vfs_syscalls.c revision 1.496
1 /* $NetBSD: vfs_syscalls.c,v 1.496 2015/04/20 23:03:08 riastradh 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 1.496 2015/04/20 23:03:08 riastradh 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/proc.h> 91 #include <sys/uio.h> 92 #include <sys/kmem.h> 93 #include <sys/dirent.h> 94 #include <sys/sysctl.h> 95 #include <sys/syscallargs.h> 96 #include <sys/vfs_syscalls.h> 97 #include <sys/quota.h> 98 #include <sys/quotactl.h> 99 #include <sys/ktrace.h> 100 #ifdef FILEASSOC 101 #include <sys/fileassoc.h> 102 #endif /* FILEASSOC */ 103 #include <sys/extattr.h> 104 #include <sys/verified_exec.h> 105 #include <sys/kauth.h> 106 #include <sys/atomic.h> 107 #include <sys/module.h> 108 #include <sys/buf.h> 109 110 #include <miscfs/genfs/genfs.h> 111 #include <miscfs/syncfs/syncfs.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 if (vfs_busy(mp, NULL)) { 286 error = EPERM; 287 goto out; 288 } 289 290 mutex_enter(&mp->mnt_updating); 291 292 mp->mnt_flag &= ~MNT_OP_FLAGS; 293 mp->mnt_flag |= flags & MNT_OP_FLAGS; 294 295 /* 296 * Set the mount level flags. 297 */ 298 if (flags & MNT_RDONLY) 299 mp->mnt_flag |= MNT_RDONLY; 300 else if (mp->mnt_flag & MNT_RDONLY) 301 mp->mnt_iflag |= IMNT_WANTRDWR; 302 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 303 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 304 error = VFS_MOUNT(mp, path, data, data_len); 305 306 if (error && data != NULL) { 307 int error2; 308 309 /* 310 * Update failed; let's try and see if it was an 311 * export request. For compat with 3.0 and earlier. 312 */ 313 error2 = vfs_hooks_reexport(mp, path, data); 314 315 /* 316 * Only update error code if the export request was 317 * understood but some problem occurred while 318 * processing it. 319 */ 320 if (error2 != EJUSTRETURN) 321 error = error2; 322 } 323 324 if (mp->mnt_iflag & IMNT_WANTRDWR) 325 mp->mnt_flag &= ~MNT_RDONLY; 326 if (error) 327 mp->mnt_flag = saved_flags; 328 mp->mnt_flag &= ~MNT_OP_FLAGS; 329 mp->mnt_iflag &= ~IMNT_WANTRDWR; 330 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 331 if (mp->mnt_syncer == NULL) 332 error = vfs_allocate_syncvnode(mp); 333 } else { 334 if (mp->mnt_syncer != NULL) 335 vfs_deallocate_syncvnode(mp); 336 } 337 mutex_exit(&mp->mnt_updating); 338 vfs_unbusy(mp, false, NULL); 339 340 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 341 (flags & MNT_EXTATTR)) { 342 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 343 NULL, 0, NULL) != 0) { 344 printf("%s: failed to start extattr, error = %d", 345 mp->mnt_stat.f_mntonname, error); 346 mp->mnt_flag &= ~MNT_EXTATTR; 347 } 348 } 349 350 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 351 !(flags & MNT_EXTATTR)) { 352 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 353 NULL, 0, NULL) != 0) { 354 printf("%s: failed to stop extattr, error = %d", 355 mp->mnt_stat.f_mntonname, error); 356 mp->mnt_flag |= MNT_RDONLY; 357 } 358 } 359 out: 360 return (error); 361 } 362 363 static int 364 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 365 { 366 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 367 int error; 368 369 /* Copy file-system type from userspace. */ 370 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 371 if (error) { 372 /* 373 * Historically, filesystem types were identified by numbers. 374 * If we get an integer for the filesystem type instead of a 375 * string, we check to see if it matches one of the historic 376 * filesystem types. 377 */ 378 u_long fsindex = (u_long)fstype; 379 if (fsindex >= nmountcompatnames || 380 mountcompatnames[fsindex] == NULL) 381 return ENODEV; 382 strlcpy(fstypename, mountcompatnames[fsindex], 383 sizeof(fstypename)); 384 } 385 386 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 387 if (strcmp(fstypename, "ufs") == 0) 388 fstypename[0] = 'f'; 389 390 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 391 return 0; 392 393 /* If we can autoload a vfs module, try again */ 394 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 395 396 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 397 return 0; 398 399 return ENODEV; 400 } 401 402 static int 403 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 404 void *data, size_t *data_len) 405 { 406 struct mount *mp; 407 int error; 408 409 /* If MNT_GETARGS is specified, it should be the only flag. */ 410 if (flags & ~MNT_GETARGS) 411 return EINVAL; 412 413 mp = vp->v_mount; 414 415 /* XXX: probably some notion of "can see" here if we want isolation. */ 416 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 417 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 418 if (error) 419 return error; 420 421 if ((vp->v_vflag & VV_ROOT) == 0) 422 return EINVAL; 423 424 if (vfs_busy(mp, NULL)) 425 return EPERM; 426 427 mutex_enter(&mp->mnt_updating); 428 mp->mnt_flag &= ~MNT_OP_FLAGS; 429 mp->mnt_flag |= MNT_GETARGS; 430 error = VFS_MOUNT(mp, path, data, data_len); 431 mp->mnt_flag &= ~MNT_OP_FLAGS; 432 mutex_exit(&mp->mnt_updating); 433 434 vfs_unbusy(mp, false, NULL); 435 return (error); 436 } 437 438 int 439 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 440 { 441 /* { 442 syscallarg(const char *) type; 443 syscallarg(const char *) path; 444 syscallarg(int) flags; 445 syscallarg(void *) data; 446 syscallarg(size_t) data_len; 447 } */ 448 449 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 450 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 451 SCARG(uap, data_len), retval); 452 } 453 454 int 455 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 456 const char *path, int flags, void *data, enum uio_seg data_seg, 457 size_t data_len, register_t *retval) 458 { 459 struct vnode *vp; 460 void *data_buf = data; 461 bool vfsopsrele = false; 462 size_t alloc_sz = 0; 463 int error; 464 465 /* XXX: The calling convention of this routine is totally bizarre */ 466 if (vfsops) 467 vfsopsrele = true; 468 469 /* 470 * Get vnode to be covered 471 */ 472 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 473 if (error != 0) { 474 vp = NULL; 475 goto done; 476 } 477 478 if (vfsops == NULL) { 479 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 480 vfsops = vp->v_mount->mnt_op; 481 } else { 482 /* 'type' is userspace */ 483 error = mount_get_vfsops(type, &vfsops); 484 if (error != 0) 485 goto done; 486 vfsopsrele = true; 487 } 488 } 489 490 /* 491 * We allow data to be NULL, even for userspace. Some fs's don't need 492 * it. The others will handle NULL. 493 */ 494 if (data != NULL && data_seg == UIO_USERSPACE) { 495 if (data_len == 0) { 496 /* No length supplied, use default for filesystem */ 497 data_len = vfsops->vfs_min_mount_data; 498 499 /* 500 * Hopefully a longer buffer won't make copyin() fail. 501 * For compatibility with 3.0 and earlier. 502 */ 503 if (flags & MNT_UPDATE 504 && data_len < sizeof (struct mnt_export_args30)) 505 data_len = sizeof (struct mnt_export_args30); 506 } 507 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 508 error = EINVAL; 509 goto done; 510 } 511 alloc_sz = data_len; 512 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 513 514 /* NFS needs the buffer even for mnt_getargs .... */ 515 error = copyin(data, data_buf, data_len); 516 if (error != 0) 517 goto done; 518 } 519 520 if (flags & MNT_GETARGS) { 521 if (data_len == 0) { 522 error = EINVAL; 523 goto done; 524 } 525 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 526 if (error != 0) 527 goto done; 528 if (data_seg == UIO_USERSPACE) 529 error = copyout(data_buf, data, data_len); 530 *retval = data_len; 531 } else if (flags & MNT_UPDATE) { 532 error = mount_update(l, vp, path, flags, data_buf, &data_len); 533 } else { 534 /* Locking is handled internally in mount_domount(). */ 535 KASSERT(vfsopsrele == true); 536 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 537 &data_len); 538 vfsopsrele = false; 539 } 540 541 done: 542 if (vfsopsrele) 543 vfs_delref(vfsops); 544 if (vp != NULL) { 545 vrele(vp); 546 } 547 if (data_buf != data) 548 kmem_free(data_buf, alloc_sz); 549 return (error); 550 } 551 552 /* 553 * Unmount a file system. 554 * 555 * Note: unmount takes a path to the vnode mounted on as argument, 556 * not special file (as before). 557 */ 558 /* ARGSUSED */ 559 int 560 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 561 { 562 /* { 563 syscallarg(const char *) path; 564 syscallarg(int) flags; 565 } */ 566 struct vnode *vp; 567 struct mount *mp; 568 int error; 569 struct pathbuf *pb; 570 struct nameidata nd; 571 572 error = pathbuf_copyin(SCARG(uap, path), &pb); 573 if (error) { 574 return error; 575 } 576 577 NDINIT(&nd, LOOKUP, LOCKLEAF | TRYEMULROOT, pb); 578 if ((error = namei(&nd)) != 0) { 579 pathbuf_destroy(pb); 580 return error; 581 } 582 vp = nd.ni_vp; 583 pathbuf_destroy(pb); 584 585 mp = vp->v_mount; 586 atomic_inc_uint(&mp->mnt_refcnt); 587 VOP_UNLOCK(vp); 588 589 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 590 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 591 if (error) { 592 vrele(vp); 593 vfs_destroy(mp); 594 return (error); 595 } 596 597 /* 598 * Don't allow unmounting the root file system. 599 */ 600 if (mp->mnt_flag & MNT_ROOTFS) { 601 vrele(vp); 602 vfs_destroy(mp); 603 return (EINVAL); 604 } 605 606 /* 607 * Must be the root of the filesystem 608 */ 609 if ((vp->v_vflag & VV_ROOT) == 0) { 610 vrele(vp); 611 vfs_destroy(mp); 612 return (EINVAL); 613 } 614 615 vrele(vp); 616 error = dounmount(mp, SCARG(uap, flags), l); 617 vfs_destroy(mp); 618 return error; 619 } 620 621 /* 622 * Sync each mounted filesystem. 623 */ 624 #ifdef DEBUG 625 int syncprt = 0; 626 struct ctldebug debug0 = { "syncprt", &syncprt }; 627 #endif 628 629 void 630 do_sys_sync(struct lwp *l) 631 { 632 struct mount *mp, *nmp; 633 int asyncflag; 634 635 mutex_enter(&mountlist_lock); 636 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 637 if (vfs_busy(mp, &nmp)) { 638 continue; 639 } 640 mutex_enter(&mp->mnt_updating); 641 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 642 asyncflag = mp->mnt_flag & MNT_ASYNC; 643 mp->mnt_flag &= ~MNT_ASYNC; 644 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 645 if (asyncflag) 646 mp->mnt_flag |= MNT_ASYNC; 647 } 648 mutex_exit(&mp->mnt_updating); 649 vfs_unbusy(mp, false, &nmp); 650 } 651 mutex_exit(&mountlist_lock); 652 #ifdef DEBUG 653 if (syncprt) 654 vfs_bufstats(); 655 #endif /* DEBUG */ 656 } 657 658 /* ARGSUSED */ 659 int 660 sys_sync(struct lwp *l, const void *v, register_t *retval) 661 { 662 do_sys_sync(l); 663 return (0); 664 } 665 666 667 /* 668 * Access or change filesystem quotas. 669 * 670 * (this is really 14 different calls bundled into one) 671 */ 672 673 static int 674 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 675 { 676 struct quotastat info_k; 677 int error; 678 679 /* ensure any padding bytes are cleared */ 680 memset(&info_k, 0, sizeof(info_k)); 681 682 error = vfs_quotactl_stat(mp, &info_k); 683 if (error) { 684 return error; 685 } 686 687 return copyout(&info_k, info_u, sizeof(info_k)); 688 } 689 690 static int 691 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 692 struct quotaidtypestat *info_u) 693 { 694 struct quotaidtypestat info_k; 695 int error; 696 697 /* ensure any padding bytes are cleared */ 698 memset(&info_k, 0, sizeof(info_k)); 699 700 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 701 if (error) { 702 return error; 703 } 704 705 return copyout(&info_k, info_u, sizeof(info_k)); 706 } 707 708 static int 709 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 710 struct quotaobjtypestat *info_u) 711 { 712 struct quotaobjtypestat info_k; 713 int error; 714 715 /* ensure any padding bytes are cleared */ 716 memset(&info_k, 0, sizeof(info_k)); 717 718 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 719 if (error) { 720 return error; 721 } 722 723 return copyout(&info_k, info_u, sizeof(info_k)); 724 } 725 726 static int 727 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 728 struct quotaval *val_u) 729 { 730 struct quotakey key_k; 731 struct quotaval val_k; 732 int error; 733 734 /* ensure any padding bytes are cleared */ 735 memset(&val_k, 0, sizeof(val_k)); 736 737 error = copyin(key_u, &key_k, sizeof(key_k)); 738 if (error) { 739 return error; 740 } 741 742 error = vfs_quotactl_get(mp, &key_k, &val_k); 743 if (error) { 744 return error; 745 } 746 747 return copyout(&val_k, val_u, sizeof(val_k)); 748 } 749 750 static int 751 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 752 const struct quotaval *val_u) 753 { 754 struct quotakey key_k; 755 struct quotaval val_k; 756 int error; 757 758 error = copyin(key_u, &key_k, sizeof(key_k)); 759 if (error) { 760 return error; 761 } 762 763 error = copyin(val_u, &val_k, sizeof(val_k)); 764 if (error) { 765 return error; 766 } 767 768 return vfs_quotactl_put(mp, &key_k, &val_k); 769 } 770 771 static int 772 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 773 { 774 struct quotakey key_k; 775 int error; 776 777 error = copyin(key_u, &key_k, sizeof(key_k)); 778 if (error) { 779 return error; 780 } 781 782 return vfs_quotactl_del(mp, &key_k); 783 } 784 785 static int 786 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 787 { 788 struct quotakcursor cursor_k; 789 int error; 790 791 /* ensure any padding bytes are cleared */ 792 memset(&cursor_k, 0, sizeof(cursor_k)); 793 794 error = vfs_quotactl_cursoropen(mp, &cursor_k); 795 if (error) { 796 return error; 797 } 798 799 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 800 } 801 802 static int 803 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 804 { 805 struct quotakcursor cursor_k; 806 int error; 807 808 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 809 if (error) { 810 return error; 811 } 812 813 return vfs_quotactl_cursorclose(mp, &cursor_k); 814 } 815 816 static int 817 do_sys_quotactl_cursorskipidtype(struct mount *mp, 818 struct quotakcursor *cursor_u, int idtype) 819 { 820 struct quotakcursor cursor_k; 821 int error; 822 823 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 824 if (error) { 825 return error; 826 } 827 828 