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 2017 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_poll(vnode_t *vp, short events, int anyyet, short *reventsp,
 117     struct pollhead **phpp, caller_context_t *ct)
 118 {
 119         return (ENOSYS);
 120 }
 121
 122
 123 /*
 124  * Does nothing but fop_fsync must not fail.
 125  */
 126 /* ARGSUSED */
 127 int
 128 fs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct)
 129 {
 130         return (0);
 131 }
 132
 133 /*
 134  * Does nothing but fop_putpage must not fail.
 135  */
 136 /* ARGSUSED */
 137 int
 138 fs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr,
 139     caller_context_t *ctp)
 140 {
 141         return (0);
 142 }
 143
 144 /*
 145  * Does nothing but fop_ioctl must not fail.
 146  */
 147 /* ARGSUSED */
 148 int
 149 fs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred,
 150     int *rvalp, caller_context_t *ct)
 151 {
 152         return (0);
 153 }
 154
 155 /*
 156  * No-op seek operation.
 157  */
 158 /* ARGSUSED */
 159 int
 160 fs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp, caller_context_t *ct)
 161 {
 162         return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0);
 163 }
 164
 165 /*
 166  * File and record locking.
 167  */
 168 /* ARGSUSED */
 169 int
 170 fs_frlock(vnode_t *vp, int cmd, struct flock64 *bfp, int flag, offset_t offset,
 171     flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct)
 172 {
 173         int frcmd;
 174         int error = 0;
 175         boolean_t skip_lock = B_FALSE;
 176         flk_callback_t serialize_callback;
 177         int serialize = 0;
 178         v_mode_t mode;
 179
 180         switch (cmd) {
 181
 182         case F_GETLK:
 183         case F_O_GETLK:
 184                 if (flag & F_REMOTELOCK) {
 185                         frcmd = RCMDLCK;
 186                 } else {
 187                         frcmd = 0;
 188                         bfp->l_pid = ttoproc(curthread)->p_pid;
 189                         bfp->l_sysid = 0;
 190                 }
 191                 break;
 192
 193         case F_OFD_GETLK:
 194                 /*
 195                  * TBD we do not support remote OFD locks at this time.
 196                  */
 197                 if (flag & F_REMOTELOCK) {
 198                         error = EINVAL;
 199                         goto done;
 200                 }
 201                 skip_lock = B_TRUE;
 202                 break;
 203
 204         case F_SETLK_NBMAND:
 205                 /*
 206                  * Are NBMAND locks allowed on this file?
 207                  */
 208                 if (!vp->v_vfsp ||
 209                     !(vp->v_vfsp->vfs_flag & VFS_NBMAND)) {
 210                         error = EINVAL;
 211                         goto done;
 212                 }
 213                 if (vp->v_type != VREG) {
 214                         error = EINVAL;
 215                         goto done;
 216                 }
 217                 /*FALLTHROUGH*/
 218
 219         case F_SETLK:
 220                 if (flag & F_REMOTELOCK) {
 221                         frcmd = SETFLCK|RCMDLCK;
 222                 } else {
 223                         frcmd = SETFLCK;
 224                         bfp->l_pid = ttoproc(curthread)->p_pid;
 225                         bfp->l_sysid = 0;
 226                 }
 227                 if (cmd == F_SETLK_NBMAND &&
 228                     (bfp->l_type == F_RDLCK || bfp->l_type == F_WRLCK)) {
 229                         frcmd |= NBMLCK;
 230                 }
 231
 232                 if (nbl_need_check(vp)) {
 233                         nbl_start_crit(vp, RW_WRITER);
 234                         serialize = 1;
 235                         if (frcmd & NBMLCK) {
 236                                 mode = (bfp->l_type == F_RDLCK) ?
