vfs_syscalls.c revision 1.519
1 /* $NetBSD: vfs_syscalls.c,v 1.519 2019/01/27 02:08:43 pgoyette 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 1.519 2019/01/27 02:08:43 pgoyette 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/fstrans.h> 91 #include <sys/proc.h> 92 #include <sys/uio.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/quota.h> 99 #include <sys/quotactl.h> 100 #include <sys/ktrace.h> 101 #ifdef FILEASSOC 102 #include <sys/fileassoc.h> 103 #endif /* FILEASSOC */ 104 #include <sys/extattr.h> 105 #include <sys/verified_exec.h> 106 #include <sys/kauth.h> 107 #include <sys/atomic.h> 108 #include <sys/module.h> 109 #include <sys/buf.h> 110 #include <sys/event.h> 111 #include <sys/compat_stub.h> 112 113 #include <miscfs/genfs/genfs.h> 114 #include <miscfs/specfs/specdev.h> 115 116 #include <nfs/rpcv2.h> 117 #include <nfs/nfsproto.h> 118 #include <nfs/nfs.h> 119 #include <nfs/nfs_var.h> 120 121 /* XXX this shouldn't be here */ 122 #ifndef OFF_T_MAX 123 #define OFF_T_MAX __type_max(off_t) 124 #endif 125 126 static int change_flags(struct vnode *, u_long, struct lwp *); 127 static int change_mode(struct vnode *, int, struct lwp *); 128 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 129 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 130 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 131 enum uio_seg); 132 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 133 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 134 enum uio_seg); 135 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 136 enum uio_seg, int); 137 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 138 size_t, register_t *); 139 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 140 141 static int fd_nameiat(struct lwp *, int, struct nameidata *); 142 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 143 namei_simple_flags_t, struct vnode **); 144 145 /* 146 * This table is used to maintain compatibility with 4.3BSD 147 * and NetBSD 0.9 mount syscalls - and possibly other systems. 148 * Note, the order is important! 149 * 150 * Do not modify this table. It should only contain filesystems 151 * supported by NetBSD 0.9 and 4.3BSD. 152 */ 153 const char * const mountcompatnames[] = { 154 NULL, /* 0 = MOUNT_NONE */ 155 MOUNT_FFS, /* 1 = MOUNT_UFS */ 156 MOUNT_NFS, /* 2 */ 157 MOUNT_MFS, /* 3 */ 158 MOUNT_MSDOS, /* 4 */ 159 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 160 MOUNT_FDESC, /* 6 */ 161 MOUNT_KERNFS, /* 7 */ 162 NULL, /* 8 = MOUNT_DEVFS */ 163 MOUNT_AFS, /* 9 */ 164 }; 165 166 const int nmountcompatnames = __arraycount(mountcompatnames); 167 168 static int 169 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 170 { 171 file_t *dfp; 172 int error; 173 174 if (fdat != AT_FDCWD) { 175 if ((error = fd_getvnode(fdat, &dfp)) != 0) 176 goto out; 177 178 NDAT(ndp, dfp->f_vnode); 179 } 180 181 error = namei(ndp); 182 183 if (fdat != AT_FDCWD) 184 fd_putfile(fdat); 185 out: 186 return error; 187 } 188 189 static int 190 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 191 namei_simple_flags_t sflags, struct vnode **vp_ret) 192 { 193 file_t *dfp; 194 struct vnode *dvp; 195 int error; 196 197 if (fdat != AT_FDCWD) { 198 if ((error = fd_getvnode(fdat, &dfp)) != 0) 199 goto out; 200 201 dvp = dfp->f_vnode; 202 } else { 203 dvp = NULL; 204 } 205 206 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 207 208 if (fdat != AT_FDCWD) 209 fd_putfile(fdat); 210 out: 211 return error; 212 } 213 214 static int 215 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 216 { 217 int error; 218 219 fp->f_flag = flags & FMASK; 220 fp->f_type = DTYPE_VNODE; 221 fp->f_ops = &vnops; 222 fp->f_vnode = vp; 223 224 if (flags & (O_EXLOCK | O_SHLOCK)) { 225 struct flock lf; 226 int type; 227 228 lf.l_whence = SEEK_SET; 229 lf.l_start = 0; 230 lf.l_len = 0; 231 if (flags & O_EXLOCK) 232 lf.l_type = F_WRLCK; 233 else 234 lf.l_type = F_RDLCK; 235 type = F_FLOCK; 236 if ((flags & FNONBLOCK) == 0) 237 type |= F_WAIT; 238 VOP_UNLOCK(vp); 239 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 240 if (error) { 241 (void) vn_close(vp, fp->f_flag, fp->f_cred); 242 fd_abort(l->l_proc, fp, indx); 243 return error; 244 } 245 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 246 atomic_or_uint(&fp->f_flag, FHASLOCK); 247 } 248 if (flags & O_CLOEXEC) 249 fd_set_exclose(l, indx, true); 250 return 0; 251 } 252 253 static int 254 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 255 void *data, size_t *data_len) 256 { 257 struct mount *mp; 258 int error = 0, saved_flags; 259 260 mp = vp->v_mount; 261 saved_flags = mp->mnt_flag; 262 263 /* We can operate only on VV_ROOT nodes. */ 264 if ((vp->v_vflag & VV_ROOT) == 0) { 265 error = EINVAL; 266 goto out; 267 } 268 269 /* 270 * We only allow the filesystem to be reloaded if it 271 * is currently mounted read-only. Additionally, we 272 * prevent read-write to read-only downgrades. 273 */ 274 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 275 (mp->mnt_flag & MNT_RDONLY) == 0 && 276 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 277 error = EOPNOTSUPP; /* Needs translation */ 278 goto out; 279 } 280 281 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 282 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 283 if (error) 284 goto out; 285 286 error = vfs_suspend(mp, 0); 287 if (error) 288 goto out; 289 290 mutex_enter(&mp->mnt_updating); 291 292 mp->mnt_flag &= ~MNT_OP_FLAGS; 293 mp->mnt_flag |= flags & MNT_OP_FLAGS; 294 295 /* 296 * Set the mount level flags. 297 */ 298 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 299 if ((flags & MNT_RDONLY)) 300 mp->mnt_iflag |= IMNT_WANTRDONLY; 301 else 302 mp->mnt_iflag |= IMNT_WANTRDWR; 303 } 304 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 305 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 306 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 307 mp->mnt_flag &= ~MNT_RDONLY; 308 309 error = VFS_MOUNT(mp, path, data, data_len); 310 311 if (error && data != NULL) { 312 int error2; 313 314 /* 315 * Update failed; let's try and see if it was an 316 * export request. For compat with 3.0 and earlier. 317 */ 318 error2 = vfs_hooks_reexport(mp, path, data); 319 320 /* 321 * Only update error code if the export request was 322 * understood but some problem occurred while 323 * processing it. 324 */ 325 if (error2 != EJUSTRETURN) 326 error = error2; 327 } 328 329 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 330 mp->mnt_flag |= MNT_RDONLY; 331 if (error) 332 mp->mnt_flag = saved_flags; 333 mp->mnt_flag &= ~MNT_OP_FLAGS; 334 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 335 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 336 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 337 vfs_syncer_add_to_worklist(mp); 338 } else { 339 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 340 vfs_syncer_remove_from_worklist(mp); 341 } 342 mutex_exit(&mp->mnt_updating); 343 vfs_resume(mp); 344 345 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 346 (flags & MNT_EXTATTR)) { 347 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 348 NULL, 0, NULL) != 0) { 349 printf("%s: failed to start extattr, error = %d", 350 mp->mnt_stat.f_mntonname, error); 351 mp->mnt_flag &= ~MNT_EXTATTR; 352 } 353 } 354 355 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 356 !(flags & MNT_EXTATTR)) { 357 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 358 NULL, 0, NULL) != 0) { 359 printf("%s: failed to stop extattr, error = %d", 360 mp->mnt_stat.f_mntonname, error); 361 mp->mnt_flag |= MNT_RDONLY; 362 } 363 } 364 out: 365 return (error); 366 } 367 368 static int 369 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 370 struct vfsops **vfsops) 371 { 372 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 373 int error; 374 375 if (type_seg == UIO_USERSPACE) { 376 /* Copy file-system type from userspace. */ 377 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 378 } else { 379 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 380 KASSERT(error == 0); 381 } 382 383 if (error) { 384 /* 385 * Historically, filesystem types were identified by numbers. 386 * If we get an integer for the filesystem type instead of a 387 * string, we check to see if it matches one of the historic 388 * filesystem types. 389 */ 390 u_long fsindex = (u_long)fstype; 391 if (fsindex >= nmountcompatnames || 392 mountcompatnames[fsindex] == NULL) 393 return ENODEV; 394 strlcpy(fstypename, mountcompatnames[fsindex], 395 sizeof(fstypename)); 396 } 397 398 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 399 if (strcmp(fstypename, "ufs") == 0) 400 fstypename[0] = 'f'; 401 402 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 403 return 0; 404 405 /* If we can autoload a vfs module, try again */ 406 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 407 408 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 409 return 0; 410 411 return ENODEV; 412 } 413 414 static int 415 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 416 void *data, size_t *data_len) 417 { 418 struct mount *mp; 419 int error; 420 421 /* If MNT_GETARGS is specified, it should be the only flag. */ 422 if (flags & ~MNT_GETARGS) 423 return EINVAL; 424 425 mp = vp->v_mount; 426 427 /* XXX: probably some notion of "can see" here if we want isolation. */ 428 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 429 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 430 if (error) 431 return error; 432 433 if ((vp->v_vflag & VV_ROOT) == 0) 434 return EINVAL; 435 436 if (vfs_busy(mp)) 437 return EPERM; 438 439 mutex_enter(&mp->mnt_updating); 440 mp->mnt_flag &= ~MNT_OP_FLAGS; 441 mp->mnt_flag |= MNT_GETARGS; 442 error = VFS_MOUNT(mp, path, data, data_len); 443 mp->mnt_flag &= ~MNT_OP_FLAGS; 444 mutex_exit(&mp->mnt_updating); 445 446 vfs_unbusy(mp); 447 return (error); 448 } 449 450 int 451 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 452 { 453 /* { 454 syscallarg(const char *) type; 455 syscallarg(const char *) path; 456 syscallarg(int) flags; 457 syscallarg(void *) data; 458 syscallarg(size_t) data_len; 459 } */ 460 461 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 462 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 463 SCARG(uap, data_len), retval); 464 } 465 466 int 467 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 468 const char *path, int flags, void *data, enum uio_seg data_seg, 469 size_t data_len, register_t *retval) 470 { 471 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 472 struct vnode *vp; 473 void *data_buf = data; 474 bool vfsopsrele = false; 475 size_t alloc_sz = 0; 476 int error; 477 478 /* 479 * Get vnode to be covered 480 */ 481 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 482 if (error != 0) { 483 vp = NULL; 484 goto done; 485 } 486 487 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 488 vfsops = vp->v_mount->mnt_op; 489 } else { 490 /* 'type' is userspace */ 491 error = mount_get_vfsops(type, type_seg, &vfsops); 492 if (error != 0) 493 goto done; 494 vfsopsrele = true; 495 } 496 497 /* 498 * We allow data to be NULL, even for userspace. Some fs's don't need 499 * it. The others will handle NULL. 500 */ 501 if (data != NULL && data_seg == UIO_USERSPACE) { 502 if (data_len == 0) { 503 /* No length supplied, use default for filesystem */ 504 data_len = vfsops->vfs_min_mount_data; 505 506 /* 507 * Hopefully a longer buffer won't make copyin() fail. 508 * For compatibility with 3.0 and earlier. 509 */ 510 if (flags & MNT_UPDATE 511 && data_len < sizeof (struct mnt_export_args30)) 512 data_len = sizeof (struct mnt_export_args30); 513 } 514 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 515 error = EINVAL; 516 goto done; 517 } 518 alloc_sz = data_len; 519 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 520 521 /* NFS needs the buffer even for mnt_getargs .... */ 522 error = copyin(data, data_buf, data_len); 523 if (error != 0) 524 goto done; 525 } 526 527 if (flags & MNT_GETARGS) { 528 if (data_len == 0) { 529 error = EINVAL; 530 goto done; 531 } 532 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 533 if (error != 0) 534 goto done; 535 if (data_seg == UIO_USERSPACE) 536 error = copyout(data_buf, data, data_len); 537 *retval = data_len; 538 } else if (flags & MNT_UPDATE) { 539 error = mount_update(l, vp, path, flags, data_buf, &data_len); 540 } else { 541 /* Locking is handled internally in mount_domount(). */ 542 KASSERT(vfsopsrele == true); 543 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 544 &data_len); 545 vfsopsrele = false; 546 } 547 if (!error) 548 KNOTE(&fs_klist, VQ_MOUNT); 549 550 done: 551 if (vfsopsrele) 552 vfs_delref(vfsops); 553 if (vp != NULL) { 554 vrele(vp); 555 } 556 if (data_buf != data) 557 kmem_free(data_buf, alloc_sz); 558 return (error); 559 } 560 561 /* 562 * Unmount a file system. 563 * 564 * Note: unmount takes a path to the vnode mounted on as argument, 565 * not special file (as before). 