kernel/fs/fs_subr.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*      Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */
  22 /*        All Rights Reserved   */
  23
  24
  25 /*
  26  * Copyright (c) 1989, 2010, Oracle and/or its affiliates. All rights reserved.
  27  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  28  * Copyright 2015 Joyent, Inc.
  29  */
  30
  31 /*
  32  * Generic vnode operations.
  33  */
  34 #include <sys/types.h>
  35 #include <sys/param.h>
  36 #include <sys/systm.h>
  37 #include <sys/errno.h>
  38 #include <sys/fcntl.h>
  39 #include <sys/flock.h>
  40 #include <sys/statvfs.h>
  41 #include <sys/vfs.h>
  42 #include <sys/vnode.h>
  43 #include <sys/proc.h>
  44 #include <sys/user.h>
  45 #include <sys/unistd.h>
  46 #include <sys/cred.h>
  47 #include <sys/poll.h>
  48 #include <sys/debug.h>
  49 #include <sys/cmn_err.h>
  50 #include <sys/stream.h>
  51 #include <sys/fs_subr.h>
  52 #include <sys/fs_reparse.h>
  53 #include <sys/door.h>
  54 #include <sys/acl.h>
  55 #include <sys/share.h>
  56 #include <sys/file.h>
  57 #include <sys/kmem.h>
  58 #include <sys/file.h>
  59 #include <sys/nbmlock.h>
  60 #include <acl/acl_common.h>
  61 #include <sys/pathname.h>
  62
  63 static callb_cpr_t *frlock_serialize_blocked(flk_cb_when_t, void *);
  64
  65 /*
  66  * Tunable to limit the number of retry to recover from STALE error.
  67  */
  68 int fs_estale_retry = 5;
  69
  70 /*
  71  * supports for reparse point door upcall
  72  */
  73 static door_handle_t reparsed_door;
  74 static kmutex_t reparsed_door_lock;
  75
  76 /*
  77  * The associated operation is not supported by the file system.
  78  */
  79 int
  80 fs_nosys()
  81 {
  82         return (ENOSYS);
  83 }
  84
  85 /*
  86  * The associated operation is invalid (on this vnode).
  87  */
  88 int
  89 fs_inval()
  90 {
  91         return (EINVAL);
  92 }
  93
  94 /*
  95  * The associated operation is valid only for directories.
  96  */
  97 int
  98 fs_notdir()
  99 {
 100         return (ENOTDIR);
 101 }
 102
 103 /*
 104  * Free the file system specific resources. For the file systems that
 105  * do not support the forced unmount, it will be a nop function.
 106  */
 107
 108 /*ARGSUSED*/
 109 void
 110 fs_freevfs(vfs_t *vfsp)
 111 {
 112 }
 113
 114 /* ARGSUSED */
 115 int
 116 fs_nosys_map(struct vnode *vp, offset_t off, struct as *as, caddr_t *addrp,
 117     size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, struct cred *cr,
 118     caller_context_t *ct)
 119 {
 120         return (ENOSYS);
 121 }
 122
 123 /* ARGSUSED */
 124 int
 125 fs_nosys_addmap(struct vnode *vp, offset_t off, struct as *as, caddr_t addr,
 126     size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, struct cred *cr,
 127     caller_context_t *ct)
 128 {
 129         return (ENOSYS);
 130 }
 131
 132 /* ARGSUSED */
 133 int
 134 fs_nosys_poll(vnode_t *vp, short events, int anyyet, short *reventsp,
 135     struct pollhead **phpp, caller_context_t *ct)
 136 {
 137         return (ENOSYS);
 138 }
 139
 140
 141 /*
 142  * The file system has nothing to sync to disk.  However, the
 143  * VFS_SYNC operation must not fail.
 144  */
 145 /* ARGSUSED */
 146 int
 147 fs_sync(struct vfs *vfspp, short flag, cred_t *cr)
 148 {
 149         return (0);
 150 }
 151
 152 /*
 153  * Does nothing but fop_fsync must not fail.
 154  */
 155 /* ARGSUSED */
 156 int
 157 fs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
 158 {
 159         return (0);
 160 }
 161
 162 /*
 163  * Does nothing but fop_putpage must not fail.
