vfs_syscalls.c revision 1.478
1 /* $NetBSD: vfs_syscalls.c,v 1.478 2014/04/04 06:47:02 maxv 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 1.478 2014/04/04 06:47:02 maxv 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 *); 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; 488 489 /* 490 * Hopefully a longer buffer won't make copyin() fail. 491 * For compatibility with 3.0 and earlier. 492 */ 493 if (flags & MNT_UPDATE 494 && data_len < sizeof (struct mnt_export_args30)) 495 data_len = sizeof (struct mnt_export_args30); 496 } 497 if (data_len > VFS_MAX_MOUNT_DATA) { 498 error = EINVAL; 499 goto done; 500 } 501 data_buf = kmem_alloc(data_len, KM_SLEEP); 502 503 /* NFS needs the buffer even for mnt_getargs .... */ 504 error = copyin(data, data_buf, data_len); 505 if (error != 0) 506 goto done; 507 } 508 509 if (flags & MNT_GETARGS) { 510 if (data_len == 0) { 511 error = EINVAL; 512 goto done; 513 } 514 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 515 if (error != 0) 516 goto done; 517 if (data_seg == UIO_USERSPACE) 518 error = copyout(data_buf, data, data_len); 519 *retval = data_len; 520 } else if (flags & MNT_UPDATE) { 521 error = mount_update(l, vp, path, flags, data_buf, &data_len); 522 } else { 523 /* Locking is handled internally in mount_domount(). */ 524 KASSERT(vfsopsrele == true); 525 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 526 &data_len); 527 vfsopsrele = false; 528 } 529 530 done: 531 if (vfsopsrele) 532 vfs_delref(vfsops); 533 if (vp != NULL) { 534 vrele(vp); 535 } 536 if (data_buf != data) 537 kmem_free(data_buf, data_len); 538 return (error); 539 } 540 541 /* 542 * Unmount a file system. 543 * 544 * Note: unmount takes a path to the vnode mounted on as argument, 545 * not special file (as before). 546 */ 547 /* ARGSUSED */ 548 int 549 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 550 { 551 /* { 552 syscallarg(const char *) path; 553 syscallarg(int) flags; 554 } */ 555 struct vnode *vp; 556 struct mount *mp; 557 int error; 558 struct pathbuf *pb; 559 struct nameidata nd; 560 561 error = pathbuf_copyin(SCARG(uap, path), &pb); 562 if (error) { 563 return error; 564 } 565 566 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 567 if ((error = namei(&nd)) != 0) { 568 pathbuf_destroy(pb); 569 return error; 570 } 571 vp = nd.ni_vp; 572 pathbuf_destroy(pb); 573 574 mp = vp->v_mount; 575 atomic_inc_uint(&mp->mnt_refcnt); 576 VOP_UNLOCK(vp); 577 578 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 579 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 580 if (error) { 581 vrele(vp); 582 vfs_destroy(mp); 583 return (error); 584 } 585 586 /* 587 * Don't allow unmounting the root file system. 588 */ 589 if (mp->mnt_flag & MNT_ROOTFS) { 590 vrele(vp); 591 vfs_destroy(mp); 592 return (EINVAL); 593 } 594 595 /* 596 * Must be the root of the filesystem 597 */ 598 if ((vp->v_vflag & VV_ROOT) == 0) { 599 vrele(vp); 600 vfs_destroy(mp); 601 return (EINVAL); 602 } 603 604 vrele(vp); 605 error = dounmount(mp, SCARG(uap, flags), l); 606 vfs_destroy(mp); 607 return error; 608 } 609 610 /* 611 * Sync each mounted filesystem. 612 */ 613 #ifdef DEBUG 614 int syncprt = 0; 615 struct ctldebug debug0 = { "syncprt", &syncprt }; 616 #endif 617 618 void 619 do_sys_sync(struct lwp *l) 620 { 621 struct mount *mp, *nmp; 622 int asyncflag; 623 624 mutex_enter(&mountlist_lock); 625 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 626 if (vfs_busy(mp, &nmp)) { 627 continue; 628 } 629 mutex_enter(&mp->mnt_updating); 630 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 631 asyncflag = mp->mnt_flag & MNT_ASYNC; 632 mp->mnt_flag &= ~MNT_ASYNC; 633 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 634 if (asyncflag) 635 mp->mnt_flag |= MNT_ASYNC; 636 } 637 mutex_exit(&mp->mnt_updating); 638 vfs_unbusy(mp, false, &nmp); 639 } 640 mutex_exit(&mountlist_lock); 641 #ifdef DEBUG 642 if (syncprt) 643 vfs_bufstats(); 644 #endif /* DEBUG */ 645 } 646 647 /* ARGSUSED */ 648 int 649 sys_sync(struct lwp *l, const void *v, register_t *retval) 650 { 651 do_sys_sync(l); 652 return (0); 653 } 654 655 656 /* 657 * Access or change filesystem quotas. 658 * 659 * (this is really 14 different calls bundled into one) 660 */ 661 662 static int 663 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 664 { 665 struct quotastat info_k; 666 int error; 667 668 /* ensure any padding bytes are cleared */ 669 memset(&info_k, 0, sizeof(info_k)); 670 671 error = vfs_quotactl_stat(mp, &info_k); 672 if (error) { 673 return error; 674 } 675 676 return copyout(&info_k, info_u, sizeof(info_k)); 677 } 678 679 static int 680 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 681 struct quotaidtypestat *info_u) 682 { 683 struct quotaidtypestat info_k; 684 int error; 685 686 /* ensure any padding bytes are cleared */ 687 memset(&info_k, 0, sizeof(info_k)); 688 689 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 690 if (error) { 691 return error; 692 } 693 694 return copyout(&info_k, info_u, sizeof(info_k)); 695 } 696 697 static int 698 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 699 struct quotaobjtypestat *info_u) 700 { 701 struct quotaobjtypestat info_k; 702 int error; 703 704 /* ensure any padding bytes are cleared */ 705 memset(&info_k, 0, sizeof(info_k)); 706 707 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 708 if (error) { 709 return error; 710 } 711 712 return copyout(&info_k, info_u, sizeof(info_k)); 713 } 714 715 static int 716 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 717 struct quotaval *val_u) 718 { 719 struct quotakey key_k; 720 struct quotaval val_k; 721 int error; 722 723 /* ensure any padding bytes are cleared */ 724 memset(&val_k, 0, sizeof(val_k)); 725 726 error = copyin(key_u, &key_k, sizeof(key_k)); 727 if (error) { 728 return error; 729 } 730 731 error = vfs_quotactl_get(mp, &key_k, &val_k); 732 if (error) { 733 return error; 734 } 735 736 return copyout(&val_k, val_u, sizeof(val_k)); 737 } 738 739 static int 740 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 741 const struct quotaval *val_u) 742 { 743 struct quotakey key_k; 744 struct quotaval val_k; 745 int error; 746 747 error = copyin(key_u, &key_k, sizeof(key_k)); 748 if (error) { 749 return error; 750 } 751 752 error = copyin(val_u, &val_k, sizeof(val_k)); 753 if (error) { 754 return error; 755 } 756 757 return vfs_quotactl_put(mp, &key_k, &val_k); 758 } 759 760 static int 761 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u) 762 { 763 struct quotakey key_k; 764 int error; 765 766 error = copyin(key_u, &key_k, sizeof(key_k)); 767 if (error) { 768 return error; 769 } 770 771 return vfs_quotactl_delete(mp, &key_k); 772 } 773 774 static int 775 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 776 { 777 struct quotakcursor cursor_k; 778 int error; 779 780 /* ensure any padding bytes are cleared */ 781 memset(&cursor_k, 0, sizeof(cursor_k)); 782 783 error = vfs_quotactl_cursoropen(mp, &cursor_k); 784 if (error) { 785 return error; 786 } 787 788 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 789 } 790 791 static int 792 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 793 { 794 struct quotakcursor cursor_k; 795 int error; 796 797 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 798 if (error) { 799 return error; 800 } 801 802 return vfs_quotactl_cursorclose(mp, &cursor_k); 803 } 804 805 static int 806 do_sys_quotactl_cursorskipidtype(struct mount *mp, 807 struct quotakcursor *cursor_u, int idtype) 808 { 809 struct quotakcursor cursor_k; 810 int error; 811 812 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 813 if (error) { 814 return error; 815 } 816 817 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 818 if (error) { 819 return error; 820 } 821 822 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 823 } 824 825 static int 826 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 827 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 828 unsigned *ret_u) 829 { 830 #define CGET_STACK_MAX 8 831 struct quotakcursor cursor_k; 832 struct quotakey stackkeys[CGET_STACK_MAX]; 833 struct quotaval stackvals[CGET_STACK_MAX]; 834 struct quotakey *keys_k; 835 struct quotaval *vals_k; 836 unsigned ret_k; 837 int error; 838 839 if (maxnum > 128) { 840 maxnum = 128; 841 } 842 843 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 844 if (error) { 845 return error; 846 } 847 848 if (maxnum <= CGET_STACK_MAX) { 849 keys_k = stackkeys; 850 vals_k = stackvals; 851 /* ensure any padding bytes are cleared */ 852 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 853 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 854 } else { 855 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 856 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 857 } 858 859 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 860 &ret_k); 861 if (error) { 862 goto fail; 863 } 864 865 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 866 if (error) { 867 goto fail; 868 } 869 870 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 871 if (error) { 872 goto fail; 873 } 874 875 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 876 if (error) { 877 goto fail; 878 } 879 880 /* do last to maximize the chance of being able to recover a failure */ 881 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 882 883 fail: 884 if (keys_k != stackkeys) { 885 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 886 } 887 if (vals_k != stackvals) { 888 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 889 } 890 return error; 891 } 892 893 static int 894 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 895 int *ret_u) 896 { 897 struct quotakcursor cursor_k; 898 int ret_k; 899 int error; 900 901 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 902 if (error) { 903 return error; 904 } 905 906 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 907 if (error) { 908 return error; 909 } 910 911 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 912 if (error) { 913 return error; 914 } 915 916 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 917 } 918 919 static int 920 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 921 { 922 struct quotakcursor cursor_k; 923 int error; 924 925 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 926 if (error) { 927 return error; 928 } 929 930 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 931 if (error) { 932 return error; 933 } 934 935 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 936 } 937 938 static int 939 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 940 { 941 char *path_k; 942 int error; 943 944 /* XXX this should probably be a struct pathbuf */ 945 path_k = PNBUF_GET(); 946 error = copyin(path_u, path_k, PATH_MAX); 947 if (error) { 948 PNBUF_PUT(path_k); 949 return error; 950 } 951 952 error = vfs_quotactl_quotaon(mp, idtype, path_k); 953 954 PNBUF_PUT(path_k); 955 return error; 956 } 957 958 static int 959 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 960 { 961 return vfs_quotactl_quotaoff(mp, idtype); 962 } 963 964 int 965 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 966 { 967 struct mount *mp; 968 struct vnode *vp; 969 int error; 970 971 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 972 if (error != 0) 973 return (error); 974 mp = vp->v_mount; 975 976 switch (args->qc_op) { 977 case QUOTACTL_STAT: 978 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 979 break; 980 case QUOTACTL_IDTYPESTAT: 981 error = do_sys_quotactl_idtypestat(mp, 982 args->u.idtypestat.qc_idtype, 983 args->u.idtypestat.qc_info); 984 break; 985 case QUOTACTL_OBJTYPESTAT: 986 error = do_sys_quotactl_objtypestat(mp, 987 args->u.objtypestat.qc_objtype, 988 args->u.objtypestat.qc_info); 989 break; 990 case QUOTACTL_GET: 991 error = do_sys_quotactl_get(mp, 992 args->u.get.qc_key, 993 args->u.get.qc_val); 994 break; 995 case QUOTACTL_PUT: 996 error = do_sys_quotactl_put(mp, 997 args->u.put.qc_key, 998 args->u.put.qc_val); 999 break; 1000 case QUOTACTL_DELETE: 1001 error = do_sys_quotactl_delete(mp, args->u.delete.qc_key); 1002 break; 1003 case QUOTACTL_CURSOROPEN: 1004 error = do_sys_quotactl_cursoropen(mp, 1005 args->u.cursoropen.qc_cursor); 1006 break; 1007 case QUOTACTL_CURSORCLOSE: 1008 error = do_sys_quotactl_cursorclose(mp, 1009 args->u.cursorclose.qc_cursor); 1010 break; 1011 case QUOTACTL_CURSORSKIPIDTYPE: 1012 error = do_sys_quotactl_cursorskipidtype(mp, 1013 args->u.cursorskipidtype.qc_cursor, 1014 args->u.cursorskipidtype.qc_idtype); 1015 break; 1016 case QUOTACTL_CURSORGET: 1017 error = do_sys_quotactl_cursorget(mp, 1018 args->u.cursorget.qc_cursor, 1019 args->u.cursorget.qc_keys, 1020 args->u.cursorget.qc_vals, 1021 args->u.cursorget.qc_maxnum, 1022 args->u.cursorget.qc_ret); 1023 break; 1024 case QUOTACTL_CURSORATEND: 1025 error = do_sys_quotactl_cursoratend(mp, 1026 args->u.cursoratend.qc_cursor, 1027 args->u.cursoratend.qc_ret); 1028 break; 1029 case QUOTACTL_CURSORREWIND: 1030 error = do_sys_quotactl_cursorrewind(mp, 1031 args->u.cursorrewind.qc_cursor); 1032 break; 1033 case QUOTACTL_QUOTAON: 1034 error = do_sys_quotactl_quotaon(mp, 1035 args->u.quotaon.qc_idtype, 1036 args->u.quotaon.qc_quotafile); 1037 break; 1038 case QUOTACTL_QUOTAOFF: 1039 error = do_sys_quotactl_quotaoff(mp, 1040 args->u.quotaoff.qc_idtype); 1041 break; 1042 default: 1043 error = EINVAL; 1044 break; 1045 } 1046 1047 vrele(vp); 1048 return error; 1049 } 1050 1051 /* ARGSUSED */ 1052 int 1053 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1054 register_t *retval) 1055 { 1056 /* { 1057 syscallarg(const char *) path; 1058 syscallarg(struct quotactl_args *) args; 1059 } */ 1060 struct quotactl_args args; 1061 int error; 1062 1063 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1064 if (error) { 1065 return error; 1066 } 1067 1068 return do_sys_quotactl(SCARG(uap, path), &args); 1069 } 1070 1071 int 1072 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1073 int root) 1074 { 1075 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1076 int error = 0; 1077 1078 /* 1079 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1080 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1081 * overrides MNT_NOWAIT. 