vfs_syscalls.c revision 1.474
1 /* $NetBSD: vfs_syscalls.c,v 1.474 2014/01/25 02:28:31 christos 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 1.474 2014/01/25 02:28:31 christos 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_sys_openat(lwp_t *, int, const char *, int, int, int *); 123 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 124 enum uio_seg); 125 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 126 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 127 enum uio_seg); 128 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 129 enum uio_seg, int); 130 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 131 size_t, register_t *); 132 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 133 134 static int fd_nameiat(struct lwp *, int, struct nameidata *); 135 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 136 namei_simple_flags_t, struct vnode **); 137 138 139 /* 140 * This table is used to maintain compatibility with 4.3BSD 141 * and NetBSD 0.9 mount syscalls - and possibly other systems. 142 * Note, the order is important! 143 * 144 * Do not modify this table. It should only contain filesystems 145 * supported by NetBSD 0.9 and 4.3BSD. 146 */ 147 const char * const mountcompatnames[] = { 148 NULL, /* 0 = MOUNT_NONE */ 149 MOUNT_FFS, /* 1 = MOUNT_UFS */ 150 MOUNT_NFS, /* 2 */ 151 MOUNT_MFS, /* 3 */ 152 MOUNT_MSDOS, /* 4 */ 153 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 154 MOUNT_FDESC, /* 6 */ 155 MOUNT_KERNFS, /* 7 */ 156 NULL, /* 8 = MOUNT_DEVFS */ 157 MOUNT_AFS, /* 9 */ 158 }; 159 160 const int nmountcompatnames = __arraycount(mountcompatnames); 161 162 static int 163 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 164 { 165 file_t *dfp; 166 int error; 167 168 if (fdat != AT_FDCWD) { 169 if ((error = fd_getvnode(fdat, &dfp)) != 0) 170 goto out; 171 172 NDAT(ndp, dfp->f_data); 173 } 174 175 error = namei(ndp); 176 177 if (fdat != AT_FDCWD) 178 fd_putfile(fdat); 179 out: 180 return error; 181 } 182 183 static int 184 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 185 namei_simple_flags_t sflags, struct vnode **vp_ret) 186 { 187 file_t *dfp; 188 struct vnode *dvp; 189 int error; 190 191 if (fdat != AT_FDCWD) { 192 if ((error = fd_getvnode(fdat, &dfp)) != 0) 193 goto out; 194 195 dvp = dfp->f_data; 196 } else { 197 dvp = NULL; 198 } 199 200 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 201 202 if (fdat != AT_FDCWD) 203 fd_putfile(fdat); 204 out: 205 return error; 206 } 207 208 static int 209 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 210 { 211 int error; 212 213 fp->f_flag = flags & FMASK; 214 fp->f_type = DTYPE_VNODE; 215 fp->f_ops = &vnops; 216 fp->f_data = vp; 217 218 if (flags & (O_EXLOCK | O_SHLOCK)) { 219 struct flock lf; 220 int type; 221 222 lf.l_whence = SEEK_SET; 223 lf.l_start = 0; 224 lf.l_len = 0; 225 if (flags & O_EXLOCK) 226 lf.l_type = F_WRLCK; 227 else 228 lf.l_type = F_RDLCK; 229 type = F_FLOCK; 230 if ((flags & FNONBLOCK) == 0) 231 type |= F_WAIT; 232 VOP_UNLOCK(vp); 233 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 234 if (error) { 235 (void) vn_close(vp, fp->f_flag, fp->f_cred); 236 fd_abort(l->l_proc, fp, indx); 237 return error; 238 } 239 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 240 atomic_or_uint(&fp->f_flag, FHASLOCK); 241 } 242 if (flags & O_CLOEXEC) 243 fd_set_exclose(l, indx, true); 244 return 0; 245 } 246 247 static int 248 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 249 void *data, size_t *data_len) 250 { 251 struct mount *mp; 252 int error = 0, saved_flags; 253 254 mp = vp->v_mount; 255 saved_flags = mp->mnt_flag; 256 257 /* We can operate only on VV_ROOT nodes. */ 258 if ((vp->v_vflag & VV_ROOT) == 0) { 259 error = EINVAL; 260 goto out; 261 } 262 263 /* 264 * We only allow the filesystem to be reloaded if it 265 * is currently mounted read-only. Additionally, we 266 * prevent read-write to read-only downgrades. 267 */ 268 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 269 (mp->mnt_flag & MNT_RDONLY) == 0 && 270 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 271 error = EOPNOTSUPP; /* Needs translation */ 272 goto out; 273 } 274 275 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 276 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 277 if (error) 278 goto out; 279 280 if (vfs_busy(mp, NULL)) { 281 error = EPERM; 282 goto out; 283 } 284 285 mutex_enter(&mp->mnt_updating); 286 287 mp->mnt_flag &= ~MNT_OP_FLAGS; 288 mp->mnt_flag |= flags & MNT_OP_FLAGS; 289 290 /* 291 * Set the mount level flags. 292 */ 293 if (flags & MNT_RDONLY) 294 mp->mnt_flag |= MNT_RDONLY; 295 else if (mp->mnt_flag & MNT_RDONLY) 296 mp->mnt_iflag |= IMNT_WANTRDWR; 297 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 298 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 299 error = VFS_MOUNT(mp, path, data, data_len); 300 301 if (error && data != NULL) { 302 int error2; 303 304 /* 305 * Update failed; let's try and see if it was an 306 * export request. For compat with 3.0 and earlier. 307 */ 308 error2 = vfs_hooks_reexport(mp, path, data); 309 310 /* 311 * Only update error code if the export request was 312 * understood but some problem occurred while 313 * processing it. 314 */ 315 if (error2 != EJUSTRETURN) 316 error = error2; 317 } 318 319 if (mp->mnt_iflag & IMNT_WANTRDWR) 320 mp->mnt_flag &= ~MNT_RDONLY; 321 if (error) 322 mp->mnt_flag = saved_flags; 323 mp->mnt_flag &= ~MNT_OP_FLAGS; 324 mp->mnt_iflag &= ~IMNT_WANTRDWR; 325 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 326 if (mp->mnt_syncer == NULL) 327 error = vfs_allocate_syncvnode(mp); 328 } else { 329 if (mp->mnt_syncer != NULL) 330 vfs_deallocate_syncvnode(mp); 331 } 332 mutex_exit(&mp->mnt_updating); 333 vfs_unbusy(mp, false, NULL); 334 335 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 336 (flags & MNT_EXTATTR)) { 337 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 338 NULL, 0, NULL) != 0) { 339 printf("%s: failed to start extattr, error = %d", 340 mp->mnt_stat.f_mntonname, error); 341 mp->mnt_flag &= ~MNT_EXTATTR; 342 } 343 } 344 345 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 346 !(flags & MNT_EXTATTR)) { 347 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 348 NULL, 0, NULL) != 0) { 349 printf("%s: failed to stop extattr, error = %d", 350 mp->mnt_stat.f_mntonname, error); 351 mp->mnt_flag |= MNT_RDONLY; 352 } 353 } 354 out: 355 return (error); 356 } 357 358 static int 359 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 360 { 361 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 362 int error; 363 364 /* Copy file-system type from userspace. */ 365 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 366 if (error) { 367 /* 368 * Historically, filesystem types were identified by numbers. 369 * If we get an integer for the filesystem type instead of a 370 * string, we check to see if it matches one of the historic 371 * filesystem types. 372 */ 373 u_long fsindex = (u_long)fstype; 374 if (fsindex >= nmountcompatnames || 375 mountcompatnames[fsindex] == NULL) 376 return ENODEV; 377 strlcpy(fstypename, mountcompatnames[fsindex], 378 sizeof(fstypename)); 379 } 380 381 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 382 if (strcmp(fstypename, "ufs") == 0) 383 fstypename[0] = 'f'; 384 385 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 386 return 0; 387 388 /* If we can autoload a vfs module, try again */ 389 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 390 391 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 392 return 0; 393 394 return ENODEV; 395 } 396 397 static int 398 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 399 void *data, size_t *data_len) 400 { 401 struct mount *mp; 402 int error; 403 404 /* If MNT_GETARGS is specified, it should be the only flag. */ 405 if (flags & ~MNT_GETARGS) 406 return EINVAL; 407 408 mp = vp->v_mount; 409 410 /* XXX: probably some notion of "can see" here if we want isolation. */ 411 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 412 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 413 if (error) 414 return error; 415 416 if ((vp->v_vflag & VV_ROOT) == 0) 417 return EINVAL; 418 419 if (vfs_busy(mp, NULL)) 420 return EPERM; 421 422 mutex_enter(&mp->mnt_updating); 423 mp->mnt_flag &= ~MNT_OP_FLAGS; 424 mp->mnt_flag |= MNT_GETARGS; 425 error = VFS_MOUNT(mp, path, data, data_len); 426 mp->mnt_flag &= ~MNT_OP_FLAGS; 427 mutex_exit(&mp->mnt_updating); 428 429 vfs_unbusy(mp, false, NULL); 430 return (error); 431 } 432 433 int 434 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 435 { 436 /* { 437 syscallarg(const char *) type; 438 syscallarg(const char *) path; 439 syscallarg(int) flags; 440 syscallarg(void *) data; 441 syscallarg(size_t) data_len; 442 } */ 443 444 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 445 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 446 SCARG(uap, data_len), retval); 447 } 448 449 int 450 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 451 const char *path, int flags, void *data, enum uio_seg data_seg, 452 size_t data_len, register_t *retval) 453 { 454 struct vnode *vp; 455 void *data_buf = data; 456 bool vfsopsrele = false; 457 int error; 458 459 /* XXX: The calling convention of this routine is totally bizarre */ 460 if (vfsops) 461 vfsopsrele = true; 462 463 /* 464 * Get vnode to be covered 465 */ 466 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 467 if (error != 0) { 468 vp = NULL; 469 goto done; 470 } 471 472 if (vfsops == NULL) { 473 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 474 vfsops = vp->v_mount->mnt_op; 475 } else { 476 /* 'type' is userspace */ 477 error = mount_get_vfsops(type, &vfsops); 478 if (error != 0) 479 goto done; 480 vfsopsrele = true; 481 } 482 } 483 484 if (data != NULL && data_seg == UIO_USERSPACE) { 485 if (data_len == 0) { 486 /* No length supplied, use default for filesystem */ 487 data_len = vfsops->vfs_min_mount_data; 488 if (data_len > VFS_MAX_MOUNT_DATA) { 489 error = EINVAL; 490 goto done; 491 } 492 /* 493 * Hopefully a longer buffer won't make copyin() fail. 494 * For compatibility with 3.0 and earlier. 495 */ 496 if (flags & MNT_UPDATE 497 && data_len < sizeof (struct mnt_export_args30)) 498 data_len = sizeof (struct mnt_export_args30); 499 } 500 data_buf = kmem_alloc(data_len, KM_SLEEP); 501 502 /* NFS needs the buffer even for mnt_getargs .... */ 503 error = copyin(data, data_buf, data_len); 504 if (error != 0) 505 goto done; 506 } 507 508 if (flags & MNT_GETARGS) { 509 if (data_len == 0) { 510 error = EINVAL; 511 goto done; 512 } 513 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 514 if (error != 0) 515 goto done; 516 if (data_seg == UIO_USERSPACE) 517 error = copyout(data_buf, data, data_len); 518 *retval = data_len; 519 } else if (flags & MNT_UPDATE) { 520 error = mount_update(l, vp, path, flags, data_buf, &data_len); 521 } else { 522 /* Locking is handled internally in mount_domount(). */ 523 KASSERT(vfsopsrele == true); 524 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 525 &data_len); 526 vfsopsrele = false; 527 } 528 529 done: 530 if (vfsopsrele) 531 vfs_delref(vfsops); 532 if (vp != NULL) { 533 vrele(vp); 534 } 535 if (data_buf != data) 536 kmem_free(data_buf, data_len); 537 return (error); 538 } 539 540 /* 541 * Unmount a file system. 542 * 543 * Note: unmount takes a path to the vnode mounted on as argument, 544 * not special file (as before). 545 */ 546 /* ARGSUSED */ 547 int 548 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 549 { 550 /* { 551 syscallarg(const char *) path; 552 syscallarg(int) flags; 553 } */ 554 struct vnode *vp; 555 struct mount *mp; 556 int error; 557 struct pathbuf *pb; 558 struct nameidata nd; 559 560 error = pathbuf_copyin(SCARG(uap, path), &pb); 561 if (error) { 562 return error; 563 } 564 565 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 566 if ((error = namei(&nd)) != 0) { 567 pathbuf_destroy(pb); 568 return error; 569 } 570 vp = nd.ni_vp; 571 pathbuf_destroy(pb); 572 573 mp = vp->v_mount; 574 atomic_inc_uint(&mp->mnt_refcnt); 575 VOP_UNLOCK(vp); 576 577 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 578 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 579 if (error) { 580 vrele(vp); 581 vfs_destroy(mp); 582 return (error); 583 } 584 585 /* 586 * Don't allow unmounting the root file system. 587 */ 588 if (mp->mnt_flag & MNT_ROOTFS) { 589 vrele(vp); 590 vfs_destroy(mp); 591 return (EINVAL); 592 } 593 594 /* 595 * Must be the root of the filesystem 596 */ 597 if ((vp->v_vflag & VV_ROOT) == 0) { 598 vrele(vp); 599 vfs_destroy(mp); 600 return (EINVAL); 601 } 602 603 vrele(vp); 604 error = dounmount(mp, SCARG(uap, flags), l); 605 vfs_destroy(mp); 606 return error; 607 } 608 609 /* 610 * Sync each mounted filesystem. 611 */ 612 #ifdef DEBUG 613 int syncprt = 0; 614 struct ctldebug debug0 = { "syncprt", &syncprt }; 615 #endif 616 617 void 618 do_sys_sync(struct lwp *l) 619 { 620 struct mount *mp, *nmp; 621 int asyncflag; 622 623 mutex_enter(&mountlist_lock); 624 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 625 if (vfs_busy(mp, &nmp)) { 626 continue; 627 } 628 mutex_enter(&mp->mnt_updating); 629 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 630 asyncflag = mp->mnt_flag & MNT_ASYNC; 631 mp->mnt_flag &= ~MNT_ASYNC; 632 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 633 if (asyncflag) 634 mp->mnt_flag |= MNT_ASYNC; 635 } 636 mutex_exit(&mp->mnt_updating); 637 vfs_unbusy(mp, false, &nmp); 638 } 639 mutex_exit(&mountlist_lock); 640 #ifdef DEBUG 641 if (syncprt) 642 vfs_bufstats(); 643 #endif /* DEBUG */ 644 } 645 646 /* ARGSUSED */ 647 int 648 sys_sync(struct lwp *l, const void *v, register_t *retval) 649 { 650 do_sys_sync(l); 651 return (0); 652 } 653 654 655 /* 656 * Access or change filesystem quotas. 