vfs_syscalls.c revision 1.348.2.1
1 /* $NetBSD: vfs_syscalls.c,v 1.348.2.1 2008/03/29 20:47:01 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 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the NetBSD 18 * Foundation, Inc. and its contributors. 19 * 4. Neither the name of The NetBSD Foundation nor the names of its 20 * contributors may be used to endorse or promote products derived 21 * from this software without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 24 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 25 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 27 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 33 * POSSIBILITY OF SUCH DAMAGE. 34 */ 35 36 /* 37 * Copyright (c) 1989, 1993 38 * The Regents of the University of California. All rights reserved. 39 * (c) UNIX System Laboratories, Inc. 40 * All or some portions of this file are derived from material licensed 41 * to the University of California by American Telephone and Telegraph 42 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 43 * the permission of UNIX System Laboratories, Inc. 44 * 45 * Redistribution and use in source and binary forms, with or without 46 * modification, are permitted provided that the following conditions 47 * are met: 48 * 1. Redistributions of source code must retain the above copyright 49 * notice, this list of conditions and the following disclaimer. 50 * 2. Redistributions in binary form must reproduce the above copyright 51 * notice, this list of conditions and the following disclaimer in the 52 * documentation and/or other materials provided with the distribution. 53 * 3. Neither the name of the University nor the names of its contributors 54 * may be used to endorse or promote products derived from this software 55 * without specific prior written permission. 56 * 57 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 58 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 59 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 60 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 61 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 62 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 63 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 64 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 65 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 66 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 67 * SUCH DAMAGE. 68 * 69 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.348.2.1 2008/03/29 20:47:01 christos Exp $"); 74 75 #include "opt_compat_netbsd.h" 76 #include "opt_compat_43.h" 77 #include "opt_fileassoc.h" 78 #include "fss.h" 79 #include "veriexec.h" 80 81 #include <sys/param.h> 82 #include <sys/systm.h> 83 #include <sys/namei.h> 84 #include <sys/filedesc.h> 85 #include <sys/kernel.h> 86 #include <sys/file.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/proc.h> 91 #include <sys/uio.h> 92 #include <sys/malloc.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/ktrace.h> 99 #ifdef FILEASSOC 100 #include <sys/fileassoc.h> 101 #endif /* FILEASSOC */ 102 #include <sys/verified_exec.h> 103 #include <sys/kauth.h> 104 #include <sys/atomic.h> 105 106 #include <miscfs/genfs/genfs.h> 107 #include <miscfs/syncfs/syncfs.h> 108 #include <miscfs/specfs/specdev.h> 109 110 #ifdef COMPAT_30 111 #include "opt_nfsserver.h" 112 #include <nfs/rpcv2.h> 113 #endif 114 #include <nfs/nfsproto.h> 115 #ifdef COMPAT_30 116 #include <nfs/nfs.h> 117 #include <nfs/nfs_var.h> 118 #endif 119 120 #if NFSS > 0 121 #include <dev/fssvar.h> 122 #endif 123 124 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 125 126 static int change_dir(struct nameidata *, struct lwp *); 127 static int change_flags(struct vnode *, u_long, struct lwp *); 128 static int change_mode(struct vnode *, int, struct lwp *l); 129 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 130 131 void checkdirs(struct vnode *); 132 133 int dovfsusermount = 0; 134 135 /* 136 * Virtual File System System Calls 137 */ 138 139 /* 140 * Mount a file system. 141 */ 142 143 #if defined(COMPAT_09) || defined(COMPAT_43) 144 /* 145 * This table is used to maintain compatibility with 4.3BSD 146 * and NetBSD 0.9 mount syscalls. Note, the order is important! 147 * 148 * Do not modify this table. It should only contain filesystems 149 * supported by NetBSD 0.9 and 4.3BSD. 150 */ 151 const char * const mountcompatnames[] = { 152 NULL, /* 0 = MOUNT_NONE */ 153 MOUNT_FFS, /* 1 = MOUNT_UFS */ 154 MOUNT_NFS, /* 2 */ 155 MOUNT_MFS, /* 3 */ 156 MOUNT_MSDOS, /* 4 */ 157 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 158 MOUNT_FDESC, /* 6 */ 159 MOUNT_KERNFS, /* 7 */ 160 NULL, /* 8 = MOUNT_DEVFS */ 161 MOUNT_AFS, /* 9 */ 162 }; 163 const int nmountcompatnames = sizeof(mountcompatnames) / 164 sizeof(mountcompatnames[0]); 165 #endif /* COMPAT_09 || COMPAT_43 */ 166 167 static int 168 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 169 void *data, size_t *data_len) 170 { 171 struct mount *mp; 172 int error = 0, saved_flags; 173 174 mp = vp->v_mount; 175 saved_flags = mp->mnt_flag; 176 177 /* We can operate only on VV_ROOT nodes. */ 178 if ((vp->v_vflag & VV_ROOT) == 0) { 179 error = EINVAL; 180 goto out; 181 } 182 183 /* 184 * We only allow the filesystem to be reloaded if it 185 * is currently mounted read-only. 186 */ 187 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 188 error = EOPNOTSUPP; /* Needs translation */ 189 goto out; 190 } 191 192 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 193 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 194 if (error) 195 goto out; 196 197 if (vfs_trybusy(mp, RW_WRITER, 0)) { 198 error = EPERM; 199 goto out; 200 } 201 202 mp->mnt_flag &= ~MNT_OP_FLAGS; 203 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 204 205 /* 206 * Set the mount level flags. 207 */ 208 if (flags & MNT_RDONLY) 209 mp->mnt_flag |= MNT_RDONLY; 210 else if (mp->mnt_flag & MNT_RDONLY) 211 mp->mnt_iflag |= IMNT_WANTRDWR; 212 mp->mnt_flag &= 213 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 214 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 215 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 216 mp->mnt_flag |= flags & 217 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 218 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 219 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 220 MNT_IGNORE); 221 222 error = VFS_MOUNT(mp, path, data, data_len); 223 224 #if defined(COMPAT_30) && defined(NFSSERVER) 225 if (error && data != NULL) { 226 int error2; 227 228 /* Update failed; let's try and see if it was an 229 * export request. */ 230 error2 = nfs_update_exports_30(mp, path, data, l); 231 232 /* Only update error code if the export request was 233 * understood but some problem occurred while 234 * processing it. */ 235 if (error2 != EJUSTRETURN) 236 error = error2; 237 } 238 #endif 239 if (mp->mnt_iflag & IMNT_WANTRDWR) 240 mp->mnt_flag &= ~MNT_RDONLY; 241 if (error) 242 mp->mnt_flag = saved_flags; 243 mp->mnt_flag &= ~MNT_OP_FLAGS; 244 mp->mnt_iflag &= ~IMNT_WANTRDWR; 245 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 246 if (mp->mnt_syncer == NULL) 247 error = vfs_allocate_syncvnode(mp); 248 } else { 249 if (mp->mnt_syncer != NULL) 250 vfs_deallocate_syncvnode(mp); 251 } 252 vfs_unbusy(mp, false); 253 254 out: 255 return (error); 256 } 257 258 static int 259 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 260 { 261 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 262 int error; 263 264 /* Copy file-system type from userspace. */ 265 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 266 if (error) { 267 #if defined(COMPAT_09) || defined(COMPAT_43) 268 /* 269 * Historically, filesystem types were identified by numbers. 270 * If we get an integer for the filesystem type instead of a 271 * string, we check to see if it matches one of the historic 272 * filesystem types. 273 */ 274 u_long fsindex = (u_long)fstype; 275 if (fsindex >= nmountcompatnames || 276 mountcompatnames[fsindex] == NULL) 277 return ENODEV; 278 strlcpy(fstypename, mountcompatnames[fsindex], 279 sizeof(fstypename)); 280 #else 281 return error; 282 #endif 283 } 284 285 #ifdef COMPAT_10 286 /* Accept `ufs' as an alias for `ffs'. */ 287 if (strcmp(fstypename, "ufs") == 0) 288 fstypename[0] = 'f'; 289 #endif 290 291 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 292 return ENODEV; 293 return 0; 294 } 295 296 static int 297 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 298 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 299 { 300 struct mount *mp = NULL; 301 struct vnode *vp = *vpp; 302 struct vattr va; 303 int error; 304 305 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 306 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 307 if (error) 308 return error; 309 310 /* Can't make a non-dir a mount-point (from here anyway). */ 311 if (vp->v_type != VDIR) 312 return ENOTDIR; 313 314 /* 315 * If the user is not root, ensure that they own the directory 316 * onto which we are attempting to mount. 317 */ 318 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 319 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 320 (error = kauth_authorize_generic(l->l_cred, 321 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 322 return error; 323 } 324 325 if (flags & MNT_EXPORTED) 326 return EINVAL; 327 328 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 329 return error; 330 331 /* 332 * Check if a file-system is not already mounted on this vnode. 333 */ 334 if (vp->v_mountedhere != NULL) 335 return EBUSY; 336 337 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP); 338 if (mp == NULL) 339 return ENOMEM; 340 341 mp->mnt_op = vfsops; 342 mp->mnt_refcnt = 1; 343 344 TAILQ_INIT(&mp->mnt_vnodelist); 345 rw_init(&mp->mnt_lock); 346 mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE); 347 (void)vfs_busy(mp, RW_WRITER, 0); 348 349 mp->mnt_vnodecovered = vp; 350 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 351 mount_initspecific(mp); 352 353 /* 354 * The underlying file system may refuse the mount for 355 * various reasons. Allow the user to force it to happen. 356 * 357 * Set the mount level flags. 