vfs_syscalls.c revision 1.348.2.2
1 /* $NetBSD: vfs_syscalls.c,v 1.348.2.2 2008/11/01 21:22:28 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 17 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 18 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 19 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 20 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 21 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 22 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 23 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 24 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 25 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 26 * POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 /* 30 * Copyright (c) 1989, 1993 31 * The Regents of the University of California. All rights reserved. 32 * (c) UNIX System Laboratories, Inc. 33 * All or some portions of this file are derived from material licensed 34 * to the University of California by American Telephone and Telegraph 35 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 36 * the permission of UNIX System Laboratories, Inc. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 63 */ 64 65 #include <sys/cdefs.h> 66 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.348.2.2 2008/11/01 21:22:28 christos Exp $"); 67 68 #include "opt_compat_netbsd.h" 69 #include "opt_compat_43.h" 70 #include "opt_fileassoc.h" 71 #include "veriexec.h" 72 73 #include <sys/param.h> 74 #include <sys/systm.h> 75 #include <sys/namei.h> 76 #include <sys/filedesc.h> 77 #include <sys/kernel.h> 78 #include <sys/file.h> 79 #include <sys/stat.h> 80 #include <sys/vnode.h> 81 #include <sys/mount.h> 82 #include <sys/proc.h> 83 #include <sys/uio.h> 84 #include <sys/malloc.h> 85 #include <sys/kmem.h> 86 #include <sys/dirent.h> 87 #include <sys/sysctl.h> 88 #include <sys/syscallargs.h> 89 #include <sys/vfs_syscalls.h> 90 #include <sys/ktrace.h> 91 #ifdef FILEASSOC 92 #include <sys/fileassoc.h> 93 #endif /* FILEASSOC */ 94 #include <sys/verified_exec.h> 95 #include <sys/kauth.h> 96 #include <sys/atomic.h> 97 #include <sys/module.h> 98 99 #include <miscfs/genfs/genfs.h> 100 #include <miscfs/syncfs/syncfs.h> 101 #include <miscfs/specfs/specdev.h> 102 103 #ifdef COMPAT_30 104 #include "opt_nfsserver.h" 105 #include <nfs/rpcv2.h> 106 #endif 107 #include <nfs/nfsproto.h> 108 #ifdef COMPAT_30 109 #include <nfs/nfs.h> 110 #include <nfs/nfs_var.h> 111 #endif 112 113 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 114 115 static int change_dir(struct nameidata *, struct lwp *); 116 static int change_flags(struct vnode *, u_long, struct lwp *); 117 static int change_mode(struct vnode *, int, struct lwp *l); 118 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 119 120 void checkdirs(struct vnode *); 121 122 int dovfsusermount = 0; 123 124 /* 125 * Virtual File System System Calls 126 */ 127 128 /* 129 * Mount a file system. 130 */ 131 132 #if defined(COMPAT_09) || defined(COMPAT_43) 133 /* 134 * This table is used to maintain compatibility with 4.3BSD 135 * and NetBSD 0.9 mount syscalls. Note, the order is important! 136 * 137 * Do not modify this table. It should only contain filesystems 138 * supported by NetBSD 0.9 and 4.3BSD. 139 */ 140 const char * const mountcompatnames[] = { 141 NULL, /* 0 = MOUNT_NONE */ 142 MOUNT_FFS, /* 1 = MOUNT_UFS */ 143 MOUNT_NFS, /* 2 */ 144 MOUNT_MFS, /* 3 */ 145 MOUNT_MSDOS, /* 4 */ 146 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 147 MOUNT_FDESC, /* 6 */ 148 MOUNT_KERNFS, /* 7 */ 149 NULL, /* 8 = MOUNT_DEVFS */ 150 MOUNT_AFS, /* 9 */ 151 }; 152 const int nmountcompatnames = sizeof(mountcompatnames) / 153 sizeof(mountcompatnames[0]); 154 #endif /* COMPAT_09 || COMPAT_43 */ 155 156 static int 157 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 158 void *data, size_t *data_len) 159 { 160 struct mount *mp; 161 int error = 0, saved_flags; 162 163 mp = vp->v_mount; 164 saved_flags = mp->mnt_flag; 165 166 /* We can operate only on VV_ROOT nodes. */ 167 if ((vp->v_vflag & VV_ROOT) == 0) { 168 error = EINVAL; 169 goto out; 170 } 171 172 /* 173 * We only allow the filesystem to be reloaded if it 174 * is currently mounted read-only. Additionally, we 175 * prevent read-write to read-only downgrades. 176 */ 177 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 178 (mp->mnt_flag & MNT_RDONLY) == 0) { 179 error = EOPNOTSUPP; /* Needs translation */ 180 goto out; 181 } 182 183 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 184 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 185 if (error) 186 goto out; 187 188 if (vfs_busy(mp, NULL)) { 189 error = EPERM; 190 goto out; 191 } 192 193 mutex_enter(&mp->mnt_updating); 194 195 mp->mnt_flag &= ~MNT_OP_FLAGS; 196 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 197 198 /* 199 * Set the mount level flags. 200 */ 201 if (flags & MNT_RDONLY) 202 mp->mnt_flag |= MNT_RDONLY; 203 else if (mp->mnt_flag & MNT_RDONLY) 204 mp->mnt_iflag |= IMNT_WANTRDWR; 205 mp->mnt_flag &= 206 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 207 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 208 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 209 MNT_LOG); 210 mp->mnt_flag |= flags & 211 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 212 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 213 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 214 MNT_LOG | MNT_IGNORE); 215 216 error = VFS_MOUNT(mp, path, data, data_len); 217 218 #if defined(COMPAT_30) && defined(NFSSERVER) 219 if (error && data != NULL) { 220 int error2; 221 222 /* Update failed; let's try and see if it was an 223 * export request. */ 224 error2 = nfs_update_exports_30(mp, path, data, l); 225 226 /* Only update error code if the export request was 227 * understood but some problem occurred while 228 * processing it. */ 229 if (error2 != EJUSTRETURN) 230 error = error2; 231 } 232 #endif 233 if (mp->mnt_iflag & IMNT_WANTRDWR) 234 mp->mnt_flag &= ~MNT_RDONLY; 235 if (error) 236 mp->mnt_flag = saved_flags; 237 mp->mnt_flag &= ~MNT_OP_FLAGS; 238 mp->mnt_iflag &= ~IMNT_WANTRDWR; 239 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 240 if (mp->mnt_syncer == NULL) 241 error = vfs_allocate_syncvnode(mp); 242 } else { 243 if (mp->mnt_syncer != NULL) 244 vfs_deallocate_syncvnode(mp); 245 } 246 mutex_exit(&mp->mnt_updating); 247 vfs_unbusy(mp, false, NULL); 248 249 out: 250 return (error); 251 } 252 253 static int 254 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 255 { 256 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 257 int error; 258 259 /* Copy file-system type from userspace. */ 260 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 261 if (error) { 262 #if defined(COMPAT_09) || defined(COMPAT_43) 263 /* 264 * Historically, filesystem types were identified by numbers. 265 * If we get an integer for the filesystem type instead of a 266 * string, we check to see if it matches one of the historic 267 * filesystem types. 268 */ 269 u_long fsindex = (u_long)fstype; 270 if (fsindex >= nmountcompatnames || 271 mountcompatnames[fsindex] == NULL) 272 return ENODEV; 273 strlcpy(fstypename, mountcompatnames[fsindex], 274 sizeof(fstypename)); 275 #else 276 return error; 277 #endif 278 } 279 280 #ifdef COMPAT_10 281 /* Accept `ufs' as an alias for `ffs'. */ 282 if (strcmp(fstypename, "ufs") == 0) 283 fstypename[0] = 'f'; 284 #endif 285 286 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 287 return 0; 288 289 /* If we can autoload a vfs module, try again */ 290 mutex_enter(&module_lock); 291 (void)module_autoload(fstype, MODULE_CLASS_VFS); 292 mutex_exit(&module_lock); 293 294 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 295 return 0; 296 297 return ENODEV; 298 } 299 300 static int 301 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 302 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 303 { 304 struct mount *mp; 305 struct vnode *vp = *vpp; 306 struct vattr va; 307 int error; 308 309 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 310 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 311 if (error) 312 return error; 313 314 /* Can't make a non-dir a mount-point (from here anyway). */ 315 if (vp->v_type != VDIR) 316 return ENOTDIR; 317 318 /* 319 * If the user is not root, ensure that they own the directory 320 * onto which we are attempting to mount. 321 */ 322 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 323 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 324 (error = kauth_authorize_generic(l->l_cred, 325 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 326 return error; 327 } 328 329 if (flags & MNT_EXPORTED) 330 return EINVAL; 331 332 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 333 return error; 334 335 /* 336 * Check if a file-system is not already mounted on this vnode. 337 */ 338 if (vp->v_mountedhere != NULL) 339 return EBUSY; 340 341 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP); 342 if (mp == NULL) 343 return ENOMEM; 344 345 mp->mnt_op = vfsops; 346 mp->mnt_refcnt = 1; 347 348 TAILQ_INIT(&mp->mnt_vnodelist); 349 rw_init(&mp->mnt_unmounting); 350 mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE); 351 mutex_init(&mp->mnt_updating, MUTEX_DEFAULT, IPL_NONE); 352 error = vfs_busy(mp, NULL); 353 KASSERT(error == 0); 354 mutex_enter(&mp->mnt_updating); 355 356 mp->mnt_vnodecovered = vp; 357 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 358 mount_initspecific(mp); 359 360 /* 361 * The underlying file system may refuse the mount for 362 * various reasons. Allow the user to force it to happen. 363 * 364 * Set the mount level flags. 365 */ 366 mp->mnt_flag = flags & 367 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 368 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 369 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 370 MNT_LOG | MNT_IGNORE | MNT_RDONLY); 371 372 error = VFS_MOUNT(mp, path, data, data_len); 373 mp->mnt_flag &= ~MNT_OP_FLAGS; 374 375 /* 376 * Put the new filesystem on the mount list after root. 377 */ 378 cache_purge(vp); 379 if (error != 0) { 380 vp->v_mountedhere = NULL; 381 mutex_exit(&mp->mnt_updating); 382 vfs_unbusy(mp, false, NULL); 383 vfs_destroy(mp); 384 return error; 385 } 386 387 mp->mnt_iflag &= ~IMNT_WANTRDWR; 388 mutex_enter(&mountlist_lock); 389 vp->v_mountedhere = mp; 390 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 391 mutex_exit(&mountlist_lock); 392 vn_restorerecurse(vp, recurse); 393 VOP_UNLOCK(vp, 0); 394 checkdirs(vp); 395 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 396 error = vfs_allocate_syncvnode(mp); 397 /* Hold an additional reference to the mount across VFS_START(). */ 398 mutex_exit(&mp->mnt_updating); 399 vfs_unbusy(mp, true, NULL); 400 (void) VFS_STATVFS(mp, &mp->mnt_stat); 401 error = VFS_START(mp, 0); 402 if (error) 403 vrele(vp); 404 /* Drop reference held for VFS_START(). */ 405 vfs_destroy(mp); 406 *vpp = NULL; 407 return error; 408 } 409 410 static int 411 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 412 void *data, size_t *data_len) 413 { 414 struct mount *mp; 415 int error; 416 417 /* If MNT_GETARGS is specified, it should be the only flag. */ 418 if (flags & ~MNT_GETARGS) 419 return EINVAL; 420 421 mp = vp->v_mount; 422 423 /* XXX: probably some notion of "can see" here if we want isolation. */ 424 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 425 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 