1 /* $NetBSD: vfs_syscalls.c,v 1.431 2011/07/03 15:25:09 hannken Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 /* 69 * Virtual File System System Calls 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.431 2011/07/03 15:25:09 hannken Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/stat.h> 87 #include <sys/vnode.h> 88 #include <sys/mount.h> 89 #include <sys/proc.h> 90 #include <sys/uio.h> 91 #include <sys/kmem.h> 92 #include <sys/dirent.h> 93 #include <sys/sysctl.h> 94 #include <sys/syscallargs.h> 95 #include <sys/vfs_syscalls.h> 96 #include <sys/ktrace.h> 97 #ifdef FILEASSOC 98 #include <sys/fileassoc.h> 99 #endif /* FILEASSOC */ 100 #include <sys/extattr.h> 101 #include <sys/verified_exec.h> 102 #include <sys/kauth.h> 103 #include <sys/atomic.h> 104 #include <sys/module.h> 105 #include <sys/buf.h> 106 107 #include <miscfs/genfs/genfs.h> 108 #include <miscfs/syncfs/syncfs.h> 109 #include <miscfs/specfs/specdev.h> 110 111 #include <nfs/rpcv2.h> 112 #include <nfs/nfsproto.h> 113 #include <nfs/nfs.h> 114 #include <nfs/nfs_var.h> 115 116 static int change_flags(struct vnode *, u_long, struct lwp *); 117 static int change_mode(struct vnode *, int, struct lwp *l); 118 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 119 120 /* 121 * This table is used to maintain compatibility with 4.3BSD 122 * and NetBSD 0.9 mount syscalls - and possibly other systems. 123 * Note, the order is important! 124 * 125 * Do not modify this table. It should only contain filesystems 126 * supported by NetBSD 0.9 and 4.3BSD. 127 */ 128 const char * const mountcompatnames[] = { 129 NULL, /* 0 = MOUNT_NONE */ 130 MOUNT_FFS, /* 1 = MOUNT_UFS */ 131 MOUNT_NFS, /* 2 */ 132 MOUNT_MFS, /* 3 */ 133 MOUNT_MSDOS, /* 4 */ 134 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 135 MOUNT_FDESC, /* 6 */ 136 MOUNT_KERNFS, /* 7 */ 137 NULL, /* 8 = MOUNT_DEVFS */ 138 MOUNT_AFS, /* 9 */ 139 }; 140 141 const int nmountcompatnames = __arraycount(mountcompatnames); 142 143 static int 144 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 145 { 146 int error; 147 148 fp->f_flag = flags & FMASK; 149 fp->f_type = DTYPE_VNODE; 150 fp->f_ops = &vnops; 151 fp->f_data = vp; 152 153 if (flags & (O_EXLOCK | O_SHLOCK)) { 154 struct flock lf; 155 int type; 156 157 lf.l_whence = SEEK_SET; 158 lf.l_start = 0; 159 lf.l_len = 0; 160 if (flags & O_EXLOCK) 161 lf.l_type = F_WRLCK; 162 else 163 lf.l_type = F_RDLCK; 164 type = F_FLOCK; 165 if ((flags & FNONBLOCK) == 0) 166 type |= F_WAIT; 167 VOP_UNLOCK(vp); 168 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 169 if (error) { 170 (void) vn_close(vp, fp->f_flag, fp->f_cred); 171 fd_abort(l->l_proc, fp, indx); 172 return error; 173 } 174 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 175 atomic_or_uint(&fp->f_flag, FHASLOCK); 176 } 177 if (flags & O_CLOEXEC) 178 fd_set_exclose(l, indx, true); 179 return 0; 180 } 181 182 static int 183 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 184 void *data, size_t *data_len) 185 { 186 struct mount *mp; 187 int error = 0, saved_flags; 188 189 mp = vp->v_mount; 190 saved_flags = mp->mnt_flag; 191 192 /* We can operate only on VV_ROOT nodes. */ 193 if ((vp->v_vflag & VV_ROOT) == 0) { 194 error = EINVAL; 195 goto out; 196 } 197 198 /* 199 * We only allow the filesystem to be reloaded if it 200 * is currently mounted read-only. Additionally, we 201 * prevent read-write to read-only downgrades. 202 */ 203 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 204 (mp->mnt_flag & MNT_RDONLY) == 0 && 205 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 206 error = EOPNOTSUPP; /* Needs translation */ 207 goto out; 208 } 209 210 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 211 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 212 if (error) 213 goto out; 214 215 if (vfs_busy(mp, NULL)) { 216 error = EPERM; 217 goto out; 218 } 219 220 mutex_enter(&mp->mnt_updating); 221 222 mp->mnt_flag &= ~MNT_OP_FLAGS; 223 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 224 225 /* 226 * Set the mount level flags. 227 */ 228 if (flags & MNT_RDONLY) 229 mp->mnt_flag |= MNT_RDONLY; 230 else if (mp->mnt_flag & MNT_RDONLY) 231 mp->mnt_iflag |= IMNT_WANTRDWR; 232 mp->mnt_flag &= 233 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 234 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 235 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 236 MNT_LOG | MNT_EXTATTR); 237 mp->mnt_flag |= flags & 238 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 239 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 240 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 241 MNT_LOG | MNT_EXTATTR | MNT_IGNORE); 242 243 error = VFS_MOUNT(mp, path, data, data_len); 244 245 if (error && data != NULL) { 246 int error2; 247 248 /* 249 * Update failed; let's try and see if it was an 250 * export request. For compat with 3.0 and earlier. 251 */ 252 error2 = vfs_hooks_reexport(mp, path, data); 253 254 /* 255 * Only update error code if the export request was 256 * understood but some problem occurred while 257 * processing it. 258 */ 259 if (error2 != EJUSTRETURN) 260 error = error2; 261 } 262 263 if (mp->mnt_iflag & IMNT_WANTRDWR) 264 mp->mnt_flag &= ~MNT_RDONLY; 265 if (error) 266 mp->mnt_flag = saved_flags; 267 mp->mnt_flag &= ~MNT_OP_FLAGS; 268 mp->mnt_iflag &= ~IMNT_WANTRDWR; 269 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 270 if (mp->mnt_syncer == NULL) 271 error = vfs_allocate_syncvnode(mp); 272 } else { 273 if (mp->mnt_syncer != NULL) 274 vfs_deallocate_syncvnode(mp); 275 } 276 mutex_exit(&mp->mnt_updating); 277 vfs_unbusy(mp, false, NULL); 278 279 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 280 (flags & MNT_EXTATTR)) { 281 if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_START, 282 NULL, 0, NULL) != 0) { 283 printf("%s: failed to start extattr, error = %d", 284 vp->v_mount->mnt_stat.f_mntonname, error); 285 mp->mnt_flag &= ~MNT_EXTATTR; 286 } 287 } 288 289 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 290 !(flags & MNT_EXTATTR)) { 291 if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_STOP, 292 NULL, 0, NULL) != 0) { 293 printf("%s: failed to stop extattr, error = %d", 294 vp->v_mount->mnt_stat.f_mntonname, error); 295 mp->mnt_flag |= MNT_RDONLY; 296 } 297 } 298 out: 299 return (error); 300 } 301 302 static int 303 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 304 { 305 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 306 int error; 307 308 /* Copy file-system type from userspace. */ 309 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 310 if (error) { 311 /* 312 * Historically, filesystem types were identified by numbers. 313 * If we get an integer for the filesystem type instead of a 314 * string, we check to see if it matches one of the historic 315 * filesystem types. 316 */ 317 u_long fsindex = (u_long)fstype; 318 if (fsindex >= nmountcompatnames || 319 mountcompatnames[fsindex] == NULL) 320 return ENODEV; 321 strlcpy(fstypename, mountcompatnames[fsindex], 322 sizeof(fstypename)); 323 } 324 325 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 326 if (strcmp(fstypename, "ufs") == 0) 327 fstypename[0] = 'f'; 328 329 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 330 return 0; 331 332 /* If we can autoload a vfs module, try again */ 333 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 334 335 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 336 return 0; 337 338 return ENODEV; 339 } 340 341 static int 342 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 343 void *data, size_t *data_len) 344 { 345 struct mount *mp; 346 int error; 347 348 /* If MNT_GETARGS is specified, it should be the only flag. */ 349 if (flags & ~MNT_GETARGS) 350 return EINVAL; 351 352 mp = vp->v_mount; 353 354 /* XXX: probably some notion of "can see" here if we want isolation. */ 355 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 356 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 357 if (error) 358 return error; 359 360 if ((vp->v_vflag & VV_ROOT) == 0) 361 return EINVAL; 362 363 if (vfs_busy(mp, NULL)) 364 return EPERM; 365 366 mutex_enter(&mp->mnt_updating); 367 mp->mnt_flag &= ~MNT_OP_FLAGS; 368 mp->mnt_flag |= MNT_GETARGS; 369 error = VFS_MOUNT(mp, path, data, data_len); 370 mp->mnt_flag &= ~MNT_OP_FLAGS; 371 mutex_exit(&mp->mnt_updating); 372 373 vfs_unbusy(mp, false, NULL); 374 return (error); 375 } 376 377 int 378 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 379 { 380 /* { 381 syscallarg(const char *) type; 382 syscallarg(const char *) path; 383 syscallarg(int) flags; 384 syscallarg(void *) data; 385 syscallarg(size_t) data_len; 386 } */ 387 388 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 389 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 390 SCARG(uap, data_len), retval); 391 } 392 393 int 394 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 395 const char *path, int flags, void *data, enum uio_seg data_seg, 396 size_t data_len, register_t *retval) 397 { 398 struct vnode *vp; 399 void *data_buf = data; 400 bool vfsopsrele = false; 401 int error; 402 403 /* XXX: The calling convention of this routine is totally bizarre */ 404 if (vfsops) 405 vfsopsrele = true; 406 407 /* 408 * Get vnode to be covered 409 */ 410 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 411 if (error != 0) { 412 vp = NULL; 413 goto done; 414 } 415 416 if (vfsops == NULL) { 417 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 418 vfsops = vp->v_mount->mnt_op; 419 } else { 420 /* 'type' is userspace */ 421 error = mount_get_vfsops(type, &vfsops); 422 if (error != 0) 423 goto done; 424 vfsopsrele = true; 425 } 426 } 427 428 if (data != NULL && data_seg == UIO_USERSPACE) { 429 if (data_len == 0) { 430 /* No length supplied, use default for filesystem */ 431 data_len = vfsops->vfs_min_mount_data; 432 if (data_len > VFS_MAX_MOUNT_DATA) { 433 error = EINVAL; 434 goto done; 435 } 436 /* 437 * Hopefully a longer buffer won't make copyin() fail. 438 * For compatibility with 3.0 and earlier. 