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 829 if (error) { 830 return error; 831 } 832 833 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 834 } 835 836 static int 837 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 838 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 839 unsigned *ret_u) 840 { 841 #define CGET_STACK_MAX 8 842 struct quotakcursor cursor_k; 843 struct quotakey stackkeys[CGET_STACK_MAX]; 844 struct quotaval stackvals[CGET_STACK_MAX]; 845 struct quotakey *keys_k; 846 struct quotaval *vals_k; 847 unsigned ret_k; 848 int error; 849 850 if (maxnum > 128) { 851 maxnum = 128; 852 } 853 854 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 855 if (error) { 856 return error; 857 } 858 859 if (maxnum <= CGET_STACK_MAX) { 860 keys_k = stackkeys; 861 vals_k = stackvals; 862 /* ensure any padding bytes are cleared */ 863 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 864 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 865 } else { 866 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 867 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 868 } 869 870 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 871 &ret_k); 872 if (error) { 873 goto fail; 874 } 875 876 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 877 if (error) { 878 goto fail; 879 } 880 881 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 882 if (error) { 883 goto fail; 884 } 885 886 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 887 if (error) { 888 goto fail; 889 } 890 891 /* do last to maximize the chance of being able to recover a failure */ 892 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 893 894 fail: 895 if (keys_k != stackkeys) { 896 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 897 } 898 if (vals_k != stackvals) { 899 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 900 } 901 return error; 902 } 903 904 static int 905 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 906 int *ret_u) 907 { 908 struct quotakcursor cursor_k; 909 int ret_k; 910 int error; 911 912 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 913 if (error) { 914 return error; 915 } 916 917 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 918 if (error) { 919 return error; 920 } 921 922 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 923 if (error) { 924 return error; 925 } 926 927 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 928 } 929 930 static int 931 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 932 { 933 struct quotakcursor cursor_k; 934 int error; 935 936 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 937 if (error) { 938 return error; 939 } 940 941 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 942 if (error) { 943 return error; 944 } 945 946 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 947 } 948 949 static int 950 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 951 { 952 char *path_k; 953 int error; 954 955 /* XXX this should probably be a struct pathbuf */ 956 path_k = PNBUF_GET(); 957 error = copyin(path_u, path_k, PATH_MAX); 958 if (error) { 959 PNBUF_PUT(path_k); 960 return error; 961 } 962 963 error = vfs_quotactl_quotaon(mp, idtype, path_k); 964 965 PNBUF_PUT(path_k); 966 return error; 967 } 968 969 static int 970 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 971 { 972 return vfs_quotactl_quotaoff(mp, idtype); 973 } 974 975 int 976 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 977 { 978 struct mount *mp; 979 struct vnode *vp; 980 int error; 981 982 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 983 if (error != 0) 984 return (error); 985 mp = vp->v_mount; 986 987 switch (args->qc_op) { 988 case QUOTACTL_STAT: 989 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 990 break; 991 case QUOTACTL_IDTYPESTAT: 992 error = do_sys_quotactl_idtypestat(mp, 993 args->u.idtypestat.qc_idtype, 994 args->u.idtypestat.qc_info); 995 break; 996 case QUOTACTL_OBJTYPESTAT: 997 error = do_sys_quotactl_objtypestat(mp, 998 args->u.objtypestat.qc_objtype, 999 args->u.objtypestat.qc_info); 1000 break; 1001 case QUOTACTL_GET: 1002 error = do_sys_quotactl_get(mp, 1003 args->u.get.qc_key, 1004 args->u.get.qc_val); 1005 break; 1006 case QUOTACTL_PUT: 1007 error = do_sys_quotactl_put(mp, 1008 args->u.put.qc_key, 1009 args->u.put.qc_val); 1010 break; 1011 case QUOTACTL_DEL: 1012 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1013 break; 1014 case QUOTACTL_CURSOROPEN: 1015 error = do_sys_quotactl_cursoropen(mp, 1016 args->u.cursoropen.qc_cursor); 1017 break; 1018 case QUOTACTL_CURSORCLOSE: 1019 error = do_sys_quotactl_cursorclose(mp, 1020 args->u.cursorclose.qc_cursor); 1021 break; 1022 case QUOTACTL_CURSORSKIPIDTYPE: 1023 error = do_sys_quotactl_cursorskipidtype(mp, 1024 args->u.cursorskipidtype.qc_cursor, 1025 args->u.cursorskipidtype.qc_idtype); 1026 break; 1027 case QUOTACTL_CURSORGET: 1028 error = do_sys_quotactl_cursorget(mp, 1029 args->u.cursorget.qc_cursor, 1030 args->u.cursorget.qc_keys, 1031 args->u.cursorget.qc_vals, 1032 args->u.cursorget.qc_maxnum, 1033 args->u.cursorget.qc_ret); 1034 break; 1035 case QUOTACTL_CURSORATEND: 1036 error = do_sys_quotactl_cursoratend(mp, 1037 args->u.cursoratend.qc_cursor, 1038 args->u.cursoratend.qc_ret); 1039 break; 1040 case QUOTACTL_CURSORREWIND: 1041 error = do_sys_quotactl_cursorrewind(mp, 1042 args->u.cursorrewind.qc_cursor); 1043 break; 1044 case QUOTACTL_QUOTAON: 1045 error = do_sys_quotactl_quotaon(mp, 1046 args->u.quotaon.qc_idtype, 1047 args->u.quotaon.qc_quotafile); 1048 break; 1049 case QUOTACTL_QUOTAOFF: 1050 error = do_sys_quotactl_quotaoff(mp, 1051 args->u.quotaoff.qc_idtype); 1052 break; 1053 default: 1054 error = EINVAL; 1055 break; 1056 } 1057 1058 vrele(vp); 1059 return error; 1060 } 1061 1062 /* ARGSUSED */ 1063 int 1064 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1065 register_t *retval) 1066 { 1067 /* { 1068 syscallarg(const char *) path; 1069 syscallarg(struct quotactl_args *) args; 1070 } */ 1071 struct quotactl_args args; 1072 int error; 1073 1074 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1075 if (error) { 1076 return error; 1077 } 1078 1079 return do_sys_quotactl(SCARG(uap, path), &args); 1080 } 1081 1082 int 1083 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1084 int root) 1085 { 1086 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1087 int error = 0; 1088 1089 /* 1090 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1091 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1092 * overrides MNT_NOWAIT. 1093 */ 1094 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1095 (flags != MNT_WAIT && flags != 0)) { 1096 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1097 goto done; 1098 } 1099 1100 /* Get the filesystem stats now */ 1101 memset(sp, 0, sizeof(*sp)); 1102 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1103 return error; 1104 } 1105 1106 if (cwdi->cwdi_rdir == NULL) 1107 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1108 done: 1109 if (cwdi->cwdi_rdir != NULL) { 1110 size_t len; 1111 char *bp; 1112 char c; 1113 char *path = PNBUF_GET(); 1114 1115 bp = path + MAXPATHLEN; 1116 *--bp = '\0'; 1117 rw_enter(&cwdi->cwdi_lock, RW_READER); 1118 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1119 MAXPATHLEN / 2, 0, l); 1120 rw_exit(&cwdi->cwdi_lock); 1121 if (error) { 1122 PNBUF_PUT(path); 1123 return error; 1124 } 1125 len = strlen(bp); 1126 if (len != 1) { 1127 /* 1128 * for mount points that are below our root, we can see 1129 * them, so we fix up the pathname and return them. The 1130 * rest we cannot see, so we don't allow viewing the 1131 * data. 1132 */ 1133 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1134 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1135 (void)strlcpy(sp->f_mntonname, 1136 c == '\0' ? "/" : &sp->f_mntonname[len], 1137 sizeof(sp->f_mntonname)); 1138 } else { 1139 if (root) 1140 (void)strlcpy(sp->f_mntonname, "/", 1141 sizeof(sp->f_mntonname)); 1142 else 1143 error = EPERM; 1144 } 1145 } 1146 PNBUF_PUT(path); 1147 } 1148 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1149 return error; 1150 } 1151 1152 /* 1153 * Get filesystem statistics by path. 1154 */ 1155 int 1156 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1157 { 1158 struct mount *mp; 1159 int error; 1160 struct vnode *vp; 1161 1162 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1163 if (error != 0) 1164 return error; 1165 mp = vp->v_mount; 1166 error = dostatvfs(mp, sb, l, flags, 1); 1167 vrele(vp); 1168 return error; 1169 } 1170 1171 /* ARGSUSED */ 1172 int 1173 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1174 { 1175 /* { 1176 syscallarg(const char *) path; 1177 syscallarg(struct statvfs *) buf; 1178 syscallarg(int) flags; 1179 } */ 1180 struct statvfs *sb; 1181 int error; 1182 1183 sb = STATVFSBUF_GET(); 1184 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1185 if (error == 0) 1186 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1187 STATVFSBUF_PUT(sb); 1188 return error; 1189 } 1190 1191 /* 1192 * Get filesystem statistics by fd. 1193 */ 1194 int 1195 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1196 { 1197 file_t *fp; 1198 struct mount *mp; 1199 int error; 1200 1201 /* fd_getvnode() will use the descriptor for us */ 1202 if ((error = fd_getvnode(fd, &fp)) != 0) 1203 return (error); 1204 mp = fp->f_vnode->v_mount; 1205 error = dostatvfs(mp, sb, curlwp, flags, 1); 1206 fd_putfile(fd); 1207 return error; 1208 } 1209 1210 /* ARGSUSED */ 1211 int 1212 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1213 { 1214 /* { 1215 syscallarg(int) fd; 1216 syscallarg(struct statvfs *) buf; 1217 syscallarg(int) flags; 1218 } */ 1219 struct statvfs *sb; 1220 int error; 1221 1222 sb = STATVFSBUF_GET(); 1223 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1224 if (error == 0) 1225 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1226 STATVFSBUF_PUT(sb); 1227 return error; 1228 } 1229 1230 1231 /* 1232 * Get statistics on all filesystems. 1233 */ 1234 int 1235 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1236 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1237 register_t *retval) 1238 { 1239 int root = 0; 1240 struct proc *p = l->l_proc; 1241 struct mount *mp, *nmp; 1242 struct statvfs *sb; 1243 size_t count, maxcount; 1244 int error = 0; 1245 1246 sb = STATVFSBUF_GET(); 1247 maxcount = bufsize / entry_sz; 1248 mutex_enter(&mountlist_lock); 1249 count = 0; 1250 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 1251 if (vfs_busy(mp, &nmp)) { 1252 continue; 1253 } 1254 if (sfsp && count < maxcount) { 1255 error = dostatvfs(mp, sb, l, flags, 0); 1256 if (error) { 1257 vfs_unbusy(mp, false, &nmp); 1258 error = 0; 1259 continue; 1260 } 1261 error = copyfn(sb, sfsp, entry_sz); 1262 if (error) { 1263 vfs_unbusy(mp, false, NULL); 1264 goto out; 1265 } 1266 sfsp = (char *)sfsp + entry_sz; 1267 root |= strcmp(sb->f_mntonname, "/") == 0; 1268 } 1269 count++; 1270 vfs_unbusy(mp, false, &nmp); 1271 } 1272 mutex_exit(&mountlist_lock); 1273 1274 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1275 /* 1276 * fake a root entry 1277 */ 1278 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1279 sb, l, flags, 1); 1280 if (error != 0) 1281 goto out; 1282 if (sfsp) { 1283 error = copyfn(sb, sfsp, entry_sz); 1284 if (error != 0) 1285 goto out; 1286 } 1287 count++; 1288 } 1289 if (sfsp && count > maxcount) 1290 *retval = maxcount; 1291 else 1292 *retval = count; 1293 out: 1294 STATVFSBUF_PUT(sb); 1295 return error; 1296 } 1297 1298 int 1299 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1300 { 1301 /* { 1302 syscallarg(struct statvfs *) buf; 1303 syscallarg(size_t) bufsize; 1304 syscallarg(int) flags; 1305 } */ 1306 1307 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1308 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1309 } 1310 1311 /* 1312 * Change current working directory to a given file descriptor. 1313 */ 1314 /* ARGSUSED */ 1315 int 1316 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1317 { 1318 /* { 1319 syscallarg(int) fd; 1320 } */ 1321 struct proc *p = l->l_proc; 1322 struct cwdinfo *cwdi; 1323 struct vnode *vp, *tdp; 1324 struct mount *mp; 1325 file_t *fp; 1326 int error, fd; 1327 1328 /* fd_getvnode() will use the descriptor for us */ 1329 fd = SCARG(uap, fd); 1330 if ((error = fd_getvnode(fd, &fp)) != 0) 1331 return (error); 1332 vp = fp->f_vnode; 1333 1334 vref(vp); 1335 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1336 if (vp->v_type != VDIR) 1337 error = ENOTDIR; 1338 else 1339 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1340 if (error) { 1341 vput(vp); 1342 goto out; 1343 } 1344 while ((mp = vp->v_mountedhere) != NULL) { 1345 error = vfs_busy(mp, NULL); 1346 vput(vp); 1347 if (error != 0) 1348 goto out; 1349 error = VFS_ROOT(mp, &tdp); 1350 vfs_unbusy(mp, false, NULL); 1351 if (error) 1352 goto out; 1353 vp = tdp; 1354 } 1355 VOP_UNLOCK(vp); 1356 1357 /* 1358 * Disallow changing to a directory not under the process's 1359 * current root directory (if there is one). 1360 */ 1361 cwdi = p->p_cwdi; 1362 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1363 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1364 vrele(vp); 1365 error = EPERM; /* operation not permitted */ 1366 } else { 1367 vrele(cwdi->cwdi_cdir); 1368 cwdi->cwdi_cdir = vp; 1369 } 1370 rw_exit(&cwdi->cwdi_lock); 1371 1372 out: 1373 fd_putfile(fd); 1374 return (error); 1375 } 1376 1377 /* 1378 * Change this process's notion of the root directory to a given file 1379 * descriptor. 1380 */ 1381 int 1382 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1383 { 1384 struct proc *p = l->l_proc; 1385 struct vnode *vp; 1386 file_t *fp; 1387 int error, fd = SCARG(uap, fd); 1388 1389 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1390 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1391 return error; 1392 /* fd_getvnode() will use the descriptor for us */ 1393 if ((error = fd_getvnode(fd, &fp)) != 0) 1394 return error; 1395 vp = fp->f_vnode; 1396 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1397 if (vp->v_type != VDIR) 1398 error = ENOTDIR; 1399 else 1400 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1401 VOP_UNLOCK(vp); 1402 if (error) 1403 goto out; 1404 vref(vp); 1405 1406 change_root(p->p_cwdi, vp, l); 1407 1408 out: 1409 fd_putfile(fd); 1410 return (error); 1411 } 1412 1413 /* 1414 * Change current working directory (``.''). 