vfs_syscalls.c revision 1.522
1 /* $NetBSD: vfs_syscalls.c,v 1.522 2019/02/05 08:53:18 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 1.522 2019/02/05 08:53:18 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 const char *pathstring = NULL; 1635 int error; 1636 1637 if (path == NULL) { 1638 MODULE_CALL_HOOK(vfs_openat_10_hook, (&pb), enosys(), error); 1639 if (error != ENOSYS) 1640 return error; 1641 } 1642 error = pathbuf_copyin(path, &pb); 1643 if (error) 1644 return error; 1645 1646 pathstring = pathbuf_stringcopy_get(pb); 1647 1648 /* 1649 * fdat is ignored if: 1650 * 1) if fdat is AT_FDCWD, which means use current directory as base. 1651 * 2) if path is absolute, then fdat is useless. 1652 */ 1653 if (fdat != AT_FDCWD && pathstring[0] != '/') { 1654 /* fd_getvnode() will use the descriptor for us */ 1655 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1656 goto out; 1657 1658 dvp = dfp->f_vnode; 1659 } 1660 1661 error = do_open(l, dvp, pb, flags, mode, fd); 1662 1663 if (dfp != NULL) 1664 fd_putfile(fdat); 1665 out: 1666 pathbuf_stringcopy_put(pb, pathstring); 1667 pathbuf_destroy(pb); 1668 return error; 1669 } 1670 1671 int 1672 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1673 { 1674 /* { 1675 syscallarg(const char *) path; 1676 syscallarg(int) flags; 1677 syscallarg(int) mode; 1678 } */ 1679 int error; 1680 int fd; 1681 1682 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1683 SCARG(uap, flags), SCARG(uap, mode), &fd); 1684 1685 if (error == 0) 1686 *retval = fd; 1687 1688 return error; 1689 } 1690 1691 int 1692 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1693 { 1694 /* { 1695 syscallarg(int) fd; 1696 syscallarg(const char *) path; 1697 syscallarg(int) oflags; 1698 syscallarg(int) mode; 1699 } */ 1700 int error; 1701 int fd; 1702 1703 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1704 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1705 1706 if (error == 0) 1707 *retval = fd; 1708 1709 return error; 1710 } 1711 1712 static void 1713 vfs__fhfree(fhandle_t *fhp) 1714 { 1715 size_t fhsize; 1716 1717 fhsize = FHANDLE_SIZE(fhp); 1718 kmem_free(fhp, fhsize); 1719 } 1720 1721 /* 1722 * vfs_composefh: compose a filehandle. 1723 */ 1724 1725 int 1726 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1727 { 1728 struct mount *mp; 1729 struct fid *fidp; 1730 int error; 1731 size_t needfhsize; 1732 size_t fidsize; 1733 1734 mp = vp->v_mount; 1735 fidp = NULL; 1736 if (*fh_size < FHANDLE_SIZE_MIN) { 1737 fidsize = 0; 1738 } else { 1739 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1740 if (fhp != NULL) { 1741 memset(fhp, 0, *fh_size); 1742 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1743 fidp = &fhp->fh_fid; 1744 } 1745 } 1746 error = VFS_VPTOFH(vp, fidp, &fidsize); 1747 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1748 if (error == 0 && *fh_size < needfhsize) { 1749 error = E2BIG; 1750 } 1751 *fh_size = needfhsize; 1752 return error; 1753 } 1754 1755 int 1756 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1757 { 1758 struct mount *mp; 1759 fhandle_t *fhp; 1760 size_t fhsize; 1761 size_t fidsize; 1762 int error; 1763 1764 mp = vp->v_mount; 1765 fidsize = 0; 1766 error = VFS_VPTOFH(vp, NULL, &fidsize); 1767 KASSERT(error != 0); 1768 if (error != E2BIG) { 1769 goto out; 1770 } 1771 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1772 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1773 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1774 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1775 if (error == 0) { 1776 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1777 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1778 *fhpp = fhp; 1779 } else { 1780 kmem_free(fhp, fhsize); 1781 } 1782 out: 1783 return error; 1784 } 1785 1786 void 1787 vfs_composefh_free(fhandle_t *fhp) 1788 { 1789 1790 vfs__fhfree(fhp); 1791 } 1792 1793 /* 1794 * vfs_fhtovp: lookup a vnode by a filehandle. 1795 */ 1796 1797 int 1798 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1799 { 1800 struct mount *mp; 1801 int error; 1802 1803 *vpp = NULL; 1804 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1805 if (mp == NULL) { 1806 error = ESTALE; 1807 goto out; 1808 } 1809 if (mp->mnt_op->vfs_fhtovp == NULL) { 1810 error = EOPNOTSUPP; 1811 goto out; 1812 } 1813 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1814 out: 1815 return error; 1816 } 1817 1818 /* 1819 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1820 * the needed size. 1821 */ 1822 1823 int 1824 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1825 { 1826 fhandle_t *fhp; 1827 int error; 1828 1829 if (fhsize > FHANDLE_SIZE_MAX) { 1830 return EINVAL; 1831 } 1832 if (fhsize < FHANDLE_SIZE_MIN) { 1833 return EINVAL; 1834 } 1835 again: 1836 fhp = kmem_alloc(fhsize, KM_SLEEP); 1837 error = copyin(ufhp, fhp, fhsize); 1838 if (error == 0) { 1839 /* XXX this check shouldn't be here */ 1840 if (FHANDLE_SIZE(fhp) == fhsize) { 1841 *fhpp = fhp; 1842 return 0; 1843 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1844 /* 1845 * a kludge for nfsv2 padded handles. 1846 */ 1847 size_t sz; 1848 1849 sz = FHANDLE_SIZE(fhp); 1850 kmem_free(fhp, fhsize); 1851 fhsize = sz; 1852 goto again; 1853 } else { 1854 /* 1855 * userland told us wrong size. 1856 */ 1857 error = EINVAL; 1858 } 1859 } 1860 kmem_free(fhp, fhsize); 1861 return error; 1862 } 1863 1864 void 1865 vfs_copyinfh_free(fhandle_t *fhp) 1866 { 1867 1868 vfs__fhfree(fhp); 1869 } 1870 1871 /* 1872 * Get file handle system call 1873 */ 1874 int 1875 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1876 { 1877 /* { 1878 syscallarg(char *) fname; 1879 syscallarg(fhandle_t *) fhp; 1880 syscallarg(size_t *) fh_size; 1881 } */ 1882 struct vnode *vp; 1883 fhandle_t *fh; 1884 int error; 1885 struct pathbuf *pb; 1886 struct nameidata nd; 1887 size_t sz; 1888 size_t usz; 1889 1890 /* 1891 * Must be super user 1892 */ 1893 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1894 0, NULL, NULL, NULL); 1895 if (error) 1896 return (error); 1897 1898 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1899 if (error) { 1900 return error; 1901 } 1902 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1903 error = namei(&nd); 1904 if (error) { 1905 pathbuf_destroy(pb); 1906 return error; 1907 } 1908 vp = nd.ni_vp; 1909 pathbuf_destroy(pb); 1910 1911 error = vfs_composefh_alloc(vp, &fh); 1912 vput(vp); 1913 if (error != 0) { 1914 return error; 1915 } 1916 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1917 if (error != 0) { 1918 goto out; 1919 } 1920 sz = FHANDLE_SIZE(fh); 1921 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1922 if (error != 0) { 1923 goto out; 1924 } 1925 if (usz >= sz) { 1926 error = copyout(fh, SCARG(uap, fhp), sz); 1927 } else { 1928 error = E2BIG; 1929 } 1930 out: 1931 vfs_composefh_free(fh); 1932 return (error); 1933 } 1934 1935 /* 1936 * Open a file given a file handle. 1937 * 1938 * Check permissions, allocate an open file structure, 1939 * and call the device open routine if any. 1940 */ 1941 1942 int 1943 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1944 register_t *retval) 1945 { 1946 file_t *fp; 1947 struct vnode *vp = NULL; 1948 kauth_cred_t cred = l->l_cred; 1949 file_t *nfp; 1950 int indx, error; 1951 struct vattr va; 1952 fhandle_t *fh; 1953 int flags; 1954 proc_t *p; 1955 1956 p = curproc; 1957 1958 /* 1959 * Must be super user 1960 */ 1961 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1962 0, NULL, NULL, NULL))) 1963 return (error); 1964 1965 if (oflags & O_SEARCH) { 1966 oflags &= ~(int)O_SEARCH; 1967 } 1968 1969 flags = FFLAGS(oflags); 1970 if ((flags & (FREAD | FWRITE)) == 0) 1971 return (EINVAL); 1972 if ((flags & O_CREAT)) 1973 return (EINVAL); 1974 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1975 return (error); 1976 fp = nfp; 1977 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1978 if (error != 0) { 1979 goto bad; 1980 } 1981 error = vfs_fhtovp(fh, &vp); 1982 vfs_copyinfh_free(fh); 1983 if (error != 0) { 1984 goto bad; 1985 } 1986 1987 /* Now do an effective vn_open */ 1988 1989 if (vp->v_type == VSOCK) { 1990 error = EOPNOTSUPP; 1991 goto bad; 1992 } 1993 error = vn_openchk(vp, cred, flags); 1994 if (error != 0) 1995 goto bad; 1996 if (flags & O_TRUNC) { 1997 VOP_UNLOCK(vp); /* XXX */ 1998 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1999 vattr_null(&va); 2000 va.va_size = 0; 2001 error = VOP_SETATTR(vp, &va, cred); 2002 if (error) 2003 goto bad; 2004 } 2005 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2006 goto bad; 2007 if (flags & FWRITE) { 2008 mutex_enter(vp->v_interlock); 2009 vp->v_writecount++; 2010 mutex_exit(vp->v_interlock); 2011 } 2012 2013 /* done with modified vn_open, now finish what sys_open does. */ 2014 if ((error = open_setfp(l, fp, vp, indx, flags))) 2015 return error; 2016 2017 VOP_UNLOCK(vp); 2018 *retval = indx; 2019 fd_affix(p, fp, indx); 2020 return (0); 2021 2022 bad: 2023 fd_abort(p, fp, indx); 2024 if (vp != NULL) 2025 vput(vp); 2026 return (error); 2027 } 2028 2029 int 2030 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2031 { 2032 /* { 2033 syscallarg(const void *) fhp; 2034 syscallarg(size_t) fh_size; 2035 syscallarg(int) flags; 2036 } */ 2037 2038 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2039 SCARG(uap, flags), retval); 2040 } 2041 2042 int 2043 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2044 { 2045 int error; 2046 fhandle_t *fh; 2047 struct vnode *vp; 2048 2049 /* 2050 * Must be super user 2051 */ 2052 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2053 0, NULL, NULL, NULL))) 2054 return (error); 2055 2056 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2057 if (error != 0) 2058 return error; 2059 2060 error = vfs_fhtovp(fh, &vp); 2061 vfs_copyinfh_free(fh); 2062 if (error != 0) 2063 return error; 2064 2065 error = vn_stat(vp, sb); 2066 