 164  */
 165 /* ARGSUSED */
 166 int
 167 fs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr,
 168     caller_context_t *ctp)
 169 {
 170         return (0);
 171 }
 172
 173 /*
 174  * Does nothing but fop_ioctl must not fail.
 175  */
 176 /* ARGSUSED */
 177 int
 178 fs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred,
 179     int *rvalp)
 180 {
 181         return (0);
 182 }
 183
 184 /*
 185  * Read/write lock/unlock.  Does nothing.
 186  */
 187 /* ARGSUSED */
 188 int
 189 fs_rwlock(vnode_t *vp, int write_lock, caller_context_t *ctp)
 190 {
 191         return (-1);
 192 }
 193
 194 /* ARGSUSED */
 195 void
 196 fs_rwunlock(vnode_t *vp, int write_lock, caller_context_t *ctp)
 197 {
 198 }
 199
 200 /*
 201  * Compare two vnodes.
 202  */
 203 /*ARGSUSED2*/
 204 int
 205 fs_cmp(vnode_t *vp1, vnode_t *vp2, caller_context_t *ct)
 206 {
 207         return (vp1 == vp2);
 208 }
 209
 210 /*
 211  * No-op seek operation.
 212  */
 213 /* ARGSUSED */
 214 int
 215 fs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp, caller_context_t *ct)
 216 {
 217         return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0);
 218 }
 219
 220 /*
 221  * File and record locking.
 222  */
 223 /* ARGSUSED */
 224 int
 225 fs_frlock(vnode_t *vp, int cmd, struct flock64 *bfp, int flag, offset_t offset,
 226     flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct)
 227 {
 228         int frcmd;
 229         int error = 0;
 230         boolean_t skip_lock = B_FALSE;
 231         flk_callback_t serialize_callback;
 232         int serialize = 0;
 233         v_mode_t mode;
 234
 235         switch (cmd) {
 236
 237         case F_GETLK:
 238         case F_O_GETLK:
 239                 if (flag & F_REMOTELOCK) {
 240                         frcmd = RCMDLCK;
 241                 } else {
 242                         frcmd = 0;
 243                         bfp->l_pid = ttoproc(curthread)->p_pid;
 244                         bfp->l_sysid = 0;
 245                 }
 246                 break;
 247
 248         case F_OFD_GETLK:
 249                 /*
 250                  * TBD we do not support remote OFD locks at this time.
 251                  */
 252                 if (flag & F_REMOTELOCK) {
 253                         error = EINVAL;
 254                         goto done;
 255                 }
 256                 skip_lock = B_TRUE;
 257                 break;
 258
 259         case F_SETLK_NBMAND:
 260                 /*
 261                  * Are NBMAND locks allowed on this file?
 262                  */
 263                 if (!vp->v_vfsp ||
 264                     !(vp->v_vfsp->vfs_flag & VFS_NBMAND)) {
 265                         error = EINVAL;
 266                         goto done;
 267                 }
 268                 if (vp->v_type != VREG) {
 269                         error = EINVAL;
 270                         goto done;
 271                 }
 272                 /*FALLTHROUGH*/
 273
 274         case F_SETLK:
 275                 if (flag & F_REMOTELOCK) {
 276                         frcmd = SETFLCK|RCMDLCK;
 277                 } else {
 278                         frcmd = SETFLCK;
 279                         bfp->l_pid = ttoproc(curthread)->p_pid;
 280                         bfp->l_sysid = 0;
 281                 }
 282                 if (cmd == F_SETLK_NBMAND &&
 283                     (bfp->l_type == F_RDLCK || bfp->l_type == F_WRLCK)) {
 284                         frcmd |= NBMLCK;
 285                 }
 286
 287                 if (nbl_need_check(vp)) {
 288                         nbl_start_crit(vp, RW_WRITER);
 289                         serialize = 1;
 290                         if (frcmd & NBMLCK) {
 291                                 mode = (bfp->l_type == F_RDLCK) ?
