vfs_syscalls.c revision 1.440.2.4
1 /* $NetBSD: vfs_syscalls.c,v 1.440.2.4 2013/01/16 05:33:45 yamt 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 1.440.2.4 2013/01/16 05:33:45 yamt 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 vn_lock(dfp->f_data, LK_EXCLUSIVE); 186 error = VOP_ACCESS(dfp->f_data, VEXEC, l->l_cred); 187 VOP_UNLOCK(dfp->f_data); 188 if (error) 189 goto cleanup; 190 } 191 192 NDAT(ndp, dfp->f_data); 193 } 194 195 error = namei(ndp); 196 197 cleanup: 198 if (fdat != AT_FDCWD) 199 fd_putfile(fdat); 200 out: 201 return error; 202 } 203 204 static int 205 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 206 namei_simple_flags_t sflags, struct vnode **vp_ret) 207 { 208 file_t *dfp; 209 struct vnode *dvp; 210 int error; 211 212 if (fdat != AT_FDCWD) { 213 if ((error = fd_getvnode(fdat, &dfp)) != 0) 214 goto out; 215 216 if (!(dfp->f_flag & FSEARCH)) { 217 vn_lock(dfp->f_data, LK_EXCLUSIVE); 218 error = VOP_ACCESS(dfp->f_data, VEXEC, l->l_cred); 219 VOP_UNLOCK(dfp->f_data); 220 if (error) 221 goto cleanup; 222 } 223 224 dvp = dfp->f_data; 225 } else { 226 dvp = NULL; 227 } 228 229 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 230 231 cleanup: 232 if (fdat != AT_FDCWD) 233 fd_putfile(fdat); 234 out: 235 return error; 236 } 237 238 static int 239 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 240 { 241 int error; 242 243 fp->f_flag = flags & FMASK; 244 fp->f_type = DTYPE_VNODE; 245 fp->f_ops = &vnops; 246 fp->f_data = vp; 247 248 if (flags & (O_EXLOCK | O_SHLOCK)) { 249 struct flock lf; 250 int type; 251 252 lf.l_whence = SEEK_SET; 253 lf.l_start = 0; 254 lf.l_len = 0; 255 if (flags & O_EXLOCK) 256 lf.l_type = F_WRLCK; 257 else 258 lf.l_type = F_RDLCK; 259 type = F_FLOCK; 260 if ((flags & FNONBLOCK) == 0) 261 type |= F_WAIT; 262 VOP_UNLOCK(vp); 263 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 264 if (error) { 265 (void) vn_close(vp, fp->f_flag, fp->f_cred); 266 fd_abort(l->l_proc, fp, indx); 267 return error; 268 } 269 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 270 atomic_or_uint(&fp->f_flag, FHASLOCK); 271 } 272 if (flags & O_CLOEXEC) 273 fd_set_exclose(l, indx, true); 274 return 0; 275 } 276 277 static int 278 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 279 void *data, size_t *data_len) 280 { 281 struct mount *mp; 282 int error = 0, saved_flags; 283 284 mp = vp->v_mount; 285 saved_flags = mp->mnt_flag; 286 287 /* We can operate only on VV_ROOT nodes. */ 288 if ((vp->v_vflag & VV_ROOT) == 0) { 289 error = EINVAL; 290 goto out; 291 } 292 293 /* 294 * We only allow the filesystem to be reloaded if it 295 * is currently mounted read-only. Additionally, we 296 * prevent read-write to read-only downgrades. 297 */ 298 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 299 (mp->mnt_flag & MNT_RDONLY) == 0 && 300 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 301 error = EOPNOTSUPP; /* Needs translation */ 302 goto out; 303 } 304 305 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 306 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 307 if (error) 308 goto out; 309 310 if (vfs_busy(mp, NULL)) { 311 error = EPERM; 312 goto out; 313 } 314 315 mutex_enter(&mp->mnt_updating); 316 317 mp->mnt_flag &= ~MNT_OP_FLAGS; 318 mp->mnt_flag |= flags & MNT_OP_FLAGS; 319 320 /* 321 * Set the mount level flags. 322 */ 323 if (flags & MNT_RDONLY) 324 mp->mnt_flag |= MNT_RDONLY; 325 else if (mp->mnt_flag & MNT_RDONLY) 326 mp->mnt_iflag |= IMNT_WANTRDWR; 327 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 328 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 329 error = VFS_MOUNT(mp, path, data, data_len); 330 331 if (error && data != NULL) { 332 int error2; 333 334 /* 335 * Update failed; let's try and see if it was an 336 * export request. For compat with 3.0 and earlier. 337 */ 338 error2 = vfs_hooks_reexport(mp, path, data); 339 340 /* 341 * Only update error code if the export request was 342 * understood but some problem occurred while 343 * processing it. 344 */ 345 if (error2 != EJUSTRETURN) 346 error = error2; 347 } 348 349 if (mp->mnt_iflag & IMNT_WANTRDWR) 350 mp->mnt_flag &= ~MNT_RDONLY; 351 if (error) 352 mp->mnt_flag = saved_flags; 353 mp->mnt_flag &= ~MNT_OP_FLAGS; 354 mp->mnt_iflag &= ~IMNT_WANTRDWR; 355 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 356 if (mp->mnt_syncer == NULL) 357 error = vfs_allocate_syncvnode(mp); 358 } else { 359 if (mp->mnt_syncer != NULL) 360 vfs_deallocate_syncvnode(mp); 361 } 362 mutex_exit(&mp->mnt_updating); 363 vfs_unbusy(mp, false, NULL); 364 365 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 366 (flags & MNT_EXTATTR)) { 367 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 368 NULL, 0, NULL) != 0) { 369 printf("%s: failed to start extattr, error = %d", 370 mp->mnt_stat.f_mntonname, error); 371 mp->mnt_flag &= ~MNT_EXTATTR; 372 } 373 } 374 375 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 376 !(flags & MNT_EXTATTR)) { 377 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 378 NULL, 0, NULL) != 0) { 379 printf("%s: failed to stop extattr, error = %d", 380 mp->mnt_stat.f_mntonname, error); 381 mp->mnt_flag |= MNT_RDONLY; 382 } 383 } 384 out: 385 return (error); 386 } 387 388 static int 389 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 390 { 391 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 392 int error; 393 394 /* Copy file-system type from userspace. */ 395 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 396 if (error) { 397 /* 398 * Historically, filesystem types were identified by numbers. 399 * If we get an integer for the filesystem type instead of a 400 * string, we check to see if it matches one of the historic 401 * filesystem types. 402 */ 403 u_long fsindex = (u_long)fstype; 404 if (fsindex >= nmountcompatnames || 405 mountcompatnames[fsindex] == NULL) 406 return ENODEV; 407 strlcpy(fstypename, mountcompatnames[fsindex], 408 sizeof(fstypename)); 409 } 410 411 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 412 if (strcmp(fstypename, "ufs") == 0) 413 fstypename[0] = 'f'; 414 415 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 416 return 0; 417 418 /* If we can autoload a vfs module, try again */ 419 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 420 421 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 422 return 0; 423 424 return ENODEV; 425 } 426 427 static int 428 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 429 void *data, size_t *data_len) 430 { 431 struct mount *mp; 432 int error; 433 434 /* If MNT_GETARGS is specified, it should be the only flag. */ 435 if (flags & ~MNT_GETARGS) 436 return EINVAL; 437 438 mp = vp->v_mount; 439 440 /* XXX: probably some notion of "can see" here if we want isolation. */ 441 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 442 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 443 if (error) 444 return error; 445 446 if ((vp->v_vflag & VV_ROOT) == 0) 447 return EINVAL; 448 449 if (vfs_busy(mp, NULL)) 450 return EPERM; 451 452 mutex_enter(&mp->mnt_updating); 453 mp->mnt_flag &= ~MNT_OP_FLAGS; 454 mp->mnt_flag |= MNT_GETARGS; 455 error = VFS_MOUNT(mp, path, data, data_len); 456 mp->mnt_flag &= ~MNT_OP_FLAGS; 457 mutex_exit(&mp->mnt_updating); 458 459 vfs_unbusy(mp, false, NULL); 460 return (error); 461 } 462 463 int 464 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 465 { 466 /* { 467 syscallarg(const char *) type; 468 syscallarg(const char *) path; 469 syscallarg(int) flags; 470 syscallarg(void *) data; 471 syscallarg(size_t) data_len; 472 } */ 473 474 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 475 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 476 SCARG(uap, data_len), retval); 477 } 478 479 int 480 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 481 const char *path, int flags, void *data, enum uio_seg data_seg, 482 size_t data_len, register_t *retval) 483 { 484 struct vnode *vp; 485 void *data_buf = data; 486 bool vfsopsrele = false; 487 int error; 488 489 /* XXX: The calling convention of this routine is totally bizarre */ 490 if (vfsops) 491 vfsopsrele = true; 492 493 /* 494 * Get vnode to be covered 495 */ 496 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 497 if (error != 0) { 498 vp = NULL; 499 goto done; 500 } 501 502 if (vfsops == NULL) { 503 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 504 vfsops = vp->v_mount->mnt_op; 505 } else { 506 /* 'type' is userspace */ 507 error = mount_get_vfsops(type, &vfsops); 508 if (error != 0) 509 goto done; 510 vfsopsrele = true; 511 } 512 } 513 514 if (data != NULL && data_seg == UIO_USERSPACE) { 515 if (data_len == 0) { 516 /* No length supplied, use default for filesystem */ 517 data_len = vfsops->vfs_min_mount_data; 518 if (data_len > VFS_MAX_MOUNT_DATA) { 519 error = EINVAL; 520 goto done; 521 } 522 /* 523 * Hopefully a longer buffer won't make copyin() fail. 524 * For compatibility with 3.0 and earlier. 525 */ 526 if (flags & MNT_UPDATE 527 && data_len < sizeof (struct mnt_export_args30)) 528 data_len = sizeof (struct mnt_export_args30); 529 } 530 data_buf = kmem_alloc(data_len, KM_SLEEP); 531 532 /* NFS needs the buffer even for mnt_getargs .... */ 533 error = copyin(data, data_buf, data_len); 534 if (error != 0) 535 goto done; 536 } 537 538 if (flags & MNT_GETARGS) { 539 if (data_len == 0) { 540 error = EINVAL; 541 goto done; 542 } 543 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 544 if (error != 0) 545 goto done; 546 if (data_seg == UIO_USERSPACE) 547 error = copyout(data_buf, data, data_len); 548 *retval = data_len; 549 } else if (flags & MNT_UPDATE) { 550 error = mount_update(l, vp, path, flags, data_buf, &data_len); 551 } else { 552 /* Locking is handled internally in mount_domount(). */ 553 KASSERT(vfsopsrele == true); 554 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 555 &data_len); 556 vfsopsrele = false; 557 } 558 559 done: 560 if (vfsopsrele) 561 vfs_delref(vfsops); 562 if (vp != NULL) { 563 vrele(vp); 564 } 565 if (data_buf != data) 566 kmem_free(data_buf, data_len); 567 return (error); 568 } 569 570 /* 571 * Unmount a file system. 572 * 573 * Note: unmount takes a path to the vnode mounted on as argument, 574 * not special file (as before). 575 */ 576 /* ARGSUSED */ 577 int 578 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 579 { 580 /* { 581 syscallarg(const char *) path; 582 syscallarg(int) flags; 583 } */ 584 struct vnode *vp; 585 struct mount *mp; 586 int error; 587 struct pathbuf *pb; 588 struct nameidata nd; 589 590 error = pathbuf_copyin(SCARG(uap, path), &pb); 591 if (error) { 592 return error; 593 } 594 595 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 596 if ((error = namei(&nd)) != 0) { 597 pathbuf_destroy(pb); 598 return error; 599 } 600 vp = nd.ni_vp; 601 pathbuf_destroy(pb); 602 603 mp = vp->v_mount; 604 atomic_inc_uint(&mp->mnt_refcnt); 605 VOP_UNLOCK(vp); 606 607 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 608 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 609 if (error) { 610 vrele(vp); 611 vfs_destroy(mp); 612 return (error); 613 } 614 615 /* 616 * Don't allow unmounting the root file system. 617 */ 618 if (mp->mnt_flag & MNT_ROOTFS) { 619 vrele(vp); 620 vfs_destroy(mp); 621 return (EINVAL); 622 } 623 624 /* 625 * Must be the root of the filesystem 626 */ 627 if ((vp->v_vflag & VV_ROOT) == 0) { 628 vrele(vp); 629 vfs_destroy(mp); 630 return (EINVAL); 631 } 632 633 vrele(vp); 634 error = dounmount(mp, SCARG(uap, flags), l); 635 vfs_destroy(mp); 636 return error; 637 } 638 639 /* 640 * Sync each mounted filesystem. 641 */ 642 #ifdef DEBUG 643 int syncprt = 0; 644 struct ctldebug debug0 = { "syncprt", &syncprt }; 645 #endif 646 647 void 648 do_sys_sync(struct lwp *l) 649 { 650 struct mount *mp, *nmp; 651 int asyncflag; 652 653 mutex_enter(&mountlist_lock); 654 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 655 mp = nmp) { 656 if (vfs_busy(mp, &nmp)) { 657 continue; 658 } 659 mutex_enter(&mp->mnt_updating); 660 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 661 asyncflag = mp->mnt_flag & MNT_ASYNC; 662 mp->mnt_flag &= ~MNT_ASYNC; 663 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 664 if (asyncflag) 665 mp->mnt_flag |= MNT_ASYNC; 666 } 667 mutex_exit(&mp->mnt_updating); 668 vfs_unbusy(mp, false, &nmp); 669 } 670 mutex_exit(&mountlist_lock); 671 #ifdef DEBUG 672 if (syncprt) 673 vfs_bufstats(); 674 #endif /* DEBUG */ 675 } 676 677 /* ARGSUSED */ 678 int 679 sys_sync(struct lwp *l, const void *v, register_t *retval) 680 { 681 do_sys_sync(l); 682 return (0); 683 } 684 685 686 /* 687 * Access or change filesystem quotas. 