vfs_syscalls.c revision 1.516.2.2
1 /* $NetBSD: vfs_syscalls.c,v 1.516.2.2 2023/03/07 20:02:57 martin 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 1.516.2.2 2023/03/07 20:02:57 martin Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/fcntl.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/fstrans.h> 91 #include <sys/proc.h> 92 #include <sys/uio.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/quota.h> 99 #include <sys/quotactl.h> 100 #include <sys/ktrace.h> 101 #ifdef FILEASSOC 102 #include <sys/fileassoc.h> 103 #endif /* FILEASSOC */ 104 #include <sys/extattr.h> 105 #include <sys/verified_exec.h> 106 #include <sys/kauth.h> 107 #include <sys/atomic.h> 108 #include <sys/module.h> 109 #include <sys/buf.h> 110 111 #include <miscfs/genfs/genfs.h> 112 #include <miscfs/specfs/specdev.h> 113 114 #include <nfs/rpcv2.h> 115 #include <nfs/nfsproto.h> 116 #include <nfs/nfs.h> 117 #include <nfs/nfs_var.h> 118 119 /* XXX this shouldn't be here */ 120 #ifndef OFF_T_MAX 121 #define OFF_T_MAX __type_max(off_t) 122 #endif 123 124 static int change_flags(struct vnode *, u_long, struct lwp *); 125 static int change_mode(struct vnode *, int, struct lwp *); 126 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 127 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 128 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 129 enum uio_seg); 130 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 131 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 132 enum uio_seg); 133 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 134 enum uio_seg, int); 135 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 136 size_t, register_t *); 137 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 138 139 static int fd_nameiat(struct lwp *, int, struct nameidata *); 140 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 141 namei_simple_flags_t, struct vnode **); 142 143 144 /* 145 * This table is used to maintain compatibility with 4.3BSD 146 * and NetBSD 0.9 mount syscalls - and possibly other systems. 147 * Note, the order is important! 148 * 149 * Do not modify this table. It should only contain filesystems 150 * supported by NetBSD 0.9 and 4.3BSD. 151 */ 152 const char * const mountcompatnames[] = { 153 NULL, /* 0 = MOUNT_NONE */ 154 MOUNT_FFS, /* 1 = MOUNT_UFS */ 155 MOUNT_NFS, /* 2 */ 156 MOUNT_MFS, /* 3 */ 157 MOUNT_MSDOS, /* 4 */ 158 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 159 MOUNT_FDESC, /* 6 */ 160 MOUNT_KERNFS, /* 7 */ 161 NULL, /* 8 = MOUNT_DEVFS */ 162 MOUNT_AFS, /* 9 */ 163 }; 164 165 const int nmountcompatnames = __arraycount(mountcompatnames); 166 167 static int 168 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 169 { 170 file_t *dfp; 171 int error; 172 173 if (fdat != AT_FDCWD) { 174 if ((error = fd_getvnode(fdat, &dfp)) != 0) 175 goto out; 176 177 NDAT(ndp, dfp->f_vnode); 178 } 179 180 error = namei(ndp); 181 182 if (fdat != AT_FDCWD) 183 fd_putfile(fdat); 184 out: 185 return error; 186 } 187 188 static int 189 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 190 namei_simple_flags_t sflags, struct vnode **vp_ret) 191 { 192 file_t *dfp; 193 struct vnode *dvp; 194 int error; 195 196 if (fdat != AT_FDCWD) { 197 if ((error = fd_getvnode(fdat, &dfp)) != 0) 198 goto out; 199 200 dvp = dfp->f_vnode; 201 } else { 202 dvp = NULL; 203 } 204 205 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 206 207 if (fdat != AT_FDCWD) 208 fd_putfile(fdat); 209 out: 210 return error; 211 } 212 213 static int 214 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 215 { 216 int error; 217 218 fp->f_flag = flags & FMASK; 219 fp->f_type = DTYPE_VNODE; 220 fp->f_ops = &vnops; 221 fp->f_vnode = vp; 222 223 if (flags & (O_EXLOCK | O_SHLOCK)) { 224 struct flock lf; 225 int type; 226 227 lf.l_whence = SEEK_SET; 228 lf.l_start = 0; 229 lf.l_len = 0; 230 if (flags & O_EXLOCK) 231 lf.l_type = F_WRLCK; 232 else 233 lf.l_type = F_RDLCK; 234 type = F_FLOCK; 235 if ((flags & FNONBLOCK) == 0) 236 type |= F_WAIT; 237 VOP_UNLOCK(vp); 238 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 239 if (error) { 240 (void) vn_close(vp, fp->f_flag, fp->f_cred); 241 fd_abort(l->l_proc, fp, indx); 242 return error; 243 } 244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 245 atomic_or_uint(&fp->f_flag, FHASLOCK); 246 } 247 if (flags & O_CLOEXEC) 248 fd_set_exclose(l, indx, true); 249 return 0; 250 } 251 252 static int 253 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 254 void *data, size_t *data_len) 255 { 256 struct mount *mp; 257 int error = 0, saved_flags; 258 259 mp = vp->v_mount; 260 saved_flags = mp->mnt_flag; 261 262 /* We can operate only on VV_ROOT nodes. */ 263 if ((vp->v_vflag & VV_ROOT) == 0) { 264 error = EINVAL; 265 goto out; 266 } 267 268 /* 269 * We only allow the filesystem to be reloaded if it 270 * is currently mounted read-only. Additionally, we 271 * prevent read-write to read-only downgrades. 272 */ 273 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 274 (mp->mnt_flag & MNT_RDONLY) == 0 && 275 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 276 error = EOPNOTSUPP; /* Needs translation */ 277 goto out; 278 } 279 280 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 281 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 282 if (error) 283 goto out; 284 285 error = vfs_suspend(mp, 0); 286 if (error) 287 goto out; 288 289 mutex_enter(&mp->mnt_updating); 290 291 mp->mnt_flag &= ~MNT_OP_FLAGS; 292 mp->mnt_flag |= flags & MNT_OP_FLAGS; 293 294 /* 295 * Set the mount level flags. 296 */ 297 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 298 if ((flags & MNT_RDONLY)) 299 mp->mnt_iflag |= IMNT_WANTRDONLY; 300 else 301 mp->mnt_iflag |= IMNT_WANTRDWR; 302 } 303 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 304 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 305 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 306 mp->mnt_flag &= ~MNT_RDONLY; 307 308 error = VFS_MOUNT(mp, path, data, data_len); 309 310 if (error && data != NULL) { 311 int error2; 312 313 /* 314 * Update failed; let's try and see if it was an 315 * export request. For compat with 3.0 and earlier. 316 */ 317 error2 = vfs_hooks_reexport(mp, path, data); 318 319 /* 320 * Only update error code if the export request was 321 * understood but some problem occurred while 322 * processing it. 323 */ 324 if (error2 != EJUSTRETURN) 325 error = error2; 326 } 327 328 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 329 mp->mnt_flag |= MNT_RDONLY; 330 if (error) 331 mp->mnt_flag = saved_flags; 332 mp->mnt_flag &= ~MNT_OP_FLAGS; 333 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 334 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 335 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 336 vfs_syncer_add_to_worklist(mp); 337 } else { 338 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 339 vfs_syncer_remove_from_worklist(mp); 340 } 341 mutex_exit(&mp->mnt_updating); 342 vfs_resume(mp); 343 344 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 345 (flags & MNT_EXTATTR)) { 346 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 347 NULL, 0, NULL) != 0) { 348 printf("%s: failed to start extattr, error = %d", 349 mp->mnt_stat.f_mntonname, error); 350 mp->mnt_flag &= ~MNT_EXTATTR; 351 } 352 } 353 354 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 355 !(flags & MNT_EXTATTR)) { 356 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 357 NULL, 0, NULL) != 0) { 358 printf("%s: failed to stop extattr, error = %d", 359 mp->mnt_stat.f_mntonname, error); 360 mp->mnt_flag |= MNT_RDONLY; 361 } 362 } 363 out: 364 return (error); 365 } 366 367 static int 368 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 369 struct vfsops **vfsops) 370 { 371 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 372 int error; 373 374 if (type_seg == UIO_USERSPACE) { 375 /* Copy file-system type from userspace. */ 376 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 377 } else { 378 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 379 KASSERT(error == 0); 380 } 381 382 if (error) { 383 /* 384 * Historically, filesystem types were identified by numbers. 385 * If we get an integer for the filesystem type instead of a 386 * string, we check to see if it matches one of the historic 387 * filesystem types. 388 */ 389 u_long fsindex = (u_long)fstype; 390 if (fsindex >= nmountcompatnames || 391 mountcompatnames[fsindex] == NULL) 392 return ENODEV; 393 strlcpy(fstypename, mountcompatnames[fsindex], 394 sizeof(fstypename)); 395 } 396 397 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 398 if (strcmp(fstypename, "ufs") == 0) 399 fstypename[0] = 'f'; 400 401 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 402 return 0; 403 404 /* If we can autoload a vfs module, try again */ 405 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 406 407 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 408 return 0; 409 410 return ENODEV; 411 } 412 413 static int 414 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 415 void *data, size_t *data_len) 416 { 417 struct mount *mp; 418 int error; 419 420 /* If MNT_GETARGS is specified, it should be the only flag. */ 421 if (flags & ~MNT_GETARGS) 422 return EINVAL; 423 424 mp = vp->v_mount; 425 426 /* XXX: probably some notion of "can see" here if we want isolation. */ 427 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 428 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 429 if (error) 430 return error; 431 432 if ((vp->v_vflag & VV_ROOT) == 0) 433 return EINVAL; 434 435 if (vfs_busy(mp)) 436 return EPERM; 437 438 mutex_enter(&mp->mnt_updating); 439 mp->mnt_flag &= ~MNT_OP_FLAGS; 440 mp->mnt_flag |= MNT_GETARGS; 441 error = VFS_MOUNT(mp, path, data, data_len); 442 mp->mnt_flag &= ~MNT_OP_FLAGS; 443 mutex_exit(&mp->mnt_updating); 444 445 vfs_unbusy(mp); 446 return (error); 447 } 448 449 int 450 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 451 { 452 /* { 453 syscallarg(const char *) type; 454 syscallarg(const char *) path; 455 syscallarg(int) flags; 456 syscallarg(void *) data; 457 syscallarg(size_t) data_len; 458 } */ 459 460 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 461 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 462 SCARG(uap, data_len), retval); 463 } 464 465 int 466 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 467 const char *path, int flags, void *data, enum uio_seg data_seg, 468 size_t data_len, register_t *retval) 469 { 470 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 471 struct vnode *vp; 472 void *data_buf = data; 473 bool vfsopsrele = false; 474 size_t alloc_sz = 0; 475 int error; 476 477 /* 478 * Get vnode to be covered 479 */ 480 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 481 if (error != 0) { 482 vp = NULL; 483 goto done; 484 } 485 486 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 487 vfsops = vp->v_mount->mnt_op; 488 } else { 489 /* 'type' is userspace */ 490 error = mount_get_vfsops(type, type_seg, &vfsops); 491 if (error != 0) 492 goto done; 493 vfsopsrele = true; 494 } 495 496 /* 497 * We allow data to be NULL, even for userspace. Some fs's don't need 498 * it. The others will handle NULL. 499 */ 500 if (data != NULL && data_seg == UIO_USERSPACE) { 501 if (data_len == 0) { 502 /* No length supplied, use default for filesystem */ 503 data_len = vfsops->vfs_min_mount_data; 504 505 /* 506 * Hopefully a longer buffer won't make copyin() fail. 507 * For compatibility with 3.0 and earlier. 508 */ 509 if (flags & MNT_UPDATE 510 && data_len < sizeof (struct mnt_export_args30)) 511 data_len = sizeof (struct mnt_export_args30); 512 } 513 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 514 error = EINVAL; 515 goto done; 516 } 517 alloc_sz = data_len; 518 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 519 520 /* NFS needs the buffer even for mnt_getargs .... */ 521 error = copyin(data, data_buf, data_len); 522 if (error != 0) 523 goto done; 524 } 525 526 if (flags & MNT_GETARGS) { 527 if (data_len == 0) { 528 error = EINVAL; 529 goto done; 530 } 531 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 532 if (error != 0) 533 goto done; 534 if (data_seg == UIO_USERSPACE) 535 error = copyout(data_buf, data, data_len); 536 *retval = data_len; 537 } else if (flags & MNT_UPDATE) { 538 error = mount_update(l, vp, path, flags, data_buf, &data_len); 539 } else { 540 /* Locking is handled internally in mount_domount(). */ 541 KASSERT(vfsopsrele == true); 542 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 543 &data_len); 544 vfsopsrele = false; 545 } 546 547 done: 548 if (vfsopsrele) 549 vfs_delref(vfsops); 550 if (vp != NULL) { 551 vrele(vp); 552 } 553 if (data_buf != data) 554 kmem_free(data_buf, alloc_sz); 555 return (error); 556 } 557 558 /* 559 * Unmount a file system. 560 * 561 * Note: unmount takes a path to the vnode mounted on as argument, 562 * not special file (as before). 563 */ 564 /* ARGSUSED */ 565 int 566 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 567 { 568 /* { 569 syscallarg(const char *) path; 570 syscallarg(int) flags; 571 } */ 572 struct vnode *vp; 573 struct mount *mp; 574 int error; 575 struct pathbuf *pb; 576 struct nameidata nd; 577 578 error = pathbuf_copyin(SCARG(uap, path), &pb); 579 if (error) { 580 return error; 581 } 582 583 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 584 if ((error = namei(&nd)) != 0) { 585 pathbuf_destroy(pb); 586 return error; 587 } 588 vp = nd.ni_vp; 589 pathbuf_destroy(pb); 590 591 mp = vp->v_mount; 592 vfs_ref(mp); 593 VOP_UNLOCK(vp); 594 595 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 596 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 597 if (error) { 598 vrele(vp); 599 vfs_rele(mp); 600 return (error); 601 } 602 603 /* 604 * Don't allow unmounting the root file system. 605 */ 606 if (mp->mnt_flag & MNT_ROOTFS) { 607 vrele(vp); 608 vfs_rele(mp); 609 return (EINVAL); 610 } 611 612 /* 613 * Must be the root of the filesystem 614 */ 615 if ((vp->v_vflag & VV_ROOT) == 0) { 616 vrele(vp); 617 vfs_rele(mp); 618 return (EINVAL); 619 } 620 621 vrele(vp); 622 error = dounmount(mp, SCARG(uap, flags), l); 623 vfs_rele(mp); 624 return error; 625 } 626 627 /* 628 * Sync each mounted filesystem. 