vfs_syscalls.c revision 1.418
1 /* $NetBSD: vfs_syscalls.c,v 1.418 2011/03/06 17:08:36 bouyer Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.418 2011/03/06 17:08:36 bouyer Exp $"); 70 71 #ifdef _KERNEL_OPT 72 #include "opt_fileassoc.h" 73 #include "veriexec.h" 74 #endif 75 76 #include <sys/param.h> 77 #include <sys/systm.h> 78 #include <sys/namei.h> 79 #include <sys/filedesc.h> 80 #include <sys/kernel.h> 81 #include <sys/file.h> 82 #include <sys/stat.h> 83 #include <sys/vnode.h> 84 #include <sys/mount.h> 85 #include <sys/proc.h> 86 #include <sys/uio.h> 87 #include <sys/kmem.h> 88 #include <sys/dirent.h> 89 #include <sys/sysctl.h> 90 #include <sys/syscallargs.h> 91 #include <sys/vfs_syscalls.h> 92 #include <sys/ktrace.h> 93 #ifdef FILEASSOC 94 #include <sys/fileassoc.h> 95 #endif /* FILEASSOC */ 96 #include <sys/verified_exec.h> 97 #include <sys/kauth.h> 98 #include <sys/atomic.h> 99 #include <sys/module.h> 100 #include <sys/buf.h> 101 102 #include <miscfs/genfs/genfs.h> 103 #include <miscfs/syncfs/syncfs.h> 104 #include <miscfs/specfs/specdev.h> 105 106 #include <nfs/rpcv2.h> 107 #include <nfs/nfsproto.h> 108 #include <nfs/nfs.h> 109 #include <nfs/nfs_var.h> 110 111 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 112 113 static int change_flags(struct vnode *, u_long, struct lwp *); 114 static int change_mode(struct vnode *, int, struct lwp *l); 115 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 116 117 void checkdirs(struct vnode *); 118 119 /* 120 * Virtual File System System Calls 121 */ 122 123 /* 124 * Mount a file system. 125 */ 126 127 /* 128 * This table is used to maintain compatibility with 4.3BSD 129 * and NetBSD 0.9 mount syscalls - and possibly other systems. 130 * Note, the order is important! 131 * 132 * Do not modify this table. It should only contain filesystems 133 * supported by NetBSD 0.9 and 4.3BSD. 134 */ 135 const char * const mountcompatnames[] = { 136 NULL, /* 0 = MOUNT_NONE */ 137 MOUNT_FFS, /* 1 = MOUNT_UFS */ 138 MOUNT_NFS, /* 2 */ 139 MOUNT_MFS, /* 3 */ 140 MOUNT_MSDOS, /* 4 */ 141 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 142 MOUNT_FDESC, /* 6 */ 143 MOUNT_KERNFS, /* 7 */ 144 NULL, /* 8 = MOUNT_DEVFS */ 145 MOUNT_AFS, /* 9 */ 146 }; 147 const int nmountcompatnames = sizeof(mountcompatnames) / 148 sizeof(mountcompatnames[0]); 149 150 static int 151 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 152 void *data, size_t *data_len) 153 { 154 struct mount *mp; 155 int error = 0, saved_flags; 156 157 mp = vp->v_mount; 158 saved_flags = mp->mnt_flag; 159 160 /* We can operate only on VV_ROOT nodes. */ 161 if ((vp->v_vflag & VV_ROOT) == 0) { 162 error = EINVAL; 163 goto out; 164 } 165 166 /* 167 * We only allow the filesystem to be reloaded if it 168 * is currently mounted read-only. Additionally, we 169 * prevent read-write to read-only downgrades. 170 */ 171 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 172 (mp->mnt_flag & MNT_RDONLY) == 0 && 173 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 174 error = EOPNOTSUPP; /* Needs translation */ 175 goto out; 176 } 177 178 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 179 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 180 if (error) 181 goto out; 182 183 if (vfs_busy(mp, NULL)) { 184 error = EPERM; 185 goto out; 186 } 187 188 mutex_enter(&mp->mnt_updating); 189 190 mp->mnt_flag &= ~MNT_OP_FLAGS; 191 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 192 193 /* 194 * Set the mount level flags. 195 */ 196 if (flags & MNT_RDONLY) 197 mp->mnt_flag |= MNT_RDONLY; 198 else if (mp->mnt_flag & MNT_RDONLY) 199 mp->mnt_iflag |= IMNT_WANTRDWR; 200 mp->mnt_flag &= 201 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 202 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 203 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 204 MNT_LOG); 205 mp->mnt_flag |= flags & 206 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 207 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 208 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 209 MNT_LOG | MNT_IGNORE); 210 211 error = VFS_MOUNT(mp, path, data, data_len); 212 213 if (error && data != NULL) { 214 int error2; 215 216 /* 217 * Update failed; let's try and see if it was an 218 * export request. For compat with 3.0 and earlier. 219 */ 220 error2 = vfs_hooks_reexport(mp, path, data); 221 222 /* 223 * Only update error code if the export request was 224 * understood but some problem occurred while 225 * processing it. 226 */ 227 if (error2 != EJUSTRETURN) 228 error = error2; 229 } 230 231 if (mp->mnt_iflag & IMNT_WANTRDWR) 232 mp->mnt_flag &= ~MNT_RDONLY; 233 if (error) 234 mp->mnt_flag = saved_flags; 235 mp->mnt_flag &= ~MNT_OP_FLAGS; 236 mp->mnt_iflag &= ~IMNT_WANTRDWR; 237 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 238 if (mp->mnt_syncer == NULL) 239 error = vfs_allocate_syncvnode(mp); 240 } else { 241 if (mp->mnt_syncer != NULL) 242 vfs_deallocate_syncvnode(mp); 243 } 244 mutex_exit(&mp->mnt_updating); 245 vfs_unbusy(mp, false, NULL); 246 247 out: 248 return (error); 249 } 250 251 static int 252 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 253 { 254 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 255 int error; 256 257 /* Copy file-system type from userspace. */ 258 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 259 if (error) { 260 /* 261 * Historically, filesystem types were identified by numbers. 262 * If we get an integer for the filesystem type instead of a 263 * string, we check to see if it matches one of the historic 264 * filesystem types. 265 */ 266 u_long fsindex = (u_long)fstype; 267 if (fsindex >= nmountcompatnames || 268 mountcompatnames[fsindex] == NULL) 269 return ENODEV; 270 strlcpy(fstypename, mountcompatnames[fsindex], 271 sizeof(fstypename)); 272 } 273 274 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 275 if (strcmp(fstypename, "ufs") == 0) 276 fstypename[0] = 'f'; 277 278 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 279 return 0; 280 281 /* If we can autoload a vfs module, try again */ 282 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 283 284 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 285 return 0; 286 287 return ENODEV; 288 } 289 290 static int 291 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 292 const char *path, int flags, void *data, size_t *data_len) 293 { 294 struct mount *mp; 295 struct vnode *vp = *vpp; 296 struct vattr va; 297 struct pathbuf *pb; 298 struct nameidata nd; 299 int error; 300 301 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 302 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 303 if (error) { 304 vfs_delref(vfsops); 305 return error; 306 } 307 308 /* Can't make a non-dir a mount-point (from here anyway). */ 309 if (vp->v_type != VDIR) { 310 vfs_delref(vfsops); 311 return ENOTDIR; 312 } 313 314 /* 315 * If the user is not root, ensure that they own the directory 316 * onto which we are attempting to mount. 317 */ 318 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 319 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 320 (error = kauth_authorize_generic(l->l_cred, 321 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 322 vfs_delref(vfsops); 323 return error; 324 } 325 326 if (flags & MNT_EXPORTED) { 327 vfs_delref(vfsops); 328 return EINVAL; 329 } 330 331 if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) { 332 vfs_delref(vfsops); 333 return ENOMEM; 334 } 335 336 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 337 338 /* 339 * The underlying file system may refuse the mount for 340 * various reasons. Allow the user to force it to happen. 341 * 342 * Set the mount level flags. 343 */ 344 mp->mnt_flag = flags & 345 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 346 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 347 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 348 MNT_LOG | MNT_IGNORE | MNT_RDONLY); 349 350 mutex_enter(&mp->mnt_updating); 351 error = VFS_MOUNT(mp, path, data, data_len); 352 mp->mnt_flag &= ~MNT_OP_FLAGS; 353 354 if (error != 0) 355 goto err_unmounted; 356 357 /* 358 * Validate and prepare the mount point. 359 */ 360 error = pathbuf_copyin(path, &pb); 361 if (error != 0) { 362 goto err_mounted; 363 } 364 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 365 error = namei(&nd); 366 pathbuf_destroy(pb); 367 if (error != 0) { 368 goto err_mounted; 369 } 370 if (nd.ni_vp != vp) { 371 vput(nd.ni_vp); 372 error = EINVAL; 373 goto err_mounted; 374 } 375 if (vp->v_mountedhere != NULL) { 376 vput(nd.ni_vp); 377 error = EBUSY; 378 goto err_mounted; 379 } 380 error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0); 381 if (error != 0) { 382 vput(nd.ni_vp); 383 goto err_mounted; 384 } 385 386 /* 387 * Put the new filesystem on the mount list after root. 388 */ 389 cache_purge(vp); 390 mp->mnt_iflag &= ~IMNT_WANTRDWR; 391 392 mutex_enter(&mountlist_lock); 393 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 394 mutex_exit(&mountlist_lock); 395 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 396 error = vfs_allocate_syncvnode(mp); 397 if (error == 0) 398 vp->v_mountedhere = mp; 399 vput(nd.ni_vp); 400 if (error != 0) 401 goto err_onmountlist; 402 403 checkdirs(vp); 404 mutex_exit(&mp->mnt_updating); 405 406 /* Hold an additional reference to the mount across VFS_START(). */ 407 vfs_unbusy(mp, true, NULL); 408 (void) VFS_STATVFS(mp, &mp->mnt_stat); 409 error = VFS_START(mp, 0); 410 if (error) 411 vrele(vp); 412 /* Drop reference held for VFS_START(). */ 413 vfs_destroy(mp); 414 *vpp = NULL; 415 return error; 416 417 err_onmountlist: 418 mutex_enter(&mountlist_lock); 419 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 420 mp->mnt_iflag |= IMNT_GONE; 421 mutex_exit(&mountlist_lock); 422 423 err_mounted: 424 if (VFS_UNMOUNT(mp, MNT_FORCE) != 0) 425 panic("Unmounting fresh file system failed"); 426 427 err_unmounted: 428 vp->v_mountedhere = NULL; 429 mutex_exit(&mp->mnt_updating); 430 vfs_unbusy(mp, false, NULL); 431 vfs_destroy(mp); 432 433 return error; 434 } 435 436 static int 437 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 438 void *data, size_t *data_len) 439 { 440 struct mount *mp; 441 int error; 442 443 /* If MNT_GETARGS is specified, it should be the only flag. */ 444 if (flags & ~MNT_GETARGS) 445 return EINVAL; 446 447 mp = vp->v_mount; 448 449 /* XXX: probably some notion of "can see" here if we want isolation. */ 450 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 451 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 452 if (error) 453 return error; 454 455 if ((vp->v_vflag & VV_ROOT) == 0) 456 return EINVAL; 457 458 if (vfs_busy(mp, NULL)) 459 return EPERM; 460 461 mutex_enter(&mp->mnt_updating); 462 mp->mnt_flag &= ~MNT_OP_FLAGS; 463 mp->mnt_flag |= MNT_GETARGS; 464 error = VFS_MOUNT(mp, path, data, data_len); 465 mp->mnt_flag &= ~MNT_OP_FLAGS; 466 mutex_exit(&mp->mnt_updating); 467 468 vfs_unbusy(mp, false, NULL); 469 return (error); 470 } 471 472 int 473 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 474 { 475 /* { 476 syscallarg(const char *) type; 477 syscallarg(const char *) path; 478 syscallarg(int) flags; 479 syscallarg(void *) data; 480 syscallarg(size_t) data_len; 481 } */ 482 483 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 484 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 485 SCARG(uap, data_len), retval); 486 } 487 488 int 489 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 490 const char *path, int flags, void *data, enum uio_seg data_seg, 491 size_t data_len, register_t *retval) 492 { 493 struct vnode *vp; 494 void *data_buf = data; 495 bool vfsopsrele = false; 496 int error; 497 498 /* XXX: The calling convention of this routine is totally bizarre */ 499 if (vfsops) 500 vfsopsrele = true; 501 502 /* 503 * Get vnode to be covered 504 */ 505 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 506 if (error != 0) { 507 vp = NULL; 508 goto done; 509 } 510 511 if (vfsops == NULL) { 512 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 513 vfsops = vp->v_mount->mnt_op; 514 } else { 515 /* 'type' is userspace */ 516 error = mount_get_vfsops(type, &vfsops); 517 if (error != 0) 518 goto done; 519 vfsopsrele = true; 520 } 521 } 522 523 if (data != NULL && data_seg == UIO_USERSPACE) { 524 if (data_len == 0) { 525 /* No length supplied, use default for filesystem */ 526 data_len = vfsops->vfs_min_mount_data; 527 if (data_len > VFS_MAX_MOUNT_DATA) { 528 error = EINVAL; 529 goto done; 530 } 531 /* 532 * Hopefully a longer buffer won't make copyin() fail. 533 * For compatibility with 3.0 and earlier. 