vfs_syscalls.c revision 1.350.2.8
1 /* $NetBSD: vfs_syscalls.c,v 1.350.2.8 2010/09/26 03:58:54 yamt Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.350.2.8 2010/09/26 03:58:54 yamt 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 error = EOPNOTSUPP; /* Needs translation */ 174 goto out; 175 } 176 177 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 178 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 179 if (error) 180 goto out; 181 182 if (vfs_busy(mp, NULL)) { 183 error = EPERM; 184 goto out; 185 } 186 187 mutex_enter(&mp->mnt_updating); 188 189 mp->mnt_flag &= ~MNT_OP_FLAGS; 190 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 191 192 /* 193 * Set the mount level flags. 194 */ 195 if (flags & MNT_RDONLY) 196 mp->mnt_flag |= MNT_RDONLY; 197 else if (mp->mnt_flag & MNT_RDONLY) 198 mp->mnt_iflag |= IMNT_WANTRDWR; 199 mp->mnt_flag &= 200 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 201 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 202 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 203 MNT_LOG); 204 mp->mnt_flag |= flags & 205 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 206 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 207 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 208 MNT_LOG | MNT_IGNORE); 209 210 error = VFS_MOUNT(mp, path, data, data_len); 211 212 if (error && data != NULL) { 213 int error2; 214 215 /* 216 * Update failed; let's try and see if it was an 217 * export request. For compat with 3.0 and earlier. 218 */ 219 error2 = vfs_hooks_reexport(mp, path, data); 220 221 /* 222 * Only update error code if the export request was 223 * understood but some problem occurred while 224 * processing it. 225 */ 226 if (error2 != EJUSTRETURN) 227 error = error2; 228 } 229 230 if (mp->mnt_iflag & IMNT_WANTRDWR) 231 mp->mnt_flag &= ~MNT_RDONLY; 232 if (error) 233 mp->mnt_flag = saved_flags; 234 mp->mnt_flag &= ~MNT_OP_FLAGS; 235 mp->mnt_iflag &= ~IMNT_WANTRDWR; 236 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 237 if (mp->mnt_syncer == NULL) 238 error = vfs_allocate_syncvnode(mp); 239 } else { 240 if (mp->mnt_syncer != NULL) 241 vfs_deallocate_syncvnode(mp); 242 } 243 mutex_exit(&mp->mnt_updating); 244 vfs_unbusy(mp, false, NULL); 245 246 out: 247 return (error); 248 } 249 250 static int 251 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 252 { 253 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 254 int error; 255 256 /* Copy file-system type from userspace. */ 257 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 258 if (error) { 259 /* 260 * Historically, filesystem types were identified by numbers. 261 * If we get an integer for the filesystem type instead of a 262 * string, we check to see if it matches one of the historic 263 * filesystem types. 264 */ 265 u_long fsindex = (u_long)fstype; 266 if (fsindex >= nmountcompatnames || 267 mountcompatnames[fsindex] == NULL) 268 return ENODEV; 269 strlcpy(fstypename, mountcompatnames[fsindex], 270 sizeof(fstypename)); 271 } 272 273 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 274 if (strcmp(fstypename, "ufs") == 0) 275 fstypename[0] = 'f'; 276 277 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 278 return 0; 279 280 /* If we can autoload a vfs module, try again */ 281 mutex_enter(&module_lock); 282 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 283 mutex_exit(&module_lock); 284 285 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 286 return 0; 287 288 return ENODEV; 289 } 290 291 static int 292 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 293 const char *path, int flags, void *data, size_t *data_len) 294 { 295 struct mount *mp; 296 struct vnode *vp = *vpp; 297 struct vattr va; 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 vn_lock(vp, LK_SHARED | LK_RETRY); 319 error = VOP_GETATTR(vp, &va, l->l_cred); 320 VOP_UNLOCK(vp); 321 if (error != 0) { 322 vfs_delref(vfsops); 323 return error; 324 } 325 if ((va.va_uid != kauth_cred_geteuid(l->l_cred) && 326 (error = kauth_authorize_generic(l->l_cred, 327 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 328 vfs_delref(vfsops); 329 return error; 330 } 331 332 if (flags & MNT_EXPORTED) { 333 vfs_delref(vfsops); 334 return EINVAL; 335 } 336 337 if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) { 338 vfs_delref(vfsops); 339 return ENOMEM; 340 } 341 342 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 343 344 /* 345 * The underlying file system may refuse the mount for 346 * various reasons. Allow the user to force it to happen. 347 * 348 * Set the mount level flags. 349 */ 350 mp->mnt_flag = flags & 351 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 352 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 353 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 354 MNT_LOG | MNT_IGNORE | MNT_RDONLY); 355 356 mutex_enter(&mp->mnt_updating); 357 error = VFS_MOUNT(mp, path, data, data_len); 358 mp->mnt_flag &= ~MNT_OP_FLAGS; 359 360 if (error != 0) 361 goto err_unmounted; 362 363 /* 364 * Validate and prepare the mount point. 365 */ 366 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, 367 UIO_USERSPACE, path); 368 error = namei(&nd); 369 if (error != 0) { 370 goto err_mounted; 371 } 372 if (nd.ni_vp != vp) { 373 vput(nd.ni_vp); 374 error = EINVAL; 375 goto err_mounted; 376 } 377 if (vp->v_mountedhere != NULL) { 378 vput(nd.ni_vp); 379 error = EBUSY; 380 goto err_mounted; 381 } 382 error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0); 383 if (error != 0) { 384 vput(nd.ni_vp); 385 goto err_mounted; 386 } 387 388 /* 389 * Put the new filesystem on the mount list after root. 390 */ 391 cache_purge(vp); 392 mp->mnt_iflag &= ~IMNT_WANTRDWR; 393 394 mutex_enter(&mountlist_lock); 395 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 396 mutex_exit(&mountlist_lock); 397 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 398 error = vfs_allocate_syncvnode(mp); 399 if (error == 0) 400 vp->v_mountedhere = mp; 401 vput(nd.ni_vp); 402 if (error != 0) 403 goto err_onmountlist; 404 405 checkdirs(vp); 406 mutex_exit(&mp->mnt_updating); 407 408 /* Hold an additional reference to the mount across VFS_START(). */ 409 vfs_unbusy(mp, true, NULL); 410 (void) VFS_STATVFS(mp, &mp->mnt_stat); 411 error = VFS_START(mp, 0); 412 if (error) 413 vrele(vp); 414 /* Drop reference held for VFS_START(). */ 415 vfs_destroy(mp); 416 *vpp = NULL; 417 return error; 418 419 err_onmountlist: 420 mutex_enter(&mountlist_lock); 421 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 422 mp->mnt_iflag |= IMNT_GONE; 423 mutex_exit(&mountlist_lock); 424 425 err_mounted: 426 if (VFS_UNMOUNT(mp, MNT_FORCE) != 0) 427 panic("Unmounting fresh file system failed"); 428 429 err_unmounted: 430 vp->v_mountedhere = NULL; 431 mutex_exit(&mp->mnt_updating); 432 vfs_unbusy(mp, false, NULL); 433 vfs_destroy(mp); 434 435 return error; 436 } 437 438 static int 439 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 440 void *data, size_t *data_len) 441 { 442 struct mount *mp; 443 int error; 444 445 /* If MNT_GETARGS is specified, it should be the only flag. */ 446 if (flags & ~MNT_GETARGS) 447 return EINVAL; 448 449 mp = vp->v_mount; 450 451 /* XXX: probably some notion of "can see" here if we want isolation. */ 452 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 453 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 454 if (error) 455 return error; 456 457 if ((vp->v_vflag & VV_ROOT) == 0) 458 return EINVAL; 459 460 if (vfs_busy(mp, NULL)) 461 return EPERM; 462 463 mutex_enter(&mp->mnt_updating); 464 mp->mnt_flag &= ~MNT_OP_FLAGS; 465 mp->mnt_flag |= MNT_GETARGS; 466 error = VFS_MOUNT(mp, path, data, data_len); 467 mp->mnt_flag &= ~MNT_OP_FLAGS; 468 mutex_exit(&mp->mnt_updating); 469 470 vfs_unbusy(mp, false, NULL); 471 return (error); 472 } 473 474 int 475 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 476 { 477 /* { 478 syscallarg(const char *) type; 479 syscallarg(const char *) path; 480 syscallarg(int) flags; 481 syscallarg(void *) data; 482 syscallarg(size_t) data_len; 483 } */ 484 485 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 486 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 487 SCARG(uap, data_len), retval); 488 } 489 490 int 491 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 492 const char *path, int flags, void *data, enum uio_seg data_seg, 493 size_t data_len, register_t *retval) 494 { 495 struct vnode *vp; 496 void *data_buf = data; 497 bool vfsopsrele = false; 498 int error; 499 500 /* XXX: The calling convention of this routine is totally bizarre */ 501 if (vfsops) 502 vfsopsrele = true; 503 504 /* 505 * Get vnode to be covered 506 */ 507 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 508 if (error != 0) { 509 vp = NULL; 510 goto done; 511 } 512 513 if (vfsops == NULL) { 514 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 515 vfsops = vp->v_mount->mnt_op; 516 } else { 517 /* 'type' is userspace */ 518 error = mount_get_vfsops(type, &vfsops); 519 if (error != 0) 520 goto done; 521 vfsopsrele = true; 522 } 523 } 524 525 if (data != NULL && data_seg == UIO_USERSPACE) { 526 if (data_len == 0) { 527 /* No length supplied, use default for filesystem */ 528 data_len = vfsops->vfs_min_mount_data; 529 if (data_len > VFS_MAX_MOUNT_DATA) { 530 error = EINVAL; 531 goto done; 532 } 533 /* 534 * Hopefully a longer buffer won't make copyin() fail. 535 * For compatibility with 3.0 and earlier. 536 */ 537 if (flags & MNT_UPDATE 538 && data_len < sizeof (struct mnt_export_args30)) 539 data_len = sizeof (struct mnt_export_args30); 540 } 541 data_buf = kmem_alloc(data_len, KM_SLEEP); 542 543 /* NFS needs the buffer even for mnt_getargs .... */ 544 error = copyin(data, data_buf, data_len); 545 if (error != 0) 546 goto done; 547 } 548 549 if (flags & MNT_GETARGS) { 550 if (data_len == 0) { 551 error = EINVAL; 552 goto done; 553 } 554 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 555 if (error != 0) 556 goto done; 557 if (data_seg == UIO_USERSPACE) 558 error = copyout(data_buf, data, data_len); 559 *retval = data_len; 560 } else if (flags & MNT_UPDATE) { 561 error = mount_update(l, vp, path, flags, data_buf, &data_len); 562 } else { 563 /* Locking is handled internally in mount_domount(). */ 564 KASSERT(vfsopsrele == true); 565 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 566 &data_len); 567 vfsopsrele = false; 568 } 569 570 done: 571 if (vfsopsrele) 572 vfs_delref(vfsops); 573 if (vp != NULL) { 574 vrele(vp); 575 } 576 if (data_buf != data) 577 kmem_free(data_buf, data_len); 578 return (error); 579 } 580 581 /* 582 * Scan all active processes to see if any of them have a current 583 * or root directory onto which the new filesystem has just been 584 * mounted. If so, replace them with the new mount point. 