426 if (error) 427 return error; 428 429 if ((vp->v_vflag & VV_ROOT) == 0) 430 return EINVAL; 431 432 if (vfs_busy(mp, NULL)) 433 return EPERM; 434 435 mutex_enter(&mp->mnt_updating); 436 mp->mnt_flag &= ~MNT_OP_FLAGS; 437 mp->mnt_flag |= MNT_GETARGS; 438 error = VFS_MOUNT(mp, path, data, data_len); 439 mp->mnt_flag &= ~MNT_OP_FLAGS; 440 mutex_exit(&mp->mnt_updating); 441 442 vfs_unbusy(mp, false, NULL); 443 return (error); 444 } 445 446 #ifdef COMPAT_40 447 /* ARGSUSED */ 448 int 449 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 450 { 451 /* { 452 syscallarg(const char *) type; 453 syscallarg(const char *) path; 454 syscallarg(int) flags; 455 syscallarg(void *) data; 456 } */ 457 register_t dummy; 458 459 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 460 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 461 } 462 #endif 463 464 int 465 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 466 { 467 /* { 468 syscallarg(const char *) type; 469 syscallarg(const char *) path; 470 syscallarg(int) flags; 471 syscallarg(void *) data; 472 syscallarg(size_t) data_len; 473 } */ 474 475 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 476 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 477 SCARG(uap, data_len), retval); 478 } 479 480 int 481 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 482 const char *path, int flags, void *data, enum uio_seg data_seg, 483 size_t data_len, register_t *retval) 484 { 485 struct vnode *vp; 486 struct nameidata nd; 487 void *data_buf = data; 488 u_int recurse; 489 int error; 490 491 /* 492 * Get vnode to be covered 493 */ 494 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 495 if ((error = namei(&nd)) != 0) 496 return (error); 497 vp = nd.ni_vp; 498 499 /* 500 * A lookup in VFS_MOUNT might result in an attempt to 501 * lock this vnode again, so make the lock recursive. 502 */ 503 if (vfsops == NULL) { 504 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 505 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 506 recurse = vn_setrecurse(vp); 507 vfsops = vp->v_mount->mnt_op; 508 } else { 509 /* 'type' is userspace */ 510 error = mount_get_vfsops(type, &vfsops); 511 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 512 recurse = vn_setrecurse(vp); 513 if (error != 0) 514 goto done; 515 } 516 } else { 517 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 518 recurse = vn_setrecurse(vp); 519 } 520 521 if (data != NULL && data_seg == UIO_USERSPACE) { 522 if (data_len == 0) { 523 /* No length supplied, use default for filesystem */ 524 data_len = vfsops->vfs_min_mount_data; 525 if (data_len > VFS_MAX_MOUNT_DATA) { 526 /* maybe a force loaded old LKM */ 527 error = EINVAL; 528 goto done; 529 } 530 #ifdef COMPAT_30 531 /* Hopefully a longer buffer won't make copyin() fail */ 532 if (flags & MNT_UPDATE 533 && data_len < sizeof (struct mnt_export_args30)) 534 data_len = sizeof (struct mnt_export_args30); 535 #endif 536 } 537 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 538 539 /* NFS needs the buffer even for mnt_getargs .... */ 540 error = copyin(data, data_buf, data_len); 541 if (error != 0) 542 goto done; 543 } 544 545 if (flags & MNT_GETARGS) { 546 if (data_len == 0) { 547 error = EINVAL; 548 goto done; 549 } 550 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 551 if (error != 0) 552 goto done; 553 if (data_seg == UIO_USERSPACE) 554 error = copyout(data_buf, data, data_len); 555 *retval = data_len; 556 } else if (flags & MNT_UPDATE) { 557 error = mount_update(l, vp, path, flags, data_buf, &data_len); 558 } else { 559 /* Locking is handled internally in mount_domount(). */ 560 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 561 &data_len, recurse); 562 } 563 564 done: 565 if (vp != NULL) { 566 vn_restorerecurse(vp, recurse); 567 vput(vp); 568 } 569 if (data_buf != data) 570 free(data_buf, M_TEMP); 571 return (error); 572 } 573 574 /* 575 * Scan all active processes to see if any of them have a current 576 * or root directory onto which the new filesystem has just been 577 * mounted. If so, replace them with the new mount point. 578 */ 579 void 580 checkdirs(struct vnode *olddp) 581 { 582 struct cwdinfo *cwdi; 583 struct vnode *newdp, *rele1, *rele2; 584 struct proc *p; 585 bool retry; 586 587 if (olddp->v_usecount == 1) 588 return; 589 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 590 panic("mount: lost mount"); 591 592 do { 593 retry = false; 594 mutex_enter(proc_lock); 595 PROCLIST_FOREACH(p, &allproc) { 596 if ((p->p_flag & PK_MARKER) != 0) 597 continue; 598 if ((cwdi = p->p_cwdi) == NULL) 599 continue; 600 /* 601 * Can't change to the old directory any more, 602 * so even if we see a stale value it's not a 603 * problem. 604 */ 605 if (cwdi->cwdi_cdir != olddp && 606 cwdi->cwdi_rdir != olddp) 607 continue; 608 retry = true; 609 rele1 = NULL; 610 rele2 = NULL; 611 atomic_inc_uint(&cwdi->cwdi_refcnt); 612 mutex_exit(proc_lock); 613 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 614 if (cwdi->cwdi_cdir == olddp) { 615 rele1 = cwdi->cwdi_cdir; 616 VREF(newdp); 617 cwdi->cwdi_cdir = newdp; 618 } 619 if (cwdi->cwdi_rdir == olddp) { 620 rele2 = cwdi->cwdi_rdir; 621 VREF(newdp); 622 cwdi->cwdi_rdir = newdp; 623 } 624 rw_exit(&cwdi->cwdi_lock); 625 cwdfree(cwdi); 626 if (rele1 != NULL) 627 vrele(rele1); 628 if (rele2 != NULL) 629 vrele(rele2); 630 mutex_enter(proc_lock); 631 break; 632 } 633 mutex_exit(proc_lock); 634 } while (retry); 635 636 if (rootvnode == olddp) { 637 vrele(rootvnode); 638 VREF(newdp); 639 rootvnode = newdp; 640 } 641 vput(newdp); 642 } 643 644 /* 645 * Unmount a file system. 646 * 647 * Note: unmount takes a path to the vnode mounted on as argument, 648 * not special file (as before). 649 */ 650 /* ARGSUSED */ 651 int 652 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 653 { 654 /* { 655 syscallarg(const char *) path; 656 syscallarg(int) flags; 657 } */ 658 struct vnode *vp; 659 struct mount *mp; 660 int error; 661 struct nameidata nd; 662 663 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 664 SCARG(uap, path)); 665 if ((error = namei(&nd)) != 0) 666 return (error); 667 vp = nd.ni_vp; 668 mp = vp->v_mount; 669 atomic_inc_uint(&mp->mnt_refcnt); 670 VOP_UNLOCK(vp, 0); 671 672 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 673 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 674 if (error) { 675 vrele(vp); 676 vfs_destroy(mp); 677 return (error); 678 } 679 680 /* 681 * Don't allow unmounting the root file system. 682 */ 683 if (mp->mnt_flag & MNT_ROOTFS) { 684 vrele(vp); 685 vfs_destroy(mp); 686 return (EINVAL); 687 } 688 689 /* 690 * Must be the root of the filesystem 691 */ 692 if ((vp->v_vflag & VV_ROOT) == 0) { 693 vrele(vp); 694 vfs_destroy(mp); 695 return (EINVAL); 696 } 697 698 vrele(vp); 699 error = dounmount(mp, SCARG(uap, flags), l); 700 vfs_destroy(mp); 701 return error; 702 } 703 704 /* 705 * Do the actual file system unmount. File system is assumed to have 706 * been locked by the caller. 707 * 708 * => Caller hold reference to the mount, explicitly for dounmount(). 709 */ 710 int 711 dounmount(struct mount *mp, int flags, struct lwp *l) 712 { 713 struct vnode *coveredvp; 714 int error; 715 int async; 716 int used_syncer; 717 718 #if NVERIEXEC > 0 719 error = veriexec_unmountchk(mp); 720 if (error) 721 return (error); 722 #endif /* NVERIEXEC > 0 */ 723 724 /* 725 * XXX Freeze syncer. Must do this before locking the 726 * mount point. See dounmount() for details. 727 */ 728 mutex_enter(&syncer_mutex); 729 rw_enter(&mp->mnt_unmounting, RW_WRITER); 730 if ((mp->mnt_iflag & IMNT_GONE) != 0) { 731 rw_exit(&mp->mnt_unmounting); 732 mutex_exit(&syncer_mutex); 733 return ENOENT; 734 } 735 736 used_syncer = (mp->mnt_syncer != NULL); 737 738 /* 739 * XXX Syncer must be frozen when we get here. This should really 740 * be done on a per-mountpoint basis, but especially the softdep 741 * code possibly called from the syncer doesn't exactly work on a 742 * per-mountpoint basis, so the softdep code would become a maze 743 * of vfs_busy() calls. 744 * 745 * The caller of dounmount() must acquire syncer_mutex because 746 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 747 * order, and we must preserve that order to avoid deadlock. 748 * 749 * So, if the file system did not use the syncer, now is 750 * the time to release the syncer_mutex. 751 */ 752 if (used_syncer == 0) 753 mutex_exit(&syncer_mutex); 754 755 mp->mnt_iflag |= IMNT_UNMOUNT; 756 async = mp->mnt_flag & MNT_ASYNC; 757 mp->mnt_flag &= ~MNT_ASYNC; 758 cache_purgevfs(mp); /* remove cache entries for this file sys */ 759 if (mp->mnt_syncer != NULL) 760 vfs_deallocate_syncvnode(mp); 761 error = 0; 762 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 763 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 764 } 765 vfs_scrubvnlist(mp); 766 if (error == 0 || (flags & MNT_FORCE)) 767 error = VFS_UNMOUNT(mp, flags); 768 if (error) { 769 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 770 (void) vfs_allocate_syncvnode(mp); 771 mp->mnt_iflag &= ~IMNT_UNMOUNT; 772 mp->mnt_flag |= async; 773 rw_exit(&mp->mnt_unmounting); 774 if (used_syncer) 775 mutex_exit(&syncer_mutex); 776 return (error); 777 } 778 vfs_scrubvnlist(mp); 779 mutex_enter(&mountlist_lock); 780 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 781 coveredvp->v_mountedhere = NULL; 782 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 783 mp->mnt_iflag |= IMNT_GONE; 784 mutex_exit(&mountlist_lock); 785 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 786 panic("unmount: dangling vnode"); 787 if (used_syncer) 788 mutex_exit(&syncer_mutex); 789 vfs_hooks_unmount(mp); 790 rw_exit(&mp->mnt_unmounting); 791 vfs_destroy(mp); /* reference from mount() */ 792 if (coveredvp != NULLVP) 793 vrele(coveredvp); 794 return (0); 795 } 796 797 /* 798 * Sync each mounted filesystem. 799 */ 800 #ifdef DEBUG 801 int syncprt = 0; 802 struct ctldebug debug0 = { "syncprt", &syncprt }; 803 #endif 804 805 /* ARGSUSED */ 806 int 807 sys_sync(struct lwp *l, const void *v, register_t *retval) 808 { 809 struct mount *mp, *nmp; 810 int asyncflag; 811 812 if (l == NULL) 813 l = &lwp0; 814 815 mutex_enter(&mountlist_lock); 816 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 817 mp = nmp) { 818 if (vfs_busy(mp, &nmp)) { 819 continue; 820 } 821 mutex_enter(&mp->mnt_updating); 822 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 823 asyncflag = mp->mnt_flag & MNT_ASYNC; 824 mp->mnt_flag &= ~MNT_ASYNC; 825 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 826 if (asyncflag) 827 mp->mnt_flag |= MNT_ASYNC; 828 } 829 mutex_exit(&mp->mnt_updating); 830 vfs_unbusy(mp, false, &nmp); 831 } 832 mutex_exit(&mountlist_lock); 833 #ifdef DEBUG 834 if (syncprt) 835 vfs_bufstats(); 836 #endif /* DEBUG */ 837 return (0); 838 } 839 840 /* 841 * Change filesystem quotas. 842 */ 843 /* ARGSUSED */ 844 int 845 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 846 { 847 /* { 848 syscallarg(const char *) path; 849 syscallarg(int) cmd; 850 syscallarg(int) uid; 851 syscallarg(void *) arg; 852 } */ 853 struct mount *mp; 854 int error; 855 struct nameidata nd; 856 857 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 858 SCARG(uap, path)); 859 if ((error = namei(&nd)) != 0) 860 return (error); 861 mp = nd.ni_vp->v_mount; 862 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 863 SCARG(uap, arg)); 864 vrele(nd.ni_vp); 865 return (error); 866 } 867 868 int 869 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 870 int root) 871 { 872 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 873 int error = 0; 874 875 /* 876 * If MNT_NOWAIT or MNT_LAZY is specified, do not 877 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 878 * overrides MNT_NOWAIT. 