 237                                     V_READ : V_RDANDWR;
 238                                 if (vn_is_mapped(vp, mode)) {
 239                                         error = EAGAIN;
 240                                         goto done;
 241                                 }
 242                         }
 243                 }
 244                 break;
 245
 246         case F_SETLKW:
 247                 if (flag & F_REMOTELOCK) {
 248                         frcmd = SETFLCK|SLPFLCK|RCMDLCK;
 249                 } else {
 250                         frcmd = SETFLCK|SLPFLCK;
 251                         bfp->l_pid = ttoproc(curthread)->p_pid;
 252                         bfp->l_sysid = 0;
 253                 }
 254
 255                 if (nbl_need_check(vp)) {
 256                         nbl_start_crit(vp, RW_WRITER);
 257                         serialize = 1;
 258                 }
 259                 break;
 260
 261         case F_OFD_SETLK:
 262         case F_OFD_SETLKW:
 263         case F_FLOCK:
 264         case F_FLOCKW:
 265                 /*
 266                  * TBD we do not support remote OFD locks at this time.
 267                  */
 268                 if (flag & F_REMOTELOCK) {
 269                         error = EINVAL;
 270                         goto done;
 271                 }
 272                 skip_lock = B_TRUE;
 273                 break;
 274
 275         case F_HASREMOTELOCKS:
 276                 l_has_rmt(bfp) = flk_has_remote_locks(vp);
 277                 goto done;
 278
 279         default:
 280                 error = EINVAL;
 281                 goto done;
 282         }
 283
 284         /*
 285          * If this is a blocking lock request and we're serializing lock
 286          * requests, modify the callback list to leave the critical region
 287          * while we're waiting for the lock.
 288          */
 289
 290         if (serialize && (frcmd & SLPFLCK) != 0) {
 291                 flk_add_callback(&serialize_callback,
 292                     frlock_serialize_blocked, vp, flk_cbp);
 293                 flk_cbp = &serialize_callback;
 294         }
 295
 296         if (!skip_lock)
 297                 error = reclock(vp, bfp, frcmd, flag, offset, flk_cbp);
 298
 299         if (serialize && (frcmd & SLPFLCK) != 0)
 300                 flk_del_callback(&serialize_callback);
 301
 302 done:
 303         if (serialize)
 304                 nbl_end_crit(vp);
 305
 306         return (error);
 307 }
 308
 309 /*
 310  * Callback when a lock request blocks and we are serializing requests.  If
 311  * before sleeping, leave the critical region.  If after wakeup, reenter
 312  * the critical region.
 313  */
 314
 315 static callb_cpr_t *
 316 frlock_serialize_blocked(flk_cb_when_t when, void *infop)
 317 {
 318         vnode_t *vp = (vnode_t *)infop;
 319
 320         if (when == FLK_BEFORE_SLEEP)
 321                 nbl_end_crit(vp);
 322         else {
 323                 nbl_start_crit(vp, RW_WRITER);
 324         }
 325
 326         return (NULL);
 327 }
 328
 329 /*
 330  * Return the answer requested to poll() for non-device files.
 331  * Only POLLIN, POLLRDNORM, and POLLOUT are recognized.
 332  */
 333 struct pollhead fs_pollhd;
 334
 335 /* ARGSUSED */
 336 int
 337 fs_poll(vnode_t *vp, short events, int anyyet, short *reventsp,
 338     struct pollhead **phpp, caller_context_t *ct)
 339 {
 340         /*
 341          * Reject all attempts for edge-triggered polling.  These should only
 342          * occur when regular files are added to a /dev/poll handle which is in
 343          * epoll mode.  The Linux epoll does not allow epoll-ing on regular
 344          * files at all, so rejecting EPOLLET requests is congruent with those
 345          * expectations.
 346          */
 347         if (events & POLLET) {
 348                 return (EPERM);
 349         }
 350
 351         *reventsp = 0;
 352         if (events & POLLIN)
 353                 *reventsp |= POLLIN;
 354         if (events & POLLRDNORM)
 355                 *reventsp |= POLLRDNORM;
 356         if (events & POLLRDBAND)
 357                 *reventsp |= POLLRDBAND;
 358         if (events & POLLOUT)
 359                 *reventsp |= POLLOUT;
 360         if (events & POLLWRBAND)
 361                 *reventsp |= POLLWRBAND;
 362         /*
 363          * Emitting a pollhead without the intention of issuing pollwakeup()
 364          * calls against it is a recipe for trouble.  It's only acceptable in
 365          * this case since the above logic matches practically all useful
 366          * events.