1082 */ 1083 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1084 (flags != MNT_WAIT && flags != 0)) { 1085 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1086 goto done; 1087 } 1088 1089 /* Get the filesystem stats now */ 1090 memset(sp, 0, sizeof(*sp)); 1091 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1092 return error; 1093 } 1094 1095 if (cwdi->cwdi_rdir == NULL) 1096 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1097 done: 1098 if (cwdi->cwdi_rdir != NULL) { 1099 size_t len; 1100 char *bp; 1101 char c; 1102 char *path = PNBUF_GET(); 1103 1104 bp = path + MAXPATHLEN; 1105 *--bp = '\0'; 1106 rw_enter(&cwdi->cwdi_lock, RW_READER); 1107 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1108 MAXPATHLEN / 2, 0, l); 1109 rw_exit(&cwdi->cwdi_lock); 1110 if (error) { 1111 PNBUF_PUT(path); 1112 return error; 1113 } 1114 len = strlen(bp); 1115 if (len != 1) { 1116 /* 1117 * for mount points that are below our root, we can see 1118 * them, so we fix up the pathname and return them. The 1119 * rest we cannot see, so we don't allow viewing the 1120 * data. 1121 */ 1122 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1123 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1124 (void)strlcpy(sp->f_mntonname, 1125 c == '\0' ? "/" : &sp->f_mntonname[len], 1126 sizeof(sp->f_mntonname)); 1127 } else { 1128 if (root) 1129 (void)strlcpy(sp->f_mntonname, "/", 1130 sizeof(sp->f_mntonname)); 1131 else 1132 error = EPERM; 1133 } 1134 } 1135 PNBUF_PUT(path); 1136 } 1137 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1138 return error; 1139 } 1140 1141 /* 1142 * Get filesystem statistics by path. 1143 */ 1144 int 1145 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1146 { 1147 struct mount *mp; 1148 int error; 1149 struct vnode *vp; 1150 1151 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1152 if (error != 0) 1153 return error; 1154 mp = vp->v_mount; 1155 error = dostatvfs(mp, sb, l, flags, 1); 1156 vrele(vp); 1157 return error; 1158 } 1159 1160 /* ARGSUSED */ 1161 int 1162 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1163 { 1164 /* { 1165 syscallarg(const char *) path; 1166 syscallarg(struct statvfs *) buf; 1167 syscallarg(int) flags; 1168 } */ 1169 struct statvfs *sb; 1170 int error; 1171 1172 sb = STATVFSBUF_GET(); 1173 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1174 if (error == 0) 1175 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1176 STATVFSBUF_PUT(sb); 1177 return error; 1178 } 1179 1180 /* 1181 * Get filesystem statistics by fd. 1182 */ 1183 int 1184 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1185 { 1186 file_t *fp; 1187 struct mount *mp; 1188 int error; 1189 1190 /* fd_getvnode() will use the descriptor for us */ 1191 if ((error = fd_getvnode(fd, &fp)) != 0) 1192 return (error); 1193 mp = ((struct vnode *)fp->f_data)->v_mount; 1194 error = dostatvfs(mp, sb, curlwp, flags, 1); 1195 fd_putfile(fd); 1196 return error; 1197 } 1198 1199 /* ARGSUSED */ 1200 int 1201 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1202 { 1203 /* { 1204 syscallarg(int) fd; 1205 syscallarg(struct statvfs *) buf; 1206 syscallarg(int) flags; 1207 } */ 1208 struct statvfs *sb; 1209 int error; 1210 1211 sb = STATVFSBUF_GET(); 1212 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1213 if (error == 0) 1214 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1215 STATVFSBUF_PUT(sb); 1216 return error; 1217 } 1218 1219 1220 /* 1221 * Get statistics on all filesystems. 1222 */ 1223 int 1224 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1225 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1226 register_t *retval) 1227 { 1228 int root = 0; 1229 struct proc *p = l->l_proc; 1230 struct mount *mp, *nmp; 1231 struct statvfs *sb; 1232 size_t count, maxcount; 1233 int error = 0; 1234 1235 sb = STATVFSBUF_GET(); 1236 maxcount = bufsize / entry_sz; 1237 mutex_enter(&mountlist_lock); 1238 count = 0; 1239 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 1240 if (vfs_busy(mp, &nmp)) { 1241 continue; 1242 } 1243 if (sfsp && count < maxcount) { 1244 error = dostatvfs(mp, sb, l, flags, 0); 1245 if (error) { 1246 vfs_unbusy(mp, false, &nmp); 1247 error = 0; 1248 continue; 1249 } 1250 error = copyfn(sb, sfsp, entry_sz); 1251 if (error) { 1252 vfs_unbusy(mp, false, NULL); 1253 goto out; 1254 } 1255 sfsp = (char *)sfsp + entry_sz; 1256 root |= strcmp(sb->f_mntonname, "/") == 0; 1257 } 1258 count++; 1259 vfs_unbusy(mp, false, &nmp); 1260 } 1261 mutex_exit(&mountlist_lock); 1262 1263 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1264 /* 1265 * fake a root entry 1266 */ 1267 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1268 sb, l, flags, 1); 1269 if (error != 0) 1270 goto out; 1271 if (sfsp) { 1272 error = copyfn(sb, sfsp, entry_sz); 1273 if (error != 0) 1274 goto out; 1275 } 1276 count++; 1277 } 1278 if (sfsp && count > maxcount) 1279 *retval = maxcount; 1280 else 1281 *retval = count; 1282 out: 1283 STATVFSBUF_PUT(sb); 1284 return error; 1285 } 1286 1287 int 1288 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1289 { 1290 /* { 1291 syscallarg(struct statvfs *) buf; 1292 syscallarg(size_t) bufsize; 1293 syscallarg(int) flags; 1294 } */ 1295 1296 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1297 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1298 } 1299 1300 /* 1301 * Change current working directory to a given file descriptor. 1302 */ 1303 /* ARGSUSED */ 1304 int 1305 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1306 { 1307 /* { 1308 syscallarg(int) fd; 1309 } */ 1310 struct proc *p = l->l_proc; 1311 struct cwdinfo *cwdi; 1312 struct vnode *vp, *tdp; 1313 struct mount *mp; 1314 file_t *fp; 1315 int error, fd; 1316 1317 /* fd_getvnode() will use the descriptor for us */ 1318 fd = SCARG(uap, fd); 1319 if ((error = fd_getvnode(fd, &fp)) != 0) 1320 return (error); 1321 vp = fp->f_data; 1322 1323 vref(vp); 1324 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1325 if (vp->v_type != VDIR) 1326 error = ENOTDIR; 1327 else 1328 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1329 if (error) { 1330 vput(vp); 1331 goto out; 1332 } 1333 while ((mp = vp->v_mountedhere) != NULL) { 1334 error = vfs_busy(mp, NULL); 1335 vput(vp); 1336 if (error != 0) 1337 goto out; 1338 error = VFS_ROOT(mp, &tdp); 1339 vfs_unbusy(mp, false, NULL); 1340 if (error) 1341 goto out; 1342 vp = tdp; 1343 } 1344 VOP_UNLOCK(vp); 1345 1346 /* 1347 * Disallow changing to a directory not under the process's 1348 * current root directory (if there is one). 1349 */ 1350 cwdi = p->p_cwdi; 1351 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1352 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1353 vrele(vp); 1354 error = EPERM; /* operation not permitted */ 1355 } else { 1356 vrele(cwdi->cwdi_cdir); 1357 cwdi->cwdi_cdir = vp; 1358 } 1359 rw_exit(&cwdi->cwdi_lock); 1360 1361 out: 1362 fd_putfile(fd); 1363 return (error); 1364 } 1365 1366 /* 1367 * Change this process's notion of the root directory to a given file 1368 * descriptor. 1369 */ 1370 int 1371 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1372 { 1373 struct proc *p = l->l_proc; 1374 struct vnode *vp; 1375 file_t *fp; 1376 int error, fd = SCARG(uap, fd); 1377 1378 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1379 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1380 return error; 1381 /* fd_getvnode() will use the descriptor for us */ 1382 if ((error = fd_getvnode(fd, &fp)) != 0) 1383 return error; 1384 vp = fp->f_data; 1385 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1386 if (vp->v_type != VDIR) 1387 error = ENOTDIR; 1388 else 1389 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1390 VOP_UNLOCK(vp); 1391 if (error) 1392 goto out; 1393 vref(vp); 1394 1395 change_root(p->p_cwdi, vp, l); 1396 1397 out: 1398 fd_putfile(fd); 1399 return (error); 1400 } 1401 1402 /* 1403 * Change current working directory (``.''). 1404 */ 1405 /* ARGSUSED */ 1406 int 1407 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1408 { 1409 /* { 1410 syscallarg(const char *) path; 1411 } */ 1412 struct proc *p = l->l_proc; 1413 struct cwdinfo *cwdi; 1414 int error; 1415 struct vnode *vp; 1416 1417 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1418 &vp, l)) != 0) 1419 return (error); 1420 cwdi = p->p_cwdi; 1421 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1422 vrele(cwdi->cwdi_cdir); 1423 cwdi->cwdi_cdir = vp; 1424 rw_exit(&cwdi->cwdi_lock); 1425 return (0); 1426 } 1427 1428 /* 1429 * Change notion of root (``/'') directory. 1430 */ 1431 /* ARGSUSED */ 1432 int 1433 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1434 { 1435 /* { 1436 syscallarg(const char *) path; 1437 } */ 1438 struct proc *p = l->l_proc; 1439 int error; 1440 struct vnode *vp; 1441 1442 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1443 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1444 return (error); 1445 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1446 &vp, l)) != 0) 1447 return (error); 1448 1449 change_root(p->p_cwdi, vp, l); 1450 1451 return (0); 1452 } 1453 1454 /* 1455 * Common routine for chroot and fchroot. 1456 * NB: callers need to properly authorize the change root operation. 1457 */ 1458 void 1459 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1460 { 1461 struct proc *p = l->l_proc; 1462 kauth_cred_t ncred; 1463 1464 ncred = kauth_cred_alloc(); 1465 1466 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1467 if (cwdi->cwdi_rdir != NULL) 1468 vrele(cwdi->cwdi_rdir); 1469 cwdi->cwdi_rdir = vp; 1470 1471 /* 1472 * Prevent escaping from chroot by putting the root under 1473 * the working directory. Silently chdir to / if we aren't 1474 * already there. 1475 */ 1476 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1477 /* 1478 * XXX would be more failsafe to change directory to a 1479 * deadfs node here instead 1480 */ 1481 vrele(cwdi->cwdi_cdir); 1482 vref(vp); 1483 cwdi->cwdi_cdir = vp; 1484 } 1485 rw_exit(&cwdi->cwdi_lock); 1486 1487 /* Get a write lock on the process credential. */ 1488 proc_crmod_enter(); 1489 1490 kauth_cred_clone(p->p_cred, ncred); 1491 kauth_proc_chroot(ncred, p->p_cwdi); 1492 1493 /* Broadcast our credentials to the process and other LWPs. */ 1494 proc_crmod_leave(ncred, p->p_cred, true); 1495 } 1496 1497 /* 1498 * Common routine for chroot and chdir. 1499 * XXX "where" should be enum uio_seg 1500 */ 1501 int 1502 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1503 { 1504 struct pathbuf *pb; 1505 struct nameidata nd; 1506 int error; 1507 1508 error = pathbuf_maybe_copyin(path, where, &pb); 1509 if (error) { 1510 return error; 1511 } 1512 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1513 if ((error = namei(&nd)) != 0) { 1514 pathbuf_destroy(pb); 1515 return error; 1516 } 1517 *vpp = nd.ni_vp; 1518 pathbuf_destroy(pb); 1519 1520 if ((*vpp)->v_type != VDIR) 1521 error = ENOTDIR; 1522 else 1523 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1524 1525 if (error) 1526 vput(*vpp); 1527 else 1528 VOP_UNLOCK(*vpp); 1529 return (error); 1530 } 1531 1532 /* 1533 * Internals of sys_open - path has already been converted into a pathbuf 1534 * (so we can easily reuse this function from other parts of the kernel, 1535 * like posix_spawn post-processing). 