879 */ 880 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 881 (flags != MNT_WAIT && flags != 0)) { 882 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 883 goto done; 884 } 885 886 /* Get the filesystem stats now */ 887 memset(sp, 0, sizeof(*sp)); 888 if ((error = VFS_STATVFS(mp, sp)) != 0) { 889 return error; 890 } 891 892 if (cwdi->cwdi_rdir == NULL) 893 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 894 done: 895 if (cwdi->cwdi_rdir != NULL) { 896 size_t len; 897 char *bp; 898 char c; 899 char *path = PNBUF_GET(); 900 901 bp = path + MAXPATHLEN; 902 *--bp = '\0'; 903 rw_enter(&cwdi->cwdi_lock, RW_READER); 904 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 905 MAXPATHLEN / 2, 0, l); 906 rw_exit(&cwdi->cwdi_lock); 907 if (error) { 908 PNBUF_PUT(path); 909 return error; 910 } 911 len = strlen(bp); 912 /* 913 * for mount points that are below our root, we can see 914 * them, so we fix up the pathname and return them. The 915 * rest we cannot see, so we don't allow viewing the 916 * data. 917 */ 918 if (strncmp(bp, sp->f_mntonname, len) == 0 && 919 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 920 (void)strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 921 sizeof(sp->f_mntonname)); 922 if (sp->f_mntonname[0] == '\0') 923 (void)strlcpy(sp->f_mntonname, "/", 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 PNBUF_PUT(path); 933 } 934 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 935 return error; 936 } 937 938 /* 939 * Get filesystem statistics by path. 940 */ 941 int 942 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 943 { 944 struct mount *mp; 945 int error; 946 struct nameidata nd; 947 948 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 949 if ((error = namei(&nd)) != 0) 950 return error; 951 mp = nd.ni_vp->v_mount; 952 error = dostatvfs(mp, sb, l, flags, 1); 953 vrele(nd.ni_vp); 954 return error; 955 } 956 957 /* ARGSUSED */ 958 int 959 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 960 { 961 /* { 962 syscallarg(const char *) path; 963 syscallarg(struct statvfs *) buf; 964 syscallarg(int) flags; 965 } */ 966 struct statvfs *sb; 967 int error; 968 969 sb = STATVFSBUF_GET(); 970 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 971 if (error == 0) 972 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 973 STATVFSBUF_PUT(sb); 974 return error; 975 } 976 977 /* 978 * Get filesystem statistics by fd. 979 */ 980 int 981 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 982 { 983 file_t *fp; 984 struct mount *mp; 985 int error; 986 987 /* fd_getvnode() will use the descriptor for us */ 988 if ((error = fd_getvnode(fd, &fp)) != 0) 989 return (error); 990 mp = ((struct vnode *)fp->f_data)->v_mount; 991 error = dostatvfs(mp, sb, curlwp, flags, 1); 992 fd_putfile(fd); 993 return error; 994 } 995 996 /* ARGSUSED */ 997 int 998 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 999 { 1000 /* { 1001 syscallarg(int) fd; 1002 syscallarg(struct statvfs *) buf; 1003 syscallarg(int) flags; 1004 } */ 1005 struct statvfs *sb; 1006 int error; 1007 1008 sb = STATVFSBUF_GET(); 1009 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1010 if (error == 0) 1011 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1012 STATVFSBUF_PUT(sb); 1013 return error; 1014 } 1015 1016 1017 /* 1018 * Get statistics on all filesystems. 1019 */ 1020 int 1021 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1022 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1023 register_t *retval) 1024 { 1025 int root = 0; 1026 struct proc *p = l->l_proc; 1027 struct mount *mp, *nmp; 1028 struct statvfs *sb; 1029 size_t count, maxcount; 1030 int error = 0; 1031 1032 sb = STATVFSBUF_GET(); 1033 maxcount = bufsize / entry_sz; 1034 mutex_enter(&mountlist_lock); 1035 count = 0; 1036 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1037 mp = nmp) { 1038 if (vfs_busy(mp, &nmp)) { 1039 continue; 1040 } 1041 if (sfsp && count < maxcount) { 1042 error = dostatvfs(mp, sb, l, flags, 0); 1043 if (error) { 1044 vfs_unbusy(mp, false, &nmp); 1045 error = 0; 1046 continue; 1047 } 1048 error = copyfn(sb, sfsp, entry_sz); 1049 if (error) { 1050 vfs_unbusy(mp, false, NULL); 1051 goto out; 1052 } 1053 sfsp = (char *)sfsp + entry_sz; 1054 root |= strcmp(sb->f_mntonname, "/") == 0; 1055 } 1056 count++; 1057 vfs_unbusy(mp, false, &nmp); 1058 } 1059 mutex_exit(&mountlist_lock); 1060 1061 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1062 /* 1063 * fake a root entry 1064 */ 1065 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1066 sb, l, flags, 1); 1067 if (error != 0) 1068 goto out; 1069 if (sfsp) { 1070 error = copyfn(sb, sfsp, entry_sz); 1071 if (error != 0) 1072 goto out; 1073 } 1074 count++; 1075 } 1076 if (sfsp && count > maxcount) 1077 *retval = maxcount; 1078 else 1079 *retval = count; 1080 out: 1081 STATVFSBUF_PUT(sb); 1082 return error; 1083 } 1084 1085 int 1086 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1087 { 1088 /* { 1089 syscallarg(struct statvfs *) buf; 1090 syscallarg(size_t) bufsize; 1091 syscallarg(int) flags; 1092 } */ 1093 1094 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1095 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1096 } 1097 1098 /* 1099 * Change current working directory to a given file descriptor. 1100 */ 1101 /* ARGSUSED */ 1102 int 1103 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1104 { 1105 /* { 1106 syscallarg(int) fd; 1107 } */ 1108 struct proc *p = l->l_proc; 1109 struct cwdinfo *cwdi; 1110 struct vnode *vp, *tdp; 1111 struct mount *mp; 1112 file_t *fp; 1113 int error, fd; 1114 1115 /* fd_getvnode() will use the descriptor for us */ 1116 fd = SCARG(uap, fd); 1117 if ((error = fd_getvnode(fd, &fp)) != 0) 1118 return (error); 1119 vp = fp->f_data; 1120 1121 VREF(vp); 1122 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1123 if (vp->v_type != VDIR) 1124 error = ENOTDIR; 1125 else 1126 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1127 if (error) { 1128 vput(vp); 1129 goto out; 1130 } 1131 while ((mp = vp->v_mountedhere) != NULL) { 1132 error = vfs_busy(mp, NULL); 1133 vput(vp); 1134 if (error != 0) 1135 goto out; 1136 error = VFS_ROOT(mp, &tdp); 1137 vfs_unbusy(mp, false, NULL); 1138 if (error) 1139 goto out; 1140 vp = tdp; 1141 } 1142 VOP_UNLOCK(vp, 0); 1143 1144 /* 1145 * Disallow changing to a directory not under the process's 1146 * current root directory (if there is one). 1147 */ 1148 cwdi = p->p_cwdi; 1149 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1150 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1151 vrele(vp); 1152 error = EPERM; /* operation not permitted */ 1153 } else { 1154 vrele(cwdi->cwdi_cdir); 1155 cwdi->cwdi_cdir = vp; 1156 } 1157 rw_exit(&cwdi->cwdi_lock); 1158 1159 out: 1160 fd_putfile(fd); 1161 return (error); 1162 } 1163 1164 /* 1165 * Change this process's notion of the root directory to a given file 1166 * descriptor. 1167 */ 1168 int 1169 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1170 { 1171 struct proc *p = l->l_proc; 1172 struct cwdinfo *cwdi; 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(SCARG(uap, 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, 0); 1190 if (error) 1191 goto out; 1192 VREF(vp); 1193 1194 /* 1195 * Prevent escaping from chroot by putting the root under 1196 * the working directory. Silently chdir to / if we aren't 1197 * already there. 1198 */ 1199 cwdi = p->p_cwdi; 1200 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1201 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1202 /* 1203 * XXX would be more failsafe to change directory to a 1204 * deadfs node here instead 1205 */ 1206 vrele(cwdi->cwdi_cdir); 1207 VREF(vp); 1208 cwdi->cwdi_cdir = vp; 1209 } 1210 1211 if (cwdi->cwdi_rdir != NULL) 1212 vrele(cwdi->cwdi_rdir); 1213 cwdi->cwdi_rdir = vp; 1214 rw_exit(&cwdi->cwdi_lock); 1215 1216 out: 1217 fd_putfile(fd); 1218 return (error); 1219 } 1220 1221 /* 1222 * Change current working directory (``.''). 1223 */ 1224 /* ARGSUSED */ 1225 int 1226 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1227 { 1228 /* { 1229 syscallarg(const char *) path; 1230 } */ 1231 struct proc *p = l->l_proc; 1232 struct cwdinfo *cwdi; 1233 int error; 1234 struct nameidata nd; 1235 1236 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1237 SCARG(uap, path)); 1238 if ((error = change_dir(&nd, l)) != 0) 1239 return (error); 1240 cwdi = p->p_cwdi; 1241 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1242 vrele(cwdi->cwdi_cdir); 1243 cwdi->cwdi_cdir = nd.ni_vp; 1244 rw_exit(&cwdi->cwdi_lock); 1245 return (0); 1246 } 1247 1248 /* 1249 * Change notion of root (``/'') directory. 1250 */ 1251 /* ARGSUSED */ 1252 int 1253 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1254 { 1255 /* { 1256 syscallarg(const char *) path; 1257 } */ 1258 struct proc *p = l->l_proc; 1259 struct cwdinfo *cwdi; 1260 struct vnode *vp; 1261 int error; 1262 struct nameidata nd; 1263 1264 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1265 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1266 return (error); 1267 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1268 SCARG(uap, path)); 1269 if ((error = change_dir(&nd, l)) != 0) 1270 return (error); 1271 1272 cwdi = p->p_cwdi; 1273 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1274 if (cwdi->cwdi_rdir != NULL) 1275 vrele(cwdi->cwdi_rdir); 1276 vp = nd.ni_vp; 1277 cwdi->cwdi_rdir = vp; 1278 1279 /* 1280 * Prevent escaping from chroot by putting the root under 1281 * the working directory. Silently chdir to / if we aren't 1282 * already there. 1283 */ 1284 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1285 /* 1286 * XXX would be more failsafe to change directory to a 1287 * deadfs node here instead 1288 */ 1289 vrele(cwdi->cwdi_cdir); 1290 VREF(vp); 1291 cwdi->cwdi_cdir = vp; 1292 } 1293 rw_exit(&cwdi->cwdi_lock); 1294 1295 return (0); 1296 } 1297 1298 /* 1299 * Common routine for chroot and chdir. 