vfs_syscalls.c revision 1.552
1 /* $NetBSD: vfs_syscalls.c,v 1.552 2021/09/11 10:08:55 riastradh Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009, 2019, 2020 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 1.552 2021/09/11 10:08:55 riastradh Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/fcntl.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/fstrans.h> 91 #include <sys/proc.h> 92 #include <sys/uio.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/quota.h> 99 #include <sys/quotactl.h> 100 #include <sys/ktrace.h> 101 #ifdef FILEASSOC 102 #include <sys/fileassoc.h> 103 #endif /* FILEASSOC */ 104 #include <sys/extattr.h> 105 #include <sys/verified_exec.h> 106 #include <sys/kauth.h> 107 #include <sys/atomic.h> 108 #include <sys/module.h> 109 #include <sys/buf.h> 110 #include <sys/event.h> 111 #include <sys/compat_stub.h> 112 113 #include <miscfs/genfs/genfs.h> 114 #include <miscfs/specfs/specdev.h> 115 116 #include <nfs/rpcv2.h> 117 #include <nfs/nfsproto.h> 118 #include <nfs/nfs.h> 119 #include <nfs/nfs_var.h> 120 121 /* XXX this shouldn't be here */ 122 #ifndef OFF_T_MAX 123 #define OFF_T_MAX __type_max(off_t) 124 #endif 125 126 static int change_flags(struct vnode *, u_long, struct lwp *); 127 static int change_mode(struct vnode *, int, struct lwp *); 128 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 129 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 130 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 131 enum uio_seg); 132 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 133 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 134 enum uio_seg); 135 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 136 enum uio_seg, int); 137 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 138 size_t, register_t *); 139 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 140 141 static int fd_nameiat(struct lwp *, int, struct nameidata *); 142 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 143 namei_simple_flags_t, struct vnode **); 144 145 /* 146 * This table is used to maintain compatibility with 4.3BSD 147 * and NetBSD 0.9 mount syscalls - and possibly other systems. 148 * Note, the order is important! 149 * 150 * Do not modify this table. It should only contain filesystems 151 * supported by NetBSD 0.9 and 4.3BSD. 152 */ 153 const char * const mountcompatnames[] = { 154 NULL, /* 0 = MOUNT_NONE */ 155 MOUNT_FFS, /* 1 = MOUNT_UFS */ 156 MOUNT_NFS, /* 2 */ 157 MOUNT_MFS, /* 3 */ 158 MOUNT_MSDOS, /* 4 */ 159 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 160 MOUNT_FDESC, /* 6 */ 161 MOUNT_KERNFS, /* 7 */ 162 NULL, /* 8 = MOUNT_DEVFS */ 163 MOUNT_AFS, /* 9 */ 164 }; 165 166 const u_int nmountcompatnames = __arraycount(mountcompatnames); 167 168 static int 169 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 170 { 171 file_t *dfp; 172 int error; 173 174 if (fdat != AT_FDCWD) { 175 if ((error = fd_getvnode(fdat, &dfp)) != 0) 176 goto out; 177 178 NDAT(ndp, dfp->f_vnode); 179 } 180 181 error = namei(ndp); 182 183 if (fdat != AT_FDCWD) 184 fd_putfile(fdat); 185 out: 186 return error; 187 } 188 189 static int 190 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 191 namei_simple_flags_t sflags, struct vnode **vp_ret) 192 { 193 file_t *dfp; 194 struct vnode *dvp; 195 int error; 196 197 if (fdat != AT_FDCWD) { 198 if ((error = fd_getvnode(fdat, &dfp)) != 0) 199 goto out; 200 201 dvp = dfp->f_vnode; 202 } else { 203 dvp = NULL; 204 } 205 206 error = nameiat_simple_user(dvp, path, sflags, vp_ret); 207 208 if (fdat != AT_FDCWD) 209 fd_putfile(fdat); 210 out: 211 return error; 212 } 213 214 static int 215 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 216 { 217 int error; 218 219 fp->f_flag = flags & FMASK; 220 fp->f_type = DTYPE_VNODE; 221 fp->f_ops = &vnops; 222 fp->f_vnode = vp; 223 224 if (flags & (O_EXLOCK | O_SHLOCK)) { 225 struct flock lf; 226 int type; 227 228 lf.l_whence = SEEK_SET; 229 lf.l_start = 0; 230 lf.l_len = 0; 231 if (flags & O_EXLOCK) 232 lf.l_type = F_WRLCK; 233 else 234 lf.l_type = F_RDLCK; 235 type = F_FLOCK; 236 if ((flags & FNONBLOCK) == 0) 237 type |= F_WAIT; 238 VOP_UNLOCK(vp); 239 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 240 if (error) { 241 (void) vn_close(vp, fp->f_flag, fp->f_cred); 242 fd_abort(l->l_proc, fp, indx); 243 return error; 244 } 245 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 246 atomic_or_uint(&fp->f_flag, FHASLOCK); 247 } 248 if (flags & O_CLOEXEC) 249 fd_set_exclose(l, indx, true); 250 return 0; 251 } 252 253 static int 254 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 255 void *data, size_t *data_len) 256 { 257 struct mount *mp; 258 int error = 0, saved_flags; 259 260 mp = vp->v_mount; 261 saved_flags = mp->mnt_flag; 262 263 /* We can operate only on VV_ROOT nodes. */ 264 if ((vp->v_vflag & VV_ROOT) == 0) { 265 error = EINVAL; 266 goto out; 267 } 268 269 /* 270 * We only allow the filesystem to be reloaded if it 271 * is currently mounted read-only. Additionally, we 272 * prevent read-write to read-only downgrades. 273 */ 274 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 275 (mp->mnt_flag & MNT_RDONLY) == 0 && 276 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 277 error = EOPNOTSUPP; /* Needs translation */ 278 goto out; 279 } 280 281 /* 282 * Enabling MNT_UNION requires a covered mountpoint and 283 * must not happen on the root mount. 284 */ 285 if ((flags & MNT_UNION) != 0 && mp->mnt_vnodecovered == NULLVP) { 286 error = EOPNOTSUPP; 287 goto out; 288 } 289 290 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 291 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 292 if (error) 293 goto out; 294 295 error = vfs_suspend(mp, 0); 296 if (error) 297 goto out; 298 299 mutex_enter(mp->mnt_updating); 300 301 mp->mnt_flag &= ~MNT_OP_FLAGS; 302 mp->mnt_flag |= flags & MNT_OP_FLAGS; 303 304 /* 305 * Set the mount level flags. 306 */ 307 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 308 if ((flags & MNT_RDONLY)) 309 mp->mnt_iflag |= IMNT_WANTRDONLY; 310 else 311 mp->mnt_iflag |= IMNT_WANTRDWR; 312 } 313 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 314 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 315 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 316 mp->mnt_flag &= ~MNT_RDONLY; 317 318 error = VFS_MOUNT(mp, path, data, data_len); 319 320 if (error && data != NULL) { 321 int error2; 322 323 /* 324 * Update failed; let's try and see if it was an 325 * export request. For compat with 3.0 and earlier. 326 */ 327 error2 = vfs_hooks_reexport(mp, path, data); 328 329 /* 330 * Only update error code if the export request was 331 * understood but some problem occurred while 332 * processing it. 333 */ 334 if (error2 != EJUSTRETURN) 335 error = error2; 336 } 337 338 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 339 mp->mnt_flag |= MNT_RDONLY; 340 if (error) 341 mp->mnt_flag = saved_flags; 342 mp->mnt_flag &= ~MNT_OP_FLAGS; 343 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 344 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 345 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 346 vfs_syncer_add_to_worklist(mp); 347 } else { 348 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 349 vfs_syncer_remove_from_worklist(mp); 350 } 351 mutex_exit(mp->mnt_updating); 352 vfs_resume(mp); 353 354 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 355 (flags & MNT_EXTATTR)) { 356 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 357 NULL, 0, NULL) != 0) { 358 printf("%s: failed to start extattr, error = %d", 359 mp->mnt_stat.f_mntonname, error); 360 mp->mnt_flag &= ~MNT_EXTATTR; 361 } 362 } 363 364 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 365 !(flags & MNT_EXTATTR)) { 366 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 367 NULL, 0, NULL) != 0) { 368 printf("%s: failed to stop extattr, error = %d", 369 mp->mnt_stat.f_mntonname, error); 370 mp->mnt_flag |= MNT_RDONLY; 371 } 372 } 373 out: 374 return (error); 375 } 376 377 static int 378 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 379 struct vfsops **vfsops) 380 { 381 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 382 int error; 383 384 if (type_seg == UIO_USERSPACE) { 385 /* Copy file-system type from userspace. */ 386 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 387 } else { 388 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 389 KASSERT(error == 0); 390 } 391 392 if (error) { 393 /* 394 * Historically, filesystem types were identified by numbers. 395 * If we get an integer for the filesystem type instead of a 396 * string, we check to see if it matches one of the historic 397 * filesystem types. 398 */ 399 u_long fsindex = (u_long)fstype; 400 if (fsindex >= nmountcompatnames || 401 mountcompatnames[fsindex] == NULL) 402 return ENODEV; 403 strlcpy(fstypename, mountcompatnames[fsindex], 404 sizeof(fstypename)); 405 } 406 407 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 408 if (strcmp(fstypename, "ufs") == 0) 409 fstypename[0] = 'f'; 410 411 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 412 return 0; 413 414 /* If we can autoload a vfs module, try again */ 415 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 416 417 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 418 return 0; 419 420 return ENODEV; 421 } 422 423 static int 424 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 425 void *data, size_t *data_len) 426 { 427 struct mount *mp; 428 int error; 429 430 /* If MNT_GETARGS is specified, it should be the only flag. */ 431 if (flags & ~MNT_GETARGS) 432 return EINVAL; 433 434 mp = vp->v_mount; 435 436 /* XXX: probably some notion of "can see" here if we want isolation. */ 437 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 438 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 439 if (error) 440 return error; 441 442 if ((vp->v_vflag & VV_ROOT) == 0) 443 return EINVAL; 444 445 if (vfs_busy(mp)) 446 return EPERM; 447 448 mutex_enter(mp->mnt_updating); 449 mp->mnt_flag &= ~MNT_OP_FLAGS; 450 mp->mnt_flag |= MNT_GETARGS; 451 error = VFS_MOUNT(mp, path, data, data_len); 452 mp->mnt_flag &= ~MNT_OP_FLAGS; 453 mutex_exit(mp->mnt_updating); 454 455 vfs_unbusy(mp); 456 return (error); 457 } 458 459 int 460 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 461 { 462 /* { 463 syscallarg(const char *) type; 464 syscallarg(const char *) path; 465 syscallarg(int) flags; 466 syscallarg(void *) data; 467 syscallarg(size_t) data_len; 468 } */ 469 470 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 471 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 472 SCARG(uap, data_len), retval); 473 } 474 475 int 476 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 477 const char *path, int flags, void *data, enum uio_seg data_seg, 478 size_t data_len, register_t *retval) 479 { 480 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 481 struct vnode *vp; 482 void *data_buf = data; 483 bool vfsopsrele = false; 484 size_t alloc_sz = 0; 485 int error; 486 487 /* 488 * Get vnode to be covered 489 */ 490 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 491 if (error != 0) { 492 vp = NULL; 493 goto done; 494 } 495 496 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 497 vfsops = vp->v_mount->mnt_op; 498 } else { 499 /* 'type' is userspace */ 500 error = mount_get_vfsops(type, type_seg, &vfsops); 501 if (error != 0) 502 goto done; 503 vfsopsrele = true; 504 } 505 506 /* 507 * We allow data to be NULL, even for userspace. Some fs's don't need 508 * it. The others will handle NULL. 509 */ 510 if (data != NULL && data_seg == UIO_USERSPACE) { 511 if (data_len == 0) { 512 /* No length supplied, use default for filesystem */ 513 data_len = vfsops->vfs_min_mount_data; 514 515 /* 516 * Hopefully a longer buffer won't make copyin() fail. 517 * For compatibility with 3.0 and earlier. 518 */ 519 if (flags & MNT_UPDATE 520 && data_len < sizeof (struct mnt_export_args30)) 521 data_len = sizeof (struct mnt_export_args30); 522 } 523 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 524 error = EINVAL; 525 goto done; 526 } 527 alloc_sz = data_len; 528 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 529 530 /* NFS needs the buffer even for mnt_getargs .... */ 531 error = copyin(data, data_buf, data_len); 532 if (error != 0) 533 goto done; 534 } 535 536 if (flags & MNT_GETARGS) { 537 if (data_len == 0) { 538 error = EINVAL; 539 goto done; 540 } 541 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 542 if (error != 0) 543 goto done; 544 if (data_seg == UIO_USERSPACE) 545 error = copyout(data_buf, data, data_len); 546 *retval = data_len; 547 } else if (flags & MNT_UPDATE) { 548 error = mount_update(l, vp, path, flags, data_buf, &data_len); 549 } else { 550 /* Locking is handled internally in mount_domount(). */ 551 KASSERT(vfsopsrele == true); 552 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 553 &data_len); 554 vfsopsrele = false; 555 } 556 if (!error) 557 KNOTE(&fs_klist, VQ_MOUNT); 558 559 done: 560 if (vfsopsrele) 561 vfs_delref(vfsops); 562 if (vp != NULL) { 563 vrele(vp); 564 } 565 if (data_buf != data) 566 kmem_free(data_buf, alloc_sz); 567 return (error); 568 } 569 570 /* 571 * Unmount a file system. 572 * 573 * Note: unmount takes a path to the vnode mounted on as argument, 574 * not special file (as before). 575 */ 576 /* ARGSUSED */ 577 int 578 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 579 { 580 /* { 581 syscallarg(const char *) path; 582 syscallarg(int) flags; 583 } */ 584 struct vnode *vp; 585 struct mount *mp; 586 int error; 587 struct pathbuf *pb; 588 struct nameidata nd; 589 590 error = pathbuf_copyin(SCARG(uap, path), &pb); 591 if (error) { 592 return error; 593 } 594 595 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 596 if ((error = namei(&nd)) != 0) { 597 pathbuf_destroy(pb); 598 return error; 599 } 600 vp = nd.ni_vp; 601 pathbuf_destroy(pb); 602 603 mp = vp->v_mount; 604 vfs_ref(mp); 605 VOP_UNLOCK(vp); 606 607 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 608 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 609 if (error) { 610 vrele(vp); 611 vfs_rele(mp); 612 return (error); 613 } 614 615 /* 616 * Don't allow unmounting the root file system. 617 */ 618 if (mp->mnt_flag & MNT_ROOTFS) { 619 vrele(vp); 620 vfs_rele(mp); 621 return (EINVAL); 622 } 623 624 /* 625 * Must be the root of the filesystem 626 */ 627 if ((vp->v_vflag & VV_ROOT) == 0) { 628 vrele(vp); 629 vfs_rele(mp); 630 return (EINVAL); 631 } 632 633 vrele(vp); 634 error = dounmount(mp, SCARG(uap, flags), l); 635 vfs_rele(mp); 636 if (!error) 637 KNOTE(&fs_klist, VQ_UNMOUNT); 638 return error; 639 } 640 641 /* 642 * Sync each mounted filesystem. 