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,
  97                               struct nfs_fattr *fattr, struct inode *inode);
  98 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  99                             struct nfs_fattr *fattr, struct iattr *sattr,
 100                             struct nfs_open_context *ctx, struct nfs4_label *ilabel);
 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
 112 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 113 static inline struct nfs4_label *
 114 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 115         struct iattr *sattr, struct nfs4_label *label)
 116 {
 117         int err;
 118
 119         if (label == NULL)
 120                 return NULL;
 121
 122         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 123                 return NULL;
 124
 125         err = security_dentry_init_security(dentry, sattr->ia_mode,
 126                                 &dentry->d_name, NULL,
 127                                 (void **)&label->label, &label->len);
 128         if (err == 0)
 129                 return label;
 130
 131         return NULL;
 132 }
 133 static inline void
 134 nfs4_label_release_security(struct nfs4_label *label)
 135 {
 136         if (label)
 137                 security_release_secctx(label->label, label->len);
 138 }
 139 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 140 {
 141         if (label)
 142                 return server->attr_bitmask;
 143
 144         return server->attr_bitmask_nl;
 145 }
 146 #else
 147 static inline struct nfs4_label *
 148 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 149         struct iattr *sattr, struct nfs4_label *l)
 150 { return NULL; }
 151 static inline void
 152 nfs4_label_release_security(struct nfs4_label *label)
 153 { return; }
 154 static inline u32 *
 155 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 156 { return server->attr_bitmask; }
 157 #endif
 158
 159 /* Prevent leaks of NFSv4 errors into userland */
 160 static int nfs4_map_errors(int err)
 161 {
 162         if (err >= -1000)
 163                 return err;
 164         switch (err) {
 165         case -NFS4ERR_RESOURCE:
 166         case -NFS4ERR_LAYOUTTRYLATER:
 167         case -NFS4ERR_RECALLCONFLICT:
 168                 return -EREMOTEIO;
 169         case -NFS4ERR_WRONGSEC:
 170         case -NFS4ERR_WRONG_CRED:
 171                 return -EPERM;
 172         case -NFS4ERR_BADOWNER:
 173         case -NFS4ERR_BADNAME:
 174                 return -EINVAL;
 175         case -NFS4ERR_SHARE_DENIED:
 176                 return -EACCES;
 177         case -NFS4ERR_MINOR_VERS_MISMATCH:
 178                 return -EPROTONOSUPPORT;
 179         case -NFS4ERR_FILE_OPEN:
 180                 return -EBUSY;
 181         case -NFS4ERR_NOT_SAME:
 182                 return -ENOTSYNC;
 183         default:
 184                 dprintk("%s could not handle NFSv4 error %d\n",
 185                                 __func__, -err);
 186                 break;
 187         }
 188         return -EIO;
 189 }
 190
 191 /*
 192  * This is our standard bitmap for GETATTR requests.
 193  */
 194 const u32 nfs4_fattr_bitmap[3] = {
 195         FATTR4_WORD0_TYPE
 196         | FATTR4_WORD0_CHANGE
 197         | FATTR4_WORD0_SIZE
 198         | FATTR4_WORD0_FSID
 199         | FATTR4_WORD0_FILEID,
 200         FATTR4_WORD1_MODE
 201         | FATTR4_WORD1_NUMLINKS
 202         | FATTR4_WORD1_OWNER
 203         | FATTR4_WORD1_OWNER_GROUP
 204         | FATTR4_WORD1_RAWDEV
 205         | FATTR4_WORD1_SPACE_USED
 206         | FATTR4_WORD1_TIME_ACCESS
 207         | FATTR4_WORD1_TIME_METADATA
 208         | FATTR4_WORD1_TIME_MODIFY
 209         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 210 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 211         FATTR4_WORD2_SECURITY_LABEL
 212 #endif
 213 };
 214
 215 static const u32 nfs4_pnfs_open_bitmap[3] = {
 216         FATTR4_WORD0_TYPE
 217         | FATTR4_WORD0_CHANGE
 218         | FATTR4_WORD0_SIZE
 219         | FATTR4_WORD0_FSID
 220         | FATTR4_WORD0_FILEID,
 221         FATTR4_WORD1_MODE
 222         | FATTR4_WORD1_NUMLINKS
 223         | FATTR4_WORD1_OWNER
 224         | FATTR4_WORD1_OWNER_GROUP
 225         | FATTR4_WORD1_RAWDEV
 226         | FATTR4_WORD1_SPACE_USED
 227         | FATTR4_WORD1_TIME_ACCESS
 228         | FATTR4_WORD1_TIME_METADATA
 229         | FATTR4_WORD1_TIME_MODIFY,
 230         FATTR4_WORD2_MDSTHRESHOLD
 231 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 232         | FATTR4_WORD2_SECURITY_LABEL
 233 #endif
 234 };
 235
 236 static const u32 nfs4_open_noattr_bitmap[3] = {
 237         FATTR4_WORD0_TYPE
 238         | FATTR4_WORD0_FILEID,
 239 };
 240
 241 const u32 nfs4_statfs_bitmap[3] = {
 242         FATTR4_WORD0_FILES_AVAIL
 243         | FATTR4_WORD0_FILES_FREE
 244         | FATTR4_WORD0_FILES_TOTAL,
 245         FATTR4_WORD1_SPACE_AVAIL
 246         | FATTR4_WORD1_SPACE_FREE
 247         | FATTR4_WORD1_SPACE_TOTAL
 248 };
 249
 250 const u32 nfs4_pathconf_bitmap[3] = {
 251         FATTR4_WORD0_MAXLINK
 252         | FATTR4_WORD0_MAXNAME,
 253         0
 254 };
 255
 256 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 257                         | FATTR4_WORD0_MAXREAD
 258                         | FATTR4_WORD0_MAXWRITE
 259                         | FATTR4_WORD0_LEASE_TIME,
 260                         FATTR4_WORD1_TIME_DELTA
 261                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 262                         FATTR4_WORD2_LAYOUT_BLKSIZE
 263                         | FATTR4_WORD2_CLONE_BLKSIZE
 264                         | FATTR4_WORD2_CHANGE_ATTR_TYPE
 265                         | FATTR4_WORD2_XATTR_SUPPORT
 266 };
 267
 268 const u32 nfs4_fs_locations_bitmap[3] = {
 269         FATTR4_WORD0_CHANGE
 270         | FATTR4_WORD0_SIZE
 271         | FATTR4_WORD0_FSID
 272         | FATTR4_WORD0_FILEID
 273         | FATTR4_WORD0_FS_LOCATIONS,
 274         FATTR4_WORD1_OWNER
 275         | FATTR4_WORD1_OWNER_GROUP
 276         | FATTR4_WORD1_RAWDEV
 277         | FATTR4_WORD1_SPACE_USED
 278         | FATTR4_WORD1_TIME_ACCESS
 279         | FATTR4_WORD1_TIME_METADATA
 280         | FATTR4_WORD1_TIME_MODIFY
 281         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 282 };
 283
 284 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 285                                     struct inode *inode, unsigned long flags)
 286 {
 287         unsigned long cache_validity;
 288
 289         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 290         if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
 291                 return;
 292
 293         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
 294
 295         /* Remove the attributes over which we have full control */
 296         dst[1] &= ~FATTR4_WORD1_RAWDEV;
 297         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 298                 dst[0] &= ~FATTR4_WORD0_SIZE;
 299
 300         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 301                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 302
 303         if (!(cache_validity & NFS_INO_INVALID_MODE))
 304                 dst[1] &= ~FATTR4_WORD1_MODE;
 305         if (!(cache_validity & NFS_INO_INVALID_OTHER))
 306                 dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
 307 }
 308
 309 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 310                 struct nfs4_readdir_arg *readdir)
 311 {
 312         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 313         __be32 *start, *p;
 314
 315         if (cookie > 2) {
 316                 readdir->cookie = cookie;
 317                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 318                 return;
 319         }
 320
 321         readdir->cookie = 0;
 322         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 323         if (cookie == 2)
 324                 return;
 325
 326         /*
 327          * NFSv4 servers do not return entries for '.' and '..'
 328          * Therefore, we fake these entries here.  We let '.'
 329          * have cookie 0 and '..' have cookie 1.  Note that
 330          * when talking to the server, we always send cookie 0
 331          * instead of 1 or 2.
 332          */
 333         start = p = kmap_atomic(*readdir->pages);
 334
 335         if (cookie == 0) {
 336                 *p++ = xdr_one;                                  /* next */
 337                 *p++ = xdr_zero;                   /* cookie, first word */
 338                 *p++ = xdr_one;                   /* cookie, second word */
 339                 *p++ = xdr_one;                             /* entry len */
 340                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 341                 p++;
 342                 *p++ = xdr_one;                         /* bitmap length */
 343                 *p++ = htonl(attrs);                           /* bitmap */
 344                 *p++ = htonl(12);             /* attribute buffer length */
 345                 *p++ = htonl(NF4DIR);
 346                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 347         }
 348
 349         *p++ = xdr_one;                                  /* next */
 350         *p++ = xdr_zero;                   /* cookie, first word */
 351         *p++ = xdr_two;                   /* cookie, second word */
 352         *p++ = xdr_two;                             /* entry len */
 353         memcpy(p, "..\0\0", 4);                         /* entry */
 354         p++;
 355         *p++ = xdr_one;                         /* bitmap length */
 356         *p++ = htonl(attrs);                           /* bitmap */
 357         *p++ = htonl(12);             /* attribute buffer length */
 358         *p++ = htonl(NF4DIR);
 359         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 360
 361         readdir->pgbase = (char *)p - (char *)start;
 362         readdir->count -= readdir->pgbase;
 363         kunmap_atomic(start);
 364 }
 365
 366 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 367                 nfs4_stateid *stateid,
 368                 const struct cred *cred)
 369 {
 370         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 371
 372         ops->test_and_free_expired(server, stateid, cred);
 373 }
 374
 375 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 376                 nfs4_stateid *stateid,
 377                 const struct cred *cred)
 378 {
 379         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 380         nfs4_test_and_free_stateid(server, stateid, cred);
 381 }
 382
 383 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 384                 const nfs4_stateid *stateid,
 385                 const struct cred *cred)
 386 {
 387         nfs4_stateid tmp;
 388
 389         nfs4_stateid_copy(&tmp, stateid);
 390         __nfs4_free_revoked_stateid(server, &tmp, cred);
 391 }
 392
 393 static long nfs4_update_delay(long *timeout)
 394 {
 395         long ret;
 396         if (!timeout)
 397                 return NFS4_POLL_RETRY_MAX;
 398         if (*timeout <= 0)
 399                 *timeout = NFS4_POLL_RETRY_MIN;
 400         if (*timeout > NFS4_POLL_RETRY_MAX)
 401                 *timeout = NFS4_POLL_RETRY_MAX;
 402         ret = *timeout;
 403         *timeout <<= 1;
 404         return ret;
 405 }
 406
 407 static int nfs4_delay_killable(long *timeout)
 408 {
 409         might_sleep();
 410
 411         freezable_schedule_timeout_killable_unsafe(
 412                 nfs4_update_delay(timeout));
 413         if (!__fatal_signal_pending(current))
 414                 return 0;
 415         return -EINTR;
 416 }
 417
 418 static int nfs4_delay_interruptible(long *timeout)
 419 {
 420         might_sleep();
 421
 422         freezable_schedule_timeout_interruptible_unsafe(nfs4_update_delay(timeout));
 423         if (!signal_pending(current))
 424                 return 0;
 425         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 426 }
 427
 428 static int nfs4_delay(long *timeout, bool interruptible)
 429 {
 430         if (interruptible)
 431                 return nfs4_delay_interruptible(timeout);
 432         return nfs4_delay_killable(timeout);
 433 }
 434
 435 static const nfs4_stateid *
 436 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 437 {
 438         if (!stateid)
 439                 return NULL;
 440         switch (stateid->type) {
 441         case NFS4_OPEN_STATEID_TYPE:
 442         case NFS4_LOCK_STATEID_TYPE:
 443         case NFS4_DELEGATION_STATEID_TYPE:
 444                 return stateid;
 445         default:
 446                 break;
 447         }
 448         return NULL;
 449 }
 450
 451 /* This is the error handling routine for processes that are allowed
 452  * to sleep.
 453  */
 454 static int nfs4_do_handle_exception(struct nfs_server *server,
 455                 int errorcode, struct nfs4_exception *exception)
 456 {
 457         struct nfs_client *clp = server->nfs_client;
 458         struct nfs4_state *state = exception->state;
 459         const nfs4_stateid *stateid;
 460         struct inode *inode = exception->inode;
 461         int ret = errorcode;
 462
 463         exception->delay = 0;
 464         exception->recovering = 0;
 465         exception->retry = 0;
 466
 467         stateid = nfs4_recoverable_stateid(exception->stateid);
 468         if (stateid == NULL && state != NULL)
 469                 stateid = nfs4_recoverable_stateid(&state->stateid);
 470
 471         switch(errorcode) {
 472                 case 0:
 473                         return 0;
 474                 case -NFS4ERR_BADHANDLE:
 475                 case -ESTALE:
 476                         if (inode != NULL && S_ISREG(inode->i_mode))
 477                                 pnfs_destroy_layout(NFS_I(inode));
 478                         break;
 479                 case -NFS4ERR_DELEG_REVOKED:
 480                 case -NFS4ERR_ADMIN_REVOKED:
 481                 case -NFS4ERR_EXPIRED:
 482                 case -NFS4ERR_BAD_STATEID:
 483                 case -NFS4ERR_PARTNER_NO_AUTH:
 484                         if (inode != NULL && stateid != NULL) {
 485                                 nfs_inode_find_state_and_recover(inode,
 486                                                 stateid);
 487                                 goto wait_on_recovery;
 488                         }
 489                         fallthrough;
 490                 case -NFS4ERR_OPENMODE:
 491                         if (inode) {
 492                                 int err;
 493
 494                                 err = nfs_async_inode_return_delegation(inode,
 495                                                 stateid);
 496                                 if (err == 0)
 497                                         goto wait_on_recovery;
 498                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 499                                         exception->retry = 1;
 500                                         break;
 501                                 }
 502                         }
 503                         if (state == NULL)
 504                                 break;
 505                         ret = nfs4_schedule_stateid_recovery(server, state);
 506                         if (ret < 0)
 507                                 break;
 508                         goto wait_on_recovery;
 509                 case -NFS4ERR_STALE_STATEID:
 510                 case -NFS4ERR_STALE_CLIENTID:
 511                         nfs4_schedule_lease_recovery(clp);
 512                         goto wait_on_recovery;
 513                 case -NFS4ERR_MOVED:
 514                         ret = nfs4_schedule_migration_recovery(server);
 515                         if (ret < 0)
 516                                 break;
 517                         goto wait_on_recovery;
 518                 case -NFS4ERR_LEASE_MOVED:
 519                         nfs4_schedule_lease_moved_recovery(clp);
 520                         goto wait_on_recovery;
 521 #if defined(CONFIG_NFS_V4_1)
 522                 case -NFS4ERR_BADSESSION:
 523                 case -NFS4ERR_BADSLOT:
 524                 case -NFS4ERR_BAD_HIGH_SLOT:
 525                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 526                 case -NFS4ERR_DEADSESSION:
 527                 case -NFS4ERR_SEQ_FALSE_RETRY:
 528                 case -NFS4ERR_SEQ_MISORDERED:
 529                         /* Handled in nfs41_sequence_process() */
 530                         goto wait_on_recovery;
 531 #endif /* defined(CONFIG_NFS_V4_1) */
 532                 case -NFS4ERR_FILE_OPEN:
 533                         if (exception->timeout > HZ) {
 534                                 /* We have retried a decent amount, time to
 535                                  * fail
 536                                  */
 537                                 ret = -EBUSY;
 538                                 break;
 539                         }
 540                         fallthrough;
 541                 case -NFS4ERR_DELAY:
 542                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 543                         fallthrough;
 544                 case -NFS4ERR_GRACE:
 545                 case -NFS4ERR_LAYOUTTRYLATER:
 546                 case -NFS4ERR_RECALLCONFLICT:
 547                         exception->delay = 1;
 548                         return 0;
 549
 550                 case -NFS4ERR_RETRY_UNCACHED_REP:
 551                 case -NFS4ERR_OLD_STATEID:
 552                         exception->retry = 1;
 553                         break;
 554                 case -NFS4ERR_BADOWNER:
 555                         /* The following works around a Linux server bug! */
 556                 case -NFS4ERR_BADNAME:
 557                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 558                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 559                                 exception->retry = 1;
 560                                 printk(KERN_WARNING "NFS: v4 server %s "
 561                                                 "does not accept raw "
 562                                                 "uid/gids. "
 563                                                 "Reenabling the idmapper.\n",
 564                                                 server->nfs_client->cl_hostname);
 565                         }
 566         }
 567         /* We failed to handle the error */
 568         return nfs4_map_errors(ret);
 569 wait_on_recovery:
 570         exception->recovering = 1;
 571         return 0;
 572 }
 573
 574 /* This is the error handling routine for processes that are allowed
 575  * to sleep.
 576  */
 577 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 578 {
 579         struct nfs_client *clp = server->nfs_client;
 580         int ret;
 581
 582         ret = nfs4_do_handle_exception(server, errorcode, exception);
 583         if (exception->delay) {
 584                 ret = nfs4_delay(&exception->timeout,
 585                                 exception->interruptible);
 586                 goto out_retry;
 587         }
 588         if (exception->recovering) {
 589                 if (exception->task_is_privileged)
 590                         return -EDEADLOCK;
 591                 ret = nfs4_wait_clnt_recover(clp);
 592                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 593                         return -EIO;
 594                 goto out_retry;
 595         }
 596         return ret;
 597 out_retry:
 598         if (ret == 0)
 599                 exception->retry = 1;
 600         return ret;
 601 }
 602
 603 static int
 604 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 605                 int errorcode, struct nfs4_exception *exception)
 606 {
 607         struct nfs_client *clp = server->nfs_client;
 608         int ret;
 609
 610         ret = nfs4_do_handle_exception(server, errorcode, exception);
 611         if (exception->delay) {
 612                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 613                 goto out_retry;
 614         }
 615         if (exception->recovering) {
 616                 if (exception->task_is_privileged)
 617                         return -EDEADLOCK;
 618                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 619                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 620                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 621                 goto out_retry;
 622         }
 623         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 624                 ret = -EIO;
 625         return ret;
 626 out_retry:
 627         if (ret == 0) {
 628                 exception->retry = 1;
 629                 /*
 630                  * For NFS4ERR_MOVED, the client transport will need to
 631                  * be recomputed after migration recovery has completed.
 632                  */
 633                 if (errorcode == -NFS4ERR_MOVED)
 634                         rpc_task_release_transport(task);
 635         }
 636         return ret;
 637 }
 638
 639 int
 640 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 641                         struct nfs4_state *state, long *timeout)
 642 {
 643         struct nfs4_exception exception = {
 644                 .state = state,
 645         };
 646
 647         if (task->tk_status >= 0)
 648                 return 0;
 649         if (timeout)
 650                 exception.timeout = *timeout;
 651         task->tk_status = nfs4_async_handle_exception(task, server,
 652                         task->tk_status,
 653                         &exception);
 654         if (exception.delay && timeout)
 655                 *timeout = exception.timeout;
 656         if (exception.retry)
 657                 return -EAGAIN;
 658         return 0;
 659 }
 660
 661 /*
 662  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 663  * or 'false' otherwise.
 664  */
 665 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 666 {
 667         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 668         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 669 }
 670
 671 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 672 {
 673         spin_lock(&clp->cl_lock);
 674         if (time_before(clp->cl_last_renewal,timestamp))
 675                 clp->cl_last_renewal = timestamp;
 676         spin_unlock(&clp->cl_lock);
 677 }
 678
 679 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 680 {
 681         struct nfs_client *clp = server->nfs_client;
 682
 683         if (!nfs4_has_session(clp))
 684                 do_renew_lease(clp, timestamp);
 685 }
 686
 687 struct nfs4_call_sync_data {
 688         const struct nfs_server *seq_server;
 689         struct nfs4_sequence_args *seq_args;
 690         struct nfs4_sequence_res *seq_res;
 691 };
 692
 693 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 694                         struct nfs4_sequence_res *res, int cache_reply,
 695                         int privileged)
 696 {
 697         args->sa_slot = NULL;
 698         args->sa_cache_this = cache_reply;
 699         args->sa_privileged = privileged;
 700
 701         res->sr_slot = NULL;
 702 }
 703
 704 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 705 {
 706         struct nfs4_slot *slot = res->sr_slot;
 707         struct nfs4_slot_table *tbl;
 708
 709         tbl = slot->table;
 710         spin_lock(&tbl->slot_tbl_lock);
 711         if (!nfs41_wake_and_assign_slot(tbl, slot))
 712                 nfs4_free_slot(tbl, slot);
 713         spin_unlock(&tbl->slot_tbl_lock);
 714
 715         res->sr_slot = NULL;
 716 }
 717
 718 static int nfs40_sequence_done(struct rpc_task *task,
 719                                struct nfs4_sequence_res *res)
 720 {
 721         if (res->sr_slot != NULL)
 722                 nfs40_sequence_free_slot(res);
 723         return 1;
 724 }
 725
 726 #if defined(CONFIG_NFS_V4_1)
 727
 728 static void nfs41_release_slot(struct nfs4_slot *slot)
 729 {
 730         struct nfs4_session *session;
 731         struct nfs4_slot_table *tbl;
 732         bool send_new_highest_used_slotid = false;
 733
 734         if (!slot)
 735                 return;
 736         tbl = slot->table;
 737         session = tbl->session;
 738
 739         /* Bump the slot sequence number */
 740         if (slot->seq_done)
 741                 slot->seq_nr++;
 742         slot->seq_done = 0;
 743
 744         spin_lock(&tbl->slot_tbl_lock);
 745         /* Be nice to the server: try to ensure that the last transmitted
 746          * value for highest_user_slotid <= target_highest_slotid
 747          */
 748         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 749                 send_new_highest_used_slotid = true;
 750
 751         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 752                 send_new_highest_used_slotid = false;
 753                 goto out_unlock;
 754         }
 755         nfs4_free_slot(tbl, slot);
 756
 757         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 758                 send_new_highest_used_slotid = false;
 759 out_unlock:
 760         spin_unlock(&tbl->slot_tbl_lock);
 761         if (send_new_highest_used_slotid)
 762                 nfs41_notify_server(session->clp);
 763         if (waitqueue_active(&tbl->slot_waitq))
 764                 wake_up_all(&tbl->slot_waitq);
 765 }
 766
 767 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 768 {
 769         nfs41_release_slot(res->sr_slot);
 770         res->sr_slot = NULL;
 771 }
 772
 773 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 774                 u32 seqnr)
 775 {
 776         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 777                 slot->seq_nr_highest_sent = seqnr;
 778 }
 779 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot,
 780                 u32 seqnr)
 781 {
 782         slot->seq_nr_highest_sent = seqnr;
 783         slot->seq_nr_last_acked = seqnr;
 784 }
 785
 786 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 787                                 struct nfs4_slot *slot)
 788 {
 789         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 790         if (!IS_ERR(task))
 791                 rpc_put_task_async(task);
 792 }
 793
 794 static int nfs41_sequence_process(struct rpc_task *task,
 795                 struct nfs4_sequence_res *res)
 796 {
 797         struct nfs4_session *session;
 798         struct nfs4_slot *slot = res->sr_slot;
 799         struct nfs_client *clp;
 800         int status;
 801         int ret = 1;
 802
 803         if (slot == NULL)
 804                 goto out_noaction;
 805         /* don't increment the sequence number if the task wasn't sent */
 806         if (!RPC_WAS_SENT(task) || slot->seq_done)
 807                 goto out;
 808
 809         session = slot->table->session;
 810         clp = session->clp;
 811
 812         trace_nfs4_sequence_done(session, res);
 813
 814         status = res->sr_status;
 815         if (task->tk_status == -NFS4ERR_DEADSESSION)
 816                 status = -NFS4ERR_DEADSESSION;
 817
 818         /* Check the SEQUENCE operation status */
 819         switch (status) {
 820         case 0:
 821                 /* Mark this sequence number as having been acked */
 822                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 823                 /* Update the slot's sequence and clientid lease timer */
 824                 slot->seq_done = 1;
 825                 do_renew_lease(clp, res->sr_timestamp);
 826                 /* Check sequence flags */
 827                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 828                                 !!slot->privileged);
 829                 nfs41_update_target_slotid(slot->table, slot, res);
 830                 break;
 831         case 1:
 832                 /*
 833                  * sr_status remains 1 if an RPC level error occurred.
 834                  * The server may or may not have processed the sequence
 835                  * operation..
 836                  */
 837                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 838                 slot->seq_done = 1;
 839                 goto out;
 840         case -NFS4ERR_DELAY:
 841                 /* The server detected a resend of the RPC call and
 842                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 843                  * of RFC5661.
 844                  */
 845                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 846                         __func__,
 847                         slot->slot_nr,
 848                         slot->seq_nr);
 849                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 850                 goto out_retry;
 851         case -NFS4ERR_RETRY_UNCACHED_REP:
 852         case -NFS4ERR_SEQ_FALSE_RETRY:
 853                 /*
 854                  * The server thinks we tried to replay a request.
 855                  * Retry the call after bumping the sequence ID.
 856                  */
 857                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 858                 goto retry_new_seq;
 859         case -NFS4ERR_BADSLOT:
 860                 /*
 861                  * The slot id we used was probably retired. Try again
 862                  * using a different slot id.
 863                  */
 864                 if (slot->slot_nr < slot->table->target_highest_slotid)
 865                         goto session_recover;
 866                 goto retry_nowait;
 867         case -NFS4ERR_SEQ_MISORDERED:
 868                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 869                 /*
 870                  * Were one or more calls using this slot interrupted?
 871                  * If the server never received the request, then our
 872                  * transmitted slot sequence number may be too high. However,
 873                  * if the server did receive the request then it might
 874                  * accidentally give us a reply with a mismatched operation.
 875                  * We can sort this out by sending a lone sequence operation
 876                  * to the server on the same slot.
 877                  */
 878                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 879                         slot->seq_nr--;
 880                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 881                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 882                                 res->sr_slot = NULL;
 883                         }
 884                         goto retry_nowait;
 885                 }
 886                 /*
 887                  * RFC5661:
 888                  * A retry might be sent while the original request is
 889                  * still in progress on the replier. The replier SHOULD
 890                  * deal with the issue by returning NFS4ERR_DELAY as the
 891                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 892                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 893                  *
 894                  * Restart the search after a delay.
 895                  */
 896                 slot->seq_nr = slot->seq_nr_highest_sent;
 897                 goto out_retry;
 898         case -NFS4ERR_BADSESSION:
 899         case -NFS4ERR_DEADSESSION:
 900         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 901                 goto session_recover;
 902         default:
 903                 /* Just update the slot sequence no. */
 904                 slot->seq_done = 1;
 905         }
 906 out:
 907         /* The session may be reset by one of the error handlers. */
 908         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 909 out_noaction:
 910         return ret;
 911 session_recover:
 912         nfs4_schedule_session_recovery(session, status);
 913         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 914         nfs41_sequence_free_slot(res);
 915         goto out;
 916 retry_new_seq:
 917         ++slot->seq_nr;
 918 retry_nowait:
 919         if (rpc_restart_call_prepare(task)) {
 920                 nfs41_sequence_free_slot(res);
 921                 task->tk_status = 0;
 922                 ret = 0;
 923         }
 924         goto out;
 925 out_retry:
 926         if (!rpc_restart_call(task))
 927                 goto out;
 928         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 929         return 0;
 930 }
 931
 932 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 933 {
 934         if (!nfs41_sequence_process(task, res))
 935                 return 0;
 936         if (res->sr_slot != NULL)
 937                 nfs41_sequence_free_slot(res);
 938         return 1;
 939
 940 }
 941 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 942
 943 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 944 {
 945         if (res->sr_slot == NULL)
 946                 return 1;
 947         if (res->sr_slot->table->session != NULL)
 948                 return nfs41_sequence_process(task, res);
 949         return nfs40_sequence_done(task, res);
 950 }
 951
 952 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
 953 {
 954         if (res->sr_slot != NULL) {
 955                 if (res->sr_slot->table->session != NULL)
 956                         nfs41_sequence_free_slot(res);
 957                 else
 958                         nfs40_sequence_free_slot(res);
 959         }
 960 }
 961
 962 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 963 {
 964         if (res->sr_slot == NULL)
 965                 return 1;
 966         if (!res->sr_slot->table->session)
 967                 return nfs40_sequence_done(task, res);
 968         return nfs41_sequence_done(task, res);
 969 }
 970 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
 971
 972 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
 973 {
 974         struct nfs4_call_sync_data *data = calldata;
 975
 976         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
 977
 978         nfs4_setup_sequence(data->seq_server->nfs_client,
 979                             data->seq_args, data->seq_res, task);
 980 }
 981
 982 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
 983 {
 984         struct nfs4_call_sync_data *data = calldata;
 985
 986         nfs41_sequence_done(task, data->seq_res);
 987 }
 988
 989 static const struct rpc_call_ops nfs41_call_sync_ops = {
 990         .rpc_call_prepare = nfs41_call_sync_prepare,
 991         .rpc_call_done = nfs41_call_sync_done,
 992 };
 993
 994 #else   /* !CONFIG_NFS_V4_1 */
 995
 996 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 997 {
 998         return nfs40_sequence_done(task, res);
 999 }
1000
1001 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1002 {
1003         if (res->sr_slot != NULL)
1004                 nfs40_sequence_free_slot(res);
1005 }
1006
1007 int nfs4_sequence_done(struct rpc_task *task,
1008                        struct nfs4_sequence_res *res)
1009 {
1010         return nfs40_sequence_done(task, res);
1011 }
1012 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1013
1014 #endif  /* !CONFIG_NFS_V4_1 */
1015
1016 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1017 {
1018         res->sr_timestamp = jiffies;
1019         res->sr_status_flags = 0;
1020         res->sr_status = 1;
1021 }
1022
1023 static
1024 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1025                 struct nfs4_sequence_res *res,
1026                 struct nfs4_slot *slot)
1027 {
1028         if (!slot)
1029                 return;
1030         slot->privileged = args->sa_privileged ? 1 : 0;
1031         args->sa_slot = slot;
1032
1033         res->sr_slot = slot;
1034 }
1035
1036 int nfs4_setup_sequence(struct nfs_client *client,
1037                         struct nfs4_sequence_args *args,
1038                         struct nfs4_sequence_res *res,
1039                         struct rpc_task *task)
1040 {
1041         struct nfs4_session *session = nfs4_get_session(client);
1042         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1043         struct nfs4_slot *slot;
1044
1045         /* slot already allocated? */
1046         if (res->sr_slot != NULL)
1047                 goto out_start;
1048
1049         if (session)
1050                 tbl = &session->fc_slot_table;
1051
1052         spin_lock(&tbl->slot_tbl_lock);
1053         /* The state manager will wait until the slot table is empty */
1054         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1055                 goto out_sleep;
1056
1057         slot = nfs4_alloc_slot(tbl);
1058         if (IS_ERR(slot)) {
1059                 if (slot == ERR_PTR(-ENOMEM))
1060                         goto out_sleep_timeout;
1061                 goto out_sleep;
1062         }
1063         spin_unlock(&tbl->slot_tbl_lock);
1064
1065         nfs4_sequence_attach_slot(args, res, slot);
1066
1067         trace_nfs4_setup_sequence(session, args);
1068 out_start:
1069         nfs41_sequence_res_init(res);
1070         rpc_call_start(task);
1071         return 0;
1072 out_sleep_timeout:
1073         /* Try again in 1/4 second */
1074         if (args->sa_privileged)
1075                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1076                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1077         else
1078                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1079                                 NULL, jiffies + (HZ >> 2));
1080         spin_unlock(&tbl->slot_tbl_lock);
1081         return -EAGAIN;
1082 out_sleep:
1083         if (args->sa_privileged)
1084                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1085                                 RPC_PRIORITY_PRIVILEGED);
1086         else
1087                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1088         spin_unlock(&tbl->slot_tbl_lock);
1089         return -EAGAIN;
1090 }
1091 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1092
1093 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1094 {
1095         struct nfs4_call_sync_data *data = calldata;
1096         nfs4_setup_sequence(data->seq_server->nfs_client,
1097                                 data->seq_args, data->seq_res, task);
1098 }
1099
1100 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1101 {
1102         struct nfs4_call_sync_data *data = calldata;
1103         nfs4_sequence_done(task, data->seq_res);
1104 }
1105
1106 static const struct rpc_call_ops nfs40_call_sync_ops = {
1107         .rpc_call_prepare = nfs40_call_sync_prepare,
1108         .rpc_call_done = nfs40_call_sync_done,
1109 };
1110
1111 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1112 {
1113         int ret;
1114         struct rpc_task *task;
1115
1116         task = rpc_run_task(task_setup);
1117         if (IS_ERR(task))
1118                 return PTR_ERR(task);
1119
1120         ret = task->tk_status;
1121         rpc_put_task(task);
1122         return ret;
1123 }
1124
1125 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1126                              struct nfs_server *server,
1127                              struct rpc_message *msg,
1128                              struct nfs4_sequence_args *args,
1129                              struct nfs4_sequence_res *res,
1130                              unsigned short task_flags)
1131 {
1132         struct nfs_client *clp = server->nfs_client;
1133         struct nfs4_call_sync_data data = {
1134                 .seq_server = server,
1135                 .seq_args = args,
1136                 .seq_res = res,
1137         };
1138         struct rpc_task_setup task_setup = {
1139                 .rpc_client = clnt,
1140                 .rpc_message = msg,
1141                 .callback_ops = clp->cl_mvops->call_sync_ops,
1142                 .callback_data = &data,
1143                 .flags = task_flags,
1144         };
1145
1146         return nfs4_call_sync_custom(&task_setup);
1147 }
1148
1149 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1150                                    struct nfs_server *server,
1151                                    struct rpc_message *msg,
1152                                    struct nfs4_sequence_args *args,
1153                                    struct nfs4_sequence_res *res)
1154 {
1155         unsigned short task_flags = 0;
1156
1157         if (server->nfs_client->cl_minorversion)
1158                 task_flags = RPC_TASK_MOVEABLE;
1159         return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1160 }
1161
1162
1163 int nfs4_call_sync(struct rpc_clnt *clnt,
1164                    struct nfs_server *server,
1165                    struct rpc_message *msg,
1166                    struct nfs4_sequence_args *args,
1167                    struct nfs4_sequence_res *res,
1168                    int cache_reply)
1169 {
1170         nfs4_init_sequence(args, res, cache_reply, 0);
1171         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1172 }
1173
1174 static void
1175 nfs4_inc_nlink_locked(struct inode *inode)
1176 {
1177         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1178                                              NFS_INO_INVALID_CTIME |
1179                                              NFS_INO_INVALID_NLINK);
1180         inc_nlink(inode);
1181 }
1182
1183 static void
1184 nfs4_inc_nlink(struct inode *inode)
1185 {
1186         spin_lock(&inode->i_lock);
1187         nfs4_inc_nlink_locked(inode);
1188         spin_unlock(&inode->i_lock);
1189 }
1190
1191 static void
1192 nfs4_dec_nlink_locked(struct inode *inode)
1193 {
1194         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1195                                              NFS_INO_INVALID_CTIME |
1196                                              NFS_INO_INVALID_NLINK);
1197         drop_nlink(inode);
1198 }
1199
1200 static void
1201 nfs4_update_changeattr_locked(struct inode *inode,
1202                 struct nfs4_change_info *cinfo,
1203                 unsigned long timestamp, unsigned long cache_validity)
1204 {
1205         struct nfs_inode *nfsi = NFS_I(inode);
1206         u64 change_attr = inode_peek_iversion_raw(inode);
1207
1208         cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1209         if (S_ISDIR(inode->i_mode))
1210                 cache_validity |= NFS_INO_INVALID_DATA;
1211
1212         switch (NFS_SERVER(inode)->change_attr_type) {
1213         case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1214                 if (cinfo->after == change_attr)
1215                         goto out;
1216                 break;
1217         default:
1218                 if ((s64)(change_attr - cinfo->after) >= 0)
1219                         goto out;
1220         }
1221
1222         inode_set_iversion_raw(inode, cinfo->after);
1223         if (!cinfo->atomic || cinfo->before != change_attr) {
1224                 if (S_ISDIR(inode->i_mode))
1225                         nfs_force_lookup_revalidate(inode);
1226
1227                 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1228                         cache_validity |=
1229                                 NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1230                                 NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1231                                 NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1232                                 NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1233                 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1234         }
1235         nfsi->attrtimeo_timestamp = jiffies;
1236         nfsi->read_cache_jiffies = timestamp;
1237         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1238         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1239 out:
1240         nfs_set_cache_invalid(inode, cache_validity);
1241 }
1242
1243 void
1244 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1245                 unsigned long timestamp, unsigned long cache_validity)
1246 {
1247         spin_lock(&dir->i_lock);
1248         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1249         spin_unlock(&dir->i_lock);
1250 }
1251
1252 struct nfs4_open_createattrs {
1253         struct nfs4_label *label;
1254         struct iattr *sattr;
1255         const __u32 verf[2];
1256 };
1257
1258 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1259                 int err, struct nfs4_exception *exception)
1260 {
1261         if (err != -EINVAL)
1262                 return false;
1263         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1264                 return false;
1265         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1266         exception->retry = 1;
1267         return true;
1268 }
1269
1270 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1271 {
1272          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1273 }
1274
1275 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1276 {
1277         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1278
1279         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1280 }
1281
1282 static u32
1283 nfs4_map_atomic_open_share(struct nfs_server *server,
1284                 fmode_t fmode, int openflags)
1285 {
1286         u32 res = 0;
1287
1288         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1289         case FMODE_READ:
1290                 res = NFS4_SHARE_ACCESS_READ;
1291                 break;
1292         case FMODE_WRITE:
1293                 res = NFS4_SHARE_ACCESS_WRITE;
1294                 break;
1295         case FMODE_READ|FMODE_WRITE:
1296                 res = NFS4_SHARE_ACCESS_BOTH;
1297         }
1298         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1299                 goto out;
1300         /* Want no delegation if we're using O_DIRECT */
1301         if (openflags & O_DIRECT)
1302                 res |= NFS4_SHARE_WANT_NO_DELEG;
1303 out:
1304         return res;
1305 }
1306
1307 static enum open_claim_type4
1308 nfs4_map_atomic_open_claim(struct nfs_server *server,
1309                 enum open_claim_type4 claim)
1310 {
1311         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1312                 return claim;
1313         switch (claim) {
1314         default:
1315                 return claim;
1316         case NFS4_OPEN_CLAIM_FH:
1317                 return NFS4_OPEN_CLAIM_NULL;
1318         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1319                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1320         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1321                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1322         }
1323 }
1324
1325 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1326 {
1327         p->o_res.f_attr = &p->f_attr;
1328         p->o_res.seqid = p->o_arg.seqid;
1329         p->c_res.seqid = p->c_arg.seqid;
1330         p->o_res.server = p->o_arg.server;
1331         p->o_res.access_request = p->o_arg.access;
1332         nfs_fattr_init(&p->f_attr);
1333         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1334 }
1335
1336 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1337                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1338                 const struct nfs4_open_createattrs *c,
1339                 enum open_claim_type4 claim,
1340                 gfp_t gfp_mask)
1341 {
1342         struct dentry *parent = dget_parent(dentry);
1343         struct inode *dir = d_inode(parent);
1344         struct nfs_server *server = NFS_SERVER(dir);
1345         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1346         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1347         struct nfs4_opendata *p;
1348
1349         p = kzalloc(sizeof(*p), gfp_mask);
1350         if (p == NULL)
1351                 goto err;
1352
1353         p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1354         if (IS_ERR(p->f_attr.label))
1355                 goto err_free_p;
1356
1357         p->a_label = nfs4_label_alloc(server, gfp_mask);
1358         if (IS_ERR(p->a_label))
1359                 goto err_free_f;
1360
1361         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1362         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1363         if (IS_ERR(p->o_arg.seqid))
1364                 goto err_free_label;
1365         nfs_sb_active(dentry->d_sb);
1366         p->dentry = dget(dentry);
1367         p->dir = parent;
1368         p->owner = sp;
1369         atomic_inc(&sp->so_count);
1370         p->o_arg.open_flags = flags;
1371         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1372         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1373         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1374                         fmode, flags);
1375         if (flags & O_CREAT) {
1376                 p->o_arg.umask = current_umask();
1377                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1378                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1379                         p->o_arg.u.attrs = &p->attrs;
1380                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1381
1382                         memcpy(p->o_arg.u.verifier.data, c->verf,
1383                                         sizeof(p->o_arg.u.verifier.data));
1384                 }
1385         }
1386         /* ask server to check for all possible rights as results
1387          * are cached */
1388         switch (p->o_arg.claim) {
1389         default:
1390                 break;
1391         case NFS4_OPEN_CLAIM_NULL:
1392         case NFS4_OPEN_CLAIM_FH:
1393                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1394                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1395                                   NFS4_ACCESS_EXECUTE;
1396 #ifdef CONFIG_NFS_V4_2
1397                 if (!(server->caps & NFS_CAP_XATTR))
1398                         break;
1399                 p->o_arg.access |= NFS4_ACCESS_XAREAD | NFS4_ACCESS_XAWRITE |
1400                                    NFS4_ACCESS_XALIST;
1401 #endif
1402         }
1403         p->o_arg.clientid = server->nfs_client->cl_clientid;
1404         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1405         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1406         p->o_arg.name = &dentry->d_name;
1407         p->o_arg.server = server;
1408         p->o_arg.bitmask = nfs4_bitmask(server, label);
1409         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1410         switch (p->o_arg.claim) {
1411         case NFS4_OPEN_CLAIM_NULL:
1412         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1413         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1414                 p->o_arg.fh = NFS_FH(dir);
1415                 break;
1416         case NFS4_OPEN_CLAIM_PREVIOUS:
1417         case NFS4_OPEN_CLAIM_FH:
1418         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1419         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1420                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1421         }
1422         p->c_arg.fh = &p->o_res.fh;
1423         p->c_arg.stateid = &p->o_res.stateid;
1424         p->c_arg.seqid = p->o_arg.seqid;
1425         nfs4_init_opendata_res(p);
1426         kref_init(&p->kref);
1427         return p;
1428
1429 err_free_label:
1430         nfs4_label_free(p->a_label);
1431 err_free_f:
1432         nfs4_label_free(p->f_attr.label);
1433 err_free_p:
1434         kfree(p);
1435 err:
1436         dput(parent);
1437         return NULL;
1438 }
1439
1440 static void nfs4_opendata_free(struct kref *kref)
1441 {
1442         struct nfs4_opendata *p = container_of(kref,
1443                         struct nfs4_opendata, kref);
1444         struct super_block *sb = p->dentry->d_sb;
1445
1446         nfs4_lgopen_release(p->lgp);
1447         nfs_free_seqid(p->o_arg.seqid);
1448         nfs4_sequence_free_slot(&p->o_res.seq_res);
1449         if (p->state != NULL)
1450                 nfs4_put_open_state(p->state);
1451         nfs4_put_state_owner(p->owner);
1452
1453         nfs4_label_free(p->a_label);
1454         nfs4_label_free(p->f_attr.label);
1455
1456         dput(p->dir);
1457         dput(p->dentry);
1458         nfs_sb_deactive(sb);
1459         nfs_fattr_free_names(&p->f_attr);
1460         kfree(p->f_attr.mdsthreshold);
1461         kfree(p);
1462 }
1463
1464 static void nfs4_opendata_put(struct nfs4_opendata *p)
1465 {
1466         if (p != NULL)
1467                 kref_put(&p->kref, nfs4_opendata_free);
1468 }
1469
1470 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1471                 fmode_t fmode)
1472 {
1473         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1474         case FMODE_READ|FMODE_WRITE:
1475                 return state->n_rdwr != 0;
1476         case FMODE_WRITE:
1477                 return state->n_wronly != 0;
1478         case FMODE_READ:
1479                 return state->n_rdonly != 0;
1480         }
1481         WARN_ON_ONCE(1);
1482         return false;
1483 }
1484
1485 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1486                 int open_mode, enum open_claim_type4 claim)
1487 {
1488         int ret = 0;
1489
1490         if (open_mode & (O_EXCL|O_TRUNC))
1491                 goto out;
1492         switch (claim) {
1493         case NFS4_OPEN_CLAIM_NULL:
1494         case NFS4_OPEN_CLAIM_FH:
1495                 goto out;
1496         default:
1497                 break;
1498         }
1499         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1500                 case FMODE_READ:
1501                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1502                                 && state->n_rdonly != 0;
1503                         break;
1504                 case FMODE_WRITE:
1505                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1506                                 && state->n_wronly != 0;
1507                         break;
1508                 case FMODE_READ|FMODE_WRITE:
1509                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1510                                 && state->n_rdwr != 0;
1511         }
1512 out:
1513         return ret;
1514 }
1515
1516 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1517                 enum open_claim_type4 claim)
1518 {
1519         if (delegation == NULL)
1520                 return 0;
1521         if ((delegation->type & fmode) != fmode)
1522                 return 0;
1523         switch (claim) {
1524         case NFS4_OPEN_CLAIM_NULL:
1525         case NFS4_OPEN_CLAIM_FH:
1526                 break;
1527         case NFS4_OPEN_CLAIM_PREVIOUS:
1528                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1529                         break;
1530                 fallthrough;
1531         default:
1532                 return 0;
1533         }
1534         nfs_mark_delegation_referenced(delegation);
1535         return 1;
1536 }
1537
1538 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1539 {
1540         switch (fmode) {
1541                 case FMODE_WRITE:
1542                         state->n_wronly++;
1543                         break;
1544                 case FMODE_READ:
1545                         state->n_rdonly++;
1546                         break;
1547                 case FMODE_READ|FMODE_WRITE:
1548                         state->n_rdwr++;
1549         }
1550         nfs4_state_set_mode_locked(state, state->state | fmode);
1551 }
1552
1553 #ifdef CONFIG_NFS_V4_1
1554 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1555 {
1556         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1557                 return true;
1558         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1559                 return true;
1560         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1561                 return true;
1562         return false;
1563 }
1564 #endif /* CONFIG_NFS_V4_1 */
1565
1566 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1567 {
1568         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1569                 wake_up_all(&state->waitq);
1570 }
1571
1572 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1573 {
1574         struct nfs_client *clp = state->owner->so_server->nfs_client;
1575         bool need_recover = false;
1576
1577         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1578                 need_recover = true;
1579         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1580                 need_recover = true;
1581         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1582                 need_recover = true;
1583         if (need_recover)
1584                 nfs4_state_mark_reclaim_nograce(clp, state);
1585 }
1586
1587 /*
1588  * Check for whether or not the caller may update the open stateid
1589  * to the value passed in by stateid.
