vfs_syscalls.c revision 1.461
1 /* $NetBSD: vfs_syscalls.c,v 1.461 2012/11/19 15:01:17 martin Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 /* 69 * Virtual File System System Calls 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.461 2012/11/19 15:01:17 martin Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/fcntl.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/proc.h> 91 #include <sys/uio.h> 92 #include <sys/kmem.h> 93 #include <sys/dirent.h> 94 #include <sys/sysctl.h> 95 #include <sys/syscallargs.h> 96 #include <sys/vfs_syscalls.h> 97 #include <sys/quota.h> 98 #include <sys/quotactl.h> 99 #include <sys/ktrace.h> 100 #ifdef FILEASSOC 101 #include <sys/fileassoc.h> 102 #endif /* FILEASSOC */ 103 #include <sys/extattr.h> 104 #include <sys/verified_exec.h> 105 #include <sys/kauth.h> 106 #include <sys/atomic.h> 107 #include <sys/module.h> 108 #include <sys/buf.h> 109 110 #include <miscfs/genfs/genfs.h> 111 #include <miscfs/syncfs/syncfs.h> 112 #include <miscfs/specfs/specdev.h> 113 114 #include <nfs/rpcv2.h> 115 #include <nfs/nfsproto.h> 116 #include <nfs/nfs.h> 117 #include <nfs/nfs_var.h> 118 119 static int change_flags(struct vnode *, u_long, struct lwp *); 120 static int change_mode(struct vnode *, int, struct lwp *l); 121 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 122 static int do_open(lwp_t *, struct vnode *, struct pathbuf *, int, int, int *); 123 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 124 static int do_sys_mknodat(struct lwp *, int, const char *, mode_t, 125 dev_t, register_t *, enum uio_seg); 126 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 127 enum uio_seg); 128 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 129 static int do_sys_chmodat(struct lwp *, int, const char *, int, int); 130 static int do_sys_chownat(struct lwp *, int, const char *, uid_t, gid_t, int); 131 static int do_sys_utimensat(struct lwp *, int, struct vnode *, 132 const char *, int, const struct timespec *, enum uio_seg); 133 static int do_sys_accessat(struct lwp *, int, const char *, int ,int); 134 static int do_sys_statat(struct lwp *, int, const char *, unsigned int, 135 struct stat *); 136 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 137 enum uio_seg); 138 static int do_sys_linkat(struct lwp *, int, const char *, int, const char *, 139 int, register_t *); 140 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 141 enum uio_seg, int); 142 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 143 size_t, register_t *); 144 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 145 146 static int fd_nameiat(struct lwp *, int, struct nameidata *); 147 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 148 namei_simple_flags_t, struct vnode **); 149 150 151 /* 152 * This table is used to maintain compatibility with 4.3BSD 153 * and NetBSD 0.9 mount syscalls - and possibly other systems. 154 * Note, the order is important! 155 * 156 * Do not modify this table. It should only contain filesystems 157 * supported by NetBSD 0.9 and 4.3BSD. 158 */ 159 const char * const mountcompatnames[] = { 160 NULL, /* 0 = MOUNT_NONE */ 161 MOUNT_FFS, /* 1 = MOUNT_UFS */ 162 MOUNT_NFS, /* 2 */ 163 MOUNT_MFS, /* 3 */ 164 MOUNT_MSDOS, /* 4 */ 165 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 166 MOUNT_FDESC, /* 6 */ 167 MOUNT_KERNFS, /* 7 */ 168 NULL, /* 8 = MOUNT_DEVFS */ 169 MOUNT_AFS, /* 9 */ 170 }; 171 172 const int nmountcompatnames = __arraycount(mountcompatnames); 173 174 static int 175 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 176 { 177 file_t *dfp; 178 int error; 179 180 if (fdat != AT_FDCWD) { 181 if ((error = fd_getvnode(fdat, &dfp)) != 0) 182 goto out; 183 184 if (!(dfp->f_flag & FSEARCH)) { 185 error = VOP_ACCESS(dfp->f_data, VEXEC, l->l_cred); 186 if (error) 187 goto cleanup; 188 } 189 190 NDAT(ndp, dfp->f_data); 191 } 192 193 error = namei(ndp); 194 195 cleanup: 196 if (fdat != AT_FDCWD) 197 fd_putfile(fdat); 198 out: 199 return error; 200 } 201 202 static int 203 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 204 namei_simple_flags_t sflags, struct vnode **vp_ret) 205 { 206 file_t *dfp; 207 struct vnode *dvp; 208 int error; 209 210 if (fdat != AT_FDCWD) { 211 if ((error = fd_getvnode(fdat, &dfp)) != 0) 212 goto out; 213 214 if (!(dfp->f_flag & FSEARCH)) { 215 error = VOP_ACCESS(dfp->f_data, VEXEC, l->l_cred); 216 if (error) 217 goto cleanup; 218 } 219 220 dvp = dfp->f_data; 221 } else { 222 dvp = NULL; 223 } 224 225 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 226 227 cleanup: 228 if (fdat != AT_FDCWD) 229 fd_putfile(fdat); 230 out: 231 return error; 232 } 233 234 static int 235 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 236 { 237 int error; 238 239 fp->f_flag = flags & FMASK; 240 fp->f_type = DTYPE_VNODE; 241 fp->f_ops = &vnops; 242 fp->f_data = vp; 243 244 if (flags & (O_EXLOCK | O_SHLOCK)) { 245 struct flock lf; 246 int type; 247 248 lf.l_whence = SEEK_SET; 249 lf.l_start = 0; 250 lf.l_len = 0; 251 if (flags & O_EXLOCK) 252 lf.l_type = F_WRLCK; 253 else 254 lf.l_type = F_RDLCK; 255 type = F_FLOCK; 256 if ((flags & FNONBLOCK) == 0) 257 type |= F_WAIT; 258 VOP_UNLOCK(vp); 259 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 260 if (error) { 261 (void) vn_close(vp, fp->f_flag, fp->f_cred); 262 fd_abort(l->l_proc, fp, indx); 263 return error; 264 } 265 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 266 atomic_or_uint(&fp->f_flag, FHASLOCK); 267 } 268 if (flags & O_CLOEXEC) 269 fd_set_exclose(l, indx, true); 270 return 0; 271 } 272 273 static int 274 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 275 void *data, size_t *data_len) 276 { 277 struct mount *mp; 278 int error = 0, saved_flags; 279 280 mp = vp->v_mount; 281 saved_flags = mp->mnt_flag; 282 283 /* We can operate only on VV_ROOT nodes. */ 284 if ((vp->v_vflag & VV_ROOT) == 0) { 285 error = EINVAL; 286 goto out; 287 } 288 289 /* 290 * We only allow the filesystem to be reloaded if it 291 * is currently mounted read-only. Additionally, we 292 * prevent read-write to read-only downgrades. 293 */ 294 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 295 (mp->mnt_flag & MNT_RDONLY) == 0 && 296 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 297 error = EOPNOTSUPP; /* Needs translation */ 298 goto out; 299 } 300 301 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 302 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 303 if (error) 304 goto out; 305 306 if (vfs_busy(mp, NULL)) { 307 error = EPERM; 308 goto out; 309 } 310 311 mutex_enter(&mp->mnt_updating); 312 313 mp->mnt_flag &= ~MNT_OP_FLAGS; 314 mp->mnt_flag |= flags & MNT_OP_FLAGS; 315 316 /* 317 * Set the mount level flags. 318 */ 319 if (flags & MNT_RDONLY) 320 mp->mnt_flag |= MNT_RDONLY; 321 else if (mp->mnt_flag & MNT_RDONLY) 322 mp->mnt_iflag |= IMNT_WANTRDWR; 323 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 324 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 325 error = VFS_MOUNT(mp, path, data, data_len); 326 327 if (error && data != NULL) { 328 int error2; 329 330 /* 331 * Update failed; let's try and see if it was an 332 * export request. For compat with 3.0 and earlier. 333 */ 334 error2 = vfs_hooks_reexport(mp, path, data); 335 336 /* 337 * Only update error code if the export request was 338 * understood but some problem occurred while 339 * processing it. 340 */ 341 if (error2 != EJUSTRETURN) 342 error = error2; 343 } 344 345 if (mp->mnt_iflag & IMNT_WANTRDWR) 346 mp->mnt_flag &= ~MNT_RDONLY; 347 if (error) 348 mp->mnt_flag = saved_flags; 349 mp->mnt_flag &= ~MNT_OP_FLAGS; 350 mp->mnt_iflag &= ~IMNT_WANTRDWR; 351 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 352 if (mp->mnt_syncer == NULL) 353 error = vfs_allocate_syncvnode(mp); 354 } else { 355 if (mp->mnt_syncer != NULL) 356 vfs_deallocate_syncvnode(mp); 357 } 358 mutex_exit(&mp->mnt_updating); 359 vfs_unbusy(mp, false, NULL); 360 361 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 362 (flags & MNT_EXTATTR)) { 363 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 364 NULL, 0, NULL) != 0) { 365 printf("%s: failed to start extattr, error = %d", 366 mp->mnt_stat.f_mntonname, error); 367 mp->mnt_flag &= ~MNT_EXTATTR; 368 } 369 } 370 371 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 372 !(flags & MNT_EXTATTR)) { 373 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 374 NULL, 0, NULL) != 0) { 375 printf("%s: failed to stop extattr, error = %d", 376 mp->mnt_stat.f_mntonname, error); 377 mp->mnt_flag |= MNT_RDONLY; 378 } 379 } 380 out: 381 return (error); 382 } 383 384 static int 385 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 386 { 387 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 388 int error; 389 390 /* Copy file-system type from userspace. */ 391 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 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, NULL)) 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, false, NULL); 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, NULL, SCARG(uap, type), 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, struct vfsops *vfsops, const char *type, 477 const char *path, int flags, void *data, enum uio_seg data_seg, 478 size_t data_len, register_t *retval) 479 { 480 struct vnode *vp; 481 void *data_buf = data; 482 bool vfsopsrele = false; 483 int error; 484 485 /* XXX: The calling convention of this routine is totally bizarre */ 486 if (vfsops) 487 vfsopsrele = true; 488 489 /* 490 * Get vnode to be covered 491 */ 492 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 493 if (error != 0) { 494 vp = NULL; 495 goto done; 496 } 497 498 if (vfsops == NULL) { 499 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 500 vfsops = vp->v_mount->mnt_op; 501 } else { 502 /* 'type' is userspace */ 503 error = mount_get_vfsops(type, &vfsops); 504 if (error != 0) 505 goto done; 506 vfsopsrele = true; 507 } 508 } 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 if (data_len > VFS_MAX_MOUNT_DATA) { 515 error = EINVAL; 516 goto done; 517 } 518 /* 519 * Hopefully a longer buffer won't make copyin() fail. 520 * For compatibility with 3.0 and earlier. 521 */ 522 if (flags & MNT_UPDATE 523 && data_len < sizeof (struct mnt_export_args30)) 524 data_len = sizeof (struct mnt_export_args30); 525 } 526 data_buf = kmem_alloc(data_len, KM_SLEEP); 527 528 /* NFS needs the buffer even for mnt_getargs .... */ 529 error = copyin(data, data_buf, data_len); 530 if (error != 0) 531 goto done; 532 } 533 534 if (flags & MNT_GETARGS) { 535 if (data_len == 0) { 536 error = EINVAL; 537 goto done; 538 } 539 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 540 if (error != 0) 541 goto done; 542 if (data_seg == UIO_USERSPACE) 543 error = copyout(data_buf, data, data_len); 544 *retval = data_len; 545 } else if (flags & MNT_UPDATE) { 546 error = mount_update(l, vp, path, flags, data_buf, &data_len); 547 } else { 548 /* Locking is handled internally in mount_domount(). */ 549 KASSERT(vfsopsrele == true); 550 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 551 &data_len); 552 vfsopsrele = false; 553 } 554 555 done: 556 if (vfsopsrele) 557 vfs_delref(vfsops); 558 if (vp != NULL) { 559 vrele(vp); 560 } 561 if (data_buf != data) 562 kmem_free(data_buf, data_len); 563 return (error); 564 } 565 566 /* 567 * Unmount a file system. 568 * 569 * Note: unmount takes a path to the vnode mounted on as argument, 570 * not special file (as before). 571 */ 572 /* ARGSUSED */ 573 int 574 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 575 { 576 /* { 577 syscallarg(const char *) path; 578 syscallarg(int) flags; 579 } */ 580 struct vnode *vp; 581 struct mount *mp; 582 int error; 583 struct pathbuf *pb; 584 struct nameidata nd; 585 586 error = pathbuf_copyin(SCARG(uap, path), &pb); 587 if (error) { 588 return error; 589 } 590 591 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 592 if ((error = namei(&nd)) != 0) { 593 pathbuf_destroy(pb); 594 return error; 595 } 596 vp = nd.ni_vp; 597 pathbuf_destroy(pb); 598 599 mp = vp->v_mount; 600 atomic_inc_uint(&mp->mnt_refcnt); 601 VOP_UNLOCK(vp); 602 603 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 604 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 605 if (error) { 606 vrele(vp); 607 vfs_destroy(mp); 608 return (error); 609 } 610 611 /* 612 * Don't allow unmounting the root file system. 613 */ 614 if (mp->mnt_flag & MNT_ROOTFS) { 615 vrele(vp); 616 vfs_destroy(mp); 617 return (EINVAL); 618 } 619 620 /* 621 * Must be the root of the filesystem 622 */ 623 if ((vp->v_vflag & VV_ROOT) == 0) { 624 vrele(vp); 625 vfs_destroy(mp); 626 return (EINVAL); 627 } 628 629 vrele(vp); 630 error = dounmount(mp, SCARG(uap, flags), l); 631 vfs_destroy(mp); 632 return error; 633 } 634 635 /* 636 * Sync each mounted filesystem. 637 */ 638 #ifdef DEBUG 639 int syncprt = 0; 640 struct ctldebug debug0 = { "syncprt", &syncprt }; 641 #endif 642 643 void 644 do_sys_sync(struct lwp *l) 645 { 646 struct mount *mp, *nmp; 647 int asyncflag; 648 649 mutex_enter(&mountlist_lock); 650 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 651 mp = nmp) { 652 if (vfs_busy(mp, &nmp)) { 653 continue; 654 } 655 mutex_enter(&mp->mnt_updating); 656 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 657 asyncflag = mp->mnt_flag & MNT_ASYNC; 658 mp->mnt_flag &= ~MNT_ASYNC; 659 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 660 if (asyncflag) 661 mp->mnt_flag |= MNT_ASYNC; 662 } 663 mutex_exit(&mp->mnt_updating); 664 vfs_unbusy(mp, false, &nmp); 665 } 666 mutex_exit(&mountlist_lock); 667 #ifdef DEBUG 668 if (syncprt) 669 vfs_bufstats(); 670 #endif /* DEBUG */ 671 } 672 673 /* ARGSUSED */ 674 int 675 sys_sync(struct lwp *l, const void *v, register_t *retval) 676 { 677 do_sys_sync(l); 678 return (0); 679 } 680 681 682 /* 683 * Access or change filesystem quotas. 