 292                                     V_READ : V_RDANDWR;
 293                                 if (vn_is_mapped(vp, mode)) {
 294                                         error = EAGAIN;
 295                                         goto done;
 296                                 }
 297                         }
 298                 }
 299                 break;
 300
 301         case F_SETLKW:
 302                 if (flag & F_REMOTELOCK) {
 303                         frcmd = SETFLCK|SLPFLCK|RCMDLCK;
 304                 } else {
 305                         frcmd = SETFLCK|SLPFLCK;
 306                         bfp->l_pid = ttoproc(curthread)->p_pid;
 307                         bfp->l_sysid = 0;
 308                 }
 309
 310                 if (nbl_need_check(vp)) {
 311                         nbl_start_crit(vp, RW_WRITER);
 312                         serialize = 1;
 313                 }
 314                 break;
 315
 316         case F_OFD_SETLK:
 317         case F_OFD_SETLKW:
 318         case F_FLOCK:
 319         case F_FLOCKW:
 320                 /*
 321                  * TBD we do not support remote OFD locks at this time.
 322                  */
 323                 if (flag & F_REMOTELOCK) {
 324                         error = EINVAL;
 325                         goto done;
 326                 }
 327                 skip_lock = B_TRUE;
 328                 break;
 329
 330         case F_HASREMOTELOCKS:
 331                 l_has_rmt(bfp) = flk_has_remote_locks(vp);
 332                 goto done;
 333
 334         default:
 335                 error = EINVAL;
 336                 goto done;
 337         }
 338
 339         /*
 340          * If this is a blocking lock request and we're serializing lock
 341          * requests, modify the callback list to leave the critical region
 342          * while we're waiting for the lock.
 343          */
 344
 345         if (serialize && (frcmd & SLPFLCK) != 0) {
 346                 flk_add_callback(&serialize_callback,
 347                     frlock_serialize_blocked, vp, flk_cbp);
 348                 flk_cbp = &serialize_callback;
 349         }
 350
 351         if (!skip_lock)
 352                 error = reclock(vp, bfp, frcmd, flag, offset, flk_cbp);
 353
 354         if (serialize && (frcmd & SLPFLCK) != 0)
 355                 flk_del_callback(&serialize_callback);
 356
 357 done:
 358         if (serialize)
 359                 nbl_end_crit(vp);
 360
 361         return (error);
 362 }
 363
 364 /*
 365  * Callback when a lock request blocks and we are serializing requests.  If
 366  * before sleeping, leave the critical region.  If after wakeup, reenter
 367  * the critical region.
 368  */
 369
 370 static callb_cpr_t *
 371 frlock_serialize_blocked(flk_cb_when_t when, void *infop)
 372 {
 373         vnode_t *vp = (vnode_t *)infop;
 374
 375         if (when == FLK_BEFORE_SLEEP)
 376                 nbl_end_crit(vp);
 377         else {
 378                 nbl_start_crit(vp, RW_WRITER);
 379         }
 380
 381         return (NULL);
 382 }
 383
 384 /*
 385  * Allow any flags.
 386  */
 387 /* ARGSUSED */
 388 int
 389 fs_setfl(vnode_t *vp, int oflags, int nflags, cred_t *cr, caller_context_t *ct)
 390 {
 391         return (0);
 392 }
 393
 394 /*
 395  * Return the answer requested to poll() for non-device files.
 396  * Only POLLIN, POLLRDNORM, and POLLOUT are recognized.
 397  */
 398 struct pollhead fs_pollhd;
 399
 400 /* ARGSUSED */
 401 int
 402 fs_poll(vnode_t *vp, short events, int anyyet, short *reventsp,
 403     struct pollhead **phpp, caller_context_t *ct)
 404 {
 405         *reventsp = 0;
 406         if (events & POLLIN)
 407                 *reventsp |= POLLIN;
 408         if (events & POLLRDNORM)
 409                 *reventsp |= POLLRDNORM;
 410         if (events & POLLRDBAND)
 411                 *reventsp |= POLLRDBAND;
 412         if (events & POLLOUT)
 413                 *reventsp |= POLLOUT;
 414         if (events & POLLWRBAND)
 415                 *reventsp |= POLLWRBAND;
 416         *phpp = !anyyet && !*reventsp ? &fs_pollhd : NULL;
 417         return (0);
 418 }
 419
 420 /*
 421  * POSIX pathconf() support.
