vfs_syscalls.c revision 1.506
1 /* $NetBSD: vfs_syscalls.c,v 1.506 2017/02/17 08:26:07 hannken 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 1.506 2017/02/17 08:26:07 hannken 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 if (vfs_busy(mp, NULL)) { 286 error = EPERM; 287 goto out; 288 } 289 290 mutex_enter(&mp->mnt_updating); 291 292 mp->mnt_flag &= ~MNT_OP_FLAGS; 293 mp->mnt_flag |= flags & MNT_OP_FLAGS; 294 295 /* 296 * Set the mount level flags. 297 */ 298 if (flags & MNT_RDONLY) 299 mp->mnt_flag |= MNT_RDONLY; 300 else if (mp->mnt_flag & MNT_RDONLY) 301 mp->mnt_iflag |= IMNT_WANTRDWR; 302 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 303 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 304 error = VFS_MOUNT(mp, path, data, data_len); 305 306 if (error && data != NULL) { 307 int error2; 308 309 /* 310 * Update failed; let's try and see if it was an 311 * export request. For compat with 3.0 and earlier. 312 */ 313 error2 = vfs_hooks_reexport(mp, path, data); 314 315 /* 316 * Only update error code if the export request was 317 * understood but some problem occurred while 318 * processing it. 319 */ 320 if (error2 != EJUSTRETURN) 321 error = error2; 322 } 323 324 if (mp->mnt_iflag & IMNT_WANTRDWR) 325 mp->mnt_flag &= ~MNT_RDONLY; 326 if (error) 327 mp->mnt_flag = saved_flags; 328 mp->mnt_flag &= ~MNT_OP_FLAGS; 329 mp->mnt_iflag &= ~IMNT_WANTRDWR; 330 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 331 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 332 vfs_syncer_add_to_worklist(mp); 333 } else { 334 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 335 vfs_syncer_remove_from_worklist(mp); 336 } 337 mutex_exit(&mp->mnt_updating); 338 vfs_unbusy(mp, false, NULL); 339 340 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 341 (flags & MNT_EXTATTR)) { 342 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 343 NULL, 0, NULL) != 0) { 344 printf("%s: failed to start extattr, error = %d", 345 mp->mnt_stat.f_mntonname, error); 346 mp->mnt_flag &= ~MNT_EXTATTR; 347 } 348 } 349 350 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 351 !(flags & MNT_EXTATTR)) { 352 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 353 NULL, 0, NULL) != 0) { 354 printf("%s: failed to stop extattr, error = %d", 355 mp->mnt_stat.f_mntonname, error); 356 mp->mnt_flag |= MNT_RDONLY; 357 } 358 } 359 out: 360 return (error); 361 } 362 363 static int 364 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 365 struct vfsops **vfsops) 366 { 367 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 368 int error; 369 370 if (type_seg == UIO_USERSPACE) { 371 /* Copy file-system type from userspace. */ 372 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 373 } else { 374 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 375 KASSERT(error == 0); 376 } 377 378 if (error) { 379 /* 380 * Historically, filesystem types were identified by numbers. 381 * If we get an integer for the filesystem type instead of a 382 * string, we check to see if it matches one of the historic 383 * filesystem types. 384 */ 385 u_long fsindex = (u_long)fstype; 386 if (fsindex >= nmountcompatnames || 387 mountcompatnames[fsindex] == NULL) 388 return ENODEV; 389 strlcpy(fstypename, mountcompatnames[fsindex], 390 sizeof(fstypename)); 391 } 392 393 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 394 if (strcmp(fstypename, "ufs") == 0) 395 fstypename[0] = 'f'; 396 397 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 398 return 0; 399 400 /* If we can autoload a vfs module, try again */ 401 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 402 403 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 404 return 0; 405 406 return ENODEV; 407 } 408 409 static int 410 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 411 void *data, size_t *data_len) 412 { 413 struct mount *mp; 414 int error; 415 416 /* If MNT_GETARGS is specified, it should be the only flag. */ 417 if (flags & ~MNT_GETARGS) 418 return EINVAL; 419 420 mp = vp->v_mount; 421 422 /* XXX: probably some notion of "can see" here if we want isolation. */ 423 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 424 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 425 if (error) 426 return error; 427 428 if ((vp->v_vflag & VV_ROOT) == 0) 429 return EINVAL; 430 431 if (vfs_busy(mp, NULL)) 432 return EPERM; 433 434 mutex_enter(&mp->mnt_updating); 435 mp->mnt_flag &= ~MNT_OP_FLAGS; 436 mp->mnt_flag |= MNT_GETARGS; 437 error = VFS_MOUNT(mp, path, data, data_len); 438 mp->mnt_flag &= ~MNT_OP_FLAGS; 439 mutex_exit(&mp->mnt_updating); 440 441 vfs_unbusy(mp, false, NULL); 442 return (error); 443 } 444 445 int 446 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 447 { 448 /* { 449 syscallarg(const char *) type; 450 syscallarg(const char *) path; 451 syscallarg(int) flags; 452 syscallarg(void *) data; 453 syscallarg(size_t) data_len; 454 } */ 455 456 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 457 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 458 SCARG(uap, data_len), retval); 459 } 460 461 int 462 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 463 const char *path, int flags, void *data, enum uio_seg data_seg, 464 size_t data_len, register_t *retval) 465 { 466 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 467 struct vnode *vp; 468 void *data_buf = data; 469 bool vfsopsrele = false; 470 size_t alloc_sz = 0; 471 int error; 472 473 /* 474 * Get vnode to be covered 475 */ 476 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 477 if (error != 0) { 478 vp = NULL; 479 goto done; 480 } 481 482 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 483 vfsops = vp->v_mount->mnt_op; 484 } else { 485 /* 'type' is userspace */ 486 error = mount_get_vfsops(type, type_seg, &vfsops); 487 if (error != 0) 488 goto done; 489 vfsopsrele = true; 490 } 491 492 /* 493 * We allow data to be NULL, even for userspace. Some fs's don't need 494 * it. The others will handle NULL. 495 */ 496 if (data != NULL && data_seg == UIO_USERSPACE) { 497 if (data_len == 0) { 498 /* No length supplied, use default for filesystem */ 499 data_len = vfsops->vfs_min_mount_data; 500 501 /* 502 * Hopefully a longer buffer won't make copyin() fail. 503 * For compatibility with 3.0 and earlier. 504 */ 505 if (flags & MNT_UPDATE 506 && data_len < sizeof (struct mnt_export_args30)) 507 data_len = sizeof (struct mnt_export_args30); 508 } 509 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 510 error = EINVAL; 511 goto done; 512 } 513 alloc_sz = data_len; 514 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 515 516 /* NFS needs the buffer even for mnt_getargs .... */ 517 error = copyin(data, data_buf, data_len); 518 if (error != 0) 519 goto done; 520 } 521 522 if (flags & MNT_GETARGS) { 523 if (data_len == 0) { 524 error = EINVAL; 525 goto done; 526 } 527 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 528 if (error != 0) 529 goto done; 530 if (data_seg == UIO_USERSPACE) 531 error = copyout(data_buf, data, data_len); 532 *retval = data_len; 533 } else if (flags & MNT_UPDATE) { 534 error = mount_update(l, vp, path, flags, data_buf, &data_len); 535 } else { 536 /* Locking is handled internally in mount_domount(). */ 537 KASSERT(vfsopsrele == true); 538 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 539 &data_len); 540 vfsopsrele = false; 541 } 542 543 done: 544 if (vfsopsrele) 545 vfs_delref(vfsops); 546 if (vp != NULL) { 547 vrele(vp); 548 } 549 if (data_buf != data) 550 kmem_free(data_buf, alloc_sz); 551 return (error); 552 } 553 554 /* 555 * Unmount a file system. 556 * 557 * Note: unmount takes a path to the vnode mounted on as argument, 558 * not special file (as before). 559 */ 560 /* ARGSUSED */ 561 int 562 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 563 { 564 /* { 565 syscallarg(const char *) path; 566 syscallarg(int) flags; 567 } */ 568 struct vnode *vp; 569 struct mount *mp; 570 int error; 571 struct pathbuf *pb; 572 struct nameidata nd; 573 574 error = pathbuf_copyin(SCARG(uap, path), &pb); 575 if (error) { 576 return error; 577 } 578 579 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 580 if ((error = namei(&nd)) != 0) { 581 pathbuf_destroy(pb); 582 return error; 583 } 584 vp = nd.ni_vp; 585 pathbuf_destroy(pb); 586 587 mp = vp->v_mount; 588 atomic_inc_uint(&mp->mnt_refcnt); 589 VOP_UNLOCK(vp); 590 591 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 592 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 593 if (error) { 594 vrele(vp); 595 vfs_destroy(mp); 596 return (error); 597 } 598 599 /* 600 * Don't allow unmounting the root file system. 601 */ 602 if (mp->mnt_flag & MNT_ROOTFS) { 603 vrele(vp); 604 vfs_destroy(mp); 605 return (EINVAL); 606 } 607 608 /* 609 * Must be the root of the filesystem 610 */ 611 if ((vp->v_vflag & VV_ROOT) == 0) { 612 vrele(vp); 613 vfs_destroy(mp); 614 return (EINVAL); 615 } 616 617 vrele(vp); 618 error = dounmount(mp, SCARG(uap, flags), l); 619 vfs_destroy(mp); 620 return error; 621 } 622 623 /* 624 * Sync each mounted filesystem. 625 */ 626 #ifdef DEBUG 627 int syncprt = 0; 628 struct ctldebug debug0 = { "syncprt", &syncprt }; 629 #endif 630 631 void 632 do_sys_sync(struct lwp *l) 633 { 634 struct mount *mp, *nmp; 635 int asyncflag; 636 637 mutex_enter(&mountlist_lock); 638 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 639 if (vfs_busy(mp, &nmp)) { 640 continue; 641 } 642 fstrans_start(mp, FSTRANS_SHARED); 643 mutex_enter(&mp->mnt_updating); 644 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 645 asyncflag = mp->mnt_flag & MNT_ASYNC; 646 mp->mnt_flag &= ~MNT_ASYNC; 647 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 648 if (asyncflag) 649 mp->mnt_flag |= MNT_ASYNC; 650 } 651 mutex_exit(&mp->mnt_updating); 652 fstrans_done(mp); 653 vfs_unbusy(mp, false, &nmp); 654 } 655 mutex_exit(&mountlist_lock); 656 #ifdef DEBUG 657 if (syncprt) 658 vfs_bufstats(); 659 #endif /* DEBUG */ 660 } 661 662 /* ARGSUSED */ 663 int 664 sys_sync(struct lwp *l, const void *v, register_t *retval) 665 { 666 do_sys_sync(l); 667 return (0); 668 } 669 670 671 /* 672 * Access or change filesystem quotas. 673 * 674 * (this is really 14 different calls bundled into one) 675 */ 676 677 static int 678 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 679 { 680 struct quotastat info_k; 681 int error; 682 683 /* ensure any padding bytes are cleared */ 684 memset(&info_k, 0, sizeof(info_k)); 685 686 error = vfs_quotactl_stat(mp, &info_k); 687 if (error) { 688 return error; 689 } 690 691 return copyout(&info_k, info_u, sizeof(info_k)); 692 } 693 694 static int 695 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 696 struct quotaidtypestat *info_u) 697 { 698 struct quotaidtypestat info_k; 699 int error; 700 701 /* ensure any padding bytes are cleared */ 702 memset(&info_k, 0, sizeof(info_k)); 703 704 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 705 if (error) { 706 return error; 707 } 708 709 return copyout(&info_k, info_u, sizeof(info_k)); 710 } 711 712 static int 713 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 714 struct quotaobjtypestat *info_u) 715 { 716 struct quotaobjtypestat info_k; 717 int error; 718 719 /* ensure any padding bytes are cleared */ 720 memset(&info_k, 0, sizeof(info_k)); 721 722 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 723 if (error) { 724 return error; 725 } 726 727 return copyout(&info_k, info_u, sizeof(info_k)); 728 } 729 730 static int 731 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 732 struct quotaval *val_u) 733 { 734 struct quotakey key_k; 735 struct quotaval val_k; 736 int error; 737 738 /* ensure any padding bytes are cleared */ 739 memset(&val_k, 0, sizeof(val_k)); 740 741 error = copyin(key_u, &key_k, sizeof(key_k)); 742 if (error) { 743 return error; 744 } 745 746 error = vfs_quotactl_get(mp, &key_k, &val_k); 747 if (error) { 748 return error; 749 } 750 751 return copyout(&val_k, val_u, sizeof(val_k)); 752 } 753 754 static int 755 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 756 const struct quotaval *val_u) 757 { 758 struct quotakey key_k; 759 struct quotaval val_k; 760 int error; 761 762 error = copyin(key_u, &key_k, sizeof(key_k)); 763 if (error) { 764 return error; 765 } 766 767 error = copyin(val_u, &val_k, sizeof(val_k)); 768 if (error) { 769 return error; 770 } 771 772 return vfs_quotactl_put(mp, &key_k, &val_k); 773 } 774 775 static int 776 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 777 { 778 struct quotakey key_k; 779 int error; 780 781 error = copyin(key_u, &key_k, sizeof(key_k)); 782 if (error) { 783 return error; 784 } 785 786 return vfs_quotactl_del(mp, &key_k); 787 } 788 789 static int 790 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 791 { 792 struct quotakcursor cursor_k; 793 int error; 794 795 /* ensure any padding bytes are cleared */ 796 memset(&cursor_k, 0, sizeof(cursor_k)); 797 798 error = vfs_quotactl_cursoropen(mp, &cursor_k); 799 if (error) { 800 return error; 801 } 802 803 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 804 } 805 806 static int 807 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 808 { 809 struct quotakcursor cursor_k; 810 int error; 811 812 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 813 if (error) { 814 return error; 815 } 816 817 return vfs_quotactl_cursorclose(mp, &cursor_k); 818 } 819 820 static int 821 do_sys_quotactl_cursorskipidtype(struct mount *mp, 822 struct quotakcursor *cursor_u, int idtype) 823 { 824 struct quotakcursor cursor_k; 825 int error; 826 827 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 828 if (error) { 829 return error; 830 } 831 832 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 833 if (error) { 834 return error; 835 } 836 837 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 838 } 839 840 static int 841 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 842 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 843 unsigned *ret_u) 844 { 845 #define CGET_STACK_MAX 8 846 struct quotakcursor cursor_k; 847 