643 */ 644 #ifdef DEBUG 645 int syncprt = 0; 646 struct ctldebug debug0 = { "syncprt", &syncprt }; 647 #endif 648 649 void 650 do_sys_sync(struct lwp *l) 651 { 652 mount_iterator_t *iter; 653 struct mount *mp; 654 int asyncflag; 655 656 mountlist_iterator_init(&iter); 657 while ((mp = mountlist_iterator_next(iter)) != NULL) { 658 mutex_enter(mp->mnt_updating); 659 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 660 asyncflag = mp->mnt_flag & MNT_ASYNC; 661 mp->mnt_flag &= ~MNT_ASYNC; 662 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 663 if (asyncflag) 664 mp->mnt_flag |= MNT_ASYNC; 665 } 666 mutex_exit(mp->mnt_updating); 667 } 668 mountlist_iterator_destroy(iter); 669 #ifdef DEBUG 670 if (syncprt) 671 vfs_bufstats(); 672 #endif /* DEBUG */ 673 } 674 675 static bool 676 sync_vnode_filter(void *cookie, vnode_t *vp) 677 { 678 679 if (vp->v_numoutput > 0) { 680 ++*(int *)cookie; 681 } 682 return false; 683 } 684 685 int 686 vfs_syncwait(void) 687 { 688 int nbusy, nbusy_prev, iter; 689 struct vnode_iterator *vniter; 690 mount_iterator_t *mpiter; 691 struct mount *mp; 692 693 for (nbusy_prev = 0, iter = 0; iter < 20;) { 694 nbusy = 0; 695 mountlist_iterator_init(&mpiter); 696 while ((mp = mountlist_iterator_next(mpiter)) != NULL) { 697 vnode_t *vp __diagused; 698 vfs_vnode_iterator_init(mp, &vniter); 699 vp = vfs_vnode_iterator_next(vniter, 700 sync_vnode_filter, &nbusy); 701 KASSERT(vp == NULL); 702 vfs_vnode_iterator_destroy(vniter); 703 } 704 mountlist_iterator_destroy(mpiter); 705 706 if (nbusy == 0) 707 break; 708 if (nbusy_prev == 0) 709 nbusy_prev = nbusy; 710 printf("%d ", nbusy); 711 kpause("syncwait", false, MAX(1, hz / 25 * iter), NULL); 712 if (nbusy >= nbusy_prev) /* we didn't flush anything */ 713 iter++; 714 else 715 nbusy_prev = nbusy; 716 } 717 718 if (nbusy) { 719 #if defined(DEBUG) || defined(DEBUG_HALT_BUSY) 720 printf("giving up\nPrinting vnodes for busy buffers\n"); 721 mountlist_iterator_init(&mpiter); 722 while ((mp = mountlist_iterator_next(mpiter)) != NULL) { 723 vnode_t *vp; 724 vfs_vnode_iterator_init(mp, &vniter); 725 vp = vfs_vnode_iterator_next(vniter, 726 NULL, NULL); 727 mutex_enter(vp->v_interlock); 728 if (vp->v_numoutput > 0) 729 vprint(NULL, vp); 730 mutex_exit(vp->v_interlock); 731 vrele(vp); 732 vfs_vnode_iterator_destroy(vniter); 733 } 734 mountlist_iterator_destroy(mpiter); 735 #endif 736 } 737 738 return nbusy; 739 } 740 741 /* ARGSUSED */ 742 int 743 sys_sync(struct lwp *l, const void *v, register_t *retval) 744 { 745 do_sys_sync(l); 746 return (0); 747 } 748 749 750 /* 751 * Access or change filesystem quotas. 752 * 753 * (this is really 14 different calls bundled into one) 754 */ 755 756 static int 757 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 758 { 759 struct quotastat info_k; 760 int error; 761 762 /* ensure any padding bytes are cleared */ 763 memset(&info_k, 0, sizeof(info_k)); 764 765 error = vfs_quotactl_stat(mp, &info_k); 766 if (error) { 767 return error; 768 } 769 770 return copyout(&info_k, info_u, sizeof(info_k)); 771 } 772 773 static int 774 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 775 struct quotaidtypestat *info_u) 776 { 777 struct quotaidtypestat info_k; 778 int error; 779 780 /* ensure any padding bytes are cleared */ 781 memset(&info_k, 0, sizeof(info_k)); 782 783 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 784 if (error) { 785 return error; 786 } 787 788 return copyout(&info_k, info_u, sizeof(info_k)); 789 } 790 791 static int 792 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 793 struct quotaobjtypestat *info_u) 794 { 795 struct quotaobjtypestat info_k; 796 int error; 797 798 /* ensure any padding bytes are cleared */ 799 memset(&info_k, 0, sizeof(info_k)); 800 801 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 802 if (error) { 803 return error; 804 } 805 806 return copyout(&info_k, info_u, sizeof(info_k)); 807 } 808 809 static int 810 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 811 struct quotaval *val_u) 812 { 813 struct quotakey key_k; 814 struct quotaval val_k; 815 int error; 816 817 /* ensure any padding bytes are cleared */ 818 memset(&val_k, 0, sizeof(val_k)); 819 820 error = copyin(key_u, &key_k, sizeof(key_k)); 821 if (error) { 822 return error; 823 } 824 825 error = vfs_quotactl_get(mp, &key_k, &val_k); 826 if (error) { 827 return error; 828 } 829 830 return copyout(&val_k, val_u, sizeof(val_k)); 831 } 832 833 static int 834 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 835 const struct quotaval *val_u) 836 { 837 struct quotakey key_k; 838 struct quotaval val_k; 839 int error; 840 841 error = copyin(key_u, &key_k, sizeof(key_k)); 842 if (error) { 843 return error; 844 } 845 846 error = copyin(val_u, &val_k, sizeof(val_k)); 847 if (error) { 848 return error; 849 } 850 851 return vfs_quotactl_put(mp, &key_k, &val_k); 852 } 853 854 static int 855 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 856 { 857 struct quotakey key_k; 858 int error; 859 860 error = copyin(key_u, &key_k, sizeof(key_k)); 861 if (error) { 862 return error; 863 } 864 865 return vfs_quotactl_del(mp, &key_k); 866 } 867 868 static int 869 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 870 { 871 struct quotakcursor cursor_k; 872 int error; 873 874 /* ensure any padding bytes are cleared */ 875 memset(&cursor_k, 0, sizeof(cursor_k)); 876 877 error = vfs_quotactl_cursoropen(mp, &cursor_k); 878 if (error) { 879 return error; 880 } 881 882 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 883 } 884 885 static int 886 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 887 { 888 struct quotakcursor cursor_k; 889 int error; 890 891 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 892 if (error) { 893 return error; 894 } 895 896 return vfs_quotactl_cursorclose(mp, &cursor_k); 897 } 898 899 static int 900 do_sys_quotactl_cursorskipidtype(struct mount *mp, 901 struct quotakcursor *cursor_u, int idtype) 902 { 903 struct quotakcursor cursor_k; 904 int error; 905 906 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 907 if (error) { 908 return error; 909 } 910 911 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 912 if (error) { 913 return error; 914 } 915 916 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 917 } 918 919 static int 920 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 921 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 922 unsigned *ret_u) 923 { 924 #define CGET_STACK_MAX 8 925 struct quotakcursor cursor_k; 926 struct quotakey stackkeys[CGET_STACK_MAX]; 927 struct quotaval stackvals[CGET_STACK_MAX]; 928 struct quotakey *keys_k; 929 struct quotaval *vals_k; 930 unsigned ret_k; 931 int error; 932 933 if (maxnum > 128) { 934 maxnum = 128; 935 } 936 937 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 938 if (error) { 939 return error; 940 } 941 942 if (maxnum <= CGET_STACK_MAX) { 943 keys_k = stackkeys; 944 vals_k = stackvals; 945 /* ensure any padding bytes are cleared */ 946 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 947 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 948 } else { 949 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 950 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 951 } 952 953 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 954 &ret_k); 955 if (error) { 956 goto fail; 957 } 958 959 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 960 if (error) { 961 goto fail; 962 } 963 964 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 965 if (error) { 966 goto fail; 967 } 968 969 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 970 if (error) { 971 goto fail; 972 } 973 974 /* do last to maximize the chance of being able to recover a failure */ 975 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 976 977 fail: 978 if (keys_k != stackkeys) { 979 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 980 } 981 if (vals_k != stackvals) { 982 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 983 } 984 return error; 985 } 986 987 static int 988 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 989 int *ret_u) 990 { 991 struct quotakcursor cursor_k; 992 int ret_k; 993 int error; 994 995 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 996 if (error) { 997 return error; 998 } 999 1000 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 1001 if (error) { 1002 return error; 1003 } 1004 1005 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 1006 if (error) { 1007 return error; 1008 } 1009 1010 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 1011 } 1012 1013 static int 1014 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 1015 { 1016 struct quotakcursor cursor_k; 1017 int error; 1018 1019 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 1020 if (error) { 1021 return error; 1022 } 1023 1024 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 1025 if (error) { 1026 return error; 1027 } 1028 1029 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 1030 } 1031 1032 static int 1033 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 1034 { 1035 char *path_k; 1036 int error; 1037 1038 /* XXX this should probably be a struct pathbuf */ 1039 path_k = PNBUF_GET(); 1040 error = copyin(path_u, path_k, PATH_MAX); 1041 if (error) { 1042 PNBUF_PUT(path_k); 1043 return error; 1044 } 1045 1046 error = vfs_quotactl_quotaon(mp, idtype, path_k); 1047 1048 PNBUF_PUT(path_k); 1049 return error; 1050 } 1051 1052 static int 1053 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 1054 { 1055 return vfs_quotactl_quotaoff(mp, idtype); 1056 } 1057 1058 int 1059 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 1060 { 1061 struct mount *mp; 1062 struct vnode *vp; 1063 int error; 1064 1065 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 1066 if (error != 0) 1067 return (error); 1068 mp = vp->v_mount; 1069 1070 switch (args->qc_op) { 1071 case QUOTACTL_STAT: 1072 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 1073 break; 1074 case QUOTACTL_IDTYPESTAT: 1075 error = do_sys_quotactl_idtypestat(mp, 1076 args->u.idtypestat.qc_idtype, 1077 args->u.idtypestat.qc_info); 1078 break; 1079 case QUOTACTL_OBJTYPESTAT: 1080 error = do_sys_quotactl_objtypestat(mp, 1081 args->u.objtypestat.qc_objtype, 1082 args->u.objtypestat.qc_info); 1083 break; 1084 case QUOTACTL_GET: 1085 error = do_sys_quotactl_get(mp, 1086 args->u.get.qc_key, 1087 args->u.get.qc_val); 1088 break; 1089 case QUOTACTL_PUT: 1090 error = do_sys_quotactl_put(mp, 1091 args->u.put.qc_key, 1092 args->u.put.qc_val); 1093 break; 1094 case QUOTACTL_DEL: 1095 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1096 break; 1097 case QUOTACTL_CURSOROPEN: 1098 error = do_sys_quotactl_cursoropen(mp, 1099 args->u.cursoropen.qc_cursor); 1100 break; 1101 case QUOTACTL_CURSORCLOSE: 1102 error = do_sys_quotactl_cursorclose(mp, 1103 args->u.cursorclose.qc_cursor); 1104 break; 1105 case QUOTACTL_CURSORSKIPIDTYPE: 1106 error = do_sys_quotactl_cursorskipidtype(mp, 1107 args->u.cursorskipidtype.qc_cursor, 1108 args->u.cursorskipidtype.qc_idtype); 1109 break; 1110 case QUOTACTL_CURSORGET: 1111 error = do_sys_quotactl_cursorget(mp, 1112 args->u.cursorget.qc_cursor, 1113 args->u.cursorget.qc_keys, 1114 args->u.cursorget.qc_vals, 1115 args->u.cursorget.qc_maxnum, 1116 args->u.cursorget.qc_ret); 1117 break; 1118 case QUOTACTL_CURSORATEND: 1119 error = do_sys_quotactl_cursoratend(mp, 1120 args->u.cursoratend.qc_cursor, 1121 args->u.cursoratend.qc_ret); 1122 break; 1123 case QUOTACTL_CURSORREWIND: 1124 error = do_sys_quotactl_cursorrewind(mp, 1125 args->u.cursorrewind.qc_cursor); 1126 break; 1127 case QUOTACTL_QUOTAON: 1128 error = do_sys_quotactl_quotaon(mp, 1129 args->u.quotaon.qc_idtype, 1130 args->u.quotaon.qc_quotafile); 1131 break; 1132 case QUOTACTL_QUOTAOFF: 1133 error = do_sys_quotactl_quotaoff(mp, 1134 args->u.quotaoff.qc_idtype); 1135 break; 1136 default: 1137 error = EINVAL; 1138 break; 1139 } 1140 1141 vrele(vp); 1142 return error; 1143 } 1144 1145 /* ARGSUSED */ 1146 int 1147 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1148 register_t *retval) 1149 { 1150 /* { 1151 syscallarg(const char *) path; 1152 syscallarg(struct quotactl_args *) args; 1153 } */ 1154 struct quotactl_args args; 1155 int error; 1156 1157 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1158 if (error) { 1159 return error; 1160 } 1161 1162 return do_sys_quotactl(SCARG(uap, path), &args); 1163 } 1164 1165 int 1166 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1167 int root) 1168 { 1169 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1170 bool chrooted; 1171 int error = 0; 1172 1173 KASSERT(l == curlwp); 1174 1175 /* 1176 * This is safe unlocked. cwdi_rdir never goes non-NULL -> NULL, 1177 * since it would imply chroots can be escaped. Just make sure this 1178 * routine is self-consistent. 1179 */ 1180 chrooted = (atomic_load_relaxed(&cwdi->cwdi_rdir) != NULL); 1181 1182 /* 1183 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1184 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1185 * overrides MNT_NOWAIT. 1186 */ 1187 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1188 (flags != MNT_WAIT && flags != 0)) { 1189 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1190 } else { 1191 /* Get the filesystem stats now */ 1192 memset(sp, 0, sizeof(*sp)); 1193 if ((error = VFS_STATVFS(mp, sp)) != 0) 1194 return error; 1195 if (!chrooted) 1196 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1197 } 1198 1199 if (chrooted) { 1200 size_t len; 1201 char *bp; 1202 char c; 1203 char *path = PNBUF_GET(); 1204 1205 bp = path + MAXPATHLEN; 1206 *--bp = '\0'; 1207 rw_enter(&cwdi->cwdi_lock, RW_READER); 1208 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1209 MAXPATHLEN / 2, 0, l); 1210 rw_exit(&cwdi->cwdi_lock); 1211 if (error) { 1212 PNBUF_PUT(path); 1213 return error; 1214 } 1215 len = strlen(bp); 1216 if (len != 1) { 1217 /* 1218 * for mount points that are below our root, we can see 1219 * them, so we fix up the pathname and return them. The 1220 * rest we cannot see, so we don't allow viewing the 1221 * data. 1222 */ 1223 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1224 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1225 (void)strlcpy(sp->f_mntonname, 1226 c == '\0' ? "/" : &sp->f_mntonname[len], 1227 sizeof(sp->f_mntonname)); 1228 } else { 1229 if (root) 1230 (void)strlcpy(sp->f_mntonname, "/", 1231 sizeof(sp->f_mntonname)); 1232 else 1233 error = EPERM; 1234 } 1235 } 1236 PNBUF_PUT(path); 1237 } 1238 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1239 return error; 1240 } 1241 1242 /* 1243 * Get filesystem statistics by path. 1244 */ 1245 int 1246 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1247 { 1248 struct mount *mp; 1249 int error; 1250 struct vnode *vp; 1251 1252 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1253 if (error != 0) 1254 return error; 1255 mp = vp->v_mount; 1256 error = dostatvfs(mp, sb, l, flags, 1); 1257 vrele(vp); 1258 return error; 1259 } 1260 1261 /* ARGSUSED */ 1262 int 1263 sys___statvfs190(struct lwp *l, const struct sys___statvfs190_args *uap, register_t *retval) 1264 { 1265 /* { 1266 syscallarg(const char *) path; 1267 syscallarg(struct statvfs *) buf; 1268 syscallarg(int) flags; 1269 } */ 1270 struct statvfs *sb; 1271 int error; 1272 1273 sb = STATVFSBUF_GET(); 1274 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1275 if (error == 0) 1276 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1277 STATVFSBUF_PUT(sb); 1278 return error; 1279 } 1280 1281 /* 1282 * Get filesystem statistics by fd. 1283 */ 1284 int 1285 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1286 { 1287 file_t *fp; 1288 struct mount *mp; 1289 int error; 1290 1291 /* fd_getvnode() will use the descriptor for us */ 1292 if ((error = fd_getvnode(fd, &fp)) != 0) 1293 return (error); 1294 mp = fp->f_vnode->v_mount; 1295 error = dostatvfs(mp, sb, curlwp, flags, 1); 1296 fd_putfile(fd); 1297 return error; 1298 } 1299 1300 /* ARGSUSED */ 1301 int 1302 sys___fstatvfs190(struct lwp *l, const struct sys___fstatvfs190_args *uap, register_t *retval) 1303 { 1304 /* { 1305 syscallarg(int) fd; 1306 syscallarg(struct statvfs *) buf; 1307 syscallarg(int) flags; 1308 } */ 1309 struct statvfs *sb; 1310 int error; 1311 1312 sb = STATVFSBUF_GET(); 1313 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1314 if (error == 0) 1315 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1316 STATVFSBUF_PUT(sb); 1317 return error; 1318 } 1319 1320 1321 /* 1322 * Get statistics on all filesystems. 