684 * 685 * (this is really 14 different calls bundled into one) 686 */ 687 688 static int 689 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 690 { 691 struct quotastat info_k; 692 int error; 693 694 /* ensure any padding bytes are cleared */ 695 memset(&info_k, 0, sizeof(info_k)); 696 697 error = vfs_quotactl_stat(mp, &info_k); 698 if (error) { 699 return error; 700 } 701 702 return copyout(&info_k, info_u, sizeof(info_k)); 703 } 704 705 static int 706 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 707 struct quotaidtypestat *info_u) 708 { 709 struct quotaidtypestat info_k; 710 int error; 711 712 /* ensure any padding bytes are cleared */ 713 memset(&info_k, 0, sizeof(info_k)); 714 715 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 716 if (error) { 717 return error; 718 } 719 720 return copyout(&info_k, info_u, sizeof(info_k)); 721 } 722 723 static int 724 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 725 struct quotaobjtypestat *info_u) 726 { 727 struct quotaobjtypestat info_k; 728 int error; 729 730 /* ensure any padding bytes are cleared */ 731 memset(&info_k, 0, sizeof(info_k)); 732 733 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 734 if (error) { 735 return error; 736 } 737 738 return copyout(&info_k, info_u, sizeof(info_k)); 739 } 740 741 static int 742 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 743 struct quotaval *val_u) 744 { 745 struct quotakey key_k; 746 struct quotaval val_k; 747 int error; 748 749 /* ensure any padding bytes are cleared */ 750 memset(&val_k, 0, sizeof(val_k)); 751 752 error = copyin(key_u, &key_k, sizeof(key_k)); 753 if (error) { 754 return error; 755 } 756 757 error = vfs_quotactl_get(mp, &key_k, &val_k); 758 if (error) { 759 return error; 760 } 761 762 return copyout(&val_k, val_u, sizeof(val_k)); 763 } 764 765 static int 766 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 767 const struct quotaval *val_u) 768 { 769 struct quotakey key_k; 770 struct quotaval val_k; 771 int error; 772 773 error = copyin(key_u, &key_k, sizeof(key_k)); 774 if (error) { 775 return error; 776 } 777 778 error = copyin(val_u, &val_k, sizeof(val_k)); 779 if (error) { 780 return error; 781 } 782 783 return vfs_quotactl_put(mp, &key_k, &val_k); 784 } 785 786 static int 787 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u) 788 { 789 struct quotakey key_k; 790 int error; 791 792 error = copyin(key_u, &key_k, sizeof(key_k)); 793 if (error) { 794 return error; 795 } 796 797 return vfs_quotactl_delete(mp, &key_k); 798 } 799 800 static int 801 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 802 { 803 struct quotakcursor cursor_k; 804 int error; 805 806 /* ensure any padding bytes are cleared */ 807 memset(&cursor_k, 0, sizeof(cursor_k)); 808 809 error = vfs_quotactl_cursoropen(mp, &cursor_k); 810 if (error) { 811 return error; 812 } 813 814 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 815 } 816 817 static int 818 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 819 { 820 struct quotakcursor cursor_k; 821 int error; 822 823 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 824 if (error) { 825 return error; 826 } 827 828 return vfs_quotactl_cursorclose(mp, &cursor_k); 829 } 830 831 static int 832 do_sys_quotactl_cursorskipidtype(struct mount *mp, 833 struct quotakcursor *cursor_u, int idtype) 834 { 835 struct quotakcursor cursor_k; 836 int error; 837 838 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 839 if (error) { 840 return error; 841 } 842 843 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 844 if (error) { 845 return error; 846 } 847 848 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 849 } 850 851 static int 852 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 853 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 854 unsigned *ret_u) 855 { 856 #define CGET_STACK_MAX 8 857 struct quotakcursor cursor_k; 858 struct quotakey stackkeys[CGET_STACK_MAX]; 859 struct quotaval stackvals[CGET_STACK_MAX]; 860 struct quotakey *keys_k; 861 struct quotaval *vals_k; 862 unsigned ret_k; 863 int error; 864 865 if (maxnum > 128) { 866 maxnum = 128; 867 } 868 869 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 870 if (error) { 871 return error; 872 } 873 874 if (maxnum <= CGET_STACK_MAX) { 875 keys_k = stackkeys; 876 vals_k = stackvals; 877 /* ensure any padding bytes are cleared */ 878 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 879 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 880 } else { 881 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 882 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 883 } 884 885 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 886 &ret_k); 887 if (error) { 888 goto fail; 889 } 890 891 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 892 if (error) { 893 goto fail; 894 } 895 896 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 897 if (error) { 898 goto fail; 899 } 900 901 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 902 if (error) { 903 goto fail; 904 } 905 906 /* do last to maximize the chance of being able to recover a failure */ 907 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 908 909 fail: 910 if (keys_k != stackkeys) { 911 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 912 } 913 if (vals_k != stackvals) { 914 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 915 } 916 return error; 917 } 918 919 static int 920 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 921 int *ret_u) 922 { 923 struct quotakcursor cursor_k; 924 int ret_k; 925 int error; 926 927 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 928 if (error) { 929 return error; 930 } 931 932 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 933 if (error) { 934 return error; 935 } 936 937 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 938 if (error) { 939 return error; 940 } 941 942 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 943 } 944 945 static int 946 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 947 { 948 struct quotakcursor cursor_k; 949 int error; 950 951 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 952 if (error) { 953 return error; 954 } 955 956 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 957 if (error) { 958 return error; 959 } 960 961 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 962 } 963 964 static int 965 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 966 { 967 char *path_k; 968 int error; 969 970 /* XXX this should probably be a struct pathbuf */ 971 path_k = PNBUF_GET(); 972 error = copyin(path_u, path_k, PATH_MAX); 973 if (error) { 974 PNBUF_PUT(path_k); 975 return error; 976 } 977 978 error = vfs_quotactl_quotaon(mp, idtype, path_k); 979 980 PNBUF_PUT(path_k); 981 return error; 982 } 983 984 static int 985 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 986 { 987 return vfs_quotactl_quotaoff(mp, idtype); 988 } 989 990 int 991 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 992 { 993 struct mount *mp; 994 struct vnode *vp; 995 int error; 996 997 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 998 if (error != 0) 999 return (error); 1000 mp = vp->v_mount; 1001 1002 switch (args->qc_op) { 1003 case QUOTACTL_STAT: 1004 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 1005 break; 1006 case QUOTACTL_IDTYPESTAT: 1007 error = do_sys_quotactl_idtypestat(mp, 1008 args->u.idtypestat.qc_idtype, 1009 args->u.idtypestat.qc_info); 1010 break; 1011 case QUOTACTL_OBJTYPESTAT: 1012 error = do_sys_quotactl_objtypestat(mp, 1013 args->u.objtypestat.qc_objtype, 1014 args->u.objtypestat.qc_info); 1015 break; 1016 case QUOTACTL_GET: 1017 error = do_sys_quotactl_get(mp, 1018 args->u.get.qc_key, 1019 args->u.get.qc_val); 1020 break; 1021 case QUOTACTL_PUT: 1022 error = do_sys_quotactl_put(mp, 1023 args->u.put.qc_key, 1024 args->u.put.qc_val); 1025 break; 1026 case QUOTACTL_DELETE: 1027 error = do_sys_quotactl_delete(mp, args->u.delete.qc_key); 1028 break; 1029 case QUOTACTL_CURSOROPEN: 1030 error = do_sys_quotactl_cursoropen(mp, 1031 args->u.cursoropen.qc_cursor); 1032 break; 1033 case QUOTACTL_CURSORCLOSE: 1034 error = do_sys_quotactl_cursorclose(mp, 1035 args->u.cursorclose.qc_cursor); 1036 break; 1037 case QUOTACTL_CURSORSKIPIDTYPE: 1038 error = do_sys_quotactl_cursorskipidtype(mp, 1039 args->u.cursorskipidtype.qc_cursor, 1040 args->u.cursorskipidtype.qc_idtype); 1041 break; 1042 case QUOTACTL_CURSORGET: 1043 error = do_sys_quotactl_cursorget(mp, 1044 args->u.cursorget.qc_cursor, 1045 args->u.cursorget.qc_keys, 1046 args->u.cursorget.qc_vals, 1047 args->u.cursorget.qc_maxnum, 1048 args->u.cursorget.qc_ret); 1049 break; 1050 case QUOTACTL_CURSORATEND: 1051 error = do_sys_quotactl_cursoratend(mp, 1052 args->u.cursoratend.qc_cursor, 1053 args->u.cursoratend.qc_ret); 1054 break; 1055 case QUOTACTL_CURSORREWIND: 1056 error = do_sys_quotactl_cursorrewind(mp, 1057 args->u.cursorrewind.qc_cursor); 1058 break; 1059 case QUOTACTL_QUOTAON: 1060 error = do_sys_quotactl_quotaon(mp, 1061 args->u.quotaon.qc_idtype, 1062 args->u.quotaon.qc_quotafile); 1063 break; 1064 case QUOTACTL_QUOTAOFF: 1065 error = do_sys_quotactl_quotaoff(mp, 1066 args->u.quotaoff.qc_idtype); 1067 break; 1068 default: 1069 error = EINVAL; 1070 break; 1071 } 1072 1073 vrele(vp); 1074 return error; 1075 } 1076 1077 /* ARGSUSED */ 1078 int 1079 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1080 register_t *retval) 1081 { 1082 /* { 1083 syscallarg(const char *) path; 1084 syscallarg(struct quotactl_args *) args; 1085 } */ 1086 struct quotactl_args args; 1087 int error; 1088 1089 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1090 if (error) { 1091 return error; 1092 } 1093 1094 return do_sys_quotactl(SCARG(uap, path), &args); 1095 } 1096 1097 int 1098 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1099 int root) 1100 { 1101 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1102 int error = 0; 1103 1104 /* 1105 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1106 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1107 * overrides MNT_NOWAIT. 1108 */ 1109 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1110 (flags != MNT_WAIT && flags != 0)) { 1111 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1112 goto done; 1113 } 1114 1115 /* Get the filesystem stats now */ 1116 memset(sp, 0, sizeof(*sp)); 1117 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1118 return error; 1119 } 1120 1121 if (cwdi->cwdi_rdir == NULL) 1122 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1123 done: 1124 if (cwdi->cwdi_rdir != NULL) { 1125 size_t len; 1126 char *bp; 1127 char c; 1128 char *path = PNBUF_GET(); 1129 1130 bp = path + MAXPATHLEN; 1131 *--bp = '\0'; 1132 rw_enter(&cwdi->cwdi_lock, RW_READER); 1133 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1134 MAXPATHLEN / 2, 0, l); 1135 rw_exit(&cwdi->cwdi_lock); 1136 if (error) { 1137 PNBUF_PUT(path); 1138 return error; 1139 } 1140 len = strlen(bp); 1141 if (len != 1) { 1142 /* 1143 * for mount points that are below our root, we can see 1144 * them, so we fix up the pathname and return them. The 1145 * rest we cannot see, so we don't allow viewing the 1146 * data. 1147 */ 1148 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1149 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1150 (void)strlcpy(sp->f_mntonname, 1151 c == '\0' ? "/" : &sp->f_mntonname[len], 1152 sizeof(sp->f_mntonname)); 1153 } else { 1154 if (root) 1155 (void)strlcpy(sp->f_mntonname, "/", 1156 sizeof(sp->f_mntonname)); 1157 else 1158 error = EPERM; 1159 } 1160 } 1161 PNBUF_PUT(path); 1162 } 1163 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1164 return error; 1165 } 1166 1167 /* 1168 * Get filesystem statistics by path. 1169 */ 1170 int 1171 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1172 { 1173 struct mount *mp; 1174 int error; 1175 struct vnode *vp; 1176 1177 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1178 if (error != 0) 1179 return error; 1180 mp = vp->v_mount; 1181 error = dostatvfs(mp, sb, l, flags, 1); 1182 vrele(vp); 1183 return error; 1184 } 1185 1186 /* ARGSUSED */ 1187 int 1188 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1189 { 1190 /* { 1191 syscallarg(const char *) path; 1192 syscallarg(struct statvfs *) buf; 1193 syscallarg(int) flags; 1194 } */ 1195 struct statvfs *sb; 1196 int error; 1197 1198 sb = STATVFSBUF_GET(); 1199 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1200 if (error == 0) 1201 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1202 STATVFSBUF_PUT(sb); 1203 return error; 1204 } 1205 1206 /* 1207 * Get filesystem statistics by fd. 1208 */ 1209 int 1210 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1211 { 1212 file_t *fp; 1213 struct mount *mp; 1214 int error; 1215 1216 /* fd_getvnode() will use the descriptor for us */ 1217 if ((error = fd_getvnode(fd, &fp)) != 0) 1218 return (error); 1219 mp = ((struct vnode *)fp->f_data)->v_mount; 1220 error = dostatvfs(mp, sb, curlwp, flags, 1); 1221 fd_putfile(fd); 1222 return error; 1223 } 1224 1225 /* ARGSUSED */ 1226 int 1227 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1228 { 1229 /* { 1230 syscallarg(int) fd; 1231 syscallarg(struct statvfs *) buf; 1232 syscallarg(int) flags; 1233 } */ 1234 struct statvfs *sb; 1235 int error; 1236 1237 sb = STATVFSBUF_GET(); 1238 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1239 if (error == 0) 1240 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1241 STATVFSBUF_PUT(sb); 1242 return error; 1243 } 1244 1245 1246 /* 1247 * Get statistics on all filesystems. 1248 */ 1249 int 1250 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1251 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1252 register_t *retval) 1253 { 1254 int root = 0; 1255 struct proc *p = l->l_proc; 1256 struct mount *mp, *nmp; 1257 struct statvfs *sb; 1258 size_t count, maxcount; 1259 int error = 0; 1260 1261 sb = STATVFSBUF_GET(); 1262 maxcount = bufsize / entry_sz; 1263 mutex_enter(&mountlist_lock); 1264 count = 0; 1265 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1266 mp = nmp) { 1267 if (vfs_busy(mp, &nmp)) { 1268 continue; 1269 } 1270 if (sfsp && count < maxcount) { 1271 error = dostatvfs(mp, sb, l, flags, 0); 1272 if (error) { 1273 vfs_unbusy(mp, false, &nmp); 1274 error = 0; 1275 continue; 1276 } 1277 error = copyfn(sb, sfsp, entry_sz); 1278 if (error) { 1279 vfs_unbusy(mp, false, NULL); 1280 goto out; 1281 } 1282 sfsp = (char *)sfsp + entry_sz; 1283 root |= strcmp(sb->f_mntonname, "/") == 0; 1284 } 1285 count++; 1286 vfs_unbusy(mp, false, &nmp); 1287 } 1288 mutex_exit(&mountlist_lock); 1289 1290 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1291 /* 1292 * fake a root entry 1293 */ 1294 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1295 sb, l, flags, 1); 1296 if (error != 0) 1297 goto out; 1298 if (sfsp) { 1299 error = copyfn(sb, sfsp, entry_sz); 1300 if (error != 0) 1301 goto out; 1302 } 1303 count++; 1304 } 1305 if (sfsp && count > maxcount) 1306 *retval = maxcount; 1307 else 1308 *retval = count; 1309 out: 1310 STATVFSBUF_PUT(sb); 1311 return error; 1312 } 1313 1314 int 1315 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1316 { 1317 /* { 1318 syscallarg(struct statvfs *) buf; 1319 syscallarg(size_t) bufsize; 1320 syscallarg(int) flags; 1321 } */ 1322 1323 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1324 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1325 } 1326 1327 /* 1328 * Change current working directory to a given file descriptor. 