358 */ 359 mp->mnt_flag = flags & 360 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 361 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 362 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 363 MNT_IGNORE | MNT_RDONLY); 364 365 error = VFS_MOUNT(mp, path, data, data_len); 366 mp->mnt_flag &= ~MNT_OP_FLAGS; 367 368 /* 369 * Put the new filesystem on the mount list after root. 370 */ 371 cache_purge(vp); 372 if (error != 0) { 373 vp->v_mountedhere = NULL; 374 mp->mnt_op->vfs_refcount--; 375 vfs_unbusy(mp, false); 376 vfs_destroy(mp); 377 return error; 378 } 379 380 mp->mnt_iflag &= ~IMNT_WANTRDWR; 381 mutex_enter(&mountlist_lock); 382 vp->v_mountedhere = mp; 383 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 384 mutex_exit(&mountlist_lock); 385 vn_restorerecurse(vp, recurse); 386 VOP_UNLOCK(vp, 0); 387 checkdirs(vp); 388 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 389 error = vfs_allocate_syncvnode(mp); 390 /* Hold an additional reference to the mount across VFS_START(). */ 391 vfs_unbusy(mp, true); 392 (void) VFS_STATVFS(mp, &mp->mnt_stat); 393 error = VFS_START(mp, 0); 394 if (error) { 395 vrele(vp); 396 vfs_destroy(mp); 397 } 398 /* Drop reference held for VFS_START(). */ 399 vfs_destroy(mp); 400 *vpp = NULL; 401 return error; 402 } 403 404 static int 405 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 406 void *data, size_t *data_len) 407 { 408 struct mount *mp; 409 int error; 410 411 /* If MNT_GETARGS is specified, it should be the only flag. */ 412 if (flags & ~MNT_GETARGS) 413 return EINVAL; 414 415 mp = vp->v_mount; 416 417 /* XXX: probably some notion of "can see" here if we want isolation. */ 418 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 419 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 420 if (error) 421 return error; 422 423 if ((vp->v_vflag & VV_ROOT) == 0) 424 return EINVAL; 425 426 if (vfs_trybusy(mp, RW_WRITER, NULL)) 427 return EPERM; 428 429 mp->mnt_flag &= ~MNT_OP_FLAGS; 430 mp->mnt_flag |= MNT_GETARGS; 431 error = VFS_MOUNT(mp, path, data, data_len); 432 mp->mnt_flag &= ~MNT_OP_FLAGS; 433 434 vfs_unbusy(mp, false); 435 return (error); 436 } 437 438 #ifdef COMPAT_40 439 /* ARGSUSED */ 440 int 441 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 442 { 443 /* { 444 syscallarg(const char *) type; 445 syscallarg(const char *) path; 446 syscallarg(int) flags; 447 syscallarg(void *) data; 448 } */ 449 register_t dummy; 450 451 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 452 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 453 } 454 #endif 455 456 int 457 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 458 { 459 /* { 460 syscallarg(const char *) type; 461 syscallarg(const char *) path; 462 syscallarg(int) flags; 463 syscallarg(void *) data; 464 syscallarg(size_t) data_len; 465 } */ 466 467 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 468 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 469 SCARG(uap, data_len), retval); 470 } 471 472 int 473 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 474 const char *path, int flags, void *data, enum uio_seg data_seg, 475 size_t data_len, register_t *retval) 476 { 477 struct vnode *vp; 478 struct nameidata nd; 479 void *data_buf = data; 480 u_int recurse; 481 int error; 482 483 /* 484 * Get vnode to be covered 485 */ 486 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 487 if ((error = namei(&nd)) != 0) 488 return (error); 489 vp = nd.ni_vp; 490 491 /* 492 * A lookup in VFS_MOUNT might result in an attempt to 493 * lock this vnode again, so make the lock recursive. 494 */ 495 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 496 recurse = vn_setrecurse(vp); 497 498 if (vfsops == NULL) { 499 if (flags & (MNT_GETARGS | MNT_UPDATE)) 500 vfsops = vp->v_mount->mnt_op; 501 else { 502 /* 'type' is userspace */ 503 error = mount_get_vfsops(type, &vfsops); 504 if (error != 0) 505 goto done; 506 } 507 } 508 509 if (data != NULL && data_seg == UIO_USERSPACE) { 510 if (data_len == 0) { 511 /* No length supplied, use default for filesystem */ 512 data_len = vfsops->vfs_min_mount_data; 513 if (data_len > VFS_MAX_MOUNT_DATA) { 514 /* maybe a force loaded old LKM */ 515 error = EINVAL; 516 goto done; 517 } 518 #ifdef COMPAT_30 519 /* Hopefully a longer buffer won't make copyin() fail */ 520 if (flags & MNT_UPDATE 521 && data_len < sizeof (struct mnt_export_args30)) 522 data_len = sizeof (struct mnt_export_args30); 523 #endif 524 } 525 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 526 527 /* NFS needs the buffer even for mnt_getargs .... */ 528 error = copyin(data, data_buf, data_len); 529 if (error != 0) 530 goto done; 531 } 532 533 if (flags & MNT_GETARGS) { 534 if (data_len == 0) { 535 error = EINVAL; 536 goto done; 537 } 538 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 539 if (error != 0) 540 goto done; 541 if (data_seg == UIO_USERSPACE) 542 error = copyout(data_buf, data, data_len); 543 *retval = data_len; 544 } else if (flags & MNT_UPDATE) { 545 error = mount_update(l, vp, path, flags, data_buf, &data_len); 546 } else { 547 /* Locking is handled internally in mount_domount(). */ 548 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 549 &data_len, recurse); 550 } 551 552 done: 553 if (vp != NULL) { 554 vn_restorerecurse(vp, recurse); 555 vput(vp); 556 } 557 if (data_buf != data) 558 free(data_buf, M_TEMP); 559 return (error); 560 } 561 562 /* 563 * Scan all active processes to see if any of them have a current 564 * or root directory onto which the new filesystem has just been 565 * mounted. If so, replace them with the new mount point. 566 */ 567 void 568 checkdirs(struct vnode *olddp) 569 { 570 struct cwdinfo *cwdi; 571 struct vnode *newdp; 572 struct proc *p; 573 574 if (olddp->v_usecount == 1) 575 return; 576 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 577 panic("mount: lost mount"); 578 mutex_enter(&proclist_lock); 579 PROCLIST_FOREACH(p, &allproc) { 580 cwdi = p->p_cwdi; 581 if (!cwdi) 582 continue; 583 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 584 if (cwdi->cwdi_cdir == olddp) { 585 vrele(cwdi->cwdi_cdir); 586 VREF(newdp); 587 cwdi->cwdi_cdir = newdp; 588 } 589 if (cwdi->cwdi_rdir == olddp) { 590 vrele(cwdi->cwdi_rdir); 591 VREF(newdp); 592 cwdi->cwdi_rdir = newdp; 593 } 594 rw_exit(&cwdi->cwdi_lock); 595 } 596 mutex_exit(&proclist_lock); 597 if (rootvnode == olddp) { 598 vrele(rootvnode); 599 VREF(newdp); 600 rootvnode = newdp; 601 } 602 vput(newdp); 603 } 604 605 /* 606 * Unmount a file system. 607 * 608 * Note: unmount takes a path to the vnode mounted on as argument, 609 * not special file (as before). 610 */ 611 /* ARGSUSED */ 612 int 613 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 614 { 615 /* { 616 syscallarg(const char *) path; 617 syscallarg(int) flags; 618 } */ 619 struct vnode *vp; 620 struct mount *mp; 621 int error; 622 struct nameidata nd; 623 624 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 625 SCARG(uap, path)); 626 if ((error = namei(&nd)) != 0) 627 return (error); 628 vp = nd.ni_vp; 629 mp = vp->v_mount; 630 VOP_UNLOCK(vp, 0); 631 632 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 633 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 634 if (error) { 635 vrele(vp); 636 return (error); 637 } 638 639 /* 640 * Don't allow unmounting the root file system. 641 */ 642 if (mp->mnt_flag & MNT_ROOTFS) { 643 vrele(vp); 644 return (EINVAL); 645 } 646 647 /* 648 * Must be the root of the filesystem 649 */ 650 if ((vp->v_vflag & VV_ROOT) == 0) { 651 vrele(vp); 652 return (EINVAL); 653 } 654 655 /* 656 * XXX Freeze syncer. Must do this before locking the 657 * mount point. See dounmount() for details. 658 */ 659 mutex_enter(&syncer_mutex); 660 error = vfs_busy(mp, RW_WRITER, NULL); 661 vrele(vp); 662 if (error != 0) { 663 mutex_exit(&syncer_mutex); 664 return (error); 665 } 666 667 return (dounmount(mp, SCARG(uap, flags), l)); 668 } 669 670 /* 671 * Do the actual file system unmount. File system is assumed to have been 672 * marked busy by the caller. 673 */ 674 int 675 dounmount(struct mount *mp, int flags, struct lwp *l) 676 { 677 struct vnode *coveredvp; 678 int error; 679 int async; 680 int used_syncer; 681 682 KASSERT(rw_write_held(&mp->mnt_lock)); 683 684 #if NVERIEXEC > 0 685 error = veriexec_unmountchk(mp); 686 if (error) 687 return (error); 688 #endif /* NVERIEXEC > 0 */ 689 690 used_syncer = (mp->mnt_syncer != NULL); 691 692 /* 693 * XXX Syncer must be frozen when we get here. This should really 694 * be done on a per-mountpoint basis, but especially the softdep 695 * code possibly called from the syncer doesn't exactly work on a 696 * per-mountpoint basis, so the softdep code would become a maze 697 * of vfs_busy() calls. 698 * 699 * The caller of dounmount() must acquire syncer_mutex because 700 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 701 * order, and we must preserve that order to avoid deadlock. 702 * 703 * So, if the file system did not use the syncer, now is 704 * the time to release the syncer_mutex. 705 */ 706 if (used_syncer == 0) 707 mutex_exit(&syncer_mutex); 708 709 mp->mnt_iflag |= IMNT_UNMOUNT; 710 async = mp->mnt_flag & MNT_ASYNC; 711 mp->mnt_flag &= ~MNT_ASYNC; 712 cache_purgevfs(mp); /* remove cache entries for this file sys */ 713 if (mp->mnt_syncer != NULL) 714 vfs_deallocate_syncvnode(mp); 715 error = 0; 716 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 717 #if NFSS > 0 718 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 719 #endif 720 if (error == 0) 721 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 722 } 723 vfs_scrubvnlist(mp); 724 if (error == 0 || (flags & MNT_FORCE)) 725 error = VFS_UNMOUNT(mp, flags); 726 if (error) { 727 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 728 (void) vfs_allocate_syncvnode(mp); 729 mp->mnt_iflag &= ~IMNT_UNMOUNT; 730 mp->mnt_flag |= async; 731 if (used_syncer) 732 mutex_exit(&syncer_mutex); 733 vfs_unbusy(mp, false); 734 return (error); 735 } 736 vfs_scrubvnlist(mp); 737 mutex_enter(&mountlist_lock); 738 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 739 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 740 coveredvp->v_mountedhere = NULL; 741 mutex_exit(&mountlist_lock); 742 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 743 panic("unmount: dangling vnode"); 744 mp->mnt_iflag |= IMNT_GONE; 745 if (used_syncer) 746 mutex_exit(&syncer_mutex); 747 vfs_hooks_unmount(mp); 748 vfs_unbusy(mp, false); 749 vfs_destroy(mp); 750 if (coveredvp != NULLVP) 751 vrele(coveredvp); 752 return (0); 753 } 754 755 /* 756 * Sync each mounted filesystem. 757 */ 758 #ifdef DEBUG 759 int syncprt = 0; 760 struct ctldebug debug0 = { "syncprt", &syncprt }; 761 #endif 762 763 /* ARGSUSED */ 764 int 765 sys_sync(struct lwp *l, const void *v, register_t *retval) 766 { 767 struct mount *mp, *nmp; 768 int asyncflag; 769 770 if (l == NULL) 771 l = &lwp0; 772 773 mutex_enter(&mountlist_lock); 774 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 775 if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) { 776 nmp = mp->mnt_list.cqe_prev; 777 continue; 778 } 779 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 780 asyncflag = mp->mnt_flag & MNT_ASYNC; 781 mp->mnt_flag &= ~MNT_ASYNC; 782 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 783 if (asyncflag) 784 mp->mnt_flag |= MNT_ASYNC; 785 } 786 mutex_enter(&mountlist_lock); 787 nmp = mp->mnt_list.cqe_prev; 788 vfs_unbusy(mp, false); 789 790 } 791 mutex_exit(&mountlist_lock); 792 #ifdef DEBUG 793 if (syncprt) 794 vfs_bufstats(); 795 #endif /* DEBUG */ 796 return (0); 797 } 798 799 /* 800 * Change filesystem quotas. 801 */ 802 /* ARGSUSED */ 803 int 804 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 805 { 806 /* { 807 syscallarg(const char *) path; 808 syscallarg(int) cmd; 809 syscallarg(int) uid; 810 syscallarg(void *) arg; 811 } */ 812 struct mount *mp; 813 int error; 814 struct nameidata nd; 815 816 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 817 SCARG(uap, path)); 818 if ((error = namei(&nd)) != 0) 819 return (error); 820 mp = nd.ni_vp->v_mount; 821 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 822 SCARG(uap, arg)); 823 vrele(nd.ni_vp); 824 return (error); 825 } 826 827 int 828 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 829 int root) 830 { 831 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 832 int error = 0; 833 834 /* 835 * If MNT_NOWAIT or MNT_LAZY is specified, do not 836 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 837 * overrides MNT_NOWAIT. 