688 * 689 * (this is really 14 different calls bundled into one) 690 */ 691 692 static int 693 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 694 { 695 struct quotastat info_k; 696 int error; 697 698 /* ensure any padding bytes are cleared */ 699 memset(&info_k, 0, sizeof(info_k)); 700 701 error = vfs_quotactl_stat(mp, &info_k); 702 if (error) { 703 return error; 704 } 705 706 return copyout(&info_k, info_u, sizeof(info_k)); 707 } 708 709 static int 710 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 711 struct quotaidtypestat *info_u) 712 { 713 struct quotaidtypestat info_k; 714 int error; 715 716 /* ensure any padding bytes are cleared */ 717 memset(&info_k, 0, sizeof(info_k)); 718 719 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 720 if (error) { 721 return error; 722 } 723 724 return copyout(&info_k, info_u, sizeof(info_k)); 725 } 726 727 static int 728 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 729 struct quotaobjtypestat *info_u) 730 { 731 struct quotaobjtypestat info_k; 732 int error; 733 734 /* ensure any padding bytes are cleared */ 735 memset(&info_k, 0, sizeof(info_k)); 736 737 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 738 if (error) { 739 return error; 740 } 741 742 return copyout(&info_k, info_u, sizeof(info_k)); 743 } 744 745 static int 746 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 747 struct quotaval *val_u) 748 { 749 struct quotakey key_k; 750 struct quotaval val_k; 751 int error; 752 753 /* ensure any padding bytes are cleared */ 754 memset(&val_k, 0, sizeof(val_k)); 755 756 error = copyin(key_u, &key_k, sizeof(key_k)); 757 if (error) { 758 return error; 759 } 760 761 error = vfs_quotactl_get(mp, &key_k, &val_k); 762 if (error) { 763 return error; 764 } 765 766 return copyout(&val_k, val_u, sizeof(val_k)); 767 } 768 769 static int 770 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 771 const struct quotaval *val_u) 772 { 773 struct quotakey key_k; 774 struct quotaval val_k; 775 int error; 776 777 error = copyin(key_u, &key_k, sizeof(key_k)); 778 if (error) { 779 return error; 780 } 781 782 error = copyin(val_u, &val_k, sizeof(val_k)); 783 if (error) { 784 return error; 785 } 786 787 return vfs_quotactl_put(mp, &key_k, &val_k); 788 } 789 790 static int 791 do_sys_quotactl_delete(struct mount *mp, const struct quotakey *key_u) 792 { 793 struct quotakey key_k; 794 int error; 795 796 error = copyin(key_u, &key_k, sizeof(key_k)); 797 if (error) { 798 return error; 799 } 800 801 return vfs_quotactl_delete(mp, &key_k); 802 } 803 804 static int 805 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 806 { 807 struct quotakcursor cursor_k; 808 int error; 809 810 /* ensure any padding bytes are cleared */ 811 memset(&cursor_k, 0, sizeof(cursor_k)); 812 813 error = vfs_quotactl_cursoropen(mp, &cursor_k); 814 if (error) { 815 return error; 816 } 817 818 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 819 } 820 821 static int 822 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 823 { 824 struct quotakcursor cursor_k; 825 int error; 826 827 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 828 if (error) { 829 return error; 830 } 831 832 return vfs_quotactl_cursorclose(mp, &cursor_k); 833 } 834 835 static int 836 do_sys_quotactl_cursorskipidtype(struct mount *mp, 837 struct quotakcursor *cursor_u, int idtype) 838 { 839 struct quotakcursor cursor_k; 840 int error; 841 842 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 843 if (error) { 844 return error; 845 } 846 847 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 848 if (error) { 849 return error; 850 } 851 852 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 853 } 854 855 static int 856 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 857 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 858 unsigned *ret_u) 859 { 860 #define CGET_STACK_MAX 8 861 struct quotakcursor cursor_k; 862 struct quotakey stackkeys[CGET_STACK_MAX]; 863 struct quotaval stackvals[CGET_STACK_MAX]; 864 struct quotakey *keys_k; 865 struct quotaval *vals_k; 866 unsigned ret_k; 867 int error; 868 869 if (maxnum > 128) { 870 maxnum = 128; 871 } 872 873 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 874 if (error) { 875 return error; 876 } 877 878 if (maxnum <= CGET_STACK_MAX) { 879 keys_k = stackkeys; 880 vals_k = stackvals; 881 /* ensure any padding bytes are cleared */ 882 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 883 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 884 } else { 885 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 886 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 887 } 888 889 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 890 &ret_k); 891 if (error) { 892 goto fail; 893 } 894 895 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 896 if (error) { 897 goto fail; 898 } 899 900 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 901 if (error) { 902 goto fail; 903 } 904 905 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 906 if (error) { 907 goto fail; 908 } 909 910 /* do last to maximize the chance of being able to recover a failure */ 911 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 912 913 fail: 914 if (keys_k != stackkeys) { 915 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 916 } 917 if (vals_k != stackvals) { 918 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 919 } 920 return error; 921 } 922 923 static int 924 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 925 int *ret_u) 926 { 927 struct quotakcursor cursor_k; 928 int ret_k; 929 int error; 930 931 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 932 if (error) { 933 return error; 934 } 935 936 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 937 if (error) { 938 return error; 939 } 940 941 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 942 if (error) { 943 return error; 944 } 945 946 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 947 } 948 949 static int 950 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 951 { 952 struct quotakcursor cursor_k; 953 int error; 954 955 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 956 if (error) { 957 return error; 958 } 959 960 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 961 if (error) { 962 return error; 963 } 964 965 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 966 } 967 968 static int 969 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 970 { 971 char *path_k; 972 int error; 973 974 /* XXX this should probably be a struct pathbuf */ 975 path_k = PNBUF_GET(); 976 error = copyin(path_u, path_k, PATH_MAX); 977 if (error) { 978 PNBUF_PUT(path_k); 979 return error; 980 } 981 982 error = vfs_quotactl_quotaon(mp, idtype, path_k); 983 984 PNBUF_PUT(path_k); 985 return error; 986 } 987 988 static int 989 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 990 { 991 return vfs_quotactl_quotaoff(mp, idtype); 992 } 993 994 int 995 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 996 { 997 struct mount *mp; 998 struct vnode *vp; 999 int error; 1000 1001 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 1002 if (error != 0) 1003 return (error); 1004 mp = vp->v_mount; 1005 1006 switch (args->qc_op) { 1007 case QUOTACTL_STAT: 1008 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 1009 break; 1010 case QUOTACTL_IDTYPESTAT: 1011 error = do_sys_quotactl_idtypestat(mp, 1012 args->u.idtypestat.qc_idtype, 1013 args->u.idtypestat.qc_info); 1014 break; 1015 case QUOTACTL_OBJTYPESTAT: 1016 error = do_sys_quotactl_objtypestat(mp, 1017 args->u.objtypestat.qc_objtype, 1018 args->u.objtypestat.qc_info); 1019 break; 1020 case QUOTACTL_GET: 1021 error = do_sys_quotactl_get(mp, 1022 args->u.get.qc_key, 1023 args->u.get.qc_val); 1024 break; 1025 case QUOTACTL_PUT: 1026 error = do_sys_quotactl_put(mp, 1027 args->u.put.qc_key, 1028 args->u.put.qc_val); 1029 break; 1030 case QUOTACTL_DELETE: 1031 error = do_sys_quotactl_delete(mp, args->u.delete.qc_key); 1032 break; 1033 case QUOTACTL_CURSOROPEN: 1034 error = do_sys_quotactl_cursoropen(mp, 1035 args->u.cursoropen.qc_cursor); 1036 break; 1037 case QUOTACTL_CURSORCLOSE: 1038 error = do_sys_quotactl_cursorclose(mp, 1039 args->u.cursorclose.qc_cursor); 1040 break; 1041 case QUOTACTL_CURSORSKIPIDTYPE: 1042 error = do_sys_quotactl_cursorskipidtype(mp, 1043 args->u.cursorskipidtype.qc_cursor, 1044 args->u.cursorskipidtype.qc_idtype); 1045 break; 1046 case QUOTACTL_CURSORGET: 1047 error = do_sys_quotactl_cursorget(mp, 1048 args->u.cursorget.qc_cursor, 1049 args->u.cursorget.qc_keys, 1050 args->u.cursorget.qc_vals, 1051 args->u.cursorget.qc_maxnum, 1052 args->u.cursorget.qc_ret); 1053 break; 1054 case QUOTACTL_CURSORATEND: 1055 error = do_sys_quotactl_cursoratend(mp, 1056 args->u.cursoratend.qc_cursor, 1057 args->u.cursoratend.qc_ret); 1058 break; 1059 case QUOTACTL_CURSORREWIND: 1060 error = do_sys_quotactl_cursorrewind(mp, 1061 args->u.cursorrewind.qc_cursor); 1062 break; 1063 case QUOTACTL_QUOTAON: 1064 error = do_sys_quotactl_quotaon(mp, 1065 args->u.quotaon.qc_idtype, 1066 args->u.quotaon.qc_quotafile); 1067 break; 1068 case QUOTACTL_QUOTAOFF: 1069 error = do_sys_quotactl_quotaoff(mp, 1070 args->u.quotaoff.qc_idtype); 1071 break; 1072 default: 1073 error = EINVAL; 1074 break; 1075 } 1076 1077 vrele(vp); 1078 return error; 1079 } 1080 1081 /* ARGSUSED */ 1082 int 1083 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1084 register_t *retval) 1085 { 1086 /* { 1087 syscallarg(const char *) path; 1088 syscallarg(struct quotactl_args *) args; 1089 } */ 1090 struct quotactl_args args; 1091 int error; 1092 1093 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1094 if (error) { 1095 return error; 1096 } 1097 1098 return do_sys_quotactl(SCARG(uap, path), &args); 1099 } 1100 1101 int 1102 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1103 int root) 1104 { 1105 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1106 int error = 0; 1107 1108 /* 1109 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1110 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1111 * overrides MNT_NOWAIT. 1112 */ 1113 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1114 (flags != MNT_WAIT && flags != 0)) { 1115 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1116 goto done; 1117 } 1118 1119 /* Get the filesystem stats now */ 1120 memset(sp, 0, sizeof(*sp)); 1121 if ((error = VFS_STATVFS(mp, sp)) != 0) { 1122 return error; 1123 } 1124 1125 if (cwdi->cwdi_rdir == NULL) 1126 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1127 done: 1128 if (cwdi->cwdi_rdir != NULL) { 1129 size_t len; 1130 char *bp; 1131 char c; 1132 char *path = PNBUF_GET(); 1133 1134 bp = path + MAXPATHLEN; 1135 *--bp = '\0'; 1136 rw_enter(&cwdi->cwdi_lock, RW_READER); 1137 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1138 MAXPATHLEN / 2, 0, l); 1139 rw_exit(&cwdi->cwdi_lock); 1140 if (error) { 1141 PNBUF_PUT(path); 1142 return error; 1143 } 1144 len = strlen(bp); 1145 if (len != 1) { 1146 /* 1147 * for mount points that are below our root, we can see 1148 * them, so we fix up the pathname and return them. The 1149 * rest we cannot see, so we don't allow viewing the 1150 * data. 1151 */ 1152 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1153 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1154 (void)strlcpy(sp->f_mntonname, 1155 c == '\0' ? "/" : &sp->f_mntonname[len], 1156 sizeof(sp->f_mntonname)); 1157 } else { 1158 if (root) 1159 (void)strlcpy(sp->f_mntonname, "/", 1160 sizeof(sp->f_mntonname)); 1161 else 1162 error = EPERM; 1163 } 1164 } 1165 PNBUF_PUT(path); 1166 } 1167 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1168 return error; 1169 } 1170 1171 /* 1172 * Get filesystem statistics by path. 1173 */ 1174 int 1175 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1176 { 1177 struct mount *mp; 1178 int error; 1179 struct vnode *vp; 1180 1181 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1182 if (error != 0) 1183 return error; 1184 mp = vp->v_mount; 1185 error = dostatvfs(mp, sb, l, flags, 1); 1186 vrele(vp); 1187 return error; 1188 } 1189 1190 /* ARGSUSED */ 1191 int 1192 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1193 { 1194 /* { 1195 syscallarg(const char *) path; 1196 syscallarg(struct statvfs *) buf; 1197 syscallarg(int) flags; 1198 } */ 1199 struct statvfs *sb; 1200 int error; 1201 1202 sb = STATVFSBUF_GET(); 1203 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1204 if (error == 0) 1205 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1206 STATVFSBUF_PUT(sb); 1207 return error; 1208 } 1209 1210 /* 1211 * Get filesystem statistics by fd. 1212 */ 1213 int 1214 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1215 { 1216 file_t *fp; 1217 struct mount *mp; 1218 int error; 1219 1220 /* fd_getvnode() will use the descriptor for us */ 1221 if ((error = fd_getvnode(fd, &fp)) != 0) 1222 return (error); 1223 mp = ((struct vnode *)fp->f_data)->v_mount; 1224 error = dostatvfs(mp, sb, curlwp, flags, 1); 1225 fd_putfile(fd); 1226 return error; 1227 } 1228 1229 /* ARGSUSED */ 1230 int 1231 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1232 { 1233 /* { 1234 syscallarg(int) fd; 1235 syscallarg(struct statvfs *) buf; 1236 syscallarg(int) flags; 1237 } */ 1238 struct statvfs *sb; 1239 int error; 1240 1241 sb = STATVFSBUF_GET(); 1242 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1243 if (error == 0) 1244 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1245 STATVFSBUF_PUT(sb); 1246 return error; 1247 } 1248 1249 1250 /* 1251 * Get statistics on all filesystems. 1252 */ 1253 int 1254 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1255 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1256 register_t *retval) 1257 { 1258 int root = 0; 1259 struct proc *p = l->l_proc; 1260 struct mount *mp, *nmp; 1261 struct statvfs *sb; 1262 size_t count, maxcount; 1263 int error = 0; 1264 1265 sb = STATVFSBUF_GET(); 1266 maxcount = bufsize / entry_sz; 1267 mutex_enter(&mountlist_lock); 1268 count = 0; 1269 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1270 mp = nmp) { 1271 if (vfs_busy(mp, &nmp)) { 1272 continue; 1273 } 1274 if (sfsp && count < maxcount) { 1275 error = dostatvfs(mp, sb, l, flags, 0); 1276 if (error) { 1277 vfs_unbusy(mp, false, &nmp); 1278 error = 0; 1279 continue; 1280 } 1281 error = copyfn(sb, sfsp, entry_sz); 1282 if (error) { 1283 vfs_unbusy(mp, false, NULL); 1284 goto out; 1285 } 1286 sfsp = (char *)sfsp + entry_sz; 1287 root |= strcmp(sb->f_mntonname, "/") == 0; 1288 } 1289 count++; 1290 vfs_unbusy(mp, false, &nmp); 1291 } 1292 mutex_exit(&mountlist_lock); 1293 1294 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1295 /* 1296 * fake a root entry 1297 */ 1298 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1299 sb, l, flags, 1); 1300 if (error != 0) 1301 goto out; 1302 if (sfsp) { 1303 error = copyfn(sb, sfsp, entry_sz); 1304 if (error != 0) 1305 goto out; 1306 } 1307 count++; 1308 } 1309 if (sfsp && count > maxcount) 1310 *retval = maxcount; 1311 else 1312 *retval = count; 1313 out: 1314 STATVFSBUF_PUT(sb); 1315 return error; 1316 } 1317 1318 int 1319 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1320 { 1321 /* { 1322 syscallarg(struct statvfs *) buf; 1323 syscallarg(size_t) bufsize; 1324 syscallarg(int) flags; 1325 } */ 1326 1327 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1328 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1329 } 1330 1331 /* 1332 * Change current working directory to a given file descriptor. 