vfs_syscalls.c revision 1.479
1 /* $NetBSD: vfs_syscalls.c,v 1.479 2014/04/16 18:55:18 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 1.479 2014/04/16 18:55:18 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 /* 485 * We allow data to be NULL, even for userspace. Some fs's don't need 486 * it. The others will handle NULL. 487 */ 488 if (data != NULL && data_seg == UIO_USERSPACE) { 489 if (data_len == 0) { 490 /* No length supplied, use default for filesystem */ 491 data_len = vfsops->vfs_min_mount_data; 492 493 /* 494 * Hopefully a longer buffer won't make copyin() fail. 495 * For compatibility with 3.0 and earlier. 496 */ 497 if (flags & MNT_UPDATE 498 && data_len < sizeof (struct mnt_export_args30)) 499 data_len = sizeof (struct mnt_export_args30); 500 } 501 if (data_len > VFS_MAX_MOUNT_DATA) { 502 error = EINVAL; 503 goto done; 504 } 505 data_buf = kmem_alloc(data_len, KM_SLEEP); 506 507 /* NFS needs the buffer even for mnt_getargs .... */ 508 error = copyin(data, data_buf, data_len); 509 if (error != 0) 510 goto done; 511 } 512 513 if (flags & MNT_GETARGS) { 514 if (data_len == 0) { 515 error = EINVAL; 516 goto done; 517 } 518 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 519 if (error != 0) 520 goto done; 521 if (data_seg == UIO_USERSPACE) 522 error = copyout(data_buf, data, data_len); 523 *retval = data_len; 524 } else if (flags & MNT_UPDATE) { 525 error = mount_update(l, vp, path, flags, data_buf, &data_len); 526 } else { 527 /* Locking is handled internally in mount_domount(). */ 528 KASSERT(vfsopsrele == true); 529 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 530 &data_len); 531 vfsopsrele = false; 532 } 533 534 done: 535 if (vfsopsrele) 536 vfs_delref(vfsops); 537 if (vp != NULL) { 538 vrele(vp); 539 } 540 if (data_buf != data) 541 kmem_free(data_buf, data_len); 542 return (error); 543 } 544 545 /* 546 * Unmount a file system. 547 * 548 * Note: unmount takes a path to the vnode mounted on as argument, 549 * not special file (as before). 550 */ 551 /* ARGSUSED */ 552 int 553 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 554 { 555 /* { 556 syscallarg(const char *) path; 557 syscallarg(int) flags; 558 } */ 559 struct vnode *vp; 560 struct mount *mp; 561 int error; 562 struct pathbuf *pb; 563 struct nameidata nd; 564 565 error = pathbuf_copyin(SCARG(uap, path), &pb); 566 if (error) { 567 return error; 568 } 569 570 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 571 if ((error = namei(&nd)) != 0) { 572 pathbuf_destroy(pb); 573 return error; 574 } 575 vp = nd.ni_vp; 576 pathbuf_destroy(pb); 577 578 mp = vp->v_mount; 579 atomic_inc_uint(&mp->mnt_refcnt); 580 VOP_UNLOCK(vp); 581 582 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 583 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 584 if (error) { 585 vrele(vp); 586 vfs_destroy(mp); 587 return (error); 588 } 589 590 /* 591 * Don't allow unmounting the root file system. 592 */ 593 if (mp->mnt_flag & MNT_ROOTFS) { 594 vrele(vp); 595 vfs_destroy(mp); 596 return (EINVAL); 597 } 598 599 /* 600 * Must be the root of the filesystem 601 */ 602 if ((vp->v_vflag & VV_ROOT) == 0) { 603 vrele(vp); 604 vfs_destroy(mp); 605 return (EINVAL); 606 } 607 608 vrele(vp); 609 error = dounmount(mp, SCARG(uap, flags), l); 610 vfs_destroy(mp); 611 return error; 612 } 613 614 /* 615 * Sync each mounted filesystem. 616 */ 617 #ifdef DEBUG 618 int syncprt = 0; 619 struct ctldebug debug0 = { "syncprt", &syncprt }; 620 #endif 621 622 void 623 do_sys_sync(struct lwp *l) 624 { 625 struct mount *mp, *nmp; 626 int asyncflag; 627 628 mutex_enter(&mountlist_lock); 629 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 630 if (vfs_busy(mp, &nmp)) { 631 continue; 632 } 633 mutex_enter(&mp->mnt_updating); 634 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 635 asyncflag = mp->mnt_flag & MNT_ASYNC; 636 mp->mnt_flag &= ~MNT_ASYNC; 637 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 638 if (asyncflag) 639 mp->mnt_flag |= MNT_ASYNC; 640 } 641 mutex_exit(&mp->mnt_updating); 642 vfs_unbusy(mp, false, &nmp); 643 } 644 mutex_exit(&mountlist_lock); 645 #ifdef DEBUG 646 if (syncprt) 647 vfs_bufstats(); 648 #endif /* DEBUG */ 649 } 650 651 /* ARGSUSED */ 652 int 653 sys_sync(struct lwp *l, const void *v, register_t *retval) 654 { 655 do_sys_sync(l); 656 return (0); 657 } 658 659 660 /* 661 * Access or change filesystem quotas. 662 * 663 * (this is really 14 different calls bundled into one) 664 */ 665 666 static int 667 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 668 { 669 struct quotastat info_k; 670 int error; 671 672 /* ensure any padding bytes are cleared */ 673 memset(&info_k, 0, sizeof(info_k)); 674 675 error = vfs_quotactl_stat(mp, &info_k); 676 if (error) { 677 return error; 678 } 679 680 return copyout(&info_k, info_u, sizeof(info_k)); 681 } 682 683 static int 684 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 685 struct quotaidtypestat *info_u) 686 { 687 struct quotaidtypestat info_k; 688 int error; 689 690 /* ensure any padding bytes are cleared */ 691 memset(&info_k, 0, sizeof(info_k)); 692 693 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 694 if (error) { 695 return error; 696 } 697 698 return copyout(&info_k, info_u, sizeof(info_k)); 699 } 700 701 static int 702 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 703 struct quotaobjtypestat *info_u) 704 { 705 struct quotaobjtypestat info_k; 706 int error; 707 708 /* ensure any padding bytes are cleared */ 709 memset(&info_k, 0, sizeof(info_k)); 710 711 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 712 if (error) { 713 return error; 714 } 715 716 return copyout(&info_k, info_u, sizeof(info_k)); 717 } 718 719 static int 720 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 721 struct quotaval *val_u) 722 { 723 struct quotakey key_k; 724 struct quotaval val_k; 725 int error; 726 727 /* ensure any padding bytes are cleared */ 728 memset(&val_k, 0, sizeof(val_k)); 729 730 error = copyin(key_u, &key_k, sizeof(key_k)); 731 if (error) { 732 return error; 733 } 734 735 error = vfs_quotactl_get(mp, &key_k, &val_k); 736 if (error) { 737 return error; 738 } 739 740 return copyout(&val_k, val_u, sizeof(val_k)); 741 } 742 743 static int 744 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 745 const struct quotaval *val_u) 746 { 747 struct quotakey key_k; 748 struct quotaval val_k; 749 int error; 750 751 error = copyin(key_u, &key_k, sizeof(key_k)); 752 if (error) { 753 return error; 754 } 755 756 error = copyin(val_u, &val_k, sizeof(val_k)); 757 if (error) { 758 return error; 759 } 760 761 return vfs_quotactl_put(mp, &key_k, &val_k); 762 } 763 764 static int 765 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u) 766 { 767 struct quotakey key_k; 768 int error; 769 770 error = copyin(key_u, &key_k, sizeof(key_k)); 771 if (error) { 772 return error; 773 } 774 775 return vfs_quotactl_delete(mp, &key_k); 776 } 777 778 static int 779 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 780 { 781 struct quotakcursor cursor_k; 782 int error; 783 784 /* ensure any padding bytes are cleared */ 785 memset(&cursor_k, 0, sizeof(cursor_k)); 786 787 error = vfs_quotactl_cursoropen(mp, &cursor_k); 788 if (error) { 789 return error; 790 } 791 792 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 793 } 794 795 static int 796 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 797 { 798 struct quotakcursor cursor_k; 799 int error; 800 801 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 802 if (error) { 803 return error; 804 } 805 806 return vfs_quotactl_cursorclose(mp, &cursor_k); 807 } 808 809 static int 810 do_sys_quotactl_cursorskipidtype(struct mount *mp, 811 struct quotakcursor *cursor_u, int idtype) 812 { 813 struct quotakcursor cursor_k; 814 int error; 815 816 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 817 if (error) { 818 return error; 819 } 820 821 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 822 if (error) { 823 return error; 824 } 825 826 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 827 } 828 829 static int 830 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 831 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 832 unsigned *ret_u) 833 { 834 #define CGET_STACK_MAX 8 835 struct quotakcursor cursor_k; 836 struct quotakey stackkeys[CGET_STACK_MAX]; 837 struct quotaval stackvals[CGET_STACK_MAX]; 838 struct quotakey *keys_k; 839 struct quotaval *vals_k; 840 unsigned ret_k; 841 int error; 842 843 if (maxnum > 128) { 844 maxnum = 128; 845 } 846 847 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 848 if (error) { 849 return error; 850 } 851 852 if (maxnum <= CGET_STACK_MAX) { 853 keys_k = stackkeys; 854 vals_k = stackvals; 855 /* ensure any padding bytes are cleared */ 856 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 857 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 858 } else { 859 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 860 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 861 } 862 863 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 864 &ret_k); 865 if (error) { 866 goto fail; 867 } 868 869 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 870 if (error) { 871 goto fail; 872 } 873 874 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 875 if (error) { 876 goto fail; 877 } 878 879 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 880 if (error) { 881 goto fail; 882 } 883 884 /* do last to maximize the chance of being able to recover a failure */ 885 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 886 887 fail: 888 if (keys_k != stackkeys) { 889 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 890 } 891 if (vals_k != stackvals) { 892 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 893 } 894 return error; 895 } 896 897 static int 898 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 899 int *ret_u) 900 { 901 struct quotakcursor cursor_k; 902 int ret_k; 903 int error; 904 905 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 906 if (error) { 907 return error; 908 } 909 910 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 911 if (error) { 912 return error; 913 } 914 915 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 916 if (error) { 917 return error; 918 } 919 920 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 921 } 922 923 static int 924 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 925 { 926 struct quotakcursor cursor_k; 927 int error; 928 929 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 930 if (error) { 931 return error; 932 } 933 934 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 935 if (error) { 936 return error; 937 } 938 939 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 940 } 941 942 static int 943 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 944 { 945 char *path_k; 946 int error; 947 948 /* XXX this should probably be a struct pathbuf */ 949 path_k = PNBUF_GET(); 950 error = copyin(path_u, path_k, PATH_MAX); 951 if (error) { 952 PNBUF_PUT(path_k); 953 return error; 954 } 955 956 error = vfs_quotactl_quotaon(mp, idtype, path_k); 957 958 PNBUF_PUT(path_k); 959 return error; 960 } 961 962 static int 963 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 964 { 965 return vfs_quotactl_quotaoff(mp, idtype); 966 } 967 968 int 969 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 970 { 971 struct mount *mp; 972 struct vnode *vp; 973 int error; 974 975 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 976 if (error != 0) 977 return (error); 978 mp = vp->v_mount; 979 980 switch (args->qc_op) { 981 case QUOTACTL_STAT: 982 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 983 break; 984 case QUOTACTL_IDTYPESTAT: 985 error = do_sys_quotactl_idtypestat(mp, 986 args->u.idtypestat.qc_idtype, 987 args->u.idtypestat.qc_info); 988 break; 989 case QUOTACTL_OBJTYPESTAT: 990 error = do_sys_quotactl_objtypestat(mp, 991 args->u.objtypestat.qc_objtype, 992 args->u.objtypestat.qc_info); 993 break; 994 case QUOTACTL_GET: 995 error = do_sys_quotactl_get(mp, 996 args->u.get.qc_key, 997 args->u.get.qc_val); 998 break; 999 case QUOTACTL_PUT: 1000 error = do_sys_quotactl_put(mp, 1001 args->u.put.qc_key, 1002 args->u.put.qc_val); 1003 break; 1004 case QUOTACTL_DELETE: 1005 error = do_sys_quotactl_delete(mp, args->u.delete.qc_key); 1006 break; 1007 case QUOTACTL_CURSOROPEN: 1008 error = do_sys_quotactl_cursoropen(mp, 1009 args->u.cursoropen.qc_cursor); 1010 break; 1011 case QUOTACTL_CURSORCLOSE: 1012 error = do_sys_quotactl_cursorclose(mp, 1013 args->u.cursorclose.qc_cursor); 1014 break; 1015 case QUOTACTL_CURSORSKIPIDTYPE: 1016 error = do_sys_quotactl_cursorskipidtype(mp, 1017 args->u.cursorskipidtype.qc_cursor, 1018 args->u.cursorskipidtype.qc_idtype); 1019 break; 1020 case QUOTACTL_CURSORGET: 1021 error = do_sys_quotactl_cursorget(mp, 1022 args->u.cursorget.qc_cursor, 1023 args->u.cursorget.qc_keys, 1024 args->u.cursorget.qc_vals, 1025 args->u.cursorget.qc_maxnum, 1026 args->u.cursorget.qc_ret); 1027 break; 1028 case QUOTACTL_CURSORATEND: 1029 error = do_sys_quotactl_cursoratend(mp, 1030 args->u.cursoratend.qc_cursor, 1031 args->u.cursoratend.qc_ret); 1032 break; 1033 case QUOTACTL_CURSORREWIND: 1034 error = do_sys_quotactl_cursorrewind(mp, 1035 args->u.cursorrewind.qc_cursor); 1036 break; 1037 case QUOTACTL_QUOTAON: 1038 error = do_sys_quotactl_quotaon(mp, 1039 args->u.quotaon.qc_idtype, 1040 args->u.quotaon.qc_quotafile); 1041 break; 1042 case QUOTACTL_QUOTAOFF: 1043 error = do_sys_quotactl_quotaoff(mp, 1044 args->u.quotaoff.qc_idtype); 1045 break; 1046 default: 1047 error = EINVAL; 1048 break; 1049 } 1050 1051 vrele(vp); 1052 return error; 1053 } 1054 1055 /* ARGSUSED */ 1056 int 1057 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1058 register_t *retval) 1059 { 1060 /* { 1061 syscallarg(const char *) path; 1062 syscallarg(struct quotactl_args *) args; 1063 } */ 1064 struct quotactl_args args; 1065 int error; 1066 1067 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1068 if (error) { 1069 return error; 1070 } 1071 1072 return do_sys_quotactl(SCARG(uap, path), &args); 1073 } 1074 1075 int 1076 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1077 int root) 1078 { 1079 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1080 int error = 0; 1081 1082 /* 1083 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1084 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1085 * overrides MNT_NOWAIT. 