vfs_syscalls.c revision 1.457.2.4
1 /* $NetBSD: vfs_syscalls.c,v 1.457.2.4 2017/12/03 11:38:45 jdolecek 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 1.457.2.4 2017/12/03 11:38:45 jdolecek 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> 110 111 #include <miscfs/genfs/genfs.h> 112 #include <miscfs/specfs/specdev.h> 113 114 #include <nfs/rpcv2.h> 115 #include <nfs/nfsproto.h> 116 #include <nfs/nfs.h> 117 #include <nfs/nfs_var.h> 118 119 /* XXX this shouldn't be here */ 120 #ifndef OFF_T_MAX 121 #define OFF_T_MAX __type_max(off_t) 122 #endif 123 124 static int change_flags(struct vnode *, u_long, struct lwp *); 125 static int change_mode(struct vnode *, int, struct lwp *); 126 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 127 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 128 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 129 enum uio_seg); 130 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 131 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 132 enum uio_seg); 133 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 134 enum uio_seg, int); 135 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 136 size_t, register_t *); 137 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 138 139 static int fd_nameiat(struct lwp *, int, struct nameidata *); 140 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 141 namei_simple_flags_t, struct vnode **); 142 143 144 /* 145 * This table is used to maintain compatibility with 4.3BSD 146 * and NetBSD 0.9 mount syscalls - and possibly other systems. 147 * Note, the order is important! 148 * 149 * Do not modify this table. It should only contain filesystems 150 * supported by NetBSD 0.9 and 4.3BSD. 151 */ 152 const char * const mountcompatnames[] = { 153 NULL, /* 0 = MOUNT_NONE */ 154 MOUNT_FFS, /* 1 = MOUNT_UFS */ 155 MOUNT_NFS, /* 2 */ 156 MOUNT_MFS, /* 3 */ 157 MOUNT_MSDOS, /* 4 */ 158 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 159 MOUNT_FDESC, /* 6 */ 160 MOUNT_KERNFS, /* 7 */ 161 NULL, /* 8 = MOUNT_DEVFS */ 162 MOUNT_AFS, /* 9 */ 163 }; 164 165 const int nmountcompatnames = __arraycount(mountcompatnames); 166 167 static int 168 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 169 { 170 file_t *dfp; 171 int error; 172 173 if (fdat != AT_FDCWD) { 174 if ((error = fd_getvnode(fdat, &dfp)) != 0) 175 goto out; 176 177 NDAT(ndp, dfp->f_vnode); 178 } 179 180 error = namei(ndp); 181 182 if (fdat != AT_FDCWD) 183 fd_putfile(fdat); 184 out: 185 return error; 186 } 187 188 static int 189 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 190 namei_simple_flags_t sflags, struct vnode **vp_ret) 191 { 192 file_t *dfp; 193 struct vnode *dvp; 194 int error; 195 196 if (fdat != AT_FDCWD) { 197 if ((error = fd_getvnode(fdat, &dfp)) != 0) 198 goto out; 199 200 dvp = dfp->f_vnode; 201 } else { 202 dvp = NULL; 203 } 204 205 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 206 207 if (fdat != AT_FDCWD) 208 fd_putfile(fdat); 209 out: 210 return error; 211 } 212 213 static int 214 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 215 { 216 int error; 217 218 fp->f_flag = flags & FMASK; 219 fp->f_type = DTYPE_VNODE; 220 fp->f_ops = &vnops; 221 fp->f_vnode = vp; 222 223 if (flags & (O_EXLOCK | O_SHLOCK)) { 224 struct flock lf; 225 int type; 226 227 lf.l_whence = SEEK_SET; 228 lf.l_start = 0; 229 lf.l_len = 0; 230 if (flags & O_EXLOCK) 231 lf.l_type = F_WRLCK; 232 else 233 lf.l_type = F_RDLCK; 234 type = F_FLOCK; 235 if ((flags & FNONBLOCK) == 0) 236 type |= F_WAIT; 237 VOP_UNLOCK(vp); 238 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 239 if (error) { 240 (void) vn_close(vp, fp->f_flag, fp->f_cred); 241 fd_abort(l->l_proc, fp, indx); 242 return error; 243 } 244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 245 atomic_or_uint(&fp->f_flag, FHASLOCK); 246 } 247 if (flags & O_CLOEXEC) 248 fd_set_exclose(l, indx, true); 249 return 0; 250 } 251 252 static int 253 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 254 void *data, size_t *data_len) 255 { 256 struct mount *mp; 257 int error = 0, saved_flags; 258 259 mp = vp->v_mount; 260 saved_flags = mp->mnt_flag; 261 262 /* We can operate only on VV_ROOT nodes. */ 263 if ((vp->v_vflag & VV_ROOT) == 0) { 264 error = EINVAL; 265 goto out; 266 } 267 268 /* 269 * We only allow the filesystem to be reloaded if it 270 * is currently mounted read-only. Additionally, we 271 * prevent read-write to read-only downgrades. 272 */ 273 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 274 (mp->mnt_flag & MNT_RDONLY) == 0 && 275 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 276 error = EOPNOTSUPP; /* Needs translation */ 277 goto out; 278 } 279 280 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 281 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 282 if (error) 283 goto out; 284 285 error = vfs_suspend(mp, 0); 286 if (error) 287 goto out; 288 289 mutex_enter(&mp->mnt_updating); 290 291 mp->mnt_flag &= ~MNT_OP_FLAGS; 292 mp->mnt_flag |= flags & MNT_OP_FLAGS; 293 294 /* 295 * Set the mount level flags. 296 */ 297 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 298 if ((flags & MNT_RDONLY)) 299 mp->mnt_iflag |= IMNT_WANTRDONLY; 300 else 301 mp->mnt_iflag |= IMNT_WANTRDWR; 302 } 303 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 304 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 305 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 306 mp->mnt_flag &= ~MNT_RDONLY; 307 308 error = VFS_MOUNT(mp, path, data, data_len); 309 310 if (error && data != NULL) { 311 int error2; 312 313 /* 314 * Update failed; let's try and see if it was an 315 * export request. For compat with 3.0 and earlier. 316 */ 317 error2 = vfs_hooks_reexport(mp, path, data); 318 319 /* 320 * Only update error code if the export request was 321 * understood but some problem occurred while 322 * processing it. 323 */ 324 if (error2 != EJUSTRETURN) 325 error = error2; 326 } 327 328 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 329 mp->mnt_flag |= MNT_RDONLY; 330 if (error) 331 mp->mnt_flag = saved_flags; 332 mp->mnt_flag &= ~MNT_OP_FLAGS; 333 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 334 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 335 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 336 vfs_syncer_add_to_worklist(mp); 337 } else { 338 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 339 vfs_syncer_remove_from_worklist(mp); 340 } 341 mutex_exit(&mp->mnt_updating); 342 vfs_resume(mp); 343 344 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 345 (flags & MNT_EXTATTR)) { 346 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 347 NULL, 0, NULL) != 0) { 348 printf("%s: failed to start extattr, error = %d", 349 mp->mnt_stat.f_mntonname, error); 350 mp->mnt_flag &= ~MNT_EXTATTR; 351 } 352 } 353 354 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 355 !(flags & MNT_EXTATTR)) { 356 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 357 NULL, 0, NULL) != 0) { 358 printf("%s: failed to stop extattr, error = %d", 359 mp->mnt_stat.f_mntonname, error); 360 mp->mnt_flag |= MNT_RDONLY; 361 } 362 } 363 out: 364 return (error); 365 } 366 367 static int 368 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 369 struct vfsops **vfsops) 370 { 371 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 372 int error; 373 374 if (type_seg == UIO_USERSPACE) { 375 /* Copy file-system type from userspace. */ 376 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 377 } else { 378 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 379 KASSERT(error == 0); 380 } 381 382 if (error) { 383 /* 384 * Historically, filesystem types were identified by numbers. 385 * If we get an integer for the filesystem type instead of a 386 * string, we check to see if it matches one of the historic 387 * filesystem types. 388 */ 389 u_long fsindex = (u_long)fstype; 390 if (fsindex >= nmountcompatnames || 391 mountcompatnames[fsindex] == NULL) 392 return ENODEV; 393 strlcpy(fstypename, mountcompatnames[fsindex], 394 sizeof(fstypename)); 395 } 396 397 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 398 if (strcmp(fstypename, "ufs") == 0) 399 fstypename[0] = 'f'; 400 401 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 402 return 0; 403 404 /* If we can autoload a vfs module, try again */ 405 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 406 407 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 408 return 0; 409 410 return ENODEV; 411 } 412 413 static int 414 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 415 void *data, size_t *data_len) 416 { 417 struct mount *mp; 418 int error; 419 420 /* If MNT_GETARGS is specified, it should be the only flag. */ 421 if (flags & ~MNT_GETARGS) 422 return EINVAL; 423 424 mp = vp->v_mount; 425 426 /* XXX: probably some notion of "can see" here if we want isolation. */ 427 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 428 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 429 if (error) 430 return error; 431 432 if ((vp->v_vflag & VV_ROOT) == 0) 433 return EINVAL; 434 435 if (vfs_busy(mp)) 436 return EPERM; 437 438 mutex_enter(&mp->mnt_updating); 439 mp->mnt_flag &= ~MNT_OP_FLAGS; 440 mp->mnt_flag |= MNT_GETARGS; 441 error = VFS_MOUNT(mp, path, data, data_len); 442 mp->mnt_flag &= ~MNT_OP_FLAGS; 443 mutex_exit(&mp->mnt_updating); 444 445 vfs_unbusy(mp); 446 return (error); 447 } 448 449 int 450 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 451 { 452 /* { 453 syscallarg(const char *) type; 454 syscallarg(const char *) path; 455 syscallarg(int) flags; 456 syscallarg(void *) data; 457 syscallarg(size_t) data_len; 458 } */ 459 460 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 461 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 462 SCARG(uap, data_len), retval); 463 } 464 465 int 466 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 467 const char *path, int flags, void *data, enum uio_seg data_seg, 468 size_t data_len, register_t *retval) 469 { 470 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 471 struct vnode *vp; 472 void *data_buf = data; 473 bool vfsopsrele = false; 474 size_t alloc_sz = 0; 475 int error; 476 477 /* 478 * Get vnode to be covered 479 */ 480 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 481 if (error != 0) { 482 vp = NULL; 483 goto done; 484 } 485 486 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 487 vfsops = vp->v_mount->mnt_op; 488 } else { 489 /* 'type' is userspace */ 490 error = mount_get_vfsops(type, type_seg, &vfsops); 491 if (error != 0) 492 goto done; 493 vfsopsrele = true; 494 } 495 496 /* 497 * We allow data to be NULL, even for userspace. Some fs's don't need 498 * it. The others will handle NULL. 499 */ 500 if (data != NULL && data_seg == UIO_USERSPACE) { 501 if (data_len == 0) { 502 /* No length supplied, use default for filesystem */ 503 data_len = vfsops->vfs_min_mount_data; 504 505 /* 506 * Hopefully a longer buffer won't make copyin() fail. 507 * For compatibility with 3.0 and earlier. 508 */ 509 if (flags & MNT_UPDATE 510 && data_len < sizeof (struct mnt_export_args30)) 511 data_len = sizeof (struct mnt_export_args30); 512 } 513 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 514 error = EINVAL; 515 goto done; 516 } 517 alloc_sz = data_len; 518 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 519 520 /* NFS needs the buffer even for mnt_getargs .... */ 521 error = copyin(data, data_buf, data_len); 522 if (error != 0) 523 goto done; 524 } 525 526 if (flags & MNT_GETARGS) { 527 if (data_len == 0) { 528 error = EINVAL; 529 goto done; 530 } 531 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 532 if (error != 0) 533 goto done; 534 if (data_seg == UIO_USERSPACE) 535 error = copyout(data_buf, data, data_len); 536 *retval = data_len; 537 } else if (flags & MNT_UPDATE) { 538 error = mount_update(l, vp, path, flags, data_buf, &data_len); 539 } else { 540 /* Locking is handled internally in mount_domount(). */ 541 KASSERT(vfsopsrele == true); 542 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 543 &data_len); 544 vfsopsrele = false; 545 } 546 547 done: 548 if (vfsopsrele) 549 vfs_delref(vfsops); 550 if (vp != NULL) { 551 vrele(vp); 552 } 553 if (data_buf != data) 554 kmem_free(data_buf, alloc_sz); 555 return (error); 556 } 557 558 /* 559 * Unmount a file system. 560 * 561 * Note: unmount takes a path to the vnode mounted on as argument, 562 * not special file (as before). 563 */ 564 /* ARGSUSED */ 565 int 566 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 567 { 568 /* { 569 syscallarg(const char *) path; 570 syscallarg(int) flags; 571 } */ 572 struct vnode *vp; 573 struct mount *mp; 574 int error; 575 struct pathbuf *pb; 576 struct nameidata nd; 577 578 error = pathbuf_copyin(SCARG(uap, path), &pb); 579 if (error) { 580 return error; 581 } 582 583 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 584 if ((error = namei(&nd)) != 0) { 585 pathbuf_destroy(pb); 586 return error; 587 } 588 vp = nd.ni_vp; 589 pathbuf_destroy(pb); 590 591 mp = vp->v_mount; 592 vfs_ref(mp); 593 VOP_UNLOCK(vp); 594 595 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 596 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 597 if (error) { 598 vrele(vp); 599 vfs_rele(mp); 600 return (error); 601 } 602 603 /* 604 * Don't allow unmounting the root file system. 605 */ 606 if (mp->mnt_flag & MNT_ROOTFS) { 607 vrele(vp); 608 vfs_rele(mp); 609 return (EINVAL); 610 } 611 612 /* 613 * Must be the root of the filesystem 614 */ 615 if ((vp->v_vflag & VV_ROOT) == 0) { 616 vrele(vp); 617 vfs_rele(mp); 618 return (EINVAL); 619 } 620 621 vrele(vp); 622 error = dounmount(mp, SCARG(uap, flags), l); 623 vfs_rele(mp); 624 return error; 625 } 626 627 /* 628 * Sync each mounted filesystem. 629 */ 630 #ifdef DEBUG 631 int syncprt = 0; 632 struct ctldebug debug0 = { "syncprt", &syncprt }; 633 #endif 634 635 void 636 do_sys_sync(struct lwp *l) 637 { 638 mount_iterator_t *iter; 639 struct mount *mp; 640 int asyncflag; 641 642 mountlist_iterator_init(&iter); 643 while ((mp = mountlist_iterator_next(iter)) != NULL) { 644 mutex_enter(&mp->mnt_updating); 645 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 646 asyncflag = mp->mnt_flag & MNT_ASYNC; 647 mp->mnt_flag &= ~MNT_ASYNC; 648 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 649 if (asyncflag) 650 mp->mnt_flag |= MNT_ASYNC; 651 } 652 mutex_exit(&mp->mnt_updating); 653 } 654 mountlist_iterator_destroy(iter); 655 #ifdef DEBUG 656 if (syncprt) 657 vfs_bufstats(); 658 #endif /* DEBUG */ 659 } 660 661 /* ARGSUSED */ 662 int 663 sys_sync(struct lwp *l, const void *v, register_t *retval) 664 { 665 do_sys_sync(l); 666 return (0); 667 } 668 669 670 /* 671 * Access or change filesystem quotas. 