vfs_syscalls.c revision 1.518.2.4
1 /* $NetBSD: vfs_syscalls.c,v 1.518.2.4 2019/01/14 13:34:28 pgoyette Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 /* 69 * Virtual File System System Calls 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.518.2.4 2019/01/14 13:34:28 pgoyette Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/fcntl.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/fstrans.h> 91 #include <sys/proc.h> 92 #include <sys/uio.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/quota.h> 99 #include <sys/quotactl.h> 100 #include <sys/ktrace.h> 101 #ifdef FILEASSOC 102 #include <sys/fileassoc.h> 103 #endif /* FILEASSOC */ 104 #include <sys/extattr.h> 105 #include <sys/verified_exec.h> 106 #include <sys/kauth.h> 107 #include <sys/atomic.h> 108 #include <sys/module.h> 109 #include <sys/buf.h> 110 #include <sys/event.h> 111 #include <sys/compat_stub.h> 112 113 #include <miscfs/genfs/genfs.h> 114 #include <miscfs/specfs/specdev.h> 115 116 #include <nfs/rpcv2.h> 117 #include <nfs/nfsproto.h> 118 #include <nfs/nfs.h> 119 #include <nfs/nfs_var.h> 120 121 /* XXX this shouldn't be here */ 122 #ifndef OFF_T_MAX 123 #define OFF_T_MAX __type_max(off_t) 124 #endif 125 126 static int change_flags(struct vnode *, u_long, struct lwp *); 127 static int change_mode(struct vnode *, int, struct lwp *); 128 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 129 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 130 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 131 enum uio_seg); 132 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 133 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 134 enum uio_seg); 135 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 136 enum uio_seg, int); 137 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 138 size_t, register_t *); 139 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 140 141 static int fd_nameiat(struct lwp *, int, struct nameidata *); 142 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 143 namei_simple_flags_t, struct vnode **); 144 145 /* 146 * This table is used to maintain compatibility with 4.3BSD 147 * and NetBSD 0.9 mount syscalls - and possibly other systems. 148 * Note, the order is important! 149 * 150 * Do not modify this table. It should only contain filesystems 151 * supported by NetBSD 0.9 and 4.3BSD. 152 */ 153 const char * const mountcompatnames[] = { 154 NULL, /* 0 = MOUNT_NONE */ 155 MOUNT_FFS, /* 1 = MOUNT_UFS */ 156 MOUNT_NFS, /* 2 */ 157 MOUNT_MFS, /* 3 */ 158 MOUNT_MSDOS, /* 4 */ 159 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 160 MOUNT_FDESC, /* 6 */ 161 MOUNT_KERNFS, /* 7 */ 162 NULL, /* 8 = MOUNT_DEVFS */ 163 MOUNT_AFS, /* 9 */ 164 }; 165 166 const int nmountcompatnames = __arraycount(mountcompatnames); 167 168 static int 169 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 170 { 171 file_t *dfp; 172 int error; 173 174 if (fdat != AT_FDCWD) { 175 if ((error = fd_getvnode(fdat, &dfp)) != 0) 176 goto out; 177 178 NDAT(ndp, dfp->f_vnode); 179 } 180 181 error = namei(ndp); 182 183 if (fdat != AT_FDCWD) 184 fd_putfile(fdat); 185 out: 186 return error; 187 } 188 189 static int 190 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 191 namei_simple_flags_t sflags, struct vnode **vp_ret) 192 { 193 file_t *dfp; 194 struct vnode *dvp; 195 int error; 196 197 if (fdat != AT_FDCWD) { 198 if ((error = fd_getvnode(fdat, &dfp)) != 0) 199 goto out; 200 201 dvp = dfp->f_vnode; 202 } else { 203 dvp = NULL; 204 } 205 206 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 207 208 if (fdat != AT_FDCWD) 209 fd_putfile(fdat); 210 out: 211 return error; 212 } 213 214 static int 215 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 216 { 217 int error; 218 219 fp->f_flag = flags & FMASK; 220 fp->f_type = DTYPE_VNODE; 221 fp->f_ops = &vnops; 222 fp->f_vnode = vp; 223 224 if (flags & (O_EXLOCK | O_SHLOCK)) { 225 struct flock lf; 226 int type; 227 228 lf.l_whence = SEEK_SET; 229 lf.l_start = 0; 230 lf.l_len = 0; 231 if (flags & O_EXLOCK) 232 lf.l_type = F_WRLCK; 233 else 234 lf.l_type = F_RDLCK; 235 type = F_FLOCK; 236 if ((flags & FNONBLOCK) == 0) 237 type |= F_WAIT; 238 VOP_UNLOCK(vp); 239 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 240 if (error) { 241 (void) vn_close(vp, fp->f_flag, fp->f_cred); 242 fd_abort(l->l_proc, fp, indx); 243 return error; 244 } 245 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 246 atomic_or_uint(&fp->f_flag, FHASLOCK); 247 } 248 if (flags & O_CLOEXEC) 249 fd_set_exclose(l, indx, true); 250 return 0; 251 } 252 253 static int 254 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 255 void *data, size_t *data_len) 256 { 257 struct mount *mp; 258 int error = 0, saved_flags; 259 260 mp = vp->v_mount; 261 saved_flags = mp->mnt_flag; 262 263 /* We can operate only on VV_ROOT nodes. */ 264 if ((vp->v_vflag & VV_ROOT) == 0) { 265 error = EINVAL; 266 goto out; 267 } 268 269 /* 270 * We only allow the filesystem to be reloaded if it 271 * is currently mounted read-only. Additionally, we 272 * prevent read-write to read-only downgrades. 273 */ 274 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 275 (mp->mnt_flag & MNT_RDONLY) == 0 && 276 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 277 error = EOPNOTSUPP; /* Needs translation */ 278 goto out; 279 } 280 281 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 282 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 283 if (error) 284 goto out; 285 286 error = vfs_suspend(mp, 0); 287 if (error) 288 goto out; 289 290 mutex_enter(&mp->mnt_updating); 291 292 mp->mnt_flag &= ~MNT_OP_FLAGS; 293 mp->mnt_flag |= flags & MNT_OP_FLAGS; 294 295 /* 296 * Set the mount level flags. 297 */ 298 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 299 if ((flags & MNT_RDONLY)) 300 mp->mnt_iflag |= IMNT_WANTRDONLY; 301 else 302 mp->mnt_iflag |= IMNT_WANTRDWR; 303 } 304 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 305 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 306 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 307 mp->mnt_flag &= ~MNT_RDONLY; 308 309 error = VFS_MOUNT(mp, path, data, data_len); 310 311 if (error && data != NULL) { 312 int error2; 313 314 /* 315 * Update failed; let's try and see if it was an 316 * export request. For compat with 3.0 and earlier. 317 */ 318 error2 = vfs_hooks_reexport(mp, path, data); 319 320 /* 321 * Only update error code if the export request was 322 * understood but some problem occurred while 323 * processing it. 324 */ 325 if (error2 != EJUSTRETURN) 326 error = error2; 327 } 328 329 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 330 mp->mnt_flag |= MNT_RDONLY; 331 if (error) 332 mp->mnt_flag = saved_flags; 333 mp->mnt_flag &= ~MNT_OP_FLAGS; 334 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 335 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 336 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 337 vfs_syncer_add_to_worklist(mp); 338 } else { 339 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 340 vfs_syncer_remove_from_worklist(mp); 341 } 342 mutex_exit(&mp->mnt_updating); 343 vfs_resume(mp); 344 345 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 346 (flags & MNT_EXTATTR)) { 347 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 348 NULL, 0, NULL) != 0) { 349 printf("%s: failed to start extattr, error = %d", 350 mp->mnt_stat.f_mntonname, error); 351 mp->mnt_flag &= ~MNT_EXTATTR; 352 } 353 } 354 355 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 356 !(flags & MNT_EXTATTR)) { 357 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 358 NULL, 0, NULL) != 0) { 359 printf("%s: failed to stop extattr, error = %d", 360 mp->mnt_stat.f_mntonname, error); 361 mp->mnt_flag |= MNT_RDONLY; 362 } 363 } 364 out: 365 return (error); 366 } 367 368 static int 369 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 370 struct vfsops **vfsops) 371 { 372 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 373 int error; 374 375 if (type_seg == UIO_USERSPACE) { 376 /* Copy file-system type from userspace. */ 377 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 378 } else { 379 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 380 KASSERT(error == 0); 381 } 382 383 if (error) { 384 /* 385 * Historically, filesystem types were identified by numbers. 386 * If we get an integer for the filesystem type instead of a 387 * string, we check to see if it matches one of the historic 388 * filesystem types. 389 */ 390 u_long fsindex = (u_long)fstype; 391 if (fsindex >= nmountcompatnames || 392 mountcompatnames[fsindex] == NULL) 393 return ENODEV; 394 strlcpy(fstypename, mountcompatnames[fsindex], 395 sizeof(fstypename)); 396 } 397 398 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 399 if (strcmp(fstypename, "ufs") == 0) 400 fstypename[0] = 'f'; 401 402 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 403 return 0; 404 405 /* If we can autoload a vfs module, try again */ 406 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 407 408 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 409 return 0; 410 411 return ENODEV; 412 } 413 414 static int 415 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 416 void *data, size_t *data_len) 417 { 418 struct mount *mp; 419 int error; 420 421 /* If MNT_GETARGS is specified, it should be the only flag. */ 422 if (flags & ~MNT_GETARGS) 423 return EINVAL; 424 425 mp = vp->v_mount; 426 427 /* XXX: probably some notion of "can see" here if we want isolation. */ 428 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 429 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 430 if (error) 431 return error; 432 433 if ((vp->v_vflag & VV_ROOT) == 0) 434 return EINVAL; 435 436 if (vfs_busy(mp)) 437 return EPERM; 438 439 mutex_enter(&mp->mnt_updating); 440 mp->mnt_flag &= ~MNT_OP_FLAGS; 441 mp->mnt_flag |= MNT_GETARGS; 442 error = VFS_MOUNT(mp, path, data, data_len); 443 mp->mnt_flag &= ~MNT_OP_FLAGS; 444 mutex_exit(&mp->mnt_updating); 445 446 vfs_unbusy(mp); 447 return (error); 448 } 449 450 int 451 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 452 { 453 /* { 454 syscallarg(const char *) type; 455 syscallarg(const char *) path; 456 syscallarg(int) flags; 457 syscallarg(void *) data; 458 syscallarg(size_t) data_len; 459 } */ 460 461 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 462 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 463 SCARG(uap, data_len), retval); 464 } 465 466 int 467 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 468 const char *path, int flags, void *data, enum uio_seg data_seg, 469 size_t data_len, register_t *retval) 470 { 471 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 472 struct vnode *vp; 473 void *data_buf = data; 474 bool vfsopsrele = false; 475 size_t alloc_sz = 0; 476 int error; 477 478 /* 479 * Get vnode to be covered 480 */ 481 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 482 if (error != 0) { 483 vp = NULL; 484 goto done; 485 } 486 487 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 488 vfsops = vp->v_mount->mnt_op; 489 } else { 490 /* 'type' is userspace */ 491 error = mount_get_vfsops(type, type_seg, &vfsops); 492 if (error != 0) 493 goto done; 494 vfsopsrele = true; 495 } 496 497 /* 498 * We allow data to be NULL, even for userspace. Some fs's don't need 499 * it. The others will handle NULL. 500 */ 501 if (data != NULL && data_seg == UIO_USERSPACE) { 502 if (data_len == 0) { 503 /* No length supplied, use default for filesystem */ 504 data_len = vfsops->vfs_min_mount_data; 505 506 /* 507 * Hopefully a longer buffer won't make copyin() fail. 508 * For compatibility with 3.0 and earlier. 509 */ 510 if (flags & MNT_UPDATE 511 && data_len < sizeof (struct mnt_export_args30)) 512 data_len = sizeof (struct mnt_export_args30); 513 } 514 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 515 error = EINVAL; 516 goto done; 517 } 518 alloc_sz = data_len; 519 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 520 521 /* NFS needs the buffer even for mnt_getargs .... */ 522 error = copyin(data, data_buf, data_len); 523 if (error != 0) 524 goto done; 525 } 526 527 if (flags & MNT_GETARGS) { 528 if (data_len == 0) { 529 error = EINVAL; 530 goto done; 531 } 532 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 533 if (error != 0) 534 goto done; 535 if (data_seg == UIO_USERSPACE) 536 error = copyout(data_buf, data, data_len); 537 *retval = data_len; 538 } else if (flags & MNT_UPDATE) { 539 error = mount_update(l, vp, path, flags, data_buf, &data_len); 540 } else { 541 /* Locking is handled internally in mount_domount(). */ 542 KASSERT(vfsopsrele == true); 543 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 544 &data_len); 545 vfsopsrele = false; 546 } 547 if (!error) 548 KNOTE(&fs_klist, VQ_MOUNT); 549 550 done: 551 if (vfsopsrele) 552 vfs_delref(vfsops); 553 if (vp != NULL) { 554 vrele(vp); 555 } 556 if (data_buf != data) 557 kmem_free(data_buf, alloc_sz); 558 return (error); 559 } 560 561 /* 562 * Unmount a file system. 563 * 564 * Note: unmount takes a path to the vnode mounted on as argument, 565 * not special file (as before). 566 */ 567 /* ARGSUSED */ 568 int 569 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 570 { 571 /* { 572 syscallarg(const char *) path; 573 syscallarg(int) flags; 574 } */ 575 struct vnode *vp; 576 struct mount *mp; 577 int error; 578 struct pathbuf *pb; 579 struct nameidata nd; 580 581 error = pathbuf_copyin(SCARG(uap, path), &pb); 582 if (error) { 583 return error; 584 } 585 586 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 587 if ((error = namei(&nd)) != 0) { 588 pathbuf_destroy(pb); 589 return error; 590 } 591 vp = nd.ni_vp; 592 pathbuf_destroy(pb); 593 594 mp = vp->v_mount; 595 vfs_ref(mp); 596 VOP_UNLOCK(vp); 597 598 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 599 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 600 if (error) { 601 vrele(vp); 602 vfs_rele(mp); 603 return (error); 604 } 605 606 /* 607 * Don't allow unmounting the root file system. 608 */ 609 if (mp->mnt_flag & MNT_ROOTFS) { 610 vrele(vp); 611 vfs_rele(mp); 612 return (EINVAL); 613 } 614 615 /* 616 * Must be the root of the filesystem 617 */ 618 if ((vp->v_vflag & VV_ROOT) == 0) { 619 vrele(vp); 620 vfs_rele(mp); 621 return (EINVAL); 622 } 623 624 vrele(vp); 625 error = dounmount(mp, SCARG(uap, flags), l); 626 vfs_rele(mp); 627 if (!error) 628 KNOTE(&fs_klist, VQ_UNMOUNT); 629 return error; 630 } 631 632 /* 633 * Sync each mounted filesystem. 