vfs_syscalls.c revision 1.536
1 /* $NetBSD: vfs_syscalls.c,v 1.536 2019/09/22 22:59:39 christos 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 1.536 2019/09/22 22:59:39 christos Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/fcntl.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/fstrans.h> 91 #include <sys/proc.h> 92 #include <sys/uio.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/quota.h> 99 #include <sys/quotactl.h> 100 #include <sys/ktrace.h> 101 #ifdef FILEASSOC 102 #include <sys/fileassoc.h> 103 #endif /* FILEASSOC */ 104 #include <sys/extattr.h> 105 #include <sys/verified_exec.h> 106 #include <sys/kauth.h> 107 #include <sys/atomic.h> 108 #include <sys/module.h> 109 #include <sys/buf.h> 110 #include <sys/event.h> 111 #include <sys/compat_stub.h> 112 113 #include <miscfs/genfs/genfs.h> 114 #include <miscfs/specfs/specdev.h> 115 116 #include <nfs/rpcv2.h> 117 #include <nfs/nfsproto.h> 118 #include <nfs/nfs.h> 119 #include <nfs/nfs_var.h> 120 121 /* XXX this shouldn't be here */ 122 #ifndef OFF_T_MAX 123 #define OFF_T_MAX __type_max(off_t) 124 #endif 125 126 static int change_flags(struct vnode *, u_long, struct lwp *); 127 static int change_mode(struct vnode *, int, struct lwp *); 128 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 129 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 130 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 131 enum uio_seg); 132 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 133 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 134 enum uio_seg); 135 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 136 enum uio_seg, int); 137 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 138 size_t, register_t *); 139 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 140 141 static int fd_nameiat(struct lwp *, int, struct nameidata *); 142 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 143 namei_simple_flags_t, struct vnode **); 144 145 /* 146 * This table is used to maintain compatibility with 4.3BSD 147 * and NetBSD 0.9 mount syscalls - and possibly other systems. 148 * Note, the order is important! 149 * 150 * Do not modify this table. It should only contain filesystems 151 * supported by NetBSD 0.9 and 4.3BSD. 152 */ 153 const char * const mountcompatnames[] = { 154 NULL, /* 0 = MOUNT_NONE */ 155 MOUNT_FFS, /* 1 = MOUNT_UFS */ 156 MOUNT_NFS, /* 2 */ 157 MOUNT_MFS, /* 3 */ 158 MOUNT_MSDOS, /* 4 */ 159 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 160 MOUNT_FDESC, /* 6 */ 161 MOUNT_KERNFS, /* 7 */ 162 NULL, /* 8 = MOUNT_DEVFS */ 163 MOUNT_AFS, /* 9 */ 164 }; 165 166 const int nmountcompatnames = __arraycount(mountcompatnames); 167 168 static int 169 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 170 { 171 file_t *dfp; 172 int error; 173 174 if (fdat != AT_FDCWD) { 175 if ((error = fd_getvnode(fdat, &dfp)) != 0) 176 goto out; 177 178 NDAT(ndp, dfp->f_vnode); 179 } 180 181 error = namei(ndp); 182 183 if (fdat != AT_FDCWD) 184 fd_putfile(fdat); 185 out: 186 return error; 187 } 188 189 static int 190 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 191 namei_simple_flags_t sflags, struct vnode **vp_ret) 192 { 193 file_t *dfp; 194 struct vnode *dvp; 195 int error; 196 197 if (fdat != AT_FDCWD) { 198 if ((error = fd_getvnode(fdat, &dfp)) != 0) 199 goto out; 200 201 dvp = dfp->f_vnode; 202 } else { 203 dvp = NULL; 204 } 205 206 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 207 208 if (fdat != AT_FDCWD) 209 fd_putfile(fdat); 210 out: 211 return error; 212 } 213 214 static int 215 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 216 { 217 int error; 218 219 fp->f_flag = flags & FMASK; 220 fp->f_type = DTYPE_VNODE; 221 fp->f_ops = &vnops; 222 fp->f_vnode = vp; 223 224 if (flags & (O_EXLOCK | O_SHLOCK)) { 225 struct flock lf; 226 int type; 227 228 lf.l_whence = SEEK_SET; 229 lf.l_start = 0; 230 lf.l_len = 0; 231 if (flags & O_EXLOCK) 232 lf.l_type = F_WRLCK; 233 else 234 lf.l_type = F_RDLCK; 235 type = F_FLOCK; 236 if ((flags & FNONBLOCK) == 0) 237 type |= F_WAIT; 238 VOP_UNLOCK(vp); 239 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 240 if (error) { 241 (void) vn_close(vp, fp->f_flag, fp->f_cred); 242 fd_abort(l->l_proc, fp, indx); 243 return error; 244 } 245 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 246 atomic_or_uint(&fp->f_flag, FHASLOCK); 247 } 248 if (flags & O_CLOEXEC) 249 fd_set_exclose(l, indx, true); 250 return 0; 251 } 252 253 static int 254 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 255 void *data, size_t *data_len) 256 { 257 struct mount *mp; 258 int error = 0, saved_flags; 259 260 mp = vp->v_mount; 261 saved_flags = mp->mnt_flag; 262 263 /* We can operate only on VV_ROOT nodes. */ 264 if ((vp->v_vflag & VV_ROOT) == 0) { 265 error = EINVAL; 266 goto out; 267 } 268 269 /* 270 * We only allow the filesystem to be reloaded if it 271 * is currently mounted read-only. Additionally, we 272 * prevent read-write to read-only downgrades. 273 */ 274 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 275 (mp->mnt_flag & MNT_RDONLY) == 0 && 276 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 277 error = EOPNOTSUPP; /* Needs translation */ 278 goto out; 279 } 280 281 /* 282 * Enabling MNT_UNION requires a covered mountpoint and 283 * must not happen on the root mount. 284 */ 285 if ((flags & MNT_UNION) != 0 && mp->mnt_vnodecovered == NULLVP) { 286 error = EOPNOTSUPP; 287 goto out; 288 } 289 290 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 291 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 292 if (error) 293 goto out; 294 295 error = vfs_suspend(mp, 0); 296 if (error) 297 goto out; 298 299 mutex_enter(&mp->mnt_updating); 300 301 mp->mnt_flag &= ~MNT_OP_FLAGS; 302 mp->mnt_flag |= flags & MNT_OP_FLAGS; 303 304 /* 305 * Set the mount level flags. 306 */ 307 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 308 if ((flags & MNT_RDONLY)) 309 mp->mnt_iflag |= IMNT_WANTRDONLY; 310 else 311 mp->mnt_iflag |= IMNT_WANTRDWR; 312 } 313 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 314 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 315 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 316 mp->mnt_flag &= ~MNT_RDONLY; 317 318 error = VFS_MOUNT(mp, path, data, data_len); 319 320 if (error && data != NULL) { 321 int error2; 322 323 /* 324 * Update failed; let's try and see if it was an 325 * export request. For compat with 3.0 and earlier. 326 */ 327 error2 = vfs_hooks_reexport(mp, path, data); 328 329 /* 330 * Only update error code if the export request was 331 * understood but some problem occurred while 332 * processing it. 333 */ 334 if (error2 != EJUSTRETURN) 335 error = error2; 336 } 337 338 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 339 mp->mnt_flag |= MNT_RDONLY; 340 if (error) 341 mp->mnt_flag = saved_flags; 342 mp->mnt_flag &= ~MNT_OP_FLAGS; 343 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 344 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 345 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 346 vfs_syncer_add_to_worklist(mp); 347 } else { 348 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 349 vfs_syncer_remove_from_worklist(mp); 350 } 351 mutex_exit(&mp->mnt_updating); 352 vfs_resume(mp); 353 354 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 355 (flags & MNT_EXTATTR)) { 356 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 357 NULL, 0, NULL) != 0) { 358 printf("%s: failed to start extattr, error = %d", 359 mp->mnt_stat.f_mntonname, error); 360 mp->mnt_flag &= ~MNT_EXTATTR; 361 } 362 } 363 364 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 365 !(flags & MNT_EXTATTR)) { 366 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 367 NULL, 0, NULL) != 0) { 368 printf("%s: failed to stop extattr, error = %d", 369 mp->mnt_stat.f_mntonname, error); 370 mp->mnt_flag |= MNT_RDONLY; 371 } 372 } 373 out: 374 return (error); 375 } 376 377 static int 378 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 379 struct vfsops **vfsops) 380 { 381 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 382 int error; 383 384 if (type_seg == UIO_USERSPACE) { 385 /* Copy file-system type from userspace. */ 386 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 387 } else { 388 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 389 KASSERT(error == 0); 390 } 391 392 if (error) { 393 /* 394 * Historically, filesystem types were identified by numbers. 395 * If we get an integer for the filesystem type instead of a 396 * string, we check to see if it matches one of the historic 397 * filesystem types. 398 */ 399 u_long fsindex = (u_long)fstype; 400 if (fsindex >= nmountcompatnames || 401 mountcompatnames[fsindex] == NULL) 402 return ENODEV; 403 strlcpy(fstypename, mountcompatnames[fsindex], 404 sizeof(fstypename)); 405 } 406 407 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 408 if (strcmp(fstypename, "ufs") == 0) 409 fstypename[0] = 'f'; 410 411 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 412 return 0; 413 414 /* If we can autoload a vfs module, try again */ 415 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 416 417 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 418 return 0; 419 420 return ENODEV; 421 } 422 423 static int 424 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 425 void *data, size_t *data_len) 426 { 427 struct mount *mp; 428 int error; 429 430 /* If MNT_GETARGS is specified, it should be the only flag. */ 431 if (flags & ~MNT_GETARGS) 432 return EINVAL; 433 434 mp = vp->v_mount; 435 436 /* XXX: probably some notion of "can see" here if we want isolation. */ 437 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 438 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 439 if (error) 440 return error; 441 442 if ((vp->v_vflag & VV_ROOT) == 0) 443 return EINVAL; 444 445 if (vfs_busy(mp)) 446 return EPERM; 447 448 mutex_enter(&mp->mnt_updating); 449 mp->mnt_flag &= ~MNT_OP_FLAGS; 450 mp->mnt_flag |= MNT_GETARGS; 451 error = VFS_MOUNT(mp, path, data, data_len); 452 mp->mnt_flag &= ~MNT_OP_FLAGS; 453 mutex_exit(&mp->mnt_updating); 454 455 vfs_unbusy(mp); 456 return (error); 457 } 458 459 int 460 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 461 { 462 /* { 463 syscallarg(const char *) type; 464 syscallarg(const char *) path; 465 syscallarg(int) flags; 466 syscallarg(void *) data; 467 syscallarg(size_t) data_len; 468 } */ 469 470 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 471 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 472 SCARG(uap, data_len), retval); 473 } 474 475 int 476 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 477 const char *path, int flags, void *data, enum uio_seg data_seg, 478 size_t data_len, register_t *retval) 479 { 480 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 481 struct vnode *vp; 482 void *data_buf = data; 483 bool vfsopsrele = false; 484 size_t alloc_sz = 0; 485 int error; 486 487 /* 488 * Get vnode to be covered 489 */ 490 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 491 if (error != 0) { 492 vp = NULL; 493 goto done; 494 } 495 496 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 497 vfsops = vp->v_mount->mnt_op; 498 } else { 499 /* 'type' is userspace */ 500 error = mount_get_vfsops(type, type_seg, &vfsops); 501 if (error != 0) 502 goto done; 503 vfsopsrele = true; 504 } 505 506 /* 507 * We allow data to be NULL, even for userspace. Some fs's don't need 508 * it. The others will handle NULL. 509 */ 510 if (data != NULL && data_seg == UIO_USERSPACE) { 511 if (data_len == 0) { 512 /* No length supplied, use default for filesystem */ 513 data_len = vfsops->vfs_min_mount_data; 514 515 /* 516 * Hopefully a longer buffer won't make copyin() fail. 517 * For compatibility with 3.0 and earlier. 518 */ 519 if (flags & MNT_UPDATE 520 && data_len < sizeof (struct mnt_export_args30)) 521 data_len = sizeof (struct mnt_export_args30); 522 } 523 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 524 error = EINVAL; 525 goto done; 526 } 527 alloc_sz = data_len; 528 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 529 530 /* NFS needs the buffer even for mnt_getargs .... */ 531 error = copyin(data, data_buf, data_len); 532 if (error != 0) 533 goto done; 534 } 535 536 if (flags & MNT_GETARGS) { 537 if (data_len == 0) { 538 error = EINVAL; 539 goto done; 540 } 541 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 542 if (error != 0) 543 goto done; 544 if (data_seg == UIO_USERSPACE) 545 error = copyout(data_buf, data, data_len); 546 *retval = data_len; 547 } else if (flags & MNT_UPDATE) { 548 error = mount_update(l, vp, path, flags, data_buf, &data_len); 549 } else { 550 /* Locking is handled internally in mount_domount(). */ 551 KASSERT(vfsopsrele == true); 552 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 553 &data_len); 554 vfsopsrele = false; 555 } 556 if (!error) 557 KNOTE(&fs_klist, VQ_MOUNT); 558 559 done: 560 if (vfsopsrele) 561 vfs_delref(vfsops); 562 if (vp != NULL) { 563 vrele(vp); 564 } 565 if (data_buf != data) 566 kmem_free(data_buf, alloc_sz); 567 return (error); 568 } 569 570 /* 571 * Unmount a file system. 572 * 573 * Note: unmount takes a path to the vnode mounted on as argument, 574 * not special file (as before). 575 */ 576 /* ARGSUSED */ 577 int 578 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 579 { 580 /* { 581 syscallarg(const char *) path; 582 syscallarg(int) flags; 583 } */ 584 struct vnode *vp; 585 struct mount *mp; 586 int error; 587 struct pathbuf *pb; 588 struct nameidata nd; 589 590 error = pathbuf_copyin(SCARG(uap, path), &pb); 591 if (error) { 592 return error; 593 } 594 595 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 596 if ((error = namei(&nd)) != 0) { 597 pathbuf_destroy(pb); 598 return error; 599 } 600 vp = nd.ni_vp; 601 pathbuf_destroy(pb); 602 603 mp = vp->v_mount; 604 vfs_ref(mp); 605 VOP_UNLOCK(vp); 606 607 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 608 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 609 if (error) { 610 vrele(vp); 611 vfs_rele(mp); 612 return (error); 613 } 614 615 /* 616 * Don't allow unmounting the root file system. 617 */ 618 if (mp->mnt_flag & MNT_ROOTFS) { 619 vrele(vp); 620 vfs_rele(mp); 621 return (EINVAL); 622 } 623 624 /* 625 * Must be the root of the filesystem 626 */ 627 if ((vp->v_vflag & VV_ROOT) == 0) { 628 vrele(vp); 629 vfs_rele(mp); 630 return (EINVAL); 631 } 632 633 vrele(vp); 634 error = dounmount(mp, SCARG(uap, flags), l); 635 vfs_rele(mp); 636 if (!error) 637 KNOTE(&fs_klist, VQ_UNMOUNT); 638 return error; 639 } 640 641 /* 642 * Sync each mounted filesystem. 