vfs_syscalls.c revision 1.550
1 /* $NetBSD: vfs_syscalls.c,v 1.550 2021/06/29 22:40:53 dholland 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 1.550 2021/06/29 22:40:53 dholland 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 == 0) { 1678 *fd = indx; 1679 return 0; 1680 } 1681 } else { 1682 error = open_setfp(l, fp, vp, indx, flags); 1683 if (error) 1684 return error; 1685 VOP_UNLOCK(vp); 1686 *fd = indx; 1687 fd_affix(p, fp, indx); 1688 } 1689 1690 return 0; 1691 } 1692 1693 int 1694 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1695 { 1696 struct pathbuf *pb; 1697 int error, oflags; 1698 1699 oflags = FFLAGS(open_flags); 1700 if ((oflags & (FREAD | FWRITE)) == 0) 1701 return EINVAL; 1702 1703 pb = pathbuf_create(path); 1704 if (pb == NULL) 1705 return ENOMEM; 1706 1707 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1708 pathbuf_destroy(pb); 1709 1710 return error; 1711 } 1712 1713 static int 1714 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1715 int mode, int *fd) 1716 { 1717 file_t *dfp = NULL; 1718 struct vnode *dvp = NULL; 1719 struct pathbuf *pb; 1720 const char *pathstring = NULL; 1721 int error; 1722 1723 if (path == NULL) { 1724 MODULE_HOOK_CALL(vfs_openat_10_hook, (&pb), enosys(), error); 1725 if (error == ENOSYS) 1726 goto no_compat; 1727 if (error) 1728 return error; 1729 } else { 1730 no_compat: 1731 error = pathbuf_copyin(path, &pb); 1732 if (error) 1733 return error; 1734 } 1735 1736 pathstring = pathbuf_stringcopy_get(pb); 1737 1738 /* 1739 * fdat is ignored if: 1740 * 1) if fdat is AT_FDCWD, which means use current directory as base. 1741 * 2) if path is absolute, then fdat is useless. 1742 */ 1743 if (fdat != AT_FDCWD && pathstring[0] != '/') { 1744 /* fd_getvnode() will use the descriptor for us */ 1745 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1746 goto out; 1747 1748 dvp = dfp->f_vnode; 1749 } 1750 1751 error = do_open(l, dvp, pb, flags, mode, fd); 1752 1753 if (dfp != NULL) 1754 fd_putfile(fdat); 1755 out: 1756 pathbuf_stringcopy_put(pb, pathstring); 1757 pathbuf_destroy(pb); 1758 return error; 1759 } 1760 1761 int 1762 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1763 { 1764 /* { 1765 syscallarg(const char *) path; 1766 syscallarg(int) flags; 1767 syscallarg(int) mode; 1768 } */ 1769 int error; 1770 int fd; 1771 1772 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1773 SCARG(uap, flags), SCARG(uap, mode), &fd); 1774 1775 if (error == 0) 1776 *retval = fd; 1777 1778 return error; 1779 } 1780 1781 int 1782 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1783 { 1784 /* { 1785 syscallarg(int) fd; 1786 syscallarg(const char *) path; 1787 syscallarg(int) oflags; 1788 syscallarg(int) mode; 1789 } */ 1790 int error; 1791 int fd; 1792 1793 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1794 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1795 1796 if (error == 0) 1797 *retval = fd; 1798 1799 return error; 1800 } 1801 1802 static void 1803 vfs__fhfree(fhandle_t *fhp) 1804 { 1805 size_t fhsize; 1806 1807 fhsize = FHANDLE_SIZE(fhp); 1808 kmem_free(fhp, fhsize); 1809 } 1810 1811 /* 1812 * vfs_composefh: compose a filehandle. 1813 */ 1814 1815 int 1816 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1817 { 1818 struct mount *mp; 1819 struct fid *fidp; 1820 int error; 1821 size_t needfhsize; 1822 size_t fidsize; 1823 1824 mp = vp->v_mount; 1825 fidp = NULL; 1826 if (*fh_size < FHANDLE_SIZE_MIN) { 1827 fidsize = 0; 1828 } else { 1829 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1830 if (fhp != NULL) { 1831 memset(fhp, 0, *fh_size); 1832 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1833 fidp = &fhp->fh_fid; 1834 } 1835 } 1836 error = VFS_VPTOFH(vp, fidp, &fidsize); 1837 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1838 if (error == 0 && *fh_size < needfhsize) { 1839 error = E2BIG; 1840 } 1841 *fh_size = needfhsize; 1842 return error; 1843 } 1844 1845 int 1846 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1847 { 1848 struct mount *mp; 1849 fhandle_t *fhp; 1850 size_t fhsize; 1851 size_t fidsize; 1852 int error; 1853 1854 mp = vp->v_mount; 1855 fidsize = 0; 1856 error = VFS_VPTOFH(vp, NULL, &fidsize); 1857 KASSERT(error != 0); 1858 if (error != E2BIG) { 1859 goto out; 1860 } 1861 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1862 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1863 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1864 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1865 if (error == 0) { 1866 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1867 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1868 *fhpp = fhp; 1869 } else { 1870 kmem_free(fhp, fhsize); 1871 } 1872 out: 1873 return error; 1874 } 1875 1876 void 1877 vfs_composefh_free(fhandle_t *fhp) 1878 { 1879 1880 vfs__fhfree(fhp); 1881 } 1882 1883 /* 1884 * vfs_fhtovp: lookup a vnode by a filehandle. 1885 */ 1886 1887 int 1888 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1889 { 1890 struct mount *mp; 1891 int error; 1892 1893 *vpp = NULL; 1894 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1895 if (mp == NULL) { 1896 error = ESTALE; 1897 goto out; 1898 } 1899 if (mp->mnt_op->vfs_fhtovp == NULL) { 1900 error = EOPNOTSUPP; 1901 goto out; 1902 } 1903 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), LK_EXCLUSIVE, vpp); 1904 out: 1905 return error; 1906 } 1907 1908 /* 1909 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1910 * the needed size. 1911 */ 1912 1913 int 1914 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1915 { 1916 fhandle_t *fhp; 1917 int error; 1918 1919 if (fhsize > FHANDLE_SIZE_MAX) { 1920 return EINVAL; 1921 } 1922 if (fhsize < FHANDLE_SIZE_MIN) { 1923 return EINVAL; 1924 } 1925 again: 1926 fhp = kmem_alloc(fhsize, KM_SLEEP); 1927 error = copyin(ufhp, fhp, fhsize); 1928 if (error == 0) { 1929 /* XXX this check shouldn't be here */ 1930 if (FHANDLE_SIZE(fhp) == fhsize) { 1931 *fhpp = fhp; 1932 return 0; 1933 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1934 /* 1935 * a kludge for nfsv2 padded handles. 1936 */ 1937 size_t sz; 1938 1939 sz = FHANDLE_SIZE(fhp); 1940 kmem_free(fhp, fhsize); 1941 fhsize = sz; 1942 goto again; 1943 } else { 1944 /* 1945 * userland told us wrong size. 1946 */ 1947 error = EINVAL; 1948 } 1949 } 1950 kmem_free(fhp, fhsize); 1951 return error; 1952 } 1953 1954 void 1955 vfs_copyinfh_free(fhandle_t *fhp) 1956 { 1957 1958 vfs__fhfree(fhp); 1959 } 1960 1961 /* 1962 * Get file handle system call 1963 */ 1964 int 1965 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1966 { 1967 /* { 1968 syscallarg(char *) fname; 1969 syscallarg(fhandle_t *) fhp; 1970 syscallarg(size_t *) fh_size; 1971 } */ 1972 struct vnode *vp; 1973 fhandle_t *fh; 1974 int error; 1975 struct pathbuf *pb; 1976 struct nameidata nd; 1977 size_t sz; 1978 size_t usz; 1979 1980 /* 1981 * Must be super user 1982 */ 1983 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1984 0, NULL, NULL, NULL); 1985 if (error) 1986 return (error); 1987 1988 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1989 if (error) { 1990 return error; 1991 } 1992 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1993 error = namei(&nd); 1994 if (error) { 1995 pathbuf_destroy(pb); 1996 return error; 1997 } 1998 vp = nd.ni_vp; 1999 pathbuf_destroy(pb); 2000 2001 error = vfs_composefh_alloc(vp, &fh); 2002 vput(vp); 2003 if (error != 0) { 2004 return error; 2005 } 2006 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 2007 if (error != 0) { 2008 goto out; 2009 } 2010 sz = FHANDLE_SIZE(fh); 2011 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 2012 if (error != 0) { 2013 goto out; 2014 } 2015 if (usz >= sz) { 2016 error = copyout(fh, SCARG(uap, fhp), sz); 2017 } else { 2018 error = E2BIG; 2019 } 2020 out: 2021 vfs_composefh_free(fh); 2022 return (error); 2023 } 2024 2025 /* 2026 * Open a file given a file handle. 2027 * 2028 * Check permissions, allocate an open file structure, 2029 * and call the device open routine if any. 2030 */ 2031 2032 int 2033 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 2034 register_t *retval) 2035 { 2036 file_t *fp; 2037 struct vnode *vp = NULL; 2038 kauth_cred_t cred = l->l_cred; 2039 file_t *nfp; 2040 int indx, error; 2041 struct vattr va; 2042 fhandle_t *fh; 2043 int flags; 2044 proc_t *p; 2045 2046 p = curproc; 2047 2048 /* 2049 * Must be super user 2050 */ 2051 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2052 0, NULL, NULL, NULL))) 2053 return (error); 2054 2055 if (oflags & O_SEARCH) { 2056 oflags &= ~(int)O_SEARCH; 2057 } 2058 2059 flags = FFLAGS(oflags); 2060 if ((flags & (FREAD | FWRITE)) == 0) 2061 return (EINVAL); 2062 if ((flags & O_CREAT)) 2063 return (EINVAL); 2064 if ((error = fd_allocfile(&nfp, &indx)) != 0) 2065 return (error); 2066 fp = nfp; 2067 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2068 if (error != 0) { 2069 goto bad; 2070 } 2071 error = vfs_fhtovp(fh, &vp); 2072 vfs_copyinfh_free(fh); 2073 if (error != 0) { 2074 goto bad; 2075 } 2076 2077 /* Now do an effective vn_open */ 2078 2079 if (vp->v_type == VSOCK) { 2080 error = EOPNOTSUPP; 2081 goto bad; 2082 } 2083 error = vn_openchk(vp, cred, flags); 2084 if (error != 0) 2085 goto bad; 2086 if (flags & O_TRUNC) { 2087 VOP_UNLOCK(vp); /* XXX */ 2088 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2089 vattr_null(&va); 2090 va.va_size = 0; 2091 error = VOP_SETATTR(vp, &va, cred); 2092 if (error) 2093 goto bad; 2094 } 2095 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2096 goto bad; 2097 if (flags & FWRITE) { 2098 mutex_enter(vp->v_interlock); 2099 vp->v_writecount++; 2100 mutex_exit(vp->v_interlock); 2101 } 2102 2103 /* done with modified vn_open, now finish what sys_open does. */ 2104 if ((error = open_setfp(l, fp, vp, indx, flags))) 2105 return error; 2106 2107 VOP_UNLOCK(vp); 2108 *retval = indx; 2109 fd_affix(p, fp, indx); 2110 return (0); 2111 2112 bad: 2113 fd_abort(p, fp, indx); 2114 if (vp != NULL) 2115 vput(vp); 2116 if (error == EDUPFD || error == EMOVEFD) { 2117 /* XXX should probably close curlwp->l_dupfd */ 2118 error = EOPNOTSUPP; 2119 } 2120 return (error); 2121 } 2122 2123 int 2124 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2125 { 2126 /* { 2127 syscallarg(const void *) fhp; 2128 syscallarg(size_t) fh_size; 2129 syscallarg(int) flags; 2130 } */ 2131 2132 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2133 SCARG(uap, flags), retval); 2134 } 2135 2136 int 2137 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2138 { 2139 int error; 2140 fhandle_t *fh; 2141 struct vnode *vp; 2142 2143 /* 2144 * Must be super user 2145 */ 2146 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2147 0, NULL, NULL, NULL))) 2148 return (error); 2149 2150 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2151 if (error != 0) 2152 return error; 2153 2154 error = vfs_fhtovp(fh, &vp); 2155 vfs_copyinfh_free(fh); 2156 if (error != 0) 2157 return error; 2158 2159 error = vn_stat(vp, sb); 2160 vput(vp); 2161 return error; 2162 } 2163 2164 2165 /* ARGSUSED */ 2166 int 2167 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2168 { 2169 /* { 2170 syscallarg(const void *) fhp; 2171 syscallarg(size_t) fh_size; 2172 syscallarg(struct stat *) sb; 2173 } */ 2174 struct stat sb; 2175 int error; 2176 2177 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2178 if (error) 2179 return error; 2180 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2181 } 2182 2183 int 2184 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2185 int flags) 2186 { 2187 fhandle_t *fh; 2188 struct mount *mp; 2189 struct vnode *vp; 2190 int error; 2191 2192 /* 2193 * Must be super user 2194 */ 2195 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2196 0, NULL, NULL, NULL))) 2197 return error; 2198 2199 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2200 if (error != 0) 2201 return error; 2202 2203 error = vfs_fhtovp(fh, &vp); 2204 vfs_copyinfh_free(fh); 2205 if (error != 0) 2206 return error; 2207 2208 mp = vp->v_mount; 2209 error = dostatvfs(mp, sb, l, flags, 1); 2210 vput(vp); 2211 return error; 2212 } 2213 2214 /* ARGSUSED */ 2215 int 2216 sys___fhstatvfs190(struct lwp *l, const struct sys___fhstatvfs190_args *uap, register_t *retval) 2217 { 2218 /* { 2219 syscallarg(const void *) fhp; 2220 syscallarg(size_t) fh_size; 2221 syscallarg(struct statvfs *) buf; 2222 syscallarg(int) flags; 2223 } */ 2224 struct statvfs *sb = STATVFSBUF_GET(); 2225 int error; 2226 2227 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2228 SCARG(uap, flags)); 2229 if (error == 0) 2230 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2231 STATVFSBUF_PUT(sb); 2232 return error; 2233 } 2234 2235 int 2236 do_posix_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2237 dev_t dev) 2238 { 2239 2240 /* 2241 * The POSIX mknod(2) call is an alias for mkfifo(2) for S_IFIFO 2242 * in mode and dev=0. 2243 * 2244 * In all the other cases it's implementation defined behavior. 2245 */ 2246 2247 if ((mode & S_IFIFO) && dev == 0) 2248 return do_sys_mkfifoat(l, fdat, pathname, mode); 2249 else 2250 return do_sys_mknodat(l, fdat, pathname, mode, dev, 2251 UIO_USERSPACE); 2252 } 2253 2254 /* 2255 * Create a special file. 2256 */ 2257 /* ARGSUSED */ 2258 int 2259 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2260 register_t *retval) 2261 { 2262 /* { 2263 syscallarg(const char *) path; 2264 syscallarg(mode_t) mode; 2265 syscallarg(dev_t) dev; 2266 } */ 2267 return do_posix_mknodat(l, AT_FDCWD, SCARG(uap, path), 2268 SCARG(uap, mode), SCARG(uap, dev)); 2269 } 2270 2271 int 2272 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2273 register_t *retval) 2274 { 2275 /* { 2276 syscallarg(int) fd; 2277 syscallarg(const char *) path; 2278 syscallarg(mode_t) mode; 2279 syscallarg(int) pad; 2280 syscallarg(dev_t) dev; 2281 } */ 2282 2283 return do_posix_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2284 SCARG(uap, mode), SCARG(uap, dev)); 2285 } 2286 2287 int 2288 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2289 enum uio_seg seg) 2290 { 2291 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, seg); 2292 } 2293 2294 int 2295 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2296 dev_t dev, enum uio_seg seg) 2297 { 2298 struct proc *p = l->l_proc; 2299 struct vnode *vp; 2300 struct vattr vattr; 2301 int error, optype; 2302 struct pathbuf *pb; 2303 struct nameidata nd; 2304 const char *pathstring; 2305 2306 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2307 0, NULL, NULL, NULL)) != 0) 2308 return (error); 2309 2310 optype = VOP_MKNOD_DESCOFFSET; 2311 2312 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2313 if (error) { 2314 return error; 2315 } 2316 pathstring = pathbuf_stringcopy_get(pb); 2317 if (pathstring == NULL) { 2318 pathbuf_destroy(pb); 2319 return ENOMEM; 2320 } 2321 2322 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2323 2324 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2325 goto out; 2326 vp = nd.ni_vp; 2327 2328 if (vp != NULL) 2329 error = EEXIST; 2330 else { 2331 vattr_null(&vattr); 2332 /* We will read cwdi->cwdi_cmask unlocked. */ 2333 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2334 vattr.va_rdev = dev; 2335 2336 switch (mode & S_IFMT) { 2337 case S_IFMT: /* used by badsect to flag bad sectors */ 2338 vattr.va_type = VBAD; 2339 break; 2340 case S_IFCHR: 2341 vattr.va_type = VCHR; 2342 break; 2343 case S_IFBLK: 2344 vattr.va_type = VBLK; 2345 break; 2346 case S_IFWHT: 2347 optype = VOP_WHITEOUT_DESCOFFSET; 2348 break; 2349 case S_IFREG: 2350 #if NVERIEXEC > 0 2351 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2352 O_CREAT); 2353 #endif /* NVERIEXEC > 0 */ 2354 vattr.va_type = VREG; 2355 vattr.va_rdev = VNOVAL; 2356 optype = VOP_CREATE_DESCOFFSET; 2357 break; 2358 default: 2359 error = EINVAL; 2360 break; 2361 } 2362 2363 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET && 2364 vattr.va_rdev == VNOVAL) 2365 error = EINVAL; 2366 } 2367 2368 if (!error) { 2369 switch (optype) { 2370 case VOP_WHITEOUT_DESCOFFSET: 2371 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2372 if (error) 2373 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2374 vput(nd.ni_dvp); 2375 break; 2376 2377 case VOP_MKNOD_DESCOFFSET: 2378 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2379 &nd.ni_cnd, &vattr); 2380 if (error == 0) 2381 vrele(nd.ni_vp); 2382 vput(nd.ni_dvp); 2383 break; 2384 2385 case VOP_CREATE_DESCOFFSET: 2386 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2387 &nd.ni_cnd, &vattr); 2388 if (error == 0) 2389 vrele(nd.ni_vp); 2390 vput(nd.ni_dvp); 2391 break; 2392 } 2393 } else { 2394 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2395 if (nd.ni_dvp == vp) 2396 vrele(nd.ni_dvp); 2397 else 2398 vput(nd.ni_dvp); 2399 if (vp) 2400 vrele(vp); 2401 } 2402 out: 2403 pathbuf_stringcopy_put(pb, pathstring); 2404 pathbuf_destroy(pb); 2405 return (error); 2406 } 2407 2408 /* 2409 * Create a named pipe. 2410 */ 2411 /* ARGSUSED */ 2412 int 2413 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2414 { 2415 /* { 2416 syscallarg(const char *) path; 2417 syscallarg(int) mode; 2418 } */ 2419 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2420 } 2421 2422 int 2423 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2424 register_t *retval) 2425 { 2426 /* { 2427 syscallarg(int) fd; 2428 syscallarg(const char *) path; 2429 syscallarg(int) mode; 2430 } */ 2431 2432 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2433 SCARG(uap, mode)); 2434 } 2435 2436 static int 2437 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2438 { 2439 struct proc *p = l->l_proc; 2440 struct vattr vattr; 2441 int error; 2442 struct pathbuf *pb; 2443 struct nameidata nd; 2444 2445 error = pathbuf_copyin(path, &pb); 2446 if (error) { 2447 return error; 2448 } 2449 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2450 2451 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2452 pathbuf_destroy(pb); 2453 return error; 2454 } 2455 if (nd.ni_vp != NULL) { 2456 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2457 if (nd.ni_dvp == nd.ni_vp) 2458 vrele(nd.ni_dvp); 2459 else 2460 vput(nd.ni_dvp); 2461 vrele(nd.ni_vp); 2462 pathbuf_destroy(pb); 2463 return (EEXIST); 2464 } 2465 vattr_null(&vattr); 2466 vattr.va_type = VFIFO; 2467 /* We will read cwdi->cwdi_cmask unlocked. */ 2468 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2469 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2470 if (error == 0) 2471 vrele(nd.ni_vp); 2472 vput(nd.ni_dvp); 2473 pathbuf_destroy(pb); 2474 return (error); 2475 } 2476 2477 /* 2478 * Make a hard file link. 2479 */ 2480 /* ARGSUSED */ 2481 int 2482 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2483 const char *link, int follow, register_t *retval) 2484 { 2485 struct vnode *vp; 2486 struct pathbuf *linkpb; 2487 struct nameidata nd; 2488 namei_simple_flags_t ns_flags; 2489 int error; 2490 2491 if (follow & AT_SYMLINK_FOLLOW) 2492 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2493 else 2494 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2495 2496 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2497 if (error != 0) 2498 return (error); 2499 error = pathbuf_copyin(link, &linkpb); 2500 if (error) { 2501 goto out1; 2502 } 2503 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2504 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2505 goto out2; 2506 if (nd.ni_vp) { 2507 error = EEXIST; 2508 goto abortop; 2509 } 2510 /* Prevent hard links on directories. */ 2511 if (vp->v_type == VDIR) { 2512 error = EPERM; 2513 goto abortop; 2514 } 2515 /* Prevent cross-mount operation. */ 2516 if (nd.ni_dvp->v_mount != vp->v_mount) { 2517 error = EXDEV; 2518 goto abortop; 2519 } 2520 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2521 VOP_UNLOCK(nd.ni_dvp); 2522 vrele(nd.ni_dvp); 2523 out2: 2524 pathbuf_destroy(linkpb); 2525 out1: 2526 vrele(vp); 2527 return (error); 2528 abortop: 2529 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2530 if (nd.ni_dvp == nd.ni_vp) 2531 vrele(nd.ni_dvp); 2532 else 2533 vput(nd.ni_dvp); 2534 if (nd.ni_vp != NULL) 2535 vrele(nd.ni_vp); 2536 goto out2; 2537 } 2538 2539 int 2540 