1323 */ 1324 int 1325 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1326 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1327 register_t *retval) 1328 { 1329 int root = 0; 1330 mount_iterator_t *iter; 1331 struct proc *p = l->l_proc; 1332 struct mount *mp; 1333 struct statvfs *sb; 1334 size_t count, maxcount; 1335 int error = 0; 1336 1337 sb = STATVFSBUF_GET(); 1338 maxcount = bufsize / entry_sz; 1339 count = 0; 1340 mountlist_iterator_init(&iter); 1341 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1342 if (sfsp && count < maxcount) { 1343 error = dostatvfs(mp, sb, l, flags, 0); 1344 if (error) { 1345 error = 0; 1346 continue; 1347 } 1348 error = copyfn(sb, sfsp, entry_sz); 1349 if (error) 1350 goto out; 1351 sfsp = (char *)sfsp + entry_sz; 1352 root |= strcmp(sb->f_mntonname, "/") == 0; 1353 } 1354 count++; 1355 } 1356 1357 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1358 /* 1359 * fake a root entry 1360 */ 1361 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1362 sb, l, flags, 1); 1363 if (error != 0) 1364 goto out; 1365 if (sfsp) { 1366 error = copyfn(sb, sfsp, entry_sz); 1367 if (error != 0) 1368 goto out; 1369 } 1370 count++; 1371 } 1372 if (sfsp && count > maxcount) 1373 *retval = maxcount; 1374 else 1375 *retval = count; 1376 out: 1377 mountlist_iterator_destroy(iter); 1378 STATVFSBUF_PUT(sb); 1379 return error; 1380 } 1381 1382 int 1383 sys___getvfsstat90(struct lwp *l, const struct sys___getvfsstat90_args *uap, 1384 register_t *retval) 1385 { 1386 /* { 1387 syscallarg(struct statvfs *) buf; 1388 syscallarg(size_t) bufsize; 1389 syscallarg(int) flags; 1390 } */ 1391 1392 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1393 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1394 } 1395 1396 /* 1397 * Change current working directory to a given file descriptor. 1398 */ 1399 /* ARGSUSED */ 1400 int 1401 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1402 { 1403 /* { 1404 syscallarg(int) fd; 1405 } */ 1406 struct proc *p = l->l_proc; 1407 struct cwdinfo *cwdi; 1408 struct vnode *vp, *tdp; 1409 struct mount *mp; 1410 file_t *fp; 1411 int error, fd; 1412 1413 /* fd_getvnode() will use the descriptor for us */ 1414 fd = SCARG(uap, fd); 1415 if ((error = fd_getvnode(fd, &fp)) != 0) 1416 return (error); 1417 vp = fp->f_vnode; 1418 1419 vref(vp); 1420 vn_lock(vp, LK_SHARED | LK_RETRY); 1421 if (vp->v_type != VDIR) 1422 error = ENOTDIR; 1423 else 1424 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1425 if (error) { 1426 vput(vp); 1427 goto out; 1428 } 1429 while ((mp = vp->v_mountedhere) != NULL) { 1430 error = vfs_busy(mp); 1431 vput(vp); 1432 if (error != 0) 1433 goto out; 1434 error = VFS_ROOT(mp, LK_SHARED, &tdp); 1435 vfs_unbusy(mp); 1436 if (error) 1437 goto out; 1438 vp = tdp; 1439 } 1440 VOP_UNLOCK(vp); 1441 1442 /* 1443 * Disallow changing to a directory not under the process's 1444 * current root directory (if there is one). 1445 */ 1446 cwdi = p->p_cwdi; 1447 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1448 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1449 vrele(vp); 1450 error = EPERM; /* operation not permitted */ 1451 } else { 1452 vrele(cwdi->cwdi_cdir); 1453 cwdi->cwdi_cdir = vp; 1454 } 1455 rw_exit(&cwdi->cwdi_lock); 1456 1457 out: 1458 fd_putfile(fd); 1459 return (error); 1460 } 1461 1462 /* 1463 * Change this process's notion of the root directory to a given file 1464 * descriptor. 1465 */ 1466 int 1467 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1468 { 1469 struct vnode *vp; 1470 file_t *fp; 1471 int error, fd = SCARG(uap, fd); 1472 1473 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1474 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1475 return error; 1476 /* fd_getvnode() will use the descriptor for us */ 1477 if ((error = fd_getvnode(fd, &fp)) != 0) 1478 return error; 1479 vp = fp->f_vnode; 1480 vn_lock(vp, LK_SHARED | LK_RETRY); 1481 if (vp->v_type != VDIR) 1482 error = ENOTDIR; 1483 else 1484 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1485 VOP_UNLOCK(vp); 1486 if (error) 1487 goto out; 1488 vref(vp); 1489 change_root(vp); 1490 1491 out: 1492 fd_putfile(fd); 1493 return (error); 1494 } 1495 1496 /* 1497 * Change current working directory (``.''). 1498 */ 1499 /* ARGSUSED */ 1500 int 1501 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1502 { 1503 /* { 1504 syscallarg(const char *) path; 1505 } */ 1506 struct proc *p = l->l_proc; 1507 struct cwdinfo *cwdi; 1508 int error; 1509 struct vnode *vp; 1510 1511 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1512 &vp, l)) != 0) 1513 return (error); 1514 cwdi = p->p_cwdi; 1515 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1516 vrele(cwdi->cwdi_cdir); 1517 cwdi->cwdi_cdir = vp; 1518 rw_exit(&cwdi->cwdi_lock); 1519 return (0); 1520 } 1521 1522 /* 1523 * Change notion of root (``/'') directory. 1524 */ 1525 /* ARGSUSED */ 1526 int 1527 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1528 { 1529 /* { 1530 syscallarg(const char *) path; 1531 } */ 1532 int error; 1533 struct vnode *vp; 1534 1535 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1536 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1537 return (error); 1538 1539 error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, &vp, l); 1540 if (error == 0) 1541 change_root(vp); 1542 return error; 1543 } 1544 1545 /* 1546 * Common routine for chroot and fchroot. 1547 * NB: callers need to properly authorize the change root operation. 1548 */ 1549 void 1550 change_root(struct vnode *vp) 1551 { 1552 kauth_cred_t ncred; 1553 struct lwp *l = curlwp; 1554 struct proc *p = l->l_proc; 1555 struct cwdinfo *cwdi = p->p_cwdi; 1556 1557 ncred = kauth_cred_alloc(); 1558 1559 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1560 if (cwdi->cwdi_rdir != NULL) 1561 vrele(cwdi->cwdi_rdir); 1562 cwdi->cwdi_rdir = vp; 1563 1564 /* 1565 * Prevent escaping from chroot by putting the root under 1566 * the working directory. Silently chdir to / if we aren't 1567 * already there. 1568 */ 1569 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1570 /* 1571 * XXX would be more failsafe to change directory to a 1572 * deadfs node here instead 1573 */ 1574 vrele(cwdi->cwdi_cdir); 1575 vref(vp); 1576 cwdi->cwdi_cdir = vp; 1577 } 1578 rw_exit(&cwdi->cwdi_lock); 1579 1580 /* Get a write lock on the process credential. */ 1581 proc_crmod_enter(); 1582 1583 kauth_cred_clone(p->p_cred, ncred); 1584 kauth_proc_chroot(ncred, p->p_cwdi); 1585 1586 /* Broadcast our credentials to the process and other LWPs. */ 1587 proc_crmod_leave(ncred, p->p_cred, true); 1588 } 1589 1590 /* 1591 * Common routine for chroot and chdir. 1592 * XXX "where" should be enum uio_seg 1593 */ 1594 int 1595 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1596 { 1597 struct pathbuf *pb; 1598 struct nameidata nd; 1599 int error; 1600 1601 error = pathbuf_maybe_copyin(path, where, &pb); 1602 if (error) { 1603 return error; 1604 } 1605 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT, pb); 1606 if ((error = namei(&nd)) != 0) { 1607 pathbuf_destroy(pb); 1608 return error; 1609 } 1610 *vpp = nd.ni_vp; 1611 pathbuf_destroy(pb); 1612 1613 if ((*vpp)->v_type != VDIR) 1614 error = ENOTDIR; 1615 else 1616 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1617 1618 if (error) 1619 vput(*vpp); 1620 else 1621 VOP_UNLOCK(*vpp); 1622 return (error); 1623 } 1624 1625 /* 1626 * Internals of sys_open - path has already been converted into a pathbuf 1627 * (so we can easily reuse this function from other parts of the kernel, 1628 * like posix_spawn post-processing). 1629 */ 1630 int 1631 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1632 int open_mode, int *fd) 1633 { 1634 struct proc *p = l->l_proc; 1635 struct cwdinfo *cwdi = p->p_cwdi; 1636 file_t *fp; 1637 struct vnode *vp; 1638 int dupfd; 1639 bool dupfd_move; 1640 int flags, cmode; 1641 int indx, error; 1642 1643 if (open_flags & O_SEARCH) { 1644 open_flags &= ~(int)O_SEARCH; 1645 } 1646 1647 /* 1648 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1649 * may be specified. 1650 */ 1651 if ((open_flags & O_EXEC) && (open_flags & O_ACCMODE)) 1652 return EINVAL; 1653 1654 flags = FFLAGS(open_flags); 1655 if ((flags & (FREAD | FWRITE)) == 0) 1656 return EINVAL; 1657 1658 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1659 return error; 1660 } 1661 1662 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1663 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1664 1665 error = vn_open(dvp, pb, TRYEMULROOT, flags, cmode, 1666 &vp, &dupfd_move, &dupfd); 1667 if (error != 0) { 1668 fd_abort(p, fp, indx); 1669 if (error == ERESTART) 1670 error = EINTR; 1671 return error; 1672 } 1673 1674 if (vp == NULL) { 1675 fd_abort(p, fp, indx); 1676 error = fd_dupopen(dupfd, dupfd_move, flags, &indx); 1677 if (error) 1678 return error; 1679 *fd = indx; 1680 } else { 1681 error = open_setfp(l, fp, vp, indx, flags); 1682 if (error) 1683 return error; 1684 VOP_UNLOCK(vp); 1685 *fd = indx; 1686 fd_affix(p, fp, indx); 1687 } 1688 1689 return 0; 1690 } 1691 1692 int 1693 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1694 { 1695 struct pathbuf *pb; 1696 int error, oflags; 1697 1698 oflags = FFLAGS(open_flags); 1699 if ((oflags & (FREAD | FWRITE)) == 0) 1700 return EINVAL; 1701 1702 pb = pathbuf_create(path); 1703 if (pb == NULL) 1704 return ENOMEM; 1705 1706 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1707 pathbuf_destroy(pb); 1708 1709 return error; 1710 } 1711 1712 static int 1713 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1714 int mode, int *fd) 1715 { 1716 file_t *dfp = NULL; 1717 struct vnode *dvp = NULL; 1718 struct pathbuf *pb; 1719 const char *pathstring = NULL; 1720 int error; 1721 1722 if (path == NULL) { 1723 MODULE_HOOK_CALL(vfs_openat_10_hook, (&pb), enosys(), error); 1724 if (error == ENOSYS) 1725 goto no_compat; 1726 if (error) 1727 return error; 1728 } else { 1729 no_compat: 1730 error = pathbuf_copyin(path, &pb); 1731 if (error) 1732 return error; 1733 } 1734 1735 pathstring = pathbuf_stringcopy_get(pb); 1736 1737 /* 1738 * fdat is ignored if: 1739 * 1) if fdat is AT_FDCWD, which means use current directory as base. 1740 * 2) if path is absolute, then fdat is useless. 1741 */ 1742 if (fdat != AT_FDCWD && pathstring[0] != '/') { 1743 /* fd_getvnode() will use the descriptor for us */ 1744 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1745 goto out; 1746 1747 dvp = dfp->f_vnode; 1748 } 1749 1750 error = do_open(l, dvp, pb, flags, mode, fd); 1751 1752 if (dfp != NULL) 1753 fd_putfile(fdat); 1754 out: 1755 pathbuf_stringcopy_put(pb, pathstring); 1756 pathbuf_destroy(pb); 1757 return error; 1758 } 1759 1760 int 1761 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1762 { 1763 /* { 1764 syscallarg(const char *) path; 1765 syscallarg(int) flags; 1766 syscallarg(int) mode; 1767 } */ 1768 int error; 1769 int fd; 1770 1771 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1772 SCARG(uap, flags), SCARG(uap, mode), &fd); 1773 1774 if (error == 0) 1775 *retval = fd; 1776 1777 return error; 1778 } 1779 1780 int 1781 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1782 { 1783 /* { 1784 syscallarg(int) fd; 1785 syscallarg(const char *) path; 1786 syscallarg(int) oflags; 1787 syscallarg(int) mode; 1788 } */ 1789 int error; 1790 int fd; 1791 1792 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1793 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1794 1795 if (error == 0) 1796 *retval = fd; 1797 1798 return error; 1799 } 1800 1801 static void 1802 vfs__fhfree(fhandle_t *fhp) 1803 { 1804 size_t fhsize; 1805 1806 fhsize = FHANDLE_SIZE(fhp); 1807 kmem_free(fhp, fhsize); 1808 } 1809 1810 /* 1811 * vfs_composefh: compose a filehandle. 1812 */ 1813 1814 int 1815 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1816 { 1817 struct mount *mp; 1818 struct fid *fidp; 1819 int error; 1820 size_t needfhsize; 1821 size_t fidsize; 1822 1823 mp = vp->v_mount; 1824 fidp = NULL; 1825 if (*fh_size < FHANDLE_SIZE_MIN) { 1826 fidsize = 0; 1827 } else { 1828 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1829 if (fhp != NULL) { 1830 memset(fhp, 0, *fh_size); 1831 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1832 fidp = &fhp->fh_fid; 1833 } 1834 } 1835 error = VFS_VPTOFH(vp, fidp, &fidsize); 1836 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1837 if (error == 0 && *fh_size < needfhsize) { 1838 error = E2BIG; 1839 } 1840 *fh_size = needfhsize; 1841 return error; 1842 } 1843 1844 int 1845 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1846 { 1847 struct mount *mp; 1848 fhandle_t *fhp; 1849 size_t fhsize; 1850 size_t fidsize; 1851 int error; 1852 1853 mp = vp->v_mount; 1854 fidsize = 0; 1855 error = VFS_VPTOFH(vp, NULL, &fidsize); 1856 KASSERT(error != 0); 1857 if (error != E2BIG) { 1858 goto out; 1859 } 1860 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1861 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1862 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1863 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1864 if (error == 0) { 1865 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1866 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1867 *fhpp = fhp; 1868 } else { 1869 kmem_free(fhp, fhsize); 1870 } 1871 out: 1872 return error; 1873 } 1874 1875 void 1876 vfs_composefh_free(fhandle_t *fhp) 1877 { 1878 1879 vfs__fhfree(fhp); 1880 } 1881 1882 /* 1883 * vfs_fhtovp: lookup a vnode by a filehandle. 