643 */ 644 #ifdef DEBUG 645 int syncprt = 0; 646 struct ctldebug debug0 = { "syncprt", &syncprt }; 647 #endif 648 649 void 650 do_sys_sync(struct lwp *l) 651 { 652 mount_iterator_t *iter; 653 struct mount *mp; 654 int asyncflag; 655 656 mountlist_iterator_init(&iter); 657 while ((mp = mountlist_iterator_next(iter)) != NULL) { 658 mutex_enter(&mp->mnt_updating); 659 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 660 asyncflag = mp->mnt_flag & MNT_ASYNC; 661 mp->mnt_flag &= ~MNT_ASYNC; 662 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 663 if (asyncflag) 664 mp->mnt_flag |= MNT_ASYNC; 665 } 666 mutex_exit(&mp->mnt_updating); 667 } 668 mountlist_iterator_destroy(iter); 669 #ifdef DEBUG 670 if (syncprt) 671 vfs_bufstats(); 672 #endif /* DEBUG */ 673 } 674 675 /* ARGSUSED */ 676 int 677 sys_sync(struct lwp *l, const void *v, register_t *retval) 678 { 679 do_sys_sync(l); 680 return (0); 681 } 682 683 684 /* 685 * Access or change filesystem quotas. 686 * 687 * (this is really 14 different calls bundled into one) 688 */ 689 690 static int 691 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 692 { 693 struct quotastat info_k; 694 int error; 695 696 /* ensure any padding bytes are cleared */ 697 memset(&info_k, 0, sizeof(info_k)); 698 699 error = vfs_quotactl_stat(mp, &info_k); 700 if (error) { 701 return error; 702 } 703 704 return copyout(&info_k, info_u, sizeof(info_k)); 705 } 706 707 static int 708 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 709 struct quotaidtypestat *info_u) 710 { 711 struct quotaidtypestat info_k; 712 int error; 713 714 /* ensure any padding bytes are cleared */ 715 memset(&info_k, 0, sizeof(info_k)); 716 717 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 718 if (error) { 719 return error; 720 } 721 722 return copyout(&info_k, info_u, sizeof(info_k)); 723 } 724 725 static int 726 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 727 struct quotaobjtypestat *info_u) 728 { 729 struct quotaobjtypestat info_k; 730 int error; 731 732 /* ensure any padding bytes are cleared */ 733 memset(&info_k, 0, sizeof(info_k)); 734 735 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 736 if (error) { 737 return error; 738 } 739 740 return copyout(&info_k, info_u, sizeof(info_k)); 741 } 742 743 static int 744 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 745 struct quotaval *val_u) 746 { 747 struct quotakey key_k; 748 struct quotaval val_k; 749 int error; 750 751 /* ensure any padding bytes are cleared */ 752 memset(&val_k, 0, sizeof(val_k)); 753 754 error = copyin(key_u, &key_k, sizeof(key_k)); 755 if (error) { 756 return error; 757 } 758 759 error = vfs_quotactl_get(mp, &key_k, &val_k); 760 if (error) { 761 return error; 762 } 763 764 return copyout(&val_k, val_u, sizeof(val_k)); 765 } 766 767 static int 768 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 769 const struct quotaval *val_u) 770 { 771 struct quotakey key_k; 772 struct quotaval val_k; 773 int error; 774 775 error = copyin(key_u, &key_k, sizeof(key_k)); 776 if (error) { 777 return error; 778 } 779 780 error = copyin(val_u, &val_k, sizeof(val_k)); 781 if (error) { 782 return error; 783 } 784 785 return vfs_quotactl_put(mp, &key_k, &val_k); 786 } 787 788 static int 789 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 790 { 791 struct quotakey key_k; 792 int error; 793 794 error = copyin(key_u, &key_k, sizeof(key_k)); 795 if (error) { 796 return error; 797 } 798 799 return vfs_quotactl_del(mp, &key_k); 800 } 801 802 static int 803 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 804 { 805 struct quotakcursor cursor_k; 806 int error; 807 808 /* ensure any padding bytes are cleared */ 809 memset(&cursor_k, 0, sizeof(cursor_k)); 810 811 error = vfs_quotactl_cursoropen(mp, &cursor_k); 812 if (error) { 813 return error; 814 } 815 816 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 817 } 818 819 static int 820 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 821 { 822 struct quotakcursor cursor_k; 823 int error; 824 825 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 826 if (error) { 827 return error; 828 } 829 830 return vfs_quotactl_cursorclose(mp, &cursor_k); 831 } 832 833 static int 834 do_sys_quotactl_cursorskipidtype(struct mount *mp, 835 struct quotakcursor *cursor_u, int idtype) 836 { 837 struct quotakcursor cursor_k; 838 int error; 839 840 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 841 if (error) { 842 return error; 843 } 844 845 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 846 if (error) { 847 return error; 848 } 849 850 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 851 } 852 853 static int 854 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 855 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 856 unsigned *ret_u) 857 { 858 #define CGET_STACK_MAX 8 859 struct quotakcursor cursor_k; 860 struct quotakey stackkeys[CGET_STACK_MAX]; 861 struct quotaval stackvals[CGET_STACK_MAX]; 862 struct quotakey *keys_k; 863 struct quotaval *vals_k; 864 unsigned ret_k; 865 int error; 866 867 if (maxnum > 128) { 868 maxnum = 128; 869 } 870 871 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 872 if (error) { 873 return error; 874 } 875 876 if (maxnum <= CGET_STACK_MAX) { 877 keys_k = stackkeys; 878 vals_k = stackvals; 879 /* ensure any padding bytes are cleared */ 880 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 881 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 882 } else { 883 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 884 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 885 } 886 887 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 888 &ret_k); 889 if (error) { 890 goto fail; 891 } 892 893 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 894 if (error) { 895 goto fail; 896 } 897 898 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 899 if (error) { 900 goto fail; 901 } 902 903 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 904 if (error) { 905 goto fail; 906 } 907 908 /* do last to maximize the chance of being able to recover a failure */ 909 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 910 911 fail: 912 if (keys_k != stackkeys) { 913 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 914 } 915 if (vals_k != stackvals) { 916 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 917 } 918 return error; 919 } 920 921 static int 922 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 923 int *ret_u) 924 { 925 struct quotakcursor cursor_k; 926 int ret_k; 927 int error; 928 929 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 930 if (error) { 931 return error; 932 } 933 934 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 935 if (error) { 936 return error; 937 } 938 939 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 940 if (error) { 941 return error; 942 } 943 944 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 945 } 946 947 static int 948 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 949 { 950 struct quotakcursor cursor_k; 951 int error; 952 953 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 954 if (error) { 955 return error; 956 } 957 958 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 959 if (error) { 960 return error; 961 } 962 963 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 964 } 965 966 static int 967 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 968 { 969 char *path_k; 970 int error; 971 972 /* XXX this should probably be a struct pathbuf */ 973 path_k = PNBUF_GET(); 974 error = copyin(path_u, path_k, PATH_MAX); 975 if (error) { 976 PNBUF_PUT(path_k); 977 return error; 978 } 979 980 error = vfs_quotactl_quotaon(mp, idtype, path_k); 981 982 PNBUF_PUT(path_k); 983 return error; 984 } 985 986 static int 987 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 988 { 989 return vfs_quotactl_quotaoff(mp, idtype); 990 } 991 992 int 993 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 994 { 995 struct mount *mp; 996 struct vnode *vp; 997 int error; 998 999 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 1000 if (error != 0) 1001 return (error); 1002 mp = vp->v_mount; 1003 1004 switch (args->qc_op) { 1005 case QUOTACTL_STAT: 1006 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 1007 break; 1008 case QUOTACTL_IDTYPESTAT: 1009 error = do_sys_quotactl_idtypestat(mp, 1010 args->u.idtypestat.qc_idtype, 1011 args->u.idtypestat.qc_info); 1012 break; 1013 case QUOTACTL_OBJTYPESTAT: 1014 error = do_sys_quotactl_objtypestat(mp, 1015 args->u.objtypestat.qc_objtype, 1016 args->u.objtypestat.qc_info); 1017 break; 1018 case QUOTACTL_GET: 1019 error = do_sys_quotactl_get(mp, 1020 args->u.get.qc_key, 1021 args->u.get.qc_val); 1022 break; 1023 case QUOTACTL_PUT: 1024 error = do_sys_quotactl_put(mp, 1025 args->u.put.qc_key, 1026 args->u.put.qc_val); 1027 break; 1028 case QUOTACTL_DEL: 1029 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1030 break; 1031 case QUOTACTL_CURSOROPEN: 1032 error = do_sys_quotactl_cursoropen(mp, 1033 args->u.cursoropen.qc_cursor); 1034 break; 1035 case QUOTACTL_CURSORCLOSE: 1036 error = do_sys_quotactl_cursorclose(mp, 1037 args->u.cursorclose.qc_cursor); 1038 break; 1039 case QUOTACTL_CURSORSKIPIDTYPE: 1040 error = do_sys_quotactl_cursorskipidtype(mp, 1041 args->u.cursorskipidtype.qc_cursor, 1042 args->u.cursorskipidtype.qc_idtype); 1043 break; 1044 case QUOTACTL_CURSORGET: 1045 error = do_sys_quotactl_cursorget(mp, 1046 args->u.cursorget.qc_cursor, 1047 args->u.cursorget.qc_keys, 1048 args->u.cursorget.qc_vals, 1049 args->u.cursorget.qc_maxnum, 1050 args->u.cursorget.qc_ret); 1051 break; 1052 case QUOTACTL_CURSORATEND: 1053 error = do_sys_quotactl_cursoratend(mp, 1054 args->u.cursoratend.qc_cursor, 1055 args->u.cursoratend.qc_ret); 1056 break; 1057 case QUOTACTL_CURSORREWIND: 1058 error = do_sys_quotactl_cursorrewind(mp, 1059 args->u.cursorrewind.qc_cursor); 1060 break; 1061 case QUOTACTL_QUOTAON: 1062 error = do_sys_quotactl_quotaon(mp, 1063 args->u.quotaon.qc_idtype, 1064 args->u.quotaon.qc_quotafile); 1065 break; 1066 case QUOTACTL_QUOTAOFF: 1067 error = do_sys_quotactl_quotaoff(mp, 1068 args->u.quotaoff.qc_idtype); 1069 break; 1070 default: 1071 error = EINVAL; 1072 break; 1073 } 1074 1075 vrele(vp); 1076 return error; 1077 } 1078 1079 /* ARGSUSED */ 1080 int 1081 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1082 register_t *retval) 1083 { 1084 /* { 1085 syscallarg(const char *) path; 1086 syscallarg(struct quotactl_args *) args; 1087 } */ 1088 struct quotactl_args args; 1089 int error; 1090 1091 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1092 if (error) { 1093 return error; 1094 } 1095 1096 return do_sys_quotactl(SCARG(uap, path), &args); 1097 } 1098 1099 int 1100 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1101 int root) 1102 { 1103 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1104 int error = 0; 1105 1106 /* 1107 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1108 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1109 * overrides MNT_NOWAIT. 1110 */ 1111 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1112 (flags != MNT_WAIT && flags != 0)) { 1113 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1114 goto done; 1115 } 1116 1117 /* Get the filesystem stats now */ 1118 memset(sp, 0, sizeof(*sp)); 1119 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1120 return error; 1121 } 1122 1123 if (cwdi->cwdi_rdir == NULL) 1124 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1125 done: 1126 if (cwdi->cwdi_rdir != NULL) { 1127 size_t len; 1128 char *bp; 1129 char c; 1130 char *path = PNBUF_GET(); 1131 1132 bp = path + MAXPATHLEN; 1133 *--bp = '\0'; 1134 rw_enter(&cwdi->cwdi_lock, RW_READER); 1135 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1136 MAXPATHLEN / 2, 0, l); 1137 rw_exit(&cwdi->cwdi_lock); 1138 if (error) { 1139 PNBUF_PUT(path); 1140 return error; 1141 } 1142 len = strlen(bp); 1143 if (len != 1) { 1144 /* 1145 * for mount points that are below our root, we can see 1146 * them, so we fix up the pathname and return them. The 1147 * rest we cannot see, so we don't allow viewing the 1148 * data. 1149 */ 1150 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1151 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1152 (void)strlcpy(sp->f_mntonname, 1153 c == '\0' ? "/" : &sp->f_mntonname[len], 1154 sizeof(sp->f_mntonname)); 1155 } else { 1156 if (root) 1157 (void)strlcpy(sp->f_mntonname, "/", 1158 sizeof(sp->f_mntonname)); 1159 else 1160 error = EPERM; 1161 } 1162 } 1163 PNBUF_PUT(path); 1164 } 1165 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1166 return error; 1167 } 1168 1169 /* 1170 * Get filesystem statistics by path. 1171 */ 1172 int 1173 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1174 { 1175 struct mount *mp; 1176 int error; 1177 struct vnode *vp; 1178 1179 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1180 if (error != 0) 1181 return error; 1182 mp = vp->v_mount; 1183 error = dostatvfs(mp, sb, l, flags, 1); 1184 vrele(vp); 1185 return error; 1186 } 1187 1188 /* ARGSUSED */ 1189 int 1190 sys___statvfs190(struct lwp *l, const struct sys___statvfs190_args *uap, register_t *retval) 1191 { 1192 /* { 1193 syscallarg(const char *) path; 1194 syscallarg(struct statvfs *) buf; 1195 syscallarg(int) flags; 1196 } */ 1197 struct statvfs *sb; 1198 int error; 1199 1200 sb = STATVFSBUF_GET(); 1201 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1202 if (error == 0) 1203 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1204 STATVFSBUF_PUT(sb); 1205 return error; 1206 } 1207 1208 /* 1209 * Get filesystem statistics by fd. 1210 */ 1211 int 1212 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1213 { 1214 file_t *fp; 1215 struct mount *mp; 1216 int error; 1217 1218 /* fd_getvnode() will use the descriptor for us */ 1219 if ((error = fd_getvnode(fd, &fp)) != 0) 1220 return (error); 1221 mp = fp->f_vnode->v_mount; 1222 error = dostatvfs(mp, sb, curlwp, flags, 1); 1223 fd_putfile(fd); 1224 return error; 1225 } 1226 1227 /* ARGSUSED */ 1228 int 1229 sys___fstatvfs190(struct lwp *l, const struct sys___fstatvfs190_args *uap, register_t *retval) 1230 { 1231 /* { 1232 syscallarg(int) fd; 1233 syscallarg(struct statvfs *) buf; 1234 syscallarg(int) flags; 1235 } */ 1236 struct statvfs *sb; 1237 int error; 1238 1239 sb = STATVFSBUF_GET(); 1240 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1241 if (error == 0) 1242 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1243 STATVFSBUF_PUT(sb); 1244 return error; 1245 } 1246 1247 1248 /* 1249 * Get statistics on all filesystems. 