1590  *
1591  * Note: This function relies heavily on the server implementing
1592  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1593  * correctly.
1594  * i.e. The stateid seqids have to be initialised to 1, and
1595  * are then incremented on every state transition.
1596  */
1597 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1598                 const nfs4_stateid *stateid)
1599 {
1600         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1601                 /* The common case - we're updating to a new sequence number */
1602                 if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1603                         if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1604                                 return true;
1605                         return false;
1606                 }
1607                 /* The server returned a new stateid */
1608         }
1609         /* This is the first OPEN in this generation */
1610         if (stateid->seqid == cpu_to_be32(1))
1611                 return true;
1612         return false;
1613 }
1614
1615 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1616 {
1617         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1618                 return;
1619         if (state->n_wronly)
1620                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1621         if (state->n_rdonly)
1622                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1623         if (state->n_rdwr)
1624                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1625         set_bit(NFS_OPEN_STATE, &state->flags);
1626 }
1627
1628 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1629                 nfs4_stateid *stateid, fmode_t fmode)
1630 {
1631         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1632         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1633         case FMODE_WRITE:
1634                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1635                 break;
1636         case FMODE_READ:
1637                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1638                 break;
1639         case 0:
1640                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1641                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1642                 clear_bit(NFS_OPEN_STATE, &state->flags);
1643         }
1644         if (stateid == NULL)
1645                 return;
1646         /* Handle OPEN+OPEN_DOWNGRADE races */
1647         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1648             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1649                 nfs_resync_open_stateid_locked(state);
1650                 goto out;
1651         }
1652         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1653                 nfs4_stateid_copy(&state->stateid, stateid);
1654         nfs4_stateid_copy(&state->open_stateid, stateid);
1655         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1656 out:
1657         nfs_state_log_update_open_stateid(state);
1658 }
1659
1660 static void nfs_clear_open_stateid(struct nfs4_state *state,
1661         nfs4_stateid *arg_stateid,
1662         nfs4_stateid *stateid, fmode_t fmode)
1663 {
1664         write_seqlock(&state->seqlock);
1665         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1666         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1667                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1668         write_sequnlock(&state->seqlock);
1669         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1670                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1671 }
1672
1673 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1674                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1675         __must_hold(&state->owner->so_lock)
1676         __must_hold(&state->seqlock)
1677         __must_hold(RCU)
1678
1679 {
1680         DEFINE_WAIT(wait);
1681         int status = 0;
1682         for (;;) {
1683
1684                 if (nfs_stateid_is_sequential(state, stateid))
1685                         break;
1686
1687                 if (status)
1688                         break;
1689                 /* Rely on seqids for serialisation with NFSv4.0 */
1690                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1691                         break;
1692
1693                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1694                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1695                 /*
1696                  * Ensure we process the state changes in the same order
1697                  * in which the server processed them by delaying the
1698                  * update of the stateid until we are in sequence.
1699                  */
1700                 write_sequnlock(&state->seqlock);
1701                 spin_unlock(&state->owner->so_lock);
1702                 rcu_read_unlock();
1703                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1704
1705                 if (!fatal_signal_pending(current)) {
1706                         if (schedule_timeout(5*HZ) == 0)
1707                                 status = -EAGAIN;
1708                         else
1709                                 status = 0;
1710                 } else
1711                         status = -EINTR;
1712                 finish_wait(&state->waitq, &wait);
1713                 rcu_read_lock();
1714                 spin_lock(&state->owner->so_lock);
1715                 write_seqlock(&state->seqlock);
1716         }
1717
1718         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1719             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1720                 nfs4_stateid_copy(freeme, &state->open_stateid);
1721                 nfs_test_and_clear_all_open_stateid(state);
1722         }
1723
1724         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1725                 nfs4_stateid_copy(&state->stateid, stateid);
1726         nfs4_stateid_copy(&state->open_stateid, stateid);
1727         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1728         nfs_state_log_update_open_stateid(state);
1729 }
1730
1731 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1732                 const nfs4_stateid *open_stateid,
1733                 fmode_t fmode,
1734                 nfs4_stateid *freeme)
1735 {
1736         /*
1737          * Protect the call to nfs4_state_set_mode_locked and
1738          * serialise the stateid update
1739          */
1740         write_seqlock(&state->seqlock);
1741         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1742         switch (fmode) {
1743         case FMODE_READ:
1744                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1745                 break;
1746         case FMODE_WRITE:
1747                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1748                 break;
1749         case FMODE_READ|FMODE_WRITE:
1750                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1751         }
1752         set_bit(NFS_OPEN_STATE, &state->flags);
1753         write_sequnlock(&state->seqlock);
1754 }
1755
1756 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1757 {
1758         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1759         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1760         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1761         clear_bit(NFS_OPEN_STATE, &state->flags);
1762 }
1763
1764 static void nfs_state_set_delegation(struct nfs4_state *state,
1765                 const nfs4_stateid *deleg_stateid,
1766                 fmode_t fmode)
1767 {
1768         /*
1769          * Protect the call to nfs4_state_set_mode_locked and
1770          * serialise the stateid update
1771          */
1772         write_seqlock(&state->seqlock);
1773         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1774         set_bit(NFS_DELEGATED_STATE, &state->flags);
1775         write_sequnlock(&state->seqlock);
1776 }
1777
1778 static void nfs_state_clear_delegation(struct nfs4_state *state)
1779 {
1780         write_seqlock(&state->seqlock);
1781         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1782         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1783         write_sequnlock(&state->seqlock);
1784 }
1785
1786 int update_open_stateid(struct nfs4_state *state,
1787                 const nfs4_stateid *open_stateid,
1788                 const nfs4_stateid *delegation,
1789                 fmode_t fmode)
1790 {
1791         struct nfs_server *server = NFS_SERVER(state->inode);
1792         struct nfs_client *clp = server->nfs_client;
1793         struct nfs_inode *nfsi = NFS_I(state->inode);
1794         struct nfs_delegation *deleg_cur;
1795         nfs4_stateid freeme = { };
1796         int ret = 0;
1797
1798         fmode &= (FMODE_READ|FMODE_WRITE);
1799
1800         rcu_read_lock();
1801         spin_lock(&state->owner->so_lock);
1802         if (open_stateid != NULL) {
1803                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1804                 ret = 1;
1805         }
1806
1807         deleg_cur = nfs4_get_valid_delegation(state->inode);
1808         if (deleg_cur == NULL)
1809                 goto no_delegation;
1810
1811         spin_lock(&deleg_cur->lock);
1812         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1813            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1814             (deleg_cur->type & fmode) != fmode)
1815                 goto no_delegation_unlock;
1816
1817         if (delegation == NULL)
1818                 delegation = &deleg_cur->stateid;
1819         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1820                 goto no_delegation_unlock;
1821
1822         nfs_mark_delegation_referenced(deleg_cur);
1823         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1824         ret = 1;
1825 no_delegation_unlock:
1826         spin_unlock(&deleg_cur->lock);
1827 no_delegation:
1828         if (ret)
1829                 update_open_stateflags(state, fmode);
1830         spin_unlock(&state->owner->so_lock);
1831         rcu_read_unlock();
1832
1833         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1834                 nfs4_schedule_state_manager(clp);
1835         if (freeme.type != 0)
1836                 nfs4_test_and_free_stateid(server, &freeme,
1837                                 state->owner->so_cred);
1838
1839         return ret;
1840 }
1841
1842 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1843                 const nfs4_stateid *stateid)
1844 {
1845         struct nfs4_state *state = lsp->ls_state;
1846         bool ret = false;
1847
1848         spin_lock(&state->state_lock);
1849         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1850                 goto out_noupdate;
1851         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1852                 goto out_noupdate;
1853         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1854         ret = true;
1855 out_noupdate:
1856         spin_unlock(&state->state_lock);
1857         return ret;
1858 }
1859
1860 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1861 {
1862         struct nfs_delegation *delegation;
1863
1864         fmode &= FMODE_READ|FMODE_WRITE;
1865         rcu_read_lock();
1866         delegation = nfs4_get_valid_delegation(inode);
1867         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1868                 rcu_read_unlock();
1869                 return;
1870         }
1871         rcu_read_unlock();
1872         nfs4_inode_return_delegation(inode);
1873 }
1874
1875 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1876 {
1877         struct nfs4_state *state = opendata->state;
1878         struct nfs_delegation *delegation;
1879         int open_mode = opendata->o_arg.open_flags;
1880         fmode_t fmode = opendata->o_arg.fmode;
1881         enum open_claim_type4 claim = opendata->o_arg.claim;
1882         nfs4_stateid stateid;
1883         int ret = -EAGAIN;
1884
1885         for (;;) {
1886                 spin_lock(&state->owner->so_lock);
1887                 if (can_open_cached(state, fmode, open_mode, claim)) {
1888                         update_open_stateflags(state, fmode);
1889                         spin_unlock(&state->owner->so_lock);
1890                         goto out_return_state;
1891                 }
1892                 spin_unlock(&state->owner->so_lock);
1893                 rcu_read_lock();
1894                 delegation = nfs4_get_valid_delegation(state->inode);
1895                 if (!can_open_delegated(delegation, fmode, claim)) {
1896                         rcu_read_unlock();
1897                         break;
1898                 }
1899                 /* Save the delegation */
1900                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1901                 rcu_read_unlock();
1902                 nfs_release_seqid(opendata->o_arg.seqid);
1903                 if (!opendata->is_recover) {
1904                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1905                         if (ret != 0)
1906                                 goto out;
1907                 }
1908                 ret = -EAGAIN;
1909
1910                 /* Try to update the stateid using the delegation */
1911                 if (update_open_stateid(state, NULL, &stateid, fmode))
1912                         goto out_return_state;
1913         }
1914 out:
1915         return ERR_PTR(ret);
1916 out_return_state:
1917         refcount_inc(&state->count);
1918         return state;
1919 }
1920
1921 static void
1922 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1923 {
1924         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1925         struct nfs_delegation *delegation;
1926         int delegation_flags = 0;
1927
1928         rcu_read_lock();
1929         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1930         if (delegation)
1931                 delegation_flags = delegation->flags;
1932         rcu_read_unlock();
1933         switch (data->o_arg.claim) {
1934         default:
1935                 break;
1936         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1937         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1938                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1939                                    "returning a delegation for "
1940                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1941                                    clp->cl_hostname);
1942                 return;
1943         }
1944         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1945                 nfs_inode_set_delegation(state->inode,
1946                                 data->owner->so_cred,
1947                                 data->o_res.delegation_type,
1948                                 &data->o_res.delegation,
1949                                 data->o_res.pagemod_limit);
1950         else
1951                 nfs_inode_reclaim_delegation(state->inode,
1952                                 data->owner->so_cred,
1953                                 data->o_res.delegation_type,
1954                                 &data->o_res.delegation,
1955                                 data->o_res.pagemod_limit);
1956
1957         if (data->o_res.do_recall)
1958                 nfs_async_inode_return_delegation(state->inode,
1959                                                   &data->o_res.delegation);
1960 }
1961
1962 /*
1963  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1964  * and update the nfs4_state.
1965  */
1966 static struct nfs4_state *
1967 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1968 {
1969         struct inode *inode = data->state->inode;
1970         struct nfs4_state *state = data->state;
1971         int ret;
1972
1973         if (!data->rpc_done) {
1974                 if (data->rpc_status)
1975                         return ERR_PTR(data->rpc_status);
1976                 /* cached opens have already been processed */
1977                 goto update;
1978         }
1979
1980         ret = nfs_refresh_inode(inode, &data->f_attr);
1981         if (ret)
1982                 return ERR_PTR(ret);
1983
1984         if (data->o_res.delegation_type != 0)
1985                 nfs4_opendata_check_deleg(data, state);
1986 update:
1987         if (!update_open_stateid(state, &data->o_res.stateid,
1988                                 NULL, data->o_arg.fmode))
1989                 return ERR_PTR(-EAGAIN);
1990         refcount_inc(&state->count);
1991
1992         return state;
1993 }
1994
1995 static struct inode *
1996 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1997 {
1998         struct inode *inode;
1999
2000         switch (data->o_arg.claim) {
2001         case NFS4_OPEN_CLAIM_NULL:
2002         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2003         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2004                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2005                         return ERR_PTR(-EAGAIN);
2006                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2007                                 &data->f_attr);
2008                 break;
2009         default:
2010                 inode = d_inode(data->dentry);
2011                 ihold(inode);
2012                 nfs_refresh_inode(inode, &data->f_attr);
2013         }
2014         return inode;
2015 }
2016
2017 static struct nfs4_state *
2018 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2019 {
2020         struct nfs4_state *state;
2021         struct inode *inode;
2022
2023         inode = nfs4_opendata_get_inode(data);
2024         if (IS_ERR(inode))
2025                 return ERR_CAST(inode);
2026         if (data->state != NULL && data->state->inode == inode) {
2027                 state = data->state;
2028                 refcount_inc(&state->count);
2029         } else
2030                 state = nfs4_get_open_state(inode, data->owner);
2031         iput(inode);
2032         if (state == NULL)
2033                 state = ERR_PTR(-ENOMEM);
2034         return state;
2035 }
2036
2037 static struct nfs4_state *
2038 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2039 {
2040         struct nfs4_state *state;
2041
2042         if (!data->rpc_done) {
2043                 state = nfs4_try_open_cached(data);
2044                 trace_nfs4_cached_open(data->state);
2045                 goto out;
2046         }
2047
2048         state = nfs4_opendata_find_nfs4_state(data);
2049         if (IS_ERR(state))
2050                 goto out;
2051
2052         if (data->o_res.delegation_type != 0)
2053                 nfs4_opendata_check_deleg(data, state);
2054         if (!update_open_stateid(state, &data->o_res.stateid,
2055                                 NULL, data->o_arg.fmode)) {
2056                 nfs4_put_open_state(state);
2057                 state = ERR_PTR(-EAGAIN);
2058         }
2059 out:
2060         nfs_release_seqid(data->o_arg.seqid);
2061         return state;
2062 }
2063
2064 static struct nfs4_state *
2065 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2066 {
2067         struct nfs4_state *ret;
2068
2069         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2070                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2071         else
2072                 ret = _nfs4_opendata_to_nfs4_state(data);
2073         nfs4_sequence_free_slot(&data->o_res.seq_res);
2074         return ret;
2075 }
2076
2077 static struct nfs_open_context *
2078 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2079 {
2080         struct nfs_inode *nfsi = NFS_I(state->inode);
2081         struct nfs_open_context *ctx;
2082
2083         rcu_read_lock();
2084         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2085                 if (ctx->state != state)
2086                         continue;
2087                 if ((ctx->mode & mode) != mode)
2088                         continue;
2089                 if (!get_nfs_open_context(ctx))
2090                         continue;
2091                 rcu_read_unlock();
2092                 return ctx;
2093         }
2094         rcu_read_unlock();
2095         return ERR_PTR(-ENOENT);
2096 }
2097
2098 static struct nfs_open_context *
2099 nfs4_state_find_open_context(struct nfs4_state *state)
2100 {
2101         struct nfs_open_context *ctx;
2102
2103         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2104         if (!IS_ERR(ctx))
2105                 return ctx;
2106         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2107         if (!IS_ERR(ctx))
2108                 return ctx;
2109         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2110 }
2111
2112 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2113                 struct nfs4_state *state, enum open_claim_type4 claim)
2114 {
2115         struct nfs4_opendata *opendata;
2116
2117         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2118                         NULL, claim, GFP_NOFS);
2119         if (opendata == NULL)
2120                 return ERR_PTR(-ENOMEM);
2121         opendata->state = state;
2122         refcount_inc(&state->count);
2123         return opendata;
2124 }
2125
2126 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2127                 fmode_t fmode)
2128 {
2129         struct nfs4_state *newstate;
2130         int ret;
2131
2132         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2133                 return 0;
2134         opendata->o_arg.open_flags = 0;
2135         opendata->o_arg.fmode = fmode;
2136         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2137                         NFS_SB(opendata->dentry->d_sb),
2138                         fmode, 0);
2139         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2140         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2141         nfs4_init_opendata_res(opendata);
2142         ret = _nfs4_recover_proc_open(opendata);
2143         if (ret != 0)
2144                 return ret;
2145         newstate = nfs4_opendata_to_nfs4_state(opendata);
2146         if (IS_ERR(newstate))
2147                 return PTR_ERR(newstate);
2148         if (newstate != opendata->state)
2149                 ret = -ESTALE;
2150         nfs4_close_state(newstate, fmode);
2151         return ret;
2152 }
2153
2154 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2155 {
2156         int ret;
2157
2158         /* memory barrier prior to reading state->n_* */
2159         smp_rmb();
2160         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2161         if (ret != 0)
2162                 return ret;
2163         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2164         if (ret != 0)
2165                 return ret;
2166         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2167         if (ret != 0)
2168                 return ret;
2169         /*
2170          * We may have performed cached opens for all three recoveries.
2171          * Check if we need to update the current stateid.
2172          */
2173         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2174             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2175                 write_seqlock(&state->seqlock);
2176                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2177                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2178                 write_sequnlock(&state->seqlock);
2179         }
2180         return 0;
2181 }
2182
2183 /*
2184  * OPEN_RECLAIM:
2185  *      reclaim state on the server after a reboot.
2186  */
2187 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2188 {
2189         struct nfs_delegation *delegation;
2190         struct nfs4_opendata *opendata;
2191         fmode_t delegation_type = 0;
2192         int status;
2193
2194         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2195                         NFS4_OPEN_CLAIM_PREVIOUS);
2196         if (IS_ERR(opendata))
2197                 return PTR_ERR(opendata);
2198         rcu_read_lock();
2199         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2200         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2201                 delegation_type = delegation->type;
2202         rcu_read_unlock();
2203         opendata->o_arg.u.delegation_type = delegation_type;
2204         status = nfs4_open_recover(opendata, state);
2205         nfs4_opendata_put(opendata);
2206         return status;
2207 }
2208
2209 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2210 {
2211         struct nfs_server *server = NFS_SERVER(state->inode);
2212         struct nfs4_exception exception = { };
2213         int err;
2214         do {
2215                 err = _nfs4_do_open_reclaim(ctx, state);
2216                 trace_nfs4_open_reclaim(ctx, 0, err);
2217                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2218                         continue;
2219                 if (err != -NFS4ERR_DELAY)
2220                         break;
2221                 nfs4_handle_exception(server, err, &exception);
2222         } while (exception.retry);
2223         return err;
2224 }
2225
2226 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2227 {
2228         struct nfs_open_context *ctx;
2229         int ret;
2230
2231         ctx = nfs4_state_find_open_context(state);
2232         if (IS_ERR(ctx))
2233                 return -EAGAIN;
2234         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2235         nfs_state_clear_open_state_flags(state);
2236         ret = nfs4_do_open_reclaim(ctx, state);
2237         put_nfs_open_context(ctx);
2238         return ret;
2239 }
2240
2241 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)
2242 {
2243         switch (err) {
2244                 default:
2245                         printk(KERN_ERR "NFS: %s: unhandled error "
2246                                         "%d.\n", __func__, err);
2247                         fallthrough;
2248                 case 0:
2249                 case -ENOENT:
2250                 case -EAGAIN:
2251                 case -ESTALE:
2252                 case -ETIMEDOUT:
2253                         break;
2254                 case -NFS4ERR_BADSESSION:
2255                 case -NFS4ERR_BADSLOT:
2256                 case -NFS4ERR_BAD_HIGH_SLOT:
2257                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2258                 case -NFS4ERR_DEADSESSION:
2259                         return -EAGAIN;
2260                 case -NFS4ERR_STALE_CLIENTID:
2261                 case -NFS4ERR_STALE_STATEID:
2262                         /* Don't recall a delegation if it was lost */
2263                         nfs4_schedule_lease_recovery(server->nfs_client);
2264                         return -EAGAIN;
2265                 case -NFS4ERR_MOVED:
2266                         nfs4_schedule_migration_recovery(server);
2267                         return -EAGAIN;
2268                 case -NFS4ERR_LEASE_MOVED:
2269                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2270                         return -EAGAIN;
2271                 case -NFS4ERR_DELEG_REVOKED:
2272                 case -NFS4ERR_ADMIN_REVOKED:
2273                 case -NFS4ERR_EXPIRED:
2274                 case -NFS4ERR_BAD_STATEID:
2275                 case -NFS4ERR_OPENMODE:
2276                         nfs_inode_find_state_and_recover(state->inode,
2277                                         stateid);
2278                         nfs4_schedule_stateid_recovery(server, state);
2279                         return -EAGAIN;
2280                 case -NFS4ERR_DELAY:
2281                 case -NFS4ERR_GRACE:
2282                         ssleep(1);
2283                         return -EAGAIN;
2284                 case -ENOMEM:
2285                 case -NFS4ERR_DENIED:
2286                         if (fl) {
2287                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2288                                 if (lsp)
2289                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2290                         }
2291                         return 0;
2292         }
2293         return err;
2294 }
2295
2296 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2297                 struct nfs4_state *state, const nfs4_stateid *stateid)
2298 {
2299         struct nfs_server *server = NFS_SERVER(state->inode);
2300         struct nfs4_opendata *opendata;
2301         int err = 0;
2302
2303         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2304                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2305         if (IS_ERR(opendata))
2306                 return PTR_ERR(opendata);
2307         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2308         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2309                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2310                 if (err)
2311                         goto out;
2312         }
2313         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2314                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2315                 if (err)
2316                         goto out;
2317         }
2318         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2319                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2320                 if (err)
2321                         goto out;
2322         }
2323         nfs_state_clear_delegation(state);
2324 out:
2325         nfs4_opendata_put(opendata);
2326         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2327 }
2328
2329 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2330 {
2331         struct nfs4_opendata *data = calldata;
2332
2333         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2334                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2335 }
2336
2337 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2338 {
2339         struct nfs4_opendata *data = calldata;
2340
2341         nfs40_sequence_done(task, &data->c_res.seq_res);
2342
2343         data->rpc_status = task->tk_status;
2344         if (data->rpc_status == 0) {
2345                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2346                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2347                 renew_lease(data->o_res.server, data->timestamp);
2348                 data->rpc_done = true;
2349         }
2350 }
2351
2352 static void nfs4_open_confirm_release(void *calldata)
2353 {
2354         struct nfs4_opendata *data = calldata;
2355         struct nfs4_state *state = NULL;
2356
2357         /* If this request hasn't been cancelled, do nothing */
2358         if (!data->cancelled)
2359                 goto out_free;
2360         /* In case of error, no cleanup! */
2361         if (!data->rpc_done)
2362                 goto out_free;
2363         state = nfs4_opendata_to_nfs4_state(data);
2364         if (!IS_ERR(state))
2365                 nfs4_close_state(state, data->o_arg.fmode);
2366 out_free:
2367         nfs4_opendata_put(data);
2368 }
2369
2370 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2371         .rpc_call_prepare = nfs4_open_confirm_prepare,
2372         .rpc_call_done = nfs4_open_confirm_done,
2373         .rpc_release = nfs4_open_confirm_release,
2374 };
2375
2376 /*
2377  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2378  */
2379 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2380 {
2381         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2382         struct rpc_task *task;
2383         struct  rpc_message msg = {
2384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2385                 .rpc_argp = &data->c_arg,
2386                 .rpc_resp = &data->c_res,
2387                 .rpc_cred = data->owner->so_cred,
2388         };
2389         struct rpc_task_setup task_setup_data = {
2390                 .rpc_client = server->client,
2391                 .rpc_message = &msg,
2392                 .callback_ops = &nfs4_open_confirm_ops,
2393                 .callback_data = data,
2394                 .workqueue = nfsiod_workqueue,
2395                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2396         };
2397         int status;
2398
2399         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2400                                 data->is_recover);
2401         kref_get(&data->kref);
2402         data->rpc_done = false;
2403         data->rpc_status = 0;
2404         data->timestamp = jiffies;
2405         task = rpc_run_task(&task_setup_data);
2406         if (IS_ERR(task))
2407                 return PTR_ERR(task);
2408         status = rpc_wait_for_completion_task(task);
2409         if (status != 0) {
2410                 data->cancelled = true;
2411                 smp_wmb();
2412         } else
2413                 status = data->rpc_status;
2414         rpc_put_task(task);
2415         return status;
2416 }
2417
2418 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2419 {
2420         struct nfs4_opendata *data = calldata;
2421         struct nfs4_state_owner *sp = data->owner;
2422         struct nfs_client *clp = sp->so_server->nfs_client;
2423         enum open_claim_type4 claim = data->o_arg.claim;
2424
2425         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2426                 goto out_wait;
2427         /*
2428          * Check if we still need to send an OPEN call, or if we can use
2429          * a delegation instead.
2430          */
2431         if (data->state != NULL) {
2432                 struct nfs_delegation *delegation;
2433
2434                 if (can_open_cached(data->state, data->o_arg.fmode,
2435                                         data->o_arg.open_flags, claim))
2436                         goto out_no_action;
2437                 rcu_read_lock();
2438                 delegation = nfs4_get_valid_delegation(data->state->inode);
2439                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2440                         goto unlock_no_action;
2441                 rcu_read_unlock();
2442         }
2443         /* Update client id. */
2444         data->o_arg.clientid = clp->cl_clientid;
2445         switch (claim) {
2446         default:
2447                 break;
2448         case NFS4_OPEN_CLAIM_PREVIOUS:
2449         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2450         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2451                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2452                 fallthrough;
2453         case NFS4_OPEN_CLAIM_FH:
2454                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2455         }
2456         data->timestamp = jiffies;
2457         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2458                                 &data->o_arg.seq_args,
2459                                 &data->o_res.seq_res,
2460                                 task) != 0)
2461                 nfs_release_seqid(data->o_arg.seqid);
2462
2463         /* Set the create mode (note dependency on the session type) */
2464         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2465         if (data->o_arg.open_flags & O_EXCL) {
2466                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2467                 if (clp->cl_mvops->minor_version == 0) {
2468                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2469                         /* don't put an ACCESS op in OPEN compound if O_EXCL,
2470                          * because ACCESS will return permission denied for
2471                          * all bits until close */
2472                         data->o_res.access_request = data->o_arg.access = 0;
2473                 } else if (nfs4_has_persistent_session(clp))
2474                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2475         }
2476         return;
2477 unlock_no_action:
2478         trace_nfs4_cached_open(data->state);
2479         rcu_read_unlock();
2480 out_no_action:
2481         task->tk_action = NULL;
2482 out_wait:
2483         nfs4_sequence_done(task, &data->o_res.seq_res);
2484 }
2485
2486 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2487 {
2488         struct nfs4_opendata *data = calldata;
2489
2490         data->rpc_status = task->tk_status;
2491
2492         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2493                 return;
2494
2495         if (task->tk_status == 0) {
2496                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2497                         switch (data->o_res.f_attr->mode & S_IFMT) {
2498                         case S_IFREG:
2499                                 break;
2500                         case S_IFLNK:
2501                                 data->rpc_status = -ELOOP;
2502                                 break;
2503                         case S_IFDIR:
2504                                 data->rpc_status = -EISDIR;
2505                                 break;
2506                         default:
2507                                 data->rpc_status = -ENOTDIR;
2508                         }
2509                 }
2510                 renew_lease(data->o_res.server, data->timestamp);
2511                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2512                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2513         }
2514         data->rpc_done = true;
2515 }
2516
2517 static void nfs4_open_release(void *calldata)
2518 {
2519         struct nfs4_opendata *data = calldata;
2520         struct nfs4_state *state = NULL;
2521
2522         /* If this request hasn't been cancelled, do nothing */
2523         if (!data->cancelled)
2524                 goto out_free;
2525         /* In case of error, no cleanup! */
2526         if (data->rpc_status != 0 || !data->rpc_done)
2527                 goto out_free;
2528         /* In case we need an open_confirm, no cleanup! */
2529         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2530                 goto out_free;
2531         state = nfs4_opendata_to_nfs4_state(data);
2532         if (!IS_ERR(state))
2533                 nfs4_close_state(state, data->o_arg.fmode);
2534 out_free:
2535         nfs4_opendata_put(data);
2536 }
2537
2538 static const struct rpc_call_ops nfs4_open_ops = {
2539         .rpc_call_prepare = nfs4_open_prepare,
2540         .rpc_call_done = nfs4_open_done,
2541         .rpc_release = nfs4_open_release,
2542 };
2543
2544 static int nfs4_run_open_task(struct nfs4_opendata *data,
2545                               struct nfs_open_context *ctx)
2546 {
2547         struct inode *dir = d_inode(data->dir);
2548         struct nfs_server *server = NFS_SERVER(dir);
2549         struct nfs_openargs *o_arg = &data->o_arg;
2550         struct nfs_openres *o_res = &data->o_res;
2551         struct rpc_task *task;
2552         struct rpc_message msg = {
2553                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2554                 .rpc_argp = o_arg,
2555                 .rpc_resp = o_res,
2556                 .rpc_cred = data->owner->so_cred,
2557         };
2558         struct rpc_task_setup task_setup_data = {
2559                 .rpc_client = server->client,
2560                 .rpc_message = &msg,
2561                 .callback_ops = &nfs4_open_ops,
2562                 .callback_data = data,
2563                 .workqueue = nfsiod_workqueue,
2564                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2565         };
2566         int status;
2567
2568         if (server->nfs_client->cl_minorversion)
2569                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
2570
2571         kref_get(&data->kref);
2572         data->rpc_done = false;
2573         data->rpc_status = 0;
2574         data->cancelled = false;
2575         data->is_recover = false;
2576         if (!ctx) {
2577                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2578                 data->is_recover = true;
2579                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2580         } else {
2581                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2582                 pnfs_lgopen_prepare(data, ctx);
2583         }
2584         task = rpc_run_task(&task_setup_data);
2585         if (IS_ERR(task))
2586                 return PTR_ERR(task);
2587         status = rpc_wait_for_completion_task(task);
2588         if (status != 0) {
2589                 data->cancelled = true;
2590                 smp_wmb();
2591         } else
2592                 status = data->rpc_status;
2593         rpc_put_task(task);
2594
2595         return status;
2596 }
2597
2598 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2599 {
2600         struct inode *dir = d_inode(data->dir);
2601         struct nfs_openres *o_res = &data->o_res;
2602         int status;
2603
2604         status = nfs4_run_open_task(data, NULL);
2605         if (status != 0 || !data->rpc_done)
2606                 return status;
2607
2608         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2609
2610         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2611                 status = _nfs4_proc_open_confirm(data);
2612
2613         return status;
2614 }
2615
2616 /*
2617  * Additional permission checks in order to distinguish between an
2618  * open for read, and an open for execute. This works around the
2619  * fact that NFSv4 OPEN treats read and execute permissions as being
2620  * the same.
2621  * Note that in the non-execute case, we want to turn off permission
2622  * checking if we just created a new file (POSIX open() semantics).
2623  */
2624 static int nfs4_opendata_access(const struct cred *cred,
2625                                 struct nfs4_opendata *opendata,
2626                                 struct nfs4_state *state, fmode_t fmode,
2627                                 int openflags)
2628 {
2629         struct nfs_access_entry cache;
2630         u32 mask, flags;
2631
2632         /* access call failed or for some reason the server doesn't
2633          * support any access modes -- defer access call until later */
2634         if (opendata->o_res.access_supported == 0)
2635                 return 0;
2636
2637         mask = 0;
2638         /*
2639          * Use openflags to check for exec, because fmode won't
2640          * always have FMODE_EXEC set when file open for exec.