vput(vp); 2067 return error; 2068 } 2069 2070 2071 /* ARGSUSED */ 2072 int 2073 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2074 { 2075 /* { 2076 syscallarg(const void *) fhp; 2077 syscallarg(size_t) fh_size; 2078 syscallarg(struct stat *) sb; 2079 } */ 2080 struct stat sb; 2081 int error; 2082 2083 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2084 if (error) 2085 return error; 2086 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2087 } 2088 2089 int 2090 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2091 int flags) 2092 { 2093 fhandle_t *fh; 2094 struct mount *mp; 2095 struct vnode *vp; 2096 int error; 2097 2098 /* 2099 * Must be super user 2100 */ 2101 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2102 0, NULL, NULL, NULL))) 2103 return error; 2104 2105 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2106 if (error != 0) 2107 return error; 2108 2109 error = vfs_fhtovp(fh, &vp); 2110 vfs_copyinfh_free(fh); 2111 if (error != 0) 2112 return error; 2113 2114 mp = vp->v_mount; 2115 error = dostatvfs(mp, sb, l, flags, 1); 2116 vput(vp); 2117 return error; 2118 } 2119 2120 /* ARGSUSED */ 2121 int 2122 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2123 { 2124 /* { 2125 syscallarg(const void *) fhp; 2126 syscallarg(size_t) fh_size; 2127 syscallarg(struct statvfs *) buf; 2128 syscallarg(int) flags; 2129 } */ 2130 struct statvfs *sb = STATVFSBUF_GET(); 2131 int error; 2132 2133 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2134 SCARG(uap, flags)); 2135 if (error == 0) 2136 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2137 STATVFSBUF_PUT(sb); 2138 return error; 2139 } 2140 2141 /* 2142 * Create a special file. 2143 */ 2144 /* ARGSUSED */ 2145 int 2146 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2147 register_t *retval) 2148 { 2149 /* { 2150 syscallarg(const char *) path; 2151 syscallarg(mode_t) mode; 2152 syscallarg(dev_t) dev; 2153 } */ 2154 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2155 SCARG(uap, dev), retval, UIO_USERSPACE); 2156 } 2157 2158 int 2159 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2160 register_t *retval) 2161 { 2162 /* { 2163 syscallarg(int) fd; 2164 syscallarg(const char *) path; 2165 syscallarg(mode_t) mode; 2166 syscallarg(int) pad; 2167 syscallarg(dev_t) dev; 2168 } */ 2169 2170 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2171 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2172 } 2173 2174 int 2175 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2176 register_t *retval, enum uio_seg seg) 2177 { 2178 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2179 } 2180 2181 int 2182 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2183 dev_t dev, register_t *retval, enum uio_seg seg) 2184 { 2185 struct proc *p = l->l_proc; 2186 struct vnode *vp; 2187 struct vattr vattr; 2188 int error, optype; 2189 struct pathbuf *pb; 2190 struct nameidata nd; 2191 const char *pathstring; 2192 2193 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2194 0, NULL, NULL, NULL)) != 0) 2195 return (error); 2196 2197 optype = VOP_MKNOD_DESCOFFSET; 2198 2199 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2200 if (error) { 2201 return error; 2202 } 2203 pathstring = pathbuf_stringcopy_get(pb); 2204 if (pathstring == NULL) { 2205 pathbuf_destroy(pb); 2206 return ENOMEM; 2207 } 2208 2209 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2210 2211 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2212 goto out; 2213 vp = nd.ni_vp; 2214 2215 if (vp != NULL) 2216 error = EEXIST; 2217 else { 2218 vattr_null(&vattr); 2219 /* We will read cwdi->cwdi_cmask unlocked. */ 2220 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2221 vattr.va_rdev = dev; 2222 2223 switch (mode & S_IFMT) { 2224 case S_IFMT: /* used by badsect to flag bad sectors */ 2225 vattr.va_type = VBAD; 2226 break; 2227 case S_IFCHR: 2228 vattr.va_type = VCHR; 2229 break; 2230 case S_IFBLK: 2231 vattr.va_type = VBLK; 2232 break; 2233 case S_IFWHT: 2234 optype = VOP_WHITEOUT_DESCOFFSET; 2235 break; 2236 case S_IFREG: 2237 #if NVERIEXEC > 0 2238 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2239 O_CREAT); 2240 #endif /* NVERIEXEC > 0 */ 2241 vattr.va_type = VREG; 2242 vattr.va_rdev = VNOVAL; 2243 optype = VOP_CREATE_DESCOFFSET; 2244 break; 2245 default: 2246 error = EINVAL; 2247 break; 2248 } 2249 } 2250 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2251 && vattr.va_rdev == VNOVAL) 2252 error = EINVAL; 2253 if (!error) { 2254 switch (optype) { 2255 case VOP_WHITEOUT_DESCOFFSET: 2256 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2257 if (error) 2258 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2259 vput(nd.ni_dvp); 2260 break; 2261 2262 case VOP_MKNOD_DESCOFFSET: 2263 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2264 &nd.ni_cnd, &vattr); 2265 if (error == 0) 2266 vrele(nd.ni_vp); 2267 vput(nd.ni_dvp); 2268 break; 2269 2270 case VOP_CREATE_DESCOFFSET: 2271 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2272 &nd.ni_cnd, &vattr); 2273 if (error == 0) 2274 vrele(nd.ni_vp); 2275 vput(nd.ni_dvp); 2276 break; 2277 } 2278 } else { 2279 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2280 if (nd.ni_dvp == vp) 2281 vrele(nd.ni_dvp); 2282 else 2283 vput(nd.ni_dvp); 2284 if (vp) 2285 vrele(vp); 2286 } 2287 out: 2288 pathbuf_stringcopy_put(pb, pathstring); 2289 pathbuf_destroy(pb); 2290 return (error); 2291 } 2292 2293 /* 2294 * Create a named pipe. 2295 */ 2296 /* ARGSUSED */ 2297 int 2298 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2299 { 2300 /* { 2301 syscallarg(const char *) path; 2302 syscallarg(int) mode; 2303 } */ 2304 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2305 } 2306 2307 int 2308 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2309 register_t *retval) 2310 { 2311 /* { 2312 syscallarg(int) fd; 2313 syscallarg(const char *) path; 2314 syscallarg(int) mode; 2315 } */ 2316 2317 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2318 SCARG(uap, mode)); 2319 } 2320 2321 static int 2322 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2323 { 2324 struct proc *p = l->l_proc; 2325 struct vattr vattr; 2326 int error; 2327 struct pathbuf *pb; 2328 struct nameidata nd; 2329 2330 error = pathbuf_copyin(path, &pb); 2331 if (error) { 2332 return error; 2333 } 2334 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2335 2336 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2337 pathbuf_destroy(pb); 2338 return error; 2339 } 2340 if (nd.ni_vp != NULL) { 2341 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2342 if (nd.ni_dvp == nd.ni_vp) 2343 vrele(nd.ni_dvp); 2344 else 2345 vput(nd.ni_dvp); 2346 vrele(nd.ni_vp); 2347 pathbuf_destroy(pb); 2348 return (EEXIST); 2349 } 2350 vattr_null(&vattr); 2351 vattr.va_type = VFIFO; 2352 /* We will read cwdi->cwdi_cmask unlocked. */ 2353 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2354 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2355 if (error == 0) 2356 vrele(nd.ni_vp); 2357 vput(nd.ni_dvp); 2358 pathbuf_destroy(pb); 2359 return (error); 2360 } 2361 2362 /* 2363 * Make a hard file link. 2364 */ 2365 /* ARGSUSED */ 2366 int 2367 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2368 const char *link, int follow, register_t *retval) 2369 { 2370 struct vnode *vp; 2371 struct pathbuf *linkpb; 2372 struct nameidata nd; 2373 namei_simple_flags_t ns_flags; 2374 int error; 2375 2376 if (follow & AT_SYMLINK_FOLLOW) 2377 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2378 else 2379 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2380 2381 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2382 if (error != 0) 2383 return (error); 2384 error = pathbuf_copyin(link, &linkpb); 2385 if (error) { 2386 goto out1; 2387 } 2388 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2389 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2390 goto out2; 2391 if (nd.ni_vp) { 2392 error = EEXIST; 2393 goto abortop; 2394 } 2395 /* Prevent hard links on directories. */ 2396 if (vp->v_type == VDIR) { 2397 error = EPERM; 2398 goto abortop; 2399 } 2400 /* Prevent cross-mount operation. */ 2401 if (nd.ni_dvp->v_mount != vp->v_mount) { 2402 error = EXDEV; 2403 goto abortop; 2404 } 2405 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2406 VOP_UNLOCK(nd.ni_dvp); 2407 vrele(nd.ni_dvp); 2408 out2: 2409 pathbuf_destroy(linkpb); 2410 out1: 2411 vrele(vp); 2412 return (error); 2413 abortop: 2414 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2415 if (nd.ni_dvp == nd.ni_vp) 2416 vrele(nd.ni_dvp); 2417 else 2418 vput(nd.ni_dvp); 2419 if (nd.ni_vp != NULL) 2420 vrele(nd.ni_vp); 2421 goto out2; 2422 } 2423 2424 int 2425 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2426 { 2427 /* { 2428 syscallarg(const char *) path; 2429 syscallarg(const char *) link; 2430 } */ 2431 const char *path = SCARG(uap, path); 2432 const char *link = SCARG(uap, link); 2433 2434 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2435 AT_SYMLINK_FOLLOW, retval); 2436 } 2437 2438 int 2439 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2440 register_t *retval) 2441 { 2442 /* { 2443 syscallarg(int) fd1; 2444 syscallarg(const char *) name1; 2445 syscallarg(int) fd2; 2446 syscallarg(const char *) name2; 2447 syscallarg(int) flags; 2448 } */ 2449 int fd1 = SCARG(uap, fd1); 2450 const char *name1 = SCARG(uap, name1); 2451 int fd2 = SCARG(uap, fd2); 2452 const char *name2 = SCARG(uap, name2); 2453 int follow; 2454 2455 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2456 2457 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2458 } 2459 2460 2461 int 2462 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2463 { 2464 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2465 } 2466 2467 static int 2468 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2469 const char *link, enum uio_seg seg) 2470 { 2471 struct proc *p = curproc; 