vfs_syscalls.c revision 1.350.2.9
1 /* $NetBSD: vfs_syscalls.c,v 1.350.2.9 2010/10/09 03:32:33 yamt Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.350.2.9 2010/10/09 03:32:33 yamt Exp $"); 70 71 #ifdef _KERNEL_OPT 72 #include "opt_fileassoc.h" 73 #include "veriexec.h" 74 #endif 75 76 #include <sys/param.h> 77 #include <sys/systm.h> 78 #include <sys/namei.h> 79 #include <sys/filedesc.h> 80 #include <sys/kernel.h> 81 #include <sys/file.h> 82 #include <sys/stat.h> 83 #include <sys/vnode.h> 84 #include <sys/mount.h> 85 #include <sys/proc.h> 86 #include <sys/uio.h> 87 #include <sys/kmem.h> 88 #include <sys/dirent.h> 89 #include <sys/sysctl.h> 90 #include <sys/syscallargs.h> 91 #include <sys/vfs_syscalls.h> 92 #include <sys/ktrace.h> 93 #ifdef FILEASSOC 94 #include <sys/fileassoc.h> 95 #endif /* FILEASSOC */ 96 #include <sys/verified_exec.h> 97 #include <sys/kauth.h> 98 #include <sys/atomic.h> 99 #include <sys/module.h> 100 #include <sys/buf.h> 101 102 #include <miscfs/genfs/genfs.h> 103 #include <miscfs/syncfs/syncfs.h> 104 #include <miscfs/specfs/specdev.h> 105 106 #include <nfs/rpcv2.h> 107 #include <nfs/nfsproto.h> 108 #include <nfs/nfs.h> 109 #include <nfs/nfs_var.h> 110 111 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 112 113 static int change_flags(struct vnode *, u_long, struct lwp *); 114 static int change_mode(struct vnode *, int, struct lwp *l); 115 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 116 117 void checkdirs(struct vnode *); 118 119 /* 120 * Virtual File System System Calls 121 */ 122 123 /* 124 * Mount a file system. 125 */ 126 127 /* 128 * This table is used to maintain compatibility with 4.3BSD 129 * and NetBSD 0.9 mount syscalls - and possibly other systems. 130 * Note, the order is important! 131 * 132 * Do not modify this table. It should only contain filesystems 133 * supported by NetBSD 0.9 and 4.3BSD. 134 */ 135 const char * const mountcompatnames[] = { 136 NULL, /* 0 = MOUNT_NONE */ 137 MOUNT_FFS, /* 1 = MOUNT_UFS */ 138 MOUNT_NFS, /* 2 */ 139 MOUNT_MFS, /* 3 */ 140 MOUNT_MSDOS, /* 4 */ 141 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 142 MOUNT_FDESC, /* 6 */ 143 MOUNT_KERNFS, /* 7 */ 144 NULL, /* 8 = MOUNT_DEVFS */ 145 MOUNT_AFS, /* 9 */ 146 }; 147 const int nmountcompatnames = sizeof(mountcompatnames) / 148 sizeof(mountcompatnames[0]); 149 150 static int 151 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 152 void *data, size_t *data_len) 153 { 154 struct mount *mp; 155 int error = 0, saved_flags; 156 157 mp = vp->v_mount; 158 saved_flags = mp->mnt_flag; 159 160 /* We can operate only on VV_ROOT nodes. */ 161 if ((vp->v_vflag & VV_ROOT) == 0) { 162 error = EINVAL; 163 goto out; 164 } 165 166 /* 167 * We only allow the filesystem to be reloaded if it 168 * is currently mounted read-only. Additionally, we 169 * prevent read-write to read-only downgrades. 170 */ 171 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 172 (mp->mnt_flag & MNT_RDONLY) == 0) { 173 error = EOPNOTSUPP; /* Needs translation */ 174 goto out; 175 } 176 177 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 178 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 179 if (error) 180 goto out; 181 182 if (vfs_busy(mp, NULL)) { 183 error = EPERM; 184 goto out; 185 } 186 187 mutex_enter(&mp->mnt_updating); 188 189 mp->mnt_flag &= ~MNT_OP_FLAGS; 190 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 191 192 /* 193 * Set the mount level flags. 194 */ 195 if (flags & MNT_RDONLY) 196 mp->mnt_flag |= MNT_RDONLY; 197 else if (mp->mnt_flag & MNT_RDONLY) 198 mp->mnt_iflag |= IMNT_WANTRDWR; 199 mp->mnt_flag &= 200 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 201 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 202 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 203 MNT_LOG); 204 mp->mnt_flag |= flags & 205 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 206 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 207 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 208 MNT_LOG | MNT_IGNORE); 209 210 error = VFS_MOUNT(mp, path, data, data_len); 211 212 if (error && data != NULL) { 213 int error2; 214 215 /* 216 * Update failed; let's try and see if it was an 217 * export request. For compat with 3.0 and earlier. 218 */ 219 error2 = vfs_hooks_reexport(mp, path, data); 220 221 /* 222 * Only update error code if the export request was 223 * understood but some problem occurred while 224 * processing it. 225 */ 226 if (error2 != EJUSTRETURN) 227 error = error2; 228 } 229 230 if (mp->mnt_iflag & IMNT_WANTRDWR) 231 mp->mnt_flag &= ~MNT_RDONLY; 232 if (error) 233 mp->mnt_flag = saved_flags; 234 mp->mnt_flag &= ~MNT_OP_FLAGS; 235 mp->mnt_iflag &= ~IMNT_WANTRDWR; 236 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 237 if (mp->mnt_syncer == NULL) 238 error = vfs_allocate_syncvnode(mp); 239 } else { 240 if (mp->mnt_syncer != NULL) 241 vfs_deallocate_syncvnode(mp); 242 } 243 mutex_exit(&mp->mnt_updating); 244 vfs_unbusy(mp, false, NULL); 245 246 out: 247 return (error); 248 } 249 250 static int 251 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 252 { 253 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 254 int error; 255 256 /* Copy file-system type from userspace. */ 257 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 258 if (error) { 259 /* 260 * Historically, filesystem types were identified by numbers. 261 * If we get an integer for the filesystem type instead of a 262 * string, we check to see if it matches one of the historic 263 * filesystem types. 264 */ 265 u_long fsindex = (u_long)fstype; 266 if (fsindex >= nmountcompatnames || 267 mountcompatnames[fsindex] == NULL) 268 return ENODEV; 269 strlcpy(fstypename, mountcompatnames[fsindex], 270 sizeof(fstypename)); 271 } 272 273 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 274 if (strcmp(fstypename, "ufs") == 0) 275 fstypename[0] = 'f'; 276 277 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 278 return 0; 279 280 /* If we can autoload a vfs module, try again */ 281 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 282 283 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 284 return 0; 285 286 return ENODEV; 287 } 288 289 static int 290 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 291 const char *path, int flags, void *data, size_t *data_len) 292 { 293 struct mount *mp; 294 struct vnode *vp = *vpp; 295 struct vattr va; 296 struct nameidata nd; 297 int error; 298 299 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 300 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 301 if (error) { 302 vfs_delref(vfsops); 303 return error; 304 } 305 306 /* Can't make a non-dir a mount-point (from here anyway). */ 307 if (vp->v_type != VDIR) { 308 vfs_delref(vfsops); 309 return ENOTDIR; 310 } 311 312 /* 313 * If the user is not root, ensure that they own the directory 314 * onto which we are attempting to mount. 315 */ 316 vn_lock(vp, LK_SHARED | LK_RETRY); 317 error = VOP_GETATTR(vp, &va, l->l_cred); 318 VOP_UNLOCK(vp); 319 if (error != 0) { 320 vfs_delref(vfsops); 321 return error; 322 } 323 if ((va.va_uid != kauth_cred_geteuid(l->l_cred) && 324 (error = kauth_authorize_generic(l->l_cred, 325 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 326 vfs_delref(vfsops); 327 return error; 328 } 329 330 if (flags & MNT_EXPORTED) { 331 vfs_delref(vfsops); 332 return EINVAL; 333 } 334 335 if ((mp = vfs_mountalloc(vfsops, vp)) == NULL) { 336 vfs_delref(vfsops); 337 return ENOMEM; 338 } 339 340 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 341 342 /* 343 * The underlying file system may refuse the mount for 344 * various reasons. Allow the user to force it to happen. 345 * 346 * Set the mount level flags. 347 */ 348 mp->mnt_flag = flags & 349 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 350 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 351 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 352 MNT_LOG | MNT_IGNORE | MNT_RDONLY); 353 354 mutex_enter(&mp->mnt_updating); 355 error = VFS_MOUNT(mp, path, data, data_len); 356 mp->mnt_flag &= ~MNT_OP_FLAGS; 357 358 if (error != 0) 359 goto err_unmounted; 360 361 /* 362 * Validate and prepare the mount point. 363 */ 364 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, 365 UIO_USERSPACE, path); 366 error = namei(&nd); 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 nameidata nd; 665 666 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 667 SCARG(uap, path)); 668 if ((error = namei(&nd)) != 0) 669 return (error); 670 vp = nd.ni_vp; 671 mp = vp->v_mount; 672 atomic_inc_uint(&mp->mnt_refcnt); 673 VOP_UNLOCK(vp); 674 675 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 676 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 677 if (error) { 678 vrele(vp); 679 vfs_destroy(mp); 680 return (error); 681 } 682 683 /* 684 * Don't allow unmounting the root file system. 685 */ 686 if (mp->mnt_flag & MNT_ROOTFS) { 687 vrele(vp); 688 vfs_destroy(mp); 689 return (EINVAL); 690 } 691 692 /* 693 * Must be the root of the filesystem 694 */ 695 if ((vp->v_vflag & VV_ROOT) == 0) { 696 vrele(vp); 697 vfs_destroy(mp); 698 return (EINVAL); 699 } 700 701 vrele(vp); 702 error = dounmount(mp, SCARG(uap, flags), l); 703 vfs_destroy(mp); 704 return error; 705 } 706 707 /* 708 * Do the actual file system unmount. File system is assumed to have 709 * been locked by the caller. 710 * 711 * => Caller hold reference to the mount, explicitly for dounmount(). 712 */ 713 int 714 dounmount(struct mount *mp, int flags, struct lwp *l) 715 { 716 struct vnode *coveredvp; 717 int error; 718 int async; 719 int used_syncer; 720 721 #if NVERIEXEC > 0 722 error = veriexec_unmountchk(mp); 723 if (error) 724 return (error); 725 #endif /* NVERIEXEC > 0 */ 726 727 /* 728 * XXX Freeze syncer. Must do this before locking the 729 * mount point. See dounmount() for details. 730 */ 731 mutex_enter(&syncer_mutex); 732 rw_enter(&mp->mnt_unmounting, RW_WRITER); 733 if ((mp->mnt_iflag & IMNT_GONE) != 0) { 734 rw_exit(&mp->mnt_unmounting); 735 mutex_exit(&syncer_mutex); 736 return ENOENT; 737 } 738 739 used_syncer = (mp->mnt_syncer != NULL); 740 741 /* 742 * XXX Syncer must be frozen when we get here. This should really 743 * be done on a per-mountpoint basis, but the syncer doesn't work 744 * like that. 745 * 746 * The caller of dounmount() must acquire syncer_mutex because 747 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 748 * order, and we must preserve that order to avoid deadlock. 