vfs_syscalls.c revision 1.518.4.2
1 /* $NetBSD: vfs_syscalls.c,v 1.518.4.2 2020/04/08 14:08:52 martin Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009, 2019, 2020 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 1.518.4.2 2020/04/08 14:08:52 martin 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 /* 282 * Enabling MNT_UNION requires a covered mountpoint and 283 * must not happen on the root mount. 284 */ 285 if ((flags & MNT_UNION) != 0 && mp->mnt_vnodecovered == NULLVP) { 286 error = EOPNOTSUPP; 287 goto out; 288 } 289 290 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 291 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 292 if (error) 293 goto out; 294 295 error = vfs_suspend(mp, 0); 296 if (error) 297 goto out; 298 299 mutex_enter(mp->mnt_updating); 300 301 mp->mnt_flag &= ~MNT_OP_FLAGS; 302 mp->mnt_flag |= flags & MNT_OP_FLAGS; 303 304 /* 305 * Set the mount level flags. 306 */ 307 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 308 if ((flags & MNT_RDONLY)) 309 mp->mnt_iflag |= IMNT_WANTRDONLY; 310 else 311 mp->mnt_iflag |= IMNT_WANTRDWR; 312 } 313 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 314 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 315 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 316 mp->mnt_flag &= ~MNT_RDONLY; 317 318 error = VFS_MOUNT(mp, path, data, data_len); 319 320 if (error && data != NULL) { 321 int error2; 322 323 /* 324 * Update failed; let's try and see if it was an 325 * export request. For compat with 3.0 and earlier. 326 */ 327 error2 = vfs_hooks_reexport(mp, path, data); 328 329 /* 330 * Only update error code if the export request was 331 * understood but some problem occurred while 332 * processing it. 333 */ 334 if (error2 != EJUSTRETURN) 335 error = error2; 336 } 337 338 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 339 mp->mnt_flag |= MNT_RDONLY; 340 if (error) 341 mp->mnt_flag = saved_flags; 342 mp->mnt_flag &= ~MNT_OP_FLAGS; 343 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 344 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 345 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 346 vfs_syncer_add_to_worklist(mp); 347 } else { 348 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 349 vfs_syncer_remove_from_worklist(mp); 350 } 351 mutex_exit(mp->mnt_updating); 352 vfs_resume(mp); 353 354 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 355 (flags & MNT_EXTATTR)) { 356 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 357 NULL, 0, NULL) != 0) { 358 printf("%s: failed to start extattr, error = %d", 359 mp->mnt_stat.f_mntonname, error); 360 mp->mnt_flag &= ~MNT_EXTATTR; 361 } 362 } 363 364 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 365 !(flags & MNT_EXTATTR)) { 366 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 367 NULL, 0, NULL) != 0) { 368 printf("%s: failed to stop extattr, error = %d", 369 mp->mnt_stat.f_mntonname, error); 370 mp->mnt_flag |= MNT_RDONLY; 371 } 372 } 373 out: 374 return (error); 375 } 376 377 static int 378 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 379 struct vfsops **vfsops) 380 { 381 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 382 int error; 383 384 if (type_seg == UIO_USERSPACE) { 385 /* Copy file-system type from userspace. */ 386 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 387 } else { 388 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 389 KASSERT(error == 0); 390 } 391 392 if (error) { 393 /* 394 * Historically, filesystem types were identified by numbers. 395 * If we get an integer for the filesystem type instead of a 396 * string, we check to see if it matches one of the historic 397 * filesystem types. 398 */ 399 u_long fsindex = (u_long)fstype; 400 if (fsindex >= nmountcompatnames || 401 mountcompatnames[fsindex] == NULL) 402 return ENODEV; 403 strlcpy(fstypename, mountcompatnames[fsindex], 404 sizeof(fstypename)); 405 } 406 407 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 408 if (strcmp(fstypename, "ufs") == 0) 409 fstypename[0] = 'f'; 410 411 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 412 return 0; 413 414 /* If we can autoload a vfs module, try again */ 415 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 416 417 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 418 return 0; 419 420 return ENODEV; 421 } 422 423 static int 424 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 425 void *data, size_t *data_len) 426 { 427 struct mount *mp; 428 int error; 429 430 /* If MNT_GETARGS is specified, it should be the only flag. */ 431 if (flags & ~MNT_GETARGS) 432 return EINVAL; 433 434 mp = vp->v_mount; 435 436 /* XXX: probably some notion of "can see" here if we want isolation. */ 437 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 438 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 439 if (error) 440 return error; 441 442 if ((vp->v_vflag & VV_ROOT) == 0) 443 return EINVAL; 444 445 if (vfs_busy(mp)) 446 return EPERM; 447 448 mutex_enter(mp->mnt_updating); 449 mp->mnt_flag &= ~MNT_OP_FLAGS; 450 mp->mnt_flag |= MNT_GETARGS; 451 error = VFS_MOUNT(mp, path, data, data_len); 452 mp->mnt_flag &= ~MNT_OP_FLAGS; 453 mutex_exit(mp->mnt_updating); 454 455 vfs_unbusy(mp); 456 return (error); 457 } 458 459 int 460 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 461 { 462 /* { 463 syscallarg(const char *) type; 464 syscallarg(const char *) path; 465 syscallarg(int) flags; 466 syscallarg(void *) data; 467 syscallarg(size_t) data_len; 468 } */ 469 470 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 471 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 472 SCARG(uap, data_len), retval); 473 } 474 475 int 476 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 477 const char *path, int flags, void *data, enum uio_seg data_seg, 478 size_t data_len, register_t *retval) 479 { 480 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 481 struct vnode *vp; 482 void *data_buf = data; 483 bool vfsopsrele = false; 484 size_t alloc_sz = 0; 485 int error; 486 487 /* 488 * Get vnode to be covered 489 */ 490 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 491 if (error != 0) { 492 vp = NULL; 493 goto done; 494 } 495 496 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 497 vfsops = vp->v_mount->mnt_op; 498 } else { 499 /* 'type' is userspace */ 500 error = mount_get_vfsops(type, type_seg, &vfsops); 501 if (error != 0) 502 goto done; 503 vfsopsrele = true; 504 } 505 506 /* 507 * We allow data to be NULL, even for userspace. Some fs's don't need 508 * it. The others will handle NULL. 509 */ 510 if (data != NULL && data_seg == UIO_USERSPACE) { 511 if (data_len == 0) { 512 /* No length supplied, use default for filesystem */ 513 data_len = vfsops->vfs_min_mount_data; 514 515 /* 516 * Hopefully a longer buffer won't make copyin() fail. 517 * For compatibility with 3.0 and earlier. 518 */ 519 if (flags & MNT_UPDATE 520 && data_len < sizeof (struct mnt_export_args30)) 521 data_len = sizeof (struct mnt_export_args30); 522 } 523 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 524 error = EINVAL; 525 goto done; 526 } 527 alloc_sz = data_len; 528 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 529 530 /* NFS needs the buffer even for mnt_getargs .... */ 531 error = copyin(data, data_buf, data_len); 532 if (error != 0) 533 goto done; 534 } 535 536 if (flags & MNT_GETARGS) { 537 if (data_len == 0) { 538 error = EINVAL; 539 goto done; 540 } 541 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 542 if (error != 0) 543 goto done; 544 if (data_seg == UIO_USERSPACE) 545 error = copyout(data_buf, data, data_len); 546 *retval = data_len; 547 } else if (flags & MNT_UPDATE) { 548 error = mount_update(l, vp, path, flags, data_buf, &data_len); 549 } else { 550 /* Locking is handled internally in mount_domount(). */ 551 KASSERT(vfsopsrele == true); 552 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 553 &data_len); 554 vfsopsrele = false; 555 } 556 if (!error) 557 KNOTE(&fs_klist, VQ_MOUNT); 558 559 done: 560 if (vfsopsrele) 561 vfs_delref(vfsops); 562 if (vp != NULL) { 563 vrele(vp); 564 } 565 if (data_buf != data) 566 kmem_free(data_buf, alloc_sz); 567 return (error); 568 } 569 570 /* 571 * Unmount a file system. 572 * 573 * Note: unmount takes a path to the vnode mounted on as argument, 574 * not special file (as before). 575 */ 576 /* ARGSUSED */ 577 int 578 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 579 { 580 /* { 581 syscallarg(const char *) path; 582 syscallarg(int) flags; 583 } */ 584 struct vnode *vp; 585 struct mount *mp; 586 int error; 587 struct pathbuf *pb; 588 struct nameidata nd; 589 590 error = pathbuf_copyin(SCARG(uap, path), &pb); 591 if (error) { 592 return error; 593 } 594 595 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 596 if ((error = namei(&nd)) != 0) { 597 pathbuf_destroy(pb); 598 return error; 599 } 600 vp = nd.ni_vp; 601 pathbuf_destroy(pb); 602 603 mp = vp->v_mount; 604 vfs_ref(mp); 605 VOP_UNLOCK(vp); 606 607 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 608 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 609 if (error) { 610 vrele(vp); 611 vfs_rele(mp); 612 return (error); 613 } 614 615 /* 616 * Don't allow unmounting the root file system. 617 */ 618 if (mp->mnt_flag & MNT_ROOTFS) { 619 vrele(vp); 620 vfs_rele(mp); 621 return (EINVAL); 622 } 623 624 /* 625 * Must be the root of the filesystem 626 */ 627 if ((vp->v_vflag & VV_ROOT) == 0) { 628 vrele(vp); 629 vfs_rele(mp); 630 return (EINVAL); 631 } 632 633 vrele(vp); 634 error = dounmount(mp, SCARG(uap, flags), l); 635 vfs_rele(mp); 636 if (!error) 637 KNOTE(&fs_klist, VQ_UNMOUNT); 638 return error; 639 } 640 641 /* 642 * Sync each mounted filesystem. 643 */ 644 #ifdef DEBUG 645 int syncprt = 0; 646 struct ctldebug debug0 = { "syncprt", &syncprt }; 647 #endif 648 649 void 650 do_sys_sync(struct lwp *l) 651 { 652 mount_iterator_t *iter; 653 struct mount *mp; 654 int asyncflag; 655 656 mountlist_iterator_init(&iter); 657 while ((mp = mountlist_iterator_next(iter)) != NULL) { 658 mutex_enter(mp->mnt_updating); 659 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 660 asyncflag = mp->mnt_flag & MNT_ASYNC; 661 mp->mnt_flag &= ~MNT_ASYNC; 662 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 663 if (asyncflag) 664 mp->mnt_flag |= MNT_ASYNC; 665 } 666 mutex_exit(mp->mnt_updating); 667 } 668 mountlist_iterator_destroy(iter); 669 #ifdef DEBUG 670 if (syncprt) 671 vfs_bufstats(); 672 #endif /* DEBUG */ 673 } 674 675 /* ARGSUSED */ 676 int 677 sys_sync(struct lwp *l, const void *v, register_t *retval) 678 { 679 do_sys_sync(l); 680 return (0); 681 } 682 683 684 /* 685 * Access or change filesystem quotas. 686 * 687 * (this is really 14 different calls bundled into one) 688 */ 689 690 static int 691 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 692 { 693 struct quotastat info_k; 694 int error; 695 696 /* ensure any padding bytes are cleared */ 697 memset(&info_k, 0, sizeof(info_k)); 698 699 error = vfs_quotactl_stat(mp, &info_k); 700 if (error) { 701 return error; 702 } 703 704 return copyout(&info_k, info_u, sizeof(info_k)); 705 } 706 707 static int 708 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 709 struct quotaidtypestat *info_u) 710 { 711 struct quotaidtypestat info_k; 712 int error; 713 714 /* ensure any padding bytes are cleared */ 715 memset(&info_k, 0, sizeof(info_k)); 716 717 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 718 if (error) { 719 return error; 720 } 721 722 return copyout(&info_k, info_u, sizeof(info_k)); 723 } 724 725 static int 726 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 727 struct quotaobjtypestat *info_u) 728 { 729 struct quotaobjtypestat info_k; 730 int error; 731 732 /* ensure any padding bytes are cleared */ 733 memset(&info_k, 0, sizeof(info_k)); 734 735 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 736 if (error) { 737 return error; 738 } 739 740 return copyout(&info_k, info_u, sizeof(info_k)); 741 } 742 743 static int 744 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 745 struct quotaval *val_u) 746 { 747 struct quotakey key_k; 748 struct quotaval val_k; 749 int error; 750 751 /* ensure any padding bytes are cleared */ 752 memset(&val_k, 0, sizeof(val_k)); 753 754 error = copyin(key_u, &key_k, sizeof(key_k)); 755 if (error) { 756 return error; 757 } 758 759 error = vfs_quotactl_get(mp, &key_k, &val_k); 760 if (error) { 761 return error; 762 } 763 764 return copyout(&val_k, val_u, sizeof(val_k)); 765 } 766 767 static int 768 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 769 const struct quotaval *val_u) 770 { 771 struct quotakey key_k; 772 struct quotaval val_k; 773 int error; 774 775 error = copyin(key_u, &key_k, sizeof(key_k)); 776 if (error) { 777 return error; 778 } 779 780 error = copyin(val_u, &val_k, sizeof(val_k)); 781 if (error) { 782 return error; 783 } 784 785 return vfs_quotactl_put(mp, &key_k, &val_k); 786 } 787 788 static int 789 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 790 { 791 struct quotakey key_k; 792 int error; 793 794 error = copyin(key_u, &key_k, sizeof(key_k)); 795 if (error) { 796 return error; 797 } 798 799 return vfs_quotactl_del(mp, &key_k); 800 } 801 802 static int 803 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 804 { 805 struct quotakcursor cursor_k; 806 int error; 807 808 /* ensure any padding bytes are cleared */ 809 memset(&cursor_k, 0, sizeof(cursor_k)); 810 811 error = vfs_quotactl_cursoropen(mp, &cursor_k); 812 if (error) { 813 return error; 814 } 815 816 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 817 } 818 819 static int 820 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 821 { 822 struct quotakcursor cursor_k; 823 int error; 824 825 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 826 if (error) { 827 return error; 828 } 829 830 return vfs_quotactl_cursorclose(mp, &cursor_k); 831 } 832 833 static int 834 do_sys_quotactl_cursorskipidtype(struct mount *mp, 835 