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2541 { 2542 /* { 2543 syscallarg(const char *) path; 2544 syscallarg(const char *) link; 2545 } */ 2546 const char *path = SCARG(uap, path); 2547 const char *link = SCARG(uap, link); 2548 2549 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2550 AT_SYMLINK_FOLLOW, retval); 2551 } 2552 2553 int 2554 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2555 register_t *retval) 2556 { 2557 /* { 2558 syscallarg(int) fd1; 2559 syscallarg(const char *) name1; 2560 syscallarg(int) fd2; 2561 syscallarg(const char *) name2; 2562 syscallarg(int) flags; 2563 } */ 2564 int fd1 = SCARG(uap, fd1); 2565 const char *name1 = SCARG(uap, name1); 2566 int fd2 = SCARG(uap, fd2); 2567 const char *name2 = SCARG(uap, name2); 2568 int follow; 2569 2570 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2571 2572 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2573 } 2574 2575 2576 int 2577 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2578 { 2579 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2580 } 2581 2582 static int 2583 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2584 const char *link, enum uio_seg seg) 2585 { 2586 struct proc *p = curproc; 2587 struct vattr vattr; 2588 char *path; 2589 int error; 2590 size_t len; 2591 struct pathbuf *linkpb; 2592 struct nameidata nd; 2593 2594 KASSERT(l != NULL || fdat == AT_FDCWD); 2595 2596 path = PNBUF_GET(); 2597 if (seg == UIO_USERSPACE) { 2598 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2599 goto out1; 2600 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2601 goto out1; 2602 } else { 2603 len = strlen(patharg) + 1; 2604 KASSERT(len <= MAXPATHLEN); 2605 memcpy(path, patharg, len); 2606 linkpb = pathbuf_create(link); 2607 if (linkpb == NULL) { 2608 error = ENOMEM; 2609 goto out1; 2610 } 2611 } 2612 ktrkuser("symlink-target", path, len - 1); 2613 2614 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2615 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2616 goto out2; 2617 if (nd.ni_vp) { 2618 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2619 if (nd.ni_dvp == nd.ni_vp) 2620 vrele(nd.ni_dvp); 2621 else 2622 vput(nd.ni_dvp); 2623 vrele(nd.ni_vp); 2624 error = EEXIST; 2625 goto out2; 2626 } 2627 vattr_null(&vattr); 2628 vattr.va_type = VLNK; 2629 /* We will read cwdi->cwdi_cmask unlocked. */ 2630 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2631 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2632 if (error == 0) 2633 vrele(nd.ni_vp); 2634 vput(nd.ni_dvp); 2635 out2: 2636 pathbuf_destroy(linkpb); 2637 out1: 2638 PNBUF_PUT(path); 2639 return (error); 2640 } 2641 2642 /* 2643 * Make a symbolic link. 2644 */ 2645 /* ARGSUSED */ 2646 int 2647 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2648 { 2649 /* { 2650 syscallarg(const char *) path; 2651 syscallarg(const char *) link; 2652 } */ 2653 2654 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2655 UIO_USERSPACE); 2656 } 2657 2658 int 2659 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2660 register_t *retval) 2661 { 2662 /* { 2663 syscallarg(const char *) path1; 2664 syscallarg(int) fd; 2665 syscallarg(const char *) path2; 2666 } */ 2667 2668 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2669 SCARG(uap, path2), UIO_USERSPACE); 2670 } 2671 2672 /* 2673 * Delete a whiteout from the filesystem. 2674 */ 2675 /* ARGSUSED */ 2676 int 2677 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2678 { 2679 /* { 2680 syscallarg(const char *) path; 2681 } */ 2682 int error; 2683 struct pathbuf *pb; 2684 struct nameidata nd; 2685 2686 error = pathbuf_copyin(SCARG(uap, path), &pb); 2687 if (error) { 2688 return error; 2689 } 2690 2691 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2692 error = namei(&nd); 2693 if (error) { 2694 pathbuf_destroy(pb); 2695 return (error); 2696 } 2697 2698 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2699 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2700 if (nd.ni_dvp == nd.ni_vp) 2701 vrele(nd.ni_dvp); 2702 else 2703 vput(nd.ni_dvp); 2704 if (nd.ni_vp) 2705 vrele(nd.ni_vp); 2706 pathbuf_destroy(pb); 2707 return (EEXIST); 2708 } 2709 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2710 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2711 vput(nd.ni_dvp); 2712 pathbuf_destroy(pb); 2713 return (error); 2714 } 2715 2716 /* 2717 * Delete a name from the filesystem. 2718 */ 2719 /* ARGSUSED */ 2720 int 2721 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2722 { 2723 /* { 2724 syscallarg(const char *) path; 2725 } */ 2726 2727 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2728 } 2729 2730 int 2731 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2732 register_t *retval) 2733 { 2734 /* { 2735 syscallarg(int) fd; 2736 syscallarg(const char *) path; 2737 syscallarg(int) flag; 2738 } */ 2739 2740 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2741 SCARG(uap, flag), UIO_USERSPACE); 2742 } 2743 2744 int 2745 do_sys_unlink(const char *arg, enum uio_seg seg) 2746 { 2747 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2748 } 2749 2750 static int 2751 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2752 enum uio_seg seg) 2753 { 2754 struct vnode *vp; 2755 int error; 2756 struct pathbuf *pb; 2757 struct nameidata nd; 2758 const char *pathstring; 2759 2760 KASSERT(l != NULL || fdat == AT_FDCWD); 2761 2762 error = pathbuf_maybe_copyin(arg, seg, &pb); 2763 if (error) { 2764 return error; 2765 } 2766 pathstring = pathbuf_stringcopy_get(pb); 2767 if (pathstring == NULL) { 2768 pathbuf_destroy(pb); 2769 return ENOMEM; 2770 } 2771 2772 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2773 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2774 goto out; 2775 vp = nd.ni_vp; 2776 2777 /* 2778 * The root of a mounted filesystem cannot be deleted. 2779 */ 2780 if ((vp->v_vflag & VV_ROOT) != 0) { 2781 error = EBUSY; 2782 goto abort; 2783 } 2784 2785 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2786 error = EBUSY; 2787 goto abort; 2788 } 2789 2790 /* 2791 * No rmdir "." please. 2792 */ 2793 if (nd.ni_dvp == vp) { 2794 error = EINVAL; 2795 goto abort; 2796 } 2797 2798 /* 2799 * AT_REMOVEDIR is required to remove a directory 2800 */ 2801 if (vp->v_type == VDIR) { 2802 if (!(flags & AT_REMOVEDIR)) { 2803 error = EPERM; 2804 goto abort; 2805 } else { 2806 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2807 vput(nd.ni_dvp); 2808 goto out; 2809 } 2810 } 2811 2812 /* 2813 * Starting here we only deal with non directories. 2814 */ 2815 if (flags & AT_REMOVEDIR) { 2816 error = ENOTDIR; 2817 goto abort; 2818 } 2819 2820 #if NVERIEXEC > 0 2821 /* Handle remove requests for veriexec entries. */ 2822 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2823 goto abort; 2824 } 2825 #endif /* NVERIEXEC > 0 */ 2826 2827 #ifdef FILEASSOC 2828 (void)fileassoc_file_delete(vp); 2829 #endif /* FILEASSOC */ 2830 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2831 vput(nd.ni_dvp); 2832 goto out; 2833 2834 abort: 2835 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2836 if (nd.ni_dvp == vp) 2837 vrele(nd.ni_dvp); 2838 else 2839 vput(nd.ni_dvp); 2840 vput(vp); 2841 2842 out: 2843 pathbuf_stringcopy_put(pb, pathstring); 2844 pathbuf_destroy(pb); 2845 return (error); 2846 } 2847 2848 /* 2849 * Reposition read/write file offset. 2850 */ 2851 int 2852 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2853 { 2854 /* { 2855 syscallarg(int) fd; 2856 syscallarg(int) pad; 2857 syscallarg(off_t) offset; 2858 syscallarg(int) whence; 2859 } */ 2860 kauth_cred_t cred = l->l_cred; 2861 file_t *fp; 2862 struct vnode *vp; 2863 struct vattr vattr; 2864 off_t newoff; 2865 int error, fd; 2866 2867 fd = SCARG(uap, fd); 2868 2869 if ((fp = fd_getfile(fd)) == NULL) 2870 return (EBADF); 2871 2872 vp = fp->f_vnode; 2873 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2874 error = ESPIPE; 2875 goto out; 2876 } 2877 2878 vn_lock(vp, LK_SHARED | LK_RETRY); 2879 2880 switch (SCARG(uap, whence)) { 2881 case SEEK_CUR: 2882 newoff = fp->f_offset + SCARG(uap, offset); 2883 break; 2884 case SEEK_END: 2885 error = VOP_GETATTR(vp, &vattr, cred); 2886 if (error) { 2887 VOP_UNLOCK(vp); 2888 goto out; 2889 } 2890 newoff = SCARG(uap, offset) + vattr.va_size; 2891 break; 2892 case SEEK_SET: 2893 newoff = SCARG(uap, offset); 2894 break; 2895 default: 2896 error = EINVAL; 2897 VOP_UNLOCK(vp); 2898 goto out; 2899 } 2900 VOP_UNLOCK(vp); 2901 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2902 *(off_t *)retval = fp->f_offset = newoff; 2903 } 2904 out: 2905 fd_putfile(fd); 2906 return (error); 2907 } 2908 2909 /* 2910 * Positional read system call. 2911 */ 2912 int 2913 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2914 { 2915 /* { 2916 syscallarg(int) fd; 2917 syscallarg(void *) buf; 2918 syscallarg(size_t) nbyte; 2919 syscallarg(off_t) offset; 2920 } */ 2921 file_t *fp; 2922 struct vnode *vp; 2923 off_t offset; 2924 int error, fd = SCARG(uap, fd); 2925 2926 if ((fp = fd_getfile(fd)) == NULL) 2927 return (EBADF); 2928 2929 if ((fp->f_flag & FREAD) == 0) { 2930 fd_putfile(fd); 2931 return (EBADF); 2932 } 2933 2934 vp = fp->f_vnode; 2935 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2936 error = ESPIPE; 2937 goto out; 2938 } 2939 2940 offset = SCARG(uap, offset); 2941 2942 /* 2943 * XXX This works because no file systems actually 2944 * XXX take any action on the seek operation. 