vfs_syscalls.c revision 1.333.2.3
1 /* $NetBSD: vfs_syscalls.c,v 1.333.2.3 2007/12/09 22:57:20 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.333.2.3 2007/12/09 22:57:20 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "fss.h" 46 #include "veriexec.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/kmem.h> 61 #include <sys/dirent.h> 62 #include <sys/sysctl.h> 63 #include <sys/syscallargs.h> 64 #include <sys/vfs_syscalls.h> 65 #include <sys/ktrace.h> 66 #ifdef FILEASSOC 67 #include <sys/fileassoc.h> 68 #endif /* FILEASSOC */ 69 #include <sys/verified_exec.h> 70 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #endif 79 #include <nfs/nfsproto.h> 80 #ifdef COMPAT_30 81 #include <nfs/nfs.h> 82 #include <nfs/nfs_var.h> 83 #endif 84 85 #if NFSS > 0 86 #include <dev/fssvar.h> 87 #endif 88 89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 90 91 static int change_dir(struct nameidata *, struct lwp *); 92 static int change_flags(struct vnode *, u_long, struct lwp *); 93 static int change_mode(struct vnode *, int, struct lwp *l); 94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 95 static int rename_files(const char *, const char *, struct lwp *, int); 96 97 void checkdirs(struct vnode *); 98 99 int dovfsusermount = 0; 100 101 /* 102 * Virtual File System System Calls 103 */ 104 105 /* 106 * Mount a file system. 107 */ 108 109 #if defined(COMPAT_09) || defined(COMPAT_43) 110 /* 111 * This table is used to maintain compatibility with 4.3BSD 112 * and NetBSD 0.9 mount syscalls. Note, the order is important! 113 * 114 * Do not modify this table. It should only contain filesystems 115 * supported by NetBSD 0.9 and 4.3BSD. 116 */ 117 const char * const mountcompatnames[] = { 118 NULL, /* 0 = MOUNT_NONE */ 119 MOUNT_FFS, /* 1 = MOUNT_UFS */ 120 MOUNT_NFS, /* 2 */ 121 MOUNT_MFS, /* 3 */ 122 MOUNT_MSDOS, /* 4 */ 123 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 124 MOUNT_FDESC, /* 6 */ 125 MOUNT_KERNFS, /* 7 */ 126 NULL, /* 8 = MOUNT_DEVFS */ 127 MOUNT_AFS, /* 9 */ 128 }; 129 const int nmountcompatnames = sizeof(mountcompatnames) / 130 sizeof(mountcompatnames[0]); 131 #endif /* COMPAT_09 || COMPAT_43 */ 132 133 static int 134 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 135 void *data, size_t *data_len) 136 { 137 struct mount *mp; 138 int error = 0, saved_flags; 139 140 mp = vp->v_mount; 141 saved_flags = mp->mnt_flag; 142 143 /* We can operate only on VV_ROOT nodes. */ 144 if ((vp->v_vflag & VV_ROOT) == 0) { 145 error = EINVAL; 146 goto out; 147 } 148 149 /* 150 * We only allow the filesystem to be reloaded if it 151 * is currently mounted read-only. 152 */ 153 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 154 error = EOPNOTSUPP; /* Needs translation */ 155 goto out; 156 } 157 158 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 159 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 160 if (error) 161 goto out; 162 163 if (vfs_busy(mp, LK_NOWAIT, 0)) { 164 error = EPERM; 165 goto out; 166 } 167 168 mp->mnt_flag &= ~MNT_OP_FLAGS; 169 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 170 171 /* 172 * Set the mount level flags. 173 */ 174 if (flags & MNT_RDONLY) 175 mp->mnt_flag |= MNT_RDONLY; 176 else if (mp->mnt_flag & MNT_RDONLY) 177 mp->mnt_iflag |= IMNT_WANTRDWR; 178 mp->mnt_flag &= 179 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 180 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 181 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 182 mp->mnt_flag |= flags & 183 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 184 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 185 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 186 MNT_IGNORE); 187 188 error = VFS_MOUNT(mp, path, data, data_len); 189 190 #if defined(COMPAT_30) && defined(NFSSERVER) 191 if (error && data != NULL) { 192 int error2; 193 194 /* Update failed; let's