749 * 750 * So, if the file system did not use the syncer, now is 751 * the time to release the syncer_mutex. 752 */ 753 if (used_syncer == 0) 754 mutex_exit(&syncer_mutex); 755 756 mp->mnt_iflag |= IMNT_UNMOUNT; 757 async = mp->mnt_flag & MNT_ASYNC; 758 mp->mnt_flag &= ~MNT_ASYNC; 759 cache_purgevfs(mp); /* remove cache entries for this file sys */ 760 if (mp->mnt_syncer != NULL) 761 vfs_deallocate_syncvnode(mp); 762 error = 0; 763 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 764 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 765 } 766 vfs_scrubvnlist(mp); 767 if (error == 0 || (flags & MNT_FORCE)) 768 error = VFS_UNMOUNT(mp, flags); 769 if (error) { 770 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 771 (void) vfs_allocate_syncvnode(mp); 772 mp->mnt_iflag &= ~IMNT_UNMOUNT; 773 mp->mnt_flag |= async; 774 rw_exit(&mp->mnt_unmounting); 775 if (used_syncer) 776 mutex_exit(&syncer_mutex); 777 return (error); 778 } 779 vfs_scrubvnlist(mp); 780 mutex_enter(&mountlist_lock); 781 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 782 coveredvp->v_mountedhere = NULL; 783 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 784 mp->mnt_iflag |= IMNT_GONE; 785 mutex_exit(&mountlist_lock); 786 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 787 panic("unmount: dangling vnode"); 788 if (used_syncer) 789 mutex_exit(&syncer_mutex); 790 vfs_hooks_unmount(mp); 791 rw_exit(&mp->mnt_unmounting); 792 vfs_destroy(mp); /* reference from mount() */ 793 if (coveredvp != NULLVP) 794 vrele(coveredvp); 795 return (0); 796 } 797 798 /* 799 * Sync each mounted filesystem. 800 */ 801 #ifdef DEBUG 802 int syncprt = 0; 803 struct ctldebug debug0 = { "syncprt", &syncprt }; 804 #endif 805 806 /* ARGSUSED */ 807 int 808 sys_sync(struct lwp *l, const void *v, register_t *retval) 809 { 810 struct mount *mp, *nmp; 811 int asyncflag; 812 813 if (l == NULL) 814 l = &lwp0; 815 816 mutex_enter(&mountlist_lock); 817 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 818 mp = nmp) { 819 if (vfs_busy(mp, &nmp)) { 820 continue; 821 } 822 mutex_enter(&mp->mnt_updating); 823 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 824 asyncflag = mp->mnt_flag & MNT_ASYNC; 825 mp->mnt_flag &= ~MNT_ASYNC; 826 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 827 if (asyncflag) 828 mp->mnt_flag |= MNT_ASYNC; 829 } 830 mutex_exit(&mp->mnt_updating); 831 vfs_unbusy(mp, false, &nmp); 832 } 833 mutex_exit(&mountlist_lock); 834 #ifdef DEBUG 835 if (syncprt) 836 vfs_bufstats(); 837 #endif /* DEBUG */ 838 return (0); 839 } 840 841 /* 842 * Change filesystem quotas. 843 */ 844 /* ARGSUSED */ 845 int 846 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 847 { 848 /* { 849 syscallarg(const char *) path; 850 syscallarg(int) cmd; 851 syscallarg(int) uid; 852 syscallarg(void *) arg; 853 } */ 854 struct mount *mp; 855 int error; 856 struct vnode *vp; 857 858 error = namei_simple_user(SCARG(uap, path), 859 NSM_FOLLOW_TRYEMULROOT, &vp); 860 if (error != 0) 861 return (error); 862 mp = vp->v_mount; 863 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 864 SCARG(uap, arg)); 865 vrele(vp); 866 return (error); 867 } 868 869 int 870 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 871 int root) 872 { 873 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 874 int error = 0; 875 876 /* 877 * If MNT_NOWAIT or MNT_LAZY is specified, do not 878 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 879 * overrides MNT_NOWAIT. 880 */ 881 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 882 (flags != MNT_WAIT && flags != 0)) { 883 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 884 goto done; 885 } 886 887 /* Get the filesystem stats now */ 888 memset(sp, 0, sizeof(*sp)); 889 if ((error = VFS_STATVFS(mp, sp)) != 0) { 890 return error; 891 } 892 893 if (cwdi->cwdi_rdir == NULL) 894 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 895 done: 896 if (cwdi->cwdi_rdir != NULL) { 897 size_t len; 898 char *bp; 899 char c; 900 char *path = PNBUF_GET(); 901 902 bp = path + MAXPATHLEN; 903 *--bp = '\0'; 904 rw_enter(&cwdi->cwdi_lock, RW_READER); 905 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 906 MAXPATHLEN / 2, 0, l); 907 rw_exit(&cwdi->cwdi_lock); 908 if (error) { 909 PNBUF_PUT(path); 910 return error; 911 } 912 len = strlen(bp); 913 if (len != 1) { 914 /* 915 * for mount points that are below our root, we can see 916 * them, so we fix up the pathname and return them. The 917 * rest we cannot see, so we don't allow viewing the 918 * data. 919 */ 920 if (strncmp(bp, sp->f_mntonname, len) == 0 && 921 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 922 (void)strlcpy(sp->f_mntonname, 923 c == '\0' ? "/" : &sp->f_mntonname[len], 924 sizeof(sp->f_mntonname)); 925 } else { 926 if (root) 927 (void)strlcpy(sp->f_mntonname, "/", 928 sizeof(sp->f_mntonname)); 929 else 930 error = EPERM; 931 } 932 } 933 PNBUF_PUT(path); 934 } 935 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 936 return error; 937 } 938 939 /* 940 * Get filesystem statistics by path. 941 */ 942 int 943 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 944 { 945 struct mount *mp; 946 int error; 947 struct vnode *vp; 948 949 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 950 if (error != 0) 951 return error; 952 mp = vp->v_mount; 953 error = dostatvfs(mp, sb, l, flags, 1); 954 vrele(vp); 955 return error; 956 } 957 958 /* ARGSUSED */ 959 int 960 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 961 { 962 /* { 963 syscallarg(const char *) path; 964 syscallarg(struct statvfs *) buf; 965 syscallarg(int) flags; 966 } */ 967 struct statvfs *sb; 968 int error; 969 970 sb = STATVFSBUF_GET(); 971 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 972 if (error == 0) 973 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 974 STATVFSBUF_PUT(sb); 975 return error; 976 } 977 978 /* 979 * Get filesystem statistics by fd. 980 */ 981 int 982 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 983 { 984 file_t *fp; 985 struct mount *mp; 986 int error; 987 988 /* fd_getvnode() will use the descriptor for us */ 989 if ((error = fd_getvnode(fd, &fp)) != 0) 990 return (error); 991 mp = ((struct vnode *)fp->f_data)->v_mount; 992 error = dostatvfs(mp, sb, curlwp, flags, 1); 993 fd_putfile(fd); 994 return error; 995 } 996 997 /* ARGSUSED */ 998 int 999 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 1000 { 1001 /* { 1002 syscallarg(int) fd; 1003 syscallarg(struct statvfs *) buf; 1004 syscallarg(int) flags; 1005 } */ 1006 struct statvfs *sb; 1007 int error; 1008 1009 sb = STATVFSBUF_GET(); 1010 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1011 if (error == 0) 1012 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1013 STATVFSBUF_PUT(sb); 1014 return error; 1015 } 1016 1017 1018 /* 1019 * Get statistics on all filesystems. 1020 */ 1021 int 1022 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1023 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1024 register_t *retval) 1025 { 1026 int root = 0; 1027 struct proc *p = l->l_proc; 1028 struct mount *mp, *nmp; 1029 struct statvfs *sb; 1030 size_t count, maxcount; 1031 int error = 0; 1032 1033 sb = STATVFSBUF_GET(); 1034 maxcount = bufsize / entry_sz; 1035 mutex_enter(&mountlist_lock); 1036 count = 0; 1037 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1038 mp = nmp) { 1039 if (vfs_busy(mp, &nmp)) { 1040 continue; 1041 } 1042 if (sfsp && count < maxcount) { 1043 error = dostatvfs(mp, sb, l, flags, 0); 1044 if (error) { 1045 vfs_unbusy(mp, false, &nmp); 1046 error = 0; 1047 continue; 1048 } 1049 error = copyfn(sb, sfsp, entry_sz); 1050 if (error) { 1051 vfs_unbusy(mp, false, NULL); 1052 goto out; 1053 } 1054 sfsp = (char *)sfsp + entry_sz; 1055 root |= strcmp(sb->f_mntonname, "/") == 0; 1056 } 1057 count++; 1058 vfs_unbusy(mp, false, &nmp); 1059 } 1060 mutex_exit(&mountlist_lock); 1061 1062 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1063 /* 1064 * fake a root entry 1065 */ 1066 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1067 sb, l, flags, 1); 1068 if (error != 0) 1069 goto out; 1070 if (sfsp) { 1071 error = copyfn(sb, sfsp, entry_sz); 1072 if (error != 0) 1073 goto out; 1074 } 1075 count++; 1076 } 1077 if (sfsp && count > maxcount) 1078 *retval = maxcount; 1079 else 1080 *retval = count; 1081 out: 1082 STATVFSBUF_PUT(sb); 1083 return error; 1084 } 1085 1086 int 1087 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1088 { 1089 /* { 1090 syscallarg(struct statvfs *) buf; 1091 syscallarg(size_t) bufsize; 1092 syscallarg(int) flags; 1093 } */ 1094 1095 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1096 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1097 } 1098 1099 /* 1100 * Change current working directory to a given file descriptor. 1101 */ 1102 /* ARGSUSED */ 1103 int 1104 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1105 { 1106 /* { 1107 syscallarg(int) fd; 1108 } */ 1109 struct proc *p = l->l_proc; 1110 struct cwdinfo *cwdi; 1111 struct vnode *vp, *tdp; 1112 struct mount *mp; 1113 file_t *fp; 1114 int error, fd; 1115 1116 /* fd_getvnode() will use the descriptor for us */ 1117 fd = SCARG(uap, fd); 1118 if ((error = fd_getvnode(fd, &fp)) != 0) 1119 return (error); 1120 vp = fp->f_data; 1121 1122 vref(vp); 1123 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1124 if (vp->v_type != VDIR) 1125 error = ENOTDIR; 1126 else 1127 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1128 if (error) { 1129 vput(vp); 1130 goto out; 1131 } 1132 while ((mp = vp->v_mountedhere) != NULL) { 1133 error = vfs_busy(mp, NULL); 1134 vput(vp); 1135 if (error != 0) 1136 goto out; 1137 error = VFS_ROOT(mp, &tdp); 1138 vfs_unbusy(mp, false, NULL); 1139 if (error) 1140 goto out; 1141 vp = tdp; 1142 } 1143 VOP_UNLOCK(vp); 1144 1145 /* 1146 * Disallow changing to a directory not under the process's 1147 * current root directory (if there is one). 1148 */ 1149 cwdi = p->p_cwdi; 1150 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1151 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1152 vrele(vp); 1153 error = EPERM; /* operation not permitted */ 1154 } else { 1155 vrele(cwdi->cwdi_cdir); 1156 cwdi->cwdi_cdir = vp; 1157 } 1158 rw_exit(&cwdi->cwdi_lock); 1159 1160 out: 1161 fd_putfile(fd); 1162 return (error); 1163 } 1164 1165 /* 1166 * Change this process's notion of the root directory to a given file 1167 * descriptor. 