657 * 658 * (this is really 14 different calls bundled into one) 659 */ 660 661 static int 662 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 663 { 664 struct quotastat info_k; 665 int error; 666 667 /* ensure any padding bytes are cleared */ 668 memset(&info_k, 0, sizeof(info_k)); 669 670 error = vfs_quotactl_stat(mp, &info_k); 671 if (error) { 672 return error; 673 } 674 675 return copyout(&info_k, info_u, sizeof(info_k)); 676 } 677 678 static int 679 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 680 struct quotaidtypestat *info_u) 681 { 682 struct quotaidtypestat info_k; 683 int error; 684 685 /* ensure any padding bytes are cleared */ 686 memset(&info_k, 0, sizeof(info_k)); 687 688 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 689 if (error) { 690 return error; 691 } 692 693 return copyout(&info_k, info_u, sizeof(info_k)); 694 } 695 696 static int 697 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 698 struct quotaobjtypestat *info_u) 699 { 700 struct quotaobjtypestat info_k; 701 int error; 702 703 /* ensure any padding bytes are cleared */ 704 memset(&info_k, 0, sizeof(info_k)); 705 706 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 707 if (error) { 708 return error; 709 } 710 711 return copyout(&info_k, info_u, sizeof(info_k)); 712 } 713 714 static int 715 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 716 struct quotaval *val_u) 717 { 718 struct quotakey key_k; 719 struct quotaval val_k; 720 int error; 721 722 /* ensure any padding bytes are cleared */ 723 memset(&val_k, 0, sizeof(val_k)); 724 725 error = copyin(key_u, &key_k, sizeof(key_k)); 726 if (error) { 727 return error; 728 } 729 730 error = vfs_quotactl_get(mp, &key_k, &val_k); 731 if (error) { 732 return error; 733 } 734 735 return copyout(&val_k, val_u, sizeof(val_k)); 736 } 737 738 static int 739 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 740 const struct quotaval *val_u) 741 { 742 struct quotakey key_k; 743 struct quotaval val_k; 744 int error; 745 746 error = copyin(key_u, &key_k, sizeof(key_k)); 747 if (error) { 748 return error; 749 } 750 751 error = copyin(val_u, &val_k, sizeof(val_k)); 752 if (error) { 753 return error; 754 } 755 756 return vfs_quotactl_put(mp, &key_k, &val_k); 757 } 758 759 static int 760 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u) 761 { 762 struct quotakey key_k; 763 int error; 764 765 error = copyin(key_u, &key_k, sizeof(key_k)); 766 if (error) { 767 return error; 768 } 769 770 return vfs_quotactl_delete(mp, &key_k); 771 } 772 773 static int 774 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 775 { 776 struct quotakcursor cursor_k; 777 int error; 778 779 /* ensure any padding bytes are cleared */ 780 memset(&cursor_k, 0, sizeof(cursor_k)); 781 782 error = vfs_quotactl_cursoropen(mp, &cursor_k); 783 if (error) { 784 return error; 785 } 786 787 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 788 } 789 790 static int 791 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 792 { 793 struct quotakcursor cursor_k; 794 int error; 795 796 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 797 if (error) { 798 return error; 799 } 800 801 return vfs_quotactl_cursorclose(mp, &cursor_k); 802 } 803 804 static int 805 do_sys_quotactl_cursorskipidtype(struct mount *mp, 806 struct quotakcursor *cursor_u, int idtype) 807 { 808 struct quotakcursor cursor_k; 809 int error; 810 811 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 812 if (error) { 813 return error; 814 } 815 816 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 817 if (error) { 818 return error; 819 } 820 821 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 822 } 823 824 static int 825 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 826 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 827 unsigned *ret_u) 828 { 829 #define CGET_STACK_MAX 8 830 struct quotakcursor cursor_k; 831 struct quotakey stackkeys[CGET_STACK_MAX]; 832 struct quotaval stackvals[CGET_STACK_MAX]; 833 struct quotakey *keys_k; 834 struct quotaval *vals_k; 835 unsigned ret_k; 836 int error; 837 838 if (maxnum > 128) { 839 maxnum = 128; 840 } 841 842 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 843 if (error) { 844 return error; 845 } 846 847 if (maxnum <= CGET_STACK_MAX) { 848 keys_k = stackkeys; 849 vals_k = stackvals; 850 /* ensure any padding bytes are cleared */ 851 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 852 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 853 } else { 854 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 855 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 856 } 857 858 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 859 &ret_k); 860 if (error) { 861 goto fail; 862 } 863 864 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 865 if (error) { 866 goto fail; 867 } 868 869 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 870 if (error) { 871 goto fail; 872 } 873 874 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 875 if (error) { 876 goto fail; 877 } 878 879 /* do last to maximize the chance of being able to recover a failure */ 880 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 881 882 fail: 883 if (keys_k != stackkeys) { 884 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 885 } 886 if (vals_k != stackvals) { 887 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 888 } 889 return error; 890 } 891 892 static int 893 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 894 int *ret_u) 895 { 896 struct quotakcursor cursor_k; 897 int ret_k; 898 int error; 899 900 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 901 if (error) { 902 return error; 903 } 904 905 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 906 if (error) { 907 return error; 908 } 909 910 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 911 if (error) { 912 return error; 913 } 914 915 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 916 } 917 918 static int 919 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 920 { 921 struct quotakcursor cursor_k; 922 int error; 923 924 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 925 if (error) { 926 return error; 927 } 928 929 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 930 if (error) { 931 return error; 932 } 933 934 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 935 } 936 937 static int 938 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 939 { 940 char *path_k; 941 int error; 942 943 /* XXX this should probably be a struct pathbuf */ 944 path_k = PNBUF_GET(); 945 error = copyin(path_u, path_k, PATH_MAX); 946 if (error) { 947 PNBUF_PUT(path_k); 948 return error; 949 } 950 951 error = vfs_quotactl_quotaon(mp, idtype, path_k); 952 953 PNBUF_PUT(path_k); 954 return error; 955 } 956 957 static int 958 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 959 { 960 return vfs_quotactl_quotaoff(mp, idtype); 961 } 962 963 int 964 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 965 { 966 struct mount *mp; 967 struct vnode *vp; 968 int error; 969 970 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 971 if (error != 0) 972 return (error); 973 mp = vp->v_mount; 974 975 switch (args->qc_op) { 976 case QUOTACTL_STAT: 977 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 978 break; 979 case QUOTACTL_IDTYPESTAT: 980 error = do_sys_quotactl_idtypestat(mp, 981 args->u.idtypestat.qc_idtype, 982 args->u.idtypestat.qc_info); 983 break; 984 case QUOTACTL_OBJTYPESTAT: 985 error = do_sys_quotactl_objtypestat(mp, 986 args->u.objtypestat.qc_objtype, 987 args->u.objtypestat.qc_info); 988 break; 989 case QUOTACTL_GET: 990 error = do_sys_quotactl_get(mp, 991 args->u.get.qc_key, 992 args->u.get.qc_val); 993 break; 994 case QUOTACTL_PUT: 995 error = do_sys_quotactl_put(mp, 996 args->u.put.qc_key, 997 args->u.put.qc_val); 998 break; 999 case QUOTACTL_DELETE: 1000 error = do_sys_quotactl_delete(mp, args->u.delete.qc_key); 1001 break; 1002 case QUOTACTL_CURSOROPEN: 1003 error = do_sys_quotactl_cursoropen(mp, 1004 args->u.cursoropen.qc_cursor); 1005 break; 1006 case QUOTACTL_CURSORCLOSE: 1007 error = do_sys_quotactl_cursorclose(mp, 1008 args->u.cursorclose.qc_cursor); 1009 break; 1010 case QUOTACTL_CURSORSKIPIDTYPE: 1011 error = do_sys_quotactl_cursorskipidtype(mp, 1012 args->u.cursorskipidtype.qc_cursor, 1013 args->u.cursorskipidtype.qc_idtype); 1014 break; 1015 case QUOTACTL_CURSORGET: 1016 error = do_sys_quotactl_cursorget(mp, 1017 args->u.cursorget.qc_cursor, 1018 args->u.cursorget.qc_keys, 1019 args->u.cursorget.qc_vals, 1020 args->u.cursorget.qc_maxnum, 1021 args->u.cursorget.qc_ret); 1022 break; 1023 case QUOTACTL_CURSORATEND: 1024 error = do_sys_quotactl_cursoratend(mp, 1025 args->u.cursoratend.qc_cursor, 1026 args->u.cursoratend.qc_ret); 1027 break; 1028 case QUOTACTL_CURSORREWIND: 1029 error = do_sys_quotactl_cursorrewind(mp, 1030 args->u.cursorrewind.qc_cursor); 1031 break; 1032 case QUOTACTL_QUOTAON: 1033 error = do_sys_quotactl_quotaon(mp, 1034 args->u.quotaon.qc_idtype, 1035 args->u.quotaon.qc_quotafile); 1036 break; 1037 case QUOTACTL_QUOTAOFF: 1038 error = do_sys_quotactl_quotaoff(mp, 1039 args->u.quotaoff.qc_idtype); 1040 break; 1041 default: 1042 error = EINVAL; 1043 break; 1044 } 1045 1046 vrele(vp); 1047 return error; 1048 } 1049 1050 /* ARGSUSED */ 1051 int 1052 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1053 register_t *retval) 1054 { 1055 /* { 1056 syscallarg(const char *) path; 1057 syscallarg(struct quotactl_args *) args; 1058 } */ 1059 struct quotactl_args args; 1060 int error; 1061 1062 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1063 if (error) { 1064 return error; 1065 } 1066 1067 return do_sys_quotactl(SCARG(uap, path), &args); 1068 } 1069 1070 int 1071 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1072 int root) 1073 { 1074 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1075 int error = 0; 1076 1077 /* 1078 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1079 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1080 * overrides MNT_NOWAIT. 1081 */ 1082 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1083 (flags != MNT_WAIT && flags != 0)) { 1084 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1085 goto done; 1086 } 1087 1088 /* Get the filesystem stats now */ 1089 memset(sp, 0, sizeof(*sp)); 1090 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1091 return error; 1092 } 1093 1094 if (cwdi->cwdi_rdir == NULL) 1095 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1096 done: 1097 if (cwdi->cwdi_rdir != NULL) { 1098 size_t len; 1099 char *bp; 1100 char c; 1101 char *path = PNBUF_GET(); 1102 1103 bp = path + MAXPATHLEN; 1104 *--bp = '\0'; 1105 rw_enter(&cwdi->cwdi_lock, RW_READER); 1106 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1107 MAXPATHLEN / 2, 0, l); 1108 rw_exit(&cwdi->cwdi_lock); 1109 if (error) { 1110 PNBUF_PUT(path); 1111 return error; 1112 } 1113 len = strlen(bp); 1114 if (len != 1) { 1115 /* 1116 * for mount points that are below our root, we can see 1117 * them, so we fix up the pathname and return them. The 1118 * rest we cannot see, so we don't allow viewing the 1119 * data. 1120 */ 1121 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1122 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1123 (void)strlcpy(sp->f_mntonname, 1124 c == '\0' ? "/" : &sp->f_mntonname[len], 1125 sizeof(sp->f_mntonname)); 1126 } else { 1127 if (root) 1128 (void)strlcpy(sp->f_mntonname, "/", 1129 sizeof(sp->f_mntonname)); 1130 else 1131 error = EPERM; 1132 } 1133 } 1134 PNBUF_PUT(path); 1135 } 1136 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1137 return error; 1138 } 1139 1140 /* 1141 * Get filesystem statistics by path. 1142 */ 1143 int 1144 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1145 { 1146 struct mount *mp; 1147 int error; 1148 struct vnode *vp; 1149 1150 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1151 if (error != 0) 1152 return error; 1153 mp = vp->v_mount; 1154 error = dostatvfs(mp, sb, l, flags, 1); 1155 vrele(vp); 1156 return error; 1157 } 1158 1159 /* ARGSUSED */ 1160 int 1161 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1162 { 1163 /* { 1164 syscallarg(const char *) path; 1165 syscallarg(struct statvfs *) buf; 1166 syscallarg(int) flags; 1167 } */ 1168 struct statvfs *sb; 1169 int error; 1170 1171 sb = STATVFSBUF_GET(); 1172 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1173 if (error == 0) 1174 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1175 STATVFSBUF_PUT(sb); 1176 return error; 1177 } 1178 1179 /* 1180 * Get filesystem statistics by fd. 1181 */ 1182 int 1183 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1184 { 1185 file_t *fp; 1186 struct mount *mp; 1187 int error; 1188 1189 /* fd_getvnode() will use the descriptor for us */ 1190 if ((error = fd_getvnode(fd, &fp)) != 0) 1191 return (error); 1192 mp = ((struct vnode *)fp->f_data)->v_mount; 1193 error = dostatvfs(mp, sb, curlwp, flags, 1); 1194 fd_putfile(fd); 1195 return error; 1196 } 1197 1198 /* ARGSUSED */ 1199 int 1200 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1201 { 1202 /* { 1203 syscallarg(int) fd; 1204 syscallarg(struct statvfs *) buf; 1205 syscallarg(int) flags; 1206 } */ 1207 struct statvfs *sb; 1208 int error; 1209 1210 sb = STATVFSBUF_GET(); 1211 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1212 if (error == 0) 1213 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1214 STATVFSBUF_PUT(sb); 1215 return error; 1216 } 1217 1218 1219 /* 1220 * Get statistics on all filesystems. 1221 */ 1222 int 1223 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1224 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1225 register_t *retval) 1226 { 1227 int root = 0; 1228 struct proc *p = l->l_proc; 1229 struct mount *mp, *nmp; 1230 struct statvfs *sb; 1231 size_t count, maxcount; 1232 int error = 0; 1233 1234 sb = STATVFSBUF_GET(); 1235 maxcount = bufsize / entry_sz; 1236 mutex_enter(&mountlist_lock); 1237 count = 0; 1238 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 1239 if (vfs_busy(mp, &nmp)) { 1240 continue; 1241 } 1242 if (sfsp && count < maxcount) { 1243 error = dostatvfs(mp, sb, l, flags, 0); 1244 if (error) { 1245 vfs_unbusy(mp, false, &nmp); 1246 error = 0; 1247 continue; 1248 } 1249 error = copyfn(sb, sfsp, entry_sz); 1250 if (error) { 1251 vfs_unbusy(mp, false, NULL); 1252 goto out; 1253 } 1254 sfsp = (char *)sfsp + entry_sz; 1255 root |= strcmp(sb->f_mntonname, "/") == 0; 1256 } 1257 count++; 1258 vfs_unbusy(mp, false, &nmp); 1259 } 1260 mutex_exit(&mountlist_lock); 1261 1262 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1263 /* 1264 * fake a root entry 1265 */ 1266 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1267 sb, l, flags, 1); 1268 if (error != 0) 1269 goto out; 1270 if (sfsp) { 1271 error = copyfn(sb, sfsp, entry_sz); 1272 if (error != 0) 1273 goto out; 1274 } 1275 count++; 1276 } 1277 if (sfsp && count > maxcount) 1278 *retval = maxcount; 1279 else 1280 *retval = count; 1281 out: 1282 STATVFSBUF_PUT(sb); 1283 return error; 1284 } 1285 1286 int 1287 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1288 { 1289 /* { 1290 syscallarg(struct statvfs *) buf; 1291 syscallarg(size_t) bufsize; 1292 syscallarg(int) flags; 1293 } */ 1294 1295 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1296 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1297 } 1298 1299 /* 1300 * Change current working directory to a given file descriptor. 