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                                 NFS_INO_REVAL_PAGECACHE;
1234                 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1235         }
1236         nfsi->attrtimeo_timestamp = jiffies;
1237         nfsi->read_cache_jiffies = timestamp;
1238         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1239         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1240 out:
1241         nfs_set_cache_invalid(inode, cache_validity);
1242 }
1243
1244 void
1245 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1246                 unsigned long timestamp, unsigned long cache_validity)
1247 {
1248         spin_lock(&dir->i_lock);
1249         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1250         spin_unlock(&dir->i_lock);
1251 }
1252
1253 struct nfs4_open_createattrs {
1254         struct nfs4_label *label;
1255         struct iattr *sattr;
1256         const __u32 verf[2];
1257 };
1258
1259 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1260                 int err, struct nfs4_exception *exception)
1261 {
1262         if (err != -EINVAL)
1263                 return false;
1264         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1265                 return false;
1266         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1267         exception->retry = 1;
1268         return true;
1269 }
1270
1271 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1272 {
1273          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1274 }
1275
1276 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1277 {
1278         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1279
1280         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1281 }
1282
1283 static u32
1284 nfs4_map_atomic_open_share(struct nfs_server *server,
1285                 fmode_t fmode, int openflags)
1286 {
1287         u32 res = 0;
1288
1289         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1290         case FMODE_READ:
1291                 res = NFS4_SHARE_ACCESS_READ;
1292                 break;
1293         case FMODE_WRITE:
1294                 res = NFS4_SHARE_ACCESS_WRITE;
1295                 break;
1296         case FMODE_READ|FMODE_WRITE:
1297                 res = NFS4_SHARE_ACCESS_BOTH;
1298         }
1299         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1300                 goto out;
1301         /* Want no delegation if we're using O_DIRECT */
1302         if (openflags & O_DIRECT)
1303                 res |= NFS4_SHARE_WANT_NO_DELEG;
1304 out:
1305         return res;
1306 }
1307
1308 static enum open_claim_type4
1309 nfs4_map_atomic_open_claim(struct nfs_server *server,
1310                 enum open_claim_type4 claim)
1311 {
1312         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1313                 return claim;
1314         switch (claim) {
1315         default:
1316                 return claim;
1317         case NFS4_OPEN_CLAIM_FH:
1318                 return NFS4_OPEN_CLAIM_NULL;
1319         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1320                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1321         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1322                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1323         }
1324 }
1325
1326 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1327 {
1328         p->o_res.f_attr = &p->f_attr;
1329         p->o_res.seqid = p->o_arg.seqid;
1330         p->c_res.seqid = p->c_arg.seqid;
1331         p->o_res.server = p->o_arg.server;
1332         p->o_res.access_request = p->o_arg.access;
1333         nfs_fattr_init(&p->f_attr);
1334         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1335 }
1336
1337 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1338                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1339                 const struct nfs4_open_createattrs *c,
1340                 enum open_claim_type4 claim,
1341                 gfp_t gfp_mask)
1342 {
1343         struct dentry *parent = dget_parent(dentry);
1344         struct inode *dir = d_inode(parent);
1345         struct nfs_server *server = NFS_SERVER(dir);
1346         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1347         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1348         struct nfs4_opendata *p;
1349
1350         p = kzalloc(sizeof(*p), gfp_mask);
1351         if (p == NULL)
1352                 goto err;
1353
1354         p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1355         if (IS_ERR(p->f_attr.label))
1356                 goto err_free_p;
1357
1358         p->a_label = nfs4_label_alloc(server, gfp_mask);
1359         if (IS_ERR(p->a_label))
1360                 goto err_free_f;
1361
1362         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1363         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1364         if (IS_ERR(p->o_arg.seqid))
1365                 goto err_free_label;
1366         nfs_sb_active(dentry->d_sb);
1367         p->dentry = dget(dentry);
1368         p->dir = parent;
1369         p->owner = sp;
1370         atomic_inc(&sp->so_count);
1371         p->o_arg.open_flags = flags;
1372         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1373         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1374         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1375                         fmode, flags);
1376         if (flags & O_CREAT) {
1377                 p->o_arg.umask = current_umask();
1378                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1379                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1380                         p->o_arg.u.attrs = &p->attrs;
1381                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1382
1383                         memcpy(p->o_arg.u.verifier.data, c->verf,
1384                                         sizeof(p->o_arg.u.verifier.data));
1385                 }
1386         }
1387         /* ask server to check for all possible rights as results
1388          * are cached */
1389         switch (p->o_arg.claim) {
1390         default:
1391                 break;
1392         case NFS4_OPEN_CLAIM_NULL:
1393         case NFS4_OPEN_CLAIM_FH:
1394                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1395                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1396                                   NFS4_ACCESS_EXECUTE;
1397 #ifdef CONFIG_NFS_V4_2
1398                 if (!(server->caps & NFS_CAP_XATTR))
1399                         break;
1400                 p->o_arg.access |= NFS4_ACCESS_XAREAD | NFS4_ACCESS_XAWRITE |
1401                                    NFS4_ACCESS_XALIST;
1402 #endif
1403         }
1404         p->o_arg.clientid = server->nfs_client->cl_clientid;
1405         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1406         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1407         p->o_arg.name = &dentry->d_name;
1408         p->o_arg.server = server;
1409         p->o_arg.bitmask = nfs4_bitmask(server, label);
1410         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1411         switch (p->o_arg.claim) {
1412         case NFS4_OPEN_CLAIM_NULL:
1413         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1414         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1415                 p->o_arg.fh = NFS_FH(dir);
1416                 break;
1417         case NFS4_OPEN_CLAIM_PREVIOUS:
1418         case NFS4_OPEN_CLAIM_FH:
1419         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1420         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1421                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1422         }
1423         p->c_arg.fh = &p->o_res.fh;
1424         p->c_arg.stateid = &p->o_res.stateid;
1425         p->c_arg.seqid = p->o_arg.seqid;
1426         nfs4_init_opendata_res(p);
1427         kref_init(&p->kref);
1428         return p;
1429
1430 err_free_label:
1431         nfs4_label_free(p->a_label);
1432 err_free_f:
1433         nfs4_label_free(p->f_attr.label);
1434 err_free_p:
1435         kfree(p);
1436 err:
1437         dput(parent);
1438         return NULL;
1439 }
1440
1441 static void nfs4_opendata_free(struct kref *kref)
1442 {
1443         struct nfs4_opendata *p = container_of(kref,
1444                         struct nfs4_opendata, kref);
1445         struct super_block *sb = p->dentry->d_sb;
1446
1447         nfs4_lgopen_release(p->lgp);
1448         nfs_free_seqid(p->o_arg.seqid);
1449         nfs4_sequence_free_slot(&p->o_res.seq_res);
1450         if (p->state != NULL)
1451                 nfs4_put_open_state(p->state);
1452         nfs4_put_state_owner(p->owner);
1453
1454         nfs4_label_free(p->a_label);
1455         nfs4_label_free(p->f_attr.label);
1456
1457         dput(p->dir);
1458         dput(p->dentry);
1459         nfs_sb_deactive(sb);
1460         nfs_fattr_free_names(&p->f_attr);
1461         kfree(p->f_attr.mdsthreshold);
1462         kfree(p);
1463 }
1464
1465 static void nfs4_opendata_put(struct nfs4_opendata *p)
1466 {
1467         if (p != NULL)
1468                 kref_put(&p->kref, nfs4_opendata_free);
1469 }
1470
1471 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1472                 fmode_t fmode)
1473 {
1474         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1475         case FMODE_READ|FMODE_WRITE:
1476                 return state->n_rdwr != 0;
1477         case FMODE_WRITE:
1478                 return state->n_wronly != 0;
1479         case FMODE_READ:
1480                 return state->n_rdonly != 0;
1481         }
1482         WARN_ON_ONCE(1);
1483         return false;
1484 }
1485
1486 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1487                 int open_mode, enum open_claim_type4 claim)
1488 {
1489         int ret = 0;
1490
1491         if (open_mode & (O_EXCL|O_TRUNC))
1492                 goto out;
1493         switch (claim) {
1494         case NFS4_OPEN_CLAIM_NULL:
1495         case NFS4_OPEN_CLAIM_FH:
1496                 goto out;
1497         default:
1498                 break;
1499         }
1500         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1501                 case FMODE_READ:
1502                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1503                                 && state->n_rdonly != 0;
1504                         break;
1505                 case FMODE_WRITE:
1506                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1507                                 && state->n_wronly != 0;
1508                         break;
1509                 case FMODE_READ|FMODE_WRITE:
1510                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1511                                 && state->n_rdwr != 0;
1512         }
1513 out:
1514         return ret;
1515 }
1516
1517 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1518                 enum open_claim_type4 claim)
1519 {
1520         if (delegation == NULL)
1521                 return 0;
1522         if ((delegation->type & fmode) != fmode)
1523                 return 0;
1524         switch (claim) {
1525         case NFS4_OPEN_CLAIM_NULL:
1526         case NFS4_OPEN_CLAIM_FH:
1527                 break;
1528         case NFS4_OPEN_CLAIM_PREVIOUS:
1529                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1530                         break;
1531                 fallthrough;
1532         default:
1533                 return 0;
1534         }
1535         nfs_mark_delegation_referenced(delegation);
1536         return 1;
1537 }
1538
1539 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1540 {
1541         switch (fmode) {
1542                 case FMODE_WRITE:
1543                         state->n_wronly++;
1544                         break;
1545                 case FMODE_READ:
1546                         state->n_rdonly++;
1547                         break;
1548                 case FMODE_READ|FMODE_WRITE:
1549                         state->n_rdwr++;
1550         }
1551         nfs4_state_set_mode_locked(state, state->state | fmode);
1552 }
1553
1554 #ifdef CONFIG_NFS_V4_1
1555 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1556 {
1557         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1558                 return true;
1559         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1560                 return true;
1561         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1562                 return true;
1563         return false;
1564 }
1565 #endif /* CONFIG_NFS_V4_1 */
1566
1567 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1568 {
1569         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1570                 wake_up_all(&state->waitq);
1571 }
1572
1573 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1574 {
1575         struct nfs_client *clp = state->owner->so_server->nfs_client;
1576         bool need_recover = false;
1577
1578         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1579                 need_recover = true;
1580         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1581                 need_recover = true;
1582         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1583                 need_recover = true;
1584         if (need_recover)
1585                 nfs4_state_mark_reclaim_nograce(clp, state);
1586 }
1587
1588 /*
1589  * Check for whether or not the caller may update the open stateid
1590  * to the value passed in by stateid.
1591  *
1592  * Note: This function relies heavily on the server implementing
1593  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1594  * correctly.
1595  * i.e. The stateid seqids have to be initialised to 1, and
1596  * are then incremented on every state transition.
1597  */
1598 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1599                 const nfs4_stateid *stateid)
1600 {
1601         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1602                 /* The common case - we're updating to a new sequence number */
1603                 if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1604                         if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1605                                 return true;
1606                         return false;
1607                 }
1608                 /* The server returned a new stateid */
1609         }
1610         /* This is the first OPEN in this generation */
1611         if (stateid->seqid == cpu_to_be32(1))
1612                 return true;
1613         return false;
1614 }
1615
1616 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1617 {
1618         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1619                 return;
1620         if (state->n_wronly)
1621                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1622         if (state->n_rdonly)
1623                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1624         if (state->n_rdwr)
1625                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1626         set_bit(NFS_OPEN_STATE, &state->flags);
1627 }
1628
1629 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1630                 nfs4_stateid *stateid, fmode_t fmode)
1631 {
1632         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1633         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1634         case FMODE_WRITE:
1635                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1636                 break;
1637         case FMODE_READ:
1638                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1639                 break;
1640         case 0:
1641                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1642                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1643                 clear_bit(NFS_OPEN_STATE, &state->flags);
1644         }
1645         if (stateid == NULL)
1646                 return;
1647         /* Handle OPEN+OPEN_DOWNGRADE races */
1648         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1649             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1650                 nfs_resync_open_stateid_locked(state);
1651                 goto out;
1652         }
1653         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1654                 nfs4_stateid_copy(&state->stateid, stateid);
1655         nfs4_stateid_copy(&state->open_stateid, stateid);
1656         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1657 out:
1658         nfs_state_log_update_open_stateid(state);
1659 }
1660
1661 static void nfs_clear_open_stateid(struct nfs4_state *state,
1662         nfs4_stateid *arg_stateid,
1663         nfs4_stateid *stateid, fmode_t fmode)
1664 {
1665         write_seqlock(&state->seqlock);
1666         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1667         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1668                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1669         write_sequnlock(&state->seqlock);
1670         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1671                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1672 }
1673
1674 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1675                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1676         __must_hold(&state->owner->so_lock)
1677         __must_hold(&state->seqlock)
1678         __must_hold(RCU)
1679
1680 {
1681         DEFINE_WAIT(wait);
1682         int status = 0;
1683         for (;;) {
1684
1685                 if (nfs_stateid_is_sequential(state, stateid))
1686                         break;
1687
1688                 if (status)
1689                         break;
1690                 /* Rely on seqids for serialisation with NFSv4.0 */
1691                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1692                         break;
1693
1694                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1695                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1696                 /*
1697                  * Ensure we process the state changes in the same order
1698                  * in which the server processed them by delaying the
1699                  * update of the stateid until we are in sequence.
1700                  */
1701                 write_sequnlock(&state->seqlock);
1702                 spin_unlock(&state->owner->so_lock);
1703                 rcu_read_unlock();
1704                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1705
1706                 if (!fatal_signal_pending(current)) {
1707                         if (schedule_timeout(5*HZ) == 0)
1708                                 status = -EAGAIN;
1709                         else
1710                                 status = 0;
1711                 } else
1712                         status = -EINTR;
1713                 finish_wait(&state->waitq, &wait);
1714                 rcu_read_lock();
1715                 spin_lock(&state->owner->so_lock);
1716                 write_seqlock(&state->seqlock);
1717         }
1718
1719         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1720             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1721                 nfs4_stateid_copy(freeme, &state->open_stateid);
1722                 nfs_test_and_clear_all_open_stateid(state);
1723         }
1724
1725         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1726                 nfs4_stateid_copy(&state->stateid, stateid);
1727         nfs4_stateid_copy(&state->open_stateid, stateid);
1728         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1729         nfs_state_log_update_open_stateid(state);
1730 }
1731
1732 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1733                 const nfs4_stateid *open_stateid,
1734                 fmode_t fmode,
1735                 nfs4_stateid *freeme)
1736 {
1737         /*
1738          * Protect the call to nfs4_state_set_mode_locked and
1739          * serialise the stateid update
1740          */
1741         write_seqlock(&state->seqlock);
1742         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1743         switch (fmode) {
1744         case FMODE_READ:
1745                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1746                 break;
1747         case FMODE_WRITE:
1748                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1749                 break;
1750         case FMODE_READ|FMODE_WRITE:
1751                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1752         }
1753         set_bit(NFS_OPEN_STATE, &state->flags);
1754         write_sequnlock(&state->seqlock);
1755 }
1756
1757 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1758 {
1759         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1760         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1761         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1762         clear_bit(NFS_OPEN_STATE, &state->flags);
1763 }
1764
1765 static void nfs_state_set_delegation(struct nfs4_state *state,
1766                 const nfs4_stateid *deleg_stateid,
1767                 fmode_t fmode)
1768 {
1769         /*
1770          * Protect the call to nfs4_state_set_mode_locked and
1771          * serialise the stateid update
1772          */
1773         write_seqlock(&state->seqlock);
1774         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1775         set_bit(NFS_DELEGATED_STATE, &state->flags);
1776         write_sequnlock(&state->seqlock);
1777 }
1778
1779 static void nfs_state_clear_delegation(struct nfs4_state *state)
1780 {
1781         write_seqlock(&state->seqlock);
1782         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1783         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1784         write_sequnlock(&state->seqlock);
1785 }
1786
1787 int update_open_stateid(struct nfs4_state *state,
1788                 const nfs4_stateid *open_stateid,
1789                 const nfs4_stateid *delegation,
1790                 fmode_t fmode)
1791 {
1792         struct nfs_server *server = NFS_SERVER(state->inode);
1793         struct nfs_client *clp = server->nfs_client;
1794         struct nfs_inode *nfsi = NFS_I(state->inode);
1795         struct nfs_delegation *deleg_cur;
1796         nfs4_stateid freeme = { };
1797         int ret = 0;
1798
1799         fmode &= (FMODE_READ|FMODE_WRITE);
1800
1801         rcu_read_lock();
1802         spin_lock(&state->owner->so_lock);
1803         if (open_stateid != NULL) {
1804                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1805                 ret = 1;
1806         }
1807
1808         deleg_cur = nfs4_get_valid_delegation(state->inode);
1809         if (deleg_cur == NULL)
1810                 goto no_delegation;
1811
1812         spin_lock(&deleg_cur->lock);
1813         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1814            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1815             (deleg_cur->type & fmode) != fmode)
1816                 goto no_delegation_unlock;
1817
1818         if (delegation == NULL)
1819                 delegation = &deleg_cur->stateid;
1820         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1821                 goto no_delegation_unlock;
1822
1823         nfs_mark_delegation_referenced(deleg_cur);
1824         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1825         ret = 1;
1826 no_delegation_unlock:
1827         spin_unlock(&deleg_cur->lock);
1828 no_delegation:
1829         if (ret)
1830                 update_open_stateflags(state, fmode);
1831         spin_unlock(&state->owner->so_lock);
1832         rcu_read_unlock();
1833
1834         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1835                 nfs4_schedule_state_manager(clp);
1836         if (freeme.type != 0)
1837                 nfs4_test_and_free_stateid(server, &freeme,
1838                                 state->owner->so_cred);
1839
1840         return ret;
1841 }
1842
1843 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1844                 const nfs4_stateid *stateid)
1845 {
1846         struct nfs4_state *state = lsp->ls_state;
1847         bool ret = false;
1848
1849         spin_lock(&state->state_lock);
1850         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1851                 goto out_noupdate;
1852         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1853                 goto out_noupdate;
1854         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1855         ret = true;
1856 out_noupdate:
1857         spin_unlock(&state->state_lock);
1858         return ret;
1859 }
1860
1861 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1862 {
1863         struct nfs_delegation *delegation;
1864
1865         fmode &= FMODE_READ|FMODE_WRITE;
1866         rcu_read_lock();
1867         delegation = nfs4_get_valid_delegation(inode);
1868         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1869                 rcu_read_unlock();
1870                 return;
1871         }
1872         rcu_read_unlock();
1873         nfs4_inode_return_delegation(inode);
1874 }
1875
1876 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1877 {
1878         struct nfs4_state *state = opendata->state;
1879         struct nfs_delegation *delegation;
1880         int open_mode = opendata->o_arg.open_flags;
1881         fmode_t fmode = opendata->o_arg.fmode;
1882         enum open_claim_type4 claim = opendata->o_arg.claim;
1883         nfs4_stateid stateid;
1884         int ret = -EAGAIN;
1885
1886         for (;;) {
1887                 spin_lock(&state->owner->so_lock);
1888                 if (can_open_cached(state, fmode, open_mode, claim)) {
1889                         update_open_stateflags(state, fmode);
1890                         spin_unlock(&state->owner->so_lock);
1891                         goto out_return_state;
1892                 }
1893                 spin_unlock(&state->owner->so_lock);
1894                 rcu_read_lock();
1895                 delegation = nfs4_get_valid_delegation(state->inode);
1896                 if (!can_open_delegated(delegation, fmode, claim)) {
1897                         rcu_read_unlock();
1898                         break;
1899                 }
1900                 /* Save the delegation */
1901                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1902                 rcu_read_unlock();
1903                 nfs_release_seqid(opendata->o_arg.seqid);
1904                 if (!opendata->is_recover) {
1905                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1906                         if (ret != 0)
1907                                 goto out;
1908                 }
1909                 ret = -EAGAIN;
1910
1911                 /* Try to update the stateid using the delegation */
1912                 if (update_open_stateid(state, NULL, &stateid, fmode))
1913                         goto out_return_state;
1914         }
1915 out:
1916         return ERR_PTR(ret);
1917 out_return_state:
1918         refcount_inc(&state->count);
1919         return state;
1920 }
1921
1922 static void
1923 nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1924 {
1925         struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1926         struct nfs_delegation *delegation;
1927         int delegation_flags = 0;
1928
1929         rcu_read_lock();
1930         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1931         if (delegation)
1932                 delegation_flags = delegation->flags;
1933         rcu_read_unlock();
1934         switch (data->o_arg.claim) {
1935         default:
1936                 break;
1937         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1938         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1939                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1940                                    "returning a delegation for "
1941                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
1942                                    clp->cl_hostname);
1943                 return;
1944         }
1945         if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1946                 nfs_inode_set_delegation(state->inode,
1947                                 data->owner->so_cred,
1948                                 data->o_res.delegation_type,
1949                                 &data->o_res.delegation,
1950                                 data->o_res.pagemod_limit);
1951         else
1952                 nfs_inode_reclaim_delegation(state->inode,
1953                                 data->owner->so_cred,
1954                                 data->o_res.delegation_type,
1955                                 &data->o_res.delegation,
1956                                 data->o_res.pagemod_limit);
1957
1958         if (data->o_res.do_recall)
1959                 nfs_async_inode_return_delegation(state->inode,
1960                                                   &data->o_res.delegation);
1961 }
1962
1963 /*
1964  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1965  * and update the nfs4_state.
1966  */
1967 static struct nfs4_state *
1968 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
1969 {
1970         struct inode *inode = data->state->inode;
1971         struct nfs4_state *state = data->state;
1972         int ret;
1973
1974         if (!data->rpc_done) {
1975                 if (data->rpc_status)
1976                         return ERR_PTR(data->rpc_status);
1977                 /* cached opens have already been processed */
1978                 goto update;
1979         }
1980
1981         ret = nfs_refresh_inode(inode, &data->f_attr);
1982         if (ret)
1983                 return ERR_PTR(ret);
1984
1985         if (data->o_res.delegation_type != 0)
1986                 nfs4_opendata_check_deleg(data, state);
1987 update:
1988         if (!update_open_stateid(state, &data->o_res.stateid,
1989                                 NULL, data->o_arg.fmode))
1990                 return ERR_PTR(-EAGAIN);
1991         refcount_inc(&state->count);
1992
1993         return state;
1994 }
1995
1996 static struct inode *
1997 nfs4_opendata_get_inode(struct nfs4_opendata *data)
1998 {
1999         struct inode *inode;
2000
2001         switch (data->o_arg.claim) {
2002         case NFS4_OPEN_CLAIM_NULL:
2003         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2004         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2005                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2006                         return ERR_PTR(-EAGAIN);
2007                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2008                                 &data->f_attr);
2009                 break;
2010         default:
2011                 inode = d_inode(data->dentry);
2012                 ihold(inode);
2013                 nfs_refresh_inode(inode, &data->f_attr);
2014         }
2015         return inode;
2016 }
2017
2018 static struct nfs4_state *
2019 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2020 {
2021         struct nfs4_state *state;
2022         struct inode *inode;
2023
2024         inode = nfs4_opendata_get_inode(data);
2025         if (IS_ERR(inode))
2026                 return ERR_CAST(inode);
2027         if (data->state != NULL && data->state->inode == inode) {
2028                 state = data->state;
2029                 refcount_inc(&state->count);
2030         } else
2031                 state = nfs4_get_open_state(inode, data->owner);
2032         iput(inode);
2033         if (state == NULL)
2034                 state = ERR_PTR(-ENOMEM);
2035         return state;
2036 }
2037
2038 static struct nfs4_state *
2039 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2040 {
2041         struct nfs4_state *state;
2042
2043         if (!data->rpc_done) {
2044                 state = nfs4_try_open_cached(data);
2045                 trace_nfs4_cached_open(data->state);
2046                 goto out;
2047         }
2048
2049         state = nfs4_opendata_find_nfs4_state(data);
2050         if (IS_ERR(state))
2051                 goto out;
2052
2053         if (data->o_res.delegation_type != 0)
2054                 nfs4_opendata_check_deleg(data, state);
2055         if (!update_open_stateid(state, &data->o_res.stateid,
2056                                 NULL, data->o_arg.fmode)) {
2057                 nfs4_put_open_state(state);
2058                 state = ERR_PTR(-EAGAIN);
2059         }
2060 out:
2061         nfs_release_seqid(data->o_arg.seqid);
2062         return state;
2063 }
2064
2065 static struct nfs4_state *
2066 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2067 {
2068         struct nfs4_state *ret;
2069
2070         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2071                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2072         else
2073                 ret = _nfs4_opendata_to_nfs4_state(data);
2074         nfs4_sequence_free_slot(&data->o_res.seq_res);
2075         return ret;
2076 }
2077
2078 static struct nfs_open_context *
2079 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2080 {
2081         struct nfs_inode *nfsi = NFS_I(state->inode);
2082         struct nfs_open_context *ctx;
2083
2084         rcu_read_lock();
2085         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2086                 if (ctx->state != state)
2087                         continue;
2088                 if ((ctx->mode & mode) != mode)
2089                         continue;
2090                 if (!get_nfs_open_context(ctx))
2091                         continue;
2092                 rcu_read_unlock();
2093                 return ctx;
2094         }
2095         rcu_read_unlock();
2096         return ERR_PTR(-ENOENT);
2097 }
2098
2099 static struct nfs_open_context *
2100 nfs4_state_find_open_context(struct nfs4_state *state)
2101 {
2102         struct nfs_open_context *ctx;
2103
2104         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2105         if (!IS_ERR(ctx))
2106                 return ctx;
2107         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2108         if (!IS_ERR(ctx))
2109                 return ctx;
2110         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2111 }
2112
2113 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2114                 struct nfs4_state *state, enum open_claim_type4 claim)
2115 {
2116         struct nfs4_opendata *opendata;
2117
2118         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2119                         NULL, claim, GFP_NOFS);
2120         if (opendata == NULL)
2121                 return ERR_PTR(-ENOMEM);
2122         opendata->state = state;
2123         refcount_inc(&state->count);
2124         return opendata;
2125 }
2126
2127 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2128                 fmode_t fmode)
2129 {
2130         struct nfs4_state *newstate;
2131         int ret;
2132
2133         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2134                 return 0;
2135         opendata->o_arg.open_flags = 0;
2136         opendata->o_arg.fmode = fmode;
2137         opendata->o_arg.share_access = nfs4_map_atomic_open_share(
2138                         NFS_SB(opendata->dentry->d_sb),
2139                         fmode, 0);
2140         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2141         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2142         nfs4_init_opendata_res(opendata);
2143         ret = _nfs4_recover_proc_open(opendata);
2144         if (ret != 0)
2145                 return ret;
2146         newstate = nfs4_opendata_to_nfs4_state(opendata);
2147         if (IS_ERR(newstate))
2148                 return PTR_ERR(newstate);
2149         if (newstate != opendata->state)
2150                 ret = -ESTALE;
2151         nfs4_close_state(newstate, fmode);
2152         return ret;
2153 }
2154
2155 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2156 {
2157         int ret;
2158
2159         /* memory barrier prior to reading state->n_* */
2160         smp_rmb();
2161         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2162         if (ret != 0)
2163                 return ret;
2164         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2165         if (ret != 0)
2166                 return ret;
2167         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2168         if (ret != 0)
2169                 return ret;
2170         /*
2171          * We may have performed cached opens for all three recoveries.
2172          * Check if we need to update the current stateid.
2173          */
2174         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2175             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2176                 write_seqlock(&state->seqlock);
2177                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2178                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2179                 write_sequnlock(&state->seqlock);
2180         }
2181         return 0;
2182 }
2183
2184 /*
2185  * OPEN_RECLAIM:
2186  *      reclaim state on the server after a reboot.
2187  */
2188 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2189 {
2190         struct nfs_delegation *delegation;
2191         struct nfs4_opendata *opendata;
2192         fmode_t delegation_type = 0;
2193         int status;
2194
2195         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2196                         NFS4_OPEN_CLAIM_PREVIOUS);
2197         if (IS_ERR(opendata))
2198                 return PTR_ERR(opendata);
2199         rcu_read_lock();
2200         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2201         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2202                 delegation_type = delegation->type;
2203         rcu_read_unlock();
2204         opendata->o_arg.u.delegation_type = delegation_type;
2205         status = nfs4_open_recover(opendata, state);
2206         nfs4_opendata_put(opendata);
2207         return status;
2208 }
2209
2210 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2211 {
2212         struct nfs_server *server = NFS_SERVER(state->inode);
2213         struct nfs4_exception exception = { };
2214         int err;
2215         do {
2216                 err = _nfs4_do_open_reclaim(ctx, state);
2217                 trace_nfs4_open_reclaim(ctx, 0, err);
2218                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2219                         continue;
2220                 if (err != -NFS4ERR_DELAY)
2221                         break;
2222                 nfs4_handle_exception(server, err, &exception);
2223         } while (exception.retry);
2224         return err;
2225 }
2226
2227 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2228 {
2229         struct nfs_open_context *ctx;
2230         int ret;
2231
2232         ctx = nfs4_state_find_open_context(state);
2233         if (IS_ERR(ctx))
2234                 return -EAGAIN;
2235         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2236         nfs_state_clear_open_state_flags(state);
2237         ret = nfs4_do_open_reclaim(ctx, state);
2238         put_nfs_open_context(ctx);
2239         return ret;
2240 }
2241
2242 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)
2243 {
2244         switch (err) {
2245                 default:
2246                         printk(KERN_ERR "NFS: %s: unhandled error "
2247                                         "%d.\n", __func__, err);
2248                         fallthrough;
2249                 case 0:
2250                 case -ENOENT:
2251                 case -EAGAIN:
2252                 case -ESTALE:
2253                 case -ETIMEDOUT:
2254                         break;
2255                 case -NFS4ERR_BADSESSION:
2256                 case -NFS4ERR_BADSLOT:
2257                 case -NFS4ERR_BAD_HIGH_SLOT:
2258                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2259                 case -NFS4ERR_DEADSESSION:
2260                         return -EAGAIN;
2261                 case -NFS4ERR_STALE_CLIENTID:
2262                 case -NFS4ERR_STALE_STATEID:
2263                         /* Don't recall a delegation if it was lost */
2264                         nfs4_schedule_lease_recovery(server->nfs_client);
2265                         return -EAGAIN;
2266                 case -NFS4ERR_MOVED:
2267                         nfs4_schedule_migration_recovery(server);
2268                         return -EAGAIN;
2269                 case -NFS4ERR_LEASE_MOVED:
2270                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2271                         return -EAGAIN;
2272                 case -NFS4ERR_DELEG_REVOKED:
2273                 case -NFS4ERR_ADMIN_REVOKED:
2274                 case -NFS4ERR_EXPIRED:
2275                 case -NFS4ERR_BAD_STATEID:
2276                 case -NFS4ERR_OPENMODE:
2277                         nfs_inode_find_state_and_recover(state->inode,
2278                                         stateid);
2279                         nfs4_schedule_stateid_recovery(server, state);
2280                         return -EAGAIN;
2281                 case -NFS4ERR_DELAY:
2282                 case -NFS4ERR_GRACE:
2283                         ssleep(1);
2284                         return -EAGAIN;
2285                 case -ENOMEM:
2286                 case -NFS4ERR_DENIED:
2287                         if (fl) {
2288                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2289                                 if (lsp)
2290                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2291                         }
2292                         return 0;
2293         }
2294         return err;
2295 }
2296
2297 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2298                 struct nfs4_state *state, const nfs4_stateid *stateid)
2299 {
2300         struct nfs_server *server = NFS_SERVER(state->inode);
2301         struct nfs4_opendata *opendata;
2302         int err = 0;
2303
2304         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2305                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2306         if (IS_ERR(opendata))
2307                 return PTR_ERR(opendata);
2308         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2309         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2310                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2311                 if (err)
2312                         goto out;
2313         }
2314         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2315                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2316                 if (err)
2317                         goto out;
2318         }
2319         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2320                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2321                 if (err)
2322                         goto out;
2323         }
2324         nfs_state_clear_delegation(state);
2325 out:
2326         nfs4_opendata_put(opendata);
2327         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2328 }
2329
2330 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2331 {
2332         struct nfs4_opendata *data = calldata;
2333
2334         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2335                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2336 }
2337
2338 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2339 {
2340         struct nfs4_opendata *data = calldata;
2341
2342         nfs40_sequence_done(task, &data->c_res.seq_res);
2343
2344         data->rpc_status = task->tk_status;
2345         if (data->rpc_status == 0) {
2346                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2347                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2348                 renew_lease(data->o_res.server, data->timestamp);
2349                 data->rpc_done = true;
2350         }
2351 }
2352
2353 static void nfs4_open_confirm_release(void *calldata)
2354 {
2355         struct nfs4_opendata *data = calldata;
2356         struct nfs4_state *state = NULL;
2357
2358         /* If this request hasn't been cancelled, do nothing */
2359         if (!data->cancelled)
2360                 goto out_free;
2361         /* In case of error, no cleanup! */
2362         if (!data->rpc_done)
2363                 goto out_free;
2364         state = nfs4_opendata_to_nfs4_state(data);
2365         if (!IS_ERR(state))
2366                 nfs4_close_state(state, data->o_arg.fmode);
2367 out_free:
2368         nfs4_opendata_put(data);
2369 }
2370
2371 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2372         .rpc_call_prepare = nfs4_open_confirm_prepare,
2373         .rpc_call_done = nfs4_open_confirm_done,
2374         .rpc_release = nfs4_open_confirm_release,
2375 };
2376
2377 /*
2378  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2379  */
2380 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2381 {
2382         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2383         struct rpc_task *task;
2384         struct  rpc_message msg = {
2385                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2386                 .rpc_argp = &data->c_arg,
2387                 .rpc_resp = &data->c_res,
2388                 .rpc_cred = data->owner->so_cred,
2389         };
2390         struct rpc_task_setup task_setup_data = {
2391                 .rpc_client = server->client,
2392                 .rpc_message = &msg,
2393                 .callback_ops = &nfs4_open_confirm_ops,
2394                 .callback_data = data,
2395                 .workqueue = nfsiod_workqueue,
2396                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2397         };
2398         int status;
2399
2400         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2401                                 data->is_recover);
2402         kref_get(&data->kref);
2403         data->rpc_done = false;
2404         data->rpc_status = 0;
2405         data->timestamp = jiffies;
2406         task = rpc_run_task(&task_setup_data);
2407         if (IS_ERR(task))
2408                 return PTR_ERR(task);
2409         status = rpc_wait_for_completion_task(task);
2410         if (status != 0) {
2411                 data->cancelled = true;
2412                 smp_wmb();
2413         } else
2414                 status = data->rpc_status;
2415         rpc_put_task(task);
2416         return status;
2417 }
2418
2419 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2420 {
2421         struct nfs4_opendata *data = calldata;
2422         struct nfs4_state_owner *sp = data->owner;
2423         struct nfs_client *clp = sp->so_server->nfs_client;
2424         enum open_claim_type4 claim = data->o_arg.claim;
2425
2426         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2427                 goto out_wait;
2428         /*
2429          * Check if we still need to send an OPEN call, or if we can use
2430          * a delegation instead.