566 */ 567 /* ARGSUSED */ 568 int 569 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 570 { 571 /* { 572 syscallarg(const char *) path; 573 syscallarg(int) flags; 574 } */ 575 struct vnode *vp; 576 struct mount *mp; 577 int error; 578 struct pathbuf *pb; 579 struct nameidata nd; 580 581 error = pathbuf_copyin(SCARG(uap, path), &pb); 582 if (error) { 583 return error; 584 } 585 586 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 587 if ((error = namei(&nd)) != 0) { 588 pathbuf_destroy(pb); 589 return error; 590 } 591 vp = nd.ni_vp; 592 pathbuf_destroy(pb); 593 594 mp = vp->v_mount; 595 vfs_ref(mp); 596 VOP_UNLOCK(vp); 597 598 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 599 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 600 if (error) { 601 vrele(vp); 602 vfs_rele(mp); 603 return (error); 604 } 605 606 /* 607 * Don't allow unmounting the root file system. 608 */ 609 if (mp->mnt_flag & MNT_ROOTFS) { 610 vrele(vp); 611 vfs_rele(mp); 612 return (EINVAL); 613 } 614 615 /* 616 * Must be the root of the filesystem 617 */ 618 if ((vp->v_vflag & VV_ROOT) == 0) { 619 vrele(vp); 620 vfs_rele(mp); 621 return (EINVAL); 622 } 623 624 vrele(vp); 625 error = dounmount(mp, SCARG(uap, flags), l); 626 vfs_rele(mp); 627 if (!error) 628 KNOTE(&fs_klist, VQ_UNMOUNT); 629 return error; 630 } 631 632 /* 633 * Sync each mounted filesystem. 634 */ 635 #ifdef DEBUG 636 int syncprt = 0; 637 struct ctldebug debug0 = { "syncprt", &syncprt }; 638 #endif 639 640 void 641 do_sys_sync(struct lwp *l) 642 { 643 mount_iterator_t *iter; 644 struct mount *mp; 645 int asyncflag; 646 647 mountlist_iterator_init(&iter); 648 while ((mp = mountlist_iterator_next(iter)) != NULL) { 649 mutex_enter(&mp->mnt_updating); 650 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 651 asyncflag = mp->mnt_flag & MNT_ASYNC; 652 mp->mnt_flag &= ~MNT_ASYNC; 653 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 654 if (asyncflag) 655 mp->mnt_flag |= MNT_ASYNC; 656 } 657 mutex_exit(&mp->mnt_updating); 658 } 659 mountlist_iterator_destroy(iter); 660 #ifdef DEBUG 661 if (syncprt) 662 vfs_bufstats(); 663 #endif /* DEBUG */ 664 } 665 666 /* ARGSUSED */ 667 int 668 sys_sync(struct lwp *l, const void *v, register_t *retval) 669 { 670 do_sys_sync(l); 671 return (0); 672 } 673 674 675 /* 676 * Access or change filesystem quotas. 677 * 678 * (this is really 14 different calls bundled into one) 679 */ 680 681 static int 682 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 683 { 684 struct quotastat info_k; 685 int error; 686 687 /* ensure any padding bytes are cleared */ 688 memset(&info_k, 0, sizeof(info_k)); 689 690 error = vfs_quotactl_stat(mp, &info_k); 691 if (error) { 692 return error; 693 } 694 695 return copyout(&info_k, info_u, sizeof(info_k)); 696 } 697 698 static int 699 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 700 struct quotaidtypestat *info_u) 701 { 702 struct quotaidtypestat info_k; 703 int error; 704 705 /* ensure any padding bytes are cleared */ 706 memset(&info_k, 0, sizeof(info_k)); 707 708 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 709 if (error) { 710 return error; 711 } 712 713 return copyout(&info_k, info_u, sizeof(info_k)); 714 } 715 716 static int 717 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 718 struct quotaobjtypestat *info_u) 719 { 720 struct quotaobjtypestat info_k; 721 int error; 722 723 /* ensure any padding bytes are cleared */ 724 memset(&info_k, 0, sizeof(info_k)); 725 726 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 727 if (error) { 728 return error; 729 } 730 731 return copyout(&info_k, info_u, sizeof(info_k)); 732 } 733 734 static int 735 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 736 struct quotaval *val_u) 737 { 738 struct quotakey key_k; 739 struct quotaval val_k; 740 int error; 741 742 /* ensure any padding bytes are cleared */ 743 memset(&val_k, 0, sizeof(val_k)); 744 745 error = copyin(key_u, &key_k, sizeof(key_k)); 746 if (error) { 747 return error; 748 } 749 750 error = vfs_quotactl_get(mp, &key_k, &val_k); 751 if (error) { 752 return error; 753 } 754 755 return copyout(&val_k, val_u, sizeof(val_k)); 756 } 757 758 static int 759 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 760 const struct quotaval *val_u) 761 { 762 struct quotakey key_k; 763 struct quotaval val_k; 764 int error; 765 766 error = copyin(key_u, &key_k, sizeof(key_k)); 767 if (error) { 768 return error; 769 } 770 771 error = copyin(val_u, &val_k, sizeof(val_k)); 772 if (error) { 773 return error; 774 } 775 776 return vfs_quotactl_put(mp, &key_k, &val_k); 777 } 778 779 static int 780 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 781 { 782 struct quotakey key_k; 783 int error; 784 785 error = copyin(key_u, &key_k, sizeof(key_k)); 786 if (error) { 787 return error; 788 } 789 790 return vfs_quotactl_del(mp, &key_k); 791 } 792 793 static int 794 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 795 { 796 struct quotakcursor cursor_k; 797 int error; 798 799 /* ensure any padding bytes are cleared */ 800 memset(&cursor_k, 0, sizeof(cursor_k)); 801 802 error = vfs_quotactl_cursoropen(mp, &cursor_k); 803 if (error) { 804 return error; 805 } 806 807 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 808 } 809 810 static int 811 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 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 return vfs_quotactl_cursorclose(mp, &cursor_k); 822 } 823 824 static int 825 do_sys_quotactl_cursorskipidtype(struct mount *mp, 826 struct quotakcursor *cursor_u, int idtype) 827 { 828 struct quotakcursor cursor_k; 829 int error; 830 831 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 832 if (error) { 833 return error; 834 } 835 836 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 837 if (error) { 838 return error; 839 } 840 841 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 842 } 843 844 static int 845 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 846 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 847 unsigned *ret_u) 848 { 849 #define CGET_STACK_MAX 8 850 struct quotakcursor cursor_k; 851 struct quotakey stackkeys[CGET_STACK_MAX]; 852 struct quotaval stackvals[CGET_STACK_MAX]; 853 struct quotakey *keys_k; 854 struct quotaval *vals_k; 855 unsigned ret_k; 856 int error; 857 858 if (maxnum > 128) { 859 maxnum = 128; 860 } 861 862 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 863 if (error) { 864 return error; 865 } 866 867 if (maxnum <= CGET_STACK_MAX) { 868 keys_k = stackkeys; 869 vals_k = stackvals; 870 /* ensure any padding bytes are cleared */ 871 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 872 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 873 } else { 874 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 875 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 876 } 877 878 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 879 &ret_k); 880 if (error) { 881 goto fail; 882 } 883 884 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 885 if (error) { 886 goto fail; 887 } 888 889 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 890 if (error) { 891 goto fail; 892 } 893 894 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 895 if (error) { 896 goto fail; 897 } 898 899 /* do last to maximize the chance of being able to recover a failure */ 900 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 901 902 fail: 903 if (keys_k != stackkeys) { 904 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 905 } 906 if (vals_k != stackvals) { 907 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 908 } 909 return error; 910 } 911 912 static int 913 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 914 int *ret_u) 915 { 916 struct quotakcursor cursor_k; 917 int ret_k; 918 int error; 919 920 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 921 if (error) { 922 return error; 923 } 924 925 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 926 if (error) { 927 return error; 928 } 929 930 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 931 if (error) { 932 return error; 933 } 934 935 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 936 } 937 938 static int 939 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 940 { 941 struct quotakcursor cursor_k; 942 int error; 943 944 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 945 if (error) { 946 return error; 947 } 948 949 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 950 if (error) { 951 return error; 952 } 953 954 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 955 } 956 957 static int 958 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 959 { 960 char *path_k; 961 int error; 962 963 /* XXX this should probably be a struct pathbuf */ 964 path_k = PNBUF_GET(); 965 error = copyin(path_u, path_k, PATH_MAX); 966 if (error) { 967 PNBUF_PUT(path_k); 968 return error; 969 } 970 971 error = vfs_quotactl_quotaon(mp, idtype, path_k); 972 973 PNBUF_PUT(path_k); 974 return error; 975 } 976 977 static int 978 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 979 { 980 return vfs_quotactl_quotaoff(mp, idtype); 981 } 982 983 int 984 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 985 { 986 struct mount *mp; 987 struct vnode *vp; 988 int error; 989 990 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 991 if (error != 0) 992 return (error); 993 mp = vp->v_mount; 994 995 switch (args->qc_op) { 996 case QUOTACTL_STAT: 997 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 998 break; 999 case QUOTACTL_IDTYPESTAT: 1000 error = do_sys_quotactl_idtypestat(mp, 1001 args->u.idtypestat.qc_idtype, 1002 args->u.idtypestat.qc_info); 1003 break; 1004 case QUOTACTL_OBJTYPESTAT: 1005 error = do_sys_quotactl_objtypestat(mp, 1006 args->u.objtypestat.qc_objtype, 1007 args->u.objtypestat.qc_info); 1008 break; 1009 case QUOTACTL_GET: 1010 error = do_sys_quotactl_get(mp, 1011 args->u.get.qc_key, 1012 args->u.get.qc_val); 1013 break; 1014 case QUOTACTL_PUT: 1015 error = do_sys_quotactl_put(mp, 1016 args->u.put.qc_key, 1017 args->u.put.qc_val); 1018 break; 1019 case QUOTACTL_DEL: 1020 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1021 break; 1022 case QUOTACTL_CURSOROPEN: 1023 error = do_sys_quotactl_cursoropen(mp, 1024 args->u.cursoropen.qc_cursor); 1025 break; 1026 case QUOTACTL_CURSORCLOSE: 1027 error = do_sys_quotactl_cursorclose(mp, 1028 args->u.cursorclose.qc_cursor); 1029 break; 1030 case QUOTACTL_CURSORSKIPIDTYPE: 1031 error = do_sys_quotactl_cursorskipidtype(mp, 1032 args->u.cursorskipidtype.qc_cursor, 1033 args->u.cursorskipidtype.qc_idtype); 1034 break; 1035 case QUOTACTL_CURSORGET: 1036 error = do_sys_quotactl_cursorget(mp, 1037 args->u.cursorget.qc_cursor, 1038 args->u.cursorget.qc_keys, 1039 args->u.cursorget.qc_vals, 1040 args->u.cursorget.qc_maxnum, 1041 args->u.cursorget.qc_ret); 1042 break; 1043 case QUOTACTL_CURSORATEND: 1044 error = do_sys_quotactl_cursoratend(mp, 1045 args->u.cursoratend.qc_cursor, 1046 args->u.cursoratend.qc_ret); 1047 break; 1048 case QUOTACTL_CURSORREWIND: 1049 error = do_sys_quotactl_cursorrewind(mp, 1050 args->u.cursorrewind.qc_cursor); 1051 break; 1052 case QUOTACTL_QUOTAON: 1053 error = do_sys_quotactl_quotaon(mp, 1054 args->u.quotaon.qc_idtype, 1055 args->u.quotaon.qc_quotafile); 1056 break; 1057 case QUOTACTL_QUOTAOFF: 1058 error = do_sys_quotactl_quotaoff(mp, 1059 args->u.quotaoff.qc_idtype); 1060 break; 1061 default: 1062 error = EINVAL; 1063 break; 1064 } 1065 1066 vrele(vp); 1067 return error; 1068 } 1069 1070 /* ARGSUSED */ 1071 int 1072 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1073 register_t *retval) 1074 { 1075 /* { 1076 syscallarg(const char *) path; 1077 syscallarg(struct quotactl_args *) args; 1078 } */ 1079 struct quotactl_args args; 1080 int error; 1081 1082 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1083 if (error) { 1084 return error; 1085 } 1086 1087 return do_sys_quotactl(SCARG(uap, path), &args); 1088 } 1089 1090 int 1091 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1092 int root) 1093 { 1094 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1095 int error = 0; 1096 1097 /* 1098 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1099 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1100 * overrides MNT_NOWAIT. 