1301 */ 1302 /* ARGSUSED */ 1303 int 1304 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1305 { 1306 /* { 1307 syscallarg(int) fd; 1308 } */ 1309 struct proc *p = l->l_proc; 1310 struct cwdinfo *cwdi; 1311 struct vnode *vp, *tdp; 1312 struct mount *mp; 1313 file_t *fp; 1314 int error, fd; 1315 1316 /* fd_getvnode() will use the descriptor for us */ 1317 fd = SCARG(uap, fd); 1318 if ((error = fd_getvnode(fd, &fp)) != 0) 1319 return (error); 1320 vp = fp->f_data; 1321 1322 vref(vp); 1323 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1324 if (vp->v_type != VDIR) 1325 error = ENOTDIR; 1326 else 1327 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1328 if (error) { 1329 vput(vp); 1330 goto out; 1331 } 1332 while ((mp = vp->v_mountedhere) != NULL) { 1333 error = vfs_busy(mp, NULL); 1334 vput(vp); 1335 if (error != 0) 1336 goto out; 1337 error = VFS_ROOT(mp, &tdp); 1338 vfs_unbusy(mp, false, NULL); 1339 if (error) 1340 goto out; 1341 vp = tdp; 1342 } 1343 VOP_UNLOCK(vp); 1344 1345 /* 1346 * Disallow changing to a directory not under the process's 1347 * current root directory (if there is one). 1348 */ 1349 cwdi = p->p_cwdi; 1350 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1351 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1352 vrele(vp); 1353 error = EPERM; /* operation not permitted */ 1354 } else { 1355 vrele(cwdi->cwdi_cdir); 1356 cwdi->cwdi_cdir = vp; 1357 } 1358 rw_exit(&cwdi->cwdi_lock); 1359 1360 out: 1361 fd_putfile(fd); 1362 return (error); 1363 } 1364 1365 /* 1366 * Change this process's notion of the root directory to a given file 1367 * descriptor. 1368 */ 1369 int 1370 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1371 { 1372 struct proc *p = l->l_proc; 1373 struct vnode *vp; 1374 file_t *fp; 1375 int error, fd = SCARG(uap, fd); 1376 1377 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1378 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1379 return error; 1380 /* fd_getvnode() will use the descriptor for us */ 1381 if ((error = fd_getvnode(fd, &fp)) != 0) 1382 return error; 1383 vp = fp->f_data; 1384 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1385 if (vp->v_type != VDIR) 1386 error = ENOTDIR; 1387 else 1388 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1389 VOP_UNLOCK(vp); 1390 if (error) 1391 goto out; 1392 vref(vp); 1393 1394 change_root(p->p_cwdi, vp, l); 1395 1396 out: 1397 fd_putfile(fd); 1398 return (error); 1399 } 1400 1401 /* 1402 * Change current working directory (``.''). 1403 */ 1404 /* ARGSUSED */ 1405 int 1406 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1407 { 1408 /* { 1409 syscallarg(const char *) path; 1410 } */ 1411 struct proc *p = l->l_proc; 1412 struct cwdinfo *cwdi; 1413 int error; 1414 struct vnode *vp; 1415 1416 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1417 &vp, l)) != 0) 1418 return (error); 1419 cwdi = p->p_cwdi; 1420 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1421 vrele(cwdi->cwdi_cdir); 1422 cwdi->cwdi_cdir = vp; 1423 rw_exit(&cwdi->cwdi_lock); 1424 return (0); 1425 } 1426 1427 /* 1428 * Change notion of root (``/'') directory. 1429 */ 1430 /* ARGSUSED */ 1431 int 1432 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1433 { 1434 /* { 1435 syscallarg(const char *) path; 1436 } */ 1437 struct proc *p = l->l_proc; 1438 int error; 1439 struct vnode *vp; 1440 1441 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1442 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1443 return (error); 1444 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1445 &vp, l)) != 0) 1446 return (error); 1447 1448 change_root(p->p_cwdi, vp, l); 1449 1450 return (0); 1451 } 1452 1453 /* 1454 * Common routine for chroot and fchroot. 1455 * NB: callers need to properly authorize the change root operation. 1456 */ 1457 void 1458 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1459 { 1460 struct proc *p = l->l_proc; 1461 kauth_cred_t ncred; 1462 1463 ncred = kauth_cred_alloc(); 1464 1465 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1466 if (cwdi->cwdi_rdir != NULL) 1467 vrele(cwdi->cwdi_rdir); 1468 cwdi->cwdi_rdir = vp; 1469 1470 /* 1471 * Prevent escaping from chroot by putting the root under 1472 * the working directory. Silently chdir to / if we aren't 1473 * already there. 1474 */ 1475 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1476 /* 1477 * XXX would be more failsafe to change directory to a 1478 * deadfs node here instead 1479 */ 1480 vrele(cwdi->cwdi_cdir); 1481 vref(vp); 1482 cwdi->cwdi_cdir = vp; 1483 } 1484 rw_exit(&cwdi->cwdi_lock); 1485 1486 /* Get a write lock on the process credential. */ 1487 proc_crmod_enter(); 1488 1489 kauth_cred_clone(p->p_cred, ncred); 1490 kauth_proc_chroot(ncred, p->p_cwdi); 1491 1492 /* Broadcast our credentials to the process and other LWPs. */ 1493 proc_crmod_leave(ncred, p->p_cred, true); 1494 } 1495 1496 /* 1497 * Common routine for chroot and chdir. 1498 * XXX "where" should be enum uio_seg 1499 */ 1500 int 1501 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1502 { 1503 struct pathbuf *pb; 1504 struct nameidata nd; 1505 int error; 1506 1507 error = pathbuf_maybe_copyin(path, where, &pb); 1508 if (error) { 1509 return error; 1510 } 1511 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1512 if ((error = namei(&nd)) != 0) { 1513 pathbuf_destroy(pb); 1514 return error; 1515 } 1516 *vpp = nd.ni_vp; 1517 pathbuf_destroy(pb); 1518 1519 if ((*vpp)->v_type != VDIR) 1520 error = ENOTDIR; 1521 else 1522 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1523 1524 if (error) 1525 vput(*vpp); 1526 else 1527 VOP_UNLOCK(*vpp); 1528 return (error); 1529 } 1530 1531 /* 1532 * Internals of sys_open - path has already been converted into a pathbuf 1533 * (so we can easily reuse this function from other parts of the kernel, 1534 * like posix_spawn post-processing). 1535 */ 1536 int 1537 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1538 int open_mode, int *fd) 1539 { 1540 struct proc *p = l->l_proc; 1541 struct cwdinfo *cwdi = p->p_cwdi; 1542 file_t *fp; 1543 struct vnode *vp; 1544 int flags, cmode; 1545 int indx, error; 1546 struct nameidata nd; 1547 1548 if (open_flags & O_SEARCH) { 1549 open_flags &= ~(int)O_SEARCH; 1550 } 1551 1552 flags = FFLAGS(open_flags); 1553 if ((flags & (FREAD | FWRITE)) == 0) 1554 return EINVAL; 1555 1556 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1557 return error; 1558 } 1559 1560 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1561 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1562 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1563 if (dvp != NULL) 1564 NDAT(&nd, dvp); 1565 1566 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1567 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1568 fd_abort(p, fp, indx); 1569 if ((error == EDUPFD || error == EMOVEFD) && 1570 l->l_dupfd >= 0 && /* XXX from fdopen */ 1571 (error = 1572 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1573 *fd = indx; 1574 return 0; 1575 } 1576 if (error == ERESTART) 1577 error = EINTR; 1578 return error; 1579 } 1580 1581 l->l_dupfd = 0; 1582 vp = nd.ni_vp; 1583 1584 if ((error = open_setfp(l, fp, vp, indx, flags))) 1585 return error; 1586 1587 VOP_UNLOCK(vp); 1588 *fd = indx; 1589 fd_affix(p, fp, indx); 1590 return 0; 1591 } 1592 1593 int 1594 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1595 { 1596 struct pathbuf *pb; 1597 int error, oflags; 1598 1599 oflags = FFLAGS(open_flags); 1600 if ((oflags & (FREAD | FWRITE)) == 0) 1601 return EINVAL; 1602 1603 pb = pathbuf_create(path); 1604 if (pb == NULL) 1605 return ENOMEM; 1606 1607 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1608 pathbuf_destroy(pb); 1609 1610 return error; 1611 } 1612 1613 /* 1614 * Check permissions, allocate an open file structure, 1615 * and call the device open routine if any. 1616 */ 1617 static int 1618 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1619 int mode, int *fd) 1620 { 1621 file_t *dfp = NULL; 1622 struct vnode *dvp = NULL; 1623 struct pathbuf *pb; 1624 int error; 1625 1626 error = pathbuf_copyin(path, &pb); 1627 if (error) 1628 return error; 1629 1630 if (fdat != AT_FDCWD) { 1631 /* fd_getvnode() will use the descriptor for us */ 1632 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1633 goto out; 1634 1635 dvp = dfp->f_data; 1636 } 1637 1638 error = do_open(l, dvp, pb, flags, mode, fd); 1639 1640 if (dfp != NULL) 1641 fd_putfile(fdat); 1642 out: 1643 pathbuf_destroy(pb); 1644 return error; 1645 } 1646 1647 int 1648 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1649 { 1650 /* { 1651 syscallarg(const char *) path; 1652 syscallarg(int) flags; 1653 syscallarg(int) mode; 1654 } */ 1655 int error; 1656 int fd; 1657 1658 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1659 SCARG(uap, flags), SCARG(uap, mode), &fd); 1660 1661 if (error == 0) 1662 *retval = fd; 1663 1664 return error; 1665 } 1666 1667 int 1668 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1669 { 1670 /* { 1671 syscallarg(int) fd; 1672 syscallarg(const char *) path; 1673 syscallarg(int) oflags; 1674 syscallarg(int) mode; 1675 } */ 1676 int error; 1677 int fd; 1678 1679 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1680 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1681 1682 if (error == 0) 1683 *retval = fd; 1684 1685 return error; 1686 } 1687 1688 static void 1689 vfs__fhfree(fhandle_t *fhp) 1690 { 1691 size_t fhsize; 1692 1693 if (fhp == NULL) { 1694 return; 1695 } 1696 fhsize = FHANDLE_SIZE(fhp); 1697 kmem_free(fhp, fhsize); 1698 } 1699 1700 /* 1701 * vfs_composefh: compose a filehandle. 1702 */ 1703 1704 int 1705 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1706 { 1707 struct mount *mp; 1708 struct fid *fidp; 1709 int error; 1710 size_t needfhsize; 1711 size_t fidsize; 1712 1713 mp = vp->v_mount; 1714 fidp = NULL; 1715 if (*fh_size < FHANDLE_SIZE_MIN) { 1716 fidsize = 0; 1717 } else { 1718 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1719 if (fhp != NULL) { 1720 memset(fhp, 0, *fh_size); 1721 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1722 fidp = &fhp->fh_fid; 1723 } 1724 } 1725 error = VFS_VPTOFH(vp, fidp, &fidsize); 1726 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1727 if (error == 0 && *fh_size < needfhsize) { 1728 error = E2BIG; 1729 } 1730 *fh_size = needfhsize; 1731 return error; 1732 } 1733 1734 int 1735 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1736 { 1737 struct mount *mp; 1738 fhandle_t *fhp; 1739 size_t fhsize; 1740 size_t fidsize; 1741 int error; 1742 1743 *fhpp = NULL; 1744 mp = vp->v_mount; 1745 fidsize = 0; 1746 error = VFS_VPTOFH(vp, NULL, &fidsize); 1747 KASSERT(error != 0); 1748 if (error != E2BIG) { 1749 goto out; 1750 } 1751 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1752 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1753 if (fhp == NULL) { 1754 error = ENOMEM; 1755 goto out; 1756 } 1757 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1758 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1759 if (error == 0) { 1760 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1761 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1762 *fhpp = fhp; 1763 } else { 1764 kmem_free(fhp, fhsize); 1765 } 1766 out: 1767 return error; 1768 } 1769 1770 void 1771 vfs_composefh_free(fhandle_t *fhp) 1772 { 1773 1774 vfs__fhfree(fhp); 1775 } 1776 1777 /* 1778 * vfs_fhtovp: lookup a vnode by a filehandle. 1779 */ 1780 1781 int 1782 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1783 { 1784 struct mount *mp; 1785 int error; 1786 1787 *vpp = NULL; 1788 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1789 if (mp == NULL) { 1790 error = ESTALE; 1791 goto out; 1792 } 1793 if (mp->mnt_op->vfs_fhtovp == NULL) { 1794 error = EOPNOTSUPP; 1795 goto out; 1796 } 1797 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1798 out: 1799 return error; 1800 } 1801 1802 /* 1803 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1804 * the needed size. 1805 */ 1806 1807 int 1808 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1809 { 1810 fhandle_t *fhp; 1811 int error; 1812 1813 *fhpp = NULL; 1814 if (fhsize > FHANDLE_SIZE_MAX) { 1815 return EINVAL; 1816 } 1817 if (fhsize < FHANDLE_SIZE_MIN) { 1818 return EINVAL; 1819 } 1820 again: 1821 fhp = kmem_alloc(fhsize, KM_SLEEP); 1822 if (fhp == NULL) { 1823 return ENOMEM; 1824 } 1825 error = copyin(ufhp, fhp, fhsize); 1826 if (error == 0) { 1827 /* XXX this check shouldn't be here */ 1828 if (FHANDLE_SIZE(fhp) == fhsize) { 1829 *fhpp = fhp; 1830 return 0; 1831 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1832 /* 1833 * a kludge for nfsv2 padded handles. 