struct quotakey stackkeys[CGET_STACK_MAX]; 848 struct quotaval stackvals[CGET_STACK_MAX]; 849 struct quotakey *keys_k; 850 struct quotaval *vals_k; 851 unsigned ret_k; 852 int error; 853 854 if (maxnum > 128) { 855 maxnum = 128; 856 } 857 858 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 859 if (error) { 860 return error; 861 } 862 863 if (maxnum <= CGET_STACK_MAX) { 864 keys_k = stackkeys; 865 vals_k = stackvals; 866 /* ensure any padding bytes are cleared */ 867 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 868 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 869 } else { 870 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 871 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 872 } 873 874 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 875 &ret_k); 876 if (error) { 877 goto fail; 878 } 879 880 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 881 if (error) { 882 goto fail; 883 } 884 885 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 886 if (error) { 887 goto fail; 888 } 889 890 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 891 if (error) { 892 goto fail; 893 } 894 895 /* do last to maximize the chance of being able to recover a failure */ 896 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 897 898 fail: 899 if (keys_k != stackkeys) { 900 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 901 } 902 if (vals_k != stackvals) { 903 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 904 } 905 return error; 906 } 907 908 static int 909 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 910 int *ret_u) 911 { 912 struct quotakcursor cursor_k; 913 int ret_k; 914 int error; 915 916 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 917 if (error) { 918 return error; 919 } 920 921 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 922 if (error) { 923 return error; 924 } 925 926 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 927 if (error) { 928 return error; 929 } 930 931 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 932 } 933 934 static int 935 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 936 { 937 struct quotakcursor cursor_k; 938 int error; 939 940 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 941 if (error) { 942 return error; 943 } 944 945 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 946 if (error) { 947 return error; 948 } 949 950 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 951 } 952 953 static int 954 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 955 { 956 char *path_k; 957 int error; 958 959 /* XXX this should probably be a struct pathbuf */ 960 path_k = PNBUF_GET(); 961 error = copyin(path_u, path_k, PATH_MAX); 962 if (error) { 963 PNBUF_PUT(path_k); 964 return error; 965 } 966 967 error = vfs_quotactl_quotaon(mp, idtype, path_k); 968 969 PNBUF_PUT(path_k); 970 return error; 971 } 972 973 static int 974 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 975 { 976 return vfs_quotactl_quotaoff(mp, idtype); 977 } 978 979 int 980 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 981 { 982 struct mount *mp; 983 struct vnode *vp; 984 int error; 985 986 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 987 if (error != 0) 988 return (error); 989 mp = vp->v_mount; 990 991 switch (args->qc_op) { 992 case QUOTACTL_STAT: 993 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 994 break; 995 case QUOTACTL_IDTYPESTAT: 996 error = do_sys_quotactl_idtypestat(mp, 997 args->u.idtypestat.qc_idtype, 998 args->u.idtypestat.qc_info); 999 break; 1000 case QUOTACTL_OBJTYPESTAT: 1001 error = do_sys_quotactl_objtypestat(mp, 1002 args->u.objtypestat.qc_objtype, 1003 args->u.objtypestat.qc_info); 1004 break; 1005 case QUOTACTL_GET: 1006 error = do_sys_quotactl_get(mp, 1007 args->u.get.qc_key, 1008 args->u.get.qc_val); 1009 break; 1010 case QUOTACTL_PUT: 1011 error = do_sys_quotactl_put(mp, 1012 args->u.put.qc_key, 1013 args->u.put.qc_val); 1014 break; 1015 case QUOTACTL_DEL: 1016 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1017 break; 1018 case QUOTACTL_CURSOROPEN: 1019 error = do_sys_quotactl_cursoropen(mp, 1020 args->u.cursoropen.qc_cursor); 1021 break; 1022 case QUOTACTL_CURSORCLOSE: 1023 error = do_sys_quotactl_cursorclose(mp, 1024 args->u.cursorclose.qc_cursor); 1025 break; 1026 case QUOTACTL_CURSORSKIPIDTYPE: 1027 error = do_sys_quotactl_cursorskipidtype(mp, 1028 args->u.cursorskipidtype.qc_cursor, 1029 args->u.cursorskipidtype.qc_idtype); 1030 break; 1031 case QUOTACTL_CURSORGET: 1032 error = do_sys_quotactl_cursorget(mp, 1033 args->u.cursorget.qc_cursor, 1034 args->u.cursorget.qc_keys, 1035 args->u.cursorget.qc_vals, 1036 args->u.cursorget.qc_maxnum, 1037 args->u.cursorget.qc_ret); 1038 break; 1039 case QUOTACTL_CURSORATEND: 1040 error = do_sys_quotactl_cursoratend(mp, 1041 args->u.cursoratend.qc_cursor, 1042 args->u.cursoratend.qc_ret); 1043 break; 1044 case QUOTACTL_CURSORREWIND: 1045 error = do_sys_quotactl_cursorrewind(mp, 1046 args->u.cursorrewind.qc_cursor); 1047 break; 1048 case QUOTACTL_QUOTAON: 1049 error = do_sys_quotactl_quotaon(mp, 1050 args->u.quotaon.qc_idtype, 1051 args->u.quotaon.qc_quotafile); 1052 break; 1053 case QUOTACTL_QUOTAOFF: 1054 error = do_sys_quotactl_quotaoff(mp, 1055 args->u.quotaoff.qc_idtype); 1056 break; 1057 default: 1058 error = EINVAL; 1059 break; 1060 } 1061 1062 vrele(vp); 1063 return error; 1064 } 1065 1066 /* ARGSUSED */ 1067 int 1068 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1069 register_t *retval) 1070 { 1071 /* { 1072 syscallarg(const char *) path; 1073 syscallarg(struct quotactl_args *) args; 1074 } */ 1075 struct quotactl_args args; 1076 int error; 1077 1078 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1079 if (error) { 1080 return error; 1081 } 1082 1083 return do_sys_quotactl(SCARG(uap, path), &args); 1084 } 1085 1086 int 1087 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1088 int root) 1089 { 1090 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1091 int error = 0; 1092 1093 /* 1094 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1095 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1096 * overrides MNT_NOWAIT. 1097 */ 1098 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1099 (flags != MNT_WAIT && flags != 0)) { 1100 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1101 goto done; 1102 } 1103 1104 /* Get the filesystem stats now */ 1105 memset(sp, 0, sizeof(*sp)); 1106 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1107 return error; 1108 } 1109 1110 if (cwdi->cwdi_rdir == NULL) 1111 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1112 done: 1113 if (cwdi->cwdi_rdir != NULL) { 1114 size_t len; 1115 char *bp; 1116 char c; 1117 char *path = PNBUF_GET(); 1118 1119 bp = path + MAXPATHLEN; 1120 *--bp = '\0'; 1121 rw_enter(&cwdi->cwdi_lock, RW_READER); 1122 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1123 MAXPATHLEN / 2, 0, l); 1124 rw_exit(&cwdi->cwdi_lock); 1125 if (error) { 1126 PNBUF_PUT(path); 1127 return error; 1128 } 1129 len = strlen(bp); 1130 if (len != 1) { 1131 /* 1132 * for mount points that are below our root, we can see 1133 * them, so we fix up the pathname and return them. The 1134 * rest we cannot see, so we don't allow viewing the 1135 * data. 1136 */ 1137 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1138 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1139 (void)strlcpy(sp->f_mntonname, 1140 c == '\0' ? "/" : &sp->f_mntonname[len], 1141 sizeof(sp->f_mntonname)); 1142 } else { 1143 if (root) 1144 (void)strlcpy(sp->f_mntonname, "/", 1145 sizeof(sp->f_mntonname)); 1146 else 1147 error = EPERM; 1148 } 1149 } 1150 PNBUF_PUT(path); 1151 } 1152 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1153 return error; 1154 } 1155 1156 /* 1157 * Get filesystem statistics by path. 1158 */ 1159 int 1160 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1161 { 1162 struct mount *mp; 1163 int error; 1164 struct vnode *vp; 1165 1166 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1167 if (error != 0) 1168 return error; 1169 mp = vp->v_mount; 1170 error = dostatvfs(mp, sb, l, flags, 1); 1171 vrele(vp); 1172 return error; 1173 } 1174 1175 /* ARGSUSED */ 1176 int 1177 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1178 { 1179 /* { 1180 syscallarg(const char *) path; 1181 syscallarg(struct statvfs *) buf; 1182 syscallarg(int) flags; 1183 } */ 1184 struct statvfs *sb; 1185 int error; 1186 1187 sb = STATVFSBUF_GET(); 1188 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1189 if (error == 0) 1190 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1191 STATVFSBUF_PUT(sb); 1192 return error; 1193 } 1194 1195 /* 1196 * Get filesystem statistics by fd. 1197 */ 1198 int 1199 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1200 { 1201 file_t *fp; 1202 struct mount *mp; 1203 int error; 1204 1205 /* fd_getvnode() will use the descriptor for us */ 1206 if ((error = fd_getvnode(fd, &fp)) != 0) 1207 return (error); 1208 mp = fp->f_vnode->v_mount; 1209 error = dostatvfs(mp, sb, curlwp, flags, 1); 1210 fd_putfile(fd); 1211 return error; 1212 } 1213 1214 /* ARGSUSED */ 1215 int 1216 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1217 { 1218 /* { 1219 syscallarg(int) fd; 1220 syscallarg(struct statvfs *) buf; 1221 syscallarg(int) flags; 1222 } */ 1223 struct statvfs *sb; 1224 int error; 1225 1226 sb = STATVFSBUF_GET(); 1227 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1228 if (error == 0) 1229 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1230 STATVFSBUF_PUT(sb); 1231 return error; 1232 } 1233 1234 1235 /* 1236 * Get statistics on all filesystems. 1237 */ 1238 int 1239 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1240 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1241 register_t *retval) 1242 { 1243 int root = 0; 1244 struct proc *p = l->l_proc; 1245 struct mount *mp, *nmp; 1246 struct statvfs *sb; 1247 size_t count, maxcount; 1248 int error = 0; 1249 1250 sb = STATVFSBUF_GET(); 1251 maxcount = bufsize / entry_sz; 1252 mutex_enter(&mountlist_lock); 1253 count = 0; 1254 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 1255 if (vfs_busy(mp, &nmp)) { 1256 continue; 1257 } 1258 if (sfsp && count < maxcount) { 1259 error = dostatvfs(mp, sb, l, flags, 0); 1260 if (error) { 1261 vfs_unbusy(mp, false, &nmp); 1262 error = 0; 1263 continue; 1264 } 1265 error = copyfn(sb, sfsp, entry_sz); 1266 if (error) { 1267 vfs_unbusy(mp, false, NULL); 1268 goto out; 1269 } 1270 sfsp = (char *)sfsp + entry_sz; 1271 root |= strcmp(sb->f_mntonname, "/") == 0; 1272 } 1273 count++; 1274 vfs_unbusy(mp, false, &nmp); 1275 } 1276 mutex_exit(&mountlist_lock); 1277 1278 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1279 /* 1280 * fake a root entry 1281 */ 1282 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1283 sb, l, flags, 1); 1284 if (error != 0) 1285 goto out; 1286 if (sfsp) { 1287 error = copyfn(sb, sfsp, entry_sz); 1288 if (error != 0) 1289 goto out; 1290 } 1291 count++; 1292 } 1293 if (sfsp && count > maxcount) 1294 *retval = maxcount; 1295 else 1296 *retval = count; 1297 out: 1298 STATVFSBUF_PUT(sb); 1299 return error; 1300 } 1301 1302 int 1303 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1304 { 1305 /* { 1306 syscallarg(struct statvfs *) buf; 1307 syscallarg(size_t) bufsize; 1308 syscallarg(int) flags; 1309 } */ 1310 1311 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1312 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1313 } 1314 1315 /* 1316 * Change current working directory to a given file descriptor. 1317 */ 1318 /* ARGSUSED */ 1319 int 1320 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1321 { 1322 /* { 1323 syscallarg(int) fd; 1324 } */ 1325 struct proc *p = l->l_proc; 1326 struct cwdinfo *cwdi; 1327 struct vnode *vp, *tdp; 1328 struct mount *mp; 1329 file_t *fp; 1330 int error, fd; 1331 1332 /* fd_getvnode() will use the descriptor for us */ 1333 fd = SCARG(uap, fd); 1334 if ((error = fd_getvnode(fd, &fp)) != 0) 1335 return (error); 1336 vp = fp->f_vnode; 1337 1338 vref(vp); 1339 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1340 if (vp->v_type != VDIR) 1341 error = ENOTDIR; 1342 else 1343 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1344 if (error) { 1345 vput(vp); 1346 goto out; 1347 } 1348 while ((mp = vp->v_mountedhere) != NULL) { 1349 error = vfs_busy(mp, NULL); 1350 vput(vp); 1351 if (error != 0) 1352 goto out; 1353 error = VFS_ROOT(mp, &tdp); 1354 vfs_unbusy(mp, false, NULL); 1355 if (error) 1356 goto out; 1357 vp = tdp; 1358 } 1359 VOP_UNLOCK(vp); 1360 1361 /* 1362 * Disallow changing to a directory not under the process's 1363 * current root directory (if there is one). 1364 */ 1365 cwdi = p->p_cwdi; 1366 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1367 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1368 vrele(vp); 1369 error = EPERM; /* operation not permitted */ 1370 } else { 1371 vrele(cwdi->cwdi_cdir); 1372 cwdi->cwdi_cdir = vp; 1373 } 1374 rw_exit(&cwdi->cwdi_lock); 1375 1376 out: 1377 fd_putfile(fd); 1378 return (error); 1379 } 1380 1381 /* 1382 * Change this process's notion of the root directory to a given file 1383 * descriptor. 1384 */ 1385 int 1386 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1387 { 1388 struct proc *p = l->l_proc; 1389 struct vnode *vp; 1390 file_t *fp; 1391 int error, fd = SCARG(uap, fd); 1392 1393 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1394 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1395 return error; 1396 /* fd_getvnode() will use the descriptor for us */ 1397 if ((error = fd_getvnode(fd, &fp)) != 0) 1398 return error; 1399 vp = fp->f_vnode; 1400 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1401 if (vp->v_type != VDIR) 1402 error = ENOTDIR; 1403 else 1404 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1405 VOP_UNLOCK(vp); 1406 if (error) 1407 goto out; 1408 vref(vp); 1409 1410 change_root(p->p_cwdi, vp, l); 1411 1412 out: 1413 fd_putfile(fd); 1414 return (error); 1415 } 1416 1417 /* 1418 * Change current working directory (``.''). 