2945 */ 2946 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2947 goto out; 2948 2949 /* dofileread() will unuse the descriptor for us */ 2950 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2951 &offset, 0, retval)); 2952 2953 out: 2954 fd_putfile(fd); 2955 return (error); 2956 } 2957 2958 /* 2959 * Positional scatter read system call. 2960 */ 2961 int 2962 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2963 { 2964 /* { 2965 syscallarg(int) fd; 2966 syscallarg(const struct iovec *) iovp; 2967 syscallarg(int) iovcnt; 2968 syscallarg(off_t) offset; 2969 } */ 2970 off_t offset = SCARG(uap, offset); 2971 2972 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2973 SCARG(uap, iovcnt), &offset, 0, retval); 2974 } 2975 2976 /* 2977 * Positional write system call. 2978 */ 2979 int 2980 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2981 { 2982 /* { 2983 syscallarg(int) fd; 2984 syscallarg(const void *) buf; 2985 syscallarg(size_t) nbyte; 2986 syscallarg(off_t) offset; 2987 } */ 2988 file_t *fp; 2989 struct vnode *vp; 2990 off_t offset; 2991 int error, fd = SCARG(uap, fd); 2992 2993 if ((fp = fd_getfile(fd)) == NULL) 2994 return (EBADF); 2995 2996 if ((fp->f_flag & FWRITE) == 0) { 2997 fd_putfile(fd); 2998 return (EBADF); 2999 } 3000 3001 vp = fp->f_vnode; 3002 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 3003 error = ESPIPE; 3004 goto out; 3005 } 3006 3007 offset = SCARG(uap, offset); 3008 3009 /* 3010 * XXX This works because no file systems actually 3011 * XXX take any action on the seek operation. 3012 */ 3013 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 3014 goto out; 3015 3016 /* dofilewrite() will unuse the descriptor for us */ 3017 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 3018 &offset, 0, retval)); 3019 3020 out: 3021 fd_putfile(fd); 3022 return (error); 3023 } 3024 3025 /* 3026 * Positional gather write system call. 3027 */ 3028 int 3029 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 3030 { 3031 /* { 3032 syscallarg(int) fd; 3033 syscallarg(const struct iovec *) iovp; 3034 syscallarg(int) iovcnt; 3035 syscallarg(off_t) offset; 3036 } */ 3037 off_t offset = SCARG(uap, offset); 3038 3039 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 3040 SCARG(uap, iovcnt), &offset, 0, retval); 3041 } 3042 3043 /* 3044 * Check access permissions. 3045 */ 3046 int 3047 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 3048 { 3049 /* { 3050 syscallarg(const char *) path; 3051 syscallarg(int) flags; 3052 } */ 3053 3054 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 3055 SCARG(uap, flags), 0); 3056 } 3057 3058 int 3059 do_sys_accessat(struct lwp *l, int fdat, const char *path, 3060 int mode, int flags) 3061 { 3062 kauth_cred_t cred; 3063 struct vnode *vp; 3064 int error, nd_flag, vmode; 3065 struct pathbuf *pb; 3066 struct nameidata nd; 3067 3068 CTASSERT(F_OK == 0); 3069 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 3070 /* nonsense mode */ 3071 return EINVAL; 3072 } 3073 3074 nd_flag = FOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT; 3075 if (flags & AT_SYMLINK_NOFOLLOW) 3076 nd_flag &= ~FOLLOW; 3077 3078 error = pathbuf_copyin(path, &pb); 3079 if (error) 3080 return error; 3081 3082 NDINIT(&nd, LOOKUP, nd_flag, pb); 3083 3084 /* Override default credentials */ 3085 cred = kauth_cred_dup(l->l_cred); 3086 if (!(flags & AT_EACCESS)) { 3087 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 3088 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 3089 } 3090 nd.ni_cnd.cn_cred = cred; 3091 3092 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3093 pathbuf_destroy(pb); 3094 goto out; 3095 } 3096 vp = nd.ni_vp; 3097 pathbuf_destroy(pb); 3098 3099 /* Flags == 0 means only check for existence. */ 3100 if (mode) { 3101 vmode = 0; 3102 if (mode & R_OK) 3103 vmode |= VREAD; 3104 if (mode & W_OK) 3105 vmode |= VWRITE; 3106 if (mode & X_OK) 3107 vmode |= VEXEC; 3108 3109 error = VOP_ACCESS(vp, vmode, cred); 3110 if (!error && (vmode & VWRITE)) 3111 error = vn_writechk(vp); 3112 } 3113 vput(vp); 3114 out: 3115 kauth_cred_free(cred); 3116 return (error); 3117 } 3118 3119 int 3120 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3121 register_t *retval) 3122 { 3123 /* { 3124 syscallarg(int) fd; 3125 syscallarg(const char *) path; 3126 syscallarg(int) amode; 3127 syscallarg(int) flag; 3128 } */ 3129 3130 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3131 SCARG(uap, amode), SCARG(uap, flag)); 3132 } 3133 3134 /* 3135 * Common code for all sys_stat functions, including compat versions. 3136 */ 3137 int 3138 do_sys_stat(const char *userpath, unsigned int nd_flag, 3139 struct stat *sb) 3140 { 3141 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3142 } 3143 3144 int 3145 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3146 unsigned int nd_flag, struct stat *sb) 3147 { 3148 int error; 3149 struct pathbuf *pb; 3150 struct nameidata nd; 3151 3152 KASSERT(l != NULL || fdat == AT_FDCWD); 3153 3154 error = pathbuf_copyin(userpath, &pb); 3155 if (error) { 3156 return error; 3157 } 3158 3159 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3160 3161 error = fd_nameiat(l, fdat, &nd); 3162 if (error != 0) { 3163 pathbuf_destroy(pb); 3164 return error; 3165 } 3166 error = vn_stat(nd.ni_vp, sb); 3167 vput(nd.ni_vp); 3168 pathbuf_destroy(pb); 3169 return error; 3170 } 3171 3172 /* 3173 * Get file status; this version follows links. 3174 */ 3175 /* ARGSUSED */ 3176 int 3177 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3178 { 3179 /* { 3180 syscallarg(const char *) path; 3181 syscallarg(struct stat *) ub; 3182 } */ 3183 struct stat sb; 3184 int error; 3185 3186 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3187 if (error) 3188 return error; 3189 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3190 } 3191 3192 /* 3193 * Get file status; this version does not follow links. 3194 */ 3195 /* ARGSUSED */ 3196 int 3197 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3198 { 3199 /* { 3200 syscallarg(const char *) path; 3201 syscallarg(struct stat *) ub; 3202 } */ 3203 struct stat sb; 3204 int error; 3205 3206 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3207 if (error) 3208 return error; 3209 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3210 } 3211 3212 int 3213 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3214 register_t *retval) 3215 { 3216 /* { 3217 syscallarg(int) fd; 3218 syscallarg(const char *) path; 3219 syscallarg(struct stat *) buf; 3220 syscallarg(int) flag; 3221 } */ 3222 unsigned int nd_flag; 3223 struct stat sb; 3224 int error; 3225 3226 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3227 nd_flag = NOFOLLOW; 3228 else 3229 nd_flag = FOLLOW; 3230 3231 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3232 &sb); 3233 if (error) 3234 return error; 3235 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3236 } 3237 3238 static int 3239 kern_pathconf(register_t *retval, const char *path, int name, int flag) 3240 { 3241 int error; 3242 struct pathbuf *pb; 3243 struct nameidata nd; 3244 3245 error = pathbuf_copyin(path, &pb); 3246 if (error) { 3247 return error; 3248 } 3249 NDINIT(&nd, LOOKUP, flag | LOCKLEAF | TRYEMULROOT, pb); 3250 if ((error = namei(&nd)) != 0) { 3251 pathbuf_destroy(pb); 3252 return error; 3253 } 3254 error = VOP_PATHCONF(nd.ni_vp, name, retval); 3255 vput(nd.ni_vp); 3256 pathbuf_destroy(pb); 3257 return error; 3258 } 3259 3260 /* 3261 * Get configurable pathname variables. 3262 */ 3263 /* ARGSUSED */ 3264 int 3265 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, 3266 register_t *retval) 3267 { 3268 /* { 3269 syscallarg(const char *) path; 3270 syscallarg(int) name; 3271 } */ 3272 return kern_pathconf(retval, SCARG(uap, path), SCARG(uap, name), 3273 FOLLOW); 3274 } 3275 3276 /* ARGSUSED */ 3277 int 3278 sys_lpathconf(struct lwp *l, const struct sys_lpathconf_args *uap, 3279 register_t *retval) 3280 { 3281 /* { 3282 syscallarg(const char *) path; 3283 syscallarg(int) name; 3284 } */ 3285 return kern_pathconf(retval, SCARG(uap, path), SCARG(uap, name), 3286 NOFOLLOW); 3287 } 3288 3289 /* 3290 * Return target name of a symbolic link. 3291 */ 3292 /* ARGSUSED */ 3293 int 3294 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3295 register_t *retval) 3296 { 3297 /* { 3298 syscallarg(const char *) path; 3299 syscallarg(char *) buf; 3300 syscallarg(size_t) count; 3301 } */ 3302 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3303 SCARG(uap, buf), SCARG(uap, count), retval); 3304 } 3305 3306 static int 3307 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3308 size_t count, register_t *retval) 3309 { 3310 struct vnode *vp; 3311 struct iovec aiov; 3312 struct uio auio; 3313 int error; 3314 struct pathbuf *pb; 3315 struct nameidata nd; 3316 3317 error = pathbuf_copyin(path, &pb); 3318 if (error) { 3319 return error; 3320 } 3321 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT, pb); 3322 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3323 pathbuf_destroy(pb); 3324 return error; 3325 } 3326 vp = nd.ni_vp; 3327 pathbuf_destroy(pb); 3328 if (vp->v_type != VLNK) 3329 error = EINVAL; 3330 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3331 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3332 aiov.iov_base = buf; 3333 aiov.iov_len = count; 3334 auio.uio_iov = &aiov; 3335 auio.uio_iovcnt = 1; 3336 auio.uio_offset = 0; 3337 auio.uio_rw = UIO_READ; 3338 KASSERT(l == curlwp); 3339 auio.uio_vmspace = l->l_proc->p_vmspace; 3340 auio.uio_resid = count; 3341 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3342 *retval = count - auio.uio_resid; 3343 } 3344 vput(vp); 3345 return (error); 3346 } 3347 3348 int 3349 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3350 register_t *retval) 3351 { 3352 /* { 3353 syscallarg(int) fd; 3354 syscallarg(const char *) path; 3355 syscallarg(char *) buf; 3356 syscallarg(size_t) bufsize; 3357 } */ 3358 3359 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3360 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3361 } 3362 3363 /* 3364 * Change flags of a file given a path name. 3365 */ 3366 /* ARGSUSED */ 3367 int 3368 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3369 { 3370 /* { 3371 syscallarg(const char *) path; 3372 syscallarg(u_long) flags; 3373 } */ 3374 struct vnode *vp; 3375 int error; 3376 3377 error = namei_simple_user(SCARG(uap, path), 3378 NSM_FOLLOW_TRYEMULROOT, &vp); 3379 if (error != 0) 3380 return (error); 3381 error = change_flags(vp, SCARG(uap, flags), l); 3382 vput(vp); 3383 return (error); 3384 } 3385 3386 /* 3387 * Change flags of a file given a file descriptor. 