vfs_syscalls.c revision 1.414.4.1
1 /* $NetBSD: vfs_syscalls.c,v 1.414.4.1 2011/01/20 14:24:58 bouyer Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.414.4.1 2011/01/20 14:24:58 bouyer Exp $"); 70 71 #ifdef _KERNEL_OPT 72 #include "opt_fileassoc.h" 73 #include "veriexec.h" 74 #endif 75 76 #include <sys/param.h> 77 #include <sys/systm.h> 78 #include <sys/namei.h> 79 #include <sys/filedesc.h> 80 #include <sys/kernel.h> 81 #include <sys/file.h> 82 #include <sys/stat.h> 83 #include <sys/vnode.h> 84 #include <sys/mount.h> 85 #include <sys/proc.h> 86 #include <sys/uio.h> 87 #include <sys/kmem.h> 88 #include <sys/dirent.h> 89 #include <sys/sysctl.h> 90 #include <sys/syscallargs.h> 91 #include <sys/vfs_syscalls.h> 92 #include <sys/ktrace.h> 93 #ifdef FILEASSOC 94 #include <sys/fileassoc.h> 95 #endif /* FILEASSOC */ 96 #include <sys/verified_exec.h> 97 #include <sys/kauth.h> 98 #include <sys/atomic.h> 99 #include <sys/module.h> 100 #include <sys/buf.h> 101 102 #include <miscfs/genfs/genfs.h> 103 #include <miscfs/syncfs/syncfs.h> 104 #include <miscfs/specfs/specdev.h> 105 106 #include <nfs/rpcv2.h> 107 #include <nfs/nfsproto.h> 108 #include <nfs/nfs.h> 109 #include <nfs/nfs_var.h> 110 111 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 112 113 static int change_flags(struct vnode *, u_long, struct lwp *); 114 static int change_mode(struct vnode *, int, struct lwp *l); 115 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 116 117 void checkdirs(struct vnode *); 118 119 /* 120 * Virtual File System System Calls 121 */ 122 123 /* 124 * Mount a file system. 125 */ 126 127 /* 128 * This table is used to maintain compatibility with 4.3BSD 129 * and NetBSD 0.9 mount syscalls - and possibly other systems. 130 * Note, the order is important! 131 * 132 * Do not modify this table. It should only contain filesystems 133 * supported by NetBSD 0.9 and 4.3BSD. 134 */ 135 const char * const mountcompatnames[] = { 136 NULL, /* 0 = MOUNT_NONE */ 137 MOUNT_FFS, /* 1 = MOUNT_UFS */ 138 MOUNT_NFS, /* 2 */ 139 MOUNT_MFS, /* 3 */ 140 MOUNT_MSDOS, /* 4 */ 141 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 142 MOUNT_FDESC, /* 6 */ 143 MOUNT_KERNFS, /* 7 */ 144 NULL, /* 8 = MOUNT_DEVFS */ 145 MOUNT_AFS, /* 9 */ 146 }; 147 const int nmountcompatnames = sizeof(mountcompatnames) / 148 sizeof(mountcompatnames[0]); 149 150 static int 151 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 152 void *data, size_t *data_len) 153 { 154 struct mount *mp; 155 int error = 0, saved_flags; 156 157 mp = vp->v_mount; 158 saved_flags = mp->mnt_flag; 159 160 /* We can operate only on VV_ROOT nodes. */ 161 if ((vp->v_vflag & VV_ROOT) == 0) { 162 error = EINVAL; 163 goto out; 164 } 165 166 /* 167 * We only allow the filesystem to be reloaded if it 168 * is currently mounted read-only. Additionally, we 169 * prevent read-write to read-only downgrades. 170 */ 171 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 172 (mp->mnt_flag & MNT_RDONLY) == 0 && 173 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 174 error = EOPNOTSUPP; /* Needs translation */ 175 goto out; 176 } 177 178 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 179 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 180 if (error) 181 goto out; 182 183 if (vfs_busy(mp, NULL)) { 184 error = EPERM; 185 goto out; 186 } 187 188 mutex_enter(&mp->mnt_updating); 189 190 mp->mnt_flag &= ~MNT_OP_FLAGS; 191 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 192 193 /* 194 * Set the mount level flags. 195 */ 196 if (flags & MNT_RDONLY) 197 mp->mnt_flag |= MNT_RDONLY; 198 else if (mp->mnt_flag & MNT_RDONLY) 199 mp->mnt_iflag |= IMNT_WANTRDWR; 200 mp->mnt_flag &= 201 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 202 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 203 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 204 MNT_LOG); 205 mp->mnt_flag |= flags & 206 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 207 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 208 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 209 MNT_LOG | MNT_IGNORE); 210 211 error = VFS_MOUNT(mp, path, data, data_len); 212 213 if (error && data != NULL) { 214 int error2; 215 216 /* 217 * Update failed; let's try and see if it was an 218 * export request. For compat with 3.0 and earlier. 219 */ 220 error2 = vfs_hooks_reexport(mp, path, data); 221 222 /* 223 * Only update error code if the export request was 224 * understood but some problem occurred while 225 * processing it. 226 */ 227 if (error2 != EJUSTRETURN) 228 error = error2; 229 } 230 231 if (mp->mnt_iflag & IMNT_WANTRDWR) 232 mp->mnt_flag &= ~MNT_RDONLY; 233 if (error) 234 mp->mnt_flag = saved_flags; 235 mp->mnt_flag &= ~MNT_OP_FLAGS; 236 mp->mnt_iflag &= ~IMNT_WANTRDWR; 237 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 238 if (mp->mnt_syncer == NULL) 239 error = vfs_allocate_syncvnode(mp); 240 } else { 241 if (mp->mnt_syncer != NULL) 242 vfs_deallocate_syncvnode(mp); 243 } 244 mutex_exit(&mp->mnt_updating); 245 vfs_unbusy(mp, false, NULL); 246 247 out: 248 return (error); 249 } 250 251 static int 252 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 253 { 254 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 255 int error; 256 257 /* Copy file-system type from userspace. */ 258 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 259 if (error) { 260 /* 261 * Historically, filesystem types were identified by numbers. 262 * If we get an integer for the filesystem type instead of a 263 * string, we check to see if it matches one of the historic 264 * filesystem types. 265 */ 266 u_long fsindex = (u_long)fstype; 267 if (fsindex >= nmountcompatnames || 268 mountcompatnames[fsindex] == NULL) 269 return ENODEV; 270 strlcpy(fstypename, mountcompatnames[fsindex], 271 sizeof(fstypename)); 272 } 273 274 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 275 if (strcmp(fstypename, "ufs") == 0) 276 fstypename[0] = 'f'; 277 278 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 279 return 0; 280 281 /* If we can autoload a vfs module, try again */ 282 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 283 284 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 285 return 0; 286 287 return ENODEV; 288 } 289 290 static int 291 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 292 const char *path, int flags, void *data, size_t *data_len) 293 { 294 struct mount *mp; 295 struct vnode *vp = *vpp; 296 struct vattr va; 297 struct pathbuf *pb; 298 struct nameidata nd; 299 int error; 300 301 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 302 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 303 if (error) { 304 vfs_delref(vfsops); 305 return error; 306 } 307 308 /* Can't make a non-dir a mount-point (from here anyway). */ 309 if (vp->v_type != VDIR) { 310 vfs_delref(vfsops); 311 return ENOTDIR; 312 } 313 314 /* 315 * If the user is not root, ensure that they own the directory 316 * onto which we are attempting to mount. 317 */ 318 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 319 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 320 (error = kauth_authorize_generic(l->l_cred, 321 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 322 vfs_delref(vfsops); 323 return error; 324 } 325 326 if (flags & MNT_EXPORTED) { 327 vfs_delref(vfsops); 328 return EINVAL; 329 } 330 331 if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) { 332 vfs_delref(vfsops); 333 return ENOMEM; 334 } 335 336 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 337 338 /* 339 * The underlying file system may refuse the mount for 340 * various reasons. Allow the user to force it to happen. 341 * 342 * Set the mount level flags. 343 */ 344 mp->mnt_flag = flags & 345 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 346 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 347 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 348 MNT_LOG | MNT_IGNORE | MNT_RDONLY); 349 350 mutex_enter(&mp->mnt_updating); 351 error = VFS_MOUNT(mp, path, data, data_len); 352 mp->mnt_flag &= ~MNT_OP_FLAGS; 353 354 if (error != 0) 355 goto err_unmounted; 356 357 /* 358 * Validate and prepare the mount point. 359 */ 360 error = pathbuf_copyin(path, &pb); 361 if (error != 0) { 362 goto err_mounted; 363 } 364 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 365 error = namei(&nd); 366 pathbuf_destroy(pb); 367 if (error != 0) { 368 goto err_mounted; 369 } 370 if (nd.ni_vp != vp) { 371 vput(nd.ni_vp); 372 error = EINVAL; 373 goto err_mounted; 374 } 375 if (vp->v_mountedhere != NULL) { 376 vput(nd.ni_vp); 377 error = EBUSY; 378 goto err_mounted; 379 } 380 error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0); 381 if (error != 0) { 382 vput(nd.ni_vp); 383 goto err_mounted; 384 } 385 386 /* 387 * Put the new filesystem on the mount list after root. 388 */ 389 cache_purge(vp); 390 mp->mnt_iflag &= ~IMNT_WANTRDWR; 391 392 mutex_enter(&mountlist_lock); 393 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 394 mutex_exit(&mountlist_lock); 395 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 396 error = vfs_allocate_syncvnode(mp); 397 if (error == 0) 398 vp->v_mountedhere = mp; 399 vput(nd.ni_vp); 400 if (error != 0) 401 goto err_onmountlist; 402 403 checkdirs(vp); 404 mutex_exit(&mp->mnt_updating); 405 406 /* Hold an additional reference to the mount across VFS_START(). */ 407 vfs_unbusy(mp, true, NULL); 408 (void) VFS_STATVFS(mp, &mp->mnt_stat); 409 error = VFS_START(mp, 0); 410 if (error) 411 vrele(vp); 412 /* Drop reference held for VFS_START(). */ 413 vfs_destroy(mp); 414 *vpp = NULL; 415 return error; 416 417 err_onmountlist: 418 mutex_enter(&mountlist_lock); 419 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 420 mp->mnt_iflag |= IMNT_GONE; 421 mutex_exit(&mountlist_lock); 422 423 err_mounted: 424 if (VFS_UNMOUNT(mp, MNT_FORCE) != 0) 425 panic("Unmounting fresh file system failed"); 426 427 err_unmounted: 428 vp->v_mountedhere = NULL; 429 mutex_exit(&mp->mnt_updating); 430 vfs_unbusy(mp, false, NULL); 431 vfs_destroy(mp); 432 433 return error; 434 } 435 436 static int 437 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 438 void *data, size_t *data_len) 439 { 440 struct mount *mp; 441 int error; 442 443 /* If MNT_GETARGS is specified, it should be the only flag. */ 444 if (flags & ~MNT_GETARGS) 445 return EINVAL; 446 447 mp = vp->v_mount; 448 449 /* XXX: probably some notion of "can see" here if we want isolation. */ 450 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 451 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 452 if (error) 453 return error; 454 455 if ((vp->v_vflag & VV_ROOT) == 0) 456 return EINVAL; 457 458 if (vfs_busy(mp, NULL)) 459 return EPERM; 460 461 mutex_enter(&mp->mnt_updating); 462 mp->mnt_flag &= ~MNT_OP_FLAGS; 463 mp->mnt_flag |= MNT_GETARGS; 464 error = VFS_MOUNT(mp, path, data, data_len); 465 mp->mnt_flag &= ~MNT_OP_FLAGS; 466 mutex_exit(&mp->mnt_updating); 467 468 vfs_unbusy(mp, false, NULL); 469 return (error); 470 } 471 472 int 473 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 474 { 475 /* { 476 syscallarg(const char *) type; 477 syscallarg(const char *) path; 478 syscallarg(int) flags; 479 syscallarg(void *) data; 480 syscallarg(size_t) data_len; 481 } */ 482 483 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 484 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 485 SCARG(uap, data_len), retval); 486 } 487 488 int 489 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 490 const char *path, int flags, void *data, enum uio_seg data_seg, 491 size_t data_len, register_t *retval) 492 { 493 struct vnode *vp; 494 void *data_buf = data; 495 bool vfsopsrele = false; 496 int error; 497 498 /* XXX: The calling convention of this routine is totally bizarre */ 499 if (vfsops) 500 vfsopsrele = true; 501 502 /* 503 * Get vnode to be covered 504 */ 505 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 506 if (error != 0) { 507 vp = NULL; 508 goto done; 509 } 510 511 if (vfsops == NULL) { 512 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 513 vfsops = vp->v_mount->mnt_op; 514 } else { 515 /* 'type' is userspace */ 516 error = mount_get_vfsops(type, &vfsops); 517 if (error != 0) 518 goto done; 519 vfsopsrele = true; 520 } 521 } 522 523 if (data != NULL && data_seg == UIO_USERSPACE) { 524 if (data_len == 0) { 525 /* No length supplied, use default for filesystem */ 526 data_len = vfsops->vfs_min_mount_data; 527 if (data_len > VFS_MAX_MOUNT_DATA) { 528 error = EINVAL; 529 goto done; 530 } 531 /* 532 * Hopefully a longer buffer won't make copyin() fail. 533 * For compatibility with 3.0 and earlier. 