1415 */ 1416 /* ARGSUSED */ 1417 int 1418 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1419 { 1420 /* { 1421 syscallarg(const char *) path; 1422 } */ 1423 struct proc *p = l->l_proc; 1424 struct cwdinfo *cwdi; 1425 int error; 1426 struct vnode *vp; 1427 1428 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1429 &vp, l)) != 0) 1430 return (error); 1431 cwdi = p->p_cwdi; 1432 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1433 vrele(cwdi->cwdi_cdir); 1434 cwdi->cwdi_cdir = vp; 1435 rw_exit(&cwdi->cwdi_lock); 1436 return (0); 1437 } 1438 1439 /* 1440 * Change notion of root (``/'') directory. 1441 */ 1442 /* ARGSUSED */ 1443 int 1444 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1445 { 1446 /* { 1447 syscallarg(const char *) path; 1448 } */ 1449 struct proc *p = l->l_proc; 1450 int error; 1451 struct vnode *vp; 1452 1453 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1454 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1455 return (error); 1456 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1457 &vp, l)) != 0) 1458 return (error); 1459 1460 change_root(p->p_cwdi, vp, l); 1461 1462 return (0); 1463 } 1464 1465 /* 1466 * Common routine for chroot and fchroot. 1467 * NB: callers need to properly authorize the change root operation. 1468 */ 1469 void 1470 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1471 { 1472 struct proc *p = l->l_proc; 1473 kauth_cred_t ncred; 1474 1475 ncred = kauth_cred_alloc(); 1476 1477 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1478 if (cwdi->cwdi_rdir != NULL) 1479 vrele(cwdi->cwdi_rdir); 1480 cwdi->cwdi_rdir = vp; 1481 1482 /* 1483 * Prevent escaping from chroot by putting the root under 1484 * the working directory. Silently chdir to / if we aren't 1485 * already there. 1486 */ 1487 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1488 /* 1489 * XXX would be more failsafe to change directory to a 1490 * deadfs node here instead 1491 */ 1492 vrele(cwdi->cwdi_cdir); 1493 vref(vp); 1494 cwdi->cwdi_cdir = vp; 1495 } 1496 rw_exit(&cwdi->cwdi_lock); 1497 1498 /* Get a write lock on the process credential. */ 1499 proc_crmod_enter(); 1500 1501 kauth_cred_clone(p->p_cred, ncred); 1502 kauth_proc_chroot(ncred, p->p_cwdi); 1503 1504 /* Broadcast our credentials to the process and other LWPs. */ 1505 proc_crmod_leave(ncred, p->p_cred, true); 1506 } 1507 1508 /* 1509 * Common routine for chroot and chdir. 1510 * XXX "where" should be enum uio_seg 1511 */ 1512 int 1513 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1514 { 1515 struct pathbuf *pb; 1516 struct nameidata nd; 1517 int error; 1518 1519 error = pathbuf_maybe_copyin(path, where, &pb); 1520 if (error) { 1521 return error; 1522 } 1523 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1524 if ((error = namei(&nd)) != 0) { 1525 pathbuf_destroy(pb); 1526 return error; 1527 } 1528 *vpp = nd.ni_vp; 1529 pathbuf_destroy(pb); 1530 1531 if ((*vpp)->v_type != VDIR) 1532 error = ENOTDIR; 1533 else 1534 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1535 1536 if (error) 1537 vput(*vpp); 1538 else 1539 VOP_UNLOCK(*vpp); 1540 return (error); 1541 } 1542 1543 /* 1544 * Internals of sys_open - path has already been converted into a pathbuf 1545 * (so we can easily reuse this function from other parts of the kernel, 1546 * like posix_spawn post-processing). 1547 */ 1548 int 1549 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1550 int open_mode, int *fd) 1551 { 1552 struct proc *p = l->l_proc; 1553 struct cwdinfo *cwdi = p->p_cwdi; 1554 file_t *fp; 1555 struct vnode *vp; 1556 int flags, cmode; 1557 int indx, error; 1558 struct nameidata nd; 1559 1560 if (open_flags & O_SEARCH) { 1561 open_flags &= ~(int)O_SEARCH; 1562 } 1563 1564 flags = FFLAGS(open_flags); 1565 if ((flags & (FREAD | FWRITE)) == 0) 1566 return EINVAL; 1567 1568 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1569 return error; 1570 } 1571 1572 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1573 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1574 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1575 if (dvp != NULL) 1576 NDAT(&nd, dvp); 1577 1578 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1579 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1580 fd_abort(p, fp, indx); 1581 if ((error == EDUPFD || error == EMOVEFD) && 1582 l->l_dupfd >= 0 && /* XXX from fdopen */ 1583 (error = 1584 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1585 *fd = indx; 1586 return 0; 1587 } 1588 if (error == ERESTART) 1589 error = EINTR; 1590 return error; 1591 } 1592 1593 l->l_dupfd = 0; 1594 vp = nd.ni_vp; 1595 1596 if ((error = open_setfp(l, fp, vp, indx, flags))) 1597 return error; 1598 1599 VOP_UNLOCK(vp); 1600 *fd = indx; 1601 fd_affix(p, fp, indx); 1602 return 0; 1603 } 1604 1605 int 1606 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1607 { 1608 struct pathbuf *pb; 1609 int error, oflags; 1610 1611 oflags = FFLAGS(open_flags); 1612 if ((oflags & (FREAD | FWRITE)) == 0) 1613 return EINVAL; 1614 1615 pb = pathbuf_create(path); 1616 if (pb == NULL) 1617 return ENOMEM; 1618 1619 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1620 pathbuf_destroy(pb); 1621 1622 return error; 1623 } 1624 1625 /* 1626 * Check permissions, allocate an open file structure, 1627 * and call the device open routine if any. 1628 */ 1629 static int 1630 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1631 int mode, int *fd) 1632 { 1633 file_t *dfp = NULL; 1634 struct vnode *dvp = NULL; 1635 struct pathbuf *pb; 1636 int error; 1637 1638 #ifdef COMPAT_10 /* XXX: and perhaps later */ 1639 if (path == NULL) { 1640 pb = pathbuf_create("."); 1641 if (pb == NULL) 1642 return ENOMEM; 1643 } else 1644 #endif 1645 { 1646 error = pathbuf_copyin(path, &pb); 1647 if (error) 1648 return error; 1649 } 1650 1651 if (fdat != AT_FDCWD) { 1652 /* fd_getvnode() will use the descriptor for us */ 1653 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1654 goto out; 1655 1656 dvp = dfp->f_vnode; 1657 } 1658 1659 error = do_open(l, dvp, pb, flags, mode, fd); 1660 1661 if (dfp != NULL) 1662 fd_putfile(fdat); 1663 out: 1664 pathbuf_destroy(pb); 1665 return error; 1666 } 1667 1668 int 1669 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1670 { 1671 /* { 1672 syscallarg(const char *) path; 1673 syscallarg(int) flags; 1674 syscallarg(int) mode; 1675 } */ 1676 int error; 1677 int fd; 1678 1679 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1680 SCARG(uap, flags), SCARG(uap, mode), &fd); 1681 1682 if (error == 0) 1683 *retval = fd; 1684 1685 return error; 1686 } 1687 1688 int 1689 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1690 { 1691 /* { 1692 syscallarg(int) fd; 1693 syscallarg(const char *) path; 1694 syscallarg(int) oflags; 1695 syscallarg(int) mode; 1696 } */ 1697 int error; 1698 int fd; 1699 1700 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1701 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1702 1703 if (error == 0) 1704 *retval = fd; 1705 1706 return error; 1707 } 1708 1709 static void 1710 vfs__fhfree(fhandle_t *fhp) 1711 { 1712 size_t fhsize; 1713 1714 fhsize = FHANDLE_SIZE(fhp); 1715 kmem_free(fhp, fhsize); 1716 } 1717 1718 /* 1719 * vfs_composefh: compose a filehandle. 1720 */ 1721 1722 int 1723 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1724 { 1725 struct mount *mp; 1726 struct fid *fidp; 1727 int error; 1728 size_t needfhsize; 1729 size_t fidsize; 1730 1731 mp = vp->v_mount; 1732 fidp = NULL; 1733 if (*fh_size < FHANDLE_SIZE_MIN) { 1734 fidsize = 0; 1735 } else { 1736 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1737 if (fhp != NULL) { 1738 memset(fhp, 0, *fh_size); 1739 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1740 fidp = &fhp->fh_fid; 1741 } 1742 } 1743 error = VFS_VPTOFH(vp, fidp, &fidsize); 1744 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1745 if (error == 0 && *fh_size < needfhsize) { 1746 error = E2BIG; 1747 } 1748 *fh_size = needfhsize; 1749 return error; 1750 } 1751 1752 int 1753 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1754 { 1755 struct mount *mp; 1756 fhandle_t *fhp; 1757 size_t fhsize; 1758 size_t fidsize; 1759 int error; 1760 1761 mp = vp->v_mount; 1762 fidsize = 0; 1763 error = VFS_VPTOFH(vp, NULL, &fidsize); 1764 KASSERT(error != 0); 1765 if (error != E2BIG) { 1766 goto out; 1767 } 1768 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1769 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1770 if (fhp == NULL) { 1771 error = ENOMEM; 1772 goto out; 1773 } 1774 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1775 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1776 if (error == 0) { 1777 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1778 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1779 *fhpp = fhp; 1780 } else { 1781 kmem_free(fhp, fhsize); 1782 } 1783 out: 1784 return error; 1785 } 1786 1787 void 1788 vfs_composefh_free(fhandle_t *fhp) 1789 { 1790 1791 vfs__fhfree(fhp); 1792 } 1793 1794 /* 1795 * vfs_fhtovp: lookup a vnode by a filehandle. 1796 */ 1797 1798 int 1799 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1800 { 1801 struct mount *mp; 1802 int error; 1803 1804 *vpp = NULL; 1805 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1806 if (mp == NULL) { 1807 error = ESTALE; 1808 goto out; 1809 } 1810 if (mp->mnt_op->vfs_fhtovp == NULL) { 1811 error = EOPNOTSUPP; 1812 goto out; 1813 } 1814 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1815 out: 1816 return error; 1817 } 1818 1819 /* 1820 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1821 * the needed size. 1822 */ 1823 1824 int 1825 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1826 { 1827 fhandle_t *fhp; 1828 int error; 1829 1830 if (fhsize > FHANDLE_SIZE_MAX) { 1831 return EINVAL; 1832 } 1833 if (fhsize < FHANDLE_SIZE_MIN) { 1834 return EINVAL; 1835 } 1836 again: 1837 fhp = kmem_alloc(fhsize, KM_SLEEP); 1838 if (fhp == NULL) { 1839 return ENOMEM; 1840 } 1841 error = copyin(ufhp, fhp, fhsize); 1842 if (error == 0) { 1843 /* XXX this check shouldn't be here */ 1844 if (FHANDLE_SIZE(fhp) == fhsize) { 1845 *fhpp = fhp; 1846 return 0; 1847 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1848 /* 1849 * a kludge for nfsv2 padded handles. 1850 */ 1851 size_t sz; 1852 1853 sz = FHANDLE_SIZE(fhp); 1854 kmem_free(fhp, fhsize); 1855 fhsize = sz; 1856 goto again; 1857 } else { 1858 /* 1859 * userland told us wrong size. 1860 */ 1861 error = EINVAL; 1862 } 1863 } 1864 kmem_free(fhp, fhsize); 1865 return error; 1866 } 1867 1868 void 1869 vfs_copyinfh_free(fhandle_t *fhp) 1870 { 1871 1872 vfs__fhfree(fhp); 1873 } 1874 1875 /* 1876 * Get file handle system call 1877 */ 1878 int 1879 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1880 { 1881 /* { 1882 syscallarg(char *) fname; 1883 syscallarg(fhandle_t *) fhp; 1884 syscallarg(size_t *) fh_size; 1885 } */ 1886 struct vnode *vp; 1887 fhandle_t *fh; 1888 int error; 1889 struct pathbuf *pb; 1890 struct nameidata nd; 1891 size_t sz; 1892 size_t usz; 1893 1894 /* 1895 * Must be super user 1896 */ 1897 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1898 0, NULL, NULL, NULL); 1899 if (error) 1900 return (error); 1901 1902 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1903 if (error) { 1904 return error; 1905 } 1906 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1907 error = namei(&nd); 1908 if (error) { 1909 pathbuf_destroy(pb); 1910 return error; 1911 } 1912 vp = nd.ni_vp; 1913 pathbuf_destroy(pb); 1914 1915 error = vfs_composefh_alloc(vp, &fh); 1916 vput(vp); 1917 if (error != 0) { 1918 return error; 1919 } 1920 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1921 if (error != 0) { 1922 goto out; 1923 } 1924 sz = FHANDLE_SIZE(fh); 1925 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1926 if (error != 0) { 1927 goto out; 1928 } 1929 if (usz >= sz) { 1930 error = copyout(fh, SCARG(uap, fhp), sz); 1931 } else { 1932 error = E2BIG; 1933 } 1934 out: 1935 vfs_composefh_free(fh); 1936 return (error); 1937 } 1938 1939 /* 1940 * Open a file given a file handle. 