 422  */
 423 /* ARGSUSED */
 424 int
 425 fs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
 426     caller_context_t *ct)
 427 {
 428         ulong_t val;
 429         int error = 0;
 430         struct statvfs64 vfsbuf;
 431
 432         switch (cmd) {
 433
 434         case _PC_LINK_MAX:
 435                 val = MAXLINK;
 436                 break;
 437
 438         case _PC_MAX_CANON:
 439                 val = MAX_CANON;
 440                 break;
 441
 442         case _PC_MAX_INPUT:
 443                 val = MAX_INPUT;
 444                 break;
 445
 446         case _PC_NAME_MAX:
 447                 bzero(&vfsbuf, sizeof (vfsbuf));
 448                 if (error = VFS_STATVFS(vp->v_vfsp, &vfsbuf))
 449                         break;
 450                 val = vfsbuf.f_namemax;
 451                 break;
 452
 453         case _PC_PATH_MAX:
 454         case _PC_SYMLINK_MAX:
 455                 val = MAXPATHLEN;
 456                 break;
 457
 458         case _PC_PIPE_BUF:
 459                 val = PIPE_BUF;
 460                 break;
 461
 462         case _PC_NO_TRUNC:
 463                 if (vp->v_vfsp->vfs_flag & VFS_NOTRUNC)
 464                         val = 1;        /* NOTRUNC is enabled for vp */
 465                 else
 466                         val = (ulong_t)-1;
 467                 break;
 468
 469         case _PC_VDISABLE:
 470                 val = _POSIX_VDISABLE;
 471                 break;
 472
 473         case _PC_CHOWN_RESTRICTED:
 474                 if (rstchown)
 475                         val = rstchown; /* chown restricted enabled */
 476                 else
 477                         val = (ulong_t)-1;
 478                 break;
 479
 480         case _PC_FILESIZEBITS:
 481
 482                 /*
 483                  * If ever we come here it means that underlying file system
 484                  * does not recognise the command and therefore this
 485                  * configurable limit cannot be determined. We return -1
 486                  * and don't change errno.
 487                  */
 488
 489                 val = (ulong_t)-1;    /* large file support */
 490                 break;
 491
 492         case _PC_ACL_ENABLED:
 493                 val = 0;
 494                 break;
 495
 496         case _PC_CASE_BEHAVIOR:
 497                 val = _CASE_SENSITIVE;
 498                 if (vfs_has_feature(vp->v_vfsp, VFSFT_CASEINSENSITIVE) == 1)
 499                         val |= _CASE_INSENSITIVE;
 500                 if (vfs_has_feature(vp->v_vfsp, VFSFT_NOCASESENSITIVE) == 1)
 501                         val &= ~_CASE_SENSITIVE;
 502                 break;
 503
 504         case _PC_SATTR_ENABLED:
 505         case _PC_SATTR_EXISTS:
 506                 val = 0;
 507                 break;
 508
 509         case _PC_ACCESS_FILTERING:
 510                 val = 0;
 511                 break;
 512
 513         default:
 514                 error = EINVAL;
 515                 break;
 516         }
 517
 518         if (error == 0)
 519                 *valp = val;
 520         return (error);
 521 }
 522
 523 /*
 524  * Dispose of a page.
 525  */
 526 /* ARGSUSED */
 527 void
 528 fs_dispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 529     caller_context_t *ct)
 530 {
 531
 532         ASSERT(fl == B_FREE || fl == B_INVAL);
 533
 534         if (fl == B_FREE)
 535                 page_free(pp, dn);
 536         else
 537                 page_destroy(pp, dn);
 538 }
 539
 540 /* ARGSUSED */
 541 void
 542 fs_nodispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 543     caller_context_t *ct)
 544 {
 545         cmn_err(CE_PANIC, "fs_nodispose invoked");
 546 }
 547
 548 /*
 549  * fabricate acls for file systems that do not support acls.