1300 */ 1301 static int 1302 change_dir(struct nameidata *ndp, struct lwp *l) 1303 { 1304 struct vnode *vp; 1305 int error; 1306 1307 if ((error = namei(ndp)) != 0) 1308 return (error); 1309 vp = ndp->ni_vp; 1310 if (vp->v_type != VDIR) 1311 error = ENOTDIR; 1312 else 1313 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1314 1315 if (error) 1316 vput(vp); 1317 else 1318 VOP_UNLOCK(vp, 0); 1319 return (error); 1320 } 1321 1322 /* 1323 * Check permissions, allocate an open file structure, 1324 * and call the device open routine if any. 1325 */ 1326 int 1327 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1328 { 1329 /* { 1330 syscallarg(const char *) path; 1331 syscallarg(int) flags; 1332 syscallarg(int) mode; 1333 } */ 1334 struct proc *p = l->l_proc; 1335 struct cwdinfo *cwdi = p->p_cwdi; 1336 file_t *fp; 1337 struct vnode *vp; 1338 int flags, cmode; 1339 int type, indx, error; 1340 struct flock lf; 1341 struct nameidata nd; 1342 1343 flags = FFLAGS(SCARG(uap, flags)); 1344 if ((flags & (FREAD | FWRITE)) == 0) 1345 return (EINVAL); 1346 if ((error = fd_allocfile(&fp, &indx)) != 0) 1347 return (error); 1348 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1349 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1350 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1351 SCARG(uap, path)); 1352 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1353 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1354 fd_abort(p, fp, indx); 1355 if ((error == EDUPFD || error == EMOVEFD) && 1356 l->l_dupfd >= 0 && /* XXX from fdopen */ 1357 (error = 1358 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1359 *retval = indx; 1360 return (0); 1361 } 1362 if (error == ERESTART) 1363 error = EINTR; 1364 return (error); 1365 } 1366 1367 l->l_dupfd = 0; 1368 vp = nd.ni_vp; 1369 fp->f_flag = flags & FMASK; 1370 fp->f_type = DTYPE_VNODE; 1371 fp->f_ops = &vnops; 1372 fp->f_data = vp; 1373 if (flags & (O_EXLOCK | O_SHLOCK)) { 1374 lf.l_whence = SEEK_SET; 1375 lf.l_start = 0; 1376 lf.l_len = 0; 1377 if (flags & O_EXLOCK) 1378 lf.l_type = F_WRLCK; 1379 else 1380 lf.l_type = F_RDLCK; 1381 type = F_FLOCK; 1382 if ((flags & FNONBLOCK) == 0) 1383 type |= F_WAIT; 1384 VOP_UNLOCK(vp, 0); 1385 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1386 if (error) { 1387 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1388 fd_abort(p, fp, indx); 1389 return (error); 1390 } 1391 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1392 atomic_or_uint(&fp->f_flag, FHASLOCK); 1393 } 1394 VOP_UNLOCK(vp, 0); 1395 *retval = indx; 1396 fd_affix(p, fp, indx); 1397 return (0); 1398 } 1399 1400 static void 1401 vfs__fhfree(fhandle_t *fhp) 1402 { 1403 size_t fhsize; 1404 1405 if (fhp == NULL) { 1406 return; 1407 } 1408 fhsize = FHANDLE_SIZE(fhp); 1409 kmem_free(fhp, fhsize); 1410 } 1411 1412 /* 1413 * vfs_composefh: compose a filehandle. 1414 */ 1415 1416 int 1417 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1418 { 1419 struct mount *mp; 1420 struct fid *fidp; 1421 int error; 1422 size_t needfhsize; 1423 size_t fidsize; 1424 1425 mp = vp->v_mount; 1426 fidp = NULL; 1427 if (*fh_size < FHANDLE_SIZE_MIN) { 1428 fidsize = 0; 1429 } else { 1430 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1431 if (fhp != NULL) { 1432 memset(fhp, 0, *fh_size); 1433 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1434 fidp = &fhp->fh_fid; 1435 } 1436 } 1437 error = VFS_VPTOFH(vp, fidp, &fidsize); 1438 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1439 if (error == 0 && *fh_size < needfhsize) { 1440 error = E2BIG; 1441 } 1442 *fh_size = needfhsize; 1443 return error; 1444 } 1445 1446 int 1447 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1448 { 1449 struct mount *mp; 1450 fhandle_t *fhp; 1451 size_t fhsize; 1452 size_t fidsize; 1453 int error; 1454 1455 *fhpp = NULL; 1456 mp = vp->v_mount; 1457 fidsize = 0; 1458 error = VFS_VPTOFH(vp, NULL, &fidsize); 1459 KASSERT(error != 0); 1460 if (error != E2BIG) { 1461 goto out; 1462 } 1463 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1464 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1465 if (fhp == NULL) { 1466 error = ENOMEM; 1467 goto out; 1468 } 1469 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1470 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1471 if (error == 0) { 1472 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1473 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1474 *fhpp = fhp; 1475 } else { 1476 kmem_free(fhp, fhsize); 1477 } 1478 out: 1479 return error; 1480 } 1481 1482 void 1483 vfs_composefh_free(fhandle_t *fhp) 1484 { 1485 1486 vfs__fhfree(fhp); 1487 } 1488 1489 /* 1490 * vfs_fhtovp: lookup a vnode by a filehandle. 1491 */ 1492 1493 int 1494 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1495 { 1496 struct mount *mp; 1497 int error; 1498 1499 *vpp = NULL; 1500 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1501 if (mp == NULL) { 1502 error = ESTALE; 1503 goto out; 1504 } 1505 if (mp->mnt_op->vfs_fhtovp == NULL) { 1506 error = EOPNOTSUPP; 1507 goto out; 1508 } 1509 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1510 out: 1511 return error; 1512 } 1513 1514 /* 1515 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1516 * the needed size. 1517 */ 1518 1519 int 1520 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1521 { 1522 fhandle_t *fhp; 1523 int error; 1524 1525 *fhpp = NULL; 1526 if (fhsize > FHANDLE_SIZE_MAX) { 1527 return EINVAL; 1528 } 1529 if (fhsize < FHANDLE_SIZE_MIN) { 1530 return EINVAL; 1531 } 1532 again: 1533 fhp = kmem_alloc(fhsize, KM_SLEEP); 1534 if (fhp == NULL) { 1535 return ENOMEM; 1536 } 1537 error = copyin(ufhp, fhp, fhsize); 1538 if (error == 0) { 1539 /* XXX this check shouldn't be here */ 1540 if (FHANDLE_SIZE(fhp) == fhsize) { 1541 *fhpp = fhp; 1542 return 0; 1543 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1544 /* 1545 * a kludge for nfsv2 padded handles. 1546 */ 1547 size_t sz; 1548 1549 sz = FHANDLE_SIZE(fhp); 1550 kmem_free(fhp, fhsize); 1551 fhsize = sz; 1552 goto again; 1553 } else { 1554 /* 1555 * userland told us wrong size. 1556 */ 1557 error = EINVAL; 1558 } 1559 } 1560 kmem_free(fhp, fhsize); 1561 return error; 1562 } 1563 1564 void 1565 vfs_copyinfh_free(fhandle_t *fhp) 1566 { 1567 1568 vfs__fhfree(fhp); 1569 } 1570 1571 /* 1572 * Get file handle system call 1573 */ 1574 int 1575 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1576 { 1577 /* { 1578 syscallarg(char *) fname; 1579 syscallarg(fhandle_t *) fhp; 1580 syscallarg(size_t *) fh_size; 1581 } */ 1582 struct vnode *vp; 1583 fhandle_t *fh; 1584 int error; 1585 struct nameidata nd; 1586 size_t sz; 1587 size_t usz; 1588 1589 /* 1590 * Must be super user 1591 */ 1592 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1593 0, NULL, NULL, NULL); 1594 if (error) 1595 return (error); 1596 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1597 SCARG(uap, fname)); 1598 error = namei(&nd); 1599 if (error) 1600 return (error); 1601 vp = nd.ni_vp; 1602 error = vfs_composefh_alloc(vp, &fh); 1603 vput(vp); 1604 if (error != 0) { 1605 goto out; 1606 } 1607 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1608 if (error != 0) { 1609 goto out; 1610 } 1611 sz = FHANDLE_SIZE(fh); 1612 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1613 if (error != 0) { 1614 goto out; 1615 } 1616 if (usz >= sz) { 1617 error = copyout(fh, SCARG(uap, fhp), sz); 1618 } else { 1619 error = E2BIG; 1620 } 1621 out: 1622 vfs_composefh_free(fh); 1623 return (error); 1624 } 1625 1626 /* 1627 * Open a file given a file handle. 1628 * 1629 * Check permissions, allocate an open file structure, 1630 * and call the device open routine if any. 1631 */ 1632 1633 int 1634 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1635 register_t *retval) 1636 { 1637 file_t *fp; 1638 struct vnode *vp = NULL; 1639 kauth_cred_t cred = l->l_cred; 1640 file_t *nfp; 1641 int type, indx, error=0; 1642 struct flock lf; 1643 struct vattr va; 1644 fhandle_t *fh; 1645 int flags; 1646 proc_t *p; 1647 1648 p = curproc; 1649 1650 /* 1651 * Must be super user 1652 */ 1653 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1654 0, NULL, NULL, NULL))) 1655 return (error); 1656 1657 flags = FFLAGS(oflags); 1658 if ((flags & (FREAD | FWRITE)) == 0) 1659 return (EINVAL); 1660 if ((flags & O_CREAT)) 1661 return (EINVAL); 1662 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1663 return (error); 1664 fp = nfp; 1665 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1666 if (error != 0) { 1667 goto bad; 1668 } 1669 error = vfs_fhtovp(fh, &vp); 1670 if (error != 0) { 1671 goto bad; 1672 } 1673 1674 /* Now do an effective vn_open */ 1675 1676 if (vp->v_type == VSOCK) { 1677 error = EOPNOTSUPP; 1678 goto bad; 1679 } 1680 error = vn_openchk(vp, cred, flags); 1681 if (error != 0) 1682 goto bad; 1683 if (flags & O_TRUNC) { 1684 VOP_UNLOCK(vp, 0); /* XXX */ 1685 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1686 VATTR_NULL(&va); 1687 va.va_size = 0; 1688 error = VOP_SETATTR(vp, &va, cred); 1689 if (error) 1690 goto bad; 1691 } 1692 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1693 goto bad; 1694 if (flags & FWRITE) { 1695 mutex_enter(&vp->v_interlock); 1696 vp->v_writecount++; 1697 mutex_exit(&vp->v_interlock); 1698 } 1699 1700 /* done with modified vn_open, now finish what sys_open does. */ 1701 1702 fp->f_flag = flags & FMASK; 1703 fp->f_type = DTYPE_VNODE; 1704 fp->f_ops = &vnops; 1705 fp->f_data = vp; 1706 if (flags & (O_EXLOCK | O_SHLOCK)) { 1707 lf.l_whence = SEEK_SET; 1708 lf.l_start = 0; 1709 lf.l_len = 0; 1710 if (flags & O_EXLOCK) 1711 lf.l_type = F_WRLCK; 1712 else 1713 lf.l_type = F_RDLCK; 1714 type = F_FLOCK; 1715 if ((flags & FNONBLOCK) == 0) 1716 type |= F_WAIT; 1717 VOP_UNLOCK(vp, 0); 1718 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1719 if (error) { 1720 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1721 fd_abort(p, fp, indx); 1722 return (error); 1723 } 1724 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1725 atomic_or_uint(&fp->f_flag, FHASLOCK); 1726 } 1727 VOP_UNLOCK(vp, 0); 1728 *retval = indx; 1729 fd_affix(p, fp, indx); 1730 vfs_copyinfh_free(fh); 1731 return (0); 1732 1733 bad: 1734 fd_abort(p, fp, indx); 1735 if (vp != NULL) 1736 vput(vp); 1737 vfs_copyinfh_free(fh); 1738 return (error); 1739 } 1740 1741 int 1742 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1743 { 1744 /* { 1745 syscallarg(const void *) fhp; 1746 syscallarg(size_t) fh_size; 1747 syscallarg(int) flags; 1748 } */ 1749 1750 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1751 SCARG(uap, flags), retval); 1752 } 1753 1754 int 1755 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1756 { 1757 int error; 1758 fhandle_t *fh; 1759 struct vnode *vp; 1760 1761 /* 1762 * Must be super user 1763 */ 1764 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1765 0, NULL, NULL, NULL))) 1766 return (error); 1767 1768 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1769 if (error != 0) 1770 return error; 1771 1772 error = vfs_fhtovp(fh, &vp); 1773 vfs_copyinfh_free(fh); 1774 if (error != 0) 1775 return error; 1776 1777 error = vn_stat(vp, sb); 1778 vput(vp); 1779 return error; 1780 } 1781 1782 1783 /* ARGSUSED */ 1784 int 1785 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1786 { 1787 /* { 1788 syscallarg(const void *) fhp; 1789 syscallarg(size_t) fh_size; 1790 syscallarg(struct stat *) sb; 1791 } */ 1792 struct stat sb; 1793 int error; 1794 1795 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1796 if (error) 1797 return error; 1798 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1799 } 1800 1801 int 1802 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1803 int flags) 1804 { 1805 fhandle_t *fh; 1806 struct mount *mp; 1807 struct vnode *vp; 1808 int error; 1809 1810 /* 1811 * Must be super user 1812 */ 1813 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1814 0, NULL, NULL, NULL))) 1815 return error; 1816 1817 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1818 if (error != 0) 1819 return error; 1820 1821 error = vfs_fhtovp(fh, &vp); 1822 vfs_copyinfh_free(fh); 1823 if (error != 0) 1824 return error; 1825 1826 mp = vp->v_mount; 1827 error = dostatvfs(mp, sb, l, flags, 1); 1828 vput(vp); 1829 return error; 1830 } 1831 1832 /* ARGSUSED */ 1833 int 1834 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1835 { 1836 /* { 1837 syscallarg(const void *) fhp; 1838 syscallarg(size_t) fh_size; 1839 syscallarg(struct statvfs *) buf; 1840 syscallarg(int) flags; 1841 } */ 1842 struct statvfs *sb = STATVFSBUF_GET(); 1843 int error; 1844 1845 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1846 