534 */ 535 if (flags & MNT_UPDATE 536 && data_len < sizeof (struct mnt_export_args30)) 537 data_len = sizeof (struct mnt_export_args30); 538 } 539 data_buf = kmem_alloc(data_len, KM_SLEEP); 540 541 /* NFS needs the buffer even for mnt_getargs .... */ 542 error = copyin(data, data_buf, data_len); 543 if (error != 0) 544 goto done; 545 } 546 547 if (flags & MNT_GETARGS) { 548 if (data_len == 0) { 549 error = EINVAL; 550 goto done; 551 } 552 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 553 if (error != 0) 554 goto done; 555 if (data_seg == UIO_USERSPACE) 556 error = copyout(data_buf, data, data_len); 557 *retval = data_len; 558 } else if (flags & MNT_UPDATE) { 559 error = mount_update(l, vp, path, flags, data_buf, &data_len); 560 } else { 561 /* Locking is handled internally in mount_domount(). */ 562 KASSERT(vfsopsrele == true); 563 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 564 &data_len); 565 vfsopsrele = false; 566 } 567 568 done: 569 if (vfsopsrele) 570 vfs_delref(vfsops); 571 if (vp != NULL) { 572 vrele(vp); 573 } 574 if (data_buf != data) 575 kmem_free(data_buf, data_len); 576 return (error); 577 } 578 579 /* 580 * Scan all active processes to see if any of them have a current 581 * or root directory onto which the new filesystem has just been 582 * mounted. If so, replace them with the new mount point. 583 */ 584 void 585 checkdirs(struct vnode *olddp) 586 { 587 struct cwdinfo *cwdi; 588 struct vnode *newdp, *rele1, *rele2; 589 struct proc *p; 590 bool retry; 591 592 if (olddp->v_usecount == 1) 593 return; 594 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 595 panic("mount: lost mount"); 596 597 do { 598 retry = false; 599 mutex_enter(proc_lock); 600 PROCLIST_FOREACH(p, &allproc) { 601 if ((cwdi = p->p_cwdi) == NULL) 602 continue; 603 /* 604 * Can't change to the old directory any more, 605 * so even if we see a stale value it's not a 606 * problem. 607 */ 608 if (cwdi->cwdi_cdir != olddp && 609 cwdi->cwdi_rdir != olddp) 610 continue; 611 retry = true; 612 rele1 = NULL; 613 rele2 = NULL; 614 atomic_inc_uint(&cwdi->cwdi_refcnt); 615 mutex_exit(proc_lock); 616 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 617 if (cwdi->cwdi_cdir == olddp) { 618 rele1 = cwdi->cwdi_cdir; 619 vref(newdp); 620 cwdi->cwdi_cdir = newdp; 621 } 622 if (cwdi->cwdi_rdir == olddp) { 623 rele2 = cwdi->cwdi_rdir; 624 vref(newdp); 625 cwdi->cwdi_rdir = newdp; 626 } 627 rw_exit(&cwdi->cwdi_lock); 628 cwdfree(cwdi); 629 if (rele1 != NULL) 630 vrele(rele1); 631 if (rele2 != NULL) 632 vrele(rele2); 633 mutex_enter(proc_lock); 634 break; 635 } 636 mutex_exit(proc_lock); 637 } while (retry); 638 639 if (rootvnode == olddp) { 640 vrele(rootvnode); 641 vref(newdp); 642 rootvnode = newdp; 643 } 644 vput(newdp); 645 } 646 647 /* 648 * Unmount a file system. 649 * 650 * Note: unmount takes a path to the vnode mounted on as argument, 651 * not special file (as before). 652 */ 653 /* ARGSUSED */ 654 int 655 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 656 { 657 /* { 658 syscallarg(const char *) path; 659 syscallarg(int) flags; 660 } */ 661 struct vnode *vp; 662 struct mount *mp; 663 int error; 664 struct pathbuf *pb; 665 struct nameidata nd; 666 667 error = pathbuf_copyin(SCARG(uap, path), &pb); 668 if (error) { 669 return error; 670 } 671 672 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 673 if ((error = namei(&nd)) != 0) { 674 pathbuf_destroy(pb); 675 return error; 676 } 677 vp = nd.ni_vp; 678 pathbuf_destroy(pb); 679 680 mp = vp->v_mount; 681 atomic_inc_uint(&mp->mnt_refcnt); 682 VOP_UNLOCK(vp); 683 684 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 685 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 686 if (error) { 687 vrele(vp); 688 vfs_destroy(mp); 689 return (error); 690 } 691 692 /* 693 * Don't allow unmounting the root file system. 694 */ 695 if (mp->mnt_flag & MNT_ROOTFS) { 696 vrele(vp); 697 vfs_destroy(mp); 698 return (EINVAL); 699 } 700 701 /* 702 * Must be the root of the filesystem 703 */ 704 if ((vp->v_vflag & VV_ROOT) == 0) { 705 vrele(vp); 706 vfs_destroy(mp); 707 return (EINVAL); 708 } 709 710 vrele(vp); 711 error = dounmount(mp, SCARG(uap, flags), l); 712 vfs_destroy(mp); 713 return error; 714 } 715 716 /* 717 * Do the actual file system unmount. File system is assumed to have 718 * been locked by the caller. 719 * 720 * => Caller hold reference to the mount, explicitly for dounmount(). 721 */ 722 int 723 dounmount(struct mount *mp, int flags, struct lwp *l) 724 { 725 struct vnode *coveredvp; 726 int error; 727 int async; 728 int used_syncer; 729 730 #if NVERIEXEC > 0 731 error = veriexec_unmountchk(mp); 732 if (error) 733 return (error); 734 #endif /* NVERIEXEC > 0 */ 735 736 /* 737 * XXX Freeze syncer. Must do this before locking the 738 * mount point. See dounmount() for details. 739 */ 740 mutex_enter(&syncer_mutex); 741 rw_enter(&mp->mnt_unmounting, RW_WRITER); 742 if ((mp->mnt_iflag & IMNT_GONE) != 0) { 743 rw_exit(&mp->mnt_unmounting); 744 mutex_exit(&syncer_mutex); 745 return ENOENT; 746 } 747 748 used_syncer = (mp->mnt_syncer != NULL); 749 750 /* 751 * XXX Syncer must be frozen when we get here. This should really 752 * be done on a per-mountpoint basis, but the syncer doesn't work 753 * like that. 754 * 755 * The caller of dounmount() must acquire syncer_mutex because 756 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 757 * order, and we must preserve that order to avoid deadlock. 758 * 759 * So, if the file system did not use the syncer, now is 760 * the time to release the syncer_mutex. 761 */ 762 if (used_syncer == 0) 763 mutex_exit(&syncer_mutex); 764 765 mp->mnt_iflag |= IMNT_UNMOUNT; 766 async = mp->mnt_flag & MNT_ASYNC; 767 mp->mnt_flag &= ~MNT_ASYNC; 768 cache_purgevfs(mp); /* remove cache entries for this file sys */ 769 if (mp->mnt_syncer != NULL) 770 vfs_deallocate_syncvnode(mp); 771 error = 0; 772 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 773 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 774 } 775 vfs_scrubvnlist(mp); 776 if (error == 0 || (flags & MNT_FORCE)) 777 error = VFS_UNMOUNT(mp, flags); 778 if (error) { 779 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 780 (void) vfs_allocate_syncvnode(mp); 781 mp->mnt_iflag &= ~IMNT_UNMOUNT; 782 mp->mnt_flag |= async; 783 rw_exit(&mp->mnt_unmounting); 784 if (used_syncer) 785 mutex_exit(&syncer_mutex); 786 return (error); 787 } 788 vfs_scrubvnlist(mp); 789 mutex_enter(&mountlist_lock); 790 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 791 coveredvp->v_mountedhere = NULL; 792 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 793 mp->mnt_iflag |= IMNT_GONE; 794 mutex_exit(&mountlist_lock); 795 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 796 panic("unmount: dangling vnode"); 797 if (used_syncer) 798 mutex_exit(&syncer_mutex); 799 vfs_hooks_unmount(mp); 800 rw_exit(&mp->mnt_unmounting); 801 vfs_destroy(mp); /* reference from mount() */ 802 if (coveredvp != NULLVP) 803 vrele(coveredvp); 804 return (0); 805 } 806 807 /* 808 * Sync each mounted filesystem. 809 */ 810 #ifdef DEBUG 811 int syncprt = 0; 812 struct ctldebug debug0 = { "syncprt", &syncprt }; 813 #endif 814 815 /* ARGSUSED */ 816 int 817 sys_sync(struct lwp *l, const void *v, register_t *retval) 818 { 819 struct mount *mp, *nmp; 820 int asyncflag; 821 822 if (l == NULL) 823 l = &lwp0; 824 825 mutex_enter(&mountlist_lock); 826 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 827 mp = nmp) { 828 if (vfs_busy(mp, &nmp)) { 829 continue; 830 } 831 mutex_enter(&mp->mnt_updating); 832 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 833 asyncflag = mp->mnt_flag & MNT_ASYNC; 834 mp->mnt_flag &= ~MNT_ASYNC; 835 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 836 if (asyncflag) 837 mp->mnt_flag |= MNT_ASYNC; 838 } 839 mutex_exit(&mp->mnt_updating); 840 vfs_unbusy(mp, false, &nmp); 841 } 842 mutex_exit(&mountlist_lock); 843 #ifdef DEBUG 844 if (syncprt) 845 vfs_bufstats(); 846 #endif /* DEBUG */ 847 return (0); 848 } 849 850 /* 851 * Change filesystem quotas. 852 */ 853 /* ARGSUSED */ 854 int 855 compat_50_sys_quotactl(struct lwp *l, const struct compat_50_sys_quotactl_args *uap, register_t *retval) 856 { 857 /* { 858 syscallarg(const char *) path; 859 syscallarg(int) cmd; 860 syscallarg(int) uid; 861 syscallarg(void *) arg; 862 } */ 863 struct mount *mp; 864 int error; 865 struct vnode *vp; 866 867 error = namei_simple_user(SCARG(uap, path), 868 NSM_FOLLOW_TRYEMULROOT, &vp); 869 if (error != 0) 870 return (error); 871 mp = vp->v_mount; 872 #if 0 873 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 874 SCARG(uap, arg)); 875 #else 876 error = EOPNOTSUPP; 877 #endif 878 vrele(vp); 879 return (error); 880 } 881 882 /* 883 * Change filesystem quotas. 884 */ 885 /* ARGSUSED */ 886 int 887 sys___quotactl50(struct lwp *l, const struct sys___quotactl50_args *uap, 888 register_t *retval) 889 { 890 /* { 891 syscallarg(const char *) path; 892 syscallarg(struct plistref *) pref; 893 } */ 894 struct mount *mp; 895 int error; 896 struct vnode *vp; 897 prop_dictionary_t dict; 898 struct plistref pref; 899 900 error = namei_simple_user(SCARG(uap, path), 901 NSM_FOLLOW_TRYEMULROOT, &vp); 902 printf("namei error %d\n", error); 903 if (error != 0) 904 return (error); 905 mp = vp->v_mount; 906 error = copyin(SCARG(uap, pref), &pref, sizeof(pref)); 907 printf("copyin error %d\n", error); 908 if (error) 909 return error; 910 error = prop_dictionary_copyin(&pref, &dict); 911 printf("prop_dictionary_copyin error %d\n", error); 912 if (error) 913 return error; 914 error = VFS_QUOTACTL(mp, dict); 915 printf("VFS_QUOTACTL error %d\n", error); 916 vrele(vp); 917 if (!error) 918 error = prop_dictionary_copyout(&pref, dict); 919 printf("prop_dictionary_copyout error %d\n", error); 920 if (!error) 921 error = copyout(&pref, SCARG(uap, pref), sizeof(pref)); 922 printf("copyout error %d\n", error); 923 prop_object_release(dict); 924 return (error); 925 } 926 927 int 928 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 929 int root) 930 { 931 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 932 int error = 0; 933 934 /* 935 * If MNT_NOWAIT or MNT_LAZY is specified, do not 936 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 937 * overrides MNT_NOWAIT. 