nfs41_free_lock_state,
10279         .call_sync_ops = &nfs41_call_sync_ops,
10280         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10281         .alloc_seqid = nfs_alloc_no_seqid,
10282         .session_trunk = nfs4_test_session_trunk,
10283         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10284         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10285         .state_renewal_ops = &nfs41_state_renewal_ops,
10286         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10287 };
10288 #endif
10289
10290 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10291         [0] = &nfs_v4_0_minor_ops,
10292 #if defined(CONFIG_NFS_V4_1)
10293         [1] = &nfs_v4_1_minor_ops,
10294 #endif
10295 #if defined(CONFIG_NFS_V4_2)
10296         [2] = &nfs_v4_2_minor_ops,
10297 #endif
10298 };
10299
10300 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10301 {
10302         ssize_t error, error2, error3;
10303
10304         error = generic_listxattr(dentry, list, size);
10305         if (error < 0)
10306                 return error;
10307         if (list) {
10308                 list += error;
10309                 size -= error;
10310         }
10311
10312         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10313         if (error2 < 0)
10314                 return error2;
10315
10316         if (list) {
10317                 list += error2;
10318                 size -= error2;
10319         }
10320
10321         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10322         if (error3 < 0)
10323                 return error3;
10324
10325         return error + error2 + error3;
10326 }
10327
10328 static const struct inode_operations nfs4_dir_inode_operations = {
10329         .create         = nfs_create,
10330         .lookup         = nfs_lookup,
10331         .atomic_open    = nfs_atomic_open,
10332         .link           = nfs_link,
10333         .unlink         = nfs_unlink,
10334         .symlink        = nfs_symlink,
10335         .mkdir          = nfs_mkdir,
10336         .rmdir          = nfs_rmdir,
10337         .mknod          = nfs_mknod,
10338         .rename         = nfs_rename,
10339         .permission     = nfs_permission,
10340         .getattr        = nfs_getattr,
10341         .setattr        = nfs_setattr,
10342         .listxattr      = nfs4_listxattr,
10343 };
10344
10345 static const struct inode_operations nfs4_file_inode_operations = {
10346         .permission     = nfs_permission,
10347         .getattr        = nfs_getattr,
10348         .setattr        = nfs_setattr,
10349         .listxattr      = nfs4_listxattr,
10350 };
10351
10352 const struct nfs_rpc_ops nfs_v4_clientops = {
10353         .version        = 4,                    /* protocol version */
10354         .dentry_ops     = &nfs4_dentry_operations,
10355         .dir_inode_ops  = &nfs4_dir_inode_operations,
10356         .file_inode_ops = &nfs4_file_inode_operations,
10357         .file_ops       = &nfs4_file_operations,
10358         .getroot        = nfs4_proc_get_root,
10359         .submount       = nfs4_submount,
10360         .try_get_tree   = nfs4_try_get_tree,
10361         .getattr        = nfs4_proc_getattr,
10362         .setattr        = nfs4_proc_setattr,
10363         .lookup         = nfs4_proc_lookup,
10364         .lookupp        = nfs4_proc_lookupp,
10365         .access         = nfs4_proc_access,
10366         .readlink       = nfs4_proc_readlink,
10367         .create         = nfs4_proc_create,
10368         .remove         = nfs4_proc_remove,
10369         .unlink_setup   = nfs4_proc_unlink_setup,
10370         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10371         .unlink_done    = nfs4_proc_unlink_done,
10372         .rename_setup   = nfs4_proc_rename_setup,
10373         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10374         .rename_done    = nfs4_proc_rename_done,
10375         .link           = nfs4_proc_link,
10376         .symlink        = nfs4_proc_symlink,
10377         .mkdir          = nfs4_proc_mkdir,
10378         .rmdir          = nfs4_proc_rmdir,
10379         .readdir        = nfs4_proc_readdir,
10380         .mknod          = nfs4_proc_mknod,
10381         .statfs         = nfs4_proc_statfs,
10382         .fsinfo         = nfs4_proc_fsinfo,
10383         .pathconf       = nfs4_proc_pathconf,
10384         .set_capabilities = nfs4_server_capabilities,
10385         .decode_dirent  = nfs4_decode_dirent,
10386         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10387         .read_setup     = nfs4_proc_read_setup,
10388         .read_done      = nfs4_read_done,
10389         .write_setup    = nfs4_proc_write_setup,
10390         .write_done     = nfs4_write_done,
10391         .commit_setup   = nfs4_proc_commit_setup,
10392         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10393         .commit_done    = nfs4_commit_done,
10394         .lock           = nfs4_proc_lock,
10395         .clear_acl_cache = nfs4_zap_acl_attr,
10396         .close_context  = nfs4_close_context,
10397         .open_context   = nfs4_atomic_open,
10398         .have_delegation = nfs4_have_delegation,
10399         .alloc_client   = nfs4_alloc_client,
10400         .init_client    = nfs4_init_client,
10401         .free_client    = nfs4_free_client,
10402         .create_server  = nfs4_create_server,
10403         .clone_server   = nfs_clone_server,
10404 };
10405
10406 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10407         .name   = XATTR_NAME_NFSV4_ACL,
10408         .list   = nfs4_xattr_list_nfs4_acl,
10409         .get    = nfs4_xattr_get_nfs4_acl,
10410         .set    = nfs4_xattr_set_nfs4_acl,
10411 };
10412
10413 #ifdef CONFIG_NFS_V4_2
10414 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10415         .prefix = XATTR_USER_PREFIX,
10416         .get    = nfs4_xattr_get_nfs4_user,
10417         .set    = nfs4_xattr_set_nfs4_user,
10418 };
10419 #endif
10420
10421 const struct xattr_handler *nfs4_xattr_handlers[] = {
10422         &nfs4_xattr_nfs4_acl_handler,
10423 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10424         &nfs4_xattr_nfs4_label_handler,
10425 #endif
10426 #ifdef CONFIG_NFS_V4_2
10427         &nfs4_xattr_nfs4_user_handler,
10428 #endif
10429         NULL
10430 };
10431
10432 /*
10433  * Local variables:
10434  *  c-basic-offset: 8
10435  * End:
10436  */