1536 */ 1537 int 1538 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1539 int open_mode, int *fd) 1540 { 1541 struct proc *p = l->l_proc; 1542 struct cwdinfo *cwdi = p->p_cwdi; 1543 file_t *fp; 1544 struct vnode *vp; 1545 int flags, cmode; 1546 int indx, error; 1547 struct nameidata nd; 1548 1549 if (open_flags & O_SEARCH) { 1550 open_flags &= ~(int)O_SEARCH; 1551 } 1552 1553 flags = FFLAGS(open_flags); 1554 if ((flags & (FREAD | FWRITE)) == 0) 1555 return EINVAL; 1556 1557 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1558 return error; 1559 } 1560 1561 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1562 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1563 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1564 if (dvp != NULL) 1565 NDAT(&nd, dvp); 1566 1567 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1568 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1569 fd_abort(p, fp, indx); 1570 if ((error == EDUPFD || error == EMOVEFD) && 1571 l->l_dupfd >= 0 && /* XXX from fdopen */ 1572 (error = 1573 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1574 *fd = indx; 1575 return 0; 1576 } 1577 if (error == ERESTART) 1578 error = EINTR; 1579 return error; 1580 } 1581 1582 l->l_dupfd = 0; 1583 vp = nd.ni_vp; 1584 1585 if ((error = open_setfp(l, fp, vp, indx, flags))) 1586 return error; 1587 1588 VOP_UNLOCK(vp); 1589 *fd = indx; 1590 fd_affix(p, fp, indx); 1591 return 0; 1592 } 1593 1594 int 1595 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1596 { 1597 struct pathbuf *pb; 1598 int error, oflags; 1599 1600 oflags = FFLAGS(open_flags); 1601 if ((oflags & (FREAD | FWRITE)) == 0) 1602 return EINVAL; 1603 1604 pb = pathbuf_create(path); 1605 if (pb == NULL) 1606 return ENOMEM; 1607 1608 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1609 pathbuf_destroy(pb); 1610 1611 return error; 1612 } 1613 1614 /* 1615 * Check permissions, allocate an open file structure, 1616 * and call the device open routine if any. 1617 */ 1618 static int 1619 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1620 int mode, int *fd) 1621 { 1622 file_t *dfp = NULL; 1623 struct vnode *dvp = NULL; 1624 struct pathbuf *pb; 1625 int error; 1626 1627 #ifdef COMPAT_10 /* XXX: and perhaps later */ 1628 if (path == NULL) { 1629 pb = pathbuf_create("."); 1630 if (pb == NULL) 1631 return ENOMEM; 1632 } else 1633 #endif 1634 { 1635 error = pathbuf_copyin(path, &pb); 1636 if (error) 1637 return error; 1638 } 1639 1640 if (fdat != AT_FDCWD) { 1641 /* fd_getvnode() will use the descriptor for us */ 1642 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1643 goto out; 1644 1645 dvp = dfp->f_data; 1646 } 1647 1648 error = do_open(l, dvp, pb, flags, mode, fd); 1649 1650 if (dfp != NULL) 1651 fd_putfile(fdat); 1652 out: 1653 pathbuf_destroy(pb); 1654 return error; 1655 } 1656 1657 int 1658 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1659 { 1660 /* { 1661 syscallarg(const char *) path; 1662 syscallarg(int) flags; 1663 syscallarg(int) mode; 1664 } */ 1665 int error; 1666 int fd; 1667 1668 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1669 SCARG(uap, flags), SCARG(uap, mode), &fd); 1670 1671 if (error == 0) 1672 *retval = fd; 1673 1674 return error; 1675 } 1676 1677 int 1678 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1679 { 1680 /* { 1681 syscallarg(int) fd; 1682 syscallarg(const char *) path; 1683 syscallarg(int) oflags; 1684 syscallarg(int) mode; 1685 } */ 1686 int error; 1687 int fd; 1688 1689 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1690 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1691 1692 if (error == 0) 1693 *retval = fd; 1694 1695 return error; 1696 } 1697 1698 static void 1699 vfs__fhfree(fhandle_t *fhp) 1700 { 1701 size_t fhsize; 1702 1703 if (fhp == NULL) { 1704 return; 1705 } 1706 fhsize = FHANDLE_SIZE(fhp); 1707 kmem_free(fhp, fhsize); 1708 } 1709 1710 /* 1711 * vfs_composefh: compose a filehandle. 1712 */ 1713 1714 int 1715 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1716 { 1717 struct mount *mp; 1718 struct fid *fidp; 1719 int error; 1720 size_t needfhsize; 1721 size_t fidsize; 1722 1723 mp = vp->v_mount; 1724 fidp = NULL; 1725 if (*fh_size < FHANDLE_SIZE_MIN) { 1726 fidsize = 0; 1727 } else { 1728 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1729 if (fhp != NULL) { 1730 memset(fhp, 0, *fh_size); 1731 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1732 fidp = &fhp->fh_fid; 1733 } 1734 } 1735 error = VFS_VPTOFH(vp, fidp, &fidsize); 1736 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1737 if (error == 0 && *fh_size < needfhsize) { 1738 error = E2BIG; 1739 } 1740 *fh_size = needfhsize; 1741 return error; 1742 } 1743 1744 int 1745 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1746 { 1747 struct mount *mp; 1748 fhandle_t *fhp; 1749 size_t fhsize; 1750 size_t fidsize; 1751 int error; 1752 1753 *fhpp = NULL; 1754 mp = vp->v_mount; 1755 fidsize = 0; 1756 error = VFS_VPTOFH(vp, NULL, &fidsize); 1757 KASSERT(error != 0); 1758 if (error != E2BIG) { 1759 goto out; 1760 } 1761 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1762 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1763 if (fhp == NULL) { 1764 error = ENOMEM; 1765 goto out; 1766 } 1767 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1768 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1769 if (error == 0) { 1770 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1771 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1772 *fhpp = fhp; 1773 } else { 1774 kmem_free(fhp, fhsize); 1775 } 1776 out: 1777 return error; 1778 } 1779 1780 void 1781 vfs_composefh_free(fhandle_t *fhp) 1782 { 1783 1784 vfs__fhfree(fhp); 1785 } 1786 1787 /* 1788 * vfs_fhtovp: lookup a vnode by a filehandle. 1789 */ 1790 1791 int 1792 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1793 { 1794 struct mount *mp; 1795 int error; 1796 1797 *vpp = NULL; 1798 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1799 if (mp == NULL) { 1800 error = ESTALE; 1801 goto out; 1802 } 1803 if (mp->mnt_op->vfs_fhtovp == NULL) { 1804 error = EOPNOTSUPP; 1805 goto out; 1806 } 1807 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1808 out: 1809 return error; 1810 } 1811 1812 /* 1813 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1814 * the needed size. 1815 */ 1816 1817 int 1818 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1819 { 1820 fhandle_t *fhp; 1821 int error; 1822 1823 *fhpp = NULL; 1824 if (fhsize > FHANDLE_SIZE_MAX) { 1825 return EINVAL; 1826 } 1827 if (fhsize < FHANDLE_SIZE_MIN) { 1828 return EINVAL; 1829 } 1830 again: 1831 fhp = kmem_alloc(fhsize, KM_SLEEP); 1832 if (fhp == NULL) { 1833 return ENOMEM; 1834 } 1835 error = copyin(ufhp, fhp, fhsize); 1836 if (error == 0) { 1837 /* XXX this check shouldn't be here */ 1838 if (FHANDLE_SIZE(fhp) == fhsize) { 1839 *fhpp = fhp; 1840 return 0; 1841 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1842 /* 1843 * a kludge for nfsv2 padded handles. 1844 */ 1845 size_t sz; 1846 1847 sz = FHANDLE_SIZE(fhp); 1848 kmem_free(fhp, fhsize); 1849 fhsize = sz; 1850 goto again; 1851 } else { 1852 /* 1853 * userland told us wrong size. 1854 */ 1855 error = EINVAL; 1856 } 1857 } 1858 kmem_free(fhp, fhsize); 1859 return error; 1860 } 1861 1862 void 1863 vfs_copyinfh_free(fhandle_t *fhp) 1864 { 1865 1866 vfs__fhfree(fhp); 1867 } 1868 1869 /* 1870 * Get file handle system call 1871 */ 1872 int 1873 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1874 { 1875 /* { 1876 syscallarg(char *) fname; 1877 syscallarg(fhandle_t *) fhp; 1878 syscallarg(size_t *) fh_size; 1879 } */ 1880 struct vnode *vp; 1881 fhandle_t *fh; 1882 int error; 1883 struct pathbuf *pb; 1884 struct nameidata nd; 1885 size_t sz; 1886 size_t usz; 1887 1888 /* 1889 * Must be super user 1890 */ 1891 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1892 0, NULL, NULL, NULL); 1893 if (error) 1894 return (error); 1895 1896 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1897 if (error) { 1898 return error; 1899 } 1900 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1901 error = namei(&nd); 1902 if (error) { 1903 pathbuf_destroy(pb); 1904 return error; 1905 } 1906 vp = nd.ni_vp; 1907 pathbuf_destroy(pb); 1908 1909 error = vfs_composefh_alloc(vp, &fh); 1910 vput(vp); 1911 if (error != 0) { 1912 goto out; 1913 } 1914 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1915 if (error != 0) { 1916 goto out; 1917 } 1918 sz = FHANDLE_SIZE(fh); 1919 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1920 if (error != 0) { 1921 goto out; 1922 } 1923 if (usz >= sz) { 1924 error = copyout(fh, SCARG(uap, fhp), sz); 1925 } else { 1926 error = E2BIG; 1927 } 1928 out: 1929 vfs_composefh_free(fh); 1930 return (error); 1931 } 1932 1933 /* 1934 * Open a file given a file handle. 1935 * 1936 * Check permissions, allocate an open file structure, 1937 * and call the device open routine if any. 1938 */ 1939 1940 int 1941 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1942 register_t *retval) 1943 { 1944 file_t *fp; 1945 struct vnode *vp = NULL; 1946 kauth_cred_t cred = l->l_cred; 1947 file_t *nfp; 1948 int indx, error = 0; 1949 struct vattr va; 1950 fhandle_t *fh; 1951 int flags; 1952 proc_t *p; 1953 1954 p = curproc; 1955 1956 /* 1957 * Must be super user 1958 */ 1959 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1960 0, NULL, NULL, NULL))) 1961 return (error); 1962 1963 if (oflags & O_SEARCH) { 1964 oflags &= ~(int)O_SEARCH; 1965 } 1966 1967 flags = FFLAGS(oflags); 1968 if ((flags & (FREAD | FWRITE)) == 0) 1969 return (EINVAL); 1970 if ((flags & O_CREAT)) 1971 return (EINVAL); 1972 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1973 return (error); 1974 fp = nfp; 1975 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1976 if (error != 0) { 1977 goto bad; 1978 } 1979 error = vfs_fhtovp(fh, &vp); 1980 if (error != 0) { 1981 goto bad; 1982 } 1983 1984 /* Now do an effective vn_open */ 1985 1986 if (vp->v_type == VSOCK) { 1987 error = EOPNOTSUPP; 1988 goto bad; 1989 } 1990 error = vn_openchk(vp, cred, flags); 1991 if (error != 0) 1992 goto bad; 1993 if (flags & O_TRUNC) { 1994 VOP_UNLOCK(vp); /* XXX */ 1995 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1996 vattr_null(&va); 1997 va.va_size = 0; 1998 error = VOP_SETATTR(vp, &va, cred); 1999 if (error) 2000 goto bad; 2001 } 2002 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2003 goto bad; 2004 if (flags & FWRITE) { 2005 mutex_enter(vp->v_interlock); 2006 vp->v_writecount++; 2007 mutex_exit(vp->v_interlock); 2008 } 2009 2010 /* done with modified vn_open, now finish what sys_open does. */ 2011 if ((error = open_setfp(l, fp, vp, indx, flags))) 2012 return error; 2013 2014 VOP_UNLOCK(vp); 2015 *retval = indx; 2016 fd_affix(p, fp, indx); 2017 vfs_copyinfh_free(fh); 2018 return (0); 2019 2020 bad: 2021 fd_abort(p, fp, indx); 2022 if (vp != NULL) 2023 vput(vp); 2024 vfs_copyinfh_free(fh); 2025 return (error); 2026 } 2027 2028 int 2029 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2030 { 2031 /* { 2032 syscallarg(const void *) fhp; 2033 syscallarg(size_t) fh_size; 2034 syscallarg(int) flags; 2035 } */ 2036 2037 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2038 SCARG(uap, flags), retval); 2039 } 2040 2041 int 2042 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2043 { 2044 int error; 2045 fhandle_t *fh; 2046 struct vnode *vp; 2047 2048 /* 2049 * Must be super user 2050 */ 2051 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2052 0, NULL, NULL, NULL))) 2053 return (error); 2054 2055 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2056 if (error != 0) 2057 return error; 2058 2059 error = vfs_fhtovp(fh, &vp); 2060 vfs_copyinfh_free(fh); 2061 if (error != 0) 2062 return error; 2063 2064 error = vn_stat(vp, sb); 2065 vput(vp); 2066 return error; 2067 } 2068 2069 2070 /* ARGSUSED */ 2071 int 2072 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2073 { 2074 /* { 2075 syscallarg(const void *) fhp; 2076 syscallarg(size_t) fh_size; 2077 syscallarg(struct stat *) sb; 2078 } */ 2079 struct stat sb; 2080 int error; 2081 2082 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2083 if (error) 2084 return error; 2085 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2086 } 2087 2088 int 2089 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2090 int flags) 2091 { 2092 fhandle_t *fh; 2093 struct mount *mp; 2094 struct vnode *vp; 2095 int error; 2096 2097 /* 2098 * Must be super user 2099 */ 2100 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2101 0, NULL, NULL, NULL))) 2102 return error; 2103 2104 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2105 if (error != 0) 2106 return error; 2107 2108 error = vfs_fhtovp(fh, &vp); 2109 vfs_copyinfh_free(fh); 2110 if (error != 0) 2111 return error; 2112 2113 mp = vp->v_mount; 2114 error = dostatvfs(mp, sb, l, flags, 1); 2115 vput(vp); 2116 return error; 2117 } 2118 2119 /* ARGSUSED */ 2120 int 2121 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2122 { 2123 /* { 2124 syscallarg(const void *) fhp; 2125 syscallarg(size_t) fh_size; 2126 syscallarg(struct statvfs *) buf; 2127 syscallarg(int) flags; 2128 } */ 2129 struct statvfs *sb = STATVFSBUF_GET(); 2130 int error; 2131 2132 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2133 SCARG(uap, flags)); 2134 if (error == 0) 2135 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2136 STATVFSBUF_PUT(sb); 2137 return error; 2138 } 2139 2140 /* 2141 * Create a special file. 2142 */ 2143 /* ARGSUSED */ 2144 int 2145 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2146 