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_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         *reventsp = 0;
 341         if (events & POLLIN)
 342                 *reventsp |= POLLIN;
 343         if (events & POLLRDNORM)
 344                 *reventsp |= POLLRDNORM;
 345         if (events & POLLRDBAND)
 346                 *reventsp |= POLLRDBAND;
 347         if (events & POLLOUT)
 348                 *reventsp |= POLLOUT;
 349         if (events & POLLWRBAND)
 350                 *reventsp |= POLLWRBAND;
 351         *phpp = !anyyet && !*reventsp ? &fs_pollhd : NULL;
 352         return (0);
 353 }
 354
 355 /*
 356  * POSIX pathconf() support.
 357  */
 358 /* ARGSUSED */
 359 int
 360 fs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
 361     caller_context_t *ct)
 362 {
 363         ulong_t val;
 364         int error = 0;
 365         struct statvfs64 vfsbuf;
 366
 367         switch (cmd) {
 368
 369         case _PC_LINK_MAX:
 370                 val = MAXLINK;
 371                 break;
 372
 373         case _PC_MAX_CANON:
 374                 val = MAX_CANON;
 375                 break;
 376
 377         case _PC_MAX_INPUT:
 378                 val = MAX_INPUT;
 379                 break;
 380
 381         case _PC_NAME_MAX:
 382                 bzero(&vfsbuf, sizeof (vfsbuf));
 383                 if (error = VFS_STATVFS(vp->v_vfsp, &vfsbuf))
 384                         break;
 385                 val = vfsbuf.f_namemax;
 386                 break;
 387
 388         case _PC_PATH_MAX:
 389         case _PC_SYMLINK_MAX:
 390                 val = MAXPATHLEN;
 391                 break;
 392
 393         case _PC_PIPE_BUF:
 394                 val = PIPE_BUF;
 395                 break;
 396
 397         case _PC_NO_TRUNC:
 398                 if (vp->v_vfsp->vfs_flag & VFS_NOTRUNC)
 399                         val = 1;        /* NOTRUNC is enabled for vp */
 400                 else
 401                         val = (ulong_t)-1;
 402                 break;
 403
 404         case _PC_VDISABLE:
 405                 val = _POSIX_VDISABLE;
 406                 break;
 407
 408         case _PC_CHOWN_RESTRICTED:
 409                 if (rstchown)
 410                         val = rstchown; /* chown restricted enabled */
 411                 else
 412                         val = (ulong_t)-1;
 413                 break;
 414
 415         case _PC_FILESIZEBITS:
 416
 417                 /*
 418                  * If ever we come here it means that underlying file system
 419                  * does not recognise the command and therefore this
 420                  * configurable limit cannot be determined. We return -1
 421                  * and don't change errno.
 422                  */
 423
 424                 val = (ulong_t)-1;    /* large file support */
 425                 break;
 426
 427         case _PC_ACL_ENABLED:
 428                 val = 0;
 429                 break;
 430
 431         case _PC_CASE_BEHAVIOR:
 432                 val = _CASE_SENSITIVE;
 433                 if (vfs_has_feature(vp->v_vfsp, VFSFT_CASEINSENSITIVE) == 1)
 434                         val |= _CASE_INSENSITIVE;
 435                 if (vfs_has_feature(vp->v_vfsp, VFSFT_NOCASESENSITIVE) == 1)
 436                         val &= ~_CASE_SENSITIVE;
 437                 break;
 438
 439         case _PC_SATTR_ENABLED:
 440         case _PC_SATTR_EXISTS:
 441                 val = 0;
 442                 break;
 443
 444         case _PC_ACCESS_FILTERING:
 445                 val = 0;
 446                 break;
 447
 448         default:
 449                 error = EINVAL;
 450                 break;
 451         }
 452
 453         if (error == 0)
 454                 *valp = val;
 455         return (error);
 456 }
 457
 458 /*
 459  * Dispose of a page.
 460  */
 461 /* ARGSUSED */
 462 void
 463 fs_dispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 464     caller_context_t *ct)
 465 {
 466
 467         ASSERT(fl == B_FREE || fl == B_INVAL);
 468
 469         if (fl == B_FREE)
 470                 page_free(pp, dn);
 471         else
 472                 page_destroy(pp, dn);
 473 }
 474
 475 /* ARGSUSED */
 476 void
 477 fs_nodispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 478     caller_context_t *ct)
 479 {
 480         cmn_err(CE_PANIC, "fs_nodispose invoked");
 481 }
 482
 483 /*
 484  * fabricate acls for file systems that do not support acls.