 367          */
 368         if (*reventsp == 0 && !anyyet)
 369                 *phpp = &fs_pollhd;
 370         return (0);
 371 }
 372
 373 /*
 374  * POSIX pathconf() support.
 375  */
 376 /* ARGSUSED */
 377 int
 378 fs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
 379     caller_context_t *ct)
 380 {
 381         ulong_t val;
 382         int error = 0;
 383         struct statvfs64 vfsbuf;
 384
 385         switch (cmd) {
 386
 387         case _PC_LINK_MAX:
 388                 val = MAXLINK;
 389                 break;
 390
 391         case _PC_MAX_CANON:
 392                 val = MAX_CANON;
 393                 break;
 394
 395         case _PC_MAX_INPUT:
 396                 val = MAX_INPUT;
 397                 break;
 398
 399         case _PC_NAME_MAX:
 400                 bzero(&vfsbuf, sizeof (vfsbuf));
 401                 if (error = VFS_STATVFS(vp->v_vfsp, &vfsbuf))
 402                         break;
 403                 val = vfsbuf.f_namemax;
 404                 break;
 405
 406         case _PC_PATH_MAX:
 407         case _PC_SYMLINK_MAX:
 408                 val = MAXPATHLEN;
 409                 break;
 410
 411         case _PC_PIPE_BUF:
 412                 val = PIPE_BUF;
 413                 break;
 414
 415         case _PC_NO_TRUNC:
 416                 if (vp->v_vfsp->vfs_flag & VFS_NOTRUNC)
 417                         val = 1;        /* NOTRUNC is enabled for vp */
 418                 else
 419                         val = (ulong_t)-1;
 420                 break;
 421
 422         case _PC_VDISABLE:
 423                 val = _POSIX_VDISABLE;
 424                 break;
 425
 426         case _PC_CHOWN_RESTRICTED:
 427                 if (rstchown)
 428                         val = rstchown; /* chown restricted enabled */
 429                 else
 430                         val = (ulong_t)-1;
 431                 break;
 432
 433         case _PC_FILESIZEBITS:
 434
 435                 /*
 436                  * If ever we come here it means that underlying file system
 437                  * does not recognise the command and therefore this
 438                  * configurable limit cannot be determined. We return -1
 439                  * and don't change errno.
 440                  */
 441
 442                 val = (ulong_t)-1;    /* large file support */
 443                 break;
 444
 445         case _PC_ACL_ENABLED:
 446                 val = 0;
 447                 break;
 448
 449         case _PC_CASE_BEHAVIOR:
 450                 val = _CASE_SENSITIVE;
 451                 if (vfs_has_feature(vp->v_vfsp, VFSFT_CASEINSENSITIVE) == 1)
 452                         val |= _CASE_INSENSITIVE;
 453                 if (vfs_has_feature(vp->v_vfsp, VFSFT_NOCASESENSITIVE) == 1)
 454                         val &= ~_CASE_SENSITIVE;
 455                 break;
 456
 457         case _PC_SATTR_ENABLED:
 458         case _PC_SATTR_EXISTS:
 459                 val = 0;
 460                 break;
 461
 462         case _PC_ACCESS_FILTERING:
 463                 val = 0;
 464                 break;
 465
 466         default:
 467                 error = EINVAL;
 468                 break;
 469         }
 470
 471         if (error == 0)
 472                 *valp = val;
 473         return (error);
 474 }
 475
 476 /*
 477  * Dispose of a page.
 478  */
 479 /* ARGSUSED */
 480 void
 481 fs_dispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 482     caller_context_t *ct)
 483 {
 484
 485         ASSERT(fl == B_FREE || fl == B_INVAL);
 486
 487         if (fl == B_FREE)
 488                 page_free(pp, dn);
 489         else
 490                 page_destroy(pp, dn);
 491 }
 492
 493 /* ARGSUSED */
 494 void
 495 fs_nodispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 496     caller_context_t *ct)
 497 {
 498         cmn_err(CE_PANIC, "fs_nodispose invoked");
 499 }
 500
 501 /*
 502  * fabricate acls for file systems that do not support acls.
