usr/src/uts/common/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 <fs/fs_subr.h>
  52 #include <fs/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 VOP_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 VOP_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 VOP_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 nlmid;
 230         int error = 0;
 231         boolean_t skip_lock = B_FALSE;
 232         flk_callback_t serialize_callback;
 233         int serialize = 0;
 234         v_mode_t mode;
 235
 236         switch (cmd) {
 237
 238         case F_GETLK:
 239         case F_O_GETLK:
 240                 if (flag & F_REMOTELOCK) {
 241                         frcmd = RCMDLCK;
 242                 } else if (flag & F_PXFSLOCK) {
 243                         frcmd = PCMDLCK;
 244                 } else {
 245                         frcmd = 0;
 246                         bfp->l_pid = ttoproc(curthread)->p_pid;
 247                         bfp->l_sysid = 0;
 248                 }
 249                 break;
 250
 251         case F_OFD_GETLK:
 252                 /*
 253                  * TBD we do not support remote OFD locks at this time.
 254                  */
 255                 if (flag & (F_REMOTELOCK | F_PXFSLOCK)) {
 256                         error = EINVAL;
 257                         goto done;
 258                 }
 259                 skip_lock = B_TRUE;
 260                 break;
 261
 262         case F_SETLK_NBMAND:
 263                 /*
 264                  * Are NBMAND locks allowed on this file?
 265                  */
 266                 if (!vp->v_vfsp ||
 267                     !(vp->v_vfsp->vfs_flag & VFS_NBMAND)) {
 268                         error = EINVAL;
 269                         goto done;
 270                 }
 271                 if (vp->v_type != VREG) {
 272                         error = EINVAL;
 273                         goto done;
 274                 }
 275                 /*FALLTHROUGH*/
 276
 277         case F_SETLK:
 278                 if (flag & F_REMOTELOCK) {
 279                         frcmd = SETFLCK|RCMDLCK;
 280                 } else if (flag & F_PXFSLOCK) {
 281                         frcmd = SETFLCK|PCMDLCK;
 282                 } else {
 283                         frcmd = SETFLCK;
 284                         bfp->l_pid = ttoproc(curthread)->p_pid;
 285                         bfp->l_sysid = 0;
 286                 }
 287                 if (cmd == F_SETLK_NBMAND &&
 288                     (bfp->l_type == F_RDLCK || bfp->l_type == F_WRLCK)) {
 289                         frcmd |= NBMLCK;
 290                 }
 291
 292                 if (nbl_need_check(vp)) {
 293                         nbl_start_crit(vp, RW_WRITER);
 294                         serialize = 1;
 295                         if (frcmd & NBMLCK) {
 296                                 mode = (bfp->l_type == F_RDLCK) ?
 297                                     V_READ : V_RDANDWR;
 298                                 if (vn_is_mapped(vp, mode)) {
 299                                         error = EAGAIN;
 300                                         goto done;
 301                                 }
 302                         }
 303                 }
 304                 break;
 305
 306         case F_SETLKW:
 307                 if (flag & F_REMOTELOCK) {
 308                         frcmd = SETFLCK|SLPFLCK|RCMDLCK;
 309                 } else if (flag & F_PXFSLOCK) {
 310                         frcmd = SETFLCK|SLPFLCK|PCMDLCK;
 311                 } else {
 312                         frcmd = SETFLCK|SLPFLCK;
 313                         bfp->l_pid = ttoproc(curthread)->p_pid;
 314                         bfp->l_sysid = 0;
 315                 }
 316
 317                 if (nbl_need_check(vp)) {
 318                         nbl_start_crit(vp, RW_WRITER);
 319                         serialize = 1;
 320                 }
 321                 break;
 322
 323         case F_OFD_SETLK:
 324         case F_OFD_SETLKW:
 325         case F_FLOCK:
 326         case F_FLOCKW:
 327                 /*
 328                  * TBD we do not support remote OFD locks at this time.