1333 */ 1334 /* ARGSUSED */ 1335 int 1336 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1337 { 1338 /* { 1339 syscallarg(int) fd; 1340 } */ 1341 struct proc *p = l->l_proc; 1342 struct cwdinfo *cwdi; 1343 struct vnode *vp, *tdp; 1344 struct mount *mp; 1345 file_t *fp; 1346 int error, fd; 1347 1348 /* fd_getvnode() will use the descriptor for us */ 1349 fd = SCARG(uap, fd); 1350 if ((error = fd_getvnode(fd, &fp)) != 0) 1351 return (error); 1352 vp = fp->f_data; 1353 1354 vref(vp); 1355 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1356 if (vp->v_type != VDIR) 1357 error = ENOTDIR; 1358 else 1359 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1360 if (error) { 1361 vput(vp); 1362 goto out; 1363 } 1364 while ((mp = vp->v_mountedhere) != NULL) { 1365 error = vfs_busy(mp, NULL); 1366 vput(vp); 1367 if (error != 0) 1368 goto out; 1369 error = VFS_ROOT(mp, &tdp); 1370 vfs_unbusy(mp, false, NULL); 1371 if (error) 1372 goto out; 1373 vp = tdp; 1374 } 1375 VOP_UNLOCK(vp); 1376 1377 /* 1378 * Disallow changing to a directory not under the process's 1379 * current root directory (if there is one). 1380 */ 1381 cwdi = p->p_cwdi; 1382 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1383 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1384 vrele(vp); 1385 error = EPERM; /* operation not permitted */ 1386 } else { 1387 vrele(cwdi->cwdi_cdir); 1388 cwdi->cwdi_cdir = vp; 1389 } 1390 rw_exit(&cwdi->cwdi_lock); 1391 1392 out: 1393 fd_putfile(fd); 1394 return (error); 1395 } 1396 1397 /* 1398 * Change this process's notion of the root directory to a given file 1399 * descriptor. 1400 */ 1401 int 1402 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1403 { 1404 struct proc *p = l->l_proc; 1405 struct vnode *vp; 1406 file_t *fp; 1407 int error, fd = SCARG(uap, fd); 1408 1409 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1410 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1411 return error; 1412 /* fd_getvnode() will use the descriptor for us */ 1413 if ((error = fd_getvnode(fd, &fp)) != 0) 1414 return error; 1415 vp = fp->f_data; 1416 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1417 if (vp->v_type != VDIR) 1418 error = ENOTDIR; 1419 else 1420 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1421 VOP_UNLOCK(vp); 1422 if (error) 1423 goto out; 1424 vref(vp); 1425 1426 change_root(p->p_cwdi, vp, l); 1427 1428 out: 1429 fd_putfile(fd); 1430 return (error); 1431 } 1432 1433 /* 1434 * Change current working directory (``.''). 1435 */ 1436 /* ARGSUSED */ 1437 int 1438 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1439 { 1440 /* { 1441 syscallarg(const char *) path; 1442 } */ 1443 struct proc *p = l->l_proc; 1444 struct cwdinfo *cwdi; 1445 int error; 1446 struct vnode *vp; 1447 1448 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1449 &vp, l)) != 0) 1450 return (error); 1451 cwdi = p->p_cwdi; 1452 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1453 vrele(cwdi->cwdi_cdir); 1454 cwdi->cwdi_cdir = vp; 1455 rw_exit(&cwdi->cwdi_lock); 1456 return (0); 1457 } 1458 1459 /* 1460 * Change notion of root (``/'') directory. 1461 */ 1462 /* ARGSUSED */ 1463 int 1464 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1465 { 1466 /* { 1467 syscallarg(const char *) path; 1468 } */ 1469 struct proc *p = l->l_proc; 1470 int error; 1471 struct vnode *vp; 1472 1473 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1474 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1475 return (error); 1476 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1477 &vp, l)) != 0) 1478 return (error); 1479 1480 change_root(p->p_cwdi, vp, l); 1481 1482 return (0); 1483 } 1484 1485 /* 1486 * Common routine for chroot and fchroot. 1487 * NB: callers need to properly authorize the change root operation. 1488 */ 1489 void 1490 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1491 { 1492 struct proc *p = l->l_proc; 1493 kauth_cred_t ncred; 1494 1495 ncred = kauth_cred_alloc(); 1496 1497 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1498 if (cwdi->cwdi_rdir != NULL) 1499 vrele(cwdi->cwdi_rdir); 1500 cwdi->cwdi_rdir = vp; 1501 1502 /* 1503 * Prevent escaping from chroot by putting the root under 1504 * the working directory. Silently chdir to / if we aren't 1505 * already there. 1506 */ 1507 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1508 /* 1509 * XXX would be more failsafe to change directory to a 1510 * deadfs node here instead 1511 */ 1512 vrele(cwdi->cwdi_cdir); 1513 vref(vp); 1514 cwdi->cwdi_cdir = vp; 1515 } 1516 rw_exit(&cwdi->cwdi_lock); 1517 1518 /* Get a write lock on the process credential. */ 1519 proc_crmod_enter(); 1520 1521 kauth_cred_clone(p->p_cred, ncred); 1522 kauth_proc_chroot(ncred, p->p_cwdi); 1523 1524 /* Broadcast our credentials to the process and other LWPs. */ 1525 proc_crmod_leave(ncred, p->p_cred, true); 1526 } 1527 1528 /* 1529 * Common routine for chroot and chdir. 1530 * XXX "where" should be enum uio_seg 1531 */ 1532 int 1533 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1534 { 1535 struct pathbuf *pb; 1536 struct nameidata nd; 1537 int error; 1538 1539 error = pathbuf_maybe_copyin(path, where, &pb); 1540 if (error) { 1541 return error; 1542 } 1543 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1544 if ((error = namei(&nd)) != 0) { 1545 pathbuf_destroy(pb); 1546 return error; 1547 } 1548 *vpp = nd.ni_vp; 1549 pathbuf_destroy(pb); 1550 1551 if ((*vpp)->v_type != VDIR) 1552 error = ENOTDIR; 1553 else 1554 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1555 1556 if (error) 1557 vput(*vpp); 1558 else 1559 VOP_UNLOCK(*vpp); 1560 return (error); 1561 } 1562 1563 /* 1564 * Internals of sys_open - path has already been converted into a pathbuf 1565 * (so we can easily reuse this function from other parts of the kernel, 1566 * like posix_spawn post-processing). 1567 */ 1568 static int 1569 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1570 int open_mode, int *fd) 1571 { 1572 struct proc *p = l->l_proc; 1573 struct cwdinfo *cwdi = p->p_cwdi; 1574 file_t *fp; 1575 struct vnode *vp; 1576 int flags, cmode; 1577 int indx, error; 1578 struct nameidata nd; 1579 1580 flags = FFLAGS(open_flags); 1581 if ((flags & (FREAD | FWRITE)) == 0) 1582 return EINVAL; 1583 1584 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1585 return error; 1586 } 1587 1588 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1589 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1590 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1591 if (dvp != NULL) 1592 NDAT(&nd, dvp); 1593 1594 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1595 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1596 fd_abort(p, fp, indx); 1597 if ((error == EDUPFD || error == EMOVEFD) && 1598 l->l_dupfd >= 0 && /* XXX from fdopen */ 1599 (error = 1600 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1601 *fd = indx; 1602 return 0; 1603 } 1604 if (error == ERESTART) 1605 error = EINTR; 1606 return error; 1607 } 1608 1609 l->l_dupfd = 0; 1610 vp = nd.ni_vp; 1611 1612 if ((error = open_setfp(l, fp, vp, indx, flags))) 1613 return error; 1614 1615 VOP_UNLOCK(vp); 1616 *fd = indx; 1617 fd_affix(p, fp, indx); 1618 return 0; 1619 } 1620 1621 int 1622 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1623 { 1624 struct pathbuf *pb; 1625 int error, oflags; 1626 1627 oflags = FFLAGS(open_flags); 1628 if ((oflags & (FREAD | FWRITE)) == 0) 1629 return EINVAL; 1630 1631 pb = pathbuf_create(path); 1632 if (pb == NULL) 1633 return ENOMEM; 1634 1635 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1636 pathbuf_destroy(pb); 1637 1638 return error; 1639 } 1640 1641 /* 1642 * Check permissions, allocate an open file structure, 1643 * and call the device open routine if any. 1644 * XXX implement O_SEARCH 1645 */ 1646 static int 1647 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1648 int mode, int *fd) 1649 { 1650 file_t *dfp = NULL; 1651 struct vnode *dvp = NULL; 1652 struct pathbuf *pb; 1653 int error; 1654 1655 error = pathbuf_copyin(path, &pb); 1656 if (error) 1657 return error; 1658 1659 if (fdat != AT_FDCWD) { 1660 /* fd_getvnode() will use the descriptor for us */ 1661 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1662 goto out; 1663 1664 dvp = dfp->f_data; 1665 1666 if (!(dfp->f_flag & FSEARCH)) { 1667 vn_lock(dfp->f_data, LK_EXCLUSIVE); 1668 error = VOP_ACCESS(dfp->f_data, VEXEC, l->l_cred); 1669 VOP_UNLOCK(dfp->f_data); 1670 if (error) 1671 goto cleanup; 1672 } 1673 } 1674 1675 error = do_open(l, dvp, pb, flags, mode, fd); 1676 1677 cleanup: 1678 if (dfp != NULL) 1679 fd_putfile(fdat); 1680 out: 1681 pathbuf_destroy(pb); 1682 return error; 1683 } 1684 1685 int 1686 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1687 { 1688 /* { 1689 syscallarg(const char *) path; 1690 syscallarg(int) flags; 1691 syscallarg(int) mode; 1692 } */ 1693 int error; 1694 int fd; 1695 1696 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1697 SCARG(uap, flags), SCARG(uap, mode), &fd); 1698 1699 if (error == 0) 1700 *retval = fd; 1701 1702 return error; 1703 } 1704 1705 int 1706 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1707 { 1708 /* { 1709 syscallarg(int) fd; 1710 syscallarg(const char *) path; 1711 syscallarg(int) oflags; 1712 syscallarg(int) mode; 1713 } */ 1714 int error; 1715 int fd; 1716 1717 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1718 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1719 1720 if (error == 0) 1721 *retval = fd; 1722 1723 return error; 1724 } 1725 1726 static void 1727 vfs__fhfree(fhandle_t *fhp) 1728 { 1729 size_t fhsize; 1730 1731 if (fhp == NULL) { 1732 return; 1733 } 1734 fhsize = FHANDLE_SIZE(fhp); 1735 kmem_free(fhp, fhsize); 1736 } 1737 1738 /* 1739 * vfs_composefh: compose a filehandle. 1740 */ 1741 1742 int 1743 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1744 { 1745 struct mount *mp; 1746 struct fid *fidp; 1747 int error; 1748 size_t needfhsize; 1749 size_t fidsize; 1750 1751 mp = vp->v_mount; 1752 fidp = NULL; 1753 if (*fh_size < FHANDLE_SIZE_MIN) { 1754 fidsize = 0; 1755 } else { 1756 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1757 if (fhp != NULL) { 1758 memset(fhp, 0, *fh_size); 1759 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1760 fidp = &fhp->fh_fid; 1761 } 1762 } 1763 error = VFS_VPTOFH(vp, fidp, &fidsize); 1764 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1765 if (error == 0 && *fh_size < needfhsize) { 1766 error = E2BIG; 1767 } 1768 *fh_size = needfhsize; 1769 return error; 1770 } 1771 1772 int 1773 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1774 { 1775 struct mount *mp; 1776 fhandle_t *fhp; 1777 size_t fhsize; 1778 size_t fidsize; 1779 int error; 1780 1781 *fhpp = NULL; 1782 mp = vp->v_mount; 1783 fidsize = 0; 1784 error = VFS_VPTOFH(vp, NULL, &fidsize); 1785 KASSERT(error != 0); 1786 if (error != E2BIG) { 1787 goto out; 1788 } 1789 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1790 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1791 if (fhp == NULL) { 1792 error = ENOMEM; 1793 goto out; 1794 } 1795 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1796 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1797 if (error == 0) { 1798 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1799 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1800 *fhpp = fhp; 1801 } else { 1802 kmem_free(fhp, fhsize); 1803 } 1804 out: 1805 return error; 1806 } 1807 1808 void 1809 vfs_composefh_free(fhandle_t *fhp) 1810 { 1811 1812 vfs__fhfree(fhp); 1813 } 1814 1815 /* 1816 * vfs_fhtovp: lookup a vnode by a filehandle. 1817 */ 1818 1819 int 1820 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1821 { 1822 struct mount *mp; 1823 int error; 1824 1825 *vpp = NULL; 1826 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1827 if (mp == NULL) { 1828 error = ESTALE; 1829 goto out; 1830 } 1831 if (mp->mnt_op->vfs_fhtovp == NULL) { 1832 error = EOPNOTSUPP; 1833 goto out; 1834 } 1835 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1836 out: 1837 return error; 1838 } 1839 1840 /* 1841 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1842 * the needed size. 1843 */ 1844 1845 int 1846 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1847 { 1848 fhandle_t *fhp; 1849 int error; 1850 1851 *fhpp = NULL; 1852 if (fhsize > FHANDLE_SIZE_MAX) { 1853 return EINVAL; 1854 } 1855 if (fhsize < FHANDLE_SIZE_MIN) { 1856 return EINVAL; 1857 } 1858 again: 1859 fhp = kmem_alloc(fhsize, KM_SLEEP); 1860 if (fhp == NULL) { 1861 return ENOMEM; 1862 } 1863 error = copyin(ufhp, fhp, fhsize); 1864 if (error == 0) { 1865 /* XXX this check shouldn't be here */ 1866 if (FHANDLE_SIZE(fhp) == fhsize) { 1867 *fhpp = fhp; 1868 return 0; 1869 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1870 /* 1871 * a kludge for nfsv2 padded handles. 1872 */ 1873 size_t sz; 1874 1875 sz = FHANDLE_SIZE(fhp); 1876 kmem_free(fhp, fhsize); 1877 fhsize = sz; 1878 goto again; 1879 } else { 1880 /* 1881 * userland told us wrong size. 1882 */ 1883 error = EINVAL; 1884 } 1885 } 1886 kmem_free(fhp, fhsize); 1887 return error; 1888 } 1889 1890 void 1891 vfs_copyinfh_free(fhandle_t *fhp) 1892 { 1893 1894 vfs__fhfree(fhp); 1895 } 1896 1897 /* 1898 * Get file handle system call 1899 */ 1900 int 1901 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1902 { 1903 /* { 1904 syscallarg(char *) fname; 1905 syscallarg(fhandle_t *) fhp; 1906 syscallarg(size_t *) fh_size; 1907 } */ 1908 struct vnode *vp; 1909 fhandle_t *fh; 1910 int error; 1911 struct pathbuf *pb; 1912 struct nameidata nd; 1913 size_t sz; 1914 size_t usz; 1915 1916 /* 1917 * Must be super user 1918 */ 1919 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1920 0, NULL, NULL, NULL); 1921 if (error) 1922 return (error); 1923 1924 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1925 if (error) { 1926 return error; 1927 } 1928 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1929 error = namei(&nd); 1930 if (error) { 1931 pathbuf_destroy(pb); 1932 return error; 1933 } 1934 vp = nd.ni_vp; 1935 pathbuf_destroy(pb); 1936 1937 error = vfs_composefh_alloc(vp, &fh); 1938 vput(vp); 1939 if (error != 0) { 1940 goto out; 1941 } 1942 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1943 if (error != 0) { 1944 goto out; 1945 } 1946 sz = FHANDLE_SIZE(fh); 1947 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1948 if (error != 0) { 1949 goto out; 1950 } 1951 if (usz >= sz) { 1952 error = copyout(fh, SCARG(uap, fhp), sz); 1953 } else { 1954 error = E2BIG; 1955 } 1956 out: 1957 vfs_composefh_free(fh); 1958 return (error); 1959 } 1960 1961 /* 1962 * Open a file given a file handle. 