1168 */ 1169 int 1170 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1171 { 1172 struct proc *p = l->l_proc; 1173 struct vnode *vp; 1174 file_t *fp; 1175 int error, fd = SCARG(uap, fd); 1176 1177 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1178 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1179 return error; 1180 /* fd_getvnode() will use the descriptor for us */ 1181 if ((error = fd_getvnode(fd, &fp)) != 0) 1182 return error; 1183 vp = fp->f_data; 1184 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1185 if (vp->v_type != VDIR) 1186 error = ENOTDIR; 1187 else 1188 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1189 VOP_UNLOCK(vp); 1190 if (error) 1191 goto out; 1192 vref(vp); 1193 1194 change_root(p->p_cwdi, vp, l); 1195 1196 out: 1197 fd_putfile(fd); 1198 return (error); 1199 } 1200 1201 /* 1202 * Change current working directory (``.''). 1203 */ 1204 /* ARGSUSED */ 1205 int 1206 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1207 { 1208 /* { 1209 syscallarg(const char *) path; 1210 } */ 1211 struct proc *p = l->l_proc; 1212 struct cwdinfo *cwdi; 1213 int error; 1214 struct vnode *vp; 1215 1216 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1217 &vp, l)) != 0) 1218 return (error); 1219 cwdi = p->p_cwdi; 1220 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1221 vrele(cwdi->cwdi_cdir); 1222 cwdi->cwdi_cdir = vp; 1223 rw_exit(&cwdi->cwdi_lock); 1224 return (0); 1225 } 1226 1227 /* 1228 * Change notion of root (``/'') directory. 1229 */ 1230 /* ARGSUSED */ 1231 int 1232 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1233 { 1234 /* { 1235 syscallarg(const char *) path; 1236 } */ 1237 struct proc *p = l->l_proc; 1238 int error; 1239 struct vnode *vp; 1240 1241 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1242 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1243 return (error); 1244 if ((error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, 1245 &vp, l)) != 0) 1246 return (error); 1247 1248 change_root(p->p_cwdi, vp, l); 1249 1250 return (0); 1251 } 1252 1253 /* 1254 * Common routine for chroot and fchroot. 1255 * NB: callers need to properly authorize the change root operation. 1256 */ 1257 void 1258 change_root(struct cwdinfo *cwdi, struct vnode *vp, struct lwp *l) 1259 { 1260 1261 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1262 if (cwdi->cwdi_rdir != NULL) 1263 vrele(cwdi->cwdi_rdir); 1264 cwdi->cwdi_rdir = vp; 1265 1266 /* 1267 * Prevent escaping from chroot by putting the root under 1268 * the working directory. Silently chdir to / if we aren't 1269 * already there. 1270 */ 1271 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1272 /* 1273 * XXX would be more failsafe to change directory to a 1274 * deadfs node here instead 1275 */ 1276 vrele(cwdi->cwdi_cdir); 1277 vref(vp); 1278 cwdi->cwdi_cdir = vp; 1279 } 1280 rw_exit(&cwdi->cwdi_lock); 1281 } 1282 1283 /* 1284 * Common routine for chroot and chdir. 1285 */ 1286 int 1287 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1288 { 1289 struct nameidata nd; 1290 int error; 1291 1292 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, where, 1293 path); 1294 if ((error = namei(&nd)) != 0) 1295 return (error); 1296 *vpp = nd.ni_vp; 1297 if ((*vpp)->v_type != VDIR) 1298 error = ENOTDIR; 1299 else 1300 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1301 1302 if (error) 1303 vput(*vpp); 1304 else 1305 VOP_UNLOCK(*vpp); 1306 return (error); 1307 } 1308 1309 /* 1310 * Check permissions, allocate an open file structure, 1311 * and call the device open routine if any. 1312 */ 1313 int 1314 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1315 { 1316 /* { 1317 syscallarg(const char *) path; 1318 syscallarg(int) flags; 1319 syscallarg(int) mode; 1320 } */ 1321 struct proc *p = l->l_proc; 1322 struct cwdinfo *cwdi = p->p_cwdi; 1323 file_t *fp; 1324 struct vnode *vp; 1325 int flags, cmode; 1326 int type, indx, error; 1327 struct flock lf; 1328 struct nameidata nd; 1329 1330 flags = FFLAGS(SCARG(uap, flags)); 1331 if ((flags & (FREAD | FWRITE)) == 0) 1332 return (EINVAL); 1333 if ((error = fd_allocfile(&fp, &indx)) != 0) 1334 return (error); 1335 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1336 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1337 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1338 SCARG(uap, path)); 1339 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1340 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1341 fd_abort(p, fp, indx); 1342 if ((error == EDUPFD || error == EMOVEFD) && 1343 l->l_dupfd >= 0 && /* XXX from fdopen */ 1344 (error = 1345 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1346 *retval = indx; 1347 return (0); 1348 } 1349 if (error == ERESTART) 1350 error = EINTR; 1351 return (error); 1352 } 1353 1354 l->l_dupfd = 0; 1355 vp = nd.ni_vp; 1356 fp->f_flag = flags & FMASK; 1357 fp->f_type = DTYPE_VNODE; 1358 fp->f_ops = &vnops; 1359 fp->f_data = vp; 1360 if (flags & (O_EXLOCK | O_SHLOCK)) { 1361 lf.l_whence = SEEK_SET; 1362 lf.l_start = 0; 1363 lf.l_len = 0; 1364 if (flags & O_EXLOCK) 1365 lf.l_type = F_WRLCK; 1366 else 1367 lf.l_type = F_RDLCK; 1368 type = F_FLOCK; 1369 if ((flags & FNONBLOCK) == 0) 1370 type |= F_WAIT; 1371 VOP_UNLOCK(vp); 1372 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1373 if (error) { 1374 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1375 fd_abort(p, fp, indx); 1376 return (error); 1377 } 1378 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1379 atomic_or_uint(&fp->f_flag, FHASLOCK); 1380 } 1381 VOP_UNLOCK(vp); 1382 *retval = indx; 1383 fd_affix(p, fp, indx); 1384 return (0); 1385 } 1386 1387 static void 1388 vfs__fhfree(fhandle_t *fhp) 1389 { 1390 size_t fhsize; 1391 1392 if (fhp == NULL) { 1393 return; 1394 } 1395 fhsize = FHANDLE_SIZE(fhp); 1396 kmem_free(fhp, fhsize); 1397 } 1398 1399 /* 1400 * vfs_composefh: compose a filehandle. 1401 */ 1402 1403 int 1404 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1405 { 1406 struct mount *mp; 1407 struct fid *fidp; 1408 int error; 1409 size_t needfhsize; 1410 size_t fidsize; 1411 1412 mp = vp->v_mount; 1413 fidp = NULL; 1414 if (*fh_size < FHANDLE_SIZE_MIN) { 1415 fidsize = 0; 1416 } else { 1417 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1418 if (fhp != NULL) { 1419 memset(fhp, 0, *fh_size); 1420 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1421 fidp = &fhp->fh_fid; 1422 } 1423 } 1424 error = VFS_VPTOFH(vp, fidp, &fidsize); 1425 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1426 if (error == 0 && *fh_size < needfhsize) { 1427 error = E2BIG; 1428 } 1429 *fh_size = needfhsize; 1430 return error; 1431 } 1432 1433 int 1434 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1435 { 1436 struct mount *mp; 1437 fhandle_t *fhp; 1438 size_t fhsize; 1439 size_t fidsize; 1440 int error; 1441 1442 *fhpp = NULL; 1443 mp = vp->v_mount; 1444 fidsize = 0; 1445 error = VFS_VPTOFH(vp, NULL, &fidsize); 1446 KASSERT(error != 0); 1447 if (error != E2BIG) { 1448 goto out; 1449 } 1450 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1451 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1452 if (fhp == NULL) { 1453 error = ENOMEM; 1454 goto out; 1455 } 1456 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1457 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1458 if (error == 0) { 1459 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1460 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1461 *fhpp = fhp; 1462 } else { 1463 kmem_free(fhp, fhsize); 1464 } 1465 out: 1466 return error; 1467 } 1468 1469 void 1470 vfs_composefh_free(fhandle_t *fhp) 1471 { 1472 1473 vfs__fhfree(fhp); 1474 } 1475 1476 /* 1477 * vfs_fhtovp: lookup a vnode by a filehandle. 1478 */ 1479 1480 int 1481 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1482 { 1483 struct mount *mp; 1484 int error; 1485 1486 *vpp = NULL; 1487 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1488 if (mp == NULL) { 1489 error = ESTALE; 1490 goto out; 1491 } 1492 if (mp->mnt_op->vfs_fhtovp == NULL) { 1493 error = EOPNOTSUPP; 1494 goto out; 1495 } 1496 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1497 out: 1498 return error; 1499 } 1500 1501 /* 1502 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1503 * the needed size. 1504 */ 1505 1506 int 1507 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1508 { 1509 fhandle_t *fhp; 1510 int error; 1511 1512 *fhpp = NULL; 1513 if (fhsize > FHANDLE_SIZE_MAX) { 1514 return EINVAL; 1515 } 1516 if (fhsize < FHANDLE_SIZE_MIN) { 1517 return EINVAL; 1518 } 1519 again: 1520 fhp = kmem_alloc(fhsize, KM_SLEEP); 1521 if (fhp == NULL) { 1522 return ENOMEM; 1523 } 1524 error = copyin(ufhp, fhp, fhsize); 1525 if (error == 0) { 1526 /* XXX this check shouldn't be here */ 1527 if (FHANDLE_SIZE(fhp) == fhsize) { 1528 *fhpp = fhp; 1529 return 0; 1530 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1531 /* 1532 * a kludge for nfsv2 padded handles. 1533 */ 1534 size_t sz; 1535 1536 sz = FHANDLE_SIZE(fhp); 1537 kmem_free(fhp, fhsize); 1538 fhsize = sz; 1539 goto again; 1540 } else { 1541 /* 1542 * userland told us wrong size. 1543 */ 1544 error = EINVAL; 1545 } 1546 } 1547 kmem_free(fhp, fhsize); 1548 return error; 1549 } 1550 1551 void 1552 vfs_copyinfh_free(fhandle_t *fhp) 1553 { 1554 1555 vfs__fhfree(fhp); 1556 } 1557 1558 /* 1559 * Get file handle system call 1560 */ 1561 int 1562 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1563 { 1564 /* { 1565 syscallarg(char *) fname; 1566 syscallarg(fhandle_t *) fhp; 1567 syscallarg(size_t *) fh_size; 1568 } */ 1569 struct vnode *vp; 1570 fhandle_t *fh; 1571 int error; 1572 struct nameidata nd; 1573 size_t sz; 1574 size_t usz; 1575 1576 /* 1577 * Must be super user 1578 */ 1579 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1580 0, NULL, NULL, NULL); 1581 if (error) 1582 return (error); 1583 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1584 SCARG(uap, fname)); 1585 error = namei(&nd); 1586 if (error) 1587 return (error); 1588 vp = nd.ni_vp; 1589 error = vfs_composefh_alloc(vp, &fh); 1590 vput(vp); 1591 if (error != 0) { 1592 goto out; 1593 } 1594 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1595 if (error != 0) { 1596 goto out; 1597 } 1598 sz = FHANDLE_SIZE(fh); 1599 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1600 if (error != 0) { 1601 goto out; 1602 } 1603 if (usz >= sz) { 1604 error = copyout(fh, SCARG(uap, fhp), sz); 1605 } else { 1606 error = E2BIG; 1607 } 1608 out: 1609 vfs_composefh_free(fh); 1610 return (error); 1611 } 1612 1613 /* 1614 * Open a file given a file handle. 