 329                  */
 330                 if (flag & (F_REMOTELOCK | F_PXFSLOCK)) {
 331                         error = EINVAL;
 332                         goto done;
 333                 }
 334                 skip_lock = B_TRUE;
 335                 break;
 336
 337         case F_HASREMOTELOCKS:
 338                 nlmid = GETNLMID(bfp->l_sysid);
 339                 if (nlmid != 0) {       /* booted as a cluster */
 340                         l_has_rmt(bfp) =
 341                             cl_flk_has_remote_locks_for_nlmid(vp, nlmid);
 342                 } else {                /* not booted as a cluster */
 343                         l_has_rmt(bfp) = flk_has_remote_locks(vp);
 344                 }
 345
 346                 goto done;
 347
 348         default:
 349                 error = EINVAL;
 350                 goto done;
 351         }
 352
 353         /*
 354          * If this is a blocking lock request and we're serializing lock
 355          * requests, modify the callback list to leave the critical region
 356          * while we're waiting for the lock.
 357          */
 358
 359         if (serialize && (frcmd & SLPFLCK) != 0) {
 360                 flk_add_callback(&serialize_callback,
 361                     frlock_serialize_blocked, vp, flk_cbp);
 362                 flk_cbp = &serialize_callback;
 363         }
 364
 365         if (!skip_lock)
 366                 error = reclock(vp, bfp, frcmd, flag, offset, flk_cbp);
 367
 368         if (serialize && (frcmd & SLPFLCK) != 0)
 369                 flk_del_callback(&serialize_callback);
 370
 371 done:
 372         if (serialize)
 373                 nbl_end_crit(vp);
 374
 375         return (error);
 376 }
 377
 378 /*
 379  * Callback when a lock request blocks and we are serializing requests.  If
 380  * before sleeping, leave the critical region.  If after wakeup, reenter
 381  * the critical region.
 382  */
 383
 384 static callb_cpr_t *
 385 frlock_serialize_blocked(flk_cb_when_t when, void *infop)
 386 {
 387         vnode_t *vp = (vnode_t *)infop;
 388
 389         if (when == FLK_BEFORE_SLEEP)
 390                 nbl_end_crit(vp);
 391         else {
 392                 nbl_start_crit(vp, RW_WRITER);
 393         }
 394
 395         return (NULL);
 396 }
 397
 398 /*
 399  * Allow any flags.
 400  */
 401 /* ARGSUSED */
 402 int
 403 fs_setfl(vnode_t *vp, int oflags, int nflags, cred_t *cr, caller_context_t *ct)
 404 {
 405         return (0);
 406 }
 407
 408 /*
 409  * Return the answer requested to poll() for non-device files.
 410  * Only POLLIN, POLLRDNORM, and POLLOUT are recognized.
 411  */
 412 struct pollhead fs_pollhd;
 413
 414 /* ARGSUSED */
 415 int
 416 fs_poll(vnode_t *vp, short events, int anyyet, short *reventsp,
 417     struct pollhead **phpp, caller_context_t *ct)
 418 {
 419         /*
 420          * Reject all attempts for edge-triggered polling.  These should only
 421          * occur when regular files are added to a /dev/poll handle which is in
 422          * epoll mode.  The Linux epoll does not allow epoll-ing on regular
 423          * files at all, so rejecting EPOLLET requests is congruent with those
 424          * expectations.
 425          */
 426         if (events & POLLET) {
 427                 return (EPERM);
 428         }
 429
 430         *reventsp = 0;
 431         if (events & POLLIN)
 432                 *reventsp |= POLLIN;
 433         if (events & POLLRDNORM)
 434                 *reventsp |= POLLRDNORM;
 435         if (events & POLLRDBAND)
 436                 *reventsp |= POLLRDBAND;
 437         if (events & POLLOUT)
 438                 *reventsp |= POLLOUT;
 439         if (events & POLLWRBAND)
 440                 *reventsp |= POLLWRBAND;
 441         /*
 442          * Emitting a pollhead without the intention of issuing pollwakeup()
 443          * calls against it is a recipe for trouble.  It's only acceptable in
 444          * this case since the above logic matches practically all useful
 445          * events.
 446          */
 447         if (*reventsp == 0 && !anyyet) {
 448                 *phpp = &fs_pollhd;
 449         }
 450         return (0);
 451 }
 452
 453 /*
 454  * POSIX pathconf() support.