439 */ 440 if (flags & MNT_UPDATE 441 && data_len < sizeof (struct mnt_export_args30)) 442 data_len = sizeof (struct mnt_export_args30); 443 } 444 data_buf = kmem_alloc(data_len, KM_SLEEP); 445 446 /* NFS needs the buffer even for mnt_getargs .... */ 447 error = copyin(data, data_buf, data_len); 448 if (error != 0) 449 goto done; 450 } 451 452 if (flags & MNT_GETARGS) { 453 if (data_len == 0) { 454 error = EINVAL; 455 goto done; 456 } 457 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 458 if (error != 0) 459 goto done; 460 if (data_seg == UIO_USERSPACE) 461 error = copyout(data_buf, data, data_len); 462 *retval = data_len; 463 } else if (flags & MNT_UPDATE) { 464 error = mount_update(l, vp, path, flags, data_buf, &data_len); 465 } else { 466 /* Locking is handled internally in mount_domount(). */ 467 KASSERT(vfsopsrele == true); 468 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 469 &data_len); 470 vfsopsrele = false; 471 472 if ((error == 0) && (flags & MNT_EXTATTR)) { 473 if (VFS_EXTATTRCTL(vp->v_mount, EXTATTR_CMD_START, 474 NULL, 0, NULL) != 0) 475 printf("%s: failed to start extattr", 476 vp->v_mount->mnt_stat.f_mntonname); 477 /* XXX remove flag */ 478 } 479 } 480 481 done: 482 if (vfsopsrele) 483 vfs_delref(vfsops); 484 if (vp != NULL) { 485 vrele(vp); 486 } 487 if (data_buf != data) 488 kmem_free(data_buf, data_len); 489 return (error); 490 } 491 492 /* 493 * Unmount a file system. 494 * 495 * Note: unmount takes a path to the vnode mounted on as argument, 496 * not special file (as before). 497 */ 498 /* ARGSUSED */ 499 int 500 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 501 { 502 /* { 503 syscallarg(const char *) path; 504 syscallarg(int) flags; 505 } */ 506 struct vnode *vp; 507 struct mount *mp; 508 int error; 509 struct pathbuf *pb; 510 struct nameidata nd; 511 512 error = pathbuf_copyin(SCARG(uap, path), &pb); 513 if (error) { 514 return error; 515 } 516 517 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 518 if ((error = namei(&nd)) != 0) { 519 pathbuf_destroy(pb); 520 return error; 521 } 522 vp = nd.ni_vp; 523 pathbuf_destroy(pb); 524 525 mp = vp->v_mount; 526 atomic_inc_uint(&mp->mnt_refcnt); 527 VOP_UNLOCK(vp); 528 529 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 530 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 531 if (error) { 532 vrele(vp); 533 vfs_destroy(mp); 534 return (error); 535 } 536 537 /* 538 * Don't allow unmounting the root file system. 539 */ 540 if (mp->mnt_flag & MNT_ROOTFS) { 541 vrele(vp); 542 vfs_destroy(mp); 543 return (EINVAL); 544 } 545 546 /* 547 * Must be the root of the filesystem 548 */ 549 if ((vp->v_vflag & VV_ROOT) == 0) { 550 vrele(vp); 551 vfs_destroy(mp); 552 return (EINVAL); 553 } 554 555 vrele(vp); 556 error = dounmount(mp, SCARG(uap, flags), l); 557 vfs_destroy(mp); 558 return error; 559 } 560 561 /* 562 * Sync each mounted filesystem. 563 */ 564 #ifdef DEBUG 565 int syncprt = 0; 566 struct ctldebug debug0 = { "syncprt", &syncprt }; 567 #endif 568 569 void 570 do_sys_sync(struct lwp *l) 571 { 572 struct mount *mp, *nmp; 573 int asyncflag; 574 575 mutex_enter(&mountlist_lock); 576 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 577 mp = nmp) { 578 if (vfs_busy(mp, &nmp)) { 579 continue; 580 } 581 mutex_enter(&mp->mnt_updating); 582 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 583 asyncflag = mp->mnt_flag & MNT_ASYNC; 584 mp->mnt_flag &= ~MNT_ASYNC; 585 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 586 if (asyncflag) 587 mp->mnt_flag |= MNT_ASYNC; 588 } 589 mutex_exit(&mp->mnt_updating); 590 vfs_unbusy(mp, false, &nmp); 591 } 592 mutex_exit(&mountlist_lock); 593 #ifdef DEBUG 594 if (syncprt) 595 vfs_bufstats(); 596 #endif /* DEBUG */ 597 } 598 599 /* ARGSUSED */ 600 int 601 sys_sync(struct lwp *l, const void *v, register_t *retval) 602 { 603 do_sys_sync(l); 604 return (0); 605 } 606 607 608 /* 609 * Change filesystem quotas. 610 */ 611 /* ARGSUSED */ 612 int 613 sys___quotactl50(struct lwp *l, const struct sys___quotactl50_args *uap, 614 register_t *retval) 615 { 616 /* { 617 syscallarg(const char *) path; 618 syscallarg(struct plistref *) pref; 619 } */ 620 struct mount *mp; 621 int error; 622 struct vnode *vp; 623 prop_dictionary_t dict; 624 struct plistref pref; 625 626 error = namei_simple_user(SCARG(uap, path), 627 NSM_FOLLOW_TRYEMULROOT, &vp); 628 if (error != 0) 629 return (error); 630 mp = vp->v_mount; 631 error = copyin(SCARG(uap, pref), &pref, sizeof(pref)); 632 if (error) 633 return error; 634 error = prop_dictionary_copyin(&pref, &dict); 635 if (error) 636 return error; 637 error = VFS_QUOTACTL(mp, dict); 638 vrele(vp); 639 if (!error) 640 error = prop_dictionary_copyout(&pref, dict); 641 if (!error) 642 error = copyout(&pref, SCARG(uap, pref), sizeof(pref)); 643 prop_object_release(dict); 644 return (error); 645 } 646 647 int 648 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 649 int root) 650 { 651 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 652 int error = 0; 653 654 /* 655 * If MNT_NOWAIT or MNT_LAZY is specified, do not 656 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 657 * overrides MNT_NOWAIT. 658 */ 659 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 660 (flags != MNT_WAIT && flags != 0)) { 661 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 662 goto done; 663 } 664 665 /* Get the filesystem stats now */ 666 memset(sp, 0, sizeof(*sp)); 667 if ((error = VFS_STATVFS(mp, sp)) != 0) { 668 return error; 669 } 670 671 if (cwdi->cwdi_rdir == NULL) 672 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 673 done: 674 if (cwdi->cwdi_rdir != NULL) { 675 size_t len; 676 char *bp; 677 char c; 678 char *path = PNBUF_GET(); 679 680 bp = path + MAXPATHLEN; 681 *--bp = '\0'; 682 rw_enter(&cwdi->cwdi_lock, RW_READER); 683 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 684 MAXPATHLEN / 2, 0, l); 685 rw_exit(&cwdi->cwdi_lock); 686 if (error) { 687 PNBUF_PUT(path); 688 return error; 689 } 690 len = strlen(bp); 691 if (len != 1) { 692 /* 693 * for mount points that are below our root, we can see 694 * them, so we fix up the pathname and return them. The 695 * rest we cannot see, so we don't allow viewing the 696 * data. 697 */ 698 if (strncmp(bp, sp->f_mntonname, len) == 0 && 699 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 700 (void)strlcpy(sp->f_mntonname, 701 c == '\0' ? "/" : &sp->f_mntonname[len], 702 sizeof(sp->f_mntonname)); 703 } else { 704 if (root) 705 (void)strlcpy(sp->f_mntonname, "/", 706 sizeof(sp->f_mntonname)); 707 else 708 error = EPERM; 709 } 710 } 711 PNBUF_PUT(path); 712 } 713 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 714 return error; 715 } 716 717 /* 718 * Get filesystem statistics by path. 719 */ 720 int 721 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 722 { 723 struct mount *mp; 724 int error; 725 struct vnode *vp; 726 727 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 728 if (error != 0) 729 return error; 730 mp = vp->v_mount; 731 error = dostatvfs(mp, sb, l, flags, 1); 732 vrele(vp); 733 return error; 734 } 735 736 /* ARGSUSED */ 737 int 738 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 739 { 740 /* { 741 syscallarg(const char *) path; 742 syscallarg(struct statvfs *) buf; 743 syscallarg(int) flags; 744 } */ 745 struct statvfs *sb; 746 int error; 747 748 sb = STATVFSBUF_GET(); 749 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 750 if (error == 0) 751 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 752 STATVFSBUF_PUT(sb); 753 return error; 754 } 755 756 /* 757 * Get filesystem statistics by fd. 758 */ 759 int 760 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 761 { 762 file_t *fp; 763 struct mount *mp; 764 int error; 765 766 /* fd_getvnode() will use the descriptor for us */ 767 if ((error = fd_getvnode(fd, &fp)) != 0) 768 return (error); 769 mp = ((struct vnode *)fp->f_data)->v_mount; 770 error = dostatvfs(mp, sb, curlwp, flags, 1); 771 fd_putfile(fd); 772 return error; 773 } 774 775 /* ARGSUSED */ 776 int 777 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 778 { 779 /* { 780 syscallarg(int) fd; 781 syscallarg(struct statvfs *) buf; 782 syscallarg(int) flags; 783 } */ 784 struct statvfs *sb; 785 int error; 786 787 sb = STATVFSBUF_GET(); 788 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 789 if (error == 0) 790 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 791 STATVFSBUF_PUT(sb); 792 return error; 793 } 794 795 796 /* 797 * Get statistics on all filesystems. 798 */ 799 int 800 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 801 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 802 register_t *retval) 803 { 804 int root = 0; 805 struct proc *p = l->l_proc; 806 struct mount *mp, *nmp; 807 struct statvfs *sb; 808 size_t count, maxcount; 809 int error = 0; 810 811 sb = STATVFSBUF_GET(); 812 maxcount = bufsize / entry_sz; 813 mutex_enter(&mountlist_lock); 814 count = 0; 815 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 816 mp = nmp) { 817 if (vfs_busy(mp, &nmp)) { 818 continue; 819 } 820 if (sfsp && count < maxcount) { 821 error = dostatvfs(mp, sb, l, flags, 0); 822 if (error) { 823 vfs_unbusy(mp, false, &nmp); 824 error = 0; 825 continue; 826 } 827 error = copyfn(sb, sfsp, entry_sz); 828 if (error) { 829 vfs_unbusy(mp, false, NULL); 830 goto out; 831 } 832 sfsp = (char *)sfsp + entry_sz; 833 root |= strcmp(sb->f_mntonname, "/") == 0; 834 } 835 count++; 836 vfs_unbusy(mp, false, &nmp); 837 } 838 mutex_exit(&mountlist_lock); 839 840 if (root == 0 && p->p_cwdi->cwdi_rdir) { 841 /* 842 * fake a root entry 843 */ 844 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 845 sb, l, flags, 1); 846 if (error != 0) 847 goto out; 848 if (sfsp) { 849 error = copyfn(sb, sfsp, entry_sz); 850 if (error != 0) 851 goto out; 852 } 853 count++; 854 } 855 if (sfsp && count > maxcount) 856 *retval = maxcount; 857 else 858 *retval = count; 859 out: 860 STATVFSBUF_PUT(sb); 861 return error; 862 } 863 864 int 865 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 866 { 867 /* { 868 syscallarg(struct statvfs *) buf; 869 syscallarg(size_t) bufsize; 870 syscallarg(int) flags; 871 } */ 872 873 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 874 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 875 } 876 877 /* 878 * Change current working directory to a given file descriptor. 879 */ 880 /* ARGSUSED */ 881 int 882 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 883 { 884 /* { 885 syscallarg(int) fd; 886 } */ 887 struct proc *p = l->l_proc; 888 struct cwdinfo *cwdi; 889 struct vnode *vp, *tdp; 890 struct mount *mp; 891 file_t *fp; 892 int error, fd; 893 894 /* fd_getvnode() will use the descriptor for us */ 895 fd = SCARG(uap, fd); 896 if ((error = fd_getvnode(fd, &fp)) != 0) 897 return (error); 898 vp = fp->f_data; 899 900 vref(vp); 901 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 902 if (vp->v_type != VDIR) 903 error = ENOTDIR; 904 else 905 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 906 if (error) { 907 vput(vp); 908 goto out; 909 } 910 while ((mp = vp->v_mountedhere) != NULL) { 911 error = vfs_busy(mp, NULL); 912 vput(vp); 913 if (error != 0) 914 goto out; 915 error = VFS_ROOT(mp, &tdp); 916 vfs_unbusy(mp, false, NULL); 917 if (error) 918 goto out; 919 vp = tdp; 920 } 921 VOP_UNLOCK(vp); 922 923 /* 924 * Disallow changing to a directory not under the process's 925 * current root directory (if there is one). 