634 */ 635 #ifdef DEBUG 636 int syncprt = 0; 637 struct ctldebug debug0 = { "syncprt", &syncprt }; 638 #endif 639 640 void 641 do_sys_sync(struct lwp *l) 642 { 643 mount_iterator_t *iter; 644 struct mount *mp; 645 int asyncflag; 646 647 mountlist_iterator_init(&iter); 648 while ((mp = mountlist_iterator_next(iter)) != NULL) { 649 mutex_enter(&mp->mnt_updating); 650 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 651 asyncflag = mp->mnt_flag & MNT_ASYNC; 652 mp->mnt_flag &= ~MNT_ASYNC; 653 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 654 if (asyncflag) 655 mp->mnt_flag |= MNT_ASYNC; 656 } 657 mutex_exit(&mp->mnt_updating); 658 } 659 mountlist_iterator_destroy(iter); 660 #ifdef DEBUG 661 if (syncprt) 662 vfs_bufstats(); 663 #endif /* DEBUG */ 664 } 665 666 /* ARGSUSED */ 667 int 668 sys_sync(struct lwp *l, const void *v, register_t *retval) 669 { 670 do_sys_sync(l); 671 return (0); 672 } 673 674 675 /* 676 * Access or change filesystem quotas. 677 * 678 * (this is really 14 different calls bundled into one) 679 */ 680 681 static int 682 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 683 { 684 struct quotastat info_k; 685 int error; 686 687 /* ensure any padding bytes are cleared */ 688 memset(&info_k, 0, sizeof(info_k)); 689 690 error = vfs_quotactl_stat(mp, &info_k); 691 if (error) { 692 return error; 693 } 694 695 return copyout(&info_k, info_u, sizeof(info_k)); 696 } 697 698 static int 699 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 700 struct quotaidtypestat *info_u) 701 { 702 struct quotaidtypestat info_k; 703 int error; 704 705 /* ensure any padding bytes are cleared */ 706 memset(&info_k, 0, sizeof(info_k)); 707 708 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 709 if (error) { 710 return error; 711 } 712 713 return copyout(&info_k, info_u, sizeof(info_k)); 714 } 715 716 static int 717 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 718 struct quotaobjtypestat *info_u) 719 { 720 struct quotaobjtypestat info_k; 721 int error; 722 723 /* ensure any padding bytes are cleared */ 724 memset(&info_k, 0, sizeof(info_k)); 725 726 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 727 if (error) { 728 return error; 729 } 730 731 return copyout(&info_k, info_u, sizeof(info_k)); 732 } 733 734 static int 735 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 736 struct quotaval *val_u) 737 { 738 struct quotakey key_k; 739 struct quotaval val_k; 740 int error; 741 742 /* ensure any padding bytes are cleared */ 743 memset(&val_k, 0, sizeof(val_k)); 744 745 error = copyin(key_u, &key_k, sizeof(key_k)); 746 if (error) { 747 return error; 748 } 749 750 error = vfs_quotactl_get(mp, &key_k, &val_k); 751 if (error) { 752 return error; 753 } 754 755 return copyout(&val_k, val_u, sizeof(val_k)); 756 } 757 758 static int 759 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 760 const struct quotaval *val_u) 761 { 762 struct quotakey key_k; 763 struct quotaval val_k; 764 int error; 765 766 error = copyin(key_u, &key_k, sizeof(key_k)); 767 if (error) { 768 return error; 769 } 770 771 error = copyin(val_u, &val_k, sizeof(val_k)); 772 if (error) { 773 return error; 774 } 775 776 return vfs_quotactl_put(mp, &key_k, &val_k); 777 } 778 779 static int 780 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 781 { 782 struct quotakey key_k; 783 int error; 784 785 error = copyin(key_u, &key_k, sizeof(key_k)); 786 if (error) { 787 return error; 788 } 789 790 return vfs_quotactl_del(mp, &key_k); 791 } 792 793 static int 794 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 795 { 796 struct quotakcursor cursor_k; 797 int error; 798 799 /* ensure any padding bytes are cleared */ 800 memset(&cursor_k, 0, sizeof(cursor_k)); 801 802 error = vfs_quotactl_cursoropen(mp, &cursor_k); 803 if (error) { 804 return error; 805 } 806 807 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 808 } 809 810 static int 811 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 812 { 813 struct quotakcursor cursor_k; 814 int error; 815 816 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 817 if (error) { 818 return error; 819 } 820 821 return vfs_quotactl_cursorclose(mp, &cursor_k); 822 } 823 824 static int 825 do_sys_quotactl_cursorskipidtype(struct mount *mp, 826 struct quotakcursor *cursor_u, int idtype) 827 { 828 struct quotakcursor cursor_k; 829 int error; 830 831 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 832 if (error) { 833 return error; 834 } 835 836 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 837 if (error) { 838 return error; 839 } 840 841 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 842 } 843 844 static int 845 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 846 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 847 unsigned *ret_u) 848 { 849 #define CGET_STACK_MAX 8 850 struct quotakcursor cursor_k; 851 struct quotakey stackkeys[CGET_STACK_MAX]; 852 struct quotaval stackvals[CGET_STACK_MAX]; 853 struct quotakey *keys_k; 854 struct quotaval *vals_k; 855 unsigned ret_k; 856 int error; 857 858 if (maxnum > 128) { 859 maxnum = 128; 860 } 861 862 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 863 if (error) { 864 return error; 865 } 866 867 if (maxnum <= CGET_STACK_MAX) { 868 keys_k = stackkeys; 869 vals_k = stackvals; 870 /* ensure any padding bytes are cleared */ 871 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 872 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 873 } else { 874 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 875 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 876 } 877 878 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 879 &ret_k); 880 if (error) { 881 goto fail; 882 } 883 884 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 885 if (error) { 886 goto fail; 887 } 888 889 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 890 if (error) { 891 goto fail; 892 } 893 894 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 895 if (error) { 896 goto fail; 897 } 898 899 /* do last to maximize the chance of being able to recover a failure */ 900 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 901 902 fail: 903 if (keys_k != stackkeys) { 904 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 905 } 906 if (vals_k != stackvals) { 907 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 908 } 909 return error; 910 } 911 912 static int 913 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 914 int *ret_u) 915 { 916 struct quotakcursor cursor_k; 917 int ret_k; 918 int error; 919 920 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 921 if (error) { 922 return error; 923 } 924 925 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 926 if (error) { 927 return error; 928 } 929 930 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 931 if (error) { 932 return error; 933 } 934 935 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 936 } 937 938 static int 939 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 940 { 941 struct quotakcursor cursor_k; 942 int error; 943 944 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 945 if (error) { 946 return error; 947 } 948 949 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 950 if (error) { 951 return error; 952 } 953 954 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 955 } 956 957 static int 958 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 959 { 960 char *path_k; 961 int error; 962 963 /* XXX this should probably be a struct pathbuf */ 964 path_k = PNBUF_GET(); 965 error = copyin(path_u, path_k, PATH_MAX); 966 if (error) { 967 PNBUF_PUT(path_k); 968 return error; 969 } 970 971 error = vfs_quotactl_quotaon(mp, idtype, path_k); 972 973 PNBUF_PUT(path_k); 974 return error; 975 } 976 977 static int 978 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 979 { 980 return vfs_quotactl_quotaoff(mp, idtype); 981 } 982 983 int 984 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 985 { 986 struct mount *mp; 987 struct vnode *vp; 988 int error; 989 990 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 991 if (error != 0) 992 return (error); 993 mp = vp->v_mount; 994 995 switch (args->qc_op) { 996 case QUOTACTL_STAT: 997 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 998 break; 999 case QUOTACTL_IDTYPESTAT: 1000 error = do_sys_quotactl_idtypestat(mp, 1001 args->u.idtypestat.qc_idtype, 1002 args->u.idtypestat.qc_info); 1003 break; 1004 case QUOTACTL_OBJTYPESTAT: 1005 error = do_sys_quotactl_objtypestat(mp, 1006 args->u.objtypestat.qc_objtype, 1007 args->u.objtypestat.qc_info); 1008 break; 1009 case QUOTACTL_GET: 1010 error = do_sys_quotactl_get(mp, 1011 args->u.get.qc_key, 1012 args->u.get.qc_val); 1013 break; 1014 case QUOTACTL_PUT: 1015 error = do_sys_quotactl_put(mp, 1016 args->u.put.qc_key, 1017 args->u.put.qc_val); 1018 break; 1019 case QUOTACTL_DEL: 1020 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1021 break; 1022 case QUOTACTL_CURSOROPEN: 1023 error = do_sys_quotactl_cursoropen(mp, 1024 args->u.cursoropen.qc_cursor); 1025 break; 1026 case QUOTACTL_CURSORCLOSE: 1027 error = do_sys_quotactl_cursorclose(mp, 1028 args->u.cursorclose.qc_cursor); 1029 break; 1030 case QUOTACTL_CURSORSKIPIDTYPE: 1031 error = do_sys_quotactl_cursorskipidtype(mp, 1032 args->u.cursorskipidtype.qc_cursor, 1033 args->u.cursorskipidtype.qc_idtype); 1034 break; 1035 case QUOTACTL_CURSORGET: 1036 error = do_sys_quotactl_cursorget(mp, 1037 args->u.cursorget.qc_cursor, 1038 args->u.cursorget.qc_keys, 1039 args->u.cursorget.qc_vals, 1040 args->u.cursorget.qc_maxnum, 1041 args->u.cursorget.qc_ret); 1042 break; 1043 case QUOTACTL_CURSORATEND: 1044 error = do_sys_quotactl_cursoratend(mp, 1045 args->u.cursoratend.qc_cursor, 1046 args->u.cursoratend.qc_ret); 1047 break; 1048 case QUOTACTL_CURSORREWIND: 1049 error = do_sys_quotactl_cursorrewind(mp, 1050 args->u.cursorrewind.qc_cursor); 1051 break; 1052 case QUOTACTL_QUOTAON: 1053 error = do_sys_quotactl_quotaon(mp, 1054 args->u.quotaon.qc_idtype, 1055 args->u.quotaon.qc_quotafile); 1056 break; 1057 case QUOTACTL_QUOTAOFF: 1058 error = do_sys_quotactl_quotaoff(mp, 1059 args->u.quotaoff.qc_idtype); 1060 break; 1061 default: 1062 error = EINVAL; 1063 break; 1064 } 1065 1066 vrele(vp); 1067 return error; 1068 } 1069 1070 /* ARGSUSED */ 1071 int 1072 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1073 register_t *retval) 1074 { 1075 /* { 1076 syscallarg(const char *) path; 1077 syscallarg(struct quotactl_args *) args; 1078 } */ 1079 struct quotactl_args args; 1080 int error; 1081 1082 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1083 if (error) { 1084 return error; 1085 } 1086 1087 return do_sys_quotactl(SCARG(uap, path), &args); 1088 } 1089 1090 int 1091 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1092 int root) 1093 { 1094 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1095 int error = 0; 1096 1097 /* 1098 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1099 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1100 * overrides MNT_NOWAIT. 1101 */ 1102 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1103 (flags != MNT_WAIT && flags != 0)) { 1104 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1105 goto done; 1106 } 1107 1108 /* Get the filesystem stats now */ 1109 memset(sp, 0, sizeof(*sp)); 1110 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1111 return error; 1112 } 1113 1114 if (cwdi->cwdi_rdir == NULL) 1115 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1116 done: 1117 if (cwdi->cwdi_rdir != NULL) { 1118 size_t len; 1119 char *bp; 1120 char c; 1121 char *path = PNBUF_GET(); 1122 1123 bp = path + MAXPATHLEN; 1124 *--bp = '\0'; 1125 rw_enter(&cwdi->cwdi_lock, RW_READER); 1126 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1127 MAXPATHLEN / 2, 0, l); 1128 rw_exit(&cwdi->cwdi_lock); 1129 if (error) { 1130 PNBUF_PUT(path); 1131 return error; 1132 } 1133 len = strlen(bp); 1134 if (len != 1) { 1135 /* 1136 * for mount points that are below our root, we can see 1137 * them, so we fix up the pathname and return them. The 1138 * rest we cannot see, so we don't allow viewing the 1139 * data. 1140 */ 1141 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1142 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1143 (void)strlcpy(sp->f_mntonname, 1144 c == '\0' ? "/" : &sp->f_mntonname[len], 1145 sizeof(sp->f_mntonname)); 1146 } else { 1147 if (root) 1148 (void)strlcpy(sp->f_mntonname, "/", 1149 sizeof(sp->f_mntonname)); 1150 else 1151 error = EPERM; 1152 } 1153 } 1154 PNBUF_PUT(path); 1155 } 1156 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1157 return error; 1158 } 1159 1160 /* 1161 * Get filesystem statistics by path. 1162 */ 1163 int 1164 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1165 { 1166 struct mount *mp; 1167 int error; 1168 struct vnode *vp; 1169 1170 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1171 if (error != 0) 1172 return error; 1173 mp = vp->v_mount; 1174 error = dostatvfs(mp, sb, l, flags, 1); 1175 vrele(vp); 1176 return error; 1177 } 1178 1179 /* ARGSUSED */ 1180 int 1181 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1182 { 1183 /* { 1184 syscallarg(const char *) path; 1185 syscallarg(struct statvfs *) buf; 1186 syscallarg(int) flags; 1187 } */ 1188 struct statvfs *sb; 1189 int error; 1190 1191 sb = STATVFSBUF_GET(); 1192 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1193 if (error == 0) 1194 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1195 STATVFSBUF_PUT(sb); 1196 return error; 1197 } 1198 1199 /* 1200 * Get filesystem statistics by fd. 1201 */ 1202 int 1203 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1204 { 1205 file_t *fp; 1206 struct mount *mp; 1207 int error; 1208 1209 /* fd_getvnode() will use the descriptor for us */ 1210 if ((error = fd_getvnode(fd, &fp)) != 0) 1211 return (error); 1212 mp = fp->f_vnode->v_mount; 1213 error = dostatvfs(mp, sb, curlwp, flags, 1); 1214 fd_putfile(fd); 1215 return error; 1216 } 1217 1218 /* ARGSUSED */ 1219 int 1220 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1221 { 1222 /* { 1223 syscallarg(int) fd; 1224 syscallarg(struct statvfs *) buf; 1225 syscallarg(int) flags; 1226 } */ 1227 struct statvfs *sb; 1228 int error; 1229 1230 sb = STATVFSBUF_GET(); 1231 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1232 if (error == 0) 1233 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1234 STATVFSBUF_PUT(sb); 1235 return error; 1236 } 1237 1238 1239 /* 1240 * Get statistics on all filesystems. 