vfs_syscalls.c revision 1.442.2.4
1 /* $NetBSD: vfs_syscalls.c,v 1.442.2.4 2012/06/02 11:09:35 mrg Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 /* 69 * Virtual File System System Calls 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.442.2.4 2012/06/02 11:09:35 mrg Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/fcntl.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/proc.h> 91 #include <sys/uio.h> 92 #include <sys/kmem.h> 93 #include <sys/dirent.h> 94 #include <sys/sysctl.h> 95 #include <sys/syscallargs.h> 96 #include <sys/vfs_syscalls.h> 97 #include <sys/quota.h> 98 #include <sys/quotactl.h> 99 #include <sys/ktrace.h> 100 #ifdef FILEASSOC 101 #include <sys/fileassoc.h> 102 #endif /* FILEASSOC */ 103 #include <sys/extattr.h> 104 #include <sys/verified_exec.h> 105 #include <sys/kauth.h> 106 #include <sys/atomic.h> 107 #include <sys/module.h> 108 #include <sys/buf.h> 109 110 #include <miscfs/genfs/genfs.h> 111 #include <miscfs/syncfs/syncfs.h> 112 #include <miscfs/specfs/specdev.h> 113 114 #include <nfs/rpcv2.h> 115 #include <nfs/nfsproto.h> 116 #include <nfs/nfs.h> 117 #include <nfs/nfs_var.h> 118 119 static int change_flags(struct vnode *, u_long, struct lwp *); 120 static int change_mode(struct vnode *, int, struct lwp *l); 121 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 122 static int do_open(lwp_t *, struct pathbuf *, int, int, int *); 123 124 /* 125 * This table is used to maintain compatibility with 4.3BSD 126 * and NetBSD 0.9 mount syscalls - and possibly other systems. 127 * Note, the order is important! 128 * 129 * Do not modify this table. It should only contain filesystems 130 * supported by NetBSD 0.9 and 4.3BSD. 131 */ 132 const char * const mountcompatnames[] = { 133 NULL, /* 0 = MOUNT_NONE */ 134 MOUNT_FFS, /* 1 = MOUNT_UFS */ 135 MOUNT_NFS, /* 2 */ 136 MOUNT_MFS, /* 3 */ 137 MOUNT_MSDOS, /* 4 */ 138 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 139 MOUNT_FDESC, /* 6 */ 140 MOUNT_KERNFS, /* 7 */ 141 NULL, /* 8 = MOUNT_DEVFS */ 142 MOUNT_AFS, /* 9 */ 143 }; 144 145 const int nmountcompatnames = __arraycount(mountcompatnames); 146 147 static int 148 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 149 { 150 int error; 151 152 fp->f_flag = flags & FMASK; 153 fp->f_type = DTYPE_VNODE; 154 fp->f_ops = &vnops; 155 fp->f_data = vp; 156 157 if (flags & (O_EXLOCK | O_SHLOCK)) { 158 struct flock lf; 159 int type; 160 161 lf.l_whence = SEEK_SET; 162 lf.l_start = 0; 163 lf.l_len = 0; 164 if (flags & O_EXLOCK) 165 lf.l_type = F_WRLCK; 166 else 167 lf.l_type = F_RDLCK; 168 type = F_FLOCK; 169 if ((flags & FNONBLOCK) == 0) 170 type |= F_WAIT; 171 VOP_UNLOCK(vp); 172 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 173 if (error) { 174 (void) vn_close(vp, fp->f_flag, fp->f_cred); 175 fd_abort(l->l_proc, fp, indx); 176 return error; 177 } 178 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 179 atomic_or_uint(&fp->f_flag, FHASLOCK); 180 } 181 if (flags & O_CLOEXEC) 182 fd_set_exclose(l, indx, true); 183 return 0; 184 } 185 186 static int 187 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 188 void *data, size_t *data_len) 189 { 190 struct mount *mp; 191 int error = 0, saved_flags; 192 193 mp = vp->v_mount; 194 saved_flags = mp->mnt_flag; 195 196 /* We can operate only on VV_ROOT nodes. */ 197 if ((vp->v_vflag & VV_ROOT) == 0) { 198 error = EINVAL; 199 goto out; 200 } 201 202 /* 203 * We only allow the filesystem to be reloaded if it 204 * is currently mounted read-only. Additionally, we 205 * prevent read-write to read-only downgrades. 206 */ 207 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 208 (mp->mnt_flag & MNT_RDONLY) == 0 && 209 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 210 error = EOPNOTSUPP; /* Needs translation */ 211 goto out; 212 } 213 214 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 215 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 216 if (error) 217 goto out; 218 219 if (vfs_busy(mp, NULL)) { 220 error = EPERM; 221 goto out; 222 } 223 224 mutex_enter(&mp->mnt_updating); 225 226 mp->mnt_flag &= ~MNT_OP_FLAGS; 227 mp->mnt_flag |= flags & MNT_OP_FLAGS; 228 229 /* 230 * Set the mount level flags. 231 */ 232 if (flags & MNT_RDONLY) 233 mp->mnt_flag |= MNT_RDONLY; 234 else if (mp->mnt_flag & MNT_RDONLY) 235 mp->mnt_iflag |= IMNT_WANTRDWR; 236 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 237 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 238 error = VFS_MOUNT(mp, path, data, data_len); 239 240 if (error && data != NULL) { 241 int error2; 242 243 /* 244 * Update failed; let's try and see if it was an 245 * export request. For compat with 3.0 and earlier. 246 */ 247 error2 = vfs_hooks_reexport(mp, path, data); 248 249 /* 250 * Only update error code if the export request was 251 * understood but some problem occurred while 252 * processing it. 253 */ 254 if (error2 != EJUSTRETURN) 255 error = error2; 256 } 257 258 if (mp->mnt_iflag & IMNT_WANTRDWR) 259 mp->mnt_flag &= ~MNT_RDONLY; 260 if (error) 261 mp->mnt_flag = saved_flags; 262 mp->mnt_flag &= ~MNT_OP_FLAGS; 263 mp->mnt_iflag &= ~IMNT_WANTRDWR; 264 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 265 if (mp->mnt_syncer == NULL) 266 error = vfs_allocate_syncvnode(mp); 267 } else { 268 if (mp->mnt_syncer != NULL) 269 vfs_deallocate_syncvnode(mp); 270 } 271 mutex_exit(&mp->mnt_updating); 272 vfs_unbusy(mp, false, NULL); 273 274 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 275 (flags & MNT_EXTATTR)) { 276 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 277 NULL, 0, NULL) != 0) { 278 printf("%s: failed to start extattr, error = %d", 279 mp->mnt_stat.f_mntonname, error); 280 mp->mnt_flag &= ~MNT_EXTATTR; 281 } 282 } 283 284 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 285 !(flags & MNT_EXTATTR)) { 286 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 287 NULL, 0, NULL) != 0) { 288 printf("%s: failed to stop extattr, error = %d", 289 mp->mnt_stat.f_mntonname, error); 290 mp->mnt_flag |= MNT_RDONLY; 291 } 292 } 293 out: 294 return (error); 295 } 296 297 static int 298 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 299 { 300 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 301 int error; 302 303 /* Copy file-system type from userspace. */ 304 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 305 if (error) { 306 /* 307 * Historically, filesystem types were identified by numbers. 308 * If we get an integer for the filesystem type instead of a 309 * string, we check to see if it matches one of the historic 310 * filesystem types. 311 */ 312 u_long fsindex = (u_long)fstype; 313 if (fsindex >= nmountcompatnames || 314 mountcompatnames[fsindex] == NULL) 315 return ENODEV; 316 strlcpy(fstypename, mountcompatnames[fsindex], 317 sizeof(fstypename)); 318 } 319 320 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 321 if (strcmp(fstypename, "ufs") == 0) 322 fstypename[0] = 'f'; 323 324 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 325 return 0; 326 327 /* If we can autoload a vfs module, try again */ 328 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 329 330 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 331 return 0; 332 333 return ENODEV; 334 } 335 336 static int 337 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 338 void *data, size_t *data_len) 339 { 340 struct mount *mp; 341 int error; 342 343 /* If MNT_GETARGS is specified, it should be the only flag. */ 344 if (flags & ~MNT_GETARGS) 345 return EINVAL; 346 347 mp = vp->v_mount; 348 349 /* XXX: probably some notion of "can see" here if we want isolation. */ 350 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 351 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 352 if (error) 353 return error; 354 355 if ((vp->v_vflag & VV_ROOT) == 0) 356 return EINVAL; 357 358 if (vfs_busy(mp, NULL)) 359 return EPERM; 360 361 mutex_enter(&mp->mnt_updating); 362 mp->mnt_flag &= ~MNT_OP_FLAGS; 363 mp->mnt_flag |= MNT_GETARGS; 364 error = VFS_MOUNT(mp, path, data, data_len); 365 mp->mnt_flag &= ~MNT_OP_FLAGS; 366 mutex_exit(&mp->mnt_updating); 367 368 vfs_unbusy(mp, false, NULL); 369 return (error); 370 } 371 372 int 373 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 374 { 375 /* { 376 syscallarg(const char *) type; 377 syscallarg(const char *) path; 378 syscallarg(int) flags; 379 syscallarg(void *) data; 380 syscallarg(size_t) data_len; 381 } */ 382 383 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 384 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 385 SCARG(uap, data_len), retval); 386 } 387 388 int 389 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 390 const char *path, int flags, void *data, enum uio_seg data_seg, 391 size_t data_len, register_t *retval) 392 { 393 struct vnode *vp; 394 void *data_buf = data; 395 bool vfsopsrele = false; 396 int error; 397 398 /* XXX: The calling convention of this routine is totally bizarre */ 399 if (vfsops) 400 vfsopsrele = true; 401 402 /* 403 * Get vnode to be covered 404 */ 405 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 406 if (error != 0) { 407 vp = NULL; 408 goto done; 409 } 410 411 if (vfsops == NULL) { 412 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 413 vfsops = vp->v_mount->mnt_op; 414 } else { 415 /* 'type' is userspace */ 416 error = mount_get_vfsops(type, &vfsops); 417 if (error != 0) 418 goto done; 419 vfsopsrele = true; 420 } 421 } 422 423 if (data != NULL && data_seg == UIO_USERSPACE) { 424 if (data_len == 0) { 425 /* No length supplied, use default for filesystem */ 426 data_len = vfsops->vfs_min_mount_data; 427 if (data_len > VFS_MAX_MOUNT_DATA) { 428 error = EINVAL; 429 goto done; 430 } 431 /* 432 * Hopefully a longer buffer won't make copyin() fail. 433 * For compatibility with 3.0 and earlier. 434 */ 435 if (flags & MNT_UPDATE 436 && data_len < sizeof (struct mnt_export_args30)) 437 data_len = sizeof (struct mnt_export_args30); 438 } 439 data_buf = kmem_alloc(data_len, KM_SLEEP); 440 441 /* NFS needs the buffer even for mnt_getargs .... */ 442 error = copyin(data, data_buf, data_len); 443 if (error != 0) 444 goto done; 445 } 446 447 if (flags & MNT_GETARGS) { 448 if (data_len == 0) { 449 error = EINVAL; 450 goto done; 451 } 452 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 453 if (error != 0) 454 goto done; 455 if (data_seg == UIO_USERSPACE) 456 error = copyout(data_buf, data, data_len); 457 *retval = data_len; 458 } else if (flags & MNT_UPDATE) { 459 error = mount_update(l, vp, path, flags, data_buf, &data_len); 460 } else { 461 /* Locking is handled internally in mount_domount(). */ 462 KASSERT(vfsopsrele == true); 463 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 464 &data_len); 465 vfsopsrele = false; 466 } 467 468 done: 469 if (vfsopsrele) 470 vfs_delref(vfsops); 471 if (vp != NULL) { 472 vrele(vp); 473 } 474 if (data_buf != data) 475 kmem_free(data_buf, data_len); 476 return (error); 477 } 478 479 /* 480 * Unmount a file system. 481 * 482 * Note: unmount takes a path to the vnode mounted on as argument, 483 * not special file (as before). 484 */ 485 /* ARGSUSED */ 486 int 487 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 488 { 489 /* { 490 syscallarg(const char *) path; 491 syscallarg(int) flags; 492 } */ 493 struct vnode *vp; 494 struct mount *mp; 495 int error; 496 struct pathbuf *pb; 497 struct nameidata nd; 498 499 error = pathbuf_copyin(SCARG(uap, path), &pb); 500 if (error) { 501 return error; 502 } 503 504 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 505 if ((error = namei(&nd)) != 0) { 506 pathbuf_destroy(pb); 507 return error; 508 } 509 vp = nd.ni_vp; 510 pathbuf_destroy(pb); 511 512 mp = vp->v_mount; 513 atomic_inc_uint(&mp->mnt_refcnt); 514 VOP_UNLOCK(vp); 515 516 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 517 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 518 if (error) { 519 vrele(vp); 520 vfs_destroy(mp); 521 return (error); 522 } 523 524 /* 525 * Don't allow unmounting the root file system. 526 */ 527 if (mp->mnt_flag & MNT_ROOTFS) { 528 vrele(vp); 529 vfs_destroy(mp); 530 return (EINVAL); 531 } 532 533 /* 534 * Must be the root of the filesystem 535 */ 536 if ((vp->v_vflag & VV_ROOT) == 0) { 537 vrele(vp); 538 vfs_destroy(mp); 539 return (EINVAL); 540 } 541 542 vrele(vp); 543 error = dounmount(mp, SCARG(uap, flags), l); 544 vfs_destroy(mp); 545 return error; 546 } 547 548 /* 549 * Sync each mounted filesystem. 