3388 */ 3389 /* ARGSUSED */ 3390 int 3391 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3392 { 3393 /* { 3394 syscallarg(int) fd; 3395 syscallarg(u_long) flags; 3396 } */ 3397 struct vnode *vp; 3398 file_t *fp; 3399 int error; 3400 3401 /* fd_getvnode() will use the descriptor for us */ 3402 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3403 return (error); 3404 vp = fp->f_vnode; 3405 error = change_flags(vp, SCARG(uap, flags), l); 3406 VOP_UNLOCK(vp); 3407 fd_putfile(SCARG(uap, fd)); 3408 return (error); 3409 } 3410 3411 /* 3412 * Change flags of a file given a path name; this version does 3413 * not follow links. 3414 */ 3415 int 3416 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3417 { 3418 /* { 3419 syscallarg(const char *) path; 3420 syscallarg(u_long) flags; 3421 } */ 3422 struct vnode *vp; 3423 int error; 3424 3425 error = namei_simple_user(SCARG(uap, path), 3426 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3427 if (error != 0) 3428 return (error); 3429 error = change_flags(vp, SCARG(uap, flags), l); 3430 vput(vp); 3431 return (error); 3432 } 3433 3434 /* 3435 * Common routine to change flags of a file. 3436 */ 3437 int 3438 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3439 { 3440 struct vattr vattr; 3441 int error; 3442 3443 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3444 3445 vattr_null(&vattr); 3446 vattr.va_flags = flags; 3447 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3448 3449 return (error); 3450 } 3451 3452 /* 3453 * Change mode of a file given path name; this version follows links. 3454 */ 3455 /* ARGSUSED */ 3456 int 3457 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3458 { 3459 /* { 3460 syscallarg(const char *) path; 3461 syscallarg(int) mode; 3462 } */ 3463 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3464 SCARG(uap, mode), 0); 3465 } 3466 3467 int 3468 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3469 { 3470 int error; 3471 struct vnode *vp; 3472 namei_simple_flags_t ns_flag; 3473 3474 if (flags & AT_SYMLINK_NOFOLLOW) 3475 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3476 else 3477 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3478 3479 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3480 if (error != 0) 3481 return error; 3482 3483 error = change_mode(vp, mode, l); 3484 3485 vrele(vp); 3486 3487 return (error); 3488 } 3489 3490 /* 3491 * Change mode of a file given a file descriptor. 3492 */ 3493 /* ARGSUSED */ 3494 int 3495 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3496 { 3497 /* { 3498 syscallarg(int) fd; 3499 syscallarg(int) mode; 3500 } */ 3501 file_t *fp; 3502 int error; 3503 3504 /* fd_getvnode() will use the descriptor for us */ 3505 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3506 return (error); 3507 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3508 fd_putfile(SCARG(uap, fd)); 3509 return (error); 3510 } 3511 3512 int 3513 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3514 register_t *retval) 3515 { 3516 /* { 3517 syscallarg(int) fd; 3518 syscallarg(const char *) path; 3519 syscallarg(int) mode; 3520 syscallarg(int) flag; 3521 } */ 3522 3523 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3524 SCARG(uap, mode), SCARG(uap, flag)); 3525 } 3526 3527 /* 3528 * Change mode of a file given path name; this version does not follow links. 3529 */ 3530 /* ARGSUSED */ 3531 int 3532 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3533 { 3534 /* { 3535 syscallarg(const char *) path; 3536 syscallarg(int) mode; 3537 } */ 3538 int error; 3539 struct vnode *vp; 3540 3541 error = namei_simple_user(SCARG(uap, path), 3542 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3543 if (error != 0) 3544 return (error); 3545 3546 error = change_mode(vp, SCARG(uap, mode), l); 3547 3548 vrele(vp); 3549 return (error); 3550 } 3551 3552 /* 3553 * Common routine to set mode given a vnode. 3554 */ 3555 static int 3556 change_mode(struct vnode *vp, int mode, struct lwp *l) 3557 { 3558 struct vattr vattr; 3559 int error; 3560 3561 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3562 vattr_null(&vattr); 3563 vattr.va_mode = mode & ALLPERMS; 3564 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3565 VOP_UNLOCK(vp); 3566 return (error); 3567 } 3568 3569 /* 3570 * Set ownership given a path name; this version follows links. 3571 */ 3572 /* ARGSUSED */ 3573 int 3574 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3575 { 3576 /* { 3577 syscallarg(const char *) path; 3578 syscallarg(uid_t) uid; 3579 syscallarg(gid_t) gid; 3580 } */ 3581 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3582 SCARG(uap, gid), 0); 3583 } 3584 3585 int 3586 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3587 gid_t gid, int flags) 3588 { 3589 int error; 3590 struct vnode *vp; 3591 namei_simple_flags_t ns_flag; 3592 3593 if (flags & AT_SYMLINK_NOFOLLOW) 3594 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3595 else 3596 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3597 3598 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3599 if (error != 0) 3600 return error; 3601 3602 error = change_owner(vp, uid, gid, l, 0); 3603 3604 vrele(vp); 3605 3606 return (error); 3607 } 3608 3609 /* 3610 * Set ownership given a path name; this version follows links. 3611 * Provides POSIX semantics. 3612 */ 3613 /* ARGSUSED */ 3614 int 3615 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3616 { 3617 /* { 3618 syscallarg(const char *) path; 3619 syscallarg(uid_t) uid; 3620 syscallarg(gid_t) gid; 3621 } */ 3622 int error; 3623 struct vnode *vp; 3624 3625 error = namei_simple_user(SCARG(uap, path), 3626 NSM_FOLLOW_TRYEMULROOT, &vp); 3627 if (error != 0) 3628 return (error); 3629 3630 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3631 3632 vrele(vp); 3633 return (error); 3634 } 3635 3636 /* 3637 * Set ownership given a file descriptor. 3638 */ 3639 /* ARGSUSED */ 3640 int 3641 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3642 { 3643 /* { 3644 syscallarg(int) fd; 3645 syscallarg(uid_t) uid; 3646 syscallarg(gid_t) gid; 3647 } */ 3648 int error; 3649 file_t *fp; 3650 3651 /* fd_getvnode() will use the descriptor for us */ 3652 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3653 return (error); 3654 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3655 l, 0); 3656 fd_putfile(SCARG(uap, fd)); 3657 return (error); 3658 } 3659 3660 int 3661 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3662 register_t *retval) 3663 { 3664 /* { 3665 syscallarg(int) fd; 3666 syscallarg(const char *) path; 3667 syscallarg(uid_t) owner; 3668 syscallarg(gid_t) group; 3669 syscallarg(int) flag; 3670 } */ 3671 3672 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3673 SCARG(uap, owner), SCARG(uap, group), 3674 SCARG(uap, flag)); 3675 } 3676 3677 /* 3678 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3679 */ 3680 /* ARGSUSED */ 3681 int 3682 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3683 { 3684 /* { 3685 syscallarg(int) fd; 3686 syscallarg(uid_t) uid; 3687 syscallarg(gid_t) gid; 3688 } */ 3689 int error; 3690 file_t *fp; 3691 3692 /* fd_getvnode() will use the descriptor for us */ 3693 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3694 return (error); 3695 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3696 l, 1); 3697 fd_putfile(SCARG(uap, fd)); 3698 return (error); 3699 } 3700 3701 /* 3702 * Set ownership given a path name; this version does not follow links. 3703 */ 3704 /* ARGSUSED */ 3705 int 3706 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3707 { 3708 /* { 3709 syscallarg(const char *) path; 3710 syscallarg(uid_t) uid; 3711 syscallarg(gid_t) gid; 3712 } */ 3713 int error; 3714 struct vnode *vp; 3715 3716 error = namei_simple_user(SCARG(uap, path), 3717 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3718 if (error != 0) 3719 return (error); 3720 3721 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3722 3723 vrele(vp); 3724 return (error); 3725 } 3726 3727 /* 3728 * Set ownership given a path name; this version does not follow links. 3729 * Provides POSIX/XPG semantics. 3730 */ 3731 /* ARGSUSED */ 3732 int 3733 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3734 { 3735 /* { 3736 syscallarg(const char *) path; 3737 syscallarg(uid_t) uid; 3738 syscallarg(gid_t) gid; 3739 } */ 3740 int error; 3741 struct vnode *vp; 3742 3743 error = namei_simple_user(SCARG(uap, path), 3744 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3745 if (error != 0) 3746 return (error); 3747 3748 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3749 3750 vrele(vp); 3751 return (error); 3752 } 3753 3754 /* 3755 * Common routine to set ownership given a vnode. 3756 */ 3757 static int 3758 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3759 int posix_semantics) 3760 { 3761 struct vattr vattr; 3762 mode_t newmode; 3763 int error; 3764 3765 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3766 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3767 goto out; 3768 3769 #define CHANGED(x) ((int)(x) != -1) 3770 newmode = vattr.va_mode; 3771 if (posix_semantics) { 3772 /* 3773 * POSIX/XPG semantics: if the caller is not the super-user, 3774 * clear set-user-id and set-group-id bits. Both POSIX and 3775 * the XPG consider the behaviour for calls by the super-user 3776 * implementation-defined; we leave the set-user-id and set- 3777 * group-id settings intact in that case. 3778 */ 3779 if (vattr.va_mode & S_ISUID) { 3780 if (kauth_authorize_vnode(l->l_cred, 3781 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3782 newmode &= ~S_ISUID; 3783 } 3784 if (vattr.va_mode & S_ISGID) { 3785 if (kauth_authorize_vnode(l->l_cred, 3786 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3787 newmode &= ~S_ISGID; 3788 } 3789 } else { 3790 /* 3791 * NetBSD semantics: when changing owner and/or group, 3792 * clear the respective bit(s). 3793 */ 3794 if (CHANGED(uid)) 3795 newmode &= ~S_ISUID; 3796 if (CHANGED(gid)) 3797 newmode &= ~S_ISGID; 3798 } 3799 /* Update va_mode iff altered. */ 3800 if (vattr.va_mode == newmode) 3801 newmode = VNOVAL; 3802 3803 vattr_null(&vattr); 3804 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3805 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3806 vattr.va_mode = newmode; 3807 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3808 #undef CHANGED 3809 3810 out: 3811 VOP_UNLOCK(vp); 3812 return (error); 3813 } 3814 3815 /* 3816 * Set the access and modification times given a path name; this 3817 * version follows links. 