1329 */ 1330 /* ARGSUSED */ 1331 int 1332 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1333 { 1334 /* { 1335 syscallarg(int) fd; 1336 } */ 1337 struct proc *p = l->l_proc; 1338 struct cwdinfo *cwdi; 1339 struct vnode *vp, *tdp; 1340 struct mount *mp; 1341 file_t *fp; 1342 int error, fd; 1343 1344 /* fd_getvnode() will use the descriptor for us */ 1345 fd = SCARG(uap, fd); 1346 if ((error = fd_getvnode(fd, &fp)) != 0) 1347 return (error); 1348 vp = fp->f_data; 1349 1350 vref(vp); 1351 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1352 if (vp->v_type != VDIR) 1353 error = ENOTDIR; 1354 else 1355 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1356 if (error) { 1357 vput(vp); 1358 goto out; 1359 } 1360 while ((mp = vp->v_mountedhere) != NULL) { 1361 error = vfs_busy(mp, NULL); 1362 vput(vp); 1363 if (error != 0) 1364 goto out; 1365 error = VFS_ROOT(mp, &tdp); 1366 vfs_unbusy(mp, false, NULL); 1367 if (error) 1368 goto out; 1369 vp = tdp; 1370 } 1371 VOP_UNLOCK(vp); 1372 1373 /* 1374 * Disallow changing to a directory not under the process's 1375 * current root directory (if there is one). 1376 */ 1377 cwdi = p->p_cwdi; 1378 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1379 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1380 vrele(vp); 1381 error = EPERM; /* operation not permitted */ 1382 } else { 1383 vrele(cwdi->cwdi_cdir); 1384 cwdi->cwdi_cdir = vp; 1385 } 1386 rw_exit(&cwdi->cwdi_lock); 1387 1388 out: 1389 fd_putfile(fd); 1390 return (error); 1391 } 1392 1393 /* 1394 * Change this process's notion of the root directory to a given file 1395 * descriptor. 1396 */ 1397 int 1398 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1399 { 1400 struct proc *p = l->l_proc; 1401 struct vnode *vp; 1402 file_t *fp; 1403 int error, fd = SCARG(uap, fd); 1404 1405 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1406 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1407 return error; 1408 /* fd_getvnode() will use the descriptor for us */ 1409 if ((error = fd_getvnode(fd, &fp)) != 0) 1410 return error; 1411 vp = fp->f_data; 1412 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1413 if (vp->v_type != VDIR) 1414 error = ENOTDIR; 1415 else 1416 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1417 VOP_UNLOCK(vp); 1418 if (error) 1419 goto out; 1420 vref(vp); 1421 1422 change_root(p->p_cwdi, vp, l); 1423 1424 out: 1425 fd_putfile(fd); 1426 return (error); 1427 } 1428 1429 /* 1430 * Change current working directory (``.''). 1431 */ 1432 /* ARGSUSED */ 1433 int 1434 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1435 { 1436 /* { 1437 syscallarg(const char *) path; 1438 } */ 1439 struct proc *p = l->l_proc; 1440 struct cwdinfo *cwdi; 1441 int error; 1442 struct vnode *vp; 1443 1444 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1445 &vp, l)) != 0) 1446 return (error); 1447 cwdi = p->p_cwdi; 1448 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1449 vrele(cwdi->cwdi_cdir); 1450 cwdi->cwdi_cdir = vp; 1451 rw_exit(&cwdi->cwdi_lock); 1452 return (0); 1453 } 1454 1455 /* 1456 * Change notion of root (``/'') directory. 1457 */ 1458 /* ARGSUSED */ 1459 int 1460 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1461 { 1462 /* { 1463 syscallarg(const char *) path; 1464 } */ 1465 struct proc *p = l->l_proc; 1466 int error; 1467 struct vnode *vp; 1468 1469 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1470 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1471 return (error); 1472 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1473 &vp, l)) != 0) 1474 return (error); 1475 1476 change_root(p->p_cwdi, vp, l); 1477 1478 return (0); 1479 } 1480 1481 /* 1482 * Common routine for chroot and fchroot. 1483 * NB: callers need to properly authorize the change root operation. 1484 */ 1485 void 1486 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1487 { 1488 struct proc *p = l->l_proc; 1489 kauth_cred_t ncred; 1490 1491 ncred = kauth_cred_alloc(); 1492 1493 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1494 if (cwdi->cwdi_rdir != NULL) 1495 vrele(cwdi->cwdi_rdir); 1496 cwdi->cwdi_rdir = vp; 1497 1498 /* 1499 * Prevent escaping from chroot by putting the root under 1500 * the working directory. Silently chdir to / if we aren't 1501 * already there. 1502 */ 1503 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1504 /* 1505 * XXX would be more failsafe to change directory to a 1506 * deadfs node here instead 1507 */ 1508 vrele(cwdi->cwdi_cdir); 1509 vref(vp); 1510 cwdi->cwdi_cdir = vp; 1511 } 1512 rw_exit(&cwdi->cwdi_lock); 1513 1514 /* Get a write lock on the process credential. */ 1515 proc_crmod_enter(); 1516 1517 kauth_cred_clone(p->p_cred, ncred); 1518 kauth_proc_chroot(ncred, p->p_cwdi); 1519 1520 /* Broadcast our credentials to the process and other LWPs. */ 1521 proc_crmod_leave(ncred, p->p_cred, true); 1522 } 1523 1524 /* 1525 * Common routine for chroot and chdir. 1526 * XXX "where" should be enum uio_seg 1527 */ 1528 int 1529 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1530 { 1531 struct pathbuf *pb; 1532 struct nameidata nd; 1533 int error; 1534 1535 error = pathbuf_maybe_copyin(path, where, &pb); 1536 if (error) { 1537 return error; 1538 } 1539 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1540 if ((error = namei(&nd)) != 0) { 1541 pathbuf_destroy(pb); 1542 return error; 1543 } 1544 *vpp = nd.ni_vp; 1545 pathbuf_destroy(pb); 1546 1547 if ((*vpp)->v_type != VDIR) 1548 error = ENOTDIR; 1549 else 1550 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1551 1552 if (error) 1553 vput(*vpp); 1554 else 1555 VOP_UNLOCK(*vpp); 1556 return (error); 1557 } 1558 1559 /* 1560 * Internals of sys_open - path has already been converted into a pathbuf 1561 * (so we can easily reuse this function from other parts of the kernel, 1562 * like posix_spawn post-processing). 1563 */ 1564 static int 1565 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1566 int open_mode, int *fd) 1567 { 1568 struct proc *p = l->l_proc; 1569 struct cwdinfo *cwdi = p->p_cwdi; 1570 file_t *fp; 1571 struct vnode *vp; 1572 int flags, cmode; 1573 int indx, error; 1574 struct nameidata nd; 1575 1576 flags = FFLAGS(open_flags); 1577 if ((flags & (FREAD | FWRITE)) == 0) 1578 return EINVAL; 1579 1580 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1581 return error; 1582 } 1583 1584 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1585 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1586 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1587 if (dvp != NULL) 1588 NDAT(&nd, dvp); 1589 1590 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1591 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1592 fd_abort(p, fp, indx); 1593 if ((error == EDUPFD || error == EMOVEFD) && 1594 l->l_dupfd >= 0 && /* XXX from fdopen */ 1595 (error = 1596 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1597 *fd = indx; 1598 return 0; 1599 } 1600 if (error == ERESTART) 1601 error = EINTR; 1602 return error; 1603 } 1604 1605 l->l_dupfd = 0; 1606 vp = nd.ni_vp; 1607 1608 if ((error = open_setfp(l, fp, vp, indx, flags))) 1609 return error; 1610 1611 VOP_UNLOCK(vp); 1612 *fd = indx; 1613 fd_affix(p, fp, indx); 1614 return 0; 1615 } 1616 1617 int 1618 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1619 { 1620 struct pathbuf *pb; 1621 int error, oflags; 1622 1623 oflags = FFLAGS(open_flags); 1624 if ((oflags & (FREAD | FWRITE)) == 0) 1625 return EINVAL; 1626 1627 pb = pathbuf_create(path); 1628 if (pb == NULL) 1629 return ENOMEM; 1630 1631 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1632 pathbuf_destroy(pb); 1633 1634 return error; 1635 } 1636 1637 /* 1638 * Check permissions, allocate an open file structure, 1639 * and call the device open routine if any. 1640 * XXX implement O_SEARCH 1641 */ 1642 static int 1643 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1644 int mode, int *fd) 1645 { 1646 file_t *dfp = NULL; 1647 struct vnode *dvp = NULL; 1648 struct pathbuf *pb; 1649 int error; 1650 1651 error = pathbuf_copyin(path, &pb); 1652 if (error) 1653 return error; 1654 1655 if (fdat != AT_FDCWD) { 1656 /* fd_getvnode() will use the descriptor for us */ 1657 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1658 goto out; 1659 1660 dvp = dfp->f_data; 1661 1662 if (!(dfp->f_flag & FSEARCH)) { 1663 error = VOP_ACCESS(dfp->f_data, VEXEC, l->l_cred); 1664 if (error) 1665 goto cleanup; 1666 } 1667 } 1668 1669 error = do_open(l, dvp, pb, flags, mode, fd); 1670 1671 cleanup: 1672 if (dfp != NULL) 1673 fd_putfile(fdat); 1674 out: 1675 pathbuf_destroy(pb); 1676 return error; 1677 } 1678 1679 int 1680 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1681 { 1682 /* { 1683 syscallarg(const char *) path; 1684 syscallarg(int) flags; 1685 syscallarg(int) mode; 1686 } */ 1687 int error; 1688 int fd; 1689 1690 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1691 SCARG(uap, flags), SCARG(uap, mode), &fd); 1692 1693 if (error == 0) 1694 *retval = fd; 1695 1696 return error; 1697 } 1698 1699 int 1700 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1701 { 1702 /* { 1703 syscallarg(int) fd; 1704 syscallarg(const char *) path; 1705 syscallarg(int) oflags; 1706 syscallarg(int) mode; 1707 } */ 1708 int error; 1709 int fd; 1710 1711 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1712 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1713 1714 if (error == 0) 1715 *retval = fd; 1716 1717 return error; 1718 } 1719 1720 static void 1721 vfs__fhfree(fhandle_t *fhp) 1722 { 1723 size_t fhsize; 1724 1725 if (fhp == NULL) { 1726 return; 1727 } 1728 fhsize = FHANDLE_SIZE(fhp); 1729 kmem_free(fhp, fhsize); 1730 } 1731 1732 /* 1733 * vfs_composefh: compose a filehandle. 1734 */ 1735 1736 int 1737 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1738 { 1739 struct mount *mp; 1740 struct fid *fidp; 1741 int error; 1742 size_t needfhsize; 1743 size_t fidsize; 1744 1745 mp = vp->v_mount; 1746 fidp = NULL; 1747 if (*fh_size < FHANDLE_SIZE_MIN) { 1748 fidsize = 0; 1749 } else { 1750 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1751 if (fhp != NULL) { 1752 memset(fhp, 0, *fh_size); 1753 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1754 fidp = &fhp->fh_fid; 1755 } 1756 } 1757 error = VFS_VPTOFH(vp, fidp, &fidsize); 1758 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1759 if (error == 0 && *fh_size < needfhsize) { 1760 error = E2BIG; 1761 } 1762 *fh_size = needfhsize; 1763 return error; 1764 } 1765 1766 int 1767 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1768 { 1769 struct mount *mp; 1770 fhandle_t *fhp; 1771 size_t fhsize; 1772 size_t fidsize; 1773 int error; 1774 1775 *fhpp = NULL; 1776 mp = vp->v_mount; 1777 fidsize = 0; 1778 error = VFS_VPTOFH(vp, NULL, &fidsize); 1779 KASSERT(error != 0); 1780 if (error != E2BIG) { 1781 goto out; 1782 } 1783 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1784 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1785 if (fhp == NULL) { 1786 error = ENOMEM; 1787 goto out; 1788 } 1789 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1790 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1791 if (error == 0) { 1792 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1793 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1794 *fhpp = fhp; 1795 } else { 1796 kmem_free(fhp, fhsize); 1797 } 1798 out: 1799 return error; 1800 } 1801 1802 void 1803 vfs_composefh_free(fhandle_t *fhp) 1804 { 1805 1806 vfs__fhfree(fhp); 1807 } 1808 1809 /* 1810 * vfs_fhtovp: lookup a vnode by a filehandle. 1811 */ 1812 1813 int 1814 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1815 { 1816 struct mount *mp; 1817 int error; 1818 1819 *vpp = NULL; 1820 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1821 if (mp == NULL) { 1822 error = ESTALE; 1823 goto out; 1824 } 1825 if (mp->mnt_op->vfs_fhtovp == NULL) { 1826 error = EOPNOTSUPP; 1827 goto out; 1828 } 1829 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1830 out: 1831 return error; 1832 } 1833 1834 /* 1835 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1836 * the needed size. 1837 */ 1838 1839 int 1840 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1841 { 1842 fhandle_t *fhp; 1843 int error; 1844 1845 *fhpp = NULL; 1846 if (fhsize > FHANDLE_SIZE_MAX) { 1847 return EINVAL; 1848 } 1849 if (fhsize < FHANDLE_SIZE_MIN) { 1850 return EINVAL; 1851 } 1852 again: 1853 fhp = kmem_alloc(fhsize, KM_SLEEP); 1854 if (fhp == NULL) { 1855 return ENOMEM; 1856 } 1857 error = copyin(ufhp, fhp, fhsize); 1858 if (error == 0) { 1859 /* XXX this check shouldn't be here */ 1860 if (FHANDLE_SIZE(fhp) == fhsize) { 1861 *fhpp = fhp; 1862 return 0; 1863 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1864 /* 1865 * a kludge for nfsv2 padded handles. 