1884 */ 1885 1886 int 1887 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1888 { 1889 struct mount *mp; 1890 int error; 1891 1892 *vpp = NULL; 1893 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1894 if (mp == NULL) { 1895 error = ESTALE; 1896 goto out; 1897 } 1898 if (mp->mnt_op->vfs_fhtovp == NULL) { 1899 error = EOPNOTSUPP; 1900 goto out; 1901 } 1902 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), LK_EXCLUSIVE, vpp); 1903 out: 1904 return error; 1905 } 1906 1907 /* 1908 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1909 * the needed size. 1910 */ 1911 1912 int 1913 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1914 { 1915 fhandle_t *fhp; 1916 int error; 1917 1918 if (fhsize > FHANDLE_SIZE_MAX) { 1919 return EINVAL; 1920 } 1921 if (fhsize < FHANDLE_SIZE_MIN) { 1922 return EINVAL; 1923 } 1924 again: 1925 fhp = kmem_alloc(fhsize, KM_SLEEP); 1926 error = copyin(ufhp, fhp, fhsize); 1927 if (error == 0) { 1928 /* XXX this check shouldn't be here */ 1929 if (FHANDLE_SIZE(fhp) == fhsize) { 1930 *fhpp = fhp; 1931 return 0; 1932 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1933 /* 1934 * a kludge for nfsv2 padded handles. 1935 */ 1936 size_t sz; 1937 1938 sz = FHANDLE_SIZE(fhp); 1939 kmem_free(fhp, fhsize); 1940 fhsize = sz; 1941 goto again; 1942 } else { 1943 /* 1944 * userland told us wrong size. 1945 */ 1946 error = EINVAL; 1947 } 1948 } 1949 kmem_free(fhp, fhsize); 1950 return error; 1951 } 1952 1953 void 1954 vfs_copyinfh_free(fhandle_t *fhp) 1955 { 1956 1957 vfs__fhfree(fhp); 1958 } 1959 1960 /* 1961 * Get file handle system call 1962 */ 1963 int 1964 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1965 { 1966 /* { 1967 syscallarg(char *) fname; 1968 syscallarg(fhandle_t *) fhp; 1969 syscallarg(size_t *) fh_size; 1970 } */ 1971 struct vnode *vp; 1972 fhandle_t *fh; 1973 int error; 1974 struct pathbuf *pb; 1975 struct nameidata nd; 1976 size_t sz; 1977 size_t usz; 1978 1979 /* 1980 * Must be super user 1981 */ 1982 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1983 0, NULL, NULL, NULL); 1984 if (error) 1985 return (error); 1986 1987 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1988 if (error) { 1989 return error; 1990 } 1991 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1992 error = namei(&nd); 1993 if (error) { 1994 pathbuf_destroy(pb); 1995 return error; 1996 } 1997 vp = nd.ni_vp; 1998 pathbuf_destroy(pb); 1999 2000 error = vfs_composefh_alloc(vp, &fh); 2001 vput(vp); 2002 if (error != 0) { 2003 return error; 2004 } 2005 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 2006 if (error != 0) { 2007 goto out; 2008 } 2009 sz = FHANDLE_SIZE(fh); 2010 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 2011 if (error != 0) { 2012 goto out; 2013 } 2014 if (usz >= sz) { 2015 error = copyout(fh, SCARG(uap, fhp), sz); 2016 } else { 2017 error = E2BIG; 2018 } 2019 out: 2020 vfs_composefh_free(fh); 2021 return (error); 2022 } 2023 2024 /* 2025 * Open a file given a file handle. 2026 * 2027 * Check permissions, allocate an open file structure, 2028 * and call the device open routine if any. 2029 */ 2030 2031 int 2032 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 2033 register_t *retval) 2034 { 2035 file_t *fp; 2036 struct vnode *vp = NULL; 2037 kauth_cred_t cred = l->l_cred; 2038 file_t *nfp; 2039 int indx, error; 2040 struct vattr va; 2041 fhandle_t *fh; 2042 int flags; 2043 proc_t *p; 2044 2045 p = curproc; 2046 2047 /* 2048 * Must be super user 2049 */ 2050 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2051 0, NULL, NULL, NULL))) 2052 return (error); 2053 2054 if (oflags & O_SEARCH) { 2055 oflags &= ~(int)O_SEARCH; 2056 } 2057 2058 flags = FFLAGS(oflags); 2059 if ((flags & (FREAD | FWRITE)) == 0) 2060 return (EINVAL); 2061 if ((flags & O_CREAT)) 2062 return (EINVAL); 2063 if ((error = fd_allocfile(&nfp, &indx)) != 0) 2064 return (error); 2065 fp = nfp; 2066 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2067 if (error != 0) { 2068 goto bad; 2069 } 2070 error = vfs_fhtovp(fh, &vp); 2071 vfs_copyinfh_free(fh); 2072 if (error != 0) { 2073 goto bad; 2074 } 2075 2076 /* Now do an effective vn_open */ 2077 2078 if (vp->v_type == VSOCK) { 2079 error = EOPNOTSUPP; 2080 goto bad; 2081 } 2082 error = vn_openchk(vp, cred, flags); 2083 if (error != 0) 2084 goto bad; 2085 if (flags & O_TRUNC) { 2086 VOP_UNLOCK(vp); /* XXX */ 2087 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2088 vattr_null(&va); 2089 va.va_size = 0; 2090 error = VOP_SETATTR(vp, &va, cred); 2091 if (error) 2092 goto bad; 2093 } 2094 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2095 goto bad; 2096 if (flags & FWRITE) { 2097 mutex_enter(vp->v_interlock); 2098 vp->v_writecount++; 2099 mutex_exit(vp->v_interlock); 2100 } 2101 2102 /* done with modified vn_open, now finish what sys_open does. */ 2103 if ((error = open_setfp(l, fp, vp, indx, flags))) 2104 return error; 2105 2106 VOP_UNLOCK(vp); 2107 *retval = indx; 2108 fd_affix(p, fp, indx); 2109 return (0); 2110 2111 bad: 2112 fd_abort(p, fp, indx); 2113 if (vp != NULL) 2114 vput(vp); 2115 if (error == EDUPFD || error == EMOVEFD) { 2116 /* XXX should probably close curlwp->l_dupfd */ 2117 error = EOPNOTSUPP; 2118 } 2119 return (error); 2120 } 2121 2122 int 2123 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2124 { 2125 /* { 2126 syscallarg(const void *) fhp; 2127 syscallarg(size_t) fh_size; 2128 syscallarg(int) flags; 2129 } */ 2130 2131 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2132 SCARG(uap, flags), retval); 2133 } 2134 2135 int 2136 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2137 { 2138 int error; 2139 fhandle_t *fh; 2140 struct vnode *vp; 2141 2142 /* 2143 * Must be super user 2144 */ 2145 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2146 0, NULL, NULL, NULL))) 2147 return (error); 2148 2149 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2150 if (error != 0) 2151 return error; 2152 2153 error = vfs_fhtovp(fh, &vp); 2154 vfs_copyinfh_free(fh); 2155 if (error != 0) 2156 return error; 2157 2158 error = vn_stat(vp, sb); 2159 vput(vp); 2160 return error; 2161 } 2162 2163 2164 /* ARGSUSED */ 2165 int 2166 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2167 { 2168 /* { 2169 syscallarg(const void *) fhp; 2170 syscallarg(size_t) fh_size; 2171 syscallarg(struct stat *) sb; 2172 } */ 2173 struct stat sb; 2174 int error; 2175 2176 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2177 if (error) 2178 return error; 2179 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2180 } 2181 2182 int 2183 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2184 int flags) 2185 { 2186 fhandle_t *fh; 2187 struct mount *mp; 2188 struct vnode *vp; 2189 int error; 2190 2191 /* 2192 * Must be super user 2193 */ 2194 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2195 0, NULL, NULL, NULL))) 2196 return error; 2197 2198 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2199 if (error != 0) 2200 return error; 2201 2202 error = vfs_fhtovp(fh, &vp); 2203 vfs_copyinfh_free(fh); 2204 if (error != 0) 2205 return error; 2206 2207 mp = vp->v_mount; 2208 error = dostatvfs(mp, sb, l, flags, 1); 2209 vput(vp); 2210 return error; 2211 } 2212 2213 /* ARGSUSED */ 2214 int 2215 sys___fhstatvfs190(struct lwp *l, const struct sys___fhstatvfs190_args *uap, register_t *retval) 2216 { 2217 /* { 2218 syscallarg(const void *) fhp; 2219 syscallarg(size_t) fh_size; 2220 syscallarg(struct statvfs *) buf; 2221 syscallarg(int) flags; 2222 } */ 2223 struct statvfs *sb = STATVFSBUF_GET(); 2224 int error; 2225 2226 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2227 SCARG(uap, flags)); 2228 if (error == 0) 2229 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2230 STATVFSBUF_PUT(sb); 2231 return error; 2232 } 2233 2234 int 2235 do_posix_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2236 dev_t dev) 2237 { 2238 2239 /* 2240 * The POSIX mknod(2) call is an alias for mkfifo(2) for S_IFIFO 2241 * in mode and dev=0. 2242 * 2243 * In all the other cases it's implementation defined behavior. 2244 */ 2245 2246 if ((mode & S_IFIFO) && dev == 0) 2247 return do_sys_mkfifoat(l, fdat, pathname, mode); 2248 else 2249 return do_sys_mknodat(l, fdat, pathname, mode, dev, 2250 UIO_USERSPACE); 2251 } 2252 2253 /* 2254 * Create a special file. 2255 */ 2256 /* ARGSUSED */ 2257 int 2258 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2259 register_t *retval) 2260 { 2261 /* { 2262 syscallarg(const char *) path; 2263 syscallarg(mode_t) mode; 2264 syscallarg(dev_t) dev; 2265 } */ 2266 return do_posix_mknodat(l, AT_FDCWD, SCARG(uap, path), 2267 SCARG(uap, mode), SCARG(uap, dev)); 2268 } 2269 2270 int 2271 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2272 register_t *retval) 2273 { 2274 /* { 2275 syscallarg(int) fd; 2276 syscallarg(const char *) path; 2277 syscallarg(mode_t) mode; 2278 syscallarg(int) pad; 2279 syscallarg(dev_t) dev; 2280 } */ 2281 2282 return do_posix_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2283 SCARG(uap, mode), SCARG(uap, dev)); 2284 } 2285 2286 int 2287 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2288 enum uio_seg seg) 2289 { 2290 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, seg); 2291 } 2292 2293 int 2294 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2295 dev_t dev, enum uio_seg seg) 2296 { 2297 struct proc *p = l->l_proc; 2298 struct vnode *vp; 2299 struct vattr vattr; 2300 int error, optype; 2301 struct pathbuf *pb; 2302 struct nameidata nd; 2303 const char *pathstring; 2304 2305 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2306 0, NULL, NULL, NULL)) != 0) 2307 return (error); 2308 2309 optype = VOP_MKNOD_DESCOFFSET; 2310 2311 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2312 if (error) { 2313 return error; 2314 } 2315 pathstring = pathbuf_stringcopy_get(pb); 2316 if (pathstring == NULL) { 2317 pathbuf_destroy(pb); 2318 return ENOMEM; 2319 } 2320 2321 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2322 2323 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2324 goto out; 2325 vp = nd.ni_vp; 2326 2327 if (vp != NULL) 2328 error = EEXIST; 2329 else { 2330 vattr_null(&vattr); 2331 /* We will read cwdi->cwdi_cmask unlocked. */ 2332 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2333 vattr.va_rdev = dev; 2334 2335 switch (mode & S_IFMT) { 2336 case S_IFMT: /* used by badsect to flag bad sectors */ 2337 vattr.va_type = VBAD; 2338 break; 2339 case S_IFCHR: 2340 vattr.va_type = VCHR; 2341 break; 2342 case S_IFBLK: 2343 vattr.va_type = VBLK; 2344 break; 2345 case S_IFWHT: 2346 optype = VOP_WHITEOUT_DESCOFFSET; 2347 break; 2348 case S_IFREG: 2349 #if NVERIEXEC > 0 2350 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2351 O_CREAT); 2352 #endif /* NVERIEXEC > 0 */ 2353 vattr.va_type = VREG; 2354 vattr.va_rdev = VNOVAL; 2355 optype = VOP_CREATE_DESCOFFSET; 2356 break; 2357 default: 2358 error = EINVAL; 2359 break; 2360 } 2361 2362 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET && 2363 vattr.va_rdev == VNOVAL) 2364 error = EINVAL; 2365 } 2366 2367 if (!error) { 2368 switch (optype) { 2369 case VOP_WHITEOUT_DESCOFFSET: 2370 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2371 if (error) 2372 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2373 vput(nd.ni_dvp); 2374 break; 2375 2376 case VOP_MKNOD_DESCOFFSET: 2377 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2378 &nd.ni_cnd, &vattr); 2379 if (error == 0) 2380 vrele(nd.ni_vp); 2381 vput(nd.ni_dvp); 2382 break; 2383 2384 case VOP_CREATE_DESCOFFSET: 2385 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2386 &nd.ni_cnd, &vattr); 2387 if (error == 0) 2388 vrele(nd.ni_vp); 2389 vput(nd.ni_dvp); 2390 break; 2391 } 2392 } else { 2393 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2394 if (nd.ni_dvp == vp) 2395 vrele(nd.ni_dvp); 2396 else 2397 vput(nd.ni_dvp); 2398 if (vp) 2399 vrele(vp); 2400 } 2401 out: 2402 pathbuf_stringcopy_put(pb, pathstring); 2403 pathbuf_destroy(pb); 2404 return (error); 2405 } 2406 2407 /* 2408 * Create a named pipe. 2409 */ 2410 /* ARGSUSED */ 2411 int 2412 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2413 { 2414 /* { 2415 syscallarg(const char *) path; 2416 syscallarg(int) mode; 2417 } */ 2418 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2419 } 2420 2421 int 2422 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2423 register_t *retval) 2424 { 2425 /* { 2426 syscallarg(int) fd; 2427 syscallarg(const char *) path; 2428 syscallarg(int) mode; 2429 } */ 2430 2431 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2432 SCARG(uap, mode)); 2433 } 2434 2435 static int 2436 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2437 { 2438 struct proc *p = l->l_proc; 2439 struct vattr vattr; 2440 int error; 2441 struct pathbuf *pb; 2442 struct nameidata nd; 2443 2444 error = pathbuf_copyin(path, &pb); 2445 if (error) { 2446 return error; 2447 } 2448 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2449 2450 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2451 pathbuf_destroy(pb); 2452 return error; 2453 } 2454 if (nd.ni_vp != NULL) { 2455 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2456 if (nd.ni_dvp == nd.ni_vp) 2457 vrele(nd.ni_dvp); 2458 else 2459 vput(nd.ni_dvp); 2460 vrele(nd.ni_vp); 2461 pathbuf_destroy(pb); 2462 return (EEXIST); 2463 } 2464 vattr_null(&vattr); 2465 vattr.va_type = VFIFO; 2466 /* We will read cwdi->cwdi_cmask unlocked. */ 2467 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2468 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2469 if (error == 0) 2470 vrele(nd.ni_vp); 2471 vput(nd.ni_dvp); 2472 pathbuf_destroy(pb); 2473 return (error); 2474 } 2475 2476 /* 2477 * Make a hard file link. 2478 */ 2479 /* ARGSUSED */ 2480 int 2481 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2482 const char *link, int follow, register_t *retval) 2483 { 2484 struct vnode *vp; 2485 struct pathbuf *linkpb; 2486 struct nameidata nd; 2487 namei_simple_flags_t ns_flags; 2488 int error; 2489 2490 if (follow & AT_SYMLINK_FOLLOW) 2491 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2492 else 2493 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2494 2495 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2496 if (error != 0) 2497 return (error); 2498 error = pathbuf_copyin(link, &linkpb); 2499 if (error) { 2500 goto out1; 2501 } 2502 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2503 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2504 goto out2; 2505 if (nd.ni_vp) { 2506 error = EEXIST; 2507 goto abortop; 2508 } 2509 /* Prevent hard links on directories. */ 2510 if (vp->v_type == VDIR) { 2511 error = EPERM; 2512 goto abortop; 2513 } 2514 /* Prevent cross-mount operation. */ 2515 if (nd.ni_dvp->v_mount != vp->v_mount) { 2516 error = EXDEV; 2517 goto abortop; 2518 } 2519 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2520 VOP_UNLOCK(nd.ni_dvp); 2521 vrele(nd.ni_dvp); 2522 out2: 2523 pathbuf_destroy(linkpb); 2524 out1: 2525 vrele(vp); 2526 return (error); 2527 abortop: 2528 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2529 if (nd.ni_dvp == nd.ni_vp) 2530 vrele(nd.ni_dvp); 2531 else 2532 vput(nd.ni_dvp); 