550 */ 551 #ifdef DEBUG 552 int syncprt = 0; 553 struct ctldebug debug0 = { "syncprt", &syncprt }; 554 #endif 555 556 void 557 do_sys_sync(struct lwp *l) 558 { 559 struct mount *mp, *nmp; 560 int asyncflag; 561 562 mutex_enter(&mountlist_lock); 563 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 564 mp = nmp) { 565 if (vfs_busy(mp, &nmp)) { 566 continue; 567 } 568 mutex_enter(&mp->mnt_updating); 569 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 570 asyncflag = mp->mnt_flag & MNT_ASYNC; 571 mp->mnt_flag &= ~MNT_ASYNC; 572 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 573 if (asyncflag) 574 mp->mnt_flag |= MNT_ASYNC; 575 } 576 mutex_exit(&mp->mnt_updating); 577 vfs_unbusy(mp, false, &nmp); 578 } 579 mutex_exit(&mountlist_lock); 580 #ifdef DEBUG 581 if (syncprt) 582 vfs_bufstats(); 583 #endif /* DEBUG */ 584 } 585 586 /* ARGSUSED */ 587 int 588 sys_sync(struct lwp *l, const void *v, register_t *retval) 589 { 590 do_sys_sync(l); 591 return (0); 592 } 593 594 595 /* 596 * Access or change filesystem quotas. 597 * 598 * (this is really 14 different calls bundled into one) 599 */ 600 601 static int 602 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 603 { 604 struct quotastat info_k; 605 int error; 606 607 /* ensure any padding bytes are cleared */ 608 memset(&info_k, 0, sizeof(info_k)); 609 610 error = vfs_quotactl_stat(mp, &info_k); 611 if (error) { 612 return error; 613 } 614 615 return copyout(&info_k, info_u, sizeof(info_k)); 616 } 617 618 static int 619 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 620 struct quotaidtypestat *info_u) 621 { 622 struct quotaidtypestat info_k; 623 int error; 624 625 /* ensure any padding bytes are cleared */ 626 memset(&info_k, 0, sizeof(info_k)); 627 628 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 629 if (error) { 630 return error; 631 } 632 633 return copyout(&info_k, info_u, sizeof(info_k)); 634 } 635 636 static int 637 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 638 struct quotaobjtypestat *info_u) 639 { 640 struct quotaobjtypestat info_k; 641 int error; 642 643 /* ensure any padding bytes are cleared */ 644 memset(&info_k, 0, sizeof(info_k)); 645 646 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 647 if (error) { 648 return error; 649 } 650 651 return copyout(&info_k, info_u, sizeof(info_k)); 652 } 653 654 static int 655 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 656 struct quotaval *val_u) 657 { 658 struct quotakey key_k; 659 struct quotaval val_k; 660 int error; 661 662 /* ensure any padding bytes are cleared */ 663 memset(&val_k, 0, sizeof(val_k)); 664 665 error = copyin(key_u, &key_k, sizeof(key_k)); 666 if (error) { 667 return error; 668 } 669 670 error = vfs_quotactl_get(mp, &key_k, &val_k); 671 if (error) { 672 return error; 673 } 674 675 return copyout(&val_k, val_u, sizeof(val_k)); 676 } 677 678 static int 679 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 680 const struct quotaval *val_u) 681 { 682 struct quotakey key_k; 683 struct quotaval val_k; 684 int error; 685 686 error = copyin(key_u, &key_k, sizeof(key_k)); 687 if (error) { 688 return error; 689 } 690 691 error = copyin(val_u, &val_k, sizeof(val_k)); 692 if (error) { 693 return error; 694 } 695 696 return vfs_quotactl_put(mp, &key_k, &val_k); 697 } 698 699 static int 700 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u) 701 { 702 struct quotakey key_k; 703 int error; 704 705 error = copyin(key_u, &key_k, sizeof(key_k)); 706 if (error) { 707 return error; 708 } 709 710 return vfs_quotactl_delete(mp, &key_k); 711 } 712 713 static int 714 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 715 { 716 struct quotakcursor cursor_k; 717 int error; 718 719 /* ensure any padding bytes are cleared */ 720 memset(&cursor_k, 0, sizeof(cursor_k)); 721 722 error = vfs_quotactl_cursoropen(mp, &cursor_k); 723 if (error) { 724 return error; 725 } 726 727 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 728 } 729 730 static int 731 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 732 { 733 struct quotakcursor cursor_k; 734 int error; 735 736 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 737 if (error) { 738 return error; 739 } 740 741 return vfs_quotactl_cursorclose(mp, &cursor_k); 742 } 743 744 static int 745 do_sys_quotactl_cursorskipidtype(struct mount *mp, 746 struct quotakcursor *cursor_u, int idtype) 747 { 748 struct quotakcursor cursor_k; 749 int error; 750 751 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 752 if (error) { 753 return error; 754 } 755 756 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 757 if (error) { 758 return error; 759 } 760 761 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 762 } 763 764 static int 765 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 766 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 767 unsigned *ret_u) 768 { 769 #define CGET_STACK_MAX 8 770 struct quotakcursor cursor_k; 771 struct quotakey stackkeys[CGET_STACK_MAX]; 772 struct quotaval stackvals[CGET_STACK_MAX]; 773 struct quotakey *keys_k; 774 struct quotaval *vals_k; 775 unsigned ret_k; 776 int error; 777 778 if (maxnum > 128) { 779 maxnum = 128; 780 } 781 782 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 783 if (error) { 784 return error; 785 } 786 787 if (maxnum <= CGET_STACK_MAX) { 788 keys_k = stackkeys; 789 vals_k = stackvals; 790 /* ensure any padding bytes are cleared */ 791 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 792 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 793 } else { 794 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 795 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 796 } 797 798 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 799 &ret_k); 800 if (error) { 801 goto fail; 802 } 803 804 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 805 if (error) { 806 goto fail; 807 } 808 809 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 810 if (error) { 811 goto fail; 812 } 813 814 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 815 if (error) { 816 goto fail; 817 } 818 819 /* do last to maximize the chance of being able to recover a failure */ 820 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 821 822 fail: 823 if (keys_k != stackkeys) { 824 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 825 } 826 if (vals_k != stackvals) { 827 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 828 } 829 return error; 830 } 831 832 static int 833 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 834 int *ret_u) 835 { 836 struct quotakcursor cursor_k; 837 int ret_k; 838 int error; 839 840 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 841 if (error) { 842 return error; 843 } 844 845 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 846 if (error) { 847 return error; 848 } 849 850 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 851 if (error) { 852 return error; 853 } 854 855 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 856 } 857 858 static int 859 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 860 { 861 struct quotakcursor cursor_k; 862 int error; 863 864 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 865 if (error) { 866 return error; 867 } 868 869 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 870 if (error) { 871 return error; 872 } 873 874 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 875 } 876 877 static int 878 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 879 { 880 char *path_k; 881 int error; 882 883 /* XXX this should probably be a struct pathbuf */ 884 path_k = PNBUF_GET(); 885 error = copyin(path_u, path_k, PATH_MAX); 886 if (error) { 887 PNBUF_PUT(path_k); 888 return error; 889 } 890 891 error = vfs_quotactl_quotaon(mp, idtype, path_k); 892 893 PNBUF_PUT(path_k); 894 return error; 895 } 896 897 static int 898 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 899 { 900 return vfs_quotactl_quotaoff(mp, idtype); 901 } 902 903 int 904 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 905 { 906 struct mount *mp; 907 struct vnode *vp; 908 int error; 909 910 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 911 if (error != 0) 912 return (error); 913 mp = vp->v_mount; 914 915 switch (args->qc_op) { 916 case QUOTACTL_STAT: 917 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 918 break; 919 case QUOTACTL_IDTYPESTAT: 920 error = do_sys_quotactl_idtypestat(mp, 921 args->u.idtypestat.qc_idtype, 922 args->u.idtypestat.qc_info); 923 break; 924 case QUOTACTL_OBJTYPESTAT: 925 error = do_sys_quotactl_objtypestat(mp, 926 args->u.objtypestat.qc_objtype, 927 args->u.objtypestat.qc_info); 928 break; 929 case QUOTACTL_GET: 930 error = do_sys_quotactl_get(mp, 931 args->u.get.qc_key, 932 args->u.get.qc_val); 933 break; 934 case QUOTACTL_PUT: 935 error = do_sys_quotactl_put(mp, 936 args->u.put.qc_key, 937 args->u.put.qc_val); 938 break; 939 case QUOTACTL_DELETE: 940 error = do_sys_quotactl_delete(mp, args->u.delete.qc_key); 941 break; 942 case QUOTACTL_CURSOROPEN: 943 error = do_sys_quotactl_cursoropen(mp, 944 args->u.cursoropen.qc_cursor); 945 break; 946 case QUOTACTL_CURSORCLOSE: 947 error = do_sys_quotactl_cursorclose(mp, 948 args->u.cursorclose.qc_cursor); 949 break; 950 case QUOTACTL_CURSORSKIPIDTYPE: 951 error = do_sys_quotactl_cursorskipidtype(mp, 952 args->u.cursorskipidtype.qc_cursor, 953 args->u.cursorskipidtype.qc_idtype); 954 break; 955 case QUOTACTL_CURSORGET: 956 error = do_sys_quotactl_cursorget(mp, 957 args->u.cursorget.qc_cursor, 958 args->u.cursorget.qc_keys, 959 args->u.cursorget.qc_vals, 960 args->u.cursorget.qc_maxnum, 961 args->u.cursorget.qc_ret); 962 break; 963 case QUOTACTL_CURSORATEND: 964 error = do_sys_quotactl_cursoratend(mp, 965 args->u.cursoratend.qc_cursor, 966 args->u.cursoratend.qc_ret); 967 break; 968 case QUOTACTL_CURSORREWIND: 969 error = do_sys_quotactl_cursorrewind(mp, 970 args->u.cursorrewind.qc_cursor); 971 break; 972 case QUOTACTL_QUOTAON: 973 error = do_sys_quotactl_quotaon(mp, 974 args->u.quotaon.qc_idtype, 975 args->u.quotaon.qc_quotafile); 976 break; 977 case QUOTACTL_QUOTAOFF: 978 error = do_sys_quotactl_quotaoff(mp, 979 args->u.quotaoff.qc_idtype); 980 break; 981 default: 982 error = EINVAL; 983 break; 984 } 985 986 vrele(vp); 987 return error; 988 } 989 990 /* ARGSUSED */ 991 int 992 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 993 register_t *retval) 994 { 995 /* { 996 syscallarg(const char *) path; 997 syscallarg(struct quotactl_args *) args; 998 } */ 999 struct quotactl_args args; 1000 int error; 1001 1002 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1003 if (error) { 1004 return error; 1005 } 1006 1007 return do_sys_quotactl(SCARG(uap, path), &args); 1008 } 1009 1010 int 1011 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1012 int root) 1013 { 1014 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1015 int error = 0; 1016 1017 /* 1018 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1019 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1020 * overrides MNT_NOWAIT. 1021 */ 1022 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1023 (flags != MNT_WAIT && flags != 0)) { 1024 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1025 goto done; 1026 } 1027 1028 /* Get the filesystem stats now */ 1029 memset(sp, 0, sizeof(*sp)); 1030 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1031 return error; 1032 } 1033 1034 if (cwdi->cwdi_rdir == NULL) 1035 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1036 done: 1037 if (cwdi->cwdi_rdir != NULL) { 1038 size_t len; 1039 char *bp; 1040 char c; 1041 char *path = PNBUF_GET(); 1042 1043 bp = path + MAXPATHLEN; 1044 *--bp = '\0'; 1045 rw_enter(&cwdi->cwdi_lock, RW_READER); 1046 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1047 MAXPATHLEN / 2, 0, l); 1048 rw_exit(&cwdi->cwdi_lock); 1049 if (error) { 1050 PNBUF_PUT(path); 1051 return error; 1052 } 1053 len = strlen(bp); 1054 if (len != 1) { 1055 /* 1056 * for mount points that are below our root, we can see 1057 * them, so we fix up the pathname and return them. The 1058 * rest we cannot see, so we don't allow viewing the 1059 * data. 1060 */ 1061 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1062 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1063 (void)strlcpy(sp->f_mntonname, 1064 c == '\0' ? "/" : &sp->f_mntonname[len], 1065 sizeof(sp->f_mntonname)); 1066 } else { 1067 if (root) 1068 (void)strlcpy(sp->f_mntonname, "/", 1069 sizeof(sp->f_mntonname)); 1070 else 1071 error = EPERM; 1072 } 1073 } 1074 PNBUF_PUT(path); 1075 } 1076 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1077 return error; 1078 } 1079 1080 /* 1081 * Get filesystem statistics by path. 1082 */ 1083 int 1084 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1085 { 1086 struct mount *mp; 1087 int error; 1088 struct vnode *vp; 1089 1090 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1091 if (error != 0) 1092 return error; 1093 mp = vp->v_mount; 1094 error = dostatvfs(mp, sb, l, flags, 1); 1095 vrele(vp); 1096 return error; 1097 } 1098 1099 /* ARGSUSED */ 1100 int 1101 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1102 { 1103 /* { 1104 syscallarg(const char *) path; 1105 syscallarg(struct statvfs *) buf; 1106 syscallarg(int) flags; 1107 } */ 1108 struct statvfs *sb; 1109 int error; 1110 1111 sb = STATVFSBUF_GET(); 1112 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1113 if (error == 0) 1114 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1115 STATVFSBUF_PUT(sb); 1116 return error; 1117 } 1118 1119 /* 1120 * Get filesystem statistics by fd. 1121 */ 1122 int 1123 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1124 { 1125 file_t *fp; 1126 struct mount *mp; 1127 int error; 1128 1129 /* fd_getvnode() will use the descriptor for us */ 1130 if ((error = fd_getvnode(fd, &fp)) != 0) 1131 return (error); 1132 mp = ((struct vnode *)fp->f_data)->v_mount; 1133 error = dostatvfs(mp, sb, curlwp, flags, 1); 1134 fd_putfile(fd); 1135 return error; 1136 } 1137 1138 /* ARGSUSED */ 1139 int 1140 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1141 { 1142 /* { 1143 syscallarg(int) fd; 1144 syscallarg(struct statvfs *) buf; 1145 syscallarg(int) flags; 1146 } */ 1147 struct statvfs *sb; 1148 int error; 1149 1150 sb = STATVFSBUF_GET(); 1151 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1152 if (error == 0) 1153 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1154 STATVFSBUF_PUT(sb); 1155 return error; 1156 } 1157 1158 1159 /* 1160 * Get statistics on all filesystems. 