534 */ 535 if (flags & MNT_UPDATE 536 && data_len < sizeof (struct mnt_export_args30)) 537 data_len = sizeof (struct mnt_export_args30); 538 } 539 data_buf = kmem_alloc(data_len, KM_SLEEP); 540 541 /* NFS needs the buffer even for mnt_getargs .... */ 542 error = copyin(data, data_buf, data_len); 543 if (error != 0) 544 goto done; 545 } 546 547 if (flags & MNT_GETARGS) { 548 if (data_len == 0) { 549 error = EINVAL; 550 goto done; 551 } 552 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 553 if (error != 0) 554 goto done; 555 if (data_seg == UIO_USERSPACE) 556 error = copyout(data_buf, data, data_len); 557 *retval = data_len; 558 } else if (flags & MNT_UPDATE) { 559 error = mount_update(l, vp, path, flags, data_buf, &data_len); 560 } else { 561 /* Locking is handled internally in mount_domount(). */ 562 KASSERT(vfsopsrele == true); 563 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 564 &data_len); 565 vfsopsrele = false; 566 } 567 568 done: 569 if (vfsopsrele) 570 vfs_delref(vfsops); 571 if (vp != NULL) { 572 vrele(vp); 573 } 574 if (data_buf != data) 575 kmem_free(data_buf, data_len); 576 return (error); 577 } 578 579 /* 580 * Scan all active processes to see if any of them have a current 581 * or root directory onto which the new filesystem has just been 582 * mounted. If so, replace them with the new mount point. 583 */ 584 void 585 checkdirs(struct vnode *olddp) 586 { 587 struct cwdinfo *cwdi; 588 struct vnode *newdp, *rele1, *rele2; 589 struct proc *p; 590 bool retry; 591 592 if (olddp->v_usecount == 1) 593 return; 594 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 595 panic("mount: lost mount"); 596 597 do { 598 retry = false; 599 mutex_enter(proc_lock); 600 PROCLIST_FOREACH(p, &allproc) { 601 if ((cwdi = p->p_cwdi) == NULL) 602 continue; 603 /* 604 * Can't change to the old directory any more, 605 * so even if we see a stale value it's not a 606 * problem. 607 */ 608 if (cwdi->cwdi_cdir != olddp && 609 cwdi->cwdi_rdir != olddp) 610 continue; 611 retry = true; 612 rele1 = NULL; 613 rele2 = NULL; 614 atomic_inc_uint(&cwdi->cwdi_refcnt); 615 mutex_exit(proc_lock); 616 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 617 if (cwdi->cwdi_cdir == olddp) { 618 rele1 = cwdi->cwdi_cdir; 619 vref(newdp); 620 cwdi->cwdi_cdir = newdp; 621 } 622 if (cwdi->cwdi_rdir == olddp) { 623 rele2 = cwdi->cwdi_rdir; 624 vref(newdp); 625 cwdi->cwdi_rdir = newdp; 626 } 627 rw_exit(&cwdi->cwdi_lock); 628 cwdfree(cwdi); 629 if (rele1 != NULL) 630 vrele(rele1); 631 if (rele2 != NULL) 632 vrele(rele2); 633 mutex_enter(proc_lock); 634 break; 635 } 636 mutex_exit(proc_lock); 637 } while (retry); 638 639 if (rootvnode == olddp) { 640 vrele(rootvnode); 641 vref(newdp); 642 rootvnode = newdp; 643 } 644 vput(newdp); 645 } 646 647 /* 648 * Unmount a file system. 649 * 650 * Note: unmount takes a path to the vnode mounted on as argument, 651 * not special file (as before). 652 */ 653 /* ARGSUSED */ 654 int 655 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 656 { 657 /* { 658 syscallarg(const char *) path; 659 syscallarg(int) flags; 660 } */ 661 struct vnode *vp; 662 struct mount *mp; 663 int error; 664 struct pathbuf *pb; 665 struct nameidata nd; 666 667 error = pathbuf_copyin(SCARG(uap, path), &pb); 668 if (error) { 669 return error; 670 } 671 672 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 673 if ((error = namei(&nd)) != 0) { 674 pathbuf_destroy(pb); 675 return error; 676 } 677 vp = nd.ni_vp; 678 pathbuf_destroy(pb); 679 680 mp = vp->v_mount; 681 atomic_inc_uint(&mp->mnt_refcnt); 682 VOP_UNLOCK(vp); 683 684 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 685 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 686 if (error) { 687 vrele(vp); 688 vfs_destroy(mp); 689 return (error); 690 } 691 692 /* 693 * Don't allow unmounting the root file system. 694 */ 695 if (mp->mnt_flag & MNT_ROOTFS) { 696 vrele(vp); 697 vfs_destroy(mp); 698 return (EINVAL); 699 } 700 701 /* 702 * Must be the root of the filesystem 703 */ 704 if ((vp->v_vflag & VV_ROOT) == 0) { 705 vrele(vp); 706 vfs_destroy(mp); 707 return (EINVAL); 708 } 709 710 vrele(vp); 711 error = dounmount(mp, SCARG(uap, flags), l); 712 vfs_destroy(mp); 713 return error; 714 } 715 716 /* 717 * Do the actual file system unmount. File system is assumed to have 718 * been locked by the caller. 719 * 720 * => Caller hold reference to the mount, explicitly for dounmount(). 721 */ 722 int 723 dounmount(struct mount *mp, int flags, struct lwp *l) 724 { 725 struct vnode *coveredvp; 726 int error; 727 int async; 728 int used_syncer; 729 730 #if NVERIEXEC > 0 731 error = veriexec_unmountchk(mp); 732 if (error) 733 return (error); 734 #endif /* NVERIEXEC > 0 */ 735 736 /* 737 * XXX Freeze syncer. Must do this before locking the 738 * mount point. See dounmount() for details. 739 */ 740 mutex_enter(&syncer_mutex); 741 rw_enter(&mp->mnt_unmounting, RW_WRITER); 742 if ((mp->mnt_iflag & IMNT_GONE) != 0) { 743 rw_exit(&mp->mnt_unmounting); 744 mutex_exit(&syncer_mutex); 745 return ENOENT; 746 } 747 748 used_syncer = (mp->mnt_syncer != NULL); 749 750 /* 751 * XXX Syncer must be frozen when we get here. This should really 752 * be done on a per-mountpoint basis, but the syncer doesn't work 753 * like that. 754 * 755 * The caller of dounmount() must acquire syncer_mutex because 756 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 757 * order, and we must preserve that order to avoid deadlock. 758 * 759 * So, if the file system did not use the syncer, now is 760 * the time to release the syncer_mutex. 761 */ 762 if (used_syncer == 0) 763 mutex_exit(&syncer_mutex); 764 765 mp->mnt_iflag |= IMNT_UNMOUNT; 766 async = mp->mnt_flag & MNT_ASYNC; 767 mp->mnt_flag &= ~MNT_ASYNC; 768 cache_purgevfs(mp); /* remove cache entries for this file sys */ 769 if (mp->mnt_syncer != NULL) 770 vfs_deallocate_syncvnode(mp); 771 error = 0; 772 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 773 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 774 } 775 vfs_scrubvnlist(mp); 776 if (error == 0 || (flags & MNT_FORCE)) 777 error = VFS_UNMOUNT(mp, flags); 778 if (error) { 779 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 780 (void) vfs_allocate_syncvnode(mp); 781 mp->mnt_iflag &= ~IMNT_UNMOUNT; 782 mp->mnt_flag |= async; 783 rw_exit(&mp->mnt_unmounting); 784 if (used_syncer) 785 mutex_exit(&syncer_mutex); 786 return (error); 787 } 788 vfs_scrubvnlist(mp); 789 mutex_enter(&mountlist_lock); 790 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 791 coveredvp->v_mountedhere = NULL; 792 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 793 mp->mnt_iflag |= IMNT_GONE; 794 mutex_exit(&mountlist_lock); 795 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 796 panic("unmount: dangling vnode"); 797 if (used_syncer) 798 mutex_exit(&syncer_mutex); 799 vfs_hooks_unmount(mp); 800 rw_exit(&mp->mnt_unmounting); 801 vfs_destroy(mp); /* reference from mount() */ 802 if (coveredvp != NULLVP) 803 vrele(coveredvp); 804 return (0); 805 } 806 807 /* 808 * Sync each mounted filesystem. 809 */ 810 #ifdef DEBUG 811 int syncprt = 0; 812 struct ctldebug debug0 = { "syncprt", &syncprt }; 813 #endif 814 815 /* ARGSUSED */ 816 int 817 sys_sync(struct lwp *l, const void *v, register_t *retval) 818 { 819 struct mount *mp, *nmp; 820 int asyncflag; 821 822 if (l == NULL) 823 l = &lwp0; 824 825 mutex_enter(&mountlist_lock); 826 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 827 mp = nmp) { 828 if (vfs_busy(mp, &nmp)) { 829 continue; 830 } 831 mutex_enter(&mp->mnt_updating); 832 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 833 asyncflag = mp->mnt_flag & MNT_ASYNC; 834 mp->mnt_flag &= ~MNT_ASYNC; 835 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 836 if (asyncflag) 837 mp->mnt_flag |= MNT_ASYNC; 838 } 839 mutex_exit(&mp->mnt_updating); 840 vfs_unbusy(mp, false, &nmp); 841 } 842 mutex_exit(&mountlist_lock); 843 #ifdef DEBUG 844 if (syncprt) 845 vfs_bufstats(); 846 #endif /* DEBUG */ 847 return (0); 848 } 849 850 /* 851 * Change filesystem quotas. 852 */ 853 /* ARGSUSED */ 854 int 855 sys___quotactl50(struct lwp *l, const struct sys___quotactl50_args *uap, 856 register_t *retval) 857 { 858 /* { 859 syscallarg(const char *) path; 860 syscallarg(struct plistref *) pref; 861 } */ 862 struct mount *mp; 863 int error; 864 struct vnode *vp; 865 prop_dictionary_t dict; 866 struct plistref pref; 867 868 error = namei_simple_user(SCARG(uap, path), 869 NSM_FOLLOW_TRYEMULROOT, &vp); 870 if (error != 0) 871 return (error); 872 mp = vp->v_mount; 873 error = copyin(SCARG(uap, pref), &pref, sizeof(pref)); 874 if (error) 875 return error; 876 error = prop_dictionary_copyin(&pref, &dict); 877 if (error) 878 return error; 879 error = VFS_QUOTACTL(mp, dict); 880 vrele(vp); 881 if (!error) 882 error = prop_dictionary_copyout(&pref, dict); 883 if (!error) 884 error = copyout(&pref, SCARG(uap, pref), sizeof(pref)); 885 prop_object_release(dict); 886 return (error); 887 } 888 889 int 890 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 891 int root) 892 { 893 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 894 int error = 0; 895 896 /* 897 * If MNT_NOWAIT or MNT_LAZY is specified, do not 898 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 899 * overrides MNT_NOWAIT. 