1834 */ 1835 size_t sz; 1836 1837 sz = FHANDLE_SIZE(fhp); 1838 kmem_free(fhp, fhsize); 1839 fhsize = sz; 1840 goto again; 1841 } else { 1842 /* 1843 * userland told us wrong size. 1844 */ 1845 error = EINVAL; 1846 } 1847 } 1848 kmem_free(fhp, fhsize); 1849 return error; 1850 } 1851 1852 void 1853 vfs_copyinfh_free(fhandle_t *fhp) 1854 { 1855 1856 vfs__fhfree(fhp); 1857 } 1858 1859 /* 1860 * Get file handle system call 1861 */ 1862 int 1863 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1864 { 1865 /* { 1866 syscallarg(char *) fname; 1867 syscallarg(fhandle_t *) fhp; 1868 syscallarg(size_t *) fh_size; 1869 } */ 1870 struct vnode *vp; 1871 fhandle_t *fh; 1872 int error; 1873 struct pathbuf *pb; 1874 struct nameidata nd; 1875 size_t sz; 1876 size_t usz; 1877 1878 /* 1879 * Must be super user 1880 */ 1881 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1882 0, NULL, NULL, NULL); 1883 if (error) 1884 return (error); 1885 1886 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1887 if (error) { 1888 return error; 1889 } 1890 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1891 error = namei(&nd); 1892 if (error) { 1893 pathbuf_destroy(pb); 1894 return error; 1895 } 1896 vp = nd.ni_vp; 1897 pathbuf_destroy(pb); 1898 1899 error = vfs_composefh_alloc(vp, &fh); 1900 vput(vp); 1901 if (error != 0) { 1902 goto out; 1903 } 1904 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1905 if (error != 0) { 1906 goto out; 1907 } 1908 sz = FHANDLE_SIZE(fh); 1909 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1910 if (error != 0) { 1911 goto out; 1912 } 1913 if (usz >= sz) { 1914 error = copyout(fh, SCARG(uap, fhp), sz); 1915 } else { 1916 error = E2BIG; 1917 } 1918 out: 1919 vfs_composefh_free(fh); 1920 return (error); 1921 } 1922 1923 /* 1924 * Open a file given a file handle. 1925 * 1926 * Check permissions, allocate an open file structure, 1927 * and call the device open routine if any. 1928 */ 1929 1930 int 1931 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1932 register_t *retval) 1933 { 1934 file_t *fp; 1935 struct vnode *vp = NULL; 1936 kauth_cred_t cred = l->l_cred; 1937 file_t *nfp; 1938 int indx, error = 0; 1939 struct vattr va; 1940 fhandle_t *fh; 1941 int flags; 1942 proc_t *p; 1943 1944 p = curproc; 1945 1946 /* 1947 * Must be super user 1948 */ 1949 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1950 0, NULL, NULL, NULL))) 1951 return (error); 1952 1953 if (oflags & O_SEARCH) { 1954 oflags &= ~(int)O_SEARCH; 1955 } 1956 1957 flags = FFLAGS(oflags); 1958 if ((flags & (FREAD | FWRITE)) == 0) 1959 return (EINVAL); 1960 if ((flags & O_CREAT)) 1961 return (EINVAL); 1962 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1963 return (error); 1964 fp = nfp; 1965 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1966 if (error != 0) { 1967 goto bad; 1968 } 1969 error = vfs_fhtovp(fh, &vp); 1970 if (error != 0) { 1971 goto bad; 1972 } 1973 1974 /* Now do an effective vn_open */ 1975 1976 if (vp->v_type == VSOCK) { 1977 error = EOPNOTSUPP; 1978 goto bad; 1979 } 1980 error = vn_openchk(vp, cred, flags); 1981 if (error != 0) 1982 goto bad; 1983 if (flags & O_TRUNC) { 1984 VOP_UNLOCK(vp); /* XXX */ 1985 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1986 vattr_null(&va); 1987 va.va_size = 0; 1988 error = VOP_SETATTR(vp, &va, cred); 1989 if (error) 1990 goto bad; 1991 } 1992 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1993 goto bad; 1994 if (flags & FWRITE) { 1995 mutex_enter(vp->v_interlock); 1996 vp->v_writecount++; 1997 mutex_exit(vp->v_interlock); 1998 } 1999 2000 /* done with modified vn_open, now finish what sys_open does. */ 2001 if ((error = open_setfp(l, fp, vp, indx, flags))) 2002 return error; 2003 2004 VOP_UNLOCK(vp); 2005 *retval = indx; 2006 fd_affix(p, fp, indx); 2007 vfs_copyinfh_free(fh); 2008 return (0); 2009 2010 bad: 2011 fd_abort(p, fp, indx); 2012 if (vp != NULL) 2013 vput(vp); 2014 vfs_copyinfh_free(fh); 2015 return (error); 2016 } 2017 2018 int 2019 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2020 { 2021 /* { 2022 syscallarg(const void *) fhp; 2023 syscallarg(size_t) fh_size; 2024 syscallarg(int) flags; 2025 } */ 2026 2027 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2028 SCARG(uap, flags), retval); 2029 } 2030 2031 int 2032 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2033 { 2034 int error; 2035 fhandle_t *fh; 2036 struct vnode *vp; 2037 2038 /* 2039 * Must be super user 2040 */ 2041 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2042 0, NULL, NULL, NULL))) 2043 return (error); 2044 2045 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2046 if (error != 0) 2047 return error; 2048 2049 error = vfs_fhtovp(fh, &vp); 2050 vfs_copyinfh_free(fh); 2051 if (error != 0) 2052 return error; 2053 2054 error = vn_stat(vp, sb); 2055 vput(vp); 2056 return error; 2057 } 2058 2059 2060 /* ARGSUSED */ 2061 int 2062 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2063 { 2064 /* { 2065 syscallarg(const void *) fhp; 2066 syscallarg(size_t) fh_size; 2067 syscallarg(struct stat *) sb; 2068 } */ 2069 struct stat sb; 2070 int error; 2071 2072 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2073 if (error) 2074 return error; 2075 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2076 } 2077 2078 int 2079 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2080 int flags) 2081 { 2082 fhandle_t *fh; 2083 struct mount *mp; 2084 struct vnode *vp; 2085 int error; 2086 2087 /* 2088 * Must be super user 2089 */ 2090 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2091 0, NULL, NULL, NULL))) 2092 return error; 2093 2094 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2095 if (error != 0) 2096 return error; 2097 2098 error = vfs_fhtovp(fh, &vp); 2099 vfs_copyinfh_free(fh); 2100 if (error != 0) 2101 return error; 2102 2103 mp = vp->v_mount; 2104 error = dostatvfs(mp, sb, l, flags, 1); 2105 vput(vp); 2106 return error; 2107 } 2108 2109 /* ARGSUSED */ 2110 int 2111 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2112 { 2113 /* { 2114 syscallarg(const void *) fhp; 2115 syscallarg(size_t) fh_size; 2116 syscallarg(struct statvfs *) buf; 2117 syscallarg(int) flags; 2118 } */ 2119 struct statvfs *sb = STATVFSBUF_GET(); 2120 int error; 2121 2122 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2123 SCARG(uap, flags)); 2124 if (error == 0) 2125 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2126 STATVFSBUF_PUT(sb); 2127 return error; 2128 } 2129 2130 /* 2131 * Create a special file. 2132 */ 2133 /* ARGSUSED */ 2134 int 2135 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2136 register_t *retval) 2137 { 2138 /* { 2139 syscallarg(const char *) path; 2140 syscallarg(mode_t) mode; 2141 syscallarg(dev_t) dev; 2142 } */ 2143 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2144 SCARG(uap, dev), retval, UIO_USERSPACE); 2145 } 2146 2147 int 2148 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2149 register_t *retval) 2150 { 2151 /* { 2152 syscallarg(int) fd; 2153 syscallarg(const char *) path; 2154 syscallarg(mode_t) mode; 2155 syscallarg(int) pad; 2156 syscallarg(dev_t) dev; 2157 } */ 2158 2159 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2160 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2161 } 2162 2163 int 2164 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2165 register_t *retval, enum uio_seg seg) 2166 { 2167 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2168 } 2169 2170 int 2171 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2172 dev_t dev, register_t *retval, enum uio_seg seg) 2173 { 2174 struct proc *p = l->l_proc; 2175 struct vnode *vp; 2176 struct vattr vattr; 2177 int error, optype; 2178 struct pathbuf *pb; 2179 struct nameidata nd; 2180 const char *pathstring; 2181 2182 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2183 0, NULL, NULL, NULL)) != 0) 2184 return (error); 2185 2186 optype = VOP_MKNOD_DESCOFFSET; 2187 2188 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2189 if (error) { 2190 return error; 2191 } 2192 pathstring = pathbuf_stringcopy_get(pb); 2193 if (pathstring == NULL) { 2194 pathbuf_destroy(pb); 2195 return ENOMEM; 2196 } 2197 2198 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2199 2200 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2201 goto out; 2202 vp = nd.ni_vp; 2203 2204 if (vp != NULL) 2205 error = EEXIST; 2206 else { 2207 vattr_null(&vattr); 2208 /* We will read cwdi->cwdi_cmask unlocked. */ 2209 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2210 vattr.va_rdev = dev; 2211 2212 switch (mode & S_IFMT) { 2213 case S_IFMT: /* used by badsect to flag bad sectors */ 2214 vattr.va_type = VBAD; 2215 break; 2216 case S_IFCHR: 2217 vattr.va_type = VCHR; 2218 break; 2219 case S_IFBLK: 2220 vattr.va_type = VBLK; 2221 break; 2222 case S_IFWHT: 2223 optype = VOP_WHITEOUT_DESCOFFSET; 2224 break; 2225 case S_IFREG: 2226 #if NVERIEXEC > 0 2227 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2228 O_CREAT); 2229 #endif /* NVERIEXEC > 0 */ 2230 vattr.va_type = VREG; 2231 vattr.va_rdev = VNOVAL; 2232 optype = VOP_CREATE_DESCOFFSET; 2233 break; 2234 default: 2235 error = EINVAL; 2236 break; 2237 } 2238 } 2239 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2240 && vattr.va_rdev == VNOVAL) 2241 error = EINVAL; 2242 if (!error) { 2243 switch (optype) { 2244 case VOP_WHITEOUT_DESCOFFSET: 2245 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2246 if (error) 2247 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2248 vput(nd.ni_dvp); 2249 break; 2250 2251 case VOP_MKNOD_DESCOFFSET: 2252 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2253 &nd.ni_cnd, &vattr); 2254 if (error == 0) 2255 vrele(nd.ni_vp); 2256 vput(nd.ni_dvp); 2257 break; 2258 2259 case VOP_CREATE_DESCOFFSET: 2260 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2261 &nd.ni_cnd, &vattr); 2262 if (error == 0) 2263 vrele(nd.ni_vp); 2264 vput(nd.ni_dvp); 2265 break; 2266 } 2267 } else { 2268 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2269 if (nd.ni_dvp == vp) 2270 vrele(nd.ni_dvp); 2271 else 2272 vput(nd.ni_dvp); 2273 if (vp) 2274 vrele(vp); 2275 } 2276 out: 2277 pathbuf_stringcopy_put(pb, pathstring); 2278 pathbuf_destroy(pb); 2279 return (error); 2280 } 2281 2282 /* 2283 * Create a named pipe. 2284 */ 2285 /* ARGSUSED */ 2286 int 2287 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2288 { 2289 /* { 2290 syscallarg(const char *) path; 2291 syscallarg(int) mode; 2292 } */ 2293 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2294 } 2295 2296 int 2297 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2298 register_t *retval) 2299 { 2300 /* { 2301 syscallarg(int) fd; 2302 syscallarg(const char *) path; 2303 syscallarg(int) mode; 2304 } */ 2305 2306 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2307 SCARG(uap, mode)); 2308 } 2309 2310 static int 2311 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2312 { 2313 struct proc *p = l->l_proc; 2314 struct vattr vattr; 2315 int error; 2316 struct pathbuf *pb; 2317 struct nameidata nd; 2318 2319 error = pathbuf_copyin(path, &pb); 2320 if (error) { 2321 return error; 2322 } 2323 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2324 2325 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2326 pathbuf_destroy(pb); 2327 return error; 2328 } 2329 if (nd.ni_vp != NULL) { 2330 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2331 if (nd.ni_dvp == nd.ni_vp) 2332 vrele(nd.ni_dvp); 2333 else 2334 vput(nd.ni_dvp); 2335 vrele(nd.ni_vp); 2336 pathbuf_destroy(pb); 2337 return (EEXIST); 2338 } 2339 vattr_null(&vattr); 2340 vattr.va_type = VFIFO; 2341 /* We will read cwdi->cwdi_cmask unlocked. */ 2342 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2343 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2344 if (error == 0) 2345 vrele(nd.ni_vp); 2346 vput(nd.ni_dvp); 2347 pathbuf_destroy(pb); 2348 return (error); 2349 } 2350 2351 /* 2352 * Make a hard file link. 2353 */ 2354 /* ARGSUSED */ 2355 int 2356 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2357 const char *link, int follow, register_t *retval) 2358 { 2359 struct vnode *vp; 2360 struct pathbuf *linkpb; 2361 struct nameidata nd; 2362 namei_simple_flags_t ns_flags; 2363 int error; 2364 2365 if (follow & AT_SYMLINK_FOLLOW) 2366 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2367 else 2368 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2369 2370 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2371 if (error != 0) 2372 return (error); 2373 error = pathbuf_copyin(link, &linkpb); 2374 if (error) { 2375 goto out1; 2376 } 2377 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2378 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2379 goto out2; 2380 if (nd.ni_vp) { 2381 error = EEXIST; 2382 goto abortop; 2383 } 2384 /* Prevent hard links on directories. */ 2385 if (vp->v_type == VDIR) { 2386 error = EPERM; 2387 goto abortop; 2388 } 2389 /* Prevent cross-mount operation. */ 2390 if (nd.ni_dvp->v_mount != vp->v_mount) { 2391 error = EXDEV; 2392 goto abortop; 2393 } 2394 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2395 out2: 2396 pathbuf_destroy(linkpb); 2397 out1: 2398 vrele(vp); 2399 return (error); 2400 abortop: 2401 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2402 if (nd.ni_dvp == nd.ni_vp) 2403 vrele(nd.ni_dvp); 2404 else 2405 vput(nd.ni_dvp); 2406 if (nd.ni_vp != NULL) 2407 vrele(nd.ni_vp); 2408 goto out2; 2409 } 2410 2411 int 2412 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2413 { 2414 /* { 2415 syscallarg(const char *) path; 2416 syscallarg(const char *) link; 2417 } */ 2418 const char *path = SCARG(uap, path); 2419 const char *link = SCARG(uap, link); 2420 2421 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2422 AT_SYMLINK_FOLLOW, retval); 2423 } 2424 2425 int 2426 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2427 register_t *retval) 2428 { 2429 /* { 2430 syscallarg(int) fd1; 2431 syscallarg(const char *) name1; 2432 syscallarg(int) fd2; 2433 syscallarg(const char *) name2; 2434 syscallarg(int) flags; 2435 } */ 2436 int fd1 = SCARG(uap, fd1); 2437 const char *name1 = SCARG(uap, name1); 2438 int fd2 = SCARG(uap, fd2); 2439 const char *name2 = SCARG(uap, name2); 2440 int follow; 2441 2442 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2443 2444 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2445 } 2446 2447 2448 int 2449 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2450 { 2451 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2452 } 2453 2454 static int 2455 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2456 const char *link, enum uio_seg seg) 2457 { 2458 struct proc *p = curproc; 2459 struct vattr vattr; 2460 char *path; 2461 int error; 2462 struct pathbuf *linkpb; 2463 struct nameidata nd; 2464 2465 KASSERT(l != NULL || fdat == AT_FDCWD); 2466 2467 path = PNBUF_GET(); 2468 if (seg == UIO_USERSPACE) { 2469 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2470 goto out1; 2471 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2472 goto out1; 2473 } else { 2474 KASSERT(strlen(patharg) < MAXPATHLEN); 2475 strcpy(path, patharg); 2476 linkpb = pathbuf_create(link); 2477 if (linkpb == NULL) { 2478 error = ENOMEM; 2479 goto out1; 2480 } 2481 } 2482 ktrkuser("symlink-target", path, strlen(path)); 2483 2484 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2485 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2486 goto out2; 2487 if (nd.ni_vp) { 2488 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2489 if (nd.ni_dvp == nd.ni_vp) 2490 vrele(nd.ni_dvp); 2491 else 2492 vput(nd.ni_dvp); 2493 vrele(nd.ni_vp); 2494 error = EEXIST; 2495 goto out2; 2496 } 2497 vattr_null(&vattr); 2498 vattr.va_type = VLNK; 2499 /* We will read cwdi->cwdi_cmask unlocked. */ 2500 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2501 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2502 if (error == 0) 2503 vrele(nd.ni_vp); 2504 vput(nd.ni_dvp); 2505 out2: 2506 pathbuf_destroy(linkpb); 2507 out1: 2508 PNBUF_PUT(path); 2509 return (error); 2510 } 2511 2512 /* 2513 * Make a symbolic link. 2514 */ 2515 /* ARGSUSED */ 2516 int 2517 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2518 { 2519 /* { 2520 syscallarg(const char *) path; 2521 syscallarg(const char *) link; 2522 } */ 2523 2524 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2525 UIO_USERSPACE); 2526 } 2527 2528 int 2529 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2530 register_t *retval) 2531 { 2532 /* { 2533 syscallarg(const char *) path1; 2534 syscallarg(int) fd; 2535 syscallarg(const char *) path2; 2536 } */ 2537 2538 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2539 SCARG(uap, path2), UIO_USERSPACE); 2540 } 2541 2542 /* 2543 * Delete a whiteout from the filesystem. 