938 */ 939 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 940 (flags != MNT_WAIT && flags != 0)) { 941 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 942 goto done; 943 } 944 945 /* Get the filesystem stats now */ 946 memset(sp, 0, sizeof(*sp)); 947 if ((error = VFS_STATVFS(mp, sp)) != 0) { 948 return error; 949 } 950 951 if (cwdi->cwdi_rdir == NULL) 952 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 953 done: 954 if (cwdi->cwdi_rdir != NULL) { 955 size_t len; 956 char *bp; 957 char c; 958 char *path = PNBUF_GET(); 959 960 bp = path + MAXPATHLEN; 961 *--bp = '\0'; 962 rw_enter(&cwdi->cwdi_lock, RW_READER); 963 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 964 MAXPATHLEN / 2, 0, l); 965 rw_exit(&cwdi->cwdi_lock); 966 if (error) { 967 PNBUF_PUT(path); 968 return error; 969 } 970 len = strlen(bp); 971 if (len != 1) { 972 /* 973 * for mount points that are below our root, we can see 974 * them, so we fix up the pathname and return them. The 975 * rest we cannot see, so we don't allow viewing the 976 * data. 977 */ 978 if (strncmp(bp, sp->f_mntonname, len) == 0 && 979 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 980 (void)strlcpy(sp->f_mntonname, 981 c == '\0' ? "/" : &sp->f_mntonname[len], 982 sizeof(sp->f_mntonname)); 983 } else { 984 if (root) 985 (void)strlcpy(sp->f_mntonname, "/", 986 sizeof(sp->f_mntonname)); 987 else 988 error = EPERM; 989 } 990 } 991 PNBUF_PUT(path); 992 } 993 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 994 return error; 995 } 996 997 /* 998 * Get filesystem statistics by path. 999 */ 1000 int 1001 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1002 { 1003 struct mount *mp; 1004 int error; 1005 struct vnode *vp; 1006 1007 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1008 if (error != 0) 1009 return error; 1010 mp = vp->v_mount; 1011 error = dostatvfs(mp, sb, l, flags, 1); 1012 vrele(vp); 1013 return error; 1014 } 1015 1016 /* ARGSUSED */ 1017 int 1018 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 1019 { 1020 /* { 1021 syscallarg(const char *) path; 1022 syscallarg(struct statvfs *) buf; 1023 syscallarg(int) flags; 1024 } */ 1025 struct statvfs *sb; 1026 int error; 1027 1028 sb = STATVFSBUF_GET(); 1029 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1030 if (error == 0) 1031 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1032 STATVFSBUF_PUT(sb); 1033 return error; 1034 } 1035 1036 /* 1037 * Get filesystem statistics by fd. 1038 */ 1039 int 1040 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1041 { 1042 file_t *fp; 1043 struct mount *mp; 1044 int error; 1045 1046 /* fd_getvnode() will use the descriptor for us */ 1047 if ((error = fd_getvnode(fd, &fp)) != 0) 1048 return (error); 1049 mp = ((struct vnode *)fp->f_data)->v_mount; 1050 error = dostatvfs(mp, sb, curlwp, flags, 1); 1051 fd_putfile(fd); 1052 return error; 1053 } 1054 1055 /* ARGSUSED */ 1056 int 1057 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1058 { 1059 /* { 1060 syscallarg(int) fd; 1061 syscallarg(struct statvfs *) buf; 1062 syscallarg(int) flags; 1063 } */ 1064 struct statvfs *sb; 1065 int error; 1066 1067 sb = STATVFSBUF_GET(); 1068 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1069 if (error == 0) 1070 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1071 STATVFSBUF_PUT(sb); 1072 return error; 1073 } 1074 1075 1076 /* 1077 * Get statistics on all filesystems. 1078 */ 1079 int 1080 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1081 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1082 register_t *retval) 1083 { 1084 int root = 0; 1085 struct proc *p = l->l_proc; 1086 struct mount *mp, *nmp; 1087 struct statvfs *sb; 1088 size_t count, maxcount; 1089 int error = 0; 1090 1091 sb = STATVFSBUF_GET(); 1092 maxcount = bufsize / entry_sz; 1093 mutex_enter(&mountlist_lock); 1094 count = 0; 1095 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1096 mp = nmp) { 1097 if (vfs_busy(mp, &nmp)) { 1098 continue; 1099 } 1100 if (sfsp && count < maxcount) { 1101 error = dostatvfs(mp, sb, l, flags, 0); 1102 if (error) { 1103 vfs_unbusy(mp, false, &nmp); 1104 error = 0; 1105 continue; 1106 } 1107 error = copyfn(sb, sfsp, entry_sz); 1108 if (error) { 1109 vfs_unbusy(mp, false, NULL); 1110 goto out; 1111 } 1112 sfsp = (char *)sfsp + entry_sz; 1113 root |= strcmp(sb->f_mntonname, "/") == 0; 1114 } 1115 count++; 1116 vfs_unbusy(mp, false, &nmp); 1117 } 1118 mutex_exit(&mountlist_lock); 1119 1120 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1121 /* 1122 * fake a root entry 1123 */ 1124 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1125 sb, l, flags, 1); 1126 if (error != 0) 1127 goto out; 1128 if (sfsp) { 1129 error = copyfn(sb, sfsp, entry_sz); 1130 if (error != 0) 1131 goto out; 1132 } 1133 count++; 1134 } 1135 if (sfsp && count > maxcount) 1136 *retval = maxcount; 1137 else 1138 *retval = count; 1139 out: 1140 STATVFSBUF_PUT(sb); 1141 return error; 1142 } 1143 1144 int 1145 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1146 { 1147 /* { 1148 syscallarg(struct statvfs *) buf; 1149 syscallarg(size_t) bufsize; 1150 syscallarg(int) flags; 1151 } */ 1152 1153 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1154 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1155 } 1156 1157 /* 1158 * Change current working directory to a given file descriptor. 1159 */ 1160 /* ARGSUSED */ 1161 int 1162 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1163 { 1164 /* { 1165 syscallarg(int) fd; 1166 } */ 1167 struct proc *p = l->l_proc; 1168 struct cwdinfo *cwdi; 1169 struct vnode *vp, *tdp; 1170 struct mount *mp; 1171 file_t *fp; 1172 int error, fd; 1173 1174 /* fd_getvnode() will use the descriptor for us */ 1175 fd = SCARG(uap, fd); 1176 if ((error = fd_getvnode(fd, &fp)) != 0) 1177 return (error); 1178 vp = fp->f_data; 1179 1180 vref(vp); 1181 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1182 if (vp->v_type != VDIR) 1183 error = ENOTDIR; 1184 else 1185 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1186 if (error) { 1187 vput(vp); 1188 goto out; 1189 } 1190 while ((mp = vp->v_mountedhere) != NULL) { 1191 error = vfs_busy(mp, NULL); 1192 vput(vp); 1193 if (error != 0) 1194 goto out; 1195 error = VFS_ROOT(mp, &tdp); 1196 vfs_unbusy(mp, false, NULL); 1197 if (error) 1198 goto out; 1199 vp = tdp; 1200 } 1201 VOP_UNLOCK(vp); 1202 1203 /* 1204 * Disallow changing to a directory not under the process's 1205 * current root directory (if there is one). 1206 */ 1207 cwdi = p->p_cwdi; 1208 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1209 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1210 vrele(vp); 1211 error = EPERM; /* operation not permitted */ 1212 } else { 1213 vrele(cwdi->cwdi_cdir); 1214 cwdi->cwdi_cdir = vp; 1215 } 1216 rw_exit(&cwdi->cwdi_lock); 1217 1218 out: 1219 fd_putfile(fd); 1220 return (error); 1221 } 1222 1223 /* 1224 * Change this process's notion of the root directory to a given file 1225 * descriptor. 1226 */ 1227 int 1228 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1229 { 1230 struct proc *p = l->l_proc; 1231 struct vnode *vp; 1232 file_t *fp; 1233 int error, fd = SCARG(uap, fd); 1234 1235 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1236 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1237 return error; 1238 /* fd_getvnode() will use the descriptor for us */ 1239 if ((error = fd_getvnode(fd, &fp)) != 0) 1240 return error; 1241 vp = fp->f_data; 1242 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1243 if (vp->v_type != VDIR) 1244 error = ENOTDIR; 1245 else 1246 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1247 VOP_UNLOCK(vp); 1248 if (error) 1249 goto out; 1250 vref(vp); 1251 1252 change_root(p->p_cwdi, vp, l); 1253 1254 out: 1255 fd_putfile(fd); 1256 return (error); 1257 } 1258 1259 /* 1260 * Change current working directory (``.''). 1261 */ 1262 /* ARGSUSED */ 1263 int 1264 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1265 { 1266 /* { 1267 syscallarg(const char *) path; 1268 } */ 1269 struct proc *p = l->l_proc; 1270 struct cwdinfo *cwdi; 1271 int error; 1272 struct vnode *vp; 1273 1274 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1275 &vp, l)) != 0) 1276 return (error); 1277 cwdi = p->p_cwdi; 1278 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1279 vrele(cwdi->cwdi_cdir); 1280 cwdi->cwdi_cdir = vp; 1281 rw_exit(&cwdi->cwdi_lock); 1282 return (0); 1283 } 1284 1285 /* 1286 * Change notion of root (``/'') directory. 1287 */ 1288 /* ARGSUSED */ 1289 int 1290 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1291 { 1292 /* { 1293 syscallarg(const char *) path; 1294 } */ 1295 struct proc *p = l->l_proc; 1296 int error; 1297 struct vnode *vp; 1298 1299 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1300 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1301 return (error); 1302 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1303 &vp, l)) != 0) 1304 return (error); 1305 1306 change_root(p->p_cwdi, vp, l); 1307 1308 return (0); 1309 } 1310 1311 /* 1312 * Common routine for chroot and fchroot. 1313 * NB: callers need to properly authorize the change root operation. 1314 */ 1315 void 1316 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1317 { 1318 1319 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1320 if (cwdi->cwdi_rdir != NULL) 1321 vrele(cwdi->cwdi_rdir); 1322 cwdi->cwdi_rdir = vp; 1323 1324 /* 1325 * Prevent escaping from chroot by putting the root under 1326 * the working directory. Silently chdir to / if we aren't 1327 * already there. 1328 */ 1329 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1330 /* 1331 * XXX would be more failsafe to change directory to a 1332 * deadfs node here instead 1333 */ 1334 vrele(cwdi->cwdi_cdir); 1335 vref(vp); 1336 cwdi->cwdi_cdir = vp; 1337 } 1338 rw_exit(&cwdi->cwdi_lock); 1339 } 1340 1341 /* 1342 * Common routine for chroot and chdir. 1343 * XXX "where" should be enum uio_seg 1344 */ 1345 int 1346 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1347 { 1348 struct pathbuf *pb; 1349 struct nameidata nd; 1350 int error; 1351 1352 error = pathbuf_maybe_copyin(path, where, &pb); 1353 if (error) { 1354 return error; 1355 } 1356 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1357 if ((error = namei(&nd)) != 0) { 1358 pathbuf_destroy(pb); 1359 return error; 1360 } 1361 *vpp = nd.ni_vp; 1362 pathbuf_destroy(pb); 1363 1364 if ((*vpp)->v_type != VDIR) 1365 error = ENOTDIR; 1366 else 1367 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1368 1369 if (error) 1370 vput(*vpp); 1371 else 1372 VOP_UNLOCK(*vpp); 1373 return (error); 1374 } 1375 1376 /* 1377 * Check permissions, allocate an open file structure, 1378 * and call the device open routine if any. 1379 */ 1380 int 1381 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1382 { 1383 /* { 1384 syscallarg(const char *) path; 1385 syscallarg(int) flags; 1386 syscallarg(int) mode; 1387 } */ 1388 struct proc *p = l->l_proc; 1389 struct cwdinfo *cwdi = p->p_cwdi; 1390 file_t *fp; 1391 struct vnode *vp; 1392 int flags, cmode; 1393 int type, indx, error; 1394 struct flock lf; 1395 struct pathbuf *pb; 1396 struct nameidata nd; 1397 1398 flags = FFLAGS(SCARG(uap, flags)); 1399 if ((flags & (FREAD | FWRITE)) == 0) 1400 return (EINVAL); 1401 1402 error = pathbuf_copyin(SCARG(uap, path), &pb); 1403 if (error) { 1404 return error; 1405 } 1406 1407 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1408 pathbuf_destroy(pb); 1409 return error; 1410 } 1411 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1412 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1413 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1414 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1415 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1416 fd_abort(p, fp, indx); 1417 if ((error == EDUPFD || error == EMOVEFD) && 1418 l->l_dupfd >= 0 && /* XXX from fdopen */ 1419 (error = 1420 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1421 *retval = indx; 1422 pathbuf_destroy(pb); 1423 return (0); 1424 } 1425 if (error == ERESTART) 1426 error = EINTR; 1427 pathbuf_destroy(pb); 1428 return (error); 1429 } 1430 1431 l->l_dupfd = 0; 1432 vp = nd.ni_vp; 1433 pathbuf_destroy(pb); 1434 1435 fp->f_flag = flags & FMASK; 1436 fp->f_type = DTYPE_VNODE; 1437 fp->f_ops = &vnops; 1438 fp->f_data = vp; 1439 if (flags & (O_EXLOCK | O_SHLOCK)) { 1440 lf.l_whence = SEEK_SET; 1441 lf.l_start = 0; 1442 lf.l_len = 0; 1443 if (flags & O_EXLOCK) 1444 lf.l_type = F_WRLCK; 1445 else 1446 lf.l_type = F_RDLCK; 1447 type = F_FLOCK; 1448 if ((flags & FNONBLOCK) == 0) 1449 type |= F_WAIT; 1450 VOP_UNLOCK(vp); 1451 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1452 if (error) { 1453 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1454 fd_abort(p, fp, indx); 1455 return (error); 1456 } 1457 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1458 atomic_or_uint(&fp->f_flag, FHASLOCK); 1459 } 1460 VOP_UNLOCK(vp); 1461 *retval = indx; 1462 fd_affix(p, fp, indx); 1463 return (0); 1464 } 1465 1466 static void 1467 vfs__fhfree(fhandle_t *fhp) 1468 { 1469 size_t fhsize; 1470 1471 if (fhp == NULL) { 1472 return; 1473 } 1474 fhsize = FHANDLE_SIZE(fhp); 1475 kmem_free(fhp, fhsize); 1476 } 1477 1478 /* 1479 * vfs_composefh: compose a filehandle. 