2431          */
2432         if (data->state != NULL) {
2433                 struct nfs_delegation *delegation;
2434
2435                 if (can_open_cached(data->state, data->o_arg.fmode,
2436                                         data->o_arg.open_flags, claim))
2437                         goto out_no_action;
2438                 rcu_read_lock();
2439                 delegation = nfs4_get_valid_delegation(data->state->inode);
2440                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2441                         goto unlock_no_action;
2442                 rcu_read_unlock();
2443         }
2444         /* Update client id. */
2445         data->o_arg.clientid = clp->cl_clientid;
2446         switch (claim) {
2447         default:
2448                 break;
2449         case NFS4_OPEN_CLAIM_PREVIOUS:
2450         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2451         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2452                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2453                 fallthrough;
2454         case NFS4_OPEN_CLAIM_FH:
2455                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2456         }
2457         data->timestamp = jiffies;
2458         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2459                                 &data->o_arg.seq_args,
2460                                 &data->o_res.seq_res,
2461                                 task) != 0)
2462                 nfs_release_seqid(data->o_arg.seqid);
2463
2464         /* Set the create mode (note dependency on the session type) */
2465         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2466         if (data->o_arg.open_flags & O_EXCL) {
2467                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2468                 if (clp->cl_mvops->minor_version == 0) {
2469                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2470                         /* don't put an ACCESS op in OPEN compound if O_EXCL,
2471                          * because ACCESS will return permission denied for
2472                          * all bits until close */
2473                         data->o_res.access_request = data->o_arg.access = 0;
2474                 } else if (nfs4_has_persistent_session(clp))
2475                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2476         }
2477         return;
2478 unlock_no_action:
2479         trace_nfs4_cached_open(data->state);
2480         rcu_read_unlock();
2481 out_no_action:
2482         task->tk_action = NULL;
2483 out_wait:
2484         nfs4_sequence_done(task, &data->o_res.seq_res);
2485 }
2486
2487 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2488 {
2489         struct nfs4_opendata *data = calldata;
2490
2491         data->rpc_status = task->tk_status;
2492
2493         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2494                 return;
2495
2496         if (task->tk_status == 0) {
2497                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2498                         switch (data->o_res.f_attr->mode & S_IFMT) {
2499                         case S_IFREG:
2500                                 break;
2501                         case S_IFLNK:
2502                                 data->rpc_status = -ELOOP;
2503                                 break;
2504                         case S_IFDIR:
2505                                 data->rpc_status = -EISDIR;
2506                                 break;
2507                         default:
2508                                 data->rpc_status = -ENOTDIR;
2509                         }
2510                 }
2511                 renew_lease(data->o_res.server, data->timestamp);
2512                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2513                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2514         }
2515         data->rpc_done = true;
2516 }
2517
2518 static void nfs4_open_release(void *calldata)
2519 {
2520         struct nfs4_opendata *data = calldata;
2521         struct nfs4_state *state = NULL;
2522
2523         /* If this request hasn't been cancelled, do nothing */
2524         if (!data->cancelled)
2525                 goto out_free;
2526         /* In case of error, no cleanup! */
2527         if (data->rpc_status != 0 || !data->rpc_done)
2528                 goto out_free;
2529         /* In case we need an open_confirm, no cleanup! */
2530         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2531                 goto out_free;
2532         state = nfs4_opendata_to_nfs4_state(data);
2533         if (!IS_ERR(state))
2534                 nfs4_close_state(state, data->o_arg.fmode);
2535 out_free:
2536         nfs4_opendata_put(data);
2537 }
2538
2539 static const struct rpc_call_ops nfs4_open_ops = {
2540         .rpc_call_prepare = nfs4_open_prepare,
2541         .rpc_call_done = nfs4_open_done,
2542         .rpc_release = nfs4_open_release,
2543 };
2544
2545 static int nfs4_run_open_task(struct nfs4_opendata *data,
2546                               struct nfs_open_context *ctx)
2547 {
2548         struct inode *dir = d_inode(data->dir);
2549         struct nfs_server *server = NFS_SERVER(dir);
2550         struct nfs_openargs *o_arg = &data->o_arg;
2551         struct nfs_openres *o_res = &data->o_res;
2552         struct rpc_task *task;
2553         struct rpc_message msg = {
2554                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2555                 .rpc_argp = o_arg,
2556                 .rpc_resp = o_res,
2557                 .rpc_cred = data->owner->so_cred,
2558         };
2559         struct rpc_task_setup task_setup_data = {
2560                 .rpc_client = server->client,
2561                 .rpc_message = &msg,
2562                 .callback_ops = &nfs4_open_ops,
2563                 .callback_data = data,
2564                 .workqueue = nfsiod_workqueue,
2565                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2566         };
2567         int status;
2568
2569         if (server->nfs_client->cl_minorversion)
2570                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
2571
2572         kref_get(&data->kref);
2573         data->rpc_done = false;
2574         data->rpc_status = 0;
2575         data->cancelled = false;
2576         data->is_recover = false;
2577         if (!ctx) {
2578                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2579                 data->is_recover = true;
2580                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2581         } else {
2582                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2583                 pnfs_lgopen_prepare(data, ctx);
2584         }
2585         task = rpc_run_task(&task_setup_data);
2586         if (IS_ERR(task))
2587                 return PTR_ERR(task);
2588         status = rpc_wait_for_completion_task(task);
2589         if (status != 0) {
2590                 data->cancelled = true;
2591                 smp_wmb();
2592         } else
2593                 status = data->rpc_status;
2594         rpc_put_task(task);
2595
2596         return status;
2597 }
2598
2599 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2600 {
2601         struct inode *dir = d_inode(data->dir);
2602         struct nfs_openres *o_res = &data->o_res;
2603         int status;
2604
2605         status = nfs4_run_open_task(data, NULL);
2606         if (status != 0 || !data->rpc_done)
2607                 return status;
2608
2609         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2610
2611         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2612                 status = _nfs4_proc_open_confirm(data);
2613
2614         return status;
2615 }
2616
2617 /*
2618  * Additional permission checks in order to distinguish between an
2619  * open for read, and an open for execute. This works around the
2620  * fact that NFSv4 OPEN treats read and execute permissions as being
2621  * the same.
2622  * Note that in the non-execute case, we want to turn off permission
2623  * checking if we just created a new file (POSIX open() semantics).
2624  */
2625 static int nfs4_opendata_access(const struct cred *cred,
2626                                 struct nfs4_opendata *opendata,
2627                                 struct nfs4_state *state, fmode_t fmode,
2628                                 int openflags)
2629 {
2630         struct nfs_access_entry cache;
2631         u32 mask, flags;
2632
2633         /* access call failed or for some reason the server doesn't
2634          * support any access modes -- defer access call until later */
2635         if (opendata->o_res.access_supported == 0)
2636                 return 0;
2637
2638         mask = 0;
2639         /*
2640          * Use openflags to check for exec, because fmode won't
2641          * always have FMODE_EXEC set when file open for exec.
2642          */
2643         if (openflags & __FMODE_EXEC) {
2644                 /* ONLY check for exec rights */
2645                 if (S_ISDIR(state->inode->i_mode))
2646                         mask = NFS4_ACCESS_LOOKUP;
2647                 else
2648                         mask = NFS4_ACCESS_EXECUTE;
2649         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2650                 mask = NFS4_ACCESS_READ;
2651
2652         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2653         nfs_access_add_cache(state->inode, &cache, cred);
2654
2655         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2656         if ((mask & ~cache.mask & flags) == 0)
2657                 return 0;
2658
2659         return -EACCES;
2660 }
2661
2662 /*
2663  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2664  */
2665 static int _nfs4_proc_open(struct nfs4_opendata *data,
2666                            struct nfs_open_context *ctx)
2667 {
2668         struct inode *dir = d_inode(data->dir);
2669         struct nfs_server *server = NFS_SERVER(dir);
2670         struct nfs_openargs *o_arg = &data->o_arg;
2671         struct nfs_openres *o_res = &data->o_res;
2672         int status;
2673
2674         status = nfs4_run_open_task(data, ctx);
2675         if (!data->rpc_done)
2676                 return status;
2677         if (status != 0) {
2678                 if (status == -NFS4ERR_BADNAME &&
2679                                 !(o_arg->open_flags & O_CREAT))
2680                         return -ENOENT;
2681                 return status;
2682         }
2683
2684         nfs_fattr_map_and_free_names(server, &data->f_attr);
2685
2686         if (o_arg->open_flags & O_CREAT) {
2687                 if (o_arg->open_flags & O_EXCL)
2688                         data->file_created = true;
2689                 else if (o_res->cinfo.before != o_res->cinfo.after)
2690                         data->file_created = true;
2691                 if (data->file_created ||
2692                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2693                         nfs4_update_changeattr(dir, &o_res->cinfo,
2694                                         o_res->f_attr->time_start,
2695                                         NFS_INO_INVALID_DATA);
2696         }
2697         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2698                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2699         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2700                 status = _nfs4_proc_open_confirm(data);
2701                 if (status != 0)
2702                         return status;
2703         }
2704         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2705                 nfs4_sequence_free_slot(&o_res->seq_res);
2706                 nfs4_proc_getattr(server, &o_res->fh, o_res->f_attr, NULL);
2707         }
2708         return 0;
2709 }
2710
2711 /*
2712  * OPEN_EXPIRED:
2713  *      reclaim state on the server after a network partition.
2714  *      Assumes caller holds the appropriate lock
2715  */
2716 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2717 {
2718         struct nfs4_opendata *opendata;
2719         int ret;
2720
2721         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2722                         NFS4_OPEN_CLAIM_FH);
2723         if (IS_ERR(opendata))
2724                 return PTR_ERR(opendata);
2725         ret = nfs4_open_recover(opendata, state);
2726         if (ret == -ESTALE)
2727                 d_drop(ctx->dentry);
2728         nfs4_opendata_put(opendata);
2729         return ret;
2730 }
2731
2732 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2733 {
2734         struct nfs_server *server = NFS_SERVER(state->inode);
2735         struct nfs4_exception exception = { };
2736         int err;
2737
2738         do {
2739                 err = _nfs4_open_expired(ctx, state);
2740                 trace_nfs4_open_expired(ctx, 0, err);
2741                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2742                         continue;
2743                 switch (err) {
2744                 default:
2745                         goto out;
2746                 case -NFS4ERR_GRACE:
2747                 case -NFS4ERR_DELAY:
2748                         nfs4_handle_exception(server, err, &exception);
2749                         err = 0;
2750                 }
2751         } while (exception.retry);
2752 out:
2753         return err;
2754 }
2755
2756 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2757 {
2758         struct nfs_open_context *ctx;
2759         int ret;
2760
2761         ctx = nfs4_state_find_open_context(state);
2762         if (IS_ERR(ctx))
2763                 return -EAGAIN;
2764         ret = nfs4_do_open_expired(ctx, state);
2765         put_nfs_open_context(ctx);
2766         return ret;
2767 }
2768
2769 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2770                 const nfs4_stateid *stateid)
2771 {
2772         nfs_remove_bad_delegation(state->inode, stateid);
2773         nfs_state_clear_delegation(state);
2774 }
2775
2776 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2777 {
2778         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2779                 nfs_finish_clear_delegation_stateid(state, NULL);
2780 }
2781
2782 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2783 {
2784         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2785         nfs40_clear_delegation_stateid(state);
2786         nfs_state_clear_open_state_flags(state);
2787         return nfs4_open_expired(sp, state);
2788 }
2789
2790 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2791                 nfs4_stateid *stateid,
2792                 const struct cred *cred)
2793 {
2794         return -NFS4ERR_BAD_STATEID;
2795 }
2796
2797 #if defined(CONFIG_NFS_V4_1)
2798 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2799                 nfs4_stateid *stateid,
2800                 const struct cred *cred)
2801 {
2802         int status;
2803
2804         switch (stateid->type) {
2805         default:
2806                 break;
2807         case NFS4_INVALID_STATEID_TYPE:
2808         case NFS4_SPECIAL_STATEID_TYPE:
2809                 return -NFS4ERR_BAD_STATEID;
2810         case NFS4_REVOKED_STATEID_TYPE:
2811                 goto out_free;
2812         }
2813
2814         status = nfs41_test_stateid(server, stateid, cred);
2815         switch (status) {
2816         case -NFS4ERR_EXPIRED:
2817         case -NFS4ERR_ADMIN_REVOKED:
2818         case -NFS4ERR_DELEG_REVOKED:
2819                 break;
2820         default:
2821                 return status;
2822         }
2823 out_free:
2824         /* Ack the revoked state to the server */
2825         nfs41_free_stateid(server, stateid, cred, true);
2826         return -NFS4ERR_EXPIRED;
2827 }
2828
2829 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2830 {
2831         struct nfs_server *server = NFS_SERVER(state->inode);
2832         nfs4_stateid stateid;
2833         struct nfs_delegation *delegation;
2834         const struct cred *cred = NULL;
2835         int status, ret = NFS_OK;
2836
2837         /* Get the delegation credential for use by test/free_stateid */
2838         rcu_read_lock();
2839         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2840         if (delegation == NULL) {
2841                 rcu_read_unlock();
2842                 nfs_state_clear_delegation(state);
2843                 return NFS_OK;
2844         }
2845
2846         spin_lock(&delegation->lock);
2847         nfs4_stateid_copy(&stateid, &delegation->stateid);
2848
2849         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2850                                 &delegation->flags)) {
2851                 spin_unlock(&delegation->lock);
2852                 rcu_read_unlock();
2853                 return NFS_OK;
2854         }
2855
2856         if (delegation->cred)
2857                 cred = get_cred(delegation->cred);
2858         spin_unlock(&delegation->lock);
2859         rcu_read_unlock();
2860         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2861         trace_nfs4_test_delegation_stateid(state, NULL, status);
2862         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2863                 nfs_finish_clear_delegation_stateid(state, &stateid);
2864         else
2865                 ret = status;
2866
2867         put_cred(cred);
2868         return ret;
2869 }
2870
2871 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2872 {
2873         nfs4_stateid tmp;
2874
2875         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2876             nfs4_copy_delegation_stateid(state->inode, state->state,
2877                                 &tmp, NULL) &&
2878             nfs4_stateid_match_other(&state->stateid, &tmp))
2879                 nfs_state_set_delegation(state, &tmp, state->state);
2880         else
2881                 nfs_state_clear_delegation(state);
2882 }
2883
2884 /**
2885  * nfs41_check_expired_locks - possibly free a lock stateid
2886  *
2887  * @state: NFSv4 state for an inode
2888  *
2889  * Returns NFS_OK if recovery for this stateid is now finished.
2890  * Otherwise a negative NFS4ERR value is returned.
2891  */
2892 static int nfs41_check_expired_locks(struct nfs4_state *state)
2893 {
2894         int status, ret = NFS_OK;
2895         struct nfs4_lock_state *lsp, *prev = NULL;
2896         struct nfs_server *server = NFS_SERVER(state->inode);
2897
2898         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2899                 goto out;
2900
2901         spin_lock(&state->state_lock);
2902         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2903                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2904                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2905
2906                         refcount_inc(&lsp->ls_count);
2907                         spin_unlock(&state->state_lock);
2908
2909                         nfs4_put_lock_state(prev);
2910                         prev = lsp;
2911
2912                         status = nfs41_test_and_free_expired_stateid(server,
2913                                         &lsp->ls_stateid,
2914                                         cred);
2915                         trace_nfs4_test_lock_stateid(state, lsp, status);
2916                         if (status == -NFS4ERR_EXPIRED ||
2917                             status == -NFS4ERR_BAD_STATEID) {
2918                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2919                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2920                                 if (!recover_lost_locks)
2921                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2922                         } else if (status != NFS_OK) {
2923                                 ret = status;
2924                                 nfs4_put_lock_state(prev);
2925                                 goto out;
2926                         }
2927                         spin_lock(&state->state_lock);
2928                 }
2929         }
2930         spin_unlock(&state->state_lock);
2931         nfs4_put_lock_state(prev);
2932 out:
2933         return ret;
2934 }
2935
2936 /**
2937  * nfs41_check_open_stateid - possibly free an open stateid
2938  *
2939  * @state: NFSv4 state for an inode
2940  *
2941  * Returns NFS_OK if recovery for this stateid is now finished.
2942  * Otherwise a negative NFS4ERR value is returned.
2943  */
2944 static int nfs41_check_open_stateid(struct nfs4_state *state)
2945 {
2946         struct nfs_server *server = NFS_SERVER(state->inode);
2947         nfs4_stateid *stateid = &state->open_stateid;
2948         const struct cred *cred = state->owner->so_cred;
2949         int status;
2950
2951         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2952                 return -NFS4ERR_BAD_STATEID;
2953         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2954         trace_nfs4_test_open_stateid(state, NULL, status);
2955         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2956                 nfs_state_clear_open_state_flags(state);
2957                 stateid->type = NFS4_INVALID_STATEID_TYPE;
2958                 return status;
2959         }
2960         if (nfs_open_stateid_recover_openmode(state))
2961                 return -NFS4ERR_OPENMODE;
2962         return NFS_OK;
2963 }
2964
2965 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2966 {
2967         int status;
2968
2969         status = nfs41_check_delegation_stateid(state);
2970         if (status != NFS_OK)
2971                 return status;
2972         nfs41_delegation_recover_stateid(state);
2973
2974         status = nfs41_check_expired_locks(state);
2975         if (status != NFS_OK)
2976                 return status;
2977         status = nfs41_check_open_stateid(state);
2978         if (status != NFS_OK)
2979                 status = nfs4_open_expired(sp, state);
2980         return status;
2981 }
2982 #endif
2983
2984 /*
2985  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
2986  * fields corresponding to attributes that were used to store the verifier.
2987  * Make sure we clobber those fields in the later setattr call
2988  */
2989 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
2990                                 struct iattr *sattr, struct nfs4_label **label)
2991 {
2992         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
2993         __u32 attrset[3];
2994         unsigned ret;
2995         unsigned i;
2996
2997         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
2998                 attrset[i] = opendata->o_res.attrset[i];
2999                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3000                         attrset[i] &= ~bitmask[i];
3001         }
3002
3003         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3004                 sattr->ia_valid : 0;
3005
3006         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3007                 if (sattr->ia_valid & ATTR_ATIME_SET)
3008                         ret |= ATTR_ATIME_SET;
3009                 else
3010                         ret |= ATTR_ATIME;
3011         }
3012
3013         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3014                 if (sattr->ia_valid & ATTR_MTIME_SET)
3015                         ret |= ATTR_MTIME_SET;
3016                 else
3017                         ret |= ATTR_MTIME;
3018         }
3019
3020         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3021                 *label = NULL;
3022         return ret;
3023 }
3024
3025 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3026                 int flags, struct nfs_open_context *ctx)
3027 {
3028         struct nfs4_state_owner *sp = opendata->owner;
3029         struct nfs_server *server = sp->so_server;
3030         struct dentry *dentry;
3031         struct nfs4_state *state;
3032         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3033         struct inode *dir = d_inode(opendata->dir);
3034         unsigned long dir_verifier;
3035         unsigned int seq;
3036         int ret;
3037
3038         seq = raw_seqcount_begin(&sp->so_reclaim_seqcount);
3039         dir_verifier = nfs_save_change_attribute(dir);
3040
3041         ret = _nfs4_proc_open(opendata, ctx);
3042         if (ret != 0)
3043                 goto out;
3044
3045         state = _nfs4_opendata_to_nfs4_state(opendata);
3046         ret = PTR_ERR(state);
3047         if (IS_ERR(state))
3048                 goto out;
3049         ctx->state = state;
3050         if (server->caps & NFS_CAP_POSIX_LOCK)
3051                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3052         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3053                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3054
3055         dentry = opendata->dentry;
3056         if (d_really_is_negative(dentry)) {
3057                 struct dentry *alias;
3058                 d_drop(dentry);
3059                 alias = d_exact_alias(dentry, state->inode);
3060                 if (!alias)
3061                         alias = d_splice_alias(igrab(state->inode), dentry);
3062                 /* d_splice_alias() can't fail here - it's a non-directory */
3063                 if (alias) {
3064                         dput(ctx->dentry);
3065                         ctx->dentry = dentry = alias;
3066                 }
3067         }
3068
3069         switch(opendata->o_arg.claim) {
3070         default:
3071                 break;
3072         case NFS4_OPEN_CLAIM_NULL:
3073         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3074         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3075                 if (!opendata->rpc_done)
3076                         break;
3077                 if (opendata->o_res.delegation_type != 0)
3078                         dir_verifier = nfs_save_change_attribute(dir);
3079                 nfs_set_verifier(dentry, dir_verifier);
3080         }
3081
3082         /* Parse layoutget results before we check for access */
3083         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3084
3085         ret = nfs4_opendata_access(sp->so_cred, opendata, state,
3086                         acc_mode, flags);
3087         if (ret != 0)
3088                 goto out;
3089
3090         if (d_inode(dentry) == state->inode) {
3091                 nfs_inode_attach_open_context(ctx);
3092                 if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq))
3093                         nfs4_schedule_stateid_recovery(server, state);
3094         }
3095
3096 out:
3097         if (!opendata->cancelled)
3098                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3099         return ret;
3100 }
3101
3102 /*
3103  * Returns a referenced nfs4_state
3104  */
3105 static int _nfs4_do_open(struct inode *dir,
3106                         struct nfs_open_context *ctx,
3107                         int flags,
3108                         const struct nfs4_open_createattrs *c,
3109                         int *opened)
3110 {
3111         struct nfs4_state_owner  *sp;
3112         struct nfs4_state     *state = NULL;
3113         struct nfs_server       *server = NFS_SERVER(dir);
3114         struct nfs4_opendata *opendata;
3115         struct dentry *dentry = ctx->dentry;
3116         const struct cred *cred = ctx->cred;
3117         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3118         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3119         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3120         struct iattr *sattr = c->sattr;
3121         struct nfs4_label *label = c->label;
3122         int status;
3123
3124         /* Protect against reboot recovery conflicts */
3125         status = -ENOMEM;
3126         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3127         if (sp == NULL) {
3128                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3129                 goto out_err;
3130         }
3131         status = nfs4_client_recover_expired_lease(server->nfs_client);
3132         if (status != 0)
3133                 goto err_put_state_owner;
3134         if (d_really_is_positive(dentry))
3135                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3136         status = -ENOMEM;
3137         if (d_really_is_positive(dentry))
3138                 claim = NFS4_OPEN_CLAIM_FH;
3139         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3140                         c, claim, GFP_KERNEL);
3141         if (opendata == NULL)
3142                 goto err_put_state_owner;
3143
3144         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3145                 if (!opendata->f_attr.mdsthreshold) {
3146                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3147                         if (!opendata->f_attr.mdsthreshold)
3148                                 goto err_opendata_put;
3149                 }
3150                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3151         }
3152         if (d_really_is_positive(dentry))
3153                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3154
3155         status = _nfs4_open_and_get_state(opendata, flags, ctx);
3156         if (status != 0)
3157                 goto err_opendata_put;
3158         state = ctx->state;
3159
3160         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3161             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3162                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3163                 /*
3164                  * send create attributes which was not set by open
3165                  * with an extra setattr.
3166                  */
3167                 if (attrs || label) {
3168                         unsigned ia_old = sattr->ia_valid;
3169
3170                         sattr->ia_valid = attrs;
3171                         nfs_fattr_init(opendata->o_res.f_attr);
3172                         status = nfs4_do_setattr(state->inode, cred,
3173                                         opendata->o_res.f_attr, sattr,
3174                                         ctx, label);
3175                         if (status == 0) {
3176                                 nfs_setattr_update_inode(state->inode, sattr,
3177                                                 opendata->o_res.f_attr);
3178                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3179                         }
3180                         sattr->ia_valid = ia_old;
3181                 }
3182         }
3183         if (opened && opendata->file_created)
3184                 *opened = 1;
3185
3186         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3187                 *ctx_th = opendata->f_attr.mdsthreshold;
3188                 opendata->f_attr.mdsthreshold = NULL;
3189         }
3190
3191         nfs4_opendata_put(opendata);
3192         nfs4_put_state_owner(sp);
3193         return 0;
3194 err_opendata_put:
3195         nfs4_opendata_put(opendata);
3196 err_put_state_owner:
3197         nfs4_put_state_owner(sp);
3198 out_err:
3199         return status;
3200 }
3201
3202
3203 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3204                                         struct nfs_open_context *ctx,
3205                                         int flags,
3206                                         struct iattr *sattr,
3207                                         struct nfs4_label *label,
3208                                         int *opened)
3209 {
3210         struct nfs_server *server = NFS_SERVER(dir);
3211         struct nfs4_exception exception = {
3212                 .interruptible = true,
3213         };
3214         struct nfs4_state *res;
3215         struct nfs4_open_createattrs c = {
3216                 .label = label,
3217                 .sattr = sattr,
3218                 .verf = {
3219                         [0] = (__u32)jiffies,
3220                         [1] = (__u32)current->pid,
3221                 },
3222         };
3223         int status;
3224
3225         do {
3226                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3227                 res = ctx->state;
3228                 trace_nfs4_open_file(ctx, flags, status);
3229                 if (status == 0)
3230                         break;
3231                 /* NOTE: BAD_SEQID means the server and client disagree about the
3232                  * book-keeping w.r.t. state-changing operations
3233                  * (OPEN/CLOSE/LOCK/LOCKU...)
3234                  * It is actually a sign of a bug on the client or on the server.
3235                  *
3236                  * If we receive a BAD_SEQID error in the particular case of
3237                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3238                  * have unhashed the old state_owner for us, and that we can
3239                  * therefore safely retry using a new one. We should still warn
3240                  * the user though...
3241                  */
3242                 if (status == -NFS4ERR_BAD_SEQID) {
3243                         pr_warn_ratelimited("NFS: v4 server %s "
3244                                         " returned a bad sequence-id error!\n",
3245                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3246                         exception.retry = 1;
3247                         continue;
3248                 }
3249                 /*
3250                  * BAD_STATEID on OPEN means that the server cancelled our
3251                  * state before it received the OPEN_CONFIRM.
3252                  * Recover by retrying the request as per the discussion
3253                  * on Page 181 of RFC3530.
3254                  */
3255                 if (status == -NFS4ERR_BAD_STATEID) {
3256                         exception.retry = 1;
3257                         continue;
3258                 }
3259                 if (status == -NFS4ERR_EXPIRED) {
3260                         nfs4_schedule_lease_recovery(server->nfs_client);
3261                         exception.retry = 1;
3262                         continue;
3263                 }
3264                 if (status == -EAGAIN) {
3265                         /* We must have found a delegation */
3266                         exception.retry = 1;
3267                         continue;
3268                 }
3269                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3270                         continue;
3271                 res = ERR_PTR(nfs4_handle_exception(server,
3272                                         status, &exception));
3273         } while (exception.retry);
3274         return res;
3275 }
3276
3277 static int _nfs4_do_setattr(struct inode *inode,
3278                             struct nfs_setattrargs *arg,
3279                             struct nfs_setattrres *res,
3280                             const struct cred *cred,
3281                             struct nfs_open_context *ctx)
3282 {
3283         struct nfs_server *server = NFS_SERVER(inode);
3284         struct rpc_message msg = {
3285                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3286                 .rpc_argp       = arg,
3287                 .rpc_resp       = res,
3288                 .rpc_cred       = cred,
3289         };
3290         const struct cred *delegation_cred = NULL;
3291         unsigned long timestamp = jiffies;
3292         bool truncate;
3293         int status;
3294
3295         nfs_fattr_init(res->fattr);
3296
3297         /* Servers should only apply open mode checks for file size changes */
3298         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3299         if (!truncate) {
3300                 nfs4_inode_make_writeable(inode);
3301                 goto zero_stateid;
3302         }
3303
3304         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3305                 /* Use that stateid */
3306         } else if (ctx != NULL && ctx->state) {
3307                 struct nfs_lock_context *l_ctx;
3308                 if (!nfs4_valid_open_stateid(ctx->state))
3309                         return -EBADF;
3310                 l_ctx = nfs_get_lock_context(ctx);
3311                 if (IS_ERR(l_ctx))
3312                         return PTR_ERR(l_ctx);
3313                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3314                                                 &arg->stateid, &delegation_cred);
3315                 nfs_put_lock_context(l_ctx);
3316                 if (status == -EIO)
3317                         return -EBADF;
3318                 else if (status == -EAGAIN)
3319                         goto zero_stateid;
3320         } else {
3321 zero_stateid:
3322                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3323         }
3324         if (delegation_cred)
3325                 msg.rpc_cred = delegation_cred;
3326
3327         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3328
3329         put_cred(delegation_cred);
3330         if (status == 0 && ctx != NULL)
3331                 renew_lease(server, timestamp);
3332         trace_nfs4_setattr(inode, &arg->stateid, status);
3333         return status;
3334 }
3335
3336 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3337                            struct nfs_fattr *fattr, struct iattr *sattr,
3338                            struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3339 {
3340         struct nfs_server *server = NFS_SERVER(inode);
3341         __u32 bitmask[NFS4_BITMASK_SZ];
3342         struct nfs4_state *state = ctx ? ctx->state : NULL;
3343         struct nfs_setattrargs  arg = {
3344                 .fh             = NFS_FH(inode),
3345                 .iap            = sattr,
3346                 .server         = server,
3347                 .bitmask = bitmask,
3348                 .label          = ilabel,
3349         };
3350         struct nfs_setattrres  res = {
3351                 .fattr          = fattr,
3352                 .server         = server,
3353         };
3354         struct nfs4_exception exception = {
3355                 .state = state,
3356                 .inode = inode,
3357                 .stateid = &arg.stateid,
3358         };
3359         unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3360         int err;
3361
3362         if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3363                 adjust_flags |= NFS_INO_INVALID_MODE;
3364         if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3365                 adjust_flags |= NFS_INO_INVALID_OTHER;
3366
3367         do {
3368                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3369                                         inode, adjust_flags);
3370
3371                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3372                 switch (err) {
3373                 case -NFS4ERR_OPENMODE:
3374                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3375                                 pr_warn_once("NFSv4: server %s is incorrectly "
3376                                                 "applying open mode checks to "
3377                                                 "a SETATTR that is not "
3378                                                 "changing file size.\n",
3379                                                 server->nfs_client->cl_hostname);
3380                         }
3381                         if (state && !(state->state & FMODE_WRITE)) {
3382                                 err = -EBADF;
3383                                 if (sattr->ia_valid & ATTR_OPEN)
3384                                         err = -EACCES;
3385                                 goto out;
3386                         }
3387                 }
3388                 err = nfs4_handle_exception(server, err, &exception);
3389         } while (exception.retry);
3390 out:
3391         return err;
3392 }
3393
3394 static bool
3395 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3396 {
3397         if (inode == NULL || !nfs_have_layout(inode))
3398                 return false;
3399
3400         return pnfs_wait_on_layoutreturn(inode, task);
3401 }
3402
3403 /*
3404  * Update the seqid of an open stateid
3405  */
3406 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3407                 struct nfs4_state *state)
3408 {
3409         __be32 seqid_open;
3410         u32 dst_seqid;
3411         int seq;
3412
3413         for (;;) {
3414                 if (!nfs4_valid_open_stateid(state))
3415                         break;
3416                 seq = read_seqbegin(&state->seqlock);
3417                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3418                         nfs4_stateid_copy(dst, &state->open_stateid);
3419                         if (read_seqretry(&state->seqlock, seq))
3420                                 continue;
3421                         break;
3422                 }
3423                 seqid_open = state->open_stateid.seqid;
3424                 if (read_seqretry(&state->seqlock, seq))
3425                         continue;
3426
3427                 dst_seqid = be32_to_cpu(dst->seqid);
3428                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3429                         dst->seqid = seqid_open;
3430                 break;
3431         }
3432 }
3433
3434 /*
3435  * Update the seqid of an open stateid after receiving
3436  * NFS4ERR_OLD_STATEID
3437  */
3438 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3439                 struct nfs4_state *state)
3440 {
3441         __be32 seqid_open;
3442         u32 dst_seqid;
3443         bool ret;
3444         int seq, status = -EAGAIN;
3445         DEFINE_WAIT(wait);
3446
3447         for (;;) {
3448                 ret = false;
3449                 if (!nfs4_valid_open_stateid(state))
3450                         break;
3451                 seq = read_seqbegin(&state->seqlock);
3452                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3453                         if (read_seqretry(&state->seqlock, seq))
3454                                 continue;
3455                         break;
3456                 }
3457
3458                 write_seqlock(&state->seqlock);
3459                 seqid_open = state->open_stateid.seqid;
3460
3461                 dst_seqid = be32_to_cpu(dst->seqid);
3462
3463                 /* Did another OPEN bump the state's seqid?  try again: */
3464                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3465                         dst->seqid = seqid_open;
3466                         write_sequnlock(&state->seqlock);
3467                         ret = true;
3468                         break;
3469                 }
3470
3471                 /* server says we're behind but we haven't seen the update yet */
3472                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3473                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3474                 write_sequnlock(&state->seqlock);
3475                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3476
3477                 if (fatal_signal_pending(current))
3478                         status = -EINTR;
3479                 else
3480                         if (schedule_timeout(5*HZ) != 0)
3481                                 status = 0;
3482
3483                 finish_wait(&state->waitq, &wait);
3484
3485                 if (!status)
3486                         continue;
3487                 if (status == -EINTR)
3488                         break;
3489
3490                 /* we slept the whole 5 seconds, we must have lost a seqid */
3491                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3492                 ret = true;
3493                 break;
3494         }
3495
3496         return ret;
3497 }
3498
3499 struct nfs4_closedata {
3500         struct inode *inode;
3501         struct nfs4_state *state;
3502         struct nfs_closeargs arg;
3503         struct nfs_closeres res;
3504         struct {
3505                 struct nfs4_layoutreturn_args arg;
3506                 struct nfs4_layoutreturn_res res;
3507                 struct nfs4_xdr_opaque_data ld_private;
3508                 u32 roc_barrier;
3509                 bool roc;
3510         } lr;
3511         struct nfs_fattr fattr;
3512         unsigned long timestamp;
3513 };
3514
3515 static void nfs4_free_closedata(void *data)
3516 {
3517         struct nfs4_closedata *calldata = data;
3518         struct nfs4_state_owner *sp = calldata->state->owner;
3519         struct super_block *sb = calldata->state->inode->i_sb;
3520
3521         if (calldata->lr.roc)
3522                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3523                                 calldata->res.lr_ret);
3524         nfs4_put_open_state(calldata->state);
3525         nfs_free_seqid(calldata->arg.seqid);
3526         nfs4_put_state_owner(sp);
3527         nfs_sb_deactive(sb);
3528         kfree(calldata);
3529 }
3530
3531 static void nfs4_close_done(struct rpc_task *task, void *data)
3532 {
3533         struct nfs4_closedata *calldata = data;
3534         struct nfs4_state *state = calldata->state;
3535         struct nfs_server *server = NFS_SERVER(calldata->inode);
3536         nfs4_stateid *res_stateid = NULL;
3537         struct nfs4_exception exception = {
3538                 .state = state,
3539                 .inode = calldata->inode,
3540                 .stateid = &calldata->arg.stateid,
3541         };
3542
3543         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3544                 return;
3545         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3546
3547         /* Handle Layoutreturn errors */
3548         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3549                           &calldata->res.lr_ret) == -EAGAIN)
3550                 goto out_restart;
3551
3552         /* hmm. we are done with the inode, and in the process of freeing
3553          * the state_owner. we keep this around to process errors
3554          */
3555         switch (task->tk_status) {
3556                 case 0:
3557                         res_stateid = &calldata->res.stateid;
3558                         renew_lease(server, calldata->timestamp);
3559                         break;
3560                 case -NFS4ERR_ACCESS:
3561                         if (calldata->arg.bitmask != NULL) {
3562                                 calldata->arg.bitmask = NULL;
3563                                 calldata->res.fattr = NULL;
3564                                 goto out_restart;
3565
3566                         }
3567                         break;
3568                 case -NFS4ERR_OLD_STATEID:
3569                         /* Did we race with OPEN? */
3570                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3571                                                 state))
3572                                 goto out_restart;
3573                         goto out_release;
3574                 case -NFS4ERR_ADMIN_REVOKED:
3575                 case -NFS4ERR_STALE_STATEID:
3576                 case -NFS4ERR_EXPIRED:
3577                         nfs4_free_revoked_stateid(server,
3578                                         &calldata->arg.stateid,
3579                                         task->tk_msg.rpc_cred);
3580                         fallthrough;
3581                 case -NFS4ERR_BAD_STATEID:
3582                         if (calldata->arg.fmode == 0)
3583                                 break;
3584                         fallthrough;
3585                 default:
3586                         task->tk_status = nfs4_async_handle_exception(task,
3587                                         server, task->tk_status, &exception);
3588                         if (exception.retry)
3589                                 goto out_restart;
3590         }
3591         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3592                         res_stateid, calldata->arg.fmode);
3593 out_release:
3594         task->tk_status = 0;
3595         nfs_release_seqid(calldata->arg.seqid);
3596         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3597         dprintk("%s: ret = %d\n", __func__, task->tk_status);
3598         return;
3599 out_restart:
3600         task->tk_status = 0;
3601         rpc_restart_call_prepare(task);
3602         goto out_release;
3603 }
3604
3605 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3606 {
3607         struct nfs4_closedata *calldata = data;
3608         struct nfs4_state *state = calldata->state;
3609         struct inode *inode = calldata->inode;
3610         struct nfs_server *server = NFS_SERVER(inode);
3611         struct pnfs_layout_hdr *lo;
3612         bool is_rdonly, is_wronly, is_rdwr;
3613         int call_close = 0;
3614
3615         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3616                 goto out_wait;
3617
3618         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3619         spin_lock(&state->owner->so_lock);
3620         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3621         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3622         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3623         /* Calculate the change in open mode */
3624         calldata->arg.fmode = 0;
3625         if (state->n_rdwr == 0) {
3626                 if (state->n_rdonly == 0)
3627                         call_close |= is_rdonly;
3628                 else if (is_rdonly)
3629                         calldata->arg.fmode |= FMODE_READ;
3630                 if (state->n_wronly == 0)
3631                         call_close |= is_wronly;
3632                 else if (is_wronly)
3633                         calldata->arg.fmode |= FMODE_WRITE;
3634                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3635                         call_close |= is_rdwr;
3636         } else if (is_rdwr)
3637                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3638
3639         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3640         if (!nfs4_valid_open_stateid(state))
3641                 call_close = 0;
3642         spin_unlock(&state->owner->so_lock);
3643
3644         if (!call_close) {
3645                 /* Note: exit _without_ calling nfs4_close_done */
3646                 goto out_no_action;
3647         }
3648
3649         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3650                 nfs_release_seqid(calldata->arg.seqid);
3651                 goto out_wait;
3652         }
3653
3654         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3655         if (lo && !pnfs_layout_is_valid(lo)) {
3656                 calldata->arg.lr_args = NULL;
3657                 calldata->res.lr_res = NULL;
3658         }
3659
3660         if (calldata->arg.fmode == 0)
3661                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3662
3663         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3664                 /* Close-to-open cache consistency revalidation */
3665                 if (!nfs4_have_delegation(inode, FMODE_READ)) {
3666                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3667                                          server->cache_consistency_bitmask,
3668                                          inode, 0);
3669                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3670                 } else
3671                         calldata->arg.bitmask = NULL;
3672         }
3673
3674         calldata->arg.share_access =
3675                 nfs4_map_atomic_open_share(NFS_SERVER(inode),
3676                                 calldata->arg.fmode, 0);
3677
3678         if (calldata->res.fattr == NULL)
3679                 calldata->arg.bitmask = NULL;
3680         else if (calldata->arg.bitmask == NULL)
3681                 calldata->res.fattr = NULL;
3682         calldata->timestamp = jiffies;
3683         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3684                                 &calldata->arg.seq_args,
3685                                 &calldata->res.seq_res,
3686                                 task) != 0)
3687                 nfs_release_seqid(calldata->arg.seqid);
3688         return;
3689 out_no_action:
3690         task->tk_action = NULL;
3691 out_wait:
3692         nfs4_sequence_done(task, &calldata->res.seq_res);
3693 }
3694
3695 static const struct rpc_call_ops nfs4_close_ops = {
3696         .rpc_call_prepare = nfs4_close_prepare,
3697         .rpc_call_done = nfs4_close_done,
3698         .rpc_release = nfs4_free_closedata,
3699 };
3700
3701 /*
3702  * It is possible for data to be read/written from a mem-mapped file
3703  * after the sys_close call (which hits the vfs layer as a flush).