try and see if it was an 195 * export request. */ 196 error2 = nfs_update_exports_30(mp, path, data, l); 197 198 /* Only update error code if the export request was 199 * understood but some problem occurred while 200 * processing it. */ 201 if (error2 != EJUSTRETURN) 202 error = error2; 203 } 204 #endif 205 if (mp->mnt_iflag & IMNT_WANTRDWR) 206 mp->mnt_flag &= ~MNT_RDONLY; 207 if (error) 208 mp->mnt_flag = saved_flags; 209 mp->mnt_flag &= ~MNT_OP_FLAGS; 210 mp->mnt_iflag &= ~IMNT_WANTRDWR; 211 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 212 if (mp->mnt_syncer == NULL) 213 error = vfs_allocate_syncvnode(mp); 214 } else { 215 if (mp->mnt_syncer != NULL) 216 vfs_deallocate_syncvnode(mp); 217 } 218 vfs_unbusy(mp); 219 220 out: 221 return (error); 222 } 223 224 static int 225 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 226 { 227 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 228 int error; 229 230 /* Copy file-system type from userspace. */ 231 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 232 if (error) { 233 #if defined(COMPAT_09) || defined(COMPAT_43) 234 /* 235 * Historically, filesystem types were identified by numbers. 236 * If we get an integer for the filesystem type instead of a 237 * string, we check to see if it matches one of the historic 238 * filesystem types. 239 */ 240 u_long fsindex = (u_long)fstype; 241 if (fsindex >= nmountcompatnames || 242 mountcompatnames[fsindex] == NULL) 243 return ENODEV; 244 strlcpy(fstypename, mountcompatnames[fsindex], 245 sizeof(fstypename)); 246 #else 247 return error; 248 #endif 249 } 250 251 #ifdef COMPAT_10 252 /* Accept `ufs' as an alias for `ffs'. */ 253 if (strcmp(fstypename, "ufs") == 0) 254 fstypename[0] = 'f'; 255 #endif 256 257 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 258 return ENODEV; 259 return 0; 260 } 261 262 static int 263 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 264 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 265 { 266 struct mount *mp = NULL; 267 struct vnode *vp = *vpp; 268 struct vattr va; 269 int error; 270 271 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 272 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 273 if (error) 274 return error; 275 276 /* Can't make a non-dir a mount-point (from here anyway). */ 277 if (vp->v_type != VDIR) 278 return ENOTDIR; 279 280 /* 281 * If the user is not root, ensure that they own the directory 282 * onto which we are attempting to mount. 283 */ 284 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 285 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 286 (error = kauth_authorize_generic(l->l_cred, 287 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 288 return error; 289 } 290 291 if (flags & MNT_EXPORTED) 292 return EINVAL; 293 294 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 295 return error; 296 297 /* 298 * Check if a file-system is not already mounted on this vnode. 299 */ 300 if (vp->v_mountedhere != NULL) 301 return EBUSY; 302 303 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 304 305 mp->mnt_op = vfsops; 306 307 TAILQ_INIT(&mp->mnt_vnodelist); 308 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 309 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 310 (void)vfs_busy(mp, LK_NOWAIT, 0); 311 312 mp->mnt_vnodecovered = vp; 313 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 314 mp->mnt_unmounter = NULL; 315 mount_initspecific(mp); 316 317 /* 318 * The underlying file system may refuse the mount for 319 * various reasons. Allow the user to force it to happen. 320 * 321 * Set the mount level flags. 