838 */ 839 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 840 (flags != MNT_WAIT && flags != 0)) { 841 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 842 goto done; 843 } 844 845 /* Get the filesystem stats now */ 846 memset(sp, 0, sizeof(*sp)); 847 if ((error = VFS_STATVFS(mp, sp)) != 0) { 848 return error; 849 } 850 851 if (cwdi->cwdi_rdir == NULL) 852 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 853 done: 854 if (cwdi->cwdi_rdir != NULL) { 855 size_t len; 856 char *bp; 857 char *path = PNBUF_GET(); 858 859 bp = path + MAXPATHLEN; 860 *--bp = '\0'; 861 rw_enter(&cwdi->cwdi_lock, RW_READER); 862 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 863 MAXPATHLEN / 2, 0, l); 864 rw_exit(&cwdi->cwdi_lock); 865 if (error) { 866 PNBUF_PUT(path); 867 return error; 868 } 869 len = strlen(bp); 870 /* 871 * for mount points that are below our root, we can see 872 * them, so we fix up the pathname and return them. The 873 * rest we cannot see, so we don't allow viewing the 874 * data. 875 */ 876 if (strncmp(bp, sp->f_mntonname, len) == 0) { 877 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 878 sizeof(sp->f_mntonname)); 879 if (sp->f_mntonname[0] == '\0') 880 (void)strlcpy(sp->f_mntonname, "/", 881 sizeof(sp->f_mntonname)); 882 } else { 883 if (root) 884 (void)strlcpy(sp->f_mntonname, "/", 885 sizeof(sp->f_mntonname)); 886 else 887 error = EPERM; 888 } 889 PNBUF_PUT(path); 890 } 891 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 892 return error; 893 } 894 895 /* 896 * Get filesystem statistics by path. 897 */ 898 int 899 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 900 { 901 struct mount *mp; 902 int error; 903 struct nameidata nd; 904 905 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 906 if ((error = namei(&nd)) != 0) 907 return error; 908 mp = nd.ni_vp->v_mount; 909 error = dostatvfs(mp, sb, l, flags, 1); 910 vrele(nd.ni_vp); 911 return error; 912 } 913 914 /* ARGSUSED */ 915 int 916 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 917 { 918 /* { 919 syscallarg(const char *) path; 920 syscallarg(struct statvfs *) buf; 921 syscallarg(int) flags; 922 } */ 923 struct statvfs *sb; 924 int error; 925 926 sb = STATVFSBUF_GET(); 927 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 928 if (error == 0) 929 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 930 STATVFSBUF_PUT(sb); 931 return error; 932 } 933 934 /* 935 * Get filesystem statistics by fd. 936 */ 937 int 938 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 939 { 940 file_t *fp; 941 struct mount *mp; 942 int error; 943 944 /* fd_getvnode() will use the descriptor for us */ 945 if ((error = fd_getvnode(fd, &fp)) != 0) 946 return (error); 947 mp = ((struct vnode *)fp->f_data)->v_mount; 948 error = dostatvfs(mp, sb, curlwp, flags, 1); 949 fd_putfile(fd); 950 return error; 951 } 952 953 /* ARGSUSED */ 954 int 955 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 956 { 957 /* { 958 syscallarg(int) fd; 959 syscallarg(struct statvfs *) buf; 960 syscallarg(int) flags; 961 } */ 962 struct statvfs *sb; 963 int error; 964 965 sb = STATVFSBUF_GET(); 966 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 967 if (error == 0) 968 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 969 STATVFSBUF_PUT(sb); 970 return error; 971 } 972 973 974 /* 975 * Get statistics on all filesystems. 976 */ 977 int 978 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 979 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 980 register_t *retval) 981 { 982 int root = 0; 983 struct proc *p = l->l_proc; 984 struct mount *mp, *nmp; 985 struct statvfs *sb; 986 size_t count, maxcount; 987 int error = 0; 988 989 sb = STATVFSBUF_GET(); 990 maxcount = bufsize / entry_sz; 991 mutex_enter(&mountlist_lock); 992 count = 0; 993 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 994 mp = nmp) { 995 if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) { 996 nmp = CIRCLEQ_NEXT(mp, mnt_list); 997 continue; 998 } 999 if (sfsp && count < maxcount) { 1000 error = dostatvfs(mp, sb, l, flags, 0); 1001 if (error) { 1002 mutex_enter(&mountlist_lock); 1003 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1004 vfs_unbusy(mp, false); 1005 continue; 1006 } 1007 error = copyfn(sb, sfsp, entry_sz); 1008 if (error) { 1009 vfs_unbusy(mp, false); 1010 goto out; 1011 } 1012 sfsp = (char *)sfsp + entry_sz; 1013 root |= strcmp(sb->f_mntonname, "/") == 0; 1014 } 1015 count++; 1016 mutex_enter(&mountlist_lock); 1017 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1018 vfs_unbusy(mp, false); 1019 } 1020 1021 mutex_exit(&mountlist_lock); 1022 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1023 /* 1024 * fake a root entry 1025 */ 1026 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1027 sb, l, flags, 1); 1028 if (error != 0) 1029 goto out; 1030 if (sfsp) 1031 error = copyfn(sb, sfsp, entry_sz); 1032 count++; 1033 } 1034 if (sfsp && count > maxcount) 1035 *retval = maxcount; 1036 else 1037 *retval = count; 1038 out: 1039 STATVFSBUF_PUT(sb); 1040 return error; 1041 } 1042 1043 int 1044 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1045 { 1046 /* { 1047 syscallarg(struct statvfs *) buf; 1048 syscallarg(size_t) bufsize; 1049 syscallarg(int) flags; 1050 } */ 1051 1052 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1053 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1054 } 1055 1056 /* 1057 * Change current working directory to a given file descriptor. 1058 */ 1059 /* ARGSUSED */ 1060 int 1061 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1062 { 1063 /* { 1064 syscallarg(int) fd; 1065 } */ 1066 struct proc *p = l->l_proc; 1067 struct cwdinfo *cwdi; 1068 struct vnode *vp, *tdp; 1069 struct mount *mp; 1070 file_t *fp; 1071 int error, fd; 1072 1073 /* fd_getvnode() will use the descriptor for us */ 1074 fd = SCARG(uap, fd); 1075 if ((error = fd_getvnode(fd, &fp)) != 0) 1076 return (error); 1077 vp = fp->f_data; 1078 1079 VREF(vp); 1080 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1081 if (vp->v_type != VDIR) 1082 error = ENOTDIR; 1083 else 1084 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1085 if (error) { 1086 vput(vp); 1087 goto out; 1088 } 1089 while ((mp = vp->v_mountedhere) != NULL) { 1090 if (vfs_busy(mp, RW_READER, NULL)) 1091 continue; 1092 vput(vp); 1093 error = VFS_ROOT(mp, &tdp); 1094 vfs_unbusy(mp, false); 1095 if (error) 1096 goto out; 1097 vp = tdp; 1098 } 1099 VOP_UNLOCK(vp, 0); 1100 1101 /* 1102 * Disallow changing to a directory not under the process's 1103 * current root directory (if there is one). 1104 */ 1105 cwdi = p->p_cwdi; 1106 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1107 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1108 vrele(vp); 1109 error = EPERM; /* operation not permitted */ 1110 } else { 1111 vrele(cwdi->cwdi_cdir); 1112 cwdi->cwdi_cdir = vp; 1113 } 1114 rw_exit(&cwdi->cwdi_lock); 1115 1116 out: 1117 fd_putfile(fd); 1118 return (error); 1119 } 1120 1121 /* 1122 * Change this process's notion of the root directory to a given file 1123 * descriptor. 1124 */ 1125 int 1126 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1127 { 1128 struct proc *p = l->l_proc; 1129 struct cwdinfo *cwdi; 1130 struct vnode *vp; 1131 file_t *fp; 1132 int error, fd = SCARG(uap, fd); 1133 1134 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1135 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1136 return error; 1137 /* fd_getvnode() will use the descriptor for us */ 1138 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 1139 return error; 1140 vp = fp->f_data; 1141 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1142 if (vp->v_type != VDIR) 1143 error = ENOTDIR; 1144 else 1145 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1146 VOP_UNLOCK(vp, 0); 1147 if (error) 1148 goto out; 1149 VREF(vp); 1150 1151 /* 1152 * Prevent escaping from chroot by putting the root under 1153 * the working directory. Silently chdir to / if we aren't 1154 * already there. 1155 */ 1156 cwdi = p->p_cwdi; 1157 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1158 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1159 /* 1160 * XXX would be more failsafe to change directory to a 1161 * deadfs node here instead 1162 */ 1163 vrele(cwdi->cwdi_cdir); 1164 VREF(vp); 1165 cwdi->cwdi_cdir = vp; 1166 } 1167 1168 if (cwdi->cwdi_rdir != NULL) 1169 vrele(cwdi->cwdi_rdir); 1170 cwdi->cwdi_rdir = vp; 1171 rw_exit(&cwdi->cwdi_lock); 1172 1173 out: 1174 fd_putfile(fd); 1175 return (error); 1176 } 1177 1178 /* 1179 * Change current working directory (``.''). 1180 */ 1181 /* ARGSUSED */ 1182 int 1183 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1184 { 1185 /* { 1186 syscallarg(const char *) path; 1187 } */ 1188 struct proc *p = l->l_proc; 1189 struct cwdinfo *cwdi; 1190 int error; 1191 struct nameidata nd; 1192 1193 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1194 SCARG(uap, path)); 1195 if ((error = change_dir(&nd, l)) != 0) 1196 return (error); 1197 cwdi = p->p_cwdi; 1198 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1199 vrele(cwdi->cwdi_cdir); 1200 cwdi->cwdi_cdir = nd.ni_vp; 1201 rw_exit(&cwdi->cwdi_lock); 1202 return (0); 1203 } 1204 1205 /* 1206 * Change notion of root (``/'') directory. 1207 */ 1208 /* ARGSUSED */ 1209 int 1210 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1211 { 1212 /* { 1213 syscallarg(const char *) path; 1214 } */ 1215 struct proc *p = l->l_proc; 1216 struct cwdinfo *cwdi; 1217 struct vnode *vp; 1218 int error; 1219 struct nameidata nd; 1220 1221 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1222 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1223 return (error); 1224 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1225 SCARG(uap, path)); 1226 if ((error = change_dir(&nd, l)) != 0) 1227 return (error); 1228 1229 cwdi = p->p_cwdi; 1230 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1231 if (cwdi->cwdi_rdir != NULL) 1232 vrele(cwdi->cwdi_rdir); 1233 vp = nd.ni_vp; 1234 cwdi->cwdi_rdir = vp; 1235 1236 /* 1237 * Prevent escaping from chroot by putting the root under 1238 * the working directory. Silently chdir to / if we aren't 1239 * already there. 1240 */ 1241 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1242 /* 1243 * XXX would be more failsafe to change directory to a 1244 * deadfs node here instead 1245 */ 1246 vrele(cwdi->cwdi_cdir); 1247 VREF(vp); 1248 cwdi->cwdi_cdir = vp; 1249 } 1250 rw_exit(&cwdi->cwdi_lock); 1251 1252 return (0); 1253 } 1254 1255 /* 1256 * Common routine for chroot and chdir. 