1241 */ 1242 int 1243 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1244 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1245 register_t *retval) 1246 { 1247 int root = 0; 1248 mount_iterator_t *iter; 1249 struct proc *p = l->l_proc; 1250 struct mount *mp; 1251 struct statvfs *sb; 1252 size_t count, maxcount; 1253 int error = 0; 1254 1255 sb = STATVFSBUF_GET(); 1256 maxcount = bufsize / entry_sz; 1257 count = 0; 1258 mountlist_iterator_init(&iter); 1259 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1260 if (sfsp && count < maxcount) { 1261 error = dostatvfs(mp, sb, l, flags, 0); 1262 if (error) { 1263 error = 0; 1264 continue; 1265 } 1266 error = copyfn(sb, sfsp, entry_sz); 1267 if (error) 1268 goto out; 1269 sfsp = (char *)sfsp + entry_sz; 1270 root |= strcmp(sb->f_mntonname, "/") == 0; 1271 } 1272 count++; 1273 } 1274 1275 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1276 /* 1277 * fake a root entry 1278 */ 1279 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1280 sb, l, flags, 1); 1281 if (error != 0) 1282 goto out; 1283 if (sfsp) { 1284 error = copyfn(sb, sfsp, entry_sz); 1285 if (error != 0) 1286 goto out; 1287 } 1288 count++; 1289 } 1290 if (sfsp && count > maxcount) 1291 *retval = maxcount; 1292 else 1293 *retval = count; 1294 out: 1295 mountlist_iterator_destroy(iter); 1296 STATVFSBUF_PUT(sb); 1297 return error; 1298 } 1299 1300 int 1301 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1302 { 1303 /* { 1304 syscallarg(struct statvfs *) buf; 1305 syscallarg(size_t) bufsize; 1306 syscallarg(int) flags; 1307 } */ 1308 1309 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1310 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1311 } 1312 1313 /* 1314 * Change current working directory to a given file descriptor. 1315 */ 1316 /* ARGSUSED */ 1317 int 1318 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1319 { 1320 /* { 1321 syscallarg(int) fd; 1322 } */ 1323 struct proc *p = l->l_proc; 1324 struct cwdinfo *cwdi; 1325 struct vnode *vp, *tdp; 1326 struct mount *mp; 1327 file_t *fp; 1328 int error, fd; 1329 1330 /* fd_getvnode() will use the descriptor for us */ 1331 fd = SCARG(uap, fd); 1332 if ((error = fd_getvnode(fd, &fp)) != 0) 1333 return (error); 1334 vp = fp->f_vnode; 1335 1336 vref(vp); 1337 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1338 if (vp->v_type != VDIR) 1339 error = ENOTDIR; 1340 else 1341 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1342 if (error) { 1343 vput(vp); 1344 goto out; 1345 } 1346 while ((mp = vp->v_mountedhere) != NULL) { 1347 error = vfs_busy(mp); 1348 vput(vp); 1349 if (error != 0) 1350 goto out; 1351 error = VFS_ROOT(mp, &tdp); 1352 vfs_unbusy(mp); 1353 if (error) 1354 goto out; 1355 vp = tdp; 1356 } 1357 VOP_UNLOCK(vp); 1358 1359 /* 1360 * Disallow changing to a directory not under the process's 1361 * current root directory (if there is one). 1362 */ 1363 cwdi = p->p_cwdi; 1364 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1365 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1366 vrele(vp); 1367 error = EPERM; /* operation not permitted */ 1368 } else { 1369 vrele(cwdi->cwdi_cdir); 1370 cwdi->cwdi_cdir = vp; 1371 } 1372 rw_exit(&cwdi->cwdi_lock); 1373 1374 out: 1375 fd_putfile(fd); 1376 return (error); 1377 } 1378 1379 /* 1380 * Change this process's notion of the root directory to a given file 1381 * descriptor. 1382 */ 1383 int 1384 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1385 { 1386 struct proc *p = l->l_proc; 1387 struct vnode *vp; 1388 file_t *fp; 1389 int error, fd = SCARG(uap, fd); 1390 1391 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1392 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1393 return error; 1394 /* fd_getvnode() will use the descriptor for us */ 1395 if ((error = fd_getvnode(fd, &fp)) != 0) 1396 return error; 1397 vp = fp->f_vnode; 1398 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1399 if (vp->v_type != VDIR) 1400 error = ENOTDIR; 1401 else 1402 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1403 VOP_UNLOCK(vp); 1404 if (error) 1405 goto out; 1406 vref(vp); 1407 1408 change_root(p->p_cwdi, vp, l); 1409 1410 out: 1411 fd_putfile(fd); 1412 return (error); 1413 } 1414 1415 /* 1416 * Change current working directory (``.''). 1417 */ 1418 /* ARGSUSED */ 1419 int 1420 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1421 { 1422 /* { 1423 syscallarg(const char *) path; 1424 } */ 1425 struct proc *p = l->l_proc; 1426 struct cwdinfo *cwdi; 1427 int error; 1428 struct vnode *vp; 1429 1430 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1431 &vp, l)) != 0) 1432 return (error); 1433 cwdi = p->p_cwdi; 1434 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1435 vrele(cwdi->cwdi_cdir); 1436 cwdi->cwdi_cdir = vp; 1437 rw_exit(&cwdi->cwdi_lock); 1438 return (0); 1439 } 1440 1441 /* 1442 * Change notion of root (``/'') directory. 1443 */ 1444 /* ARGSUSED */ 1445 int 1446 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1447 { 1448 /* { 1449 syscallarg(const char *) path; 1450 } */ 1451 struct proc *p = l->l_proc; 1452 int error; 1453 struct vnode *vp; 1454 1455 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1456 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1457 return (error); 1458 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1459 &vp, l)) != 0) 1460 return (error); 1461 1462 change_root(p->p_cwdi, vp, l); 1463 1464 return (0); 1465 } 1466 1467 /* 1468 * Common routine for chroot and fchroot. 1469 * NB: callers need to properly authorize the change root operation. 1470 */ 1471 void 1472 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1473 { 1474 struct proc *p = l->l_proc; 1475 kauth_cred_t ncred; 1476 1477 ncred = kauth_cred_alloc(); 1478 1479 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1480 if (cwdi->cwdi_rdir != NULL) 1481 vrele(cwdi->cwdi_rdir); 1482 cwdi->cwdi_rdir = vp; 1483 1484 /* 1485 * Prevent escaping from chroot by putting the root under 1486 * the working directory. Silently chdir to / if we aren't 1487 * already there. 1488 */ 1489 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1490 /* 1491 * XXX would be more failsafe to change directory to a 1492 * deadfs node here instead 1493 */ 1494 vrele(cwdi->cwdi_cdir); 1495 vref(vp); 1496 cwdi->cwdi_cdir = vp; 1497 } 1498 rw_exit(&cwdi->cwdi_lock); 1499 1500 /* Get a write lock on the process credential. */ 1501 proc_crmod_enter(); 1502 1503 kauth_cred_clone(p->p_cred, ncred); 1504 kauth_proc_chroot(ncred, p->p_cwdi); 1505 1506 /* Broadcast our credentials to the process and other LWPs. */ 1507 proc_crmod_leave(ncred, p->p_cred, true); 1508 } 1509 1510 /* 1511 * Common routine for chroot and chdir. 1512 * XXX "where" should be enum uio_seg 1513 */ 1514 int 1515 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1516 { 1517 struct pathbuf *pb; 1518 struct nameidata nd; 1519 int error; 1520 1521 error = pathbuf_maybe_copyin(path, where, &pb); 1522 if (error) { 1523 return error; 1524 } 1525 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1526 if ((error = namei(&nd)) != 0) { 1527 pathbuf_destroy(pb); 1528 return error; 1529 } 1530 *vpp = nd.ni_vp; 1531 pathbuf_destroy(pb); 1532 1533 if ((*vpp)->v_type != VDIR) 1534 error = ENOTDIR; 1535 else 1536 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1537 1538 if (error) 1539 vput(*vpp); 1540 else 1541 VOP_UNLOCK(*vpp); 1542 return (error); 1543 } 1544 1545 /* 1546 * Internals of sys_open - path has already been converted into a pathbuf 1547 * (so we can easily reuse this function from other parts of the kernel, 1548 * like posix_spawn post-processing). 1549 */ 1550 int 1551 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1552 int open_mode, int *fd) 1553 { 1554 struct proc *p = l->l_proc; 1555 struct cwdinfo *cwdi = p->p_cwdi; 1556 file_t *fp; 1557 struct vnode *vp; 1558 int flags, cmode; 1559 int indx, error; 1560 struct nameidata nd; 1561 1562 if (open_flags & O_SEARCH) { 1563 open_flags &= ~(int)O_SEARCH; 1564 } 1565 1566 flags = FFLAGS(open_flags); 1567 if ((flags & (FREAD | FWRITE)) == 0) 1568 return EINVAL; 1569 1570 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1571 return error; 1572 } 1573 1574 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1575 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1576 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1577 if (dvp != NULL) 1578 NDAT(&nd, dvp); 1579 1580 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1581 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1582 fd_abort(p, fp, indx); 1583 if ((error == EDUPFD || error == EMOVEFD) && 1584 l->l_dupfd >= 0 && /* XXX from fdopen */ 1585 (error = 1586 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1587 *fd = indx; 1588 return 0; 1589 } 1590 if (error == ERESTART) 1591 error = EINTR; 1592 return error; 1593 } 1594 1595 l->l_dupfd = 0; 1596 vp = nd.ni_vp; 1597 1598 if ((error = open_setfp(l, fp, vp, indx, flags))) 1599 return error; 1600 1601 VOP_UNLOCK(vp); 1602 *fd = indx; 1603 fd_affix(p, fp, indx); 1604 return 0; 1605 } 1606 1607 int 1608 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1609 { 1610 struct pathbuf *pb; 1611 int error, oflags; 1612 1613 oflags = FFLAGS(open_flags); 1614 if ((oflags & (FREAD | FWRITE)) == 0) 1615 return EINVAL; 1616 1617 pb = pathbuf_create(path); 1618 if (pb == NULL) 1619 return ENOMEM; 1620 1621 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1622 pathbuf_destroy(pb); 1623 1624 return error; 1625 } 1626 1627 /* 1628 * Check permissions, allocate an open file structure, 1629 * and call the device open routine if any. 1630 */ 1631 static int 1632 stub_sys_openat_10(struct pathbuf **pb) 1633 { 1634 1635 return 0; 1636 } 1637 1638 MODULE_CALL_INT_HOOK_DECL(compat_10_openat_hook, (struct pathbuf **)); 1639 MODULE_CALL_INT_HOOK(compat_10_openat_hook, (struct pathbuf **pb), (pb), 1640 stub_sys_openat_10(pb)); 1641 1642 static int 1643 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1644 int mode, int *fd) 1645 { 1646 file_t *dfp = NULL; 1647 struct vnode *dvp = NULL; 1648 struct pathbuf *pb; 1649 int error; 1650 1651 if (path == NULL) { 1652 error = compat_10_openat_hook_call(&pb); 1653 if (error) 1654 return error; 1655 } else { 1656 error = pathbuf_copyin(path, &pb); 1657 if (error) 1658 return error; 1659 } 1660 1661 if (fdat != AT_FDCWD) { 1662 /* fd_getvnode() will use the descriptor for us */ 1663 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1664 goto out; 1665 1666 dvp = dfp->f_vnode; 1667 } 1668 1669 error = do_open(l, dvp, pb, flags, mode, fd); 1670 1671 if (dfp != NULL) 1672 fd_putfile(fdat); 1673 out: 1674 pathbuf_destroy(pb); 1675 return error; 1676 } 1677 1678 int 1679 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1680 { 1681 /* { 1682 syscallarg(const char *) path; 1683 syscallarg(int) flags; 1684 syscallarg(int) mode; 1685 } */ 1686 int error; 1687 int fd; 1688 1689 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1690 SCARG(uap, flags), SCARG(uap, mode), &fd); 1691 1692 if (error == 0) 1693 *retval = fd; 1694 1695 return error; 1696 } 1697 1698 int 1699 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1700 { 1701 /* { 1702 syscallarg(int) fd; 1703 syscallarg(const char *) path; 1704 syscallarg(int) oflags; 1705 syscallarg(int) mode; 1706 } */ 1707 int error; 1708 int fd; 1709 1710 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1711 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1712 1713 if (error == 0) 1714 *retval = fd; 1715 1716 return error; 1717 } 1718 1719 static void 1720 vfs__fhfree(fhandle_t *fhp) 1721 { 1722 size_t fhsize; 1723 1724 fhsize = FHANDLE_SIZE(fhp); 1725 kmem_free(fhp, fhsize); 1726 } 1727 1728 /* 1729 * vfs_composefh: compose a filehandle. 1730 */ 1731 1732 int 1733 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1734 { 1735 struct mount *mp; 1736 struct fid *fidp; 1737 int error; 1738 size_t needfhsize; 1739 size_t fidsize; 1740 1741 mp = vp->v_mount; 1742 fidp = NULL; 1743 if (*fh_size < FHANDLE_SIZE_MIN) { 1744 fidsize = 0; 1745 } else { 1746 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1747 if (fhp != NULL) { 1748 memset(fhp, 0, *fh_size); 1749 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1750 fidp = &fhp->fh_fid; 1751 } 1752 } 1753 error = VFS_VPTOFH(vp, fidp, &fidsize); 1754 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1755 if (error == 0 && *fh_size < needfhsize) { 1756 error = E2BIG; 1757 } 1758 *fh_size = needfhsize; 1759 return error; 1760 } 1761 1762 int 1763 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1764 { 1765 struct mount *mp; 1766 fhandle_t *fhp; 1767 size_t fhsize; 1768 size_t fidsize; 1769 int error; 1770 1771 mp = vp->v_mount; 1772 fidsize = 0; 1773 error = VFS_VPTOFH(vp, NULL, &fidsize); 1774 KASSERT(error != 0); 1775 if (error != E2BIG) { 1776 goto out; 1777 } 1778 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1779 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1780 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1781 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1782 if (error == 0) { 1783 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1784 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1785 *fhpp = fhp; 1786 } else { 1787 kmem_free(fhp, fhsize); 1788 } 1789 out: 1790 return error; 1791 } 1792 1793 void 1794 vfs_composefh_free(fhandle_t *fhp) 1795 { 1796 1797 vfs__fhfree(fhp); 1798 } 1799 1800 /* 1801 * vfs_fhtovp: lookup a vnode by a filehandle. 