1480 */ 1481 1482 int 1483 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1484 { 1485 struct mount *mp; 1486 struct fid *fidp; 1487 int error; 1488 size_t needfhsize; 1489 size_t fidsize; 1490 1491 mp = vp->v_mount; 1492 fidp = NULL; 1493 if (*fh_size < FHANDLE_SIZE_MIN) { 1494 fidsize = 0; 1495 } else { 1496 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1497 if (fhp != NULL) { 1498 memset(fhp, 0, *fh_size); 1499 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1500 fidp = &fhp->fh_fid; 1501 } 1502 } 1503 error = VFS_VPTOFH(vp, fidp, &fidsize); 1504 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1505 if (error == 0 && *fh_size < needfhsize) { 1506 error = E2BIG; 1507 } 1508 *fh_size = needfhsize; 1509 return error; 1510 } 1511 1512 int 1513 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1514 { 1515 struct mount *mp; 1516 fhandle_t *fhp; 1517 size_t fhsize; 1518 size_t fidsize; 1519 int error; 1520 1521 *fhpp = NULL; 1522 mp = vp->v_mount; 1523 fidsize = 0; 1524 error = VFS_VPTOFH(vp, NULL, &fidsize); 1525 KASSERT(error != 0); 1526 if (error != E2BIG) { 1527 goto out; 1528 } 1529 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1530 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1531 if (fhp == NULL) { 1532 error = ENOMEM; 1533 goto out; 1534 } 1535 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1536 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1537 if (error == 0) { 1538 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1539 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1540 *fhpp = fhp; 1541 } else { 1542 kmem_free(fhp, fhsize); 1543 } 1544 out: 1545 return error; 1546 } 1547 1548 void 1549 vfs_composefh_free(fhandle_t *fhp) 1550 { 1551 1552 vfs__fhfree(fhp); 1553 } 1554 1555 /* 1556 * vfs_fhtovp: lookup a vnode by a filehandle. 1557 */ 1558 1559 int 1560 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1561 { 1562 struct mount *mp; 1563 int error; 1564 1565 *vpp = NULL; 1566 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1567 if (mp == NULL) { 1568 error = ESTALE; 1569 goto out; 1570 } 1571 if (mp->mnt_op->vfs_fhtovp == NULL) { 1572 error = EOPNOTSUPP; 1573 goto out; 1574 } 1575 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1576 out: 1577 return error; 1578 } 1579 1580 /* 1581 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1582 * the needed size. 1583 */ 1584 1585 int 1586 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1587 { 1588 fhandle_t *fhp; 1589 int error; 1590 1591 *fhpp = NULL; 1592 if (fhsize > FHANDLE_SIZE_MAX) { 1593 return EINVAL; 1594 } 1595 if (fhsize < FHANDLE_SIZE_MIN) { 1596 return EINVAL; 1597 } 1598 again: 1599 fhp = kmem_alloc(fhsize, KM_SLEEP); 1600 if (fhp == NULL) { 1601 return ENOMEM; 1602 } 1603 error = copyin(ufhp, fhp, fhsize); 1604 if (error == 0) { 1605 /* XXX this check shouldn't be here */ 1606 if (FHANDLE_SIZE(fhp) == fhsize) { 1607 *fhpp = fhp; 1608 return 0; 1609 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1610 /* 1611 * a kludge for nfsv2 padded handles. 1612 */ 1613 size_t sz; 1614 1615 sz = FHANDLE_SIZE(fhp); 1616 kmem_free(fhp, fhsize); 1617 fhsize = sz; 1618 goto again; 1619 } else { 1620 /* 1621 * userland told us wrong size. 1622 */ 1623 error = EINVAL; 1624 } 1625 } 1626 kmem_free(fhp, fhsize); 1627 return error; 1628 } 1629 1630 void 1631 vfs_copyinfh_free(fhandle_t *fhp) 1632 { 1633 1634 vfs__fhfree(fhp); 1635 } 1636 1637 /* 1638 * Get file handle system call 1639 */ 1640 int 1641 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1642 { 1643 /* { 1644 syscallarg(char *) fname; 1645 syscallarg(fhandle_t *) fhp; 1646 syscallarg(size_t *) fh_size; 1647 } */ 1648 struct vnode *vp; 1649 fhandle_t *fh; 1650 int error; 1651 struct pathbuf *pb; 1652 struct nameidata nd; 1653 size_t sz; 1654 size_t usz; 1655 1656 /* 1657 * Must be super user 1658 */ 1659 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1660 0, NULL, NULL, NULL); 1661 if (error) 1662 return (error); 1663 1664 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1665 if (error) { 1666 return error; 1667 } 1668 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1669 error = namei(&nd); 1670 if (error) { 1671 pathbuf_destroy(pb); 1672 return error; 1673 } 1674 vp = nd.ni_vp; 1675 pathbuf_destroy(pb); 1676 1677 error = vfs_composefh_alloc(vp, &fh); 1678 vput(vp); 1679 if (error != 0) { 1680 goto out; 1681 } 1682 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1683 if (error != 0) { 1684 goto out; 1685 } 1686 sz = FHANDLE_SIZE(fh); 1687 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1688 if (error != 0) { 1689 goto out; 1690 } 1691 if (usz >= sz) { 1692 error = copyout(fh, SCARG(uap, fhp), sz); 1693 } else { 1694 error = E2BIG; 1695 } 1696 out: 1697 vfs_composefh_free(fh); 1698 return (error); 1699 } 1700 1701 /* 1702 * Open a file given a file handle. 1703 * 1704 * Check permissions, allocate an open file structure, 1705 * and call the device open routine if any. 1706 */ 1707 1708 int 1709 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1710 register_t *retval) 1711 { 1712 file_t *fp; 1713 struct vnode *vp = NULL; 1714 kauth_cred_t cred = l->l_cred; 1715 file_t *nfp; 1716 int type, indx, error=0; 1717 struct flock lf; 1718 struct vattr va; 1719 fhandle_t *fh; 1720 int flags; 1721 proc_t *p; 1722 1723 p = curproc; 1724 1725 /* 1726 * Must be super user 1727 */ 1728 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1729 0, NULL, NULL, NULL))) 1730 return (error); 1731 1732 flags = FFLAGS(oflags); 1733 if ((flags & (FREAD | FWRITE)) == 0) 1734 return (EINVAL); 1735 if ((flags & O_CREAT)) 1736 return (EINVAL); 1737 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1738 return (error); 1739 fp = nfp; 1740 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1741 if (error != 0) { 1742 goto bad; 1743 } 1744 error = vfs_fhtovp(fh, &vp); 1745 if (error != 0) { 1746 goto bad; 1747 } 1748 1749 /* Now do an effective vn_open */ 1750 1751 if (vp->v_type == VSOCK) { 1752 error = EOPNOTSUPP; 1753 goto bad; 1754 } 1755 error = vn_openchk(vp, cred, flags); 1756 if (error != 0) 1757 goto bad; 1758 if (flags & O_TRUNC) { 1759 VOP_UNLOCK(vp); /* XXX */ 1760 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1761 vattr_null(&va); 1762 va.va_size = 0; 1763 error = VOP_SETATTR(vp, &va, cred); 1764 if (error) 1765 goto bad; 1766 } 1767 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1768 goto bad; 1769 if (flags & FWRITE) { 1770 mutex_enter(&vp->v_interlock); 1771 vp->v_writecount++; 1772 mutex_exit(&vp->v_interlock); 1773 } 1774 1775 /* done with modified vn_open, now finish what sys_open does. */ 1776 1777 fp->f_flag = flags & FMASK; 1778 fp->f_type = DTYPE_VNODE; 1779 fp->f_ops = &vnops; 1780 fp->f_data = vp; 1781 if (flags & (O_EXLOCK | O_SHLOCK)) { 1782 lf.l_whence = SEEK_SET; 1783 lf.l_start = 0; 1784 lf.l_len = 0; 1785 if (flags & O_EXLOCK) 1786 lf.l_type = F_WRLCK; 1787 else 1788 lf.l_type = F_RDLCK; 1789 type = F_FLOCK; 1790 if ((flags & FNONBLOCK) == 0) 1791 type |= F_WAIT; 1792 VOP_UNLOCK(vp); 1793 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1794 if (error) { 1795 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1796 fd_abort(p, fp, indx); 1797 return (error); 1798 } 1799 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1800 atomic_or_uint(&fp->f_flag, FHASLOCK); 1801 } 1802 VOP_UNLOCK(vp); 1803 *retval = indx; 1804 fd_affix(p, fp, indx); 1805 vfs_copyinfh_free(fh); 1806 return (0); 1807 1808 bad: 1809 fd_abort(p, fp, indx); 1810 if (vp != NULL) 1811 vput(vp); 1812 vfs_copyinfh_free(fh); 1813 return (error); 1814 } 1815 1816 int 1817 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1818 { 1819 /* { 1820 syscallarg(const void *) fhp; 1821 syscallarg(size_t) fh_size; 1822 syscallarg(int) flags; 1823 } */ 1824 1825 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1826 SCARG(uap, flags), retval); 1827 } 1828 1829 int 1830 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1831 { 1832 int error; 1833 fhandle_t *fh; 1834 struct vnode *vp; 1835 1836 /* 1837 * Must be super user 1838 */ 1839 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1840 0, NULL, NULL, NULL))) 1841 return (error); 1842 1843 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1844 if (error != 0) 1845 return error; 1846 1847 error = vfs_fhtovp(fh, &vp); 1848 vfs_copyinfh_free(fh); 1849 if (error != 0) 1850 return error; 1851 1852 error = vn_stat(vp, sb); 1853 vput(vp); 1854 return error; 1855 } 1856 1857 1858 /* ARGSUSED */ 1859 int 1860 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 1861 { 1862 /* { 1863 syscallarg(const void *) fhp; 1864 syscallarg(size_t) fh_size; 1865 syscallarg(struct stat *) sb; 1866 } */ 1867 struct stat sb; 1868 int error; 1869 1870 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1871 if (error) 1872 return error; 1873 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1874 } 1875 1876 int 1877 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1878 int flags) 1879 { 1880 fhandle_t *fh; 1881 struct mount *mp; 1882 struct vnode *vp; 1883 int error; 1884 1885 /* 1886 * Must be super user 1887 */ 1888 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1889 0, NULL, NULL, NULL))) 1890 return error; 1891 1892 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1893 if (error != 0) 1894 return error; 1895 1896 error = vfs_fhtovp(fh, &vp); 1897 vfs_copyinfh_free(fh); 1898 if (error != 0) 1899 return error; 1900 1901 mp = vp->v_mount; 1902 error = dostatvfs(mp, sb, l, flags, 1); 1903 vput(vp); 1904 return error; 1905 } 1906 1907 /* ARGSUSED */ 1908 int 1909 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1910 { 1911 /* { 1912 syscallarg(const void *) fhp; 1913 syscallarg(size_t) fh_size; 1914 syscallarg(struct statvfs *) buf; 1915 syscallarg(int) flags; 1916 } */ 1917 struct statvfs *sb = STATVFSBUF_GET(); 1918 int error; 1919 1920 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1921 SCARG(uap, flags)); 1922 if (error == 0) 1923 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1924 STATVFSBUF_PUT(sb); 1925 return error; 1926 } 1927 1928 /* 1929 * Create a special file. 1930 */ 1931 /* ARGSUSED */ 1932 int 1933 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 1934 register_t *retval) 1935 { 1936 /* { 1937 syscallarg(const char *) path; 1938 syscallarg(mode_t) mode; 1939 syscallarg(dev_t) dev; 1940 } */ 1941 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 1942 SCARG(uap, dev), retval, UIO_USERSPACE); 1943 } 1944 1945 int 1946 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 1947 register_t *retval, enum uio_seg seg) 1948 { 1949 struct proc *p = l->l_proc; 1950 struct vnode *vp; 1951 struct vattr vattr; 1952 int error, optype; 1953 struct pathbuf *pb; 1954 struct nameidata nd; 1955 const char *pathstring; 1956 1957 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1958 0, NULL, NULL, NULL)) != 0) 1959 return (error); 1960 1961 optype = VOP_MKNOD_DESCOFFSET; 1962 1963 error = pathbuf_maybe_copyin(pathname, seg, &pb); 