register_t *retval) 2147 { 2148 /* { 2149 syscallarg(const char *) path; 2150 syscallarg(mode_t) mode; 2151 syscallarg(dev_t) dev; 2152 } */ 2153 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2154 SCARG(uap, dev), retval, UIO_USERSPACE); 2155 } 2156 2157 int 2158 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2159 register_t *retval) 2160 { 2161 /* { 2162 syscallarg(int) fd; 2163 syscallarg(const char *) path; 2164 syscallarg(mode_t) mode; 2165 syscallarg(int) pad; 2166 syscallarg(dev_t) dev; 2167 } */ 2168 2169 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2170 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2171 } 2172 2173 int 2174 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2175 register_t *retval, enum uio_seg seg) 2176 { 2177 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2178 } 2179 2180 int 2181 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2182 dev_t dev, register_t *retval, enum uio_seg seg) 2183 { 2184 struct proc *p = l->l_proc; 2185 struct vnode *vp; 2186 struct vattr vattr; 2187 int error, optype; 2188 struct pathbuf *pb; 2189 struct nameidata nd; 2190 const char *pathstring; 2191 2192 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2193 0, NULL, NULL, NULL)) != 0) 2194 return (error); 2195 2196 optype = VOP_MKNOD_DESCOFFSET; 2197 2198 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2199 if (error) { 2200 return error; 2201 } 2202 pathstring = pathbuf_stringcopy_get(pb); 2203 if (pathstring == NULL) { 2204 pathbuf_destroy(pb); 2205 return ENOMEM; 2206 } 2207 2208 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2209 2210 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2211 goto out; 2212 vp = nd.ni_vp; 2213 2214 if (vp != NULL) 2215 error = EEXIST; 2216 else { 2217 vattr_null(&vattr); 2218 /* We will read cwdi->cwdi_cmask unlocked. */ 2219 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2220 vattr.va_rdev = dev; 2221 2222 switch (mode & S_IFMT) { 2223 case S_IFMT: /* used by badsect to flag bad sectors */ 2224 vattr.va_type = VBAD; 2225 break; 2226 case S_IFCHR: 2227 vattr.va_type = VCHR; 2228 break; 2229 case S_IFBLK: 2230 vattr.va_type = VBLK; 2231 break; 2232 case S_IFWHT: 2233 optype = VOP_WHITEOUT_DESCOFFSET; 2234 break; 2235 case S_IFREG: 2236 #if NVERIEXEC > 0 2237 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2238 O_CREAT); 2239 #endif /* NVERIEXEC > 0 */ 2240 vattr.va_type = VREG; 2241 vattr.va_rdev = VNOVAL; 2242 optype = VOP_CREATE_DESCOFFSET; 2243 break; 2244 default: 2245 error = EINVAL; 2246 break; 2247 } 2248 } 2249 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2250 && vattr.va_rdev == VNOVAL) 2251 error = EINVAL; 2252 if (!error) { 2253 switch (optype) { 2254 case VOP_WHITEOUT_DESCOFFSET: 2255 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2256 if (error) 2257 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2258 vput(nd.ni_dvp); 2259 break; 2260 2261 case VOP_MKNOD_DESCOFFSET: 2262 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2263 &nd.ni_cnd, &vattr); 2264 if (error == 0) 2265 vrele(nd.ni_vp); 2266 vput(nd.ni_dvp); 2267 break; 2268 2269 case VOP_CREATE_DESCOFFSET: 2270 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2271 &nd.ni_cnd, &vattr); 2272 if (error == 0) 2273 vrele(nd.ni_vp); 2274 vput(nd.ni_dvp); 2275 break; 2276 } 2277 } else { 2278 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2279 if (nd.ni_dvp == vp) 2280 vrele(nd.ni_dvp); 2281 else 2282 vput(nd.ni_dvp); 2283 if (vp) 2284 vrele(vp); 2285 } 2286 out: 2287 pathbuf_stringcopy_put(pb, pathstring); 2288 pathbuf_destroy(pb); 2289 return (error); 2290 } 2291 2292 /* 2293 * Create a named pipe. 2294 */ 2295 /* ARGSUSED */ 2296 int 2297 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2298 { 2299 /* { 2300 syscallarg(const char *) path; 2301 syscallarg(int) mode; 2302 } */ 2303 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2304 } 2305 2306 int 2307 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2308 register_t *retval) 2309 { 2310 /* { 2311 syscallarg(int) fd; 2312 syscallarg(const char *) path; 2313 syscallarg(int) mode; 2314 } */ 2315 2316 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2317 SCARG(uap, mode)); 2318 } 2319 2320 static int 2321 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2322 { 2323 struct proc *p = l->l_proc; 2324 struct vattr vattr; 2325 int error; 2326 struct pathbuf *pb; 2327 struct nameidata nd; 2328 2329 error = pathbuf_copyin(path, &pb); 2330 if (error) { 2331 return error; 2332 } 2333 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2334 2335 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2336 pathbuf_destroy(pb); 2337 return error; 2338 } 2339 if (nd.ni_vp != NULL) { 2340 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2341 if (nd.ni_dvp == nd.ni_vp) 2342 vrele(nd.ni_dvp); 2343 else 2344 vput(nd.ni_dvp); 2345 vrele(nd.ni_vp); 2346 pathbuf_destroy(pb); 2347 return (EEXIST); 2348 } 2349 vattr_null(&vattr); 2350 vattr.va_type = VFIFO; 2351 /* We will read cwdi->cwdi_cmask unlocked. */ 2352 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2353 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2354 if (error == 0) 2355 vrele(nd.ni_vp); 2356 vput(nd.ni_dvp); 2357 pathbuf_destroy(pb); 2358 return (error); 2359 } 2360 2361 /* 2362 * Make a hard file link. 2363 */ 2364 /* ARGSUSED */ 2365 int 2366 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2367 const char *link, int follow, register_t *retval) 2368 { 2369 struct vnode *vp; 2370 struct pathbuf *linkpb; 2371 struct nameidata nd; 2372 namei_simple_flags_t ns_flags; 2373 int error; 2374 2375 if (follow & AT_SYMLINK_FOLLOW) 2376 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2377 else 2378 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2379 2380 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2381 if (error != 0) 2382 return (error); 2383 error = pathbuf_copyin(link, &linkpb); 2384 if (error) { 2385 goto out1; 2386 } 2387 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2388 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2389 goto out2; 2390 if (nd.ni_vp) { 2391 error = EEXIST; 2392 goto abortop; 2393 } 2394 /* Prevent hard links on directories. */ 2395 if (vp->v_type == VDIR) { 2396 error = EPERM; 2397 goto abortop; 2398 } 2399 /* Prevent cross-mount operation. */ 2400 if (nd.ni_dvp->v_mount != vp->v_mount) { 2401 error = EXDEV; 2402 goto abortop; 2403 } 2404 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2405 out2: 2406 pathbuf_destroy(linkpb); 2407 out1: 2408 vrele(vp); 2409 return (error); 2410 abortop: 2411 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2412 if (nd.ni_dvp == nd.ni_vp) 2413 vrele(nd.ni_dvp); 2414 else 2415 vput(nd.ni_dvp); 2416 if (nd.ni_vp != NULL) 2417 vrele(nd.ni_vp); 2418 goto out2; 2419 } 2420 2421 int 2422 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2423 { 2424 /* { 2425 syscallarg(const char *) path; 2426 syscallarg(const char *) link; 2427 } */ 2428 const char *path = SCARG(uap, path); 2429 const char *link = SCARG(uap, link); 2430 2431 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2432 AT_SYMLINK_FOLLOW, retval); 2433 } 2434 2435 int 2436 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2437 register_t *retval) 2438 { 2439 /* { 2440 syscallarg(int) fd1; 2441 syscallarg(const char *) name1; 2442 syscallarg(int) fd2; 2443 syscallarg(const char *) name2; 2444 syscallarg(int) flags; 2445 } */ 2446 int fd1 = SCARG(uap, fd1); 2447 const char *name1 = SCARG(uap, name1); 2448 int fd2 = SCARG(uap, fd2); 2449 const char *name2 = SCARG(uap, name2); 2450 int follow; 2451 2452 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2453 2454 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2455 } 2456 2457 2458 int 2459 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2460 { 2461 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2462 } 2463 2464 static int 2465 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2466 const char *link, enum uio_seg seg) 2467 { 2468 struct proc *p = curproc; 2469 struct vattr vattr; 2470 char *path; 2471 int error; 2472 struct pathbuf *linkpb; 2473 struct nameidata nd; 2474 2475 KASSERT(l != NULL || fdat == AT_FDCWD); 2476 2477 path = PNBUF_GET(); 2478 if (seg == UIO_USERSPACE) { 2479 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2480 goto out1; 2481 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2482 goto out1; 2483 } else { 2484 KASSERT(strlen(patharg) < MAXPATHLEN); 2485 strcpy(path, patharg); 2486 linkpb = pathbuf_create(link); 2487 if (linkpb == NULL) { 2488 error = ENOMEM; 2489 goto out1; 2490 } 2491 } 2492 ktrkuser("symlink-target", path, strlen(path)); 2493 2494 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2495 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2496 goto out2; 2497 if (nd.ni_vp) { 2498 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2499 if (nd.ni_dvp == nd.ni_vp) 2500 vrele(nd.ni_dvp); 2501 else 2502 vput(nd.ni_dvp); 2503 vrele(nd.ni_vp); 2504 error = EEXIST; 2505 goto out2; 2506 } 2507 vattr_null(&vattr); 2508 vattr.va_type = VLNK; 2509 /* We will read cwdi->cwdi_cmask unlocked. */ 2510 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2511 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2512 if (error == 0) 2513 vrele(nd.ni_vp); 2514 vput(nd.ni_dvp); 2515 out2: 2516 pathbuf_destroy(linkpb); 2517 out1: 2518 PNBUF_PUT(path); 2519 return (error); 2520 } 2521 2522 /* 2523 * Make a symbolic link. 2524 */ 2525 /* ARGSUSED */ 2526 int 2527 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2528 { 2529 /* { 2530 syscallarg(const char *) path; 2531 syscallarg(const char *) link; 2532 } */ 2533 2534 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2535 UIO_USERSPACE); 2536 } 2537 2538 int 2539 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2540 register_t *retval) 2541 { 2542 /* { 2543 syscallarg(const char *) path1; 2544 syscallarg(int) fd; 2545 syscallarg(const char *) path2; 2546 } */ 2547 2548 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2549 SCARG(uap, path2), UIO_USERSPACE); 2550 } 2551 2552 /* 2553 * Delete a whiteout from the filesystem. 2554 */ 2555 /* ARGSUSED */ 2556 int 2557 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2558 { 2559 /* { 2560 syscallarg(const char *) path; 2561 } */ 2562 int error; 2563 struct pathbuf *pb; 2564 struct nameidata nd; 2565 2566 error = pathbuf_copyin(SCARG(uap, path), &pb); 2567 if (error) { 2568 return error; 2569 } 2570 2571 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2572 error = namei(&nd); 2573 if (error) { 2574 pathbuf_destroy(pb); 2575 return (error); 2576 } 2577 2578 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2579 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2580 if (nd.ni_dvp == nd.ni_vp) 2581 vrele(nd.ni_dvp); 2582 else 2583 vput(nd.ni_dvp); 2584 if (nd.ni_vp) 2585 vrele(nd.ni_vp); 2586 pathbuf_destroy(pb); 2587 return (EEXIST); 2588 } 2589 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2590 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2591 vput(nd.ni_dvp); 2592 pathbuf_destroy(pb); 2593 return (error); 2594 } 2595 2596 /* 2597 * Delete a name from the filesystem. 2598 */ 2599 /* ARGSUSED */ 2600 int 2601 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2602 { 2603 /* { 2604 syscallarg(const char *) path; 2605 } */ 2606 2607 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2608 } 2609 2610 int 2611 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2612 register_t *retval) 2613 { 2614 /* { 2615 syscallarg(int) fd; 2616 syscallarg(const char *) path; 2617 syscallarg(int) flag; 2618 } */ 2619 2620 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2621 SCARG(uap, flag), UIO_USERSPACE); 2622 } 2623 2624 int 2625 do_sys_unlink(const char *arg, enum uio_seg seg) 2626 { 2627 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2628 } 2629 2630 static int 2631 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2632 enum uio_seg seg) 2633 { 2634 struct vnode *vp; 2635 int error; 2636 struct pathbuf *pb; 2637 struct nameidata nd; 2638 const char *pathstring; 2639 2640 KASSERT(l != NULL || fdat == AT_FDCWD); 2641 2642 error = pathbuf_maybe_copyin(arg, seg, &pb); 2643 if (error) { 2644 return error; 2645 } 2646 pathstring = pathbuf_stringcopy_get(pb); 2647 if (pathstring == NULL) { 2648 pathbuf_destroy(pb); 2649 return ENOMEM; 2650 } 2651 2652 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2653 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2654 goto out; 2655 vp = nd.ni_vp; 2656 2657 /* 2658 * The root of a mounted filesystem cannot be deleted. 2659 */ 2660 if ((vp->v_vflag & VV_ROOT) != 0) { 2661 error = EBUSY; 2662 goto abort; 2663 } 2664 2665 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2666 error = EBUSY; 2667 goto abort; 2668 } 2669 2670 /* 2671 * No rmdir "." please. 2672 */ 2673 if (nd.ni_dvp == vp) { 2674 error = EINVAL; 2675 goto abort; 2676 } 2677 2678 /* 2679 * AT_REMOVEDIR is required to remove a directory 2680 */ 2681 if (vp->v_type == VDIR) { 2682 if (!