2533 if (nd.ni_vp != NULL) 2534 vrele(nd.ni_vp); 2535 goto out2; 2536 } 2537 2538 int 2539 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2540 { 2541 /* { 2542 syscallarg(const char *) path; 2543 syscallarg(const char *) link; 2544 } */ 2545 const char *path = SCARG(uap, path); 2546 const char *link = SCARG(uap, link); 2547 2548 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2549 AT_SYMLINK_FOLLOW, retval); 2550 } 2551 2552 int 2553 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2554 register_t *retval) 2555 { 2556 /* { 2557 syscallarg(int) fd1; 2558 syscallarg(const char *) name1; 2559 syscallarg(int) fd2; 2560 syscallarg(const char *) name2; 2561 syscallarg(int) flags; 2562 } */ 2563 int fd1 = SCARG(uap, fd1); 2564 const char *name1 = SCARG(uap, name1); 2565 int fd2 = SCARG(uap, fd2); 2566 const char *name2 = SCARG(uap, name2); 2567 int follow; 2568 2569 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2570 2571 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2572 } 2573 2574 2575 int 2576 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2577 { 2578 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2579 } 2580 2581 static int 2582 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2583 const char *link, enum uio_seg seg) 2584 { 2585 struct proc *p = curproc; 2586 struct vattr vattr; 2587 char *path; 2588 int error; 2589 size_t len; 2590 struct pathbuf *linkpb; 2591 struct nameidata nd; 2592 2593 KASSERT(l != NULL || fdat == AT_FDCWD); 2594 2595 path = PNBUF_GET(); 2596 if (seg == UIO_USERSPACE) { 2597 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2598 goto out1; 2599 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2600 goto out1; 2601 } else { 2602 len = strlen(patharg) + 1; 2603 KASSERT(len <= MAXPATHLEN); 2604 memcpy(path, patharg, len); 2605 linkpb = pathbuf_create(link); 2606 if (linkpb == NULL) { 2607 error = ENOMEM; 2608 goto out1; 2609 } 2610 } 2611 ktrkuser("symlink-target", path, len - 1); 2612 2613 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2614 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2615 goto out2; 2616 if (nd.ni_vp) { 2617 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2618 if (nd.ni_dvp == nd.ni_vp) 2619 vrele(nd.ni_dvp); 2620 else 2621 vput(nd.ni_dvp); 2622 vrele(nd.ni_vp); 2623 error = EEXIST; 2624 goto out2; 2625 } 2626 vattr_null(&vattr); 2627 vattr.va_type = VLNK; 2628 /* We will read cwdi->cwdi_cmask unlocked. */ 2629 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2630 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2631 if (error == 0) 2632 vrele(nd.ni_vp); 2633 vput(nd.ni_dvp); 2634 out2: 2635 pathbuf_destroy(linkpb); 2636 out1: 2637 PNBUF_PUT(path); 2638 return (error); 2639 } 2640 2641 /* 2642 * Make a symbolic link. 2643 */ 2644 /* ARGSUSED */ 2645 int 2646 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2647 { 2648 /* { 2649 syscallarg(const char *) path; 2650 syscallarg(const char *) link; 2651 } */ 2652 2653 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2654 UIO_USERSPACE); 2655 } 2656 2657 int 2658 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2659 register_t *retval) 2660 { 2661 /* { 2662 syscallarg(const char *) path1; 2663 syscallarg(int) fd; 2664 syscallarg(const char *) path2; 2665 } */ 2666 2667 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2668 SCARG(uap, path2), UIO_USERSPACE); 2669 } 2670 2671 /* 2672 * Delete a whiteout from the filesystem. 2673 */ 2674 /* ARGSUSED */ 2675 int 2676 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2677 { 2678 /* { 2679 syscallarg(const char *) path; 2680 } */ 2681 int error; 2682 struct pathbuf *pb; 2683 struct nameidata nd; 2684 2685 error = pathbuf_copyin(SCARG(uap, path), &pb); 2686 if (error) { 2687 return error; 2688 } 2689 2690 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2691 error = namei(&nd); 2692 if (error) { 2693 pathbuf_destroy(pb); 2694 return (error); 2695 } 2696 2697 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2698 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2699 if (nd.ni_dvp == nd.ni_vp) 2700 vrele(nd.ni_dvp); 2701 else 2702 vput(nd.ni_dvp); 2703 if (nd.ni_vp) 2704 vrele(nd.ni_vp); 2705 pathbuf_destroy(pb); 2706 return (EEXIST); 2707 } 2708 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2709 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2710 vput(nd.ni_dvp); 2711 pathbuf_destroy(pb); 2712 return (error); 2713 } 2714 2715 /* 2716 * Delete a name from the filesystem. 2717 */ 2718 /* ARGSUSED */ 2719 int 2720 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2721 { 2722 /* { 2723 syscallarg(const char *) path; 2724 } */ 2725 2726 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2727 } 2728 2729 int 2730 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2731 register_t *retval) 2732 { 2733 /* { 2734 syscallarg(int) fd; 2735 syscallarg(const char *) path; 2736 syscallarg(int) flag; 2737 } */ 2738 2739 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2740 SCARG(uap, flag), UIO_USERSPACE); 2741 } 2742 2743 int 2744 do_sys_unlink(const char *arg, enum uio_seg seg) 2745 { 2746 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2747 } 2748 2749 static int 2750 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2751 enum uio_seg seg) 2752 { 2753 struct vnode *vp; 2754 int error; 2755 struct pathbuf *pb; 2756 struct nameidata nd; 2757 const char *pathstring; 2758 2759 KASSERT(l != NULL || fdat == AT_FDCWD); 2760 2761 error = pathbuf_maybe_copyin(arg, seg, &pb); 2762 if (error) { 2763 return error; 2764 } 2765 pathstring = pathbuf_stringcopy_get(pb); 2766 if (pathstring == NULL) { 2767 pathbuf_destroy(pb); 2768 return ENOMEM; 2769 } 2770 2771 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2772 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2773 goto out; 2774 vp = nd.ni_vp; 2775 2776 /* 2777 * The root of a mounted filesystem cannot be deleted. 2778 */ 2779 if ((vp->v_vflag & VV_ROOT) != 0) { 2780 error = EBUSY; 2781 goto abort; 2782 } 2783 2784 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2785 error = EBUSY; 2786 goto abort; 2787 } 2788 2789 /* 2790 * No rmdir "." please. 2791 */ 2792 if (nd.ni_dvp == vp) { 2793 error = EINVAL; 2794 goto abort; 2795 } 2796 2797 /* 2798 * AT_REMOVEDIR is required to remove a directory 2799 */ 2800 if (vp->v_type == VDIR) { 2801 if (!(flags & AT_REMOVEDIR)) { 2802 error = EPERM; 2803 goto abort; 2804 } else { 2805 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2806 vput(nd.ni_dvp); 2807 goto out; 2808 } 2809 } 2810 2811 /* 2812 * Starting here we only deal with non directories. 2813 */ 2814 if (flags & AT_REMOVEDIR) { 2815 error = ENOTDIR; 2816 goto abort; 2817 } 2818 2819 #if NVERIEXEC > 0 2820 /* Handle remove requests for veriexec entries. */ 2821 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2822 goto abort; 2823 } 2824 #endif /* NVERIEXEC > 0 */ 2825 2826 #ifdef FILEASSOC 2827 (void)fileassoc_file_delete(vp); 2828 #endif /* FILEASSOC */ 2829 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2830 vput(nd.ni_dvp); 2831 goto out; 2832 2833 abort: 2834 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2835 if (nd.ni_dvp == vp) 2836 vrele(nd.ni_dvp); 2837 else 2838 vput(nd.ni_dvp); 2839 vput(vp); 2840 2841 out: 2842 pathbuf_stringcopy_put(pb, pathstring); 2843 pathbuf_destroy(pb); 2844 return (error); 2845 } 2846 2847 /* 2848 * Reposition read/write file offset. 2849 */ 2850 int 2851 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2852 { 2853 /* { 2854 syscallarg(int) fd; 2855 syscallarg(int) pad; 2856 syscallarg(off_t) offset; 2857 syscallarg(int) whence; 2858 } */ 2859 file_t *fp; 2860 int error, fd; 2861 2862 switch (SCARG(uap, whence)) { 2863 case SEEK_CUR: 2864 case SEEK_END: 2865 case SEEK_SET: 2866 break; 2867 default: 2868 return EINVAL; 2869 } 2870 2871 fd = SCARG(uap, fd); 2872 2873 if ((fp = fd_getfile(fd)) == NULL) 2874 return (EBADF); 2875 2876 if (fp->f_ops->fo_seek == NULL) { 2877 error = ESPIPE; 2878 goto out; 2879 } 2880 2881 error = (*fp->f_ops->fo_seek)(fp, SCARG(uap, offset), 2882 SCARG(uap, whence), (off_t *)retval, FOF_UPDATE_OFFSET); 2883 out: 2884 fd_putfile(fd); 2885 return (error); 2886 } 2887 2888 /* 2889 * Positional read system call. 2890 */ 2891 int 2892 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2893 { 2894 /* { 2895 syscallarg(int) fd; 2896 syscallarg(void *) buf; 2897 syscallarg(size_t) nbyte; 2898 syscallarg(off_t) offset; 2899 } */ 2900 file_t *fp; 2901 off_t offset; 2902 int error, fd = SCARG(uap, fd); 2903 2904 if ((fp = fd_getfile(fd)) == NULL) 2905 return (EBADF); 2906 2907 if ((fp->f_flag & FREAD) == 0) { 2908 fd_putfile(fd); 2909 return (EBADF); 2910 } 2911 2912 if (fp->f_ops->fo_seek == NULL) { 2913 error = ESPIPE; 2914 goto out; 2915 } 2916 2917 offset = SCARG(uap, offset); 2918 error = (*fp->f_ops->fo_seek)(fp, offset, SEEK_SET, &offset, 0); 2919 if (error) 2920 goto out; 2921 2922 /* dofileread() will unuse the descriptor for us */ 2923 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2924 &offset, 0, retval)); 2925 2926 out: 2927 fd_putfile(fd); 2928 return (error); 2929 } 2930 2931 /* 2932 * Positional scatter read system call. 2933 */ 2934 int 2935 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2936 { 2937 /* { 2938 syscallarg(int) fd; 2939 syscallarg(const struct iovec *) iovp; 2940 syscallarg(int) iovcnt; 2941 syscallarg(off_t) offset; 2942 } */ 2943 off_t offset = SCARG(uap, offset); 2944 2945 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2946 SCARG(uap, iovcnt), &offset, 0, retval); 2947 } 2948 2949 /* 2950 * Positional write system call. 2951 */ 2952 int 2953 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2954 { 2955 /* { 2956 syscallarg(int) fd; 2957 syscallarg(const void *) buf; 2958 syscallarg(size_t) nbyte; 2959 syscallarg(off_t) offset; 2960 } */ 2961 file_t *fp; 2962 off_t offset; 2963 int error, fd = SCARG(uap, fd); 2964 2965 if ((fp = fd_getfile(fd)) == NULL) 2966 return (EBADF); 2967 2968 if ((fp->f_flag & FWRITE) == 0) { 2969 fd_putfile(fd); 2970 return (EBADF); 2971 } 2972 2973 if (fp->f_ops->fo_seek == NULL) { 2974 error = ESPIPE; 2975 goto out; 2976 } 2977 2978 offset = SCARG(uap, offset); 2979 error = (*fp->f_ops->fo_seek)(fp, offset, SEEK_SET, &offset, 0); 2980 if (error) 2981 goto out; 2982 2983 /* dofilewrite() will unuse the descriptor for us */ 2984 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2985 &offset, 0, retval)); 2986 2987 out: 2988 fd_putfile(fd); 2989 return (error); 2990 } 2991 2992 /* 2993 * Positional gather write system call. 2994 */ 2995 int 2996 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2997 { 2998 /* { 2999 syscallarg(int) fd; 3000 syscallarg(const struct iovec *) iovp; 3001 syscallarg(int) iovcnt; 3002 syscallarg(off_t) offset; 3003 } */ 3004 off_t offset = SCARG(uap, offset); 3005 3006 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 3007 SCARG(uap, iovcnt), &offset, 0, retval); 3008 } 3009 3010 /* 3011 * Check access permissions. 3012 */ 3013 int 3014 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 3015 { 3016 /* { 3017 syscallarg(const char *) path; 3018 syscallarg(int) flags; 3019 } */ 3020 3021 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 3022 SCARG(uap, flags), 0); 3023 } 3024 3025 int 3026 do_sys_accessat(struct lwp *l, int fdat, const char *path, 3027 int mode, int flags) 3028 { 3029 kauth_cred_t cred; 3030 struct vnode *vp; 3031 int error, nd_flag, vmode; 3032 struct pathbuf *pb; 3033 struct nameidata nd; 3034 3035 CTASSERT(F_OK == 0); 3036 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 3037 /* nonsense mode */ 3038 return EINVAL; 3039 } 3040 3041 nd_flag = FOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT; 3042 if (flags & AT_SYMLINK_NOFOLLOW) 3043 nd_flag &= ~FOLLOW; 3044 3045 error = pathbuf_copyin(path, &pb); 3046 if (error) 3047 return error; 3048 3049 NDINIT(&nd, LOOKUP, nd_flag, pb); 3050 3051 /* Override default credentials */ 3052 cred = kauth_cred_dup(l->l_cred); 3053 if (!(flags & AT_EACCESS)) { 3054 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 3055 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 3056 } 3057 nd.ni_cnd.cn_cred = cred; 3058 3059 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3060 pathbuf_destroy(pb); 3061 goto out; 3062 } 3063 vp = nd.ni_vp; 3064 pathbuf_destroy(pb); 3065 3066 /* Flags == 0 means only check for existence. */ 3067 if (mode) { 3068 vmode = 0; 3069 if (mode & R_OK) 3070 vmode |= VREAD; 3071 if (mode & W_OK) 3072 vmode |= VWRITE; 3073 if (mode & X_OK) 3074 vmode |= VEXEC; 3075 3076 error = VOP_ACCESS(vp, vmode, cred); 3077 if (!error && (vmode & VWRITE)) 3078 error = vn_writechk(vp); 3079 } 3080 vput(vp); 3081 out: 3082 kauth_cred_free(cred); 3083 return (error); 3084 } 3085 3086 int 3087 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3088 register_t *retval) 3089 { 3090 /* { 3091 syscallarg(int) fd; 3092 syscallarg(const char *) path; 3093 syscallarg(int) amode; 3094 syscallarg(int) flag; 3095 } */ 3096 3097 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3098 SCARG(uap, amode), SCARG(uap, flag)); 3099 } 3100 3101 /* 3102 * Common code for all sys_stat functions, including compat versions. 3103 */ 3104 int 3105 do_sys_stat(const char *userpath, unsigned int nd_flag, 3106 struct stat *sb) 3107 { 3108 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3109 } 3110 3111 int 3112 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3113 unsigned int nd_flag, struct stat *sb) 3114 { 3115 int error; 3116 struct pathbuf *pb; 3117 struct nameidata nd; 3118 3119 KASSERT(l != NULL || fdat == AT_FDCWD); 3120 3121 error = pathbuf_copyin(userpath, &pb); 3122 if (error) { 3123 return error; 3124 } 3125 3126 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3127 3128 error = fd_nameiat(l, fdat, &nd); 3129 if (error != 0) { 3130 pathbuf_destroy(pb); 3131 return error; 3132 } 3133 error = vn_stat(nd.ni_vp, sb); 3134 vput(nd.ni_vp); 3135 pathbuf_destroy(pb); 3136 return error; 3137 } 3138 3139 /* 3140 * Get file status; this version follows links. 3141 */ 3142 /* ARGSUSED */ 3143 int 3144 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3145 { 3146 /* { 3147 syscallarg(const char *) path; 3148 syscallarg(struct stat *) ub; 3149 } */ 3150 struct stat sb; 3151 int error; 3152 3153 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3154 if (error) 3155 return error; 3156 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3157 } 3158 3159 /* 3160 * Get file status; this version does not follow links. 