629 */ 630 #ifdef DEBUG 631 int syncprt = 0; 632 struct ctldebug debug0 = { "syncprt", &syncprt }; 633 #endif 634 635 void 636 do_sys_sync(struct lwp *l) 637 { 638 mount_iterator_t *iter; 639 struct mount *mp; 640 int asyncflag; 641 642 mountlist_iterator_init(&iter); 643 while ((mp = mountlist_iterator_next(iter)) != NULL) { 644 mutex_enter(&mp->mnt_updating); 645 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 646 asyncflag = mp->mnt_flag & MNT_ASYNC; 647 mp->mnt_flag &= ~MNT_ASYNC; 648 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 649 if (asyncflag) 650 mp->mnt_flag |= MNT_ASYNC; 651 } 652 mutex_exit(&mp->mnt_updating); 653 } 654 mountlist_iterator_destroy(iter); 655 #ifdef DEBUG 656 if (syncprt) 657 vfs_bufstats(); 658 #endif /* DEBUG */ 659 } 660 661 /* ARGSUSED */ 662 int 663 sys_sync(struct lwp *l, const void *v, register_t *retval) 664 { 665 do_sys_sync(l); 666 return (0); 667 } 668 669 670 /* 671 * Access or change filesystem quotas. 672 * 673 * (this is really 14 different calls bundled into one) 674 */ 675 676 static int 677 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 678 { 679 struct quotastat info_k; 680 int error; 681 682 /* ensure any padding bytes are cleared */ 683 memset(&info_k, 0, sizeof(info_k)); 684 685 error = vfs_quotactl_stat(mp, &info_k); 686 if (error) { 687 return error; 688 } 689 690 return copyout(&info_k, info_u, sizeof(info_k)); 691 } 692 693 static int 694 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 695 struct quotaidtypestat *info_u) 696 { 697 struct quotaidtypestat info_k; 698 int error; 699 700 /* ensure any padding bytes are cleared */ 701 memset(&info_k, 0, sizeof(info_k)); 702 703 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 704 if (error) { 705 return error; 706 } 707 708 return copyout(&info_k, info_u, sizeof(info_k)); 709 } 710 711 static int 712 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 713 struct quotaobjtypestat *info_u) 714 { 715 struct quotaobjtypestat info_k; 716 int error; 717 718 /* ensure any padding bytes are cleared */ 719 memset(&info_k, 0, sizeof(info_k)); 720 721 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 722 if (error) { 723 return error; 724 } 725 726 return copyout(&info_k, info_u, sizeof(info_k)); 727 } 728 729 static int 730 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 731 struct quotaval *val_u) 732 { 733 struct quotakey key_k; 734 struct quotaval val_k; 735 int error; 736 737 /* ensure any padding bytes are cleared */ 738 memset(&val_k, 0, sizeof(val_k)); 739 740 error = copyin(key_u, &key_k, sizeof(key_k)); 741 if (error) { 742 return error; 743 } 744 745 error = vfs_quotactl_get(mp, &key_k, &val_k); 746 if (error) { 747 return error; 748 } 749 750 return copyout(&val_k, val_u, sizeof(val_k)); 751 } 752 753 static int 754 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 755 const struct quotaval *val_u) 756 { 757 struct quotakey key_k; 758 struct quotaval val_k; 759 int error; 760 761 error = copyin(key_u, &key_k, sizeof(key_k)); 762 if (error) { 763 return error; 764 } 765 766 error = copyin(val_u, &val_k, sizeof(val_k)); 767 if (error) { 768 return error; 769 } 770 771 return vfs_quotactl_put(mp, &key_k, &val_k); 772 } 773 774 static int 775 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 776 { 777 struct quotakey key_k; 778 int error; 779 780 error = copyin(key_u, &key_k, sizeof(key_k)); 781 if (error) { 782 return error; 783 } 784 785 return vfs_quotactl_del(mp, &key_k); 786 } 787 788 static int 789 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 790 { 791 struct quotakcursor cursor_k; 792 int error; 793 794 /* ensure any padding bytes are cleared */ 795 memset(&cursor_k, 0, sizeof(cursor_k)); 796 797 error = vfs_quotactl_cursoropen(mp, &cursor_k); 798 if (error) { 799 return error; 800 } 801 802 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 803 } 804 805 static int 806 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 807 { 808 struct quotakcursor cursor_k; 809 int error; 810 811 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 812 if (error) { 813 return error; 814 } 815 816 return vfs_quotactl_cursorclose(mp, &cursor_k); 817 } 818 819 static int 820 do_sys_quotactl_cursorskipidtype(struct mount *mp, 821 struct quotakcursor *cursor_u, int idtype) 822 { 823 struct quotakcursor cursor_k; 824 int error; 825 826 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 827 if (error) { 828 return error; 829 } 830 831 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 832 if (error) { 833 return error; 834 } 835 836 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 837 } 838 839 static int 840 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 841 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 842 unsigned *ret_u) 843 { 844 #define CGET_STACK_MAX 8 845 struct quotakcursor cursor_k; 846 struct quotakey stackkeys[CGET_STACK_MAX]; 847 struct quotaval stackvals[CGET_STACK_MAX]; 848 struct quotakey *keys_k; 849 struct quotaval *vals_k; 850 unsigned ret_k; 851 int error; 852 853 if (maxnum > 128) { 854 maxnum = 128; 855 } 856 857 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 858 if (error) { 859 return error; 860 } 861 862 if (maxnum <= CGET_STACK_MAX) { 863 keys_k = stackkeys; 864 vals_k = stackvals; 865 /* ensure any padding bytes are cleared */ 866 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 867 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 868 } else { 869 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 870 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 871 } 872 873 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 874 &ret_k); 875 if (error) { 876 goto fail; 877 } 878 879 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 880 if (error) { 881 goto fail; 882 } 883 884 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 885 if (error) { 886 goto fail; 887 } 888 889 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 890 if (error) { 891 goto fail; 892 } 893 894 /* do last to maximize the chance of being able to recover a failure */ 895 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 896 897 fail: 898 if (keys_k != stackkeys) { 899 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 900 } 901 if (vals_k != stackvals) { 902 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 903 } 904 return error; 905 } 906 907 static int 908 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 909 int *ret_u) 910 { 911 struct quotakcursor cursor_k; 912 int ret_k; 913 int error; 914 915 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 916 if (error) { 917 return error; 918 } 919 920 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 921 if (error) { 922 return error; 923 } 924 925 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 926 if (error) { 927 return error; 928 } 929 930 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 931 } 932 933 static int 934 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 935 { 936 struct quotakcursor cursor_k; 937 int error; 938 939 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 940 if (error) { 941 return error; 942 } 943 944 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 945 if (error) { 946 return error; 947 } 948 949 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 950 } 951 952 static int 953 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 954 { 955 char *path_k; 956 int error; 957 958 /* XXX this should probably be a struct pathbuf */ 959 path_k = PNBUF_GET(); 960 error = copyin(path_u, path_k, PATH_MAX); 961 if (error) { 962 PNBUF_PUT(path_k); 963 return error; 964 } 965 966 error = vfs_quotactl_quotaon(mp, idtype, path_k); 967 968 PNBUF_PUT(path_k); 969 return error; 970 } 971 972 static int 973 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 974 { 975 return vfs_quotactl_quotaoff(mp, idtype); 976 } 977 978 int 979 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 980 { 981 struct mount *mp; 982 struct vnode *vp; 983 int error; 984 985 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 986 if (error != 0) 987 return (error); 988 mp = vp->v_mount; 989 990 switch (args->qc_op) { 991 case QUOTACTL_STAT: 992 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 993 break; 994 case QUOTACTL_IDTYPESTAT: 995 error = do_sys_quotactl_idtypestat(mp, 996 args->u.idtypestat.qc_idtype, 997 args->u.idtypestat.qc_info); 998 break; 999 case QUOTACTL_OBJTYPESTAT: 1000 error = do_sys_quotactl_objtypestat(mp, 1001 args->u.objtypestat.qc_objtype, 1002 args->u.objtypestat.qc_info); 1003 break; 1004 case QUOTACTL_GET: 1005 error = do_sys_quotactl_get(mp, 1006 args->u.get.qc_key, 1007 args->u.get.qc_val); 1008 break; 1009 case QUOTACTL_PUT: 1010 error = do_sys_quotactl_put(mp, 1011 args->u.put.qc_key, 1012 args->u.put.qc_val); 1013 break; 1014 case QUOTACTL_DEL: 1015 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1016 break; 1017 case QUOTACTL_CURSOROPEN: 1018 error = do_sys_quotactl_cursoropen(mp, 1019 args->u.cursoropen.qc_cursor); 1020 break; 1021 case QUOTACTL_CURSORCLOSE: 1022 error = do_sys_quotactl_cursorclose(mp, 1023 args->u.cursorclose.qc_cursor); 1024 break; 1025 case QUOTACTL_CURSORSKIPIDTYPE: 1026 error = do_sys_quotactl_cursorskipidtype(mp, 1027 args->u.cursorskipidtype.qc_cursor, 1028 args->u.cursorskipidtype.qc_idtype); 1029 break; 1030 case QUOTACTL_CURSORGET: 1031 error = do_sys_quotactl_cursorget(mp, 1032 args->u.cursorget.qc_cursor, 1033 args->u.cursorget.qc_keys, 1034 args->u.cursorget.qc_vals, 1035 args->u.cursorget.qc_maxnum, 1036 args->u.cursorget.qc_ret); 1037 break; 1038 case QUOTACTL_CURSORATEND: 1039 error = do_sys_quotactl_cursoratend(mp, 1040 args->u.cursoratend.qc_cursor, 1041 args->u.cursoratend.qc_ret); 1042 break; 1043 case QUOTACTL_CURSORREWIND: 1044 error = do_sys_quotactl_cursorrewind(mp, 1045 args->u.cursorrewind.qc_cursor); 1046 break; 1047 case QUOTACTL_QUOTAON: 1048 error = do_sys_quotactl_quotaon(mp, 1049 args->u.quotaon.qc_idtype, 1050 args->u.quotaon.qc_quotafile); 1051 break; 1052 case QUOTACTL_QUOTAOFF: 1053 error = do_sys_quotactl_quotaoff(mp, 1054 args->u.quotaoff.qc_idtype); 1055 break; 1056 default: 1057 error = EINVAL; 1058 break; 1059 } 1060 1061 vrele(vp); 1062 return error; 1063 } 1064 1065 /* ARGSUSED */ 1066 int 1067 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1068 register_t *retval) 1069 { 1070 /* { 1071 syscallarg(const char *) path; 1072 syscallarg(struct quotactl_args *) args; 1073 } */ 1074 struct quotactl_args args; 1075 int error; 1076 1077 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1078 if (error) { 1079 return error; 1080 } 1081 1082 return do_sys_quotactl(SCARG(uap, path), &args); 1083 } 1084 1085 int 1086 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1087 int root) 1088 { 1089 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1090 int error = 0; 1091 1092 /* 1093 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1094 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1095 * overrides MNT_NOWAIT. 1096 */ 1097 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1098 (flags != MNT_WAIT && flags != 0)) { 1099 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1100 goto done; 1101 } 1102 1103 /* Get the filesystem stats now */ 1104 memset(sp, 0, sizeof(*sp)); 1105 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1106 return error; 1107 } 1108 1109 if (cwdi->cwdi_rdir == NULL) 1110 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1111 done: 1112 if (cwdi->cwdi_rdir != NULL) { 1113 size_t len; 1114 char *bp; 1115 char c; 1116 char *path = PNBUF_GET(); 1117 1118 bp = path + MAXPATHLEN; 1119 *--bp = '\0'; 1120 rw_enter(&cwdi->cwdi_lock, RW_READER); 1121 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1122 MAXPATHLEN / 2, 0, l); 1123 rw_exit(&cwdi->cwdi_lock); 1124 if (error) { 1125 PNBUF_PUT(path); 1126 return error; 1127 } 1128 len = strlen(bp); 1129 if (len != 1) { 1130 /* 1131 * for mount points that are below our root, we can see 1132 * them, so we fix up the pathname and return them. The 1133 * rest we cannot see, so we don't allow viewing the 1134 * data. 1135 */ 1136 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1137 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1138 (void)strlcpy(sp->f_mntonname, 1139 c == '\0' ? "/" : &sp->f_mntonname[len], 1140 sizeof(sp->f_mntonname)); 1141 } else { 1142 if (root) 1143 (void)strlcpy(sp->f_mntonname, "/", 1144 sizeof(sp->f_mntonname)); 1145 else 1146 error = EPERM; 1147 } 1148 } 1149 PNBUF_PUT(path); 1150 } 1151 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1152 return error; 1153 } 1154 1155 /* 1156 * Get filesystem statistics by path. 1157 */ 1158 int 1159 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1160 { 1161 struct mount *mp; 1162 int error; 1163 struct vnode *vp; 1164 1165 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1166 if (error != 0) 1167 return error; 1168 mp = vp->v_mount; 1169 error = dostatvfs(mp, sb, l, flags, 1); 1170 vrele(vp); 1171 return error; 1172 } 1173 1174 /* ARGSUSED */ 1175 int 1176 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1177 { 1178 /* { 1179 syscallarg(const char *) path; 1180 syscallarg(struct statvfs *) buf; 1181 syscallarg(int) flags; 1182 } */ 1183 struct statvfs *sb; 1184 int error; 1185 1186 sb = STATVFSBUF_GET(); 1187 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1188 if (error == 0) 1189 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1190 STATVFSBUF_PUT(sb); 1191 return error; 1192 } 1193 1194 /* 1195 * Get filesystem statistics by fd. 1196 */ 1197 int 1198 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1199 { 1200 file_t *fp; 1201 struct mount *mp; 1202 int error; 1203 1204 /* fd_getvnode() will use the descriptor for us */ 1205 if ((error = fd_getvnode(fd, &fp)) != 0) 1206 return (error); 1207 mp = fp->f_vnode->v_mount; 1208 error = dostatvfs(mp, sb, curlwp, flags, 1); 1209 fd_putfile(fd); 1210 return error; 1211 } 1212 1213 /* ARGSUSED */ 1214 int 1215 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1216 { 1217 /* { 1218 syscallarg(int) fd; 1219 syscallarg(struct statvfs *) buf; 1220 syscallarg(int) flags; 1221 } */ 1222 struct statvfs *sb; 1223 int error; 1224 1225 sb = STATVFSBUF_GET(); 1226 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1227 if (error == 0) 1228 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1229 STATVFSBUF_PUT(sb); 1230 return error; 1231 } 1232 1233 1234 /* 1235 * Get statistics on all filesystems. 