926 */ 927 cwdi = p->p_cwdi; 928 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 929 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 930 vrele(vp); 931 error = EPERM; /* operation not permitted */ 932 } else { 933 vrele(cwdi->cwdi_cdir); 934 cwdi->cwdi_cdir = vp; 935 } 936 rw_exit(&cwdi->cwdi_lock); 937 938 out: 939 fd_putfile(fd); 940 return (error); 941 } 942 943 /* 944 * Change this process's notion of the root directory to a given file 945 * descriptor. 946 */ 947 int 948 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 949 { 950 struct proc *p = l->l_proc; 951 struct vnode *vp; 952 file_t *fp; 953 int error, fd = SCARG(uap, fd); 954 955 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 956 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 957 return error; 958 /* fd_getvnode() will use the descriptor for us */ 959 if ((error = fd_getvnode(fd, &fp)) != 0) 960 return error; 961 vp = fp->f_data; 962 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 963 if (vp->v_type != VDIR) 964 error = ENOTDIR; 965 else 966 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 967 VOP_UNLOCK(vp); 968 if (error) 969 goto out; 970 vref(vp); 971 972 change_root(p->p_cwdi, vp, l); 973 974 out: 975 fd_putfile(fd); 976 return (error); 977 } 978 979 /* 980 * Change current working directory (``.''). 981 */ 982 /* ARGSUSED */ 983 int 984 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 985 { 986 /* { 987 syscallarg(const char *) path; 988 } */ 989 struct proc *p = l->l_proc; 990 struct cwdinfo *cwdi; 991 int error; 992 struct vnode *vp; 993 994 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 995 &vp, l)) != 0) 996 return (error); 997 cwdi = p->p_cwdi; 998 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 999 vrele(cwdi->cwdi_cdir); 1000 cwdi->cwdi_cdir = vp; 1001 rw_exit(&cwdi->cwdi_lock); 1002 return (0); 1003 } 1004 1005 /* 1006 * Change notion of root (``/'') directory. 1007 */ 1008 /* ARGSUSED */ 1009 int 1010 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1011 { 1012 /* { 1013 syscallarg(const char *) path; 1014 } */ 1015 struct proc *p = l->l_proc; 1016 int error; 1017 struct vnode *vp; 1018 1019 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1020 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1021 return (error); 1022 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1023 &vp, l)) != 0) 1024 return (error); 1025 1026 change_root(p->p_cwdi, vp, l); 1027 1028 return (0); 1029 } 1030 1031 /* 1032 * Common routine for chroot and fchroot. 1033 * NB: callers need to properly authorize the change root operation. 1034 */ 1035 void 1036 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1037 { 1038 1039 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1040 if (cwdi->cwdi_rdir != NULL) 1041 vrele(cwdi->cwdi_rdir); 1042 cwdi->cwdi_rdir = vp; 1043 1044 /* 1045 * Prevent escaping from chroot by putting the root under 1046 * the working directory. Silently chdir to / if we aren't 1047 * already there. 1048 */ 1049 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1050 /* 1051 * XXX would be more failsafe to change directory to a 1052 * deadfs node here instead 1053 */ 1054 vrele(cwdi->cwdi_cdir); 1055 vref(vp); 1056 cwdi->cwdi_cdir = vp; 1057 } 1058 rw_exit(&cwdi->cwdi_lock); 1059 } 1060 1061 /* 1062 * Common routine for chroot and chdir. 1063 * XXX "where" should be enum uio_seg 1064 */ 1065 int 1066 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1067 { 1068 struct pathbuf *pb; 1069 struct nameidata nd; 1070 int error; 1071 1072 error = pathbuf_maybe_copyin(path, where, &pb); 1073 if (error) { 1074 return error; 1075 } 1076 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1077 if ((error = namei(&nd)) != 0) { 1078 pathbuf_destroy(pb); 1079 return error; 1080 } 1081 *vpp = nd.ni_vp; 1082 pathbuf_destroy(pb); 1083 1084 if ((*vpp)->v_type != VDIR) 1085 error = ENOTDIR; 1086 else 1087 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1088 1089 if (error) 1090 vput(*vpp); 1091 else 1092 VOP_UNLOCK(*vpp); 1093 return (error); 1094 } 1095 1096 /* 1097 * Check permissions, allocate an open file structure, 1098 * and call the device open routine if any. 1099 */ 1100 int 1101 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1102 { 1103 /* { 1104 syscallarg(const char *) path; 1105 syscallarg(int) flags; 1106 syscallarg(int) mode; 1107 } */ 1108 struct proc *p = l->l_proc; 1109 struct cwdinfo *cwdi = p->p_cwdi; 1110 file_t *fp; 1111 struct vnode *vp; 1112 int flags, cmode; 1113 int indx, error; 1114 struct pathbuf *pb; 1115 struct nameidata nd; 1116 1117 flags = FFLAGS(SCARG(uap, flags)); 1118 if ((flags & (FREAD | FWRITE)) == 0) 1119 return (EINVAL); 1120 1121 error = pathbuf_copyin(SCARG(uap, path), &pb); 1122 if (error) { 1123 return error; 1124 } 1125 1126 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1127 pathbuf_destroy(pb); 1128 return error; 1129 } 1130 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1131 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1132 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1133 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1134 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1135 fd_abort(p, fp, indx); 1136 if ((error == EDUPFD || error == EMOVEFD) && 1137 l->l_dupfd >= 0 && /* XXX from fdopen */ 1138 (error = 1139 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1140 *retval = indx; 1141 pathbuf_destroy(pb); 1142 return (0); 1143 } 1144 if (error == ERESTART) 1145 error = EINTR; 1146 pathbuf_destroy(pb); 1147 return (error); 1148 } 1149 1150 l->l_dupfd = 0; 1151 vp = nd.ni_vp; 1152 pathbuf_destroy(pb); 1153 1154 if ((error = open_setfp(l, fp, vp, indx, flags))) 1155 return error; 1156 1157 VOP_UNLOCK(vp); 1158 *retval = indx; 1159 fd_affix(p, fp, indx); 1160 return (0); 1161 } 1162 1163 static void 1164 vfs__fhfree(fhandle_t *fhp) 1165 { 1166 size_t fhsize; 1167 1168 if (fhp == NULL) { 1169 return; 1170 } 1171 fhsize = FHANDLE_SIZE(fhp); 1172 kmem_free(fhp, fhsize); 1173 } 1174 1175 /* 1176 * vfs_composefh: compose a filehandle. 1177 */ 1178 1179 int 1180 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1181 { 1182 struct mount *mp; 1183 struct fid *fidp; 1184 int error; 1185 size_t needfhsize; 1186 size_t fidsize; 1187 1188 mp = vp->v_mount; 1189 fidp = NULL; 1190 if (*fh_size < FHANDLE_SIZE_MIN) { 1191 fidsize = 0; 1192 } else { 1193 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1194 if (fhp != NULL) { 1195 memset(fhp, 0, *fh_size); 1196 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1197 fidp = &fhp->fh_fid; 1198 } 1199 } 1200 error = VFS_VPTOFH(vp, fidp, &fidsize); 1201 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1202 if (error == 0 && *fh_size < needfhsize) { 1203 error = E2BIG; 1204 } 1205 *fh_size = needfhsize; 1206 return error; 1207 } 1208 1209 int 1210 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1211 { 1212 struct mount *mp; 1213 fhandle_t *fhp; 1214 size_t fhsize; 1215 size_t fidsize; 1216 int error; 1217 1218 *fhpp = NULL; 1219 mp = vp->v_mount; 1220 fidsize = 0; 1221 error = VFS_VPTOFH(vp, NULL, &fidsize); 1222 KASSERT(error != 0); 1223 if (error != E2BIG) { 1224 goto out; 1225 } 1226 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1227 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1228 if (fhp == NULL) { 1229 error = ENOMEM; 1230 goto out; 1231 } 1232 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1233 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1234 if (error == 0) { 1235 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1236 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1237 *fhpp = fhp; 1238 } else { 1239 kmem_free(fhp, fhsize); 1240 } 1241 out: 1242 return error; 1243 } 1244 1245 void 1246 vfs_composefh_free(fhandle_t *fhp) 1247 { 1248 1249 vfs__fhfree(fhp); 1250 } 1251 1252 /* 1253 * vfs_fhtovp: lookup a vnode by a filehandle. 1254 */ 1255 1256 int 1257 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1258 { 1259 struct mount *mp; 1260 int error; 1261 1262 *vpp = NULL; 1263 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1264 if (mp == NULL) { 1265 error = ESTALE; 1266 goto out; 1267 } 1268 if (mp->mnt_op->vfs_fhtovp == NULL) { 1269 error = EOPNOTSUPP; 1270 goto out; 1271 } 1272 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1273 out: 1274 return error; 1275 } 1276 1277 /* 1278 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1279 * the needed size. 1280 */ 1281 1282 int 1283 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1284 { 1285 fhandle_t *fhp; 1286 int error; 1287 1288 *fhpp = NULL; 1289 if (fhsize > FHANDLE_SIZE_MAX) { 1290 return EINVAL; 1291 } 1292 if (fhsize < FHANDLE_SIZE_MIN) { 1293 return EINVAL; 1294 } 1295 again: 1296 fhp = kmem_alloc(fhsize, KM_SLEEP); 1297 if (fhp == NULL) { 1298 return ENOMEM; 1299 } 1300 error = copyin(ufhp, fhp, fhsize); 1301 if (error == 0) { 1302 /* XXX this check shouldn't be here */ 1303 if (FHANDLE_SIZE(fhp) == fhsize) { 1304 *fhpp = fhp; 1305 return 0; 1306 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1307 /* 1308 * a kludge for nfsv2 padded handles. 1309 */ 1310 size_t sz; 1311 1312 sz = FHANDLE_SIZE(fhp); 1313 kmem_free(fhp, fhsize); 1314 fhsize = sz; 1315 goto again; 1316 } else { 1317 /* 1318 * userland told us wrong size. 1319 */ 1320 error = EINVAL; 1321 } 1322 } 1323 kmem_free(fhp, fhsize); 1324 return error; 1325 } 1326 1327 void 1328 vfs_copyinfh_free(fhandle_t *fhp) 1329 { 1330 1331 vfs__fhfree(fhp); 1332 } 1333 1334 /* 1335 * Get file handle system call 1336 */ 1337 int 1338 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1339 { 1340 /* { 1341 syscallarg(char *) fname; 1342 syscallarg(fhandle_t *) fhp; 1343 syscallarg(size_t *) fh_size; 1344 } */ 1345 struct vnode *vp; 1346 fhandle_t *fh; 1347 int error; 1348 struct pathbuf *pb; 1349 struct nameidata nd; 1350 size_t sz; 1351 size_t usz; 1352 1353 /* 1354 * Must be super user 1355 */ 1356 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1357 0, NULL, NULL, NULL); 1358 if (error) 1359 return (error); 1360 1361 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1362 if (error) { 1363 return error; 1364 } 1365 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1366 error = namei(&nd); 1367 if (error) { 1368 pathbuf_destroy(pb); 1369 return error; 1370 } 1371 vp = nd.ni_vp; 1372 pathbuf_destroy(pb); 1373 1374 error = vfs_composefh_alloc(vp, &fh); 1375 vput(vp); 1376 if (error != 0) { 1377 goto out; 1378 } 1379 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1380 if (error != 0) { 1381 goto out; 1382 } 1383 sz = FHANDLE_SIZE(fh); 1384 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1385 if (error != 0) { 1386 goto out; 1387 } 1388 if (usz >= sz) { 1389 error = copyout(fh, SCARG(uap, fhp), sz); 1390 } else { 1391 error = E2BIG; 1392 } 1393 out: 1394 vfs_composefh_free(fh); 1395 return (error); 1396 } 1397 1398 /* 1399 * Open a file given a file handle. 