1964 if (error) { 1965 return error; 1966 } 1967 pathstring = pathbuf_stringcopy_get(pb); 1968 if (pathstring == NULL) { 1969 pathbuf_destroy(pb); 1970 return ENOMEM; 1971 } 1972 1973 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 1974 if ((error = namei(&nd)) != 0) 1975 goto out; 1976 vp = nd.ni_vp; 1977 1978 if (vp != NULL) 1979 error = EEXIST; 1980 else { 1981 vattr_null(&vattr); 1982 /* We will read cwdi->cwdi_cmask unlocked. */ 1983 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1984 vattr.va_rdev = dev; 1985 1986 switch (mode & S_IFMT) { 1987 case S_IFMT: /* used by badsect to flag bad sectors */ 1988 vattr.va_type = VBAD; 1989 break; 1990 case S_IFCHR: 1991 vattr.va_type = VCHR; 1992 break; 1993 case S_IFBLK: 1994 vattr.va_type = VBLK; 1995 break; 1996 case S_IFWHT: 1997 optype = VOP_WHITEOUT_DESCOFFSET; 1998 break; 1999 case S_IFREG: 2000 #if NVERIEXEC > 0 2001 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2002 O_CREAT); 2003 #endif /* NVERIEXEC > 0 */ 2004 vattr.va_type = VREG; 2005 vattr.va_rdev = VNOVAL; 2006 optype = VOP_CREATE_DESCOFFSET; 2007 break; 2008 default: 2009 error = EINVAL; 2010 break; 2011 } 2012 } 2013 if (!error) { 2014 switch (optype) { 2015 case VOP_WHITEOUT_DESCOFFSET: 2016 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2017 if (error) 2018 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2019 vput(nd.ni_dvp); 2020 break; 2021 2022 case VOP_MKNOD_DESCOFFSET: 2023 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2024 &nd.ni_cnd, &vattr); 2025 if (error == 0) 2026 vput(nd.ni_vp); 2027 break; 2028 2029 case VOP_CREATE_DESCOFFSET: 2030 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2031 &nd.ni_cnd, &vattr); 2032 if (error == 0) 2033 vput(nd.ni_vp); 2034 break; 2035 } 2036 } else { 2037 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2038 if (nd.ni_dvp == vp) 2039 vrele(nd.ni_dvp); 2040 else 2041 vput(nd.ni_dvp); 2042 if (vp) 2043 vrele(vp); 2044 } 2045 out: 2046 pathbuf_stringcopy_put(pb, pathstring); 2047 pathbuf_destroy(pb); 2048 return (error); 2049 } 2050 2051 /* 2052 * Create a named pipe. 2053 */ 2054 /* ARGSUSED */ 2055 int 2056 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2057 { 2058 /* { 2059 syscallarg(const char *) path; 2060 syscallarg(int) mode; 2061 } */ 2062 struct proc *p = l->l_proc; 2063 struct vattr vattr; 2064 int error; 2065 struct pathbuf *pb; 2066 struct nameidata nd; 2067 2068 error = pathbuf_copyin(SCARG(uap, path), &pb); 2069 if (error) { 2070 return error; 2071 } 2072 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2073 if ((error = namei(&nd)) != 0) { 2074 pathbuf_destroy(pb); 2075 return error; 2076 } 2077 if (nd.ni_vp != NULL) { 2078 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2079 if (nd.ni_dvp == nd.ni_vp) 2080 vrele(nd.ni_dvp); 2081 else 2082 vput(nd.ni_dvp); 2083 vrele(nd.ni_vp); 2084 pathbuf_destroy(pb); 2085 return (EEXIST); 2086 } 2087 vattr_null(&vattr); 2088 vattr.va_type = VFIFO; 2089 /* We will read cwdi->cwdi_cmask unlocked. */ 2090 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2091 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2092 if (error == 0) 2093 vput(nd.ni_vp); 2094 pathbuf_destroy(pb); 2095 return (error); 2096 } 2097 2098 /* 2099 * Make a hard file link. 2100 */ 2101 /* ARGSUSED */ 2102 int 2103 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2104 { 2105 /* { 2106 syscallarg(const char *) path; 2107 syscallarg(const char *) link; 2108 } */ 2109 struct vnode *vp; 2110 struct pathbuf *linkpb; 2111 struct nameidata nd; 2112 int error; 2113 2114 error = namei_simple_user(SCARG(uap, path), 2115 NSM_FOLLOW_TRYEMULROOT, &vp); 2116 if (error != 0) 2117 return (error); 2118 error = pathbuf_copyin(SCARG(uap, link), &linkpb); 2119 if (error) { 2120 goto out1; 2121 } 2122 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2123 if ((error = namei(&nd)) != 0) 2124 goto out2; 2125 if (nd.ni_vp) { 2126 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2127 if (nd.ni_dvp == nd.ni_vp) 2128 vrele(nd.ni_dvp); 2129 else 2130 vput(nd.ni_dvp); 2131 vrele(nd.ni_vp); 2132 error = EEXIST; 2133 goto out2; 2134 } 2135 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2136 out2: 2137 pathbuf_destroy(linkpb); 2138 out1: 2139 vrele(vp); 2140 return (error); 2141 } 2142 2143 int 2144 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2145 { 2146 struct proc *p = curproc; 2147 struct vattr vattr; 2148 char *path; 2149 int error; 2150 struct pathbuf *linkpb; 2151 struct nameidata nd; 2152 2153 path = PNBUF_GET(); 2154 if (seg == UIO_USERSPACE) { 2155 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2156 goto out1; 2157 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2158 goto out1; 2159 } else { 2160 KASSERT(strlen(patharg) < MAXPATHLEN); 2161 strcpy(path, patharg); 2162 linkpb = pathbuf_create(link); 2163 if (linkpb == NULL) { 2164 error = ENOMEM; 2165 goto out1; 2166 } 2167 } 2168 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2169 if ((error = namei(&nd)) != 0) 2170 goto out2; 2171 if (nd.ni_vp) { 2172 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2173 if (nd.ni_dvp == nd.ni_vp) 2174 vrele(nd.ni_dvp); 2175 else 2176 vput(nd.ni_dvp); 2177 vrele(nd.ni_vp); 2178 error = EEXIST; 2179 goto out2; 2180 } 2181 vattr_null(&vattr); 2182 vattr.va_type = VLNK; 2183 /* We will read cwdi->cwdi_cmask unlocked. */ 2184 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2185 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2186 if (error == 0) 2187 vput(nd.ni_vp); 2188 out2: 2189 pathbuf_destroy(linkpb); 2190 out1: 2191 PNBUF_PUT(path); 2192 return (error); 2193 } 2194 2195 /* 2196 * Make a symbolic link. 2197 */ 2198 /* ARGSUSED */ 2199 int 2200 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2201 { 2202 /* { 2203 syscallarg(const char *) path; 2204 syscallarg(const char *) link; 2205 } */ 2206 2207 return do_sys_symlink(SCARG(uap, path), SCARG(uap, link), 2208 UIO_USERSPACE); 2209 } 2210 2211 /* 2212 * Delete a whiteout from the filesystem. 2213 */ 2214 /* ARGSUSED */ 2215 int 2216 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2217 { 2218 /* { 2219 syscallarg(const char *) path; 2220 } */ 2221 int error; 2222 struct pathbuf *pb; 2223 struct nameidata nd; 2224 2225 error = pathbuf_copyin(SCARG(uap, path), &pb); 2226 if (error) { 2227 return error; 2228 } 2229 2230 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2231 error = namei(&nd); 2232 if (error) { 2233 pathbuf_destroy(pb); 2234 return (error); 2235 } 2236 2237 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2238 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2239 if (nd.ni_dvp == nd.ni_vp) 2240 vrele(nd.ni_dvp); 2241 else 2242 vput(nd.ni_dvp); 2243 if (nd.ni_vp) 2244 vrele(nd.ni_vp); 2245 pathbuf_destroy(pb); 2246 return (EEXIST); 2247 } 2248 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2249 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2250 vput(nd.ni_dvp); 2251 pathbuf_destroy(pb); 2252 return (error); 2253 } 2254 2255 /* 2256 * Delete a name from the filesystem. 2257 */ 2258 /* ARGSUSED */ 2259 int 2260 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2261 { 2262 /* { 2263 syscallarg(const char *) path; 2264 } */ 2265 2266 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2267 } 2268 2269 int 2270 do_sys_unlink(const char *arg, enum uio_seg seg) 2271 { 2272 struct vnode *vp; 2273 int error; 2274 struct pathbuf *pb; 2275 struct nameidata nd; 2276 const char *pathstring; 2277 2278 error = pathbuf_maybe_copyin(arg, seg, &pb); 2279 if (error) { 2280 return error; 2281 } 2282 pathstring = pathbuf_stringcopy_get(pb); 2283 if (pathstring == NULL) { 2284 pathbuf_destroy(pb); 2285 return ENOMEM; 2286 } 2287 2288 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2289 if ((error = namei(&nd)) != 0) 2290 goto out; 2291 vp = nd.ni_vp; 2292 2293 /* 2294 * The root of a mounted filesystem cannot be deleted. 2295 */ 2296 if (vp->v_vflag & VV_ROOT) { 2297 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2298 if (nd.ni_dvp == vp) 2299 vrele(nd.ni_dvp); 2300 else 2301 vput(nd.ni_dvp); 2302 vput(vp); 2303 error = EBUSY; 2304 goto out; 2305 } 2306 2307 #if NVERIEXEC > 0 2308 /* Handle remove requests for veriexec entries. */ 2309 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2310 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2311 if (nd.ni_dvp == vp) 2312 vrele(nd.ni_dvp); 2313 else 2314 vput(nd.ni_dvp); 2315 vput(vp); 2316 goto out; 2317 } 2318 #endif /* NVERIEXEC > 0 */ 2319 2320 #ifdef FILEASSOC 2321 (void)fileassoc_file_delete(vp); 2322 #endif /* FILEASSOC */ 2323 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2324 out: 2325 pathbuf_stringcopy_put(pb, pathstring); 2326 pathbuf_destroy(pb); 2327 return (error); 2328 } 2329 2330 /* 2331 * Reposition read/write file offset. 2332 */ 2333 int 2334 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2335 { 2336 /* { 2337 syscallarg(int) fd; 2338 syscallarg(int) pad; 2339 syscallarg(off_t) offset; 2340 syscallarg(int) whence; 2341 } */ 2342 kauth_cred_t cred = l->l_cred; 2343 file_t *fp; 2344 struct vnode *vp; 2345 struct vattr vattr; 2346 off_t newoff; 2347 int error, fd; 2348 2349 fd = SCARG(uap, fd); 2350 2351 if ((fp = fd_getfile(fd)) == NULL) 2352 return (EBADF); 2353 2354 vp = fp->f_data; 2355 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2356 error = ESPIPE; 2357 goto out; 2358 } 2359 2360 switch (SCARG(uap, whence)) { 2361 case SEEK_CUR: 2362 newoff = fp->f_offset + SCARG(uap, offset); 2363 break; 2364 case SEEK_END: 2365 error = VOP_GETATTR(vp, &vattr, cred); 2366 if (error) { 2367 goto out; 2368 } 2369 newoff = SCARG(uap, offset) + vattr.va_size; 2370 break; 2371 case SEEK_SET: 2372 newoff = SCARG(uap, offset); 2373 break; 2374 default: 2375 error = EINVAL; 2376 goto out; 2377 } 2378 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2379 *(off_t *)retval = fp->f_offset = newoff; 2380 } 2381 out: 2382 fd_putfile(fd); 2383 return (error); 2384 } 2385 2386 /* 2387 * Positional read system call. 2388 */ 2389 int 2390 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2391 { 2392 /* { 2393 syscallarg(int) fd; 2394 syscallarg(void *) buf; 2395 syscallarg(size_t) nbyte; 2396 syscallarg(off_t) offset; 2397 } */ 2398 file_t *fp; 2399 struct vnode *vp; 2400 off_t offset; 2401 int error, fd = SCARG(uap, fd); 2402 2403 if ((fp = fd_getfile(fd)) == NULL) 2404 return (EBADF); 2405 2406 if ((fp->f_flag & FREAD) == 0) { 2407 fd_putfile(fd); 2408 return (EBADF); 2409 } 2410 2411 vp = fp->f_data; 2412 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2413 error = ESPIPE; 2414 goto out; 2415 } 2416 2417 offset = SCARG(uap, offset); 2418 2419 /* 2420 * XXX This works because no file systems actually 2421 * XXX take any action on the seek operation. 2422 */ 2423 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2424 goto out; 2425 2426 /* dofileread() will unuse the descriptor for us */ 2427 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2428 &offset, 0, retval)); 2429 2430 out: 2431 fd_putfile(fd); 2432 return (error); 2433 } 2434 2435 /* 2436 * Positional scatter read system call. 2437 */ 2438 int 2439 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2440 { 2441 /* { 2442 syscallarg(int) fd; 2443 syscallarg(const struct iovec *) iovp; 2444 syscallarg(int) iovcnt; 2445 syscallarg(off_t) offset; 2446 } */ 2447 off_t offset = SCARG(uap, offset); 2448 2449 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2450 SCARG(uap, iovcnt), &offset, 0, retval); 2451 } 2452 2453 /* 2454 * Positional write system call. 2455 */ 2456 int 2457 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2458 { 2459 /* { 2460 syscallarg(int) fd; 2461 syscallarg(const void *) buf; 2462 syscallarg(size_t) nbyte; 2463 syscallarg(off_t) offset; 2464 } */ 2465 file_t *fp; 2466 struct vnode *vp; 2467 off_t offset; 2468 int error, fd = SCARG(uap, fd); 2469 2470 if ((fp = fd_getfile(fd)) == NULL) 2471 return (EBADF); 2472 2473 if ((fp->f_flag & FWRITE) == 0) { 2474 fd_putfile(fd); 2475 return (EBADF); 2476 } 2477 2478 vp = fp->f_data; 2479 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2480 error = ESPIPE; 2481 goto out; 2482 } 2483 2484 offset = SCARG(uap, offset); 2485 2486 /* 2487 * XXX This works because no file systems actually 2488 * XXX take any action on the seek operation. 