 455  */
 456 /* ARGSUSED */
 457 int
 458 fs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
 459     caller_context_t *ct)
 460 {
 461         ulong_t val;
 462         int error = 0;
 463         struct statvfs64 vfsbuf;
 464
 465         switch (cmd) {
 466
 467         case _PC_LINK_MAX:
 468                 val = MAXLINK;
 469                 break;
 470
 471         case _PC_MAX_CANON:
 472                 val = MAX_CANON;
 473                 break;
 474
 475         case _PC_MAX_INPUT:
 476                 val = MAX_INPUT;
 477                 break;
 478
 479         case _PC_NAME_MAX:
 480                 bzero(&vfsbuf, sizeof (vfsbuf));
 481                 if (error = VFS_STATVFS(vp->v_vfsp, &vfsbuf))
 482                         break;
 483                 val = vfsbuf.f_namemax;
 484                 break;
 485
 486         case _PC_PATH_MAX:
 487         case _PC_SYMLINK_MAX:
 488                 val = MAXPATHLEN;
 489                 break;
 490
 491         case _PC_PIPE_BUF:
 492                 val = PIPE_BUF;
 493                 break;
 494
 495         case _PC_NO_TRUNC:
 496                 if (vp->v_vfsp->vfs_flag & VFS_NOTRUNC)
 497                         val = 1;        /* NOTRUNC is enabled for vp */
 498                 else
 499                         val = (ulong_t)-1;
 500                 break;
 501
 502         case _PC_VDISABLE:
 503                 val = _POSIX_VDISABLE;
 504                 break;
 505
 506         case _PC_CHOWN_RESTRICTED:
 507                 if (rstchown)
 508                         val = rstchown; /* chown restricted enabled */
 509                 else
 510                         val = (ulong_t)-1;
 511                 break;
 512
 513         case _PC_FILESIZEBITS:
 514
 515                 /*
 516                  * If ever we come here it means that underlying file system
 517                  * does not recognise the command and therefore this
 518                  * configurable limit cannot be determined. We return -1
 519                  * and don't change errno.
 520                  */
 521
 522                 val = (ulong_t)-1;    /* large file support */
 523                 break;
 524
 525         case _PC_ACL_ENABLED:
 526                 val = 0;
 527                 break;
 528
 529         case _PC_CASE_BEHAVIOR:
 530                 val = _CASE_SENSITIVE;
 531                 if (vfs_has_feature(vp->v_vfsp, VFSFT_CASEINSENSITIVE) == 1)
 532                         val |= _CASE_INSENSITIVE;
 533                 if (vfs_has_feature(vp->v_vfsp, VFSFT_NOCASESENSITIVE) == 1)
 534                         val &= ~_CASE_SENSITIVE;
 535                 break;
 536
 537         case _PC_SATTR_ENABLED:
 538         case _PC_SATTR_EXISTS:
 539                 val = 0;
 540                 break;
 541
 542         case _PC_ACCESS_FILTERING:
 543                 val = 0;
 544                 break;
 545
 546         default:
 547                 error = EINVAL;
 548                 break;
 549         }
 550
 551         if (error == 0)
 552                 *valp = val;
 553         return (error);
 554 }
 555
 556 /*
 557  * Dispose of a page.
 558  */
 559 /* ARGSUSED */
 560 void
 561 fs_dispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 562     caller_context_t *ct)
 563 {
 564
 565         ASSERT(fl == B_FREE || fl == B_INVAL);
 566
 567         if (fl == B_FREE)
 568                 page_free(pp, dn);
 569         else
 570                 page_destroy(pp, dn);
 571 }
 572
 573 /* ARGSUSED */
 574 void
 575 fs_nodispose(struct vnode *vp, page_t *pp, int fl, int dn, struct cred *cr,
 576     caller_context_t *ct)
 577 {
 578         cmn_err(CE_PANIC, "fs_nodispose invoked");
 579 }
 580
 581 /*
 582  * fabricate acls for file systems that do not support acls.