 485  */
 486 /* ARGSUSED */
 487 int
 488 fs_fab_acl(vnode_t *vp, vsecattr_t *vsecattr, int flag, cred_t *cr,
 489     caller_context_t *ct)
 490 {
 491         aclent_t        *aclentp;
 492         struct vattr    vattr;
 493         int             error;
 494         size_t          aclsize;
 495
 496         vsecattr->vsa_aclcnt    = 0;
 497         vsecattr->vsa_aclentsz  = 0;
 498         vsecattr->vsa_aclentp   = NULL;
 499         vsecattr->vsa_dfaclcnt  = 0;    /* Default ACLs are not fabricated */
 500         vsecattr->vsa_dfaclentp = NULL;
 501
 502         vattr.va_mask = AT_MODE | AT_UID | AT_GID;
 503         if (error = fop_getattr(vp, &vattr, 0, cr, ct))
 504                 return (error);
 505
 506         if (vsecattr->vsa_mask & (VSA_ACLCNT | VSA_ACL)) {
 507                 aclsize = 4 * sizeof (aclent_t);
 508                 vsecattr->vsa_aclcnt    = 4; /* USER, GROUP, OTHER, and CLASS */
 509                 vsecattr->vsa_aclentp = kmem_zalloc(aclsize, KM_SLEEP);
 510                 aclentp = vsecattr->vsa_aclentp;
 511
 512                 aclentp->a_type = USER_OBJ;     /* Owner */
 513                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0700)) >> 6;
 514                 aclentp->a_id = vattr.va_uid;   /* Really undefined */
 515                 aclentp++;
 516
 517                 aclentp->a_type = GROUP_OBJ;    /* Group */
 518                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0070)) >> 3;
 519                 aclentp->a_id = vattr.va_gid;   /* Really undefined */
 520                 aclentp++;
 521
 522                 aclentp->a_type = OTHER_OBJ;    /* Other */
 523                 aclentp->a_perm = vattr.va_mode & 0007;
 524                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 525                 aclentp++;
 526
 527                 aclentp->a_type = CLASS_OBJ;    /* Class */
 528                 aclentp->a_perm = (ushort_t)(0007);
 529                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 530         } else if (vsecattr->vsa_mask & (VSA_ACECNT | VSA_ACE)) {
 531                 VERIFY(0 == acl_trivial_create(vattr.va_mode,
 532                     (vp->v_type == VDIR), (ace_t **)&vsecattr->vsa_aclentp,
 533                     &vsecattr->vsa_aclcnt));
 534                 vsecattr->vsa_aclentsz = vsecattr->vsa_aclcnt * sizeof (ace_t);
 535         }
 536
 537         return (error);
 538 }
 539
 540 /*
 541  * Common code for implementing DOS share reservations
 542  */
 543 /* ARGSUSED4 */
 544 int
 545 fs_shrlock(struct vnode *vp, int cmd, struct shrlock *shr, int flag, cred_t *cr,
 546     caller_context_t *ct)
 547 {
 548         int error;
 549
 550         /*
 551          * Make sure that the file was opened with permissions appropriate
 552          * for the request, and make sure the caller isn't trying to sneak
 553          * in an NBMAND request.
 554          */
 555         if (cmd == F_SHARE) {
 556                 if (((shr->s_access & F_RDACC) && (flag & FREAD) == 0) ||
 557                     ((shr->s_access & F_WRACC) && (flag & FWRITE) == 0))
 558                         return (EBADF);
 559                 if (shr->s_access & (F_RMACC | F_MDACC))
 560                         return (EINVAL);
 561                 if (shr->s_deny & (F_MANDDNY | F_RMDNY))
 562                         return (EINVAL);
 563         }
 564         if (cmd == F_SHARE_NBMAND) {
 565                 /* make sure nbmand is allowed on the file */
 566                 if (!vp->v_vfsp ||
 567                     !(vp->v_vfsp->vfs_flag & VFS_NBMAND)) {
 568                         return (EINVAL);
 569                 }
 570                 if (vp->v_type != VREG) {
 571                         return (EINVAL);
 572                 }
 573         }
 574
 575         nbl_start_crit(vp, RW_WRITER);
 576
 577         switch (cmd) {
 578
 579         case F_SHARE_NBMAND:
 580                 shr->s_deny |= F_MANDDNY;
 581                 /*FALLTHROUGH*/
 582         case F_SHARE:
 583                 error = add_share(vp, shr);
 584                 break;
 585
 586         case F_UNSHARE:
 587                 error = del_share(vp, shr);
 588                 break;
 589
 590         case F_HASREMOTELOCKS:
 591                 /*
 592                  * We are overloading this command to refer to remote
 593                  * shares as well as remote locks, despite its name.