1615 * 1616 * Check permissions, allocate an open file structure, 1617 * and call the device open routine if any. 1618 */ 1619 1620 int 1621 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1622 register_t *retval) 1623 { 1624 file_t *fp; 1625 struct vnode *vp = NULL; 1626 kauth_cred_t cred = l->l_cred; 1627 file_t *nfp; 1628 int type, indx, error=0; 1629 struct flock lf; 1630 struct vattr va; 1631 fhandle_t *fh; 1632 int flags; 1633 proc_t *p; 1634 1635 p = curproc; 1636 1637 /* 1638 * Must be super user 1639 */ 1640 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1641 0, NULL, NULL, NULL))) 1642 return (error); 1643 1644 flags = FFLAGS(oflags); 1645 if ((flags & (FREAD | FWRITE)) == 0) 1646 return (EINVAL); 1647 if ((flags & O_CREAT)) 1648 return (EINVAL); 1649 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1650 return (error); 1651 fp = nfp; 1652 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1653 if (error != 0) { 1654 goto bad; 1655 } 1656 error = vfs_fhtovp(fh, &vp); 1657 if (error != 0) { 1658 goto bad; 1659 } 1660 1661 /* Now do an effective vn_open */ 1662 1663 if (vp->v_type == VSOCK) { 1664 error = EOPNOTSUPP; 1665 goto bad; 1666 } 1667 error = vn_openchk(vp, cred, flags); 1668 if (error != 0) 1669 goto bad; 1670 if (flags & O_TRUNC) { 1671 VOP_UNLOCK(vp); /* XXX */ 1672 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1673 vattr_null(&va); 1674 va.va_size = 0; 1675 error = VOP_SETATTR(vp, &va, cred); 1676 if (error) 1677 goto bad; 1678 } 1679 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1680 goto bad; 1681 if (flags & FWRITE) { 1682 mutex_enter(&vp->v_interlock); 1683 vp->v_writecount++; 1684 mutex_exit(&vp->v_interlock); 1685 } 1686 1687 /* done with modified vn_open, now finish what sys_open does. */ 1688 1689 fp->f_flag = flags & FMASK; 1690 fp->f_type = DTYPE_VNODE; 1691 fp->f_ops = &vnops; 1692 fp->f_data = vp; 1693 if (flags & (O_EXLOCK | O_SHLOCK)) { 1694 lf.l_whence = SEEK_SET; 1695 lf.l_start = 0; 1696 lf.l_len = 0; 1697 if (flags & O_EXLOCK) 1698 lf.l_type = F_WRLCK; 1699 else 1700 lf.l_type = F_RDLCK; 1701 type = F_FLOCK; 1702 if ((flags & FNONBLOCK) == 0) 1703 type |= F_WAIT; 1704 VOP_UNLOCK(vp); 1705 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1706 if (error) { 1707 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1708 fd_abort(p, fp, indx); 1709 return (error); 1710 } 1711 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1712 atomic_or_uint(&fp->f_flag, FHASLOCK); 1713 } 1714 VOP_UNLOCK(vp); 1715 *retval = indx; 1716 fd_affix(p, fp, indx); 1717 vfs_copyinfh_free(fh); 1718 return (0); 1719 1720 bad: 1721 fd_abort(p, fp, indx); 1722 if (vp != NULL) 1723 vput(vp); 1724 vfs_copyinfh_free(fh); 1725 return (error); 1726 } 1727 1728 int 1729 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1730 { 1731 /* { 1732 syscallarg(const void *) fhp; 1733 syscallarg(size_t) fh_size; 1734 syscallarg(int) flags; 1735 } */ 1736 1737 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1738 SCARG(uap, flags), retval); 1739 } 1740 1741 int 1742 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1743 { 1744 int error; 1745 fhandle_t *fh; 1746 struct vnode *vp; 1747 1748 /* 1749 * Must be super user 1750 */ 1751 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1752 0, NULL, NULL, NULL))) 1753 return (error); 1754 1755 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1756 if (error != 0) 1757 return error; 1758 1759 error = vfs_fhtovp(fh, &vp); 1760 vfs_copyinfh_free(fh); 1761 if (error != 0) 1762 return error; 1763 1764 error = vn_stat(vp, sb); 1765 vput(vp); 1766 return error; 1767 } 1768 1769 1770 /* ARGSUSED */ 1771 int 1772 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 1773 { 1774 /* { 1775 syscallarg(const void *) fhp; 1776 syscallarg(size_t) fh_size; 1777 syscallarg(struct stat *) sb; 1778 } */ 1779 struct stat sb; 1780 int error; 1781 1782 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1783 if (error) 1784 return error; 1785 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1786 } 1787 1788 int 1789 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1790 int flags) 1791 { 1792 fhandle_t *fh; 1793 struct mount *mp; 1794 struct vnode *vp; 1795 int error; 1796 1797 /* 1798 * Must be super user 1799 */ 1800 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1801 0, NULL, NULL, NULL))) 1802 return error; 1803 1804 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1805 if (error != 0) 1806 return error; 1807 1808 error = vfs_fhtovp(fh, &vp); 1809 vfs_copyinfh_free(fh); 1810 if (error != 0) 1811 return error; 1812 1813 mp = vp->v_mount; 1814 error = dostatvfs(mp, sb, l, flags, 1); 1815 vput(vp); 1816 return error; 1817 } 1818 1819 /* ARGSUSED */ 1820 int 1821 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1822 { 1823 /* { 1824 syscallarg(const void *) fhp; 1825 syscallarg(size_t) fh_size; 1826 syscallarg(struct statvfs *) buf; 1827 syscallarg(int) flags; 1828 } */ 1829 struct statvfs *sb = STATVFSBUF_GET(); 1830 int error; 1831 1832 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1833 SCARG(uap, flags)); 1834 if (error == 0) 1835 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1836 STATVFSBUF_PUT(sb); 1837 return error; 1838 } 1839 1840 /* 1841 * Create a special file. 1842 */ 1843 /* ARGSUSED */ 1844 int 1845 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 1846 register_t *retval) 1847 { 1848 /* { 1849 syscallarg(const char *) path; 1850 syscallarg(mode_t) mode; 1851 syscallarg(dev_t) dev; 1852 } */ 1853 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 1854 SCARG(uap, dev), retval, UIO_USERSPACE); 1855 } 1856 1857 int 1858 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 1859 register_t *retval, enum uio_seg seg) 1860 { 1861 struct proc *p = l->l_proc; 1862 struct vnode *vp; 1863 struct vattr vattr; 1864 int error, optype; 1865 struct nameidata nd; 1866 char *path; 1867 const char *cpath; 1868 1869 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1870 0, NULL, NULL, NULL)) != 0) 1871 return (error); 1872 1873 optype = VOP_MKNOD_DESCOFFSET; 1874 1875 VERIEXEC_PATH_GET(pathname, seg, cpath, path); 1876 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1877 1878 if ((error = namei(&nd)) != 0) 1879 goto out; 1880 vp = nd.ni_vp; 1881 if (vp != NULL) 1882 error = EEXIST; 1883 else { 1884 vattr_null(&vattr); 1885 /* We will read cwdi->cwdi_cmask unlocked. */ 1886 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1887 vattr.va_rdev = dev; 1888 1889 switch (mode & S_IFMT) { 1890 case S_IFMT: /* used by badsect to flag bad sectors */ 1891 vattr.va_type = VBAD; 1892 break; 1893 case S_IFCHR: 1894 vattr.va_type = VCHR; 1895 break; 1896 case S_IFBLK: 1897 vattr.va_type = VBLK; 1898 break; 1899 case S_IFWHT: 1900 optype = VOP_WHITEOUT_DESCOFFSET; 1901 break; 1902 case S_IFREG: 1903 #if NVERIEXEC > 0 1904 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1905 O_CREAT); 1906 #endif /* NVERIEXEC > 0 */ 1907 vattr.va_type = VREG; 1908 vattr.va_rdev = VNOVAL; 1909 optype = VOP_CREATE_DESCOFFSET; 1910 break; 1911 default: 1912 error = EINVAL; 1913 break; 1914 } 1915 } 1916 if (!error) { 1917 switch (optype) { 1918 case VOP_WHITEOUT_DESCOFFSET: 1919 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1920 if (error) 1921 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1922 vput(nd.ni_dvp); 1923 break; 1924 1925 case VOP_MKNOD_DESCOFFSET: 1926 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1927 &nd.ni_cnd, &vattr); 1928 if (error == 0) 1929 vput(nd.ni_vp); 1930 break; 1931 1932 case VOP_CREATE_DESCOFFSET: 1933 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1934 &nd.ni_cnd, &vattr); 1935 if (error == 0) 1936 vput(nd.ni_vp); 1937 break; 1938 } 1939 } else { 1940 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1941 if (nd.ni_dvp == vp) 1942 vrele(nd.ni_dvp); 1943 else 1944 vput(nd.ni_dvp); 1945 if (vp) 1946 vrele(vp); 1947 } 1948 out: 1949 VERIEXEC_PATH_PUT(path); 1950 return (error); 1951 } 1952 1953 /* 1954 * Create a named pipe. 1955 */ 1956 /* ARGSUSED */ 1957 int 1958 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1959 { 1960 /* { 1961 syscallarg(const char *) path; 1962 syscallarg(int) mode; 1963 } */ 1964 struct proc *p = l->l_proc; 1965 struct vattr vattr; 1966 int error; 1967 struct nameidata nd; 1968 1969 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1970 SCARG(uap, path)); 1971 if ((error = namei(&nd)) != 0) 1972 return (error); 1973 if (nd.ni_vp != NULL) { 1974 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1975 if (nd.ni_dvp == nd.ni_vp) 1976 vrele(nd.ni_dvp); 1977 else 1978 vput(nd.ni_dvp); 1979 vrele(nd.ni_vp); 1980 return (EEXIST); 1981 } 1982 vattr_null(&vattr); 1983 vattr.va_type = VFIFO; 1984 /* We will read cwdi->cwdi_cmask unlocked. */ 1985 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1986 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1987 if (error == 0) 1988 vput(nd.ni_vp); 1989 return (error); 1990 } 1991 1992 /* 1993 * Make a hard file link. 1994 */ 1995 /* ARGSUSED */ 1996 int 1997 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1998 { 1999 /* { 2000 syscallarg(const char *) path; 2001 syscallarg(const char *) link; 2002 } */ 2003 struct vnode *vp; 2004 struct nameidata nd; 2005 int error; 2006 2007 error = namei_simple_user(SCARG(uap, path), 2008 NSM_FOLLOW_TRYEMULROOT, &vp); 2009 if (error != 0) 2010 return (error); 2011 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2012 SCARG(uap, link)); 2013 if ((error = namei(&nd)) != 0) 2014 goto out; 2015 if (nd.ni_vp) { 2016 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2017 if (nd.ni_dvp == nd.ni_vp) 2018 vrele(nd.ni_dvp); 2019 else 2020 vput(nd.ni_dvp); 2021 vrele(nd.ni_vp); 2022 error = EEXIST; 2023 goto out; 2024 } 2025 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2026 out: 2027 vrele(vp); 2028 return (error); 2029 } 2030 2031 int 2032 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2033 { 2034 struct proc *p = curproc; 2035 struct vattr vattr; 2036 char *path; 2037 int error; 2038 struct nameidata nd; 2039 2040 path = PNBUF_GET(); 2041 if (seg == UIO_USERSPACE) { 2042 if ((error = copyinstr(patharg, path, MAXPATHLEN, NULL)) != 0) 2043 goto out; 2044 } else { 2045 KASSERT(strlen(patharg) < MAXPATHLEN); 2046 strcpy(path, patharg); 2047 } 2048 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, link); 2049 if ((error = namei(&nd)) != 0) 2050 goto out; 2051 if (nd.ni_vp) { 2052 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2053 if (nd.ni_dvp == nd.ni_vp) 2054 vrele(nd.ni_dvp); 2055 else 2056 vput(nd.ni_dvp); 2057 vrele(nd.ni_vp); 2058 error = EEXIST; 2059 goto out; 2060 } 2061 vattr_null(&vattr); 2062 vattr.va_type = VLNK; 2063 /* We will read cwdi->cwdi_cmask unlocked. */ 2064 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2065 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2066 if (error == 0) 2067 vput(nd.ni_vp); 2068 out: 2069 PNBUF_PUT(path); 2070 return (error); 2071 } 2072 2073 /* 2074 * Make a symbolic link. 2075 */ 2076 /* ARGSUSED */ 2077 int 2078 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2079 { 2080 /* { 2081 syscallarg(const char *) path; 2082 syscallarg(const char *) link; 2083 } */ 2084 2085 return do_sys_symlink(SCARG(uap, path), SCARG(uap, link), 2086 UIO_USERSPACE); 2087 } 2088 2089 /* 2090 * Delete a whiteout from the filesystem. 