1419 */ 1420 /* ARGSUSED */ 1421 int 1422 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1423 { 1424 /* { 1425 syscallarg(const char *) path; 1426 } */ 1427 struct proc *p = l->l_proc; 1428 struct cwdinfo *cwdi; 1429 int error; 1430 struct vnode *vp; 1431 1432 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1433 &vp, l)) != 0) 1434 return (error); 1435 cwdi = p->p_cwdi; 1436 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1437 vrele(cwdi->cwdi_cdir); 1438 cwdi->cwdi_cdir = vp; 1439 rw_exit(&cwdi->cwdi_lock); 1440 return (0); 1441 } 1442 1443 /* 1444 * Change notion of root (``/'') directory. 1445 */ 1446 /* ARGSUSED */ 1447 int 1448 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1449 { 1450 /* { 1451 syscallarg(const char *) path; 1452 } */ 1453 struct proc *p = l->l_proc; 1454 int error; 1455 struct vnode *vp; 1456 1457 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1458 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1459 return (error); 1460 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1461 &vp, l)) != 0) 1462 return (error); 1463 1464 change_root(p->p_cwdi, vp, l); 1465 1466 return (0); 1467 } 1468 1469 /* 1470 * Common routine for chroot and fchroot. 1471 * NB: callers need to properly authorize the change root operation. 1472 */ 1473 void 1474 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1475 { 1476 struct proc *p = l->l_proc; 1477 kauth_cred_t ncred; 1478 1479 ncred = kauth_cred_alloc(); 1480 1481 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1482 if (cwdi->cwdi_rdir != NULL) 1483 vrele(cwdi->cwdi_rdir); 1484 cwdi->cwdi_rdir = vp; 1485 1486 /* 1487 * Prevent escaping from chroot by putting the root under 1488 * the working directory. Silently chdir to / if we aren't 1489 * already there. 1490 */ 1491 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1492 /* 1493 * XXX would be more failsafe to change directory to a 1494 * deadfs node here instead 1495 */ 1496 vrele(cwdi->cwdi_cdir); 1497 vref(vp); 1498 cwdi->cwdi_cdir = vp; 1499 } 1500 rw_exit(&cwdi->cwdi_lock); 1501 1502 /* Get a write lock on the process credential. */ 1503 proc_crmod_enter(); 1504 1505 kauth_cred_clone(p->p_cred, ncred); 1506 kauth_proc_chroot(ncred, p->p_cwdi); 1507 1508 /* Broadcast our credentials to the process and other LWPs. */ 1509 proc_crmod_leave(ncred, p->p_cred, true); 1510 } 1511 1512 /* 1513 * Common routine for chroot and chdir. 1514 * XXX "where" should be enum uio_seg 1515 */ 1516 int 1517 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1518 { 1519 struct pathbuf *pb; 1520 struct nameidata nd; 1521 int error; 1522 1523 error = pathbuf_maybe_copyin(path, where, &pb); 1524 if (error) { 1525 return error; 1526 } 1527 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1528 if ((error = namei(&nd)) != 0) { 1529 pathbuf_destroy(pb); 1530 return error; 1531 } 1532 *vpp = nd.ni_vp; 1533 pathbuf_destroy(pb); 1534 1535 if ((*vpp)->v_type != VDIR) 1536 error = ENOTDIR; 1537 else 1538 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1539 1540 if (error) 1541 vput(*vpp); 1542 else 1543 VOP_UNLOCK(*vpp); 1544 return (error); 1545 } 1546 1547 /* 1548 * Internals of sys_open - path has already been converted into a pathbuf 1549 * (so we can easily reuse this function from other parts of the kernel, 1550 * like posix_spawn post-processing). 1551 */ 1552 int 1553 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1554 int open_mode, int *fd) 1555 { 1556 struct proc *p = l->l_proc; 1557 struct cwdinfo *cwdi = p->p_cwdi; 1558 file_t *fp; 1559 struct vnode *vp; 1560 int flags, cmode; 1561 int indx, error; 1562 struct nameidata nd; 1563 1564 if (open_flags & O_SEARCH) { 1565 open_flags &= ~(int)O_SEARCH; 1566 } 1567 1568 flags = FFLAGS(open_flags); 1569 if ((flags & (FREAD | FWRITE)) == 0) 1570 return EINVAL; 1571 1572 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1573 return error; 1574 } 1575 1576 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1577 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1578 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1579 if (dvp != NULL) 1580 NDAT(&nd, dvp); 1581 1582 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1583 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1584 fd_abort(p, fp, indx); 1585 if ((error == EDUPFD || error == EMOVEFD) && 1586 l->l_dupfd >= 0 && /* XXX from fdopen */ 1587 (error = 1588 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1589 *fd = indx; 1590 return 0; 1591 } 1592 if (error == ERESTART) 1593 error = EINTR; 1594 return error; 1595 } 1596 1597 l->l_dupfd = 0; 1598 vp = nd.ni_vp; 1599 1600 if ((error = open_setfp(l, fp, vp, indx, flags))) 1601 return error; 1602 1603 VOP_UNLOCK(vp); 1604 *fd = indx; 1605 fd_affix(p, fp, indx); 1606 return 0; 1607 } 1608 1609 int 1610 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1611 { 1612 struct pathbuf *pb; 1613 int error, oflags; 1614 1615 oflags = FFLAGS(open_flags); 1616 if ((oflags & (FREAD | FWRITE)) == 0) 1617 return EINVAL; 1618 1619 pb = pathbuf_create(path); 1620 if (pb == NULL) 1621 return ENOMEM; 1622 1623 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1624 pathbuf_destroy(pb); 1625 1626 return error; 1627 } 1628 1629 /* 1630 * Check permissions, allocate an open file structure, 1631 * and call the device open routine if any. 1632 */ 1633 static int 1634 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1635 int mode, int *fd) 1636 { 1637 file_t *dfp = NULL; 1638 struct vnode *dvp = NULL; 1639 struct pathbuf *pb; 1640 int error; 1641 1642 #ifdef COMPAT_10 /* XXX: and perhaps later */ 1643 if (path == NULL) { 1644 pb = pathbuf_create("."); 1645 if (pb == NULL) 1646 return ENOMEM; 1647 } else 1648 #endif 1649 { 1650 error = pathbuf_copyin(path, &pb); 1651 if (error) 1652 return error; 1653 } 1654 1655 if (fdat != AT_FDCWD) { 1656 /* fd_getvnode() will use the descriptor for us */ 1657 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1658 goto out; 1659 1660 dvp = dfp->f_vnode; 1661 } 1662 1663 error = do_open(l, dvp, pb, flags, mode, fd); 1664 1665 if (dfp != NULL) 1666 fd_putfile(fdat); 1667 out: 1668 pathbuf_destroy(pb); 1669 return error; 1670 } 1671 1672 int 1673 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1674 { 1675 /* { 1676 syscallarg(const char *) path; 1677 syscallarg(int) flags; 1678 syscallarg(int) mode; 1679 } */ 1680 int error; 1681 int fd; 1682 1683 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1684 SCARG(uap, flags), SCARG(uap, mode), &fd); 1685 1686 if (error == 0) 1687 *retval = fd; 1688 1689 return error; 1690 } 1691 1692 int 1693 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1694 { 1695 /* { 1696 syscallarg(int) fd; 1697 syscallarg(const char *) path; 1698 syscallarg(int) oflags; 1699 syscallarg(int) mode; 1700 } */ 1701 int error; 1702 int fd; 1703 1704 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1705 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1706 1707 if (error == 0) 1708 *retval = fd; 1709 1710 return error; 1711 } 1712 1713 static void 1714 vfs__fhfree(fhandle_t *fhp) 1715 { 1716 size_t fhsize; 1717 1718 fhsize = FHANDLE_SIZE(fhp); 1719 kmem_free(fhp, fhsize); 1720 } 1721 1722 /* 1723 * vfs_composefh: compose a filehandle. 1724 */ 1725 1726 int 1727 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1728 { 1729 struct mount *mp; 1730 struct fid *fidp; 1731 int error; 1732 size_t needfhsize; 1733 size_t fidsize; 1734 1735 mp = vp->v_mount; 1736 fidp = NULL; 1737 if (*fh_size < FHANDLE_SIZE_MIN) { 1738 fidsize = 0; 1739 } else { 1740 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1741 if (fhp != NULL) { 1742 memset(fhp, 0, *fh_size); 1743 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1744 fidp = &fhp->fh_fid; 1745 } 1746 } 1747 error = VFS_VPTOFH(vp, fidp, &fidsize); 1748 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1749 if (error == 0 && *fh_size < needfhsize) { 1750 error = E2BIG; 1751 } 1752 *fh_size = needfhsize; 1753 return error; 1754 } 1755 1756 int 1757 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1758 { 1759 struct mount *mp; 1760 fhandle_t *fhp; 1761 size_t fhsize; 1762 size_t fidsize; 1763 int error; 1764 1765 mp = vp->v_mount; 1766 fidsize = 0; 1767 error = VFS_VPTOFH(vp, NULL, &fidsize); 1768 KASSERT(error != 0); 1769 if (error != E2BIG) { 1770 goto out; 1771 } 1772 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1773 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1774 if (fhp == NULL) { 1775 error = ENOMEM; 1776 goto out; 1777 } 1778 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1779 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1780 if (error == 0) { 1781 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1782 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1783 *fhpp = fhp; 1784 } else { 1785 kmem_free(fhp, fhsize); 1786 } 1787 out: 1788 return error; 1789 } 1790 1791 void 1792 vfs_composefh_free(fhandle_t *fhp) 1793 { 1794 1795 vfs__fhfree(fhp); 1796 } 1797 1798 /* 1799 * vfs_fhtovp: lookup a vnode by a filehandle. 1800 */ 1801 1802 int 1803 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1804 { 1805 struct mount *mp; 1806 int error; 1807 1808 *vpp = NULL; 1809 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1810 if (mp == NULL) { 1811 error = ESTALE; 1812 goto out; 1813 } 1814 if (mp->mnt_op->vfs_fhtovp == NULL) { 1815 error = EOPNOTSUPP; 1816 goto out; 1817 } 1818 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1819 out: 1820 return error; 1821 } 1822 1823 /* 1824 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1825 * the needed size. 1826 */ 1827 1828 int 1829 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1830 { 1831 fhandle_t *fhp; 1832 int error; 1833 1834 if (fhsize > FHANDLE_SIZE_MAX) { 1835 return EINVAL; 1836 } 1837 if (fhsize < FHANDLE_SIZE_MIN) { 1838 return EINVAL; 1839 } 1840 again: 1841 fhp = kmem_alloc(fhsize, KM_SLEEP); 1842 if (fhp == NULL) { 1843 return ENOMEM; 1844 } 1845 error = copyin(ufhp, fhp, fhsize); 1846 if (error == 0) { 1847 /* XXX this check shouldn't be here */ 1848 if (FHANDLE_SIZE(fhp) == fhsize) { 1849 *fhpp = fhp; 1850 return 0; 1851 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1852 /* 1853 * a kludge for nfsv2 padded handles. 1854 */ 1855 size_t sz; 1856 1857 sz = FHANDLE_SIZE(fhp); 1858 kmem_free(fhp, fhsize); 1859 fhsize = sz; 1860 goto again; 1861 } else { 1862 /* 1863 * userland told us wrong size. 1864 */ 1865 error = EINVAL; 1866 } 1867 } 1868 kmem_free(fhp, fhsize); 1869 return error; 1870 } 1871 1872 void 1873 vfs_copyinfh_free(fhandle_t *fhp) 1874 { 1875 1876 vfs__fhfree(fhp); 1877 } 1878 1879 /* 1880 * Get file handle system call 1881 */ 1882 int 1883 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1884 { 1885 /* { 1886 syscallarg(char *) fname; 1887 syscallarg(fhandle_t *) fhp; 1888 syscallarg(size_t *) fh_size; 1889 } */ 1890 struct vnode *vp; 1891 fhandle_t *fh; 1892 int error; 1893 struct pathbuf *pb; 1894 struct nameidata nd; 1895 size_t sz; 1896 size_t usz; 1897 1898 /* 1899 * Must be super user 1900 */ 1901 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1902 0, NULL, NULL, NULL); 1903 if (error) 1904 return (error); 1905 1906 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1907 if (error) { 1908 return error; 1909 } 1910 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1911 error = namei(&nd); 1912 if (error) { 1913 pathbuf_destroy(pb); 1914 return error; 1915 } 1916 vp = nd.ni_vp; 1917 pathbuf_destroy(pb); 1918 1919 error = vfs_composefh_alloc(vp, &fh); 1920 vput(vp); 1921 if (error != 0) { 1922 return error; 1923 } 1924 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1925 if (error != 0) { 1926 goto out; 1927 } 1928 sz = FHANDLE_SIZE(fh); 1929 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1930 if (error != 0) { 1931 goto out; 1932 } 1933 if (usz >= sz) { 1934 error = copyout(fh, SCARG(uap, fhp), sz); 1935 } else { 1936 error = E2BIG; 1937 } 1938 out: 1939 vfs_composefh_free(fh); 1940 return (error); 1941 } 1942 1943 /* 1944 * Open a file given a file handle. 