2472 struct vattr vattr; 2473 char *path; 2474 int error; 2475 size_t len; 2476 struct pathbuf *linkpb; 2477 struct nameidata nd; 2478 2479 KASSERT(l != NULL || fdat == AT_FDCWD); 2480 2481 path = PNBUF_GET(); 2482 if (seg == UIO_USERSPACE) { 2483 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2484 goto out1; 2485 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2486 goto out1; 2487 } else { 2488 len = strlen(patharg) + 1; 2489 KASSERT(len <= MAXPATHLEN); 2490 memcpy(path, patharg, len); 2491 linkpb = pathbuf_create(link); 2492 if (linkpb == NULL) { 2493 error = ENOMEM; 2494 goto out1; 2495 } 2496 } 2497 ktrkuser("symlink-target", path, len - 1); 2498 2499 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2500 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2501 goto out2; 2502 if (nd.ni_vp) { 2503 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2504 if (nd.ni_dvp == nd.ni_vp) 2505 vrele(nd.ni_dvp); 2506 else 2507 vput(nd.ni_dvp); 2508 vrele(nd.ni_vp); 2509 error = EEXIST; 2510 goto out2; 2511 } 2512 vattr_null(&vattr); 2513 vattr.va_type = VLNK; 2514 /* We will read cwdi->cwdi_cmask unlocked. */ 2515 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2516 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2517 if (error == 0) 2518 vrele(nd.ni_vp); 2519 vput(nd.ni_dvp); 2520 out2: 2521 pathbuf_destroy(linkpb); 2522 out1: 2523 PNBUF_PUT(path); 2524 return (error); 2525 } 2526 2527 /* 2528 * Make a symbolic link. 2529 */ 2530 /* ARGSUSED */ 2531 int 2532 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2533 { 2534 /* { 2535 syscallarg(const char *) path; 2536 syscallarg(const char *) link; 2537 } */ 2538 2539 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2540 UIO_USERSPACE); 2541 } 2542 2543 int 2544 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2545 register_t *retval) 2546 { 2547 /* { 2548 syscallarg(const char *) path1; 2549 syscallarg(int) fd; 2550 syscallarg(const char *) path2; 2551 } */ 2552 2553 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2554 SCARG(uap, path2), UIO_USERSPACE); 2555 } 2556 2557 /* 2558 * Delete a whiteout from the filesystem. 2559 */ 2560 /* ARGSUSED */ 2561 int 2562 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2563 { 2564 /* { 2565 syscallarg(const char *) path; 2566 } */ 2567 int error; 2568 struct pathbuf *pb; 2569 struct nameidata nd; 2570 2571 error = pathbuf_copyin(SCARG(uap, path), &pb); 2572 if (error) { 2573 return error; 2574 } 2575 2576 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2577 error = namei(&nd); 2578 if (error) { 2579 pathbuf_destroy(pb); 2580 return (error); 2581 } 2582 2583 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2584 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2585 if (nd.ni_dvp == nd.ni_vp) 2586 vrele(nd.ni_dvp); 2587 else 2588 vput(nd.ni_dvp); 2589 if (nd.ni_vp) 2590 vrele(nd.ni_vp); 2591 pathbuf_destroy(pb); 2592 return (EEXIST); 2593 } 2594 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2595 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2596 vput(nd.ni_dvp); 2597 pathbuf_destroy(pb); 2598 return (error); 2599 } 2600 2601 /* 2602 * Delete a name from the filesystem. 2603 */ 2604 /* ARGSUSED */ 2605 int 2606 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2607 { 2608 /* { 2609 syscallarg(const char *) path; 2610 } */ 2611 2612 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2613 } 2614 2615 int 2616 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2617 register_t *retval) 2618 { 2619 /* { 2620 syscallarg(int) fd; 2621 syscallarg(const char *) path; 2622 syscallarg(int) flag; 2623 } */ 2624 2625 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2626 SCARG(uap, flag), UIO_USERSPACE); 2627 } 2628 2629 int 2630 do_sys_unlink(const char *arg, enum uio_seg seg) 2631 { 2632 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2633 } 2634 2635 static int 2636 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2637 enum uio_seg seg) 2638 { 2639 struct vnode *vp; 2640 int error; 2641 struct pathbuf *pb; 2642 struct nameidata nd; 2643 const char *pathstring; 2644 2645 KASSERT(l != NULL || fdat == AT_FDCWD); 2646 2647 error = pathbuf_maybe_copyin(arg, seg, &pb); 2648 if (error) { 2649 return error; 2650 } 2651 pathstring = pathbuf_stringcopy_get(pb); 2652 if (pathstring == NULL) { 2653 pathbuf_destroy(pb); 2654 return ENOMEM; 2655 } 2656 2657 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2658 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2659 goto out; 2660 vp = nd.ni_vp; 2661 2662 /* 2663 * The root of a mounted filesystem cannot be deleted. 2664 */ 2665 if ((vp->v_vflag & VV_ROOT) != 0) { 2666 error = EBUSY; 2667 goto abort; 2668 } 2669 2670 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2671 error = EBUSY; 2672 goto abort; 2673 } 2674 2675 /* 2676 * No rmdir "." please. 2677 */ 2678 if (nd.ni_dvp == vp) { 2679 error = EINVAL; 2680 goto abort; 2681 } 2682 2683 /* 2684 * AT_REMOVEDIR is required to remove a directory 2685 */ 2686 if (vp->v_type == VDIR) { 2687 if (!(flags & AT_REMOVEDIR)) { 2688 error = EPERM; 2689 goto abort; 2690 } else { 2691 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2692 vput(nd.ni_dvp); 2693 goto out; 2694 } 2695 } 2696 2697 /* 2698 * Starting here we only deal with non directories. 2699 */ 2700 if (flags & AT_REMOVEDIR) { 2701 error = ENOTDIR; 2702 goto abort; 2703 } 2704 2705 #if NVERIEXEC > 0 2706 /* Handle remove requests for veriexec entries. */ 2707 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2708 goto abort; 2709 } 2710 #endif /* NVERIEXEC > 0 */ 2711 2712 #ifdef FILEASSOC 2713 (void)fileassoc_file_delete(vp); 2714 #endif /* FILEASSOC */ 2715 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2716 vput(nd.ni_dvp); 2717 goto out; 2718 2719 abort: 2720 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2721 if (nd.ni_dvp == vp) 2722 vrele(nd.ni_dvp); 2723 else 2724 vput(nd.ni_dvp); 2725 vput(vp); 2726 2727 out: 2728 pathbuf_stringcopy_put(pb, pathstring); 2729 pathbuf_destroy(pb); 2730 return (error); 2731 } 2732 2733 /* 2734 * Reposition read/write file offset. 2735 */ 2736 int 2737 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2738 { 2739 /* { 2740 syscallarg(int) fd; 2741 syscallarg(int) pad; 2742 syscallarg(off_t) offset; 2743 syscallarg(int) whence; 2744 } */ 2745 kauth_cred_t cred = l->l_cred; 2746 file_t *fp; 2747 struct vnode *vp; 2748 struct vattr vattr; 2749 off_t newoff; 2750 int error, fd; 2751 2752 fd = SCARG(uap, fd); 2753 2754 if ((fp = fd_getfile(fd)) == NULL) 2755 return (EBADF); 2756 2757 vp = fp->f_vnode; 2758 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2759 error = ESPIPE; 2760 goto out; 2761 } 2762 2763 vn_lock(vp, LK_SHARED | LK_RETRY); 2764 2765 switch (SCARG(uap, whence)) { 2766 case SEEK_CUR: 2767 newoff = fp->f_offset + SCARG(uap, offset); 2768 break; 2769 case SEEK_END: 2770 error = VOP_GETATTR(vp, &vattr, cred); 2771 if (error) { 2772 VOP_UNLOCK(vp); 2773 goto out; 2774 } 2775 newoff = SCARG(uap, offset) + vattr.va_size; 2776 break; 2777 case SEEK_SET: 2778 newoff = SCARG(uap, offset); 2779 break; 2780 default: 2781 error = EINVAL; 2782 VOP_UNLOCK(vp); 2783 goto out; 2784 } 2785 VOP_UNLOCK(vp); 2786 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2787 *(off_t *)retval = fp->f_offset = newoff; 2788 } 2789 out: 2790 fd_putfile(fd); 2791 return (error); 2792 } 2793 2794 /* 2795 * Positional read system call. 2796 */ 2797 int 2798 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2799 { 2800 /* { 2801 syscallarg(int) fd; 2802 syscallarg(void *) buf; 2803 syscallarg(size_t) nbyte; 2804 syscallarg(off_t) offset; 2805 } */ 2806 file_t *fp; 2807 struct vnode *vp; 2808 off_t offset; 2809 int error, fd = SCARG(uap, fd); 2810 2811 if ((fp = fd_getfile(fd)) == NULL) 2812 return (EBADF); 2813 2814 if ((fp->f_flag & FREAD) == 0) { 2815 fd_putfile(fd); 2816 return (EBADF); 2817 } 2818 2819 vp = fp->f_vnode; 2820 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2821 error = ESPIPE; 2822 goto out; 2823 } 2824 2825 offset = SCARG(uap, offset); 2826 2827 /* 2828 * XXX This works because no file systems actually 2829 * XXX take any action on the seek operation. 2830 */ 2831 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2832 goto out; 2833 2834 /* dofileread() will unuse the descriptor for us */ 2835 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2836 &offset, 0, retval)); 2837 2838 out: 2839 fd_putfile(fd); 2840 return (error); 2841 } 2842 2843 /* 2844 * Positional scatter read system call. 2845 */ 2846 int 2847 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2848 { 2849 /* { 2850 syscallarg(int) fd; 2851 syscallarg(const struct iovec *) iovp; 2852 syscallarg(int) iovcnt; 2853 syscallarg(off_t) offset; 2854 } */ 2855 off_t offset = SCARG(uap, offset); 2856 2857 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2858 SCARG(uap, iovcnt), &offset, 0, retval); 2859 } 2860 2861 /* 2862 * Positional write system call. 2863 */ 2864 int 2865 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2866 { 2867 /* { 2868 syscallarg(int) fd; 2869 syscallarg(const void *) buf; 2870 syscallarg(size_t) nbyte; 2871 syscallarg(off_t) offset; 2872 } */ 2873 file_t *fp; 2874 struct vnode *vp; 2875 off_t offset; 2876 int error, fd = SCARG(uap, fd); 2877 2878 if ((fp = fd_getfile(fd)) == NULL) 2879 return (EBADF); 2880 2881 if ((fp->f_flag & FWRITE) == 0) { 2882 fd_putfile(fd); 2883 return (EBADF); 2884 } 2885 2886 vp = fp->f_vnode; 2887 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2888 error = ESPIPE; 2889 goto out; 2890 } 2891 2892 offset = SCARG(uap, offset); 2893 2894 /* 2895 * XXX This works because no file systems actually 2896 * XXX take any action on the seek operation. 