322 */ 323 mp->mnt_flag = flags & 324 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 325 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 326 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 327 MNT_IGNORE | MNT_RDONLY); 328 329 error = VFS_MOUNT(mp, path, data, data_len); 330 mp->mnt_flag &= ~MNT_OP_FLAGS; 331 332 /* 333 * Put the new filesystem on the mount list after root. 334 */ 335 cache_purge(vp); 336 if (error != 0) { 337 vp->v_mountedhere = NULL; 338 mp->mnt_op->vfs_refcount--; 339 vfs_unbusy(mp); 340 vfs_destroy(mp); 341 return error; 342 } 343 344 mp->mnt_iflag &= ~IMNT_WANTRDWR; 345 vp->v_mountedhere = mp; 346 mutex_enter(&mountlist_lock); 347 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 348 mutex_exit(&mountlist_lock); 349 vn_restorerecurse(vp, recurse); 350 VOP_UNLOCK(vp, 0); 351 checkdirs(vp); 352 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 353 error = vfs_allocate_syncvnode(mp); 354 vfs_unbusy(mp); 355 (void) VFS_STATVFS(mp, &mp->mnt_stat); 356 error = VFS_START(mp, 0); 357 if (error) 358 vrele(vp); 359 *vpp = NULL; 360 return error; 361 } 362 363 static int 364 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 365 void *data, size_t *data_len) 366 { 367 struct mount *mp; 368 int error; 369 370 /* If MNT_GETARGS is specified, it should be the only flag. */ 371 if (flags & ~MNT_GETARGS) 372 return EINVAL; 373 374 mp = vp->v_mount; 375 376 /* XXX: probably some notion of "can see" here if we want isolation. */ 377 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 378 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 379 if (error) 380 return error; 381 382 if ((vp->v_vflag & VV_ROOT) == 0) 383 return EINVAL; 384 385 if (vfs_busy(mp, LK_NOWAIT, 0)) 386 return EPERM; 387 388 mp->mnt_flag &= ~MNT_OP_FLAGS; 389 mp->mnt_flag |= MNT_GETARGS; 390 error = VFS_MOUNT(mp, path, data, data_len); 391 mp->mnt_flag &= ~MNT_OP_FLAGS; 392 393 vfs_unbusy(mp); 394 return (error); 395 } 396 397 #ifdef COMPAT_40 398 /* ARGSUSED */ 399 int 400 compat_40_sys_mount(struct lwp *l, void *v, register_t *retval) 401 { 402 struct compat_40_sys_mount_args /* { 403 syscallarg(const char *) type; 404 syscallarg(const char *) path; 405 syscallarg(int) flags; 406 syscallarg(void *) data; 407 } */ *uap = v; 408 register_t dummy; 409 410 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 411 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 412 } 413 #endif 414 415 int 416 sys___mount50(struct lwp *l, void *v, register_t *retval) 417 { 418 struct sys___mount50_args /* { 419 syscallarg(const char *) type; 420 syscallarg(const char *) path; 421 syscallarg(int) flags; 422 syscallarg(void *) data; 423 syscallarg(size_t) data_len; 424 } */ *uap = v; 425 426 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 427 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 428 SCARG(uap, data_len), retval); 429 } 430 431 int 432 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 433 const char *path, int flags, void *data, enum uio_seg data_seg, 434 size_t data_len, register_t *retval) 435 { 436 struct vnode *vp; 437 struct nameidata nd; 438 void *data_buf = data; 439 u_int recurse; 440 int error; 441 442 /* 443 * Get vnode to be covered 444 */ 445 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 446 if ((error = namei(&nd)) != 0) 447 return (error); 448 vp = nd.ni_vp; 449 450 /* 451 * A lookup in VFS_MOUNT might result in an attempt to 452 * lock this vnode again, so make the lock recursive. 453 */ 454 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 455 recurse = vn_setrecurse(vp); 456 457 if (vfsops == NULL) { 458 if (flags & (MNT_GETARGS | MNT_UPDATE)) 459 vfsops = vp->v_mount->mnt_op; 460 else { 461 /* 'type' is userspace */ 462 error = mount_get_vfsops(type, &vfsops); 463 if (error != 0) 464 goto done; 465 } 466 } 467 468 if (data != NULL && data_seg == UIO_USERSPACE) { 469 if (data_len == 0) { 470 /* No length supplied, use default for filesystem */ 471 data_len = vfsops->vfs_min_mount_data; 472 if (data_len > VFS_MAX_MOUNT_DATA) { 473 /* maybe a force loaded old LKM */ 474 error = EINVAL; 475 goto done; 476 } 477 #ifdef