1086 */ 1087 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1088 (flags != MNT_WAIT && flags != 0)) { 1089 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1090 goto done; 1091 } 1092 1093 /* Get the filesystem stats now */ 1094 memset(sp, 0, sizeof(*sp)); 1095 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1096 return error; 1097 } 1098 1099 if (cwdi->cwdi_rdir == NULL) 1100 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1101 done: 1102 if (cwdi->cwdi_rdir != NULL) { 1103 size_t len; 1104 char *bp; 1105 char c; 1106 char *path = PNBUF_GET(); 1107 1108 bp = path + MAXPATHLEN; 1109 *--bp = '\0'; 1110 rw_enter(&cwdi->cwdi_lock, RW_READER); 1111 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1112 MAXPATHLEN / 2, 0, l); 1113 rw_exit(&cwdi->cwdi_lock); 1114 if (error) { 1115 PNBUF_PUT(path); 1116 return error; 1117 } 1118 len = strlen(bp); 1119 if (len != 1) { 1120 /* 1121 * for mount points that are below our root, we can see 1122 * them, so we fix up the pathname and return them. The 1123 * rest we cannot see, so we don't allow viewing the 1124 * data. 1125 */ 1126 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1127 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1128 (void)strlcpy(sp->f_mntonname, 1129 c == '\0' ? "/" : &sp->f_mntonname[len], 1130 sizeof(sp->f_mntonname)); 1131 } else { 1132 if (root) 1133 (void)strlcpy(sp->f_mntonname, "/", 1134 sizeof(sp->f_mntonname)); 1135 else 1136 error = EPERM; 1137 } 1138 } 1139 PNBUF_PUT(path); 1140 } 1141 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1142 return error; 1143 } 1144 1145 /* 1146 * Get filesystem statistics by path. 1147 */ 1148 int 1149 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1150 { 1151 struct mount *mp; 1152 int error; 1153 struct vnode *vp; 1154 1155 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1156 if (error != 0) 1157 return error; 1158 mp = vp->v_mount; 1159 error = dostatvfs(mp, sb, l, flags, 1); 1160 vrele(vp); 1161 return error; 1162 } 1163 1164 /* ARGSUSED */ 1165 int 1166 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1167 { 1168 /* { 1169 syscallarg(const char *) path; 1170 syscallarg(struct statvfs *) buf; 1171 syscallarg(int) flags; 1172 } */ 1173 struct statvfs *sb; 1174 int error; 1175 1176 sb = STATVFSBUF_GET(); 1177 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1178 if (error == 0) 1179 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1180 STATVFSBUF_PUT(sb); 1181 return error; 1182 } 1183 1184 /* 1185 * Get filesystem statistics by fd. 1186 */ 1187 int 1188 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1189 { 1190 file_t *fp; 1191 struct mount *mp; 1192 int error; 1193 1194 /* fd_getvnode() will use the descriptor for us */ 1195 if ((error = fd_getvnode(fd, &fp)) != 0) 1196 return (error); 1197 mp = ((struct vnode *)fp->f_data)->v_mount; 1198 error = dostatvfs(mp, sb, curlwp, flags, 1); 1199 fd_putfile(fd); 1200 return error; 1201 } 1202 1203 /* ARGSUSED */ 1204 int 1205 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1206 { 1207 /* { 1208 syscallarg(int) fd; 1209 syscallarg(struct statvfs *) buf; 1210 syscallarg(int) flags; 1211 } */ 1212 struct statvfs *sb; 1213 int error; 1214 1215 sb = STATVFSBUF_GET(); 1216 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1217 if (error == 0) 1218 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1219 STATVFSBUF_PUT(sb); 1220 return error; 1221 } 1222 1223 1224 /* 1225 * Get statistics on all filesystems. 1226 */ 1227 int 1228 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1229 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1230 register_t *retval) 1231 { 1232 int root = 0; 1233 struct proc *p = l->l_proc; 1234 struct mount *mp, *nmp; 1235 struct statvfs *sb; 1236 size_t count, maxcount; 1237 int error = 0; 1238 1239 sb = STATVFSBUF_GET(); 1240 maxcount = bufsize / entry_sz; 1241 mutex_enter(&mountlist_lock); 1242 count = 0; 1243 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 1244 if (vfs_busy(mp, &nmp)) { 1245 continue; 1246 } 1247 if (sfsp && count < maxcount) { 1248 error = dostatvfs(mp, sb, l, flags, 0); 1249 if (error) { 1250 vfs_unbusy(mp, false, &nmp); 1251 error = 0; 1252 continue; 1253 } 1254 error = copyfn(sb, sfsp, entry_sz); 1255 if (error) { 1256 vfs_unbusy(mp, false, NULL); 1257 goto out; 1258 } 1259 sfsp = (char *)sfsp + entry_sz; 1260 root |= strcmp(sb->f_mntonname, "/") == 0; 1261 } 1262 count++; 1263 vfs_unbusy(mp, false, &nmp); 1264 } 1265 mutex_exit(&mountlist_lock); 1266 1267 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1268 /* 1269 * fake a root entry 1270 */ 1271 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1272 sb, l, flags, 1); 1273 if (error != 0) 1274 goto out; 1275 if (sfsp) { 1276 error = copyfn(sb, sfsp, entry_sz); 1277 if (error != 0) 1278 goto out; 1279 } 1280 count++; 1281 } 1282 if (sfsp && count > maxcount) 1283 *retval = maxcount; 1284 else 1285 *retval = count; 1286 out: 1287 STATVFSBUF_PUT(sb); 1288 return error; 1289 } 1290 1291 int 1292 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1293 { 1294 /* { 1295 syscallarg(struct statvfs *) buf; 1296 syscallarg(size_t) bufsize; 1297 syscallarg(int) flags; 1298 } */ 1299 1300 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1301 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1302 } 1303 1304 /* 1305 * Change current working directory to a given file descriptor. 1306 */ 1307 /* ARGSUSED */ 1308 int 1309 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1310 { 1311 /* { 1312 syscallarg(int) fd; 1313 } */ 1314 struct proc *p = l->l_proc; 1315 struct cwdinfo *cwdi; 1316 struct vnode *vp, *tdp; 1317 struct mount *mp; 1318 file_t *fp; 1319 int error, fd; 1320 1321 /* fd_getvnode() will use the descriptor for us */ 1322 fd = SCARG(uap, fd); 1323 if ((error = fd_getvnode(fd, &fp)) != 0) 1324 return (error); 1325 vp = fp->f_data; 1326 1327 vref(vp); 1328 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1329 if (vp->v_type != VDIR) 1330 error = ENOTDIR; 1331 else 1332 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1333 if (error) { 1334 vput(vp); 1335 goto out; 1336 } 1337 while ((mp = vp->v_mountedhere) != NULL) { 1338 error = vfs_busy(mp, NULL); 1339 vput(vp); 1340 if (error != 0) 1341 goto out; 1342 error = VFS_ROOT(mp, &tdp); 1343 vfs_unbusy(mp, false, NULL); 1344 if (error) 1345 goto out; 1346 vp = tdp; 1347 } 1348 VOP_UNLOCK(vp); 1349 1350 /* 1351 * Disallow changing to a directory not under the process's 1352 * current root directory (if there is one). 1353 */ 1354 cwdi = p->p_cwdi; 1355 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1356 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1357 vrele(vp); 1358 error = EPERM; /* operation not permitted */ 1359 } else { 1360 vrele(cwdi->cwdi_cdir); 1361 cwdi->cwdi_cdir = vp; 1362 } 1363 rw_exit(&cwdi->cwdi_lock); 1364 1365 out: 1366 fd_putfile(fd); 1367 return (error); 1368 } 1369 1370 /* 1371 * Change this process's notion of the root directory to a given file 1372 * descriptor. 1373 */ 1374 int 1375 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1376 { 1377 struct proc *p = l->l_proc; 1378 struct vnode *vp; 1379 file_t *fp; 1380 int error, fd = SCARG(uap, fd); 1381 1382 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1383 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1384 return error; 1385 /* fd_getvnode() will use the descriptor for us */ 1386 if ((error = fd_getvnode(fd, &fp)) != 0) 1387 return error; 1388 vp = fp->f_data; 1389 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1390 if (vp->v_type != VDIR) 1391 error = ENOTDIR; 1392 else 1393 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1394 VOP_UNLOCK(vp); 1395 if (error) 1396 goto out; 1397 vref(vp); 1398 1399 change_root(p->p_cwdi, vp, l); 1400 1401 out: 1402 fd_putfile(fd); 1403 return (error); 1404 } 1405 1406 /* 1407 * Change current working directory (``.''). 1408 */ 1409 /* ARGSUSED */ 1410 int 1411 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1412 { 1413 /* { 1414 syscallarg(const char *) path; 1415 } */ 1416 struct proc *p = l->l_proc; 1417 struct cwdinfo *cwdi; 1418 int error; 1419 struct vnode *vp; 1420 1421 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1422 &vp, l)) != 0) 1423 return (error); 1424 cwdi = p->p_cwdi; 1425 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1426 vrele(cwdi->cwdi_cdir); 1427 cwdi->cwdi_cdir = vp; 1428 rw_exit(&cwdi->cwdi_lock); 1429 return (0); 1430 } 1431 1432 /* 1433 * Change notion of root (``/'') directory. 1434 */ 1435 /* ARGSUSED */ 1436 int 1437 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1438 { 1439 /* { 1440 syscallarg(const char *) path; 1441 } */ 1442 struct proc *p = l->l_proc; 1443 int error; 1444 struct vnode *vp; 1445 1446 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1447 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1448 return (error); 1449 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1450 &vp, l)) != 0) 1451 return (error); 1452 1453 change_root(p->p_cwdi, vp, l); 1454 1455 return (0); 1456 } 1457 1458 /* 1459 * Common routine for chroot and fchroot. 1460 * NB: callers need to properly authorize the change root operation. 1461 */ 1462 void 1463 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1464 { 1465 struct proc *p = l->l_proc; 1466 kauth_cred_t ncred; 1467 1468 ncred = kauth_cred_alloc(); 1469 1470 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1471 if (cwdi->cwdi_rdir != NULL) 1472 vrele(cwdi->cwdi_rdir); 1473 cwdi->cwdi_rdir = vp; 1474 1475 /* 1476 * Prevent escaping from chroot by putting the root under 1477 * the working directory. Silently chdir to / if we aren't 1478 * already there. 1479 */ 1480 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1481 /* 1482 * XXX would be more failsafe to change directory to a 1483 * deadfs node here instead 1484 */ 1485 vrele(cwdi->cwdi_cdir); 1486 vref(vp); 1487 cwdi->cwdi_cdir = vp; 1488 } 1489 rw_exit(&cwdi->cwdi_lock); 1490 1491 /* Get a write lock on the process credential. */ 1492 proc_crmod_enter(); 1493 1494 kauth_cred_clone(p->p_cred, ncred); 1495 kauth_proc_chroot(ncred, p->p_cwdi); 1496 1497 /* Broadcast our credentials to the process and other LWPs. */ 1498 proc_crmod_leave(ncred, p->p_cred, true); 1499 } 1500 1501 /* 1502 * Common routine for chroot and chdir. 1503 * XXX "where" should be enum uio_seg 1504 */ 1505 int 1506 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1507 { 1508 struct pathbuf *pb; 1509 struct nameidata nd; 1510 int error; 1511 1512 error = pathbuf_maybe_copyin(path, where, &pb); 1513 if (error) { 1514 return error; 1515 } 1516 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1517 if ((error = namei(&nd)) != 0) { 1518 pathbuf_destroy(pb); 1519 return error; 1520 } 1521 *vpp = nd.ni_vp; 1522 pathbuf_destroy(pb); 1523 1524 if ((*vpp)->v_type != VDIR) 1525 error = ENOTDIR; 1526 else 1527 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1528 1529 if (error) 1530 vput(*vpp); 1531 else 1532 VOP_UNLOCK(*vpp); 1533 return (error); 1534 } 1535 1536 /* 1537 * Internals of sys_open - path has already been converted into a pathbuf 1538 * (so we can easily reuse this function from other parts of the kernel, 1539 * like posix_spawn post-processing). 