 503  */
 504 /* ARGSUSED */
 505 int
 506 fs_fab_acl(vnode_t *vp, vsecattr_t *vsecattr, int flag, cred_t *cr,
 507     caller_context_t *ct)
 508 {
 509         aclent_t        *aclentp;
 510         struct vattr    vattr;
 511         int             error;
 512         size_t          aclsize;
 513
 514         vsecattr->vsa_aclcnt    = 0;
 515         vsecattr->vsa_aclentsz  = 0;
 516         vsecattr->vsa_aclentp   = NULL;
 517         vsecattr->vsa_dfaclcnt  = 0;    /* Default ACLs are not fabricated */
 518         vsecattr->vsa_dfaclentp = NULL;
 519
 520         vattr.va_mask = AT_MODE | AT_UID | AT_GID;
 521         if (error = fop_getattr(vp, &vattr, 0, cr, ct))
 522                 return (error);
 523
 524         if (vsecattr->vsa_mask & (VSA_ACLCNT | VSA_ACL)) {
 525                 aclsize = 4 * sizeof (aclent_t);
 526                 vsecattr->vsa_aclcnt    = 4; /* USER, GROUP, OTHER, and CLASS */
 527                 vsecattr->vsa_aclentp = kmem_zalloc(aclsize, KM_SLEEP);
 528                 aclentp = vsecattr->vsa_aclentp;
 529
 530                 aclentp->a_type = USER_OBJ;     /* Owner */
 531                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0700)) >> 6;
 532                 aclentp->a_id = vattr.va_uid;   /* Really undefined */
 533                 aclentp++;
 534
 535                 aclentp->a_type = GROUP_OBJ;    /* Group */
 536                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0070)) >> 3;
 537                 aclentp->a_id = vattr.va_gid;   /* Really undefined */
 538                 aclentp++;
 539
 540                 aclentp->a_type = OTHER_OBJ;    /* Other */
 541                 aclentp->a_perm = vattr.va_mode & 0007;
 542                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 543                 aclentp++;
 544
 545                 aclentp->a_type = CLASS_OBJ;    /* Class */
 546                 aclentp->a_perm = (ushort_t)(0007);
 547                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 548         } else if (vsecattr->vsa_mask & (VSA_ACECNT | VSA_ACE)) {
 549                 VERIFY(0 == acl_trivial_create(vattr.va_mode,
 550                     (vp->v_type == VDIR), (ace_t **)&vsecattr->vsa_aclentp,
 551                     &vsecattr->vsa_aclcnt));
 552                 vsecattr->vsa_aclentsz = vsecattr->vsa_aclcnt * sizeof (ace_t);
 553         }
 554
 555         return (error);
 556 }
 557
 558 /*
 559  * Common code for implementing DOS share reservations
 560  */
 561 /* ARGSUSED4 */
 562 int
 563 fs_shrlock(struct vnode *vp, int cmd, struct shrlock *shr, int flag, cred_t *cr,
 564     caller_context_t *ct)
 565 {
 566         int error;
 567
 568         /*
 569          * Make sure that the file was opened with permissions appropriate
 570          * for the request, and make sure the caller isn't trying to sneak
 571          * in an NBMAND request.
 572          */
 573         if (cmd == F_SHARE) {
 574                 if (((shr->s_access & F_RDACC) && (flag & FREAD) == 0) ||
 575                     ((shr->s_access & F_WRACC) && (flag & FWRITE) == 0))
 576                         return (EBADF);
 577                 if (shr->s_access & (F_RMACC | F_MDACC))
 578                         return (EINVAL);
 579                 if (shr->s_deny & (F_MANDDNY | F_RMDNY))
 580                         return (EINVAL);
 581         }
 582         if (cmd == F_SHARE_NBMAND) {
 583                 /* make sure nbmand is allowed on the file */
 584                 if (!vp->v_vfsp ||
 585                     !(vp->v_vfsp->vfs_flag & VFS_NBMAND)) {
 586                         return (EINVAL);
 587                 }
 588                 if (vp->v_type != VREG) {
 589                         return (EINVAL);
 590                 }
 591         }
 592
 593         nbl_start_crit(vp, RW_WRITER);
 594
 595         switch (cmd) {
 596
 597         case F_SHARE_NBMAND:
 598                 shr->s_deny |= F_MANDDNY;
 599                 /*FALLTHROUGH*/
 600         case F_SHARE:
 601                 error = add_share(vp, shr);
 602                 break;
 603
 604         case F_UNSHARE:
 605                 error = del_share(vp, shr);
 606                 break;
 607
 608         case F_HASREMOTELOCKS:
 609                 /*
 610                  * We are overloading this command to refer to remote
 611                  * shares as well as remote locks, despite its name.