struct quotakcursor *cursor_u, int idtype) 836 { 837 struct quotakcursor cursor_k; 838 int error; 839 840 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 841 if (error) { 842 return error; 843 } 844 845 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 846 if (error) { 847 return error; 848 } 849 850 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 851 } 852 853 static int 854 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 855 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 856 unsigned *ret_u) 857 { 858 #define CGET_STACK_MAX 8 859 struct quotakcursor cursor_k; 860 struct quotakey stackkeys[CGET_STACK_MAX]; 861 struct quotaval stackvals[CGET_STACK_MAX]; 862 struct quotakey *keys_k; 863 struct quotaval *vals_k; 864 unsigned ret_k; 865 int error; 866 867 if (maxnum > 128) { 868 maxnum = 128; 869 } 870 871 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 872 if (error) { 873 return error; 874 } 875 876 if (maxnum <= CGET_STACK_MAX) { 877 keys_k = stackkeys; 878 vals_k = stackvals; 879 /* ensure any padding bytes are cleared */ 880 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 881 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 882 } else { 883 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 884 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 885 } 886 887 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 888 &ret_k); 889 if (error) { 890 goto fail; 891 } 892 893 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 894 if (error) { 895 goto fail; 896 } 897 898 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 899 if (error) { 900 goto fail; 901 } 902 903 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 904 if (error) { 905 goto fail; 906 } 907 908 /* do last to maximize the chance of being able to recover a failure */ 909 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 910 911 fail: 912 if (keys_k != stackkeys) { 913 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 914 } 915 if (vals_k != stackvals) { 916 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 917 } 918 return error; 919 } 920 921 static int 922 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 923 int *ret_u) 924 { 925 struct quotakcursor cursor_k; 926 int ret_k; 927 int error; 928 929 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 930 if (error) { 931 return error; 932 } 933 934 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 935 if (error) { 936 return error; 937 } 938 939 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 940 if (error) { 941 return error; 942 } 943 944 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 945 } 946 947 static int 948 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 949 { 950 struct quotakcursor cursor_k; 951 int error; 952 953 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 954 if (error) { 955 return error; 956 } 957 958 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 959 if (error) { 960 return error; 961 } 962 963 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 964 } 965 966 static int 967 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 968 { 969 char *path_k; 970 int error; 971 972 /* XXX this should probably be a struct pathbuf */ 973 path_k = PNBUF_GET(); 974 error = copyin(path_u, path_k, PATH_MAX); 975 if (error) { 976 PNBUF_PUT(path_k); 977 return error; 978 } 979 980 error = vfs_quotactl_quotaon(mp, idtype, path_k); 981 982 PNBUF_PUT(path_k); 983 return error; 984 } 985 986 static int 987 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 988 { 989 return vfs_quotactl_quotaoff(mp, idtype); 990 } 991 992 int 993 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 994 { 995 struct mount *mp; 996 struct vnode *vp; 997 int error; 998 999 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 1000 if (error != 0) 1001 return (error); 1002 mp = vp->v_mount; 1003 1004 switch (args->qc_op) { 1005 case QUOTACTL_STAT: 1006 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 1007 break; 1008 case QUOTACTL_IDTYPESTAT: 1009 error = do_sys_quotactl_idtypestat(mp, 1010 args->u.idtypestat.qc_idtype, 1011 args->u.idtypestat.qc_info); 1012 break; 1013 case QUOTACTL_OBJTYPESTAT: 1014 error = do_sys_quotactl_objtypestat(mp, 1015 args->u.objtypestat.qc_objtype, 1016 args->u.objtypestat.qc_info); 1017 break; 1018 case QUOTACTL_GET: 1019 error = do_sys_quotactl_get(mp, 1020 args->u.get.qc_key, 1021 args->u.get.qc_val); 1022 break; 1023 case QUOTACTL_PUT: 1024 error = do_sys_quotactl_put(mp, 1025 args->u.put.qc_key, 1026 args->u.put.qc_val); 1027 break; 1028 case QUOTACTL_DEL: 1029 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1030 break; 1031 case QUOTACTL_CURSOROPEN: 1032 error = do_sys_quotactl_cursoropen(mp, 1033 args->u.cursoropen.qc_cursor); 1034 break; 1035 case QUOTACTL_CURSORCLOSE: 1036 error = do_sys_quotactl_cursorclose(mp, 1037 args->u.cursorclose.qc_cursor); 1038 break; 1039 case QUOTACTL_CURSORSKIPIDTYPE: 1040 error = do_sys_quotactl_cursorskipidtype(mp, 1041 args->u.cursorskipidtype.qc_cursor, 1042 args->u.cursorskipidtype.qc_idtype); 1043 break; 1044 case QUOTACTL_CURSORGET: 1045 error = do_sys_quotactl_cursorget(mp, 1046 args->u.cursorget.qc_cursor, 1047 args->u.cursorget.qc_keys, 1048 args->u.cursorget.qc_vals, 1049 args->u.cursorget.qc_maxnum, 1050 args->u.cursorget.qc_ret); 1051 break; 1052 case QUOTACTL_CURSORATEND: 1053 error = do_sys_quotactl_cursoratend(mp, 1054 args->u.cursoratend.qc_cursor, 1055 args->u.cursoratend.qc_ret); 1056 break; 1057 case QUOTACTL_CURSORREWIND: 1058 error = do_sys_quotactl_cursorrewind(mp, 1059 args->u.cursorrewind.qc_cursor); 1060 break; 1061 case QUOTACTL_QUOTAON: 1062 error = do_sys_quotactl_quotaon(mp, 1063 args->u.quotaon.qc_idtype, 1064 args->u.quotaon.qc_quotafile); 1065 break; 1066 case QUOTACTL_QUOTAOFF: 1067 error = do_sys_quotactl_quotaoff(mp, 1068 args->u.quotaoff.qc_idtype); 1069 break; 1070 default: 1071 error = EINVAL; 1072 break; 1073 } 1074 1075 vrele(vp); 1076 return error; 1077 } 1078 1079 /* ARGSUSED */ 1080 int 1081 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1082 register_t *retval) 1083 { 1084 /* { 1085 syscallarg(const char *) path; 1086 syscallarg(struct quotactl_args *) args; 1087 } */ 1088 struct quotactl_args args; 1089 int error; 1090 1091 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1092 if (error) { 1093 return error; 1094 } 1095 1096 return do_sys_quotactl(SCARG(uap, path), &args); 1097 } 1098 1099 int 1100 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1101 int root) 1102 { 1103 struct vnode *rvp; 1104 int error = 0; 1105 1106 /* 1107 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1108 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1109 * overrides MNT_NOWAIT. 1110 */ 1111 KASSERT(l == curlwp); 1112 rvp = cwdrdir(); 1113 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1114 (flags != MNT_WAIT && flags != 0)) { 1115 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1116 } else { 1117 /* Get the filesystem stats now */ 1118 memset(sp, 0, sizeof(*sp)); 1119 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1120 if (rvp) 1121 vrele(rvp); 1122 return error; 1123 } 1124 if (rvp == NULL) 1125 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1126 } 1127 1128 if (rvp != NULL) { 1129 size_t len; 1130 char *bp; 1131 char c; 1132 char *path = PNBUF_GET(); 1133 1134 bp = path + MAXPATHLEN; 1135 *--bp = '\0'; 1136 error = getcwd_common(rvp, rootvnode, &bp, path, 1137 MAXPATHLEN / 2, 0, l); 1138 if (error) { 1139 PNBUF_PUT(path); 1140 vrele(rvp); 1141 return error; 1142 } 1143 len = strlen(bp); 1144 if (len != 1) { 1145 /* 1146 * for mount points that are below our root, we can see 1147 * them, so we fix up the pathname and return them. The 1148 * rest we cannot see, so we don't allow viewing the 1149 * data. 1150 */ 1151 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1152 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1153 (void)strlcpy(sp->f_mntonname, 1154 c == '\0' ? "/" : &sp->f_mntonname[len], 1155 sizeof(sp->f_mntonname)); 1156 } else { 1157 if (root) 1158 (void)strlcpy(sp->f_mntonname, "/", 1159 sizeof(sp->f_mntonname)); 1160 else 1161 error = EPERM; 1162 } 1163 } 1164 PNBUF_PUT(path); 1165 vrele(rvp); 1166 } 1167 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1168 return error; 1169 } 1170 1171 /* 1172 * Get filesystem statistics by path. 1173 */ 1174 int 1175 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1176 { 1177 struct mount *mp; 1178 int error; 1179 struct vnode *vp; 1180 1181 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1182 if (error != 0) 1183 return error; 1184 mp = vp->v_mount; 1185 error = dostatvfs(mp, sb, l, flags, 1); 1186 vrele(vp); 1187 return error; 1188 } 1189 1190 /* ARGSUSED */ 1191 int 1192 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1193 { 1194 /* { 1195 syscallarg(const char *) path; 1196 syscallarg(struct statvfs *) buf; 1197 syscallarg(int) flags; 1198 } */ 1199 struct statvfs *sb; 1200 int error; 1201 1202 sb = STATVFSBUF_GET(); 1203 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1204 if (error == 0) 1205 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1206 STATVFSBUF_PUT(sb); 1207 return error; 1208 } 1209 1210 /* 1211 * Get filesystem statistics by fd. 1212 */ 1213 int 1214 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1215 { 1216 file_t *fp; 1217 struct mount *mp; 1218 int error; 1219 1220 /* fd_getvnode() will use the descriptor for us */ 1221 if ((error = fd_getvnode(fd, &fp)) != 0) 1222 return (error); 1223 mp = fp->f_vnode->v_mount; 1224 error = dostatvfs(mp, sb, curlwp, flags, 1); 1225 fd_putfile(fd); 1226 return error; 1227 } 1228 1229 /* ARGSUSED */ 1230 int 1231 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1232 { 1233 /* { 1234 syscallarg(int) fd; 1235 syscallarg(struct statvfs *) buf; 1236 syscallarg(int) flags; 1237 } */ 1238 struct statvfs *sb; 1239 int error; 1240 1241 sb = STATVFSBUF_GET(); 1242 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1243 if (error == 0) 1244 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1245 STATVFSBUF_PUT(sb); 1246 return error; 1247 } 1248 1249 1250 /* 1251 * Get statistics on all filesystems. 1252 */ 1253 int 1254 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1255 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1256 register_t *retval) 1257 { 1258 int root = 0; 1259 mount_iterator_t *iter; 1260 struct proc *p = l->l_proc; 1261 struct mount *mp; 1262 struct statvfs *sb; 1263 size_t count, maxcount; 1264 int error = 0; 1265 1266 sb = STATVFSBUF_GET(); 1267 maxcount = bufsize / entry_sz; 1268 count = 0; 1269 mountlist_iterator_init(&iter); 1270 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1271 if (sfsp && count < maxcount) { 1272 error = dostatvfs(mp, sb, l, flags, 0); 1273 if (error) { 1274 error = 0; 1275 continue; 1276 } 1277 error = copyfn(sb, sfsp, entry_sz); 1278 if (error) 1279 goto out; 1280 sfsp = (char *)sfsp + entry_sz; 1281 root |= strcmp(sb->f_mntonname, "/") == 0; 1282 } 1283 count++; 1284 } 1285 1286 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1287 /* 1288 * fake a root entry 1289 */ 1290 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1291 sb, l, flags, 1); 1292 if (error != 0) 1293 goto out; 1294 if (sfsp) { 1295 error = copyfn(sb, sfsp, entry_sz); 1296 if (error != 0) 1297 goto out; 1298 } 1299 count++; 1300 } 1301 if (sfsp && count > maxcount) 1302 *retval = maxcount; 1303 else 1304 *retval = count; 1305 out: 1306 mountlist_iterator_destroy(iter); 1307 STATVFSBUF_PUT(sb); 1308 return error; 1309 } 1310 1311 int 1312 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1313 { 1314 /* { 1315 syscallarg(struct statvfs *) buf; 1316 syscallarg(size_t) bufsize; 1317 syscallarg(int) flags; 1318 } */ 1319 1320 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1321 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1322 } 1323 1324 /* 1325 * Change current working directory to a given file descriptor. 1326 */ 1327 /* ARGSUSED */ 1328 int 1329 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1330 { 1331 /* { 1332 syscallarg(int) fd; 1333 } */ 1334 struct cwdinfo *cwdi; 1335 struct vnode *vp, *tdp; 1336 struct mount *mp; 1337 file_t *fp; 1338 int error, fd; 1339 1340 /* fd_getvnode() will use the descriptor for us */ 1341 fd = SCARG(uap, fd); 1342 if ((error = fd_getvnode(fd, &fp)) != 0) 1343 return (error); 1344 vp = fp->f_vnode; 1345 1346 vref(vp); 1347 vn_lock(vp, LK_SHARED | LK_RETRY); 1348 if (vp->v_type != VDIR) 1349 error = ENOTDIR; 1350 else 1351 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1352 if (error) { 1353 vput(vp); 1354 goto out; 1355 } 1356 while ((mp = vp->v_mountedhere) != NULL) { 1357 error = vfs_busy(mp); 1358 vput(vp); 1359 if (error != 0) 1360 goto out; 1361 error = VFS_ROOT(mp, LK_SHARED, &tdp); 1362 vfs_unbusy(mp); 1363 if (error) 1364 goto out; 1365 vp = tdp; 1366 } 1367 VOP_UNLOCK(vp); 1368 1369 /* 1370 * Disallow changing to a directory not under the process's 1371 * current root directory (if there is one). 1372 */ 1373 cwdi = cwdenter(RW_WRITER); 1374 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1375 vrele(vp); 1376 error = EPERM; /* operation not permitted */ 1377 } else { 1378 vrele(cwdi->cwdi_cdir); 1379 cwdi->cwdi_cdir = vp; 1380 } 1381 cwdexit(cwdi); 1382 1383 out: 1384 fd_putfile(fd); 1385 return (error); 1386 } 1387 1388 /* 1389 * Change this process's notion of the root directory to a given file 1390 * descriptor. 1391 */ 1392 int 1393 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1394 { 1395 struct vnode *vp; 1396 file_t *fp; 1397 int error, fd = SCARG(uap, fd); 1398 1399 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1400 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1401 return error; 1402 /* fd_getvnode() will use the descriptor for us */ 1403 if ((error = fd_getvnode(fd, &fp)) != 0) 1404 return error; 1405 vp = fp->f_vnode; 1406 vn_lock(vp, LK_SHARED | LK_RETRY); 1407 if (vp->v_type != VDIR) 1408 error = ENOTDIR; 1409 else 1410 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1411 VOP_UNLOCK(vp); 1412 if (error) 1413 goto out; 1414 vref(vp); 1415 change_root(vp); 1416 1417 out: 1418 fd_putfile(fd); 1419 return (error); 1420 } 1421 1422 /* 1423 * Change current working directory (``.''). 