3161 */ 3162 /* ARGSUSED */ 3163 int 3164 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3165 { 3166 /* { 3167 syscallarg(const char *) path; 3168 syscallarg(struct stat *) ub; 3169 } */ 3170 struct stat sb; 3171 int error; 3172 3173 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3174 if (error) 3175 return error; 3176 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3177 } 3178 3179 int 3180 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3181 register_t *retval) 3182 { 3183 /* { 3184 syscallarg(int) fd; 3185 syscallarg(const char *) path; 3186 syscallarg(struct stat *) buf; 3187 syscallarg(int) flag; 3188 } */ 3189 unsigned int nd_flag; 3190 struct stat sb; 3191 int error; 3192 3193 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3194 nd_flag = NOFOLLOW; 3195 else 3196 nd_flag = FOLLOW; 3197 3198 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3199 &sb); 3200 if (error) 3201 return error; 3202 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3203 } 3204 3205 static int 3206 kern_pathconf(register_t *retval, const char *path, int name, int flag) 3207 { 3208 int error; 3209 struct pathbuf *pb; 3210 struct nameidata nd; 3211 3212 error = pathbuf_copyin(path, &pb); 3213 if (error) { 3214 return error; 3215 } 3216 NDINIT(&nd, LOOKUP, flag | LOCKLEAF | TRYEMULROOT, pb); 3217 if ((error = namei(&nd)) != 0) { 3218 pathbuf_destroy(pb); 3219 return error; 3220 } 3221 error = VOP_PATHCONF(nd.ni_vp, name, retval); 3222 vput(nd.ni_vp); 3223 pathbuf_destroy(pb); 3224 return error; 3225 } 3226 3227 /* 3228 * Get configurable pathname variables. 3229 */ 3230 /* ARGSUSED */ 3231 int 3232 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, 3233 register_t *retval) 3234 { 3235 /* { 3236 syscallarg(const char *) path; 3237 syscallarg(int) name; 3238 } */ 3239 return kern_pathconf(retval, SCARG(uap, path), SCARG(uap, name), 3240 FOLLOW); 3241 } 3242 3243 /* ARGSUSED */ 3244 int 3245 sys_lpathconf(struct lwp *l, const struct sys_lpathconf_args *uap, 3246 register_t *retval) 3247 { 3248 /* { 3249 syscallarg(const char *) path; 3250 syscallarg(int) name; 3251 } */ 3252 return kern_pathconf(retval, SCARG(uap, path), SCARG(uap, name), 3253 NOFOLLOW); 3254 } 3255 3256 /* 3257 * Return target name of a symbolic link. 3258 */ 3259 /* ARGSUSED */ 3260 int 3261 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3262 register_t *retval) 3263 { 3264 /* { 3265 syscallarg(const char *) path; 3266 syscallarg(char *) buf; 3267 syscallarg(size_t) count; 3268 } */ 3269 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3270 SCARG(uap, buf), SCARG(uap, count), retval); 3271 } 3272 3273 static int 3274 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3275 size_t count, register_t *retval) 3276 { 3277 struct vnode *vp; 3278 struct iovec aiov; 3279 struct uio auio; 3280 int error; 3281 struct pathbuf *pb; 3282 struct nameidata nd; 3283 3284 error = pathbuf_copyin(path, &pb); 3285 if (error) { 3286 return error; 3287 } 3288 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT, pb); 3289 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3290 pathbuf_destroy(pb); 3291 return error; 3292 } 3293 vp = nd.ni_vp; 3294 pathbuf_destroy(pb); 3295 if (vp->v_type != VLNK) 3296 error = EINVAL; 3297 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3298 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3299 aiov.iov_base = buf; 3300 aiov.iov_len = count; 3301 auio.uio_iov = &aiov; 3302 auio.uio_iovcnt = 1; 3303 auio.uio_offset = 0; 3304 auio.uio_rw = UIO_READ; 3305 KASSERT(l == curlwp); 3306 auio.uio_vmspace = l->l_proc->p_vmspace; 3307 auio.uio_resid = count; 3308 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3309 *retval = count - auio.uio_resid; 3310 } 3311 vput(vp); 3312 return (error); 3313 } 3314 3315 int 3316 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3317 register_t *retval) 3318 { 3319 /* { 3320 syscallarg(int) fd; 3321 syscallarg(const char *) path; 3322 syscallarg(char *) buf; 3323 syscallarg(size_t) bufsize; 3324 } */ 3325 3326 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3327 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3328 } 3329 3330 /* 3331 * Change flags of a file given a path name. 3332 */ 3333 /* ARGSUSED */ 3334 int 3335 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3336 { 3337 /* { 3338 syscallarg(const char *) path; 3339 syscallarg(u_long) flags; 3340 } */ 3341 struct vnode *vp; 3342 int error; 3343 3344 error = namei_simple_user(SCARG(uap, path), 3345 NSM_FOLLOW_TRYEMULROOT, &vp); 3346 if (error != 0) 3347 return (error); 3348 error = change_flags(vp, SCARG(uap, flags), l); 3349 vput(vp); 3350 return (error); 3351 } 3352 3353 /* 3354 * Change flags of a file given a file descriptor. 3355 */ 3356 /* ARGSUSED */ 3357 int 3358 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3359 { 3360 /* { 3361 syscallarg(int) fd; 3362 syscallarg(u_long) flags; 3363 } */ 3364 struct vnode *vp; 3365 file_t *fp; 3366 int error; 3367 3368 /* fd_getvnode() will use the descriptor for us */ 3369 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3370 return (error); 3371 vp = fp->f_vnode; 3372 error = change_flags(vp, SCARG(uap, flags), l); 3373 VOP_UNLOCK(vp); 3374 fd_putfile(SCARG(uap, fd)); 3375 return (error); 3376 } 3377 3378 /* 3379 * Change flags of a file given a path name; this version does 3380 * not follow links. 3381 */ 3382 int 3383 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3384 { 3385 /* { 3386 syscallarg(const char *) path; 3387 syscallarg(u_long) flags; 3388 } */ 3389 struct vnode *vp; 3390 int error; 3391 3392 error = namei_simple_user(SCARG(uap, path), 3393 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3394 if (error != 0) 3395 return (error); 3396 error = change_flags(vp, SCARG(uap, flags), l); 3397 vput(vp); 3398 return (error); 3399 } 3400 3401 /* 3402 * Common routine to change flags of a file. 3403 */ 3404 int 3405 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3406 { 3407 struct vattr vattr; 3408 int error; 3409 3410 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3411 3412 vattr_null(&vattr); 3413 vattr.va_flags = flags; 3414 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3415 3416 return (error); 3417 } 3418 3419 /* 3420 * Change mode of a file given path name; this version follows links. 3421 */ 3422 /* ARGSUSED */ 3423 int 3424 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3425 { 3426 /* { 3427 syscallarg(const char *) path; 3428 syscallarg(int) mode; 3429 } */ 3430 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3431 SCARG(uap, mode), 0); 3432 } 3433 3434 int 3435 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3436 { 3437 int error; 3438 struct vnode *vp; 3439 namei_simple_flags_t ns_flag; 3440 3441 if (flags & AT_SYMLINK_NOFOLLOW) 3442 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3443 else 3444 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3445 3446 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3447 if (error != 0) 3448 return error; 3449 3450 error = change_mode(vp, mode, l); 3451 3452 vrele(vp); 3453 3454 return (error); 3455 } 3456 3457 /* 3458 * Change mode of a file given a file descriptor. 3459 */ 3460 /* ARGSUSED */ 3461 int 3462 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3463 { 3464 /* { 3465 syscallarg(int) fd; 3466 syscallarg(int) mode; 3467 } */ 3468 file_t *fp; 3469 int error; 3470 3471 /* fd_getvnode() will use the descriptor for us */ 3472 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3473 return (error); 3474 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3475 fd_putfile(SCARG(uap, fd)); 3476 return (error); 3477 } 3478 3479 int 3480 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3481 register_t *retval) 3482 { 3483 /* { 3484 syscallarg(int) fd; 3485 syscallarg(const char *) path; 3486 syscallarg(int) mode; 3487 syscallarg(int) flag; 3488 } */ 3489 3490 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3491 SCARG(uap, mode), SCARG(uap, flag)); 3492 } 3493 3494 /* 3495 * Change mode of a file given path name; this version does not follow links. 3496 */ 3497 /* ARGSUSED */ 3498 int 3499 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3500 { 3501 /* { 3502 syscallarg(const char *) path; 3503 syscallarg(int) mode; 3504 } */ 3505 int error; 3506 struct vnode *vp; 3507 3508 error = namei_simple_user(SCARG(uap, path), 3509 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3510 if (error != 0) 3511 return (error); 3512 3513 error = change_mode(vp, SCARG(uap, mode), l); 3514 3515 vrele(vp); 3516 return (error); 3517 } 3518 3519 /* 3520 * Common routine to set mode given a vnode. 3521 */ 3522 static int 3523 change_mode(struct vnode *vp, int mode, struct lwp *l) 3524 { 3525 struct vattr vattr; 3526 int error; 3527 3528 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3529 vattr_null(&vattr); 3530 vattr.va_mode = mode & ALLPERMS; 3531 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3532 VOP_UNLOCK(vp); 3533 return (error); 3534 } 3535 3536 /* 3537 * Set ownership given a path name; this version follows links. 3538 */ 3539 /* ARGSUSED */ 3540 int 3541 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3542 { 3543 /* { 3544 syscallarg(const char *) path; 3545 syscallarg(uid_t) uid; 3546 syscallarg(gid_t) gid; 3547 } */ 3548 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3549 SCARG(uap, gid), 0); 3550 } 3551 3552 int 3553 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3554 gid_t gid, int flags) 3555 { 3556 int error; 3557 struct vnode *vp; 3558 namei_simple_flags_t ns_flag; 3559 3560 if (flags & AT_SYMLINK_NOFOLLOW) 3561 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3562 else 3563 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3564 3565 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3566 if (error != 0) 3567 return error; 3568 3569 error = change_owner(vp, uid, gid, l, 0); 3570 3571 vrele(vp); 3572 3573 return (error); 3574 } 3575 3576 /* 3577 * Set ownership given a path name; this version follows links. 3578 * Provides POSIX semantics. 3579 */ 3580 /* ARGSUSED */ 3581 int 3582 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3583 { 3584 /* { 3585 syscallarg(const char *) path; 3586 syscallarg(uid_t) uid; 3587 syscallarg(gid_t) gid; 3588 } */ 3589 int error; 3590 struct vnode *vp; 3591 3592 error = namei_simple_user(SCARG(uap, path), 3593 NSM_FOLLOW_TRYEMULROOT, &vp); 3594 if (error != 0) 3595 return (error); 3596 3597 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3598 3599 vrele(vp); 3600 return (error); 3601 } 3602 3603 /* 3604 * Set ownership given a file descriptor. 3605 */ 3606 /* ARGSUSED */ 3607 int 3608 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3609 { 3610 /* { 3611 syscallarg(int) fd; 3612 syscallarg(uid_t) uid; 3613 syscallarg(gid_t) gid; 3614 } */ 3615 int error; 3616 file_t *fp; 3617 3618 /* fd_getvnode() will use the descriptor for us */ 3619 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3620 return (error); 3621 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3622 l, 0); 3623 fd_putfile(SCARG(uap, fd)); 3624 return (error); 3625 } 3626 3627 int 3628 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3629 register_t *retval) 3630 { 3631 /* { 3632 syscallarg(int) fd; 3633 syscallarg(const char *) path; 3634 syscallarg(uid_t) owner; 3635 syscallarg(gid_t) group; 3636 syscallarg(int) flag; 3637 } */ 3638 3639 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3640 SCARG(uap, owner), SCARG(uap, group), 3641 SCARG(uap, flag)); 3642 } 3643 3644 /* 3645 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3646 */ 3647 /* ARGSUSED */ 3648 int 3649 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3650 { 3651 /* { 3652 syscallarg(int) fd; 3653 syscallarg(uid_t) uid; 3654 syscallarg(gid_t) gid; 3655 } */ 3656 int error; 3657 file_t *fp; 3658 3659 /* fd_getvnode() will use the descriptor for us */ 3660 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3661 return (error); 3662 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3663 l, 1); 3664 fd_putfile(SCARG(uap, fd)); 3665 return (error); 3666 } 3667 3668 /* 3669 * Set ownership given a path name; this version does not follow links. 3670 */ 3671 /* ARGSUSED */ 3672 int 3673 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3674 { 3675 /* { 3676 syscallarg(const char *) path; 3677 syscallarg(uid_t) uid; 3678 syscallarg(gid_t) gid; 3679 } */ 3680 int error; 3681 struct vnode *vp; 3682 3683 error = namei_simple_user(SCARG(uap, path), 3684 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3685 if (error != 0) 3686 return (error); 3687 3688 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3689 3690 vrele(vp); 3691 return (error); 3692 } 3693 3694 /* 3695 * Set ownership given a path name; this version does not follow links. 3696 * Provides POSIX/XPG semantics. 3697 */ 3698 /* ARGSUSED */ 3699 int 3700 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3701 { 3702 /* { 3703 syscallarg(const char *) path; 3704 syscallarg(uid_t) uid; 3705 syscallarg(gid_t) gid; 3706 } */ 3707 int error; 3708 struct vnode *vp; 3709 3710 error = namei_simple_user(SCARG(uap, path), 3711 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3712 if (error != 0) 3713 return (error); 3714 3715 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3716 3717 vrele(vp); 3718 return (error); 3719 } 3720 3721 /* 3722 * Common routine to set ownership given a vnode. 3723 */ 3724 static int 3725 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3726 int posix_semantics) 3727 { 3728 struct vattr vattr; 3729 mode_t newmode; 3730 int error; 3731 3732 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3733 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3734 goto out; 3735 3736 #define CHANGED(x) ((int)(x) != -1) 3737 newmode = vattr.va_mode; 3738 if (posix_semantics) { 3739 /* 3740 * POSIX/XPG semantics: if the caller is not the super-user, 3741 * clear set-user-id and set-group-id bits. Both POSIX and 3742 * the XPG consider the behaviour for calls by the super-user 3743 * implementation-defined; we leave the set-user-id and set- 3744 * group-id settings intact in that case. 3745 */ 3746 if (vattr.va_mode & S_ISUID) { 3747 if (kauth_authorize_vnode(l->l_cred, 3748 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3749 newmode &= ~S_ISUID; 3750 } 3751 if (vattr.va_mode & S_ISGID) { 3752 if (kauth_authorize_vnode(l->l_cred, 3753 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3754 newmode &= ~S_ISGID; 3755 } 3756 } else { 3757 /* 3758 * NetBSD semantics: when changing owner and/or group, 3759 * clear the respective bit(s). 3760 */ 3761 if (CHANGED(uid)) 3762 newmode &= ~S_ISUID; 3763 if (CHANGED(gid)) 3764 newmode &= ~S_ISGID; 3765 } 3766 /* Update va_mode iff altered. */ 3767 if (vattr.va_mode == newmode) 3768 newmode = VNOVAL; 3769 3770 vattr_null(&vattr); 3771 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3772 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3773 vattr.va_mode = newmode; 3774 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3775 #undef CHANGED 3776 3777 out: 3778 VOP_UNLOCK(vp); 3779 return (error); 3780 } 3781 3782 /* 3783 * Set the access and modification times given a path name; this 3784 * version follows links. 