SCARG(uap, flags)); 1847 if (error == 0) 1848 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1849 STATVFSBUF_PUT(sb); 1850 return error; 1851 } 1852 1853 /* 1854 * Create a special file. 1855 */ 1856 /* ARGSUSED */ 1857 int 1858 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 1859 register_t *retval) 1860 { 1861 /* { 1862 syscallarg(const char *) path; 1863 syscallarg(mode_t) mode; 1864 syscallarg(dev_t) dev; 1865 } */ 1866 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 1867 SCARG(uap, dev), retval); 1868 } 1869 1870 int 1871 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 1872 register_t *retval) 1873 { 1874 struct proc *p = l->l_proc; 1875 struct vnode *vp; 1876 struct vattr vattr; 1877 int error, optype; 1878 struct nameidata nd; 1879 char *path; 1880 const char *cpath; 1881 enum uio_seg seg = UIO_USERSPACE; 1882 1883 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1884 0, NULL, NULL, NULL)) != 0) 1885 return (error); 1886 1887 optype = VOP_MKNOD_DESCOFFSET; 1888 1889 VERIEXEC_PATH_GET(pathname, seg, cpath, path); 1890 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1891 1892 if ((error = namei(&nd)) != 0) 1893 goto out; 1894 vp = nd.ni_vp; 1895 if (vp != NULL) 1896 error = EEXIST; 1897 else { 1898 VATTR_NULL(&vattr); 1899 /* We will read cwdi->cwdi_cmask unlocked. */ 1900 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1901 vattr.va_rdev = dev; 1902 1903 switch (mode & S_IFMT) { 1904 case S_IFMT: /* used by badsect to flag bad sectors */ 1905 vattr.va_type = VBAD; 1906 break; 1907 case S_IFCHR: 1908 vattr.va_type = VCHR; 1909 break; 1910 case S_IFBLK: 1911 vattr.va_type = VBLK; 1912 break; 1913 case S_IFWHT: 1914 optype = VOP_WHITEOUT_DESCOFFSET; 1915 break; 1916 case S_IFREG: 1917 #if NVERIEXEC > 0 1918 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1919 O_CREAT); 1920 #endif /* NVERIEXEC > 0 */ 1921 vattr.va_type = VREG; 1922 vattr.va_rdev = VNOVAL; 1923 optype = VOP_CREATE_DESCOFFSET; 1924 break; 1925 default: 1926 error = EINVAL; 1927 break; 1928 } 1929 } 1930 if (!error) { 1931 switch (optype) { 1932 case VOP_WHITEOUT_DESCOFFSET: 1933 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1934 if (error) 1935 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1936 vput(nd.ni_dvp); 1937 break; 1938 1939 case VOP_MKNOD_DESCOFFSET: 1940 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1941 &nd.ni_cnd, &vattr); 1942 if (error == 0) 1943 vput(nd.ni_vp); 1944 break; 1945 1946 case VOP_CREATE_DESCOFFSET: 1947 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1948 &nd.ni_cnd, &vattr); 1949 if (error == 0) 1950 vput(nd.ni_vp); 1951 break; 1952 } 1953 } else { 1954 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1955 if (nd.ni_dvp == vp) 1956 vrele(nd.ni_dvp); 1957 else 1958 vput(nd.ni_dvp); 1959 if (vp) 1960 vrele(vp); 1961 } 1962 out: 1963 VERIEXEC_PATH_PUT(path); 1964 return (error); 1965 } 1966 1967 /* 1968 * Create a named pipe. 1969 */ 1970 /* ARGSUSED */ 1971 int 1972 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1973 { 1974 /* { 1975 syscallarg(const char *) path; 1976 syscallarg(int) mode; 1977 } */ 1978 struct proc *p = l->l_proc; 1979 struct vattr vattr; 1980 int error; 1981 struct nameidata nd; 1982 1983 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1984 SCARG(uap, path)); 1985 if ((error = namei(&nd)) != 0) 1986 return (error); 1987 if (nd.ni_vp != NULL) { 1988 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1989 if (nd.ni_dvp == nd.ni_vp) 1990 vrele(nd.ni_dvp); 1991 else 1992 vput(nd.ni_dvp); 1993 vrele(nd.ni_vp); 1994 return (EEXIST); 1995 } 1996 VATTR_NULL(&vattr); 1997 vattr.va_type = VFIFO; 1998 /* We will read cwdi->cwdi_cmask unlocked. */ 1999 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2000 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2001 if (error == 0) 2002 vput(nd.ni_vp); 2003 return (error); 2004 } 2005 2006 /* 2007 * Make a hard file link. 2008 */ 2009 /* ARGSUSED */ 2010 int 2011 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2012 { 2013 /* { 2014 syscallarg(const char *) path; 2015 syscallarg(const char *) link; 2016 } */ 2017 struct vnode *vp; 2018 struct nameidata nd; 2019 int error; 2020 2021 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2022 SCARG(uap, path)); 2023 if ((error = namei(&nd)) != 0) 2024 return (error); 2025 vp = nd.ni_vp; 2026 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2027 SCARG(uap, link)); 2028 if ((error = namei(&nd)) != 0) 2029 goto out; 2030 if (nd.ni_vp) { 2031 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2032 if (nd.ni_dvp == nd.ni_vp) 2033 vrele(nd.ni_dvp); 2034 else 2035 vput(nd.ni_dvp); 2036 vrele(nd.ni_vp); 2037 error = EEXIST; 2038 goto out; 2039 } 2040 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2041 out: 2042 vrele(vp); 2043 return (error); 2044 } 2045 2046 /* 2047 * Make a symbolic link. 2048 */ 2049 /* ARGSUSED */ 2050 int 2051 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2052 { 2053 /* { 2054 syscallarg(const char *) path; 2055 syscallarg(const char *) link; 2056 } */ 2057 struct proc *p = l->l_proc; 2058 struct vattr vattr; 2059 char *path; 2060 int error; 2061 struct nameidata nd; 2062 2063 path = PNBUF_GET(); 2064 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2065 if (error) 2066 goto out; 2067 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2068 SCARG(uap, link)); 2069 if ((error = namei(&nd)) != 0) 2070 goto out; 2071 if (nd.ni_vp) { 2072 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2073 if (nd.ni_dvp == nd.ni_vp) 2074 vrele(nd.ni_dvp); 2075 else 2076 vput(nd.ni_dvp); 2077 vrele(nd.ni_vp); 2078 error = EEXIST; 2079 goto out; 2080 } 2081 VATTR_NULL(&vattr); 2082 vattr.va_type = VLNK; 2083 /* We will read cwdi->cwdi_cmask unlocked. */ 2084 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2085 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2086 if (error == 0) 2087 vput(nd.ni_vp); 2088 out: 2089 PNBUF_PUT(path); 2090 return (error); 2091 } 2092 2093 /* 2094 * Delete a whiteout from the filesystem. 2095 */ 2096 /* ARGSUSED */ 2097 int 2098 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2099 { 2100 /* { 2101 syscallarg(const char *) path; 2102 } */ 2103 int error; 2104 struct nameidata nd; 2105 2106 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2107 UIO_USERSPACE, SCARG(uap, path)); 2108 error = namei(&nd); 2109 if (error) 2110 return (error); 2111 2112 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2113 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2114 if (nd.ni_dvp == nd.ni_vp) 2115 vrele(nd.ni_dvp); 2116 else 2117 vput(nd.ni_dvp); 2118 if (nd.ni_vp) 2119 vrele(nd.ni_vp); 2120 return (EEXIST); 2121 } 2122 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2123 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2124 vput(nd.ni_dvp); 2125 return (error); 2126 } 2127 2128 /* 2129 * Delete a name from the filesystem. 2130 */ 2131 /* ARGSUSED */ 2132 int 2133 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2134 { 2135 /* { 2136 syscallarg(const char *) path; 2137 } */ 2138 2139 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2140 } 2141 2142 int 2143 do_sys_unlink(const char *arg, enum uio_seg seg) 2144 { 2145 struct vnode *vp; 2146 int error; 2147 struct nameidata nd; 2148 kauth_cred_t cred; 2149 char *path; 2150 const char *cpath; 2151 2152 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2153 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2154 2155 if ((error = namei(&nd)) != 0) 2156 goto out; 2157 vp = nd.ni_vp; 2158 2159 /* 2160 * The root of a mounted filesystem cannot be deleted. 2161 */ 2162 if (vp->v_vflag & VV_ROOT) { 2163 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2164 if (nd.ni_dvp == vp) 2165 vrele(nd.ni_dvp); 2166 else 2167 vput(nd.ni_dvp); 2168 vput(vp); 2169 error = EBUSY; 2170 goto out; 2171 } 2172 2173 #if NVERIEXEC > 0 2174 /* Handle remove requests for veriexec entries. */ 2175 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2176 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2177 if (nd.ni_dvp == vp) 2178 vrele(nd.ni_dvp); 2179 else 2180 vput(nd.ni_dvp); 2181 vput(vp); 2182 goto out; 2183 } 2184 #endif /* NVERIEXEC > 0 */ 2185 2186 cred = kauth_cred_get(); 2187 #ifdef FILEASSOC 2188 (void)fileassoc_file_delete(vp); 2189 #endif /* FILEASSOC */ 2190 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2191 out: 2192 VERIEXEC_PATH_PUT(path); 2193 return (error); 2194 } 2195 2196 /* 2197 * Reposition read/write file offset. 2198 */ 2199 int 2200 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2201 { 2202 /* { 2203 syscallarg(int) fd; 2204 syscallarg(int) pad; 2205 syscallarg(off_t) offset; 2206 syscallarg(int) whence; 2207 } */ 2208 kauth_cred_t cred = l->l_cred; 2209 file_t *fp; 2210 struct vnode *vp; 2211 struct vattr vattr; 2212 off_t newoff; 2213 int error, fd; 2214 2215 fd = SCARG(uap, fd); 2216 2217 if ((fp = fd_getfile(fd)) == NULL) 2218 return (EBADF); 2219 2220 vp = fp->f_data; 2221 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2222 error = ESPIPE; 2223 goto out; 2224 } 2225 2226 switch (SCARG(uap, whence)) { 2227 case SEEK_CUR: 2228 newoff = fp->f_offset + SCARG(uap, offset); 2229 break; 2230 case SEEK_END: 2231 error = VOP_GETATTR(vp, &vattr, cred); 2232 if (error) { 2233 goto out; 2234 } 2235 newoff = SCARG(uap, offset) + vattr.va_size; 2236 break; 2237 case SEEK_SET: 2238 newoff = SCARG(uap, offset); 2239 break; 2240 default: 2241 error = EINVAL; 2242 goto out; 2243 } 2244 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2245 *(off_t *)retval = fp->f_offset = newoff; 2246 } 2247 out: 2248 fd_putfile(fd); 2249 return (error); 2250 } 2251 2252 /* 2253 * Positional read system call. 2254 */ 2255 int 2256 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2257 { 2258 /* { 2259 syscallarg(int) fd; 2260 syscallarg(void *) buf; 2261 syscallarg(size_t) nbyte; 2262 syscallarg(off_t) offset; 2263 } */ 2264 file_t *fp; 2265 struct vnode *vp; 2266 off_t offset; 2267 int error, fd = SCARG(uap, fd); 2268 2269 if ((fp = fd_getfile(fd)) == NULL) 2270 return (EBADF); 2271 2272 if ((fp->f_flag & FREAD) == 0) { 2273 fd_putfile(fd); 2274 return (EBADF); 2275 } 2276 2277 vp = fp->f_data; 2278 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2279 error = ESPIPE; 2280 goto out; 2281 } 2282 2283 offset = SCARG(uap, offset); 2284 2285 /* 2286 * XXX This works because no file systems actually 2287 * XXX take any action on the seek operation. 2288 */ 2289 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2290 goto out; 2291 2292 /* dofileread() will unuse the descriptor for us */ 2293 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2294 &offset, 0, retval)); 2295 2296 out: 2297 fd_putfile(fd); 2298 return (error); 2299 } 2300 2301 /* 2302 * Positional scatter read system call. 2303 */ 2304 int 2305 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2306 { 2307 /* { 2308 syscallarg(int) fd; 2309 syscallarg(const struct iovec *) iovp; 2310 syscallarg(int) iovcnt; 2311 syscallarg(off_t) offset; 2312 } */ 2313 off_t offset = SCARG(uap, offset); 2314 2315 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2316 SCARG(uap, iovcnt), &offset, 0, retval); 2317 } 2318 2319 /* 2320 * Positional write system call. 2321 */ 2322 int 2323 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2324 { 2325 /* { 2326 syscallarg(int) fd; 2327 syscallarg(const void *) buf; 2328 syscallarg(size_t) nbyte; 2329 syscallarg(off_t) offset; 2330 } */ 2331 file_t *fp; 2332 struct vnode *vp; 2333 off_t offset; 2334 int error, fd = SCARG(uap, fd); 2335 2336 if ((fp = fd_getfile(fd)) == NULL) 2337 return (EBADF); 2338 2339 if ((fp->f_flag & FWRITE) == 0) { 2340 fd_putfile(fd); 2341 return (EBADF); 2342 } 2343 2344 vp = fp->f_data; 2345 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2346 error = ESPIPE; 2347 goto out; 2348 } 2349 2350 offset = SCARG(uap, offset); 2351 2352 /* 2353 * XXX This works because no file systems actually 2354 * XXX take any action on the seek operation. 