2641          */
2642         if (openflags & __FMODE_EXEC) {
2643                 /* ONLY check for exec rights */
2644                 if (S_ISDIR(state->inode->i_mode))
2645                         mask = NFS4_ACCESS_LOOKUP;
2646                 else
2647                         mask = NFS4_ACCESS_EXECUTE;
2648         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2649                 mask = NFS4_ACCESS_READ;
2650
2651         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2652         nfs_access_add_cache(state->inode, &cache, cred);
2653
2654         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2655         if ((mask & ~cache.mask & flags) == 0)
2656                 return 0;
2657
2658         return -EACCES;
2659 }
2660
2661 /*
2662  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2663  */
2664 static int _nfs4_proc_open(struct nfs4_opendata *data,
2665                            struct nfs_open_context *ctx)
2666 {
2667         struct inode *dir = d_inode(data->dir);
2668         struct nfs_server *server = NFS_SERVER(dir);
2669         struct nfs_openargs *o_arg = &data->o_arg;
2670         struct nfs_openres *o_res = &data->o_res;
2671         int status;
2672
2673         status = nfs4_run_open_task(data, ctx);
2674         if (!data->rpc_done)
2675                 return status;
2676         if (status != 0) {
2677                 if (status == -NFS4ERR_BADNAME &&
2678                                 !(o_arg->open_flags & O_CREAT))
2679                         return -ENOENT;
2680                 return status;
2681         }
2682
2683         nfs_fattr_map_and_free_names(server, &data->f_attr);
2684
2685         if (o_arg->open_flags & O_CREAT) {
2686                 if (o_arg->open_flags & O_EXCL)
2687                         data->file_created = true;
2688                 else if (o_res->cinfo.before != o_res->cinfo.after)
2689                         data->file_created = true;
2690                 if (data->file_created ||
2691                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2692                         nfs4_update_changeattr(dir, &o_res->cinfo,
2693                                         o_res->f_attr->time_start,
2694                                         NFS_INO_INVALID_DATA);
2695         }
2696         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2697                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2698         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2699                 status = _nfs4_proc_open_confirm(data);
2700                 if (status != 0)
2701                         return status;
2702         }
2703         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2704                 nfs4_sequence_free_slot(&o_res->seq_res);
2705                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, NULL);
2706         }
2707         return 0;
2708 }
2709
2710 /*
2711  * OPEN_EXPIRED:
2712  *      reclaim state on the server after a network partition.
2713  *      Assumes caller holds the appropriate lock
2714  */
2715 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2716 {
2717         struct nfs4_opendata *opendata;
2718         int ret;
2719
2720         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2721                         NFS4_OPEN_CLAIM_FH);
2722         if (IS_ERR(opendata))
2723                 return PTR_ERR(opendata);
2724         ret = nfs4_open_recover(opendata, state);
2725         if (ret == -ESTALE)
2726                 d_drop(ctx->dentry);
2727         nfs4_opendata_put(opendata);
2728         return ret;
2729 }
2730
2731 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2732 {
2733         struct nfs_server *server = NFS_SERVER(state->inode);
2734         struct nfs4_exception exception = { };
2735         int err;
2736
2737         do {
2738                 err = _nfs4_open_expired(ctx, state);
2739                 trace_nfs4_open_expired(ctx, 0, err);
2740                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2741                         continue;
2742                 switch (err) {
2743                 default:
2744                         goto out;
2745                 case -NFS4ERR_GRACE:
2746                 case -NFS4ERR_DELAY:
2747                         nfs4_handle_exception(server, err, &exception);
2748                         err = 0;
2749                 }
2750         } while (exception.retry);
2751 out:
2752         return err;
2753 }
2754
2755 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2756 {
2757         struct nfs_open_context *ctx;
2758         int ret;
2759
2760         ctx = nfs4_state_find_open_context(state);
2761         if (IS_ERR(ctx))
2762                 return -EAGAIN;
2763         ret = nfs4_do_open_expired(ctx, state);
2764         put_nfs_open_context(ctx);
2765         return ret;
2766 }
2767
2768 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2769                 const nfs4_stateid *stateid)
2770 {
2771         nfs_remove_bad_delegation(state->inode, stateid);
2772         nfs_state_clear_delegation(state);
2773 }
2774
2775 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2776 {
2777         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2778                 nfs_finish_clear_delegation_stateid(state, NULL);
2779 }
2780
2781 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2782 {
2783         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2784         nfs40_clear_delegation_stateid(state);
2785         nfs_state_clear_open_state_flags(state);
2786         return nfs4_open_expired(sp, state);
2787 }
2788
2789 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2790                 nfs4_stateid *stateid,
2791                 const struct cred *cred)
2792 {
2793         return -NFS4ERR_BAD_STATEID;
2794 }
2795
2796 #if defined(CONFIG_NFS_V4_1)
2797 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2798                 nfs4_stateid *stateid,
2799                 const struct cred *cred)
2800 {
2801         int status;
2802
2803         switch (stateid->type) {
2804         default:
2805                 break;
2806         case NFS4_INVALID_STATEID_TYPE:
2807         case NFS4_SPECIAL_STATEID_TYPE:
2808                 return -NFS4ERR_BAD_STATEID;
2809         case NFS4_REVOKED_STATEID_TYPE:
2810                 goto out_free;
2811         }
2812
2813         status = nfs41_test_stateid(server, stateid, cred);
2814         switch (status) {
2815         case -NFS4ERR_EXPIRED:
2816         case -NFS4ERR_ADMIN_REVOKED:
2817         case -NFS4ERR_DELEG_REVOKED:
2818                 break;
2819         default:
2820                 return status;
2821         }
2822 out_free:
2823         /* Ack the revoked state to the server */
2824         nfs41_free_stateid(server, stateid, cred, true);
2825         return -NFS4ERR_EXPIRED;
2826 }
2827
2828 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2829 {
2830         struct nfs_server *server = NFS_SERVER(state->inode);
2831         nfs4_stateid stateid;
2832         struct nfs_delegation *delegation;
2833         const struct cred *cred = NULL;
2834         int status, ret = NFS_OK;
2835
2836         /* Get the delegation credential for use by test/free_stateid */
2837         rcu_read_lock();
2838         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2839         if (delegation == NULL) {
2840                 rcu_read_unlock();
2841                 nfs_state_clear_delegation(state);
2842                 return NFS_OK;
2843         }
2844
2845         spin_lock(&delegation->lock);
2846         nfs4_stateid_copy(&stateid, &delegation->stateid);
2847
2848         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2849                                 &delegation->flags)) {
2850                 spin_unlock(&delegation->lock);
2851                 rcu_read_unlock();
2852                 return NFS_OK;
2853         }
2854
2855         if (delegation->cred)
2856                 cred = get_cred(delegation->cred);
2857         spin_unlock(&delegation->lock);
2858         rcu_read_unlock();
2859         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2860         trace_nfs4_test_delegation_stateid(state, NULL, status);
2861         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2862                 nfs_finish_clear_delegation_stateid(state, &stateid);
2863         else
2864                 ret = status;
2865
2866         put_cred(cred);
2867         return ret;
2868 }
2869
2870 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2871 {
2872         nfs4_stateid tmp;
2873
2874         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2875             nfs4_copy_delegation_stateid(state->inode, state->state,
2876                                 &tmp, NULL) &&
2877             nfs4_stateid_match_other(&state->stateid, &tmp))
2878                 nfs_state_set_delegation(state, &tmp, state->state);
2879         else
2880                 nfs_state_clear_delegation(state);
2881 }
2882
2883 /**
2884  * nfs41_check_expired_locks - possibly free a lock stateid
2885  *
2886  * @state: NFSv4 state for an inode
2887  *
2888  * Returns NFS_OK if recovery for this stateid is now finished.
2889  * Otherwise a negative NFS4ERR value is returned.
2890  */
2891 static int nfs41_check_expired_locks(struct nfs4_state *state)
2892 {
2893         int status, ret = NFS_OK;
2894         struct nfs4_lock_state *lsp, *prev = NULL;
2895         struct nfs_server *server = NFS_SERVER(state->inode);
2896
2897         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2898                 goto out;
2899
2900         spin_lock(&state->state_lock);
2901         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2902                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2903                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2904
2905                         refcount_inc(&lsp->ls_count);
2906                         spin_unlock(&state->state_lock);
2907
2908                         nfs4_put_lock_state(prev);
2909                         prev = lsp;
2910
2911                         status = nfs41_test_and_free_expired_stateid(server,
2912                                         &lsp->ls_stateid,
2913                                         cred);
2914                         trace_nfs4_test_lock_stateid(state, lsp, status);
2915                         if (status == -NFS4ERR_EXPIRED ||
2916                             status == -NFS4ERR_BAD_STATEID) {
2917                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2918                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2919                                 if (!recover_lost_locks)
2920                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2921                         } else if (status != NFS_OK) {
2922                                 ret = status;
2923                                 nfs4_put_lock_state(prev);
2924                                 goto out;
2925                         }
2926                         spin_lock(&state->state_lock);
2927                 }
2928         }
2929         spin_unlock(&state->state_lock);
2930         nfs4_put_lock_state(prev);
2931 out:
2932         return ret;
2933 }
2934
2935 /**
2936  * nfs41_check_open_stateid - possibly free an open stateid
2937  *
2938  * @state: NFSv4 state for an inode
2939  *
2940  * Returns NFS_OK if recovery for this stateid is now finished.
2941  * Otherwise a negative NFS4ERR value is returned.
2942  */
2943 static int nfs41_check_open_stateid(struct nfs4_state *state)
2944 {
2945         struct nfs_server *server = NFS_SERVER(state->inode);
2946         nfs4_stateid *stateid = &state->open_stateid;
2947         const struct cred *cred = state->owner->so_cred;
2948         int status;
2949
2950         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2951                 return -NFS4ERR_BAD_STATEID;
2952         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2953         trace_nfs4_test_open_stateid(state, NULL, status);
2954         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2955                 nfs_state_clear_open_state_flags(state);
2956                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2957                 return status;
2958         }
2959         if (nfs_open_stateid_recover_openmode(state))
2960                 return -NFS4ERR_OPENMODE;
2961         return NFS_OK;
2962 }
2963
2964 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2965 {
2966         int status;
2967
2968         status = nfs41_check_delegation_stateid(state);
2969         if (status != NFS_OK)
2970                 return status;
2971         nfs41_delegation_recover_stateid(state);
2972
2973         status = nfs41_check_expired_locks(state);
2974         if (status != NFS_OK)
2975                 return status;
2976         status = nfs41_check_open_stateid(state);
2977         if (status != NFS_OK)
2978                 status = nfs4_open_expired(sp, state);
2979         return status;
2980 }
2981 #endif
2982
2983 /*
2984  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2985  * fields corresponding to attributes that were used to store the verifier.
2986  * Make sure we clobber those fields in the later setattr call
2987  */
2988 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2989                                 struct iattr *sattr, struct nfs4_label **label)
2990 {
2991         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2992         __u32 attrset[3];
2993         unsigned ret;
2994         unsigned i;
2995
2996         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2997                 attrset[i] = opendata->o_res.attrset[i];
2998                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
2999                         attrset[i] &= ~bitmask[i];
3000         }
3001
3002         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3003                 sattr->ia_valid : 0;
3004
3005         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3006                 if (sattr->ia_valid & ATTR_ATIME_SET)
3007                         ret |= ATTR_ATIME_SET;
3008                 else
3009                         ret |= ATTR_ATIME;
3010         }
3011
3012         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3013                 if (sattr->ia_valid & ATTR_MTIME_SET)
3014                         ret |= ATTR_MTIME_SET;
3015                 else
3016                         ret |= ATTR_MTIME;
3017         }
3018
3019         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3020                 *label = NULL;
3021         return ret;
3022 }
3023
3024 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3025                 int flags, struct nfs_open_context *ctx)
3026 {
3027         struct nfs4_state_owner *sp = opendata->owner;
3028         struct nfs_server *server = sp->so_server;
3029         struct dentry *dentry;
3030         struct nfs4_state *state;
3031         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3032         struct inode *dir = d_inode(opendata->dir);
3033         unsigned long dir_verifier;
3034         unsigned int seq;
3035         int ret;
3036
3037         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3038         dir_verifier = nfs_save_change_attribute(dir);
3039
3040         ret = _nfs4_proc_open(opendata, ctx);
3041         if (ret != 0)
3042                 goto out;
3043
3044         state = _nfs4_opendata_to_nfs4_state(opendata);
3045         ret = PTR_ERR(state);
3046         if (IS_ERR(state))
3047                 goto out;
3048         ctx->state = state;
3049         if (server->caps & NFS_CAP_POSIX_LOCK)
3050                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3051         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3052                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3053
3054         dentry = opendata->dentry;
3055         if (d_really_is_negative(dentry)) {
3056                 struct dentry *alias;
3057                 d_drop(dentry);
3058                 alias = d_exact_alias(dentry, state->inode);
3059                 if (!alias)
3060                         alias = d_splice_alias(igrab(state->inode), dentry);
3061                 /* d_splice_alias() can't fail here - it's a non-directory */
3062                 if (alias) {
3063                         dput(ctx->dentry);
3064                         ctx->dentry = dentry = alias;
3065                 }
3066         }
3067
3068         switch(opendata->o_arg.claim) {
3069         default:
3070                 break;
3071         case NFS4_OPEN_CLAIM_NULL:
3072         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3073         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3074                 if (!opendata->rpc_done)
3075                         break;
3076                 if (opendata->o_res.delegation_type != 0)
3077                         dir_verifier = nfs_save_change_attribute(dir);
3078                 nfs_set_verifier(dentry, dir_verifier);
3079         }
3080
3081         /* Parse layoutget results before we check for access */
3082         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3083
3084         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3085                         acc_mode, flags);
3086         if (ret != 0)
3087                 goto out;
3088
3089         if (d_inode(dentry) == state->inode) {
3090                 nfs_inode_attach_open_context(ctx);
3091                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3092                         nfs4_schedule_stateid_recovery(server, state);
3093         }
3094
3095 out:
3096         if (!opendata->cancelled)
3097                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3098         return ret;
3099 }
3100
3101 /*
3102  * Returns a referenced nfs4_state
3103  */
3104 static int _nfs4_do_open(struct inode *dir,
3105                         struct nfs_open_context *ctx,
3106                         int flags,
3107                         const struct nfs4_open_createattrs *c,
3108                         int *opened)
3109 {
3110         struct nfs4_state_owner  *sp;
3111         struct nfs4_state     *state = NULL;
3112         struct nfs_server       *server = NFS_SERVER(dir);
3113         struct nfs4_opendata *opendata;
3114         struct dentry *dentry = ctx->dentry;
3115         const struct cred *cred = ctx->cred;
3116         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3117         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3118         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3119         struct iattr *sattr = c->sattr;
3120         struct nfs4_label *label = c->label;
3121         int status;
3122
3123         /* Protect against reboot recovery conflicts */
3124         status = -ENOMEM;
3125         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3126         if (sp == NULL) {
3127                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3128                 goto out_err;
3129         }
3130         status = nfs4_client_recover_expired_lease(server->nfs_client);
3131         if (status != 0)
3132                 goto err_put_state_owner;
3133         if (d_really_is_positive(dentry))
3134                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3135         status = -ENOMEM;
3136         if (d_really_is_positive(dentry))
3137                 claim = NFS4_OPEN_CLAIM_FH;
3138         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3139                         c, claim, GFP_KERNEL);
3140         if (opendata == NULL)
3141                 goto err_put_state_owner;
3142
3143         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3144                 if (!opendata->f_attr.mdsthreshold) {
3145                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3146                         if (!opendata->f_attr.mdsthreshold)
3147                                 goto err_opendata_put;
3148                 }
3149                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3150         }
3151         if (d_really_is_positive(dentry))
3152                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3153
3154         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3155         if (status != 0)
3156                 goto err_opendata_put;
3157         state = ctx->state;
3158
3159         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3160             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3161                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3162                 /*
3163                  * send create attributes which was not set by open
3164                  * with an extra setattr.
3165                  */
3166                 if (attrs || label) {
3167                         unsigned ia_old = sattr->ia_valid;
3168
3169                         sattr->ia_valid = attrs;
3170                         nfs_fattr_init(opendata->o_res.f_attr);
3171                         status = nfs4_do_setattr(state->inode, cred,
3172                                         opendata->o_res.f_attr, sattr,
3173                                         ctx, label);
3174                         if (status == 0) {
3175                                 nfs_setattr_update_inode(state->inode, sattr,
3176                                                 opendata->o_res.f_attr);
3177                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3178                         }
3179                         sattr->ia_valid = ia_old;
3180                 }
3181         }
3182         if (opened && opendata->file_created)
3183                 *opened = 1;
3184
3185         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3186                 *ctx_th = opendata->f_attr.mdsthreshold;
3187                 opendata->f_attr.mdsthreshold = NULL;
3188         }
3189
3190         nfs4_opendata_put(opendata);
3191         nfs4_put_state_owner(sp);
3192         return 0;
3193 err_opendata_put:
3194         nfs4_opendata_put(opendata);
3195 err_put_state_owner:
3196         nfs4_put_state_owner(sp);
3197 out_err:
3198         return status;
3199 }
3200
3201
3202 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3203                                         struct nfs_open_context *ctx,
3204                                         int flags,
3205                                         struct iattr *sattr,
3206                                         struct nfs4_label *label,
3207                                         int *opened)
3208 {
3209         struct nfs_server *server = NFS_SERVER(dir);
3210         struct nfs4_exception exception = {
3211                 .interruptible = true,
3212         };
3213         struct nfs4_state *res;
3214         struct nfs4_open_createattrs c = {
3215                 .label = label,
3216                 .sattr = sattr,
3217                 .verf = {
3218                         [0] = (__u32)jiffies,
3219                         [1] = (__u32)current->pid,
3220                 },
3221         };
3222         int status;
3223
3224         do {
3225                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3226                 res = ctx->state;
3227                 trace_nfs4_open_file(ctx, flags, status);
3228                 if (status == 0)
3229                         break;
3230                 /* NOTE: BAD_SEQID means the server and client disagree about the
3231                  * book-keeping w.r.t. state-changing operations
3232                  * (OPEN/CLOSE/LOCK/LOCKU...)
3233                  * It is actually a sign of a bug on the client or on the server.
3234                  *
3235                  * If we receive a BAD_SEQID error in the particular case of
3236                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3237                  * have unhashed the old state_owner for us, and that we can
3238                  * therefore safely retry using a new one. We should still warn
3239                  * the user though...
3240                  */
3241                 if (status == -NFS4ERR_BAD_SEQID) {
3242                         pr_warn_ratelimited("NFS: v4 server %s "
3243                                         " returned a bad sequence-id error!\n",
3244                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3245                         exception.retry = 1;
3246                         continue;
3247                 }
3248                 /*
3249                  * BAD_STATEID on OPEN means that the server cancelled our
3250                  * state before it received the OPEN_CONFIRM.
3251                  * Recover by retrying the request as per the discussion
3252                  * on Page 181 of RFC3530.
3253                  */
3254                 if (status == -NFS4ERR_BAD_STATEID) {
3255                         exception.retry = 1;
3256                         continue;
3257                 }
3258                 if (status == -NFS4ERR_EXPIRED) {
3259                         nfs4_schedule_lease_recovery(server->nfs_client);
3260                         exception.retry = 1;
3261                         continue;
3262                 }
3263                 if (status == -EAGAIN) {
3264                         /* We must have found a delegation */
3265                         exception.retry = 1;
3266                         continue;
3267                 }
3268                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3269                         continue;
3270                 res = ERR_PTR(nfs4_handle_exception(server,
3271                                         status, &exception));
3272         } while (exception.retry);
3273         return res;
3274 }
3275
3276 static int _nfs4_do_setattr(struct inode *inode,
3277                             struct nfs_setattrargs *arg,
3278                             struct nfs_setattrres *res,
3279                             const struct cred *cred,
3280                             struct nfs_open_context *ctx)
3281 {
3282         struct nfs_server *server = NFS_SERVER(inode);
3283         struct rpc_message msg = {
3284                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3285                 .rpc_argp       = arg,
3286                 .rpc_resp       = res,
3287                 .rpc_cred       = cred,
3288         };
3289         const struct cred *delegation_cred = NULL;
3290         unsigned long timestamp = jiffies;
3291         bool truncate;
3292         int status;
3293
3294         nfs_fattr_init(res->fattr);
3295
3296         /* Servers should only apply open mode checks for file size changes */
3297         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3298         if (!truncate) {
3299                 nfs4_inode_make_writeable(inode);
3300                 goto zero_stateid;
3301         }
3302
3303         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3304                 /* Use that stateid */
3305         } else if (ctx != NULL && ctx->state) {
3306                 struct nfs_lock_context *l_ctx;
3307                 if (!nfs4_valid_open_stateid(ctx->state))
3308                         return -EBADF;
3309                 l_ctx = nfs_get_lock_context(ctx);
3310                 if (IS_ERR(l_ctx))
3311                         return PTR_ERR(l_ctx);
3312                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3313                                                 &arg->stateid, &delegation_cred);
3314                 nfs_put_lock_context(l_ctx);
3315                 if (status == -EIO)
3316                         return -EBADF;
3317                 else if (status == -EAGAIN)
3318                         goto zero_stateid;
3319         } else {
3320 zero_stateid:
3321                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3322         }
3323         if (delegation_cred)
3324                 msg.rpc_cred = delegation_cred;
3325
3326         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3327
3328         put_cred(delegation_cred);
3329         if (status == 0 && ctx != NULL)
3330                 renew_lease(server, timestamp);
3331         trace_nfs4_setattr(inode, &arg->stateid, status);
3332         return status;
3333 }
3334
3335 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3336                            struct nfs_fattr *fattr, struct iattr *sattr,
3337                            struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3338 {
3339         struct nfs_server *server = NFS_SERVER(inode);
3340         __u32 bitmask[NFS4_BITMASK_SZ];
3341         struct nfs4_state *state = ctx ? ctx->state : NULL;
3342         struct nfs_setattrargs  arg = {
3343                 .fh             = NFS_FH(inode),
3344                 .iap            = sattr,
3345                 .server         = server,
3346                 .bitmask = bitmask,
3347                 .label          = ilabel,
3348         };
3349         struct nfs_setattrres  res = {
3350                 .fattr          = fattr,
3351                 .server         = server,
3352         };
3353         struct nfs4_exception exception = {
3354                 .state = state,
3355                 .inode = inode,
3356                 .stateid = &arg.stateid,
3357         };
3358         unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3359         int err;
3360
3361         if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3362                 adjust_flags |= NFS_INO_INVALID_MODE;
3363         if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3364                 adjust_flags |= NFS_INO_INVALID_OTHER;
3365
3366         do {
3367                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3368                                         inode, adjust_flags);
3369
3370                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3371                 switch (err) {
3372                 case -NFS4ERR_OPENMODE:
3373                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3374                                 pr_warn_once("NFSv4: server %s is incorrectly "
3375                                                 "applying open mode checks to "
3376                                                 "a SETATTR that is not "
3377                                                 "changing file size.\n",
3378                                                 server->nfs_client->cl_hostname);
3379                         }
3380                         if (state && !(state->state & FMODE_WRITE)) {
3381                                 err = -EBADF;
3382                                 if (sattr->ia_valid & ATTR_OPEN)
3383                                         err = -EACCES;
3384                                 goto out;
3385                         }
3386                 }
3387                 err = nfs4_handle_exception(server, err, &exception);
3388         } while (exception.retry);
3389 out:
3390         return err;
3391 }
3392
3393 static bool
3394 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3395 {
3396         if (inode == NULL || !nfs_have_layout(inode))
3397                 return false;
3398
3399         return pnfs_wait_on_layoutreturn(inode, task);
3400 }
3401
3402 /*
3403  * Update the seqid of an open stateid
3404  */
3405 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3406                 struct nfs4_state *state)
3407 {
3408         __be32 seqid_open;
3409         u32 dst_seqid;
3410         int seq;
3411
3412         for (;;) {
3413                 if (!nfs4_valid_open_stateid(state))
3414                         break;
3415                 seq = read_seqbegin(&state->seqlock);
3416                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3417                         nfs4_stateid_copy(dst, &state->open_stateid);
3418                         if (read_seqretry(&state->seqlock, seq))
3419                                 continue;
3420                         break;
3421                 }
3422                 seqid_open = state->open_stateid.seqid;
3423                 if (read_seqretry(&state->seqlock, seq))
3424                         continue;
3425
3426                 dst_seqid = be32_to_cpu(dst->seqid);
3427                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3428                         dst->seqid = seqid_open;
3429                 break;
3430         }
3431 }
3432
3433 /*
3434  * Update the seqid of an open stateid after receiving
3435  * NFS4ERR_OLD_STATEID
3436  */
3437 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3438                 struct nfs4_state *state)
3439 {
3440         __be32 seqid_open;
3441         u32 dst_seqid;
3442         bool ret;
3443         int seq, status = -EAGAIN;
3444         DEFINE_WAIT(wait);
3445
3446         for (;;) {
3447                 ret = false;
3448                 if (!nfs4_valid_open_stateid(state))
3449                         break;
3450                 seq = read_seqbegin(&state->seqlock);
3451                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3452                         if (read_seqretry(&state->seqlock, seq))
3453                                 continue;
3454                         break;
3455                 }
3456
3457                 write_seqlock(&state->seqlock);
3458                 seqid_open = state->open_stateid.seqid;
3459
3460                 dst_seqid = be32_to_cpu(dst->seqid);
3461
3462                 /* Did another OPEN bump the state's seqid?  try again: */
3463                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3464                         dst->seqid = seqid_open;
3465                         write_sequnlock(&state->seqlock);
3466                         ret = true;
3467                         break;
3468                 }
3469
3470                 /* server says we're behind but we haven't seen the update yet */
3471                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3472                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3473                 write_sequnlock(&state->seqlock);
3474                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3475
3476                 if (fatal_signal_pending(current))
3477                         status = -EINTR;
3478                 else
3479                         if (schedule_timeout(5*HZ) != 0)
3480                                 status = 0;
3481
3482                 finish_wait(&state->waitq, &wait);
3483
3484                 if (!status)
3485                         continue;
3486                 if (status == -EINTR)
3487                         break;
3488
3489                 /* we slept the whole 5 seconds, we must have lost a seqid */
3490                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3491                 ret = true;
3492                 break;
3493         }
3494
3495         return ret;
3496 }
3497
3498 struct nfs4_closedata {
3499         struct inode *inode;
3500         struct nfs4_state *state;
3501         struct nfs_closeargs arg;
3502         struct nfs_closeres res;
3503         struct {
3504                 struct nfs4_layoutreturn_args arg;
3505                 struct nfs4_layoutreturn_res res;
3506                 struct nfs4_xdr_opaque_data ld_private;
3507                 u32 roc_barrier;
3508                 bool roc;
3509         } lr;
3510         struct nfs_fattr fattr;
3511         unsigned long timestamp;
3512 };
3513
3514 static void nfs4_free_closedata(void *data)
3515 {
3516         struct nfs4_closedata *calldata = data;
3517         struct nfs4_state_owner *sp = calldata->state->owner;
3518         struct super_block *sb = calldata->state->inode->i_sb;
3519
3520         if (calldata->lr.roc)
3521                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3522                                 calldata->res.lr_ret);
3523         nfs4_put_open_state(calldata->state);
3524         nfs_free_seqid(calldata->arg.seqid);
3525         nfs4_put_state_owner(sp);
3526         nfs_sb_deactive(sb);
3527         kfree(calldata);
3528 }
3529
3530 static void nfs4_close_done(struct rpc_task *task, void *data)
3531 {
3532         struct nfs4_closedata *calldata = data;
3533         struct nfs4_state *state = calldata->state;
3534         struct nfs_server *server = NFS_SERVER(calldata->inode);
3535         nfs4_stateid *res_stateid = NULL;
3536         struct nfs4_exception exception = {
3537                 .state = state,
3538                 .inode = calldata->inode,
3539                 .stateid = &calldata->arg.stateid,
3540         };
3541
3542         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3543                 return;
3544         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3545
3546         /* Handle Layoutreturn errors */
3547         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3548                           &calldata->res.lr_ret) == -EAGAIN)
3549                 goto out_restart;
3550
3551         /* hmm. we are done with the inode, and in the process of freeing
3552          * the state_owner. we keep this around to process errors
3553          */
3554         switch (task->tk_status) {
3555                 case 0:
3556                         res_stateid = &calldata->res.stateid;
3557                         renew_lease(server, calldata->timestamp);
3558                         break;
3559                 case -NFS4ERR_ACCESS:
3560                         if (calldata->arg.bitmask != NULL) {
3561                                 calldata->arg.bitmask = NULL;
3562                                 calldata->res.fattr = NULL;
3563                                 goto out_restart;
3564
3565                         }
3566                         break;
3567                 case -NFS4ERR_OLD_STATEID:
3568                         /* Did we race with OPEN? */
3569                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3570                                                 state))
3571                                 goto out_restart;
3572                         goto out_release;
3573                 case -NFS4ERR_ADMIN_REVOKED:
3574                 case -NFS4ERR_STALE_STATEID:
3575                 case -NFS4ERR_EXPIRED:
3576                         nfs4_free_revoked_stateid(server,
3577                                         &calldata->arg.stateid,
3578                                         task->tk_msg.rpc_cred);
3579                         fallthrough;
3580                 case -NFS4ERR_BAD_STATEID:
3581                         if (calldata->arg.fmode == 0)
3582                                 break;
3583                         fallthrough;
3584                 default:
3585                         task->tk_status = nfs4_async_handle_exception(task,
3586                                         server, task->tk_status, &exception);
3587                         if (exception.retry)
3588                                 goto out_restart;
3589         }
3590         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3591                         res_stateid, calldata->arg.fmode);
3592 out_release:
3593         task->tk_status = 0;
3594         nfs_release_seqid(calldata->arg.seqid);
3595         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3596         dprintk("%s: ret = %d\n", __func__, task->tk_status);
3597         return;
3598 out_restart:
3599         task->tk_status = 0;
3600         rpc_restart_call_prepare(task);
3601         goto out_release;
3602 }
3603
3604 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3605 {
3606         struct nfs4_closedata *calldata = data;
3607         struct nfs4_state *state = calldata->state;
3608         struct inode *inode = calldata->inode;
3609         struct nfs_server *server = NFS_SERVER(inode);
3610         struct pnfs_layout_hdr *lo;
3611         bool is_rdonly, is_wronly, is_rdwr;
3612         int call_close = 0;
3613
3614         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3615                 goto out_wait;
3616
3617         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3618         spin_lock(&state->owner->so_lock);
3619         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3620         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3621         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3622         /* Calculate the change in open mode */
3623         calldata->arg.fmode = 0;
3624         if (state->n_rdwr == 0) {
3625                 if (state->n_rdonly == 0)
3626                         call_close |= is_rdonly;
3627                 else if (is_rdonly)
3628                         calldata->arg.fmode |= FMODE_READ;
3629                 if (state->n_wronly == 0)
3630                         call_close |= is_wronly;
3631                 else if (is_wronly)
3632                         calldata->arg.fmode |= FMODE_WRITE;
3633                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3634                         call_close |= is_rdwr;
3635         } else if (is_rdwr)
3636                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3637
3638         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3639         if (!nfs4_valid_open_stateid(state))
3640                 call_close = 0;
3641         spin_unlock(&state->owner->so_lock);
3642
3643         if (!call_close) {
3644                 /* Note: exit _without_ calling nfs4_close_done */
3645                 goto out_no_action;
3646         }
3647
3648         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3649                 nfs_release_seqid(calldata->arg.seqid);
3650                 goto out_wait;
3651         }
3652
3653         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3654         if (lo && !pnfs_layout_is_valid(lo)) {
3655                 calldata->arg.lr_args = NULL;
3656                 calldata->res.lr_res = NULL;
3657         }
3658
3659         if (calldata->arg.fmode == 0)
3660                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3661
3662         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3663                 /* Close-to-open cache consistency revalidation */
3664                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3665                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3666                                          server->cache_consistency_bitmask,
3667                                          inode, 0);
3668                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3669                 } else
3670                         calldata->arg.bitmask = NULL;
3671         }
3672
3673         calldata->arg.share_access =
3674                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3675                                 calldata->arg.fmode, 0);
3676
3677         if (calldata->res.fattr == NULL)
3678                 calldata->arg.bitmask = NULL;
3679         else if (calldata->arg.bitmask == NULL)
3680                 calldata->res.fattr = NULL;
3681         calldata->timestamp = jiffies;
3682         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3683                                 &calldata->arg.seq_args,
3684                                 &calldata->res.seq_res,
3685                                 task) != 0)
3686                 nfs_release_seqid(calldata->arg.seqid);
3687         return;
3688 out_no_action:
3689         task->tk_action = NULL;
3690 out_wait:
3691         nfs4_sequence_done(task, &calldata->res.seq_res);
3692 }
3693
3694 static const struct rpc_call_ops nfs4_close_ops = {
3695         .rpc_call_prepare = nfs4_close_prepare,
3696         .rpc_call_done = nfs4_close_done,
3697         .rpc_release = nfs4_free_closedata,
3698 };
3699
3700 /*
3701  * It is possible for data to be read/written from a mem-mapped file
3702  * after the sys_close call (which hits the vfs layer as a flush).
3703  * This means that we can't safely call nfsv4 close on a file until
3704  * the inode is cleared. This in turn means that we are not good
3705  * NFSv4 citizens - we do not indicate to the server to update the file's
3706  * share state even when we are done with one of the three share
3707  * stateid's in the inode.
3708  *
3709  * NOTE: Caller must be holding the sp->so_owner semaphore!