1400 * 1401 * Check permissions, allocate an open file structure, 1402 * and call the device open routine if any. 1403 */ 1404 1405 int 1406 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1407 register_t *retval) 1408 { 1409 file_t *fp; 1410 struct vnode *vp = NULL; 1411 kauth_cred_t cred = l->l_cred; 1412 file_t *nfp; 1413 int indx, error = 0; 1414 struct vattr va; 1415 fhandle_t *fh; 1416 int flags; 1417 proc_t *p; 1418 1419 p = curproc; 1420 1421 /* 1422 * Must be super user 1423 */ 1424 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1425 0, NULL, NULL, NULL))) 1426 return (error); 1427 1428 flags = FFLAGS(oflags); 1429 if ((flags & (FREAD | FWRITE)) == 0) 1430 return (EINVAL); 1431 if ((flags & O_CREAT)) 1432 return (EINVAL); 1433 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1434 return (error); 1435 fp = nfp; 1436 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1437 if (error != 0) { 1438 goto bad; 1439 } 1440 error = vfs_fhtovp(fh, &vp); 1441 if (error != 0) { 1442 goto bad; 1443 } 1444 1445 /* Now do an effective vn_open */ 1446 1447 if (vp->v_type == VSOCK) { 1448 error = EOPNOTSUPP; 1449 goto bad; 1450 } 1451 error = vn_openchk(vp, cred, flags); 1452 if (error != 0) 1453 goto bad; 1454 if (flags & O_TRUNC) { 1455 VOP_UNLOCK(vp); /* XXX */ 1456 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1457 vattr_null(&va); 1458 va.va_size = 0; 1459 error = VOP_SETATTR(vp, &va, cred); 1460 if (error) 1461 goto bad; 1462 } 1463 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1464 goto bad; 1465 if (flags & FWRITE) { 1466 mutex_enter(vp->v_interlock); 1467 vp->v_writecount++; 1468 mutex_exit(vp->v_interlock); 1469 } 1470 1471 /* done with modified vn_open, now finish what sys_open does. */ 1472 if ((error = open_setfp(l, fp, vp, indx, flags))) 1473 return error; 1474 1475 VOP_UNLOCK(vp); 1476 *retval = indx; 1477 fd_affix(p, fp, indx); 1478 vfs_copyinfh_free(fh); 1479 return (0); 1480 1481 bad: 1482 fd_abort(p, fp, indx); 1483 if (vp != NULL) 1484 vput(vp); 1485 vfs_copyinfh_free(fh); 1486 return (error); 1487 } 1488 1489 int 1490 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1491 { 1492 /* { 1493 syscallarg(const void *) fhp; 1494 syscallarg(size_t) fh_size; 1495 syscallarg(int) flags; 1496 } */ 1497 1498 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1499 SCARG(uap, flags), retval); 1500 } 1501 1502 int 1503 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1504 { 1505 int error; 1506 fhandle_t *fh; 1507 struct vnode *vp; 1508 1509 /* 1510 * Must be super user 1511 */ 1512 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1513 0, NULL, NULL, NULL))) 1514 return (error); 1515 1516 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1517 if (error != 0) 1518 return error; 1519 1520 error = vfs_fhtovp(fh, &vp); 1521 vfs_copyinfh_free(fh); 1522 if (error != 0) 1523 return error; 1524 1525 error = vn_stat(vp, sb); 1526 vput(vp); 1527 return error; 1528 } 1529 1530 1531 /* ARGSUSED */ 1532 int 1533 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 1534 { 1535 /* { 1536 syscallarg(const void *) fhp; 1537 syscallarg(size_t) fh_size; 1538 syscallarg(struct stat *) sb; 1539 } */ 1540 struct stat sb; 1541 int error; 1542 1543 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1544 if (error) 1545 return error; 1546 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1547 } 1548 1549 int 1550 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1551 int flags) 1552 { 1553 fhandle_t *fh; 1554 struct mount *mp; 1555 struct vnode *vp; 1556 int error; 1557 1558 /* 1559 * Must be super user 1560 */ 1561 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1562 0, NULL, NULL, NULL))) 1563 return error; 1564 1565 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1566 if (error != 0) 1567 return error; 1568 1569 error = vfs_fhtovp(fh, &vp); 1570 vfs_copyinfh_free(fh); 1571 if (error != 0) 1572 return error; 1573 1574 mp = vp->v_mount; 1575 error = dostatvfs(mp, sb, l, flags, 1); 1576 vput(vp); 1577 return error; 1578 } 1579 1580 /* ARGSUSED */ 1581 int 1582 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1583 { 1584 /* { 1585 syscallarg(const void *) fhp; 1586 syscallarg(size_t) fh_size; 1587 syscallarg(struct statvfs *) buf; 1588 syscallarg(int) flags; 1589 } */ 1590 struct statvfs *sb = STATVFSBUF_GET(); 1591 int error; 1592 1593 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1594 SCARG(uap, flags)); 1595 if (error == 0) 1596 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1597 STATVFSBUF_PUT(sb); 1598 return error; 1599 } 1600 1601 /* 1602 * Create a special file. 1603 */ 1604 /* ARGSUSED */ 1605 int 1606 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 1607 register_t *retval) 1608 { 1609 /* { 1610 syscallarg(const char *) path; 1611 syscallarg(mode_t) mode; 1612 syscallarg(dev_t) dev; 1613 } */ 1614 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 1615 SCARG(uap, dev), retval, UIO_USERSPACE); 1616 } 1617 1618 int 1619 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 1620 register_t *retval, enum uio_seg seg) 1621 { 1622 struct proc *p = l->l_proc; 1623 struct vnode *vp; 1624 struct vattr vattr; 1625 int error, optype; 1626 struct pathbuf *pb; 1627 struct nameidata nd; 1628 const char *pathstring; 1629 1630 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1631 0, NULL, NULL, NULL)) != 0) 1632 return (error); 1633 1634 optype = VOP_MKNOD_DESCOFFSET; 1635 1636 error = pathbuf_maybe_copyin(pathname, seg, &pb); 1637 if (error) { 1638 return error; 1639 } 1640 pathstring = pathbuf_stringcopy_get(pb); 1641 if (pathstring == NULL) { 1642 pathbuf_destroy(pb); 1643 return ENOMEM; 1644 } 1645 1646 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 1647 if ((error = namei(&nd)) != 0) 1648 goto out; 1649 vp = nd.ni_vp; 1650 1651 if (vp != NULL) 1652 error = EEXIST; 1653 else { 1654 vattr_null(&vattr); 1655 /* We will read cwdi->cwdi_cmask unlocked. */ 1656 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1657 vattr.va_rdev = dev; 1658 1659 switch (mode & S_IFMT) { 1660 case S_IFMT: /* used by badsect to flag bad sectors */ 1661 vattr.va_type = VBAD; 1662 break; 1663 case S_IFCHR: 1664 vattr.va_type = VCHR; 1665 break; 1666 case S_IFBLK: 1667 vattr.va_type = VBLK; 1668 break; 1669 case S_IFWHT: 1670 optype = VOP_WHITEOUT_DESCOFFSET; 1671 break; 1672 case S_IFREG: 1673 #if NVERIEXEC > 0 1674 error = veriexec_openchk(l, nd.ni_vp, pathstring, 1675 O_CREAT); 1676 #endif /* NVERIEXEC > 0 */ 1677 vattr.va_type = VREG; 1678 vattr.va_rdev = VNOVAL; 1679 optype = VOP_CREATE_DESCOFFSET; 1680 break; 1681 default: 1682 error = EINVAL; 1683 break; 1684 } 1685 } 1686 if (optype == VOP_MKNOD_DESCOFFSET && vattr.va_rdev == VNOVAL) 1687 error = EINVAL; 1688 if (!error) { 1689 switch (optype) { 1690 case VOP_WHITEOUT_DESCOFFSET: 1691 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1692 if (error) 1693 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1694 vput(nd.ni_dvp); 1695 break; 1696 1697 case VOP_MKNOD_DESCOFFSET: 1698 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1699 &nd.ni_cnd, &vattr); 1700 if (error == 0) 1701 vput(nd.ni_vp); 1702 break; 1703 1704 case VOP_CREATE_DESCOFFSET: 1705 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1706 &nd.ni_cnd, &vattr); 1707 if (error == 0) 1708 vput(nd.ni_vp); 1709 break; 1710 } 1711 } else { 1712 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1713 if (nd.ni_dvp == vp) 1714 vrele(nd.ni_dvp); 1715 else 1716 vput(nd.ni_dvp); 1717 if (vp) 1718 vrele(vp); 1719 } 1720 out: 1721 pathbuf_stringcopy_put(pb, pathstring); 1722 pathbuf_destroy(pb); 1723 return (error); 1724 } 1725 1726 /* 1727 * Create a named pipe. 1728 */ 1729 /* ARGSUSED */ 1730 int 1731 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1732 { 1733 /* { 1734 syscallarg(const char *) path; 1735 syscallarg(int) mode; 1736 } */ 1737 struct proc *p = l->l_proc; 1738 struct vattr vattr; 1739 int error; 1740 struct pathbuf *pb; 1741 struct nameidata nd; 1742 1743 error = pathbuf_copyin(SCARG(uap, path), &pb); 1744 if (error) { 1745 return error; 1746 } 1747 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 1748 if ((error = namei(&nd)) != 0) { 1749 pathbuf_destroy(pb); 1750 return error; 1751 } 1752 if (nd.ni_vp != NULL) { 1753 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1754 if (nd.ni_dvp == nd.ni_vp) 1755 vrele(nd.ni_dvp); 1756 else 1757 vput(nd.ni_dvp); 1758 vrele(nd.ni_vp); 1759 pathbuf_destroy(pb); 1760 return (EEXIST); 1761 } 1762 vattr_null(&vattr); 1763 vattr.va_type = VFIFO; 1764 /* We will read cwdi->cwdi_cmask unlocked. */ 1765 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1766 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1767 if (error == 0) 1768 vput(nd.ni_vp); 1769 pathbuf_destroy(pb); 1770 return (error); 1771 } 1772 1773 /* 1774 * Make a hard file link. 1775 */ 1776 /* ARGSUSED */ 1777 int 1778 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1779 { 1780 /* { 1781 syscallarg(const char *) path; 1782 syscallarg(const char *) link; 1783 } */ 1784 struct vnode *vp; 1785 struct pathbuf *linkpb; 1786 struct nameidata nd; 1787 int error; 1788 1789 error = namei_simple_user(SCARG(uap, path), 1790 NSM_FOLLOW_TRYEMULROOT, &vp); 1791 if (error != 0) 1792 return (error); 1793 error = pathbuf_copyin(SCARG(uap, link), &linkpb); 1794 if (error) { 1795 goto out1; 1796 } 1797 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 1798 if ((error = namei(&nd)) != 0) 1799 goto out2; 1800 if (nd.ni_vp) { 1801 error = EEXIST; 1802 goto abortop; 1803 } 1804 /* Prevent hard links on directories. */ 1805 if (vp->v_type == VDIR) { 1806 error = EPERM; 1807 goto abortop; 1808 } 1809 /* Prevent cross-mount operation. */ 1810 if (nd.ni_dvp->v_mount != vp->v_mount) { 1811 error = EXDEV; 1812 goto abortop; 1813 } 1814 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1815 out2: 1816 pathbuf_destroy(linkpb); 1817 out1: 1818 vrele(vp); 1819 return (error); 1820 abortop: 1821 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1822 if (nd.ni_dvp == nd.ni_vp) 1823 vrele(nd.ni_dvp); 1824 else 1825 vput(nd.ni_dvp); 1826 if (nd.ni_vp != NULL) 1827 vrele(nd.ni_vp); 1828 goto out2; 1829 } 1830 1831 int 1832 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 1833 { 1834 struct proc *p = curproc; 1835 struct vattr vattr; 1836 char *path; 1837 int error; 1838 struct pathbuf *linkpb; 1839 struct nameidata nd; 1840 1841 path = PNBUF_GET(); 1842 if (seg == UIO_USERSPACE) { 1843 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 1844 goto out1; 1845 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 1846 goto out1; 1847 } else { 1848 KASSERT(strlen(patharg) < MAXPATHLEN); 1849 strcpy(path, patharg); 1850 linkpb = pathbuf_create(link); 1851 if (linkpb == NULL) { 1852 error = ENOMEM; 1853 goto out1; 1854 } 1855 } 1856 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 1857 if ((error = namei(&nd)) != 0) 1858 goto out2; 1859 if (nd.ni_vp) { 1860 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1861 if (nd.ni_dvp == nd.ni_vp) 1862 vrele(nd.ni_dvp); 1863 else 1864 vput(nd.ni_dvp); 1865 vrele(nd.ni_vp); 1866 error = EEXIST; 1867 goto out2; 1868 } 1869 vattr_null(&vattr); 1870 vattr.va_type = VLNK; 1871 /* We will read cwdi->cwdi_cmask unlocked. */ 1872 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1873 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1874 if (error == 0) 1875 vput(nd.ni_vp); 1876 out2: 1877 pathbuf_destroy(linkpb); 1878 out1: 1879 PNBUF_PUT(path); 1880 return (error); 1881 } 1882 1883 /* 1884 * Make a symbolic link. 1885 */ 1886 /* ARGSUSED */ 1887 int 1888 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1889 { 1890 /* { 1891 syscallarg(const char *) path; 1892 syscallarg(const char *) link; 1893 } */ 1894 1895 return do_sys_symlink(SCARG(uap, path), SCARG(uap, link), 1896 UIO_USERSPACE); 1897 } 1898 1899 /* 1900 * Delete a whiteout from the filesystem. 