 594                  */
 595                 shr->s_access = shr_has_remote_shares(vp, shr->s_sysid);
 596                 error = 0;
 597                 break;
 598
 599         default:
 600                 error = EINVAL;
 601                 break;
 602         }
 603
 604         nbl_end_crit(vp);
 605         return (error);
 606 }
 607
 608 /*ARGSUSED1*/
 609 int
 610 fs_vnevent_support(vnode_t *vp, vnevent_t e, vnode_t *dvp, char *fnm,
 611     caller_context_t *ct)
 612 {
 613         ASSERT(vp != NULL);
 614         return (0);
 615 }
 616
 617 /*
 618  * return 1 for non-trivial ACL.
 619  *
 620  * NB: It is not necessary for the caller to fop_rwlock since
 621  *      we only issue fop_getsecattr.
 622  *
 623  * Returns 0 == trivial
 624  *         1 == NOT Trivial
 625  *         <0 could not determine.
 626  */
 627 int
 628 fs_acl_nontrivial(vnode_t *vp, cred_t *cr)
 629 {
 630         ulong_t         acl_styles;
 631         ulong_t         acl_flavor;
 632         vsecattr_t      vsecattr;
 633         int             error;
 634         int             isnontrivial;
 635
 636         /* determine the forms of ACLs maintained */
 637         error = fop_pathconf(vp, _PC_ACL_ENABLED, &acl_styles, cr, NULL);
 638
 639         /* clear bits we don't understand and establish default acl_style */
 640         acl_styles &= (_ACL_ACLENT_ENABLED | _ACL_ACE_ENABLED);
 641         if (error || (acl_styles == 0))
 642                 acl_styles = _ACL_ACLENT_ENABLED;
 643
 644         vsecattr.vsa_aclentp = NULL;
 645         vsecattr.vsa_dfaclentp = NULL;
 646         vsecattr.vsa_aclcnt = 0;
 647         vsecattr.vsa_dfaclcnt = 0;
 648
 649         while (acl_styles) {
 650                 /* select one of the styles as current flavor */
 651                 acl_flavor = 0;
 652                 if (acl_styles & _ACL_ACLENT_ENABLED) {
 653                         acl_flavor = _ACL_ACLENT_ENABLED;
 654                         vsecattr.vsa_mask = VSA_ACLCNT | VSA_DFACLCNT;
 655                 } else if (acl_styles & _ACL_ACE_ENABLED) {
 656                         acl_flavor = _ACL_ACE_ENABLED;
 657                         vsecattr.vsa_mask = VSA_ACECNT | VSA_ACE;
 658                 }
 659
 660                 ASSERT(vsecattr.vsa_mask && acl_flavor);
 661                 error = fop_getsecattr(vp, &vsecattr, 0, cr, NULL);
 662                 if (error == 0)
 663                         break;
 664
 665                 /* that flavor failed */
 666                 acl_styles &= ~acl_flavor;
 667         }
 668
 669         /* if all styles fail then assume trivial */
 670         if (acl_styles == 0)
 671                 return (0);
 672
 673         /* process the flavor that worked */
 674         isnontrivial = 0;
 675         if (acl_flavor & _ACL_ACLENT_ENABLED) {
 676                 if (vsecattr.vsa_aclcnt > MIN_ACL_ENTRIES)
 677                         isnontrivial = 1;
 678                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 679                         kmem_free(vsecattr.vsa_aclentp,
 680                             vsecattr.vsa_aclcnt * sizeof (aclent_t));
 681                 if (vsecattr.vsa_dfaclcnt && vsecattr.vsa_dfaclentp != NULL)
 682                         kmem_free(vsecattr.vsa_dfaclentp,
 683                             vsecattr.vsa_dfaclcnt * sizeof (aclent_t));
 684         }
 685         if (acl_flavor & _ACL_ACE_ENABLED) {
 686                 isnontrivial = ace_trivial(vsecattr.vsa_aclentp,
 687                     vsecattr.vsa_aclcnt);
 688
 689                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 690                         kmem_free(vsecattr.vsa_aclentp,
 691                             vsecattr.vsa_aclcnt * sizeof (ace_t));
 692                 /* ACE has no vsecattr.vsa_dfaclcnt */
 693         }
 694         return (isnontrivial);
 695 }
 696
 697 /*
 698  * Check whether we need a retry to recover from STALE error.