3710  */
3711 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3712 {
3713         struct nfs_server *server = NFS_SERVER(state->inode);
3714         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3715         struct nfs4_closedata *calldata;
3716         struct nfs4_state_owner *sp = state->owner;
3717         struct rpc_task *task;
3718         struct rpc_message msg = {
3719                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3720                 .rpc_cred = state->owner->so_cred,
3721         };
3722         struct rpc_task_setup task_setup_data = {
3723                 .rpc_client = server->client,
3724                 .rpc_message = &msg,
3725                 .callback_ops = &nfs4_close_ops,
3726                 .workqueue = nfsiod_workqueue,
3727                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3728         };
3729         int status = -ENOMEM;
3730
3731         if (server->nfs_client->cl_minorversion)
3732                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
3733
3734         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3735                 &task_setup_data.rpc_client, &msg);
3736
3737         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3738         if (calldata == NULL)
3739                 goto out;
3740         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3741         calldata->inode = state->inode;
3742         calldata->state = state;
3743         calldata->arg.fh = NFS_FH(state->inode);
3744         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3745                 goto out_free_calldata;
3746         /* Serialization for the sequence id */
3747         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3748         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3749         if (IS_ERR(calldata->arg.seqid))
3750                 goto out_free_calldata;
3751         nfs_fattr_init(&calldata->fattr);
3752         calldata->arg.fmode = 0;
3753         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3754         calldata->res.fattr = &calldata->fattr;
3755         calldata->res.seqid = calldata->arg.seqid;
3756         calldata->res.server = server;
3757         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3758         calldata->lr.roc = pnfs_roc(state->inode,
3759                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3760         if (calldata->lr.roc) {
3761                 calldata->arg.lr_args = &calldata->lr.arg;
3762                 calldata->res.lr_res = &calldata->lr.res;
3763         }
3764         nfs_sb_active(calldata->inode->i_sb);
3765
3766         msg.rpc_argp = &calldata->arg;
3767         msg.rpc_resp = &calldata->res;
3768         task_setup_data.callback_data = calldata;
3769         task = rpc_run_task(&task_setup_data);
3770         if (IS_ERR(task))
3771                 return PTR_ERR(task);
3772         status = 0;
3773         if (wait)
3774                 status = rpc_wait_for_completion_task(task);
3775         rpc_put_task(task);
3776         return status;
3777 out_free_calldata:
3778         kfree(calldata);
3779 out:
3780         nfs4_put_open_state(state);
3781         nfs4_put_state_owner(sp);
3782         return status;
3783 }
3784
3785 static struct inode *
3786 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3787                 int open_flags, struct iattr *attr, int *opened)
3788 {
3789         struct nfs4_state *state;
3790         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3791
3792         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3793
3794         /* Protect against concurrent sillydeletes */
3795         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3796
3797         nfs4_label_release_security(label);
3798
3799         if (IS_ERR(state))
3800                 return ERR_CAST(state);
3801         return state->inode;
3802 }
3803
3804 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3805 {
3806         if (ctx->state == NULL)
3807                 return;
3808         if (is_sync)
3809                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3810         else
3811                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3812 }
3813
3814 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3815 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3816 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3817
3818 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3819 {
3820         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3821         struct nfs4_server_caps_arg args = {
3822                 .fhandle = fhandle,
3823                 .bitmask = bitmask,
3824         };
3825         struct nfs4_server_caps_res res = {};
3826         struct rpc_message msg = {
3827                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3828                 .rpc_argp = &args,
3829                 .rpc_resp = &res,
3830         };
3831         int status;
3832         int i;
3833
3834         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3835                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3836                      FATTR4_WORD0_LINK_SUPPORT |
3837                      FATTR4_WORD0_SYMLINK_SUPPORT |
3838                      FATTR4_WORD0_ACLSUPPORT |
3839                      FATTR4_WORD0_CASE_INSENSITIVE |
3840                      FATTR4_WORD0_CASE_PRESERVING;
3841         if (minorversion)
3842                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3843
3844         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3845         if (status == 0) {
3846                 /* Sanity check the server answers */
3847                 switch (minorversion) {
3848                 case 0:
3849                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3850                         res.attr_bitmask[2] = 0;
3851                         break;
3852                 case 1:
3853                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3854                         break;
3855                 case 2:
3856                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3857                 }
3858                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3859                 server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3860                                   NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3861                 server->fattr_valid = NFS_ATTR_FATTR_V4;
3862                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3863                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3864                         server->caps |= NFS_CAP_ACLS;
3865                 if (res.has_links != 0)
3866                         server->caps |= NFS_CAP_HARDLINKS;
3867                 if (res.has_symlinks != 0)
3868                         server->caps |= NFS_CAP_SYMLINKS;
3869                 if (res.case_insensitive)
3870                         server->caps |= NFS_CAP_CASE_INSENSITIVE;
3871                 if (res.case_preserving)
3872                         server->caps |= NFS_CAP_CASE_PRESERVING;
3873 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3874                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3875                         server->caps |= NFS_CAP_SECURITY_LABEL;
3876 #endif
3877                 if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3878                         server->caps |= NFS_CAP_FS_LOCATIONS;
3879                 if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3880                         server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3881                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3882                         server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3883                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3884                         server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3885                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3886                         server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3887                                 NFS_ATTR_FATTR_OWNER_NAME);
3888                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3889                         server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3890                                 NFS_ATTR_FATTR_GROUP_NAME);
3891                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3892                         server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3893                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3894                         server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3895                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3896                         server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3897                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3898                         server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3899                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3900                                 sizeof(server->attr_bitmask));
3901                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3902
3903                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3904                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3905                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3906                 server->cache_consistency_bitmask[2] = 0;
3907
3908                 /* Avoid a regression due to buggy server */
3909                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3910                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3911                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3912                         sizeof(server->exclcreat_bitmask));
3913
3914                 server->acl_bitmask = res.acl_bitmask;
3915                 server->fh_expire_type = res.fh_expire_type;
3916         }
3917
3918         return status;
3919 }
3920
3921 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3922 {
3923         struct nfs4_exception exception = {
3924                 .interruptible = true,
3925         };
3926         int err;
3927
3928         nfs4_server_set_init_caps(server);
3929         do {
3930                 err = nfs4_handle_exception(server,
3931                                 _nfs4_server_capabilities(server, fhandle),
3932                                 &exception);
3933         } while (exception.retry);
3934         return err;
3935 }
3936
3937 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3938                                           struct nfs_client *clp,
3939                                           struct nfs_server *server)
3940 {
3941         int i;
3942
3943         for (i = 0; i < location->nservers; i++) {
3944                 struct nfs4_string *srv_loc = &location->servers[i];
3945                 struct sockaddr addr;
3946                 size_t addrlen;
3947                 struct xprt_create xprt_args = {
3948                         .ident = 0,
3949                         .net = clp->cl_net,
3950                 };
3951                 struct nfs4_add_xprt_data xprtdata = {
3952                         .clp = clp,
3953                 };
3954                 struct rpc_add_xprt_test rpcdata = {
3955                         .add_xprt_test = clp->cl_mvops->session_trunk,
3956                         .data = &xprtdata,
3957                 };
3958                 char *servername = NULL;
3959
3960                 if (!srv_loc->len)
3961                         continue;
3962
3963                 addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
3964                                                 &addr, sizeof(addr),
3965                                                 clp->cl_net, server->port);
3966                 if (!addrlen)
3967                         return;
3968                 xprt_args.dstaddr = &addr;
3969                 xprt_args.addrlen = addrlen;
3970                 servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
3971                 if (!servername)
3972                         return;
3973                 memcpy(servername, srv_loc->data, srv_loc->len);
3974                 servername[srv_loc->len] = '\0';
3975                 xprt_args.servername = servername;
3976
3977                 xprtdata.cred = nfs4_get_clid_cred(clp);
3978                 rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
3979                                   rpc_clnt_setup_test_and_add_xprt,
3980                                   &rpcdata);
3981                 if (xprtdata.cred)
3982                         put_cred(xprtdata.cred);
3983                 kfree(servername);
3984         }
3985 }
3986
3987 static int _nfs4_discover_trunking(struct nfs_server *server,
3988                                    struct nfs_fh *fhandle)
3989 {
3990         struct nfs4_fs_locations *locations = NULL;
3991         struct page *page;
3992         const struct cred *cred;
3993         struct nfs_client *clp = server->nfs_client;
3994         const struct nfs4_state_maintenance_ops *ops =
3995                 clp->cl_mvops->state_renewal_ops;
3996         int status = -ENOMEM, i;
3997
3998         cred = ops->get_state_renewal_cred(clp);
3999         if (cred == NULL) {
4000                 cred = nfs4_get_clid_cred(clp);
4001                 if (cred == NULL)
4002                         return -ENOKEY;
4003         }
4004
4005         page = alloc_page(GFP_KERNEL);
4006         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4007         if (page == NULL || locations == NULL)
4008                 goto out;
4009
4010         status = nfs4_proc_get_locations(server, fhandle, locations, page,
4011                                          cred);
4012         if (status)
4013                 goto out;
4014
4015         for (i = 0; i < locations->nlocations; i++)
4016                 test_fs_location_for_trunking(&locations->locations[i], clp,
4017                                               server);
4018 out:
4019         if (page)
4020                 __free_page(page);
4021         kfree(locations);
4022         return status;
4023 }
4024
4025 static int nfs4_discover_trunking(struct nfs_server *server,
4026                                   struct nfs_fh *fhandle)
4027 {
4028         struct nfs4_exception exception = {
4029                 .interruptible = true,
4030         };
4031         struct nfs_client *clp = server->nfs_client;
4032         int err = 0;
4033
4034         if (!nfs4_has_session(clp))
4035                 goto out;
4036         do {
4037                 err = nfs4_handle_exception(server,
4038                                 _nfs4_discover_trunking(server, fhandle),
4039                                 &exception);
4040         } while (exception.retry);
4041 out:
4042         return err;
4043 }
4044
4045 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4046                 struct nfs_fsinfo *info)
4047 {
4048         u32 bitmask[3];
4049         struct nfs4_lookup_root_arg args = {
4050                 .bitmask = bitmask,
4051         };
4052         struct nfs4_lookup_res res = {
4053                 .server = server,
4054                 .fattr = info->fattr,
4055                 .fh = fhandle,
4056         };
4057         struct rpc_message msg = {
4058                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4059                 .rpc_argp = &args,
4060                 .rpc_resp = &res,
4061         };
4062
4063         bitmask[0] = nfs4_fattr_bitmap[0];
4064         bitmask[1] = nfs4_fattr_bitmap[1];
4065         /*
4066          * Process the label in the upcoming getfattr
4067          */
4068         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4069
4070         nfs_fattr_init(info->fattr);
4071         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4072 }
4073
4074 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4075                 struct nfs_fsinfo *info)
4076 {
4077         struct nfs4_exception exception = {
4078                 .interruptible = true,
4079         };
4080         int err;
4081         do {
4082                 err = _nfs4_lookup_root(server, fhandle, info);
4083                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4084                 switch (err) {
4085                 case 0:
4086                 case -NFS4ERR_WRONGSEC:
4087                         goto out;
4088                 default:
4089                         err = nfs4_handle_exception(server, err, &exception);
4090                 }
4091         } while (exception.retry);
4092 out:
4093         return err;
4094 }
4095
4096 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4097                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4098 {
4099         struct rpc_auth_create_args auth_args = {
4100                 .pseudoflavor = flavor,
4101         };
4102         struct rpc_auth *auth;
4103
4104         auth = rpcauth_create(&auth_args, server->client);
4105         if (IS_ERR(auth))
4106                 return -EACCES;
4107         return nfs4_lookup_root(server, fhandle, info);
4108 }
4109
4110 /*
4111  * Retry pseudoroot lookup with various security flavors.  We do this when:
4112  *
4113  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4114  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4115  *
4116  * Returns zero on success, or a negative NFS4ERR value, or a
4117  * negative errno value.
4118  */
4119 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4120                               struct nfs_fsinfo *info)
4121 {
4122         /* Per 3530bis 15.33.5 */
4123         static const rpc_authflavor_t flav_array[] = {
4124                 RPC_AUTH_GSS_KRB5P,
4125                 RPC_AUTH_GSS_KRB5I,
4126                 RPC_AUTH_GSS_KRB5,
4127                 RPC_AUTH_UNIX,                  /* courtesy */
4128                 RPC_AUTH_NULL,
4129         };
4130         int status = -EPERM;
4131         size_t i;
4132
4133         if (server->auth_info.flavor_len > 0) {
4134                 /* try each flavor specified by user */
4135                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4136                         status = nfs4_lookup_root_sec(server, fhandle, info,
4137                                                 server->auth_info.flavors[i]);
4138                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4139                                 continue;
4140                         break;
4141                 }
4142         } else {
4143                 /* no flavors specified by user, try default list */
4144                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4145                         status = nfs4_lookup_root_sec(server, fhandle, info,
4146                                                       flav_array[i]);
4147                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4148                                 continue;
4149                         break;
4150                 }
4151         }
4152
4153         /*
4154          * -EACCES could mean that the user doesn't have correct permissions
4155          * to access the mount.  It could also mean that we tried to mount
4156          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4157          * existing mount programs don't handle -EACCES very well so it should
4158          * be mapped to -EPERM instead.
4159          */
4160         if (status == -EACCES)
4161                 status = -EPERM;
4162         return status;
4163 }
4164
4165 /**
4166  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4167  * @server: initialized nfs_server handle
4168  * @fhandle: we fill in the pseudo-fs root file handle
4169  * @info: we fill in an FSINFO struct
4170  * @auth_probe: probe the auth flavours
4171  *
4172  * Returns zero on success, or a negative errno.
4173  */
4174 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4175                          struct nfs_fsinfo *info,
4176                          bool auth_probe)
4177 {
4178         int status = 0;
4179
4180         if (!auth_probe)
4181                 status = nfs4_lookup_root(server, fhandle, info);
4182
4183         if (auth_probe || status == NFS4ERR_WRONGSEC)
4184                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4185                                 fhandle, info);
4186
4187         if (status == 0)
4188                 status = nfs4_server_capabilities(server, fhandle);
4189         if (status == 0)
4190                 status = nfs4_do_fsinfo(server, fhandle, info);
4191
4192         return nfs4_map_errors(status);
4193 }
4194
4195 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4196                               struct nfs_fsinfo *info)
4197 {
4198         int error;
4199         struct nfs_fattr *fattr = info->fattr;
4200
4201         error = nfs4_server_capabilities(server, mntfh);
4202         if (error < 0) {
4203                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4204                 return error;
4205         }
4206
4207         error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4208         if (error < 0) {
4209                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4210                 goto out;
4211         }
4212
4213         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4214             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4215                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4216
4217 out:
4218         return error;
4219 }
4220
4221 /*
4222  * Get locations and (maybe) other attributes of a referral.
4223  * Note that we'll actually follow the referral later when
4224  * we detect fsid mismatch in inode revalidation
4225  */
4226 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4227                              const struct qstr *name, struct nfs_fattr *fattr,
4228                              struct nfs_fh *fhandle)
4229 {
4230         int status = -ENOMEM;
4231         struct page *page = NULL;
4232         struct nfs4_fs_locations *locations = NULL;
4233
4234         page = alloc_page(GFP_KERNEL);
4235         if (page == NULL)
4236                 goto out;
4237         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4238         if (locations == NULL)
4239                 goto out;
4240
4241         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4242         if (status != 0)
4243                 goto out;
4244
4245         /*
4246          * If the fsid didn't change, this is a migration event, not a
4247          * referral.  Cause us to drop into the exception handler, which
4248          * will kick off migration recovery.
4249          */
4250         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4251                 dprintk("%s: server did not return a different fsid for"
4252                         " a referral at %s\n", __func__, name->name);
4253                 status = -NFS4ERR_MOVED;
4254                 goto out;
4255         }
4256         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4257         nfs_fixup_referral_attributes(&locations->fattr);
4258
4259         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4260         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4261         memset(fhandle, 0, sizeof(struct nfs_fh));
4262 out:
4263         if (page)
4264                 __free_page(page);
4265         kfree(locations);
4266         return status;
4267 }
4268
4269 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4270                                 struct nfs_fattr *fattr, struct inode *inode)
4271 {
4272         __u32 bitmask[NFS4_BITMASK_SZ];
4273         struct nfs4_getattr_arg args = {
4274                 .fh = fhandle,
4275                 .bitmask = bitmask,
4276         };
4277         struct nfs4_getattr_res res = {
4278                 .fattr = fattr,
4279                 .server = server,
4280         };
4281         struct rpc_message msg = {
4282                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4283                 .rpc_argp = &args,
4284                 .rpc_resp = &res,
4285         };
4286         unsigned short task_flags = 0;
4287
4288         if (nfs4_has_session(server->nfs_client))
4289                 task_flags = RPC_TASK_MOVEABLE;
4290
4291         /* Is this is an attribute revalidation, subject to softreval? */
4292         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4293                 task_flags |= RPC_TASK_TIMEOUT;
4294
4295         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4296         nfs_fattr_init(fattr);
4297         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4298         return nfs4_do_call_sync(server->client, server, &msg,
4299                         &args.seq_args, &res.seq_res, task_flags);
4300 }
4301
4302 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4303                                 struct nfs_fattr *fattr, struct inode *inode)
4304 {
4305         struct nfs4_exception exception = {
4306                 .interruptible = true,
4307         };
4308         int err;
4309         do {
4310                 err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4311                 trace_nfs4_getattr(server, fhandle, fattr, err);
4312                 err = nfs4_handle_exception(server, err,
4313                                 &exception);
4314         } while (exception.retry);
4315         return err;
4316 }
4317
4318 /*
4319  * The file is not closed if it is opened due to the a request to change
4320  * the size of the file. The open call will not be needed once the
4321  * VFS layer lookup-intents are implemented.
4322  *
4323  * Close is called when the inode is destroyed.
4324  * If we haven't opened the file for O_WRONLY, we
4325  * need to in the size_change case to obtain a stateid.
4326  *
4327  * Got race?
4328  * Because OPEN is always done by name in nfsv4, it is
4329  * possible that we opened a different file by the same
4330  * name.  We can recognize this race condition, but we
4331  * can't do anything about it besides returning an error.
4332  *
4333  * This will be fixed with VFS changes (lookup-intent).
4334  */
4335 static int
4336 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4337                   struct iattr *sattr)
4338 {
4339         struct inode *inode = d_inode(dentry);
4340         const struct cred *cred = NULL;
4341         struct nfs_open_context *ctx = NULL;
4342         int status;
4343
4344         if (pnfs_ld_layoutret_on_setattr(inode) &&
4345             sattr->ia_valid & ATTR_SIZE &&
4346             sattr->ia_size < i_size_read(inode))
4347                 pnfs_commit_and_return_layout(inode);
4348
4349         nfs_fattr_init(fattr);
4350
4351         /* Deal with open(O_TRUNC) */
4352         if (sattr->ia_valid & ATTR_OPEN)
4353                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4354
4355         /* Optimization: if the end result is no change, don't RPC */
4356         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4357                 return 0;
4358
4359         /* Search for an existing open(O_WRITE) file */
4360         if (sattr->ia_valid & ATTR_FILE) {
4361
4362                 ctx = nfs_file_open_context(sattr->ia_file);
4363                 if (ctx)
4364                         cred = ctx->cred;
4365         }
4366
4367         /* Return any delegations if we're going to change ACLs */
4368         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4369                 nfs4_inode_make_writeable(inode);
4370
4371         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4372         if (status == 0) {
4373                 nfs_setattr_update_inode(inode, sattr, fattr);
4374                 nfs_setsecurity(inode, fattr);
4375         }
4376         return status;
4377 }
4378
4379 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4380                 struct dentry *dentry, struct nfs_fh *fhandle,
4381                 struct nfs_fattr *fattr)
4382 {
4383         struct nfs_server *server = NFS_SERVER(dir);
4384         int                    status;
4385         struct nfs4_lookup_arg args = {
4386                 .bitmask = server->attr_bitmask,
4387                 .dir_fh = NFS_FH(dir),
4388                 .name = &dentry->d_name,
4389         };
4390         struct nfs4_lookup_res res = {
4391                 .server = server,
4392                 .fattr = fattr,
4393                 .fh = fhandle,
4394         };
4395         struct rpc_message msg = {
4396                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4397                 .rpc_argp = &args,
4398                 .rpc_resp = &res,
4399         };
4400         unsigned short task_flags = 0;
4401
4402         if (server->nfs_client->cl_minorversion)
4403                 task_flags = RPC_TASK_MOVEABLE;
4404
4405         /* Is this is an attribute revalidation, subject to softreval? */
4406         if (nfs_lookup_is_soft_revalidate(dentry))
4407                 task_flags |= RPC_TASK_TIMEOUT;
4408
4409         args.bitmask = nfs4_bitmask(server, fattr->label);
4410
4411         nfs_fattr_init(fattr);
4412
4413         dprintk("NFS call  lookup %pd2\n", dentry);
4414         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4415         status = nfs4_do_call_sync(clnt, server, &msg,
4416                         &args.seq_args, &res.seq_res, task_flags);
4417         dprintk("NFS reply lookup: %d\n", status);
4418         return status;
4419 }
4420
4421 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4422 {
4423         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4424                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4425         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4426         fattr->nlink = 2;
4427 }
4428
4429 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4430                                    struct dentry *dentry, struct nfs_fh *fhandle,
4431                                    struct nfs_fattr *fattr)
4432 {
4433         struct nfs4_exception exception = {
4434                 .interruptible = true,
4435         };
4436         struct rpc_clnt *client = *clnt;
4437         const struct qstr *name = &dentry->d_name;
4438         int err;
4439         do {
4440                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4441                 trace_nfs4_lookup(dir, name, err);
4442                 switch (err) {
4443                 case -NFS4ERR_BADNAME:
4444                         err = -ENOENT;
4445                         goto out;
4446                 case -NFS4ERR_MOVED:
4447                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4448                         if (err == -NFS4ERR_MOVED)
4449                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4450                         goto out;
4451                 case -NFS4ERR_WRONGSEC:
4452                         err = -EPERM;
4453                         if (client != *clnt)
4454                                 goto out;
4455                         client = nfs4_negotiate_security(client, dir, name);
4456                         if (IS_ERR(client))
4457                                 return PTR_ERR(client);
4458
4459                         exception.retry = 1;
4460                         break;
4461                 default:
4462                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4463                 }
4464         } while (exception.retry);
4465
4466 out:
4467         if (err == 0)
4468                 *clnt = client;
4469         else if (client != *clnt)
4470                 rpc_shutdown_client(client);
4471
4472         return err;
4473 }
4474
4475 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4476                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4477 {
4478         int status;
4479         struct rpc_clnt *client = NFS_CLIENT(dir);
4480
4481         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4482         if (client != NFS_CLIENT(dir)) {
4483                 rpc_shutdown_client(client);
4484                 nfs_fixup_secinfo_attributes(fattr);
4485         }
4486         return status;
4487 }
4488
4489 struct rpc_clnt *
4490 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4491                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4492 {
4493         struct rpc_clnt *client = NFS_CLIENT(dir);
4494         int status;
4495
4496         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4497         if (status < 0)
4498                 return ERR_PTR(status);
4499         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4500 }
4501
4502 static int _nfs4_proc_lookupp(struct inode *inode,
4503                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4504 {
4505         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4506         struct nfs_server *server = NFS_SERVER(inode);
4507         int                    status;
4508         struct nfs4_lookupp_arg args = {
4509                 .bitmask = server->attr_bitmask,
4510                 .fh = NFS_FH(inode),
4511         };
4512         struct nfs4_lookupp_res res = {
4513                 .server = server,
4514                 .fattr = fattr,
4515                 .fh = fhandle,
4516         };
4517         struct rpc_message msg = {
4518                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4519                 .rpc_argp = &args,
4520                 .rpc_resp = &res,
4521         };
4522         unsigned short task_flags = 0;
4523
4524         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4525                 task_flags |= RPC_TASK_TIMEOUT;
4526
4527         args.bitmask = nfs4_bitmask(server, fattr->label);
4528
4529         nfs_fattr_init(fattr);
4530
4531         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4532         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4533                                 &res.seq_res, task_flags);
4534         dprintk("NFS reply lookupp: %d\n", status);
4535         return status;
4536 }
4537
4538 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4539                              struct nfs_fattr *fattr)
4540 {
4541         struct nfs4_exception exception = {
4542                 .interruptible = true,
4543         };
4544         int err;
4545         do {
4546                 err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4547                 trace_nfs4_lookupp(inode, err);
4548                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4549                                 &exception);
4550         } while (exception.retry);
4551         return err;
4552 }
4553
4554 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4555                              const struct cred *cred)
4556 {
4557         struct nfs_server *server = NFS_SERVER(inode);
4558         struct nfs4_accessargs args = {
4559                 .fh = NFS_FH(inode),
4560                 .access = entry->mask,
4561         };
4562         struct nfs4_accessres res = {
4563                 .server = server,
4564         };
4565         struct rpc_message msg = {
4566                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4567                 .rpc_argp = &args,
4568                 .rpc_resp = &res,
4569                 .rpc_cred = cred,
4570         };
4571         int status = 0;
4572
4573         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4574                 res.fattr = nfs_alloc_fattr();
4575                 if (res.fattr == NULL)
4576                         return -ENOMEM;
4577                 args.bitmask = server->cache_consistency_bitmask;
4578         }
4579         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4580         if (!status) {
4581                 nfs_access_set_mask(entry, res.access);
4582                 if (res.fattr)
4583                         nfs_refresh_inode(inode, res.fattr);
4584         }
4585         nfs_free_fattr(res.fattr);
4586         return status;
4587 }
4588
4589 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4590                             const struct cred *cred)
4591 {
4592         struct nfs4_exception exception = {
4593                 .interruptible = true,
4594         };
4595         int err;
4596         do {
4597                 err = _nfs4_proc_access(inode, entry, cred);
4598                 trace_nfs4_access(inode, err);
4599                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4600                                 &exception);
4601         } while (exception.retry);
4602         return err;
4603 }
4604
4605 /*
4606  * TODO: For the time being, we don't try to get any attributes
4607  * along with any of the zero-copy operations READ, READDIR,
4608  * READLINK, WRITE.
4609  *
4610  * In the case of the first three, we want to put the GETATTR
4611  * after the read-type operation -- this is because it is hard
4612  * to predict the length of a GETATTR response in v4, and thus
4613  * align the READ data correctly.  This means that the GETATTR
4614  * may end up partially falling into the page cache, and we should
4615  * shift it into the 'tail' of the xdr_buf before processing.
4616  * To do this efficiently, we need to know the total length
4617  * of data received, which doesn't seem to be available outside
4618  * of the RPC layer.
4619  *
4620  * In the case of WRITE, we also want to put the GETATTR after
4621  * the operation -- in this case because we want to make sure
4622  * we get the post-operation mtime and size.
4623  *
4624  * Both of these changes to the XDR layer would in fact be quite
4625  * minor, but I decided to leave them for a subsequent patch.
4626  */
4627 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4628                 unsigned int pgbase, unsigned int pglen)
4629 {
4630         struct nfs4_readlink args = {
4631                 .fh       = NFS_FH(inode),
4632                 .pgbase   = pgbase,
4633                 .pglen    = pglen,
4634                 .pages    = &page,
4635         };
4636         struct nfs4_readlink_res res;
4637         struct rpc_message msg = {
4638                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4639                 .rpc_argp = &args,
4640                 .rpc_resp = &res,
4641         };
4642
4643         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4644 }
4645
4646 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4647                 unsigned int pgbase, unsigned int pglen)
4648 {
4649         struct nfs4_exception exception = {
4650                 .interruptible = true,
4651         };
4652         int err;
4653         do {
4654                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4655                 trace_nfs4_readlink(inode, err);
4656                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4657                                 &exception);
4658         } while (exception.retry);
4659         return err;
4660 }
4661
4662 /*
4663  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4664  */
4665 static int
4666 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4667                  int flags)
4668 {
4669         struct nfs_server *server = NFS_SERVER(dir);
4670         struct nfs4_label l, *ilabel = NULL;
4671         struct nfs_open_context *ctx;
4672         struct nfs4_state *state;
4673         int status = 0;
4674
4675         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4676         if (IS_ERR(ctx))
4677                 return PTR_ERR(ctx);
4678
4679         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4680
4681         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4682                 sattr->ia_mode &= ~current_umask();
4683         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4684         if (IS_ERR(state)) {
4685                 status = PTR_ERR(state);
4686                 goto out;
4687         }
4688 out:
4689         nfs4_label_release_security(ilabel);
4690         put_nfs_open_context(ctx);
4691         return status;
4692 }
4693
4694 static int
4695 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4696 {
4697         struct nfs_server *server = NFS_SERVER(dir);
4698         struct nfs_removeargs args = {
4699                 .fh = NFS_FH(dir),
4700                 .name = *name,
4701         };
4702         struct nfs_removeres res = {
4703                 .server = server,
4704         };
4705         struct rpc_message msg = {
4706                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4707                 .rpc_argp = &args,
4708                 .rpc_resp = &res,
4709         };
4710         unsigned long timestamp = jiffies;
4711         int status;
4712
4713         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4714         if (status == 0) {
4715                 spin_lock(&dir->i_lock);
4716                 /* Removing a directory decrements nlink in the parent */
4717                 if (ftype == NF4DIR && dir->i_nlink > 2)
4718                         nfs4_dec_nlink_locked(dir);
4719                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4720                                               NFS_INO_INVALID_DATA);
4721                 spin_unlock(&dir->i_lock);
4722         }
4723         return status;
4724 }
4725
4726 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4727 {
4728         struct nfs4_exception exception = {
4729                 .interruptible = true,
4730         };
4731         struct inode *inode = d_inode(dentry);
4732         int err;
4733
4734         if (inode) {
4735                 if (inode->i_nlink == 1)
4736                         nfs4_inode_return_delegation(inode);
4737                 else
4738                         nfs4_inode_make_writeable(inode);
4739         }
4740         do {
4741                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4742                 trace_nfs4_remove(dir, &dentry->d_name, err);
4743                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4744                                 &exception);
4745         } while (exception.retry);
4746         return err;
4747 }
4748
4749 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4750 {
4751         struct nfs4_exception exception = {
4752                 .interruptible = true,
4753         };
4754         int err;
4755
4756         do {
4757                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4758                 trace_nfs4_remove(dir, name, err);
4759                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4760                                 &exception);
4761         } while (exception.retry);
4762         return err;
4763 }
4764
4765 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4766                 struct dentry *dentry,
4767                 struct inode *inode)
4768 {
4769         struct nfs_removeargs *args = msg->rpc_argp;
4770         struct nfs_removeres *res = msg->rpc_resp;
4771
4772         res->server = NFS_SB(dentry->d_sb);
4773         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4774         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4775
4776         nfs_fattr_init(res->dir_attr);
4777
4778         if (inode) {
4779                 nfs4_inode_return_delegation(inode);
4780                 nfs_d_prune_case_insensitive_aliases(inode);
4781         }
4782 }
4783
4784 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4785 {
4786         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4787                         &data->args.seq_args,
4788                         &data->res.seq_res,
4789                         task);
4790 }
4791
4792 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4793 {
4794         struct nfs_unlinkdata *data = task->tk_calldata;
4795         struct nfs_removeres *res = &data->res;
4796
4797         if (!nfs4_sequence_done(task, &res->seq_res))
4798                 return 0;
4799         if (nfs4_async_handle_error(task, res->server, NULL,
4800                                     &data->timeout) == -EAGAIN)
4801                 return 0;
4802         if (task->tk_status == 0)
4803                 nfs4_update_changeattr(dir, &res->cinfo,
4804                                 res->dir_attr->time_start,
4805                                 NFS_INO_INVALID_DATA);
4806         return 1;
4807 }
4808
4809 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4810                 struct dentry *old_dentry,
4811                 struct dentry *new_dentry)
4812 {
4813         struct nfs_renameargs *arg = msg->rpc_argp;
4814         struct nfs_renameres *res = msg->rpc_resp;
4815         struct inode *old_inode = d_inode(old_dentry);
4816         struct inode *new_inode = d_inode(new_dentry);
4817
4818         if (old_inode)
4819                 nfs4_inode_make_writeable(old_inode);
4820         if (new_inode)
4821                 nfs4_inode_return_delegation(new_inode);
4822         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4823         res->server = NFS_SB(old_dentry->d_sb);
4824         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4825 }
4826
4827 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4828 {
4829         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4830                         &data->args.seq_args,
4831                         &data->res.seq_res,
4832                         task);
4833 }
4834
4835 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4836                                  struct inode *new_dir)
4837 {
4838         struct nfs_renamedata *data = task->tk_calldata;
4839         struct nfs_renameres *res = &data->res;
4840
4841         if (!nfs4_sequence_done(task, &res->seq_res))
4842                 return 0;
4843         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4844                 return 0;
4845
4846         if (task->tk_status == 0) {
4847                 nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4848                 if (new_dir != old_dir) {
4849                         /* Note: If we moved a directory, nlink will change */
4850                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4851                                         res->old_fattr->time_start,
4852                                         NFS_INO_INVALID_NLINK |
4853                                             NFS_INO_INVALID_DATA);
4854                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4855                                         res->new_fattr->time_start,
4856                                         NFS_INO_INVALID_NLINK |
4857                                             NFS_INO_INVALID_DATA);
4858                 } else
4859                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4860                                         res->old_fattr->time_start,
4861                                         NFS_INO_INVALID_DATA);
4862         }
4863         return 1;
4864 }
4865
4866 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4867 {
4868         struct nfs_server *server = NFS_SERVER(inode);
4869         __u32 bitmask[NFS4_BITMASK_SZ];
4870         struct nfs4_link_arg arg = {
4871                 .fh     = NFS_FH(inode),
4872                 .dir_fh = NFS_FH(dir),
4873                 .name   = name,
4874                 .bitmask = bitmask,
4875         };
4876         struct nfs4_link_res res = {
4877                 .server = server,
4878         };
4879         struct rpc_message msg = {
4880                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4881                 .rpc_argp = &arg,
4882                 .rpc_resp = &res,
4883         };
4884         int status = -ENOMEM;
4885
4886         res.fattr = nfs_alloc_fattr_with_label(server);
4887         if (res.fattr == NULL)
4888                 goto out;
4889
4890         nfs4_inode_make_writeable(inode);
4891         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4892                                 NFS_INO_INVALID_CHANGE);
4893         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4894         if (!status) {
4895                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4896                                        NFS_INO_INVALID_DATA);
4897                 nfs4_inc_nlink(inode);
4898                 status = nfs_post_op_update_inode(inode, res.fattr);
4899                 if (!status)
4900                         nfs_setsecurity(inode, res.fattr);
4901         }
4902
4903 out:
4904         nfs_free_fattr(res.fattr);
4905         return status;
4906 }
4907
4908 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4909 {
4910         struct nfs4_exception exception = {
4911                 .interruptible = true,
4912         };
4913         int err;
4914         do {
4915                 err = nfs4_handle_exception(NFS_SERVER(inode),
4916                                 _nfs4_proc_link(inode, dir, name),
4917                                 &exception);
4918         } while (exception.retry);
4919         return err;
4920 }
4921
4922 struct nfs4_createdata {
4923         struct rpc_message msg;
4924         struct nfs4_create_arg arg;
4925         struct nfs4_create_res res;
4926         struct nfs_fh fh;
4927         struct nfs_fattr fattr;
4928 };
4929
4930 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4931                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4932 {
4933         struct nfs4_createdata *data;
4934
4935         data = kzalloc(sizeof(*data), GFP_KERNEL);
4936         if (data != NULL) {
4937                 struct nfs_server *server = NFS_SERVER(dir);
4938
4939                 data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4940                 if (IS_ERR(data->fattr.label))
4941                         goto out_free;
4942
4943                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4944                 data->msg.rpc_argp = &data->arg;
4945                 data->msg.rpc_resp = &data->res;
4946                 data->arg.dir_fh = NFS_FH(dir);
4947                 data->arg.server = server;
4948                 data->arg.name = name;
4949                 data->arg.attrs = sattr;
4950                 data->arg.ftype = ftype;
4951                 data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4952                 data->arg.umask = current_umask();
4953                 data->res.server = server;
4954                 data->res.fh = &data->fh;
4955                 data->res.fattr = &data->fattr;
4956                 nfs_fattr_init(data->res.fattr);
4957         }
4958         return data;
4959 out_free:
4960         kfree(data);
4961         return NULL;
4962 }
4963
4964 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4965 {
4966         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4967                                     &data->arg.seq_args, &data->res.seq_res, 1);
4968         if (status == 0) {
4969                 spin_lock(&dir->i_lock);
4970                 /* Creating a directory bumps nlink in the parent */
4971                 if (data->arg.ftype == NF4DIR)
4972                         nfs4_inc_nlink_locked(dir);
4973                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4974                                               data->res.fattr->time_start,
4975                                               NFS_INO_INVALID_DATA);
4976                 spin_unlock(&dir->i_lock);
4977                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4978         }
4979         return status;
4980 }
4981
4982 static void nfs4_free_createdata(struct nfs4_createdata *data)
4983 {
4984         nfs4_label_free(data->fattr.label);
4985         kfree(data);
4986 }
4987
4988 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4989                 struct page *page, unsigned int len, struct iattr *sattr,
4990                 struct nfs4_label *label)
4991 {
4992         struct nfs4_createdata *data;
4993         int status = -ENAMETOOLONG;
4994
4995         if (len > NFS4_MAXPATHLEN)
4996                 goto out;
4997
4998         status = -ENOMEM;
4999         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5000         if (data == NULL)
5001                 goto out;
5002
5003         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5004         data->arg.u.symlink.pages = &page;
5005         data->arg.u.symlink.len = len;
5006         data->arg.label = label;
5007
5008         status = nfs4_do_create(dir, dentry, data);
5009
5010         nfs4_free_createdata(data);
5011 out:
5012         return status;
5013 }
5014
5015 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5016                 struct page *page, unsigned int len, struct iattr *sattr)
5017 {
5018         struct nfs4_exception exception = {
5019                 .interruptible = true,
5020         };
5021         struct nfs4_label l, *label = NULL;
5022         int err;
5023
5024         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5025
5026         do {
5027                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5028                 trace_nfs4_symlink(dir, &dentry->d_name, err);
5029                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5030                                 &exception);
5031         } while (exception.retry);
5032
5033         nfs4_label_release_security(label);
5034         return err;
5035 }
5036
5037 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5038                 struct iattr *sattr, struct nfs4_label *label)
5039 {
5040         struct nfs4_createdata *data;
5041         int status = -ENOMEM;
5042
5043         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5044         if (data == NULL)
5045                 goto out;
5046
5047         data->arg.label = label;
5048         status = nfs4_do_create(dir, dentry, data);
5049
5050         nfs4_free_createdata(data);
5051 out:
5052         return status;
5053 }
5054
5055 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5056                 struct iattr *sattr)
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_mkdir(dir, dentry, sattr, label);
5071                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
5072                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5073                                 &exception);
5074         } while (exception.retry);
5075         nfs4_label_release_security(label);
5076
5077         return err;
5078 }
5079
5080 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5081                               struct nfs_readdir_res *nr_res)
5082 {
5083         struct inode            *dir = d_inode(nr_arg->dentry);
5084         struct nfs_server       *server = NFS_SERVER(dir);
5085         struct nfs4_readdir_arg args = {
5086                 .fh = NFS_FH(dir),
5087                 .pages = nr_arg->pages,
5088                 .pgbase = 0,
5089                 .count = nr_arg->page_len,
5090                 .plus = nr_arg->plus,
5091         };
5092         struct nfs4_readdir_res res;
5093         struct rpc_message msg = {
5094                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5095                 .rpc_argp = &args,
5096                 .rpc_resp = &res,
5097                 .rpc_cred = nr_arg->cred,
5098         };
5099         int                     status;
5100
5101         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5102                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5103         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5104                 args.bitmask = server->attr_bitmask_nl;
5105         else
5106                 args.bitmask = server->attr_bitmask;
5107
5108         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5109         res.pgbase = args.pgbase;
5110         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5111                         &res.seq_res, 0);
5112         if (status >= 0) {
5113                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5114                 status += args.pgbase;
5115         }
5116
5117         nfs_invalidate_atime(dir);
5118
5119         dprintk("%s: returns %d\n", __func__, status);
5120         return status;
5121 }
5122
5123 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5124                              struct nfs_readdir_res *res)
5125 {
5126         struct nfs4_exception exception = {
5127                 .interruptible = true,
5128         };
5129         int err;
5130         do {
5131                 err = _nfs4_proc_readdir(arg, res);
5132                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5133                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5134                                             err, &exception);
5135         } while (exception.retry);
5136         return err;
5137 }
5138
5139 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5140                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5141 {
5142         struct nfs4_createdata *data;
5143         int mode = sattr->ia_mode;
5144         int status = -ENOMEM;
5145
5146         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5147         if (data == NULL)
5148                 goto out;
5149
5150         if (S_ISFIFO(mode))
5151                 data->arg.ftype = NF4FIFO;
5152         else if (S_ISBLK(mode)) {
5153                 data->arg.ftype = NF4BLK;
5154                 data->arg.u.device.specdata1 = MAJOR(rdev);
5155                 data->arg.u.device.specdata2 = MINOR(rdev);
5156         }
5157         else if (S_ISCHR(mode)) {
5158                 data->arg.ftype = NF4CHR;
5159                 data->arg.u.device.specdata1 = MAJOR(rdev);
5160                 data->arg.u.device.specdata2 = MINOR(rdev);
5161         } else if (!S_ISSOCK(mode)) {
5162                 status = -EINVAL;
5163                 goto out_free;
5164         }
5165
5166         data->arg.label = label;
5167         status = nfs4_do_create(dir, dentry, data);
5168 out_free:
5169         nfs4_free_createdata(data);
5170 out:
5171         return status;
5172 }
5173
5174 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5175                 struct iattr *sattr, dev_t rdev)
5176 {
5177         struct nfs_server *server = NFS_SERVER(dir);
5178         struct nfs4_exception exception = {
5179                 .interruptible = true,
5180         };
5181         struct nfs4_label l, *label = NULL;
5182         int err;
5183
5184         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5185
5186         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5187                 sattr->ia_mode &= ~current_umask();
5188         do {
5189                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5190                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5191                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5192                                 &exception);
5193         } while (exception.retry);
5194
5195         nfs4_label_release_security(label);
5196
5197         return err;
5198 }
5199
5200 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5201                  struct nfs_fsstat *fsstat)
5202 {
5203         struct nfs4_statfs_arg args = {
5204                 .fh = fhandle,
5205                 .bitmask = server->attr_bitmask,
5206         };
5207         struct nfs4_statfs_res res = {
5208                 .fsstat = fsstat,
5209         };
5210         struct rpc_message msg = {
5211                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5212                 .rpc_argp = &args,
5213                 .rpc_resp = &res,
5214         };
5215
5216         nfs_fattr_init(fsstat->fattr);
5217         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5218 }
5219
5220 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5221 {
5222         struct nfs4_exception exception = {
5223                 .interruptible = true,
5224         };
5225         int err;
5226         do {
5227                 err = nfs4_handle_exception(server,
5228                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5229                                 &exception);
5230         } while (exception.retry);
5231         return err;
5232 }
5233
5234 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5235                 struct nfs_fsinfo *fsinfo)
5236 {
5237         struct nfs4_fsinfo_arg args = {
5238                 .fh = fhandle,
5239                 .bitmask = server->attr_bitmask,
5240         };
5241         struct nfs4_fsinfo_res res = {
5242                 .fsinfo = fsinfo,
5243         };
5244         struct rpc_message msg = {
5245                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5246                 .rpc_argp = &args,
5247                 .rpc_resp = &res,
5248         };
5249
5250         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5251 }
5252
5253 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5254 {
5255         struct nfs4_exception exception = {
5256                 .interruptible = true,
5257         };
5258         int err;
5259
5260         do {
5261                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5262                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5263                 if (err == 0) {
5264                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5265                         break;
5266                 }
5267                 err = nfs4_handle_exception(server, err, &exception);
5268         } while (exception.retry);
5269         return err;
5270 }
5271
5272 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5273 {
5274         int error;
5275
5276         nfs_fattr_init(fsinfo->fattr);
5277         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5278         if (error == 0) {
5279                 /* block layout checks this! */
5280                 server->pnfs_blksize = fsinfo->blksize;
5281                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5282         }
5283
5284         return error;
5285 }
5286
5287 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5288                 struct nfs_pathconf *pathconf)
5289 {
5290         struct nfs4_pathconf_arg args = {
5291                 .fh = fhandle,
5292                 .bitmask = server->attr_bitmask,
5293         };
5294         struct nfs4_pathconf_res res = {
5295                 .pathconf = pathconf,
5296         };
5297         struct rpc_message msg = {
5298                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5299                 .rpc_argp = &args,
5300                 .rpc_resp = &res,
5301         };
5302
5303         /* None of the pathconf attributes are mandatory to implement */
5304         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5305                 memset(pathconf, 0, sizeof(*pathconf));
5306                 return 0;
5307         }
5308
5309         nfs_fattr_init(pathconf->fattr);
5310         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5311 }
5312
5313 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5314                 struct nfs_pathconf *pathconf)
5315 {
5316         struct nfs4_exception exception = {
5317                 .interruptible = true,
5318         };
5319         int err;
5320
5321         do {
5322                 err = nfs4_handle_exception(server,
5323                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5324                                 &exception);
5325         } while (exception.retry);
5326         return err;
5327 }
5328
5329 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5330                 const struct nfs_open_context *ctx,
5331                 const struct nfs_lock_context *l_ctx,
5332                 fmode_t fmode)
5333 {
5334         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5335 }
5336 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5337
5338 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5339                 const struct nfs_open_context *ctx,
5340                 const struct nfs_lock_context *l_ctx,
5341                 fmode_t fmode)
5342 {
5343         nfs4_stateid _current_stateid;
5344
5345         /* If the current stateid represents a lost lock, then exit */
5346         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5347                 return true;
5348         return nfs4_stateid_match(stateid, &_current_stateid);
5349 }
5350
5351 static bool nfs4_error_stateid_expired(int err)
5352 {
5353         switch (err) {
5354         case -NFS4ERR_DELEG_REVOKED:
5355         case -NFS4ERR_ADMIN_REVOKED:
5356         case -NFS4ERR_BAD_STATEID:
5357         case -NFS4ERR_STALE_STATEID:
5358         case -NFS4ERR_OLD_STATEID:
5359         case -NFS4ERR_OPENMODE:
5360         case -NFS4ERR_EXPIRED:
5361                 return true;
5362         }
5363         return false;
5364 }
5365
5366 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5367 {
5368         struct nfs_server *server = NFS_SERVER(hdr->inode);
5369
5370         trace_nfs4_read(hdr, task->tk_status);
5371         if (task->tk_status < 0) {
5372                 struct nfs4_exception exception = {
5373                         .inode = hdr->inode,
5374                         .state = hdr->args.context->state,
5375                         .stateid = &hdr->args.stateid,
5376                 };
5377                 task->tk_status = nfs4_async_handle_exception(task,
5378                                 server, task->tk_status, &exception);
5379                 if (exception.retry) {
5380                         rpc_restart_call_prepare(task);
5381                         return -EAGAIN;
5382                 }
5383         }
5384
5385         if (task->tk_status > 0)
5386                 renew_lease(server, hdr->timestamp);
5387         return 0;
5388 }
5389
5390 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5391                 struct nfs_pgio_args *args)
5392 {
5393
5394         if (!nfs4_error_stateid_expired(task->tk_status) ||
5395                 nfs4_stateid_is_current(&args->stateid,
5396                                 args->context,
5397                                 args->lock_context,
5398                                 FMODE_READ))
5399                 return false;
5400         rpc_restart_call_prepare(task);
5401         return true;
5402 }
5403
5404 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5405                                          struct nfs_pgio_header *hdr)
5406 {
5407         struct nfs_server *server = NFS_SERVER(hdr->inode);
5408         struct rpc_message *msg = &task->tk_msg;
5409
5410         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5411             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5412                 server->caps &= ~NFS_CAP_READ_PLUS;
5413                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5414                 rpc_restart_call_prepare(task);
5415                 return true;
5416         }
5417         return false;
5418 }
5419
5420 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5421 {
5422         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5423                 return -EAGAIN;
5424         if (nfs4_read_stateid_changed(task, &hdr->args))
5425                 return -EAGAIN;
5426         if (nfs4_read_plus_not_supported(task, hdr))
5427                 return -EAGAIN;
5428         if (task->tk_status > 0)
5429                 nfs_invalidate_atime(hdr->inode);
5430         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5431                                     nfs4_read_done_cb(task, hdr);
5432 }
5433
5434 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5435 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5436                                     struct rpc_message *msg)
5437 {
5438         /* Note: We don't use READ_PLUS with pNFS yet */
5439         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5440                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5441 }
5442 #else
5443 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5444                                     struct rpc_message *msg)
5445 {
5446 }
5447 #endif /* CONFIG_NFS_V4_2 */
5448
5449 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5450                                  struct rpc_message *msg)
5451 {
5452         hdr->timestamp   = jiffies;
5453         if (!hdr->pgio_done_cb)
5454                 hdr->pgio_done_cb = nfs4_read_done_cb;
5455         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5456         nfs42_read_plus_support(hdr, msg);
5457         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5458 }
5459
5460 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5461                                       struct nfs_pgio_header *hdr)
5462 {
5463         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5464                         &hdr->args.seq_args,
5465                         &hdr->res.seq_res,
5466                         task))
5467                 return 0;
5468         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5469                                 hdr->args.lock_context,
5470                                 hdr->rw_mode) == -EIO)
5471                 return -EIO;
5472         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5473                 return -EIO;
5474         return 0;
5475 }
5476
5477 static int nfs4_write_done_cb(struct rpc_task *task,
5478                               struct nfs_pgio_header *hdr)
5479 {
5480         struct inode *inode = hdr->inode;
5481
5482         trace_nfs4_write(hdr, task->tk_status);
5483         if (task->tk_status < 0) {
5484                 struct nfs4_exception exception = {
5485                         .inode = hdr->inode,
5486                         .state = hdr->args.context->state,
5487                         .stateid = &hdr->args.stateid,
5488                 };
5489                 task->tk_status = nfs4_async_handle_exception(task,
5490                                 NFS_SERVER(inode), task->tk_status,
5491                                 &exception);
5492                 if (exception.retry) {
5493                         rpc_restart_call_prepare(task);
5494                         return -EAGAIN;
5495                 }
5496         }
5497         if (task->tk_status >= 0) {
5498                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5499                 nfs_writeback_update_inode(hdr);
5500         }
5501         return 0;
5502 }
5503
5504 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5505                 struct nfs_pgio_args *args)
5506 {
5507
5508         if (!nfs4_error_stateid_expired(task->tk_status) ||
5509                 nfs4_stateid_is_current(&args->stateid,
5510                                 args->context,
5511                                 args->lock_context,
5512                                 FMODE_WRITE))
5513                 return false;
5514         rpc_restart_call_prepare(task);
5515         return true;
5516 }
5517
5518 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5519 {
5520         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5521                 return -EAGAIN;
5522         if (nfs4_write_stateid_changed(task, &hdr->args))
5523                 return -EAGAIN;
5524         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5525                 nfs4_write_done_cb(task, hdr);
5526 }
5527
5528 static
5529 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5530 {
5531         /* Don't request attributes for pNFS or O_DIRECT writes */
5532         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5533                 return false;
5534         /* Otherwise, request attributes if and only if we don't hold
5535          * a delegation
5536          */
5537         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5538 }
5539
5540 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5541                       struct inode *inode, unsigned long cache_validity)
5542 {
5543         struct nfs_server *server = NFS_SERVER(inode);
5544         unsigned int i;
5545
5546         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5547         cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5548
5549         if (cache_validity & NFS_INO_INVALID_CHANGE)
5550                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5551         if (cache_validity & NFS_INO_INVALID_ATIME)
5552                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5553         if (cache_validity & NFS_INO_INVALID_MODE)
5554                 bitmask[1] |= FATTR4_WORD1_MODE;
5555         if (cache_validity & NFS_INO_INVALID_OTHER)
5556                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5557         if (cache_validity & NFS_INO_INVALID_NLINK)
5558                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5559         if (cache_validity & NFS_INO_INVALID_CTIME)
5560                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5561         if (cache_validity & NFS_INO_INVALID_MTIME)
5562                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5563         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5564                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5565
5566         if (cache_validity & NFS_INO_INVALID_SIZE)
5567                 bitmask[0] |= FATTR4_WORD0_SIZE;
5568
5569         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5570                 bitmask[i] &= server->attr_bitmask[i];
5571 }
5572
5573 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5574                                   struct rpc_message *msg,
5575                                   struct rpc_clnt **clnt)
5576 {
5577         struct nfs_server *server = NFS_SERVER(hdr->inode);
5578
5579         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5580                 hdr->args.bitmask = NULL;
5581                 hdr->res.fattr = NULL;
5582         } else {
5583                 nfs4_bitmask_set(hdr->args.bitmask_store,
5584                                  server->cache_consistency_bitmask,
5585                                  hdr->inode, NFS_INO_INVALID_BLOCKS);
5586                 hdr->args.bitmask = hdr->args.bitmask_store;
5587         }
5588
5589         if (!hdr->pgio_done_cb)
5590                 hdr->pgio_done_cb = nfs4_write_done_cb;
5591         hdr->res.server = server;
5592         hdr->timestamp   = jiffies;
5593
5594         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5595         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5596         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5597 }
5598
5599 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5600 {
5601         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5602                         &data->args.seq_args,
5603                         &data->res.seq_res,
5604                         task);
5605 }
5606
5607 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5608 {
5609         struct inode *inode = data->inode;
5610
5611         trace_nfs4_commit(data, task->tk_status);
5612         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5613                                     NULL, NULL) == -EAGAIN) {
5614                 rpc_restart_call_prepare(task);
5615                 return -EAGAIN;
5616         }
5617         return 0;
5618 }
5619
5620 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5621 {
5622         if (!nfs4_sequence_done(task, &data->res.seq_res))
5623                 return -EAGAIN;
5624         return data->commit_done_cb(task, data);
5625 }
5626
5627 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5628                                    struct rpc_clnt **clnt)
5629 {
5630         struct nfs_server *server = NFS_SERVER(data->inode);
5631
5632         if (data->commit_done_cb == NULL)
5633                 data->commit_done_cb = nfs4_commit_done_cb;
5634         data->res.server = server;
5635         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5636         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5637         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5638 }
5639
5640 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5641                                 struct nfs_commitres *res)
5642 {
5643         struct inode *dst_inode = file_inode(dst);
5644         struct nfs_server *server = NFS_SERVER(dst_inode);
5645         struct rpc_message msg = {
5646                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5647                 .rpc_argp = args,
5648                 .rpc_resp = res,
5649         };
5650
5651         args->fh = NFS_FH(dst_inode);
5652         return nfs4_call_sync(server->client, server, &msg,
5653                         &args->seq_args, &res->seq_res, 1);
5654 }
5655
5656 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5657 {
5658         struct nfs_commitargs args = {
5659                 .offset = offset,
5660                 .count = count,
5661         };
5662         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5663         struct nfs4_exception exception = { };
5664         int status;
5665
5666         do {
5667                 status = _nfs4_proc_commit(dst, &args, res);
5668                 status = nfs4_handle_exception(dst_server, status, &exception);
5669         } while (exception.retry);
5670
5671         return status;
5672 }
5673
5674 struct nfs4_renewdata {
5675         struct nfs_client       *client;
5676         unsigned long           timestamp;
5677 };
5678
5679 /*
5680  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5681  * standalone procedure for queueing an asynchronous RENEW.