1963 * 1964 * Check permissions, allocate an open file structure, 1965 * and call the device open routine if any. 1966 */ 1967 1968 int 1969 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1970 register_t *retval) 1971 { 1972 file_t *fp; 1973 struct vnode *vp = NULL; 1974 kauth_cred_t cred = l->l_cred; 1975 file_t *nfp; 1976 int indx, error = 0; 1977 struct vattr va; 1978 fhandle_t *fh; 1979 int flags; 1980 proc_t *p; 1981 1982 p = curproc; 1983 1984 /* 1985 * Must be super user 1986 */ 1987 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1988 0, NULL, NULL, NULL))) 1989 return (error); 1990 1991 flags = FFLAGS(oflags); 1992 if ((flags & (FREAD | FWRITE)) == 0) 1993 return (EINVAL); 1994 if ((flags & O_CREAT)) 1995 return (EINVAL); 1996 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1997 return (error); 1998 fp = nfp; 1999 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2000 if (error != 0) { 2001 goto bad; 2002 } 2003 error = vfs_fhtovp(fh, &vp); 2004 if (error != 0) { 2005 goto bad; 2006 } 2007 2008 /* Now do an effective vn_open */ 2009 2010 if (vp->v_type == VSOCK) { 2011 error = EOPNOTSUPP; 2012 goto bad; 2013 } 2014 error = vn_openchk(vp, cred, flags); 2015 if (error != 0) 2016 goto bad; 2017 if (flags & O_TRUNC) { 2018 VOP_UNLOCK(vp); /* XXX */ 2019 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2020 vattr_null(&va); 2021 va.va_size = 0; 2022 error = VOP_SETATTR(vp, &va, cred); 2023 if (error) 2024 goto bad; 2025 } 2026 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2027 goto bad; 2028 if (flags & FWRITE) { 2029 mutex_enter(vp->v_interlock); 2030 vp->v_writecount++; 2031 mutex_exit(vp->v_interlock); 2032 } 2033 2034 /* done with modified vn_open, now finish what sys_open does. */ 2035 if ((error = open_setfp(l, fp, vp, indx, flags))) 2036 return error; 2037 2038 VOP_UNLOCK(vp); 2039 *retval = indx; 2040 fd_affix(p, fp, indx); 2041 vfs_copyinfh_free(fh); 2042 return (0); 2043 2044 bad: 2045 fd_abort(p, fp, indx); 2046 if (vp != NULL) 2047 vput(vp); 2048 vfs_copyinfh_free(fh); 2049 return (error); 2050 } 2051 2052 int 2053 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2054 { 2055 /* { 2056 syscallarg(const void *) fhp; 2057 syscallarg(size_t) fh_size; 2058 syscallarg(int) flags; 2059 } */ 2060 2061 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2062 SCARG(uap, flags), retval); 2063 } 2064 2065 int 2066 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2067 { 2068 int error; 2069 fhandle_t *fh; 2070 struct vnode *vp; 2071 2072 /* 2073 * Must be super user 2074 */ 2075 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2076 0, NULL, NULL, NULL))) 2077 return (error); 2078 2079 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2080 if (error != 0) 2081 return error; 2082 2083 error = vfs_fhtovp(fh, &vp); 2084 vfs_copyinfh_free(fh); 2085 if (error != 0) 2086 return error; 2087 2088 error = vn_stat(vp, sb); 2089 vput(vp); 2090 return error; 2091 } 2092 2093 2094 /* ARGSUSED */ 2095 int 2096 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2097 { 2098 /* { 2099 syscallarg(const void *) fhp; 2100 syscallarg(size_t) fh_size; 2101 syscallarg(struct stat *) sb; 2102 } */ 2103 struct stat sb; 2104 int error; 2105 2106 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2107 if (error) 2108 return error; 2109 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2110 } 2111 2112 int 2113 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2114 int flags) 2115 { 2116 fhandle_t *fh; 2117 struct mount *mp; 2118 struct vnode *vp; 2119 int error; 2120 2121 /* 2122 * Must be super user 2123 */ 2124 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2125 0, NULL, NULL, NULL))) 2126 return error; 2127 2128 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2129 if (error != 0) 2130 return error; 2131 2132 error = vfs_fhtovp(fh, &vp); 2133 vfs_copyinfh_free(fh); 2134 if (error != 0) 2135 return error; 2136 2137 mp = vp->v_mount; 2138 error = dostatvfs(mp, sb, l, flags, 1); 2139 vput(vp); 2140 return error; 2141 } 2142 2143 /* ARGSUSED */ 2144 int 2145 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 2146 { 2147 /* { 2148 syscallarg(const void *) fhp; 2149 syscallarg(size_t) fh_size; 2150 syscallarg(struct statvfs *) buf; 2151 syscallarg(int) flags; 2152 } */ 2153 struct statvfs *sb = STATVFSBUF_GET(); 2154 int error; 2155 2156 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2157 SCARG(uap, flags)); 2158 if (error == 0) 2159 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2160 STATVFSBUF_PUT(sb); 2161 return error; 2162 } 2163 2164 /* 2165 * Create a special file. 2166 */ 2167 /* ARGSUSED */ 2168 int 2169 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2170 register_t *retval) 2171 { 2172 /* { 2173 syscallarg(const char *) path; 2174 syscallarg(mode_t) mode; 2175 syscallarg(dev_t) dev; 2176 } */ 2177 return do_sys_mknodat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode), 2178 SCARG(uap, dev), retval, UIO_USERSPACE); 2179 } 2180 2181 int 2182 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2183 register_t *retval) 2184 { 2185 /* { 2186 syscallarg(int) fd; 2187 syscallarg(const char *) path; 2188 syscallarg(mode_t) mode; 2189 syscallarg(uint32_t) dev; 2190 } */ 2191 2192 return do_sys_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2193 SCARG(uap, mode), SCARG(uap, dev), retval, UIO_USERSPACE); 2194 } 2195 2196 int 2197 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2198 register_t *retval, enum uio_seg seg) 2199 { 2200 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, retval, seg); 2201 } 2202 2203 int 2204 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2205 dev_t dev, register_t *retval, enum uio_seg seg) 2206 { 2207 struct proc *p = l->l_proc; 2208 struct vnode *vp; 2209 struct vattr vattr; 2210 int error, optype; 2211 struct pathbuf *pb; 2212 struct nameidata nd; 2213 const char *pathstring; 2214 2215 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2216 0, NULL, NULL, NULL)) != 0) 2217 return (error); 2218 2219 optype = VOP_MKNOD_DESCOFFSET; 2220 2221 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2222 if (error) { 2223 return error; 2224 } 2225 pathstring = pathbuf_stringcopy_get(pb); 2226 if (pathstring == NULL) { 2227 pathbuf_destroy(pb); 2228 return ENOMEM; 2229 } 2230 2231 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2232 2233 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2234 goto out; 2235 vp = nd.ni_vp; 2236 2237 if (vp != NULL) 2238 error = EEXIST; 2239 else { 2240 vattr_null(&vattr); 2241 /* We will read cwdi->cwdi_cmask unlocked. */ 2242 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2243 vattr.va_rdev = dev; 2244 2245 switch (mode & S_IFMT) { 2246 case S_IFMT: /* used by badsect to flag bad sectors */ 2247 vattr.va_type = VBAD; 2248 break; 2249 case S_IFCHR: 2250 vattr.va_type = VCHR; 2251 break; 2252 case S_IFBLK: 2253 vattr.va_type = VBLK; 2254 break; 2255 case S_IFWHT: 2256 optype = VOP_WHITEOUT_DESCOFFSET; 2257 break; 2258 case S_IFREG: 2259 #if NVERIEXEC > 0 2260 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2261 O_CREAT); 2262 #endif /* NVERIEXEC > 0 */ 2263 vattr.va_type = VREG; 2264 vattr.va_rdev = VNOVAL; 2265 optype = VOP_CREATE_DESCOFFSET; 2266 break; 2267 default: 2268 error = EINVAL; 2269 break; 2270 } 2271 } 2272 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET 2273 && vattr.va_rdev == VNOVAL) 2274 error = EINVAL; 2275 if (!error) { 2276 switch (optype) { 2277 case VOP_WHITEOUT_DESCOFFSET: 2278 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2279 if (error) 2280 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2281 vput(nd.ni_dvp); 2282 break; 2283 2284 case VOP_MKNOD_DESCOFFSET: 2285 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2286 &nd.ni_cnd, &vattr); 2287 if (error == 0) 2288 vput(nd.ni_vp); 2289 break; 2290 2291 case VOP_CREATE_DESCOFFSET: 2292 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2293 &nd.ni_cnd, &vattr); 2294 if (error == 0) 2295 vput(nd.ni_vp); 2296 break; 2297 } 2298 } else { 2299 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2300 if (nd.ni_dvp == vp) 2301 vrele(nd.ni_dvp); 2302 else 2303 vput(nd.ni_dvp); 2304 if (vp) 2305 vrele(vp); 2306 } 2307 out: 2308 pathbuf_stringcopy_put(pb, pathstring); 2309 pathbuf_destroy(pb); 2310 return (error); 2311 } 2312 2313 /* 2314 * Create a named pipe. 2315 */ 2316 /* ARGSUSED */ 2317 int 2318 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2319 { 2320 /* { 2321 syscallarg(const char *) path; 2322 syscallarg(int) mode; 2323 } */ 2324 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2325 } 2326 2327 int 2328 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2329 register_t *retval) 2330 { 2331 /* { 2332 syscallarg(int) fd; 2333 syscallarg(const char *) path; 2334 syscallarg(int) mode; 2335 } */ 2336 2337 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2338 SCARG(uap, mode)); 2339 } 2340 2341 static int 2342 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2343 { 2344 struct proc *p = l->l_proc; 2345 struct vattr vattr; 2346 int error; 2347 struct pathbuf *pb; 2348 struct nameidata nd; 2349 2350 error = pathbuf_copyin(path, &pb); 2351 if (error) { 2352 return error; 2353 } 2354 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2355 2356 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2357 pathbuf_destroy(pb); 2358 return error; 2359 } 2360 if (nd.ni_vp != NULL) { 2361 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2362 if (nd.ni_dvp == nd.ni_vp) 2363 vrele(nd.ni_dvp); 2364 else 2365 vput(nd.ni_dvp); 2366 vrele(nd.ni_vp); 2367 pathbuf_destroy(pb); 2368 return (EEXIST); 2369 } 2370 vattr_null(&vattr); 2371 vattr.va_type = VFIFO; 2372 /* We will read cwdi->cwdi_cmask unlocked. */ 2373 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2374 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2375 if (error == 0) 2376 vput(nd.ni_vp); 2377 pathbuf_destroy(pb); 2378 return (error); 2379 } 2380 2381 /* 2382 * Make a hard file link. 2383 */ 2384 /* ARGSUSED */ 2385 static int 2386 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2387 const char *link, int follow, register_t *retval) 2388 { 2389 struct vnode *vp; 2390 struct pathbuf *linkpb; 2391 struct nameidata nd; 2392 namei_simple_flags_t ns_flags; 2393 int error; 2394 2395 if (follow & AT_SYMLINK_FOLLOW) 2396 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2397 else 2398 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2399 2400 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2401 if (error != 0) 2402 return (error); 2403 error = pathbuf_copyin(link, &linkpb); 2404 if (error) { 2405 goto out1; 2406 } 2407 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2408 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2409 goto out2; 2410 if (nd.ni_vp) { 2411 error = EEXIST; 2412 goto abortop; 2413 } 2414 /* Prevent hard links on directories. */ 2415 if (vp->v_type == VDIR) { 2416 error = EPERM; 2417 goto abortop; 2418 } 2419 /* Prevent cross-mount operation. */ 2420 if (nd.ni_dvp->v_mount != vp->v_mount) { 2421 error = EXDEV; 2422 goto abortop; 2423 } 2424 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2425 out2: 2426 pathbuf_destroy(linkpb); 2427 out1: 2428 vrele(vp); 2429 return (error); 2430 abortop: 2431 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2432 if (nd.ni_dvp == nd.ni_vp) 2433 vrele(nd.ni_dvp); 2434 else 2435 vput(nd.ni_dvp); 2436 if (nd.ni_vp != NULL) 2437 vrele(nd.ni_vp); 2438 goto out2; 2439 } 2440 2441 int 2442 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2443 { 2444 /* { 2445 syscallarg(const char *) path; 2446 syscallarg(const char *) link; 2447 } */ 2448 const char *path = SCARG(uap, path); 2449 const char *link = SCARG(uap, link); 2450 2451 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2452 AT_SYMLINK_FOLLOW, retval); 2453 } 2454 2455 int 2456 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2457 register_t *retval) 2458 { 2459 /* { 2460 syscallarg(int) fd1; 2461 syscallarg(const char *) name1; 2462 syscallarg(int) fd2; 2463 syscallarg(const char *) name2; 2464 syscallarg(int) flags; 2465 } */ 2466 int fd1 = SCARG(uap, fd1); 2467 const char *name1 = SCARG(uap, name1); 2468 int fd2 = SCARG(uap, fd2); 2469 const char *name2 = SCARG(uap, name2); 2470 int follow; 2471 2472 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2473 2474 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2475 } 2476 2477 2478 int 2479 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2480 { 2481 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2482 } 2483 2484 static int 2485 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2486 const char *link, enum uio_seg seg) 2487 { 2488 struct proc *p = curproc; 2489 struct vattr vattr; 2490 char *path; 2491 int error; 2492 struct pathbuf *linkpb; 2493 struct nameidata nd; 2494 2495 KASSERT(l != NULL || fdat == AT_FDCWD); 2496 2497 path = PNBUF_GET(); 2498 if (seg == UIO_USERSPACE) { 2499 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2500 goto out1; 2501 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2502 goto out1; 2503 } else { 2504 KASSERT(strlen(patharg) < MAXPATHLEN); 2505 strcpy(path, patharg); 2506 linkpb = pathbuf_create(link); 2507 if (linkpb == NULL) { 2508 error = ENOMEM; 2509 goto out1; 2510 } 2511 } 2512 ktrkuser("symlink-target", path, strlen(path)); 2513 2514 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2515 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2516 goto out2; 2517 if (nd.ni_vp) { 2518 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2519 if (nd.ni_dvp == nd.ni_vp) 2520 vrele(nd.ni_dvp); 2521 else 2522 vput(nd.ni_dvp); 2523 vrele(nd.ni_vp); 2524 error = EEXIST; 2525 goto out2; 2526 } 2527 vattr_null(&vattr); 2528 vattr.va_type = VLNK; 2529 /* We will read cwdi->cwdi_cmask unlocked. */ 2530 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2531 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2532 if (error == 0) 2533 vput(nd.ni_vp); 2534 out2: 2535 pathbuf_destroy(linkpb); 2536 out1: 2537 PNBUF_PUT(path); 2538 return (error); 2539 } 2540 2541 /* 2542 * Make a symbolic link. 2543 */ 2544 /* ARGSUSED */ 2545 int 2546 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2547 { 2548 /* { 2549 syscallarg(const char *) path; 2550 syscallarg(const char *) link; 2551 } */ 2552 2553 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2554 UIO_USERSPACE); 2555 } 2556 2557 int 2558 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2559 register_t *retval) 2560 { 2561 /* { 2562 syscallarg(const char *) path1; 2563 syscallarg(int) fd; 2564 syscallarg(const char *) path2; 2565 } */ 2566 2567 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2568 SCARG(uap, path2), UIO_USERSPACE); 2569 } 2570 2571 /* 2572 * Delete a whiteout from the filesystem. 