2355 */ 2356 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2357 goto out; 2358 2359 /* dofilewrite() will unuse the descriptor for us */ 2360 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2361 &offset, 0, retval)); 2362 2363 out: 2364 fd_putfile(fd); 2365 return (error); 2366 } 2367 2368 /* 2369 * Positional gather write system call. 2370 */ 2371 int 2372 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2373 { 2374 /* { 2375 syscallarg(int) fd; 2376 syscallarg(const struct iovec *) iovp; 2377 syscallarg(int) iovcnt; 2378 syscallarg(off_t) offset; 2379 } */ 2380 off_t offset = SCARG(uap, offset); 2381 2382 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2383 SCARG(uap, iovcnt), &offset, 0, retval); 2384 } 2385 2386 /* 2387 * Check access permissions. 2388 */ 2389 int 2390 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2391 { 2392 /* { 2393 syscallarg(const char *) path; 2394 syscallarg(int) flags; 2395 } */ 2396 kauth_cred_t cred; 2397 struct vnode *vp; 2398 int error, flags; 2399 struct nameidata nd; 2400 2401 cred = kauth_cred_dup(l->l_cred); 2402 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2403 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2404 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2405 SCARG(uap, path)); 2406 /* Override default credentials */ 2407 nd.ni_cnd.cn_cred = cred; 2408 if ((error = namei(&nd)) != 0) 2409 goto out; 2410 vp = nd.ni_vp; 2411 2412 /* Flags == 0 means only check for existence. */ 2413 if (SCARG(uap, flags)) { 2414 flags = 0; 2415 if (SCARG(uap, flags) & R_OK) 2416 flags |= VREAD; 2417 if (SCARG(uap, flags) & W_OK) 2418 flags |= VWRITE; 2419 if (SCARG(uap, flags) & X_OK) 2420 flags |= VEXEC; 2421 2422 error = VOP_ACCESS(vp, flags, cred); 2423 if (!error && (flags & VWRITE)) 2424 error = vn_writechk(vp); 2425 } 2426 vput(vp); 2427 out: 2428 kauth_cred_free(cred); 2429 return (error); 2430 } 2431 2432 /* 2433 * Common code for all sys_stat functions, including compat versions. 2434 */ 2435 int 2436 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb) 2437 { 2438 int error; 2439 struct nameidata nd; 2440 2441 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2442 UIO_USERSPACE, path); 2443 error = namei(&nd); 2444 if (error != 0) 2445 return error; 2446 error = vn_stat(nd.ni_vp, sb); 2447 vput(nd.ni_vp); 2448 return error; 2449 } 2450 2451 /* 2452 * Get file status; this version follows links. 2453 */ 2454 /* ARGSUSED */ 2455 int 2456 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2457 { 2458 /* { 2459 syscallarg(const char *) path; 2460 syscallarg(struct stat *) ub; 2461 } */ 2462 struct stat sb; 2463 int error; 2464 2465 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2466 if (error) 2467 return error; 2468 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2469 } 2470 2471 /* 2472 * Get file status; this version does not follow links. 2473 */ 2474 /* ARGSUSED */ 2475 int 2476 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2477 { 2478 /* { 2479 syscallarg(const char *) path; 2480 syscallarg(struct stat *) ub; 2481 } */ 2482 struct stat sb; 2483 int error; 2484 2485 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2486 if (error) 2487 return error; 2488 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2489 } 2490 2491 /* 2492 * Get configurable pathname variables. 2493 */ 2494 /* ARGSUSED */ 2495 int 2496 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2497 { 2498 /* { 2499 syscallarg(const char *) path; 2500 syscallarg(int) name; 2501 } */ 2502 int error; 2503 struct nameidata nd; 2504 2505 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2506 SCARG(uap, path)); 2507 if ((error = namei(&nd)) != 0) 2508 return (error); 2509 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2510 vput(nd.ni_vp); 2511 return (error); 2512 } 2513 2514 /* 2515 * Return target name of a symbolic link. 2516 */ 2517 /* ARGSUSED */ 2518 int 2519 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2520 { 2521 /* { 2522 syscallarg(const char *) path; 2523 syscallarg(char *) buf; 2524 syscallarg(size_t) count; 2525 } */ 2526 struct vnode *vp; 2527 struct iovec aiov; 2528 struct uio auio; 2529 int error; 2530 struct nameidata nd; 2531 2532 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2533 SCARG(uap, path)); 2534 if ((error = namei(&nd)) != 0) 2535 return (error); 2536 vp = nd.ni_vp; 2537 if (vp->v_type != VLNK) 2538 error = EINVAL; 2539 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2540 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2541 aiov.iov_base = SCARG(uap, buf); 2542 aiov.iov_len = SCARG(uap, count); 2543 auio.uio_iov = &aiov; 2544 auio.uio_iovcnt = 1; 2545 auio.uio_offset = 0; 2546 auio.uio_rw = UIO_READ; 2547 KASSERT(l == curlwp); 2548 auio.uio_vmspace = l->l_proc->p_vmspace; 2549 auio.uio_resid = SCARG(uap, count); 2550 error = VOP_READLINK(vp, &auio, l->l_cred); 2551 } 2552 vput(vp); 2553 *retval = SCARG(uap, count) - auio.uio_resid; 2554 return (error); 2555 } 2556 2557 /* 2558 * Change flags of a file given a path name. 2559 */ 2560 /* ARGSUSED */ 2561 int 2562 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2563 { 2564 /* { 2565 syscallarg(const char *) path; 2566 syscallarg(u_long) flags; 2567 } */ 2568 struct vnode *vp; 2569 int error; 2570 struct nameidata nd; 2571 2572 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2573 SCARG(uap, path)); 2574 if ((error = namei(&nd)) != 0) 2575 return (error); 2576 vp = nd.ni_vp; 2577 error = change_flags(vp, SCARG(uap, flags), l); 2578 vput(vp); 2579 return (error); 2580 } 2581 2582 /* 2583 * Change flags of a file given a file descriptor. 2584 */ 2585 /* ARGSUSED */ 2586 int 2587 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2588 { 2589 /* { 2590 syscallarg(int) fd; 2591 syscallarg(u_long) flags; 2592 } */ 2593 struct vnode *vp; 2594 file_t *fp; 2595 int error; 2596 2597 /* fd_getvnode() will use the descriptor for us */ 2598 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2599 return (error); 2600 vp = fp->f_data; 2601 error = change_flags(vp, SCARG(uap, flags), l); 2602 VOP_UNLOCK(vp, 0); 2603 fd_putfile(SCARG(uap, fd)); 2604 return (error); 2605 } 2606 2607 /* 2608 * Change flags of a file given a path name; this version does 2609 * not follow links. 2610 */ 2611 int 2612 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2613 { 2614 /* { 2615 syscallarg(const char *) path; 2616 syscallarg(u_long) flags; 2617 } */ 2618 struct vnode *vp; 2619 int error; 2620 struct nameidata nd; 2621 2622 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2623 SCARG(uap, path)); 2624 if ((error = namei(&nd)) != 0) 2625 return (error); 2626 vp = nd.ni_vp; 2627 error = change_flags(vp, SCARG(uap, flags), l); 2628 vput(vp); 2629 return (error); 2630 } 2631 2632 /* 2633 * Common routine to change flags of a file. 2634 */ 2635 int 2636 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2637 { 2638 struct vattr vattr; 2639 int error; 2640 2641 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2642 /* 2643 * Non-superusers cannot change the flags on devices, even if they 2644 * own them. 2645 */ 2646 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2647 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2648 goto out; 2649 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2650 error = EINVAL; 2651 goto out; 2652 } 2653 } 2654 VATTR_NULL(&vattr); 2655 vattr.va_flags = flags; 2656 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2657 out: 2658 return (error); 2659 } 2660 2661 /* 2662 * Change mode of a file given path name; this version follows links. 2663 */ 2664 /* ARGSUSED */ 2665 int 2666 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2667 { 2668 /* { 2669 syscallarg(const char *) path; 2670 syscallarg(int) mode; 2671 } */ 2672 int error; 2673 struct nameidata nd; 2674 2675 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2676 SCARG(uap, path)); 2677 if ((error = namei(&nd)) != 0) 2678 return (error); 2679 2680 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2681 2682 vrele(nd.ni_vp); 2683 return (error); 2684 } 2685 2686 /* 2687 * Change mode of a file given a file descriptor. 2688 */ 2689 /* ARGSUSED */ 2690 int 2691 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2692 { 2693 /* { 2694 syscallarg(int) fd; 2695 syscallarg(int) mode; 2696 } */ 2697 file_t *fp; 2698 int error; 2699 2700 /* fd_getvnode() will use the descriptor for us */ 2701 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2702 return (error); 2703 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2704 fd_putfile(SCARG(uap, fd)); 2705 return (error); 2706 } 2707 2708 /* 2709 * Change mode of a file given path name; this version does not follow links. 2710 */ 2711 /* ARGSUSED */ 2712 int 2713 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2714 { 2715 /* { 2716 syscallarg(const char *) path; 2717 syscallarg(int) mode; 2718 } */ 2719 int error; 2720 struct nameidata nd; 2721 2722 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2723 SCARG(uap, path)); 2724 if ((error = namei(&nd)) != 0) 2725 return (error); 2726 2727 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2728 2729 vrele(nd.ni_vp); 2730 return (error); 2731 } 2732 2733 /* 2734 * Common routine to set mode given a vnode. 2735 */ 2736 static int 2737 change_mode(struct vnode *vp, int mode, struct lwp *l) 2738 { 2739 struct vattr vattr; 2740 int error; 2741 2742 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2743 VATTR_NULL(&vattr); 2744 vattr.va_mode = mode & ALLPERMS; 2745 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2746 VOP_UNLOCK(vp, 0); 2747 return (error); 2748 } 2749 2750 /* 2751 * Set ownership given a path name; this version follows links. 2752 */ 2753 /* ARGSUSED */ 2754 int 2755 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2756 { 2757 /* { 2758 syscallarg(const char *) path; 2759 syscallarg(uid_t) uid; 2760 syscallarg(gid_t) gid; 2761 } */ 2762 int error; 2763 struct nameidata nd; 2764 2765 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2766 SCARG(uap, path)); 2767 if ((error = namei(&nd)) != 0) 2768 return (error); 2769 2770 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2771 2772 vrele(nd.ni_vp); 2773 return (error); 2774 } 2775 2776 /* 2777 * Set ownership given a path name; this version follows links. 2778 * Provides POSIX semantics. 2779 */ 2780 /* ARGSUSED */ 2781 int 2782 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2783 { 2784 /* { 2785 syscallarg(const char *) path; 2786 syscallarg(uid_t) uid; 2787 syscallarg(gid_t) gid; 2788 } */ 2789 int error; 2790 struct nameidata nd; 2791 2792 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2793 SCARG(uap, path)); 2794 if ((error = namei(&nd)) != 0) 2795 return (error); 2796 2797 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2798 2799 vrele(nd.ni_vp); 2800 return (error); 2801 } 2802 2803 /* 2804 * Set ownership given a file descriptor. 