1161 */ 1162 int 1163 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1164 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1165 register_t *retval) 1166 { 1167 int root = 0; 1168 struct proc *p = l->l_proc; 1169 struct mount *mp, *nmp; 1170 struct statvfs *sb; 1171 size_t count, maxcount; 1172 int error = 0; 1173 1174 sb = STATVFSBUF_GET(); 1175 maxcount = bufsize / entry_sz; 1176 mutex_enter(&mountlist_lock); 1177 count = 0; 1178 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1179 mp = nmp) { 1180 if (vfs_busy(mp, &nmp)) { 1181 continue; 1182 } 1183 if (sfsp && count < maxcount) { 1184 error = dostatvfs(mp, sb, l, flags, 0); 1185 if (error) { 1186 vfs_unbusy(mp, false, &nmp); 1187 error = 0; 1188 continue; 1189 } 1190 error = copyfn(sb, sfsp, entry_sz); 1191 if (error) { 1192 vfs_unbusy(mp, false, NULL); 1193 goto out; 1194 } 1195 sfsp = (char *)sfsp + entry_sz; 1196 root |= strcmp(sb->f_mntonname, "/") == 0; 1197 } 1198 count++; 1199 vfs_unbusy(mp, false, &nmp); 1200 } 1201 mutex_exit(&mountlist_lock); 1202 1203 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1204 /* 1205 * fake a root entry 1206 */ 1207 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1208 sb, l, flags, 1); 1209 if (error != 0) 1210 goto out; 1211 if (sfsp) { 1212 error = copyfn(sb, sfsp, entry_sz); 1213 if (error != 0) 1214 goto out; 1215 } 1216 count++; 1217 } 1218 if (sfsp && count > maxcount) 1219 *retval = maxcount; 1220 else 1221 *retval = count; 1222 out: 1223 STATVFSBUF_PUT(sb); 1224 return error; 1225 } 1226 1227 int 1228 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1229 { 1230 /* { 1231 syscallarg(struct statvfs *) buf; 1232 syscallarg(size_t) bufsize; 1233 syscallarg(int) flags; 1234 } */ 1235 1236 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1237 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1238 } 1239 1240 /* 1241 * Change current working directory to a given file descriptor. 1242 */ 1243 /* ARGSUSED */ 1244 int 1245 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1246 { 1247 /* { 1248 syscallarg(int) fd; 1249 } */ 1250 struct proc *p = l->l_proc; 1251 struct cwdinfo *cwdi; 1252 struct vnode *vp, *tdp; 1253 struct mount *mp; 1254 file_t *fp; 1255 int error, fd; 1256 1257 /* fd_getvnode() will use the descriptor for us */ 1258 fd = SCARG(uap, fd); 1259 if ((error = fd_getvnode(fd, &fp)) != 0) 1260 return (error); 1261 vp = fp->f_data; 1262 1263 vref(vp); 1264 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1265 if (vp->v_type != VDIR) 1266 error = ENOTDIR; 1267 else 1268 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1269 if (error) { 1270 vput(vp); 1271 goto out; 1272 } 1273 while ((mp = vp->v_mountedhere) != NULL) { 1274 error = vfs_busy(mp, NULL); 1275 vput(vp); 1276 if (error != 0) 1277 goto out; 1278 error = VFS_ROOT(mp, &tdp); 1279 vfs_unbusy(mp, false, NULL); 1280 if (error) 1281 goto out; 1282 vp = tdp; 1283 } 1284 VOP_UNLOCK(vp); 1285 1286 /* 1287 * Disallow changing to a directory not under the process's 1288 * current root directory (if there is one). 1289 */ 1290 cwdi = p->p_cwdi; 1291 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1292 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1293 vrele(vp); 1294 error = EPERM; /* operation not permitted */ 1295 } else { 1296 vrele(cwdi->cwdi_cdir); 1297 cwdi->cwdi_cdir = vp; 1298 } 1299 rw_exit(&cwdi->cwdi_lock); 1300 1301 out: 1302 fd_putfile(fd); 1303 return (error); 1304 } 1305 1306 /* 1307 * Change this process's notion of the root directory to a given file 1308 * descriptor. 1309 */ 1310 int 1311 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1312 { 1313 struct proc *p = l->l_proc; 1314 struct vnode *vp; 1315 file_t *fp; 1316 int error, fd = SCARG(uap, fd); 1317 1318 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1319 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1320 return error; 1321 /* fd_getvnode() will use the descriptor for us */ 1322 if ((error = fd_getvnode(fd, &fp)) != 0) 1323 return error; 1324 vp = fp->f_data; 1325 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1326 if (vp->v_type != VDIR) 1327 error = ENOTDIR; 1328 else 1329 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1330 VOP_UNLOCK(vp); 1331 if (error) 1332 goto out; 1333 vref(vp); 1334 1335 change_root(p->p_cwdi, vp, l); 1336 1337 out: 1338 fd_putfile(fd); 1339 return (error); 1340 } 1341 1342 /* 1343 * Change current working directory (``.''). 1344 */ 1345 /* ARGSUSED */ 1346 int 1347 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1348 { 1349 /* { 1350 syscallarg(const char *) path; 1351 } */ 1352 struct proc *p = l->l_proc; 1353 struct cwdinfo *cwdi; 1354 int error; 1355 struct vnode *vp; 1356 1357 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1358 &vp, l)) != 0) 1359 return (error); 1360 cwdi = p->p_cwdi; 1361 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1362 vrele(cwdi->cwdi_cdir); 1363 cwdi->cwdi_cdir = vp; 1364 rw_exit(&cwdi->cwdi_lock); 1365 return (0); 1366 } 1367 1368 /* 1369 * Change notion of root (``/'') directory. 1370 */ 1371 /* ARGSUSED */ 1372 int 1373 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1374 { 1375 /* { 1376 syscallarg(const char *) path; 1377 } */ 1378 struct proc *p = l->l_proc; 1379 int error; 1380 struct vnode *vp; 1381 1382 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1383 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1384 return (error); 1385 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1386 &vp, l)) != 0) 1387 return (error); 1388 1389 change_root(p->p_cwdi, vp, l); 1390 1391 return (0); 1392 } 1393 1394 /* 1395 * Common routine for chroot and fchroot. 1396 * NB: callers need to properly authorize the change root operation. 1397 */ 1398 void 1399 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1400 { 1401 1402 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1403 if (cwdi->cwdi_rdir != NULL) 1404 vrele(cwdi->cwdi_rdir); 1405 cwdi->cwdi_rdir = vp; 1406 1407 /* 1408 * Prevent escaping from chroot by putting the root under 1409 * the working directory. Silently chdir to / if we aren't 1410 * already there. 1411 */ 1412 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1413 /* 1414 * XXX would be more failsafe to change directory to a 1415 * deadfs node here instead 1416 */ 1417 vrele(cwdi->cwdi_cdir); 1418 vref(vp); 1419 cwdi->cwdi_cdir = vp; 1420 } 1421 rw_exit(&cwdi->cwdi_lock); 1422 } 1423 1424 /* 1425 * Common routine for chroot and chdir. 1426 * XXX "where" should be enum uio_seg 1427 */ 1428 int 1429 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1430 { 1431 struct pathbuf *pb; 1432 struct nameidata nd; 1433 int error; 1434 1435 error = pathbuf_maybe_copyin(path, where, &pb); 1436 if (error) { 1437 return error; 1438 } 1439 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1440 if ((error = namei(&nd)) != 0) { 1441 pathbuf_destroy(pb); 1442 return error; 1443 } 1444 *vpp = nd.ni_vp; 1445 pathbuf_destroy(pb); 1446 1447 if ((*vpp)->v_type != VDIR) 1448 error = ENOTDIR; 1449 else 1450 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1451 1452 if (error) 1453 vput(*vpp); 1454 else 1455 VOP_UNLOCK(*vpp); 1456 return (error); 1457 } 1458 1459 /* 1460 * Internals of sys_open - path has already been converted into a pathbuf 1461 * (so we can easily reuse this function from other parts of the kernel, 1462 * like posix_spawn post-processing). 1463 */ 1464 static int 1465 do_open(lwp_t *l, struct pathbuf *pb, int open_flags, int open_mode, int *fd) 1466 { 1467 struct proc *p = l->l_proc; 1468 struct cwdinfo *cwdi = p->p_cwdi; 1469 file_t *fp; 1470 struct vnode *vp; 1471 int flags, cmode; 1472 int indx, error; 1473 struct nameidata nd; 1474 1475 flags = FFLAGS(open_flags); 1476 if ((flags & (FREAD | FWRITE)) == 0) 1477 return EINVAL; 1478 1479 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1480 return error; 1481 } 1482 1483 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1484 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1485 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1486 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1487 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1488 fd_abort(p, fp, indx); 1489 if ((error == EDUPFD || error == EMOVEFD) && 1490 l->l_dupfd >= 0 && /* XXX from fdopen */ 1491 (error = 1492 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1493 *fd = indx; 1494 return 0; 1495 } 1496 if (error == ERESTART) 1497 error = EINTR; 1498 return error; 1499 } 1500 1501 l->l_dupfd = 0; 1502 vp = nd.ni_vp; 1503 1504 if ((error = open_setfp(l, fp, vp, indx, flags))) 1505 return error; 1506 1507 VOP_UNLOCK(vp); 1508 *fd = indx; 1509 fd_affix(p, fp, indx); 1510 return 0; 1511 } 1512 1513 int 1514 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1515 { 1516 struct pathbuf *pb; 1517 int error, oflags; 1518 1519 oflags = FFLAGS(open_flags); 1520 if ((oflags & (FREAD | FWRITE)) == 0) 1521 return EINVAL; 1522 1523 pb = pathbuf_create(path); 1524 if (pb == NULL) 1525 return ENOMEM; 1526 1527 error = do_open(curlwp, pb, open_flags, open_mode, fd); 1528 pathbuf_destroy(pb); 1529 1530 return error; 1531 } 1532 1533 /* 1534 * Check permissions, allocate an open file structure, 1535 * and call the device open routine if any. 1536 */ 1537 int 1538 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1539 { 1540 /* { 1541 syscallarg(const char *) path; 1542 syscallarg(int) flags; 1543 syscallarg(int) mode; 1544 } */ 1545 struct pathbuf *pb; 1546 int result, flags, error; 1547 1548 flags = FFLAGS(SCARG(uap, flags)); 1549 if ((flags & (FREAD | FWRITE)) == 0) 1550 return EINVAL; 1551 1552 error = pathbuf_copyin(SCARG(uap, path), &pb); 1553 if (error) 1554 return error; 1555 1556 error = do_open(l, pb, SCARG(uap, flags), SCARG(uap, mode), &result); 1557 pathbuf_destroy(pb); 1558 1559 *retval = result; 1560 return error; 1561 } 1562 1563 int 1564 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1565 { 1566 /* { 1567 syscallarg(int) fd; 1568 syscallarg(const char *) path; 1569 syscallarg(int) flags; 1570 syscallarg(int) mode; 1571 } */ 1572 1573 return ENOSYS; 1574 } 1575 1576 static void 1577 vfs__fhfree(fhandle_t *fhp) 1578 { 1579 size_t fhsize; 1580 1581 if (fhp == NULL) { 1582 return; 1583 } 1584 fhsize = FHANDLE_SIZE(fhp); 1585 kmem_free(fhp, fhsize); 1586 } 1587 1588 /* 1589 * vfs_composefh: compose a filehandle. 