1540 */ 1541 int 1542 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1543 int open_mode, int *fd) 1544 { 1545 struct proc *p = l->l_proc; 1546 struct cwdinfo *cwdi = p->p_cwdi; 1547 file_t *fp; 1548 struct vnode *vp; 1549 int flags, cmode; 1550 int indx, error; 1551 struct nameidata nd; 1552 1553 if (open_flags & O_SEARCH) { 1554 open_flags &= ~(int)O_SEARCH; 1555 } 1556 1557 flags = FFLAGS(open_flags); 1558 if ((flags & (FREAD | FWRITE)) == 0) 1559 return EINVAL; 1560 1561 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1562 return error; 1563 } 1564 1565 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1566 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1567 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1568 if (dvp != NULL) 1569 NDAT(&nd, dvp); 1570 1571 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1572 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1573 fd_abort(p, fp, indx); 1574 if ((error == EDUPFD || error == EMOVEFD) && 1575 l->l_dupfd >= 0 && /* XXX from fdopen */ 1576 (error = 1577 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1578 *fd = indx; 1579 return 0; 1580 } 1581 if (error == ERESTART) 1582 error = EINTR; 1583 return error; 1584 } 1585 1586 l->l_dupfd = 0; 1587 vp = nd.ni_vp; 1588 1589 if ((error = open_setfp(l, fp, vp, indx, flags))) 1590 return error; 1591 1592 VOP_UNLOCK(vp); 1593 *fd = indx; 1594 fd_affix(p, fp, indx); 1595 return 0; 1596 } 1597 1598 int 1599 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1600 { 1601 struct pathbuf *pb; 1602 int error, oflags; 1603 1604 oflags = FFLAGS(open_flags); 1605 if ((oflags & (FREAD | FWRITE)) == 0) 1606 return EINVAL; 1607 1608 pb = pathbuf_create(path); 1609 if (pb == NULL) 1610 return ENOMEM; 1611 1612 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1613 pathbuf_destroy(pb); 1614 1615 return error; 1616 } 1617 1618 /* 1619 * Check permissions, allocate an open file structure, 1620 * and call the device open routine if any. 1621 */ 1622 static int 1623 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1624 int mode, int *fd) 1625 { 1626 file_t *dfp = NULL; 1627 struct vnode *dvp = NULL; 1628 struct pathbuf *pb; 1629 int error; 1630 1631 #ifdef COMPAT_10 /* XXX: and perhaps later */ 1632 if (path == NULL) { 1633 pb = pathbuf_create("."); 1634 if (pb == NULL) 1635 return ENOMEM; 1636 } else 1637 #endif 1638 { 1639 error = pathbuf_copyin(path, &pb); 1640 if (error) 1641 return error; 1642 } 1643 1644 if (fdat != AT_FDCWD) { 1645 /* fd_getvnode() will use the descriptor for us */ 1646 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1647 goto out; 1648 1649 dvp = dfp->f_data; 1650 } 1651 1652 error = do_open(l, dvp, pb, flags, mode, fd); 1653 1654 if (dfp != NULL) 1655 fd_putfile(fdat); 1656 out: 1657 pathbuf_destroy(pb); 1658 return error; 1659 } 1660 1661 int 1662 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1663 { 1664 /* { 1665 syscallarg(const char *) path; 1666 syscallarg(int) flags; 1667 syscallarg(int) mode; 1668 } */ 1669 int error; 1670 int fd; 1671 1672 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1673 SCARG(uap, flags), SCARG(uap, mode), &fd); 1674 1675 if (error == 0) 1676 *retval = fd; 1677 1678 return error; 1679 } 1680 1681 int 1682 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1683 { 1684 /* { 1685 syscallarg(int) fd; 1686 syscallarg(const char *) path; 1687 syscallarg(int) oflags; 1688 syscallarg(int) mode; 1689 } */ 1690 int error; 1691 int fd; 1692 1693 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1694 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1695 1696 if (error == 0) 1697 *retval = fd; 1698 1699 return error; 1700 } 1701 1702 static void 1703 vfs__fhfree(fhandle_t *fhp) 1704 { 1705 size_t fhsize; 1706 1707 if (fhp == NULL) { 1708 return; 1709 } 1710 fhsize = FHANDLE_SIZE(fhp); 1711 kmem_free(fhp, fhsize); 1712 } 1713 1714 /* 1715 * vfs_composefh: compose a filehandle. 1716 */ 1717 1718 int 1719 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1720 { 1721 struct mount *mp; 1722 struct fid *fidp; 1723 int error; 1724 size_t needfhsize; 1725 size_t fidsize; 1726 1727 mp = vp->v_mount; 1728 fidp = NULL; 1729 if (*fh_size < FHANDLE_SIZE_MIN) { 1730 fidsize = 0; 1731 } else { 1732 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1733 if (fhp != NULL) { 1734 memset(fhp, 0, *fh_size); 1735 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1736 fidp = &fhp->fh_fid; 1737 } 1738 } 1739 error = VFS_VPTOFH(vp, fidp, &fidsize); 1740 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1741 if (error == 0 && *fh_size < needfhsize) { 1742 error = E2BIG; 1743 } 1744 *fh_size = needfhsize; 1745 return error; 1746 } 1747 1748 int 1749 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1750 { 1751 struct mount *mp; 1752 fhandle_t *fhp; 1753 size_t fhsize; 1754 size_t fidsize; 1755 int error; 1756 1757 *fhpp = NULL; 1758 mp = vp->v_mount; 1759 fidsize = 0; 1760 error = VFS_VPTOFH(vp, NULL, &fidsize); 1761 KASSERT(error != 0); 1762 if (error != E2BIG) { 1763 goto out; 1764 } 1765 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1766 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1767 if (fhp == NULL) { 1768 error = ENOMEM; 1769 goto out; 1770 } 1771 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1772 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1773 if (error == 0) { 1774 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1775 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1776 *fhpp = fhp; 1777 } else { 1778 kmem_free(fhp, fhsize); 1779 } 1780 out: 1781 return error; 1782 } 1783 1784 void 1785 vfs_composefh_free(fhandle_t *fhp) 1786 { 1787 1788 vfs__fhfree(fhp); 1789 } 1790 1791 /* 1792 * vfs_fhtovp: lookup a vnode by a filehandle. 1793 */ 1794 1795 int 1796 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1797 { 1798 struct mount *mp; 1799 int error; 1800 1801 *vpp = NULL; 1802 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1803 if (mp == NULL) { 1804 error = ESTALE; 1805 goto out; 1806 } 1807 if (mp->mnt_op->vfs_fhtovp == NULL) { 1808 error = EOPNOTSUPP; 1809 goto out; 1810 } 1811 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1812 out: 1813 return error; 1814 } 1815 1816 /* 1817 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1818 * the needed size. 1819 */ 1820 1821 int 1822 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1823 { 1824 fhandle_t *fhp; 1825 int error; 1826 1827 *fhpp = NULL; 1828 if (fhsize > FHANDLE_SIZE_MAX) { 1829 return EINVAL; 1830 } 1831 if (fhsize < FHANDLE_SIZE_MIN) { 1832 return EINVAL; 1833 } 1834 again: 1835 fhp = kmem_alloc(fhsize, KM_SLEEP); 1836 if (fhp == NULL) { 1837 return ENOMEM; 1838 } 1839 error = copyin(ufhp, fhp, fhsize); 1840 if (error == 0) { 1841 /* XXX this check shouldn't be here */ 1842 if (FHANDLE_SIZE(fhp) == fhsize) { 1843 *fhpp = fhp; 1844 return 0; 1845 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1846 /* 1847 * a kludge for nfsv2 padded handles. 1848 */ 1849 size_t sz; 1850 1851 sz = FHANDLE_SIZE(fhp); 1852 kmem_free(fhp, fhsize); 1853 fhsize = sz; 1854 goto again; 1855 } else { 1856 /* 1857 * userland told us wrong size. 1858 */ 1859 error = EINVAL; 1860 } 1861 } 1862 kmem_free(fhp, fhsize); 1863 return error; 1864 } 1865 1866 void 1867 vfs_copyinfh_free(fhandle_t *fhp) 1868 { 1869 1870 vfs__fhfree(fhp); 1871 } 1872 1873 /* 1874 * Get file handle system call 1875 */ 1876 int 1877 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1878 { 1879 /* { 1880 syscallarg(char *) fname; 1881 syscallarg(fhandle_t *) fhp; 1882 syscallarg(size_t *) fh_size; 1883 } */ 1884 struct vnode *vp; 1885 fhandle_t *fh; 1886 int error; 1887 struct pathbuf *pb; 1888 struct nameidata nd; 1889 size_t sz; 1890 size_t usz; 1891 1892 /* 1893 * Must be super user 1894 */ 1895 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1896 0, NULL, NULL, NULL); 1897 if (error) 1898 return (error); 1899 1900 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1901 if (error) { 1902 return error; 1903 } 1904 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1905 error = namei(&nd); 1906 if (error) { 1907 pathbuf_destroy(pb); 1908 return error; 1909 } 1910 vp = nd.ni_vp; 1911 pathbuf_destroy(pb); 1912 1913 error = vfs_composefh_alloc(vp, &fh); 1914 vput(vp); 1915 if (error != 0) { 1916 goto out; 1917 } 1918 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1919 if (error != 0) { 1920 goto out; 1921 } 1922 sz = FHANDLE_SIZE(fh); 1923 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1924 if (error != 0) { 1925 goto out; 1926 } 1927 if (usz >= sz) { 1928 error = copyout(fh, SCARG(uap, fhp), sz); 1929 } else { 1930 error = E2BIG; 1931 } 1932 out: 1933 vfs_composefh_free(fh); 1934 return (error); 1935 } 1936 1937 /* 1938 * Open a file given a file handle. 1939 * 1940 * Check permissions, allocate an open file structure, 1941 * and call the device open routine if any. 1942 */ 1943 1944 int 1945 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1946 register_t *retval) 1947 { 1948 file_t *fp; 1949 struct vnode *vp = NULL; 1950 kauth_cred_t cred = l->l_cred; 1951 file_t *nfp; 1952 int indx, error = 0; 1953 struct vattr va; 1954 fhandle_t *fh; 1955 int flags; 1956 proc_t *p; 1957 1958 p = curproc; 1959 1960 /* 1961 * Must be super user 1962 */ 1963 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1964 0, NULL, NULL, NULL))) 1965 return (error); 1966 1967 if (oflags & O_SEARCH) { 1968 oflags &= ~(int)O_SEARCH; 1969 } 1970 1971 flags = FFLAGS(oflags); 1972 if ((flags & (FREAD | FWRITE)) == 0) 1973 return (EINVAL); 1974 if ((flags & O_CREAT)) 1975 return (EINVAL); 1976 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1977 return (error); 1978 fp = nfp; 1979 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1980 if (error != 0) { 1981 goto bad; 1982 } 1983 error = vfs_fhtovp(fh, &vp); 1984 if (error != 0) { 1985 goto bad; 1986 } 1987 1988 /* Now do an effective vn_open */ 1989 1990 if (vp->v_type == VSOCK) { 1991 error = EOPNOTSUPP; 1992 goto bad; 1993 } 1994 error = vn_openchk(vp, cred, flags); 1995 if (error != 0) 1996 goto bad; 1997 if (flags & O_TRUNC) { 1998 VOP_UNLOCK(vp); /* XXX */ 1999 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2000 vattr_null(&va); 2001 va.va_size = 0; 2002 error = VOP_SETATTR(vp, &va, cred); 2003 if (error) 2004 goto bad; 2005 } 2006 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2007 goto bad; 2008 if (flags & FWRITE) { 2009 mutex_enter(vp->v_interlock); 2010 vp->v_writecount++; 2011 mutex_exit(vp->v_interlock); 2012 } 2013 2014 /* done with modified vn_open, now finish what sys_open does. */ 2015 if ((error = open_setfp(l, fp, vp, indx, flags))) 2016 return error; 2017 2018 VOP_UNLOCK(vp); 2019 *retval = indx; 2020 fd_affix(p, fp, indx); 2021 vfs_copyinfh_free(fh); 2022 return (0); 2023 2024 bad: 2025 fd_abort(p, fp, indx); 2026 if (vp != NULL) 2027 vput(vp); 2028 vfs_copyinfh_free(fh); 2029 return (error); 2030 } 2031 2032 int 2033 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2034 { 2035 /* { 2036 syscallarg(const void *) fhp; 2037 syscallarg(size_t) fh_size; 2038 syscallarg(int) flags; 2039 } */ 2040 2041 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2042 SCARG(uap, flags), retval); 2043 } 2044 2045 int 2046 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2047 { 2048 int error; 2049 fhandle_t *fh; 2050 struct vnode *vp; 2051 2052 /* 2053 * Must be super user 2054 */ 2055 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2056 0, NULL, NULL, NULL))) 2057 return (error); 2058 2059 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2060 if (error != 0) 2061 return error; 2062 2063 error = vfs_fhtovp(fh, &vp); 2064 vfs_copyinfh_free(fh); 2065 if (error != 0) 2066 return error; 2067 2068 error = vn_stat(vp, sb); 2069 vput(vp); 2070 return error; 2071 } 2072 2073 2074 /* ARGSUSED */ 2075 int 2076 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2077 { 2078 /* { 2079 syscallarg(const void *) fhp; 2080 syscallarg(size_t) fh_size; 2081 syscallarg(struct stat *) sb; 2082 } */ 2083 struct stat sb; 2084 int error; 2085 2086 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2087 if (error) 2088 return error; 2089 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2090 } 2091 2092 int 2093 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2094 int flags) 2095 { 2096 fhandle_t *fh; 2097 struct mount *mp; 2098 struct vnode *vp; 2099 int error; 2100 2101 /* 2102 * Must be super user 2103 */ 2104 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2105 0, NULL, NULL, NULL))) 2106 return error; 2107 2108 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2109 if (error != 0) 2110 return error; 2111 2112 error = vfs_fhtovp(fh, &vp); 2113 vfs_copyinfh_free(fh); 2114 if (error != 0) 2115 return error; 2116 2117 mp = vp->v_mount; 2118 error = dostatvfs(mp, sb, l, flags, 1); 2119 vput(vp); 2120 return error; 2121 } 2122 2123 /* ARGSUSED */ 2124 int 2125 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2126 { 2127 /* { 2128 syscallarg(const void *) fhp; 2129 syscallarg(size_t) fh_size; 2130 syscallarg(struct statvfs *) buf; 2131 syscallarg(int) flags; 2132 } */ 2133 struct statvfs *sb = STATVFSBUF_GET(); 2134 int error; 2135 2136 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2137 SCARG(uap, flags)); 2138 if (error == 0) 2139 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2140 STATVFSBUF_PUT(sb); 2141 return error; 2142 } 2143 2144 /* 2145 * Create a special file. 2146 */ 2147 /* ARGSUSED */ 2148 int 2149 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2150 register_t *retval) 2151 { 2152 /* { 2153 syscallarg(const char *) path; 2154 syscallarg(mode_t) mode; 2155 syscallarg(dev_t) dev; 2156 } */ 2157 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2158 SCARG(uap, dev), retval, UIO_USERSPACE); 2159 } 2160 2161 int 2162 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2163 register_t *retval) 2164 { 2165 /* { 2166 syscallarg(int) fd; 2167 syscallarg(const char *) path; 2168 syscallarg(mode_t) mode; 2169 syscallarg(int) pad; 2170 syscallarg(dev_t) dev; 2171 } */ 2172 2173 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2174 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2175 } 2176 2177 int 2178 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2179 register_t *retval, enum uio_seg seg) 2180 { 2181 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2182 } 2183 2184 int 2185 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2186 dev_t dev, register_t *retval, enum uio_seg seg) 2187 { 2188 struct proc *p = l->l_proc; 2189 struct vnode *vp; 2190 struct vattr vattr; 2191 int error, optype; 2192 struct pathbuf *pb; 2193 struct nameidata nd; 2194 const char *pathstring; 2195 2196 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2197 0, NULL, NULL, NULL)) != 0) 2198 return (error); 2199 2200 optype = VOP_MKNOD_DESCOFFSET; 2201 2202 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2203 if (error) { 2204 return error; 2205 } 2206 pathstring = pathbuf_stringcopy_get(pb); 2207 if (pathstring == NULL) { 2208 pathbuf_destroy(pb); 2209 return ENOMEM; 2210 } 2211 2212 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2213 2214 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2215 goto out; 2216 vp = nd.ni_vp; 2217 2218 if (vp != NULL) 2219 error = EEXIST; 2220 else { 2221 vattr_null(&vattr); 2222 /* We will read cwdi->cwdi_cmask unlocked. */ 2223 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2224 vattr.va_rdev = dev; 2225 2226 switch (mode & S_IFMT) { 2227 case S_IFMT: /* used by badsect to flag bad sectors */ 2228 vattr.va_type = VBAD; 2229 break; 2230 case S_IFCHR: 2231 vattr.va_type = VCHR; 2232 break; 2233 case S_IFBLK: 2234 vattr.va_type = VBLK; 2235 break; 2236 case S_IFWHT: 2237 optype = VOP_WHITEOUT_DESCOFFSET; 2238 break; 2239 case S_IFREG: 2240 #if NVERIEXEC > 0 2241 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2242 O_CREAT); 2243 #endif /* NVERIEXEC > 0 */ 2244 vattr.va_type = VREG; 2245 vattr.va_rdev = VNOVAL; 2246 optype = VOP_CREATE_DESCOFFSET; 2247 break; 2248 default: 2249 error = EINVAL; 2250 break; 2251 } 2252 } 2253 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2254 && vattr.va_rdev == VNOVAL) 2255 error = EINVAL; 2256 if (!error) { 2257 switch (optype) { 2258 case VOP_WHITEOUT_DESCOFFSET: 2259 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2260 if (error) 2261 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2262 vput(nd.ni_dvp); 2263 break; 2264 2265 case VOP_MKNOD_DESCOFFSET: 2266 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2267 &nd.ni_cnd, &vattr); 2268 if (error == 0) 2269 vrele(nd.ni_vp); 2270 vput(nd.ni_dvp); 2271 break; 2272 2273 case VOP_CREATE_DESCOFFSET: 2274 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2275 &nd.ni_cnd, &vattr); 2276 if (error == 0) 2277 vrele(nd.ni_vp); 2278 vput(nd.ni_dvp); 2279 break; 2280 } 2281 } else { 2282 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2283 if (nd.ni_dvp == vp) 2284 vrele(nd.ni_dvp); 2285 else 2286 vput(nd.ni_dvp); 2287 if (vp) 2288 vrele(vp); 2289 } 2290 out: 2291 pathbuf_stringcopy_put(pb, pathstring); 2292 pathbuf_destroy(pb); 2293 return (error); 2294 } 2295 2296 /* 2297 * Create a named pipe. 2298 */ 2299 /* ARGSUSED */ 2300 int 2301 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2302 { 2303 /* { 2304 syscallarg(const char *) path; 2305 syscallarg(int) mode; 2306 } */ 2307 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2308 } 2309 2310 int 2311 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2312 register_t *retval) 2313 { 2314 /* { 2315 syscallarg(int) fd; 2316 syscallarg(const char *) path; 2317 syscallarg(int) mode; 2318 } */ 2319 2320 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2321 SCARG(uap, mode)); 2322 } 2323 2324 static int 2325 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2326 { 2327 struct proc *p = l->l_proc; 2328 struct vattr vattr; 2329 int error; 2330 struct pathbuf *pb; 2331 struct nameidata nd; 2332 2333 error = pathbuf_copyin(path, &pb); 2334 if (error) { 2335 return error; 2336 } 2337 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2338 2339 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2340 pathbuf_destroy(pb); 2341 return error; 2342 } 2343 if (nd.ni_vp != NULL) { 2344 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2345 if (nd.ni_dvp == nd.ni_vp) 2346 vrele(nd.ni_dvp); 2347 else 2348 vput(nd.ni_dvp); 2349 vrele(nd.ni_vp); 2350 pathbuf_destroy(pb); 2351 return (EEXIST); 2352 } 2353 vattr_null(&vattr); 2354 vattr.va_type = VFIFO; 2355 /* We will read cwdi->cwdi_cmask unlocked. */ 2356 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2357 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2358 if (error == 0) 2359 vrele(nd.ni_vp); 2360 vput(nd.ni_dvp); 2361 pathbuf_destroy(pb); 2362 return (error); 2363 } 2364 2365 /* 2366 * Make a hard file link. 2367 */ 2368 /* ARGSUSED */ 2369 int 2370 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2371 const char *link, int follow, register_t *retval) 2372 { 2373 struct vnode *vp; 2374 struct pathbuf *linkpb; 2375 struct nameidata nd; 2376 namei_simple_flags_t ns_flags; 2377 int error; 2378 2379 if (follow & AT_SYMLINK_FOLLOW) 2380 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2381 else 2382 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2383 2384 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2385 if (error != 0) 2386 return (error); 2387 error = pathbuf_copyin(link, &linkpb); 2388 if (error) { 2389 goto out1; 2390 } 2391 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2392 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2393 goto out2; 2394 if (nd.ni_vp) { 2395 error = EEXIST; 2396 goto abortop; 2397 } 2398 /* Prevent hard links on directories. */ 2399 if (vp->v_type == VDIR) { 2400 error = EPERM; 2401 goto abortop; 2402 } 2403 /* Prevent cross-mount operation. */ 2404 if (nd.ni_dvp->v_mount != vp->v_mount) { 2405 error = EXDEV; 2406 goto abortop; 2407 } 2408 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2409 out2: 2410 pathbuf_destroy(linkpb); 2411 out1: 2412 vrele(vp); 2413 return (error); 2414 abortop: 2415 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2416 if (nd.ni_dvp == nd.ni_vp) 2417 vrele(nd.ni_dvp); 2418 else 2419 vput(nd.ni_dvp); 2420 if (nd.ni_vp != NULL) 2421 vrele(nd.ni_vp); 2422 goto out2; 2423 } 2424 2425 int 2426 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2427 { 2428 /* { 2429 syscallarg(const char *) path; 2430 syscallarg(const char *) link; 2431 } */ 2432 const char *path = SCARG(uap, path); 2433 const char *link = SCARG(uap, link); 2434 2435 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2436 AT_SYMLINK_FOLLOW, retval); 2437 } 2438 2439 int 2440 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2441 register_t *retval) 2442 { 2443 /* { 2444 syscallarg(int) fd1; 2445 syscallarg(const char *) name1; 2446 syscallarg(int) fd2; 2447 syscallarg(const char *) name2; 2448 syscallarg(int) flags; 2449 } */ 2450 int fd1 = SCARG(uap, fd1); 2451 const char *name1 = SCARG(uap, name1); 2452 int fd2 = SCARG(uap, fd2); 2453 const char *name2 = SCARG(uap, name2); 2454 int follow; 2455 2456 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2457 2458 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2459 } 2460 2461 2462 int 2463 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2464 { 2465 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2466 } 2467 2468 static int 2469 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2470 const char *link, enum uio_seg seg) 2471 { 2472 struct proc *p = curproc; 2473 struct vattr vattr; 2474 char *path; 2475 int error; 2476 struct pathbuf *linkpb; 2477 struct nameidata nd; 2478 2479 KASSERT(l != NULL || fdat == AT_FDCWD); 2480 2481 path = PNBUF_GET(); 2482 if (seg == UIO_USERSPACE) { 2483 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2484 goto out1; 2485 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2486 goto out1; 2487 } else { 2488 KASSERT(strlen(patharg) < MAXPATHLEN); 2489 strcpy(path, patharg); 2490 linkpb = pathbuf_create(link); 2491 if (linkpb == NULL) { 2492 error = ENOMEM; 2493 goto out1; 2494 } 2495 } 2496 ktrkuser("symlink-target", path, strlen(path)); 2497 2498 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2499 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2500 goto out2; 2501 if (nd.ni_vp) { 2502 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2503 if (nd.ni_dvp == nd.ni_vp) 2504 vrele(nd.ni_dvp); 2505 else 2506 vput(nd.ni_dvp); 2507 vrele(nd.ni_vp); 2508 error = EEXIST; 2509 goto out2; 2510 } 2511 vattr_null(&vattr); 2512 vattr.va_type = VLNK; 2513 /* We will read cwdi->cwdi_cmask unlocked. */ 2514 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2515 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2516 if (error == 0) 2517 vrele(nd.ni_vp); 2518 vput(nd.ni_dvp); 2519 out2: 2520 pathbuf_destroy(linkpb); 2521 out1: 2522 PNBUF_PUT(path); 2523 return (error); 2524 } 2525 2526 /* 2527 * Make a symbolic link. 2528 */ 2529 /* ARGSUSED */ 2530 int 2531 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2532 { 2533 /* { 2534 syscallarg(const char *) path; 2535 syscallarg(const char *) link; 2536 } */ 2537 2538 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2539 UIO_USERSPACE); 2540 } 2541 2542 int 2543 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2544 register_t *retval) 2545 { 2546 /* { 2547 syscallarg(const char *) path1; 2548 syscallarg(int) fd; 2549 syscallarg(const char *) path2; 2550 } */ 2551 2552 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2553 SCARG(uap, path2), UIO_USERSPACE); 2554 } 2555 2556 /* 2557 * Delete a whiteout from the filesystem. 2558 */ 2559 /* ARGSUSED */ 2560 int 2561 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2562 { 2563 /* { 2564 syscallarg(const char *) path; 2565 } */ 2566 int error; 2567 struct pathbuf *pb; 2568 struct nameidata nd; 2569 2570 error = pathbuf_copyin(SCARG(uap, path), &pb); 2571 if (error) { 2572 return error; 2573 } 2574 2575 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2576 error = namei(&nd); 2577 if (error) { 2578 pathbuf_destroy(pb); 2579 return (error); 2580 } 2581 2582 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2583 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2584 if (nd.ni_dvp == nd.ni_vp) 2585 vrele(nd.ni_dvp); 2586 else 2587 vput(nd.ni_dvp); 2588 if (nd.ni_vp) 2589 vrele(nd.ni_vp); 2590 pathbuf_destroy(pb); 2591 return (EEXIST); 2592 } 2593 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2594 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2595 vput(nd.ni_dvp); 2596 pathbuf_destroy(pb); 2597 return (error); 2598 } 2599 2600 /* 2601 * Delete a name from the filesystem. 2602 */ 2603 /* ARGSUSED */ 2604 int 2605 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2606 { 2607 /* { 2608 syscallarg(const char *) path; 2609 } */ 2610 2611 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2612 } 2613 2614 int 2615 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2616 register_t *retval) 2617 { 2618 /* { 2619 syscallarg(int) fd; 2620 syscallarg(const char *) path; 2621 syscallarg(int) flag; 2622 } */ 2623 2624 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2625 SCARG(uap, flag), UIO_USERSPACE); 2626 } 2627 2628 int 2629 do_sys_unlink(const char *arg, enum uio_seg seg) 2630 { 2631 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2632 } 2633 2634 static int 2635 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2636 enum uio_seg seg) 2637 { 2638 struct vnode *vp; 2639 int error; 2640 struct pathbuf *pb; 2641 struct nameidata nd; 2642 const char *pathstring; 2643 2644 KASSERT(l != NULL || fdat == AT_FDCWD); 2645 2646 error = pathbuf_maybe_copyin(arg, seg, &pb); 2647 if (error) { 2648 return error; 2649 } 2650 pathstring = pathbuf_stringcopy_get(pb); 2651 if (pathstring == NULL) { 2652 pathbuf_destroy(pb); 2653 return ENOMEM; 2654 } 2655 2656 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2657 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2658 goto out; 2659 vp = nd.ni_vp; 2660 2661 /* 2662 * The root of a mounted filesystem cannot be deleted. 2663 */ 2664 if ((vp->v_vflag & VV_ROOT) != 0) { 2665 error = EBUSY; 2666 goto abort; 2667 } 2668 2669 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2670 error = EBUSY; 2671 goto abort; 2672 } 2673 2674 /* 2675 * No rmdir "." please. 