 699  */
 700 int
 701 fs_need_estale_retry(int retry_count)
 702 {
 703         if (retry_count < fs_estale_retry)
 704                 return (1);
 705         else
 706                 return (0);
 707 }
 708
 709
 710 static int (*fs_av_scan)(vnode_t *, cred_t *, int) = NULL;
 711
 712 /*
 713  * Routine for anti-virus scanner to call to register its scanning routine.
 714  */
 715 void
 716 fs_vscan_register(int (*av_scan)(vnode_t *, cred_t *, int))
 717 {
 718         fs_av_scan = av_scan;
 719 }
 720
 721 /*
 722  * Routine for file systems to call to initiate anti-virus scanning.
 723  * Scanning will only be done on REGular files (currently).
 724  */
 725 int
 726 fs_vscan(vnode_t *vp, cred_t *cr, int async)
 727 {
 728         int ret = 0;
 729
 730         if (fs_av_scan && vp->v_type == VREG)
 731                 ret = (*fs_av_scan)(vp, cr, async);
 732
 733         return (ret);
 734 }
 735
 736 /*
 737  * support functions for reparse point
 738  */
 739 /*
 740  * reparse_vnode_parse
 741  *
 742  * Read the symlink data of a reparse point specified by the vnode
 743  * and return the reparse data as name-value pair in the nvlist.
 744  */
 745 int
 746 reparse_vnode_parse(vnode_t *vp, nvlist_t *nvl)
 747 {
 748         int err;
 749         char *lkdata;
 750         struct uio uio;
 751         struct iovec iov;
 752
 753         if (vp == NULL || nvl == NULL)
 754                 return (EINVAL);
 755
 756         lkdata = kmem_alloc(MAXREPARSELEN, KM_SLEEP);
 757
 758         /*
 759          * Set up io vector to read sym link data
 760          */
 761         iov.iov_base = lkdata;
 762         iov.iov_len = MAXREPARSELEN;
 763         uio.uio_iov = &iov;
 764         uio.uio_iovcnt = 1;
 765         uio.uio_segflg = UIO_SYSSPACE;
 766         uio.uio_extflg = UIO_COPY_CACHED;
 767         uio.uio_loffset = 0;
 768         uio.uio_resid = MAXREPARSELEN;
 769
 770         if ((err = fop_readlink(vp, &uio, kcred, NULL)) == 0) {
 771                 *(lkdata + MAXREPARSELEN - uio.uio_resid) = '\0';
 772                 err = reparse_parse(lkdata, nvl);
 773         }
 774         kmem_free(lkdata, MAXREPARSELEN);       /* done with lkdata */
 775
 776         return (err);
 777 }
 778
 779 void
 780 reparse_point_init()
 781 {
 782         mutex_init(&reparsed_door_lock, NULL, MUTEX_DEFAULT, NULL);
 783 }
 784
 785 static door_handle_t
 786 reparse_door_get_handle()
 787 {
 788         door_handle_t dh;
 789
 790         mutex_enter(&reparsed_door_lock);
 791         if ((dh = reparsed_door) == NULL) {
 792                 if (door_ki_open(REPARSED_DOOR, &reparsed_door) != 0) {
 793                         reparsed_door = NULL;
 794                         dh = NULL;
 795                 } else
 796                         dh = reparsed_door;
 797         }
 798         mutex_exit(&reparsed_door_lock);
 799         return (dh);
 800 }
 801
 802 static void
 803 reparse_door_reset_handle()
 804 {
 805         mutex_enter(&reparsed_door_lock);
 806         reparsed_door = NULL;
 807         mutex_exit(&reparsed_door_lock);
 808 }
 809
 810 /*
 811  * reparse_kderef
 812  *
 813  * Accepts the service-specific item from the reparse point and returns
 814  * the service-specific data requested.  The caller specifies the size of
 815  * the buffer provided via *bufsz; the routine will fail with EOVERFLOW
 816  * if the results will not fit in the buffer, in which case, *bufsz will
 817  * contain the number of bytes needed to hold the results.