1424 */ 1425 /* ARGSUSED */ 1426 int 1427 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1428 { 1429 /* { 1430 syscallarg(const char *) path; 1431 } */ 1432 struct cwdinfo *cwdi; 1433 int error; 1434 struct vnode *vp, *ovp; 1435 1436 error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, &vp, l); 1437 if (error != 0) 1438 return (error); 1439 1440 cwdi = cwdenter(RW_WRITER); 1441 ovp = cwdi->cwdi_cdir; 1442 cwdi->cwdi_cdir = vp; 1443 cwdexit(cwdi); 1444 vrele(ovp); 1445 return (0); 1446 } 1447 1448 /* 1449 * Change notion of root (``/'') directory. 1450 */ 1451 /* ARGSUSED */ 1452 int 1453 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1454 { 1455 /* { 1456 syscallarg(const char *) path; 1457 } */ 1458 int error; 1459 struct vnode *vp; 1460 1461 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1462 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1463 return (error); 1464 1465 error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, &vp, l); 1466 if (error == 0) 1467 change_root(vp); 1468 return error; 1469 } 1470 1471 /* 1472 * Common routine for chroot and fchroot. 1473 * NB: callers need to properly authorize the change root operation. 1474 */ 1475 void 1476 change_root(struct vnode *vp) 1477 { 1478 struct cwdinfo *cwdi; 1479 kauth_cred_t ncred; 1480 struct lwp *l = curlwp; 1481 struct proc *p = l->l_proc; 1482 1483 ncred = kauth_cred_alloc(); 1484 1485 cwdi = cwdenter(RW_WRITER); 1486 if (cwdi->cwdi_rdir != NULL) 1487 vrele(cwdi->cwdi_rdir); 1488 cwdi->cwdi_rdir = vp; 1489 1490 /* 1491 * Prevent escaping from chroot by putting the root under 1492 * the working directory. Silently chdir to / if we aren't 1493 * already there. 1494 */ 1495 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1496 /* 1497 * XXX would be more failsafe to change directory to a 1498 * deadfs node here instead 1499 */ 1500 vrele(cwdi->cwdi_cdir); 1501 vref(vp); 1502 cwdi->cwdi_cdir = vp; 1503 } 1504 cwdexit(cwdi); 1505 1506 /* Get a write lock on the process credential. */ 1507 proc_crmod_enter(); 1508 1509 kauth_cred_clone(p->p_cred, ncred); 1510 kauth_proc_chroot(ncred, p->p_cwdi); 1511 1512 /* Broadcast our credentials to the process and other LWPs. */ 1513 proc_crmod_leave(ncred, p->p_cred, true); 1514 } 1515 1516 /* 1517 * Common routine for chroot and chdir. 1518 * XXX "where" should be enum uio_seg 1519 */ 1520 int 1521 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1522 { 1523 struct pathbuf *pb; 1524 struct nameidata nd; 1525 int error; 1526 1527 error = pathbuf_maybe_copyin(path, where, &pb); 1528 if (error) { 1529 return error; 1530 } 1531 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT, pb); 1532 if ((error = namei(&nd)) != 0) { 1533 pathbuf_destroy(pb); 1534 return error; 1535 } 1536 *vpp = nd.ni_vp; 1537 pathbuf_destroy(pb); 1538 1539 if ((*vpp)->v_type != VDIR) 1540 error = ENOTDIR; 1541 else 1542 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1543 1544 if (error) 1545 vput(*vpp); 1546 else 1547 VOP_UNLOCK(*vpp); 1548 return (error); 1549 } 1550 1551 /* 1552 * Internals of sys_open - path has already been converted into a pathbuf 1553 * (so we can easily reuse this function from other parts of the kernel, 1554 * like posix_spawn post-processing). 1555 */ 1556 int 1557 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1558 int open_mode, int *fd) 1559 { 1560 struct proc *p = l->l_proc; 1561 struct cwdinfo *cwdi = p->p_cwdi; 1562 file_t *fp; 1563 struct vnode *vp; 1564 int flags, cmode; 1565 int indx, error; 1566 struct nameidata nd; 1567 1568 if (open_flags & O_SEARCH) { 1569 open_flags &= ~(int)O_SEARCH; 1570 } 1571 1572 flags = FFLAGS(open_flags); 1573 if ((flags & (FREAD | FWRITE)) == 0) 1574 return EINVAL; 1575 1576 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1577 return error; 1578 } 1579 1580 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1581 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1582 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1583 if (dvp != NULL) 1584 NDAT(&nd, dvp); 1585 1586 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1587 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1588 fd_abort(p, fp, indx); 1589 if ((error == EDUPFD || error == EMOVEFD) && 1590 l->l_dupfd >= 0 && /* XXX from fdopen */ 1591 (error = 1592 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1593 *fd = indx; 1594 return 0; 1595 } 1596 if (error == ERESTART) 1597 error = EINTR; 1598 return error; 1599 } 1600 1601 l->l_dupfd = 0; 1602 vp = nd.ni_vp; 1603 1604 if ((error = open_setfp(l, fp, vp, indx, flags))) 1605 return error; 1606 1607 VOP_UNLOCK(vp); 1608 *fd = indx; 1609 fd_affix(p, fp, indx); 1610 return 0; 1611 } 1612 1613 int 1614 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1615 { 1616 struct pathbuf *pb; 1617 int error, oflags; 1618 1619 oflags = FFLAGS(open_flags); 1620 if ((oflags & (FREAD | FWRITE)) == 0) 1621 return EINVAL; 1622 1623 pb = pathbuf_create(path); 1624 if (pb == NULL) 1625 return ENOMEM; 1626 1627 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1628 pathbuf_destroy(pb); 1629 1630 return error; 1631 } 1632 1633 static int 1634 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1635 int mode, int *fd) 1636 { 1637 file_t *dfp = NULL; 1638 struct vnode *dvp = NULL; 1639 struct pathbuf *pb; 1640 const char *pathstring = NULL; 1641 int error; 1642 1643 if (path == NULL) { 1644 MODULE_HOOK_CALL(vfs_openat_10_hook, (&pb), enosys(), error); 1645 if (error == ENOSYS) 1646 goto no_compat; 1647 if (error) 1648 return error; 1649 } else { 1650 no_compat: 1651 error = pathbuf_copyin(path, &pb); 1652 if (error) 1653 return error; 1654 } 1655 1656 pathstring = pathbuf_stringcopy_get(pb); 1657 1658 /* 1659 * fdat is ignored if: 1660 * 1) if fdat is AT_FDCWD, which means use current directory as base. 1661 * 2) if path is absolute, then fdat is useless. 1662 */ 1663 if (fdat != AT_FDCWD && pathstring[0] != '/') { 1664 /* fd_getvnode() will use the descriptor for us */ 1665 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1666 goto out; 1667 1668 dvp = dfp->f_vnode; 1669 } 1670 1671 error = do_open(l, dvp, pb, flags, mode, fd); 1672 1673 if (dfp != NULL) 1674 fd_putfile(fdat); 1675 out: 1676 pathbuf_stringcopy_put(pb, pathstring); 1677 pathbuf_destroy(pb); 1678 return error; 1679 } 1680 1681 int 1682 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1683 { 1684 /* { 1685 syscallarg(const char *) path; 1686 syscallarg(int) flags; 1687 syscallarg(int) mode; 1688 } */ 1689 int error; 1690 int fd; 1691 1692 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1693 SCARG(uap, flags), SCARG(uap, mode), &fd); 1694 1695 if (error == 0) 1696 *retval = fd; 1697 1698 return error; 1699 } 1700 1701 int 1702 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1703 { 1704 /* { 1705 syscallarg(int) fd; 1706 syscallarg(const char *) path; 1707 syscallarg(int) oflags; 1708 syscallarg(int) mode; 1709 } */ 1710 int error; 1711 int fd; 1712 1713 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1714 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1715 1716 if (error == 0) 1717 *retval = fd; 1718 1719 return error; 1720 } 1721 1722 static void 1723 vfs__fhfree(fhandle_t *fhp) 1724 { 1725 size_t fhsize; 1726 1727 fhsize = FHANDLE_SIZE(fhp); 1728 kmem_free(fhp, fhsize); 1729 } 1730 1731 /* 1732 * vfs_composefh: compose a filehandle. 1733 */ 1734 1735 int 1736 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1737 { 1738 struct mount *mp; 1739 struct fid *fidp; 1740 int error; 1741 size_t needfhsize; 1742 size_t fidsize; 1743 1744 mp = vp->v_mount; 1745 fidp = NULL; 1746 if (*fh_size < FHANDLE_SIZE_MIN) { 1747 fidsize = 0; 1748 } else { 1749 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1750 if (fhp != NULL) { 1751 memset(fhp, 0, *fh_size); 1752 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1753 fidp = &fhp->fh_fid; 1754 } 1755 } 1756 error = VFS_VPTOFH(vp, fidp, &fidsize); 1757 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1758 if (error == 0 && *fh_size < needfhsize) { 1759 error = E2BIG; 1760 } 1761 *fh_size = needfhsize; 1762 return error; 1763 } 1764 1765 int 1766 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1767 { 1768 struct mount *mp; 1769 fhandle_t *fhp; 1770 size_t fhsize; 1771 size_t fidsize; 1772 int error; 1773 1774 mp = vp->v_mount; 1775 fidsize = 0; 1776 error = VFS_VPTOFH(vp, NULL, &fidsize); 1777 KASSERT(error != 0); 1778 if (error != E2BIG) { 1779 goto out; 1780 } 1781 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1782 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1783 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1784 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1785 if (error == 0) { 1786 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1787 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1788 *fhpp = fhp; 1789 } else { 1790 kmem_free(fhp, fhsize); 1791 } 1792 out: 1793 return error; 1794 } 1795 1796 void 1797 vfs_composefh_free(fhandle_t *fhp) 1798 { 1799 1800 vfs__fhfree(fhp); 1801 } 1802 1803 /* 1804 * vfs_fhtovp: lookup a vnode by a filehandle. 1805 */ 1806 1807 int 1808 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1809 { 1810 struct mount *mp; 1811 int error; 1812 1813 *vpp = NULL; 1814 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1815 if (mp == NULL) { 1816 error = ESTALE; 1817 goto out; 1818 } 1819 if (mp->mnt_op->vfs_fhtovp == NULL) { 1820 error = EOPNOTSUPP; 1821 goto out; 1822 } 1823 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), LK_EXCLUSIVE, vpp); 1824 out: 1825 return error; 1826 } 1827 1828 /* 1829 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1830 * the needed size. 1831 */ 1832 1833 int 1834 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1835 { 1836 fhandle_t *fhp; 1837 int error; 1838 1839 if (fhsize > FHANDLE_SIZE_MAX) { 1840 return EINVAL; 1841 } 1842 if (fhsize < FHANDLE_SIZE_MIN) { 1843 return EINVAL; 1844 } 1845 again: 1846 fhp = kmem_alloc(fhsize, KM_SLEEP); 1847 error = copyin(ufhp, fhp, fhsize); 1848 if (error == 0) { 1849 /* XXX this check shouldn't be here */ 1850 if (FHANDLE_SIZE(fhp) == fhsize) { 1851 *fhpp = fhp; 1852 return 0; 1853 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1854 /* 1855 * a kludge for nfsv2 padded handles. 1856 */ 1857 size_t sz; 1858 1859 sz = FHANDLE_SIZE(fhp); 1860 kmem_free(fhp, fhsize); 1861 fhsize = sz; 1862 goto again; 1863 } else { 1864 /* 1865 * userland told us wrong size. 1866 */ 1867 error = EINVAL; 1868 } 1869 } 1870 kmem_free(fhp, fhsize); 1871 return error; 1872 } 1873 1874 void 1875 vfs_copyinfh_free(fhandle_t *fhp) 1876 { 1877 1878 vfs__fhfree(fhp); 1879 } 1880 1881 /* 1882 * Get file handle system call 1883 */ 1884 int 1885 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1886 { 1887 /* { 1888 syscallarg(char *) fname; 1889 syscallarg(fhandle_t *) fhp; 1890 syscallarg(size_t *) fh_size; 1891 } */ 1892 struct vnode *vp; 1893 fhandle_t *fh; 1894 int error; 1895 struct pathbuf *pb; 1896 struct nameidata nd; 1897 size_t sz; 1898 size_t usz; 1899 1900 /* 1901 * Must be super user 1902 */ 1903 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1904 0, NULL, NULL, NULL); 1905 if (error) 1906 return (error); 1907 1908 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1909 if (error) { 1910 return error; 1911 } 1912 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1913 error = namei(&nd); 1914 if (error) { 1915 pathbuf_destroy(pb); 1916 return error; 1917 } 1918 vp = nd.ni_vp; 1919 pathbuf_destroy(pb); 1920 1921 error = vfs_composefh_alloc(vp, &fh); 1922 vput(vp); 1923 if (error != 0) { 1924 return error; 1925 } 1926 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1927 if (error != 0) { 1928 goto out; 1929 } 1930 sz = FHANDLE_SIZE(fh); 1931 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1932 if (error != 0) { 1933 goto out; 1934 } 1935 if (usz >= sz) { 1936 error = copyout(fh, SCARG(uap, fhp), sz); 1937 } else { 1938 error = E2BIG; 1939 } 1940 out: 1941 vfs_composefh_free(fh); 1942 return (error); 1943 } 1944 1945 /* 1946 * Open a file given a file handle. 