2544 */ 2545 /* ARGSUSED */ 2546 int 2547 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2548 { 2549 /* { 2550 syscallarg(const char *) path; 2551 } */ 2552 int error; 2553 struct pathbuf *pb; 2554 struct nameidata nd; 2555 2556 error = pathbuf_copyin(SCARG(uap, path), &pb); 2557 if (error) { 2558 return error; 2559 } 2560 2561 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2562 error = namei(&nd); 2563 if (error) { 2564 pathbuf_destroy(pb); 2565 return (error); 2566 } 2567 2568 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2569 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2570 if (nd.ni_dvp == nd.ni_vp) 2571 vrele(nd.ni_dvp); 2572 else 2573 vput(nd.ni_dvp); 2574 if (nd.ni_vp) 2575 vrele(nd.ni_vp); 2576 pathbuf_destroy(pb); 2577 return (EEXIST); 2578 } 2579 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2580 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2581 vput(nd.ni_dvp); 2582 pathbuf_destroy(pb); 2583 return (error); 2584 } 2585 2586 /* 2587 * Delete a name from the filesystem. 2588 */ 2589 /* ARGSUSED */ 2590 int 2591 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2592 { 2593 /* { 2594 syscallarg(const char *) path; 2595 } */ 2596 2597 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2598 } 2599 2600 int 2601 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2602 register_t *retval) 2603 { 2604 /* { 2605 syscallarg(int) fd; 2606 syscallarg(const char *) path; 2607 syscallarg(int) flag; 2608 } */ 2609 2610 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2611 SCARG(uap, flag), UIO_USERSPACE); 2612 } 2613 2614 int 2615 do_sys_unlink(const char *arg, enum uio_seg seg) 2616 { 2617 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2618 } 2619 2620 static int 2621 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2622 enum uio_seg seg) 2623 { 2624 struct vnode *vp; 2625 int error; 2626 struct pathbuf *pb; 2627 struct nameidata nd; 2628 const char *pathstring; 2629 2630 KASSERT(l != NULL || fdat == AT_FDCWD); 2631 2632 error = pathbuf_maybe_copyin(arg, seg, &pb); 2633 if (error) { 2634 return error; 2635 } 2636 pathstring = pathbuf_stringcopy_get(pb); 2637 if (pathstring == NULL) { 2638 pathbuf_destroy(pb); 2639 return ENOMEM; 2640 } 2641 2642 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2643 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2644 goto out; 2645 vp = nd.ni_vp; 2646 2647 /* 2648 * The root of a mounted filesystem cannot be deleted. 2649 */ 2650 if ((vp->v_vflag & VV_ROOT) != 0) { 2651 error = EBUSY; 2652 goto abort; 2653 } 2654 2655 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2656 error = EBUSY; 2657 goto abort; 2658 } 2659 2660 /* 2661 * No rmdir "." please. 2662 */ 2663 if (nd.ni_dvp == vp) { 2664 error = EINVAL; 2665 goto abort; 2666 } 2667 2668 /* 2669 * AT_REMOVEDIR is required to remove a directory 2670 */ 2671 if (vp->v_type == VDIR) { 2672 if (!(flags & AT_REMOVEDIR)) { 2673 error = EPERM; 2674 goto abort; 2675 } else { 2676 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2677 goto out; 2678 } 2679 } 2680 2681 /* 2682 * Starting here we only deal with non directories. 2683 */ 2684 if (flags & AT_REMOVEDIR) { 2685 error = ENOTDIR; 2686 goto abort; 2687 } 2688 2689 2690 #if NVERIEXEC > 0 2691 /* Handle remove requests for veriexec entries. */ 2692 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2693 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2694 if (nd.ni_dvp == vp) 2695 vrele(nd.ni_dvp); 2696 else 2697 vput(nd.ni_dvp); 2698 vput(vp); 2699 goto out; 2700 } 2701 #endif /* NVERIEXEC > 0 */ 2702 2703 #ifdef FILEASSOC 2704 (void)fileassoc_file_delete(vp); 2705 #endif /* FILEASSOC */ 2706 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2707 goto out; 2708 2709 abort: 2710 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2711 if (nd.ni_dvp == vp) 2712 vrele(nd.ni_dvp); 2713 else 2714 vput(nd.ni_dvp); 2715 vput(vp); 2716 2717 out: 2718 pathbuf_stringcopy_put(pb, pathstring); 2719 pathbuf_destroy(pb); 2720 return (error); 2721 } 2722 2723 /* 2724 * Reposition read/write file offset. 2725 */ 2726 int 2727 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2728 { 2729 /* { 2730 syscallarg(int) fd; 2731 syscallarg(int) pad; 2732 syscallarg(off_t) offset; 2733 syscallarg(int) whence; 2734 } */ 2735 kauth_cred_t cred = l->l_cred; 2736 file_t *fp; 2737 struct vnode *vp; 2738 struct vattr vattr; 2739 off_t newoff; 2740 int error, fd; 2741 2742 fd = SCARG(uap, fd); 2743 2744 if ((fp = fd_getfile(fd)) == NULL) 2745 return (EBADF); 2746 2747 vp = fp->f_data; 2748 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2749 error = ESPIPE; 2750 goto out; 2751 } 2752 2753 switch (SCARG(uap, whence)) { 2754 case SEEK_CUR: 2755 newoff = fp->f_offset + SCARG(uap, offset); 2756 break; 2757 case SEEK_END: 2758 vn_lock(vp, LK_SHARED | LK_RETRY); 2759 error = VOP_GETATTR(vp, &vattr, cred); 2760 VOP_UNLOCK(vp); 2761 if (error) { 2762 goto out; 2763 } 2764 newoff = SCARG(uap, offset) + vattr.va_size; 2765 break; 2766 case SEEK_SET: 2767 newoff = SCARG(uap, offset); 2768 break; 2769 default: 2770 error = EINVAL; 2771 goto out; 2772 } 2773 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2774 *(off_t *)retval = fp->f_offset = newoff; 2775 } 2776 out: 2777 fd_putfile(fd); 2778 return (error); 2779 } 2780 2781 /* 2782 * Positional read system call. 2783 */ 2784 int 2785 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2786 { 2787 /* { 2788 syscallarg(int) fd; 2789 syscallarg(void *) buf; 2790 syscallarg(size_t) nbyte; 2791 syscallarg(off_t) offset; 2792 } */ 2793 file_t *fp; 2794 struct vnode *vp; 2795 off_t offset; 2796 int error, fd = SCARG(uap, fd); 2797 2798 if ((fp = fd_getfile(fd)) == NULL) 2799 return (EBADF); 2800 2801 if ((fp->f_flag & FREAD) == 0) { 2802 fd_putfile(fd); 2803 return (EBADF); 2804 } 2805 2806 vp = fp->f_data; 2807 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2808 error = ESPIPE; 2809 goto out; 2810 } 2811 2812 offset = SCARG(uap, offset); 2813 2814 /* 2815 * XXX This works because no file systems actually 2816 * XXX take any action on the seek operation. 2817 */ 2818 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2819 goto out; 2820 2821 /* dofileread() will unuse the descriptor for us */ 2822 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2823 &offset, 0, retval)); 2824 2825 out: 2826 fd_putfile(fd); 2827 return (error); 2828 } 2829 2830 /* 2831 * Positional scatter read system call. 2832 */ 2833 int 2834 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2835 { 2836 /* { 2837 syscallarg(int) fd; 2838 syscallarg(const struct iovec *) iovp; 2839 syscallarg(int) iovcnt; 2840 syscallarg(off_t) offset; 2841 } */ 2842 off_t offset = SCARG(uap, offset); 2843 2844 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2845 SCARG(uap, iovcnt), &offset, 0, retval); 2846 } 2847 2848 /* 2849 * Positional write system call. 2850 */ 2851 int 2852 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2853 { 2854 /* { 2855 syscallarg(int) fd; 2856 syscallarg(const void *) buf; 2857 syscallarg(size_t) nbyte; 2858 syscallarg(off_t) offset; 2859 } */ 2860 file_t *fp; 2861 struct vnode *vp; 2862 off_t offset; 2863 int error, fd = SCARG(uap, fd); 2864 2865 if ((fp = fd_getfile(fd)) == NULL) 2866 return (EBADF); 2867 2868 if ((fp->f_flag & FWRITE) == 0) { 2869 fd_putfile(fd); 2870 return (EBADF); 2871 } 2872 2873 vp = fp->f_data; 2874 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2875 error = ESPIPE; 2876 goto out; 2877 } 2878 2879 offset = SCARG(uap, offset); 2880 2881 /* 2882 * XXX This works because no file systems actually 2883 * XXX take any action on the seek operation. 2884 */ 2885 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2886 goto out; 2887 2888 /* dofilewrite() will unuse the descriptor for us */ 2889 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2890 &offset, 0, retval)); 2891 2892 out: 2893 fd_putfile(fd); 2894 return (error); 2895 } 2896 2897 /* 2898 * Positional gather write system call. 2899 */ 2900 int 2901 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2902 { 2903 /* { 2904 syscallarg(int) fd; 2905 syscallarg(const struct iovec *) iovp; 2906 syscallarg(int) iovcnt; 2907 syscallarg(off_t) offset; 2908 } */ 2909 off_t offset = SCARG(uap, offset); 2910 2911 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2912 SCARG(uap, iovcnt), &offset, 0, retval); 2913 } 2914 2915 /* 2916 * Check access permissions. 2917 */ 2918 int 2919 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2920 { 2921 /* { 2922 syscallarg(const char *) path; 2923 syscallarg(int) flags; 2924 } */ 2925 2926 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2927 SCARG(uap, flags), 0); 2928 } 2929 2930 int 2931 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2932 int mode, int flags) 2933 { 2934 kauth_cred_t cred; 2935 struct vnode *vp; 2936 int error, nd_flag, vmode; 2937 struct pathbuf *pb; 2938 struct nameidata nd; 2939 2940 CTASSERT(F_OK == 0); 2941 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2942 /* nonsense mode */ 2943 return EINVAL; 2944 } 2945 2946 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2947 if (flags & AT_SYMLINK_NOFOLLOW) 2948 nd_flag &= ~FOLLOW; 2949 2950 error = pathbuf_copyin(path, &pb); 2951 if (error) 2952 return error; 2953 2954 NDINIT(&nd, LOOKUP, nd_flag, pb); 2955 2956 /* Override default credentials */ 2957 cred = kauth_cred_dup(l->l_cred); 2958 if (!(flags & AT_EACCESS)) { 2959 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2960 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2961 } 2962 nd.ni_cnd.cn_cred = cred; 2963 2964 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2965 pathbuf_destroy(pb); 2966 goto out; 2967 } 2968 vp = nd.ni_vp; 2969 pathbuf_destroy(pb); 2970 2971 /* Flags == 0 means only check for existence. */ 2972 if (mode) { 2973 vmode = 0; 2974 if (mode & R_OK) 2975 vmode |= VREAD; 2976 if (mode & W_OK) 2977 vmode |= VWRITE; 2978 if (mode & X_OK) 2979 vmode |= VEXEC; 2980 2981 error = VOP_ACCESS(vp, vmode, cred); 2982 if (!error && (vmode & VWRITE)) 2983 error = vn_writechk(vp); 2984 } 2985 vput(vp); 2986 out: 2987 kauth_cred_free(cred); 2988 return (error); 2989 } 2990 2991 int 2992 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 2993 register_t *retval) 2994 { 2995 /* { 2996 syscallarg(int) fd; 2997 syscallarg(const char *) path; 2998 syscallarg(int) amode; 2999 syscallarg(int) flag; 3000 } */ 3001 3002 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3003 SCARG(uap, amode), SCARG(uap, flag)); 3004 } 3005 3006 /* 3007 * Common code for all sys_stat functions, including compat versions. 3008 */ 3009 int 3010 do_sys_stat(const char *userpath, unsigned int nd_flag, 3011 struct stat *sb) 3012 { 3013 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3014 } 3015 3016 int 3017 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3018 unsigned int nd_flag, struct stat *sb) 3019 { 3020 int error; 3021 struct pathbuf *pb; 3022 struct nameidata nd; 3023 3024 KASSERT(l != NULL || fdat == AT_FDCWD); 3025 3026 error = pathbuf_copyin(userpath, &pb); 3027 if (error) { 3028 return error; 3029 } 3030 3031 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3032 3033 error = fd_nameiat(l, fdat, &nd); 3034 if (error != 0) { 3035 pathbuf_destroy(pb); 3036 return error; 3037 } 3038 error = vn_stat(nd.ni_vp, sb); 3039 vput(nd.ni_vp); 3040 pathbuf_destroy(pb); 3041 return error; 3042 } 3043 3044 /* 3045 * Get file status; this version follows links. 3046 */ 3047 /* ARGSUSED */ 3048 int 3049 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3050 { 3051 /* { 3052 syscallarg(const char *) path; 3053 syscallarg(struct stat *) ub; 3054 } */ 3055 struct stat sb; 3056 int error; 3057 3058 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3059 if (error) 3060 return error; 3061 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3062 } 3063 3064 /* 3065 * Get file status; this version does not follow links. 