3818 */ 3819 /* ARGSUSED */ 3820 int 3821 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3822 register_t *retval) 3823 { 3824 /* { 3825 syscallarg(const char *) path; 3826 syscallarg(const struct timeval *) tptr; 3827 } */ 3828 3829 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3830 SCARG(uap, tptr), UIO_USERSPACE); 3831 } 3832 3833 /* 3834 * Set the access and modification times given a file descriptor. 3835 */ 3836 /* ARGSUSED */ 3837 int 3838 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3839 register_t *retval) 3840 { 3841 /* { 3842 syscallarg(int) fd; 3843 syscallarg(const struct timeval *) tptr; 3844 } */ 3845 int error; 3846 file_t *fp; 3847 3848 /* fd_getvnode() will use the descriptor for us */ 3849 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3850 return (error); 3851 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3852 UIO_USERSPACE); 3853 fd_putfile(SCARG(uap, fd)); 3854 return (error); 3855 } 3856 3857 int 3858 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3859 register_t *retval) 3860 { 3861 /* { 3862 syscallarg(int) fd; 3863 syscallarg(const struct timespec *) tptr; 3864 } */ 3865 int error; 3866 file_t *fp; 3867 3868 /* fd_getvnode() will use the descriptor for us */ 3869 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3870 return (error); 3871 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3872 SCARG(uap, tptr), UIO_USERSPACE); 3873 fd_putfile(SCARG(uap, fd)); 3874 return (error); 3875 } 3876 3877 /* 3878 * Set the access and modification times given a path name; this 3879 * version does not follow links. 3880 */ 3881 int 3882 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3883 register_t *retval) 3884 { 3885 /* { 3886 syscallarg(const char *) path; 3887 syscallarg(const struct timeval *) tptr; 3888 } */ 3889 3890 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3891 SCARG(uap, tptr), UIO_USERSPACE); 3892 } 3893 3894 int 3895 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3896 register_t *retval) 3897 { 3898 /* { 3899 syscallarg(int) fd; 3900 syscallarg(const char *) path; 3901 syscallarg(const struct timespec *) tptr; 3902 syscallarg(int) flag; 3903 } */ 3904 int follow; 3905 const struct timespec *tptr; 3906 int error; 3907 3908 tptr = SCARG(uap, tptr); 3909 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3910 3911 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3912 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3913 3914 return error; 3915 } 3916 3917 /* 3918 * Common routine to set access and modification times given a vnode. 3919 */ 3920 int 3921 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3922 const struct timespec *tptr, enum uio_seg seg) 3923 { 3924 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3925 } 3926 3927 int 3928 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 3929 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 3930 { 3931 struct vattr vattr; 3932 int error, dorele = 0; 3933 namei_simple_flags_t sflags; 3934 bool vanull, setbirthtime; 3935 struct timespec ts[2]; 3936 3937 KASSERT(l != NULL || fdat == AT_FDCWD); 3938 3939 /* 3940 * I have checked all callers and they pass either FOLLOW, 3941 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3942 * is 0. More to the point, they don't pass anything else. 3943 * Let's keep it that way at least until the namei interfaces 3944 * are fully sanitized. 3945 */ 3946 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3947 sflags = (flag == FOLLOW) ? 3948 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3949 3950 if (tptr == NULL) { 3951 vanull = true; 3952 nanotime(&ts[0]); 3953 ts[1] = ts[0]; 3954 } else { 3955 vanull = false; 3956 if (seg != UIO_SYSSPACE) { 3957 error = copyin(tptr, ts, sizeof (ts)); 3958 if (error != 0) 3959 return error; 3960 } else { 3961 ts[0] = tptr[0]; 3962 ts[1] = tptr[1]; 3963 } 3964 } 3965 3966 if (ts[0].tv_nsec == UTIME_NOW) { 3967 nanotime(&ts[0]); 3968 if (ts[1].tv_nsec == UTIME_NOW) { 3969 vanull = true; 3970 ts[1] = ts[0]; 3971 } 3972 } else if (ts[1].tv_nsec == UTIME_NOW) 3973 nanotime(&ts[1]); 3974 3975 if (vp == NULL) { 3976 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3977 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 3978 if (error != 0) 3979 return error; 3980 dorele = 1; 3981 } 3982 3983 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3984 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3985 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3986 vattr_null(&vattr); 3987 3988 if (ts[0].tv_nsec != UTIME_OMIT) 3989 vattr.va_atime = ts[0]; 3990 3991 if (ts[1].tv_nsec != UTIME_OMIT) { 3992 vattr.va_mtime = ts[1]; 3993 if (setbirthtime) 3994 vattr.va_birthtime = ts[1]; 3995 } 3996 3997 if (vanull) 3998 vattr.va_vaflags |= VA_UTIMES_NULL; 3999 error = VOP_SETATTR(vp, &vattr, l->l_cred); 4000 VOP_UNLOCK(vp); 4001 4002 if (dorele != 0) 4003 vrele(vp); 4004 4005 return error; 4006 } 4007 4008 int 4009 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 4010 const struct timeval *tptr, enum uio_seg seg) 4011 { 4012 struct timespec ts[2]; 4013 struct timespec *tsptr = NULL; 4014 int error; 4015 4016 if (tptr != NULL) { 4017 struct timeval tv[2]; 4018 4019 if (seg != UIO_SYSSPACE) { 4020 error = copyin(tptr, tv, sizeof(tv)); 4021 if (error != 0) 4022 return error; 4023 tptr = tv; 4024 } 4025 4026 if ((tptr[0].tv_usec == UTIME_NOW) || 4027 (tptr[0].tv_usec == UTIME_OMIT)) 4028 ts[0].tv_nsec = tptr[0].tv_usec; 4029 else { 4030 if (tptr[0].tv_usec < 0 || tptr[0].tv_usec >= 1000000) 4031 return EINVAL; 4032 4033 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 4034 } 4035 4036 if ((tptr[1].tv_usec == UTIME_NOW) || 4037 (tptr[1].tv_usec == UTIME_OMIT)) 4038 ts[1].tv_nsec = tptr[1].tv_usec; 4039 else { 4040 if (tptr[1].tv_usec < 0 || tptr[1].tv_usec >= 1000000) 4041 return EINVAL; 4042 4043 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 4044 } 4045 4046 tsptr = &ts[0]; 4047 } 4048 4049 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 4050 } 4051 4052 /* 4053 * Truncate a file given its path name. 4054 */ 4055 /* ARGSUSED */ 4056 int 4057 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 4058 { 4059 /* { 4060 syscallarg(const char *) path; 4061 syscallarg(int) pad; 4062 syscallarg(off_t) length; 4063 } */ 4064 struct vnode *vp; 4065 struct vattr vattr; 4066 int error; 4067 4068 if (SCARG(uap, length) < 0) 4069 return EINVAL; 4070 4071 error = namei_simple_user(SCARG(uap, path), 4072 NSM_FOLLOW_TRYEMULROOT, &vp); 4073 if (error != 0) 4074 return (error); 4075 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4076 if (vp->v_type == VDIR) 4077 error = EISDIR; 4078 else if ((error = vn_writechk(vp)) == 0 && 4079 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 4080 vattr_null(&vattr); 4081 vattr.va_size = SCARG(uap, length); 4082 error = VOP_SETATTR(vp, &vattr, l->l_cred); 4083 } 4084 vput(vp); 4085 return (error); 4086 } 4087 4088 /* 4089 * Truncate a file given a file descriptor. 4090 */ 4091 /* ARGSUSED */ 4092 int 4093 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 4094 { 4095 /* { 4096 syscallarg(int) fd; 4097 syscallarg(int) pad; 4098 syscallarg(off_t) length; 4099 } */ 4100 struct vattr vattr; 4101 struct vnode *vp; 4102 file_t *fp; 4103 int error; 4104 4105 if (SCARG(uap, length) < 0) 4106 return EINVAL; 4107 4108 /* fd_getvnode() will use the descriptor for us */ 4109 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4110 return (error); 4111 if ((fp->f_flag & FWRITE) == 0) { 4112 error = EINVAL; 4113 goto out; 4114 } 4115 vp = fp->f_vnode; 4116 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4117 if (vp->v_type == VDIR) 4118 error = EISDIR; 4119 else if ((error = vn_writechk(vp)) == 0) { 4120 vattr_null(&vattr); 4121 vattr.va_size = SCARG(uap, length); 4122 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 4123 } 4124 VOP_UNLOCK(vp); 4125 out: 4126 fd_putfile(SCARG(uap, fd)); 4127 return (error); 4128 } 4129 4130 /* 4131 * Sync an open file. 4132 */ 4133 /* ARGSUSED */ 4134 int 4135 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 4136 { 4137 /* { 4138 syscallarg(int) fd; 4139 } */ 4140 struct vnode *vp; 4141 file_t *fp; 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 vp = fp->f_vnode; 4148 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4149 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4150 VOP_UNLOCK(vp); 4151 fd_putfile(SCARG(uap, fd)); 4152 return (error); 4153 } 4154 4155 /* 4156 * Sync a range of file data. API modeled after that found in AIX. 4157 * 4158 * FDATASYNC indicates that we need only save enough metadata to be able 4159 * to re-read the written data. 4160 */ 4161 /* ARGSUSED */ 4162 int 4163 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4164 { 4165 /* { 4166 syscallarg(int) fd; 4167 syscallarg(int) flags; 4168 syscallarg(off_t) start; 4169 syscallarg(off_t) length; 4170 } */ 4171 struct vnode *vp; 4172 file_t *fp; 4173 int flags, nflags; 4174 off_t s, e, len; 4175 int error; 4176 4177 /* fd_getvnode() will use the descriptor for us */ 4178 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4179 return (error); 4180 4181 if ((fp->f_flag & FWRITE) == 0) { 4182 error = EBADF; 4183 goto out; 4184 } 4185 4186 flags = SCARG(uap, flags); 4187 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4188 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4189 error = EINVAL; 4190 goto out; 4191 } 4192 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4193 if (flags & FDATASYNC) 4194 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4195 else 4196 nflags = FSYNC_WAIT; 4197 if (flags & FDISKSYNC) 4198 nflags |= FSYNC_CACHE; 4199 4200 len = SCARG(uap, length); 4201 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4202 if (len) { 4203 s = SCARG(uap, start); 4204 if (s < 0 || len < 0 || len > OFF_T_MAX - s) { 4205 error = EINVAL; 4206 goto out; 4207 } 4208 e = s + len; 4209 KASSERT(s <= e); 4210 } else { 4211 e = 0; 4212 s = 0; 4213 } 4214 4215 vp = fp->f_vnode; 4216 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4217 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4218 VOP_UNLOCK(vp); 4219 out: 4220 fd_putfile(SCARG(uap, fd)); 4221 return (error); 4222 } 4223 4224 /* 4225 * Sync the data of an open file. 4226 */ 4227 /* ARGSUSED */ 4228 int 4229 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4230 { 4231 /* { 4232 syscallarg(int) fd; 4233 } */ 4234 struct vnode *vp; 4235 file_t *fp; 4236 int error; 4237 4238 /* fd_getvnode() will use the descriptor for us */ 4239 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4240 return (error); 4241 vp = fp->f_vnode; 4242 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4243 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4244 VOP_UNLOCK(vp); 4245 fd_putfile(SCARG(uap, fd)); 4246 return (error); 4247 } 4248 4249 /* 4250 * Rename files, (standard) BSD semantics frontend. 