1257 */ 1258 static int 1259 change_dir(struct nameidata *ndp, struct lwp *l) 1260 { 1261 struct vnode *vp; 1262 int error; 1263 1264 if ((error = namei(ndp)) != 0) 1265 return (error); 1266 vp = ndp->ni_vp; 1267 if (vp->v_type != VDIR) 1268 error = ENOTDIR; 1269 else 1270 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1271 1272 if (error) 1273 vput(vp); 1274 else 1275 VOP_UNLOCK(vp, 0); 1276 return (error); 1277 } 1278 1279 /* 1280 * Check permissions, allocate an open file structure, 1281 * and call the device open routine if any. 1282 */ 1283 int 1284 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1285 { 1286 /* { 1287 syscallarg(const char *) path; 1288 syscallarg(int) flags; 1289 syscallarg(int) mode; 1290 } */ 1291 struct proc *p = l->l_proc; 1292 struct cwdinfo *cwdi = p->p_cwdi; 1293 file_t *fp; 1294 struct vnode *vp; 1295 int flags, cmode; 1296 int type, indx, error; 1297 struct flock lf; 1298 struct nameidata nd; 1299 1300 flags = FFLAGS(SCARG(uap, flags)); 1301 if ((flags & (FREAD | FWRITE)) == 0) 1302 return (EINVAL); 1303 if ((error = fd_allocfile(&fp, &indx)) != 0) 1304 return (error); 1305 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1306 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1307 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1308 SCARG(uap, path)); 1309 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1310 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1311 fd_abort(p, fp, indx); 1312 if ((error == EDUPFD || error == EMOVEFD) && 1313 l->l_dupfd >= 0 && /* XXX from fdopen */ 1314 (error = 1315 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1316 *retval = indx; 1317 return (0); 1318 } 1319 if (error == ERESTART) 1320 error = EINTR; 1321 return (error); 1322 } 1323 1324 l->l_dupfd = 0; 1325 vp = nd.ni_vp; 1326 fp->f_flag = flags & FMASK; 1327 fp->f_type = DTYPE_VNODE; 1328 fp->f_ops = &vnops; 1329 fp->f_data = vp; 1330 if (flags & (O_EXLOCK | O_SHLOCK)) { 1331 lf.l_whence = SEEK_SET; 1332 lf.l_start = 0; 1333 lf.l_len = 0; 1334 if (flags & O_EXLOCK) 1335 lf.l_type = F_WRLCK; 1336 else 1337 lf.l_type = F_RDLCK; 1338 type = F_FLOCK; 1339 if ((flags & FNONBLOCK) == 0) 1340 type |= F_WAIT; 1341 VOP_UNLOCK(vp, 0); 1342 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1343 if (error) { 1344 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1345 fd_abort(p, fp, indx); 1346 return (error); 1347 } 1348 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1349 atomic_or_uint(&fp->f_flag, FHASLOCK); 1350 } 1351 VOP_UNLOCK(vp, 0); 1352 *retval = indx; 1353 fd_affix(p, fp, indx); 1354 return (0); 1355 } 1356 1357 static void 1358 vfs__fhfree(fhandle_t *fhp) 1359 { 1360 size_t fhsize; 1361 1362 if (fhp == NULL) { 1363 return; 1364 } 1365 fhsize = FHANDLE_SIZE(fhp); 1366 kmem_free(fhp, fhsize); 1367 } 1368 1369 /* 1370 * vfs_composefh: compose a filehandle. 1371 */ 1372 1373 int 1374 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1375 { 1376 struct mount *mp; 1377 struct fid *fidp; 1378 int error; 1379 size_t needfhsize; 1380 size_t fidsize; 1381 1382 mp = vp->v_mount; 1383 fidp = NULL; 1384 if (*fh_size < FHANDLE_SIZE_MIN) { 1385 fidsize = 0; 1386 } else { 1387 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1388 if (fhp != NULL) { 1389 memset(fhp, 0, *fh_size); 1390 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1391 fidp = &fhp->fh_fid; 1392 } 1393 } 1394 error = VFS_VPTOFH(vp, fidp, &fidsize); 1395 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1396 if (error == 0 && *fh_size < needfhsize) { 1397 error = E2BIG; 1398 } 1399 *fh_size = needfhsize; 1400 return error; 1401 } 1402 1403 int 1404 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1405 { 1406 struct mount *mp; 1407 fhandle_t *fhp; 1408 size_t fhsize; 1409 size_t fidsize; 1410 int error; 1411 1412 *fhpp = NULL; 1413 mp = vp->v_mount; 1414 fidsize = 0; 1415 error = VFS_VPTOFH(vp, NULL, &fidsize); 1416 KASSERT(error != 0); 1417 if (error != E2BIG) { 1418 goto out; 1419 } 1420 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1421 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1422 if (fhp == NULL) { 1423 error = ENOMEM; 1424 goto out; 1425 } 1426 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1427 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1428 if (error == 0) { 1429 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1430 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1431 *fhpp = fhp; 1432 } else { 1433 kmem_free(fhp, fhsize); 1434 } 1435 out: 1436 return error; 1437 } 1438 1439 void 1440 vfs_composefh_free(fhandle_t *fhp) 1441 { 1442 1443 vfs__fhfree(fhp); 1444 } 1445 1446 /* 1447 * vfs_fhtovp: lookup a vnode by a filehandle. 1448 */ 1449 1450 int 1451 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1452 { 1453 struct mount *mp; 1454 int error; 1455 1456 *vpp = NULL; 1457 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1458 if (mp == NULL) { 1459 error = ESTALE; 1460 goto out; 1461 } 1462 if (mp->mnt_op->vfs_fhtovp == NULL) { 1463 error = EOPNOTSUPP; 1464 goto out; 1465 } 1466 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1467 out: 1468 return error; 1469 } 1470 1471 /* 1472 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1473 * the needed size. 1474 */ 1475 1476 int 1477 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1478 { 1479 fhandle_t *fhp; 1480 int error; 1481 1482 *fhpp = NULL; 1483 if (fhsize > FHANDLE_SIZE_MAX) { 1484 return EINVAL; 1485 } 1486 if (fhsize < FHANDLE_SIZE_MIN) { 1487 return EINVAL; 1488 } 1489 again: 1490 fhp = kmem_alloc(fhsize, KM_SLEEP); 1491 if (fhp == NULL) { 1492 return ENOMEM; 1493 } 1494 error = copyin(ufhp, fhp, fhsize); 1495 if (error == 0) { 1496 /* XXX this check shouldn't be here */ 1497 if (FHANDLE_SIZE(fhp) == fhsize) { 1498 *fhpp = fhp; 1499 return 0; 1500 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1501 /* 1502 * a kludge for nfsv2 padded handles. 1503 */ 1504 size_t sz; 1505 1506 sz = FHANDLE_SIZE(fhp); 1507 kmem_free(fhp, fhsize); 1508 fhsize = sz; 1509 goto again; 1510 } else { 1511 /* 1512 * userland told us wrong size. 1513 */ 1514 error = EINVAL; 1515 } 1516 } 1517 kmem_free(fhp, fhsize); 1518 return error; 1519 } 1520 1521 void 1522 vfs_copyinfh_free(fhandle_t *fhp) 1523 { 1524 1525 vfs__fhfree(fhp); 1526 } 1527 1528 /* 1529 * Get file handle system call 1530 */ 1531 int 1532 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1533 { 1534 /* { 1535 syscallarg(char *) fname; 1536 syscallarg(fhandle_t *) fhp; 1537 syscallarg(size_t *) fh_size; 1538 } */ 1539 struct vnode *vp; 1540 fhandle_t *fh; 1541 int error; 1542 struct nameidata nd; 1543 size_t sz; 1544 size_t usz; 1545 1546 /* 1547 * Must be super user 1548 */ 1549 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1550 0, NULL, NULL, NULL); 1551 if (error) 1552 return (error); 1553 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1554 SCARG(uap, fname)); 1555 error = namei(&nd); 1556 if (error) 1557 return (error); 1558 vp = nd.ni_vp; 1559 error = vfs_composefh_alloc(vp, &fh); 1560 vput(vp); 1561 if (error != 0) { 1562 goto out; 1563 } 1564 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1565 if (error != 0) { 1566 goto out; 1567 } 1568 sz = FHANDLE_SIZE(fh); 1569 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1570 if (error != 0) { 1571 goto out; 1572 } 1573 if (usz >= sz) { 1574 error = copyout(fh, SCARG(uap, fhp), sz); 1575 } else { 1576 error = E2BIG; 1577 } 1578 out: 1579 vfs_composefh_free(fh); 1580 return (error); 1581 } 1582 1583 /* 1584 * Open a file given a file handle. 1585 * 1586 * Check permissions, allocate an open file structure, 1587 * and call the device open routine if any. 1588 */ 1589 1590 int 1591 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1592 register_t *retval) 1593 { 1594 file_t *fp; 1595 struct vnode *vp = NULL; 1596 kauth_cred_t cred = l->l_cred; 1597 file_t *nfp; 1598 int type, indx, error=0; 1599 struct flock lf; 1600 struct vattr va; 1601 fhandle_t *fh; 1602 int flags; 1603 proc_t *p; 1604 1605 p = curproc; 1606 1607 /* 1608 * Must be super user 1609 */ 1610 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1611 0, NULL, NULL, NULL))) 1612 return (error); 1613 1614 flags = FFLAGS(oflags); 1615 if ((flags & (FREAD | FWRITE)) == 0) 1616 return (EINVAL); 1617 if ((flags & O_CREAT)) 1618 return (EINVAL); 1619 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1620 return (error); 1621 fp = nfp; 1622 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1623 if (error != 0) { 1624 goto bad; 1625 } 1626 error = vfs_fhtovp(fh, &vp); 1627 if (error != 0) { 1628 goto bad; 1629 } 1630 1631 /* Now do an effective vn_open */ 1632 1633 if (vp->v_type == VSOCK) { 1634 error = EOPNOTSUPP; 1635 goto bad; 1636 } 1637 error = vn_openchk(vp, cred, flags); 1638 if (error != 0) 1639 goto bad; 1640 if (flags & O_TRUNC) { 1641 VOP_UNLOCK(vp, 0); /* XXX */ 1642 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1643 VATTR_NULL(&va); 1644 va.va_size = 0; 1645 error = VOP_SETATTR(vp, &va, cred); 1646 if (error) 1647 goto bad; 1648 } 1649 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1650 goto bad; 1651 if (flags & FWRITE) { 1652 mutex_enter(&vp->v_interlock); 1653 vp->v_writecount++; 1654 mutex_exit(&vp->v_interlock); 1655 } 1656 1657 /* done with modified vn_open, now finish what sys_open does. */ 1658 1659 fp->f_flag = flags & FMASK; 1660 fp->f_type = DTYPE_VNODE; 1661 fp->f_ops = &vnops; 1662 fp->f_data = vp; 1663 if (flags & (O_EXLOCK | O_SHLOCK)) { 1664 lf.l_whence = SEEK_SET; 1665 lf.l_start = 0; 1666 lf.l_len = 0; 1667 if (flags & O_EXLOCK) 1668 lf.l_type = F_WRLCK; 1669 else 1670 lf.l_type = F_RDLCK; 1671 type = F_FLOCK; 1672 if ((flags & FNONBLOCK) == 0) 1673 type |= F_WAIT; 1674 VOP_UNLOCK(vp, 0); 1675 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1676 if (error) { 1677 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1678 fd_abort(p, fp, indx); 1679 return (error); 1680 } 1681 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1682 atomic_or_uint(&fp->f_flag, FHASLOCK); 1683 } 1684 VOP_UNLOCK(vp, 0); 1685 *retval = indx; 1686 fd_affix(p, fp, indx); 1687 vfs_copyinfh_free(fh); 1688 return (0); 1689 1690 bad: 1691 fd_abort(p, fp, indx); 1692 if (vp != NULL) 1693 vput(vp); 1694 vfs_copyinfh_free(fh); 1695 return (error); 1696 } 1697 1698 int 1699 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1700 { 1701 /* { 1702 syscallarg(const void *) fhp; 1703 syscallarg(size_t) fh_size; 1704 syscallarg(int) flags; 1705 } */ 1706 1707 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1708 SCARG(uap, flags), retval); 1709 } 1710 1711 int 1712 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1713 { 1714 int error; 1715 fhandle_t *fh; 1716 struct vnode *vp; 1717 1718 /* 1719 * Must be super user 1720 */ 1721 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1722 0, NULL, NULL, NULL))) 1723 return (error); 1724 1725 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1726 if (error != 0) 1727 return error; 1728 1729 error = vfs_fhtovp(fh, &vp); 1730 vfs_copyinfh_free(fh); 1731 if (error != 0) 1732 return error; 1733 1734 error = vn_stat(vp, sb); 1735 vput(vp); 1736 return error; 1737 } 1738 1739 1740 /* ARGSUSED */ 1741 int 1742 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1743 { 1744 /* { 1745 syscallarg(const void *) fhp; 1746 syscallarg(size_t) fh_size; 1747 syscallarg(struct stat *) sb; 1748 } */ 1749 struct stat sb; 1750 int error; 1751 1752 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1753 if (error) 1754 return error; 1755 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1756 } 1757 1758 int 1759 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1760 int flags) 1761 { 1762 fhandle_t *fh; 1763 struct mount *mp; 1764 struct vnode *vp; 1765 int error; 1766 1767 /* 1768 * Must be super user 1769 */ 1770 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1771 0, NULL, NULL, NULL))) 1772 return error; 1773 1774 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1775 if (error != 0) 1776 return error; 1777 1778 error = vfs_fhtovp(fh, &vp); 1779 vfs_copyinfh_free(fh); 1780 if (error != 0) 1781 return error; 1782 1783 mp = vp->v_mount; 1784 error = dostatvfs(mp, sb, l, flags, 1); 1785 vput(vp); 1786 return error; 1787 } 1788 1789 /* ARGSUSED */ 1790 int 1791 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1792 { 1793 /* { 1794 syscallarg(const void *) fhp; 1795 syscallarg(size_t) fh_size; 1796 syscallarg(struct statvfs *) buf; 1797 syscallarg(int) flags; 1798 } */ 1799 struct statvfs *sb = STATVFSBUF_GET(); 1800 int error; 1801 1802 