1947 * 1948 * Check permissions, allocate an open file structure, 1949 * and call the device open routine if any. 1950 */ 1951 1952 int 1953 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1954 register_t *retval) 1955 { 1956 file_t *fp; 1957 struct vnode *vp = NULL; 1958 kauth_cred_t cred = l->l_cred; 1959 file_t *nfp; 1960 int indx, error; 1961 struct vattr va; 1962 fhandle_t *fh; 1963 int flags; 1964 proc_t *p; 1965 1966 p = curproc; 1967 1968 /* 1969 * Must be super user 1970 */ 1971 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1972 0, NULL, NULL, NULL))) 1973 return (error); 1974 1975 if (oflags & O_SEARCH) { 1976 oflags &= ~(int)O_SEARCH; 1977 } 1978 1979 flags = FFLAGS(oflags); 1980 if ((flags & (FREAD | FWRITE)) == 0) 1981 return (EINVAL); 1982 if ((flags & O_CREAT)) 1983 return (EINVAL); 1984 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1985 return (error); 1986 fp = nfp; 1987 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1988 if (error != 0) { 1989 goto bad; 1990 } 1991 error = vfs_fhtovp(fh, &vp); 1992 vfs_copyinfh_free(fh); 1993 if (error != 0) { 1994 goto bad; 1995 } 1996 1997 /* Now do an effective vn_open */ 1998 1999 if (vp->v_type == VSOCK) { 2000 error = EOPNOTSUPP; 2001 goto bad; 2002 } 2003 error = vn_openchk(vp, cred, flags); 2004 if (error != 0) 2005 goto bad; 2006 if (flags & O_TRUNC) { 2007 VOP_UNLOCK(vp); /* XXX */ 2008 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2009 vattr_null(&va); 2010 va.va_size = 0; 2011 error = VOP_SETATTR(vp, &va, cred); 2012 if (error) 2013 goto bad; 2014 } 2015 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2016 goto bad; 2017 if (flags & FWRITE) { 2018 mutex_enter(vp->v_interlock); 2019 vp->v_writecount++; 2020 mutex_exit(vp->v_interlock); 2021 } 2022 2023 /* done with modified vn_open, now finish what sys_open does. */ 2024 if ((error = open_setfp(l, fp, vp, indx, flags))) 2025 return error; 2026 2027 VOP_UNLOCK(vp); 2028 *retval = indx; 2029 fd_affix(p, fp, indx); 2030 return (0); 2031 2032 bad: 2033 fd_abort(p, fp, indx); 2034 if (vp != NULL) 2035 vput(vp); 2036 return (error); 2037 } 2038 2039 int 2040 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2041 { 2042 /* { 2043 syscallarg(const void *) fhp; 2044 syscallarg(size_t) fh_size; 2045 syscallarg(int) flags; 2046 } */ 2047 2048 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2049 SCARG(uap, flags), retval); 2050 } 2051 2052 int 2053 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2054 { 2055 int error; 2056 fhandle_t *fh; 2057 struct vnode *vp; 2058 2059 /* 2060 * Must be super user 2061 */ 2062 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2063 0, NULL, NULL, NULL))) 2064 return (error); 2065 2066 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2067 if (error != 0) 2068 return error; 2069 2070 error = vfs_fhtovp(fh, &vp); 2071 vfs_copyinfh_free(fh); 2072 if (error != 0) 2073 return error; 2074 2075 error = vn_stat(vp, sb); 2076 vput(vp); 2077 return error; 2078 } 2079 2080 2081 /* ARGSUSED */ 2082 int 2083 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2084 { 2085 /* { 2086 syscallarg(const void *) fhp; 2087 syscallarg(size_t) fh_size; 2088 syscallarg(struct stat *) sb; 2089 } */ 2090 struct stat sb; 2091 int error; 2092 2093 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2094 if (error) 2095 return error; 2096 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2097 } 2098 2099 int 2100 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2101 int flags) 2102 { 2103 fhandle_t *fh; 2104 struct mount *mp; 2105 struct vnode *vp; 2106 int error; 2107 2108 /* 2109 * Must be super user 2110 */ 2111 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2112 0, NULL, NULL, NULL))) 2113 return error; 2114 2115 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2116 if (error != 0) 2117 return error; 2118 2119 error = vfs_fhtovp(fh, &vp); 2120 vfs_copyinfh_free(fh); 2121 if (error != 0) 2122 return error; 2123 2124 mp = vp->v_mount; 2125 error = dostatvfs(mp, sb, l, flags, 1); 2126 vput(vp); 2127 return error; 2128 } 2129 2130 /* ARGSUSED */ 2131 int 2132 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2133 { 2134 /* { 2135 syscallarg(const void *) fhp; 2136 syscallarg(size_t) fh_size; 2137 syscallarg(struct statvfs *) buf; 2138 syscallarg(int) flags; 2139 } */ 2140 struct statvfs *sb = STATVFSBUF_GET(); 2141 int error; 2142 2143 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2144 SCARG(uap, flags)); 2145 if (error == 0) 2146 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2147 STATVFSBUF_PUT(sb); 2148 return error; 2149 } 2150 2151 /* 2152 * Create a special file. 2153 */ 2154 /* ARGSUSED */ 2155 int 2156 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2157 register_t *retval) 2158 { 2159 /* { 2160 syscallarg(const char *) path; 2161 syscallarg(mode_t) mode; 2162 syscallarg(dev_t) dev; 2163 } */ 2164 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2165 SCARG(uap, dev), retval, UIO_USERSPACE); 2166 } 2167 2168 int 2169 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2170 register_t *retval) 2171 { 2172 /* { 2173 syscallarg(int) fd; 2174 syscallarg(const char *) path; 2175 syscallarg(mode_t) mode; 2176 syscallarg(int) pad; 2177 syscallarg(dev_t) dev; 2178 } */ 2179 2180 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2181 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2182 } 2183 2184 int 2185 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2186 register_t *retval, enum uio_seg seg) 2187 { 2188 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2189 } 2190 2191 int 2192 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2193 dev_t dev, register_t *retval, enum uio_seg seg) 2194 { 2195 struct proc *p = l->l_proc; 2196 struct vnode *vp; 2197 struct vattr vattr; 2198 int error, optype; 2199 struct pathbuf *pb; 2200 struct nameidata nd; 2201 const char *pathstring; 2202 2203 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2204 0, NULL, NULL, NULL)) != 0) 2205 return (error); 2206 2207 optype = VOP_MKNOD_DESCOFFSET; 2208 2209 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2210 if (error) { 2211 return error; 2212 } 2213 pathstring = pathbuf_stringcopy_get(pb); 2214 if (pathstring == NULL) { 2215 pathbuf_destroy(pb); 2216 return ENOMEM; 2217 } 2218 2219 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2220 2221 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2222 goto out; 2223 vp = nd.ni_vp; 2224 2225 if (vp != NULL) 2226 error = EEXIST; 2227 else { 2228 vattr_null(&vattr); 2229 /* We will read cwdi->cwdi_cmask unlocked. */ 2230 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2231 vattr.va_rdev = dev; 2232 2233 switch (mode & S_IFMT) { 2234 case S_IFMT: /* used by badsect to flag bad sectors */ 2235 vattr.va_type = VBAD; 2236 break; 2237 case S_IFCHR: 2238 vattr.va_type = VCHR; 2239 break; 2240 case S_IFBLK: 2241 vattr.va_type = VBLK; 2242 break; 2243 case S_IFWHT: 2244 optype = VOP_WHITEOUT_DESCOFFSET; 2245 break; 2246 case S_IFREG: 2247 #if NVERIEXEC > 0 2248 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2249 O_CREAT); 2250 #endif /* NVERIEXEC > 0 */ 2251 vattr.va_type = VREG; 2252 vattr.va_rdev = VNOVAL; 2253 optype = VOP_CREATE_DESCOFFSET; 2254 break; 2255 default: 2256 error = EINVAL; 2257 break; 2258 } 2259 2260 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET && 2261 vattr.va_rdev == VNOVAL) 2262 error = EINVAL; 2263 } 2264 2265 if (!error) { 2266 switch (optype) { 2267 case VOP_WHITEOUT_DESCOFFSET: 2268 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2269 if (error) 2270 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2271 vput(nd.ni_dvp); 2272 break; 2273 2274 case VOP_MKNOD_DESCOFFSET: 2275 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2276 &nd.ni_cnd, &vattr); 2277 if (error == 0) 2278 vrele(nd.ni_vp); 2279 vput(nd.ni_dvp); 2280 break; 2281 2282 case VOP_CREATE_DESCOFFSET: 2283 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2284 &nd.ni_cnd, &vattr); 2285 if (error == 0) 2286 vrele(nd.ni_vp); 2287 vput(nd.ni_dvp); 2288 break; 2289 } 2290 } else { 2291 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2292 if (nd.ni_dvp == vp) 2293 vrele(nd.ni_dvp); 2294 else 2295 vput(nd.ni_dvp); 2296 if (vp) 2297 vrele(vp); 2298 } 2299 out: 2300 pathbuf_stringcopy_put(pb, pathstring); 2301 pathbuf_destroy(pb); 2302 return (error); 2303 } 2304 2305 /* 2306 * Create a named pipe. 2307 */ 2308 /* ARGSUSED */ 2309 int 2310 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2311 { 2312 /* { 2313 syscallarg(const char *) path; 2314 syscallarg(int) mode; 2315 } */ 2316 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2317 } 2318 2319 int 2320 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2321 register_t *retval) 2322 { 2323 /* { 2324 syscallarg(int) fd; 2325 syscallarg(const char *) path; 2326 syscallarg(int) mode; 2327 } */ 2328 2329 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2330 SCARG(uap, mode)); 2331 } 2332 2333 static int 2334 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2335 { 2336 struct proc *p = l->l_proc; 2337 struct vattr vattr; 2338 int error; 2339 struct pathbuf *pb; 2340 struct nameidata nd; 2341 2342 error = pathbuf_copyin(path, &pb); 2343 if (error) { 2344 return error; 2345 } 2346 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2347 2348 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2349 pathbuf_destroy(pb); 2350 return error; 2351 } 2352 if (nd.ni_vp != NULL) { 2353 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2354 if (nd.ni_dvp == nd.ni_vp) 2355 vrele(nd.ni_dvp); 2356 else 2357 vput(nd.ni_dvp); 2358 vrele(nd.ni_vp); 2359 pathbuf_destroy(pb); 2360 return (EEXIST); 2361 } 2362 vattr_null(&vattr); 2363 vattr.va_type = VFIFO; 2364 /* We will read cwdi->cwdi_cmask unlocked. */ 2365 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2366 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2367 if (error == 0) 2368 vrele(nd.ni_vp); 2369 vput(nd.ni_dvp); 2370 pathbuf_destroy(pb); 2371 return (error); 2372 } 2373 2374 /* 2375 * Make a hard file link. 2376 */ 2377 /* ARGSUSED */ 2378 int 2379 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2380 const char *link, int follow, register_t *retval) 2381 { 2382 struct vnode *vp; 2383 struct pathbuf *linkpb; 2384 struct nameidata nd; 2385 namei_simple_flags_t ns_flags; 2386 int error; 2387 2388 if (follow & AT_SYMLINK_FOLLOW) 2389 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2390 else 2391 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2392 2393 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2394 if (error != 0) 2395 return (error); 2396 error = pathbuf_copyin(link, &linkpb); 2397 if (error) { 2398 goto out1; 2399 } 2400 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2401 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2402 goto out2; 2403 if (nd.ni_vp) { 2404 error = EEXIST; 2405 goto abortop; 2406 } 2407 /* Prevent hard links on directories. */ 2408 if (vp->v_type == VDIR) { 2409 error = EPERM; 2410 goto abortop; 2411 } 2412 /* Prevent cross-mount operation. */ 2413 if (nd.ni_dvp->v_mount != vp->v_mount) { 2414 error = EXDEV; 2415 goto abortop; 2416 } 2417 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2418 VOP_UNLOCK(nd.ni_dvp); 2419 vrele(nd.ni_dvp); 2420 out2: 2421 pathbuf_destroy(linkpb); 2422 out1: 2423 vrele(vp); 2424 return (error); 2425 abortop: 2426 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2427 if (nd.ni_dvp == nd.ni_vp) 2428 vrele(nd.ni_dvp); 2429 else 2430 vput(nd.ni_dvp); 2431 if (nd.ni_vp != NULL) 2432 vrele(nd.ni_vp); 2433 goto out2; 2434 } 2435 2436 int 2437 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2438 { 2439 /* { 2440 syscallarg(const char *) path; 2441 syscallarg(const char *) link; 2442 } */ 2443 const char *path = SCARG(uap, path); 2444 const char *link = SCARG(uap, link); 2445 2446 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2447 AT_SYMLINK_FOLLOW, retval); 2448 } 2449 2450 int 2451 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2452 register_t *retval) 2453 { 2454 /* { 2455 syscallarg(int) fd1; 2456 syscallarg(const char *) name1; 2457 syscallarg(int) fd2; 2458 syscallarg(const char *) name2; 2459 syscallarg(int) flags; 2460 } */ 2461 int fd1 = SCARG(uap, fd1); 2462 const char *name1 = SCARG(uap, name1); 2463 int fd2 = SCARG(uap, fd2); 2464 const char *name2 = SCARG(uap, name2); 2465 int follow; 2466 2467 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2468 2469 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2470 } 2471 2472 2473 int 2474 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2475 { 2476 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2477 } 2478 2479 static int 2480 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2481 const char *link, enum uio_seg seg) 2482 { 2483 struct proc *p = curproc; 2484 struct vattr vattr; 2485 char *path; 2486 int error; 2487 size_t len; 2488 struct pathbuf *linkpb; 2489 struct nameidata nd; 2490 2491 KASSERT(l != NULL || fdat == AT_FDCWD); 2492 2493 path = PNBUF_GET(); 2494 if (seg == UIO_USERSPACE) { 2495 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2496 goto out1; 2497 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2498 goto out1; 2499 } else { 2500 len = strlen(patharg) + 1; 2501 KASSERT(len <= MAXPATHLEN); 2502 memcpy(path, patharg, len); 2503 linkpb = pathbuf_create(link); 2504 if (linkpb == NULL) { 2505 error = ENOMEM; 2506 goto out1; 2507 } 2508 } 2509 ktrkuser("symlink-target", path, len - 1); 2510 2511 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2512 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2513 goto out2; 2514 if (nd.ni_vp) { 2515 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2516 if (nd.ni_dvp == nd.ni_vp) 2517 vrele(nd.ni_dvp); 2518 else 2519 vput(nd.ni_dvp); 2520 vrele(nd.ni_vp); 2521 error = EEXIST; 2522 goto out2; 2523 } 2524 vattr_null(&vattr); 2525 vattr.va_type = VLNK; 2526 /* We will read cwdi->cwdi_cmask unlocked. */ 2527 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2528 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2529 if (error == 0) 2530 vrele(nd.ni_vp); 2531 vput(nd.ni_dvp); 2532 out2: 2533 pathbuf_destroy(linkpb); 2534 out1: 2535 PNBUF_PUT(path); 2536 return (error); 2537 } 2538 2539 /* 2540 * Make a symbolic link. 2541 */ 2542 /* ARGSUSED */ 2543 int 2544 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2545 { 2546 /* { 2547 syscallarg(const char *) path; 2548 syscallarg(const char *) link; 2549 } */ 2550 2551 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2552 UIO_USERSPACE); 2553 } 2554 2555 int 2556 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2557 register_t *retval) 2558 { 2559 /* { 2560 syscallarg(const char *) path1; 2561 syscallarg(int) fd; 2562 syscallarg(const char *) path2; 2563 } */ 2564 2565 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2566 SCARG(uap, path2), UIO_USERSPACE); 2567 } 2568 2569 /* 2570 * Delete a whiteout from the filesystem. 