900 */ 901 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 902 (flags != MNT_WAIT && flags != 0)) { 903 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 904 goto done; 905 } 906 907 /* Get the filesystem stats now */ 908 memset(sp, 0, sizeof(*sp)); 909 if ((error = VFS_STATVFS(mp, sp)) != 0) { 910 return error; 911 } 912 913 if (cwdi->cwdi_rdir == NULL) 914 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 915 done: 916 if (cwdi->cwdi_rdir != NULL) { 917 size_t len; 918 char *bp; 919 char c; 920 char *path = PNBUF_GET(); 921 922 bp = path + MAXPATHLEN; 923 *--bp = '\0'; 924 rw_enter(&cwdi->cwdi_lock, RW_READER); 925 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 926 MAXPATHLEN / 2, 0, l); 927 rw_exit(&cwdi->cwdi_lock); 928 if (error) { 929 PNBUF_PUT(path); 930 return error; 931 } 932 len = strlen(bp); 933 if (len != 1) { 934 /* 935 * for mount points that are below our root, we can see 936 * them, so we fix up the pathname and return them. The 937 * rest we cannot see, so we don't allow viewing the 938 * data. 939 */ 940 if (strncmp(bp, sp->f_mntonname, len) == 0 && 941 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 942 (void)strlcpy(sp->f_mntonname, 943 c == '\0' ? "/" : &sp->f_mntonname[len], 944 sizeof(sp->f_mntonname)); 945 } else { 946 if (root) 947 (void)strlcpy(sp->f_mntonname, "/", 948 sizeof(sp->f_mntonname)); 949 else 950 error = EPERM; 951 } 952 } 953 PNBUF_PUT(path); 954 } 955 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 956 return error; 957 } 958 959 /* 960 * Get filesystem statistics by path. 961 */ 962 int 963 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 964 { 965 struct mount *mp; 966 int error; 967 struct vnode *vp; 968 969 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 970 if (error != 0) 971 return error; 972 mp = vp->v_mount; 973 error = dostatvfs(mp, sb, l, flags, 1); 974 vrele(vp); 975 return error; 976 } 977 978 /* ARGSUSED */ 979 int 980 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 981 { 982 /* { 983 syscallarg(const char *) path; 984 syscallarg(struct statvfs *) buf; 985 syscallarg(int) flags; 986 } */ 987 struct statvfs *sb; 988 int error; 989 990 sb = STATVFSBUF_GET(); 991 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 992 if (error == 0) 993 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 994 STATVFSBUF_PUT(sb); 995 return error; 996 } 997 998 /* 999 * Get filesystem statistics by fd. 1000 */ 1001 int 1002 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1003 { 1004 file_t *fp; 1005 struct mount *mp; 1006 int error; 1007 1008 /* fd_getvnode() will use the descriptor for us */ 1009 if ((error = fd_getvnode(fd, &fp)) != 0) 1010 return (error); 1011 mp = ((struct vnode *)fp->f_data)->v_mount; 1012 error = dostatvfs(mp, sb, curlwp, flags, 1); 1013 fd_putfile(fd); 1014 return error; 1015 } 1016 1017 /* ARGSUSED */ 1018 int 1019 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1020 { 1021 /* { 1022 syscallarg(int) fd; 1023 syscallarg(struct statvfs *) buf; 1024 syscallarg(int) flags; 1025 } */ 1026 struct statvfs *sb; 1027 int error; 1028 1029 sb = STATVFSBUF_GET(); 1030 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1031 if (error == 0) 1032 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1033 STATVFSBUF_PUT(sb); 1034 return error; 1035 } 1036 1037 1038 /* 1039 * Get statistics on all filesystems. 1040 */ 1041 int 1042 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1043 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1044 register_t *retval) 1045 { 1046 int root = 0; 1047 struct proc *p = l->l_proc; 1048 struct mount *mp, *nmp; 1049 struct statvfs *sb; 1050 size_t count, maxcount; 1051 int error = 0; 1052 1053 sb = STATVFSBUF_GET(); 1054 maxcount = bufsize / entry_sz; 1055 mutex_enter(&mountlist_lock); 1056 count = 0; 1057 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1058 mp = nmp) { 1059 if (vfs_busy(mp, &nmp)) { 1060 continue; 1061 } 1062 if (sfsp && count < maxcount) { 1063 error = dostatvfs(mp, sb, l, flags, 0); 1064 if (error) { 1065 vfs_unbusy(mp, false, &nmp); 1066 error = 0; 1067 continue; 1068 } 1069 error = copyfn(sb, sfsp, entry_sz); 1070 if (error) { 1071 vfs_unbusy(mp, false, NULL); 1072 goto out; 1073 } 1074 sfsp = (char *)sfsp + entry_sz; 1075 root |= strcmp(sb->f_mntonname, "/") == 0; 1076 } 1077 count++; 1078 vfs_unbusy(mp, false, &nmp); 1079 } 1080 mutex_exit(&mountlist_lock); 1081 1082 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1083 /* 1084 * fake a root entry 1085 */ 1086 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1087 sb, l, flags, 1); 1088 if (error != 0) 1089 goto out; 1090 if (sfsp) { 1091 error = copyfn(sb, sfsp, entry_sz); 1092 if (error != 0) 1093 goto out; 1094 } 1095 count++; 1096 } 1097 if (sfsp && count > maxcount) 1098 *retval = maxcount; 1099 else 1100 *retval = count; 1101 out: 1102 STATVFSBUF_PUT(sb); 1103 return error; 1104 } 1105 1106 int 1107 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1108 { 1109 /* { 1110 syscallarg(struct statvfs *) buf; 1111 syscallarg(size_t) bufsize; 1112 syscallarg(int) flags; 1113 } */ 1114 1115 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1116 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1117 } 1118 1119 /* 1120 * Change current working directory to a given file descriptor. 1121 */ 1122 /* ARGSUSED */ 1123 int 1124 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1125 { 1126 /* { 1127 syscallarg(int) fd; 1128 } */ 1129 struct proc *p = l->l_proc; 1130 struct cwdinfo *cwdi; 1131 struct vnode *vp, *tdp; 1132 struct mount *mp; 1133 file_t *fp; 1134 int error, fd; 1135 1136 /* fd_getvnode() will use the descriptor for us */ 1137 fd = SCARG(uap, fd); 1138 if ((error = fd_getvnode(fd, &fp)) != 0) 1139 return (error); 1140 vp = fp->f_data; 1141 1142 vref(vp); 1143 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1144 if (vp->v_type != VDIR) 1145 error = ENOTDIR; 1146 else 1147 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1148 if (error) { 1149 vput(vp); 1150 goto out; 1151 } 1152 while ((mp = vp->v_mountedhere) != NULL) { 1153 error = vfs_busy(mp, NULL); 1154 vput(vp); 1155 if (error != 0) 1156 goto out; 1157 error = VFS_ROOT(mp, &tdp); 1158 vfs_unbusy(mp, false, NULL); 1159 if (error) 1160 goto out; 1161 vp = tdp; 1162 } 1163 VOP_UNLOCK(vp); 1164 1165 /* 1166 * Disallow changing to a directory not under the process's 1167 * current root directory (if there is one). 1168 */ 1169 cwdi = p->p_cwdi; 1170 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1171 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1172 vrele(vp); 1173 error = EPERM; /* operation not permitted */ 1174 } else { 1175 vrele(cwdi->cwdi_cdir); 1176 cwdi->cwdi_cdir = vp; 1177 } 1178 rw_exit(&cwdi->cwdi_lock); 1179 1180 out: 1181 fd_putfile(fd); 1182 return (error); 1183 } 1184 1185 /* 1186 * Change this process's notion of the root directory to a given file 1187 * descriptor. 1188 */ 1189 int 1190 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1191 { 1192 struct proc *p = l->l_proc; 1193 struct vnode *vp; 1194 file_t *fp; 1195 int error, fd = SCARG(uap, fd); 1196 1197 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1198 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1199 return error; 1200 /* fd_getvnode() will use the descriptor for us */ 1201 if ((error = fd_getvnode(fd, &fp)) != 0) 1202 return error; 1203 vp = fp->f_data; 1204 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1205 if (vp->v_type != VDIR) 1206 error = ENOTDIR; 1207 else 1208 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1209 VOP_UNLOCK(vp); 1210 if (error) 1211 goto out; 1212 vref(vp); 1213 1214 change_root(p->p_cwdi, vp, l); 1215 1216 out: 1217 fd_putfile(fd); 1218 return (error); 1219 } 1220 1221 /* 1222 * Change current working directory (``.''). 1223 */ 1224 /* ARGSUSED */ 1225 int 1226 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1227 { 1228 /* { 1229 syscallarg(const char *) path; 1230 } */ 1231 struct proc *p = l->l_proc; 1232 struct cwdinfo *cwdi; 1233 int error; 1234 struct vnode *vp; 1235 1236 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1237 &vp, l)) != 0) 1238 return (error); 1239 cwdi = p->p_cwdi; 1240 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1241 vrele(cwdi->cwdi_cdir); 1242 cwdi->cwdi_cdir = vp; 1243 rw_exit(&cwdi->cwdi_lock); 1244 return (0); 1245 } 1246 1247 /* 1248 * Change notion of root (``/'') directory. 1249 */ 1250 /* ARGSUSED */ 1251 int 1252 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1253 { 1254 /* { 1255 syscallarg(const char *) path; 1256 } */ 1257 struct proc *p = l->l_proc; 1258 int error; 1259 struct vnode *vp; 1260 1261 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1262 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1263 return (error); 1264 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1265 &vp, l)) != 0) 1266 return (error); 1267 1268 change_root(p->p_cwdi, vp, l); 1269 1270 return (0); 1271 } 1272 1273 /* 1274 * Common routine for chroot and fchroot. 1275 * NB: callers need to properly authorize the change root operation. 1276 */ 1277 void 1278 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1279 { 1280 1281 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1282 if (cwdi->cwdi_rdir != NULL) 1283 vrele(cwdi->cwdi_rdir); 1284 cwdi->cwdi_rdir = vp; 1285 1286 /* 1287 * Prevent escaping from chroot by putting the root under 1288 * the working directory. Silently chdir to / if we aren't 1289 * already there. 1290 */ 1291 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1292 /* 1293 * XXX would be more failsafe to change directory to a 1294 * deadfs node here instead 1295 */ 1296 vrele(cwdi->cwdi_cdir); 1297 vref(vp); 1298 cwdi->cwdi_cdir = vp; 1299 } 1300 rw_exit(&cwdi->cwdi_lock); 1301 } 1302 1303 /* 1304 * Common routine for chroot and chdir. 1305 * XXX "where" should be enum uio_seg 1306 */ 1307 int 1308 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1309 { 1310 struct pathbuf *pb; 1311 struct nameidata nd; 1312 int error; 1313 1314 error = pathbuf_maybe_copyin(path, where, &pb); 1315 if (error) { 1316 return error; 1317 } 1318 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1319 if ((error = namei(&nd)) != 0) { 1320 pathbuf_destroy(pb); 1321 return error; 1322 } 1323 *vpp = nd.ni_vp; 1324 pathbuf_destroy(pb); 1325 1326 if ((*vpp)->v_type != VDIR) 1327 error = ENOTDIR; 1328 else 1329 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1330 1331 if (error) 1332 vput(*vpp); 1333 else 1334 VOP_UNLOCK(*vpp); 1335 return (error); 1336 } 1337 1338 /* 1339 * Check permissions, allocate an open file structure, 1340 * and call the device open routine if any. 1341 */ 1342 int 1343 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1344 { 1345 /* { 1346 syscallarg(const char *) path; 1347 syscallarg(int) flags; 1348 syscallarg(int) mode; 1349 } */ 1350 struct proc *p = l->l_proc; 1351 struct cwdinfo *cwdi = p->p_cwdi; 1352 file_t *fp; 1353 struct vnode *vp; 1354 int flags, cmode; 1355 int type, indx, error; 1356 struct flock lf; 1357 struct pathbuf *pb; 1358 struct nameidata nd; 1359 1360 flags = FFLAGS(SCARG(uap, flags)); 1361 if ((flags & (FREAD | FWRITE)) == 0) 1362 return (EINVAL); 1363 1364 error = pathbuf_copyin(SCARG(uap, path), &pb); 1365 if (error) { 1366 return error; 1367 } 1368 1369 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1370 pathbuf_destroy(pb); 1371 return error; 1372 } 1373 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1374 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1375 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 1376 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1377 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1378 fd_abort(p, fp, indx); 1379 if ((error == EDUPFD || error == EMOVEFD) && 1380 l->l_dupfd >= 0 && /* XXX from fdopen */ 1381 (error = 1382 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1383 *retval = indx; 1384 pathbuf_destroy(pb); 1385 return (0); 1386 } 1387 if (error == ERESTART) 1388 error = EINTR; 1389 pathbuf_destroy(pb); 1390 return (error); 1391 } 1392 1393 l->l_dupfd = 0; 1394 vp = nd.ni_vp; 1395 pathbuf_destroy(pb); 1396 1397 fp->f_flag = flags & FMASK; 1398 fp->f_type = DTYPE_VNODE; 1399 fp->f_ops = &vnops; 1400 fp->f_data = vp; 1401 if (flags & (O_EXLOCK | O_SHLOCK)) { 1402 lf.l_whence = SEEK_SET; 1403 lf.l_start = 0; 1404 lf.l_len = 0; 1405 if (flags & O_EXLOCK) 1406 lf.l_type = F_WRLCK; 1407 else 1408 lf.l_type = F_RDLCK; 1409 type = F_FLOCK; 1410 if ((flags & FNONBLOCK) == 0) 1411 type |= F_WAIT; 1412 VOP_UNLOCK(vp); 1413 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1414 if (error) { 1415 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1416 fd_abort(p, fp, indx); 1417 return (error); 1418 } 1419 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1420 atomic_or_uint(&fp->f_flag, FHASLOCK); 1421 } 1422 VOP_UNLOCK(vp); 1423 *retval = indx; 1424 fd_affix(p, fp, indx); 1425 return (0); 1426 } 1427 1428 static void 1429 vfs__fhfree(fhandle_t *fhp) 1430 { 1431 size_t fhsize; 1432 1433 if (fhp == NULL) { 1434 return; 1435 } 1436 fhsize = FHANDLE_SIZE(fhp); 1437 kmem_free(fhp, fhsize); 1438 } 1439 1440 /* 1441 * vfs_composefh: compose a filehandle. 