COMPAT_30 478 /* Hopefully a longer buffer won't make copyin() fail */ 479 if (flags & MNT_UPDATE 480 && data_len < sizeof (struct mnt_export_args30)) 481 data_len = sizeof (struct mnt_export_args30); 482 #endif 483 } 484 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 485 486 /* NFS needs the buffer even for mnt_getargs .... */ 487 error = copyin(data, data_buf, data_len); 488 if (error != 0) 489 goto done; 490 } 491 492 if (flags & MNT_GETARGS) { 493 if (data_len == 0) { 494 error = EINVAL; 495 goto done; 496 } 497 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 498 if (error != 0) 499 goto done; 500 if (data_seg == UIO_USERSPACE) 501 error = copyout(data_buf, data, data_len); 502 *retval = data_len; 503 } else if (flags & MNT_UPDATE) { 504 error = mount_update(l, vp, path, flags, data_buf, &data_len); 505 } else { 506 /* Locking is handled internally in mount_domount(). */ 507 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 508 &data_len, recurse); 509 } 510 511 done: 512 if (vp != NULL) { 513 vn_restorerecurse(vp, recurse); 514 vput(vp); 515 } 516 if (data_buf != data) 517 free(data_buf, M_TEMP); 518 return (error); 519 } 520 521 /* 522 * Scan all active processes to see if any of them have a current 523 * or root directory onto which the new filesystem has just been 524 * mounted. If so, replace them with the new mount point. 525 */ 526 void 527 checkdirs(struct vnode *olddp) 528 { 529 struct cwdinfo *cwdi; 530 struct vnode *newdp; 531 struct proc *p; 532 533 if (olddp->v_usecount == 1) 534 return; 535 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 536 panic("mount: lost mount"); 537 mutex_enter(&proclist_lock); 538 PROCLIST_FOREACH(p, &allproc) { 539 cwdi = p->p_cwdi; 540 if (!cwdi) 541 continue; 542 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 543 if (cwdi->cwdi_cdir == olddp) { 544 vrele(cwdi->cwdi_cdir); 545 VREF(newdp); 546 cwdi->cwdi_cdir = newdp; 547 } 548 if (cwdi->cwdi_rdir == olddp) { 549 vrele(cwdi->cwdi_rdir); 550 VREF(newdp); 551 cwdi->cwdi_rdir = newdp; 552 } 553 rw_exit(&cwdi->cwdi_lock); 554 } 555 mutex_exit(&proclist_lock); 556 if (rootvnode == olddp) { 557 vrele(rootvnode); 558 VREF(newdp); 559 rootvnode = newdp; 560 } 561 vput(newdp); 562 } 563 564 /* 565 * Unmount a file system. 566 * 567 * Note: unmount takes a path to the vnode mounted on as argument, 568 * not special file (as before). 569 */ 570 /* ARGSUSED */ 571 int 572 sys_unmount(struct lwp *l, void *v, register_t *retval) 573 { 574 struct sys_unmount_args /* { 575 syscallarg(const char *) path; 576 syscallarg(int) flags; 577 } */ *uap = v; 578 struct vnode *vp; 579 struct mount *mp; 580 int error; 581 struct nameidata nd; 582 583 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 584 SCARG(uap, path), l); 585 if ((error = namei(&nd)) != 0) 586 return (error); 587 vp = nd.ni_vp; 588 mp = vp->v_mount; 589 590 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 591 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 592 if (error) { 593 vput(vp); 594 return (error); 595 } 596 597 /* 598 * Don't allow unmounting the root file system. 599 */ 600 if (mp->mnt_flag & MNT_ROOTFS) { 601 vput(vp); 602 return (EINVAL); 603 } 604 605 /* 606 * Must be the root of the filesystem 607 */ 608 if ((vp->v_vflag & VV_ROOT) == 0) { 609 vput(vp); 610 return (EINVAL); 611 } 612 vput(vp); 613 614 /* 615 * XXX Freeze syncer. Must do this before locking the 616 * mount point. See dounmount() for details. 617 */ 618 mutex_enter(&syncer_mutex); 619 620 if (vfs_busy(mp, 0, 0)) { 621 mutex_exit(&syncer_mutex); 622 return (EBUSY); 623 } 624 625 return (dounmount(mp, SCARG(uap, flags), l)); 626 } 627 628 /* 629 * Do the actual file system unmount. File system is assumed to have been 630 * marked busy by the caller. 