1802 */ 1803 1804 int 1805 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1806 { 1807 struct mount *mp; 1808 int error; 1809 1810 *vpp = NULL; 1811 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1812 if (mp == NULL) { 1813 error = ESTALE; 1814 goto out; 1815 } 1816 if (mp->mnt_op->vfs_fhtovp == NULL) { 1817 error = EOPNOTSUPP; 1818 goto out; 1819 } 1820 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1821 out: 1822 return error; 1823 } 1824 1825 /* 1826 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1827 * the needed size. 1828 */ 1829 1830 int 1831 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1832 { 1833 fhandle_t *fhp; 1834 int error; 1835 1836 if (fhsize > FHANDLE_SIZE_MAX) { 1837 return EINVAL; 1838 } 1839 if (fhsize < FHANDLE_SIZE_MIN) { 1840 return EINVAL; 1841 } 1842 again: 1843 fhp = kmem_alloc(fhsize, KM_SLEEP); 1844 error = copyin(ufhp, fhp, fhsize); 1845 if (error == 0) { 1846 /* XXX this check shouldn't be here */ 1847 if (FHANDLE_SIZE(fhp) == fhsize) { 1848 *fhpp = fhp; 1849 return 0; 1850 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1851 /* 1852 * a kludge for nfsv2 padded handles. 1853 */ 1854 size_t sz; 1855 1856 sz = FHANDLE_SIZE(fhp); 1857 kmem_free(fhp, fhsize); 1858 fhsize = sz; 1859 goto again; 1860 } else { 1861 /* 1862 * userland told us wrong size. 1863 */ 1864 error = EINVAL; 1865 } 1866 } 1867 kmem_free(fhp, fhsize); 1868 return error; 1869 } 1870 1871 void 1872 vfs_copyinfh_free(fhandle_t *fhp) 1873 { 1874 1875 vfs__fhfree(fhp); 1876 } 1877 1878 /* 1879 * Get file handle system call 1880 */ 1881 int 1882 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1883 { 1884 /* { 1885 syscallarg(char *) fname; 1886 syscallarg(fhandle_t *) fhp; 1887 syscallarg(size_t *) fh_size; 1888 } */ 1889 struct vnode *vp; 1890 fhandle_t *fh; 1891 int error; 1892 struct pathbuf *pb; 1893 struct nameidata nd; 1894 size_t sz; 1895 size_t usz; 1896 1897 /* 1898 * Must be super user 1899 */ 1900 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1901 0, NULL, NULL, NULL); 1902 if (error) 1903 return (error); 1904 1905 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1906 if (error) { 1907 return error; 1908 } 1909 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1910 error = namei(&nd); 1911 if (error) { 1912 pathbuf_destroy(pb); 1913 return error; 1914 } 1915 vp = nd.ni_vp; 1916 pathbuf_destroy(pb); 1917 1918 error = vfs_composefh_alloc(vp, &fh); 1919 vput(vp); 1920 if (error != 0) { 1921 return error; 1922 } 1923 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1924 if (error != 0) { 1925 goto out; 1926 } 1927 sz = FHANDLE_SIZE(fh); 1928 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1929 if (error != 0) { 1930 goto out; 1931 } 1932 if (usz >= sz) { 1933 error = copyout(fh, SCARG(uap, fhp), sz); 1934 } else { 1935 error = E2BIG; 1936 } 1937 out: 1938 vfs_composefh_free(fh); 1939 return (error); 1940 } 1941 1942 /* 1943 * Open a file given a file handle. 1944 * 1945 * Check permissions, allocate an open file structure, 1946 * and call the device open routine if any. 1947 */ 1948 1949 int 1950 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1951 register_t *retval) 1952 { 1953 file_t *fp; 1954 struct vnode *vp = NULL; 1955 kauth_cred_t cred = l->l_cred; 1956 file_t *nfp; 1957 int indx, error; 1958 struct vattr va; 1959 fhandle_t *fh; 1960 int flags; 1961 proc_t *p; 1962 1963 p = curproc; 1964 1965 /* 1966 * Must be super user 1967 */ 1968 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1969 0, NULL, NULL, NULL))) 1970 return (error); 1971 1972 if (oflags & O_SEARCH) { 1973 oflags &= ~(int)O_SEARCH; 1974 } 1975 1976 flags = FFLAGS(oflags); 1977 if ((flags & (FREAD | FWRITE)) == 0) 1978 return (EINVAL); 1979 if ((flags & O_CREAT)) 1980 return (EINVAL); 1981 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1982 return (error); 1983 fp = nfp; 1984 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1985 if (error != 0) { 1986 goto bad; 1987 } 1988 error = vfs_fhtovp(fh, &vp); 1989 vfs_copyinfh_free(fh); 1990 if (error != 0) { 1991 goto bad; 1992 } 1993 1994 /* Now do an effective vn_open */ 1995 1996 if (vp->v_type == VSOCK) { 1997 error = EOPNOTSUPP; 1998 goto bad; 1999 } 2000 error = vn_openchk(vp, cred, flags); 2001 if (error != 0) 2002 goto bad; 2003 if (flags & O_TRUNC) { 2004 VOP_UNLOCK(vp); /* XXX */ 2005 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2006 vattr_null(&va); 2007 va.va_size = 0; 2008 error = VOP_SETATTR(vp, &va, cred); 2009 if (error) 2010 goto bad; 2011 } 2012 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2013 goto bad; 2014 if (flags & FWRITE) { 2015 mutex_enter(vp->v_interlock); 2016 vp->v_writecount++; 2017 mutex_exit(vp->v_interlock); 2018 } 2019 2020 /* done with modified vn_open, now finish what sys_open does. */ 2021 if ((error = open_setfp(l, fp, vp, indx, flags))) 2022 return error; 2023 2024 VOP_UNLOCK(vp); 2025 *retval = indx; 2026 fd_affix(p, fp, indx); 2027 return (0); 2028 2029 bad: 2030 fd_abort(p, fp, indx); 2031 if (vp != NULL) 2032 vput(vp); 2033 return (error); 2034 } 2035 2036 int 2037 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2038 { 2039 /* { 2040 syscallarg(const void *) fhp; 2041 syscallarg(size_t) fh_size; 2042 syscallarg(int) flags; 2043 } */ 2044 2045 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2046 SCARG(uap, flags), retval); 2047 } 2048 2049 int 2050 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2051 { 2052 int error; 2053 fhandle_t *fh; 2054 struct vnode *vp; 2055 2056 /* 2057 * Must be super user 2058 */ 2059 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2060 0, NULL, NULL, NULL))) 2061 return (error); 2062 2063 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2064 if (error != 0) 2065 return error; 2066 2067 error = vfs_fhtovp(fh, &vp); 2068 vfs_copyinfh_free(fh); 2069 if (error != 0) 2070 return error; 2071 2072 error = vn_stat(vp, sb); 2073 vput(vp); 2074 return error; 2075 } 2076 2077 2078 /* ARGSUSED */ 2079 int 2080 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2081 { 2082 /* { 2083 syscallarg(const void *) fhp; 2084 syscallarg(size_t) fh_size; 2085 syscallarg(struct stat *) sb; 2086 } */ 2087 struct stat sb; 2088 int error; 2089 2090 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2091 if (error) 2092 return error; 2093 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2094 } 2095 2096 int 2097 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2098 int flags) 2099 { 2100 fhandle_t *fh; 2101 struct mount *mp; 2102 struct vnode *vp; 2103 int error; 2104 2105 /* 2106 * Must be super user 2107 */ 2108 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2109 0, NULL, NULL, NULL))) 2110 return error; 2111 2112 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2113 if (error != 0) 2114 return error; 2115 2116 error = vfs_fhtovp(fh, &vp); 2117 vfs_copyinfh_free(fh); 2118 if (error != 0) 2119 return error; 2120 2121 mp = vp->v_mount; 2122 error = dostatvfs(mp, sb, l, flags, 1); 2123 vput(vp); 2124 return error; 2125 } 2126 2127 /* ARGSUSED */ 2128 int 2129 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2130 { 2131 /* { 2132 syscallarg(const void *) fhp; 2133 syscallarg(size_t) fh_size; 2134 syscallarg(struct statvfs *) buf; 2135 syscallarg(int) flags; 2136 } */ 2137 struct statvfs *sb = STATVFSBUF_GET(); 2138 int error; 2139 2140 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2141 SCARG(uap, flags)); 2142 if (error == 0) 2143 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2144 STATVFSBUF_PUT(sb); 2145 return error; 2146 } 2147 2148 /* 2149 * Create a special file. 2150 */ 2151 /* ARGSUSED */ 2152 int 2153 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2154 register_t *retval) 2155 { 2156 /* { 2157 syscallarg(const char *) path; 2158 syscallarg(mode_t) mode; 2159 syscallarg(dev_t) dev; 2160 } */ 2161 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2162 SCARG(uap, dev), retval, UIO_USERSPACE); 2163 } 2164 2165 int 2166 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2167 register_t *retval) 2168 { 2169 /* { 2170 syscallarg(int) fd; 2171 syscallarg(const char *) path; 2172 syscallarg(mode_t) mode; 2173 syscallarg(int) pad; 2174 syscallarg(dev_t) dev; 2175 } */ 2176 2177 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2178 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2179 } 2180 2181 int 2182 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2183 register_t *retval, enum uio_seg seg) 2184 { 2185 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2186 } 2187 2188 int 2189 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2190 dev_t dev, register_t *retval, enum uio_seg seg) 2191 { 2192 struct proc *p = l->l_proc; 2193 struct vnode *vp; 2194 struct vattr vattr; 2195 int error, optype; 2196 struct pathbuf *pb; 2197 struct nameidata nd; 2198 const char *pathstring; 2199 2200 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2201 0, NULL, NULL, NULL)) != 0) 2202 return (error); 2203 2204 optype = VOP_MKNOD_DESCOFFSET; 2205 2206 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2207 if (error) { 2208 return error; 2209 } 2210 pathstring = pathbuf_stringcopy_get(pb); 2211 if (pathstring == NULL) { 2212 pathbuf_destroy(pb); 2213 return ENOMEM; 2214 } 2215 2216 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2217 2218 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2219 goto out; 2220 vp = nd.ni_vp; 2221 2222 if (vp != NULL) 2223 error = EEXIST; 2224 else { 2225 vattr_null(&vattr); 2226 /* We will read cwdi->cwdi_cmask unlocked. */ 2227 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2228 vattr.va_rdev = dev; 2229 2230 switch (mode & S_IFMT) { 2231 case S_IFMT: /* used by badsect to flag bad sectors */ 2232 vattr.va_type = VBAD; 2233 break; 2234 case S_IFCHR: 2235 vattr.va_type = VCHR; 2236 break; 2237 case S_IFBLK: 2238 vattr.va_type = VBLK; 2239 break; 2240 case S_IFWHT: 2241 optype = VOP_WHITEOUT_DESCOFFSET; 2242 break; 2243 case S_IFREG: 2244 #if NVERIEXEC > 0 2245 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2246 O_CREAT); 2247 #endif /* NVERIEXEC > 0 */ 2248 vattr.va_type = VREG; 2249 vattr.va_rdev = VNOVAL; 2250 optype = VOP_CREATE_DESCOFFSET; 2251 break; 2252 default: 2253 error = EINVAL; 2254 break; 2255 } 2256 } 2257 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2258 && vattr.va_rdev == VNOVAL) 2259 error = EINVAL; 2260 if (!error) { 2261 switch (optype) { 2262 case VOP_WHITEOUT_DESCOFFSET: 2263 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2264 if (error) 2265 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2266 vput(nd.ni_dvp); 2267 break; 2268 2269 case VOP_MKNOD_DESCOFFSET: 2270 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2271 &nd.ni_cnd, &vattr); 2272 if (error == 0) 2273 vrele(nd.ni_vp); 2274 vput(nd.ni_dvp); 2275 break; 2276 2277 case VOP_CREATE_DESCOFFSET: 2278 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2279 &nd.ni_cnd, &vattr); 2280 if (error == 0) 2281 vrele(nd.ni_vp); 2282 vput(nd.ni_dvp); 2283 break; 2284 } 2285 } else { 2286 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2287 if (nd.ni_dvp == vp) 2288 vrele(nd.ni_dvp); 2289 else 2290 vput(nd.ni_dvp); 2291 if (vp) 2292 vrele(vp); 2293 } 2294 out: 2295 pathbuf_stringcopy_put(pb, pathstring); 2296 pathbuf_destroy(pb); 2297 return (error); 2298 } 2299 2300 /* 2301 * Create a named pipe. 2302 */ 2303 /* ARGSUSED */ 2304 int 2305 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2306 { 2307 /* { 2308 syscallarg(const char *) path; 2309 syscallarg(int) mode; 2310 } */ 2311 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2312 } 2313 2314 int 2315 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2316 register_t *retval) 2317 { 2318 /* { 2319 syscallarg(int) fd; 2320 syscallarg(const char *) path; 2321 syscallarg(int) mode; 2322 } */ 2323 2324 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2325 SCARG(uap, mode)); 2326 } 2327 2328 static int 2329 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2330 { 2331 struct proc *p = l->l_proc; 2332 struct vattr vattr; 2333 int error; 2334 struct pathbuf *pb; 2335 struct nameidata nd; 2336 2337 error = pathbuf_copyin(path, &pb); 2338 if (error) { 2339 return error; 2340 } 2341 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2342 2343 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2344 pathbuf_destroy(pb); 2345 return error; 2346 } 2347 if (nd.ni_vp != NULL) { 2348 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2349 if (nd.ni_dvp == nd.ni_vp) 2350 vrele(nd.ni_dvp); 2351 else 2352 vput(nd.ni_dvp); 2353 vrele(nd.ni_vp); 2354 pathbuf_destroy(pb); 2355 return (EEXIST); 2356 } 2357 vattr_null(&vattr); 2358 vattr.va_type = VFIFO; 2359 /* We will read cwdi->cwdi_cmask unlocked. */ 2360 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2361 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2362 if (error == 0) 2363 vrele(nd.ni_vp); 2364 vput(nd.ni_dvp); 2365 pathbuf_destroy(pb); 2366 return (error); 2367 } 2368 2369 /* 2370 * Make a hard file link. 2371 */ 2372 /* ARGSUSED */ 2373 int 2374 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2375 const char *link, int follow, register_t *retval) 2376 { 2377 struct vnode *vp; 2378 struct pathbuf *linkpb; 2379 struct nameidata nd; 2380 namei_simple_flags_t ns_flags; 2381 int error; 2382 2383 if (follow & AT_SYMLINK_FOLLOW) 2384 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2385 else 2386 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2387 2388 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2389 if (error != 0) 2390 return (error); 2391 error = pathbuf_copyin(link, &linkpb); 2392 if (error) { 2393 goto out1; 2394 } 2395 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2396 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2397 goto out2; 2398 if (nd.ni_vp) { 2399 error = EEXIST; 2400 goto abortop; 2401 } 2402 /* Prevent hard links on directories. */ 2403 if (vp->v_type == VDIR) { 2404 error = EPERM; 2405 goto abortop; 2406 } 2407 /* Prevent cross-mount operation. */ 2408 if (nd.ni_dvp->v_mount != vp->v_mount) { 2409 error = EXDEV; 2410 goto abortop; 2411 } 2412 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2413 VOP_UNLOCK(nd.ni_dvp); 2414 vrele(nd.ni_dvp); 2415 out2: 2416 pathbuf_destroy(linkpb); 2417 out1: 2418 vrele(vp); 2419 return (error); 2420 abortop: 2421 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2422 if (nd.ni_dvp == nd.ni_vp) 2423 vrele(nd.ni_dvp); 2424 else 2425 vput(nd.ni_dvp); 2426 if (nd.ni_vp != NULL) 2427 vrele(nd.ni_vp); 2428 goto out2; 2429 } 2430 2431 int 2432 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2433 { 2434 /* { 2435 syscallarg(const char *) path; 2436 syscallarg(const char *) link; 2437 } */ 2438 const