585 */ 586 void 587 checkdirs(struct vnode *olddp) 588 { 589 struct cwdinfo *cwdi; 590 struct vnode *newdp, *rele1, *rele2; 591 struct proc *p; 592 bool retry; 593 594 if (olddp->v_usecount == 1) 595 return; 596 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 597 panic("mount: lost mount"); 598 599 do { 600 retry = false; 601 mutex_enter(proc_lock); 602 PROCLIST_FOREACH(p, &allproc) { 603 if ((cwdi = p->p_cwdi) == NULL) 604 continue; 605 /* 606 * Can't change to the old directory any more, 607 * so even if we see a stale value it's not a 608 * problem. 609 */ 610 if (cwdi->cwdi_cdir != olddp && 611 cwdi->cwdi_rdir != olddp) 612 continue; 613 retry = true; 614 rele1 = NULL; 615 rele2 = NULL; 616 atomic_inc_uint(&cwdi->cwdi_refcnt); 617 mutex_exit(proc_lock); 618 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 619 if (cwdi->cwdi_cdir == olddp) { 620 rele1 = cwdi->cwdi_cdir; 621 vref(newdp); 622 cwdi->cwdi_cdir = newdp; 623 } 624 if (cwdi->cwdi_rdir == olddp) { 625 rele2 = cwdi->cwdi_rdir; 626 vref(newdp); 627 cwdi->cwdi_rdir = newdp; 628 } 629 rw_exit(&cwdi->cwdi_lock); 630 cwdfree(cwdi); 631 if (rele1 != NULL) 632 vrele(rele1); 633 if (rele2 != NULL) 634 vrele(rele2); 635 mutex_enter(proc_lock); 636 break; 637 } 638 mutex_exit(proc_lock); 639 } while (retry); 640 641 if (rootvnode == olddp) { 642 vrele(rootvnode); 643 vref(newdp); 644 rootvnode = newdp; 645 } 646 vput(newdp); 647 } 648 649 /* 650 * Unmount a file system. 651 * 652 * Note: unmount takes a path to the vnode mounted on as argument, 653 * not special file (as before). 654 */ 655 /* ARGSUSED */ 656 int 657 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 658 { 659 /* { 660 syscallarg(const char *) path; 661 syscallarg(int) flags; 662 } */ 663 struct vnode *vp; 664 struct mount *mp; 665 int error; 666 struct nameidata nd; 667 668 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 669 SCARG(uap, path)); 670 if ((error = namei(&nd)) != 0) 671 return (error); 672 vp = nd.ni_vp; 673 mp = vp->v_mount; 674 atomic_inc_uint(&mp->mnt_refcnt); 675 VOP_UNLOCK(vp); 676 677 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 678 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 679 if (error) { 680 vrele(vp); 681 vfs_destroy(mp); 682 return (error); 683 } 684 685 /* 686 * Don't allow unmounting the root file system. 687 */ 688 if (mp->mnt_flag & MNT_ROOTFS) { 689 vrele(vp); 690 vfs_destroy(mp); 691 return (EINVAL); 692 } 693 694 /* 695 * Must be the root of the filesystem 696 */ 697 if ((vp->v_vflag & VV_ROOT) == 0) { 698 vrele(vp); 699 vfs_destroy(mp); 700 return (EINVAL); 701 } 702 703 vrele(vp); 704 error = dounmount(mp, SCARG(uap, flags), l); 705 vfs_destroy(mp); 706 return error; 707 } 708 709 /* 710 * Do the actual file system unmount. File system is assumed to have 711 * been locked by the caller. 712 * 713 * => Caller hold reference to the mount, explicitly for dounmount(). 714 */ 715 int 716 dounmount(struct mount *mp, int flags, struct lwp *l) 717 { 718 struct vnode *coveredvp; 719 int error; 720 int async; 721 int used_syncer; 722 723 #if NVERIEXEC > 0 724 error = veriexec_unmountchk(mp); 725 if (error) 726 return (error); 727 #endif /* NVERIEXEC > 0 */ 728 729 /* 730 * XXX Freeze syncer. Must do this before locking the 731 * mount point. See dounmount() for details. 732 */ 733 mutex_enter(&syncer_mutex); 734 rw_enter(&mp->mnt_unmounting, RW_WRITER); 735 if ((mp->mnt_iflag & IMNT_GONE) != 0) { 736 rw_exit(&mp->mnt_unmounting); 737 mutex_exit(&syncer_mutex); 738 return ENOENT; 739 } 740 741 used_syncer = (mp->mnt_syncer != NULL); 742 743 /* 744 * XXX Syncer must be frozen when we get here. This should really 745 * be done on a per-mountpoint basis, but the syncer doesn't work 746 * like that. 747 * 748 * The caller of dounmount() must acquire syncer_mutex because 749 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 750 * order, and we must preserve that order to avoid deadlock. 751 * 752 * So, if the file system did not use the syncer, now is 753 * the time to release the syncer_mutex. 754 */ 755 if (used_syncer == 0) 756 mutex_exit(&syncer_mutex); 757 758 mp->mnt_iflag |= IMNT_UNMOUNT; 759 async = mp->mnt_flag & MNT_ASYNC; 760 mp->mnt_flag &= ~MNT_ASYNC; 761 cache_purgevfs(mp); /* remove cache entries for this file sys */ 762 if (mp->mnt_syncer != NULL) 763 vfs_deallocate_syncvnode(mp); 764 error = 0; 765 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 766 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 767 } 768 vfs_scrubvnlist(mp); 769 if (error == 0 || (flags & MNT_FORCE)) 770 error = VFS_UNMOUNT(mp, flags); 771 if (error) { 772 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 773 (void) vfs_allocate_syncvnode(mp); 774 mp->mnt_iflag &= ~IMNT_UNMOUNT; 775 mp->mnt_flag |= async; 776 rw_exit(&mp->mnt_unmounting); 777 if (used_syncer) 778 mutex_exit(&syncer_mutex); 779 return (error); 780 } 781 vfs_scrubvnlist(mp); 782 mutex_enter(&mountlist_lock); 783 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 784 coveredvp->v_mountedhere = NULL; 785 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 786 mp->mnt_iflag |= IMNT_GONE; 787 mutex_exit(&mountlist_lock); 788 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 789 panic("unmount: dangling vnode"); 790 if (used_syncer) 791 mutex_exit(&syncer_mutex); 792 vfs_hooks_unmount(mp); 793 rw_exit(&mp->mnt_unmounting); 794 vfs_destroy(mp); /* reference from mount() */ 795 if (coveredvp != NULLVP) 796 vrele(coveredvp); 797 return (0); 798 } 799 800 /* 801 * Sync each mounted filesystem. 802 */ 803 #ifdef DEBUG 804 int syncprt = 0; 805 struct ctldebug debug0 = { "syncprt", &syncprt }; 806 #endif 807 808 /* ARGSUSED */ 809 int 810 sys_sync(struct lwp *l, const void *v, register_t *retval) 811 { 812 struct mount *mp, *nmp; 813 int asyncflag; 814 815 if (l == NULL) 816 l = &lwp0; 817 818 mutex_enter(&mountlist_lock); 819 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 820 mp = nmp) { 821 if (vfs_busy(mp, &nmp)) { 822 continue; 823 } 824 mutex_enter(&mp->mnt_updating); 825 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 826 asyncflag = mp->mnt_flag & MNT_ASYNC; 827 mp->mnt_flag &= ~MNT_ASYNC; 828 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 829 if (asyncflag) 830 mp->mnt_flag |= MNT_ASYNC; 831 } 832 mutex_exit(&mp->mnt_updating); 833 vfs_unbusy(mp, false, &nmp); 834 } 835 mutex_exit(&mountlist_lock); 836 #ifdef DEBUG 837 if (syncprt) 838 vfs_bufstats(); 839 #endif /* DEBUG */ 840 return (0); 841 } 842 843 /* 844 * Change filesystem quotas. 845 */ 846 /* ARGSUSED */ 847 int 848 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 849 { 850 /* { 851 syscallarg(const char *) path; 852 syscallarg(int) cmd; 853 syscallarg(int) uid; 854 syscallarg(void *) arg; 855 } */ 856 struct mount *mp; 857 int error; 858 struct vnode *vp; 859 860 error = namei_simple_user(SCARG(uap, path), 861 NSM_FOLLOW_TRYEMULROOT, &vp); 862 if (error != 0) 863 return (error); 864 mp = vp->v_mount; 865 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 866 SCARG(uap, arg)); 867 vrele(vp); 868 return (error); 869 } 870 871 int 872 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 873 int root) 874 { 875 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 876 int error = 0; 877 878 /* 879 * If MNT_NOWAIT or MNT_LAZY is specified, do not 880 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 881 * overrides MNT_NOWAIT. 882 */ 883 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 884 (flags != MNT_WAIT && flags != 0)) { 885 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 886 goto done; 887 } 888 889 /* Get the filesystem stats now */ 890 memset(sp, 0, sizeof(*sp)); 891 if ((error = VFS_STATVFS(mp, sp)) != 0) { 892 return error; 893 } 894 895 if (cwdi->cwdi_rdir == NULL) 896 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 897 done: 898 if (cwdi->cwdi_rdir != NULL) { 899 size_t len; 900 char *bp; 901 char c; 902 char *path = PNBUF_GET(); 903 904 bp = path + MAXPATHLEN; 905 *--bp = '\0'; 906 rw_enter(&cwdi->cwdi_lock, RW_READER); 907 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 908 MAXPATHLEN / 2, 0, l); 909 rw_exit(&cwdi->cwdi_lock); 910 if (error) { 911 PNBUF_PUT(path); 912 return error; 913 } 914 len = strlen(bp); 915 if (len != 1) { 916 /* 917 * for mount points that are below our root, we can see 918 * them, so we fix up the pathname and return them. The 919 * rest we cannot see, so we don't allow viewing the 920 * data. 921 */ 922 if (strncmp(bp, sp->f_mntonname, len) == 0 && 923 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 924 (void)strlcpy(sp->f_mntonname, 925 c == '\0' ? "/" : &sp->f_mntonname[len], 926 sizeof(sp->f_mntonname)); 927 } else { 928 if (root) 929 (void)strlcpy(sp->f_mntonname, "/", 930 sizeof(sp->f_mntonname)); 931 else 932 error = EPERM; 933 } 934 } 935 PNBUF_PUT(path); 936 } 937 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 938 return error; 939 } 940 941 /* 942 * Get filesystem statistics by path. 943 */ 944 int 945 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 946 { 947 struct mount *mp; 948 int error; 949 struct vnode *vp; 950 951 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 952 if (error != 0) 953 return error; 954 mp = vp->v_mount; 955 error = dostatvfs(mp, sb, l, flags, 1); 956 vrele(vp); 957 return error; 958 } 959 960 /* ARGSUSED */ 961 int 962 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 963 { 964 /* { 965 syscallarg(const char *) path; 966 syscallarg(struct statvfs *) buf; 967 syscallarg(int) flags; 968 } */ 969 struct statvfs *sb; 970 int error; 971 972 sb = STATVFSBUF_GET(); 973 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 974 if (error == 0) 975 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 976 STATVFSBUF_PUT(sb); 977 return error; 978 } 979 980 /* 981 * Get filesystem statistics by fd. 982 */ 983 int 984 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 985 { 986 file_t *fp; 987 struct mount *mp; 988 int error; 989 990 /* fd_getvnode() will use the descriptor for us */ 991 if ((error = fd_getvnode(fd, &fp)) != 0) 992 return (error); 993 mp = ((struct vnode *)fp->f_data)->v_mount; 994 error = dostatvfs(mp, sb, curlwp, flags, 1); 995 fd_putfile(fd); 996 return error; 997 } 998 999 /* ARGSUSED */ 1000 int 1001 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1002 { 1003 /* { 1004 syscallarg(int) fd; 1005 syscallarg(struct statvfs *) buf; 1006 syscallarg(int) flags; 1007 } */ 1008 struct statvfs *sb; 1009 int error; 1010 1011 sb = STATVFSBUF_GET(); 1012 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1013 if (error == 0) 1014 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1015 STATVFSBUF_PUT(sb); 1016 return error; 1017 } 1018 1019 1020 /* 1021 * Get statistics on all filesystems. 