2897 */ 2898 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2899 goto out; 2900 2901 /* dofilewrite() will unuse the descriptor for us */ 2902 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2903 &offset, 0, retval)); 2904 2905 out: 2906 fd_putfile(fd); 2907 return (error); 2908 } 2909 2910 /* 2911 * Positional gather write system call. 2912 */ 2913 int 2914 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2915 { 2916 /* { 2917 syscallarg(int) fd; 2918 syscallarg(const struct iovec *) iovp; 2919 syscallarg(int) iovcnt; 2920 syscallarg(off_t) offset; 2921 } */ 2922 off_t offset = SCARG(uap, offset); 2923 2924 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2925 SCARG(uap, iovcnt), &offset, 0, retval); 2926 } 2927 2928 /* 2929 * Check access permissions. 2930 */ 2931 int 2932 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2933 { 2934 /* { 2935 syscallarg(const char *) path; 2936 syscallarg(int) flags; 2937 } */ 2938 2939 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2940 SCARG(uap, flags), 0); 2941 } 2942 2943 int 2944 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2945 int mode, int flags) 2946 { 2947 kauth_cred_t cred; 2948 struct vnode *vp; 2949 int error, nd_flag, vmode; 2950 struct pathbuf *pb; 2951 struct nameidata nd; 2952 2953 CTASSERT(F_OK == 0); 2954 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2955 /* nonsense mode */ 2956 return EINVAL; 2957 } 2958 2959 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2960 if (flags & AT_SYMLINK_NOFOLLOW) 2961 nd_flag &= ~FOLLOW; 2962 2963 error = pathbuf_copyin(path, &pb); 2964 if (error) 2965 return error; 2966 2967 NDINIT(&nd, LOOKUP, nd_flag, pb); 2968 2969 /* Override default credentials */ 2970 cred = kauth_cred_dup(l->l_cred); 2971 if (!(flags & AT_EACCESS)) { 2972 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2973 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2974 } 2975 nd.ni_cnd.cn_cred = cred; 2976 2977 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2978 pathbuf_destroy(pb); 2979 goto out; 2980 } 2981 vp = nd.ni_vp; 2982 pathbuf_destroy(pb); 2983 2984 /* Flags == 0 means only check for existence. */ 2985 if (mode) { 2986 vmode = 0; 2987 if (mode & R_OK) 2988 vmode |= VREAD; 2989 if (mode & W_OK) 2990 vmode |= VWRITE; 2991 if (mode & X_OK) 2992 vmode |= VEXEC; 2993 2994 error = VOP_ACCESS(vp, vmode, cred); 2995 if (!error && (vmode & VWRITE)) 2996 error = vn_writechk(vp); 2997 } 2998 vput(vp); 2999 out: 3000 kauth_cred_free(cred); 3001 return (error); 3002 } 3003 3004 int 3005 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3006 register_t *retval) 3007 { 3008 /* { 3009 syscallarg(int) fd; 3010 syscallarg(const char *) path; 3011 syscallarg(int) amode; 3012 syscallarg(int) flag; 3013 } */ 3014 3015 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3016 SCARG(uap, amode), SCARG(uap, flag)); 3017 } 3018 3019 /* 3020 * Common code for all sys_stat functions, including compat versions. 3021 */ 3022 int 3023 do_sys_stat(const char *userpath, unsigned int nd_flag, 3024 struct stat *sb) 3025 { 3026 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3027 } 3028 3029 int 3030 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3031 unsigned int nd_flag, struct stat *sb) 3032 { 3033 int error; 3034 struct pathbuf *pb; 3035 struct nameidata nd; 3036 3037 KASSERT(l != NULL || fdat == AT_FDCWD); 3038 3039 error = pathbuf_copyin(userpath, &pb); 3040 if (error) { 3041 return error; 3042 } 3043 3044 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3045 3046 error = fd_nameiat(l, fdat, &nd); 3047 if (error != 0) { 3048 pathbuf_destroy(pb); 3049 return error; 3050 } 3051 error = vn_stat(nd.ni_vp, sb); 3052 vput(nd.ni_vp); 3053 pathbuf_destroy(pb); 3054 return error; 3055 } 3056 3057 /* 3058 * Get file status; this version follows links. 3059 */ 3060 /* ARGSUSED */ 3061 int 3062 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3063 { 3064 /* { 3065 syscallarg(const char *) path; 3066 syscallarg(struct stat *) ub; 3067 } */ 3068 struct stat sb; 3069 int error; 3070 3071 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3072 if (error) 3073 return error; 3074 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3075 } 3076 3077 /* 3078 * Get file status; this version does not follow links. 3079 */ 3080 /* ARGSUSED */ 3081 int 3082 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3083 { 3084 /* { 3085 syscallarg(const char *) path; 3086 syscallarg(struct stat *) ub; 3087 } */ 3088 struct stat sb; 3089 int error; 3090 3091 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3092 if (error) 3093 return error; 3094 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3095 } 3096 3097 int 3098 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3099 register_t *retval) 3100 { 3101 /* { 3102 syscallarg(int) fd; 3103 syscallarg(const char *) path; 3104 syscallarg(struct stat *) buf; 3105 syscallarg(int) flag; 3106 } */ 3107 unsigned int nd_flag; 3108 struct stat sb; 3109 int error; 3110 3111 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3112 nd_flag = NOFOLLOW; 3113 else 3114 nd_flag = FOLLOW; 3115 3116 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3117 &sb); 3118 if (error) 3119 return error; 3120 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3121 } 3122 3123 /* 3124 * Get configurable pathname variables. 3125 */ 3126 /* ARGSUSED */ 3127 int 3128 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3129 { 3130 /* { 3131 syscallarg(const char *) path; 3132 syscallarg(int) name; 3133 } */ 3134 int error; 3135 struct pathbuf *pb; 3136 struct nameidata nd; 3137 3138 error = pathbuf_copyin(SCARG(uap, path), &pb); 3139 if (error) { 3140 return error; 3141 } 3142 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3143 if ((error = namei(&nd)) != 0) { 3144 pathbuf_destroy(pb); 3145 return (error); 3146 } 3147 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3148 vput(nd.ni_vp); 3149 pathbuf_destroy(pb); 3150 return (error); 3151 } 3152 3153 /* 3154 * Return target name of a symbolic link. 3155 */ 3156 /* ARGSUSED */ 3157 int 3158 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3159 register_t *retval) 3160 { 3161 /* { 3162 syscallarg(const char *) path; 3163 syscallarg(char *) buf; 3164 syscallarg(size_t) count; 3165 } */ 3166 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3167 SCARG(uap, buf), SCARG(uap, count), retval); 3168 } 3169 3170 static int 3171 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3172 size_t count, register_t *retval) 3173 { 3174 struct vnode *vp; 3175 struct iovec aiov; 3176 struct uio auio; 3177 int error; 3178 struct pathbuf *pb; 3179 struct nameidata nd; 3180 3181 error = pathbuf_copyin(path, &pb); 3182 if (error) { 3183 return error; 3184 } 3185 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3186 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3187 pathbuf_destroy(pb); 3188 return error; 3189 } 3190 vp = nd.ni_vp; 3191 pathbuf_destroy(pb); 3192 if (vp->v_type != VLNK) 3193 error = EINVAL; 3194 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3195 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3196 aiov.iov_base = buf; 3197 aiov.iov_len = count; 3198 auio.uio_iov = &aiov; 3199 auio.uio_iovcnt = 1; 3200 auio.uio_offset = 0; 3201 auio.uio_rw = UIO_READ; 3202 KASSERT(l == curlwp); 3203 auio.uio_vmspace = l->l_proc->p_vmspace; 3204 auio.uio_resid = count; 3205 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3206 *retval = count - auio.uio_resid; 3207 } 3208 vput(vp); 3209 return (error); 3210 } 3211 3212 int 3213 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3214 register_t *retval) 3215 { 3216 /* { 3217 syscallarg(int) fd; 3218 syscallarg(const char *) path; 3219 syscallarg(char *) buf; 3220 syscallarg(size_t) bufsize; 3221 } */ 3222 3223 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3224 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3225 } 3226 3227 /* 3228 * Change flags of a file given a path name. 3229 */ 3230 /* ARGSUSED */ 3231 int 3232 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3233 { 3234 /* { 3235 syscallarg(const char *) path; 3236 syscallarg(u_long) flags; 3237 } */ 3238 struct vnode *vp; 3239 int error; 3240 3241 error = namei_simple_user(SCARG(uap, path), 3242 NSM_FOLLOW_TRYEMULROOT, &vp); 3243 if (error != 0) 3244 return (error); 3245 error = change_flags(vp, SCARG(uap, flags), l); 3246 vput(vp); 3247 return (error); 3248 } 3249 3250 /* 3251 * Change flags of a file given a file descriptor. 3252 */ 3253 /* ARGSUSED */ 3254 int 3255 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3256 { 3257 /* { 3258 syscallarg(int) fd; 3259 syscallarg(u_long) flags; 3260 } */ 3261 struct vnode *vp; 3262 file_t *fp; 3263 int error; 3264 3265 /* fd_getvnode() will use the descriptor for us */ 3266 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3267 return (error); 3268 vp = fp->f_vnode; 3269 error = change_flags(vp, SCARG(uap, flags), l); 3270 VOP_UNLOCK(vp); 3271 fd_putfile(SCARG(uap, fd)); 3272 return (error); 3273 } 3274 3275 /* 3276 * Change flags of a file given a path name; this version does 3277 * not follow links. 3278 */ 3279 int 3280 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3281 { 3282 /* { 3283 syscallarg(const char *) path; 3284 syscallarg(u_long) flags; 3285 } */ 3286 struct vnode *vp; 3287 int error; 3288 3289 error = namei_simple_user(SCARG(uap, path), 3290 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3291 if (error != 0) 3292 return (error); 3293 error = change_flags(vp, SCARG(uap, flags), l); 3294 vput(vp); 3295 return (error); 3296 } 3297 3298 /* 3299 * Common routine to change flags of a file. 3300 */ 3301 int 3302 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3303 { 3304 struct vattr vattr; 3305 int error; 3306 3307 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3308 3309 vattr_null(&vattr); 3310 vattr.va_flags = flags; 3311 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3312 3313 return (error); 3314 } 3315 3316 /* 3317 * Change mode of a file given path name; this version follows links. 