672 * 673 * (this is really 14 different calls bundled into one) 674 */ 675 676 static int 677 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 678 { 679 struct quotastat info_k; 680 int error; 681 682 /* ensure any padding bytes are cleared */ 683 memset(&info_k, 0, sizeof(info_k)); 684 685 error = vfs_quotactl_stat(mp, &info_k); 686 if (error) { 687 return error; 688 } 689 690 return copyout(&info_k, info_u, sizeof(info_k)); 691 } 692 693 static int 694 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 695 struct quotaidtypestat *info_u) 696 { 697 struct quotaidtypestat info_k; 698 int error; 699 700 /* ensure any padding bytes are cleared */ 701 memset(&info_k, 0, sizeof(info_k)); 702 703 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 704 if (error) { 705 return error; 706 } 707 708 return copyout(&info_k, info_u, sizeof(info_k)); 709 } 710 711 static int 712 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 713 struct quotaobjtypestat *info_u) 714 { 715 struct quotaobjtypestat info_k; 716 int error; 717 718 /* ensure any padding bytes are cleared */ 719 memset(&info_k, 0, sizeof(info_k)); 720 721 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 722 if (error) { 723 return error; 724 } 725 726 return copyout(&info_k, info_u, sizeof(info_k)); 727 } 728 729 static int 730 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 731 struct quotaval *val_u) 732 { 733 struct quotakey key_k; 734 struct quotaval val_k; 735 int error; 736 737 /* ensure any padding bytes are cleared */ 738 memset(&val_k, 0, sizeof(val_k)); 739 740 error = copyin(key_u, &key_k, sizeof(key_k)); 741 if (error) { 742 return error; 743 } 744 745 error = vfs_quotactl_get(mp, &key_k, &val_k); 746 if (error) { 747 return error; 748 } 749 750 return copyout(&val_k, val_u, sizeof(val_k)); 751 } 752 753 static int 754 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 755 const struct quotaval *val_u) 756 { 757 struct quotakey key_k; 758 struct quotaval val_k; 759 int error; 760 761 error = copyin(key_u, &key_k, sizeof(key_k)); 762 if (error) { 763 return error; 764 } 765 766 error = copyin(val_u, &val_k, sizeof(val_k)); 767 if (error) { 768 return error; 769 } 770 771 return vfs_quotactl_put(mp, &key_k, &val_k); 772 } 773 774 static int 775 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 776 { 777 struct quotakey key_k; 778 int error; 779 780 error = copyin(key_u, &key_k, sizeof(key_k)); 781 if (error) { 782 return error; 783 } 784 785 return vfs_quotactl_del(mp, &key_k); 786 } 787 788 static int 789 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 790 { 791 struct quotakcursor cursor_k; 792 int error; 793 794 /* ensure any padding bytes are cleared */ 795 memset(&cursor_k, 0, sizeof(cursor_k)); 796 797 error = vfs_quotactl_cursoropen(mp, &cursor_k); 798 if (error) { 799 return error; 800 } 801 802 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 803 } 804 805 static int 806 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 807 { 808 struct quotakcursor cursor_k; 809 int error; 810 811 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 812 if (error) { 813 return error; 814 } 815 816 return vfs_quotactl_cursorclose(mp, &cursor_k); 817 } 818 819 static int 820 do_sys_quotactl_cursorskipidtype(struct mount *mp, 821 struct quotakcursor *cursor_u, int idtype) 822 { 823 struct quotakcursor cursor_k; 824 int error; 825 826 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 827 if (error) { 828 return error; 829 } 830 831 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 832 if (error) { 833 return error; 834 } 835 836 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 837 } 838 839 static int 840 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 841 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 842 unsigned *ret_u) 843 { 844 #define CGET_STACK_MAX 8 845 struct quotakcursor cursor_k; 846 struct quotakey stackkeys[CGET_STACK_MAX]; 847 struct quotaval stackvals[CGET_STACK_MAX]; 848 struct quotakey *keys_k; 849 struct quotaval *vals_k; 850 unsigned ret_k; 851 int error; 852 853 if (maxnum > 128) { 854 maxnum = 128; 855 } 856 857 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 858 if (error) { 859 return error; 860 } 861 862 if (maxnum <= CGET_STACK_MAX) { 863 keys_k = stackkeys; 864 vals_k = stackvals; 865 /* ensure any padding bytes are cleared */ 866 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 867 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 868 } else { 869 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 870 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 871 } 872 873 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 874 &ret_k); 875 if (error) { 876 goto fail; 877 } 878 879 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 880 if (error) { 881 goto fail; 882 } 883 884 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 885 if (error) { 886 goto fail; 887 } 888 889 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 890 if (error) { 891 goto fail; 892 } 893 894 /* do last to maximize the chance of being able to recover a failure */ 895 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 896 897 fail: 898 if (keys_k != stackkeys) { 899 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 900 } 901 if (vals_k != stackvals) { 902 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 903 } 904 return error; 905 } 906 907 static int 908 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 909 int *ret_u) 910 { 911 struct quotakcursor cursor_k; 912 int ret_k; 913 int error; 914 915 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 916 if (error) { 917 return error; 918 } 919 920 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 921 if (error) { 922 return error; 923 } 924 925 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 926 if (error) { 927 return error; 928 } 929 930 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 931 } 932 933 static int 934 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 935 { 936 struct quotakcursor cursor_k; 937 int error; 938 939 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 940 if (error) { 941 return error; 942 } 943 944 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 945 if (error) { 946 return error; 947 } 948 949 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 950 } 951 952 static int 953 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 954 { 955 char *path_k; 956 int error; 957 958 /* XXX this should probably be a struct pathbuf */ 959 path_k = PNBUF_GET(); 960 error = copyin(path_u, path_k, PATH_MAX); 961 if (error) { 962 PNBUF_PUT(path_k); 963 return error; 964 } 965 966 error = vfs_quotactl_quotaon(mp, idtype, path_k); 967 968 PNBUF_PUT(path_k); 969 return error; 970 } 971 972 static int 973 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 974 { 975 return vfs_quotactl_quotaoff(mp, idtype); 976 } 977 978 int 979 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 980 { 981 struct mount *mp; 982 struct vnode *vp; 983 int error; 984 985 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 986 if (error != 0) 987 return (error); 988 mp = vp->v_mount; 989 990 switch (args->qc_op) { 991 case QUOTACTL_STAT: 992 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 993 break; 994 case QUOTACTL_IDTYPESTAT: 995 error = do_sys_quotactl_idtypestat(mp, 996 args->u.idtypestat.qc_idtype, 997 args->u.idtypestat.qc_info); 998 break; 999 case QUOTACTL_OBJTYPESTAT: 1000 error = do_sys_quotactl_objtypestat(mp, 1001 args->u.objtypestat.qc_objtype, 1002 args->u.objtypestat.qc_info); 1003 break; 1004 case QUOTACTL_GET: 1005 error = do_sys_quotactl_get(mp, 1006 args->u.get.qc_key, 1007 args->u.get.qc_val); 1008 break; 1009 case QUOTACTL_PUT: 1010 error = do_sys_quotactl_put(mp, 1011 args->u.put.qc_key, 1012 args->u.put.qc_val); 1013 break; 1014 case QUOTACTL_DEL: 1015 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1016 break; 1017 case QUOTACTL_CURSOROPEN: 1018 error = do_sys_quotactl_cursoropen(mp, 1019 args->u.cursoropen.qc_cursor); 1020 break; 1021 case QUOTACTL_CURSORCLOSE: 1022 error = do_sys_quotactl_cursorclose(mp, 1023 args->u.cursorclose.qc_cursor); 1024 break; 1025 case QUOTACTL_CURSORSKIPIDTYPE: 1026 error = do_sys_quotactl_cursorskipidtype(mp, 1027 args->u.cursorskipidtype.qc_cursor, 1028 args->u.cursorskipidtype.qc_idtype); 1029 break; 1030 case QUOTACTL_CURSORGET: 1031 error = do_sys_quotactl_cursorget(mp, 1032 args->u.cursorget.qc_cursor, 1033 args->u.cursorget.qc_keys, 1034 args->u.cursorget.qc_vals, 1035 args->u.cursorget.qc_maxnum, 1036 args->u.cursorget.qc_ret); 1037 break; 1038 case QUOTACTL_CURSORATEND: 1039 error = do_sys_quotactl_cursoratend(mp, 1040 args->u.cursoratend.qc_cursor, 1041 args->u.cursoratend.qc_ret); 1042 break; 1043 case QUOTACTL_CURSORREWIND: 1044 error = do_sys_quotactl_cursorrewind(mp, 1045 args->u.cursorrewind.qc_cursor); 1046 break; 1047 case QUOTACTL_QUOTAON: 1048 error = do_sys_quotactl_quotaon(mp, 1049 args->u.quotaon.qc_idtype, 1050 args->u.quotaon.qc_quotafile); 1051 break; 1052 case QUOTACTL_QUOTAOFF: 1053 error = do_sys_quotactl_quotaoff(mp, 1054 args->u.quotaoff.qc_idtype); 1055 break; 1056 default: 1057 error = EINVAL; 1058 break; 1059 } 1060 1061 vrele(vp); 1062 return error; 1063 } 1064 1065 /* ARGSUSED */ 1066 int 1067 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1068 register_t *retval) 1069 { 1070 /* { 1071 syscallarg(const char *) path; 1072 syscallarg(struct quotactl_args *) args; 1073 } */ 1074 struct quotactl_args args; 1075 int error; 1076 1077 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1078 if (error) { 1079 return error; 1080 } 1081 1082 return do_sys_quotactl(SCARG(uap, path), &args); 1083 } 1084 1085 int 1086 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1087 int root) 1088 { 1089 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1090 int error = 0; 1091 1092 /* 1093 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1094 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1095 * overrides MNT_NOWAIT. 1096 */ 1097 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1098 (flags != MNT_WAIT && flags != 0)) { 1099 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1100 goto done; 1101 } 1102 1103 /* Get the filesystem stats now */ 1104 memset(sp, 0, sizeof(*sp)); 1105 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1106 return error; 1107 } 1108 1109 if (cwdi->cwdi_rdir == NULL) 1110 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1111 done: 1112 if (cwdi->cwdi_rdir != NULL) { 1113 size_t len; 1114 char *bp; 1115 char c; 1116 char *path = PNBUF_GET(); 1117 1118 bp = path + MAXPATHLEN; 1119 *--bp = '\0'; 1120 rw_enter(&cwdi->cwdi_lock, RW_READER); 1121 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1122 MAXPATHLEN / 2, 0, l); 1123 rw_exit(&cwdi->cwdi_lock); 1124 if (error) { 1125 PNBUF_PUT(path); 1126 return error; 1127 } 1128 len = strlen(bp); 1129 if (len != 1) { 1130 /* 1131 * for mount points that are below our root, we can see 1132 * them, so we fix up the pathname and return them. The 1133 * rest we cannot see, so we don't allow viewing the 1134 * data. 1135 */ 1136 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1137 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1138 (void)strlcpy(sp->f_mntonname, 1139 c == '\0' ? "/" : &sp->f_mntonname[len], 1140 sizeof(sp->f_mntonname)); 1141 } else { 1142 if (root) 1143 (void)strlcpy(sp->f_mntonname, "/", 1144 sizeof(sp->f_mntonname)); 1145 else 1146 error = EPERM; 1147 } 1148 } 1149 PNBUF_PUT(path); 1150 } 1151 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1152 return error; 1153 } 1154 1155 /* 1156 * Get filesystem statistics by path. 1157 */ 1158 int 1159 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1160 { 1161 struct mount *mp; 1162 int error; 1163 struct vnode *vp; 1164 1165 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1166 if (error != 0) 1167 return error; 1168 mp = vp->v_mount; 1169 error = dostatvfs(mp, sb, l, flags, 1); 1170 vrele(vp); 1171 return error; 1172 } 1173 1174 /* ARGSUSED */ 1175 int 1176 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1177 { 1178 /* { 1179 syscallarg(const char *) path; 1180 syscallarg(struct statvfs *) buf; 1181 syscallarg(int) flags; 1182 } */ 1183 struct statvfs *sb; 1184 int error; 1185 1186 sb = STATVFSBUF_GET(); 1187 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1188 if (error == 0) 1189 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1190 STATVFSBUF_PUT(sb); 1191 return error; 1192 } 1193 1194 /* 1195 * Get filesystem statistics by fd. 1196 */ 1197 int 1198 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1199 { 1200 file_t *fp; 1201 struct mount *mp; 1202 int error; 1203 1204 /* fd_getvnode() will use the descriptor for us */ 1205 if ((error = fd_getvnode(fd, &fp)) != 0) 1206 return (error); 1207 mp = fp->f_vnode->v_mount; 1208 error = dostatvfs(mp, sb, curlwp, flags, 1); 1209 fd_putfile(fd); 1210 return error; 1211 } 1212 1213 /* ARGSUSED */ 1214 int 1215 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1216 { 1217 /* { 1218 syscallarg(int) fd; 1219 syscallarg(struct statvfs *) buf; 1220 syscallarg(int) flags; 1221 } */ 1222 struct statvfs *sb; 1223 int error; 1224 1225 sb = STATVFSBUF_GET(); 1226 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1227 if (error == 0) 1228 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1229 STATVFSBUF_PUT(sb); 1230 return error; 1231 } 1232 1233 1234 /* 1235 * Get statistics on all filesystems. 1236 */ 1237 int 1238 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1239 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1240 register_t *retval) 1241 { 1242 int root = 0; 1243 mount_iterator_t *iter; 1244 struct proc *p = l->l_proc; 1245 struct mount *mp; 1246 struct statvfs *sb; 1247 size_t count, maxcount; 1248 int error = 0; 1249 1250 sb = STATVFSBUF_GET(); 1251 maxcount = bufsize / entry_sz; 1252 count = 0; 1253 mountlist_iterator_init(&iter); 1254 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1255 if (sfsp && count < maxcount) { 1256 error = dostatvfs(mp, sb, l, flags, 0); 1257 if (error) { 1258 error = 0; 1259 continue; 1260 } 1261 error = copyfn(sb, sfsp, entry_sz); 1262 if (error) 1263 goto out; 1264 sfsp = (char *)sfsp + entry_sz; 1265 root |= strcmp(sb->f_mntonname, "/") == 0; 1266 } 1267 count++; 1268 } 1269 1270 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1271 /* 1272 * fake a root entry 1273 */ 1274 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1275 sb, l, flags, 1); 1276 if (error != 0) 1277 goto out; 1278 if (sfsp) { 1279 error = copyfn(sb, sfsp, entry_sz); 1280 if (error != 0) 1281 goto out; 1282 } 1283 count++; 1284 } 1285 if (sfsp && count > maxcount) 1286 *retval = maxcount; 1287 else 1288 *retval = count; 1289 out: 1290 mountlist_iterator_destroy(iter); 1291 STATVFSBUF_PUT(sb); 1292 return error; 1293 } 1294 1295 int 1296 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1297 { 1298 /* { 1299 syscallarg(struct statvfs *) buf; 1300 syscallarg(size_t) bufsize; 1301 syscallarg(int) flags; 1302 } */ 1303 1304 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1305 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1306 } 1307 1308 /* 1309 * Change current working directory to a given file descriptor. 