3704  * This means that we can't safely call nfsv4 close on a file until
3705  * the inode is cleared. This in turn means that we are not good
3706  * NFSv4 citizens - we do not indicate to the server to update the file's
3707  * share state even when we are done with one of the three share
3708  * stateid's in the inode.
3709  *
3710  * NOTE: Caller must be holding the sp->so_owner semaphore!
3711  */
3712 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3713 {
3714         struct nfs_server *server = NFS_SERVER(state->inode);
3715         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3716         struct nfs4_closedata *calldata;
3717         struct nfs4_state_owner *sp = state->owner;
3718         struct rpc_task *task;
3719         struct rpc_message msg = {
3720                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3721                 .rpc_cred = state->owner->so_cred,
3722         };
3723         struct rpc_task_setup task_setup_data = {
3724                 .rpc_client = server->client,
3725                 .rpc_message = &msg,
3726                 .callback_ops = &nfs4_close_ops,
3727                 .workqueue = nfsiod_workqueue,
3728                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3729         };
3730         int status = -ENOMEM;
3731
3732         if (server->nfs_client->cl_minorversion)
3733                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
3734
3735         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3736                 &task_setup_data.rpc_client, &msg);
3737
3738         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3739         if (calldata == NULL)
3740                 goto out;
3741         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3742         calldata->inode = state->inode;
3743         calldata->state = state;
3744         calldata->arg.fh = NFS_FH(state->inode);
3745         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3746                 goto out_free_calldata;
3747         /* Serialization for the sequence id */
3748         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3749         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3750         if (IS_ERR(calldata->arg.seqid))
3751                 goto out_free_calldata;
3752         nfs_fattr_init(&calldata->fattr);
3753         calldata->arg.fmode = 0;
3754         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3755         calldata->res.fattr = &calldata->fattr;
3756         calldata->res.seqid = calldata->arg.seqid;
3757         calldata->res.server = server;
3758         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3759         calldata->lr.roc = pnfs_roc(state->inode,
3760                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3761         if (calldata->lr.roc) {
3762                 calldata->arg.lr_args = &calldata->lr.arg;
3763                 calldata->res.lr_res = &calldata->lr.res;
3764         }
3765         nfs_sb_active(calldata->inode->i_sb);
3766
3767         msg.rpc_argp = &calldata->arg;
3768         msg.rpc_resp = &calldata->res;
3769         task_setup_data.callback_data = calldata;
3770         task = rpc_run_task(&task_setup_data);
3771         if (IS_ERR(task))
3772                 return PTR_ERR(task);
3773         status = 0;
3774         if (wait)
3775                 status = rpc_wait_for_completion_task(task);
3776         rpc_put_task(task);
3777         return status;
3778 out_free_calldata:
3779         kfree(calldata);
3780 out:
3781         nfs4_put_open_state(state);
3782         nfs4_put_state_owner(sp);
3783         return status;
3784 }
3785
3786 static struct inode *
3787 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3788                 int open_flags, struct iattr *attr, int *opened)
3789 {
3790         struct nfs4_state *state;
3791         struct nfs4_label l = {0, 0, 0, NULL}, *label = NULL;
3792
3793         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3794
3795         /* Protect against concurrent sillydeletes */
3796         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3797
3798         nfs4_label_release_security(label);
3799
3800         if (IS_ERR(state))
3801                 return ERR_CAST(state);
3802         return state->inode;
3803 }
3804
3805 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3806 {
3807         if (ctx->state == NULL)
3808                 return;
3809         if (is_sync)
3810                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3811         else
3812                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3813 }
3814
3815 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3816 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3817 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3818
3819 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3820 {
3821         u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3822         struct nfs4_server_caps_arg args = {
3823                 .fhandle = fhandle,
3824                 .bitmask = bitmask,
3825         };
3826         struct nfs4_server_caps_res res = {};
3827         struct rpc_message msg = {
3828                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3829                 .rpc_argp = &args,
3830                 .rpc_resp = &res,
3831         };
3832         int status;
3833         int i;
3834
3835         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3836                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3837                      FATTR4_WORD0_LINK_SUPPORT |
3838                      FATTR4_WORD0_SYMLINK_SUPPORT |
3839                      FATTR4_WORD0_ACLSUPPORT |
3840                      FATTR4_WORD0_CASE_INSENSITIVE |
3841                      FATTR4_WORD0_CASE_PRESERVING;
3842         if (minorversion)
3843                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3844
3845         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3846         if (status == 0) {
3847                 /* Sanity check the server answers */
3848                 switch (minorversion) {
3849                 case 0:
3850                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3851                         res.attr_bitmask[2] = 0;
3852                         break;
3853                 case 1:
3854                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3855                         break;
3856                 case 2:
3857                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3858                 }
3859                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3860                 server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3861                                   NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3862                 server->fattr_valid = NFS_ATTR_FATTR_V4;
3863                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3864                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3865                         server->caps |= NFS_CAP_ACLS;
3866                 if (res.has_links != 0)
3867                         server->caps |= NFS_CAP_HARDLINKS;
3868                 if (res.has_symlinks != 0)
3869                         server->caps |= NFS_CAP_SYMLINKS;
3870                 if (res.case_insensitive)
3871                         server->caps |= NFS_CAP_CASE_INSENSITIVE;
3872                 if (res.case_preserving)
3873                         server->caps |= NFS_CAP_CASE_PRESERVING;
3874 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3875                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3876                         server->caps |= NFS_CAP_SECURITY_LABEL;
3877 #endif
3878                 if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3879                         server->caps |= NFS_CAP_FS_LOCATIONS;
3880                 if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3881                         server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3882                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3883                         server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3884                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3885                         server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3886                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3887                         server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3888                                 NFS_ATTR_FATTR_OWNER_NAME);
3889                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3890                         server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3891                                 NFS_ATTR_FATTR_GROUP_NAME);
3892                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3893                         server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3894                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3895                         server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3896                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3897                         server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3898                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3899                         server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3900                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3901                                 sizeof(server->attr_bitmask));
3902                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3903
3904                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3905                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3906                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3907                 server->cache_consistency_bitmask[2] = 0;
3908
3909                 /* Avoid a regression due to buggy server */
3910                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3911                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3912                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3913                         sizeof(server->exclcreat_bitmask));
3914
3915                 server->acl_bitmask = res.acl_bitmask;
3916                 server->fh_expire_type = res.fh_expire_type;
3917         }
3918
3919         return status;
3920 }
3921
3922 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3923 {
3924         struct nfs4_exception exception = {
3925                 .interruptible = true,
3926         };
3927         int err;
3928
3929         nfs4_server_set_init_caps(server);
3930         do {
3931                 err = nfs4_handle_exception(server,
3932                                 _nfs4_server_capabilities(server, fhandle),
3933                                 &exception);
3934         } while (exception.retry);
3935         return err;
3936 }
3937
3938 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3939                                           struct nfs_client *clp,
3940                                           struct nfs_server *server)
3941 {
3942         int i;
3943
3944         for (i = 0; i < location->nservers; i++) {
3945                 struct nfs4_string *srv_loc = &location->servers[i];
3946                 struct sockaddr addr;
3947                 size_t addrlen;
3948                 struct xprt_create xprt_args = {
3949                         .ident = 0,
3950                         .net = clp->cl_net,
3951                 };
3952                 struct nfs4_add_xprt_data xprtdata = {
3953                         .clp = clp,
3954                 };
3955                 struct rpc_add_xprt_test rpcdata = {
3956                         .add_xprt_test = clp->cl_mvops->session_trunk,
3957                         .data = &xprtdata,
3958                 };
3959                 char *servername = NULL;
3960
3961                 if (!srv_loc->len)
3962                         continue;
3963
3964                 addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
3965                                                 &addr, sizeof(addr),
3966                                                 clp->cl_net, server->port);
3967                 if (!addrlen)
3968                         return;
3969                 xprt_args.dstaddr = &addr;
3970                 xprt_args.addrlen = addrlen;
3971                 servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
3972                 if (!servername)
3973                         return;
3974                 memcpy(servername, srv_loc->data, srv_loc->len);
3975                 servername[srv_loc->len] = '\0';
3976                 xprt_args.servername = servername;
3977
3978                 xprtdata.cred = nfs4_get_clid_cred(clp);
3979                 rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
3980                                   rpc_clnt_setup_test_and_add_xprt,
3981                                   &rpcdata);
3982                 if (xprtdata.cred)
3983                         put_cred(xprtdata.cred);
3984                 kfree(servername);
3985         }
3986 }
3987
3988 static int _nfs4_discover_trunking(struct nfs_server *server,
3989                                    struct nfs_fh *fhandle)
3990 {
3991         struct nfs4_fs_locations *locations = NULL;
3992         struct page *page;
3993         const struct cred *cred;
3994         struct nfs_client *clp = server->nfs_client;
3995         const struct nfs4_state_maintenance_ops *ops =
3996                 clp->cl_mvops->state_renewal_ops;
3997         int status = -ENOMEM, i;
3998
3999         cred = ops->get_state_renewal_cred(clp);
4000         if (cred == NULL) {
4001                 cred = nfs4_get_clid_cred(clp);
4002                 if (cred == NULL)
4003                         return -ENOKEY;
4004         }
4005
4006         page = alloc_page(GFP_KERNEL);
4007         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4008         if (page == NULL || locations == NULL)
4009                 goto out;
4010
4011         status = nfs4_proc_get_locations(server, fhandle, locations, page,
4012                                          cred);
4013         if (status)
4014                 goto out;
4015
4016         for (i = 0; i < locations->nlocations; i++)
4017                 test_fs_location_for_trunking(&locations->locations[i], clp,
4018                                               server);
4019 out:
4020         if (page)
4021                 __free_page(page);
4022         kfree(locations);
4023         return status;
4024 }
4025
4026 static int nfs4_discover_trunking(struct nfs_server *server,
4027                                   struct nfs_fh *fhandle)
4028 {
4029         struct nfs4_exception exception = {
4030                 .interruptible = true,
4031         };
4032         struct nfs_client *clp = server->nfs_client;
4033         int err = 0;
4034
4035         if (!nfs4_has_session(clp))
4036                 goto out;
4037         do {
4038                 err = nfs4_handle_exception(server,
4039                                 _nfs4_discover_trunking(server, fhandle),
4040                                 &exception);
4041         } while (exception.retry);
4042 out:
4043         return err;
4044 }
4045
4046 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4047                 struct nfs_fsinfo *info)
4048 {
4049         u32 bitmask[3];
4050         struct nfs4_lookup_root_arg args = {
4051                 .bitmask = bitmask,
4052         };
4053         struct nfs4_lookup_res res = {
4054                 .server = server,
4055                 .fattr = info->fattr,
4056                 .fh = fhandle,
4057         };
4058         struct rpc_message msg = {
4059                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4060                 .rpc_argp = &args,
4061                 .rpc_resp = &res,
4062         };
4063
4064         bitmask[0] = nfs4_fattr_bitmap[0];
4065         bitmask[1] = nfs4_fattr_bitmap[1];
4066         /*
4067          * Process the label in the upcoming getfattr
4068          */
4069         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4070
4071         nfs_fattr_init(info->fattr);
4072         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4073 }
4074
4075 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4076                 struct nfs_fsinfo *info)
4077 {
4078         struct nfs4_exception exception = {
4079                 .interruptible = true,
4080         };
4081         int err;
4082         do {
4083                 err = _nfs4_lookup_root(server, fhandle, info);
4084                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4085                 switch (err) {
4086                 case 0:
4087                 case -NFS4ERR_WRONGSEC:
4088                         goto out;
4089                 default:
4090                         err = nfs4_handle_exception(server, err, &exception);
4091                 }
4092         } while (exception.retry);
4093 out:
4094         return err;
4095 }
4096
4097 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4098                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4099 {
4100         struct rpc_auth_create_args auth_args = {
4101                 .pseudoflavor = flavor,
4102         };
4103         struct rpc_auth *auth;
4104
4105         auth = rpcauth_create(&auth_args, server->client);
4106         if (IS_ERR(auth))
4107                 return -EACCES;
4108         return nfs4_lookup_root(server, fhandle, info);
4109 }
4110
4111 /*
4112  * Retry pseudoroot lookup with various security flavors.  We do this when:
4113  *
4114  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4115  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4116  *
4117  * Returns zero on success, or a negative NFS4ERR value, or a
4118  * negative errno value.
4119  */
4120 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4121                               struct nfs_fsinfo *info)
4122 {
4123         /* Per 3530bis 15.33.5 */
4124         static const rpc_authflavor_t flav_array[] = {
4125                 RPC_AUTH_GSS_KRB5P,
4126                 RPC_AUTH_GSS_KRB5I,
4127                 RPC_AUTH_GSS_KRB5,
4128                 RPC_AUTH_UNIX,                  /* courtesy */
4129                 RPC_AUTH_NULL,
4130         };
4131         int status = -EPERM;
4132         size_t i;
4133
4134         if (server->auth_info.flavor_len > 0) {
4135                 /* try each flavor specified by user */
4136                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4137                         status = nfs4_lookup_root_sec(server, fhandle, info,
4138                                                 server->auth_info.flavors[i]);
4139                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4140                                 continue;
4141                         break;
4142                 }
4143         } else {
4144                 /* no flavors specified by user, try default list */
4145                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4146                         status = nfs4_lookup_root_sec(server, fhandle, info,
4147                                                       flav_array[i]);
4148                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4149                                 continue;
4150                         break;
4151                 }
4152         }
4153
4154         /*
4155          * -EACCES could mean that the user doesn't have correct permissions
4156          * to access the mount.  It could also mean that we tried to mount
4157          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4158          * existing mount programs don't handle -EACCES very well so it should
4159          * be mapped to -EPERM instead.
4160          */
4161         if (status == -EACCES)
4162                 status = -EPERM;
4163         return status;
4164 }
4165
4166 /**
4167  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4168  * @server: initialized nfs_server handle
4169  * @fhandle: we fill in the pseudo-fs root file handle
4170  * @info: we fill in an FSINFO struct
4171  * @auth_probe: probe the auth flavours
4172  *
4173  * Returns zero on success, or a negative errno.
4174  */
4175 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4176                          struct nfs_fsinfo *info,
4177                          bool auth_probe)
4178 {
4179         int status = 0;
4180
4181         if (!auth_probe)
4182                 status = nfs4_lookup_root(server, fhandle, info);
4183
4184         if (auth_probe || status == NFS4ERR_WRONGSEC)
4185                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4186                                 fhandle, info);
4187
4188         if (status == 0)
4189                 status = nfs4_server_capabilities(server, fhandle);
4190         if (status == 0)
4191                 status = nfs4_do_fsinfo(server, fhandle, info);
4192
4193         return nfs4_map_errors(status);
4194 }
4195
4196 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4197                               struct nfs_fsinfo *info)
4198 {
4199         int error;
4200         struct nfs_fattr *fattr = info->fattr;
4201
4202         error = nfs4_server_capabilities(server, mntfh);
4203         if (error < 0) {
4204                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4205                 return error;
4206         }
4207
4208         error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4209         if (error < 0) {
4210                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4211                 goto out;
4212         }
4213
4214         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4215             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4216                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4217
4218 out:
4219         return error;
4220 }
4221
4222 /*
4223  * Get locations and (maybe) other attributes of a referral.
4224  * Note that we'll actually follow the referral later when
4225  * we detect fsid mismatch in inode revalidation
4226  */
4227 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4228                              const struct qstr *name, struct nfs_fattr *fattr,
4229                              struct nfs_fh *fhandle)
4230 {
4231         int status = -ENOMEM;
4232         struct page *page = NULL;
4233         struct nfs4_fs_locations *locations = NULL;
4234
4235         page = alloc_page(GFP_KERNEL);
4236         if (page == NULL)
4237                 goto out;
4238         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4239         if (locations == NULL)
4240                 goto out;
4241
4242         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4243         if (status != 0)
4244                 goto out;
4245
4246         /*
4247          * If the fsid didn't change, this is a migration event, not a
4248          * referral.  Cause us to drop into the exception handler, which
4249          * will kick off migration recovery.
4250          */
4251         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &locations->fattr.fsid)) {
4252                 dprintk("%s: server did not return a different fsid for"
4253                         " a referral at %s\n", __func__, name->name);
4254                 status = -NFS4ERR_MOVED;
4255                 goto out;
4256         }
4257         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4258         nfs_fixup_referral_attributes(&locations->fattr);
4259
4260         /* replace the lookup nfs_fattr with the locations nfs_fattr */
4261         memcpy(fattr, &locations->fattr, sizeof(struct nfs_fattr));
4262         memset(fhandle, 0, sizeof(struct nfs_fh));
4263 out:
4264         if (page)
4265                 __free_page(page);
4266         kfree(locations);
4267         return status;
4268 }
4269
4270 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4271                                 struct nfs_fattr *fattr, struct inode *inode)
4272 {
4273         __u32 bitmask[NFS4_BITMASK_SZ];
4274         struct nfs4_getattr_arg args = {
4275                 .fh = fhandle,
4276                 .bitmask = bitmask,
4277         };
4278         struct nfs4_getattr_res res = {
4279                 .fattr = fattr,
4280                 .server = server,
4281         };
4282         struct rpc_message msg = {
4283                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4284                 .rpc_argp = &args,
4285                 .rpc_resp = &res,
4286         };
4287         unsigned short task_flags = 0;
4288
4289         if (nfs4_has_session(server->nfs_client))
4290                 task_flags = RPC_TASK_MOVEABLE;
4291
4292         /* Is this is an attribute revalidation, subject to softreval? */
4293         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4294                 task_flags |= RPC_TASK_TIMEOUT;
4295
4296         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4297         nfs_fattr_init(fattr);
4298         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4299         return nfs4_do_call_sync(server->client, server, &msg,
4300                         &args.seq_args, &res.seq_res, task_flags);
4301 }
4302
4303 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4304                                 struct nfs_fattr *fattr, struct inode *inode)
4305 {
4306         struct nfs4_exception exception = {
4307                 .interruptible = true,
4308         };
4309         int err;
4310         do {
4311                 err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4312                 trace_nfs4_getattr(server, fhandle, fattr, err);
4313                 err = nfs4_handle_exception(server, err,
4314                                 &exception);
4315         } while (exception.retry);
4316         return err;
4317 }
4318
4319 /*
4320  * The file is not closed if it is opened due to the a request to change
4321  * the size of the file. The open call will not be needed once the
4322  * VFS layer lookup-intents are implemented.
4323  *
4324  * Close is called when the inode is destroyed.
4325  * If we haven't opened the file for O_WRONLY, we
4326  * need to in the size_change case to obtain a stateid.
4327  *
4328  * Got race?
4329  * Because OPEN is always done by name in nfsv4, it is
4330  * possible that we opened a different file by the same
4331  * name.  We can recognize this race condition, but we
4332  * can't do anything about it besides returning an error.
4333  *
4334  * This will be fixed with VFS changes (lookup-intent).
4335  */
4336 static int
4337 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4338                   struct iattr *sattr)
4339 {
4340         struct inode *inode = d_inode(dentry);
4341         const struct cred *cred = NULL;
4342         struct nfs_open_context *ctx = NULL;
4343         int status;
4344
4345         if (pnfs_ld_layoutret_on_setattr(inode) &&
4346             sattr->ia_valid & ATTR_SIZE &&
4347             sattr->ia_size < i_size_read(inode))
4348                 pnfs_commit_and_return_layout(inode);
4349
4350         nfs_fattr_init(fattr);
4351
4352         /* Deal with open(O_TRUNC) */
4353         if (sattr->ia_valid & ATTR_OPEN)
4354                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4355
4356         /* Optimization: if the end result is no change, don't RPC */
4357         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4358                 return 0;
4359
4360         /* Search for an existing open(O_WRITE) file */
4361         if (sattr->ia_valid & ATTR_FILE) {
4362
4363                 ctx = nfs_file_open_context(sattr->ia_file);
4364                 if (ctx)
4365                         cred = ctx->cred;
4366         }
4367
4368         /* Return any delegations if we're going to change ACLs */
4369         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4370                 nfs4_inode_make_writeable(inode);
4371
4372         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4373         if (status == 0) {
4374                 nfs_setattr_update_inode(inode, sattr, fattr);
4375                 nfs_setsecurity(inode, fattr);
4376         }
4377         return status;
4378 }
4379
4380 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4381                 struct dentry *dentry, struct nfs_fh *fhandle,
4382                 struct nfs_fattr *fattr)
4383 {
4384         struct nfs_server *server = NFS_SERVER(dir);
4385         int                    status;
4386         struct nfs4_lookup_arg args = {
4387                 .bitmask = server->attr_bitmask,
4388                 .dir_fh = NFS_FH(dir),
4389                 .name = &dentry->d_name,
4390         };
4391         struct nfs4_lookup_res res = {
4392                 .server = server,
4393                 .fattr = fattr,
4394                 .fh = fhandle,
4395         };
4396         struct rpc_message msg = {
4397                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4398                 .rpc_argp = &args,
4399                 .rpc_resp = &res,
4400         };
4401         unsigned short task_flags = 0;
4402
4403         if (server->nfs_client->cl_minorversion)
4404                 task_flags = RPC_TASK_MOVEABLE;
4405
4406         /* Is this is an attribute revalidation, subject to softreval? */
4407         if (nfs_lookup_is_soft_revalidate(dentry))
4408                 task_flags |= RPC_TASK_TIMEOUT;
4409
4410         args.bitmask = nfs4_bitmask(server, fattr->label);
4411
4412         nfs_fattr_init(fattr);
4413
4414         dprintk("NFS call  lookup %pd2\n", dentry);
4415         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4416         status = nfs4_do_call_sync(clnt, server, &msg,
4417                         &args.seq_args, &res.seq_res, task_flags);
4418         dprintk("NFS reply lookup: %d\n", status);
4419         return status;
4420 }
4421
4422 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4423 {
4424         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4425                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4426         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4427         fattr->nlink = 2;
4428 }
4429
4430 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4431                                    struct dentry *dentry, struct nfs_fh *fhandle,
4432                                    struct nfs_fattr *fattr)
4433 {
4434         struct nfs4_exception exception = {
4435                 .interruptible = true,
4436         };
4437         struct rpc_clnt *client = *clnt;
4438         const struct qstr *name = &dentry->d_name;
4439         int err;
4440         do {
4441                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4442                 trace_nfs4_lookup(dir, name, err);
4443                 switch (err) {
4444                 case -NFS4ERR_BADNAME:
4445                         err = -ENOENT;
4446                         goto out;
4447                 case -NFS4ERR_MOVED:
4448                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4449                         if (err == -NFS4ERR_MOVED)
4450                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4451                         goto out;
4452                 case -NFS4ERR_WRONGSEC:
4453                         err = -EPERM;
4454                         if (client != *clnt)
4455                                 goto out;
4456                         client = nfs4_negotiate_security(client, dir, name);
4457                         if (IS_ERR(client))
4458                                 return PTR_ERR(client);
4459
4460                         exception.retry = 1;
4461                         break;
4462                 default:
4463                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4464                 }
4465         } while (exception.retry);
4466
4467 out:
4468         if (err == 0)
4469                 *clnt = client;
4470         else if (client != *clnt)
4471                 rpc_shutdown_client(client);
4472
4473         return err;
4474 }
4475
4476 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4477                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4478 {
4479         int status;
4480         struct rpc_clnt *client = NFS_CLIENT(dir);
4481
4482         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4483         if (client != NFS_CLIENT(dir)) {
4484                 rpc_shutdown_client(client);
4485                 nfs_fixup_secinfo_attributes(fattr);
4486         }
4487         return status;
4488 }
4489
4490 struct rpc_clnt *
4491 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4492                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4493 {
4494         struct rpc_clnt *client = NFS_CLIENT(dir);
4495         int status;
4496
4497         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4498         if (status < 0)
4499                 return ERR_PTR(status);
4500         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4501 }
4502
4503 static int _nfs4_proc_lookupp(struct inode *inode,
4504                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4505 {
4506         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4507         struct nfs_server *server = NFS_SERVER(inode);
4508         int                    status;
4509         struct nfs4_lookupp_arg args = {
4510                 .bitmask = server->attr_bitmask,
4511                 .fh = NFS_FH(inode),
4512         };
4513         struct nfs4_lookupp_res res = {
4514                 .server = server,
4515                 .fattr = fattr,
4516                 .fh = fhandle,
4517         };
4518         struct rpc_message msg = {
4519                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4520                 .rpc_argp = &args,
4521                 .rpc_resp = &res,
4522         };
4523         unsigned short task_flags = 0;
4524
4525         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4526                 task_flags |= RPC_TASK_TIMEOUT;
4527
4528         args.bitmask = nfs4_bitmask(server, fattr->label);
4529
4530         nfs_fattr_init(fattr);
4531
4532         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4533         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4534                                 &res.seq_res, task_flags);
4535         dprintk("NFS reply lookupp: %d\n", status);
4536         return status;
4537 }
4538
4539 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4540                              struct nfs_fattr *fattr)
4541 {
4542         struct nfs4_exception exception = {
4543                 .interruptible = true,
4544         };
4545         int err;
4546         do {
4547                 err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4548                 trace_nfs4_lookupp(inode, err);
4549                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4550                                 &exception);
4551         } while (exception.retry);
4552         return err;
4553 }
4554
4555 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4556                              const struct cred *cred)
4557 {
4558         struct nfs_server *server = NFS_SERVER(inode);
4559         struct nfs4_accessargs args = {
4560                 .fh = NFS_FH(inode),
4561                 .access = entry->mask,
4562         };
4563         struct nfs4_accessres res = {
4564                 .server = server,
4565         };
4566         struct rpc_message msg = {
4567                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4568                 .rpc_argp = &args,
4569                 .rpc_resp = &res,
4570                 .rpc_cred = cred,
4571         };
4572         int status = 0;
4573
4574         if (!nfs4_have_delegation(inode, FMODE_READ)) {
4575                 res.fattr = nfs_alloc_fattr();
4576                 if (res.fattr == NULL)
4577                         return -ENOMEM;
4578                 args.bitmask = server->cache_consistency_bitmask;
4579         }
4580         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4581         if (!status) {
4582                 nfs_access_set_mask(entry, res.access);
4583                 if (res.fattr)
4584                         nfs_refresh_inode(inode, res.fattr);
4585         }
4586         nfs_free_fattr(res.fattr);
4587         return status;
4588 }
4589
4590 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4591                             const struct cred *cred)
4592 {
4593         struct nfs4_exception exception = {
4594                 .interruptible = true,
4595         };
4596         int err;
4597         do {
4598                 err = _nfs4_proc_access(inode, entry, cred);
4599                 trace_nfs4_access(inode, err);
4600                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4601                                 &exception);
4602         } while (exception.retry);
4603         return err;
4604 }
4605
4606 /*
4607  * TODO: For the time being, we don't try to get any attributes
4608  * along with any of the zero-copy operations READ, READDIR,
4609  * READLINK, WRITE.
4610  *
4611  * In the case of the first three, we want to put the GETATTR
4612  * after the read-type operation -- this is because it is hard
4613  * to predict the length of a GETATTR response in v4, and thus
4614  * align the READ data correctly.  This means that the GETATTR
4615  * may end up partially falling into the page cache, and we should
4616  * shift it into the 'tail' of the xdr_buf before processing.
4617  * To do this efficiently, we need to know the total length
4618  * of data received, which doesn't seem to be available outside
4619  * of the RPC layer.
4620  *
4621  * In the case of WRITE, we also want to put the GETATTR after
4622  * the operation -- in this case because we want to make sure
4623  * we get the post-operation mtime and size.
4624  *
4625  * Both of these changes to the XDR layer would in fact be quite
4626  * minor, but I decided to leave them for a subsequent patch.