2676 */ 2677 if (nd.ni_dvp == vp) { 2678 error = EINVAL; 2679 goto abort; 2680 } 2681 2682 /* 2683 * AT_REMOVEDIR is required to remove a directory 2684 */ 2685 if (vp->v_type == VDIR) { 2686 if (!(flags & AT_REMOVEDIR)) { 2687 error = EPERM; 2688 goto abort; 2689 } else { 2690 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2691 goto out; 2692 } 2693 } 2694 2695 /* 2696 * Starting here we only deal with non directories. 2697 */ 2698 if (flags & AT_REMOVEDIR) { 2699 error = ENOTDIR; 2700 goto abort; 2701 } 2702 2703 2704 #if NVERIEXEC > 0 2705 /* Handle remove requests for veriexec entries. */ 2706 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2707 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2708 if (nd.ni_dvp == vp) 2709 vrele(nd.ni_dvp); 2710 else 2711 vput(nd.ni_dvp); 2712 vput(vp); 2713 goto out; 2714 } 2715 #endif /* NVERIEXEC > 0 */ 2716 2717 #ifdef FILEASSOC 2718 (void)fileassoc_file_delete(vp); 2719 #endif /* FILEASSOC */ 2720 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2721 goto out; 2722 2723 abort: 2724 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2725 if (nd.ni_dvp == vp) 2726 vrele(nd.ni_dvp); 2727 else 2728 vput(nd.ni_dvp); 2729 vput(vp); 2730 2731 out: 2732 pathbuf_stringcopy_put(pb, pathstring); 2733 pathbuf_destroy(pb); 2734 return (error); 2735 } 2736 2737 /* 2738 * Reposition read/write file offset. 2739 */ 2740 int 2741 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2742 { 2743 /* { 2744 syscallarg(int) fd; 2745 syscallarg(int) pad; 2746 syscallarg(off_t) offset; 2747 syscallarg(int) whence; 2748 } */ 2749 kauth_cred_t cred = l->l_cred; 2750 file_t *fp; 2751 struct vnode *vp; 2752 struct vattr vattr; 2753 off_t newoff; 2754 int error, fd; 2755 2756 fd = SCARG(uap, fd); 2757 2758 if ((fp = fd_getfile(fd)) == NULL) 2759 return (EBADF); 2760 2761 vp = fp->f_data; 2762 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2763 error = ESPIPE; 2764 goto out; 2765 } 2766 2767 switch (SCARG(uap, whence)) { 2768 case SEEK_CUR: 2769 newoff = fp->f_offset + SCARG(uap, offset); 2770 break; 2771 case SEEK_END: 2772 vn_lock(vp, LK_SHARED | LK_RETRY); 2773 error = VOP_GETATTR(vp, &vattr, cred); 2774 VOP_UNLOCK(vp); 2775 if (error) { 2776 goto out; 2777 } 2778 newoff = SCARG(uap, offset) + vattr.va_size; 2779 break; 2780 case SEEK_SET: 2781 newoff = SCARG(uap, offset); 2782 break; 2783 default: 2784 error = EINVAL; 2785 goto out; 2786 } 2787 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2788 *(off_t *)retval = fp->f_offset = newoff; 2789 } 2790 out: 2791 fd_putfile(fd); 2792 return (error); 2793 } 2794 2795 /* 2796 * Positional read system call. 2797 */ 2798 int 2799 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2800 { 2801 /* { 2802 syscallarg(int) fd; 2803 syscallarg(void *) buf; 2804 syscallarg(size_t) nbyte; 2805 syscallarg(off_t) offset; 2806 } */ 2807 file_t *fp; 2808 struct vnode *vp; 2809 off_t offset; 2810 int error, fd = SCARG(uap, fd); 2811 2812 if ((fp = fd_getfile(fd)) == NULL) 2813 return (EBADF); 2814 2815 if ((fp->f_flag & FREAD) == 0) { 2816 fd_putfile(fd); 2817 return (EBADF); 2818 } 2819 2820 vp = fp->f_data; 2821 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2822 error = ESPIPE; 2823 goto out; 2824 } 2825 2826 offset = SCARG(uap, offset); 2827 2828 /* 2829 * XXX This works because no file systems actually 2830 * XXX take any action on the seek operation. 2831 */ 2832 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2833 goto out; 2834 2835 /* dofileread() will unuse the descriptor for us */ 2836 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2837 &offset, 0, retval)); 2838 2839 out: 2840 fd_putfile(fd); 2841 return (error); 2842 } 2843 2844 /* 2845 * Positional scatter read system call. 2846 */ 2847 int 2848 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2849 { 2850 /* { 2851 syscallarg(int) fd; 2852 syscallarg(const struct iovec *) iovp; 2853 syscallarg(int) iovcnt; 2854 syscallarg(off_t) offset; 2855 } */ 2856 off_t offset = SCARG(uap, offset); 2857 2858 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2859 SCARG(uap, iovcnt), &offset, 0, retval); 2860 } 2861 2862 /* 2863 * Positional write system call. 2864 */ 2865 int 2866 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2867 { 2868 /* { 2869 syscallarg(int) fd; 2870 syscallarg(const void *) buf; 2871 syscallarg(size_t) nbyte; 2872 syscallarg(off_t) offset; 2873 } */ 2874 file_t *fp; 2875 struct vnode *vp; 2876 off_t offset; 2877 int error, fd = SCARG(uap, fd); 2878 2879 if ((fp = fd_getfile(fd)) == NULL) 2880 return (EBADF); 2881 2882 if ((fp->f_flag & FWRITE) == 0) { 2883 fd_putfile(fd); 2884 return (EBADF); 2885 } 2886 2887 vp = fp->f_data; 2888 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2889 error = ESPIPE; 2890 goto out; 2891 } 2892 2893 offset = SCARG(uap, offset); 2894 2895 /* 2896 * XXX This works because no file systems actually 2897 * XXX take any action on the seek operation. 2898 */ 2899 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2900 goto out; 2901 2902 /* dofilewrite() will unuse the descriptor for us */ 2903 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2904 &offset, 0, retval)); 2905 2906 out: 2907 fd_putfile(fd); 2908 return (error); 2909 } 2910 2911 /* 2912 * Positional gather write system call. 2913 */ 2914 int 2915 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2916 { 2917 /* { 2918 syscallarg(int) fd; 2919 syscallarg(const struct iovec *) iovp; 2920 syscallarg(int) iovcnt; 2921 syscallarg(off_t) offset; 2922 } */ 2923 off_t offset = SCARG(uap, offset); 2924 2925 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2926 SCARG(uap, iovcnt), &offset, 0, retval); 2927 } 2928 2929 /* 2930 * Check access permissions. 2931 */ 2932 int 2933 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2934 { 2935 /* { 2936 syscallarg(const char *) path; 2937 syscallarg(int) flags; 2938 } */ 2939 2940 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2941 SCARG(uap, flags), 0); 2942 } 2943 2944 int 2945 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2946 int mode, int flags) 2947 { 2948 kauth_cred_t cred; 2949 struct vnode *vp; 2950 int error, nd_flag, vmode; 2951 struct pathbuf *pb; 2952 struct nameidata nd; 2953 2954 CTASSERT(F_OK == 0); 2955 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2956 /* nonsense mode */ 2957 return EINVAL; 2958 } 2959 2960 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2961 if (flags & AT_SYMLINK_NOFOLLOW) 2962 nd_flag &= ~FOLLOW; 2963 2964 error = pathbuf_copyin(path, &pb); 2965 if (error) 2966 return error; 2967 2968 NDINIT(&nd, LOOKUP, nd_flag, pb); 2969 2970 /* Override default credentials */ 2971 cred = kauth_cred_dup(l->l_cred); 2972 if (!(flags & AT_EACCESS)) { 2973 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2974 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2975 } 2976 nd.ni_cnd.cn_cred = cred; 2977 2978 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2979 pathbuf_destroy(pb); 2980 goto out; 2981 } 2982 vp = nd.ni_vp; 2983 pathbuf_destroy(pb); 2984 2985 /* Flags == 0 means only check for existence. */ 2986 if (mode) { 2987 vmode = 0; 2988 if (mode & R_OK) 2989 vmode |= VREAD; 2990 if (mode & W_OK) 2991 vmode |= VWRITE; 2992 if (mode & X_OK) 2993 vmode |= VEXEC; 2994 2995 error = VOP_ACCESS(vp, vmode, cred); 2996 if (!error && (vmode & VWRITE)) 2997 error = vn_writechk(vp); 2998 } 2999 vput(vp); 3000 out: 3001 kauth_cred_free(cred); 3002 return (error); 3003 } 3004 3005 int 3006 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3007 register_t *retval) 3008 { 3009 /* { 3010 syscallarg(int) fd; 3011 syscallarg(const char *) path; 3012 syscallarg(int) amode; 3013 syscallarg(int) flag; 3014 } */ 3015 3016 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3017 SCARG(uap, amode), SCARG(uap, flag)); 3018 } 3019 3020 /* 3021 * Common code for all sys_stat functions, including compat versions. 3022 */ 3023 int 3024 do_sys_stat(const char *userpath, unsigned int nd_flag, 3025 struct stat *sb) 3026 { 3027 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3028 } 3029 3030 int 3031 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3032 unsigned int nd_flag, struct stat *sb) 3033 { 3034 int error; 3035 struct pathbuf *pb; 3036 struct nameidata nd; 3037 3038 KASSERT(l != NULL || fdat == AT_FDCWD); 3039 3040 error = pathbuf_copyin(userpath, &pb); 3041 if (error) { 3042 return error; 3043 } 3044 3045 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3046 3047 error = fd_nameiat(l, fdat, &nd); 3048 if (error != 0) { 3049 pathbuf_destroy(pb); 3050 return error; 3051 } 3052 error = vn_stat(nd.ni_vp, sb); 3053 vput(nd.ni_vp); 3054 pathbuf_destroy(pb); 3055 return error; 3056 } 3057 3058 /* 3059 * Get file status; this version follows links. 3060 */ 3061 /* ARGSUSED */ 3062 int 3063 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3064 { 3065 /* { 3066 syscallarg(const char *) path; 3067 syscallarg(struct stat *) ub; 3068 } */ 3069 struct stat sb; 3070 int error; 3071 3072 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3073 if (error) 3074 return error; 3075 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3076 } 3077 3078 /* 3079 * Get file status; this version does not follow links. 3080 */ 3081 /* ARGSUSED */ 3082 int 3083 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3084 { 3085 /* { 3086 syscallarg(const char *) path; 3087 syscallarg(struct stat *) ub; 3088 } */ 3089 struct stat sb; 3090 int error; 3091 3092 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3093 if (error) 3094 return error; 3095 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3096 } 3097 3098 int 3099 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3100 register_t *retval) 3101 { 3102 /* { 3103 syscallarg(int) fd; 3104 syscallarg(const char *) path; 3105 syscallarg(struct stat *) buf; 3106 syscallarg(int) flag; 3107 } */ 3108 unsigned int nd_flag; 3109 struct stat sb; 3110 int error; 3111 3112 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3113 nd_flag = NOFOLLOW; 3114 else 3115 nd_flag = FOLLOW; 3116 3117 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3118 &sb); 3119 if (error) 3120 return error; 3121 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3122 } 3123 3124 /* 3125 * Get configurable pathname variables. 3126 */ 3127 /* ARGSUSED */ 3128 int 3129 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3130 { 3131 /* { 3132 syscallarg(const char *) path; 3133 syscallarg(int) name; 3134 } */ 3135 int error; 3136 struct pathbuf *pb; 3137 struct nameidata nd; 3138 3139 error = pathbuf_copyin(SCARG(uap, path), &pb); 3140 if (error) { 3141 return error; 3142 } 3143 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3144 if ((error = namei(&nd)) != 0) { 3145 pathbuf_destroy(pb); 3146 return (error); 3147 } 3148 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3149 vput(nd.ni_vp); 3150 pathbuf_destroy(pb); 3151 return (error); 3152 } 3153 3154 /* 3155 * Return target name of a symbolic link. 3156 */ 3157 /* ARGSUSED */ 3158 int 3159 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3160 register_t *retval) 3161 { 3162 /* { 3163 syscallarg(const char *) path; 3164 syscallarg(char *) buf; 3165 syscallarg(size_t) count; 3166 } */ 3167 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3168 SCARG(uap, buf), SCARG(uap, count), retval); 3169 } 3170 3171 static int 3172 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3173 size_t count, register_t *retval) 3174 { 3175 struct vnode *vp; 3176 struct iovec aiov; 3177 struct uio auio; 3178 int error; 3179 struct pathbuf *pb; 3180 struct nameidata nd; 3181 3182 error = pathbuf_copyin(path, &pb); 3183 if (error) { 3184 return error; 3185 } 3186 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3187 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3188 pathbuf_destroy(pb); 3189 return error; 3190 } 3191 vp = nd.ni_vp; 3192 pathbuf_destroy(pb); 3193 if (vp->v_type != VLNK) 3194 error = EINVAL; 3195 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3196 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3197 aiov.iov_base = buf; 3198 aiov.iov_len = count; 3199 auio.uio_iov = &aiov; 3200 auio.uio_iovcnt = 1; 3201 auio.uio_offset = 0; 3202 auio.uio_rw = UIO_READ; 3203 KASSERT(l == curlwp); 3204 auio.uio_vmspace = l->l_proc->p_vmspace; 3205 auio.uio_resid = count; 3206 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3207 *retval = count - auio.uio_resid; 3208 } 3209 vput(vp); 3210 return (error); 3211 } 3212 3213 int 3214 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3215 register_t *retval) 3216 { 3217 /* { 3218 syscallarg(int) fd; 3219 syscallarg(const char *) path; 3220 syscallarg(char *) buf; 3221 syscallarg(size_t) bufsize; 3222 } */ 3223 3224 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3225 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3226 } 3227 3228 /* 3229 * Change flags of a file given a path name. 3230 */ 3231 /* ARGSUSED */ 3232 int 3233 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3234 { 3235 /* { 3236 syscallarg(const char *) path; 3237 syscallarg(u_long) flags; 3238 } */ 3239 struct vnode *vp; 3240 int error; 3241 3242 error = namei_simple_user(SCARG(uap, path), 3243 NSM_FOLLOW_TRYEMULROOT, &vp); 3244 if (error != 0) 3245 return (error); 3246 error = change_flags(vp, SCARG(uap, flags), l); 3247 vput(vp); 3248 return (error); 3249 } 3250 3251 /* 3252 * Change flags of a file given a file descriptor. 