631 */ 632 int 633 dounmount(struct mount *mp, int flags, struct lwp *l) 634 { 635 struct vnode *coveredvp; 636 int error; 637 int async; 638 int used_syncer; 639 640 #if NVERIEXEC > 0 641 error = veriexec_unmountchk(mp); 642 if (error) 643 return (error); 644 #endif /* NVERIEXEC > 0 */ 645 646 mutex_enter(&mountlist_lock); 647 vfs_unbusy(mp); 648 used_syncer = (mp->mnt_syncer != NULL); 649 650 /* 651 * XXX Syncer must be frozen when we get here. This should really 652 * be done on a per-mountpoint basis, but especially the softdep 653 * code possibly called from the syncer doesn't exactly work on a 654 * per-mountpoint basis, so the softdep code would become a maze 655 * of vfs_busy() calls. 656 * 657 * The caller of dounmount() must acquire syncer_mutex because 658 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 659 * order, and we must preserve that order to avoid deadlock. 660 * 661 * So, if the file system did not use the syncer, now is 662 * the time to release the syncer_mutex. 663 */ 664 if (used_syncer == 0) 665 mutex_exit(&syncer_mutex); 666 667 mp->mnt_iflag |= IMNT_UNMOUNT; 668 mp->mnt_unmounter = l; 669 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 670 671 async = mp->mnt_flag & MNT_ASYNC; 672 mp->mnt_flag &= ~MNT_ASYNC; 673 cache_purgevfs(mp); /* remove cache entries for this file sys */ 674 if (mp->mnt_syncer != NULL) 675 vfs_deallocate_syncvnode(mp); 676 error = 0; 677 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 678 #if NFSS > 0 679 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 680 #endif 681 if (error == 0) 682 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 683 } 684 if (error == 0 || (flags & MNT_FORCE)) 685 error = VFS_UNMOUNT(mp, flags); 686 if (error) { 687 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 688 (void) vfs_allocate_syncvnode(mp); 689 mutex_enter(&mountlist_lock); 690 mp->mnt_iflag &= ~IMNT_UNMOUNT; 691 mp->mnt_unmounter = NULL; 692 mp->mnt_flag |= async; 693 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 694 &mountlist_lock); 695 if (used_syncer) 696 mutex_exit(&syncer_mutex); 697 mutex_enter(&mp->mnt_mutex); 698 while (mp->mnt_wcnt > 0) { 699 wakeup(mp); 700 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 701 0, &mp->mnt_mutex); 702 } 703 mutex_exit(&mp->mnt_mutex); 704 return (error); 705 } 706 vfs_scrubvnlist(mp); 707 mutex_enter(&mountlist_lock); 708 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 709 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 710 coveredvp->v_mountedhere = NULL; 711 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 712 panic("unmount: dangling vnode"); 713 mp->mnt_iflag |= IMNT_GONE; 714 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 715 if (coveredvp != NULLVP) 716 vrele(coveredvp); 717 if (used_syncer) 718 mutex_exit(&syncer_mutex); 719 mutex_enter(&mp->mnt_mutex); 720 while (mp->mnt_wcnt > 0) { 721 wakeup(mp); 722 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 723 } 724 mutex_exit(&mp->mnt_mutex); 725 vfs_hooks_unmount(mp); 726 vfs_delref(mp->mnt_op); 727 vfs_destroy(mp); 728 return (0); 729 } 730 731 /* 732 * Sync each mounted filesystem. 733 */ 734 #ifdef DEBUG 735 int syncprt = 0; 736 struct ctldebug debug0 = { "syncprt", &syncprt }; 737 #endif 738 739 /* ARGSUSED */ 740 int 741 sys_sync(struct lwp *l, void *v, register_t *retval) 742 { 743 struct mount *mp, *nmp; 744 int asyncflag; 745 746 if (l == NULL) 747 l = &lwp0; 748 749 mutex_enter(&mountlist_lock); 750 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 751 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 752 nmp = mp->mnt_list.cqe_prev; 753 continue; 754 } 755 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 756 asyncflag = mp->mnt_flag & MNT_ASYNC; 757 mp->mnt_flag &= ~MNT_ASYNC; 758 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 759 if (asyncflag) 760 mp->mnt_flag |= MNT_ASYNC; 761 } 762 mutex_enter(&mountlist_lock); 763 nmp = mp->mnt_list.cqe_prev; 764 vfs_unbusy(mp); 765 766 } 767 mutex_exit(&mountlist_lock); 768 #ifdef DEBUG 769 if (syncprt) 770 vfs_bufstats(); 771 #endif /* DEBUG */ 772 return (0); 773 } 774 775 /* 776 * Change filesystem quotas. 