1590 */ 1591 1592 int 1593 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1594 { 1595 struct mount *mp; 1596 struct fid *fidp; 1597 int error; 1598 size_t needfhsize; 1599 size_t fidsize; 1600 1601 mp = vp->v_mount; 1602 fidp = NULL; 1603 if (*fh_size < FHANDLE_SIZE_MIN) { 1604 fidsize = 0; 1605 } else { 1606 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1607 if (fhp != NULL) { 1608 memset(fhp, 0, *fh_size); 1609 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1610 fidp = &fhp->fh_fid; 1611 } 1612 } 1613 error = VFS_VPTOFH(vp, fidp, &fidsize); 1614 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1615 if (error == 0 && *fh_size < needfhsize) { 1616 error = E2BIG; 1617 } 1618 *fh_size = needfhsize; 1619 return error; 1620 } 1621 1622 int 1623 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1624 { 1625 struct mount *mp; 1626 fhandle_t *fhp; 1627 size_t fhsize; 1628 size_t fidsize; 1629 int error; 1630 1631 *fhpp = NULL; 1632 mp = vp->v_mount; 1633 fidsize = 0; 1634 error = VFS_VPTOFH(vp, NULL, &fidsize); 1635 KASSERT(error != 0); 1636 if (error != E2BIG) { 1637 goto out; 1638 } 1639 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1640 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1641 if (fhp == NULL) { 1642 error = ENOMEM; 1643 goto out; 1644 } 1645 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1646 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1647 if (error == 0) { 1648 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1649 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1650 *fhpp = fhp; 1651 } else { 1652 kmem_free(fhp, fhsize); 1653 } 1654 out: 1655 return error; 1656 } 1657 1658 void 1659 vfs_composefh_free(fhandle_t *fhp) 1660 { 1661 1662 vfs__fhfree(fhp); 1663 } 1664 1665 /* 1666 * vfs_fhtovp: lookup a vnode by a filehandle. 1667 */ 1668 1669 int 1670 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1671 { 1672 struct mount *mp; 1673 int error; 1674 1675 *vpp = NULL; 1676 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1677 if (mp == NULL) { 1678 error = ESTALE; 1679 goto out; 1680 } 1681 if (mp->mnt_op->vfs_fhtovp == NULL) { 1682 error = EOPNOTSUPP; 1683 goto out; 1684 } 1685 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1686 out: 1687 return error; 1688 } 1689 1690 /* 1691 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1692 * the needed size. 1693 */ 1694 1695 int 1696 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1697 { 1698 fhandle_t *fhp; 1699 int error; 1700 1701 *fhpp = NULL; 1702 if (fhsize > FHANDLE_SIZE_MAX) { 1703 return EINVAL; 1704 } 1705 if (fhsize < FHANDLE_SIZE_MIN) { 1706 return EINVAL; 1707 } 1708 again: 1709 fhp = kmem_alloc(fhsize, KM_SLEEP); 1710 if (fhp == NULL) { 1711 return ENOMEM; 1712 } 1713 error = copyin(ufhp, fhp, fhsize); 1714 if (error == 0) { 1715 /* XXX this check shouldn't be here */ 1716 if (FHANDLE_SIZE(fhp) == fhsize) { 1717 *fhpp = fhp; 1718 return 0; 1719 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1720 /* 1721 * a kludge for nfsv2 padded handles. 1722 */ 1723 size_t sz; 1724 1725 sz = FHANDLE_SIZE(fhp); 1726 kmem_free(fhp, fhsize); 1727 fhsize = sz; 1728 goto again; 1729 } else { 1730 /* 1731 * userland told us wrong size. 1732 */ 1733 error = EINVAL; 1734 } 1735 } 1736 kmem_free(fhp, fhsize); 1737 return error; 1738 } 1739 1740 void 1741 vfs_copyinfh_free(fhandle_t *fhp) 1742 { 1743 1744 vfs__fhfree(fhp); 1745 } 1746 1747 /* 1748 * Get file handle system call 1749 */ 1750 int 1751 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1752 { 1753 /* { 1754 syscallarg(char *) fname; 1755 syscallarg(fhandle_t *) fhp; 1756 syscallarg(size_t *) fh_size; 1757 } */ 1758 struct vnode *vp; 1759 fhandle_t *fh; 1760 int error; 1761 struct pathbuf *pb; 1762 struct nameidata nd; 1763 size_t sz; 1764 size_t usz; 1765 1766 /* 1767 * Must be super user 1768 */ 1769 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1770 0, NULL, NULL, NULL); 1771 if (error) 1772 return (error); 1773 1774 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1775 if (error) { 1776 return error; 1777 } 1778 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1779 error = namei(&nd); 1780 if (error) { 1781 pathbuf_destroy(pb); 1782 return error; 1783 } 1784 vp = nd.ni_vp; 1785 pathbuf_destroy(pb); 1786 1787 error = vfs_composefh_alloc(vp, &fh); 1788 vput(vp); 1789 if (error != 0) { 1790 goto out; 1791 } 1792 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1793 if (error != 0) { 1794 goto out; 1795 } 1796 sz = FHANDLE_SIZE(fh); 1797 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1798 if (error != 0) { 1799 goto out; 1800 } 1801 if (usz >= sz) { 1802 error = copyout(fh, SCARG(uap, fhp), sz); 1803 } else { 1804 error = E2BIG; 1805 } 1806 out: 1807 vfs_composefh_free(fh); 1808 return (error); 1809 } 1810 1811 /* 1812 * Open a file given a file handle. 1813 * 1814 * Check permissions, allocate an open file structure, 1815 * and call the device open routine if any. 1816 */ 1817 1818 int 1819 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1820 register_t *retval) 1821 { 1822 file_t *fp; 1823 struct vnode *vp = NULL; 1824 kauth_cred_t cred = l->l_cred; 1825 file_t *nfp; 1826 int indx, error = 0; 1827 struct vattr va; 1828 fhandle_t *fh; 1829 int flags; 1830 proc_t *p; 1831 1832 p = curproc; 1833 1834 /* 1835 * Must be super user 1836 */ 1837 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1838 0, NULL, NULL, NULL))) 1839 return (error); 1840 1841 flags = FFLAGS(oflags); 1842 if ((flags & (FREAD | FWRITE)) == 0) 1843 return (EINVAL); 1844 if ((flags & O_CREAT)) 1845 return (EINVAL); 1846 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1847 return (error); 1848 fp = nfp; 1849 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1850 if (error != 0) { 1851 goto bad; 1852 } 1853 error = vfs_fhtovp(fh, &vp); 1854 if (error != 0) { 1855 goto bad; 1856 } 1857 1858 /* Now do an effective vn_open */ 1859 1860 if (vp->v_type == VSOCK) { 1861 error = EOPNOTSUPP; 1862 goto bad; 1863 } 1864 error = vn_openchk(vp, cred, flags); 1865 if (error != 0) 1866 goto bad; 1867 if (flags & O_TRUNC) { 1868 VOP_UNLOCK(vp); /* XXX */ 1869 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1870 vattr_null(&va); 1871 va.va_size = 0; 1872 error = VOP_SETATTR(vp, &va, cred); 1873 if (error) 1874 goto bad; 1875 } 1876 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1877 goto bad; 1878 if (flags & FWRITE) { 1879 mutex_enter(vp->v_interlock); 1880 vp->v_writecount++; 1881 mutex_exit(vp->v_interlock); 1882 } 1883 1884 /* done with modified vn_open, now finish what sys_open does. */ 1885 if ((error = open_setfp(l, fp, vp, indx, flags))) 1886 return error; 1887 1888 VOP_UNLOCK(vp); 1889 *retval = indx; 1890 fd_affix(p, fp, indx); 1891 vfs_copyinfh_free(fh); 1892 return (0); 1893 1894 bad: 1895 fd_abort(p, fp, indx); 1896 if (vp != NULL) 1897 vput(vp); 1898 vfs_copyinfh_free(fh); 1899 return (error); 1900 } 1901 1902 int 1903 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1904 { 1905 /* { 1906 syscallarg(const void *) fhp; 1907 syscallarg(size_t) fh_size; 1908 syscallarg(int) flags; 1909 } */ 1910 1911 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1912 SCARG(uap, flags), retval); 1913 } 1914 1915 int 1916 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1917 { 1918 int error; 1919 fhandle_t *fh; 1920 struct vnode *vp; 1921 1922 /* 1923 * Must be super user 1924 */ 1925 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1926 0, NULL, NULL, NULL))) 1927 return (error); 1928 1929 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1930 if (error != 0) 1931 return error; 1932 1933 error = vfs_fhtovp(fh, &vp); 1934 vfs_copyinfh_free(fh); 1935 if (error != 0) 1936 return error; 1937 1938 error = vn_stat(vp, sb); 1939 vput(vp); 1940 return error; 1941 } 1942 1943 1944 /* ARGSUSED */ 1945 int 1946 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 1947 { 1948 /* { 1949 syscallarg(const void *) fhp; 1950 syscallarg(size_t) fh_size; 1951 syscallarg(struct stat *) sb; 1952 } */ 1953 struct stat sb; 1954 int error; 1955 1956 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1957 if (error) 1958 return error; 1959 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1960 } 1961 1962 int 1963 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1964 int flags) 1965 { 1966 fhandle_t *fh; 1967 struct mount *mp; 1968 struct vnode *vp; 1969 int error; 1970 1971 /* 1972 * Must be super user 1973 */ 1974 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1975 0, NULL, NULL, NULL))) 1976 return error; 1977 1978 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1979 if (error != 0) 1980 return error; 1981 1982 error = vfs_fhtovp(fh, &vp); 1983 vfs_copyinfh_free(fh); 1984 if (error != 0) 1985 return error; 1986 1987 mp = vp->v_mount; 1988 error = dostatvfs(mp, sb, l, flags, 1); 1989 vput(vp); 1990 return error; 1991 } 1992 1993 /* ARGSUSED */ 1994 int 1995 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1996 { 1997 /* { 1998 syscallarg(const void *) fhp; 1999 syscallarg(size_t) fh_size; 2000 syscallarg(struct statvfs *) buf; 2001 syscallarg(int) flags; 2002 } */ 2003 struct statvfs *sb = STATVFSBUF_GET(); 2004 int error; 2005 2006 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2007 SCARG(uap, flags)); 2008 if (error == 0) 2009 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2010 STATVFSBUF_PUT(sb); 2011 return error; 2012 } 2013 2014 /* 2015 * Create a special file. 2016 */ 2017 /* ARGSUSED */ 2018 int 2019 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2020 register_t *retval) 2021 { 2022 /* { 2023 syscallarg(const char *) path; 2024 syscallarg(mode_t) mode; 2025 syscallarg(dev_t) dev; 2026 } */ 2027 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 2028 SCARG(uap, dev), retval, UIO_USERSPACE); 2029 } 2030 2031 int 2032 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2033 register_t *retval) 2034 { 2035 /* { 2036 syscallarg(int) fd; 2037 syscallarg(const char *) path; 2038 syscallarg(mode_t) mode; 2039 syscallarg(uint32_t) dev; 2040 } */ 2041 2042 return ENOSYS; 2043 } 2044 2045 int 2046 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2047 register_t *retval, enum uio_seg seg) 2048 { 2049 struct proc *p = l->l_proc; 2050 struct vnode *vp; 2051 struct vattr vattr; 2052 int error, optype; 2053 struct pathbuf *pb; 2054 struct nameidata nd; 2055 const char *pathstring; 2056 2057 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2058 0, NULL, NULL, NULL)) != 0) 2059 return (error); 2060 2061 optype = VOP_MKNOD_DESCOFFSET; 2062 2063 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2064 if (error) { 2065 return error; 2066 } 2067 pathstring = pathbuf_stringcopy_get(pb); 2068 if (pathstring == NULL) { 2069 pathbuf_destroy(pb); 2070 return ENOMEM; 2071 } 2072 2073 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2074 if ((error = namei(&nd)) != 0) 2075 goto out; 2076 vp = nd.ni_vp; 2077 2078 if (vp != NULL) 2079 error = EEXIST; 2080 else { 2081 vattr_null(&vattr); 2082 /* We will read cwdi->cwdi_cmask unlocked. */ 2083 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2084 vattr.va_rdev = dev; 2085 2086 switch (mode & S_IFMT) { 2087 case S_IFMT: /* used by badsect to flag bad sectors */ 2088 vattr.va_type = VBAD; 2089 break; 2090 case S_IFCHR: 2091 vattr.va_type = VCHR; 2092 break; 2093 case S_IFBLK: 2094 vattr.va_type = VBLK; 2095 break; 2096 case S_IFWHT: 2097 optype = VOP_WHITEOUT_DESCOFFSET; 2098 break; 2099 case S_IFREG: 2100 #if NVERIEXEC > 0 2101 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2102 O_CREAT); 2103 #endif /* NVERIEXEC > 0 */ 2104 vattr.va_type = VREG; 2105 vattr.va_rdev = VNOVAL; 2106 optype = VOP_CREATE_DESCOFFSET; 2107 break; 2108 default: 2109 error = EINVAL; 2110 break; 2111 } 2112 } 2113 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2114 && vattr.va_rdev == VNOVAL) 2115 error = EINVAL; 2116 if (!error) { 2117 switch (optype) { 2118 case VOP_WHITEOUT_DESCOFFSET: 2119 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2120 if (error) 2121 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2122 vput(nd.ni_dvp); 2123 break; 2124 2125 case VOP_MKNOD_DESCOFFSET: 2126 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2127 &nd.ni_cnd, &vattr); 2128 if (error == 0) 2129 vput(nd.ni_vp); 2130 break; 2131 2132 case VOP_CREATE_DESCOFFSET: 2133 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2134 &nd.ni_cnd, &vattr); 2135 if (error == 0) 2136 vput(nd.ni_vp); 2137 break; 2138 } 2139 } else { 2140 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2141 if (nd.ni_dvp == vp) 2142 vrele(nd.ni_dvp); 2143 else 2144 vput(nd.ni_dvp); 2145 if (vp) 2146 vrele(vp); 2147 } 2148 out: 2149 pathbuf_stringcopy_put(pb, pathstring); 2150 pathbuf_destroy(pb); 2151 return (error); 2152 } 2153 2154 /* 2155 * Create a named pipe. 2156 */ 2157 /* ARGSUSED */ 2158 int 2159 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2160 { 2161 /* { 2162 syscallarg(const char *) path; 2163 syscallarg(int) mode; 2164 } */ 2165 struct proc *p = l->l_proc; 2166 struct vattr vattr; 2167 int error; 2168 struct pathbuf *pb; 2169 struct nameidata nd; 2170 2171 error = pathbuf_copyin(SCARG(uap, path), &pb); 2172 if (error) { 2173 return error; 2174 } 2175 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2176 if ((error = namei(&nd)) != 0) { 2177 pathbuf_destroy(pb); 2178 return error; 2179 } 2180 if (nd.ni_vp != NULL) { 2181 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2182 if (nd.ni_dvp == nd.ni_vp) 2183 vrele(nd.ni_dvp); 2184 else 2185 vput(nd.ni_dvp); 2186 vrele(nd.ni_vp); 2187 pathbuf_destroy(pb); 2188 