1250 */ 1251 int 1252 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1253 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1254 register_t *retval) 1255 { 1256 int root = 0; 1257 mount_iterator_t *iter; 1258 struct proc *p = l->l_proc; 1259 struct mount *mp; 1260 struct statvfs *sb; 1261 size_t count, maxcount; 1262 int error = 0; 1263 1264 sb = STATVFSBUF_GET(); 1265 maxcount = bufsize / entry_sz; 1266 count = 0; 1267 mountlist_iterator_init(&iter); 1268 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1269 if (sfsp && count < maxcount) { 1270 error = dostatvfs(mp, sb, l, flags, 0); 1271 if (error) { 1272 error = 0; 1273 continue; 1274 } 1275 error = copyfn(sb, sfsp, entry_sz); 1276 if (error) 1277 goto out; 1278 sfsp = (char *)sfsp + entry_sz; 1279 root |= strcmp(sb->f_mntonname, "/") == 0; 1280 } 1281 count++; 1282 } 1283 1284 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1285 /* 1286 * fake a root entry 1287 */ 1288 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1289 sb, l, flags, 1); 1290 if (error != 0) 1291 goto out; 1292 if (sfsp) { 1293 error = copyfn(sb, sfsp, entry_sz); 1294 if (error != 0) 1295 goto out; 1296 } 1297 count++; 1298 } 1299 if (sfsp && count > maxcount) 1300 *retval = maxcount; 1301 else 1302 *retval = count; 1303 out: 1304 mountlist_iterator_destroy(iter); 1305 STATVFSBUF_PUT(sb); 1306 return error; 1307 } 1308 1309 int 1310 sys___getvfsstat90(struct lwp *l, const struct sys___getvfsstat90_args *uap, 1311 register_t *retval) 1312 { 1313 /* { 1314 syscallarg(struct statvfs *) buf; 1315 syscallarg(size_t) bufsize; 1316 syscallarg(int) flags; 1317 } */ 1318 1319 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1320 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1321 } 1322 1323 /* 1324 * Change current working directory to a given file descriptor. 1325 */ 1326 /* ARGSUSED */ 1327 int 1328 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1329 { 1330 /* { 1331 syscallarg(int) fd; 1332 } */ 1333 struct proc *p = l->l_proc; 1334 struct cwdinfo *cwdi; 1335 struct vnode *vp, *tdp; 1336 struct mount *mp; 1337 file_t *fp; 1338 int error, fd; 1339 1340 /* fd_getvnode() will use the descriptor for us */ 1341 fd = SCARG(uap, fd); 1342 if ((error = fd_getvnode(fd, &fp)) != 0) 1343 return (error); 1344 vp = fp->f_vnode; 1345 1346 vref(vp); 1347 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1348 if (vp->v_type != VDIR) 1349 error = ENOTDIR; 1350 else 1351 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1352 if (error) { 1353 vput(vp); 1354 goto out; 1355 } 1356 while ((mp = vp->v_mountedhere) != NULL) { 1357 error = vfs_busy(mp); 1358 vput(vp); 1359 if (error != 0) 1360 goto out; 1361 error = VFS_ROOT(mp, &tdp); 1362 vfs_unbusy(mp); 1363 if (error) 1364 goto out; 1365 vp = tdp; 1366 } 1367 VOP_UNLOCK(vp); 1368 1369 /* 1370 * Disallow changing to a directory not under the process's 1371 * current root directory (if there is one). 1372 */ 1373 cwdi = p->p_cwdi; 1374 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1375 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1376 vrele(vp); 1377 error = EPERM; /* operation not permitted */ 1378 } else { 1379 vrele(cwdi->cwdi_cdir); 1380 cwdi->cwdi_cdir = vp; 1381 } 1382 rw_exit(&cwdi->cwdi_lock); 1383 1384 out: 1385 fd_putfile(fd); 1386 return (error); 1387 } 1388 1389 /* 1390 * Change this process's notion of the root directory to a given file 1391 * descriptor. 1392 */ 1393 int 1394 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1395 { 1396 struct proc *p = l->l_proc; 1397 struct vnode *vp; 1398 file_t *fp; 1399 int error, fd = SCARG(uap, fd); 1400 1401 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1402 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1403 return error; 1404 /* fd_getvnode() will use the descriptor for us */ 1405 if ((error = fd_getvnode(fd, &fp)) != 0) 1406 return error; 1407 vp = fp->f_vnode; 1408 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1409 if (vp->v_type != VDIR) 1410 error = ENOTDIR; 1411 else 1412 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1413 VOP_UNLOCK(vp); 1414 if (error) 1415 goto out; 1416 vref(vp); 1417 1418 change_root(p->p_cwdi, vp, l); 1419 1420 out: 1421 fd_putfile(fd); 1422 return (error); 1423 } 1424 1425 /* 1426 * Change current working directory (``.''). 1427 */ 1428 /* ARGSUSED */ 1429 int 1430 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1431 { 1432 /* { 1433 syscallarg(const char *) path; 1434 } */ 1435 struct proc *p = l->l_proc; 1436 struct cwdinfo *cwdi; 1437 int error; 1438 struct vnode *vp; 1439 1440 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1441 &vp, l)) != 0) 1442 return (error); 1443 cwdi = p->p_cwdi; 1444 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1445 vrele(cwdi->cwdi_cdir); 1446 cwdi->cwdi_cdir = vp; 1447 rw_exit(&cwdi->cwdi_lock); 1448 return (0); 1449 } 1450 1451 /* 1452 * Change notion of root (``/'') directory. 1453 */ 1454 /* ARGSUSED */ 1455 int 1456 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1457 { 1458 /* { 1459 syscallarg(const char *) path; 1460 } */ 1461 struct proc *p = l->l_proc; 1462 int error; 1463 struct vnode *vp; 1464 1465 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1466 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1467 return (error); 1468 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1469 &vp, l)) != 0) 1470 return (error); 1471 1472 change_root(p->p_cwdi, vp, l); 1473 1474 return (0); 1475 } 1476 1477 /* 1478 * Common routine for chroot and fchroot. 1479 * NB: callers need to properly authorize the change root operation. 1480 */ 1481 void 1482 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1483 { 1484 struct proc *p = l->l_proc; 1485 kauth_cred_t ncred; 1486 1487 ncred = kauth_cred_alloc(); 1488 1489 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1490 if (cwdi->cwdi_rdir != NULL) 1491 vrele(cwdi->cwdi_rdir); 1492 cwdi->cwdi_rdir = vp; 1493 1494 /* 1495 * Prevent escaping from chroot by putting the root under 1496 * the working directory. Silently chdir to / if we aren't 1497 * already there. 1498 */ 1499 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1500 /* 1501 * XXX would be more failsafe to change directory to a 1502 * deadfs node here instead 1503 */ 1504 vrele(cwdi->cwdi_cdir); 1505 vref(vp); 1506 cwdi->cwdi_cdir = vp; 1507 } 1508 rw_exit(&cwdi->cwdi_lock); 1509 1510 /* Get a write lock on the process credential. */ 1511 proc_crmod_enter(); 1512 1513 kauth_cred_clone(p->p_cred, ncred); 1514 kauth_proc_chroot(ncred, p->p_cwdi); 1515 1516 /* Broadcast our credentials to the process and other LWPs. */ 1517 proc_crmod_leave(ncred, p->p_cred, true); 1518 } 1519 1520 /* 1521 * Common routine for chroot and chdir. 1522 * XXX "where" should be enum uio_seg 1523 */ 1524 int 1525 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1526 { 1527 struct pathbuf *pb; 1528 struct nameidata nd; 1529 int error; 1530 1531 error = pathbuf_maybe_copyin(path, where, &pb); 1532 if (error) { 1533 return error; 1534 } 1535 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1536 if ((error = namei(&nd)) != 0) { 1537 pathbuf_destroy(pb); 1538 return error; 1539 } 1540 *vpp = nd.ni_vp; 1541 pathbuf_destroy(pb); 1542 1543 if ((*vpp)->v_type != VDIR) 1544 error = ENOTDIR; 1545 else 1546 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1547 1548 if (error) 1549 vput(*vpp); 1550 else 1551 VOP_UNLOCK(*vpp); 1552 return (error); 1553 } 1554 1555 /* 1556 * Internals of sys_open - path has already been converted into a pathbuf 1557 * (so we can easily reuse this function from other parts of the kernel, 1558 * like posix_spawn post-processing). 1559 */ 1560 int 1561 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1562 int open_mode, int *fd) 1563 { 1564 struct proc *p = l->l_proc; 1565 struct cwdinfo *cwdi = p->p_cwdi; 1566 file_t *fp; 1567 struct vnode *vp; 1568 int flags, cmode; 1569 int indx, error; 1570 struct nameidata nd; 1571 1572 if (open_flags & O_SEARCH) { 1573 open_flags &= ~(int)O_SEARCH; 1574 } 1575 1576 /* 1577 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1578 * may be specified. 1579 */ 1580 if ((open_flags & O_EXEC) && (open_flags & O_ACCMODE)) 1581 return EINVAL; 1582 1583 flags = FFLAGS(open_flags); 1584 if ((flags & (FREAD | FWRITE)) == 0) 1585 return EINVAL; 1586 1587 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1588 return error; 1589 } 1590 1591 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1592 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1593 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1594 if (dvp != NULL) 1595 NDAT(&nd, dvp); 1596 1597 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1598 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1599 fd_abort(p, fp, indx); 1600 if ((error == EDUPFD || error == EMOVEFD) && 1601 l->l_dupfd >= 0 && /* XXX from fdopen */ 1602 (error = 1603 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1604 *fd = indx; 1605 return 0; 1606 } 1607 if (error == ERESTART) 1608 error = EINTR; 1609 return error; 1610 } 1611 1612 l->l_dupfd = 0; 1613 vp = nd.ni_vp; 1614 1615 if ((error = open_setfp(l, fp, vp, indx, flags))) 1616 return error; 1617 1618 VOP_UNLOCK(vp); 1619 *fd = indx; 1620 fd_affix(p, fp, indx); 1621 return 0; 1622 } 1623 1624 int 1625 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1626 { 1627 struct pathbuf *pb; 1628 int error, oflags; 1629 1630 oflags = FFLAGS(open_flags); 1631 if ((oflags & (FREAD | FWRITE)) == 0) 1632 return EINVAL; 1633 1634 pb = pathbuf_create(path); 1635 if (pb == NULL) 1636 return ENOMEM; 1637 1638 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1639 pathbuf_destroy(pb); 1640 1641 return error; 1642 } 1643 1644 static int 1645 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1646 int mode, int *fd) 1647 { 1648 file_t *dfp = NULL; 1649 struct vnode *dvp = NULL; 1650 struct pathbuf *pb; 1651 const char *pathstring = NULL; 1652 int error; 1653 1654 if (path == NULL) { 1655 MODULE_HOOK_CALL(vfs_openat_10_hook, (&pb), enosys(), error); 1656 if (error == ENOSYS) 1657 goto no_compat; 1658 if (error) 1659 return error; 1660 } else { 1661 no_compat: 1662 error = pathbuf_copyin(path, &pb); 1663 if (error) 1664 return error; 1665 } 1666 1667 pathstring = pathbuf_stringcopy_get(pb); 1668 1669 /* 1670 * fdat is ignored if: 1671 * 1) if fdat is AT_FDCWD, which means use current directory as base. 1672 * 2) if path is absolute, then fdat is useless. 1673 */ 1674 if (fdat != AT_FDCWD && pathstring[0] != '/') { 1675 /* fd_getvnode() will use the descriptor for us */ 1676 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1677 goto out; 1678 1679 dvp = dfp->f_vnode; 1680 } 1681 1682 error = do_open(l, dvp, pb, flags, mode, fd); 1683 1684 if (dfp != NULL) 1685 fd_putfile(fdat); 1686 out: 1687 pathbuf_stringcopy_put(pb, pathstring); 1688 pathbuf_destroy(pb); 1689 return error; 1690 } 1691 1692 int 1693 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1694 { 1695 /* { 1696 syscallarg(const char *) path; 1697 syscallarg(int) flags; 1698 syscallarg(int) mode; 1699 } */ 1700 int error; 1701 int fd; 1702 1703 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1704 SCARG(uap, flags), SCARG(uap, mode), &fd); 1705 1706 if (error == 0) 1707 *retval = fd; 1708 1709 return error; 1710 } 1711 1712 int 1713 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1714 { 1715 /* { 1716 syscallarg(int) fd; 1717 syscallarg(const char *) path; 1718 syscallarg(int) oflags; 1719 syscallarg(int) mode; 1720 } */ 1721 int error; 1722 int fd; 1723 1724 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1725 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1726 1727 if (error == 0) 1728 *retval = fd; 1729 1730 return error; 1731 } 1732 1733 static void 1734 vfs__fhfree(fhandle_t *fhp) 1735 { 1736 size_t fhsize; 1737 1738 fhsize = FHANDLE_SIZE(fhp); 1739 kmem_free(fhp, fhsize); 1740 } 1741 1742 /* 1743 * vfs_composefh: compose a filehandle. 1744 */ 1745 1746 int 1747 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1748 { 1749 struct mount *mp; 1750 struct fid *fidp; 1751 int error; 1752 size_t needfhsize; 1753 size_t fidsize; 1754 1755 mp = vp->v_mount; 1756 fidp = NULL; 1757 if (*fh_size < FHANDLE_SIZE_MIN) { 1758 fidsize = 0; 1759 } else { 1760 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1761 if (fhp != NULL) { 1762 memset(fhp, 0, *fh_size); 1763 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1764 fidp = &fhp->fh_fid; 1765 } 1766 } 1767 error = VFS_VPTOFH(vp, fidp, &fidsize); 1768 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1769 if (error == 0 && *fh_size < needfhsize) { 1770 error = E2BIG; 1771 } 1772 *fh_size = needfhsize; 1773 return error; 1774 } 1775 1776 int 1777 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1778 { 1779 struct mount *mp; 1780 fhandle_t *fhp; 1781 size_t fhsize; 1782 size_t fidsize; 1783 int error; 1784 1785 mp = vp->v_mount; 1786 fidsize = 0; 1787 error = VFS_VPTOFH(vp, NULL, &fidsize); 1788 KASSERT(error != 0); 1789 if (error != E2BIG) { 1790 goto out; 1791 } 1792 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1793 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1794 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1795 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1796 if (error == 0) { 1797 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1798 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1799 *fhpp = fhp; 1800 } else { 1801 kmem_free(fhp, fhsize); 1802 } 1803 out: 1804 return error; 1805 } 1806 1807 void 1808 vfs_composefh_free(fhandle_t *fhp) 1809 { 1810 1811 vfs__fhfree(fhp); 1812 } 1813 1814 /* 1815 * vfs_fhtovp: lookup a vnode by a filehandle. 1816 */ 1817 1818 int 1819 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1820 { 1821 struct mount *mp; 1822 int error; 1823 1824 *vpp = NULL; 1825 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1826 if (mp == NULL) { 1827 error = ESTALE; 1828 goto out; 1829 } 1830 if (mp->mnt_op->vfs_fhtovp == NULL) { 1831 error = EOPNOTSUPP; 1832 goto out; 1833 } 1834 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1835 out: 1836 return error; 1837 } 1838 1839 /* 1840 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1841 * the needed size. 1842 */ 1843 1844 int 1845 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1846 { 1847 fhandle_t *fhp; 1848 int error; 1849 1850 if (fhsize > FHANDLE_SIZE_MAX) { 1851 return EINVAL; 1852 } 1853 if (fhsize < FHANDLE_SIZE_MIN) { 1854 return EINVAL; 1855 } 1856 again: 1857 fhp = kmem_alloc(fhsize, KM_SLEEP); 1858 error = copyin(ufhp, fhp, fhsize); 1859 if (error == 0) { 1860 /* XXX this check shouldn't be here */ 1861 if (FHANDLE_SIZE(fhp) == fhsize) { 1862 *fhpp = fhp; 1863 return 0; 1864 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1865 /* 1866 * a kludge for nfsv2 padded handles. 