3318 */ 3319 /* ARGSUSED */ 3320 int 3321 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3322 { 3323 /* { 3324 syscallarg(const char *) path; 3325 syscallarg(int) mode; 3326 } */ 3327 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3328 SCARG(uap, mode), 0); 3329 } 3330 3331 int 3332 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3333 { 3334 int error; 3335 struct vnode *vp; 3336 namei_simple_flags_t ns_flag; 3337 3338 if (flags & AT_SYMLINK_NOFOLLOW) 3339 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3340 else 3341 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3342 3343 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3344 if (error != 0) 3345 return error; 3346 3347 error = change_mode(vp, mode, l); 3348 3349 vrele(vp); 3350 3351 return (error); 3352 } 3353 3354 /* 3355 * Change mode of a file given a file descriptor. 3356 */ 3357 /* ARGSUSED */ 3358 int 3359 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3360 { 3361 /* { 3362 syscallarg(int) fd; 3363 syscallarg(int) mode; 3364 } */ 3365 file_t *fp; 3366 int error; 3367 3368 /* fd_getvnode() will use the descriptor for us */ 3369 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3370 return (error); 3371 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3372 fd_putfile(SCARG(uap, fd)); 3373 return (error); 3374 } 3375 3376 int 3377 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3378 register_t *retval) 3379 { 3380 /* { 3381 syscallarg(int) fd; 3382 syscallarg(const char *) path; 3383 syscallarg(int) mode; 3384 syscallarg(int) flag; 3385 } */ 3386 3387 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3388 SCARG(uap, mode), SCARG(uap, flag)); 3389 } 3390 3391 /* 3392 * Change mode of a file given path name; this version does not follow links. 3393 */ 3394 /* ARGSUSED */ 3395 int 3396 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3397 { 3398 /* { 3399 syscallarg(const char *) path; 3400 syscallarg(int) mode; 3401 } */ 3402 int error; 3403 struct vnode *vp; 3404 3405 error = namei_simple_user(SCARG(uap, path), 3406 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3407 if (error != 0) 3408 return (error); 3409 3410 error = change_mode(vp, SCARG(uap, mode), l); 3411 3412 vrele(vp); 3413 return (error); 3414 } 3415 3416 /* 3417 * Common routine to set mode given a vnode. 3418 */ 3419 static int 3420 change_mode(struct vnode *vp, int mode, struct lwp *l) 3421 { 3422 struct vattr vattr; 3423 int error; 3424 3425 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3426 vattr_null(&vattr); 3427 vattr.va_mode = mode & ALLPERMS; 3428 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3429 VOP_UNLOCK(vp); 3430 return (error); 3431 } 3432 3433 /* 3434 * Set ownership given a path name; this version follows links. 3435 */ 3436 /* ARGSUSED */ 3437 int 3438 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3439 { 3440 /* { 3441 syscallarg(const char *) path; 3442 syscallarg(uid_t) uid; 3443 syscallarg(gid_t) gid; 3444 } */ 3445 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3446 SCARG(uap, gid), 0); 3447 } 3448 3449 int 3450 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3451 gid_t gid, int flags) 3452 { 3453 int error; 3454 struct vnode *vp; 3455 namei_simple_flags_t ns_flag; 3456 3457 if (flags & AT_SYMLINK_NOFOLLOW) 3458 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3459 else 3460 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3461 3462 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3463 if (error != 0) 3464 return error; 3465 3466 error = change_owner(vp, uid, gid, l, 0); 3467 3468 vrele(vp); 3469 3470 return (error); 3471 } 3472 3473 /* 3474 * Set ownership given a path name; this version follows links. 3475 * Provides POSIX semantics. 3476 */ 3477 /* ARGSUSED */ 3478 int 3479 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3480 { 3481 /* { 3482 syscallarg(const char *) path; 3483 syscallarg(uid_t) uid; 3484 syscallarg(gid_t) gid; 3485 } */ 3486 int error; 3487 struct vnode *vp; 3488 3489 error = namei_simple_user(SCARG(uap, path), 3490 NSM_FOLLOW_TRYEMULROOT, &vp); 3491 if (error != 0) 3492 return (error); 3493 3494 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3495 3496 vrele(vp); 3497 return (error); 3498 } 3499 3500 /* 3501 * Set ownership given a file descriptor. 3502 */ 3503 /* ARGSUSED */ 3504 int 3505 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3506 { 3507 /* { 3508 syscallarg(int) fd; 3509 syscallarg(uid_t) uid; 3510 syscallarg(gid_t) gid; 3511 } */ 3512 int error; 3513 file_t *fp; 3514 3515 /* fd_getvnode() will use the descriptor for us */ 3516 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3517 return (error); 3518 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3519 l, 0); 3520 fd_putfile(SCARG(uap, fd)); 3521 return (error); 3522 } 3523 3524 int 3525 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3526 register_t *retval) 3527 { 3528 /* { 3529 syscallarg(int) fd; 3530 syscallarg(const char *) path; 3531 syscallarg(uid_t) owner; 3532 syscallarg(gid_t) group; 3533 syscallarg(int) flag; 3534 } */ 3535 3536 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3537 SCARG(uap, owner), SCARG(uap, group), 3538 SCARG(uap, flag)); 3539 } 3540 3541 /* 3542 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3543 */ 3544 /* ARGSUSED */ 3545 int 3546 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3547 { 3548 /* { 3549 syscallarg(int) fd; 3550 syscallarg(uid_t) uid; 3551 syscallarg(gid_t) gid; 3552 } */ 3553 int error; 3554 file_t *fp; 3555 3556 /* fd_getvnode() will use the descriptor for us */ 3557 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3558 return (error); 3559 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3560 l, 1); 3561 fd_putfile(SCARG(uap, fd)); 3562 return (error); 3563 } 3564 3565 /* 3566 * Set ownership given a path name; this version does not follow links. 3567 */ 3568 /* ARGSUSED */ 3569 int 3570 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3571 { 3572 /* { 3573 syscallarg(const char *) path; 3574 syscallarg(uid_t) uid; 3575 syscallarg(gid_t) gid; 3576 } */ 3577 int error; 3578 struct vnode *vp; 3579 3580 error = namei_simple_user(SCARG(uap, path), 3581 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3582 if (error != 0) 3583 return (error); 3584 3585 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3586 3587 vrele(vp); 3588 return (error); 3589 } 3590 3591 /* 3592 * Set ownership given a path name; this version does not follow links. 3593 * Provides POSIX/XPG semantics. 3594 */ 3595 /* ARGSUSED */ 3596 int 3597 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3598 { 3599 /* { 3600 syscallarg(const char *) path; 3601 syscallarg(uid_t) uid; 3602 syscallarg(gid_t) gid; 3603 } */ 3604 int error; 3605 struct vnode *vp; 3606 3607 error = namei_simple_user(SCARG(uap, path), 3608 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3609 if (error != 0) 3610 return (error); 3611 3612 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3613 3614 vrele(vp); 3615 return (error); 3616 } 3617 3618 /* 3619 * Common routine to set ownership given a vnode. 3620 */ 3621 static int 3622 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3623 int posix_semantics) 3624 { 3625 struct vattr vattr; 3626 mode_t newmode; 3627 int error; 3628 3629 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3630 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3631 goto out; 3632 3633 #define CHANGED(x) ((int)(x) != -1) 3634 newmode = vattr.va_mode; 3635 if (posix_semantics) { 3636 /* 3637 * POSIX/XPG semantics: if the caller is not the super-user, 3638 * clear set-user-id and set-group-id bits. Both POSIX and 3639 * the XPG consider the behaviour for calls by the super-user 3640 * implementation-defined; we leave the set-user-id and set- 3641 * group-id settings intact in that case. 3642 */ 3643 if (vattr.va_mode & S_ISUID) { 3644 if (kauth_authorize_vnode(l->l_cred, 3645 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3646 newmode &= ~S_ISUID; 3647 } 3648 if (vattr.va_mode & S_ISGID) { 3649 if (kauth_authorize_vnode(l->l_cred, 3650 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3651 newmode &= ~S_ISGID; 3652 } 3653 } else { 3654 /* 3655 * NetBSD semantics: when changing owner and/or group, 3656 * clear the respective bit(s). 3657 */ 3658 if (CHANGED(uid)) 3659 newmode &= ~S_ISUID; 3660 if (CHANGED(gid)) 3661 newmode &= ~S_ISGID; 3662 } 3663 /* Update va_mode iff altered. */ 3664 if (vattr.va_mode == newmode) 3665 newmode = VNOVAL; 3666 3667 vattr_null(&vattr); 3668 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3669 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3670 vattr.va_mode = newmode; 3671 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3672 #undef CHANGED 3673 3674 out: 3675 VOP_UNLOCK(vp); 3676 return (error); 3677 } 3678 3679 /* 3680 * Set the access and modification times given a path name; this 3681 * version follows links. 3682 */ 3683 /* ARGSUSED */ 3684 int 3685 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3686 register_t *retval) 3687 { 3688 /* { 3689 syscallarg(const char *) path; 3690 syscallarg(const struct timeval *) tptr; 3691 } */ 3692 3693 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3694 SCARG(uap, tptr), UIO_USERSPACE); 3695 } 3696 3697 /* 3698 * Set the access and modification times given a file descriptor. 