1022 */ 1023 int 1024 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1025 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1026 register_t *retval) 1027 { 1028 int root = 0; 1029 struct proc *p = l->l_proc; 1030 struct mount *mp, *nmp; 1031 struct statvfs *sb; 1032 size_t count, maxcount; 1033 int error = 0; 1034 1035 sb = STATVFSBUF_GET(); 1036 maxcount = bufsize / entry_sz; 1037 mutex_enter(&mountlist_lock); 1038 count = 0; 1039 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1040 mp = nmp) { 1041 if (vfs_busy(mp, &nmp)) { 1042 continue; 1043 } 1044 if (sfsp && count < maxcount) { 1045 error = dostatvfs(mp, sb, l, flags, 0); 1046 if (error) { 1047 vfs_unbusy(mp, false, &nmp); 1048 error = 0; 1049 continue; 1050 } 1051 error = copyfn(sb, sfsp, entry_sz); 1052 if (error) { 1053 vfs_unbusy(mp, false, NULL); 1054 goto out; 1055 } 1056 sfsp = (char *)sfsp + entry_sz; 1057 root |= strcmp(sb->f_mntonname, "/") == 0; 1058 } 1059 count++; 1060 vfs_unbusy(mp, false, &nmp); 1061 } 1062 mutex_exit(&mountlist_lock); 1063 1064 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1065 /* 1066 * fake a root entry 1067 */ 1068 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1069 sb, l, flags, 1); 1070 if (error != 0) 1071 goto out; 1072 if (sfsp) { 1073 error = copyfn(sb, sfsp, entry_sz); 1074 if (error != 0) 1075 goto out; 1076 } 1077 count++; 1078 } 1079 if (sfsp && count > maxcount) 1080 *retval = maxcount; 1081 else 1082 *retval = count; 1083 out: 1084 STATVFSBUF_PUT(sb); 1085 return error; 1086 } 1087 1088 int 1089 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1090 { 1091 /* { 1092 syscallarg(struct statvfs *) buf; 1093 syscallarg(size_t) bufsize; 1094 syscallarg(int) flags; 1095 } */ 1096 1097 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1098 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1099 } 1100 1101 /* 1102 * Change current working directory to a given file descriptor. 1103 */ 1104 /* ARGSUSED */ 1105 int 1106 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1107 { 1108 /* { 1109 syscallarg(int) fd; 1110 } */ 1111 struct proc *p = l->l_proc; 1112 struct cwdinfo *cwdi; 1113 struct vnode *vp, *tdp; 1114 struct mount *mp; 1115 file_t *fp; 1116 int error, fd; 1117 1118 /* fd_getvnode() will use the descriptor for us */ 1119 fd = SCARG(uap, fd); 1120 if ((error = fd_getvnode(fd, &fp)) != 0) 1121 return (error); 1122 vp = fp->f_data; 1123 1124 vref(vp); 1125 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1126 if (vp->v_type != VDIR) 1127 error = ENOTDIR; 1128 else 1129 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1130 if (error) { 1131 vput(vp); 1132 goto out; 1133 } 1134 while ((mp = vp->v_mountedhere) != NULL) { 1135 error = vfs_busy(mp, NULL); 1136 vput(vp); 1137 if (error != 0) 1138 goto out; 1139 error = VFS_ROOT(mp, &tdp); 1140 vfs_unbusy(mp, false, NULL); 1141 if (error) 1142 goto out; 1143 vp = tdp; 1144 } 1145 VOP_UNLOCK(vp); 1146 1147 /* 1148 * Disallow changing to a directory not under the process's 1149 * current root directory (if there is one). 1150 */ 1151 cwdi = p->p_cwdi; 1152 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1153 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1154 vrele(vp); 1155 error = EPERM; /* operation not permitted */ 1156 } else { 1157 vrele(cwdi->cwdi_cdir); 1158 cwdi->cwdi_cdir = vp; 1159 } 1160 rw_exit(&cwdi->cwdi_lock); 1161 1162 out: 1163 fd_putfile(fd); 1164 return (error); 1165 } 1166 1167 /* 1168 * Change this process's notion of the root directory to a given file 1169 * descriptor. 1170 */ 1171 int 1172 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1173 { 1174 struct proc *p = l->l_proc; 1175 struct vnode *vp; 1176 file_t *fp; 1177 int error, fd = SCARG(uap, fd); 1178 1179 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1180 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1181 return error; 1182 /* fd_getvnode() will use the descriptor for us */ 1183 if ((error = fd_getvnode(fd, &fp)) != 0) 1184 return error; 1185 vp = fp->f_data; 1186 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1187 if (vp->v_type != VDIR) 1188 error = ENOTDIR; 1189 else 1190 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1191 VOP_UNLOCK(vp); 1192 if (error) 1193 goto out; 1194 vref(vp); 1195 1196 change_root(p->p_cwdi, vp, l); 1197 1198 out: 1199 fd_putfile(fd); 1200 return (error); 1201 } 1202 1203 /* 1204 * Change current working directory (``.''). 1205 */ 1206 /* ARGSUSED */ 1207 int 1208 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1209 { 1210 /* { 1211 syscallarg(const char *) path; 1212 } */ 1213 struct proc *p = l->l_proc; 1214 struct cwdinfo *cwdi; 1215 int error; 1216 struct vnode *vp; 1217 1218 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1219 &vp, l)) != 0) 1220 return (error); 1221 cwdi = p->p_cwdi; 1222 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1223 vrele(cwdi->cwdi_cdir); 1224 cwdi->cwdi_cdir = vp; 1225 rw_exit(&cwdi->cwdi_lock); 1226 return (0); 1227 } 1228 1229 /* 1230 * Change notion of root (``/'') directory. 1231 */ 1232 /* ARGSUSED */ 1233 int 1234 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1235 { 1236 /* { 1237 syscallarg(const char *) path; 1238 } */ 1239 struct proc *p = l->l_proc; 1240 int error; 1241 struct vnode *vp; 1242 1243 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1244 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1245 return (error); 1246 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1247 &vp, l)) != 0) 1248 return (error); 1249 1250 change_root(p->p_cwdi, vp, l); 1251 1252 return (0); 1253 } 1254 1255 /* 1256 * Common routine for chroot and fchroot. 1257 * NB: callers need to properly authorize the change root operation. 1258 */ 1259 void 1260 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1261 { 1262 1263 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1264 if (cwdi->cwdi_rdir != NULL) 1265 vrele(cwdi->cwdi_rdir); 1266 cwdi->cwdi_rdir = vp; 1267 1268 /* 1269 * Prevent escaping from chroot by putting the root under 1270 * the working directory. Silently chdir to / if we aren't 1271 * already there. 1272 */ 1273 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1274 /* 1275 * XXX would be more failsafe to change directory to a 1276 * deadfs node here instead 1277 */ 1278 vrele(cwdi->cwdi_cdir); 1279 vref(vp); 1280 cwdi->cwdi_cdir = vp; 1281 } 1282 rw_exit(&cwdi->cwdi_lock); 1283 } 1284 1285 /* 1286 * Common routine for chroot and chdir. 1287 */ 1288 int 1289 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1290 { 1291 struct nameidata nd; 1292 int error; 1293 1294 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, where, 1295 path); 1296 if ((error = namei(&nd)) != 0) 1297 return (error); 1298 *vpp = nd.ni_vp; 1299 if ((*vpp)->v_type != VDIR) 1300 error = ENOTDIR; 1301 else 1302 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1303 1304 if (error) 1305 vput(*vpp); 1306 else 1307 VOP_UNLOCK(*vpp); 1308 return (error); 1309 } 1310 1311 /* 1312 * Check permissions, allocate an open file structure, 1313 * and call the device open routine if any. 1314 */ 1315 int 1316 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1317 { 1318 /* { 1319 syscallarg(const char *) path; 1320 syscallarg(int) flags; 1321 syscallarg(int) mode; 1322 } */ 1323 struct proc *p = l->l_proc; 1324 struct cwdinfo *cwdi = p->p_cwdi; 1325 file_t *fp; 1326 struct vnode *vp; 1327 int flags, cmode; 1328 int type, indx, error; 1329 struct flock lf; 1330 struct nameidata nd; 1331 1332 flags = FFLAGS(SCARG(uap, flags)); 1333 if ((flags & (FREAD | FWRITE)) == 0) 1334 return (EINVAL); 1335 if ((error = fd_allocfile(&fp, &indx)) != 0) 1336 return (error); 1337 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1338 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1339 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1340 SCARG(uap, path)); 1341 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1342 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1343 fd_abort(p, fp, indx); 1344 if ((error == EDUPFD || error == EMOVEFD) && 1345 l->l_dupfd >= 0 && /* XXX from fdopen */ 1346 (error = 1347 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1348 *retval = indx; 1349 return (0); 1350 } 1351 if (error == ERESTART) 1352 error = EINTR; 1353 return (error); 1354 } 1355 1356 l->l_dupfd = 0; 1357 vp = nd.ni_vp; 1358 fp->f_flag = flags & FMASK; 1359 fp->f_type = DTYPE_VNODE; 1360 fp->f_ops = &vnops; 1361 fp->f_data = vp; 1362 if (flags & (O_EXLOCK | O_SHLOCK)) { 1363 lf.l_whence = SEEK_SET; 1364 lf.l_start = 0; 1365 lf.l_len = 0; 1366 if (flags & O_EXLOCK) 1367 lf.l_type = F_WRLCK; 1368 else 1369 lf.l_type = F_RDLCK; 1370 type = F_FLOCK; 1371 if ((flags & FNONBLOCK) == 0) 1372 type |= F_WAIT; 1373 VOP_UNLOCK(vp); 1374 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1375 if (error) { 1376 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1377 fd_abort(p, fp, indx); 1378 return (error); 1379 } 1380 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1381 atomic_or_uint(&fp->f_flag, FHASLOCK); 1382 } 1383 VOP_UNLOCK(vp); 1384 *retval = indx; 1385 fd_affix(p, fp, indx); 1386 return (0); 1387 } 1388 1389 static void 1390 vfs__fhfree(fhandle_t *fhp) 1391 { 1392 size_t fhsize; 1393 1394 if (fhp == NULL) { 1395 return; 1396 } 1397 fhsize = FHANDLE_SIZE(fhp); 1398 kmem_free(fhp, fhsize); 1399 } 1400 1401 /* 1402 * vfs_composefh: compose a filehandle. 1403 */ 1404 1405 int 1406 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1407 { 1408 struct mount *mp; 1409 struct fid *fidp; 1410 int error; 1411 size_t needfhsize; 1412 size_t fidsize; 1413 1414 mp = vp->v_mount; 1415 fidp = NULL; 1416 if (*fh_size < FHANDLE_SIZE_MIN) { 1417 fidsize = 0; 1418 } else { 1419 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1420 if (fhp != NULL) { 1421 memset(fhp, 0, *fh_size); 1422 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1423 fidp = &fhp->fh_fid; 1424 } 1425 } 1426 error = VFS_VPTOFH(vp, fidp, &fidsize); 1427 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1428 if (error == 0 && *fh_size < needfhsize) { 1429 error = E2BIG; 1430 } 1431 *fh_size = needfhsize; 1432 return error; 1433 } 1434 1435 int 1436 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1437 { 1438 struct mount *mp; 1439 fhandle_t *fhp; 1440 size_t fhsize; 1441 size_t fidsize; 1442 int error; 1443 1444 *fhpp = NULL; 1445 mp = vp->v_mount; 1446 fidsize = 0; 1447 error = VFS_VPTOFH(vp, NULL, &fidsize); 1448 KASSERT(error != 0); 1449 if (error != E2BIG) { 1450 goto out; 1451 } 1452 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1453 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1454 if (fhp == NULL) { 1455 error = ENOMEM; 1456 goto out; 1457 } 1458 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1459 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1460 if (error == 0) { 1461 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1462 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1463 *fhpp = fhp; 1464 } else { 1465 kmem_free(fhp, fhsize); 1466 } 1467 out: 1468 return error; 1469 } 1470 1471 void 1472 vfs_composefh_free(fhandle_t *fhp) 1473 { 1474 1475 vfs__fhfree(fhp); 1476 } 1477 1478 /* 1479 * vfs_fhtovp: lookup a vnode by a filehandle. 