2489 */ 2490 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2491 goto out; 2492 2493 /* dofilewrite() will unuse the descriptor for us */ 2494 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2495 &offset, 0, retval)); 2496 2497 out: 2498 fd_putfile(fd); 2499 return (error); 2500 } 2501 2502 /* 2503 * Positional gather write system call. 2504 */ 2505 int 2506 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2507 { 2508 /* { 2509 syscallarg(int) fd; 2510 syscallarg(const struct iovec *) iovp; 2511 syscallarg(int) iovcnt; 2512 syscallarg(off_t) offset; 2513 } */ 2514 off_t offset = SCARG(uap, offset); 2515 2516 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2517 SCARG(uap, iovcnt), &offset, 0, retval); 2518 } 2519 2520 /* 2521 * Check access permissions. 2522 */ 2523 int 2524 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2525 { 2526 /* { 2527 syscallarg(const char *) path; 2528 syscallarg(int) flags; 2529 } */ 2530 kauth_cred_t cred; 2531 struct vnode *vp; 2532 int error, flags; 2533 struct pathbuf *pb; 2534 struct nameidata nd; 2535 2536 error = pathbuf_copyin(SCARG(uap, path), &pb); 2537 if (error) { 2538 return error; 2539 } 2540 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2541 2542 /* Override default credentials */ 2543 cred = kauth_cred_dup(l->l_cred); 2544 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2545 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2546 nd.ni_cnd.cn_cred = cred; 2547 2548 if ((error = namei(&nd)) != 0) { 2549 pathbuf_destroy(pb); 2550 goto out; 2551 } 2552 vp = nd.ni_vp; 2553 pathbuf_destroy(pb); 2554 2555 /* Flags == 0 means only check for existence. */ 2556 if (SCARG(uap, flags)) { 2557 flags = 0; 2558 if (SCARG(uap, flags) & R_OK) 2559 flags |= VREAD; 2560 if (SCARG(uap, flags) & W_OK) 2561 flags |= VWRITE; 2562 if (SCARG(uap, flags) & X_OK) 2563 flags |= VEXEC; 2564 2565 error = VOP_ACCESS(vp, flags, cred); 2566 if (!error && (flags & VWRITE)) 2567 error = vn_writechk(vp); 2568 } 2569 vput(vp); 2570 out: 2571 kauth_cred_free(cred); 2572 return (error); 2573 } 2574 2575 /* 2576 * Common code for all sys_stat functions, including compat versions. 2577 */ 2578 int 2579 do_sys_stat(const char *userpath, unsigned int nd_flags, struct stat *sb) 2580 { 2581 int error; 2582 struct pathbuf *pb; 2583 struct nameidata nd; 2584 2585 error = pathbuf_copyin(userpath, &pb); 2586 if (error) { 2587 return error; 2588 } 2589 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, pb); 2590 error = namei(&nd); 2591 if (error != 0) { 2592 pathbuf_destroy(pb); 2593 return error; 2594 } 2595 error = vn_stat(nd.ni_vp, sb); 2596 vput(nd.ni_vp); 2597 pathbuf_destroy(pb); 2598 return error; 2599 } 2600 2601 /* 2602 * Get file status; this version follows links. 2603 */ 2604 /* ARGSUSED */ 2605 int 2606 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2607 { 2608 /* { 2609 syscallarg(const char *) path; 2610 syscallarg(struct stat *) ub; 2611 } */ 2612 struct stat sb; 2613 int error; 2614 2615 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2616 if (error) 2617 return error; 2618 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2619 } 2620 2621 /* 2622 * Get file status; this version does not follow links. 2623 */ 2624 /* ARGSUSED */ 2625 int 2626 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2627 { 2628 /* { 2629 syscallarg(const char *) path; 2630 syscallarg(struct stat *) ub; 2631 } */ 2632 struct stat sb; 2633 int error; 2634 2635 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2636 if (error) 2637 return error; 2638 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2639 } 2640 2641 /* 2642 * Get configurable pathname variables. 2643 */ 2644 /* ARGSUSED */ 2645 int 2646 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2647 { 2648 /* { 2649 syscallarg(const char *) path; 2650 syscallarg(int) name; 2651 } */ 2652 int error; 2653 struct pathbuf *pb; 2654 struct nameidata nd; 2655 2656 error = pathbuf_copyin(SCARG(uap, path), &pb); 2657 if (error) { 2658 return error; 2659 } 2660 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2661 if ((error = namei(&nd)) != 0) { 2662 pathbuf_destroy(pb); 2663 return (error); 2664 } 2665 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2666 vput(nd.ni_vp); 2667 pathbuf_destroy(pb); 2668 return (error); 2669 } 2670 2671 /* 2672 * Return target name of a symbolic link. 2673 */ 2674 /* ARGSUSED */ 2675 int 2676 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2677 { 2678 /* { 2679 syscallarg(const char *) path; 2680 syscallarg(char *) buf; 2681 syscallarg(size_t) count; 2682 } */ 2683 struct vnode *vp; 2684 struct iovec aiov; 2685 struct uio auio; 2686 int error; 2687 struct pathbuf *pb; 2688 struct nameidata nd; 2689 2690 error = pathbuf_copyin(SCARG(uap, path), &pb); 2691 if (error) { 2692 return error; 2693 } 2694 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2695 if ((error = namei(&nd)) != 0) { 2696 pathbuf_destroy(pb); 2697 return error; 2698 } 2699 vp = nd.ni_vp; 2700 pathbuf_destroy(pb); 2701 if (vp->v_type != VLNK) 2702 error = EINVAL; 2703 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2704 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2705 aiov.iov_base = SCARG(uap, buf); 2706 aiov.iov_len = SCARG(uap, count); 2707 auio.uio_iov = &aiov; 2708 auio.uio_iovcnt = 1; 2709 auio.uio_offset = 0; 2710 auio.uio_rw = UIO_READ; 2711 KASSERT(l == curlwp); 2712 auio.uio_vmspace = l->l_proc->p_vmspace; 2713 auio.uio_resid = SCARG(uap, count); 2714 error = VOP_READLINK(vp, &auio, l->l_cred); 2715 } 2716 vput(vp); 2717 *retval = SCARG(uap, count) - auio.uio_resid; 2718 return (error); 2719 } 2720 2721 /* 2722 * Change flags of a file given a path name. 2723 */ 2724 /* ARGSUSED */ 2725 int 2726 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2727 { 2728 /* { 2729 syscallarg(const char *) path; 2730 syscallarg(u_long) flags; 2731 } */ 2732 struct vnode *vp; 2733 int error; 2734 2735 error = namei_simple_user(SCARG(uap, path), 2736 NSM_FOLLOW_TRYEMULROOT, &vp); 2737 if (error != 0) 2738 return (error); 2739 error = change_flags(vp, SCARG(uap, flags), l); 2740 vput(vp); 2741 return (error); 2742 } 2743 2744 /* 2745 * Change flags of a file given a file descriptor. 2746 */ 2747 /* ARGSUSED */ 2748 int 2749 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2750 { 2751 /* { 2752 syscallarg(int) fd; 2753 syscallarg(u_long) flags; 2754 } */ 2755 struct vnode *vp; 2756 file_t *fp; 2757 int error; 2758 2759 /* fd_getvnode() will use the descriptor for us */ 2760 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2761 return (error); 2762 vp = fp->f_data; 2763 error = change_flags(vp, SCARG(uap, flags), l); 2764 VOP_UNLOCK(vp); 2765 fd_putfile(SCARG(uap, fd)); 2766 return (error); 2767 } 2768 2769 /* 2770 * Change flags of a file given a path name; this version does 2771 * not follow links. 2772 */ 2773 int 2774 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2775 { 2776 /* { 2777 syscallarg(const char *) path; 2778 syscallarg(u_long) flags; 2779 } */ 2780 struct vnode *vp; 2781 int error; 2782 2783 error = namei_simple_user(SCARG(uap, path), 2784 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2785 if (error != 0) 2786 return (error); 2787 error = change_flags(vp, SCARG(uap, flags), l); 2788 vput(vp); 2789 return (error); 2790 } 2791 2792 /* 2793 * Common routine to change flags of a file. 2794 */ 2795 int 2796 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2797 { 2798 struct vattr vattr; 2799 int error; 2800 2801 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2802 /* 2803 * Non-superusers cannot change the flags on devices, even if they 2804 * own them. 2805 */ 2806 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2807 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2808 goto out; 2809 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2810 error = EINVAL; 2811 goto out; 2812 } 2813 } 2814 vattr_null(&vattr); 2815 vattr.va_flags = flags; 2816 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2817 out: 2818 return (error); 2819 } 2820 2821 /* 2822 * Change mode of a file given path name; this version follows links. 2823 */ 2824 /* ARGSUSED */ 2825 int 2826 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2827 { 2828 /* { 2829 syscallarg(const char *) path; 2830 syscallarg(int) mode; 2831 } */ 2832 int error; 2833 struct vnode *vp; 2834 2835 error = namei_simple_user(SCARG(uap, path), 2836 NSM_FOLLOW_TRYEMULROOT, &vp); 2837 if (error != 0) 2838 return (error); 2839 2840 error = change_mode(vp, SCARG(uap, mode), l); 2841 2842 vrele(vp); 2843 return (error); 2844 } 2845 2846 /* 2847 * Change mode of a file given a file descriptor. 2848 */ 2849 /* ARGSUSED */ 2850 int 2851 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2852 { 2853 /* { 2854 syscallarg(int) fd; 2855 syscallarg(int) mode; 2856 } */ 2857 file_t *fp; 2858 int error; 2859 2860 /* fd_getvnode() will use the descriptor for us */ 2861 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2862 return (error); 2863 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2864 fd_putfile(SCARG(uap, fd)); 2865 return (error); 2866 } 2867 2868 /* 2869 * Change mode of a file given path name; this version does not follow links. 2870 */ 2871 /* ARGSUSED */ 2872 int 2873 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2874 { 2875 /* { 2876 syscallarg(const char *) path; 2877 syscallarg(int) mode; 2878 } */ 2879 int error; 2880 struct vnode *vp; 2881 2882 error = namei_simple_user(SCARG(uap, path), 2883 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2884 if (error != 0) 2885 return (error); 2886 2887 error = change_mode(vp, SCARG(uap, mode), l); 2888 2889 vrele(vp); 2890 return (error); 2891 } 2892 2893 /* 2894 * Common routine to set mode given a vnode. 2895 */ 2896 static int 2897 change_mode(struct vnode *vp, int mode, struct lwp *l) 2898 { 2899 struct vattr vattr; 2900 int error; 2901 2902 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2903 vattr_null(&vattr); 2904 vattr.va_mode = mode & ALLPERMS; 2905 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2906 VOP_UNLOCK(vp); 2907 return (error); 2908 } 2909 2910 /* 2911 * Set ownership given a path name; this version follows links. 2912 */ 2913 /* ARGSUSED */ 2914 int 2915 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2916 { 2917 /* { 2918 syscallarg(const char *) path; 2919 syscallarg(uid_t) uid; 2920 syscallarg(gid_t) gid; 2921 } */ 2922 int error; 2923 struct vnode *vp; 2924 2925 error = namei_simple_user(SCARG(uap, path), 2926 NSM_FOLLOW_TRYEMULROOT, &vp); 2927 if (error != 0) 2928 return (error); 2929 2930 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2931 2932 vrele(vp); 2933 return (error); 2934 } 2935 2936 /* 2937 * Set ownership given a path name; this version follows links. 2938 * Provides POSIX semantics. 2939 */ 2940 /* ARGSUSED */ 2941 int 2942 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2943 { 2944 /* { 2945 syscallarg(const char *) path; 2946 syscallarg(uid_t) uid; 2947 syscallarg(gid_t) gid; 2948 } */ 2949 int error; 2950 struct vnode *vp; 2951 2952 error = namei_simple_user(SCARG(uap, path), 2953 NSM_FOLLOW_TRYEMULROOT, &vp); 2954 if (error != 0) 2955 return (error); 2956 2957 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2958 2959 vrele(vp); 2960 return (error); 2961 } 2962 2963 /* 2964 * Set ownership given a file descriptor. 