1310 */ 1311 /* ARGSUSED */ 1312 int 1313 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1314 { 1315 /* { 1316 syscallarg(int) fd; 1317 } */ 1318 struct proc *p = l->l_proc; 1319 struct cwdinfo *cwdi; 1320 struct vnode *vp, *tdp; 1321 struct mount *mp; 1322 file_t *fp; 1323 int error, fd; 1324 1325 /* fd_getvnode() will use the descriptor for us */ 1326 fd = SCARG(uap, fd); 1327 if ((error = fd_getvnode(fd, &fp)) != 0) 1328 return (error); 1329 vp = fp->f_vnode; 1330 1331 vref(vp); 1332 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1333 if (vp->v_type != VDIR) 1334 error = ENOTDIR; 1335 else 1336 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1337 if (error) { 1338 vput(vp); 1339 goto out; 1340 } 1341 while ((mp = vp->v_mountedhere) != NULL) { 1342 error = vfs_busy(mp); 1343 vput(vp); 1344 if (error != 0) 1345 goto out; 1346 error = VFS_ROOT(mp, &tdp); 1347 vfs_unbusy(mp); 1348 if (error) 1349 goto out; 1350 vp = tdp; 1351 } 1352 VOP_UNLOCK(vp); 1353 1354 /* 1355 * Disallow changing to a directory not under the process's 1356 * current root directory (if there is one). 1357 */ 1358 cwdi = p->p_cwdi; 1359 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1360 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1361 vrele(vp); 1362 error = EPERM; /* operation not permitted */ 1363 } else { 1364 vrele(cwdi->cwdi_cdir); 1365 cwdi->cwdi_cdir = vp; 1366 } 1367 rw_exit(&cwdi->cwdi_lock); 1368 1369 out: 1370 fd_putfile(fd); 1371 return (error); 1372 } 1373 1374 /* 1375 * Change this process's notion of the root directory to a given file 1376 * descriptor. 1377 */ 1378 int 1379 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1380 { 1381 struct proc *p = l->l_proc; 1382 struct vnode *vp; 1383 file_t *fp; 1384 int error, fd = SCARG(uap, fd); 1385 1386 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1387 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1388 return error; 1389 /* fd_getvnode() will use the descriptor for us */ 1390 if ((error = fd_getvnode(fd, &fp)) != 0) 1391 return error; 1392 vp = fp->f_vnode; 1393 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1394 if (vp->v_type != VDIR) 1395 error = ENOTDIR; 1396 else 1397 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1398 VOP_UNLOCK(vp); 1399 if (error) 1400 goto out; 1401 vref(vp); 1402 1403 change_root(p->p_cwdi, vp, l); 1404 1405 out: 1406 fd_putfile(fd); 1407 return (error); 1408 } 1409 1410 /* 1411 * Change current working directory (``.''). 1412 */ 1413 /* ARGSUSED */ 1414 int 1415 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1416 { 1417 /* { 1418 syscallarg(const char *) path; 1419 } */ 1420 struct proc *p = l->l_proc; 1421 struct cwdinfo *cwdi; 1422 int error; 1423 struct vnode *vp; 1424 1425 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1426 &vp, l)) != 0) 1427 return (error); 1428 cwdi = p->p_cwdi; 1429 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1430 vrele(cwdi->cwdi_cdir); 1431 cwdi->cwdi_cdir = vp; 1432 rw_exit(&cwdi->cwdi_lock); 1433 return (0); 1434 } 1435 1436 /* 1437 * Change notion of root (``/'') directory. 1438 */ 1439 /* ARGSUSED */ 1440 int 1441 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1442 { 1443 /* { 1444 syscallarg(const char *) path; 1445 } */ 1446 struct proc *p = l->l_proc; 1447 int error; 1448 struct vnode *vp; 1449 1450 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1451 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1452 return (error); 1453 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1454 &vp, l)) != 0) 1455 return (error); 1456 1457 change_root(p->p_cwdi, vp, l); 1458 1459 return (0); 1460 } 1461 1462 /* 1463 * Common routine for chroot and fchroot. 1464 * NB: callers need to properly authorize the change root operation. 1465 */ 1466 void 1467 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1468 { 1469 struct proc *p = l->l_proc; 1470 kauth_cred_t ncred; 1471 1472 ncred = kauth_cred_alloc(); 1473 1474 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1475 if (cwdi->cwdi_rdir != NULL) 1476 vrele(cwdi->cwdi_rdir); 1477 cwdi->cwdi_rdir = vp; 1478 1479 /* 1480 * Prevent escaping from chroot by putting the root under 1481 * the working directory. Silently chdir to / if we aren't 1482 * already there. 1483 */ 1484 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1485 /* 1486 * XXX would be more failsafe to change directory to a 1487 * deadfs node here instead 1488 */ 1489 vrele(cwdi->cwdi_cdir); 1490 vref(vp); 1491 cwdi->cwdi_cdir = vp; 1492 } 1493 rw_exit(&cwdi->cwdi_lock); 1494 1495 /* Get a write lock on the process credential. */ 1496 proc_crmod_enter(); 1497 1498 kauth_cred_clone(p->p_cred, ncred); 1499 kauth_proc_chroot(ncred, p->p_cwdi); 1500 1501 /* Broadcast our credentials to the process and other LWPs. */ 1502 proc_crmod_leave(ncred, p->p_cred, true); 1503 } 1504 1505 /* 1506 * Common routine for chroot and chdir. 1507 * XXX "where" should be enum uio_seg 1508 */ 1509 int 1510 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1511 { 1512 struct pathbuf *pb; 1513 struct nameidata nd; 1514 int error; 1515 1516 error = pathbuf_maybe_copyin(path, where, &pb); 1517 if (error) { 1518 return error; 1519 } 1520 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1521 if ((error = namei(&nd)) != 0) { 1522 pathbuf_destroy(pb); 1523 return error; 1524 } 1525 *vpp = nd.ni_vp; 1526 pathbuf_destroy(pb); 1527 1528 if ((*vpp)->v_type != VDIR) 1529 error = ENOTDIR; 1530 else 1531 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1532 1533 if (error) 1534 vput(*vpp); 1535 else 1536 VOP_UNLOCK(*vpp); 1537 return (error); 1538 } 1539 1540 /* 1541 * Internals of sys_open - path has already been converted into a pathbuf 1542 * (so we can easily reuse this function from other parts of the kernel, 1543 * like posix_spawn post-processing). 1544 */ 1545 int 1546 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1547 int open_mode, int *fd) 1548 { 1549 struct proc *p = l->l_proc; 1550 struct cwdinfo *cwdi = p->p_cwdi; 1551 file_t *fp; 1552 struct vnode *vp; 1553 int flags, cmode; 1554 int indx, error; 1555 struct nameidata nd; 1556 1557 if (open_flags & O_SEARCH) { 1558 open_flags &= ~(int)O_SEARCH; 1559 } 1560 1561 flags = FFLAGS(open_flags); 1562 if ((flags & (FREAD | FWRITE)) == 0) 1563 return EINVAL; 1564 1565 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1566 return error; 1567 } 1568 1569 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1570 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1571 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1572 if (dvp != NULL) 1573 NDAT(&nd, dvp); 1574 1575 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1576 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1577 fd_abort(p, fp, indx); 1578 if ((error == EDUPFD || error == EMOVEFD) && 1579 l->l_dupfd >= 0 && /* XXX from fdopen */ 1580 (error = 1581 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1582 *fd = indx; 1583 return 0; 1584 } 1585 if (error == ERESTART) 1586 error = EINTR; 1587 return error; 1588 } 1589 1590 l->l_dupfd = 0; 1591 vp = nd.ni_vp; 1592 1593 if ((error = open_setfp(l, fp, vp, indx, flags))) 1594 return error; 1595 1596 VOP_UNLOCK(vp); 1597 *fd = indx; 1598 fd_affix(p, fp, indx); 1599 return 0; 1600 } 1601 1602 int 1603 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1604 { 1605 struct pathbuf *pb; 1606 int error, oflags; 1607 1608 oflags = FFLAGS(open_flags); 1609 if ((oflags & (FREAD | FWRITE)) == 0) 1610 return EINVAL; 1611 1612 pb = pathbuf_create(path); 1613 if (pb == NULL) 1614 return ENOMEM; 1615 1616 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1617 pathbuf_destroy(pb); 1618 1619 return error; 1620 } 1621 1622 /* 1623 * Check permissions, allocate an open file structure, 1624 * and call the device open routine if any. 1625 */ 1626 static int 1627 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1628 int mode, int *fd) 1629 { 1630 file_t *dfp = NULL; 1631 struct vnode *dvp = NULL; 1632 struct pathbuf *pb; 1633 int error; 1634 1635 #ifdef COMPAT_10 /* XXX: and perhaps later */ 1636 if (path == NULL) { 1637 pb = pathbuf_create("."); 1638 if (pb == NULL) 1639 return ENOMEM; 1640 } else 1641 #endif 1642 { 1643 error = pathbuf_copyin(path, &pb); 1644 if (error) 1645 return error; 1646 } 1647 1648 if (fdat != AT_FDCWD) { 1649 /* fd_getvnode() will use the descriptor for us */ 1650 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1651 goto out; 1652 1653 dvp = dfp->f_vnode; 1654 } 1655 1656 error = do_open(l, dvp, pb, flags, mode, fd); 1657 1658 if (dfp != NULL) 1659 fd_putfile(fdat); 1660 out: 1661 pathbuf_destroy(pb); 1662 return error; 1663 } 1664 1665 int 1666 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1667 { 1668 /* { 1669 syscallarg(const char *) path; 1670 syscallarg(int) flags; 1671 syscallarg(int) mode; 1672 } */ 1673 int error; 1674 int fd; 1675 1676 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1677 SCARG(uap, flags), SCARG(uap, mode), &fd); 1678 1679 if (error == 0) 1680 *retval = fd; 1681 1682 return error; 1683 } 1684 1685 int 1686 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1687 { 1688 /* { 1689 syscallarg(int) fd; 1690 syscallarg(const char *) path; 1691 syscallarg(int) oflags; 1692 syscallarg(int) mode; 1693 } */ 1694 int error; 1695 int fd; 1696 1697 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1698 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1699 1700 if (error == 0) 1701 *retval = fd; 1702 1703 return error; 1704 } 1705 1706 static void 1707 vfs__fhfree(fhandle_t *fhp) 1708 { 1709 size_t fhsize; 1710 1711 fhsize = FHANDLE_SIZE(fhp); 1712 kmem_free(fhp, fhsize); 1713 } 1714 1715 /* 1716 * vfs_composefh: compose a filehandle. 1717 */ 1718 1719 int 1720 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1721 { 1722 struct mount *mp; 1723 struct fid *fidp; 1724 int error; 1725 size_t needfhsize; 1726 size_t fidsize; 1727 1728 mp = vp->v_mount; 1729 fidp = NULL; 1730 if (*fh_size < FHANDLE_SIZE_MIN) { 1731 fidsize = 0; 1732 } else { 1733 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1734 if (fhp != NULL) { 1735 memset(fhp, 0, *fh_size); 1736 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1737 fidp = &fhp->fh_fid; 1738 } 1739 } 1740 error = VFS_VPTOFH(vp, fidp, &fidsize); 1741 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1742 if (error == 0 && *fh_size < needfhsize) { 1743 error = E2BIG; 1744 } 1745 *fh_size = needfhsize; 1746 return error; 1747 } 1748 1749 int 1750 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1751 { 1752 struct mount *mp; 1753 fhandle_t *fhp; 1754 size_t fhsize; 1755 size_t fidsize; 1756 int error; 1757 1758 mp = vp->v_mount; 1759 fidsize = 0; 1760 error = VFS_VPTOFH(vp, NULL, &fidsize); 1761 KASSERT(error != 0); 1762 if (error != E2BIG) { 1763 goto out; 1764 } 1765 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1766 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1767 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1768 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1769 if (error == 0) { 1770 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1771 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1772 *fhpp = fhp; 1773 } else { 1774 kmem_free(fhp, fhsize); 1775 } 1776 out: 1777 return error; 1778 } 1779 1780 void 1781 vfs_composefh_free(fhandle_t *fhp) 1782 { 1783 1784 vfs__fhfree(fhp); 1785 } 1786 1787 /* 1788 * vfs_fhtovp: lookup a vnode by a filehandle. 1789 */ 1790 1791 int 1792 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1793 { 1794 struct mount *mp; 1795 int error; 1796 1797 *vpp = NULL; 1798 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1799 if (mp == NULL) { 1800 error = ESTALE; 1801 goto out; 1802 } 1803 if (mp->mnt_op->vfs_fhtovp == NULL) { 1804 error = EOPNOTSUPP; 1805 goto out; 1806 } 1807 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1808 out: 1809 return error; 1810 } 1811 1812 /* 1813 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1814 * the needed size. 1815 */ 1816 1817 int 1818 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1819 { 1820 fhandle_t *fhp; 1821 int error; 1822 1823 if (fhsize > FHANDLE_SIZE_MAX) { 1824 return EINVAL; 1825 } 1826 if (fhsize < FHANDLE_SIZE_MIN) { 1827 return EINVAL; 1828 } 1829 again: 1830 fhp = kmem_alloc(fhsize, KM_SLEEP); 1831 error = copyin(ufhp, fhp, fhsize); 1832 if (error == 0) { 1833 /* XXX this check shouldn't be here */ 1834 if (FHANDLE_SIZE(fhp) == fhsize) { 1835 *fhpp = fhp; 1836 return 0; 1837 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1838 /* 1839 * a kludge for nfsv2 padded handles. 1840 */ 1841 size_t sz; 1842 1843 sz = FHANDLE_SIZE(fhp); 1844 kmem_free(fhp, fhsize); 1845 fhsize = sz; 1846 goto again; 1847 } else { 1848 /* 1849 * userland told us wrong size. 1850 */ 1851 error = EINVAL; 1852 } 1853 } 1854 kmem_free(fhp, fhsize); 1855 return error; 1856 } 1857 1858 void 1859 vfs_copyinfh_free(fhandle_t *fhp) 1860 { 1861 1862 vfs__fhfree(fhp); 1863 } 1864 1865 /* 1866 * Get file handle system call 1867 */ 1868 int 1869 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1870 { 1871 /* { 1872 syscallarg(char *) fname; 1873 syscallarg(fhandle_t *) fhp; 1874 syscallarg(size_t *) fh_size; 1875 } */ 1876 struct vnode *vp; 1877 fhandle_t *fh; 1878 int error; 1879 struct pathbuf *pb; 1880 struct nameidata nd; 1881 size_t sz; 1882 size_t usz; 1883 1884 /* 1885 * Must be super user 1886 */ 1887 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1888 0, NULL, NULL, NULL); 1889 if (error) 1890 return (error); 1891 1892 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1893 if (error) { 1894 return error; 1895 } 1896 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1897 error = namei(&nd); 1898 if (error) { 1899 pathbuf_destroy(pb); 1900 return error; 1901 } 1902 vp = nd.ni_vp; 1903 pathbuf_destroy(pb); 1904 1905 error = vfs_composefh_alloc(vp, &fh); 1906 vput(vp); 1907 if (error != 0) { 1908 return error; 1909 } 1910 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1911 if (error != 0) { 1912 goto out; 1913 } 1914 sz = FHANDLE_SIZE(fh); 1915 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1916 if (error != 0) { 1917 goto out; 1918 } 1919 if (usz >= sz) { 1920 error = copyout(fh, SCARG(uap, fhp), sz); 1921 } else { 1922 error = E2BIG; 1923 } 1924 out: 1925 vfs_composefh_free(fh); 1926 return (error); 1927 } 1928 1929 /* 1930 * Open a file given a file handle. 