 818  *
 819  * if ok return 0 and update *bufsize with length of actual result
 820  * else return error code.
 821  */
 822 int
 823 reparse_kderef(const char *svc_type, const char *svc_data, char *buf,
 824     size_t *bufsize)
 825 {
 826         int err, retries, need_free, retried_doorhd;
 827         size_t dlen, res_len;
 828         char *darg;
 829         door_arg_t door_args;
 830         reparsed_door_res_t *resp;
 831         door_handle_t rp_door;
 832
 833         if (svc_type == NULL || svc_data == NULL || buf == NULL ||
 834             bufsize == NULL)
 835                 return (EINVAL);
 836
 837         /* get reparsed's door handle */
 838         if ((rp_door = reparse_door_get_handle()) == NULL)
 839                 return (EBADF);
 840
 841         /* setup buffer for door_call args and results */
 842         dlen = strlen(svc_type) + strlen(svc_data) + 2;
 843         if (*bufsize < dlen) {
 844                 darg = kmem_alloc(dlen, KM_SLEEP);
 845                 need_free = 1;
 846         } else {
 847                 darg = buf;     /* use same buffer for door's args & results */
 848                 need_free = 0;
 849         }
 850
 851         /* build argument string of door call */
 852         (void) snprintf(darg, dlen, "%s:%s", svc_type, svc_data);
 853
 854         /* setup args for door call */
 855         door_args.data_ptr = darg;
 856         door_args.data_size = dlen;
 857         door_args.desc_ptr = NULL;
 858         door_args.desc_num = 0;
 859         door_args.rbuf = buf;
 860         door_args.rsize = *bufsize;
 861
 862         /* do the door_call */
 863         retried_doorhd = 0;
 864         retries = 0;
 865         door_ki_hold(rp_door);
 866         while ((err = door_ki_upcall_limited(rp_door, &door_args,
 867             NULL, SIZE_MAX, 0)) != 0) {
 868                 if (err == EAGAIN || err == EINTR) {
 869                         if (++retries < REPARSED_DOORCALL_MAX_RETRY) {
 870                                 ddi_sleep(1);
 871                                 continue;
 872                         }
 873                 } else if (err == EBADF) {
 874                         /* door server goes away... */
 875                         reparse_door_reset_handle();
 876
 877                         if (retried_doorhd == 0) {
 878                                 door_ki_rele(rp_door);
 879                                 retried_doorhd++;
 880                                 rp_door = reparse_door_get_handle();
 881                                 if (rp_door != NULL) {
 882                                         door_ki_hold(rp_door);
 883                                         continue;
 884                                 }
 885                         }
 886                 }
 887                 break;
 888         }
 889
 890         if (rp_door)
 891                 door_ki_rele(rp_door);
 892
 893         if (need_free)
 894                 kmem_free(darg, dlen);          /* done with args buffer */
 895
 896         if (err != 0)
 897                 return (err);
 898
 899         resp = (reparsed_door_res_t *)door_args.rbuf;
 900         if ((err = resp->res_status) == 0) {
 901                 /*
 902                  * have to save the length of the results before the
 903                  * bcopy below since it's can be an overlap copy that
 904                  * overwrites the reparsed_door_res_t structure at
 905                  * the beginning of the buffer.
 906                  */
 907                 res_len = (size_t)resp->res_len;
 908
 909                 /* deref call is ok */
 910                 if (res_len > *bufsize)
 911                         err = EOVERFLOW;
 912                 else
 913                         bcopy(resp->res_data, buf, res_len);
 914                 *bufsize = res_len;
 915         }
 916         if (door_args.rbuf != buf)
 917                 kmem_free(door_args.rbuf, door_args.rsize);
 918
 919         return (err);
 920 }