777 */ 778 /* ARGSUSED */ 779 int 780 sys_quotactl(struct lwp *l, void *v, register_t *retval) 781 { 782 struct sys_quotactl_args /* { 783 syscallarg(const char *) path; 784 syscallarg(int) cmd; 785 syscallarg(int) uid; 786 syscallarg(void *) arg; 787 } */ *uap = v; 788 struct mount *mp; 789 int error; 790 struct nameidata nd; 791 792 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 793 SCARG(uap, path), l); 794 if ((error = namei(&nd)) != 0) 795 return (error); 796 mp = nd.ni_vp->v_mount; 797 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 798 SCARG(uap, arg)); 799 vrele(nd.ni_vp); 800 return (error); 801 } 802 803 int 804 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 805 int root) 806 { 807 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 808 int error = 0; 809 810 /* 811 * If MNT_NOWAIT or MNT_LAZY is specified, do not 812 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 813 * overrides MNT_NOWAIT. 814 */ 815 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 816 (flags != MNT_WAIT && flags != 0)) { 817 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 818 goto done; 819 } 820 821 /* Get the filesystem stats now */ 822 memset(sp, 0, sizeof(*sp)); 823 if ((error = VFS_STATVFS(mp, sp)) != 0) { 824 return error; 825 } 826 827 if (cwdi->cwdi_rdir == NULL) 828 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 829 done: 830 if (cwdi->cwdi_rdir != NULL) { 831 size_t len; 832 char *bp; 833 char *path = PNBUF_GET(); 834 835 bp = path + MAXPATHLEN; 836 *--bp = '\0'; 837 rw_enter(&cwdi->cwdi_lock, RW_READER); 838 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 839 MAXPATHLEN / 2, 0, l); 840 rw_exit(&cwdi->cwdi_lock); 841 if (error) { 842 PNBUF_PUT(path); 843 return error; 844 } 845 len = strlen(bp); 846 /* 847 * for mount points that are below our root, we can see 848 * them, so we fix up the pathname and return them. The 849 * rest we cannot see, so we don't allow viewing the 850 * data. 851 */ 852 if (strncmp(bp, sp->f_mntonname, len) == 0) { 853 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 854 sizeof(sp->f_mntonname)); 855 if (sp->f_mntonname[0] == '\0') 856 (void)strlcpy(sp->f_mntonname, "/", 857 sizeof(sp->f_mntonname)); 858 } else { 859 if (root) 860 (void)strlcpy(sp->f_mntonname, "/", 861 sizeof(sp->f_mntonname)); 862 else 863 error = EPERM; 864 } 865 PNBUF_PUT(path); 866 } 867 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 868 return error; 869 } 870 871 /* 872 * Get filesystem statistics by path. 873 */ 874 int 875 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 876 { 877 struct mount *mp; 878 int error; 879 struct nameidata nd; 880 881 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 882 if ((error = namei(&nd)) != 0) 883 return error; 884 mp = nd.ni_vp->v_mount; 885 error = dostatvfs(mp, sb, l, flags, 1); 886 vrele(nd.ni_vp); 887 return error; 888 } 889 890 /* ARGSUSED */ 891 int 892 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 893 { 894 struct sys_statvfs1_args /* { 895 syscallarg(const char *) path; 896 syscallarg(struct statvfs *) buf; 897 syscallarg(int) flags; 898 } */ *uap = v; 899 struct statvfs *sb; 900 int error; 901 902 sb = STATVFSBUF_GET(); 903 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 904 if (error == 0) 905 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 906 STATVFSBUF_PUT(sb); 907 return error; 908 } 909 910 /* 911 * Get filesystem statistics by fd. 912 */ 913 int 914 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 915 { 916 struct proc *p = l->l_proc; 917 struct file *fp; 918 struct mount *mp; 919 int error; 920 921 /* getvnode() will use the descriptor for us */ 922 