1941 * 1942 * Check permissions, allocate an open file structure, 1943 * and call the device open routine if any. 1944 */ 1945 1946 int 1947 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1948 register_t *retval) 1949 { 1950 file_t *fp; 1951 struct vnode *vp = NULL; 1952 kauth_cred_t cred = l->l_cred; 1953 file_t *nfp; 1954 int indx, error = 0; 1955 struct vattr va; 1956 fhandle_t *fh; 1957 int flags; 1958 proc_t *p; 1959 1960 p = curproc; 1961 1962 /* 1963 * Must be super user 1964 */ 1965 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1966 0, NULL, NULL, NULL))) 1967 return (error); 1968 1969 if (oflags & O_SEARCH) { 1970 oflags &= ~(int)O_SEARCH; 1971 } 1972 1973 flags = FFLAGS(oflags); 1974 if ((flags & (FREAD | FWRITE)) == 0) 1975 return (EINVAL); 1976 if ((flags & O_CREAT)) 1977 return (EINVAL); 1978 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1979 return (error); 1980 fp = nfp; 1981 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1982 if (error != 0) { 1983 goto bad; 1984 } 1985 error = vfs_fhtovp(fh, &vp); 1986 vfs_copyinfh_free(fh); 1987 if (error != 0) { 1988 goto bad; 1989 } 1990 1991 /* Now do an effective vn_open */ 1992 1993 if (vp->v_type == VSOCK) { 1994 error = EOPNOTSUPP; 1995 goto bad; 1996 } 1997 error = vn_openchk(vp, cred, flags); 1998 if (error != 0) 1999 goto bad; 2000 if (flags & O_TRUNC) { 2001 VOP_UNLOCK(vp); /* XXX */ 2002 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2003 vattr_null(&va); 2004 va.va_size = 0; 2005 error = VOP_SETATTR(vp, &va, cred); 2006 if (error) 2007 goto bad; 2008 } 2009 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2010 goto bad; 2011 if (flags & FWRITE) { 2012 mutex_enter(vp->v_interlock); 2013 vp->v_writecount++; 2014 mutex_exit(vp->v_interlock); 2015 } 2016 2017 /* done with modified vn_open, now finish what sys_open does. */ 2018 if ((error = open_setfp(l, fp, vp, indx, flags))) 2019 return error; 2020 2021 VOP_UNLOCK(vp); 2022 *retval = indx; 2023 fd_affix(p, fp, indx); 2024 return (0); 2025 2026 bad: 2027 fd_abort(p, fp, indx); 2028 if (vp != NULL) 2029 vput(vp); 2030 return (error); 2031 } 2032 2033 int 2034 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2035 { 2036 /* { 2037 syscallarg(const void *) fhp; 2038 syscallarg(size_t) fh_size; 2039 syscallarg(int) flags; 2040 } */ 2041 2042 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2043 SCARG(uap, flags), retval); 2044 } 2045 2046 int 2047 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2048 { 2049 int error; 2050 fhandle_t *fh; 2051 struct vnode *vp; 2052 2053 /* 2054 * Must be super user 2055 */ 2056 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2057 0, NULL, NULL, NULL))) 2058 return (error); 2059 2060 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2061 if (error != 0) 2062 return error; 2063 2064 error = vfs_fhtovp(fh, &vp); 2065 vfs_copyinfh_free(fh); 2066 if (error != 0) 2067 return error; 2068 2069 error = vn_stat(vp, sb); 2070 vput(vp); 2071 return error; 2072 } 2073 2074 2075 /* ARGSUSED */ 2076 int 2077 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2078 { 2079 /* { 2080 syscallarg(const void *) fhp; 2081 syscallarg(size_t) fh_size; 2082 syscallarg(struct stat *) sb; 2083 } */ 2084 struct stat sb; 2085 int error; 2086 2087 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2088 if (error) 2089 return error; 2090 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2091 } 2092 2093 int 2094 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2095 int flags) 2096 { 2097 fhandle_t *fh; 2098 struct mount *mp; 2099 struct vnode *vp; 2100 int error; 2101 2102 /* 2103 * Must be super user 2104 */ 2105 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2106 0, NULL, NULL, NULL))) 2107 return error; 2108 2109 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2110 if (error != 0) 2111 return error; 2112 2113 error = vfs_fhtovp(fh, &vp); 2114 vfs_copyinfh_free(fh); 2115 if (error != 0) 2116 return error; 2117 2118 mp = vp->v_mount; 2119 error = dostatvfs(mp, sb, l, flags, 1); 2120 vput(vp); 2121 return error; 2122 } 2123 2124 /* ARGSUSED */ 2125 int 2126 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2127 { 2128 /* { 2129 syscallarg(const void *) fhp; 2130 syscallarg(size_t) fh_size; 2131 syscallarg(struct statvfs *) buf; 2132 syscallarg(int) flags; 2133 } */ 2134 struct statvfs *sb = STATVFSBUF_GET(); 2135 int error; 2136 2137 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2138 SCARG(uap, flags)); 2139 if (error == 0) 2140 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2141 STATVFSBUF_PUT(sb); 2142 return error; 2143 } 2144 2145 /* 2146 * Create a special file. 2147 */ 2148 /* ARGSUSED */ 2149 int 2150 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2151 register_t *retval) 2152 { 2153 /* { 2154 syscallarg(const char *) path; 2155 syscallarg(mode_t) mode; 2156 syscallarg(dev_t) dev; 2157 } */ 2158 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2159 SCARG(uap, dev), retval, UIO_USERSPACE); 2160 } 2161 2162 int 2163 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2164 register_t *retval) 2165 { 2166 /* { 2167 syscallarg(int) fd; 2168 syscallarg(const char *) path; 2169 syscallarg(mode_t) mode; 2170 syscallarg(int) pad; 2171 syscallarg(dev_t) dev; 2172 } */ 2173 2174 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2175 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2176 } 2177 2178 int 2179 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2180 register_t *retval, enum uio_seg seg) 2181 { 2182 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2183 } 2184 2185 int 2186 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2187 dev_t dev, register_t *retval, enum uio_seg seg) 2188 { 2189 struct proc *p = l->l_proc; 2190 struct vnode *vp; 2191 struct vattr vattr; 2192 int error, optype; 2193 struct pathbuf *pb; 2194 struct nameidata nd; 2195 const char *pathstring; 2196 2197 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2198 0, NULL, NULL, NULL)) != 0) 2199 return (error); 2200 2201 optype = VOP_MKNOD_DESCOFFSET; 2202 2203 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2204 if (error) { 2205 return error; 2206 } 2207 pathstring = pathbuf_stringcopy_get(pb); 2208 if (pathstring == NULL) { 2209 pathbuf_destroy(pb); 2210 return ENOMEM; 2211 } 2212 2213 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2214 2215 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2216 goto out; 2217 vp = nd.ni_vp; 2218 2219 if (vp != NULL) 2220 error = EEXIST; 2221 else { 2222 vattr_null(&vattr); 2223 /* We will read cwdi->cwdi_cmask unlocked. */ 2224 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2225 vattr.va_rdev = dev; 2226 2227 switch (mode & S_IFMT) { 2228 case S_IFMT: /* used by badsect to flag bad sectors */ 2229 vattr.va_type = VBAD; 2230 break; 2231 case S_IFCHR: 2232 vattr.va_type = VCHR; 2233 break; 2234 case S_IFBLK: 2235 vattr.va_type = VBLK; 2236 break; 2237 case S_IFWHT: 2238 optype = VOP_WHITEOUT_DESCOFFSET; 2239 break; 2240 case S_IFREG: 2241 #if NVERIEXEC > 0 2242 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2243 O_CREAT); 2244 #endif /* NVERIEXEC > 0 */ 2245 vattr.va_type = VREG; 2246 vattr.va_rdev = VNOVAL; 2247 optype = VOP_CREATE_DESCOFFSET; 2248 break; 2249 default: 2250 error = EINVAL; 2251 break; 2252 } 2253 } 2254 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2255 && vattr.va_rdev == VNOVAL) 2256 error = EINVAL; 2257 if (!error) { 2258 switch (optype) { 2259 case VOP_WHITEOUT_DESCOFFSET: 2260 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2261 if (error) 2262 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2263 vput(nd.ni_dvp); 2264 break; 2265 2266 case VOP_MKNOD_DESCOFFSET: 2267 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2268 &nd.ni_cnd, &vattr); 2269 if (error == 0) 2270 vrele(nd.ni_vp); 2271 vput(nd.ni_dvp); 2272 break; 2273 2274 case VOP_CREATE_DESCOFFSET: 2275 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2276 &nd.ni_cnd, &vattr); 2277 if (error == 0) 2278 vrele(nd.ni_vp); 2279 vput(nd.ni_dvp); 2280 break; 2281 } 2282 } else { 2283 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2284 if (nd.ni_dvp == vp) 2285 vrele(nd.ni_dvp); 2286 else 2287 vput(nd.ni_dvp); 2288 if (vp) 2289 vrele(vp); 2290 } 2291 out: 2292 pathbuf_stringcopy_put(pb, pathstring); 2293 pathbuf_destroy(pb); 2294 return (error); 2295 } 2296 2297 /* 2298 * Create a named pipe. 2299 */ 2300 /* ARGSUSED */ 2301 int 2302 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2303 { 2304 /* { 2305 syscallarg(const char *) path; 2306 syscallarg(int) mode; 2307 } */ 2308 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2309 } 2310 2311 int 2312 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2313 register_t *retval) 2314 { 2315 /* { 2316 syscallarg(int) fd; 2317 syscallarg(const char *) path; 2318 syscallarg(int) mode; 2319 } */ 2320 2321 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2322 SCARG(uap, mode)); 2323 } 2324 2325 static int 2326 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2327 { 2328 struct proc *p = l->l_proc; 2329 struct vattr vattr; 2330 int error; 2331 struct pathbuf *pb; 2332 struct nameidata nd; 2333 2334 error = pathbuf_copyin(path, &pb); 2335 if (error) { 2336 return error; 2337 } 2338 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2339 2340 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2341 pathbuf_destroy(pb); 2342 return error; 2343 } 2344 if (nd.ni_vp != NULL) { 2345 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2346 if (nd.ni_dvp == nd.ni_vp) 2347 vrele(nd.ni_dvp); 2348 else 2349 vput(nd.ni_dvp); 2350 vrele(nd.ni_vp); 2351 pathbuf_destroy(pb); 2352 return (EEXIST); 2353 } 2354 vattr_null(&vattr); 2355 vattr.va_type = VFIFO; 2356 /* We will read cwdi->cwdi_cmask unlocked. */ 2357 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2358 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2359 if (error == 0) 2360 vrele(nd.ni_vp); 2361 vput(nd.ni_dvp); 2362 pathbuf_destroy(pb); 2363 return (error); 2364 } 2365 2366 /* 2367 * Make a hard file link. 2368 */ 2369 /* ARGSUSED */ 2370 int 2371 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2372 const char *link, int follow, register_t *retval) 2373 { 2374 struct vnode *vp; 2375 struct pathbuf *linkpb; 2376 struct nameidata nd; 2377 namei_simple_flags_t ns_flags; 2378 int error; 2379 2380 if (follow & AT_SYMLINK_FOLLOW) 2381 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2382 else 2383 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2384 2385 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2386 if (error != 0) 2387 return (error); 2388 error = pathbuf_copyin(link, &linkpb); 2389 if (error) { 2390 goto out1; 2391 } 2392 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2393 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2394 goto out2; 2395 if (nd.ni_vp) { 2396 error = EEXIST; 2397 goto abortop; 2398 } 2399 /* Prevent hard links on directories. */ 2400 if (vp->v_type == VDIR) { 2401 error = EPERM; 2402 goto abortop; 2403 } 2404 /* Prevent cross-mount operation. */ 2405 if (nd.ni_dvp->v_mount != vp->v_mount) { 2406 error = EXDEV; 2407 goto abortop; 2408 } 2409 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2410 VOP_UNLOCK(nd.ni_dvp); 2411 vrele(nd.ni_dvp); 2412 out2: 2413 pathbuf_destroy(linkpb); 2414 out1: 2415 vrele(vp); 2416 return (error); 2417 abortop: 2418 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2419 if (nd.ni_dvp == nd.ni_vp) 2420 vrele(nd.ni_dvp); 2421 else 2422 vput(nd.ni_dvp); 2423 if (nd.ni_vp != NULL) 2424 vrele(nd.ni_vp); 2425 goto out2; 2426 } 2427 2428 int 2429 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2430 { 2431 /* { 2432 syscallarg(const char *) path; 2433 syscallarg(const char *) link; 2434 } */ 2435 const char *path = SCARG(uap, path); 2436 const char *link = SCARG(uap, link); 2437 2438 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2439 AT_SYMLINK_FOLLOW, retval); 2440 } 2441 2442 int 2443 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2444 register_t *retval) 2445 { 2446 /* { 2447 syscallarg(int) fd1; 2448 syscallarg(const char *) name1; 2449 syscallarg(int) fd2; 2450 syscallarg(const char *) name2; 2451 syscallarg(int) flags; 2452 } */ 2453 int fd1 = SCARG(uap, fd1); 2454 const char *name1 = SCARG(uap, name1); 2455 int fd2 = SCARG(uap, fd2); 2456 const char *name2 = SCARG(uap, name2); 2457 int follow; 2458 2459 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2460 2461 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2462 } 2463 2464 2465 int 2466 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2467 { 2468 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2469 } 2470 2471 static int 2472 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2473 const char *link, enum uio_seg seg) 2474 { 2475 struct proc *p = curproc; 2476 struct vattr vattr; 2477 char *path; 2478 int error; 2479 struct pathbuf *linkpb; 2480 struct nameidata nd; 2481 2482 KASSERT(l != NULL || fdat == AT_FDCWD); 2483 2484 path = PNBUF_GET(); 2485 if (seg == UIO_USERSPACE) { 2486 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2487 goto out1; 2488 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2489 goto out1; 2490 } else { 2491 KASSERT(strlen(patharg) < MAXPATHLEN); 2492 strcpy(path, patharg); 2493 linkpb = pathbuf_create(link); 2494 if (linkpb == NULL) { 2495 error = ENOMEM; 2496 goto out1; 2497 } 2498 } 2499 ktrkuser("symlink-target", path, strlen(path)); 2500 2501 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2502 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2503 goto out2; 2504 if (nd.ni_vp) { 2505 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2506 if (nd.ni_dvp == nd.ni_vp) 2507 vrele(nd.ni_dvp); 2508 else 2509 vput(nd.ni_dvp); 2510 vrele(nd.ni_vp); 2511 error = EEXIST; 2512 goto out2; 2513 } 2514 vattr_null(&vattr); 2515 vattr.va_type = VLNK; 2516 /* We will read cwdi->cwdi_cmask unlocked. */ 2517 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2518 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2519 if (error == 0) 2520 vrele(nd.ni_vp); 2521 vput(nd.ni_dvp); 2522 out2: 2523 pathbuf_destroy(linkpb); 2524 out1: 2525 PNBUF_PUT(path); 2526 return (error); 2527 } 2528 2529 /* 2530 * Make a symbolic link. 2531 */ 2532 /* ARGSUSED */ 2533 int 2534 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2535 { 2536 /* { 2537 syscallarg(const char *) path; 2538 syscallarg(const char *) link; 2539 } */ 2540 2541 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2542 UIO_USERSPACE); 2543 } 2544 2545 int 2546 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2547 register_t *retval) 2548 { 2549 /* { 2550 syscallarg(const char *) path1; 2551 syscallarg(int) fd; 2552 syscallarg(const char *) path2; 2553 } */ 2554 2555 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2556 SCARG(uap, path2), UIO_USERSPACE); 2557 } 2558 2559 /* 2560 * Delete a whiteout from the filesystem. 2561 */ 2562 /* ARGSUSED */ 2563 int 2564 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2565 { 2566 /* { 2567 syscallarg(const char *) path; 2568 } */ 2569 int error; 2570 struct pathbuf *pb; 2571 struct nameidata nd; 2572 2573 error = pathbuf_copyin(SCARG(uap, path), &pb); 2574 if (error) { 2575 return error; 2576 } 2577 2578 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2579 error = namei(&nd); 2580 if (error) { 2581 pathbuf_destroy(pb); 2582 return (error); 2583 } 2584 2585 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2586 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2587 if (nd.ni_dvp == nd.ni_vp) 2588 vrele(nd.ni_dvp); 2589 else 2590 vput(nd.ni_dvp); 2591 if (nd.ni_vp) 2592 vrele(nd.ni_vp); 2593 pathbuf_destroy(pb); 2594 return (EEXIST); 2595 } 2596 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2597 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2598 vput(nd.ni_dvp); 2599 pathbuf_destroy(pb); 2600 return (error); 2601 } 2602 2603 /* 2604 * Delete a name from the filesystem. 