3066 */ 3067 /* ARGSUSED */ 3068 int 3069 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3070 { 3071 /* { 3072 syscallarg(const char *) path; 3073 syscallarg(struct stat *) ub; 3074 } */ 3075 struct stat sb; 3076 int error; 3077 3078 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3079 if (error) 3080 return error; 3081 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3082 } 3083 3084 int 3085 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3086 register_t *retval) 3087 { 3088 /* { 3089 syscallarg(int) fd; 3090 syscallarg(const char *) path; 3091 syscallarg(struct stat *) buf; 3092 syscallarg(int) flag; 3093 } */ 3094 unsigned int nd_flag; 3095 struct stat sb; 3096 int error; 3097 3098 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3099 nd_flag = NOFOLLOW; 3100 else 3101 nd_flag = FOLLOW; 3102 3103 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3104 &sb); 3105 if (error) 3106 return error; 3107 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3108 } 3109 3110 /* 3111 * Get configurable pathname variables. 3112 */ 3113 /* ARGSUSED */ 3114 int 3115 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3116 { 3117 /* { 3118 syscallarg(const char *) path; 3119 syscallarg(int) name; 3120 } */ 3121 int error; 3122 struct pathbuf *pb; 3123 struct nameidata nd; 3124 3125 error = pathbuf_copyin(SCARG(uap, path), &pb); 3126 if (error) { 3127 return error; 3128 } 3129 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3130 if ((error = namei(&nd)) != 0) { 3131 pathbuf_destroy(pb); 3132 return (error); 3133 } 3134 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3135 vput(nd.ni_vp); 3136 pathbuf_destroy(pb); 3137 return (error); 3138 } 3139 3140 /* 3141 * Return target name of a symbolic link. 3142 */ 3143 /* ARGSUSED */ 3144 int 3145 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3146 register_t *retval) 3147 { 3148 /* { 3149 syscallarg(const char *) path; 3150 syscallarg(char *) buf; 3151 syscallarg(size_t) count; 3152 } */ 3153 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3154 SCARG(uap, buf), SCARG(uap, count), retval); 3155 } 3156 3157 static int 3158 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3159 size_t count, register_t *retval) 3160 { 3161 struct vnode *vp; 3162 struct iovec aiov; 3163 struct uio auio; 3164 int error; 3165 struct pathbuf *pb; 3166 struct nameidata nd; 3167 3168 error = pathbuf_copyin(path, &pb); 3169 if (error) { 3170 return error; 3171 } 3172 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3173 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3174 pathbuf_destroy(pb); 3175 return error; 3176 } 3177 vp = nd.ni_vp; 3178 pathbuf_destroy(pb); 3179 if (vp->v_type != VLNK) 3180 error = EINVAL; 3181 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3182 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3183 aiov.iov_base = buf; 3184 aiov.iov_len = count; 3185 auio.uio_iov = &aiov; 3186 auio.uio_iovcnt = 1; 3187 auio.uio_offset = 0; 3188 auio.uio_rw = UIO_READ; 3189 KASSERT(l == curlwp); 3190 auio.uio_vmspace = l->l_proc->p_vmspace; 3191 auio.uio_resid = count; 3192 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3193 *retval = count - auio.uio_resid; 3194 } 3195 vput(vp); 3196 return (error); 3197 } 3198 3199 int 3200 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3201 register_t *retval) 3202 { 3203 /* { 3204 syscallarg(int) fd; 3205 syscallarg(const char *) path; 3206 syscallarg(char *) buf; 3207 syscallarg(size_t) bufsize; 3208 } */ 3209 3210 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3211 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3212 } 3213 3214 /* 3215 * Change flags of a file given a path name. 3216 */ 3217 /* ARGSUSED */ 3218 int 3219 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3220 { 3221 /* { 3222 syscallarg(const char *) path; 3223 syscallarg(u_long) flags; 3224 } */ 3225 struct vnode *vp; 3226 int error; 3227 3228 error = namei_simple_user(SCARG(uap, path), 3229 NSM_FOLLOW_TRYEMULROOT, &vp); 3230 if (error != 0) 3231 return (error); 3232 error = change_flags(vp, SCARG(uap, flags), l); 3233 vput(vp); 3234 return (error); 3235 } 3236 3237 /* 3238 * Change flags of a file given a file descriptor. 3239 */ 3240 /* ARGSUSED */ 3241 int 3242 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3243 { 3244 /* { 3245 syscallarg(int) fd; 3246 syscallarg(u_long) flags; 3247 } */ 3248 struct vnode *vp; 3249 file_t *fp; 3250 int error; 3251 3252 /* fd_getvnode() will use the descriptor for us */ 3253 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3254 return (error); 3255 vp = fp->f_data; 3256 error = change_flags(vp, SCARG(uap, flags), l); 3257 VOP_UNLOCK(vp); 3258 fd_putfile(SCARG(uap, fd)); 3259 return (error); 3260 } 3261 3262 /* 3263 * Change flags of a file given a path name; this version does 3264 * not follow links. 3265 */ 3266 int 3267 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3268 { 3269 /* { 3270 syscallarg(const char *) path; 3271 syscallarg(u_long) flags; 3272 } */ 3273 struct vnode *vp; 3274 int error; 3275 3276 error = namei_simple_user(SCARG(uap, path), 3277 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3278 if (error != 0) 3279 return (error); 3280 error = change_flags(vp, SCARG(uap, flags), l); 3281 vput(vp); 3282 return (error); 3283 } 3284 3285 /* 3286 * Common routine to change flags of a file. 3287 */ 3288 int 3289 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3290 { 3291 struct vattr vattr; 3292 int error; 3293 3294 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3295 3296 vattr_null(&vattr); 3297 vattr.va_flags = flags; 3298 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3299 3300 return (error); 3301 } 3302 3303 /* 3304 * Change mode of a file given path name; this version follows links. 3305 */ 3306 /* ARGSUSED */ 3307 int 3308 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3309 { 3310 /* { 3311 syscallarg(const char *) path; 3312 syscallarg(int) mode; 3313 } */ 3314 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3315 SCARG(uap, mode), 0); 3316 } 3317 3318 int 3319 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3320 { 3321 int error; 3322 struct vnode *vp; 3323 namei_simple_flags_t ns_flag; 3324 3325 if (flags & AT_SYMLINK_NOFOLLOW) 3326 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3327 else 3328 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3329 3330 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3331 if (error != 0) 3332 return error; 3333 3334 error = change_mode(vp, mode, l); 3335 3336 vrele(vp); 3337 3338 return (error); 3339 } 3340 3341 /* 3342 * Change mode of a file given a file descriptor. 3343 */ 3344 /* ARGSUSED */ 3345 int 3346 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3347 { 3348 /* { 3349 syscallarg(int) fd; 3350 syscallarg(int) mode; 3351 } */ 3352 file_t *fp; 3353 int error; 3354 3355 /* fd_getvnode() will use the descriptor for us */ 3356 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3357 return (error); 3358 error = change_mode(fp->f_data, SCARG(uap, mode), l); 3359 fd_putfile(SCARG(uap, fd)); 3360 return (error); 3361 } 3362 3363 int 3364 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3365 register_t *retval) 3366 { 3367 /* { 3368 syscallarg(int) fd; 3369 syscallarg(const char *) path; 3370 syscallarg(int) mode; 3371 syscallarg(int) flag; 3372 } */ 3373 3374 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3375 SCARG(uap, mode), SCARG(uap, flag)); 3376 } 3377 3378 /* 3379 * Change mode of a file given path name; this version does not follow links. 3380 */ 3381 /* ARGSUSED */ 3382 int 3383 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3384 { 3385 /* { 3386 syscallarg(const char *) path; 3387 syscallarg(int) mode; 3388 } */ 3389 int error; 3390 struct vnode *vp; 3391 3392 error = namei_simple_user(SCARG(uap, path), 3393 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3394 if (error != 0) 3395 return (error); 3396 3397 error = change_mode(vp, SCARG(uap, mode), l); 3398 3399 vrele(vp); 3400 return (error); 3401 } 3402 3403 /* 3404 * Common routine to set mode given a vnode. 3405 */ 3406 static int 3407 change_mode(struct vnode *vp, int mode, struct lwp *l) 3408 { 3409 struct vattr vattr; 3410 int error; 3411 3412 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3413 vattr_null(&vattr); 3414 vattr.va_mode = mode & ALLPERMS; 3415 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3416 VOP_UNLOCK(vp); 3417 return (error); 3418 } 3419 3420 /* 3421 * Set ownership given a path name; this version follows links. 3422 */ 3423 /* ARGSUSED */ 3424 int 3425 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3426 { 3427 /* { 3428 syscallarg(const char *) path; 3429 syscallarg(uid_t) uid; 3430 syscallarg(gid_t) gid; 3431 } */ 3432 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3433 SCARG(uap, gid), 0); 3434 } 3435 3436 int 3437 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3438 gid_t gid, int flags) 3439 { 3440 int error; 3441 struct vnode *vp; 3442 namei_simple_flags_t ns_flag; 3443 3444 if (flags & AT_SYMLINK_NOFOLLOW) 3445 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3446 else 3447 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3448 3449 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3450 if (error != 0) 3451 return error; 3452 3453 error = change_owner(vp, uid, gid, l, 0); 3454 3455 vrele(vp); 3456 3457 return (error); 3458 } 3459 3460 /* 3461 * Set ownership given a path name; this version follows links. 3462 * Provides POSIX semantics. 3463 */ 3464 /* ARGSUSED */ 3465 int 3466 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3467 { 3468 /* { 3469 syscallarg(const char *) path; 3470 syscallarg(uid_t) uid; 3471 syscallarg(gid_t) gid; 3472 } */ 3473 int error; 3474 struct vnode *vp; 3475 3476 error = namei_simple_user(SCARG(uap, path), 3477 NSM_FOLLOW_TRYEMULROOT, &vp); 3478 if (error != 0) 3479 return (error); 3480 3481 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3482 3483 vrele(vp); 3484 return (error); 3485 } 3486 3487 /* 3488 * Set ownership given a file descriptor. 3489 */ 3490 /* ARGSUSED */ 3491 int 3492 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3493 { 3494 /* { 3495 syscallarg(int) fd; 3496 syscallarg(uid_t) uid; 3497 syscallarg(gid_t) gid; 3498 } */ 3499 int error; 3500 file_t *fp; 3501 3502 /* fd_getvnode() will use the descriptor for us */ 3503 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3504 return (error); 3505 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3506 l, 0); 3507 fd_putfile(SCARG(uap, fd)); 3508 return (error); 3509 } 3510 3511 int 3512 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3513 register_t *retval) 3514 { 3515 /* { 3516 syscallarg(int) fd; 3517 syscallarg(const char *) path; 3518 syscallarg(uid_t) owner; 3519 syscallarg(gid_t) group; 3520 syscallarg(int) flag; 3521 } */ 3522 3523 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3524 SCARG(uap, owner), SCARG(uap, group), 3525 SCARG(uap, flag)); 3526 } 3527 3528 /* 3529 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3530 */ 3531 /* ARGSUSED */ 3532 int 3533 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3534 { 3535 /* { 3536 syscallarg(int) fd; 3537 syscallarg(uid_t) uid; 3538 syscallarg(gid_t) gid; 3539 } */ 3540 int error; 3541 file_t *fp; 3542 3543 /* fd_getvnode() will use the descriptor for us */ 3544 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3545 return (error); 3546 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3547 l, 1); 3548 fd_putfile(SCARG(uap, fd)); 3549 return (error); 3550 } 3551 3552 /* 3553 * Set ownership given a path name; this version does not follow links. 3554 */ 3555 /* ARGSUSED */ 3556 int 3557 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3558 { 3559 /* { 3560 syscallarg(const char *) path; 3561 syscallarg(uid_t) uid; 3562 syscallarg(gid_t) gid; 3563 } */ 3564 int error; 3565 struct vnode *vp; 3566 3567 error = namei_simple_user(SCARG(uap, path), 3568 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3569 if (error != 0) 3570 return (error); 3571 3572 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3573 3574 vrele(vp); 3575 return (error); 3576 } 3577 3578 /* 3579 * Set ownership given a path name; this version does not follow links. 3580 * Provides POSIX/XPG semantics. 3581 */ 3582 /* ARGSUSED */ 3583 int 3584 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3585 { 3586 /* { 3587 syscallarg(const char *) path; 3588 syscallarg(uid_t) uid; 3589 syscallarg(gid_t) gid; 3590 } */ 3591 int error; 3592 struct vnode *vp; 3593 3594 error = namei_simple_user(SCARG(uap, path), 3595 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3596 if (error != 0) 3597 return (error); 3598 3599 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3600 3601 vrele(vp); 3602 return (error); 3603 } 3604 3605 /* 3606 * Common routine to set ownership given a vnode. 3607 */ 3608 static int 3609 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3610 int posix_semantics) 3611 { 3612 struct vattr vattr; 3613 mode_t newmode; 3614 int error; 3615 3616 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3617 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3618 goto out; 3619 3620 #define CHANGED(x) ((int)(x) != -1) 3621 newmode = vattr.va_mode; 3622 if (posix_semantics) { 3623 /* 3624 * POSIX/XPG semantics: if the caller is not the super-user, 3625 * clear set-user-id and set-group-id bits. Both POSIX and 3626 * the XPG consider the behaviour for calls by the super-user 3627 * implementation-defined; we leave the set-user-id and set- 3628 * group-id settings intact in that case. 3629 */ 3630 if (vattr.va_mode & S_ISUID) { 3631 if (kauth_authorize_vnode(l->l_cred, 3632 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3633 newmode &= ~S_ISUID; 3634 } 3635 if (vattr.va_mode & S_ISGID) { 3636 if (kauth_authorize_vnode(l->l_cred, 3637 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3638 newmode &= ~S_ISGID; 3639 } 3640 } else { 3641 /* 3642 * NetBSD semantics: when changing owner and/or group, 3643 * clear the respective bit(s). 3644 */ 3645 if (CHANGED(uid)) 3646 newmode &= ~S_ISUID; 3647 if (CHANGED(gid)) 3648 newmode &= ~S_ISGID; 3649 } 3650 /* Update va_mode iff altered. */ 3651 if (vattr.va_mode == newmode) 3652 newmode = VNOVAL; 3653 3654 vattr_null(&vattr); 3655 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3656 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3657 vattr.va_mode = newmode; 3658 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3659 #undef CHANGED 3660 3661 out: 3662 VOP_UNLOCK(vp); 3663 return (error); 3664 } 3665 3666 /* 3667 * Set the access and modification times given a path name; this 3668 * version follows links. 