3253 */ 3254 /* ARGSUSED */ 3255 int 3256 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3257 { 3258 /* { 3259 syscallarg(int) fd; 3260 syscallarg(u_long) flags; 3261 } */ 3262 struct vnode *vp; 3263 file_t *fp; 3264 int error; 3265 3266 /* fd_getvnode() will use the descriptor for us */ 3267 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3268 return (error); 3269 vp = fp->f_data; 3270 error = change_flags(vp, SCARG(uap, flags), l); 3271 VOP_UNLOCK(vp); 3272 fd_putfile(SCARG(uap, fd)); 3273 return (error); 3274 } 3275 3276 /* 3277 * Change flags of a file given a path name; this version does 3278 * not follow links. 3279 */ 3280 int 3281 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3282 { 3283 /* { 3284 syscallarg(const char *) path; 3285 syscallarg(u_long) flags; 3286 } */ 3287 struct vnode *vp; 3288 int error; 3289 3290 error = namei_simple_user(SCARG(uap, path), 3291 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3292 if (error != 0) 3293 return (error); 3294 error = change_flags(vp, SCARG(uap, flags), l); 3295 vput(vp); 3296 return (error); 3297 } 3298 3299 /* 3300 * Common routine to change flags of a file. 3301 */ 3302 int 3303 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3304 { 3305 struct vattr vattr; 3306 int error; 3307 3308 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3309 3310 vattr_null(&vattr); 3311 vattr.va_flags = flags; 3312 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3313 3314 return (error); 3315 } 3316 3317 /* 3318 * Change mode of a file given path name; this version follows links. 3319 */ 3320 /* ARGSUSED */ 3321 int 3322 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3323 { 3324 /* { 3325 syscallarg(const char *) path; 3326 syscallarg(int) mode; 3327 } */ 3328 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3329 SCARG(uap, mode), 0); 3330 } 3331 3332 int 3333 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3334 { 3335 int error; 3336 struct vnode *vp; 3337 namei_simple_flags_t ns_flag; 3338 3339 if (flags & AT_SYMLINK_NOFOLLOW) 3340 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3341 else 3342 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3343 3344 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3345 if (error != 0) 3346 return error; 3347 3348 error = change_mode(vp, mode, l); 3349 3350 vrele(vp); 3351 3352 return (error); 3353 } 3354 3355 /* 3356 * Change mode of a file given a file descriptor. 3357 */ 3358 /* ARGSUSED */ 3359 int 3360 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3361 { 3362 /* { 3363 syscallarg(int) fd; 3364 syscallarg(int) mode; 3365 } */ 3366 file_t *fp; 3367 int error; 3368 3369 /* fd_getvnode() will use the descriptor for us */ 3370 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3371 return (error); 3372 error = change_mode(fp->f_data, SCARG(uap, mode), l); 3373 fd_putfile(SCARG(uap, fd)); 3374 return (error); 3375 } 3376 3377 int 3378 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3379 register_t *retval) 3380 { 3381 /* { 3382 syscallarg(int) fd; 3383 syscallarg(const char *) path; 3384 syscallarg(int) mode; 3385 syscallarg(int) flag; 3386 } */ 3387 3388 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3389 SCARG(uap, mode), SCARG(uap, flag)); 3390 } 3391 3392 /* 3393 * Change mode of a file given path name; this version does not follow links. 3394 */ 3395 /* ARGSUSED */ 3396 int 3397 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3398 { 3399 /* { 3400 syscallarg(const char *) path; 3401 syscallarg(int) mode; 3402 } */ 3403 int error; 3404 struct vnode *vp; 3405 3406 error = namei_simple_user(SCARG(uap, path), 3407 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3408 if (error != 0) 3409 return (error); 3410 3411 error = change_mode(vp, SCARG(uap, mode), l); 3412 3413 vrele(vp); 3414 return (error); 3415 } 3416 3417 /* 3418 * Common routine to set mode given a vnode. 3419 */ 3420 static int 3421 change_mode(struct vnode *vp, int mode, struct lwp *l) 3422 { 3423 struct vattr vattr; 3424 int error; 3425 3426 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3427 vattr_null(&vattr); 3428 vattr.va_mode = mode & ALLPERMS; 3429 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3430 VOP_UNLOCK(vp); 3431 return (error); 3432 } 3433 3434 /* 3435 * Set ownership given a path name; this version follows links. 3436 */ 3437 /* ARGSUSED */ 3438 int 3439 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3440 { 3441 /* { 3442 syscallarg(const char *) path; 3443 syscallarg(uid_t) uid; 3444 syscallarg(gid_t) gid; 3445 } */ 3446 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3447 SCARG(uap, gid), 0); 3448 } 3449 3450 int 3451 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3452 gid_t gid, int flags) 3453 { 3454 int error; 3455 struct vnode *vp; 3456 namei_simple_flags_t ns_flag; 3457 3458 if (flags & AT_SYMLINK_NOFOLLOW) 3459 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3460 else 3461 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3462 3463 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3464 if (error != 0) 3465 return error; 3466 3467 error = change_owner(vp, uid, gid, l, 0); 3468 3469 vrele(vp); 3470 3471 return (error); 3472 } 3473 3474 /* 3475 * Set ownership given a path name; this version follows links. 3476 * Provides POSIX semantics. 3477 */ 3478 /* ARGSUSED */ 3479 int 3480 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3481 { 3482 /* { 3483 syscallarg(const char *) path; 3484 syscallarg(uid_t) uid; 3485 syscallarg(gid_t) gid; 3486 } */ 3487 int error; 3488 struct vnode *vp; 3489 3490 error = namei_simple_user(SCARG(uap, path), 3491 NSM_FOLLOW_TRYEMULROOT, &vp); 3492 if (error != 0) 3493 return (error); 3494 3495 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3496 3497 vrele(vp); 3498 return (error); 3499 } 3500 3501 /* 3502 * Set ownership given a file descriptor. 3503 */ 3504 /* ARGSUSED */ 3505 int 3506 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3507 { 3508 /* { 3509 syscallarg(int) fd; 3510 syscallarg(uid_t) uid; 3511 syscallarg(gid_t) gid; 3512 } */ 3513 int error; 3514 file_t *fp; 3515 3516 /* fd_getvnode() will use the descriptor for us */ 3517 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3518 return (error); 3519 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3520 l, 0); 3521 fd_putfile(SCARG(uap, fd)); 3522 return (error); 3523 } 3524 3525 int 3526 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3527 register_t *retval) 3528 { 3529 /* { 3530 syscallarg(int) fd; 3531 syscallarg(const char *) path; 3532 syscallarg(uid_t) owner; 3533 syscallarg(gid_t) group; 3534 syscallarg(int) flag; 3535 } */ 3536 3537 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3538 SCARG(uap, owner), SCARG(uap, group), 3539 SCARG(uap, flag)); 3540 } 3541 3542 /* 3543 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3544 */ 3545 /* ARGSUSED */ 3546 int 3547 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3548 { 3549 /* { 3550 syscallarg(int) fd; 3551 syscallarg(uid_t) uid; 3552 syscallarg(gid_t) gid; 3553 } */ 3554 int error; 3555 file_t *fp; 3556 3557 /* fd_getvnode() will use the descriptor for us */ 3558 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3559 return (error); 3560 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3561 l, 1); 3562 fd_putfile(SCARG(uap, fd)); 3563 return (error); 3564 } 3565 3566 /* 3567 * Set ownership given a path name; this version does not follow links. 3568 */ 3569 /* ARGSUSED */ 3570 int 3571 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3572 { 3573 /* { 3574 syscallarg(const char *) path; 3575 syscallarg(uid_t) uid; 3576 syscallarg(gid_t) gid; 3577 } */ 3578 int error; 3579 struct vnode *vp; 3580 3581 error = namei_simple_user(SCARG(uap, path), 3582 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3583 if (error != 0) 3584 return (error); 3585 3586 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3587 3588 vrele(vp); 3589 return (error); 3590 } 3591 3592 /* 3593 * Set ownership given a path name; this version does not follow links. 3594 * Provides POSIX/XPG semantics. 3595 */ 3596 /* ARGSUSED */ 3597 int 3598 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3599 { 3600 /* { 3601 syscallarg(const char *) path; 3602 syscallarg(uid_t) uid; 3603 syscallarg(gid_t) gid; 3604 } */ 3605 int error; 3606 struct vnode *vp; 3607 3608 error = namei_simple_user(SCARG(uap, path), 3609 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3610 if (error != 0) 3611 return (error); 3612 3613 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3614 3615 vrele(vp); 3616 return (error); 3617 } 3618 3619 /* 3620 * Common routine to set ownership given a vnode. 3621 */ 3622 static int 3623 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3624 int posix_semantics) 3625 { 3626 struct vattr vattr; 3627 mode_t newmode; 3628 int error; 3629 3630 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3631 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3632 goto out; 3633 3634 #define CHANGED(x) ((int)(x) != -1) 3635 newmode = vattr.va_mode; 3636 if (posix_semantics) { 3637 /* 3638 * POSIX/XPG semantics: if the caller is not the super-user, 3639 * clear set-user-id and set-group-id bits. Both POSIX and 3640 * the XPG consider the behaviour for calls by the super-user 3641 * implementation-defined; we leave the set-user-id and set- 3642 * group-id settings intact in that case. 3643 */ 3644 if (vattr.va_mode & S_ISUID) { 3645 if (kauth_authorize_vnode(l->l_cred, 3646 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3647 newmode &= ~S_ISUID; 3648 } 3649 if (vattr.va_mode & S_ISGID) { 3650 if (kauth_authorize_vnode(l->l_cred, 3651 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3652 newmode &= ~S_ISGID; 3653 } 3654 } else { 3655 /* 3656 * NetBSD semantics: when changing owner and/or group, 3657 * clear the respective bit(s). 3658 */ 3659 if (CHANGED(uid)) 3660 newmode &= ~S_ISUID; 3661 if (CHANGED(gid)) 3662 newmode &= ~S_ISGID; 3663 } 3664 /* Update va_mode iff altered. */ 3665 if (vattr.va_mode == newmode) 3666 newmode = VNOVAL; 3667 3668 vattr_null(&vattr); 3669 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3670 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3671 vattr.va_mode = newmode; 3672 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3673 #undef CHANGED 3674 3675 out: 3676 VOP_UNLOCK(vp); 3677 return (error); 3678 } 3679 3680 /* 3681 * Set the access and modification times given a path name; this 3682 * version follows links. 3683 */ 3684 /* ARGSUSED */ 3685 int 3686 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3687 register_t *retval) 3688 { 3689 /* { 3690 syscallarg(const char *) path; 3691 syscallarg(const struct timeval *) tptr; 3692 } */ 3693 3694 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3695 SCARG(uap, tptr), UIO_USERSPACE); 3696 } 3697 3698 /* 3699 * Set the access and modification times given a file descriptor. 3700 */ 3701 /* ARGSUSED */ 3702 int 3703 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3704 register_t *retval) 3705 { 3706 /* { 3707 syscallarg(int) fd; 3708 syscallarg(const struct timeval *) tptr; 3709 } */ 3710 int error; 3711 file_t *fp; 3712 3713 /* fd_getvnode() will use the descriptor for us */ 3714 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3715 return (error); 3716 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3717 UIO_USERSPACE); 3718 fd_putfile(SCARG(uap, fd)); 3719 return (error); 3720 } 3721 3722 int 3723 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3724 register_t *retval) 3725 { 3726 /* { 3727 syscallarg(int) fd; 3728 syscallarg(const struct timespec *) tptr; 3729 } */ 3730 int error; 3731 file_t *fp; 3732 3733 /* fd_getvnode() will use the descriptor for us */ 3734 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3735 return (error); 3736 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0, 3737 SCARG(uap, tptr), UIO_USERSPACE); 3738 fd_putfile(SCARG(uap, fd)); 3739 return (error); 3740 } 3741 3742 /* 3743 * Set the access and modification times given a path name; this 3744 * version does not follow links. 3745 */ 3746 int 3747 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3748 register_t *retval) 3749 { 3750 /* { 3751 syscallarg(const char *) path; 3752 syscallarg(const struct timeval *) tptr; 3753 } */ 3754 3755 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3756 SCARG(uap, tptr), UIO_USERSPACE); 3757 } 3758 3759 int 3760 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3761 register_t *retval) 3762 { 3763 /* { 3764 syscallarg(int) fd; 3765 syscallarg(const char *) path; 3766 syscallarg(const struct timespec *) tptr; 3767 syscallarg(int) flag; 3768 } */ 3769 int follow; 3770 const struct timespec *tptr; 3771 int error; 3772 3773 tptr = SCARG(uap, tptr); 3774 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3775 3776 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3777 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3778 3779 return error; 3780 } 3781 3782 /* 3783 * Common routine to set access and modification times given a vnode. 