2605 */ 2606 /* ARGSUSED */ 2607 int 2608 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2609 { 2610 /* { 2611 syscallarg(const char *) path; 2612 } */ 2613 2614 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2615 } 2616 2617 int 2618 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2619 register_t *retval) 2620 { 2621 /* { 2622 syscallarg(int) fd; 2623 syscallarg(const char *) path; 2624 syscallarg(int) flag; 2625 } */ 2626 2627 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2628 SCARG(uap, flag), UIO_USERSPACE); 2629 } 2630 2631 int 2632 do_sys_unlink(const char *arg, enum uio_seg seg) 2633 { 2634 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2635 } 2636 2637 static int 2638 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2639 enum uio_seg seg) 2640 { 2641 struct vnode *vp; 2642 int error; 2643 struct pathbuf *pb; 2644 struct nameidata nd; 2645 const char *pathstring; 2646 2647 KASSERT(l != NULL || fdat == AT_FDCWD); 2648 2649 error = pathbuf_maybe_copyin(arg, seg, &pb); 2650 if (error) { 2651 return error; 2652 } 2653 pathstring = pathbuf_stringcopy_get(pb); 2654 if (pathstring == NULL) { 2655 pathbuf_destroy(pb); 2656 return ENOMEM; 2657 } 2658 2659 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2660 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2661 goto out; 2662 vp = nd.ni_vp; 2663 2664 /* 2665 * The root of a mounted filesystem cannot be deleted. 2666 */ 2667 if ((vp->v_vflag & VV_ROOT) != 0) { 2668 error = EBUSY; 2669 goto abort; 2670 } 2671 2672 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2673 error = EBUSY; 2674 goto abort; 2675 } 2676 2677 /* 2678 * No rmdir "." please. 2679 */ 2680 if (nd.ni_dvp == vp) { 2681 error = EINVAL; 2682 goto abort; 2683 } 2684 2685 /* 2686 * AT_REMOVEDIR is required to remove a directory 2687 */ 2688 if (vp->v_type == VDIR) { 2689 if (!(flags & AT_REMOVEDIR)) { 2690 error = EPERM; 2691 goto abort; 2692 } else { 2693 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2694 goto out; 2695 } 2696 } 2697 2698 /* 2699 * Starting here we only deal with non directories. 2700 */ 2701 if (flags & AT_REMOVEDIR) { 2702 error = ENOTDIR; 2703 goto abort; 2704 } 2705 2706 #if NVERIEXEC > 0 2707 /* Handle remove requests for veriexec entries. */ 2708 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2709 goto abort; 2710 } 2711 #endif /* NVERIEXEC > 0 */ 2712 2713 #ifdef FILEASSOC 2714 (void)fileassoc_file_delete(vp); 2715 #endif /* FILEASSOC */ 2716 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2717 goto out; 2718 2719 abort: 2720 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2721 if (nd.ni_dvp == vp) 2722 vrele(nd.ni_dvp); 2723 else 2724 vput(nd.ni_dvp); 2725 vput(vp); 2726 2727 out: 2728 pathbuf_stringcopy_put(pb, pathstring); 2729 pathbuf_destroy(pb); 2730 return (error); 2731 } 2732 2733 /* 2734 * Reposition read/write file offset. 2735 */ 2736 int 2737 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2738 { 2739 /* { 2740 syscallarg(int) fd; 2741 syscallarg(int) pad; 2742 syscallarg(off_t) offset; 2743 syscallarg(int) whence; 2744 } */ 2745 kauth_cred_t cred = l->l_cred; 2746 file_t *fp; 2747 struct vnode *vp; 2748 struct vattr vattr; 2749 off_t newoff; 2750 int error, fd; 2751 2752 fd = SCARG(uap, fd); 2753 2754 if ((fp = fd_getfile(fd)) == NULL) 2755 return (EBADF); 2756 2757 vp = fp->f_vnode; 2758 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2759 error = ESPIPE; 2760 goto out; 2761 } 2762 2763 switch (SCARG(uap, whence)) { 2764 case SEEK_CUR: 2765 newoff = fp->f_offset + SCARG(uap, offset); 2766 break; 2767 case SEEK_END: 2768 vn_lock(vp, LK_SHARED | LK_RETRY); 2769 error = VOP_GETATTR(vp, &vattr, cred); 2770 VOP_UNLOCK(vp); 2771 if (error) { 2772 goto out; 2773 } 2774 newoff = SCARG(uap, offset) + vattr.va_size; 2775 break; 2776 case SEEK_SET: 2777 newoff = SCARG(uap, offset); 2778 break; 2779 default: 2780 error = EINVAL; 2781 goto out; 2782 } 2783 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2784 *(off_t *)retval = fp->f_offset = newoff; 2785 } 2786 out: 2787 fd_putfile(fd); 2788 return (error); 2789 } 2790 2791 /* 2792 * Positional read system call. 2793 */ 2794 int 2795 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2796 { 2797 /* { 2798 syscallarg(int) fd; 2799 syscallarg(void *) buf; 2800 syscallarg(size_t) nbyte; 2801 syscallarg(off_t) offset; 2802 } */ 2803 file_t *fp; 2804 struct vnode *vp; 2805 off_t offset; 2806 int error, fd = SCARG(uap, fd); 2807 2808 if ((fp = fd_getfile(fd)) == NULL) 2809 return (EBADF); 2810 2811 if ((fp->f_flag & FREAD) == 0) { 2812 fd_putfile(fd); 2813 return (EBADF); 2814 } 2815 2816 vp = fp->f_vnode; 2817 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2818 error = ESPIPE; 2819 goto out; 2820 } 2821 2822 offset = SCARG(uap, offset); 2823 2824 /* 2825 * XXX This works because no file systems actually 2826 * XXX take any action on the seek operation. 2827 */ 2828 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2829 goto out; 2830 2831 /* dofileread() will unuse the descriptor for us */ 2832 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2833 &offset, 0, retval)); 2834 2835 out: 2836 fd_putfile(fd); 2837 return (error); 2838 } 2839 2840 /* 2841 * Positional scatter read system call. 2842 */ 2843 int 2844 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2845 { 2846 /* { 2847 syscallarg(int) fd; 2848 syscallarg(const struct iovec *) iovp; 2849 syscallarg(int) iovcnt; 2850 syscallarg(off_t) offset; 2851 } */ 2852 off_t offset = SCARG(uap, offset); 2853 2854 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2855 SCARG(uap, iovcnt), &offset, 0, retval); 2856 } 2857 2858 /* 2859 * Positional write system call. 2860 */ 2861 int 2862 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2863 { 2864 /* { 2865 syscallarg(int) fd; 2866 syscallarg(const void *) buf; 2867 syscallarg(size_t) nbyte; 2868 syscallarg(off_t) offset; 2869 } */ 2870 file_t *fp; 2871 struct vnode *vp; 2872 off_t offset; 2873 int error, fd = SCARG(uap, fd); 2874 2875 if ((fp = fd_getfile(fd)) == NULL) 2876 return (EBADF); 2877 2878 if ((fp->f_flag & FWRITE) == 0) { 2879 fd_putfile(fd); 2880 return (EBADF); 2881 } 2882 2883 vp = fp->f_vnode; 2884 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2885 error = ESPIPE; 2886 goto out; 2887 } 2888 2889 offset = SCARG(uap, offset); 2890 2891 /* 2892 * XXX This works because no file systems actually 2893 * XXX take any action on the seek operation. 2894 */ 2895 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2896 goto out; 2897 2898 /* dofilewrite() will unuse the descriptor for us */ 2899 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2900 &offset, 0, retval)); 2901 2902 out: 2903 fd_putfile(fd); 2904 return (error); 2905 } 2906 2907 /* 2908 * Positional gather write system call. 2909 */ 2910 int 2911 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2912 { 2913 /* { 2914 syscallarg(int) fd; 2915 syscallarg(const struct iovec *) iovp; 2916 syscallarg(int) iovcnt; 2917 syscallarg(off_t) offset; 2918 } */ 2919 off_t offset = SCARG(uap, offset); 2920 2921 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2922 SCARG(uap, iovcnt), &offset, 0, retval); 2923 } 2924 2925 /* 2926 * Check access permissions. 2927 */ 2928 int 2929 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2930 { 2931 /* { 2932 syscallarg(const char *) path; 2933 syscallarg(int) flags; 2934 } */ 2935 2936 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2937 SCARG(uap, flags), 0); 2938 } 2939 2940 int 2941 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2942 int mode, int flags) 2943 { 2944 kauth_cred_t cred; 2945 struct vnode *vp; 2946 int error, nd_flag, vmode; 2947 struct pathbuf *pb; 2948 struct nameidata nd; 2949 2950 CTASSERT(F_OK == 0); 2951 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2952 /* nonsense mode */ 2953 return EINVAL; 2954 } 2955 2956 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2957 if (flags & AT_SYMLINK_NOFOLLOW) 2958 nd_flag &= ~FOLLOW; 2959 2960 error = pathbuf_copyin(path, &pb); 2961 if (error) 2962 return error; 2963 2964 NDINIT(&nd, LOOKUP, nd_flag, pb); 2965 2966 /* Override default credentials */ 2967 cred = kauth_cred_dup(l->l_cred); 2968 if (!(flags & AT_EACCESS)) { 2969 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2970 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2971 } 2972 nd.ni_cnd.cn_cred = cred; 2973 2974 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2975 pathbuf_destroy(pb); 2976 goto out; 2977 } 2978 vp = nd.ni_vp; 2979 pathbuf_destroy(pb); 2980 2981 /* Flags == 0 means only check for existence. */ 2982 if (mode) { 2983 vmode = 0; 2984 if (mode & R_OK) 2985 vmode |= VREAD; 2986 if (mode & W_OK) 2987 vmode |= VWRITE; 2988 if (mode & X_OK) 2989 vmode |= VEXEC; 2990 2991 error = VOP_ACCESS(vp, vmode, cred); 2992 if (!error && (vmode & VWRITE)) 2993 error = vn_writechk(vp); 2994 } 2995 vput(vp); 2996 out: 2997 kauth_cred_free(cred); 2998 return (error); 2999 } 3000 3001 int 3002 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3003 register_t *retval) 3004 { 3005 /* { 3006 syscallarg(int) fd; 3007 syscallarg(const char *) path; 3008 syscallarg(int) amode; 3009 syscallarg(int) flag; 3010 } */ 3011 3012 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3013 SCARG(uap, amode), SCARG(uap, flag)); 3014 } 3015 3016 /* 3017 * Common code for all sys_stat functions, including compat versions. 3018 */ 3019 int 3020 do_sys_stat(const char *userpath, unsigned int nd_flag, 3021 struct stat *sb) 3022 { 3023 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3024 } 3025 3026 int 3027 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3028 unsigned int nd_flag, struct stat *sb) 3029 { 3030 int error; 3031 struct pathbuf *pb; 3032 struct nameidata nd; 3033 3034 KASSERT(l != NULL || fdat == AT_FDCWD); 3035 3036 error = pathbuf_copyin(userpath, &pb); 3037 if (error) { 3038 return error; 3039 } 3040 3041 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3042 3043 error = fd_nameiat(l, fdat, &nd); 3044 if (error != 0) { 3045 pathbuf_destroy(pb); 3046 return error; 3047 } 3048 error = vn_stat(nd.ni_vp, sb); 3049 vput(nd.ni_vp); 3050 pathbuf_destroy(pb); 3051 return error; 3052 } 3053 3054 /* 3055 * Get file status; this version follows links. 3056 */ 3057 /* ARGSUSED */ 3058 int 3059 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3060 { 3061 /* { 3062 syscallarg(const char *) path; 3063 syscallarg(struct stat *) ub; 3064 } */ 3065 struct stat sb; 3066 int error; 3067 3068 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3069 if (error) 3070 return error; 3071 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3072 } 3073 3074 /* 3075 * Get file status; this version does not follow links. 3076 */ 3077 /* ARGSUSED */ 3078 int 3079 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3080 { 3081 /* { 3082 syscallarg(const char *) path; 3083 syscallarg(struct stat *) ub; 3084 } */ 3085 struct stat sb; 3086 int error; 3087 3088 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3089 if (error) 3090 return error; 3091 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3092 } 3093 3094 int 3095 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3096 register_t *retval) 3097 { 3098 /* { 3099 syscallarg(int) fd; 3100 syscallarg(const char *) path; 3101 syscallarg(struct stat *) buf; 3102 syscallarg(int) flag; 3103 } */ 3104 unsigned int nd_flag; 3105 struct stat sb; 3106 int error; 3107 3108 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3109 nd_flag = NOFOLLOW; 3110 else 3111 nd_flag = FOLLOW; 3112 3113 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3114 &sb); 3115 if (error) 3116 return error; 3117 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3118 } 3119 3120 /* 3121 * Get configurable pathname variables. 3122 */ 3123 /* ARGSUSED */ 3124 int 3125 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3126 { 3127 /* { 3128 syscallarg(const char *) path; 3129 syscallarg(int) name; 3130 } */ 3131 int error; 3132 struct pathbuf *pb; 3133 struct nameidata nd; 3134 3135 error = pathbuf_copyin(SCARG(uap, path), &pb); 3136 if (error) { 3137 return error; 3138 } 3139 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3140 if ((error = namei(&nd)) != 0) { 3141 pathbuf_destroy(pb); 3142 return (error); 3143 } 3144 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3145 vput(nd.ni_vp); 3146 pathbuf_destroy(pb); 3147 return (error); 3148 } 3149 3150 /* 3151 * Return target name of a symbolic link. 3152 */ 3153 /* ARGSUSED */ 3154 int 3155 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3156 register_t *retval) 3157 { 3158 /* { 3159 syscallarg(const char *) path; 3160 syscallarg(char *) buf; 3161 syscallarg(size_t) count; 3162 } */ 3163 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3164 SCARG(uap, buf), SCARG(uap, count), retval); 3165 } 3166 3167 static int 3168 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3169 size_t count, register_t *retval) 3170 { 3171 struct vnode *vp; 3172 struct iovec aiov; 3173 struct uio auio; 3174 int error; 3175 struct pathbuf *pb; 3176 struct nameidata nd; 3177 3178 error = pathbuf_copyin(path, &pb); 3179 if (error) { 3180 return error; 3181 } 3182 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3183 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3184 pathbuf_destroy(pb); 3185 return error; 3186 } 3187 vp = nd.ni_vp; 3188 pathbuf_destroy(pb); 3189 if (vp->v_type != VLNK) 3190 error = EINVAL; 3191 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3192 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3193 aiov.iov_base = buf; 3194 aiov.iov_len = count; 3195 auio.uio_iov = &aiov; 3196 auio.uio_iovcnt = 1; 3197 auio.uio_offset = 0; 3198 auio.uio_rw = UIO_READ; 3199 KASSERT(l == curlwp); 3200 auio.uio_vmspace = l->l_proc->p_vmspace; 3201 auio.uio_resid = count; 3202 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3203 *retval = count - auio.uio_resid; 3204 } 3205 vput(vp); 3206 return (error); 3207 } 3208 3209 int 3210 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3211 register_t *retval) 3212 { 3213 /* { 3214 syscallarg(int) fd; 3215 syscallarg(const char *) path; 3216 syscallarg(char *) buf; 3217 syscallarg(size_t) bufsize; 3218 } */ 3219 3220 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3221 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3222 } 3223 3224 /* 3225 * Change flags of a file given a path name. 3226 */ 3227 /* ARGSUSED */ 3228 int 3229 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3230 { 3231 /* { 3232 syscallarg(const char *) path; 3233 syscallarg(u_long) flags; 3234 } */ 3235 struct vnode *vp; 3236 int error; 3237 3238 error = namei_simple_user(SCARG(uap, path), 3239 NSM_FOLLOW_TRYEMULROOT, &vp); 3240 if (error != 0) 3241 return (error); 3242 error = change_flags(vp, SCARG(uap, flags), l); 3243 vput(vp); 3244 return (error); 3245 } 3246 3247 /* 3248 * Change flags of a file given a file descriptor. 