4627  */
4628 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4629                 unsigned int pgbase, unsigned int pglen)
4630 {
4631         struct nfs4_readlink args = {
4632                 .fh       = NFS_FH(inode),
4633                 .pgbase   = pgbase,
4634                 .pglen    = pglen,
4635                 .pages    = &page,
4636         };
4637         struct nfs4_readlink_res res;
4638         struct rpc_message msg = {
4639                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4640                 .rpc_argp = &args,
4641                 .rpc_resp = &res,
4642         };
4643
4644         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4645 }
4646
4647 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4648                 unsigned int pgbase, unsigned int pglen)
4649 {
4650         struct nfs4_exception exception = {
4651                 .interruptible = true,
4652         };
4653         int err;
4654         do {
4655                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4656                 trace_nfs4_readlink(inode, err);
4657                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4658                                 &exception);
4659         } while (exception.retry);
4660         return err;
4661 }
4662
4663 /*
4664  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4665  */
4666 static int
4667 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4668                  int flags)
4669 {
4670         struct nfs_server *server = NFS_SERVER(dir);
4671         struct nfs4_label l, *ilabel = NULL;
4672         struct nfs_open_context *ctx;
4673         struct nfs4_state *state;
4674         int status = 0;
4675
4676         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4677         if (IS_ERR(ctx))
4678                 return PTR_ERR(ctx);
4679
4680         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4681
4682         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4683                 sattr->ia_mode &= ~current_umask();
4684         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4685         if (IS_ERR(state)) {
4686                 status = PTR_ERR(state);
4687                 goto out;
4688         }
4689 out:
4690         nfs4_label_release_security(ilabel);
4691         put_nfs_open_context(ctx);
4692         return status;
4693 }
4694
4695 static int
4696 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4697 {
4698         struct nfs_server *server = NFS_SERVER(dir);
4699         struct nfs_removeargs args = {
4700                 .fh = NFS_FH(dir),
4701                 .name = *name,
4702         };
4703         struct nfs_removeres res = {
4704                 .server = server,
4705         };
4706         struct rpc_message msg = {
4707                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4708                 .rpc_argp = &args,
4709                 .rpc_resp = &res,
4710         };
4711         unsigned long timestamp = jiffies;
4712         int status;
4713
4714         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4715         if (status == 0) {
4716                 spin_lock(&dir->i_lock);
4717                 /* Removing a directory decrements nlink in the parent */
4718                 if (ftype == NF4DIR && dir->i_nlink > 2)
4719                         nfs4_dec_nlink_locked(dir);
4720                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4721                                               NFS_INO_INVALID_DATA);
4722                 spin_unlock(&dir->i_lock);
4723         }
4724         return status;
4725 }
4726
4727 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4728 {
4729         struct nfs4_exception exception = {
4730                 .interruptible = true,
4731         };
4732         struct inode *inode = d_inode(dentry);
4733         int err;
4734
4735         if (inode) {
4736                 if (inode->i_nlink == 1)
4737                         nfs4_inode_return_delegation(inode);
4738                 else
4739                         nfs4_inode_make_writeable(inode);
4740         }
4741         do {
4742                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4743                 trace_nfs4_remove(dir, &dentry->d_name, err);
4744                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4745                                 &exception);
4746         } while (exception.retry);
4747         return err;
4748 }
4749
4750 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4751 {
4752         struct nfs4_exception exception = {
4753                 .interruptible = true,
4754         };
4755         int err;
4756
4757         do {
4758                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4759                 trace_nfs4_remove(dir, name, err);
4760                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4761                                 &exception);
4762         } while (exception.retry);
4763         return err;
4764 }
4765
4766 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4767                 struct dentry *dentry,
4768                 struct inode *inode)
4769 {
4770         struct nfs_removeargs *args = msg->rpc_argp;
4771         struct nfs_removeres *res = msg->rpc_resp;
4772
4773         res->server = NFS_SB(dentry->d_sb);
4774         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4775         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4776
4777         nfs_fattr_init(res->dir_attr);
4778
4779         if (inode) {
4780                 nfs4_inode_return_delegation(inode);
4781                 nfs_d_prune_case_insensitive_aliases(inode);
4782         }
4783 }
4784
4785 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4786 {
4787         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4788                         &data->args.seq_args,
4789                         &data->res.seq_res,
4790                         task);
4791 }
4792
4793 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4794 {
4795         struct nfs_unlinkdata *data = task->tk_calldata;
4796         struct nfs_removeres *res = &data->res;
4797
4798         if (!nfs4_sequence_done(task, &res->seq_res))
4799                 return 0;
4800         if (nfs4_async_handle_error(task, res->server, NULL,
4801                                     &data->timeout) == -EAGAIN)
4802                 return 0;
4803         if (task->tk_status == 0)
4804                 nfs4_update_changeattr(dir, &res->cinfo,
4805                                 res->dir_attr->time_start,
4806                                 NFS_INO_INVALID_DATA);
4807         return 1;
4808 }
4809
4810 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4811                 struct dentry *old_dentry,
4812                 struct dentry *new_dentry)
4813 {
4814         struct nfs_renameargs *arg = msg->rpc_argp;
4815         struct nfs_renameres *res = msg->rpc_resp;
4816         struct inode *old_inode = d_inode(old_dentry);
4817         struct inode *new_inode = d_inode(new_dentry);
4818
4819         if (old_inode)
4820                 nfs4_inode_make_writeable(old_inode);
4821         if (new_inode)
4822                 nfs4_inode_return_delegation(new_inode);
4823         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4824         res->server = NFS_SB(old_dentry->d_sb);
4825         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4826 }
4827
4828 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4829 {
4830         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4831                         &data->args.seq_args,
4832                         &data->res.seq_res,
4833                         task);
4834 }
4835
4836 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4837                                  struct inode *new_dir)
4838 {
4839         struct nfs_renamedata *data = task->tk_calldata;
4840         struct nfs_renameres *res = &data->res;
4841
4842         if (!nfs4_sequence_done(task, &res->seq_res))
4843                 return 0;
4844         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4845                 return 0;
4846
4847         if (task->tk_status == 0) {
4848                 nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4849                 if (new_dir != old_dir) {
4850                         /* Note: If we moved a directory, nlink will change */
4851                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4852                                         res->old_fattr->time_start,
4853                                         NFS_INO_INVALID_NLINK |
4854                                             NFS_INO_INVALID_DATA);
4855                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
4856                                         res->new_fattr->time_start,
4857                                         NFS_INO_INVALID_NLINK |
4858                                             NFS_INO_INVALID_DATA);
4859                 } else
4860                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
4861                                         res->old_fattr->time_start,
4862                                         NFS_INO_INVALID_DATA);
4863         }
4864         return 1;
4865 }
4866
4867 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4868 {
4869         struct nfs_server *server = NFS_SERVER(inode);
4870         __u32 bitmask[NFS4_BITMASK_SZ];
4871         struct nfs4_link_arg arg = {
4872                 .fh     = NFS_FH(inode),
4873                 .dir_fh = NFS_FH(dir),
4874                 .name   = name,
4875                 .bitmask = bitmask,
4876         };
4877         struct nfs4_link_res res = {
4878                 .server = server,
4879         };
4880         struct rpc_message msg = {
4881                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4882                 .rpc_argp = &arg,
4883                 .rpc_resp = &res,
4884         };
4885         int status = -ENOMEM;
4886
4887         res.fattr = nfs_alloc_fattr_with_label(server);
4888         if (res.fattr == NULL)
4889                 goto out;
4890
4891         nfs4_inode_make_writeable(inode);
4892         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4893                                 NFS_INO_INVALID_CHANGE);
4894         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4895         if (!status) {
4896                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4897                                        NFS_INO_INVALID_DATA);
4898                 nfs4_inc_nlink(inode);
4899                 status = nfs_post_op_update_inode(inode, res.fattr);
4900                 if (!status)
4901                         nfs_setsecurity(inode, res.fattr);
4902         }
4903
4904 out:
4905         nfs_free_fattr(res.fattr);
4906         return status;
4907 }
4908
4909 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4910 {
4911         struct nfs4_exception exception = {
4912                 .interruptible = true,
4913         };
4914         int err;
4915         do {
4916                 err = nfs4_handle_exception(NFS_SERVER(inode),
4917                                 _nfs4_proc_link(inode, dir, name),
4918                                 &exception);
4919         } while (exception.retry);
4920         return err;
4921 }
4922
4923 struct nfs4_createdata {
4924         struct rpc_message msg;
4925         struct nfs4_create_arg arg;
4926         struct nfs4_create_res res;
4927         struct nfs_fh fh;
4928         struct nfs_fattr fattr;
4929 };
4930
4931 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4932                 const struct qstr *name, struct iattr *sattr, u32 ftype)
4933 {
4934         struct nfs4_createdata *data;
4935
4936         data = kzalloc(sizeof(*data), GFP_KERNEL);
4937         if (data != NULL) {
4938                 struct nfs_server *server = NFS_SERVER(dir);
4939
4940                 data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4941                 if (IS_ERR(data->fattr.label))
4942                         goto out_free;
4943
4944                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4945                 data->msg.rpc_argp = &data->arg;
4946                 data->msg.rpc_resp = &data->res;
4947                 data->arg.dir_fh = NFS_FH(dir);
4948                 data->arg.server = server;
4949                 data->arg.name = name;
4950                 data->arg.attrs = sattr;
4951                 data->arg.ftype = ftype;
4952                 data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4953                 data->arg.umask = current_umask();
4954                 data->res.server = server;
4955                 data->res.fh = &data->fh;
4956                 data->res.fattr = &data->fattr;
4957                 nfs_fattr_init(data->res.fattr);
4958         }
4959         return data;
4960 out_free:
4961         kfree(data);
4962         return NULL;
4963 }
4964
4965 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
4966 {
4967         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
4968                                     &data->arg.seq_args, &data->res.seq_res, 1);
4969         if (status == 0) {
4970                 spin_lock(&dir->i_lock);
4971                 /* Creating a directory bumps nlink in the parent */
4972                 if (data->arg.ftype == NF4DIR)
4973                         nfs4_inc_nlink_locked(dir);
4974                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
4975                                               data->res.fattr->time_start,
4976                                               NFS_INO_INVALID_DATA);
4977                 spin_unlock(&dir->i_lock);
4978                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
4979         }
4980         return status;
4981 }
4982
4983 static void nfs4_free_createdata(struct nfs4_createdata *data)
4984 {
4985         nfs4_label_free(data->fattr.label);
4986         kfree(data);
4987 }
4988
4989 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
4990                 struct page *page, unsigned int len, struct iattr *sattr,
4991                 struct nfs4_label *label)
4992 {
4993         struct nfs4_createdata *data;
4994         int status = -ENAMETOOLONG;
4995
4996         if (len > NFS4_MAXPATHLEN)
4997                 goto out;
4998
4999         status = -ENOMEM;
5000         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5001         if (data == NULL)
5002                 goto out;
5003
5004         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5005         data->arg.u.symlink.pages = &page;
5006         data->arg.u.symlink.len = len;
5007         data->arg.label = label;
5008
5009         status = nfs4_do_create(dir, dentry, data);
5010
5011         nfs4_free_createdata(data);
5012 out:
5013         return status;
5014 }
5015
5016 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5017                 struct page *page, unsigned int len, struct iattr *sattr)
5018 {
5019         struct nfs4_exception exception = {
5020                 .interruptible = true,
5021         };
5022         struct nfs4_label l, *label = NULL;
5023         int err;
5024
5025         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5026
5027         do {
5028                 err = _nfs4_proc_symlink(dir, dentry, page, len, sattr, label);
5029                 trace_nfs4_symlink(dir, &dentry->d_name, err);
5030                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5031                                 &exception);
5032         } while (exception.retry);
5033
5034         nfs4_label_release_security(label);
5035         return err;
5036 }
5037
5038 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5039                 struct iattr *sattr, struct nfs4_label *label)
5040 {
5041         struct nfs4_createdata *data;
5042         int status = -ENOMEM;
5043
5044         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5045         if (data == NULL)
5046                 goto out;
5047
5048         data->arg.label = label;
5049         status = nfs4_do_create(dir, dentry, data);
5050
5051         nfs4_free_createdata(data);
5052 out:
5053         return status;
5054 }
5055
5056 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5057                 struct iattr *sattr)
5058 {
5059         struct nfs_server *server = NFS_SERVER(dir);
5060         struct nfs4_exception exception = {
5061                 .interruptible = true,
5062         };
5063         struct nfs4_label l, *label = NULL;
5064         int err;
5065
5066         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5067
5068         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5069                 sattr->ia_mode &= ~current_umask();
5070         do {
5071                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5072                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
5073                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5074                                 &exception);
5075         } while (exception.retry);
5076         nfs4_label_release_security(label);
5077
5078         return err;
5079 }
5080
5081 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5082                               struct nfs_readdir_res *nr_res)
5083 {
5084         struct inode            *dir = d_inode(nr_arg->dentry);
5085         struct nfs_server       *server = NFS_SERVER(dir);
5086         struct nfs4_readdir_arg args = {
5087                 .fh = NFS_FH(dir),
5088                 .pages = nr_arg->pages,
5089                 .pgbase = 0,
5090                 .count = nr_arg->page_len,
5091                 .plus = nr_arg->plus,
5092         };
5093         struct nfs4_readdir_res res;
5094         struct rpc_message msg = {
5095                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5096                 .rpc_argp = &args,
5097                 .rpc_resp = &res,
5098                 .rpc_cred = nr_arg->cred,
5099         };
5100         int                     status;
5101
5102         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5103                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5104         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5105                 args.bitmask = server->attr_bitmask_nl;
5106         else
5107                 args.bitmask = server->attr_bitmask;
5108
5109         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5110         res.pgbase = args.pgbase;
5111         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5112                         &res.seq_res, 0);
5113         if (status >= 0) {
5114                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5115                 status += args.pgbase;
5116         }
5117
5118         nfs_invalidate_atime(dir);
5119
5120         dprintk("%s: returns %d\n", __func__, status);
5121         return status;
5122 }
5123
5124 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5125                              struct nfs_readdir_res *res)
5126 {
5127         struct nfs4_exception exception = {
5128                 .interruptible = true,
5129         };
5130         int err;
5131         do {
5132                 err = _nfs4_proc_readdir(arg, res);
5133                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5134                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5135                                             err, &exception);
5136         } while (exception.retry);
5137         return err;
5138 }
5139
5140 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5141                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5142 {
5143         struct nfs4_createdata *data;
5144         int mode = sattr->ia_mode;
5145         int status = -ENOMEM;
5146
5147         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5148         if (data == NULL)
5149                 goto out;
5150
5151         if (S_ISFIFO(mode))
5152                 data->arg.ftype = NF4FIFO;
5153         else if (S_ISBLK(mode)) {
5154                 data->arg.ftype = NF4BLK;
5155                 data->arg.u.device.specdata1 = MAJOR(rdev);
5156                 data->arg.u.device.specdata2 = MINOR(rdev);
5157         }
5158         else if (S_ISCHR(mode)) {
5159                 data->arg.ftype = NF4CHR;
5160                 data->arg.u.device.specdata1 = MAJOR(rdev);
5161                 data->arg.u.device.specdata2 = MINOR(rdev);
5162         } else if (!S_ISSOCK(mode)) {
5163                 status = -EINVAL;
5164                 goto out_free;
5165         }
5166
5167         data->arg.label = label;
5168         status = nfs4_do_create(dir, dentry, data);
5169 out_free:
5170         nfs4_free_createdata(data);
5171 out:
5172         return status;
5173 }
5174
5175 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5176                 struct iattr *sattr, dev_t rdev)
5177 {
5178         struct nfs_server *server = NFS_SERVER(dir);
5179         struct nfs4_exception exception = {
5180                 .interruptible = true,
5181         };
5182         struct nfs4_label l, *label = NULL;
5183         int err;
5184
5185         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5186
5187         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5188                 sattr->ia_mode &= ~current_umask();
5189         do {
5190                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5191                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5192                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5193                                 &exception);
5194         } while (exception.retry);
5195
5196         nfs4_label_release_security(label);
5197
5198         return err;
5199 }
5200
5201 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5202                  struct nfs_fsstat *fsstat)
5203 {
5204         struct nfs4_statfs_arg args = {
5205                 .fh = fhandle,
5206                 .bitmask = server->attr_bitmask,
5207         };
5208         struct nfs4_statfs_res res = {
5209                 .fsstat = fsstat,
5210         };
5211         struct rpc_message msg = {
5212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5213                 .rpc_argp = &args,
5214                 .rpc_resp = &res,
5215         };
5216
5217         nfs_fattr_init(fsstat->fattr);
5218         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5219 }
5220
5221 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5222 {
5223         struct nfs4_exception exception = {
5224                 .interruptible = true,
5225         };
5226         int err;
5227         do {
5228                 err = nfs4_handle_exception(server,
5229                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5230                                 &exception);
5231         } while (exception.retry);
5232         return err;
5233 }
5234
5235 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5236                 struct nfs_fsinfo *fsinfo)
5237 {
5238         struct nfs4_fsinfo_arg args = {
5239                 .fh = fhandle,
5240                 .bitmask = server->attr_bitmask,
5241         };
5242         struct nfs4_fsinfo_res res = {
5243                 .fsinfo = fsinfo,
5244         };
5245         struct rpc_message msg = {
5246                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5247                 .rpc_argp = &args,
5248                 .rpc_resp = &res,
5249         };
5250
5251         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5252 }
5253
5254 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5255 {
5256         struct nfs4_exception exception = {
5257                 .interruptible = true,
5258         };
5259         int err;
5260
5261         do {
5262                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5263                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5264                 if (err == 0) {
5265                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5266                         break;
5267                 }
5268                 err = nfs4_handle_exception(server, err, &exception);
5269         } while (exception.retry);
5270         return err;
5271 }
5272
5273 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5274 {
5275         int error;
5276
5277         nfs_fattr_init(fsinfo->fattr);
5278         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5279         if (error == 0) {
5280                 /* block layout checks this! */
5281                 server->pnfs_blksize = fsinfo->blksize;
5282                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5283         }
5284
5285         return error;
5286 }
5287
5288 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5289                 struct nfs_pathconf *pathconf)
5290 {
5291         struct nfs4_pathconf_arg args = {
5292                 .fh = fhandle,
5293                 .bitmask = server->attr_bitmask,
5294         };
5295         struct nfs4_pathconf_res res = {
5296                 .pathconf = pathconf,
5297         };
5298         struct rpc_message msg = {
5299                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5300                 .rpc_argp = &args,
5301                 .rpc_resp = &res,
5302         };
5303
5304         /* None of the pathconf attributes are mandatory to implement */
5305         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5306                 memset(pathconf, 0, sizeof(*pathconf));
5307                 return 0;
5308         }
5309
5310         nfs_fattr_init(pathconf->fattr);
5311         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5312 }
5313
5314 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5315                 struct nfs_pathconf *pathconf)
5316 {
5317         struct nfs4_exception exception = {
5318                 .interruptible = true,
5319         };
5320         int err;
5321
5322         do {
5323                 err = nfs4_handle_exception(server,
5324                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5325                                 &exception);
5326         } while (exception.retry);
5327         return err;
5328 }
5329
5330 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5331                 const struct nfs_open_context *ctx,
5332                 const struct nfs_lock_context *l_ctx,
5333                 fmode_t fmode)
5334 {
5335         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5336 }
5337 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5338
5339 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5340                 const struct nfs_open_context *ctx,
5341                 const struct nfs_lock_context *l_ctx,
5342                 fmode_t fmode)
5343 {
5344         nfs4_stateid _current_stateid;
5345
5346         /* If the current stateid represents a lost lock, then exit */
5347         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5348                 return true;
5349         return nfs4_stateid_match(stateid, &_current_stateid);
5350 }
5351
5352 static bool nfs4_error_stateid_expired(int err)
5353 {
5354         switch (err) {
5355         case -NFS4ERR_DELEG_REVOKED:
5356         case -NFS4ERR_ADMIN_REVOKED:
5357         case -NFS4ERR_BAD_STATEID:
5358         case -NFS4ERR_STALE_STATEID:
5359         case -NFS4ERR_OLD_STATEID:
5360         case -NFS4ERR_OPENMODE:
5361         case -NFS4ERR_EXPIRED:
5362                 return true;
5363         }
5364         return false;
5365 }
5366
5367 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5368 {
5369         struct nfs_server *server = NFS_SERVER(hdr->inode);
5370
5371         trace_nfs4_read(hdr, task->tk_status);
5372         if (task->tk_status < 0) {
5373                 struct nfs4_exception exception = {
5374                         .inode = hdr->inode,
5375                         .state = hdr->args.context->state,
5376                         .stateid = &hdr->args.stateid,
5377                 };
5378                 task->tk_status = nfs4_async_handle_exception(task,
5379                                 server, task->tk_status, &exception);
5380                 if (exception.retry) {
5381                         rpc_restart_call_prepare(task);
5382                         return -EAGAIN;
5383                 }
5384         }
5385
5386         if (task->tk_status > 0)
5387                 renew_lease(server, hdr->timestamp);
5388         return 0;
5389 }
5390
5391 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5392                 struct nfs_pgio_args *args)
5393 {
5394
5395         if (!nfs4_error_stateid_expired(task->tk_status) ||
5396                 nfs4_stateid_is_current(&args->stateid,
5397                                 args->context,
5398                                 args->lock_context,
5399                                 FMODE_READ))
5400                 return false;
5401         rpc_restart_call_prepare(task);
5402         return true;
5403 }
5404
5405 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5406                                          struct nfs_pgio_header *hdr)
5407 {
5408         struct nfs_server *server = NFS_SERVER(hdr->inode);
5409         struct rpc_message *msg = &task->tk_msg;
5410
5411         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5412             server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5413                 server->caps &= ~NFS_CAP_READ_PLUS;
5414                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5415                 rpc_restart_call_prepare(task);
5416                 return true;
5417         }
5418         return false;
5419 }
5420
5421 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5422 {
5423         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5424                 return -EAGAIN;
5425         if (nfs4_read_stateid_changed(task, &hdr->args))
5426                 return -EAGAIN;
5427         if (nfs4_read_plus_not_supported(task, hdr))
5428                 return -EAGAIN;
5429         if (task->tk_status > 0)
5430                 nfs_invalidate_atime(hdr->inode);
5431         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5432                                     nfs4_read_done_cb(task, hdr);
5433 }
5434
5435 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5436 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5437                                     struct rpc_message *msg)
5438 {
5439         /* Note: We don't use READ_PLUS with pNFS yet */
5440         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp)
5441                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5442 }
5443 #else
5444 static void nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5445                                     struct rpc_message *msg)
5446 {
5447 }
5448 #endif /* CONFIG_NFS_V4_2 */
5449
5450 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5451                                  struct rpc_message *msg)
5452 {
5453         hdr->timestamp   = jiffies;
5454         if (!hdr->pgio_done_cb)
5455                 hdr->pgio_done_cb = nfs4_read_done_cb;
5456         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5457         nfs42_read_plus_support(hdr, msg);
5458         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5459 }
5460
5461 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5462                                       struct nfs_pgio_header *hdr)
5463 {
5464         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5465                         &hdr->args.seq_args,
5466                         &hdr->res.seq_res,
5467                         task))
5468                 return 0;
5469         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5470                                 hdr->args.lock_context,
5471                                 hdr->rw_mode) == -EIO)
5472                 return -EIO;
5473         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5474                 return -EIO;
5475         return 0;
5476 }
5477
5478 static int nfs4_write_done_cb(struct rpc_task *task,
5479                               struct nfs_pgio_header *hdr)
5480 {
5481         struct inode *inode = hdr->inode;
5482
5483         trace_nfs4_write(hdr, task->tk_status);
5484         if (task->tk_status < 0) {
5485                 struct nfs4_exception exception = {
5486                         .inode = hdr->inode,
5487                         .state = hdr->args.context->state,
5488                         .stateid = &hdr->args.stateid,
5489                 };
5490                 task->tk_status = nfs4_async_handle_exception(task,
5491                                 NFS_SERVER(inode), task->tk_status,
5492                                 &exception);
5493                 if (exception.retry) {
5494                         rpc_restart_call_prepare(task);
5495                         return -EAGAIN;
5496                 }
5497         }
5498         if (task->tk_status >= 0) {
5499                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5500                 nfs_writeback_update_inode(hdr);
5501         }
5502         return 0;
5503 }
5504
5505 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5506                 struct nfs_pgio_args *args)
5507 {
5508
5509         if (!nfs4_error_stateid_expired(task->tk_status) ||
5510                 nfs4_stateid_is_current(&args->stateid,
5511                                 args->context,
5512                                 args->lock_context,
5513                                 FMODE_WRITE))
5514                 return false;
5515         rpc_restart_call_prepare(task);
5516         return true;
5517 }
5518
5519 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5520 {
5521         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5522                 return -EAGAIN;
5523         if (nfs4_write_stateid_changed(task, &hdr->args))
5524                 return -EAGAIN;
5525         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5526                 nfs4_write_done_cb(task, hdr);
5527 }
5528
5529 static
5530 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5531 {
5532         /* Don't request attributes for pNFS or O_DIRECT writes */
5533         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5534                 return false;
5535         /* Otherwise, request attributes if and only if we don't hold
5536          * a delegation
5537          */
5538         return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5539 }
5540
5541 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5542                       struct inode *inode, unsigned long cache_validity)
5543 {
5544         struct nfs_server *server = NFS_SERVER(inode);
5545         unsigned int i;
5546
5547         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5548         cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5549
5550         if (cache_validity & NFS_INO_INVALID_CHANGE)
5551                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5552         if (cache_validity & NFS_INO_INVALID_ATIME)
5553                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5554         if (cache_validity & NFS_INO_INVALID_MODE)
5555                 bitmask[1] |= FATTR4_WORD1_MODE;
5556         if (cache_validity & NFS_INO_INVALID_OTHER)
5557                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5558         if (cache_validity & NFS_INO_INVALID_NLINK)
5559                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5560         if (cache_validity & NFS_INO_INVALID_CTIME)
5561                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5562         if (cache_validity & NFS_INO_INVALID_MTIME)
5563                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5564         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5565                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5566
5567         if (cache_validity & NFS_INO_INVALID_SIZE)
5568                 bitmask[0] |= FATTR4_WORD0_SIZE;
5569
5570         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5571                 bitmask[i] &= server->attr_bitmask[i];
5572 }
5573
5574 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5575                                   struct rpc_message *msg,
5576                                   struct rpc_clnt **clnt)
5577 {
5578         struct nfs_server *server = NFS_SERVER(hdr->inode);
5579
5580         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5581                 hdr->args.bitmask = NULL;
5582                 hdr->res.fattr = NULL;
5583         } else {
5584                 nfs4_bitmask_set(hdr->args.bitmask_store,
5585                                  server->cache_consistency_bitmask,
5586                                  hdr->inode, NFS_INO_INVALID_BLOCKS);
5587                 hdr->args.bitmask = hdr->args.bitmask_store;
5588         }
5589
5590         if (!hdr->pgio_done_cb)
5591                 hdr->pgio_done_cb = nfs4_write_done_cb;
5592         hdr->res.server = server;
5593         hdr->timestamp   = jiffies;
5594
5595         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5596         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5597         nfs4_state_protect_write(server->nfs_client, clnt, msg, hdr);
5598 }
5599
5600 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5601 {
5602         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5603                         &data->args.seq_args,
5604                         &data->res.seq_res,
5605                         task);
5606 }
5607
5608 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5609 {
5610         struct inode *inode = data->inode;
5611
5612         trace_nfs4_commit(data, task->tk_status);
5613         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5614                                     NULL, NULL) == -EAGAIN) {
5615                 rpc_restart_call_prepare(task);
5616                 return -EAGAIN;
5617         }
5618         return 0;
5619 }
5620
5621 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5622 {
5623         if (!nfs4_sequence_done(task, &data->res.seq_res))
5624                 return -EAGAIN;
5625         return data->commit_done_cb(task, data);
5626 }
5627
5628 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5629                                    struct rpc_clnt **clnt)
5630 {
5631         struct nfs_server *server = NFS_SERVER(data->inode);
5632
5633         if (data->commit_done_cb == NULL)
5634                 data->commit_done_cb = nfs4_commit_done_cb;
5635         data->res.server = server;
5636         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5637         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5638         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5639 }
5640
5641 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5642                                 struct nfs_commitres *res)
5643 {
5644         struct inode *dst_inode = file_inode(dst);
5645         struct nfs_server *server = NFS_SERVER(dst_inode);
5646         struct rpc_message msg = {
5647                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5648                 .rpc_argp = args,
5649                 .rpc_resp = res,
5650         };
5651
5652         args->fh = NFS_FH(dst_inode);
5653         return nfs4_call_sync(server->client, server, &msg,
5654                         &args->seq_args, &res->seq_res, 1);
5655 }
5656
5657 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5658 {
5659         struct nfs_commitargs args = {
5660                 .offset = offset,
5661                 .count = count,
5662         };
5663         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5664         struct nfs4_exception exception = { };
5665         int status;
5666
5667         do {
5668                 status = _nfs4_proc_commit(dst, &args, res);
5669                 status = nfs4_handle_exception(dst_server, status, &exception);
5670         } while (exception.retry);
5671
5672         return status;
5673 }
5674
5675 struct nfs4_renewdata {
5676         struct nfs_client       *client;
5677         unsigned long           timestamp;
5678 };
5679
5680 /*
5681  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5682  * standalone procedure for queueing an asynchronous RENEW.
5683  */
5684 static void nfs4_renew_release(void *calldata)
5685 {
5686         struct nfs4_renewdata *data = calldata;
5687         struct nfs_client *clp = data->client;
5688
5689         if (refcount_read(&clp->cl_count) > 1)
5690                 nfs4_schedule_state_renewal(clp);
5691         nfs_put_client(clp);
5692         kfree(data);
5693 }
5694
5695 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5696 {
5697         struct nfs4_renewdata *data = calldata;
5698         struct nfs_client *clp = data->client;
5699         unsigned long timestamp = data->timestamp;
5700
5701         trace_nfs4_renew_async(clp, task->tk_status);
5702         switch (task->tk_status) {
5703         case 0:
5704                 break;
5705         case -NFS4ERR_LEASE_MOVED:
5706                 nfs4_schedule_lease_moved_recovery(clp);
5707                 break;
5708         default:
5709                 /* Unless we're shutting down, schedule state recovery! */
5710                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5711                         return;
5712                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5713                         nfs4_schedule_lease_recovery(clp);
5714                         return;
5715                 }
5716                 nfs4_schedule_path_down_recovery(clp);
5717         }
5718         do_renew_lease(clp, timestamp);
5719 }
5720
5721 static const struct rpc_call_ops nfs4_renew_ops = {
5722         .rpc_call_done = nfs4_renew_done,
5723         .rpc_release = nfs4_renew_release,
5724 };
5725
5726 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5727 {
5728         struct rpc_message msg = {
5729                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5730                 .rpc_argp       = clp,
5731                 .rpc_cred       = cred,
5732         };
5733         struct nfs4_renewdata *data;
5734
5735         if (renew_flags == 0)
5736                 return 0;
5737         if (!refcount_inc_not_zero(&clp->cl_count))
5738                 return -EIO;
5739         data = kmalloc(sizeof(*data), GFP_NOFS);
5740         if (data == NULL) {
5741                 nfs_put_client(clp);
5742                 return -ENOMEM;
5743         }
5744         data->client = clp;
5745         data->timestamp = jiffies;
5746         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5747                         &nfs4_renew_ops, data);
5748 }
5749
5750 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5751 {
5752         struct rpc_message msg = {
5753                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5754                 .rpc_argp       = clp,
5755                 .rpc_cred       = cred,
5756         };
5757         unsigned long now = jiffies;
5758         int status;
5759
5760         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5761         if (status < 0)
5762                 return status;
5763         do_renew_lease(clp, now);
5764         return 0;
5765 }
5766
5767 static inline int nfs4_server_supports_acls(struct nfs_server *server)
5768 {
5769         return server->caps & NFS_CAP_ACLS;
5770 }
5771
5772 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5773  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5774  * the stack.
5775  */
5776 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5777
5778 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5779                 struct page **pages)
5780 {
5781         struct page *newpage, **spages;
5782         int rc = 0;
5783         size_t len;
5784         spages = pages;
5785
5786         do {
5787                 len = min_t(size_t, PAGE_SIZE, buflen);
5788                 newpage = alloc_page(GFP_KERNEL);
5789
5790                 if (newpage == NULL)
5791                         goto unwind;
5792                 memcpy(page_address(newpage), buf, len);
5793                 buf += len;
5794                 buflen -= len;
5795                 *pages++ = newpage;
5796                 rc++;
5797         } while (buflen != 0);
5798
5799         return rc;
5800
5801 unwind:
5802         for(; rc > 0; rc--)
5803                 __free_page(spages[rc-1]);
5804         return -ENOMEM;
5805 }
5806
5807 struct nfs4_cached_acl {
5808         int cached;
5809         size_t len;
5810         char data[];
5811 };
5812
5813 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5814 {
5815         struct nfs_inode *nfsi = NFS_I(inode);
5816
5817         spin_lock(&inode->i_lock);
5818         kfree(nfsi->nfs4_acl);
5819         nfsi->nfs4_acl = acl;
5820         spin_unlock(&inode->i_lock);
5821 }
5822
5823 static void nfs4_zap_acl_attr(struct inode *inode)
5824 {
5825         nfs4_set_cached_acl(inode, NULL);
5826 }
5827
5828 static inline ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf, size_t buflen)
5829 {
5830         struct nfs_inode *nfsi = NFS_I(inode);
5831         struct nfs4_cached_acl *acl;
5832         int ret = -ENOENT;
5833
5834         spin_lock(&inode->i_lock);
5835         acl = nfsi->nfs4_acl;
5836         if (acl == NULL)
5837                 goto out;
5838         if (buf == NULL) /* user is just asking for length */
5839                 goto out_len;
5840         if (acl->cached == 0)
5841                 goto out;
5842         ret = -ERANGE; /* see getxattr(2) man page */
5843         if (acl->len > buflen)
5844                 goto out;
5845         memcpy(buf, acl->data, acl->len);
5846 out_len:
5847         ret = acl->len;
5848 out:
5849         spin_unlock(&inode->i_lock);
5850         return ret;
5851 }
5852
5853 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages, size_t pgbase, size_t acl_len)
5854 {
5855         struct nfs4_cached_acl *acl;
5856         size_t buflen = sizeof(*acl) + acl_len;
5857
5858         if (buflen <= PAGE_SIZE) {
5859                 acl = kmalloc(buflen, GFP_KERNEL);
5860                 if (acl == NULL)
5861                         goto out;
5862                 acl->cached = 1;
5863                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
5864         } else {
5865                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5866                 if (acl == NULL)
5867                         goto out;
5868                 acl->cached = 0;
5869         }
5870         acl->len = acl_len;
5871 out:
5872         nfs4_set_cached_acl(inode, acl);
5873 }
5874
5875 /*
5876  * The getxattr API returns the required buffer length when called with a
5877  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5878  * the required buf.  On a NULL buf, we send a page of data to the server
5879  * guessing that the ACL request can be serviced by a page. If so, we cache
5880  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5881  * the cache. If not so, we throw away the page, and cache the required
5882  * length. The next getxattr call will then produce another round trip to
5883  * the server, this time with the input buf of the required size.