1236 */ 1237 int 1238 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1239 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1240 register_t *retval) 1241 { 1242 int root = 0; 1243 mount_iterator_t *iter; 1244 struct proc *p = l->l_proc; 1245 struct mount *mp; 1246 struct statvfs *sb; 1247 size_t count, maxcount; 1248 int error = 0; 1249 1250 sb = STATVFSBUF_GET(); 1251 maxcount = bufsize / entry_sz; 1252 count = 0; 1253 mountlist_iterator_init(&iter); 1254 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1255 if (sfsp && count < maxcount) { 1256 error = dostatvfs(mp, sb, l, flags, 0); 1257 if (error) { 1258 error = 0; 1259 continue; 1260 } 1261 error = copyfn(sb, sfsp, entry_sz); 1262 if (error) 1263 goto out; 1264 sfsp = (char *)sfsp + entry_sz; 1265 root |= strcmp(sb->f_mntonname, "/") == 0; 1266 } 1267 count++; 1268 } 1269 1270 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1271 /* 1272 * fake a root entry 1273 */ 1274 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1275 sb, l, flags, 1); 1276 if (error != 0) 1277 goto out; 1278 if (sfsp) { 1279 error = copyfn(sb, sfsp, entry_sz); 1280 if (error != 0) 1281 goto out; 1282 } 1283 count++; 1284 } 1285 if (sfsp && count > maxcount) 1286 *retval = maxcount; 1287 else 1288 *retval = count; 1289 out: 1290 mountlist_iterator_destroy(iter); 1291 STATVFSBUF_PUT(sb); 1292 return error; 1293 } 1294 1295 int 1296 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1297 { 1298 /* { 1299 syscallarg(struct statvfs *) buf; 1300 syscallarg(size_t) bufsize; 1301 syscallarg(int) flags; 1302 } */ 1303 1304 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1305 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1306 } 1307 1308 /* 1309 * Change current working directory to a given file descriptor. 1310 */ 1311 /* ARGSUSED */ 1312 int 1313 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1314 { 1315 /* { 1316 syscallarg(int) fd; 1317 } */ 1318 struct proc *p = l->l_proc; 1319 struct cwdinfo *cwdi; 1320 struct vnode *vp, *tdp; 1321 struct mount *mp; 1322 file_t *fp; 1323 int error, fd; 1324 1325 /* fd_getvnode() will use the descriptor for us */ 1326 fd = SCARG(uap, fd); 1327 if ((error = fd_getvnode(fd, &fp)) != 0) 1328 return (error); 1329 vp = fp->f_vnode; 1330 1331 vref(vp); 1332 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1333 if (vp->v_type != VDIR) 1334 error = ENOTDIR; 1335 else 1336 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1337 if (error) { 1338 vput(vp); 1339 goto out; 1340 } 1341 while ((mp = vp->v_mountedhere) != NULL) { 1342 error = vfs_busy(mp); 1343 vput(vp); 1344 if (error != 0) 1345 goto out; 1346 error = VFS_ROOT(mp, &tdp); 1347 vfs_unbusy(mp); 1348 if (error) 1349 goto out; 1350 vp = tdp; 1351 } 1352 VOP_UNLOCK(vp); 1353 1354 /* 1355 * Disallow changing to a directory not under the process's 1356 * current root directory (if there is one). 1357 */ 1358 cwdi = p->p_cwdi; 1359 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1360 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1361 vrele(vp); 1362 error = EPERM; /* operation not permitted */ 1363 } else { 1364 vrele(cwdi->cwdi_cdir); 1365 cwdi->cwdi_cdir = vp; 1366 } 1367 rw_exit(&cwdi->cwdi_lock); 1368 1369 out: 1370 fd_putfile(fd); 1371 return (error); 1372 } 1373 1374 /* 1375 * Change this process's notion of the root directory to a given file 1376 * descriptor. 1377 */ 1378 int 1379 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1380 { 1381 struct proc *p = l->l_proc; 1382 struct vnode *vp; 1383 file_t *fp; 1384 int error, fd = SCARG(uap, fd); 1385 1386 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1387 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1388 return error; 1389 /* fd_getvnode() will use the descriptor for us */ 1390 if ((error = fd_getvnode(fd, &fp)) != 0) 1391 return error; 1392 vp = fp->f_vnode; 1393 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1394 if (vp->v_type != VDIR) 1395 error = ENOTDIR; 1396 else 1397 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1398 VOP_UNLOCK(vp); 1399 if (error) 1400 goto out; 1401 vref(vp); 1402 1403 change_root(p->p_cwdi, vp, l); 1404 1405 out: 1406 fd_putfile(fd); 1407 return (error); 1408 } 1409 1410 /* 1411 * Change current working directory (``.''). 1412 */ 1413 /* ARGSUSED */ 1414 int 1415 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1416 { 1417 /* { 1418 syscallarg(const char *) path; 1419 } */ 1420 struct proc *p = l->l_proc; 1421 struct cwdinfo *cwdi; 1422 int error; 1423 struct vnode *vp; 1424 1425 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1426 &vp, l)) != 0) 1427 return (error); 1428 cwdi = p->p_cwdi; 1429 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1430 vrele(cwdi->cwdi_cdir); 1431 cwdi->cwdi_cdir = vp; 1432 rw_exit(&cwdi->cwdi_lock); 1433 return (0); 1434 } 1435 1436 /* 1437 * Change notion of root (``/'') directory. 1438 */ 1439 /* ARGSUSED */ 1440 int 1441 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1442 { 1443 /* { 1444 syscallarg(const char *) path; 1445 } */ 1446 struct proc *p = l->l_proc; 1447 int error; 1448 struct vnode *vp; 1449 1450 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1451 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1452 return (error); 1453 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1454 &vp, l)) != 0) 1455 return (error); 1456 1457 change_root(p->p_cwdi, vp, l); 1458 1459 return (0); 1460 } 1461 1462 /* 1463 * Common routine for chroot and fchroot. 1464 * NB: callers need to properly authorize the change root operation. 1465 */ 1466 void 1467 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1468 { 1469 struct proc *p = l->l_proc; 1470 kauth_cred_t ncred; 1471 1472 ncred = kauth_cred_alloc(); 1473 1474 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1475 if (cwdi->cwdi_rdir != NULL) 1476 vrele(cwdi->cwdi_rdir); 1477 cwdi->cwdi_rdir = vp; 1478 1479 /* 1480 * Prevent escaping from chroot by putting the root under 1481 * the working directory. Silently chdir to / if we aren't 1482 * already there. 1483 */ 1484 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1485 /* 1486 * XXX would be more failsafe to change directory to a 1487 * deadfs node here instead 1488 */ 1489 vrele(cwdi->cwdi_cdir); 1490 vref(vp); 1491 cwdi->cwdi_cdir = vp; 1492 } 1493 rw_exit(&cwdi->cwdi_lock); 1494 1495 /* Get a write lock on the process credential. */ 1496 proc_crmod_enter(); 1497 1498 kauth_cred_clone(p->p_cred, ncred); 1499 kauth_proc_chroot(ncred, p->p_cwdi); 1500 1501 /* Broadcast our credentials to the process and other LWPs. */ 1502 proc_crmod_leave(ncred, p->p_cred, true); 1503 } 1504 1505 /* 1506 * Common routine for chroot and chdir. 1507 * XXX "where" should be enum uio_seg 1508 */ 1509 int 1510 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1511 { 1512 struct pathbuf *pb; 1513 struct nameidata nd; 1514 int error; 1515 1516 error = pathbuf_maybe_copyin(path, where, &pb); 1517 if (error) { 1518 return error; 1519 } 1520 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1521 if ((error = namei(&nd)) != 0) { 1522 pathbuf_destroy(pb); 1523 return error; 1524 } 1525 *vpp = nd.ni_vp; 1526 pathbuf_destroy(pb); 1527 1528 if ((*vpp)->v_type != VDIR) 1529 error = ENOTDIR; 1530 else 1531 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1532 1533 if (error) 1534 vput(*vpp); 1535 else 1536 VOP_UNLOCK(*vpp); 1537 return (error); 1538 } 1539 1540 /* 1541 * Internals of sys_open - path has already been converted into a pathbuf 1542 * (so we can easily reuse this function from other parts of the kernel, 1543 * like posix_spawn post-processing). 1544 */ 1545 int 1546 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1547 int open_mode, int *fd) 1548 { 1549 struct proc *p = l->l_proc; 1550 struct cwdinfo *cwdi = p->p_cwdi; 1551 file_t *fp; 1552 struct vnode *vp; 1553 int flags, cmode; 1554 int indx, error; 1555 struct nameidata nd; 1556 1557 if (open_flags & O_SEARCH) { 1558 open_flags &= ~(int)O_SEARCH; 1559 } 1560 1561 flags = FFLAGS(open_flags); 1562 if ((flags & (FREAD | FWRITE)) == 0) 1563 return EINVAL; 1564 1565 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1566 return error; 1567 } 1568 1569 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1570 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1571 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1572 if (dvp != NULL) 1573 NDAT(&nd, dvp); 1574 1575 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1576 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1577 fd_abort(p, fp, indx); 1578 if ((error == EDUPFD || error == EMOVEFD) && 1579 l->l_dupfd >= 0 && /* XXX from fdopen */ 1580 (error = 1581 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1582 *fd = indx; 1583 return 0; 1584 } 1585 return error; 1586 } 1587 1588 l->l_dupfd = 0; 1589 vp = nd.ni_vp; 1590 1591 if ((error = open_setfp(l, fp, vp, indx, flags))) 1592 return error; 1593 1594 VOP_UNLOCK(vp); 1595 *fd = indx; 1596 fd_affix(p, fp, indx); 1597 return 0; 1598 } 1599 1600 int 1601 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1602 { 1603 struct pathbuf *pb; 1604 int error, oflags; 1605 1606 oflags = FFLAGS(open_flags); 1607 if ((oflags & (FREAD | FWRITE)) == 0) 1608 return EINVAL; 1609 1610 pb = pathbuf_create(path); 1611 if (pb == NULL) 1612 return ENOMEM; 1613 1614 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1615 pathbuf_destroy(pb); 1616 1617 return error; 1618 } 1619 1620 /* 1621 * Check permissions, allocate an open file structure, 1622 * and call the device open routine if any. 1623 */ 1624 static int 1625 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1626 int mode, int *fd) 1627 { 1628 file_t *dfp = NULL; 1629 struct vnode *dvp = NULL; 1630 struct pathbuf *pb; 1631 int error; 1632 1633 #ifdef COMPAT_10 /* XXX: and perhaps later */ 1634 if (path == NULL) { 1635 pb = pathbuf_create("."); 1636 if (pb == NULL) 1637 return ENOMEM; 1638 } else 1639 #endif 1640 { 1641 error = pathbuf_copyin(path, &pb); 1642 if (error) 1643 return error; 1644 } 1645 1646 if (fdat != AT_FDCWD) { 1647 /* fd_getvnode() will use the descriptor for us */ 1648 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1649 goto out; 1650 1651 dvp = dfp->f_vnode; 1652 } 1653 1654 error = do_open(l, dvp, pb, flags, mode, fd); 1655 1656 if (dfp != NULL) 1657 fd_putfile(fdat); 1658 out: 1659 pathbuf_destroy(pb); 1660 return error; 1661 } 1662 1663 int 1664 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1665 { 1666 /* { 1667 syscallarg(const char *) path; 1668 syscallarg(int) flags; 1669 syscallarg(int) mode; 1670 } */ 1671 int error; 1672 int fd; 1673 1674 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1675 SCARG(uap, flags), SCARG(uap, mode), &fd); 1676 1677 if (error == 0) 1678 *retval = fd; 1679 1680 return error; 1681 } 1682 1683 int 1684 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1685 { 1686 /* { 1687 syscallarg(int) fd; 1688 syscallarg(const char *) path; 1689 syscallarg(int) oflags; 1690 syscallarg(int) mode; 1691 } */ 1692 int error; 1693 int fd; 1694 1695 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1696 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1697 1698 if (error == 0) 1699 *retval = fd; 1700 1701 return error; 1702 } 1703 1704 static void 1705 vfs__fhfree(fhandle_t *fhp) 1706 { 1707 size_t fhsize; 1708 1709 fhsize = FHANDLE_SIZE(fhp); 1710 kmem_free(fhp, fhsize); 1711 } 1712 1713 /* 1714 * vfs_composefh: compose a filehandle. 1715 */ 1716 1717 int 1718 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1719 { 1720 struct mount *mp; 1721 struct fid *fidp; 1722 int error; 1723 size_t needfhsize; 1724 size_t fidsize; 1725 1726 mp = vp->v_mount; 1727 fidp = NULL; 1728 if (*fh_size < FHANDLE_SIZE_MIN) { 1729 fidsize = 0; 1730 } else { 1731 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1732 if (fhp != NULL) { 1733 memset(fhp, 0, *fh_size); 1734 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1735 fidp = &fhp->fh_fid; 1736 } 1737 } 1738 error = VFS_VPTOFH(vp, fidp, &fidsize); 1739 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1740 if (error == 0 && *fh_size < needfhsize) { 1741 error = E2BIG; 1742 } 1743 *fh_size = needfhsize; 1744 return error; 1745 } 1746 1747 int 1748 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1749 { 1750 struct mount *mp; 1751 fhandle_t *fhp; 1752 size_t fhsize; 1753 size_t fidsize; 1754 int error; 1755 1756 mp = vp->v_mount; 1757 fidsize = 0; 1758 error = VFS_VPTOFH(vp, NULL, &fidsize); 1759 KASSERT(error != 0); 1760 if (error != E2BIG) { 1761 goto out; 1762 } 1763 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1764 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1765 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1766 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1767 if (error == 0) { 1768 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1769 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1770 *fhpp = fhp; 1771 } else { 1772 kmem_free(fhp, fhsize); 1773 } 1774 out: 1775 return error; 1776 } 1777 1778 void 1779 vfs_composefh_free(fhandle_t *fhp) 1780 { 1781 1782 vfs__fhfree(fhp); 1783 } 1784 1785 /* 1786 * vfs_fhtovp: lookup a vnode by a filehandle. 1787 */ 1788 1789 int 1790 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1791 { 1792 struct mount *mp; 1793 int error; 1794 1795 *vpp = NULL; 1796 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1797 if (mp == NULL) { 1798 error = ESTALE; 1799 goto out; 1800 } 1801 if (mp->mnt_op->vfs_fhtovp == NULL) { 1802 error = EOPNOTSUPP; 1803 goto out; 1804 } 1805 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1806 out: 1807 return error; 1808 } 1809 1810 /* 1811 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1812 * the needed size. 1813 */ 1814 1815 int 1816 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1817 { 1818 fhandle_t *fhp; 1819 int error; 1820 1821 if (fhsize > FHANDLE_SIZE_MAX) { 1822 return EINVAL; 1823 } 1824 if (fhsize < FHANDLE_SIZE_MIN) { 1825 return EINVAL; 1826 } 1827 again: 1828 fhp = kmem_alloc(fhsize, KM_SLEEP); 1829 error = copyin(ufhp, fhp, fhsize); 1830 if (error == 0) { 1831 /* XXX this check shouldn't be here */ 1832 if (FHANDLE_SIZE(fhp) == fhsize) { 1833 *fhpp = fhp; 1834 return 0; 1835 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1836 /* 1837 * a kludge for nfsv2 padded handles. 