1945 * 1946 * Check permissions, allocate an open file structure, 1947 * and call the device open routine if any. 1948 */ 1949 1950 int 1951 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1952 register_t *retval) 1953 { 1954 file_t *fp; 1955 struct vnode *vp = NULL; 1956 kauth_cred_t cred = l->l_cred; 1957 file_t *nfp; 1958 int indx, error; 1959 struct vattr va; 1960 fhandle_t *fh; 1961 int flags; 1962 proc_t *p; 1963 1964 p = curproc; 1965 1966 /* 1967 * Must be super user 1968 */ 1969 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1970 0, NULL, NULL, NULL))) 1971 return (error); 1972 1973 if (oflags & O_SEARCH) { 1974 oflags &= ~(int)O_SEARCH; 1975 } 1976 1977 flags = FFLAGS(oflags); 1978 if ((flags & (FREAD | FWRITE)) == 0) 1979 return (EINVAL); 1980 if ((flags & O_CREAT)) 1981 return (EINVAL); 1982 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1983 return (error); 1984 fp = nfp; 1985 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1986 if (error != 0) { 1987 goto bad; 1988 } 1989 error = vfs_fhtovp(fh, &vp); 1990 vfs_copyinfh_free(fh); 1991 if (error != 0) { 1992 goto bad; 1993 } 1994 1995 /* Now do an effective vn_open */ 1996 1997 if (vp->v_type == VSOCK) { 1998 error = EOPNOTSUPP; 1999 goto bad; 2000 } 2001 error = vn_openchk(vp, cred, flags); 2002 if (error != 0) 2003 goto bad; 2004 if (flags & O_TRUNC) { 2005 VOP_UNLOCK(vp); /* XXX */ 2006 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2007 vattr_null(&va); 2008 va.va_size = 0; 2009 error = VOP_SETATTR(vp, &va, cred); 2010 if (error) 2011 goto bad; 2012 } 2013 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2014 goto bad; 2015 if (flags & FWRITE) { 2016 mutex_enter(vp->v_interlock); 2017 vp->v_writecount++; 2018 mutex_exit(vp->v_interlock); 2019 } 2020 2021 /* done with modified vn_open, now finish what sys_open does. */ 2022 if ((error = open_setfp(l, fp, vp, indx, flags))) 2023 return error; 2024 2025 VOP_UNLOCK(vp); 2026 *retval = indx; 2027 fd_affix(p, fp, indx); 2028 return (0); 2029 2030 bad: 2031 fd_abort(p, fp, indx); 2032 if (vp != NULL) 2033 vput(vp); 2034 return (error); 2035 } 2036 2037 int 2038 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2039 { 2040 /* { 2041 syscallarg(const void *) fhp; 2042 syscallarg(size_t) fh_size; 2043 syscallarg(int) flags; 2044 } */ 2045 2046 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2047 SCARG(uap, flags), retval); 2048 } 2049 2050 int 2051 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2052 { 2053 int error; 2054 fhandle_t *fh; 2055 struct vnode *vp; 2056 2057 /* 2058 * Must be super user 2059 */ 2060 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2061 0, NULL, NULL, NULL))) 2062 return (error); 2063 2064 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2065 if (error != 0) 2066 return error; 2067 2068 error = vfs_fhtovp(fh, &vp); 2069 vfs_copyinfh_free(fh); 2070 if (error != 0) 2071 return error; 2072 2073 error = vn_stat(vp, sb); 2074 vput(vp); 2075 return error; 2076 } 2077 2078 2079 /* ARGSUSED */ 2080 int 2081 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2082 { 2083 /* { 2084 syscallarg(const void *) fhp; 2085 syscallarg(size_t) fh_size; 2086 syscallarg(struct stat *) sb; 2087 } */ 2088 struct stat sb; 2089 int error; 2090 2091 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2092 if (error) 2093 return error; 2094 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2095 } 2096 2097 int 2098 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2099 int flags) 2100 { 2101 fhandle_t *fh; 2102 struct mount *mp; 2103 struct vnode *vp; 2104 int error; 2105 2106 /* 2107 * Must be super user 2108 */ 2109 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2110 0, NULL, NULL, NULL))) 2111 return error; 2112 2113 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2114 if (error != 0) 2115 return error; 2116 2117 error = vfs_fhtovp(fh, &vp); 2118 vfs_copyinfh_free(fh); 2119 if (error != 0) 2120 return error; 2121 2122 mp = vp->v_mount; 2123 error = dostatvfs(mp, sb, l, flags, 1); 2124 vput(vp); 2125 return error; 2126 } 2127 2128 /* ARGSUSED */ 2129 int 2130 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2131 { 2132 /* { 2133 syscallarg(const void *) fhp; 2134 syscallarg(size_t) fh_size; 2135 syscallarg(struct statvfs *) buf; 2136 syscallarg(int) flags; 2137 } */ 2138 struct statvfs *sb = STATVFSBUF_GET(); 2139 int error; 2140 2141 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2142 SCARG(uap, flags)); 2143 if (error == 0) 2144 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2145 STATVFSBUF_PUT(sb); 2146 return error; 2147 } 2148 2149 /* 2150 * Create a special file. 2151 */ 2152 /* ARGSUSED */ 2153 int 2154 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2155 register_t *retval) 2156 { 2157 /* { 2158 syscallarg(const char *) path; 2159 syscallarg(mode_t) mode; 2160 syscallarg(dev_t) dev; 2161 } */ 2162 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2163 SCARG(uap, dev), retval, UIO_USERSPACE); 2164 } 2165 2166 int 2167 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2168 register_t *retval) 2169 { 2170 /* { 2171 syscallarg(int) fd; 2172 syscallarg(const char *) path; 2173 syscallarg(mode_t) mode; 2174 syscallarg(int) pad; 2175 syscallarg(dev_t) dev; 2176 } */ 2177 2178 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2179 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2180 } 2181 2182 int 2183 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2184 register_t *retval, enum uio_seg seg) 2185 { 2186 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2187 } 2188 2189 int 2190 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2191 dev_t dev, register_t *retval, enum uio_seg seg) 2192 { 2193 struct proc *p = l->l_proc; 2194 struct vnode *vp; 2195 struct vattr vattr; 2196 int error, optype; 2197 struct pathbuf *pb; 2198 struct nameidata nd; 2199 const char *pathstring; 2200 2201 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2202 0, NULL, NULL, NULL)) != 0) 2203 return (error); 2204 2205 optype = VOP_MKNOD_DESCOFFSET; 2206 2207 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2208 if (error) { 2209 return error; 2210 } 2211 pathstring = pathbuf_stringcopy_get(pb); 2212 if (pathstring == NULL) { 2213 pathbuf_destroy(pb); 2214 return ENOMEM; 2215 } 2216 2217 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2218 2219 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2220 goto out; 2221 vp = nd.ni_vp; 2222 2223 if (vp != NULL) 2224 error = EEXIST; 2225 else { 2226 vattr_null(&vattr); 2227 /* We will read cwdi->cwdi_cmask unlocked. */ 2228 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2229 vattr.va_rdev = dev; 2230 2231 switch (mode & S_IFMT) { 2232 case S_IFMT: /* used by badsect to flag bad sectors */ 2233 vattr.va_type = VBAD; 2234 break; 2235 case S_IFCHR: 2236 vattr.va_type = VCHR; 2237 break; 2238 case S_IFBLK: 2239 vattr.va_type = VBLK; 2240 break; 2241 case S_IFWHT: 2242 optype = VOP_WHITEOUT_DESCOFFSET; 2243 break; 2244 case S_IFREG: 2245 #if NVERIEXEC > 0 2246 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2247 O_CREAT); 2248 #endif /* NVERIEXEC > 0 */ 2249 vattr.va_type = VREG; 2250 vattr.va_rdev = VNOVAL; 2251 optype = VOP_CREATE_DESCOFFSET; 2252 break; 2253 default: 2254 error = EINVAL; 2255 break; 2256 } 2257 } 2258 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2259 && vattr.va_rdev == VNOVAL) 2260 error = EINVAL; 2261 if (!error) { 2262 switch (optype) { 2263 case VOP_WHITEOUT_DESCOFFSET: 2264 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2265 if (error) 2266 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2267 vput(nd.ni_dvp); 2268 break; 2269 2270 case VOP_MKNOD_DESCOFFSET: 2271 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2272 &nd.ni_cnd, &vattr); 2273 if (error == 0) 2274 vrele(nd.ni_vp); 2275 vput(nd.ni_dvp); 2276 break; 2277 2278 case VOP_CREATE_DESCOFFSET: 2279 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2280 &nd.ni_cnd, &vattr); 2281 if (error == 0) 2282 vrele(nd.ni_vp); 2283 vput(nd.ni_dvp); 2284 break; 2285 } 2286 } else { 2287 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2288 if (nd.ni_dvp == vp) 2289 vrele(nd.ni_dvp); 2290 else 2291 vput(nd.ni_dvp); 2292 if (vp) 2293 vrele(vp); 2294 } 2295 out: 2296 pathbuf_stringcopy_put(pb, pathstring); 2297 pathbuf_destroy(pb); 2298 return (error); 2299 } 2300 2301 /* 2302 * Create a named pipe. 2303 */ 2304 /* ARGSUSED */ 2305 int 2306 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2307 { 2308 /* { 2309 syscallarg(const char *) path; 2310 syscallarg(int) mode; 2311 } */ 2312 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2313 } 2314 2315 int 2316 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2317 register_t *retval) 2318 { 2319 /* { 2320 syscallarg(int) fd; 2321 syscallarg(const char *) path; 2322 syscallarg(int) mode; 2323 } */ 2324 2325 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2326 SCARG(uap, mode)); 2327 } 2328 2329 static int 2330 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2331 { 2332 struct proc *p = l->l_proc; 2333 struct vattr vattr; 2334 int error; 2335 struct pathbuf *pb; 2336 struct nameidata nd; 2337 2338 error = pathbuf_copyin(path, &pb); 2339 if (error) { 2340 return error; 2341 } 2342 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2343 2344 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2345 pathbuf_destroy(pb); 2346 return error; 2347 } 2348 if (nd.ni_vp != NULL) { 2349 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2350 if (nd.ni_dvp == nd.ni_vp) 2351 vrele(nd.ni_dvp); 2352 else 2353 vput(nd.ni_dvp); 2354 vrele(nd.ni_vp); 2355 pathbuf_destroy(pb); 2356 return (EEXIST); 2357 } 2358 vattr_null(&vattr); 2359 vattr.va_type = VFIFO; 2360 /* We will read cwdi->cwdi_cmask unlocked. */ 2361 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2362 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2363 if (error == 0) 2364 vrele(nd.ni_vp); 2365 vput(nd.ni_dvp); 2366 pathbuf_destroy(pb); 2367 return (error); 2368 } 2369 2370 /* 2371 * Make a hard file link. 2372 */ 2373 /* ARGSUSED */ 2374 int 2375 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2376 const char *link, int follow, register_t *retval) 2377 { 2378 struct vnode *vp; 2379 struct pathbuf *linkpb; 2380 struct nameidata nd; 2381 namei_simple_flags_t ns_flags; 2382 int error; 2383 2384 if (follow & AT_SYMLINK_FOLLOW) 2385 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2386 else 2387 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2388 2389 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2390 if (error != 0) 2391 return (error); 2392 error = pathbuf_copyin(link, &linkpb); 2393 if (error) { 2394 goto out1; 2395 } 2396 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2397 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2398 goto out2; 2399 if (nd.ni_vp) { 2400 error = EEXIST; 2401 goto abortop; 2402 } 2403 /* Prevent hard links on directories. */ 2404 if (vp->v_type == VDIR) { 2405 error = EPERM; 2406 goto abortop; 2407 } 2408 /* Prevent cross-mount operation. */ 2409 if (nd.ni_dvp->v_mount != vp->v_mount) { 2410 error = EXDEV; 2411 goto abortop; 2412 } 2413 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2414 VOP_UNLOCK(nd.ni_dvp); 2415 vrele(nd.ni_dvp); 2416 out2: 2417 pathbuf_destroy(linkpb); 2418 out1: 2419 vrele(vp); 2420 return (error); 2421 abortop: 2422 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2423 if (nd.ni_dvp == nd.ni_vp) 2424 vrele(nd.ni_dvp); 2425 else 2426 vput(nd.ni_dvp); 2427 if (nd.ni_vp != NULL) 2428 vrele(nd.ni_vp); 2429 goto out2; 2430 } 2431 2432 int 2433 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2434 { 2435 /* { 2436 syscallarg(const char *) path; 2437 syscallarg(const char *) link; 2438 } */ 2439 const char *path = SCARG(uap, path); 2440 const char *link = SCARG(uap, link); 2441 2442 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2443 AT_SYMLINK_FOLLOW, retval); 2444 } 2445 2446 int 2447 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2448 register_t *retval) 2449 { 2450 /* { 2451 syscallarg(int) fd1; 2452 syscallarg(const char *) name1; 2453 syscallarg(int) fd2; 2454 syscallarg(const char *) name2; 2455 syscallarg(int) flags; 2456 } */ 2457 int fd1 = SCARG(uap, fd1); 2458 const char *name1 = SCARG(uap, name1); 2459 int fd2 = SCARG(uap, fd2); 2460 const char *name2 = SCARG(uap, name2); 2461 int follow; 2462 2463 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2464 2465 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2466 } 2467 2468 2469 int 2470 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2471 { 2472 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2473 } 2474 2475 static int 2476 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2477 const char *link, enum uio_seg seg) 2478 { 2479 struct proc *p = curproc; 2480 struct vattr vattr; 2481 char *path; 2482 int error; 2483 struct pathbuf *linkpb; 2484 struct nameidata nd; 2485 2486 KASSERT(l != NULL || fdat == AT_FDCWD); 2487 2488 path = PNBUF_GET(); 2489 if (seg == UIO_USERSPACE) { 2490 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2491 goto out1; 2492 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2493 goto out1; 2494 } else { 2495 KASSERT(strlen(patharg) < MAXPATHLEN); 2496 strcpy(path, patharg); 2497 linkpb = pathbuf_create(link); 2498 if (linkpb == NULL) { 2499 error = ENOMEM; 2500 goto out1; 2501 } 2502 } 2503 ktrkuser("symlink-target", path, strlen(path)); 2504 2505 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2506 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2507 goto out2; 2508 if (nd.ni_vp) { 2509 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2510 if (nd.ni_dvp == nd.ni_vp) 2511 vrele(nd.ni_dvp); 2512 else 2513 vput(nd.ni_dvp); 2514 vrele(nd.ni_vp); 2515 error = EEXIST; 2516 goto out2; 2517 } 2518 vattr_null(&vattr); 2519 vattr.va_type = VLNK; 2520 /* We will read cwdi->cwdi_cmask unlocked. */ 2521 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2522 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2523 if (error == 0) 2524 vrele(nd.ni_vp); 2525 vput(nd.ni_dvp); 2526 out2: 2527 pathbuf_destroy(linkpb); 2528 out1: 2529 PNBUF_PUT(path); 2530 return (error); 2531 } 2532 2533 /* 2534 * Make a symbolic link. 2535 */ 2536 /* ARGSUSED */ 2537 int 2538 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2539 { 2540 /* { 2541 syscallarg(const char *) path; 2542 syscallarg(const char *) link; 2543 } */ 2544 2545 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2546 UIO_USERSPACE); 2547 } 2548 2549 int 2550 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2551 register_t *retval) 2552 { 2553 /* { 2554 syscallarg(const char *) path1; 2555 syscallarg(int) fd; 2556 syscallarg(const char *) path2; 2557 } */ 2558 2559 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2560 SCARG(uap, path2), UIO_USERSPACE); 2561 } 2562 2563 /* 2564 * Delete a whiteout from the filesystem. 2565 */ 2566 /* ARGSUSED */ 2567 int 2568 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2569 { 2570 /* { 2571 syscallarg(const char *) path; 2572 } */ 2573 int error; 2574 struct pathbuf *pb; 2575 struct nameidata nd; 2576 2577 error = pathbuf_copyin(SCARG(uap, path), &pb); 2578 if (error) { 2579 return error; 2580 } 2581 2582 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2583 error = namei(&nd); 2584 if (error) { 2585 pathbuf_destroy(pb); 2586 return (error); 2587 } 2588 2589 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2590 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2591 if (nd.ni_dvp == nd.ni_vp) 2592 vrele(nd.ni_dvp); 2593 else 2594 vput(nd.ni_dvp); 2595 if (nd.ni_vp) 2596 vrele(nd.ni_vp); 2597 pathbuf_destroy(pb); 2598 return (EEXIST); 2599 } 2600 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2601 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2602 vput(nd.ni_dvp); 2603 pathbuf_destroy(pb); 2604 return (error); 2605 } 2606 2607 /* 2608 * Delete a name from the filesystem. 