 550  */
 551 /* ARGSUSED */
 552 int
 553 fs_fab_acl(vnode_t *vp, vsecattr_t *vsecattr, int flag, cred_t *cr,
 554     caller_context_t *ct)
 555 {
 556         aclent_t        *aclentp;
 557         struct vattr    vattr;
 558         int             error;
 559         size_t          aclsize;
 560
 561         vsecattr->vsa_aclcnt    = 0;
 562         vsecattr->vsa_aclentsz  = 0;
 563         vsecattr->vsa_aclentp   = NULL;
 564         vsecattr->vsa_dfaclcnt  = 0;    /* Default ACLs are not fabricated */
 565         vsecattr->vsa_dfaclentp = NULL;
 566
 567         vattr.va_mask = AT_MODE | AT_UID | AT_GID;
 568         if (error = fop_getattr(vp, &vattr, 0, cr, ct))
 569                 return (error);
 570
 571         if (vsecattr->vsa_mask & (VSA_ACLCNT | VSA_ACL)) {
 572                 aclsize = 4 * sizeof (aclent_t);
 573                 vsecattr->vsa_aclcnt    = 4; /* USER, GROUP, OTHER, and CLASS */
 574                 vsecattr->vsa_aclentp = kmem_zalloc(aclsize, KM_SLEEP);
 575                 aclentp = vsecattr->vsa_aclentp;
 576
 577                 aclentp->a_type = USER_OBJ;     /* Owner */
 578                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0700)) >> 6;
 579                 aclentp->a_id = vattr.va_uid;   /* Really undefined */
 580                 aclentp++;
 581
 582                 aclentp->a_type = GROUP_OBJ;    /* Group */
 583                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0070)) >> 3;
 584                 aclentp->a_id = vattr.va_gid;   /* Really undefined */
 585                 aclentp++;
 586
 587                 aclentp->a_type = OTHER_OBJ;    /* Other */
 588                 aclentp->a_perm = vattr.va_mode & 0007;
 589                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 590                 aclentp++;
 591
 592                 aclentp->a_type = CLASS_OBJ;    /* Class */
 593                 aclentp->a_perm = (ushort_t)(0007);
 594                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 595         } else if (vsecattr->vsa_mask & (VSA_ACECNT | VSA_ACE)) {
 596                 VERIFY(0 == acl_trivial_create(vattr.va_mode,
 597                     (vp->v_type == VDIR), (ace_t **)&vsecattr->vsa_aclentp,
 598                     &vsecattr->vsa_aclcnt));
 599                 vsecattr->vsa_aclentsz = vsecattr->vsa_aclcnt * sizeof (ace_t);
 600         }
 601
 602         return (error);
 603 }
 604
 605 /*
 606  * Common code for implementing DOS share reservations
 607  */
 608 /* ARGSUSED4 */
 609 int
 610 fs_shrlock(struct vnode *vp, int cmd, struct shrlock *shr, int flag, cred_t *cr,
 611     caller_context_t *ct)
 612 {
 613         int error;
 614
 615         /*
 616          * Make sure that the file was opened with permissions appropriate
 617          * for the request, and make sure the caller isn't trying to sneak
 618          * in an NBMAND request.
 619          */
 620         if (cmd == F_SHARE) {
 621                 if (((shr->s_access & F_RDACC) && (flag & FREAD) == 0) ||
 622                     ((shr->s_access & F_WRACC) && (flag & FWRITE) == 0))
 623                         return (EBADF);
 624                 if (shr->s_access & (F_RMACC | F_MDACC))
 625                         return (EINVAL);
 626                 if (shr->s_deny & (F_MANDDNY | F_RMDNY))
 627                         return (EINVAL);
 628         }
 629         if (cmd == F_SHARE_NBMAND) {
 630                 /* make sure nbmand is allowed on the file */
 631                 if (!vp->v_vfsp ||
 632                     !(vp->v_vfsp->vfs_flag & VFS_NBMAND)) {
 633                         return (EINVAL);
 634                 }
 635                 if (vp->v_type != VREG) {
 636                         return (EINVAL);
 637                 }
 638         }
 639
 640         nbl_start_crit(vp, RW_WRITER);
 641
 642         switch (cmd) {
 643
 644         case F_SHARE_NBMAND:
 645                 shr->s_deny |= F_MANDDNY;
 646                 /*FALLTHROUGH*/
 647         case F_SHARE:
 648                 error = add_share(vp, shr);
 649                 break;
 650
 651         case F_UNSHARE:
 652                 error = del_share(vp, shr);
 653                 break;
 654
 655         case F_HASREMOTELOCKS:
 656                 /*
 657                  * We are overloading this command to refer to remote
 658                  * shares as well as remote locks, despite its name.