1101 */ 1102 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1103 (flags != MNT_WAIT && flags != 0)) { 1104 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1105 goto done; 1106 } 1107 1108 /* Get the filesystem stats now */ 1109 memset(sp, 0, sizeof(*sp)); 1110 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1111 return error; 1112 } 1113 1114 if (cwdi->cwdi_rdir == NULL) 1115 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1116 done: 1117 if (cwdi->cwdi_rdir != NULL) { 1118 size_t len; 1119 char *bp; 1120 char c; 1121 char *path = PNBUF_GET(); 1122 1123 bp = path + MAXPATHLEN; 1124 *--bp = '\0'; 1125 rw_enter(&cwdi->cwdi_lock, RW_READER); 1126 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1127 MAXPATHLEN / 2, 0, l); 1128 rw_exit(&cwdi->cwdi_lock); 1129 if (error) { 1130 PNBUF_PUT(path); 1131 return error; 1132 } 1133 len = strlen(bp); 1134 if (len != 1) { 1135 /* 1136 * for mount points that are below our root, we can see 1137 * them, so we fix up the pathname and return them. The 1138 * rest we cannot see, so we don't allow viewing the 1139 * data. 1140 */ 1141 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1142 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1143 (void)strlcpy(sp->f_mntonname, 1144 c == '\0' ? "/" : &sp->f_mntonname[len], 1145 sizeof(sp->f_mntonname)); 1146 } else { 1147 if (root) 1148 (void)strlcpy(sp->f_mntonname, "/", 1149 sizeof(sp->f_mntonname)); 1150 else 1151 error = EPERM; 1152 } 1153 } 1154 PNBUF_PUT(path); 1155 } 1156 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1157 return error; 1158 } 1159 1160 /* 1161 * Get filesystem statistics by path. 1162 */ 1163 int 1164 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1165 { 1166 struct mount *mp; 1167 int error; 1168 struct vnode *vp; 1169 1170 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1171 if (error != 0) 1172 return error; 1173 mp = vp->v_mount; 1174 error = dostatvfs(mp, sb, l, flags, 1); 1175 vrele(vp); 1176 return error; 1177 } 1178 1179 /* ARGSUSED */ 1180 int 1181 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1182 { 1183 /* { 1184 syscallarg(const char *) path; 1185 syscallarg(struct statvfs *) buf; 1186 syscallarg(int) flags; 1187 } */ 1188 struct statvfs *sb; 1189 int error; 1190 1191 sb = STATVFSBUF_GET(); 1192 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1193 if (error == 0) 1194 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1195 STATVFSBUF_PUT(sb); 1196 return error; 1197 } 1198 1199 /* 1200 * Get filesystem statistics by fd. 1201 */ 1202 int 1203 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1204 { 1205 file_t *fp; 1206 struct mount *mp; 1207 int error; 1208 1209 /* fd_getvnode() will use the descriptor for us */ 1210 if ((error = fd_getvnode(fd, &fp)) != 0) 1211 return (error); 1212 mp = fp->f_vnode->v_mount; 1213 error = dostatvfs(mp, sb, curlwp, flags, 1); 1214 fd_putfile(fd); 1215 return error; 1216 } 1217 1218 /* ARGSUSED */ 1219 int 1220 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1221 { 1222 /* { 1223 syscallarg(int) fd; 1224 syscallarg(struct statvfs *) buf; 1225 syscallarg(int) flags; 1226 } */ 1227 struct statvfs *sb; 1228 int error; 1229 1230 sb = STATVFSBUF_GET(); 1231 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1232 if (error == 0) 1233 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1234 STATVFSBUF_PUT(sb); 1235 return error; 1236 } 1237 1238 1239 /* 1240 * Get statistics on all filesystems. 1241 */ 1242 int 1243 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1244 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1245 register_t *retval) 1246 { 1247 int root = 0; 1248 mount_iterator_t *iter; 1249 struct proc *p = l->l_proc; 1250 struct mount *mp; 1251 struct statvfs *sb; 1252 size_t count, maxcount; 1253 int error = 0; 1254 1255 sb = STATVFSBUF_GET(); 1256 maxcount = bufsize / entry_sz; 1257 count = 0; 1258 mountlist_iterator_init(&iter); 1259 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1260 if (sfsp && count < maxcount) { 1261 error = dostatvfs(mp, sb, l, flags, 0); 1262 if (error) { 1263 error = 0; 1264 continue; 1265 } 1266 error = copyfn(sb, sfsp, entry_sz); 1267 if (error) 1268 goto out; 1269 sfsp = (char *)sfsp + entry_sz; 1270 root |= strcmp(sb->f_mntonname, "/") == 0; 1271 } 1272 count++; 1273 } 1274 1275 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1276 /* 1277 * fake a root entry 1278 */ 1279 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1280 sb, l, flags, 1); 1281 if (error != 0) 1282 goto out; 1283 if (sfsp) { 1284 error = copyfn(sb, sfsp, entry_sz); 1285 if (error != 0) 1286 goto out; 1287 } 1288 count++; 1289 } 1290 if (sfsp && count > maxcount) 1291 *retval = maxcount; 1292 else 1293 *retval = count; 1294 out: 1295 mountlist_iterator_destroy(iter); 1296 STATVFSBUF_PUT(sb); 1297 return error; 1298 } 1299 1300 int 1301 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1302 { 1303 /* { 1304 syscallarg(struct statvfs *) buf; 1305 syscallarg(size_t) bufsize; 1306 syscallarg(int) flags; 1307 } */ 1308 1309 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1310 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1311 } 1312 1313 /* 1314 * Change current working directory to a given file descriptor. 1315 */ 1316 /* ARGSUSED */ 1317 int 1318 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1319 { 1320 /* { 1321 syscallarg(int) fd; 1322 } */ 1323 struct proc *p = l->l_proc; 1324 struct cwdinfo *cwdi; 1325 struct vnode *vp, *tdp; 1326 struct mount *mp; 1327 file_t *fp; 1328 int error, fd; 1329 1330 /* fd_getvnode() will use the descriptor for us */ 1331 fd = SCARG(uap, fd); 1332 if ((error = fd_getvnode(fd, &fp)) != 0) 1333 return (error); 1334 vp = fp->f_vnode; 1335 1336 vref(vp); 1337 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1338 if (vp->v_type != VDIR) 1339 error = ENOTDIR; 1340 else 1341 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1342 if (error) { 1343 vput(vp); 1344 goto out; 1345 } 1346 while ((mp = vp->v_mountedhere) != NULL) { 1347 error = vfs_busy(mp); 1348 vput(vp); 1349 if (error != 0) 1350 goto out; 1351 error = VFS_ROOT(mp, &tdp); 1352 vfs_unbusy(mp); 1353 if (error) 1354 goto out; 1355 vp = tdp; 1356 } 1357 VOP_UNLOCK(vp); 1358 1359 /* 1360 * Disallow changing to a directory not under the process's 1361 * current root directory (if there is one). 1362 */ 1363 cwdi = p->p_cwdi; 1364 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1365 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1366 vrele(vp); 1367 error = EPERM; /* operation not permitted */ 1368 } else { 1369 vrele(cwdi->cwdi_cdir); 1370 cwdi->cwdi_cdir = vp; 1371 } 1372 rw_exit(&cwdi->cwdi_lock); 1373 1374 out: 1375 fd_putfile(fd); 1376 return (error); 1377 } 1378 1379 /* 1380 * Change this process's notion of the root directory to a given file 1381 * descriptor. 1382 */ 1383 int 1384 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1385 { 1386 struct proc *p = l->l_proc; 1387 struct vnode *vp; 1388 file_t *fp; 1389 int error, fd = SCARG(uap, fd); 1390 1391 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1392 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1393 return error; 1394 /* fd_getvnode() will use the descriptor for us */ 1395 if ((error = fd_getvnode(fd, &fp)) != 0) 1396 return error; 1397 vp = fp->f_vnode; 1398 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1399 if (vp->v_type != VDIR) 1400 error = ENOTDIR; 1401 else 1402 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1403 VOP_UNLOCK(vp); 1404 if (error) 1405 goto out; 1406 vref(vp); 1407 1408 change_root(p->p_cwdi, vp, l); 1409 1410 out: 1411 fd_putfile(fd); 1412 return (error); 1413 } 1414 1415 /* 1416 * Change current working directory (``.''). 1417 */ 1418 /* ARGSUSED */ 1419 int 1420 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1421 { 1422 /* { 1423 syscallarg(const char *) path; 1424 } */ 1425 struct proc *p = l->l_proc; 1426 struct cwdinfo *cwdi; 1427 int error; 1428 struct vnode *vp; 1429 1430 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1431 &vp, l)) != 0) 1432 return (error); 1433 cwdi = p->p_cwdi; 1434 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1435 vrele(cwdi->cwdi_cdir); 1436 cwdi->cwdi_cdir = vp; 1437 rw_exit(&cwdi->cwdi_lock); 1438 return (0); 1439 } 1440 1441 /* 1442 * Change notion of root (``/'') directory. 1443 */ 1444 /* ARGSUSED */ 1445 int 1446 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1447 { 1448 /* { 1449 syscallarg(const char *) path; 1450 } */ 1451 struct proc *p = l->l_proc; 1452 int error; 1453 struct vnode *vp; 1454 1455 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1456 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1457 return (error); 1458 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1459 &vp, l)) != 0) 1460 return (error); 1461 1462 change_root(p->p_cwdi, vp, l); 1463 1464 return (0); 1465 } 1466 1467 /* 1468 * Common routine for chroot and fchroot. 1469 * NB: callers need to properly authorize the change root operation. 1470 */ 1471 void 1472 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1473 { 1474 struct proc *p = l->l_proc; 1475 kauth_cred_t ncred; 1476 1477 ncred = kauth_cred_alloc(); 1478 1479 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1480 if (cwdi->cwdi_rdir != NULL) 1481 vrele(cwdi->cwdi_rdir); 1482 cwdi->cwdi_rdir = vp; 1483 1484 /* 1485 * Prevent escaping from chroot by putting the root under 1486 * the working directory. Silently chdir to / if we aren't 1487 * already there. 1488 */ 1489 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1490 /* 1491 * XXX would be more failsafe to change directory to a 1492 * deadfs node here instead 1493 */ 1494 vrele(cwdi->cwdi_cdir); 1495 vref(vp); 1496 cwdi->cwdi_cdir = vp; 1497 } 1498 rw_exit(&cwdi->cwdi_lock); 1499 1500 /* Get a write lock on the process credential. */ 1501 proc_crmod_enter(); 1502 1503 kauth_cred_clone(p->p_cred, ncred); 1504 kauth_proc_chroot(ncred, p->p_cwdi); 1505 1506 /* Broadcast our credentials to the process and other LWPs. */ 1507 proc_crmod_leave(ncred, p->p_cred, true); 1508 } 1509 1510 /* 1511 * Common routine for chroot and chdir. 1512 * XXX "where" should be enum uio_seg 1513 */ 1514 int 1515 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1516 { 1517 struct pathbuf *pb; 1518 struct nameidata nd; 1519 int error; 1520 1521 error = pathbuf_maybe_copyin(path, where, &pb); 1522 if (error) { 1523 return error; 1524 } 1525 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1526 if ((error = namei(&nd)) != 0) { 1527 pathbuf_destroy(pb); 1528 return error; 1529 } 1530 *vpp = nd.ni_vp; 1531 pathbuf_destroy(pb); 1532 1533 if ((*vpp)->v_type != VDIR) 1534 error = ENOTDIR; 1535 else 1536 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1537 1538 if (error) 1539 vput(*vpp); 1540 else 1541 VOP_UNLOCK(*vpp); 1542 return (error); 1543 } 1544 1545 /* 1546 * Internals of sys_open - path has already been converted into a pathbuf 1547 * (so we can easily reuse this function from other parts of the kernel, 1548 * like posix_spawn post-processing). 1549 */ 1550 int 1551 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1552 int open_mode, int *fd) 1553 { 1554 struct proc *p = l->l_proc; 1555 struct cwdinfo *cwdi = p->p_cwdi; 1556 file_t *fp; 1557 struct vnode *vp; 1558 int flags, cmode; 1559 int indx, error; 1560 struct nameidata nd; 1561 1562 if (open_flags & O_SEARCH) { 1563 open_flags &= ~(int)O_SEARCH; 1564 } 1565 1566 flags = FFLAGS(open_flags); 1567 if ((flags & (FREAD | FWRITE)) == 0) 1568 return EINVAL; 1569 1570 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1571 return error; 1572 } 1573 1574 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1575 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1576 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1577 if (dvp != NULL) 1578 NDAT(&nd, dvp); 1579 1580 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1581 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1582 fd_abort(p, fp, indx); 1583 if ((error == EDUPFD || error == EMOVEFD) && 1584 l->l_dupfd >= 0 && /* XXX from fdopen */ 1585 (error = 1586 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1587 *fd = indx; 1588 return 0; 1589 } 1590 if (error == ERESTART) 1591 error = EINTR; 1592 return error; 1593 } 1594 1595 l->l_dupfd = 0; 1596 vp = nd.ni_vp; 1597 1598 if ((error = open_setfp(l, fp, vp, indx, flags))) 1599 return error; 1600 1601 VOP_UNLOCK(vp); 1602 *fd = indx; 1603 fd_affix(p, fp, indx); 1604 return 0; 1605 } 1606 1607 int 1608 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1609 { 1610 struct pathbuf *pb; 1611 int error, oflags; 1612 1613 oflags = FFLAGS(open_flags); 1614 if ((oflags & (FREAD | FWRITE)) == 0) 1615 return EINVAL; 1616 1617 pb = pathbuf_create(path); 1618 if (pb == NULL) 1619 return ENOMEM; 1620 1621 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1622 pathbuf_destroy(pb); 1623 1624 return error; 1625 } 1626 1627 static int 1628 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1629 int mode, int *fd) 1630 { 1631 file_t *dfp = NULL; 1632 struct vnode *dvp = NULL; 1633 struct pathbuf *pb; 1634 int error; 1635 1636 if (path == NULL) { 1637 MODULE_CALL_HOOK(compat_10_openat_hook, (&pb), 1638 0, error); 1639 if (error) 1640 return error; 1641 } else { 1642 error = pathbuf_copyin(path, &pb); 1643 if (error) 1644 return error; 1645 } 1646 1647 if (fdat != AT_FDCWD) { 1648 /* fd_getvnode() will use the descriptor for us */ 1649 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1650 goto out; 1651 1652 dvp = dfp->f_vnode; 1653 } 1654 1655 error = do_open(l, dvp, pb, flags, mode, fd); 1656 1657 if (dfp != NULL) 1658 fd_putfile(fdat); 1659 out: 1660 pathbuf_destroy(pb); 1661 return error; 1662 } 1663 1664 int 1665 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1666 { 1667 /* { 1668 syscallarg(const char *) path; 1669 syscallarg(int) flags; 1670 syscallarg(int) mode; 1671 } */ 1672 int error; 1673 int fd; 1674 1675 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1676 SCARG(uap, flags), SCARG(uap, mode), &fd); 1677 1678 if (error == 0) 1679 *retval = fd; 1680 1681 return error; 1682 } 1683 1684 int 1685 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1686 { 1687 /* { 1688 syscallarg(int) fd; 1689 syscallarg(const char *) path; 1690 syscallarg(int) oflags; 1691 syscallarg(int) mode; 1692 } */ 1693 int error; 1694 int fd; 1695 1696 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1697 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1698 1699 if (error == 0) 1700 *retval = fd; 1701 1702 return error; 1703 } 1704 1705 static void 1706 vfs__fhfree(fhandle_t *fhp) 1707 { 1708 size_t fhsize; 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 mp = vp->v_mount; 1758 fidsize = 0; 1759 error = VFS_VPTOFH(vp, NULL, &fidsize); 1760 KASSERT(error != 0); 1761 if (error != E2BIG) { 1762 goto out; 1763 } 1764 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1765 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1766 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1767 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1768 if (error == 0) { 1769 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1770 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1771 *fhpp = fhp; 1772 } else { 1773 kmem_free(fhp, fhsize); 1774 } 1775 out: 1776 return error; 1777 } 1778 1779 void 1780 vfs_composefh_free(fhandle_t *fhp) 1781 { 1782 1783 vfs__fhfree(fhp); 1784 } 1785 1786 /* 1787 * vfs_fhtovp: lookup a vnode by a filehandle. 