char *path = SCARG(uap, path); 2439 const char *link = SCARG(uap, link); 2440 2441 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2442 AT_SYMLINK_FOLLOW, retval); 2443 } 2444 2445 int 2446 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2447 register_t *retval) 2448 { 2449 /* { 2450 syscallarg(int) fd1; 2451 syscallarg(const char *) name1; 2452 syscallarg(int) fd2; 2453 syscallarg(const char *) name2; 2454 syscallarg(int) flags; 2455 } */ 2456 int fd1 = SCARG(uap, fd1); 2457 const char *name1 = SCARG(uap, name1); 2458 int fd2 = SCARG(uap, fd2); 2459 const char *name2 = SCARG(uap, name2); 2460 int follow; 2461 2462 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2463 2464 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2465 } 2466 2467 2468 int 2469 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2470 { 2471 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2472 } 2473 2474 static int 2475 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2476 const char *link, enum uio_seg seg) 2477 { 2478 struct proc *p = curproc; 2479 struct vattr vattr; 2480 char *path; 2481 int error; 2482 size_t len; 2483 struct pathbuf *linkpb; 2484 struct nameidata nd; 2485 2486 KASSERT(l != NULL || fdat == AT_FDCWD); 2487 2488 path = PNBUF_GET(); 2489 if (seg == UIO_USERSPACE) { 2490 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2491 goto out1; 2492 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2493 goto out1; 2494 } else { 2495 len = strlen(patharg) + 1; 2496 KASSERT(len <= MAXPATHLEN); 2497 memcpy(path, patharg, len); 2498 linkpb = pathbuf_create(link); 2499 if (linkpb == NULL) { 2500 error = ENOMEM; 2501 goto out1; 2502 } 2503 } 2504 ktrkuser("symlink-target", path, len - 1); 2505 2506 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2507 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2508 goto out2; 2509 if (nd.ni_vp) { 2510 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2511 if (nd.ni_dvp == nd.ni_vp) 2512 vrele(nd.ni_dvp); 2513 else 2514 vput(nd.ni_dvp); 2515 vrele(nd.ni_vp); 2516 error = EEXIST; 2517 goto out2; 2518 } 2519 vattr_null(&vattr); 2520 vattr.va_type = VLNK; 2521 /* We will read cwdi->cwdi_cmask unlocked. */ 2522 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2523 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2524 if (error == 0) 2525 vrele(nd.ni_vp); 2526 vput(nd.ni_dvp); 2527 out2: 2528 pathbuf_destroy(linkpb); 2529 out1: 2530 PNBUF_PUT(path); 2531 return (error); 2532 } 2533 2534 /* 2535 * Make a symbolic link. 2536 */ 2537 /* ARGSUSED */ 2538 int 2539 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2540 { 2541 /* { 2542 syscallarg(const char *) path; 2543 syscallarg(const char *) link; 2544 } */ 2545 2546 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2547 UIO_USERSPACE); 2548 } 2549 2550 int 2551 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2552 register_t *retval) 2553 { 2554 /* { 2555 syscallarg(const char *) path1; 2556 syscallarg(int) fd; 2557 syscallarg(const char *) path2; 2558 } */ 2559 2560 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2561 SCARG(uap, path2), UIO_USERSPACE); 2562 } 2563 2564 /* 2565 * Delete a whiteout from the filesystem. 2566 */ 2567 /* ARGSUSED */ 2568 int 2569 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2570 { 2571 /* { 2572 syscallarg(const char *) path; 2573 } */ 2574 int error; 2575 struct pathbuf *pb; 2576 struct nameidata nd; 2577 2578 error = pathbuf_copyin(SCARG(uap, path), &pb); 2579 if (error) { 2580 return error; 2581 } 2582 2583 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2584 error = namei(&nd); 2585 if (error) { 2586 pathbuf_destroy(pb); 2587 return (error); 2588 } 2589 2590 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2591 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2592 if (nd.ni_dvp == nd.ni_vp) 2593 vrele(nd.ni_dvp); 2594 else 2595 vput(nd.ni_dvp); 2596 if (nd.ni_vp) 2597 vrele(nd.ni_vp); 2598 pathbuf_destroy(pb); 2599 return (EEXIST); 2600 } 2601 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2602 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2603 vput(nd.ni_dvp); 2604 pathbuf_destroy(pb); 2605 return (error); 2606 } 2607 2608 /* 2609 * Delete a name from the filesystem. 2610 */ 2611 /* ARGSUSED */ 2612 int 2613 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2614 { 2615 /* { 2616 syscallarg(const char *) path; 2617 } */ 2618 2619 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2620 } 2621 2622 int 2623 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2624 register_t *retval) 2625 { 2626 /* { 2627 syscallarg(int) fd; 2628 syscallarg(const char *) path; 2629 syscallarg(int) flag; 2630 } */ 2631 2632 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2633 SCARG(uap, flag), UIO_USERSPACE); 2634 } 2635 2636 int 2637 do_sys_unlink(const char *arg, enum uio_seg seg) 2638 { 2639 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2640 } 2641 2642 static int 2643 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2644 enum uio_seg seg) 2645 { 2646 struct vnode *vp; 2647 int error; 2648 struct pathbuf *pb; 2649 struct nameidata nd; 2650 const char *pathstring; 2651 2652 KASSERT(l != NULL || fdat == AT_FDCWD); 2653 2654 error = pathbuf_maybe_copyin(arg, seg, &pb); 2655 if (error) { 2656 return error; 2657 } 2658 pathstring = pathbuf_stringcopy_get(pb); 2659 if (pathstring == NULL) { 2660 pathbuf_destroy(pb); 2661 return ENOMEM; 2662 } 2663 2664 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2665 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2666 goto out; 2667 vp = nd.ni_vp; 2668 2669 /* 2670 * The root of a mounted filesystem cannot be deleted. 2671 */ 2672 if ((vp->v_vflag & VV_ROOT) != 0) { 2673 error = EBUSY; 2674 goto abort; 2675 } 2676 2677 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2678 error = EBUSY; 2679 goto abort; 2680 } 2681 2682 /* 2683 * No rmdir "." please. 2684 */ 2685 if (nd.ni_dvp == vp) { 2686 error = EINVAL; 2687 goto abort; 2688 } 2689 2690 /* 2691 * AT_REMOVEDIR is required to remove a directory 2692 */ 2693 if (vp->v_type == VDIR) { 2694 if (!(flags & AT_REMOVEDIR)) { 2695 error = EPERM; 2696 goto abort; 2697 } else { 2698 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2699 vput(nd.ni_dvp); 2700 goto out; 2701 } 2702 } 2703 2704 /* 2705 * Starting here we only deal with non directories. 2706 */ 2707 if (flags & AT_REMOVEDIR) { 2708 error = ENOTDIR; 2709 goto abort; 2710 } 2711 2712 #if NVERIEXEC > 0 2713 /* Handle remove requests for veriexec entries. */ 2714 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2715 goto abort; 2716 } 2717 #endif /* NVERIEXEC > 0 */ 2718 2719 #ifdef FILEASSOC 2720 (void)fileassoc_file_delete(vp); 2721 #endif /* FILEASSOC */ 2722 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2723 vput(nd.ni_dvp); 2724 goto out; 2725 2726 abort: 2727 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2728 if (nd.ni_dvp == vp) 2729 vrele(nd.ni_dvp); 2730 else 2731 vput(nd.ni_dvp); 2732 vput(vp); 2733 2734 out: 2735 pathbuf_stringcopy_put(pb, pathstring); 2736 pathbuf_destroy(pb); 2737 return (error); 2738 } 2739 2740 /* 2741 * Reposition read/write file offset. 2742 */ 2743 int 2744 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2745 { 2746 /* { 2747 syscallarg(int) fd; 2748 syscallarg(int) pad; 2749 syscallarg(off_t) offset; 2750 syscallarg(int) whence; 2751 } */ 2752 kauth_cred_t cred = l->l_cred; 2753 file_t *fp; 2754 struct vnode *vp; 2755 struct vattr vattr; 2756 off_t newoff; 2757 int error, fd; 2758 2759 fd = SCARG(uap, fd); 2760 2761 if ((fp = fd_getfile(fd)) == NULL) 2762 return (EBADF); 2763 2764 vp = fp->f_vnode; 2765 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2766 error = ESPIPE; 2767 goto out; 2768 } 2769 2770 vn_lock(vp, LK_SHARED | LK_RETRY); 2771 2772 switch (SCARG(uap, whence)) { 2773 case SEEK_CUR: 2774 newoff = fp->f_offset + SCARG(uap, offset); 2775 break; 2776 case SEEK_END: 2777 error = VOP_GETATTR(vp, &vattr, cred); 2778 if (error) { 2779 VOP_UNLOCK(vp); 2780 goto out; 2781 } 2782 newoff = SCARG(uap, offset) + vattr.va_size; 2783 break; 2784 case SEEK_SET: 2785 newoff = SCARG(uap, offset); 2786 break; 2787 default: 2788 error = EINVAL; 2789 VOP_UNLOCK(vp); 2790 goto out; 2791 } 2792 VOP_UNLOCK(vp); 2793 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2794 *(off_t *)retval = fp->f_offset = newoff; 2795 } 2796 out: 2797 fd_putfile(fd); 2798 return (error); 2799 } 2800 2801 /* 2802 * Positional read system call. 2803 */ 2804 int 2805 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2806 { 2807 /* { 2808 syscallarg(int) fd; 2809 syscallarg(void *) buf; 2810 syscallarg(size_t) nbyte; 2811 syscallarg(off_t) offset; 2812 } */ 2813 file_t *fp; 2814 struct vnode *vp; 2815 off_t offset; 2816 int error, fd = SCARG(uap, fd); 2817 2818 if ((fp = fd_getfile(fd)) == NULL) 2819 return (EBADF); 2820 2821 if ((fp->f_flag & FREAD) == 0) { 2822 fd_putfile(fd); 2823 return (EBADF); 2824 } 2825 2826 vp = fp->f_vnode; 2827 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2828 error = ESPIPE; 2829 goto out; 2830 } 2831 2832 offset = SCARG(uap, offset); 2833 2834 /* 2835 * XXX This works because no file systems actually 2836 * XXX take any action on the seek operation. 2837 */ 2838 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2839 goto out; 2840 2841 /* dofileread() will unuse the descriptor for us */ 2842 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2843 &offset, 0, retval)); 2844 2845 out: 2846 fd_putfile(fd); 2847 return (error); 2848 } 2849 2850 /* 2851 * Positional scatter read system call. 2852 */ 2853 int 2854 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2855 { 2856 /* { 2857 syscallarg(int) fd; 2858 syscallarg(const struct iovec *) iovp; 2859 syscallarg(int) iovcnt; 2860 syscallarg(off_t) offset; 2861 } */ 2862 off_t offset = SCARG(uap, offset); 2863 2864 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2865 SCARG(uap, iovcnt), &offset, 0, retval); 2866 } 2867 2868 /* 2869 * Positional write system call. 2870 */ 2871 int 2872 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2873 { 2874 /* { 2875 syscallarg(int) fd; 2876 syscallarg(const void *) buf; 2877 syscallarg(size_t) nbyte; 2878 syscallarg(off_t) offset; 2879 } */ 2880 file_t *fp; 2881 struct vnode *vp; 2882 off_t offset; 2883 int error, fd = SCARG(uap, fd); 2884 2885 if ((fp = fd_getfile(fd)) == NULL) 2886 return (EBADF); 2887 2888 if ((fp->f_flag & FWRITE) == 0) { 2889 fd_putfile(fd); 2890 return (EBADF); 2891 } 2892 2893 vp = fp->f_vnode; 2894 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2895 error = ESPIPE; 2896 goto out; 2897 } 2898 2899 offset = SCARG(uap, offset); 2900 2901 /* 2902 * XXX This works because no file systems actually 2903 * XXX take any action on the seek operation. 2904 */ 2905 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2906 goto out; 2907 2908 /* dofilewrite() will unuse the descriptor for us */ 2909 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2910 &offset, 0, retval)); 2911 2912 out: 2913 fd_putfile(fd); 2914 return (error); 2915 } 2916 2917 /* 2918 * Positional gather write system call. 2919 */ 2920 int 2921 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2922 { 2923 /* { 2924 syscallarg(int) fd; 2925 syscallarg(const struct iovec *) iovp; 2926 syscallarg(int) iovcnt; 2927 syscallarg(off_t) offset; 2928 } */ 2929 off_t offset = SCARG(uap, offset); 2930 2931 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2932 SCARG(uap, iovcnt), &offset, 0, retval); 2933 } 2934 2935 /* 2936 * Check access permissions. 2937 */ 2938 int 2939 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2940 { 2941 /* { 2942 syscallarg(const char *) path; 2943 syscallarg(int) flags; 2944 } */ 2945 2946 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2947 SCARG(uap, flags), 0); 2948 } 2949 2950 int 2951 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2952 int mode, int flags) 2953 { 2954 kauth_cred_t cred; 2955 struct vnode *vp; 2956 int error, nd_flag, vmode; 2957 struct pathbuf *pb; 2958 struct nameidata nd; 2959 2960 CTASSERT(F_OK == 0); 2961 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2962 /* nonsense mode */ 2963 return EINVAL; 2964 } 2965 2966 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2967 if (flags & AT_SYMLINK_NOFOLLOW) 2968 nd_flag &= ~FOLLOW; 2969 2970 error = pathbuf_copyin(path, &pb); 2971 if (error) 2972 return error; 2973 2974 NDINIT(&nd, LOOKUP, nd_flag, pb); 2975 2976 /* Override default credentials */ 2977 cred = kauth_cred_dup(l->l_cred); 2978 if (!(flags & AT_EACCESS)) { 2979 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2980 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2981 } 2982 nd.ni_cnd.cn_cred = cred; 2983 2984 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2985 pathbuf_destroy(pb); 2986 goto out; 2987 } 2988 vp = nd.ni_vp; 2989 pathbuf_destroy(pb); 2990 2991 /* Flags == 0 means only check for existence. */ 2992 if (mode) { 2993 vmode = 0; 2994 if (mode & R_OK) 2995 vmode |= VREAD; 2996 if (mode & W_OK) 2997 vmode |= VWRITE; 2998 if (mode & X_OK) 2999 vmode |= VEXEC; 3000 3001 error = VOP_ACCESS(vp, vmode, cred); 3002 if (!error && (vmode & VWRITE)) 3003 error = vn_writechk(vp); 3004 } 3005 vput(vp); 3006 out: 3007 kauth_cred_free(cred); 3008 return (error); 3009 } 3010 3011 int 3012 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3013 register_t *retval) 3014 { 3015 /* { 3016 syscallarg(int) fd; 3017 syscallarg(const char *) path; 3018 syscallarg(int) amode; 3019 syscallarg(int) flag; 3020 } */ 3021 3022 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3023 SCARG(uap, amode), SCARG(uap, flag)); 3024 } 3025 3026 /* 3027 * Common code for all sys_stat functions, including compat versions. 3028 */ 3029 int 3030 do_sys_stat(const char *userpath, unsigned int nd_flag, 3031 struct stat *sb) 3032 { 3033 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3034 } 3035 3036 int 3037 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3038 unsigned int nd_flag, struct stat *sb) 3039 { 3040 int error; 3041 struct pathbuf *pb; 3042 struct nameidata nd; 3043 3044 KASSERT(l != NULL || fdat == AT_FDCWD); 3045 3046 error = pathbuf_copyin(userpath, &pb); 3047 if (error) { 3048 return error; 3049 } 3050 3051 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3052 3053 error = fd_nameiat(l, fdat, &nd); 3054 if (error != 0) { 3055 pathbuf_destroy(pb); 3056 return error; 3057 } 3058 error = vn_stat(nd.ni_vp, sb); 3059 vput(nd.ni_vp); 3060 pathbuf_destroy(pb); 3061 return error; 3062 } 3063 3064 /* 3065 * Get file status; this version follows links. 