5682  */
5683 static void nfs4_renew_release(void *calldata)
5684 {
5685         struct nfs4_renewdata *data = calldata;
5686         struct nfs_client *clp = data->client;
5687
5688         if (refcount_read(&clp->cl_count) > 1)
5689                 nfs4_schedule_state_renewal(clp);
5690         nfs_put_client(clp);
5691         kfree(data);
5692 }
5693
5694 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5695 {
5696         struct nfs4_renewdata *data = calldata;
5697         struct nfs_client *clp = data->client;
5698         unsigned long timestamp = data->timestamp;
5699
5700         trace_nfs4_renew_async(clp, task->tk_status);
5701         switch (task->tk_status) {
5702         case 0:
5703                 break;
5704         case -NFS4ERR_LEASE_MOVED:
5705                 nfs4_schedule_lease_moved_recovery(clp);
5706                 break;
5707         default:
5708                 /* Unless we're shutting down, schedule state recovery! */
5709                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5710                         return;
5711                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5712                         nfs4_schedule_lease_recovery(clp);
5713                         return;
5714                 }
5715                 nfs4_schedule_path_down_recovery(clp);
5716         }
5717         do_renew_lease(clp, timestamp);
5718 }
5719
5720 static const struct rpc_call_ops nfs4_renew_ops = {
5721         .rpc_call_done = nfs4_renew_done,
5722         .rpc_release = nfs4_renew_release,
5723 };
5724
5725 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5726 {
5727         struct rpc_message msg = {
5728                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5729                 .rpc_argp       = clp,
5730                 .rpc_cred       = cred,
5731         };
5732         struct nfs4_renewdata *data;
5733
5734         if (renew_flags == 0)
5735                 return 0;
5736         if (!refcount_inc_not_zero(&clp->cl_count))
5737                 return -EIO;
5738         data = kmalloc(sizeof(*data), GFP_NOFS);
5739         if (data == NULL) {
5740                 nfs_put_client(clp);
5741                 return -ENOMEM;
5742         }
5743         data->client = clp;
5744         data->timestamp = jiffies;
5745         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5746                         &nfs4_renew_ops, data);
5747 }
5748
5749 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5750 {
5751         struct rpc_message msg = {
5752                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5753                 .rpc_argp       = clp,
5754                 .rpc_cred       = cred,
5755         };
5756         unsigned long now = jiffies;
5757         int status;
5758
5759         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5760         if (status < 0)
5761                 return status;
5762         do_renew_lease(clp, now);
5763         return 0;
5764 }
5765
5766 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5767 {
5768         return server->caps & NFS_CAP_ACLS;
5769 }
5770
5771 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5772  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5773  * the stack.
5774  */
5775 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5776
5777 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5778                 struct page **pages)
5779 {
5780         struct page *newpage, **spages;
5781         int rc = 0;
5782         size_t len;
5783         spages = pages;
5784
5785         do {
5786                 len = min_t(size_t, PAGE_SIZE, buflen);
5787                 newpage = alloc_page(GFP_KERNEL);
5788
5789                 if (newpage == NULL)
5790                         goto unwind;
5791                 memcpy(page_address(newpage), buf, len);
5792                 buf += len;
5793                 buflen -= len;
5794                 *pages++ = newpage;
5795                 rc++;
5796         } while (buflen != 0);
5797
5798         return rc;
5799
5800 unwind:
5801         for(; rc > 0; rc--)
5802                 __free_page(spages[rc-1]);
5803         return -ENOMEM;
5804 }
5805
5806 struct nfs4_cached_acl {
5807         int cached;
5808         size_t len;
5809         char data[];
5810 };
5811
5812 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5813 {
5814         struct nfs_inode *nfsi = NFS_I(inode);
5815
5816         spin_lock(&inode->i_lock);
5817         kfree(nfsi->nfs4_acl);
5818         nfsi->nfs4_acl = acl;
5819         spin_unlock(&inode->i_lock);
5820 }
5821
5822 static void nfs4_zap_acl_attr(struct inode *inode)
5823 {
5824         nfs4_set_cached_acl(inode, NULL);
5825 }
5826
5827 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5828 {
5829         struct nfs_inode *nfsi = NFS_I(inode);
5830         struct nfs4_cached_acl *acl;
5831         int ret = -ENOENT;
5832
5833         spin_lock(&inode->i_lock);
5834         acl = nfsi->nfs4_acl;
5835         if (acl == NULL)
5836                 goto out;
5837         if (buf == NULL) /* user is just asking for length */
5838                 goto out_len;
5839         if (acl->cached == 0)
5840                 goto out;
5841         ret = -ERANGE; /* see getxattr(2) man page */
5842         if (acl->len > buflen)
5843                 goto out;
5844         memcpy(buf, acl->data, acl->len);
5845 out_len:
5846         ret = acl->len;
5847 out:
5848         spin_unlock(&inode->i_lock);
5849         return ret;
5850 }
5851
5852 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5853 {
5854         struct nfs4_cached_acl *acl;
5855         size_t buflen = sizeof(*acl) + acl_len;
5856
5857         if (buflen <= PAGE_SIZE) {
5858                 acl = kmalloc(buflen, GFP_KERNEL);
5859                 if (acl == NULL)
5860                         goto out;
5861                 acl->cached = 1;
5862                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5863         } else {
5864                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5865                 if (acl == NULL)
5866                         goto out;
5867                 acl->cached = 0;
5868         }
5869         acl->len = acl_len;
5870 out:
5871         nfs4_set_cached_acl(inode, acl);
5872 }
5873
5874 /*
5875  * The getxattr API returns the required buffer length when called with a
5876  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5877  * the required buf.  On a NULL buf, we send a page of data to the server
5878  * guessing that the ACL request can be serviced by a page. If so, we cache
5879  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5880  * the cache. If not so, we throw away the page, and cache the required
5881  * length. The next getxattr call will then produce another round trip to
5882  * the server, this time with the input buf of the required size.
5883  */
5884 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5885 {
5886         struct page **pages;
5887         struct nfs_getaclargs args = {
5888                 .fh = NFS_FH(inode),
5889                 .acl_len = buflen,
5890         };
5891         struct nfs_getaclres res = {
5892                 .acl_len = buflen,
5893         };
5894         struct rpc_message msg = {
5895                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5896                 .rpc_argp = &args,
5897                 .rpc_resp = &res,
5898         };
5899         unsigned int npages;
5900         int ret = -ENOMEM, i;
5901         struct nfs_server *server = NFS_SERVER(inode);
5902
5903         if (buflen == 0)
5904                 buflen = server->rsize;
5905
5906         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5907         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5908         if (!pages)
5909                 return -ENOMEM;
5910
5911         args.acl_pages = pages;
5912
5913         for (i = 0; i < npages; i++) {
5914                 pages[i] = alloc_page(GFP_KERNEL);
5915                 if (!pages[i])
5916                         goto out_free;
5917         }
5918
5919         /* for decoding across pages */
5920         res.acl_scratch = alloc_page(GFP_KERNEL);
5921         if (!res.acl_scratch)
5922                 goto out_free;
5923
5924         args.acl_len = npages * PAGE_SIZE;
5925
5926         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5927                 __func__, buf, buflen, npages, args.acl_len);
5928         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5929                              &msg, &args.seq_args, &res.seq_res, 0);
5930         if (ret)
5931                 goto out_free;
5932
5933         /* Handle the case where the passed-in buffer is too short */
5934         if (res.acl_flags & NFS4_ACL_TRUNC) {
5935                 /* Did the user only issue a request for the acl length? */
5936                 if (buf == NULL)
5937                         goto out_ok;
5938                 ret = -ERANGE;
5939                 goto out_free;
5940         }
5941         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5942         if (buf) {
5943                 if (res.acl_len > buflen) {
5944                         ret = -ERANGE;
5945                         goto out_free;
5946                 }
5947                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5948         }
5949 out_ok:
5950         ret = res.acl_len;
5951 out_free:
5952         for (i = 0; i < npages; i++)
5953                 if (pages[i])
5954                         __free_page(pages[i]);
5955         if (res.acl_scratch)
5956                 __free_page(res.acl_scratch);
5957         kfree(pages);
5958         return ret;
5959 }
5960
5961 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5962 {
5963         struct nfs4_exception exception = {
5964                 .interruptible = true,
5965         };
5966         ssize_t ret;
5967         do {
5968                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5969                 trace_nfs4_get_acl(inode, ret);
5970                 if (ret >= 0)
5971                         break;
5972                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5973         } while (exception.retry);
5974         return ret;
5975 }
5976
5977 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5978 {
5979         struct nfs_server *server = NFS_SERVER(inode);
5980         int ret;
5981
5982         if (!nfs4_server_supports_acls(server))
5983                 return -EOPNOTSUPP;
5984         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
5985         if (ret < 0)
5986                 return ret;
5987         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5988                 nfs_zap_acl_cache(inode);
5989         ret = nfs4_read_cached_acl(inode, buf, buflen);
5990         if (ret != -ENOENT)
5991                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5992                  * but no cached acl data, just the acl length */
5993                 return ret;
5994         return nfs4_get_acl_uncached(inode, buf, buflen);
5995 }
5996
5997 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5998 {
5999         struct nfs_server *server = NFS_SERVER(inode);
6000         struct page *pages[NFS4ACL_MAXPAGES];
6001         struct nfs_setaclargs arg = {
6002                 .fh             = NFS_FH(inode),
6003                 .acl_pages      = pages,
6004                 .acl_len        = buflen,
6005         };
6006         struct nfs_setaclres res;
6007         struct rpc_message msg = {
6008                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6009                 .rpc_argp       = &arg,
6010                 .rpc_resp       = &res,
6011         };
6012         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6013         int ret, i;
6014
6015         /* You can't remove system.nfs4_acl: */
6016         if (buflen == 0)
6017                 return -EINVAL;
6018         if (!nfs4_server_supports_acls(server))
6019                 return -EOPNOTSUPP;
6020         if (npages > ARRAY_SIZE(pages))
6021                 return -ERANGE;
6022         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6023         if (i < 0)
6024                 return i;
6025         nfs4_inode_make_writeable(inode);
6026         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6027
6028         /*
6029          * Free each page after tx, so the only ref left is
6030          * held by the network stack
6031          */
6032         for (; i > 0; i--)
6033                 put_page(pages[i-1]);
6034
6035         /*
6036          * Acl update can result in inode attribute update.
6037          * so mark the attribute cache invalid.
6038          */
6039         spin_lock(&inode->i_lock);
6040         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6041                                              NFS_INO_INVALID_CTIME |
6042                                              NFS_INO_REVAL_FORCED);
6043         spin_unlock(&inode->i_lock);
6044         nfs_access_zap_cache(inode);
6045         nfs_zap_acl_cache(inode);
6046         return ret;
6047 }
6048
6049 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
6050 {
6051         struct nfs4_exception exception = { };
6052         int err;
6053         do {
6054                 err = __nfs4_proc_set_acl(inode, buf, buflen);
6055                 trace_nfs4_set_acl(inode, err);
6056                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6057                         /*
6058                          * no need to retry since the kernel
6059                          * isn't involved in encoding the ACEs.
6060                          */
6061                         err = -EINVAL;
6062                         break;
6063                 }
6064                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6065                                 &exception);
6066         } while (exception.retry);
6067         return err;
6068 }
6069
6070 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6071 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6072                                         size_t buflen)
6073 {
6074         struct nfs_server *server = NFS_SERVER(inode);
6075         struct nfs4_label label = {0, 0, buflen, buf};
6076
6077         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6078         struct nfs_fattr fattr = {
6079                 .label = &label,
6080         };
6081         struct nfs4_getattr_arg arg = {
6082                 .fh             = NFS_FH(inode),
6083                 .bitmask        = bitmask,
6084         };
6085         struct nfs4_getattr_res res = {
6086                 .fattr          = &fattr,
6087                 .server         = server,
6088         };
6089         struct rpc_message msg = {
6090                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6091                 .rpc_argp       = &arg,
6092                 .rpc_resp       = &res,
6093         };
6094         int ret;
6095
6096         nfs_fattr_init(&fattr);
6097
6098         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6099         if (ret)
6100                 return ret;
6101         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6102                 return -ENOENT;
6103         return label.len;
6104 }
6105
6106 static int nfs4_get_security_label(struct inode *inode, void *buf,
6107                                         size_t buflen)
6108 {
6109         struct nfs4_exception exception = {
6110                 .interruptible = true,
6111         };
6112         int err;
6113
6114         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6115                 return -EOPNOTSUPP;
6116
6117         do {
6118                 err = _nfs4_get_security_label(inode, buf, buflen);
6119                 trace_nfs4_get_security_label(inode, err);
6120                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6121                                 &exception);
6122         } while (exception.retry);
6123         return err;
6124 }
6125
6126 static int _nfs4_do_set_security_label(struct inode *inode,
6127                 struct nfs4_label *ilabel,
6128                 struct nfs_fattr *fattr)
6129 {
6130
6131         struct iattr sattr = {0};
6132         struct nfs_server *server = NFS_SERVER(inode);
6133         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6134         struct nfs_setattrargs arg = {
6135                 .fh             = NFS_FH(inode),
6136                 .iap            = &sattr,
6137                 .server         = server,
6138                 .bitmask        = bitmask,
6139                 .label          = ilabel,
6140         };
6141         struct nfs_setattrres res = {
6142                 .fattr          = fattr,
6143                 .server         = server,
6144         };
6145         struct rpc_message msg = {
6146                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6147                 .rpc_argp       = &arg,
6148                 .rpc_resp       = &res,
6149         };
6150         int status;
6151
6152         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6153
6154         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6155         if (status)
6156                 dprintk("%s failed: %d\n", __func__, status);
6157
6158         return status;
6159 }
6160
6161 static int nfs4_do_set_security_label(struct inode *inode,
6162                 struct nfs4_label *ilabel,
6163                 struct nfs_fattr *fattr)
6164 {
6165         struct nfs4_exception exception = { };
6166         int err;
6167
6168         do {
6169                 err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6170                 trace_nfs4_set_security_label(inode, err);
6171                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6172                                 &exception);
6173         } while (exception.retry);
6174         return err;
6175 }
6176
6177 static int
6178 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6179 {
6180         struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6181         struct nfs_fattr *fattr;
6182         int status;
6183
6184         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6185                 return -EOPNOTSUPP;
6186
6187         fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6188         if (fattr == NULL)
6189                 return -ENOMEM;
6190
6191         status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6192         if (status == 0)
6193                 nfs_setsecurity(inode, fattr);
6194
6195         return status;
6196 }
6197 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6198
6199
6200 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6201                                     nfs4_verifier *bootverf)
6202 {
6203         __be32 verf[2];
6204
6205         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6206                 /* An impossible timestamp guarantees this value
6207                  * will never match a generated boot time. */
6208                 verf[0] = cpu_to_be32(U32_MAX);
6209                 verf[1] = cpu_to_be32(U32_MAX);
6210         } else {
6211                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6212                 u64 ns = ktime_to_ns(nn->boot_time);
6213
6214                 verf[0] = cpu_to_be32(ns >> 32);
6215                 verf[1] = cpu_to_be32(ns);
6216         }
6217         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6218 }
6219
6220 static size_t
6221 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6222 {
6223         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6224         struct nfs_netns_client *nn_clp = nn->nfs_client;
6225         const char *id;
6226
6227         buf[0] = '\0';
6228
6229         if (nn_clp) {
6230                 rcu_read_lock();
6231                 id = rcu_dereference(nn_clp->identifier);
6232                 if (id)
6233                         strscpy(buf, id, buflen);
6234                 rcu_read_unlock();
6235         }
6236
6237         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6238                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6239
6240         return strlen(buf);
6241 }
6242
6243 static int
6244 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6245 {
6246         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6247         size_t buflen;
6248         size_t len;
6249         char *str;
6250
6251         if (clp->cl_owner_id != NULL)
6252                 return 0;
6253
6254         rcu_read_lock();
6255         len = 14 +
6256                 strlen(clp->cl_rpcclient->cl_nodename) +
6257                 1 +
6258                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6259                 1;
6260         rcu_read_unlock();
6261
6262         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6263         if (buflen)
6264                 len += buflen + 1;
6265
6266         if (len > NFS4_OPAQUE_LIMIT + 1)
6267                 return -EINVAL;
6268
6269         /*
6270          * Since this string is allocated at mount time, and held until the
6271          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6272          * about a memory-reclaim deadlock.
6273          */
6274         str = kmalloc(len, GFP_KERNEL);
6275         if (!str)
6276                 return -ENOMEM;
6277
6278         rcu_read_lock();
6279         if (buflen)
6280                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6281                           clp->cl_rpcclient->cl_nodename, buf,
6282                           rpc_peeraddr2str(clp->cl_rpcclient,
6283                                            RPC_DISPLAY_ADDR));
6284         else
6285                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6286                           clp->cl_rpcclient->cl_nodename,
6287                           rpc_peeraddr2str(clp->cl_rpcclient,
6288                                            RPC_DISPLAY_ADDR));
6289         rcu_read_unlock();
6290
6291         clp->cl_owner_id = str;
6292         return 0;
6293 }
6294
6295 static int
6296 nfs4_init_uniform_client_string(struct nfs_client *clp)
6297 {
6298         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6299         size_t buflen;
6300         size_t len;
6301         char *str;
6302
6303         if (clp->cl_owner_id != NULL)
6304                 return 0;
6305
6306         len = 10 + 10 + 1 + 10 + 1 +
6307                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6308
6309         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6310         if (buflen)
6311                 len += buflen + 1;
6312
6313         if (len > NFS4_OPAQUE_LIMIT + 1)
6314                 return -EINVAL;
6315
6316         /*
6317          * Since this string is allocated at mount time, and held until the
6318          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6319          * about a memory-reclaim deadlock.
6320          */
6321         str = kmalloc(len, GFP_KERNEL);
6322         if (!str)
6323                 return -ENOMEM;
6324
6325         if (buflen)
6326                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6327                           clp->rpc_ops->version, clp->cl_minorversion,
6328                           buf, clp->cl_rpcclient->cl_nodename);
6329         else
6330                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6331                           clp->rpc_ops->version, clp->cl_minorversion,
6332                           clp->cl_rpcclient->cl_nodename);
6333         clp->cl_owner_id = str;
6334         return 0;
6335 }
6336
6337 /*
6338  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6339  * services.  Advertise one based on the address family of the
6340  * clientaddr.
6341  */
6342 static unsigned int
6343 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6344 {
6345         if (strchr(clp->cl_ipaddr, ':') != NULL)
6346                 return scnprintf(buf, len, "tcp6");
6347         else
6348                 return scnprintf(buf, len, "tcp");
6349 }
6350
6351 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6352 {
6353         struct nfs4_setclientid *sc = calldata;
6354
6355         if (task->tk_status == 0)
6356                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6357 }
6358
6359 static const struct rpc_call_ops nfs4_setclientid_ops = {
6360         .rpc_call_done = nfs4_setclientid_done,
6361 };
6362
6363 /**
6364  * nfs4_proc_setclientid - Negotiate client ID
6365  * @clp: state data structure
6366  * @program: RPC program for NFSv4 callback service
6367  * @port: IP port number for NFS4 callback service
6368  * @cred: credential to use for this call
6369  * @res: where to place the result
6370  *
6371  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6372  */
6373 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6374                 unsigned short port, const struct cred *cred,
6375                 struct nfs4_setclientid_res *res)
6376 {
6377         nfs4_verifier sc_verifier;
6378         struct nfs4_setclientid setclientid = {
6379                 .sc_verifier = &sc_verifier,
6380                 .sc_prog = program,
6381                 .sc_clnt = clp,
6382         };
6383         struct rpc_message msg = {
6384                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6385                 .rpc_argp = &setclientid,
6386                 .rpc_resp = res,
6387                 .rpc_cred = cred,
6388         };
6389         struct rpc_task_setup task_setup_data = {
6390                 .rpc_client = clp->cl_rpcclient,
6391                 .rpc_message = &msg,
6392                 .callback_ops = &nfs4_setclientid_ops,
6393                 .callback_data = &setclientid,
6394                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6395         };
6396         unsigned long now = jiffies;
6397         int status;
6398
6399         /* nfs_client_id4 */
6400         nfs4_init_boot_verifier(clp, &sc_verifier);
6401
6402         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6403                 status = nfs4_init_uniform_client_string(clp);
6404         else
6405                 status = nfs4_init_nonuniform_client_string(clp);
6406
6407         if (status)
6408                 goto out;
6409
6410         /* cb_client4 */
6411         setclientid.sc_netid_len =
6412                                 nfs4_init_callback_netid(clp,
6413                                                 setclientid.sc_netid,
6414                                                 sizeof(setclientid.sc_netid));
6415         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6416                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6417                                 clp->cl_ipaddr, port >> 8, port & 255);
6418
6419         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6420                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6421                 clp->cl_owner_id);
6422
6423         status = nfs4_call_sync_custom(&task_setup_data);
6424         if (setclientid.sc_cred) {
6425                 kfree(clp->cl_acceptor);
6426                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6427                 put_rpccred(setclientid.sc_cred);
6428         }
6429
6430         if (status == 0)
6431                 do_renew_lease(clp, now);
6432 out:
6433         trace_nfs4_setclientid(clp, status);
6434         dprintk("NFS reply setclientid: %d\n", status);
6435         return status;
6436 }
6437
6438 /**
6439  * nfs4_proc_setclientid_confirm - Confirm client ID
6440  * @clp: state data structure
6441  * @arg: result of a previous SETCLIENTID
6442  * @cred: credential to use for this call
6443  *
6444  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6445  */
6446 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6447                 struct nfs4_setclientid_res *arg,
6448                 const struct cred *cred)
6449 {
6450         struct rpc_message msg = {
6451                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6452                 .rpc_argp = arg,
6453                 .rpc_cred = cred,
6454         };
6455         int status;
6456
6457         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6458                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6459                 clp->cl_clientid);
6460         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6461                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6462         trace_nfs4_setclientid_confirm(clp, status);
6463         dprintk("NFS reply setclientid_confirm: %d\n", status);
6464         return status;
6465 }
6466
6467 struct nfs4_delegreturndata {
6468         struct nfs4_delegreturnargs args;
6469         struct nfs4_delegreturnres res;
6470         struct nfs_fh fh;
6471         nfs4_stateid stateid;
6472         unsigned long timestamp;
6473         struct {
6474                 struct nfs4_layoutreturn_args arg;
6475                 struct nfs4_layoutreturn_res res;
6476                 struct nfs4_xdr_opaque_data ld_private;
6477                 u32 roc_barrier;
6478                 bool roc;
6479         } lr;
6480         struct nfs_fattr fattr;
6481         int rpc_status;
6482         struct inode *inode;
6483 };
6484
6485 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6486 {
6487         struct nfs4_delegreturndata *data = calldata;
6488         struct nfs4_exception exception = {
6489                 .inode = data->inode,
6490                 .stateid = &data->stateid,
6491                 .task_is_privileged = data->args.seq_args.sa_privileged,
6492         };
6493
6494         if (!nfs4_sequence_done(task, &data->res.seq_res))
6495                 return;
6496
6497         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6498
6499         /* Handle Layoutreturn errors */
6500         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6501                           &data->res.lr_ret) == -EAGAIN)
6502                 goto out_restart;
6503
6504         switch (task->tk_status) {
6505         case 0:
6506                 renew_lease(data->res.server, data->timestamp);
6507                 break;
6508         case -NFS4ERR_ADMIN_REVOKED:
6509         case -NFS4ERR_DELEG_REVOKED:
6510         case -NFS4ERR_EXPIRED:
6511                 nfs4_free_revoked_stateid(data->res.server,
6512                                 data->args.stateid,
6513                                 task->tk_msg.rpc_cred);
6514                 fallthrough;
6515         case -NFS4ERR_BAD_STATEID:
6516         case -NFS4ERR_STALE_STATEID:
6517         case -ETIMEDOUT:
6518                 task->tk_status = 0;
6519                 break;
6520         case -NFS4ERR_OLD_STATEID:
6521                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6522                         nfs4_stateid_seqid_inc(&data->stateid);
6523                 if (data->args.bitmask) {
6524                         data->args.bitmask = NULL;
6525                         data->res.fattr = NULL;
6526                 }
6527                 goto out_restart;
6528         case -NFS4ERR_ACCESS:
6529                 if (data->args.bitmask) {
6530                         data->args.bitmask = NULL;
6531                         data->res.fattr = NULL;
6532                         goto out_restart;
6533                 }
6534                 fallthrough;
6535         default:
6536                 task->tk_status = nfs4_async_handle_exception(task,
6537                                 data->res.server, task->tk_status,
6538                                 &exception);
6539                 if (exception.retry)
6540                         goto out_restart;
6541         }
6542         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6543         data->rpc_status = task->tk_status;
6544         return;
6545 out_restart:
6546         task->tk_status = 0;
6547         rpc_restart_call_prepare(task);
6548 }
6549
6550 static void nfs4_delegreturn_release(void *calldata)
6551 {
6552         struct nfs4_delegreturndata *data = calldata;
6553         struct inode *inode = data->inode;
6554
6555         if (data->lr.roc)
6556                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6557                                  data->res.lr_ret);
6558         if (inode) {
6559                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6560                 nfs_iput_and_deactive(inode);
6561         }
6562         kfree(calldata);
6563 }
6564
6565 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6566 {
6567         struct nfs4_delegreturndata *d_data;
6568         struct pnfs_layout_hdr *lo;
6569
6570         d_data = (struct nfs4_delegreturndata *)data;
6571
6572         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6573                 nfs4_sequence_done(task, &d_data->res.seq_res);
6574                 return;
6575         }
6576
6577         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6578         if (lo && !pnfs_layout_is_valid(lo)) {
6579                 d_data->args.lr_args = NULL;
6580                 d_data->res.lr_res = NULL;
6581         }
6582
6583         nfs4_setup_sequence(d_data->res.server->nfs_client,
6584                         &d_data->args.seq_args,
6585                         &d_data->res.seq_res,
6586                         task);
6587 }
6588
6589 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6590         .rpc_call_prepare = nfs4_delegreturn_prepare,
6591         .rpc_call_done = nfs4_delegreturn_done,
6592         .rpc_release = nfs4_delegreturn_release,
6593 };
6594
6595 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6596 {
6597         struct nfs4_delegreturndata *data;
6598         struct nfs_server *server = NFS_SERVER(inode);
6599         struct rpc_task *task;
6600         struct rpc_message msg = {
6601                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6602                 .rpc_cred = cred,
6603         };
6604         struct rpc_task_setup task_setup_data = {
6605                 .rpc_client = server->client,
6606                 .rpc_message = &msg,
6607                 .callback_ops = &nfs4_delegreturn_ops,
6608                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
6609         };
6610         int status = 0;
6611
6612         data = kzalloc(sizeof(*data), GFP_NOFS);
6613         if (data == NULL)
6614                 return -ENOMEM;
6615
6616         nfs4_state_protect(server->nfs_client,
6617                         NFS_SP4_MACH_CRED_CLEANUP,
6618                         &task_setup_data.rpc_client, &msg);
6619
6620         data->args.fhandle = &data->fh;
6621         data->args.stateid = &data->stateid;
6622         nfs4_bitmask_set(data->args.bitmask_store,
6623                          server->cache_consistency_bitmask, inode, 0);
6624         data->args.bitmask = data->args.bitmask_store;
6625         nfs_copy_fh(&data->fh, NFS_FH(inode));
6626         nfs4_stateid_copy(&data->stateid, stateid);
6627         data->res.fattr = &data->fattr;
6628         data->res.server = server;
6629         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6630         data->lr.arg.ld_private = &data->lr.ld_private;
6631         nfs_fattr_init(data->res.fattr);
6632         data->timestamp = jiffies;
6633         data->rpc_status = 0;
6634         data->inode = nfs_igrab_and_active(inode);
6635         if (data->inode || issync) {
6636                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6637                                         cred);
6638                 if (data->lr.roc) {
6639                         data->args.lr_args = &data->lr.arg;
6640                         data->res.lr_res = &data->lr.res;
6641                 }
6642         }
6643
6644         if (!data->inode)
6645                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6646                                    1);
6647         else
6648                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6649                                    0);
6650         task_setup_data.callback_data = data;
6651         msg.rpc_argp = &data->args;
6652         msg.rpc_resp = &data->res;
6653         task = rpc_run_task(&task_setup_data);
6654         if (IS_ERR(task))
6655                 return PTR_ERR(task);
6656         if (!issync)
6657                 goto out;
6658         status = rpc_wait_for_completion_task(task);
6659         if (status != 0)
6660                 goto out;
6661         status = data->rpc_status;
6662 out:
6663         rpc_put_task(task);
6664         return status;
6665 }
6666
6667 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6668 {
6669         struct nfs_server *server = NFS_SERVER(inode);
6670         struct nfs4_exception exception = { };
6671         int err;
6672         do {
6673                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6674                 trace_nfs4_delegreturn(inode, stateid, err);
6675                 switch (err) {
6676                         case -NFS4ERR_STALE_STATEID:
6677                         case -NFS4ERR_EXPIRED:
6678                         case 0:
6679                                 return 0;
6680                 }
6681                 err = nfs4_handle_exception(server, err, &exception);
6682         } while (exception.retry);
6683         return err;
6684 }
6685
6686 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6687 {
6688         struct inode *inode = state->inode;
6689         struct nfs_server *server = NFS_SERVER(inode);
6690         struct nfs_client *clp = server->nfs_client;
6691         struct nfs_lockt_args arg = {
6692                 .fh = NFS_FH(inode),
6693                 .fl = request,
6694         };
6695         struct nfs_lockt_res res = {
6696                 .denied = request,
6697         };
6698         struct rpc_message msg = {
6699                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6700                 .rpc_argp       = &arg,
6701                 .rpc_resp       = &res,
6702                 .rpc_cred       = state->owner->so_cred,
6703         };
6704         struct nfs4_lock_state *lsp;
6705         int status;
6706
6707         arg.lock_owner.clientid = clp->cl_clientid;
6708         status = nfs4_set_lock_state(state, request);
6709         if (status != 0)
6710                 goto out;
6711         lsp = request->fl_u.nfs4_fl.owner;
6712         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6713         arg.lock_owner.s_dev = server->s_dev;
6714         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6715         switch (status) {
6716                 case 0:
6717                         request->fl_type = F_UNLCK;
6718                         break;
6719                 case -NFS4ERR_DENIED:
6720                         status = 0;
6721         }
6722         request->fl_ops->fl_release_private(request);
6723         request->fl_ops = NULL;
6724 out:
6725         return status;
6726 }
6727
6728 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6729 {
6730         struct nfs4_exception exception = {
6731                 .interruptible = true,
6732         };
6733         int err;
6734
6735         do {
6736                 err = _nfs4_proc_getlk(state, cmd, request);
6737                 trace_nfs4_get_lock(request, state, cmd, err);
6738                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6739                                 &exception);
6740         } while (exception.retry);
6741         return err;
6742 }
6743
6744 /*
6745  * Update the seqid of a lock stateid after receiving
6746  * NFS4ERR_OLD_STATEID
6747  */
6748 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6749                 struct nfs4_lock_state *lsp)
6750 {
6751         struct nfs4_state *state = lsp->ls_state;
6752         bool ret = false;
6753
6754         spin_lock(&state->state_lock);
6755         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6756                 goto out;
6757         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6758                 nfs4_stateid_seqid_inc(dst);
6759         else
6760                 dst->seqid = lsp->ls_stateid.seqid;
6761         ret = true;
6762 out:
6763         spin_unlock(&state->state_lock);
6764         return ret;
6765 }
6766
6767 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6768                 struct nfs4_lock_state *lsp)
6769 {
6770         struct nfs4_state *state = lsp->ls_state;
6771         bool ret;
6772
6773         spin_lock(&state->state_lock);
6774         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6775         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6776         spin_unlock(&state->state_lock);
6777         return ret;
6778 }
6779
6780 struct nfs4_unlockdata {
6781         struct nfs_locku_args arg;
6782         struct nfs_locku_res res;
6783         struct nfs4_lock_state *lsp;
6784         struct nfs_open_context *ctx;
6785         struct nfs_lock_context *l_ctx;
6786         struct file_lock fl;
6787         struct nfs_server *server;
6788         unsigned long timestamp;
6789 };
6790
6791 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6792                 struct nfs_open_context *ctx,
6793                 struct nfs4_lock_state *lsp,
6794                 struct nfs_seqid *seqid)
6795 {
6796         struct nfs4_unlockdata *p;
6797         struct nfs4_state *state = lsp->ls_state;
6798         struct inode *inode = state->inode;
6799
6800         p = kzalloc(sizeof(*p), GFP_NOFS);
6801         if (p == NULL)
6802                 return NULL;
6803         p->arg.fh = NFS_FH(inode);
6804         p->arg.fl = &p->fl;
6805         p->arg.seqid = seqid;
6806         p->res.seqid = seqid;
6807         p->lsp = lsp;
6808         /* Ensure we don't close file until we're done freeing locks! */
6809         p->ctx = get_nfs_open_context(ctx);
6810         p->l_ctx = nfs_get_lock_context(ctx);
6811         locks_init_lock(&p->fl);
6812         locks_copy_lock(&p->fl, fl);
6813         p->server = NFS_SERVER(inode);
6814         spin_lock(&state->state_lock);
6815         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6816         spin_unlock(&state->state_lock);
6817         return p;
6818 }
6819
6820 static void nfs4_locku_release_calldata(void *data)
6821 {
6822         struct nfs4_unlockdata *calldata = data;
6823         nfs_free_seqid(calldata->arg.seqid);
6824         nfs4_put_lock_state(calldata->lsp);
6825         nfs_put_lock_context(calldata->l_ctx);
6826         put_nfs_open_context(calldata->ctx);
6827         kfree(calldata);
6828 }
6829
6830 static void nfs4_locku_done(struct rpc_task *task, void *data)
6831 {
6832         struct nfs4_unlockdata *calldata = data;
6833         struct nfs4_exception exception = {
6834                 .inode = calldata->lsp->ls_state->inode,
6835                 .stateid = &calldata->arg.stateid,
6836         };
6837
6838         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6839                 return;
6840         switch (task->tk_status) {
6841                 case 0:
6842                         renew_lease(calldata->server, calldata->timestamp);
6843                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6844                         if (nfs4_update_lock_stateid(calldata->lsp,
6845                                         &calldata->res.stateid))
6846                                 break;
6847                         fallthrough;
6848                 case -NFS4ERR_ADMIN_REVOKED:
6849                 case -NFS4ERR_EXPIRED:
6850                         nfs4_free_revoked_stateid(calldata->server,
6851                                         &calldata->arg.stateid,
6852                                         task->tk_msg.rpc_cred);
6853                         fallthrough;
6854                 case -NFS4ERR_BAD_STATEID:
6855                 case -NFS4ERR_STALE_STATEID:
6856                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6857                                                 calldata->lsp))
6858                                 rpc_restart_call_prepare(task);
6859                         break;
6860                 case -NFS4ERR_OLD_STATEID:
6861                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6862                                                 calldata->lsp))
6863                                 rpc_restart_call_prepare(task);
6864                         break;
6865                 default:
6866                         task->tk_status = nfs4_async_handle_exception(task,
6867                                         calldata->server, task->tk_status,
6868                                         &exception);
6869                         if (exception.retry)
6870                                 rpc_restart_call_prepare(task);
6871         }
6872         nfs_release_seqid(calldata->arg.seqid);
6873 }
6874
6875 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6876 {
6877         struct nfs4_unlockdata *calldata = data;
6878
6879         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6880                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6881                 return;
6882
6883         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6884                 goto out_wait;
6885         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6886                 /* Note: exit _without_ running nfs4_locku_done */
6887                 goto out_no_action;
6888         }
6889         calldata->timestamp = jiffies;
6890         if (nfs4_setup_sequence(calldata->server->nfs_client,
6891                                 &calldata->arg.seq_args,
6892                                 &calldata->res.seq_res,
6893                                 task) != 0)
6894                 nfs_release_seqid(calldata->arg.seqid);
6895         return;
6896 out_no_action:
6897         task->tk_action = NULL;
6898 out_wait:
6899         nfs4_sequence_done(task, &calldata->res.seq_res);
6900 }
6901
6902 static const struct rpc_call_ops nfs4_locku_ops = {
6903         .rpc_call_prepare = nfs4_locku_prepare,
6904         .rpc_call_done = nfs4_locku_done,
6905         .rpc_release = nfs4_locku_release_calldata,
6906 };
6907
6908 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6909                 struct nfs_open_context *ctx,
6910                 struct nfs4_lock_state *lsp,
6911                 struct nfs_seqid *seqid)
6912 {
6913         struct nfs4_unlockdata *data;
6914         struct rpc_message msg = {
6915                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6916                 .rpc_cred = ctx->cred,
6917         };
6918         struct rpc_task_setup task_setup_data = {
6919                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6920                 .rpc_message = &msg,
6921                 .callback_ops = &nfs4_locku_ops,
6922                 .workqueue = nfsiod_workqueue,
6923                 .flags = RPC_TASK_ASYNC,
6924         };
6925         struct nfs_client *client =
6926                 NFS_SERVER(lsp->ls_state->inode)->nfs_client;
6927
6928         if (client->cl_minorversion)
6929                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6930
6931         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6932                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6933
6934         /* Ensure this is an unlock - when canceling a lock, the
6935          * canceled lock is passed in, and it won't be an unlock.