3699 */ 3700 /* ARGSUSED */ 3701 int 3702 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3703 register_t *retval) 3704 { 3705 /* { 3706 syscallarg(int) fd; 3707 syscallarg(const struct timeval *) tptr; 3708 } */ 3709 int error; 3710 file_t *fp; 3711 3712 /* fd_getvnode() will use the descriptor for us */ 3713 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3714 return (error); 3715 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3716 UIO_USERSPACE); 3717 fd_putfile(SCARG(uap, fd)); 3718 return (error); 3719 } 3720 3721 int 3722 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3723 register_t *retval) 3724 { 3725 /* { 3726 syscallarg(int) fd; 3727 syscallarg(const struct timespec *) tptr; 3728 } */ 3729 int error; 3730 file_t *fp; 3731 3732 /* fd_getvnode() will use the descriptor for us */ 3733 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3734 return (error); 3735 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3736 SCARG(uap, tptr), UIO_USERSPACE); 3737 fd_putfile(SCARG(uap, fd)); 3738 return (error); 3739 } 3740 3741 /* 3742 * Set the access and modification times given a path name; this 3743 * version does not follow links. 3744 */ 3745 int 3746 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3747 register_t *retval) 3748 { 3749 /* { 3750 syscallarg(const char *) path; 3751 syscallarg(const struct timeval *) tptr; 3752 } */ 3753 3754 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3755 SCARG(uap, tptr), UIO_USERSPACE); 3756 } 3757 3758 int 3759 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3760 register_t *retval) 3761 { 3762 /* { 3763 syscallarg(int) fd; 3764 syscallarg(const char *) path; 3765 syscallarg(const struct timespec *) tptr; 3766 syscallarg(int) flag; 3767 } */ 3768 int follow; 3769 const struct timespec *tptr; 3770 int error; 3771 3772 tptr = SCARG(uap, tptr); 3773 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3774 3775 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3776 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3777 3778 return error; 3779 } 3780 3781 /* 3782 * Common routine to set access and modification times given a vnode. 3783 */ 3784 int 3785 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3786 const struct timespec *tptr, enum uio_seg seg) 3787 { 3788 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3789 } 3790 3791 int 3792 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3793 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3794 { 3795 struct vattr vattr; 3796 int error, dorele = 0; 3797 namei_simple_flags_t sflags; 3798 bool vanull, setbirthtime; 3799 struct timespec ts[2]; 3800 3801 KASSERT(l != NULL || fdat == AT_FDCWD); 3802 3803 /* 3804 * I have checked all callers and they pass either FOLLOW, 3805 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3806 * is 0. More to the point, they don't pass anything else. 3807 * Let's keep it that way at least until the namei interfaces 3808 * are fully sanitized. 3809 */ 3810 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3811 sflags = (flag == FOLLOW) ? 3812 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3813 3814 if (tptr == NULL) { 3815 vanull = true; 3816 nanotime(&ts[0]); 3817 ts[1] = ts[0]; 3818 } else { 3819 vanull = false; 3820 if (seg != UIO_SYSSPACE) { 3821 error = copyin(tptr, ts, sizeof (ts)); 3822 if (error != 0) 3823 return error; 3824 } else { 3825 ts[0] = tptr[0]; 3826 ts[1] = tptr[1]; 3827 } 3828 } 3829 3830 if (ts[0].tv_nsec == UTIME_NOW) { 3831 nanotime(&ts[0]); 3832 if (ts[1].tv_nsec == UTIME_NOW) { 3833 vanull = true; 3834 ts[1] = ts[0]; 3835 } 3836 } else if (ts[1].tv_nsec == UTIME_NOW) 3837 nanotime(&ts[1]); 3838 3839 if (vp == NULL) { 3840 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3841 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3842 if (error != 0) 3843 return error; 3844 dorele = 1; 3845 } 3846 3847 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3848 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3849 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3850 vattr_null(&vattr); 3851 3852 if (ts[0].tv_nsec != UTIME_OMIT) 3853 vattr.va_atime = ts[0]; 3854 3855 if (ts[1].tv_nsec != UTIME_OMIT) { 3856 vattr.va_mtime = ts[1]; 3857 if (setbirthtime) 3858 vattr.va_birthtime = ts[1]; 3859 } 3860 3861 if (vanull) 3862 vattr.va_vaflags |= VA_UTIMES_NULL; 3863 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3864 VOP_UNLOCK(vp); 3865 3866 if (dorele != 0) 3867 vrele(vp); 3868 3869 return error; 3870 } 3871 3872 int 3873 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3874 const struct timeval *tptr, enum uio_seg seg) 3875 { 3876 struct timespec ts[2]; 3877 struct timespec *tsptr = NULL; 3878 int error; 3879 3880 if (tptr != NULL) { 3881 struct timeval tv[2]; 3882 3883 if (seg != UIO_SYSSPACE) { 3884 error = copyin(tptr, tv, sizeof (tv)); 3885 if (error != 0) 3886 return error; 3887 tptr = tv; 3888 } 3889 3890 if ((tv[0].tv_usec == UTIME_NOW) || 3891 (tv[0].tv_usec == UTIME_OMIT)) 3892 ts[0].tv_nsec = tv[0].tv_usec; 3893 else 3894 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3895 3896 if ((tv[1].tv_usec == UTIME_NOW) || 3897 (tv[1].tv_usec == UTIME_OMIT)) 3898 ts[1].tv_nsec = tv[1].tv_usec; 3899 else 3900 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3901 3902 tsptr = &ts[0]; 3903 } 3904 3905 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3906 } 3907 3908 /* 3909 * Truncate a file given its path name. 3910 */ 3911 /* ARGSUSED */ 3912 int 3913 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3914 { 3915 /* { 3916 syscallarg(const char *) path; 3917 syscallarg(int) pad; 3918 syscallarg(off_t) length; 3919 } */ 3920 struct vnode *vp; 3921 struct vattr vattr; 3922 int error; 3923 3924 if (SCARG(uap, length) < 0) 3925 return EINVAL; 3926 3927 error = namei_simple_user(SCARG(uap, path), 3928 NSM_FOLLOW_TRYEMULROOT, &vp); 3929 if (error != 0) 3930 return (error); 3931 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3932 if (vp->v_type == VDIR) 3933 error = EISDIR; 3934 else if ((error = vn_writechk(vp)) == 0 && 3935 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3936 vattr_null(&vattr); 3937 vattr.va_size = SCARG(uap, length); 3938 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3939 } 3940 vput(vp); 3941 return (error); 3942 } 3943 3944 /* 3945 * Truncate a file given a file descriptor. 3946 */ 3947 /* ARGSUSED */ 3948 int 3949 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3950 { 3951 /* { 3952 syscallarg(int) fd; 3953 syscallarg(int) pad; 3954 syscallarg(off_t) length; 3955 } */ 3956 struct vattr vattr; 3957 struct vnode *vp; 3958 file_t *fp; 3959 int error; 3960 3961 if (SCARG(uap, length) < 0) 3962 return EINVAL; 3963 3964 /* fd_getvnode() will use the descriptor for us */ 3965 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3966 return (error); 3967 if ((fp->f_flag & FWRITE) == 0) { 3968 error = EINVAL; 3969 goto out; 3970 } 3971 vp = fp->f_vnode; 3972 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3973 if (vp->v_type == VDIR) 3974 error = EISDIR; 3975 else if ((error = vn_writechk(vp)) == 0) { 3976 vattr_null(&vattr); 3977 vattr.va_size = SCARG(uap, length); 3978 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3979 } 3980 VOP_UNLOCK(vp); 3981 out: 3982 fd_putfile(SCARG(uap, fd)); 3983 return (error); 3984 } 3985 3986 /* 3987 * Sync an open file. 3988 */ 3989 /* ARGSUSED */ 3990 int 3991 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3992 { 3993 /* { 3994 syscallarg(int) fd; 3995 } */ 3996 struct vnode *vp; 3997 file_t *fp; 3998 int error; 3999 4000 /* fd_getvnode() will use the descriptor for us */ 4001 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4002 return (error); 4003 vp = fp->f_vnode; 4004 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4005 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4006 VOP_UNLOCK(vp); 4007 fd_putfile(SCARG(uap, fd)); 4008 return (error); 4009 } 4010 4011 /* 4012 * Sync a range of file data. API modeled after that found in AIX. 4013 * 4014 * FDATASYNC indicates that we need only save enough metadata to be able 4015 * to re-read the written data. Note we duplicate AIX's requirement that 4016 * the file be open for writing. 4017 */ 4018 /* ARGSUSED */ 4019 int 4020 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4021 { 4022 /* { 4023 syscallarg(int) fd; 4024 syscallarg(int) flags; 4025 syscallarg(off_t) start; 4026 syscallarg(off_t) length; 4027 } */ 4028 struct vnode *vp; 4029 file_t *fp; 4030 int flags, nflags; 4031 off_t s, e, len; 4032 int error; 4033 4034 /* fd_getvnode() will use the descriptor for us */ 4035 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4036 return (error); 4037 4038 if ((fp->f_flag & FWRITE) == 0) { 4039 error = EBADF; 4040 goto out; 4041 } 4042 4043 flags = SCARG(uap, flags); 4044 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4045 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4046 error = EINVAL; 4047 goto out; 4048 } 4049 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4050 if (flags & FDATASYNC) 4051 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4052 else 4053 nflags = FSYNC_WAIT; 4054 if (flags & FDISKSYNC) 4055 nflags |= FSYNC_CACHE; 4056 4057 len = SCARG(uap, length); 4058 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4059 if (len) { 4060 s = SCARG(uap, start); 4061 e = s + len; 4062 if (e < s) { 4063 error = EINVAL; 4064 goto out; 4065 } 4066 } else { 4067 e = 0; 4068 s = 0; 4069 } 4070 4071 vp = fp->f_vnode; 4072 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4073 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4074 VOP_UNLOCK(vp); 4075 out: 4076 fd_putfile(SCARG(uap, fd)); 4077 return (error); 4078 } 4079 4080 /* 4081 * Sync the data of an open file. 4082 */ 4083 /* ARGSUSED */ 4084 int 4085 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4086 { 4087 /* { 4088 syscallarg(int) fd; 4089 } */ 4090 struct vnode *vp; 4091 file_t *fp; 4092 int error; 4093 4094 /* fd_getvnode() will use the descriptor for us */ 4095 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4096 return (error); 4097 if ((fp->f_flag & FWRITE) == 0) { 4098 fd_putfile(SCARG(uap, fd)); 4099 return (EBADF); 4100 } 4101 vp = fp->f_vnode; 4102 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4103 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4104 VOP_UNLOCK(vp); 4105 fd_putfile(SCARG(uap, fd)); 4106 return (error); 4107 } 4108 4109 /* 4110 * Rename files, (standard) BSD semantics frontend. 