 659                  */
 660                 shr->s_access = shr_has_remote_shares(vp, shr->s_sysid);
 661                 error = 0;
 662                 break;
 663
 664         default:
 665                 error = EINVAL;
 666                 break;
 667         }
 668
 669         nbl_end_crit(vp);
 670         return (error);
 671 }
 672
 673 /*ARGSUSED1*/
 674 int
 675 fs_vnevent_nosupport(vnode_t *vp, vnevent_t e, vnode_t *dvp, char *fnm,
 676     caller_context_t *ct)
 677 {
 678         ASSERT(vp != NULL);
 679         return (ENOTSUP);
 680 }
 681
 682 /*ARGSUSED1*/
 683 int
 684 fs_vnevent_support(vnode_t *vp, vnevent_t e, vnode_t *dvp, char *fnm,
 685     caller_context_t *ct)
 686 {
 687         ASSERT(vp != NULL);
 688         return (0);
 689 }
 690
 691 /*
 692  * return 1 for non-trivial ACL.
 693  *
 694  * NB: It is not necessary for the caller to fop_rwlock since
 695  *      we only issue fop_getsecattr.
 696  *
 697  * Returns 0 == trivial
 698  *         1 == NOT Trivial
 699  *         <0 could not determine.
 700  */
 701 int
 702 fs_acl_nontrivial(vnode_t *vp, cred_t *cr)
 703 {
 704         ulong_t         acl_styles;
 705         ulong_t         acl_flavor;
 706         vsecattr_t      vsecattr;
 707         int             error;
 708         int             isnontrivial;
 709
 710         /* determine the forms of ACLs maintained */
 711         error = fop_pathconf(vp, _PC_ACL_ENABLED, &acl_styles, cr, NULL);
 712
 713         /* clear bits we don't understand and establish default acl_style */
 714         acl_styles &= (_ACL_ACLENT_ENABLED | _ACL_ACE_ENABLED);
 715         if (error || (acl_styles == 0))
 716                 acl_styles = _ACL_ACLENT_ENABLED;
 717
 718         vsecattr.vsa_aclentp = NULL;
 719         vsecattr.vsa_dfaclentp = NULL;
 720         vsecattr.vsa_aclcnt = 0;
 721         vsecattr.vsa_dfaclcnt = 0;
 722
 723         while (acl_styles) {
 724                 /* select one of the styles as current flavor */
 725                 acl_flavor = 0;
 726                 if (acl_styles & _ACL_ACLENT_ENABLED) {
 727                         acl_flavor = _ACL_ACLENT_ENABLED;
 728                         vsecattr.vsa_mask = VSA_ACLCNT | VSA_DFACLCNT;
 729                 } else if (acl_styles & _ACL_ACE_ENABLED) {
 730                         acl_flavor = _ACL_ACE_ENABLED;
 731                         vsecattr.vsa_mask = VSA_ACECNT | VSA_ACE;
 732                 }
 733
 734                 ASSERT(vsecattr.vsa_mask && acl_flavor);
 735                 error = fop_getsecattr(vp, &vsecattr, 0, cr, NULL);
 736                 if (error == 0)
 737                         break;
 738
 739                 /* that flavor failed */
 740                 acl_styles &= ~acl_flavor;
 741         }
 742
 743         /* if all styles fail then assume trivial */
 744         if (acl_styles == 0)
 745                 return (0);
 746
 747         /* process the flavor that worked */
 748         isnontrivial = 0;
 749         if (acl_flavor & _ACL_ACLENT_ENABLED) {
 750                 if (vsecattr.vsa_aclcnt > MIN_ACL_ENTRIES)
 751                         isnontrivial = 1;
 752                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 753                         kmem_free(vsecattr.vsa_aclentp,
 754                             vsecattr.vsa_aclcnt * sizeof (aclent_t));
 755                 if (vsecattr.vsa_dfaclcnt && vsecattr.vsa_dfaclentp != NULL)
 756                         kmem_free(vsecattr.vsa_dfaclentp,
 757                             vsecattr.vsa_dfaclcnt * sizeof (aclent_t));
 758         }
 759         if (acl_flavor & _ACL_ACE_ENABLED) {
 760                 isnontrivial = ace_trivial(vsecattr.vsa_aclentp,
 761                     vsecattr.vsa_aclcnt);
 762
 763                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 764                         kmem_free(vsecattr.vsa_aclentp,
 765                             vsecattr.vsa_aclcnt * sizeof (ace_t));
 766                 /* ACE has no vsecattr.vsa_dfaclcnt */
 767         }
 768         return (isnontrivial);
 769 }
 770
 771 /*
 772  * Check whether we need a retry to recover from STALE error.