1901 */ 1902 /* ARGSUSED */ 1903 int 1904 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 1905 { 1906 /* { 1907 syscallarg(const char *) path; 1908 } */ 1909 int error; 1910 struct pathbuf *pb; 1911 struct nameidata nd; 1912 1913 error = pathbuf_copyin(SCARG(uap, path), &pb); 1914 if (error) { 1915 return error; 1916 } 1917 1918 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 1919 error = namei(&nd); 1920 if (error) { 1921 pathbuf_destroy(pb); 1922 return (error); 1923 } 1924 1925 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1926 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1927 if (nd.ni_dvp == nd.ni_vp) 1928 vrele(nd.ni_dvp); 1929 else 1930 vput(nd.ni_dvp); 1931 if (nd.ni_vp) 1932 vrele(nd.ni_vp); 1933 pathbuf_destroy(pb); 1934 return (EEXIST); 1935 } 1936 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1937 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1938 vput(nd.ni_dvp); 1939 pathbuf_destroy(pb); 1940 return (error); 1941 } 1942 1943 /* 1944 * Delete a name from the filesystem. 1945 */ 1946 /* ARGSUSED */ 1947 int 1948 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 1949 { 1950 /* { 1951 syscallarg(const char *) path; 1952 } */ 1953 1954 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 1955 } 1956 1957 int 1958 do_sys_unlink(const char *arg, enum uio_seg seg) 1959 { 1960 struct vnode *vp; 1961 int error; 1962 struct pathbuf *pb; 1963 struct nameidata nd; 1964 const char *pathstring; 1965 1966 error = pathbuf_maybe_copyin(arg, seg, &pb); 1967 if (error) { 1968 return error; 1969 } 1970 pathstring = pathbuf_stringcopy_get(pb); 1971 if (pathstring == NULL) { 1972 pathbuf_destroy(pb); 1973 return ENOMEM; 1974 } 1975 1976 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 1977 if ((error = namei(&nd)) != 0) 1978 goto out; 1979 vp = nd.ni_vp; 1980 1981 /* 1982 * The root of a mounted filesystem cannot be deleted. 1983 */ 1984 if (vp->v_vflag & VV_ROOT) { 1985 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1986 if (nd.ni_dvp == vp) 1987 vrele(nd.ni_dvp); 1988 else 1989 vput(nd.ni_dvp); 1990 vput(vp); 1991 error = EBUSY; 1992 goto out; 1993 } 1994 1995 #if NVERIEXEC > 0 1996 /* Handle remove requests for veriexec entries. */ 1997 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 1998 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1999 if (nd.ni_dvp == vp) 2000 vrele(nd.ni_dvp); 2001 else 2002 vput(nd.ni_dvp); 2003 vput(vp); 2004 goto out; 2005 } 2006 #endif /* NVERIEXEC > 0 */ 2007 2008 #ifdef FILEASSOC 2009 (void)fileassoc_file_delete(vp); 2010 #endif /* FILEASSOC */ 2011 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2012 out: 2013 pathbuf_stringcopy_put(pb, pathstring); 2014 pathbuf_destroy(pb); 2015 return (error); 2016 } 2017 2018 /* 2019 * Reposition read/write file offset. 2020 */ 2021 int 2022 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2023 { 2024 /* { 2025 syscallarg(int) fd; 2026 syscallarg(int) pad; 2027 syscallarg(off_t) offset; 2028 syscallarg(int) whence; 2029 } */ 2030 kauth_cred_t cred = l->l_cred; 2031 file_t *fp; 2032 struct vnode *vp; 2033 struct vattr vattr; 2034 off_t newoff; 2035 int error, fd; 2036 2037 fd = SCARG(uap, fd); 2038 2039 if ((fp = fd_getfile(fd)) == NULL) 2040 return (EBADF); 2041 2042 vp = fp->f_data; 2043 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2044 error = ESPIPE; 2045 goto out; 2046 } 2047 2048 switch (SCARG(uap, whence)) { 2049 case SEEK_CUR: 2050 newoff = fp->f_offset + SCARG(uap, offset); 2051 break; 2052 case SEEK_END: 2053 error = VOP_GETATTR(vp, &vattr, cred); 2054 if (error) { 2055 goto out; 2056 } 2057 newoff = SCARG(uap, offset) + vattr.va_size; 2058 break; 2059 case SEEK_SET: 2060 newoff = SCARG(uap, offset); 2061 break; 2062 default: 2063 error = EINVAL; 2064 goto out; 2065 } 2066 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2067 *(off_t *)retval = fp->f_offset = newoff; 2068 } 2069 out: 2070 fd_putfile(fd); 2071 return (error); 2072 } 2073 2074 /* 2075 * Positional read system call. 2076 */ 2077 int 2078 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2079 { 2080 /* { 2081 syscallarg(int) fd; 2082 syscallarg(void *) buf; 2083 syscallarg(size_t) nbyte; 2084 syscallarg(off_t) offset; 2085 } */ 2086 file_t *fp; 2087 struct vnode *vp; 2088 off_t offset; 2089 int error, fd = SCARG(uap, fd); 2090 2091 if ((fp = fd_getfile(fd)) == NULL) 2092 return (EBADF); 2093 2094 if ((fp->f_flag & FREAD) == 0) { 2095 fd_putfile(fd); 2096 return (EBADF); 2097 } 2098 2099 vp = fp->f_data; 2100 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2101 error = ESPIPE; 2102 goto out; 2103 } 2104 2105 offset = SCARG(uap, offset); 2106 2107 /* 2108 * XXX This works because no file systems actually 2109 * XXX take any action on the seek operation. 2110 */ 2111 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2112 goto out; 2113 2114 /* dofileread() will unuse the descriptor for us */ 2115 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2116 &offset, 0, retval)); 2117 2118 out: 2119 fd_putfile(fd); 2120 return (error); 2121 } 2122 2123 /* 2124 * Positional scatter read system call. 2125 */ 2126 int 2127 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2128 { 2129 /* { 2130 syscallarg(int) fd; 2131 syscallarg(const struct iovec *) iovp; 2132 syscallarg(int) iovcnt; 2133 syscallarg(off_t) offset; 2134 } */ 2135 off_t offset = SCARG(uap, offset); 2136 2137 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2138 SCARG(uap, iovcnt), &offset, 0, retval); 2139 } 2140 2141 /* 2142 * Positional write system call. 2143 */ 2144 int 2145 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2146 { 2147 /* { 2148 syscallarg(int) fd; 2149 syscallarg(const void *) buf; 2150 syscallarg(size_t) nbyte; 2151 syscallarg(off_t) offset; 2152 } */ 2153 file_t *fp; 2154 struct vnode *vp; 2155 off_t offset; 2156 int error, fd = SCARG(uap, fd); 2157 2158 if ((fp = fd_getfile(fd)) == NULL) 2159 return (EBADF); 2160 2161 if ((fp->f_flag & FWRITE) == 0) { 2162 fd_putfile(fd); 2163 return (EBADF); 2164 } 2165 2166 vp = fp->f_data; 2167 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2168 error = ESPIPE; 2169 goto out; 2170 } 2171 2172 offset = SCARG(uap, offset); 2173 2174 /* 2175 * XXX This works because no file systems actually 2176 * XXX take any action on the seek operation. 2177 */ 2178 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2179 goto out; 2180 2181 /* dofilewrite() will unuse the descriptor for us */ 2182 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2183 &offset, 0, retval)); 2184 2185 out: 2186 fd_putfile(fd); 2187 return (error); 2188 } 2189 2190 /* 2191 * Positional gather write system call. 2192 */ 2193 int 2194 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2195 { 2196 /* { 2197 syscallarg(int) fd; 2198 syscallarg(const struct iovec *) iovp; 2199 syscallarg(int) iovcnt; 2200 syscallarg(off_t) offset; 2201 } */ 2202 off_t offset = SCARG(uap, offset); 2203 2204 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2205 SCARG(uap, iovcnt), &offset, 0, retval); 2206 } 2207 2208 /* 2209 * Check access permissions. 2210 */ 2211 int 2212 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2213 { 2214 /* { 2215 syscallarg(const char *) path; 2216 syscallarg(int) flags; 2217 } */ 2218 kauth_cred_t cred; 2219 struct vnode *vp; 2220 int error, flags; 2221 struct pathbuf *pb; 2222 struct nameidata nd; 2223 2224 CTASSERT(F_OK == 0); 2225 if ((SCARG(uap, flags) & ~(R_OK | W_OK | X_OK)) != 0) { 2226 /* nonsense flags */ 2227 return EINVAL; 2228 } 2229 2230 error = pathbuf_copyin(SCARG(uap, path), &pb); 2231 if (error) { 2232 return error; 2233 } 2234 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2235 2236 /* Override default credentials */ 2237 cred = kauth_cred_dup(l->l_cred); 2238 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2239 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2240 nd.ni_cnd.cn_cred = cred; 2241 2242 if ((error = namei(&nd)) != 0) { 2243 pathbuf_destroy(pb); 2244 goto out; 2245 } 2246 vp = nd.ni_vp; 2247 pathbuf_destroy(pb); 2248 2249 /* Flags == 0 means only check for existence. */ 2250 if (SCARG(uap, flags)) { 2251 flags = 0; 2252 if (SCARG(uap, flags) & R_OK) 2253 flags |= VREAD; 2254 if (SCARG(uap, flags) & W_OK) 2255 flags |= VWRITE; 2256 if (SCARG(uap, flags) & X_OK) 2257 flags |= VEXEC; 2258 2259 error = VOP_ACCESS(vp, flags, cred); 2260 if (!error && (flags & VWRITE)) 2261 error = vn_writechk(vp); 2262 } 2263 vput(vp); 2264 out: 2265 kauth_cred_free(cred); 2266 return (error); 2267 } 2268 2269 /* 2270 * Common code for all sys_stat functions, including compat versions. 2271 */ 2272 int 2273 do_sys_stat(const char *userpath, unsigned int nd_flags, struct stat *sb) 2274 { 2275 int error; 2276 struct pathbuf *pb; 2277 struct nameidata nd; 2278 2279 error = pathbuf_copyin(userpath, &pb); 2280 if (error) { 2281 return error; 2282 } 2283 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, pb); 2284 error = namei(&nd); 2285 if (error != 0) { 2286 pathbuf_destroy(pb); 2287 return error; 2288 } 2289 error = vn_stat(nd.ni_vp, sb); 2290 vput(nd.ni_vp); 2291 pathbuf_destroy(pb); 2292 return error; 2293 } 2294 2295 /* 2296 * Get file status; this version follows links. 