5884  */
5885 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5886 {
5887         struct page **pages;
5888         struct nfs_getaclargs args = {
5889                 .fh = NFS_FH(inode),
5890                 .acl_len = buflen,
5891         };
5892         struct nfs_getaclres res = {
5893                 .acl_len = buflen,
5894         };
5895         struct rpc_message msg = {
5896                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5897                 .rpc_argp = &args,
5898                 .rpc_resp = &res,
5899         };
5900         unsigned int npages;
5901         int ret = -ENOMEM, i;
5902         struct nfs_server *server = NFS_SERVER(inode);
5903
5904         if (buflen == 0)
5905                 buflen = server->rsize;
5906
5907         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5908         pages = kmalloc_array(npages, sizeof(struct page *), GFP_NOFS);
5909         if (!pages)
5910                 return -ENOMEM;
5911
5912         args.acl_pages = pages;
5913
5914         for (i = 0; i < npages; i++) {
5915                 pages[i] = alloc_page(GFP_KERNEL);
5916                 if (!pages[i])
5917                         goto out_free;
5918         }
5919
5920         /* for decoding across pages */
5921         res.acl_scratch = alloc_page(GFP_KERNEL);
5922         if (!res.acl_scratch)
5923                 goto out_free;
5924
5925         args.acl_len = npages * PAGE_SIZE;
5926
5927         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5928                 __func__, buf, buflen, npages, args.acl_len);
5929         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
5930                              &msg, &args.seq_args, &res.seq_res, 0);
5931         if (ret)
5932                 goto out_free;
5933
5934         /* Handle the case where the passed-in buffer is too short */
5935         if (res.acl_flags & NFS4_ACL_TRUNC) {
5936                 /* Did the user only issue a request for the acl length? */
5937                 if (buf == NULL)
5938                         goto out_ok;
5939                 ret = -ERANGE;
5940                 goto out_free;
5941         }
5942         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len);
5943         if (buf) {
5944                 if (res.acl_len > buflen) {
5945                         ret = -ERANGE;
5946                         goto out_free;
5947                 }
5948                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
5949         }
5950 out_ok:
5951         ret = res.acl_len;
5952 out_free:
5953         for (i = 0; i < npages; i++)
5954                 if (pages[i])
5955                         __free_page(pages[i]);
5956         if (res.acl_scratch)
5957                 __free_page(res.acl_scratch);
5958         kfree(pages);
5959         return ret;
5960 }
5961
5962 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf, size_t buflen)
5963 {
5964         struct nfs4_exception exception = {
5965                 .interruptible = true,
5966         };
5967         ssize_t ret;
5968         do {
5969                 ret = __nfs4_get_acl_uncached(inode, buf, buflen);
5970                 trace_nfs4_get_acl(inode, ret);
5971                 if (ret >= 0)
5972                         break;
5973                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
5974         } while (exception.retry);
5975         return ret;
5976 }
5977
5978 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen)
5979 {
5980         struct nfs_server *server = NFS_SERVER(inode);
5981         int ret;
5982
5983         if (!nfs4_server_supports_acls(server))
5984                 return -EOPNOTSUPP;
5985         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
5986         if (ret < 0)
5987                 return ret;
5988         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
5989                 nfs_zap_acl_cache(inode);
5990         ret = nfs4_read_cached_acl(inode, buf, buflen);
5991         if (ret != -ENOENT)
5992                 /* -ENOENT is returned if there is no ACL or if there is an ACL
5993                  * but no cached acl data, just the acl length */
5994                 return ret;
5995         return nfs4_get_acl_uncached(inode, buf, buflen);
5996 }
5997
5998 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
5999 {
6000         struct nfs_server *server = NFS_SERVER(inode);
6001         struct page *pages[NFS4ACL_MAXPAGES];
6002         struct nfs_setaclargs arg = {
6003                 .fh             = NFS_FH(inode),
6004                 .acl_pages      = pages,
6005                 .acl_len        = buflen,
6006         };
6007         struct nfs_setaclres res;
6008         struct rpc_message msg = {
6009                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6010                 .rpc_argp       = &arg,
6011                 .rpc_resp       = &res,
6012         };
6013         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6014         int ret, i;
6015
6016         /* You can't remove system.nfs4_acl: */
6017         if (buflen == 0)
6018                 return -EINVAL;
6019         if (!nfs4_server_supports_acls(server))
6020                 return -EOPNOTSUPP;
6021         if (npages > ARRAY_SIZE(pages))
6022                 return -ERANGE;
6023         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6024         if (i < 0)
6025                 return i;
6026         nfs4_inode_make_writeable(inode);
6027         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6028
6029         /*
6030          * Free each page after tx, so the only ref left is
6031          * held by the network stack
6032          */
6033         for (; i > 0; i--)
6034                 put_page(pages[i-1]);
6035
6036         /*
6037          * Acl update can result in inode attribute update.
6038          * so mark the attribute cache invalid.
6039          */
6040         spin_lock(&inode->i_lock);
6041         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6042                                              NFS_INO_INVALID_CTIME |
6043                                              NFS_INO_REVAL_FORCED);
6044         spin_unlock(&inode->i_lock);
6045         nfs_access_zap_cache(inode);
6046         nfs_zap_acl_cache(inode);
6047         return ret;
6048 }
6049
6050 static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen)
6051 {
6052         struct nfs4_exception exception = { };
6053         int err;
6054         do {
6055                 err = __nfs4_proc_set_acl(inode, buf, buflen);
6056                 trace_nfs4_set_acl(inode, err);
6057                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6058                         /*
6059                          * no need to retry since the kernel
6060                          * isn't involved in encoding the ACEs.
6061                          */
6062                         err = -EINVAL;
6063                         break;
6064                 }
6065                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6066                                 &exception);
6067         } while (exception.retry);
6068         return err;
6069 }
6070
6071 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6072 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6073                                         size_t buflen)
6074 {
6075         struct nfs_server *server = NFS_SERVER(inode);
6076         struct nfs4_label label = {0, 0, buflen, buf};
6077
6078         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6079         struct nfs_fattr fattr = {
6080                 .label = &label,
6081         };
6082         struct nfs4_getattr_arg arg = {
6083                 .fh             = NFS_FH(inode),
6084                 .bitmask        = bitmask,
6085         };
6086         struct nfs4_getattr_res res = {
6087                 .fattr          = &fattr,
6088                 .server         = server,
6089         };
6090         struct rpc_message msg = {
6091                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6092                 .rpc_argp       = &arg,
6093                 .rpc_resp       = &res,
6094         };
6095         int ret;
6096
6097         nfs_fattr_init(&fattr);
6098
6099         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6100         if (ret)
6101                 return ret;
6102         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6103                 return -ENOENT;
6104         return label.len;
6105 }
6106
6107 static int nfs4_get_security_label(struct inode *inode, void *buf,
6108                                         size_t buflen)
6109 {
6110         struct nfs4_exception exception = {
6111                 .interruptible = true,
6112         };
6113         int err;
6114
6115         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6116                 return -EOPNOTSUPP;
6117
6118         do {
6119                 err = _nfs4_get_security_label(inode, buf, buflen);
6120                 trace_nfs4_get_security_label(inode, err);
6121                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6122                                 &exception);
6123         } while (exception.retry);
6124         return err;
6125 }
6126
6127 static int _nfs4_do_set_security_label(struct inode *inode,
6128                 struct nfs4_label *ilabel,
6129                 struct nfs_fattr *fattr)
6130 {
6131
6132         struct iattr sattr = {0};
6133         struct nfs_server *server = NFS_SERVER(inode);
6134         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6135         struct nfs_setattrargs arg = {
6136                 .fh             = NFS_FH(inode),
6137                 .iap            = &sattr,
6138                 .server         = server,
6139                 .bitmask        = bitmask,
6140                 .label          = ilabel,
6141         };
6142         struct nfs_setattrres res = {
6143                 .fattr          = fattr,
6144                 .server         = server,
6145         };
6146         struct rpc_message msg = {
6147                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6148                 .rpc_argp       = &arg,
6149                 .rpc_resp       = &res,
6150         };
6151         int status;
6152
6153         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6154
6155         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6156         if (status)
6157                 dprintk("%s failed: %d\n", __func__, status);
6158
6159         return status;
6160 }
6161
6162 static int nfs4_do_set_security_label(struct inode *inode,
6163                 struct nfs4_label *ilabel,
6164                 struct nfs_fattr *fattr)
6165 {
6166         struct nfs4_exception exception = { };
6167         int err;
6168
6169         do {
6170                 err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6171                 trace_nfs4_set_security_label(inode, err);
6172                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6173                                 &exception);
6174         } while (exception.retry);
6175         return err;
6176 }
6177
6178 static int
6179 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6180 {
6181         struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6182         struct nfs_fattr *fattr;
6183         int status;
6184
6185         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6186                 return -EOPNOTSUPP;
6187
6188         fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6189         if (fattr == NULL)
6190                 return -ENOMEM;
6191
6192         status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6193         if (status == 0)
6194                 nfs_setsecurity(inode, fattr);
6195
6196         return status;
6197 }
6198 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6199
6200
6201 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6202                                     nfs4_verifier *bootverf)
6203 {
6204         __be32 verf[2];
6205
6206         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6207                 /* An impossible timestamp guarantees this value
6208                  * will never match a generated boot time. */
6209                 verf[0] = cpu_to_be32(U32_MAX);
6210                 verf[1] = cpu_to_be32(U32_MAX);
6211         } else {
6212                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6213                 u64 ns = ktime_to_ns(nn->boot_time);
6214
6215                 verf[0] = cpu_to_be32(ns >> 32);
6216                 verf[1] = cpu_to_be32(ns);
6217         }
6218         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6219 }
6220
6221 static size_t
6222 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6223 {
6224         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6225         struct nfs_netns_client *nn_clp = nn->nfs_client;
6226         const char *id;
6227
6228         buf[0] = '\0';
6229
6230         if (nn_clp) {
6231                 rcu_read_lock();
6232                 id = rcu_dereference(nn_clp->identifier);
6233                 if (id)
6234                         strscpy(buf, id, buflen);
6235                 rcu_read_unlock();
6236         }
6237
6238         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6239                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6240
6241         return strlen(buf);
6242 }
6243
6244 static int
6245 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6246 {
6247         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6248         size_t buflen;
6249         size_t len;
6250         char *str;
6251
6252         if (clp->cl_owner_id != NULL)
6253                 return 0;
6254
6255         rcu_read_lock();
6256         len = 14 +
6257                 strlen(clp->cl_rpcclient->cl_nodename) +
6258                 1 +
6259                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6260                 1;
6261         rcu_read_unlock();
6262
6263         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6264         if (buflen)
6265                 len += buflen + 1;
6266
6267         if (len > NFS4_OPAQUE_LIMIT + 1)
6268                 return -EINVAL;
6269
6270         /*
6271          * Since this string is allocated at mount time, and held until the
6272          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6273          * about a memory-reclaim deadlock.
6274          */
6275         str = kmalloc(len, GFP_KERNEL);
6276         if (!str)
6277                 return -ENOMEM;
6278
6279         rcu_read_lock();
6280         if (buflen)
6281                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6282                           clp->cl_rpcclient->cl_nodename, buf,
6283                           rpc_peeraddr2str(clp->cl_rpcclient,
6284                                            RPC_DISPLAY_ADDR));
6285         else
6286                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6287                           clp->cl_rpcclient->cl_nodename,
6288                           rpc_peeraddr2str(clp->cl_rpcclient,
6289                                            RPC_DISPLAY_ADDR));
6290         rcu_read_unlock();
6291
6292         clp->cl_owner_id = str;
6293         return 0;
6294 }
6295
6296 static int
6297 nfs4_init_uniform_client_string(struct nfs_client *clp)
6298 {
6299         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6300         size_t buflen;
6301         size_t len;
6302         char *str;
6303
6304         if (clp->cl_owner_id != NULL)
6305                 return 0;
6306
6307         len = 10 + 10 + 1 + 10 + 1 +
6308                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6309
6310         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6311         if (buflen)
6312                 len += buflen + 1;
6313
6314         if (len > NFS4_OPAQUE_LIMIT + 1)
6315                 return -EINVAL;
6316
6317         /*
6318          * Since this string is allocated at mount time, and held until the
6319          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6320          * about a memory-reclaim deadlock.
6321          */
6322         str = kmalloc(len, GFP_KERNEL);
6323         if (!str)
6324                 return -ENOMEM;
6325
6326         if (buflen)
6327                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6328                           clp->rpc_ops->version, clp->cl_minorversion,
6329                           buf, clp->cl_rpcclient->cl_nodename);
6330         else
6331                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6332                           clp->rpc_ops->version, clp->cl_minorversion,
6333                           clp->cl_rpcclient->cl_nodename);
6334         clp->cl_owner_id = str;
6335         return 0;
6336 }
6337
6338 /*
6339  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6340  * services.  Advertise one based on the address family of the
6341  * clientaddr.
6342  */
6343 static unsigned int
6344 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6345 {
6346         if (strchr(clp->cl_ipaddr, ':') != NULL)
6347                 return scnprintf(buf, len, "tcp6");
6348         else
6349                 return scnprintf(buf, len, "tcp");
6350 }
6351
6352 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6353 {
6354         struct nfs4_setclientid *sc = calldata;
6355
6356         if (task->tk_status == 0)
6357                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6358 }
6359
6360 static const struct rpc_call_ops nfs4_setclientid_ops = {
6361         .rpc_call_done = nfs4_setclientid_done,
6362 };
6363
6364 /**
6365  * nfs4_proc_setclientid - Negotiate client ID
6366  * @clp: state data structure
6367  * @program: RPC program for NFSv4 callback service
6368  * @port: IP port number for NFS4 callback service
6369  * @cred: credential to use for this call
6370  * @res: where to place the result
6371  *
6372  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6373  */
6374 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6375                 unsigned short port, const struct cred *cred,
6376                 struct nfs4_setclientid_res *res)
6377 {
6378         nfs4_verifier sc_verifier;
6379         struct nfs4_setclientid setclientid = {
6380                 .sc_verifier = &sc_verifier,
6381                 .sc_prog = program,
6382                 .sc_clnt = clp,
6383         };
6384         struct rpc_message msg = {
6385                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6386                 .rpc_argp = &setclientid,
6387                 .rpc_resp = res,
6388                 .rpc_cred = cred,
6389         };
6390         struct rpc_task_setup task_setup_data = {
6391                 .rpc_client = clp->cl_rpcclient,
6392                 .rpc_message = &msg,
6393                 .callback_ops = &nfs4_setclientid_ops,
6394                 .callback_data = &setclientid,
6395                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6396         };
6397         unsigned long now = jiffies;
6398         int status;
6399
6400         /* nfs_client_id4 */
6401         nfs4_init_boot_verifier(clp, &sc_verifier);
6402
6403         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6404                 status = nfs4_init_uniform_client_string(clp);
6405         else
6406                 status = nfs4_init_nonuniform_client_string(clp);
6407
6408         if (status)
6409                 goto out;
6410
6411         /* cb_client4 */
6412         setclientid.sc_netid_len =
6413                                 nfs4_init_callback_netid(clp,
6414                                                 setclientid.sc_netid,
6415                                                 sizeof(setclientid.sc_netid));
6416         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6417                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6418                                 clp->cl_ipaddr, port >> 8, port & 255);
6419
6420         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6421                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6422                 clp->cl_owner_id);
6423
6424         status = nfs4_call_sync_custom(&task_setup_data);
6425         if (setclientid.sc_cred) {
6426                 kfree(clp->cl_acceptor);
6427                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6428                 put_rpccred(setclientid.sc_cred);
6429         }
6430
6431         if (status == 0)
6432                 do_renew_lease(clp, now);
6433 out:
6434         trace_nfs4_setclientid(clp, status);
6435         dprintk("NFS reply setclientid: %d\n", status);
6436         return status;
6437 }
6438
6439 /**
6440  * nfs4_proc_setclientid_confirm - Confirm client ID
6441  * @clp: state data structure
6442  * @arg: result of a previous SETCLIENTID
6443  * @cred: credential to use for this call
6444  *
6445  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6446  */
6447 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6448                 struct nfs4_setclientid_res *arg,
6449                 const struct cred *cred)
6450 {
6451         struct rpc_message msg = {
6452                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6453                 .rpc_argp = arg,
6454                 .rpc_cred = cred,
6455         };
6456         int status;
6457
6458         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6459                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6460                 clp->cl_clientid);
6461         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6462                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6463         trace_nfs4_setclientid_confirm(clp, status);
6464         dprintk("NFS reply setclientid_confirm: %d\n", status);
6465         return status;
6466 }
6467
6468 struct nfs4_delegreturndata {
6469         struct nfs4_delegreturnargs args;
6470         struct nfs4_delegreturnres res;
6471         struct nfs_fh fh;
6472         nfs4_stateid stateid;
6473         unsigned long timestamp;
6474         struct {
6475                 struct nfs4_layoutreturn_args arg;
6476                 struct nfs4_layoutreturn_res res;
6477                 struct nfs4_xdr_opaque_data ld_private;
6478                 u32 roc_barrier;
6479                 bool roc;
6480         } lr;
6481         struct nfs_fattr fattr;
6482         int rpc_status;
6483         struct inode *inode;
6484 };
6485
6486 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6487 {
6488         struct nfs4_delegreturndata *data = calldata;
6489         struct nfs4_exception exception = {
6490                 .inode = data->inode,
6491                 .stateid = &data->stateid,
6492                 .task_is_privileged = data->args.seq_args.sa_privileged,
6493         };
6494
6495         if (!nfs4_sequence_done(task, &data->res.seq_res))
6496                 return;
6497
6498         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6499
6500         /* Handle Layoutreturn errors */
6501         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6502                           &data->res.lr_ret) == -EAGAIN)
6503                 goto out_restart;
6504
6505         switch (task->tk_status) {
6506         case 0:
6507                 renew_lease(data->res.server, data->timestamp);
6508                 break;
6509         case -NFS4ERR_ADMIN_REVOKED:
6510         case -NFS4ERR_DELEG_REVOKED:
6511         case -NFS4ERR_EXPIRED:
6512                 nfs4_free_revoked_stateid(data->res.server,
6513                                 data->args.stateid,
6514                                 task->tk_msg.rpc_cred);
6515                 fallthrough;
6516         case -NFS4ERR_BAD_STATEID:
6517         case -NFS4ERR_STALE_STATEID:
6518         case -ETIMEDOUT:
6519                 task->tk_status = 0;
6520                 break;
6521         case -NFS4ERR_OLD_STATEID:
6522                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6523                         nfs4_stateid_seqid_inc(&data->stateid);
6524                 if (data->args.bitmask) {
6525                         data->args.bitmask = NULL;
6526                         data->res.fattr = NULL;
6527                 }
6528                 goto out_restart;
6529         case -NFS4ERR_ACCESS:
6530                 if (data->args.bitmask) {
6531                         data->args.bitmask = NULL;
6532                         data->res.fattr = NULL;
6533                         goto out_restart;
6534                 }
6535                 fallthrough;
6536         default:
6537                 task->tk_status = nfs4_async_handle_exception(task,
6538                                 data->res.server, task->tk_status,
6539                                 &exception);
6540                 if (exception.retry)
6541                         goto out_restart;
6542         }
6543         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6544         data->rpc_status = task->tk_status;
6545         return;
6546 out_restart:
6547         task->tk_status = 0;
6548         rpc_restart_call_prepare(task);
6549 }
6550
6551 static void nfs4_delegreturn_release(void *calldata)
6552 {
6553         struct nfs4_delegreturndata *data = calldata;
6554         struct inode *inode = data->inode;
6555
6556         if (data->lr.roc)
6557                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6558                                  data->res.lr_ret);
6559         if (inode) {
6560                 nfs_post_op_update_inode_force_wcc(inode, &data->fattr);
6561                 nfs_iput_and_deactive(inode);
6562         }
6563         kfree(calldata);
6564 }
6565
6566 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6567 {
6568         struct nfs4_delegreturndata *d_data;
6569         struct pnfs_layout_hdr *lo;
6570
6571         d_data = (struct nfs4_delegreturndata *)data;
6572
6573         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6574                 nfs4_sequence_done(task, &d_data->res.seq_res);
6575                 return;
6576         }
6577
6578         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6579         if (lo && !pnfs_layout_is_valid(lo)) {
6580                 d_data->args.lr_args = NULL;
6581                 d_data->res.lr_res = NULL;
6582         }
6583
6584         nfs4_setup_sequence(d_data->res.server->nfs_client,
6585                         &d_data->args.seq_args,
6586                         &d_data->res.seq_res,
6587                         task);
6588 }
6589
6590 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6591         .rpc_call_prepare = nfs4_delegreturn_prepare,
6592         .rpc_call_done = nfs4_delegreturn_done,
6593         .rpc_release = nfs4_delegreturn_release,
6594 };
6595
6596 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6597 {
6598         struct nfs4_delegreturndata *data;
6599         struct nfs_server *server = NFS_SERVER(inode);
6600         struct rpc_task *task;
6601         struct rpc_message msg = {
6602                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6603                 .rpc_cred = cred,
6604         };
6605         struct rpc_task_setup task_setup_data = {
6606                 .rpc_client = server->client,
6607                 .rpc_message = &msg,
6608                 .callback_ops = &nfs4_delegreturn_ops,
6609                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
6610         };
6611         int status = 0;
6612
6613         data = kzalloc(sizeof(*data), GFP_NOFS);
6614         if (data == NULL)
6615                 return -ENOMEM;
6616
6617         nfs4_state_protect(server->nfs_client,
6618                         NFS_SP4_MACH_CRED_CLEANUP,
6619                         &task_setup_data.rpc_client, &msg);
6620
6621         data->args.fhandle = &data->fh;
6622         data->args.stateid = &data->stateid;
6623         nfs4_bitmask_set(data->args.bitmask_store,
6624                          server->cache_consistency_bitmask, inode, 0);
6625         data->args.bitmask = data->args.bitmask_store;
6626         nfs_copy_fh(&data->fh, NFS_FH(inode));
6627         nfs4_stateid_copy(&data->stateid, stateid);
6628         data->res.fattr = &data->fattr;
6629         data->res.server = server;
6630         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6631         data->lr.arg.ld_private = &data->lr.ld_private;
6632         nfs_fattr_init(data->res.fattr);
6633         data->timestamp = jiffies;
6634         data->rpc_status = 0;
6635         data->inode = nfs_igrab_and_active(inode);
6636         if (data->inode || issync) {
6637                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6638                                         cred);
6639                 if (data->lr.roc) {
6640                         data->args.lr_args = &data->lr.arg;
6641                         data->res.lr_res = &data->lr.res;
6642                 }
6643         }
6644
6645         if (!data->inode)
6646                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6647                                    1);
6648         else
6649                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6650                                    0);
6651         task_setup_data.callback_data = data;
6652         msg.rpc_argp = &data->args;
6653         msg.rpc_resp = &data->res;
6654         task = rpc_run_task(&task_setup_data);
6655         if (IS_ERR(task))
6656                 return PTR_ERR(task);
6657         if (!issync)
6658                 goto out;
6659         status = rpc_wait_for_completion_task(task);
6660         if (status != 0)
6661                 goto out;
6662         status = data->rpc_status;
6663 out:
6664         rpc_put_task(task);
6665         return status;
6666 }
6667
6668 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6669 {
6670         struct nfs_server *server = NFS_SERVER(inode);
6671         struct nfs4_exception exception = { };
6672         int err;
6673         do {
6674                 err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6675                 trace_nfs4_delegreturn(inode, stateid, err);
6676                 switch (err) {
6677                         case -NFS4ERR_STALE_STATEID:
6678                         case -NFS4ERR_EXPIRED:
6679                         case 0:
6680                                 return 0;
6681                 }
6682                 err = nfs4_handle_exception(server, err, &exception);
6683         } while (exception.retry);
6684         return err;
6685 }
6686
6687 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6688 {
6689         struct inode *inode = state->inode;
6690         struct nfs_server *server = NFS_SERVER(inode);
6691         struct nfs_client *clp = server->nfs_client;
6692         struct nfs_lockt_args arg = {
6693                 .fh = NFS_FH(inode),
6694                 .fl = request,
6695         };
6696         struct nfs_lockt_res res = {
6697                 .denied = request,
6698         };
6699         struct rpc_message msg = {
6700                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6701                 .rpc_argp       = &arg,
6702                 .rpc_resp       = &res,
6703                 .rpc_cred       = state->owner->so_cred,
6704         };
6705         struct nfs4_lock_state *lsp;
6706         int status;
6707
6708         arg.lock_owner.clientid = clp->cl_clientid;
6709         status = nfs4_set_lock_state(state, request);
6710         if (status != 0)
6711                 goto out;
6712         lsp = request->fl_u.nfs4_fl.owner;
6713         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6714         arg.lock_owner.s_dev = server->s_dev;
6715         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6716         switch (status) {
6717                 case 0:
6718                         request->fl_type = F_UNLCK;
6719                         break;
6720                 case -NFS4ERR_DENIED:
6721                         status = 0;
6722         }
6723         request->fl_ops->fl_release_private(request);
6724         request->fl_ops = NULL;
6725 out:
6726         return status;
6727 }
6728
6729 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6730 {
6731         struct nfs4_exception exception = {
6732                 .interruptible = true,
6733         };
6734         int err;
6735
6736         do {
6737                 err = _nfs4_proc_getlk(state, cmd, request);
6738                 trace_nfs4_get_lock(request, state, cmd, err);
6739                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6740                                 &exception);
6741         } while (exception.retry);
6742         return err;
6743 }
6744
6745 /*
6746  * Update the seqid of a lock stateid after receiving
6747  * NFS4ERR_OLD_STATEID
6748  */
6749 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6750                 struct nfs4_lock_state *lsp)
6751 {
6752         struct nfs4_state *state = lsp->ls_state;
6753         bool ret = false;
6754
6755         spin_lock(&state->state_lock);
6756         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6757                 goto out;
6758         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6759                 nfs4_stateid_seqid_inc(dst);
6760         else
6761                 dst->seqid = lsp->ls_stateid.seqid;
6762         ret = true;
6763 out:
6764         spin_unlock(&state->state_lock);
6765         return ret;
6766 }
6767
6768 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6769                 struct nfs4_lock_state *lsp)
6770 {
6771         struct nfs4_state *state = lsp->ls_state;
6772         bool ret;
6773
6774         spin_lock(&state->state_lock);
6775         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6776         nfs4_stateid_copy(dst, &lsp->ls_stateid);
6777         spin_unlock(&state->state_lock);
6778         return ret;
6779 }
6780
6781 struct nfs4_unlockdata {
6782         struct nfs_locku_args arg;
6783         struct nfs_locku_res res;
6784         struct nfs4_lock_state *lsp;
6785         struct nfs_open_context *ctx;
6786         struct nfs_lock_context *l_ctx;
6787         struct file_lock fl;
6788         struct nfs_server *server;
6789         unsigned long timestamp;
6790 };
6791
6792 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6793                 struct nfs_open_context *ctx,
6794                 struct nfs4_lock_state *lsp,
6795                 struct nfs_seqid *seqid)
6796 {
6797         struct nfs4_unlockdata *p;
6798         struct nfs4_state *state = lsp->ls_state;
6799         struct inode *inode = state->inode;
6800
6801         p = kzalloc(sizeof(*p), GFP_NOFS);
6802         if (p == NULL)
6803                 return NULL;
6804         p->arg.fh = NFS_FH(inode);
6805         p->arg.fl = &p->fl;
6806         p->arg.seqid = seqid;
6807         p->res.seqid = seqid;
6808         p->lsp = lsp;
6809         /* Ensure we don't close file until we're done freeing locks! */
6810         p->ctx = get_nfs_open_context(ctx);
6811         p->l_ctx = nfs_get_lock_context(ctx);
6812         locks_init_lock(&p->fl);
6813         locks_copy_lock(&p->fl, fl);
6814         p->server = NFS_SERVER(inode);
6815         spin_lock(&state->state_lock);
6816         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6817         spin_unlock(&state->state_lock);
6818         return p;
6819 }
6820
6821 static void nfs4_locku_release_calldata(void *data)
6822 {
6823         struct nfs4_unlockdata *calldata = data;
6824         nfs_free_seqid(calldata->arg.seqid);
6825         nfs4_put_lock_state(calldata->lsp);
6826         nfs_put_lock_context(calldata->l_ctx);
6827         put_nfs_open_context(calldata->ctx);
6828         kfree(calldata);
6829 }
6830
6831 static void nfs4_locku_done(struct rpc_task *task, void *data)
6832 {
6833         struct nfs4_unlockdata *calldata = data;
6834         struct nfs4_exception exception = {
6835                 .inode = calldata->lsp->ls_state->inode,
6836                 .stateid = &calldata->arg.stateid,
6837         };
6838
6839         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6840                 return;
6841         switch (task->tk_status) {
6842                 case 0:
6843                         renew_lease(calldata->server, calldata->timestamp);
6844                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6845                         if (nfs4_update_lock_stateid(calldata->lsp,
6846                                         &calldata->res.stateid))
6847                                 break;
6848                         fallthrough;
6849                 case -NFS4ERR_ADMIN_REVOKED:
6850                 case -NFS4ERR_EXPIRED:
6851                         nfs4_free_revoked_stateid(calldata->server,
6852                                         &calldata->arg.stateid,
6853                                         task->tk_msg.rpc_cred);
6854                         fallthrough;
6855                 case -NFS4ERR_BAD_STATEID:
6856                 case -NFS4ERR_STALE_STATEID:
6857                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6858                                                 calldata->lsp))
6859                                 rpc_restart_call_prepare(task);
6860                         break;
6861                 case -NFS4ERR_OLD_STATEID:
6862                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6863                                                 calldata->lsp))
6864                                 rpc_restart_call_prepare(task);
6865                         break;
6866                 default:
6867                         task->tk_status = nfs4_async_handle_exception(task,
6868                                         calldata->server, task->tk_status,
6869                                         &exception);
6870                         if (exception.retry)
6871                                 rpc_restart_call_prepare(task);
6872         }
6873         nfs_release_seqid(calldata->arg.seqid);
6874 }
6875
6876 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6877 {
6878         struct nfs4_unlockdata *calldata = data;
6879
6880         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6881                 nfs_async_iocounter_wait(task, calldata->l_ctx))
6882                 return;
6883
6884         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6885                 goto out_wait;
6886         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6887                 /* Note: exit _without_ running nfs4_locku_done */
6888                 goto out_no_action;
6889         }
6890         calldata->timestamp = jiffies;
6891         if (nfs4_setup_sequence(calldata->server->nfs_client,
6892                                 &calldata->arg.seq_args,
6893                                 &calldata->res.seq_res,
6894                                 task) != 0)
6895                 nfs_release_seqid(calldata->arg.seqid);
6896         return;
6897 out_no_action:
6898         task->tk_action = NULL;
6899 out_wait:
6900         nfs4_sequence_done(task, &calldata->res.seq_res);
6901 }
6902
6903 static const struct rpc_call_ops nfs4_locku_ops = {
6904         .rpc_call_prepare = nfs4_locku_prepare,
6905         .rpc_call_done = nfs4_locku_done,
6906         .rpc_release = nfs4_locku_release_calldata,
6907 };
6908
6909 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6910                 struct nfs_open_context *ctx,
6911                 struct nfs4_lock_state *lsp,
6912                 struct nfs_seqid *seqid)
6913 {
6914         struct nfs4_unlockdata *data;
6915         struct rpc_message msg = {
6916                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6917                 .rpc_cred = ctx->cred,
6918         };
6919         struct rpc_task_setup task_setup_data = {
6920                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
6921                 .rpc_message = &msg,
6922                 .callback_ops = &nfs4_locku_ops,
6923                 .workqueue = nfsiod_workqueue,
6924                 .flags = RPC_TASK_ASYNC,
6925         };
6926         struct nfs_client *client =
6927                 NFS_SERVER(lsp->ls_state->inode)->nfs_client;
6928
6929         if (client->cl_minorversion)
6930                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6931
6932         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
6933                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
6934
6935         /* Ensure this is an unlock - when canceling a lock, the
6936          * canceled lock is passed in, and it won't be an unlock.