1838 */ 1839 size_t sz; 1840 1841 sz = FHANDLE_SIZE(fhp); 1842 kmem_free(fhp, fhsize); 1843 fhsize = sz; 1844 goto again; 1845 } else { 1846 /* 1847 * userland told us wrong size. 1848 */ 1849 error = EINVAL; 1850 } 1851 } 1852 kmem_free(fhp, fhsize); 1853 return error; 1854 } 1855 1856 void 1857 vfs_copyinfh_free(fhandle_t *fhp) 1858 { 1859 1860 vfs__fhfree(fhp); 1861 } 1862 1863 /* 1864 * Get file handle system call 1865 */ 1866 int 1867 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1868 { 1869 /* { 1870 syscallarg(char *) fname; 1871 syscallarg(fhandle_t *) fhp; 1872 syscallarg(size_t *) fh_size; 1873 } */ 1874 struct vnode *vp; 1875 fhandle_t *fh; 1876 int error; 1877 struct pathbuf *pb; 1878 struct nameidata nd; 1879 size_t sz; 1880 size_t usz; 1881 1882 /* 1883 * Must be super user 1884 */ 1885 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1886 0, NULL, NULL, NULL); 1887 if (error) 1888 return (error); 1889 1890 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1891 if (error) { 1892 return error; 1893 } 1894 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1895 error = namei(&nd); 1896 if (error) { 1897 pathbuf_destroy(pb); 1898 return error; 1899 } 1900 vp = nd.ni_vp; 1901 pathbuf_destroy(pb); 1902 1903 error = vfs_composefh_alloc(vp, &fh); 1904 vput(vp); 1905 if (error != 0) { 1906 return error; 1907 } 1908 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1909 if (error != 0) { 1910 goto out; 1911 } 1912 sz = FHANDLE_SIZE(fh); 1913 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1914 if (error != 0) { 1915 goto out; 1916 } 1917 if (usz >= sz) { 1918 error = copyout(fh, SCARG(uap, fhp), sz); 1919 } else { 1920 error = E2BIG; 1921 } 1922 out: 1923 vfs_composefh_free(fh); 1924 return (error); 1925 } 1926 1927 /* 1928 * Open a file given a file handle. 1929 * 1930 * Check permissions, allocate an open file structure, 1931 * and call the device open routine if any. 1932 */ 1933 1934 int 1935 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1936 register_t *retval) 1937 { 1938 file_t *fp; 1939 struct vnode *vp = NULL; 1940 kauth_cred_t cred = l->l_cred; 1941 file_t *nfp; 1942 int indx, error; 1943 struct vattr va; 1944 fhandle_t *fh; 1945 int flags; 1946 proc_t *p; 1947 1948 p = curproc; 1949 1950 /* 1951 * Must be super user 1952 */ 1953 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1954 0, NULL, NULL, NULL))) 1955 return (error); 1956 1957 if (oflags & O_SEARCH) { 1958 oflags &= ~(int)O_SEARCH; 1959 } 1960 1961 flags = FFLAGS(oflags); 1962 if ((flags & (FREAD | FWRITE)) == 0) 1963 return (EINVAL); 1964 if ((flags & O_CREAT)) 1965 return (EINVAL); 1966 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1967 return (error); 1968 fp = nfp; 1969 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1970 if (error != 0) { 1971 goto bad; 1972 } 1973 error = vfs_fhtovp(fh, &vp); 1974 vfs_copyinfh_free(fh); 1975 if (error != 0) { 1976 goto bad; 1977 } 1978 1979 /* Now do an effective vn_open */ 1980 1981 if (vp->v_type == VSOCK) { 1982 error = EOPNOTSUPP; 1983 goto bad; 1984 } 1985 error = vn_openchk(vp, cred, flags); 1986 if (error != 0) 1987 goto bad; 1988 if (flags & O_TRUNC) { 1989 VOP_UNLOCK(vp); /* XXX */ 1990 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1991 vattr_null(&va); 1992 va.va_size = 0; 1993 error = VOP_SETATTR(vp, &va, cred); 1994 if (error) 1995 goto bad; 1996 } 1997 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1998 goto bad; 1999 if (flags & FWRITE) { 2000 mutex_enter(vp->v_interlock); 2001 vp->v_writecount++; 2002 mutex_exit(vp->v_interlock); 2003 } 2004 2005 /* done with modified vn_open, now finish what sys_open does. */ 2006 if ((error = open_setfp(l, fp, vp, indx, flags))) 2007 return error; 2008 2009 VOP_UNLOCK(vp); 2010 *retval = indx; 2011 fd_affix(p, fp, indx); 2012 return (0); 2013 2014 bad: 2015 fd_abort(p, fp, indx); 2016 if (vp != NULL) 2017 vput(vp); 2018 return (error); 2019 } 2020 2021 int 2022 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2023 { 2024 /* { 2025 syscallarg(const void *) fhp; 2026 syscallarg(size_t) fh_size; 2027 syscallarg(int) flags; 2028 } */ 2029 2030 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2031 SCARG(uap, flags), retval); 2032 } 2033 2034 int 2035 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2036 { 2037 int error; 2038 fhandle_t *fh; 2039 struct vnode *vp; 2040 2041 /* 2042 * Must be super user 2043 */ 2044 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2045 0, NULL, NULL, NULL))) 2046 return (error); 2047 2048 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2049 if (error != 0) 2050 return error; 2051 2052 error = vfs_fhtovp(fh, &vp); 2053 vfs_copyinfh_free(fh); 2054 if (error != 0) 2055 return error; 2056 2057 error = vn_stat(vp, sb); 2058 vput(vp); 2059 return error; 2060 } 2061 2062 2063 /* ARGSUSED */ 2064 int 2065 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2066 { 2067 /* { 2068 syscallarg(const void *) fhp; 2069 syscallarg(size_t) fh_size; 2070 syscallarg(struct stat *) sb; 2071 } */ 2072 struct stat sb; 2073 int error; 2074 2075 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2076 if (error) 2077 return error; 2078 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2079 } 2080 2081 int 2082 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2083 int flags) 2084 { 2085 fhandle_t *fh; 2086 struct mount *mp; 2087 struct vnode *vp; 2088 int error; 2089 2090 /* 2091 * Must be super user 2092 */ 2093 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2094 0, NULL, NULL, NULL))) 2095 return error; 2096 2097 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2098 if (error != 0) 2099 return error; 2100 2101 error = vfs_fhtovp(fh, &vp); 2102 vfs_copyinfh_free(fh); 2103 if (error != 0) 2104 return error; 2105 2106 mp = vp->v_mount; 2107 error = dostatvfs(mp, sb, l, flags, 1); 2108 vput(vp); 2109 return error; 2110 } 2111 2112 /* ARGSUSED */ 2113 int 2114 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2115 { 2116 /* { 2117 syscallarg(const void *) fhp; 2118 syscallarg(size_t) fh_size; 2119 syscallarg(struct statvfs *) buf; 2120 syscallarg(int) flags; 2121 } */ 2122 struct statvfs *sb = STATVFSBUF_GET(); 2123 int error; 2124 2125 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2126 SCARG(uap, flags)); 2127 if (error == 0) 2128 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2129 STATVFSBUF_PUT(sb); 2130 return error; 2131 } 2132 2133 /* 2134 * Create a special file. 2135 */ 2136 /* ARGSUSED */ 2137 int 2138 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2139 register_t *retval) 2140 { 2141 /* { 2142 syscallarg(const char *) path; 2143 syscallarg(mode_t) mode; 2144 syscallarg(dev_t) dev; 2145 } */ 2146 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2147 SCARG(uap, dev), retval, UIO_USERSPACE); 2148 } 2149 2150 int 2151 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2152 register_t *retval) 2153 { 2154 /* { 2155 syscallarg(int) fd; 2156 syscallarg(const char *) path; 2157 syscallarg(mode_t) mode; 2158 syscallarg(int) pad; 2159 syscallarg(dev_t) dev; 2160 } */ 2161 2162 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2163 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2164 } 2165 2166 int 2167 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2168 register_t *retval, enum uio_seg seg) 2169 { 2170 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2171 } 2172 2173 int 2174 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2175 dev_t dev, register_t *retval, enum uio_seg seg) 2176 { 2177 struct proc *p = l->l_proc; 2178 struct vnode *vp; 2179 struct vattr vattr; 2180 int error, optype; 2181 struct pathbuf *pb; 2182 struct nameidata nd; 2183 const char *pathstring; 2184 2185 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2186 0, NULL, NULL, NULL)) != 0) 2187 return (error); 2188 2189 optype = VOP_MKNOD_DESCOFFSET; 2190 2191 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2192 if (error) { 2193 return error; 2194 } 2195 pathstring = pathbuf_stringcopy_get(pb); 2196 if (pathstring == NULL) { 2197 pathbuf_destroy(pb); 2198 return ENOMEM; 2199 } 2200 2201 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2202 2203 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2204 goto out; 2205 vp = nd.ni_vp; 2206 2207 if (vp != NULL) 2208 error = EEXIST; 2209 else { 2210 vattr_null(&vattr); 2211 /* We will read cwdi->cwdi_cmask unlocked. */ 2212 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2213 vattr.va_rdev = dev; 2214 2215 switch (mode & S_IFMT) { 2216 case S_IFMT: /* used by badsect to flag bad sectors */ 2217 vattr.va_type = VBAD; 2218 break; 2219 case S_IFCHR: 2220 vattr.va_type = VCHR; 2221 break; 2222 case S_IFBLK: 2223 vattr.va_type = VBLK; 2224 break; 2225 case S_IFWHT: 2226 optype = VOP_WHITEOUT_DESCOFFSET; 2227 break; 2228 case S_IFREG: 2229 #if NVERIEXEC > 0 2230 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2231 O_CREAT); 2232 #endif /* NVERIEXEC > 0 */ 2233 vattr.va_type = VREG; 2234 vattr.va_rdev = VNOVAL; 2235 optype = VOP_CREATE_DESCOFFSET; 2236 break; 2237 default: 2238 error = EINVAL; 2239 break; 2240 } 2241 } 2242 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2243 && vattr.va_rdev == VNOVAL) 2244 error = EINVAL; 2245 if (!error) { 2246 switch (optype) { 2247 case VOP_WHITEOUT_DESCOFFSET: 2248 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2249 if (error) 2250 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2251 vput(nd.ni_dvp); 2252 break; 2253 2254 case VOP_MKNOD_DESCOFFSET: 2255 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2256 &nd.ni_cnd, &vattr); 2257 if (error == 0) 2258 vrele(nd.ni_vp); 2259 vput(nd.ni_dvp); 2260 break; 2261 2262 case VOP_CREATE_DESCOFFSET: 2263 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2264 &nd.ni_cnd, &vattr); 2265 if (error == 0) 2266 vrele(nd.ni_vp); 2267 vput(nd.ni_dvp); 2268 break; 2269 } 2270 } else { 2271 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2272 if (nd.ni_dvp == vp) 2273 vrele(nd.ni_dvp); 2274 else 2275 vput(nd.ni_dvp); 2276 if (vp) 2277 vrele(vp); 2278 } 2279 out: 2280 pathbuf_stringcopy_put(pb, pathstring); 2281 pathbuf_destroy(pb); 2282 return (error); 2283 } 2284 2285 /* 2286 * Create a named pipe. 2287 */ 2288 /* ARGSUSED */ 2289 int 2290 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2291 { 2292 /* { 2293 syscallarg(const char *) path; 2294 syscallarg(int) mode; 2295 } */ 2296 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2297 } 2298 2299 int 2300 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2301 register_t *retval) 2302 { 2303 /* { 2304 syscallarg(int) fd; 2305 syscallarg(const char *) path; 2306 syscallarg(int) mode; 2307 } */ 2308 2309 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2310 SCARG(uap, mode)); 2311 } 2312 2313 static int 2314 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2315 { 2316 struct proc *p = l->l_proc; 2317 struct vattr vattr; 2318 int error; 2319 struct pathbuf *pb; 2320 struct nameidata nd; 2321 2322 error = pathbuf_copyin(path, &pb); 2323 if (error) { 2324 return error; 2325 } 2326 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2327 2328 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2329 pathbuf_destroy(pb); 2330 return error; 2331 } 2332 if (nd.ni_vp != NULL) { 2333 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2334 if (nd.ni_dvp == nd.ni_vp) 2335 vrele(nd.ni_dvp); 2336 else 2337 vput(nd.ni_dvp); 2338 vrele(nd.ni_vp); 2339 pathbuf_destroy(pb); 2340 return (EEXIST); 2341 } 2342 vattr_null(&vattr); 2343 vattr.va_type = VFIFO; 2344 /* We will read cwdi->cwdi_cmask unlocked. */ 2345 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2346 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2347 if (error == 0) 2348 vrele(nd.ni_vp); 2349 vput(nd.ni_dvp); 2350 pathbuf_destroy(pb); 2351 return (error); 2352 } 2353 2354 /* 2355 * Make a hard file link. 2356 */ 2357 /* ARGSUSED */ 2358 int 2359 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2360 const char *link, int follow, register_t *retval) 2361 { 2362 struct vnode *vp; 2363 struct pathbuf *linkpb; 2364 struct nameidata nd; 2365 namei_simple_flags_t ns_flags; 2366 int error; 2367 2368 if (follow & AT_SYMLINK_FOLLOW) 2369 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2370 else 2371 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2372 2373 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2374 if (error != 0) 2375 return (error); 2376 error = pathbuf_copyin(link, &linkpb); 2377 if (error) { 2378 goto out1; 2379 } 2380 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2381 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2382 goto out2; 2383 if (nd.ni_vp) { 2384 error = EEXIST; 2385 goto abortop; 2386 } 2387 /* Prevent hard links on directories. */ 2388 if (vp->v_type == VDIR) { 2389 error = EPERM; 2390 goto abortop; 2391 } 2392 /* Prevent cross-mount operation. */ 2393 if (nd.ni_dvp->v_mount != vp->v_mount) { 2394 error = EXDEV; 2395 goto abortop; 2396 } 2397 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2398 VOP_UNLOCK(nd.ni_dvp); 2399 vrele(nd.ni_dvp); 2400 out2: 2401 pathbuf_destroy(linkpb); 2402 out1: 2403 vrele(vp); 2404 return (error); 2405 abortop: 2406 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2407 if (nd.ni_dvp == nd.ni_vp) 2408 vrele(nd.ni_dvp); 2409 else 2410 vput(nd.ni_dvp); 2411 if (nd.ni_vp != NULL) 2412 vrele(nd.ni_vp); 2413 goto out2; 2414 } 2415 2416 int 2417 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2418 { 2419 /* { 2420 syscallarg(const char *) path; 2421 syscallarg(const char *) link; 2422 } */ 2423 const char *path = SCARG(uap, path); 2424 const char *link = SCARG(uap, link); 2425 2426 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2427 AT_SYMLINK_FOLLOW, retval); 2428 } 2429 2430 int 2431 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2432 register_t *retval) 2433 { 2434 /* { 2435 syscallarg(int) fd1; 2436 syscallarg(const char *) name1; 2437 syscallarg(int) fd2; 2438 syscallarg(const char *) name2; 2439 syscallarg(int) flags; 2440 } */ 2441 int fd1 = SCARG(uap, fd1); 2442 const char *name1 = SCARG(uap, name1); 2443 int fd2 = SCARG(uap, fd2); 2444 const char *name2 = SCARG(uap, name2); 2445 int follow; 2446 2447 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2448 2449 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2450 } 2451 2452 2453 int 2454 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2455 { 2456 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2457 } 2458 2459 static int 2460 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2461 const char *link, enum uio_seg seg) 2462 { 2463 struct proc *p = curproc; 2464 struct vattr vattr; 2465 char *path; 2466 int error; 2467 struct pathbuf *linkpb; 2468 struct nameidata nd; 2469 2470 KASSERT(l != NULL || fdat == AT_FDCWD); 2471 2472 path = PNBUF_GET(); 2473 if (seg == UIO_USERSPACE) { 2474 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2475 goto out1; 2476 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2477 goto out1; 2478 } else { 2479 KASSERT(strlen(patharg) < MAXPATHLEN); 2480 strcpy(path, patharg); 2481 linkpb = pathbuf_create(link); 2482 if (linkpb == NULL) { 2483 error = ENOMEM; 2484 goto out1; 2485 } 2486 } 2487 ktrkuser("symlink-target", path, strlen(path)); 2488 2489 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2490 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2491 goto out2; 2492 if (nd.ni_vp) { 2493 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2494 if (nd.ni_dvp == nd.ni_vp) 2495 vrele(nd.ni_dvp); 2496 else 2497 vput(nd.ni_dvp); 2498 vrele(nd.ni_vp); 2499 error = EEXIST; 2500 goto out2; 2501 } 2502 vattr_null(&vattr); 2503 vattr.va_type = VLNK; 2504 /* We will read cwdi->cwdi_cmask unlocked. */ 2505 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2506 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2507 if (error == 0) 2508 vrele(nd.ni_vp); 2509 vput(nd.ni_dvp); 2510 out2: 2511 pathbuf_destroy(linkpb); 2512 out1: 2513 PNBUF_PUT(path); 2514 return (error); 2515 } 2516 2517 /* 2518 * Make a symbolic link. 2519 */ 2520 /* ARGSUSED */ 2521 int 2522 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2523 { 2524 /* { 2525 syscallarg(const char *) path; 2526 syscallarg(const char *) link; 2527 } */ 2528 2529 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2530 UIO_USERSPACE); 2531 } 2532 2533 int 2534 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2535 register_t *retval) 2536 { 2537 /* { 2538 syscallarg(const char *) path1; 2539 syscallarg(int) fd; 2540 syscallarg(const char *) path2; 2541 } */ 2542 2543 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2544 SCARG(uap, path2), UIO_USERSPACE); 2545 } 2546 2547 /* 2548 * Delete a whiteout from the filesystem. 