 583  */
 584 /* ARGSUSED */
 585 int
 586 fs_fab_acl(vnode_t *vp, vsecattr_t *vsecattr, int flag, cred_t *cr,
 587     caller_context_t *ct)
 588 {
 589         aclent_t        *aclentp;
 590         struct vattr    vattr;
 591         int             error;
 592         size_t          aclsize;
 593
 594         vsecattr->vsa_aclcnt    = 0;
 595         vsecattr->vsa_aclentsz  = 0;
 596         vsecattr->vsa_aclentp   = NULL;
 597         vsecattr->vsa_dfaclcnt  = 0;    /* Default ACLs are not fabricated */
 598         vsecattr->vsa_dfaclentp = NULL;
 599
 600         vattr.va_mask = AT_MODE | AT_UID | AT_GID;
 601         if (error = VOP_GETATTR(vp, &vattr, 0, cr, ct))
 602                 return (error);
 603
 604         if (vsecattr->vsa_mask & (VSA_ACLCNT | VSA_ACL)) {
 605                 aclsize = 4 * sizeof (aclent_t);
 606                 vsecattr->vsa_aclcnt    = 4; /* USER, GROUP, OTHER, and CLASS */
 607                 vsecattr->vsa_aclentp = kmem_zalloc(aclsize, KM_SLEEP);
 608                 aclentp = vsecattr->vsa_aclentp;
 609
 610                 aclentp->a_type = USER_OBJ;     /* Owner */
 611                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0700)) >> 6;
 612                 aclentp->a_id = vattr.va_uid;   /* Really undefined */
 613                 aclentp++;
 614
 615                 aclentp->a_type = GROUP_OBJ;    /* Group */
 616                 aclentp->a_perm = ((ushort_t)(vattr.va_mode & 0070)) >> 3;
 617                 aclentp->a_id = vattr.va_gid;   /* Really undefined */
 618                 aclentp++;
 619
 620                 aclentp->a_type = OTHER_OBJ;    /* Other */
 621                 aclentp->a_perm = vattr.va_mode & 0007;
 622                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 623                 aclentp++;
 624
 625                 aclentp->a_type = CLASS_OBJ;    /* Class */
 626                 aclentp->a_perm = (ushort_t)(0007);
 627                 aclentp->a_id = (gid_t)-1;      /* Really undefined */
 628         } else if (vsecattr->vsa_mask & (VSA_ACECNT | VSA_ACE)) {
 629                 VERIFY(0 == acl_trivial_create(vattr.va_mode,
 630                     (vp->v_type == VDIR), (ace_t **)&vsecattr->vsa_aclentp,
 631                     &vsecattr->vsa_aclcnt));
 632                 vsecattr->vsa_aclentsz = vsecattr->vsa_aclcnt * sizeof (ace_t);
 633         }
 634
 635         return (error);
 636 }
 637
 638 /*
 639  * Common code for implementing DOS share reservations
 640  */
 641 /* ARGSUSED4 */
 642 int
 643 fs_shrlock(struct vnode *vp, int cmd, struct shrlock *shr, int flag, cred_t *cr,
 644     caller_context_t *ct)
 645 {
 646         int error;
 647
 648         /*
 649          * Make sure that the file was opened with permissions appropriate
 650          * for the request, and make sure the caller isn't trying to sneak
 651          * in an NBMAND request.
 652          */
 653         if (cmd == F_SHARE) {
 654                 if (((shr->s_access & F_RDACC) && (flag & FREAD) == 0) ||
 655                     ((shr->s_access & F_WRACC) && (flag & FWRITE) == 0))
 656                         return (EBADF);
 657                 if (shr->s_access & (F_RMACC | F_MDACC))
 658                         return (EINVAL);
 659                 if (shr->s_deny & (F_MANDDNY | F_RMDNY))
 660                         return (EINVAL);
 661         }
 662         if (cmd == F_SHARE_NBMAND) {
 663                 /* make sure nbmand is allowed on the file */
 664                 if (!vp->v_vfsp ||
 665                     !(vp->v_vfsp->vfs_flag & VFS_NBMAND)) {
 666                         return (EINVAL);
 667                 }
 668                 if (vp->v_type != VREG) {
 669                         return (EINVAL);
 670                 }
 671         }
 672
 673         nbl_start_crit(vp, RW_WRITER);
 674
 675         switch (cmd) {
 676
 677         case F_SHARE_NBMAND:
 678                 shr->s_deny |= F_MANDDNY;
 679                 /*FALLTHROUGH*/
 680         case F_SHARE:
 681                 error = add_share(vp, shr);
 682                 break;
 683
 684         case F_UNSHARE:
 685                 error = del_share(vp, shr);
 686                 break;
 687
 688         case F_HASREMOTELOCKS:
 689                 /*
 690                  * We are overloading this command to refer to remote
 691                  * shares as well as remote locks, despite its name.