6937          */
6938         fl->fl_type = F_UNLCK;
6939         if (fl->fl_flags & FL_CLOSE)
6940                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
6941
6942         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
6943         if (data == NULL) {
6944                 nfs_free_seqid(seqid);
6945                 return ERR_PTR(-ENOMEM);
6946         }
6947
6948         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
6949         msg.rpc_argp = &data->arg;
6950         msg.rpc_resp = &data->res;
6951         task_setup_data.callback_data = data;
6952         return rpc_run_task(&task_setup_data);
6953 }
6954
6955 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
6956 {
6957         struct inode *inode = state->inode;
6958         struct nfs4_state_owner *sp = state->owner;
6959         struct nfs_inode *nfsi = NFS_I(inode);
6960         struct nfs_seqid *seqid;
6961         struct nfs4_lock_state *lsp;
6962         struct rpc_task *task;
6963         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
6964         int status = 0;
6965         unsigned char fl_flags = request->fl_flags;
6966
6967         status = nfs4_set_lock_state(state, request);
6968         /* Unlock _before_ we do the RPC call */
6969         request->fl_flags |= FL_EXISTS;
6970         /* Exclude nfs_delegation_claim_locks() */
6971         mutex_lock(&sp->so_delegreturn_mutex);
6972         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
6973         down_read(&nfsi->rwsem);
6974         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
6975                 up_read(&nfsi->rwsem);
6976                 mutex_unlock(&sp->so_delegreturn_mutex);
6977                 goto out;
6978         }
6979         up_read(&nfsi->rwsem);
6980         mutex_unlock(&sp->so_delegreturn_mutex);
6981         if (status != 0)
6982                 goto out;
6983         /* Is this a delegated lock? */
6984         lsp = request->fl_u.nfs4_fl.owner;
6985         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
6986                 goto out;
6987         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
6988         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
6989         status = -ENOMEM;
6990         if (IS_ERR(seqid))
6991                 goto out;
6992         task = nfs4_do_unlck(request, nfs_file_open_context(request->fl_file), lsp, seqid);
6993         status = PTR_ERR(task);
6994         if (IS_ERR(task))
6995                 goto out;
6996         status = rpc_wait_for_completion_task(task);
6997         rpc_put_task(task);
6998 out:
6999         request->fl_flags = fl_flags;
7000         trace_nfs4_unlock(request, state, F_SETLK, status);
7001         return status;
7002 }
7003
7004 struct nfs4_lockdata {
7005         struct nfs_lock_args arg;
7006         struct nfs_lock_res res;
7007         struct nfs4_lock_state *lsp;
7008         struct nfs_open_context *ctx;
7009         struct file_lock fl;
7010         unsigned long timestamp;
7011         int rpc_status;
7012         int cancelled;
7013         struct nfs_server *server;
7014 };
7015
7016 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7017                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7018                 gfp_t gfp_mask)
7019 {
7020         struct nfs4_lockdata *p;
7021         struct inode *inode = lsp->ls_state->inode;
7022         struct nfs_server *server = NFS_SERVER(inode);
7023         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7024
7025         p = kzalloc(sizeof(*p), gfp_mask);
7026         if (p == NULL)
7027                 return NULL;
7028
7029         p->arg.fh = NFS_FH(inode);
7030         p->arg.fl = &p->fl;
7031         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7032         if (IS_ERR(p->arg.open_seqid))
7033                 goto out_free;
7034         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7035         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7036         if (IS_ERR(p->arg.lock_seqid))
7037                 goto out_free_seqid;
7038         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7039         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7040         p->arg.lock_owner.s_dev = server->s_dev;
7041         p->res.lock_seqid = p->arg.lock_seqid;
7042         p->lsp = lsp;
7043         p->server = server;
7044         p->ctx = get_nfs_open_context(ctx);
7045         locks_init_lock(&p->fl);
7046         locks_copy_lock(&p->fl, fl);
7047         return p;
7048 out_free_seqid:
7049         nfs_free_seqid(p->arg.open_seqid);
7050 out_free:
7051         kfree(p);
7052         return NULL;
7053 }
7054
7055 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7056 {
7057         struct nfs4_lockdata *data = calldata;
7058         struct nfs4_state *state = data->lsp->ls_state;
7059
7060         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7061                 goto out_wait;
7062         /* Do we need to do an open_to_lock_owner? */
7063         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7064                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7065                         goto out_release_lock_seqid;
7066                 }
7067                 nfs4_stateid_copy(&data->arg.open_stateid,
7068                                 &state->open_stateid);
7069                 data->arg.new_lock_owner = 1;
7070                 data->res.open_seqid = data->arg.open_seqid;
7071         } else {
7072                 data->arg.new_lock_owner = 0;
7073                 nfs4_stateid_copy(&data->arg.lock_stateid,
7074                                 &data->lsp->ls_stateid);
7075         }
7076         if (!nfs4_valid_open_stateid(state)) {
7077                 data->rpc_status = -EBADF;
7078                 task->tk_action = NULL;
7079                 goto out_release_open_seqid;
7080         }
7081         data->timestamp = jiffies;
7082         if (nfs4_setup_sequence(data->server->nfs_client,
7083                                 &data->arg.seq_args,
7084                                 &data->res.seq_res,
7085                                 task) == 0)
7086                 return;
7087 out_release_open_seqid:
7088         nfs_release_seqid(data->arg.open_seqid);
7089 out_release_lock_seqid:
7090         nfs_release_seqid(data->arg.lock_seqid);
7091 out_wait:
7092         nfs4_sequence_done(task, &data->res.seq_res);
7093         dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7094 }
7095
7096 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7097 {
7098         struct nfs4_lockdata *data = calldata;
7099         struct nfs4_lock_state *lsp = data->lsp;
7100
7101         if (!nfs4_sequence_done(task, &data->res.seq_res))
7102                 return;
7103
7104         data->rpc_status = task->tk_status;
7105         switch (task->tk_status) {
7106         case 0:
7107                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7108                                 data->timestamp);
7109                 if (data->arg.new_lock && !data->cancelled) {
7110                         data->fl.fl_flags &= ~(FL_SLEEP | FL_ACCESS);
7111                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7112                                 goto out_restart;
7113                 }
7114                 if (data->arg.new_lock_owner != 0) {
7115                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7116                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7117                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7118                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7119                         goto out_restart;
7120                 break;
7121         case -NFS4ERR_BAD_STATEID:
7122         case -NFS4ERR_OLD_STATEID:
7123         case -NFS4ERR_STALE_STATEID:
7124         case -NFS4ERR_EXPIRED:
7125                 if (data->arg.new_lock_owner != 0) {
7126                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7127                                                 &lsp->ls_state->open_stateid))
7128                                 goto out_restart;
7129                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7130                                                 &lsp->ls_stateid))
7131                                 goto out_restart;
7132         }
7133 out_done:
7134         dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7135         return;
7136 out_restart:
7137         if (!data->cancelled)
7138                 rpc_restart_call_prepare(task);
7139         goto out_done;
7140 }
7141
7142 static void nfs4_lock_release(void *calldata)
7143 {
7144         struct nfs4_lockdata *data = calldata;
7145
7146         nfs_free_seqid(data->arg.open_seqid);
7147         if (data->cancelled && data->rpc_status == 0) {
7148                 struct rpc_task *task;
7149                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7150                                 data->arg.lock_seqid);
7151                 if (!IS_ERR(task))
7152                         rpc_put_task_async(task);
7153                 dprintk("%s: cancelling lock!\n", __func__);
7154         } else
7155                 nfs_free_seqid(data->arg.lock_seqid);
7156         nfs4_put_lock_state(data->lsp);
7157         put_nfs_open_context(data->ctx);
7158         kfree(data);
7159 }
7160
7161 static const struct rpc_call_ops nfs4_lock_ops = {
7162         .rpc_call_prepare = nfs4_lock_prepare,
7163         .rpc_call_done = nfs4_lock_done,
7164         .rpc_release = nfs4_lock_release,
7165 };
7166
7167 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7168 {
7169         switch (error) {
7170         case -NFS4ERR_ADMIN_REVOKED:
7171         case -NFS4ERR_EXPIRED:
7172         case -NFS4ERR_BAD_STATEID:
7173                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7174                 if (new_lock_owner != 0 ||
7175                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7176                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7177                 break;
7178         case -NFS4ERR_STALE_STATEID:
7179                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7180                 nfs4_schedule_lease_recovery(server->nfs_client);
7181         }
7182 }
7183
7184 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7185 {
7186         struct nfs4_lockdata *data;
7187         struct rpc_task *task;
7188         struct rpc_message msg = {
7189                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7190                 .rpc_cred = state->owner->so_cred,
7191         };
7192         struct rpc_task_setup task_setup_data = {
7193                 .rpc_client = NFS_CLIENT(state->inode),
7194                 .rpc_message = &msg,
7195                 .callback_ops = &nfs4_lock_ops,
7196                 .workqueue = nfsiod_workqueue,
7197                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7198         };
7199         int ret;
7200         struct nfs_client *client = NFS_SERVER(state->inode)->nfs_client;
7201
7202         if (client->cl_minorversion)
7203                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7204
7205         data = nfs4_alloc_lockdata(fl, nfs_file_open_context(fl->fl_file),
7206                         fl->fl_u.nfs4_fl.owner,
7207                         recovery_type == NFS_LOCK_NEW ? GFP_KERNEL : GFP_NOFS);
7208         if (data == NULL)
7209                 return -ENOMEM;
7210         if (IS_SETLKW(cmd))
7211                 data->arg.block = 1;
7212         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7213                                 recovery_type > NFS_LOCK_NEW);
7214         msg.rpc_argp = &data->arg;
7215         msg.rpc_resp = &data->res;
7216         task_setup_data.callback_data = data;
7217         if (recovery_type > NFS_LOCK_NEW) {
7218                 if (recovery_type == NFS_LOCK_RECLAIM)
7219                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7220         } else
7221                 data->arg.new_lock = 1;
7222         task = rpc_run_task(&task_setup_data);
7223         if (IS_ERR(task))
7224                 return PTR_ERR(task);
7225         ret = rpc_wait_for_completion_task(task);
7226         if (ret == 0) {
7227                 ret = data->rpc_status;
7228                 if (ret)
7229                         nfs4_handle_setlk_error(data->server, data->lsp,
7230                                         data->arg.new_lock_owner, ret);
7231         } else
7232                 data->cancelled = true;
7233         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7234         rpc_put_task(task);
7235         dprintk("%s: ret = %d\n", __func__, ret);
7236         return ret;
7237 }
7238
7239 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7240 {
7241         struct nfs_server *server = NFS_SERVER(state->inode);
7242         struct nfs4_exception exception = {
7243                 .inode = state->inode,
7244         };
7245         int err;
7246
7247         do {
7248                 /* Cache the lock if possible... */
7249                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7250                         return 0;
7251                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7252                 if (err != -NFS4ERR_DELAY)
7253                         break;
7254                 nfs4_handle_exception(server, err, &exception);
7255         } while (exception.retry);
7256         return err;
7257 }
7258
7259 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7260 {
7261         struct nfs_server *server = NFS_SERVER(state->inode);
7262         struct nfs4_exception exception = {
7263                 .inode = state->inode,
7264         };
7265         int err;
7266
7267         err = nfs4_set_lock_state(state, request);
7268         if (err != 0)
7269                 return err;
7270         if (!recover_lost_locks) {
7271                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7272                 return 0;
7273         }
7274         do {
7275                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7276                         return 0;
7277                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7278                 switch (err) {
7279                 default:
7280                         goto out;
7281                 case -NFS4ERR_GRACE:
7282                 case -NFS4ERR_DELAY:
7283                         nfs4_handle_exception(server, err, &exception);
7284                         err = 0;
7285                 }
7286         } while (exception.retry);
7287 out:
7288         return err;
7289 }
7290
7291 #if defined(CONFIG_NFS_V4_1)
7292 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7293 {
7294         struct nfs4_lock_state *lsp;
7295         int status;
7296
7297         status = nfs4_set_lock_state(state, request);
7298         if (status != 0)
7299                 return status;
7300         lsp = request->fl_u.nfs4_fl.owner;
7301         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7302             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7303                 return 0;
7304         return nfs4_lock_expired(state, request);
7305 }
7306 #endif
7307
7308 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7309 {
7310         struct nfs_inode *nfsi = NFS_I(state->inode);
7311         struct nfs4_state_owner *sp = state->owner;
7312         unsigned char fl_flags = request->fl_flags;
7313         int status;
7314
7315         request->fl_flags |= FL_ACCESS;
7316         status = locks_lock_inode_wait(state->inode, request);
7317         if (status < 0)
7318                 goto out;
7319         mutex_lock(&sp->so_delegreturn_mutex);
7320         down_read(&nfsi->rwsem);
7321         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7322                 /* Yes: cache locks! */
7323                 /* ...but avoid races with delegation recall... */
7324                 request->fl_flags = fl_flags & ~FL_SLEEP;
7325                 status = locks_lock_inode_wait(state->inode, request);
7326                 up_read(&nfsi->rwsem);
7327                 mutex_unlock(&sp->so_delegreturn_mutex);
7328                 goto out;
7329         }
7330         up_read(&nfsi->rwsem);
7331         mutex_unlock(&sp->so_delegreturn_mutex);
7332         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7333 out:
7334         request->fl_flags = fl_flags;
7335         return status;
7336 }
7337
7338 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7339 {
7340         struct nfs4_exception exception = {
7341                 .state = state,
7342                 .inode = state->inode,
7343                 .interruptible = true,
7344         };
7345         int err;
7346
7347         do {
7348                 err = _nfs4_proc_setlk(state, cmd, request);
7349                 if (err == -NFS4ERR_DENIED)
7350                         err = -EAGAIN;
7351                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7352                                 err, &exception);
7353         } while (exception.retry);
7354         return err;
7355 }
7356
7357 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7358 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7359
7360 static int
7361 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7362                         struct file_lock *request)
7363 {
7364         int             status = -ERESTARTSYS;
7365         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7366
7367         while(!signalled()) {
7368                 status = nfs4_proc_setlk(state, cmd, request);
7369                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7370                         break;
7371                 freezable_schedule_timeout_interruptible(timeout);
7372                 timeout *= 2;
7373                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7374                 status = -ERESTARTSYS;
7375         }
7376         return status;
7377 }
7378
7379 #ifdef CONFIG_NFS_V4_1
7380 struct nfs4_lock_waiter {
7381         struct inode            *inode;
7382         struct nfs_lowner       owner;
7383         wait_queue_entry_t      wait;
7384 };
7385
7386 static int
7387 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7388 {
7389         struct nfs4_lock_waiter *waiter =
7390                 container_of(wait, struct nfs4_lock_waiter, wait);
7391
7392         /* NULL key means to wake up everyone */
7393         if (key) {
7394                 struct cb_notify_lock_args      *cbnl = key;
7395                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7396                                                 *wowner = &waiter->owner;
7397
7398                 /* Only wake if the callback was for the same owner. */
7399                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7400                         return 0;
7401
7402                 /* Make sure it's for the right inode */
7403                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7404                         return 0;
7405         }
7406
7407         return woken_wake_function(wait, mode, flags, key);
7408 }
7409
7410 static int
7411 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7412 {
7413         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7414         struct nfs_server *server = NFS_SERVER(state->inode);
7415         struct nfs_client *clp = server->nfs_client;
7416         wait_queue_head_t *q = &clp->cl_lock_waitq;
7417         struct nfs4_lock_waiter waiter = {
7418                 .inode = state->inode,
7419                 .owner = { .clientid = clp->cl_clientid,
7420                            .id = lsp->ls_seqid.owner_id,
7421                            .s_dev = server->s_dev },
7422         };
7423         int status;
7424
7425         /* Don't bother with waitqueue if we don't expect a callback */
7426         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7427                 return nfs4_retry_setlk_simple(state, cmd, request);
7428
7429         init_wait(&waiter.wait);
7430         waiter.wait.func = nfs4_wake_lock_waiter;
7431         add_wait_queue(q, &waiter.wait);
7432
7433         do {
7434                 status = nfs4_proc_setlk(state, cmd, request);
7435                 if (status != -EAGAIN || IS_SETLK(cmd))
7436                         break;
7437
7438                 status = -ERESTARTSYS;
7439                 freezer_do_not_count();
7440                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE,
7441                            NFS4_LOCK_MAXTIMEOUT);
7442                 freezer_count();
7443         } while (!signalled());
7444
7445         remove_wait_queue(q, &waiter.wait);
7446
7447         return status;
7448 }
7449 #else /* !CONFIG_NFS_V4_1 */
7450 static inline int
7451 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7452 {
7453         return nfs4_retry_setlk_simple(state, cmd, request);
7454 }
7455 #endif
7456
7457 static int
7458 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7459 {
7460         struct nfs_open_context *ctx;
7461         struct nfs4_state *state;
7462         int status;
7463
7464         /* verify open state */
7465         ctx = nfs_file_open_context(filp);
7466         state = ctx->state;
7467
7468         if (IS_GETLK(cmd)) {
7469                 if (state != NULL)
7470                         return nfs4_proc_getlk(state, F_GETLK, request);
7471                 return 0;
7472         }
7473
7474         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7475                 return -EINVAL;
7476
7477         if (request->fl_type == F_UNLCK) {
7478                 if (state != NULL)
7479                         return nfs4_proc_unlck(state, cmd, request);
7480                 return 0;
7481         }
7482
7483         if (state == NULL)
7484                 return -ENOLCK;
7485
7486         if ((request->fl_flags & FL_POSIX) &&
7487             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7488                 return -ENOLCK;
7489
7490         /*
7491          * Don't rely on the VFS having checked the file open mode,
7492          * since it won't do this for flock() locks.
7493          */
7494         switch (request->fl_type) {
7495         case F_RDLCK:
7496                 if (!(filp->f_mode & FMODE_READ))
7497                         return -EBADF;
7498                 break;
7499         case F_WRLCK:
7500                 if (!(filp->f_mode & FMODE_WRITE))
7501                         return -EBADF;
7502         }
7503
7504         status = nfs4_set_lock_state(state, request);
7505         if (status != 0)
7506                 return status;
7507
7508         return nfs4_retry_setlk(state, cmd, request);
7509 }
7510
7511 static int nfs4_delete_lease(struct file *file, void **priv)
7512 {
7513         return generic_setlease(file, F_UNLCK, NULL, priv);
7514 }
7515
7516 static int nfs4_add_lease(struct file *file, long arg, struct file_lock **lease,
7517                           void **priv)
7518 {
7519         struct inode *inode = file_inode(file);
7520         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7521         int ret;
7522
7523         /* No delegation, no lease */
7524         if (!nfs4_have_delegation(inode, type))
7525                 return -EAGAIN;
7526         ret = generic_setlease(file, arg, lease, priv);
7527         if (ret || nfs4_have_delegation(inode, type))
7528                 return ret;
7529         /* We raced with a delegation return */
7530         nfs4_delete_lease(file, priv);
7531         return -EAGAIN;
7532 }
7533
7534 int nfs4_proc_setlease(struct file *file, long arg, struct file_lock **lease,
7535                        void **priv)
7536 {
7537         switch (arg) {
7538         case F_RDLCK:
7539         case F_WRLCK:
7540                 return nfs4_add_lease(file, arg, lease, priv);
7541         case F_UNLCK:
7542                 return nfs4_delete_lease(file, priv);
7543         default:
7544                 return -EINVAL;
7545         }
7546 }
7547
7548 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7549 {
7550         struct nfs_server *server = NFS_SERVER(state->inode);
7551         int err;
7552
7553         err = nfs4_set_lock_state(state, fl);
7554         if (err != 0)
7555                 return err;
7556         do {
7557                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7558                 if (err != -NFS4ERR_DELAY)
7559                         break;
7560                 ssleep(1);
7561         } while (err == -NFS4ERR_DELAY);
7562         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7563 }
7564
7565 struct nfs_release_lockowner_data {
7566         struct nfs4_lock_state *lsp;
7567         struct nfs_server *server;
7568         struct nfs_release_lockowner_args args;
7569         struct nfs_release_lockowner_res res;
7570         unsigned long timestamp;
7571 };
7572
7573 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7574 {
7575         struct nfs_release_lockowner_data *data = calldata;
7576         struct nfs_server *server = data->server;
7577         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7578                            &data->res.seq_res, task);
7579         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7580         data->timestamp = jiffies;
7581 }
7582
7583 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7584 {
7585         struct nfs_release_lockowner_data *data = calldata;
7586         struct nfs_server *server = data->server;
7587
7588         nfs40_sequence_done(task, &data->res.seq_res);
7589
7590         switch (task->tk_status) {
7591         case 0:
7592                 renew_lease(server, data->timestamp);
7593                 break;
7594         case -NFS4ERR_STALE_CLIENTID:
7595         case -NFS4ERR_EXPIRED:
7596                 nfs4_schedule_lease_recovery(server->nfs_client);
7597                 break;
7598         case -NFS4ERR_LEASE_MOVED:
7599         case -NFS4ERR_DELAY:
7600                 if (nfs4_async_handle_error(task, server,
7601                                             NULL, NULL) == -EAGAIN)
7602                         rpc_restart_call_prepare(task);
7603         }
7604 }
7605
7606 static void nfs4_release_lockowner_release(void *calldata)
7607 {
7608         struct nfs_release_lockowner_data *data = calldata;
7609         nfs4_free_lock_state(data->server, data->lsp);
7610         kfree(calldata);
7611 }
7612
7613 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7614         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7615         .rpc_call_done = nfs4_release_lockowner_done,
7616         .rpc_release = nfs4_release_lockowner_release,
7617 };
7618
7619 static void
7620 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7621 {
7622         struct nfs_release_lockowner_data *data;
7623         struct rpc_message msg = {
7624                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7625         };
7626
7627         if (server->nfs_client->cl_mvops->minor_version != 0)
7628                 return;
7629
7630         data = kmalloc(sizeof(*data), GFP_NOFS);
7631         if (!data)
7632                 return;
7633         data->lsp = lsp;
7634         data->server = server;
7635         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7636         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7637         data->args.lock_owner.s_dev = server->s_dev;
7638
7639         msg.rpc_argp = &data->args;
7640         msg.rpc_resp = &data->res;
7641         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7642         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7643 }
7644
7645 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7646
7647 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7648                                    struct user_namespace *mnt_userns,
7649                                    struct dentry *unused, struct inode *inode,
7650                                    const char *key, const void *buf,
7651                                    size_t buflen, int flags)
7652 {
7653         return nfs4_proc_set_acl(inode, buf, buflen);
7654 }
7655
7656 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7657                                    struct dentry *unused, struct inode *inode,
7658                                    const char *key, void *buf, size_t buflen)
7659 {
7660         return nfs4_proc_get_acl(inode, buf, buflen);
7661 }
7662
7663 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7664 {
7665         return nfs4_server_supports_acls(NFS_SERVER(d_inode(dentry)));
7666 }
7667
7668 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7669
7670 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7671                                      struct user_namespace *mnt_userns,
7672                                      struct dentry *unused, struct inode *inode,
7673                                      const char *key, const void *buf,
7674                                      size_t buflen, int flags)
7675 {
7676         if (security_ismaclabel(key))
7677                 return nfs4_set_security_label(inode, buf, buflen);
7678
7679         return -EOPNOTSUPP;
7680 }
7681
7682 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7683                                      struct dentry *unused, struct inode *inode,
7684                                      const char *key, void *buf, size_t buflen)
7685 {
7686         if (security_ismaclabel(key))
7687                 return nfs4_get_security_label(inode, buf, buflen);
7688         return -EOPNOTSUPP;
7689 }
7690
7691 static ssize_t
7692 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7693 {
7694         int len = 0;
7695
7696         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7697                 len = security_inode_listsecurity(inode, list, list_len);
7698                 if (len >= 0 && list_len && len > list_len)
7699                         return -ERANGE;
7700         }
7701         return len;
7702 }
7703
7704 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7705         .prefix = XATTR_SECURITY_PREFIX,
7706         .get    = nfs4_xattr_get_nfs4_label,
7707         .set    = nfs4_xattr_set_nfs4_label,
7708 };
7709
7710 #else
7711
7712 static ssize_t
7713 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7714 {
7715         return 0;
7716 }
7717
7718 #endif
7719
7720 #ifdef CONFIG_NFS_V4_2
7721 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7722                                     struct user_namespace *mnt_userns,
7723                                     struct dentry *unused, struct inode *inode,
7724                                     const char *key, const void *buf,
7725                                     size_t buflen, int flags)
7726 {
7727         u32 mask;
7728         int ret;
7729
7730         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7731                 return -EOPNOTSUPP;
7732
7733         /*
7734          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7735          * flags right now. Handling of xattr operations use the normal
7736          * file read/write permissions.
7737          *
7738          * Just in case the server has other ideas (which RFC 8276 allows),
7739          * do a cached access check for the XA* flags to possibly avoid
7740          * doing an RPC and getting EACCES back.
7741          */
7742         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7743                 if (!(mask & NFS_ACCESS_XAWRITE))
7744                         return -EACCES;
7745         }
7746
7747         if (buf == NULL) {
7748                 ret = nfs42_proc_removexattr(inode, key);
7749                 if (!ret)
7750                         nfs4_xattr_cache_remove(inode, key);
7751         } else {
7752                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7753                 if (!ret)
7754                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7755         }
7756
7757         return ret;
7758 }
7759
7760 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7761                                     struct dentry *unused, struct inode *inode,
7762                                     const char *key, void *buf, size_t buflen)
7763 {
7764         u32 mask;
7765         ssize_t ret;
7766
7767         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7768                 return -EOPNOTSUPP;
7769
7770         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7771                 if (!(mask & NFS_ACCESS_XAREAD))
7772                         return -EACCES;
7773         }
7774
7775         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7776         if (ret)
7777                 return ret;
7778
7779         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7780         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7781                 return ret;
7782
7783         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7784
7785         return ret;
7786 }
7787
7788 static ssize_t
7789 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7790 {
7791         u64 cookie;
7792         bool eof;
7793         ssize_t ret, size;
7794         char *buf;
7795         size_t buflen;
7796         u32 mask;
7797
7798         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7799                 return 0;
7800
7801         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7802                 if (!(mask & NFS_ACCESS_XALIST))
7803                         return 0;
7804         }
7805
7806         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7807         if (ret)
7808                 return ret;
7809
7810         ret = nfs4_xattr_cache_list(inode, list, list_len);
7811         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7812                 return ret;
7813
7814         cookie = 0;
7815         eof = false;
7816         buflen = list_len ? list_len : XATTR_LIST_MAX;
7817         buf = list_len ? list : NULL;
7818         size = 0;
7819
7820         while (!eof) {
7821                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
7822                     &cookie, &eof);
7823                 if (ret < 0)
7824                         return ret;
7825
7826                 if (list_len) {
7827                         buf += ret;
7828                         buflen -= ret;
7829                 }
7830                 size += ret;
7831         }
7832
7833         if (list_len)
7834                 nfs4_xattr_cache_set_list(inode, list, size);
7835
7836         return size;
7837 }
7838
7839 #else
7840
7841 static ssize_t
7842 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7843 {
7844         return 0;
7845 }
7846 #endif /* CONFIG_NFS_V4_2 */
7847
7848 /*
7849  * nfs_fhget will use either the mounted_on_fileid or the fileid
7850  */
7851 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7852 {
7853         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7854                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7855               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7856               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7857                 return;
7858
7859         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7860                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7861         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7862         fattr->nlink = 2;
7863 }
7864
7865 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7866                                    const struct qstr *name,
7867                                    struct nfs4_fs_locations *fs_locations,
7868                                    struct page *page)
7869 {
7870         struct nfs_server *server = NFS_SERVER(dir);
7871         u32 bitmask[3];
7872         struct nfs4_fs_locations_arg args = {
7873                 .dir_fh = NFS_FH(dir),
7874                 .name = name,
7875                 .page = page,
7876                 .bitmask = bitmask,
7877         };
7878         struct nfs4_fs_locations_res res = {
7879                 .fs_locations = fs_locations,
7880         };
7881         struct rpc_message msg = {
7882                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7883                 .rpc_argp = &args,
7884                 .rpc_resp = &res,
7885         };
7886         int status;
7887
7888         dprintk("%s: start\n", __func__);
7889
7890         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
7891         bitmask[1] = nfs4_fattr_bitmap[1];
7892
7893         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
7894          * is not supported */
7895         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
7896                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
7897         else
7898                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
7899
7900         nfs_fattr_init(&fs_locations->fattr);
7901         fs_locations->server = server;
7902         fs_locations->nlocations = 0;
7903         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
7904         dprintk("%s: returned status = %d\n", __func__, status);
7905         return status;
7906 }
7907
7908 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
7909                            const struct qstr *name,
7910                            struct nfs4_fs_locations *fs_locations,
7911                            struct page *page)
7912 {
7913         struct nfs4_exception exception = {
7914                 .interruptible = true,
7915         };
7916         int err;
7917         do {
7918                 err = _nfs4_proc_fs_locations(client, dir, name,
7919                                 fs_locations, page);
7920                 trace_nfs4_get_fs_locations(dir, name, err);
7921                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
7922                                 &exception);
7923         } while (exception.retry);
7924         return err;
7925 }
7926
7927 /*
7928  * This operation also signals the server that this client is
7929  * performing migration recovery.  The server can stop returning
7930  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
7931  * appended to this compound to identify the client ID which is
7932  * performing recovery.
7933  */
7934 static int _nfs40_proc_get_locations(struct nfs_server *server,
7935                                      struct nfs_fh *fhandle,
7936                                      struct nfs4_fs_locations *locations,
7937                                      struct page *page, const struct cred *cred)
7938 {
7939         struct rpc_clnt *clnt = server->client;
7940         u32 bitmask[2] = {
7941                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7942         };
7943         struct nfs4_fs_locations_arg args = {
7944                 .clientid       = server->nfs_client->cl_clientid,
7945                 .fh             = fhandle,
7946                 .page           = page,
7947                 .bitmask        = bitmask,
7948                 .migration      = 1,            /* skip LOOKUP */
7949                 .renew          = 1,            /* append RENEW */
7950         };
7951         struct nfs4_fs_locations_res res = {
7952                 .fs_locations   = locations,
7953                 .migration      = 1,
7954                 .renew          = 1,
7955         };
7956         struct rpc_message msg = {
7957                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
7958                 .rpc_argp       = &args,
7959                 .rpc_resp       = &res,
7960                 .rpc_cred       = cred,
7961         };
7962         unsigned long now = jiffies;
7963         int status;
7964
7965         nfs_fattr_init(&locations->fattr);
7966         locations->server = server;
7967         locations->nlocations = 0;
7968
7969         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
7970         status = nfs4_call_sync_sequence(clnt, server, &msg,
7971                                         &args.seq_args, &res.seq_res);
7972         if (status)
7973                 return status;
7974
7975         renew_lease(server, now);
7976         return 0;
7977 }
7978
7979 #ifdef CONFIG_NFS_V4_1
7980
7981 /*
7982  * This operation also signals the server that this client is
7983  * performing migration recovery.  The server can stop asserting
7984  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
7985  * performing this operation is identified in the SEQUENCE
7986  * operation in this compound.
7987  *
7988  * When the client supports GETATTR(fs_locations_info), it can
7989  * be plumbed in here.
7990  */
7991 static int _nfs41_proc_get_locations(struct nfs_server *server,
7992                                      struct nfs_fh *fhandle,
7993                                      struct nfs4_fs_locations *locations,
7994                                      struct page *page, const struct cred *cred)
7995 {
7996         struct rpc_clnt *clnt = server->client;
7997         u32 bitmask[2] = {
7998                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
7999         };
8000         struct nfs4_fs_locations_arg args = {
8001                 .fh             = fhandle,
8002                 .page           = page,
8003                 .bitmask        = bitmask,
8004                 .migration      = 1,            /* skip LOOKUP */
8005         };
8006         struct nfs4_fs_locations_res res = {
8007                 .fs_locations   = locations,
8008                 .migration      = 1,
8009         };
8010         struct rpc_message msg = {
8011                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8012                 .rpc_argp       = &args,
8013                 .rpc_resp       = &res,
8014                 .rpc_cred       = cred,
8015         };
8016         int status;
8017
8018         nfs_fattr_init(&locations->fattr);
8019         locations->server = server;
8020         locations->nlocations = 0;
8021
8022         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8023         status = nfs4_call_sync_sequence(clnt, server, &msg,
8024                                         &args.seq_args, &res.seq_res);
8025         if (status == NFS4_OK &&
8026             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8027                 status = -NFS4ERR_LEASE_MOVED;
8028         return status;
8029 }
8030
8031 #endif  /* CONFIG_NFS_V4_1 */
8032
8033 /**
8034  * nfs4_proc_get_locations - discover locations for a migrated FSID
8035  * @server: pointer to nfs_server to process
8036  * @fhandle: pointer to the kernel NFS client file handle
8037  * @locations: result of query
8038  * @page: buffer
8039  * @cred: credential to use for this operation
8040  *
8041  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8042  * operation failed, or a negative errno if a local error occurred.
8043  *
8044  * On success, "locations" is filled in, but if the server has
8045  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8046  * asserted.
8047  *
8048  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8049  * from this client that require migration recovery.
8050  */
8051 int nfs4_proc_get_locations(struct nfs_server *server,
8052                             struct nfs_fh *fhandle,
8053                             struct nfs4_fs_locations *locations,
8054                             struct page *page, const struct cred *cred)
8055 {
8056         struct nfs_client *clp = server->nfs_client;
8057         const struct nfs4_mig_recovery_ops *ops =
8058                                         clp->cl_mvops->mig_recovery_ops;
8059         struct nfs4_exception exception = {
8060                 .interruptible = true,
8061         };
8062         int status;
8063
8064         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8065                 (unsigned long long)server->fsid.major,
8066                 (unsigned long long)server->fsid.minor,
8067                 clp->cl_hostname);
8068         nfs_display_fhandle(fhandle, __func__);
8069
8070         do {
8071                 status = ops->get_locations(server, fhandle, locations, page,
8072                                             cred);
8073                 if (status != -NFS4ERR_DELAY)
8074                         break;
8075                 nfs4_handle_exception(server, status, &exception);
8076         } while (exception.retry);
8077         return status;
8078 }
8079
8080 /*
8081  * This operation also signals the server that this client is
8082  * performing "lease moved" recovery.  The server can stop
8083  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8084  * is appended to this compound to identify the client ID which is
8085  * performing recovery.
8086  */
8087 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8088 {
8089         struct nfs_server *server = NFS_SERVER(inode);
8090         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8091         struct rpc_clnt *clnt = server->client;
8092         struct nfs4_fsid_present_arg args = {
8093                 .fh             = NFS_FH(inode),
8094                 .clientid       = clp->cl_clientid,
8095                 .renew          = 1,            /* append RENEW */
8096         };
8097         struct nfs4_fsid_present_res res = {
8098                 .renew          = 1,
8099         };
8100         struct rpc_message msg = {
8101                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8102                 .rpc_argp       = &args,
8103                 .rpc_resp       = &res,
8104                 .rpc_cred       = cred,
8105         };
8106         unsigned long now = jiffies;
8107         int status;
8108
8109         res.fh = nfs_alloc_fhandle();
8110         if (res.fh == NULL)
8111                 return -ENOMEM;
8112
8113         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8114         status = nfs4_call_sync_sequence(clnt, server, &msg,
8115                                                 &args.seq_args, &res.seq_res);
8116         nfs_free_fhandle(res.fh);
8117         if (status)
8118                 return status;
8119
8120         do_renew_lease(clp, now);
8121         return 0;
8122 }
8123
8124 #ifdef CONFIG_NFS_V4_1
8125
8126 /*
8127  * This operation also signals the server that this client is
8128  * performing "lease moved" recovery.  The server can stop asserting
8129  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8130  * this operation is identified in the SEQUENCE operation in this
8131  * compound.
8132  */
8133 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8134 {
8135         struct nfs_server *server = NFS_SERVER(inode);
8136         struct rpc_clnt *clnt = server->client;
8137         struct nfs4_fsid_present_arg args = {
8138                 .fh             = NFS_FH(inode),
8139         };
8140         struct nfs4_fsid_present_res res = {
8141         };
8142         struct rpc_message msg = {
8143                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8144                 .rpc_argp       = &args,
8145                 .rpc_resp       = &res,
8146                 .rpc_cred       = cred,
8147         };
8148         int status;
8149
8150         res.fh = nfs_alloc_fhandle();
8151         if (res.fh == NULL)
8152                 return -ENOMEM;
8153
8154         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8155         status = nfs4_call_sync_sequence(clnt, server, &msg,
8156                                                 &args.seq_args, &res.seq_res);
8157         nfs_free_fhandle(res.fh);
8158         if (status == NFS4_OK &&
8159             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8160                 status = -NFS4ERR_LEASE_MOVED;
8161         return status;
8162 }
8163
8164 #endif  /* CONFIG_NFS_V4_1 */
8165
8166 /**
8167  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8168  * @inode: inode on FSID to check
8169  * @cred: credential to use for this operation
8170  *
8171  * Server indicates whether the FSID is present, moved, or not
8172  * recognized.  This operation is necessary to clear a LEASE_MOVED
8173  * condition for this client ID.
8174  *
8175  * Returns NFS4_OK if the FSID is present on this server,
8176  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8177  *  NFS4ERR code if some error occurred on the server, or a
8178  *  negative errno if a local failure occurred.
8179  */
8180 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8181 {
8182         struct nfs_server *server = NFS_SERVER(inode);
8183         struct nfs_client *clp = server->nfs_client;
8184         const struct nfs4_mig_recovery_ops *ops =
8185                                         clp->cl_mvops->mig_recovery_ops;
8186         struct nfs4_exception exception = {
8187                 .interruptible = true,
8188         };
8189         int status;
8190
8191         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8192                 (unsigned long long)server->fsid.major,
8193                 (unsigned long long)server->fsid.minor,
8194                 clp->cl_hostname);
8195         nfs_display_fhandle(NFS_FH(inode), __func__);
8196
8197         do {
8198                 status = ops->fsid_present(inode, cred);
8199                 if (status != -NFS4ERR_DELAY)
8200                         break;
8201                 nfs4_handle_exception(server, status, &exception);
8202         } while (exception.retry);
8203         return status;
8204 }
8205
8206 /*
8207  * If 'use_integrity' is true and the state managment nfs_client
8208  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8209  * and the machine credential as per RFC3530bis and RFC5661 Security
8210  * Considerations sections. Otherwise, just use the user cred with the
8211  * filesystem's rpc_client.
8212  */
8213 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8214 {
8215         int status;
8216         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8217         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8218         struct nfs4_secinfo_arg args = {
8219                 .dir_fh = NFS_FH(dir),
8220                 .name   = name,
8221         };
8222         struct nfs4_secinfo_res res = {
8223                 .flavors     = flavors,
8224         };
8225         struct rpc_message msg = {
8226                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8227                 .rpc_argp = &args,
8228                 .rpc_resp = &res,
8229         };
8230         struct nfs4_call_sync_data data = {
8231                 .seq_server = NFS_SERVER(dir),
8232                 .seq_args = &args.seq_args,
8233                 .seq_res = &res.seq_res,
8234         };
8235         struct rpc_task_setup task_setup = {
8236                 .rpc_client = clnt,
8237                 .rpc_message = &msg,
8238                 .callback_ops = clp->cl_mvops->call_sync_ops,
8239                 .callback_data = &data,
8240                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8241         };
8242         const struct cred *cred = NULL;
8243
8244         if (use_integrity) {
8245                 clnt = clp->cl_rpcclient;
8246                 task_setup.rpc_client = clnt;
8247
8248                 cred = nfs4_get_clid_cred(clp);
8249                 msg.rpc_cred = cred;
8250         }
8251
8252         dprintk("NFS call  secinfo %s\n", name->name);
8253
8254         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8255         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8256         status = nfs4_call_sync_custom(&task_setup);
8257
8258         dprintk("NFS reply  secinfo: %d\n", status);
8259
8260         put_cred(cred);
8261         return status;
8262 }
8263
8264 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8265                       struct nfs4_secinfo_flavors *flavors)
8266 {
8267         struct nfs4_exception exception = {
8268                 .interruptible = true,
8269         };
8270         int err;
8271         do {
8272                 err = -NFS4ERR_WRONGSEC;
8273
8274                 /* try to use integrity protection with machine cred */
8275                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8276                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8277
8278                 /*
8279                  * if unable to use integrity protection, or SECINFO with
8280                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8281                  * disallowed by spec, but exists in deployed servers) use
8282                  * the current filesystem's rpc_client and the user cred.