2965 */ 2966 /* ARGSUSED */ 2967 int 2968 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2969 { 2970 /* { 2971 syscallarg(int) fd; 2972 syscallarg(uid_t) uid; 2973 syscallarg(gid_t) gid; 2974 } */ 2975 int error; 2976 file_t *fp; 2977 2978 /* fd_getvnode() will use the descriptor for us */ 2979 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2980 return (error); 2981 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2982 l, 0); 2983 fd_putfile(SCARG(uap, fd)); 2984 return (error); 2985 } 2986 2987 /* 2988 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2989 */ 2990 /* ARGSUSED */ 2991 int 2992 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2993 { 2994 /* { 2995 syscallarg(int) fd; 2996 syscallarg(uid_t) uid; 2997 syscallarg(gid_t) gid; 2998 } */ 2999 int error; 3000 file_t *fp; 3001 3002 /* fd_getvnode() will use the descriptor for us */ 3003 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3004 return (error); 3005 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 3006 l, 1); 3007 fd_putfile(SCARG(uap, fd)); 3008 return (error); 3009 } 3010 3011 /* 3012 * Set ownership given a path name; this version does not follow links. 3013 */ 3014 /* ARGSUSED */ 3015 int 3016 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3017 { 3018 /* { 3019 syscallarg(const char *) path; 3020 syscallarg(uid_t) uid; 3021 syscallarg(gid_t) gid; 3022 } */ 3023 int error; 3024 struct vnode *vp; 3025 3026 error = namei_simple_user(SCARG(uap, path), 3027 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3028 if (error != 0) 3029 return (error); 3030 3031 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3032 3033 vrele(vp); 3034 return (error); 3035 } 3036 3037 /* 3038 * Set ownership given a path name; this version does not follow links. 3039 * Provides POSIX/XPG semantics. 3040 */ 3041 /* ARGSUSED */ 3042 int 3043 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3044 { 3045 /* { 3046 syscallarg(const char *) path; 3047 syscallarg(uid_t) uid; 3048 syscallarg(gid_t) gid; 3049 } */ 3050 int error; 3051 struct vnode *vp; 3052 3053 error = namei_simple_user(SCARG(uap, path), 3054 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3055 if (error != 0) 3056 return (error); 3057 3058 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3059 3060 vrele(vp); 3061 return (error); 3062 } 3063 3064 /* 3065 * Common routine to set ownership given a vnode. 3066 */ 3067 static int 3068 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3069 int posix_semantics) 3070 { 3071 struct vattr vattr; 3072 mode_t newmode; 3073 int error; 3074 3075 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3076 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3077 goto out; 3078 3079 #define CHANGED(x) ((int)(x) != -1) 3080 newmode = vattr.va_mode; 3081 if (posix_semantics) { 3082 /* 3083 * POSIX/XPG semantics: if the caller is not the super-user, 3084 * clear set-user-id and set-group-id bits. Both POSIX and 3085 * the XPG consider the behaviour for calls by the super-user 3086 * implementation-defined; we leave the set-user-id and set- 3087 * group-id settings intact in that case. 3088 */ 3089 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 3090 NULL) != 0) 3091 newmode &= ~(S_ISUID | S_ISGID); 3092 } else { 3093 /* 3094 * NetBSD semantics: when changing owner and/or group, 3095 * clear the respective bit(s). 3096 */ 3097 if (CHANGED(uid)) 3098 newmode &= ~S_ISUID; 3099 if (CHANGED(gid)) 3100 newmode &= ~S_ISGID; 3101 } 3102 /* Update va_mode iff altered. */ 3103 if (vattr.va_mode == newmode) 3104 newmode = VNOVAL; 3105 3106 vattr_null(&vattr); 3107 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3108 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3109 vattr.va_mode = newmode; 3110 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3111 #undef CHANGED 3112 3113 out: 3114 VOP_UNLOCK(vp); 3115 return (error); 3116 } 3117 3118 /* 3119 * Set the access and modification times given a path name; this 3120 * version follows links. 3121 */ 3122 /* ARGSUSED */ 3123 int 3124 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3125 register_t *retval) 3126 { 3127 /* { 3128 syscallarg(const char *) path; 3129 syscallarg(const struct timeval *) tptr; 3130 } */ 3131 3132 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3133 SCARG(uap, tptr), UIO_USERSPACE); 3134 } 3135 3136 /* 3137 * Set the access and modification times given a file descriptor. 3138 */ 3139 /* ARGSUSED */ 3140 int 3141 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3142 register_t *retval) 3143 { 3144 /* { 3145 syscallarg(int) fd; 3146 syscallarg(const struct timeval *) tptr; 3147 } */ 3148 int error; 3149 file_t *fp; 3150 3151 /* fd_getvnode() will use the descriptor for us */ 3152 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3153 return (error); 3154 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3155 UIO_USERSPACE); 3156 fd_putfile(SCARG(uap, fd)); 3157 return (error); 3158 } 3159 3160 /* 3161 * Set the access and modification times given a path name; this 3162 * version does not follow links. 3163 */ 3164 int 3165 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3166 register_t *retval) 3167 { 3168 /* { 3169 syscallarg(const char *) path; 3170 syscallarg(const struct timeval *) tptr; 3171 } */ 3172 3173 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3174 SCARG(uap, tptr), UIO_USERSPACE); 3175 } 3176 3177 /* 3178 * Common routine to set access and modification times given a vnode. 3179 */ 3180 int 3181 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3182 const struct timeval *tptr, enum uio_seg seg) 3183 { 3184 struct vattr vattr; 3185 int error, dorele = 0; 3186 namei_simple_flags_t sflags; 3187 3188 bool vanull, setbirthtime; 3189 struct timespec ts[2]; 3190 3191 /* 3192 * I have checked all callers and they pass either FOLLOW, 3193 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3194 * is 0. More to the point, they don't pass anything else. 3195 * Let's keep it that way at least until the namei interfaces 3196 * are fully sanitized. 3197 */ 3198 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3199 sflags = (flag == FOLLOW) ? 3200 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3201 3202 if (tptr == NULL) { 3203 vanull = true; 3204 nanotime(&ts[0]); 3205 ts[1] = ts[0]; 3206 } else { 3207 struct timeval tv[2]; 3208 3209 vanull = false; 3210 if (seg != UIO_SYSSPACE) { 3211 error = copyin(tptr, tv, sizeof (tv)); 3212 if (error != 0) 3213 return error; 3214 tptr = tv; 3215 } 3216 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3217 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3218 } 3219 3220 if (vp == NULL) { 3221 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3222 error = namei_simple_user(path, sflags, &vp); 3223 if (error != 0) 3224 return error; 3225 dorele = 1; 3226 } 3227 3228 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3229 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3230 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3231 vattr_null(&vattr); 3232 vattr.va_atime = ts[0]; 3233 vattr.va_mtime = ts[1]; 3234 if (setbirthtime) 3235 vattr.va_birthtime = ts[1]; 3236 if (vanull) 3237 vattr.va_vaflags |= VA_UTIMES_NULL; 3238 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3239 VOP_UNLOCK(vp); 3240 3241 if (dorele != 0) 3242 vrele(vp); 3243 3244 return error; 3245 } 3246 3247 /* 3248 * Truncate a file given its path name. 3249 */ 3250 /* ARGSUSED */ 3251 int 3252 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3253 { 3254 /* { 3255 syscallarg(const char *) path; 3256 syscallarg(int) pad; 3257 syscallarg(off_t) length; 3258 } */ 3259 struct vnode *vp; 3260 struct vattr vattr; 3261 int error; 3262 3263 error = namei_simple_user(SCARG(uap, path), 3264 NSM_FOLLOW_TRYEMULROOT, &vp); 3265 if (error != 0) 3266 return (error); 3267 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3268 if (vp->v_type == VDIR) 3269 error = EISDIR; 3270 else if ((error = vn_writechk(vp)) == 0 && 3271 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3272 vattr_null(&vattr); 3273 vattr.va_size = SCARG(uap, length); 3274 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3275 } 3276 vput(vp); 3277 return (error); 3278 } 3279 3280 /* 3281 * Truncate a file given a file descriptor. 3282 */ 3283 /* ARGSUSED */ 3284 int 3285 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3286 { 3287 /* { 3288 syscallarg(int) fd; 3289 syscallarg(int) pad; 3290 syscallarg(off_t) length; 3291 } */ 3292 struct vattr vattr; 3293 struct vnode *vp; 3294 file_t *fp; 3295 int error; 3296 3297 /* fd_getvnode() will use the descriptor for us */ 3298 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3299 return (error); 3300 if ((fp->f_flag & FWRITE) == 0) { 3301 error = EINVAL; 3302 goto out; 3303 } 3304 vp = fp->f_data; 3305 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3306 if (vp->v_type == VDIR) 3307 error = EISDIR; 3308 else if ((error = vn_writechk(vp)) == 0) { 3309 vattr_null(&vattr); 3310 vattr.va_size = SCARG(uap, length); 3311 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3312 } 3313 VOP_UNLOCK(vp); 3314 out: 3315 fd_putfile(SCARG(uap, fd)); 3316 return (error); 3317 } 3318 3319 /* 3320 * Sync an open file. 3321 */ 3322 /* ARGSUSED */ 3323 int 3324 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3325 { 3326 /* { 3327 syscallarg(int) fd; 3328 } */ 3329 struct vnode *vp; 3330 file_t *fp; 3331 int error; 3332 3333 /* fd_getvnode() will use the descriptor for us */ 3334 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3335 return (error); 3336 vp = fp->f_data; 3337 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3338 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3339 VOP_UNLOCK(vp); 3340 fd_putfile(SCARG(uap, fd)); 3341 return (error); 3342 } 3343 3344 /* 3345 * Sync a range of file data. API modeled after that found in AIX. 3346 * 3347 * FDATASYNC indicates that we need only save enough metadata to be able 3348 * to re-read the written data. Note we duplicate AIX's requirement that 3349 * the file be open for writing. 3350 */ 3351 /* ARGSUSED */ 3352 int 3353 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3354 { 3355 /* { 3356 syscallarg(int) fd; 3357 syscallarg(int) flags; 3358 syscallarg(off_t) start; 3359 syscallarg(off_t) length; 3360 } */ 3361 struct vnode *vp; 3362 file_t *fp; 3363 int flags, nflags; 3364 off_t s, e, len; 3365 int error; 3366 3367 /* fd_getvnode() will use the descriptor for us */ 3368 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3369 return (error); 3370 3371 if ((fp->f_flag & FWRITE) == 0) { 3372 error = EBADF; 3373 goto out; 3374 } 3375 3376 flags = SCARG(uap, flags); 3377 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3378 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3379 error = EINVAL; 3380 goto out; 3381 } 3382 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3383 if (flags & FDATASYNC) 3384 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3385 else 3386 nflags = FSYNC_WAIT; 3387 if (flags & FDISKSYNC) 3388 nflags |= FSYNC_CACHE; 3389 3390 len = SCARG(uap, length); 3391 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3392 if (len) { 3393 s = SCARG(uap, start); 3394 e = s + len; 3395 if (e < s) { 3396 error = EINVAL; 3397 goto out; 3398 } 3399 } else { 3400 e = 0; 3401 s = 0; 3402 } 3403 3404 vp = fp->f_data; 3405 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3406 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3407 VOP_UNLOCK(vp); 3408 out: 3409 fd_putfile(SCARG(uap, fd)); 3410 return (error); 3411 } 3412 3413 /* 3414 * Sync the data of an open file. 3415 */ 3416 /* ARGSUSED */ 3417 int 3418 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3419 { 3420 /* { 3421 syscallarg(int) fd; 3422 } */ 3423 struct vnode *vp; 3424 file_t *fp; 3425 int error; 3426 3427 /* fd_getvnode() will use the descriptor for us */ 3428 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3429 return (error); 3430 if ((fp->f_flag & FWRITE) == 0) { 3431 fd_putfile(SCARG(uap, fd)); 3432 return (EBADF); 3433 } 3434 vp = fp->f_data; 3435 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3436 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3437 VOP_UNLOCK(vp); 3438 fd_putfile(SCARG(uap, fd)); 3439 return (error); 3440 } 3441 3442 /* 3443 * Rename files, (standard) BSD semantics frontend. 