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1803 SCARG(uap, flags)); 1804 if (error == 0) 1805 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1806 STATVFSBUF_PUT(sb); 1807 return error; 1808 } 1809 1810 /* 1811 * Create a special file. 1812 */ 1813 /* ARGSUSED */ 1814 int 1815 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 1816 register_t *retval) 1817 { 1818 /* { 1819 syscallarg(const char *) path; 1820 syscallarg(mode_t) mode; 1821 syscallarg(dev_t) dev; 1822 } */ 1823 return do_sys_mknod(l, SCARG(uap, path), SCARG(uap, mode), 1824 SCARG(uap, dev), retval); 1825 } 1826 1827 int 1828 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 1829 register_t *retval) 1830 { 1831 struct proc *p = l->l_proc; 1832 struct vnode *vp; 1833 struct vattr vattr; 1834 int error, optype; 1835 struct nameidata nd; 1836 char *path; 1837 const char *cpath; 1838 enum uio_seg seg = UIO_USERSPACE; 1839 1840 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1841 0, NULL, NULL, NULL)) != 0) 1842 return (error); 1843 1844 optype = VOP_MKNOD_DESCOFFSET; 1845 1846 VERIEXEC_PATH_GET(pathname, seg, cpath, path); 1847 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1848 1849 if ((error = namei(&nd)) != 0) 1850 goto out; 1851 vp = nd.ni_vp; 1852 if (vp != NULL) 1853 error = EEXIST; 1854 else { 1855 VATTR_NULL(&vattr); 1856 /* We will read cwdi->cwdi_cmask unlocked. */ 1857 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1858 vattr.va_rdev = dev; 1859 1860 switch (mode & S_IFMT) { 1861 case S_IFMT: /* used by badsect to flag bad sectors */ 1862 vattr.va_type = VBAD; 1863 break; 1864 case S_IFCHR: 1865 vattr.va_type = VCHR; 1866 break; 1867 case S_IFBLK: 1868 vattr.va_type = VBLK; 1869 break; 1870 case S_IFWHT: 1871 optype = VOP_WHITEOUT_DESCOFFSET; 1872 break; 1873 case S_IFREG: 1874 #if NVERIEXEC > 0 1875 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1876 O_CREAT); 1877 #endif /* NVERIEXEC > 0 */ 1878 vattr.va_type = VREG; 1879 vattr.va_rdev = VNOVAL; 1880 optype = VOP_CREATE_DESCOFFSET; 1881 break; 1882 default: 1883 error = EINVAL; 1884 break; 1885 } 1886 } 1887 if (!error) { 1888 switch (optype) { 1889 case VOP_WHITEOUT_DESCOFFSET: 1890 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1891 if (error) 1892 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1893 vput(nd.ni_dvp); 1894 break; 1895 1896 case VOP_MKNOD_DESCOFFSET: 1897 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1898 &nd.ni_cnd, &vattr); 1899 if (error == 0) 1900 vput(nd.ni_vp); 1901 break; 1902 1903 case VOP_CREATE_DESCOFFSET: 1904 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1905 &nd.ni_cnd, &vattr); 1906 if (error == 0) 1907 vput(nd.ni_vp); 1908 break; 1909 } 1910 } else { 1911 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1912 if (nd.ni_dvp == vp) 1913 vrele(nd.ni_dvp); 1914 else 1915 vput(nd.ni_dvp); 1916 if (vp) 1917 vrele(vp); 1918 } 1919 out: 1920 VERIEXEC_PATH_PUT(path); 1921 return (error); 1922 } 1923 1924 /* 1925 * Create a named pipe. 1926 */ 1927 /* ARGSUSED */ 1928 int 1929 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1930 { 1931 /* { 1932 syscallarg(const char *) path; 1933 syscallarg(int) mode; 1934 } */ 1935 struct proc *p = l->l_proc; 1936 struct vattr vattr; 1937 int error; 1938 struct nameidata nd; 1939 1940 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1941 SCARG(uap, path)); 1942 if ((error = namei(&nd)) != 0) 1943 return (error); 1944 if (nd.ni_vp != NULL) { 1945 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1946 if (nd.ni_dvp == nd.ni_vp) 1947 vrele(nd.ni_dvp); 1948 else 1949 vput(nd.ni_dvp); 1950 vrele(nd.ni_vp); 1951 return (EEXIST); 1952 } 1953 VATTR_NULL(&vattr); 1954 vattr.va_type = VFIFO; 1955 /* We will read cwdi->cwdi_cmask unlocked. */ 1956 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1957 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1958 if (error == 0) 1959 vput(nd.ni_vp); 1960 return (error); 1961 } 1962 1963 /* 1964 * Make a hard file link. 1965 */ 1966 /* ARGSUSED */ 1967 int 1968 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1969 { 1970 /* { 1971 syscallarg(const char *) path; 1972 syscallarg(const char *) link; 1973 } */ 1974 struct vnode *vp; 1975 struct nameidata nd; 1976 int error; 1977 1978 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1979 SCARG(uap, path)); 1980 if ((error = namei(&nd)) != 0) 1981 return (error); 1982 vp = nd.ni_vp; 1983 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1984 SCARG(uap, link)); 1985 if ((error = namei(&nd)) != 0) 1986 goto out; 1987 if (nd.ni_vp) { 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 error = EEXIST; 1995 goto out; 1996 } 1997 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1998 out: 1999 vrele(vp); 2000 return (error); 2001 } 2002 2003 /* 2004 * Make a symbolic link. 2005 */ 2006 /* ARGSUSED */ 2007 int 2008 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2009 { 2010 /* { 2011 syscallarg(const char *) path; 2012 syscallarg(const char *) link; 2013 } */ 2014 struct proc *p = l->l_proc; 2015 struct vattr vattr; 2016 char *path; 2017 int error; 2018 struct nameidata nd; 2019 2020 path = PNBUF_GET(); 2021 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2022 if (error) 2023 goto out; 2024 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2025 SCARG(uap, link)); 2026 if ((error = namei(&nd)) != 0) 2027 goto out; 2028 if (nd.ni_vp) { 2029 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2030 if (nd.ni_dvp == nd.ni_vp) 2031 vrele(nd.ni_dvp); 2032 else 2033 vput(nd.ni_dvp); 2034 vrele(nd.ni_vp); 2035 error = EEXIST; 2036 goto out; 2037 } 2038 VATTR_NULL(&vattr); 2039 vattr.va_type = VLNK; 2040 /* We will read cwdi->cwdi_cmask unlocked. */ 2041 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2042 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2043 if (error == 0) 2044 vput(nd.ni_vp); 2045 out: 2046 PNBUF_PUT(path); 2047 return (error); 2048 } 2049 2050 /* 2051 * Delete a whiteout from the filesystem. 2052 */ 2053 /* ARGSUSED */ 2054 int 2055 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2056 { 2057 /* { 2058 syscallarg(const char *) path; 2059 } */ 2060 int error; 2061 struct nameidata nd; 2062 2063 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2064 UIO_USERSPACE, SCARG(uap, path)); 2065 error = namei(&nd); 2066 if (error) 2067 return (error); 2068 2069 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2070 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2071 if (nd.ni_dvp == nd.ni_vp) 2072 vrele(nd.ni_dvp); 2073 else 2074 vput(nd.ni_dvp); 2075 if (nd.ni_vp) 2076 vrele(nd.ni_vp); 2077 return (EEXIST); 2078 } 2079 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2080 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2081 vput(nd.ni_dvp); 2082 return (error); 2083 } 2084 2085 /* 2086 * Delete a name from the filesystem. 2087 */ 2088 /* ARGSUSED */ 2089 int 2090 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2091 { 2092 /* { 2093 syscallarg(const char *) path; 2094 } */ 2095 2096 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2097 } 2098 2099 int 2100 do_sys_unlink(const char *arg, enum uio_seg seg) 2101 { 2102 struct vnode *vp; 2103 int error; 2104 struct nameidata nd; 2105 kauth_cred_t cred; 2106 char *path; 2107 const char *cpath; 2108 2109 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2110 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2111 2112 if ((error = namei(&nd)) != 0) 2113 goto out; 2114 vp = nd.ni_vp; 2115 2116 /* 2117 * The root of a mounted filesystem cannot be deleted. 2118 */ 2119 if (vp->v_vflag & VV_ROOT) { 2120 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2121 if (nd.ni_dvp == vp) 2122 vrele(nd.ni_dvp); 2123 else 2124 vput(nd.ni_dvp); 2125 vput(vp); 2126 error = EBUSY; 2127 goto out; 2128 } 2129 2130 #if NVERIEXEC > 0 2131 /* Handle remove requests for veriexec entries. */ 2132 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2133 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2134 if (nd.ni_dvp == vp) 2135 vrele(nd.ni_dvp); 2136 else 2137 vput(nd.ni_dvp); 2138 vput(vp); 2139 goto out; 2140 } 2141 #endif /* NVERIEXEC > 0 */ 2142 2143 cred = kauth_cred_get(); 2144 #ifdef FILEASSOC 2145 (void)fileassoc_file_delete(vp); 2146 #endif /* FILEASSOC */ 2147 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2148 out: 2149 VERIEXEC_PATH_PUT(path); 2150 return (error); 2151 } 2152 2153 /* 2154 * Reposition read/write file offset. 2155 */ 2156 int 2157 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2158 { 2159 /* { 2160 syscallarg(int) fd; 2161 syscallarg(int) pad; 2162 syscallarg(off_t) offset; 2163 syscallarg(int) whence; 2164 } */ 2165 kauth_cred_t cred = l->l_cred; 2166 file_t *fp; 2167 struct vnode *vp; 2168 struct vattr vattr; 2169 off_t newoff; 2170 int error, fd; 2171 2172 fd = SCARG(uap, fd); 2173 2174 if ((fp = fd_getfile(fd)) == NULL) 2175 return (EBADF); 2176 2177 vp = fp->f_data; 2178 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2179 error = ESPIPE; 2180 goto out; 2181 } 2182 2183 switch (SCARG(uap, whence)) { 2184 case SEEK_CUR: 2185 newoff = fp->f_offset + SCARG(uap, offset); 2186 break; 2187 case SEEK_END: 2188 error = VOP_GETATTR(vp, &vattr, cred); 2189 if (error) { 2190 goto out; 2191 } 2192 newoff = SCARG(uap, offset) + vattr.va_size; 2193 break; 2194 case SEEK_SET: 2195 newoff = SCARG(uap, offset); 2196 break; 2197 default: 2198 error = EINVAL; 2199 goto out; 2200 } 2201 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2202 *(off_t *)retval = fp->f_offset = newoff; 2203 } 2204 out: 2205 fd_putfile(fd); 2206 return (error); 2207 } 2208 2209 /* 2210 * Positional read system call. 2211 */ 2212 int 2213 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2214 { 2215 /* { 2216 syscallarg(int) fd; 2217 syscallarg(void *) buf; 2218 syscallarg(size_t) nbyte; 2219 syscallarg(off_t) offset; 2220 } */ 2221 file_t *fp; 2222 struct vnode *vp; 2223 off_t offset; 2224 int error, fd = SCARG(uap, fd); 2225 2226 if ((fp = fd_getfile(fd)) == NULL) 2227 return (EBADF); 2228 2229 if ((fp->f_flag & FREAD) == 0) { 2230 fd_putfile(fd); 2231 return (EBADF); 2232 } 2233 2234 vp = fp->f_data; 2235 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2236 error = ESPIPE; 2237 goto out; 2238 } 2239 2240 offset = SCARG(uap, offset); 2241 2242 /* 2243 * XXX This works because no file systems actually 2244 * XXX take any action on the seek operation. 2245 */ 2246 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2247 goto out; 2248 2249 /* dofileread() will unuse the descriptor for us */ 2250 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2251 &offset, 0, retval)); 2252 2253 out: 2254 fd_putfile(fd); 2255 return (error); 2256 } 2257 2258 /* 2259 * Positional scatter read system call. 2260 */ 2261 int 2262 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2263 { 2264 /* { 2265 syscallarg(int) fd; 2266 syscallarg(const struct iovec *) iovp; 2267 syscallarg(int) iovcnt; 2268 syscallarg(off_t) offset; 2269 } */ 2270 off_t offset = SCARG(uap, offset); 2271 2272 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2273 SCARG(uap, iovcnt), &offset, 0, retval); 2274 } 2275 2276 /* 2277 * Positional write system call. 2278 */ 2279 int 2280 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2281 { 2282 /* { 2283 syscallarg(int) fd; 2284 syscallarg(const void *) buf; 2285 syscallarg(size_t) nbyte; 2286 syscallarg(off_t) offset; 2287 } */ 2288 file_t *fp; 2289 struct vnode *vp; 2290 off_t offset; 2291 int error, fd = SCARG(uap, fd); 2292 2293 if ((fp = fd_getfile(fd)) == NULL) 2294 return (EBADF); 2295 2296 if ((fp->f_flag & FWRITE) == 0) { 2297 fd_putfile(fd); 2298 return (EBADF); 2299 } 2300 2301 vp = fp->f_data; 2302 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2303 error = ESPIPE; 2304 goto out; 2305 } 2306 2307 offset = SCARG(uap, offset); 2308 2309 /* 2310 * XXX This works because no file systems actually 2311 * XXX take any action on the seek operation. 