return (EEXIST); 2189 } 2190 vattr_null(&vattr); 2191 vattr.va_type = VFIFO; 2192 /* We will read cwdi->cwdi_cmask unlocked. */ 2193 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2194 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2195 if (error == 0) 2196 vput(nd.ni_vp); 2197 pathbuf_destroy(pb); 2198 return (error); 2199 } 2200 2201 int 2202 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2203 register_t *retval) 2204 { 2205 /* { 2206 syscallarg(int) fd; 2207 syscallarg(const char *) path; 2208 syscallarg(int) mode; 2209 } */ 2210 2211 return ENOSYS; 2212 } 2213 /* 2214 * Make a hard file link. 2215 */ 2216 /* ARGSUSED */ 2217 static int 2218 do_sys_link(struct lwp *l, const char *path, const char *link, 2219 int follow, register_t *retval) 2220 { 2221 struct vnode *vp; 2222 struct pathbuf *linkpb; 2223 struct nameidata nd; 2224 namei_simple_flags_t namei_simple_flags; 2225 int error; 2226 2227 if (follow) 2228 namei_simple_flags = NSM_FOLLOW_TRYEMULROOT; 2229 else 2230 namei_simple_flags = NSM_NOFOLLOW_TRYEMULROOT; 2231 2232 error = namei_simple_user(path, namei_simple_flags, &vp); 2233 if (error != 0) 2234 return (error); 2235 error = pathbuf_copyin(link, &linkpb); 2236 if (error) { 2237 goto out1; 2238 } 2239 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2240 if ((error = namei(&nd)) != 0) 2241 goto out2; 2242 if (nd.ni_vp) { 2243 error = EEXIST; 2244 goto abortop; 2245 } 2246 /* Prevent hard links on directories. */ 2247 if (vp->v_type == VDIR) { 2248 error = EPERM; 2249 goto abortop; 2250 } 2251 /* Prevent cross-mount operation. */ 2252 if (nd.ni_dvp->v_mount != vp->v_mount) { 2253 error = EXDEV; 2254 goto abortop; 2255 } 2256 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2257 out2: 2258 pathbuf_destroy(linkpb); 2259 out1: 2260 vrele(vp); 2261 return (error); 2262 abortop: 2263 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2264 if (nd.ni_dvp == nd.ni_vp) 2265 vrele(nd.ni_dvp); 2266 else 2267 vput(nd.ni_dvp); 2268 if (nd.ni_vp != NULL) 2269 vrele(nd.ni_vp); 2270 goto out2; 2271 } 2272 2273 int 2274 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2275 { 2276 /* { 2277 syscallarg(const char *) path; 2278 syscallarg(const char *) link; 2279 } */ 2280 const char *path = SCARG(uap, path); 2281 const char *link = SCARG(uap, link); 2282 2283 return do_sys_link(l, path, link, 1, retval); 2284 } 2285 2286 int 2287 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2288 register_t *retval) 2289 { 2290 /* { 2291 syscallarg(int) fd1; 2292 syscallarg(const char *) name1; 2293 syscallarg(int) fd2; 2294 syscallarg(const char *) name2; 2295 syscallarg(int) flags; 2296 } */ 2297 const char *name1 = SCARG(uap, name1); 2298 const char *name2 = SCARG(uap, name2); 2299 int follow; 2300 2301 /* 2302 * Specified fd1 and fd2 are not yet implemented 2303 */ 2304 if ((SCARG(uap, fd1) != AT_FDCWD) || (SCARG(uap, fd2) != AT_FDCWD)) 2305 return ENOSYS; 2306 2307 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2308 2309 return do_sys_link(l, name1, name2, follow, retval); 2310 } 2311 2312 2313 int 2314 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2315 { 2316 struct proc *p = curproc; 2317 struct vattr vattr; 2318 char *path; 2319 int error; 2320 struct pathbuf *linkpb; 2321 struct nameidata nd; 2322 2323 path = PNBUF_GET(); 2324 if (seg == UIO_USERSPACE) { 2325 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2326 goto out1; 2327 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2328 goto out1; 2329 } else { 2330 KASSERT(strlen(patharg) < MAXPATHLEN); 2331 strcpy(path, patharg); 2332 linkpb = pathbuf_create(link); 2333 if (linkpb == NULL) { 2334 error = ENOMEM; 2335 goto out1; 2336 } 2337 } 2338 ktrkuser("symlink-target", path, strlen(path)); 2339 2340 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2341 if ((error = namei(&nd)) != 0) 2342 goto out2; 2343 if (nd.ni_vp) { 2344 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2345 if (nd.ni_dvp == nd.ni_vp) 2346 vrele(nd.ni_dvp); 2347 else 2348 vput(nd.ni_dvp); 2349 vrele(nd.ni_vp); 2350 error = EEXIST; 2351 goto out2; 2352 } 2353 vattr_null(&vattr); 2354 vattr.va_type = VLNK; 2355 /* We will read cwdi->cwdi_cmask unlocked. */ 2356 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2357 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2358 if (error == 0) 2359 vput(nd.ni_vp); 2360 out2: 2361 pathbuf_destroy(linkpb); 2362 out1: 2363 PNBUF_PUT(path); 2364 return (error); 2365 } 2366 2367 /* 2368 * Make a symbolic link. 2369 */ 2370 /* ARGSUSED */ 2371 int 2372 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2373 { 2374 /* { 2375 syscallarg(const char *) path; 2376 syscallarg(const char *) link; 2377 } */ 2378 2379 return do_sys_symlink(SCARG(uap, path), SCARG(uap, link), 2380 UIO_USERSPACE); 2381 } 2382 2383 int 2384 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2385 register_t *retval) 2386 { 2387 /* { 2388 syscallarg(int) fd; 2389 syscallarg(const char *) path; 2390 syscallarg(const char *) link; 2391 } */ 2392 2393 return ENOSYS; 2394 } 2395 2396 /* 2397 * Delete a whiteout from the filesystem. 2398 */ 2399 /* ARGSUSED */ 2400 int 2401 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2402 { 2403 /* { 2404 syscallarg(const char *) path; 2405 } */ 2406 int error; 2407 struct pathbuf *pb; 2408 struct nameidata nd; 2409 2410 error = pathbuf_copyin(SCARG(uap, path), &pb); 2411 if (error) { 2412 return error; 2413 } 2414 2415 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2416 error = namei(&nd); 2417 if (error) { 2418 pathbuf_destroy(pb); 2419 return (error); 2420 } 2421 2422 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2423 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2424 if (nd.ni_dvp == nd.ni_vp) 2425 vrele(nd.ni_dvp); 2426 else 2427 vput(nd.ni_dvp); 2428 if (nd.ni_vp) 2429 vrele(nd.ni_vp); 2430 pathbuf_destroy(pb); 2431 return (EEXIST); 2432 } 2433 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2434 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2435 vput(nd.ni_dvp); 2436 pathbuf_destroy(pb); 2437 return (error); 2438 } 2439 2440 /* 2441 * Delete a name from the filesystem. 2442 */ 2443 /* ARGSUSED */ 2444 int 2445 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2446 { 2447 /* { 2448 syscallarg(const char *) path; 2449 } */ 2450 2451 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2452 } 2453 2454 int 2455 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2456 register_t *retval) 2457 { 2458 /* { 2459 syscallarg(int) fd; 2460 syscallarg(const char *) path; 2461 } */ 2462 2463 return ENOSYS; 2464 } 2465 2466 int 2467 do_sys_unlink(const char *arg, enum uio_seg seg) 2468 { 2469 struct vnode *vp; 2470 int error; 2471 struct pathbuf *pb; 2472 struct nameidata nd; 2473 const char *pathstring; 2474 2475 error = pathbuf_maybe_copyin(arg, seg, &pb); 2476 if (error) { 2477 return error; 2478 } 2479 pathstring = pathbuf_stringcopy_get(pb); 2480 if (pathstring == NULL) { 2481 pathbuf_destroy(pb); 2482 return ENOMEM; 2483 } 2484 2485 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2486 if ((error = namei(&nd)) != 0) 2487 goto out; 2488 vp = nd.ni_vp; 2489 2490 /* 2491 * The root of a mounted filesystem cannot be deleted. 2492 */ 2493 if (vp->v_vflag & VV_ROOT) { 2494 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2495 if (nd.ni_dvp == vp) 2496 vrele(nd.ni_dvp); 2497 else 2498 vput(nd.ni_dvp); 2499 vput(vp); 2500 error = EBUSY; 2501 goto out; 2502 } 2503 2504 #if NVERIEXEC > 0 2505 /* Handle remove requests for veriexec entries. */ 2506 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2507 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2508 if (nd.ni_dvp == vp) 2509 vrele(nd.ni_dvp); 2510 else 2511 vput(nd.ni_dvp); 2512 vput(vp); 2513 goto out; 2514 } 2515 #endif /* NVERIEXEC > 0 */ 2516 2517 #ifdef FILEASSOC 2518 (void)fileassoc_file_delete(vp); 2519 #endif /* FILEASSOC */ 2520 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2521 out: 2522 pathbuf_stringcopy_put(pb, pathstring); 2523 pathbuf_destroy(pb); 2524 return (error); 2525 } 2526 2527 /* 2528 * Reposition read/write file offset. 2529 */ 2530 int 2531 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2532 { 2533 /* { 2534 syscallarg(int) fd; 2535 syscallarg(int) pad; 2536 syscallarg(off_t) offset; 2537 syscallarg(int) whence; 2538 } */ 2539 kauth_cred_t cred = l->l_cred; 2540 file_t *fp; 2541 struct vnode *vp; 2542 struct vattr vattr; 2543 off_t newoff; 2544 int error, fd; 2545 2546 fd = SCARG(uap, fd); 2547 2548 if ((fp = fd_getfile(fd)) == NULL) 2549 return (EBADF); 2550 2551 vp = fp->f_data; 2552 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2553 error = ESPIPE; 2554 goto out; 2555 } 2556 2557 switch (SCARG(uap, whence)) { 2558 case SEEK_CUR: 2559 newoff = fp->f_offset + SCARG(uap, offset); 2560 break; 2561 case SEEK_END: 2562 vn_lock(vp, LK_SHARED | LK_RETRY); 2563 error = VOP_GETATTR(vp, &vattr, cred); 2564 VOP_UNLOCK(vp); 2565 if (error) { 2566 goto out; 2567 } 2568 newoff = SCARG(uap, offset) + vattr.va_size; 2569 break; 2570 case SEEK_SET: 2571 newoff = SCARG(uap, offset); 2572 break; 2573 default: 2574 error = EINVAL; 2575 goto out; 2576 } 2577 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2578 *(off_t *)retval = fp->f_offset = newoff; 2579 } 2580 out: 2581 fd_putfile(fd); 2582 return (error); 2583 } 2584 2585 /* 2586 * Positional read system call. 2587 */ 2588 int 2589 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2590 { 2591 /* { 2592 syscallarg(int) fd; 2593 syscallarg(void *) buf; 2594 syscallarg(size_t) nbyte; 2595 syscallarg(off_t) offset; 2596 } */ 2597 file_t *fp; 2598 struct vnode *vp; 2599 off_t offset; 2600 int error, fd = SCARG(uap, fd); 2601 2602 if ((fp = fd_getfile(fd)) == NULL) 2603 return (EBADF); 2604 2605 if ((fp->f_flag & FREAD) == 0) { 2606 fd_putfile(fd); 2607 return (EBADF); 2608 } 2609 2610 vp = fp->f_data; 2611 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2612 error = ESPIPE; 2613 goto out; 2614 } 2615 2616 offset = SCARG(uap, offset); 2617 2618 /* 2619 * XXX This works because no file systems actually 2620 * XXX take any action on the seek operation. 2621 */ 2622 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2623 goto out; 2624 2625 /* dofileread() will unuse the descriptor for us */ 2626 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2627 &offset, 0, retval)); 2628 2629 out: 2630 fd_putfile(fd); 2631 return (error); 2632 } 2633 2634 /* 2635 * Positional scatter read system call. 2636 */ 2637 int 2638 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2639 { 2640 /* { 2641 syscallarg(int) fd; 2642 syscallarg(const struct iovec *) iovp; 2643 syscallarg(int) iovcnt; 2644 syscallarg(off_t) offset; 2645 } */ 2646 off_t offset = SCARG(uap, offset); 2647 2648 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2649 SCARG(uap, iovcnt), &offset, 0, retval); 2650 } 2651 2652 /* 2653 * Positional write system call. 2654 */ 2655 int 2656 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2657 { 2658 /* { 2659 syscallarg(int) fd; 2660 syscallarg(const void *) buf; 2661 syscallarg(size_t) nbyte; 2662 syscallarg(off_t) offset; 2663 } */ 2664 file_t *fp; 2665 struct vnode *vp; 2666 off_t offset; 2667 int error, fd = SCARG(uap, fd); 2668 2669 if ((fp = fd_getfile(fd)) == NULL) 2670 return (EBADF); 2671 2672 if ((fp->f_flag & FWRITE) == 0) { 2673 fd_putfile(fd); 2674 return (EBADF); 2675 } 2676 2677 vp = fp->f_data; 2678 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2679 error = ESPIPE; 2680 goto out; 2681 } 2682 2683 offset = SCARG(uap, offset); 2684 2685 /* 2686 * XXX This works because no file systems actually 2687 * XXX take any action on the seek operation. 2688 */ 2689 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2690 goto out; 2691 2692 /* dofilewrite() will unuse the descriptor for us */ 2693 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2694 &offset, 0, retval)); 2695 2696 out: 2697 fd_putfile(fd); 2698 return (error); 2699 } 2700 2701 /* 2702 * Positional gather write system call. 2703 */ 2704 int 2705 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2706 { 2707 /* { 2708 syscallarg(int) fd; 2709 syscallarg(const struct iovec *) iovp; 2710 syscallarg(int) iovcnt; 2711 syscallarg(off_t) offset; 2712 } */ 2713 off_t offset = SCARG(uap, offset); 2714 2715 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2716 SCARG(uap, iovcnt), &offset, 0, retval); 2717 } 2718 2719 /* 2720 * Check access permissions. 