 773  */
 774 int
 775 fs_need_estale_retry(int retry_count)
 776 {
 777         if (retry_count < fs_estale_retry)
 778                 return (1);
 779         else
 780                 return (0);
 781 }
 782
 783
 784 static int (*fs_av_scan)(vnode_t *, cred_t *, int) = NULL;
 785
 786 /*
 787  * Routine for anti-virus scanner to call to register its scanning routine.
 788  */
 789 void
 790 fs_vscan_register(int (*av_scan)(vnode_t *, cred_t *, int))
 791 {
 792         fs_av_scan = av_scan;
 793 }
 794
 795 /*
 796  * Routine for file systems to call to initiate anti-virus scanning.
 797  * Scanning will only be done on REGular files (currently).
 798  */
 799 int
 800 fs_vscan(vnode_t *vp, cred_t *cr, int async)
 801 {
 802         int ret = 0;
 803
 804         if (fs_av_scan && vp->v_type == VREG)
 805                 ret = (*fs_av_scan)(vp, cr, async);
 806
 807         return (ret);
 808 }
 809
 810 /*
 811  * support functions for reparse point
 812  */
 813 /*
 814  * reparse_vnode_parse
 815  *
 816  * Read the symlink data of a reparse point specified by the vnode
 817  * and return the reparse data as name-value pair in the nvlist.
 818  */
 819 int
 820 reparse_vnode_parse(vnode_t *vp, nvlist_t *nvl)
 821 {
 822         int err;
 823         char *lkdata;
 824         struct uio uio;
 825         struct iovec iov;
 826
 827         if (vp == NULL || nvl == NULL)
 828                 return (EINVAL);
 829
 830         lkdata = kmem_alloc(MAXREPARSELEN, KM_SLEEP);
 831
 832         /*
 833          * Set up io vector to read sym link data
 834          */
 835         iov.iov_base = lkdata;
 836         iov.iov_len = MAXREPARSELEN;
 837         uio.uio_iov = &iov;
 838         uio.uio_iovcnt = 1;
 839         uio.uio_segflg = UIO_SYSSPACE;
 840         uio.uio_extflg = UIO_COPY_CACHED;
 841         uio.uio_loffset = (offset_t)0;
 842         uio.uio_resid = MAXREPARSELEN;
 843
 844         if ((err = fop_readlink(vp, &uio, kcred, NULL)) == 0) {
 845                 *(lkdata + MAXREPARSELEN - uio.uio_resid) = '\0';
 846                 err = reparse_parse(lkdata, nvl);
 847         }
 848         kmem_free(lkdata, MAXREPARSELEN);       /* done with lkdata */
 849
 850         return (err);
 851 }
 852
 853 void
 854 reparse_point_init()
 855 {
 856         mutex_init(&reparsed_door_lock, NULL, MUTEX_DEFAULT, NULL);
 857 }
 858
 859 static door_handle_t
 860 reparse_door_get_handle()
 861 {
 862         door_handle_t dh;
 863
 864         mutex_enter(&reparsed_door_lock);
 865         if ((dh = reparsed_door) == NULL) {
 866                 if (door_ki_open(REPARSED_DOOR, &reparsed_door) != 0) {
 867                         reparsed_door = NULL;
 868                         dh = NULL;
 869                 } else
 870                         dh = reparsed_door;
 871         }
 872         mutex_exit(&reparsed_door_lock);
 873         return (dh);
 874 }
 875
 876 static void
 877 reparse_door_reset_handle()
 878 {
 879         mutex_enter(&reparsed_door_lock);
 880         reparsed_door = NULL;
 881         mutex_exit(&reparsed_door_lock);
 882 }
 883
 884 /*
 885  * reparse_kderef
 886  *
 887  * Accepts the service-specific item from the reparse point and returns
 888  * the service-specific data requested.  The caller specifies the size of
 889  * the buffer provided via *bufsz; the routine will fail with EOVERFLOW
 890  * if the results will not fit in the buffer, in which case, *bufsz will
 891  * contain the number of bytes needed to hold the results.