1442 */ 1443 1444 int 1445 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1446 { 1447 struct mount *mp; 1448 struct fid *fidp; 1449 int error; 1450 size_t needfhsize; 1451 size_t fidsize; 1452 1453 mp = vp->v_mount; 1454 fidp = NULL; 1455 if (*fh_size < FHANDLE_SIZE_MIN) { 1456 fidsize = 0; 1457 } else { 1458 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1459 if (fhp != NULL) { 1460 memset(fhp, 0, *fh_size); 1461 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1462 fidp = &fhp->fh_fid; 1463 } 1464 } 1465 error = VFS_VPTOFH(vp, fidp, &fidsize); 1466 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1467 if (error == 0 && *fh_size < needfhsize) { 1468 error = E2BIG; 1469 } 1470 *fh_size = needfhsize; 1471 return error; 1472 } 1473 1474 int 1475 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1476 { 1477 struct mount *mp; 1478 fhandle_t *fhp; 1479 size_t fhsize; 1480 size_t fidsize; 1481 int error; 1482 1483 *fhpp = NULL; 1484 mp = vp->v_mount; 1485 fidsize = 0; 1486 error = VFS_VPTOFH(vp, NULL, &fidsize); 1487 KASSERT(error != 0); 1488 if (error != E2BIG) { 1489 goto out; 1490 } 1491 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1492 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1493 if (fhp == NULL) { 1494 error = ENOMEM; 1495 goto out; 1496 } 1497 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1498 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1499 if (error == 0) { 1500 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1501 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1502 *fhpp = fhp; 1503 } else { 1504 kmem_free(fhp, fhsize); 1505 } 1506 out: 1507 return error; 1508 } 1509 1510 void 1511 vfs_composefh_free(fhandle_t *fhp) 1512 { 1513 1514 vfs__fhfree(fhp); 1515 } 1516 1517 /* 1518 * vfs_fhtovp: lookup a vnode by a filehandle. 1519 */ 1520 1521 int 1522 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1523 { 1524 struct mount *mp; 1525 int error; 1526 1527 *vpp = NULL; 1528 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1529 if (mp == NULL) { 1530 error = ESTALE; 1531 goto out; 1532 } 1533 if (mp->mnt_op->vfs_fhtovp == NULL) { 1534 error = EOPNOTSUPP; 1535 goto out; 1536 } 1537 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1538 out: 1539 return error; 1540 } 1541 1542 /* 1543 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1544 * the needed size. 1545 */ 1546 1547 int 1548 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1549 { 1550 fhandle_t *fhp; 1551 int error; 1552 1553 *fhpp = NULL; 1554 if (fhsize > FHANDLE_SIZE_MAX) { 1555 return EINVAL; 1556 } 1557 if (fhsize < FHANDLE_SIZE_MIN) { 1558 return EINVAL; 1559 } 1560 again: 1561 fhp = kmem_alloc(fhsize, KM_SLEEP); 1562 if (fhp == NULL) { 1563 return ENOMEM; 1564 } 1565 error = copyin(ufhp, fhp, fhsize); 1566 if (error == 0) { 1567 /* XXX this check shouldn't be here */ 1568 if (FHANDLE_SIZE(fhp) == fhsize) { 1569 *fhpp = fhp; 1570 return 0; 1571 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1572 /* 1573 * a kludge for nfsv2 padded handles. 1574 */ 1575 size_t sz; 1576 1577 sz = FHANDLE_SIZE(fhp); 1578 kmem_free(fhp, fhsize); 1579 fhsize = sz; 1580 goto again; 1581 } else { 1582 /* 1583 * userland told us wrong size. 1584 */ 1585 error = EINVAL; 1586 } 1587 } 1588 kmem_free(fhp, fhsize); 1589 return error; 1590 } 1591 1592 void 1593 vfs_copyinfh_free(fhandle_t *fhp) 1594 { 1595 1596 vfs__fhfree(fhp); 1597 } 1598 1599 /* 1600 * Get file handle system call 1601 */ 1602 int 1603 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1604 { 1605 /* { 1606 syscallarg(char *) fname; 1607 syscallarg(fhandle_t *) fhp; 1608 syscallarg(size_t *) fh_size; 1609 } */ 1610 struct vnode *vp; 1611 fhandle_t *fh; 1612 int error; 1613 struct pathbuf *pb; 1614 struct nameidata nd; 1615 size_t sz; 1616 size_t usz; 1617 1618 /* 1619 * Must be super user 1620 */ 1621 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1622 0, NULL, NULL, NULL); 1623 if (error) 1624 return (error); 1625 1626 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1627 if (error) { 1628 return error; 1629 } 1630 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 1631 error = namei(&nd); 1632 if (error) { 1633 pathbuf_destroy(pb); 1634 return error; 1635 } 1636 vp = nd.ni_vp; 1637 pathbuf_destroy(pb); 1638 1639 error = vfs_composefh_alloc(vp, &fh); 1640 vput(vp); 1641 if (error != 0) { 1642 goto out; 1643 } 1644 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1645 if (error != 0) { 1646 goto out; 1647 } 1648 sz = FHANDLE_SIZE(fh); 1649 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1650 if (error != 0) { 1651 goto out; 1652 } 1653 if (usz >= sz) { 1654 error = copyout(fh, SCARG(uap, fhp), sz); 1655 } else { 1656 error = E2BIG; 1657 } 1658 out: 1659 vfs_composefh_free(fh); 1660 return (error); 1661 } 1662 1663 /* 1664 * Open a file given a file handle. 1665 * 1666 * Check permissions, allocate an open file structure, 1667 * and call the device open routine if any. 1668 */ 1669 1670 int 1671 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1672 register_t *retval) 1673 { 1674 file_t *fp; 1675 struct vnode *vp = NULL; 1676 kauth_cred_t cred = l->l_cred; 1677 file_t *nfp; 1678 int type, indx, error=0; 1679 struct flock lf; 1680 struct vattr va; 1681 fhandle_t *fh; 1682 int flags; 1683 proc_t *p; 1684 1685 p = curproc; 1686 1687 /* 1688 * Must be super user 1689 */ 1690 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1691 0, NULL, NULL, NULL))) 1692 return (error); 1693 1694 flags = FFLAGS(oflags); 1695 if ((flags & (FREAD | FWRITE)) == 0) 1696 return (EINVAL); 1697 if ((flags & O_CREAT)) 1698 return (EINVAL); 1699 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1700 return (error); 1701 fp = nfp; 1702 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1703 if (error != 0) { 1704 goto bad; 1705 } 1706 error = vfs_fhtovp(fh, &vp); 1707 if (error != 0) { 1708 goto bad; 1709 } 1710 1711 /* Now do an effective vn_open */ 1712 1713 if (vp->v_type == VSOCK) { 1714 error = EOPNOTSUPP; 1715 goto bad; 1716 } 1717 error = vn_openchk(vp, cred, flags); 1718 if (error != 0) 1719 goto bad; 1720 if (flags & O_TRUNC) { 1721 VOP_UNLOCK(vp); /* XXX */ 1722 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1723 vattr_null(&va); 1724 va.va_size = 0; 1725 error = VOP_SETATTR(vp, &va, cred); 1726 if (error) 1727 goto bad; 1728 } 1729 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1730 goto bad; 1731 if (flags & FWRITE) { 1732 mutex_enter(&vp->v_interlock); 1733 vp->v_writecount++; 1734 mutex_exit(&vp->v_interlock); 1735 } 1736 1737 /* done with modified vn_open, now finish what sys_open does. */ 1738 1739 fp->f_flag = flags & FMASK; 1740 fp->f_type = DTYPE_VNODE; 1741 fp->f_ops = &vnops; 1742 fp->f_data = vp; 1743 if (flags & (O_EXLOCK | O_SHLOCK)) { 1744 lf.l_whence = SEEK_SET; 1745 lf.l_start = 0; 1746 lf.l_len = 0; 1747 if (flags & O_EXLOCK) 1748 lf.l_type = F_WRLCK; 1749 else 1750 lf.l_type = F_RDLCK; 1751 type = F_FLOCK; 1752 if ((flags & FNONBLOCK) == 0) 1753 type |= F_WAIT; 1754 VOP_UNLOCK(vp); 1755 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1756 if (error) { 1757 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1758 fd_abort(p, fp, indx); 1759 return (error); 1760 } 1761 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1762 atomic_or_uint(&fp->f_flag, FHASLOCK); 1763 } 1764 VOP_UNLOCK(vp); 1765 *retval = indx; 1766 fd_affix(p, fp, indx); 1767 vfs_copyinfh_free(fh); 1768 return (0); 1769 1770 bad: 1771 fd_abort(p, fp, indx); 1772 if (vp != NULL) 1773 vput(vp); 1774 vfs_copyinfh_free(fh); 1775 return (error); 1776 } 1777 1778 int 1779 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1780 { 1781 /* { 1782 syscallarg(const void *) fhp; 1783 syscallarg(size_t) fh_size; 1784 syscallarg(int) flags; 1785 } */ 1786 1787 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1788 SCARG(uap, flags), retval); 1789 } 1790 1791 int 1792 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1793 { 1794 int error; 1795 fhandle_t *fh; 1796 struct vnode *vp; 1797 1798 /* 1799 * Must be super user 1800 */ 1801 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1802 0, NULL, NULL, NULL))) 1803 return (error); 1804 1805 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1806 if (error != 0) 1807 return error; 1808 1809 error = vfs_fhtovp(fh, &vp); 1810 vfs_copyinfh_free(fh); 1811 if (error != 0) 1812 return error; 1813 1814 error = vn_stat(vp, sb); 1815 vput(vp); 1816 return error; 1817 } 1818 1819 1820 /* ARGSUSED */ 1821 int 1822 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 1823 { 1824 /* { 1825 syscallarg(const void *) fhp; 1826 syscallarg(size_t) fh_size; 1827 syscallarg(struct stat *) sb; 1828 } */ 1829 struct stat sb; 1830 int error; 1831 1832 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1833 if (error) 1834 return error; 1835 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1836 } 1837 1838 int 1839 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1840 int flags) 1841 { 1842 fhandle_t *fh; 1843 struct mount *mp; 1844 struct vnode *vp; 1845 int error; 1846 1847 /* 1848 * Must be super user 1849 */ 1850 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1851 0, NULL, NULL, NULL))) 1852 return error; 1853 1854 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1855 if (error != 0) 1856 return error; 1857 1858 error = vfs_fhtovp(fh, &vp); 1859 vfs_copyinfh_free(fh); 1860 if (error != 0) 1861 return error; 1862 1863 mp = vp->v_mount; 1864 error = dostatvfs(mp, sb, l, flags, 1); 1865 vput(vp); 1866 return error; 1867 } 1868 1869 /* ARGSUSED */ 1870 int 1871 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1872 { 1873 /* { 1874 syscallarg(const void *) fhp; 1875 syscallarg(size_t) fh_size; 1876 syscallarg(struct statvfs *) buf; 1877 syscallarg(int) flags; 1878 } */ 1879 struct statvfs *sb = STATVFSBUF_GET(); 1880 int error; 1881 1882 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1883 SCARG(uap, flags)); 1884 if (error == 0) 1885 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1886 STATVFSBUF_PUT(sb); 1887 return error; 1888 } 1889 1890 /* 1891 * Create a special file. 1892 */ 1893 /* ARGSUSED */ 1894 int 1895 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 1896 register_t *retval) 1897 { 1898 /* { 1899 syscallarg(const char *) path; 1900 syscallarg(mode_t) mode; 1901 syscallarg(dev_t) dev; 1902 } */ 1903 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 1904 SCARG(uap, dev), retval, UIO_USERSPACE); 1905 } 1906 1907 int 1908 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 1909 register_t *retval, enum uio_seg seg) 1910 { 1911 struct proc *p = l->l_proc; 1912 struct vnode *vp; 1913 struct vattr vattr; 1914 int error, optype; 1915 struct pathbuf *pb; 1916 struct nameidata nd; 1917 const char *pathstring; 1918 1919 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1920 0, NULL, NULL, NULL)) != 0) 1921 return (error); 1922 1923 optype = VOP_MKNOD_DESCOFFSET; 1924 1925 error = pathbuf_maybe_copyin(pathname, seg, &pb); 1926 if (error) { 1927 return error; 1928 } 1929 pathstring = pathbuf_stringcopy_get(pb); 1930 if (pathstring == NULL) { 1931 pathbuf_destroy(pb); 1932 return ENOMEM; 1933 } 1934 1935 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 1936 if ((error = namei(&nd)) != 0) 1937 goto out; 1938 vp = nd.ni_vp; 1939 1940 if (vp != NULL) 1941 error = EEXIST; 1942 else { 1943 vattr_null(&vattr); 1944 /* We will read cwdi->cwdi_cmask unlocked. */ 1945 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1946 vattr.va_rdev = dev; 1947 1948 switch (mode & S_IFMT) { 1949 case S_IFMT: /* used by badsect to flag bad sectors */ 1950 vattr.va_type = VBAD; 1951 break; 1952 case S_IFCHR: 1953 vattr.va_type = VCHR; 1954 break; 1955 case S_IFBLK: 1956 vattr.va_type = VBLK; 1957 break; 1958 case S_IFWHT: 1959 optype = VOP_WHITEOUT_DESCOFFSET; 1960 break; 1961 case S_IFREG: 1962 #if NVERIEXEC > 0 1963 error = veriexec_openchk(l, nd.ni_vp, pathstring, 1964 O_CREAT); 1965 #endif /* NVERIEXEC > 0 */ 1966 vattr.va_type = VREG; 1967 vattr.va_rdev = VNOVAL; 1968 optype = VOP_CREATE_DESCOFFSET; 1969 break; 1970 default: 1971 error = EINVAL; 1972 break; 1973 } 1974 } 1975 if (!error) { 1976 switch (optype) { 1977 case VOP_WHITEOUT_DESCOFFSET: 1978 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1979 if (error) 1980 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1981 vput(nd.ni_dvp); 1982 break; 1983 1984 case VOP_MKNOD_DESCOFFSET: 1985 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1986 &nd.ni_cnd, &vattr); 1987 if (error == 0) 1988 vput(nd.ni_vp); 1989 break; 1990 1991 case VOP_CREATE_DESCOFFSET: 1992 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1993 &nd.ni_cnd, &vattr); 1994 if (error == 0) 1995 vput(nd.ni_vp); 1996 break; 1997 } 1998 } else { 1999 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2000 if (nd.ni_dvp == vp) 2001 vrele(nd.ni_dvp); 2002 else 2003 vput(nd.ni_dvp); 2004 if (vp) 2005 vrele(vp); 2006 } 2007 out: 2008 pathbuf_stringcopy_put(pb, pathstring); 2009 pathbuf_destroy(pb); 2010 return (error); 2011 } 2012 2013 /* 2014 * Create a named pipe. 2015 */ 2016 /* ARGSUSED */ 2017 int 2018 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2019 { 2020 /* { 2021 syscallarg(const char *) path; 2022 syscallarg(int) mode; 2023 } */ 2024 struct proc *p = l->l_proc; 2025 struct vattr vattr; 2026 int error; 2027 struct pathbuf *pb; 2028 struct nameidata nd; 2029 2030 error = pathbuf_copyin(SCARG(uap, path), &pb); 2031 if (error) { 2032 return error; 2033 } 2034 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2035 if ((error = namei(&nd)) != 0) { 2036 pathbuf_destroy(pb); 2037 return error; 2038 } 2039 if (nd.ni_vp != NULL) { 2040 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2041 if (nd.ni_dvp == nd.ni_vp) 2042 vrele(nd.ni_dvp); 2043 else 2044 vput(nd.ni_dvp); 2045 vrele(nd.ni_vp); 2046 pathbuf_destroy(pb); 2047 return (EEXIST); 2048 } 2049 vattr_null(&vattr); 2050 vattr.va_type = VFIFO; 2051 /* We will read cwdi->cwdi_cmask unlocked. */ 2052 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2053 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2054 if (error == 0) 2055 vput(nd.ni_vp); 2056 pathbuf_destroy(pb); 2057 return (error); 2058 } 2059 2060 /* 2061 * Make a hard file link. 2062 */ 2063 /* ARGSUSED */ 2064 int 2065 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2066 { 2067 /* { 2068 syscallarg(const char *) path; 2069 syscallarg(const char *) link; 2070 } */ 2071 struct vnode *vp; 2072 struct pathbuf *linkpb; 2073 struct nameidata nd; 2074 int error; 2075 2076 error = namei_simple_user(SCARG(uap, path), 2077 NSM_FOLLOW_TRYEMULROOT, &vp); 2078 if (error != 0) 2079 return (error); 2080 error = pathbuf_copyin(SCARG(uap, link), &linkpb); 2081 if (error) { 2082 goto out1; 2083 } 2084 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2085 if ((error = namei(&nd)) != 0) 2086 goto out2; 2087 if (nd.ni_vp) { 2088 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2089 if (nd.ni_dvp == nd.ni_vp) 2090 vrele(nd.ni_dvp); 2091 else 2092 vput(nd.ni_dvp); 2093 vrele(nd.ni_vp); 2094 error = EEXIST; 2095 goto out2; 2096 } 2097 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2098 out2: 2099 pathbuf_destroy(linkpb); 2100 out1: 2101 vrele(vp); 2102 return (error); 2103 } 2104 2105 int 2106 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2107 { 2108 struct proc *p = curproc; 2109 struct vattr vattr; 2110 char *path; 2111 int error; 2112 struct pathbuf *linkpb; 2113 struct nameidata nd; 2114 2115 path = PNBUF_GET(); 2116 if (seg == UIO_USERSPACE) { 2117 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2118 goto out1; 2119 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2120 goto out1; 2121 } else { 2122 KASSERT(strlen(patharg) < MAXPATHLEN); 2123 strcpy(path, patharg); 2124 linkpb = pathbuf_create(link); 2125 if (linkpb == NULL) { 2126 error = ENOMEM; 2127 goto out1; 2128 } 2129 } 2130 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2131 if ((error = namei(&nd)) != 0) 2132 goto out2; 2133 if (nd.ni_vp) { 2134 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2135 if (nd.ni_dvp == nd.ni_vp) 2136 vrele(nd.ni_dvp); 2137 else 2138 vput(nd.ni_dvp); 2139 vrele(nd.ni_vp); 2140 error = EEXIST; 2141 goto out2; 2142 } 2143 vattr_null(&vattr); 2144 vattr.va_type = VLNK; 2145 /* We will read cwdi->cwdi_cmask unlocked. */ 2146 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2147 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2148 if (error == 0) 2149 vput(nd.ni_vp); 2150 out2: 2151 pathbuf_destroy(linkpb); 2152 out1: 2153 PNBUF_PUT(path); 2154 return (error); 2155 } 2156 2157 /* 2158 * Make a symbolic link. 2159 */ 2160 /* ARGSUSED */ 2161 int 2162 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2163 { 2164 /* { 2165 syscallarg(const char *) path; 2166 syscallarg(const char *) link; 2167 } */ 2168 2169 return do_sys_symlink(SCARG(uap, path), SCARG(uap, link), 2170 UIO_USERSPACE); 2171 } 2172 2173 /* 2174 * Delete a whiteout from the filesystem. 2175 */ 2176 /* ARGSUSED */ 2177 int 2178 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2179 { 2180 /* { 2181 syscallarg(const char *) path; 2182 } */ 2183 int error; 2184 struct pathbuf *pb; 2185 struct nameidata nd; 2186 2187 error = pathbuf_copyin(SCARG(uap, path), &pb); 2188 if (error) { 2189 return error; 2190 } 2191 2192 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2193 error = namei(&nd); 2194 if (error) { 2195 pathbuf_destroy(pb); 2196 return (error); 2197 } 2198 2199 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2200 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2201 if (nd.ni_dvp == nd.ni_vp) 2202 vrele(nd.ni_dvp); 2203 else 2204 vput(nd.ni_dvp); 2205 if (nd.ni_vp) 2206 vrele(nd.ni_vp); 2207 pathbuf_destroy(pb); 2208 return (EEXIST); 2209 } 2210 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2211 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2212 vput(nd.ni_dvp); 2213 pathbuf_destroy(pb); 2214 return (error); 2215 } 2216 2217 /* 2218 * Delete a name from the filesystem. 2219 */ 2220 /* ARGSUSED */ 2221 int 2222 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2223 { 2224 /* { 2225 syscallarg(const char *) path; 2226 } */ 2227 2228 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2229 } 2230 2231 int 2232 do_sys_unlink(const char *arg, enum uio_seg seg) 2233 { 2234 struct vnode *vp; 2235 int error; 2236 struct pathbuf *pb; 2237 struct nameidata nd; 2238 const char *pathstring; 2239 2240 error = pathbuf_maybe_copyin(arg, seg, &pb); 2241 if (error) { 2242 return error; 2243 } 2244 pathstring = pathbuf_stringcopy_get(pb); 2245 if (pathstring == NULL) { 2246 pathbuf_destroy(pb); 2247 return ENOMEM; 2248 } 2249 2250 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2251 if ((error = namei(&nd)) != 0) 2252 goto out; 2253 vp = nd.ni_vp; 2254 2255 /* 2256 * The root of a mounted filesystem cannot be deleted. 2257 */ 2258 if (vp->v_vflag & VV_ROOT) { 2259 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2260 if (nd.ni_dvp == vp) 2261 vrele(nd.ni_dvp); 2262 else 2263 vput(nd.ni_dvp); 2264 vput(vp); 2265 error = EBUSY; 2266 goto out; 2267 } 2268 2269 #if NVERIEXEC > 0 2270 /* Handle remove requests for veriexec entries. */ 2271 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2272 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2273 if (nd.ni_dvp == vp) 2274 vrele(nd.ni_dvp); 2275 else 2276 vput(nd.ni_dvp); 2277 vput(vp); 2278 goto out; 2279 } 2280 #endif /* NVERIEXEC > 0 */ 2281 2282 #ifdef FILEASSOC 2283 (void)fileassoc_file_delete(vp); 2284 #endif /* FILEASSOC */ 2285 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2286 out: 2287 pathbuf_stringcopy_put(pb, pathstring); 2288 pathbuf_destroy(pb); 2289 return (error); 2290 } 2291 2292 /* 2293 * Reposition read/write file offset. 2294 */ 2295 int 2296 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2297 { 2298 /* { 2299 syscallarg(int) fd; 2300 syscallarg(int) pad; 2301 syscallarg(off_t) offset; 2302 syscallarg(int) whence; 2303 } */ 2304 kauth_cred_t cred = l->l_cred; 2305 file_t *fp; 2306 struct vnode *vp; 2307 struct vattr vattr; 2308 off_t newoff; 2309 int error, fd; 2310 2311 fd = SCARG(uap, fd); 2312 2313 if ((fp = fd_getfile(fd)) == NULL) 2314 return (EBADF); 2315 2316 vp = fp->f_data; 2317 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2318 error = ESPIPE; 2319 goto out; 2320 } 2321 2322 switch (SCARG(uap, whence)) { 2323 case SEEK_CUR: 2324 newoff = fp->f_offset + SCARG(uap, offset); 2325 break; 2326 case SEEK_END: 2327 error = VOP_GETATTR(vp, &vattr, cred); 2328 if (error) { 2329 goto out; 2330 } 2331 newoff = SCARG(uap, offset) + vattr.va_size; 2332 break; 2333 case SEEK_SET: 2334 newoff = SCARG(uap, offset); 2335 break; 2336 default: 2337 error = EINVAL; 2338 goto out; 2339 } 2340 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2341 *(off_t *)retval = fp->f_offset = newoff; 2342 } 2343 out: 2344 fd_putfile(fd); 2345 return (error); 2346 } 2347 2348 /* 2349 * Positional read system call. 2350 */ 2351 int 2352 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2353 { 2354 /* { 2355 syscallarg(int) fd; 2356 syscallarg(void *) buf; 2357 syscallarg(size_t) nbyte; 2358 syscallarg(off_t) offset; 2359 } */ 2360 file_t *fp; 2361 struct vnode *vp; 2362 off_t offset; 2363 int error, fd = SCARG(uap, fd); 2364 2365 if ((fp = fd_getfile(fd)) == NULL) 2366 return (EBADF); 2367 2368 if ((fp->f_flag & FREAD) == 0) { 2369 fd_putfile(fd); 2370 return (EBADF); 2371 } 2372 2373 vp = fp->f_data; 2374 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2375 error = ESPIPE; 2376 goto out; 2377 } 2378 2379 offset = SCARG(uap, offset); 2380 2381 /* 2382 * XXX This works because no file systems actually 2383 * XXX take any action on the seek operation. 2384 */ 2385 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2386 goto out; 2387 2388 /* dofileread() will unuse the descriptor for us */ 2389 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2390 &offset, 0, retval)); 2391 2392 out: 2393 fd_putfile(fd); 2394 return (error); 2395 } 2396 2397 /* 2398 * Positional scatter read system call. 2399 */ 2400 int 2401 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2402 { 2403 /* { 2404 syscallarg(int) fd; 2405 syscallarg(const struct iovec *) iovp; 2406 syscallarg(int) iovcnt; 2407 syscallarg(off_t) offset; 2408 } */ 2409 off_t offset = SCARG(uap, offset); 2410 2411 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2412 SCARG(uap, iovcnt), &offset, 0, retval); 2413 } 2414 2415 /* 2416 * Positional write system call. 2417 */ 2418 int 2419 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2420 { 2421 /* { 2422 syscallarg(int) fd; 2423 syscallarg(const void *) buf; 2424 syscallarg(size_t) nbyte; 2425 syscallarg(off_t) offset; 2426 } */ 2427 file_t *fp; 2428 struct vnode *vp; 2429 off_t offset; 2430 int error, fd = SCARG(uap, fd); 2431 2432 if ((fp = fd_getfile(fd)) == NULL) 2433 return (EBADF); 2434 2435 if ((fp->f_flag & FWRITE) == 0) { 2436 fd_putfile(fd); 2437 return (EBADF); 2438 } 2439 2440 vp = fp->f_data; 2441 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2442 error = ESPIPE; 2443 goto out; 2444 } 2445 2446 offset = SCARG(uap, offset); 2447 2448 /* 2449 * XXX This works because no file systems actually 2450 * XXX take any action on the seek operation. 