(flags & AT_REMOVEDIR)) { 2683 error = EPERM; 2684 goto abort; 2685 } else { 2686 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2687 goto out; 2688 } 2689 } 2690 2691 /* 2692 * Starting here we only deal with non directories. 2693 */ 2694 if (flags & AT_REMOVEDIR) { 2695 error = ENOTDIR; 2696 goto abort; 2697 } 2698 2699 2700 #if NVERIEXEC > 0 2701 /* Handle remove requests for veriexec entries. */ 2702 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2703 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2704 if (nd.ni_dvp == vp) 2705 vrele(nd.ni_dvp); 2706 else 2707 vput(nd.ni_dvp); 2708 vput(vp); 2709 goto out; 2710 } 2711 #endif /* NVERIEXEC > 0 */ 2712 2713 #ifdef FILEASSOC 2714 (void)fileassoc_file_delete(vp); 2715 #endif /* FILEASSOC */ 2716 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2717 goto out; 2718 2719 abort: 2720 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2721 if (nd.ni_dvp == vp) 2722 vrele(nd.ni_dvp); 2723 else 2724 vput(nd.ni_dvp); 2725 vput(vp); 2726 2727 out: 2728 pathbuf_stringcopy_put(pb, pathstring); 2729 pathbuf_destroy(pb); 2730 return (error); 2731 } 2732 2733 /* 2734 * Reposition read/write file offset. 2735 */ 2736 int 2737 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2738 { 2739 /* { 2740 syscallarg(int) fd; 2741 syscallarg(int) pad; 2742 syscallarg(off_t) offset; 2743 syscallarg(int) whence; 2744 } */ 2745 kauth_cred_t cred = l->l_cred; 2746 file_t *fp; 2747 struct vnode *vp; 2748 struct vattr vattr; 2749 off_t newoff; 2750 int error, fd; 2751 2752 fd = SCARG(uap, fd); 2753 2754 if ((fp = fd_getfile(fd)) == NULL) 2755 return (EBADF); 2756 2757 vp = fp->f_data; 2758 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2759 error = ESPIPE; 2760 goto out; 2761 } 2762 2763 switch (SCARG(uap, whence)) { 2764 case SEEK_CUR: 2765 newoff = fp->f_offset + SCARG(uap, offset); 2766 break; 2767 case SEEK_END: 2768 vn_lock(vp, LK_SHARED | LK_RETRY); 2769 error = VOP_GETATTR(vp, &vattr, cred); 2770 VOP_UNLOCK(vp); 2771 if (error) { 2772 goto out; 2773 } 2774 newoff = SCARG(uap, offset) + vattr.va_size; 2775 break; 2776 case SEEK_SET: 2777 newoff = SCARG(uap, offset); 2778 break; 2779 default: 2780 error = EINVAL; 2781 goto out; 2782 } 2783 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2784 *(off_t *)retval = fp->f_offset = newoff; 2785 } 2786 out: 2787 fd_putfile(fd); 2788 return (error); 2789 } 2790 2791 /* 2792 * Positional read system call. 2793 */ 2794 int 2795 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2796 { 2797 /* { 2798 syscallarg(int) fd; 2799 syscallarg(void *) buf; 2800 syscallarg(size_t) nbyte; 2801 syscallarg(off_t) offset; 2802 } */ 2803 file_t *fp; 2804 struct vnode *vp; 2805 off_t offset; 2806 int error, fd = SCARG(uap, fd); 2807 2808 if ((fp = fd_getfile(fd)) == NULL) 2809 return (EBADF); 2810 2811 if ((fp->f_flag & FREAD) == 0) { 2812 fd_putfile(fd); 2813 return (EBADF); 2814 } 2815 2816 vp = fp->f_data; 2817 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2818 error = ESPIPE; 2819 goto out; 2820 } 2821 2822 offset = SCARG(uap, offset); 2823 2824 /* 2825 * XXX This works because no file systems actually 2826 * XXX take any action on the seek operation. 2827 */ 2828 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2829 goto out; 2830 2831 /* dofileread() will unuse the descriptor for us */ 2832 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2833 &offset, 0, retval)); 2834 2835 out: 2836 fd_putfile(fd); 2837 return (error); 2838 } 2839 2840 /* 2841 * Positional scatter read system call. 2842 */ 2843 int 2844 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2845 { 2846 /* { 2847 syscallarg(int) fd; 2848 syscallarg(const struct iovec *) iovp; 2849 syscallarg(int) iovcnt; 2850 syscallarg(off_t) offset; 2851 } */ 2852 off_t offset = SCARG(uap, offset); 2853 2854 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2855 SCARG(uap, iovcnt), &offset, 0, retval); 2856 } 2857 2858 /* 2859 * Positional write system call. 2860 */ 2861 int 2862 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2863 { 2864 /* { 2865 syscallarg(int) fd; 2866 syscallarg(const void *) buf; 2867 syscallarg(size_t) nbyte; 2868 syscallarg(off_t) offset; 2869 } */ 2870 file_t *fp; 2871 struct vnode *vp; 2872 off_t offset; 2873 int error, fd = SCARG(uap, fd); 2874 2875 if ((fp = fd_getfile(fd)) == NULL) 2876 return (EBADF); 2877 2878 if ((fp->f_flag & FWRITE) == 0) { 2879 fd_putfile(fd); 2880 return (EBADF); 2881 } 2882 2883 vp = fp->f_data; 2884 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2885 error = ESPIPE; 2886 goto out; 2887 } 2888 2889 offset = SCARG(uap, offset); 2890 2891 /* 2892 * XXX This works because no file systems actually 2893 * XXX take any action on the seek operation. 2894 */ 2895 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2896 goto out; 2897 2898 /* dofilewrite() will unuse the descriptor for us */ 2899 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2900 &offset, 0, retval)); 2901 2902 out: 2903 fd_putfile(fd); 2904 return (error); 2905 } 2906 2907 /* 2908 * Positional gather write system call. 2909 */ 2910 int 2911 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2912 { 2913 /* { 2914 syscallarg(int) fd; 2915 syscallarg(const struct iovec *) iovp; 2916 syscallarg(int) iovcnt; 2917 syscallarg(off_t) offset; 2918 } */ 2919 off_t offset = SCARG(uap, offset); 2920 2921 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2922 SCARG(uap, iovcnt), &offset, 0, retval); 2923 } 2924 2925 /* 2926 * Check access permissions. 2927 */ 2928 int 2929 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2930 { 2931 /* { 2932 syscallarg(const char *) path; 2933 syscallarg(int) flags; 2934 } */ 2935 2936 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2937 SCARG(uap, flags), 0); 2938 } 2939 2940 int 2941 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2942 int mode, int flags) 2943 { 2944 kauth_cred_t cred; 2945 struct vnode *vp; 2946 int error, nd_flag, vmode; 2947 struct pathbuf *pb; 2948 struct nameidata nd; 2949 2950 CTASSERT(F_OK == 0); 2951 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2952 /* nonsense mode */ 2953 return EINVAL; 2954 } 2955 2956 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2957 if (flags & AT_SYMLINK_NOFOLLOW) 2958 nd_flag &= ~FOLLOW; 2959 2960 error = pathbuf_copyin(path, &pb); 2961 if (error) 2962 return error; 2963 2964 NDINIT(&nd, LOOKUP, nd_flag, pb); 2965 2966 /* Override default credentials */ 2967 cred = kauth_cred_dup(l->l_cred); 2968 if (!(flags & AT_EACCESS)) { 2969 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2970 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2971 } 2972 nd.ni_cnd.cn_cred = cred; 2973 2974 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2975 pathbuf_destroy(pb); 2976 goto out; 2977 } 2978 vp = nd.ni_vp; 2979 pathbuf_destroy(pb); 2980 2981 /* Flags == 0 means only check for existence. */ 2982 if (mode) { 2983 vmode = 0; 2984 if (mode & R_OK) 2985 vmode |= VREAD; 2986 if (mode & W_OK) 2987 vmode |= VWRITE; 2988 if (mode & X_OK) 2989 vmode |= VEXEC; 2990 2991 error = VOP_ACCESS(vp, vmode, cred); 2992 if (!error && (vmode & VWRITE)) 2993 error = vn_writechk(vp); 2994 } 2995 vput(vp); 2996 out: 2997 kauth_cred_free(cred); 2998 return (error); 2999 } 3000 3001 int 3002 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3003 register_t *retval) 3004 { 3005 /* { 3006 syscallarg(int) fd; 3007 syscallarg(const char *) path; 3008 syscallarg(int) amode; 3009 syscallarg(int) flag; 3010 } */ 3011 3012 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3013 SCARG(uap, amode), SCARG(uap, flag)); 3014 } 3015 3016 /* 3017 * Common code for all sys_stat functions, including compat versions. 3018 */ 3019 int 3020 do_sys_stat(const char *userpath, unsigned int nd_flag, 3021 struct stat *sb) 3022 { 3023 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3024 } 3025 3026 int 3027 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3028 unsigned int nd_flag, struct stat *sb) 3029 { 3030 int error; 3031 struct pathbuf *pb; 3032 struct nameidata nd; 3033 3034 KASSERT(l != NULL || fdat == AT_FDCWD); 3035 3036 error = pathbuf_copyin(userpath, &pb); 3037 if (error) { 3038 return error; 3039 } 3040 3041 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3042 3043 error = fd_nameiat(l, fdat, &nd); 3044 if (error != 0) { 3045 pathbuf_destroy(pb); 3046 return error; 3047 } 3048 error = vn_stat(nd.ni_vp, sb); 3049 vput(nd.ni_vp); 3050 pathbuf_destroy(pb); 3051 return error; 3052 } 3053 3054 /* 3055 * Get file status; this version follows links. 3056 */ 3057 /* ARGSUSED */ 3058 int 3059 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3060 { 3061 /* { 3062 syscallarg(const char *) path; 3063 syscallarg(struct stat *) ub; 3064 } */ 3065 struct stat sb; 3066 int error; 3067 3068 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3069 if (error) 3070 return error; 3071 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3072 } 3073 3074 /* 3075 * Get file status; this version does not follow links. 3076 */ 3077 /* ARGSUSED */ 3078 int 3079 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3080 { 3081 /* { 3082 syscallarg(const char *) path; 3083 syscallarg(struct stat *) ub; 3084 } */ 3085 struct stat sb; 3086 int error; 3087 3088 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3089 if (error) 3090 return error; 3091 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3092 } 3093 3094 int 3095 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3096 register_t *retval) 3097 { 3098 /* { 3099 syscallarg(int) fd; 3100 syscallarg(const char *) path; 3101 syscallarg(struct stat *) buf; 3102 syscallarg(int) flag; 3103 } */ 3104 unsigned int nd_flag; 3105 struct stat sb; 3106 int error; 3107 3108 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3109 nd_flag = NOFOLLOW; 3110 else 3111 nd_flag = FOLLOW; 3112 3113 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3114 &sb); 3115 if (error) 3116 return error; 3117 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3118 } 3119 3120 /* 3121 * Get configurable pathname variables. 3122 */ 3123 /* ARGSUSED */ 3124 int 3125 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3126 { 3127 /* { 3128 syscallarg(const char *) path; 3129 syscallarg(int) name; 3130 } */ 3131 int error; 3132 struct pathbuf *pb; 3133 struct nameidata nd; 3134 3135 error = pathbuf_copyin(SCARG(uap, path), &pb); 3136 if (error) { 3137 return error; 3138 } 3139 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3140 if ((error = namei(&nd)) != 0) { 3141 pathbuf_destroy(pb); 3142 return (error); 3143 } 3144 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3145 vput(nd.ni_vp); 3146 pathbuf_destroy(pb); 3147 return (error); 3148 } 3149 3150 /* 3151 * Return target name of a symbolic link. 3152 */ 3153 /* ARGSUSED */ 3154 int 3155 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3156 register_t *retval) 3157 { 3158 /* { 3159 syscallarg(const char *) path; 3160 syscallarg(char *) buf; 3161 syscallarg(size_t) count; 3162 } */ 3163 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3164 SCARG(uap, buf), SCARG(uap, count), retval); 3165 } 3166 3167 static int 3168 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3169 size_t count, register_t *retval) 3170 { 3171 struct vnode *vp; 3172 struct iovec aiov; 3173 struct uio auio; 3174 int error; 3175 struct pathbuf *pb; 3176 struct nameidata nd; 3177 3178 error = pathbuf_copyin(path, &pb); 3179 if (error) { 3180 return error; 3181 } 3182 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3183 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3184 pathbuf_destroy(pb); 3185 return error; 3186 } 3187 vp = nd.ni_vp; 3188 pathbuf_destroy(pb); 3189 if (vp->v_type != VLNK) 3190 error = EINVAL; 3191 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3192 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3193 aiov.iov_base = buf; 3194 aiov.iov_len = count; 3195 auio.uio_iov = &aiov; 3196 auio.uio_iovcnt = 1; 3197 auio.uio_offset = 0; 3198 auio.uio_rw = UIO_READ; 3199 KASSERT(l == curlwp); 3200 auio.uio_vmspace = l->l_proc->p_vmspace; 3201 auio.uio_resid = count; 3202 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3203 *retval = count - auio.uio_resid; 3204 } 3205 vput(vp); 3206 return (error); 3207 } 3208 3209 int 3210 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3211 register_t *retval) 3212 { 3213 /* { 3214 syscallarg(int) fd; 3215 syscallarg(const char *) path; 3216 syscallarg(char *) buf; 3217 syscallarg(size_t) bufsize; 3218 } */ 3219 3220 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3221 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3222 } 3223 3224 /* 3225 * Change flags of a file given a path name. 3226 */ 3227 /* ARGSUSED */ 3228 int 3229 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3230 { 3231 /* { 3232 syscallarg(const char *) path; 3233 syscallarg(u_long) flags; 3234 } */ 3235 struct vnode *vp; 3236 int error; 3237 3238 error = namei_simple_user(SCARG(uap, path), 3239 NSM_FOLLOW_TRYEMULROOT, &vp); 3240 if (error != 0) 3241 return (error); 3242 error = change_flags(vp, SCARG(uap, flags), l); 3243 vput(vp); 3244 return (error); 3245 } 3246 3247 /* 3248 * Change flags of a file given a file descriptor. 