2571 */ 2572 /* ARGSUSED */ 2573 int 2574 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2575 { 2576 /* { 2577 syscallarg(const char *) path; 2578 } */ 2579 int error; 2580 struct pathbuf *pb; 2581 struct nameidata nd; 2582 2583 error = pathbuf_copyin(SCARG(uap, path), &pb); 2584 if (error) { 2585 return error; 2586 } 2587 2588 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2589 error = namei(&nd); 2590 if (error) { 2591 pathbuf_destroy(pb); 2592 return (error); 2593 } 2594 2595 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2596 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2597 if (nd.ni_dvp == nd.ni_vp) 2598 vrele(nd.ni_dvp); 2599 else 2600 vput(nd.ni_dvp); 2601 if (nd.ni_vp) 2602 vrele(nd.ni_vp); 2603 pathbuf_destroy(pb); 2604 return (EEXIST); 2605 } 2606 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2607 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2608 vput(nd.ni_dvp); 2609 pathbuf_destroy(pb); 2610 return (error); 2611 } 2612 2613 /* 2614 * Delete a name from the filesystem. 2615 */ 2616 /* ARGSUSED */ 2617 int 2618 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2619 { 2620 /* { 2621 syscallarg(const char *) path; 2622 } */ 2623 2624 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2625 } 2626 2627 int 2628 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2629 register_t *retval) 2630 { 2631 /* { 2632 syscallarg(int) fd; 2633 syscallarg(const char *) path; 2634 syscallarg(int) flag; 2635 } */ 2636 2637 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2638 SCARG(uap, flag), UIO_USERSPACE); 2639 } 2640 2641 int 2642 do_sys_unlink(const char *arg, enum uio_seg seg) 2643 { 2644 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2645 } 2646 2647 static int 2648 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2649 enum uio_seg seg) 2650 { 2651 struct vnode *vp; 2652 int error; 2653 struct pathbuf *pb; 2654 struct nameidata nd; 2655 const char *pathstring; 2656 2657 KASSERT(l != NULL || fdat == AT_FDCWD); 2658 2659 error = pathbuf_maybe_copyin(arg, seg, &pb); 2660 if (error) { 2661 return error; 2662 } 2663 pathstring = pathbuf_stringcopy_get(pb); 2664 if (pathstring == NULL) { 2665 pathbuf_destroy(pb); 2666 return ENOMEM; 2667 } 2668 2669 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2670 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2671 goto out; 2672 vp = nd.ni_vp; 2673 2674 /* 2675 * The root of a mounted filesystem cannot be deleted. 2676 */ 2677 if ((vp->v_vflag & VV_ROOT) != 0) { 2678 error = EBUSY; 2679 goto abort; 2680 } 2681 2682 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2683 error = EBUSY; 2684 goto abort; 2685 } 2686 2687 /* 2688 * No rmdir "." please. 2689 */ 2690 if (nd.ni_dvp == vp) { 2691 error = EINVAL; 2692 goto abort; 2693 } 2694 2695 /* 2696 * AT_REMOVEDIR is required to remove a directory 2697 */ 2698 if (vp->v_type == VDIR) { 2699 if (!(flags & AT_REMOVEDIR)) { 2700 error = EPERM; 2701 goto abort; 2702 } else { 2703 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2704 vput(nd.ni_dvp); 2705 goto out; 2706 } 2707 } 2708 2709 /* 2710 * Starting here we only deal with non directories. 2711 */ 2712 if (flags & AT_REMOVEDIR) { 2713 error = ENOTDIR; 2714 goto abort; 2715 } 2716 2717 #if NVERIEXEC > 0 2718 /* Handle remove requests for veriexec entries. */ 2719 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2720 goto abort; 2721 } 2722 #endif /* NVERIEXEC > 0 */ 2723 2724 #ifdef FILEASSOC 2725 (void)fileassoc_file_delete(vp); 2726 #endif /* FILEASSOC */ 2727 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2728 vput(nd.ni_dvp); 2729 goto out; 2730 2731 abort: 2732 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2733 if (nd.ni_dvp == vp) 2734 vrele(nd.ni_dvp); 2735 else 2736 vput(nd.ni_dvp); 2737 vput(vp); 2738 2739 out: 2740 pathbuf_stringcopy_put(pb, pathstring); 2741 pathbuf_destroy(pb); 2742 return (error); 2743 } 2744 2745 /* 2746 * Reposition read/write file offset. 2747 */ 2748 int 2749 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2750 { 2751 /* { 2752 syscallarg(int) fd; 2753 syscallarg(int) pad; 2754 syscallarg(off_t) offset; 2755 syscallarg(int) whence; 2756 } */ 2757 kauth_cred_t cred = l->l_cred; 2758 file_t *fp; 2759 struct vnode *vp; 2760 struct vattr vattr; 2761 off_t newoff; 2762 int error, fd; 2763 2764 fd = SCARG(uap, fd); 2765 2766 if ((fp = fd_getfile(fd)) == NULL) 2767 return (EBADF); 2768 2769 vp = fp->f_vnode; 2770 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2771 error = ESPIPE; 2772 goto out; 2773 } 2774 2775 vn_lock(vp, LK_SHARED | LK_RETRY); 2776 2777 switch (SCARG(uap, whence)) { 2778 case SEEK_CUR: 2779 newoff = fp->f_offset + SCARG(uap, offset); 2780 break; 2781 case SEEK_END: 2782 error = VOP_GETATTR(vp, &vattr, cred); 2783 if (error) { 2784 VOP_UNLOCK(vp); 2785 goto out; 2786 } 2787 newoff = SCARG(uap, offset) + vattr.va_size; 2788 break; 2789 case SEEK_SET: 2790 newoff = SCARG(uap, offset); 2791 break; 2792 default: 2793 error = EINVAL; 2794 VOP_UNLOCK(vp); 2795 goto out; 2796 } 2797 VOP_UNLOCK(vp); 2798 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2799 *(off_t *)retval = fp->f_offset = newoff; 2800 } 2801 out: 2802 fd_putfile(fd); 2803 return (error); 2804 } 2805 2806 /* 2807 * Positional read system call. 2808 */ 2809 int 2810 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2811 { 2812 /* { 2813 syscallarg(int) fd; 2814 syscallarg(void *) buf; 2815 syscallarg(size_t) nbyte; 2816 syscallarg(off_t) offset; 2817 } */ 2818 file_t *fp; 2819 struct vnode *vp; 2820 off_t offset; 2821 int error, fd = SCARG(uap, fd); 2822 2823 if ((fp = fd_getfile(fd)) == NULL) 2824 return (EBADF); 2825 2826 if ((fp->f_flag & FREAD) == 0) { 2827 fd_putfile(fd); 2828 return (EBADF); 2829 } 2830 2831 vp = fp->f_vnode; 2832 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2833 error = ESPIPE; 2834 goto out; 2835 } 2836 2837 offset = SCARG(uap, offset); 2838 2839 /* 2840 * XXX This works because no file systems actually 2841 * XXX take any action on the seek operation. 2842 */ 2843 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2844 goto out; 2845 2846 /* dofileread() will unuse the descriptor for us */ 2847 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2848 &offset, 0, retval)); 2849 2850 out: 2851 fd_putfile(fd); 2852 return (error); 2853 } 2854 2855 /* 2856 * Positional scatter read system call. 2857 */ 2858 int 2859 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2860 { 2861 /* { 2862 syscallarg(int) fd; 2863 syscallarg(const struct iovec *) iovp; 2864 syscallarg(int) iovcnt; 2865 syscallarg(off_t) offset; 2866 } */ 2867 off_t offset = SCARG(uap, offset); 2868 2869 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2870 SCARG(uap, iovcnt), &offset, 0, retval); 2871 } 2872 2873 /* 2874 * Positional write system call. 2875 */ 2876 int 2877 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2878 { 2879 /* { 2880 syscallarg(int) fd; 2881 syscallarg(const void *) buf; 2882 syscallarg(size_t) nbyte; 2883 syscallarg(off_t) offset; 2884 } */ 2885 file_t *fp; 2886 struct vnode *vp; 2887 off_t offset; 2888 int error, fd = SCARG(uap, fd); 2889 2890 if ((fp = fd_getfile(fd)) == NULL) 2891 return (EBADF); 2892 2893 if ((fp->f_flag & FWRITE) == 0) { 2894 fd_putfile(fd); 2895 return (EBADF); 2896 } 2897 2898 vp = fp->f_vnode; 2899 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2900 error = ESPIPE; 2901 goto out; 2902 } 2903 2904 offset = SCARG(uap, offset); 2905 2906 /* 2907 * XXX This works because no file systems actually 2908 * XXX take any action on the seek operation. 2909 */ 2910 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2911 goto out; 2912 2913 /* dofilewrite() will unuse the descriptor for us */ 2914 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2915 &offset, 0, retval)); 2916 2917 out: 2918 fd_putfile(fd); 2919 return (error); 2920 } 2921 2922 /* 2923 * Positional gather write system call. 2924 */ 2925 int 2926 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2927 { 2928 /* { 2929 syscallarg(int) fd; 2930 syscallarg(const struct iovec *) iovp; 2931 syscallarg(int) iovcnt; 2932 syscallarg(off_t) offset; 2933 } */ 2934 off_t offset = SCARG(uap, offset); 2935 2936 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2937 SCARG(uap, iovcnt), &offset, 0, retval); 2938 } 2939 2940 /* 2941 * Check access permissions. 2942 */ 2943 int 2944 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2945 { 2946 /* { 2947 syscallarg(const char *) path; 2948 syscallarg(int) flags; 2949 } */ 2950 2951 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2952 SCARG(uap, flags), 0); 2953 } 2954 2955 int 2956 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2957 int mode, int flags) 2958 { 2959 kauth_cred_t cred; 2960 struct vnode *vp; 2961 int error, nd_flag, vmode; 2962 struct pathbuf *pb; 2963 struct nameidata nd; 2964 2965 CTASSERT(F_OK == 0); 2966 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2967 /* nonsense mode */ 2968 return EINVAL; 2969 } 2970 2971 nd_flag = FOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT; 2972 if (flags & AT_SYMLINK_NOFOLLOW) 2973 nd_flag &= ~FOLLOW; 2974 2975 error = pathbuf_copyin(path, &pb); 2976 if (error) 2977 return error; 2978 2979 NDINIT(&nd, LOOKUP, nd_flag, pb); 2980 2981 /* Override default credentials */ 2982 cred = kauth_cred_dup(l->l_cred); 2983 if (!(flags & AT_EACCESS)) { 2984 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2985 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2986 } 2987 nd.ni_cnd.cn_cred = cred; 2988 2989 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2990 pathbuf_destroy(pb); 2991 goto out; 2992 } 2993 vp = nd.ni_vp; 2994 pathbuf_destroy(pb); 2995 2996 /* Flags == 0 means only check for existence. */ 2997 if (mode) { 2998 vmode = 0; 2999 if (mode & R_OK) 3000 vmode |= VREAD; 3001 if (mode & W_OK) 3002 vmode |= VWRITE; 3003 if (mode & X_OK) 3004 vmode |= VEXEC; 3005 3006 error = VOP_ACCESS(vp, vmode, cred); 3007 if (!error && (vmode & VWRITE)) 3008 error = vn_writechk(vp); 3009 } 3010 vput(vp); 3011 out: 3012 kauth_cred_free(cred); 3013 return (error); 3014 } 3015 3016 int 3017 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3018 register_t *retval) 3019 { 3020 /* { 3021 syscallarg(int) fd; 3022 syscallarg(const char *) path; 3023 syscallarg(int) amode; 3024 syscallarg(int) flag; 3025 } */ 3026 3027 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3028 SCARG(uap, amode), SCARG(uap, flag)); 3029 } 3030 3031 /* 3032 * Common code for all sys_stat functions, including compat versions. 3033 */ 3034 int 3035 do_sys_stat(const char *userpath, unsigned int nd_flag, 3036 struct stat *sb) 3037 { 3038 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3039 } 3040 3041 int 3042 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3043 unsigned int nd_flag, struct stat *sb) 3044 { 3045 int error; 3046 struct pathbuf *pb; 3047 struct nameidata nd; 3048 3049 KASSERT(l != NULL || fdat == AT_FDCWD); 3050 3051 error = pathbuf_copyin(userpath, &pb); 3052 if (error) { 3053 return error; 3054 } 3055 3056 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3057 3058 error = fd_nameiat(l, fdat, &nd); 3059 if (error != 0) { 3060 pathbuf_destroy(pb); 3061 return error; 3062 } 3063 error = vn_stat(nd.ni_vp, sb); 3064 vput(nd.ni_vp); 3065 pathbuf_destroy(pb); 3066 return error; 3067 } 3068 3069 /* 3070 * Get file status; this version follows links. 3071 */ 3072 /* ARGSUSED */ 3073 int 3074 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3075 { 3076 /* { 3077 syscallarg(const char *) path; 3078 syscallarg(struct stat *) ub; 3079 } */ 3080 struct stat sb; 3081 int error; 3082 3083 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3084 if (error) 3085 return error; 3086 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3087 } 3088 3089 /* 3090 * Get file status; this version does not follow links. 3091 */ 3092 /* ARGSUSED */ 3093 int 3094 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3095 { 3096 /* { 3097 syscallarg(const char *) path; 3098 syscallarg(struct stat *) ub; 3099 } */ 3100 struct stat sb; 3101 int error; 3102 3103 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3104 if (error) 3105 return error; 3106 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3107 } 3108 3109 int 3110 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3111 register_t *retval) 3112 { 3113 /* { 3114 syscallarg(int) fd; 3115 syscallarg(const char *) path; 3116 syscallarg(struct stat *) buf; 3117 syscallarg(int) flag; 3118 } */ 3119 unsigned int nd_flag; 3120 struct stat sb; 3121 int error; 3122 3123 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3124 nd_flag = NOFOLLOW; 3125 else 3126 nd_flag = FOLLOW; 3127 3128 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3129 &sb); 3130 if (error) 3131 return error; 3132 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3133 } 3134 3135 /* 3136 * Get configurable pathname variables. 