1867 */ 1868 size_t sz; 1869 1870 sz = FHANDLE_SIZE(fhp); 1871 kmem_free(fhp, fhsize); 1872 fhsize = sz; 1873 goto again; 1874 } else { 1875 /* 1876 * userland told us wrong size. 1877 */ 1878 error = EINVAL; 1879 } 1880 } 1881 kmem_free(fhp, fhsize); 1882 return error; 1883 } 1884 1885 void 1886 vfs_copyinfh_free(fhandle_t *fhp) 1887 { 1888 1889 vfs__fhfree(fhp); 1890 } 1891 1892 /* 1893 * Get file handle system call 1894 */ 1895 int 1896 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1897 { 1898 /* { 1899 syscallarg(char *) fname; 1900 syscallarg(fhandle_t *) fhp; 1901 syscallarg(size_t *) fh_size; 1902 } */ 1903 struct vnode *vp; 1904 fhandle_t *fh; 1905 int error; 1906 struct pathbuf *pb; 1907 struct nameidata nd; 1908 size_t sz; 1909 size_t usz; 1910 1911 /* 1912 * Must be super user 1913 */ 1914 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1915 0, NULL, NULL, NULL); 1916 if (error) 1917 return (error); 1918 1919 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1920 if (error) { 1921 return error; 1922 } 1923 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1924 error = namei(&nd); 1925 if (error) { 1926 pathbuf_destroy(pb); 1927 return error; 1928 } 1929 vp = nd.ni_vp; 1930 pathbuf_destroy(pb); 1931 1932 error = vfs_composefh_alloc(vp, &fh); 1933 vput(vp); 1934 if (error != 0) { 1935 return error; 1936 } 1937 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1938 if (error != 0) { 1939 goto out; 1940 } 1941 sz = FHANDLE_SIZE(fh); 1942 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1943 if (error != 0) { 1944 goto out; 1945 } 1946 if (usz >= sz) { 1947 error = copyout(fh, SCARG(uap, fhp), sz); 1948 } else { 1949 error = E2BIG; 1950 } 1951 out: 1952 vfs_composefh_free(fh); 1953 return (error); 1954 } 1955 1956 /* 1957 * Open a file given a file handle. 1958 * 1959 * Check permissions, allocate an open file structure, 1960 * and call the device open routine if any. 1961 */ 1962 1963 int 1964 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1965 register_t *retval) 1966 { 1967 file_t *fp; 1968 struct vnode *vp = NULL; 1969 kauth_cred_t cred = l->l_cred; 1970 file_t *nfp; 1971 int indx, error; 1972 struct vattr va; 1973 fhandle_t *fh; 1974 int flags; 1975 proc_t *p; 1976 1977 p = curproc; 1978 1979 /* 1980 * Must be super user 1981 */ 1982 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1983 0, NULL, NULL, NULL))) 1984 return (error); 1985 1986 if (oflags & O_SEARCH) { 1987 oflags &= ~(int)O_SEARCH; 1988 } 1989 1990 flags = FFLAGS(oflags); 1991 if ((flags & (FREAD | FWRITE)) == 0) 1992 return (EINVAL); 1993 if ((flags & O_CREAT)) 1994 return (EINVAL); 1995 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1996 return (error); 1997 fp = nfp; 1998 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1999 if (error != 0) { 2000 goto bad; 2001 } 2002 error = vfs_fhtovp(fh, &vp); 2003 vfs_copyinfh_free(fh); 2004 if (error != 0) { 2005 goto bad; 2006 } 2007 2008 /* Now do an effective vn_open */ 2009 2010 if (vp->v_type == VSOCK) { 2011 error = EOPNOTSUPP; 2012 goto bad; 2013 } 2014 error = vn_openchk(vp, cred, flags); 2015 if (error != 0) 2016 goto bad; 2017 if (flags & O_TRUNC) { 2018 VOP_UNLOCK(vp); /* XXX */ 2019 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2020 vattr_null(&va); 2021 va.va_size = 0; 2022 error = VOP_SETATTR(vp, &va, cred); 2023 if (error) 2024 goto bad; 2025 } 2026 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2027 goto bad; 2028 if (flags & FWRITE) { 2029 mutex_enter(vp->v_interlock); 2030 vp->v_writecount++; 2031 mutex_exit(vp->v_interlock); 2032 } 2033 2034 /* done with modified vn_open, now finish what sys_open does. */ 2035 if ((error = open_setfp(l, fp, vp, indx, flags))) 2036 return error; 2037 2038 VOP_UNLOCK(vp); 2039 *retval = indx; 2040 fd_affix(p, fp, indx); 2041 return (0); 2042 2043 bad: 2044 fd_abort(p, fp, indx); 2045 if (vp != NULL) 2046 vput(vp); 2047 return (error); 2048 } 2049 2050 int 2051 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2052 { 2053 /* { 2054 syscallarg(const void *) fhp; 2055 syscallarg(size_t) fh_size; 2056 syscallarg(int) flags; 2057 } */ 2058 2059 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2060 SCARG(uap, flags), retval); 2061 } 2062 2063 int 2064 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2065 { 2066 int error; 2067 fhandle_t *fh; 2068 struct vnode *vp; 2069 2070 /* 2071 * Must be super user 2072 */ 2073 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2074 0, NULL, NULL, NULL))) 2075 return (error); 2076 2077 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2078 if (error != 0) 2079 return error; 2080 2081 error = vfs_fhtovp(fh, &vp); 2082 vfs_copyinfh_free(fh); 2083 if (error != 0) 2084 return error; 2085 2086 error = vn_stat(vp, sb); 2087 vput(vp); 2088 return error; 2089 } 2090 2091 2092 /* ARGSUSED */ 2093 int 2094 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2095 { 2096 /* { 2097 syscallarg(const void *) fhp; 2098 syscallarg(size_t) fh_size; 2099 syscallarg(struct stat *) sb; 2100 } */ 2101 struct stat sb; 2102 int error; 2103 2104 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2105 if (error) 2106 return error; 2107 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2108 } 2109 2110 int 2111 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2112 int flags) 2113 { 2114 fhandle_t *fh; 2115 struct mount *mp; 2116 struct vnode *vp; 2117 int error; 2118 2119 /* 2120 * Must be super user 2121 */ 2122 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2123 0, NULL, NULL, NULL))) 2124 return error; 2125 2126 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2127 if (error != 0) 2128 return error; 2129 2130 error = vfs_fhtovp(fh, &vp); 2131 vfs_copyinfh_free(fh); 2132 if (error != 0) 2133 return error; 2134 2135 mp = vp->v_mount; 2136 error = dostatvfs(mp, sb, l, flags, 1); 2137 vput(vp); 2138 return error; 2139 } 2140 2141 /* ARGSUSED */ 2142 int 2143 sys___fhstatvfs190(struct lwp *l, const struct sys___fhstatvfs190_args *uap, register_t *retval) 2144 { 2145 /* { 2146 syscallarg(const void *) fhp; 2147 syscallarg(size_t) fh_size; 2148 syscallarg(struct statvfs *) buf; 2149 syscallarg(int) flags; 2150 } */ 2151 struct statvfs *sb = STATVFSBUF_GET(); 2152 int error; 2153 2154 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2155 SCARG(uap, flags)); 2156 if (error == 0) 2157 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2158 STATVFSBUF_PUT(sb); 2159 return error; 2160 } 2161 2162 int 2163 do_posix_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2164 dev_t dev) 2165 { 2166 2167 /* 2168 * The POSIX mknod(2) call is an alias for mkfifo(2) for S_IFIFO 2169 * in mode and dev=0. 2170 * 2171 * In all the other cases it's implementation defined behavior. 2172 */ 2173 2174 if ((mode & S_IFIFO) && dev == 0) 2175 return do_sys_mkfifoat(l, fdat, pathname, mode); 2176 else 2177 return do_sys_mknodat(l, fdat, pathname, mode, dev, 2178 UIO_USERSPACE); 2179 } 2180 2181 /* 2182 * Create a special file. 2183 */ 2184 /* ARGSUSED */ 2185 int 2186 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2187 register_t *retval) 2188 { 2189 /* { 2190 syscallarg(const char *) path; 2191 syscallarg(mode_t) mode; 2192 syscallarg(dev_t) dev; 2193 } */ 2194 return do_posix_mknodat(l, AT_FDCWD, SCARG(uap, path), 2195 SCARG(uap, mode), SCARG(uap, dev)); 2196 } 2197 2198 int 2199 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2200 register_t *retval) 2201 { 2202 /* { 2203 syscallarg(int) fd; 2204 syscallarg(const char *) path; 2205 syscallarg(mode_t) mode; 2206 syscallarg(int) pad; 2207 syscallarg(dev_t) dev; 2208 } */ 2209 2210 return do_posix_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2211 SCARG(uap, mode), SCARG(uap, dev)); 2212 } 2213 2214 int 2215 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2216 enum uio_seg seg) 2217 { 2218 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, seg); 2219 } 2220 2221 int 2222 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2223 dev_t dev, enum uio_seg seg) 2224 { 2225 struct proc *p = l->l_proc; 2226 struct vnode *vp; 2227 struct vattr vattr; 2228 int error, optype; 2229 struct pathbuf *pb; 2230 struct nameidata nd; 2231 const char *pathstring; 2232 2233 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2234 0, NULL, NULL, NULL)) != 0) 2235 return (error); 2236 2237 optype = VOP_MKNOD_DESCOFFSET; 2238 2239 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2240 if (error) { 2241 return error; 2242 } 2243 pathstring = pathbuf_stringcopy_get(pb); 2244 if (pathstring == NULL) { 2245 pathbuf_destroy(pb); 2246 return ENOMEM; 2247 } 2248 2249 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2250 2251 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2252 goto out; 2253 vp = nd.ni_vp; 2254 2255 if (vp != NULL) 2256 error = EEXIST; 2257 else { 2258 vattr_null(&vattr); 2259 /* We will read cwdi->cwdi_cmask unlocked. */ 2260 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2261 vattr.va_rdev = dev; 2262 2263 switch (mode & S_IFMT) { 2264 case S_IFMT: /* used by badsect to flag bad sectors */ 2265 vattr.va_type = VBAD; 2266 break; 2267 case S_IFCHR: 2268 vattr.va_type = VCHR; 2269 break; 2270 case S_IFBLK: 2271 vattr.va_type = VBLK; 2272 break; 2273 case S_IFWHT: 2274 optype = VOP_WHITEOUT_DESCOFFSET; 2275 break; 2276 case S_IFREG: 2277 #if NVERIEXEC > 0 2278 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2279 O_CREAT); 2280 #endif /* NVERIEXEC > 0 */ 2281 vattr.va_type = VREG; 2282 vattr.va_rdev = VNOVAL; 2283 optype = VOP_CREATE_DESCOFFSET; 2284 break; 2285 default: 2286 error = EINVAL; 2287 break; 2288 } 2289 } 2290 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2291 && vattr.va_rdev == VNOVAL) 2292 error = EINVAL; 2293 if (!error) { 2294 switch (optype) { 2295 case VOP_WHITEOUT_DESCOFFSET: 2296 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2297 if (error) 2298 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2299 vput(nd.ni_dvp); 2300 break; 2301 2302 case VOP_MKNOD_DESCOFFSET: 2303 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2304 &nd.ni_cnd, &vattr); 2305 if (error == 0) 2306 vrele(nd.ni_vp); 2307 vput(nd.ni_dvp); 2308 break; 2309 2310 case VOP_CREATE_DESCOFFSET: 2311 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2312 &nd.ni_cnd, &vattr); 2313 if (error == 0) 2314 vrele(nd.ni_vp); 2315 vput(nd.ni_dvp); 2316 break; 2317 } 2318 } else { 2319 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2320 if (nd.ni_dvp == vp) 2321 vrele(nd.ni_dvp); 2322 else 2323 vput(nd.ni_dvp); 2324 if (vp) 2325 vrele(vp); 2326 } 2327 out: 2328 pathbuf_stringcopy_put(pb, pathstring); 2329 pathbuf_destroy(pb); 2330 return (error); 2331 } 2332 2333 /* 2334 * Create a named pipe. 2335 */ 2336 /* ARGSUSED */ 2337 int 2338 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2339 { 2340 /* { 2341 syscallarg(const char *) path; 2342 syscallarg(int) mode; 2343 } */ 2344 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2345 } 2346 2347 int 2348 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2349 register_t *retval) 2350 { 2351 /* { 2352 syscallarg(int) fd; 2353 syscallarg(const char *) path; 2354 syscallarg(int) mode; 2355 } */ 2356 2357 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2358 SCARG(uap, mode)); 2359 } 2360 2361 static int 2362 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2363 { 2364 struct proc *p = l->l_proc; 2365 struct vattr vattr; 2366 int error; 2367 struct pathbuf *pb; 2368 struct nameidata nd; 2369 2370 error = pathbuf_copyin(path, &pb); 2371 if (error) { 2372 return error; 2373 } 2374 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2375 2376 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2377 pathbuf_destroy(pb); 2378 return error; 2379 } 2380 if (nd.ni_vp != NULL) { 2381 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2382 if (nd.ni_dvp == nd.ni_vp) 2383 vrele(nd.ni_dvp); 2384 else 2385 vput(nd.ni_dvp); 2386 vrele(nd.ni_vp); 2387 pathbuf_destroy(pb); 2388 return (EEXIST); 2389 } 2390 vattr_null(&vattr); 2391 vattr.va_type = VFIFO; 2392 /* We will read cwdi->cwdi_cmask unlocked. */ 2393 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2394 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2395 if (error == 0) 2396 vrele(nd.ni_vp); 2397 vput(nd.ni_dvp); 2398 pathbuf_destroy(pb); 2399 return (error); 2400 } 2401 2402 /* 2403 * Make a hard file link. 2404 */ 2405 /* ARGSUSED */ 2406 int 2407 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2408 const char *link, int follow, register_t *retval) 2409 { 2410 struct vnode *vp; 2411 struct pathbuf *linkpb; 2412 struct nameidata nd; 2413 namei_simple_flags_t ns_flags; 2414 int error; 2415 2416 if (follow & AT_SYMLINK_FOLLOW) 2417 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2418 else 2419 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2420 2421 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2422 if (error != 0) 2423 return (error); 2424 error = pathbuf_copyin(link, &linkpb); 2425 if (error) { 2426 goto out1; 2427 } 2428 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2429 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2430 goto out2; 2431 if (nd.ni_vp) { 2432 error = EEXIST; 2433 goto abortop; 2434 } 2435 /* Prevent hard links on directories. */ 2436 if (vp->v_type == VDIR) { 2437 error = EPERM; 2438 goto abortop; 2439 } 2440 /* Prevent cross-mount operation. */ 2441 if (nd.ni_dvp->v_mount != vp->v_mount) { 2442 error = EXDEV; 2443 goto abortop; 2444 } 2445 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2446 VOP_UNLOCK(nd.ni_dvp); 2447 vrele(nd.ni_dvp); 2448 out2: 2449 pathbuf_destroy(linkpb); 2450 out1: 2451 vrele(vp); 2452 return (error); 2453 abortop: 2454 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2455 if (nd.ni_dvp == nd.ni_vp) 2456 vrele(nd.ni_dvp); 2457 else 2458 vput(nd.ni_dvp); 2459 if (nd.ni_vp != NULL) 2460 vrele(nd.ni_vp); 2461 goto out2; 2462 } 2463 2464 int 2465 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2466 { 2467 /* { 2468 syscallarg(const char *) path; 2469 syscallarg(const char *) link; 2470 } */ 2471 const char *path = SCARG(uap, path); 2472 const char *link = SCARG(uap, link); 2473 2474 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2475 AT_SYMLINK_FOLLOW, retval); 2476 } 2477 2478 int 2479 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2480 register_t *retval) 2481 { 2482 /* { 2483 syscallarg(int) fd1; 2484 syscallarg(const char *) name1; 2485 syscallarg(int) fd2; 2486 syscallarg(const char *) name2; 2487 syscallarg(int) flags; 2488 } */ 2489 int fd1 = SCARG(uap, fd1); 2490 const char *name1 = SCARG(uap, name1); 2491 int fd2 = SCARG(uap, fd2); 2492 const char *name2 = SCARG(uap, name2); 2493 int follow; 2494 2495 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2496 2497 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2498 } 2499 2500 2501 int 2502 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2503 { 2504 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2505 } 2506 2507 static int 2508 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2509 