1931 * 1932 * Check permissions, allocate an open file structure, 1933 * and call the device open routine if any. 1934 */ 1935 1936 int 1937 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1938 register_t *retval) 1939 { 1940 file_t *fp; 1941 struct vnode *vp = NULL; 1942 kauth_cred_t cred = l->l_cred; 1943 file_t *nfp; 1944 int indx, error; 1945 struct vattr va; 1946 fhandle_t *fh; 1947 int flags; 1948 proc_t *p; 1949 1950 p = curproc; 1951 1952 /* 1953 * Must be super user 1954 */ 1955 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1956 0, NULL, NULL, NULL))) 1957 return (error); 1958 1959 if (oflags & O_SEARCH) { 1960 oflags &= ~(int)O_SEARCH; 1961 } 1962 1963 flags = FFLAGS(oflags); 1964 if ((flags & (FREAD | FWRITE)) == 0) 1965 return (EINVAL); 1966 if ((flags & O_CREAT)) 1967 return (EINVAL); 1968 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1969 return (error); 1970 fp = nfp; 1971 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1972 if (error != 0) { 1973 goto bad; 1974 } 1975 error = vfs_fhtovp(fh, &vp); 1976 vfs_copyinfh_free(fh); 1977 if (error != 0) { 1978 goto bad; 1979 } 1980 1981 /* Now do an effective vn_open */ 1982 1983 if (vp->v_type == VSOCK) { 1984 error = EOPNOTSUPP; 1985 goto bad; 1986 } 1987 error = vn_openchk(vp, cred, flags); 1988 if (error != 0) 1989 goto bad; 1990 if (flags & O_TRUNC) { 1991 VOP_UNLOCK(vp); /* XXX */ 1992 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1993 vattr_null(&va); 1994 va.va_size = 0; 1995 error = VOP_SETATTR(vp, &va, cred); 1996 if (error) 1997 goto bad; 1998 } 1999 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2000 goto bad; 2001 if (flags & FWRITE) { 2002 mutex_enter(vp->v_interlock); 2003 vp->v_writecount++; 2004 mutex_exit(vp->v_interlock); 2005 } 2006 2007 /* done with modified vn_open, now finish what sys_open does. */ 2008 if ((error = open_setfp(l, fp, vp, indx, flags))) 2009 return error; 2010 2011 VOP_UNLOCK(vp); 2012 *retval = indx; 2013 fd_affix(p, fp, indx); 2014 return (0); 2015 2016 bad: 2017 fd_abort(p, fp, indx); 2018 if (vp != NULL) 2019 vput(vp); 2020 return (error); 2021 } 2022 2023 int 2024 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2025 { 2026 /* { 2027 syscallarg(const void *) fhp; 2028 syscallarg(size_t) fh_size; 2029 syscallarg(int) flags; 2030 } */ 2031 2032 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2033 SCARG(uap, flags), retval); 2034 } 2035 2036 int 2037 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2038 { 2039 int error; 2040 fhandle_t *fh; 2041 struct vnode *vp; 2042 2043 /* 2044 * Must be super user 2045 */ 2046 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2047 0, NULL, NULL, NULL))) 2048 return (error); 2049 2050 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2051 if (error != 0) 2052 return error; 2053 2054 error = vfs_fhtovp(fh, &vp); 2055 vfs_copyinfh_free(fh); 2056 if (error != 0) 2057 return error; 2058 2059 error = vn_stat(vp, sb); 2060 vput(vp); 2061 return error; 2062 } 2063 2064 2065 /* ARGSUSED */ 2066 int 2067 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2068 { 2069 /* { 2070 syscallarg(const void *) fhp; 2071 syscallarg(size_t) fh_size; 2072 syscallarg(struct stat *) sb; 2073 } */ 2074 struct stat sb; 2075 int error; 2076 2077 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2078 if (error) 2079 return error; 2080 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2081 } 2082 2083 int 2084 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2085 int flags) 2086 { 2087 fhandle_t *fh; 2088 struct mount *mp; 2089 struct vnode *vp; 2090 int error; 2091 2092 /* 2093 * Must be super user 2094 */ 2095 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2096 0, NULL, NULL, NULL))) 2097 return error; 2098 2099 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2100 if (error != 0) 2101 return error; 2102 2103 error = vfs_fhtovp(fh, &vp); 2104 vfs_copyinfh_free(fh); 2105 if (error != 0) 2106 return error; 2107 2108 mp = vp->v_mount; 2109 error = dostatvfs(mp, sb, l, flags, 1); 2110 vput(vp); 2111 return error; 2112 } 2113 2114 /* ARGSUSED */ 2115 int 2116 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2117 { 2118 /* { 2119 syscallarg(const void *) fhp; 2120 syscallarg(size_t) fh_size; 2121 syscallarg(struct statvfs *) buf; 2122 syscallarg(int) flags; 2123 } */ 2124 struct statvfs *sb = STATVFSBUF_GET(); 2125 int error; 2126 2127 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2128 SCARG(uap, flags)); 2129 if (error == 0) 2130 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2131 STATVFSBUF_PUT(sb); 2132 return error; 2133 } 2134 2135 /* 2136 * Create a special file. 2137 */ 2138 /* ARGSUSED */ 2139 int 2140 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2141 register_t *retval) 2142 { 2143 /* { 2144 syscallarg(const char *) path; 2145 syscallarg(mode_t) mode; 2146 syscallarg(dev_t) dev; 2147 } */ 2148 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2149 SCARG(uap, dev), retval, UIO_USERSPACE); 2150 } 2151 2152 int 2153 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2154 register_t *retval) 2155 { 2156 /* { 2157 syscallarg(int) fd; 2158 syscallarg(const char *) path; 2159 syscallarg(mode_t) mode; 2160 syscallarg(int) pad; 2161 syscallarg(dev_t) dev; 2162 } */ 2163 2164 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2165 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2166 } 2167 2168 int 2169 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2170 register_t *retval, enum uio_seg seg) 2171 { 2172 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2173 } 2174 2175 int 2176 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2177 dev_t dev, register_t *retval, enum uio_seg seg) 2178 { 2179 struct proc *p = l->l_proc; 2180 struct vnode *vp; 2181 struct vattr vattr; 2182 int error, optype; 2183 struct pathbuf *pb; 2184 struct nameidata nd; 2185 const char *pathstring; 2186 2187 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2188 0, NULL, NULL, NULL)) != 0) 2189 return (error); 2190 2191 optype = VOP_MKNOD_DESCOFFSET; 2192 2193 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2194 if (error) { 2195 return error; 2196 } 2197 pathstring = pathbuf_stringcopy_get(pb); 2198 if (pathstring == NULL) { 2199 pathbuf_destroy(pb); 2200 return ENOMEM; 2201 } 2202 2203 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2204 2205 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2206 goto out; 2207 vp = nd.ni_vp; 2208 2209 if (vp != NULL) 2210 error = EEXIST; 2211 else { 2212 vattr_null(&vattr); 2213 /* We will read cwdi->cwdi_cmask unlocked. */ 2214 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2215 vattr.va_rdev = dev; 2216 2217 switch (mode & S_IFMT) { 2218 case S_IFMT: /* used by badsect to flag bad sectors */ 2219 vattr.va_type = VBAD; 2220 break; 2221 case S_IFCHR: 2222 vattr.va_type = VCHR; 2223 break; 2224 case S_IFBLK: 2225 vattr.va_type = VBLK; 2226 break; 2227 case S_IFWHT: 2228 optype = VOP_WHITEOUT_DESCOFFSET; 2229 break; 2230 case S_IFREG: 2231 #if NVERIEXEC > 0 2232 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2233 O_CREAT); 2234 #endif /* NVERIEXEC > 0 */ 2235 vattr.va_type = VREG; 2236 vattr.va_rdev = VNOVAL; 2237 optype = VOP_CREATE_DESCOFFSET; 2238 break; 2239 default: 2240 error = EINVAL; 2241 break; 2242 } 2243 } 2244 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2245 && vattr.va_rdev == VNOVAL) 2246 error = EINVAL; 2247 if (!error) { 2248 switch (optype) { 2249 case VOP_WHITEOUT_DESCOFFSET: 2250 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2251 if (error) 2252 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2253 vput(nd.ni_dvp); 2254 break; 2255 2256 case VOP_MKNOD_DESCOFFSET: 2257 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2258 &nd.ni_cnd, &vattr); 2259 if (error == 0) 2260 vrele(nd.ni_vp); 2261 vput(nd.ni_dvp); 2262 break; 2263 2264 case VOP_CREATE_DESCOFFSET: 2265 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2266 &nd.ni_cnd, &vattr); 2267 if (error == 0) 2268 vrele(nd.ni_vp); 2269 vput(nd.ni_dvp); 2270 break; 2271 } 2272 } else { 2273 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2274 if (nd.ni_dvp == vp) 2275 vrele(nd.ni_dvp); 2276 else 2277 vput(nd.ni_dvp); 2278 if (vp) 2279 vrele(vp); 2280 } 2281 out: 2282 pathbuf_stringcopy_put(pb, pathstring); 2283 pathbuf_destroy(pb); 2284 return (error); 2285 } 2286 2287 /* 2288 * Create a named pipe. 2289 */ 2290 /* ARGSUSED */ 2291 int 2292 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2293 { 2294 /* { 2295 syscallarg(const char *) path; 2296 syscallarg(int) mode; 2297 } */ 2298 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2299 } 2300 2301 int 2302 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2303 register_t *retval) 2304 { 2305 /* { 2306 syscallarg(int) fd; 2307 syscallarg(const char *) path; 2308 syscallarg(int) mode; 2309 } */ 2310 2311 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2312 SCARG(uap, mode)); 2313 } 2314 2315 static int 2316 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2317 { 2318 struct proc *p = l->l_proc; 2319 struct vattr vattr; 2320 int error; 2321 struct pathbuf *pb; 2322 struct nameidata nd; 2323 2324 error = pathbuf_copyin(path, &pb); 2325 if (error) { 2326 return error; 2327 } 2328 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2329 2330 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2331 pathbuf_destroy(pb); 2332 return error; 2333 } 2334 if (nd.ni_vp != NULL) { 2335 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2336 if (nd.ni_dvp == nd.ni_vp) 2337 vrele(nd.ni_dvp); 2338 else 2339 vput(nd.ni_dvp); 2340 vrele(nd.ni_vp); 2341 pathbuf_destroy(pb); 2342 return (EEXIST); 2343 } 2344 vattr_null(&vattr); 2345 vattr.va_type = VFIFO; 2346 /* We will read cwdi->cwdi_cmask unlocked. */ 2347 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2348 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2349 if (error == 0) 2350 vrele(nd.ni_vp); 2351 vput(nd.ni_dvp); 2352 pathbuf_destroy(pb); 2353 return (error); 2354 } 2355 2356 /* 2357 * Make a hard file link. 2358 */ 2359 /* ARGSUSED */ 2360 int 2361 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2362 const char *link, int follow, register_t *retval) 2363 { 2364 struct vnode *vp; 2365 struct pathbuf *linkpb; 2366 struct nameidata nd; 2367 namei_simple_flags_t ns_flags; 2368 int error; 2369 2370 if (follow & AT_SYMLINK_FOLLOW) 2371 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2372 else 2373 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2374 2375 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2376 if (error != 0) 2377 return (error); 2378 error = pathbuf_copyin(link, &linkpb); 2379 if (error) { 2380 goto out1; 2381 } 2382 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2383 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2384 goto out2; 2385 if (nd.ni_vp) { 2386 error = EEXIST; 2387 goto abortop; 2388 } 2389 /* Prevent hard links on directories. */ 2390 if (vp->v_type == VDIR) { 2391 error = EPERM; 2392 goto abortop; 2393 } 2394 /* Prevent cross-mount operation. */ 2395 if (nd.ni_dvp->v_mount != vp->v_mount) { 2396 error = EXDEV; 2397 goto abortop; 2398 } 2399 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2400 VOP_UNLOCK(nd.ni_dvp); 2401 vrele(nd.ni_dvp); 2402 out2: 2403 pathbuf_destroy(linkpb); 2404 out1: 2405 vrele(vp); 2406 return (error); 2407 abortop: 2408 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2409 if (nd.ni_dvp == nd.ni_vp) 2410 vrele(nd.ni_dvp); 2411 else 2412 vput(nd.ni_dvp); 2413 if (nd.ni_vp != NULL) 2414 vrele(nd.ni_vp); 2415 goto out2; 2416 } 2417 2418 int 2419 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2420 { 2421 /* { 2422 syscallarg(const char *) path; 2423 syscallarg(const char *) link; 2424 } */ 2425 const char *path = SCARG(uap, path); 2426 const char *link = SCARG(uap, link); 2427 2428 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2429 AT_SYMLINK_FOLLOW, retval); 2430 } 2431 2432 int 2433 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2434 register_t *retval) 2435 { 2436 /* { 2437 syscallarg(int) fd1; 2438 syscallarg(const char *) name1; 2439 syscallarg(int) fd2; 2440 syscallarg(const char *) name2; 2441 syscallarg(int) flags; 2442 } */ 2443 int fd1 = SCARG(uap, fd1); 2444 const char *name1 = SCARG(uap, name1); 2445 int fd2 = SCARG(uap, fd2); 2446 const char *name2 = SCARG(uap, name2); 2447 int follow; 2448 2449 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2450 2451 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2452 } 2453 2454 2455 int 2456 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2457 { 2458 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2459 } 2460 2461 static int 2462 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2463 const char *link, enum uio_seg seg) 2464 { 2465 struct proc *p = curproc; 2466 struct vattr vattr; 2467 char *path; 2468 int error; 2469 size_t len; 2470 struct pathbuf *linkpb; 2471 struct nameidata nd; 2472 2473 KASSERT(l != NULL || fdat == AT_FDCWD); 2474 2475 path = PNBUF_GET(); 2476 if (seg == UIO_USERSPACE) { 2477 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2478 goto out1; 2479 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2480 goto out1; 2481 } else { 2482 len = strlen(patharg) + 1; 2483 KASSERT(len <= MAXPATHLEN); 2484 memcpy(path, patharg, len); 2485 linkpb = pathbuf_create(link); 2486 if (linkpb == NULL) { 2487 error = ENOMEM; 2488 goto out1; 2489 } 2490 } 2491 ktrkuser("symlink-target", path, len - 1); 2492 2493 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2494 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2495 goto out2; 2496 if (nd.ni_vp) { 2497 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2498 if (nd.ni_dvp == nd.ni_vp) 2499 vrele(nd.ni_dvp); 2500 else 2501 vput(nd.ni_dvp); 2502 vrele(nd.ni_vp); 2503 error = EEXIST; 2504 goto out2; 2505 } 2506 vattr_null(&vattr); 2507 vattr.va_type = VLNK; 2508 /* We will read cwdi->cwdi_cmask unlocked. */ 2509 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2510 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2511 if (error == 0) 2512 vrele(nd.ni_vp); 2513 vput(nd.ni_dvp); 2514 out2: 2515 pathbuf_destroy(linkpb); 2516 out1: 2517 PNBUF_PUT(path); 2518 return (error); 2519 } 2520 2521 /* 2522 * Make a symbolic link. 2523 */ 2524 /* ARGSUSED */ 2525 int 2526 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2527 { 2528 /* { 2529 syscallarg(const char *) path; 2530 syscallarg(const char *) link; 2531 } */ 2532 2533 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2534 UIO_USERSPACE); 2535 } 2536 2537 int 2538 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2539 register_t *retval) 2540 { 2541 /* { 2542 syscallarg(const char *) path1; 2543 syscallarg(int) fd; 2544 syscallarg(const char *) path2; 2545 } */ 2546 2547 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2548 SCARG(uap, path2), UIO_USERSPACE); 2549 } 2550 2551 /* 2552 * Delete a whiteout from the filesystem. 