3249 */ 3250 /* ARGSUSED */ 3251 int 3252 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3253 { 3254 /* { 3255 syscallarg(int) fd; 3256 syscallarg(u_long) flags; 3257 } */ 3258 struct vnode *vp; 3259 file_t *fp; 3260 int error; 3261 3262 /* fd_getvnode() will use the descriptor for us */ 3263 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3264 return (error); 3265 vp = fp->f_vnode; 3266 error = change_flags(vp, SCARG(uap, flags), l); 3267 VOP_UNLOCK(vp); 3268 fd_putfile(SCARG(uap, fd)); 3269 return (error); 3270 } 3271 3272 /* 3273 * Change flags of a file given a path name; this version does 3274 * not follow links. 3275 */ 3276 int 3277 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3278 { 3279 /* { 3280 syscallarg(const char *) path; 3281 syscallarg(u_long) flags; 3282 } */ 3283 struct vnode *vp; 3284 int error; 3285 3286 error = namei_simple_user(SCARG(uap, path), 3287 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3288 if (error != 0) 3289 return (error); 3290 error = change_flags(vp, SCARG(uap, flags), l); 3291 vput(vp); 3292 return (error); 3293 } 3294 3295 /* 3296 * Common routine to change flags of a file. 3297 */ 3298 int 3299 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3300 { 3301 struct vattr vattr; 3302 int error; 3303 3304 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3305 3306 vattr_null(&vattr); 3307 vattr.va_flags = flags; 3308 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3309 3310 return (error); 3311 } 3312 3313 /* 3314 * Change mode of a file given path name; this version follows links. 3315 */ 3316 /* ARGSUSED */ 3317 int 3318 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3319 { 3320 /* { 3321 syscallarg(const char *) path; 3322 syscallarg(int) mode; 3323 } */ 3324 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3325 SCARG(uap, mode), 0); 3326 } 3327 3328 int 3329 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3330 { 3331 int error; 3332 struct vnode *vp; 3333 namei_simple_flags_t ns_flag; 3334 3335 if (flags & AT_SYMLINK_NOFOLLOW) 3336 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3337 else 3338 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3339 3340 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3341 if (error != 0) 3342 return error; 3343 3344 error = change_mode(vp, mode, l); 3345 3346 vrele(vp); 3347 3348 return (error); 3349 } 3350 3351 /* 3352 * Change mode of a file given a file descriptor. 3353 */ 3354 /* ARGSUSED */ 3355 int 3356 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3357 { 3358 /* { 3359 syscallarg(int) fd; 3360 syscallarg(int) mode; 3361 } */ 3362 file_t *fp; 3363 int error; 3364 3365 /* fd_getvnode() will use the descriptor for us */ 3366 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3367 return (error); 3368 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3369 fd_putfile(SCARG(uap, fd)); 3370 return (error); 3371 } 3372 3373 int 3374 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3375 register_t *retval) 3376 { 3377 /* { 3378 syscallarg(int) fd; 3379 syscallarg(const char *) path; 3380 syscallarg(int) mode; 3381 syscallarg(int) flag; 3382 } */ 3383 3384 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3385 SCARG(uap, mode), SCARG(uap, flag)); 3386 } 3387 3388 /* 3389 * Change mode of a file given path name; this version does not follow links. 3390 */ 3391 /* ARGSUSED */ 3392 int 3393 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3394 { 3395 /* { 3396 syscallarg(const char *) path; 3397 syscallarg(int) mode; 3398 } */ 3399 int error; 3400 struct vnode *vp; 3401 3402 error = namei_simple_user(SCARG(uap, path), 3403 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3404 if (error != 0) 3405 return (error); 3406 3407 error = change_mode(vp, SCARG(uap, mode), l); 3408 3409 vrele(vp); 3410 return (error); 3411 } 3412 3413 /* 3414 * Common routine to set mode given a vnode. 3415 */ 3416 static int 3417 change_mode(struct vnode *vp, int mode, struct lwp *l) 3418 { 3419 struct vattr vattr; 3420 int error; 3421 3422 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3423 vattr_null(&vattr); 3424 vattr.va_mode = mode & ALLPERMS; 3425 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3426 VOP_UNLOCK(vp); 3427 return (error); 3428 } 3429 3430 /* 3431 * Set ownership given a path name; this version follows links. 3432 */ 3433 /* ARGSUSED */ 3434 int 3435 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3436 { 3437 /* { 3438 syscallarg(const char *) path; 3439 syscallarg(uid_t) uid; 3440 syscallarg(gid_t) gid; 3441 } */ 3442 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3443 SCARG(uap, gid), 0); 3444 } 3445 3446 int 3447 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3448 gid_t gid, int flags) 3449 { 3450 int error; 3451 struct vnode *vp; 3452 namei_simple_flags_t ns_flag; 3453 3454 if (flags & AT_SYMLINK_NOFOLLOW) 3455 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3456 else 3457 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3458 3459 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3460 if (error != 0) 3461 return error; 3462 3463 error = change_owner(vp, uid, gid, l, 0); 3464 3465 vrele(vp); 3466 3467 return (error); 3468 } 3469 3470 /* 3471 * Set ownership given a path name; this version follows links. 3472 * Provides POSIX semantics. 3473 */ 3474 /* ARGSUSED */ 3475 int 3476 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3477 { 3478 /* { 3479 syscallarg(const char *) path; 3480 syscallarg(uid_t) uid; 3481 syscallarg(gid_t) gid; 3482 } */ 3483 int error; 3484 struct vnode *vp; 3485 3486 error = namei_simple_user(SCARG(uap, path), 3487 NSM_FOLLOW_TRYEMULROOT, &vp); 3488 if (error != 0) 3489 return (error); 3490 3491 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3492 3493 vrele(vp); 3494 return (error); 3495 } 3496 3497 /* 3498 * Set ownership given a file descriptor. 3499 */ 3500 /* ARGSUSED */ 3501 int 3502 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3503 { 3504 /* { 3505 syscallarg(int) fd; 3506 syscallarg(uid_t) uid; 3507 syscallarg(gid_t) gid; 3508 } */ 3509 int error; 3510 file_t *fp; 3511 3512 /* fd_getvnode() will use the descriptor for us */ 3513 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3514 return (error); 3515 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3516 l, 0); 3517 fd_putfile(SCARG(uap, fd)); 3518 return (error); 3519 } 3520 3521 int 3522 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3523 register_t *retval) 3524 { 3525 /* { 3526 syscallarg(int) fd; 3527 syscallarg(const char *) path; 3528 syscallarg(uid_t) owner; 3529 syscallarg(gid_t) group; 3530 syscallarg(int) flag; 3531 } */ 3532 3533 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3534 SCARG(uap, owner), SCARG(uap, group), 3535 SCARG(uap, flag)); 3536 } 3537 3538 /* 3539 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3540 */ 3541 /* ARGSUSED */ 3542 int 3543 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3544 { 3545 /* { 3546 syscallarg(int) fd; 3547 syscallarg(uid_t) uid; 3548 syscallarg(gid_t) gid; 3549 } */ 3550 int error; 3551 file_t *fp; 3552 3553 /* fd_getvnode() will use the descriptor for us */ 3554 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3555 return (error); 3556 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3557 l, 1); 3558 fd_putfile(SCARG(uap, fd)); 3559 return (error); 3560 } 3561 3562 /* 3563 * Set ownership given a path name; this version does not follow links. 3564 */ 3565 /* ARGSUSED */ 3566 int 3567 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3568 { 3569 /* { 3570 syscallarg(const char *) path; 3571 syscallarg(uid_t) uid; 3572 syscallarg(gid_t) gid; 3573 } */ 3574 int error; 3575 struct vnode *vp; 3576 3577 error = namei_simple_user(SCARG(uap, path), 3578 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3579 if (error != 0) 3580 return (error); 3581 3582 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3583 3584 vrele(vp); 3585 return (error); 3586 } 3587 3588 /* 3589 * Set ownership given a path name; this version does not follow links. 3590 * Provides POSIX/XPG semantics. 3591 */ 3592 /* ARGSUSED */ 3593 int 3594 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3595 { 3596 /* { 3597 syscallarg(const char *) path; 3598 syscallarg(uid_t) uid; 3599 syscallarg(gid_t) gid; 3600 } */ 3601 int error; 3602 struct vnode *vp; 3603 3604 error = namei_simple_user(SCARG(uap, path), 3605 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3606 if (error != 0) 3607 return (error); 3608 3609 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3610 3611 vrele(vp); 3612 return (error); 3613 } 3614 3615 /* 3616 * Common routine to set ownership given a vnode. 3617 */ 3618 static int 3619 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3620 int posix_semantics) 3621 { 3622 struct vattr vattr; 3623 mode_t newmode; 3624 int error; 3625 3626 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3627 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3628 goto out; 3629 3630 #define CHANGED(x) ((int)(x) != -1) 3631 newmode = vattr.va_mode; 3632 if (posix_semantics) { 3633 /* 3634 * POSIX/XPG semantics: if the caller is not the super-user, 3635 * clear set-user-id and set-group-id bits. Both POSIX and 3636 * the XPG consider the behaviour for calls by the super-user 3637 * implementation-defined; we leave the set-user-id and set- 3638 * group-id settings intact in that case. 3639 */ 3640 if (vattr.va_mode & S_ISUID) { 3641 if (kauth_authorize_vnode(l->l_cred, 3642 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3643 newmode &= ~S_ISUID; 3644 } 3645 if (vattr.va_mode & S_ISGID) { 3646 if (kauth_authorize_vnode(l->l_cred, 3647 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3648 newmode &= ~S_ISGID; 3649 } 3650 } else { 3651 /* 3652 * NetBSD semantics: when changing owner and/or group, 3653 * clear the respective bit(s). 3654 */ 3655 if (CHANGED(uid)) 3656 newmode &= ~S_ISUID; 3657 if (CHANGED(gid)) 3658 newmode &= ~S_ISGID; 3659 } 3660 /* Update va_mode iff altered. */ 3661 if (vattr.va_mode == newmode) 3662 newmode = VNOVAL; 3663 3664 vattr_null(&vattr); 3665 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3666 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3667 vattr.va_mode = newmode; 3668 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3669 #undef CHANGED 3670 3671 out: 3672 VOP_UNLOCK(vp); 3673 return (error); 3674 } 3675 3676 /* 3677 * Set the access and modification times given a path name; this 3678 * version follows links. 3679 */ 3680 /* ARGSUSED */ 3681 int 3682 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3683 register_t *retval) 3684 { 3685 /* { 3686 syscallarg(const char *) path; 3687 syscallarg(const struct timeval *) tptr; 3688 } */ 3689 3690 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3691 SCARG(uap, tptr), UIO_USERSPACE); 3692 } 3693 3694 /* 3695 * Set the access and modification times given a file descriptor. 3696 */ 3697 /* ARGSUSED */ 3698 int 3699 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3700 register_t *retval) 3701 { 3702 /* { 3703 syscallarg(int) fd; 3704 syscallarg(const struct timeval *) tptr; 3705 } */ 3706 int error; 3707 file_t *fp; 3708 3709 /* fd_getvnode() will use the descriptor for us */ 3710 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3711 return (error); 3712 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3713 UIO_USERSPACE); 3714 fd_putfile(SCARG(uap, fd)); 3715 return (error); 3716 } 3717 3718 int 3719 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3720 register_t *retval) 3721 { 3722 /* { 3723 syscallarg(int) fd; 3724 syscallarg(const struct timespec *) tptr; 3725 } */ 3726 int error; 3727 file_t *fp; 3728 3729 /* fd_getvnode() will use the descriptor for us */ 3730 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3731 return (error); 3732 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3733 SCARG(uap, tptr), UIO_USERSPACE); 3734 fd_putfile(SCARG(uap, fd)); 3735 return (error); 3736 } 3737 3738 /* 3739 * Set the access and modification times given a path name; this 3740 * version does not follow links. 3741 */ 3742 int 3743 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3744 register_t *retval) 3745 { 3746 /* { 3747 syscallarg(const char *) path; 3748 syscallarg(const struct timeval *) tptr; 3749 } */ 3750 3751 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3752 SCARG(uap, tptr), UIO_USERSPACE); 3753 } 3754 3755 int 3756 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3757 register_t *retval) 3758 { 3759 /* { 3760 syscallarg(int) fd; 3761 syscallarg(const char *) path; 3762 syscallarg(const struct timespec *) tptr; 3763 syscallarg(int) flag; 3764 } */ 3765 int follow; 3766 const struct timespec *tptr; 3767 int error; 3768 3769 tptr = SCARG(uap, tptr); 3770 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3771 3772 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3773 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3774 3775 return error; 3776 } 3777 3778 /* 3779 * Common routine to set access and modification times given a vnode. 