3785 */ 3786 /* ARGSUSED */ 3787 int 3788 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3789 register_t *retval) 3790 { 3791 /* { 3792 syscallarg(const char *) path; 3793 syscallarg(const struct timeval *) tptr; 3794 } */ 3795 3796 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3797 SCARG(uap, tptr), UIO_USERSPACE); 3798 } 3799 3800 /* 3801 * Set the access and modification times given a file descriptor. 3802 */ 3803 /* ARGSUSED */ 3804 int 3805 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3806 register_t *retval) 3807 { 3808 /* { 3809 syscallarg(int) fd; 3810 syscallarg(const struct timeval *) tptr; 3811 } */ 3812 int error; 3813 file_t *fp; 3814 3815 /* fd_getvnode() will use the descriptor for us */ 3816 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3817 return (error); 3818 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3819 UIO_USERSPACE); 3820 fd_putfile(SCARG(uap, fd)); 3821 return (error); 3822 } 3823 3824 int 3825 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3826 register_t *retval) 3827 { 3828 /* { 3829 syscallarg(int) fd; 3830 syscallarg(const struct timespec *) tptr; 3831 } */ 3832 int error; 3833 file_t *fp; 3834 3835 /* fd_getvnode() will use the descriptor for us */ 3836 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3837 return (error); 3838 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3839 SCARG(uap, tptr), UIO_USERSPACE); 3840 fd_putfile(SCARG(uap, fd)); 3841 return (error); 3842 } 3843 3844 /* 3845 * Set the access and modification times given a path name; this 3846 * version does not follow links. 3847 */ 3848 int 3849 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3850 register_t *retval) 3851 { 3852 /* { 3853 syscallarg(const char *) path; 3854 syscallarg(const struct timeval *) tptr; 3855 } */ 3856 3857 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3858 SCARG(uap, tptr), UIO_USERSPACE); 3859 } 3860 3861 int 3862 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3863 register_t *retval) 3864 { 3865 /* { 3866 syscallarg(int) fd; 3867 syscallarg(const char *) path; 3868 syscallarg(const struct timespec *) tptr; 3869 syscallarg(int) flag; 3870 } */ 3871 int follow; 3872 const struct timespec *tptr; 3873 int error; 3874 3875 tptr = SCARG(uap, tptr); 3876 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3877 3878 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3879 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3880 3881 return error; 3882 } 3883 3884 /* 3885 * Common routine to set access and modification times given a vnode. 3886 */ 3887 int 3888 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3889 const struct timespec *tptr, enum uio_seg seg) 3890 { 3891 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3892 } 3893 3894 int 3895 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3896 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3897 { 3898 struct vattr vattr; 3899 int error, dorele = 0; 3900 namei_simple_flags_t sflags; 3901 bool vanull, setbirthtime; 3902 struct timespec ts[2]; 3903 3904 KASSERT(l != NULL || fdat == AT_FDCWD); 3905 3906 /* 3907 * I have checked all callers and they pass either FOLLOW, 3908 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3909 * is 0. More to the point, they don't pass anything else. 3910 * Let's keep it that way at least until the namei interfaces 3911 * are fully sanitized. 3912 */ 3913 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3914 sflags = (flag == FOLLOW) ? 3915 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3916 3917 if (tptr == NULL) { 3918 vanull = true; 3919 nanotime(&ts[0]); 3920 ts[1] = ts[0]; 3921 } else { 3922 vanull = false; 3923 if (seg != UIO_SYSSPACE) { 3924 error = copyin(tptr, ts, sizeof (ts)); 3925 if (error != 0) 3926 return error; 3927 } else { 3928 ts[0] = tptr[0]; 3929 ts[1] = tptr[1]; 3930 } 3931 } 3932 3933 if (ts[0].tv_nsec == UTIME_NOW) { 3934 nanotime(&ts[0]); 3935 if (ts[1].tv_nsec == UTIME_NOW) { 3936 vanull = true; 3937 ts[1] = ts[0]; 3938 } 3939 } else if (ts[1].tv_nsec == UTIME_NOW) 3940 nanotime(&ts[1]); 3941 3942 if (vp == NULL) { 3943 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3944 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3945 if (error != 0) 3946 return error; 3947 dorele = 1; 3948 } 3949 3950 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3951 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3952 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3953 vattr_null(&vattr); 3954 3955 if (ts[0].tv_nsec != UTIME_OMIT) 3956 vattr.va_atime = ts[0]; 3957 3958 if (ts[1].tv_nsec != UTIME_OMIT) { 3959 vattr.va_mtime = ts[1]; 3960 if (setbirthtime) 3961 vattr.va_birthtime = ts[1]; 3962 } 3963 3964 if (vanull) 3965 vattr.va_vaflags |= VA_UTIMES_NULL; 3966 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3967 VOP_UNLOCK(vp); 3968 3969 if (dorele != 0) 3970 vrele(vp); 3971 3972 return error; 3973 } 3974 3975 int 3976 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3977 const struct timeval *tptr, enum uio_seg seg) 3978 { 3979 struct timespec ts[2]; 3980 struct timespec *tsptr = NULL; 3981 int error; 3982 3983 if (tptr != NULL) { 3984 struct timeval tv[2]; 3985 3986 if (seg != UIO_SYSSPACE) { 3987 error = copyin(tptr, tv, sizeof(tv)); 3988 if (error != 0) 3989 return error; 3990 tptr = tv; 3991 } 3992 3993 if ((tptr[0].tv_usec == UTIME_NOW) || 3994 (tptr[0].tv_usec == UTIME_OMIT)) 3995 ts[0].tv_nsec = tptr[0].tv_usec; 3996 else { 3997 if (tptr[0].tv_usec < 0 || tptr[0].tv_usec >= 1000000) 3998 return EINVAL; 3999 4000 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 4001 } 4002 4003 if ((tptr[1].tv_usec == UTIME_NOW) || 4004 (tptr[1].tv_usec == UTIME_OMIT)) 4005 ts[1].tv_nsec = tptr[1].tv_usec; 4006 else { 4007 if (tptr[1].tv_usec < 0 || tptr[1].tv_usec >= 1000000) 4008 return EINVAL; 4009 4010 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 4011 } 4012 4013 tsptr = &ts[0]; 4014 } 4015 4016 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 4017 } 4018 4019 /* 4020 * Truncate a file given its path name. 4021 */ 4022 /* ARGSUSED */ 4023 int 4024 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 4025 { 4026 /* { 4027 syscallarg(const char *) path; 4028 syscallarg(int) pad; 4029 syscallarg(off_t) length; 4030 } */ 4031 struct vnode *vp; 4032 struct vattr vattr; 4033 int error; 4034 4035 if (SCARG(uap, length) < 0) 4036 return EINVAL; 4037 4038 error = namei_simple_user(SCARG(uap, path), 4039 NSM_FOLLOW_TRYEMULROOT, &vp); 4040 if (error != 0) 4041 return (error); 4042 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4043 if (vp->v_type == VDIR) 4044 error = EISDIR; 4045 else if ((error = vn_writechk(vp)) == 0 && 4046 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 4047 vattr_null(&vattr); 4048 vattr.va_size = SCARG(uap, length); 4049 error = VOP_SETATTR(vp, &vattr, l->l_cred); 4050 } 4051 vput(vp); 4052 return (error); 4053 } 4054 4055 /* 4056 * Truncate a file given a file descriptor. 4057 */ 4058 /* ARGSUSED */ 4059 int 4060 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 4061 { 4062 /* { 4063 syscallarg(int) fd; 4064 syscallarg(int) pad; 4065 syscallarg(off_t) length; 4066 } */ 4067 struct vattr vattr; 4068 struct vnode *vp; 4069 file_t *fp; 4070 int error; 4071 4072 if (SCARG(uap, length) < 0) 4073 return EINVAL; 4074 4075 /* fd_getvnode() will use the descriptor for us */ 4076 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4077 return (error); 4078 if ((fp->f_flag & FWRITE) == 0) { 4079 error = EINVAL; 4080 goto out; 4081 } 4082 vp = fp->f_vnode; 4083 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4084 if (vp->v_type == VDIR) 4085 error = EISDIR; 4086 else if ((error = vn_writechk(vp)) == 0) { 4087 vattr_null(&vattr); 4088 vattr.va_size = SCARG(uap, length); 4089 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 4090 } 4091 VOP_UNLOCK(vp); 4092 out: 4093 fd_putfile(SCARG(uap, fd)); 4094 return (error); 4095 } 4096 4097 /* 4098 * Sync an open file. 4099 */ 4100 /* ARGSUSED */ 4101 int 4102 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 4103 { 4104 /* { 4105 syscallarg(int) fd; 4106 } */ 4107 struct vnode *vp; 4108 file_t *fp; 4109 int error; 4110 4111 /* fd_getvnode() will use the descriptor for us */ 4112 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4113 return (error); 4114 vp = fp->f_vnode; 4115 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4116 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4117 VOP_UNLOCK(vp); 4118 fd_putfile(SCARG(uap, fd)); 4119 return (error); 4120 } 4121 4122 /* 4123 * Sync a range of file data. API modeled after that found in AIX. 4124 * 4125 * FDATASYNC indicates that we need only save enough metadata to be able 4126 * to re-read the written data. 4127 */ 4128 /* ARGSUSED */ 4129 int 4130 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4131 { 4132 /* { 4133 syscallarg(int) fd; 4134 syscallarg(int) flags; 4135 syscallarg(off_t) start; 4136 syscallarg(off_t) length; 4137 } */ 4138 struct vnode *vp; 4139 file_t *fp; 4140 int flags, nflags; 4141 off_t s, e, len; 4142 int error; 4143 4144 /* fd_getvnode() will use the descriptor for us */ 4145 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4146 return (error); 4147 4148 if ((fp->f_flag & FWRITE) == 0) { 4149 error = EBADF; 4150 goto out; 4151 } 4152 4153 flags = SCARG(uap, flags); 4154 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4155 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4156 error = EINVAL; 4157 goto out; 4158 } 4159 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4160 if (flags & FDATASYNC) 4161 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4162 else 4163 nflags = FSYNC_WAIT; 4164 if (flags & FDISKSYNC) 4165 nflags |= FSYNC_CACHE; 4166 4167 len = SCARG(uap, length); 4168 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4169 if (len) { 4170 s = SCARG(uap, start); 4171 if (s < 0 || len < 0 || len > OFF_T_MAX - s) { 4172 error = EINVAL; 4173 goto out; 4174 } 4175 e = s + len; 4176 KASSERT(s <= e); 4177 } else { 4178 e = 0; 4179 s = 0; 4180 } 4181 4182 vp = fp->f_vnode; 4183 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4184 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4185 VOP_UNLOCK(vp); 4186 out: 4187 fd_putfile(SCARG(uap, fd)); 4188 return (error); 4189 } 4190 4191 /* 4192 * Sync the data of an open file. 4193 */ 4194 /* ARGSUSED */ 4195 int 4196 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4197 { 4198 /* { 4199 syscallarg(int) fd; 4200 } */ 4201 struct vnode *vp; 4202 file_t *fp; 4203 int error; 4204 4205 /* fd_getvnode() will use the descriptor for us */ 4206 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4207 return (error); 4208 vp = fp->f_vnode; 4209 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4210 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4211 VOP_UNLOCK(vp); 4212 fd_putfile(SCARG(uap, fd)); 4213 return (error); 4214 } 4215 4216 /* 4217 * Rename files, (standard) BSD semantics frontend. 4218 */ 4219 /* ARGSUSED */ 4220 int 4221 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4222 { 4223 /* { 4224 syscallarg(const char *) from; 4225 syscallarg(const char *) to; 4226 } */ 4227 4228 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4229 SCARG(uap, to), UIO_USERSPACE, 0)); 4230 } 4231 4232 int 4233 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4234 register_t *retval) 4235 { 4236 /* { 4237 syscallarg(int) fromfd; 4238 syscallarg(const char *) from; 4239 syscallarg(int) tofd; 4240 syscallarg(const char *) to; 4241 } */ 4242 4243 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4244 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4245 } 4246 4247 /* 4248 * Rename files, POSIX semantics frontend. 4249 */ 4250 /* ARGSUSED */ 4251 int 4252 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4253 { 4254 /* { 4255 syscallarg(const char *) from; 4256 syscallarg(const char *) to; 4257 } */ 4258 4259 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4260 SCARG(uap, to), UIO_USERSPACE, 1)); 4261 } 4262 4263 /* 4264 * Rename files. Source and destination must either both be directories, 4265 * or both not be directories. If target is a directory, it must be empty. 4266 * If `from' and `to' refer to the same object, the value of the `retain' 4267 * argument is used to determine whether `from' will be 4268 * 4269 * (retain == 0) deleted unless `from' and `to' refer to the same 4270 * object in the file system's name space (BSD). 4271 * (retain == 1) always retained (POSIX). 4272 * 4273 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4274 */ 4275 int 4276 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4277 { 4278 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4279 } 4280 4281 static int 4282 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4283 const char *to, enum uio_seg seg, int retain) 4284 { 4285 struct pathbuf *fpb, *tpb; 4286 struct nameidata fnd, tnd; 4287 struct vnode *fdvp, *fvp; 4288 struct vnode *tdvp, *tvp; 4289 struct mount *mp, *tmp; 4290 int error; 4291 4292 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4293 4294 error = pathbuf_maybe_copyin(from, seg, &fpb); 4295 if (error) 4296 goto out0; 4297 KASSERT(fpb != NULL); 4298 4299 error = pathbuf_maybe_copyin(to, seg, &tpb); 4300 if (error) 4301 goto out1; 4302 KASSERT(tpb != NULL); 4303 4304 /* 4305 * Lookup from. 4306 * 4307 * XXX LOCKPARENT is wrong because we don't actually want it 4308 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4309 * insane, so for the time being we need to leave it like this. 4310 */ 4311 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4312 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4313 goto out2; 4314 4315 /* 4316 * Pull out the important results of the lookup, fdvp and fvp. 4317 * Of course, fvp is bogus because we're about to unlock fdvp. 4318 */ 4319 fdvp = fnd.ni_dvp; 4320 fvp = fnd.ni_vp; 4321 mp = fdvp->v_mount; 4322 KASSERT(fdvp != NULL); 4323 KASSERT(fvp != NULL); 4324 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4325 /* 4326 * Bracket the operation with fstrans_start()/fstrans_done(). 4327 * 4328 * Inside the bracket this file system cannot be unmounted so 4329 * a vnode on this file system cannot change its v_mount. 4330 * A vnode on another file system may still change to dead mount. 4331 */ 4332 fstrans_start(mp); 4333 4334 /* 4335 * Make sure neither fdvp nor fvp is locked. 4336 */ 4337 if (fdvp != fvp) 4338 VOP_UNLOCK(fdvp); 4339 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4340 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4341 4342 /* 4343 * Reject renaming `.' and `..'. Can't do this until after 4344 * namei because we need namei's parsing to find the final 4345 * component name. (namei should just leave us with the final 4346 * component name and not look it up itself, but anyway...) 4347 * 4348 * This was here before because we used to relookup from 4349 * instead of to and relookup requires the caller to check 4350 * this, but now file systems may depend on this check, so we 4351 * must retain it until the file systems are all rototilled. 