2312 */ 2313 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2314 goto out; 2315 2316 /* dofilewrite() will unuse the descriptor for us */ 2317 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2318 &offset, 0, retval)); 2319 2320 out: 2321 fd_putfile(fd); 2322 return (error); 2323 } 2324 2325 /* 2326 * Positional gather write system call. 2327 */ 2328 int 2329 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2330 { 2331 /* { 2332 syscallarg(int) fd; 2333 syscallarg(const struct iovec *) iovp; 2334 syscallarg(int) iovcnt; 2335 syscallarg(off_t) offset; 2336 } */ 2337 off_t offset = SCARG(uap, offset); 2338 2339 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2340 SCARG(uap, iovcnt), &offset, 0, retval); 2341 } 2342 2343 /* 2344 * Check access permissions. 2345 */ 2346 int 2347 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2348 { 2349 /* { 2350 syscallarg(const char *) path; 2351 syscallarg(int) flags; 2352 } */ 2353 kauth_cred_t cred; 2354 struct vnode *vp; 2355 int error, flags; 2356 struct nameidata nd; 2357 2358 cred = kauth_cred_dup(l->l_cred); 2359 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2360 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2361 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2362 SCARG(uap, path)); 2363 /* Override default credentials */ 2364 nd.ni_cnd.cn_cred = cred; 2365 if ((error = namei(&nd)) != 0) 2366 goto out; 2367 vp = nd.ni_vp; 2368 2369 /* Flags == 0 means only check for existence. */ 2370 if (SCARG(uap, flags)) { 2371 flags = 0; 2372 if (SCARG(uap, flags) & R_OK) 2373 flags |= VREAD; 2374 if (SCARG(uap, flags) & W_OK) 2375 flags |= VWRITE; 2376 if (SCARG(uap, flags) & X_OK) 2377 flags |= VEXEC; 2378 2379 error = VOP_ACCESS(vp, flags, cred); 2380 if (!error && (flags & VWRITE)) 2381 error = vn_writechk(vp); 2382 } 2383 vput(vp); 2384 out: 2385 kauth_cred_free(cred); 2386 return (error); 2387 } 2388 2389 /* 2390 * Common code for all sys_stat functions, including compat versions. 2391 */ 2392 int 2393 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb) 2394 { 2395 int error; 2396 struct nameidata nd; 2397 2398 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2399 UIO_USERSPACE, path); 2400 error = namei(&nd); 2401 if (error != 0) 2402 return error; 2403 error = vn_stat(nd.ni_vp, sb); 2404 vput(nd.ni_vp); 2405 return error; 2406 } 2407 2408 /* 2409 * Get file status; this version follows links. 2410 */ 2411 /* ARGSUSED */ 2412 int 2413 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 2414 { 2415 /* { 2416 syscallarg(const char *) path; 2417 syscallarg(struct stat *) ub; 2418 } */ 2419 struct stat sb; 2420 int error; 2421 2422 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2423 if (error) 2424 return error; 2425 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2426 } 2427 2428 /* 2429 * Get file status; this version does not follow links. 2430 */ 2431 /* ARGSUSED */ 2432 int 2433 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 2434 { 2435 /* { 2436 syscallarg(const char *) path; 2437 syscallarg(struct stat *) ub; 2438 } */ 2439 struct stat sb; 2440 int error; 2441 2442 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2443 if (error) 2444 return error; 2445 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2446 } 2447 2448 /* 2449 * Get configurable pathname variables. 2450 */ 2451 /* ARGSUSED */ 2452 int 2453 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2454 { 2455 /* { 2456 syscallarg(const char *) path; 2457 syscallarg(int) name; 2458 } */ 2459 int error; 2460 struct nameidata nd; 2461 2462 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2463 SCARG(uap, path)); 2464 if ((error = namei(&nd)) != 0) 2465 return (error); 2466 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2467 vput(nd.ni_vp); 2468 return (error); 2469 } 2470 2471 /* 2472 * Return target name of a symbolic link. 2473 */ 2474 /* ARGSUSED */ 2475 int 2476 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2477 { 2478 /* { 2479 syscallarg(const char *) path; 2480 syscallarg(char *) buf; 2481 syscallarg(size_t) count; 2482 } */ 2483 struct vnode *vp; 2484 struct iovec aiov; 2485 struct uio auio; 2486 int error; 2487 struct nameidata nd; 2488 2489 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2490 SCARG(uap, path)); 2491 if ((error = namei(&nd)) != 0) 2492 return (error); 2493 vp = nd.ni_vp; 2494 if (vp->v_type != VLNK) 2495 error = EINVAL; 2496 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2497 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2498 aiov.iov_base = SCARG(uap, buf); 2499 aiov.iov_len = SCARG(uap, count); 2500 auio.uio_iov = &aiov; 2501 auio.uio_iovcnt = 1; 2502 auio.uio_offset = 0; 2503 auio.uio_rw = UIO_READ; 2504 KASSERT(l == curlwp); 2505 auio.uio_vmspace = l->l_proc->p_vmspace; 2506 auio.uio_resid = SCARG(uap, count); 2507 error = VOP_READLINK(vp, &auio, l->l_cred); 2508 } 2509 vput(vp); 2510 *retval = SCARG(uap, count) - auio.uio_resid; 2511 return (error); 2512 } 2513 2514 /* 2515 * Change flags of a file given a path name. 2516 */ 2517 /* ARGSUSED */ 2518 int 2519 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2520 { 2521 /* { 2522 syscallarg(const char *) path; 2523 syscallarg(u_long) flags; 2524 } */ 2525 struct vnode *vp; 2526 int error; 2527 struct nameidata nd; 2528 2529 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2530 SCARG(uap, path)); 2531 if ((error = namei(&nd)) != 0) 2532 return (error); 2533 vp = nd.ni_vp; 2534 error = change_flags(vp, SCARG(uap, flags), l); 2535 vput(vp); 2536 return (error); 2537 } 2538 2539 /* 2540 * Change flags of a file given a file descriptor. 2541 */ 2542 /* ARGSUSED */ 2543 int 2544 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2545 { 2546 /* { 2547 syscallarg(int) fd; 2548 syscallarg(u_long) flags; 2549 } */ 2550 struct vnode *vp; 2551 file_t *fp; 2552 int error; 2553 2554 /* fd_getvnode() will use the descriptor for us */ 2555 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2556 return (error); 2557 vp = fp->f_data; 2558 error = change_flags(vp, SCARG(uap, flags), l); 2559 VOP_UNLOCK(vp, 0); 2560 fd_putfile(SCARG(uap, fd)); 2561 return (error); 2562 } 2563 2564 /* 2565 * Change flags of a file given a path name; this version does 2566 * not follow links. 2567 */ 2568 int 2569 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2570 { 2571 /* { 2572 syscallarg(const char *) path; 2573 syscallarg(u_long) flags; 2574 } */ 2575 struct vnode *vp; 2576 int error; 2577 struct nameidata nd; 2578 2579 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2580 SCARG(uap, path)); 2581 if ((error = namei(&nd)) != 0) 2582 return (error); 2583 vp = nd.ni_vp; 2584 error = change_flags(vp, SCARG(uap, flags), l); 2585 vput(vp); 2586 return (error); 2587 } 2588 2589 /* 2590 * Common routine to change flags of a file. 2591 */ 2592 int 2593 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2594 { 2595 struct vattr vattr; 2596 int error; 2597 2598 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2599 /* 2600 * Non-superusers cannot change the flags on devices, even if they 2601 * own them. 2602 */ 2603 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2604 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2605 goto out; 2606 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2607 error = EINVAL; 2608 goto out; 2609 } 2610 } 2611 VATTR_NULL(&vattr); 2612 vattr.va_flags = flags; 2613 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2614 out: 2615 return (error); 2616 } 2617 2618 /* 2619 * Change mode of a file given path name; this version follows links. 2620 */ 2621 /* ARGSUSED */ 2622 int 2623 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2624 { 2625 /* { 2626 syscallarg(const char *) path; 2627 syscallarg(int) mode; 2628 } */ 2629 int error; 2630 struct nameidata nd; 2631 2632 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2633 SCARG(uap, path)); 2634 if ((error = namei(&nd)) != 0) 2635 return (error); 2636 2637 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2638 2639 vrele(nd.ni_vp); 2640 return (error); 2641 } 2642 2643 /* 2644 * Change mode of a file given a file descriptor. 2645 */ 2646 /* ARGSUSED */ 2647 int 2648 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2649 { 2650 /* { 2651 syscallarg(int) fd; 2652 syscallarg(int) mode; 2653 } */ 2654 file_t *fp; 2655 int error; 2656 2657 /* fd_getvnode() will use the descriptor for us */ 2658 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2659 return (error); 2660 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2661 fd_putfile(SCARG(uap, fd)); 2662 return (error); 2663 } 2664 2665 /* 2666 * Change mode of a file given path name; this version does not follow links. 2667 */ 2668 /* ARGSUSED */ 2669 int 2670 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2671 { 2672 /* { 2673 syscallarg(const char *) path; 2674 syscallarg(int) mode; 2675 } */ 2676 int error; 2677 struct nameidata nd; 2678 2679 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2680 SCARG(uap, path)); 2681 if ((error = namei(&nd)) != 0) 2682 return (error); 2683 2684 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2685 2686 vrele(nd.ni_vp); 2687 return (error); 2688 } 2689 2690 /* 2691 * Common routine to set mode given a vnode. 2692 */ 2693 static int 2694 change_mode(struct vnode *vp, int mode, struct lwp *l) 2695 { 2696 struct vattr vattr; 2697 int error; 2698 2699 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2700 VATTR_NULL(&vattr); 2701 vattr.va_mode = mode & ALLPERMS; 2702 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2703 VOP_UNLOCK(vp, 0); 2704 return (error); 2705 } 2706 2707 /* 2708 * Set ownership given a path name; this version follows links. 2709 */ 2710 /* ARGSUSED */ 2711 int 2712 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2713 { 2714 /* { 2715 syscallarg(const char *) path; 2716 syscallarg(uid_t) uid; 2717 syscallarg(gid_t) gid; 2718 } */ 2719 int error; 2720 struct nameidata nd; 2721 2722 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2723 SCARG(uap, path)); 2724 if ((error = namei(&nd)) != 0) 2725 return (error); 2726 2727 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2728 2729 vrele(nd.ni_vp); 2730 return (error); 2731 } 2732 2733 /* 2734 * Set ownership given a path name; this version follows links. 2735 * Provides POSIX semantics. 2736 */ 2737 /* ARGSUSED */ 2738 int 2739 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2740 { 2741 /* { 2742 syscallarg(const char *) path; 2743 syscallarg(uid_t) uid; 2744 syscallarg(gid_t) gid; 2745 } */ 2746 int error; 2747 struct nameidata nd; 2748 2749 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2750 SCARG(uap, path)); 2751 if ((error = namei(&nd)) != 0) 2752 return (error); 2753 2754 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2755 2756 vrele(nd.ni_vp); 2757 return (error); 2758 } 2759 2760 /* 2761 * Set ownership given a file descriptor. 