1480 */ 1481 1482 int 1483 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1484 { 1485 struct mount *mp; 1486 int error; 1487 1488 *vpp = NULL; 1489 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1490 if (mp == NULL) { 1491 error = ESTALE; 1492 goto out; 1493 } 1494 if (mp->mnt_op->vfs_fhtovp == NULL) { 1495 error = EOPNOTSUPP; 1496 goto out; 1497 } 1498 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1499 out: 1500 return error; 1501 } 1502 1503 /* 1504 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1505 * the needed size. 1506 */ 1507 1508 int 1509 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1510 { 1511 fhandle_t *fhp; 1512 int error; 1513 1514 *fhpp = NULL; 1515 if (fhsize > FHANDLE_SIZE_MAX) { 1516 return EINVAL; 1517 } 1518 if (fhsize < FHANDLE_SIZE_MIN) { 1519 return EINVAL; 1520 } 1521 again: 1522 fhp = kmem_alloc(fhsize, KM_SLEEP); 1523 if (fhp == NULL) { 1524 return ENOMEM; 1525 } 1526 error = copyin(ufhp, fhp, fhsize); 1527 if (error == 0) { 1528 /* XXX this check shouldn't be here */ 1529 if (FHANDLE_SIZE(fhp) == fhsize) { 1530 *fhpp = fhp; 1531 return 0; 1532 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1533 /* 1534 * a kludge for nfsv2 padded handles. 1535 */ 1536 size_t sz; 1537 1538 sz = FHANDLE_SIZE(fhp); 1539 kmem_free(fhp, fhsize); 1540 fhsize = sz; 1541 goto again; 1542 } else { 1543 /* 1544 * userland told us wrong size. 1545 */ 1546 error = EINVAL; 1547 } 1548 } 1549 kmem_free(fhp, fhsize); 1550 return error; 1551 } 1552 1553 void 1554 vfs_copyinfh_free(fhandle_t *fhp) 1555 { 1556 1557 vfs__fhfree(fhp); 1558 } 1559 1560 /* 1561 * Get file handle system call 1562 */ 1563 int 1564 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1565 { 1566 /* { 1567 syscallarg(char *) fname; 1568 syscallarg(fhandle_t *) fhp; 1569 syscallarg(size_t *) fh_size; 1570 } */ 1571 struct vnode *vp; 1572 fhandle_t *fh; 1573 int error; 1574 struct nameidata nd; 1575 size_t sz; 1576 size_t usz; 1577 1578 /* 1579 * Must be super user 1580 */ 1581 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1582 0, NULL, NULL, NULL); 1583 if (error) 1584 return (error); 1585 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1586 SCARG(uap, fname)); 1587 error = namei(&nd); 1588 if (error) 1589 return (error); 1590 vp = nd.ni_vp; 1591 error = vfs_composefh_alloc(vp, &fh); 1592 vput(vp); 1593 if (error != 0) { 1594 goto out; 1595 } 1596 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1597 if (error != 0) { 1598 goto out; 1599 } 1600 sz = FHANDLE_SIZE(fh); 1601 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1602 if (error != 0) { 1603 goto out; 1604 } 1605 if (usz >= sz) { 1606 error = copyout(fh, SCARG(uap, fhp), sz); 1607 } else { 1608 error = E2BIG; 1609 } 1610 out: 1611 vfs_composefh_free(fh); 1612 return (error); 1613 } 1614 1615 /* 1616 * Open a file given a file handle. 1617 * 1618 * Check permissions, allocate an open file structure, 1619 * and call the device open routine if any. 1620 */ 1621 1622 int 1623 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1624 register_t *retval) 1625 { 1626 file_t *fp; 1627 struct vnode *vp = NULL; 1628 kauth_cred_t cred = l->l_cred; 1629 file_t *nfp; 1630 int type, indx, error=0; 1631 struct flock lf; 1632 struct vattr va; 1633 fhandle_t *fh; 1634 int flags; 1635 proc_t *p; 1636 1637 p = curproc; 1638 1639 /* 1640 * Must be super user 1641 */ 1642 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1643 0, NULL, NULL, NULL))) 1644 return (error); 1645 1646 flags = FFLAGS(oflags); 1647 if ((flags & (FREAD | FWRITE)) == 0) 1648 return (EINVAL); 1649 if ((flags & O_CREAT)) 1650 return (EINVAL); 1651 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1652 return (error); 1653 fp = nfp; 1654 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1655 if (error != 0) { 1656 goto bad; 1657 } 1658 error = vfs_fhtovp(fh, &vp); 1659 if (error != 0) { 1660 goto bad; 1661 } 1662 1663 /* Now do an effective vn_open */ 1664 1665 if (vp->v_type == VSOCK) { 1666 error = EOPNOTSUPP; 1667 goto bad; 1668 } 1669 error = vn_openchk(vp, cred, flags); 1670 if (error != 0) 1671 goto bad; 1672 if (flags & O_TRUNC) { 1673 VOP_UNLOCK(vp); /* XXX */ 1674 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1675 vattr_null(&va); 1676 va.va_size = 0; 1677 error = VOP_SETATTR(vp, &va, cred); 1678 if (error) 1679 goto bad; 1680 } 1681 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1682 goto bad; 1683 if (flags & FWRITE) { 1684 mutex_enter(&vp->v_interlock); 1685 vp->v_writecount++; 1686 mutex_exit(&vp->v_interlock); 1687 } 1688 1689 /* done with modified vn_open, now finish what sys_open does. */ 1690 1691 fp->f_flag = flags & FMASK; 1692 fp->f_type = DTYPE_VNODE; 1693 fp->f_ops = &vnops; 1694 fp->f_data = vp; 1695 if (flags & (O_EXLOCK | O_SHLOCK)) { 1696 lf.l_whence = SEEK_SET; 1697 lf.l_start = 0; 1698 lf.l_len = 0; 1699 if (flags & O_EXLOCK) 1700 lf.l_type = F_WRLCK; 1701 else 1702 lf.l_type = F_RDLCK; 1703 type = F_FLOCK; 1704 if ((flags & FNONBLOCK) == 0) 1705 type |= F_WAIT; 1706 VOP_UNLOCK(vp); 1707 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1708 if (error) { 1709 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1710 fd_abort(p, fp, indx); 1711 return (error); 1712 } 1713 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1714 atomic_or_uint(&fp->f_flag, FHASLOCK); 1715 } 1716 VOP_UNLOCK(vp); 1717 *retval = indx; 1718 fd_affix(p, fp, indx); 1719 vfs_copyinfh_free(fh); 1720 return (0); 1721 1722 bad: 1723 fd_abort(p, fp, indx); 1724 if (vp != NULL) 1725 vput(vp); 1726 vfs_copyinfh_free(fh); 1727 return (error); 1728 } 1729 1730 int 1731 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1732 { 1733 /* { 1734 syscallarg(const void *) fhp; 1735 syscallarg(size_t) fh_size; 1736 syscallarg(int) flags; 1737 } */ 1738 1739 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1740 SCARG(uap, flags), retval); 1741 } 1742 1743 int 1744 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1745 { 1746 int error; 1747 fhandle_t *fh; 1748 struct vnode *vp; 1749 1750 /* 1751 * Must be super user 1752 */ 1753 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1754 0, NULL, NULL, NULL))) 1755 return (error); 1756 1757 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1758 if (error != 0) 1759 return error; 1760 1761 error = vfs_fhtovp(fh, &vp); 1762 vfs_copyinfh_free(fh); 1763 if (error != 0) 1764 return error; 1765 1766 error = vn_stat(vp, sb); 1767 vput(vp); 1768 return error; 1769 } 1770 1771 1772 /* ARGSUSED */ 1773 int 1774 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 1775 { 1776 /* { 1777 syscallarg(const void *) fhp; 1778 syscallarg(size_t) fh_size; 1779 syscallarg(struct stat *) sb; 1780 } */ 1781 struct stat sb; 1782 int error; 1783 1784 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1785 if (error) 1786 return error; 1787 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1788 } 1789 1790 int 1791 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1792 int flags) 1793 { 1794 fhandle_t *fh; 1795 struct mount *mp; 1796 struct vnode *vp; 1797 int error; 1798 1799 /* 1800 * Must be super user 1801 */ 1802 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1803 0, NULL, NULL, NULL))) 1804 return error; 1805 1806 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1807 if (error != 0) 1808 return error; 1809 1810 error = vfs_fhtovp(fh, &vp); 1811 vfs_copyinfh_free(fh); 1812 if (error != 0) 1813 return error; 1814 1815 mp = vp->v_mount; 1816 error = dostatvfs(mp, sb, l, flags, 1); 1817 vput(vp); 1818 return error; 1819 } 1820 1821 /* ARGSUSED */ 1822 int 1823 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1824 { 1825 /* { 1826 syscallarg(const void *) fhp; 1827 syscallarg(size_t) fh_size; 1828 syscallarg(struct statvfs *) buf; 1829 syscallarg(int) flags; 1830 } */ 1831 struct statvfs *sb = STATVFSBUF_GET(); 1832 int error; 1833 1834 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1835 SCARG(uap, flags)); 1836 if (error == 0) 1837 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1838 STATVFSBUF_PUT(sb); 1839 return error; 1840 } 1841 1842 /* 1843 * Create a special file. 1844 */ 1845 /* ARGSUSED */ 1846 int 1847 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 1848 register_t *retval) 1849 { 1850 /* { 1851 syscallarg(const char *) path; 1852 syscallarg(mode_t) mode; 1853 syscallarg(dev_t) dev; 1854 } */ 1855 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 1856 SCARG(uap, dev), retval, UIO_USERSPACE); 1857 } 1858 1859 int 1860 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 1861 register_t *retval, enum uio_seg seg) 1862 { 1863 struct proc *p = l->l_proc; 1864 struct vnode *vp; 1865 struct vattr vattr; 1866 int error, optype; 1867 struct nameidata nd; 1868 char *path; 1869 const char *cpath; 1870 1871 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1872 0, NULL, NULL, NULL)) != 0) 1873 return (error); 1874 1875 optype = VOP_MKNOD_DESCOFFSET; 1876 1877 VERIEXEC_PATH_GET(pathname, seg, cpath, path); 1878 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1879 1880 if ((error = namei(&nd)) != 0) 1881 goto out; 1882 vp = nd.ni_vp; 1883 if (vp != NULL) 1884 error = EEXIST; 1885 else { 1886 vattr_null(&vattr); 1887 /* We will read cwdi->cwdi_cmask unlocked. */ 1888 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1889 vattr.va_rdev = dev; 1890 1891 switch (mode & S_IFMT) { 1892 case S_IFMT: /* used by badsect to flag bad sectors */ 1893 vattr.va_type = VBAD; 1894 break; 1895 case S_IFCHR: 1896 vattr.va_type = VCHR; 1897 break; 1898 case S_IFBLK: 1899 vattr.va_type = VBLK; 1900 break; 1901 case S_IFWHT: 1902 optype = VOP_WHITEOUT_DESCOFFSET; 1903 break; 1904 case S_IFREG: 1905 #if NVERIEXEC > 0 1906 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1907 O_CREAT); 1908 #endif /* NVERIEXEC > 0 */ 1909 vattr.va_type = VREG; 1910 vattr.va_rdev = VNOVAL; 1911 optype = VOP_CREATE_DESCOFFSET; 1912 break; 1913 default: 1914 error = EINVAL; 1915 break; 1916 } 1917 } 1918 if (!error) { 1919 switch (optype) { 1920 case VOP_WHITEOUT_DESCOFFSET: 1921 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1922 if (error) 1923 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1924 vput(nd.ni_dvp); 1925 break; 1926 1927 case VOP_MKNOD_DESCOFFSET: 1928 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1929 &nd.ni_cnd, &vattr); 1930 if (error == 0) 1931 vput(nd.ni_vp); 1932 break; 1933 1934 case VOP_CREATE_DESCOFFSET: 1935 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1936 &nd.ni_cnd, &vattr); 1937 if (error == 0) 1938 vput(nd.ni_vp); 1939 break; 1940 } 1941 } else { 1942 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1943 if (nd.ni_dvp == vp) 1944 vrele(nd.ni_dvp); 1945 else 1946 vput(nd.ni_dvp); 1947 if (vp) 1948 vrele(vp); 1949 } 1950 out: 1951 VERIEXEC_PATH_PUT(path); 1952 return (error); 1953 } 1954 1955 /* 1956 * Create a named pipe. 1957 */ 1958 /* ARGSUSED */ 1959 int 1960 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1961 { 1962 /* { 1963 syscallarg(const char *) path; 1964 syscallarg(int) mode; 1965 } */ 1966 struct proc *p = l->l_proc; 1967 struct vattr vattr; 1968 int error; 1969 struct nameidata nd; 1970 1971 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1972 SCARG(uap, path)); 1973 if ((error = namei(&nd)) != 0) 1974 return (error); 1975 if (nd.ni_vp != NULL) { 1976 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1977 if (nd.ni_dvp == nd.ni_vp) 1978 vrele(nd.ni_dvp); 1979 else 1980 vput(nd.ni_dvp); 1981 vrele(nd.ni_vp); 1982 return (EEXIST); 1983 } 1984 vattr_null(&vattr); 1985 vattr.va_type = VFIFO; 1986 /* We will read cwdi->cwdi_cmask unlocked. */ 1987 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1988 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1989 if (error == 0) 1990 vput(nd.ni_vp); 1991 return (error); 1992 } 1993 1994 /* 1995 * Make a hard file link. 1996 */ 1997 /* ARGSUSED */ 1998 int 1999 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2000 { 2001 /* { 2002 syscallarg(const char *) path; 2003 syscallarg(const char *) link; 2004 } */ 2005 struct vnode *vp; 2006 struct nameidata nd; 2007 int error; 2008 2009 error = namei_simple_user(SCARG(uap, path), 2010 NSM_FOLLOW_TRYEMULROOT, &vp); 2011 if (error != 0) 2012 return (error); 2013 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2014 SCARG(uap, link)); 2015 if ((error = namei(&nd)) != 0) 2016 goto out; 2017 if (nd.ni_vp) { 2018 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2019 if (nd.ni_dvp == nd.ni_vp) 2020 vrele(nd.ni_dvp); 2021 else 2022 vput(nd.ni_dvp); 2023 vrele(nd.ni_vp); 2024 error = EEXIST; 2025 goto out; 2026 } 2027 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2028 out: 2029 vrele(vp); 2030 return (error); 2031 } 2032 2033 int 2034 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2035 { 2036 struct proc *p = curproc; 2037 struct vattr vattr; 2038 char *path; 2039 int error; 2040 struct nameidata nd; 2041 2042 path = PNBUF_GET(); 2043 if (seg == UIO_USERSPACE) { 2044 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2045 goto out; 2046 } else { 2047 KASSERT(strlen(patharg) < MAXPATHLEN); 2048 strcpy(path, patharg); 2049 } 2050 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, link); 2051 if ((error = namei(&nd)) != 0) 2052 goto out; 2053 if (nd.ni_vp) { 2054 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2055 if (nd.ni_dvp == nd.ni_vp) 2056 vrele(nd.ni_dvp); 2057 else 2058 vput(nd.ni_dvp); 2059 vrele(nd.ni_vp); 2060 error = EEXIST; 2061 goto out; 2062 } 2063 vattr_null(&vattr); 2064 vattr.va_type = VLNK; 2065 /* We will read cwdi->cwdi_cmask unlocked. */ 2066 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2067 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2068 if (error == 0) 2069 vput(nd.ni_vp); 2070 out: 2071 PNBUF_PUT(path); 2072 return (error); 2073 } 2074 2075 /* 2076 * Make a symbolic link. 2077 */ 2078 /* ARGSUSED */ 2079 int 2080 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2081 { 2082 /* { 2083 syscallarg(const char *) path; 2084 syscallarg(const char *) link; 2085 } */ 2086 2087 return do_sys_symlink(SCARG(uap, path), SCARG(uap, link), 2088 UIO_USERSPACE); 2089 } 2090 2091 /* 2092 * Delete a whiteout from the filesystem. 2093 */ 2094 /* ARGSUSED */ 2095 int 2096 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2097 { 2098 /* { 2099 syscallarg(const char *) path; 2100 } */ 2101 int error; 2102 struct nameidata nd; 2103 2104 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2105 UIO_USERSPACE, SCARG(uap, path)); 2106 error = namei(&nd); 2107 if (error) 2108 return (error); 2109 2110 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2111 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2112 if (nd.ni_dvp == nd.ni_vp) 2113 vrele(nd.ni_dvp); 2114 else 2115 vput(nd.ni_dvp); 2116 if (nd.ni_vp) 2117 vrele(nd.ni_vp); 2118 return (EEXIST); 2119 } 2120 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2121 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2122 vput(nd.ni_dvp); 2123 return (error); 2124 } 2125 2126 /* 2127 * Delete a name from the filesystem. 2128 */ 2129 /* ARGSUSED */ 2130 int 2131 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2132 { 2133 /* { 2134 syscallarg(const char *) path; 2135 } */ 2136 2137 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2138 } 2139 2140 int 2141 do_sys_unlink(const char *arg, enum uio_seg seg) 2142 { 2143 struct vnode *vp; 2144 int error; 2145 struct nameidata nd; 2146 char *path; 2147 const char *cpath; 2148 2149 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2150 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2151 2152 if ((error = namei(&nd)) != 0) 2153 goto out; 2154 vp = nd.ni_vp; 2155 2156 /* 2157 * The root of a mounted filesystem cannot be deleted. 2158 */ 2159 if (vp->v_vflag & VV_ROOT) { 2160 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2161 if (nd.ni_dvp == vp) 2162 vrele(nd.ni_dvp); 2163 else 2164 vput(nd.ni_dvp); 2165 vput(vp); 2166 error = EBUSY; 2167 goto out; 2168 } 2169 2170 #if NVERIEXEC > 0 2171 /* Handle remove requests for veriexec entries. */ 2172 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2173 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2174 if (nd.ni_dvp == vp) 2175 vrele(nd.ni_dvp); 2176 else 2177 vput(nd.ni_dvp); 2178 vput(vp); 2179 goto out; 2180 } 2181 #endif /* NVERIEXEC > 0 */ 2182 2183 #ifdef FILEASSOC 2184 (void)fileassoc_file_delete(vp); 2185 #endif /* FILEASSOC */ 2186 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2187 out: 2188 VERIEXEC_PATH_PUT(path); 2189 return (error); 2190 } 2191 2192 /* 2193 * Reposition read/write file offset. 2194 */ 2195 int 2196 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2197 { 2198 /* { 2199 syscallarg(int) fd; 2200 syscallarg(int) pad; 2201 syscallarg(off_t) offset; 2202 syscallarg(int) whence; 2203 } */ 2204 kauth_cred_t cred = l->l_cred; 2205 file_t *fp; 2206 struct vnode *vp; 2207 struct vattr vattr; 2208 off_t newoff; 2209 int error, fd; 2210 2211 fd = SCARG(uap, fd); 2212 2213 if ((fp = fd_getfile(fd)) == NULL) 2214 return (EBADF); 2215 2216 vp = fp->f_data; 2217 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2218 error = ESPIPE; 2219 goto out; 2220 } 2221 2222 switch (SCARG(uap, whence)) { 2223 case SEEK_CUR: 2224 newoff = fp->f_offset + SCARG(uap, offset); 2225 break; 2226 case SEEK_END: 2227 vn_lock(vp, LK_SHARED | LK_RETRY); 2228 error = VOP_GETATTR(vp, &vattr, cred); 2229 VOP_UNLOCK(vp); 2230 if (error) { 2231 goto out; 2232 } 2233 newoff = SCARG(uap, offset) + vattr.va_size; 2234 break; 2235 case SEEK_SET: 2236 newoff = SCARG(uap, offset); 2237 break; 2238 default: 2239 error = EINVAL; 2240 goto out; 2241 } 2242 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2243 *(off_t *)retval = fp->f_offset = newoff; 2244 } 2245 out: 2246 fd_putfile(fd); 2247 return (error); 2248 } 2249 2250 /* 2251 * Positional read system call. 2252 */ 2253 int 2254 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2255 { 2256 /* { 2257 syscallarg(int) fd; 2258 syscallarg(void *) buf; 2259 syscallarg(size_t) nbyte; 2260 syscallarg(off_t) offset; 2261 } */ 2262 file_t *fp; 2263 struct vnode *vp; 2264 off_t offset; 2265 int error, fd = SCARG(uap, fd); 2266 2267 if ((fp = fd_getfile(fd)) == NULL) 2268 return (EBADF); 2269 2270 if ((fp->f_flag & FREAD) == 0) { 2271 fd_putfile(fd); 2272 return (EBADF); 2273 } 2274 2275 vp = fp->f_data; 2276 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2277 error = ESPIPE; 2278 goto out; 2279 } 2280 2281 offset = SCARG(uap, offset); 2282 2283 /* 2284 * XXX This works because no file systems actually 2285 * XXX take any action on the seek operation. 2286 */ 2287 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2288 goto out; 2289 2290 /* dofileread() will unuse the descriptor for us */ 2291 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2292 &offset, 0, retval)); 2293 2294 out: 2295 fd_putfile(fd); 2296 return (error); 2297 } 2298 2299 /* 2300 * Positional scatter read system call. 2301 */ 2302 int 2303 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2304 { 2305 /* { 2306 syscallarg(int) fd; 2307 syscallarg(const struct iovec *) iovp; 2308 syscallarg(int) iovcnt; 2309 syscallarg(off_t) offset; 2310 } */ 2311 off_t offset = SCARG(uap, offset); 2312 2313 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2314 SCARG(uap, iovcnt), &offset, 0, retval); 2315 } 2316 2317 /* 2318 * Positional write system call. 2319 */ 2320 int 2321 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2322 { 2323 /* { 2324 syscallarg(int) fd; 2325 syscallarg(const void *) buf; 2326 syscallarg(size_t) nbyte; 2327 syscallarg(off_t) offset; 2328 } */ 2329 file_t *fp; 2330 struct vnode *vp; 2331 off_t offset; 2332 int error, fd = SCARG(uap, fd); 2333 2334 if ((fp = fd_getfile(fd)) == NULL) 2335 return (EBADF); 2336 2337 if ((fp->f_flag & FWRITE) == 0) { 2338 fd_putfile(fd); 2339 return (EBADF); 2340 } 2341 2342 vp = fp->f_data; 2343 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2344 error = ESPIPE; 2345 goto out; 2346 } 2347 2348 offset = SCARG(uap, offset); 2349 2350 /* 2351 * XXX This works because no file systems actually 2352 * XXX take any action on the seek operation. 2353 */ 2354 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2355 goto out; 2356 2357 /* dofilewrite() will unuse the descriptor for us */ 2358 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2359 &offset, 0, retval)); 2360 2361 out: 2362 fd_putfile(fd); 2363 return (error); 2364 } 2365 2366 /* 2367 * Positional gather write system call. 2368 */ 2369 int 2370 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2371 { 2372 /* { 2373 syscallarg(int) fd; 2374 syscallarg(const struct iovec *) iovp; 2375 syscallarg(int) iovcnt; 2376 syscallarg(off_t) offset; 2377 } */ 2378 off_t offset = SCARG(uap, offset); 2379 2380 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2381 SCARG(uap, iovcnt), &offset, 0, retval); 2382 } 2383 2384 /* 2385 * Check access permissions. 