8283                  */
8284                 if (err == -NFS4ERR_WRONGSEC)
8285                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8286
8287                 trace_nfs4_secinfo(dir, name, err);
8288                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8289                                 &exception);
8290         } while (exception.retry);
8291         return err;
8292 }
8293
8294 #ifdef CONFIG_NFS_V4_1
8295 /*
8296  * Check the exchange flags returned by the server for invalid flags, having
8297  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8298  * DS flags set.
8299  */
8300 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8301 {
8302         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8303                 goto out_inval;
8304         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8305                 goto out_inval;
8306         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8307             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8308                 goto out_inval;
8309         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8310                 goto out_inval;
8311         return NFS_OK;
8312 out_inval:
8313         return -NFS4ERR_INVAL;
8314 }
8315
8316 static bool
8317 nfs41_same_server_scope(struct nfs41_server_scope *a,
8318                         struct nfs41_server_scope *b)
8319 {
8320         if (a->server_scope_sz != b->server_scope_sz)
8321                 return false;
8322         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8323 }
8324
8325 static void
8326 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8327 {
8328         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8329         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8330         struct nfs_client *clp = args->client;
8331
8332         switch (task->tk_status) {
8333         case -NFS4ERR_BADSESSION:
8334         case -NFS4ERR_DEADSESSION:
8335                 nfs4_schedule_session_recovery(clp->cl_session,
8336                                 task->tk_status);
8337         }
8338         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8339                         res->dir != NFS4_CDFS4_BOTH) {
8340                 rpc_task_close_connection(task);
8341                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8342                         rpc_restart_call(task);
8343         }
8344 }
8345
8346 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8347         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8348 };
8349
8350 /*
8351  * nfs4_proc_bind_one_conn_to_session()
8352  *
8353  * The 4.1 client currently uses the same TCP connection for the
8354  * fore and backchannel.
8355  */
8356 static
8357 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8358                 struct rpc_xprt *xprt,
8359                 struct nfs_client *clp,
8360                 const struct cred *cred)
8361 {
8362         int status;
8363         struct nfs41_bind_conn_to_session_args args = {
8364                 .client = clp,
8365                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8366                 .retries = 0,
8367         };
8368         struct nfs41_bind_conn_to_session_res res;
8369         struct rpc_message msg = {
8370                 .rpc_proc =
8371                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8372                 .rpc_argp = &args,
8373                 .rpc_resp = &res,
8374                 .rpc_cred = cred,
8375         };
8376         struct rpc_task_setup task_setup_data = {
8377                 .rpc_client = clnt,
8378                 .rpc_xprt = xprt,
8379                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8380                 .rpc_message = &msg,
8381                 .flags = RPC_TASK_TIMEOUT,
8382         };
8383         struct rpc_task *task;
8384
8385         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8386         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8387                 args.dir = NFS4_CDFC4_FORE;
8388
8389         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8390         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8391                 args.dir = NFS4_CDFC4_FORE;
8392
8393         task = rpc_run_task(&task_setup_data);
8394         if (!IS_ERR(task)) {
8395                 status = task->tk_status;
8396                 rpc_put_task(task);
8397         } else
8398                 status = PTR_ERR(task);
8399         trace_nfs4_bind_conn_to_session(clp, status);
8400         if (status == 0) {
8401                 if (memcmp(res.sessionid.data,
8402                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8403                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8404                         return -EIO;
8405                 }
8406                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8407                         dprintk("NFS: %s: Unexpected direction from server\n",
8408                                 __func__);
8409                         return -EIO;
8410                 }
8411                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8412                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8413                                 __func__);
8414                         return -EIO;
8415                 }
8416         }
8417
8418         return status;
8419 }
8420
8421 struct rpc_bind_conn_calldata {
8422         struct nfs_client *clp;
8423         const struct cred *cred;
8424 };
8425
8426 static int
8427 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8428                 struct rpc_xprt *xprt,
8429                 void *calldata)
8430 {
8431         struct rpc_bind_conn_calldata *p = calldata;
8432
8433         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8434 }
8435
8436 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8437 {
8438         struct rpc_bind_conn_calldata data = {
8439                 .clp = clp,
8440                 .cred = cred,
8441         };
8442         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8443                         nfs4_proc_bind_conn_to_session_callback, &data);
8444 }
8445
8446 /*
8447  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8448  * and operations we'd like to see to enable certain features in the allow map
8449  */
8450 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8451         .how = SP4_MACH_CRED,
8452         .enforce.u.words = {
8453                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8454                       1 << (OP_EXCHANGE_ID - 32) |
8455                       1 << (OP_CREATE_SESSION - 32) |
8456                       1 << (OP_DESTROY_SESSION - 32) |
8457                       1 << (OP_DESTROY_CLIENTID - 32)
8458         },
8459         .allow.u.words = {
8460                 [0] = 1 << (OP_CLOSE) |
8461                       1 << (OP_OPEN_DOWNGRADE) |
8462                       1 << (OP_LOCKU) |
8463                       1 << (OP_DELEGRETURN) |
8464                       1 << (OP_COMMIT),
8465                 [1] = 1 << (OP_SECINFO - 32) |
8466                       1 << (OP_SECINFO_NO_NAME - 32) |
8467                       1 << (OP_LAYOUTRETURN - 32) |
8468                       1 << (OP_TEST_STATEID - 32) |
8469                       1 << (OP_FREE_STATEID - 32) |
8470                       1 << (OP_WRITE - 32)
8471         }
8472 };
8473
8474 /*
8475  * Select the state protection mode for client `clp' given the server results
8476  * from exchange_id in `sp'.
8477  *
8478  * Returns 0 on success, negative errno otherwise.
8479  */
8480 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8481                                  struct nfs41_state_protection *sp)
8482 {
8483         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8484                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8485                       1 << (OP_EXCHANGE_ID - 32) |
8486                       1 << (OP_CREATE_SESSION - 32) |
8487                       1 << (OP_DESTROY_SESSION - 32) |
8488                       1 << (OP_DESTROY_CLIENTID - 32)
8489         };
8490         unsigned long flags = 0;
8491         unsigned int i;
8492         int ret = 0;
8493
8494         if (sp->how == SP4_MACH_CRED) {
8495                 /* Print state protect result */
8496                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8497                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8498                         if (test_bit(i, sp->enforce.u.longs))
8499                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8500                         if (test_bit(i, sp->allow.u.longs))
8501                                 dfprintk(MOUNT, "  allow op %d\n", i);
8502                 }
8503
8504                 /* make sure nothing is on enforce list that isn't supported */
8505                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8506                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8507                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8508                                 ret = -EINVAL;
8509                                 goto out;
8510                         }
8511                 }
8512
8513                 /*
8514                  * Minimal mode - state operations are allowed to use machine
8515                  * credential.  Note this already happens by default, so the
8516                  * client doesn't have to do anything more than the negotiation.
8517                  *
8518                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8519                  *       we're already using the machine cred for exchange_id
8520                  *       and will never use a different cred.
8521                  */
8522                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8523                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8524                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8525                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8526                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8527                         dfprintk(MOUNT, "  minimal mode enabled\n");
8528                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8529                 } else {
8530                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8531                         ret = -EINVAL;
8532                         goto out;
8533                 }
8534
8535                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8536                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8537                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8538                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8539                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8540                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8541                 }
8542
8543                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8544                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8545                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8546                 }
8547
8548                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8549                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8550                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8551                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8552                 }
8553
8554                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8555                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8556                         dfprintk(MOUNT, "  stateid mode enabled\n");
8557                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8558                 }
8559
8560                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8561                         dfprintk(MOUNT, "  write mode enabled\n");
8562                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8563                 }
8564
8565                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8566                         dfprintk(MOUNT, "  commit mode enabled\n");
8567                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8568                 }
8569         }
8570 out:
8571         clp->cl_sp4_flags = flags;
8572         return ret;
8573 }
8574
8575 struct nfs41_exchange_id_data {
8576         struct nfs41_exchange_id_res res;
8577         struct nfs41_exchange_id_args args;
8578 };
8579
8580 static void nfs4_exchange_id_release(void *data)
8581 {
8582         struct nfs41_exchange_id_data *cdata =
8583                                         (struct nfs41_exchange_id_data *)data;
8584
8585         nfs_put_client(cdata->args.client);
8586         kfree(cdata->res.impl_id);
8587         kfree(cdata->res.server_scope);
8588         kfree(cdata->res.server_owner);
8589         kfree(cdata);
8590 }
8591
8592 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8593         .rpc_release = nfs4_exchange_id_release,
8594 };
8595
8596 /*
8597  * _nfs4_proc_exchange_id()
8598  *
8599  * Wrapper for EXCHANGE_ID operation.
8600  */
8601 static struct rpc_task *
8602 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8603                         u32 sp4_how, struct rpc_xprt *xprt)
8604 {
8605         struct rpc_message msg = {
8606                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8607                 .rpc_cred = cred,
8608         };
8609         struct rpc_task_setup task_setup_data = {
8610                 .rpc_client = clp->cl_rpcclient,
8611                 .callback_ops = &nfs4_exchange_id_call_ops,
8612                 .rpc_message = &msg,
8613                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8614         };
8615         struct nfs41_exchange_id_data *calldata;
8616         int status;
8617
8618         if (!refcount_inc_not_zero(&clp->cl_count))
8619                 return ERR_PTR(-EIO);
8620
8621         status = -ENOMEM;
8622         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8623         if (!calldata)
8624                 goto out;
8625
8626         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8627
8628         status = nfs4_init_uniform_client_string(clp);
8629         if (status)
8630                 goto out_calldata;
8631
8632         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8633                                                 GFP_NOFS);
8634         status = -ENOMEM;
8635         if (unlikely(calldata->res.server_owner == NULL))
8636                 goto out_calldata;
8637
8638         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8639                                         GFP_NOFS);
8640         if (unlikely(calldata->res.server_scope == NULL))
8641                 goto out_server_owner;
8642
8643         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8644         if (unlikely(calldata->res.impl_id == NULL))
8645                 goto out_server_scope;
8646
8647         switch (sp4_how) {
8648         case SP4_NONE:
8649                 calldata->args.state_protect.how = SP4_NONE;
8650                 break;
8651
8652         case SP4_MACH_CRED:
8653                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8654                 break;
8655
8656         default:
8657                 /* unsupported! */
8658                 WARN_ON_ONCE(1);
8659                 status = -EINVAL;
8660                 goto out_impl_id;
8661         }
8662         if (xprt) {
8663                 task_setup_data.rpc_xprt = xprt;
8664                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8665                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8666                                 sizeof(calldata->args.verifier.data));
8667         }
8668         calldata->args.client = clp;
8669         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8670         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8671 #ifdef CONFIG_NFS_V4_1_MIGRATION
8672         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8673 #endif
8674         msg.rpc_argp = &calldata->args;
8675         msg.rpc_resp = &calldata->res;
8676         task_setup_data.callback_data = calldata;
8677
8678         return rpc_run_task(&task_setup_data);
8679
8680 out_impl_id:
8681         kfree(calldata->res.impl_id);
8682 out_server_scope:
8683         kfree(calldata->res.server_scope);
8684 out_server_owner:
8685         kfree(calldata->res.server_owner);
8686 out_calldata:
8687         kfree(calldata);
8688 out:
8689         nfs_put_client(clp);
8690         return ERR_PTR(status);
8691 }
8692
8693 /*
8694  * _nfs4_proc_exchange_id()
8695  *
8696  * Wrapper for EXCHANGE_ID operation.
8697  */
8698 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8699                         u32 sp4_how)
8700 {
8701         struct rpc_task *task;
8702         struct nfs41_exchange_id_args *argp;
8703         struct nfs41_exchange_id_res *resp;
8704         unsigned long now = jiffies;
8705         int status;
8706
8707         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8708         if (IS_ERR(task))
8709                 return PTR_ERR(task);
8710
8711         argp = task->tk_msg.rpc_argp;
8712         resp = task->tk_msg.rpc_resp;
8713         status = task->tk_status;
8714         if (status  != 0)
8715                 goto out;
8716
8717         status = nfs4_check_cl_exchange_flags(resp->flags,
8718                         clp->cl_mvops->minor_version);
8719         if (status  != 0)
8720                 goto out;
8721
8722         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8723         if (status != 0)
8724                 goto out;
8725
8726         do_renew_lease(clp, now);
8727
8728         clp->cl_clientid = resp->clientid;
8729         clp->cl_exchange_flags = resp->flags;
8730         clp->cl_seqid = resp->seqid;
8731         /* Client ID is not confirmed */
8732         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8733                 clear_bit(NFS4_SESSION_ESTABLISHED,
8734                           &clp->cl_session->session_state);
8735
8736         if (clp->cl_serverscope != NULL &&
8737             !nfs41_same_server_scope(clp->cl_serverscope,
8738                                 resp->server_scope)) {
8739                 dprintk("%s: server_scope mismatch detected\n",
8740                         __func__);
8741                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8742         }
8743
8744         swap(clp->cl_serverowner, resp->server_owner);
8745         swap(clp->cl_serverscope, resp->server_scope);
8746         swap(clp->cl_implid, resp->impl_id);
8747
8748         /* Save the EXCHANGE_ID verifier session trunk tests */
8749         memcpy(clp->cl_confirm.data, argp->verifier.data,
8750                sizeof(clp->cl_confirm.data));
8751 out:
8752         trace_nfs4_exchange_id(clp, status);
8753         rpc_put_task(task);
8754         return status;
8755 }
8756
8757 /*
8758  * nfs4_proc_exchange_id()
8759  *
8760  * Returns zero, a negative errno, or a negative NFS4ERR status code.
8761  *
8762  * Since the clientid has expired, all compounds using sessions
8763  * associated with the stale clientid will be returning
8764  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8765  * be in some phase of session reset.
8766  *
8767  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8768  */
8769 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8770 {
8771         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8772         int status;
8773
8774         /* try SP4_MACH_CRED if krb5i/p */
8775         if (authflavor == RPC_AUTH_GSS_KRB5I ||
8776             authflavor == RPC_AUTH_GSS_KRB5P) {
8777                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8778                 if (!status)
8779                         return 0;
8780         }
8781
8782         /* try SP4_NONE */
8783         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8784 }
8785
8786 /**
8787  * nfs4_test_session_trunk
8788  *
8789  * This is an add_xprt_test() test function called from
8790  * rpc_clnt_setup_test_and_add_xprt.
8791  *
8792  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8793  * and is dereferrenced in nfs4_exchange_id_release
8794  *
8795  * Upon success, add the new transport to the rpc_clnt
8796  *
8797  * @clnt: struct rpc_clnt to get new transport
8798  * @xprt: the rpc_xprt to test
8799  * @data: call data for _nfs4_proc_exchange_id.
8800  */
8801 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8802                             void *data)
8803 {
8804         struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
8805         struct rpc_task *task;
8806         int status;
8807
8808         u32 sp4_how;
8809
8810         dprintk("--> %s try %s\n", __func__,
8811                 xprt->address_strings[RPC_DISPLAY_ADDR]);
8812
8813         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8814
8815         /* Test connection for session trunking. Async exchange_id call */
8816         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8817         if (IS_ERR(task))
8818                 return;
8819
8820         status = task->tk_status;
8821         if (status == 0)
8822                 status = nfs4_detect_session_trunking(adata->clp,
8823                                 task->tk_msg.rpc_resp, xprt);
8824
8825         if (status == 0)
8826                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8827
8828         rpc_put_task(task);
8829 }
8830 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8831
8832 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8833                 const struct cred *cred)
8834 {
8835         struct rpc_message msg = {
8836                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8837                 .rpc_argp = clp,
8838                 .rpc_cred = cred,
8839         };
8840         int status;
8841
8842         status = rpc_call_sync(clp->cl_rpcclient, &msg,
8843                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
8844         trace_nfs4_destroy_clientid(clp, status);
8845         if (status)
8846                 dprintk("NFS: Got error %d from the server %s on "
8847                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
8848         return status;
8849 }
8850
8851 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
8852                 const struct cred *cred)
8853 {
8854         unsigned int loop;
8855         int ret;
8856
8857         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
8858                 ret = _nfs4_proc_destroy_clientid(clp, cred);
8859                 switch (ret) {
8860                 case -NFS4ERR_DELAY:
8861                 case -NFS4ERR_CLIENTID_BUSY:
8862                         ssleep(1);
8863                         break;
8864                 default:
8865                         return ret;
8866                 }
8867         }
8868         return 0;
8869 }
8870
8871 int nfs4_destroy_clientid(struct nfs_client *clp)
8872 {
8873         const struct cred *cred;
8874         int ret = 0;
8875
8876         if (clp->cl_mvops->minor_version < 1)
8877                 goto out;
8878         if (clp->cl_exchange_flags == 0)
8879                 goto out;
8880         if (clp->cl_preserve_clid)
8881                 goto out;
8882         cred = nfs4_get_clid_cred(clp);
8883         ret = nfs4_proc_destroy_clientid(clp, cred);
8884         put_cred(cred);
8885         switch (ret) {
8886         case 0:
8887         case -NFS4ERR_STALE_CLIENTID:
8888                 clp->cl_exchange_flags = 0;
8889         }
8890 out:
8891         return ret;
8892 }
8893
8894 #endif /* CONFIG_NFS_V4_1 */
8895
8896 struct nfs4_get_lease_time_data {
8897         struct nfs4_get_lease_time_args *args;
8898         struct nfs4_get_lease_time_res *res;
8899         struct nfs_client *clp;
8900 };
8901
8902 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
8903                                         void *calldata)
8904 {
8905         struct nfs4_get_lease_time_data *data =
8906                         (struct nfs4_get_lease_time_data *)calldata;
8907
8908         /* just setup sequence, do not trigger session recovery
8909            since we're invoked within one */
8910         nfs4_setup_sequence(data->clp,
8911                         &data->args->la_seq_args,
8912                         &data->res->lr_seq_res,
8913                         task);
8914 }
8915
8916 /*
8917  * Called from nfs4_state_manager thread for session setup, so don't recover
8918  * from sequence operation or clientid errors.
8919  */
8920 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
8921 {
8922         struct nfs4_get_lease_time_data *data =
8923                         (struct nfs4_get_lease_time_data *)calldata;
8924
8925         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
8926                 return;
8927         switch (task->tk_status) {
8928         case -NFS4ERR_DELAY:
8929         case -NFS4ERR_GRACE:
8930                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
8931                 task->tk_status = 0;
8932                 fallthrough;
8933         case -NFS4ERR_RETRY_UNCACHED_REP:
8934                 rpc_restart_call_prepare(task);
8935                 return;
8936         }
8937 }
8938
8939 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
8940         .rpc_call_prepare = nfs4_get_lease_time_prepare,
8941         .rpc_call_done = nfs4_get_lease_time_done,
8942 };
8943
8944 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
8945 {
8946         struct nfs4_get_lease_time_args args;
8947         struct nfs4_get_lease_time_res res = {
8948                 .lr_fsinfo = fsinfo,
8949         };
8950         struct nfs4_get_lease_time_data data = {
8951                 .args = &args,
8952                 .res = &res,
8953                 .clp = clp,
8954         };
8955         struct rpc_message msg = {
8956                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
8957                 .rpc_argp = &args,
8958                 .rpc_resp = &res,
8959         };
8960         struct rpc_task_setup task_setup = {
8961                 .rpc_client = clp->cl_rpcclient,
8962                 .rpc_message = &msg,
8963                 .callback_ops = &nfs4_get_lease_time_ops,
8964                 .callback_data = &data,
8965                 .flags = RPC_TASK_TIMEOUT,
8966         };
8967
8968         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
8969         return nfs4_call_sync_custom(&task_setup);
8970 }
8971
8972 #ifdef CONFIG_NFS_V4_1
8973
8974 /*
8975  * Initialize the values to be used by the client in CREATE_SESSION
8976  * If nfs4_init_session set the fore channel request and response sizes,
8977  * use them.
8978  *
8979  * Set the back channel max_resp_sz_cached to zero to force the client to
8980  * always set csa_cachethis to FALSE because the current implementation
8981  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
8982  */
8983 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
8984                                     struct rpc_clnt *clnt)
8985 {
8986         unsigned int max_rqst_sz, max_resp_sz;
8987         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
8988         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
8989
8990         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
8991         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
8992
8993         /* Fore channel attributes */
8994         args->fc_attrs.max_rqst_sz = max_rqst_sz;
8995         args->fc_attrs.max_resp_sz = max_resp_sz;
8996         args->fc_attrs.max_ops = NFS4_MAX_OPS;
8997         args->fc_attrs.max_reqs = max_session_slots;
8998
8999         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9000                 "max_ops=%u max_reqs=%u\n",
9001                 __func__,
9002                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9003                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9004
9005         /* Back channel attributes */
9006         args->bc_attrs.max_rqst_sz = max_bc_payload;
9007         args->bc_attrs.max_resp_sz = max_bc_payload;
9008         args->bc_attrs.max_resp_sz_cached = 0;
9009         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9010         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9011         if (args->bc_attrs.max_reqs > max_bc_slots)
9012                 args->bc_attrs.max_reqs = max_bc_slots;
9013
9014         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9015                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9016                 __func__,
9017                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9018                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9019                 args->bc_attrs.max_reqs);
9020 }
9021
9022 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9023                 struct nfs41_create_session_res *res)
9024 {
9025         struct nfs4_channel_attrs *sent = &args->fc_attrs;
9026         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9027
9028         if (rcvd->max_resp_sz > sent->max_resp_sz)
9029                 return -EINVAL;
9030         /*
9031          * Our requested max_ops is the minimum we need; we're not
9032          * prepared to break up compounds into smaller pieces than that.
9033          * So, no point even trying to continue if the server won't
9034          * cooperate:
9035          */
9036         if (rcvd->max_ops < sent->max_ops)
9037                 return -EINVAL;
9038         if (rcvd->max_reqs == 0)
9039                 return -EINVAL;
9040         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9041                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9042         return 0;
9043 }
9044
9045 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9046                 struct nfs41_create_session_res *res)
9047 {
9048         struct nfs4_channel_attrs *sent = &args->bc_attrs;
9049         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9050
9051         if (!(res->flags & SESSION4_BACK_CHAN))
9052                 goto out;
9053         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9054                 return -EINVAL;
9055         if (rcvd->max_resp_sz < sent->max_resp_sz)
9056                 return -EINVAL;
9057         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9058                 return -EINVAL;
9059         if (rcvd->max_ops > sent->max_ops)
9060                 return -EINVAL;
9061         if (rcvd->max_reqs > sent->max_reqs)
9062                 return -EINVAL;
9063 out:
9064         return 0;
9065 }
9066
9067 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9068                                      struct nfs41_create_session_res *res)
9069 {
9070         int ret;
9071
9072         ret = nfs4_verify_fore_channel_attrs(args, res);
9073         if (ret)
9074                 return ret;
9075         return nfs4_verify_back_channel_attrs(args, res);
9076 }
9077
9078 static void nfs4_update_session(struct nfs4_session *session,
9079                 struct nfs41_create_session_res *res)
9080 {
9081         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9082         /* Mark client id and session as being confirmed */
9083         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9084         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9085         session->flags = res->flags;
9086         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9087         if (res->flags & SESSION4_BACK_CHAN)
9088                 memcpy(&session->bc_attrs, &res->bc_attrs,
9089                                 sizeof(session->bc_attrs));
9090 }
9091
9092 static int _nfs4_proc_create_session(struct nfs_client *clp,
9093                 const struct cred *cred)
9094 {
9095         struct nfs4_session *session = clp->cl_session;
9096         struct nfs41_create_session_args args = {
9097                 .client = clp,
9098                 .clientid = clp->cl_clientid,
9099                 .seqid = clp->cl_seqid,
9100                 .cb_program = NFS4_CALLBACK,
9101         };
9102         struct nfs41_create_session_res res;
9103
9104         struct rpc_message msg = {
9105                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9106                 .rpc_argp = &args,
9107                 .rpc_resp = &res,
9108                 .rpc_cred = cred,
9109         };
9110         int status;
9111
9112         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9113         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9114
9115         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9116                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9117         trace_nfs4_create_session(clp, status);
9118
9119         switch (status) {
9120         case -NFS4ERR_STALE_CLIENTID:
9121         case -NFS4ERR_DELAY:
9122         case -ETIMEDOUT:
9123         case -EACCES:
9124         case -EAGAIN:
9125                 goto out;
9126         }
9127
9128         clp->cl_seqid++;
9129         if (!status) {
9130                 /* Verify the session's negotiated channel_attrs values */
9131                 status = nfs4_verify_channel_attrs(&args, &res);
9132                 /* Increment the clientid slot sequence id */
9133                 if (status)
9134                         goto out;
9135                 nfs4_update_session(session, &res);
9136         }
9137 out:
9138         return status;
9139 }
9140
9141 /*
9142  * Issues a CREATE_SESSION operation to the server.
9143  * It is the responsibility of the caller to verify the session is
9144  * expired before calling this routine.
9145  */
9146 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9147 {
9148         int status;
9149         unsigned *ptr;
9150         struct nfs4_session *session = clp->cl_session;
9151
9152         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9153
9154         status = _nfs4_proc_create_session(clp, cred);
9155         if (status)
9156                 goto out;
9157
9158         /* Init or reset the session slot tables */
9159         status = nfs4_setup_session_slot_tables(session);
9160         dprintk("slot table setup returned %d\n", status);
9161         if (status)
9162                 goto out;
9163
9164         ptr = (unsigned *)&session->sess_id.data[0];
9165         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9166                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9167 out:
9168         return status;
9169 }
9170
9171 /*
9172  * Issue the over-the-wire RPC DESTROY_SESSION.
9173  * The caller must serialize access to this routine.
9174  */
9175 int nfs4_proc_destroy_session(struct nfs4_session *session,
9176                 const struct cred *cred)
9177 {
9178         struct rpc_message msg = {
9179                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9180                 .rpc_argp = session,
9181                 .rpc_cred = cred,
9182         };
9183         int status = 0;
9184
9185         /* session is still being setup */
9186         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9187                 return 0;
9188
9189         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9190                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9191         trace_nfs4_destroy_session(session->clp, status);
9192
9193         if (status)
9194                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9195                         "Session has been destroyed regardless...\n", status);
9196         return status;
9197 }
9198
9199 /*
9200  * Renew the cl_session lease.
9201  */
9202 struct nfs4_sequence_data {
9203         struct nfs_client *clp;
9204         struct nfs4_sequence_args args;
9205         struct nfs4_sequence_res res;
9206 };
9207
9208 static void nfs41_sequence_release(void *data)
9209 {
9210         struct nfs4_sequence_data *calldata = data;
9211         struct nfs_client *clp = calldata->clp;
9212
9213         if (refcount_read(&clp->cl_count) > 1)
9214                 nfs4_schedule_state_renewal(clp);
9215         nfs_put_client(clp);
9216         kfree(calldata);
9217 }
9218
9219 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9220 {
9221         switch(task->tk_status) {
9222         case -NFS4ERR_DELAY:
9223                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9224                 return -EAGAIN;
9225         default:
9226                 nfs4_schedule_lease_recovery(clp);
9227         }
9228         return 0;
9229 }
9230
9231 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9232 {
9233         struct nfs4_sequence_data *calldata = data;
9234         struct nfs_client *clp = calldata->clp;
9235
9236         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9237                 return;
9238
9239         trace_nfs4_sequence(clp, task->tk_status);
9240         if (task->tk_status < 0) {
9241                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9242                 if (refcount_read(&clp->cl_count) == 1)
9243                         return;
9244
9245                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9246                         rpc_restart_call_prepare(task);
9247                         return;
9248                 }
9249         }
9250         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9251 }
9252
9253 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9254 {
9255         struct nfs4_sequence_data *calldata = data;
9256         struct nfs_client *clp = calldata->clp;
9257         struct nfs4_sequence_args *args;
9258         struct nfs4_sequence_res *res;
9259
9260         args = task->tk_msg.rpc_argp;
9261         res = task->tk_msg.rpc_resp;
9262
9263         nfs4_setup_sequence(clp, args, res, task);
9264 }
9265
9266 static const struct rpc_call_ops nfs41_sequence_ops = {
9267         .rpc_call_done = nfs41_sequence_call_done,
9268         .rpc_call_prepare = nfs41_sequence_prepare,
9269         .rpc_release = nfs41_sequence_release,
9270 };
9271
9272 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9273                 const struct cred *cred,
9274                 struct nfs4_slot *slot,
9275                 bool is_privileged)
9276 {
9277         struct nfs4_sequence_data *calldata;
9278         struct rpc_message msg = {
9279                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9280                 .rpc_cred = cred,
9281         };
9282         struct rpc_task_setup task_setup_data = {
9283                 .rpc_client = clp->cl_rpcclient,
9284                 .rpc_message = &msg,
9285                 .callback_ops = &nfs41_sequence_ops,
9286                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9287         };
9288         struct rpc_task *ret;
9289
9290         ret = ERR_PTR(-EIO);
9291         if (!refcount_inc_not_zero(&clp->cl_count))
9292                 goto out_err;
9293
9294         ret = ERR_PTR(-ENOMEM);
9295         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9296         if (calldata == NULL)
9297                 goto out_put_clp;
9298         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9299         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9300         msg.rpc_argp = &calldata->args;
9301         msg.rpc_resp = &calldata->res;
9302         calldata->clp = clp;
9303         task_setup_data.callback_data = calldata;
9304
9305         ret = rpc_run_task(&task_setup_data);
9306         if (IS_ERR(ret))
9307                 goto out_err;
9308         return ret;
9309 out_put_clp:
9310         nfs_put_client(clp);
9311 out_err:
9312         nfs41_release_slot(slot);
9313         return ret;
9314 }
9315
9316 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9317 {
9318         struct rpc_task *task;
9319         int ret = 0;
9320
9321         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9322                 return -EAGAIN;
9323         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9324         if (IS_ERR(task))
9325                 ret = PTR_ERR(task);
9326         else
9327                 rpc_put_task_async(task);
9328         dprintk("<-- %s status=%d\n", __func__, ret);
9329         return ret;
9330 }
9331
9332 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9333 {
9334         struct rpc_task *task;
9335         int ret;
9336
9337         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9338         if (IS_ERR(task)) {
9339                 ret = PTR_ERR(task);
9340                 goto out;
9341         }
9342         ret = rpc_wait_for_completion_task(task);
9343         if (!ret)
9344                 ret = task->tk_status;
9345         rpc_put_task(task);
9346 out:
9347         dprintk("<-- %s status=%d\n", __func__, ret);
9348         return ret;
9349 }
9350
9351 struct nfs4_reclaim_complete_data {
9352         struct nfs_client *clp;
9353         struct nfs41_reclaim_complete_args arg;
9354         struct nfs41_reclaim_complete_res res;
9355 };
9356
9357 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9358 {
9359         struct nfs4_reclaim_complete_data *calldata = data;
9360
9361         nfs4_setup_sequence(calldata->clp,
9362                         &calldata->arg.seq_args,
9363                         &calldata->res.seq_res,
9364                         task);
9365 }
9366
9367 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9368 {
9369         switch(task->tk_status) {
9370         case 0:
9371                 wake_up_all(&clp->cl_lock_waitq);
9372                 fallthrough;
9373         case -NFS4ERR_COMPLETE_ALREADY:
9374         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9375                 break;
9376         case -NFS4ERR_DELAY:
9377                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9378                 fallthrough;
9379         case -NFS4ERR_RETRY_UNCACHED_REP:
9380                 return -EAGAIN;
9381         case -NFS4ERR_BADSESSION:
9382         case -NFS4ERR_DEADSESSION:
9383         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9384                 break;
9385         default:
9386                 nfs4_schedule_lease_recovery(clp);
9387         }
9388         return 0;
9389 }
9390
9391 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9392 {
9393         struct nfs4_reclaim_complete_data *calldata = data;
9394         struct nfs_client *clp = calldata->clp;
9395         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9396
9397         if (!nfs41_sequence_done(task, res))
9398                 return;
9399
9400         trace_nfs4_reclaim_complete(clp, task->tk_status);
9401         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9402                 rpc_restart_call_prepare(task);
9403                 return;
9404         }
9405 }
9406
9407 static void nfs4_free_reclaim_complete_data(void *data)
9408 {
9409         struct nfs4_reclaim_complete_data *calldata = data;
9410
9411         kfree(calldata);
9412 }
9413
9414 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9415         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9416         .rpc_call_done = nfs4_reclaim_complete_done,
9417         .rpc_release = nfs4_free_reclaim_complete_data,
9418 };
9419
9420 /*
9421  * Issue a global reclaim complete.
9422  */
9423 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9424                 const struct cred *cred)
9425 {
9426         struct nfs4_reclaim_complete_data *calldata;
9427         struct rpc_message msg = {
9428                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9429                 .rpc_cred = cred,
9430         };
9431         struct rpc_task_setup task_setup_data = {
9432                 .rpc_client = clp->cl_rpcclient,
9433                 .rpc_message = &msg,
9434                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9435                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9436         };
9437         int status = -ENOMEM;
9438
9439         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9440         if (calldata == NULL)
9441                 goto out;
9442         calldata->clp = clp;
9443         calldata->arg.one_fs = 0;
9444
9445         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9446         msg.rpc_argp = &calldata->arg;
9447         msg.rpc_resp = &calldata->res;
9448         task_setup_data.callback_data = calldata;
9449         status = nfs4_call_sync_custom(&task_setup_data);
9450 out:
9451         dprintk("<-- %s status=%d\n", __func__, status);
9452         return status;
9453 }
9454
9455 static void
9456 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9457 {
9458         struct nfs4_layoutget *lgp = calldata;
9459         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9460
9461         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9462                                 &lgp->res.seq_res, task);
9463 }
9464
9465 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9466 {
9467         struct nfs4_layoutget *lgp = calldata;
9468
9469         nfs41_sequence_process(task, &lgp->res.seq_res);
9470 }
9471
9472 static int
9473 nfs4_layoutget_handle_exception(struct rpc_task *task,
9474                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9475 {
9476         struct inode *inode = lgp->args.inode;
9477         struct nfs_server *server = NFS_SERVER(inode);
9478         struct pnfs_layout_hdr *lo = lgp->lo;
9479         int nfs4err = task->tk_status;
9480         int err, status = 0;
9481         LIST_HEAD(head);
9482
9483         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9484
9485         nfs4_sequence_free_slot(&lgp->res.seq_res);
9486
9487         switch (nfs4err) {
9488         case 0:
9489                 goto out;
9490
9491         /*
9492          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9493          * on the file. set tk_status to -ENODATA to tell upper layer to
9494          * retry go inband.