 692                  */
 693                 shr->s_access = shr_has_remote_shares(vp, shr->s_sysid);
 694                 error = 0;
 695                 break;
 696
 697         default:
 698                 error = EINVAL;
 699                 break;
 700         }
 701
 702         nbl_end_crit(vp);
 703         return (error);
 704 }
 705
 706 /*ARGSUSED1*/
 707 int
 708 fs_vnevent_nosupport(vnode_t *vp, vnevent_t e, vnode_t *dvp, char *fnm,
 709     caller_context_t *ct)
 710 {
 711         ASSERT(vp != NULL);
 712         return (ENOTSUP);
 713 }
 714
 715 /*ARGSUSED1*/
 716 int
 717 fs_vnevent_support(vnode_t *vp, vnevent_t e, vnode_t *dvp, char *fnm,
 718     caller_context_t *ct)
 719 {
 720         ASSERT(vp != NULL);
 721         return (0);
 722 }
 723
 724 /*
 725  * return 1 for non-trivial ACL.
 726  *
 727  * NB: It is not necessary for the caller to VOP_RWLOCK since
 728  *      we only issue VOP_GETSECATTR.
 729  *
 730  * Returns 0 == trivial
 731  *         1 == NOT Trivial
 732  *         <0 could not determine.
 733  */
 734 int
 735 fs_acl_nontrivial(vnode_t *vp, cred_t *cr)
 736 {
 737         ulong_t         acl_styles;
 738         ulong_t         acl_flavor;
 739         vsecattr_t      vsecattr;
 740         int             error;
 741         int             isnontrivial;
 742
 743         /* determine the forms of ACLs maintained */
 744         error = VOP_PATHCONF(vp, _PC_ACL_ENABLED, &acl_styles, cr, NULL);
 745
 746         /* clear bits we don't understand and establish default acl_style */
 747         acl_styles &= (_ACL_ACLENT_ENABLED | _ACL_ACE_ENABLED);
 748         if (error || (acl_styles == 0))
 749                 acl_styles = _ACL_ACLENT_ENABLED;
 750
 751         vsecattr.vsa_aclentp = NULL;
 752         vsecattr.vsa_dfaclentp = NULL;
 753         vsecattr.vsa_aclcnt = 0;
 754         vsecattr.vsa_dfaclcnt = 0;
 755
 756         while (acl_styles) {
 757                 /* select one of the styles as current flavor */
 758                 acl_flavor = 0;
 759                 if (acl_styles & _ACL_ACLENT_ENABLED) {
 760                         acl_flavor = _ACL_ACLENT_ENABLED;
 761                         vsecattr.vsa_mask = VSA_ACLCNT | VSA_DFACLCNT;
 762                 } else if (acl_styles & _ACL_ACE_ENABLED) {
 763                         acl_flavor = _ACL_ACE_ENABLED;
 764                         vsecattr.vsa_mask = VSA_ACECNT | VSA_ACE;
 765                 }
 766
 767                 ASSERT(vsecattr.vsa_mask && acl_flavor);
 768                 error = VOP_GETSECATTR(vp, &vsecattr, 0, cr, NULL);
 769                 if (error == 0)
 770                         break;
 771
 772                 /* that flavor failed */
 773                 acl_styles &= ~acl_flavor;
 774         }
 775
 776         /* if all styles fail then assume trivial */
 777         if (acl_styles == 0)
 778                 return (0);
 779
 780         /* process the flavor that worked */
 781         isnontrivial = 0;
 782         if (acl_flavor & _ACL_ACLENT_ENABLED) {
 783                 if (vsecattr.vsa_aclcnt > MIN_ACL_ENTRIES)
 784                         isnontrivial = 1;
 785                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 786                         kmem_free(vsecattr.vsa_aclentp,
 787                             vsecattr.vsa_aclcnt * sizeof (aclent_t));
 788                 if (vsecattr.vsa_dfaclcnt && vsecattr.vsa_dfaclentp != NULL)
 789                         kmem_free(vsecattr.vsa_dfaclentp,
 790                             vsecattr.vsa_dfaclcnt * sizeof (aclent_t));
 791         }
 792         if (acl_flavor & _ACL_ACE_ENABLED) {
 793                 isnontrivial = ace_trivial(vsecattr.vsa_aclentp,
 794                     vsecattr.vsa_aclcnt);
 795
 796                 if (vsecattr.vsa_aclcnt && vsecattr.vsa_aclentp != NULL)
 797                         kmem_free(vsecattr.vsa_aclentp,
 798                             vsecattr.vsa_aclcnt * sizeof (ace_t));
 799                 /* ACE has no vsecattr.vsa_dfaclcnt */
 800         }
 801         return (isnontrivial);
 802 }
 803
 804 /*
 805  * Check whether we need a retry to recover from STALE error.