2549 */ 2550 /* ARGSUSED */ 2551 int 2552 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2553 { 2554 /* { 2555 syscallarg(const char *) path; 2556 } */ 2557 int error; 2558 struct pathbuf *pb; 2559 struct nameidata nd; 2560 2561 error = pathbuf_copyin(SCARG(uap, path), &pb); 2562 if (error) { 2563 return error; 2564 } 2565 2566 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2567 error = namei(&nd); 2568 if (error) { 2569 pathbuf_destroy(pb); 2570 return (error); 2571 } 2572 2573 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2574 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2575 if (nd.ni_dvp == nd.ni_vp) 2576 vrele(nd.ni_dvp); 2577 else 2578 vput(nd.ni_dvp); 2579 if (nd.ni_vp) 2580 vrele(nd.ni_vp); 2581 pathbuf_destroy(pb); 2582 return (EEXIST); 2583 } 2584 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2585 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2586 vput(nd.ni_dvp); 2587 pathbuf_destroy(pb); 2588 return (error); 2589 } 2590 2591 /* 2592 * Delete a name from the filesystem. 2593 */ 2594 /* ARGSUSED */ 2595 int 2596 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2597 { 2598 /* { 2599 syscallarg(const char *) path; 2600 } */ 2601 2602 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2603 } 2604 2605 int 2606 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2607 register_t *retval) 2608 { 2609 /* { 2610 syscallarg(int) fd; 2611 syscallarg(const char *) path; 2612 syscallarg(int) flag; 2613 } */ 2614 2615 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2616 SCARG(uap, flag), UIO_USERSPACE); 2617 } 2618 2619 int 2620 do_sys_unlink(const char *arg, enum uio_seg seg) 2621 { 2622 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2623 } 2624 2625 static int 2626 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2627 enum uio_seg seg) 2628 { 2629 struct vnode *vp; 2630 int error; 2631 struct pathbuf *pb; 2632 struct nameidata nd; 2633 const char *pathstring; 2634 2635 KASSERT(l != NULL || fdat == AT_FDCWD); 2636 2637 error = pathbuf_maybe_copyin(arg, seg, &pb); 2638 if (error) { 2639 return error; 2640 } 2641 pathstring = pathbuf_stringcopy_get(pb); 2642 if (pathstring == NULL) { 2643 pathbuf_destroy(pb); 2644 return ENOMEM; 2645 } 2646 2647 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2648 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2649 goto out; 2650 vp = nd.ni_vp; 2651 2652 /* 2653 * The root of a mounted filesystem cannot be deleted. 2654 */ 2655 if ((vp->v_vflag & VV_ROOT) != 0) { 2656 error = EBUSY; 2657 goto abort; 2658 } 2659 2660 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2661 error = EBUSY; 2662 goto abort; 2663 } 2664 2665 /* 2666 * No rmdir "." please. 2667 */ 2668 if (nd.ni_dvp == vp) { 2669 error = EINVAL; 2670 goto abort; 2671 } 2672 2673 /* 2674 * AT_REMOVEDIR is required to remove a directory 2675 */ 2676 if (vp->v_type == VDIR) { 2677 if (!(flags & AT_REMOVEDIR)) { 2678 error = EPERM; 2679 goto abort; 2680 } else { 2681 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2682 vput(nd.ni_dvp); 2683 goto out; 2684 } 2685 } 2686 2687 /* 2688 * Starting here we only deal with non directories. 2689 */ 2690 if (flags & AT_REMOVEDIR) { 2691 error = ENOTDIR; 2692 goto abort; 2693 } 2694 2695 #if NVERIEXEC > 0 2696 /* Handle remove requests for veriexec entries. */ 2697 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2698 goto abort; 2699 } 2700 #endif /* NVERIEXEC > 0 */ 2701 2702 #ifdef FILEASSOC 2703 (void)fileassoc_file_delete(vp); 2704 #endif /* FILEASSOC */ 2705 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2706 vput(nd.ni_dvp); 2707 goto out; 2708 2709 abort: 2710 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2711 if (nd.ni_dvp == vp) 2712 vrele(nd.ni_dvp); 2713 else 2714 vput(nd.ni_dvp); 2715 vput(vp); 2716 2717 out: 2718 pathbuf_stringcopy_put(pb, pathstring); 2719 pathbuf_destroy(pb); 2720 return (error); 2721 } 2722 2723 /* 2724 * Reposition read/write file offset. 2725 */ 2726 int 2727 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2728 { 2729 /* { 2730 syscallarg(int) fd; 2731 syscallarg(int) pad; 2732 syscallarg(off_t) offset; 2733 syscallarg(int) whence; 2734 } */ 2735 kauth_cred_t cred = l->l_cred; 2736 file_t *fp; 2737 struct vnode *vp; 2738 struct vattr vattr; 2739 off_t newoff; 2740 int error, fd; 2741 2742 fd = SCARG(uap, fd); 2743 2744 if ((fp = fd_getfile(fd)) == NULL) 2745 return (EBADF); 2746 2747 vp = fp->f_vnode; 2748 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2749 error = ESPIPE; 2750 goto out; 2751 } 2752 2753 vn_lock(vp, LK_SHARED | LK_RETRY); 2754 2755 switch (SCARG(uap, whence)) { 2756 case SEEK_CUR: 2757 newoff = fp->f_offset + SCARG(uap, offset); 2758 break; 2759 case SEEK_END: 2760 error = VOP_GETATTR(vp, &vattr, cred); 2761 if (error) { 2762 VOP_UNLOCK(vp); 2763 goto out; 2764 } 2765 newoff = SCARG(uap, offset) + vattr.va_size; 2766 break; 2767 case SEEK_SET: 2768 newoff = SCARG(uap, offset); 2769 break; 2770 default: 2771 error = EINVAL; 2772 VOP_UNLOCK(vp); 2773 goto out; 2774 } 2775 VOP_UNLOCK(vp); 2776 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2777 *(off_t *)retval = fp->f_offset = newoff; 2778 } 2779 out: 2780 fd_putfile(fd); 2781 return (error); 2782 } 2783 2784 /* 2785 * Positional read system call. 2786 */ 2787 int 2788 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2789 { 2790 /* { 2791 syscallarg(int) fd; 2792 syscallarg(void *) buf; 2793 syscallarg(size_t) nbyte; 2794 syscallarg(off_t) offset; 2795 } */ 2796 file_t *fp; 2797 struct vnode *vp; 2798 off_t offset; 2799 int error, fd = SCARG(uap, fd); 2800 2801 if ((fp = fd_getfile(fd)) == NULL) 2802 return (EBADF); 2803 2804 if ((fp->f_flag & FREAD) == 0) { 2805 fd_putfile(fd); 2806 return (EBADF); 2807 } 2808 2809 vp = fp->f_vnode; 2810 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2811 error = ESPIPE; 2812 goto out; 2813 } 2814 2815 offset = SCARG(uap, offset); 2816 2817 /* 2818 * XXX This works because no file systems actually 2819 * XXX take any action on the seek operation. 2820 */ 2821 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2822 goto out; 2823 2824 /* dofileread() will unuse the descriptor for us */ 2825 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2826 &offset, 0, retval)); 2827 2828 out: 2829 fd_putfile(fd); 2830 return (error); 2831 } 2832 2833 /* 2834 * Positional scatter read system call. 2835 */ 2836 int 2837 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2838 { 2839 /* { 2840 syscallarg(int) fd; 2841 syscallarg(const struct iovec *) iovp; 2842 syscallarg(int) iovcnt; 2843 syscallarg(off_t) offset; 2844 } */ 2845 off_t offset = SCARG(uap, offset); 2846 2847 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2848 SCARG(uap, iovcnt), &offset, 0, retval); 2849 } 2850 2851 /* 2852 * Positional write system call. 2853 */ 2854 int 2855 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2856 { 2857 /* { 2858 syscallarg(int) fd; 2859 syscallarg(const void *) buf; 2860 syscallarg(size_t) nbyte; 2861 syscallarg(off_t) offset; 2862 } */ 2863 file_t *fp; 2864 struct vnode *vp; 2865 off_t offset; 2866 int error, fd = SCARG(uap, fd); 2867 2868 if ((fp = fd_getfile(fd)) == NULL) 2869 return (EBADF); 2870 2871 if ((fp->f_flag & FWRITE) == 0) { 2872 fd_putfile(fd); 2873 return (EBADF); 2874 } 2875 2876 vp = fp->f_vnode; 2877 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2878 error = ESPIPE; 2879 goto out; 2880 } 2881 2882 offset = SCARG(uap, offset); 2883 2884 /* 2885 * XXX This works because no file systems actually 2886 * XXX take any action on the seek operation. 2887 */ 2888 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2889 goto out; 2890 2891 /* dofilewrite() will unuse the descriptor for us */ 2892 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2893 &offset, 0, retval)); 2894 2895 out: 2896 fd_putfile(fd); 2897 return (error); 2898 } 2899 2900 /* 2901 * Positional gather write system call. 2902 */ 2903 int 2904 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2905 { 2906 /* { 2907 syscallarg(int) fd; 2908 syscallarg(const struct iovec *) iovp; 2909 syscallarg(int) iovcnt; 2910 syscallarg(off_t) offset; 2911 } */ 2912 off_t offset = SCARG(uap, offset); 2913 2914 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2915 SCARG(uap, iovcnt), &offset, 0, retval); 2916 } 2917 2918 /* 2919 * Check access permissions. 2920 */ 2921 int 2922 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2923 { 2924 /* { 2925 syscallarg(const char *) path; 2926 syscallarg(int) flags; 2927 } */ 2928 2929 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2930 SCARG(uap, flags), 0); 2931 } 2932 2933 int 2934 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2935 int mode, int flags) 2936 { 2937 kauth_cred_t cred; 2938 struct vnode *vp; 2939 int error, nd_flag, vmode; 2940 struct pathbuf *pb; 2941 struct nameidata nd; 2942 2943 CTASSERT(F_OK == 0); 2944 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2945 /* nonsense mode */ 2946 return EINVAL; 2947 } 2948 2949 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2950 if (flags & AT_SYMLINK_NOFOLLOW) 2951 nd_flag &= ~FOLLOW; 2952 2953 error = pathbuf_copyin(path, &pb); 2954 if (error) 2955 return error; 2956 2957 NDINIT(&nd, LOOKUP, nd_flag, pb); 2958 2959 /* Override default credentials */ 2960 cred = kauth_cred_dup(l->l_cred); 2961 if (!(flags & AT_EACCESS)) { 2962 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2963 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2964 } 2965 nd.ni_cnd.cn_cred = cred; 2966 2967 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2968 pathbuf_destroy(pb); 2969 goto out; 2970 } 2971 vp = nd.ni_vp; 2972 pathbuf_destroy(pb); 2973 2974 /* Flags == 0 means only check for existence. */ 2975 if (mode) { 2976 vmode = 0; 2977 if (mode & R_OK) 2978 vmode |= VREAD; 2979 if (mode & W_OK) 2980 vmode |= VWRITE; 2981 if (mode & X_OK) 2982 vmode |= VEXEC; 2983 2984 error = VOP_ACCESS(vp, vmode, cred); 2985 if (!error && (vmode & VWRITE)) 2986 error = vn_writechk(vp); 2987 } 2988 vput(vp); 2989 out: 2990 kauth_cred_free(cred); 2991 return (error); 2992 } 2993 2994 int 2995 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 2996 register_t *retval) 2997 { 2998 /* { 2999 syscallarg(int) fd; 3000 syscallarg(const char *) path; 3001 syscallarg(int) amode; 3002 syscallarg(int) flag; 3003 } */ 3004 3005 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3006 SCARG(uap, amode), SCARG(uap, flag)); 3007 } 3008 3009 /* 3010 * Common code for all sys_stat functions, including compat versions. 3011 */ 3012 int 3013 do_sys_stat(const char *userpath, unsigned int nd_flag, 3014 struct stat *sb) 3015 { 3016 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3017 } 3018 3019 int 3020 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3021 unsigned int nd_flag, struct stat *sb) 3022 { 3023 int error; 3024 struct pathbuf *pb; 3025 struct nameidata nd; 3026 3027 KASSERT(l != NULL || fdat == AT_FDCWD); 3028 3029 error = pathbuf_copyin(userpath, &pb); 3030 if (error) { 3031 return error; 3032 } 3033 3034 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3035 3036 error = fd_nameiat(l, fdat, &nd); 3037 if (error != 0) { 3038 pathbuf_destroy(pb); 3039 return error; 3040 } 3041 error = vn_stat(nd.ni_vp, sb); 3042 vput(nd.ni_vp); 3043 pathbuf_destroy(pb); 3044 return error; 3045 } 3046 3047 /* 3048 * Get file status; this version follows links. 3049 */ 3050 /* ARGSUSED */ 3051 int 3052 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3053 { 3054 /* { 3055 syscallarg(const char *) path; 3056 syscallarg(struct stat *) ub; 3057 } */ 3058 struct stat sb; 3059 int error; 3060 3061 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3062 if (error) 3063 return error; 3064 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3065 } 3066 3067 /* 3068 * Get file status; this version does not follow links. 3069 */ 3070 /* ARGSUSED */ 3071 int 3072 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3073 { 3074 /* { 3075 syscallarg(const char *) path; 3076 syscallarg(struct stat *) ub; 3077 } */ 3078 struct stat sb; 3079 int error; 3080 3081 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3082 if (error) 3083 return error; 3084 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3085 } 3086 3087 int 3088 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3089 register_t *retval) 3090 { 3091 /* { 3092 syscallarg(int) fd; 3093 syscallarg(const char *) path; 3094 syscallarg(struct stat *) buf; 3095 syscallarg(int) flag; 3096 } */ 3097 unsigned int nd_flag; 3098 struct stat sb; 3099 int error; 3100 3101 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3102 nd_flag = NOFOLLOW; 3103 else 3104 nd_flag = FOLLOW; 3105 3106 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3107 &sb); 3108 if (error) 3109 return error; 3110 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3111 } 3112 3113 /* 3114 * Get configurable pathname variables. 