3249 */ 3250 /* ARGSUSED */ 3251 int 3252 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3253 { 3254 /* { 3255 syscallarg(int) fd; 3256 syscallarg(u_long) flags; 3257 } */ 3258 struct vnode *vp; 3259 file_t *fp; 3260 int error; 3261 3262 /* fd_getvnode() will use the descriptor for us */ 3263 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3264 return (error); 3265 vp = fp->f_data; 3266 error = change_flags(vp, SCARG(uap, flags), l); 3267 VOP_UNLOCK(vp); 3268 fd_putfile(SCARG(uap, fd)); 3269 return (error); 3270 } 3271 3272 /* 3273 * Change flags of a file given a path name; this version does 3274 * not follow links. 3275 */ 3276 int 3277 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3278 { 3279 /* { 3280 syscallarg(const char *) path; 3281 syscallarg(u_long) flags; 3282 } */ 3283 struct vnode *vp; 3284 int error; 3285 3286 error = namei_simple_user(SCARG(uap, path), 3287 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3288 if (error != 0) 3289 return (error); 3290 error = change_flags(vp, SCARG(uap, flags), l); 3291 vput(vp); 3292 return (error); 3293 } 3294 3295 /* 3296 * Common routine to change flags of a file. 3297 */ 3298 int 3299 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3300 { 3301 struct vattr vattr; 3302 int error; 3303 3304 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3305 3306 vattr_null(&vattr); 3307 vattr.va_flags = flags; 3308 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3309 3310 return (error); 3311 } 3312 3313 /* 3314 * Change mode of a file given path name; this version follows links. 3315 */ 3316 /* ARGSUSED */ 3317 int 3318 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3319 { 3320 /* { 3321 syscallarg(const char *) path; 3322 syscallarg(int) mode; 3323 } */ 3324 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3325 SCARG(uap, mode), 0); 3326 } 3327 3328 int 3329 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3330 { 3331 int error; 3332 struct vnode *vp; 3333 namei_simple_flags_t ns_flag; 3334 3335 if (flags & AT_SYMLINK_NOFOLLOW) 3336 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3337 else 3338 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3339 3340 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3341 if (error != 0) 3342 return error; 3343 3344 error = change_mode(vp, mode, l); 3345 3346 vrele(vp); 3347 3348 return (error); 3349 } 3350 3351 /* 3352 * Change mode of a file given a file descriptor. 3353 */ 3354 /* ARGSUSED */ 3355 int 3356 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3357 { 3358 /* { 3359 syscallarg(int) fd; 3360 syscallarg(int) mode; 3361 } */ 3362 file_t *fp; 3363 int error; 3364 3365 /* fd_getvnode() will use the descriptor for us */ 3366 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3367 return (error); 3368 error = change_mode(fp->f_data, SCARG(uap, mode), l); 3369 fd_putfile(SCARG(uap, fd)); 3370 return (error); 3371 } 3372 3373 int 3374 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3375 register_t *retval) 3376 { 3377 /* { 3378 syscallarg(int) fd; 3379 syscallarg(const char *) path; 3380 syscallarg(int) mode; 3381 syscallarg(int) flag; 3382 } */ 3383 3384 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3385 SCARG(uap, mode), SCARG(uap, flag)); 3386 } 3387 3388 /* 3389 * Change mode of a file given path name; this version does not follow links. 3390 */ 3391 /* ARGSUSED */ 3392 int 3393 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3394 { 3395 /* { 3396 syscallarg(const char *) path; 3397 syscallarg(int) mode; 3398 } */ 3399 int error; 3400 struct vnode *vp; 3401 3402 error = namei_simple_user(SCARG(uap, path), 3403 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3404 if (error != 0) 3405 return (error); 3406 3407 error = change_mode(vp, SCARG(uap, mode), l); 3408 3409 vrele(vp); 3410 return (error); 3411 } 3412 3413 /* 3414 * Common routine to set mode given a vnode. 3415 */ 3416 static int 3417 change_mode(struct vnode *vp, int mode, struct lwp *l) 3418 { 3419 struct vattr vattr; 3420 int error; 3421 3422 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3423 vattr_null(&vattr); 3424 vattr.va_mode = mode & ALLPERMS; 3425 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3426 VOP_UNLOCK(vp); 3427 return (error); 3428 } 3429 3430 /* 3431 * Set ownership given a path name; this version follows links. 3432 */ 3433 /* ARGSUSED */ 3434 int 3435 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3436 { 3437 /* { 3438 syscallarg(const char *) path; 3439 syscallarg(uid_t) uid; 3440 syscallarg(gid_t) gid; 3441 } */ 3442 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3443 SCARG(uap, gid), 0); 3444 } 3445 3446 int 3447 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3448 gid_t gid, int flags) 3449 { 3450 int error; 3451 struct vnode *vp; 3452 namei_simple_flags_t ns_flag; 3453 3454 if (flags & AT_SYMLINK_NOFOLLOW) 3455 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3456 else 3457 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3458 3459 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3460 if (error != 0) 3461 return error; 3462 3463 error = change_owner(vp, uid, gid, l, 0); 3464 3465 vrele(vp); 3466 3467 return (error); 3468 } 3469 3470 /* 3471 * Set ownership given a path name; this version follows links. 3472 * Provides POSIX semantics. 3473 */ 3474 /* ARGSUSED */ 3475 int 3476 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3477 { 3478 /* { 3479 syscallarg(const char *) path; 3480 syscallarg(uid_t) uid; 3481 syscallarg(gid_t) gid; 3482 } */ 3483 int error; 3484 struct vnode *vp; 3485 3486 error = namei_simple_user(SCARG(uap, path), 3487 NSM_FOLLOW_TRYEMULROOT, &vp); 3488 if (error != 0) 3489 return (error); 3490 3491 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3492 3493 vrele(vp); 3494 return (error); 3495 } 3496 3497 /* 3498 * Set ownership given a file descriptor. 3499 */ 3500 /* ARGSUSED */ 3501 int 3502 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3503 { 3504 /* { 3505 syscallarg(int) fd; 3506 syscallarg(uid_t) uid; 3507 syscallarg(gid_t) gid; 3508 } */ 3509 int error; 3510 file_t *fp; 3511 3512 /* fd_getvnode() will use the descriptor for us */ 3513 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3514 return (error); 3515 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3516 l, 0); 3517 fd_putfile(SCARG(uap, fd)); 3518 return (error); 3519 } 3520 3521 int 3522 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3523 register_t *retval) 3524 { 3525 /* { 3526 syscallarg(int) fd; 3527 syscallarg(const char *) path; 3528 syscallarg(uid_t) owner; 3529 syscallarg(gid_t) group; 3530 syscallarg(int) flag; 3531 } */ 3532 3533 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3534 SCARG(uap, owner), SCARG(uap, group), 3535 SCARG(uap, flag)); 3536 } 3537 3538 /* 3539 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3540 */ 3541 /* ARGSUSED */ 3542 int 3543 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3544 { 3545 /* { 3546 syscallarg(int) fd; 3547 syscallarg(uid_t) uid; 3548 syscallarg(gid_t) gid; 3549 } */ 3550 int error; 3551 file_t *fp; 3552 3553 /* fd_getvnode() will use the descriptor for us */ 3554 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3555 return (error); 3556 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3557 l, 1); 3558 fd_putfile(SCARG(uap, fd)); 3559 return (error); 3560 } 3561 3562 /* 3563 * Set ownership given a path name; this version does not follow links. 3564 */ 3565 /* ARGSUSED */ 3566 int 3567 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3568 { 3569 /* { 3570 syscallarg(const char *) path; 3571 syscallarg(uid_t) uid; 3572 syscallarg(gid_t) gid; 3573 } */ 3574 int error; 3575 struct vnode *vp; 3576 3577 error = namei_simple_user(SCARG(uap, path), 3578 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3579 if (error != 0) 3580 return (error); 3581 3582 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3583 3584 vrele(vp); 3585 return (error); 3586 } 3587 3588 /* 3589 * Set ownership given a path name; this version does not follow links. 3590 * Provides POSIX/XPG semantics. 3591 */ 3592 /* ARGSUSED */ 3593 int 3594 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3595 { 3596 /* { 3597 syscallarg(const char *) path; 3598 syscallarg(uid_t) uid; 3599 syscallarg(gid_t) gid; 3600 } */ 3601 int error; 3602 struct vnode *vp; 3603 3604 error = namei_simple_user(SCARG(uap, path), 3605 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3606 if (error != 0) 3607 return (error); 3608 3609 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3610 3611 vrele(vp); 3612 return (error); 3613 } 3614 3615 /* 3616 * Common routine to set ownership given a vnode. 3617 */ 3618 static int 3619 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3620 int posix_semantics) 3621 { 3622 struct vattr vattr; 3623 mode_t newmode; 3624 int error; 3625 3626 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3627 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3628 goto out; 3629 3630 #define CHANGED(x) ((int)(x) != -1) 3631 newmode = vattr.va_mode; 3632 if (posix_semantics) { 3633 /* 3634 * POSIX/XPG semantics: if the caller is not the super-user, 3635 * clear set-user-id and set-group-id bits. Both POSIX and 3636 * the XPG consider the behaviour for calls by the super-user 3637 * implementation-defined; we leave the set-user-id and set- 3638 * group-id settings intact in that case. 3639 */ 3640 if (vattr.va_mode & S_ISUID) { 3641 if (kauth_authorize_vnode(l->l_cred, 3642 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3643 newmode &= ~S_ISUID; 3644 } 3645 if (vattr.va_mode & S_ISGID) { 3646 if (kauth_authorize_vnode(l->l_cred, 3647 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3648 newmode &= ~S_ISGID; 3649 } 3650 } else { 3651 /* 3652 * NetBSD semantics: when changing owner and/or group, 3653 * clear the respective bit(s). 3654 */ 3655 if (CHANGED(uid)) 3656 newmode &= ~S_ISUID; 3657 if (CHANGED(gid)) 3658 newmode &= ~S_ISGID; 3659 } 3660 /* Update va_mode iff altered. */ 3661 if (vattr.va_mode == newmode) 3662 newmode = VNOVAL; 3663 3664 vattr_null(&vattr); 3665 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3666 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3667 vattr.va_mode = newmode; 3668 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3669 #undef CHANGED 3670 3671 out: 3672 VOP_UNLOCK(vp); 3673 return (error); 3674 } 3675 3676 /* 3677 * Set the access and modification times given a path name; this 3678 * version follows links. 3679 */ 3680 /* ARGSUSED */ 3681 int 3682 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3683 register_t *retval) 3684 { 3685 /* { 3686 syscallarg(const char *) path; 3687 syscallarg(const struct timeval *) tptr; 3688 } */ 3689 3690 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3691 SCARG(uap, tptr), UIO_USERSPACE); 3692 } 3693 3694 /* 3695 * Set the access and modification times given a file descriptor. 3696 */ 3697 /* ARGSUSED */ 3698 int 3699 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3700 register_t *retval) 3701 { 3702 /* { 3703 syscallarg(int) fd; 3704 syscallarg(const struct timeval *) tptr; 3705 } */ 3706 int error; 3707 file_t *fp; 3708 3709 /* fd_getvnode() will use the descriptor for us */ 3710 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3711 return (error); 3712 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3713 UIO_USERSPACE); 3714 fd_putfile(SCARG(uap, fd)); 3715 return (error); 3716 } 3717 3718 int 3719 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3720 register_t *retval) 3721 { 3722 /* { 3723 syscallarg(int) fd; 3724 syscallarg(const struct timespec *) tptr; 3725 } */ 3726 int error; 3727 file_t *fp; 3728 3729 /* fd_getvnode() will use the descriptor for us */ 3730 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3731 return (error); 3732 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0, 3733 SCARG(uap, tptr), UIO_USERSPACE); 3734 fd_putfile(SCARG(uap, fd)); 3735 return (error); 3736 } 3737 3738 /* 3739 * Set the access and modification times given a path name; this 3740 * version does not follow links. 3741 */ 3742 int 3743 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3744 register_t *retval) 3745 { 3746 /* { 3747 syscallarg(const char *) path; 3748 syscallarg(const struct timeval *) tptr; 3749 } */ 3750 3751 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3752 SCARG(uap, tptr), UIO_USERSPACE); 3753 } 3754 3755 int 3756 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3757 register_t *retval) 3758 { 3759 /* { 3760 syscallarg(int) fd; 3761 syscallarg(const char *) path; 3762 syscallarg(const struct timespec *) tptr; 3763 syscallarg(int) flag; 3764 } */ 3765 int follow; 3766 const struct timespec *tptr; 3767 int error; 3768 3769 tptr = SCARG(uap, tptr); 3770 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3771 3772 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3773 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3774 3775 return error; 3776 } 3777 3778 /* 3779 * Common routine to set access and modification times given a vnode. 