 806  */
 807 int
 808 fs_need_estale_retry(int retry_count)
 809 {
 810         if (retry_count < fs_estale_retry)
 811                 return (1);
 812         else
 813                 return (0);
 814 }
 815
 816
 817 static int (*fs_av_scan)(vnode_t *, cred_t *, int) = NULL;
 818
 819 /*
 820  * Routine for anti-virus scanner to call to register its scanning routine.
 821  */
 822 void
 823 fs_vscan_register(int (*av_scan)(vnode_t *, cred_t *, int))
 824 {
 825         fs_av_scan = av_scan;
 826 }
 827
 828 /*
 829  * Routine for file systems to call to initiate anti-virus scanning.
 830  * Scanning will only be done on REGular files (currently).
 831  */
 832 int
 833 fs_vscan(vnode_t *vp, cred_t *cr, int async)
 834 {
 835         int ret = 0;
 836
 837         if (fs_av_scan && vp->v_type == VREG)
 838                 ret = (*fs_av_scan)(vp, cr, async);
 839
 840         return (ret);
 841 }
 842
 843 /*
 844  * support functions for reparse point
 845  */
 846 /*
 847  * reparse_vnode_parse
 848  *
 849  * Read the symlink data of a reparse point specified by the vnode
 850  * and return the reparse data as name-value pair in the nvlist.
 851  */
 852 int
 853 reparse_vnode_parse(vnode_t *vp, nvlist_t *nvl)
 854 {
 855         int err;
 856         char *lkdata;
 857         struct uio uio;
 858         struct iovec iov;
 859
 860         if (vp == NULL || nvl == NULL)
 861                 return (EINVAL);
 862
 863         lkdata = kmem_alloc(MAXREPARSELEN, KM_SLEEP);
 864
 865         /*
 866          * Set up io vector to read sym link data
 867          */
 868         iov.iov_base = lkdata;
 869         iov.iov_len = MAXREPARSELEN;
 870         uio.uio_iov = &iov;
 871         uio.uio_iovcnt = 1;
 872         uio.uio_segflg = UIO_SYSSPACE;
 873         uio.uio_extflg = UIO_COPY_CACHED;
 874         uio.uio_loffset = (offset_t)0;
 875         uio.uio_resid = MAXREPARSELEN;
 876
 877         if ((err = VOP_READLINK(vp, &uio, kcred, NULL)) == 0) {
 878                 *(lkdata + MAXREPARSELEN - uio.uio_resid) = '\0';
 879                 err = reparse_parse(lkdata, nvl);
 880         }
 881         kmem_free(lkdata, MAXREPARSELEN);       /* done with lkdata */
 882
 883         return (err);
 884 }
 885
 886 void
 887 reparse_point_init()
 888 {
 889         mutex_init(&reparsed_door_lock, NULL, MUTEX_DEFAULT, NULL);
 890 }
 891
 892 static door_handle_t
 893 reparse_door_get_handle()
 894 {
 895         door_handle_t dh;
 896
 897         mutex_enter(&reparsed_door_lock);
 898         if ((dh = reparsed_door) == NULL) {
 899                 if (door_ki_open(REPARSED_DOOR, &reparsed_door) != 0) {
 900                         reparsed_door = NULL;
 901                         dh = NULL;
 902                 } else
 903                         dh = reparsed_door;
 904         }
 905         mutex_exit(&reparsed_door_lock);
 906         return (dh);
 907 }
 908
 909 static void
 910 reparse_door_reset_handle()
 911 {
 912         mutex_enter(&reparsed_door_lock);
 913         reparsed_door = NULL;
 914         mutex_exit(&reparsed_door_lock);
 915 }
 916
 917 /*
 918  * reparse_kderef
 919  *
 920  * Accepts the service-specific item from the reparse point and returns
 921  * the service-specific data requested.  The caller specifies the size of
 922  * the buffer provided via *bufsz; the routine will fail with EOVERFLOW
 923  * if the results will not fit in the buffer, in which case, *bufsz will
 924  * contain the number of bytes needed to hold the results.