6936          */
6937         fl->fl_type = F_UNLCK;
6938         if (fl->fl_flags & FL_CLOSE)
6939                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6940
6941         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6942         if (data == NULL) {
6943                 nfs_free_seqid(seqid);
6944                 return ERR_PTR(-ENOMEM);
6945         }
6946
6947         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6948         msg.rpc_argp = &data->arg;
6949         msg.rpc_resp = &data->res;
6950         task_setup_data.callback_data = data;
6951         return rpc_run_task(&task_setup_data);
6952 }
6953
6954 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6955 {
6956         struct inode *inode = state->inode;
6957         struct nfs4_state_owner *sp = state->owner;
6958         struct nfs_inode *nfsi = NFS_I(inode);
6959         struct nfs_seqid *seqid;
6960         struct nfs4_lock_state *lsp;
6961         struct rpc_task *task;
6962         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6963         int status = 0;
6964         unsigned char fl_flags = request->fl_flags;
6965
6966         status = nfs4_set_lock_state(state, request);
6967         /* Unlock _before_ we do the RPC call */
6968         request->fl_flags |= FL_EXISTS;
6969         /* Exclude nfs_delegation_claim_locks() */
6970         mutex_lock(&sp->so_delegreturn_mutex);
6971         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6972         down_read(&nfsi->rwsem);
6973         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6974                 up_read(&nfsi->rwsem);
6975                 mutex_unlock(&sp->so_delegreturn_mutex);
6976                 goto out;
6977         }
6978         up_read(&nfsi->rwsem);
6979         mutex_unlock(&sp->so_delegreturn_mutex);
6980         if (status != 0)
6981                 goto out;
6982         /* Is this a delegated lock? */
6983         lsp = request->fl_u.nfs4_fl.owner;
6984         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6985                 goto out;
6986         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6987         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6988         status = -ENOMEM;
6989         if (IS_ERR(seqid))
6990                 goto out;
6991         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6992         status = PTR_ERR(task);
6993         if (IS_ERR(task))
6994                 goto out;
6995         status = rpc_wait_for_completion_task(task);
6996         rpc_put_task(task);
6997 out:
6998         request->fl_flags = fl_flags;
6999         trace_nfs4_unlock(request, state, F_SETLK, status);
7000         return status;
7001 }
7002
7003 struct nfs4_lockdata {
7004         struct nfs_lock_args arg;
7005         struct nfs_lock_res res;
7006         struct nfs4_lock_state *lsp;
7007         struct nfs_open_context *ctx;
7008         struct file_lock fl;
7009         unsigned long timestamp;
7010         int rpc_status;
7011         int cancelled;
7012         struct nfs_server *server;
7013 };
7014
7015 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7016                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7017                 gfp_t gfp_mask)
7018 {
7019         struct nfs4_lockdata *p;
7020         struct inode *inode = lsp->ls_state->inode;
7021         struct nfs_server *server = NFS_SERVER(inode);
7022         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7023
7024         p = kzalloc(sizeof(*p), gfp_mask);
7025         if (p == NULL)
7026                 return NULL;
7027
7028         p->arg.fh = NFS_FH(inode);
7029         p->arg.fl = &p->fl;
7030         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7031         if (IS_ERR(p->arg.open_seqid))
7032                 goto out_free;
7033         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7034         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7035         if (IS_ERR(p->arg.lock_seqid))
7036                 goto out_free_seqid;
7037         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7038         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7039         p->arg.lock_owner.s_dev = server->s_dev;
7040         p->res.lock_seqid = p->arg.lock_seqid;
7041         p->lsp = lsp;
7042         p->server = server;
7043         p->ctx = get_nfs_open_context(ctx);
7044         locks_init_lock(&p->fl);
7045         locks_copy_lock(&p->fl, fl);
7046         return p;
7047 out_free_seqid:
7048         nfs_free_seqid(p->arg.open_seqid);
7049 out_free:
7050         kfree(p);
7051         return NULL;
7052 }
7053
7054 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7055 {
7056         struct nfs4_lockdata *data = calldata;
7057         struct nfs4_state *state = data->lsp->ls_state;
7058
7059         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7060                 goto out_wait;
7061         /* Do we need to do an open_to_lock_owner? */
7062         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7063                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7064                         goto out_release_lock_seqid;
7065                 }
7066                 nfs4_stateid_copy(&data->arg.open_stateid,
7067                                 &state->open_stateid);
7068                 data->arg.new_lock_owner = 1;
7069                 data->res.open_seqid = data->arg.open_seqid;
7070         } else {
7071                 data->arg.new_lock_owner = 0;
7072                 nfs4_stateid_copy(&data->arg.lock_stateid,
7073                                 &data->lsp->ls_stateid);
7074         }
7075         if (!nfs4_valid_open_stateid(state)) {
7076                 data->rpc_status = -EBADF;
7077                 task->tk_action = NULL;
7078                 goto out_release_open_seqid;
7079         }
7080         data->timestamp = jiffies;
7081         if (nfs4_setup_sequence(data->server->nfs_client,
7082                                 &data->arg.seq_args,
7083                                 &data->res.seq_res,
7084                                 task) == 0)
7085                 return;
7086 out_release_open_seqid:
7087         nfs_release_seqid(data->arg.open_seqid);
7088 out_release_lock_seqid:
7089         nfs_release_seqid(data->arg.lock_seqid);
7090 out_wait:
7091         nfs4_sequence_done(task, &data->res.seq_res);
7092         dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7093 }
7094
7095 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7096 {
7097         struct nfs4_lockdata *data = calldata;
7098         struct nfs4_lock_state *lsp = data->lsp;
7099
7100         if (!nfs4_sequence_done(task, &data->res.seq_res))
7101                 return;
7102
7103         data->rpc_status = task->tk_status;
7104         switch (task->tk_status) {
7105         case 0:
7106                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7107                                 data->timestamp);
7108                 if (data->arg.new_lock && !data->cancelled) {
7109                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7110                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7111                                 goto out_restart;
7112                 }
7113                 if (data->arg.new_lock_owner != 0) {
7114                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7115                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7116                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7117                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7118                         goto out_restart;
7119                 break;
7120         case -NFS4ERR_BAD_STATEID:
7121         case -NFS4ERR_OLD_STATEID:
7122         case -NFS4ERR_STALE_STATEID:
7123         case -NFS4ERR_EXPIRED:
7124                 if (data->arg.new_lock_owner != 0) {
7125                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7126                                                 &lsp->ls_state->open_stateid))
7127                                 goto out_restart;
7128                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7129                                                 &lsp->ls_stateid))
7130                                 goto out_restart;
7131         }
7132 out_done:
7133         dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7134         return;
7135 out_restart:
7136         if (!data->cancelled)
7137                 rpc_restart_call_prepare(task);
7138         goto out_done;
7139 }
7140
7141 static void nfs4_lock_release(void *calldata)
7142 {
7143         struct nfs4_lockdata *data = calldata;
7144
7145         nfs_free_seqid(data->arg.open_seqid);
7146         if (data->cancelled && data->rpc_status == 0) {
7147                 struct rpc_task *task;
7148                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7149                                 data->arg.lock_seqid);
7150                 if (!IS_ERR(task))
7151                         rpc_put_task_async(task);
7152                 dprintk("%s: cancelling lock!\n", __func__);
7153         } else
7154                 nfs_free_seqid(data->arg.lock_seqid);
7155         nfs4_put_lock_state(data->lsp);
7156         put_nfs_open_context(data->ctx);
7157         kfree(data);
7158 }
7159
7160 static const struct rpc_call_ops nfs4_lock_ops = {
7161         .rpc_call_prepare = nfs4_lock_prepare,
7162         .rpc_call_done = nfs4_lock_done,
7163         .rpc_release = nfs4_lock_release,
7164 };
7165
7166 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7167 {
7168         switch (error) {
7169         case -NFS4ERR_ADMIN_REVOKED:
7170         case -NFS4ERR_EXPIRED:
7171         case -NFS4ERR_BAD_STATEID:
7172                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7173                 if (new_lock_owner != 0 ||
7174                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7175                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7176                 break;
7177         case -NFS4ERR_STALE_STATEID:
7178                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7179                 nfs4_schedule_lease_recovery(server->nfs_client);
7180         }
7181 }
7182
7183 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7184 {
7185         struct nfs4_lockdata *data;
7186         struct rpc_task *task;
7187         struct rpc_message msg = {
7188                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7189                 .rpc_cred = state->owner->so_cred,
7190         };
7191         struct rpc_task_setup task_setup_data = {
7192                 .rpc_client = NFS_CLIENT(state->inode),
7193                 .rpc_message = &msg,
7194                 .callback_ops = &nfs4_lock_ops,
7195                 .workqueue = nfsiod_workqueue,
7196                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7197         };
7198         int ret;
7199         struct nfs_client *client = NFS_SERVER(state->inode)->nfs_client;
7200
7201         if (client->cl_minorversion)
7202                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7203
7204         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7205                         fl->fl_u.nfs4_fl.owner,
7206                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7207         if (data == NULL)
7208                 return -ENOMEM;
7209         if (IS_SETLKW(cmd))
7210                 data->arg.block = 1;
7211         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7212                                 recovery_type > NFS_LOCK_NEW);
7213         msg.rpc_argp = &data->arg;
7214         msg.rpc_resp = &data->res;
7215         task_setup_data.callback_data = data;
7216         if (recovery_type > NFS_LOCK_NEW) {
7217                 if (recovery_type == NFS_LOCK_RECLAIM)
7218                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7219         } else
7220                 data->arg.new_lock = 1;
7221         task = rpc_run_task(&task_setup_data);
7222         if (IS_ERR(task))
7223                 return PTR_ERR(task);
7224         ret = rpc_wait_for_completion_task(task);
7225         if (ret == 0) {
7226                 ret = data->rpc_status;
7227                 if (ret)
7228                         nfs4_handle_setlk_error(data->server, data->lsp,
7229                                         data->arg.new_lock_owner, ret);
7230         } else
7231                 data->cancelled = true;
7232         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7233         rpc_put_task(task);
7234         dprintk("%s: ret = %d\n", __func__, ret);
7235         return ret;
7236 }
7237
7238 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7239 {
7240         struct nfs_server *server = NFS_SERVER(state->inode);
7241         struct nfs4_exception exception = {
7242                 .inode = state->inode,
7243         };
7244         int err;
7245
7246         do {
7247                 /* Cache the lock if possible... */
7248                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7249                         return 0;
7250                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7251                 if (err != -NFS4ERR_DELAY)
7252                         break;
7253                 nfs4_handle_exception(server, err, &exception);
7254         } while (exception.retry);
7255         return err;
7256 }
7257
7258 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7259 {
7260         struct nfs_server *server = NFS_SERVER(state->inode);
7261         struct nfs4_exception exception = {
7262                 .inode = state->inode,
7263         };
7264         int err;
7265
7266         err = nfs4_set_lock_state(state, request);
7267         if (err != 0)
7268                 return err;
7269         if (!recover_lost_locks) {
7270                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7271                 return 0;
7272         }
7273         do {
7274                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7275                         return 0;
7276                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7277                 switch (err) {
7278                 default:
7279                         goto out;
7280                 case -NFS4ERR_GRACE:
7281                 case -NFS4ERR_DELAY:
7282                         nfs4_handle_exception(server, err, &exception);
7283                         err = 0;
7284                 }
7285         } while (exception.retry);
7286 out:
7287         return err;
7288 }
7289
7290 #if defined(CONFIG_NFS_V4_1)
7291 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7292 {
7293         struct nfs4_lock_state *lsp;
7294         int status;
7295
7296         status = nfs4_set_lock_state(state, request);
7297         if (status != 0)
7298                 return status;
7299         lsp = request->fl_u.nfs4_fl.owner;
7300         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7301             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7302                 return 0;
7303         return nfs4_lock_expired(state, request);
7304 }
7305 #endif
7306
7307 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7308 {
7309         struct nfs_inode *nfsi = NFS_I(state->inode);
7310         struct nfs4_state_owner *sp = state->owner;
7311         unsigned char fl_flags = request->fl_flags;
7312         int status;
7313
7314         request->fl_flags |= FL_ACCESS;
7315         status = locks_lock_inode_wait(state->inode, request);
7316         if (status < 0)
7317                 goto out;
7318         mutex_lock(&sp->so_delegreturn_mutex);
7319         down_read(&nfsi->rwsem);
7320         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7321                 /* Yes: cache locks! */
7322                 /* ...but avoid races with delegation recall... */
7323                 request->fl_flags = fl_flags & ~FL_SLEEP;
7324                 status = locks_lock_inode_wait(state->inode, request);
7325                 up_read(&nfsi->rwsem);
7326                 mutex_unlock(&sp->so_delegreturn_mutex);
7327                 goto out;
7328         }
7329         up_read(&nfsi->rwsem);
7330         mutex_unlock(&sp->so_delegreturn_mutex);
7331         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7332 out:
7333         request->fl_flags = fl_flags;
7334         return status;
7335 }
7336
7337 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7338 {
7339         struct nfs4_exception exception = {
7340                 .state = state,
7341                 .inode = state->inode,
7342                 .interruptible = true,
7343         };
7344         int err;
7345
7346         do {
7347                 err = _nfs4_proc_setlk(state, cmd, request);
7348                 if (err == -NFS4ERR_DENIED)
7349                         err = -EAGAIN;
7350                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7351                                 err, &exception);
7352         } while (exception.retry);
7353         return err;
7354 }
7355
7356 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7357 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7358
7359 static int
7360 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7361                         struct file_lock *request)
7362 {
7363         int             status = -ERESTARTSYS;
7364         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7365
7366         while(!signalled()) {
7367                 status = nfs4_proc_setlk(state, cmd, request);
7368                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7369                         break;
7370                 freezable_schedule_timeout_interruptible(timeout);
7371                 timeout *= 2;
7372                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7373                 status = -ERESTARTSYS;
7374         }
7375         return status;
7376 }
7377
7378 #ifdef CONFIG_NFS_V4_1
7379 struct nfs4_lock_waiter {
7380         struct inode            *inode;
7381         struct nfs_lowner       owner;
7382         wait_queue_entry_t      wait;
7383 };
7384
7385 static int
7386 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7387 {
7388         struct nfs4_lock_waiter *waiter =
7389                 container_of(wait, struct nfs4_lock_waiter, wait);
7390
7391         /* NULL key means to wake up everyone */
7392         if (key) {
7393                 struct cb_notify_lock_args      *cbnl = key;
7394                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7395                                                 *wowner = &waiter->owner;
7396
7397                 /* Only wake if the callback was for the same owner. */
7398                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7399                         return 0;
7400
7401                 /* Make sure it's for the right inode */
7402                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7403                         return 0;
7404         }
7405
7406         return woken_wake_function(wait, mode, flags, key);
7407 }
7408
7409 static int
7410 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7411 {
7412         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7413         struct nfs_server *server = NFS_SERVER(state->inode);
7414         struct nfs_client *clp = server->nfs_client;
7415         wait_queue_head_t *q = &clp->cl_lock_waitq;
7416         struct nfs4_lock_waiter waiter = {
7417                 .inode = state->inode,
7418                 .owner = { .clientid = clp->cl_clientid,
7419                            .id = lsp->ls_seqid.owner_id,
7420                            .s_dev = server->s_dev },
7421         };
7422         int status;
7423
7424         /* Don't bother with waitqueue if we don't expect a callback */
7425         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7426                 return nfs4_retry_setlk_simple(state, cmd, request);
7427
7428         init_wait(&waiter.wait);
7429         waiter.wait.func = nfs4_wake_lock_waiter;
7430         add_wait_queue(q, &waiter.wait);
7431
7432         do {
7433                 status = nfs4_proc_setlk(state, cmd, request);
7434                 if (status != -EAGAIN || IS_SETLK(cmd))
7435                         break;
7436
7437                 status = -ERESTARTSYS;
7438                 freezer_do_not_count();
7439                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7440                            NFS4_LOCK_MAXTIMEOUT);
7441                 freezer_count();
7442         } while (!signalled());
7443
7444         remove_wait_queue(q, &waiter.wait);
7445
7446         return status;
7447 }
7448 #else /* !CONFIG_NFS_V4_1 */
7449 static inline int
7450 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7451 {
7452         return nfs4_retry_setlk_simple(state, cmd, request);
7453 }
7454 #endif
7455
7456 static int
7457 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7458 {
7459         struct nfs_open_context *ctx;
7460         struct nfs4_state *state;
7461         int status;
7462
7463         /* verify open state */
7464         ctx = nfs_file_open_context(filp);
7465         state = ctx->state;
7466
7467         if (IS_GETLK(cmd)) {
7468                 if (state != NULL)
7469                         return nfs4_proc_getlk(state, F_GETLK, request);
7470                 return 0;
7471         }
7472
7473         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7474                 return -EINVAL;
7475
7476         if (request->fl_type == F_UNLCK) {
7477                 if (state != NULL)
7478                         return nfs4_proc_unlck(state, cmd, request);
7479                 return 0;
7480         }
7481
7482         if (state == NULL)
7483                 return -ENOLCK;
7484
7485         if ((request->fl_flags & FL_POSIX) &&
7486             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7487                 return -ENOLCK;
7488
7489         /*
7490          * Don't rely on the VFS having checked the file open mode,
7491          * since it won't do this for flock() locks.
7492          */
7493         switch (request->fl_type) {
7494         case F_RDLCK:
7495                 if (!(filp->f_mode & FMODE_READ))
7496                         return -EBADF;
7497                 break;
7498         case F_WRLCK:
7499                 if (!(filp->f_mode & FMODE_WRITE))
7500                         return -EBADF;
7501         }
7502
7503         status = nfs4_set_lock_state(state, request);
7504         if (status != 0)
7505                 return status;
7506
7507         return nfs4_retry_setlk(state, cmd, request);
7508 }
7509
7510 static int nfs4_delete_lease(struct file *file, void **priv)
7511 {
7512         return generic_setlease(file, F_UNLCK, NULL, priv);
7513 }
7514
7515 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7516                           void **priv)
7517 {
7518         struct inode *inode = file_inode(file);
7519         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7520         int ret;
7521
7522         /* No delegation, no lease */
7523         if (!nfs4_have_delegation(inode, type))
7524                 return -EAGAIN;
7525         ret = generic_setlease(file, arg, lease, priv);
7526         if (ret || nfs4_have_delegation(inode, type))
7527                 return ret;
7528         /* We raced with a delegation return */
7529         nfs4_delete_lease(file, priv);
7530         return -EAGAIN;
7531 }
7532
7533 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7534                        void **priv)
7535 {
7536         switch (arg) {
7537         case F_RDLCK:
7538         case F_WRLCK:
7539                 return nfs4_add_lease(file, arg, lease, priv);
7540         case F_UNLCK:
7541                 return nfs4_delete_lease(file, priv);
7542         default:
7543                 return -EINVAL;
7544         }
7545 }
7546
7547 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7548 {
7549         struct nfs_server *server = NFS_SERVER(state->inode);
7550         int err;
7551
7552         err = nfs4_set_lock_state(state, fl);
7553         if (err != 0)
7554                 return err;
7555         do {
7556                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7557                 if (err != -NFS4ERR_DELAY)
7558                         break;
7559                 ssleep(1);
7560         } while (err == -NFS4ERR_DELAY);
7561         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7562 }
7563
7564 struct nfs_release_lockowner_data {
7565         struct nfs4_lock_state *lsp;
7566         struct nfs_server *server;
7567         struct nfs_release_lockowner_args args;
7568         struct nfs_release_lockowner_res res;
7569         unsigned long timestamp;
7570 };
7571
7572 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7573 {
7574         struct nfs_release_lockowner_data *data = calldata;
7575         struct nfs_server *server = data->server;
7576         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7577                            &data->res.seq_res, task);
7578         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7579         data->timestamp = jiffies;
7580 }
7581
7582 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7583 {
7584         struct nfs_release_lockowner_data *data = calldata;
7585         struct nfs_server *server = data->server;
7586
7587         nfs40_sequence_done(task, &data->res.seq_res);
7588
7589         switch (task->tk_status) {
7590         case 0:
7591                 renew_lease(server, data->timestamp);
7592                 break;
7593         case -NFS4ERR_STALE_CLIENTID:
7594         case -NFS4ERR_EXPIRED:
7595                 nfs4_schedule_lease_recovery(server->nfs_client);
7596                 break;
7597         case -NFS4ERR_LEASE_MOVED:
7598         case -NFS4ERR_DELAY:
7599                 if (nfs4_async_handle_error(task, server,
7600                                             NULL, NULL) == -EAGAIN)
7601                         rpc_restart_call_prepare(task);
7602         }
7603 }
7604
7605 static void nfs4_release_lockowner_release(void *calldata)
7606 {
7607         struct nfs_release_lockowner_data *data = calldata;
7608         nfs4_free_lock_state(data->server, data->lsp);
7609         kfree(calldata);
7610 }
7611
7612 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7613         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7614         .rpc_call_done = nfs4_release_lockowner_done,
7615         .rpc_release = nfs4_release_lockowner_release,
7616 };
7617
7618 static void
7619 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7620 {
7621         struct nfs_release_lockowner_data *data;
7622         struct rpc_message msg = {
7623                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7624         };
7625
7626         if (server->nfs_client->cl_mvops->minor_version != 0)
7627                 return;
7628
7629         data = kmalloc(sizeof(*data), GFP_NOFS);
7630         if (!data)
7631                 return;
7632         data->lsp = lsp;
7633         data->server = server;
7634         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7635         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7636         data->args.lock_owner.s_dev = server->s_dev;
7637
7638         msg.rpc_argp = &data->args;
7639         msg.rpc_resp = &data->res;
7640         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7641         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7642 }
7643
7644 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7645
7646 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7647                                    struct user_namespace *mnt_userns,
7648                                    struct dentry *unused, struct inode *inode,
7649                                    const char *key, const void *buf,
7650                                    size_t buflen, int flags)
7651 {
7652         return nfs4_proc_set_acl(inode, buf, buflen);
7653 }
7654
7655 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7656                                    struct dentry *unused, struct inode *inode,
7657                                    const char *key, void *buf, size_t buflen)
7658 {
7659         return nfs4_proc_get_acl(inode, buf, buflen);
7660 }
7661
7662 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7663 {
7664         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7665 }
7666
7667 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7668
7669 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7670                                      struct user_namespace *mnt_userns,
7671                                      struct dentry *unused, struct inode *inode,
7672                                      const char *key, const void *buf,
7673                                      size_t buflen, int flags)
7674 {
7675         if (security_ismaclabel(key))
7676                 return nfs4_set_security_label(inode, buf, buflen);
7677
7678         return -EOPNOTSUPP;
7679 }
7680
7681 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7682                                      struct dentry *unused, struct inode *inode,
7683                                      const char *key, void *buf, size_t buflen)
7684 {
7685         if (security_ismaclabel(key))
7686                 return nfs4_get_security_label(inode, buf, buflen);
7687         return -EOPNOTSUPP;
7688 }
7689
7690 static ssize_t
7691 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7692 {
7693         int len = 0;
7694
7695         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7696                 len = security_inode_listsecurity(inode, list, list_len);
7697                 if (len >= 0 && list_len && len > list_len)
7698                         return -ERANGE;
7699         }
7700         return len;
7701 }
7702
7703 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7704         .prefix = XATTR_SECURITY_PREFIX,
7705         .get    = nfs4_xattr_get_nfs4_label,
7706         .set    = nfs4_xattr_set_nfs4_label,
7707 };
7708
7709 #else
7710
7711 static ssize_t
7712 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7713 {
7714         return 0;
7715 }
7716
7717 #endif
7718
7719 #ifdef CONFIG_NFS_V4_2
7720 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7721                                     struct user_namespace *mnt_userns,
7722                                     struct dentry *unused, struct inode *inode,
7723                                     const char *key, const void *buf,
7724                                     size_t buflen, int flags)
7725 {
7726         u32 mask;
7727         int ret;
7728
7729         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7730                 return -EOPNOTSUPP;
7731
7732         /*
7733          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7734          * flags right now. Handling of xattr operations use the normal
7735          * file read/write permissions.
7736          *
7737          * Just in case the server has other ideas (which RFC 8276 allows),
7738          * do a cached access check for the XA* flags to possibly avoid
7739          * doing an RPC and getting EACCES back.
7740          */
7741         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7742                 if (!(mask & NFS_ACCESS_XAWRITE))
7743                         return -EACCES;
7744         }
7745
7746         if (buf == NULL) {
7747                 ret = nfs42_proc_removexattr(inode, key);
7748                 if (!ret)
7749                         nfs4_xattr_cache_remove(inode, key);
7750         } else {
7751                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7752                 if (!ret)
7753                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7754         }
7755
7756         return ret;
7757 }
7758
7759 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7760                                     struct dentry *unused, struct inode *inode,
7761                                     const char *key, void *buf, size_t buflen)
7762 {
7763         u32 mask;
7764         ssize_t ret;
7765
7766         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7767                 return -EOPNOTSUPP;
7768
7769         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7770                 if (!(mask & NFS_ACCESS_XAREAD))
7771                         return -EACCES;
7772         }
7773
7774         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7775         if (ret)
7776                 return ret;
7777
7778         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7779         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7780                 return ret;
7781
7782         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7783
7784         return ret;
7785 }
7786
7787 static ssize_t
7788 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7789 {
7790         u64 cookie;
7791         bool eof;
7792         ssize_t ret, size;
7793         char *buf;
7794         size_t buflen;
7795         u32 mask;
7796
7797         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7798                 return 0;
7799
7800         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7801                 if (!(mask & NFS_ACCESS_XALIST))
7802                         return 0;
7803         }
7804
7805         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7806         if (ret)
7807                 return ret;
7808
7809         ret = nfs4_xattr_cache_list(inode, list, list_len);
7810         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7811                 return ret;
7812
7813         cookie = 0;
7814         eof = false;
7815         buflen = list_len ? list_len : XATTR_LIST_MAX;
7816         buf = list_len ? list : NULL;
7817         size = 0;
7818
7819         while (!eof) {
7820                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7821                     &cookie, &eof);
7822                 if (ret < 0)
7823                         return ret;
7824
7825                 if (list_len) {
7826                         buf += ret;
7827                         buflen -= ret;
7828                 }
7829                 size += ret;
7830         }
7831
7832         if (list_len)
7833                 nfs4_xattr_cache_set_list(inode, list, size);
7834
7835         return size;
7836 }
7837
7838 #else
7839
7840 static ssize_t
7841 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7842 {
7843         return 0;
7844 }
7845 #endif /* CONFIG_NFS_V4_2 */
7846
7847 /*
7848  * nfs_fhget will use either the mounted_on_fileid or the fileid
7849  */
7850 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7851 {
7852         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7853                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7854               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7855               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7856                 return;
7857
7858         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7859                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7860         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7861         fattr->nlink = 2;
7862 }
7863
7864 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7865                                    const struct qstr *name,
7866                                    struct nfs4_fs_locations *fs_locations,
7867                                    struct page *page)
7868 {
7869         struct nfs_server *server = NFS_SERVER(dir);
7870         u32 bitmask[3];
7871         struct nfs4_fs_locations_arg args = {
7872                 .dir_fh = NFS_FH(dir),
7873                 .name = name,
7874                 .page = page,
7875                 .bitmask = bitmask,
7876         };
7877         struct nfs4_fs_locations_res res = {
7878                 .fs_locations = fs_locations,
7879         };
7880         struct rpc_message msg = {
7881                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7882                 .rpc_argp = &args,
7883                 .rpc_resp = &res,
7884         };
7885         int status;
7886
7887         dprintk("%s: start\n", __func__);
7888
7889         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7890         bitmask[1] = nfs4_fattr_bitmap[1];
7891
7892         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7893          * is not supported */
7894         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7895                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7896         else
7897                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7898
7899         nfs_fattr_init(&fs_locations->fattr);
7900         fs_locations->server = server;
7901         fs_locations->nlocations = 0;
7902         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7903         dprintk("%s: returned status = %d\n", __func__, status);
7904         return status;
7905 }
7906
7907 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7908                            const struct qstr *name,
7909                            struct nfs4_fs_locations *fs_locations,
7910                            struct page *page)
7911 {
7912         struct nfs4_exception exception = {
7913                 .interruptible = true,
7914         };
7915         int err;
7916         do {
7917                 err = _nfs4_proc_fs_locations(client, dir, name,
7918                                 fs_locations, page);
7919                 trace_nfs4_get_fs_locations(dir, name, err);
7920                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7921                                 &exception);
7922         } while (exception.retry);
7923         return err;
7924 }
7925
7926 /*
7927  * This operation also signals the server that this client is
7928  * performing migration recovery.  The server can stop returning
7929  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7930  * appended to this compound to identify the client ID which is
7931  * performing recovery.
7932  */
7933 static int _nfs40_proc_get_locations(struct nfs_server *server,
7934                                      struct nfs_fh *fhandle,
7935                                      struct nfs4_fs_locations *locations,
7936                                      struct page *page, const struct cred *cred)
7937 {
7938         struct rpc_clnt *clnt = server->client;
7939         u32 bitmask[2] = {
7940                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7941         };
7942         struct nfs4_fs_locations_arg args = {
7943                 .clientid       = server->nfs_client->cl_clientid,
7944                 .fh             = fhandle,
7945                 .page           = page,
7946                 .bitmask        = bitmask,
7947                 .migration      = 1,            /* skip LOOKUP */
7948                 .renew          = 1,            /* append RENEW */
7949         };
7950         struct nfs4_fs_locations_res res = {
7951                 .fs_locations   = locations,
7952                 .migration      = 1,
7953                 .renew          = 1,
7954         };
7955         struct rpc_message msg = {
7956                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7957                 .rpc_argp       = &args,
7958                 .rpc_resp       = &res,
7959                 .rpc_cred       = cred,
7960         };
7961         unsigned long now = jiffies;
7962         int status;
7963
7964         nfs_fattr_init(&locations->fattr);
7965         locations->server = server;
7966         locations->nlocations = 0;
7967
7968         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7969         status = nfs4_call_sync_sequence(clnt, server, &msg,
7970                                         &args.seq_args, &res.seq_res);
7971         if (status)
7972                 return status;
7973
7974         renew_lease(server, now);
7975         return 0;
7976 }
7977
7978 #ifdef CONFIG_NFS_V4_1
7979
7980 /*
7981  * This operation also signals the server that this client is
7982  * performing migration recovery.  The server can stop asserting
7983  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7984  * performing this operation is identified in the SEQUENCE
7985  * operation in this compound.