3115 */ 3116 /* ARGSUSED */ 3117 int 3118 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3119 { 3120 /* { 3121 syscallarg(const char *) path; 3122 syscallarg(int) name; 3123 } */ 3124 int error; 3125 struct pathbuf *pb; 3126 struct nameidata nd; 3127 3128 error = pathbuf_copyin(SCARG(uap, path), &pb); 3129 if (error) { 3130 return error; 3131 } 3132 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3133 if ((error = namei(&nd)) != 0) { 3134 pathbuf_destroy(pb); 3135 return (error); 3136 } 3137 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3138 vput(nd.ni_vp); 3139 pathbuf_destroy(pb); 3140 return (error); 3141 } 3142 3143 /* 3144 * Return target name of a symbolic link. 3145 */ 3146 /* ARGSUSED */ 3147 int 3148 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3149 register_t *retval) 3150 { 3151 /* { 3152 syscallarg(const char *) path; 3153 syscallarg(char *) buf; 3154 syscallarg(size_t) count; 3155 } */ 3156 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3157 SCARG(uap, buf), SCARG(uap, count), retval); 3158 } 3159 3160 static int 3161 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3162 size_t count, register_t *retval) 3163 { 3164 struct vnode *vp; 3165 struct iovec aiov; 3166 struct uio auio; 3167 int error; 3168 struct pathbuf *pb; 3169 struct nameidata nd; 3170 3171 error = pathbuf_copyin(path, &pb); 3172 if (error) { 3173 return error; 3174 } 3175 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3176 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3177 pathbuf_destroy(pb); 3178 return error; 3179 } 3180 vp = nd.ni_vp; 3181 pathbuf_destroy(pb); 3182 if (vp->v_type != VLNK) 3183 error = EINVAL; 3184 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3185 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3186 aiov.iov_base = buf; 3187 aiov.iov_len = count; 3188 auio.uio_iov = &aiov; 3189 auio.uio_iovcnt = 1; 3190 auio.uio_offset = 0; 3191 auio.uio_rw = UIO_READ; 3192 KASSERT(l == curlwp); 3193 auio.uio_vmspace = l->l_proc->p_vmspace; 3194 auio.uio_resid = count; 3195 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3196 *retval = count - auio.uio_resid; 3197 } 3198 vput(vp); 3199 return (error); 3200 } 3201 3202 int 3203 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3204 register_t *retval) 3205 { 3206 /* { 3207 syscallarg(int) fd; 3208 syscallarg(const char *) path; 3209 syscallarg(char *) buf; 3210 syscallarg(size_t) bufsize; 3211 } */ 3212 3213 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3214 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3215 } 3216 3217 /* 3218 * Change flags of a file given a path name. 3219 */ 3220 /* ARGSUSED */ 3221 int 3222 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3223 { 3224 /* { 3225 syscallarg(const char *) path; 3226 syscallarg(u_long) flags; 3227 } */ 3228 struct vnode *vp; 3229 int error; 3230 3231 error = namei_simple_user(SCARG(uap, path), 3232 NSM_FOLLOW_TRYEMULROOT, &vp); 3233 if (error != 0) 3234 return (error); 3235 error = change_flags(vp, SCARG(uap, flags), l); 3236 vput(vp); 3237 return (error); 3238 } 3239 3240 /* 3241 * Change flags of a file given a file descriptor. 3242 */ 3243 /* ARGSUSED */ 3244 int 3245 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3246 { 3247 /* { 3248 syscallarg(int) fd; 3249 syscallarg(u_long) flags; 3250 } */ 3251 struct vnode *vp; 3252 file_t *fp; 3253 int error; 3254 3255 /* fd_getvnode() will use the descriptor for us */ 3256 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3257 return (error); 3258 vp = fp->f_vnode; 3259 error = change_flags(vp, SCARG(uap, flags), l); 3260 VOP_UNLOCK(vp); 3261 fd_putfile(SCARG(uap, fd)); 3262 return (error); 3263 } 3264 3265 /* 3266 * Change flags of a file given a path name; this version does 3267 * not follow links. 3268 */ 3269 int 3270 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3271 { 3272 /* { 3273 syscallarg(const char *) path; 3274 syscallarg(u_long) flags; 3275 } */ 3276 struct vnode *vp; 3277 int error; 3278 3279 error = namei_simple_user(SCARG(uap, path), 3280 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3281 if (error != 0) 3282 return (error); 3283 error = change_flags(vp, SCARG(uap, flags), l); 3284 vput(vp); 3285 return (error); 3286 } 3287 3288 /* 3289 * Common routine to change flags of a file. 3290 */ 3291 int 3292 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3293 { 3294 struct vattr vattr; 3295 int error; 3296 3297 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3298 3299 vattr_null(&vattr); 3300 vattr.va_flags = flags; 3301 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3302 3303 return (error); 3304 } 3305 3306 /* 3307 * Change mode of a file given path name; this version follows links. 3308 */ 3309 /* ARGSUSED */ 3310 int 3311 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3312 { 3313 /* { 3314 syscallarg(const char *) path; 3315 syscallarg(int) mode; 3316 } */ 3317 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3318 SCARG(uap, mode), 0); 3319 } 3320 3321 int 3322 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3323 { 3324 int error; 3325 struct vnode *vp; 3326 namei_simple_flags_t ns_flag; 3327 3328 if (flags & AT_SYMLINK_NOFOLLOW) 3329 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3330 else 3331 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3332 3333 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3334 if (error != 0) 3335 return error; 3336 3337 error = change_mode(vp, mode, l); 3338 3339 vrele(vp); 3340 3341 return (error); 3342 } 3343 3344 /* 3345 * Change mode of a file given a file descriptor. 3346 */ 3347 /* ARGSUSED */ 3348 int 3349 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3350 { 3351 /* { 3352 syscallarg(int) fd; 3353 syscallarg(int) mode; 3354 } */ 3355 file_t *fp; 3356 int error; 3357 3358 /* fd_getvnode() will use the descriptor for us */ 3359 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3360 return (error); 3361 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3362 fd_putfile(SCARG(uap, fd)); 3363 return (error); 3364 } 3365 3366 int 3367 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3368 register_t *retval) 3369 { 3370 /* { 3371 syscallarg(int) fd; 3372 syscallarg(const char *) path; 3373 syscallarg(int) mode; 3374 syscallarg(int) flag; 3375 } */ 3376 3377 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3378 SCARG(uap, mode), SCARG(uap, flag)); 3379 } 3380 3381 /* 3382 * Change mode of a file given path name; this version does not follow links. 3383 */ 3384 /* ARGSUSED */ 3385 int 3386 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3387 { 3388 /* { 3389 syscallarg(const char *) path; 3390 syscallarg(int) mode; 3391 } */ 3392 int error; 3393 struct vnode *vp; 3394 3395 error = namei_simple_user(SCARG(uap, path), 3396 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3397 if (error != 0) 3398 return (error); 3399 3400 error = change_mode(vp, SCARG(uap, mode), l); 3401 3402 vrele(vp); 3403 return (error); 3404 } 3405 3406 /* 3407 * Common routine to set mode given a vnode. 3408 */ 3409 static int 3410 change_mode(struct vnode *vp, int mode, struct lwp *l) 3411 { 3412 struct vattr vattr; 3413 int error; 3414 3415 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3416 vattr_null(&vattr); 3417 vattr.va_mode = mode & ALLPERMS; 3418 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3419 VOP_UNLOCK(vp); 3420 return (error); 3421 } 3422 3423 /* 3424 * Set ownership given a path name; this version follows links. 3425 */ 3426 /* ARGSUSED */ 3427 int 3428 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3429 { 3430 /* { 3431 syscallarg(const char *) path; 3432 syscallarg(uid_t) uid; 3433 syscallarg(gid_t) gid; 3434 } */ 3435 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3436 SCARG(uap, gid), 0); 3437 } 3438 3439 int 3440 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3441 gid_t gid, int flags) 3442 { 3443 int error; 3444 struct vnode *vp; 3445 namei_simple_flags_t ns_flag; 3446 3447 if (flags & AT_SYMLINK_NOFOLLOW) 3448 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3449 else 3450 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3451 3452 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3453 if (error != 0) 3454 return error; 3455 3456 error = change_owner(vp, uid, gid, l, 0); 3457 3458 vrele(vp); 3459 3460 return (error); 3461 } 3462 3463 /* 3464 * Set ownership given a path name; this version follows links. 3465 * Provides POSIX semantics. 3466 */ 3467 /* ARGSUSED */ 3468 int 3469 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3470 { 3471 /* { 3472 syscallarg(const char *) path; 3473 syscallarg(uid_t) uid; 3474 syscallarg(gid_t) gid; 3475 } */ 3476 int error; 3477 struct vnode *vp; 3478 3479 error = namei_simple_user(SCARG(uap, path), 3480 NSM_FOLLOW_TRYEMULROOT, &vp); 3481 if (error != 0) 3482 return (error); 3483 3484 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3485 3486 vrele(vp); 3487 return (error); 3488 } 3489 3490 /* 3491 * Set ownership given a file descriptor. 3492 */ 3493 /* ARGSUSED */ 3494 int 3495 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3496 { 3497 /* { 3498 syscallarg(int) fd; 3499 syscallarg(uid_t) uid; 3500 syscallarg(gid_t) gid; 3501 } */ 3502 int error; 3503 file_t *fp; 3504 3505 /* fd_getvnode() will use the descriptor for us */ 3506 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3507 return (error); 3508 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3509 l, 0); 3510 fd_putfile(SCARG(uap, fd)); 3511 return (error); 3512 } 3513 3514 int 3515 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3516 register_t *retval) 3517 { 3518 /* { 3519 syscallarg(int) fd; 3520 syscallarg(const char *) path; 3521 syscallarg(uid_t) owner; 3522 syscallarg(gid_t) group; 3523 syscallarg(int) flag; 3524 } */ 3525 3526 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3527 SCARG(uap, owner), SCARG(uap, group), 3528 SCARG(uap, flag)); 3529 } 3530 3531 /* 3532 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3533 */ 3534 /* ARGSUSED */ 3535 int 3536 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3537 { 3538 /* { 3539 syscallarg(int) fd; 3540 syscallarg(uid_t) uid; 3541 syscallarg(gid_t) gid; 3542 } */ 3543 int error; 3544 file_t *fp; 3545 3546 /* fd_getvnode() will use the descriptor for us */ 3547 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3548 return (error); 3549 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3550 l, 1); 3551 fd_putfile(SCARG(uap, fd)); 3552 return (error); 3553 } 3554 3555 /* 3556 * Set ownership given a path name; this version does not follow links. 3557 */ 3558 /* ARGSUSED */ 3559 int 3560 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3561 { 3562 /* { 3563 syscallarg(const char *) path; 3564 syscallarg(uid_t) uid; 3565 syscallarg(gid_t) gid; 3566 } */ 3567 int error; 3568 struct vnode *vp; 3569 3570 error = namei_simple_user(SCARG(uap, path), 3571 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3572 if (error != 0) 3573 return (error); 3574 3575 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3576 3577 vrele(vp); 3578 return (error); 3579 } 3580 3581 /* 3582 * Set ownership given a path name; this version does not follow links. 3583 * Provides POSIX/XPG semantics. 3584 */ 3585 /* ARGSUSED */ 3586 int 3587 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3588 { 3589 /* { 3590 syscallarg(const char *) path; 3591 syscallarg(uid_t) uid; 3592 syscallarg(gid_t) gid; 3593 } */ 3594 int error; 3595 struct vnode *vp; 3596 3597 error = namei_simple_user(SCARG(uap, path), 3598 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3599 if (error != 0) 3600 return (error); 3601 3602 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3603 3604 vrele(vp); 3605 return (error); 3606 } 3607 3608 /* 3609 * Common routine to set ownership given a vnode. 3610 */ 3611 static int 3612 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3613 int posix_semantics) 3614 { 3615 struct vattr vattr; 3616 mode_t newmode; 3617 int error; 3618 3619 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3620 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3621 goto out; 3622 3623 #define CHANGED(x) ((int)(x) != -1) 3624 newmode = vattr.va_mode; 3625 if (posix_semantics) { 3626 /* 3627 * POSIX/XPG semantics: if the caller is not the super-user, 3628 * clear set-user-id and set-group-id bits. Both POSIX and 3629 * the XPG consider the behaviour for calls by the super-user 3630 * implementation-defined; we leave the set-user-id and set- 3631 * group-id settings intact in that case. 3632 */ 3633 if (vattr.va_mode & S_ISUID) { 3634 if (kauth_authorize_vnode(l->l_cred, 3635 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3636 newmode &= ~S_ISUID; 3637 } 3638 if (vattr.va_mode & S_ISGID) { 3639 if (kauth_authorize_vnode(l->l_cred, 3640 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3641 newmode &= ~S_ISGID; 3642 } 3643 } else { 3644 /* 3645 * NetBSD semantics: when changing owner and/or group, 3646 * clear the respective bit(s). 3647 */ 3648 if (CHANGED(uid)) 3649 newmode &= ~S_ISUID; 3650 if (CHANGED(gid)) 3651 newmode &= ~S_ISGID; 3652 } 3653 /* Update va_mode iff altered. */ 3654 if (vattr.va_mode == newmode) 3655 newmode = VNOVAL; 3656 3657 vattr_null(&vattr); 3658 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3659 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3660 vattr.va_mode = newmode; 3661 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3662 #undef CHANGED 3663 3664 out: 3665 VOP_UNLOCK(vp); 3666 return (error); 3667 } 3668 3669 /* 3670 * Set the access and modification times given a path name; this 3671 * version follows links. 3672 */ 3673 /* ARGSUSED */ 3674 int 3675 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3676 register_t *retval) 3677 { 3678 /* { 3679 syscallarg(const char *) path; 3680 syscallarg(const struct timeval *) tptr; 3681 } */ 3682 3683 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3684 SCARG(uap, tptr), UIO_USERSPACE); 3685 } 3686 3687 /* 3688 * Set the access and modification times given a file descriptor. 3689 */ 3690 /* ARGSUSED */ 3691 int 3692 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3693 register_t *retval) 3694 { 3695 /* { 3696 syscallarg(int) fd; 3697 syscallarg(const struct timeval *) tptr; 3698 } */ 3699 int error; 3700 file_t *fp; 3701 3702 /* fd_getvnode() will use the descriptor for us */ 3703 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3704 return (error); 3705 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3706 UIO_USERSPACE); 3707 fd_putfile(SCARG(uap, fd)); 3708 return (error); 3709 } 3710 3711 int 3712 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3713 register_t *retval) 3714 { 3715 /* { 3716 syscallarg(int) fd; 3717 syscallarg(const struct timespec *) tptr; 3718 } */ 3719 int error; 3720 file_t *fp; 3721 3722 /* fd_getvnode() will use the descriptor for us */ 3723 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3724 return (error); 3725 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3726 SCARG(uap, tptr), UIO_USERSPACE); 3727 fd_putfile(SCARG(uap, fd)); 3728 return (error); 3729 } 3730 3731 /* 3732 * Set the access and modification times given a path name; this 3733 * version does not follow links. 