2451 */ 2452 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2453 goto out; 2454 2455 /* dofilewrite() will unuse the descriptor for us */ 2456 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2457 &offset, 0, retval)); 2458 2459 out: 2460 fd_putfile(fd); 2461 return (error); 2462 } 2463 2464 /* 2465 * Positional gather write system call. 2466 */ 2467 int 2468 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2469 { 2470 /* { 2471 syscallarg(int) fd; 2472 syscallarg(const struct iovec *) iovp; 2473 syscallarg(int) iovcnt; 2474 syscallarg(off_t) offset; 2475 } */ 2476 off_t offset = SCARG(uap, offset); 2477 2478 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2479 SCARG(uap, iovcnt), &offset, 0, retval); 2480 } 2481 2482 /* 2483 * Check access permissions. 2484 */ 2485 int 2486 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2487 { 2488 /* { 2489 syscallarg(const char *) path; 2490 syscallarg(int) flags; 2491 } */ 2492 kauth_cred_t cred; 2493 struct vnode *vp; 2494 int error, flags; 2495 struct pathbuf *pb; 2496 struct nameidata nd; 2497 2498 CTASSERT(F_OK == 0); 2499 if ((SCARG(uap, flags) & ~(R_OK | W_OK | X_OK)) != 0) { 2500 /* nonsense flags */ 2501 return EINVAL; 2502 } 2503 2504 error = pathbuf_copyin(SCARG(uap, path), &pb); 2505 if (error) { 2506 return error; 2507 } 2508 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2509 2510 /* Override default credentials */ 2511 cred = kauth_cred_dup(l->l_cred); 2512 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2513 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2514 nd.ni_cnd.cn_cred = cred; 2515 2516 if ((error = namei(&nd)) != 0) { 2517 pathbuf_destroy(pb); 2518 goto out; 2519 } 2520 vp = nd.ni_vp; 2521 pathbuf_destroy(pb); 2522 2523 /* Flags == 0 means only check for existence. */ 2524 if (SCARG(uap, flags)) { 2525 flags = 0; 2526 if (SCARG(uap, flags) & R_OK) 2527 flags |= VREAD; 2528 if (SCARG(uap, flags) & W_OK) 2529 flags |= VWRITE; 2530 if (SCARG(uap, flags) & X_OK) 2531 flags |= VEXEC; 2532 2533 error = VOP_ACCESS(vp, flags, cred); 2534 if (!error && (flags & VWRITE)) 2535 error = vn_writechk(vp); 2536 } 2537 vput(vp); 2538 out: 2539 kauth_cred_free(cred); 2540 return (error); 2541 } 2542 2543 /* 2544 * Common code for all sys_stat functions, including compat versions. 2545 */ 2546 int 2547 do_sys_stat(const char *userpath, unsigned int nd_flags, struct stat *sb) 2548 { 2549 int error; 2550 struct pathbuf *pb; 2551 struct nameidata nd; 2552 2553 error = pathbuf_copyin(userpath, &pb); 2554 if (error) { 2555 return error; 2556 } 2557 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, pb); 2558 error = namei(&nd); 2559 if (error != 0) { 2560 pathbuf_destroy(pb); 2561 return error; 2562 } 2563 error = vn_stat(nd.ni_vp, sb); 2564 vput(nd.ni_vp); 2565 pathbuf_destroy(pb); 2566 return error; 2567 } 2568 2569 /* 2570 * Get file status; this version follows links. 2571 */ 2572 /* ARGSUSED */ 2573 int 2574 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2575 { 2576 /* { 2577 syscallarg(const char *) path; 2578 syscallarg(struct stat *) ub; 2579 } */ 2580 struct stat sb; 2581 int error; 2582 2583 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2584 if (error) 2585 return error; 2586 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2587 } 2588 2589 /* 2590 * Get file status; this version does not follow links. 2591 */ 2592 /* ARGSUSED */ 2593 int 2594 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2595 { 2596 /* { 2597 syscallarg(const char *) path; 2598 syscallarg(struct stat *) ub; 2599 } */ 2600 struct stat sb; 2601 int error; 2602 2603 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2604 if (error) 2605 return error; 2606 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2607 } 2608 2609 /* 2610 * Get configurable pathname variables. 2611 */ 2612 /* ARGSUSED */ 2613 int 2614 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2615 { 2616 /* { 2617 syscallarg(const char *) path; 2618 syscallarg(int) name; 2619 } */ 2620 int error; 2621 struct pathbuf *pb; 2622 struct nameidata nd; 2623 2624 error = pathbuf_copyin(SCARG(uap, path), &pb); 2625 if (error) { 2626 return error; 2627 } 2628 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2629 if ((error = namei(&nd)) != 0) { 2630 pathbuf_destroy(pb); 2631 return (error); 2632 } 2633 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2634 vput(nd.ni_vp); 2635 pathbuf_destroy(pb); 2636 return (error); 2637 } 2638 2639 /* 2640 * Return target name of a symbolic link. 2641 */ 2642 /* ARGSUSED */ 2643 int 2644 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2645 { 2646 /* { 2647 syscallarg(const char *) path; 2648 syscallarg(char *) buf; 2649 syscallarg(size_t) count; 2650 } */ 2651 struct vnode *vp; 2652 struct iovec aiov; 2653 struct uio auio; 2654 int error; 2655 struct pathbuf *pb; 2656 struct nameidata nd; 2657 2658 error = pathbuf_copyin(SCARG(uap, path), &pb); 2659 if (error) { 2660 return error; 2661 } 2662 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2663 if ((error = namei(&nd)) != 0) { 2664 pathbuf_destroy(pb); 2665 return error; 2666 } 2667 vp = nd.ni_vp; 2668 pathbuf_destroy(pb); 2669 if (vp->v_type != VLNK) 2670 error = EINVAL; 2671 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2672 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2673 aiov.iov_base = SCARG(uap, buf); 2674 aiov.iov_len = SCARG(uap, count); 2675 auio.uio_iov = &aiov; 2676 auio.uio_iovcnt = 1; 2677 auio.uio_offset = 0; 2678 auio.uio_rw = UIO_READ; 2679 KASSERT(l == curlwp); 2680 auio.uio_vmspace = l->l_proc->p_vmspace; 2681 auio.uio_resid = SCARG(uap, count); 2682 error = VOP_READLINK(vp, &auio, l->l_cred); 2683 } 2684 vput(vp); 2685 *retval = SCARG(uap, count) - auio.uio_resid; 2686 return (error); 2687 } 2688 2689 /* 2690 * Change flags of a file given a path name. 2691 */ 2692 /* ARGSUSED */ 2693 int 2694 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2695 { 2696 /* { 2697 syscallarg(const char *) path; 2698 syscallarg(u_long) flags; 2699 } */ 2700 struct vnode *vp; 2701 int error; 2702 2703 error = namei_simple_user(SCARG(uap, path), 2704 NSM_FOLLOW_TRYEMULROOT, &vp); 2705 if (error != 0) 2706 return (error); 2707 error = change_flags(vp, SCARG(uap, flags), l); 2708 vput(vp); 2709 return (error); 2710 } 2711 2712 /* 2713 * Change flags of a file given a file descriptor. 2714 */ 2715 /* ARGSUSED */ 2716 int 2717 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2718 { 2719 /* { 2720 syscallarg(int) fd; 2721 syscallarg(u_long) flags; 2722 } */ 2723 struct vnode *vp; 2724 file_t *fp; 2725 int error; 2726 2727 /* fd_getvnode() will use the descriptor for us */ 2728 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2729 return (error); 2730 vp = fp->f_data; 2731 error = change_flags(vp, SCARG(uap, flags), l); 2732 VOP_UNLOCK(vp); 2733 fd_putfile(SCARG(uap, fd)); 2734 return (error); 2735 } 2736 2737 /* 2738 * Change flags of a file given a path name; this version does 2739 * not follow links. 2740 */ 2741 int 2742 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2743 { 2744 /* { 2745 syscallarg(const char *) path; 2746 syscallarg(u_long) flags; 2747 } */ 2748 struct vnode *vp; 2749 int error; 2750 2751 error = namei_simple_user(SCARG(uap, path), 2752 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2753 if (error != 0) 2754 return (error); 2755 error = change_flags(vp, SCARG(uap, flags), l); 2756 vput(vp); 2757 return (error); 2758 } 2759 2760 /* 2761 * Common routine to change flags of a file. 2762 */ 2763 int 2764 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2765 { 2766 struct vattr vattr; 2767 int error; 2768 2769 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2770 /* 2771 * Non-superusers cannot change the flags on devices, even if they 2772 * own them. 2773 */ 2774 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2775 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2776 goto out; 2777 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2778 error = EINVAL; 2779 goto out; 2780 } 2781 } 2782 vattr_null(&vattr); 2783 vattr.va_flags = flags; 2784 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2785 out: 2786 return (error); 2787 } 2788 2789 /* 2790 * Change mode of a file given path name; this version follows links. 2791 */ 2792 /* ARGSUSED */ 2793 int 2794 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2795 { 2796 /* { 2797 syscallarg(const char *) path; 2798 syscallarg(int) mode; 2799 } */ 2800 int error; 2801 struct vnode *vp; 2802 2803 error = namei_simple_user(SCARG(uap, path), 2804 NSM_FOLLOW_TRYEMULROOT, &vp); 2805 if (error != 0) 2806 return (error); 2807 2808 error = change_mode(vp, SCARG(uap, mode), l); 2809 2810 vrele(vp); 2811 return (error); 2812 } 2813 2814 /* 2815 * Change mode of a file given a file descriptor. 2816 */ 2817 /* ARGSUSED */ 2818 int 2819 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2820 { 2821 /* { 2822 syscallarg(int) fd; 2823 syscallarg(int) mode; 2824 } */ 2825 file_t *fp; 2826 int error; 2827 2828 /* fd_getvnode() will use the descriptor for us */ 2829 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2830 return (error); 2831 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2832 fd_putfile(SCARG(uap, fd)); 2833 return (error); 2834 } 2835 2836 /* 2837 * Change mode of a file given path name; this version does not follow links. 2838 */ 2839 /* ARGSUSED */ 2840 int 2841 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2842 { 2843 /* { 2844 syscallarg(const char *) path; 2845 syscallarg(int) mode; 2846 } */ 2847 int error; 2848 struct vnode *vp; 2849 2850 error = namei_simple_user(SCARG(uap, path), 2851 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2852 if (error != 0) 2853 return (error); 2854 2855 error = change_mode(vp, SCARG(uap, mode), l); 2856 2857 vrele(vp); 2858 return (error); 2859 } 2860 2861 /* 2862 * Common routine to set mode given a vnode. 2863 */ 2864 static int 2865 change_mode(struct vnode *vp, int mode, struct lwp *l) 2866 { 2867 struct vattr vattr; 2868 int error; 2869 2870 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2871 vattr_null(&vattr); 2872 vattr.va_mode = mode & ALLPERMS; 2873 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2874 VOP_UNLOCK(vp); 2875 return (error); 2876 } 2877 2878 /* 2879 * Set ownership given a path name; this version follows links. 2880 */ 2881 /* ARGSUSED */ 2882 int 2883 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2884 { 2885 /* { 2886 syscallarg(const char *) path; 2887 syscallarg(uid_t) uid; 2888 syscallarg(gid_t) gid; 2889 } */ 2890 int error; 2891 struct vnode *vp; 2892 2893 error = namei_simple_user(SCARG(uap, path), 2894 NSM_FOLLOW_TRYEMULROOT, &vp); 2895 if (error != 0) 2896 return (error); 2897 2898 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2899 2900 vrele(vp); 2901 return (error); 2902 } 2903 2904 /* 2905 * Set ownership given a path name; this version follows links. 2906 * Provides POSIX semantics. 2907 */ 2908 /* ARGSUSED */ 2909 int 2910 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2911 { 2912 /* { 2913 syscallarg(const char *) path; 2914 syscallarg(uid_t) uid; 2915 syscallarg(gid_t) gid; 2916 } */ 2917 int error; 2918 struct vnode *vp; 2919 2920 error = namei_simple_user(SCARG(uap, path), 2921 NSM_FOLLOW_TRYEMULROOT, &vp); 2922 if (error != 0) 2923 return (error); 2924 2925 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2926 2927 vrele(vp); 2928 return (error); 2929 } 2930 2931 /* 2932 * Set ownership given a file descriptor. 