7986  *
7987  * When the client supports GETATTR(fs_locations_info), it can
7988  * be plumbed in here.
7989  */
7990 static int _nfs41_proc_get_locations(struct nfs_server *server,
7991                                      struct nfs_fh *fhandle,
7992                                      struct nfs4_fs_locations *locations,
7993                                      struct page *page, const struct cred *cred)
7994 {
7995         struct rpc_clnt *clnt = server->client;
7996         u32 bitmask[2] = {
7997                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7998         };
7999         struct nfs4_fs_locations_arg args = {
8000                 .fh             = fhandle,
8001                 .page           = page,
8002                 .bitmask        = bitmask,
8003                 .migration      = 1,            /* skip LOOKUP */
8004         };
8005         struct nfs4_fs_locations_res res = {
8006                 .fs_locations   = locations,
8007                 .migration      = 1,
8008         };
8009         struct rpc_message msg = {
8010                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8011                 .rpc_argp       = &args,
8012                 .rpc_resp       = &res,
8013                 .rpc_cred       = cred,
8014         };
8015         int status;
8016
8017         nfs_fattr_init(&locations->fattr);
8018         locations->server = server;
8019         locations->nlocations = 0;
8020
8021         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8022         status = nfs4_call_sync_sequence(clnt, server, &msg,
8023                                         &args.seq_args, &res.seq_res);
8024         if (status == NFS4_OK &&
8025             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8026                 status = -NFS4ERR_LEASE_MOVED;
8027         return status;
8028 }
8029
8030 #endif  /* CONFIG_NFS_V4_1 */
8031
8032 /**
8033  * nfs4_proc_get_locations - discover locations for a migrated FSID
8034  * @server: pointer to nfs_server to process
8035  * @fhandle: pointer to the kernel NFS client file handle
8036  * @locations: result of query
8037  * @page: buffer
8038  * @cred: credential to use for this operation
8039  *
8040  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8041  * operation failed, or a negative errno if a local error occurred.
8042  *
8043  * On success, "locations" is filled in, but if the server has
8044  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8045  * asserted.
8046  *
8047  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8048  * from this client that require migration recovery.
8049  */
8050 int nfs4_proc_get_locations(struct nfs_server *server,
8051                             struct nfs_fh *fhandle,
8052                             struct nfs4_fs_locations *locations,
8053                             struct page *page, const struct cred *cred)
8054 {
8055         struct nfs_client *clp = server->nfs_client;
8056         const struct nfs4_mig_recovery_ops *ops =
8057                                         clp->cl_mvops->mig_recovery_ops;
8058         struct nfs4_exception exception = {
8059                 .interruptible = true,
8060         };
8061         int status;
8062
8063         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8064                 (unsigned long long)server->fsid.major,
8065                 (unsigned long long)server->fsid.minor,
8066                 clp->cl_hostname);
8067         nfs_display_fhandle(fhandle, __func__);
8068
8069         do {
8070                 status = ops->get_locations(server, fhandle, locations, page,
8071                                             cred);
8072                 if (status != -NFS4ERR_DELAY)
8073                         break;
8074                 nfs4_handle_exception(server, status, &exception);
8075         } while (exception.retry);
8076         return status;
8077 }
8078
8079 /*
8080  * This operation also signals the server that this client is
8081  * performing "lease moved" recovery.  The server can stop
8082  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8083  * is appended to this compound to identify the client ID which is
8084  * performing recovery.
8085  */
8086 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8087 {
8088         struct nfs_server *server = NFS_SERVER(inode);
8089         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8090         struct rpc_clnt *clnt = server->client;
8091         struct nfs4_fsid_present_arg args = {
8092                 .fh             = NFS_FH(inode),
8093                 .clientid       = clp->cl_clientid,
8094                 .renew          = 1,            /* append RENEW */
8095         };
8096         struct nfs4_fsid_present_res res = {
8097                 .renew          = 1,
8098         };
8099         struct rpc_message msg = {
8100                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8101                 .rpc_argp       = &args,
8102                 .rpc_resp       = &res,
8103                 .rpc_cred       = cred,
8104         };
8105         unsigned long now = jiffies;
8106         int status;
8107
8108         res.fh = nfs_alloc_fhandle();
8109         if (res.fh == NULL)
8110                 return -ENOMEM;
8111
8112         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8113         status = nfs4_call_sync_sequence(clnt, server, &msg,
8114                                                 &args.seq_args, &res.seq_res);
8115         nfs_free_fhandle(res.fh);
8116         if (status)
8117                 return status;
8118
8119         do_renew_lease(clp, now);
8120         return 0;
8121 }
8122
8123 #ifdef CONFIG_NFS_V4_1
8124
8125 /*
8126  * This operation also signals the server that this client is
8127  * performing "lease moved" recovery.  The server can stop asserting
8128  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8129  * this operation is identified in the SEQUENCE operation in this
8130  * compound.
8131  */
8132 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8133 {
8134         struct nfs_server *server = NFS_SERVER(inode);
8135         struct rpc_clnt *clnt = server->client;
8136         struct nfs4_fsid_present_arg args = {
8137                 .fh             = NFS_FH(inode),
8138         };
8139         struct nfs4_fsid_present_res res = {
8140         };
8141         struct rpc_message msg = {
8142                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8143                 .rpc_argp       = &args,
8144                 .rpc_resp       = &res,
8145                 .rpc_cred       = cred,
8146         };
8147         int status;
8148
8149         res.fh = nfs_alloc_fhandle();
8150         if (res.fh == NULL)
8151                 return -ENOMEM;
8152
8153         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8154         status = nfs4_call_sync_sequence(clnt, server, &msg,
8155                                                 &args.seq_args, &res.seq_res);
8156         nfs_free_fhandle(res.fh);
8157         if (status == NFS4_OK &&
8158             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8159                 status = -NFS4ERR_LEASE_MOVED;
8160         return status;
8161 }
8162
8163 #endif  /* CONFIG_NFS_V4_1 */
8164
8165 /**
8166  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8167  * @inode: inode on FSID to check
8168  * @cred: credential to use for this operation
8169  *
8170  * Server indicates whether the FSID is present, moved, or not
8171  * recognized.  This operation is necessary to clear a LEASE_MOVED
8172  * condition for this client ID.
8173  *
8174  * Returns NFS4_OK if the FSID is present on this server,
8175  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8176  *  NFS4ERR code if some error occurred on the server, or a
8177  *  negative errno if a local failure occurred.
8178  */
8179 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8180 {
8181         struct nfs_server *server = NFS_SERVER(inode);
8182         struct nfs_client *clp = server->nfs_client;
8183         const struct nfs4_mig_recovery_ops *ops =
8184                                         clp->cl_mvops->mig_recovery_ops;
8185         struct nfs4_exception exception = {
8186                 .interruptible = true,
8187         };
8188         int status;
8189
8190         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8191                 (unsigned long long)server->fsid.major,
8192                 (unsigned long long)server->fsid.minor,
8193                 clp->cl_hostname);
8194         nfs_display_fhandle(NFS_FH(inode), __func__);
8195
8196         do {
8197                 status = ops->fsid_present(inode, cred);
8198                 if (status != -NFS4ERR_DELAY)
8199                         break;
8200                 nfs4_handle_exception(server, status, &exception);
8201         } while (exception.retry);
8202         return status;
8203 }
8204
8205 /*
8206  * If 'use_integrity' is true and the state managment nfs_client
8207  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8208  * and the machine credential as per RFC3530bis and RFC5661 Security
8209  * Considerations sections. Otherwise, just use the user cred with the
8210  * filesystem's rpc_client.
8211  */
8212 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8213 {
8214         int status;
8215         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8216         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8217         struct nfs4_secinfo_arg args = {
8218                 .dir_fh = NFS_FH(dir),
8219                 .name   = name,
8220         };
8221         struct nfs4_secinfo_res res = {
8222                 .flavors     = flavors,
8223         };
8224         struct rpc_message msg = {
8225                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8226                 .rpc_argp = &args,
8227                 .rpc_resp = &res,
8228         };
8229         struct nfs4_call_sync_data data = {
8230                 .seq_server = NFS_SERVER(dir),
8231                 .seq_args = &args.seq_args,
8232                 .seq_res = &res.seq_res,
8233         };
8234         struct rpc_task_setup task_setup = {
8235                 .rpc_client = clnt,
8236                 .rpc_message = &msg,
8237                 .callback_ops = clp->cl_mvops->call_sync_ops,
8238                 .callback_data = &data,
8239                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8240         };
8241         const struct cred *cred = NULL;
8242
8243         if (use_integrity) {
8244                 clnt = clp->cl_rpcclient;
8245                 task_setup.rpc_client = clnt;
8246
8247                 cred = nfs4_get_clid_cred(clp);
8248                 msg.rpc_cred = cred;
8249         }
8250
8251         dprintk("NFS call  secinfo %s\n", name->name);
8252
8253         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8254         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8255         status = nfs4_call_sync_custom(&task_setup);
8256
8257         dprintk("NFS reply  secinfo: %d\n", status);
8258
8259         put_cred(cred);
8260         return status;
8261 }
8262
8263 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8264                       struct nfs4_secinfo_flavors *flavors)
8265 {
8266         struct nfs4_exception exception = {
8267                 .interruptible = true,
8268         };
8269         int err;
8270         do {
8271                 err = -NFS4ERR_WRONGSEC;
8272
8273                 /* try to use integrity protection with machine cred */
8274                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8275                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8276
8277                 /*
8278                  * if unable to use integrity protection, or SECINFO with
8279                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8280                  * disallowed by spec, but exists in deployed servers) use
8281                  * the current filesystem's rpc_client and the user cred.
8282                  */
8283                 if (err == -NFS4ERR_WRONGSEC)
8284                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8285
8286                 trace_nfs4_secinfo(dir, name, err);
8287                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8288                                 &exception);
8289         } while (exception.retry);
8290         return err;
8291 }
8292
8293 #ifdef CONFIG_NFS_V4_1
8294 /*
8295  * Check the exchange flags returned by the server for invalid flags, having
8296  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8297  * DS flags set.
8298  */
8299 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8300 {
8301         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8302                 goto out_inval;
8303         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8304                 goto out_inval;
8305         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8306             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8307                 goto out_inval;
8308         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8309                 goto out_inval;
8310         return NFS_OK;
8311 out_inval:
8312         return -NFS4ERR_INVAL;
8313 }
8314
8315 static bool
8316 nfs41_same_server_scope(struct nfs41_server_scope *a,
8317                         struct nfs41_server_scope *b)
8318 {
8319         if (a->server_scope_sz != b->server_scope_sz)
8320                 return false;
8321         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8322 }
8323
8324 static void
8325 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8326 {
8327         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8328         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8329         struct nfs_client *clp = args->client;
8330
8331         switch (task->tk_status) {
8332         case -NFS4ERR_BADSESSION:
8333         case -NFS4ERR_DEADSESSION:
8334                 nfs4_schedule_session_recovery(clp->cl_session,
8335                                 task->tk_status);
8336         }
8337         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8338                         res->dir != NFS4_CDFS4_BOTH) {
8339                 rpc_task_close_connection(task);
8340                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8341                         rpc_restart_call(task);
8342         }
8343 }
8344
8345 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8346         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8347 };
8348
8349 /*
8350  * nfs4_proc_bind_one_conn_to_session()
8351  *
8352  * The 4.1 client currently uses the same TCP connection for the
8353  * fore and backchannel.
8354  */
8355 static
8356 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8357                 struct rpc_xprt *xprt,
8358                 struct nfs_client *clp,
8359                 const struct cred *cred)
8360 {
8361         int status;
8362         struct nfs41_bind_conn_to_session_args args = {
8363                 .client = clp,
8364                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8365                 .retries = 0,
8366         };
8367         struct nfs41_bind_conn_to_session_res res;
8368         struct rpc_message msg = {
8369                 .rpc_proc =
8370                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8371                 .rpc_argp = &args,
8372                 .rpc_resp = &res,
8373                 .rpc_cred = cred,
8374         };
8375         struct rpc_task_setup task_setup_data = {
8376                 .rpc_client = clnt,
8377                 .rpc_xprt = xprt,
8378                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8379                 .rpc_message = &msg,
8380                 .flags = RPC_TASK_TIMEOUT,
8381         };
8382         struct rpc_task *task;
8383
8384         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8385         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8386                 args.dir = NFS4_CDFC4_FORE;
8387
8388         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8389         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8390                 args.dir = NFS4_CDFC4_FORE;
8391
8392         task = rpc_run_task(&task_setup_data);
8393         if (!IS_ERR(task)) {
8394                 status = task->tk_status;
8395                 rpc_put_task(task);
8396         } else
8397                 status = PTR_ERR(task);
8398         trace_nfs4_bind_conn_to_session(clp, status);
8399         if (status == 0) {
8400                 if (memcmp(res.sessionid.data,
8401                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8402                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8403                         return -EIO;
8404                 }
8405                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8406                         dprintk("NFS: %s: Unexpected direction from server\n",
8407                                 __func__);
8408                         return -EIO;
8409                 }
8410                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8411                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8412                                 __func__);
8413                         return -EIO;
8414                 }
8415         }
8416
8417         return status;
8418 }
8419
8420 struct rpc_bind_conn_calldata {
8421         struct nfs_client *clp;
8422         const struct cred *cred;
8423 };
8424
8425 static int
8426 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8427                 struct rpc_xprt *xprt,
8428                 void *calldata)
8429 {
8430         struct rpc_bind_conn_calldata *p = calldata;
8431
8432         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8433 }
8434
8435 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8436 {
8437         struct rpc_bind_conn_calldata data = {
8438                 .clp = clp,
8439                 .cred = cred,
8440         };
8441         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8442                         nfs4_proc_bind_conn_to_session_callback, &data);
8443 }
8444
8445 /*
8446  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8447  * and operations we'd like to see to enable certain features in the allow map
8448  */
8449 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8450         .how = SP4_MACH_CRED,
8451         .enforce.u.words = {
8452                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8453                       1 << (OP_EXCHANGE_ID - 32) |
8454                       1 << (OP_CREATE_SESSION - 32) |
8455                       1 << (OP_DESTROY_SESSION - 32) |
8456                       1 << (OP_DESTROY_CLIENTID - 32)
8457         },
8458         .allow.u.words = {
8459                 [0] = 1 << (OP_CLOSE) |
8460                       1 << (OP_OPEN_DOWNGRADE) |
8461                       1 << (OP_LOCKU) |
8462                       1 << (OP_DELEGRETURN) |
8463                       1 << (OP_COMMIT),
8464                 [1] = 1 << (OP_SECINFO - 32) |
8465                       1 << (OP_SECINFO_NO_NAME - 32) |
8466                       1 << (OP_LAYOUTRETURN - 32) |
8467                       1 << (OP_TEST_STATEID - 32) |
8468                       1 << (OP_FREE_STATEID - 32) |
8469                       1 << (OP_WRITE - 32)
8470         }
8471 };
8472
8473 /*
8474  * Select the state protection mode for client `clp' given the server results
8475  * from exchange_id in `sp'.
8476  *
8477  * Returns 0 on success, negative errno otherwise.
8478  */
8479 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8480                                  struct nfs41_state_protection *sp)
8481 {
8482         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8483                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8484                       1 << (OP_EXCHANGE_ID - 32) |
8485                       1 << (OP_CREATE_SESSION - 32) |
8486                       1 << (OP_DESTROY_SESSION - 32) |
8487                       1 << (OP_DESTROY_CLIENTID - 32)
8488         };
8489         unsigned long flags = 0;
8490         unsigned int i;
8491         int ret = 0;
8492
8493         if (sp->how == SP4_MACH_CRED) {
8494                 /* Print state protect result */
8495                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8496                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8497                         if (test_bit(i, sp->enforce.u.longs))
8498                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8499                         if (test_bit(i, sp->allow.u.longs))
8500                                 dfprintk(MOUNT, "  allow op %d\n", i);
8501                 }
8502
8503                 /* make sure nothing is on enforce list that isn't supported */
8504                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8505                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8506                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8507                                 ret = -EINVAL;
8508                                 goto out;
8509                         }
8510                 }
8511
8512                 /*
8513                  * Minimal mode - state operations are allowed to use machine
8514                  * credential.  Note this already happens by default, so the
8515                  * client doesn't have to do anything more than the negotiation.
8516                  *
8517                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8518                  *       we're already using the machine cred for exchange_id
8519                  *       and will never use a different cred.
8520                  */
8521                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8522                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8523                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8524                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8525                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8526                         dfprintk(MOUNT, "  minimal mode enabled\n");
8527                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8528                 } else {
8529                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8530                         ret = -EINVAL;
8531                         goto out;
8532                 }
8533
8534                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8535                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8536                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8537                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8538                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8539                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8540                 }
8541
8542                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8543                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8544                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8545                 }
8546
8547                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8548                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8549                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8550                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8551                 }
8552
8553                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8554                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8555                         dfprintk(MOUNT, "  stateid mode enabled\n");
8556                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8557                 }
8558
8559                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8560                         dfprintk(MOUNT, "  write mode enabled\n");
8561                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8562                 }
8563
8564                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8565                         dfprintk(MOUNT, "  commit mode enabled\n");
8566                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8567                 }
8568         }
8569 out:
8570         clp->cl_sp4_flags = flags;
8571         return ret;
8572 }
8573
8574 struct nfs41_exchange_id_data {
8575         struct nfs41_exchange_id_res res;
8576         struct nfs41_exchange_id_args args;
8577 };
8578
8579 static void nfs4_exchange_id_release(void *data)
8580 {
8581         struct nfs41_exchange_id_data *cdata =
8582                                         (struct nfs41_exchange_id_data *)data;
8583
8584         nfs_put_client(cdata->args.client);
8585         kfree(cdata->res.impl_id);
8586         kfree(cdata->res.server_scope);
8587         kfree(cdata->res.server_owner);
8588         kfree(cdata);
8589 }
8590
8591 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8592         .rpc_release = nfs4_exchange_id_release,
8593 };
8594
8595 /*
8596  * _nfs4_proc_exchange_id()
8597  *
8598  * Wrapper for EXCHANGE_ID operation.
8599  */
8600 static struct rpc_task *
8601 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8602                         u32 sp4_how, struct rpc_xprt *xprt)
8603 {
8604         struct rpc_message msg = {
8605                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8606                 .rpc_cred = cred,
8607         };
8608         struct rpc_task_setup task_setup_data = {
8609                 .rpc_client = clp->cl_rpcclient,
8610                 .callback_ops = &nfs4_exchange_id_call_ops,
8611                 .rpc_message = &msg,
8612                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8613         };
8614         struct nfs41_exchange_id_data *calldata;
8615         int status;
8616
8617         if (!refcount_inc_not_zero(&clp->cl_count))
8618                 return ERR_PTR(-EIO);
8619
8620         status = -ENOMEM;
8621         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8622         if (!calldata)
8623                 goto out;
8624
8625         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8626
8627         status = nfs4_init_uniform_client_string(clp);
8628         if (status)
8629                 goto out_calldata;
8630
8631         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8632                                                 GFP_NOFS);
8633         status = -ENOMEM;
8634         if (unlikely(calldata->res.server_owner == NULL))
8635                 goto out_calldata;
8636
8637         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8638                                         GFP_NOFS);
8639         if (unlikely(calldata->res.server_scope == NULL))
8640                 goto out_server_owner;
8641
8642         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8643         if (unlikely(calldata->res.impl_id == NULL))
8644                 goto out_server_scope;
8645
8646         switch (sp4_how) {
8647         case SP4_NONE:
8648                 calldata->args.state_protect.how = SP4_NONE;
8649                 break;
8650
8651         case SP4_MACH_CRED:
8652                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8653                 break;
8654
8655         default:
8656                 /* unsupported! */
8657                 WARN_ON_ONCE(1);
8658                 status = -EINVAL;
8659                 goto out_impl_id;
8660         }
8661         if (xprt) {
8662                 task_setup_data.rpc_xprt = xprt;
8663                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8664                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8665                                 sizeof(calldata->args.verifier.data));
8666         }
8667         calldata->args.client = clp;
8668         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8669         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8670 #ifdef CONFIG_NFS_V4_1_MIGRATION
8671         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8672 #endif
8673         msg.rpc_argp = &calldata->args;
8674         msg.rpc_resp = &calldata->res;
8675         task_setup_data.callback_data = calldata;
8676
8677         return rpc_run_task(&task_setup_data);
8678
8679 out_impl_id:
8680         kfree(calldata->res.impl_id);
8681 out_server_scope:
8682         kfree(calldata->res.server_scope);
8683 out_server_owner:
8684         kfree(calldata->res.server_owner);
8685 out_calldata:
8686         kfree(calldata);
8687 out:
8688         nfs_put_client(clp);
8689         return ERR_PTR(status);
8690 }
8691
8692 /*
8693  * _nfs4_proc_exchange_id()
8694  *
8695  * Wrapper for EXCHANGE_ID operation.
8696  */
8697 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8698                         u32 sp4_how)
8699 {
8700         struct rpc_task *task;
8701         struct nfs41_exchange_id_args *argp;
8702         struct nfs41_exchange_id_res *resp;
8703         unsigned long now = jiffies;
8704         int status;
8705
8706         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8707         if (IS_ERR(task))
8708                 return PTR_ERR(task);
8709
8710         argp = task->tk_msg.rpc_argp;
8711         resp = task->tk_msg.rpc_resp;
8712         status = task->tk_status;
8713         if (status  != 0)
8714                 goto out;
8715
8716         status = nfs4_check_cl_exchange_flags(resp->flags,
8717                         clp->cl_mvops->minor_version);
8718         if (status  != 0)
8719                 goto out;
8720
8721         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8722         if (status != 0)
8723                 goto out;
8724
8725         do_renew_lease(clp, now);
8726
8727         clp->cl_clientid = resp->clientid;
8728         clp->cl_exchange_flags = resp->flags;
8729         clp->cl_seqid = resp->seqid;
8730         /* Client ID is not confirmed */
8731         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8732                 clear_bit(NFS4_SESSION_ESTABLISHED,
8733                           &clp->cl_session->session_state);
8734
8735         if (clp->cl_serverscope != NULL &&
8736             !nfs41_same_server_scope(clp->cl_serverscope,
8737                                 resp->server_scope)) {
8738                 dprintk("%s: server_scope mismatch detected\n",
8739                         __func__);
8740                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8741         }
8742
8743         swap(clp->cl_serverowner, resp->server_owner);
8744         swap(clp->cl_serverscope, resp->server_scope);
8745         swap(clp->cl_implid, resp->impl_id);
8746
8747         /* Save the EXCHANGE_ID verifier session trunk tests */
8748         memcpy(clp->cl_confirm.data, argp->verifier.data,
8749                sizeof(clp->cl_confirm.data));
8750 out:
8751         trace_nfs4_exchange_id(clp, status);
8752         rpc_put_task(task);
8753         return status;
8754 }
8755
8756 /*
8757  * nfs4_proc_exchange_id()
8758  *
8759  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8760  *
8761  * Since the clientid has expired, all compounds using sessions
8762  * associated with the stale clientid will be returning
8763  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8764  * be in some phase of session reset.
8765  *
8766  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8767  */
8768 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8769 {
8770         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8771         int status;
8772
8773         /* try SP4_MACH_CRED if krb5i/p */
8774         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8775             authflavor == RPC_AUTH_GSS_KRB5P) {
8776                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8777                 if (!status)
8778                         return 0;
8779         }
8780
8781         /* try SP4_NONE */
8782         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8783 }
8784
8785 /**
8786  * nfs4_test_session_trunk
8787  *
8788  * This is an add_xprt_test() test function called from
8789  * rpc_clnt_setup_test_and_add_xprt.
8790  *
8791  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8792  * and is dereferrenced in nfs4_exchange_id_release
8793  *
8794  * Upon success, add the new transport to the rpc_clnt
8795  *
8796  * @clnt: struct rpc_clnt to get new transport
8797  * @xprt: the rpc_xprt to test
8798  * @data: call data for _nfs4_proc_exchange_id.
8799  */
8800 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8801                             void *data)
8802 {
8803         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8804         struct rpc_task *task;
8805         int status;
8806
8807         u32 sp4_how;
8808
8809         dprintk("--> %s try %s\n", __func__,
8810                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8811
8812         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8813
8814         /* Test connection for session trunking. Async exchange_id call */
8815         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8816         if (IS_ERR(task))
8817                 return;
8818
8819         status = task->tk_status;
8820         if (status == 0)
8821                 status = nfs4_detect_session_trunking(adata->clp,
8822                                 task->tk_msg.rpc_resp, xprt);
8823
8824         if (status == 0)
8825                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8826
8827         rpc_put_task(task);
8828 }
8829 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8830
8831 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8832                 const struct cred *cred)
8833 {
8834         struct rpc_message msg = {
8835                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8836                 .rpc_argp = clp,
8837                 .rpc_cred = cred,
8838         };
8839         int status;
8840
8841         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8842                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8843         trace_nfs4_destroy_clientid(clp, status);
8844         if (status)
8845                 dprintk("NFS: Got error %d from the server %s on "
8846                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8847         return status;
8848 }
8849
8850 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8851                 const struct cred *cred)
8852 {
8853         unsigned int loop;
8854         int ret;
8855
8856         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8857                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8858                 switch (ret) {
8859                 case -NFS4ERR_DELAY:
8860                 case -NFS4ERR_CLIENTID_BUSY:
8861                         ssleep(1);
8862                         break;
8863                 default:
8864                         return ret;
8865                 }
8866         }
8867         return 0;
8868 }
8869
8870 int nfs4_destroy_clientid(struct nfs_client *clp)
8871 {
8872         const struct cred *cred;
8873         int ret = 0;
8874
8875         if (clp->cl_mvops->minor_version < 1)
8876                 goto out;
8877         if (clp->cl_exchange_flags == 0)
8878                 goto out;
8879         if (clp->cl_preserve_clid)
8880                 goto out;
8881         cred = nfs4_get_clid_cred(clp);
8882         ret = nfs4_proc_destroy_clientid(clp, cred);
8883         put_cred(cred);
8884         switch (ret) {
8885         case 0:
8886         case -NFS4ERR_STALE_CLIENTID:
8887                 clp->cl_exchange_flags = 0;
8888         }
8889 out:
8890         return ret;
8891 }
8892
8893 #endif /* CONFIG_NFS_V4_1 */
8894
8895 struct nfs4_get_lease_time_data {
8896         struct nfs4_get_lease_time_args *args;
8897         struct nfs4_get_lease_time_res *res;
8898         struct nfs_client *clp;
8899 };
8900
8901 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8902                                         void *calldata)
8903 {
8904         struct nfs4_get_lease_time_data *data =
8905                         (struct nfs4_get_lease_time_data *)calldata;
8906
8907         /* just setup sequence, do not trigger session recovery
8908            since we're invoked within one */
8909         nfs4_setup_sequence(data->clp,
8910                         &data->args->la_seq_args,
8911                         &data->res->lr_seq_res,
8912                         task);
8913 }
8914
8915 /*
8916  * Called from nfs4_state_manager thread for session setup, so don't recover
8917  * from sequence operation or clientid errors.
8918  */
8919 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8920 {
8921         struct nfs4_get_lease_time_data *data =
8922                         (struct nfs4_get_lease_time_data *)calldata;
8923
8924         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8925                 return;
8926         switch (task->tk_status) {
8927         case -NFS4ERR_DELAY:
8928         case -NFS4ERR_GRACE:
8929                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8930                 task->tk_status = 0;
8931                 fallthrough;
8932         case -NFS4ERR_RETRY_UNCACHED_REP:
8933                 rpc_restart_call_prepare(task);
8934                 return;
8935         }
8936 }
8937
8938 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8939         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8940         .rpc_call_done = nfs4_get_lease_time_done,
8941 };
8942
8943 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8944 {
8945         struct nfs4_get_lease_time_args args;
8946         struct nfs4_get_lease_time_res res = {
8947                 .lr_fsinfo = fsinfo,
8948         };
8949         struct nfs4_get_lease_time_data data = {
8950                 .args = &args,
8951                 .res = &res,
8952                 .clp = clp,
8953         };
8954         struct rpc_message msg = {
8955                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8956                 .rpc_argp = &args,
8957                 .rpc_resp = &res,
8958         };
8959         struct rpc_task_setup task_setup = {
8960                 .rpc_client = clp->cl_rpcclient,
8961                 .rpc_message = &msg,
8962                 .callback_ops = &nfs4_get_lease_time_ops,
8963                 .callback_data = &data,
8964                 .flags = RPC_TASK_TIMEOUT,
8965         };
8966
8967         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8968         return nfs4_call_sync_custom(&task_setup);
8969 }
8970
8971 #ifdef CONFIG_NFS_V4_1
8972
8973 /*
8974  * Initialize the values to be used by the client in CREATE_SESSION
8975  * If nfs4_init_session set the fore channel request and response sizes,
8976  * use them.
8977  *
8978  * Set the back channel max_resp_sz_cached to zero to force the client to
8979  * always set csa_cachethis to FALSE because the current implementation
8980  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8981  */
8982 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8983                                     struct rpc_clnt *clnt)
8984 {
8985         unsigned int max_rqst_sz, max_resp_sz;
8986         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8987         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8988
8989         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8990         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8991
8992         /* Fore channel attributes */
8993         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8994         args->fc_attrs.max_resp_sz = max_resp_sz;
8995         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8996         args->fc_attrs.max_reqs = max_session_slots;
8997
8998         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
8999                 "max_ops=%u max_reqs=%u\n",
9000                 __func__,
9001                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9002                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9003
9004         /* Back channel attributes */
9005         args->bc_attrs.max_rqst_sz = max_bc_payload;
9006         args->bc_attrs.max_resp_sz = max_bc_payload;
9007         args->bc_attrs.max_resp_sz_cached = 0;
9008         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9009         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9010         if (args->bc_attrs.max_reqs > max_bc_slots)
9011                 args->bc_attrs.max_reqs = max_bc_slots;
9012
9013         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9014                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9015                 __func__,
9016                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9017                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9018                 args->bc_attrs.max_reqs);
9019 }
9020
9021 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9022                 struct nfs41_create_session_res *res)
9023 {
9024         struct nfs4_channel_attrs *sent = &args->fc_attrs;
9025         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9026
9027         if (rcvd->max_resp_sz > sent->max_resp_sz)
9028                 return -EINVAL;
9029         /*
9030          * Our requested max_ops is the minimum we need; we're not
9031          * prepared to break up compounds into smaller pieces than that.
9032          * So, no point even trying to continue if the server won't
9033          * cooperate:
9034          */
9035         if (rcvd->max_ops < sent->max_ops)
9036                 return -EINVAL;
9037         if (rcvd->max_reqs == 0)
9038                 return -EINVAL;
9039         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9040                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9041         return 0;
9042 }
9043
9044 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9045                 struct nfs41_create_session_res *res)
9046 {
9047         struct nfs4_channel_attrs *sent = &args->bc_attrs;
9048         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9049
9050         if (!(res->flags & SESSION4_BACK_CHAN))
9051                 goto out;
9052         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9053                 return -EINVAL;
9054         if (rcvd->max_resp_sz < sent->max_resp_sz)
9055                 return -EINVAL;
9056         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9057                 return -EINVAL;
9058         if (rcvd->max_ops > sent->max_ops)
9059                 return -EINVAL;
9060         if (rcvd->max_reqs > sent->max_reqs)
9061                 return -EINVAL;
9062 out:
9063         return 0;
9064 }
9065
9066 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9067                                      struct nfs41_create_session_res *res)
9068 {
9069         int ret;
9070
9071         ret = nfs4_verify_fore_channel_attrs(args, res);
9072         if (ret)
9073                 return ret;
9074         return nfs4_verify_back_channel_attrs(args, res);
9075 }
9076
9077 static void nfs4_update_session(struct nfs4_session *session,
9078                 struct nfs41_create_session_res *res)
9079 {
9080         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9081         /* Mark client id and session as being confirmed */
9082         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9083         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9084         session->flags = res->flags;
9085         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9086         if (res->flags & SESSION4_BACK_CHAN)
9087                 memcpy(&session->bc_attrs, &res->bc_attrs,
9088                                 sizeof(session->bc_attrs));
9089 }
9090
9091 static int _nfs4_proc_create_session(struct nfs_client *clp,
9092                 const struct cred *cred)
9093 {
9094         struct nfs4_session *session = clp->cl_session;
9095         struct nfs41_create_session_args args = {
9096                 .client = clp,
9097                 .clientid = clp->cl_clientid,
9098                 .seqid = clp->cl_seqid,
9099                 .cb_program = NFS4_CALLBACK,
9100         };
9101         struct nfs41_create_session_res res;
9102
9103         struct rpc_message msg = {
9104                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9105                 .rpc_argp = &args,
9106                 .rpc_resp = &res,
9107                 .rpc_cred = cred,
9108         };
9109         int status;
9110
9111         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9112         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9113
9114         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9115                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9116         trace_nfs4_create_session(clp, status);
9117
9118         switch (status) {
9119         case -NFS4ERR_STALE_CLIENTID:
9120         case -NFS4ERR_DELAY:
9121         case -ETIMEDOUT:
9122         case -EACCES:
9123         case -EAGAIN:
9124                 goto out;
9125         }
9126
9127         clp->cl_seqid++;
9128         if (!status) {
9129                 /* Verify the session's negotiated channel_attrs values */
9130                 status = nfs4_verify_channel_attrs(&args, &res);
9131                 /* Increment the clientid slot sequence id */
9132                 if (status)
9133                         goto out;
9134                 nfs4_update_session(session, &res);
9135         }
9136 out:
9137         return status;
9138 }
9139
9140 /*
9141  * Issues a CREATE_SESSION operation to the server.
9142  * It is the responsibility of the caller to verify the session is
9143  * expired before calling this routine.
9144  */
9145 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9146 {
9147         int status;
9148         unsigned *ptr;
9149         struct nfs4_session *session = clp->cl_session;
9150
9151         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9152
9153         status = _nfs4_proc_create_session(clp, cred);
9154         if (status)
9155                 goto out;
9156
9157         /* Init or reset the session slot tables */
9158         status = nfs4_setup_session_slot_tables(session);
9159         dprintk("slot table setup returned %d\n", status);
9160         if (status)
9161                 goto out;
9162
9163         ptr = (unsigned *)&session->sess_id.data[0];
9164         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9165                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9166 out:
9167         return status;
9168 }
9169
9170 /*
9171  * Issue the over-the-wire RPC DESTROY_SESSION.
9172  * The caller must serialize access to this routine.
9173  */
9174 int nfs4_proc_destroy_session(struct nfs4_session *session,
9175                 const struct cred *cred)
9176 {
9177         struct rpc_message msg = {
9178                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9179                 .rpc_argp = session,
9180                 .rpc_cred = cred,
9181         };
9182         int status = 0;
9183
9184         /* session is still being setup */
9185         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9186                 return 0;
9187
9188         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9189                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9190         trace_nfs4_destroy_session(session->clp, status);
9191
9192         if (status)
9193                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9194                         "Session has been destroyed regardless...\n", status);
9195         return status;
9196 }
9197
9198 /*
9199  * Renew the cl_session lease.
9200  */
9201 struct nfs4_sequence_data {
9202         struct nfs_client *clp;
9203         struct nfs4_sequence_args args;
9204         struct nfs4_sequence_res res;
9205 };
9206
9207 static void nfs41_sequence_release(void *data)
9208 {
9209         struct nfs4_sequence_data *calldata = data;
9210         struct nfs_client *clp = calldata->clp;
9211
9212         if (refcount_read(&clp->cl_count) > 1)
9213                 nfs4_schedule_state_renewal(clp);
9214         nfs_put_client(clp);
9215         kfree(calldata);
9216 }
9217
9218 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9219 {
9220         switch(task->tk_status) {
9221         case -NFS4ERR_DELAY:
9222                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9223                 return -EAGAIN;
9224         default:
9225                 nfs4_schedule_lease_recovery(clp);
9226         }
9227         return 0;
9228 }
9229
9230 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9231 {
9232         struct nfs4_sequence_data *calldata = data;
9233         struct nfs_client *clp = calldata->clp;
9234
9235         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9236                 return;
9237
9238         trace_nfs4_sequence(clp, task->tk_status);
9239         if (task->tk_status < 0) {
9240                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9241                 if (refcount_read(&clp->cl_count) == 1)
9242                         return;
9243
9244                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9245                         rpc_restart_call_prepare(task);
9246                         return;
9247                 }
9248         }
9249         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9250 }
9251
9252 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9253 {
9254         struct nfs4_sequence_data *calldata = data;
9255         struct nfs_client *clp = calldata->clp;
9256         struct nfs4_sequence_args *args;
9257         struct nfs4_sequence_res *res;
9258
9259         args = task->tk_msg.rpc_argp;
9260         res = task->tk_msg.rpc_resp;
9261
9262         nfs4_setup_sequence(clp, args, res, task);
9263 }
9264
9265 static const struct rpc_call_ops nfs41_sequence_ops = {
9266         .rpc_call_done = nfs41_sequence_call_done,
9267         .rpc_call_prepare = nfs41_sequence_prepare,
9268         .rpc_release = nfs41_sequence_release,
9269 };
9270
9271 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9272                 const struct cred *cred,
9273                 struct nfs4_slot *slot,
9274                 bool is_privileged)
9275 {
9276         struct nfs4_sequence_data *calldata;
9277         struct rpc_message msg = {
9278                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9279                 .rpc_cred = cred,
9280         };
9281         struct rpc_task_setup task_setup_data = {
9282                 .rpc_client = clp->cl_rpcclient,
9283                 .rpc_message = &msg,
9284                 .callback_ops = &nfs41_sequence_ops,
9285                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9286         };
9287         struct rpc_task *ret;
9288
9289         ret = ERR_PTR(-EIO);
9290         if (!refcount_inc_not_zero(&clp->cl_count))
9291                 goto out_err;
9292
9293         ret = ERR_PTR(-ENOMEM);
9294         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9295         if (calldata == NULL)
9296                 goto out_put_clp;
9297         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9298         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9299         msg.rpc_argp = &calldata->args;
9300         msg.rpc_resp = &calldata->res;
9301         calldata->clp = clp;
9302         task_setup_data.callback_data = calldata;
9303
9304         ret = rpc_run_task(&task_setup_data);
9305         if (IS_ERR(ret))
9306                 goto out_err;
9307         return ret;
9308 out_put_clp:
9309         nfs_put_client(clp);
9310 out_err:
9311         nfs41_release_slot(slot);
9312         return ret;
9313 }
9314
9315 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9316 {
9317         struct rpc_task *task;
9318         int ret = 0;
9319
9320         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9321                 return -EAGAIN;
9322         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9323         if (IS_ERR(task))
9324                 ret = PTR_ERR(task);
9325         else
9326                 rpc_put_task_async(task);
9327         dprintk("<-- %s status=%d\n", __func__, ret);
9328         return ret;
9329 }
9330
9331 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9332 {
9333         struct rpc_task *task;
9334         int ret;
9335
9336         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9337         if (IS_ERR(task)) {
9338                 ret = PTR_ERR(task);
9339                 goto out;
9340         }
9341         ret = rpc_wait_for_completion_task(task);
9342         if (!ret)
9343                 ret = task->tk_status;
9344         rpc_put_task(task);
9345 out:
9346         dprintk("<-- %s status=%d\n", __func__, ret);
9347         return ret;
9348 }
9349
9350 struct nfs4_reclaim_complete_data {
9351         struct nfs_client *clp;
9352         struct nfs41_reclaim_complete_args arg;
9353         struct nfs41_reclaim_complete_res res;
9354 };
9355
9356 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9357 {
9358         struct nfs4_reclaim_complete_data *calldata = data;
9359
9360         nfs4_setup_sequence(calldata->clp,
9361                         &calldata->arg.seq_args,
9362                         &calldata->res.seq_res,
9363                         task);
9364 }
9365
9366 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9367 {
9368         switch(task->tk_status) {
9369         case 0:
9370                 wake_up_all(&clp->cl_lock_waitq);
9371                 fallthrough;
9372         case -NFS4ERR_COMPLETE_ALREADY:
9373         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9374                 break;
9375         case -NFS4ERR_DELAY:
9376                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9377                 fallthrough;
9378         case -NFS4ERR_RETRY_UNCACHED_REP:
9379                 return -EAGAIN;
9380         case -NFS4ERR_BADSESSION:
9381         case -NFS4ERR_DEADSESSION:
9382         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9383                 break;
9384         default:
9385                 nfs4_schedule_lease_recovery(clp);
9386         }
9387         return 0;
9388 }
9389
9390 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9391 {
9392         struct nfs4_reclaim_complete_data *calldata = data;
9393         struct nfs_client *clp = calldata->clp;
9394         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9395
9396         if (!nfs41_sequence_done(task, res))
9397                 return;
9398
9399         trace_nfs4_reclaim_complete(clp, task->tk_status);
9400         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9401                 rpc_restart_call_prepare(task);
9402                 return;
9403         }
9404 }
9405
9406 static void nfs4_free_reclaim_complete_data(void *data)
9407 {
9408         struct nfs4_reclaim_complete_data *calldata = data;
9409
9410         kfree(calldata);
9411 }
9412
9413 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9414         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9415         .rpc_call_done = nfs4_reclaim_complete_done,
9416         .rpc_release = nfs4_free_reclaim_complete_data,
9417 };
9418
9419 /*
9420  * Issue a global reclaim complete.