3066 */ 3067 /* ARGSUSED */ 3068 int 3069 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3070 { 3071 /* { 3072 syscallarg(const char *) path; 3073 syscallarg(struct stat *) ub; 3074 } */ 3075 struct stat sb; 3076 int error; 3077 3078 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3079 if (error) 3080 return error; 3081 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3082 } 3083 3084 /* 3085 * Get file status; this version does not follow links. 3086 */ 3087 /* ARGSUSED */ 3088 int 3089 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3090 { 3091 /* { 3092 syscallarg(const char *) path; 3093 syscallarg(struct stat *) ub; 3094 } */ 3095 struct stat sb; 3096 int error; 3097 3098 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3099 if (error) 3100 return error; 3101 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3102 } 3103 3104 int 3105 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3106 register_t *retval) 3107 { 3108 /* { 3109 syscallarg(int) fd; 3110 syscallarg(const char *) path; 3111 syscallarg(struct stat *) buf; 3112 syscallarg(int) flag; 3113 } */ 3114 unsigned int nd_flag; 3115 struct stat sb; 3116 int error; 3117 3118 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3119 nd_flag = NOFOLLOW; 3120 else 3121 nd_flag = FOLLOW; 3122 3123 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3124 &sb); 3125 if (error) 3126 return error; 3127 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3128 } 3129 3130 /* 3131 * Get configurable pathname variables. 3132 */ 3133 /* ARGSUSED */ 3134 int 3135 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3136 { 3137 /* { 3138 syscallarg(const char *) path; 3139 syscallarg(int) name; 3140 } */ 3141 int error; 3142 struct pathbuf *pb; 3143 struct nameidata nd; 3144 3145 error = pathbuf_copyin(SCARG(uap, path), &pb); 3146 if (error) { 3147 return error; 3148 } 3149 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3150 if ((error = namei(&nd)) != 0) { 3151 pathbuf_destroy(pb); 3152 return (error); 3153 } 3154 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3155 vput(nd.ni_vp); 3156 pathbuf_destroy(pb); 3157 return (error); 3158 } 3159 3160 /* 3161 * Return target name of a symbolic link. 3162 */ 3163 /* ARGSUSED */ 3164 int 3165 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3166 register_t *retval) 3167 { 3168 /* { 3169 syscallarg(const char *) path; 3170 syscallarg(char *) buf; 3171 syscallarg(size_t) count; 3172 } */ 3173 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3174 SCARG(uap, buf), SCARG(uap, count), retval); 3175 } 3176 3177 static int 3178 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3179 size_t count, register_t *retval) 3180 { 3181 struct vnode *vp; 3182 struct iovec aiov; 3183 struct uio auio; 3184 int error; 3185 struct pathbuf *pb; 3186 struct nameidata nd; 3187 3188 error = pathbuf_copyin(path, &pb); 3189 if (error) { 3190 return error; 3191 } 3192 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3193 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3194 pathbuf_destroy(pb); 3195 return error; 3196 } 3197 vp = nd.ni_vp; 3198 pathbuf_destroy(pb); 3199 if (vp->v_type != VLNK) 3200 error = EINVAL; 3201 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3202 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3203 aiov.iov_base = buf; 3204 aiov.iov_len = count; 3205 auio.uio_iov = &aiov; 3206 auio.uio_iovcnt = 1; 3207 auio.uio_offset = 0; 3208 auio.uio_rw = UIO_READ; 3209 KASSERT(l == curlwp); 3210 auio.uio_vmspace = l->l_proc->p_vmspace; 3211 auio.uio_resid = count; 3212 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3213 *retval = count - auio.uio_resid; 3214 } 3215 vput(vp); 3216 return (error); 3217 } 3218 3219 int 3220 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3221 register_t *retval) 3222 { 3223 /* { 3224 syscallarg(int) fd; 3225 syscallarg(const char *) path; 3226 syscallarg(char *) buf; 3227 syscallarg(size_t) bufsize; 3228 } */ 3229 3230 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3231 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3232 } 3233 3234 /* 3235 * Change flags of a file given a path name. 3236 */ 3237 /* ARGSUSED */ 3238 int 3239 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3240 { 3241 /* { 3242 syscallarg(const char *) path; 3243 syscallarg(u_long) flags; 3244 } */ 3245 struct vnode *vp; 3246 int error; 3247 3248 error = namei_simple_user(SCARG(uap, path), 3249 NSM_FOLLOW_TRYEMULROOT, &vp); 3250 if (error != 0) 3251 return (error); 3252 error = change_flags(vp, SCARG(uap, flags), l); 3253 vput(vp); 3254 return (error); 3255 } 3256 3257 /* 3258 * Change flags of a file given a file descriptor. 3259 */ 3260 /* ARGSUSED */ 3261 int 3262 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3263 { 3264 /* { 3265 syscallarg(int) fd; 3266 syscallarg(u_long) flags; 3267 } */ 3268 struct vnode *vp; 3269 file_t *fp; 3270 int error; 3271 3272 /* fd_getvnode() will use the descriptor for us */ 3273 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3274 return (error); 3275 vp = fp->f_vnode; 3276 error = change_flags(vp, SCARG(uap, flags), l); 3277 VOP_UNLOCK(vp); 3278 fd_putfile(SCARG(uap, fd)); 3279 return (error); 3280 } 3281 3282 /* 3283 * Change flags of a file given a path name; this version does 3284 * not follow links. 3285 */ 3286 int 3287 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3288 { 3289 /* { 3290 syscallarg(const char *) path; 3291 syscallarg(u_long) flags; 3292 } */ 3293 struct vnode *vp; 3294 int error; 3295 3296 error = namei_simple_user(SCARG(uap, path), 3297 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3298 if (error != 0) 3299 return (error); 3300 error = change_flags(vp, SCARG(uap, flags), l); 3301 vput(vp); 3302 return (error); 3303 } 3304 3305 /* 3306 * Common routine to change flags of a file. 3307 */ 3308 int 3309 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3310 { 3311 struct vattr vattr; 3312 int error; 3313 3314 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3315 3316 vattr_null(&vattr); 3317 vattr.va_flags = flags; 3318 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3319 3320 return (error); 3321 } 3322 3323 /* 3324 * Change mode of a file given path name; this version follows links. 3325 */ 3326 /* ARGSUSED */ 3327 int 3328 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3329 { 3330 /* { 3331 syscallarg(const char *) path; 3332 syscallarg(int) mode; 3333 } */ 3334 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3335 SCARG(uap, mode), 0); 3336 } 3337 3338 int 3339 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3340 { 3341 int error; 3342 struct vnode *vp; 3343 namei_simple_flags_t ns_flag; 3344 3345 if (flags & AT_SYMLINK_NOFOLLOW) 3346 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3347 else 3348 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3349 3350 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3351 if (error != 0) 3352 return error; 3353 3354 error = change_mode(vp, mode, l); 3355 3356 vrele(vp); 3357 3358 return (error); 3359 } 3360 3361 /* 3362 * Change mode of a file given a file descriptor. 3363 */ 3364 /* ARGSUSED */ 3365 int 3366 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3367 { 3368 /* { 3369 syscallarg(int) fd; 3370 syscallarg(int) mode; 3371 } */ 3372 file_t *fp; 3373 int error; 3374 3375 /* fd_getvnode() will use the descriptor for us */ 3376 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3377 return (error); 3378 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3379 fd_putfile(SCARG(uap, fd)); 3380 return (error); 3381 } 3382 3383 int 3384 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3385 register_t *retval) 3386 { 3387 /* { 3388 syscallarg(int) fd; 3389 syscallarg(const char *) path; 3390 syscallarg(int) mode; 3391 syscallarg(int) flag; 3392 } */ 3393 3394 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3395 SCARG(uap, mode), SCARG(uap, flag)); 3396 } 3397 3398 /* 3399 * Change mode of a file given path name; this version does not follow links. 3400 */ 3401 /* ARGSUSED */ 3402 int 3403 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3404 { 3405 /* { 3406 syscallarg(const char *) path; 3407 syscallarg(int) mode; 3408 } */ 3409 int error; 3410 struct vnode *vp; 3411 3412 error = namei_simple_user(SCARG(uap, path), 3413 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3414 if (error != 0) 3415 return (error); 3416 3417 error = change_mode(vp, SCARG(uap, mode), l); 3418 3419 vrele(vp); 3420 return (error); 3421 } 3422 3423 /* 3424 * Common routine to set mode given a vnode. 3425 */ 3426 static int 3427 change_mode(struct vnode *vp, int mode, struct lwp *l) 3428 { 3429 struct vattr vattr; 3430 int error; 3431 3432 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3433 vattr_null(&vattr); 3434 vattr.va_mode = mode & ALLPERMS; 3435 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3436 VOP_UNLOCK(vp); 3437 return (error); 3438 } 3439 3440 /* 3441 * Set ownership given a path name; this version follows links. 3442 */ 3443 /* ARGSUSED */ 3444 int 3445 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3446 { 3447 /* { 3448 syscallarg(const char *) path; 3449 syscallarg(uid_t) uid; 3450 syscallarg(gid_t) gid; 3451 } */ 3452 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3453 SCARG(uap, gid), 0); 3454 } 3455 3456 int 3457 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3458 gid_t gid, int flags) 3459 { 3460 int error; 3461 struct vnode *vp; 3462 namei_simple_flags_t ns_flag; 3463 3464 if (flags & AT_SYMLINK_NOFOLLOW) 3465 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3466 else 3467 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3468 3469 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3470 if (error != 0) 3471 return error; 3472 3473 error = change_owner(vp, uid, gid, l, 0); 3474 3475 vrele(vp); 3476 3477 return (error); 3478 } 3479 3480 /* 3481 * Set ownership given a path name; this version follows links. 3482 * Provides POSIX semantics. 3483 */ 3484 /* ARGSUSED */ 3485 int 3486 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3487 { 3488 /* { 3489 syscallarg(const char *) path; 3490 syscallarg(uid_t) uid; 3491 syscallarg(gid_t) gid; 3492 } */ 3493 int error; 3494 struct vnode *vp; 3495 3496 error = namei_simple_user(SCARG(uap, path), 3497 NSM_FOLLOW_TRYEMULROOT, &vp); 3498 if (error != 0) 3499 return (error); 3500 3501 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3502 3503 vrele(vp); 3504 return (error); 3505 } 3506 3507 /* 3508 * Set ownership given a file descriptor. 3509 */ 3510 /* ARGSUSED */ 3511 int 3512 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3513 { 3514 /* { 3515 syscallarg(int) fd; 3516 syscallarg(uid_t) uid; 3517 syscallarg(gid_t) gid; 3518 } */ 3519 int error; 3520 file_t *fp; 3521 3522 /* fd_getvnode() will use the descriptor for us */ 3523 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3524 return (error); 3525 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3526 l, 0); 3527 fd_putfile(SCARG(uap, fd)); 3528 return (error); 3529 } 3530 3531 int 3532 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3533 register_t *retval) 3534 { 3535 /* { 3536 syscallarg(int) fd; 3537 syscallarg(const char *) path; 3538 syscallarg(uid_t) owner; 3539 syscallarg(gid_t) group; 3540 syscallarg(int) flag; 3541 } */ 3542 3543 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3544 SCARG(uap, owner), SCARG(uap, group), 3545 SCARG(uap, flag)); 3546 } 3547 3548 /* 3549 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3550 */ 3551 /* ARGSUSED */ 3552 int 3553 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3554 { 3555 /* { 3556 syscallarg(int) fd; 3557 syscallarg(uid_t) uid; 3558 syscallarg(gid_t) gid; 3559 } */ 3560 int error; 3561 file_t *fp; 3562 3563 /* fd_getvnode() will use the descriptor for us */ 3564 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3565 return (error); 3566 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3567 l, 1); 3568 fd_putfile(SCARG(uap, fd)); 3569 return (error); 3570 } 3571 3572 /* 3573 * Set ownership given a path name; this version does not follow links. 3574 */ 3575 /* ARGSUSED */ 3576 int 3577 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3578 { 3579 /* { 3580 syscallarg(const char *) path; 3581 syscallarg(uid_t) uid; 3582 syscallarg(gid_t) gid; 3583 } */ 3584 int error; 3585 struct vnode *vp; 3586 3587 error = namei_simple_user(SCARG(uap, path), 3588 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3589 if (error != 0) 3590 return (error); 3591 3592 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3593 3594 vrele(vp); 3595 return (error); 3596 } 3597 3598 /* 3599 * Set ownership given a path name; this version does not follow links. 3600 * Provides POSIX/XPG semantics. 3601 */ 3602 /* ARGSUSED */ 3603 int 3604 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3605 { 3606 /* { 3607 syscallarg(const char *) path; 3608 syscallarg(uid_t) uid; 3609 syscallarg(gid_t) gid; 3610 } */ 3611 int error; 3612 struct vnode *vp; 3613 3614 error = namei_simple_user(SCARG(uap, path), 3615 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3616 if (error != 0) 3617 return (error); 3618 3619 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3620 3621 vrele(vp); 3622 return (error); 3623 } 3624 3625 /* 3626 * Common routine to set ownership given a vnode. 3627 */ 3628 static int 3629 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3630 int posix_semantics) 3631 { 3632 struct vattr vattr; 3633 mode_t newmode; 3634 int error; 3635 3636 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3637 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3638 goto out; 3639 3640 #define CHANGED(x) ((int)(x) != -1) 3641 newmode = vattr.va_mode; 3642 if (posix_semantics) { 3643 /* 3644 * POSIX/XPG semantics: if the caller is not the super-user, 3645 * clear set-user-id and set-group-id bits. Both POSIX and 3646 * the XPG consider the behaviour for calls by the super-user 3647 * implementation-defined; we leave the set-user-id and set- 3648 * group-id settings intact in that case. 3649 */ 3650 if (vattr.va_mode & S_ISUID) { 3651 if (kauth_authorize_vnode(l->l_cred, 3652 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3653 newmode &= ~S_ISUID; 3654 } 3655 if (vattr.va_mode & S_ISGID) { 3656 if (kauth_authorize_vnode(l->l_cred, 3657 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3658 newmode &= ~S_ISGID; 3659 } 3660 } else { 3661 /* 3662 * NetBSD semantics: when changing owner and/or group, 3663 * clear the respective bit(s). 