2573 */ 2574 /* ARGSUSED */ 2575 int 2576 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2577 { 2578 /* { 2579 syscallarg(const char *) path; 2580 } */ 2581 int error; 2582 struct pathbuf *pb; 2583 struct nameidata nd; 2584 2585 error = pathbuf_copyin(SCARG(uap, path), &pb); 2586 if (error) { 2587 return error; 2588 } 2589 2590 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2591 error = namei(&nd); 2592 if (error) { 2593 pathbuf_destroy(pb); 2594 return (error); 2595 } 2596 2597 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2598 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2599 if (nd.ni_dvp == nd.ni_vp) 2600 vrele(nd.ni_dvp); 2601 else 2602 vput(nd.ni_dvp); 2603 if (nd.ni_vp) 2604 vrele(nd.ni_vp); 2605 pathbuf_destroy(pb); 2606 return (EEXIST); 2607 } 2608 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2609 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2610 vput(nd.ni_dvp); 2611 pathbuf_destroy(pb); 2612 return (error); 2613 } 2614 2615 /* 2616 * Delete a name from the filesystem. 2617 */ 2618 /* ARGSUSED */ 2619 int 2620 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2621 { 2622 /* { 2623 syscallarg(const char *) path; 2624 } */ 2625 2626 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2627 } 2628 2629 int 2630 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2631 register_t *retval) 2632 { 2633 /* { 2634 syscallarg(int) fd; 2635 syscallarg(const char *) path; 2636 syscallarg(int) flag; 2637 } */ 2638 2639 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2640 SCARG(uap, flag), UIO_USERSPACE); 2641 } 2642 2643 int 2644 do_sys_unlink(const char *arg, enum uio_seg seg) 2645 { 2646 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2647 } 2648 2649 static int 2650 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2651 enum uio_seg seg) 2652 { 2653 struct vnode *vp; 2654 int error; 2655 struct pathbuf *pb; 2656 struct nameidata nd; 2657 const char *pathstring; 2658 2659 KASSERT(l != NULL || fdat == AT_FDCWD); 2660 2661 error = pathbuf_maybe_copyin(arg, seg, &pb); 2662 if (error) { 2663 return error; 2664 } 2665 pathstring = pathbuf_stringcopy_get(pb); 2666 if (pathstring == NULL) { 2667 pathbuf_destroy(pb); 2668 return ENOMEM; 2669 } 2670 2671 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2672 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2673 goto out; 2674 vp = nd.ni_vp; 2675 2676 /* 2677 * The root of a mounted filesystem cannot be deleted. 2678 */ 2679 if ((vp->v_vflag & VV_ROOT) != 0) { 2680 error = EBUSY; 2681 goto abort; 2682 } 2683 2684 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2685 error = EBUSY; 2686 goto abort; 2687 } 2688 2689 /* 2690 * No rmdir "." please. 2691 */ 2692 if (nd.ni_dvp == vp) { 2693 error = EINVAL; 2694 goto abort; 2695 } 2696 2697 /* 2698 * AT_REMOVEDIR is required to remove a directory 2699 */ 2700 if (vp->v_type == VDIR) { 2701 if (!(flags & AT_REMOVEDIR)) { 2702 error = EPERM; 2703 goto abort; 2704 } else { 2705 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2706 goto out; 2707 } 2708 } 2709 2710 /* 2711 * Starting here we only deal with non directories. 2712 */ 2713 if (flags & AT_REMOVEDIR) { 2714 error = ENOTDIR; 2715 goto abort; 2716 } 2717 2718 2719 #if NVERIEXEC > 0 2720 /* Handle remove requests for veriexec entries. */ 2721 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2722 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2723 if (nd.ni_dvp == vp) 2724 vrele(nd.ni_dvp); 2725 else 2726 vput(nd.ni_dvp); 2727 vput(vp); 2728 goto out; 2729 } 2730 #endif /* NVERIEXEC > 0 */ 2731 2732 #ifdef FILEASSOC 2733 (void)fileassoc_file_delete(vp); 2734 #endif /* FILEASSOC */ 2735 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2736 goto out; 2737 2738 abort: 2739 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2740 if (nd.ni_dvp == vp) 2741 vrele(nd.ni_dvp); 2742 else 2743 vput(nd.ni_dvp); 2744 vput(vp); 2745 2746 out: 2747 pathbuf_stringcopy_put(pb, pathstring); 2748 pathbuf_destroy(pb); 2749 return (error); 2750 } 2751 2752 /* 2753 * Reposition read/write file offset. 2754 */ 2755 int 2756 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2757 { 2758 /* { 2759 syscallarg(int) fd; 2760 syscallarg(int) pad; 2761 syscallarg(off_t) offset; 2762 syscallarg(int) whence; 2763 } */ 2764 kauth_cred_t cred = l->l_cred; 2765 file_t *fp; 2766 struct vnode *vp; 2767 struct vattr vattr; 2768 off_t newoff; 2769 int error, fd; 2770 2771 fd = SCARG(uap, fd); 2772 2773 if ((fp = fd_getfile(fd)) == NULL) 2774 return (EBADF); 2775 2776 vp = fp->f_data; 2777 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2778 error = ESPIPE; 2779 goto out; 2780 } 2781 2782 switch (SCARG(uap, whence)) { 2783 case SEEK_CUR: 2784 newoff = fp->f_offset + SCARG(uap, offset); 2785 break; 2786 case SEEK_END: 2787 vn_lock(vp, LK_SHARED | LK_RETRY); 2788 error = VOP_GETATTR(vp, &vattr, cred); 2789 VOP_UNLOCK(vp); 2790 if (error) { 2791 goto out; 2792 } 2793 newoff = SCARG(uap, offset) + vattr.va_size; 2794 break; 2795 case SEEK_SET: 2796 newoff = SCARG(uap, offset); 2797 break; 2798 default: 2799 error = EINVAL; 2800 goto out; 2801 } 2802 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2803 *(off_t *)retval = fp->f_offset = newoff; 2804 } 2805 out: 2806 fd_putfile(fd); 2807 return (error); 2808 } 2809 2810 /* 2811 * Positional read system call. 2812 */ 2813 int 2814 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2815 { 2816 /* { 2817 syscallarg(int) fd; 2818 syscallarg(void *) buf; 2819 syscallarg(size_t) nbyte; 2820 syscallarg(off_t) offset; 2821 } */ 2822 file_t *fp; 2823 struct vnode *vp; 2824 off_t offset; 2825 int error, fd = SCARG(uap, fd); 2826 2827 if ((fp = fd_getfile(fd)) == NULL) 2828 return (EBADF); 2829 2830 if ((fp->f_flag & FREAD) == 0) { 2831 fd_putfile(fd); 2832 return (EBADF); 2833 } 2834 2835 vp = fp->f_data; 2836 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2837 error = ESPIPE; 2838 goto out; 2839 } 2840 2841 offset = SCARG(uap, offset); 2842 2843 /* 2844 * XXX This works because no file systems actually 2845 * XXX take any action on the seek operation. 2846 */ 2847 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2848 goto out; 2849 2850 /* dofileread() will unuse the descriptor for us */ 2851 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2852 &offset, 0, retval)); 2853 2854 out: 2855 fd_putfile(fd); 2856 return (error); 2857 } 2858 2859 /* 2860 * Positional scatter read system call. 2861 */ 2862 int 2863 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2864 { 2865 /* { 2866 syscallarg(int) fd; 2867 syscallarg(const struct iovec *) iovp; 2868 syscallarg(int) iovcnt; 2869 syscallarg(off_t) offset; 2870 } */ 2871 off_t offset = SCARG(uap, offset); 2872 2873 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2874 SCARG(uap, iovcnt), &offset, 0, retval); 2875 } 2876 2877 /* 2878 * Positional write system call. 2879 */ 2880 int 2881 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2882 { 2883 /* { 2884 syscallarg(int) fd; 2885 syscallarg(const void *) buf; 2886 syscallarg(size_t) nbyte; 2887 syscallarg(off_t) offset; 2888 } */ 2889 file_t *fp; 2890 struct vnode *vp; 2891 off_t offset; 2892 int error, fd = SCARG(uap, fd); 2893 2894 if ((fp = fd_getfile(fd)) == NULL) 2895 return (EBADF); 2896 2897 if ((fp->f_flag & FWRITE) == 0) { 2898 fd_putfile(fd); 2899 return (EBADF); 2900 } 2901 2902 vp = fp->f_data; 2903 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2904 error = ESPIPE; 2905 goto out; 2906 } 2907 2908 offset = SCARG(uap, offset); 2909 2910 /* 2911 * XXX This works because no file systems actually 2912 * XXX take any action on the seek operation. 2913 */ 2914 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2915 goto out; 2916 2917 /* dofilewrite() will unuse the descriptor for us */ 2918 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2919 &offset, 0, retval)); 2920 2921 out: 2922 fd_putfile(fd); 2923 return (error); 2924 } 2925 2926 /* 2927 * Positional gather write system call. 2928 */ 2929 int 2930 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2931 { 2932 /* { 2933 syscallarg(int) fd; 2934 syscallarg(const struct iovec *) iovp; 2935 syscallarg(int) iovcnt; 2936 syscallarg(off_t) offset; 2937 } */ 2938 off_t offset = SCARG(uap, offset); 2939 2940 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2941 SCARG(uap, iovcnt), &offset, 0, retval); 2942 } 2943 2944 /* 2945 * Check access permissions. 2946 */ 2947 int 2948 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2949 { 2950 /* { 2951 syscallarg(const char *) path; 2952 syscallarg(int) flags; 2953 } */ 2954 2955 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 2956 SCARG(uap, flags), 0); 2957 } 2958 2959 static int 2960 do_sys_accessat(struct lwp *l, int fdat, const char *path, 2961 int mode, int flags) 2962 { 2963 kauth_cred_t cred; 2964 struct vnode *vp; 2965 int error, nd_flag, vmode; 2966 struct pathbuf *pb; 2967 struct nameidata nd; 2968 2969 CTASSERT(F_OK == 0); 2970 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 2971 /* nonsense mode */ 2972 return EINVAL; 2973 } 2974 2975 nd_flag = FOLLOW | LOCKLEAF | TRYEMULROOT; 2976 if (flags & AT_SYMLINK_NOFOLLOW) 2977 nd_flag &= ~FOLLOW; 2978 2979 error = pathbuf_copyin(path, &pb); 2980 if (error) 2981 return error; 2982 2983 NDINIT(&nd, LOOKUP, nd_flag, pb); 2984 2985 /* Override default credentials */ 2986 cred = kauth_cred_dup(l->l_cred); 2987 if (!(flags & AT_EACCESS)) { 2988 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2989 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2990 } 2991 nd.ni_cnd.cn_cred = cred; 2992 2993 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2994 pathbuf_destroy(pb); 2995 goto out; 2996 } 2997 vp = nd.ni_vp; 2998 pathbuf_destroy(pb); 2999 3000 /* Flags == 0 means only check for existence. */ 3001 if (mode) { 3002 vmode = 0; 3003 if (mode & R_OK) 3004 vmode |= VREAD; 3005 if (mode & W_OK) 3006 vmode |= VWRITE; 3007 if (mode & X_OK) 3008 vmode |= VEXEC; 3009 3010 error = VOP_ACCESS(vp, vmode, cred); 3011 if (!error && (vmode & VWRITE)) 3012 error = vn_writechk(vp); 3013 } 3014 vput(vp); 3015 out: 3016 kauth_cred_free(cred); 3017 return (error); 3018 } 3019 3020 int 3021 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3022 register_t *retval) 3023 { 3024 /* { 3025 syscallarg(int) fd; 3026 syscallarg(const char *) path; 3027 syscallarg(int) amode; 3028 syscallarg(int) flag; 3029 } */ 3030 3031 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3032 SCARG(uap, amode), SCARG(uap, flag)); 3033 } 3034 3035 /* 3036 * Common code for all sys_stat functions, including compat versions. 3037 */ 3038 int 3039 do_sys_stat(const char *userpath, unsigned int nd_flag, 3040 struct stat *sb) 3041 { 3042 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3043 } 3044 3045 static int 3046 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3047 unsigned int nd_flag, struct stat *sb) 3048 { 3049 int error; 3050 struct pathbuf *pb; 3051 struct nameidata nd; 3052 3053 KASSERT(l != NULL || fdat == AT_FDCWD); 3054 3055 error = pathbuf_copyin(userpath, &pb); 3056 if (error) { 3057 return error; 3058 } 3059 3060 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3061 3062 error = fd_nameiat(l, fdat, &nd); 3063 if (error != 0) { 3064 pathbuf_destroy(pb); 3065 return error; 3066 } 3067 error = vn_stat(nd.ni_vp, sb); 3068 vput(nd.ni_vp); 3069 pathbuf_destroy(pb); 3070 return error; 3071 } 3072 3073 /* 3074 * Get file status; this version follows links. 3075 */ 3076 /* ARGSUSED */ 3077 int 3078 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3079 { 3080 /* { 3081 syscallarg(const char *) path; 3082 syscallarg(struct stat *) ub; 3083 } */ 3084 struct stat sb; 3085 int error; 3086 3087 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3088 if (error) 3089 return error; 3090 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3091 } 3092 3093 /* 3094 * Get file status; this version does not follow links. 3095 */ 3096 /* ARGSUSED */ 3097 int 3098 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3099 { 3100 /* { 3101 syscallarg(const char *) path; 3102 syscallarg(struct stat *) ub; 3103 } */ 3104 struct stat sb; 3105 int error; 3106 3107 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3108 if (error) 3109 return error; 3110 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3111 } 3112 3113 int 3114 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3115 register_t *retval) 3116 { 3117 /* { 3118 syscallarg(int) fd; 3119 syscallarg(const char *) path; 3120 syscallarg(struct stat *) buf; 3121 syscallarg(int) flag; 3122 } */ 3123 unsigned int nd_flag; 3124 struct stat sb; 3125 int error; 3126 3127 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3128 nd_flag = NOFOLLOW; 3129 else 3130 nd_flag = FOLLOW; 3131 3132 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3133 &sb); 3134 if (error) 3135 return error; 3136 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3137 } 3138 3139 /* 3140 * Get configurable pathname variables. 