2386 */ 2387 int 2388 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2389 { 2390 /* { 2391 syscallarg(const char *) path; 2392 syscallarg(int) flags; 2393 } */ 2394 kauth_cred_t cred; 2395 struct vnode *vp; 2396 int error, flags; 2397 struct nameidata nd; 2398 2399 cred = kauth_cred_dup(l->l_cred); 2400 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2401 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2402 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2403 SCARG(uap, path)); 2404 /* Override default credentials */ 2405 nd.ni_cnd.cn_cred = cred; 2406 if ((error = namei(&nd)) != 0) 2407 goto out; 2408 vp = nd.ni_vp; 2409 2410 /* Flags == 0 means only check for existence. */ 2411 if (SCARG(uap, flags)) { 2412 flags = 0; 2413 if (SCARG(uap, flags) & R_OK) 2414 flags |= VREAD; 2415 if (SCARG(uap, flags) & W_OK) 2416 flags |= VWRITE; 2417 if (SCARG(uap, flags) & X_OK) 2418 flags |= VEXEC; 2419 2420 error = VOP_ACCESS(vp, flags, cred); 2421 if (!error && (flags & VWRITE)) 2422 error = vn_writechk(vp); 2423 } 2424 vput(vp); 2425 out: 2426 kauth_cred_free(cred); 2427 return (error); 2428 } 2429 2430 /* 2431 * Common code for all sys_stat functions, including compat versions. 2432 */ 2433 int 2434 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb) 2435 { 2436 int error; 2437 struct nameidata nd; 2438 2439 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2440 UIO_USERSPACE, path); 2441 error = namei(&nd); 2442 if (error != 0) 2443 return error; 2444 error = vn_stat(nd.ni_vp, sb); 2445 vput(nd.ni_vp); 2446 return error; 2447 } 2448 2449 /* 2450 * Get file status; this version follows links. 2451 */ 2452 /* ARGSUSED */ 2453 int 2454 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2455 { 2456 /* { 2457 syscallarg(const char *) path; 2458 syscallarg(struct stat *) ub; 2459 } */ 2460 struct stat sb; 2461 int error; 2462 2463 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2464 if (error) 2465 return error; 2466 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2467 } 2468 2469 /* 2470 * Get file status; this version does not follow links. 2471 */ 2472 /* ARGSUSED */ 2473 int 2474 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2475 { 2476 /* { 2477 syscallarg(const char *) path; 2478 syscallarg(struct stat *) ub; 2479 } */ 2480 struct stat sb; 2481 int error; 2482 2483 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2484 if (error) 2485 return error; 2486 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2487 } 2488 2489 /* 2490 * Get configurable pathname variables. 2491 */ 2492 /* ARGSUSED */ 2493 int 2494 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2495 { 2496 /* { 2497 syscallarg(const char *) path; 2498 syscallarg(int) name; 2499 } */ 2500 int error; 2501 struct nameidata nd; 2502 2503 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2504 SCARG(uap, path)); 2505 if ((error = namei(&nd)) != 0) 2506 return (error); 2507 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2508 vput(nd.ni_vp); 2509 return (error); 2510 } 2511 2512 /* 2513 * Return target name of a symbolic link. 2514 */ 2515 /* ARGSUSED */ 2516 int 2517 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2518 { 2519 /* { 2520 syscallarg(const char *) path; 2521 syscallarg(char *) buf; 2522 syscallarg(size_t) count; 2523 } */ 2524 struct vnode *vp; 2525 struct iovec aiov; 2526 struct uio auio; 2527 int error; 2528 struct nameidata nd; 2529 2530 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2531 SCARG(uap, path)); 2532 if ((error = namei(&nd)) != 0) 2533 return (error); 2534 vp = nd.ni_vp; 2535 if (vp->v_type != VLNK) 2536 error = EINVAL; 2537 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2538 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2539 aiov.iov_base = SCARG(uap, buf); 2540 aiov.iov_len = SCARG(uap, count); 2541 auio.uio_iov = &aiov; 2542 auio.uio_iovcnt = 1; 2543 auio.uio_offset = 0; 2544 auio.uio_rw = UIO_READ; 2545 KASSERT(l == curlwp); 2546 auio.uio_vmspace = l->l_proc->p_vmspace; 2547 auio.uio_resid = SCARG(uap, count); 2548 error = VOP_READLINK(vp, &auio, l->l_cred); 2549 } 2550 vput(vp); 2551 *retval = SCARG(uap, count) - auio.uio_resid; 2552 return (error); 2553 } 2554 2555 /* 2556 * Change flags of a file given a path name. 2557 */ 2558 /* ARGSUSED */ 2559 int 2560 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2561 { 2562 /* { 2563 syscallarg(const char *) path; 2564 syscallarg(u_long) flags; 2565 } */ 2566 struct vnode *vp; 2567 int error; 2568 2569 error = namei_simple_user(SCARG(uap, path), 2570 NSM_FOLLOW_TRYEMULROOT, &vp); 2571 if (error != 0) 2572 return (error); 2573 error = change_flags(vp, SCARG(uap, flags), l); 2574 vput(vp); 2575 return (error); 2576 } 2577 2578 /* 2579 * Change flags of a file given a file descriptor. 2580 */ 2581 /* ARGSUSED */ 2582 int 2583 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2584 { 2585 /* { 2586 syscallarg(int) fd; 2587 syscallarg(u_long) flags; 2588 } */ 2589 struct vnode *vp; 2590 file_t *fp; 2591 int error; 2592 2593 /* fd_getvnode() will use the descriptor for us */ 2594 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2595 return (error); 2596 vp = fp->f_data; 2597 error = change_flags(vp, SCARG(uap, flags), l); 2598 VOP_UNLOCK(vp); 2599 fd_putfile(SCARG(uap, fd)); 2600 return (error); 2601 } 2602 2603 /* 2604 * Change flags of a file given a path name; this version does 2605 * not follow links. 2606 */ 2607 int 2608 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2609 { 2610 /* { 2611 syscallarg(const char *) path; 2612 syscallarg(u_long) flags; 2613 } */ 2614 struct vnode *vp; 2615 int error; 2616 2617 error = namei_simple_user(SCARG(uap, path), 2618 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2619 if (error != 0) 2620 return (error); 2621 error = change_flags(vp, SCARG(uap, flags), l); 2622 vput(vp); 2623 return (error); 2624 } 2625 2626 /* 2627 * Common routine to change flags of a file. 2628 */ 2629 int 2630 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2631 { 2632 struct vattr vattr; 2633 int error; 2634 2635 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2636 /* 2637 * Non-superusers cannot change the flags on devices, even if they 2638 * own them. 2639 */ 2640 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2641 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2642 goto out; 2643 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2644 error = EINVAL; 2645 goto out; 2646 } 2647 } 2648 vattr_null(&vattr); 2649 vattr.va_flags = flags; 2650 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2651 out: 2652 return (error); 2653 } 2654 2655 /* 2656 * Change mode of a file given path name; this version follows links. 2657 */ 2658 /* ARGSUSED */ 2659 int 2660 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2661 { 2662 /* { 2663 syscallarg(const char *) path; 2664 syscallarg(int) mode; 2665 } */ 2666 int error; 2667 struct vnode *vp; 2668 2669 error = namei_simple_user(SCARG(uap, path), 2670 NSM_FOLLOW_TRYEMULROOT, &vp); 2671 if (error != 0) 2672 return (error); 2673 2674 error = change_mode(vp, SCARG(uap, mode), l); 2675 2676 vrele(vp); 2677 return (error); 2678 } 2679 2680 /* 2681 * Change mode of a file given a file descriptor. 2682 */ 2683 /* ARGSUSED */ 2684 int 2685 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2686 { 2687 /* { 2688 syscallarg(int) fd; 2689 syscallarg(int) mode; 2690 } */ 2691 file_t *fp; 2692 int error; 2693 2694 /* fd_getvnode() will use the descriptor for us */ 2695 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2696 return (error); 2697 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2698 fd_putfile(SCARG(uap, fd)); 2699 return (error); 2700 } 2701 2702 /* 2703 * Change mode of a file given path name; this version does not follow links. 2704 */ 2705 /* ARGSUSED */ 2706 int 2707 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2708 { 2709 /* { 2710 syscallarg(const char *) path; 2711 syscallarg(int) mode; 2712 } */ 2713 int error; 2714 struct vnode *vp; 2715 2716 error = namei_simple_user(SCARG(uap, path), 2717 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2718 if (error != 0) 2719 return (error); 2720 2721 error = change_mode(vp, SCARG(uap, mode), l); 2722 2723 vrele(vp); 2724 return (error); 2725 } 2726 2727 /* 2728 * Common routine to set mode given a vnode. 2729 */ 2730 static int 2731 change_mode(struct vnode *vp, int mode, struct lwp *l) 2732 { 2733 struct vattr vattr; 2734 int error; 2735 2736 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2737 vattr_null(&vattr); 2738 vattr.va_mode = mode & ALLPERMS; 2739 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2740 VOP_UNLOCK(vp); 2741 return (error); 2742 } 2743 2744 /* 2745 * Set ownership given a path name; this version follows links. 2746 */ 2747 /* ARGSUSED */ 2748 int 2749 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2750 { 2751 /* { 2752 syscallarg(const char *) path; 2753 syscallarg(uid_t) uid; 2754 syscallarg(gid_t) gid; 2755 } */ 2756 int error; 2757 struct vnode *vp; 2758 2759 error = namei_simple_user(SCARG(uap, path), 2760 NSM_FOLLOW_TRYEMULROOT, &vp); 2761 if (error != 0) 2762 return (error); 2763 2764 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2765 2766 vrele(vp); 2767 return (error); 2768 } 2769 2770 /* 2771 * Set ownership given a path name; this version follows links. 2772 * Provides POSIX semantics. 2773 */ 2774 /* ARGSUSED */ 2775 int 2776 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2777 { 2778 /* { 2779 syscallarg(const char *) path; 2780 syscallarg(uid_t) uid; 2781 syscallarg(gid_t) gid; 2782 } */ 2783 int error; 2784 struct vnode *vp; 2785 2786 error = namei_simple_user(SCARG(uap, path), 2787 NSM_FOLLOW_TRYEMULROOT, &vp); 2788 if (error != 0) 2789 return (error); 2790 2791 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2792 2793 vrele(vp); 2794 return (error); 2795 } 2796 2797 /* 2798 * Set ownership given a file descriptor. 2799 */ 2800 /* ARGSUSED */ 2801 int 2802 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2803 { 2804 /* { 2805 syscallarg(int) fd; 2806 syscallarg(uid_t) uid; 2807 syscallarg(gid_t) gid; 2808 } */ 2809 int error; 2810 file_t *fp; 2811 2812 /* fd_getvnode() will use the descriptor for us */ 2813 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2814 return (error); 2815 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2816 l, 0); 2817 fd_putfile(SCARG(uap, fd)); 2818 return (error); 2819 } 2820 2821 /* 2822 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2823 */ 2824 /* ARGSUSED */ 2825 int 2826 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2827 { 2828 /* { 2829 syscallarg(int) fd; 2830 syscallarg(uid_t) uid; 2831 syscallarg(gid_t) gid; 2832 } */ 2833 int error; 2834 file_t *fp; 2835 2836 /* fd_getvnode() will use the descriptor for us */ 2837 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2838 return (error); 2839 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2840 l, 1); 2841 fd_putfile(SCARG(uap, fd)); 2842 return (error); 2843 } 2844 2845 /* 2846 * Set ownership given a path name; this version does not follow links. 