3137 */ 3138 /* ARGSUSED */ 3139 int 3140 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3141 { 3142 /* { 3143 syscallarg(const char *) path; 3144 syscallarg(int) name; 3145 } */ 3146 int error; 3147 struct pathbuf *pb; 3148 struct nameidata nd; 3149 3150 error = pathbuf_copyin(SCARG(uap, path), &pb); 3151 if (error) { 3152 return error; 3153 } 3154 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3155 if ((error = namei(&nd)) != 0) { 3156 pathbuf_destroy(pb); 3157 return (error); 3158 } 3159 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3160 vput(nd.ni_vp); 3161 pathbuf_destroy(pb); 3162 return (error); 3163 } 3164 3165 /* 3166 * Return target name of a symbolic link. 3167 */ 3168 /* ARGSUSED */ 3169 int 3170 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3171 register_t *retval) 3172 { 3173 /* { 3174 syscallarg(const char *) path; 3175 syscallarg(char *) buf; 3176 syscallarg(size_t) count; 3177 } */ 3178 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3179 SCARG(uap, buf), SCARG(uap, count), retval); 3180 } 3181 3182 static int 3183 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3184 size_t count, register_t *retval) 3185 { 3186 struct vnode *vp; 3187 struct iovec aiov; 3188 struct uio auio; 3189 int error; 3190 struct pathbuf *pb; 3191 struct nameidata nd; 3192 3193 error = pathbuf_copyin(path, &pb); 3194 if (error) { 3195 return error; 3196 } 3197 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT, pb); 3198 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3199 pathbuf_destroy(pb); 3200 return error; 3201 } 3202 vp = nd.ni_vp; 3203 pathbuf_destroy(pb); 3204 if (vp->v_type != VLNK) 3205 error = EINVAL; 3206 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3207 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3208 aiov.iov_base = buf; 3209 aiov.iov_len = count; 3210 auio.uio_iov = &aiov; 3211 auio.uio_iovcnt = 1; 3212 auio.uio_offset = 0; 3213 auio.uio_rw = UIO_READ; 3214 KASSERT(l == curlwp); 3215 auio.uio_vmspace = l->l_proc->p_vmspace; 3216 auio.uio_resid = count; 3217 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3218 *retval = count - auio.uio_resid; 3219 } 3220 vput(vp); 3221 return (error); 3222 } 3223 3224 int 3225 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3226 register_t *retval) 3227 { 3228 /* { 3229 syscallarg(int) fd; 3230 syscallarg(const char *) path; 3231 syscallarg(char *) buf; 3232 syscallarg(size_t) bufsize; 3233 } */ 3234 3235 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3236 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3237 } 3238 3239 /* 3240 * Change flags of a file given a path name. 3241 */ 3242 /* ARGSUSED */ 3243 int 3244 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3245 { 3246 /* { 3247 syscallarg(const char *) path; 3248 syscallarg(u_long) flags; 3249 } */ 3250 struct vnode *vp; 3251 int error; 3252 3253 error = namei_simple_user(SCARG(uap, path), 3254 NSM_FOLLOW_TRYEMULROOT, &vp); 3255 if (error != 0) 3256 return (error); 3257 error = change_flags(vp, SCARG(uap, flags), l); 3258 vput(vp); 3259 return (error); 3260 } 3261 3262 /* 3263 * Change flags of a file given a file descriptor. 3264 */ 3265 /* ARGSUSED */ 3266 int 3267 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3268 { 3269 /* { 3270 syscallarg(int) fd; 3271 syscallarg(u_long) flags; 3272 } */ 3273 struct vnode *vp; 3274 file_t *fp; 3275 int error; 3276 3277 /* fd_getvnode() will use the descriptor for us */ 3278 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3279 return (error); 3280 vp = fp->f_vnode; 3281 error = change_flags(vp, SCARG(uap, flags), l); 3282 VOP_UNLOCK(vp); 3283 fd_putfile(SCARG(uap, fd)); 3284 return (error); 3285 } 3286 3287 /* 3288 * Change flags of a file given a path name; this version does 3289 * not follow links. 3290 */ 3291 int 3292 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3293 { 3294 /* { 3295 syscallarg(const char *) path; 3296 syscallarg(u_long) flags; 3297 } */ 3298 struct vnode *vp; 3299 int error; 3300 3301 error = namei_simple_user(SCARG(uap, path), 3302 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3303 if (error != 0) 3304 return (error); 3305 error = change_flags(vp, SCARG(uap, flags), l); 3306 vput(vp); 3307 return (error); 3308 } 3309 3310 /* 3311 * Common routine to change flags of a file. 3312 */ 3313 int 3314 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3315 { 3316 struct vattr vattr; 3317 int error; 3318 3319 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3320 3321 vattr_null(&vattr); 3322 vattr.va_flags = flags; 3323 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3324 3325 return (error); 3326 } 3327 3328 /* 3329 * Change mode of a file given path name; this version follows links. 3330 */ 3331 /* ARGSUSED */ 3332 int 3333 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3334 { 3335 /* { 3336 syscallarg(const char *) path; 3337 syscallarg(int) mode; 3338 } */ 3339 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3340 SCARG(uap, mode), 0); 3341 } 3342 3343 int 3344 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3345 { 3346 int error; 3347 struct vnode *vp; 3348 namei_simple_flags_t ns_flag; 3349 3350 if (flags & AT_SYMLINK_NOFOLLOW) 3351 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3352 else 3353 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3354 3355 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3356 if (error != 0) 3357 return error; 3358 3359 error = change_mode(vp, mode, l); 3360 3361 vrele(vp); 3362 3363 return (error); 3364 } 3365 3366 /* 3367 * Change mode of a file given a file descriptor. 3368 */ 3369 /* ARGSUSED */ 3370 int 3371 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3372 { 3373 /* { 3374 syscallarg(int) fd; 3375 syscallarg(int) mode; 3376 } */ 3377 file_t *fp; 3378 int error; 3379 3380 /* fd_getvnode() will use the descriptor for us */ 3381 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3382 return (error); 3383 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3384 fd_putfile(SCARG(uap, fd)); 3385 return (error); 3386 } 3387 3388 int 3389 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3390 register_t *retval) 3391 { 3392 /* { 3393 syscallarg(int) fd; 3394 syscallarg(const char *) path; 3395 syscallarg(int) mode; 3396 syscallarg(int) flag; 3397 } */ 3398 3399 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3400 SCARG(uap, mode), SCARG(uap, flag)); 3401 } 3402 3403 /* 3404 * Change mode of a file given path name; this version does not follow links. 3405 */ 3406 /* ARGSUSED */ 3407 int 3408 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3409 { 3410 /* { 3411 syscallarg(const char *) path; 3412 syscallarg(int) mode; 3413 } */ 3414 int error; 3415 struct vnode *vp; 3416 3417 error = namei_simple_user(SCARG(uap, path), 3418 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3419 if (error != 0) 3420 return (error); 3421 3422 error = change_mode(vp, SCARG(uap, mode), l); 3423 3424 vrele(vp); 3425 return (error); 3426 } 3427 3428 /* 3429 * Common routine to set mode given a vnode. 3430 */ 3431 static int 3432 change_mode(struct vnode *vp, int mode, struct lwp *l) 3433 { 3434 struct vattr vattr; 3435 int error; 3436 3437 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3438 vattr_null(&vattr); 3439 vattr.va_mode = mode & ALLPERMS; 3440 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3441 VOP_UNLOCK(vp); 3442 return (error); 3443 } 3444 3445 /* 3446 * Set ownership given a path name; this version follows links. 3447 */ 3448 /* ARGSUSED */ 3449 int 3450 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3451 { 3452 /* { 3453 syscallarg(const char *) path; 3454 syscallarg(uid_t) uid; 3455 syscallarg(gid_t) gid; 3456 } */ 3457 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3458 SCARG(uap, gid), 0); 3459 } 3460 3461 int 3462 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3463 gid_t gid, int flags) 3464 { 3465 int error; 3466 struct vnode *vp; 3467 namei_simple_flags_t ns_flag; 3468 3469 if (flags & AT_SYMLINK_NOFOLLOW) 3470 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3471 else 3472 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3473 3474 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3475 if (error != 0) 3476 return error; 3477 3478 error = change_owner(vp, uid, gid, l, 0); 3479 3480 vrele(vp); 3481 3482 return (error); 3483 } 3484 3485 /* 3486 * Set ownership given a path name; this version follows links. 3487 * Provides POSIX semantics. 3488 */ 3489 /* ARGSUSED */ 3490 int 3491 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3492 { 3493 /* { 3494 syscallarg(const char *) path; 3495 syscallarg(uid_t) uid; 3496 syscallarg(gid_t) gid; 3497 } */ 3498 int error; 3499 struct vnode *vp; 3500 3501 error = namei_simple_user(SCARG(uap, path), 3502 NSM_FOLLOW_TRYEMULROOT, &vp); 3503 if (error != 0) 3504 return (error); 3505 3506 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3507 3508 vrele(vp); 3509 return (error); 3510 } 3511 3512 /* 3513 * Set ownership given a file descriptor. 3514 */ 3515 /* ARGSUSED */ 3516 int 3517 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3518 { 3519 /* { 3520 syscallarg(int) fd; 3521 syscallarg(uid_t) uid; 3522 syscallarg(gid_t) gid; 3523 } */ 3524 int error; 3525 file_t *fp; 3526 3527 /* fd_getvnode() will use the descriptor for us */ 3528 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3529 return (error); 3530 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3531 l, 0); 3532 fd_putfile(SCARG(uap, fd)); 3533 return (error); 3534 } 3535 3536 int 3537 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3538 register_t *retval) 3539 { 3540 /* { 3541 syscallarg(int) fd; 3542 syscallarg(const char *) path; 3543 syscallarg(uid_t) owner; 3544 syscallarg(gid_t) group; 3545 syscallarg(int) flag; 3546 } */ 3547 3548 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3549 SCARG(uap, owner), SCARG(uap, group), 3550 SCARG(uap, flag)); 3551 } 3552 3553 /* 3554 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3555 */ 3556 /* ARGSUSED */ 3557 int 3558 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3559 { 3560 /* { 3561 syscallarg(int) fd; 3562 syscallarg(uid_t) uid; 3563 syscallarg(gid_t) gid; 3564 } */ 3565 int error; 3566 file_t *fp; 3567 3568 /* fd_getvnode() will use the descriptor for us */ 3569 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3570 return (error); 3571 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3572 l, 1); 3573 fd_putfile(SCARG(uap, fd)); 3574 return (error); 3575 } 3576 3577 /* 3578 * Set ownership given a path name; this version does not follow links. 3579 */ 3580 /* ARGSUSED */ 3581 int 3582 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3583 { 3584 /* { 3585 syscallarg(const char *) path; 3586 syscallarg(uid_t) uid; 3587 syscallarg(gid_t) gid; 3588 } */ 3589 int error; 3590 struct vnode *vp; 3591 3592 error = namei_simple_user(SCARG(uap, path), 3593 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3594 if (error != 0) 3595 return (error); 3596 3597 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3598 3599 vrele(vp); 3600 return (error); 3601 } 3602 3603 /* 3604 * Set ownership given a path name; this version does not follow links. 3605 * Provides POSIX/XPG semantics. 3606 */ 3607 /* ARGSUSED */ 3608 int 3609 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3610 { 3611 /* { 3612 syscallarg(const char *) path; 3613 syscallarg(uid_t) uid; 3614 syscallarg(gid_t) gid; 3615 } */ 3616 int error; 3617 struct vnode *vp; 3618 3619 error = namei_simple_user(SCARG(uap, path), 3620 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3621 if (error != 0) 3622 return (error); 3623 3624 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3625 3626 vrele(vp); 3627 return (error); 3628 } 3629 3630 /* 3631 * Common routine to set ownership given a vnode. 3632 */ 3633 static int 3634 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3635 int posix_semantics) 3636 { 3637 struct vattr vattr; 3638 mode_t newmode; 3639 int error; 3640 3641 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3642 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3643 goto out; 3644 3645 #define CHANGED(x) ((int)(x) != -1) 3646 newmode = vattr.va_mode; 3647 if (posix_semantics) { 3648 /* 3649 * POSIX/XPG semantics: if the caller is not the super-user, 3650 * clear set-user-id and set-group-id bits. Both POSIX and 3651 * the XPG consider the behaviour for calls by the super-user 3652 * implementation-defined; we leave the set-user-id and set- 3653 * group-id settings intact in that case. 3654 */ 3655 if (vattr.va_mode & S_ISUID) { 3656 if (kauth_authorize_vnode(l->l_cred, 3657 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3658 newmode &= ~S_ISUID; 3659 } 3660 if (vattr.va_mode & S_ISGID) { 3661 if (kauth_authorize_vnode(l->l_cred, 3662 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3663 newmode &= ~S_ISGID; 3664 } 3665 } else { 3666 /* 3667 * NetBSD semantics: when changing owner and/or group, 3668 * clear the respective bit(s). 3669 */ 3670 if (CHANGED(uid)) 3671 newmode &= ~S_ISUID; 3672 if (CHANGED(gid)) 3673 newmode &= ~S_ISGID; 3674 } 3675 /* Update va_mode iff altered. */ 3676 if (vattr.va_mode == newmode) 3677 newmode = VNOVAL; 3678 3679 vattr_null(&vattr); 3680 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3681 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3682 vattr.va_mode = newmode; 3683 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3684 #undef CHANGED 3685 3686 out: 3687 VOP_UNLOCK(vp); 3688 return (error); 3689 } 3690 3691 /* 3692 * Set the access and modification times given a path name; this 3693 * version follows links. 3694 */ 3695 /* ARGSUSED */ 3696 int 3697 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3698 register_t *retval) 3699 { 3700 /* { 3701 syscallarg(const char *) path; 3702 syscallarg(const struct timeval *) tptr; 3703 } */ 3704 3705 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3706 SCARG(uap, tptr), UIO_USERSPACE); 3707 } 3708 3709 /* 3710 * Set the access and modification times given a file descriptor. 3711 */ 3712 /* ARGSUSED */ 3713 int 3714 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3715 register_t *retval) 3716 { 3717 /* { 3718 syscallarg(int) fd; 3719 syscallarg(const struct timeval *) tptr; 3720 } */ 3721 int error; 3722 file_t *fp; 3723 3724 /* fd_getvnode() will use the descriptor for us */ 3725 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3726 return (error); 3727 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3728 UIO_USERSPACE); 3729 fd_putfile(SCARG(uap, fd)); 3730 return (error); 3731 } 3732 3733 int 3734 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3735 register_t *retval) 3736 { 3737 /* { 3738 syscallarg(int) fd; 3739 syscallarg(const struct timespec *) tptr; 3740 } */ 3741 int error; 3742 file_t *fp; 3743 3744 /* fd_getvnode() will use the descriptor for us */ 3745 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3746 return (error); 3747 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3748 SCARG(uap, tptr), UIO_USERSPACE); 3749 fd_putfile(SCARG(uap, fd)); 3750 return (error); 3751 } 3752 3753 /* 3754 * Set the access and modification times given a path name; this 3755 * version does not follow links. 