3780 */ 3781 int 3782 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3783 const struct timespec *tptr, enum uio_seg seg) 3784 { 3785 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3786 } 3787 3788 int 3789 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3790 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3791 { 3792 struct vattr vattr; 3793 int error, dorele = 0; 3794 namei_simple_flags_t sflags; 3795 bool vanull, setbirthtime; 3796 struct timespec ts[2]; 3797 3798 KASSERT(l != NULL || fdat == AT_FDCWD); 3799 3800 /* 3801 * I have checked all callers and they pass either FOLLOW, 3802 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3803 * is 0. More to the point, they don't pass anything else. 3804 * Let's keep it that way at least until the namei interfaces 3805 * are fully sanitized. 3806 */ 3807 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3808 sflags = (flag == FOLLOW) ? 3809 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3810 3811 if (tptr == NULL) { 3812 vanull = true; 3813 nanotime(&ts[0]); 3814 ts[1] = ts[0]; 3815 } else { 3816 vanull = false; 3817 if (seg != UIO_SYSSPACE) { 3818 error = copyin(tptr, ts, sizeof (ts)); 3819 if (error != 0) 3820 return error; 3821 } else { 3822 ts[0] = tptr[0]; 3823 ts[1] = tptr[1]; 3824 } 3825 } 3826 3827 if (ts[0].tv_nsec == UTIME_NOW) { 3828 nanotime(&ts[0]); 3829 if (ts[1].tv_nsec == UTIME_NOW) { 3830 vanull = true; 3831 ts[1] = ts[0]; 3832 } 3833 } else if (ts[1].tv_nsec == UTIME_NOW) 3834 nanotime(&ts[1]); 3835 3836 if (vp == NULL) { 3837 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3838 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3839 if (error != 0) 3840 return error; 3841 dorele = 1; 3842 } 3843 3844 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3845 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3846 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3847 vattr_null(&vattr); 3848 3849 if (ts[0].tv_nsec != UTIME_OMIT) 3850 vattr.va_atime = ts[0]; 3851 3852 if (ts[1].tv_nsec != UTIME_OMIT) { 3853 vattr.va_mtime = ts[1]; 3854 if (setbirthtime) 3855 vattr.va_birthtime = ts[1]; 3856 } 3857 3858 if (vanull) 3859 vattr.va_vaflags |= VA_UTIMES_NULL; 3860 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3861 VOP_UNLOCK(vp); 3862 3863 if (dorele != 0) 3864 vrele(vp); 3865 3866 return error; 3867 } 3868 3869 int 3870 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3871 const struct timeval *tptr, enum uio_seg seg) 3872 { 3873 struct timespec ts[2]; 3874 struct timespec *tsptr = NULL; 3875 int error; 3876 3877 if (tptr != NULL) { 3878 struct timeval tv[2]; 3879 3880 if (seg != UIO_SYSSPACE) { 3881 error = copyin(tptr, tv, sizeof (tv)); 3882 if (error != 0) 3883 return error; 3884 tptr = tv; 3885 } 3886 3887 if ((tv[0].tv_usec == UTIME_NOW) || 3888 (tv[0].tv_usec == UTIME_OMIT)) 3889 ts[0].tv_nsec = tv[0].tv_usec; 3890 else 3891 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3892 3893 if ((tv[1].tv_usec == UTIME_NOW) || 3894 (tv[1].tv_usec == UTIME_OMIT)) 3895 ts[1].tv_nsec = tv[1].tv_usec; 3896 else 3897 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3898 3899 tsptr = &ts[0]; 3900 } 3901 3902 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3903 } 3904 3905 /* 3906 * Truncate a file given its path name. 3907 */ 3908 /* ARGSUSED */ 3909 int 3910 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3911 { 3912 /* { 3913 syscallarg(const char *) path; 3914 syscallarg(int) pad; 3915 syscallarg(off_t) length; 3916 } */ 3917 struct vnode *vp; 3918 struct vattr vattr; 3919 int error; 3920 3921 error = namei_simple_user(SCARG(uap, path), 3922 NSM_FOLLOW_TRYEMULROOT, &vp); 3923 if (error != 0) 3924 return (error); 3925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3926 if (vp->v_type == VDIR) 3927 error = EISDIR; 3928 else if ((error = vn_writechk(vp)) == 0 && 3929 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3930 vattr_null(&vattr); 3931 vattr.va_size = SCARG(uap, length); 3932 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3933 } 3934 vput(vp); 3935 return (error); 3936 } 3937 3938 /* 3939 * Truncate a file given a file descriptor. 3940 */ 3941 /* ARGSUSED */ 3942 int 3943 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3944 { 3945 /* { 3946 syscallarg(int) fd; 3947 syscallarg(int) pad; 3948 syscallarg(off_t) length; 3949 } */ 3950 struct vattr vattr; 3951 struct vnode *vp; 3952 file_t *fp; 3953 int error; 3954 3955 /* fd_getvnode() will use the descriptor for us */ 3956 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3957 return (error); 3958 if ((fp->f_flag & FWRITE) == 0) { 3959 error = EINVAL; 3960 goto out; 3961 } 3962 vp = fp->f_data; 3963 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3964 if (vp->v_type == VDIR) 3965 error = EISDIR; 3966 else if ((error = vn_writechk(vp)) == 0) { 3967 vattr_null(&vattr); 3968 vattr.va_size = SCARG(uap, length); 3969 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3970 } 3971 VOP_UNLOCK(vp); 3972 out: 3973 fd_putfile(SCARG(uap, fd)); 3974 return (error); 3975 } 3976 3977 /* 3978 * Sync an open file. 3979 */ 3980 /* ARGSUSED */ 3981 int 3982 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3983 { 3984 /* { 3985 syscallarg(int) fd; 3986 } */ 3987 struct vnode *vp; 3988 file_t *fp; 3989 int error; 3990 3991 /* fd_getvnode() will use the descriptor for us */ 3992 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3993 return (error); 3994 vp = fp->f_data; 3995 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3996 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3997 VOP_UNLOCK(vp); 3998 fd_putfile(SCARG(uap, fd)); 3999 return (error); 4000 } 4001 4002 /* 4003 * Sync a range of file data. API modeled after that found in AIX. 4004 * 4005 * FDATASYNC indicates that we need only save enough metadata to be able 4006 * to re-read the written data. Note we duplicate AIX's requirement that 4007 * the file be open for writing. 4008 */ 4009 /* ARGSUSED */ 4010 int 4011 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4012 { 4013 /* { 4014 syscallarg(int) fd; 4015 syscallarg(int) flags; 4016 syscallarg(off_t) start; 4017 syscallarg(off_t) length; 4018 } */ 4019 struct vnode *vp; 4020 file_t *fp; 4021 int flags, nflags; 4022 off_t s, e, len; 4023 int error; 4024 4025 /* fd_getvnode() will use the descriptor for us */ 4026 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4027 return (error); 4028 4029 if ((fp->f_flag & FWRITE) == 0) { 4030 error = EBADF; 4031 goto out; 4032 } 4033 4034 flags = SCARG(uap, flags); 4035 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4036 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4037 error = EINVAL; 4038 goto out; 4039 } 4040 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4041 if (flags & FDATASYNC) 4042 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4043 else 4044 nflags = FSYNC_WAIT; 4045 if (flags & FDISKSYNC) 4046 nflags |= FSYNC_CACHE; 4047 4048 len = SCARG(uap, length); 4049 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4050 if (len) { 4051 s = SCARG(uap, start); 4052 e = s + len; 4053 if (e < s) { 4054 error = EINVAL; 4055 goto out; 4056 } 4057 } else { 4058 e = 0; 4059 s = 0; 4060 } 4061 4062 vp = fp->f_data; 4063 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4064 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4065 VOP_UNLOCK(vp); 4066 out: 4067 fd_putfile(SCARG(uap, fd)); 4068 return (error); 4069 } 4070 4071 /* 4072 * Sync the data of an open file. 4073 */ 4074 /* ARGSUSED */ 4075 int 4076 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4077 { 4078 /* { 4079 syscallarg(int) fd; 4080 } */ 4081 struct vnode *vp; 4082 file_t *fp; 4083 int error; 4084 4085 /* fd_getvnode() will use the descriptor for us */ 4086 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4087 return (error); 4088 if ((fp->f_flag & FWRITE) == 0) { 4089 fd_putfile(SCARG(uap, fd)); 4090 return (EBADF); 4091 } 4092 vp = fp->f_data; 4093 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4094 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4095 VOP_UNLOCK(vp); 4096 fd_putfile(SCARG(uap, fd)); 4097 return (error); 4098 } 4099 4100 /* 4101 * Rename files, (standard) BSD semantics frontend. 4102 */ 4103 /* ARGSUSED */ 4104 int 4105 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4106 { 4107 /* { 4108 syscallarg(const char *) from; 4109 syscallarg(const char *) to; 4110 } */ 4111 4112 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4113 SCARG(uap, to), UIO_USERSPACE, 0)); 4114 } 4115 4116 int 4117 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4118 register_t *retval) 4119 { 4120 /* { 4121 syscallarg(int) fromfd; 4122 syscallarg(const char *) from; 4123 syscallarg(int) tofd; 4124 syscallarg(const char *) to; 4125 } */ 4126 4127 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4128 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4129 } 4130 4131 /* 4132 * Rename files, POSIX semantics frontend. 4133 */ 4134 /* ARGSUSED */ 4135 int 4136 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4137 { 4138 /* { 4139 syscallarg(const char *) from; 4140 syscallarg(const char *) to; 4141 } */ 4142 4143 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4144 SCARG(uap, to), UIO_USERSPACE, 1)); 4145 } 4146 4147 /* 4148 * Rename files. Source and destination must either both be directories, 4149 * or both not be directories. If target is a directory, it must be empty. 4150 * If `from' and `to' refer to the same object, the value of the `retain' 4151 * argument is used to determine whether `from' will be 4152 * 4153 * (retain == 0) deleted unless `from' and `to' refer to the same 4154 * object in the file system's name space (BSD). 4155 * (retain == 1) always retained (POSIX). 4156 * 4157 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4158 */ 4159 int 4160 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4161 { 4162 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4163 } 4164 4165 static int 4166 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4167 const char *to, enum uio_seg seg, int retain) 4168 { 4169 struct pathbuf *fpb, *tpb; 4170 struct nameidata fnd, tnd; 4171 struct vnode *fdvp, *fvp; 4172 struct vnode *tdvp, *tvp; 4173 struct mount *mp, *tmp; 4174 int error; 4175 4176 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4177 4178 error = pathbuf_maybe_copyin(from, seg, &fpb); 4179 if (error) 4180 goto out0; 4181 KASSERT(fpb != NULL); 4182 4183 error = pathbuf_maybe_copyin(to, seg, &tpb); 4184 if (error) 4185 goto out1; 4186 KASSERT(tpb != NULL); 4187 4188 /* 4189 * Lookup from. 4190 * 4191 * XXX LOCKPARENT is wrong because we don't actually want it 4192 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4193 * insane, so for the time being we need to leave it like this. 4194 */ 4195 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT | INRENAME), fpb); 4196 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4197 goto out2; 4198 4199 /* 4200 * Pull out the important results of the lookup, fdvp and fvp. 4201 * Of course, fvp is bogus because we're about to unlock fdvp. 4202 */ 4203 fdvp = fnd.ni_dvp; 4204 fvp = fnd.ni_vp; 4205 KASSERT(fdvp != NULL); 4206 KASSERT(fvp != NULL); 4207 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4208 4209 /* 4210 * Make sure neither fdvp nor fvp is locked. 4211 */ 4212 if (fdvp != fvp) 4213 VOP_UNLOCK(fdvp); 4214 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4215 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4216 4217 /* 4218 * Reject renaming `.' and `..'. Can't do this until after 4219 * namei because we need namei's parsing to find the final 4220 * component name. (namei should just leave us with the final 4221 * component name and not look it up itself, but anyway...) 4222 * 4223 * This was here before because we used to relookup from 4224 * instead of to and relookup requires the caller to check 4225 * this, but now file systems may depend on this check, so we 4226 * must retain it until the file systems are all rototilled. 4227 */ 4228 if (((fnd.ni_cnd.cn_namelen == 1) && 4229 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4230 ((fnd.ni_cnd.cn_namelen == 2) && 4231 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4232 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4233 error = EINVAL; /* XXX EISDIR? */ 4234 goto abort0; 4235 } 4236 4237 /* 4238 * Lookup to. 4239 * 4240 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4241 * fvp here to decide whether to add CREATEDIR is a load of 4242 * bollocks because fvp might be the wrong node by now, since 4243 * fdvp is unlocked. 