4111 */ 4112 /* ARGSUSED */ 4113 int 4114 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4115 { 4116 /* { 4117 syscallarg(const char *) from; 4118 syscallarg(const char *) to; 4119 } */ 4120 4121 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4122 SCARG(uap, to), UIO_USERSPACE, 0)); 4123 } 4124 4125 int 4126 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4127 register_t *retval) 4128 { 4129 /* { 4130 syscallarg(int) fromfd; 4131 syscallarg(const char *) from; 4132 syscallarg(int) tofd; 4133 syscallarg(const char *) to; 4134 } */ 4135 4136 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4137 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4138 } 4139 4140 /* 4141 * Rename files, POSIX semantics frontend. 4142 */ 4143 /* ARGSUSED */ 4144 int 4145 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4146 { 4147 /* { 4148 syscallarg(const char *) from; 4149 syscallarg(const char *) to; 4150 } */ 4151 4152 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4153 SCARG(uap, to), UIO_USERSPACE, 1)); 4154 } 4155 4156 /* 4157 * Rename files. Source and destination must either both be directories, 4158 * or both not be directories. If target is a directory, it must be empty. 4159 * If `from' and `to' refer to the same object, the value of the `retain' 4160 * argument is used to determine whether `from' will be 4161 * 4162 * (retain == 0) deleted unless `from' and `to' refer to the same 4163 * object in the file system's name space (BSD). 4164 * (retain == 1) always retained (POSIX). 4165 * 4166 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4167 */ 4168 int 4169 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4170 { 4171 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4172 } 4173 4174 static int 4175 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4176 const char *to, enum uio_seg seg, int retain) 4177 { 4178 struct pathbuf *fpb, *tpb; 4179 struct nameidata fnd, tnd; 4180 struct vnode *fdvp, *fvp; 4181 struct vnode *tdvp, *tvp; 4182 struct mount *mp, *tmp; 4183 int error; 4184 4185 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4186 4187 error = pathbuf_maybe_copyin(from, seg, &fpb); 4188 if (error) 4189 goto out0; 4190 KASSERT(fpb != NULL); 4191 4192 error = pathbuf_maybe_copyin(to, seg, &tpb); 4193 if (error) 4194 goto out1; 4195 KASSERT(tpb != NULL); 4196 4197 /* 4198 * Lookup from. 4199 * 4200 * XXX LOCKPARENT is wrong because we don't actually want it 4201 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4202 * insane, so for the time being we need to leave it like this. 4203 */ 4204 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4205 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4206 goto out2; 4207 4208 /* 4209 * Pull out the important results of the lookup, fdvp and fvp. 4210 * Of course, fvp is bogus because we're about to unlock fdvp. 4211 */ 4212 fdvp = fnd.ni_dvp; 4213 fvp = fnd.ni_vp; 4214 KASSERT(fdvp != NULL); 4215 KASSERT(fvp != NULL); 4216 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4217 4218 /* 4219 * Make sure neither fdvp nor fvp is locked. 4220 */ 4221 if (fdvp != fvp) 4222 VOP_UNLOCK(fdvp); 4223 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4224 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4225 4226 /* 4227 * Reject renaming `.' and `..'. Can't do this until after 4228 * namei because we need namei's parsing to find the final 4229 * component name. (namei should just leave us with the final 4230 * component name and not look it up itself, but anyway...) 4231 * 4232 * This was here before because we used to relookup from 4233 * instead of to and relookup requires the caller to check 4234 * this, but now file systems may depend on this check, so we 4235 * must retain it until the file systems are all rototilled. 4236 */ 4237 if (((fnd.ni_cnd.cn_namelen == 1) && 4238 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4239 ((fnd.ni_cnd.cn_namelen == 2) && 4240 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4241 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4242 error = EINVAL; /* XXX EISDIR? */ 4243 goto abort0; 4244 } 4245 4246 /* 4247 * Lookup to. 4248 * 4249 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4250 * fvp here to decide whether to add CREATEDIR is a load of 4251 * bollocks because fvp might be the wrong node by now, since 4252 * fdvp is unlocked. 4253 * 4254 * XXX Why not pass CREATEDIR always? 4255 */ 4256 NDINIT(&tnd, RENAME, 4257 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4258 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4259 tpb); 4260 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4261 goto abort0; 4262 4263 /* 4264 * Pull out the important results of the lookup, tdvp and tvp. 4265 * Of course, tvp is bogus because we're about to unlock tdvp. 4266 */ 4267 tdvp = tnd.ni_dvp; 4268 tvp = tnd.ni_vp; 4269 KASSERT(tdvp != NULL); 4270 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4271 4272 /* 4273 * Make sure neither tdvp nor tvp is locked. 4274 */ 4275 if (tdvp != tvp) 4276 VOP_UNLOCK(tdvp); 4277 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4278 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4279 4280 /* 4281 * Reject renaming onto `.' or `..'. relookup is unhappy with 4282 * these, which is why we must do this here. Once upon a time 4283 * we relooked up from instead of to, and consequently didn't 4284 * need this check, but now that we relookup to instead of 4285 * from, we need this; and we shall need it forever forward 4286 * until the VOP_RENAME protocol changes, because file systems 4287 * will no doubt begin to depend on this check. 4288 */ 4289 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4290 error = EISDIR; 4291 goto abort1; 4292 } 4293 if ((tnd.ni_cnd.cn_namelen == 2) && 4294 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4295 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4296 error = EINVAL; 4297 goto abort1; 4298 } 4299 4300 /* 4301 * Get the mount point. If the file system has been unmounted, 4302 * which it may be because we're not holding any vnode locks, 4303 * then v_mount will be NULL. We're not really supposed to 4304 * read v_mount without holding the vnode lock, but since we 4305 * have fdvp referenced, if fdvp->v_mount changes then at worst 4306 * it will be set to NULL, not changed to another mount point. 4307 * And, of course, since it is up to the file system to 4308 * determine the real lock order, we can't lock both fdvp and 4309 * tdvp at the same time. 4310 */ 4311 mp = fdvp->v_mount; 4312 if (mp == NULL) { 4313 error = ENOENT; 4314 goto abort1; 4315 } 4316 4317 /* 4318 * Make sure the mount points match. Again, although we don't 4319 * hold any vnode locks, the v_mount fields may change -- but 4320 * at worst they will change to NULL, so this will never become 4321 * a cross-device rename, because we hold vnode references. 4322 * 4323 * XXX Because nothing is locked and the compiler may reorder 4324 * things here, unmounting the file system at an inopportune 4325 * moment may cause rename to fail with EXDEV when it really 4326 * should fail with ENOENT. 4327 */ 4328 tmp = tdvp->v_mount; 4329 if (tmp == NULL) { 4330 error = ENOENT; 4331 goto abort1; 4332 } 4333 4334 if (mp != tmp) { 4335 error = EXDEV; 4336 goto abort1; 4337 } 4338 4339 /* 4340 * Take the vfs rename lock to avoid cross-directory screw cases. 4341 * Nothing is locked currently, so taking this lock is safe. 4342 */ 4343 error = VFS_RENAMELOCK_ENTER(mp); 4344 if (error) 4345 goto abort1; 4346 4347 /* 4348 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4349 * and nothing is locked except for the vfs rename lock. 4350 * 4351 * The next step is a little rain dance to conform to the 4352 * insane lock protocol, even though it does nothing to ward 4353 * off race conditions. 4354 * 4355 * We need tdvp and tvp to be locked. However, because we have 4356 * unlocked tdvp in order to hold no locks while we take the 4357 * vfs rename lock, tvp may be wrong here, and we can't safely 4358 * lock it even if the sensible file systems will just unlock 4359 * it straight away. Consequently, we must lock tdvp and then 4360 * relookup tvp to get it locked. 4361 * 4362 * Finally, because the VOP_RENAME protocol is brain-damaged 4363 * and various file systems insanely depend on the semantics of 4364 * this brain damage, the lookup of to must be the last lookup 4365 * before VOP_RENAME. 4366 */ 4367 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4368 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4369 if (error) 4370 goto abort2; 4371 4372 /* 4373 * Drop the old tvp and pick up the new one -- which might be 4374 * the same, but that doesn't matter to us. After this, tdvp 4375 * and tvp should both be locked. 4376 */ 4377 if (tvp != NULL) 4378 vrele(tvp); 4379 tvp = tnd.ni_vp; 4380 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4381 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4382 4383 /* 4384 * The old do_sys_rename had various consistency checks here 4385 * involving fvp and tvp. fvp is bogus already here, and tvp 4386 * will become bogus soon in any sensible file system, so the 4387 * only purpose in putting these checks here is to give lip 4388 * service to these screw cases and to acknowledge that they 4389 * exist, not actually to handle them, but here you go 4390 * anyway... 4391 */ 4392 4393 /* 4394 * Acknowledge that directories and non-directories aren't 4395 * suposed to mix. 4396 */ 4397 if (tvp != NULL) { 4398 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4399 error = ENOTDIR; 4400 goto abort3; 4401 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4402 error = EISDIR; 4403 goto abort3; 4404 } 4405 } 4406 4407 /* 4408 * Acknowledge some random screw case, among the dozens that 4409 * might arise. 