3756 */ 3757 int 3758 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3759 register_t *retval) 3760 { 3761 /* { 3762 syscallarg(const char *) path; 3763 syscallarg(const struct timeval *) tptr; 3764 } */ 3765 3766 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3767 SCARG(uap, tptr), UIO_USERSPACE); 3768 } 3769 3770 int 3771 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3772 register_t *retval) 3773 { 3774 /* { 3775 syscallarg(int) fd; 3776 syscallarg(const char *) path; 3777 syscallarg(const struct timespec *) tptr; 3778 syscallarg(int) flag; 3779 } */ 3780 int follow; 3781 const struct timespec *tptr; 3782 int error; 3783 3784 tptr = SCARG(uap, tptr); 3785 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3786 3787 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3788 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3789 3790 return error; 3791 } 3792 3793 /* 3794 * Common routine to set access and modification times given a vnode. 3795 */ 3796 int 3797 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3798 const struct timespec *tptr, enum uio_seg seg) 3799 { 3800 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3801 } 3802 3803 int 3804 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3805 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3806 { 3807 struct vattr vattr; 3808 int error, dorele = 0; 3809 namei_simple_flags_t sflags; 3810 bool vanull, setbirthtime; 3811 struct timespec ts[2]; 3812 3813 KASSERT(l != NULL || fdat == AT_FDCWD); 3814 3815 /* 3816 * I have checked all callers and they pass either FOLLOW, 3817 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3818 * is 0. More to the point, they don't pass anything else. 3819 * Let's keep it that way at least until the namei interfaces 3820 * are fully sanitized. 3821 */ 3822 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3823 sflags = (flag == FOLLOW) ? 3824 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3825 3826 if (tptr == NULL) { 3827 vanull = true; 3828 nanotime(&ts[0]); 3829 ts[1] = ts[0]; 3830 } else { 3831 vanull = false; 3832 if (seg != UIO_SYSSPACE) { 3833 error = copyin(tptr, ts, sizeof (ts)); 3834 if (error != 0) 3835 return error; 3836 } else { 3837 ts[0] = tptr[0]; 3838 ts[1] = tptr[1]; 3839 } 3840 } 3841 3842 if (ts[0].tv_nsec == UTIME_NOW) { 3843 nanotime(&ts[0]); 3844 if (ts[1].tv_nsec == UTIME_NOW) { 3845 vanull = true; 3846 ts[1] = ts[0]; 3847 } 3848 } else if (ts[1].tv_nsec == UTIME_NOW) 3849 nanotime(&ts[1]); 3850 3851 if (vp == NULL) { 3852 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3853 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3854 if (error != 0) 3855 return error; 3856 dorele = 1; 3857 } 3858 3859 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3860 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3861 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3862 vattr_null(&vattr); 3863 3864 if (ts[0].tv_nsec != UTIME_OMIT) 3865 vattr.va_atime = ts[0]; 3866 3867 if (ts[1].tv_nsec != UTIME_OMIT) { 3868 vattr.va_mtime = ts[1]; 3869 if (setbirthtime) 3870 vattr.va_birthtime = ts[1]; 3871 } 3872 3873 if (vanull) 3874 vattr.va_vaflags |= VA_UTIMES_NULL; 3875 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3876 VOP_UNLOCK(vp); 3877 3878 if (dorele != 0) 3879 vrele(vp); 3880 3881 return error; 3882 } 3883 3884 int 3885 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3886 const struct timeval *tptr, enum uio_seg seg) 3887 { 3888 struct timespec ts[2]; 3889 struct timespec *tsptr = NULL; 3890 int error; 3891 3892 if (tptr != NULL) { 3893 struct timeval tv[2]; 3894 3895 if (seg != UIO_SYSSPACE) { 3896 error = copyin(tptr, tv, sizeof (tv)); 3897 if (error != 0) 3898 return error; 3899 tptr = tv; 3900 } 3901 3902 if ((tv[0].tv_usec == UTIME_NOW) || 3903 (tv[0].tv_usec == UTIME_OMIT)) 3904 ts[0].tv_nsec = tv[0].tv_usec; 3905 else 3906 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3907 3908 if ((tv[1].tv_usec == UTIME_NOW) || 3909 (tv[1].tv_usec == UTIME_OMIT)) 3910 ts[1].tv_nsec = tv[1].tv_usec; 3911 else 3912 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3913 3914 tsptr = &ts[0]; 3915 } 3916 3917 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3918 } 3919 3920 /* 3921 * Truncate a file given its path name. 3922 */ 3923 /* ARGSUSED */ 3924 int 3925 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3926 { 3927 /* { 3928 syscallarg(const char *) path; 3929 syscallarg(int) pad; 3930 syscallarg(off_t) length; 3931 } */ 3932 struct vnode *vp; 3933 struct vattr vattr; 3934 int error; 3935 3936 if (SCARG(uap, length) < 0) 3937 return EINVAL; 3938 3939 error = namei_simple_user(SCARG(uap, path), 3940 NSM_FOLLOW_TRYEMULROOT, &vp); 3941 if (error != 0) 3942 return (error); 3943 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3944 if (vp->v_type == VDIR) 3945 error = EISDIR; 3946 else if ((error = vn_writechk(vp)) == 0 && 3947 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3948 vattr_null(&vattr); 3949 vattr.va_size = SCARG(uap, length); 3950 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3951 } 3952 vput(vp); 3953 return (error); 3954 } 3955 3956 /* 3957 * Truncate a file given a file descriptor. 3958 */ 3959 /* ARGSUSED */ 3960 int 3961 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3962 { 3963 /* { 3964 syscallarg(int) fd; 3965 syscallarg(int) pad; 3966 syscallarg(off_t) length; 3967 } */ 3968 struct vattr vattr; 3969 struct vnode *vp; 3970 file_t *fp; 3971 int error; 3972 3973 if (SCARG(uap, length) < 0) 3974 return EINVAL; 3975 3976 /* fd_getvnode() will use the descriptor for us */ 3977 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3978 return (error); 3979 if ((fp->f_flag & FWRITE) == 0) { 3980 error = EINVAL; 3981 goto out; 3982 } 3983 vp = fp->f_vnode; 3984 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3985 if (vp->v_type == VDIR) 3986 error = EISDIR; 3987 else if ((error = vn_writechk(vp)) == 0) { 3988 vattr_null(&vattr); 3989 vattr.va_size = SCARG(uap, length); 3990 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3991 } 3992 VOP_UNLOCK(vp); 3993 out: 3994 fd_putfile(SCARG(uap, fd)); 3995 return (error); 3996 } 3997 3998 /* 3999 * Sync an open file. 4000 */ 4001 /* ARGSUSED */ 4002 int 4003 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 4004 { 4005 /* { 4006 syscallarg(int) fd; 4007 } */ 4008 struct vnode *vp; 4009 file_t *fp; 4010 int error; 4011 4012 /* fd_getvnode() will use the descriptor for us */ 4013 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4014 return (error); 4015 vp = fp->f_vnode; 4016 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4017 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4018 VOP_UNLOCK(vp); 4019 fd_putfile(SCARG(uap, fd)); 4020 return (error); 4021 } 4022 4023 /* 4024 * Sync a range of file data. API modeled after that found in AIX. 4025 * 4026 * FDATASYNC indicates that we need only save enough metadata to be able 4027 * to re-read the written data. 4028 */ 4029 /* ARGSUSED */ 4030 int 4031 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4032 { 4033 /* { 4034 syscallarg(int) fd; 4035 syscallarg(int) flags; 4036 syscallarg(off_t) start; 4037 syscallarg(off_t) length; 4038 } */ 4039 struct vnode *vp; 4040 file_t *fp; 4041 int flags, nflags; 4042 off_t s, e, len; 4043 int error; 4044 4045 /* fd_getvnode() will use the descriptor for us */ 4046 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4047 return (error); 4048 4049 if ((fp->f_flag & FWRITE) == 0) { 4050 error = EBADF; 4051 goto out; 4052 } 4053 4054 flags = SCARG(uap, flags); 4055 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4056 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4057 error = EINVAL; 4058 goto out; 4059 } 4060 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4061 if (flags & FDATASYNC) 4062 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4063 else 4064 nflags = FSYNC_WAIT; 4065 if (flags & FDISKSYNC) 4066 nflags |= FSYNC_CACHE; 4067 4068 len = SCARG(uap, length); 4069 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4070 if (len) { 4071 s = SCARG(uap, start); 4072 e = s + len; 4073 if (e < s) { 4074 error = EINVAL; 4075 goto out; 4076 } 4077 } else { 4078 e = 0; 4079 s = 0; 4080 } 4081 4082 vp = fp->f_vnode; 4083 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4084 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4085 VOP_UNLOCK(vp); 4086 out: 4087 fd_putfile(SCARG(uap, fd)); 4088 return (error); 4089 } 4090 4091 /* 4092 * Sync the data of an open file. 4093 */ 4094 /* ARGSUSED */ 4095 int 4096 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4097 { 4098 /* { 4099 syscallarg(int) fd; 4100 } */ 4101 struct vnode *vp; 4102 file_t *fp; 4103 int error; 4104 4105 /* fd_getvnode() will use the descriptor for us */ 4106 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4107 return (error); 4108 vp = fp->f_vnode; 4109 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4110 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4111 VOP_UNLOCK(vp); 4112 fd_putfile(SCARG(uap, fd)); 4113 return (error); 4114 } 4115 4116 /* 4117 * Rename files, (standard) BSD semantics frontend. 4118 */ 4119 /* ARGSUSED */ 4120 int 4121 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4122 { 4123 /* { 4124 syscallarg(const char *) from; 4125 syscallarg(const char *) to; 4126 } */ 4127 4128 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4129 SCARG(uap, to), UIO_USERSPACE, 0)); 4130 } 4131 4132 int 4133 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4134 register_t *retval) 4135 { 4136 /* { 4137 syscallarg(int) fromfd; 4138 syscallarg(const char *) from; 4139 syscallarg(int) tofd; 4140 syscallarg(const char *) to; 4141 } */ 4142 4143 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4144 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4145 } 4146 4147 /* 4148 * Rename files, POSIX semantics frontend. 4149 */ 4150 /* ARGSUSED */ 4151 int 4152 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4153 { 4154 /* { 4155 syscallarg(const char *) from; 4156 syscallarg(const char *) to; 4157 } */ 4158 4159 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4160 SCARG(uap, to), UIO_USERSPACE, 1)); 4161 } 4162 4163 /* 4164 * Rename files. Source and destination must either both be directories, 4165 * or both not be directories. If target is a directory, it must be empty. 4166 * If `from' and `to' refer to the same object, the value of the `retain' 4167 * argument is used to determine whether `from' will be 4168 * 4169 * (retain == 0) deleted unless `from' and `to' refer to the same 4170 * object in the file system's name space (BSD). 4171 * (retain == 1) always retained (POSIX). 4172 * 4173 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4174 */ 4175 int 4176 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4177 { 4178 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4179 } 4180 4181 static int 4182 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4183 const char *to, enum uio_seg seg, int retain) 4184 { 4185 struct pathbuf *fpb, *tpb; 4186 struct nameidata fnd, tnd; 4187 struct vnode *fdvp, *fvp; 4188 struct vnode *tdvp, *tvp; 4189 struct mount *mp, *tmp; 4190 int error; 4191 4192 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4193 4194 error = pathbuf_maybe_copyin(from, seg, &fpb); 4195 if (error) 4196 goto out0; 4197 KASSERT(fpb != NULL); 4198 4199 error = pathbuf_maybe_copyin(to, seg, &tpb); 4200 if (error) 4201 goto out1; 4202 KASSERT(tpb != NULL); 4203 4204 /* 4205 * Lookup from. 4206 * 4207 * XXX LOCKPARENT is wrong because we don't actually want it 4208 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4209 * insane, so for the time being we need to leave it like this. 4210 */ 4211 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4212 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4213 goto out2; 4214 4215 /* 4216 * Pull out the important results of the lookup, fdvp and fvp. 4217 * Of course, fvp is bogus because we're about to unlock fdvp. 4218 */ 4219 fdvp = fnd.ni_dvp; 4220 fvp = fnd.ni_vp; 4221 mp = fdvp->v_mount; 4222 KASSERT(fdvp != NULL); 4223 KASSERT(fvp != NULL); 4224 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4225 /* 4226 * Bracket the operation with fstrans_start()/fstrans_done(). 4227 * 4228 * Inside the bracket this file system cannot be unmounted so 4229 * a vnode on this file system cannot change its v_mount. 4230 * A vnode on another file system may still change to dead mount. 4231 */ 4232 fstrans_start(mp); 4233 4234 /* 4235 * Make sure neither fdvp nor fvp is locked. 4236 */ 4237 if (fdvp != fvp) 4238 VOP_UNLOCK(fdvp); 4239 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4240 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4241 4242 /* 4243 * Reject renaming `.' and `..'. Can't do this until after 4244 * namei because we need namei's parsing to find the final 4245 * component name. (namei should just leave us with the final 4246 * component name and not look it up itself, but anyway...) 4247 * 4248 * This was here before because we used to relookup from 4249 * instead of to and relookup requires the caller to check 4250 * this, but now file systems may depend on this check, so we 4251 * must retain it until the file systems are all rototilled. 4252 */ 4253 if (((fnd.ni_cnd.cn_namelen == 1) && 4254 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4255 ((fnd.ni_cnd.cn_namelen == 2) && 4256 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4257 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4258 error = EINVAL; /* XXX EISDIR? */ 4259 goto abort0; 4260 } 4261 4262 /* 4263 * Lookup to. 4264 * 4265 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4266 * fvp here to decide whether to add CREATEDIR is a load of 4267 * bollocks because fvp might be the wrong node by now, since 4268 * fdvp is unlocked. 4269 * 4270 * XXX Why not pass CREATEDIR always? 