3664 */ 3665 if (CHANGED(uid)) 3666 newmode &= ~S_ISUID; 3667 if (CHANGED(gid)) 3668 newmode &= ~S_ISGID; 3669 } 3670 /* Update va_mode iff altered. */ 3671 if (vattr.va_mode == newmode) 3672 newmode = VNOVAL; 3673 3674 vattr_null(&vattr); 3675 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3676 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3677 vattr.va_mode = newmode; 3678 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3679 #undef CHANGED 3680 3681 out: 3682 VOP_UNLOCK(vp); 3683 return (error); 3684 } 3685 3686 /* 3687 * Set the access and modification times given a path name; this 3688 * version follows links. 3689 */ 3690 /* ARGSUSED */ 3691 int 3692 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3693 register_t *retval) 3694 { 3695 /* { 3696 syscallarg(const char *) path; 3697 syscallarg(const struct timeval *) tptr; 3698 } */ 3699 3700 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3701 SCARG(uap, tptr), UIO_USERSPACE); 3702 } 3703 3704 /* 3705 * Set the access and modification times given a file descriptor. 3706 */ 3707 /* ARGSUSED */ 3708 int 3709 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3710 register_t *retval) 3711 { 3712 /* { 3713 syscallarg(int) fd; 3714 syscallarg(const struct timeval *) tptr; 3715 } */ 3716 int error; 3717 file_t *fp; 3718 3719 /* fd_getvnode() will use the descriptor for us */ 3720 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3721 return (error); 3722 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3723 UIO_USERSPACE); 3724 fd_putfile(SCARG(uap, fd)); 3725 return (error); 3726 } 3727 3728 int 3729 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3730 register_t *retval) 3731 { 3732 /* { 3733 syscallarg(int) fd; 3734 syscallarg(const struct timespec *) tptr; 3735 } */ 3736 int error; 3737 file_t *fp; 3738 3739 /* fd_getvnode() will use the descriptor for us */ 3740 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3741 return (error); 3742 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3743 SCARG(uap, tptr), UIO_USERSPACE); 3744 fd_putfile(SCARG(uap, fd)); 3745 return (error); 3746 } 3747 3748 /* 3749 * Set the access and modification times given a path name; this 3750 * version does not follow links. 3751 */ 3752 int 3753 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3754 register_t *retval) 3755 { 3756 /* { 3757 syscallarg(const char *) path; 3758 syscallarg(const struct timeval *) tptr; 3759 } */ 3760 3761 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3762 SCARG(uap, tptr), UIO_USERSPACE); 3763 } 3764 3765 int 3766 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3767 register_t *retval) 3768 { 3769 /* { 3770 syscallarg(int) fd; 3771 syscallarg(const char *) path; 3772 syscallarg(const struct timespec *) tptr; 3773 syscallarg(int) flag; 3774 } */ 3775 int follow; 3776 const struct timespec *tptr; 3777 int error; 3778 3779 tptr = SCARG(uap, tptr); 3780 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3781 3782 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3783 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3784 3785 return error; 3786 } 3787 3788 /* 3789 * Common routine to set access and modification times given a vnode. 3790 */ 3791 int 3792 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3793 const struct timespec *tptr, enum uio_seg seg) 3794 { 3795 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3796 } 3797 3798 int 3799 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3800 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3801 { 3802 struct vattr vattr; 3803 int error, dorele = 0; 3804 namei_simple_flags_t sflags; 3805 bool vanull, setbirthtime; 3806 struct timespec ts[2]; 3807 3808 KASSERT(l != NULL || fdat == AT_FDCWD); 3809 3810 /* 3811 * I have checked all callers and they pass either FOLLOW, 3812 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3813 * is 0. More to the point, they don't pass anything else. 3814 * Let's keep it that way at least until the namei interfaces 3815 * are fully sanitized. 3816 */ 3817 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3818 sflags = (flag == FOLLOW) ? 3819 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3820 3821 if (tptr == NULL) { 3822 vanull = true; 3823 nanotime(&ts[0]); 3824 ts[1] = ts[0]; 3825 } else { 3826 vanull = false; 3827 if (seg != UIO_SYSSPACE) { 3828 error = copyin(tptr, ts, sizeof (ts)); 3829 if (error != 0) 3830 return error; 3831 } else { 3832 ts[0] = tptr[0]; 3833 ts[1] = tptr[1]; 3834 } 3835 } 3836 3837 if (ts[0].tv_nsec == UTIME_NOW) { 3838 nanotime(&ts[0]); 3839 if (ts[1].tv_nsec == UTIME_NOW) { 3840 vanull = true; 3841 ts[1] = ts[0]; 3842 } 3843 } else if (ts[1].tv_nsec == UTIME_NOW) 3844 nanotime(&ts[1]); 3845 3846 if (vp == NULL) { 3847 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3848 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3849 if (error != 0) 3850 return error; 3851 dorele = 1; 3852 } 3853 3854 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3855 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3856 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3857 vattr_null(&vattr); 3858 3859 if (ts[0].tv_nsec != UTIME_OMIT) 3860 vattr.va_atime = ts[0]; 3861 3862 if (ts[1].tv_nsec != UTIME_OMIT) { 3863 vattr.va_mtime = ts[1]; 3864 if (setbirthtime) 3865 vattr.va_birthtime = ts[1]; 3866 } 3867 3868 if (vanull) 3869 vattr.va_vaflags |= VA_UTIMES_NULL; 3870 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3871 VOP_UNLOCK(vp); 3872 3873 if (dorele != 0) 3874 vrele(vp); 3875 3876 return error; 3877 } 3878 3879 int 3880 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3881 const struct timeval *tptr, enum uio_seg seg) 3882 { 3883 struct timespec ts[2]; 3884 struct timespec *tsptr = NULL; 3885 int error; 3886 3887 if (tptr != NULL) { 3888 struct timeval tv[2]; 3889 3890 if (seg != UIO_SYSSPACE) { 3891 error = copyin(tptr, tv, sizeof (tv)); 3892 if (error != 0) 3893 return error; 3894 tptr = tv; 3895 } 3896 3897 if ((tv[0].tv_usec == UTIME_NOW) || 3898 (tv[0].tv_usec == UTIME_OMIT)) 3899 ts[0].tv_nsec = tv[0].tv_usec; 3900 else 3901 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3902 3903 if ((tv[1].tv_usec == UTIME_NOW) || 3904 (tv[1].tv_usec == UTIME_OMIT)) 3905 ts[1].tv_nsec = tv[1].tv_usec; 3906 else 3907 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3908 3909 tsptr = &ts[0]; 3910 } 3911 3912 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3913 } 3914 3915 /* 3916 * Truncate a file given its path name. 3917 */ 3918 /* ARGSUSED */ 3919 int 3920 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3921 { 3922 /* { 3923 syscallarg(const char *) path; 3924 syscallarg(int) pad; 3925 syscallarg(off_t) length; 3926 } */ 3927 struct vnode *vp; 3928 struct vattr vattr; 3929 int error; 3930 3931 if (SCARG(uap, length) < 0) 3932 return EINVAL; 3933 3934 error = namei_simple_user(SCARG(uap, path), 3935 NSM_FOLLOW_TRYEMULROOT, &vp); 3936 if (error != 0) 3937 return (error); 3938 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3939 if (vp->v_type == VDIR) 3940 error = EISDIR; 3941 else if ((error = vn_writechk(vp)) == 0 && 3942 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3943 vattr_null(&vattr); 3944 vattr.va_size = SCARG(uap, length); 3945 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3946 } 3947 vput(vp); 3948 return (error); 3949 } 3950 3951 /* 3952 * Truncate a file given a file descriptor. 3953 */ 3954 /* ARGSUSED */ 3955 int 3956 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3957 { 3958 /* { 3959 syscallarg(int) fd; 3960 syscallarg(int) pad; 3961 syscallarg(off_t) length; 3962 } */ 3963 struct vattr vattr; 3964 struct vnode *vp; 3965 file_t *fp; 3966 int error; 3967 3968 if (SCARG(uap, length) < 0) 3969 return EINVAL; 3970 3971 /* fd_getvnode() will use the descriptor for us */ 3972 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3973 return (error); 3974 if ((fp->f_flag & FWRITE) == 0) { 3975 error = EINVAL; 3976 goto out; 3977 } 3978 vp = fp->f_vnode; 3979 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3980 if (vp->v_type == VDIR) 3981 error = EISDIR; 3982 else if ((error = vn_writechk(vp)) == 0) { 3983 vattr_null(&vattr); 3984 vattr.va_size = SCARG(uap, length); 3985 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3986 } 3987 VOP_UNLOCK(vp); 3988 out: 3989 fd_putfile(SCARG(uap, fd)); 3990 return (error); 3991 } 3992 3993 /* 3994 * Sync an open file. 3995 */ 3996 /* ARGSUSED */ 3997 int 3998 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3999 { 4000 /* { 4001 syscallarg(int) fd; 4002 } */ 4003 struct vnode *vp; 4004 file_t *fp; 4005 int error; 4006 4007 /* fd_getvnode() will use the descriptor for us */ 4008 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4009 return (error); 4010 vp = fp->f_vnode; 4011 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4012 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4013 VOP_UNLOCK(vp); 4014 fd_putfile(SCARG(uap, fd)); 4015 return (error); 4016 } 4017 4018 /* 4019 * Sync a range of file data. API modeled after that found in AIX. 4020 * 4021 * FDATASYNC indicates that we need only save enough metadata to be able 4022 * to re-read the written data. Note we duplicate AIX's requirement that 4023 * the file be open for writing. 4024 */ 4025 /* ARGSUSED */ 4026 int 4027 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4028 { 4029 /* { 4030 syscallarg(int) fd; 4031 syscallarg(int) flags; 4032 syscallarg(off_t) start; 4033 syscallarg(off_t) length; 4034 } */ 4035 struct vnode *vp; 4036 file_t *fp; 4037 int flags, nflags; 4038 off_t s, e, len; 4039 int error; 4040 4041 /* fd_getvnode() will use the descriptor for us */ 4042 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4043 return (error); 4044 4045 if ((fp->f_flag & FWRITE) == 0) { 4046 error = EBADF; 4047 goto out; 4048 } 4049 4050 flags = SCARG(uap, flags); 4051 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4052 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4053 error = EINVAL; 4054 goto out; 4055 } 4056 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4057 if (flags & FDATASYNC) 4058 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4059 else 4060 nflags = FSYNC_WAIT; 4061 if (flags & FDISKSYNC) 4062 nflags |= FSYNC_CACHE; 4063 4064 len = SCARG(uap, length); 4065 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4066 if (len) { 4067 s = SCARG(uap, start); 4068 e = s + len; 4069 if (e < s) { 4070 error = EINVAL; 4071 goto out; 4072 } 4073 } else { 4074 e = 0; 4075 s = 0; 4076 } 4077 4078 vp = fp->f_vnode; 4079 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4080 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4081 VOP_UNLOCK(vp); 4082 out: 4083 fd_putfile(SCARG(uap, fd)); 4084 return (error); 4085 } 4086 4087 /* 4088 * Sync the data of an open file. 4089 */ 4090 /* ARGSUSED */ 4091 int 4092 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4093 { 4094 /* { 4095 syscallarg(int) fd; 4096 } */ 4097 struct vnode *vp; 4098 file_t *fp; 4099 int error; 4100 4101 /* fd_getvnode() will use the descriptor for us */ 4102 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4103 return (error); 4104 if ((fp->f_flag & FWRITE) == 0) { 4105 fd_putfile(SCARG(uap, fd)); 4106 return (EBADF); 4107 } 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 KASSERT(fdvp != NULL); 4222 KASSERT(fvp != NULL); 4223 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4224 4225 /* 4226 * Make sure neither fdvp nor fvp is locked. 4227 */ 4228 if (fdvp != fvp) 4229 VOP_UNLOCK(fdvp); 4230 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4231 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4232 4233 /* 4234 * Reject renaming `.' and `..'. Can't do this until after 4235 * namei because we need namei's parsing to find the final 4236 * component name. (namei should just leave us with the final 4237 * component name and not look it up itself, but anyway...) 4238 * 4239 * This was here before because we used to relookup from 4240 * instead of to and relookup requires the caller to check 4241 * this, but now file systems may depend on this check, so we 4242 * must retain it until the file systems are all rototilled. 4243 */ 4244 if (((fnd.ni_cnd.cn_namelen == 1) && 4245 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4246 ((fnd.ni_cnd.cn_namelen == 2) && 4247 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4248 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4249 error = EINVAL; /* XXX EISDIR? */ 4250 goto abort0; 4251 } 4252 4253 /* 4254 * Lookup to. 4255 * 4256 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4257 * fvp here to decide whether to add CREATEDIR is a load of 4258 * bollocks because fvp might be the wrong node by now, since 4259 * fdvp is unlocked. 4260 * 4261 * XXX Why not pass CREATEDIR always? 4262 */ 4263 NDINIT(&tnd, RENAME, 4264 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4265 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4266 tpb); 4267 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4268 goto abort0; 4269 4270 /* 4271 * Pull out the important results of the lookup, tdvp and tvp. 4272 * Of course, tvp is bogus because we're about to unlock tdvp. 4273 */ 4274 tdvp = tnd.ni_dvp; 4275 tvp = tnd.ni_vp; 4276 KASSERT(tdvp != NULL); 4277 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4278 4279 /* 4280 * Make sure neither tdvp nor tvp is locked. 4281 */ 4282 if (tdvp != tvp) 4283 VOP_UNLOCK(tdvp); 4284 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4285 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4286 4287 /* 4288 * Reject renaming onto `.' or `..'