3141 */ 3142 /* ARGSUSED */ 3143 int 3144 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 3145 { 3146 /* { 3147 syscallarg(const char *) path; 3148 syscallarg(int) name; 3149 } */ 3150 int error; 3151 struct pathbuf *pb; 3152 struct nameidata nd; 3153 3154 error = pathbuf_copyin(SCARG(uap, path), &pb); 3155 if (error) { 3156 return error; 3157 } 3158 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3159 if ((error = namei(&nd)) != 0) { 3160 pathbuf_destroy(pb); 3161 return (error); 3162 } 3163 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 3164 vput(nd.ni_vp); 3165 pathbuf_destroy(pb); 3166 return (error); 3167 } 3168 3169 /* 3170 * Return target name of a symbolic link. 3171 */ 3172 /* ARGSUSED */ 3173 int 3174 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3175 register_t *retval) 3176 { 3177 /* { 3178 syscallarg(const char *) path; 3179 syscallarg(char *) buf; 3180 syscallarg(size_t) count; 3181 } */ 3182 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3183 SCARG(uap, buf), SCARG(uap, count), retval); 3184 } 3185 3186 static int 3187 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3188 size_t count, register_t *retval) 3189 { 3190 struct vnode *vp; 3191 struct iovec aiov; 3192 struct uio auio; 3193 int error; 3194 struct pathbuf *pb; 3195 struct nameidata nd; 3196 3197 error = pathbuf_copyin(path, &pb); 3198 if (error) { 3199 return error; 3200 } 3201 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 3202 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3203 pathbuf_destroy(pb); 3204 return error; 3205 } 3206 vp = nd.ni_vp; 3207 pathbuf_destroy(pb); 3208 if (vp->v_type != VLNK) 3209 error = EINVAL; 3210 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3211 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3212 aiov.iov_base = buf; 3213 aiov.iov_len = count; 3214 auio.uio_iov = &aiov; 3215 auio.uio_iovcnt = 1; 3216 auio.uio_offset = 0; 3217 auio.uio_rw = UIO_READ; 3218 KASSERT(l == curlwp); 3219 auio.uio_vmspace = l->l_proc->p_vmspace; 3220 auio.uio_resid = count; 3221 error = VOP_READLINK(vp, &auio, l->l_cred); 3222 } 3223 vput(vp); 3224 *retval = count - auio.uio_resid; 3225 return (error); 3226 } 3227 3228 int 3229 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3230 register_t *retval) 3231 { 3232 /* { 3233 syscallarg(int) fd; 3234 syscallarg(const char *) path; 3235 syscallarg(char *) buf; 3236 syscallarg(size_t) bufsize; 3237 } */ 3238 3239 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3240 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3241 } 3242 3243 /* 3244 * Change flags of a file given a path name. 3245 */ 3246 /* ARGSUSED */ 3247 int 3248 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3249 { 3250 /* { 3251 syscallarg(const char *) path; 3252 syscallarg(u_long) flags; 3253 } */ 3254 struct vnode *vp; 3255 int error; 3256 3257 error = namei_simple_user(SCARG(uap, path), 3258 NSM_FOLLOW_TRYEMULROOT, &vp); 3259 if (error != 0) 3260 return (error); 3261 error = change_flags(vp, SCARG(uap, flags), l); 3262 vput(vp); 3263 return (error); 3264 } 3265 3266 /* 3267 * Change flags of a file given a file descriptor. 3268 */ 3269 /* ARGSUSED */ 3270 int 3271 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3272 { 3273 /* { 3274 syscallarg(int) fd; 3275 syscallarg(u_long) flags; 3276 } */ 3277 struct vnode *vp; 3278 file_t *fp; 3279 int error; 3280 3281 /* fd_getvnode() will use the descriptor for us */ 3282 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3283 return (error); 3284 vp = fp->f_data; 3285 error = change_flags(vp, SCARG(uap, flags), l); 3286 VOP_UNLOCK(vp); 3287 fd_putfile(SCARG(uap, fd)); 3288 return (error); 3289 } 3290 3291 /* 3292 * Change flags of a file given a path name; this version does 3293 * not follow links. 3294 */ 3295 int 3296 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3297 { 3298 /* { 3299 syscallarg(const char *) path; 3300 syscallarg(u_long) flags; 3301 } */ 3302 struct vnode *vp; 3303 int error; 3304 3305 error = namei_simple_user(SCARG(uap, path), 3306 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3307 if (error != 0) 3308 return (error); 3309 error = change_flags(vp, SCARG(uap, flags), l); 3310 vput(vp); 3311 return (error); 3312 } 3313 3314 /* 3315 * Common routine to change flags of a file. 3316 */ 3317 int 3318 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3319 { 3320 struct vattr vattr; 3321 int error; 3322 3323 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3324 3325 vattr_null(&vattr); 3326 vattr.va_flags = flags; 3327 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3328 3329 return (error); 3330 } 3331 3332 /* 3333 * Change mode of a file given path name; this version follows links. 3334 */ 3335 /* ARGSUSED */ 3336 int 3337 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3338 { 3339 /* { 3340 syscallarg(const char *) path; 3341 syscallarg(int) mode; 3342 } */ 3343 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3344 SCARG(uap, mode), 0); 3345 } 3346 3347 static int 3348 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3349 { 3350 int error; 3351 struct vnode *vp; 3352 namei_simple_flags_t ns_flag; 3353 3354 if (flags & AT_SYMLINK_NOFOLLOW) 3355 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3356 else 3357 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3358 3359 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3360 if (error != 0) 3361 return error; 3362 3363 error = change_mode(vp, mode, l); 3364 3365 vrele(vp); 3366 3367 return (error); 3368 } 3369 3370 /* 3371 * Change mode of a file given a file descriptor. 3372 */ 3373 /* ARGSUSED */ 3374 int 3375 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3376 { 3377 /* { 3378 syscallarg(int) fd; 3379 syscallarg(int) mode; 3380 } */ 3381 file_t *fp; 3382 int error; 3383 3384 /* fd_getvnode() will use the descriptor for us */ 3385 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3386 return (error); 3387 error = change_mode(fp->f_data, SCARG(uap, mode), l); 3388 fd_putfile(SCARG(uap, fd)); 3389 return (error); 3390 } 3391 3392 int 3393 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3394 register_t *retval) 3395 { 3396 /* { 3397 syscallarg(int) fd; 3398 syscallarg(const char *) path; 3399 syscallarg(int) mode; 3400 syscallarg(int) flag; 3401 } */ 3402 3403 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3404 SCARG(uap, mode), SCARG(uap, flag)); 3405 } 3406 3407 /* 3408 * Change mode of a file given path name; this version does not follow links. 3409 */ 3410 /* ARGSUSED */ 3411 int 3412 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3413 { 3414 /* { 3415 syscallarg(const char *) path; 3416 syscallarg(int) mode; 3417 } */ 3418 int error; 3419 struct vnode *vp; 3420 3421 error = namei_simple_user(SCARG(uap, path), 3422 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3423 if (error != 0) 3424 return (error); 3425 3426 error = change_mode(vp, SCARG(uap, mode), l); 3427 3428 vrele(vp); 3429 return (error); 3430 } 3431 3432 /* 3433 * Common routine to set mode given a vnode. 3434 */ 3435 static int 3436 change_mode(struct vnode *vp, int mode, struct lwp *l) 3437 { 3438 struct vattr vattr; 3439 int error; 3440 3441 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3442 vattr_null(&vattr); 3443 vattr.va_mode = mode & ALLPERMS; 3444 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3445 VOP_UNLOCK(vp); 3446 return (error); 3447 } 3448 3449 /* 3450 * Set ownership given a path name; this version follows links. 3451 */ 3452 /* ARGSUSED */ 3453 int 3454 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3455 { 3456 /* { 3457 syscallarg(const char *) path; 3458 syscallarg(uid_t) uid; 3459 syscallarg(gid_t) gid; 3460 } */ 3461 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3462 SCARG(uap, gid), 0); 3463 } 3464 3465 static int 3466 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3467 gid_t gid, int flags) 3468 { 3469 int error; 3470 struct vnode *vp; 3471 namei_simple_flags_t ns_flag; 3472 3473 if (flags & AT_SYMLINK_NOFOLLOW) 3474 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3475 else 3476 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3477 3478 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3479 if (error != 0) 3480 return error; 3481 3482 error = change_owner(vp, uid, gid, l, 0); 3483 3484 vrele(vp); 3485 3486 return (error); 3487 } 3488 3489 /* 3490 * Set ownership given a path name; this version follows links. 3491 * Provides POSIX semantics. 3492 */ 3493 /* ARGSUSED */ 3494 int 3495 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3496 { 3497 /* { 3498 syscallarg(const char *) path; 3499 syscallarg(uid_t) uid; 3500 syscallarg(gid_t) gid; 3501 } */ 3502 int error; 3503 struct vnode *vp; 3504 3505 error = namei_simple_user(SCARG(uap, path), 3506 NSM_FOLLOW_TRYEMULROOT, &vp); 3507 if (error != 0) 3508 return (error); 3509 3510 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3511 3512 vrele(vp); 3513 return (error); 3514 } 3515 3516 /* 3517 * Set ownership given a file descriptor. 3518 */ 3519 /* ARGSUSED */ 3520 int 3521 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3522 { 3523 /* { 3524 syscallarg(int) fd; 3525 syscallarg(uid_t) uid; 3526 syscallarg(gid_t) gid; 3527 } */ 3528 int error; 3529 file_t *fp; 3530 3531 /* fd_getvnode() will use the descriptor for us */ 3532 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3533 return (error); 3534 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3535 l, 0); 3536 fd_putfile(SCARG(uap, fd)); 3537 return (error); 3538 } 3539 3540 int 3541 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3542 register_t *retval) 3543 { 3544 /* { 3545 syscallarg(int) fd; 3546 syscallarg(const char *) path; 3547 syscallarg(uid_t) owner; 3548 syscallarg(gid_t) group; 3549 syscallarg(int) flag; 3550 } */ 3551 3552 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3553 SCARG(uap, owner), SCARG(uap, group), 3554 SCARG(uap, flag)); 3555 } 3556 3557 /* 3558 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3559 */ 3560 /* ARGSUSED */ 3561 int 3562 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3563 { 3564 /* { 3565 syscallarg(int) fd; 3566 syscallarg(uid_t) uid; 3567 syscallarg(gid_t) gid; 3568 } */ 3569 int error; 3570 file_t *fp; 3571 3572 /* fd_getvnode() will use the descriptor for us */ 3573 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3574 return (error); 3575 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3576 l, 1); 3577 fd_putfile(SCARG(uap, fd)); 3578 return (error); 3579 } 3580 3581 /* 3582 * Set ownership given a path name; this version does not follow links. 3583 */ 3584 /* ARGSUSED */ 3585 int 3586 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3587 { 3588 /* { 3589 syscallarg(const char *) path; 3590 syscallarg(uid_t) uid; 3591 syscallarg(gid_t) gid; 3592 } */ 3593 int error; 3594 struct vnode *vp; 3595 3596 error = namei_simple_user(SCARG(uap, path), 3597 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3598 if (error != 0) 3599 return (error); 3600 3601 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3602 3603 vrele(vp); 3604 return (error); 3605 } 3606 3607 /* 3608 * Set ownership given a path name; this version does not follow links. 3609 * Provides POSIX/XPG semantics. 3610 */ 3611 /* ARGSUSED */ 3612 int 3613 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3614 { 3615 /* { 3616 syscallarg(const char *) path; 3617 syscallarg(uid_t) uid; 3618 syscallarg(gid_t) gid; 3619 } */ 3620 int error; 3621 struct vnode *vp; 3622 3623 error = namei_simple_user(SCARG(uap, path), 3624 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3625 if (error != 0) 3626 return (error); 3627 3628 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3629 3630 vrele(vp); 3631 return (error); 3632 } 3633 3634 /* 3635 * Common routine to set ownership given a vnode. 3636 */ 3637 static int 3638 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3639 int posix_semantics) 3640 { 3641 struct vattr vattr; 3642 mode_t newmode; 3643 int error; 3644 3645 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3646 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3647 goto out; 3648 3649 #define CHANGED(x) ((int)(x) != -1) 3650 newmode = vattr.va_mode; 3651 if (posix_semantics) { 3652 /* 3653 * POSIX/XPG semantics: if the caller is not the super-user, 3654 * clear set-user-id and set-group-id bits. Both POSIX and 3655 * the XPG consider the behaviour for calls by the super-user 3656 * implementation-defined; we leave the set-user-id and set- 3657 * group-id settings intact in that case. 3658 */ 3659 if (vattr.va_mode & S_ISUID) { 3660 if (kauth_authorize_vnode(l->l_cred, 3661 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3662 newmode &= ~S_ISUID; 3663 } 3664 if (vattr.va_mode & S_ISGID) { 3665 if (kauth_authorize_vnode(l->l_cred, 3666 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3667 newmode &= ~S_ISGID; 3668 } 3669 } else { 3670 /* 3671 * NetBSD semantics: when changing owner and/or group, 3672 * clear the respective bit(s). 3673 */ 3674 if (CHANGED(uid)) 3675 newmode &= ~S_ISUID; 3676 if (CHANGED(gid)) 3677 newmode &= ~S_ISGID; 3678 } 3679 /* Update va_mode iff altered. */ 3680 if (vattr.va_mode == newmode) 3681 newmode = VNOVAL; 3682 3683 vattr_null(&vattr); 3684 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3685 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3686 vattr.va_mode = newmode; 3687 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3688 #undef CHANGED 3689 3690 out: 3691 VOP_UNLOCK(vp); 3692 return (error); 3693 } 3694 3695 /* 3696 * Set the access and modification times given a path name; this 3697 * version follows links. 