9421  */
9422 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9423                 const struct cred *cred)
9424 {
9425         struct nfs4_reclaim_complete_data *calldata;
9426         struct rpc_message msg = {
9427                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9428                 .rpc_cred = cred,
9429         };
9430         struct rpc_task_setup task_setup_data = {
9431                 .rpc_client = clp->cl_rpcclient,
9432                 .rpc_message = &msg,
9433                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9434                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9435         };
9436         int status = -ENOMEM;
9437
9438         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9439         if (calldata == NULL)
9440                 goto out;
9441         calldata->clp = clp;
9442         calldata->arg.one_fs = 0;
9443
9444         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9445         msg.rpc_argp = &calldata->arg;
9446         msg.rpc_resp = &calldata->res;
9447         task_setup_data.callback_data = calldata;
9448         status = nfs4_call_sync_custom(&task_setup_data);
9449 out:
9450         dprintk("<-- %s status=%d\n", __func__, status);
9451         return status;
9452 }
9453
9454 static void
9455 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9456 {
9457         struct nfs4_layoutget *lgp = calldata;
9458         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9459
9460         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9461                                 &lgp->res.seq_res, task);
9462 }
9463
9464 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9465 {
9466         struct nfs4_layoutget *lgp = calldata;
9467
9468         nfs41_sequence_process(task, &lgp->res.seq_res);
9469 }
9470
9471 static int
9472 nfs4_layoutget_handle_exception(struct rpc_task *task,
9473                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9474 {
9475         struct inode *inode = lgp->args.inode;
9476         struct nfs_server *server = NFS_SERVER(inode);
9477         struct pnfs_layout_hdr *lo = lgp->lo;
9478         int nfs4err = task->tk_status;
9479         int err, status = 0;
9480         LIST_HEAD(head);
9481
9482         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9483
9484         nfs4_sequence_free_slot(&lgp->res.seq_res);
9485
9486         switch (nfs4err) {
9487         case 0:
9488                 goto out;
9489
9490         /*
9491          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9492          * on the file. set tk_status to -ENODATA to tell upper layer to
9493          * retry go inband.
9494          */
9495         case -NFS4ERR_LAYOUTUNAVAILABLE:
9496                 status = -ENODATA;
9497                 goto out;
9498         /*
9499          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9500          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9501          */
9502         case -NFS4ERR_BADLAYOUT:
9503                 status = -EOVERFLOW;
9504                 goto out;
9505         /*
9506          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9507          * (or clients) writing to the same RAID stripe except when
9508          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9509          *
9510          * Treat it like we would RECALLCONFLICT -- we retry for a little
9511          * while, and then eventually give up.
9512          */
9513         case -NFS4ERR_LAYOUTTRYLATER:
9514                 if (lgp->args.minlength == 0) {
9515                         status = -EOVERFLOW;
9516                         goto out;
9517                 }
9518                 status = -EBUSY;
9519                 break;
9520         case -NFS4ERR_RECALLCONFLICT:
9521                 status = -ERECALLCONFLICT;
9522                 break;
9523         case -NFS4ERR_DELEG_REVOKED:
9524         case -NFS4ERR_ADMIN_REVOKED:
9525         case -NFS4ERR_EXPIRED:
9526         case -NFS4ERR_BAD_STATEID:
9527                 exception->timeout = 0;
9528                 spin_lock(&inode->i_lock);
9529                 /* If the open stateid was bad, then recover it. */
9530                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9531                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9532                         spin_unlock(&inode->i_lock);
9533                         exception->state = lgp->args.ctx->state;
9534                         exception->stateid = &lgp->args.stateid;
9535                         break;
9536                 }
9537
9538                 /*
9539                  * Mark the bad layout state as invalid, then retry
9540                  */
9541                 pnfs_mark_layout_stateid_invalid(lo, &head);
9542                 spin_unlock(&inode->i_lock);
9543                 nfs_commit_inode(inode, 0);
9544                 pnfs_free_lseg_list(&head);
9545                 status = -EAGAIN;
9546                 goto out;
9547         }
9548
9549         err = nfs4_handle_exception(server, nfs4err, exception);
9550         if (!status) {
9551                 if (exception->retry)
9552                         status = -EAGAIN;
9553                 else
9554                         status = err;
9555         }
9556 out:
9557         return status;
9558 }
9559
9560 size_t max_response_pages(struct nfs_server *server)
9561 {
9562         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9563         return nfs_page_array_len(0, max_resp_sz);
9564 }
9565
9566 static void nfs4_layoutget_release(void *calldata)
9567 {
9568         struct nfs4_layoutget *lgp = calldata;
9569
9570         nfs4_sequence_free_slot(&lgp->res.seq_res);
9571         pnfs_layoutget_free(lgp);
9572 }
9573
9574 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9575         .rpc_call_prepare = nfs4_layoutget_prepare,
9576         .rpc_call_done = nfs4_layoutget_done,
9577         .rpc_release = nfs4_layoutget_release,
9578 };
9579
9580 struct pnfs_layout_segment *
9581 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9582 {
9583         struct inode *inode = lgp->args.inode;
9584         struct nfs_server *server = NFS_SERVER(inode);
9585         struct rpc_task *task;
9586         struct rpc_message msg = {
9587                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9588                 .rpc_argp = &lgp->args,
9589                 .rpc_resp = &lgp->res,
9590                 .rpc_cred = lgp->cred,
9591         };
9592         struct rpc_task_setup task_setup_data = {
9593                 .rpc_client = server->client,
9594                 .rpc_message = &msg,
9595                 .callback_ops = &nfs4_layoutget_call_ops,
9596                 .callback_data = lgp,
9597                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9598                          RPC_TASK_MOVEABLE,
9599         };
9600         struct pnfs_layout_segment *lseg = NULL;
9601         struct nfs4_exception exception = {
9602                 .inode = inode,
9603                 .timeout = *timeout,
9604         };
9605         int status = 0;
9606
9607         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9608
9609         task = rpc_run_task(&task_setup_data);
9610
9611         status = rpc_wait_for_completion_task(task);
9612         if (status != 0)
9613                 goto out;
9614
9615         if (task->tk_status < 0) {
9616                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9617                 *timeout = exception.timeout;
9618         } else if (lgp->res.layoutp->len == 0) {
9619                 status = -EAGAIN;
9620                 *timeout = nfs4_update_delay(&exception.timeout);
9621         } else
9622                 lseg = pnfs_layout_process(lgp);
9623 out:
9624         trace_nfs4_layoutget(lgp->args.ctx,
9625                         &lgp->args.range,
9626                         &lgp->res.range,
9627                         &lgp->res.stateid,
9628                         status);
9629
9630         rpc_put_task(task);
9631         dprintk("<-- %s status=%d\n", __func__, status);
9632         if (status)
9633                 return ERR_PTR(status);
9634         return lseg;
9635 }
9636
9637 static void
9638 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9639 {
9640         struct nfs4_layoutreturn *lrp = calldata;
9641
9642         nfs4_setup_sequence(lrp->clp,
9643                         &lrp->args.seq_args,
9644                         &lrp->res.seq_res,
9645                         task);
9646         if (!pnfs_layout_is_valid(lrp->args.layout))
9647                 rpc_exit(task, 0);
9648 }
9649
9650 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9651 {
9652         struct nfs4_layoutreturn *lrp = calldata;
9653         struct nfs_server *server;
9654
9655         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9656                 return;
9657
9658         /*
9659          * Was there an RPC level error? Assume the call succeeded,
9660          * and that we need to release the layout
9661          */
9662         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9663                 lrp->res.lrs_present = 0;
9664                 return;
9665         }
9666
9667         server = NFS_SERVER(lrp->args.inode);
9668         switch (task->tk_status) {
9669         case -NFS4ERR_OLD_STATEID:
9670                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9671                                         &lrp->args.range,
9672                                         lrp->args.inode))
9673                         goto out_restart;
9674                 fallthrough;
9675         default:
9676                 task->tk_status = 0;
9677                 fallthrough;
9678         case 0:
9679                 break;
9680         case -NFS4ERR_DELAY:
9681                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9682                         break;
9683                 goto out_restart;
9684         }
9685         return;
9686 out_restart:
9687         task->tk_status = 0;
9688         nfs4_sequence_free_slot(&lrp->res.seq_res);
9689         rpc_restart_call_prepare(task);
9690 }
9691
9692 static void nfs4_layoutreturn_release(void *calldata)
9693 {
9694         struct nfs4_layoutreturn *lrp = calldata;
9695         struct pnfs_layout_hdr *lo = lrp->args.layout;
9696
9697         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9698                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9699         nfs4_sequence_free_slot(&lrp->res.seq_res);
9700         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9701                 lrp->ld_private.ops->free(&lrp->ld_private);
9702         pnfs_put_layout_hdr(lrp->args.layout);
9703         nfs_iput_and_deactive(lrp->inode);
9704         put_cred(lrp->cred);
9705         kfree(calldata);
9706 }
9707
9708 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9709         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9710         .rpc_call_done = nfs4_layoutreturn_done,
9711         .rpc_release = nfs4_layoutreturn_release,
9712 };
9713
9714 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9715 {
9716         struct rpc_task *task;
9717         struct rpc_message msg = {
9718                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9719                 .rpc_argp = &lrp->args,
9720                 .rpc_resp = &lrp->res,
9721                 .rpc_cred = lrp->cred,
9722         };
9723         struct rpc_task_setup task_setup_data = {
9724                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9725                 .rpc_message = &msg,
9726                 .callback_ops = &nfs4_layoutreturn_call_ops,
9727                 .callback_data = lrp,
9728                 .flags = RPC_TASK_MOVEABLE,
9729         };
9730         int status = 0;
9731
9732         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9733                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9734                         &task_setup_data.rpc_client, &msg);
9735
9736         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9737         if (!sync) {
9738                 if (!lrp->inode) {
9739                         nfs4_layoutreturn_release(lrp);
9740                         return -EAGAIN;
9741                 }
9742                 task_setup_data.flags |= RPC_TASK_ASYNC;
9743         }
9744         if (!lrp->inode)
9745                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9746                                    1);
9747         else
9748                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9749                                    0);
9750         task = rpc_run_task(&task_setup_data);
9751         if (IS_ERR(task))
9752                 return PTR_ERR(task);
9753         if (sync)
9754                 status = task->tk_status;
9755         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9756         dprintk("<-- %s status=%d\n", __func__, status);
9757         rpc_put_task(task);
9758         return status;
9759 }
9760
9761 static int
9762 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9763                 struct pnfs_device *pdev,
9764                 const struct cred *cred)
9765 {
9766         struct nfs4_getdeviceinfo_args args = {
9767                 .pdev = pdev,
9768                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9769                         NOTIFY_DEVICEID4_DELETE,
9770         };
9771         struct nfs4_getdeviceinfo_res res = {
9772                 .pdev = pdev,
9773         };
9774         struct rpc_message msg = {
9775                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9776                 .rpc_argp = &args,
9777                 .rpc_resp = &res,
9778                 .rpc_cred = cred,
9779         };
9780         int status;
9781
9782         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9783         if (res.notification & ~args.notify_types)
9784                 dprintk("%s: unsupported notification\n", __func__);
9785         if (res.notification != args.notify_types)
9786                 pdev->nocache = 1;
9787
9788         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9789
9790         dprintk("<-- %s status=%d\n", __func__, status);
9791
9792         return status;
9793 }
9794
9795 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9796                 struct pnfs_device *pdev,
9797                 const struct cred *cred)
9798 {
9799         struct nfs4_exception exception = { };
9800         int err;
9801
9802         do {
9803                 err = nfs4_handle_exception(server,
9804                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9805                                         &exception);
9806         } while (exception.retry);
9807         return err;
9808 }
9809 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9810
9811 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9812 {
9813         struct nfs4_layoutcommit_data *data = calldata;
9814         struct nfs_server *server = NFS_SERVER(data->args.inode);
9815
9816         nfs4_setup_sequence(server->nfs_client,
9817                         &data->args.seq_args,
9818                         &data->res.seq_res,
9819                         task);
9820 }
9821
9822 static void
9823 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9824 {
9825         struct nfs4_layoutcommit_data *data = calldata;
9826         struct nfs_server *server = NFS_SERVER(data->args.inode);
9827
9828         if (!nfs41_sequence_done(task, &data->res.seq_res))
9829                 return;
9830
9831         switch (task->tk_status) { /* Just ignore these failures */
9832         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9833         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9834         case -NFS4ERR_BADLAYOUT:     /* no layout */
9835         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9836                 task->tk_status = 0;
9837                 break;
9838         case 0:
9839                 break;
9840         default:
9841                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9842                         rpc_restart_call_prepare(task);
9843                         return;
9844                 }
9845         }
9846 }
9847
9848 static void nfs4_layoutcommit_release(void *calldata)
9849 {
9850         struct nfs4_layoutcommit_data *data = calldata;
9851
9852         pnfs_cleanup_layoutcommit(data);
9853         nfs_post_op_update_inode_force_wcc(data->args.inode,
9854                                            data->res.fattr);
9855         put_cred(data->cred);
9856         nfs_iput_and_deactive(data->inode);
9857         kfree(data);
9858 }
9859
9860 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9861         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9862         .rpc_call_done = nfs4_layoutcommit_done,
9863         .rpc_release = nfs4_layoutcommit_release,
9864 };
9865
9866 int
9867 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9868 {
9869         struct rpc_message msg = {
9870                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9871                 .rpc_argp = &data->args,
9872                 .rpc_resp = &data->res,
9873                 .rpc_cred = data->cred,
9874         };
9875         struct rpc_task_setup task_setup_data = {
9876                 .task = &data->task,
9877                 .rpc_client = NFS_CLIENT(data->args.inode),
9878                 .rpc_message = &msg,
9879                 .callback_ops = &nfs4_layoutcommit_ops,
9880                 .callback_data = data,
9881                 .flags = RPC_TASK_MOVEABLE,
9882         };
9883         struct rpc_task *task;
9884         int status = 0;
9885
9886         dprintk("NFS: initiating layoutcommit call. sync %d "
9887                 "lbw: %llu inode %lu\n", sync,
9888                 data->args.lastbytewritten,
9889                 data->args.inode->i_ino);
9890
9891         if (!sync) {
9892                 data->inode = nfs_igrab_and_active(data->args.inode);
9893                 if (data->inode == NULL) {
9894                         nfs4_layoutcommit_release(data);
9895                         return -EAGAIN;
9896                 }
9897                 task_setup_data.flags = RPC_TASK_ASYNC;
9898         }
9899         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9900         task = rpc_run_task(&task_setup_data);
9901         if (IS_ERR(task))
9902                 return PTR_ERR(task);
9903         if (sync)
9904                 status = task->tk_status;
9905         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9906         dprintk("%s: status %d\n", __func__, status);
9907         rpc_put_task(task);
9908         return status;
9909 }
9910
9911 /*
9912  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9913  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9914  */
9915 static int
9916 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9917                     struct nfs_fsinfo *info,
9918                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9919 {
9920         struct nfs41_secinfo_no_name_args args = {
9921                 .style = SECINFO_STYLE_CURRENT_FH,
9922         };
9923         struct nfs4_secinfo_res res = {
9924                 .flavors = flavors,
9925         };
9926         struct rpc_message msg = {
9927                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9928                 .rpc_argp = &args,
9929                 .rpc_resp = &res,
9930         };
9931         struct nfs4_call_sync_data data = {
9932                 .seq_server = server,
9933                 .seq_args = &args.seq_args,
9934                 .seq_res = &res.seq_res,
9935         };
9936         struct rpc_task_setup task_setup = {
9937                 .rpc_client = server->client,
9938                 .rpc_message = &msg,
9939                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9940                 .callback_data = &data,
9941                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9942         };
9943         const struct cred *cred = NULL;
9944         int status;
9945
9946         if (use_integrity) {
9947                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9948
9949                 cred = nfs4_get_clid_cred(server->nfs_client);
9950                 msg.rpc_cred = cred;
9951         }
9952
9953         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9954         status = nfs4_call_sync_custom(&task_setup);
9955         dprintk("<-- %s status=%d\n", __func__, status);
9956
9957         put_cred(cred);
9958
9959         return status;
9960 }
9961
9962 static int
9963 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9964                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9965 {
9966         struct nfs4_exception exception = {
9967                 .interruptible = true,
9968         };
9969         int err;
9970         do {
9971                 /* first try using integrity protection */
9972                 err = -NFS4ERR_WRONGSEC;
9973
9974                 /* try to use integrity protection with machine cred */
9975                 if (_nfs4_is_integrity_protected(server->nfs_client))
9976                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9977                                                           flavors, true);
9978
9979                 /*
9980                  * if unable to use integrity protection, or SECINFO with
9981                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9982                  * disallowed by spec, but exists in deployed servers) use
9983                  * the current filesystem's rpc_client and the user cred.
9984                  */
9985                 if (err == -NFS4ERR_WRONGSEC)
9986                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9987                                                           flavors, false);
9988
9989                 switch (err) {
9990                 case 0:
9991                 case -NFS4ERR_WRONGSEC:
9992                 case -ENOTSUPP:
9993                         goto out;
9994                 default:
9995                         err = nfs4_handle_exception(server, err, &exception);
9996                 }
9997         } while (exception.retry);
9998 out:
9999         return err;
10000 }
10001
10002 static int
10003 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10004                     struct nfs_fsinfo *info)
10005 {
10006         int err;
10007         struct page *page;
10008         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10009         struct nfs4_secinfo_flavors *flavors;
10010         struct nfs4_secinfo4 *secinfo;
10011         int i;
10012
10013         page = alloc_page(GFP_KERNEL);
10014         if (!page) {
10015                 err = -ENOMEM;
10016                 goto out;
10017         }
10018
10019         flavors = page_address(page);
10020         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10021
10022         /*
10023          * Fall back on "guess and check" method if
10024          * the server doesn't support SECINFO_NO_NAME
10025          */
10026         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10027                 err = nfs4_find_root_sec(server, fhandle, info);
10028                 goto out_freepage;
10029         }
10030         if (err)
10031                 goto out_freepage;
10032
10033         for (i = 0; i < flavors->num_flavors; i++) {
10034                 secinfo = &flavors->flavors[i];
10035
10036                 switch (secinfo->flavor) {
10037                 case RPC_AUTH_NULL:
10038                 case RPC_AUTH_UNIX:
10039                 case RPC_AUTH_GSS:
10040                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10041                                         &secinfo->flavor_info);
10042                         break;
10043                 default:
10044                         flavor = RPC_AUTH_MAXFLAVOR;
10045                         break;
10046                 }
10047
10048                 if (!nfs_auth_info_match(&server->auth_info, flavor))
10049                         flavor = RPC_AUTH_MAXFLAVOR;
10050
10051                 if (flavor != RPC_AUTH_MAXFLAVOR) {
10052                         err = nfs4_lookup_root_sec(server, fhandle,
10053                                                    info, flavor);
10054                         if (!err)
10055                                 break;
10056                 }
10057         }
10058
10059         if (flavor == RPC_AUTH_MAXFLAVOR)
10060                 err = -EPERM;
10061
10062 out_freepage:
10063         put_page(page);
10064         if (err == -EACCES)
10065                 return -EPERM;
10066 out:
10067         return err;
10068 }
10069
10070 static int _nfs41_test_stateid(struct nfs_server *server,
10071                 nfs4_stateid *stateid,
10072                 const struct cred *cred)
10073 {
10074         int status;
10075         struct nfs41_test_stateid_args args = {
10076                 .stateid = stateid,
10077         };
10078         struct nfs41_test_stateid_res res;
10079         struct rpc_message msg = {
10080                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10081                 .rpc_argp = &args,
10082                 .rpc_resp = &res,
10083                 .rpc_cred = cred,
10084         };
10085         struct rpc_clnt *rpc_client = server->client;
10086
10087         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10088                 &rpc_client, &msg);
10089
10090         dprintk("NFS call  test_stateid %p\n", stateid);
10091         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10092         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10093                         &args.seq_args, &res.seq_res);
10094         if (status != NFS_OK) {
10095                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10096                 return status;
10097         }
10098         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10099         return -res.status;
10100 }
10101
10102 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10103                 int err, struct nfs4_exception *exception)
10104 {
10105         exception->retry = 0;
10106         switch(err) {
10107         case -NFS4ERR_DELAY:
10108         case -NFS4ERR_RETRY_UNCACHED_REP:
10109                 nfs4_handle_exception(server, err, exception);
10110                 break;
10111         case -NFS4ERR_BADSESSION:
10112         case -NFS4ERR_BADSLOT:
10113         case -NFS4ERR_BAD_HIGH_SLOT:
10114         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10115         case -NFS4ERR_DEADSESSION:
10116                 nfs4_do_handle_exception(server, err, exception);
10117         }
10118 }
10119
10120 /**
10121  * nfs41_test_stateid - perform a TEST_STATEID operation
10122  *
10123  * @server: server / transport on which to perform the operation
10124  * @stateid: state ID to test
10125  * @cred: credential
10126  *
10127  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10128  * Otherwise a negative NFS4ERR value is returned if the operation
10129  * failed or the state ID is not currently valid.
10130  */
10131 static int nfs41_test_stateid(struct nfs_server *server,
10132                 nfs4_stateid *stateid,
10133                 const struct cred *cred)
10134 {
10135         struct nfs4_exception exception = {
10136                 .interruptible = true,
10137         };
10138         int err;
10139         do {
10140                 err = _nfs41_test_stateid(server, stateid, cred);
10141                 nfs4_handle_delay_or_session_error(server, err, &exception);
10142         } while (exception.retry);
10143         return err;
10144 }
10145
10146 struct nfs_free_stateid_data {
10147         struct nfs_server *server;
10148         struct nfs41_free_stateid_args args;
10149         struct nfs41_free_stateid_res res;
10150 };
10151
10152 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10153 {
10154         struct nfs_free_stateid_data *data = calldata;
10155         nfs4_setup_sequence(data->server->nfs_client,
10156                         &data->args.seq_args,
10157                         &data->res.seq_res,
10158                         task);
10159 }
10160
10161 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10162 {
10163         struct nfs_free_stateid_data *data = calldata;
10164
10165         nfs41_sequence_done(task, &data->res.seq_res);
10166
10167         switch (task->tk_status) {
10168         case -NFS4ERR_DELAY:
10169                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10170                         rpc_restart_call_prepare(task);
10171         }
10172 }
10173
10174 static void nfs41_free_stateid_release(void *calldata)
10175 {
10176         struct nfs_free_stateid_data *data = calldata;
10177         struct nfs_client *clp = data->server->nfs_client;
10178
10179         nfs_put_client(clp);
10180         kfree(calldata);
10181 }
10182
10183 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10184         .rpc_call_prepare = nfs41_free_stateid_prepare,
10185         .rpc_call_done = nfs41_free_stateid_done,
10186         .rpc_release = nfs41_free_stateid_release,
10187 };
10188
10189 /**
10190  * nfs41_free_stateid - perform a FREE_STATEID operation
10191  *
10192  * @server: server / transport on which to perform the operation
10193  * @stateid: state ID to release
10194  * @cred: credential
10195  * @privileged: set to true if this call needs to be privileged
10196  *
10197  * Note: this function is always asynchronous.
10198  */
10199 static int nfs41_free_stateid(struct nfs_server *server,
10200                 const nfs4_stateid *stateid,
10201                 const struct cred *cred,
10202                 bool privileged)
10203 {
10204         struct rpc_message msg = {
10205                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10206                 .rpc_cred = cred,
10207         };
10208         struct rpc_task_setup task_setup = {
10209                 .rpc_client = server->client,
10210                 .rpc_message = &msg,
10211                 .callback_ops = &nfs41_free_stateid_ops,
10212                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10213         };
10214         struct nfs_free_stateid_data *data;
10215         struct rpc_task *task;
10216         struct nfs_client *clp = server->nfs_client;
10217
10218         if (!refcount_inc_not_zero(&clp->cl_count))
10219                 return -EIO;
10220
10221         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10222                 &task_setup.rpc_client, &msg);
10223
10224         dprintk("NFS call  free_stateid %p\n", stateid);
10225         data = kmalloc(sizeof(*data), GFP_NOFS);
10226         if (!data)
10227                 return -ENOMEM;
10228         data->server = server;
10229         nfs4_stateid_copy(&data->args.stateid, stateid);
10230
10231         task_setup.callback_data = data;
10232
10233         msg.rpc_argp = &data->args;
10234         msg.rpc_resp = &data->res;
10235         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10236         task = rpc_run_task(&task_setup);
10237         if (IS_ERR(task))
10238                 return PTR_ERR(task);
10239         rpc_put_task(task);
10240         return 0;
10241 }
10242
10243 static void
10244 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10245 {
10246         const struct cred *cred = lsp->ls_state->owner->so_cred;
10247
10248         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10249         nfs4_free_lock_state(server, lsp);
10250 }
10251
10252 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10253                 const nfs4_stateid *s2)
10254 {
10255         if (s1->type != s2->type)
10256                 return false;
10257
10258         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10259                 return false;
10260
10261         if (s1->seqid == s2->seqid)
10262                 return true;
10263
10264         return s1->seqid == 0 || s2->seqid == 0;
10265 }
10266
10267 #endif /* CONFIG_NFS_V4_1 */
10268
10269 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10270                 const nfs4_stateid *s2)
10271 {
10272         return nfs4_stateid_match(s1, s2);
10273 }
10274
10275
10276 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10277         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10278         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10279         .recover_open   = nfs4_open_reclaim,
10280         .recover_lock   = nfs4_lock_reclaim,
10281         .establish_clid = nfs4_init_clientid,
10282         .detect_trunking = nfs40_discover_server_trunking,
10283 };
10284
10285 #if defined(CONFIG_NFS_V4_1)
10286 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10287         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10288         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10289         .recover_open   = nfs4_open_reclaim,
10290         .recover_lock   = nfs4_lock_reclaim,
10291         .establish_clid = nfs41_init_clientid,
10292         .reclaim_complete = nfs41_proc_reclaim_complete,
10293         .detect_trunking = nfs41_discover_server_trunking,
10294 };
10295 #endif /* CONFIG_NFS_V4_1 */
10296
10297 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10298         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10299         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10300         .recover_open   = nfs40_open_expired,
10301         .recover_lock   = nfs4_lock_expired,
10302         .establish_clid = nfs4_init_clientid,
10303 };
10304
10305 #if defined(CONFIG_NFS_V4_1)
10306 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10307         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10308         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10309         .recover_open   = nfs41_open_expired,
10310         .recover_lock   = nfs41_lock_expired,
10311         .establish_clid = nfs41_init_clientid,
10312 };
10313 #endif /* CONFIG_NFS_V4_1 */
10314
10315 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10316         .sched_state_renewal = nfs4_proc_async_renew,
10317         .get_state_renewal_cred = nfs4_get_renew_cred,
10318         .renew_lease = nfs4_proc_renew,
10319 };
10320
10321 #if defined(CONFIG_NFS_V4_1)
10322 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10323         .sched_state_renewal = nfs41_proc_async_sequence,
10324         .get_state_renewal_cred = nfs4_get_machine_cred,
10325         .renew_lease = nfs4_proc_sequence,
10326 };
10327 #endif
10328
10329 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10330         .get_locations = _nfs40_proc_get_locations,
10331         .fsid_present = _nfs40_proc_fsid_present,
10332 };
10333
10334 #if defined(CONFIG_NFS_V4_1)
10335 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10336         .get_locations = _nfs41_proc_get_locations,
10337         .fsid_present = _nfs41_proc_fsid_present,
10338 };
10339 #endif  /* CONFIG_NFS_V4_1 */
10340
10341 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10342         .minor_version = 0,
10343         .init_caps = NFS_CAP_READDIRPLUS
10344                 | NFS_CAP_ATOMIC_OPEN
10345                 | NFS_CAP_POSIX_LOCK,
10346         .init_client = nfs40_init_client,
10347         .shutdown_client = nfs40_shutdown_client,
10348         .match_stateid = nfs4_match_stateid,
10349         .find_root_sec = nfs4_find_root_sec,
10350         .free_lock_state = nfs4_release_lockowner,
10351         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10352         .alloc_seqid = nfs_alloc_seqid,
10353         .call_sync_ops = &nfs40_call_sync_ops,
10354         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10355         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10356         .state_renewal_ops = &nfs40_state_renewal_ops,
10357         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10358 };
10359
10360 #if defined(CONFIG_NFS_V4_1)
10361 static struct nfs_seqid *
10362 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10363 {
10364         return NULL;
10365 }
10366
10367 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10368         .minor_version = 1,
10369         .init_caps = NFS_CAP_READDIRPLUS
10370                 | NFS_CAP_ATOMIC_OPEN
10371                 | NFS_CAP_POSIX_LOCK
10372                 | NFS_CAP_STATEID_NFSV41
10373                 | NFS_CAP_ATOMIC_OPEN_V1
10374                 | NFS_CAP_LGOPEN,
10375         .init_client = nfs41_init_client,
10376         .shutdown_client = nfs41_shutdown_client,
10377         .match_stateid = nfs41_match_stateid,
10378         .find_root_sec = nfs41_find_root_sec,
10379         .free_lock_state = nfs41_free_lock_state,
10380         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10381         .alloc_seqid = nfs_alloc_no_seqid,
10382         .session_trunk = nfs4_test_session_trunk,
10383         .call_sync_ops = &nfs41_call_sync_ops,
10384         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10385         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10386         .state_renewal_ops = &nfs41_state_renewal_ops,
10387         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10388 };
10389 #endif
10390
10391 #if defined(CONFIG_NFS_V4_2)
10392 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10393         .minor_version = 2,
10394         .init_caps = NFS_CAP_READDIRPLUS
10395                 | NFS_CAP_ATOMIC_OPEN
10396                 | NFS_CAP_POSIX_LOCK
10397                 | NFS_CAP_STATEID_NFSV41
10398                 | NFS_CAP_ATOMIC_OPEN_V1
10399                 | NFS_CAP_LGOPEN
10400                 | NFS_CAP_ALLOCATE
10401                 | NFS_CAP_COPY
10402                 | NFS_CAP_OFFLOAD_CANCEL
10403                 | NFS_CAP_COPY_NOTIFY
10404                 | NFS_CAP_DEALLOCATE
10405                 | NFS_CAP_SEEK
10406                 | NFS_CAP_LAYOUTSTATS
10407                 | NFS_CAP_CLONE
10408                 | NFS_CAP_LAYOUTERROR
10409                 | NFS_CAP_READ_PLUS,
10410         .init_client = nfs41_init_client,
10411         .shutdown_client = nfs41_shutdown_client,
10412         .match_stateid = nfs41_match_stateid,
10413         .find_root_sec = nfs41_find_root_sec,
10414         .free_lock_state = nfs41_free_lock_state,
10415         .call_sync_ops = &nfs41_call_sync_ops,
10416         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10417         .alloc_seqid = nfs_alloc_no_seqid,
10418         .session_trunk = nfs4_test_session_trunk,
10419         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10420         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10421         .state_renewal_ops = &nfs41_state_renewal_ops,
10422         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10423 };
10424 #endif
10425
10426 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10427         [0] = &nfs_v4_0_minor_ops,
10428 #if defined(CONFIG_NFS_V4_1)
10429         [1] = &nfs_v4_1_minor_ops,
10430 #endif
10431 #if defined(CONFIG_NFS_V4_2)
10432         [2] = &nfs_v4_2_minor_ops,
10433 #endif
10434 };
10435
10436 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10437 {
10438         ssize_t error, error2, error3;
10439
10440         error = generic_listxattr(dentry, list, size);
10441         if (error < 0)
10442                 return error;
10443         if (list) {
10444                 list += error;
10445                 size -= error;
10446         }
10447
10448         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10449         if (error2 < 0)
10450                 return error2;
10451
10452         if (list) {
10453                 list += error2;
10454                 size -= error2;
10455         }
10456
10457         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10458         if (error3 < 0)
10459                 return error3;
10460
10461         return error + error2 + error3;
10462 }
10463
10464 static const struct inode_operations nfs4_dir_inode_operations = {
10465         .create         = nfs_create,
10466         .lookup         = nfs_lookup,
10467         .atomic_open    = nfs_atomic_open,
10468         .link           = nfs_link,
10469         .unlink         = nfs_unlink,
10470         .symlink        = nfs_symlink,
10471         .mkdir          = nfs_mkdir,
10472         .rmdir          = nfs_rmdir,
10473         .mknod          = nfs_mknod,
10474         .rename         = nfs_rename,
10475         .permission     = nfs_permission,
10476         .getattr        = nfs_getattr,
10477         .setattr        = nfs_setattr,
10478         .listxattr      = nfs4_listxattr,
10479 };
10480
10481 static const struct inode_operations nfs4_file_inode_operations = {
10482         .permission     = nfs_permission,
10483         .getattr        = nfs_getattr,
10484         .setattr        = nfs_setattr,
10485         .listxattr      = nfs4_listxattr,
10486 };
10487
10488 const struct nfs_rpc_ops nfs_v4_clientops = {
10489         .version        = 4,                    /* protocol version */
10490         .dentry_ops     = &nfs4_dentry_operations,
10491         .dir_inode_ops  = &nfs4_dir_inode_operations,
10492         .file_inode_ops = &nfs4_file_inode_operations,
10493         .file_ops       = &nfs4_file_operations,
10494         .getroot        = nfs4_proc_get_root,
10495         .submount       = nfs4_submount,
10496         .try_get_tree   = nfs4_try_get_tree,
10497         .getattr        = nfs4_proc_getattr,
10498         .setattr        = nfs4_proc_setattr,
10499         .lookup         = nfs4_proc_lookup,
10500         .lookupp        = nfs4_proc_lookupp,
10501         .access         = nfs4_proc_access,
10502         .readlink       = nfs4_proc_readlink,
10503         .create         = nfs4_proc_create,
10504         .remove         = nfs4_proc_remove,
10505         .unlink_setup   = nfs4_proc_unlink_setup,
10506         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10507         .unlink_done    = nfs4_proc_unlink_done,
10508         .rename_setup   = nfs4_proc_rename_setup,
10509         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10510         .rename_done    = nfs4_proc_rename_done,
10511         .link           = nfs4_proc_link,
10512         .symlink        = nfs4_proc_symlink,
10513         .mkdir          = nfs4_proc_mkdir,
10514         .rmdir          = nfs4_proc_rmdir,
10515         .readdir        = nfs4_proc_readdir,
10516         .mknod          = nfs4_proc_mknod,
10517         .statfs         = nfs4_proc_statfs,
10518         .fsinfo         = nfs4_proc_fsinfo,
10519         .pathconf       = nfs4_proc_pathconf,
10520         .set_capabilities = nfs4_server_capabilities,
10521         .decode_dirent  = nfs4_decode_dirent,
10522         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10523         .read_setup     = nfs4_proc_read_setup,
10524         .read_done      = nfs4_read_done,
10525         .write_setup    = nfs4_proc_write_setup,
10526         .write_done     = nfs4_write_done,
10527         .commit_setup   = nfs4_proc_commit_setup,
10528         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10529         .commit_done    = nfs4_commit_done,
10530         .lock           = nfs4_proc_lock,
10531         .clear_acl_cache = nfs4_zap_acl_attr,
10532         .close_context  = nfs4_close_context,
10533         .open_context   = nfs4_atomic_open,
10534         .have_delegation = nfs4_have_delegation,
10535         .alloc_client   = nfs4_alloc_client,
10536         .init_client    = nfs4_init_client,
10537         .free_client    = nfs4_free_client,
10538         .create_server  = nfs4_create_server,
10539         .clone_server   = nfs_clone_server,
10540         .discover_trunking = nfs4_discover_trunking,
10541 };
10542
10543 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10544         .name   = XATTR_NAME_NFSV4_ACL,
10545         .list   = nfs4_xattr_list_nfs4_acl,
10546         .get    = nfs4_xattr_get_nfs4_acl,
10547         .set    = nfs4_xattr_set_nfs4_acl,
10548 };
10549
10550 #ifdef CONFIG_NFS_V4_2
10551 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10552         .prefix = XATTR_USER_PREFIX,
10553         .get    = nfs4_xattr_get_nfs4_user,
10554         .set    = nfs4_xattr_set_nfs4_user,
10555 };
10556 #endif
10557
10558 const struct xattr_handler *nfs4_xattr_handlers[] = {
10559         &nfs4_xattr_nfs4_acl_handler,
10560 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10561         &nfs4_xattr_nfs4_label_handler,
10562 #endif
10563 #ifdef CONFIG_NFS_V4_2
10564         &nfs4_xattr_nfs4_user_handler,
10565 #endif
10566         NULL
10567 };