1866 */ 1867 size_t sz; 1868 1869 sz = FHANDLE_SIZE(fhp); 1870 kmem_free(fhp, fhsize); 1871 fhsize = sz; 1872 goto again; 1873 } else { 1874 /* 1875 * userland told us wrong size. 1876 */ 1877 error = EINVAL; 1878 } 1879 } 1880 kmem_free(fhp, fhsize); 1881 return error; 1882 } 1883 1884 void 1885 vfs_copyinfh_free(fhandle_t *fhp) 1886 { 1887 1888 vfs__fhfree(fhp); 1889 } 1890 1891 /* 1892 * Get file handle system call 1893 */ 1894 int 1895 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1896 { 1897 /* { 1898 syscallarg(char *) fname; 1899 syscallarg(fhandle_t *) fhp; 1900 syscallarg(size_t *) fh_size; 1901 } */ 1902 struct vnode *vp; 1903 fhandle_t *fh; 1904 int error; 1905 struct pathbuf *pb; 1906 struct nameidata nd; 1907 size_t sz; 1908 size_t usz; 1909 1910 /* 1911 * Must be super user 1912 */ 1913 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1914 0, NULL, NULL, NULL); 1915 if (error) 1916 return (error); 1917 1918 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1919 if (error) { 1920 return error; 1921 } 1922 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1923 error = namei(&nd); 1924 if (error) { 1925 pathbuf_destroy(pb); 1926 return error; 1927 } 1928 vp = nd.ni_vp; 1929 pathbuf_destroy(pb); 1930 1931 error = vfs_composefh_alloc(vp, &fh); 1932 vput(vp); 1933 if (error != 0) { 1934 goto out; 1935 } 1936 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1937 if (error != 0) { 1938 goto out; 1939 } 1940 sz = FHANDLE_SIZE(fh); 1941 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1942 if (error != 0) { 1943 goto out; 1944 } 1945 if (usz >= sz) { 1946 error = copyout(fh, SCARG(uap, fhp), sz); 1947 } else { 1948 error = E2BIG; 1949 } 1950 out: 1951 vfs_composefh_free(fh); 1952 return (error); 1953 } 1954 1955 /* 1956 * Open a file given a file handle. 1957 * 1958 * Check permissions, allocate an open file structure, 1959 * and call the device open routine if any. 1960 */ 1961 1962 int 1963 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1964 register_t *retval) 1965 { 1966 file_t *fp; 1967 struct vnode *vp = NULL; 1968 kauth_cred_t cred = l->l_cred; 1969 file_t *nfp; 1970 int indx, error = 0; 1971 struct vattr va; 1972 fhandle_t *fh; 1973 int flags; 1974 proc_t *p; 1975 1976 p = curproc; 1977 1978 /* 1979 * Must be super user 1980 */ 1981 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1982 0, NULL, NULL, NULL))) 1983 return (error); 1984 1985 flags = FFLAGS(oflags); 1986 if ((flags & (FREAD | FWRITE)) == 0) 1987 return (EINVAL); 1988 if ((flags & O_CREAT)) 1989 return (EINVAL); 1990 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1991 return (error); 1992 fp = nfp; 1993 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1994 if (error != 0) { 1995 goto bad; 1996 } 1997 error = vfs_fhtovp(fh, &vp); 1998 if (error != 0) { 1999 goto bad; 2000 } 2001 2002 /* Now do an effective vn_open */ 2003 2004 if (vp->v_type == VSOCK) { 2005 error = EOPNOTSUPP; 2006 goto bad; 2007 } 2008 error = vn_openchk(vp, cred, flags); 2009 if (error != 0) 2010 goto bad; 2011 if (flags & O_TRUNC) { 2012 VOP_UNLOCK(vp); /* XXX */ 2013 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2014 vattr_null(&va); 2015 va.va_size = 0; 2016 error = VOP_SETATTR(vp, &va, cred); 2017 if (error) 2018 goto bad; 2019 } 2020 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2021 goto bad; 2022 if (flags & FWRITE) { 2023 mutex_enter(vp->v_interlock); 2024 vp->v_writecount++; 2025 mutex_exit(vp->v_interlock); 2026 } 2027 2028 /* done with modified vn_open, now finish what sys_open does. */ 2029 if ((error = open_setfp(l, fp, vp, indx, flags))) 2030 return error; 2031 2032 VOP_UNLOCK(vp); 2033 *retval = indx; 2034 fd_affix(p, fp, indx); 2035 vfs_copyinfh_free(fh); 2036 return (0); 2037 2038 bad: 2039 fd_abort(p, fp, indx); 2040 if (vp != NULL) 2041 vput(vp); 2042 vfs_copyinfh_free(fh); 2043 return (error); 2044 } 2045 2046 int 2047 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2048 { 2049 /* { 2050 syscallarg(const void *) fhp; 2051 syscallarg(size_t) fh_size; 2052 syscallarg(int) flags; 2053 } */ 2054 2055 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2056 SCARG(uap, flags), retval); 2057 } 2058 2059 int 2060 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2061 { 2062 int error; 2063 fhandle_t *fh; 2064 struct vnode *vp; 2065 2066 /* 2067 * Must be super user 2068 */ 2069 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2070 0, NULL, NULL, NULL))) 2071 return (error); 2072 2073 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2074 if (error != 0) 2075 return error; 2076 2077 error = vfs_fhtovp(fh, &vp); 2078 vfs_copyinfh_free(fh); 2079 if (error != 0) 2080 return error; 2081 2082 error = vn_stat(vp, sb); 2083 vput(vp); 2084 return error; 2085 } 2086 2087 2088 /* ARGSUSED */ 2089 int 2090 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2091 { 2092 /* { 2093 syscallarg(const void *) fhp; 2094 syscallarg(size_t) fh_size; 2095 syscallarg(struct stat *) sb; 2096 } */ 2097 struct stat sb; 2098 int error; 2099 2100 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2101 if (error) 2102 return error; 2103 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2104 } 2105 2106 int 2107 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2108 int flags) 2109 { 2110 fhandle_t *fh; 2111 struct mount *mp; 2112 struct vnode *vp; 2113 int error; 2114 2115 /* 2116 * Must be super user 2117 */ 2118 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2119 0, NULL, NULL, NULL))) 2120 return error; 2121 2122 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2123 if (error != 0) 2124 return error; 2125 2126 error = vfs_fhtovp(fh, &vp); 2127 vfs_copyinfh_free(fh); 2128 if (error != 0) 2129 return error; 2130 2131 mp = vp->v_mount; 2132 error = dostatvfs(mp, sb, l, flags, 1); 2133 vput(vp); 2134 return error; 2135 } 2136 2137 /* ARGSUSED */ 2138 int 2139 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2140 { 2141 /* { 2142 syscallarg(const void *) fhp; 2143 syscallarg(size_t) fh_size; 2144 syscallarg(struct statvfs *) buf; 2145 syscallarg(int) flags; 2146 } */ 2147 struct statvfs *sb = STATVFSBUF_GET(); 2148 int error; 2149 2150 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2151 SCARG(uap, flags)); 2152 if (error == 0) 2153 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2154 STATVFSBUF_PUT(sb); 2155 return error; 2156 } 2157 2158 /* 2159 * Create a special file. 2160 */ 2161 /* ARGSUSED */ 2162 int 2163 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2164 register_t *retval) 2165 { 2166 /* { 2167 syscallarg(const char *) path; 2168 syscallarg(mode_t) mode; 2169 syscallarg(dev_t) dev; 2170 } */ 2171 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2172 SCARG(uap, dev), retval, UIO_USERSPACE); 2173 } 2174 2175 int 2176 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2177 register_t *retval) 2178 { 2179 /* { 2180 syscallarg(int) fd; 2181 syscallarg(const char *) path; 2182 syscallarg(mode_t) mode; 2183 syscallarg(uint32_t) dev; 2184 } */ 2185 2186 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2187 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2188 } 2189 2190 int 2191 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2192 register_t *retval, enum uio_seg seg) 2193 { 2194 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2195 } 2196 2197 int 2198 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2199 dev_t dev, register_t *retval, enum uio_seg seg) 2200 { 2201 struct proc *p = l->l_proc; 2202 struct vnode *vp; 2203 struct vattr vattr; 2204 int error, optype; 2205 struct pathbuf *pb; 2206 struct nameidata nd; 2207 const char *pathstring; 2208 2209 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2210 0, NULL, NULL, NULL)) != 0) 2211 return (error); 2212 2213 optype = VOP_MKNOD_DESCOFFSET; 2214 2215 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2216 if (error) { 2217 return error; 2218 } 2219 pathstring = pathbuf_stringcopy_get(pb); 2220 if (pathstring == NULL) { 2221 pathbuf_destroy(pb); 2222 return ENOMEM; 2223 } 2224 2225 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2226 2227 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2228 goto out; 2229 vp = nd.ni_vp; 2230 2231 if (vp != NULL) 2232 error = EEXIST; 2233 else { 2234 vattr_null(&vattr); 2235 /* We will read cwdi->cwdi_cmask unlocked. */ 2236 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2237 vattr.va_rdev = dev; 2238 2239 switch (mode & S_IFMT) { 2240 case S_IFMT: /* used by badsect to flag bad sectors */ 2241 vattr.va_type = VBAD; 2242 break; 2243 case S_IFCHR: 2244 vattr.va_type = VCHR; 2245 break; 2246 case S_IFBLK: 2247 vattr.va_type = VBLK; 2248 break; 2249 case S_IFWHT: 2250 optype = VOP_WHITEOUT_DESCOFFSET; 2251 break; 2252 case S_IFREG: 2253 #if NVERIEXEC > 0 2254 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2255 O_CREAT); 2256 #endif /* NVERIEXEC > 0 */ 2257 vattr.va_type = VREG; 2258 vattr.va_rdev = VNOVAL; 2259 optype = VOP_CREATE_DESCOFFSET; 2260 break; 2261 default: 2262 error = EINVAL; 2263 break; 2264 } 2265 } 2266 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2267 && vattr.va_rdev == VNOVAL) 2268 error = EINVAL; 2269 if (!error) { 2270 switch (optype) { 2271 case VOP_WHITEOUT_DESCOFFSET: 2272 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2273 if (error) 2274 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2275 vput(nd.ni_dvp); 2276 break; 2277 2278 case VOP_MKNOD_DESCOFFSET: 2279 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2280 &nd.ni_cnd, &vattr); 2281 if (error == 0) 2282 vput(nd.ni_vp); 2283 break; 2284 2285 case VOP_CREATE_DESCOFFSET: 2286 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2287 &nd.ni_cnd, &vattr); 2288 if (error == 0) 2289 vput(nd.ni_vp); 2290 break; 2291 } 2292 } else { 2293 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2294 if (nd.ni_dvp == vp) 2295 vrele(nd.ni_dvp); 2296 else 2297 vput(nd.ni_dvp); 2298 if (vp) 2299 vrele(vp); 2300 } 2301 out: 2302 pathbuf_stringcopy_put(pb, pathstring); 2303 pathbuf_destroy(pb); 2304 return (error); 2305 } 2306 2307 /* 2308 * Create a named pipe. 2309 */ 2310 /* ARGSUSED */ 2311 int 2312 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2313 { 2314 /* { 2315 syscallarg(const char *) path; 2316 syscallarg(int) mode; 2317 } */ 2318 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2319 } 2320 2321 int 2322 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2323 register_t *retval) 2324 { 2325 /* { 2326 syscallarg(int) fd; 2327 syscallarg(const char *) path; 2328 syscallarg(int) mode; 2329 } */ 2330 2331 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2332 SCARG(uap, mode)); 2333 } 2334 2335 static int 2336 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2337 { 2338 struct proc *p = l->l_proc; 2339 struct vattr vattr; 2340 int error; 2341 struct pathbuf *pb; 2342 struct nameidata nd; 2343 2344 error = pathbuf_copyin(path, &pb); 2345 if (error) { 2346 return error; 2347 } 2348 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2349 2350 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2351 pathbuf_destroy(pb); 2352 return error; 2353 } 2354 if (nd.ni_vp != NULL) { 2355 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2356 if (nd.ni_dvp == nd.ni_vp) 2357 vrele(nd.ni_dvp); 2358 else 2359 vput(nd.ni_dvp); 2360 vrele(nd.ni_vp); 2361 pathbuf_destroy(pb); 2362 return (EEXIST); 2363 } 2364 vattr_null(&vattr); 2365 vattr.va_type = VFIFO; 2366 /* We will read cwdi->cwdi_cmask unlocked. */ 2367 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2368 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2369 if (error == 0) 2370 vput(nd.ni_vp); 2371 pathbuf_destroy(pb); 2372 return (error); 2373 } 2374 2375 /* 2376 * Make a hard file link. 2377 */ 2378 /* ARGSUSED */ 2379 static int 2380 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2381 const char *link, int follow, register_t *retval) 2382 { 2383 struct vnode *vp; 2384 struct pathbuf *linkpb; 2385 struct nameidata nd; 2386 namei_simple_flags_t ns_flags; 2387 int error; 2388 2389 if (follow & AT_SYMLINK_FOLLOW) 2390 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2391 else 2392 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2393 2394 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2395 if (error != 0) 2396 return (error); 2397 error = pathbuf_copyin(link, &linkpb); 2398 if (error) { 2399 goto out1; 2400 } 2401 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2402 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2403 goto out2; 2404 if (nd.ni_vp) { 2405 error = EEXIST; 2406 goto abortop; 2407 } 2408 /* Prevent hard links on directories. */ 2409 if (vp->v_type == VDIR) { 2410 error = EPERM; 2411 goto abortop; 2412 } 2413 /* Prevent cross-mount operation. */ 2414 if (nd.ni_dvp->v_mount != vp->v_mount) { 2415 error = EXDEV; 2416 goto abortop; 2417 } 2418 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2419 out2: 2420 pathbuf_destroy(linkpb); 2421 out1: 2422 vrele(vp); 2423 return (error); 2424 abortop: 2425 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2426 if (nd.ni_dvp == nd.ni_vp) 2427 vrele(nd.ni_dvp); 2428 else 2429 vput(nd.ni_dvp); 2430 if (nd.ni_vp != NULL) 2431 vrele(nd.ni_vp); 2432 goto out2; 2433 } 2434 2435 int 2436 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2437 { 2438 /* { 2439 syscallarg(const char *) path; 2440 syscallarg(const char *) link; 2441 } */ 2442 const char *path = SCARG(uap, path); 2443 const char *link = SCARG(uap, link); 2444 2445 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2446 AT_SYMLINK_FOLLOW, retval); 2447 } 2448 2449 int 2450 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2451 register_t *retval) 2452 { 2453 /* { 2454 syscallarg(int) fd1; 2455 syscallarg(const char *) name1; 2456 syscallarg(int) fd2; 2457 syscallarg(const char *) name2; 2458 syscallarg(int) flags; 2459 } */ 2460 int fd1 = SCARG(uap, fd1); 2461 const char *name1 = SCARG(uap, name1); 2462 int fd2 = SCARG(uap, fd2); 2463 const char *name2 = SCARG(uap, name2); 2464 int follow; 2465 2466 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2467 2468 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2469 } 2470 2471 2472 int 2473 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2474 { 2475 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2476 } 2477 2478 static int 2479 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2480 const char *link, enum uio_seg seg) 2481 { 2482 struct proc *p = curproc; 2483 struct vattr vattr; 2484 char *path; 2485 int error; 2486 struct pathbuf *linkpb; 2487 struct nameidata nd; 2488 2489 KASSERT(l != NULL || fdat == AT_FDCWD); 2490 2491 path = PNBUF_GET(); 2492 if (seg == UIO_USERSPACE) { 2493 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2494 goto out1; 2495 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2496 goto out1; 2497 } else { 2498 KASSERT(strlen(patharg) < MAXPATHLEN); 2499 strcpy(path, patharg); 2500 linkpb = pathbuf_create(link); 2501 if (linkpb == NULL) { 2502 error = ENOMEM; 2503 goto out1; 2504 } 2505 } 2506 ktrkuser("symlink-target", path, strlen(path)); 2507 2508 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2509 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2510 goto out2; 2511 if (nd.ni_vp) { 2512 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2513 if (nd.ni_dvp == nd.ni_vp) 2514 vrele(nd.ni_dvp); 2515 else 2516 vput(nd.ni_dvp); 2517 vrele(nd.ni_vp); 2518 error = EEXIST; 2519 goto out2; 2520 } 2521 vattr_null(&vattr); 2522 vattr.va_type = VLNK; 2523 /* We will read cwdi->cwdi_cmask unlocked. */ 2524 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2525 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2526 if (error == 0) 2527 vput(nd.ni_vp); 2528 out2: 2529 pathbuf_destroy(linkpb); 2530 out1: 2531 PNBUF_PUT(path); 2532 return (error); 2533 } 2534 2535 /* 2536 * Make a symbolic link. 2537 */ 2538 /* ARGSUSED */ 2539 int 2540 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2541 { 2542 /* { 2543 syscallarg(const char *) path; 2544 syscallarg(const char *) link; 2545 } */ 2546 2547 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2548 UIO_USERSPACE); 2549 } 2550 2551 int 2552 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2553 register_t *retval) 2554 { 2555 /* { 2556 syscallarg(const char *) path1; 2557 syscallarg(int) fd; 2558 syscallarg(const char *) path2; 2559 } */ 2560 2561 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2562 SCARG(uap, path2), UIO_USERSPACE); 2563 } 2564 2565 /* 2566 * Delete a whiteout from the filesystem. 