2297 */ 2298 /* ARGSUSED */ 2299 int 2300 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2301 { 2302 /* { 2303 syscallarg(const char *) path; 2304 syscallarg(struct stat *) ub; 2305 } */ 2306 struct stat sb; 2307 int error; 2308 2309 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2310 if (error) 2311 return error; 2312 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2313 } 2314 2315 /* 2316 * Get file status; this version does not follow links. 2317 */ 2318 /* ARGSUSED */ 2319 int 2320 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2321 { 2322 /* { 2323 syscallarg(const char *) path; 2324 syscallarg(struct stat *) ub; 2325 } */ 2326 struct stat sb; 2327 int error; 2328 2329 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2330 if (error) 2331 return error; 2332 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2333 } 2334 2335 /* 2336 * Get configurable pathname variables. 2337 */ 2338 /* ARGSUSED */ 2339 int 2340 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2341 { 2342 /* { 2343 syscallarg(const char *) path; 2344 syscallarg(int) name; 2345 } */ 2346 int error; 2347 struct pathbuf *pb; 2348 struct nameidata nd; 2349 2350 error = pathbuf_copyin(SCARG(uap, path), &pb); 2351 if (error) { 2352 return error; 2353 } 2354 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2355 if ((error = namei(&nd)) != 0) { 2356 pathbuf_destroy(pb); 2357 return (error); 2358 } 2359 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2360 vput(nd.ni_vp); 2361 pathbuf_destroy(pb); 2362 return (error); 2363 } 2364 2365 /* 2366 * Return target name of a symbolic link. 2367 */ 2368 /* ARGSUSED */ 2369 int 2370 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2371 { 2372 /* { 2373 syscallarg(const char *) path; 2374 syscallarg(char *) buf; 2375 syscallarg(size_t) count; 2376 } */ 2377 struct vnode *vp; 2378 struct iovec aiov; 2379 struct uio auio; 2380 int error; 2381 struct pathbuf *pb; 2382 struct nameidata nd; 2383 2384 error = pathbuf_copyin(SCARG(uap, path), &pb); 2385 if (error) { 2386 return error; 2387 } 2388 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2389 if ((error = namei(&nd)) != 0) { 2390 pathbuf_destroy(pb); 2391 return error; 2392 } 2393 vp = nd.ni_vp; 2394 pathbuf_destroy(pb); 2395 if (vp->v_type != VLNK) 2396 error = EINVAL; 2397 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2398 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2399 aiov.iov_base = SCARG(uap, buf); 2400 aiov.iov_len = SCARG(uap, count); 2401 auio.uio_iov = &aiov; 2402 auio.uio_iovcnt = 1; 2403 auio.uio_offset = 0; 2404 auio.uio_rw = UIO_READ; 2405 KASSERT(l == curlwp); 2406 auio.uio_vmspace = l->l_proc->p_vmspace; 2407 auio.uio_resid = SCARG(uap, count); 2408 error = VOP_READLINK(vp, &auio, l->l_cred); 2409 } 2410 vput(vp); 2411 *retval = SCARG(uap, count) - auio.uio_resid; 2412 return (error); 2413 } 2414 2415 /* 2416 * Change flags of a file given a path name. 2417 */ 2418 /* ARGSUSED */ 2419 int 2420 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2421 { 2422 /* { 2423 syscallarg(const char *) path; 2424 syscallarg(u_long) flags; 2425 } */ 2426 struct vnode *vp; 2427 int error; 2428 2429 error = namei_simple_user(SCARG(uap, path), 2430 NSM_FOLLOW_TRYEMULROOT, &vp); 2431 if (error != 0) 2432 return (error); 2433 error = change_flags(vp, SCARG(uap, flags), l); 2434 vput(vp); 2435 return (error); 2436 } 2437 2438 /* 2439 * Change flags of a file given a file descriptor. 2440 */ 2441 /* ARGSUSED */ 2442 int 2443 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2444 { 2445 /* { 2446 syscallarg(int) fd; 2447 syscallarg(u_long) flags; 2448 } */ 2449 struct vnode *vp; 2450 file_t *fp; 2451 int error; 2452 2453 /* fd_getvnode() will use the descriptor for us */ 2454 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2455 return (error); 2456 vp = fp->f_data; 2457 error = change_flags(vp, SCARG(uap, flags), l); 2458 VOP_UNLOCK(vp); 2459 fd_putfile(SCARG(uap, fd)); 2460 return (error); 2461 } 2462 2463 /* 2464 * Change flags of a file given a path name; this version does 2465 * not follow links. 2466 */ 2467 int 2468 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2469 { 2470 /* { 2471 syscallarg(const char *) path; 2472 syscallarg(u_long) flags; 2473 } */ 2474 struct vnode *vp; 2475 int error; 2476 2477 error = namei_simple_user(SCARG(uap, path), 2478 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2479 if (error != 0) 2480 return (error); 2481 error = change_flags(vp, SCARG(uap, flags), l); 2482 vput(vp); 2483 return (error); 2484 } 2485 2486 /* 2487 * Common routine to change flags of a file. 2488 */ 2489 int 2490 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2491 { 2492 struct vattr vattr; 2493 int error; 2494 2495 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2496 /* 2497 * Non-superusers cannot change the flags on devices, even if they 2498 * own them. 2499 */ 2500 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2501 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2502 goto out; 2503 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2504 error = EINVAL; 2505 goto out; 2506 } 2507 } 2508 vattr_null(&vattr); 2509 vattr.va_flags = flags; 2510 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2511 out: 2512 return (error); 2513 } 2514 2515 /* 2516 * Change mode of a file given path name; this version follows links. 2517 */ 2518 /* ARGSUSED */ 2519 int 2520 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2521 { 2522 /* { 2523 syscallarg(const char *) path; 2524 syscallarg(int) mode; 2525 } */ 2526 int error; 2527 struct vnode *vp; 2528 2529 error = namei_simple_user(SCARG(uap, path), 2530 NSM_FOLLOW_TRYEMULROOT, &vp); 2531 if (error != 0) 2532 return (error); 2533 2534 error = change_mode(vp, SCARG(uap, mode), l); 2535 2536 vrele(vp); 2537 return (error); 2538 } 2539 2540 /* 2541 * Change mode of a file given a file descriptor. 2542 */ 2543 /* ARGSUSED */ 2544 int 2545 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2546 { 2547 /* { 2548 syscallarg(int) fd; 2549 syscallarg(int) mode; 2550 } */ 2551 file_t *fp; 2552 int error; 2553 2554 /* fd_getvnode() will use the descriptor for us */ 2555 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2556 return (error); 2557 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2558 fd_putfile(SCARG(uap, fd)); 2559 return (error); 2560 } 2561 2562 /* 2563 * Change mode of a file given path name; this version does not follow links. 2564 */ 2565 /* ARGSUSED */ 2566 int 2567 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2568 { 2569 /* { 2570 syscallarg(const char *) path; 2571 syscallarg(int) mode; 2572 } */ 2573 int error; 2574 struct vnode *vp; 2575 2576 error = namei_simple_user(SCARG(uap, path), 2577 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2578 if (error != 0) 2579 return (error); 2580 2581 error = change_mode(vp, SCARG(uap, mode), l); 2582 2583 vrele(vp); 2584 return (error); 2585 } 2586 2587 /* 2588 * Common routine to set mode given a vnode. 2589 */ 2590 static int 2591 change_mode(struct vnode *vp, int mode, struct lwp *l) 2592 { 2593 struct vattr vattr; 2594 int error; 2595 2596 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2597 vattr_null(&vattr); 2598 vattr.va_mode = mode & ALLPERMS; 2599 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2600 VOP_UNLOCK(vp); 2601 return (error); 2602 } 2603 2604 /* 2605 * Set ownership given a path name; this version follows links. 2606 */ 2607 /* ARGSUSED */ 2608 int 2609 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2610 { 2611 /* { 2612 syscallarg(const char *) path; 2613 syscallarg(uid_t) uid; 2614 syscallarg(gid_t) gid; 2615 } */ 2616 int error; 2617 struct vnode *vp; 2618 2619 error = namei_simple_user(SCARG(uap, path), 2620 NSM_FOLLOW_TRYEMULROOT, &vp); 2621 if (error != 0) 2622 return (error); 2623 2624 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2625 2626 vrele(vp); 2627 return (error); 2628 } 2629 2630 /* 2631 * Set ownership given a path name; this version follows links. 2632 * Provides POSIX semantics. 2633 */ 2634 /* ARGSUSED */ 2635 int 2636 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2637 { 2638 /* { 2639 syscallarg(const char *) path; 2640 syscallarg(uid_t) uid; 2641 syscallarg(gid_t) gid; 2642 } */ 2643 int error; 2644 struct vnode *vp; 2645 2646 error = namei_simple_user(SCARG(uap, path), 2647 NSM_FOLLOW_TRYEMULROOT, &vp); 2648 if (error != 0) 2649 return (error); 2650 2651 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2652 2653 vrele(vp); 2654 return (error); 2655 } 2656 2657 /* 2658 * Set ownership given a file descriptor. 2659 */ 2660 /* ARGSUSED */ 2661 int 2662 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2663 { 2664 /* { 2665 syscallarg(int) fd; 2666 syscallarg(uid_t) uid; 2667 syscallarg(gid_t) gid; 2668 } */ 2669 int error; 2670 file_t *fp; 2671 2672 /* fd_getvnode() will use the descriptor for us */ 2673 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2674 return (error); 2675 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2676 l, 0); 2677 fd_putfile(SCARG(uap, fd)); 2678 return (error); 2679 } 2680 2681 /* 2682 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2683 */ 2684 /* ARGSUSED */ 2685 int 2686 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2687 { 2688 /* { 2689 syscallarg(int) fd; 2690 syscallarg(uid_t) uid; 2691 syscallarg(gid_t) gid; 2692 } */ 2693 int error; 2694 file_t *fp; 2695 2696 /* fd_getvnode() will use the descriptor for us */ 2697 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2698 return (error); 2699 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2700 l, 1); 2701 fd_putfile(SCARG(uap, fd)); 2702 return (error); 2703 } 2704 2705 /* 2706 * Set ownership given a path name; this version does not follow links. 2707 */ 2708 /* ARGSUSED */ 2709 int 2710 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2711 { 2712 /* { 2713 syscallarg(const char *) path; 2714 syscallarg(uid_t) uid; 2715 syscallarg(gid_t) gid; 2716 } */ 2717 int error; 2718 struct vnode *vp; 2719 2720 error = namei_simple_user(SCARG(uap, path), 2721 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2722 if (error != 0) 2723 return (error); 2724 2725 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2726 2727 vrele(vp); 2728 return (error); 2729 } 2730 2731 /* 2732 * Set ownership given a path name; this version does not follow links. 2733 * Provides POSIX/XPG semantics. 