3780 */ 3781 int 3782 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3783 const struct timespec *tptr, enum uio_seg seg) 3784 { 3785 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3786 } 3787 3788 int 3789 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3790 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3791 { 3792 struct vattr vattr; 3793 int error, dorele = 0; 3794 namei_simple_flags_t sflags; 3795 bool vanull, setbirthtime; 3796 struct timespec ts[2]; 3797 3798 KASSERT(l != NULL || fdat == AT_FDCWD); 3799 3800 /* 3801 * I have checked all callers and they pass either FOLLOW, 3802 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3803 * is 0. More to the point, they don't pass anything else. 3804 * Let's keep it that way at least until the namei interfaces 3805 * are fully sanitized. 3806 */ 3807 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3808 sflags = (flag == FOLLOW) ? 3809 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3810 3811 if (tptr == NULL) { 3812 vanull = true; 3813 nanotime(&ts[0]); 3814 ts[1] = ts[0]; 3815 } else { 3816 vanull = false; 3817 if (seg != UIO_SYSSPACE) { 3818 error = copyin(tptr, ts, sizeof (ts)); 3819 if (error != 0) 3820 return error; 3821 } else { 3822 ts[0] = tptr[0]; 3823 ts[1] = tptr[1]; 3824 } 3825 } 3826 3827 if (ts[0].tv_nsec == UTIME_NOW) { 3828 nanotime(&ts[0]); 3829 if (ts[1].tv_nsec == UTIME_NOW) { 3830 vanull = true; 3831 ts[1] = ts[0]; 3832 } 3833 } else if (ts[1].tv_nsec == UTIME_NOW) 3834 nanotime(&ts[1]); 3835 3836 if (vp == NULL) { 3837 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3838 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3839 if (error != 0) 3840 return error; 3841 dorele = 1; 3842 } 3843 3844 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3845 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3846 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3847 vattr_null(&vattr); 3848 3849 if (ts[0].tv_nsec != UTIME_OMIT) 3850 vattr.va_atime = ts[0]; 3851 3852 if (ts[1].tv_nsec != UTIME_OMIT) { 3853 vattr.va_mtime = ts[1]; 3854 if (setbirthtime) 3855 vattr.va_birthtime = ts[1]; 3856 } 3857 3858 if (vanull) 3859 vattr.va_vaflags |= VA_UTIMES_NULL; 3860 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3861 VOP_UNLOCK(vp); 3862 3863 if (dorele != 0) 3864 vrele(vp); 3865 3866 return error; 3867 } 3868 3869 int 3870 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3871 const struct timeval *tptr, enum uio_seg seg) 3872 { 3873 struct timespec ts[2]; 3874 struct timespec *tsptr = NULL; 3875 int error; 3876 3877 if (tptr != NULL) { 3878 struct timeval tv[2]; 3879 3880 if (seg != UIO_SYSSPACE) { 3881 error = copyin(tptr, tv, sizeof (tv)); 3882 if (error != 0) 3883 return error; 3884 tptr = tv; 3885 } 3886 3887 if ((tv[0].tv_usec == UTIME_NOW) || 3888 (tv[0].tv_usec == UTIME_OMIT)) 3889 ts[0].tv_nsec = tv[0].tv_usec; 3890 else 3891 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3892 3893 if ((tv[1].tv_usec == UTIME_NOW) || 3894 (tv[1].tv_usec == UTIME_OMIT)) 3895 ts[1].tv_nsec = tv[1].tv_usec; 3896 else 3897 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3898 3899 tsptr = &ts[0]; 3900 } 3901 3902 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3903 } 3904 3905 /* 3906 * Truncate a file given its path name. 3907 */ 3908 /* ARGSUSED */ 3909 int 3910 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3911 { 3912 /* { 3913 syscallarg(const char *) path; 3914 syscallarg(int) pad; 3915 syscallarg(off_t) length; 3916 } */ 3917 struct vnode *vp; 3918 struct vattr vattr; 3919 int error; 3920 3921 if (SCARG(uap, length) < 0) 3922 return EINVAL; 3923 3924 error = namei_simple_user(SCARG(uap, path), 3925 NSM_FOLLOW_TRYEMULROOT, &vp); 3926 if (error != 0) 3927 return (error); 3928 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3929 if (vp->v_type == VDIR) 3930 error = EISDIR; 3931 else if ((error = vn_writechk(vp)) == 0 && 3932 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3933 vattr_null(&vattr); 3934 vattr.va_size = SCARG(uap, length); 3935 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3936 } 3937 vput(vp); 3938 return (error); 3939 } 3940 3941 /* 3942 * Truncate a file given a file descriptor. 3943 */ 3944 /* ARGSUSED */ 3945 int 3946 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3947 { 3948 /* { 3949 syscallarg(int) fd; 3950 syscallarg(int) pad; 3951 syscallarg(off_t) length; 3952 } */ 3953 struct vattr vattr; 3954 struct vnode *vp; 3955 file_t *fp; 3956 int error; 3957 3958 if (SCARG(uap, length) < 0) 3959 return EINVAL; 3960 3961 /* fd_getvnode() will use the descriptor for us */ 3962 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3963 return (error); 3964 if ((fp->f_flag & FWRITE) == 0) { 3965 error = EINVAL; 3966 goto out; 3967 } 3968 vp = fp->f_vnode; 3969 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3970 if (vp->v_type == VDIR) 3971 error = EISDIR; 3972 else if ((error = vn_writechk(vp)) == 0) { 3973 vattr_null(&vattr); 3974 vattr.va_size = SCARG(uap, length); 3975 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3976 } 3977 VOP_UNLOCK(vp); 3978 out: 3979 fd_putfile(SCARG(uap, fd)); 3980 return (error); 3981 } 3982 3983 /* 3984 * Sync an open file. 3985 */ 3986 /* ARGSUSED */ 3987 int 3988 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3989 { 3990 /* { 3991 syscallarg(int) fd; 3992 } */ 3993 struct vnode *vp; 3994 file_t *fp; 3995 int error; 3996 3997 /* fd_getvnode() will use the descriptor for us */ 3998 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3999 return (error); 4000 vp = fp->f_vnode; 4001 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4002 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4003 VOP_UNLOCK(vp); 4004 fd_putfile(SCARG(uap, fd)); 4005 return (error); 4006 } 4007 4008 /* 4009 * Sync a range of file data. API modeled after that found in AIX. 4010 * 4011 * FDATASYNC indicates that we need only save enough metadata to be able 4012 * to re-read the written data. Note we duplicate AIX's requirement that 4013 * the file be open for writing. 4014 */ 4015 /* ARGSUSED */ 4016 int 4017 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4018 { 4019 /* { 4020 syscallarg(int) fd; 4021 syscallarg(int) flags; 4022 syscallarg(off_t) start; 4023 syscallarg(off_t) length; 4024 } */ 4025 struct vnode *vp; 4026 file_t *fp; 4027 int flags, nflags; 4028 off_t s, e, len; 4029 int error; 4030 4031 /* fd_getvnode() will use the descriptor for us */ 4032 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4033 return (error); 4034 4035 if ((fp->f_flag & FWRITE) == 0) { 4036 error = EBADF; 4037 goto out; 4038 } 4039 4040 flags = SCARG(uap, flags); 4041 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4042 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4043 error = EINVAL; 4044 goto out; 4045 } 4046 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4047 if (flags & FDATASYNC) 4048 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4049 else 4050 nflags = FSYNC_WAIT; 4051 if (flags & FDISKSYNC) 4052 nflags |= FSYNC_CACHE; 4053 4054 len = SCARG(uap, length); 4055 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4056 if (len) { 4057 s = SCARG(uap, start); 4058 e = s + len; 4059 if (e < s) { 4060 error = EINVAL; 4061 goto out; 4062 } 4063 } else { 4064 e = 0; 4065 s = 0; 4066 } 4067 4068 vp = fp->f_vnode; 4069 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4070 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4071 VOP_UNLOCK(vp); 4072 out: 4073 fd_putfile(SCARG(uap, fd)); 4074 return (error); 4075 } 4076 4077 /* 4078 * Sync the data of an open file. 4079 */ 4080 /* ARGSUSED */ 4081 int 4082 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4083 { 4084 /* { 4085 syscallarg(int) fd; 4086 } */ 4087 struct vnode *vp; 4088 file_t *fp; 4089 int error; 4090 4091 /* fd_getvnode() will use the descriptor for us */ 4092 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4093 return (error); 4094 if ((fp->f_flag & FWRITE) == 0) { 4095 fd_putfile(SCARG(uap, fd)); 4096 return (EBADF); 4097 } 4098 vp = fp->f_vnode; 4099 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4100 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4101 VOP_UNLOCK(vp); 4102 fd_putfile(SCARG(uap, fd)); 4103 return (error); 4104 } 4105 4106 /* 4107 * Rename files, (standard) BSD semantics frontend. 4108 */ 4109 /* ARGSUSED */ 4110 int 4111 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4112 { 4113 /* { 4114 syscallarg(const char *) from; 4115 syscallarg(const char *) to; 4116 } */ 4117 4118 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4119 SCARG(uap, to), UIO_USERSPACE, 0)); 4120 } 4121 4122 int 4123 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4124 register_t *retval) 4125 { 4126 /* { 4127 syscallarg(int) fromfd; 4128 syscallarg(const char *) from; 4129 syscallarg(int) tofd; 4130 syscallarg(const char *) to; 4131 } */ 4132 4133 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4134 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4135 } 4136 4137 /* 4138 * Rename files, POSIX semantics frontend. 4139 */ 4140 /* ARGSUSED */ 4141 int 4142 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4143 { 4144 /* { 4145 syscallarg(const char *) from; 4146 syscallarg(const char *) to; 4147 } */ 4148 4149 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4150 SCARG(uap, to), UIO_USERSPACE, 1)); 4151 } 4152 4153 /* 4154 * Rename files. Source and destination must either both be directories, 4155 * or both not be directories. If target is a directory, it must be empty. 4156 * If `from' and `to' refer to the same object, the value of the `retain' 4157 * argument is used to determine whether `from' will be 4158 * 4159 * (retain == 0) deleted unless `from' and `to' refer to the same 4160 * object in the file system's name space (BSD). 4161 * (retain == 1) always retained (POSIX). 4162 * 4163 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4164 */ 4165 int 4166 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4167 { 4168 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4169 } 4170 4171 static int 4172 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4173 const char *to, enum uio_seg seg, int retain) 4174 { 4175 struct pathbuf *fpb, *tpb; 4176 struct nameidata fnd, tnd; 4177 struct vnode *fdvp, *fvp; 4178 struct vnode *tdvp, *tvp; 4179 struct mount *mp, *tmp; 4180 int error; 4181 4182 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4183 4184 error = pathbuf_maybe_copyin(from, seg, &fpb); 4185 if (error) 4186 goto out0; 4187 KASSERT(fpb != NULL); 4188 4189 error = pathbuf_maybe_copyin(to, seg, &tpb); 4190 if (error) 4191 goto out1; 4192 KASSERT(tpb != NULL); 4193 4194 /* 4195 * Lookup from. 4196 * 4197 * XXX LOCKPARENT is wrong because we don't actually want it 4198 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4199 * insane, so for the time being we need to leave it like this. 4200 */ 4201 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT | INRENAME), fpb); 4202 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4203 goto out2; 4204 4205 /* 4206 * Pull out the important results of the lookup, fdvp and fvp. 4207 * Of course, fvp is bogus because we're about to unlock fdvp. 4208 */ 4209 fdvp = fnd.ni_dvp; 4210 fvp = fnd.ni_vp; 4211 KASSERT(fdvp != NULL); 4212 KASSERT(fvp != NULL); 4213 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4214 4215 /* 4216 * Make sure neither fdvp nor fvp is locked. 4217 */ 4218 if (fdvp != fvp) 4219 VOP_UNLOCK(fdvp); 4220 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4221 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4222 4223 /* 4224 * Reject renaming `.' and `..'. Can't do this until after 4225 * namei because we need namei's parsing to find the final 4226 * component name. (namei should just leave us with the final 4227 * component name and not look it up itself, but anyway...) 4228 * 4229 * This was here before because we used to relookup from 4230 * instead of to and relookup requires the caller to check 4231 * this, but now file systems may depend on this check, so we 4232 * must retain it until the file systems are all rototilled. 4233 */ 4234 if (((fnd.ni_cnd.cn_namelen == 1) && 4235 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4236 ((fnd.ni_cnd.cn_namelen == 2) && 4237 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4238 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4239 error = EINVAL; /* XXX EISDIR? */ 4240 goto abort0; 4241 } 4242 4243 /* 4244 * Lookup to. 4245 * 4246 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4247 * fvp here to decide whether to add CREATEDIR is a load of 4248 * bollocks because fvp might be the wrong node by now, since 4249 * fdvp is unlocked. 4250 * 4251 * XXX Why not pass CREATEDIR always? 4252 */ 4253 NDINIT(&tnd, RENAME, 4254 (LOCKPARENT | NOCACHE | TRYEMULROOT | INRENAME | 4255 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4256 tpb); 4257 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4258 goto abort0; 4259 4260 /* 4261 * Pull out the important results of the lookup, tdvp and tvp. 4262 * Of course, tvp is bogus because we're about to unlock tdvp. 4263 */ 4264 tdvp = tnd.ni_dvp; 4265 tvp = tnd.ni_vp; 4266 KASSERT(tdvp != NULL); 4267 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4268 4269 /* 4270 * Make sure neither tdvp nor tvp is locked. 4271 */ 4272 if (tdvp != tvp) 4273 VOP_UNLOCK(tdvp); 4274 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4275 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4276 4277 /* 4278 * Reject renaming onto `.' or `..'. relookup is unhappy with 4279 * these, which is why we must do this here. Once upon a time 4280 * we relooked up from instead of to, and consequently didn't 4281 * need this check, but now that we relookup to instead of 4282 * from, we need this; and we shall need it forever forward 4283 * until the VOP_RENAME protocol changes, because file systems 4284 * will no doubt begin to depend on this check. 