2567 */ 2568 /* ARGSUSED */ 2569 int 2570 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2571 { 2572 /* { 2573 syscallarg(const char *) path; 2574 } */ 2575 int error; 2576 struct pathbuf *pb; 2577 struct nameidata nd; 2578 2579 error = pathbuf_copyin(SCARG(uap, path), &pb); 2580 if (error) { 2581 return error; 2582 } 2583 2584 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2585 error = namei(&nd); 2586 if (error) { 2587 pathbuf_destroy(pb); 2588 return (error); 2589 } 2590 2591 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2592 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2593 if (nd.ni_dvp == nd.ni_vp) 2594 vrele(nd.ni_dvp); 2595 else 2596 vput(nd.ni_dvp); 2597 if (nd.ni_vp) 2598 vrele(nd.ni_vp); 2599 pathbuf_destroy(pb); 2600 return (EEXIST); 2601 } 2602 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2603 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2604 vput(nd.ni_dvp); 2605 pathbuf_destroy(pb); 2606 return (error); 2607 } 2608 2609 /* 2610 * Delete a name from the filesystem. 2611 */ 2612 /* ARGSUSED */ 2613 int 2614 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2615 { 2616 /* { 2617 syscallarg(const char *) path; 2618 } */ 2619 2620 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2621 } 2622 2623 int 2624 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2625 register_t *retval) 2626 { 2627 /* { 2628 syscallarg(int) fd; 2629 syscallarg(const char *) path; 2630 syscallarg(int) flag; 2631 } */ 2632 2633 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2634 SCARG(uap, flag), UIO_USERSPACE); 2635 } 2636 2637 int 2638 do_sys_unlink(const char *arg, enum uio_seg seg) 2639 { 2640 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2641 } 2642 2643 static int 2644 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2645 enum uio_seg seg) 2646 { 2647 struct vnode *vp; 2648 int error; 2649 struct pathbuf *pb; 2650 struct nameidata nd; 2651 const char *pathstring; 2652 2653 KASSERT(l != NULL || fdat == AT_FDCWD); 2654 2655 error = pathbuf_maybe_copyin(arg, seg, &pb); 2656 if (error) { 2657 return error; 2658 } 2659 pathstring = pathbuf_stringcopy_get(pb); 2660 if (pathstring == NULL) { 2661 pathbuf_destroy(pb); 2662 return ENOMEM; 2663 } 2664 2665 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2666 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2667 goto out; 2668 vp = nd.ni_vp; 2669 2670 /* 2671 * The root of a mounted filesystem cannot be deleted. 2672 */ 2673 if ((vp->v_vflag & VV_ROOT) != 0) { 2674 error = EBUSY; 2675 goto abort; 2676 } 2677 2678 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2679 error = EBUSY; 2680 goto abort; 2681 } 2682 2683 /* 2684 * No rmdir "." please. 2685 */ 2686 if (nd.ni_dvp == vp) { 2687 error = EINVAL; 2688 goto abort; 2689 } 2690 2691 /* 2692 * AT_REMOVEDIR is required to remove a directory 2693 */ 2694 if (vp->v_type == VDIR) { 2695 if (!(flags & AT_REMOVEDIR)) { 2696 error = EPERM; 2697 goto abort; 2698 } else { 2699 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2700 goto out; 2701 } 2702 } 2703 2704 /* 2705 * Starting here we only deal with non directories. 2706 */ 2707 if (flags & AT_REMOVEDIR) { 2708 error = ENOTDIR; 2709 goto abort; 2710 } 2711 2712 2713 #if NVERIEXEC > 0 2714 /* Handle remove requests for veriexec entries. */ 2715 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2716 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2717 if (nd.ni_dvp == vp) 2718 vrele(nd.ni_dvp); 2719 else 2720 vput(nd.ni_dvp); 2721 vput(vp); 2722 goto out; 2723 } 2724 #endif /* NVERIEXEC > 0 */ 2725 2726 #ifdef FILEASSOC 2727 (void)fileassoc_file_delete(vp); 2728 #endif /* FILEASSOC */ 2729 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2730 goto out; 2731 2732 abort: 2733 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2734 if (nd.ni_dvp == vp) 2735 vrele(nd.ni_dvp); 2736 else 2737 vput(nd.ni_dvp); 2738 vput(vp); 2739 2740 out: 2741 pathbuf_stringcopy_put(pb, pathstring); 2742 pathbuf_destroy(pb); 2743 return (error); 2744 } 2745 2746 /* 2747 * Reposition read/write file offset. 2748 */ 2749 int 2750 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2751 { 2752 /* { 2753 syscallarg(int) fd; 2754 syscallarg(int) pad; 2755 syscallarg(off_t) offset; 2756 syscallarg(int) whence; 2757 } */ 2758 kauth_cred_t cred = l->l_cred; 2759 file_t *fp; 2760 struct vnode *vp; 2761 struct vattr vattr; 2762 off_t newoff; 2763 int error, fd; 2764 2765 fd = SCARG(uap, fd); 2766 2767 if ((fp = fd_getfile(fd)) == NULL) 2768 return (EBADF); 2769 2770 vp = fp->f_data; 2771 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2772 error = ESPIPE; 2773 goto out; 2774 } 2775 2776 switch (SCARG(uap, whence)) { 2777 case SEEK_CUR: 2778 newoff = fp->f_offset + SCARG(uap, offset); 2779 break; 2780 case SEEK_END: 2781 vn_lock(vp, LK_SHARED | LK_RETRY); 2782 error = VOP_GETATTR(vp, &vattr, cred); 2783 VOP_UNLOCK(vp); 2784 if (error) { 2785 goto out; 2786 } 2787 newoff = SCARG(uap, offset) + vattr.va_size; 2788 break; 2789 case SEEK_SET: 2790 newoff = SCARG(uap, offset); 2791 break; 2792 default: 2793 error = EINVAL; 2794 goto out; 2795 } 2796 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2797 *(off_t *)retval = fp->f_offset = newoff; 2798 } 2799 out: 2800 fd_putfile(fd); 2801 return (error); 2802 } 2803 2804 /* 2805 * Positional read system call. 2806 */ 2807 int 2808 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2809 { 2810 /* { 2811 syscallarg(int) fd; 2812 syscallarg(void *) buf; 2813 syscallarg(size_t) nbyte; 2814 syscallarg(off_t) offset; 2815 } */ 2816 file_t *fp; 2817 struct vnode *vp; 2818 off_t offset; 2819 int error, fd = SCARG(uap, fd); 2820 2821 if ((fp = fd_getfile(fd)) == NULL) 2822 return (EBADF); 2823 2824 if ((fp->f_flag & FREAD) == 0) { 2825 fd_putfile(fd); 2826 return (EBADF); 2827 } 2828 2829 vp = fp->f_data; 2830 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2831 error = ESPIPE; 2832 goto out; 2833 } 2834 2835 offset = SCARG(uap, offset); 2836 2837 /* 2838 * XXX This works because no file systems actually 2839 * XXX take any action on the seek operation. 2840 */ 2841 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2842 goto out; 2843 2844 /* dofileread() will unuse the descriptor for us */ 2845 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2846 &offset, 0, retval)); 2847 2848 out: 2849 fd_putfile(fd); 2850 return (error); 2851 } 2852 2853 /* 2854 * Positional scatter read system call. 2855 */ 2856 int 2857 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2858 { 2859 /* { 2860 syscallarg(int) fd; 2861 syscallarg(const struct iovec *) iovp; 2862 syscallarg(int) iovcnt; 2863 syscallarg(off_t) offset; 2864 } */ 2865 off_t offset = SCARG(uap, offset); 2866 2867 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2868 SCARG(uap, iovcnt), &offset, 0, retval); 2869 } 2870 2871 /* 2872 * Positional write system call. 2873 */ 2874 int 2875 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2876 { 2877 /* { 2878 syscallarg(int) fd; 2879 syscallarg(const void *) buf; 2880 syscallarg(size_t) nbyte; 2881 syscallarg(off_t) offset; 2882 } */ 2883 file_t *fp; 2884 struct vnode *vp; 2885 off_t offset; 2886 int error, fd = SCARG(uap, fd); 2887 2888 if ((fp = fd_getfile(fd)) == NULL) 2889 return (EBADF); 2890 2891 if ((fp->f_flag & FWRITE) == 0) { 2892 fd_putfile(fd); 2893 return (EBADF); 2894 } 2895 2896 vp = fp->f_data; 2897 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2898 error = ESPIPE; 2899 goto out; 2900 } 2901 2902 offset = SCARG(uap, offset); 2903 2904 /* 2905 * XXX This works because no file systems actually 2906 * XXX take any action on the seek operation. 2907 */ 2908 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2909 goto out; 2910 2911 /* dofilewrite() will unuse the descriptor for us */ 2912 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2913 &offset, 0, retval)); 2914 2915 out: 2916 fd_putfile(fd); 2917 return (error); 2918 } 2919 2920 /* 2921 * Positional gather write system call. 2922 */ 2923 int 2924 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2925 { 2926 /* { 2927 syscallarg(int) fd; 2928 syscallarg(const struct iovec *) iovp; 2929 syscallarg(int) iovcnt; 2930 syscallarg(off_t) offset; 2931 } */ 2932 off_t offset = SCARG(uap, offset); 2933 2934 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2935 SCARG(uap, iovcnt), &offset, 0, retval); 2936 } 2937 2938 /* 2939 * Check access permissions. 2940 */ 2941 int 2942 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2943 { 2944 /* { 2945 syscallarg(const char *) path; 2946 syscallarg(int) flags; 2947 } */ 2948 2949 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2950 SCARG(uap, flags), 0); 2951 } 2952 2953 static int 2954 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2955 int mode, int flags) 2956 { 2957 kauth_cred_t cred; 2958 struct vnode *vp; 2959 int error, nd_flag, vmode; 2960 struct pathbuf *pb; 2961 struct nameidata nd; 2962 2963 CTASSERT(F_OK == 0); 2964 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2965 /* nonsense mode */ 2966 return EINVAL; 2967 } 2968 2969 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2970 if (flags & AT_SYMLINK_NOFOLLOW) 2971 nd_flag &= ~FOLLOW; 2972 2973 error = pathbuf_copyin(path, &pb); 2974 if (error) 2975 return error; 2976 2977 NDINIT(&nd, LOOKUP, nd_flag, pb); 2978 2979 /* Override default credentials */ 2980 cred = kauth_cred_dup(l->l_cred); 2981 if (!(flags & AT_EACCESS)) { 2982 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2983 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2984 } 2985 nd.ni_cnd.cn_cred = cred; 2986 2987 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2988 pathbuf_destroy(pb); 2989 goto out; 2990 } 2991 vp = nd.ni_vp; 2992 pathbuf_destroy(pb); 2993 2994 /* Flags == 0 means only check for existence. */ 2995 if (mode) { 2996 vmode = 0; 2997 if (mode & R_OK) 2998 vmode |= VREAD; 2999 if (mode & W_OK) 3000 vmode |= VWRITE; 3001 if (mode & X_OK) 3002 vmode |= VEXEC; 3003 3004 error = VOP_ACCESS(vp, vmode, cred); 3005 if (!error && (vmode & VWRITE)) 3006 error = vn_writechk(vp); 3007 } 3008 vput(vp); 3009 out: 3010 kauth_cred_free(cred); 3011 return (error); 3012 } 3013 3014 int 3015 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3016 register_t *retval) 3017 { 3018 /* { 3019 syscallarg(int) fd; 3020 syscallarg(const char *) path; 3021 syscallarg(int) amode; 3022 syscallarg(int) flag; 3023 } */ 3024 3025 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3026 SCARG(uap, amode), SCARG(uap, flag)); 3027 } 3028 3029 /* 3030 * Common code for all sys_stat functions, including compat versions. 3031 */ 3032 int 3033 do_sys_stat(const char *userpath, unsigned int nd_flag, 3034 struct stat *sb) 3035 { 3036 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3037 } 3038 3039 static int 3040 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3041 unsigned int nd_flag, struct stat *sb) 3042 { 3043 int error; 3044 struct pathbuf *pb; 3045 struct nameidata nd; 3046 3047 KASSERT(l != NULL || fdat == AT_FDCWD); 3048 3049 error = pathbuf_copyin(userpath, &pb); 3050 if (error) { 3051 return error; 3052 } 3053 3054 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3055 3056 error = fd_nameiat(l, fdat, &nd); 3057 if (error != 0) { 3058 pathbuf_destroy(pb); 3059 return error; 3060 } 3061 error = vn_stat(nd.ni_vp, sb); 3062 vput(nd.ni_vp); 3063 pathbuf_destroy(pb); 3064 return error; 3065 } 3066 3067 /* 3068 * Get file status; this version follows links. 3069 */ 3070 /* ARGSUSED */ 3071 int 3072 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3073 { 3074 /* { 3075 syscallarg(const char *) path; 3076 syscallarg(struct stat *) ub; 3077 } */ 3078 struct stat sb; 3079 int error; 3080 3081 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3082 if (error) 3083 return error; 3084 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3085 } 3086 3087 /* 3088 * Get file status; this version does not follow links. 