9495          */
9496         case -NFS4ERR_LAYOUTUNAVAILABLE:
9497                 status = -ENODATA;
9498                 goto out;
9499         /*
9500          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9501          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9502          */
9503         case -NFS4ERR_BADLAYOUT:
9504                 status = -EOVERFLOW;
9505                 goto out;
9506         /*
9507          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9508          * (or clients) writing to the same RAID stripe except when
9509          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9510          *
9511          * Treat it like we would RECALLCONFLICT -- we retry for a little
9512          * while, and then eventually give up.
9513          */
9514         case -NFS4ERR_LAYOUTTRYLATER:
9515                 if (lgp->args.minlength == 0) {
9516                         status = -EOVERFLOW;
9517                         goto out;
9518                 }
9519                 status = -EBUSY;
9520                 break;
9521         case -NFS4ERR_RECALLCONFLICT:
9522                 status = -ERECALLCONFLICT;
9523                 break;
9524         case -NFS4ERR_DELEG_REVOKED:
9525         case -NFS4ERR_ADMIN_REVOKED:
9526         case -NFS4ERR_EXPIRED:
9527         case -NFS4ERR_BAD_STATEID:
9528                 exception->timeout = 0;
9529                 spin_lock(&inode->i_lock);
9530                 /* If the open stateid was bad, then recover it. */
9531                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9532                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9533                         spin_unlock(&inode->i_lock);
9534                         exception->state = lgp->args.ctx->state;
9535                         exception->stateid = &lgp->args.stateid;
9536                         break;
9537                 }
9538
9539                 /*
9540                  * Mark the bad layout state as invalid, then retry
9541                  */
9542                 pnfs_mark_layout_stateid_invalid(lo, &head);
9543                 spin_unlock(&inode->i_lock);
9544                 nfs_commit_inode(inode, 0);
9545                 pnfs_free_lseg_list(&head);
9546                 status = -EAGAIN;
9547                 goto out;
9548         }
9549
9550         err = nfs4_handle_exception(server, nfs4err, exception);
9551         if (!status) {
9552                 if (exception->retry)
9553                         status = -EAGAIN;
9554                 else
9555                         status = err;
9556         }
9557 out:
9558         return status;
9559 }
9560
9561 size_t max_response_pages(struct nfs_server *server)
9562 {
9563         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9564         return nfs_page_array_len(0, max_resp_sz);
9565 }
9566
9567 static void nfs4_layoutget_release(void *calldata)
9568 {
9569         struct nfs4_layoutget *lgp = calldata;
9570
9571         nfs4_sequence_free_slot(&lgp->res.seq_res);
9572         pnfs_layoutget_free(lgp);
9573 }
9574
9575 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9576         .rpc_call_prepare = nfs4_layoutget_prepare,
9577         .rpc_call_done = nfs4_layoutget_done,
9578         .rpc_release = nfs4_layoutget_release,
9579 };
9580
9581 struct pnfs_layout_segment *
9582 nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout)
9583 {
9584         struct inode *inode = lgp->args.inode;
9585         struct nfs_server *server = NFS_SERVER(inode);
9586         struct rpc_task *task;
9587         struct rpc_message msg = {
9588                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9589                 .rpc_argp = &lgp->args,
9590                 .rpc_resp = &lgp->res,
9591                 .rpc_cred = lgp->cred,
9592         };
9593         struct rpc_task_setup task_setup_data = {
9594                 .rpc_client = server->client,
9595                 .rpc_message = &msg,
9596                 .callback_ops = &nfs4_layoutget_call_ops,
9597                 .callback_data = lgp,
9598                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9599                          RPC_TASK_MOVEABLE,
9600         };
9601         struct pnfs_layout_segment *lseg = NULL;
9602         struct nfs4_exception exception = {
9603                 .inode = inode,
9604                 .timeout = *timeout,
9605         };
9606         int status = 0;
9607
9608         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9609
9610         task = rpc_run_task(&task_setup_data);
9611
9612         status = rpc_wait_for_completion_task(task);
9613         if (status != 0)
9614                 goto out;
9615
9616         if (task->tk_status < 0) {
9617                 status = nfs4_layoutget_handle_exception(task, lgp, &exception);
9618                 *timeout = exception.timeout;
9619         } else if (lgp->res.layoutp->len == 0) {
9620                 status = -EAGAIN;
9621                 *timeout = nfs4_update_delay(&exception.timeout);
9622         } else
9623                 lseg = pnfs_layout_process(lgp);
9624 out:
9625         trace_nfs4_layoutget(lgp->args.ctx,
9626                         &lgp->args.range,
9627                         &lgp->res.range,
9628                         &lgp->res.stateid,
9629                         status);
9630
9631         rpc_put_task(task);
9632         dprintk("<-- %s status=%d\n", __func__, status);
9633         if (status)
9634                 return ERR_PTR(status);
9635         return lseg;
9636 }
9637
9638 static void
9639 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9640 {
9641         struct nfs4_layoutreturn *lrp = calldata;
9642
9643         nfs4_setup_sequence(lrp->clp,
9644                         &lrp->args.seq_args,
9645                         &lrp->res.seq_res,
9646                         task);
9647         if (!pnfs_layout_is_valid(lrp->args.layout))
9648                 rpc_exit(task, 0);
9649 }
9650
9651 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9652 {
9653         struct nfs4_layoutreturn *lrp = calldata;
9654         struct nfs_server *server;
9655
9656         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9657                 return;
9658
9659         /*
9660          * Was there an RPC level error? Assume the call succeeded,
9661          * and that we need to release the layout
9662          */
9663         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9664                 lrp->res.lrs_present = 0;
9665                 return;
9666         }
9667
9668         server = NFS_SERVER(lrp->args.inode);
9669         switch (task->tk_status) {
9670         case -NFS4ERR_OLD_STATEID:
9671                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9672                                         &lrp->args.range,
9673                                         lrp->args.inode))
9674                         goto out_restart;
9675                 fallthrough;
9676         default:
9677                 task->tk_status = 0;
9678                 fallthrough;
9679         case 0:
9680                 break;
9681         case -NFS4ERR_DELAY:
9682                 if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9683                         break;
9684                 goto out_restart;
9685         }
9686         return;
9687 out_restart:
9688         task->tk_status = 0;
9689         nfs4_sequence_free_slot(&lrp->res.seq_res);
9690         rpc_restart_call_prepare(task);
9691 }
9692
9693 static void nfs4_layoutreturn_release(void *calldata)
9694 {
9695         struct nfs4_layoutreturn *lrp = calldata;
9696         struct pnfs_layout_hdr *lo = lrp->args.layout;
9697
9698         pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9699                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9700         nfs4_sequence_free_slot(&lrp->res.seq_res);
9701         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9702                 lrp->ld_private.ops->free(&lrp->ld_private);
9703         pnfs_put_layout_hdr(lrp->args.layout);
9704         nfs_iput_and_deactive(lrp->inode);
9705         put_cred(lrp->cred);
9706         kfree(calldata);
9707 }
9708
9709 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9710         .rpc_call_prepare = nfs4_layoutreturn_prepare,
9711         .rpc_call_done = nfs4_layoutreturn_done,
9712         .rpc_release = nfs4_layoutreturn_release,
9713 };
9714
9715 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9716 {
9717         struct rpc_task *task;
9718         struct rpc_message msg = {
9719                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9720                 .rpc_argp = &lrp->args,
9721                 .rpc_resp = &lrp->res,
9722                 .rpc_cred = lrp->cred,
9723         };
9724         struct rpc_task_setup task_setup_data = {
9725                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
9726                 .rpc_message = &msg,
9727                 .callback_ops = &nfs4_layoutreturn_call_ops,
9728                 .callback_data = lrp,
9729                 .flags = RPC_TASK_MOVEABLE,
9730         };
9731         int status = 0;
9732
9733         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9734                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9735                         &task_setup_data.rpc_client, &msg);
9736
9737         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9738         if (!sync) {
9739                 if (!lrp->inode) {
9740                         nfs4_layoutreturn_release(lrp);
9741                         return -EAGAIN;
9742                 }
9743                 task_setup_data.flags |= RPC_TASK_ASYNC;
9744         }
9745         if (!lrp->inode)
9746                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9747                                    1);
9748         else
9749                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9750                                    0);
9751         task = rpc_run_task(&task_setup_data);
9752         if (IS_ERR(task))
9753                 return PTR_ERR(task);
9754         if (sync)
9755                 status = task->tk_status;
9756         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9757         dprintk("<-- %s status=%d\n", __func__, status);
9758         rpc_put_task(task);
9759         return status;
9760 }
9761
9762 static int
9763 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
9764                 struct pnfs_device *pdev,
9765                 const struct cred *cred)
9766 {
9767         struct nfs4_getdeviceinfo_args args = {
9768                 .pdev = pdev,
9769                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
9770                         NOTIFY_DEVICEID4_DELETE,
9771         };
9772         struct nfs4_getdeviceinfo_res res = {
9773                 .pdev = pdev,
9774         };
9775         struct rpc_message msg = {
9776                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9777                 .rpc_argp = &args,
9778                 .rpc_resp = &res,
9779                 .rpc_cred = cred,
9780         };
9781         int status;
9782
9783         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9784         if (res.notification & ~args.notify_types)
9785                 dprintk("%s: unsupported notification\n", __func__);
9786         if (res.notification != args.notify_types)
9787                 pdev->nocache = 1;
9788
9789         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9790
9791         dprintk("<-- %s status=%d\n", __func__, status);
9792
9793         return status;
9794 }
9795
9796 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9797                 struct pnfs_device *pdev,
9798                 const struct cred *cred)
9799 {
9800         struct nfs4_exception exception = { };
9801         int err;
9802
9803         do {
9804                 err = nfs4_handle_exception(server,
9805                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
9806                                         &exception);
9807         } while (exception.retry);
9808         return err;
9809 }
9810 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9811
9812 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9813 {
9814         struct nfs4_layoutcommit_data *data = calldata;
9815         struct nfs_server *server = NFS_SERVER(data->args.inode);
9816
9817         nfs4_setup_sequence(server->nfs_client,
9818                         &data->args.seq_args,
9819                         &data->res.seq_res,
9820                         task);
9821 }
9822
9823 static void
9824 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
9825 {
9826         struct nfs4_layoutcommit_data *data = calldata;
9827         struct nfs_server *server = NFS_SERVER(data->args.inode);
9828
9829         if (!nfs41_sequence_done(task, &data->res.seq_res))
9830                 return;
9831
9832         switch (task->tk_status) { /* Just ignore these failures */
9833         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
9834         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
9835         case -NFS4ERR_BADLAYOUT:     /* no layout */
9836         case -NFS4ERR_GRACE:        /* loca_recalim always false */
9837                 task->tk_status = 0;
9838                 break;
9839         case 0:
9840                 break;
9841         default:
9842                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
9843                         rpc_restart_call_prepare(task);
9844                         return;
9845                 }
9846         }
9847 }
9848
9849 static void nfs4_layoutcommit_release(void *calldata)
9850 {
9851         struct nfs4_layoutcommit_data *data = calldata;
9852
9853         pnfs_cleanup_layoutcommit(data);
9854         nfs_post_op_update_inode_force_wcc(data->args.inode,
9855                                            data->res.fattr);
9856         put_cred(data->cred);
9857         nfs_iput_and_deactive(data->inode);
9858         kfree(data);
9859 }
9860
9861 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
9862         .rpc_call_prepare = nfs4_layoutcommit_prepare,
9863         .rpc_call_done = nfs4_layoutcommit_done,
9864         .rpc_release = nfs4_layoutcommit_release,
9865 };
9866
9867 int
9868 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
9869 {
9870         struct rpc_message msg = {
9871                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
9872                 .rpc_argp = &data->args,
9873                 .rpc_resp = &data->res,
9874                 .rpc_cred = data->cred,
9875         };
9876         struct rpc_task_setup task_setup_data = {
9877                 .task = &data->task,
9878                 .rpc_client = NFS_CLIENT(data->args.inode),
9879                 .rpc_message = &msg,
9880                 .callback_ops = &nfs4_layoutcommit_ops,
9881                 .callback_data = data,
9882                 .flags = RPC_TASK_MOVEABLE,
9883         };
9884         struct rpc_task *task;
9885         int status = 0;
9886
9887         dprintk("NFS: initiating layoutcommit call. sync %d "
9888                 "lbw: %llu inode %lu\n", sync,
9889                 data->args.lastbytewritten,
9890                 data->args.inode->i_ino);
9891
9892         if (!sync) {
9893                 data->inode = nfs_igrab_and_active(data->args.inode);
9894                 if (data->inode == NULL) {
9895                         nfs4_layoutcommit_release(data);
9896                         return -EAGAIN;
9897                 }
9898                 task_setup_data.flags = RPC_TASK_ASYNC;
9899         }
9900         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
9901         task = rpc_run_task(&task_setup_data);
9902         if (IS_ERR(task))
9903                 return PTR_ERR(task);
9904         if (sync)
9905                 status = task->tk_status;
9906         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
9907         dprintk("%s: status %d\n", __func__, status);
9908         rpc_put_task(task);
9909         return status;
9910 }
9911
9912 /*
9913  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
9914  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
9915  */
9916 static int
9917 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9918                     struct nfs_fsinfo *info,
9919                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
9920 {
9921         struct nfs41_secinfo_no_name_args args = {
9922                 .style = SECINFO_STYLE_CURRENT_FH,
9923         };
9924         struct nfs4_secinfo_res res = {
9925                 .flavors = flavors,
9926         };
9927         struct rpc_message msg = {
9928                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
9929                 .rpc_argp = &args,
9930                 .rpc_resp = &res,
9931         };
9932         struct nfs4_call_sync_data data = {
9933                 .seq_server = server,
9934                 .seq_args = &args.seq_args,
9935                 .seq_res = &res.seq_res,
9936         };
9937         struct rpc_task_setup task_setup = {
9938                 .rpc_client = server->client,
9939                 .rpc_message = &msg,
9940                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
9941                 .callback_data = &data,
9942                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9943         };
9944         const struct cred *cred = NULL;
9945         int status;
9946
9947         if (use_integrity) {
9948                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
9949
9950                 cred = nfs4_get_clid_cred(server->nfs_client);
9951                 msg.rpc_cred = cred;
9952         }
9953
9954         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
9955         status = nfs4_call_sync_custom(&task_setup);
9956         dprintk("<-- %s status=%d\n", __func__, status);
9957
9958         put_cred(cred);
9959
9960         return status;
9961 }
9962
9963 static int
9964 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
9965                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
9966 {
9967         struct nfs4_exception exception = {
9968                 .interruptible = true,
9969         };
9970         int err;
9971         do {
9972                 /* first try using integrity protection */
9973                 err = -NFS4ERR_WRONGSEC;
9974
9975                 /* try to use integrity protection with machine cred */
9976                 if (_nfs4_is_integrity_protected(server->nfs_client))
9977                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9978                                                           flavors, true);
9979
9980                 /*
9981                  * if unable to use integrity protection, or SECINFO with
9982                  * integrity protection returns NFS4ERR_WRONGSEC (which is
9983                  * disallowed by spec, but exists in deployed servers) use
9984                  * the current filesystem's rpc_client and the user cred.
9985                  */
9986                 if (err == -NFS4ERR_WRONGSEC)
9987                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
9988                                                           flavors, false);
9989
9990                 switch (err) {
9991                 case 0:
9992                 case -NFS4ERR_WRONGSEC:
9993                 case -ENOTSUPP:
9994                         goto out;
9995                 default:
9996                         err = nfs4_handle_exception(server, err, &exception);
9997                 }
9998         } while (exception.retry);
9999 out:
10000         return err;
10001 }
10002
10003 static int
10004 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10005                     struct nfs_fsinfo *info)
10006 {
10007         int err;
10008         struct page *page;
10009         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10010         struct nfs4_secinfo_flavors *flavors;
10011         struct nfs4_secinfo4 *secinfo;
10012         int i;
10013
10014         page = alloc_page(GFP_KERNEL);
10015         if (!page) {
10016                 err = -ENOMEM;
10017                 goto out;
10018         }
10019
10020         flavors = page_address(page);
10021         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10022
10023         /*
10024          * Fall back on "guess and check" method if
10025          * the server doesn't support SECINFO_NO_NAME
10026          */
10027         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10028                 err = nfs4_find_root_sec(server, fhandle, info);
10029                 goto out_freepage;
10030         }
10031         if (err)
10032                 goto out_freepage;
10033
10034         for (i = 0; i < flavors->num_flavors; i++) {
10035                 secinfo = &flavors->flavors[i];
10036
10037                 switch (secinfo->flavor) {
10038                 case RPC_AUTH_NULL:
10039                 case RPC_AUTH_UNIX:
10040                 case RPC_AUTH_GSS:
10041                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10042                                         &secinfo->flavor_info);
10043                         break;
10044                 default:
10045                         flavor = RPC_AUTH_MAXFLAVOR;
10046                         break;
10047                 }
10048
10049                 if (!nfs_auth_info_match(&server->auth_info, flavor))
10050                         flavor = RPC_AUTH_MAXFLAVOR;
10051
10052                 if (flavor != RPC_AUTH_MAXFLAVOR) {
10053                         err = nfs4_lookup_root_sec(server, fhandle,
10054                                                    info, flavor);
10055                         if (!err)
10056                                 break;
10057                 }
10058         }
10059
10060         if (flavor == RPC_AUTH_MAXFLAVOR)
10061                 err = -EPERM;
10062
10063 out_freepage:
10064         put_page(page);
10065         if (err == -EACCES)
10066                 return -EPERM;
10067 out:
10068         return err;
10069 }
10070
10071 static int _nfs41_test_stateid(struct nfs_server *server,
10072                 nfs4_stateid *stateid,
10073                 const struct cred *cred)
10074 {
10075         int status;
10076         struct nfs41_test_stateid_args args = {
10077                 .stateid = stateid,
10078         };
10079         struct nfs41_test_stateid_res res;
10080         struct rpc_message msg = {
10081                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10082                 .rpc_argp = &args,
10083                 .rpc_resp = &res,
10084                 .rpc_cred = cred,
10085         };
10086         struct rpc_clnt *rpc_client = server->client;
10087
10088         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10089                 &rpc_client, &msg);
10090
10091         dprintk("NFS call  test_stateid %p\n", stateid);
10092         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10093         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10094                         &args.seq_args, &res.seq_res);
10095         if (status != NFS_OK) {
10096                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10097                 return status;
10098         }
10099         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10100         return -res.status;
10101 }
10102
10103 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10104                 int err, struct nfs4_exception *exception)
10105 {
10106         exception->retry = 0;
10107         switch(err) {
10108         case -NFS4ERR_DELAY:
10109         case -NFS4ERR_RETRY_UNCACHED_REP:
10110                 nfs4_handle_exception(server, err, exception);
10111                 break;
10112         case -NFS4ERR_BADSESSION:
10113         case -NFS4ERR_BADSLOT:
10114         case -NFS4ERR_BAD_HIGH_SLOT:
10115         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10116         case -NFS4ERR_DEADSESSION:
10117                 nfs4_do_handle_exception(server, err, exception);
10118         }
10119 }
10120
10121 /**
10122  * nfs41_test_stateid - perform a TEST_STATEID operation
10123  *
10124  * @server: server / transport on which to perform the operation
10125  * @stateid: state ID to test
10126  * @cred: credential
10127  *
10128  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10129  * Otherwise a negative NFS4ERR value is returned if the operation
10130  * failed or the state ID is not currently valid.
10131  */
10132 static int nfs41_test_stateid(struct nfs_server *server,
10133                 nfs4_stateid *stateid,
10134                 const struct cred *cred)
10135 {
10136         struct nfs4_exception exception = {
10137                 .interruptible = true,
10138         };
10139         int err;
10140         do {
10141                 err = _nfs41_test_stateid(server, stateid, cred);
10142                 nfs4_handle_delay_or_session_error(server, err, &exception);
10143         } while (exception.retry);
10144         return err;
10145 }
10146
10147 struct nfs_free_stateid_data {
10148         struct nfs_server *server;
10149         struct nfs41_free_stateid_args args;
10150         struct nfs41_free_stateid_res res;
10151 };
10152
10153 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10154 {
10155         struct nfs_free_stateid_data *data = calldata;
10156         nfs4_setup_sequence(data->server->nfs_client,
10157                         &data->args.seq_args,
10158                         &data->res.seq_res,
10159                         task);
10160 }
10161
10162 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10163 {
10164         struct nfs_free_stateid_data *data = calldata;
10165
10166         nfs41_sequence_done(task, &data->res.seq_res);
10167
10168         switch (task->tk_status) {
10169         case -NFS4ERR_DELAY:
10170                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10171                         rpc_restart_call_prepare(task);
10172         }
10173 }
10174
10175 static void nfs41_free_stateid_release(void *calldata)
10176 {
10177         struct nfs_free_stateid_data *data = calldata;
10178         struct nfs_client *clp = data->server->nfs_client;
10179
10180         nfs_put_client(clp);
10181         kfree(calldata);
10182 }
10183
10184 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10185         .rpc_call_prepare = nfs41_free_stateid_prepare,
10186         .rpc_call_done = nfs41_free_stateid_done,
10187         .rpc_release = nfs41_free_stateid_release,
10188 };
10189
10190 /**
10191  * nfs41_free_stateid - perform a FREE_STATEID operation
10192  *
10193  * @server: server / transport on which to perform the operation
10194  * @stateid: state ID to release
10195  * @cred: credential
10196  * @privileged: set to true if this call needs to be privileged
10197  *
10198  * Note: this function is always asynchronous.
10199  */
10200 static int nfs41_free_stateid(struct nfs_server *server,
10201                 const nfs4_stateid *stateid,
10202                 const struct cred *cred,
10203                 bool privileged)
10204 {
10205         struct rpc_message msg = {
10206                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10207                 .rpc_cred = cred,
10208         };
10209         struct rpc_task_setup task_setup = {
10210                 .rpc_client = server->client,
10211                 .rpc_message = &msg,
10212                 .callback_ops = &nfs41_free_stateid_ops,
10213                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10214         };
10215         struct nfs_free_stateid_data *data;
10216         struct rpc_task *task;
10217         struct nfs_client *clp = server->nfs_client;
10218
10219         if (!refcount_inc_not_zero(&clp->cl_count))
10220                 return -EIO;
10221
10222         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10223                 &task_setup.rpc_client, &msg);
10224
10225         dprintk("NFS call  free_stateid %p\n", stateid);
10226         data = kmalloc(sizeof(*data), GFP_NOFS);
10227         if (!data)
10228                 return -ENOMEM;
10229         data->server = server;
10230         nfs4_stateid_copy(&data->args.stateid, stateid);
10231
10232         task_setup.callback_data = data;
10233
10234         msg.rpc_argp = &data->args;
10235         msg.rpc_resp = &data->res;
10236         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10237         task = rpc_run_task(&task_setup);
10238         if (IS_ERR(task))
10239                 return PTR_ERR(task);
10240         rpc_put_task(task);
10241         return 0;
10242 }
10243
10244 static void
10245 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10246 {
10247         const struct cred *cred = lsp->ls_state->owner->so_cred;
10248
10249         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10250         nfs4_free_lock_state(server, lsp);
10251 }
10252
10253 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10254                 const nfs4_stateid *s2)
10255 {
10256         if (s1->type != s2->type)
10257                 return false;
10258
10259         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10260                 return false;
10261
10262         if (s1->seqid == s2->seqid)
10263                 return true;
10264
10265         return s1->seqid == 0 || s2->seqid == 0;
10266 }
10267
10268 #endif /* CONFIG_NFS_V4_1 */
10269
10270 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10271                 const nfs4_stateid *s2)
10272 {
10273         return nfs4_stateid_match(s1, s2);
10274 }
10275
10276
10277 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10278         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10279         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10280         .recover_open   = nfs4_open_reclaim,
10281         .recover_lock   = nfs4_lock_reclaim,
10282         .establish_clid = nfs4_init_clientid,
10283         .detect_trunking = nfs40_discover_server_trunking,
10284 };
10285
10286 #if defined(CONFIG_NFS_V4_1)
10287 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10288         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10289         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10290         .recover_open   = nfs4_open_reclaim,
10291         .recover_lock   = nfs4_lock_reclaim,
10292         .establish_clid = nfs41_init_clientid,
10293         .reclaim_complete = nfs41_proc_reclaim_complete,
10294         .detect_trunking = nfs41_discover_server_trunking,
10295 };
10296 #endif /* CONFIG_NFS_V4_1 */
10297
10298 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10299         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10300         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10301         .recover_open   = nfs40_open_expired,
10302         .recover_lock   = nfs4_lock_expired,
10303         .establish_clid = nfs4_init_clientid,
10304 };
10305
10306 #if defined(CONFIG_NFS_V4_1)
10307 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10308         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10309         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10310         .recover_open   = nfs41_open_expired,
10311         .recover_lock   = nfs41_lock_expired,
10312         .establish_clid = nfs41_init_clientid,
10313 };
10314 #endif /* CONFIG_NFS_V4_1 */
10315
10316 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10317         .sched_state_renewal = nfs4_proc_async_renew,
10318         .get_state_renewal_cred = nfs4_get_renew_cred,
10319         .renew_lease = nfs4_proc_renew,
10320 };
10321
10322 #if defined(CONFIG_NFS_V4_1)
10323 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10324         .sched_state_renewal = nfs41_proc_async_sequence,
10325         .get_state_renewal_cred = nfs4_get_machine_cred,
10326         .renew_lease = nfs4_proc_sequence,
10327 };
10328 #endif
10329
10330 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10331         .get_locations = _nfs40_proc_get_locations,
10332         .fsid_present = _nfs40_proc_fsid_present,
10333 };
10334
10335 #if defined(CONFIG_NFS_V4_1)
10336 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10337         .get_locations = _nfs41_proc_get_locations,
10338         .fsid_present = _nfs41_proc_fsid_present,
10339 };
10340 #endif  /* CONFIG_NFS_V4_1 */
10341
10342 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10343         .minor_version = 0,
10344         .init_caps = NFS_CAP_READDIRPLUS
10345                 | NFS_CAP_ATOMIC_OPEN
10346                 | NFS_CAP_POSIX_LOCK,
10347         .init_client = nfs40_init_client,
10348         .shutdown_client = nfs40_shutdown_client,
10349         .match_stateid = nfs4_match_stateid,
10350         .find_root_sec = nfs4_find_root_sec,
10351         .free_lock_state = nfs4_release_lockowner,
10352         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10353         .alloc_seqid = nfs_alloc_seqid,
10354         .call_sync_ops = &nfs40_call_sync_ops,
10355         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10356         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10357         .state_renewal_ops = &nfs40_state_renewal_ops,
10358         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10359 };
10360
10361 #if defined(CONFIG_NFS_V4_1)
10362 static struct nfs_seqid *
10363 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10364 {
10365         return NULL;
10366 }
10367
10368 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10369         .minor_version = 1,
10370         .init_caps = NFS_CAP_READDIRPLUS
10371                 | NFS_CAP_ATOMIC_OPEN
10372                 | NFS_CAP_POSIX_LOCK
10373                 | NFS_CAP_STATEID_NFSV41
10374                 | NFS_CAP_ATOMIC_OPEN_V1
10375                 | NFS_CAP_LGOPEN,
10376         .init_client = nfs41_init_client,
10377         .shutdown_client = nfs41_shutdown_client,
10378         .match_stateid = nfs41_match_stateid,
10379         .find_root_sec = nfs41_find_root_sec,
10380         .free_lock_state = nfs41_free_lock_state,
10381         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10382         .alloc_seqid = nfs_alloc_no_seqid,
10383         .session_trunk = nfs4_test_session_trunk,
10384         .call_sync_ops = &nfs41_call_sync_ops,
10385         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10386         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10387         .state_renewal_ops = &nfs41_state_renewal_ops,
10388         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10389 };
10390 #endif
10391
10392 #if defined(CONFIG_NFS_V4_2)
10393 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10394         .minor_version = 2,
10395         .init_caps = NFS_CAP_READDIRPLUS
10396                 | NFS_CAP_ATOMIC_OPEN
10397                 | NFS_CAP_POSIX_LOCK
10398                 | NFS_CAP_STATEID_NFSV41
10399                 | NFS_CAP_ATOMIC_OPEN_V1
10400                 | NFS_CAP_LGOPEN
10401                 | NFS_CAP_ALLOCATE
10402                 | NFS_CAP_COPY
10403                 | NFS_CAP_OFFLOAD_CANCEL
10404                 | NFS_CAP_COPY_NOTIFY
10405                 | NFS_CAP_DEALLOCATE
10406                 | NFS_CAP_SEEK
10407                 | NFS_CAP_LAYOUTSTATS
10408                 | NFS_CAP_CLONE
10409                 | NFS_CAP_LAYOUTERROR
10410                 | NFS_CAP_READ_PLUS,
10411         .init_client = nfs41_init_client,
10412         .shutdown_client = nfs41_shutdown_client,
10413         .match_stateid = nfs41_match_stateid,
10414         .find_root_sec = nfs41_find_root_sec,
10415         .free_lock_state = nfs41_free_lock_state,
10416         .call_sync_ops = &nfs41_call_sync_ops,
10417         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10418         .alloc_seqid = nfs_alloc_no_seqid,
10419         .session_trunk = nfs4_test_session_trunk,
10420         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10421         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10422         .state_renewal_ops = &nfs41_state_renewal_ops,
10423         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10424 };
10425 #endif
10426
10427 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10428         [0] = &nfs_v4_0_minor_ops,
10429 #if defined(CONFIG_NFS_V4_1)
10430         [1] = &nfs_v4_1_minor_ops,
10431 #endif
10432 #if defined(CONFIG_NFS_V4_2)
10433         [2] = &nfs_v4_2_minor_ops,
10434 #endif
10435 };
10436
10437 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10438 {
10439         ssize_t error, error2, error3;
10440
10441         error = generic_listxattr(dentry, list, size);
10442         if (error < 0)
10443                 return error;
10444         if (list) {
10445                 list += error;
10446                 size -= error;
10447         }
10448
10449         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, size);
10450         if (error2 < 0)
10451                 return error2;
10452
10453         if (list) {
10454                 list += error2;
10455                 size -= error2;
10456         }
10457
10458         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, size);
10459         if (error3 < 0)
10460                 return error3;
10461
10462         return error + error2 + error3;
10463 }
10464
10465 static const struct inode_operations nfs4_dir_inode_operations = {
10466         .create         = nfs_create,
10467         .lookup         = nfs_lookup,
10468         .atomic_open    = nfs_atomic_open,
10469         .link           = nfs_link,
10470         .unlink         = nfs_unlink,
10471         .symlink        = nfs_symlink,
10472         .mkdir          = nfs_mkdir,
10473         .rmdir          = nfs_rmdir,
10474         .mknod          = nfs_mknod,
10475         .rename         = nfs_rename,
10476         .permission     = nfs_permission,
10477         .getattr        = nfs_getattr,
10478         .setattr        = nfs_setattr,
10479         .listxattr      = nfs4_listxattr,
10480 };
10481
10482 static const struct inode_operations nfs4_file_inode_operations = {
10483         .permission     = nfs_permission,
10484         .getattr        = nfs_getattr,
10485         .setattr        = nfs_setattr,
10486         .listxattr      = nfs4_listxattr,
10487 };
10488
10489 const struct nfs_rpc_ops nfs_v4_clientops = {
10490         .version        = 4,                    /* protocol version */
10491         .dentry_ops     = &nfs4_dentry_operations,
10492         .dir_inode_ops  = &nfs4_dir_inode_operations,
10493         .file_inode_ops = &nfs4_file_inode_operations,
10494         .file_ops       = &nfs4_file_operations,
10495         .getroot        = nfs4_proc_get_root,
10496         .submount       = nfs4_submount,
10497         .try_get_tree   = nfs4_try_get_tree,
10498         .getattr        = nfs4_proc_getattr,
10499         .setattr        = nfs4_proc_setattr,
10500         .lookup         = nfs4_proc_lookup,
10501         .lookupp        = nfs4_proc_lookupp,
10502         .access         = nfs4_proc_access,
10503         .readlink       = nfs4_proc_readlink,
10504         .create         = nfs4_proc_create,
10505         .remove         = nfs4_proc_remove,
10506         .unlink_setup   = nfs4_proc_unlink_setup,
10507         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10508         .unlink_done    = nfs4_proc_unlink_done,
10509         .rename_setup   = nfs4_proc_rename_setup,
10510         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10511         .rename_done    = nfs4_proc_rename_done,
10512         .link           = nfs4_proc_link,
10513         .symlink        = nfs4_proc_symlink,
10514         .mkdir          = nfs4_proc_mkdir,
10515         .rmdir          = nfs4_proc_rmdir,
10516         .readdir        = nfs4_proc_readdir,
10517         .mknod          = nfs4_proc_mknod,
10518         .statfs         = nfs4_proc_statfs,
10519         .fsinfo         = nfs4_proc_fsinfo,
10520         .pathconf       = nfs4_proc_pathconf,
10521         .set_capabilities = nfs4_server_capabilities,
10522         .decode_dirent  = nfs4_decode_dirent,
10523         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10524         .read_setup     = nfs4_proc_read_setup,
10525         .read_done      = nfs4_read_done,
10526         .write_setup    = nfs4_proc_write_setup,
10527         .write_done     = nfs4_write_done,
10528         .commit_setup   = nfs4_proc_commit_setup,
10529         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10530         .commit_done    = nfs4_commit_done,
10531         .lock           = nfs4_proc_lock,
10532         .clear_acl_cache = nfs4_zap_acl_attr,
10533         .close_context  = nfs4_close_context,
10534         .open_context   = nfs4_atomic_open,
10535         .have_delegation = nfs4_have_delegation,
10536         .alloc_client   = nfs4_alloc_client,
10537         .init_client    = nfs4_init_client,
10538         .free_client    = nfs4_free_client,
10539         .create_server  = nfs4_create_server,
10540         .clone_server   = nfs_clone_server,
10541         .discover_trunking = nfs4_discover_trunking,
10542 };
10543
10544 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10545         .name   = XATTR_NAME_NFSV4_ACL,
10546         .list   = nfs4_xattr_list_nfs4_acl,
10547         .get    = nfs4_xattr_get_nfs4_acl,
10548         .set    = nfs4_xattr_set_nfs4_acl,
10549 };
10550
10551 #ifdef CONFIG_NFS_V4_2
10552 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10553         .prefix = XATTR_USER_PREFIX,
10554         .get    = nfs4_xattr_get_nfs4_user,
10555         .set    = nfs4_xattr_set_nfs4_user,
10556 };
10557 #endif
10558
10559 const struct xattr_handler *nfs4_xattr_handlers[] = {
10560         &nfs4_xattr_nfs4_acl_handler,
10561 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10562         &nfs4_xattr_nfs4_label_handler,
10563 #endif
10564 #ifdef CONFIG_NFS_V4_2
10565         &nfs4_xattr_nfs4_user_handler,
10566 #endif
10567         NULL
10568 };