3784 */ 3785 int 3786 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3787 const struct timespec *tptr, enum uio_seg seg) 3788 { 3789 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3790 } 3791 3792 int 3793 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3794 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3795 { 3796 struct vattr vattr; 3797 int error, dorele = 0; 3798 namei_simple_flags_t sflags; 3799 bool vanull, setbirthtime; 3800 struct timespec ts[2]; 3801 3802 KASSERT(l != NULL || fdat == AT_FDCWD); 3803 3804 /* 3805 * I have checked all callers and they pass either FOLLOW, 3806 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3807 * is 0. More to the point, they don't pass anything else. 3808 * Let's keep it that way at least until the namei interfaces 3809 * are fully sanitized. 3810 */ 3811 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3812 sflags = (flag == FOLLOW) ? 3813 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3814 3815 if (tptr == NULL) { 3816 vanull = true; 3817 nanotime(&ts[0]); 3818 ts[1] = ts[0]; 3819 } else { 3820 vanull = false; 3821 if (seg != UIO_SYSSPACE) { 3822 error = copyin(tptr, ts, sizeof (ts)); 3823 if (error != 0) 3824 return error; 3825 } else { 3826 ts[0] = tptr[0]; 3827 ts[1] = tptr[1]; 3828 } 3829 } 3830 3831 if (ts[0].tv_nsec == UTIME_NOW) { 3832 nanotime(&ts[0]); 3833 if (ts[1].tv_nsec == UTIME_NOW) { 3834 vanull = true; 3835 ts[1] = ts[0]; 3836 } 3837 } else if (ts[1].tv_nsec == UTIME_NOW) 3838 nanotime(&ts[1]); 3839 3840 if (vp == NULL) { 3841 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3842 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3843 if (error != 0) 3844 return error; 3845 dorele = 1; 3846 } 3847 3848 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3849 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3850 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3851 vattr_null(&vattr); 3852 3853 if (ts[0].tv_nsec != UTIME_OMIT) 3854 vattr.va_atime = ts[0]; 3855 3856 if (ts[1].tv_nsec != UTIME_OMIT) { 3857 vattr.va_mtime = ts[1]; 3858 if (setbirthtime) 3859 vattr.va_birthtime = ts[1]; 3860 } 3861 3862 if (vanull) 3863 vattr.va_vaflags |= VA_UTIMES_NULL; 3864 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3865 VOP_UNLOCK(vp); 3866 3867 if (dorele != 0) 3868 vrele(vp); 3869 3870 return error; 3871 } 3872 3873 int 3874 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3875 const struct timeval *tptr, enum uio_seg seg) 3876 { 3877 struct timespec ts[2]; 3878 struct timespec *tsptr = NULL; 3879 int error; 3880 3881 if (tptr != NULL) { 3882 struct timeval tv[2]; 3883 3884 if (seg != UIO_SYSSPACE) { 3885 error = copyin(tptr, tv, sizeof (tv)); 3886 if (error != 0) 3887 return error; 3888 tptr = tv; 3889 } 3890 3891 if ((tv[0].tv_usec == UTIME_NOW) || 3892 (tv[0].tv_usec == UTIME_OMIT)) 3893 ts[0].tv_nsec = tv[0].tv_usec; 3894 else 3895 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3896 3897 if ((tv[1].tv_usec == UTIME_NOW) || 3898 (tv[1].tv_usec == UTIME_OMIT)) 3899 ts[1].tv_nsec = tv[1].tv_usec; 3900 else 3901 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3902 3903 tsptr = &ts[0]; 3904 } 3905 3906 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3907 } 3908 3909 /* 3910 * Truncate a file given its path name. 3911 */ 3912 /* ARGSUSED */ 3913 int 3914 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3915 { 3916 /* { 3917 syscallarg(const char *) path; 3918 syscallarg(int) pad; 3919 syscallarg(off_t) length; 3920 } */ 3921 struct vnode *vp; 3922 struct vattr vattr; 3923 int error; 3924 3925 error = namei_simple_user(SCARG(uap, path), 3926 NSM_FOLLOW_TRYEMULROOT, &vp); 3927 if (error != 0) 3928 return (error); 3929 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3930 if (vp->v_type == VDIR) 3931 error = EISDIR; 3932 else if ((error = vn_writechk(vp)) == 0 && 3933 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3934 vattr_null(&vattr); 3935 vattr.va_size = SCARG(uap, length); 3936 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3937 } 3938 vput(vp); 3939 return (error); 3940 } 3941 3942 /* 3943 * Truncate a file given a file descriptor. 3944 */ 3945 /* ARGSUSED */ 3946 int 3947 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3948 { 3949 /* { 3950 syscallarg(int) fd; 3951 syscallarg(int) pad; 3952 syscallarg(off_t) length; 3953 } */ 3954 struct vattr vattr; 3955 struct vnode *vp; 3956 file_t *fp; 3957 int error; 3958 3959 /* fd_getvnode() will use the descriptor for us */ 3960 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3961 return (error); 3962 if ((fp->f_flag & FWRITE) == 0) { 3963 error = EINVAL; 3964 goto out; 3965 } 3966 vp = fp->f_data; 3967 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3968 if (vp->v_type == VDIR) 3969 error = EISDIR; 3970 else if ((error = vn_writechk(vp)) == 0) { 3971 vattr_null(&vattr); 3972 vattr.va_size = SCARG(uap, length); 3973 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3974 } 3975 VOP_UNLOCK(vp); 3976 out: 3977 fd_putfile(SCARG(uap, fd)); 3978 return (error); 3979 } 3980 3981 /* 3982 * Sync an open file. 3983 */ 3984 /* ARGSUSED */ 3985 int 3986 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3987 { 3988 /* { 3989 syscallarg(int) fd; 3990 } */ 3991 struct vnode *vp; 3992 file_t *fp; 3993 int error; 3994 3995 /* fd_getvnode() will use the descriptor for us */ 3996 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3997 return (error); 3998 vp = fp->f_data; 3999 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4000 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4001 VOP_UNLOCK(vp); 4002 fd_putfile(SCARG(uap, fd)); 4003 return (error); 4004 } 4005 4006 /* 4007 * Sync a range of file data. API modeled after that found in AIX. 4008 * 4009 * FDATASYNC indicates that we need only save enough metadata to be able 4010 * to re-read the written data. Note we duplicate AIX's requirement that 4011 * the file be open for writing. 4012 */ 4013 /* ARGSUSED */ 4014 int 4015 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4016 { 4017 /* { 4018 syscallarg(int) fd; 4019 syscallarg(int) flags; 4020 syscallarg(off_t) start; 4021 syscallarg(off_t) length; 4022 } */ 4023 struct vnode *vp; 4024 file_t *fp; 4025 int flags, nflags; 4026 off_t s, e, len; 4027 int error; 4028 4029 /* fd_getvnode() will use the descriptor for us */ 4030 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4031 return (error); 4032 4033 if ((fp->f_flag & FWRITE) == 0) { 4034 error = EBADF; 4035 goto out; 4036 } 4037 4038 flags = SCARG(uap, flags); 4039 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4040 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4041 error = EINVAL; 4042 goto out; 4043 } 4044 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4045 if (flags & FDATASYNC) 4046 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4047 else 4048 nflags = FSYNC_WAIT; 4049 if (flags & FDISKSYNC) 4050 nflags |= FSYNC_CACHE; 4051 4052 len = SCARG(uap, length); 4053 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4054 if (len) { 4055 s = SCARG(uap, start); 4056 e = s + len; 4057 if (e < s) { 4058 error = EINVAL; 4059 goto out; 4060 } 4061 } else { 4062 e = 0; 4063 s = 0; 4064 } 4065 4066 vp = fp->f_data; 4067 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4068 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4069 VOP_UNLOCK(vp); 4070 out: 4071 fd_putfile(SCARG(uap, fd)); 4072 return (error); 4073 } 4074 4075 /* 4076 * Sync the data of an open file. 4077 */ 4078 /* ARGSUSED */ 4079 int 4080 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4081 { 4082 /* { 4083 syscallarg(int) fd; 4084 } */ 4085 struct vnode *vp; 4086 file_t *fp; 4087 int error; 4088 4089 /* fd_getvnode() will use the descriptor for us */ 4090 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4091 return (error); 4092 if ((fp->f_flag & FWRITE) == 0) { 4093 fd_putfile(SCARG(uap, fd)); 4094 return (EBADF); 4095 } 4096 vp = fp->f_data; 4097 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4098 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4099 VOP_UNLOCK(vp); 4100 fd_putfile(SCARG(uap, fd)); 4101 return (error); 4102 } 4103 4104 /* 4105 * Rename files, (standard) BSD semantics frontend. 4106 */ 4107 /* ARGSUSED */ 4108 int 4109 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4110 { 4111 /* { 4112 syscallarg(const char *) from; 4113 syscallarg(const char *) to; 4114 } */ 4115 4116 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4117 SCARG(uap, to), UIO_USERSPACE, 0)); 4118 } 4119 4120 int 4121 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4122 register_t *retval) 4123 { 4124 /* { 4125 syscallarg(int) fromfd; 4126 syscallarg(const char *) from; 4127 syscallarg(int) tofd; 4128 syscallarg(const char *) to; 4129 } */ 4130 4131 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4132 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4133 } 4134 4135 /* 4136 * Rename files, POSIX semantics frontend. 4137 */ 4138 /* ARGSUSED */ 4139 int 4140 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4141 { 4142 /* { 4143 syscallarg(const char *) from; 4144 syscallarg(const char *) to; 4145 } */ 4146 4147 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4148 SCARG(uap, to), UIO_USERSPACE, 1)); 4149 } 4150 4151 /* 4152 * Rename files. Source and destination must either both be directories, 4153 * or both not be directories. If target is a directory, it must be empty. 4154 * If `from' and `to' refer to the same object, the value of the `retain' 4155 * argument is used to determine whether `from' will be 4156 * 4157 * (retain == 0) deleted unless `from' and `to' refer to the same 4158 * object in the file system's name space (BSD). 4159 * (retain == 1) always retained (POSIX). 4160 * 4161 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4162 */ 4163 int 4164 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4165 { 4166 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4167 } 4168 4169 static int 4170 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4171 const char *to, enum uio_seg seg, int retain) 4172 { 4173 struct pathbuf *fpb, *tpb; 4174 struct nameidata fnd, tnd; 4175 struct vnode *fdvp, *fvp; 4176 struct vnode *tdvp, *tvp; 4177 struct mount *mp, *tmp; 4178 int error; 4179 4180 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4181 4182 error = pathbuf_maybe_copyin(from, seg, &fpb); 4183 if (error) 4184 goto out0; 4185 KASSERT(fpb != NULL); 4186 4187 error = pathbuf_maybe_copyin(to, seg, &tpb); 4188 if (error) 4189 goto out1; 4190 KASSERT(tpb != NULL); 4191 4192 /* 4193 * Lookup from. 4194 * 4195 * XXX LOCKPARENT is wrong because we don't actually want it 4196 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4197 * insane, so for the time being we need to leave it like this. 4198 */ 4199 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT | INRENAME), fpb); 4200 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4201 goto out2; 4202 4203 /* 4204 * Pull out the important results of the lookup, fdvp and fvp. 4205 * Of course, fvp is bogus because we're about to unlock fdvp. 4206 */ 4207 fdvp = fnd.ni_dvp; 4208 fvp = fnd.ni_vp; 4209 KASSERT(fdvp != NULL); 4210 KASSERT(fvp != NULL); 4211 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4212 4213 /* 4214 * Make sure neither fdvp nor fvp is locked. 4215 */ 4216 if (fdvp != fvp) 4217 VOP_UNLOCK(fdvp); 4218 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4219 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4220 4221 /* 4222 * Reject renaming `.' and `..'. Can't do this until after 4223 * namei because we need namei's parsing to find the final 4224 * component name. (namei should just leave us with the final 4225 * component name and not look it up itself, but anyway...) 4226 * 4227 * This was here before because we used to relookup from 4228 * instead of to and relookup requires the caller to check 4229 * this, but now file systems may depend on this check, so we 4230 * must retain it until the file systems are all rototilled. 4231 */ 4232 if (((fnd.ni_cnd.cn_namelen == 1) && 4233 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4234 ((fnd.ni_cnd.cn_namelen == 2) && 4235 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4236 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4237 error = EINVAL; /* XXX EISDIR? */ 4238 goto abort0; 4239 } 4240 4241 /* 4242 * Lookup to. 4243 * 4244 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4245 * fvp here to decide whether to add CREATEDIR is a load of 4246 * bollocks because fvp might be the wrong node by now, since 4247 * fdvp is unlocked. 