 612                  */
 613                 shr->s_access = shr_has_remote_shares(vp, shr->s_sysid);
 614                 error = 0;
 615                 break;
 616
 617         default:
 618                 error = EINVAL;
 619                 break;
 620         }
 621
 622         nbl_end_crit(vp);
 623         return (error);
 624 }
 625
 626 /*ARGSUSED1*/
 627 int
 628 fs_vnevent_support(vnode_t *vp, vnevent_t e, vnode_t *dvp, char *fnm,
 629     caller_context_t *ct)
 630 {
 631         ASSERT(vp != NULL);
 632         return (0);
 633 }
 634
 635 /*
 636  * return 1 for non-trivial ACL.
 637  *
 638  * NB: It is not necessary for the caller to fop_rwlock since
 639  *      we only issue fop_getsecattr.
 640  *
 641  * Returns 0 == trivial
 642  *         1 == NOT Trivial
 643  *         <0 could not determine.
 644  */
 645 int
 646 fs_acl_nontrivial(vnode_t *vp, cred_t *cr)
 647 {
 648         ulong_t         acl_styles;
 649         ulong_t         acl_flavor;
 650         vsecattr_t      vsecattr;
 651         int             error;
 652         int             isnontrivial;
 653
 654         /* determine the forms of ACLs maintained */
 655         error = fop_pathconf(vp, _PC_ACL_ENABLED, &acl_styles, cr, NULL);
 656
 657         /* clear bits we don't understand and establish default acl_style */
 658         acl_styles &= (_ACL_ACLENT_ENABLED | _ACL_ACE_ENABLED);
 659         if (error || (acl_styles == 0))
 660                 acl_styles = _ACL_ACLENT_ENABLED;
 661
 662         vsecattr.vsa_aclentp = NULL;
 663         vsecattr.vsa_dfaclentp = NULL;
 664         vsecattr.vsa_aclcnt = 0;
 665         vsecattr.vsa_dfaclcnt = 0;
 666
 667         while (acl_styles) {
 668                 /* select one of the styles as current flavor */
 669                 acl_flavor = 0;
 670                 if (acl_styles & _ACL_ACLENT_ENABLED) {
 671                         acl_flavor = _ACL_ACLENT_ENABLED;
 672                         vsecattr.vsa_mask = VSA_ACLCNT | VSA_DFACLCNT;
 673                 } else if (acl_styles & _ACL_ACE_ENABLED) {
 674                         acl_flavor = _ACL_ACE_ENABLED;
 675                         vsecattr.vsa_mask = VSA_ACECNT | VSA_ACE;
 676                 }
 677
 678                 ASSERT(vsecattr.vsa_mask && acl_flavor);
 679                 error = fop_getsecattr(vp, &vsecattr, 0, cr, NULL);
 680                 if (error == 0)
 681                         break;
 682
 683                 /* that flavor failed */
 684                 acl_styles &= ~acl_flavor;
 685         }
 686
 687         /* if all styles fail then assume trivial */
 688         if (acl_styles == 0)
 689                 return (0);
 690
 691         /* process the flavor that worked */
 692         isnontrivial = 0;
 693         if (acl_flavor & _ACL_ACLENT_ENABLED) {
 694                 if (vsecattr.vsa_aclcnt > MIN_ACL_ENTRIES)
 695                         isnontrivial = 1;
 696                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 697                         kmem_free(vsecattr.vsa_aclentp,
 698                             vsecattr.vsa_aclcnt * sizeof (aclent_t));
 699                 if (vsecattr.vsa_dfaclcnt && vsecattr.vsa_dfaclentp != NULL)
 700                         kmem_free(vsecattr.vsa_dfaclentp,
 701                             vsecattr.vsa_dfaclcnt * sizeof (aclent_t));
 702         }
 703         if (acl_flavor & _ACL_ACE_ENABLED) {
 704                 isnontrivial = ace_trivial(vsecattr.vsa_aclentp,
 705                     vsecattr.vsa_aclcnt);
 706
 707                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 708                         kmem_free(vsecattr.vsa_aclentp,
 709                             vsecattr.vsa_aclcnt * sizeof (ace_t));
 710                 /* ACE has no vsecattr.vsa_dfaclcnt */
 711         }
 712         return (isnontrivial);
 713 }
 714
 715 /*
 716  * Check whether we need a retry to recover from STALE error.