const char *link, enum uio_seg seg) 2510 { 2511 struct proc *p = curproc; 2512 struct vattr vattr; 2513 char *path; 2514 int error; 2515 size_t len; 2516 struct pathbuf *linkpb; 2517 struct nameidata nd; 2518 2519 KASSERT(l != NULL || fdat == AT_FDCWD); 2520 2521 path = PNBUF_GET(); 2522 if (seg == UIO_USERSPACE) { 2523 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2524 goto out1; 2525 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2526 goto out1; 2527 } else { 2528 len = strlen(patharg) + 1; 2529 KASSERT(len <= MAXPATHLEN); 2530 memcpy(path, patharg, len); 2531 linkpb = pathbuf_create(link); 2532 if (linkpb == NULL) { 2533 error = ENOMEM; 2534 goto out1; 2535 } 2536 } 2537 ktrkuser("symlink-target", path, len - 1); 2538 2539 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2540 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2541 goto out2; 2542 if (nd.ni_vp) { 2543 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2544 if (nd.ni_dvp == nd.ni_vp) 2545 vrele(nd.ni_dvp); 2546 else 2547 vput(nd.ni_dvp); 2548 vrele(nd.ni_vp); 2549 error = EEXIST; 2550 goto out2; 2551 } 2552 vattr_null(&vattr); 2553 vattr.va_type = VLNK; 2554 /* We will read cwdi->cwdi_cmask unlocked. */ 2555 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2556 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2557 if (error == 0) 2558 vrele(nd.ni_vp); 2559 vput(nd.ni_dvp); 2560 out2: 2561 pathbuf_destroy(linkpb); 2562 out1: 2563 PNBUF_PUT(path); 2564 return (error); 2565 } 2566 2567 /* 2568 * Make a symbolic link. 2569 */ 2570 /* ARGSUSED */ 2571 int 2572 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2573 { 2574 /* { 2575 syscallarg(const char *) path; 2576 syscallarg(const char *) link; 2577 } */ 2578 2579 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2580 UIO_USERSPACE); 2581 } 2582 2583 int 2584 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2585 register_t *retval) 2586 { 2587 /* { 2588 syscallarg(const char *) path1; 2589 syscallarg(int) fd; 2590 syscallarg(const char *) path2; 2591 } */ 2592 2593 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2594 SCARG(uap, path2), UIO_USERSPACE); 2595 } 2596 2597 /* 2598 * Delete a whiteout from the filesystem. 2599 */ 2600 /* ARGSUSED */ 2601 int 2602 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2603 { 2604 /* { 2605 syscallarg(const char *) path; 2606 } */ 2607 int error; 2608 struct pathbuf *pb; 2609 struct nameidata nd; 2610 2611 error = pathbuf_copyin(SCARG(uap, path), &pb); 2612 if (error) { 2613 return error; 2614 } 2615 2616 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2617 error = namei(&nd); 2618 if (error) { 2619 pathbuf_destroy(pb); 2620 return (error); 2621 } 2622 2623 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2624 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2625 if (nd.ni_dvp == nd.ni_vp) 2626 vrele(nd.ni_dvp); 2627 else 2628 vput(nd.ni_dvp); 2629 if (nd.ni_vp) 2630 vrele(nd.ni_vp); 2631 pathbuf_destroy(pb); 2632 return (EEXIST); 2633 } 2634 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2635 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2636 vput(nd.ni_dvp); 2637 pathbuf_destroy(pb); 2638 return (error); 2639 } 2640 2641 /* 2642 * Delete a name from the filesystem. 2643 */ 2644 /* ARGSUSED */ 2645 int 2646 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2647 { 2648 /* { 2649 syscallarg(const char *) path; 2650 } */ 2651 2652 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2653 } 2654 2655 int 2656 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2657 register_t *retval) 2658 { 2659 /* { 2660 syscallarg(int) fd; 2661 syscallarg(const char *) path; 2662 syscallarg(int) flag; 2663 } */ 2664 2665 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2666 SCARG(uap, flag), UIO_USERSPACE); 2667 } 2668 2669 int 2670 do_sys_unlink(const char *arg, enum uio_seg seg) 2671 { 2672 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2673 } 2674 2675 static int 2676 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2677 enum uio_seg seg) 2678 { 2679 struct vnode *vp; 2680 int error; 2681 struct pathbuf *pb; 2682 struct nameidata nd; 2683 const char *pathstring; 2684 2685 KASSERT(l != NULL || fdat == AT_FDCWD); 2686 2687 error = pathbuf_maybe_copyin(arg, seg, &pb); 2688 if (error) { 2689 return error; 2690 } 2691 pathstring = pathbuf_stringcopy_get(pb); 2692 if (pathstring == NULL) { 2693 pathbuf_destroy(pb); 2694 return ENOMEM; 2695 } 2696 2697 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2698 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2699 goto out; 2700 vp = nd.ni_vp; 2701 2702 /* 2703 * The root of a mounted filesystem cannot be deleted. 2704 */ 2705 if ((vp->v_vflag & VV_ROOT) != 0) { 2706 error = EBUSY; 2707 goto abort; 2708 } 2709 2710 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2711 error = EBUSY; 2712 goto abort; 2713 } 2714 2715 /* 2716 * No rmdir "." please. 2717 */ 2718 if (nd.ni_dvp == vp) { 2719 error = EINVAL; 2720 goto abort; 2721 } 2722 2723 /* 2724 * AT_REMOVEDIR is required to remove a directory 2725 */ 2726 if (vp->v_type == VDIR) { 2727 if (!(flags & AT_REMOVEDIR)) { 2728 error = EPERM; 2729 goto abort; 2730 } else { 2731 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2732 vput(nd.ni_dvp); 2733 goto out; 2734 } 2735 } 2736 2737 /* 2738 * Starting here we only deal with non directories. 2739 */ 2740 if (flags & AT_REMOVEDIR) { 2741 error = ENOTDIR; 2742 goto abort; 2743 } 2744 2745 #if NVERIEXEC > 0 2746 /* Handle remove requests for veriexec entries. */ 2747 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2748 goto abort; 2749 } 2750 #endif /* NVERIEXEC > 0 */ 2751 2752 #ifdef FILEASSOC 2753 (void)fileassoc_file_delete(vp); 2754 #endif /* FILEASSOC */ 2755 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2756 vput(nd.ni_dvp); 2757 goto out; 2758 2759 abort: 2760 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2761 if (nd.ni_dvp == vp) 2762 vrele(nd.ni_dvp); 2763 else 2764 vput(nd.ni_dvp); 2765 vput(vp); 2766 2767 out: 2768 pathbuf_stringcopy_put(pb, pathstring); 2769 pathbuf_destroy(pb); 2770 return (error); 2771 } 2772 2773 /* 2774 * Reposition read/write file offset. 2775 */ 2776 int 2777 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2778 { 2779 /* { 2780 syscallarg(int) fd; 2781 syscallarg(int) pad; 2782 syscallarg(off_t) offset; 2783 syscallarg(int) whence; 2784 } */ 2785 kauth_cred_t cred = l->l_cred; 2786 file_t *fp; 2787 struct vnode *vp; 2788 struct vattr vattr; 2789 off_t newoff; 2790 int error, fd; 2791 2792 fd = SCARG(uap, fd); 2793 2794 if ((fp = fd_getfile(fd)) == NULL) 2795 return (EBADF); 2796 2797 vp = fp->f_vnode; 2798 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2799 error = ESPIPE; 2800 goto out; 2801 } 2802 2803 vn_lock(vp, LK_SHARED | LK_RETRY); 2804 2805 switch (SCARG(uap, whence)) { 2806 case SEEK_CUR: 2807 newoff = fp->f_offset + SCARG(uap, offset); 2808 break; 2809 case SEEK_END: 2810 error = VOP_GETATTR(vp, &vattr, cred); 2811 if (error) { 2812 VOP_UNLOCK(vp); 2813 goto out; 2814 } 2815 newoff = SCARG(uap, offset) + vattr.va_size; 2816 break; 2817 case SEEK_SET: 2818 newoff = SCARG(uap, offset); 2819 break; 2820 default: 2821 error = EINVAL; 2822 VOP_UNLOCK(vp); 2823 goto out; 2824 } 2825 VOP_UNLOCK(vp); 2826 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2827 *(off_t *)retval = fp->f_offset = newoff; 2828 } 2829 out: 2830 fd_putfile(fd); 2831 return (error); 2832 } 2833 2834 /* 2835 * Positional read system call. 2836 */ 2837 int 2838 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2839 { 2840 /* { 2841 syscallarg(int) fd; 2842 syscallarg(void *) buf; 2843 syscallarg(size_t) nbyte; 2844 syscallarg(off_t) offset; 2845 } */ 2846 file_t *fp; 2847 struct vnode *vp; 2848 off_t offset; 2849 int error, fd = SCARG(uap, fd); 2850 2851 if ((fp = fd_getfile(fd)) == NULL) 2852 return (EBADF); 2853 2854 if ((fp->f_flag & FREAD) == 0) { 2855 fd_putfile(fd); 2856 return (EBADF); 2857 } 2858 2859 vp = fp->f_vnode; 2860 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2861 error = ESPIPE; 2862 goto out; 2863 } 2864 2865 offset = SCARG(uap, offset); 2866 2867 /* 2868 * XXX This works because no file systems actually 2869 * XXX take any action on the seek operation. 2870 */ 2871 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2872 goto out; 2873 2874 /* dofileread() will unuse the descriptor for us */ 2875 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2876 &offset, 0, retval)); 2877 2878 out: 2879 fd_putfile(fd); 2880 return (error); 2881 } 2882 2883 /* 2884 * Positional scatter read system call. 2885 */ 2886 int 2887 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2888 { 2889 /* { 2890 syscallarg(int) fd; 2891 syscallarg(const struct iovec *) iovp; 2892 syscallarg(int) iovcnt; 2893 syscallarg(off_t) offset; 2894 } */ 2895 off_t offset = SCARG(uap, offset); 2896 2897 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2898 SCARG(uap, iovcnt), &offset, 0, retval); 2899 } 2900 2901 /* 2902 * Positional write system call. 2903 */ 2904 int 2905 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2906 { 2907 /* { 2908 syscallarg(int) fd; 2909 syscallarg(const void *) buf; 2910 syscallarg(size_t) nbyte; 2911 syscallarg(off_t) offset; 2912 } */ 2913 file_t *fp; 2914 struct vnode *vp; 2915 off_t offset; 2916 int error, fd = SCARG(uap, fd); 2917 2918 if ((fp = fd_getfile(fd)) == NULL) 2919 return (EBADF); 2920 2921 if ((fp->f_flag & FWRITE) == 0) { 2922 fd_putfile(fd); 2923 return (EBADF); 2924 } 2925 2926 vp = fp->f_vnode; 2927 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2928 error = ESPIPE; 2929 goto out; 2930 } 2931 2932 offset = SCARG(uap, offset); 2933 2934 /* 2935 * XXX This works because no file systems actually 2936 * XXX take any action on the seek operation. 2937 */ 2938 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2939 goto out; 2940 2941 /* dofilewrite() will unuse the descriptor for us */ 2942 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2943 &offset, 0, retval)); 2944 2945 out: 2946 fd_putfile(fd); 2947 return (error); 2948 } 2949 2950 /* 2951 * Positional gather write system call. 2952 */ 2953 int 2954 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2955 { 2956 /* { 2957 syscallarg(int) fd; 2958 syscallarg(const struct iovec *) iovp; 2959 syscallarg(int) iovcnt; 2960 syscallarg(off_t) offset; 2961 } */ 2962 off_t offset = SCARG(uap, offset); 2963 2964 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2965 SCARG(uap, iovcnt), &offset, 0, retval); 2966 } 2967 2968 /* 2969 * Check access permissions. 2970 */ 2971 int 2972 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2973 { 2974 /* { 2975 syscallarg(const char *) path; 2976 syscallarg(int) flags; 2977 } */ 2978 2979 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2980 SCARG(uap, flags), 0); 2981 } 2982 2983 int 2984 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2985 int mode, int flags) 2986 { 2987 kauth_cred_t cred; 2988 struct vnode *vp; 2989 int error, nd_flag, vmode; 2990 struct pathbuf *pb; 2991 struct nameidata nd; 2992 2993 CTASSERT(F_OK == 0); 2994 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2995 /* nonsense mode */ 2996 return EINVAL; 2997 } 2998 2999 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 3000 if (flags & AT_SYMLINK_NOFOLLOW) 3001 nd_flag &= ~FOLLOW; 3002 3003 error = pathbuf_copyin(path, &pb); 3004 if (error) 3005 return error; 3006 3007 NDINIT(&nd, LOOKUP, nd_flag, pb); 3008 3009 /* Override default credentials */ 3010 cred = kauth_cred_dup(l->l_cred); 3011 if (!(flags & AT_EACCESS)) { 3012 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 3013 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 3014 } 3015 nd.ni_cnd.cn_cred = cred; 3016 3017 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3018 pathbuf_destroy(pb); 3019 goto out; 3020 } 3021 vp = nd.ni_vp; 3022 pathbuf_destroy(pb); 3023 3024 /* Flags == 0 means only check for existence. */ 3025 if (mode) { 3026 vmode = 0; 3027 if (mode & R_OK) 3028 vmode |= VREAD; 3029 if (mode & W_OK) 3030 vmode |= VWRITE; 3031 if (mode & X_OK) 3032 vmode |= VEXEC; 3033 3034 error = VOP_ACCESS(vp, vmode, cred); 3035 if (!error && (vmode & VWRITE)) 3036 error = vn_writechk(vp); 3037 } 3038 vput(vp); 3039 out: 3040 kauth_cred_free(cred); 3041 return (error); 3042 } 3043 3044 int 3045 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3046 register_t *retval) 3047 { 3048 /* { 3049 syscallarg(int) fd; 3050 syscallarg(const char *) path; 3051 syscallarg(int) amode; 3052 syscallarg(int) flag; 3053 } */ 3054 3055 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3056 SCARG(uap, amode), SCARG(uap, flag)); 3057 } 3058 3059 /* 3060 * Common code for all sys_stat functions, including compat versions. 3061 */ 3062 int 3063 do_sys_stat(const char *userpath, unsigned int nd_flag, 3064 struct stat *sb) 3065 { 3066 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3067 } 3068 3069 int 3070 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3071 unsigned int nd_flag, struct stat *sb) 3072 { 3073 int error; 3074 struct pathbuf *pb; 3075 struct nameidata nd; 3076 3077 KASSERT(l != NULL || fdat == AT_FDCWD); 3078 3079 error = pathbuf_copyin(userpath, &pb); 3080 if (error) { 3081 return error; 3082 } 3083 3084 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3085 3086 error = fd_nameiat(l, fdat, &nd); 3087 if (error != 0) { 3088 pathbuf_destroy(pb); 3089 return error; 3090 } 3091 error = vn_stat(nd.ni_vp, sb); 3092 vput(nd.ni_vp); 3093 pathbuf_destroy(pb); 3094 return error; 3095 } 3096 3097 /* 3098 * Get file status; this version follows links. 3099 */ 3100 /* ARGSUSED */ 3101 int 3102 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3103 { 3104 /* { 3105 syscallarg(const char *) path; 3106 syscallarg(struct stat *) ub; 3107 } */ 3108 struct stat sb; 3109 int error; 3110 3111 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3112 if (error) 3113 return error; 3114 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3115 } 3116 3117 /* 3118 * Get file status; this version does not follow links. 3119 */ 3120 /* ARGSUSED */ 3121 int 3122 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3123 { 3124 /* { 3125 syscallarg(const char *) path; 3126 syscallarg(struct stat *) ub; 3127 } */ 3128 struct stat sb; 3129 int error; 3130 3131 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3132 if (error) 3133 return error; 3134 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3135 } 3136 3137 int 3138 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3139 register_t *retval) 3140 { 3141 /* { 3142 syscallarg(int) fd; 3143 syscallarg(const char *) path; 3144 syscallarg(struct stat *) buf; 3145 syscallarg(int) flag; 3146 } */ 3147 unsigned int nd_flag; 3148 struct stat sb; 3149 int error; 3150 3151 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3152 nd_flag = NOFOLLOW; 3153 else 3154 nd_flag = FOLLOW; 3155 3156 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3157 &sb); 3158 if (error) 3159 return error; 3160 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3161 } 3162 3163 /* 3164 * Get configurable pathname variables. 