2091 */ 2092 /* ARGSUSED */ 2093 int 2094 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2095 { 2096 /* { 2097 syscallarg(const char *) path; 2098 } */ 2099 int error; 2100 struct nameidata nd; 2101 2102 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2103 UIO_USERSPACE, SCARG(uap, path)); 2104 error = namei(&nd); 2105 if (error) 2106 return (error); 2107 2108 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2109 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2110 if (nd.ni_dvp == nd.ni_vp) 2111 vrele(nd.ni_dvp); 2112 else 2113 vput(nd.ni_dvp); 2114 if (nd.ni_vp) 2115 vrele(nd.ni_vp); 2116 return (EEXIST); 2117 } 2118 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2119 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2120 vput(nd.ni_dvp); 2121 return (error); 2122 } 2123 2124 /* 2125 * Delete a name from the filesystem. 2126 */ 2127 /* ARGSUSED */ 2128 int 2129 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2130 { 2131 /* { 2132 syscallarg(const char *) path; 2133 } */ 2134 2135 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2136 } 2137 2138 int 2139 do_sys_unlink(const char *arg, enum uio_seg seg) 2140 { 2141 struct vnode *vp; 2142 int error; 2143 struct nameidata nd; 2144 char *path; 2145 const char *cpath; 2146 2147 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2148 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2149 2150 if ((error = namei(&nd)) != 0) 2151 goto out; 2152 vp = nd.ni_vp; 2153 2154 /* 2155 * The root of a mounted filesystem cannot be deleted. 2156 */ 2157 if (vp->v_vflag & VV_ROOT) { 2158 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2159 if (nd.ni_dvp == vp) 2160 vrele(nd.ni_dvp); 2161 else 2162 vput(nd.ni_dvp); 2163 vput(vp); 2164 error = EBUSY; 2165 goto out; 2166 } 2167 2168 #if NVERIEXEC > 0 2169 /* Handle remove requests for veriexec entries. */ 2170 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2171 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2172 if (nd.ni_dvp == vp) 2173 vrele(nd.ni_dvp); 2174 else 2175 vput(nd.ni_dvp); 2176 vput(vp); 2177 goto out; 2178 } 2179 #endif /* NVERIEXEC > 0 */ 2180 2181 #ifdef FILEASSOC 2182 (void)fileassoc_file_delete(vp); 2183 #endif /* FILEASSOC */ 2184 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2185 out: 2186 VERIEXEC_PATH_PUT(path); 2187 return (error); 2188 } 2189 2190 /* 2191 * Reposition read/write file offset. 2192 */ 2193 int 2194 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2195 { 2196 /* { 2197 syscallarg(int) fd; 2198 syscallarg(int) pad; 2199 syscallarg(off_t) offset; 2200 syscallarg(int) whence; 2201 } */ 2202 kauth_cred_t cred = l->l_cred; 2203 file_t *fp; 2204 struct vnode *vp; 2205 struct vattr vattr; 2206 off_t newoff; 2207 int error, fd; 2208 2209 fd = SCARG(uap, fd); 2210 2211 if ((fp = fd_getfile(fd)) == NULL) 2212 return (EBADF); 2213 2214 vp = fp->f_data; 2215 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2216 error = ESPIPE; 2217 goto out; 2218 } 2219 2220 switch (SCARG(uap, whence)) { 2221 case SEEK_CUR: 2222 newoff = fp->f_offset + SCARG(uap, offset); 2223 break; 2224 case SEEK_END: 2225 vn_lock(vp, LK_SHARED | LK_RETRY); 2226 error = VOP_GETATTR(vp, &vattr, cred); 2227 VOP_UNLOCK(vp); 2228 if (error) { 2229 goto out; 2230 } 2231 newoff = SCARG(uap, offset) + vattr.va_size; 2232 break; 2233 case SEEK_SET: 2234 newoff = SCARG(uap, offset); 2235 break; 2236 default: 2237 error = EINVAL; 2238 goto out; 2239 } 2240 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2241 *(off_t *)retval = fp->f_offset = newoff; 2242 } 2243 out: 2244 fd_putfile(fd); 2245 return (error); 2246 } 2247 2248 /* 2249 * Positional read system call. 2250 */ 2251 int 2252 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2253 { 2254 /* { 2255 syscallarg(int) fd; 2256 syscallarg(void *) buf; 2257 syscallarg(size_t) nbyte; 2258 syscallarg(off_t) offset; 2259 } */ 2260 file_t *fp; 2261 struct vnode *vp; 2262 off_t offset; 2263 int error, fd = SCARG(uap, fd); 2264 2265 if ((fp = fd_getfile(fd)) == NULL) 2266 return (EBADF); 2267 2268 if ((fp->f_flag & FREAD) == 0) { 2269 fd_putfile(fd); 2270 return (EBADF); 2271 } 2272 2273 vp = fp->f_data; 2274 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2275 error = ESPIPE; 2276 goto out; 2277 } 2278 2279 offset = SCARG(uap, offset); 2280 2281 /* 2282 * XXX This works because no file systems actually 2283 * XXX take any action on the seek operation. 2284 */ 2285 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2286 goto out; 2287 2288 /* dofileread() will unuse the descriptor for us */ 2289 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2290 &offset, 0, retval)); 2291 2292 out: 2293 fd_putfile(fd); 2294 return (error); 2295 } 2296 2297 /* 2298 * Positional scatter read system call. 2299 */ 2300 int 2301 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2302 { 2303 /* { 2304 syscallarg(int) fd; 2305 syscallarg(const struct iovec *) iovp; 2306 syscallarg(int) iovcnt; 2307 syscallarg(off_t) offset; 2308 } */ 2309 off_t offset = SCARG(uap, offset); 2310 2311 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2312 SCARG(uap, iovcnt), &offset, 0, retval); 2313 } 2314 2315 /* 2316 * Positional write system call. 2317 */ 2318 int 2319 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2320 { 2321 /* { 2322 syscallarg(int) fd; 2323 syscallarg(const void *) buf; 2324 syscallarg(size_t) nbyte; 2325 syscallarg(off_t) offset; 2326 } */ 2327 file_t *fp; 2328 struct vnode *vp; 2329 off_t offset; 2330 int error, fd = SCARG(uap, fd); 2331 2332 if ((fp = fd_getfile(fd)) == NULL) 2333 return (EBADF); 2334 2335 if ((fp->f_flag & FWRITE) == 0) { 2336 fd_putfile(fd); 2337 return (EBADF); 2338 } 2339 2340 vp = fp->f_data; 2341 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2342 error = ESPIPE; 2343 goto out; 2344 } 2345 2346 offset = SCARG(uap, offset); 2347 2348 /* 2349 * XXX This works because no file systems actually 2350 * XXX take any action on the seek operation. 2351 */ 2352 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2353 goto out; 2354 2355 /* dofilewrite() will unuse the descriptor for us */ 2356 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2357 &offset, 0, retval)); 2358 2359 out: 2360 fd_putfile(fd); 2361 return (error); 2362 } 2363 2364 /* 2365 * Positional gather write system call. 2366 */ 2367 int 2368 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2369 { 2370 /* { 2371 syscallarg(int) fd; 2372 syscallarg(const struct iovec *) iovp; 2373 syscallarg(int) iovcnt; 2374 syscallarg(off_t) offset; 2375 } */ 2376 off_t offset = SCARG(uap, offset); 2377 2378 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2379 SCARG(uap, iovcnt), &offset, 0, retval); 2380 } 2381 2382 /* 2383 * Check access permissions. 2384 */ 2385 int 2386 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2387 { 2388 /* { 2389 syscallarg(const char *) path; 2390 syscallarg(int) flags; 2391 } */ 2392 kauth_cred_t cred; 2393 struct vnode *vp; 2394 int error, flags; 2395 struct nameidata nd; 2396 2397 cred = kauth_cred_dup(l->l_cred); 2398 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2399 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2400 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2401 SCARG(uap, path)); 2402 /* Override default credentials */ 2403 nd.ni_cnd.cn_cred = cred; 2404 if ((error = namei(&nd)) != 0) 2405 goto out; 2406 vp = nd.ni_vp; 2407 2408 /* Flags == 0 means only check for existence. */ 2409 if (SCARG(uap, flags)) { 2410 flags = 0; 2411 if (SCARG(uap, flags) & R_OK) 2412 flags |= VREAD; 2413 if (SCARG(uap, flags) & W_OK) 2414 flags |= VWRITE; 2415 if (SCARG(uap, flags) & X_OK) 2416 flags |= VEXEC; 2417 2418 error = VOP_ACCESS(vp, flags, cred); 2419 if (!error && (flags & VWRITE)) 2420 error = vn_writechk(vp); 2421 } 2422 vput(vp); 2423 out: 2424 kauth_cred_free(cred); 2425 return (error); 2426 } 2427 2428 /* 2429 * Common code for all sys_stat functions, including compat versions. 2430 */ 2431 int 2432 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb) 2433 { 2434 int error; 2435 struct nameidata nd; 2436 2437 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2438 UIO_USERSPACE, path); 2439 error = namei(&nd); 2440 if (error != 0) 2441 return error; 2442 error = vn_stat(nd.ni_vp, sb); 2443 vput(nd.ni_vp); 2444 return error; 2445 } 2446 2447 /* 2448 * Get file status; this version follows links. 2449 */ 2450 /* ARGSUSED */ 2451 int 2452 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2453 { 2454 /* { 2455 syscallarg(const char *) path; 2456 syscallarg(struct stat *) ub; 2457 } */ 2458 struct stat sb; 2459 int error; 2460 2461 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2462 if (error) 2463 return error; 2464 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2465 } 2466 2467 /* 2468 * Get file status; this version does not follow links. 