4251 */ 4252 /* ARGSUSED */ 4253 int 4254 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4255 { 4256 /* { 4257 syscallarg(const char *) from; 4258 syscallarg(const char *) to; 4259 } */ 4260 4261 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4262 SCARG(uap, to), UIO_USERSPACE, 0)); 4263 } 4264 4265 int 4266 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4267 register_t *retval) 4268 { 4269 /* { 4270 syscallarg(int) fromfd; 4271 syscallarg(const char *) from; 4272 syscallarg(int) tofd; 4273 syscallarg(const char *) to; 4274 } */ 4275 4276 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4277 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4278 } 4279 4280 /* 4281 * Rename files, POSIX semantics frontend. 4282 */ 4283 /* ARGSUSED */ 4284 int 4285 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4286 { 4287 /* { 4288 syscallarg(const char *) from; 4289 syscallarg(const char *) to; 4290 } */ 4291 4292 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4293 SCARG(uap, to), UIO_USERSPACE, 1)); 4294 } 4295 4296 /* 4297 * Rename files. Source and destination must either both be directories, 4298 * or both not be directories. If target is a directory, it must be empty. 4299 * If `from' and `to' refer to the same object, the value of the `retain' 4300 * argument is used to determine whether `from' will be 4301 * 4302 * (retain == 0) deleted unless `from' and `to' refer to the same 4303 * object in the file system's name space (BSD). 4304 * (retain == 1) always retained (POSIX). 4305 * 4306 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4307 */ 4308 int 4309 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4310 { 4311 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4312 } 4313 4314 static int 4315 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4316 const char *to, enum uio_seg seg, int retain) 4317 { 4318 struct pathbuf *fpb, *tpb; 4319 struct nameidata fnd, tnd; 4320 struct vnode *fdvp, *fvp; 4321 struct vnode *tdvp, *tvp; 4322 struct mount *mp, *tmp; 4323 int error; 4324 4325 KASSERT(l != NULL || (fromfd == AT_FDCWD && tofd == AT_FDCWD)); 4326 4327 error = pathbuf_maybe_copyin(from, seg, &fpb); 4328 if (error) 4329 goto out0; 4330 KASSERT(fpb != NULL); 4331 4332 error = pathbuf_maybe_copyin(to, seg, &tpb); 4333 if (error) 4334 goto out1; 4335 KASSERT(tpb != NULL); 4336 4337 /* 4338 * Lookup from. 4339 * 4340 * XXX LOCKPARENT is wrong because we don't actually want it 4341 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4342 * insane, so for the time being we need to leave it like this. 4343 */ 4344 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4345 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4346 goto out2; 4347 4348 /* 4349 * Pull out the important results of the lookup, fdvp and fvp. 4350 * Of course, fvp is bogus because we're about to unlock fdvp. 4351 */ 4352 fdvp = fnd.ni_dvp; 4353 fvp = fnd.ni_vp; 4354 mp = fdvp->v_mount; 4355 KASSERT(fdvp != NULL); 4356 KASSERT(fvp != NULL); 4357 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4358 /* 4359 * Bracket the operation with fstrans_start()/fstrans_done(). 4360 * 4361 * Inside the bracket this file system cannot be unmounted so 4362 * a vnode on this file system cannot change its v_mount. 4363 * A vnode on another file system may still change to dead mount. 4364 */ 4365 fstrans_start(mp); 4366 4367 /* 4368 * Make sure neither fdvp nor fvp is locked. 4369 */ 4370 if (fdvp != fvp) 4371 VOP_UNLOCK(fdvp); 4372 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4373 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4374 4375 /* 4376 * Reject renaming `.' and `..'. Can't do this until after 4377 * namei because we need namei's parsing to find the final 4378 * component name. (namei should just leave us with the final 4379 * component name and not look it up itself, but anyway...) 4380 * 4381 * This was here before because we used to relookup from 4382 * instead of to and relookup requires the caller to check 4383 * this, but now file systems may depend on this check, so we 4384 * must retain it until the file systems are all rototilled. 4385 */ 4386 if (((fnd.ni_cnd.cn_namelen == 1) && 4387 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4388 ((fnd.ni_cnd.cn_namelen == 2) && 4389 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4390 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4391 error = EINVAL; /* XXX EISDIR? */ 4392 goto abort0; 4393 } 4394 4395 /* 4396 * Lookup to. 4397 * 4398 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4399 * fvp here to decide whether to add CREATEDIR is a load of 4400 * bollocks because fvp might be the wrong node by now, since 4401 * fdvp is unlocked. 4402 * 4403 * XXX Why not pass CREATEDIR always? 4404 */ 4405 NDINIT(&tnd, RENAME, 4406 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4407 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4408 tpb); 4409 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4410 goto abort0; 4411 4412 /* 4413 * Pull out the important results of the lookup, tdvp and tvp. 4414 * Of course, tvp is bogus because we're about to unlock tdvp. 4415 */ 4416 tdvp = tnd.ni_dvp; 4417 tvp = tnd.ni_vp; 4418 KASSERT(tdvp != NULL); 4419 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4420 4421 if (fvp->v_type == VDIR) 4422 tnd.ni_cnd.cn_flags |= WILLBEDIR; 4423 /* 4424 * Make sure neither tdvp nor tvp is locked. 4425 */ 4426 if (tdvp != tvp) 4427 VOP_UNLOCK(tdvp); 4428 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4429 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4430 4431 /* 4432 * Reject renaming onto `.' or `..'. relookup is unhappy with 4433 * these, which is why we must do this here. Once upon a time 4434 * we relooked up from instead of to, and consequently didn't 4435 * need this check, but now that we relookup to instead of 4436 * from, we need this; and we shall need it forever forward 4437 * until the VOP_RENAME protocol changes, because file systems 4438 * will no doubt begin to depend on this check. 4439 */ 4440 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4441 error = EISDIR; 4442 goto abort1; 4443 } 4444 if ((tnd.ni_cnd.cn_namelen == 2) && 4445 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4446 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4447 error = EINVAL; 4448 goto abort1; 4449 } 4450 4451 /* 4452 * Make sure the mount points match. Although we don't hold 4453 * any vnode locks, the v_mount on fdvp file system are stable. 4454 * 4455 * Unmounting another file system at an inopportune moment may 4456 * cause tdvp to disappear and change its v_mount to dead. 4457 * 4458 * So in either case different v_mount means cross-device rename. 4459 */ 4460 KASSERT(mp != NULL); 4461 tmp = tdvp->v_mount; 4462 4463 if (mp != tmp) { 4464 error = EXDEV; 4465 goto abort1; 4466 } 4467 4468 /* 4469 * Take the vfs rename lock to avoid cross-directory screw cases. 4470 * Nothing is locked currently, so taking this lock is safe. 4471 */ 4472 error = VFS_RENAMELOCK_ENTER(mp); 4473 if (error) 4474 goto abort1; 4475 4476 /* 4477 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4478 * and nothing is locked except for the vfs rename lock. 4479 * 4480 * The next step is a little rain dance to conform to the 4481 * insane lock protocol, even though it does nothing to ward 4482 * off race conditions. 4483 * 4484 * We need tdvp and tvp to be locked. However, because we have 4485 * unlocked tdvp in order to hold no locks while we take the 4486 * vfs rename lock, tvp may be wrong here, and we can't safely 4487 * lock it even if the sensible file systems will just unlock 4488 * it straight away. Consequently, we must lock tdvp and then 4489 * relookup tvp to get it locked. 4490 * 4491 * Finally, because the VOP_RENAME protocol is brain-damaged 4492 * and various file systems insanely depend on the semantics of 4493 * this brain damage, the lookup of to must be the last lookup 4494 * before VOP_RENAME. 4495 */ 4496 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4497 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4498 if (error) 4499 goto abort2; 4500 4501 /* 4502 * Drop the old tvp and pick up the new one -- which might be 4503 * the same, but that doesn't matter to us. After this, tdvp 4504 * and tvp should both be locked. 4505 */ 4506 if (tvp != NULL) 4507 vrele(tvp); 4508 tvp = tnd.ni_vp; 4509 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4510 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4511 4512 /* 4513 * The old do_sys_rename had various consistency checks here 4514 * involving fvp and tvp. fvp is bogus already here, and tvp 4515 * will become bogus soon in any sensible file system, so the 4516 * only purpose in putting these checks here is to give lip 4517 * service to these screw cases and to acknowledge that they 4518 * exist, not actually to handle them, but here you go 4519 * anyway... 