2734 */ 2735 /* ARGSUSED */ 2736 int 2737 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2738 { 2739 /* { 2740 syscallarg(const char *) path; 2741 syscallarg(uid_t) uid; 2742 syscallarg(gid_t) gid; 2743 } */ 2744 int error; 2745 struct vnode *vp; 2746 2747 error = namei_simple_user(SCARG(uap, path), 2748 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2749 if (error != 0) 2750 return (error); 2751 2752 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2753 2754 vrele(vp); 2755 return (error); 2756 } 2757 2758 /* 2759 * Common routine to set ownership given a vnode. 2760 */ 2761 static int 2762 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2763 int posix_semantics) 2764 { 2765 struct vattr vattr; 2766 mode_t newmode; 2767 int error; 2768 2769 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2770 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2771 goto out; 2772 2773 #define CHANGED(x) ((int)(x) != -1) 2774 newmode = vattr.va_mode; 2775 if (posix_semantics) { 2776 /* 2777 * POSIX/XPG semantics: if the caller is not the super-user, 2778 * clear set-user-id and set-group-id bits. Both POSIX and 2779 * the XPG consider the behaviour for calls by the super-user 2780 * implementation-defined; we leave the set-user-id and set- 2781 * group-id settings intact in that case. 2782 */ 2783 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2784 NULL) != 0) 2785 newmode &= ~(S_ISUID | S_ISGID); 2786 } else { 2787 /* 2788 * NetBSD semantics: when changing owner and/or group, 2789 * clear the respective bit(s). 2790 */ 2791 if (CHANGED(uid)) 2792 newmode &= ~S_ISUID; 2793 if (CHANGED(gid)) 2794 newmode &= ~S_ISGID; 2795 } 2796 /* Update va_mode iff altered. */ 2797 if (vattr.va_mode == newmode) 2798 newmode = VNOVAL; 2799 2800 vattr_null(&vattr); 2801 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2802 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2803 vattr.va_mode = newmode; 2804 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2805 #undef CHANGED 2806 2807 out: 2808 VOP_UNLOCK(vp); 2809 return (error); 2810 } 2811 2812 /* 2813 * Set the access and modification times given a path name; this 2814 * version follows links. 2815 */ 2816 /* ARGSUSED */ 2817 int 2818 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 2819 register_t *retval) 2820 { 2821 /* { 2822 syscallarg(const char *) path; 2823 syscallarg(const struct timeval *) tptr; 2824 } */ 2825 2826 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2827 SCARG(uap, tptr), UIO_USERSPACE); 2828 } 2829 2830 /* 2831 * Set the access and modification times given a file descriptor. 2832 */ 2833 /* ARGSUSED */ 2834 int 2835 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 2836 register_t *retval) 2837 { 2838 /* { 2839 syscallarg(int) fd; 2840 syscallarg(const struct timeval *) tptr; 2841 } */ 2842 int error; 2843 file_t *fp; 2844 2845 /* fd_getvnode() will use the descriptor for us */ 2846 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2847 return (error); 2848 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2849 UIO_USERSPACE); 2850 fd_putfile(SCARG(uap, fd)); 2851 return (error); 2852 } 2853 2854 /* 2855 * Set the access and modification times given a path name; this 2856 * version does not follow links. 2857 */ 2858 int 2859 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 2860 register_t *retval) 2861 { 2862 /* { 2863 syscallarg(const char *) path; 2864 syscallarg(const struct timeval *) tptr; 2865 } */ 2866 2867 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2868 SCARG(uap, tptr), UIO_USERSPACE); 2869 } 2870 2871 /* 2872 * Common routine to set access and modification times given a vnode. 2873 */ 2874 int 2875 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2876 const struct timeval *tptr, enum uio_seg seg) 2877 { 2878 struct vattr vattr; 2879 int error, dorele = 0; 2880 namei_simple_flags_t sflags; 2881 2882 bool vanull, setbirthtime; 2883 struct timespec ts[2]; 2884 2885 /* 2886 * I have checked all callers and they pass either FOLLOW, 2887 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 2888 * is 0. More to the point, they don't pass anything else. 2889 * Let's keep it that way at least until the namei interfaces 2890 * are fully sanitized. 2891 */ 2892 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 2893 sflags = (flag == FOLLOW) ? 2894 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 2895 2896 if (tptr == NULL) { 2897 vanull = true; 2898 nanotime(&ts[0]); 2899 ts[1] = ts[0]; 2900 } else { 2901 struct timeval tv[2]; 2902 2903 vanull = false; 2904 if (seg != UIO_SYSSPACE) { 2905 error = copyin(tptr, tv, sizeof (tv)); 2906 if (error != 0) 2907 return error; 2908 tptr = tv; 2909 } 2910 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 2911 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 2912 } 2913 2914 if (vp == NULL) { 2915 /* note: SEG describes TPTR, not PATH; PATH is always user */ 2916 error = namei_simple_user(path, sflags, &vp); 2917 if (error != 0) 2918 return error; 2919 dorele = 1; 2920 } 2921 2922 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2923 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 2924 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 2925 vattr_null(&vattr); 2926 vattr.va_atime = ts[0]; 2927 vattr.va_mtime = ts[1]; 2928 if (setbirthtime) 2929 vattr.va_birthtime = ts[1]; 2930 if (vanull) 2931 vattr.va_vaflags |= VA_UTIMES_NULL; 2932 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2933 VOP_UNLOCK(vp); 2934 2935 if (dorele != 0) 2936 vrele(vp); 2937 2938 return error; 2939 } 2940 2941 /* 2942 * Truncate a file given its path name. 2943 */ 2944 /* ARGSUSED */ 2945 int 2946 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 2947 { 2948 /* { 2949 syscallarg(const char *) path; 2950 syscallarg(int) pad; 2951 syscallarg(off_t) length; 2952 } */ 2953 struct vnode *vp; 2954 struct vattr vattr; 2955 int error; 2956 2957 error = namei_simple_user(SCARG(uap, path), 2958 NSM_FOLLOW_TRYEMULROOT, &vp); 2959 if (error != 0) 2960 return (error); 2961 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2962 if (vp->v_type == VDIR) 2963 error = EISDIR; 2964 else if ((error = vn_writechk(vp)) == 0 && 2965 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 2966 vattr_null(&vattr); 2967 vattr.va_size = SCARG(uap, length); 2968 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2969 } 2970 vput(vp); 2971 return (error); 2972 } 2973 2974 /* 2975 * Truncate a file given a file descriptor. 2976 */ 2977 /* ARGSUSED */ 2978 int 2979 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 2980 { 2981 /* { 2982 syscallarg(int) fd; 2983 syscallarg(int) pad; 2984 syscallarg(off_t) length; 2985 } */ 2986 struct vattr vattr; 2987 struct vnode *vp; 2988 file_t *fp; 2989 int error; 2990 2991 /* fd_getvnode() will use the descriptor for us */ 2992 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2993 return (error); 2994 if ((fp->f_flag & FWRITE) == 0) { 2995 error = EINVAL; 2996 goto out; 2997 } 2998 vp = fp->f_data; 2999 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3000 if (vp->v_type == VDIR) 3001 error = EISDIR; 3002 else if ((error = vn_writechk(vp)) == 0) { 3003 vattr_null(&vattr); 3004 vattr.va_size = SCARG(uap, length); 3005 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3006 } 3007 VOP_UNLOCK(vp); 3008 out: 3009 fd_putfile(SCARG(uap, fd)); 3010 return (error); 3011 } 3012 3013 /* 3014 * Sync an open file. 3015 */ 3016 /* ARGSUSED */ 3017 int 3018 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3019 { 3020 /* { 3021 syscallarg(int) fd; 3022 } */ 3023 struct vnode *vp; 3024 file_t *fp; 3025 int error; 3026 3027 /* fd_getvnode() will use the descriptor for us */ 3028 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3029 return (error); 3030 vp = fp->f_data; 3031 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3032 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3033 VOP_UNLOCK(vp); 3034 fd_putfile(SCARG(uap, fd)); 3035 return (error); 3036 } 3037 3038 /* 3039 * Sync a range of file data. API modeled after that found in AIX. 3040 * 3041 * FDATASYNC indicates that we need only save enough metadata to be able 3042 * to re-read the written data. Note we duplicate AIX's requirement that 3043 * the file be open for writing. 3044 */ 3045 /* ARGSUSED */ 3046 int 3047 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3048 { 3049 /* { 3050 syscallarg(int) fd; 3051 syscallarg(int) flags; 3052 syscallarg(off_t) start; 3053 syscallarg(off_t) length; 3054 } */ 3055 struct vnode *vp; 3056 file_t *fp; 3057 int flags, nflags; 3058 off_t s, e, len; 3059 int error; 3060 3061 /* fd_getvnode() will use the descriptor for us */ 3062 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3063 return (error); 3064 3065 if ((fp->f_flag & FWRITE) == 0) { 3066 error = EBADF; 3067 goto out; 3068 } 3069 3070 flags = SCARG(uap, flags); 3071 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3072 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3073 error = EINVAL; 3074 goto out; 3075 } 3076 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3077 if (flags & FDATASYNC) 3078 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3079 else 3080 nflags = FSYNC_WAIT; 3081 if (flags & FDISKSYNC) 3082 nflags |= FSYNC_CACHE; 3083 3084 len = SCARG(uap, length); 3085 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 3086 if (len) { 3087 s = SCARG(uap, start); 3088 e = s + len; 3089 if (e < s) { 3090 error = EINVAL; 3091 goto out; 3092 } 3093 } else { 3094 e = 0; 3095 s = 0; 3096 } 3097 3098 vp = fp->f_data; 3099 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3100 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3101 VOP_UNLOCK(vp); 3102 out: 3103 fd_putfile(SCARG(uap, fd)); 3104 return (error); 3105 } 3106 3107 /* 3108 * Sync the data of an open file. 3109 */ 3110 /* ARGSUSED */ 3111 int 3112 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3113 { 3114 /* { 3115 syscallarg(int) fd; 3116 } */ 3117 struct vnode *vp; 3118 file_t *fp; 3119 int error; 3120 3121 /* fd_getvnode() will use the descriptor for us */ 3122 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3123 return (error); 3124 if ((fp->f_flag & FWRITE) == 0) { 3125 fd_putfile(SCARG(uap, fd)); 3126 return (EBADF); 3127 } 3128 vp = fp->f_data; 3129 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3130 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3131 VOP_UNLOCK(vp); 3132 fd_putfile(SCARG(uap, fd)); 3133 return (error); 3134 } 3135 3136 /* 3137 * Rename files, (standard) BSD semantics frontend. 3138 */ 3139 /* ARGSUSED */ 3140 int 3141 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3142 { 3143 /* { 3144 syscallarg(const char *) from; 3145 syscallarg(const char *) to; 3146 } */ 3147 3148 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3149 } 3150 3151 /* 3152 * Rename files, POSIX semantics frontend. 