3669 */ 3670 /* ARGSUSED */ 3671 int 3672 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3673 register_t *retval) 3674 { 3675 /* { 3676 syscallarg(const char *) path; 3677 syscallarg(const struct timeval *) tptr; 3678 } */ 3679 3680 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3681 SCARG(uap, tptr), UIO_USERSPACE); 3682 } 3683 3684 /* 3685 * Set the access and modification times given a file descriptor. 3686 */ 3687 /* ARGSUSED */ 3688 int 3689 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3690 register_t *retval) 3691 { 3692 /* { 3693 syscallarg(int) fd; 3694 syscallarg(const struct timeval *) tptr; 3695 } */ 3696 int error; 3697 file_t *fp; 3698 3699 /* fd_getvnode() will use the descriptor for us */ 3700 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3701 return (error); 3702 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3703 UIO_USERSPACE); 3704 fd_putfile(SCARG(uap, fd)); 3705 return (error); 3706 } 3707 3708 int 3709 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3710 register_t *retval) 3711 { 3712 /* { 3713 syscallarg(int) fd; 3714 syscallarg(const struct timespec *) tptr; 3715 } */ 3716 int error; 3717 file_t *fp; 3718 3719 /* fd_getvnode() will use the descriptor for us */ 3720 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3721 return (error); 3722 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0, 3723 SCARG(uap, tptr), UIO_USERSPACE); 3724 fd_putfile(SCARG(uap, fd)); 3725 return (error); 3726 } 3727 3728 /* 3729 * Set the access and modification times given a path name; this 3730 * version does not follow links. 3731 */ 3732 int 3733 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3734 register_t *retval) 3735 { 3736 /* { 3737 syscallarg(const char *) path; 3738 syscallarg(const struct timeval *) tptr; 3739 } */ 3740 3741 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3742 SCARG(uap, tptr), UIO_USERSPACE); 3743 } 3744 3745 int 3746 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3747 register_t *retval) 3748 { 3749 /* { 3750 syscallarg(int) fd; 3751 syscallarg(const char *) path; 3752 syscallarg(const struct timespec *) tptr; 3753 syscallarg(int) flag; 3754 } */ 3755 int follow; 3756 const struct timespec *tptr; 3757 int error; 3758 3759 tptr = SCARG(uap, tptr); 3760 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3761 3762 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3763 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3764 3765 return error; 3766 } 3767 3768 /* 3769 * Common routine to set access and modification times given a vnode. 3770 */ 3771 int 3772 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3773 const struct timespec *tptr, enum uio_seg seg) 3774 { 3775 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3776 } 3777 3778 int 3779 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3780 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3781 { 3782 struct vattr vattr; 3783 int error, dorele = 0; 3784 namei_simple_flags_t sflags; 3785 bool vanull, setbirthtime; 3786 struct timespec ts[2]; 3787 3788 KASSERT(l != NULL || fdat == AT_FDCWD); 3789 3790 /* 3791 * I have checked all callers and they pass either FOLLOW, 3792 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3793 * is 0. More to the point, they don't pass anything else. 3794 * Let's keep it that way at least until the namei interfaces 3795 * are fully sanitized. 3796 */ 3797 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3798 sflags = (flag == FOLLOW) ? 3799 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3800 3801 if (tptr == NULL) { 3802 vanull = true; 3803 nanotime(&ts[0]); 3804 ts[1] = ts[0]; 3805 } else { 3806 vanull = false; 3807 if (seg != UIO_SYSSPACE) { 3808 error = copyin(tptr, ts, sizeof (ts)); 3809 if (error != 0) 3810 return error; 3811 } else { 3812 ts[0] = tptr[0]; 3813 ts[1] = tptr[1]; 3814 } 3815 } 3816 3817 if (ts[0].tv_nsec == UTIME_NOW) { 3818 nanotime(&ts[0]); 3819 if (ts[1].tv_nsec == UTIME_NOW) { 3820 vanull = true; 3821 ts[1] = ts[0]; 3822 } 3823 } else if (ts[1].tv_nsec == UTIME_NOW) 3824 nanotime(&ts[1]); 3825 3826 if (vp == NULL) { 3827 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3828 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3829 if (error != 0) 3830 return error; 3831 dorele = 1; 3832 } 3833 3834 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3835 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3836 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3837 vattr_null(&vattr); 3838 3839 if (ts[0].tv_nsec != UTIME_OMIT) 3840 vattr.va_atime = ts[0]; 3841 3842 if (ts[1].tv_nsec != UTIME_OMIT) { 3843 vattr.va_mtime = ts[1]; 3844 if (setbirthtime) 3845 vattr.va_birthtime = ts[1]; 3846 } 3847 3848 if (vanull) 3849 vattr.va_vaflags |= VA_UTIMES_NULL; 3850 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3851 VOP_UNLOCK(vp); 3852 3853 if (dorele != 0) 3854 vrele(vp); 3855 3856 return error; 3857 } 3858 3859 int 3860 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3861 const struct timeval *tptr, enum uio_seg seg) 3862 { 3863 struct timespec ts[2]; 3864 struct timespec *tsptr = NULL; 3865 int error; 3866 3867 if (tptr != NULL) { 3868 struct timeval tv[2]; 3869 3870 if (seg != UIO_SYSSPACE) { 3871 error = copyin(tptr, tv, sizeof (tv)); 3872 if (error != 0) 3873 return error; 3874 tptr = tv; 3875 } 3876 3877 if ((tv[0].tv_usec == UTIME_NOW) || 3878 (tv[0].tv_usec == UTIME_OMIT)) 3879 ts[0].tv_nsec = tv[0].tv_usec; 3880 else 3881 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3882 3883 if ((tv[1].tv_usec == UTIME_NOW) || 3884 (tv[1].tv_usec == UTIME_OMIT)) 3885 ts[1].tv_nsec = tv[1].tv_usec; 3886 else 3887 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3888 3889 tsptr = &ts[0]; 3890 } 3891 3892 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3893 } 3894 3895 /* 3896 * Truncate a file given its path name. 3897 */ 3898 /* ARGSUSED */ 3899 int 3900 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3901 { 3902 /* { 3903 syscallarg(const char *) path; 3904 syscallarg(int) pad; 3905 syscallarg(off_t) length; 3906 } */ 3907 struct vnode *vp; 3908 struct vattr vattr; 3909 int error; 3910 3911 error = namei_simple_user(SCARG(uap, path), 3912 NSM_FOLLOW_TRYEMULROOT, &vp); 3913 if (error != 0) 3914 return (error); 3915 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3916 if (vp->v_type == VDIR) 3917 error = EISDIR; 3918 else if ((error = vn_writechk(vp)) == 0 && 3919 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3920 vattr_null(&vattr); 3921 vattr.va_size = SCARG(uap, length); 3922 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3923 } 3924 vput(vp); 3925 return (error); 3926 } 3927 3928 /* 3929 * Truncate a file given a file descriptor. 3930 */ 3931 /* ARGSUSED */ 3932 int 3933 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3934 { 3935 /* { 3936 syscallarg(int) fd; 3937 syscallarg(int) pad; 3938 syscallarg(off_t) length; 3939 } */ 3940 struct vattr vattr; 3941 struct vnode *vp; 3942 file_t *fp; 3943 int error; 3944 3945 /* fd_getvnode() will use the descriptor for us */ 3946 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3947 return (error); 3948 if ((fp->f_flag & FWRITE) == 0) { 3949 error = EINVAL; 3950 goto out; 3951 } 3952 vp = fp->f_data; 3953 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3954 if (vp->v_type == VDIR) 3955 error = EISDIR; 3956 else if ((error = vn_writechk(vp)) == 0) { 3957 vattr_null(&vattr); 3958 vattr.va_size = SCARG(uap, length); 3959 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3960 } 3961 VOP_UNLOCK(vp); 3962 out: 3963 fd_putfile(SCARG(uap, fd)); 3964 return (error); 3965 } 3966 3967 /* 3968 * Sync an open file. 3969 */ 3970 /* ARGSUSED */ 3971 int 3972 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3973 { 3974 /* { 3975 syscallarg(int) fd; 3976 } */ 3977 struct vnode *vp; 3978 file_t *fp; 3979 int error; 3980 3981 /* fd_getvnode() will use the descriptor for us */ 3982 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3983 return (error); 3984 vp = fp->f_data; 3985 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3986 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3987 VOP_UNLOCK(vp); 3988 fd_putfile(SCARG(uap, fd)); 3989 return (error); 3990 } 3991 3992 /* 3993 * Sync a range of file data. API modeled after that found in AIX. 3994 * 3995 * FDATASYNC indicates that we need only save enough metadata to be able 3996 * to re-read the written data. Note we duplicate AIX's requirement that 3997 * the file be open for writing. 3998 */ 3999 /* ARGSUSED */ 4000 int 4001 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4002 { 4003 /* { 4004 syscallarg(int) fd; 4005 syscallarg(int) flags; 4006 syscallarg(off_t) start; 4007 syscallarg(off_t) length; 4008 } */ 4009 struct vnode *vp; 4010 file_t *fp; 4011 int flags, nflags; 4012 off_t s, e, len; 4013 int error; 4014 4015 /* fd_getvnode() will use the descriptor for us */ 4016 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4017 return (error); 4018 4019 if ((fp->f_flag & FWRITE) == 0) { 4020 error = EBADF; 4021 goto out; 4022 } 4023 4024 flags = SCARG(uap, flags); 4025 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4026 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4027 error = EINVAL; 4028 goto out; 4029 } 4030 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4031 if (flags & FDATASYNC) 4032 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4033 else 4034 nflags = FSYNC_WAIT; 4035 if (flags & FDISKSYNC) 4036 nflags |= FSYNC_CACHE; 4037 4038 len = SCARG(uap, length); 4039 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4040 if (len) { 4041 s = SCARG(uap, start); 4042 e = s + len; 4043 if (e < s) { 4044 error = EINVAL; 4045 goto out; 4046 } 4047 } else { 4048 e = 0; 4049 s = 0; 4050 } 4051 4052 vp = fp->f_data; 4053 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4054 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4055 VOP_UNLOCK(vp); 4056 out: 4057 fd_putfile(SCARG(uap, fd)); 4058 return (error); 4059 } 4060 4061 /* 4062 * Sync the data of an open file. 4063 */ 4064 /* ARGSUSED */ 4065 int 4066 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4067 { 4068 /* { 4069 syscallarg(int) fd; 4070 } */ 4071 struct vnode *vp; 4072 file_t *fp; 4073 int error; 4074 4075 /* fd_getvnode() will use the descriptor for us */ 4076 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4077 return (error); 4078 if ((fp->f_flag & FWRITE) == 0) { 4079 fd_putfile(SCARG(uap, fd)); 4080 return (EBADF); 4081 } 4082 vp = fp->f_data; 4083 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4084 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4085 VOP_UNLOCK(vp); 4086 fd_putfile(SCARG(uap, fd)); 4087 return (error); 4088 } 4089 4090 /* 4091 * Rename files, (standard) BSD semantics frontend. 4092 */ 4093 /* ARGSUSED */ 4094 int 4095 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4096 { 4097 /* { 4098 syscallarg(const char *) from; 4099 syscallarg(const char *) to; 4100 } */ 4101 4102 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4103 SCARG(uap, to), UIO_USERSPACE, 0)); 4104 } 4105 4106 int 4107 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4108 register_t *retval) 4109 { 4110 /* { 4111 syscallarg(int) fromfd; 4112 syscallarg(const char *) from; 4113 syscallarg(int) tofd; 4114 syscallarg(const char *) to; 4115 } */ 4116 4117 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4118 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4119 } 4120 4121 /* 4122 * Rename files, POSIX semantics frontend. 4123 */ 4124 /* ARGSUSED */ 4125 int 4126 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4127 { 4128 /* { 4129 syscallarg(const char *) from; 4130 syscallarg(const char *) to; 4131 } */ 4132 4133 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4134 SCARG(uap, to), UIO_USERSPACE, 1)); 4135 } 4136 4137 /* 4138 * Rename files. Source and destination must either both be directories, 4139 * or both not be directories. If target is a directory, it must be empty. 4140 * If `from' and `to' refer to the same object, the value of the `retain' 4141 * argument is used to determine whether `from' will be 4142 * 4143 * (retain == 0) deleted unless `from' and `to' refer to the same 4144 * object in the file system's name space (BSD). 4145 * (retain == 1) always retained (POSIX). 4146 * 4147 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4148 */ 4149 int 4150 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4151 { 4152 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4153 } 4154 4155 static int 4156 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4157 const char *to, enum uio_seg seg, int retain) 4158 { 4159 struct pathbuf *fpb, *tpb; 4160 struct nameidata fnd, tnd; 4161 struct vnode *fdvp, *fvp; 4162 struct vnode *tdvp, *tvp; 4163 struct mount *mp, *tmp; 4164 int error; 4165 4166 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4167 4168 error = pathbuf_maybe_copyin(from, seg, &fpb); 4169 if (error) 4170 goto out0; 4171 KASSERT(fpb != NULL); 4172 4173 error = pathbuf_maybe_copyin(to, seg, &tpb); 4174 if (error) 4175 goto out1; 4176 KASSERT(tpb != NULL); 4177 4178 /* 4179 * Lookup from. 4180 * 4181 * XXX LOCKPARENT is wrong because we don't actually want it 4182 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4183 * insane, so for the time being we need to leave it like this. 4184 */ 4185 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT | INRENAME), fpb); 4186 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4187 goto out2; 4188 4189 /* 4190 * Pull out the important results of the lookup, fdvp and fvp. 4191 * Of course, fvp is bogus because we're about to unlock fdvp. 4192 */ 4193 fdvp = fnd.ni_dvp; 4194 fvp = fnd.ni_vp; 4195 KASSERT(fdvp != NULL); 4196 KASSERT(fvp != NULL); 4197 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4198 4199 /* 4200 * Make sure neither fdvp nor fvp is locked. 