2847 */ 2848 /* ARGSUSED */ 2849 int 2850 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2851 { 2852 /* { 2853 syscallarg(const char *) path; 2854 syscallarg(uid_t) uid; 2855 syscallarg(gid_t) gid; 2856 } */ 2857 int error; 2858 struct vnode *vp; 2859 2860 error = namei_simple_user(SCARG(uap, path), 2861 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2862 if (error != 0) 2863 return (error); 2864 2865 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2866 2867 vrele(vp); 2868 return (error); 2869 } 2870 2871 /* 2872 * Set ownership given a path name; this version does not follow links. 2873 * Provides POSIX/XPG semantics. 2874 */ 2875 /* ARGSUSED */ 2876 int 2877 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2878 { 2879 /* { 2880 syscallarg(const char *) path; 2881 syscallarg(uid_t) uid; 2882 syscallarg(gid_t) gid; 2883 } */ 2884 int error; 2885 struct vnode *vp; 2886 2887 error = namei_simple_user(SCARG(uap, path), 2888 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2889 if (error != 0) 2890 return (error); 2891 2892 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2893 2894 vrele(vp); 2895 return (error); 2896 } 2897 2898 /* 2899 * Common routine to set ownership given a vnode. 2900 */ 2901 static int 2902 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2903 int posix_semantics) 2904 { 2905 struct vattr vattr; 2906 mode_t newmode; 2907 int error; 2908 2909 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2910 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2911 goto out; 2912 2913 #define CHANGED(x) ((int)(x) != -1) 2914 newmode = vattr.va_mode; 2915 if (posix_semantics) { 2916 /* 2917 * POSIX/XPG semantics: if the caller is not the super-user, 2918 * clear set-user-id and set-group-id bits. Both POSIX and 2919 * the XPG consider the behaviour for calls by the super-user 2920 * implementation-defined; we leave the set-user-id and set- 2921 * group-id settings intact in that case. 2922 */ 2923 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2924 NULL) != 0) 2925 newmode &= ~(S_ISUID | S_ISGID); 2926 } else { 2927 /* 2928 * NetBSD semantics: when changing owner and/or group, 2929 * clear the respective bit(s). 2930 */ 2931 if (CHANGED(uid)) 2932 newmode &= ~S_ISUID; 2933 if (CHANGED(gid)) 2934 newmode &= ~S_ISGID; 2935 } 2936 /* Update va_mode iff altered. */ 2937 if (vattr.va_mode == newmode) 2938 newmode = VNOVAL; 2939 2940 vattr_null(&vattr); 2941 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2942 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2943 vattr.va_mode = newmode; 2944 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2945 #undef CHANGED 2946 2947 out: 2948 VOP_UNLOCK(vp); 2949 return (error); 2950 } 2951 2952 /* 2953 * Set the access and modification times given a path name; this 2954 * version follows links. 2955 */ 2956 /* ARGSUSED */ 2957 int 2958 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 2959 register_t *retval) 2960 { 2961 /* { 2962 syscallarg(const char *) path; 2963 syscallarg(const struct timeval *) tptr; 2964 } */ 2965 2966 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2967 SCARG(uap, tptr), UIO_USERSPACE); 2968 } 2969 2970 /* 2971 * Set the access and modification times given a file descriptor. 2972 */ 2973 /* ARGSUSED */ 2974 int 2975 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 2976 register_t *retval) 2977 { 2978 /* { 2979 syscallarg(int) fd; 2980 syscallarg(const struct timeval *) tptr; 2981 } */ 2982 int error; 2983 file_t *fp; 2984 2985 /* fd_getvnode() will use the descriptor for us */ 2986 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2987 return (error); 2988 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2989 UIO_USERSPACE); 2990 fd_putfile(SCARG(uap, fd)); 2991 return (error); 2992 } 2993 2994 /* 2995 * Set the access and modification times given a path name; this 2996 * version does not follow links. 2997 */ 2998 int 2999 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3000 register_t *retval) 3001 { 3002 /* { 3003 syscallarg(const char *) path; 3004 syscallarg(const struct timeval *) tptr; 3005 } */ 3006 3007 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3008 SCARG(uap, tptr), UIO_USERSPACE); 3009 } 3010 3011 /* 3012 * Common routine to set access and modification times given a vnode. 3013 */ 3014 int 3015 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3016 const struct timeval *tptr, enum uio_seg seg) 3017 { 3018 struct vattr vattr; 3019 int error, dorele = 0; 3020 namei_simple_flags_t sflags; 3021 3022 bool vanull, setbirthtime; 3023 struct timespec ts[2]; 3024 3025 /* 3026 * I have checked all callers and they pass either FOLLOW, 3027 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3028 * is 0. More to the point, they don't pass anything else. 3029 * Let's keep it that way at least until the namei interfaces 3030 * are fully sanitized. 3031 */ 3032 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3033 sflags = (flag == FOLLOW) ? 3034 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3035 3036 if (tptr == NULL) { 3037 vanull = true; 3038 nanotime(&ts[0]); 3039 ts[1] = ts[0]; 3040 } else { 3041 struct timeval tv[2]; 3042 3043 vanull = false; 3044 if (seg != UIO_SYSSPACE) { 3045 error = copyin(tptr, tv, sizeof (tv)); 3046 if (error != 0) 3047 return error; 3048 tptr = tv; 3049 } 3050 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3051 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3052 } 3053 3054 if (vp == NULL) { 3055 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3056 error = namei_simple_user(path, sflags, &vp); 3057 if (error != 0) 3058 return error; 3059 dorele = 1; 3060 } 3061 3062 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3063 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3064 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3065 vattr_null(&vattr); 3066 vattr.va_atime = ts[0]; 3067 vattr.va_mtime = ts[1]; 3068 if (setbirthtime) 3069 vattr.va_birthtime = ts[1]; 3070 if (vanull) 3071 vattr.va_vaflags |= VA_UTIMES_NULL; 3072 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3073 VOP_UNLOCK(vp); 3074 3075 if (dorele != 0) 3076 vrele(vp); 3077 3078 return error; 3079 } 3080 3081 /* 3082 * Truncate a file given its path name. 3083 */ 3084 /* ARGSUSED */ 3085 int 3086 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3087 { 3088 /* { 3089 syscallarg(const char *) path; 3090 syscallarg(int) pad; 3091 syscallarg(off_t) length; 3092 } */ 3093 struct vnode *vp; 3094 struct vattr vattr; 3095 int error; 3096 3097 error = namei_simple_user(SCARG(uap, path), 3098 NSM_FOLLOW_TRYEMULROOT, &vp); 3099 if (error != 0) 3100 return (error); 3101 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3102 if (vp->v_type == VDIR) 3103 error = EISDIR; 3104 else if ((error = vn_writechk(vp)) == 0 && 3105 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3106 vattr_null(&vattr); 3107 vattr.va_size = SCARG(uap, length); 3108 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3109 } 3110 vput(vp); 3111 return (error); 3112 } 3113 3114 /* 3115 * Truncate a file given a file descriptor. 3116 */ 3117 /* ARGSUSED */ 3118 int 3119 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3120 { 3121 /* { 3122 syscallarg(int) fd; 3123 syscallarg(int) pad; 3124 syscallarg(off_t) length; 3125 } */ 3126 struct vattr vattr; 3127 struct vnode *vp; 3128 file_t *fp; 3129 int error; 3130 3131 /* fd_getvnode() will use the descriptor for us */ 3132 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3133 return (error); 3134 if ((fp->f_flag & FWRITE) == 0) { 3135 error = EINVAL; 3136 goto out; 3137 } 3138 vp = fp->f_data; 3139 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3140 if (vp->v_type == VDIR) 3141 error = EISDIR; 3142 else if ((error = vn_writechk(vp)) == 0) { 3143 vattr_null(&vattr); 3144 vattr.va_size = SCARG(uap, length); 3145 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3146 } 3147 VOP_UNLOCK(vp); 3148 out: 3149 fd_putfile(SCARG(uap, fd)); 3150 return (error); 3151 } 3152 3153 /* 3154 * Sync an open file. 3155 */ 3156 /* ARGSUSED */ 3157 int 3158 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3159 { 3160 /* { 3161 syscallarg(int) fd; 3162 } */ 3163 struct vnode *vp; 3164 file_t *fp; 3165 int error; 3166 3167 /* fd_getvnode() will use the descriptor for us */ 3168 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3169 return (error); 3170 vp = fp->f_data; 3171 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3172 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3173 VOP_UNLOCK(vp); 3174 fd_putfile(SCARG(uap, fd)); 3175 return (error); 3176 } 3177 3178 /* 3179 * Sync a range of file data. API modeled after that found in AIX. 3180 * 3181 * FDATASYNC indicates that we need only save enough metadata to be able 3182 * to re-read the written data. Note we duplicate AIX's requirement that 3183 * the file be open for writing. 3184 */ 3185 /* ARGSUSED */ 3186 int 3187 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3188 { 3189 /* { 3190 syscallarg(int) fd; 3191 syscallarg(int) flags; 3192 syscallarg(off_t) start; 3193 syscallarg(off_t) length; 3194 } */ 3195 struct vnode *vp; 3196 file_t *fp; 3197 int flags, nflags; 3198 off_t s, e, len; 3199 int error; 3200 3201 /* fd_getvnode() will use the descriptor for us */ 3202 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3203 return (error); 3204 3205 if ((fp->f_flag & FWRITE) == 0) { 3206 error = EBADF; 3207 goto out; 3208 } 3209 3210 flags = SCARG(uap, flags); 3211 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3212 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3213 error = EINVAL; 3214 goto out; 3215 } 3216 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3217 if (flags & FDATASYNC) 3218 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3219 else 3220 nflags = FSYNC_WAIT; 3221 if (flags & FDISKSYNC) 3222 nflags |= FSYNC_CACHE; 3223 3224 len = SCARG(uap, length); 3225 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3226 if (len) { 3227 s = SCARG(uap, start); 3228 e = s + len; 3229 if (e < s) { 3230 error = EINVAL; 3231 goto out; 3232 } 3233 } else { 3234 e = 0; 3235 s = 0; 3236 } 3237 3238 vp = fp->f_data; 3239 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3240 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3241 VOP_UNLOCK(vp); 3242 out: 3243 fd_putfile(SCARG(uap, fd)); 3244 return (error); 3245 } 3246 3247 /* 3248 * Sync the data of an open file. 3249 */ 3250 /* ARGSUSED */ 3251 int 3252 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3253 { 3254 /* { 3255 syscallarg(int) fd; 3256 } */ 3257 struct vnode *vp; 3258 file_t *fp; 3259 int error; 3260 3261 /* fd_getvnode() will use the descriptor for us */ 3262 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3263 return (error); 3264 if ((fp->f_flag & FWRITE) == 0) { 3265 fd_putfile(SCARG(uap, fd)); 3266 return (EBADF); 3267 } 3268 vp = fp->f_data; 3269 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3270 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3271 VOP_UNLOCK(vp); 3272 fd_putfile(SCARG(uap, fd)); 3273 return (error); 3274 } 3275 3276 /* 3277 * Rename files, (standard) BSD semantics frontend. 