3734 */ 3735 int 3736 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3737 register_t *retval) 3738 { 3739 /* { 3740 syscallarg(const char *) path; 3741 syscallarg(const struct timeval *) tptr; 3742 } */ 3743 3744 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3745 SCARG(uap, tptr), UIO_USERSPACE); 3746 } 3747 3748 int 3749 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3750 register_t *retval) 3751 { 3752 /* { 3753 syscallarg(int) fd; 3754 syscallarg(const char *) path; 3755 syscallarg(const struct timespec *) tptr; 3756 syscallarg(int) flag; 3757 } */ 3758 int follow; 3759 const struct timespec *tptr; 3760 int error; 3761 3762 tptr = SCARG(uap, tptr); 3763 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3764 3765 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3766 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3767 3768 return error; 3769 } 3770 3771 /* 3772 * Common routine to set access and modification times given a vnode. 3773 */ 3774 int 3775 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3776 const struct timespec *tptr, enum uio_seg seg) 3777 { 3778 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3779 } 3780 3781 int 3782 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3783 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3784 { 3785 struct vattr vattr; 3786 int error, dorele = 0; 3787 namei_simple_flags_t sflags; 3788 bool vanull, setbirthtime; 3789 struct timespec ts[2]; 3790 3791 KASSERT(l != NULL || fdat == AT_FDCWD); 3792 3793 /* 3794 * I have checked all callers and they pass either FOLLOW, 3795 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3796 * is 0. More to the point, they don't pass anything else. 3797 * Let's keep it that way at least until the namei interfaces 3798 * are fully sanitized. 3799 */ 3800 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3801 sflags = (flag == FOLLOW) ? 3802 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3803 3804 if (tptr == NULL) { 3805 vanull = true; 3806 nanotime(&ts[0]); 3807 ts[1] = ts[0]; 3808 } else { 3809 vanull = false; 3810 if (seg != UIO_SYSSPACE) { 3811 error = copyin(tptr, ts, sizeof (ts)); 3812 if (error != 0) 3813 return error; 3814 } else { 3815 ts[0] = tptr[0]; 3816 ts[1] = tptr[1]; 3817 } 3818 } 3819 3820 if (ts[0].tv_nsec == UTIME_NOW) { 3821 nanotime(&ts[0]); 3822 if (ts[1].tv_nsec == UTIME_NOW) { 3823 vanull = true; 3824 ts[1] = ts[0]; 3825 } 3826 } else if (ts[1].tv_nsec == UTIME_NOW) 3827 nanotime(&ts[1]); 3828 3829 if (vp == NULL) { 3830 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3831 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3832 if (error != 0) 3833 return error; 3834 dorele = 1; 3835 } 3836 3837 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3838 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3839 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3840 vattr_null(&vattr); 3841 3842 if (ts[0].tv_nsec != UTIME_OMIT) 3843 vattr.va_atime = ts[0]; 3844 3845 if (ts[1].tv_nsec != UTIME_OMIT) { 3846 vattr.va_mtime = ts[1]; 3847 if (setbirthtime) 3848 vattr.va_birthtime = ts[1]; 3849 } 3850 3851 if (vanull) 3852 vattr.va_vaflags |= VA_UTIMES_NULL; 3853 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3854 VOP_UNLOCK(vp); 3855 3856 if (dorele != 0) 3857 vrele(vp); 3858 3859 return error; 3860 } 3861 3862 int 3863 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3864 const struct timeval *tptr, enum uio_seg seg) 3865 { 3866 struct timespec ts[2]; 3867 struct timespec *tsptr = NULL; 3868 int error; 3869 3870 if (tptr != NULL) { 3871 struct timeval tv[2]; 3872 3873 if (seg != UIO_SYSSPACE) { 3874 error = copyin(tptr, tv, sizeof (tv)); 3875 if (error != 0) 3876 return error; 3877 tptr = tv; 3878 } 3879 3880 if ((tv[0].tv_usec == UTIME_NOW) || 3881 (tv[0].tv_usec == UTIME_OMIT)) 3882 ts[0].tv_nsec = tv[0].tv_usec; 3883 else 3884 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3885 3886 if ((tv[1].tv_usec == UTIME_NOW) || 3887 (tv[1].tv_usec == UTIME_OMIT)) 3888 ts[1].tv_nsec = tv[1].tv_usec; 3889 else 3890 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3891 3892 tsptr = &ts[0]; 3893 } 3894 3895 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3896 } 3897 3898 /* 3899 * Truncate a file given its path name. 3900 */ 3901 /* ARGSUSED */ 3902 int 3903 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3904 { 3905 /* { 3906 syscallarg(const char *) path; 3907 syscallarg(int) pad; 3908 syscallarg(off_t) length; 3909 } */ 3910 struct vnode *vp; 3911 struct vattr vattr; 3912 int error; 3913 3914 if (SCARG(uap, length) < 0) 3915 return EINVAL; 3916 3917 error = namei_simple_user(SCARG(uap, path), 3918 NSM_FOLLOW_TRYEMULROOT, &vp); 3919 if (error != 0) 3920 return (error); 3921 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3922 if (vp->v_type == VDIR) 3923 error = EISDIR; 3924 else if ((error = vn_writechk(vp)) == 0 && 3925 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3926 vattr_null(&vattr); 3927 vattr.va_size = SCARG(uap, length); 3928 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3929 } 3930 vput(vp); 3931 return (error); 3932 } 3933 3934 /* 3935 * Truncate a file given a file descriptor. 3936 */ 3937 /* ARGSUSED */ 3938 int 3939 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3940 { 3941 /* { 3942 syscallarg(int) fd; 3943 syscallarg(int) pad; 3944 syscallarg(off_t) length; 3945 } */ 3946 struct vattr vattr; 3947 struct vnode *vp; 3948 file_t *fp; 3949 int error; 3950 3951 if (SCARG(uap, length) < 0) 3952 return EINVAL; 3953 3954 /* fd_getvnode() will use the descriptor for us */ 3955 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3956 return (error); 3957 if ((fp->f_flag & FWRITE) == 0) { 3958 error = EINVAL; 3959 goto out; 3960 } 3961 vp = fp->f_vnode; 3962 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3963 if (vp->v_type == VDIR) 3964 error = EISDIR; 3965 else if ((error = vn_writechk(vp)) == 0) { 3966 vattr_null(&vattr); 3967 vattr.va_size = SCARG(uap, length); 3968 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3969 } 3970 VOP_UNLOCK(vp); 3971 out: 3972 fd_putfile(SCARG(uap, fd)); 3973 return (error); 3974 } 3975 3976 /* 3977 * Sync an open file. 3978 */ 3979 /* ARGSUSED */ 3980 int 3981 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3982 { 3983 /* { 3984 syscallarg(int) fd; 3985 } */ 3986 struct vnode *vp; 3987 file_t *fp; 3988 int error; 3989 3990 /* fd_getvnode() will use the descriptor for us */ 3991 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3992 return (error); 3993 vp = fp->f_vnode; 3994 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3995 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3996 VOP_UNLOCK(vp); 3997 fd_putfile(SCARG(uap, fd)); 3998 return (error); 3999 } 4000 4001 /* 4002 * Sync a range of file data. API modeled after that found in AIX. 4003 * 4004 * FDATASYNC indicates that we need only save enough metadata to be able 4005 * to re-read the written data. 4006 */ 4007 /* ARGSUSED */ 4008 int 4009 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4010 { 4011 /* { 4012 syscallarg(int) fd; 4013 syscallarg(int) flags; 4014 syscallarg(off_t) start; 4015 syscallarg(off_t) length; 4016 } */ 4017 struct vnode *vp; 4018 file_t *fp; 4019 int flags, nflags; 4020 off_t s, e, len; 4021 int error; 4022 4023 /* fd_getvnode() will use the descriptor for us */ 4024 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4025 return (error); 4026 4027 if ((fp->f_flag & FWRITE) == 0) { 4028 error = EBADF; 4029 goto out; 4030 } 4031 4032 flags = SCARG(uap, flags); 4033 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4034 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4035 error = EINVAL; 4036 goto out; 4037 } 4038 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4039 if (flags & FDATASYNC) 4040 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4041 else 4042 nflags = FSYNC_WAIT; 4043 if (flags & FDISKSYNC) 4044 nflags |= FSYNC_CACHE; 4045 4046 len = SCARG(uap, length); 4047 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4048 if (len) { 4049 s = SCARG(uap, start); 4050 e = s + len; 4051 if (e < s) { 4052 error = EINVAL; 4053 goto out; 4054 } 4055 } else { 4056 e = 0; 4057 s = 0; 4058 } 4059 4060 vp = fp->f_vnode; 4061 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4062 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4063 VOP_UNLOCK(vp); 4064 out: 4065 fd_putfile(SCARG(uap, fd)); 4066 return (error); 4067 } 4068 4069 /* 4070 * Sync the data of an open file. 4071 */ 4072 /* ARGSUSED */ 4073 int 4074 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4075 { 4076 /* { 4077 syscallarg(int) fd; 4078 } */ 4079 struct vnode *vp; 4080 file_t *fp; 4081 int error; 4082 4083 /* fd_getvnode() will use the descriptor for us */ 4084 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4085 return (error); 4086 vp = fp->f_vnode; 4087 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4088 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4089 VOP_UNLOCK(vp); 4090 fd_putfile(SCARG(uap, fd)); 4091 return (error); 4092 } 4093 4094 /* 4095 * Rename files, (standard) BSD semantics frontend. 4096 */ 4097 /* ARGSUSED */ 4098 int 4099 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4100 { 4101 /* { 4102 syscallarg(const char *) from; 4103 syscallarg(const char *) to; 4104 } */ 4105 4106 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4107 SCARG(uap, to), UIO_USERSPACE, 0)); 4108 } 4109 4110 int 4111 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4112 register_t *retval) 4113 { 4114 /* { 4115 syscallarg(int) fromfd; 4116 syscallarg(const char *) from; 4117 syscallarg(int) tofd; 4118 syscallarg(const char *) to; 4119 } */ 4120 4121 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4122 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4123 } 4124 4125 /* 4126 * Rename files, POSIX semantics frontend. 4127 */ 4128 /* ARGSUSED */ 4129 int 4130 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4131 { 4132 /* { 4133 syscallarg(const char *) from; 4134 syscallarg(const char *) to; 4135 } */ 4136 4137 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4138 SCARG(uap, to), UIO_USERSPACE, 1)); 4139 } 4140 4141 /* 4142 * Rename files. Source and destination must either both be directories, 4143 * or both not be directories. If target is a directory, it must be empty. 4144 * If `from' and `to' refer to the same object, the value of the `retain' 4145 * argument is used to determine whether `from' will be 4146 * 4147 * (retain == 0) deleted unless `from' and `to' refer to the same 4148 * object in the file system's name space (BSD). 4149 * (retain == 1) always retained (POSIX). 4150 * 4151 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4152 */ 4153 int 4154 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4155 { 4156 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4157 } 4158 4159 static int 4160 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4161 const char *to, enum uio_seg seg, int retain) 4162 { 4163 struct pathbuf *fpb, *tpb; 4164 struct nameidata fnd, tnd; 4165 struct vnode *fdvp, *fvp; 4166 struct vnode *tdvp, *tvp; 4167 struct mount *mp, *tmp; 4168 int error; 4169 4170 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4171 4172 error = pathbuf_maybe_copyin(from, seg, &fpb); 4173 if (error) 4174 goto out0; 4175 KASSERT(fpb != NULL); 4176 4177 error = pathbuf_maybe_copyin(to, seg, &tpb); 4178 if (error) 4179 goto out1; 4180 KASSERT(tpb != NULL); 4181 4182 /* 4183 * Lookup from. 4184 * 4185 * XXX LOCKPARENT is wrong because we don't actually want it 4186 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4187 * insane, so for the time being we need to leave it like this. 4188 */ 4189 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4190 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4191 goto out2; 4192 4193 /* 4194 * Pull out the important results of the lookup, fdvp and fvp. 4195 * Of course, fvp is bogus because we're about to unlock fdvp. 4196 */ 4197 fdvp = fnd.ni_dvp; 4198 fvp = fnd.ni_vp; 4199 KASSERT(fdvp != NULL); 4200 KASSERT(fvp != NULL); 4201 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4202 4203 /* 4204 * Make sure neither fdvp nor fvp is locked. 4205 */ 4206 if (fdvp != fvp) 4207 VOP_UNLOCK(fdvp); 4208 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4209 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4210 4211 /* 4212 * Reject renaming `.' and `..'. Can't do this until after 4213 * namei because we need namei's parsing to find the final 4214 * component name. (namei should just leave us with the final 4215 * component name and not look it up itself, but anyway...) 4216 * 4217 * This was here before because we used to relookup from 4218 * instead of to and relookup requires the caller to check 4219 * this, but now file systems may depend on this check, so we 4220 * must retain it until the file systems are all rototilled. 4221 */ 4222 if (((fnd.ni_cnd.cn_namelen == 1) && 4223 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4224 ((fnd.ni_cnd.cn_namelen == 2) && 4225 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4226 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4227 error = EINVAL; /* XXX EISDIR? */ 4228 goto abort0; 4229 } 4230 4231 /* 4232 * Lookup to. 4233 * 4234 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4235 * fvp here to decide whether to add CREATEDIR is a load of 4236 * bollocks because fvp might be the wrong node by now, since 4237 * fdvp is unlocked. 4238 * 4239 * XXX Why not pass CREATEDIR always? 4240 */ 4241 NDINIT(&tnd, RENAME, 4242 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4243 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4244 tpb); 4245 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4246 goto abort0; 4247 4248 /* 4249 * Pull out the important results of the lookup, tdvp and tvp. 4250 * Of course, tvp is bogus because we're about to unlock tdvp. 4251 */ 4252 tdvp = tnd.ni_dvp; 4253 tvp = tnd.ni_vp; 4254 KASSERT(tdvp != NULL); 4255 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4256 4257 /* 4258 * Make sure neither tdvp nor tvp is locked. 