3153 */ 3154 /* ARGSUSED */ 3155 int 3156 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3157 { 3158 /* { 3159 syscallarg(const char *) from; 3160 syscallarg(const char *) to; 3161 } */ 3162 3163 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3164 } 3165 3166 /* 3167 * Rename files. Source and destination must either both be directories, 3168 * or both not be directories. If target is a directory, it must be empty. 3169 * If `from' and `to' refer to the same object, the value of the `retain' 3170 * argument is used to determine whether `from' will be 3171 * 3172 * (retain == 0) deleted unless `from' and `to' refer to the same 3173 * object in the file system's name space (BSD). 3174 * (retain == 1) always retained (POSIX). 3175 */ 3176 int 3177 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3178 { 3179 struct vnode *tvp, *fvp, *tdvp; 3180 struct pathbuf *frompb, *topb; 3181 struct nameidata fromnd, tond; 3182 struct mount *fs; 3183 int error; 3184 3185 error = pathbuf_maybe_copyin(from, seg, &frompb); 3186 if (error) { 3187 return error; 3188 } 3189 error = pathbuf_maybe_copyin(to, seg, &topb); 3190 if (error) { 3191 pathbuf_destroy(frompb); 3192 return error; 3193 } 3194 3195 NDINIT(&fromnd, DELETE, LOCKPARENT | TRYEMULROOT | INRENAME, 3196 frompb); 3197 if ((error = namei(&fromnd)) != 0) { 3198 pathbuf_destroy(frompb); 3199 pathbuf_destroy(topb); 3200 return (error); 3201 } 3202 if (fromnd.ni_dvp != fromnd.ni_vp) 3203 VOP_UNLOCK(fromnd.ni_dvp); 3204 fvp = fromnd.ni_vp; 3205 3206 fs = fvp->v_mount; 3207 error = VFS_RENAMELOCK_ENTER(fs); 3208 if (error) { 3209 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3210 vrele(fromnd.ni_dvp); 3211 vrele(fvp); 3212 goto out1; 3213 } 3214 3215 /* 3216 * close, partially, yet another race - ideally we should only 3217 * go as far as getting fromnd.ni_dvp before getting the per-fs 3218 * lock, and then continue to get fromnd.ni_vp, but we can't do 3219 * that with namei as it stands. 3220 * 3221 * This still won't prevent rmdir from nuking fromnd.ni_vp 3222 * under us. The real fix is to get the locks in the right 3223 * order and do the lookups in the right places, but that's a 3224 * major rototill. 3225 * 3226 * Note: this logic (as well as this whole function) is cloned 3227 * in nfs_serv.c. Proceed accordingly. 3228 */ 3229 vrele(fvp); 3230 if ((fromnd.ni_cnd.cn_namelen == 1 && 3231 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3232 (fromnd.ni_cnd.cn_namelen == 2 && 3233 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3234 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3235 error = EINVAL; 3236 VFS_RENAMELOCK_EXIT(fs); 3237 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3238 vrele(fromnd.ni_dvp); 3239 goto out1; 3240 } 3241 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3242 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0); 3243 if (error) { 3244 VOP_UNLOCK(fromnd.ni_dvp); 3245 VFS_RENAMELOCK_EXIT(fs); 3246 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3247 vrele(fromnd.ni_dvp); 3248 goto out1; 3249 } 3250 VOP_UNLOCK(fromnd.ni_vp); 3251 if (fromnd.ni_dvp != fromnd.ni_vp) 3252 VOP_UNLOCK(fromnd.ni_dvp); 3253 fvp = fromnd.ni_vp; 3254 3255 NDINIT(&tond, RENAME, 3256 LOCKPARENT | LOCKLEAF | NOCACHE | TRYEMULROOT 3257 | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0), 3258 topb); 3259 if ((error = namei(&tond)) != 0) { 3260 VFS_RENAMELOCK_EXIT(fs); 3261 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3262 vrele(fromnd.ni_dvp); 3263 vrele(fvp); 3264 goto out1; 3265 } 3266 tdvp = tond.ni_dvp; 3267 tvp = tond.ni_vp; 3268 3269 if (tvp != NULL) { 3270 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3271 error = ENOTDIR; 3272 goto out; 3273 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3274 error = EISDIR; 3275 goto out; 3276 } 3277 } 3278 3279 if (fvp == tdvp) 3280 error = EINVAL; 3281 3282 /* 3283 * Source and destination refer to the same object. 3284 */ 3285 if (fvp == tvp) { 3286 if (retain) 3287 error = -1; 3288 else if (fromnd.ni_dvp == tdvp && 3289 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3290 !memcmp(fromnd.ni_cnd.cn_nameptr, 3291 tond.ni_cnd.cn_nameptr, 3292 fromnd.ni_cnd.cn_namelen)) 3293 error = -1; 3294 } 3295 /* 3296 * Prevent cross-mount operation. 3297 */ 3298 if (error == 0) { 3299 if (tond.ni_dvp->v_mount != fromnd.ni_dvp->v_mount) { 3300 error = EXDEV; 3301 } 3302 } 3303 #if NVERIEXEC > 0 3304 if (!error) { 3305 char *f1, *f2; 3306 size_t f1_len; 3307 size_t f2_len; 3308 3309 f1_len = fromnd.ni_cnd.cn_namelen + 1; 3310 f1 = kmem_alloc(f1_len, KM_SLEEP); 3311 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len); 3312 3313 f2_len = tond.ni_cnd.cn_namelen + 1; 3314 f2 = kmem_alloc(f2_len, KM_SLEEP); 3315 strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len); 3316 3317 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 3318 3319 kmem_free(f1, f1_len); 3320 kmem_free(f2, f2_len); 3321 } 3322 #endif /* NVERIEXEC > 0 */ 3323 3324 out: 3325 if (!error) { 3326 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3327 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3328 VFS_RENAMELOCK_EXIT(fs); 3329 } else { 3330 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3331 if (tdvp == tvp) 3332 vrele(tdvp); 3333 else 3334 vput(tdvp); 3335 if (tvp) 3336 vput(tvp); 3337 VFS_RENAMELOCK_EXIT(fs); 3338 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3339 vrele(fromnd.ni_dvp); 3340 vrele(fvp); 3341 } 3342 out1: 3343 pathbuf_destroy(frompb); 3344 pathbuf_destroy(topb); 3345 return (error == -1 ? 0 : error); 3346 } 3347 3348 /* 3349 * Make a directory file. 3350 */ 3351 /* ARGSUSED */ 3352 int 3353 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3354 { 3355 /* { 3356 syscallarg(const char *) path; 3357 syscallarg(int) mode; 3358 } */ 3359 3360 return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE); 3361 } 3362 3363 int 3364 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 3365 { 3366 struct proc *p = curlwp->l_proc; 3367 struct vnode *vp; 3368 struct vattr vattr; 3369 int error; 3370 struct pathbuf *pb; 3371 struct nameidata nd; 3372 3373 /* XXX bollocks, should pass in a pathbuf */ 3374 error = pathbuf_maybe_copyin(path, seg, &pb); 3375 if (error) { 3376 return error; 3377 } 3378 3379 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 3380 if ((error = namei(&nd)) != 0) { 3381 pathbuf_destroy(pb); 3382 return (error); 3383 } 3384 vp = nd.ni_vp; 3385 if (vp != NULL) { 3386 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3387 if (nd.ni_dvp == vp) 3388 vrele(nd.ni_dvp); 3389 else 3390 vput(nd.ni_dvp); 3391 vrele(vp); 3392 pathbuf_destroy(pb); 3393 return (EEXIST); 3394 } 3395 vattr_null(&vattr); 3396 vattr.va_type = VDIR; 3397 /* We will read cwdi->cwdi_cmask unlocked. */ 3398 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3399 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3400 if (!error) 3401 vput(nd.ni_vp); 3402 pathbuf_destroy(pb); 3403 return (error); 3404 } 3405 3406 /* 3407 * Remove a directory file. 3408 */ 3409 /* ARGSUSED */ 3410 int 3411 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3412 { 3413 /* { 3414 syscallarg(const char *) path; 3415 } */ 3416 struct vnode *vp; 3417 int error; 3418 struct pathbuf *pb; 3419 struct nameidata nd; 3420 3421 error = pathbuf_copyin(SCARG(uap, path), &pb); 3422 if (error) { 3423 return error; 3424 } 3425 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 3426 if ((error = namei(&nd)) != 0) { 3427 pathbuf_destroy(pb); 3428 return error; 3429 } 3430 vp = nd.ni_vp; 3431 if (vp->v_type != VDIR) { 3432 error = ENOTDIR; 3433 goto out; 3434 } 3435 /* 3436 * No rmdir "." please. 3437 */ 3438 if (nd.ni_dvp == vp) { 3439 error = EINVAL; 3440 goto out; 3441 } 3442 /* 3443 * The root of a mounted filesystem cannot be deleted. 3444 */ 3445 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3446 error = EBUSY; 3447 goto out; 3448 } 3449 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3450 pathbuf_destroy(pb); 3451 return (error); 3452 3453 out: 3454 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3455 if (nd.ni_dvp == vp) 3456 vrele(nd.ni_dvp); 3457 else 3458 vput(nd.ni_dvp); 3459 vput(vp); 3460 pathbuf_destroy(pb); 3461 return (error); 3462 } 3463 3464 /* 3465 * Read a block of directory entries in a file system independent format. 3466 */ 3467 int 3468 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3469 { 3470 /* { 3471 syscallarg(int) fd; 3472 syscallarg(char *) buf; 3473 syscallarg(size_t) count; 3474 } */ 3475 file_t *fp; 3476 int error, done; 3477 3478 /* fd_getvnode() will use the descriptor for us */ 3479 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3480 return (error); 3481 if ((fp->f_flag & FREAD) == 0) { 3482 error = EBADF; 3483 goto out; 3484 } 3485 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3486 SCARG(uap, count), &done, l, 0, 0); 3487 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3488 *retval = done; 3489 out: 3490 fd_putfile(SCARG(uap, fd)); 3491 return (error); 3492 } 3493 3494 /* 3495 * Set the mode mask for creation of filesystem nodes. 3496 */ 3497 int 3498 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3499 { 3500 /* { 3501 syscallarg(mode_t) newmask; 3502 } */ 3503 struct proc *p = l->l_proc; 3504 struct cwdinfo *cwdi; 3505 3506 /* 3507 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3508 * important is that we serialize changes to the mask. The 3509 * rw_exit() will issue a write memory barrier on our behalf, 3510 * and force the changes out to other CPUs (as it must use an 3511 * atomic operation, draining the local CPU's store buffers). 3512 */ 3513 cwdi = p->p_cwdi; 3514 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3515 *retval = cwdi->cwdi_cmask; 3516 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3517 rw_exit(&cwdi->cwdi_lock); 3518 3519 return (0); 3520 } 3521 3522 int 3523 dorevoke(struct vnode *vp, kauth_cred_t cred) 3524 { 3525 struct vattr vattr; 3526 int error; 3527 3528 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3529 return error; 3530 if (kauth_cred_geteuid(cred) == vattr.va_uid || 3531 (error = kauth_authorize_generic(cred, 3532 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3533 VOP_REVOKE(vp, REVOKEALL); 3534 return (error); 3535 } 3536 3537 /* 3538 * Void all references to file by ripping underlying filesystem 3539 * away from vnode. 3540 */ 3541 /* ARGSUSED */ 3542 int 3543 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3544 { 3545 /* { 3546 syscallarg(const char *) path; 3547 } */ 3548 struct vnode *vp; 3549 int error; 3550 3551 error = namei_simple_user(SCARG(uap, path), 3552 NSM_FOLLOW_TRYEMULROOT, &vp); 3553 if (error != 0) 3554 return (error); 3555 error = dorevoke(vp, l->l_cred); 3556 vrele(vp); 3557 return (error); 3558 } 3559