4352 */ 4353 if (((fnd.ni_cnd.cn_namelen == 1) && 4354 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4355 ((fnd.ni_cnd.cn_namelen == 2) && 4356 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4357 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4358 error = EINVAL; /* XXX EISDIR? */ 4359 goto abort0; 4360 } 4361 4362 /* 4363 * Lookup to. 4364 * 4365 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4366 * fvp here to decide whether to add CREATEDIR is a load of 4367 * bollocks because fvp might be the wrong node by now, since 4368 * fdvp is unlocked. 4369 * 4370 * XXX Why not pass CREATEDIR always? 4371 */ 4372 NDINIT(&tnd, RENAME, 4373 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4374 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4375 tpb); 4376 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4377 goto abort0; 4378 4379 /* 4380 * Pull out the important results of the lookup, tdvp and tvp. 4381 * Of course, tvp is bogus because we're about to unlock tdvp. 4382 */ 4383 tdvp = tnd.ni_dvp; 4384 tvp = tnd.ni_vp; 4385 KASSERT(tdvp != NULL); 4386 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4387 4388 if (fvp->v_type == VDIR) 4389 tnd.ni_cnd.cn_flags |= WILLBEDIR; 4390 /* 4391 * Make sure neither tdvp nor tvp is locked. 4392 */ 4393 if (tdvp != tvp) 4394 VOP_UNLOCK(tdvp); 4395 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4396 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4397 4398 /* 4399 * Reject renaming onto `.' or `..'. relookup is unhappy with 4400 * these, which is why we must do this here. Once upon a time 4401 * we relooked up from instead of to, and consequently didn't 4402 * need this check, but now that we relookup to instead of 4403 * from, we need this; and we shall need it forever forward 4404 * until the VOP_RENAME protocol changes, because file systems 4405 * will no doubt begin to depend on this check. 4406 */ 4407 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4408 error = EISDIR; 4409 goto abort1; 4410 } 4411 if ((tnd.ni_cnd.cn_namelen == 2) && 4412 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4413 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4414 error = EINVAL; 4415 goto abort1; 4416 } 4417 4418 /* 4419 * Make sure the mount points match. Although we don't hold 4420 * any vnode locks, the v_mount on fdvp file system are stable. 4421 * 4422 * Unmounting another file system at an inopportune moment may 4423 * cause tdvp to disappear and change its v_mount to dead. 4424 * 4425 * So in either case different v_mount means cross-device rename. 4426 */ 4427 KASSERT(mp != NULL); 4428 tmp = tdvp->v_mount; 4429 4430 if (mp != tmp) { 4431 error = EXDEV; 4432 goto abort1; 4433 } 4434 4435 /* 4436 * Take the vfs rename lock to avoid cross-directory screw cases. 4437 * Nothing is locked currently, so taking this lock is safe. 4438 */ 4439 error = VFS_RENAMELOCK_ENTER(mp); 4440 if (error) 4441 goto abort1; 4442 4443 /* 4444 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4445 * and nothing is locked except for the vfs rename lock. 4446 * 4447 * The next step is a little rain dance to conform to the 4448 * insane lock protocol, even though it does nothing to ward 4449 * off race conditions. 4450 * 4451 * We need tdvp and tvp to be locked. However, because we have 4452 * unlocked tdvp in order to hold no locks while we take the 4453 * vfs rename lock, tvp may be wrong here, and we can't safely 4454 * lock it even if the sensible file systems will just unlock 4455 * it straight away. Consequently, we must lock tdvp and then 4456 * relookup tvp to get it locked. 4457 * 4458 * Finally, because the VOP_RENAME protocol is brain-damaged 4459 * and various file systems insanely depend on the semantics of 4460 * this brain damage, the lookup of to must be the last lookup 4461 * before VOP_RENAME. 4462 */ 4463 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4464 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4465 if (error) 4466 goto abort2; 4467 4468 /* 4469 * Drop the old tvp and pick up the new one -- which might be 4470 * the same, but that doesn't matter to us. After this, tdvp 4471 * and tvp should both be locked. 4472 */ 4473 if (tvp != NULL) 4474 vrele(tvp); 4475 tvp = tnd.ni_vp; 4476 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4477 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4478 4479 /* 4480 * The old do_sys_rename had various consistency checks here 4481 * involving fvp and tvp. fvp is bogus already here, and tvp 4482 * will become bogus soon in any sensible file system, so the 4483 * only purpose in putting these checks here is to give lip 4484 * service to these screw cases and to acknowledge that they 4485 * exist, not actually to handle them, but here you go 4486 * anyway... 4487 */ 4488 4489 /* 4490 * Acknowledge that directories and non-directories aren't 4491 * suposed to mix. 4492 */ 4493 if (tvp != NULL) { 4494 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4495 error = ENOTDIR; 4496 goto abort3; 4497 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4498 error = EISDIR; 4499 goto abort3; 4500 } 4501 } 4502 4503 /* 4504 * Acknowledge some random screw case, among the dozens that 4505 * might arise. 4506 */ 4507 if (fvp == tdvp) { 4508 error = EINVAL; 4509 goto abort3; 4510 } 4511 4512 /* 4513 * Acknowledge that POSIX has a wacky screw case. 4514 * 4515 * XXX Eventually the retain flag needs to be passed on to 4516 * VOP_RENAME. 4517 */ 4518 if (fvp == tvp) { 4519 if (retain) { 4520 error = 0; 4521 goto abort3; 4522 } else if ((fdvp == tdvp) && 4523 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4524 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4525 fnd.ni_cnd.cn_namelen))) { 4526 error = 0; 4527 goto abort3; 4528 } 4529 } 4530 4531 /* 4532 * Make sure veriexec can screw us up. (But a race can screw 4533 * up veriexec, of course -- remember, fvp and (soon) tvp are 4534 * bogus.) 4535 */ 4536 #if NVERIEXEC > 0 4537 { 4538 char *f1, *f2; 4539 size_t f1_len; 4540 size_t f2_len; 4541 4542 f1_len = fnd.ni_cnd.cn_namelen + 1; 4543 f1 = kmem_alloc(f1_len, KM_SLEEP); 4544 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4545 4546 f2_len = tnd.ni_cnd.cn_namelen + 1; 4547 f2 = kmem_alloc(f2_len, KM_SLEEP); 4548 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4549 4550 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4551 4552 kmem_free(f1, f1_len); 4553 kmem_free(f2, f2_len); 4554 4555 if (error) 4556 goto abort3; 4557 } 4558 #endif /* NVERIEXEC > 0 */ 4559 4560 /* 4561 * All ready. Incant the rename vop. 4562 */ 4563 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4564 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4565 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4566 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4567 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4568 4569 /* 4570 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4571 * tdvp and tvp. But we can't assert any of that. 4572 */ 4573 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4574 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4575 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4576 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4577 4578 /* 4579 * So all we have left to do is to drop the rename lock and 4580 * destroy the pathbufs. 4581 */ 4582 VFS_RENAMELOCK_EXIT(mp); 4583 fstrans_done(mp); 4584 goto out2; 4585 4586 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4587 VOP_UNLOCK(tvp); 4588 abort2: VOP_UNLOCK(tdvp); 4589 VFS_RENAMELOCK_EXIT(mp); 4590 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4591 vrele(tdvp); 4592 if (tvp != NULL) 4593 vrele(tvp); 4594 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4595 vrele(fdvp); 4596 vrele(fvp); 4597 fstrans_done(mp); 4598 out2: pathbuf_destroy(tpb); 4599 out1: pathbuf_destroy(fpb); 4600 out0: return error; 4601 } 4602 4603 /* 4604 * Make a directory file. 4605 */ 4606 /* ARGSUSED */ 4607 int 4608 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4609 { 4610 /* { 4611 syscallarg(const char *) path; 4612 syscallarg(int) mode; 4613 } */ 4614 4615 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4616 SCARG(uap, mode), UIO_USERSPACE); 4617 } 4618 4619 int 4620 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4621 register_t *retval) 4622 { 4623 /* { 4624 syscallarg(int) fd; 4625 syscallarg(const char *) path; 4626 syscallarg(int) mode; 4627 } */ 4628 4629 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4630 SCARG(uap, mode), UIO_USERSPACE); 4631 } 4632 4633 4634 int 4635 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4636 { 4637 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, seg); 4638 } 4639 4640 static int 4641 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4642 enum uio_seg seg) 4643 { 4644 struct proc *p = curlwp->l_proc; 4645 struct vnode *vp; 4646 struct vattr vattr; 4647 int error; 4648 struct pathbuf *pb; 4649 struct nameidata nd; 4650 4651 KASSERT(l != NULL || fdat == AT_FDCWD); 4652 4653 /* XXX bollocks, should pass in a pathbuf */ 4654 error = pathbuf_maybe_copyin(path, seg, &pb); 4655 if (error) { 4656 return error; 4657 } 4658 4659 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4660 4661 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4662 pathbuf_destroy(pb); 4663 return (error); 4664 } 4665 vp = nd.ni_vp; 4666 if (vp != NULL) { 4667 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4668 if (nd.ni_dvp == vp) 4669 vrele(nd.ni_dvp); 4670 else 4671 vput(nd.ni_dvp); 4672 vrele(vp); 4673 pathbuf_destroy(pb); 4674 return (EEXIST); 4675 } 4676 vattr_null(&vattr); 4677 vattr.va_type = VDIR; 4678 /* We will read cwdi->cwdi_cmask unlocked. */ 4679 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4680 nd.ni_cnd.cn_flags |= WILLBEDIR; 4681 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4682 if (!error) 4683 vrele(nd.ni_vp); 4684 vput(nd.ni_dvp); 4685 pathbuf_destroy(pb); 4686 return (error); 4687 } 4688 4689 /* 4690 * Remove a directory file. 4691 */ 4692 /* ARGSUSED */ 4693 int 4694 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4695 { 4696 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4697 AT_REMOVEDIR, UIO_USERSPACE); 4698 } 4699 4700 /* 4701 * Read a block of directory entries in a file system independent format. 4702 */ 4703 int 4704 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4705 { 4706 /* { 4707 syscallarg(int) fd; 4708 syscallarg(char *) buf; 4709 syscallarg(size_t) count; 4710 } */ 4711 file_t *fp; 4712 int error, done; 4713 4714 /* fd_getvnode() will use the descriptor for us */ 4715 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4716 return (error); 4717 if ((fp->f_flag & FREAD) == 0) { 4718 error = EBADF; 4719 goto out; 4720 } 4721 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4722 SCARG(uap, count), &done, l, 0, 0); 4723 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4724 *retval = done; 4725 out: 4726 fd_putfile(SCARG(uap, fd)); 4727 return (error); 4728 } 4729 4730 /* 4731 * Set the mode mask for creation of filesystem nodes. 4732 */ 4733 int 4734 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4735 { 4736 /* { 4737 syscallarg(mode_t) newmask; 4738 } */ 4739 4740 /* 4741 * cwdi->cwdi_cmask will be read unlocked elsewhere, and no kind of 4742 * serialization with those reads is required. It's important to 4743 * return a coherent answer for the caller of umask() though, and 4744 * the atomic operation accomplishes that. 4745 */ 4746 *retval = atomic_swap_uint(&curproc->p_cwdi->cwdi_cmask, 4747 SCARG(uap, newmask) & ALLPERMS); 4748 4749 return (0); 4750 } 4751 4752 int 4753 dorevoke(struct vnode *vp, kauth_cred_t cred) 4754 { 4755 struct vattr vattr; 4756 int error, fs_decision; 4757 4758 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4759 error = VOP_GETATTR(vp, &vattr, cred); 4760 VOP_UNLOCK(vp); 4761 if (error != 0) 4762 return error; 4763 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4764 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4765 fs_decision); 4766 if (!error) 4767 VOP_REVOKE(vp, REVOKEALL); 4768 return (error); 4769 } 4770 4771 /* 4772 * Void all references to file by ripping underlying filesystem 4773 * away from vnode. 4774 */ 4775 /* ARGSUSED */ 4776 int 4777 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4778 { 4779 /* { 4780 syscallarg(const char *) path; 4781 } */ 4782 struct vnode *vp; 4783 int error; 4784 4785 error = namei_simple_user(SCARG(uap, path), 4786 NSM_FOLLOW_TRYEMULROOT, &vp); 4787 if (error != 0) 4788 return (error); 4789 error = dorevoke(vp, l->l_cred); 4790 vrele(vp); 4791 return (error); 4792 } 4793 4794 /* 4795 * Allocate backing store for a file, filling a hole without having to 4796 * explicitly write anything out. 4797 */ 4798 /* ARGSUSED */ 4799 int 4800 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4801 register_t *retval) 4802 { 4803 /* { 4804 syscallarg(int) fd; 4805 syscallarg(off_t) pos; 4806 syscallarg(off_t) len; 4807 } */ 4808 int fd; 4809 off_t pos, len; 4810 struct file *fp; 4811 struct vnode *vp; 4812 int error; 4813 4814 fd = SCARG(uap, fd); 4815 pos = SCARG(uap, pos); 4816 len = SCARG(uap, len); 4817 4818 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4819 *retval = EINVAL; 4820 return 0; 4821 } 4822 4823 error = fd_getvnode(fd, &fp); 4824 if (error) { 4825 *retval = error; 4826 return 0; 4827 } 4828 if ((fp->f_flag & FWRITE) == 0) { 4829 error = EBADF; 4830 goto fail; 4831 } 4832 vp = fp->f_vnode; 4833 4834 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4835 if (vp->v_type == VDIR) { 4836 error = EISDIR; 4837 } else { 4838 error = VOP_FALLOCATE(vp, pos, len); 4839 } 4840 VOP_UNLOCK(vp); 4841 4842 fail: 4843 fd_putfile(fd); 4844 *retval = error; 4845 return 0; 4846 } 4847 4848 /* 4849 * Deallocate backing store for a file, creating a hole. Also used for 4850 * invoking TRIM on disks. 4851 */ 4852 /* ARGSUSED */ 4853 int 4854 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4855 register_t *retval) 4856 { 4857 /* { 4858 syscallarg(int) fd; 4859 syscallarg(off_t) pos; 4860 syscallarg(off_t) len; 4861 } */ 4862 int fd; 4863 off_t pos, len; 4864 struct file *fp; 4865 struct vnode *vp; 4866 int error; 4867 4868 fd = SCARG(uap, fd); 4869 pos = SCARG(uap, pos); 4870 len = SCARG(uap, len); 4871 4872 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4873 return EINVAL; 4874 } 4875 4876 error = fd_getvnode(fd, &fp); 4877 if (error) { 4878 return error; 4879 } 4880 if ((fp->f_flag & FWRITE) == 0) { 4881 error = EBADF; 4882 goto fail; 4883 } 4884 vp = fp->f_vnode; 4885 4886 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4887 if (vp->v_type == VDIR) { 4888 error = EISDIR; 4889 } else { 4890 error = VOP_FDISCARD(vp, pos, len); 4891 } 4892 VOP_UNLOCK(vp); 4893 4894 fail: 4895 fd_putfile(fd); 4896 return error; 4897 } 4898
Indexes created Thu Sep 25 06:09:39 GMT 2025