3165 */ 3166 /* ARGSUSED */ 3167 int 3168 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3169 { 3170 /* { 3171 syscallarg(const char *) path; 3172 syscallarg(int) name; 3173 } */ 3174 int error; 3175 struct pathbuf *pb; 3176 struct nameidata nd; 3177 3178 error = pathbuf_copyin(SCARG(uap, path), &pb); 3179 if (error) { 3180 return error; 3181 } 3182 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3183 if ((error = namei(&nd)) != 0) { 3184 pathbuf_destroy(pb); 3185 return (error); 3186 } 3187 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3188 vput(nd.ni_vp); 3189 pathbuf_destroy(pb); 3190 return (error); 3191 } 3192 3193 /* 3194 * Return target name of a symbolic link. 3195 */ 3196 /* ARGSUSED */ 3197 int 3198 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3199 register_t *retval) 3200 { 3201 /* { 3202 syscallarg(const char *) path; 3203 syscallarg(char *) buf; 3204 syscallarg(size_t) count; 3205 } */ 3206 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3207 SCARG(uap, buf), SCARG(uap, count), retval); 3208 } 3209 3210 static int 3211 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3212 size_t count, register_t *retval) 3213 { 3214 struct vnode *vp; 3215 struct iovec aiov; 3216 struct uio auio; 3217 int error; 3218 struct pathbuf *pb; 3219 struct nameidata nd; 3220 3221 error = pathbuf_copyin(path, &pb); 3222 if (error) { 3223 return error; 3224 } 3225 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3226 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3227 pathbuf_destroy(pb); 3228 return error; 3229 } 3230 vp = nd.ni_vp; 3231 pathbuf_destroy(pb); 3232 if (vp->v_type != VLNK) 3233 error = EINVAL; 3234 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3235 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3236 aiov.iov_base = buf; 3237 aiov.iov_len = count; 3238 auio.uio_iov = &aiov; 3239 auio.uio_iovcnt = 1; 3240 auio.uio_offset = 0; 3241 auio.uio_rw = UIO_READ; 3242 KASSERT(l == curlwp); 3243 auio.uio_vmspace = l->l_proc->p_vmspace; 3244 auio.uio_resid = count; 3245 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3246 *retval = count - auio.uio_resid; 3247 } 3248 vput(vp); 3249 return (error); 3250 } 3251 3252 int 3253 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3254 register_t *retval) 3255 { 3256 /* { 3257 syscallarg(int) fd; 3258 syscallarg(const char *) path; 3259 syscallarg(char *) buf; 3260 syscallarg(size_t) bufsize; 3261 } */ 3262 3263 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3264 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3265 } 3266 3267 /* 3268 * Change flags of a file given a path name. 3269 */ 3270 /* ARGSUSED */ 3271 int 3272 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3273 { 3274 /* { 3275 syscallarg(const char *) path; 3276 syscallarg(u_long) flags; 3277 } */ 3278 struct vnode *vp; 3279 int error; 3280 3281 error = namei_simple_user(SCARG(uap, path), 3282 NSM_FOLLOW_TRYEMULROOT, &vp); 3283 if (error != 0) 3284 return (error); 3285 error = change_flags(vp, SCARG(uap, flags), l); 3286 vput(vp); 3287 return (error); 3288 } 3289 3290 /* 3291 * Change flags of a file given a file descriptor. 3292 */ 3293 /* ARGSUSED */ 3294 int 3295 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3296 { 3297 /* { 3298 syscallarg(int) fd; 3299 syscallarg(u_long) flags; 3300 } */ 3301 struct vnode *vp; 3302 file_t *fp; 3303 int error; 3304 3305 /* fd_getvnode() will use the descriptor for us */ 3306 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3307 return (error); 3308 vp = fp->f_vnode; 3309 error = change_flags(vp, SCARG(uap, flags), l); 3310 VOP_UNLOCK(vp); 3311 fd_putfile(SCARG(uap, fd)); 3312 return (error); 3313 } 3314 3315 /* 3316 * Change flags of a file given a path name; this version does 3317 * not follow links. 3318 */ 3319 int 3320 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3321 { 3322 /* { 3323 syscallarg(const char *) path; 3324 syscallarg(u_long) flags; 3325 } */ 3326 struct vnode *vp; 3327 int error; 3328 3329 error = namei_simple_user(SCARG(uap, path), 3330 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3331 if (error != 0) 3332 return (error); 3333 error = change_flags(vp, SCARG(uap, flags), l); 3334 vput(vp); 3335 return (error); 3336 } 3337 3338 /* 3339 * Common routine to change flags of a file. 3340 */ 3341 int 3342 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3343 { 3344 struct vattr vattr; 3345 int error; 3346 3347 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3348 3349 vattr_null(&vattr); 3350 vattr.va_flags = flags; 3351 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3352 3353 return (error); 3354 } 3355 3356 /* 3357 * Change mode of a file given path name; this version follows links. 3358 */ 3359 /* ARGSUSED */ 3360 int 3361 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3362 { 3363 /* { 3364 syscallarg(const char *) path; 3365 syscallarg(int) mode; 3366 } */ 3367 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3368 SCARG(uap, mode), 0); 3369 } 3370 3371 int 3372 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3373 { 3374 int error; 3375 struct vnode *vp; 3376 namei_simple_flags_t ns_flag; 3377 3378 if (flags & AT_SYMLINK_NOFOLLOW) 3379 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3380 else 3381 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3382 3383 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3384 if (error != 0) 3385 return error; 3386 3387 error = change_mode(vp, mode, l); 3388 3389 vrele(vp); 3390 3391 return (error); 3392 } 3393 3394 /* 3395 * Change mode of a file given a file descriptor. 3396 */ 3397 /* ARGSUSED */ 3398 int 3399 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3400 { 3401 /* { 3402 syscallarg(int) fd; 3403 syscallarg(int) mode; 3404 } */ 3405 file_t *fp; 3406 int error; 3407 3408 /* fd_getvnode() will use the descriptor for us */ 3409 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3410 return (error); 3411 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3412 fd_putfile(SCARG(uap, fd)); 3413 return (error); 3414 } 3415 3416 int 3417 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3418 register_t *retval) 3419 { 3420 /* { 3421 syscallarg(int) fd; 3422 syscallarg(const char *) path; 3423 syscallarg(int) mode; 3424 syscallarg(int) flag; 3425 } */ 3426 3427 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3428 SCARG(uap, mode), SCARG(uap, flag)); 3429 } 3430 3431 /* 3432 * Change mode of a file given path name; this version does not follow links. 3433 */ 3434 /* ARGSUSED */ 3435 int 3436 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3437 { 3438 /* { 3439 syscallarg(const char *) path; 3440 syscallarg(int) mode; 3441 } */ 3442 int error; 3443 struct vnode *vp; 3444 3445 error = namei_simple_user(SCARG(uap, path), 3446 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3447 if (error != 0) 3448 return (error); 3449 3450 error = change_mode(vp, SCARG(uap, mode), l); 3451 3452 vrele(vp); 3453 return (error); 3454 } 3455 3456 /* 3457 * Common routine to set mode given a vnode. 3458 */ 3459 static int 3460 change_mode(struct vnode *vp, int mode, struct lwp *l) 3461 { 3462 struct vattr vattr; 3463 int error; 3464 3465 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3466 vattr_null(&vattr); 3467 vattr.va_mode = mode & ALLPERMS; 3468 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3469 VOP_UNLOCK(vp); 3470 return (error); 3471 } 3472 3473 /* 3474 * Set ownership given a path name; this version follows links. 3475 */ 3476 /* ARGSUSED */ 3477 int 3478 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3479 { 3480 /* { 3481 syscallarg(const char *) path; 3482 syscallarg(uid_t) uid; 3483 syscallarg(gid_t) gid; 3484 } */ 3485 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3486 SCARG(uap, gid), 0); 3487 } 3488 3489 int 3490 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3491 gid_t gid, int flags) 3492 { 3493 int error; 3494 struct vnode *vp; 3495 namei_simple_flags_t ns_flag; 3496 3497 if (flags & AT_SYMLINK_NOFOLLOW) 3498 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3499 else 3500 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3501 3502 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3503 if (error != 0) 3504 return error; 3505 3506 error = change_owner(vp, uid, gid, l, 0); 3507 3508 vrele(vp); 3509 3510 return (error); 3511 } 3512 3513 /* 3514 * Set ownership given a path name; this version follows links. 3515 * Provides POSIX semantics. 3516 */ 3517 /* ARGSUSED */ 3518 int 3519 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3520 { 3521 /* { 3522 syscallarg(const char *) path; 3523 syscallarg(uid_t) uid; 3524 syscallarg(gid_t) gid; 3525 } */ 3526 int error; 3527 struct vnode *vp; 3528 3529 error = namei_simple_user(SCARG(uap, path), 3530 NSM_FOLLOW_TRYEMULROOT, &vp); 3531 if (error != 0) 3532 return (error); 3533 3534 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3535 3536 vrele(vp); 3537 return (error); 3538 } 3539 3540 /* 3541 * Set ownership given a file descriptor. 3542 */ 3543 /* ARGSUSED */ 3544 int 3545 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3546 { 3547 /* { 3548 syscallarg(int) fd; 3549 syscallarg(uid_t) uid; 3550 syscallarg(gid_t) gid; 3551 } */ 3552 int error; 3553 file_t *fp; 3554 3555 /* fd_getvnode() will use the descriptor for us */ 3556 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3557 return (error); 3558 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3559 l, 0); 3560 fd_putfile(SCARG(uap, fd)); 3561 return (error); 3562 } 3563 3564 int 3565 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3566 register_t *retval) 3567 { 3568 /* { 3569 syscallarg(int) fd; 3570 syscallarg(const char *) path; 3571 syscallarg(uid_t) owner; 3572 syscallarg(gid_t) group; 3573 syscallarg(int) flag; 3574 } */ 3575 3576 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3577 SCARG(uap, owner), SCARG(uap, group), 3578 SCARG(uap, flag)); 3579 } 3580 3581 /* 3582 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3583 */ 3584 /* ARGSUSED */ 3585 int 3586 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3587 { 3588 /* { 3589 syscallarg(int) fd; 3590 syscallarg(uid_t) uid; 3591 syscallarg(gid_t) gid; 3592 } */ 3593 int error; 3594 file_t *fp; 3595 3596 /* fd_getvnode() will use the descriptor for us */ 3597 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3598 return (error); 3599 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3600 l, 1); 3601 fd_putfile(SCARG(uap, fd)); 3602 return (error); 3603 } 3604 3605 /* 3606 * Set ownership given a path name; this version does not follow links. 3607 */ 3608 /* ARGSUSED */ 3609 int 3610 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3611 { 3612 /* { 3613 syscallarg(const char *) path; 3614 syscallarg(uid_t) uid; 3615 syscallarg(gid_t) gid; 3616 } */ 3617 int error; 3618 struct vnode *vp; 3619 3620 error = namei_simple_user(SCARG(uap, path), 3621 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3622 if (error != 0) 3623 return (error); 3624 3625 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3626 3627 vrele(vp); 3628 return (error); 3629 } 3630 3631 /* 3632 * Set ownership given a path name; this version does not follow links. 3633 * Provides POSIX/XPG semantics. 3634 */ 3635 /* ARGSUSED */ 3636 int 3637 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3638 { 3639 /* { 3640 syscallarg(const char *) path; 3641 syscallarg(uid_t) uid; 3642 syscallarg(gid_t) gid; 3643 } */ 3644 int error; 3645 struct vnode *vp; 3646 3647 error = namei_simple_user(SCARG(uap, path), 3648 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3649 if (error != 0) 3650 return (error); 3651 3652 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3653 3654 vrele(vp); 3655 return (error); 3656 } 3657 3658 /* 3659 * Common routine to set ownership given a vnode. 3660 */ 3661 static int 3662 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3663 int posix_semantics) 3664 { 3665 struct vattr vattr; 3666 mode_t newmode; 3667 int error; 3668 3669 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3670 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3671 goto out; 3672 3673 #define CHANGED(x) ((int)(x) != -1) 3674 newmode = vattr.va_mode; 3675 if (posix_semantics) { 3676 /* 3677 * POSIX/XPG semantics: if the caller is not the super-user, 3678 * clear set-user-id and set-group-id bits. Both POSIX and 3679 * the XPG consider the behaviour for calls by the super-user 3680 * implementation-defined; we leave the set-user-id and set- 3681 * group-id settings intact in that case. 3682 */ 3683 if (vattr.va_mode & S_ISUID) { 3684 if (kauth_authorize_vnode(l->l_cred, 3685 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3686 newmode &= ~S_ISUID; 3687 } 3688 if (vattr.va_mode & S_ISGID) { 3689 if (kauth_authorize_vnode(l->l_cred, 3690 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3691 newmode &= ~S_ISGID; 3692 } 3693 } else { 3694 /* 3695 * NetBSD semantics: when changing owner and/or group, 3696 * clear the respective bit(s). 3697 */ 3698 if (CHANGED(uid)) 3699 newmode &= ~S_ISUID; 3700 if (CHANGED(gid)) 3701 newmode &= ~S_ISGID; 3702 } 3703 /* Update va_mode iff altered. */ 3704 if (vattr.va_mode == newmode) 3705 newmode = VNOVAL; 3706 3707 vattr_null(&vattr); 3708 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3709 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3710 vattr.va_mode = newmode; 3711 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3712 #undef CHANGED 3713 3714 out: 3715 VOP_UNLOCK(vp); 3716 return (error); 3717 } 3718 3719 /* 3720 * Set the access and modification times given a path name; this 3721 * version follows links. 3722 */ 3723 /* ARGSUSED */ 3724 int 3725 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3726 register_t *retval) 3727 { 3728 /* { 3729 syscallarg(const char *) path; 3730 syscallarg(const struct timeval *) tptr; 3731 } */ 3732 3733 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3734 SCARG(uap, tptr), UIO_USERSPACE); 3735 } 3736 3737 /* 3738 * Set the access and modification times given a file descriptor. 