1788 */ 1789 1790 int 1791 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1792 { 1793 struct mount *mp; 1794 int error; 1795 1796 *vpp = NULL; 1797 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1798 if (mp == NULL) { 1799 error = ESTALE; 1800 goto out; 1801 } 1802 if (mp->mnt_op->vfs_fhtovp == NULL) { 1803 error = EOPNOTSUPP; 1804 goto out; 1805 } 1806 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1807 out: 1808 return error; 1809 } 1810 1811 /* 1812 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1813 * the needed size. 1814 */ 1815 1816 int 1817 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1818 { 1819 fhandle_t *fhp; 1820 int error; 1821 1822 if (fhsize > FHANDLE_SIZE_MAX) { 1823 return EINVAL; 1824 } 1825 if (fhsize < FHANDLE_SIZE_MIN) { 1826 return EINVAL; 1827 } 1828 again: 1829 fhp = kmem_alloc(fhsize, KM_SLEEP); 1830 error = copyin(ufhp, fhp, fhsize); 1831 if (error == 0) { 1832 /* XXX this check shouldn't be here */ 1833 if (FHANDLE_SIZE(fhp) == fhsize) { 1834 *fhpp = fhp; 1835 return 0; 1836 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1837 /* 1838 * a kludge for nfsv2 padded handles. 1839 */ 1840 size_t sz; 1841 1842 sz = FHANDLE_SIZE(fhp); 1843 kmem_free(fhp, fhsize); 1844 fhsize = sz; 1845 goto again; 1846 } else { 1847 /* 1848 * userland told us wrong size. 1849 */ 1850 error = EINVAL; 1851 } 1852 } 1853 kmem_free(fhp, fhsize); 1854 return error; 1855 } 1856 1857 void 1858 vfs_copyinfh_free(fhandle_t *fhp) 1859 { 1860 1861 vfs__fhfree(fhp); 1862 } 1863 1864 /* 1865 * Get file handle system call 1866 */ 1867 int 1868 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1869 { 1870 /* { 1871 syscallarg(char *) fname; 1872 syscallarg(fhandle_t *) fhp; 1873 syscallarg(size_t *) fh_size; 1874 } */ 1875 struct vnode *vp; 1876 fhandle_t *fh; 1877 int error; 1878 struct pathbuf *pb; 1879 struct nameidata nd; 1880 size_t sz; 1881 size_t usz; 1882 1883 /* 1884 * Must be super user 1885 */ 1886 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1887 0, NULL, NULL, NULL); 1888 if (error) 1889 return (error); 1890 1891 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1892 if (error) { 1893 return error; 1894 } 1895 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1896 error = namei(&nd); 1897 if (error) { 1898 pathbuf_destroy(pb); 1899 return error; 1900 } 1901 vp = nd.ni_vp; 1902 pathbuf_destroy(pb); 1903 1904 error = vfs_composefh_alloc(vp, &fh); 1905 vput(vp); 1906 if (error != 0) { 1907 return error; 1908 } 1909 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1910 if (error != 0) { 1911 goto out; 1912 } 1913 sz = FHANDLE_SIZE(fh); 1914 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1915 if (error != 0) { 1916 goto out; 1917 } 1918 if (usz >= sz) { 1919 error = copyout(fh, SCARG(uap, fhp), sz); 1920 } else { 1921 error = E2BIG; 1922 } 1923 out: 1924 vfs_composefh_free(fh); 1925 return (error); 1926 } 1927 1928 /* 1929 * Open a file given a file handle. 1930 * 1931 * Check permissions, allocate an open file structure, 1932 * and call the device open routine if any. 1933 */ 1934 1935 int 1936 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1937 register_t *retval) 1938 { 1939 file_t *fp; 1940 struct vnode *vp = NULL; 1941 kauth_cred_t cred = l->l_cred; 1942 file_t *nfp; 1943 int indx, error; 1944 struct vattr va; 1945 fhandle_t *fh; 1946 int flags; 1947 proc_t *p; 1948 1949 p = curproc; 1950 1951 /* 1952 * Must be super user 1953 */ 1954 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1955 0, NULL, NULL, NULL))) 1956 return (error); 1957 1958 if (oflags & O_SEARCH) { 1959 oflags &= ~(int)O_SEARCH; 1960 } 1961 1962 flags = FFLAGS(oflags); 1963 if ((flags & (FREAD | FWRITE)) == 0) 1964 return (EINVAL); 1965 if ((flags & O_CREAT)) 1966 return (EINVAL); 1967 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1968 return (error); 1969 fp = nfp; 1970 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1971 if (error != 0) { 1972 goto bad; 1973 } 1974 error = vfs_fhtovp(fh, &vp); 1975 vfs_copyinfh_free(fh); 1976 if (error != 0) { 1977 goto bad; 1978 } 1979 1980 /* Now do an effective vn_open */ 1981 1982 if (vp->v_type == VSOCK) { 1983 error = EOPNOTSUPP; 1984 goto bad; 1985 } 1986 error = vn_openchk(vp, cred, flags); 1987 if (error != 0) 1988 goto bad; 1989 if (flags & O_TRUNC) { 1990 VOP_UNLOCK(vp); /* XXX */ 1991 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1992 vattr_null(&va); 1993 va.va_size = 0; 1994 error = VOP_SETATTR(vp, &va, cred); 1995 if (error) 1996 goto bad; 1997 } 1998 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1999 goto bad; 2000 if (flags & FWRITE) { 2001 mutex_enter(vp->v_interlock); 2002 vp->v_writecount++; 2003 mutex_exit(vp->v_interlock); 2004 } 2005 2006 /* done with modified vn_open, now finish what sys_open does. */ 2007 if ((error = open_setfp(l, fp, vp, indx, flags))) 2008 return error; 2009 2010 VOP_UNLOCK(vp); 2011 *retval = indx; 2012 fd_affix(p, fp, indx); 2013 return (0); 2014 2015 bad: 2016 fd_abort(p, fp, indx); 2017 if (vp != NULL) 2018 vput(vp); 2019 return (error); 2020 } 2021 2022 int 2023 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2024 { 2025 /* { 2026 syscallarg(const void *) fhp; 2027 syscallarg(size_t) fh_size; 2028 syscallarg(int) flags; 2029 } */ 2030 2031 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2032 SCARG(uap, flags), retval); 2033 } 2034 2035 int 2036 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2037 { 2038 int error; 2039 fhandle_t *fh; 2040 struct vnode *vp; 2041 2042 /* 2043 * Must be super user 2044 */ 2045 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2046 0, NULL, NULL, NULL))) 2047 return (error); 2048 2049 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2050 if (error != 0) 2051 return error; 2052 2053 error = vfs_fhtovp(fh, &vp); 2054 vfs_copyinfh_free(fh); 2055 if (error != 0) 2056 return error; 2057 2058 error = vn_stat(vp, sb); 2059 vput(vp); 2060 return error; 2061 } 2062 2063 2064 /* ARGSUSED */ 2065 int 2066 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2067 { 2068 /* { 2069 syscallarg(const void *) fhp; 2070 syscallarg(size_t) fh_size; 2071 syscallarg(struct stat *) sb; 2072 } */ 2073 struct stat sb; 2074 int error; 2075 2076 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2077 if (error) 2078 return error; 2079 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2080 } 2081 2082 int 2083 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2084 int flags) 2085 { 2086 fhandle_t *fh; 2087 struct mount *mp; 2088 struct vnode *vp; 2089 int error; 2090 2091 /* 2092 * Must be super user 2093 */ 2094 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2095 0, NULL, NULL, NULL))) 2096 return error; 2097 2098 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2099 if (error != 0) 2100 return error; 2101 2102 error = vfs_fhtovp(fh, &vp); 2103 vfs_copyinfh_free(fh); 2104 if (error != 0) 2105 return error; 2106 2107 mp = vp->v_mount; 2108 error = dostatvfs(mp, sb, l, flags, 1); 2109 vput(vp); 2110 return error; 2111 } 2112 2113 /* ARGSUSED */ 2114 int 2115 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2116 { 2117 /* { 2118 syscallarg(const void *) fhp; 2119 syscallarg(size_t) fh_size; 2120 syscallarg(struct statvfs *) buf; 2121 syscallarg(int) flags; 2122 } */ 2123 struct statvfs *sb = STATVFSBUF_GET(); 2124 int error; 2125 2126 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2127 SCARG(uap, flags)); 2128 if (error == 0) 2129 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2130 STATVFSBUF_PUT(sb); 2131 return error; 2132 } 2133 2134 /* 2135 * Create a special file. 2136 */ 2137 /* ARGSUSED */ 2138 int 2139 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2140 register_t *retval) 2141 { 2142 /* { 2143 syscallarg(const char *) path; 2144 syscallarg(mode_t) mode; 2145 syscallarg(dev_t) dev; 2146 } */ 2147 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2148 SCARG(uap, dev), retval, UIO_USERSPACE); 2149 } 2150 2151 int 2152 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2153 register_t *retval) 2154 { 2155 /* { 2156 syscallarg(int) fd; 2157 syscallarg(const char *) path; 2158 syscallarg(mode_t) mode; 2159 syscallarg(int) pad; 2160 syscallarg(dev_t) dev; 2161 } */ 2162 2163 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2164 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2165 } 2166 2167 int 2168 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2169 register_t *retval, enum uio_seg seg) 2170 { 2171 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2172 } 2173 2174 int 2175 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2176 dev_t dev, register_t *retval, enum uio_seg seg) 2177 { 2178 struct proc *p = l->l_proc; 2179 struct vnode *vp; 2180 struct vattr vattr; 2181 int error, optype; 2182 struct pathbuf *pb; 2183 struct nameidata nd; 2184 const char *pathstring; 2185 2186 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2187 0, NULL, NULL, NULL)) != 0) 2188 return (error); 2189 2190 optype = VOP_MKNOD_DESCOFFSET; 2191 2192 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2193 if (error) { 2194 return error; 2195 } 2196 pathstring = pathbuf_stringcopy_get(pb); 2197 if (pathstring == NULL) { 2198 pathbuf_destroy(pb); 2199 return ENOMEM; 2200 } 2201 2202 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2203 2204 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2205 goto out; 2206 vp = nd.ni_vp; 2207 2208 if (vp != NULL) 2209 error = EEXIST; 2210 else { 2211 vattr_null(&vattr); 2212 /* We will read cwdi->cwdi_cmask unlocked. */ 2213 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2214 vattr.va_rdev = dev; 2215 2216 switch (mode & S_IFMT) { 2217 case S_IFMT: /* used by badsect to flag bad sectors */ 2218 vattr.va_type = VBAD; 2219 break; 2220 case S_IFCHR: 2221 vattr.va_type = VCHR; 2222 break; 2223 case S_IFBLK: 2224 vattr.va_type = VBLK; 2225 break; 2226 case S_IFWHT: 2227 optype = VOP_WHITEOUT_DESCOFFSET; 2228 break; 2229 case S_IFREG: 2230 #if NVERIEXEC > 0 2231 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2232 O_CREAT); 2233 #endif /* NVERIEXEC > 0 */ 2234 vattr.va_type = VREG; 2235 vattr.va_rdev = VNOVAL; 2236 optype = VOP_CREATE_DESCOFFSET; 2237 break; 2238 default: 2239 error = EINVAL; 2240 break; 2241 } 2242 } 2243 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2244 && vattr.va_rdev == VNOVAL) 2245 error = EINVAL; 2246 if (!error) { 2247 switch (optype) { 2248 case VOP_WHITEOUT_DESCOFFSET: 2249 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2250 if (error) 2251 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2252 vput(nd.ni_dvp); 2253 break; 2254 2255 case VOP_MKNOD_DESCOFFSET: 2256 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2257 &nd.ni_cnd, &vattr); 2258 if (error == 0) 2259 vrele(nd.ni_vp); 2260 vput(nd.ni_dvp); 2261 break; 2262 2263 case VOP_CREATE_DESCOFFSET: 2264 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2265 &nd.ni_cnd, &vattr); 2266 if (error == 0) 2267 vrele(nd.ni_vp); 2268 vput(nd.ni_dvp); 2269 break; 2270 } 2271 } else { 2272 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2273 if (nd.ni_dvp == vp) 2274 vrele(nd.ni_dvp); 2275 else 2276 vput(nd.ni_dvp); 2277 if (vp) 2278 vrele(vp); 2279 } 2280 out: 2281 pathbuf_stringcopy_put(pb, pathstring); 2282 pathbuf_destroy(pb); 2283 return (error); 2284 } 2285 2286 /* 2287 * Create a named pipe. 2288 */ 2289 /* ARGSUSED */ 2290 int 2291 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2292 { 2293 /* { 2294 syscallarg(const