2609 */ 2610 /* ARGSUSED */ 2611 int 2612 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2613 { 2614 /* { 2615 syscallarg(const char *) path; 2616 } */ 2617 2618 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2619 } 2620 2621 int 2622 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2623 register_t *retval) 2624 { 2625 /* { 2626 syscallarg(int) fd; 2627 syscallarg(const char *) path; 2628 syscallarg(int) flag; 2629 } */ 2630 2631 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2632 SCARG(uap, flag), UIO_USERSPACE); 2633 } 2634 2635 int 2636 do_sys_unlink(const char *arg, enum uio_seg seg) 2637 { 2638 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2639 } 2640 2641 static int 2642 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2643 enum uio_seg seg) 2644 { 2645 struct vnode *vp; 2646 int error; 2647 struct pathbuf *pb; 2648 struct nameidata nd; 2649 const char *pathstring; 2650 2651 KASSERT(l != NULL || fdat == AT_FDCWD); 2652 2653 error = pathbuf_maybe_copyin(arg, seg, &pb); 2654 if (error) { 2655 return error; 2656 } 2657 pathstring = pathbuf_stringcopy_get(pb); 2658 if (pathstring == NULL) { 2659 pathbuf_destroy(pb); 2660 return ENOMEM; 2661 } 2662 2663 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2664 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2665 goto out; 2666 vp = nd.ni_vp; 2667 2668 /* 2669 * The root of a mounted filesystem cannot be deleted. 2670 */ 2671 if ((vp->v_vflag & VV_ROOT) != 0) { 2672 error = EBUSY; 2673 goto abort; 2674 } 2675 2676 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2677 error = EBUSY; 2678 goto abort; 2679 } 2680 2681 /* 2682 * No rmdir "." please. 2683 */ 2684 if (nd.ni_dvp == vp) { 2685 error = EINVAL; 2686 goto abort; 2687 } 2688 2689 /* 2690 * AT_REMOVEDIR is required to remove a directory 2691 */ 2692 if (vp->v_type == VDIR) { 2693 if (!(flags & AT_REMOVEDIR)) { 2694 error = EPERM; 2695 goto abort; 2696 } else { 2697 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2698 goto out; 2699 } 2700 } 2701 2702 /* 2703 * Starting here we only deal with non directories. 2704 */ 2705 if (flags & AT_REMOVEDIR) { 2706 error = ENOTDIR; 2707 goto abort; 2708 } 2709 2710 #if NVERIEXEC > 0 2711 /* Handle remove requests for veriexec entries. */ 2712 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2713 goto abort; 2714 } 2715 #endif /* NVERIEXEC > 0 */ 2716 2717 #ifdef FILEASSOC 2718 (void)fileassoc_file_delete(vp); 2719 #endif /* FILEASSOC */ 2720 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2721 goto out; 2722 2723 abort: 2724 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2725 if (nd.ni_dvp == vp) 2726 vrele(nd.ni_dvp); 2727 else 2728 vput(nd.ni_dvp); 2729 vput(vp); 2730 2731 out: 2732 pathbuf_stringcopy_put(pb, pathstring); 2733 pathbuf_destroy(pb); 2734 return (error); 2735 } 2736 2737 /* 2738 * Reposition read/write file offset. 2739 */ 2740 int 2741 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2742 { 2743 /* { 2744 syscallarg(int) fd; 2745 syscallarg(int) pad; 2746 syscallarg(off_t) offset; 2747 syscallarg(int) whence; 2748 } */ 2749 kauth_cred_t cred = l->l_cred; 2750 file_t *fp; 2751 struct vnode *vp; 2752 struct vattr vattr; 2753 off_t newoff; 2754 int error, fd; 2755 2756 fd = SCARG(uap, fd); 2757 2758 if ((fp = fd_getfile(fd)) == NULL) 2759 return (EBADF); 2760 2761 vp = fp->f_vnode; 2762 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2763 error = ESPIPE; 2764 goto out; 2765 } 2766 2767 vn_lock(vp, LK_SHARED | LK_RETRY); 2768 2769 switch (SCARG(uap, whence)) { 2770 case SEEK_CUR: 2771 newoff = fp->f_offset + SCARG(uap, offset); 2772 break; 2773 case SEEK_END: 2774 error = VOP_GETATTR(vp, &vattr, cred); 2775 if (error) { 2776 VOP_UNLOCK(vp); 2777 goto out; 2778 } 2779 newoff = SCARG(uap, offset) + vattr.va_size; 2780 break; 2781 case SEEK_SET: 2782 newoff = SCARG(uap, offset); 2783 break; 2784 default: 2785 error = EINVAL; 2786 VOP_UNLOCK(vp); 2787 goto out; 2788 } 2789 VOP_UNLOCK(vp); 2790 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2791 *(off_t *)retval = fp->f_offset = newoff; 2792 } 2793 out: 2794 fd_putfile(fd); 2795 return (error); 2796 } 2797 2798 /* 2799 * Positional read system call. 2800 */ 2801 int 2802 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2803 { 2804 /* { 2805 syscallarg(int) fd; 2806 syscallarg(void *) buf; 2807 syscallarg(size_t) nbyte; 2808 syscallarg(off_t) offset; 2809 } */ 2810 file_t *fp; 2811 struct vnode *vp; 2812 off_t offset; 2813 int error, fd = SCARG(uap, fd); 2814 2815 if ((fp = fd_getfile(fd)) == NULL) 2816 return (EBADF); 2817 2818 if ((fp->f_flag & FREAD) == 0) { 2819 fd_putfile(fd); 2820 return (EBADF); 2821 } 2822 2823 vp = fp->f_vnode; 2824 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2825 error = ESPIPE; 2826 goto out; 2827 } 2828 2829 offset = SCARG(uap, offset); 2830 2831 /* 2832 * XXX This works because no file systems actually 2833 * XXX take any action on the seek operation. 2834 */ 2835 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2836 goto out; 2837 2838 /* dofileread() will unuse the descriptor for us */ 2839 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2840 &offset, 0, retval)); 2841 2842 out: 2843 fd_putfile(fd); 2844 return (error); 2845 } 2846 2847 /* 2848 * Positional scatter read system call. 2849 */ 2850 int 2851 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2852 { 2853 /* { 2854 syscallarg(int) fd; 2855 syscallarg(const struct iovec *) iovp; 2856 syscallarg(int) iovcnt; 2857 syscallarg(off_t) offset; 2858 } */ 2859 off_t offset = SCARG(uap, offset); 2860 2861 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2862 SCARG(uap, iovcnt), &offset, 0, retval); 2863 } 2864 2865 /* 2866 * Positional write system call. 2867 */ 2868 int 2869 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2870 { 2871 /* { 2872 syscallarg(int) fd; 2873 syscallarg(const void *) buf; 2874 syscallarg(size_t) nbyte; 2875 syscallarg(off_t) offset; 2876 } */ 2877 file_t *fp; 2878 struct vnode *vp; 2879 off_t offset; 2880 int error, fd = SCARG(uap, fd); 2881 2882 if ((fp = fd_getfile(fd)) == NULL) 2883 return (EBADF); 2884 2885 if ((fp->f_flag & FWRITE) == 0) { 2886 fd_putfile(fd); 2887 return (EBADF); 2888 } 2889 2890 vp = fp->f_vnode; 2891 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2892 error = ESPIPE; 2893 goto out; 2894 } 2895 2896 offset = SCARG(uap, offset); 2897 2898 /* 2899 * XXX This works because no file systems actually 2900 * XXX take any action on the seek operation. 2901 */ 2902 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2903 goto out; 2904 2905 /* dofilewrite() will unuse the descriptor for us */ 2906 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2907 &offset, 0, retval)); 2908 2909 out: 2910 fd_putfile(fd); 2911 return (error); 2912 } 2913 2914 /* 2915 * Positional gather write system call. 2916 */ 2917 int 2918 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2919 { 2920 /* { 2921 syscallarg(int) fd; 2922 syscallarg(const struct iovec *) iovp; 2923 syscallarg(int) iovcnt; 2924 syscallarg(off_t) offset; 2925 } */ 2926 off_t offset = SCARG(uap, offset); 2927 2928 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2929 SCARG(uap, iovcnt), &offset, 0, retval); 2930 } 2931 2932 /* 2933 * Check access permissions. 2934 */ 2935 int 2936 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2937 { 2938 /* { 2939 syscallarg(const char *) path; 2940 syscallarg(int) flags; 2941 } */ 2942 2943 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2944 SCARG(uap, flags), 0); 2945 } 2946 2947 int 2948 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2949 int mode, int flags) 2950 { 2951 kauth_cred_t cred; 2952 struct vnode *vp; 2953 int error, nd_flag, vmode; 2954 struct pathbuf *pb; 2955 struct nameidata nd; 2956 2957 CTASSERT(F_OK == 0); 2958 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2959 /* nonsense mode */ 2960 return EINVAL; 2961 } 2962 2963 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2964 if (flags & AT_SYMLINK_NOFOLLOW) 2965 nd_flag &= ~FOLLOW; 2966 2967 error = pathbuf_copyin(path, &pb); 2968 if (error) 2969 return error; 2970 2971 NDINIT(&nd, LOOKUP, nd_flag, pb); 2972 2973 /* Override default credentials */ 2974 cred = kauth_cred_dup(l->l_cred); 2975 if (!(flags & AT_EACCESS)) { 2976 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2977 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2978 } 2979 nd.ni_cnd.cn_cred = cred; 2980 2981 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2982 pathbuf_destroy(pb); 2983 goto out; 2984 } 2985 vp = nd.ni_vp; 2986 pathbuf_destroy(pb); 2987 2988 /* Flags == 0 means only check for existence. */ 2989 if (mode) { 2990 vmode = 0; 2991 if (mode & R_OK) 2992 vmode |= VREAD; 2993 if (mode & W_OK) 2994 vmode |= VWRITE; 2995 if (mode & X_OK) 2996 vmode |= VEXEC; 2997 2998 error = VOP_ACCESS(vp, vmode, cred); 2999 if (!error && (vmode & VWRITE)) 3000 error = vn_writechk(vp); 3001 } 3002 vput(vp); 3003 out: 3004 kauth_cred_free(cred); 3005 return (error); 3006 } 3007 3008 int 3009 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3010 register_t *retval) 3011 { 3012 /* { 3013 syscallarg(int) fd; 3014 syscallarg(const char *) path; 3015 syscallarg(int) amode; 3016 syscallarg(int) flag; 3017 } */ 3018 3019 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3020 SCARG(uap, amode), SCARG(uap, flag)); 3021 } 3022 3023 /* 3024 * Common code for all sys_stat functions, including compat versions. 3025 */ 3026 int 3027 do_sys_stat(const char *userpath, unsigned int nd_flag, 3028 struct stat *sb) 3029 { 3030 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3031 } 3032 3033 int 3034 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3035 unsigned int nd_flag, struct stat *sb) 3036 { 3037 int error; 3038 struct pathbuf *pb; 3039 struct nameidata nd; 3040 3041 KASSERT(l != NULL || fdat == AT_FDCWD); 3042 3043 error = pathbuf_copyin(userpath, &pb); 3044 if (error) { 3045 return error; 3046 } 3047 3048 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3049 3050 error = fd_nameiat(l, fdat, &nd); 3051 if (error != 0) { 3052 pathbuf_destroy(pb); 3053 return error; 3054 } 3055 error = vn_stat(nd.ni_vp, sb); 3056 vput(nd.ni_vp); 3057 pathbuf_destroy(pb); 3058 return error; 3059 } 3060 3061 /* 3062 * Get file status; this version follows links. 3063 */ 3064 /* ARGSUSED */ 3065 int 3066 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3067 { 3068 /* { 3069 syscallarg(const char *) path; 3070 syscallarg(struct stat *) ub; 3071 } */ 3072 struct stat sb; 3073 int error; 3074 3075 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3076 if (error) 3077 return error; 3078 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3079 } 3080 3081 /* 3082 * Get file status; this version does not follow links. 3083 */ 3084 /* ARGSUSED */ 3085 int 3086 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3087 { 3088 /* { 3089 syscallarg(const char *) path; 3090 syscallarg(struct stat *) ub; 3091 } */ 3092 struct stat sb; 3093 int error; 3094 3095 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3096 if (error) 3097 return error; 3098 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3099 } 3100 3101 int 3102 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3103 register_t *retval) 3104 { 3105 /* { 3106 syscallarg(int) fd; 3107 syscallarg(const char *) path; 3108 syscallarg(struct stat *) buf; 3109 syscallarg(int) flag; 3110 } */ 3111 unsigned int nd_flag; 3112 struct stat sb; 3113 int error; 3114 3115 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3116 nd_flag = NOFOLLOW; 3117 else 3118 nd_flag = FOLLOW; 3119 3120 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3121 &sb); 3122 if (error) 3123 return error; 3124 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3125 } 3126 3127 /* 3128 * Get configurable pathname variables. 3129 */ 3130 /* ARGSUSED */ 3131 int 3132 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3133 { 3134 /* { 3135 syscallarg(const char *) path; 3136 syscallarg(int) name; 3137 } */ 3138 int error; 3139 struct pathbuf *pb; 3140 struct nameidata nd; 3141 3142 error = pathbuf_copyin(SCARG(uap, path), &pb); 3143 if (error) { 3144 return error; 3145 } 3146 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3147 if ((error = namei(&nd)) != 0) { 3148 pathbuf_destroy(pb); 3149 return (error); 3150 } 3151 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3152 vput(nd.ni_vp); 3153 pathbuf_destroy(pb); 3154 return (error); 3155 } 3156 3157 /* 3158 * Return target name of a symbolic link. 3159 */ 3160 /* ARGSUSED */ 3161 int 3162 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3163 register_t *retval) 3164 { 3165 /* { 3166 syscallarg(const char *) path; 3167 syscallarg(char *) buf; 3168 syscallarg(size_t) count; 3169 } */ 3170 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3171 SCARG(uap, buf), SCARG(uap, count), retval); 3172 } 3173 3174 static int 3175 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3176 size_t count, register_t *retval) 3177 { 3178 struct vnode *vp; 3179 struct iovec aiov; 3180 struct uio auio; 3181 int error; 3182 struct pathbuf *pb; 3183 struct nameidata nd; 3184 3185 error = pathbuf_copyin(path, &pb); 3186 if (error) { 3187 return error; 3188 } 3189 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3190 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3191 pathbuf_destroy(pb); 3192 return error; 3193 } 3194 vp = nd.ni_vp; 3195 pathbuf_destroy(pb); 3196 if (vp->v_type != VLNK) 3197 error = EINVAL; 3198 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3199 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3200 aiov.iov_base = buf; 3201 aiov.iov_len = count; 3202 auio.uio_iov = &aiov; 3203 auio.uio_iovcnt = 1; 3204 auio.uio_offset = 0; 3205 auio.uio_rw = UIO_READ; 3206 KASSERT(l == curlwp); 3207 auio.uio_vmspace = l->l_proc->p_vmspace; 3208 auio.uio_resid = count; 3209 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3210 *retval = count - auio.uio_resid; 3211 } 3212 vput(vp); 3213 return (error); 3214 } 3215 3216 int 3217 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3218 register_t *retval) 3219 { 3220 /* { 3221 syscallarg(int) fd; 3222 syscallarg(const char *) path; 3223 syscallarg(char *) buf; 3224 syscallarg(size_t) bufsize; 3225 } */ 3226 3227 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3228 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3229 } 3230 3231 /* 3232 * Change flags of a file given a path name. 3233 */ 3234 /* ARGSUSED */ 3235 int 3236 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3237 { 3238 /* { 3239 syscallarg(const char *) path; 3240 syscallarg(u_long) flags; 3241 } */ 3242 struct vnode *vp; 3243 int error; 3244 3245 error = namei_simple_user(SCARG(uap, path), 3246 NSM_FOLLOW_TRYEMULROOT, &vp); 3247 if (error != 0) 3248 return (error); 3249 error = change_flags(vp, SCARG(uap, flags), l); 3250 vput(vp); 3251 return (error); 3252 } 3253 3254 /* 3255 * Change flags of a file given a file descriptor. 