4248 * 4249 * XXX Why not pass CREATEDIR always? 4250 */ 4251 NDINIT(&tnd, RENAME, 4252 (LOCKPARENT | NOCACHE | TRYEMULROOT | INRENAME | 4253 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4254 tpb); 4255 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4256 goto abort0; 4257 4258 /* 4259 * Pull out the important results of the lookup, tdvp and tvp. 4260 * Of course, tvp is bogus because we're about to unlock tdvp. 4261 */ 4262 tdvp = tnd.ni_dvp; 4263 tvp = tnd.ni_vp; 4264 KASSERT(tdvp != NULL); 4265 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4266 4267 /* 4268 * Make sure neither tdvp nor tvp is locked. 4269 */ 4270 if (tdvp != tvp) 4271 VOP_UNLOCK(tdvp); 4272 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4273 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4274 4275 /* 4276 * Reject renaming onto `.' or `..'. relookup is unhappy with 4277 * these, which is why we must do this here. Once upon a time 4278 * we relooked up from instead of to, and consequently didn't 4279 * need this check, but now that we relookup to instead of 4280 * from, we need this; and we shall need it forever forward 4281 * until the VOP_RENAME protocol changes, because file systems 4282 * will no doubt begin to depend on this check. 4283 */ 4284 if (((tnd.ni_cnd.cn_namelen == 1) && 4285 (tnd.ni_cnd.cn_nameptr[0] == '.')) || 4286 ((tnd.ni_cnd.cn_namelen == 2) && 4287 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4288 (tnd.ni_cnd.cn_nameptr[1] == '.'))) { 4289 error = EINVAL; /* XXX EISDIR? */ 4290 goto abort1; 4291 } 4292 4293 /* 4294 * Get the mount point. If the file system has been unmounted, 4295 * which it may be because we're not holding any vnode locks, 4296 * then v_mount will be NULL. We're not really supposed to 4297 * read v_mount without holding the vnode lock, but since we 4298 * have fdvp referenced, if fdvp->v_mount changes then at worst 4299 * it will be set to NULL, not changed to another mount point. 4300 * And, of course, since it is up to the file system to 4301 * determine the real lock order, we can't lock both fdvp and 4302 * tdvp at the same time. 4303 */ 4304 mp = fdvp->v_mount; 4305 if (mp == NULL) { 4306 error = ENOENT; 4307 goto abort1; 4308 } 4309 4310 /* 4311 * Make sure the mount points match. Again, although we don't 4312 * hold any vnode locks, the v_mount fields may change -- but 4313 * at worst they will change to NULL, so this will never become 4314 * a cross-device rename, because we hold vnode references. 4315 * 4316 * XXX Because nothing is locked and the compiler may reorder 4317 * things here, unmounting the file system at an inopportune 4318 * moment may cause rename to fail with ENXDEV when it really 4319 * should fail with ENOENT. 4320 */ 4321 tmp = tdvp->v_mount; 4322 if (tmp == NULL) { 4323 error = ENOENT; 4324 goto abort1; 4325 } 4326 4327 if (mp != tmp) { 4328 error = EXDEV; 4329 goto abort1; 4330 } 4331 4332 /* 4333 * Take the vfs rename lock to avoid cross-directory screw cases. 4334 * Nothing is locked currently, so taking this lock is safe. 4335 */ 4336 error = VFS_RENAMELOCK_ENTER(mp); 4337 if (error) 4338 goto abort1; 4339 4340 /* 4341 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4342 * and nothing is locked except for the vfs rename lock. 4343 * 4344 * The next step is a little rain dance to conform to the 4345 * insane lock protocol, even though it does nothing to ward 4346 * off race conditions. 4347 * 4348 * We need tdvp and tvp to be locked. However, because we have 4349 * unlocked tdvp in order to hold no locks while we take the 4350 * vfs rename lock, tvp may be wrong here, and we can't safely 4351 * lock it even if the sensible file systems will just unlock 4352 * it straight away. Consequently, we must lock tdvp and then 4353 * relookup tvp to get it locked. 4354 * 4355 * Finally, because the VOP_RENAME protocol is brain-damaged 4356 * and various file systems insanely depend on the semantics of 4357 * this brain damage, the lookup of to must be the last lookup 4358 * before VOP_RENAME. 4359 */ 4360 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4361 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4362 if (error) 4363 goto abort2; 4364 4365 /* 4366 * Drop the old tvp and pick up the new one -- which might be 4367 * the same, but that doesn't matter to us. After this, tdvp 4368 * and tvp should both be locked. 4369 */ 4370 if (tvp != NULL) 4371 vrele(tvp); 4372 tvp = tnd.ni_vp; 4373 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4374 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4375 4376 /* 4377 * The old do_sys_rename had various consistency checks here 4378 * involving fvp and tvp. fvp is bogus already here, and tvp 4379 * will become bogus soon in any sensible file system, so the 4380 * only purpose in putting these checks here is to give lip 4381 * service to these screw cases and to acknowledge that they 4382 * exist, not actually to handle them, but here you go 4383 * anyway... 4384 */ 4385 4386 /* 4387 * Acknowledge that directories and non-directories aren't 4388 * suposed to mix. 4389 */ 4390 if (tvp != NULL) { 4391 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4392 error = ENOTDIR; 4393 goto abort3; 4394 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4395 error = EISDIR; 4396 goto abort3; 4397 } 4398 } 4399 4400 /* 4401 * Acknowledge some random screw case, among the dozens that 4402 * might arise. 4403 */ 4404 if (fvp == tdvp) { 4405 error = EINVAL; 4406 goto abort3; 4407 } 4408 4409 /* 4410 * Acknowledge that POSIX has a wacky screw case. 4411 * 4412 * XXX Eventually the retain flag needs to be passed on to 4413 * VOP_RENAME. 4414 */ 4415 if (fvp == tvp) { 4416 if (retain) { 4417 error = 0; 4418 goto abort3; 4419 } else if ((fdvp == tdvp) && 4420 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4421 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4422 fnd.ni_cnd.cn_namelen))) { 4423 error = 0; 4424 goto abort3; 4425 } 4426 } 4427 4428 /* 4429 * Make sure veriexec can screw us up. (But a race can screw 4430 * up veriexec, of course -- remember, fvp and (soon) tvp are 4431 * bogus.) 4432 */ 4433 #if NVERIEXEC > 0 4434 { 4435 char *f1, *f2; 4436 size_t f1_len; 4437 size_t f2_len; 4438 4439 f1_len = fnd.ni_cnd.cn_namelen + 1; 4440 f1 = kmem_alloc(f1_len, KM_SLEEP); 4441 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4442 4443 f2_len = tnd.ni_cnd.cn_namelen + 1; 4444 f2 = kmem_alloc(f2_len, KM_SLEEP); 4445 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4446 4447 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4448 4449 kmem_free(f1, f1_len); 4450 kmem_free(f2, f2_len); 4451 4452 if (error) 4453 goto abort3; 4454 } 4455 #endif /* NVERIEXEC > 0 */ 4456 4457 /* 4458 * All ready. Incant the rename vop. 4459 */ 4460 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4461 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4462 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4463 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4464 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4465 4466 /* 4467 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4468 * tdvp and tvp. But we can't assert any of that. 4469 */ 4470 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4471 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4472 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4473 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4474 4475 /* 4476 * So all we have left to do is to drop the rename lock and 4477 * destroy the pathbufs. 4478 */ 4479 VFS_RENAMELOCK_EXIT(mp); 4480 goto out2; 4481 4482 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4483 VOP_UNLOCK(tvp); 4484 abort2: VOP_UNLOCK(tdvp); 4485 VFS_RENAMELOCK_EXIT(mp); 4486 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4487 vrele(tdvp); 4488 if (tvp != NULL) 4489 vrele(tvp); 4490 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4491 vrele(fdvp); 4492 vrele(fvp); 4493 out2: pathbuf_destroy(tpb); 4494 out1: pathbuf_destroy(fpb); 4495 out0: return error; 4496 } 4497 4498 /* 4499 * Make a directory file. 4500 */ 4501 /* ARGSUSED */ 4502 int 4503 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4504 { 4505 /* { 4506 syscallarg(const char *) path; 4507 syscallarg(int) mode; 4508 } */ 4509 4510 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4511 SCARG(uap, mode), UIO_USERSPACE); 4512 } 4513 4514 int 4515 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4516 register_t *retval) 4517 { 4518 /* { 4519 syscallarg(int) fd; 4520 syscallarg(const char *) path; 4521 syscallarg(int) mode; 4522 } */ 4523 4524 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4525 SCARG(uap, mode), UIO_USERSPACE); 4526 } 4527 4528 4529 int 4530 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4531 { 4532 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4533 } 4534 4535 static int 4536 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4537 enum uio_seg seg) 4538 { 4539 struct proc *p = curlwp->l_proc; 4540 struct vnode *vp; 4541 struct vattr vattr; 4542 int error; 4543 struct pathbuf *pb; 4544 struct nameidata nd; 4545 4546 KASSERT(l != NULL || fdat == AT_FDCWD); 4547 4548 /* XXX bollocks, should pass in a pathbuf */ 4549 error = pathbuf_maybe_copyin(path, seg, &pb); 4550 if (error) { 4551 return error; 4552 } 4553 4554 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4555 4556 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4557 pathbuf_destroy(pb); 4558 return (error); 4559 } 4560 vp = nd.ni_vp; 4561 if (vp != NULL) { 4562 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4563 if (nd.ni_dvp == vp) 4564 vrele(nd.ni_dvp); 4565 else 4566 vput(nd.ni_dvp); 4567 vrele(vp); 4568 pathbuf_destroy(pb); 4569 return (EEXIST); 4570 } 4571 vattr_null(&vattr); 4572 vattr.va_type = VDIR; 4573 /* We will read cwdi->cwdi_cmask unlocked. */ 4574 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4575 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4576 if (!error) 4577 vrele(nd.ni_vp); 4578 vput(nd.ni_dvp); 4579 pathbuf_destroy(pb); 4580 return (error); 4581 } 4582 4583 /* 4584 * Remove a directory file. 4585 */ 4586 /* ARGSUSED */ 4587 int 4588 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4589 { 4590 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4591 AT_REMOVEDIR, UIO_USERSPACE); 4592 } 4593 4594 /* 4595 * Read a block of directory entries in a file system independent format. 4596 */ 4597 int 4598 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4599 { 4600 /* { 4601 syscallarg(int) fd; 4602 syscallarg(char *) buf; 4603 syscallarg(size_t) count; 4604 } */ 4605 file_t *fp; 4606 int error, done; 4607 4608 /* fd_getvnode() will use the descriptor for us */ 4609 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4610 return (error); 4611 if ((fp->f_flag & FREAD) == 0) { 4612 error = EBADF; 4613 goto out; 4614 } 4615 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4616 SCARG(uap, count), &done, l, 0, 0); 4617 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4618 *retval = done; 4619 out: 4620 fd_putfile(SCARG(uap, fd)); 4621 return (error); 4622 } 4623 4624 /* 4625 * Set the mode mask for creation of filesystem nodes. 4626 */ 4627 int 4628 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4629 { 4630 /* { 4631 syscallarg(mode_t) newmask; 4632 } */ 4633 struct proc *p = l->l_proc; 4634 struct cwdinfo *cwdi; 4635 4636 /* 4637 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4638 * important is that we serialize changes to the mask. The 4639 * rw_exit() will issue a write memory barrier on our behalf, 4640 * and force the changes out to other CPUs (as it must use an 4641 * atomic operation, draining the local CPU's store buffers). 4642 */ 4643 cwdi = p->p_cwdi; 4644 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4645 *retval = cwdi->cwdi_cmask; 4646 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4647 rw_exit(&cwdi->cwdi_lock); 4648 4649 return (0); 4650 } 4651 4652 int 4653 dorevoke(struct vnode *vp, kauth_cred_t cred) 4654 { 4655 struct vattr vattr; 4656 int error, fs_decision; 4657 4658 vn_lock(vp, LK_SHARED | LK_RETRY); 4659 error = VOP_GETATTR(vp, &vattr, cred); 4660 VOP_UNLOCK(vp); 4661 if (error != 0) 4662 return error; 4663 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4664 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4665 fs_decision); 4666 if (!error) 4667 VOP_REVOKE(vp, REVOKEALL); 4668 return (error); 4669 } 4670 4671 /* 4672 * Void all references to file by ripping underlying filesystem 4673 * away from vnode. 4674 */ 4675 /* ARGSUSED */ 4676 int 4677 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4678 { 4679 /* { 4680 syscallarg(const char *) path; 4681 } */ 4682 struct vnode *vp; 4683 int error; 4684 4685 error = namei_simple_user(SCARG(uap, path), 4686 NSM_FOLLOW_TRYEMULROOT, &vp); 4687 if (error != 0) 4688 return (error); 4689 error = dorevoke(vp, l->l_cred); 4690 vrele(vp); 4691 return (error); 4692 } 4693
Indexes created Fri Sep 26 20:09:58 GMT 2025