2469 */ 2470 /* ARGSUSED */ 2471 int 2472 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2473 { 2474 /* { 2475 syscallarg(const char *) path; 2476 syscallarg(struct stat *) ub; 2477 } */ 2478 struct stat sb; 2479 int error; 2480 2481 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2482 if (error) 2483 return error; 2484 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2485 } 2486 2487 /* 2488 * Get configurable pathname variables. 2489 */ 2490 /* ARGSUSED */ 2491 int 2492 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2493 { 2494 /* { 2495 syscallarg(const char *) path; 2496 syscallarg(int) name; 2497 } */ 2498 int error; 2499 struct nameidata nd; 2500 2501 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2502 SCARG(uap, path)); 2503 if ((error = namei(&nd)) != 0) 2504 return (error); 2505 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2506 vput(nd.ni_vp); 2507 return (error); 2508 } 2509 2510 /* 2511 * Return target name of a symbolic link. 2512 */ 2513 /* ARGSUSED */ 2514 int 2515 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2516 { 2517 /* { 2518 syscallarg(const char *) path; 2519 syscallarg(char *) buf; 2520 syscallarg(size_t) count; 2521 } */ 2522 struct vnode *vp; 2523 struct iovec aiov; 2524 struct uio auio; 2525 int error; 2526 struct nameidata nd; 2527 2528 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2529 SCARG(uap, path)); 2530 if ((error = namei(&nd)) != 0) 2531 return (error); 2532 vp = nd.ni_vp; 2533 if (vp->v_type != VLNK) 2534 error = EINVAL; 2535 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2536 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2537 aiov.iov_base = SCARG(uap, buf); 2538 aiov.iov_len = SCARG(uap, count); 2539 auio.uio_iov = &aiov; 2540 auio.uio_iovcnt = 1; 2541 auio.uio_offset = 0; 2542 auio.uio_rw = UIO_READ; 2543 KASSERT(l == curlwp); 2544 auio.uio_vmspace = l->l_proc->p_vmspace; 2545 auio.uio_resid = SCARG(uap, count); 2546 error = VOP_READLINK(vp, &auio, l->l_cred); 2547 } 2548 vput(vp); 2549 *retval = SCARG(uap, count) - auio.uio_resid; 2550 return (error); 2551 } 2552 2553 /* 2554 * Change flags of a file given a path name. 2555 */ 2556 /* ARGSUSED */ 2557 int 2558 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2559 { 2560 /* { 2561 syscallarg(const char *) path; 2562 syscallarg(u_long) flags; 2563 } */ 2564 struct vnode *vp; 2565 int error; 2566 2567 error = namei_simple_user(SCARG(uap, path), 2568 NSM_FOLLOW_TRYEMULROOT, &vp); 2569 if (error != 0) 2570 return (error); 2571 error = change_flags(vp, SCARG(uap, flags), l); 2572 vput(vp); 2573 return (error); 2574 } 2575 2576 /* 2577 * Change flags of a file given a file descriptor. 2578 */ 2579 /* ARGSUSED */ 2580 int 2581 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2582 { 2583 /* { 2584 syscallarg(int) fd; 2585 syscallarg(u_long) flags; 2586 } */ 2587 struct vnode *vp; 2588 file_t *fp; 2589 int error; 2590 2591 /* fd_getvnode() will use the descriptor for us */ 2592 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2593 return (error); 2594 vp = fp->f_data; 2595 error = change_flags(vp, SCARG(uap, flags), l); 2596 VOP_UNLOCK(vp); 2597 fd_putfile(SCARG(uap, fd)); 2598 return (error); 2599 } 2600 2601 /* 2602 * Change flags of a file given a path name; this version does 2603 * not follow links. 2604 */ 2605 int 2606 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2607 { 2608 /* { 2609 syscallarg(const char *) path; 2610 syscallarg(u_long) flags; 2611 } */ 2612 struct vnode *vp; 2613 int error; 2614 2615 error = namei_simple_user(SCARG(uap, path), 2616 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2617 if (error != 0) 2618 return (error); 2619 error = change_flags(vp, SCARG(uap, flags), l); 2620 vput(vp); 2621 return (error); 2622 } 2623 2624 /* 2625 * Common routine to change flags of a file. 2626 */ 2627 int 2628 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2629 { 2630 struct vattr vattr; 2631 int error; 2632 2633 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2634 /* 2635 * Non-superusers cannot change the flags on devices, even if they 2636 * own them. 2637 */ 2638 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2639 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2640 goto out; 2641 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2642 error = EINVAL; 2643 goto out; 2644 } 2645 } 2646 vattr_null(&vattr); 2647 vattr.va_flags = flags; 2648 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2649 out: 2650 return (error); 2651 } 2652 2653 /* 2654 * Change mode of a file given path name; this version follows links. 2655 */ 2656 /* ARGSUSED */ 2657 int 2658 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2659 { 2660 /* { 2661 syscallarg(const char *) path; 2662 syscallarg(int) mode; 2663 } */ 2664 int error; 2665 struct vnode *vp; 2666 2667 error = namei_simple_user(SCARG(uap, path), 2668 NSM_FOLLOW_TRYEMULROOT, &vp); 2669 if (error != 0) 2670 return (error); 2671 2672 error = change_mode(vp, SCARG(uap, mode), l); 2673 2674 vrele(vp); 2675 return (error); 2676 } 2677 2678 /* 2679 * Change mode of a file given a file descriptor. 2680 */ 2681 /* ARGSUSED */ 2682 int 2683 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2684 { 2685 /* { 2686 syscallarg(int) fd; 2687 syscallarg(int) mode; 2688 } */ 2689 file_t *fp; 2690 int error; 2691 2692 /* fd_getvnode() will use the descriptor for us */ 2693 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2694 return (error); 2695 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2696 fd_putfile(SCARG(uap, fd)); 2697 return (error); 2698 } 2699 2700 /* 2701 * Change mode of a file given path name; this version does not follow links. 2702 */ 2703 /* ARGSUSED */ 2704 int 2705 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2706 { 2707 /* { 2708 syscallarg(const char *) path; 2709 syscallarg(int) mode; 2710 } */ 2711 int error; 2712 struct vnode *vp; 2713 2714 error = namei_simple_user(SCARG(uap, path), 2715 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2716 if (error != 0) 2717 return (error); 2718 2719 error = change_mode(vp, SCARG(uap, mode), l); 2720 2721 vrele(vp); 2722 return (error); 2723 } 2724 2725 /* 2726 * Common routine to set mode given a vnode. 2727 */ 2728 static int 2729 change_mode(struct vnode *vp, int mode, struct lwp *l) 2730 { 2731 struct vattr vattr; 2732 int error; 2733 2734 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2735 vattr_null(&vattr); 2736 vattr.va_mode = mode & ALLPERMS; 2737 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2738 VOP_UNLOCK(vp); 2739 return (error); 2740 } 2741 2742 /* 2743 * Set ownership given a path name; this version follows links. 2744 */ 2745 /* ARGSUSED */ 2746 int 2747 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2748 { 2749 /* { 2750 syscallarg(const char *) path; 2751 syscallarg(uid_t) uid; 2752 syscallarg(gid_t) gid; 2753 } */ 2754 int error; 2755 struct vnode *vp; 2756 2757 error = namei_simple_user(SCARG(uap, path), 2758 NSM_FOLLOW_TRYEMULROOT, &vp); 2759 if (error != 0) 2760 return (error); 2761 2762 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2763 2764 vrele(vp); 2765 return (error); 2766 } 2767 2768 /* 2769 * Set ownership given a path name; this version follows links. 2770 * Provides POSIX semantics. 2771 */ 2772 /* ARGSUSED */ 2773 int 2774 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2775 { 2776 /* { 2777 syscallarg(const char *) path; 2778 syscallarg(uid_t) uid; 2779 syscallarg(gid_t) gid; 2780 } */ 2781 int error; 2782 struct vnode *vp; 2783 2784 error = namei_simple_user(SCARG(uap, path), 2785 NSM_FOLLOW_TRYEMULROOT, &vp); 2786 if (error != 0) 2787 return (error); 2788 2789 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2790 2791 vrele(vp); 2792 return (error); 2793 } 2794 2795 /* 2796 * Set ownership given a file descriptor. 2797 */ 2798 /* ARGSUSED */ 2799 int 2800 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2801 { 2802 /* { 2803 syscallarg(int) fd; 2804 syscallarg(uid_t) uid; 2805 syscallarg(gid_t) gid; 2806 } */ 2807 int error; 2808 file_t *fp; 2809 2810 /* fd_getvnode() will use the descriptor for us */ 2811 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2812 return (error); 2813 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2814 l, 0); 2815 fd_putfile(SCARG(uap, fd)); 2816 return (error); 2817 } 2818 2819 /* 2820 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2821 */ 2822 /* ARGSUSED */ 2823 int 2824 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2825 { 2826 /* { 2827 syscallarg(int) fd; 2828 syscallarg(uid_t) uid; 2829 syscallarg(gid_t) gid; 2830 } */ 2831 int error; 2832 file_t *fp; 2833 2834 /* fd_getvnode() will use the descriptor for us */ 2835 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2836 return (error); 2837 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2838 l, 1); 2839 fd_putfile(SCARG(uap, fd)); 2840 return (error); 2841 } 2842 2843 /* 2844 * Set ownership given a path name; this version does not follow links. 2845 */ 2846 /* ARGSUSED */ 2847 int 2848 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2849 { 2850 /* { 2851 syscallarg(const char *) path; 2852 syscallarg(uid_t) uid; 2853 syscallarg(gid_t) gid; 2854 } */ 2855 int error; 2856 struct vnode *vp; 2857 2858 error = namei_simple_user(SCARG(uap, path), 2859 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2860 if (error != 0) 2861 return (error); 2862 2863 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2864 2865 vrele(vp); 2866 return (error); 2867 } 2868 2869 /* 2870 * Set ownership given a path name; this version does not follow links. 2871 * Provides POSIX/XPG semantics. 2872 */ 2873 /* ARGSUSED */ 2874 int 2875 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2876 { 2877 /* { 2878 syscallarg(const char *) path; 2879 syscallarg(uid_t) uid; 2880 syscallarg(gid_t) gid; 2881 } */ 2882 int error; 2883 struct vnode *vp; 2884 2885 error = namei_simple_user(SCARG(uap, path), 2886 NSM_NOFOLLOW_TRYEMULROOT, &vp); 2887 if (error != 0) 2888 return (error); 2889 2890 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2891 2892 vrele(vp); 2893 return (error); 2894 } 2895 2896 /* 2897 * Common routine to set ownership given a vnode. 