3256 */ 3257 /* ARGSUSED */ 3258 int 3259 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3260 { 3261 /* { 3262 syscallarg(int) fd; 3263 syscallarg(u_long) flags; 3264 } */ 3265 struct vnode *vp; 3266 file_t *fp; 3267 int error; 3268 3269 /* fd_getvnode() will use the descriptor for us */ 3270 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3271 return (error); 3272 vp = fp->f_vnode; 3273 error = change_flags(vp, SCARG(uap, flags), l); 3274 VOP_UNLOCK(vp); 3275 fd_putfile(SCARG(uap, fd)); 3276 return (error); 3277 } 3278 3279 /* 3280 * Change flags of a file given a path name; this version does 3281 * not follow links. 3282 */ 3283 int 3284 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3285 { 3286 /* { 3287 syscallarg(const char *) path; 3288 syscallarg(u_long) flags; 3289 } */ 3290 struct vnode *vp; 3291 int error; 3292 3293 error = namei_simple_user(SCARG(uap, path), 3294 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3295 if (error != 0) 3296 return (error); 3297 error = change_flags(vp, SCARG(uap, flags), l); 3298 vput(vp); 3299 return (error); 3300 } 3301 3302 /* 3303 * Common routine to change flags of a file. 3304 */ 3305 int 3306 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3307 { 3308 struct vattr vattr; 3309 int error; 3310 3311 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3312 3313 vattr_null(&vattr); 3314 vattr.va_flags = flags; 3315 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3316 3317 return (error); 3318 } 3319 3320 /* 3321 * Change mode of a file given path name; this version follows links. 3322 */ 3323 /* ARGSUSED */ 3324 int 3325 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3326 { 3327 /* { 3328 syscallarg(const char *) path; 3329 syscallarg(int) mode; 3330 } */ 3331 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3332 SCARG(uap, mode), 0); 3333 } 3334 3335 int 3336 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3337 { 3338 int error; 3339 struct vnode *vp; 3340 namei_simple_flags_t ns_flag; 3341 3342 if (flags & AT_SYMLINK_NOFOLLOW) 3343 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3344 else 3345 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3346 3347 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3348 if (error != 0) 3349 return error; 3350 3351 error = change_mode(vp, mode, l); 3352 3353 vrele(vp); 3354 3355 return (error); 3356 } 3357 3358 /* 3359 * Change mode of a file given a file descriptor. 3360 */ 3361 /* ARGSUSED */ 3362 int 3363 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3364 { 3365 /* { 3366 syscallarg(int) fd; 3367 syscallarg(int) mode; 3368 } */ 3369 file_t *fp; 3370 int error; 3371 3372 /* fd_getvnode() will use the descriptor for us */ 3373 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3374 return (error); 3375 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3376 fd_putfile(SCARG(uap, fd)); 3377 return (error); 3378 } 3379 3380 int 3381 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3382 register_t *retval) 3383 { 3384 /* { 3385 syscallarg(int) fd; 3386 syscallarg(const char *) path; 3387 syscallarg(int) mode; 3388 syscallarg(int) flag; 3389 } */ 3390 3391 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3392 SCARG(uap, mode), SCARG(uap, flag)); 3393 } 3394 3395 /* 3396 * Change mode of a file given path name; this version does not follow links. 3397 */ 3398 /* ARGSUSED */ 3399 int 3400 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3401 { 3402 /* { 3403 syscallarg(const char *) path; 3404 syscallarg(int) mode; 3405 } */ 3406 int error; 3407 struct vnode *vp; 3408 3409 error = namei_simple_user(SCARG(uap, path), 3410 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3411 if (error != 0) 3412 return (error); 3413 3414 error = change_mode(vp, SCARG(uap, mode), l); 3415 3416 vrele(vp); 3417 return (error); 3418 } 3419 3420 /* 3421 * Common routine to set mode given a vnode. 3422 */ 3423 static int 3424 change_mode(struct vnode *vp, int mode, struct lwp *l) 3425 { 3426 struct vattr vattr; 3427 int error; 3428 3429 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3430 vattr_null(&vattr); 3431 vattr.va_mode = mode & ALLPERMS; 3432 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3433 VOP_UNLOCK(vp); 3434 return (error); 3435 } 3436 3437 /* 3438 * Set ownership given a path name; this version follows links. 3439 */ 3440 /* ARGSUSED */ 3441 int 3442 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3443 { 3444 /* { 3445 syscallarg(const char *) path; 3446 syscallarg(uid_t) uid; 3447 syscallarg(gid_t) gid; 3448 } */ 3449 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3450 SCARG(uap, gid), 0); 3451 } 3452 3453 int 3454 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3455 gid_t gid, int flags) 3456 { 3457 int error; 3458 struct vnode *vp; 3459 namei_simple_flags_t ns_flag; 3460 3461 if (flags & AT_SYMLINK_NOFOLLOW) 3462 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3463 else 3464 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3465 3466 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3467 if (error != 0) 3468 return error; 3469 3470 error = change_owner(vp, uid, gid, l, 0); 3471 3472 vrele(vp); 3473 3474 return (error); 3475 } 3476 3477 /* 3478 * Set ownership given a path name; this version follows links. 3479 * Provides POSIX semantics. 3480 */ 3481 /* ARGSUSED */ 3482 int 3483 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3484 { 3485 /* { 3486 syscallarg(const char *) path; 3487 syscallarg(uid_t) uid; 3488 syscallarg(gid_t) gid; 3489 } */ 3490 int error; 3491 struct vnode *vp; 3492 3493 error = namei_simple_user(SCARG(uap, path), 3494 NSM_FOLLOW_TRYEMULROOT, &vp); 3495 if (error != 0) 3496 return (error); 3497 3498 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3499 3500 vrele(vp); 3501 return (error); 3502 } 3503 3504 /* 3505 * Set ownership given a file descriptor. 3506 */ 3507 /* ARGSUSED */ 3508 int 3509 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3510 { 3511 /* { 3512 syscallarg(int) fd; 3513 syscallarg(uid_t) uid; 3514 syscallarg(gid_t) gid; 3515 } */ 3516 int error; 3517 file_t *fp; 3518 3519 /* fd_getvnode() will use the descriptor for us */ 3520 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3521 return (error); 3522 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3523 l, 0); 3524 fd_putfile(SCARG(uap, fd)); 3525 return (error); 3526 } 3527 3528 int 3529 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3530 register_t *retval) 3531 { 3532 /* { 3533 syscallarg(int) fd; 3534 syscallarg(const char *) path; 3535 syscallarg(uid_t) owner; 3536 syscallarg(gid_t) group; 3537 syscallarg(int) flag; 3538 } */ 3539 3540 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3541 SCARG(uap, owner), SCARG(uap, group), 3542 SCARG(uap, flag)); 3543 } 3544 3545 /* 3546 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3547 */ 3548 /* ARGSUSED */ 3549 int 3550 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3551 { 3552 /* { 3553 syscallarg(int) fd; 3554 syscallarg(uid_t) uid; 3555 syscallarg(gid_t) gid; 3556 } */ 3557 int error; 3558 file_t *fp; 3559 3560 /* fd_getvnode() will use the descriptor for us */ 3561 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3562 return (error); 3563 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3564 l, 1); 3565 fd_putfile(SCARG(uap, fd)); 3566 return (error); 3567 } 3568 3569 /* 3570 * Set ownership given a path name; this version does not follow links. 3571 */ 3572 /* ARGSUSED */ 3573 int 3574 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3575 { 3576 /* { 3577 syscallarg(const char *) path; 3578 syscallarg(uid_t) uid; 3579 syscallarg(gid_t) gid; 3580 } */ 3581 int error; 3582 struct vnode *vp; 3583 3584 error = namei_simple_user(SCARG(uap, path), 3585 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3586 if (error != 0) 3587 return (error); 3588 3589 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3590 3591 vrele(vp); 3592 return (error); 3593 } 3594 3595 /* 3596 * Set ownership given a path name; this version does not follow links. 3597 * Provides POSIX/XPG semantics. 3598 */ 3599 /* ARGSUSED */ 3600 int 3601 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3602 { 3603 /* { 3604 syscallarg(const char *) path; 3605 syscallarg(uid_t) uid; 3606 syscallarg(gid_t) gid; 3607 } */ 3608 int error; 3609 struct vnode *vp; 3610 3611 error = namei_simple_user(SCARG(uap, path), 3612 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3613 if (error != 0) 3614 return (error); 3615 3616 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3617 3618 vrele(vp); 3619 return (error); 3620 } 3621 3622 /* 3623 * Common routine to set ownership given a vnode. 3624 */ 3625 static int 3626 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3627 int posix_semantics) 3628 { 3629 struct vattr vattr; 3630 mode_t newmode; 3631 int error; 3632 3633 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3634 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3635 goto out; 3636 3637 #define CHANGED(x) ((int)(x) != -1) 3638 newmode = vattr.va_mode; 3639 if (posix_semantics) { 3640 /* 3641 * POSIX/XPG semantics: if the caller is not the super-user, 3642 * clear set-user-id and set-group-id bits. Both POSIX and 3643 * the XPG consider the behaviour for calls by the super-user 3644 * implementation-defined; we leave the set-user-id and set- 3645 * group-id settings intact in that case. 3646 */ 3647 if (vattr.va_mode & S_ISUID) { 3648 if (kauth_authorize_vnode(l->l_cred, 3649 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3650 newmode &= ~S_ISUID; 3651 } 3652 if (vattr.va_mode & S_ISGID) { 3653 if (kauth_authorize_vnode(l->l_cred, 3654 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3655 newmode &= ~S_ISGID; 3656 } 3657 } else { 3658 /* 3659 * NetBSD semantics: when changing owner and/or group, 3660 * clear the respective bit(s). 3661 */ 3662 if (CHANGED(uid)) 3663 newmode &= ~S_ISUID; 3664 if (CHANGED(gid)) 3665 newmode &= ~S_ISGID; 3666 } 3667 /* Update va_mode iff altered. */ 3668 if (vattr.va_mode == newmode) 3669 newmode = VNOVAL; 3670 3671 vattr_null(&vattr); 3672 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3673 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3674 vattr.va_mode = newmode; 3675 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3676 #undef CHANGED 3677 3678 out: 3679 VOP_UNLOCK(vp); 3680 return (error); 3681 } 3682 3683 /* 3684 * Set the access and modification times given a path name; this 3685 * version follows links. 3686 */ 3687 /* ARGSUSED */ 3688 int 3689 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3690 register_t *retval) 3691 { 3692 /* { 3693 syscallarg(const char *) path; 3694 syscallarg(const struct timeval *) tptr; 3695 } */ 3696 3697 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3698 SCARG(uap, tptr), UIO_USERSPACE); 3699 } 3700 3701 /* 3702 * Set the access and modification times given a file descriptor. 3703 */ 3704 /* ARGSUSED */ 3705 int 3706 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3707 register_t *retval) 3708 { 3709 /* { 3710 syscallarg(int) fd; 3711 syscallarg(const struct timeval *) tptr; 3712 } */ 3713 int error; 3714 file_t *fp; 3715 3716 /* fd_getvnode() will use the descriptor for us */ 3717 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3718 return (error); 3719 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3720 UIO_USERSPACE); 3721 fd_putfile(SCARG(uap, fd)); 3722 return (error); 3723 } 3724 3725 int 3726 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3727 register_t *retval) 3728 { 3729 /* { 3730 syscallarg(int) fd; 3731 syscallarg(const struct timespec *) tptr; 3732 } */ 3733 int error; 3734 file_t *fp; 3735 3736 /* fd_getvnode() will use the descriptor for us */ 3737 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3738 return (error); 3739 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3740 SCARG(uap, tptr), UIO_USERSPACE); 3741 fd_putfile(SCARG(uap, fd)); 3742 return (error); 3743 } 3744 3745 /* 3746 * Set the access and modification times given a path name; this 3747 * version does not follow links. 3748 */ 3749 int 3750 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3751 register_t *retval) 3752 { 3753 /* { 3754 syscallarg(const char *) path; 3755 syscallarg(const struct timeval *) tptr; 3756 } */ 3757 3758 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3759 SCARG(uap, tptr), UIO_USERSPACE); 3760 } 3761 3762 int 3763 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3764 register_t *retval) 3765 { 3766 /* { 3767 syscallarg(int) fd; 3768 syscallarg(const char *) path; 3769 syscallarg(const struct timespec *) tptr; 3770 syscallarg(int) flag; 3771 } */ 3772 int follow; 3773 const struct timespec *tptr; 3774 int error; 3775 3776 tptr = SCARG(uap, tptr); 3777 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3778 3779 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3780 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3781 3782 return error; 3783 } 3784 3785 /* 3786 * Common routine to set access and modification times given a vnode. 