 717  */
 718 int
 719 fs_need_estale_retry(int retry_count)
 720 {
 721         if (retry_count < fs_estale_retry)
 722                 return (1);
 723         else
 724                 return (0);
 725 }
 726
 727
 728 static int (*fs_av_scan)(vnode_t *, cred_t *, int) = NULL;
 729
 730 /*
 731  * Routine for anti-virus scanner to call to register its scanning routine.
 732  */
 733 void
 734 fs_vscan_register(int (*av_scan)(vnode_t *, cred_t *, int))
 735 {
 736         fs_av_scan = av_scan;
 737 }
 738
 739 /*
 740  * Routine for file systems to call to initiate anti-virus scanning.
 741  * Scanning will only be done on REGular files (currently).
 742  */
 743 int
 744 fs_vscan(vnode_t *vp, cred_t *cr, int async)
 745 {
 746         int ret = 0;
 747
 748         if (fs_av_scan && vp->v_type == VREG)
 749                 ret = (*fs_av_scan)(vp, cr, async);
 750
 751         return (ret);
 752 }
 753
 754 /*
 755  * support functions for reparse point
 756  */
 757 /*
 758  * reparse_vnode_parse
 759  *
 760  * Read the symlink data of a reparse point specified by the vnode
 761  * and return the reparse data as name-value pair in the nvlist.
 762  */
 763 int
 764 reparse_vnode_parse(vnode_t *vp, nvlist_t *nvl)
 765 {
 766         int err;
 767         char *lkdata;
 768         struct uio uio;
 769         struct iovec iov;
 770
 771         if (vp == NULL || nvl == NULL)
 772                 return (EINVAL);
 773
 774         lkdata = kmem_alloc(MAXREPARSELEN, KM_SLEEP);
 775
 776         /*
 777          * Set up io vector to read sym link data
 778          */
 779         iov.iov_base = lkdata;
 780         iov.iov_len = MAXREPARSELEN;
 781         uio.uio_iov = &iov;
 782         uio.uio_iovcnt = 1;
 783         uio.uio_segflg = UIO_SYSSPACE;
 784         uio.uio_extflg = UIO_COPY_CACHED;
 785         uio.uio_loffset = 0;
 786         uio.uio_resid = MAXREPARSELEN;
 787
 788         if ((err = fop_readlink(vp, &uio, kcred, NULL)) == 0) {
 789                 *(lkdata + MAXREPARSELEN - uio.uio_resid) = '\0';
 790                 err = reparse_parse(lkdata, nvl);
 791         }
 792         kmem_free(lkdata, MAXREPARSELEN);       /* done with lkdata */
 793
 794         return (err);
 795 }
 796
 797 void
 798 reparse_point_init()
 799 {
 800         mutex_init(&reparsed_door_lock, NULL, MUTEX_DEFAULT, NULL);
 801 }
 802
 803 static door_handle_t
 804 reparse_door_get_handle()
 805 {
 806         door_handle_t dh;
 807
 808         mutex_enter(&reparsed_door_lock);
 809         if ((dh = reparsed_door) == NULL) {
 810                 if (door_ki_open(REPARSED_DOOR, &reparsed_door) != 0) {
 811                         reparsed_door = NULL;
 812                         dh = NULL;
 813                 } else
 814                         dh = reparsed_door;
 815         }
 816         mutex_exit(&reparsed_door_lock);
 817         return (dh);
 818 }
 819
 820 static void
 821 reparse_door_reset_handle()
 822 {
 823         mutex_enter(&reparsed_door_lock);
 824         reparsed_door = NULL;
 825         mutex_exit(&reparsed_door_lock);
 826 }
 827
 828 /*
 829  * reparse_kderef
 830  *
 831  * Accepts the service-specific item from the reparse point and returns
 832  * the service-specific data requested.  The caller specifies the size of
 833  * the buffer provided via *bufsz; the routine will fail with EOVERFLOW
 834  * if the results will not fit in the buffer, in which case, *bufsz will
 835  * contain the number of bytes needed to hold the results.