4259 */ 4260 if (tdvp != tvp) 4261 VOP_UNLOCK(tdvp); 4262 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4263 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4264 4265 /* 4266 * Reject renaming onto `.' or `..'. relookup is unhappy with 4267 * these, which is why we must do this here. Once upon a time 4268 * we relooked up from instead of to, and consequently didn't 4269 * need this check, but now that we relookup to instead of 4270 * from, we need this; and we shall need it forever forward 4271 * until the VOP_RENAME protocol changes, because file systems 4272 * will no doubt begin to depend on this check. 4273 */ 4274 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4275 error = EISDIR; 4276 goto abort1; 4277 } 4278 if ((tnd.ni_cnd.cn_namelen == 2) && 4279 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4280 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4281 error = EINVAL; 4282 goto abort1; 4283 } 4284 4285 /* 4286 * Get the mount point. If the file system has been unmounted, 4287 * which it may be because we're not holding any vnode locks, 4288 * then v_mount will be NULL. We're not really supposed to 4289 * read v_mount without holding the vnode lock, but since we 4290 * have fdvp referenced, if fdvp->v_mount changes then at worst 4291 * it will be set to NULL, not changed to another mount point. 4292 * And, of course, since it is up to the file system to 4293 * determine the real lock order, we can't lock both fdvp and 4294 * tdvp at the same time. 4295 */ 4296 mp = fdvp->v_mount; 4297 if (mp == NULL) { 4298 error = ENOENT; 4299 goto abort1; 4300 } 4301 4302 /* 4303 * Make sure the mount points match. Again, although we don't 4304 * hold any vnode locks, the v_mount fields may change -- but 4305 * at worst they will change to NULL, so this will never become 4306 * a cross-device rename, because we hold vnode references. 4307 * 4308 * XXX Because nothing is locked and the compiler may reorder 4309 * things here, unmounting the file system at an inopportune 4310 * moment may cause rename to fail with EXDEV when it really 4311 * should fail with ENOENT. 4312 */ 4313 tmp = tdvp->v_mount; 4314 if (tmp == NULL) { 4315 error = ENOENT; 4316 goto abort1; 4317 } 4318 4319 if (mp != tmp) { 4320 error = EXDEV; 4321 goto abort1; 4322 } 4323 4324 /* 4325 * Take the vfs rename lock to avoid cross-directory screw cases. 4326 * Nothing is locked currently, so taking this lock is safe. 4327 */ 4328 error = VFS_RENAMELOCK_ENTER(mp); 4329 if (error) 4330 goto abort1; 4331 4332 /* 4333 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4334 * and nothing is locked except for the vfs rename lock. 4335 * 4336 * The next step is a little rain dance to conform to the 4337 * insane lock protocol, even though it does nothing to ward 4338 * off race conditions. 4339 * 4340 * We need tdvp and tvp to be locked. However, because we have 4341 * unlocked tdvp in order to hold no locks while we take the 4342 * vfs rename lock, tvp may be wrong here, and we can't safely 4343 * lock it even if the sensible file systems will just unlock 4344 * it straight away. Consequently, we must lock tdvp and then 4345 * relookup tvp to get it locked. 4346 * 4347 * Finally, because the VOP_RENAME protocol is brain-damaged 4348 * and various file systems insanely depend on the semantics of 4349 * this brain damage, the lookup of to must be the last lookup 4350 * before VOP_RENAME. 4351 */ 4352 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4353 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4354 if (error) 4355 goto abort2; 4356 4357 /* 4358 * Drop the old tvp and pick up the new one -- which might be 4359 * the same, but that doesn't matter to us. After this, tdvp 4360 * and tvp should both be locked. 4361 */ 4362 if (tvp != NULL) 4363 vrele(tvp); 4364 tvp = tnd.ni_vp; 4365 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4366 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4367 4368 /* 4369 * The old do_sys_rename had various consistency checks here 4370 * involving fvp and tvp. fvp is bogus already here, and tvp 4371 * will become bogus soon in any sensible file system, so the 4372 * only purpose in putting these checks here is to give lip 4373 * service to these screw cases and to acknowledge that they 4374 * exist, not actually to handle them, but here you go 4375 * anyway... 4376 */ 4377 4378 /* 4379 * Acknowledge that directories and non-directories aren't 4380 * suposed to mix. 4381 */ 4382 if (tvp != NULL) { 4383 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4384 error = ENOTDIR; 4385 goto abort3; 4386 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4387 error = EISDIR; 4388 goto abort3; 4389 } 4390 } 4391 4392 /* 4393 * Acknowledge some random screw case, among the dozens that 4394 * might arise. 4395 */ 4396 if (fvp == tdvp) { 4397 error = EINVAL; 4398 goto abort3; 4399 } 4400 4401 /* 4402 * Acknowledge that POSIX has a wacky screw case. 4403 * 4404 * XXX Eventually the retain flag needs to be passed on to 4405 * VOP_RENAME. 4406 */ 4407 if (fvp == tvp) { 4408 if (retain) { 4409 error = 0; 4410 goto abort3; 4411 } else if ((fdvp == tdvp) && 4412 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4413 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4414 fnd.ni_cnd.cn_namelen))) { 4415 error = 0; 4416 goto abort3; 4417 } 4418 } 4419 4420 /* 4421 * Make sure veriexec can screw us up. (But a race can screw 4422 * up veriexec, of course -- remember, fvp and (soon) tvp are 4423 * bogus.) 4424 */ 4425 #if NVERIEXEC > 0 4426 { 4427 char *f1, *f2; 4428 size_t f1_len; 4429 size_t f2_len; 4430 4431 f1_len = fnd.ni_cnd.cn_namelen + 1; 4432 f1 = kmem_alloc(f1_len, KM_SLEEP); 4433 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4434 4435 f2_len = tnd.ni_cnd.cn_namelen + 1; 4436 f2 = kmem_alloc(f2_len, KM_SLEEP); 4437 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4438 4439 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4440 4441 kmem_free(f1, f1_len); 4442 kmem_free(f2, f2_len); 4443 4444 if (error) 4445 goto abort3; 4446 } 4447 #endif /* NVERIEXEC > 0 */ 4448 4449 /* 4450 * All ready. Incant the rename vop. 4451 */ 4452 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4453 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4454 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4455 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4456 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4457 4458 /* 4459 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4460 * tdvp and tvp. But we can't assert any of that. 4461 */ 4462 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4463 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4464 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4465 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4466 4467 /* 4468 * So all we have left to do is to drop the rename lock and 4469 * destroy the pathbufs. 4470 */ 4471 VFS_RENAMELOCK_EXIT(mp); 4472 goto out2; 4473 4474 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4475 VOP_UNLOCK(tvp); 4476 abort2: VOP_UNLOCK(tdvp); 4477 VFS_RENAMELOCK_EXIT(mp); 4478 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4479 vrele(tdvp); 4480 if (tvp != NULL) 4481 vrele(tvp); 4482 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4483 vrele(fdvp); 4484 vrele(fvp); 4485 out2: pathbuf_destroy(tpb); 4486 out1: pathbuf_destroy(fpb); 4487 out0: return error; 4488 } 4489 4490 /* 4491 * Make a directory file. 4492 */ 4493 /* ARGSUSED */ 4494 int 4495 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4496 { 4497 /* { 4498 syscallarg(const char *) path; 4499 syscallarg(int) mode; 4500 } */ 4501 4502 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4503 SCARG(uap, mode), UIO_USERSPACE); 4504 } 4505 4506 int 4507 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4508 register_t *retval) 4509 { 4510 /* { 4511 syscallarg(int) fd; 4512 syscallarg(const char *) path; 4513 syscallarg(int) mode; 4514 } */ 4515 4516 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4517 SCARG(uap, mode), UIO_USERSPACE); 4518 } 4519 4520 4521 int 4522 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4523 { 4524 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4525 } 4526 4527 static int 4528 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4529 enum uio_seg seg) 4530 { 4531 struct proc *p = curlwp->l_proc; 4532 struct vnode *vp; 4533 struct vattr vattr; 4534 int error; 4535 struct pathbuf *pb; 4536 struct nameidata nd; 4537 4538 KASSERT(l != NULL || fdat == AT_FDCWD); 4539 4540 /* XXX bollocks, should pass in a pathbuf */ 4541 error = pathbuf_maybe_copyin(path, seg, &pb); 4542 if (error) { 4543 return error; 4544 } 4545 4546 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4547 4548 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4549 pathbuf_destroy(pb); 4550 return (error); 4551 } 4552 vp = nd.ni_vp; 4553 if (vp != NULL) { 4554 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4555 if (nd.ni_dvp == vp) 4556 vrele(nd.ni_dvp); 4557 else 4558 vput(nd.ni_dvp); 4559 vrele(vp); 4560 pathbuf_destroy(pb); 4561 return (EEXIST); 4562 } 4563 vattr_null(&vattr); 4564 vattr.va_type = VDIR; 4565 /* We will read cwdi->cwdi_cmask unlocked. */ 4566 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4567 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4568 if (!error) 4569 vrele(nd.ni_vp); 4570 vput(nd.ni_dvp); 4571 pathbuf_destroy(pb); 4572 return (error); 4573 } 4574 4575 /* 4576 * Remove a directory file. 4577 */ 4578 /* ARGSUSED */ 4579 int 4580 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4581 { 4582 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4583 AT_REMOVEDIR, UIO_USERSPACE); 4584 } 4585 4586 /* 4587 * Read a block of directory entries in a file system independent format. 4588 */ 4589 int 4590 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4591 { 4592 /* { 4593 syscallarg(int) fd; 4594 syscallarg(char *) buf; 4595 syscallarg(size_t) count; 4596 } */ 4597 file_t *fp; 4598 int error, done; 4599 4600 /* fd_getvnode() will use the descriptor for us */ 4601 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4602 return (error); 4603 if ((fp->f_flag & FREAD) == 0) { 4604 error = EBADF; 4605 goto out; 4606 } 4607 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4608 SCARG(uap, count), &done, l, 0, 0); 4609 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4610 *retval = done; 4611 out: 4612 fd_putfile(SCARG(uap, fd)); 4613 return (error); 4614 } 4615 4616 /* 4617 * Set the mode mask for creation of filesystem nodes. 4618 */ 4619 int 4620 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4621 { 4622 /* { 4623 syscallarg(mode_t) newmask; 4624 } */ 4625 struct proc *p = l->l_proc; 4626 struct cwdinfo *cwdi; 4627 4628 /* 4629 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4630 * important is that we serialize changes to the mask. The 4631 * rw_exit() will issue a write memory barrier on our behalf, 4632 * and force the changes out to other CPUs (as it must use an 4633 * atomic operation, draining the local CPU's store buffers). 4634 */ 4635 cwdi = p->p_cwdi; 4636 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4637 *retval = cwdi->cwdi_cmask; 4638 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4639 rw_exit(&cwdi->cwdi_lock); 4640 4641 return (0); 4642 } 4643 4644 int 4645 dorevoke(struct vnode *vp, kauth_cred_t cred) 4646 { 4647 struct vattr vattr; 4648 int error, fs_decision; 4649 4650 vn_lock(vp, LK_SHARED | LK_RETRY); 4651 error = VOP_GETATTR(vp, &vattr, cred); 4652 VOP_UNLOCK(vp); 4653 if (error != 0) 4654 return error; 4655 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4656 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4657 fs_decision); 4658 if (!error) 4659 VOP_REVOKE(vp, REVOKEALL); 4660 return (error); 4661 } 4662 4663 /* 4664 * Void all references to file by ripping underlying filesystem 4665 * away from vnode. 4666 */ 4667 /* ARGSUSED */ 4668 int 4669 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4670 { 4671 /* { 4672 syscallarg(const char *) path; 4673 } */ 4674 struct vnode *vp; 4675 int error; 4676 4677 error = namei_simple_user(SCARG(uap, path), 4678 NSM_FOLLOW_TRYEMULROOT, &vp); 4679 if (error != 0) 4680 return (error); 4681 error = dorevoke(vp, l->l_cred); 4682 vrele(vp); 4683 return (error); 4684 } 4685 4686 /* 4687 * Allocate backing store for a file, filling a hole without having to 4688 * explicitly write anything out. 4689 */ 4690 /* ARGSUSED */ 4691 int 4692 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4693 register_t *retval) 4694 { 4695 /* { 4696 syscallarg(int) fd; 4697 syscallarg(off_t) pos; 4698 syscallarg(off_t) len; 4699 } */ 4700 int fd; 4701 off_t pos, len; 4702 struct file *fp; 4703 struct vnode *vp; 4704 int error; 4705 4706 fd = SCARG(uap, fd); 4707 pos = SCARG(uap, pos); 4708 len = SCARG(uap, len); 4709 4710 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4711 *retval = EINVAL; 4712 return 0; 4713 } 4714 4715 error = fd_getvnode(fd, &fp); 4716 if (error) { 4717 *retval = error; 4718 return 0; 4719 } 4720 if ((fp->f_flag & FWRITE) == 0) { 4721 error = EBADF; 4722 goto fail; 4723 } 4724 vp = fp->f_vnode; 4725 4726 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4727 if (vp->v_type == VDIR) { 4728 error = EISDIR; 4729 } else { 4730 error = VOP_FALLOCATE(vp, pos, len); 4731 } 4732 VOP_UNLOCK(vp); 4733 4734 fail: 4735 fd_putfile(fd); 4736 *retval = error; 4737 return 0; 4738 } 4739 4740 /* 4741 * Deallocate backing store for a file, creating a hole. Also used for 4742 * invoking TRIM on disks. 4743 */ 4744 /* ARGSUSED */ 4745 int 4746 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4747 register_t *retval) 4748 { 4749 /* { 4750 syscallarg(int) fd; 4751 syscallarg(off_t) pos; 4752 syscallarg(off_t) len; 4753 } */ 4754 int fd; 4755 off_t pos, len; 4756 struct file *fp; 4757 struct vnode *vp; 4758 int error; 4759 4760 fd = SCARG(uap, fd); 4761 pos = SCARG(uap, pos); 4762 len = SCARG(uap, len); 4763 4764 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4765 return EINVAL; 4766 } 4767 4768 error = fd_getvnode(fd, &fp); 4769 if (error) { 4770 return error; 4771 } 4772 if ((fp->f_flag & FWRITE) == 0) { 4773 error = EBADF; 4774 goto fail; 4775 } 4776 vp = fp->f_vnode; 4777 4778 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4779 if (vp->v_type == VDIR) { 4780 error = EISDIR; 4781 } else { 4782 error = VOP_FDISCARD(vp, pos, len); 4783 } 4784 VOP_UNLOCK(vp); 4785 4786 fail: 4787 fd_putfile(fd); 4788 return error; 4789 } 4790
Indexes created Fri Oct 03 02:09:57 GMT 2025