char *) path; 2295 syscallarg(int) mode; 2296 } */ 2297 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2298 } 2299 2300 int 2301 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2302 register_t *retval) 2303 { 2304 /* { 2305 syscallarg(int) fd; 2306 syscallarg(const char *) path; 2307 syscallarg(int) mode; 2308 } */ 2309 2310 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2311 SCARG(uap, mode)); 2312 } 2313 2314 static int 2315 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2316 { 2317 struct proc *p = l->l_proc; 2318 struct vattr vattr; 2319 int error; 2320 struct pathbuf *pb; 2321 struct nameidata nd; 2322 2323 error = pathbuf_copyin(path, &pb); 2324 if (error) { 2325 return error; 2326 } 2327 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2328 2329 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2330 pathbuf_destroy(pb); 2331 return error; 2332 } 2333 if (nd.ni_vp != NULL) { 2334 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2335 if (nd.ni_dvp == nd.ni_vp) 2336 vrele(nd.ni_dvp); 2337 else 2338 vput(nd.ni_dvp); 2339 vrele(nd.ni_vp); 2340 pathbuf_destroy(pb); 2341 return (EEXIST); 2342 } 2343 vattr_null(&vattr); 2344 vattr.va_type = VFIFO; 2345 /* We will read cwdi->cwdi_cmask unlocked. */ 2346 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2347 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2348 if (error == 0) 2349 vrele(nd.ni_vp); 2350 vput(nd.ni_dvp); 2351 pathbuf_destroy(pb); 2352 return (error); 2353 } 2354 2355 /* 2356 * Make a hard file link. 2357 */ 2358 /* ARGSUSED */ 2359 int 2360 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2361 const char *link, int follow, register_t *retval) 2362 { 2363 struct vnode *vp; 2364 struct pathbuf *linkpb; 2365 struct nameidata nd; 2366 namei_simple_flags_t ns_flags; 2367 int error; 2368 2369 if (follow & AT_SYMLINK_FOLLOW) 2370 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2371 else 2372 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2373 2374 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2375 if (error != 0) 2376 return (error); 2377 error = pathbuf_copyin(link, &linkpb); 2378 if (error) { 2379 goto out1; 2380 } 2381 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2382 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2383 goto out2; 2384 if (nd.ni_vp) { 2385 error = EEXIST; 2386 goto abortop; 2387 } 2388 /* Prevent hard links on directories. */ 2389 if (vp->v_type == VDIR) { 2390 error = EPERM; 2391 goto abortop; 2392 } 2393 /* Prevent cross-mount operation. */ 2394 if (nd.ni_dvp->v_mount != vp->v_mount) { 2395 error = EXDEV; 2396 goto abortop; 2397 } 2398 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2399 VOP_UNLOCK(nd.ni_dvp); 2400 vrele(nd.ni_dvp); 2401 out2: 2402 pathbuf_destroy(linkpb); 2403 out1: 2404 vrele(vp); 2405 return (error); 2406 abortop: 2407 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2408 if (nd.ni_dvp == nd.ni_vp) 2409 vrele(nd.ni_dvp); 2410 else 2411 vput(nd.ni_dvp); 2412 if (nd.ni_vp != NULL) 2413 vrele(nd.ni_vp); 2414 goto out2; 2415 } 2416 2417 int 2418 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2419 { 2420 /* { 2421 syscallarg(const char *) path; 2422 syscallarg(const char *) link; 2423 } */ 2424 const char *path = SCARG(uap, path); 2425 const char *link = SCARG(uap, link); 2426 2427 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2428 AT_SYMLINK_FOLLOW, retval); 2429 } 2430 2431 int 2432 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2433 register_t *retval) 2434 { 2435 /* { 2436 syscallarg(int) fd1; 2437 syscallarg(const char *) name1; 2438 syscallarg(int) fd2; 2439 syscallarg(const char *) name2; 2440 syscallarg(int) flags; 2441 } */ 2442 int fd1 = SCARG(uap, fd1); 2443 const char *name1 = SCARG(uap, name1); 2444 int fd2 = SCARG(uap, fd2); 2445 const char *name2 = SCARG(uap, name2); 2446 int follow; 2447 2448 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2449 2450 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2451 } 2452 2453 2454 int 2455 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2456 { 2457 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2458 } 2459 2460 static int 2461 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2462 const char *link, enum uio_seg seg) 2463 { 2464 struct proc *p = curproc; 2465 struct vattr vattr; 2466 char *path; 2467 int error; 2468 size_t len; 2469 struct pathbuf *linkpb; 2470 struct nameidata nd; 2471 2472 KASSERT(l != NULL || fdat == AT_FDCWD); 2473 2474 path = PNBUF_GET(); 2475 if (seg == UIO_USERSPACE) { 2476 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2477 goto out1; 2478 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2479 goto out1; 2480 } else { 2481 len = strlen(patharg) + 1; 2482 KASSERT(len <= MAXPATHLEN); 2483 memcpy(path, patharg, len); 2484 linkpb = pathbuf_create(link); 2485 if (linkpb == NULL) { 2486 error = ENOMEM; 2487 goto out1; 2488 } 2489 } 2490 ktrkuser("symlink-target", path, len - 1); 2491 2492 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2493 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2494 goto out2; 2495 if (nd.ni_vp) { 2496 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2497 if (nd.ni_dvp == nd.ni_vp) 2498 vrele(nd.ni_dvp); 2499 else 2500 vput(nd.ni_dvp); 2501 vrele(nd.ni_vp); 2502 error = EEXIST; 2503 goto out2; 2504 } 2505 vattr_null(&vattr); 2506 vattr.va_type = VLNK; 2507 /* We will read cwdi->cwdi_cmask unlocked. */ 2508 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2509 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2510 if (error == 0) 2511 vrele(nd.ni_vp); 2512 vput(nd.ni_dvp); 2513 out2: 2514 pathbuf_destroy(linkpb); 2515 out1: 2516 PNBUF_PUT(path); 2517 return (error); 2518 } 2519 2520 /* 2521 * Make a symbolic link. 2522 */ 2523 /* ARGSUSED */ 2524 int 2525 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2526 { 2527 /* { 2528 syscallarg(const char *) path; 2529 syscallarg(const char *) link; 2530 } */ 2531 2532 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2533 UIO_USERSPACE); 2534 } 2535 2536 int 2537 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2538 register_t *retval) 2539 { 2540 /* { 2541 syscallarg(const char *) path1; 2542 syscallarg(int) fd; 2543 syscallarg(const char *) path2; 2544 } */ 2545 2546 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2547 SCARG(uap, path2), UIO_USERSPACE); 2548 } 2549 2550 /* 2551 * Delete a whiteout from the filesystem. 2552 */ 2553 /* ARGSUSED */ 2554 int 2555 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2556 { 2557 /* { 2558 syscallarg(const char *) path; 2559 } */ 2560 int error; 2561 struct pathbuf *pb; 2562 struct nameidata nd; 2563 2564 error = pathbuf_copyin(SCARG(uap, path), &pb); 2565 if (error) { 2566 return error; 2567 } 2568 2569 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2570 error = namei(&nd); 2571 if (error) { 2572 pathbuf_destroy(pb); 2573 return (error); 2574 } 2575 2576 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2577 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2578 if (nd.ni_dvp == nd.ni_vp) 2579 vrele(nd.ni_dvp); 2580 else 2581 vput(nd.ni_dvp); 2582 if (nd.ni_vp) 2583 vrele(nd.ni_vp); 2584 pathbuf_destroy(pb); 2585 return (EEXIST); 2586 } 2587 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2588 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2589 vput(nd.ni_dvp); 2590 pathbuf_destroy(pb); 2591 return (error); 2592 } 2593 2594 /* 2595 * Delete a name from the filesystem. 2596 */ 2597 /* ARGSUSED */ 2598 int 2599 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2600 { 2601 /* { 2602 syscallarg(const char *) path; 2603 } */ 2604 2605 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2606 } 2607 2608 int 2609 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2610 register_t *retval) 2611 { 2612 /* { 2613 syscallarg(int) fd; 2614 syscallarg(const char *) path; 2615 syscallarg(int) flag; 2616 } */ 2617 2618 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2619 SCARG(uap, flag), UIO_USERSPACE); 2620 } 2621 2622 int 2623 do_sys_unlink(const char *arg, enum uio_seg seg) 2624 { 2625 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2626 } 2627 2628 static int 2629 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2630 enum uio_seg seg) 2631 { 2632 struct vnode *vp; 2633 int error; 2634 struct pathbuf *pb; 2635 struct nameidata nd; 2636 const char *pathstring; 2637 2638 KASSERT(l != NULL || fdat == AT_FDCWD); 2639 2640 error = pathbuf_maybe_copyin(arg, seg, &pb); 2641 if (error) { 2642 return error; 2643 } 2644 pathstring = pathbuf_stringcopy_get(pb); 2645 if (pathstring == NULL) { 2646 pathbuf_destroy(pb); 2647 return ENOMEM; 2648 } 2649 2650 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2651 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2652 goto out; 2653 vp = nd.ni_vp; 2654 2655 /* 2656 * The root of a mounted filesystem cannot be deleted. 2657 */ 2658 if ((vp->v_vflag & VV_ROOT) != 0) { 2659 error = EBUSY; 2660 goto abort; 2661 } 2662 2663 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2664 error = EBUSY; 2665 goto abort; 2666 } 2667 2668 /* 2669 * No rmdir "." please. 2670 */ 2671 if (nd.ni_dvp == vp) { 2672 error = EINVAL; 2673 goto abort; 2674 } 2675 2676 /* 2677 * AT_REMOVEDIR is required to remove a directory 2678 */ 2679 if (vp->v_type == VDIR) { 2680 if (!(flags & AT_REMOVEDIR)) { 2681 error = EPERM; 2682 goto abort; 2683 } else { 2684 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2685 vput(nd.ni_dvp); 2686 goto out; 2687 } 2688 } 2689 2690 /* 2691 * Starting here we only deal with non directories. 2692 */ 2693 if (flags & AT_REMOVEDIR) { 2694 error = ENOTDIR; 2695 goto abort; 2696 } 2697 2698 #if NVERIEXEC > 0 2699 /* Handle remove requests for veriexec entries. */ 2700 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2701 goto abort; 2702 } 2703 #endif /* NVERIEXEC > 0 */ 2704 2705 #ifdef FILEASSOC 2706 (void)fileassoc_file_delete(vp); 2707 #endif /* FILEASSOC */ 2708 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2709 vput(nd.ni_dvp); 2710 goto out; 2711 2712 abort: 2713 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2714 if (nd.ni_dvp == vp) 2715 vrele(nd.ni_dvp); 2716 else 2717 vput(nd.ni_dvp); 2718 vput(vp); 2719 2720 out: 2721 pathbuf_stringcopy_put(pb, pathstring); 2722 pathbuf_destroy(pb); 2723 return (error); 2724 } 2725 2726 /* 2727 * Reposition read/write file offset. 2728 */ 2729 int 2730 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2731 { 2732 /* { 2733 syscallarg(int) fd; 2734 syscallarg(int) pad; 2735 syscallarg(off_t) offset; 2736 syscallarg(int) whence; 2737 } */ 2738 kauth_cred_t cred = l->l_cred; 2739 file_t *fp; 2740 struct vnode *vp; 2741 struct vattr vattr; 2742 off_t newoff; 2743 int error, fd; 2744 2745 fd = SCARG(uap, fd); 2746 2747 if ((fp = fd_getfile(fd)) == NULL) 2748 return (EBADF); 2749 2750 vp = fp->f_vnode; 2751 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2752 error = ESPIPE; 2753 goto out; 2754 } 2755 2756 vn_lock(vp, LK_SHARED | LK_RETRY); 2757 2758 switch (SCARG(uap, whence)) { 2759 case SEEK_CUR: 2760 newoff = fp->f_offset + SCARG(uap, offset); 2761 break; 2762 case SEEK_END: 2763 error = VOP_GETATTR(vp, &vattr, cred); 2764 if (error) { 2765 VOP_UNLOCK(vp); 2766 goto out; 2767 } 2768 newoff = SCARG(uap, offset) + vattr.va_size; 2769 break; 2770 case SEEK_SET: 2771 newoff = SCARG(uap, offset); 2772 break; 2773 default: 2774 error = EINVAL; 2775 VOP_UNLOCK(vp); 2776 goto out; 2777 } 2778 VOP_UNLOCK(vp); 2779 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2780 *(off_t *)retval = fp->f_offset = newoff; 2781 } 2782 out: 2783 fd_putfile(fd); 2784 return (error); 2785 } 2786 2787 /* 2788 * Positional read system call. 2789 */ 2790 int 2791 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2792 { 2793 /* { 2794 syscallarg(int) fd; 2795 syscallarg(void *) buf; 2796 syscallarg(size_t) nbyte; 2797 syscallarg(off_t) offset; 2798 } */ 2799 file_t *fp; 2800 struct vnode *vp; 2801 off_t offset; 2802 int error, fd = SCARG(uap, fd); 2803 2804 if ((fp = fd_getfile(fd)) == NULL) 2805 return (EBADF); 2806 2807 if ((fp->f_flag & FREAD) == 0) { 2808 fd_putfile(fd); 2809 return (EBADF); 2810 } 2811 2812 vp = fp->f_vnode; 2813 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2814 error = ESPIPE; 2815 goto out; 2816 } 2817 2818 offset = SCARG(uap, offset); 2819 2820 /* 2821 * XXX This works because no file systems actually 2822 * XXX take any action on the seek operation. 