 892  *
 893  * if ok return 0 and update *bufsize with length of actual result
 894  * else return error code.
 895  */
 896 int
 897 reparse_kderef(const char *svc_type, const char *svc_data, char *buf,
 898     size_t *bufsize)
 899 {
 900         int err, retries, need_free, retried_doorhd;
 901         size_t dlen, res_len;
 902         char *darg;
 903         door_arg_t door_args;
 904         reparsed_door_res_t *resp;
 905         door_handle_t rp_door;
 906
 907         if (svc_type == NULL || svc_data == NULL || buf == NULL ||
 908             bufsize == NULL)
 909                 return (EINVAL);
 910
 911         /* get reparsed's door handle */
 912         if ((rp_door = reparse_door_get_handle()) == NULL)
 913                 return (EBADF);
 914
 915         /* setup buffer for door_call args and results */
 916         dlen = strlen(svc_type) + strlen(svc_data) + 2;
 917         if (*bufsize < dlen) {
 918                 darg = kmem_alloc(dlen, KM_SLEEP);
 919                 need_free = 1;
 920         } else {
 921                 darg = buf;     /* use same buffer for door's args & results */
 922                 need_free = 0;
 923         }
 924
 925         /* build argument string of door call */
 926         (void) snprintf(darg, dlen, "%s:%s", svc_type, svc_data);
 927
 928         /* setup args for door call */
 929         door_args.data_ptr = darg;
 930         door_args.data_size = dlen;
 931         door_args.desc_ptr = NULL;
 932         door_args.desc_num = 0;
 933         door_args.rbuf = buf;
 934         door_args.rsize = *bufsize;
 935
 936         /* do the door_call */
 937         retried_doorhd = 0;
 938         retries = 0;
 939         door_ki_hold(rp_door);
 940         while ((err = door_ki_upcall_limited(rp_door, &door_args,
 941             NULL, SIZE_MAX, 0)) != 0) {
 942                 if (err == EAGAIN || err == EINTR) {
 943                         if (++retries < REPARSED_DOORCALL_MAX_RETRY) {
 944                                 delay(SEC_TO_TICK(1));
 945                                 continue;
 946                         }
 947                 } else if (err == EBADF) {
 948                         /* door server goes away... */
 949                         reparse_door_reset_handle();
 950
 951                         if (retried_doorhd == 0) {
 952                                 door_ki_rele(rp_door);
 953                                 retried_doorhd++;
 954                                 rp_door = reparse_door_get_handle();
 955                                 if (rp_door != NULL) {
 956                                         door_ki_hold(rp_door);
 957                                         continue;
 958                                 }
 959                         }
 960                 }
 961                 break;
 962         }
 963
 964         if (rp_door)
 965                 door_ki_rele(rp_door);
 966
 967         if (need_free)
 968                 kmem_free(darg, dlen);          /* done with args buffer */
 969
 970         if (err != 0)
 971                 return (err);
 972
 973         resp = (reparsed_door_res_t *)door_args.rbuf;
 974         if ((err = resp->res_status) == 0) {
 975                 /*
 976                  * have to save the length of the results before the
 977                  * bcopy below since it's can be an overlap copy that
 978                  * overwrites the reparsed_door_res_t structure at
 979                  * the beginning of the buffer.
 980                  */
 981                 res_len = (size_t)resp->res_len;
 982
 983                 /* deref call is ok */
 984                 if (res_len > *bufsize)
 985                         err = EOVERFLOW;
 986                 else
 987                         bcopy(resp->res_data, buf, res_len);
 988                 *bufsize = res_len;
 989         }
 990         if (door_args.rbuf != buf)
 991                 kmem_free(door_args.rbuf, door_args.rsize);
 992
 993         return (err);
 994 }