2762 */ 2763 /* ARGSUSED */ 2764 int 2765 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2766 { 2767 /* { 2768 syscallarg(int) fd; 2769 syscallarg(uid_t) uid; 2770 syscallarg(gid_t) gid; 2771 } */ 2772 int error; 2773 file_t *fp; 2774 2775 /* fd_getvnode() will use the descriptor for us */ 2776 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2777 return (error); 2778 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2779 l, 0); 2780 fd_putfile(SCARG(uap, fd)); 2781 return (error); 2782 } 2783 2784 /* 2785 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2786 */ 2787 /* ARGSUSED */ 2788 int 2789 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2790 { 2791 /* { 2792 syscallarg(int) fd; 2793 syscallarg(uid_t) uid; 2794 syscallarg(gid_t) gid; 2795 } */ 2796 int error; 2797 file_t *fp; 2798 2799 /* fd_getvnode() will use the descriptor for us */ 2800 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2801 return (error); 2802 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2803 l, 1); 2804 fd_putfile(SCARG(uap, fd)); 2805 return (error); 2806 } 2807 2808 /* 2809 * Set ownership given a path name; this version does not follow links. 2810 */ 2811 /* ARGSUSED */ 2812 int 2813 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2814 { 2815 /* { 2816 syscallarg(const char *) path; 2817 syscallarg(uid_t) uid; 2818 syscallarg(gid_t) gid; 2819 } */ 2820 int error; 2821 struct nameidata nd; 2822 2823 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2824 SCARG(uap, path)); 2825 if ((error = namei(&nd)) != 0) 2826 return (error); 2827 2828 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2829 2830 vrele(nd.ni_vp); 2831 return (error); 2832 } 2833 2834 /* 2835 * Set ownership given a path name; this version does not follow links. 2836 * Provides POSIX/XPG semantics. 2837 */ 2838 /* ARGSUSED */ 2839 int 2840 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2841 { 2842 /* { 2843 syscallarg(const char *) path; 2844 syscallarg(uid_t) uid; 2845 syscallarg(gid_t) gid; 2846 } */ 2847 int error; 2848 struct nameidata nd; 2849 2850 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2851 SCARG(uap, path)); 2852 if ((error = namei(&nd)) != 0) 2853 return (error); 2854 2855 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2856 2857 vrele(nd.ni_vp); 2858 return (error); 2859 } 2860 2861 /* 2862 * Common routine to set ownership given a vnode. 2863 */ 2864 static int 2865 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2866 int posix_semantics) 2867 { 2868 struct vattr vattr; 2869 mode_t newmode; 2870 int error; 2871 2872 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2873 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2874 goto out; 2875 2876 #define CHANGED(x) ((int)(x) != -1) 2877 newmode = vattr.va_mode; 2878 if (posix_semantics) { 2879 /* 2880 * POSIX/XPG semantics: if the caller is not the super-user, 2881 * clear set-user-id and set-group-id bits. Both POSIX and 2882 * the XPG consider the behaviour for calls by the super-user 2883 * implementation-defined; we leave the set-user-id and set- 2884 * group-id settings intact in that case. 2885 */ 2886 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2887 NULL) != 0) 2888 newmode &= ~(S_ISUID | S_ISGID); 2889 } else { 2890 /* 2891 * NetBSD semantics: when changing owner and/or group, 2892 * clear the respective bit(s). 2893 */ 2894 if (CHANGED(uid)) 2895 newmode &= ~S_ISUID; 2896 if (CHANGED(gid)) 2897 newmode &= ~S_ISGID; 2898 } 2899 /* Update va_mode iff altered. */ 2900 if (vattr.va_mode == newmode) 2901 newmode = VNOVAL; 2902 2903 VATTR_NULL(&vattr); 2904 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2905 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2906 vattr.va_mode = newmode; 2907 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2908 #undef CHANGED 2909 2910 out: 2911 VOP_UNLOCK(vp, 0); 2912 return (error); 2913 } 2914 2915 /* 2916 * Set the access and modification times given a path name; this 2917 * version follows links. 2918 */ 2919 /* ARGSUSED */ 2920 int 2921 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 2922 register_t *retval) 2923 { 2924 /* { 2925 syscallarg(const char *) path; 2926 syscallarg(const struct timeval *) tptr; 2927 } */ 2928 2929 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2930 SCARG(uap, tptr), UIO_USERSPACE); 2931 } 2932 2933 /* 2934 * Set the access and modification times given a file descriptor. 2935 */ 2936 /* ARGSUSED */ 2937 int 2938 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 2939 register_t *retval) 2940 { 2941 /* { 2942 syscallarg(int) fd; 2943 syscallarg(const struct timeval *) tptr; 2944 } */ 2945 int error; 2946 file_t *fp; 2947 2948 /* fd_getvnode() will use the descriptor for us */ 2949 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2950 return (error); 2951 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2952 UIO_USERSPACE); 2953 fd_putfile(SCARG(uap, fd)); 2954 return (error); 2955 } 2956 2957 /* 2958 * Set the access and modification times given a path name; this 2959 * version does not follow links. 2960 */ 2961 int 2962 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 2963 register_t *retval) 2964 { 2965 /* { 2966 syscallarg(const char *) path; 2967 syscallarg(const struct timeval *) tptr; 2968 } */ 2969 2970 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2971 SCARG(uap, tptr), UIO_USERSPACE); 2972 } 2973 2974 /* 2975 * Common routine to set access and modification times given a vnode. 2976 */ 2977 int 2978 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2979 const struct timeval *tptr, enum uio_seg seg) 2980 { 2981 struct vattr vattr; 2982 struct nameidata nd; 2983 int error; 2984 2985 VATTR_NULL(&vattr); 2986 if (tptr == NULL) { 2987 nanotime(&vattr.va_atime); 2988 vattr.va_mtime = vattr.va_atime; 2989 vattr.va_vaflags |= VA_UTIMES_NULL; 2990 } else { 2991 struct timeval tv[2]; 2992 2993 if (seg != UIO_SYSSPACE) { 2994 error = copyin(tptr, &tv, sizeof (tv)); 2995 if (error != 0) 2996 return error; 2997 tptr = tv; 2998 } 2999 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3000 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3001 } 3002 3003 if (vp == NULL) { 3004 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3005 if ((error = namei(&nd)) != 0) 3006 return (error); 3007 vp = nd.ni_vp; 3008 } else 3009 nd.ni_vp = NULL; 3010 3011 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3012 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3013 VOP_UNLOCK(vp, 0); 3014 3015 if (nd.ni_vp != NULL) 3016 vrele(nd.ni_vp); 3017 3018 return (error); 3019 } 3020 3021 /* 3022 * Truncate a file given its path name. 3023 */ 3024 /* ARGSUSED */ 3025 int 3026 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3027 { 3028 /* { 3029 syscallarg(const char *) path; 3030 syscallarg(int) pad; 3031 syscallarg(off_t) length; 3032 } */ 3033 struct vnode *vp; 3034 struct vattr vattr; 3035 int error; 3036 struct nameidata nd; 3037 3038 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3039 SCARG(uap, path)); 3040 if ((error = namei(&nd)) != 0) 3041 return (error); 3042 vp = nd.ni_vp; 3043 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3044 if (vp->v_type == VDIR) 3045 error = EISDIR; 3046 else if ((error = vn_writechk(vp)) == 0 && 3047 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3048 VATTR_NULL(&vattr); 3049 vattr.va_size = SCARG(uap, length); 3050 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3051 } 3052 vput(vp); 3053 return (error); 3054 } 3055 3056 /* 3057 * Truncate a file given a file descriptor. 3058 */ 3059 /* ARGSUSED */ 3060 int 3061 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3062 { 3063 /* { 3064 syscallarg(int) fd; 3065 syscallarg(int) pad; 3066 syscallarg(off_t) length; 3067 } */ 3068 struct vattr vattr; 3069 struct vnode *vp; 3070 file_t *fp; 3071 int error; 3072 3073 /* fd_getvnode() will use the descriptor for us */ 3074 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3075 return (error); 3076 if ((fp->f_flag & FWRITE) == 0) { 3077 error = EINVAL; 3078 goto out; 3079 } 3080 vp = fp->f_data; 3081 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3082 if (vp->v_type == VDIR) 3083 error = EISDIR; 3084 else if ((error = vn_writechk(vp)) == 0) { 3085 VATTR_NULL(&vattr); 3086 vattr.va_size = SCARG(uap, length); 3087 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3088 } 3089 VOP_UNLOCK(vp, 0); 3090 out: 3091 fd_putfile(SCARG(uap, fd)); 3092 return (error); 3093 } 3094 3095 /* 3096 * Sync an open file. 3097 */ 3098 /* ARGSUSED */ 3099 int 3100 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3101 { 3102 /* { 3103 syscallarg(int) fd; 3104 } */ 3105 struct vnode *vp; 3106 file_t *fp; 3107 int error; 3108 3109 /* fd_getvnode() will use the descriptor for us */ 3110 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3111 return (error); 3112 vp = fp->f_data; 3113 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3114 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3115 if (error == 0 && bioopsp != NULL && 3116 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3117 (*bioopsp->io_fsync)(vp, 0); 3118 VOP_UNLOCK(vp, 0); 3119 fd_putfile(SCARG(uap, fd)); 3120 return (error); 3121 } 3122 3123 /* 3124 * Sync a range of file data. API modeled after that found in AIX. 3125 * 3126 * FDATASYNC indicates that we need only save enough metadata to be able 3127 * to re-read the written data. Note we duplicate AIX's requirement that 3128 * the file be open for writing. 3129 */ 3130 /* ARGSUSED */ 3131 int 3132 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3133 { 3134 /* { 3135 syscallarg(int) fd; 3136 syscallarg(int) flags; 3137 syscallarg(off_t) start; 3138 syscallarg(off_t) length; 3139 } */ 3140 struct vnode *vp; 3141 file_t *fp; 3142 int flags, nflags; 3143 off_t s, e, len; 3144 int error; 3145 3146 /* fd_getvnode() will use the descriptor for us */ 3147 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3148 return (error); 3149 3150 if ((fp->f_flag & FWRITE) == 0) { 3151 error = EBADF; 3152 goto out; 3153 } 3154 3155 flags = SCARG(uap, flags); 3156 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3157 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3158 error = EINVAL; 3159 goto out; 3160 } 3161 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3162 if (flags & FDATASYNC) 3163 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3164 else 3165 nflags = FSYNC_WAIT; 3166 if (flags & FDISKSYNC) 3167 nflags |= FSYNC_CACHE; 3168 3169 len = SCARG(uap, length); 3170 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3171 if (len) { 3172 s = SCARG(uap, start); 3173 e = s + len; 3174 if (e < s) { 3175 error = EINVAL; 3176 goto out; 3177 } 3178 } else { 3179 e = 0; 3180 s = 0; 3181 } 3182 3183 vp = fp->f_data; 3184 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3185 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3186 3187 if (error == 0 && bioopsp != NULL && 3188 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3189 (*bioopsp->io_fsync)(vp, nflags); 3190 3191 VOP_UNLOCK(vp, 0); 3192 out: 3193 fd_putfile(SCARG(uap, fd)); 3194 return (error); 3195 } 3196 3197 /* 3198 * Sync the data of an open file. 3199 */ 3200 /* ARGSUSED */ 3201 int 3202 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3203 { 3204 /* { 3205 syscallarg(int) fd; 3206 } */ 3207 struct vnode *vp; 3208 file_t *fp; 3209 int error; 3210 3211 /* fd_getvnode() will use the descriptor for us */ 3212 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3213 return (error); 3214 if ((fp->f_flag & FWRITE) == 0) { 3215 fd_putfile(SCARG(uap, fd)); 3216 return (EBADF); 3217 } 3218 vp = fp->f_data; 3219 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3220 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3221 VOP_UNLOCK(vp, 0); 3222 fd_putfile(SCARG(uap, fd)); 3223 return (error); 3224 } 3225 3226 /* 3227 * Rename files, (standard) BSD semantics frontend. 