 925  *
 926  * if ok return 0 and update *bufsize with length of actual result
 927  * else return error code.
 928  */
 929 int
 930 reparse_kderef(const char *svc_type, const char *svc_data, char *buf,
 931     size_t *bufsize)
 932 {
 933         int err, retries, need_free, retried_doorhd;
 934         size_t dlen, res_len;
 935         char *darg;
 936         door_arg_t door_args;
 937         reparsed_door_res_t *resp;
 938         door_handle_t rp_door;
 939
 940         if (svc_type == NULL || svc_data == NULL || buf == NULL ||
 941             bufsize == NULL)
 942                 return (EINVAL);
 943
 944         /* get reparsed's door handle */
 945         if ((rp_door = reparse_door_get_handle()) == NULL)
 946                 return (EBADF);
 947
 948         /* setup buffer for door_call args and results */
 949         dlen = strlen(svc_type) + strlen(svc_data) + 2;
 950         if (*bufsize < dlen) {
 951                 darg = kmem_alloc(dlen, KM_SLEEP);
 952                 need_free = 1;
 953         } else {
 954                 darg = buf;     /* use same buffer for door's args & results */
 955                 need_free = 0;
 956         }
 957
 958         /* build argument string of door call */
 959         (void) snprintf(darg, dlen, "%s:%s", svc_type, svc_data);
 960
 961         /* setup args for door call */
 962         door_args.data_ptr = darg;
 963         door_args.data_size = dlen;
 964         door_args.desc_ptr = NULL;
 965         door_args.desc_num = 0;
 966         door_args.rbuf = buf;
 967         door_args.rsize = *bufsize;
 968
 969         /* do the door_call */
 970         retried_doorhd = 0;
 971         retries = 0;
 972         door_ki_hold(rp_door);
 973         while ((err = door_ki_upcall_limited(rp_door, &door_args,
 974             NULL, SIZE_MAX, 0)) != 0) {
 975                 if (err == EAGAIN || err == EINTR) {
 976                         if (++retries < REPARSED_DOORCALL_MAX_RETRY) {
 977                                 delay(SEC_TO_TICK(1));
 978                                 continue;
 979                         }
 980                 } else if (err == EBADF) {
 981                         /* door server goes away... */
 982                         reparse_door_reset_handle();
 983
 984                         if (retried_doorhd == 0) {
 985                                 door_ki_rele(rp_door);
 986                                 retried_doorhd++;
 987                                 rp_door = reparse_door_get_handle();
 988                                 if (rp_door != NULL) {
 989                                         door_ki_hold(rp_door);
 990                                         continue;
 991                                 }
 992                         }
 993                 }
 994                 break;
 995         }
 996
 997         if (rp_door)
 998                 door_ki_rele(rp_door);
 999
1000         if (need_free)
1001                 kmem_free(darg, dlen);          /* done with args buffer */
1002
1003         if (err != 0)
1004                 return (err);
1005
1006         resp = (reparsed_door_res_t *)door_args.rbuf;
1007         if ((err = resp->res_status) == 0) {
1008                 /*
1009                  * have to save the length of the results before the
1010                  * bcopy below since it's can be an overlap copy that
1011                  * overwrites the reparsed_door_res_t structure at
1012                  * the beginning of the buffer.
1013                  */
1014                 res_len = (size_t)resp->res_len;
1015
1016                 /* deref call is ok */
1017                 if (res_len > *bufsize)
1018                         err = EOVERFLOW;
1019                 else
1020                         bcopy(resp->res_data, buf, res_len);
1021                 *bufsize = res_len;
1022         }
1023         if (door_args.rbuf != buf)
1024                 kmem_free(door_args.rbuf, door_args.rsize);
1025
1026         return (err);
1027 }