3089 */ 3090 /* ARGSUSED */ 3091 int 3092 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3093 { 3094 /* { 3095 syscallarg(const char *) path; 3096 syscallarg(struct stat *) ub; 3097 } */ 3098 struct stat sb; 3099 int error; 3100 3101 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3102 if (error) 3103 return error; 3104 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3105 } 3106 3107 int 3108 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3109 register_t *retval) 3110 { 3111 /* { 3112 syscallarg(int) fd; 3113 syscallarg(const char *) path; 3114 syscallarg(struct stat *) buf; 3115 syscallarg(int) flag; 3116 } */ 3117 unsigned int nd_flag; 3118 struct stat sb; 3119 int error; 3120 3121 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3122 nd_flag = NOFOLLOW; 3123 else 3124 nd_flag = FOLLOW; 3125 3126 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3127 &sb); 3128 if (error) 3129 return error; 3130 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3131 } 3132 3133 /* 3134 * Get configurable pathname variables. 3135 */ 3136 /* ARGSUSED */ 3137 int 3138 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3139 { 3140 /* { 3141 syscallarg(const char *) path; 3142 syscallarg(int) name; 3143 } */ 3144 int error; 3145 struct pathbuf *pb; 3146 struct nameidata nd; 3147 3148 error = pathbuf_copyin(SCARG(uap, path), &pb); 3149 if (error) { 3150 return error; 3151 } 3152 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3153 if ((error = namei(&nd)) != 0) { 3154 pathbuf_destroy(pb); 3155 return (error); 3156 } 3157 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3158 vput(nd.ni_vp); 3159 pathbuf_destroy(pb); 3160 return (error); 3161 } 3162 3163 /* 3164 * Return target name of a symbolic link. 3165 */ 3166 /* ARGSUSED */ 3167 int 3168 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3169 register_t *retval) 3170 { 3171 /* { 3172 syscallarg(const char *) path; 3173 syscallarg(char *) buf; 3174 syscallarg(size_t) count; 3175 } */ 3176 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3177 SCARG(uap, buf), SCARG(uap, count), retval); 3178 } 3179 3180 static int 3181 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3182 size_t count, register_t *retval) 3183 { 3184 struct vnode *vp; 3185 struct iovec aiov; 3186 struct uio auio; 3187 int error; 3188 struct pathbuf *pb; 3189 struct nameidata nd; 3190 3191 error = pathbuf_copyin(path, &pb); 3192 if (error) { 3193 return error; 3194 } 3195 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3196 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3197 pathbuf_destroy(pb); 3198 return error; 3199 } 3200 vp = nd.ni_vp; 3201 pathbuf_destroy(pb); 3202 if (vp->v_type != VLNK) 3203 error = EINVAL; 3204 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3205 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3206 aiov.iov_base = buf; 3207 aiov.iov_len = count; 3208 auio.uio_iov = &aiov; 3209 auio.uio_iovcnt = 1; 3210 auio.uio_offset = 0; 3211 auio.uio_rw = UIO_READ; 3212 KASSERT(l == curlwp); 3213 auio.uio_vmspace = l->l_proc->p_vmspace; 3214 auio.uio_resid = count; 3215 error = VOP_READLINK(vp, &auio, l->l_cred); 3216 } 3217 vput(vp); 3218 *retval = count - auio.uio_resid; 3219 return (error); 3220 } 3221 3222 int 3223 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3224 register_t *retval) 3225 { 3226 /* { 3227 syscallarg(int) fd; 3228 syscallarg(const char *) path; 3229 syscallarg(char *) buf; 3230 syscallarg(size_t) bufsize; 3231 } */ 3232 3233 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3234 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3235 } 3236 3237 /* 3238 * Change flags of a file given a path name. 3239 */ 3240 /* ARGSUSED */ 3241 int 3242 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3243 { 3244 /* { 3245 syscallarg(const char *) path; 3246 syscallarg(u_long) flags; 3247 } */ 3248 struct vnode *vp; 3249 int error; 3250 3251 error = namei_simple_user(SCARG(uap, path), 3252 NSM_FOLLOW_TRYEMULROOT, &vp); 3253 if (error != 0) 3254 return (error); 3255 error = change_flags(vp, SCARG(uap, flags), l); 3256 vput(vp); 3257 return (error); 3258 } 3259 3260 /* 3261 * Change flags of a file given a file descriptor. 3262 */ 3263 /* ARGSUSED */ 3264 int 3265 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3266 { 3267 /* { 3268 syscallarg(int) fd; 3269 syscallarg(u_long) flags; 3270 } */ 3271 struct vnode *vp; 3272 file_t *fp; 3273 int error; 3274 3275 /* fd_getvnode() will use the descriptor for us */ 3276 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3277 return (error); 3278 vp = fp->f_data; 3279 error = change_flags(vp, SCARG(uap, flags), l); 3280 VOP_UNLOCK(vp); 3281 fd_putfile(SCARG(uap, fd)); 3282 return (error); 3283 } 3284 3285 /* 3286 * Change flags of a file given a path name; this version does 3287 * not follow links. 3288 */ 3289 int 3290 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3291 { 3292 /* { 3293 syscallarg(const char *) path; 3294 syscallarg(u_long) flags; 3295 } */ 3296 struct vnode *vp; 3297 int error; 3298 3299 error = namei_simple_user(SCARG(uap, path), 3300 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3301 if (error != 0) 3302 return (error); 3303 error = change_flags(vp, SCARG(uap, flags), l); 3304 vput(vp); 3305 return (error); 3306 } 3307 3308 /* 3309 * Common routine to change flags of a file. 3310 */ 3311 int 3312 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3313 { 3314 struct vattr vattr; 3315 int error; 3316 3317 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3318 3319 vattr_null(&vattr); 3320 vattr.va_flags = flags; 3321 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3322 3323 return (error); 3324 } 3325 3326 /* 3327 * Change mode of a file given path name; this version follows links. 3328 */ 3329 /* ARGSUSED */ 3330 int 3331 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3332 { 3333 /* { 3334 syscallarg(const char *) path; 3335 syscallarg(int) mode; 3336 } */ 3337 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3338 SCARG(uap, mode), 0); 3339 } 3340 3341 static int 3342 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3343 { 3344 int error; 3345 struct vnode *vp; 3346 namei_simple_flags_t ns_flag; 3347 3348 if (flags & AT_SYMLINK_NOFOLLOW) 3349 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3350 else 3351 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3352 3353 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3354 if (error != 0) 3355 return error; 3356 3357 error = change_mode(vp, mode, l); 3358 3359 vrele(vp); 3360 3361 return (error); 3362 } 3363 3364 /* 3365 * Change mode of a file given a file descriptor. 3366 */ 3367 /* ARGSUSED */ 3368 int 3369 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3370 { 3371 /* { 3372 syscallarg(int) fd; 3373 syscallarg(int) mode; 3374 } */ 3375 file_t *fp; 3376 int error; 3377 3378 /* fd_getvnode() will use the descriptor for us */ 3379 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3380 return (error); 3381 error = change_mode(fp->f_data, SCARG(uap, mode), l); 3382 fd_putfile(SCARG(uap, fd)); 3383 return (error); 3384 } 3385 3386 int 3387 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3388 register_t *retval) 3389 { 3390 /* { 3391 syscallarg(int) fd; 3392 syscallarg(const char *) path; 3393 syscallarg(int) mode; 3394 syscallarg(int) flag; 3395 } */ 3396 3397 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3398 SCARG(uap, mode), SCARG(uap, flag)); 3399 } 3400 3401 /* 3402 * Change mode of a file given path name; this version does not follow links. 3403 */ 3404 /* ARGSUSED */ 3405 int 3406 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3407 { 3408 /* { 3409 syscallarg(const char *) path; 3410 syscallarg(int) mode; 3411 } */ 3412 int error; 3413 struct vnode *vp; 3414 3415 error = namei_simple_user(SCARG(uap, path), 3416 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3417 if (error != 0) 3418 return (error); 3419 3420 error = change_mode(vp, SCARG(uap, mode), l); 3421 3422 vrele(vp); 3423 return (error); 3424 } 3425 3426 /* 3427 * Common routine to set mode given a vnode. 3428 */ 3429 static int 3430 change_mode(struct vnode *vp, int mode, struct lwp *l) 3431 { 3432 struct vattr vattr; 3433 int error; 3434 3435 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3436 vattr_null(&vattr); 3437 vattr.va_mode = mode & ALLPERMS; 3438 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3439 VOP_UNLOCK(vp); 3440 return (error); 3441 } 3442 3443 /* 3444 * Set ownership given a path name; this version follows links. 3445 */ 3446 /* ARGSUSED */ 3447 int 3448 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3449 { 3450 /* { 3451 syscallarg(const char *) path; 3452 syscallarg(uid_t) uid; 3453 syscallarg(gid_t) gid; 3454 } */ 3455 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3456 SCARG(uap, gid), 0); 3457 } 3458 3459 static int 3460 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3461 gid_t gid, int flags) 3462 { 3463 int error; 3464 struct vnode *vp; 3465 namei_simple_flags_t ns_flag; 3466 3467 if (flags & AT_SYMLINK_NOFOLLOW) 3468 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3469 else 3470 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3471 3472 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3473 if (error != 0) 3474 return error; 3475 3476 error = change_owner(vp, uid, gid, l, 0); 3477 3478 vrele(vp); 3479 3480 return (error); 3481 } 3482 3483 /* 3484 * Set ownership given a path name; this version follows links. 3485 * Provides POSIX semantics. 3486 */ 3487 /* ARGSUSED */ 3488 int 3489 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3490 { 3491 /* { 3492 syscallarg(const char *) path; 3493 syscallarg(uid_t) uid; 3494 syscallarg(gid_t) gid; 3495 } */ 3496 int error; 3497 struct vnode *vp; 3498 3499 error = namei_simple_user(SCARG(uap, path), 3500 NSM_FOLLOW_TRYEMULROOT, &vp); 3501 if (error != 0) 3502 return (error); 3503 3504 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3505 3506 vrele(vp); 3507 return (error); 3508 } 3509 3510 /* 3511 * Set ownership given a file descriptor. 3512 */ 3513 /* ARGSUSED */ 3514 int 3515 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3516 { 3517 /* { 3518 syscallarg(int) fd; 3519 syscallarg(uid_t) uid; 3520 syscallarg(gid_t) gid; 3521 } */ 3522 int error; 3523 file_t *fp; 3524 3525 /* fd_getvnode() will use the descriptor for us */ 3526 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3527 return (error); 3528 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3529 l, 0); 3530 fd_putfile(SCARG(uap, fd)); 3531 return (error); 3532 } 3533 3534 int 3535 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3536 register_t *retval) 3537 { 3538 /* { 3539 syscallarg(int) fd; 3540 syscallarg(const char *) path; 3541 syscallarg(uid_t) owner; 3542 syscallarg(gid_t) group; 3543 syscallarg(int) flag; 3544 } */ 3545 3546 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3547 SCARG(uap, owner), SCARG(uap, group), 3548 SCARG(uap, flag)); 3549 } 3550 3551 /* 3552 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3553 */ 3554 /* ARGSUSED */ 3555 int 3556 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3557 { 3558 /* { 3559 syscallarg(int) fd; 3560 syscallarg(uid_t) uid; 3561 syscallarg(gid_t) gid; 3562 } */ 3563 int error; 3564 file_t *fp; 3565 3566 /* fd_getvnode() will use the descriptor for us */ 3567 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3568 return (error); 3569 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3570 l, 1); 3571 fd_putfile(SCARG(uap, fd)); 3572 return (error); 3573 } 3574 3575 /* 3576 * Set ownership given a path name; this version does not follow links. 3577 */ 3578 /* ARGSUSED */ 3579 int 3580 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3581 { 3582 /* { 3583 syscallarg(const char *) path; 3584 syscallarg(uid_t) uid; 3585 syscallarg(gid_t) gid; 3586 } */ 3587 int error; 3588 struct vnode *vp; 3589 3590 error = namei_simple_user(SCARG(uap, path), 3591 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3592 if (error != 0) 3593 return (error); 3594 3595 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3596 3597 vrele(vp); 3598 return (error); 3599 } 3600 3601 /* 3602 * Set ownership given a path name; this version does not follow links. 3603 * Provides POSIX/XPG semantics. 3604 */ 3605 /* ARGSUSED */ 3606 int 3607 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3608 { 3609 /* { 3610 syscallarg(const char *) path; 3611 syscallarg(uid_t) uid; 3612 syscallarg(gid_t) gid; 3613 } */ 3614 int error; 3615 struct vnode *vp; 3616 3617 error = namei_simple_user(SCARG(uap, path), 3618 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3619 if (error != 0) 3620 return (error); 3621 3622 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3623 3624 vrele(vp); 3625 return (error); 3626 } 3627 3628 /* 3629 * Common routine to set ownership given a vnode. 3630 */ 3631 static int 3632 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3633 int posix_semantics) 3634 { 3635 struct vattr vattr; 3636 mode_t newmode; 3637 int error; 3638 3639 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3640 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3641 goto out; 3642 3643 #define CHANGED(x) ((int)(x) != -1) 3644 newmode = vattr.va_mode; 3645 if (posix_semantics) { 3646 /* 3647 * POSIX/XPG semantics: if the caller is not the super-user, 3648 * clear set-user-id and set-group-id bits. Both POSIX and 3649 * the XPG consider the behaviour for calls by the super-user 3650 * implementation-defined; we leave the set-user-id and set- 3651 * group-id settings intact in that case. 3652 */ 3653 if (vattr.va_mode & S_ISUID) { 3654 if (kauth_authorize_vnode(l->l_cred, 3655 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3656 newmode &= ~S_ISUID; 3657 } 3658 if (vattr.va_mode & S_ISGID) { 3659 if (kauth_authorize_vnode(l->l_cred, 3660 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3661 newmode &= ~S_ISGID; 3662 } 3663 } else { 3664 /* 3665 * NetBSD semantics: when changing owner and/or group, 3666 * clear the respective bit(s). 3667 */ 3668 if (CHANGED(uid)) 3669 newmode &= ~S_ISUID; 3670 if (CHANGED(gid)) 3671 newmode &= ~S_ISGID; 3672 } 3673 /* Update va_mode iff altered. */ 3674 if (vattr.va_mode == newmode) 3675 newmode = VNOVAL; 3676 3677 vattr_null(&vattr); 3678 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3679 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3680 vattr.va_mode = newmode; 3681 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3682 #undef CHANGED 3683 3684 out: 3685 VOP_UNLOCK(vp); 3686 return (error); 3687 } 3688 3689 /* 3690 * Set the access and modification times given a path name; this 3691 * version follows links. 