4410 */ 4411 if (fvp == tdvp) { 4412 error = EINVAL; 4413 goto abort3; 4414 } 4415 4416 /* 4417 * Acknowledge that POSIX has a wacky screw case. 4418 * 4419 * XXX Eventually the retain flag needs to be passed on to 4420 * VOP_RENAME. 4421 */ 4422 if (fvp == tvp) { 4423 if (retain) { 4424 error = 0; 4425 goto abort3; 4426 } else if ((fdvp == tdvp) && 4427 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4428 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4429 fnd.ni_cnd.cn_namelen))) { 4430 error = 0; 4431 goto abort3; 4432 } 4433 } 4434 4435 /* 4436 * Make sure veriexec can screw us up. (But a race can screw 4437 * up veriexec, of course -- remember, fvp and (soon) tvp are 4438 * bogus.) 4439 */ 4440 #if NVERIEXEC > 0 4441 { 4442 char *f1, *f2; 4443 size_t f1_len; 4444 size_t f2_len; 4445 4446 f1_len = fnd.ni_cnd.cn_namelen + 1; 4447 f1 = kmem_alloc(f1_len, KM_SLEEP); 4448 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4449 4450 f2_len = tnd.ni_cnd.cn_namelen + 1; 4451 f2 = kmem_alloc(f2_len, KM_SLEEP); 4452 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4453 4454 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4455 4456 kmem_free(f1, f1_len); 4457 kmem_free(f2, f2_len); 4458 4459 if (error) 4460 goto abort3; 4461 } 4462 #endif /* NVERIEXEC > 0 */ 4463 4464 /* 4465 * All ready. Incant the rename vop. 4466 */ 4467 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4468 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4469 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4470 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4471 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4472 4473 /* 4474 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4475 * tdvp and tvp. But we can't assert any of that. 4476 */ 4477 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4478 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4479 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4480 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4481 4482 /* 4483 * So all we have left to do is to drop the rename lock and 4484 * destroy the pathbufs. 4485 */ 4486 VFS_RENAMELOCK_EXIT(mp); 4487 goto out2; 4488 4489 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4490 VOP_UNLOCK(tvp); 4491 abort2: VOP_UNLOCK(tdvp); 4492 VFS_RENAMELOCK_EXIT(mp); 4493 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4494 vrele(tdvp); 4495 if (tvp != NULL) 4496 vrele(tvp); 4497 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4498 vrele(fdvp); 4499 vrele(fvp); 4500 out2: pathbuf_destroy(tpb); 4501 out1: pathbuf_destroy(fpb); 4502 out0: return error; 4503 } 4504 4505 /* 4506 * Make a directory file. 4507 */ 4508 /* ARGSUSED */ 4509 int 4510 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4511 { 4512 /* { 4513 syscallarg(const char *) path; 4514 syscallarg(int) mode; 4515 } */ 4516 4517 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4518 SCARG(uap, mode), UIO_USERSPACE); 4519 } 4520 4521 int 4522 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4523 register_t *retval) 4524 { 4525 /* { 4526 syscallarg(int) fd; 4527 syscallarg(const char *) path; 4528 syscallarg(int) mode; 4529 } */ 4530 4531 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4532 SCARG(uap, mode), UIO_USERSPACE); 4533 } 4534 4535 4536 int 4537 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4538 { 4539 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4540 } 4541 4542 static int 4543 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4544 enum uio_seg seg) 4545 { 4546 struct proc *p = curlwp->l_proc; 4547 struct vnode *vp; 4548 struct vattr vattr; 4549 int error; 4550 struct pathbuf *pb; 4551 struct nameidata nd; 4552 4553 KASSERT(l != NULL || fdat == AT_FDCWD); 4554 4555 /* XXX bollocks, should pass in a pathbuf */ 4556 error = pathbuf_maybe_copyin(path, seg, &pb); 4557 if (error) { 4558 return error; 4559 } 4560 4561 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4562 4563 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4564 pathbuf_destroy(pb); 4565 return (error); 4566 } 4567 vp = nd.ni_vp; 4568 if (vp != NULL) { 4569 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4570 if (nd.ni_dvp == vp) 4571 vrele(nd.ni_dvp); 4572 else 4573 vput(nd.ni_dvp); 4574 vrele(vp); 4575 pathbuf_destroy(pb); 4576 return (EEXIST); 4577 } 4578 vattr_null(&vattr); 4579 vattr.va_type = VDIR; 4580 /* We will read cwdi->cwdi_cmask unlocked. */ 4581 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4582 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4583 if (!error) 4584 vrele(nd.ni_vp); 4585 vput(nd.ni_dvp); 4586 pathbuf_destroy(pb); 4587 return (error); 4588 } 4589 4590 /* 4591 * Remove a directory file. 4592 */ 4593 /* ARGSUSED */ 4594 int 4595 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4596 { 4597 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4598 AT_REMOVEDIR, UIO_USERSPACE); 4599 } 4600 4601 /* 4602 * Read a block of directory entries in a file system independent format. 4603 */ 4604 int 4605 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4606 { 4607 /* { 4608 syscallarg(int) fd; 4609 syscallarg(char *) buf; 4610 syscallarg(size_t) count; 4611 } */ 4612 file_t *fp; 4613 int error, done; 4614 4615 /* fd_getvnode() will use the descriptor for us */ 4616 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4617 return (error); 4618 if ((fp->f_flag & FREAD) == 0) { 4619 error = EBADF; 4620 goto out; 4621 } 4622 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4623 SCARG(uap, count), &done, l, 0, 0); 4624 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4625 *retval = done; 4626 out: 4627 fd_putfile(SCARG(uap, fd)); 4628 return (error); 4629 } 4630 4631 /* 4632 * Set the mode mask for creation of filesystem nodes. 4633 */ 4634 int 4635 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4636 { 4637 /* { 4638 syscallarg(mode_t) newmask; 4639 } */ 4640 struct proc *p = l->l_proc; 4641 struct cwdinfo *cwdi; 4642 4643 /* 4644 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4645 * important is that we serialize changes to the mask. The 4646 * rw_exit() will issue a write memory barrier on our behalf, 4647 * and force the changes out to other CPUs (as it must use an 4648 * atomic operation, draining the local CPU's store buffers). 4649 */ 4650 cwdi = p->p_cwdi; 4651 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4652 *retval = cwdi->cwdi_cmask; 4653 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4654 rw_exit(&cwdi->cwdi_lock); 4655 4656 return (0); 4657 } 4658 4659 int 4660 dorevoke(struct vnode *vp, kauth_cred_t cred) 4661 { 4662 struct vattr vattr; 4663 int error, fs_decision; 4664 4665 vn_lock(vp, LK_SHARED | LK_RETRY); 4666 error = VOP_GETATTR(vp, &vattr, cred); 4667 VOP_UNLOCK(vp); 4668 if (error != 0) 4669 return error; 4670 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4671 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4672 fs_decision); 4673 if (!error) 4674 VOP_REVOKE(vp, REVOKEALL); 4675 return (error); 4676 } 4677 4678 /* 4679 * Void all references to file by ripping underlying filesystem 4680 * away from vnode. 4681 */ 4682 /* ARGSUSED */ 4683 int 4684 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4685 { 4686 /* { 4687 syscallarg(const char *) path; 4688 } */ 4689 struct vnode *vp; 4690 int error; 4691 4692 error = namei_simple_user(SCARG(uap, path), 4693 NSM_FOLLOW_TRYEMULROOT, &vp); 4694 if (error != 0) 4695 return (error); 4696 error = dorevoke(vp, l->l_cred); 4697 vrele(vp); 4698 return (error); 4699 } 4700 4701 /* 4702 * Allocate backing store for a file, filling a hole without having to 4703 * explicitly write anything out. 4704 */ 4705 /* ARGSUSED */ 4706 int 4707 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4708 register_t *retval) 4709 { 4710 /* { 4711 syscallarg(int) fd; 4712 syscallarg(off_t) pos; 4713 syscallarg(off_t) len; 4714 } */ 4715 int fd; 4716 off_t pos, len; 4717 struct file *fp; 4718 struct vnode *vp; 4719 int error; 4720 4721 fd = SCARG(uap, fd); 4722 pos = SCARG(uap, pos); 4723 len = SCARG(uap, len); 4724 4725 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4726 *retval = EINVAL; 4727 return 0; 4728 } 4729 4730 error = fd_getvnode(fd, &fp); 4731 if (error) { 4732 *retval = error; 4733 return 0; 4734 } 4735 if ((fp->f_flag & FWRITE) == 0) { 4736 error = EBADF; 4737 goto fail; 4738 } 4739 vp = fp->f_vnode; 4740 4741 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4742 if (vp->v_type == VDIR) { 4743 error = EISDIR; 4744 } else { 4745 error = VOP_FALLOCATE(vp, pos, len); 4746 } 4747 VOP_UNLOCK(vp); 4748 4749 fail: 4750 fd_putfile(fd); 4751 *retval = error; 4752 return 0; 4753 } 4754 4755 /* 4756 * Deallocate backing store for a file, creating a hole. Also used for 4757 * invoking TRIM on disks. 4758 */ 4759 /* ARGSUSED */ 4760 int 4761 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4762 register_t *retval) 4763 { 4764 /* { 4765 syscallarg(int) fd; 4766 syscallarg(off_t) pos; 4767 syscallarg(off_t) len; 4768 } */ 4769 int fd; 4770 off_t pos, len; 4771 struct file *fp; 4772 struct vnode *vp; 4773 int error; 4774 4775 fd = SCARG(uap, fd); 4776 pos = SCARG(uap, pos); 4777 len = SCARG(uap, len); 4778 4779 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4780 return EINVAL; 4781 } 4782 4783 error = fd_getvnode(fd, &fp); 4784 if (error) { 4785 return error; 4786 } 4787 if ((fp->f_flag & FWRITE) == 0) { 4788 error = EBADF; 4789 goto fail; 4790 } 4791 vp = fp->f_vnode; 4792 4793 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4794 if (vp->v_type == VDIR) { 4795 error = EISDIR; 4796 } else { 4797 error = VOP_FDISCARD(vp, pos, len); 4798 } 4799 VOP_UNLOCK(vp); 4800 4801 fail: 4802 fd_putfile(fd); 4803 return error; 4804 } 4805
Indexes created Sat Sep 27 14:09:57 GMT 2025