2721 */ 2722 int 2723 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2724 { 2725 /* { 2726 syscallarg(const char *) path; 2727 syscallarg(int) flags; 2728 } */ 2729 kauth_cred_t cred; 2730 struct vnode *vp; 2731 int error, flags; 2732 struct pathbuf *pb; 2733 struct nameidata nd; 2734 2735 CTASSERT(F_OK == 0); 2736 if ((SCARG(uap, flags) & ~(R_OK | W_OK | X_OK)) != 0) { 2737 /* nonsense flags */ 2738 return EINVAL; 2739 } 2740 2741 error = pathbuf_copyin(SCARG(uap, path), &pb); 2742 if (error) { 2743 return error; 2744 } 2745 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2746 2747 /* Override default credentials */ 2748 cred = kauth_cred_dup(l->l_cred); 2749 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2750 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2751 nd.ni_cnd.cn_cred = cred; 2752 2753 if ((error = namei(&nd)) != 0) { 2754 pathbuf_destroy(pb); 2755 goto out; 2756 } 2757 vp = nd.ni_vp; 2758 pathbuf_destroy(pb); 2759 2760 /* Flags == 0 means only check for existence. */ 2761 if (SCARG(uap, flags)) { 2762 flags = 0; 2763 if (SCARG(uap, flags) & R_OK) 2764 flags |= VREAD; 2765 if (SCARG(uap, flags) & W_OK) 2766 flags |= VWRITE; 2767 if (SCARG(uap, flags) & X_OK) 2768 flags |= VEXEC; 2769 2770 error = VOP_ACCESS(vp, flags, cred); 2771 if (!error && (flags & VWRITE)) 2772 error = vn_writechk(vp); 2773 } 2774 vput(vp); 2775 out: 2776 kauth_cred_free(cred); 2777 return (error); 2778 } 2779 2780 int 2781 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 2782 register_t *retval) 2783 { 2784 /* { 2785 syscallarg(int) fd; 2786 syscallarg(const char *) path; 2787 syscallarg(int) amode; 2788 syscallarg(int) flag; 2789 } */ 2790 2791 return ENOSYS; 2792 } 2793 2794 /* 2795 * Common code for all sys_stat functions, including compat versions. 2796 */ 2797 int 2798 do_sys_stat(const char *userpath, unsigned int nd_flags, struct stat *sb) 2799 { 2800 int error; 2801 struct pathbuf *pb; 2802 struct nameidata nd; 2803 2804 error = pathbuf_copyin(userpath, &pb); 2805 if (error) { 2806 return error; 2807 } 2808 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, pb); 2809 error = namei(&nd); 2810 if (error != 0) { 2811 pathbuf_destroy(pb); 2812 return error; 2813 } 2814 error = vn_stat(nd.ni_vp, sb); 2815 vput(nd.ni_vp); 2816 pathbuf_destroy(pb); 2817 return error; 2818 } 2819 2820 /* 2821 * Get file status; this version follows links. 2822 */ 2823 /* ARGSUSED */ 2824 int 2825 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2826 { 2827 /* { 2828 syscallarg(const char *) path; 2829 syscallarg(struct stat *) ub; 2830 } */ 2831 struct stat sb; 2832 int error; 2833 2834 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2835 if (error) 2836 return error; 2837 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2838 } 2839 2840 /* 2841 * Get file status; this version does not follow links. 2842 */ 2843 /* ARGSUSED */ 2844 int 2845 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2846 { 2847 /* { 2848 syscallarg(const char *) path; 2849 syscallarg(struct stat *) ub; 2850 } */ 2851 struct stat sb; 2852 int error; 2853 2854 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2855 if (error) 2856 return error; 2857 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2858 } 2859 2860 int 2861 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 2862 register_t *retval) 2863 { 2864 /* { 2865 syscallarg(int) fd; 2866 syscallarg(const char *) path; 2867 syscallarg(struct stat *) ub; 2868 syscallarg(int) flag; 2869 } */ 2870 2871 return ENOSYS; 2872 } 2873 /* 2874 * Get configurable pathname variables. 2875 */ 2876 /* ARGSUSED */ 2877 int 2878 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2879 { 2880 /* { 2881 syscallarg(const char *) path; 2882 syscallarg(int) name; 2883 } */ 2884 int error; 2885 struct pathbuf *pb; 2886 struct nameidata nd; 2887 2888 error = pathbuf_copyin(SCARG(uap, path), &pb); 2889 if (error) { 2890 return error; 2891 } 2892 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2893 if ((error = namei(&nd)) != 0) { 2894 pathbuf_destroy(pb); 2895 return (error); 2896 } 2897 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2898 vput(nd.ni_vp); 2899 pathbuf_destroy(pb); 2900 return (error); 2901 } 2902 2903 /* 2904 * Return target name of a symbolic link. 2905 */ 2906 /* ARGSUSED */ 2907 int 2908 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2909 { 2910 /* { 2911 syscallarg(const char *) path; 2912 syscallarg(char *) buf; 2913 syscallarg(size_t) count; 2914 } */ 2915 struct vnode *vp; 2916 struct iovec aiov; 2917 struct uio auio; 2918 int error; 2919 struct pathbuf *pb; 2920 struct nameidata nd; 2921 2922 error = pathbuf_copyin(SCARG(uap, path), &pb); 2923 if (error) { 2924 return error; 2925 } 2926 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2927 if ((error = namei(&nd)) != 0) { 2928 pathbuf_destroy(pb); 2929 return error; 2930 } 2931 vp = nd.ni_vp; 2932 pathbuf_destroy(pb); 2933 if (vp->v_type != VLNK) 2934 error = EINVAL; 2935 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2936 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2937 aiov.iov_base = SCARG(uap, buf); 2938 aiov.iov_len = SCARG(uap, count); 2939 auio.uio_iov = &aiov; 2940 auio.uio_iovcnt = 1; 2941 auio.uio_offset = 0; 2942 auio.uio_rw = UIO_READ; 2943 KASSERT(l == curlwp); 2944 auio.uio_vmspace = l->l_proc->p_vmspace; 2945 auio.uio_resid = SCARG(uap, count); 2946 error = VOP_READLINK(vp, &auio, l->l_cred); 2947 } 2948 vput(vp); 2949 *retval = SCARG(uap, count) - auio.uio_resid; 2950 return (error); 2951 } 2952 2953 int 2954 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 2955 register_t *retval) 2956 { 2957 /* { 2958 syscallarg(int) fd; 2959 syscallarg(const char *) path; 2960 syscallarg(char *) buf; 2961 syscallarg(size_t) count; 2962 } */ 2963 2964 return ENOSYS; 2965 } 2966 2967 /* 2968 * Change flags of a file given a path name. 2969 */ 2970 /* ARGSUSED */ 2971 int 2972 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2973 { 2974 /* { 2975 syscallarg(const char *) path; 2976 syscallarg(u_long) flags; 2977 } */ 2978 struct vnode *vp; 2979 int error; 2980 2981 error = namei_simple_user(SCARG(uap, path), 2982 NSM_FOLLOW_TRYEMULROOT, &vp); 2983 if (error != 0) 2984 return (error); 2985 error = change_flags(vp, SCARG(uap, flags), l); 2986 vput(vp); 2987 return (error); 2988 } 2989 2990 /* 2991 * Change flags of a file given a file descriptor. 2992 */ 2993 /* ARGSUSED */ 2994 int 2995 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2996 { 2997 /* { 2998 syscallarg(int) fd; 2999 syscallarg(u_long) flags; 3000 } */ 3001 struct vnode *vp; 3002 file_t *fp; 3003 int error; 3004 3005 /* fd_getvnode() will use the descriptor for us */ 3006 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3007 return (error); 3008 vp = fp->f_data; 3009 error = change_flags(vp, SCARG(uap, flags), l); 3010 VOP_UNLOCK(vp); 3011 fd_putfile(SCARG(uap, fd)); 3012 return (error); 3013 } 3014 3015 /* 3016 * Change flags of a file given a path name; this version does 3017 * not follow links. 3018 */ 3019 int 3020 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3021 { 3022 /* { 3023 syscallarg(const char *) path; 3024 syscallarg(u_long) flags; 3025 } */ 3026 struct vnode *vp; 3027 int error; 3028 3029 error = namei_simple_user(SCARG(uap, path), 3030 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3031 if (error != 0) 3032 return (error); 3033 error = change_flags(vp, SCARG(uap, flags), l); 3034 vput(vp); 3035 return (error); 3036 } 3037 3038 /* 3039 * Common routine to change flags of a file. 3040 */ 3041 int 3042 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3043 { 3044 struct vattr vattr; 3045 int error; 3046 3047 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3048 3049 vattr_null(&vattr); 3050 vattr.va_flags = flags; 3051 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3052 3053 return (error); 3054 } 3055 3056 /* 3057 * Change mode of a file given path name; this version follows links. 3058 */ 3059 /* ARGSUSED */ 3060 int 3061 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3062 { 3063 /* { 3064 syscallarg(const char *) path; 3065 syscallarg(int) mode; 3066 } */ 3067 int error; 3068 struct vnode *vp; 3069 3070 error = namei_simple_user(SCARG(uap, path), 3071 NSM_FOLLOW_TRYEMULROOT, &vp); 3072 if (error != 0) 3073 return (error); 3074 3075 error = change_mode(vp, SCARG(uap, mode), l); 3076 3077 vrele(vp); 3078 return (error); 3079 } 3080 3081 /* 3082 * Change mode of a file given a file descriptor. 3083 */ 3084 /* ARGSUSED */ 3085 int 3086 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3087 { 3088 /* { 3089 syscallarg(int) fd; 3090 syscallarg(int) mode; 3091 } */ 3092 file_t *fp; 3093 int error; 3094 3095 /* fd_getvnode() will use the descriptor for us */ 3096 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3097 return (error); 3098 error = change_mode(fp->f_data, SCARG(uap, mode), l); 3099 fd_putfile(SCARG(uap, fd)); 3100 return (error); 3101 } 3102 3103 int 3104 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3105 register_t *retval) 3106 { 3107 /* { 3108 syscallarg(int) fd; 3109 syscallarg(const char *) path; 3110 syscallarg(int) mode; 3111 syscallarg(int) flag; 3112 } */ 3113 3114 return ENOSYS; 3115 } 3116 3117 /* 3118 * Change mode of a file given path name; this version does not follow links. 3119 */ 3120 /* ARGSUSED */ 3121 int 3122 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3123 { 3124 /* { 3125 syscallarg(const char *) path; 3126 syscallarg(int) mode; 3127 } */ 3128 int error; 3129 struct vnode *vp; 3130 3131 error = namei_simple_user(SCARG(uap, path), 3132 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3133 if (error != 0) 3134 return (error); 3135 3136 error = change_mode(vp, SCARG(uap, mode), l); 3137 3138 vrele(vp); 3139 return (error); 3140 } 3141 3142 /* 3143 * Common routine to set mode given a vnode. 3144 */ 3145 static int 3146 change_mode(struct vnode *vp, int mode, struct lwp *l) 3147 { 3148 struct vattr vattr; 3149 int error; 3150 3151 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3152 vattr_null(&vattr); 3153 vattr.va_mode = mode & ALLPERMS; 3154 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3155 VOP_UNLOCK(vp); 3156 return (error); 3157 } 3158 3159 /* 3160 * Set ownership given a path name; this version follows links. 3161 */ 3162 /* ARGSUSED */ 3163 int 3164 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3165 { 3166 /* { 3167 syscallarg(const char *) path; 3168 syscallarg(uid_t) uid; 3169 syscallarg(gid_t) gid; 3170 } */ 3171 int error; 3172 struct vnode *vp; 3173 3174 error = namei_simple_user(SCARG(uap, path), 3175 NSM_FOLLOW_TRYEMULROOT, &vp); 3176 if (error != 0) 3177 return (error); 3178 3179 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3180 3181 vrele(vp); 3182 return (error); 3183 } 3184 3185 /* 3186 * Set ownership given a path name; this version follows links. 3187 * Provides POSIX semantics. 3188 */ 3189 /* ARGSUSED */ 3190 int 3191 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3192 { 3193 /* { 3194 syscallarg(const char *) path; 3195 syscallarg(uid_t) uid; 3196 syscallarg(gid_t) gid; 3197 } */ 3198 int error; 3199 struct vnode *vp; 3200 3201 error = namei_simple_user(SCARG(uap, path), 3202 NSM_FOLLOW_TRYEMULROOT, &vp); 3203 if (error != 0) 3204 return (error); 3205 3206 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3207 3208 vrele(vp); 3209 return (error); 3210 } 3211 3212 /* 3213 * Set ownership given a file descriptor. 3214 */ 3215 /* ARGSUSED */ 3216 int 3217 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3218 { 3219 /* { 3220 syscallarg(int) fd; 3221 syscallarg(uid_t) uid; 3222 syscallarg(gid_t) gid; 3223 } */ 3224 int error; 3225 file_t *fp; 3226 3227 /* fd_getvnode() will use the descriptor for us */ 3228 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3229 return (error); 3230 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3231 l, 0); 3232 fd_putfile(SCARG(uap, fd)); 3233 return (error); 3234 } 3235 3236 int 3237 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3238 register_t *retval) 3239 { 3240 /* { 3241 syscallarg(int) fd; 3242 syscallarg(const char *) path; 3243 syscallarg(uid_t) uid; 3244 syscallarg(gid_t) gid; 3245 syscallarg(int) flag; 3246 } */ 3247 3248 return ENOSYS; 3249 } 3250 3251 /* 3252 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3253 */ 3254 /* ARGSUSED */ 3255 int 3256 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3257 { 3258 /* { 3259 syscallarg(int) fd; 3260 syscallarg(uid_t) uid; 3261 syscallarg(gid_t) gid; 3262 } */ 3263 int error; 3264 file_t *fp; 3265 3266 /* fd_getvnode() will use the descriptor for us */ 3267 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3268 return (error); 3269 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3270 l, 1); 3271 fd_putfile(SCARG(uap, fd)); 3272 return (error); 3273 } 3274 3275 /* 3276 * Set ownership given a path name; this version does not follow links. 