 836  *
 837  * if ok return 0 and update *bufsize with length of actual result
 838  * else return error code.
 839  */
 840 int
 841 reparse_kderef(const char *svc_type, const char *svc_data, char *buf,
 842     size_t *bufsize)
 843 {
 844         int err, retries, need_free, retried_doorhd;
 845         size_t dlen, res_len;
 846         char *darg;
 847         door_arg_t door_args;
 848         reparsed_door_res_t *resp;
 849         door_handle_t rp_door;
 850
 851         if (svc_type == NULL || svc_data == NULL || buf == NULL ||
 852             bufsize == NULL)
 853                 return (EINVAL);
 854
 855         /* get reparsed's door handle */
 856         if ((rp_door = reparse_door_get_handle()) == NULL)
 857                 return (EBADF);
 858
 859         /* setup buffer for door_call args and results */
 860         dlen = strlen(svc_type) + strlen(svc_data) + 2;
 861         if (*bufsize < dlen) {
 862                 darg = kmem_alloc(dlen, KM_SLEEP);
 863                 need_free = 1;
 864         } else {
 865                 darg = buf;     /* use same buffer for door's args & results */
 866                 need_free = 0;
 867         }
 868
 869         /* build argument string of door call */
 870         (void) snprintf(darg, dlen, "%s:%s", svc_type, svc_data);
 871
 872         /* setup args for door call */
 873         door_args.data_ptr = darg;
 874         door_args.data_size = dlen;
 875         door_args.desc_ptr = NULL;
 876         door_args.desc_num = 0;
 877         door_args.rbuf = buf;
 878         door_args.rsize = *bufsize;
 879
 880         /* do the door_call */
 881         retried_doorhd = 0;
 882         retries = 0;
 883         door_ki_hold(rp_door);
 884         while ((err = door_ki_upcall_limited(rp_door, &door_args,
 885             NULL, SIZE_MAX, 0)) != 0) {
 886                 if (err == EAGAIN || err == EINTR) {
 887                         if (++retries < REPARSED_DOORCALL_MAX_RETRY) {
 888                                 ddi_sleep(1);
 889                                 continue;
 890                         }
 891                 } else if (err == EBADF) {
 892                         /* door server goes away... */
 893                         reparse_door_reset_handle();
 894
 895                         if (retried_doorhd == 0) {
 896                                 door_ki_rele(rp_door);
 897                                 retried_doorhd++;
 898                                 rp_door = reparse_door_get_handle();
 899                                 if (rp_door != NULL) {
 900                                         door_ki_hold(rp_door);
 901                                         continue;
 902                                 }
 903                         }
 904                 }
 905                 break;
 906         }
 907
 908         if (rp_door)
 909                 door_ki_rele(rp_door);
 910
 911         if (need_free)
 912                 kmem_free(darg, dlen);          /* done with args buffer */
 913
 914         if (err != 0)
 915                 return (err);
 916
 917         resp = (reparsed_door_res_t *)door_args.rbuf;
 918         if ((err = resp->res_status) == 0) {
 919                 /*
 920                  * have to save the length of the results before the
 921                  * bcopy below since it's can be an overlap copy that
 922                  * overwrites the reparsed_door_res_t structure at
 923                  * the beginning of the buffer.
 924                  */
 925                 res_len = (size_t)resp->res_len;
 926
 927                 /* deref call is ok */
 928                 if (res_len > *bufsize)
 929                         err = EOVERFLOW;
 930                 else
 931                         bcopy(resp->res_data, buf, res_len);
 932                 *bufsize = res_len;
 933         }
 934         if (door_args.rbuf != buf)
 935                 kmem_free(door_args.rbuf, door_args.rsize);
 936
 937         return (err);
 938 }