4520 */ 4521 4522 /* 4523 * Acknowledge that directories and non-directories aren't 4524 * suposed to mix. 4525 */ 4526 if (tvp != NULL) { 4527 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4528 error = ENOTDIR; 4529 goto abort3; 4530 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4531 error = EISDIR; 4532 goto abort3; 4533 } 4534 } 4535 4536 /* 4537 * Acknowledge some random screw case, among the dozens that 4538 * might arise. 4539 */ 4540 if (fvp == tdvp) { 4541 error = EINVAL; 4542 goto abort3; 4543 } 4544 4545 /* 4546 * Acknowledge that POSIX has a wacky screw case. 4547 * 4548 * XXX Eventually the retain flag needs to be passed on to 4549 * VOP_RENAME. 4550 */ 4551 if (fvp == tvp) { 4552 if (retain) { 4553 error = 0; 4554 goto abort3; 4555 } else if ((fdvp == tdvp) && 4556 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4557 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4558 fnd.ni_cnd.cn_namelen))) { 4559 error = 0; 4560 goto abort3; 4561 } 4562 } 4563 4564 /* 4565 * Make sure veriexec can screw us up. (But a race can screw 4566 * up veriexec, of course -- remember, fvp and (soon) tvp are 4567 * bogus.) 4568 */ 4569 #if NVERIEXEC > 0 4570 { 4571 char *f1, *f2; 4572 size_t f1_len; 4573 size_t f2_len; 4574 4575 f1_len = fnd.ni_cnd.cn_namelen + 1; 4576 f1 = kmem_alloc(f1_len, KM_SLEEP); 4577 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4578 4579 f2_len = tnd.ni_cnd.cn_namelen + 1; 4580 f2 = kmem_alloc(f2_len, KM_SLEEP); 4581 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4582 4583 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4584 4585 kmem_free(f1, f1_len); 4586 kmem_free(f2, f2_len); 4587 4588 if (error) 4589 goto abort3; 4590 } 4591 #endif /* NVERIEXEC > 0 */ 4592 4593 /* 4594 * All ready. Incant the rename vop. 4595 */ 4596 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4597 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4598 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4599 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4600 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4601 4602 /* 4603 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4604 * tdvp and tvp. But we can't assert any of that. 4605 */ 4606 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4607 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4608 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4609 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4610 4611 /* 4612 * So all we have left to do is to drop the rename lock and 4613 * destroy the pathbufs. 4614 */ 4615 VFS_RENAMELOCK_EXIT(mp); 4616 fstrans_done(mp); 4617 goto out2; 4618 4619 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4620 VOP_UNLOCK(tvp); 4621 abort2: VOP_UNLOCK(tdvp); 4622 VFS_RENAMELOCK_EXIT(mp); 4623 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4624 vrele(tdvp); 4625 if (tvp != NULL) 4626 vrele(tvp); 4627 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4628 vrele(fdvp); 4629 vrele(fvp); 4630 fstrans_done(mp); 4631 out2: pathbuf_destroy(tpb); 4632 out1: pathbuf_destroy(fpb); 4633 out0: return error; 4634 } 4635 4636 /* 4637 * Make a directory file. 4638 */ 4639 /* ARGSUSED */ 4640 int 4641 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4642 { 4643 /* { 4644 syscallarg(const char *) path; 4645 syscallarg(int) mode; 4646 } */ 4647 4648 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4649 SCARG(uap, mode), UIO_USERSPACE); 4650 } 4651 4652 int 4653 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4654 register_t *retval) 4655 { 4656 /* { 4657 syscallarg(int) fd; 4658 syscallarg(const char *) path; 4659 syscallarg(int) mode; 4660 } */ 4661 4662 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4663 SCARG(uap, mode), UIO_USERSPACE); 4664 } 4665 4666 4667 int 4668 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4669 { 4670 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, seg); 4671 } 4672 4673 static int 4674 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4675 enum uio_seg seg) 4676 { 4677 struct proc *p = curlwp->l_proc; 4678 struct vnode *vp; 4679 struct vattr vattr; 4680 int error; 4681 struct pathbuf *pb; 4682 struct nameidata nd; 4683 4684 KASSERT(l != NULL || fdat == AT_FDCWD); 4685 4686 /* XXX bollocks, should pass in a pathbuf */ 4687 error = pathbuf_maybe_copyin(path, seg, &pb); 4688 if (error) { 4689 return error; 4690 } 4691 4692 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4693 4694 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4695 pathbuf_destroy(pb); 4696 return (error); 4697 } 4698 vp = nd.ni_vp; 4699 if (vp != NULL) { 4700 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4701 if (nd.ni_dvp == vp) 4702 vrele(nd.ni_dvp); 4703 else 4704 vput(nd.ni_dvp); 4705 vrele(vp); 4706 pathbuf_destroy(pb); 4707 return (EEXIST); 4708 } 4709 vattr_null(&vattr); 4710 vattr.va_type = VDIR; 4711 /* We will read cwdi->cwdi_cmask unlocked. */ 4712 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4713 nd.ni_cnd.cn_flags |= WILLBEDIR; 4714 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4715 if (!error) 4716 vrele(nd.ni_vp); 4717 vput(nd.ni_dvp); 4718 pathbuf_destroy(pb); 4719 return (error); 4720 } 4721 4722 /* 4723 * Remove a directory file. 4724 */ 4725 /* ARGSUSED */ 4726 int 4727 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4728 { 4729 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4730 AT_REMOVEDIR, UIO_USERSPACE); 4731 } 4732 4733 /* 4734 * Read a block of directory entries in a file system independent format. 4735 */ 4736 int 4737 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4738 { 4739 /* { 4740 syscallarg(int) fd; 4741 syscallarg(char *) buf; 4742 syscallarg(size_t) count; 4743 } */ 4744 file_t *fp; 4745 int error, done; 4746 4747 /* fd_getvnode() will use the descriptor for us */ 4748 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4749 return (error); 4750 if ((fp->f_flag & FREAD) == 0) { 4751 error = EBADF; 4752 goto out; 4753 } 4754 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4755 SCARG(uap, count), &done, l, 0, 0); 4756 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4757 *retval = done; 4758 out: 4759 fd_putfile(SCARG(uap, fd)); 4760 return (error); 4761 } 4762 4763 /* 4764 * Set the mode mask for creation of filesystem nodes. 4765 */ 4766 int 4767 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4768 { 4769 /* { 4770 syscallarg(mode_t) newmask; 4771 } */ 4772 4773 /* 4774 * cwdi->cwdi_cmask will be read unlocked elsewhere, and no kind of 4775 * serialization with those reads is required. It's important to 4776 * return a coherent answer for the caller of umask() though, and 4777 * the atomic operation accomplishes that. 4778 */ 4779 *retval = atomic_swap_uint(&curproc->p_cwdi->cwdi_cmask, 4780 SCARG(uap, newmask) & ALLPERMS); 4781 4782 return (0); 4783 } 4784 4785 int 4786 dorevoke(struct vnode *vp, kauth_cred_t cred) 4787 { 4788 struct vattr vattr; 4789 int error, fs_decision; 4790 4791 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4792 error = VOP_GETATTR(vp, &vattr, cred); 4793 VOP_UNLOCK(vp); 4794 if (error != 0) 4795 return error; 4796 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4797 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4798 fs_decision); 4799 if (!error) 4800 VOP_REVOKE(vp, REVOKEALL); 4801 return (error); 4802 } 4803 4804 /* 4805 * Void all references to file by ripping underlying filesystem 4806 * away from vnode. 4807 */ 4808 /* ARGSUSED */ 4809 int 4810 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4811 { 4812 /* { 4813 syscallarg(const char *) path; 4814 } */ 4815 struct vnode *vp; 4816 int error; 4817 4818 error = namei_simple_user(SCARG(uap, path), 4819 NSM_FOLLOW_TRYEMULROOT, &vp); 4820 if (error != 0) 4821 return (error); 4822 error = dorevoke(vp, l->l_cred); 4823 vrele(vp); 4824 return (error); 4825 } 4826 4827 /* 4828 * Allocate backing store for a file, filling a hole without having to 4829 * explicitly write anything out. 4830 */ 4831 /* ARGSUSED */ 4832 int 4833 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4834 register_t *retval) 4835 { 4836 /* { 4837 syscallarg(int) fd; 4838 syscallarg(off_t) pos; 4839 syscallarg(off_t) len; 4840 } */ 4841 int fd; 4842 off_t pos, len; 4843 struct file *fp; 4844 struct vnode *vp; 4845 int error; 4846 4847 fd = SCARG(uap, fd); 4848 pos = SCARG(uap, pos); 4849 len = SCARG(uap, len); 4850 4851 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4852 *retval = EINVAL; 4853 return 0; 4854 } 4855 4856 error = fd_getvnode(fd, &fp); 4857 if (error) { 4858 *retval = error; 4859 return 0; 4860 } 4861 if ((fp->f_flag & FWRITE) == 0) { 4862 error = EBADF; 4863 goto fail; 4864 } 4865 vp = fp->f_vnode; 4866 4867 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4868 if (vp->v_type == VDIR) { 4869 error = EISDIR; 4870 } else { 4871 error = VOP_FALLOCATE(vp, pos, len); 4872 } 4873 VOP_UNLOCK(vp); 4874 4875 fail: 4876 fd_putfile(fd); 4877 *retval = error; 4878 return 0; 4879 } 4880 4881 /* 4882 * Deallocate backing store for a file, creating a hole. Also used for 4883 * invoking TRIM on disks. 4884 */ 4885 /* ARGSUSED */ 4886 int 4887 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4888 register_t *retval) 4889 { 4890 /* { 4891 syscallarg(int) fd; 4892 syscallarg(off_t) pos; 4893 syscallarg(off_t) len; 4894 } */ 4895 int fd; 4896 off_t pos, len; 4897 struct file *fp; 4898 struct vnode *vp; 4899 int error; 4900 4901 fd = SCARG(uap, fd); 4902 pos = SCARG(uap, pos); 4903 len = SCARG(uap, len); 4904 4905 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4906 return EINVAL; 4907 } 4908 4909 error = fd_getvnode(fd, &fp); 4910 if (error) { 4911 return error; 4912 } 4913 if ((fp->f_flag & FWRITE) == 0) { 4914 error = EBADF; 4915 goto fail; 4916 } 4917 vp = fp->f_vnode; 4918 4919 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4920 if (vp->v_type == VDIR) { 4921 error = EISDIR; 4922 } else { 4923 error = VOP_FDISCARD(vp, pos, len); 4924 } 4925 VOP_UNLOCK(vp); 4926 4927 fail: 4928 fd_putfile(fd); 4929 return error; 4930 } 4931
Indexes created Fri Oct 03 02:09:57 GMT 2025