2553 */ 2554 /* ARGSUSED */ 2555 int 2556 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2557 { 2558 /* { 2559 syscallarg(const char *) path; 2560 } */ 2561 int error; 2562 struct pathbuf *pb; 2563 struct nameidata nd; 2564 2565 error = pathbuf_copyin(SCARG(uap, path), &pb); 2566 if (error) { 2567 return error; 2568 } 2569 2570 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2571 error = namei(&nd); 2572 if (error) { 2573 pathbuf_destroy(pb); 2574 return (error); 2575 } 2576 2577 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2578 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2579 if (nd.ni_dvp == nd.ni_vp) 2580 vrele(nd.ni_dvp); 2581 else 2582 vput(nd.ni_dvp); 2583 if (nd.ni_vp) 2584 vrele(nd.ni_vp); 2585 pathbuf_destroy(pb); 2586 return (EEXIST); 2587 } 2588 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2589 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2590 vput(nd.ni_dvp); 2591 pathbuf_destroy(pb); 2592 return (error); 2593 } 2594 2595 /* 2596 * Delete a name from the filesystem. 2597 */ 2598 /* ARGSUSED */ 2599 int 2600 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2601 { 2602 /* { 2603 syscallarg(const char *) path; 2604 } */ 2605 2606 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2607 } 2608 2609 int 2610 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2611 register_t *retval) 2612 { 2613 /* { 2614 syscallarg(int) fd; 2615 syscallarg(const char *) path; 2616 syscallarg(int) flag; 2617 } */ 2618 2619 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2620 SCARG(uap, flag), UIO_USERSPACE); 2621 } 2622 2623 int 2624 do_sys_unlink(const char *arg, enum uio_seg seg) 2625 { 2626 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2627 } 2628 2629 static int 2630 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2631 enum uio_seg seg) 2632 { 2633 struct vnode *vp; 2634 int error; 2635 struct pathbuf *pb; 2636 struct nameidata nd; 2637 const char *pathstring; 2638 2639 KASSERT(l != NULL || fdat == AT_FDCWD); 2640 2641 error = pathbuf_maybe_copyin(arg, seg, &pb); 2642 if (error) { 2643 return error; 2644 } 2645 pathstring = pathbuf_stringcopy_get(pb); 2646 if (pathstring == NULL) { 2647 pathbuf_destroy(pb); 2648 return ENOMEM; 2649 } 2650 2651 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2652 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2653 goto out; 2654 vp = nd.ni_vp; 2655 2656 /* 2657 * The root of a mounted filesystem cannot be deleted. 2658 */ 2659 if ((vp->v_vflag & VV_ROOT) != 0) { 2660 error = EBUSY; 2661 goto abort; 2662 } 2663 2664 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2665 error = EBUSY; 2666 goto abort; 2667 } 2668 2669 /* 2670 * No rmdir "." please. 2671 */ 2672 if (nd.ni_dvp == vp) { 2673 error = EINVAL; 2674 goto abort; 2675 } 2676 2677 /* 2678 * AT_REMOVEDIR is required to remove a directory 2679 */ 2680 if (vp->v_type == VDIR) { 2681 if (!(flags & AT_REMOVEDIR)) { 2682 error = EPERM; 2683 goto abort; 2684 } else { 2685 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2686 vput(nd.ni_dvp); 2687 goto out; 2688 } 2689 } 2690 2691 /* 2692 * Starting here we only deal with non directories. 2693 */ 2694 if (flags & AT_REMOVEDIR) { 2695 error = ENOTDIR; 2696 goto abort; 2697 } 2698 2699 #if NVERIEXEC > 0 2700 /* Handle remove requests for veriexec entries. */ 2701 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2702 goto abort; 2703 } 2704 #endif /* NVERIEXEC > 0 */ 2705 2706 #ifdef FILEASSOC 2707 (void)fileassoc_file_delete(vp); 2708 #endif /* FILEASSOC */ 2709 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2710 vput(nd.ni_dvp); 2711 goto out; 2712 2713 abort: 2714 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2715 if (nd.ni_dvp == vp) 2716 vrele(nd.ni_dvp); 2717 else 2718 vput(nd.ni_dvp); 2719 vput(vp); 2720 2721 out: 2722 pathbuf_stringcopy_put(pb, pathstring); 2723 pathbuf_destroy(pb); 2724 return (error); 2725 } 2726 2727 /* 2728 * Reposition read/write file offset. 2729 */ 2730 int 2731 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2732 { 2733 /* { 2734 syscallarg(int) fd; 2735 syscallarg(int) pad; 2736 syscallarg(off_t) offset; 2737 syscallarg(int) whence; 2738 } */ 2739 kauth_cred_t cred = l->l_cred; 2740 file_t *fp; 2741 struct vnode *vp; 2742 struct vattr vattr; 2743 off_t newoff; 2744 int error, fd; 2745 2746 fd = SCARG(uap, fd); 2747 2748 if ((fp = fd_getfile(fd)) == NULL) 2749 return (EBADF); 2750 2751 vp = fp->f_vnode; 2752 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2753 error = ESPIPE; 2754 goto out; 2755 } 2756 2757 vn_lock(vp, LK_SHARED | LK_RETRY); 2758 2759 switch (SCARG(uap, whence)) { 2760 case SEEK_CUR: 2761 newoff = fp->f_offset + SCARG(uap, offset); 2762 break; 2763 case SEEK_END: 2764 error = VOP_GETATTR(vp, &vattr, cred); 2765 if (error) { 2766 VOP_UNLOCK(vp); 2767 goto out; 2768 } 2769 newoff = SCARG(uap, offset) + vattr.va_size; 2770 break; 2771 case SEEK_SET: 2772 newoff = SCARG(uap, offset); 2773 break; 2774 default: 2775 error = EINVAL; 2776 VOP_UNLOCK(vp); 2777 goto out; 2778 } 2779 VOP_UNLOCK(vp); 2780 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2781 *(off_t *)retval = fp->f_offset = newoff; 2782 } 2783 out: 2784 fd_putfile(fd); 2785 return (error); 2786 } 2787 2788 /* 2789 * Positional read system call. 2790 */ 2791 int 2792 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2793 { 2794 /* { 2795 syscallarg(int) fd; 2796 syscallarg(void *) buf; 2797 syscallarg(size_t) nbyte; 2798 syscallarg(off_t) offset; 2799 } */ 2800 file_t *fp; 2801 struct vnode *vp; 2802 off_t offset; 2803 int error, fd = SCARG(uap, fd); 2804 2805 if ((fp = fd_getfile(fd)) == NULL) 2806 return (EBADF); 2807 2808 if ((fp->f_flag & FREAD) == 0) { 2809 fd_putfile(fd); 2810 return (EBADF); 2811 } 2812 2813 vp = fp->f_vnode; 2814 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2815 error = ESPIPE; 2816 goto out; 2817 } 2818 2819 offset = SCARG(uap, offset); 2820 2821 /* 2822 * XXX This works because no file systems actually 2823 * XXX take any action on the seek operation. 2824 */ 2825 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2826 goto out; 2827 2828 /* dofileread() will unuse the descriptor for us */ 2829 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2830 &offset, 0, retval)); 2831 2832 out: 2833 fd_putfile(fd); 2834 return (error); 2835 } 2836 2837 /* 2838 * Positional scatter read system call. 2839 */ 2840 int 2841 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2842 { 2843 /* { 2844 syscallarg(int) fd; 2845 syscallarg(const struct iovec *) iovp; 2846 syscallarg(int) iovcnt; 2847 syscallarg(off_t) offset; 2848 } */ 2849 off_t offset = SCARG(uap, offset); 2850 2851 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2852 SCARG(uap, iovcnt), &offset, 0, retval); 2853 } 2854 2855 /* 2856 * Positional write system call. 2857 */ 2858 int 2859 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2860 { 2861 /* { 2862 syscallarg(int) fd; 2863 syscallarg(const void *) buf; 2864 syscallarg(size_t) nbyte; 2865 syscallarg(off_t) offset; 2866 } */ 2867 file_t *fp; 2868 struct vnode *vp; 2869 off_t offset; 2870 int error, fd = SCARG(uap, fd); 2871 2872 if ((fp = fd_getfile(fd)) == NULL) 2873 return (EBADF); 2874 2875 if ((fp->f_flag & FWRITE) == 0) { 2876 fd_putfile(fd); 2877 return (EBADF); 2878 } 2879 2880 vp = fp->f_vnode; 2881 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2882 error = ESPIPE; 2883 goto out; 2884 } 2885 2886 offset = SCARG(uap, offset); 2887 2888 /* 2889 * XXX This works because no file systems actually 2890 * XXX take any action on the seek operation. 2891 */ 2892 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2893 goto out; 2894 2895 /* dofilewrite() will unuse the descriptor for us */ 2896 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2897 &offset, 0, retval)); 2898 2899 out: 2900 fd_putfile(fd); 2901 return (error); 2902 } 2903 2904 /* 2905 * Positional gather write system call. 2906 */ 2907 int 2908 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2909 { 2910 /* { 2911 syscallarg(int) fd; 2912 syscallarg(const struct iovec *) iovp; 2913 syscallarg(int) iovcnt; 2914 syscallarg(off_t) offset; 2915 } */ 2916 off_t offset = SCARG(uap, offset); 2917 2918 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2919 SCARG(uap, iovcnt), &offset, 0, retval); 2920 } 2921 2922 /* 2923 * Check access permissions. 2924 */ 2925 int 2926 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2927 { 2928 /* { 2929 syscallarg(const char *) path; 2930 syscallarg(int) flags; 2931 } */ 2932 2933 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2934 SCARG(uap, flags), 0); 2935 } 2936 2937 int 2938 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2939 int mode, int flags) 2940 { 2941 kauth_cred_t cred; 2942 struct vnode *vp; 2943 int error, nd_flag, vmode; 2944 struct pathbuf *pb; 2945 struct nameidata nd; 2946 2947 CTASSERT(F_OK == 0); 2948 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2949 /* nonsense mode */ 2950 return EINVAL; 2951 } 2952 2953 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2954 if (flags & AT_SYMLINK_NOFOLLOW) 2955 nd_flag &= ~FOLLOW; 2956 2957 error = pathbuf_copyin(path, &pb); 2958 if (error) 2959 return error; 2960 2961 NDINIT(&nd, LOOKUP, nd_flag, pb); 2962 2963 /* Override default credentials */ 2964 cred = kauth_cred_dup(l->l_cred); 2965 if (!(flags & AT_EACCESS)) { 2966 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2967 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2968 } 2969 nd.ni_cnd.cn_cred = cred; 2970 2971 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2972 pathbuf_destroy(pb); 2973 goto out; 2974 } 2975 vp = nd.ni_vp; 2976 pathbuf_destroy(pb); 2977 2978 /* Flags == 0 means only check for existence. */ 2979 if (mode) { 2980 vmode = 0; 2981 if (mode & R_OK) 2982 vmode |= VREAD; 2983 if (mode & W_OK) 2984 vmode |= VWRITE; 2985 if (mode & X_OK) 2986 vmode |= VEXEC; 2987 2988 error = VOP_ACCESS(vp, vmode, cred); 2989 if (!error && (vmode & VWRITE)) 2990 error = vn_writechk(vp); 2991 } 2992 vput(vp); 2993 out: 2994 kauth_cred_free(cred); 2995 return (error); 2996 } 2997 2998 int 2999 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3000 register_t *retval) 3001 { 3002 /* { 3003 syscallarg(int) fd; 3004 syscallarg(const char *) path; 3005 syscallarg(int) amode; 3006 syscallarg(int) flag; 3007 } */ 3008 3009 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3010 SCARG(uap, amode), SCARG(uap, flag)); 3011 } 3012 3013 /* 3014 * Common code for all sys_stat functions, including compat versions. 3015 */ 3016 int 3017 do_sys_stat(const char *userpath, unsigned int nd_flag, 3018 struct stat *sb) 3019 { 3020 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3021 } 3022 3023 int 3024 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3025 unsigned int nd_flag, struct stat *sb) 3026 { 3027 int error; 3028 struct pathbuf *pb; 3029 struct nameidata nd; 3030 3031 KASSERT(l != NULL || fdat == AT_FDCWD); 3032 3033 error = pathbuf_copyin(userpath, &pb); 3034 if (error) { 3035 return error; 3036 } 3037 3038 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3039 3040 error = fd_nameiat(l, fdat, &nd); 3041 if (error != 0) { 3042 pathbuf_destroy(pb); 3043 return error; 3044 } 3045 error = vn_stat(nd.ni_vp, sb); 3046 vput(nd.ni_vp); 3047 pathbuf_destroy(pb); 3048 return error; 3049 } 3050 3051 /* 3052 * Get file status; this version follows links. 3053 */ 3054 /* ARGSUSED */ 3055 int 3056 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3057 { 3058 /* { 3059 syscallarg(const char *) path; 3060 syscallarg(struct stat *) ub; 3061 } */ 3062 struct stat sb; 3063 int error; 3064 3065 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3066 if (error) 3067 return error; 3068 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3069 } 3070 3071 /* 3072 * Get file status; this version does not follow links. 3073 */ 3074 /* ARGSUSED */ 3075 int 3076 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3077 { 3078 /* { 3079 syscallarg(const char *) path; 3080 syscallarg(struct stat *) ub; 3081 } */ 3082 struct stat sb; 3083 int error; 3084 3085 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3086 if (error) 3087 return error; 3088 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3089 } 3090 3091 int 3092 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3093 register_t *retval) 3094 { 3095 /* { 3096 syscallarg(int) fd; 3097 syscallarg(const char *) path; 3098 syscallarg(struct stat *) buf; 3099 syscallarg(int) flag; 3100 } */ 3101 unsigned int nd_flag; 3102 struct stat sb; 3103 int error; 3104 3105 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3106 nd_flag = NOFOLLOW; 3107 else 3108 nd_flag = FOLLOW; 3109 3110 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3111 &sb); 3112 if (error) 3113 return error; 3114 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3115 } 3116 3117 /* 3118 * Get configurable pathname variables. 