3787 */ 3788 int 3789 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3790 const struct timespec *tptr, enum uio_seg seg) 3791 { 3792 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3793 } 3794 3795 int 3796 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3797 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3798 { 3799 struct vattr vattr; 3800 int error, dorele = 0; 3801 namei_simple_flags_t sflags; 3802 bool vanull, setbirthtime; 3803 struct timespec ts[2]; 3804 3805 KASSERT(l != NULL || fdat == AT_FDCWD); 3806 3807 /* 3808 * I have checked all callers and they pass either FOLLOW, 3809 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3810 * is 0. More to the point, they don't pass anything else. 3811 * Let's keep it that way at least until the namei interfaces 3812 * are fully sanitized. 3813 */ 3814 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3815 sflags = (flag == FOLLOW) ? 3816 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3817 3818 if (tptr == NULL) { 3819 vanull = true; 3820 nanotime(&ts[0]); 3821 ts[1] = ts[0]; 3822 } else { 3823 vanull = false; 3824 if (seg != UIO_SYSSPACE) { 3825 error = copyin(tptr, ts, sizeof (ts)); 3826 if (error != 0) 3827 return error; 3828 } else { 3829 ts[0] = tptr[0]; 3830 ts[1] = tptr[1]; 3831 } 3832 } 3833 3834 if (ts[0].tv_nsec == UTIME_NOW) { 3835 nanotime(&ts[0]); 3836 if (ts[1].tv_nsec == UTIME_NOW) { 3837 vanull = true; 3838 ts[1] = ts[0]; 3839 } 3840 } else if (ts[1].tv_nsec == UTIME_NOW) 3841 nanotime(&ts[1]); 3842 3843 if (vp == NULL) { 3844 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3845 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3846 if (error != 0) 3847 return error; 3848 dorele = 1; 3849 } 3850 3851 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3852 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3853 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3854 vattr_null(&vattr); 3855 3856 if (ts[0].tv_nsec != UTIME_OMIT) 3857 vattr.va_atime = ts[0]; 3858 3859 if (ts[1].tv_nsec != UTIME_OMIT) { 3860 vattr.va_mtime = ts[1]; 3861 if (setbirthtime) 3862 vattr.va_birthtime = ts[1]; 3863 } 3864 3865 if (vanull) 3866 vattr.va_vaflags |= VA_UTIMES_NULL; 3867 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3868 VOP_UNLOCK(vp); 3869 3870 if (dorele != 0) 3871 vrele(vp); 3872 3873 return error; 3874 } 3875 3876 int 3877 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3878 const struct timeval *tptr, enum uio_seg seg) 3879 { 3880 struct timespec ts[2]; 3881 struct timespec *tsptr = NULL; 3882 int error; 3883 3884 if (tptr != NULL) { 3885 struct timeval tv[2]; 3886 3887 if (seg != UIO_SYSSPACE) { 3888 error = copyin(tptr, tv, sizeof (tv)); 3889 if (error != 0) 3890 return error; 3891 tptr = tv; 3892 } 3893 3894 if ((tv[0].tv_usec == UTIME_NOW) || 3895 (tv[0].tv_usec == UTIME_OMIT)) 3896 ts[0].tv_nsec = tv[0].tv_usec; 3897 else 3898 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3899 3900 if ((tv[1].tv_usec == UTIME_NOW) || 3901 (tv[1].tv_usec == UTIME_OMIT)) 3902 ts[1].tv_nsec = tv[1].tv_usec; 3903 else 3904 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3905 3906 tsptr = &ts[0]; 3907 } 3908 3909 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3910 } 3911 3912 /* 3913 * Truncate a file given its path name. 3914 */ 3915 /* ARGSUSED */ 3916 int 3917 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3918 { 3919 /* { 3920 syscallarg(const char *) path; 3921 syscallarg(int) pad; 3922 syscallarg(off_t) length; 3923 } */ 3924 struct vnode *vp; 3925 struct vattr vattr; 3926 int error; 3927 3928 if (SCARG(uap, length) < 0) 3929 return EINVAL; 3930 3931 error = namei_simple_user(SCARG(uap, path), 3932 NSM_FOLLOW_TRYEMULROOT, &vp); 3933 if (error != 0) 3934 return (error); 3935 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3936 if (vp->v_type == VDIR) 3937 error = EISDIR; 3938 else if ((error = vn_writechk(vp)) == 0 && 3939 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3940 vattr_null(&vattr); 3941 vattr.va_size = SCARG(uap, length); 3942 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3943 } 3944 vput(vp); 3945 return (error); 3946 } 3947 3948 /* 3949 * Truncate a file given a file descriptor. 3950 */ 3951 /* ARGSUSED */ 3952 int 3953 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3954 { 3955 /* { 3956 syscallarg(int) fd; 3957 syscallarg(int) pad; 3958 syscallarg(off_t) length; 3959 } */ 3960 struct vattr vattr; 3961 struct vnode *vp; 3962 file_t *fp; 3963 int error; 3964 3965 if (SCARG(uap, length) < 0) 3966 return EINVAL; 3967 3968 /* fd_getvnode() will use the descriptor for us */ 3969 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3970 return (error); 3971 if ((fp->f_flag & FWRITE) == 0) { 3972 error = EINVAL; 3973 goto out; 3974 } 3975 vp = fp->f_vnode; 3976 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3977 if (vp->v_type == VDIR) 3978 error = EISDIR; 3979 else if ((error = vn_writechk(vp)) == 0) { 3980 vattr_null(&vattr); 3981 vattr.va_size = SCARG(uap, length); 3982 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3983 } 3984 VOP_UNLOCK(vp); 3985 out: 3986 fd_putfile(SCARG(uap, fd)); 3987 return (error); 3988 } 3989 3990 /* 3991 * Sync an open file. 3992 */ 3993 /* ARGSUSED */ 3994 int 3995 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3996 { 3997 /* { 3998 syscallarg(int) fd; 3999 } */ 4000 struct vnode *vp; 4001 file_t *fp; 4002 int error; 4003 4004 /* fd_getvnode() will use the descriptor for us */ 4005 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4006 return (error); 4007 vp = fp->f_vnode; 4008 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4009 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4010 VOP_UNLOCK(vp); 4011 fd_putfile(SCARG(uap, fd)); 4012 return (error); 4013 } 4014 4015 /* 4016 * Sync a range of file data. API modeled after that found in AIX. 4017 * 4018 * FDATASYNC indicates that we need only save enough metadata to be able 4019 * to re-read the written data. Note we duplicate AIX's requirement that 4020 * the file be open for writing. 4021 */ 4022 /* ARGSUSED */ 4023 int 4024 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4025 { 4026 /* { 4027 syscallarg(int) fd; 4028 syscallarg(int) flags; 4029 syscallarg(off_t) start; 4030 syscallarg(off_t) length; 4031 } */ 4032 struct vnode *vp; 4033 file_t *fp; 4034 int flags, nflags; 4035 off_t s, e, len; 4036 int error; 4037 4038 /* fd_getvnode() will use the descriptor for us */ 4039 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4040 return (error); 4041 4042 if ((fp->f_flag & FWRITE) == 0) { 4043 error = EBADF; 4044 goto out; 4045 } 4046 4047 flags = SCARG(uap, flags); 4048 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4049 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4050 error = EINVAL; 4051 goto out; 4052 } 4053 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4054 if (flags & FDATASYNC) 4055 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4056 else 4057 nflags = FSYNC_WAIT; 4058 if (flags & FDISKSYNC) 4059 nflags |= FSYNC_CACHE; 4060 4061 len = SCARG(uap, length); 4062 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4063 if (len) { 4064 s = SCARG(uap, start); 4065 e = s + len; 4066 if (e < s) { 4067 error = EINVAL; 4068 goto out; 4069 } 4070 } else { 4071 e = 0; 4072 s = 0; 4073 } 4074 4075 vp = fp->f_vnode; 4076 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4077 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4078 VOP_UNLOCK(vp); 4079 out: 4080 fd_putfile(SCARG(uap, fd)); 4081 return (error); 4082 } 4083 4084 /* 4085 * Sync the data of an open file. 4086 */ 4087 /* ARGSUSED */ 4088 int 4089 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4090 { 4091 /* { 4092 syscallarg(int) fd; 4093 } */ 4094 struct vnode *vp; 4095 file_t *fp; 4096 int error; 4097 4098 /* fd_getvnode() will use the descriptor for us */ 4099 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4100 return (error); 4101 if ((fp->f_flag & FWRITE) == 0) { 4102 fd_putfile(SCARG(uap, fd)); 4103 return (EBADF); 4104 } 4105 vp = fp->f_vnode; 4106 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4107 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4108 VOP_UNLOCK(vp); 4109 fd_putfile(SCARG(uap, fd)); 4110 return (error); 4111 } 4112 4113 /* 4114 * Rename files, (standard) BSD semantics frontend. 4115 */ 4116 /* ARGSUSED */ 4117 int 4118 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4119 { 4120 /* { 4121 syscallarg(const char *) from; 4122 syscallarg(const char *) to; 4123 } */ 4124 4125 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4126 SCARG(uap, to), UIO_USERSPACE, 0)); 4127 } 4128 4129 int 4130 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4131 register_t *retval) 4132 { 4133 /* { 4134 syscallarg(int) fromfd; 4135 syscallarg(const char *) from; 4136 syscallarg(int) tofd; 4137 syscallarg(const char *) to; 4138 } */ 4139 4140 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4141 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4142 } 4143 4144 /* 4145 * Rename files, POSIX semantics frontend. 4146 */ 4147 /* ARGSUSED */ 4148 int 4149 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4150 { 4151 /* { 4152 syscallarg(const char *) from; 4153 syscallarg(const char *) to; 4154 } */ 4155 4156 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4157 SCARG(uap, to), UIO_USERSPACE, 1)); 4158 } 4159 4160 /* 4161 * Rename files. Source and destination must either both be directories, 4162 * or both not be directories. If target is a directory, it must be empty. 4163 * If `from' and `to' refer to the same object, the value of the `retain' 4164 * argument is used to determine whether `from' will be 4165 * 4166 * (retain == 0) deleted unless `from' and `to' refer to the same 4167 * object in the file system's name space (BSD). 4168 * (retain == 1) always retained (POSIX). 4169 * 4170 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4171 */ 4172 int 4173 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4174 { 4175 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4176 } 4177 4178 static int 4179 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4180 const char *to, enum uio_seg seg, int retain) 4181 { 4182 struct pathbuf *fpb, *tpb; 4183 struct nameidata fnd, tnd; 4184 struct vnode *fdvp, *fvp; 4185 struct vnode *tdvp, *tvp; 4186 struct mount *mp, *tmp; 4187 int error; 4188 4189 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4190 4191 error = pathbuf_maybe_copyin(from, seg, &fpb); 4192 if (error) 4193 goto out0; 4194 KASSERT(fpb != NULL); 4195 4196 error = pathbuf_maybe_copyin(to, seg, &tpb); 4197 if (error) 4198 goto out1; 4199 KASSERT(tpb != NULL); 4200 4201 /* 4202 * Lookup from. 4203 * 4204 * XXX LOCKPARENT is wrong because we don't actually want it 4205 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4206 * insane, so for the time being we need to leave it like this. 4207 */ 4208 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4209 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4210 goto out2; 4211 4212 /* 4213 * Pull out the important results of the lookup, fdvp and fvp. 4214 * Of course, fvp is bogus because we're about to unlock fdvp. 4215 */ 4216 fdvp = fnd.ni_dvp; 4217 fvp = fnd.ni_vp; 4218 KASSERT(fdvp != NULL); 4219 KASSERT(fvp != NULL); 4220 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4221 4222 /* 4223 * Make sure neither fdvp nor fvp is locked. 4224 */ 4225 if (fdvp != fvp) 4226 VOP_UNLOCK(fdvp); 4227 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4228 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4229 4230 /* 4231 * Reject renaming `.' and `..'. Can't do this until after 4232 * namei because we need namei's parsing to find the final 4233 * component name. (namei should just leave us with the final 4234 * component name and not look it up itself, but anyway...) 4235 * 4236 * This was here before because we used to relookup from 4237 * instead of to and relookup requires the caller to check 4238 * this, but now file systems may depend on this check, so we 4239 * must retain it until the file systems are all rototilled. 4240 */ 4241 if (((fnd.ni_cnd.cn_namelen == 1) && 4242 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4243 ((fnd.ni_cnd.cn_namelen == 2) && 4244 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4245 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4246 error = EINVAL; /* XXX EISDIR? */ 4247 goto abort0; 4248 } 4249 4250 /* 4251 * Lookup to. 4252 * 4253 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4254 * fvp here to decide whether to add CREATEDIR is a load of 4255 * bollocks because fvp might be the wrong node by now, since 4256 * fdvp is unlocked. 4257 * 4258 * XXX Why not pass CREATEDIR always? 4259 */ 4260 NDINIT(&tnd, RENAME, 4261 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4262 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4263 tpb); 4264 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4265 goto abort0; 4266 4267 /* 4268 * Pull out the important results of the lookup, tdvp and tvp. 4269 * Of course, tvp is bogus because we're about to unlock tdvp. 4270 */ 4271 tdvp = tnd.ni_dvp; 4272 tvp = tnd.ni_vp; 4273 KASSERT(tdvp != NULL); 4274 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4275 4276 /* 4277 * Make sure neither tdvp nor tvp is locked. 4278 */ 4279 if (tdvp != tvp) 4280 VOP_UNLOCK(tdvp); 4281 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4282 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4283 4284 /* 4285 * Reject renaming onto `.' or `..'. relookup is unhappy with 4286 * these, which is why we must do this here. Once upon a time 4287 * we relooked up from instead of to, and consequently didn't 4288 * need this check, but now that we relookup to instead of 4289 * from, we need this; and we shall need it forever forward 4290 * until the VOP_RENAME protocol changes, because file systems 4291 * will no doubt begin to depend on this check. 