4271 */ 4272 NDINIT(&tnd, RENAME, 4273 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4274 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4275 tpb); 4276 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4277 goto abort0; 4278 4279 /* 4280 * Pull out the important results of the lookup, tdvp and tvp. 4281 * Of course, tvp is bogus because we're about to unlock tdvp. 4282 */ 4283 tdvp = tnd.ni_dvp; 4284 tvp = tnd.ni_vp; 4285 KASSERT(tdvp != NULL); 4286 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4287 4288 /* 4289 * Make sure neither tdvp nor tvp is locked. 4290 */ 4291 if (tdvp != tvp) 4292 VOP_UNLOCK(tdvp); 4293 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4294 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4295 4296 /* 4297 * Reject renaming onto `.' or `..'. relookup is unhappy with 4298 * these, which is why we must do this here. Once upon a time 4299 * we relooked up from instead of to, and consequently didn't 4300 * need this check, but now that we relookup to instead of 4301 * from, we need this; and we shall need it forever forward 4302 * until the VOP_RENAME protocol changes, because file systems 4303 * will no doubt begin to depend on this check. 4304 */ 4305 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4306 error = EISDIR; 4307 goto abort1; 4308 } 4309 if ((tnd.ni_cnd.cn_namelen == 2) && 4310 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4311 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4312 error = EINVAL; 4313 goto abort1; 4314 } 4315 4316 /* 4317 * Make sure the mount points match. Although we don't hold 4318 * any vnode locks, the v_mount on fdvp file system are stable. 4319 * 4320 * Unmounting another file system at an inopportune moment may 4321 * cause tdvp to disappear and change its v_mount to dead. 4322 * 4323 * So in either case different v_mount means cross-device rename. 4324 */ 4325 KASSERT(mp != NULL); 4326 tmp = tdvp->v_mount; 4327 4328 if (mp != tmp) { 4329 error = EXDEV; 4330 goto abort1; 4331 } 4332 4333 /* 4334 * Take the vfs rename lock to avoid cross-directory screw cases. 4335 * Nothing is locked currently, so taking this lock is safe. 4336 */ 4337 error = VFS_RENAMELOCK_ENTER(mp); 4338 if (error) 4339 goto abort1; 4340 4341 /* 4342 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4343 * and nothing is locked except for the vfs rename lock. 4344 * 4345 * The next step is a little rain dance to conform to the 4346 * insane lock protocol, even though it does nothing to ward 4347 * off race conditions. 4348 * 4349 * We need tdvp and tvp to be locked. However, because we have 4350 * unlocked tdvp in order to hold no locks while we take the 4351 * vfs rename lock, tvp may be wrong here, and we can't safely 4352 * lock it even if the sensible file systems will just unlock 4353 * it straight away. Consequently, we must lock tdvp and then 4354 * relookup tvp to get it locked. 4355 * 4356 * Finally, because the VOP_RENAME protocol is brain-damaged 4357 * and various file systems insanely depend on the semantics of 4358 * this brain damage, the lookup of to must be the last lookup 4359 * before VOP_RENAME. 4360 */ 4361 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4362 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4363 if (error) 4364 goto abort2; 4365 4366 /* 4367 * Drop the old tvp and pick up the new one -- which might be 4368 * the same, but that doesn't matter to us. After this, tdvp 4369 * and tvp should both be locked. 4370 */ 4371 if (tvp != NULL) 4372 vrele(tvp); 4373 tvp = tnd.ni_vp; 4374 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4375 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4376 4377 /* 4378 * The old do_sys_rename had various consistency checks here 4379 * involving fvp and tvp. fvp is bogus already here, and tvp 4380 * will become bogus soon in any sensible file system, so the 4381 * only purpose in putting these checks here is to give lip 4382 * service to these screw cases and to acknowledge that they 4383 * exist, not actually to handle them, but here you go 4384 * anyway... 4385 */ 4386 4387 /* 4388 * Acknowledge that directories and non-directories aren't 4389 * suposed to mix. 4390 */ 4391 if (tvp != NULL) { 4392 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4393 error = ENOTDIR; 4394 goto abort3; 4395 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4396 error = EISDIR; 4397 goto abort3; 4398 } 4399 } 4400 4401 /* 4402 * Acknowledge some random screw case, among the dozens that 4403 * might arise. 4404 */ 4405 if (fvp == tdvp) { 4406 error = EINVAL; 4407 goto abort3; 4408 } 4409 4410 /* 4411 * Acknowledge that POSIX has a wacky screw case. 4412 * 4413 * XXX Eventually the retain flag needs to be passed on to 4414 * VOP_RENAME. 4415 */ 4416 if (fvp == tvp) { 4417 if (retain) { 4418 error = 0; 4419 goto abort3; 4420 } else if ((fdvp == tdvp) && 4421 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4422 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4423 fnd.ni_cnd.cn_namelen))) { 4424 error = 0; 4425 goto abort3; 4426 } 4427 } 4428 4429 /* 4430 * Make sure veriexec can screw us up. (But a race can screw 4431 * up veriexec, of course -- remember, fvp and (soon) tvp are 4432 * bogus.) 4433 */ 4434 #if NVERIEXEC > 0 4435 { 4436 char *f1, *f2; 4437 size_t f1_len; 4438 size_t f2_len; 4439 4440 f1_len = fnd.ni_cnd.cn_namelen + 1; 4441 f1 = kmem_alloc(f1_len, KM_SLEEP); 4442 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4443 4444 f2_len = tnd.ni_cnd.cn_namelen + 1; 4445 f2 = kmem_alloc(f2_len, KM_SLEEP); 4446 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4447 4448 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4449 4450 kmem_free(f1, f1_len); 4451 kmem_free(f2, f2_len); 4452 4453 if (error) 4454 goto abort3; 4455 } 4456 #endif /* NVERIEXEC > 0 */ 4457 4458 /* 4459 * All ready. Incant the rename vop. 4460 */ 4461 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4462 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4463 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4464 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4465 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4466 4467 /* 4468 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4469 * tdvp and tvp. But we can't assert any of that. 4470 */ 4471 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4472 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4473 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4474 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4475 4476 /* 4477 * So all we have left to do is to drop the rename lock and 4478 * destroy the pathbufs. 4479 */ 4480 VFS_RENAMELOCK_EXIT(mp); 4481 fstrans_done(mp); 4482 goto out2; 4483 4484 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4485 VOP_UNLOCK(tvp); 4486 abort2: VOP_UNLOCK(tdvp); 4487 VFS_RENAMELOCK_EXIT(mp); 4488 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4489 vrele(tdvp); 4490 if (tvp != NULL) 4491 vrele(tvp); 4492 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4493 vrele(fdvp); 4494 vrele(fvp); 4495 fstrans_done(mp); 4496 out2: pathbuf_destroy(tpb); 4497 out1: pathbuf_destroy(fpb); 4498 out0: return error; 4499 } 4500 4501 /* 4502 * Make a directory file. 4503 */ 4504 /* ARGSUSED */ 4505 int 4506 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4507 { 4508 /* { 4509 syscallarg(const char *) path; 4510 syscallarg(int) mode; 4511 } */ 4512 4513 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4514 SCARG(uap, mode), UIO_USERSPACE); 4515 } 4516 4517 int 4518 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4519 register_t *retval) 4520 { 4521 /* { 4522 syscallarg(int) fd; 4523 syscallarg(const char *) path; 4524 syscallarg(int) mode; 4525 } */ 4526 4527 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4528 SCARG(uap, mode), UIO_USERSPACE); 4529 } 4530 4531 4532 int 4533 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4534 { 4535 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, seg); 4536 } 4537 4538 static int 4539 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4540 enum uio_seg seg) 4541 { 4542 struct proc *p = curlwp->l_proc; 4543 struct vnode *vp; 4544 struct vattr vattr; 4545 int error; 4546 struct pathbuf *pb; 4547 struct nameidata nd; 4548 4549 KASSERT(l != NULL || fdat == AT_FDCWD); 4550 4551 /* XXX bollocks, should pass in a pathbuf */ 4552 error = pathbuf_maybe_copyin(path, seg, &pb); 4553 if (error) { 4554 return error; 4555 } 4556 4557 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4558 4559 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4560 pathbuf_destroy(pb); 4561 return (error); 4562 } 4563 vp = nd.ni_vp; 4564 if (vp != NULL) { 4565 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4566 if (nd.ni_dvp == vp) 4567 vrele(nd.ni_dvp); 4568 else 4569 vput(nd.ni_dvp); 4570 vrele(vp); 4571 pathbuf_destroy(pb); 4572 return (EEXIST); 4573 } 4574 vattr_null(&vattr); 4575 vattr.va_type = VDIR; 4576 /* We will read cwdi->cwdi_cmask unlocked. */ 4577 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4578 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4579 if (!error) 4580 vrele(nd.ni_vp); 4581 vput(nd.ni_dvp); 4582 pathbuf_destroy(pb); 4583 return (error); 4584 } 4585 4586 /* 4587 * Remove a directory file. 4588 */ 4589 /* ARGSUSED */ 4590 int 4591 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4592 { 4593 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4594 AT_REMOVEDIR, UIO_USERSPACE); 4595 } 4596 4597 /* 4598 * Read a block of directory entries in a file system independent format. 4599 */ 4600 int 4601 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4602 { 4603 /* { 4604 syscallarg(int) fd; 4605 syscallarg(char *) buf; 4606 syscallarg(size_t) count; 4607 } */ 4608 file_t *fp; 4609 int error, done; 4610 4611 /* fd_getvnode() will use the descriptor for us */ 4612 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4613 return (error); 4614 if ((fp->f_flag & FREAD) == 0) { 4615 error = EBADF; 4616 goto out; 4617 } 4618 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4619 SCARG(uap, count), &done, l, 0, 0); 4620 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4621 *retval = done; 4622 out: 4623 fd_putfile(SCARG(uap, fd)); 4624 return (error); 4625 } 4626 4627 /* 4628 * Set the mode mask for creation of filesystem nodes. 4629 */ 4630 int 4631 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4632 { 4633 /* { 4634 syscallarg(mode_t) newmask; 4635 } */ 4636 4637 /* 4638 * cwdi->cwdi_cmask will be read unlocked elsewhere, and no kind of 4639 * serialization with those reads is required. It's important to 4640 * return a coherent answer for the caller of umask() though, and 4641 * the atomic operation accomplishes that. 4642 */ 4643 *retval = atomic_swap_uint(&curproc->p_cwdi->cwdi_cmask, 4644 SCARG(uap, newmask) & ALLPERMS); 4645 4646 return (0); 4647 } 4648 4649 int 4650 dorevoke(struct vnode *vp, kauth_cred_t cred) 4651 { 4652 struct vattr vattr; 4653 int error, fs_decision; 4654 4655 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4656 error = VOP_GETATTR(vp, &vattr, cred); 4657 VOP_UNLOCK(vp); 4658 if (error != 0) 4659 return error; 4660 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4661 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4662 fs_decision); 4663 if (!error) 4664 VOP_REVOKE(vp, REVOKEALL); 4665 return (error); 4666 } 4667 4668 /* 4669 * Void all references to file by ripping underlying filesystem 4670 * away from vnode. 4671 */ 4672 /* ARGSUSED */ 4673 int 4674 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4675 { 4676 /* { 4677 syscallarg(const char *) path; 4678 } */ 4679 struct vnode *vp; 4680 int error; 4681 4682 error = namei_simple_user(SCARG(uap, path), 4683 NSM_FOLLOW_TRYEMULROOT, &vp); 4684 if (error != 0) 4685 return (error); 4686 error = dorevoke(vp, l->l_cred); 4687 vrele(vp); 4688 return (error); 4689 } 4690 4691 /* 4692 * Allocate backing store for a file, filling a hole without having to 4693 * explicitly write anything out. 4694 */ 4695 /* ARGSUSED */ 4696 int 4697 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4698 register_t *retval) 4699 { 4700 /* { 4701 syscallarg(int) fd; 4702 syscallarg(off_t) pos; 4703 syscallarg(off_t) len; 4704 } */ 4705 int fd; 4706 off_t pos, len; 4707 struct file *fp; 4708 struct vnode *vp; 4709 int error; 4710 4711 fd = SCARG(uap, fd); 4712 pos = SCARG(uap, pos); 4713 len = SCARG(uap, len); 4714 4715 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4716 *retval = EINVAL; 4717 return 0; 4718 } 4719 4720 error = fd_getvnode(fd, &fp); 4721 if (error) { 4722 *retval = error; 4723 return 0; 4724 } 4725 if ((fp->f_flag & FWRITE) == 0) { 4726 error = EBADF; 4727 goto fail; 4728 } 4729 vp = fp->f_vnode; 4730 4731 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4732 if (vp->v_type == VDIR) { 4733 error = EISDIR; 4734 } else { 4735 error = VOP_FALLOCATE(vp, pos, len); 4736 } 4737 VOP_UNLOCK(vp); 4738 4739 fail: 4740 fd_putfile(fd); 4741 *retval = error; 4742 return 0; 4743 } 4744 4745 /* 4746 * Deallocate backing store for a file, creating a hole. Also used for 4747 * invoking TRIM on disks. 4748 */ 4749 /* ARGSUSED */ 4750 int 4751 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4752 register_t *retval) 4753 { 4754 /* { 4755 syscallarg(int) fd; 4756 syscallarg(off_t) pos; 4757 syscallarg(off_t) len; 4758 } */ 4759 int fd; 4760 off_t pos, len; 4761 struct file *fp; 4762 struct vnode *vp; 4763 int error; 4764 4765 fd = SCARG(uap, fd); 4766 pos = SCARG(uap, pos); 4767 len = SCARG(uap, len); 4768 4769 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4770 return EINVAL; 4771 } 4772 4773 error = fd_getvnode(fd, &fp); 4774 if (error) { 4775 return error; 4776 } 4777 if ((fp->f_flag & FWRITE) == 0) { 4778 error = EBADF; 4779 goto fail; 4780 } 4781 vp = fp->f_vnode; 4782 4783 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4784 if (vp->v_type == VDIR) { 4785 error = EISDIR; 4786 } else { 4787 error = VOP_FDISCARD(vp, pos, len); 4788 } 4789 VOP_UNLOCK(vp); 4790 4791 fail: 4792 fd_putfile(fd); 4793 return error; 4794 } 4795
Indexes created Fri Oct 03 02:09:57 GMT 2025