3119 */ 3120 /* ARGSUSED */ 3121 int 3122 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3123 { 3124 /* { 3125 syscallarg(const char *) path; 3126 syscallarg(int) name; 3127 } */ 3128 int error; 3129 struct pathbuf *pb; 3130 struct nameidata nd; 3131 3132 error = pathbuf_copyin(SCARG(uap, path), &pb); 3133 if (error) { 3134 return error; 3135 } 3136 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3137 if ((error = namei(&nd)) != 0) { 3138 pathbuf_destroy(pb); 3139 return (error); 3140 } 3141 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3142 vput(nd.ni_vp); 3143 pathbuf_destroy(pb); 3144 return (error); 3145 } 3146 3147 /* 3148 * Return target name of a symbolic link. 3149 */ 3150 /* ARGSUSED */ 3151 int 3152 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3153 register_t *retval) 3154 { 3155 /* { 3156 syscallarg(const char *) path; 3157 syscallarg(char *) buf; 3158 syscallarg(size_t) count; 3159 } */ 3160 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3161 SCARG(uap, buf), SCARG(uap, count), retval); 3162 } 3163 3164 static int 3165 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3166 size_t count, register_t *retval) 3167 { 3168 struct vnode *vp; 3169 struct iovec aiov; 3170 struct uio auio; 3171 int error; 3172 struct pathbuf *pb; 3173 struct nameidata nd; 3174 3175 error = pathbuf_copyin(path, &pb); 3176 if (error) { 3177 return error; 3178 } 3179 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3180 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3181 pathbuf_destroy(pb); 3182 return error; 3183 } 3184 vp = nd.ni_vp; 3185 pathbuf_destroy(pb); 3186 if (vp->v_type != VLNK) 3187 error = EINVAL; 3188 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3189 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3190 aiov.iov_base = buf; 3191 aiov.iov_len = count; 3192 auio.uio_iov = &aiov; 3193 auio.uio_iovcnt = 1; 3194 auio.uio_offset = 0; 3195 auio.uio_rw = UIO_READ; 3196 KASSERT(l == curlwp); 3197 auio.uio_vmspace = l->l_proc->p_vmspace; 3198 auio.uio_resid = count; 3199 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3200 *retval = count - auio.uio_resid; 3201 } 3202 vput(vp); 3203 return (error); 3204 } 3205 3206 int 3207 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3208 register_t *retval) 3209 { 3210 /* { 3211 syscallarg(int) fd; 3212 syscallarg(const char *) path; 3213 syscallarg(char *) buf; 3214 syscallarg(size_t) bufsize; 3215 } */ 3216 3217 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3218 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3219 } 3220 3221 /* 3222 * Change flags of a file given a path name. 3223 */ 3224 /* ARGSUSED */ 3225 int 3226 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3227 { 3228 /* { 3229 syscallarg(const char *) path; 3230 syscallarg(u_long) flags; 3231 } */ 3232 struct vnode *vp; 3233 int error; 3234 3235 error = namei_simple_user(SCARG(uap, path), 3236 NSM_FOLLOW_TRYEMULROOT, &vp); 3237 if (error != 0) 3238 return (error); 3239 error = change_flags(vp, SCARG(uap, flags), l); 3240 vput(vp); 3241 return (error); 3242 } 3243 3244 /* 3245 * Change flags of a file given a file descriptor. 3246 */ 3247 /* ARGSUSED */ 3248 int 3249 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3250 { 3251 /* { 3252 syscallarg(int) fd; 3253 syscallarg(u_long) flags; 3254 } */ 3255 struct vnode *vp; 3256 file_t *fp; 3257 int error; 3258 3259 /* fd_getvnode() will use the descriptor for us */ 3260 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3261 return (error); 3262 vp = fp->f_vnode; 3263 error = change_flags(vp, SCARG(uap, flags), l); 3264 VOP_UNLOCK(vp); 3265 fd_putfile(SCARG(uap, fd)); 3266 return (error); 3267 } 3268 3269 /* 3270 * Change flags of a file given a path name; this version does 3271 * not follow links. 3272 */ 3273 int 3274 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3275 { 3276 /* { 3277 syscallarg(const char *) path; 3278 syscallarg(u_long) flags; 3279 } */ 3280 struct vnode *vp; 3281 int error; 3282 3283 error = namei_simple_user(SCARG(uap, path), 3284 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3285 if (error != 0) 3286 return (error); 3287 error = change_flags(vp, SCARG(uap, flags), l); 3288 vput(vp); 3289 return (error); 3290 } 3291 3292 /* 3293 * Common routine to change flags of a file. 3294 */ 3295 int 3296 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3297 { 3298 struct vattr vattr; 3299 int error; 3300 3301 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3302 3303 vattr_null(&vattr); 3304 vattr.va_flags = flags; 3305 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3306 3307 return (error); 3308 } 3309 3310 /* 3311 * Change mode of a file given path name; this version follows links. 3312 */ 3313 /* ARGSUSED */ 3314 int 3315 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3316 { 3317 /* { 3318 syscallarg(const char *) path; 3319 syscallarg(int) mode; 3320 } */ 3321 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3322 SCARG(uap, mode), 0); 3323 } 3324 3325 int 3326 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3327 { 3328 int error; 3329 struct vnode *vp; 3330 namei_simple_flags_t ns_flag; 3331 3332 if (flags & AT_SYMLINK_NOFOLLOW) 3333 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3334 else 3335 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3336 3337 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3338 if (error != 0) 3339 return error; 3340 3341 error = change_mode(vp, mode, l); 3342 3343 vrele(vp); 3344 3345 return (error); 3346 } 3347 3348 /* 3349 * Change mode of a file given a file descriptor. 3350 */ 3351 /* ARGSUSED */ 3352 int 3353 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3354 { 3355 /* { 3356 syscallarg(int) fd; 3357 syscallarg(int) mode; 3358 } */ 3359 file_t *fp; 3360 int error; 3361 3362 /* fd_getvnode() will use the descriptor for us */ 3363 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3364 return (error); 3365 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3366 fd_putfile(SCARG(uap, fd)); 3367 return (error); 3368 } 3369 3370 int 3371 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3372 register_t *retval) 3373 { 3374 /* { 3375 syscallarg(int) fd; 3376 syscallarg(const char *) path; 3377 syscallarg(int) mode; 3378 syscallarg(int) flag; 3379 } */ 3380 3381 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3382 SCARG(uap, mode), SCARG(uap, flag)); 3383 } 3384 3385 /* 3386 * Change mode of a file given path name; this version does not follow links. 3387 */ 3388 /* ARGSUSED */ 3389 int 3390 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3391 { 3392 /* { 3393 syscallarg(const char *) path; 3394 syscallarg(int) mode; 3395 } */ 3396 int error; 3397 struct vnode *vp; 3398 3399 error = namei_simple_user(SCARG(uap, path), 3400 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3401 if (error != 0) 3402 return (error); 3403 3404 error = change_mode(vp, SCARG(uap, mode), l); 3405 3406 vrele(vp); 3407 return (error); 3408 } 3409 3410 /* 3411 * Common routine to set mode given a vnode. 3412 */ 3413 static int 3414 change_mode(struct vnode *vp, int mode, struct lwp *l) 3415 { 3416 struct vattr vattr; 3417 int error; 3418 3419 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3420 vattr_null(&vattr); 3421 vattr.va_mode = mode & ALLPERMS; 3422 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3423 VOP_UNLOCK(vp); 3424 return (error); 3425 } 3426 3427 /* 3428 * Set ownership given a path name; this version follows links. 3429 */ 3430 /* ARGSUSED */ 3431 int 3432 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3433 { 3434 /* { 3435 syscallarg(const char *) path; 3436 syscallarg(uid_t) uid; 3437 syscallarg(gid_t) gid; 3438 } */ 3439 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3440 SCARG(uap, gid), 0); 3441 } 3442 3443 int 3444 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3445 gid_t gid, int flags) 3446 { 3447 int error; 3448 struct vnode *vp; 3449 namei_simple_flags_t ns_flag; 3450 3451 if (flags & AT_SYMLINK_NOFOLLOW) 3452 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3453 else 3454 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3455 3456 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3457 if (error != 0) 3458 return error; 3459 3460 error = change_owner(vp, uid, gid, l, 0); 3461 3462 vrele(vp); 3463 3464 return (error); 3465 } 3466 3467 /* 3468 * Set ownership given a path name; this version follows links. 3469 * Provides POSIX semantics. 3470 */ 3471 /* ARGSUSED */ 3472 int 3473 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3474 { 3475 /* { 3476 syscallarg(const char *) path; 3477 syscallarg(uid_t) uid; 3478 syscallarg(gid_t) gid; 3479 } */ 3480 int error; 3481 struct vnode *vp; 3482 3483 error = namei_simple_user(SCARG(uap, path), 3484 NSM_FOLLOW_TRYEMULROOT, &vp); 3485 if (error != 0) 3486 return (error); 3487 3488 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3489 3490 vrele(vp); 3491 return (error); 3492 } 3493 3494 /* 3495 * Set ownership given a file descriptor. 3496 */ 3497 /* ARGSUSED */ 3498 int 3499 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3500 { 3501 /* { 3502 syscallarg(int) fd; 3503 syscallarg(uid_t) uid; 3504 syscallarg(gid_t) gid; 3505 } */ 3506 int error; 3507 file_t *fp; 3508 3509 /* fd_getvnode() will use the descriptor for us */ 3510 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3511 return (error); 3512 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3513 l, 0); 3514 fd_putfile(SCARG(uap, fd)); 3515 return (error); 3516 } 3517 3518 int 3519 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3520 register_t *retval) 3521 { 3522 /* { 3523 syscallarg(int) fd; 3524 syscallarg(const char *) path; 3525 syscallarg(uid_t) owner; 3526 syscallarg(gid_t) group; 3527 syscallarg(int) flag; 3528 } */ 3529 3530 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3531 SCARG(uap, owner), SCARG(uap, group), 3532 SCARG(uap, flag)); 3533 } 3534 3535 /* 3536 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3537 */ 3538 /* ARGSUSED */ 3539 int 3540 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3541 { 3542 /* { 3543 syscallarg(int) fd; 3544 syscallarg(uid_t) uid; 3545 syscallarg(gid_t) gid; 3546 } */ 3547 int error; 3548 file_t *fp; 3549 3550 /* fd_getvnode() will use the descriptor for us */ 3551 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3552 return (error); 3553 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3554 l, 1); 3555 fd_putfile(SCARG(uap, fd)); 3556 return (error); 3557 } 3558 3559 /* 3560 * Set ownership given a path name; this version does not follow links. 3561 */ 3562 /* ARGSUSED */ 3563 int 3564 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3565 { 3566 /* { 3567 syscallarg(const char *) path; 3568 syscallarg(uid_t) uid; 3569 syscallarg(gid_t) gid; 3570 } */ 3571 int error; 3572 struct vnode *vp; 3573 3574 error = namei_simple_user(SCARG(uap, path), 3575 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3576 if (error != 0) 3577 return (error); 3578 3579 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3580 3581 vrele(vp); 3582 return (error); 3583 } 3584 3585 /* 3586 * Set ownership given a path name; this version does not follow links. 3587 * Provides POSIX/XPG semantics. 3588 */ 3589 /* ARGSUSED */ 3590 int 3591 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3592 { 3593 /* { 3594 syscallarg(const char *) path; 3595 syscallarg(uid_t) uid; 3596 syscallarg(gid_t) gid; 3597 } */ 3598 int error; 3599 struct vnode *vp; 3600 3601 error = namei_simple_user(SCARG(uap, path), 3602 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3603 if (error != 0) 3604 return (error); 3605 3606 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3607 3608 vrele(vp); 3609 return (error); 3610 } 3611 3612 /* 3613 * Common routine to set ownership given a vnode. 3614 */ 3615 static int 3616 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3617 int posix_semantics) 3618 { 3619 struct vattr vattr; 3620 mode_t newmode; 3621 int error; 3622 3623 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3624 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3625 goto out; 3626 3627 #define CHANGED(x) ((int)(x) != -1) 3628 newmode = vattr.va_mode; 3629 if (posix_semantics) { 3630 /* 3631 * POSIX/XPG semantics: if the caller is not the super-user, 3632 * clear set-user-id and set-group-id bits. Both POSIX and 3633 * the XPG consider the behaviour for calls by the super-user 3634 * implementation-defined; we leave the set-user-id and set- 3635 * group-id settings intact in that case. 3636 */ 3637 if (vattr.va_mode & S_ISUID) { 3638 if (kauth_authorize_vnode(l->l_cred, 3639 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3640 newmode &= ~S_ISUID; 3641 } 3642 if (vattr.va_mode & S_ISGID) { 3643 if (kauth_authorize_vnode(l->l_cred, 3644 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3645 newmode &= ~S_ISGID; 3646 } 3647 } else { 3648 /* 3649 * NetBSD semantics: when changing owner and/or group, 3650 * clear the respective bit(s). 3651 */ 3652 if (CHANGED(uid)) 3653 newmode &= ~S_ISUID; 3654 if (CHANGED(gid)) 3655 newmode &= ~S_ISGID; 3656 } 3657 /* Update va_mode iff altered. */ 3658 if (vattr.va_mode == newmode) 3659 newmode = VNOVAL; 3660 3661 vattr_null(&vattr); 3662 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3663 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3664 vattr.va_mode = newmode; 3665 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3666 #undef CHANGED 3667 3668 out: 3669 VOP_UNLOCK(vp); 3670 return (error); 3671 } 3672 3673 /* 3674 * Set the access and modification times given a path name; this 3675 * version follows links. 3676 */ 3677 /* ARGSUSED */ 3678 int 3679 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3680 register_t *retval) 3681 { 3682 /* { 3683 syscallarg(const char *) path; 3684 syscallarg(const struct timeval *) tptr; 3685 } */ 3686 3687 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3688 SCARG(uap, tptr), UIO_USERSPACE); 3689 } 3690 3691 /* 3692 * Set the access and modification times given a file descriptor. 3693 */ 3694 /* ARGSUSED */ 3695 int 3696 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3697 register_t *retval) 3698 { 3699 /* { 3700 syscallarg(int) fd; 3701 syscallarg(const struct timeval *) tptr; 3702 } */ 3703 int error; 3704 file_t *fp; 3705 3706 /* fd_getvnode() will use the descriptor for us */ 3707 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3708 return (error); 3709 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3710 UIO_USERSPACE); 3711 fd_putfile(SCARG(uap, fd)); 3712 return (error); 3713 } 3714 3715 int 3716 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3717 register_t *retval) 3718 { 3719 /* { 3720 syscallarg(int) fd; 3721 syscallarg(const struct timespec *) tptr; 3722 } */ 3723 int error; 3724 file_t *fp; 3725 3726 /* fd_getvnode() will use the descriptor for us */ 3727 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3728 return (error); 3729 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3730 SCARG(uap, tptr), UIO_USERSPACE); 3731 fd_putfile(SCARG(uap, fd)); 3732 return (error); 3733 } 3734 3735 /* 3736 * Set the access and modification times given a path name; this 3737 * version does not follow links. 