3692 */ 3693 /* ARGSUSED */ 3694 int 3695 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3696 register_t *retval) 3697 { 3698 /* { 3699 syscallarg(const char *) path; 3700 syscallarg(const struct timeval *) tptr; 3701 } */ 3702 3703 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3704 SCARG(uap, tptr), UIO_USERSPACE); 3705 } 3706 3707 /* 3708 * Set the access and modification times given a file descriptor. 3709 */ 3710 /* ARGSUSED */ 3711 int 3712 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3713 register_t *retval) 3714 { 3715 /* { 3716 syscallarg(int) fd; 3717 syscallarg(const struct timeval *) tptr; 3718 } */ 3719 int error; 3720 file_t *fp; 3721 3722 /* fd_getvnode() will use the descriptor for us */ 3723 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3724 return (error); 3725 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3726 UIO_USERSPACE); 3727 fd_putfile(SCARG(uap, fd)); 3728 return (error); 3729 } 3730 3731 int 3732 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3733 register_t *retval) 3734 { 3735 /* { 3736 syscallarg(int) fd; 3737 syscallarg(const struct timespec *) tptr; 3738 } */ 3739 int error; 3740 file_t *fp; 3741 3742 /* fd_getvnode() will use the descriptor for us */ 3743 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3744 return (error); 3745 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0, 3746 SCARG(uap, tptr), UIO_USERSPACE); 3747 fd_putfile(SCARG(uap, fd)); 3748 return (error); 3749 } 3750 3751 /* 3752 * Set the access and modification times given a path name; this 3753 * version does not follow links. 3754 */ 3755 int 3756 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3757 register_t *retval) 3758 { 3759 /* { 3760 syscallarg(const char *) path; 3761 syscallarg(const struct timeval *) tptr; 3762 } */ 3763 3764 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3765 SCARG(uap, tptr), UIO_USERSPACE); 3766 } 3767 3768 int 3769 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3770 register_t *retval) 3771 { 3772 /* { 3773 syscallarg(int) fd; 3774 syscallarg(const char *) path; 3775 syscallarg(const struct timespec *) tptr; 3776 syscallarg(int) flag; 3777 } */ 3778 int follow; 3779 const struct timespec *tptr; 3780 int error; 3781 3782 tptr = SCARG(uap, tptr); 3783 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3784 3785 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3786 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3787 3788 return error; 3789 } 3790 3791 /* 3792 * Common routine to set access and modification times given a vnode. 3793 */ 3794 int 3795 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3796 const struct timespec *tptr, enum uio_seg seg) 3797 { 3798 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3799 } 3800 3801 static int 3802 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3803 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3804 { 3805 struct vattr vattr; 3806 int error, dorele = 0; 3807 namei_simple_flags_t sflags; 3808 bool vanull, setbirthtime; 3809 struct timespec ts[2]; 3810 3811 KASSERT(l != NULL || fdat == AT_FDCWD); 3812 3813 /* 3814 * I have checked all callers and they pass either FOLLOW, 3815 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3816 * is 0. More to the point, they don't pass anything else. 3817 * Let's keep it that way at least until the namei interfaces 3818 * are fully sanitized. 3819 */ 3820 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3821 sflags = (flag == FOLLOW) ? 3822 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3823 3824 if (tptr == NULL) { 3825 vanull = true; 3826 nanotime(&ts[0]); 3827 ts[1] = ts[0]; 3828 } else { 3829 vanull = false; 3830 if (seg != UIO_SYSSPACE) { 3831 error = copyin(tptr, ts, sizeof (ts)); 3832 if (error != 0) 3833 return error; 3834 } else { 3835 ts[0] = tptr[0]; 3836 ts[1] = tptr[1]; 3837 } 3838 } 3839 3840 if (ts[0].tv_nsec == UTIME_NOW) { 3841 nanotime(&ts[0]); 3842 if (ts[1].tv_nsec == UTIME_NOW) { 3843 vanull = true; 3844 ts[1] = ts[0]; 3845 } 3846 } else if (ts[1].tv_nsec == UTIME_NOW) 3847 nanotime(&ts[1]); 3848 3849 if (vp == NULL) { 3850 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3851 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3852 if (error != 0) 3853 return error; 3854 dorele = 1; 3855 } 3856 3857 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3858 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3859 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3860 vattr_null(&vattr); 3861 3862 if (ts[0].tv_nsec != UTIME_OMIT) 3863 vattr.va_atime = ts[0]; 3864 3865 if (ts[1].tv_nsec != UTIME_OMIT) { 3866 vattr.va_mtime = ts[1]; 3867 if (setbirthtime) 3868 vattr.va_birthtime = ts[1]; 3869 } 3870 3871 if (vanull) 3872 vattr.va_vaflags |= VA_UTIMES_NULL; 3873 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3874 VOP_UNLOCK(vp); 3875 3876 if (dorele != 0) 3877 vrele(vp); 3878 3879 return error; 3880 } 3881 3882 int 3883 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3884 const struct timeval *tptr, enum uio_seg seg) 3885 { 3886 struct timespec ts[2]; 3887 struct timespec *tsptr = NULL; 3888 int error; 3889 3890 if (tptr != NULL) { 3891 struct timeval tv[2]; 3892 3893 if (seg != UIO_SYSSPACE) { 3894 error = copyin(tptr, tv, sizeof (tv)); 3895 if (error != 0) 3896 return error; 3897 tptr = tv; 3898 } 3899 3900 if ((tv[0].tv_usec == UTIME_NOW) || 3901 (tv[0].tv_usec == UTIME_OMIT)) 3902 ts[0].tv_nsec = tv[0].tv_usec; 3903 else 3904 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3905 3906 if ((tv[1].tv_usec == UTIME_NOW) || 3907 (tv[1].tv_usec == UTIME_OMIT)) 3908 ts[1].tv_nsec = tv[1].tv_usec; 3909 else 3910 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3911 3912 tsptr = &ts[0]; 3913 } 3914 3915 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3916 } 3917 3918 /* 3919 * Truncate a file given its path name. 3920 */ 3921 /* ARGSUSED */ 3922 int 3923 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3924 { 3925 /* { 3926 syscallarg(const char *) path; 3927 syscallarg(int) pad; 3928 syscallarg(off_t) length; 3929 } */ 3930 struct vnode *vp; 3931 struct vattr vattr; 3932 int error; 3933 3934 error = namei_simple_user(SCARG(uap, path), 3935 NSM_FOLLOW_TRYEMULROOT, &vp); 3936 if (error != 0) 3937 return (error); 3938 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3939 if (vp->v_type == VDIR) 3940 error = EISDIR; 3941 else if ((error = vn_writechk(vp)) == 0 && 3942 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3943 vattr_null(&vattr); 3944 vattr.va_size = SCARG(uap, length); 3945 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3946 } 3947 vput(vp); 3948 return (error); 3949 } 3950 3951 /* 3952 * Truncate a file given a file descriptor. 3953 */ 3954 /* ARGSUSED */ 3955 int 3956 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3957 { 3958 /* { 3959 syscallarg(int) fd; 3960 syscallarg(int) pad; 3961 syscallarg(off_t) length; 3962 } */ 3963 struct vattr vattr; 3964 struct vnode *vp; 3965 file_t *fp; 3966 int error; 3967 3968 /* fd_getvnode() will use the descriptor for us */ 3969 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3970 return (error); 3971 if ((fp->f_flag & FWRITE) == 0) { 3972 error = EINVAL; 3973 goto out; 3974 } 3975 vp = fp->f_data; 3976 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3977 if (vp->v_type == VDIR) 3978 error = EISDIR; 3979 else if ((error = vn_writechk(vp)) == 0) { 3980 vattr_null(&vattr); 3981 vattr.va_size = SCARG(uap, length); 3982 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3983 } 3984 VOP_UNLOCK(vp); 3985 out: 3986 fd_putfile(SCARG(uap, fd)); 3987 return (error); 3988 } 3989 3990 /* 3991 * Sync an open file. 3992 */ 3993 /* ARGSUSED */ 3994 int 3995 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3996 { 3997 /* { 3998 syscallarg(int) fd; 3999 } */ 4000 struct vnode *vp; 4001 file_t *fp; 4002 int error; 4003 4004 /* fd_getvnode() will use the descriptor for us */ 4005 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4006 return (error); 4007 vp = fp->f_data; 4008 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4009 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4010 VOP_UNLOCK(vp); 4011 fd_putfile(SCARG(uap, fd)); 4012 return (error); 4013 } 4014 4015 /* 4016 * Sync a range of file data. API modeled after that found in AIX. 4017 * 4018 * FDATASYNC indicates that we need only save enough metadata to be able 4019 * to re-read the written data. Note we duplicate AIX's requirement that 4020 * the file be open for writing. 4021 */ 4022 /* ARGSUSED */ 4023 int 4024 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4025 { 4026 /* { 4027 syscallarg(int) fd; 4028 syscallarg(int) flags; 4029 syscallarg(off_t) start; 4030 syscallarg(off_t) length; 4031 } */ 4032 struct vnode *vp; 4033 file_t *fp; 4034 int flags, nflags; 4035 off_t s, e, len; 4036 int error; 4037 4038 /* fd_getvnode() will use the descriptor for us */ 4039 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4040 return (error); 4041 4042 if ((fp->f_flag & FWRITE) == 0) { 4043 error = EBADF; 4044 goto out; 4045 } 4046 4047 flags = SCARG(uap, flags); 4048 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4049 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4050 error = EINVAL; 4051 goto out; 4052 } 4053 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4054 if (flags & FDATASYNC) 4055 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4056 else 4057 nflags = FSYNC_WAIT; 4058 if (flags & FDISKSYNC) 4059 nflags |= FSYNC_CACHE; 4060 4061 len = SCARG(uap, length); 4062 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4063 if (len) { 4064 s = SCARG(uap, start); 4065 e = s + len; 4066 if (e < s) { 4067 error = EINVAL; 4068 goto out; 4069 } 4070 } else { 4071 e = 0; 4072 s = 0; 4073 } 4074 4075 vp = fp->f_data; 4076 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4077 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4078 VOP_UNLOCK(vp); 4079 out: 4080 fd_putfile(SCARG(uap, fd)); 4081 return (error); 4082 } 4083 4084 /* 4085 * Sync the data of an open file. 4086 */ 4087 /* ARGSUSED */ 4088 int 4089 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4090 { 4091 /* { 4092 syscallarg(int) fd; 4093 } */ 4094 struct vnode *vp; 4095 file_t *fp; 4096 int error; 4097 4098 /* fd_getvnode() will use the descriptor for us */ 4099 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4100 return (error); 4101 if ((fp->f_flag & FWRITE) == 0) { 4102 fd_putfile(SCARG(uap, fd)); 4103 return (EBADF); 4104 } 4105 vp = fp->f_data; 4106 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4107 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4108 VOP_UNLOCK(vp); 4109 fd_putfile(SCARG(uap, fd)); 4110 return (error); 4111 } 4112 4113 /* 4114 * Rename files, (standard) BSD semantics frontend. 4115 */ 4116 /* ARGSUSED */ 4117 int 4118 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4119 { 4120 /* { 4121 syscallarg(const char *) from; 4122 syscallarg(const char *) to; 4123 } */ 4124 4125 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4126 SCARG(uap, to), UIO_USERSPACE, 0)); 4127 } 4128 4129 int 4130 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4131 register_t *retval) 4132 { 4133 /* { 4134 syscallarg(int) fromfd; 4135 syscallarg(const char *) from; 4136 syscallarg(int) tofd; 4137 syscallarg(const char *) to; 4138 } */ 4139 4140 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4141 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4142 } 4143 4144 /* 4145 * Rename files, POSIX semantics frontend. 4146 */ 4147 /* ARGSUSED */ 4148 int 4149 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4150 { 4151 /* { 4152 syscallarg(const char *) from; 4153 syscallarg(const char *) to; 4154 } */ 4155 4156 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4157 SCARG(uap, to), UIO_USERSPACE, 1)); 4158 } 4159 4160 /* 4161 * Rename files. Source and destination must either both be directories, 4162 * or both not be directories. If target is a directory, it must be empty. 4163 * If `from' and `to' refer to the same object, the value of the `retain' 4164 * argument is used to determine whether `from' will be 4165 * 4166 * (retain == 0) deleted unless `from' and `to' refer to the same 4167 * object in the file system's name space (BSD). 4168 * (retain == 1) always retained (POSIX). 4169 * 4170 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4171 */ 4172 int 4173 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4174 { 4175 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4176 } 4177 4178 static int 4179 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4180 const char *to, enum uio_seg seg, int retain) 4181 { 4182 struct pathbuf *fpb, *tpb; 4183 struct nameidata fnd, tnd; 4184 struct vnode *fdvp, *fvp; 4185 struct vnode *tdvp, *tvp; 4186 struct mount *mp, *tmp; 4187 int error; 4188 4189 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4190 KASSERT(from != NULL); 4191 KASSERT(to != NULL); 4192 4193 error = pathbuf_maybe_copyin(from, seg, &fpb); 4194 if (error) 4195 goto out0; 4196 KASSERT(fpb != NULL); 4197 4198 error = pathbuf_maybe_copyin(to, seg, &tpb); 4199 if (error) 4200 goto out1; 4201 KASSERT(tpb != NULL); 4202 4203 /* 4204 * Lookup from. 4205 * 4206 * XXX LOCKPARENT is wrong because we don't actually want it 4207 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4208 * insane, so for the time being we need to leave it like this. 4209 */ 4210 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT | INRENAME), fpb); 4211 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4212 goto out2; 4213 4214 /* 4215 * Pull out the important results of the lookup, fdvp and fvp. 4216 * Of course, fvp is bogus because we're about to unlock fdvp. 4217 */ 4218 fdvp = fnd.ni_dvp; 4219 fvp = fnd.ni_vp; 4220 KASSERT(fdvp != NULL); 4221 KASSERT(fvp != NULL); 4222 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4223 4224 /* 4225 * Make sure neither fdvp nor fvp is locked. 