3739 */ 3740 /* ARGSUSED */ 3741 int 3742 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3743 register_t *retval) 3744 { 3745 /* { 3746 syscallarg(int) fd; 3747 syscallarg(const struct timeval *) tptr; 3748 } */ 3749 int error; 3750 file_t *fp; 3751 3752 /* fd_getvnode() will use the descriptor for us */ 3753 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3754 return (error); 3755 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3756 UIO_USERSPACE); 3757 fd_putfile(SCARG(uap, fd)); 3758 return (error); 3759 } 3760 3761 int 3762 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3763 register_t *retval) 3764 { 3765 /* { 3766 syscallarg(int) fd; 3767 syscallarg(const struct timespec *) tptr; 3768 } */ 3769 int error; 3770 file_t *fp; 3771 3772 /* fd_getvnode() will use the descriptor for us */ 3773 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3774 return (error); 3775 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3776 SCARG(uap, tptr), UIO_USERSPACE); 3777 fd_putfile(SCARG(uap, fd)); 3778 return (error); 3779 } 3780 3781 /* 3782 * Set the access and modification times given a path name; this 3783 * version does not follow links. 3784 */ 3785 int 3786 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3787 register_t *retval) 3788 { 3789 /* { 3790 syscallarg(const char *) path; 3791 syscallarg(const struct timeval *) tptr; 3792 } */ 3793 3794 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3795 SCARG(uap, tptr), UIO_USERSPACE); 3796 } 3797 3798 int 3799 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3800 register_t *retval) 3801 { 3802 /* { 3803 syscallarg(int) fd; 3804 syscallarg(const char *) path; 3805 syscallarg(const struct timespec *) tptr; 3806 syscallarg(int) flag; 3807 } */ 3808 int follow; 3809 const struct timespec *tptr; 3810 int error; 3811 3812 tptr = SCARG(uap, tptr); 3813 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3814 3815 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3816 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3817 3818 return error; 3819 } 3820 3821 /* 3822 * Common routine to set access and modification times given a vnode. 3823 */ 3824 int 3825 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3826 const struct timespec *tptr, enum uio_seg seg) 3827 { 3828 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3829 } 3830 3831 int 3832 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3833 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3834 { 3835 struct vattr vattr; 3836 int error, dorele = 0; 3837 namei_simple_flags_t sflags; 3838 bool vanull, setbirthtime; 3839 struct timespec ts[2]; 3840 3841 KASSERT(l != NULL || fdat == AT_FDCWD); 3842 3843 /* 3844 * I have checked all callers and they pass either FOLLOW, 3845 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3846 * is 0. More to the point, they don't pass anything else. 3847 * Let's keep it that way at least until the namei interfaces 3848 * are fully sanitized. 3849 */ 3850 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3851 sflags = (flag == FOLLOW) ? 3852 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3853 3854 if (tptr == NULL) { 3855 vanull = true; 3856 nanotime(&ts[0]); 3857 ts[1] = ts[0]; 3858 } else { 3859 vanull = false; 3860 if (seg != UIO_SYSSPACE) { 3861 error = copyin(tptr, ts, sizeof (ts)); 3862 if (error != 0) 3863 return error; 3864 } else { 3865 ts[0] = tptr[0]; 3866 ts[1] = tptr[1]; 3867 } 3868 } 3869 3870 if (ts[0].tv_nsec == UTIME_NOW) { 3871 nanotime(&ts[0]); 3872 if (ts[1].tv_nsec == UTIME_NOW) { 3873 vanull = true; 3874 ts[1] = ts[0]; 3875 } 3876 } else if (ts[1].tv_nsec == UTIME_NOW) 3877 nanotime(&ts[1]); 3878 3879 if (vp == NULL) { 3880 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3881 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3882 if (error != 0) 3883 return error; 3884 dorele = 1; 3885 } 3886 3887 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3888 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3889 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3890 vattr_null(&vattr); 3891 3892 if (ts[0].tv_nsec != UTIME_OMIT) 3893 vattr.va_atime = ts[0]; 3894 3895 if (ts[1].tv_nsec != UTIME_OMIT) { 3896 vattr.va_mtime = ts[1]; 3897 if (setbirthtime) 3898 vattr.va_birthtime = ts[1]; 3899 } 3900 3901 if (vanull) 3902 vattr.va_vaflags |= VA_UTIMES_NULL; 3903 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3904 VOP_UNLOCK(vp); 3905 3906 if (dorele != 0) 3907 vrele(vp); 3908 3909 return error; 3910 } 3911 3912 int 3913 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3914 const struct timeval *tptr, enum uio_seg seg) 3915 { 3916 struct timespec ts[2]; 3917 struct timespec *tsptr = NULL; 3918 int error; 3919 3920 if (tptr != NULL) { 3921 struct timeval tv[2]; 3922 3923 if (seg != UIO_SYSSPACE) { 3924 error = copyin(tptr, tv, sizeof(tv)); 3925 if (error != 0) 3926 return error; 3927 tptr = tv; 3928 } 3929 3930 if ((tptr[0].tv_usec == UTIME_NOW) || 3931 (tptr[0].tv_usec == UTIME_OMIT)) 3932 ts[0].tv_nsec = tptr[0].tv_usec; 3933 else { 3934 if (tptr[0].tv_usec < 0 || tptr[0].tv_usec >= 1000000) 3935 return EINVAL; 3936 3937 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3938 } 3939 3940 if ((tptr[1].tv_usec == UTIME_NOW) || 3941 (tptr[1].tv_usec == UTIME_OMIT)) 3942 ts[1].tv_nsec = tptr[1].tv_usec; 3943 else { 3944 if (tptr[1].tv_usec < 0 || tptr[1].tv_usec >= 1000000) 3945 return EINVAL; 3946 3947 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3948 } 3949 3950 tsptr = &ts[0]; 3951 } 3952 3953 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3954 } 3955 3956 /* 3957 * Truncate a file given its path name. 3958 */ 3959 /* ARGSUSED */ 3960 int 3961 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3962 { 3963 /* { 3964 syscallarg(const char *) path; 3965 syscallarg(int) pad; 3966 syscallarg(off_t) length; 3967 } */ 3968 struct vnode *vp; 3969 struct vattr vattr; 3970 int error; 3971 3972 if (SCARG(uap, length) < 0) 3973 return EINVAL; 3974 3975 error = namei_simple_user(SCARG(uap, path), 3976 NSM_FOLLOW_TRYEMULROOT, &vp); 3977 if (error != 0) 3978 return (error); 3979 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3980 if (vp->v_type == VDIR) 3981 error = EISDIR; 3982 else if ((error = vn_writechk(vp)) == 0 && 3983 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3984 vattr_null(&vattr); 3985 vattr.va_size = SCARG(uap, length); 3986 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3987 } 3988 vput(vp); 3989 return (error); 3990 } 3991 3992 /* 3993 * Truncate a file given a file descriptor. 3994 */ 3995 /* ARGSUSED */ 3996 int 3997 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3998 { 3999 /* { 4000 syscallarg(int) fd; 4001 syscallarg(int) pad; 4002 syscallarg(off_t) length; 4003 } */ 4004 struct vattr vattr; 4005 struct vnode *vp; 4006 file_t *fp; 4007 int error; 4008 4009 if (SCARG(uap, length) < 0) 4010 return EINVAL; 4011 4012 /* fd_getvnode() will use the descriptor for us */ 4013 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4014 return (error); 4015 if ((fp->f_flag & FWRITE) == 0) { 4016 error = EINVAL; 4017 goto out; 4018 } 4019 vp = fp->f_vnode; 4020 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4021 if (vp->v_type == VDIR) 4022 error = EISDIR; 4023 else if ((error = vn_writechk(vp)) == 0) { 4024 vattr_null(&vattr); 4025 vattr.va_size = SCARG(uap, length); 4026 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 4027 } 4028 VOP_UNLOCK(vp); 4029 out: 4030 fd_putfile(SCARG(uap, fd)); 4031 return (error); 4032 } 4033 4034 /* 4035 * Sync an open file. 4036 */ 4037 /* ARGSUSED */ 4038 int 4039 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 4040 { 4041 /* { 4042 syscallarg(int) fd; 4043 } */ 4044 struct vnode *vp; 4045 file_t *fp; 4046 int error; 4047 4048 /* fd_getvnode() will use the descriptor for us */ 4049 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4050 return (error); 4051 vp = fp->f_vnode; 4052 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4053 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4054 VOP_UNLOCK(vp); 4055 fd_putfile(SCARG(uap, fd)); 4056 return (error); 4057 } 4058 4059 /* 4060 * Sync a range of file data. API modeled after that found in AIX. 4061 * 4062 * FDATASYNC indicates that we need only save enough metadata to be able 4063 * to re-read the written data. Note we duplicate AIX's requirement that 4064 * the file be open for writing. 4065 */ 4066 /* ARGSUSED */ 4067 int 4068 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4069 { 4070 /* { 4071 syscallarg(int) fd; 4072 syscallarg(int) flags; 4073 syscallarg(off_t) start; 4074 syscallarg(off_t) length; 4075 } */ 4076 struct vnode *vp; 4077 file_t *fp; 4078 int flags, nflags; 4079 off_t s, e, len; 4080 int error; 4081 4082 /* fd_getvnode() will use the descriptor for us */ 4083 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4084 return (error); 4085 4086 if ((fp->f_flag & FWRITE) == 0) { 4087 error = EBADF; 4088 goto out; 4089 } 4090 4091 flags = SCARG(uap, flags); 4092 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4093 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4094 error = EINVAL; 4095 goto out; 4096 } 4097 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4098 if (flags & FDATASYNC) 4099 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4100 else 4101 nflags = FSYNC_WAIT; 4102 if (flags & FDISKSYNC) 4103 nflags |= FSYNC_CACHE; 4104 4105 len = SCARG(uap, length); 4106 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4107 if (len) { 4108 s = SCARG(uap, start); 4109 e = s + len; 4110 if (e < s) { 4111 error = EINVAL; 4112 goto out; 4113 } 4114 } else { 4115 e = 0; 4116 s = 0; 4117 } 4118 4119 vp = fp->f_vnode; 4120 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4121 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4122 VOP_UNLOCK(vp); 4123 out: 4124 fd_putfile(SCARG(uap, fd)); 4125 return (error); 4126 } 4127 4128 /* 4129 * Sync the data of an open file. 4130 */ 4131 /* ARGSUSED */ 4132 int 4133 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4134 { 4135 /* { 4136 syscallarg(int) fd; 4137 } */ 4138 struct vnode *vp; 4139 file_t *fp; 4140 int error; 4141 4142 /* fd_getvnode() will use the descriptor for us */ 4143 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4144 return (error); 4145 if ((fp->f_flag & FWRITE) == 0) { 4146 fd_putfile(SCARG(uap, fd)); 4147 return (EBADF); 4148 } 4149 vp = fp->f_vnode; 4150 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4151 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4152 VOP_UNLOCK(vp); 4153 fd_putfile(SCARG(uap, fd)); 4154 return (error); 4155 } 4156 4157 /* 4158 * Rename files, (standard) BSD semantics frontend. 4159 */ 4160 /* ARGSUSED */ 4161 int 4162 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4163 { 4164 /* { 4165 syscallarg(const char *) from; 4166 syscallarg(const char *) to; 4167 } */ 4168 4169 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4170 SCARG(uap, to), UIO_USERSPACE, 0)); 4171 } 4172 4173 int 4174 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4175 register_t *retval) 4176 { 4177 /* { 4178 syscallarg(int) fromfd; 4179 syscallarg(const char *) from; 4180 syscallarg(int) tofd; 4181 syscallarg(const char *) to; 4182 } */ 4183 4184 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4185 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4186 } 4187 4188 /* 4189 * Rename files, POSIX semantics frontend. 4190 */ 4191 /* ARGSUSED */ 4192 int 4193 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4194 { 4195 /* { 4196 syscallarg(const char *) from; 4197 syscallarg(const char *) to; 4198 } */ 4199 4200 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4201 SCARG(uap, to), UIO_USERSPACE, 1)); 4202 } 4203 4204 /* 4205 * Rename files. Source and destination must either both be directories, 4206 * or both not be directories. If target is a directory, it must be empty. 4207 * If `from' and `to' refer to the same object, the value of the `retain' 4208 * argument is used to determine whether `from' will be 4209 * 4210 * (retain == 0) deleted unless `from' and `to' refer to the same 4211 * object in the file system's name space (BSD). 4212 * (retain == 1) always retained (POSIX). 4213 * 4214 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4215 */ 4216 int 4217 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4218 { 4219 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4220 } 4221 4222 static int 4223 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4224 const char *to, enum uio_seg seg, int retain) 4225 { 4226 struct pathbuf *fpb, *tpb; 4227 struct nameidata fnd, tnd; 4228 struct vnode *fdvp, *fvp; 4229 struct vnode *tdvp, *tvp; 4230 struct mount *mp, *tmp; 4231 int error; 4232 4233 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4234 4235 error = pathbuf_maybe_copyin(from, seg, &fpb); 4236 if (error) 4237 goto out0; 4238 KASSERT(fpb != NULL); 4239 4240 error = pathbuf_maybe_copyin(to, seg, &tpb); 4241 if (error) 4242 goto out1; 4243 KASSERT(tpb != NULL); 4244 4245 /* 4246 * Lookup from. 4247 * 4248 * XXX LOCKPARENT is wrong because we don't actually want it 4249 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4250 * insane, so for the time being we need to leave it like this. 4251 */ 4252 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4253 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4254 goto out2; 4255 4256 /* 4257 * Pull out the important results of the lookup, fdvp and fvp. 4258 * Of course, fvp is bogus because we're about to unlock fdvp. 4259 */ 4260 fdvp = fnd.ni_dvp; 4261 fvp = fnd.ni_vp; 4262 mp = fdvp->v_mount; 4263 KASSERT(fdvp != NULL); 4264 KASSERT(fvp != NULL); 4265 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4266 /* 4267 * Bracket the operation with fstrans_start()/fstrans_done(). 4268 * 4269 * Inside the bracket this file system cannot be unmounted so 4270 * a vnode on this file system cannot change its v_mount. 4271 * A vnode on another file system may still change to dead mount. 4272 */ 4273 fstrans_start(mp); 4274 4275 /* 4276 * Make sure neither fdvp nor fvp is locked. 4277 */ 4278 if (fdvp != fvp) 4279 VOP_UNLOCK(fdvp); 4280 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4281 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4282 4283 /* 4284 * Reject renaming `.' and `..'. Can't do this until after 4285 * namei because we need namei's parsing to find the final 4286 * component name. (namei should just leave us with the final 4287 * component name and not look it up itself, but anyway...) 4288 * 4289 * This was here before because we used to relookup from 4290 * instead of to and relookup requires the caller to check 4291 * this, but now file systems may depend on this check, so we 4292 * must retain it until the file systems are all rototilled. 