2933 */ 2934 /* ARGSUSED */ 2935 int 2936 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2937 { 2938 /* { 2939 syscallarg(int) fd; 2940 syscallarg(uid_t) uid; 2941 syscallarg(gid_t) gid; 2942 } */ 2943 int error; 2944 file_t *fp; 2945 2946 /* fd_getvnode() will use the descriptor for us */ 2947 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2948 return (error); 2949 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2950 l, 0); 2951 fd_putfile(SCARG(uap, fd)); 2952 return (error); 2953 } 2954 2955 /* 2956 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2957 */ 2958 /* ARGSUSED */ 2959 int 2960 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2961 { 2962 /* { 2963 syscallarg(int) fd; 2964 syscallarg(uid_t) uid; 2965 syscallarg(gid_t) gid; 2966 } */ 2967 int error; 2968 file_t *fp; 2969 2970 /* fd_getvnode() will use the descriptor for us */ 2971 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2972 return (error); 2973 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2974 l, 1); 2975 fd_putfile(SCARG(uap, fd)); 2976 return (error); 2977 } 2978 2979 /* 2980 * Set ownership given a path name; this version does not follow links. 2981 */ 2982 /* ARGSUSED */ 2983 int 2984 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2985 { 2986 /* { 2987 syscallarg(const char *) path; 2988 syscallarg(uid_t) uid; 2989 syscallarg(gid_t) gid; 2990 } */ 2991 int error; 2992 struct vnode *vp; 2993 2994 error = namei_simple_user(SCARG(uap, path), 2995 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2996 if (error != 0) 2997 return (error); 2998 2999 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3000 3001 vrele(vp); 3002 return (error); 3003 } 3004 3005 /* 3006 * Set ownership given a path name; this version does not follow links. 3007 * Provides POSIX/XPG semantics. 3008 */ 3009 /* ARGSUSED */ 3010 int 3011 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3012 { 3013 /* { 3014 syscallarg(const char *) path; 3015 syscallarg(uid_t) uid; 3016 syscallarg(gid_t) gid; 3017 } */ 3018 int error; 3019 struct vnode *vp; 3020 3021 error = namei_simple_user(SCARG(uap, path), 3022 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3023 if (error != 0) 3024 return (error); 3025 3026 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3027 3028 vrele(vp); 3029 return (error); 3030 } 3031 3032 /* 3033 * Common routine to set ownership given a vnode. 3034 */ 3035 static int 3036 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3037 int posix_semantics) 3038 { 3039 struct vattr vattr; 3040 mode_t newmode; 3041 int error; 3042 3043 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3044 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3045 goto out; 3046 3047 #define CHANGED(x) ((int)(x) != -1) 3048 newmode = vattr.va_mode; 3049 if (posix_semantics) { 3050 /* 3051 * POSIX/XPG semantics: if the caller is not the super-user, 3052 * clear set-user-id and set-group-id bits. Both POSIX and 3053 * the XPG consider the behaviour for calls by the super-user 3054 * implementation-defined; we leave the set-user-id and set- 3055 * group-id settings intact in that case. 3056 */ 3057 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 3058 NULL) != 0) 3059 newmode &= ~(S_ISUID | S_ISGID); 3060 } else { 3061 /* 3062 * NetBSD semantics: when changing owner and/or group, 3063 * clear the respective bit(s). 3064 */ 3065 if (CHANGED(uid)) 3066 newmode &= ~S_ISUID; 3067 if (CHANGED(gid)) 3068 newmode &= ~S_ISGID; 3069 } 3070 /* Update va_mode iff altered. */ 3071 if (vattr.va_mode == newmode) 3072 newmode = VNOVAL; 3073 3074 vattr_null(&vattr); 3075 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3076 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3077 vattr.va_mode = newmode; 3078 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3079 #undef CHANGED 3080 3081 out: 3082 VOP_UNLOCK(vp); 3083 return (error); 3084 } 3085 3086 /* 3087 * Set the access and modification times given a path name; this 3088 * version follows links. 3089 */ 3090 /* ARGSUSED */ 3091 int 3092 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3093 register_t *retval) 3094 { 3095 /* { 3096 syscallarg(const char *) path; 3097 syscallarg(const struct timeval *) tptr; 3098 } */ 3099 3100 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3101 SCARG(uap, tptr), UIO_USERSPACE); 3102 } 3103 3104 /* 3105 * Set the access and modification times given a file descriptor. 3106 */ 3107 /* ARGSUSED */ 3108 int 3109 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3110 register_t *retval) 3111 { 3112 /* { 3113 syscallarg(int) fd; 3114 syscallarg(const struct timeval *) tptr; 3115 } */ 3116 int error; 3117 file_t *fp; 3118 3119 /* fd_getvnode() will use the descriptor for us */ 3120 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3121 return (error); 3122 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 3123 UIO_USERSPACE); 3124 fd_putfile(SCARG(uap, fd)); 3125 return (error); 3126 } 3127 3128 /* 3129 * Set the access and modification times given a path name; this 3130 * version does not follow links. 3131 */ 3132 int 3133 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3134 register_t *retval) 3135 { 3136 /* { 3137 syscallarg(const char *) path; 3138 syscallarg(const struct timeval *) tptr; 3139 } */ 3140 3141 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3142 SCARG(uap, tptr), UIO_USERSPACE); 3143 } 3144 3145 /* 3146 * Common routine to set access and modification times given a vnode. 3147 */ 3148 int 3149 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3150 const struct timeval *tptr, enum uio_seg seg) 3151 { 3152 struct vattr vattr; 3153 int error, dorele = 0; 3154 namei_simple_flags_t sflags; 3155 3156 bool vanull, setbirthtime; 3157 struct timespec ts[2]; 3158 3159 /* 3160 * I have checked all callers and they pass either FOLLOW, 3161 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3162 * is 0. More to the point, they don't pass anything else. 3163 * Let's keep it that way at least until the namei interfaces 3164 * are fully sanitized. 3165 */ 3166 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3167 sflags = (flag == FOLLOW) ? 3168 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3169 3170 if (tptr == NULL) { 3171 vanull = true; 3172 nanotime(&ts[0]); 3173 ts[1] = ts[0]; 3174 } else { 3175 struct timeval tv[2]; 3176 3177 vanull = false; 3178 if (seg != UIO_SYSSPACE) { 3179 error = copyin(tptr, tv, sizeof (tv)); 3180 if (error != 0) 3181 return error; 3182 tptr = tv; 3183 } 3184 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3185 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3186 } 3187 3188 if (vp == NULL) { 3189 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3190 error = namei_simple_user(path, sflags, &vp); 3191 if (error != 0) 3192 return error; 3193 dorele = 1; 3194 } 3195 3196 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3197 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3198 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3199 vattr_null(&vattr); 3200 vattr.va_atime = ts[0]; 3201 vattr.va_mtime = ts[1]; 3202 if (setbirthtime) 3203 vattr.va_birthtime = ts[1]; 3204 if (vanull) 3205 vattr.va_vaflags |= VA_UTIMES_NULL; 3206 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3207 VOP_UNLOCK(vp); 3208 3209 if (dorele != 0) 3210 vrele(vp); 3211 3212 return error; 3213 } 3214 3215 /* 3216 * Truncate a file given its path name. 3217 */ 3218 /* ARGSUSED */ 3219 int 3220 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3221 { 3222 /* { 3223 syscallarg(const char *) path; 3224 syscallarg(int) pad; 3225 syscallarg(off_t) length; 3226 } */ 3227 struct vnode *vp; 3228 struct vattr vattr; 3229 int error; 3230 3231 error = namei_simple_user(SCARG(uap, path), 3232 NSM_FOLLOW_TRYEMULROOT, &vp); 3233 if (error != 0) 3234 return (error); 3235 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3236 if (vp->v_type == VDIR) 3237 error = EISDIR; 3238 else if ((error = vn_writechk(vp)) == 0 && 3239 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3240 vattr_null(&vattr); 3241 vattr.va_size = SCARG(uap, length); 3242 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3243 } 3244 vput(vp); 3245 return (error); 3246 } 3247 3248 /* 3249 * Truncate a file given a file descriptor. 3250 */ 3251 /* ARGSUSED */ 3252 int 3253 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3254 { 3255 /* { 3256 syscallarg(int) fd; 3257 syscallarg(int) pad; 3258 syscallarg(off_t) length; 3259 } */ 3260 struct vattr vattr; 3261 struct vnode *vp; 3262 file_t *fp; 3263 int error; 3264 3265 /* fd_getvnode() will use the descriptor for us */ 3266 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3267 return (error); 3268 if ((fp->f_flag & FWRITE) == 0) { 3269 error = EINVAL; 3270 goto out; 3271 } 3272 vp = fp->f_data; 3273 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3274 if (vp->v_type == VDIR) 3275 error = EISDIR; 3276 else if ((error = vn_writechk(vp)) == 0) { 3277 vattr_null(&vattr); 3278 vattr.va_size = SCARG(uap, length); 3279 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3280 } 3281 VOP_UNLOCK(vp); 3282 out: 3283 fd_putfile(SCARG(uap, fd)); 3284 return (error); 3285 } 3286 3287 /* 3288 * Sync an open file. 3289 */ 3290 /* ARGSUSED */ 3291 int 3292 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3293 { 3294 /* { 3295 syscallarg(int) fd; 3296 } */ 3297 struct vnode *vp; 3298 file_t *fp; 3299 int error; 3300 3301 /* fd_getvnode() will use the descriptor for us */ 3302 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3303 return (error); 3304 vp = fp->f_data; 3305 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3306 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3307 VOP_UNLOCK(vp); 3308 fd_putfile(SCARG(uap, fd)); 3309 return (error); 3310 } 3311 3312 /* 3313 * Sync a range of file data. API modeled after that found in AIX. 3314 * 3315 * FDATASYNC indicates that we need only save enough metadata to be able 3316 * to re-read the written data. Note we duplicate AIX's requirement that 3317 * the file be open for writing. 3318 */ 3319 /* ARGSUSED */ 3320 int 3321 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3322 { 3323 /* { 3324 syscallarg(int) fd; 3325 syscallarg(int) flags; 3326 syscallarg(off_t) start; 3327 syscallarg(off_t) length; 3328 } */ 3329 struct vnode *vp; 3330 file_t *fp; 3331 int flags, nflags; 3332 off_t s, e, len; 3333 int error; 3334 3335 /* fd_getvnode() will use the descriptor for us */ 3336 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3337 return (error); 3338 3339 if ((fp->f_flag & FWRITE) == 0) { 3340 error = EBADF; 3341 goto out; 3342 } 3343 3344 flags = SCARG(uap, flags); 3345 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3346 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3347 error = EINVAL; 3348 goto out; 3349 } 3350 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3351 if (flags & FDATASYNC) 3352 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3353 else 3354 nflags = FSYNC_WAIT; 3355 if (flags & FDISKSYNC) 3356 nflags |= FSYNC_CACHE; 3357 3358 len = SCARG(uap, length); 3359 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3360 if (len) { 3361 s = SCARG(uap, start); 3362 e = s + len; 3363 if (e < s) { 3364 error = EINVAL; 3365 goto out; 3366 } 3367 } else { 3368 e = 0; 3369 s = 0; 3370 } 3371 3372 vp = fp->f_data; 3373 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3374 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3375 VOP_UNLOCK(vp); 3376 out: 3377 fd_putfile(SCARG(uap, fd)); 3378 return (error); 3379 } 3380 3381 /* 3382 * Sync the data of an open file. 3383 */ 3384 /* ARGSUSED */ 3385 int 3386 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3387 { 3388 /* { 3389 syscallarg(int) fd; 3390 } */ 3391 struct vnode *vp; 3392 file_t *fp; 3393 int error; 3394 3395 /* fd_getvnode() will use the descriptor for us */ 3396 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3397 return (error); 3398 if ((fp->f_flag & FWRITE) == 0) { 3399 fd_putfile(SCARG(uap, fd)); 3400 return (EBADF); 3401 } 3402 vp = fp->f_data; 3403 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3404 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3405 VOP_UNLOCK(vp); 3406 fd_putfile(SCARG(uap, fd)); 3407 return (error); 3408 } 3409 3410 /* 3411 * Rename files, (standard) BSD semantics frontend. 