4285 */ 4286 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4287 error = EISDIR; 4288 goto abort1; 4289 } 4290 if ((tnd.ni_cnd.cn_namelen == 2) && 4291 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4292 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4293 error = EINVAL; 4294 goto abort1; 4295 } 4296 4297 /* 4298 * Get the mount point. If the file system has been unmounted, 4299 * which it may be because we're not holding any vnode locks, 4300 * then v_mount will be NULL. We're not really supposed to 4301 * read v_mount without holding the vnode lock, but since we 4302 * have fdvp referenced, if fdvp->v_mount changes then at worst 4303 * it will be set to NULL, not changed to another mount point. 4304 * And, of course, since it is up to the file system to 4305 * determine the real lock order, we can't lock both fdvp and 4306 * tdvp at the same time. 4307 */ 4308 mp = fdvp->v_mount; 4309 if (mp == NULL) { 4310 error = ENOENT; 4311 goto abort1; 4312 } 4313 4314 /* 4315 * Make sure the mount points match. Again, although we don't 4316 * hold any vnode locks, the v_mount fields may change -- but 4317 * at worst they will change to NULL, so this will never become 4318 * a cross-device rename, because we hold vnode references. 4319 * 4320 * XXX Because nothing is locked and the compiler may reorder 4321 * things here, unmounting the file system at an inopportune 4322 * moment may cause rename to fail with ENXDEV when it really 4323 * should fail with ENOENT. 4324 */ 4325 tmp = tdvp->v_mount; 4326 if (tmp == NULL) { 4327 error = ENOENT; 4328 goto abort1; 4329 } 4330 4331 if (mp != tmp) { 4332 error = EXDEV; 4333 goto abort1; 4334 } 4335 4336 /* 4337 * Take the vfs rename lock to avoid cross-directory screw cases. 4338 * Nothing is locked currently, so taking this lock is safe. 4339 */ 4340 error = VFS_RENAMELOCK_ENTER(mp); 4341 if (error) 4342 goto abort1; 4343 4344 /* 4345 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4346 * and nothing is locked except for the vfs rename lock. 4347 * 4348 * The next step is a little rain dance to conform to the 4349 * insane lock protocol, even though it does nothing to ward 4350 * off race conditions. 4351 * 4352 * We need tdvp and tvp to be locked. However, because we have 4353 * unlocked tdvp in order to hold no locks while we take the 4354 * vfs rename lock, tvp may be wrong here, and we can't safely 4355 * lock it even if the sensible file systems will just unlock 4356 * it straight away. Consequently, we must lock tdvp and then 4357 * relookup tvp to get it locked. 4358 * 4359 * Finally, because the VOP_RENAME protocol is brain-damaged 4360 * and various file systems insanely depend on the semantics of 4361 * this brain damage, the lookup of to must be the last lookup 4362 * before VOP_RENAME. 4363 */ 4364 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4365 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4366 if (error) 4367 goto abort2; 4368 4369 /* 4370 * Drop the old tvp and pick up the new one -- which might be 4371 * the same, but that doesn't matter to us. After this, tdvp 4372 * and tvp should both be locked. 4373 */ 4374 if (tvp != NULL) 4375 vrele(tvp); 4376 tvp = tnd.ni_vp; 4377 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4378 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4379 4380 /* 4381 * The old do_sys_rename had various consistency checks here 4382 * involving fvp and tvp. fvp is bogus already here, and tvp 4383 * will become bogus soon in any sensible file system, so the 4384 * only purpose in putting these checks here is to give lip 4385 * service to these screw cases and to acknowledge that they 4386 * exist, not actually to handle them, but here you go 4387 * anyway... 4388 */ 4389 4390 /* 4391 * Acknowledge that directories and non-directories aren't 4392 * suposed to mix. 4393 */ 4394 if (tvp != NULL) { 4395 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4396 error = ENOTDIR; 4397 goto abort3; 4398 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4399 error = EISDIR; 4400 goto abort3; 4401 } 4402 } 4403 4404 /* 4405 * Acknowledge some random screw case, among the dozens that 4406 * might arise. 4407 */ 4408 if (fvp == tdvp) { 4409 error = EINVAL; 4410 goto abort3; 4411 } 4412 4413 /* 4414 * Acknowledge that POSIX has a wacky screw case. 4415 * 4416 * XXX Eventually the retain flag needs to be passed on to 4417 * VOP_RENAME. 4418 */ 4419 if (fvp == tvp) { 4420 if (retain) { 4421 error = 0; 4422 goto abort3; 4423 } else if ((fdvp == tdvp) && 4424 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4425 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4426 fnd.ni_cnd.cn_namelen))) { 4427 error = 0; 4428 goto abort3; 4429 } 4430 } 4431 4432 /* 4433 * Make sure veriexec can screw us up. (But a race can screw 4434 * up veriexec, of course -- remember, fvp and (soon) tvp are 4435 * bogus.) 4436 */ 4437 #if NVERIEXEC > 0 4438 { 4439 char *f1, *f2; 4440 size_t f1_len; 4441 size_t f2_len; 4442 4443 f1_len = fnd.ni_cnd.cn_namelen + 1; 4444 f1 = kmem_alloc(f1_len, KM_SLEEP); 4445 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4446 4447 f2_len = tnd.ni_cnd.cn_namelen + 1; 4448 f2 = kmem_alloc(f2_len, KM_SLEEP); 4449 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4450 4451 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4452 4453 kmem_free(f1, f1_len); 4454 kmem_free(f2, f2_len); 4455 4456 if (error) 4457 goto abort3; 4458 } 4459 #endif /* NVERIEXEC > 0 */ 4460 4461 /* 4462 * All ready. Incant the rename vop. 4463 */ 4464 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4465 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4466 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4467 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4468 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4469 4470 /* 4471 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4472 * tdvp and tvp. But we can't assert any of that. 4473 */ 4474 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4475 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4476 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4477 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4478 4479 /* 4480 * So all we have left to do is to drop the rename lock and 4481 * destroy the pathbufs. 4482 */ 4483 VFS_RENAMELOCK_EXIT(mp); 4484 goto out2; 4485 4486 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4487 VOP_UNLOCK(tvp); 4488 abort2: VOP_UNLOCK(tdvp); 4489 VFS_RENAMELOCK_EXIT(mp); 4490 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4491 vrele(tdvp); 4492 if (tvp != NULL) 4493 vrele(tvp); 4494 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4495 vrele(fdvp); 4496 vrele(fvp); 4497 out2: pathbuf_destroy(tpb); 4498 out1: pathbuf_destroy(fpb); 4499 out0: return error; 4500 } 4501 4502 /* 4503 * Make a directory file. 4504 */ 4505 /* ARGSUSED */ 4506 int 4507 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4508 { 4509 /* { 4510 syscallarg(const char *) path; 4511 syscallarg(int) mode; 4512 } */ 4513 4514 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4515 SCARG(uap, mode), UIO_USERSPACE); 4516 } 4517 4518 int 4519 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4520 register_t *retval) 4521 { 4522 /* { 4523 syscallarg(int) fd; 4524 syscallarg(const char *) path; 4525 syscallarg(int) mode; 4526 } */ 4527 4528 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4529 SCARG(uap, mode), UIO_USERSPACE); 4530 } 4531 4532 4533 int 4534 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4535 { 4536 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4537 } 4538 4539 static int 4540 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4541 enum uio_seg seg) 4542 { 4543 struct proc *p = curlwp->l_proc; 4544 struct vnode *vp; 4545 struct vattr vattr; 4546 int error; 4547 struct pathbuf *pb; 4548 struct nameidata nd; 4549 4550 KASSERT(l != NULL || fdat == AT_FDCWD); 4551 4552 /* XXX bollocks, should pass in a pathbuf */ 4553 error = pathbuf_maybe_copyin(path, seg, &pb); 4554 if (error) { 4555 return error; 4556 } 4557 4558 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4559 4560 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4561 pathbuf_destroy(pb); 4562 return (error); 4563 } 4564 vp = nd.ni_vp; 4565 if (vp != NULL) { 4566 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4567 if (nd.ni_dvp == vp) 4568 vrele(nd.ni_dvp); 4569 else 4570 vput(nd.ni_dvp); 4571 vrele(vp); 4572 pathbuf_destroy(pb); 4573 return (EEXIST); 4574 } 4575 vattr_null(&vattr); 4576 vattr.va_type = VDIR; 4577 /* We will read cwdi->cwdi_cmask unlocked. */ 4578 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4579 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4580 if (!error) 4581 vrele(nd.ni_vp); 4582 vput(nd.ni_dvp); 4583 pathbuf_destroy(pb); 4584 return (error); 4585 } 4586 4587 /* 4588 * Remove a directory file. 4589 */ 4590 /* ARGSUSED */ 4591 int 4592 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4593 { 4594 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4595 AT_REMOVEDIR, UIO_USERSPACE); 4596 } 4597 4598 /* 4599 * Read a block of directory entries in a file system independent format. 4600 */ 4601 int 4602 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4603 { 4604 /* { 4605 syscallarg(int) fd; 4606 syscallarg(char *) buf; 4607 syscallarg(size_t) count; 4608 } */ 4609 file_t *fp; 4610 int error, done; 4611 4612 /* fd_getvnode() will use the descriptor for us */ 4613 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4614 return (error); 4615 if ((fp->f_flag & FREAD) == 0) { 4616 error = EBADF; 4617 goto out; 4618 } 4619 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4620 SCARG(uap, count), &done, l, 0, 0); 4621 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4622 *retval = done; 4623 out: 4624 fd_putfile(SCARG(uap, fd)); 4625 return (error); 4626 } 4627 4628 /* 4629 * Set the mode mask for creation of filesystem nodes. 4630 */ 4631 int 4632 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4633 { 4634 /* { 4635 syscallarg(mode_t) newmask; 4636 } */ 4637 struct proc *p = l->l_proc; 4638 struct cwdinfo *cwdi; 4639 4640 /* 4641 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4642 * important is that we serialize changes to the mask. The 4643 * rw_exit() will issue a write memory barrier on our behalf, 4644 * and force the changes out to other CPUs (as it must use an 4645 * atomic operation, draining the local CPU's store buffers). 4646 */ 4647 cwdi = p->p_cwdi; 4648 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4649 *retval = cwdi->cwdi_cmask; 4650 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4651 rw_exit(&cwdi->cwdi_lock); 4652 4653 return (0); 4654 } 4655 4656 int 4657 dorevoke(struct vnode *vp, kauth_cred_t cred) 4658 { 4659 struct vattr vattr; 4660 int error, fs_decision; 4661 4662 vn_lock(vp, LK_SHARED | LK_RETRY); 4663 error = VOP_GETATTR(vp, &vattr, cred); 4664 VOP_UNLOCK(vp); 4665 if (error != 0) 4666 return error; 4667 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4668 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4669 fs_decision); 4670 if (!error) 4671 VOP_REVOKE(vp, REVOKEALL); 4672 return (error); 4673 } 4674 4675 /* 4676 * Void all references to file by ripping underlying filesystem 4677 * away from vnode. 4678 */ 4679 /* ARGSUSED */ 4680 int 4681 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4682 { 4683 /* { 4684 syscallarg(const char *) path; 4685 } */ 4686 struct vnode *vp; 4687 int error; 4688 4689 error = namei_simple_user(SCARG(uap, path), 4690 NSM_FOLLOW_TRYEMULROOT, &vp); 4691 if (error != 0) 4692 return (error); 4693 error = dorevoke(vp, l->l_cred); 4694 vrele(vp); 4695 return (error); 4696 } 4697 4698 /* 4699 * Allocate backing store for a file, filling a hole without having to 4700 * explicitly write anything out. 4701 */ 4702 /* ARGSUSED */ 4703 int 4704 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4705 register_t *retval) 4706 { 4707 /* { 4708 syscallarg(int) fd; 4709 syscallarg(off_t) pos; 4710 syscallarg(off_t) len; 4711 } */ 4712 int fd; 4713 off_t pos, len; 4714 struct file *fp; 4715 struct vnode *vp; 4716 int error; 4717 4718 fd = SCARG(uap, fd); 4719 pos = SCARG(uap, pos); 4720 len = SCARG(uap, len); 4721 4722 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4723 *retval = EINVAL; 4724 return 0; 4725 } 4726 4727 error = fd_getvnode(fd, &fp); 4728 if (error) { 4729 *retval = error; 4730 return 0; 4731 } 4732 if ((fp->f_flag & FWRITE) == 0) { 4733 error = EBADF; 4734 goto fail; 4735 } 4736 vp = fp->f_vnode; 4737 4738 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4739 if (vp->v_type == VDIR) { 4740 error = EISDIR; 4741 } else { 4742 error = VOP_FALLOCATE(vp, pos, len); 4743 } 4744 VOP_UNLOCK(vp); 4745 4746 fail: 4747 fd_putfile(fd); 4748 *retval = error; 4749 return 0; 4750 } 4751 4752 /* 4753 * Deallocate backing store for a file, creating a hole. Also used for 4754 * invoking TRIM on disks. 4755 */ 4756 /* ARGSUSED */ 4757 int 4758 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4759 register_t *retval) 4760 { 4761 /* { 4762 syscallarg(int) fd; 4763 syscallarg(off_t) pos; 4764 syscallarg(off_t) len; 4765 } */ 4766 int fd; 4767 off_t pos, len; 4768 struct file *fp; 4769 struct vnode *vp; 4770 int error; 4771 4772 fd = SCARG(uap, fd); 4773 pos = SCARG(uap, pos); 4774 len = SCARG(uap, len); 4775 4776 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4777 return EINVAL; 4778 } 4779 4780 error = fd_getvnode(fd, &fp); 4781 if (error) { 4782 return error; 4783 } 4784 if ((fp->f_flag & FWRITE) == 0) { 4785 error = EBADF; 4786 goto fail; 4787 } 4788 vp = fp->f_vnode; 4789 4790 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4791 if (vp->v_type == VDIR) { 4792 error = EISDIR; 4793 } else { 4794 error = VOP_FDISCARD(vp, pos, len); 4795 } 4796 VOP_UNLOCK(vp); 4797 4798 fail: 4799 fd_putfile(fd); 4800 return error; 4801 } 4802
Indexes created Sat Sep 27 14:09:57 GMT 2025