3698 */ 3699 /* ARGSUSED */ 3700 int 3701 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3702 register_t *retval) 3703 { 3704 /* { 3705 syscallarg(const char *) path; 3706 syscallarg(const struct timeval *) tptr; 3707 } */ 3708 3709 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3710 SCARG(uap, tptr), UIO_USERSPACE); 3711 } 3712 3713 /* 3714 * Set the access and modification times given a file descriptor. 3715 */ 3716 /* ARGSUSED */ 3717 int 3718 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3719 register_t *retval) 3720 { 3721 /* { 3722 syscallarg(int) fd; 3723 syscallarg(const struct timeval *) tptr; 3724 } */ 3725 int error; 3726 file_t *fp; 3727 3728 /* fd_getvnode() will use the descriptor for us */ 3729 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3730 return (error); 3731 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3732 UIO_USERSPACE); 3733 fd_putfile(SCARG(uap, fd)); 3734 return (error); 3735 } 3736 3737 int 3738 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3739 register_t *retval) 3740 { 3741 /* { 3742 syscallarg(int) fd; 3743 syscallarg(const struct timespec *) tptr; 3744 } */ 3745 int error; 3746 file_t *fp; 3747 3748 /* fd_getvnode() will use the descriptor for us */ 3749 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3750 return (error); 3751 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0, 3752 SCARG(uap, tptr), UIO_USERSPACE); 3753 fd_putfile(SCARG(uap, fd)); 3754 return (error); 3755 } 3756 3757 /* 3758 * Set the access and modification times given a path name; this 3759 * version does not follow links. 3760 */ 3761 int 3762 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3763 register_t *retval) 3764 { 3765 /* { 3766 syscallarg(const char *) path; 3767 syscallarg(const struct timeval *) tptr; 3768 } */ 3769 3770 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3771 SCARG(uap, tptr), UIO_USERSPACE); 3772 } 3773 3774 int 3775 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3776 register_t *retval) 3777 { 3778 /* { 3779 syscallarg(int) fd; 3780 syscallarg(const char *) path; 3781 syscallarg(const struct timespec *) tptr; 3782 syscallarg(int) flag; 3783 } */ 3784 int follow; 3785 const struct timespec *tptr; 3786 int error; 3787 3788 tptr = SCARG(uap, tptr); 3789 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3790 3791 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3792 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3793 3794 return error; 3795 } 3796 3797 /* 3798 * Common routine to set access and modification times given a vnode. 3799 */ 3800 int 3801 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3802 const struct timespec *tptr, enum uio_seg seg) 3803 { 3804 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3805 } 3806 3807 static int 3808 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3809 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3810 { 3811 struct vattr vattr; 3812 int error, dorele = 0; 3813 namei_simple_flags_t sflags; 3814 bool vanull, setbirthtime; 3815 struct timespec ts[2]; 3816 3817 KASSERT(l != NULL || fdat == AT_FDCWD); 3818 3819 /* 3820 * I have checked all callers and they pass either FOLLOW, 3821 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3822 * is 0. More to the point, they don't pass anything else. 3823 * Let's keep it that way at least until the namei interfaces 3824 * are fully sanitized. 3825 */ 3826 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3827 sflags = (flag == FOLLOW) ? 3828 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3829 3830 if (tptr == NULL) { 3831 vanull = true; 3832 nanotime(&ts[0]); 3833 ts[1] = ts[0]; 3834 } else { 3835 vanull = false; 3836 if (seg != UIO_SYSSPACE) { 3837 error = copyin(tptr, ts, sizeof (ts)); 3838 if (error != 0) 3839 return error; 3840 } else { 3841 ts[0] = tptr[0]; 3842 ts[1] = tptr[1]; 3843 } 3844 } 3845 3846 if (ts[0].tv_nsec == UTIME_NOW) { 3847 nanotime(&ts[0]); 3848 if (ts[1].tv_nsec == UTIME_NOW) { 3849 vanull = true; 3850 ts[1] = ts[0]; 3851 } 3852 } else if (ts[1].tv_nsec == UTIME_NOW) 3853 nanotime(&ts[1]); 3854 3855 if (vp == NULL) { 3856 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3857 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3858 if (error != 0) 3859 return error; 3860 dorele = 1; 3861 } 3862 3863 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3864 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3865 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3866 vattr_null(&vattr); 3867 3868 if (ts[0].tv_nsec != UTIME_OMIT) 3869 vattr.va_atime = ts[0]; 3870 3871 if (ts[1].tv_nsec != UTIME_OMIT) { 3872 vattr.va_mtime = ts[1]; 3873 if (setbirthtime) 3874 vattr.va_birthtime = ts[1]; 3875 } 3876 3877 if (vanull) 3878 vattr.va_vaflags |= VA_UTIMES_NULL; 3879 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3880 VOP_UNLOCK(vp); 3881 3882 if (dorele != 0) 3883 vrele(vp); 3884 3885 return error; 3886 } 3887 3888 int 3889 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3890 const struct timeval *tptr, enum uio_seg seg) 3891 { 3892 struct timespec ts[2]; 3893 struct timespec *tsptr = NULL; 3894 int error; 3895 3896 if (tptr != NULL) { 3897 struct timeval tv[2]; 3898 3899 if (seg != UIO_SYSSPACE) { 3900 error = copyin(tptr, tv, sizeof (tv)); 3901 if (error != 0) 3902 return error; 3903 tptr = tv; 3904 } 3905 3906 if ((tv[0].tv_usec == UTIME_NOW) || 3907 (tv[0].tv_usec == UTIME_OMIT)) 3908 ts[0].tv_nsec = tv[0].tv_usec; 3909 else 3910 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3911 3912 if ((tv[1].tv_usec == UTIME_NOW) || 3913 (tv[1].tv_usec == UTIME_OMIT)) 3914 ts[1].tv_nsec = tv[1].tv_usec; 3915 else 3916 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3917 3918 tsptr = &ts[0]; 3919 } 3920 3921 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 3922 } 3923 3924 /* 3925 * Truncate a file given its path name. 3926 */ 3927 /* ARGSUSED */ 3928 int 3929 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3930 { 3931 /* { 3932 syscallarg(const char *) path; 3933 syscallarg(int) pad; 3934 syscallarg(off_t) length; 3935 } */ 3936 struct vnode *vp; 3937 struct vattr vattr; 3938 int error; 3939 3940 error = namei_simple_user(SCARG(uap, path), 3941 NSM_FOLLOW_TRYEMULROOT, &vp); 3942 if (error != 0) 3943 return (error); 3944 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3945 if (vp->v_type == VDIR) 3946 error = EISDIR; 3947 else if ((error = vn_writechk(vp)) == 0 && 3948 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3949 vattr_null(&vattr); 3950 vattr.va_size = SCARG(uap, length); 3951 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3952 } 3953 vput(vp); 3954 return (error); 3955 } 3956 3957 /* 3958 * Truncate a file given a file descriptor. 3959 */ 3960 /* ARGSUSED */ 3961 int 3962 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3963 { 3964 /* { 3965 syscallarg(int) fd; 3966 syscallarg(int) pad; 3967 syscallarg(off_t) length; 3968 } */ 3969 struct vattr vattr; 3970 struct vnode *vp; 3971 file_t *fp; 3972 int error; 3973 3974 /* fd_getvnode() will use the descriptor for us */ 3975 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3976 return (error); 3977 if ((fp->f_flag & FWRITE) == 0) { 3978 error = EINVAL; 3979 goto out; 3980 } 3981 vp = fp->f_data; 3982 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3983 if (vp->v_type == VDIR) 3984 error = EISDIR; 3985 else if ((error = vn_writechk(vp)) == 0) { 3986 vattr_null(&vattr); 3987 vattr.va_size = SCARG(uap, length); 3988 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3989 } 3990 VOP_UNLOCK(vp); 3991 out: 3992 fd_putfile(SCARG(uap, fd)); 3993 return (error); 3994 } 3995 3996 /* 3997 * Sync an open file. 3998 */ 3999 /* ARGSUSED */ 4000 int 4001 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 4002 { 4003 /* { 4004 syscallarg(int) fd; 4005 } */ 4006 struct vnode *vp; 4007 file_t *fp; 4008 int error; 4009 4010 /* fd_getvnode() will use the descriptor for us */ 4011 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4012 return (error); 4013 vp = fp->f_data; 4014 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4015 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4016 VOP_UNLOCK(vp); 4017 fd_putfile(SCARG(uap, fd)); 4018 return (error); 4019 } 4020 4021 /* 4022 * Sync a range of file data. API modeled after that found in AIX. 4023 * 4024 * FDATASYNC indicates that we need only save enough metadata to be able 4025 * to re-read the written data. Note we duplicate AIX's requirement that 4026 * the file be open for writing. 4027 */ 4028 /* ARGSUSED */ 4029 int 4030 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4031 { 4032 /* { 4033 syscallarg(int) fd; 4034 syscallarg(int) flags; 4035 syscallarg(off_t) start; 4036 syscallarg(off_t) length; 4037 } */ 4038 struct vnode *vp; 4039 file_t *fp; 4040 int flags, nflags; 4041 off_t s, e, len; 4042 int error; 4043 4044 /* fd_getvnode() will use the descriptor for us */ 4045 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4046 return (error); 4047 4048 if ((fp->f_flag & FWRITE) == 0) { 4049 error = EBADF; 4050 goto out; 4051 } 4052 4053 flags = SCARG(uap, flags); 4054 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4055 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4056 error = EINVAL; 4057 goto out; 4058 } 4059 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4060 if (flags & FDATASYNC) 4061 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4062 else 4063 nflags = FSYNC_WAIT; 4064 if (flags & FDISKSYNC) 4065 nflags |= FSYNC_CACHE; 4066 4067 len = SCARG(uap, length); 4068 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4069 if (len) { 4070 s = SCARG(uap, start); 4071 e = s + len; 4072 if (e < s) { 4073 error = EINVAL; 4074 goto out; 4075 } 4076 } else { 4077 e = 0; 4078 s = 0; 4079 } 4080 4081 vp = fp->f_data; 4082 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4083 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4084 VOP_UNLOCK(vp); 4085 out: 4086 fd_putfile(SCARG(uap, fd)); 4087 return (error); 4088 } 4089 4090 /* 4091 * Sync the data of an open file. 4092 */ 4093 /* ARGSUSED */ 4094 int 4095 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4096 { 4097 /* { 4098 syscallarg(int) fd; 4099 } */ 4100 struct vnode *vp; 4101 file_t *fp; 4102 int error; 4103 4104 /* fd_getvnode() will use the descriptor for us */ 4105 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4106 return (error); 4107 if ((fp->f_flag & FWRITE) == 0) { 4108 fd_putfile(SCARG(uap, fd)); 4109 return (EBADF); 4110 } 4111 vp = fp->f_data; 4112 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4113 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4114 VOP_UNLOCK(vp); 4115 fd_putfile(SCARG(uap, fd)); 4116 return (error); 4117 } 4118 4119 /* 4120 * Rename files, (standard) BSD semantics frontend. 4121 */ 4122 /* ARGSUSED */ 4123 int 4124 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4125 { 4126 /* { 4127 syscallarg(const char *) from; 4128 syscallarg(const char *) to; 4129 } */ 4130 4131 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4132 SCARG(uap, to), UIO_USERSPACE, 0)); 4133 } 4134 4135 int 4136 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4137 register_t *retval) 4138 { 4139 /* { 4140 syscallarg(int) fromfd; 4141 syscallarg(const char *) from; 4142 syscallarg(int) tofd; 4143 syscallarg(const char *) to; 4144 } */ 4145 4146 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4147 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4148 } 4149 4150 /* 4151 * Rename files, POSIX semantics frontend. 4152 */ 4153 /* ARGSUSED */ 4154 int 4155 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4156 { 4157 /* { 4158 syscallarg(const char *) from; 4159 syscallarg(const char *) to; 4160 } */ 4161 4162 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4163 SCARG(uap, to), UIO_USERSPACE, 1)); 4164 } 4165 4166 /* 4167 * Rename files. Source and destination must either both be directories, 4168 * or both not be directories. If target is a directory, it must be empty. 4169 * If `from' and `to' refer to the same object, the value of the `retain' 4170 * argument is used to determine whether `from' will be 4171 * 4172 * (retain == 0) deleted unless `from' and `to' refer to the same 4173 * object in the file system's name space (BSD). 4174 * (retain == 1) always retained (POSIX). 4175 * 4176 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4177 */ 4178 int 4179 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4180 { 4181 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4182 } 4183 4184 static int 4185 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4186 const char *to, enum uio_seg seg, int retain) 4187 { 4188 struct pathbuf *fpb, *tpb; 4189 struct nameidata fnd, tnd; 4190 struct vnode *fdvp, *fvp; 4191 struct vnode *tdvp, *tvp; 4192 struct mount *mp, *tmp; 4193 int error; 4194 4195 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4196 KASSERT(from != NULL); 4197 KASSERT(to != NULL); 4198 4199 error = pathbuf_maybe_copyin(from, seg, &fpb); 4200 if (error) 4201 goto out0; 4202 KASSERT(fpb != NULL); 4203 4204 error = pathbuf_maybe_copyin(to, seg, &tpb); 4205 if (error) 4206 goto out1; 4207 KASSERT(tpb != NULL); 4208 4209 /* 4210 * Lookup from. 4211 * 4212 * XXX LOCKPARENT is wrong because we don't actually want it 4213 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4214 * insane, so for the time being we need to leave it like this. 4215 */ 4216 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT | INRENAME), fpb); 4217 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4218 goto out2; 4219 4220 /* 4221 * Pull out the important results of the lookup, fdvp and fvp. 4222 * Of course, fvp is bogus because we're about to unlock fdvp. 4223 */ 4224 fdvp = fnd.ni_dvp; 4225 fvp = fnd.ni_vp; 4226 KASSERT(fdvp != NULL); 4227 KASSERT(fvp != NULL); 4228 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4229 4230 /* 4231 * Make sure neither fdvp nor fvp is locked. 4232 */ 4233 if (fdvp != fvp) 4234 VOP_UNLOCK(fdvp); 4235 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4236 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4237 4238 /* 4239 * Reject renaming `.' and `..'