2823 */ 2824 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2825 goto out; 2826 2827 /* dofileread() will unuse the descriptor for us */ 2828 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2829 &offset, 0, retval)); 2830 2831 out: 2832 fd_putfile(fd); 2833 return (error); 2834 } 2835 2836 /* 2837 * Positional scatter read system call. 2838 */ 2839 int 2840 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2841 { 2842 /* { 2843 syscallarg(int) fd; 2844 syscallarg(const struct iovec *) iovp; 2845 syscallarg(int) iovcnt; 2846 syscallarg(off_t) offset; 2847 } */ 2848 off_t offset = SCARG(uap, offset); 2849 2850 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2851 SCARG(uap, iovcnt), &offset, 0, retval); 2852 } 2853 2854 /* 2855 * Positional write system call. 2856 */ 2857 int 2858 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2859 { 2860 /* { 2861 syscallarg(int) fd; 2862 syscallarg(const void *) buf; 2863 syscallarg(size_t) nbyte; 2864 syscallarg(off_t) offset; 2865 } */ 2866 file_t *fp; 2867 struct vnode *vp; 2868 off_t offset; 2869 int error, fd = SCARG(uap, fd); 2870 2871 if ((fp = fd_getfile(fd)) == NULL) 2872 return (EBADF); 2873 2874 if ((fp->f_flag & FWRITE) == 0) { 2875 fd_putfile(fd); 2876 return (EBADF); 2877 } 2878 2879 vp = fp->f_vnode; 2880 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2881 error = ESPIPE; 2882 goto out; 2883 } 2884 2885 offset = SCARG(uap, offset); 2886 2887 /* 2888 * XXX This works because no file systems actually 2889 * XXX take any action on the seek operation. 2890 */ 2891 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2892 goto out; 2893 2894 /* dofilewrite() will unuse the descriptor for us */ 2895 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2896 &offset, 0, retval)); 2897 2898 out: 2899 fd_putfile(fd); 2900 return (error); 2901 } 2902 2903 /* 2904 * Positional gather write system call. 2905 */ 2906 int 2907 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2908 { 2909 /* { 2910 syscallarg(int) fd; 2911 syscallarg(const struct iovec *) iovp; 2912 syscallarg(int) iovcnt; 2913 syscallarg(off_t) offset; 2914 } */ 2915 off_t offset = SCARG(uap, offset); 2916 2917 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2918 SCARG(uap, iovcnt), &offset, 0, retval); 2919 } 2920 2921 /* 2922 * Check access permissions. 2923 */ 2924 int 2925 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2926 { 2927 /* { 2928 syscallarg(const char *) path; 2929 syscallarg(int) flags; 2930 } */ 2931 2932 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2933 SCARG(uap, flags), 0); 2934 } 2935 2936 int 2937 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2938 int mode, int flags) 2939 { 2940 kauth_cred_t cred; 2941 struct vnode *vp; 2942 int error, nd_flag, vmode; 2943 struct pathbuf *pb; 2944 struct nameidata nd; 2945 2946 CTASSERT(F_OK == 0); 2947 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2948 /* nonsense mode */ 2949 return EINVAL; 2950 } 2951 2952 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2953 if (flags & AT_SYMLINK_NOFOLLOW) 2954 nd_flag &= ~FOLLOW; 2955 2956 error = pathbuf_copyin(path, &pb); 2957 if (error) 2958 return error; 2959 2960 NDINIT(&nd, LOOKUP, nd_flag, pb); 2961 2962 /* Override default credentials */ 2963 cred = kauth_cred_dup(l->l_cred); 2964 if (!(flags & AT_EACCESS)) { 2965 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2966 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2967 } 2968 nd.ni_cnd.cn_cred = cred; 2969 2970 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2971 pathbuf_destroy(pb); 2972 goto out; 2973 } 2974 vp = nd.ni_vp; 2975 pathbuf_destroy(pb); 2976 2977 /* Flags == 0 means only check for existence. */ 2978 if (mode) { 2979 vmode = 0; 2980 if (mode & R_OK) 2981 vmode |= VREAD; 2982 if (mode & W_OK) 2983 vmode |= VWRITE; 2984 if (mode & X_OK) 2985 vmode |= VEXEC; 2986 2987 error = VOP_ACCESS(vp, vmode, cred); 2988 if (!error && (vmode & VWRITE)) 2989 error = vn_writechk(vp); 2990 } 2991 vput(vp); 2992 out: 2993 kauth_cred_free(cred); 2994 return (error); 2995 } 2996 2997 int 2998 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 2999 register_t *retval) 3000 { 3001 /* { 3002 syscallarg(int) fd; 3003 syscallarg(const char *) path; 3004 syscallarg(int) amode; 3005 syscallarg(int) flag; 3006 } */ 3007 3008 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3009 SCARG(uap, amode), SCARG(uap, flag)); 3010 } 3011 3012 /* 3013 * Common code for all sys_stat functions, including compat versions. 3014 */ 3015 int 3016 do_sys_stat(const char *userpath, unsigned int nd_flag, 3017 struct stat *sb) 3018 { 3019 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3020 } 3021 3022 int 3023 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3024 unsigned int nd_flag, struct stat *sb) 3025 { 3026 int error; 3027 struct pathbuf *pb; 3028 struct nameidata nd; 3029 3030 KASSERT(l != NULL || fdat == AT_FDCWD); 3031 3032 error = pathbuf_copyin(userpath, &pb); 3033 if (error) { 3034 return error; 3035 } 3036 3037 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3038 3039 error = fd_nameiat(l, fdat, &nd); 3040 if (error != 0) { 3041 pathbuf_destroy(pb); 3042 return error; 3043 } 3044 error = vn_stat(nd.ni_vp, sb); 3045 vput(nd.ni_vp); 3046 pathbuf_destroy(pb); 3047 return error; 3048 } 3049 3050 /* 3051 * Get file status; this version follows links. 3052 */ 3053 /* ARGSUSED */ 3054 int 3055 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3056 { 3057 /* { 3058 syscallarg(const char *) path; 3059 syscallarg(struct stat *) ub; 3060 } */ 3061 struct stat sb; 3062 int error; 3063 3064 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3065 if (error) 3066 return error; 3067 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3068 } 3069 3070 /* 3071 * Get file status; this version does not follow links. 3072 */ 3073 /* ARGSUSED */ 3074 int 3075 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3076 { 3077 /* { 3078 syscallarg(const char *) path; 3079 syscallarg(struct stat *) ub; 3080 } */ 3081 struct stat sb; 3082 int error; 3083 3084 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3085 if (error) 3086 return error; 3087 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3088 } 3089 3090 int 3091 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3092 register_t *retval) 3093 { 3094 /* { 3095 syscallarg(int) fd; 3096 syscallarg(const char *) path; 3097 syscallarg(struct stat *) buf; 3098 syscallarg(int) flag; 3099 } */ 3100 unsigned int nd_flag; 3101 struct stat sb; 3102 int error; 3103 3104 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3105 nd_flag = NOFOLLOW; 3106 else 3107 nd_flag = FOLLOW; 3108 3109 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3110 &sb); 3111 if (error) 3112 return error; 3113 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3114 } 3115 3116 /* 3117 * Get configurable pathname variables. 3118 */ 3119 /* ARGSUSED */ 3120 int 3121 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3122 { 3123 /* { 3124 syscallarg(const char *) path; 3125 syscallarg(int) name; 3126 } */ 3127 int error; 3128 struct pathbuf *pb; 3129 struct nameidata nd; 3130 3131 error = pathbuf_copyin(SCARG(uap, path), &pb); 3132 if (error) { 3133 return error; 3134 } 3135 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3136 if ((error = namei(&nd)) != 0) { 3137 pathbuf_destroy(pb); 3138 return (error); 3139 } 3140 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3141 vput(nd.ni_vp); 3142 pathbuf_destroy(pb); 3143 return (error); 3144 } 3145 3146 /* 3147 * Return target name of a symbolic link. 3148 */ 3149 /* ARGSUSED */ 3150 int 3151 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3152 register_t *retval) 3153 { 3154 /* { 3155 syscallarg(const char *) path; 3156 syscallarg(char *) buf; 3157 syscallarg(size_t) count; 3158 } */ 3159 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3160 SCARG(uap, buf), SCARG(uap, count), retval); 3161 } 3162 3163 static int 3164 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3165 size_t count, register_t *retval) 3166 { 3167 struct vnode *vp; 3168 struct iovec aiov; 3169 struct uio auio; 3170 int error; 3171 struct pathbuf *pb; 3172 struct nameidata nd; 3173 3174 error = pathbuf_copyin(path, &pb); 3175 if (error) { 3176 return error; 3177 } 3178 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3179 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3180 pathbuf_destroy(pb); 3181 return error; 3182 } 3183 vp = nd.ni_vp; 3184 pathbuf_destroy(pb); 3185 if (vp->v_type != VLNK) 3186 error = EINVAL; 3187 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3188 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3189 aiov.iov_base = buf; 3190 aiov.iov_len = count; 3191 auio.uio_iov = &aiov; 3192 auio.uio_iovcnt = 1; 3193 auio.uio_offset = 0; 3194 auio.uio_rw = UIO_READ; 3195 KASSERT(l == curlwp); 3196 auio.uio_vmspace = l->l_proc->p_vmspace; 3197 auio.uio_resid = count; 3198 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3199 *retval = count - auio.uio_resid; 3200 } 3201 vput(vp); 3202 return (error); 3203 } 3204 3205 int 3206 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3207 register_t *retval) 3208 { 3209 /* { 3210 syscallarg(int) fd; 3211 syscallarg(const char *) path; 3212 syscallarg(char *) buf; 3213 syscallarg(size_t) bufsize; 3214 } */ 3215 3216 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3217 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3218 } 3219 3220 /* 3221 * Change flags of a file given a path name. 3222 */ 3223 /* ARGSUSED */ 3224 int 3225 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3226 { 3227 /* { 3228 syscallarg(const char *) path; 3229 syscallarg(u_long) flags; 3230 } */ 3231 struct vnode *vp; 3232 int error; 3233 3234 error = namei_simple_user(SCARG(uap, path), 3235 NSM_FOLLOW_TRYEMULROOT, &vp); 3236 if (error != 0) 3237 return (error); 3238 error = change_flags(vp, SCARG(uap, flags), l); 3239 vput(vp); 3240 return (error); 3241 } 3242 3243 /* 3244 * Change flags of a file given a file descriptor. 3245 */ 3246 /* ARGSUSED */ 3247 int 3248 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3249 { 3250 /* { 3251 syscallarg(int) fd; 3252 syscallarg(u_long) flags; 3253 } */ 3254 struct vnode *vp; 3255 file_t *fp; 3256 int error; 3257 3258 /* fd_getvnode() will use the descriptor for us */ 3259 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3260 return (error); 3261 vp = fp->f_vnode; 3262 error = change_flags(vp, SCARG(uap, flags), l); 3263 VOP_UNLOCK(vp); 3264 fd_putfile(SCARG(uap, fd)); 3265 return (error); 3266 } 3267 3268 /* 3269 * Change flags of a file given a path name; this version does 3270 * not follow links. 3271 */ 3272 int 3273 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3274 { 3275 /* { 3276 syscallarg(const char *) path; 3277 syscallarg(u_long) flags; 3278 } */ 3279 struct vnode *vp; 3280 int error; 3281 3282 error = namei_simple_user(SCARG(uap, path), 3283 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3284 if (error != 0) 3285 return (error); 3286 error = change_flags(vp, SCARG(uap, flags), l); 3287 vput(vp); 3288 return (error); 3289 } 3290 3291 /* 3292 * Common routine to change flags of a file. 3293 */ 3294 int 3295 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3296 { 3297 struct vattr vattr; 3298 int error; 3299 3300 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3301 3302 vattr_null(&vattr); 3303 vattr.va_flags = flags; 3304 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3305 3306 return (error); 3307 } 3308 3309 /* 3310 * Change mode of a file given path name; this version follows links. 3311 */ 3312 /* ARGSUSED */ 3313 int 3314 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3315 { 3316 /* { 3317 syscallarg(const char *) path; 3318 syscallarg(int) mode; 3319 } */ 3320 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3321 SCARG(uap, mode), 0); 3322 } 3323 3324 int 3325 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3326 { 3327 int error; 3328 struct vnode *vp; 3329 namei_simple_flags_t ns_flag; 3330 3331 if (flags & AT_SYMLINK_NOFOLLOW) 3332 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3333 else 3334 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3335 3336 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3337 if (error != 0) 3338 return error; 3339 3340 error = change_mode(vp, mode, l); 3341 3342 vrele(vp); 3343 3344 return (error); 3345 } 3346 3347 /* 3348 * Change mode of a file given a file descriptor. 3349 */ 3350 /* ARGSUSED */ 3351 int 3352 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3353 { 3354 /* { 3355 syscallarg(int) fd; 3356 syscallarg(int) mode; 3357 } */ 3358 file_t *fp; 3359 int error; 3360 3361 /* fd_getvnode() will use the descriptor for us */ 3362 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3363 return (error); 3364 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3365 fd_putfile(SCARG(uap, fd)); 3366 return (error); 3367 } 3368 3369 int 3370 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3371 register_t *retval) 3372 { 3373 /* { 3374 syscallarg(int) fd; 3375 syscallarg(const char *) path; 3376 syscallarg(int) mode; 3377 syscallarg(int) flag; 3378 } */ 3379 3380 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3381 SCARG(uap, mode), SCARG(uap, flag)); 3382 } 3383 3384 /* 3385 * Change mode of a file given path name; this version does not follow links. 