3277 */ 3278 /* ARGSUSED */ 3279 int 3280 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3281 { 3282 /* { 3283 syscallarg(const char *) path; 3284 syscallarg(uid_t) uid; 3285 syscallarg(gid_t) gid; 3286 } */ 3287 int error; 3288 struct vnode *vp; 3289 3290 error = namei_simple_user(SCARG(uap, path), 3291 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3292 if (error != 0) 3293 return (error); 3294 3295 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3296 3297 vrele(vp); 3298 return (error); 3299 } 3300 3301 /* 3302 * Set ownership given a path name; this version does not follow links. 3303 * Provides POSIX/XPG semantics. 3304 */ 3305 /* ARGSUSED */ 3306 int 3307 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3308 { 3309 /* { 3310 syscallarg(const char *) path; 3311 syscallarg(uid_t) uid; 3312 syscallarg(gid_t) gid; 3313 } */ 3314 int error; 3315 struct vnode *vp; 3316 3317 error = namei_simple_user(SCARG(uap, path), 3318 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3319 if (error != 0) 3320 return (error); 3321 3322 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3323 3324 vrele(vp); 3325 return (error); 3326 } 3327 3328 /* 3329 * Common routine to set ownership given a vnode. 3330 */ 3331 static int 3332 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3333 int posix_semantics) 3334 { 3335 struct vattr vattr; 3336 mode_t newmode; 3337 int error; 3338 3339 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3340 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3341 goto out; 3342 3343 #define CHANGED(x) ((int)(x) != -1) 3344 newmode = vattr.va_mode; 3345 if (posix_semantics) { 3346 /* 3347 * POSIX/XPG semantics: if the caller is not the super-user, 3348 * clear set-user-id and set-group-id bits. Both POSIX and 3349 * the XPG consider the behaviour for calls by the super-user 3350 * implementation-defined; we leave the set-user-id and set- 3351 * group-id settings intact in that case. 3352 */ 3353 if (vattr.va_mode & S_ISUID) { 3354 if (kauth_authorize_vnode(l->l_cred, 3355 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3356 newmode &= ~S_ISUID; 3357 } 3358 if (vattr.va_mode & S_ISGID) { 3359 if (kauth_authorize_vnode(l->l_cred, 3360 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3361 newmode &= ~S_ISGID; 3362 } 3363 } else { 3364 /* 3365 * NetBSD semantics: when changing owner and/or group, 3366 * clear the respective bit(s). 3367 */ 3368 if (CHANGED(uid)) 3369 newmode &= ~S_ISUID; 3370 if (CHANGED(gid)) 3371 newmode &= ~S_ISGID; 3372 } 3373 /* Update va_mode iff altered. */ 3374 if (vattr.va_mode == newmode) 3375 newmode = VNOVAL; 3376 3377 vattr_null(&vattr); 3378 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3379 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3380 vattr.va_mode = newmode; 3381 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3382 #undef CHANGED 3383 3384 out: 3385 VOP_UNLOCK(vp); 3386 return (error); 3387 } 3388 3389 /* 3390 * Set the access and modification times given a path name; this 3391 * version follows links. 3392 */ 3393 /* ARGSUSED */ 3394 int 3395 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3396 register_t *retval) 3397 { 3398 /* { 3399 syscallarg(const char *) path; 3400 syscallarg(const struct timeval *) tptr; 3401 } */ 3402 3403 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3404 SCARG(uap, tptr), UIO_USERSPACE); 3405 } 3406 3407 /* 3408 * Set the access and modification times given a file descriptor. 3409 */ 3410 /* ARGSUSED */ 3411 int 3412 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3413 register_t *retval) 3414 { 3415 /* { 3416 syscallarg(int) fd; 3417 syscallarg(const struct timeval *) tptr; 3418 } */ 3419 int error; 3420 file_t *fp; 3421 3422 /* fd_getvnode() will use the descriptor for us */ 3423 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3424 return (error); 3425 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3426 UIO_USERSPACE); 3427 fd_putfile(SCARG(uap, fd)); 3428 return (error); 3429 } 3430 3431 int 3432 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3433 register_t *retval) 3434 { 3435 /* { 3436 syscallarg(int) fd; 3437 syscallarg(const struct timespec *) tptr; 3438 } */ 3439 int error; 3440 file_t *fp; 3441 3442 /* fd_getvnode() will use the descriptor for us */ 3443 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3444 return (error); 3445 error = do_sys_utimens(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3446 UIO_USERSPACE); 3447 fd_putfile(SCARG(uap, fd)); 3448 return (error); 3449 } 3450 3451 /* 3452 * Set the access and modification times given a path name; this 3453 * version does not follow links. 3454 */ 3455 int 3456 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3457 register_t *retval) 3458 { 3459 /* { 3460 syscallarg(const char *) path; 3461 syscallarg(const struct timeval *) tptr; 3462 } */ 3463 3464 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3465 SCARG(uap, tptr), UIO_USERSPACE); 3466 } 3467 3468 int 3469 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3470 register_t *retval) 3471 { 3472 /* { 3473 syscallarg(int) fd; 3474 syscallarg(const char *) path; 3475 syscallarg(const struct timespec *) tptr; 3476 syscallarg(int) flag; 3477 } */ 3478 int follow; 3479 const struct timespec *tptr; 3480 3481 /* 3482 * Specified fd is not yet implemented 3483 */ 3484 if (SCARG(uap, fd) != AT_FDCWD) 3485 return ENOSYS; 3486 3487 tptr = SCARG(uap, tptr); 3488 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3489 3490 return do_sys_utimens(l, NULL, SCARG(uap, path), follow, 3491 tptr, UIO_USERSPACE); 3492 } 3493 3494 /* 3495 * Common routine to set access and modification times given a vnode. 3496 */ 3497 int 3498 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3499 const struct timespec *tptr, enum uio_seg seg) 3500 { 3501 struct vattr vattr; 3502 int error, dorele = 0; 3503 namei_simple_flags_t sflags; 3504 3505 bool vanull, setbirthtime; 3506 struct timespec ts[2]; 3507 3508 /* 3509 * I have checked all callers and they pass either FOLLOW, 3510 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3511 * is 0. More to the point, they don't pass anything else. 3512 * Let's keep it that way at least until the namei interfaces 3513 * are fully sanitized. 3514 */ 3515 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3516 sflags = (flag == FOLLOW) ? 3517 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3518 3519 if (tptr == NULL) { 3520 vanull = true; 3521 nanotime(&ts[0]); 3522 ts[1] = ts[0]; 3523 } else { 3524 vanull = false; 3525 if (seg != UIO_SYSSPACE) { 3526 error = copyin(tptr, ts, sizeof (ts)); 3527 if (error != 0) 3528 return error; 3529 } else { 3530 ts[0] = tptr[0]; 3531 ts[1] = tptr[1]; 3532 } 3533 } 3534 3535 if (ts[0].tv_nsec == UTIME_NOW) { 3536 nanotime(&ts[0]); 3537 if (ts[1].tv_nsec == UTIME_NOW) { 3538 vanull = true; 3539 ts[1] = ts[0]; 3540 } 3541 } else if (ts[1].tv_nsec == UTIME_NOW) 3542 nanotime(&ts[1]); 3543 3544 if (vp == NULL) { 3545 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3546 error = namei_simple_user(path, sflags, &vp); 3547 if (error != 0) 3548 return error; 3549 dorele = 1; 3550 } 3551 3552 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3553 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3554 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3555 vattr_null(&vattr); 3556 3557 if (ts[0].tv_nsec != UTIME_OMIT) 3558 vattr.va_atime = ts[0]; 3559 3560 if (ts[1].tv_nsec != UTIME_OMIT) { 3561 vattr.va_mtime = ts[1]; 3562 if (setbirthtime) 3563 vattr.va_birthtime = ts[1]; 3564 } 3565 3566 if (vanull) 3567 vattr.va_vaflags |= VA_UTIMES_NULL; 3568 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3569 VOP_UNLOCK(vp); 3570 3571 if (dorele != 0) 3572 vrele(vp); 3573 3574 return error; 3575 } 3576 3577 int 3578 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3579 const struct timeval *tptr, enum uio_seg seg) 3580 { 3581 struct timespec ts[2]; 3582 struct timespec *tsptr = NULL; 3583 int error; 3584 3585 if (tptr != NULL) { 3586 struct timeval tv[2]; 3587 3588 if (seg != UIO_SYSSPACE) { 3589 error = copyin(tptr, tv, sizeof (tv)); 3590 if (error != 0) 3591 return error; 3592 tptr = tv; 3593 } 3594 3595 if ((tv[0].tv_usec == UTIME_NOW) || 3596 (tv[0].tv_usec == UTIME_OMIT)) 3597 ts[0].tv_nsec = tv[0].tv_usec; 3598 else 3599 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3600 3601 if ((tv[1].tv_usec == UTIME_NOW) || 3602 (tv[1].tv_usec == UTIME_OMIT)) 3603 ts[1].tv_nsec = tv[1].tv_usec; 3604 else 3605 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3606 3607 tsptr = &ts[0]; 3608 } 3609 3610 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3611 } 3612 3613 /* 3614 * Truncate a file given its path name. 3615 */ 3616 /* ARGSUSED */ 3617 int 3618 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3619 { 3620 /* { 3621 syscallarg(const char *) path; 3622 syscallarg(int) pad; 3623 syscallarg(off_t) length; 3624 } */ 3625 struct vnode *vp; 3626 struct vattr vattr; 3627 int error; 3628 3629 error = namei_simple_user(SCARG(uap, path), 3630 NSM_FOLLOW_TRYEMULROOT, &vp); 3631 if (error != 0) 3632 return (error); 3633 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3634 if (vp->v_type == VDIR) 3635 error = EISDIR; 3636 else if ((error = vn_writechk(vp)) == 0 && 3637 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3638 vattr_null(&vattr); 3639 vattr.va_size = SCARG(uap, length); 3640 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3641 } 3642 vput(vp); 3643 return (error); 3644 } 3645 3646 /* 3647 * Truncate a file given a file descriptor. 3648 */ 3649 /* ARGSUSED */ 3650 int 3651 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3652 { 3653 /* { 3654 syscallarg(int) fd; 3655 syscallarg(int) pad; 3656 syscallarg(off_t) length; 3657 } */ 3658 struct vattr vattr; 3659 struct vnode *vp; 3660 file_t *fp; 3661 int error; 3662 3663 /* fd_getvnode() will use the descriptor for us */ 3664 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3665 return (error); 3666 if ((fp->f_flag & FWRITE) == 0) { 3667 error = EINVAL; 3668 goto out; 3669 } 3670 vp = fp->f_data; 3671 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3672 if (vp->v_type == VDIR) 3673 error = EISDIR; 3674 else if ((error = vn_writechk(vp)) == 0) { 3675 vattr_null(&vattr); 3676 vattr.va_size = SCARG(uap, length); 3677 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3678 } 3679 VOP_UNLOCK(vp); 3680 out: 3681 fd_putfile(SCARG(uap, fd)); 3682 return (error); 3683 } 3684 3685 /* 3686 * Sync an open file. 3687 */ 3688 /* ARGSUSED */ 3689 int 3690 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3691 { 3692 /* { 3693 syscallarg(int) fd; 3694 } */ 3695 struct vnode *vp; 3696 file_t *fp; 3697 int error; 3698 3699 /* fd_getvnode() will use the descriptor for us */ 3700 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3701 return (error); 3702 vp = fp->f_data; 3703 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3704 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3705 VOP_UNLOCK(vp); 3706 fd_putfile(SCARG(uap, fd)); 3707 return (error); 3708 } 3709 3710 /* 3711 * Sync a range of file data. API modeled after that found in AIX. 3712 * 3713 * FDATASYNC indicates that we need only save enough metadata to be able 3714 * to re-read the written data. Note we duplicate AIX's requirement that 3715 * the file be open for writing. 3716 */ 3717 /* ARGSUSED */ 3718 int 3719 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3720 { 3721 /* { 3722 syscallarg(int) fd; 3723 syscallarg(int) flags; 3724 syscallarg(off_t) start; 3725 syscallarg(off_t) length; 3726 } */ 3727 struct vnode *vp; 3728 file_t *fp; 3729 int flags, nflags; 3730 off_t s, e, len; 3731 int error; 3732 3733 /* fd_getvnode() will use the descriptor for us */ 3734 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3735 return (error); 3736 3737 if ((fp->f_flag & FWRITE) == 0) { 3738 error = EBADF; 3739 goto out; 3740 } 3741 3742 flags = SCARG(uap, flags); 3743 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3744 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3745 error = EINVAL; 3746 goto out; 3747 } 3748 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3749 if (flags & FDATASYNC) 3750 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3751 else 3752 nflags = FSYNC_WAIT; 3753 if (flags & FDISKSYNC) 3754 nflags |= FSYNC_CACHE; 3755 3756 len = SCARG(uap, length); 3757 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 3758 if (len) { 3759 s = SCARG(uap, start); 3760 e = s + len; 3761 if (e < s) { 3762 error = EINVAL; 3763 goto out; 3764 } 3765 } else { 3766 e = 0; 3767 s = 0; 3768 } 3769 3770 vp = fp->f_data; 3771 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3772 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3773 VOP_UNLOCK(vp); 3774 out: 3775 fd_putfile(SCARG(uap, fd)); 3776 return (error); 3777 } 3778 3779 /* 3780 * Sync the data of an open file. 3781 */ 3782 /* ARGSUSED */ 3783 int 3784 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3785 { 3786 /* { 3787 syscallarg(int) fd; 3788 } */ 3789 struct vnode *vp; 3790 file_t *fp; 3791 int error; 3792 3793 /* fd_getvnode() will use the descriptor for us */ 3794 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3795 return (error); 3796 if ((fp->f_flag & FWRITE) == 0) { 3797 fd_putfile(SCARG(uap, fd)); 3798 return (EBADF); 3799 } 3800 vp = fp->f_data; 3801 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3802 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3803 VOP_UNLOCK(vp); 3804 fd_putfile(SCARG(uap, fd)); 3805 return (error); 3806 } 3807 3808 /* 3809 * Rename files, (standard) BSD semantics frontend. 