3412 */ 3413 /* ARGSUSED */ 3414 int 3415 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3416 { 3417 /* { 3418 syscallarg(const char *) from; 3419 syscallarg(const char *) to; 3420 } */ 3421 3422 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3423 } 3424 3425 /* 3426 * Rename files, POSIX semantics frontend. 3427 */ 3428 /* ARGSUSED */ 3429 int 3430 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3431 { 3432 /* { 3433 syscallarg(const char *) from; 3434 syscallarg(const char *) to; 3435 } */ 3436 3437 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3438 } 3439 3440 /* 3441 * Rename files. Source and destination must either both be directories, 3442 * or both not be directories. If target is a directory, it must be empty. 3443 * If `from' and `to' refer to the same object, the value of the `retain' 3444 * argument is used to determine whether `from' will be 3445 * 3446 * (retain == 0) deleted unless `from' and `to' refer to the same 3447 * object in the file system's name space (BSD). 3448 * (retain == 1) always retained (POSIX). 3449 */ 3450 int 3451 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3452 { 3453 struct vnode *tvp, *fvp, *tdvp; 3454 struct pathbuf *frompb, *topb; 3455 struct nameidata fromnd, tond; 3456 struct mount *fs; 3457 struct lwp *l = curlwp; 3458 struct proc *p; 3459 int error; 3460 3461 error = pathbuf_maybe_copyin(from, seg, &frompb); 3462 if (error) { 3463 return error; 3464 } 3465 error = pathbuf_maybe_copyin(to, seg, &topb); 3466 if (error) { 3467 pathbuf_destroy(frompb); 3468 return error; 3469 } 3470 3471 NDINIT(&fromnd, DELETE, LOCKPARENT | TRYEMULROOT | INRENAME, 3472 frompb); 3473 if ((error = namei(&fromnd)) != 0) { 3474 pathbuf_destroy(frompb); 3475 pathbuf_destroy(topb); 3476 return (error); 3477 } 3478 if (fromnd.ni_dvp != fromnd.ni_vp) 3479 VOP_UNLOCK(fromnd.ni_dvp); 3480 fvp = fromnd.ni_vp; 3481 3482 fs = fvp->v_mount; 3483 error = VFS_RENAMELOCK_ENTER(fs); 3484 if (error) { 3485 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3486 vrele(fromnd.ni_dvp); 3487 vrele(fvp); 3488 goto out1; 3489 } 3490 3491 /* 3492 * close, partially, yet another race - ideally we should only 3493 * go as far as getting fromnd.ni_dvp before getting the per-fs 3494 * lock, and then continue to get fromnd.ni_vp, but we can't do 3495 * that with namei as it stands. 3496 * 3497 * This still won't prevent rmdir from nuking fromnd.ni_vp 3498 * under us. The real fix is to get the locks in the right 3499 * order and do the lookups in the right places, but that's a 3500 * major rototill. 3501 * 3502 * Note: this logic (as well as this whole function) is cloned 3503 * in nfs_serv.c. Proceed accordingly. 3504 */ 3505 vrele(fvp); 3506 if ((fromnd.ni_cnd.cn_namelen == 1 && 3507 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3508 (fromnd.ni_cnd.cn_namelen == 2 && 3509 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3510 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3511 error = EINVAL; 3512 VFS_RENAMELOCK_EXIT(fs); 3513 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3514 vrele(fromnd.ni_dvp); 3515 goto out1; 3516 } 3517 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3518 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd, 0); 3519 if (error) { 3520 VOP_UNLOCK(fromnd.ni_dvp); 3521 VFS_RENAMELOCK_EXIT(fs); 3522 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3523 vrele(fromnd.ni_dvp); 3524 goto out1; 3525 } 3526 VOP_UNLOCK(fromnd.ni_vp); 3527 if (fromnd.ni_dvp != fromnd.ni_vp) 3528 VOP_UNLOCK(fromnd.ni_dvp); 3529 fvp = fromnd.ni_vp; 3530 3531 NDINIT(&tond, RENAME, 3532 LOCKPARENT | LOCKLEAF | NOCACHE | TRYEMULROOT 3533 | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0), 3534 topb); 3535 if ((error = namei(&tond)) != 0) { 3536 VFS_RENAMELOCK_EXIT(fs); 3537 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3538 vrele(fromnd.ni_dvp); 3539 vrele(fvp); 3540 goto out1; 3541 } 3542 tdvp = tond.ni_dvp; 3543 tvp = tond.ni_vp; 3544 3545 if (tvp != NULL) { 3546 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3547 error = ENOTDIR; 3548 goto out; 3549 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3550 error = EISDIR; 3551 goto out; 3552 } 3553 } 3554 3555 if (fvp == tdvp) 3556 error = EINVAL; 3557 3558 /* 3559 * Source and destination refer to the same object. 3560 */ 3561 if (fvp == tvp) { 3562 if (retain) 3563 error = -1; 3564 else if (fromnd.ni_dvp == tdvp && 3565 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3566 !memcmp(fromnd.ni_cnd.cn_nameptr, 3567 tond.ni_cnd.cn_nameptr, 3568 fromnd.ni_cnd.cn_namelen)) 3569 error = -1; 3570 } 3571 3572 #if NVERIEXEC > 0 3573 if (!error) { 3574 char *f1, *f2; 3575 size_t f1_len; 3576 size_t f2_len; 3577 3578 f1_len = fromnd.ni_cnd.cn_namelen + 1; 3579 f1 = kmem_alloc(f1_len, KM_SLEEP); 3580 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len); 3581 3582 f2_len = tond.ni_cnd.cn_namelen + 1; 3583 f2 = kmem_alloc(f2_len, KM_SLEEP); 3584 strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len); 3585 3586 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3587 3588 kmem_free(f1, f1_len); 3589 kmem_free(f2, f2_len); 3590 } 3591 #endif /* NVERIEXEC > 0 */ 3592 3593 out: 3594 p = l->l_proc; 3595 if (!error) { 3596 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3597 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3598 VFS_RENAMELOCK_EXIT(fs); 3599 } else { 3600 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3601 if (tdvp == tvp) 3602 vrele(tdvp); 3603 else 3604 vput(tdvp); 3605 if (tvp) 3606 vput(tvp); 3607 VFS_RENAMELOCK_EXIT(fs); 3608 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3609 vrele(fromnd.ni_dvp); 3610 vrele(fvp); 3611 } 3612 out1: 3613 pathbuf_destroy(frompb); 3614 pathbuf_destroy(topb); 3615 return (error == -1 ? 0 : error); 3616 } 3617 3618 /* 3619 * Make a directory file. 3620 */ 3621 /* ARGSUSED */ 3622 int 3623 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3624 { 3625 /* { 3626 syscallarg(const char *) path; 3627 syscallarg(int) mode; 3628 } */ 3629 3630 return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE); 3631 } 3632 3633 int 3634 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 3635 { 3636 struct proc *p = curlwp->l_proc; 3637 struct vnode *vp; 3638 struct vattr vattr; 3639 int error; 3640 struct pathbuf *pb; 3641 struct nameidata nd; 3642 3643 /* XXX bollocks, should pass in a pathbuf */ 3644 error = pathbuf_maybe_copyin(path, seg, &pb); 3645 if (error) { 3646 return error; 3647 } 3648 3649 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 3650 if ((error = namei(&nd)) != 0) { 3651 pathbuf_destroy(pb); 3652 return (error); 3653 } 3654 vp = nd.ni_vp; 3655 if (vp != NULL) { 3656 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3657 if (nd.ni_dvp == vp) 3658 vrele(nd.ni_dvp); 3659 else 3660 vput(nd.ni_dvp); 3661 vrele(vp); 3662 pathbuf_destroy(pb); 3663 return (EEXIST); 3664 } 3665 vattr_null(&vattr); 3666 vattr.va_type = VDIR; 3667 /* We will read cwdi->cwdi_cmask unlocked. */ 3668 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3669 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3670 if (!error) 3671 vput(nd.ni_vp); 3672 pathbuf_destroy(pb); 3673 return (error); 3674 } 3675 3676 /* 3677 * Remove a directory file. 3678 */ 3679 /* ARGSUSED */ 3680 int 3681 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3682 { 3683 /* { 3684 syscallarg(const char *) path; 3685 } */ 3686 struct vnode *vp; 3687 int error; 3688 struct pathbuf *pb; 3689 struct nameidata nd; 3690 3691 error = pathbuf_copyin(SCARG(uap, path), &pb); 3692 if (error) { 3693 return error; 3694 } 3695 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 3696 if ((error = namei(&nd)) != 0) { 3697 pathbuf_destroy(pb); 3698 return error; 3699 } 3700 vp = nd.ni_vp; 3701 if (vp->v_type != VDIR) { 3702 error = ENOTDIR; 3703 goto out; 3704 } 3705 /* 3706 * No rmdir "." please. 3707 */ 3708 if (nd.ni_dvp == vp) { 3709 error = EINVAL; 3710 goto out; 3711 } 3712 /* 3713 * The root of a mounted filesystem cannot be deleted. 3714 */ 3715 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3716 error = EBUSY; 3717 goto out; 3718 } 3719 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3720 pathbuf_destroy(pb); 3721 return (error); 3722 3723 out: 3724 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3725 if (nd.ni_dvp == vp) 3726 vrele(nd.ni_dvp); 3727 else 3728 vput(nd.ni_dvp); 3729 vput(vp); 3730 pathbuf_destroy(pb); 3731 return (error); 3732 } 3733 3734 /* 3735 * Read a block of directory entries in a file system independent format. 3736 */ 3737 int 3738 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3739 { 3740 /* { 3741 syscallarg(int) fd; 3742 syscallarg(char *) buf; 3743 syscallarg(size_t) count; 3744 } */ 3745 file_t *fp; 3746 int error, done; 3747 3748 /* fd_getvnode() will use the descriptor for us */ 3749 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3750 return (error); 3751 if ((fp->f_flag & FREAD) == 0) { 3752 error = EBADF; 3753 goto out; 3754 } 3755 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3756 SCARG(uap, count), &done, l, 0, 0); 3757 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3758 *retval = done; 3759 out: 3760 fd_putfile(SCARG(uap, fd)); 3761 return (error); 3762 } 3763 3764 /* 3765 * Set the mode mask for creation of filesystem nodes. 3766 */ 3767 int 3768 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3769 { 3770 /* { 3771 syscallarg(mode_t) newmask; 3772 } */ 3773 struct proc *p = l->l_proc; 3774 struct cwdinfo *cwdi; 3775 3776 /* 3777 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3778 * important is that we serialize changes to the mask. The 3779 * rw_exit() will issue a write memory barrier on our behalf, 3780 * and force the changes out to other CPUs (as it must use an 3781 * atomic operation, draining the local CPU's store buffers). 3782 */ 3783 cwdi = p->p_cwdi; 3784 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3785 *retval = cwdi->cwdi_cmask; 3786 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3787 rw_exit(&cwdi->cwdi_lock); 3788 3789 return (0); 3790 } 3791 3792 int 3793 dorevoke(struct vnode *vp, kauth_cred_t cred) 3794 { 3795 struct vattr vattr; 3796 int error; 3797 3798 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3799 return error; 3800 if (kauth_cred_geteuid(cred) == vattr.va_uid || 3801 (error = kauth_authorize_generic(cred, 3802 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3803 VOP_REVOKE(vp, REVOKEALL); 3804 return (error); 3805 } 3806 3807 /* 3808 * Void all references to file by ripping underlying filesystem 3809 * away from vnode. 3810 */ 3811 /* ARGSUSED */ 3812 int 3813 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3814 { 3815 /* { 3816 syscallarg(const char *) path; 3817 } */ 3818 struct vnode *vp; 3819 int error; 3820 3821 error = namei_simple_user(SCARG(uap, path), 3822 NSM_FOLLOW_TRYEMULROOT, &vp); 3823 if (error != 0) 3824 return (error); 3825 error = dorevoke(vp, l->l_cred); 3826 vrele(vp); 3827 return (error); 3828 } 3829
Indexes created Fri Sep 26 20:09:58 GMT 2025