. relookup is unhappy with 4289 * these, which is why we must do this here. Once upon a time 4290 * we relooked up from instead of to, and consequently didn't 4291 * need this check, but now that we relookup to instead of 4292 * from, we need this; and we shall need it forever forward 4293 * until the VOP_RENAME protocol changes, because file systems 4294 * will no doubt begin to depend on this check. 4295 */ 4296 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4297 error = EISDIR; 4298 goto abort1; 4299 } 4300 if ((tnd.ni_cnd.cn_namelen == 2) && 4301 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4302 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4303 error = EINVAL; 4304 goto abort1; 4305 } 4306 4307 /* 4308 * Get the mount point. If the file system has been unmounted, 4309 * which it may be because we're not holding any vnode locks, 4310 * then v_mount will be NULL. We're not really supposed to 4311 * read v_mount without holding the vnode lock, but since we 4312 * have fdvp referenced, if fdvp->v_mount changes then at worst 4313 * it will be set to NULL, not changed to another mount point. 4314 * And, of course, since it is up to the file system to 4315 * determine the real lock order, we can't lock both fdvp and 4316 * tdvp at the same time. 4317 */ 4318 mp = fdvp->v_mount; 4319 if (mp == NULL) { 4320 error = ENOENT; 4321 goto abort1; 4322 } 4323 4324 /* 4325 * Make sure the mount points match. Again, although we don't 4326 * hold any vnode locks, the v_mount fields may change -- but 4327 * at worst they will change to NULL, so this will never become 4328 * a cross-device rename, because we hold vnode references. 4329 * 4330 * XXX Because nothing is locked and the compiler may reorder 4331 * things here, unmounting the file system at an inopportune 4332 * moment may cause rename to fail with EXDEV when it really 4333 * should fail with ENOENT. 4334 */ 4335 tmp = tdvp->v_mount; 4336 if (tmp == NULL) { 4337 error = ENOENT; 4338 goto abort1; 4339 } 4340 4341 if (mp != tmp) { 4342 error = EXDEV; 4343 goto abort1; 4344 } 4345 4346 /* 4347 * Take the vfs rename lock to avoid cross-directory screw cases. 4348 * Nothing is locked currently, so taking this lock is safe. 4349 */ 4350 error = VFS_RENAMELOCK_ENTER(mp); 4351 if (error) 4352 goto abort1; 4353 4354 /* 4355 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4356 * and nothing is locked except for the vfs rename lock. 4357 * 4358 * The next step is a little rain dance to conform to the 4359 * insane lock protocol, even though it does nothing to ward 4360 * off race conditions. 4361 * 4362 * We need tdvp and tvp to be locked. However, because we have 4363 * unlocked tdvp in order to hold no locks while we take the 4364 * vfs rename lock, tvp may be wrong here, and we can't safely 4365 * lock it even if the sensible file systems will just unlock 4366 * it straight away. Consequently, we must lock tdvp and then 4367 * relookup tvp to get it locked. 4368 * 4369 * Finally, because the VOP_RENAME protocol is brain-damaged 4370 * and various file systems insanely depend on the semantics of 4371 * this brain damage, the lookup of to must be the last lookup 4372 * before VOP_RENAME. 4373 */ 4374 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4375 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4376 if (error) 4377 goto abort2; 4378 4379 /* 4380 * Drop the old tvp and pick up the new one -- which might be 4381 * the same, but that doesn't matter to us. After this, tdvp 4382 * and tvp should both be locked. 4383 */ 4384 if (tvp != NULL) 4385 vrele(tvp); 4386 tvp = tnd.ni_vp; 4387 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4388 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4389 4390 /* 4391 * The old do_sys_rename had various consistency checks here 4392 * involving fvp and tvp. fvp is bogus already here, and tvp 4393 * will become bogus soon in any sensible file system, so the 4394 * only purpose in putting these checks here is to give lip 4395 * service to these screw cases and to acknowledge that they 4396 * exist, not actually to handle them, but here you go 4397 * anyway... 4398 */ 4399 4400 /* 4401 * Acknowledge that directories and non-directories aren't 4402 * suposed to mix. 4403 */ 4404 if (tvp != NULL) { 4405 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4406 error = ENOTDIR; 4407 goto abort3; 4408 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4409 error = EISDIR; 4410 goto abort3; 4411 } 4412 } 4413 4414 /* 4415 * Acknowledge some random screw case, among the dozens that 4416 * might arise. 4417 */ 4418 if (fvp == tdvp) { 4419 error = EINVAL; 4420 goto abort3; 4421 } 4422 4423 /* 4424 * Acknowledge that POSIX has a wacky screw case. 4425 * 4426 * XXX Eventually the retain flag needs to be passed on to 4427 * VOP_RENAME. 4428 */ 4429 if (fvp == tvp) { 4430 if (retain) { 4431 error = 0; 4432 goto abort3; 4433 } else if ((fdvp == tdvp) && 4434 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4435 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4436 fnd.ni_cnd.cn_namelen))) { 4437 error = 0; 4438 goto abort3; 4439 } 4440 } 4441 4442 /* 4443 * Make sure veriexec can screw us up. (But a race can screw 4444 * up veriexec, of course -- remember, fvp and (soon) tvp are 4445 * bogus.) 4446 */ 4447 #if NVERIEXEC > 0 4448 { 4449 char *f1, *f2; 4450 size_t f1_len; 4451 size_t f2_len; 4452 4453 f1_len = fnd.ni_cnd.cn_namelen + 1; 4454 f1 = kmem_alloc(f1_len, KM_SLEEP); 4455 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4456 4457 f2_len = tnd.ni_cnd.cn_namelen + 1; 4458 f2 = kmem_alloc(f2_len, KM_SLEEP); 4459 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4460 4461 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4462 4463 kmem_free(f1, f1_len); 4464 kmem_free(f2, f2_len); 4465 4466 if (error) 4467 goto abort3; 4468 } 4469 #endif /* NVERIEXEC > 0 */ 4470 4471 /* 4472 * All ready. Incant the rename vop. 4473 */ 4474 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4475 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4476 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4477 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4478 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4479 4480 /* 4481 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4482 * tdvp and tvp. But we can't assert any of that. 4483 */ 4484 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4485 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4486 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4487 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4488 4489 /* 4490 * So all we have left to do is to drop the rename lock and 4491 * destroy the pathbufs. 4492 */ 4493 VFS_RENAMELOCK_EXIT(mp); 4494 goto out2; 4495 4496 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4497 VOP_UNLOCK(tvp); 4498 abort2: VOP_UNLOCK(tdvp); 4499 VFS_RENAMELOCK_EXIT(mp); 4500 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4501 vrele(tdvp); 4502 if (tvp != NULL) 4503 vrele(tvp); 4504 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4505 vrele(fdvp); 4506 vrele(fvp); 4507 out2: pathbuf_destroy(tpb); 4508 out1: pathbuf_destroy(fpb); 4509 out0: return error; 4510 } 4511 4512 /* 4513 * Make a directory file. 4514 */ 4515 /* ARGSUSED */ 4516 int 4517 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4518 { 4519 /* { 4520 syscallarg(const char *) path; 4521 syscallarg(int) mode; 4522 } */ 4523 4524 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4525 SCARG(uap, mode), UIO_USERSPACE); 4526 } 4527 4528 int 4529 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4530 register_t *retval) 4531 { 4532 /* { 4533 syscallarg(int) fd; 4534 syscallarg(const char *) path; 4535 syscallarg(int) mode; 4536 } */ 4537 4538 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4539 SCARG(uap, mode), UIO_USERSPACE); 4540 } 4541 4542 4543 int 4544 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4545 { 4546 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4547 } 4548 4549 static int 4550 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4551 enum uio_seg seg) 4552 { 4553 struct proc *p = curlwp->l_proc; 4554 struct vnode *vp; 4555 struct vattr vattr; 4556 int error; 4557 struct pathbuf *pb; 4558 struct nameidata nd; 4559 4560 KASSERT(l != NULL || fdat == AT_FDCWD); 4561 4562 /* XXX bollocks, should pass in a pathbuf */ 4563 error = pathbuf_maybe_copyin(path, seg, &pb); 4564 if (error) { 4565 return error; 4566 } 4567 4568 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4569 4570 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4571 pathbuf_destroy(pb); 4572 return (error); 4573 } 4574 vp = nd.ni_vp; 4575 if (vp != NULL) { 4576 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4577 if (nd.ni_dvp == vp) 4578 vrele(nd.ni_dvp); 4579 else 4580 vput(nd.ni_dvp); 4581 vrele(vp); 4582 pathbuf_destroy(pb); 4583 return (EEXIST); 4584 } 4585 vattr_null(&vattr); 4586 vattr.va_type = VDIR; 4587 /* We will read cwdi->cwdi_cmask unlocked. */ 4588 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4589 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4590 if (!error) 4591 vrele(nd.ni_vp); 4592 vput(nd.ni_dvp); 4593 pathbuf_destroy(pb); 4594 return (error); 4595 } 4596 4597 /* 4598 * Remove a directory file. 4599 */ 4600 /* ARGSUSED */ 4601 int 4602 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4603 { 4604 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4605 AT_REMOVEDIR, UIO_USERSPACE); 4606 } 4607 4608 /* 4609 * Read a block of directory entries in a file system independent format. 4610 */ 4611 int 4612 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4613 { 4614 /* { 4615 syscallarg(int) fd; 4616 syscallarg(char *) buf; 4617 syscallarg(size_t) count; 4618 } */ 4619 file_t *fp; 4620 int error, done; 4621 4622 /* fd_getvnode() will use the descriptor for us */ 4623 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4624 return (error); 4625 if ((fp->f_flag & FREAD) == 0) { 4626 error = EBADF; 4627 goto out; 4628 } 4629 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4630 SCARG(uap, count), &done, l, 0, 0); 4631 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4632 *retval = done; 4633 out: 4634 fd_putfile(SCARG(uap, fd)); 4635 return (error); 4636 } 4637 4638 /* 4639 * Set the mode mask for creation of filesystem nodes. 4640 */ 4641 int 4642 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4643 { 4644 /* { 4645 syscallarg(mode_t) newmask; 4646 } */ 4647 struct proc *p = l->l_proc; 4648 struct cwdinfo *cwdi; 4649 4650 /* 4651 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4652 * important is that we serialize changes to the mask. The 4653 * rw_exit() will issue a write memory barrier on our behalf, 4654 * and force the changes out to other CPUs (as it must use an 4655 * atomic operation, draining the local CPU's store buffers). 4656 */ 4657 cwdi = p->p_cwdi; 4658 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4659 *retval = cwdi->cwdi_cmask; 4660 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4661 rw_exit(&cwdi->cwdi_lock); 4662 4663 return (0); 4664 } 4665 4666 int 4667 dorevoke(struct vnode *vp, kauth_cred_t cred) 4668 { 4669 struct vattr vattr; 4670 int error, fs_decision; 4671 4672 vn_lock(vp, LK_SHARED | LK_RETRY); 4673 error = VOP_GETATTR(vp, &vattr, cred); 4674 VOP_UNLOCK(vp); 4675 if (error != 0) 4676 return error; 4677 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4678 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4679 fs_decision); 4680 if (!error) 4681 VOP_REVOKE(vp, REVOKEALL); 4682 return (error); 4683 } 4684 4685 /* 4686 * Void all references to file by ripping underlying filesystem 4687 * away from vnode. 4688 */ 4689 /* ARGSUSED */ 4690 int 4691 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4692 { 4693 /* { 4694 syscallarg(const char *) path; 4695 } */ 4696 struct vnode *vp; 4697 int error; 4698 4699 error = namei_simple_user(SCARG(uap, path), 4700 NSM_FOLLOW_TRYEMULROOT, &vp); 4701 if (error != 0) 4702 return (error); 4703 error = dorevoke(vp, l->l_cred); 4704 vrele(vp); 4705 return (error); 4706 } 4707 4708 /* 4709 * Allocate backing store for a file, filling a hole without having to 4710 * explicitly write anything out. 4711 */ 4712 /* ARGSUSED */ 4713 int 4714 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4715 register_t *retval) 4716 { 4717 /* { 4718 syscallarg(int) fd; 4719 syscallarg(off_t) pos; 4720 syscallarg(off_t) len; 4721 } */ 4722 int fd; 4723 off_t pos, len; 4724 struct file *fp; 4725 struct vnode *vp; 4726 int error; 4727 4728 fd = SCARG(uap, fd); 4729 pos = SCARG(uap, pos); 4730 len = SCARG(uap, len); 4731 4732 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4733 *retval = EINVAL; 4734 return 0; 4735 } 4736 4737 error = fd_getvnode(fd, &fp); 4738 if (error) { 4739 *retval = error; 4740 return 0; 4741 } 4742 if ((fp->f_flag & FWRITE) == 0) { 4743 error = EBADF; 4744 goto fail; 4745 } 4746 vp = fp->f_vnode; 4747 4748 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4749 if (vp->v_type == VDIR) { 4750 error = EISDIR; 4751 } else { 4752 error = VOP_FALLOCATE(vp, pos, len); 4753 } 4754 VOP_UNLOCK(vp); 4755 4756 fail: 4757 fd_putfile(fd); 4758 *retval = error; 4759 return 0; 4760 } 4761 4762 /* 4763 * Deallocate backing store for a file, creating a hole. Also used for 4764 * invoking TRIM on disks. 4765 */ 4766 /* ARGSUSED */ 4767 int 4768 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4769 register_t *retval) 4770 { 4771 /* { 4772 syscallarg(int) fd; 4773 syscallarg(off_t) pos; 4774 syscallarg(off_t) len; 4775 } */ 4776 int fd; 4777 off_t pos, len; 4778 struct file *fp; 4779 struct vnode *vp; 4780 int error; 4781 4782 fd = SCARG(uap, fd); 4783 pos = SCARG(uap, pos); 4784 len = SCARG(uap, len); 4785 4786 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4787 return EINVAL; 4788 } 4789 4790 error = fd_getvnode(fd, &fp); 4791 if (error) { 4792 return error; 4793 } 4794 if ((fp->f_flag & FWRITE) == 0) { 4795 error = EBADF; 4796 goto fail; 4797 } 4798 vp = fp->f_vnode; 4799 4800 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4801 if (vp->v_type == VDIR) { 4802 error = EISDIR; 4803 } else { 4804 error = VOP_FDISCARD(vp, pos, len); 4805 } 4806 VOP_UNLOCK(vp); 4807 4808 fail: 4809 fd_putfile(fd); 4810 return error; 4811 } 4812
Indexes created Fri Oct 03 02:09:57 GMT 2025