2805 */ 2806 /* ARGSUSED */ 2807 int 2808 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2809 { 2810 /* { 2811 syscallarg(int) fd; 2812 syscallarg(uid_t) uid; 2813 syscallarg(gid_t) gid; 2814 } */ 2815 int error; 2816 file_t *fp; 2817 2818 /* fd_getvnode() will use the descriptor for us */ 2819 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2820 return (error); 2821 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2822 l, 0); 2823 fd_putfile(SCARG(uap, fd)); 2824 return (error); 2825 } 2826 2827 /* 2828 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2829 */ 2830 /* ARGSUSED */ 2831 int 2832 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2833 { 2834 /* { 2835 syscallarg(int) fd; 2836 syscallarg(uid_t) uid; 2837 syscallarg(gid_t) gid; 2838 } */ 2839 int error; 2840 file_t *fp; 2841 2842 /* fd_getvnode() will use the descriptor for us */ 2843 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2844 return (error); 2845 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2846 l, 1); 2847 fd_putfile(SCARG(uap, fd)); 2848 return (error); 2849 } 2850 2851 /* 2852 * Set ownership given a path name; this version does not follow links. 2853 */ 2854 /* ARGSUSED */ 2855 int 2856 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2857 { 2858 /* { 2859 syscallarg(const char *) path; 2860 syscallarg(uid_t) uid; 2861 syscallarg(gid_t) gid; 2862 } */ 2863 int error; 2864 struct nameidata nd; 2865 2866 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2867 SCARG(uap, path)); 2868 if ((error = namei(&nd)) != 0) 2869 return (error); 2870 2871 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2872 2873 vrele(nd.ni_vp); 2874 return (error); 2875 } 2876 2877 /* 2878 * Set ownership given a path name; this version does not follow links. 2879 * Provides POSIX/XPG semantics. 2880 */ 2881 /* ARGSUSED */ 2882 int 2883 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2884 { 2885 /* { 2886 syscallarg(const char *) path; 2887 syscallarg(uid_t) uid; 2888 syscallarg(gid_t) gid; 2889 } */ 2890 int error; 2891 struct nameidata nd; 2892 2893 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2894 SCARG(uap, path)); 2895 if ((error = namei(&nd)) != 0) 2896 return (error); 2897 2898 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2899 2900 vrele(nd.ni_vp); 2901 return (error); 2902 } 2903 2904 /* 2905 * Common routine to set ownership given a vnode. 2906 */ 2907 static int 2908 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2909 int posix_semantics) 2910 { 2911 struct vattr vattr; 2912 mode_t newmode; 2913 int error; 2914 2915 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2916 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2917 goto out; 2918 2919 #define CHANGED(x) ((int)(x) != -1) 2920 newmode = vattr.va_mode; 2921 if (posix_semantics) { 2922 /* 2923 * POSIX/XPG semantics: if the caller is not the super-user, 2924 * clear set-user-id and set-group-id bits. Both POSIX and 2925 * the XPG consider the behaviour for calls by the super-user 2926 * implementation-defined; we leave the set-user-id and set- 2927 * group-id settings intact in that case. 2928 */ 2929 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2930 NULL) != 0) 2931 newmode &= ~(S_ISUID | S_ISGID); 2932 } else { 2933 /* 2934 * NetBSD semantics: when changing owner and/or group, 2935 * clear the respective bit(s). 2936 */ 2937 if (CHANGED(uid)) 2938 newmode &= ~S_ISUID; 2939 if (CHANGED(gid)) 2940 newmode &= ~S_ISGID; 2941 } 2942 /* Update va_mode iff altered. */ 2943 if (vattr.va_mode == newmode) 2944 newmode = VNOVAL; 2945 2946 VATTR_NULL(&vattr); 2947 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2948 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2949 vattr.va_mode = newmode; 2950 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2951 #undef CHANGED 2952 2953 out: 2954 VOP_UNLOCK(vp, 0); 2955 return (error); 2956 } 2957 2958 /* 2959 * Set the access and modification times given a path name; this 2960 * version follows links. 2961 */ 2962 /* ARGSUSED */ 2963 int 2964 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 2965 register_t *retval) 2966 { 2967 /* { 2968 syscallarg(const char *) path; 2969 syscallarg(const struct timeval *) tptr; 2970 } */ 2971 2972 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2973 SCARG(uap, tptr), UIO_USERSPACE); 2974 } 2975 2976 /* 2977 * Set the access and modification times given a file descriptor. 2978 */ 2979 /* ARGSUSED */ 2980 int 2981 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 2982 register_t *retval) 2983 { 2984 /* { 2985 syscallarg(int) fd; 2986 syscallarg(const struct timeval *) tptr; 2987 } */ 2988 int error; 2989 file_t *fp; 2990 2991 /* fd_getvnode() will use the descriptor for us */ 2992 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2993 return (error); 2994 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2995 UIO_USERSPACE); 2996 fd_putfile(SCARG(uap, fd)); 2997 return (error); 2998 } 2999 3000 /* 3001 * Set the access and modification times given a path name; this 3002 * version does not follow links. 3003 */ 3004 int 3005 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3006 register_t *retval) 3007 { 3008 /* { 3009 syscallarg(const char *) path; 3010 syscallarg(const struct timeval *) tptr; 3011 } */ 3012 3013 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3014 SCARG(uap, tptr), UIO_USERSPACE); 3015 } 3016 3017 /* 3018 * Common routine to set access and modification times given a vnode. 3019 */ 3020 int 3021 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3022 const struct timeval *tptr, enum uio_seg seg) 3023 { 3024 struct vattr vattr; 3025 struct nameidata nd; 3026 int error; 3027 bool vanull, setbirthtime; 3028 struct timespec ts[2]; 3029 3030 if (tptr == NULL) { 3031 vanull = true; 3032 nanotime(&ts[0]); 3033 ts[1] = ts[0]; 3034 } else { 3035 struct timeval tv[2]; 3036 3037 vanull = false; 3038 if (seg != UIO_SYSSPACE) { 3039 error = copyin(tptr, &tv, sizeof (tv)); 3040 if (error != 0) 3041 return error; 3042 tptr = tv; 3043 } 3044 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3045 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3046 } 3047 3048 if (vp == NULL) { 3049 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3050 if ((error = namei(&nd)) != 0) 3051 return error; 3052 vp = nd.ni_vp; 3053 } else 3054 nd.ni_vp = NULL; 3055 3056 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3057 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3058 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3059 VATTR_NULL(&vattr); 3060 vattr.va_atime = ts[0]; 3061 vattr.va_mtime = ts[1]; 3062 if (setbirthtime) 3063 vattr.va_birthtime = ts[1]; 3064 if (vanull) 3065 vattr.va_flags |= VA_UTIMES_NULL; 3066 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3067 VOP_UNLOCK(vp, 0); 3068 3069 if (nd.ni_vp != NULL) 3070 vrele(nd.ni_vp); 3071 3072 return error; 3073 } 3074 3075 /* 3076 * Truncate a file given its path name. 3077 */ 3078 /* ARGSUSED */ 3079 int 3080 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3081 { 3082 /* { 3083 syscallarg(const char *) path; 3084 syscallarg(int) pad; 3085 syscallarg(off_t) length; 3086 } */ 3087 struct vnode *vp; 3088 struct vattr vattr; 3089 int error; 3090 struct nameidata nd; 3091 3092 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3093 SCARG(uap, path)); 3094 if ((error = namei(&nd)) != 0) 3095 return (error); 3096 vp = nd.ni_vp; 3097 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3098 if (vp->v_type == VDIR) 3099 error = EISDIR; 3100 else if ((error = vn_writechk(vp)) == 0 && 3101 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3102 VATTR_NULL(&vattr); 3103 vattr.va_size = SCARG(uap, length); 3104 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3105 } 3106 vput(vp); 3107 return (error); 3108 } 3109 3110 /* 3111 * Truncate a file given a file descriptor. 3112 */ 3113 /* ARGSUSED */ 3114 int 3115 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3116 { 3117 /* { 3118 syscallarg(int) fd; 3119 syscallarg(int) pad; 3120 syscallarg(off_t) length; 3121 } */ 3122 struct vattr vattr; 3123 struct vnode *vp; 3124 file_t *fp; 3125 int error; 3126 3127 /* fd_getvnode() will use the descriptor for us */ 3128 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3129 return (error); 3130 if ((fp->f_flag & FWRITE) == 0) { 3131 error = EINVAL; 3132 goto out; 3133 } 3134 vp = fp->f_data; 3135 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3136 if (vp->v_type == VDIR) 3137 error = EISDIR; 3138 else if ((error = vn_writechk(vp)) == 0) { 3139 VATTR_NULL(&vattr); 3140 vattr.va_size = SCARG(uap, length); 3141 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3142 } 3143 VOP_UNLOCK(vp, 0); 3144 out: 3145 fd_putfile(SCARG(uap, fd)); 3146 return (error); 3147 } 3148 3149 /* 3150 * Sync an open file. 3151 */ 3152 /* ARGSUSED */ 3153 int 3154 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3155 { 3156 /* { 3157 syscallarg(int) fd; 3158 } */ 3159 struct vnode *vp; 3160 file_t *fp; 3161 int error; 3162 3163 /* fd_getvnode() will use the descriptor for us */ 3164 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3165 return (error); 3166 vp = fp->f_data; 3167 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3168 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3169 if (error == 0 && bioopsp != NULL && 3170 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3171 (*bioopsp->io_fsync)(vp, 0); 3172 VOP_UNLOCK(vp, 0); 3173 fd_putfile(SCARG(uap, fd)); 3174 return (error); 3175 } 3176 3177 /* 3178 * Sync a range of file data. API modeled after that found in AIX. 3179 * 3180 * FDATASYNC indicates that we need only save enough metadata to be able 3181 * to re-read the written data. Note we duplicate AIX's requirement that 3182 * the file be open for writing. 3183 */ 3184 /* ARGSUSED */ 3185 int 3186 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3187 { 3188 /* { 3189 syscallarg(int) fd; 3190 syscallarg(int) flags; 3191 syscallarg(off_t) start; 3192 syscallarg(off_t) length; 3193 } */ 3194 struct vnode *vp; 3195 file_t *fp; 3196 int flags, nflags; 3197 off_t s, e, len; 3198 int error; 3199 3200 /* fd_getvnode() will use the descriptor for us */ 3201 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3202 return (error); 3203 3204 if ((fp->f_flag & FWRITE) == 0) { 3205 error = EBADF; 3206 goto out; 3207 } 3208 3209 flags = SCARG(uap, flags); 3210 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3211 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3212 error = EINVAL; 3213 goto out; 3214 } 3215 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3216 if (flags & FDATASYNC) 3217 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3218 else 3219 nflags = FSYNC_WAIT; 3220 if (flags & FDISKSYNC) 3221 nflags |= FSYNC_CACHE; 3222 3223 len = SCARG(uap, length); 3224 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3225 if (len) { 3226 s = SCARG(uap, start); 3227 e = s + len; 3228 if (e < s) { 3229 error = EINVAL; 3230 goto out; 3231 } 3232 } else { 3233 e = 0; 3234 s = 0; 3235 } 3236 3237 vp = fp->f_data; 3238 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3239 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3240 3241 if (error == 0 && bioopsp != NULL && 3242 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3243 (*bioopsp->io_fsync)(vp, nflags); 3244 3245 VOP_UNLOCK(vp, 0); 3246 out: 3247 fd_putfile(SCARG(uap, fd)); 3248 return (error); 3249 } 3250 3251 /* 3252 * Sync the data of an open file. 3253 */ 3254 /* ARGSUSED */ 3255 int 3256 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3257 { 3258 /* { 3259 syscallarg(int) fd; 3260 } */ 3261 struct vnode *vp; 3262 file_t *fp; 3263 int error; 3264 3265 /* fd_getvnode() will use the descriptor for us */ 3266 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3267 return (error); 3268 if ((fp->f_flag & FWRITE) == 0) { 3269 fd_putfile(SCARG(uap, fd)); 3270 return (EBADF); 3271 } 3272 vp = fp->f_data; 3273 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3274 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3275 VOP_UNLOCK(vp, 0); 3276 fd_putfile(SCARG(uap, fd)); 3277 return (error); 3278 } 3279 3280 /* 3281 * Rename files, (standard) BSD semantics frontend. 