2898 */ 2899 static int 2900 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2901 int posix_semantics) 2902 { 2903 struct vattr vattr; 2904 mode_t newmode; 2905 int error; 2906 2907 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2908 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2909 goto out; 2910 2911 #define CHANGED(x) ((int)(x) != -1) 2912 newmode = vattr.va_mode; 2913 if (posix_semantics) { 2914 /* 2915 * POSIX/XPG semantics: if the caller is not the super-user, 2916 * clear set-user-id and set-group-id bits. Both POSIX and 2917 * the XPG consider the behaviour for calls by the super-user 2918 * implementation-defined; we leave the set-user-id and set- 2919 * group-id settings intact in that case. 2920 */ 2921 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2922 NULL) != 0) 2923 newmode &= ~(S_ISUID | S_ISGID); 2924 } else { 2925 /* 2926 * NetBSD semantics: when changing owner and/or group, 2927 * clear the respective bit(s). 2928 */ 2929 if (CHANGED(uid)) 2930 newmode &= ~S_ISUID; 2931 if (CHANGED(gid)) 2932 newmode &= ~S_ISGID; 2933 } 2934 /* Update va_mode iff altered. */ 2935 if (vattr.va_mode == newmode) 2936 newmode = VNOVAL; 2937 2938 vattr_null(&vattr); 2939 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2940 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2941 vattr.va_mode = newmode; 2942 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2943 #undef CHANGED 2944 2945 out: 2946 VOP_UNLOCK(vp); 2947 return (error); 2948 } 2949 2950 /* 2951 * Set the access and modification times given a path name; this 2952 * version follows links. 2953 */ 2954 /* ARGSUSED */ 2955 int 2956 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 2957 register_t *retval) 2958 { 2959 /* { 2960 syscallarg(const char *) path; 2961 syscallarg(const struct timeval *) tptr; 2962 } */ 2963 2964 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2965 SCARG(uap, tptr), UIO_USERSPACE); 2966 } 2967 2968 /* 2969 * Set the access and modification times given a file descriptor. 2970 */ 2971 /* ARGSUSED */ 2972 int 2973 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 2974 register_t *retval) 2975 { 2976 /* { 2977 syscallarg(int) fd; 2978 syscallarg(const struct timeval *) tptr; 2979 } */ 2980 int error; 2981 file_t *fp; 2982 2983 /* fd_getvnode() will use the descriptor for us */ 2984 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2985 return (error); 2986 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2987 UIO_USERSPACE); 2988 fd_putfile(SCARG(uap, fd)); 2989 return (error); 2990 } 2991 2992 /* 2993 * Set the access and modification times given a path name; this 2994 * version does not follow links. 2995 */ 2996 int 2997 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 2998 register_t *retval) 2999 { 3000 /* { 3001 syscallarg(const char *) path; 3002 syscallarg(const struct timeval *) tptr; 3003 } */ 3004 3005 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3006 SCARG(uap, tptr), UIO_USERSPACE); 3007 } 3008 3009 /* 3010 * Common routine to set access and modification times given a vnode. 3011 */ 3012 int 3013 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3014 const struct timeval *tptr, enum uio_seg seg) 3015 { 3016 struct vattr vattr; 3017 int error, dorele = 0; 3018 namei_simple_flags_t sflags; 3019 3020 bool vanull, setbirthtime; 3021 struct timespec ts[2]; 3022 3023 /* 3024 * I have checked all callers and they pass either FOLLOW, 3025 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 3026 * is 0. More to the point, they don't pass anything else. 3027 * Let's keep it that way at least until the namei interfaces 3028 * are fully sanitized. 3029 */ 3030 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 3031 sflags = (flag == FOLLOW) ? 3032 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 3033 3034 if (tptr == NULL) { 3035 vanull = true; 3036 nanotime(&ts[0]); 3037 ts[1] = ts[0]; 3038 } else { 3039 struct timeval tv[2]; 3040 3041 vanull = false; 3042 if (seg != UIO_SYSSPACE) { 3043 error = copyin(tptr, tv, sizeof (tv)); 3044 if (error != 0) 3045 return error; 3046 tptr = tv; 3047 } 3048 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3049 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3050 } 3051 3052 if (vp == NULL) { 3053 /* note: SEG describes TPTR, not PATH; PATH is always user */ 3054 error = namei_simple_user(path, sflags, &vp); 3055 if (error != 0) 3056 return error; 3057 dorele = 1; 3058 } 3059 3060 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3061 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3062 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3063 vattr_null(&vattr); 3064 vattr.va_atime = ts[0]; 3065 vattr.va_mtime = ts[1]; 3066 if (setbirthtime) 3067 vattr.va_birthtime = ts[1]; 3068 if (vanull) 3069 vattr.va_vaflags |= VA_UTIMES_NULL; 3070 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3071 VOP_UNLOCK(vp); 3072 3073 if (dorele != 0) 3074 vrele(vp); 3075 3076 return error; 3077 } 3078 3079 /* 3080 * Truncate a file given its path name. 3081 */ 3082 /* ARGSUSED */ 3083 int 3084 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3085 { 3086 /* { 3087 syscallarg(const char *) path; 3088 syscallarg(int) pad; 3089 syscallarg(off_t) length; 3090 } */ 3091 struct vnode *vp; 3092 struct vattr vattr; 3093 int error; 3094 3095 error = namei_simple_user(SCARG(uap, path), 3096 NSM_FOLLOW_TRYEMULROOT, &vp); 3097 if (error != 0) 3098 return (error); 3099 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3100 if (vp->v_type == VDIR) 3101 error = EISDIR; 3102 else if ((error = vn_writechk(vp)) == 0 && 3103 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3104 vattr_null(&vattr); 3105 vattr.va_size = SCARG(uap, length); 3106 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3107 } 3108 vput(vp); 3109 return (error); 3110 } 3111 3112 /* 3113 * Truncate a file given a file descriptor. 3114 */ 3115 /* ARGSUSED */ 3116 int 3117 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3118 { 3119 /* { 3120 syscallarg(int) fd; 3121 syscallarg(int) pad; 3122 syscallarg(off_t) length; 3123 } */ 3124 struct vattr vattr; 3125 struct vnode *vp; 3126 file_t *fp; 3127 int error; 3128 3129 /* fd_getvnode() will use the descriptor for us */ 3130 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3131 return (error); 3132 if ((fp->f_flag & FWRITE) == 0) { 3133 error = EINVAL; 3134 goto out; 3135 } 3136 vp = fp->f_data; 3137 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3138 if (vp->v_type == VDIR) 3139 error = EISDIR; 3140 else if ((error = vn_writechk(vp)) == 0) { 3141 vattr_null(&vattr); 3142 vattr.va_size = SCARG(uap, length); 3143 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3144 } 3145 VOP_UNLOCK(vp); 3146 out: 3147 fd_putfile(SCARG(uap, fd)); 3148 return (error); 3149 } 3150 3151 /* 3152 * Sync an open file. 3153 */ 3154 /* ARGSUSED */ 3155 int 3156 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3157 { 3158 /* { 3159 syscallarg(int) fd; 3160 } */ 3161 struct vnode *vp; 3162 file_t *fp; 3163 int error; 3164 3165 /* fd_getvnode() will use the descriptor for us */ 3166 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3167 return (error); 3168 vp = fp->f_data; 3169 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3170 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3171 VOP_UNLOCK(vp); 3172 fd_putfile(SCARG(uap, fd)); 3173 return (error); 3174 } 3175 3176 /* 3177 * Sync a range of file data. API modeled after that found in AIX. 3178 * 3179 * FDATASYNC indicates that we need only save enough metadata to be able 3180 * to re-read the written data. Note we duplicate AIX's requirement that 3181 * the file be open for writing. 3182 */ 3183 /* ARGSUSED */ 3184 int 3185 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3186 { 3187 /* { 3188 syscallarg(int) fd; 3189 syscallarg(int) flags; 3190 syscallarg(off_t) start; 3191 syscallarg(off_t) length; 3192 } */ 3193 struct vnode *vp; 3194 file_t *fp; 3195 int flags, nflags; 3196 off_t s, e, len; 3197 int error; 3198 3199 /* fd_getvnode() will use the descriptor for us */ 3200 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3201 return (error); 3202 3203 if ((fp->f_flag & FWRITE) == 0) { 3204 error = EBADF; 3205 goto out; 3206 } 3207 3208 flags = SCARG(uap, flags); 3209 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3210 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3211 error = EINVAL; 3212 goto out; 3213 } 3214 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3215 if (flags & FDATASYNC) 3216 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3217 else 3218 nflags = FSYNC_WAIT; 3219 if (flags & FDISKSYNC) 3220 nflags |= FSYNC_CACHE; 3221 3222 len = SCARG(uap, length); 3223 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3224 if (len) { 3225 s = SCARG(uap, start); 3226 e = s + len; 3227 if (e < s) { 3228 error = EINVAL; 3229 goto out; 3230 } 3231 } else { 3232 e = 0; 3233 s = 0; 3234 } 3235 3236 vp = fp->f_data; 3237 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3238 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3239 VOP_UNLOCK(vp); 3240 out: 3241 fd_putfile(SCARG(uap, fd)); 3242 return (error); 3243 } 3244 3245 /* 3246 * Sync the data of an open file. 3247 */ 3248 /* ARGSUSED */ 3249 int 3250 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3251 { 3252 /* { 3253 syscallarg(int) fd; 3254 } */ 3255 struct vnode *vp; 3256 file_t *fp; 3257 int error; 3258 3259 /* fd_getvnode() will use the descriptor for us */ 3260 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3261 return (error); 3262 if ((fp->f_flag & FWRITE) == 0) { 3263 fd_putfile(SCARG(uap, fd)); 3264 return (EBADF); 3265 } 3266 vp = fp->f_data; 3267 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3268 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3269 VOP_UNLOCK(vp); 3270 fd_putfile(SCARG(uap, fd)); 3271 return (error); 3272 } 3273 3274 /* 3275 * Rename files, (standard) BSD semantics frontend. 3276 */ 3277 /* ARGSUSED */ 3278 int 3279 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3280 { 3281 /* { 3282 syscallarg(const char *) from; 3283 syscallarg(const char *) to; 3284 } */ 3285 3286 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3287 } 3288 3289 /* 3290 * Rename files, POSIX semantics frontend. 