3444 */ 3445 /* ARGSUSED */ 3446 int 3447 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3448 { 3449 /* { 3450 syscallarg(const char *) from; 3451 syscallarg(const char *) to; 3452 } */ 3453 3454 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3455 } 3456 3457 /* 3458 * Rename files, POSIX semantics frontend. 3459 */ 3460 /* ARGSUSED */ 3461 int 3462 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3463 { 3464 /* { 3465 syscallarg(const char *) from; 3466 syscallarg(const char *) to; 3467 } */ 3468 3469 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3470 } 3471 3472 /* 3473 * Rename files. Source and destination must either both be directories, 3474 * or both not be directories. If target is a directory, it must be empty. 3475 * If `from' and `to' refer to the same object, the value of the `retain' 3476 * argument is used to determine whether `from' will be 3477 * 3478 * (retain == 0) deleted unless `from' and `to' refer to the same 3479 * object in the file system's name space (BSD). 3480 * (retain == 1) always retained (POSIX). 3481 */ 3482 int 3483 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3484 { 3485 struct vnode *tvp, *fvp, *tdvp; 3486 struct pathbuf *frompb, *topb; 3487 struct nameidata fromnd, tond; 3488 struct mount *fs; 3489 struct lwp *l = curlwp; 3490 struct proc *p; 3491 int error; 3492 3493 error = pathbuf_maybe_copyin(from, seg, &frompb); 3494 if (error) { 3495 return error; 3496 } 3497 error = pathbuf_maybe_copyin(to, seg, &topb); 3498 if (error) { 3499 pathbuf_destroy(frompb); 3500 return error; 3501 } 3502 3503 NDINIT(&fromnd, DELETE, LOCKPARENT | TRYEMULROOT | INRENAME, 3504 frompb); 3505 if ((error = namei(&fromnd)) != 0) { 3506 pathbuf_destroy(frompb); 3507 pathbuf_destroy(topb); 3508 return (error); 3509 } 3510 if (fromnd.ni_dvp != fromnd.ni_vp) 3511 VOP_UNLOCK(fromnd.ni_dvp); 3512 fvp = fromnd.ni_vp; 3513 3514 fs = fvp->v_mount; 3515 error = VFS_RENAMELOCK_ENTER(fs); 3516 if (error) { 3517 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3518 vrele(fromnd.ni_dvp); 3519 vrele(fvp); 3520 goto out1; 3521 } 3522 3523 /* 3524 * close, partially, yet another race - ideally we should only 3525 * go as far as getting fromnd.ni_dvp before getting the per-fs 3526 * lock, and then continue to get fromnd.ni_vp, but we can't do 3527 * that with namei as it stands. 3528 * 3529 * This still won't prevent rmdir from nuking fromnd.ni_vp 3530 * under us. The real fix is to get the locks in the right 3531 * order and do the lookups in the right places, but that's a 3532 * major rototill. 3533 * 3534 * Note: this logic (as well as this whole function) is cloned 3535 * in nfs_serv.c. Proceed accordingly. 3536 */ 3537 vrele(fvp); 3538 if ((fromnd.ni_cnd.cn_namelen == 1 && 3539 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3540 (fromnd.ni_cnd.cn_namelen == 2 && 3541 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3542 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3543 error = EINVAL; 3544 VFS_RENAMELOCK_EXIT(fs); 3545 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3546 vrele(fromnd.ni_dvp); 3547 goto out1; 3548 } 3549 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3550 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0); 3551 if (error) { 3552 VOP_UNLOCK(fromnd.ni_dvp); 3553 VFS_RENAMELOCK_EXIT(fs); 3554 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3555 vrele(fromnd.ni_dvp); 3556 goto out1; 3557 } 3558 VOP_UNLOCK(fromnd.ni_vp); 3559 if (fromnd.ni_dvp != fromnd.ni_vp) 3560 VOP_UNLOCK(fromnd.ni_dvp); 3561 fvp = fromnd.ni_vp; 3562 3563 NDINIT(&tond, RENAME, 3564 LOCKPARENT | LOCKLEAF | NOCACHE | TRYEMULROOT 3565 | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0), 3566 topb); 3567 if ((error = namei(&tond)) != 0) { 3568 VFS_RENAMELOCK_EXIT(fs); 3569 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3570 vrele(fromnd.ni_dvp); 3571 vrele(fvp); 3572 goto out1; 3573 } 3574 tdvp = tond.ni_dvp; 3575 tvp = tond.ni_vp; 3576 3577 if (tvp != NULL) { 3578 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3579 error = ENOTDIR; 3580 goto out; 3581 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3582 error = EISDIR; 3583 goto out; 3584 } 3585 } 3586 3587 if (fvp == tdvp) 3588 error = EINVAL; 3589 3590 /* 3591 * Source and destination refer to the same object. 3592 */ 3593 if (fvp == tvp) { 3594 if (retain) 3595 error = -1; 3596 else if (fromnd.ni_dvp == tdvp && 3597 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3598 !memcmp(fromnd.ni_cnd.cn_nameptr, 3599 tond.ni_cnd.cn_nameptr, 3600 fromnd.ni_cnd.cn_namelen)) 3601 error = -1; 3602 } 3603 3604 #if NVERIEXEC > 0 3605 if (!error) { 3606 char *f1, *f2; 3607 size_t f1_len; 3608 size_t f2_len; 3609 3610 f1_len = fromnd.ni_cnd.cn_namelen + 1; 3611 f1 = kmem_alloc(f1_len, KM_SLEEP); 3612 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len); 3613 3614 f2_len = tond.ni_cnd.cn_namelen + 1; 3615 f2 = kmem_alloc(f2_len, KM_SLEEP); 3616 strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len); 3617 3618 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3619 3620 kmem_free(f1, f1_len); 3621 kmem_free(f2, f2_len); 3622 } 3623 #endif /* NVERIEXEC > 0 */ 3624 3625 out: 3626 p = l->l_proc; 3627 if (!error) { 3628 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3629 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3630 VFS_RENAMELOCK_EXIT(fs); 3631 } else { 3632 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3633 if (tdvp == tvp) 3634 vrele(tdvp); 3635 else 3636 vput(tdvp); 3637 if (tvp) 3638 vput(tvp); 3639 VFS_RENAMELOCK_EXIT(fs); 3640 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3641 vrele(fromnd.ni_dvp); 3642 vrele(fvp); 3643 } 3644 out1: 3645 pathbuf_destroy(frompb); 3646 pathbuf_destroy(topb); 3647 return (error == -1 ? 0 : error); 3648 } 3649 3650 /* 3651 * Make a directory file. 3652 */ 3653 /* ARGSUSED */ 3654 int 3655 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3656 { 3657 /* { 3658 syscallarg(const char *) path; 3659 syscallarg(int) mode; 3660 } */ 3661 3662 return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE); 3663 } 3664 3665 int 3666 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 3667 { 3668 struct proc *p = curlwp->l_proc; 3669 struct vnode *vp; 3670 struct vattr vattr; 3671 int error; 3672 struct pathbuf *pb; 3673 struct nameidata nd; 3674 3675 /* XXX bollocks, should pass in a pathbuf */ 3676 error = pathbuf_maybe_copyin(path, seg, &pb); 3677 if (error) { 3678 return error; 3679 } 3680 3681 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 3682 if ((error = namei(&nd)) != 0) { 3683 pathbuf_destroy(pb); 3684 return (error); 3685 } 3686 vp = nd.ni_vp; 3687 if (vp != NULL) { 3688 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3689 if (nd.ni_dvp == vp) 3690 vrele(nd.ni_dvp); 3691 else 3692 vput(nd.ni_dvp); 3693 vrele(vp); 3694 pathbuf_destroy(pb); 3695 return (EEXIST); 3696 } 3697 vattr_null(&vattr); 3698 vattr.va_type = VDIR; 3699 /* We will read cwdi->cwdi_cmask unlocked. */ 3700 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3701 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3702 if (!error) 3703 vput(nd.ni_vp); 3704 pathbuf_destroy(pb); 3705 return (error); 3706 } 3707 3708 /* 3709 * Remove a directory file. 3710 */ 3711 /* ARGSUSED */ 3712 int 3713 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3714 { 3715 /* { 3716 syscallarg(const char *) path; 3717 } */ 3718 struct vnode *vp; 3719 int error; 3720 struct pathbuf *pb; 3721 struct nameidata nd; 3722 3723 error = pathbuf_copyin(SCARG(uap, path), &pb); 3724 if (error) { 3725 return error; 3726 } 3727 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 3728 if ((error = namei(&nd)) != 0) { 3729 pathbuf_destroy(pb); 3730 return error; 3731 } 3732 vp = nd.ni_vp; 3733 if (vp->v_type != VDIR) { 3734 error = ENOTDIR; 3735 goto out; 3736 } 3737 /* 3738 * No rmdir "." please. 3739 */ 3740 if (nd.ni_dvp == vp) { 3741 error = EINVAL; 3742 goto out; 3743 } 3744 /* 3745 * The root of a mounted filesystem cannot be deleted. 3746 */ 3747 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3748 error = EBUSY; 3749 goto out; 3750 } 3751 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3752 pathbuf_destroy(pb); 3753 return (error); 3754 3755 out: 3756 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3757 if (nd.ni_dvp == vp) 3758 vrele(nd.ni_dvp); 3759 else 3760 vput(nd.ni_dvp); 3761 vput(vp); 3762 pathbuf_destroy(pb); 3763 return (error); 3764 } 3765 3766 /* 3767 * Read a block of directory entries in a file system independent format. 3768 */ 3769 int 3770 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3771 { 3772 /* { 3773 syscallarg(int) fd; 3774 syscallarg(char *) buf; 3775 syscallarg(size_t) count; 3776 } */ 3777 file_t *fp; 3778 int error, done; 3779 3780 /* fd_getvnode() will use the descriptor for us */ 3781 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3782 return (error); 3783 if ((fp->f_flag & FREAD) == 0) { 3784 error = EBADF; 3785 goto out; 3786 } 3787 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3788 SCARG(uap, count), &done, l, 0, 0); 3789 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3790 *retval = done; 3791 out: 3792 fd_putfile(SCARG(uap, fd)); 3793 return (error); 3794 } 3795 3796 /* 3797 * Set the mode mask for creation of filesystem nodes. 3798 */ 3799 int 3800 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3801 { 3802 /* { 3803 syscallarg(mode_t) newmask; 3804 } */ 3805 struct proc *p = l->l_proc; 3806 struct cwdinfo *cwdi; 3807 3808 /* 3809 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3810 * important is that we serialize changes to the mask. The 3811 * rw_exit() will issue a write memory barrier on our behalf, 3812 * and force the changes out to other CPUs (as it must use an 3813 * atomic operation, draining the local CPU's store buffers). 3814 */ 3815 cwdi = p->p_cwdi; 3816 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3817 *retval = cwdi->cwdi_cmask; 3818 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3819 rw_exit(&cwdi->cwdi_lock); 3820 3821 return (0); 3822 } 3823 3824 int 3825 dorevoke(struct vnode *vp, kauth_cred_t cred) 3826 { 3827 struct vattr vattr; 3828 int error; 3829 3830 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3831 return error; 3832 if (kauth_cred_geteuid(cred) == vattr.va_uid || 3833 (error = kauth_authorize_generic(cred, 3834 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3835 VOP_REVOKE(vp, REVOKEALL); 3836 return (error); 3837 } 3838 3839 /* 3840 * Void all references to file by ripping underlying filesystem 3841 * away from vnode. 3842 */ 3843 /* ARGSUSED */ 3844 int 3845 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3846 { 3847 /* { 3848 syscallarg(const char *) path; 3849 } */ 3850 struct vnode *vp; 3851 int error; 3852 3853 error = namei_simple_user(SCARG(uap, path), 3854 NSM_FOLLOW_TRYEMULROOT, &vp); 3855 if (error != 0) 3856 return (error); 3857 error = dorevoke(vp, l->l_cred); 3858 vrele(vp); 3859 return (error); 3860 } 3861
Indexes created Fri Oct 03 02:09:57 GMT 2025