4292 */ 4293 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4294 error = EISDIR; 4295 goto abort1; 4296 } 4297 if ((tnd.ni_cnd.cn_namelen == 2) && 4298 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4299 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4300 error = EINVAL; 4301 goto abort1; 4302 } 4303 4304 /* 4305 * Get the mount point. If the file system has been unmounted, 4306 * which it may be because we're not holding any vnode locks, 4307 * then v_mount will be NULL. We're not really supposed to 4308 * read v_mount without holding the vnode lock, but since we 4309 * have fdvp referenced, if fdvp->v_mount changes then at worst 4310 * it will be set to NULL, not changed to another mount point. 4311 * And, of course, since it is up to the file system to 4312 * determine the real lock order, we can't lock both fdvp and 4313 * tdvp at the same time. 4314 */ 4315 mp = fdvp->v_mount; 4316 if (mp == NULL) { 4317 error = ENOENT; 4318 goto abort1; 4319 } 4320 4321 /* 4322 * Make sure the mount points match. Again, although we don't 4323 * hold any vnode locks, the v_mount fields may change -- but 4324 * at worst they will change to NULL, so this will never become 4325 * a cross-device rename, because we hold vnode references. 4326 * 4327 * XXX Because nothing is locked and the compiler may reorder 4328 * things here, unmounting the file system at an inopportune 4329 * moment may cause rename to fail with EXDEV when it really 4330 * should fail with ENOENT. 4331 */ 4332 tmp = tdvp->v_mount; 4333 if (tmp == NULL) { 4334 error = ENOENT; 4335 goto abort1; 4336 } 4337 4338 if (mp != tmp) { 4339 error = EXDEV; 4340 goto abort1; 4341 } 4342 4343 /* 4344 * Take the vfs rename lock to avoid cross-directory screw cases. 4345 * Nothing is locked currently, so taking this lock is safe. 4346 */ 4347 error = VFS_RENAMELOCK_ENTER(mp); 4348 if (error) 4349 goto abort1; 4350 4351 /* 4352 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4353 * and nothing is locked except for the vfs rename lock. 4354 * 4355 * The next step is a little rain dance to conform to the 4356 * insane lock protocol, even though it does nothing to ward 4357 * off race conditions. 4358 * 4359 * We need tdvp and tvp to be locked. However, because we have 4360 * unlocked tdvp in order to hold no locks while we take the 4361 * vfs rename lock, tvp may be wrong here, and we can't safely 4362 * lock it even if the sensible file systems will just unlock 4363 * it straight away. Consequently, we must lock tdvp and then 4364 * relookup tvp to get it locked. 4365 * 4366 * Finally, because the VOP_RENAME protocol is brain-damaged 4367 * and various file systems insanely depend on the semantics of 4368 * this brain damage, the lookup of to must be the last lookup 4369 * before VOP_RENAME. 4370 */ 4371 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4372 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4373 if (error) 4374 goto abort2; 4375 4376 /* 4377 * Drop the old tvp and pick up the new one -- which might be 4378 * the same, but that doesn't matter to us. After this, tdvp 4379 * and tvp should both be locked. 4380 */ 4381 if (tvp != NULL) 4382 vrele(tvp); 4383 tvp = tnd.ni_vp; 4384 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4385 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4386 4387 /* 4388 * The old do_sys_rename had various consistency checks here 4389 * involving fvp and tvp. fvp is bogus already here, and tvp 4390 * will become bogus soon in any sensible file system, so the 4391 * only purpose in putting these checks here is to give lip 4392 * service to these screw cases and to acknowledge that they 4393 * exist, not actually to handle them, but here you go 4394 * anyway... 4395 */ 4396 4397 /* 4398 * Acknowledge that directories and non-directories aren't 4399 * suposed to mix. 4400 */ 4401 if (tvp != NULL) { 4402 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4403 error = ENOTDIR; 4404 goto abort3; 4405 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4406 error = EISDIR; 4407 goto abort3; 4408 } 4409 } 4410 4411 /* 4412 * Acknowledge some random screw case, among the dozens that 4413 * might arise. 4414 */ 4415 if (fvp == tdvp) { 4416 error = EINVAL; 4417 goto abort3; 4418 } 4419 4420 /* 4421 * Acknowledge that POSIX has a wacky screw case. 4422 * 4423 * XXX Eventually the retain flag needs to be passed on to 4424 * VOP_RENAME. 4425 */ 4426 if (fvp == tvp) { 4427 if (retain) { 4428 error = 0; 4429 goto abort3; 4430 } else if ((fdvp == tdvp) && 4431 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4432 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4433 fnd.ni_cnd.cn_namelen))) { 4434 error = 0; 4435 goto abort3; 4436 } 4437 } 4438 4439 /* 4440 * Make sure veriexec can screw us up. (But a race can screw 4441 * up veriexec, of course -- remember, fvp and (soon) tvp are 4442 * bogus.) 4443 */ 4444 #if NVERIEXEC > 0 4445 { 4446 char *f1, *f2; 4447 size_t f1_len; 4448 size_t f2_len; 4449 4450 f1_len = fnd.ni_cnd.cn_namelen + 1; 4451 f1 = kmem_alloc(f1_len, KM_SLEEP); 4452 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4453 4454 f2_len = tnd.ni_cnd.cn_namelen + 1; 4455 f2 = kmem_alloc(f2_len, KM_SLEEP); 4456 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4457 4458 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4459 4460 kmem_free(f1, f1_len); 4461 kmem_free(f2, f2_len); 4462 4463 if (error) 4464 goto abort3; 4465 } 4466 #endif /* NVERIEXEC > 0 */ 4467 4468 /* 4469 * All ready. Incant the rename vop. 4470 */ 4471 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4472 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4473 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4474 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4475 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4476 4477 /* 4478 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4479 * tdvp and tvp. But we can't assert any of that. 4480 */ 4481 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4482 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4483 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4484 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4485 4486 /* 4487 * So all we have left to do is to drop the rename lock and 4488 * destroy the pathbufs. 4489 */ 4490 VFS_RENAMELOCK_EXIT(mp); 4491 goto out2; 4492 4493 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4494 VOP_UNLOCK(tvp); 4495 abort2: VOP_UNLOCK(tdvp); 4496 VFS_RENAMELOCK_EXIT(mp); 4497 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4498 vrele(tdvp); 4499 if (tvp != NULL) 4500 vrele(tvp); 4501 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4502 vrele(fdvp); 4503 vrele(fvp); 4504 out2: pathbuf_destroy(tpb); 4505 out1: pathbuf_destroy(fpb); 4506 out0: return error; 4507 } 4508 4509 /* 4510 * Make a directory file. 4511 */ 4512 /* ARGSUSED */ 4513 int 4514 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4515 { 4516 /* { 4517 syscallarg(const char *) path; 4518 syscallarg(int) mode; 4519 } */ 4520 4521 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4522 SCARG(uap, mode), UIO_USERSPACE); 4523 } 4524 4525 int 4526 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4527 register_t *retval) 4528 { 4529 /* { 4530 syscallarg(int) fd; 4531 syscallarg(const char *) path; 4532 syscallarg(int) mode; 4533 } */ 4534 4535 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4536 SCARG(uap, mode), UIO_USERSPACE); 4537 } 4538 4539 4540 int 4541 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4542 { 4543 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4544 } 4545 4546 static int 4547 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4548 enum uio_seg seg) 4549 { 4550 struct proc *p = curlwp->l_proc; 4551 struct vnode *vp; 4552 struct vattr vattr; 4553 int error; 4554 struct pathbuf *pb; 4555 struct nameidata nd; 4556 4557 KASSERT(l != NULL || fdat == AT_FDCWD); 4558 4559 /* XXX bollocks, should pass in a pathbuf */ 4560 error = pathbuf_maybe_copyin(path, seg, &pb); 4561 if (error) { 4562 return error; 4563 } 4564 4565 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4566 4567 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4568 pathbuf_destroy(pb); 4569 return (error); 4570 } 4571 vp = nd.ni_vp; 4572 if (vp != NULL) { 4573 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4574 if (nd.ni_dvp == vp) 4575 vrele(nd.ni_dvp); 4576 else 4577 vput(nd.ni_dvp); 4578 vrele(vp); 4579 pathbuf_destroy(pb); 4580 return (EEXIST); 4581 } 4582 vattr_null(&vattr); 4583 vattr.va_type = VDIR; 4584 /* We will read cwdi->cwdi_cmask unlocked. */ 4585 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4586 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4587 if (!error) 4588 vrele(nd.ni_vp); 4589 vput(nd.ni_dvp); 4590 pathbuf_destroy(pb); 4591 return (error); 4592 } 4593 4594 /* 4595 * Remove a directory file. 4596 */ 4597 /* ARGSUSED */ 4598 int 4599 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4600 { 4601 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4602 AT_REMOVEDIR, UIO_USERSPACE); 4603 } 4604 4605 /* 4606 * Read a block of directory entries in a file system independent format. 4607 */ 4608 int 4609 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4610 { 4611 /* { 4612 syscallarg(int) fd; 4613 syscallarg(char *) buf; 4614 syscallarg(size_t) count; 4615 } */ 4616 file_t *fp; 4617 int error, done; 4618 4619 /* fd_getvnode() will use the descriptor for us */ 4620 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4621 return (error); 4622 if ((fp->f_flag & FREAD) == 0) { 4623 error = EBADF; 4624 goto out; 4625 } 4626 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4627 SCARG(uap, count), &done, l, 0, 0); 4628 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4629 *retval = done; 4630 out: 4631 fd_putfile(SCARG(uap, fd)); 4632 return (error); 4633 } 4634 4635 /* 4636 * Set the mode mask for creation of filesystem nodes. 4637 */ 4638 int 4639 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4640 { 4641 /* { 4642 syscallarg(mode_t) newmask; 4643 } */ 4644 struct proc *p = l->l_proc; 4645 struct cwdinfo *cwdi; 4646 4647 /* 4648 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4649 * important is that we serialize changes to the mask. The 4650 * rw_exit() will issue a write memory barrier on our behalf, 4651 * and force the changes out to other CPUs (as it must use an 4652 * atomic operation, draining the local CPU's store buffers). 4653 */ 4654 cwdi = p->p_cwdi; 4655 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4656 *retval = cwdi->cwdi_cmask; 4657 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4658 rw_exit(&cwdi->cwdi_lock); 4659 4660 return (0); 4661 } 4662 4663 int 4664 dorevoke(struct vnode *vp, kauth_cred_t cred) 4665 { 4666 struct vattr vattr; 4667 int error, fs_decision; 4668 4669 vn_lock(vp, LK_SHARED | LK_RETRY); 4670 error = VOP_GETATTR(vp, &vattr, cred); 4671 VOP_UNLOCK(vp); 4672 if (error != 0) 4673 return error; 4674 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4675 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4676 fs_decision); 4677 if (!error) 4678 VOP_REVOKE(vp, REVOKEALL); 4679 return (error); 4680 } 4681 4682 /* 4683 * Void all references to file by ripping underlying filesystem 4684 * away from vnode. 4685 */ 4686 /* ARGSUSED */ 4687 int 4688 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4689 { 4690 /* { 4691 syscallarg(const char *) path; 4692 } */ 4693 struct vnode *vp; 4694 int error; 4695 4696 error = namei_simple_user(SCARG(uap, path), 4697 NSM_FOLLOW_TRYEMULROOT, &vp); 4698 if (error != 0) 4699 return (error); 4700 error = dorevoke(vp, l->l_cred); 4701 vrele(vp); 4702 return (error); 4703 } 4704 4705 /* 4706 * Allocate backing store for a file, filling a hole without having to 4707 * explicitly write anything out. 4708 */ 4709 /* ARGSUSED */ 4710 int 4711 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4712 register_t *retval) 4713 { 4714 /* { 4715 syscallarg(int) fd; 4716 syscallarg(off_t) pos; 4717 syscallarg(off_t) len; 4718 } */ 4719 int fd; 4720 off_t pos, len; 4721 struct file *fp; 4722 struct vnode *vp; 4723 int error; 4724 4725 fd = SCARG(uap, fd); 4726 pos = SCARG(uap, pos); 4727 len = SCARG(uap, len); 4728 4729 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4730 *retval = EINVAL; 4731 return 0; 4732 } 4733 4734 error = fd_getvnode(fd, &fp); 4735 if (error) { 4736 *retval = error; 4737 return 0; 4738 } 4739 if ((fp->f_flag & FWRITE) == 0) { 4740 error = EBADF; 4741 goto fail; 4742 } 4743 vp = fp->f_vnode; 4744 4745 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4746 if (vp->v_type == VDIR) { 4747 error = EISDIR; 4748 } else { 4749 error = VOP_FALLOCATE(vp, pos, len); 4750 } 4751 VOP_UNLOCK(vp); 4752 4753 fail: 4754 fd_putfile(fd); 4755 *retval = error; 4756 return 0; 4757 } 4758 4759 /* 4760 * Deallocate backing store for a file, creating a hole. Also used for 4761 * invoking TRIM on disks. 4762 */ 4763 /* ARGSUSED */ 4764 int 4765 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4766 register_t *retval) 4767 { 4768 /* { 4769 syscallarg(int) fd; 4770 syscallarg(off_t) pos; 4771 syscallarg(off_t) len; 4772 } */ 4773 int fd; 4774 off_t pos, len; 4775 struct file *fp; 4776 struct vnode *vp; 4777 int error; 4778 4779 fd = SCARG(uap, fd); 4780 pos = SCARG(uap, pos); 4781 len = SCARG(uap, len); 4782 4783 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4784 return EINVAL; 4785 } 4786 4787 error = fd_getvnode(fd, &fp); 4788 if (error) { 4789 return error; 4790 } 4791 if ((fp->f_flag & FWRITE) == 0) { 4792 error = EBADF; 4793 goto fail; 4794 } 4795 vp = fp->f_vnode; 4796 4797 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4798 if (vp->v_type == VDIR) { 4799 error = EISDIR; 4800 } else { 4801 error = VOP_FDISCARD(vp, pos, len); 4802 } 4803 VOP_UNLOCK(vp); 4804 4805 fail: 4806 fd_putfile(fd); 4807 return error; 4808 } 4809
Indexes created Fri Sep 26 20:09:58 GMT 2025