3291 */ 3292 /* ARGSUSED */ 3293 int 3294 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3295 { 3296 /* { 3297 syscallarg(const char *) from; 3298 syscallarg(const char *) to; 3299 } */ 3300 3301 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3302 } 3303 3304 /* 3305 * Rename files. Source and destination must either both be directories, 3306 * or both not be directories. If target is a directory, it must be empty. 3307 * If `from' and `to' refer to the same object, the value of the `retain' 3308 * argument is used to determine whether `from' will be 3309 * 3310 * (retain == 0) deleted unless `from' and `to' refer to the same 3311 * object in the file system's name space (BSD). 3312 * (retain == 1) always retained (POSIX). 3313 */ 3314 int 3315 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3316 { 3317 struct vnode *tvp, *fvp, *tdvp; 3318 struct nameidata fromnd, tond; 3319 struct mount *fs; 3320 struct lwp *l = curlwp; 3321 struct proc *p; 3322 uint32_t saveflag; 3323 int error; 3324 3325 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT | INRENAME, 3326 seg, from); 3327 if ((error = namei(&fromnd)) != 0) 3328 return (error); 3329 if (fromnd.ni_dvp != fromnd.ni_vp) 3330 VOP_UNLOCK(fromnd.ni_dvp); 3331 fvp = fromnd.ni_vp; 3332 3333 fs = fvp->v_mount; 3334 error = VFS_RENAMELOCK_ENTER(fs); 3335 if (error) { 3336 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3337 vrele(fromnd.ni_dvp); 3338 vrele(fvp); 3339 goto out1; 3340 } 3341 3342 /* 3343 * close, partially, yet another race - ideally we should only 3344 * go as far as getting fromnd.ni_dvp before getting the per-fs 3345 * lock, and then continue to get fromnd.ni_vp, but we can't do 3346 * that with namei as it stands. 3347 * 3348 * This still won't prevent rmdir from nuking fromnd.ni_vp 3349 * under us. The real fix is to get the locks in the right 3350 * order and do the lookups in the right places, but that's a 3351 * major rototill. 3352 * 3353 * Preserve the SAVESTART in cn_flags, because who knows what 3354 * might happen if we don't. 3355 * 3356 * Note: this logic (as well as this whole function) is cloned 3357 * in nfs_serv.c. Proceed accordingly. 3358 */ 3359 vrele(fvp); 3360 if ((fromnd.ni_cnd.cn_namelen == 1 && 3361 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3362 (fromnd.ni_cnd.cn_namelen == 2 && 3363 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3364 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3365 error = EINVAL; 3366 VFS_RENAMELOCK_EXIT(fs); 3367 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3368 vrele(fromnd.ni_dvp); 3369 goto out1; 3370 } 3371 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3372 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3373 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3374 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3375 fromnd.ni_cnd.cn_flags |= saveflag; 3376 if (error) { 3377 VOP_UNLOCK(fromnd.ni_dvp); 3378 VFS_RENAMELOCK_EXIT(fs); 3379 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3380 vrele(fromnd.ni_dvp); 3381 goto out1; 3382 } 3383 VOP_UNLOCK(fromnd.ni_vp); 3384 if (fromnd.ni_dvp != fromnd.ni_vp) 3385 VOP_UNLOCK(fromnd.ni_dvp); 3386 fvp = fromnd.ni_vp; 3387 3388 NDINIT(&tond, RENAME, 3389 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3390 | INRENAME | (fvp->v_type == VDIR ? CREATEDIR : 0), 3391 seg, to); 3392 if ((error = namei(&tond)) != 0) { 3393 VFS_RENAMELOCK_EXIT(fs); 3394 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3395 vrele(fromnd.ni_dvp); 3396 vrele(fvp); 3397 goto out1; 3398 } 3399 tdvp = tond.ni_dvp; 3400 tvp = tond.ni_vp; 3401 3402 if (tvp != NULL) { 3403 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3404 error = ENOTDIR; 3405 goto out; 3406 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3407 error = EISDIR; 3408 goto out; 3409 } 3410 } 3411 3412 if (fvp == tdvp) 3413 error = EINVAL; 3414 3415 /* 3416 * Source and destination refer to the same object. 3417 */ 3418 if (fvp == tvp) { 3419 if (retain) 3420 error = -1; 3421 else if (fromnd.ni_dvp == tdvp && 3422 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3423 !memcmp(fromnd.ni_cnd.cn_nameptr, 3424 tond.ni_cnd.cn_nameptr, 3425 fromnd.ni_cnd.cn_namelen)) 3426 error = -1; 3427 } 3428 3429 #if NVERIEXEC > 0 3430 if (!error) { 3431 char *f1, *f2; 3432 size_t f1_len; 3433 size_t f2_len; 3434 3435 f1_len = fromnd.ni_cnd.cn_namelen + 1; 3436 f1 = kmem_alloc(f1_len, KM_SLEEP); 3437 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, f1_len); 3438 3439 f2_len = tond.ni_cnd.cn_namelen + 1; 3440 f2 = kmem_alloc(f2_len, KM_SLEEP); 3441 strlcpy(f2, tond.ni_cnd.cn_nameptr, f2_len); 3442 3443 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3444 3445 kmem_free(f1, f1_len); 3446 kmem_free(f2, f2_len); 3447 } 3448 #endif /* NVERIEXEC > 0 */ 3449 3450 out: 3451 p = l->l_proc; 3452 if (!error) { 3453 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3454 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3455 VFS_RENAMELOCK_EXIT(fs); 3456 } else { 3457 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3458 if (tdvp == tvp) 3459 vrele(tdvp); 3460 else 3461 vput(tdvp); 3462 if (tvp) 3463 vput(tvp); 3464 VFS_RENAMELOCK_EXIT(fs); 3465 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3466 vrele(fromnd.ni_dvp); 3467 vrele(fvp); 3468 } 3469 vrele(tond.ni_startdir); 3470 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3471 out1: 3472 if (fromnd.ni_startdir) 3473 vrele(fromnd.ni_startdir); 3474 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3475 return (error == -1 ? 0 : error); 3476 } 3477 3478 /* 3479 * Make a directory file. 3480 */ 3481 /* ARGSUSED */ 3482 int 3483 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3484 { 3485 /* { 3486 syscallarg(const char *) path; 3487 syscallarg(int) mode; 3488 } */ 3489 3490 return do_sys_mkdir(SCARG(uap, path), SCARG(uap, mode), UIO_USERSPACE); 3491 } 3492 3493 int 3494 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 3495 { 3496 struct proc *p = curlwp->l_proc; 3497 struct vnode *vp; 3498 struct vattr vattr; 3499 int error; 3500 struct nameidata nd; 3501 3502 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, 3503 seg, path); 3504 if ((error = namei(&nd)) != 0) 3505 return (error); 3506 vp = nd.ni_vp; 3507 if (vp != NULL) { 3508 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3509 if (nd.ni_dvp == vp) 3510 vrele(nd.ni_dvp); 3511 else 3512 vput(nd.ni_dvp); 3513 vrele(vp); 3514 return (EEXIST); 3515 } 3516 vattr_null(&vattr); 3517 vattr.va_type = VDIR; 3518 /* We will read cwdi->cwdi_cmask unlocked. */ 3519 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3520 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3521 if (!error) 3522 vput(nd.ni_vp); 3523 return (error); 3524 } 3525 3526 /* 3527 * Remove a directory file. 3528 */ 3529 /* ARGSUSED */ 3530 int 3531 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3532 { 3533 /* { 3534 syscallarg(const char *) path; 3535 } */ 3536 struct vnode *vp; 3537 int error; 3538 struct nameidata nd; 3539 3540 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3541 SCARG(uap, path)); 3542 if ((error = namei(&nd)) != 0) 3543 return (error); 3544 vp = nd.ni_vp; 3545 if (vp->v_type != VDIR) { 3546 error = ENOTDIR; 3547 goto out; 3548 } 3549 /* 3550 * No rmdir "." please. 3551 */ 3552 if (nd.ni_dvp == vp) { 3553 error = EINVAL; 3554 goto out; 3555 } 3556 /* 3557 * The root of a mounted filesystem cannot be deleted. 3558 */ 3559 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3560 error = EBUSY; 3561 goto out; 3562 } 3563 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3564 return (error); 3565 3566 out: 3567 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3568 if (nd.ni_dvp == vp) 3569 vrele(nd.ni_dvp); 3570 else 3571 vput(nd.ni_dvp); 3572 vput(vp); 3573 return (error); 3574 } 3575 3576 /* 3577 * Read a block of directory entries in a file system independent format. 3578 */ 3579 int 3580 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3581 { 3582 /* { 3583 syscallarg(int) fd; 3584 syscallarg(char *) buf; 3585 syscallarg(size_t) count; 3586 } */ 3587 file_t *fp; 3588 int error, done; 3589 3590 /* fd_getvnode() will use the descriptor for us */ 3591 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3592 return (error); 3593 if ((fp->f_flag & FREAD) == 0) { 3594 error = EBADF; 3595 goto out; 3596 } 3597 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3598 SCARG(uap, count), &done, l, 0, 0); 3599 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3600 *retval = done; 3601 out: 3602 fd_putfile(SCARG(uap, fd)); 3603 return (error); 3604 } 3605 3606 /* 3607 * Set the mode mask for creation of filesystem nodes. 3608 */ 3609 int 3610 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3611 { 3612 /* { 3613 syscallarg(mode_t) newmask; 3614 } */ 3615 struct proc *p = l->l_proc; 3616 struct cwdinfo *cwdi; 3617 3618 /* 3619 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3620 * important is that we serialize changes to the mask. The 3621 * rw_exit() will issue a write memory barrier on our behalf, 3622 * and force the changes out to other CPUs (as it must use an 3623 * atomic operation, draining the local CPU's store buffers). 3624 */ 3625 cwdi = p->p_cwdi; 3626 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3627 *retval = cwdi->cwdi_cmask; 3628 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3629 rw_exit(&cwdi->cwdi_lock); 3630 3631 return (0); 3632 } 3633 3634 int 3635 dorevoke(struct vnode *vp, kauth_cred_t cred) 3636 { 3637 struct vattr vattr; 3638 int error; 3639 3640 vn_lock(vp, LK_SHARED | LK_RETRY); 3641 error = VOP_GETATTR(vp, &vattr, cred); 3642 VOP_UNLOCK(vp); 3643 if (error != 0) 3644 return error; 3645 if (kauth_cred_geteuid(cred) == vattr.va_uid || 3646 (error = kauth_authorize_generic(cred, 3647 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3648 VOP_REVOKE(vp, REVOKEALL); 3649 return (error); 3650 } 3651 3652 /* 3653 * Void all references to file by ripping underlying filesystem 3654 * away from vnode. 3655 */ 3656 /* ARGSUSED */ 3657 int 3658 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3659 { 3660 /* { 3661 syscallarg(const char *) path; 3662 } */ 3663 struct vnode *vp; 3664 int error; 3665 3666 error = namei_simple_user(SCARG(uap, path), 3667 NSM_FOLLOW_TRYEMULROOT, &vp); 3668 if (error != 0) 3669 return (error); 3670 error = dorevoke(vp, l->l_cred); 3671 vrele(vp); 3672 return (error); 3673 } 3674
Indexes created Fri Oct 03 02:09:57 GMT 2025