3810 */ 3811 /* ARGSUSED */ 3812 int 3813 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3814 { 3815 /* { 3816 syscallarg(const char *) from; 3817 syscallarg(const char *) to; 3818 } */ 3819 3820 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3821 } 3822 3823 int 3824 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 3825 register_t *retval) 3826 { 3827 /* { 3828 syscallarg(int) fromfd; 3829 syscallarg(const char *) from; 3830 syscallarg(int) tofd; 3831 syscallarg(const char *) to; 3832 } */ 3833 3834 return ENOSYS; 3835 } 3836 3837 /* 3838 * Rename files, POSIX semantics frontend. 3839 */ 3840 /* ARGSUSED */ 3841 int 3842 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3843 { 3844 /* { 3845 syscallarg(const char *) from; 3846 syscallarg(const char *) to; 3847 } */ 3848 3849 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3850 } 3851 3852 /* 3853 * Rename files. Source and destination must either both be directories, 3854 * or both not be directories. If target is a directory, it must be empty. 3855 * If `from' and `to' refer to the same object, the value of the `retain' 3856 * argument is used to determine whether `from' will be 3857 * 3858 * (retain == 0) deleted unless `from' and `to' refer to the same 3859 * object in the file system's name space (BSD). 3860 * (retain == 1) always retained (POSIX). 3861 */ 3862 int 3863 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3864 { 3865 struct vnode *tvp, *fvp, *tdvp; 3866 struct pathbuf *frompb, *topb; 3867 struct nameidata fromnd, tond; 3868 struct mount *fs; 3869 int error; 3870 3871 error = pathbuf_maybe_copyin(from, seg, &frompb); 3872 if (error) { 3873 return error; 3874 } 3875 error = pathbuf_maybe_copyin(to, seg, &topb); 3876 if (error) { 3877 pathbuf_destroy(frompb); 3878 return error; 3879 } 3880 3881 NDINIT(&fromnd, DELETE, LOCKPARENT | TRYEMULROOT | INRENAME, 3882 frompb); 3883 if ((error = namei(&fromnd)) != 0) { 3884 pathbuf_destroy(frompb); 3885 pathbuf_destroy(topb); 3886 return (error); 3887 } 3888 if (fromnd.ni_dvp != fromnd.ni_vp) 3889 VOP_UNLOCK(fromnd.ni_dvp); 3890 fvp = fromnd.ni_vp; 3891 3892 fs = fvp->v_mount; 3893 error = VFS_RENAMELOCK_ENTER(fs); 3894 if (error) { 3895 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3896 vrele(fromnd.ni_dvp); 3897 vrele(fvp); 3898 goto out1; 3899 } 3900 3901 /* 3902 * close, partially, yet another race - ideally we should only 3903 * go as far as getting fromnd.ni_dvp before getting the per-fs 3904 * lock, and then continue to get fromnd.ni_vp, but we can't do 3905 * that with namei as it stands. 3906 * 3907 * This still won't prevent rmdir from nuking fromnd.ni_vp 3908 * under us. The real fix is to get the locks in the right 3909 * order and do the lookups in the right places, but that's a 3910 * major rototill. 3911 * 3912 * Note: this logic (as well as this whole function) is cloned 3913 * in nfs_serv.c. Proceed accordingly. 3914 */ 3915 vrele(fvp); 3916 if ((fromnd.ni_cnd.cn_namelen == 1 && 3917 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3918 (fromnd.ni_cnd.cn_namelen == 2 && 3919 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3920 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3921 error = EINVAL; 3922 VFS_RENAMELOCK_EXIT(fs); 3923 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3924 vrele(fromnd.ni_dvp); 3925 goto out1; 3926 } 3927 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3928 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0); 3929 if (error) { 3930 VOP_UNLOCK(fromnd.ni_dvp); 3931 VFS_RENAMELOCK_EXIT(fs); 3932 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3933 vrele(fromnd.ni_dvp); 3934 goto out1; 3935 } 3936 VOP_UNLOCK(fromnd.ni_vp); 3937 if (fromnd.ni_dvp != fromnd.ni_vp) 3938 VOP_UNLOCK(fromnd.ni_dvp); 3939 fvp = fromnd.ni_vp; 3940 3941 NDINIT(&tond, RENAME, 3942 LOCKPARENT | LOCKLEAF | NOCACHE | TRYEMULROOT 3943 | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0), 3944 topb); 3945 if ((error = namei(&tond)) != 0) { 3946 VFS_RENAMELOCK_EXIT(fs); 3947 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3948 vrele(fromnd.ni_dvp); 3949 vrele(fvp); 3950 goto out1; 3951 } 3952 tdvp = tond.ni_dvp; 3953 tvp = tond.ni_vp; 3954 3955 if (tvp != NULL) { 3956 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3957 error = ENOTDIR; 3958 goto out; 3959 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3960 error = EISDIR; 3961 goto out; 3962 } 3963 } 3964 3965 if (fvp == tdvp) 3966 error = EINVAL; 3967 3968 /* 3969 * Source and destination refer to the same object. 3970 */ 3971 if (fvp == tvp) { 3972 if (retain) 3973 error = -1; 3974 else if (fromnd.ni_dvp == tdvp && 3975 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3976 !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr, 3977 fromnd.ni_cnd.cn_namelen)) 3978 error = -1; 3979 } 3980 /* 3981 * Prevent cross-mount operation. 3982 */ 3983 if (error == 0) { 3984 if (tond.ni_dvp->v_mount != fromnd.ni_dvp->v_mount) { 3985 error = EXDEV; 3986 } 3987 } 3988 #if NVERIEXEC > 0 3989 if (!error) { 3990 char *f1, *f2; 3991 size_t f1_len; 3992 size_t f2_len; 3993 3994 f1_len = fromnd.ni_cnd.cn_namelen + 1; 3995 f1 = kmem_alloc(f1_len, KM_SLEEP); 3996 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len); 3997 3998 f2_len = tond.ni_cnd.cn_namelen + 1; 3999 f2 = kmem_alloc(f2_len, KM_SLEEP); 4000 strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len); 4001 4002 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4003 4004 kmem_free(f1, f1_len); 4005 kmem_free(f2, f2_len); 4006 } 4007 #endif /* NVERIEXEC > 0 */ 4008 4009 out: 4010 if (!error) { 4011 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 4012 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 4013 VFS_RENAMELOCK_EXIT(fs); 4014 } else { 4015 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 4016 if (tdvp == tvp) 4017 vrele(tdvp); 4018 else 4019 vput(tdvp); 4020 if (tvp) 4021 vput(tvp); 4022 VFS_RENAMELOCK_EXIT(fs); 4023 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 4024 vrele(fromnd.ni_dvp); 4025 vrele(fvp); 4026 } 4027 out1: 4028 pathbuf_destroy(frompb); 4029 pathbuf_destroy(topb); 4030 return (error == -1 ? 0 : error); 4031 } 4032 4033 /* 4034 * Make a directory file. 4035 */ 4036 /* ARGSUSED */ 4037 int 4038 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4039 { 4040 /* { 4041 syscallarg(const char *) path; 4042 syscallarg(int) mode; 4043 } */ 4044 4045 return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE); 4046 } 4047 4048 int 4049 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4050 register_t *retval) 4051 { 4052 /* { 4053 syscallarg(int) fd; 4054 syscallarg(const char *) path; 4055 syscallarg(int) mode; 4056 } */ 4057 4058 return ENOSYS; 4059 } 4060 4061 4062 int 4063 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4064 { 4065 struct proc *p = curlwp->l_proc; 4066 struct vnode *vp; 4067 struct vattr vattr; 4068 int error; 4069 struct pathbuf *pb; 4070 struct nameidata nd; 4071 4072 /* XXX bollocks, should pass in a pathbuf */ 4073 error = pathbuf_maybe_copyin(path, seg, &pb); 4074 if (error) { 4075 return error; 4076 } 4077 4078 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4079 if ((error = namei(&nd)) != 0) { 4080 pathbuf_destroy(pb); 4081 return (error); 4082 } 4083 vp = nd.ni_vp; 4084 if (vp != NULL) { 4085 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4086 if (nd.ni_dvp == vp) 4087 vrele(nd.ni_dvp); 4088 else 4089 vput(nd.ni_dvp); 4090 vrele(vp); 4091 pathbuf_destroy(pb); 4092 return (EEXIST); 4093 } 4094 vattr_null(&vattr); 4095 vattr.va_type = VDIR; 4096 /* We will read cwdi->cwdi_cmask unlocked. */ 4097 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4098 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4099 if (!error) 4100 vput(nd.ni_vp); 4101 pathbuf_destroy(pb); 4102 return (error); 4103 } 4104 4105 /* 4106 * Remove a directory file. 4107 */ 4108 /* ARGSUSED */ 4109 int 4110 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4111 { 4112 /* { 4113 syscallarg(const char *) path; 4114 } */ 4115 struct vnode *vp; 4116 int error; 4117 struct pathbuf *pb; 4118 struct nameidata nd; 4119 4120 error = pathbuf_copyin(SCARG(uap, path), &pb); 4121 if (error) { 4122 return error; 4123 } 4124 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 4125 if ((error = namei(&nd)) != 0) { 4126 pathbuf_destroy(pb); 4127 return error; 4128 } 4129 vp = nd.ni_vp; 4130 if (vp->v_type != VDIR) { 4131 error = ENOTDIR; 4132 goto out; 4133 } 4134 /* 4135 * No rmdir "." please. 4136 */ 4137 if (nd.ni_dvp == vp) { 4138 error = EINVAL; 4139 goto out; 4140 } 4141 /* 4142 * The root of a mounted filesystem cannot be deleted. 4143 */ 4144 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 4145 error = EBUSY; 4146 goto out; 4147 } 4148 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 4149 pathbuf_destroy(pb); 4150 return (error); 4151 4152 out: 4153 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4154 if (nd.ni_dvp == vp) 4155 vrele(nd.ni_dvp); 4156 else 4157 vput(nd.ni_dvp); 4158 vput(vp); 4159 pathbuf_destroy(pb); 4160 return (error); 4161 } 4162 4163 /* 4164 * Read a block of directory entries in a file system independent format. 4165 */ 4166 int 4167 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4168 { 4169 /* { 4170 syscallarg(int) fd; 4171 syscallarg(char *) buf; 4172 syscallarg(size_t) count; 4173 } */ 4174 file_t *fp; 4175 int error, done; 4176 4177 /* fd_getvnode() will use the descriptor for us */ 4178 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4179 return (error); 4180 if ((fp->f_flag & FREAD) == 0) { 4181 error = EBADF; 4182 goto out; 4183 } 4184 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4185 SCARG(uap, count), &done, l, 0, 0); 4186 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4187 *retval = done; 4188 out: 4189 fd_putfile(SCARG(uap, fd)); 4190 return (error); 4191 } 4192 4193 /* 4194 * Set the mode mask for creation of filesystem nodes. 4195 */ 4196 int 4197 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4198 { 4199 /* { 4200 syscallarg(mode_t) newmask; 4201 } */ 4202 struct proc *p = l->l_proc; 4203 struct cwdinfo *cwdi; 4204 4205 /* 4206 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4207 * important is that we serialize changes to the mask. The 4208 * rw_exit() will issue a write memory barrier on our behalf, 4209 * and force the changes out to other CPUs (as it must use an 4210 * atomic operation, draining the local CPU's store buffers). 4211 */ 4212 cwdi = p->p_cwdi; 4213 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4214 *retval = cwdi->cwdi_cmask; 4215 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4216 rw_exit(&cwdi->cwdi_lock); 4217 4218 return (0); 4219 } 4220 4221 int 4222 dorevoke(struct vnode *vp, kauth_cred_t cred) 4223 { 4224 struct vattr vattr; 4225 int error, fs_decision; 4226 4227 vn_lock(vp, LK_SHARED | LK_RETRY); 4228 error = VOP_GETATTR(vp, &vattr, cred); 4229 VOP_UNLOCK(vp); 4230 if (error != 0) 4231 return error; 4232 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4233 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4234 fs_decision); 4235 if (!error) 4236 VOP_REVOKE(vp, REVOKEALL); 4237 return (error); 4238 } 4239 4240 /* 4241 * Void all references to file by ripping underlying filesystem 4242 * away from vnode. 4243 */ 4244 /* ARGSUSED */ 4245 int 4246 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4247 { 4248 /* { 4249 syscallarg(const char *) path; 4250 } */ 4251 struct vnode *vp; 4252 int error; 4253 4254 error = namei_simple_user(SCARG(uap, path), 4255 NSM_FOLLOW_TRYEMULROOT, &vp); 4256 if (error != 0) 4257 return (error); 4258 error = dorevoke(vp, l->l_cred); 4259 vrele(vp); 4260 return (error); 4261 } 4262
Indexes created Fri Oct 03 02:09:57 GMT 2025