3386 */ 3387 /* ARGSUSED */ 3388 int 3389 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3390 { 3391 /* { 3392 syscallarg(const char *) path; 3393 syscallarg(int) mode; 3394 } */ 3395 int error; 3396 struct vnode *vp; 3397 3398 error = namei_simple_user(SCARG(uap, path), 3399 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3400 if (error != 0) 3401 return (error); 3402 3403 error = change_mode(vp, SCARG(uap, mode), l); 3404 3405 vrele(vp); 3406 return (error); 3407 } 3408 3409 /* 3410 * Common routine to set mode given a vnode. 3411 */ 3412 static int 3413 change_mode(struct vnode *vp, int mode, struct lwp *l) 3414 { 3415 struct vattr vattr; 3416 int error; 3417 3418 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3419 vattr_null(&vattr); 3420 vattr.va_mode = mode & ALLPERMS; 3421 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3422 VOP_UNLOCK(vp); 3423 return (error); 3424 } 3425 3426 /* 3427 * Set ownership given a path name; this version follows links. 3428 */ 3429 /* ARGSUSED */ 3430 int 3431 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3432 { 3433 /* { 3434 syscallarg(const char *) path; 3435 syscallarg(uid_t) uid; 3436 syscallarg(gid_t) gid; 3437 } */ 3438 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3439 SCARG(uap, gid), 0); 3440 } 3441 3442 int 3443 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3444 gid_t gid, int flags) 3445 { 3446 int error; 3447 struct vnode *vp; 3448 namei_simple_flags_t ns_flag; 3449 3450 if (flags & AT_SYMLINK_NOFOLLOW) 3451 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3452 else 3453 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3454 3455 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3456 if (error != 0) 3457 return error; 3458 3459 error = change_owner(vp, uid, gid, l, 0); 3460 3461 vrele(vp); 3462 3463 return (error); 3464 } 3465 3466 /* 3467 * Set ownership given a path name; this version follows links. 3468 * Provides POSIX semantics. 3469 */ 3470 /* ARGSUSED */ 3471 int 3472 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3473 { 3474 /* { 3475 syscallarg(const char *) path; 3476 syscallarg(uid_t) uid; 3477 syscallarg(gid_t) gid; 3478 } */ 3479 int error; 3480 struct vnode *vp; 3481 3482 error = namei_simple_user(SCARG(uap, path), 3483 NSM_FOLLOW_TRYEMULROOT, &vp); 3484 if (error != 0) 3485 return (error); 3486 3487 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3488 3489 vrele(vp); 3490 return (error); 3491 } 3492 3493 /* 3494 * Set ownership given a file descriptor. 3495 */ 3496 /* ARGSUSED */ 3497 int 3498 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3499 { 3500 /* { 3501 syscallarg(int) fd; 3502 syscallarg(uid_t) uid; 3503 syscallarg(gid_t) gid; 3504 } */ 3505 int error; 3506 file_t *fp; 3507 3508 /* fd_getvnode() will use the descriptor for us */ 3509 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3510 return (error); 3511 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3512 l, 0); 3513 fd_putfile(SCARG(uap, fd)); 3514 return (error); 3515 } 3516 3517 int 3518 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3519 register_t *retval) 3520 { 3521 /* { 3522 syscallarg(int) fd; 3523 syscallarg(const char *) path; 3524 syscallarg(uid_t) owner; 3525 syscallarg(gid_t) group; 3526 syscallarg(int) flag; 3527 } */ 3528 3529 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3530 SCARG(uap, owner), SCARG(uap, group), 3531 SCARG(uap, flag)); 3532 } 3533 3534 /* 3535 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3536 */ 3537 /* ARGSUSED */ 3538 int 3539 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3540 { 3541 /* { 3542 syscallarg(int) fd; 3543 syscallarg(uid_t) uid; 3544 syscallarg(gid_t) gid; 3545 } */ 3546 int error; 3547 file_t *fp; 3548 3549 /* fd_getvnode() will use the descriptor for us */ 3550 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3551 return (error); 3552 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3553 l, 1); 3554 fd_putfile(SCARG(uap, fd)); 3555 return (error); 3556 } 3557 3558 /* 3559 * Set ownership given a path name; this version does not follow links. 3560 */ 3561 /* ARGSUSED */ 3562 int 3563 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3564 { 3565 /* { 3566 syscallarg(const char *) path; 3567 syscallarg(uid_t) uid; 3568 syscallarg(gid_t) gid; 3569 } */ 3570 int error; 3571 struct vnode *vp; 3572 3573 error = namei_simple_user(SCARG(uap, path), 3574 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3575 if (error != 0) 3576 return (error); 3577 3578 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3579 3580 vrele(vp); 3581 return (error); 3582 } 3583 3584 /* 3585 * Set ownership given a path name; this version does not follow links. 3586 * Provides POSIX/XPG semantics. 3587 */ 3588 /* ARGSUSED */ 3589 int 3590 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3591 { 3592 /* { 3593 syscallarg(const char *) path; 3594 syscallarg(uid_t) uid; 3595 syscallarg(gid_t) gid; 3596 } */ 3597 int error; 3598 struct vnode *vp; 3599 3600 error = namei_simple_user(SCARG(uap, path), 3601 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3602 if (error != 0) 3603 return (error); 3604 3605 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3606 3607 vrele(vp); 3608 return (error); 3609 } 3610 3611 /* 3612 * Common routine to set ownership given a vnode. 3613 */ 3614 static int 3615 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3616 int posix_semantics) 3617 { 3618 struct vattr vattr; 3619 mode_t newmode; 3620 int error; 3621 3622 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3623 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3624 goto out; 3625 3626 #define CHANGED(x) ((int)(x) != -1) 3627 newmode = vattr.va_mode; 3628 if (posix_semantics) { 3629 /* 3630 * POSIX/XPG semantics: if the caller is not the super-user, 3631 * clear set-user-id and set-group-id bits. Both POSIX and 3632 * the XPG consider the behaviour for calls by the super-user 3633 * implementation-defined; we leave the set-user-id and set- 3634 * group-id settings intact in that case. 3635 */ 3636 if (vattr.va_mode & S_ISUID) { 3637 if (kauth_authorize_vnode(l->l_cred, 3638 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3639 newmode &= ~S_ISUID; 3640 } 3641 if (vattr.va_mode & S_ISGID) { 3642 if (kauth_authorize_vnode(l->l_cred, 3643 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3644 newmode &= ~S_ISGID; 3645 } 3646 } else { 3647 /* 3648 * NetBSD semantics: when changing owner and/or group, 3649 * clear the respective bit(s). 3650 */ 3651 if (CHANGED(uid)) 3652 newmode &= ~S_ISUID; 3653 if (CHANGED(gid)) 3654 newmode &= ~S_ISGID; 3655 } 3656 /* Update va_mode iff altered. */ 3657 if (vattr.va_mode == newmode) 3658 newmode = VNOVAL; 3659 3660 vattr_null(&vattr); 3661 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3662 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3663 vattr.va_mode = newmode; 3664 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3665 #undef CHANGED 3666 3667 out: 3668 VOP_UNLOCK(vp); 3669 return (error); 3670 } 3671 3672 /* 3673 * Set the access and modification times given a path name; this 3674 * version follows links. 3675 */ 3676 /* ARGSUSED */ 3677 int 3678 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3679 register_t *retval) 3680 { 3681 /* { 3682 syscallarg(const char *) path; 3683 syscallarg(const struct timeval *) tptr; 3684 } */ 3685 3686 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3687 SCARG(uap, tptr), UIO_USERSPACE); 3688 } 3689 3690 /* 3691 * Set the access and modification times given a file descriptor. 3692 */ 3693 /* ARGSUSED */ 3694 int 3695 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3696 register_t *retval) 3697 { 3698 /* { 3699 syscallarg(int) fd; 3700 syscallarg(const struct timeval *) tptr; 3701 } */ 3702 int error; 3703 file_t *fp; 3704 3705 /* fd_getvnode() will use the descriptor for us */ 3706 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3707 return (error); 3708 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3709 UIO_USERSPACE); 3710 fd_putfile(SCARG(uap, fd)); 3711 return (error); 3712 } 3713 3714 int 3715 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3716 register_t *retval) 3717 { 3718 /* { 3719 syscallarg(int) fd; 3720 syscallarg(const struct timespec *) tptr; 3721 } */ 3722 int error; 3723 file_t *fp; 3724 3725 /* fd_getvnode() will use the descriptor for us */ 3726 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3727 return (error); 3728 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3729 SCARG(uap, tptr), UIO_USERSPACE); 3730 fd_putfile(SCARG(uap, fd)); 3731 return (error); 3732 } 3733 3734 /* 3735 * Set the access and modification times given a path name; this 3736 * version does not follow links. 3737 */ 3738 int 3739 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3740 register_t *retval) 3741 { 3742 /* { 3743 syscallarg(const char *) path; 3744 syscallarg(const struct timeval *) tptr; 3745 } */ 3746 3747 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3748 SCARG(uap, tptr), UIO_USERSPACE); 3749 } 3750 3751 int 3752 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3753 register_t *retval) 3754 { 3755 /* { 3756 syscallarg(int) fd; 3757 syscallarg(const char *) path; 3758 syscallarg(const struct timespec *) tptr; 3759 syscallarg(int) flag; 3760 } */ 3761 int follow; 3762 const struct timespec *tptr; 3763 int error; 3764 3765 tptr = SCARG(uap, tptr); 3766 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3767 3768 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3769 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3770 3771 return error; 3772 } 3773 3774 /* 3775 * Common routine to set access and modification times given a vnode. 3776 */ 3777 int 3778 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3779 const struct timespec *tptr, enum uio_seg seg) 3780 { 3781 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3782 } 3783 3784 int 3785 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3786 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3787 { 3788 struct vattr vattr; 3789 int error, dorele = 0; 3790 namei_simple_flags_t sflags; 3791 bool vanull, setbirthtime; 3792 struct timespec ts[2]; 3793 3794 KASSERT(l != NULL || fdat == AT_FDCWD); 3795 3796 /* 3797 * I have checked all callers and they pass either FOLLOW, 3798 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3799 * is 0. More to the point, they don't pass anything else. 3800 * Let's keep it that way at least until the namei interfaces 3801 * are fully sanitized. 3802 */ 3803 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3804 sflags = (flag == FOLLOW) ? 3805 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3806 3807 if (tptr == NULL) { 3808 vanull = true; 3809 nanotime(&ts[0]); 3810 ts[1] = ts[0]; 3811 } else { 3812 vanull = false; 3813 if (seg != UIO_SYSSPACE) { 3814 error = copyin(tptr, ts, sizeof (ts)); 3815 if (error != 0) 3816 return error; 3817 } else { 3818 ts[0] = tptr[0]; 3819 ts[1] = tptr[1]; 3820 } 3821 } 3822 3823 if (ts[0].tv_nsec == UTIME_NOW) { 3824 nanotime(&ts[0]); 3825 if (ts[1].tv_nsec == UTIME_NOW) { 3826 vanull = true; 3827 ts[1] = ts[0]; 3828 } 3829 } else if (ts[1].tv_nsec == UTIME_NOW) 3830 nanotime(&ts[1]); 3831 3832 if (vp == NULL) { 3833 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3834 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3835 if (error != 0) 3836 return error; 3837 dorele = 1; 3838 } 3839 3840 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3841 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3842 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3843 vattr_null(&vattr); 3844 3845 if (ts[0].tv_nsec != UTIME_OMIT) 3846 vattr.va_atime = ts[0]; 3847 3848 if (ts[1].tv_nsec != UTIME_OMIT) { 3849 vattr.va_mtime = ts[1]; 3850 if (setbirthtime) 3851 vattr.va_birthtime = ts[1]; 3852 } 3853 3854 if (vanull) 3855 vattr.va_vaflags |= VA_UTIMES_NULL; 3856 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3857 VOP_UNLOCK(vp); 3858 3859 if (dorele != 0) 3860 vrele(vp); 3861 3862 return error; 3863 } 3864 3865 int 3866 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3867 const struct timeval *tptr, enum uio_seg seg) 3868 { 3869 struct timespec ts[2]; 3870 struct timespec *tsptr = NULL; 3871 int error; 3872 3873 if (tptr != NULL) { 3874 struct timeval tv[2]; 3875 3876 if (seg != UIO_SYSSPACE) { 3877 error = copyin(tptr, tv, sizeof (tv)); 3878 if (error != 0) 3879 return error; 3880 tptr = tv; 3881 } 3882 3883 if ((tv[0].tv_usec == UTIME_NOW) || 3884 (tv[0].tv_usec == UTIME_OMIT)) 3885 ts[0].tv_nsec = tv[0].tv_usec; 3886 else 3887 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3888 3889 if ((tv[1].tv_usec == UTIME_NOW) || 3890 (tv[1].tv_usec == UTIME_OMIT)) 3891 ts[1].tv_nsec = tv[1].tv_usec; 3892 else 3893 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3894 3895 tsptr = &ts[0]; 3896 } 3897 3898 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3899 } 3900 3901 /* 3902 * Truncate a file given its path name. 3903 */ 3904 /* ARGSUSED */ 3905 int 3906 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3907 { 3908 /* { 3909 syscallarg(const char *) path; 3910 syscallarg(int) pad; 3911 syscallarg(off_t) length; 3912 } */ 3913 struct vnode *vp; 3914 struct vattr vattr; 3915 int error; 3916 3917 if (SCARG(uap, length) < 0) 3918 return EINVAL; 3919 3920 error = namei_simple_user(SCARG(uap, path), 3921 NSM_FOLLOW_TRYEMULROOT, &vp); 3922 if (error != 0) 3923 return (error); 3924 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3925 if (vp->v_type == VDIR) 3926 error = EISDIR; 3927 else if ((error = vn_writechk(vp)) == 0 && 3928 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3929 vattr_null(&vattr); 3930 vattr.va_size = SCARG(uap, length); 3931 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3932 } 3933 vput(vp); 3934 return (error); 3935 } 3936 3937 /* 3938 * Truncate a file given a file descriptor. 