4201 */ 4202 if (fdvp != fvp) 4203 VOP_UNLOCK(fdvp); 4204 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4205 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4206 4207 /* 4208 * Reject renaming `.' and `..'. Can't do this until after 4209 * namei because we need namei's parsing to find the final 4210 * component name. (namei should just leave us with the final 4211 * component name and not look it up itself, but anyway...) 4212 * 4213 * This was here before because we used to relookup from 4214 * instead of to and relookup requires the caller to check 4215 * this, but now file systems may depend on this check, so we 4216 * must retain it until the file systems are all rototilled. 4217 */ 4218 if (((fnd.ni_cnd.cn_namelen == 1) && 4219 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4220 ((fnd.ni_cnd.cn_namelen == 2) && 4221 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4222 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4223 error = EINVAL; /* XXX EISDIR? */ 4224 goto abort0; 4225 } 4226 4227 /* 4228 * Lookup to. 4229 * 4230 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4231 * fvp here to decide whether to add CREATEDIR is a load of 4232 * bollocks because fvp might be the wrong node by now, since 4233 * fdvp is unlocked. 4234 * 4235 * XXX Why not pass CREATEDIR always? 4236 */ 4237 NDINIT(&tnd, RENAME, 4238 (LOCKPARENT | NOCACHE | TRYEMULROOT | INRENAME | 4239 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4240 tpb); 4241 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4242 goto abort0; 4243 4244 /* 4245 * Pull out the important results of the lookup, tdvp and tvp. 4246 * Of course, tvp is bogus because we're about to unlock tdvp. 4247 */ 4248 tdvp = tnd.ni_dvp; 4249 tvp = tnd.ni_vp; 4250 KASSERT(tdvp != NULL); 4251 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4252 4253 /* 4254 * Make sure neither tdvp nor tvp is locked. 4255 */ 4256 if (tdvp != tvp) 4257 VOP_UNLOCK(tdvp); 4258 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4259 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4260 4261 /* 4262 * Reject renaming onto `.' or `..'. relookup is unhappy with 4263 * these, which is why we must do this here. Once upon a time 4264 * we relooked up from instead of to, and consequently didn't 4265 * need this check, but now that we relookup to instead of 4266 * from, we need this; and we shall need it forever forward 4267 * until the VOP_RENAME protocol changes, because file systems 4268 * will no doubt begin to depend on this check. 4269 */ 4270 if (((tnd.ni_cnd.cn_namelen == 1) && 4271 (tnd.ni_cnd.cn_nameptr[0] == '.')) || 4272 ((tnd.ni_cnd.cn_namelen == 2) && 4273 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4274 (tnd.ni_cnd.cn_nameptr[1] == '.'))) { 4275 error = EINVAL; /* XXX EISDIR? */ 4276 goto abort1; 4277 } 4278 4279 /* 4280 * Get the mount point. If the file system has been unmounted, 4281 * which it may be because we're not holding any vnode locks, 4282 * then v_mount will be NULL. We're not really supposed to 4283 * read v_mount without holding the vnode lock, but since we 4284 * have fdvp referenced, if fdvp->v_mount changes then at worst 4285 * it will be set to NULL, not changed to another mount point. 4286 * And, of course, since it is up to the file system to 4287 * determine the real lock order, we can't lock both fdvp and 4288 * tdvp at the same time. 4289 */ 4290 mp = fdvp->v_mount; 4291 if (mp == NULL) { 4292 error = ENOENT; 4293 goto abort1; 4294 } 4295 4296 /* 4297 * Make sure the mount points match. Again, although we don't 4298 * hold any vnode locks, the v_mount fields may change -- but 4299 * at worst they will change to NULL, so this will never become 4300 * a cross-device rename, because we hold vnode references. 4301 * 4302 * XXX Because nothing is locked and the compiler may reorder 4303 * things here, unmounting the file system at an inopportune 4304 * moment may cause rename to fail with ENXDEV when it really 4305 * should fail with ENOENT. 4306 */ 4307 tmp = tdvp->v_mount; 4308 if (tmp == NULL) { 4309 error = ENOENT; 4310 goto abort1; 4311 } 4312 4313 if (mp != tmp) { 4314 error = EXDEV; 4315 goto abort1; 4316 } 4317 4318 /* 4319 * Take the vfs rename lock to avoid cross-directory screw cases. 4320 * Nothing is locked currently, so taking this lock is safe. 4321 */ 4322 error = VFS_RENAMELOCK_ENTER(mp); 4323 if (error) 4324 goto abort1; 4325 4326 /* 4327 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4328 * and nothing is locked except for the vfs rename lock. 4329 * 4330 * The next step is a little rain dance to conform to the 4331 * insane lock protocol, even though it does nothing to ward 4332 * off race conditions. 4333 * 4334 * We need tdvp and tvp to be locked. However, because we have 4335 * unlocked tdvp in order to hold no locks while we take the 4336 * vfs rename lock, tvp may be wrong here, and we can't safely 4337 * lock it even if the sensible file systems will just unlock 4338 * it straight away. Consequently, we must lock tdvp and then 4339 * relookup tvp to get it locked. 4340 * 4341 * Finally, because the VOP_RENAME protocol is brain-damaged 4342 * and various file systems insanely depend on the semantics of 4343 * this brain damage, the lookup of to must be the last lookup 4344 * before VOP_RENAME. 4345 */ 4346 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4347 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4348 if (error) 4349 goto abort2; 4350 4351 /* 4352 * Drop the old tvp and pick up the new one -- which might be 4353 * the same, but that doesn't matter to us. After this, tdvp 4354 * and tvp should both be locked. 4355 */ 4356 if (tvp != NULL) 4357 vrele(tvp); 4358 tvp = tnd.ni_vp; 4359 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4360 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4361 4362 /* 4363 * The old do_sys_rename had various consistency checks here 4364 * involving fvp and tvp. fvp is bogus already here, and tvp 4365 * will become bogus soon in any sensible file system, so the 4366 * only purpose in putting these checks here is to give lip 4367 * service to these screw cases and to acknowledge that they 4368 * exist, not actually to handle them, but here you go 4369 * anyway... 4370 */ 4371 4372 /* 4373 * Acknowledge that directories and non-directories aren't 4374 * suposed to mix. 4375 */ 4376 if (tvp != NULL) { 4377 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4378 error = ENOTDIR; 4379 goto abort3; 4380 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4381 error = EISDIR; 4382 goto abort3; 4383 } 4384 } 4385 4386 /* 4387 * Acknowledge some random screw case, among the dozens that 4388 * might arise. 4389 */ 4390 if (fvp == tdvp) { 4391 error = EINVAL; 4392 goto abort3; 4393 } 4394 4395 /* 4396 * Acknowledge that POSIX has a wacky screw case. 4397 * 4398 * XXX Eventually the retain flag needs to be passed on to 4399 * VOP_RENAME. 4400 */ 4401 if (fvp == tvp) { 4402 if (retain) { 4403 error = 0; 4404 goto abort3; 4405 } else if ((fdvp == tdvp) && 4406 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4407 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4408 fnd.ni_cnd.cn_namelen))) { 4409 error = 0; 4410 goto abort3; 4411 } 4412 } 4413 4414 /* 4415 * Make sure veriexec can screw us up. (But a race can screw 4416 * up veriexec, of course -- remember, fvp and (soon) tvp are 4417 * bogus.) 4418 */ 4419 #if NVERIEXEC > 0 4420 { 4421 char *f1, *f2; 4422 size_t f1_len; 4423 size_t f2_len; 4424 4425 f1_len = fnd.ni_cnd.cn_namelen + 1; 4426 f1 = kmem_alloc(f1_len, KM_SLEEP); 4427 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4428 4429 f2_len = tnd.ni_cnd.cn_namelen + 1; 4430 f2 = kmem_alloc(f2_len, KM_SLEEP); 4431 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4432 4433 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4434 4435 kmem_free(f1, f1_len); 4436 kmem_free(f2, f2_len); 4437 4438 if (error) 4439 goto abort3; 4440 } 4441 #endif /* NVERIEXEC > 0 */ 4442 4443 /* 4444 * All ready. Incant the rename vop. 4445 */ 4446 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4447 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4448 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4449 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4450 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4451 4452 /* 4453 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4454 * tdvp and tvp. But we can't assert any of that. 4455 */ 4456 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4457 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4458 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4459 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4460 4461 /* 4462 * So all we have left to do is to drop the rename lock and 4463 * destroy the pathbufs. 4464 */ 4465 VFS_RENAMELOCK_EXIT(mp); 4466 goto out2; 4467 4468 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4469 VOP_UNLOCK(tvp); 4470 abort2: VOP_UNLOCK(tdvp); 4471 VFS_RENAMELOCK_EXIT(mp); 4472 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4473 vrele(tdvp); 4474 if (tvp != NULL) 4475 vrele(tvp); 4476 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4477 vrele(fdvp); 4478 vrele(fvp); 4479 out2: pathbuf_destroy(tpb); 4480 out1: pathbuf_destroy(fpb); 4481 out0: return error; 4482 } 4483 4484 /* 4485 * Make a directory file. 4486 */ 4487 /* ARGSUSED */ 4488 int 4489 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4490 { 4491 /* { 4492 syscallarg(const char *) path; 4493 syscallarg(int) mode; 4494 } */ 4495 4496 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4497 SCARG(uap, mode), UIO_USERSPACE); 4498 } 4499 4500 int 4501 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4502 register_t *retval) 4503 { 4504 /* { 4505 syscallarg(int) fd; 4506 syscallarg(const char *) path; 4507 syscallarg(int) mode; 4508 } */ 4509 4510 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4511 SCARG(uap, mode), UIO_USERSPACE); 4512 } 4513 4514 4515 int 4516 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4517 { 4518 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4519 } 4520 4521 static int 4522 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4523 enum uio_seg seg) 4524 { 4525 struct proc *p = curlwp->l_proc; 4526 struct vnode *vp; 4527 struct vattr vattr; 4528 int error; 4529 struct pathbuf *pb; 4530 struct nameidata nd; 4531 4532 KASSERT(l != NULL || fdat == AT_FDCWD); 4533 4534 /* XXX bollocks, should pass in a pathbuf */ 4535 error = pathbuf_maybe_copyin(path, seg, &pb); 4536 if (error) { 4537 return error; 4538 } 4539 4540 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4541 4542 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4543 pathbuf_destroy(pb); 4544 return (error); 4545 } 4546 vp = nd.ni_vp; 4547 if (vp != NULL) { 4548 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4549 if (nd.ni_dvp == vp) 4550 vrele(nd.ni_dvp); 4551 else 4552 vput(nd.ni_dvp); 4553 vrele(vp); 4554 pathbuf_destroy(pb); 4555 return (EEXIST); 4556 } 4557 vattr_null(&vattr); 4558 vattr.va_type = VDIR; 4559 /* We will read cwdi->cwdi_cmask unlocked. */ 4560 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4561 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4562 if (!error) 4563 vrele(nd.ni_vp); 4564 vput(nd.ni_dvp); 4565 pathbuf_destroy(pb); 4566 return (error); 4567 } 4568 4569 /* 4570 * Remove a directory file. 4571 */ 4572 /* ARGSUSED */ 4573 int 4574 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4575 { 4576 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4577 AT_REMOVEDIR, UIO_USERSPACE); 4578 } 4579 4580 /* 4581 * Read a block of directory entries in a file system independent format. 4582 */ 4583 int 4584 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4585 { 4586 /* { 4587 syscallarg(int) fd; 4588 syscallarg(char *) buf; 4589 syscallarg(size_t) count; 4590 } */ 4591 file_t *fp; 4592 int error, done; 4593 4594 /* fd_getvnode() will use the descriptor for us */ 4595 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4596 return (error); 4597 if ((fp->f_flag & FREAD) == 0) { 4598 error = EBADF; 4599 goto out; 4600 } 4601 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4602 SCARG(uap, count), &done, l, 0, 0); 4603 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4604 *retval = done; 4605 out: 4606 fd_putfile(SCARG(uap, fd)); 4607 return (error); 4608 } 4609 4610 /* 4611 * Set the mode mask for creation of filesystem nodes. 4612 */ 4613 int 4614 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4615 { 4616 /* { 4617 syscallarg(mode_t) newmask; 4618 } */ 4619 struct proc *p = l->l_proc; 4620 struct cwdinfo *cwdi; 4621 4622 /* 4623 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4624 * important is that we serialize changes to the mask. The 4625 * rw_exit() will issue a write memory barrier on our behalf, 4626 * and force the changes out to other CPUs (as it must use an 4627 * atomic operation, draining the local CPU's store buffers). 4628 */ 4629 cwdi = p->p_cwdi; 4630 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4631 *retval = cwdi->cwdi_cmask; 4632 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4633 rw_exit(&cwdi->cwdi_lock); 4634 4635 return (0); 4636 } 4637 4638 int 4639 dorevoke(struct vnode *vp, kauth_cred_t cred) 4640 { 4641 struct vattr vattr; 4642 int error, fs_decision; 4643 4644 vn_lock(vp, LK_SHARED | LK_RETRY); 4645 error = VOP_GETATTR(vp, &vattr, cred); 4646 VOP_UNLOCK(vp); 4647 if (error != 0) 4648 return error; 4649 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4650 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4651 fs_decision); 4652 if (!error) 4653 VOP_REVOKE(vp, REVOKEALL); 4654 return (error); 4655 } 4656 4657 /* 4658 * Void all references to file by ripping underlying filesystem 4659 * away from vnode. 4660 */ 4661 /* ARGSUSED */ 4662 int 4663 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4664 { 4665 /* { 4666 syscallarg(const char *) path; 4667 } */ 4668 struct vnode *vp; 4669 int error; 4670 4671 error = namei_simple_user(SCARG(uap, path), 4672 NSM_FOLLOW_TRYEMULROOT, &vp); 4673 if (error != 0) 4674 return (error); 4675 error = dorevoke(vp, l->l_cred); 4676 vrele(vp); 4677 return (error); 4678 } 4679
Indexes created Fri Sep 26 20:09:58 GMT 2025