4244 * 4245 * XXX Why not pass CREATEDIR always? 4246 */ 4247 NDINIT(&tnd, RENAME, 4248 (LOCKPARENT | NOCACHE | TRYEMULROOT | INRENAME | 4249 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4250 tpb); 4251 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4252 goto abort0; 4253 4254 /* 4255 * Pull out the important results of the lookup, tdvp and tvp. 4256 * Of course, tvp is bogus because we're about to unlock tdvp. 4257 */ 4258 tdvp = tnd.ni_dvp; 4259 tvp = tnd.ni_vp; 4260 KASSERT(tdvp != NULL); 4261 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4262 4263 /* 4264 * Make sure neither tdvp nor tvp is locked. 4265 */ 4266 if (tdvp != tvp) 4267 VOP_UNLOCK(tdvp); 4268 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4269 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4270 4271 /* 4272 * Reject renaming onto `.' or `..'. relookup is unhappy with 4273 * these, which is why we must do this here. Once upon a time 4274 * we relooked up from instead of to, and consequently didn't 4275 * need this check, but now that we relookup to instead of 4276 * from, we need this; and we shall need it forever forward 4277 * until the VOP_RENAME protocol changes, because file systems 4278 * will no doubt begin to depend on this check. 4279 */ 4280 if (((tnd.ni_cnd.cn_namelen == 1) && 4281 (tnd.ni_cnd.cn_nameptr[0] == '.')) || 4282 ((tnd.ni_cnd.cn_namelen == 2) && 4283 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4284 (tnd.ni_cnd.cn_nameptr[1] == '.'))) { 4285 error = EINVAL; /* XXX EISDIR? */ 4286 goto abort1; 4287 } 4288 4289 /* 4290 * Get the mount point. If the file system has been unmounted, 4291 * which it may be because we're not holding any vnode locks, 4292 * then v_mount will be NULL. We're not really supposed to 4293 * read v_mount without holding the vnode lock, but since we 4294 * have fdvp referenced, if fdvp->v_mount changes then at worst 4295 * it will be set to NULL, not changed to another mount point. 4296 * And, of course, since it is up to the file system to 4297 * determine the real lock order, we can't lock both fdvp and 4298 * tdvp at the same time. 4299 */ 4300 mp = fdvp->v_mount; 4301 if (mp == NULL) { 4302 error = ENOENT; 4303 goto abort1; 4304 } 4305 4306 /* 4307 * Make sure the mount points match. Again, although we don't 4308 * hold any vnode locks, the v_mount fields may change -- but 4309 * at worst they will change to NULL, so this will never become 4310 * a cross-device rename, because we hold vnode references. 4311 * 4312 * XXX Because nothing is locked and the compiler may reorder 4313 * things here, unmounting the file system at an inopportune 4314 * moment may cause rename to fail with ENXDEV when it really 4315 * should fail with ENOENT. 4316 */ 4317 tmp = tdvp->v_mount; 4318 if (tmp == NULL) { 4319 error = ENOENT; 4320 goto abort1; 4321 } 4322 4323 if (mp != tmp) { 4324 error = EXDEV; 4325 goto abort1; 4326 } 4327 4328 /* 4329 * Take the vfs rename lock to avoid cross-directory screw cases. 4330 * Nothing is locked currently, so taking this lock is safe. 4331 */ 4332 error = VFS_RENAMELOCK_ENTER(mp); 4333 if (error) 4334 goto abort1; 4335 4336 /* 4337 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4338 * and nothing is locked except for the vfs rename lock. 4339 * 4340 * The next step is a little rain dance to conform to the 4341 * insane lock protocol, even though it does nothing to ward 4342 * off race conditions. 4343 * 4344 * We need tdvp and tvp to be locked. However, because we have 4345 * unlocked tdvp in order to hold no locks while we take the 4346 * vfs rename lock, tvp may be wrong here, and we can't safely 4347 * lock it even if the sensible file systems will just unlock 4348 * it straight away. Consequently, we must lock tdvp and then 4349 * relookup tvp to get it locked. 4350 * 4351 * Finally, because the VOP_RENAME protocol is brain-damaged 4352 * and various file systems insanely depend on the semantics of 4353 * this brain damage, the lookup of to must be the last lookup 4354 * before VOP_RENAME. 4355 */ 4356 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4357 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4358 if (error) 4359 goto abort2; 4360 4361 /* 4362 * Drop the old tvp and pick up the new one -- which might be 4363 * the same, but that doesn't matter to us. After this, tdvp 4364 * and tvp should both be locked. 4365 */ 4366 if (tvp != NULL) 4367 vrele(tvp); 4368 tvp = tnd.ni_vp; 4369 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4370 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4371 4372 /* 4373 * The old do_sys_rename had various consistency checks here 4374 * involving fvp and tvp. fvp is bogus already here, and tvp 4375 * will become bogus soon in any sensible file system, so the 4376 * only purpose in putting these checks here is to give lip 4377 * service to these screw cases and to acknowledge that they 4378 * exist, not actually to handle them, but here you go 4379 * anyway... 4380 */ 4381 4382 /* 4383 * Acknowledge that directories and non-directories aren't 4384 * suposed to mix. 4385 */ 4386 if (tvp != NULL) { 4387 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4388 error = ENOTDIR; 4389 goto abort3; 4390 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4391 error = EISDIR; 4392 goto abort3; 4393 } 4394 } 4395 4396 /* 4397 * Acknowledge some random screw case, among the dozens that 4398 * might arise. 4399 */ 4400 if (fvp == tdvp) { 4401 error = EINVAL; 4402 goto abort3; 4403 } 4404 4405 /* 4406 * Acknowledge that POSIX has a wacky screw case. 4407 * 4408 * XXX Eventually the retain flag needs to be passed on to 4409 * VOP_RENAME. 4410 */ 4411 if (fvp == tvp) { 4412 if (retain) { 4413 error = 0; 4414 goto abort3; 4415 } else if ((fdvp == tdvp) && 4416 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4417 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4418 fnd.ni_cnd.cn_namelen))) { 4419 error = 0; 4420 goto abort3; 4421 } 4422 } 4423 4424 /* 4425 * Make sure veriexec can screw us up. (But a race can screw 4426 * up veriexec, of course -- remember, fvp and (soon) tvp are 4427 * bogus.) 4428 */ 4429 #if NVERIEXEC > 0 4430 { 4431 char *f1, *f2; 4432 size_t f1_len; 4433 size_t f2_len; 4434 4435 f1_len = fnd.ni_cnd.cn_namelen + 1; 4436 f1 = kmem_alloc(f1_len, KM_SLEEP); 4437 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4438 4439 f2_len = tnd.ni_cnd.cn_namelen + 1; 4440 f2 = kmem_alloc(f2_len, KM_SLEEP); 4441 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4442 4443 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4444 4445 kmem_free(f1, f1_len); 4446 kmem_free(f2, f2_len); 4447 4448 if (error) 4449 goto abort3; 4450 } 4451 #endif /* NVERIEXEC > 0 */ 4452 4453 /* 4454 * All ready. Incant the rename vop. 4455 */ 4456 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4457 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4458 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4459 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4460 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4461 4462 /* 4463 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4464 * tdvp and tvp. But we can't assert any of that. 4465 */ 4466 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4467 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4468 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4469 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4470 4471 /* 4472 * So all we have left to do is to drop the rename lock and 4473 * destroy the pathbufs. 4474 */ 4475 VFS_RENAMELOCK_EXIT(mp); 4476 goto out2; 4477 4478 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4479 VOP_UNLOCK(tvp); 4480 abort2: VOP_UNLOCK(tdvp); 4481 VFS_RENAMELOCK_EXIT(mp); 4482 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4483 vrele(tdvp); 4484 if (tvp != NULL) 4485 vrele(tvp); 4486 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4487 vrele(fdvp); 4488 vrele(fvp); 4489 out2: pathbuf_destroy(tpb); 4490 out1: pathbuf_destroy(fpb); 4491 out0: return error; 4492 } 4493 4494 /* 4495 * Make a directory file. 4496 */ 4497 /* ARGSUSED */ 4498 int 4499 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4500 { 4501 /* { 4502 syscallarg(const char *) path; 4503 syscallarg(int) mode; 4504 } */ 4505 4506 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4507 SCARG(uap, mode), UIO_USERSPACE); 4508 } 4509 4510 int 4511 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4512 register_t *retval) 4513 { 4514 /* { 4515 syscallarg(int) fd; 4516 syscallarg(const char *) path; 4517 syscallarg(int) mode; 4518 } */ 4519 4520 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4521 SCARG(uap, mode), UIO_USERSPACE); 4522 } 4523 4524 4525 int 4526 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4527 { 4528 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4529 } 4530 4531 static int 4532 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4533 enum uio_seg seg) 4534 { 4535 struct proc *p = curlwp->l_proc; 4536 struct vnode *vp; 4537 struct vattr vattr; 4538 int error; 4539 struct pathbuf *pb; 4540 struct nameidata nd; 4541 4542 KASSERT(l != NULL || fdat == AT_FDCWD); 4543 4544 /* XXX bollocks, should pass in a pathbuf */ 4545 error = pathbuf_maybe_copyin(path, seg, &pb); 4546 if (error) { 4547 return error; 4548 } 4549 4550 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4551 4552 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4553 pathbuf_destroy(pb); 4554 return (error); 4555 } 4556 vp = nd.ni_vp; 4557 if (vp != NULL) { 4558 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4559 if (nd.ni_dvp == vp) 4560 vrele(nd.ni_dvp); 4561 else 4562 vput(nd.ni_dvp); 4563 vrele(vp); 4564 pathbuf_destroy(pb); 4565 return (EEXIST); 4566 } 4567 vattr_null(&vattr); 4568 vattr.va_type = VDIR; 4569 /* We will read cwdi->cwdi_cmask unlocked. */ 4570 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4571 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4572 if (!error) 4573 vrele(nd.ni_vp); 4574 vput(nd.ni_dvp); 4575 pathbuf_destroy(pb); 4576 return (error); 4577 } 4578 4579 /* 4580 * Remove a directory file. 4581 */ 4582 /* ARGSUSED */ 4583 int 4584 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4585 { 4586 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4587 AT_REMOVEDIR, UIO_USERSPACE); 4588 } 4589 4590 /* 4591 * Read a block of directory entries in a file system independent format. 4592 */ 4593 int 4594 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4595 { 4596 /* { 4597 syscallarg(int) fd; 4598 syscallarg(char *) buf; 4599 syscallarg(size_t) count; 4600 } */ 4601 file_t *fp; 4602 int error, done; 4603 4604 /* fd_getvnode() will use the descriptor for us */ 4605 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4606 return (error); 4607 if ((fp->f_flag & FREAD) == 0) { 4608 error = EBADF; 4609 goto out; 4610 } 4611 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4612 SCARG(uap, count), &done, l, 0, 0); 4613 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4614 *retval = done; 4615 out: 4616 fd_putfile(SCARG(uap, fd)); 4617 return (error); 4618 } 4619 4620 /* 4621 * Set the mode mask for creation of filesystem nodes. 4622 */ 4623 int 4624 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4625 { 4626 /* { 4627 syscallarg(mode_t) newmask; 4628 } */ 4629 struct proc *p = l->l_proc; 4630 struct cwdinfo *cwdi; 4631 4632 /* 4633 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4634 * important is that we serialize changes to the mask. The 4635 * rw_exit() will issue a write memory barrier on our behalf, 4636 * and force the changes out to other CPUs (as it must use an 4637 * atomic operation, draining the local CPU's store buffers). 4638 */ 4639 cwdi = p->p_cwdi; 4640 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4641 *retval = cwdi->cwdi_cmask; 4642 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4643 rw_exit(&cwdi->cwdi_lock); 4644 4645 return (0); 4646 } 4647 4648 int 4649 dorevoke(struct vnode *vp, kauth_cred_t cred) 4650 { 4651 struct vattr vattr; 4652 int error, fs_decision; 4653 4654 vn_lock(vp, LK_SHARED | LK_RETRY); 4655 error = VOP_GETATTR(vp, &vattr, cred); 4656 VOP_UNLOCK(vp); 4657 if (error != 0) 4658 return error; 4659 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4660 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4661 fs_decision); 4662 if (!error) 4663 VOP_REVOKE(vp, REVOKEALL); 4664 return (error); 4665 } 4666 4667 /* 4668 * Void all references to file by ripping underlying filesystem 4669 * away from vnode. 4670 */ 4671 /* ARGSUSED */ 4672 int 4673 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4674 { 4675 /* { 4676 syscallarg(const char *) path; 4677 } */ 4678 struct vnode *vp; 4679 int error; 4680 4681 error = namei_simple_user(SCARG(uap, path), 4682 NSM_FOLLOW_TRYEMULROOT, &vp); 4683 if (error != 0) 4684 return (error); 4685 error = dorevoke(vp, l->l_cred); 4686 vrele(vp); 4687 return (error); 4688 } 4689
Indexes created Wed Sep 24 05:09:52 GMT 2025