4226 */ 4227 if (fdvp != fvp) 4228 VOP_UNLOCK(fdvp); 4229 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4230 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4231 4232 /* 4233 * Reject renaming `.' and `..'. Can't do this until after 4234 * namei because we need namei's parsing to find the final 4235 * component name. (namei should just leave us with the final 4236 * component name and not look it up itself, but anyway...) 4237 * 4238 * This was here before because we used to relookup from 4239 * instead of to and relookup requires the caller to check 4240 * this, but now file systems may depend on this check, so we 4241 * must retain it until the file systems are all rototilled. 4242 */ 4243 if (((fnd.ni_cnd.cn_namelen == 1) && 4244 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4245 ((fnd.ni_cnd.cn_namelen == 2) && 4246 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4247 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4248 error = EINVAL; /* XXX EISDIR? */ 4249 goto abort0; 4250 } 4251 4252 /* 4253 * Lookup to. 4254 * 4255 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4256 * fvp here to decide whether to add CREATEDIR is a load of 4257 * bollocks because fvp might be the wrong node by now, since 4258 * fdvp is unlocked. 4259 * 4260 * XXX Why not pass CREATEDIR always? 4261 */ 4262 NDINIT(&tnd, RENAME, 4263 (LOCKPARENT | NOCACHE | TRYEMULROOT | INRENAME | 4264 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4265 tpb); 4266 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4267 goto abort0; 4268 4269 /* 4270 * Pull out the important results of the lookup, tdvp and tvp. 4271 * Of course, tvp is bogus because we're about to unlock tdvp. 4272 */ 4273 tdvp = tnd.ni_dvp; 4274 tvp = tnd.ni_vp; 4275 KASSERT(tdvp != NULL); 4276 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4277 4278 /* 4279 * Make sure neither tdvp nor tvp is locked. 4280 */ 4281 if (tdvp != tvp) 4282 VOP_UNLOCK(tdvp); 4283 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4284 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4285 4286 /* 4287 * Reject renaming onto `.' or `..'. relookup is unhappy with 4288 * these, which is why we must do this here. Once upon a time 4289 * we relooked up from instead of to, and consequently didn't 4290 * need this check, but now that we relookup to instead of 4291 * from, we need this; and we shall need it forever forward 4292 * until the VOP_RENAME protocol changes, because file systems 4293 * will no doubt begin to depend on this check. 4294 */ 4295 if (((tnd.ni_cnd.cn_namelen == 1) && 4296 (tnd.ni_cnd.cn_nameptr[0] == '.')) || 4297 ((tnd.ni_cnd.cn_namelen == 2) && 4298 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4299 (tnd.ni_cnd.cn_nameptr[1] == '.'))) { 4300 error = EINVAL; /* XXX EISDIR? */ 4301 goto abort1; 4302 } 4303 4304 /* 4305 * Get the mount point. If the file system has been unmounted, 4306 * which it may be because we're not holding any vnode locks, 4307 * then v_mount will be NULL. We're not really supposed to 4308 * read v_mount without holding the vnode lock, but since we 4309 * have fdvp referenced, if fdvp->v_mount changes then at worst 4310 * it will be set to NULL, not changed to another mount point. 4311 * And, of course, since it is up to the file system to 4312 * determine the real lock order, we can't lock both fdvp and 4313 * tdvp at the same time. 4314 */ 4315 mp = fdvp->v_mount; 4316 if (mp == NULL) { 4317 error = ENOENT; 4318 goto abort1; 4319 } 4320 4321 /* 4322 * Make sure the mount points match. Again, although we don't 4323 * hold any vnode locks, the v_mount fields may change -- but 4324 * at worst they will change to NULL, so this will never become 4325 * a cross-device rename, because we hold vnode references. 4326 * 4327 * XXX Because nothing is locked and the compiler may reorder 4328 * things here, unmounting the file system at an inopportune 4329 * moment may cause rename to fail with ENXDEV when it really 4330 * should fail with ENOENT. 4331 */ 4332 tmp = tdvp->v_mount; 4333 if (tmp == NULL) { 4334 error = ENOENT; 4335 goto abort1; 4336 } 4337 4338 if (mp != tmp) { 4339 error = EXDEV; 4340 goto abort1; 4341 } 4342 4343 /* 4344 * Take the vfs rename lock to avoid cross-directory screw cases. 4345 * Nothing is locked currently, so taking this lock is safe. 4346 */ 4347 error = VFS_RENAMELOCK_ENTER(mp); 4348 if (error) 4349 goto abort1; 4350 4351 /* 4352 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4353 * and nothing is locked except for the vfs rename lock. 4354 * 4355 * The next step is a little rain dance to conform to the 4356 * insane lock protocol, even though it does nothing to ward 4357 * off race conditions. 4358 * 4359 * We need tdvp and tvp to be locked. However, because we have 4360 * unlocked tdvp in order to hold no locks while we take the 4361 * vfs rename lock, tvp may be wrong here, and we can't safely 4362 * lock it even if the sensible file systems will just unlock 4363 * it straight away. Consequently, we must lock tdvp and then 4364 * relookup tvp to get it locked. 4365 * 4366 * Finally, because the VOP_RENAME protocol is brain-damaged 4367 * and various file systems insanely depend on the semantics of 4368 * this brain damage, the lookup of to must be the last lookup 4369 * before VOP_RENAME. 4370 */ 4371 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4372 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4373 if (error) 4374 goto abort2; 4375 4376 /* 4377 * Drop the old tvp and pick up the new one -- which might be 4378 * the same, but that doesn't matter to us. After this, tdvp 4379 * and tvp should both be locked. 4380 */ 4381 if (tvp != NULL) 4382 vrele(tvp); 4383 tvp = tnd.ni_vp; 4384 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4385 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4386 4387 /* 4388 * The old do_sys_rename had various consistency checks here 4389 * involving fvp and tvp. fvp is bogus already here, and tvp 4390 * will become bogus soon in any sensible file system, so the 4391 * only purpose in putting these checks here is to give lip 4392 * service to these screw cases and to acknowledge that they 4393 * exist, not actually to handle them, but here you go 4394 * anyway... 4395 */ 4396 4397 /* 4398 * Acknowledge that directories and non-directories aren't 4399 * suposed to mix. 4400 */ 4401 if (tvp != NULL) { 4402 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4403 error = ENOTDIR; 4404 goto abort3; 4405 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4406 error = EISDIR; 4407 goto abort3; 4408 } 4409 } 4410 4411 /* 4412 * Acknowledge some random screw case, among the dozens that 4413 * might arise. 4414 */ 4415 if (fvp == tdvp) { 4416 error = EINVAL; 4417 goto abort3; 4418 } 4419 4420 /* 4421 * Acknowledge that POSIX has a wacky screw case. 4422 * 4423 * XXX Eventually the retain flag needs to be passed on to 4424 * VOP_RENAME. 4425 */ 4426 if (fvp == tvp) { 4427 if (retain) { 4428 error = 0; 4429 goto abort3; 4430 } else if ((fdvp == tdvp) && 4431 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4432 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4433 fnd.ni_cnd.cn_namelen))) { 4434 error = 0; 4435 goto abort3; 4436 } 4437 } 4438 4439 /* 4440 * Make sure veriexec can screw us up. (But a race can screw 4441 * up veriexec, of course -- remember, fvp and (soon) tvp are 4442 * bogus.) 4443 */ 4444 #if NVERIEXEC > 0 4445 { 4446 char *f1, *f2; 4447 size_t f1_len; 4448 size_t f2_len; 4449 4450 f1_len = fnd.ni_cnd.cn_namelen + 1; 4451 f1 = kmem_alloc(f1_len, KM_SLEEP); 4452 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4453 4454 f2_len = tnd.ni_cnd.cn_namelen + 1; 4455 f2 = kmem_alloc(f2_len, KM_SLEEP); 4456 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4457 4458 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4459 4460 kmem_free(f1, f1_len); 4461 kmem_free(f2, f2_len); 4462 4463 if (error) 4464 goto abort3; 4465 } 4466 #endif /* NVERIEXEC > 0 */ 4467 4468 /* 4469 * All ready. Incant the rename vop. 4470 */ 4471 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4472 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4473 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4474 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4475 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4476 4477 /* 4478 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4479 * tdvp and tvp. But we can't assert any of that. 4480 */ 4481 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4482 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4483 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4484 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4485 4486 /* 4487 * So all we have left to do is to drop the rename lock and 4488 * destroy the pathbufs. 4489 */ 4490 VFS_RENAMELOCK_EXIT(mp); 4491 goto out2; 4492 4493 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4494 VOP_UNLOCK(tvp); 4495 abort2: VOP_UNLOCK(tdvp); 4496 VFS_RENAMELOCK_EXIT(mp); 4497 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4498 vrele(tdvp); 4499 if (tvp != NULL) 4500 vrele(tvp); 4501 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4502 vrele(fdvp); 4503 vrele(fvp); 4504 out2: pathbuf_destroy(tpb); 4505 out1: pathbuf_destroy(fpb); 4506 out0: return error; 4507 } 4508 4509 /* 4510 * Make a directory file. 4511 */ 4512 /* ARGSUSED */ 4513 int 4514 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4515 { 4516 /* { 4517 syscallarg(const char *) path; 4518 syscallarg(int) mode; 4519 } */ 4520 4521 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4522 SCARG(uap, mode), UIO_USERSPACE); 4523 } 4524 4525 int 4526 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4527 register_t *retval) 4528 { 4529 /* { 4530 syscallarg(int) fd; 4531 syscallarg(const char *) path; 4532 syscallarg(int) mode; 4533 } */ 4534 4535 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4536 SCARG(uap, mode), UIO_USERSPACE); 4537 } 4538 4539 4540 int 4541 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4542 { 4543 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4544 } 4545 4546 static int 4547 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4548 enum uio_seg seg) 4549 { 4550 struct proc *p = curlwp->l_proc; 4551 struct vnode *vp; 4552 struct vattr vattr; 4553 int error; 4554 struct pathbuf *pb; 4555 struct nameidata nd; 4556 4557 KASSERT(l != NULL || fdat == AT_FDCWD); 4558 4559 /* XXX bollocks, should pass in a pathbuf */ 4560 error = pathbuf_maybe_copyin(path, seg, &pb); 4561 if (error) { 4562 return error; 4563 } 4564 4565 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4566 4567 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4568 pathbuf_destroy(pb); 4569 return (error); 4570 } 4571 vp = nd.ni_vp; 4572 if (vp != NULL) { 4573 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4574 if (nd.ni_dvp == vp) 4575 vrele(nd.ni_dvp); 4576 else 4577 vput(nd.ni_dvp); 4578 vrele(vp); 4579 pathbuf_destroy(pb); 4580 return (EEXIST); 4581 } 4582 vattr_null(&vattr); 4583 vattr.va_type = VDIR; 4584 /* We will read cwdi->cwdi_cmask unlocked. */ 4585 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4586 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4587 if (!error) 4588 vput(nd.ni_vp); 4589 pathbuf_destroy(pb); 4590 return (error); 4591 } 4592 4593 /* 4594 * Remove a directory file. 4595 */ 4596 /* ARGSUSED */ 4597 int 4598 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4599 { 4600 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4601 AT_REMOVEDIR, UIO_USERSPACE); 4602 } 4603 4604 /* 4605 * Read a block of directory entries in a file system independent format. 4606 */ 4607 int 4608 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4609 { 4610 /* { 4611 syscallarg(int) fd; 4612 syscallarg(char *) buf; 4613 syscallarg(size_t) count; 4614 } */ 4615 file_t *fp; 4616 int error, done; 4617 4618 /* fd_getvnode() will use the descriptor for us */ 4619 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4620 return (error); 4621 if ((fp->f_flag & FREAD) == 0) { 4622 error = EBADF; 4623 goto out; 4624 } 4625 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4626 SCARG(uap, count), &done, l, 0, 0); 4627 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4628 *retval = done; 4629 out: 4630 fd_putfile(SCARG(uap, fd)); 4631 return (error); 4632 } 4633 4634 /* 4635 * Set the mode mask for creation of filesystem nodes. 4636 */ 4637 int 4638 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4639 { 4640 /* { 4641 syscallarg(mode_t) newmask; 4642 } */ 4643 struct proc *p = l->l_proc; 4644 struct cwdinfo *cwdi; 4645 4646 /* 4647 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4648 * important is that we serialize changes to the mask. The 4649 * rw_exit() will issue a write memory barrier on our behalf, 4650 * and force the changes out to other CPUs (as it must use an 4651 * atomic operation, draining the local CPU's store buffers). 4652 */ 4653 cwdi = p->p_cwdi; 4654 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4655 *retval = cwdi->cwdi_cmask; 4656 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4657 rw_exit(&cwdi->cwdi_lock); 4658 4659 return (0); 4660 } 4661 4662 int 4663 dorevoke(struct vnode *vp, kauth_cred_t cred) 4664 { 4665 struct vattr vattr; 4666 int error, fs_decision; 4667 4668 vn_lock(vp, LK_SHARED | LK_RETRY); 4669 error = VOP_GETATTR(vp, &vattr, cred); 4670 VOP_UNLOCK(vp); 4671 if (error != 0) 4672 return error; 4673 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4674 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4675 fs_decision); 4676 if (!error) 4677 VOP_REVOKE(vp, REVOKEALL); 4678 return (error); 4679 } 4680 4681 /* 4682 * Void all references to file by ripping underlying filesystem 4683 * away from vnode. 4684 */ 4685 /* ARGSUSED */ 4686 int 4687 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4688 { 4689 /* { 4690 syscallarg(const char *) path; 4691 } */ 4692 struct vnode *vp; 4693 int error; 4694 4695 error = namei_simple_user(SCARG(uap, path), 4696 NSM_FOLLOW_TRYEMULROOT, &vp); 4697 if (error != 0) 4698 return (error); 4699 error = dorevoke(vp, l->l_cred); 4700 vrele(vp); 4701 return (error); 4702 } 4703
Indexes created Thu Sep 25 16:09:42 GMT 2025