3738 */ 3739 int 3740 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3741 register_t *retval) 3742 { 3743 /* { 3744 syscallarg(const char *) path; 3745 syscallarg(const struct timeval *) tptr; 3746 } */ 3747 3748 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3749 SCARG(uap, tptr), UIO_USERSPACE); 3750 } 3751 3752 int 3753 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3754 register_t *retval) 3755 { 3756 /* { 3757 syscallarg(int) fd; 3758 syscallarg(const char *) path; 3759 syscallarg(const struct timespec *) tptr; 3760 syscallarg(int) flag; 3761 } */ 3762 int follow; 3763 const struct timespec *tptr; 3764 int error; 3765 3766 tptr = SCARG(uap, tptr); 3767 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3768 3769 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3770 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3771 3772 return error; 3773 } 3774 3775 /* 3776 * Common routine to set access and modification times given a vnode. 3777 */ 3778 int 3779 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3780 const struct timespec *tptr, enum uio_seg seg) 3781 { 3782 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3783 } 3784 3785 int 3786 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3787 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3788 { 3789 struct vattr vattr; 3790 int error, dorele = 0; 3791 namei_simple_flags_t sflags; 3792 bool vanull, setbirthtime; 3793 struct timespec ts[2]; 3794 3795 KASSERT(l != NULL || fdat == AT_FDCWD); 3796 3797 /* 3798 * I have checked all callers and they pass either FOLLOW, 3799 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3800 * is 0. More to the point, they don't pass anything else. 3801 * Let's keep it that way at least until the namei interfaces 3802 * are fully sanitized. 3803 */ 3804 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3805 sflags = (flag == FOLLOW) ? 3806 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3807 3808 if (tptr == NULL) { 3809 vanull = true; 3810 nanotime(&ts[0]); 3811 ts[1] = ts[0]; 3812 } else { 3813 vanull = false; 3814 if (seg != UIO_SYSSPACE) { 3815 error = copyin(tptr, ts, sizeof (ts)); 3816 if (error != 0) 3817 return error; 3818 } else { 3819 ts[0] = tptr[0]; 3820 ts[1] = tptr[1]; 3821 } 3822 } 3823 3824 if (ts[0].tv_nsec == UTIME_NOW) { 3825 nanotime(&ts[0]); 3826 if (ts[1].tv_nsec == UTIME_NOW) { 3827 vanull = true; 3828 ts[1] = ts[0]; 3829 } 3830 } else if (ts[1].tv_nsec == UTIME_NOW) 3831 nanotime(&ts[1]); 3832 3833 if (vp == NULL) { 3834 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3835 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3836 if (error != 0) 3837 return error; 3838 dorele = 1; 3839 } 3840 3841 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3842 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3843 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3844 vattr_null(&vattr); 3845 3846 if (ts[0].tv_nsec != UTIME_OMIT) 3847 vattr.va_atime = ts[0]; 3848 3849 if (ts[1].tv_nsec != UTIME_OMIT) { 3850 vattr.va_mtime = ts[1]; 3851 if (setbirthtime) 3852 vattr.va_birthtime = ts[1]; 3853 } 3854 3855 if (vanull) 3856 vattr.va_vaflags |= VA_UTIMES_NULL; 3857 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3858 VOP_UNLOCK(vp); 3859 3860 if (dorele != 0) 3861 vrele(vp); 3862 3863 return error; 3864 } 3865 3866 int 3867 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3868 const struct timeval *tptr, enum uio_seg seg) 3869 { 3870 struct timespec ts[2]; 3871 struct timespec *tsptr = NULL; 3872 int error; 3873 3874 if (tptr != NULL) { 3875 struct timeval tv[2]; 3876 3877 if (seg != UIO_SYSSPACE) { 3878 error = copyin(tptr, tv, sizeof (tv)); 3879 if (error != 0) 3880 return error; 3881 tptr = tv; 3882 } 3883 3884 if ((tv[0].tv_usec == UTIME_NOW) || 3885 (tv[0].tv_usec == UTIME_OMIT)) 3886 ts[0].tv_nsec = tv[0].tv_usec; 3887 else 3888 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3889 3890 if ((tv[1].tv_usec == UTIME_NOW) || 3891 (tv[1].tv_usec == UTIME_OMIT)) 3892 ts[1].tv_nsec = tv[1].tv_usec; 3893 else 3894 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3895 3896 tsptr = &ts[0]; 3897 } 3898 3899 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3900 } 3901 3902 /* 3903 * Truncate a file given its path name. 3904 */ 3905 /* ARGSUSED */ 3906 int 3907 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3908 { 3909 /* { 3910 syscallarg(const char *) path; 3911 syscallarg(int) pad; 3912 syscallarg(off_t) length; 3913 } */ 3914 struct vnode *vp; 3915 struct vattr vattr; 3916 int error; 3917 3918 if (SCARG(uap, length) < 0) 3919 return EINVAL; 3920 3921 error = namei_simple_user(SCARG(uap, path), 3922 NSM_FOLLOW_TRYEMULROOT, &vp); 3923 if (error != 0) 3924 return (error); 3925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3926 if (vp->v_type == VDIR) 3927 error = EISDIR; 3928 else if ((error = vn_writechk(vp)) == 0 && 3929 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3930 vattr_null(&vattr); 3931 vattr.va_size = SCARG(uap, length); 3932 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3933 } 3934 vput(vp); 3935 return (error); 3936 } 3937 3938 /* 3939 * Truncate a file given a file descriptor. 3940 */ 3941 /* ARGSUSED */ 3942 int 3943 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3944 { 3945 /* { 3946 syscallarg(int) fd; 3947 syscallarg(int) pad; 3948 syscallarg(off_t) length; 3949 } */ 3950 struct vattr vattr; 3951 struct vnode *vp; 3952 file_t *fp; 3953 int error; 3954 3955 if (SCARG(uap, length) < 0) 3956 return EINVAL; 3957 3958 /* fd_getvnode() will use the descriptor for us */ 3959 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3960 return (error); 3961 if ((fp->f_flag & FWRITE) == 0) { 3962 error = EINVAL; 3963 goto out; 3964 } 3965 vp = fp->f_vnode; 3966 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3967 if (vp->v_type == VDIR) 3968 error = EISDIR; 3969 else if ((error = vn_writechk(vp)) == 0) { 3970 vattr_null(&vattr); 3971 vattr.va_size = SCARG(uap, length); 3972 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3973 } 3974 VOP_UNLOCK(vp); 3975 out: 3976 fd_putfile(SCARG(uap, fd)); 3977 return (error); 3978 } 3979 3980 /* 3981 * Sync an open file. 3982 */ 3983 /* ARGSUSED */ 3984 int 3985 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3986 { 3987 /* { 3988 syscallarg(int) fd; 3989 } */ 3990 struct vnode *vp; 3991 file_t *fp; 3992 int error; 3993 3994 /* fd_getvnode() will use the descriptor for us */ 3995 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3996 return (error); 3997 vp = fp->f_vnode; 3998 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3999 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4000 VOP_UNLOCK(vp); 4001 fd_putfile(SCARG(uap, fd)); 4002 return (error); 4003 } 4004 4005 /* 4006 * Sync a range of file data. API modeled after that found in AIX. 4007 * 4008 * FDATASYNC indicates that we need only save enough metadata to be able 4009 * to re-read the written data. Note we duplicate AIX's requirement that 4010 * the file be open for writing. 4011 */ 4012 /* ARGSUSED */ 4013 int 4014 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4015 { 4016 /* { 4017 syscallarg(int) fd; 4018 syscallarg(int) flags; 4019 syscallarg(off_t) start; 4020 syscallarg(off_t) length; 4021 } */ 4022 struct vnode *vp; 4023 file_t *fp; 4024 int flags, nflags; 4025 off_t s, e, len; 4026 int error; 4027 4028 /* fd_getvnode() will use the descriptor for us */ 4029 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4030 return (error); 4031 4032 if ((fp->f_flag & FWRITE) == 0) { 4033 error = EBADF; 4034 goto out; 4035 } 4036 4037 flags = SCARG(uap, flags); 4038 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4039 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4040 error = EINVAL; 4041 goto out; 4042 } 4043 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4044 if (flags & FDATASYNC) 4045 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4046 else 4047 nflags = FSYNC_WAIT; 4048 if (flags & FDISKSYNC) 4049 nflags |= FSYNC_CACHE; 4050 4051 len = SCARG(uap, length); 4052 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4053 if (len) { 4054 s = SCARG(uap, start); 4055 e = s + len; 4056 if (e < s) { 4057 error = EINVAL; 4058 goto out; 4059 } 4060 } else { 4061 e = 0; 4062 s = 0; 4063 } 4064 4065 vp = fp->f_vnode; 4066 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4067 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4068 VOP_UNLOCK(vp); 4069 out: 4070 fd_putfile(SCARG(uap, fd)); 4071 return (error); 4072 } 4073 4074 /* 4075 * Sync the data of an open file. 4076 */ 4077 /* ARGSUSED */ 4078 int 4079 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4080 { 4081 /* { 4082 syscallarg(int) fd; 4083 } */ 4084 struct vnode *vp; 4085 file_t *fp; 4086 int error; 4087 4088 /* fd_getvnode() will use the descriptor for us */ 4089 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4090 return (error); 4091 if ((fp->f_flag & FWRITE) == 0) { 4092 fd_putfile(SCARG(uap, fd)); 4093 return (EBADF); 4094 } 4095 vp = fp->f_vnode; 4096 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4097 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4098 VOP_UNLOCK(vp); 4099 fd_putfile(SCARG(uap, fd)); 4100 return (error); 4101 } 4102 4103 /* 4104 * Rename files, (standard) BSD semantics frontend. 4105 */ 4106 /* ARGSUSED */ 4107 int 4108 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4109 { 4110 /* { 4111 syscallarg(const char *) from; 4112 syscallarg(const char *) to; 4113 } */ 4114 4115 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4116 SCARG(uap, to), UIO_USERSPACE, 0)); 4117 } 4118 4119 int 4120 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4121 register_t *retval) 4122 { 4123 /* { 4124 syscallarg(int) fromfd; 4125 syscallarg(const char *) from; 4126 syscallarg(int) tofd; 4127 syscallarg(const char *) to; 4128 } */ 4129 4130 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4131 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4132 } 4133 4134 /* 4135 * Rename files, POSIX semantics frontend. 4136 */ 4137 /* ARGSUSED */ 4138 int 4139 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4140 { 4141 /* { 4142 syscallarg(const char *) from; 4143 syscallarg(const char *) to; 4144 } */ 4145 4146 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4147 SCARG(uap, to), UIO_USERSPACE, 1)); 4148 } 4149 4150 /* 4151 * Rename files. Source and destination must either both be directories, 4152 * or both not be directories. If target is a directory, it must be empty. 4153 * If `from' and `to' refer to the same object, the value of the `retain' 4154 * argument is used to determine whether `from' will be 4155 * 4156 * (retain == 0) deleted unless `from' and `to' refer to the same 4157 * object in the file system's name space (BSD). 4158 * (retain == 1) always retained (POSIX). 4159 * 4160 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4161 */ 4162 int 4163 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4164 { 4165 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4166 } 4167 4168 static int 4169 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4170 const char *to, enum uio_seg seg, int retain) 4171 { 4172 struct pathbuf *fpb, *tpb; 4173 struct nameidata fnd, tnd; 4174 struct vnode *fdvp, *fvp; 4175 struct vnode *tdvp, *tvp; 4176 struct mount *mp, *tmp; 4177 int error; 4178 4179 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4180 4181 error = pathbuf_maybe_copyin(from, seg, &fpb); 4182 if (error) 4183 goto out0; 4184 KASSERT(fpb != NULL); 4185 4186 error = pathbuf_maybe_copyin(to, seg, &tpb); 4187 if (error) 4188 goto out1; 4189 KASSERT(tpb != NULL); 4190 4191 /* 4192 * Lookup from. 4193 * 4194 * XXX LOCKPARENT is wrong because we don't actually want it 4195 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4196 * insane, so for the time being we need to leave it like this. 4197 */ 4198 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4199 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4200 goto out2; 4201 4202 /* 4203 * Pull out the important results of the lookup, fdvp and fvp. 4204 * Of course, fvp is bogus because we're about to unlock fdvp. 4205 */ 4206 fdvp = fnd.ni_dvp; 4207 fvp = fnd.ni_vp; 4208 KASSERT(fdvp != NULL); 4209 KASSERT(fvp != NULL); 4210 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4211 4212 /* 4213 * Make sure neither fdvp nor fvp is locked. 4214 */ 4215 if (fdvp != fvp) 4216 VOP_UNLOCK(fdvp); 4217 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4218 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4219 4220 /* 4221 * Reject renaming `.' and `..'. Can't do this until after 4222 * namei because we need namei's parsing to find the final 4223 * component name. (namei should just leave us with the final 4224 * component name and not look it up itself, but anyway...) 4225 * 4226 * This was here before because we used to relookup from 4227 * instead of to and relookup requires the caller to check 4228 * this, but now file systems may depend on this check, so we 4229 * must retain it until the file systems are all rototilled. 4230 */ 4231 if (((fnd.ni_cnd.cn_namelen == 1) && 4232 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4233 ((fnd.ni_cnd.cn_namelen == 2) && 4234 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4235 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4236 error = EINVAL; /* XXX EISDIR? */ 4237 goto abort0; 4238 } 4239 4240 /* 4241 * Lookup to. 4242 * 4243 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4244 * fvp here to decide whether to add CREATEDIR is a load of 4245 * bollocks because fvp might be the wrong node by now, since 4246 * fdvp is unlocked. 4247 * 4248 * XXX Why not pass CREATEDIR always? 4249 */ 4250 NDINIT(&tnd, RENAME, 4251 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4252 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4253 tpb); 4254 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4255 goto abort0; 4256 4257 /* 4258 * Pull out the important results of the lookup, tdvp and tvp. 4259 * Of course, tvp is bogus because we're about to unlock tdvp. 4260 */ 4261 tdvp = tnd.ni_dvp; 4262 tvp = tnd.ni_vp; 4263 KASSERT(tdvp != NULL); 4264 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4265 4266 /* 4267 * Make sure neither tdvp nor tvp is locked. 