3278 */ 3279 /* ARGSUSED */ 3280 int 3281 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3282 { 3283 /* { 3284 syscallarg(const char *) from; 3285 syscallarg(const char *) to; 3286 } */ 3287 3288 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3289 } 3290 3291 /* 3292 * Rename files, POSIX semantics frontend. 3293 */ 3294 /* ARGSUSED */ 3295 int 3296 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3297 { 3298 /* { 3299 syscallarg(const char *) from; 3300 syscallarg(const char *) to; 3301 } */ 3302 3303 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3304 } 3305 3306 /* 3307 * Rename files. Source and destination must either both be directories, 3308 * or both not be directories. If target is a directory, it must be empty. 3309 * If `from' and `to' refer to the same object, the value of the `retain' 3310 * argument is used to determine whether `from' will be 3311 * 3312 * (retain == 0) deleted unless `from' and `to' refer to the same 3313 * object in the file system's name space (BSD). 3314 * (retain == 1) always retained (POSIX). 3315 */ 3316 int 3317 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3318 { 3319 struct vnode *tvp, *fvp, *tdvp; 3320 struct nameidata fromnd, tond; 3321 struct mount *fs; 3322 struct lwp *l = curlwp; 3323 struct proc *p; 3324 uint32_t saveflag; 3325 int error; 3326 3327 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT | INRENAME, 3328 seg, from); 3329 if ((error = namei(&fromnd)) != 0) 3330 return (error); 3331 if (fromnd.ni_dvp != fromnd.ni_vp) 3332 VOP_UNLOCK(fromnd.ni_dvp); 3333 fvp = fromnd.ni_vp; 3334 3335 fs = fvp->v_mount; 3336 error = VFS_RENAMELOCK_ENTER(fs); 3337 if (error) { 3338 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3339 vrele(fromnd.ni_dvp); 3340 vrele(fvp); 3341 goto out1; 3342 } 3343 3344 /* 3345 * close, partially, yet another race - ideally we should only 3346 * go as far as getting fromnd.ni_dvp before getting the per-fs 3347 * lock, and then continue to get fromnd.ni_vp, but we can't do 3348 * that with namei as it stands. 3349 * 3350 * This still won't prevent rmdir from nuking fromnd.ni_vp 3351 * under us. The real fix is to get the locks in the right 3352 * order and do the lookups in the right places, but that's a 3353 * major rototill. 3354 * 3355 * Preserve the SAVESTART in cn_flags, because who knows what 3356 * might happen if we don't. 3357 * 3358 * Note: this logic (as well as this whole function) is cloned 3359 * in nfs_serv.c. Proceed accordingly. 3360 */ 3361 vrele(fvp); 3362 if ((fromnd.ni_cnd.cn_namelen == 1 && 3363 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3364 (fromnd.ni_cnd.cn_namelen == 2 && 3365 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3366 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3367 error = EINVAL; 3368 VFS_RENAMELOCK_EXIT(fs); 3369 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3370 vrele(fromnd.ni_dvp); 3371 goto out1; 3372 } 3373 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3374 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3375 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3376 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3377 fromnd.ni_cnd.cn_flags |= saveflag; 3378 if (error) { 3379 VOP_UNLOCK(fromnd.ni_dvp); 3380 VFS_RENAMELOCK_EXIT(fs); 3381 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3382 vrele(fromnd.ni_dvp); 3383 goto out1; 3384 } 3385 VOP_UNLOCK(fromnd.ni_vp); 3386 if (fromnd.ni_dvp != fromnd.ni_vp) 3387 VOP_UNLOCK(fromnd.ni_dvp); 3388 fvp = fromnd.ni_vp; 3389 3390 NDINIT(&tond, RENAME, 3391 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3392 | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0), 3393 seg, to); 3394 if ((error = namei(&tond)) != 0) { 3395 VFS_RENAMELOCK_EXIT(fs); 3396 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3397 vrele(fromnd.ni_dvp); 3398 vrele(fvp); 3399 goto out1; 3400 } 3401 tdvp = tond.ni_dvp; 3402 tvp = tond.ni_vp; 3403 3404 if (tvp != NULL) { 3405 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3406 error = ENOTDIR; 3407 goto out; 3408 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3409 error = EISDIR; 3410 goto out; 3411 } 3412 } 3413 3414 if (fvp == tdvp) 3415 error = EINVAL; 3416 3417 /* 3418 * Source and destination refer to the same object. 3419 */ 3420 if (fvp == tvp) { 3421 if (retain) 3422 error = -1; 3423 else if (fromnd.ni_dvp == tdvp && 3424 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3425 !memcmp(fromnd.ni_cnd.cn_nameptr, 3426 tond.ni_cnd.cn_nameptr, 3427 fromnd.ni_cnd.cn_namelen)) 3428 error = -1; 3429 } 3430 3431 #if NVERIEXEC > 0 3432 if (!error) { 3433 char *f1, *f2; 3434 size_t f1_len; 3435 size_t f2_len; 3436 3437 f1_len = fromnd.ni_cnd.cn_namelen + 1; 3438 f1 = kmem_alloc(f1_len, KM_SLEEP); 3439 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len); 3440 3441 f2_len = tond.ni_cnd.cn_namelen + 1; 3442 f2 = kmem_alloc(f2_len, KM_SLEEP); 3443 strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len); 3444 3445 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3446 3447 kmem_free(f1, f1_len); 3448 kmem_free(f2, f2_len); 3449 } 3450 #endif /* NVERIEXEC > 0 */ 3451 3452 out: 3453 p = l->l_proc; 3454 if (!error) { 3455 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3456 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3457 VFS_RENAMELOCK_EXIT(fs); 3458 } else { 3459 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3460 if (tdvp == tvp) 3461 vrele(tdvp); 3462 else 3463 vput(tdvp); 3464 if (tvp) 3465 vput(tvp); 3466 VFS_RENAMELOCK_EXIT(fs); 3467 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3468 vrele(fromnd.ni_dvp); 3469 vrele(fvp); 3470 } 3471 vrele(tond.ni_startdir); 3472 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3473 out1: 3474 if (fromnd.ni_startdir) 3475 vrele(fromnd.ni_startdir); 3476 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3477 return (error == -1 ? 0 : error); 3478 } 3479 3480 /* 3481 * Make a directory file. 3482 */ 3483 /* ARGSUSED */ 3484 int 3485 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3486 { 3487 /* { 3488 syscallarg(const char *) path; 3489 syscallarg(int) mode; 3490 } */ 3491 3492 return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE); 3493 } 3494 3495 int 3496 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 3497 { 3498 struct proc *p = curlwp->l_proc; 3499 struct vnode *vp; 3500 struct vattr vattr; 3501 int error; 3502 struct nameidata nd; 3503 3504 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, 3505 seg, path); 3506 if ((error = namei(&nd)) != 0) 3507 return (error); 3508 vp = nd.ni_vp; 3509 if (vp != NULL) { 3510 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3511 if (nd.ni_dvp == vp) 3512 vrele(nd.ni_dvp); 3513 else 3514 vput(nd.ni_dvp); 3515 vrele(vp); 3516 return (EEXIST); 3517 } 3518 vattr_null(&vattr); 3519 vattr.va_type = VDIR; 3520 /* We will read cwdi->cwdi_cmask unlocked. */ 3521 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3522 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3523 if (!error) 3524 vput(nd.ni_vp); 3525 return (error); 3526 } 3527 3528 /* 3529 * Remove a directory file. 3530 */ 3531 /* ARGSUSED */ 3532 int 3533 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3534 { 3535 /* { 3536 syscallarg(const char *) path; 3537 } */ 3538 struct vnode *vp; 3539 int error; 3540 struct nameidata nd; 3541 3542 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3543 SCARG(uap, path)); 3544 if ((error = namei(&nd)) != 0) 3545 return (error); 3546 vp = nd.ni_vp; 3547 if (vp->v_type != VDIR) { 3548 error = ENOTDIR; 3549 goto out; 3550 } 3551 /* 3552 * No rmdir "." please. 3553 */ 3554 if (nd.ni_dvp == vp) { 3555 error = EINVAL; 3556 goto out; 3557 } 3558 /* 3559 * The root of a mounted filesystem cannot be deleted. 3560 */ 3561 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3562 error = EBUSY; 3563 goto out; 3564 } 3565 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3566 return (error); 3567 3568 out: 3569 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3570 if (nd.ni_dvp == vp) 3571 vrele(nd.ni_dvp); 3572 else 3573 vput(nd.ni_dvp); 3574 vput(vp); 3575 return (error); 3576 } 3577 3578 /* 3579 * Read a block of directory entries in a file system independent format. 3580 */ 3581 int 3582 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3583 { 3584 /* { 3585 syscallarg(int) fd; 3586 syscallarg(char *) buf; 3587 syscallarg(size_t) count; 3588 } */ 3589 file_t *fp; 3590 int error, done; 3591 3592 /* fd_getvnode() will use the descriptor for us */ 3593 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3594 return (error); 3595 if ((fp->f_flag & FREAD) == 0) { 3596 error = EBADF; 3597 goto out; 3598 } 3599 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3600 SCARG(uap, count), &done, l, 0, 0); 3601 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3602 *retval = done; 3603 out: 3604 fd_putfile(SCARG(uap, fd)); 3605 return (error); 3606 } 3607 3608 /* 3609 * Set the mode mask for creation of filesystem nodes. 3610 */ 3611 int 3612 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3613 { 3614 /* { 3615 syscallarg(mode_t) newmask; 3616 } */ 3617 struct proc *p = l->l_proc; 3618 struct cwdinfo *cwdi; 3619 3620 /* 3621 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3622 * important is that we serialize changes to the mask. The 3623 * rw_exit() will issue a write memory barrier on our behalf, 3624 * and force the changes out to other CPUs (as it must use an 3625 * atomic operation, draining the local CPU's store buffers). 3626 */ 3627 cwdi = p->p_cwdi; 3628 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3629 *retval = cwdi->cwdi_cmask; 3630 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3631 rw_exit(&cwdi->cwdi_lock); 3632 3633 return (0); 3634 } 3635 3636 int 3637 dorevoke(struct vnode *vp, kauth_cred_t cred) 3638 { 3639 struct vattr vattr; 3640 int error; 3641 3642 vn_lock(vp, LK_SHARED | LK_RETRY); 3643 error = VOP_GETATTR(vp, &vattr, cred); 3644 VOP_UNLOCK(vp); 3645 if (error != 0) 3646 return error; 3647 if (kauth_cred_geteuid(cred) == vattr.va_uid || 3648 (error = kauth_authorize_generic(cred, 3649 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3650 VOP_REVOKE(vp, REVOKEALL); 3651 return (error); 3652 } 3653 3654 /* 3655 * Void all references to file by ripping underlying filesystem 3656 * away from vnode. 3657 */ 3658 /* ARGSUSED */ 3659 int 3660 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3661 { 3662 /* { 3663 syscallarg(const char *) path; 3664 } */ 3665 struct vnode *vp; 3666 int error; 3667 3668 error = namei_simple_user(SCARG(uap, path), 3669 NSM_FOLLOW_TRYEMULROOT, &vp); 3670 if (error != 0) 3671 return (error); 3672 error = dorevoke(vp, l->l_cred); 3673 vrele(vp); 3674 return (error); 3675 } 3676
Indexes created Fri Oct 03 02:09:57 GMT 2025