3282 */ 3283 /* ARGSUSED */ 3284 int 3285 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3286 { 3287 /* { 3288 syscallarg(const char *) from; 3289 syscallarg(const char *) to; 3290 } */ 3291 3292 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3293 } 3294 3295 /* 3296 * Rename files, POSIX semantics frontend. 3297 */ 3298 /* ARGSUSED */ 3299 int 3300 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3301 { 3302 /* { 3303 syscallarg(const char *) from; 3304 syscallarg(const char *) to; 3305 } */ 3306 3307 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3308 } 3309 3310 /* 3311 * Rename files. Source and destination must either both be directories, 3312 * or both not be directories. If target is a directory, it must be empty. 3313 * If `from' and `to' refer to the same object, the value of the `retain' 3314 * argument is used to determine whether `from' will be 3315 * 3316 * (retain == 0) deleted unless `from' and `to' refer to the same 3317 * object in the file system's name space (BSD). 3318 * (retain == 1) always retained (POSIX). 3319 */ 3320 int 3321 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3322 { 3323 struct vnode *tvp, *fvp, *tdvp; 3324 struct nameidata fromnd, tond; 3325 struct mount *fs; 3326 struct lwp *l = curlwp; 3327 struct proc *p; 3328 uint32_t saveflag; 3329 int error; 3330 3331 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3332 seg, from); 3333 if ((error = namei(&fromnd)) != 0) 3334 return (error); 3335 if (fromnd.ni_dvp != fromnd.ni_vp) 3336 VOP_UNLOCK(fromnd.ni_dvp, 0); 3337 fvp = fromnd.ni_vp; 3338 3339 fs = fvp->v_mount; 3340 error = VFS_RENAMELOCK_ENTER(fs); 3341 if (error) { 3342 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3343 vrele(fromnd.ni_dvp); 3344 vrele(fvp); 3345 goto out1; 3346 } 3347 3348 /* 3349 * close, partially, yet another race - ideally we should only 3350 * go as far as getting fromnd.ni_dvp before getting the per-fs 3351 * lock, and then continue to get fromnd.ni_vp, but we can't do 3352 * that with namei as it stands. 3353 * 3354 * This still won't prevent rmdir from nuking fromnd.ni_vp 3355 * under us. The real fix is to get the locks in the right 3356 * order and do the lookups in the right places, but that's a 3357 * major rototill. 3358 * 3359 * Preserve the SAVESTART in cn_flags, because who knows what 3360 * might happen if we don't. 3361 * 3362 * Note: this logic (as well as this whole function) is cloned 3363 * in nfs_serv.c. Proceed accordingly. 3364 */ 3365 vrele(fvp); 3366 if ((fromnd.ni_cnd.cn_namelen == 1 && 3367 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3368 (fromnd.ni_cnd.cn_namelen == 2 && 3369 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3370 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3371 error = EINVAL; 3372 VFS_RENAMELOCK_EXIT(fs); 3373 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3374 vrele(fromnd.ni_dvp); 3375 goto out1; 3376 } 3377 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3378 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3379 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3380 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3381 fromnd.ni_cnd.cn_flags |= saveflag; 3382 if (error) { 3383 VOP_UNLOCK(fromnd.ni_dvp, 0); 3384 VFS_RENAMELOCK_EXIT(fs); 3385 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3386 vrele(fromnd.ni_dvp); 3387 goto out1; 3388 } 3389 VOP_UNLOCK(fromnd.ni_vp, 0); 3390 if (fromnd.ni_dvp != fromnd.ni_vp) 3391 VOP_UNLOCK(fromnd.ni_dvp, 0); 3392 fvp = fromnd.ni_vp; 3393 3394 NDINIT(&tond, RENAME, 3395 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3396 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3397 seg, to); 3398 if ((error = namei(&tond)) != 0) { 3399 VFS_RENAMELOCK_EXIT(fs); 3400 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3401 vrele(fromnd.ni_dvp); 3402 vrele(fvp); 3403 goto out1; 3404 } 3405 tdvp = tond.ni_dvp; 3406 tvp = tond.ni_vp; 3407 3408 if (tvp != NULL) { 3409 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3410 error = ENOTDIR; 3411 goto out; 3412 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3413 error = EISDIR; 3414 goto out; 3415 } 3416 } 3417 3418 if (fvp == tdvp) 3419 error = EINVAL; 3420 3421 /* 3422 * Source and destination refer to the same object. 3423 */ 3424 if (fvp == tvp) { 3425 if (retain) 3426 error = -1; 3427 else if (fromnd.ni_dvp == tdvp && 3428 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3429 !memcmp(fromnd.ni_cnd.cn_nameptr, 3430 tond.ni_cnd.cn_nameptr, 3431 fromnd.ni_cnd.cn_namelen)) 3432 error = -1; 3433 } 3434 3435 #if NVERIEXEC > 0 3436 if (!error) { 3437 char *f1, *f2; 3438 3439 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3440 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3441 3442 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3443 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3444 3445 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3446 3447 free(f1, M_TEMP); 3448 free(f2, M_TEMP); 3449 } 3450 #endif /* NVERIEXEC > 0 */ 3451 3452 out: 3453 p = l->l_proc; 3454 if (!error) { 3455 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3456 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3457 VFS_RENAMELOCK_EXIT(fs); 3458 } else { 3459 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3460 if (tdvp == tvp) 3461 vrele(tdvp); 3462 else 3463 vput(tdvp); 3464 if (tvp) 3465 vput(tvp); 3466 VFS_RENAMELOCK_EXIT(fs); 3467 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3468 vrele(fromnd.ni_dvp); 3469 vrele(fvp); 3470 } 3471 vrele(tond.ni_startdir); 3472 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3473 out1: 3474 if (fromnd.ni_startdir) 3475 vrele(fromnd.ni_startdir); 3476 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3477 return (error == -1 ? 0 : error); 3478 } 3479 3480 /* 3481 * Make a directory file. 3482 */ 3483 /* ARGSUSED */ 3484 int 3485 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3486 { 3487 /* { 3488 syscallarg(const char *) path; 3489 syscallarg(int) mode; 3490 } */ 3491 struct proc *p = l->l_proc; 3492 struct vnode *vp; 3493 struct vattr vattr; 3494 int error; 3495 struct nameidata nd; 3496 3497 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3498 SCARG(uap, path)); 3499 if ((error = namei(&nd)) != 0) 3500 return (error); 3501 vp = nd.ni_vp; 3502 if (vp != NULL) { 3503 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3504 if (nd.ni_dvp == vp) 3505 vrele(nd.ni_dvp); 3506 else 3507 vput(nd.ni_dvp); 3508 vrele(vp); 3509 return (EEXIST); 3510 } 3511 VATTR_NULL(&vattr); 3512 vattr.va_type = VDIR; 3513 /* We will read cwdi->cwdi_cmask unlocked. */ 3514 vattr.va_mode = 3515 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3516 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3517 if (!error) 3518 vput(nd.ni_vp); 3519 return (error); 3520 } 3521 3522 /* 3523 * Remove a directory file. 3524 */ 3525 /* ARGSUSED */ 3526 int 3527 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3528 { 3529 /* { 3530 syscallarg(const char *) path; 3531 } */ 3532 struct vnode *vp; 3533 int error; 3534 struct nameidata nd; 3535 3536 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3537 SCARG(uap, path)); 3538 if ((error = namei(&nd)) != 0) 3539 return (error); 3540 vp = nd.ni_vp; 3541 if (vp->v_type != VDIR) { 3542 error = ENOTDIR; 3543 goto out; 3544 } 3545 /* 3546 * No rmdir "." please. 3547 */ 3548 if (nd.ni_dvp == vp) { 3549 error = EINVAL; 3550 goto out; 3551 } 3552 /* 3553 * The root of a mounted filesystem cannot be deleted. 3554 */ 3555 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3556 error = EBUSY; 3557 goto out; 3558 } 3559 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3560 return (error); 3561 3562 out: 3563 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3564 if (nd.ni_dvp == vp) 3565 vrele(nd.ni_dvp); 3566 else 3567 vput(nd.ni_dvp); 3568 vput(vp); 3569 return (error); 3570 } 3571 3572 /* 3573 * Read a block of directory entries in a file system independent format. 3574 */ 3575 int 3576 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3577 { 3578 /* { 3579 syscallarg(int) fd; 3580 syscallarg(char *) buf; 3581 syscallarg(size_t) count; 3582 } */ 3583 file_t *fp; 3584 int error, done; 3585 3586 /* fd_getvnode() will use the descriptor for us */ 3587 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3588 return (error); 3589 if ((fp->f_flag & FREAD) == 0) { 3590 error = EBADF; 3591 goto out; 3592 } 3593 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3594 SCARG(uap, count), &done, l, 0, 0); 3595 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3596 *retval = done; 3597 out: 3598 fd_putfile(SCARG(uap, fd)); 3599 return (error); 3600 } 3601 3602 /* 3603 * Set the mode mask for creation of filesystem nodes. 3604 */ 3605 int 3606 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3607 { 3608 /* { 3609 syscallarg(mode_t) newmask; 3610 } */ 3611 struct proc *p = l->l_proc; 3612 struct cwdinfo *cwdi; 3613 3614 /* 3615 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3616 * important is that we serialize changes to the mask. The 3617 * rw_exit() will issue a write memory barrier on our behalf, 3618 * and force the changes out to other CPUs (as it must use an 3619 * atomic operation, draining the local CPU's store buffers). 3620 */ 3621 cwdi = p->p_cwdi; 3622 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3623 *retval = cwdi->cwdi_cmask; 3624 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3625 rw_exit(&cwdi->cwdi_lock); 3626 3627 return (0); 3628 } 3629 3630 int 3631 dorevoke(struct vnode *vp, kauth_cred_t cred) 3632 { 3633 struct vattr vattr; 3634 int error; 3635 3636 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3637 return error; 3638 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3639 (error = kauth_authorize_generic(cred, 3640 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3641 VOP_REVOKE(vp, REVOKEALL); 3642 return (error); 3643 } 3644 3645 /* 3646 * Void all references to file by ripping underlying filesystem 3647 * away from vnode. 3648 */ 3649 /* ARGSUSED */ 3650 int 3651 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3652 { 3653 /* { 3654 syscallarg(const char *) path; 3655 } */ 3656 struct vnode *vp; 3657 int error; 3658 struct nameidata nd; 3659 3660 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3661 SCARG(uap, path)); 3662 if ((error = namei(&nd)) != 0) 3663 return (error); 3664 vp = nd.ni_vp; 3665 error = dorevoke(vp, l->l_cred); 3666 vrele(vp); 3667 return (error); 3668 } 3669
Indexes created Fri Oct 03 02:09:57 GMT 2025