3228 */ 3229 /* ARGSUSED */ 3230 int 3231 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3232 { 3233 /* { 3234 syscallarg(const char *) from; 3235 syscallarg(const char *) to; 3236 } */ 3237 3238 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3239 } 3240 3241 /* 3242 * Rename files, POSIX semantics frontend. 3243 */ 3244 /* ARGSUSED */ 3245 int 3246 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3247 { 3248 /* { 3249 syscallarg(const char *) from; 3250 syscallarg(const char *) to; 3251 } */ 3252 3253 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3254 } 3255 3256 /* 3257 * Rename files. Source and destination must either both be directories, 3258 * or both not be directories. If target is a directory, it must be empty. 3259 * If `from' and `to' refer to the same object, the value of the `retain' 3260 * argument is used to determine whether `from' will be 3261 * 3262 * (retain == 0) deleted unless `from' and `to' refer to the same 3263 * object in the file system's name space (BSD). 3264 * (retain == 1) always retained (POSIX). 3265 */ 3266 int 3267 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3268 { 3269 struct vnode *tvp, *fvp, *tdvp; 3270 struct nameidata fromnd, tond; 3271 struct mount *fs; 3272 struct lwp *l = curlwp; 3273 struct proc *p; 3274 uint32_t saveflag; 3275 int error; 3276 3277 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3278 seg, from); 3279 if ((error = namei(&fromnd)) != 0) 3280 return (error); 3281 if (fromnd.ni_dvp != fromnd.ni_vp) 3282 VOP_UNLOCK(fromnd.ni_dvp, 0); 3283 fvp = fromnd.ni_vp; 3284 3285 fs = fvp->v_mount; 3286 error = VFS_RENAMELOCK_ENTER(fs); 3287 if (error) { 3288 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3289 vrele(fromnd.ni_dvp); 3290 vrele(fvp); 3291 goto out1; 3292 } 3293 3294 /* 3295 * close, partially, yet another race - ideally we should only 3296 * go as far as getting fromnd.ni_dvp before getting the per-fs 3297 * lock, and then continue to get fromnd.ni_vp, but we can't do 3298 * that with namei as it stands. 3299 * 3300 * This still won't prevent rmdir from nuking fromnd.ni_vp 3301 * under us. The real fix is to get the locks in the right 3302 * order and do the lookups in the right places, but that's a 3303 * major rototill. 3304 * 3305 * Preserve the SAVESTART in cn_flags, because who knows what 3306 * might happen if we don't. 3307 * 3308 * Note: this logic (as well as this whole function) is cloned 3309 * in nfs_serv.c. Proceed accordingly. 3310 */ 3311 vrele(fvp); 3312 if ((fromnd.ni_cnd.cn_namelen == 1 && 3313 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3314 (fromnd.ni_cnd.cn_namelen == 2 && 3315 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3316 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3317 error = EINVAL; 3318 VFS_RENAMELOCK_EXIT(fs); 3319 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3320 vrele(fromnd.ni_dvp); 3321 goto out1; 3322 } 3323 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3324 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3325 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3326 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3327 fromnd.ni_cnd.cn_flags |= saveflag; 3328 if (error) { 3329 VOP_UNLOCK(fromnd.ni_dvp, 0); 3330 VFS_RENAMELOCK_EXIT(fs); 3331 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3332 vrele(fromnd.ni_dvp); 3333 goto out1; 3334 } 3335 VOP_UNLOCK(fromnd.ni_vp, 0); 3336 if (fromnd.ni_dvp != fromnd.ni_vp) 3337 VOP_UNLOCK(fromnd.ni_dvp, 0); 3338 fvp = fromnd.ni_vp; 3339 3340 NDINIT(&tond, RENAME, 3341 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3342 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3343 seg, to); 3344 if ((error = namei(&tond)) != 0) { 3345 VFS_RENAMELOCK_EXIT(fs); 3346 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3347 vrele(fromnd.ni_dvp); 3348 vrele(fvp); 3349 goto out1; 3350 } 3351 tdvp = tond.ni_dvp; 3352 tvp = tond.ni_vp; 3353 3354 if (tvp != NULL) { 3355 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3356 error = ENOTDIR; 3357 goto out; 3358 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3359 error = EISDIR; 3360 goto out; 3361 } 3362 } 3363 3364 if (fvp == tdvp) 3365 error = EINVAL; 3366 3367 /* 3368 * Source and destination refer to the same object. 3369 */ 3370 if (fvp == tvp) { 3371 if (retain) 3372 error = -1; 3373 else if (fromnd.ni_dvp == tdvp && 3374 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3375 !memcmp(fromnd.ni_cnd.cn_nameptr, 3376 tond.ni_cnd.cn_nameptr, 3377 fromnd.ni_cnd.cn_namelen)) 3378 error = -1; 3379 } 3380 3381 #if NVERIEXEC > 0 3382 if (!error) { 3383 char *f1, *f2; 3384 3385 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3386 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3387 3388 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3389 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3390 3391 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3392 3393 free(f1, M_TEMP); 3394 free(f2, M_TEMP); 3395 } 3396 #endif /* NVERIEXEC > 0 */ 3397 3398 out: 3399 p = l->l_proc; 3400 if (!error) { 3401 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3402 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3403 VFS_RENAMELOCK_EXIT(fs); 3404 } else { 3405 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3406 if (tdvp == tvp) 3407 vrele(tdvp); 3408 else 3409 vput(tdvp); 3410 if (tvp) 3411 vput(tvp); 3412 VFS_RENAMELOCK_EXIT(fs); 3413 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3414 vrele(fromnd.ni_dvp); 3415 vrele(fvp); 3416 } 3417 vrele(tond.ni_startdir); 3418 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3419 out1: 3420 if (fromnd.ni_startdir) 3421 vrele(fromnd.ni_startdir); 3422 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3423 return (error == -1 ? 0 : error); 3424 } 3425 3426 /* 3427 * Make a directory file. 3428 */ 3429 /* ARGSUSED */ 3430 int 3431 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3432 { 3433 /* { 3434 syscallarg(const char *) path; 3435 syscallarg(int) mode; 3436 } */ 3437 struct proc *p = l->l_proc; 3438 struct vnode *vp; 3439 struct vattr vattr; 3440 int error; 3441 struct nameidata nd; 3442 3443 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3444 SCARG(uap, path)); 3445 if ((error = namei(&nd)) != 0) 3446 return (error); 3447 vp = nd.ni_vp; 3448 if (vp != NULL) { 3449 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3450 if (nd.ni_dvp == vp) 3451 vrele(nd.ni_dvp); 3452 else 3453 vput(nd.ni_dvp); 3454 vrele(vp); 3455 return (EEXIST); 3456 } 3457 VATTR_NULL(&vattr); 3458 vattr.va_type = VDIR; 3459 /* We will read cwdi->cwdi_cmask unlocked. */ 3460 vattr.va_mode = 3461 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3462 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3463 if (!error) 3464 vput(nd.ni_vp); 3465 return (error); 3466 } 3467 3468 /* 3469 * Remove a directory file. 3470 */ 3471 /* ARGSUSED */ 3472 int 3473 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3474 { 3475 /* { 3476 syscallarg(const char *) path; 3477 } */ 3478 struct vnode *vp; 3479 int error; 3480 struct nameidata nd; 3481 3482 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3483 SCARG(uap, path)); 3484 if ((error = namei(&nd)) != 0) 3485 return (error); 3486 vp = nd.ni_vp; 3487 if (vp->v_type != VDIR) { 3488 error = ENOTDIR; 3489 goto out; 3490 } 3491 /* 3492 * No rmdir "." please. 3493 */ 3494 if (nd.ni_dvp == vp) { 3495 error = EINVAL; 3496 goto out; 3497 } 3498 /* 3499 * The root of a mounted filesystem cannot be deleted. 3500 */ 3501 if (vp->v_vflag & VV_ROOT) { 3502 error = EBUSY; 3503 goto out; 3504 } 3505 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3506 return (error); 3507 3508 out: 3509 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3510 if (nd.ni_dvp == vp) 3511 vrele(nd.ni_dvp); 3512 else 3513 vput(nd.ni_dvp); 3514 vput(vp); 3515 return (error); 3516 } 3517 3518 /* 3519 * Read a block of directory entries in a file system independent format. 3520 */ 3521 int 3522 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3523 { 3524 /* { 3525 syscallarg(int) fd; 3526 syscallarg(char *) buf; 3527 syscallarg(size_t) count; 3528 } */ 3529 file_t *fp; 3530 int error, done; 3531 3532 /* fd_getvnode() will use the descriptor for us */ 3533 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3534 return (error); 3535 if ((fp->f_flag & FREAD) == 0) { 3536 error = EBADF; 3537 goto out; 3538 } 3539 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3540 SCARG(uap, count), &done, l, 0, 0); 3541 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3542 *retval = done; 3543 out: 3544 fd_putfile(SCARG(uap, fd)); 3545 return (error); 3546 } 3547 3548 /* 3549 * Set the mode mask for creation of filesystem nodes. 3550 */ 3551 int 3552 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3553 { 3554 /* { 3555 syscallarg(mode_t) newmask; 3556 } */ 3557 struct proc *p = l->l_proc; 3558 struct cwdinfo *cwdi; 3559 3560 /* 3561 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3562 * important is that we serialize changes to the mask. The 3563 * rw_exit() will issue a write memory barrier on our behalf, 3564 * and force the changes out to other CPUs (as it must use an 3565 * atomic operation, draining the local CPU's store buffers). 3566 */ 3567 cwdi = p->p_cwdi; 3568 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3569 *retval = cwdi->cwdi_cmask; 3570 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3571 rw_exit(&cwdi->cwdi_lock); 3572 3573 return (0); 3574 } 3575 3576 int 3577 dorevoke(struct vnode *vp, kauth_cred_t cred) 3578 { 3579 struct vattr vattr; 3580 int error; 3581 3582 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3583 return error; 3584 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3585 (error = kauth_authorize_generic(cred, 3586 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3587 VOP_REVOKE(vp, REVOKEALL); 3588 return (error); 3589 } 3590 3591 /* 3592 * Void all references to file by ripping underlying filesystem 3593 * away from vnode. 3594 */ 3595 /* ARGSUSED */ 3596 int 3597 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3598 { 3599 /* { 3600 syscallarg(const char *) path; 3601 } */ 3602 struct vnode *vp; 3603 int error; 3604 struct nameidata nd; 3605 3606 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3607 SCARG(uap, path)); 3608 if ((error = namei(&nd)) != 0) 3609 return (error); 3610 vp = nd.ni_vp; 3611 error = dorevoke(vp, l->l_cred); 3612 vrele(vp); 3613 return (error); 3614 } 3615 3616 /* 3617 * Convert a user file descriptor to a kernel file entry. 3618 */ 3619 int 3620 getvnode(int fd, file_t **fpp) 3621 { 3622 struct vnode *vp; 3623 file_t *fp; 3624 3625 if ((fp = fd_getfile(fd)) == NULL) 3626 return (EBADF); 3627 3628 if (fp->f_type != DTYPE_VNODE) { 3629 fd_putfile(fd); 3630 return (EINVAL); 3631 } 3632 3633 vp = fp->f_data; 3634 if (vp->v_type == VBAD) { 3635 fd_putfile(fd); 3636 return (EBADF); 3637 } 3638 3639 *fpp = fp; 3640 return (0); 3641 } 3642
Indexes created Fri Oct 03 02:09:57 GMT 2025