. Can't do this until after 4240 * namei because we need namei's parsing to find the final 4241 * component name. (namei should just leave us with the final 4242 * component name and not look it up itself, but anyway...) 4243 * 4244 * This was here before because we used to relookup from 4245 * instead of to and relookup requires the caller to check 4246 * this, but now file systems may depend on this check, so we 4247 * must retain it until the file systems are all rototilled. 4248 */ 4249 if (((fnd.ni_cnd.cn_namelen == 1) && 4250 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4251 ((fnd.ni_cnd.cn_namelen == 2) && 4252 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4253 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4254 error = EINVAL; /* XXX EISDIR? */ 4255 goto abort0; 4256 } 4257 4258 /* 4259 * Lookup to. 4260 * 4261 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4262 * fvp here to decide whether to add CREATEDIR is a load of 4263 * bollocks because fvp might be the wrong node by now, since 4264 * fdvp is unlocked. 4265 * 4266 * XXX Why not pass CREATEDIR always? 4267 */ 4268 NDINIT(&tnd, RENAME, 4269 (LOCKPARENT | NOCACHE | TRYEMULROOT | INRENAME | 4270 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4271 tpb); 4272 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4273 goto abort0; 4274 4275 /* 4276 * Pull out the important results of the lookup, tdvp and tvp. 4277 * Of course, tvp is bogus because we're about to unlock tdvp. 4278 */ 4279 tdvp = tnd.ni_dvp; 4280 tvp = tnd.ni_vp; 4281 KASSERT(tdvp != NULL); 4282 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4283 4284 /* 4285 * Make sure neither tdvp nor tvp is locked. 4286 */ 4287 if (tdvp != tvp) 4288 VOP_UNLOCK(tdvp); 4289 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4290 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4291 4292 /* 4293 * Reject renaming onto `.' or `..'. relookup is unhappy with 4294 * these, which is why we must do this here. Once upon a time 4295 * we relooked up from instead of to, and consequently didn't 4296 * need this check, but now that we relookup to instead of 4297 * from, we need this; and we shall need it forever forward 4298 * until the VOP_RENAME protocol changes, because file systems 4299 * will no doubt begin to depend on this check. 4300 */ 4301 if (((tnd.ni_cnd.cn_namelen == 1) && 4302 (tnd.ni_cnd.cn_nameptr[0] == '.')) || 4303 ((tnd.ni_cnd.cn_namelen == 2) && 4304 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4305 (tnd.ni_cnd.cn_nameptr[1] == '.'))) { 4306 error = EINVAL; /* XXX EISDIR? */ 4307 goto abort1; 4308 } 4309 4310 /* 4311 * Get the mount point. If the file system has been unmounted, 4312 * which it may be because we're not holding any vnode locks, 4313 * then v_mount will be NULL. We're not really supposed to 4314 * read v_mount without holding the vnode lock, but since we 4315 * have fdvp referenced, if fdvp->v_mount changes then at worst 4316 * it will be set to NULL, not changed to another mount point. 4317 * And, of course, since it is up to the file system to 4318 * determine the real lock order, we can't lock both fdvp and 4319 * tdvp at the same time. 4320 */ 4321 mp = fdvp->v_mount; 4322 if (mp == NULL) { 4323 error = ENOENT; 4324 goto abort1; 4325 } 4326 4327 /* 4328 * Make sure the mount points match. Again, although we don't 4329 * hold any vnode locks, the v_mount fields may change -- but 4330 * at worst they will change to NULL, so this will never become 4331 * a cross-device rename, because we hold vnode references. 4332 * 4333 * XXX Because nothing is locked and the compiler may reorder 4334 * things here, unmounting the file system at an inopportune 4335 * moment may cause rename to fail with ENXDEV when it really 4336 * should fail with ENOENT. 4337 */ 4338 tmp = tdvp->v_mount; 4339 if (tmp == NULL) { 4340 error = ENOENT; 4341 goto abort1; 4342 } 4343 4344 if (mp != tmp) { 4345 error = EXDEV; 4346 goto abort1; 4347 } 4348 4349 /* 4350 * Take the vfs rename lock to avoid cross-directory screw cases. 4351 * Nothing is locked currently, so taking this lock is safe. 4352 */ 4353 error = VFS_RENAMELOCK_ENTER(mp); 4354 if (error) 4355 goto abort1; 4356 4357 /* 4358 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4359 * and nothing is locked except for the vfs rename lock. 4360 * 4361 * The next step is a little rain dance to conform to the 4362 * insane lock protocol, even though it does nothing to ward 4363 * off race conditions. 4364 * 4365 * We need tdvp and tvp to be locked. However, because we have 4366 * unlocked tdvp in order to hold no locks while we take the 4367 * vfs rename lock, tvp may be wrong here, and we can't safely 4368 * lock it even if the sensible file systems will just unlock 4369 * it straight away. Consequently, we must lock tdvp and then 4370 * relookup tvp to get it locked. 4371 * 4372 * Finally, because the VOP_RENAME protocol is brain-damaged 4373 * and various file systems insanely depend on the semantics of 4374 * this brain damage, the lookup of to must be the last lookup 4375 * before VOP_RENAME. 4376 */ 4377 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4378 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4379 if (error) 4380 goto abort2; 4381 4382 /* 4383 * Drop the old tvp and pick up the new one -- which might be 4384 * the same, but that doesn't matter to us. After this, tdvp 4385 * and tvp should both be locked. 4386 */ 4387 if (tvp != NULL) 4388 vrele(tvp); 4389 tvp = tnd.ni_vp; 4390 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4391 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4392 4393 /* 4394 * The old do_sys_rename had various consistency checks here 4395 * involving fvp and tvp. fvp is bogus already here, and tvp 4396 * will become bogus soon in any sensible file system, so the 4397 * only purpose in putting these checks here is to give lip 4398 * service to these screw cases and to acknowledge that they 4399 * exist, not actually to handle them, but here you go 4400 * anyway... 4401 */ 4402 4403 /* 4404 * Acknowledge that directories and non-directories aren't 4405 * suposed to mix. 4406 */ 4407 if (tvp != NULL) { 4408 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4409 error = ENOTDIR; 4410 goto abort3; 4411 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4412 error = EISDIR; 4413 goto abort3; 4414 } 4415 } 4416 4417 /* 4418 * Acknowledge some random screw case, among the dozens that 4419 * might arise. 4420 */ 4421 if (fvp == tdvp) { 4422 error = EINVAL; 4423 goto abort3; 4424 } 4425 4426 /* 4427 * Acknowledge that POSIX has a wacky screw case. 4428 * 4429 * XXX Eventually the retain flag needs to be passed on to 4430 * VOP_RENAME. 4431 */ 4432 if (fvp == tvp) { 4433 if (retain) { 4434 error = 0; 4435 goto abort3; 4436 } else if ((fdvp == tdvp) && 4437 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4438 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4439 fnd.ni_cnd.cn_namelen))) { 4440 error = 0; 4441 goto abort3; 4442 } 4443 } 4444 4445 /* 4446 * Make sure veriexec can screw us up. (But a race can screw 4447 * up veriexec, of course -- remember, fvp and (soon) tvp are 4448 * bogus.) 4449 */ 4450 #if NVERIEXEC > 0 4451 { 4452 char *f1, *f2; 4453 size_t f1_len; 4454 size_t f2_len; 4455 4456 f1_len = fnd.ni_cnd.cn_namelen + 1; 4457 f1 = kmem_alloc(f1_len, KM_SLEEP); 4458 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4459 4460 f2_len = tnd.ni_cnd.cn_namelen + 1; 4461 f2 = kmem_alloc(f2_len, KM_SLEEP); 4462 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4463 4464 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4465 4466 kmem_free(f1, f1_len); 4467 kmem_free(f2, f2_len); 4468 4469 if (error) 4470 goto abort3; 4471 } 4472 #endif /* NVERIEXEC > 0 */ 4473 4474 /* 4475 * All ready. Incant the rename vop. 4476 */ 4477 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4478 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4479 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4480 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4481 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4482 4483 /* 4484 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4485 * tdvp and tvp. But we can't assert any of that. 4486 */ 4487 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4488 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4489 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4490 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4491 4492 /* 4493 * So all we have left to do is to drop the rename lock and 4494 * destroy the pathbufs. 4495 */ 4496 VFS_RENAMELOCK_EXIT(mp); 4497 goto out2; 4498 4499 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4500 VOP_UNLOCK(tvp); 4501 abort2: VOP_UNLOCK(tdvp); 4502 VFS_RENAMELOCK_EXIT(mp); 4503 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4504 vrele(tdvp); 4505 if (tvp != NULL) 4506 vrele(tvp); 4507 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4508 vrele(fdvp); 4509 vrele(fvp); 4510 out2: pathbuf_destroy(tpb); 4511 out1: pathbuf_destroy(fpb); 4512 out0: return error; 4513 } 4514 4515 /* 4516 * Make a directory file. 4517 */ 4518 /* ARGSUSED */ 4519 int 4520 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4521 { 4522 /* { 4523 syscallarg(const char *) path; 4524 syscallarg(int) mode; 4525 } */ 4526 4527 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4528 SCARG(uap, mode), UIO_USERSPACE); 4529 } 4530 4531 int 4532 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4533 register_t *retval) 4534 { 4535 /* { 4536 syscallarg(int) fd; 4537 syscallarg(const char *) path; 4538 syscallarg(int) mode; 4539 } */ 4540 4541 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4542 SCARG(uap, mode), UIO_USERSPACE); 4543 } 4544 4545 4546 int 4547 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4548 { 4549 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, UIO_USERSPACE); 4550 } 4551 4552 static int 4553 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4554 enum uio_seg seg) 4555 { 4556 struct proc *p = curlwp->l_proc; 4557 struct vnode *vp; 4558 struct vattr vattr; 4559 int error; 4560 struct pathbuf *pb; 4561 struct nameidata nd; 4562 4563 KASSERT(l != NULL || fdat == AT_FDCWD); 4564 4565 /* XXX bollocks, should pass in a pathbuf */ 4566 error = pathbuf_maybe_copyin(path, seg, &pb); 4567 if (error) { 4568 return error; 4569 } 4570 4571 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4572 4573 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4574 pathbuf_destroy(pb); 4575 return (error); 4576 } 4577 vp = nd.ni_vp; 4578 if (vp != NULL) { 4579 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4580 if (nd.ni_dvp == vp) 4581 vrele(nd.ni_dvp); 4582 else 4583 vput(nd.ni_dvp); 4584 vrele(vp); 4585 pathbuf_destroy(pb); 4586 return (EEXIST); 4587 } 4588 vattr_null(&vattr); 4589 vattr.va_type = VDIR; 4590 /* We will read cwdi->cwdi_cmask unlocked. */ 4591 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4592 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4593 if (!error) 4594 vput(nd.ni_vp); 4595 pathbuf_destroy(pb); 4596 return (error); 4597 } 4598 4599 /* 4600 * Remove a directory file. 4601 */ 4602 /* ARGSUSED */ 4603 int 4604 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4605 { 4606 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4607 AT_REMOVEDIR, UIO_USERSPACE); 4608 } 4609 4610 /* 4611 * Read a block of directory entries in a file system independent format. 4612 */ 4613 int 4614 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4615 { 4616 /* { 4617 syscallarg(int) fd; 4618 syscallarg(char *) buf; 4619 syscallarg(size_t) count; 4620 } */ 4621 file_t *fp; 4622 int error, done; 4623 4624 /* fd_getvnode() will use the descriptor for us */ 4625 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4626 return (error); 4627 if ((fp->f_flag & FREAD) == 0) { 4628 error = EBADF; 4629 goto out; 4630 } 4631 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4632 SCARG(uap, count), &done, l, 0, 0); 4633 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4634 *retval = done; 4635 out: 4636 fd_putfile(SCARG(uap, fd)); 4637 return (error); 4638 } 4639 4640 /* 4641 * Set the mode mask for creation of filesystem nodes. 4642 */ 4643 int 4644 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4645 { 4646 /* { 4647 syscallarg(mode_t) newmask; 4648 } */ 4649 struct proc *p = l->l_proc; 4650 struct cwdinfo *cwdi; 4651 4652 /* 4653 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 4654 * important is that we serialize changes to the mask. The 4655 * rw_exit() will issue a write memory barrier on our behalf, 4656 * and force the changes out to other CPUs (as it must use an 4657 * atomic operation, draining the local CPU's store buffers). 4658 */ 4659 cwdi = p->p_cwdi; 4660 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 4661 *retval = cwdi->cwdi_cmask; 4662 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 4663 rw_exit(&cwdi->cwdi_lock); 4664 4665 return (0); 4666 } 4667 4668 int 4669 dorevoke(struct vnode *vp, kauth_cred_t cred) 4670 { 4671 struct vattr vattr; 4672 int error, fs_decision; 4673 4674 vn_lock(vp, LK_SHARED | LK_RETRY); 4675 error = VOP_GETATTR(vp, &vattr, cred); 4676 VOP_UNLOCK(vp); 4677 if (error != 0) 4678 return error; 4679 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4680 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4681 fs_decision); 4682 if (!error) 4683 VOP_REVOKE(vp, REVOKEALL); 4684 return (error); 4685 } 4686 4687 /* 4688 * Void all references to file by ripping underlying filesystem 4689 * away from vnode. 4690 */ 4691 /* ARGSUSED */ 4692 int 4693 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4694 { 4695 /* { 4696 syscallarg(const char *) path; 4697 } */ 4698 struct vnode *vp; 4699 int error; 4700 4701 error = namei_simple_user(SCARG(uap, path), 4702 NSM_FOLLOW_TRYEMULROOT, &vp); 4703 if (error != 0) 4704 return (error); 4705 error = dorevoke(vp, l->l_cred); 4706 vrele(vp); 4707 return (error); 4708 } 4709
Indexes created Fri Oct 03 02:09:57 GMT 2025