3939 */ 3940 /* ARGSUSED */ 3941 int 3942 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3943 { 3944 /* { 3945 syscallarg(int) fd; 3946 syscallarg(int) pad; 3947 syscallarg(off_t) length; 3948 } */ 3949 struct vattr vattr; 3950 struct vnode *vp; 3951 file_t *fp; 3952 int error; 3953 3954 if (SCARG(uap, length) < 0) 3955 return EINVAL; 3956 3957 /* fd_getvnode() will use the descriptor for us */ 3958 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3959 return (error); 3960 if ((fp->f_flag & FWRITE) == 0) { 3961 error = EINVAL; 3962 goto out; 3963 } 3964 vp = fp->f_vnode; 3965 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3966 if (vp->v_type == VDIR) 3967 error = EISDIR; 3968 else if ((error = vn_writechk(vp)) == 0) { 3969 vattr_null(&vattr); 3970 vattr.va_size = SCARG(uap, length); 3971 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3972 } 3973 VOP_UNLOCK(vp); 3974 out: 3975 fd_putfile(SCARG(uap, fd)); 3976 return (error); 3977 } 3978 3979 /* 3980 * Sync an open file. 3981 */ 3982 /* ARGSUSED */ 3983 int 3984 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3985 { 3986 /* { 3987 syscallarg(int) fd; 3988 } */ 3989 struct vnode *vp; 3990 file_t *fp; 3991 int error; 3992 3993 /* fd_getvnode() will use the descriptor for us */ 3994 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3995 return (error); 3996 vp = fp->f_vnode; 3997 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3998 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3999 VOP_UNLOCK(vp); 4000 fd_putfile(SCARG(uap, fd)); 4001 return (error); 4002 } 4003 4004 /* 4005 * Sync a range of file data. API modeled after that found in AIX. 4006 * 4007 * FDATASYNC indicates that we need only save enough metadata to be able 4008 * to re-read the written data. Note we duplicate AIX's requirement that 4009 * the file be open for writing. 4010 */ 4011 /* ARGSUSED */ 4012 int 4013 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4014 { 4015 /* { 4016 syscallarg(int) fd; 4017 syscallarg(int) flags; 4018 syscallarg(off_t) start; 4019 syscallarg(off_t) length; 4020 } */ 4021 struct vnode *vp; 4022 file_t *fp; 4023 int flags, nflags; 4024 off_t s, e, len; 4025 int error; 4026 4027 /* fd_getvnode() will use the descriptor for us */ 4028 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4029 return (error); 4030 4031 if ((fp->f_flag & FWRITE) == 0) { 4032 error = EBADF; 4033 goto out; 4034 } 4035 4036 flags = SCARG(uap, flags); 4037 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4038 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4039 error = EINVAL; 4040 goto out; 4041 } 4042 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4043 if (flags & FDATASYNC) 4044 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4045 else 4046 nflags = FSYNC_WAIT; 4047 if (flags & FDISKSYNC) 4048 nflags |= FSYNC_CACHE; 4049 4050 len = SCARG(uap, length); 4051 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4052 if (len) { 4053 s = SCARG(uap, start); 4054 e = s + len; 4055 if (e < s) { 4056 error = EINVAL; 4057 goto out; 4058 } 4059 } else { 4060 e = 0; 4061 s = 0; 4062 } 4063 4064 vp = fp->f_vnode; 4065 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4066 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4067 VOP_UNLOCK(vp); 4068 out: 4069 fd_putfile(SCARG(uap, fd)); 4070 return (error); 4071 } 4072 4073 /* 4074 * Sync the data of an open file. 4075 */ 4076 /* ARGSUSED */ 4077 int 4078 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4079 { 4080 /* { 4081 syscallarg(int) fd; 4082 } */ 4083 struct vnode *vp; 4084 file_t *fp; 4085 int error; 4086 4087 /* fd_getvnode() will use the descriptor for us */ 4088 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4089 return (error); 4090 if ((fp->f_flag & FWRITE) == 0) { 4091 fd_putfile(SCARG(uap, fd)); 4092 return (EBADF); 4093 } 4094 vp = fp->f_vnode; 4095 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4096 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4097 VOP_UNLOCK(vp); 4098 fd_putfile(SCARG(uap, fd)); 4099 return (error); 4100 } 4101 4102 /* 4103 * Rename files, (standard) BSD semantics frontend. 4104 */ 4105 /* ARGSUSED */ 4106 int 4107 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4108 { 4109 /* { 4110 syscallarg(const char *) from; 4111 syscallarg(const char *) to; 4112 } */ 4113 4114 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4115 SCARG(uap, to), UIO_USERSPACE, 0)); 4116 } 4117 4118 int 4119 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4120 register_t *retval) 4121 { 4122 /* { 4123 syscallarg(int) fromfd; 4124 syscallarg(const char *) from; 4125 syscallarg(int) tofd; 4126 syscallarg(const char *) to; 4127 } */ 4128 4129 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4130 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4131 } 4132 4133 /* 4134 * Rename files, POSIX semantics frontend. 4135 */ 4136 /* ARGSUSED */ 4137 int 4138 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4139 { 4140 /* { 4141 syscallarg(const char *) from; 4142 syscallarg(const char *) to; 4143 } */ 4144 4145 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4146 SCARG(uap, to), UIO_USERSPACE, 1)); 4147 } 4148 4149 /* 4150 * Rename files. Source and destination must either both be directories, 4151 * or both not be directories. If target is a directory, it must be empty. 4152 * If `from' and `to' refer to the same object, the value of the `retain' 4153 * argument is used to determine whether `from' will be 4154 * 4155 * (retain == 0) deleted unless `from' and `to' refer to the same 4156 * object in the file system's name space (BSD). 4157 * (retain == 1) always retained (POSIX). 4158 * 4159 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4160 */ 4161 int 4162 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4163 { 4164 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4165 } 4166 4167 static int 4168 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4169 const char *to, enum uio_seg seg, int retain) 4170 { 4171 struct pathbuf *fpb, *tpb; 4172 struct nameidata fnd, tnd; 4173 struct vnode *fdvp, *fvp; 4174 struct vnode *tdvp, *tvp; 4175 struct mount *mp, *tmp; 4176 int error; 4177 4178 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4179 4180 error = pathbuf_maybe_copyin(from, seg, &fpb); 4181 if (error) 4182 goto out0; 4183 KASSERT(fpb != NULL); 4184 4185 error = pathbuf_maybe_copyin(to, seg, &tpb); 4186 if (error) 4187 goto out1; 4188 KASSERT(tpb != NULL); 4189 4190 /* 4191 * Lookup from. 4192 * 4193 * XXX LOCKPARENT is wrong because we don't actually want it 4194 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4195 * insane, so for the time being we need to leave it like this. 4196 */ 4197 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4198 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4199 goto out2; 4200 4201 /* 4202 * Pull out the important results of the lookup, fdvp and fvp. 4203 * Of course, fvp is bogus because we're about to unlock fdvp. 4204 */ 4205 fdvp = fnd.ni_dvp; 4206 fvp = fnd.ni_vp; 4207 KASSERT(fdvp != NULL); 4208 KASSERT(fvp != NULL); 4209 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4210 4211 /* 4212 * Make sure neither fdvp nor fvp is locked. 4213 */ 4214 if (fdvp != fvp) 4215 VOP_UNLOCK(fdvp); 4216 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4217 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4218 4219 /* 4220 * Reject renaming `.' and `..'. Can't do this until after 4221 * namei because we need namei's parsing to find the final 4222 * component name. (namei should just leave us with the final 4223 * component name and not look it up itself, but anyway...) 4224 * 4225 * This was here before because we used to relookup from 4226 * instead of to and relookup requires the caller to check 4227 * this, but now file systems may depend on this check, so we 4228 * must retain it until the file systems are all rototilled. 4229 */ 4230 if (((fnd.ni_cnd.cn_namelen == 1) && 4231 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4232 ((fnd.ni_cnd.cn_namelen == 2) && 4233 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4234 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4235 error = EINVAL; /* XXX EISDIR? */ 4236 goto abort0; 4237 } 4238 4239 /* 4240 * Lookup to. 4241 * 4242 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4243 * fvp here to decide whether to add CREATEDIR is a load of 4244 * bollocks because fvp might be the wrong node by now, since 4245 * fdvp is unlocked. 4246 * 4247 * XXX Why not pass CREATEDIR always? 4248 */ 4249 NDINIT(&tnd, RENAME, 4250 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4251 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4252 tpb); 4253 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4254 goto abort0; 4255 4256 /* 4257 * Pull out the important results of the lookup, tdvp and tvp. 4258 * Of course, tvp is bogus because we're about to unlock tdvp. 4259 */ 4260 tdvp = tnd.ni_dvp; 4261 tvp = tnd.ni_vp; 4262 KASSERT(tdvp != NULL); 4263 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4264 4265 /* 4266 * Make sure neither tdvp nor tvp is locked. 4267 */ 4268 if (tdvp != tvp) 4269 VOP_UNLOCK(tdvp); 4270 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4271 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4272 4273 /* 4274 * Reject renaming onto `.' or `..'. relookup is unhappy with 4275 * these, which is why we must do this here. Once upon a time 4276 * we relooked up from instead of to, and consequently didn't 4277 * need this check, but now that we relookup to instead of 4278 * from, we need this; and we shall need it forever forward 4279 * until the VOP_RENAME protocol changes, because file systems 4280 * will no doubt begin to depend on this check. 4281 */ 4282 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4283 error = EISDIR; 4284 goto abort1; 4285 } 4286 if ((tnd.ni_cnd.cn_namelen == 2) && 4287 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4288 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4289 error = EINVAL; 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 EXDEV 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 4694 /* 4695 * Allocate backing store for a file, filling a hole without having to 4696 * explicitly write anything out. 4697 */ 4698 /* ARGSUSED */ 4699 int 4700 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4701 register_t *retval) 4702 { 4703 /* { 4704 syscallarg(int) fd; 4705 syscallarg(off_t) pos; 4706 syscallarg(off_t) len; 4707 } */ 4708 int fd; 4709 off_t pos, len; 4710 struct file *fp; 4711 struct vnode *vp; 4712 int error; 4713 4714 fd = SCARG(uap, fd); 4715 pos = SCARG(uap, pos); 4716 len = SCARG(uap, len); 4717 4718 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4719 *retval = EINVAL; 4720 return 0; 4721 } 4722 4723 error = fd_getvnode(fd, &fp); 4724 if (error) { 4725 *retval = error; 4726 return 0; 4727 } 4728 if ((fp->f_flag & FWRITE) == 0) { 4729 error = EBADF; 4730 goto fail; 4731 } 4732 vp = fp->f_vnode; 4733 4734 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4735 if (vp->v_type == VDIR) { 4736 error = EISDIR; 4737 } else { 4738 error = VOP_FALLOCATE(vp, pos, len); 4739 } 4740 VOP_UNLOCK(vp); 4741 4742 fail: 4743 fd_putfile(fd); 4744 *retval = error; 4745 return 0; 4746 } 4747 4748 /* 4749 * Deallocate backing store for a file, creating a hole. Also used for 4750 * invoking TRIM on disks. 4751 */ 4752 /* ARGSUSED */ 4753 int 4754 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4755 register_t *retval) 4756 { 4757 /* { 4758 syscallarg(int) fd; 4759 syscallarg(off_t) pos; 4760 syscallarg(off_t) len; 4761 } */ 4762 int fd; 4763 off_t pos, len; 4764 struct file *fp; 4765 struct vnode *vp; 4766 int error; 4767 4768 fd = SCARG(uap, fd); 4769 pos = SCARG(uap, pos); 4770 len = SCARG(uap, len); 4771 4772 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4773 return EINVAL; 4774 } 4775 4776 error = fd_getvnode(fd, &fp); 4777 if (error) { 4778 return error; 4779 } 4780 if ((fp->f_flag & FWRITE) == 0) { 4781 error = EBADF; 4782 goto fail; 4783 } 4784 vp = fp->f_vnode; 4785 4786 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4787 if (vp->v_type == VDIR) { 4788 error = EISDIR; 4789 } else { 4790 error = VOP_FDISCARD(vp, pos, len); 4791 } 4792 VOP_UNLOCK(vp); 4793 4794 fail: 4795 fd_putfile(fd); 4796 return error; 4797 } 4798
Indexes created Sun Sep 28 16:09:52 GMT 2025