4268 */ 4269 if (tdvp != tvp) 4270 VOP_UNLOCK(tdvp); 4271 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4272 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4273 4274 /* 4275 * Reject renaming onto `.' or `..'. relookup is unhappy with 4276 * these, which is why we must do this here. Once upon a time 4277 * we relooked up from instead of to, and consequently didn't 4278 * need this check, but now that we relookup to instead of 4279 * from, we need this; and we shall need it forever forward 4280 * until the VOP_RENAME protocol changes, because file systems 4281 * will no doubt begin to depend on this check. 4282 */ 4283 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4284 error = EISDIR; 4285 goto abort1; 4286 } 4287 if ((tnd.ni_cnd.cn_namelen == 2) && 4288 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4289 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4290 error = EINVAL; 4291 goto abort1; 4292 } 4293 4294 /* 4295 * Get the mount point. If the file system has been unmounted, 4296 * which it may be because we're not holding any vnode locks, 4297 * then v_mount will be NULL. We're not really supposed to 4298 * read v_mount without holding the vnode lock, but since we 4299 * have fdvp referenced, if fdvp->v_mount changes then at worst 4300 * it will be set to NULL, not changed to another mount point. 4301 * And, of course, since it is up to the file system to 4302 * determine the real lock order, we can't lock both fdvp and 4303 * tdvp at the same time. 4304 */ 4305 mp = fdvp->v_mount; 4306 if (mp == NULL) { 4307 error = ENOENT; 4308 goto abort1; 4309 } 4310 4311 /* 4312 * Make sure the mount points match. Again, although we don't 4313 * hold any vnode locks, the v_mount fields may change -- but 4314 * at worst they will change to NULL, so this will never become 4315 * a cross-device rename, because we hold vnode references. 4316 * 4317 * XXX Because nothing is locked and the compiler may reorder 4318 * things here, unmounting the file system at an inopportune 4319 * moment may cause rename to fail with EXDEV when it really 4320 * should fail with ENOENT. 4321 */ 4322 tmp = tdvp->v_mount; 4323 if (tmp == NULL) { 4324 error = ENOENT; 4325 goto abort1; 4326 } 4327 4328 if (mp != tmp) { 4329 error = EXDEV; 4330 goto abort1; 4331 } 4332 4333 /* 4334 * Take the vfs rename lock to avoid cross-directory screw cases. 4335 * Nothing is locked currently, so taking this lock is safe. 4336 */ 4337 error = VFS_RENAMELOCK_ENTER(mp); 4338 if (error) 4339 goto abort1; 4340 4341 /* 4342 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4343 * and nothing is locked except for the vfs rename lock. 4344 * 4345 * The next step is a little rain dance to conform to the 4346 * insane lock protocol, even though it does nothing to ward 4347 * off race conditions. 4348 * 4349 * We need tdvp and tvp to be locked. However, because we have 4350 * unlocked tdvp in order to hold no locks while we take the 4351 * vfs rename lock, tvp may be wrong here, and we can't safely 4352 * lock it even if the sensible file systems will just unlock 4353 * it straight away. Consequently, we must lock tdvp and then 4354 * relookup tvp to get it locked. 4355 * 4356 * Finally, because the VOP_RENAME protocol is brain-damaged 4357 * and various file systems insanely depend on the semantics of 4358 * this brain damage, the lookup of to must be the last lookup 4359 * before VOP_RENAME. 4360 */ 4361 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4362 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4363 if (error) 4364 goto abort2; 4365 4366 /* 4367 * Drop the old tvp and pick up the new one -- which might be 4368 * the same, but that doesn't matter to us. After this, tdvp 4369 * and tvp should both be locked. 4370 */ 4371 if (tvp != NULL) 4372 vrele(tvp); 4373 tvp = tnd.ni_vp; 4374 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4375 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4376 4377 /* 4378 * The old do_sys_rename had various consistency checks here 4379 * involving fvp and tvp. fvp is bogus already here, and tvp 4380 * will become bogus soon in any sensible file system, so the 4381 * only purpose in putting these checks here is to give lip 4382 * service to these screw cases and to acknowledge that they 4383 * exist, not actually to handle them, but here you go 4384 * anyway... 4385 */ 4386 4387 /* 4388 * Acknowledge that directories and non-directories aren't 4389 * suposed to mix. 4390 */ 4391 if (tvp != NULL) { 4392 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4393 error = ENOTDIR; 4394 goto abort3; 4395 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4396 error = EISDIR; 4397 goto abort3; 4398 } 4399 } 4400 4401 /* 4402 * Acknowledge some random screw case, among the dozens that 4403 * might arise. 4404 */ 4405 if (fvp == tdvp) { 4406 error = EINVAL; 4407 goto abort3; 4408 } 4409 4410 /* 4411 * Acknowledge that POSIX has a wacky screw case. 4412 * 4413 * XXX Eventually the retain flag needs to be passed on to 4414 * VOP_RENAME. 4415 */ 4416 if (fvp == tvp) { 4417 if (retain) { 4418 error = 0; 4419 goto abort3; 4420 } else if ((fdvp == tdvp) && 4421 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4422 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4423 fnd.ni_cnd.cn_namelen))) { 4424 error = 0; 4425 goto abort3; 4426 } 4427 } 4428 4429 /* 4430 * Make sure veriexec can screw us up. (But a race can screw 4431 * up veriexec, of course -- remember, fvp and (soon) tvp are 4432 * bogus.) 4433 */ 4434 #if NVERIEXEC > 0 4435 { 4436 char *f1, *f2; 4437 size_t f1_len; 4438 size_t f2_len; 4439 4440 f1_len = fnd.ni_cnd.cn_namelen + 1; 4441 f1 = kmem_alloc(f1_len, KM_SLEEP); 4442 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4443 4444 f2_len = tnd.ni_cnd.cn_namelen + 1; 4445 f2 = kmem_alloc(f2_len, KM_SLEEP); 4446 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4447 4448 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4449 4450 kmem_free(f1, f1_len); 4451 kmem_free(f2, f2_len); 4452 4453 if (error) 4454 goto abort3; 4455 } 4456 #endif /* NVERIEXEC > 0 */ 4457 4458 /* 4459 * All ready. Incant the rename vop. 4460 */ 4461 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4462 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4463 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4464 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4465 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4466 4467 /* 4468 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4469 * tdvp and tvp. But we can't assert any of that. 4470 */ 4471 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4472 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4473 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4474 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4475 4476 /* 4477 * So all we have left to do is to drop the rename lock and 4478 * destroy the pathbufs. 4479 */ 4480 VFS_RENAMELOCK_EXIT(mp); 4481 goto out2; 4482 4483 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4484 VOP_UNLOCK(tvp); 4485 abort2: VOP_UNLOCK(tdvp); 4486 VFS_RENAMELOCK_EXIT(mp); 4487 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4488 vrele(tdvp); 4489 if (tvp != NULL) 4490 vrele(tvp); 4491 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4492 vrele(fdvp); 4493 vrele(fvp); 4494 out2: pathbuf_destroy(tpb); 4495 out1: pathbuf_destroy(fpb); 4496 out0: return error; 4497 } 4498 4499 /* 4500 * Make a directory file. 4501 */ 4502 /* ARGSUSED */ 4503 int 4504 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4505 { 4506 /* { 4507 syscallarg(const char *) path; 4508 syscallarg(int) mode; 4509 } */ 4510 4511 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4512 SCARG(uap, mode), UIO_USERSPACE); 4513 } 4514 4515 int 4516 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4517 register_t *retval) 4518 { 4519 /* { 4520 syscallarg(int) fd; 4521 syscallarg(const char *) path; 4522 syscallarg(int) mode; 4523 } */ 4524 4525 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4526 SCARG(uap, mode), UIO_USERSPACE); 4527 } 4528 4529 4530 int 4531 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4532 { 4533 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4534 } 4535 4536 static int 4537 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4538 enum uio_seg seg) 4539 { 4540 struct proc *p = curlwp->l_proc; 4541 struct vnode *vp; 4542 struct vattr vattr; 4543 int error; 4544 struct pathbuf *pb; 4545 struct nameidata nd; 4546 4547 KASSERT(l != NULL || fdat == AT_FDCWD); 4548 4549 /* XXX bollocks, should pass in a pathbuf */ 4550 error = pathbuf_maybe_copyin(path, seg, &pb); 4551 if (error) { 4552 return error; 4553 } 4554 4555 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4556 4557 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4558 pathbuf_destroy(pb); 4559 return (error); 4560 } 4561 vp = nd.ni_vp; 4562 if (vp != NULL) { 4563 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4564 if (nd.ni_dvp == vp) 4565 vrele(nd.ni_dvp); 4566 else 4567 vput(nd.ni_dvp); 4568 vrele(vp); 4569 pathbuf_destroy(pb); 4570 return (EEXIST); 4571 } 4572 vattr_null(&vattr); 4573 vattr.va_type = VDIR; 4574 /* We will read cwdi->cwdi_cmask unlocked. */ 4575 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4576 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4577 if (!error) 4578 vrele(nd.ni_vp); 4579 vput(nd.ni_dvp); 4580 pathbuf_destroy(pb); 4581 return (error); 4582 } 4583 4584 /* 4585 * Remove a directory file. 4586 */ 4587 /* ARGSUSED */ 4588 int 4589 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4590 { 4591 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4592 AT_REMOVEDIR, UIO_USERSPACE); 4593 } 4594 4595 /* 4596 * Read a block of directory entries in a file system independent format. 4597 */ 4598 int 4599 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4600 { 4601 /* { 4602 syscallarg(int) fd; 4603 syscallarg(char *) buf; 4604 syscallarg(size_t) count; 4605 } */ 4606 file_t *fp; 4607 int error, done; 4608 4609 /* fd_getvnode() will use the descriptor for us */ 4610 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4611 return (error); 4612 if ((fp->f_flag & FREAD) == 0) { 4613 error = EBADF; 4614 goto out; 4615 } 4616 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4617 SCARG(uap, count), &done, l, 0, 0); 4618 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4619 *retval = done; 4620 out: 4621 fd_putfile(SCARG(uap, fd)); 4622 return (error); 4623 } 4624 4625 /* 4626 * Set the mode mask for creation of filesystem nodes. 4627 */ 4628 int 4629 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4630 { 4631 /* { 4632 syscallarg(mode_t) newmask; 4633 } */ 4634 struct proc *p = l->l_proc; 4635 struct cwdinfo *cwdi; 4636 4637 /* 4638 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4639 * important is that we serialize changes to the mask. The 4640 * rw_exit() will issue a write memory barrier on our behalf, 4641 * and force the changes out to other CPUs (as it must use an 4642 * atomic operation, draining the local CPU's store buffers). 4643 */ 4644 cwdi = p->p_cwdi; 4645 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4646 *retval = cwdi->cwdi_cmask; 4647 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4648 rw_exit(&cwdi->cwdi_lock); 4649 4650 return (0); 4651 } 4652 4653 int 4654 dorevoke(struct vnode *vp, kauth_cred_t cred) 4655 { 4656 struct vattr vattr; 4657 int error, fs_decision; 4658 4659 vn_lock(vp, LK_SHARED | LK_RETRY); 4660 error = VOP_GETATTR(vp, &vattr, cred); 4661 VOP_UNLOCK(vp); 4662 if (error != 0) 4663 return error; 4664 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4665 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4666 fs_decision); 4667 if (!error) 4668 VOP_REVOKE(vp, REVOKEALL); 4669 return (error); 4670 } 4671 4672 /* 4673 * Void all references to file by ripping underlying filesystem 4674 * away from vnode. 4675 */ 4676 /* ARGSUSED */ 4677 int 4678 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4679 { 4680 /* { 4681 syscallarg(const char *) path; 4682 } */ 4683 struct vnode *vp; 4684 int error; 4685 4686 error = namei_simple_user(SCARG(uap, path), 4687 NSM_FOLLOW_TRYEMULROOT, &vp); 4688 if (error != 0) 4689 return (error); 4690 error = dorevoke(vp, l->l_cred); 4691 vrele(vp); 4692 return (error); 4693 } 4694 4695 /* 4696 * Allocate backing store for a file, filling a hole without having to 4697 * explicitly write anything out. 4698 */ 4699 /* ARGSUSED */ 4700 int 4701 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4702 register_t *retval) 4703 { 4704 /* { 4705 syscallarg(int) fd; 4706 syscallarg(off_t) pos; 4707 syscallarg(off_t) len; 4708 } */ 4709 int fd; 4710 off_t pos, len; 4711 struct file *fp; 4712 struct vnode *vp; 4713 int error; 4714 4715 fd = SCARG(uap, fd); 4716 pos = SCARG(uap, pos); 4717 len = SCARG(uap, len); 4718 4719 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4720 *retval = EINVAL; 4721 return 0; 4722 } 4723 4724 error = fd_getvnode(fd, &fp); 4725 if (error) { 4726 *retval = error; 4727 return 0; 4728 } 4729 if ((fp->f_flag & FWRITE) == 0) { 4730 error = EBADF; 4731 goto fail; 4732 } 4733 vp = fp->f_vnode; 4734 4735 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4736 if (vp->v_type == VDIR) { 4737 error = EISDIR; 4738 } else { 4739 error = VOP_FALLOCATE(vp, pos, len); 4740 } 4741 VOP_UNLOCK(vp); 4742 4743 fail: 4744 fd_putfile(fd); 4745 *retval = error; 4746 return 0; 4747 } 4748 4749 /* 4750 * Deallocate backing store for a file, creating a hole. Also used for 4751 * invoking TRIM on disks. 4752 */ 4753 /* ARGSUSED */ 4754 int 4755 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4756 register_t *retval) 4757 { 4758 /* { 4759 syscallarg(int) fd; 4760 syscallarg(off_t) pos; 4761 syscallarg(off_t) len; 4762 } */ 4763 int fd; 4764 off_t pos, len; 4765 struct file *fp; 4766 struct vnode *vp; 4767 int error; 4768 4769 fd = SCARG(uap, fd); 4770 pos = SCARG(uap, pos); 4771 len = SCARG(uap, len); 4772 4773 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4774 return EINVAL; 4775 } 4776 4777 error = fd_getvnode(fd, &fp); 4778 if (error) { 4779 return error; 4780 } 4781 if ((fp->f_flag & FWRITE) == 0) { 4782 error = EBADF; 4783 goto fail; 4784 } 4785 vp = fp->f_vnode; 4786 4787 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4788 if (vp->v_type == VDIR) { 4789 error = EISDIR; 4790 } else { 4791 error = VOP_FDISCARD(vp, pos, len); 4792 } 4793 VOP_UNLOCK(vp); 4794 4795 fail: 4796 fd_putfile(fd); 4797 return error; 4798 } 4799
Indexes created Fri Oct 03 02:09:57 GMT 2025