4293 */ 4294 if (((fnd.ni_cnd.cn_namelen == 1) && 4295 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4296 ((fnd.ni_cnd.cn_namelen == 2) && 4297 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4298 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4299 error = EINVAL; /* XXX EISDIR? */ 4300 goto abort0; 4301 } 4302 4303 /* 4304 * Lookup to. 4305 * 4306 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4307 * fvp here to decide whether to add CREATEDIR is a load of 4308 * bollocks because fvp might be the wrong node by now, since 4309 * fdvp is unlocked. 4310 * 4311 * XXX Why not pass CREATEDIR always? 4312 */ 4313 NDINIT(&tnd, RENAME, 4314 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4315 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4316 tpb); 4317 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4318 goto abort0; 4319 4320 /* 4321 * Pull out the important results of the lookup, tdvp and tvp. 4322 * Of course, tvp is bogus because we're about to unlock tdvp. 4323 */ 4324 tdvp = tnd.ni_dvp; 4325 tvp = tnd.ni_vp; 4326 KASSERT(tdvp != NULL); 4327 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4328 4329 /* 4330 * Make sure neither tdvp nor tvp is locked. 4331 */ 4332 if (tdvp != tvp) 4333 VOP_UNLOCK(tdvp); 4334 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4335 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4336 4337 /* 4338 * Reject renaming onto `.' or `..'. relookup is unhappy with 4339 * these, which is why we must do this here. Once upon a time 4340 * we relooked up from instead of to, and consequently didn't 4341 * need this check, but now that we relookup to instead of 4342 * from, we need this; and we shall need it forever forward 4343 * until the VOP_RENAME protocol changes, because file systems 4344 * will no doubt begin to depend on this check. 4345 */ 4346 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4347 error = EISDIR; 4348 goto abort1; 4349 } 4350 if ((tnd.ni_cnd.cn_namelen == 2) && 4351 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4352 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4353 error = EINVAL; 4354 goto abort1; 4355 } 4356 4357 /* 4358 * Make sure the mount points match. Although we don't hold 4359 * any vnode locks, the v_mount on fdvp file system are stable. 4360 * 4361 * Unmounting another file system at an inopportune moment may 4362 * cause tdvp to disappear and change its v_mount to dead. 4363 * 4364 * So in either case different v_mount means cross-device rename. 4365 */ 4366 KASSERT(mp != NULL); 4367 tmp = tdvp->v_mount; 4368 4369 if (mp != tmp) { 4370 error = EXDEV; 4371 goto abort1; 4372 } 4373 4374 /* 4375 * Take the vfs rename lock to avoid cross-directory screw cases. 4376 * Nothing is locked currently, so taking this lock is safe. 4377 */ 4378 error = VFS_RENAMELOCK_ENTER(mp); 4379 if (error) 4380 goto abort1; 4381 4382 /* 4383 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4384 * and nothing is locked except for the vfs rename lock. 4385 * 4386 * The next step is a little rain dance to conform to the 4387 * insane lock protocol, even though it does nothing to ward 4388 * off race conditions. 4389 * 4390 * We need tdvp and tvp to be locked. However, because we have 4391 * unlocked tdvp in order to hold no locks while we take the 4392 * vfs rename lock, tvp may be wrong here, and we can't safely 4393 * lock it even if the sensible file systems will just unlock 4394 * it straight away. Consequently, we must lock tdvp and then 4395 * relookup tvp to get it locked. 4396 * 4397 * Finally, because the VOP_RENAME protocol is brain-damaged 4398 * and various file systems insanely depend on the semantics of 4399 * this brain damage, the lookup of to must be the last lookup 4400 * before VOP_RENAME. 4401 */ 4402 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4403 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4404 if (error) 4405 goto abort2; 4406 4407 /* 4408 * Drop the old tvp and pick up the new one -- which might be 4409 * the same, but that doesn't matter to us. After this, tdvp 4410 * and tvp should both be locked. 4411 */ 4412 if (tvp != NULL) 4413 vrele(tvp); 4414 tvp = tnd.ni_vp; 4415 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4416 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4417 4418 /* 4419 * The old do_sys_rename had various consistency checks here 4420 * involving fvp and tvp. fvp is bogus already here, and tvp 4421 * will become bogus soon in any sensible file system, so the 4422 * only purpose in putting these checks here is to give lip 4423 * service to these screw cases and to acknowledge that they 4424 * exist, not actually to handle them, but here you go 4425 * anyway... 4426 */ 4427 4428 /* 4429 * Acknowledge that directories and non-directories aren't 4430 * suposed to mix. 4431 */ 4432 if (tvp != NULL) { 4433 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4434 error = ENOTDIR; 4435 goto abort3; 4436 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4437 error = EISDIR; 4438 goto abort3; 4439 } 4440 } 4441 4442 /* 4443 * Acknowledge some random screw case, among the dozens that 4444 * might arise. 4445 */ 4446 if (fvp == tdvp) { 4447 error = EINVAL; 4448 goto abort3; 4449 } 4450 4451 /* 4452 * Acknowledge that POSIX has a wacky screw case. 4453 * 4454 * XXX Eventually the retain flag needs to be passed on to 4455 * VOP_RENAME. 4456 */ 4457 if (fvp == tvp) { 4458 if (retain) { 4459 error = 0; 4460 goto abort3; 4461 } else if ((fdvp == tdvp) && 4462 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4463 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4464 fnd.ni_cnd.cn_namelen))) { 4465 error = 0; 4466 goto abort3; 4467 } 4468 } 4469 4470 /* 4471 * Make sure veriexec can screw us up. (But a race can screw 4472 * up veriexec, of course -- remember, fvp and (soon) tvp are 4473 * bogus.) 4474 */ 4475 #if NVERIEXEC > 0 4476 { 4477 char *f1, *f2; 4478 size_t f1_len; 4479 size_t f2_len; 4480 4481 f1_len = fnd.ni_cnd.cn_namelen + 1; 4482 f1 = kmem_alloc(f1_len, KM_SLEEP); 4483 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4484 4485 f2_len = tnd.ni_cnd.cn_namelen + 1; 4486 f2 = kmem_alloc(f2_len, KM_SLEEP); 4487 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4488 4489 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4490 4491 kmem_free(f1, f1_len); 4492 kmem_free(f2, f2_len); 4493 4494 if (error) 4495 goto abort3; 4496 } 4497 #endif /* NVERIEXEC > 0 */ 4498 4499 /* 4500 * All ready. Incant the rename vop. 4501 */ 4502 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4503 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4504 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4505 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4506 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4507 4508 /* 4509 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4510 * tdvp and tvp. But we can't assert any of that. 4511 */ 4512 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4513 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4514 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4515 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4516 4517 /* 4518 * So all we have left to do is to drop the rename lock and 4519 * destroy the pathbufs. 4520 */ 4521 VFS_RENAMELOCK_EXIT(mp); 4522 fstrans_done(mp); 4523 goto out2; 4524 4525 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4526 VOP_UNLOCK(tvp); 4527 abort2: VOP_UNLOCK(tdvp); 4528 VFS_RENAMELOCK_EXIT(mp); 4529 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4530 vrele(tdvp); 4531 if (tvp != NULL) 4532 vrele(tvp); 4533 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4534 vrele(fdvp); 4535 vrele(fvp); 4536 fstrans_done(mp); 4537 out2: pathbuf_destroy(tpb); 4538 out1: pathbuf_destroy(fpb); 4539 out0: return error; 4540 } 4541 4542 /* 4543 * Make a directory file. 4544 */ 4545 /* ARGSUSED */ 4546 int 4547 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4548 { 4549 /* { 4550 syscallarg(const char *) path; 4551 syscallarg(int) mode; 4552 } */ 4553 4554 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4555 SCARG(uap, mode), UIO_USERSPACE); 4556 } 4557 4558 int 4559 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4560 register_t *retval) 4561 { 4562 /* { 4563 syscallarg(int) fd; 4564 syscallarg(const char *) path; 4565 syscallarg(int) mode; 4566 } */ 4567 4568 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4569 SCARG(uap, mode), UIO_USERSPACE); 4570 } 4571 4572 4573 int 4574 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4575 { 4576 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, seg); 4577 } 4578 4579 static int 4580 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4581 enum uio_seg seg) 4582 { 4583 struct proc *p = curlwp->l_proc; 4584 struct vnode *vp; 4585 struct vattr vattr; 4586 int error; 4587 struct pathbuf *pb; 4588 struct nameidata nd; 4589 4590 KASSERT(l != NULL || fdat == AT_FDCWD); 4591 4592 /* XXX bollocks, should pass in a pathbuf */ 4593 error = pathbuf_maybe_copyin(path, seg, &pb); 4594 if (error) { 4595 return error; 4596 } 4597 4598 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4599 4600 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4601 pathbuf_destroy(pb); 4602 return (error); 4603 } 4604 vp = nd.ni_vp; 4605 if (vp != NULL) { 4606 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4607 if (nd.ni_dvp == vp) 4608 vrele(nd.ni_dvp); 4609 else 4610 vput(nd.ni_dvp); 4611 vrele(vp); 4612 pathbuf_destroy(pb); 4613 return (EEXIST); 4614 } 4615 vattr_null(&vattr); 4616 vattr.va_type = VDIR; 4617 /* We will read cwdi->cwdi_cmask unlocked. */ 4618 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4619 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4620 if (!error) 4621 vrele(nd.ni_vp); 4622 vput(nd.ni_dvp); 4623 pathbuf_destroy(pb); 4624 return (error); 4625 } 4626 4627 /* 4628 * Remove a directory file. 4629 */ 4630 /* ARGSUSED */ 4631 int 4632 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4633 { 4634 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4635 AT_REMOVEDIR, UIO_USERSPACE); 4636 } 4637 4638 /* 4639 * Read a block of directory entries in a file system independent format. 4640 */ 4641 int 4642 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4643 { 4644 /* { 4645 syscallarg(int) fd; 4646 syscallarg(char *) buf; 4647 syscallarg(size_t) count; 4648 } */ 4649 file_t *fp; 4650 int error, done; 4651 4652 /* fd_getvnode() will use the descriptor for us */ 4653 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4654 return (error); 4655 if ((fp->f_flag & FREAD) == 0) { 4656 error = EBADF; 4657 goto out; 4658 } 4659 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4660 SCARG(uap, count), &done, l, 0, 0); 4661 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4662 *retval = done; 4663 out: 4664 fd_putfile(SCARG(uap, fd)); 4665 return (error); 4666 } 4667 4668 /* 4669 * Set the mode mask for creation of filesystem nodes. 4670 */ 4671 int 4672 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4673 { 4674 /* { 4675 syscallarg(mode_t) newmask; 4676 } */ 4677 struct proc *p = l->l_proc; 4678 struct cwdinfo *cwdi; 4679 4680 /* 4681 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4682 * important is that we serialize changes to the mask. The 4683 * rw_exit() will issue a write memory barrier on our behalf, 4684 * and force the changes out to other CPUs (as it must use an 4685 * atomic operation, draining the local CPU's store buffers). 4686 */ 4687 cwdi = p->p_cwdi; 4688 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4689 *retval = cwdi->cwdi_cmask; 4690 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4691 rw_exit(&cwdi->cwdi_lock); 4692 4693 return (0); 4694 } 4695 4696 int 4697 dorevoke(struct vnode *vp, kauth_cred_t cred) 4698 { 4699 struct vattr vattr; 4700 int error, fs_decision; 4701 4702 vn_lock(vp, LK_SHARED | LK_RETRY); 4703 error = VOP_GETATTR(vp, &vattr, cred); 4704 VOP_UNLOCK(vp); 4705 if (error != 0) 4706 return error; 4707 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4708 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4709 fs_decision); 4710 if (!error) 4711 VOP_REVOKE(vp, REVOKEALL); 4712 return (error); 4713 } 4714 4715 /* 4716 * Void all references to file by ripping underlying filesystem 4717 * away from vnode. 4718 */ 4719 /* ARGSUSED */ 4720 int 4721 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4722 { 4723 /* { 4724 syscallarg(const char *) path; 4725 } */ 4726 struct vnode *vp; 4727 int error; 4728 4729 error = namei_simple_user(SCARG(uap, path), 4730 NSM_FOLLOW_TRYEMULROOT, &vp); 4731 if (error != 0) 4732 return (error); 4733 error = dorevoke(vp, l->l_cred); 4734 vrele(vp); 4735 return (error); 4736 } 4737 4738 /* 4739 * Allocate backing store for a file, filling a hole without having to 4740 * explicitly write anything out. 4741 */ 4742 /* ARGSUSED */ 4743 int 4744 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4745 register_t *retval) 4746 { 4747 /* { 4748 syscallarg(int) fd; 4749 syscallarg(off_t) pos; 4750 syscallarg(off_t) len; 4751 } */ 4752 int fd; 4753 off_t pos, len; 4754 struct file *fp; 4755 struct vnode *vp; 4756 int error; 4757 4758 fd = SCARG(uap, fd); 4759 pos = SCARG(uap, pos); 4760 len = SCARG(uap, len); 4761 4762 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4763 *retval = EINVAL; 4764 return 0; 4765 } 4766 4767 error = fd_getvnode(fd, &fp); 4768 if (error) { 4769 *retval = error; 4770 return 0; 4771 } 4772 if ((fp->f_flag & FWRITE) == 0) { 4773 error = EBADF; 4774 goto fail; 4775 } 4776 vp = fp->f_vnode; 4777 4778 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4779 if (vp->v_type == VDIR) { 4780 error = EISDIR; 4781 } else { 4782 error = VOP_FALLOCATE(vp, pos, len); 4783 } 4784 VOP_UNLOCK(vp); 4785 4786 fail: 4787 fd_putfile(fd); 4788 *retval = error; 4789 return 0; 4790 } 4791 4792 /* 4793 * Deallocate backing store for a file, creating a hole. Also used for 4794 * invoking TRIM on disks. 4795 */ 4796 /* ARGSUSED */ 4797 int 4798 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4799 register_t *retval) 4800 { 4801 /* { 4802 syscallarg(int) fd; 4803 syscallarg(off_t) pos; 4804 syscallarg(off_t) len; 4805 } */ 4806 int fd; 4807 off_t pos, len; 4808 struct file *fp; 4809 struct vnode *vp; 4810 int error; 4811 4812 fd = SCARG(uap, fd); 4813 pos = SCARG(uap, pos); 4814 len = SCARG(uap, len); 4815 4816 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4817 return EINVAL; 4818 } 4819 4820 error = fd_getvnode(fd, &fp); 4821 if (error) { 4822 return error; 4823 } 4824 if ((fp->f_flag & FWRITE) == 0) { 4825 error = EBADF; 4826 goto fail; 4827 } 4828 vp = fp->f_vnode; 4829 4830 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4831 if (vp->v_type == VDIR) { 4832 error = EISDIR; 4833 } else { 4834 error = VOP_FDISCARD(vp, pos, len); 4835 } 4836 VOP_UNLOCK(vp); 4837 4838 fail: 4839 fd_putfile(fd); 4840 return error; 4841 } 4842
Indexes created Thu Sep 25 06:09:39 GMT 2025