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 923 return (error); 924 mp = ((struct vnode *)fp->f_data)->v_mount; 925 error = dostatvfs(mp, sb, l, flags, 1); 926 FILE_UNUSE(fp, l); 927 return error; 928 } 929 930 /* ARGSUSED */ 931 int 932 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 933 { 934 struct sys_fstatvfs1_args /* { 935 syscallarg(int) fd; 936 syscallarg(struct statvfs *) buf; 937 syscallarg(int) flags; 938 } */ *uap = v; 939 struct statvfs *sb; 940 int error; 941 942 sb = STATVFSBUF_GET(); 943 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 944 if (error == 0) 945 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 946 STATVFSBUF_PUT(sb); 947 return error; 948 } 949 950 951 /* 952 * Get statistics on all filesystems. 953 */ 954 int 955 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 956 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 957 register_t *retval) 958 { 959 int root = 0; 960 struct proc *p = l->l_proc; 961 struct mount *mp, *nmp; 962 struct statvfs *sb; 963 size_t count, maxcount; 964 int error = 0; 965 966 sb = STATVFSBUF_GET(); 967 maxcount = bufsize / entry_sz; 968 mutex_enter(&mountlist_lock); 969 count = 0; 970 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 971 mp = nmp) { 972 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 973 nmp = CIRCLEQ_NEXT(mp, mnt_list); 974 continue; 975 } 976 if (sfsp && count < maxcount) { 977 error = dostatvfs(mp, sb, l, flags, 0); 978 if (error) { 979 mutex_enter(&mountlist_lock); 980 nmp = CIRCLEQ_NEXT(mp, mnt_list); 981 vfs_unbusy(mp); 982 continue; 983 } 984 error = copyfn(sb, sfsp, entry_sz); 985 if (error) { 986 vfs_unbusy(mp); 987 goto out; 988 } 989 sfsp = (char *)sfsp + entry_sz; 990 root |= strcmp(sb->f_mntonname, "/") == 0; 991 } 992 count++; 993 mutex_enter(&mountlist_lock); 994 nmp = CIRCLEQ_NEXT(mp, mnt_list); 995 vfs_unbusy(mp); 996 } 997 998 mutex_exit(&mountlist_lock); 999 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1000 /* 1001 * fake a root entry 1002 */ 1003 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1004 sb, l, flags, 1); 1005 if (error != 0) 1006 goto out; 1007 if (sfsp) 1008 error = copyfn(sb, sfsp, entry_sz); 1009 count++; 1010 } 1011 if (sfsp && count > maxcount) 1012 *retval = maxcount; 1013 else 1014 *retval = count; 1015 out: 1016 STATVFSBUF_PUT(sb); 1017 return error; 1018 } 1019 1020 int 1021 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 1022 { 1023 struct sys_getvfsstat_args /* { 1024 syscallarg(struct statvfs *) buf; 1025 syscallarg(size_t) bufsize; 1026 syscallarg(int) flags; 1027 } */ *uap = v; 1028 1029 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1030 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1031 } 1032 1033 /* 1034 * Change current working directory to a given file descriptor. 1035 */ 1036 /* ARGSUSED */ 1037 int 1038 sys_fchdir(struct lwp *l, void *v, register_t *retval) 1039 { 1040 struct sys_fchdir_args /* { 1041 syscallarg(int) fd; 1042 } */ *uap = v; 1043 struct proc *p = l->l_proc; 1044 struct filedesc *fdp = p->p_fd; 1045 struct cwdinfo *cwdi; 1046 struct vnode *vp, *tdp; 1047 struct mount *mp; 1048 struct file *fp; 1049 int error; 1050 1051 /* getvnode() will use the descriptor for us */ 1052 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1053 return (error); 1054 vp = (struct vnode *)fp->f_data; 1055 1056 VREF(vp); 1057 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1058 if (vp->v_type != VDIR) 1059 error = ENOTDIR; 1060 else 1061 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1062 if (error) { 1063 vput(vp); 1064 goto out; 1065 } 1066 while ((mp = vp->v_mountedhere) != NULL) { 1067 if (vfs_busy(mp, 0, 0)) 1068 continue; 1069 1070 vput(vp); 1071 error = VFS_ROOT(mp, &tdp); 1072 vfs_unbusy(mp); 1073 if (error) 1074 goto out; 1075 vp = tdp; 1076 } 1077 VOP_UNLOCK(vp, 0); 1078 1079 /* 1080 * Disallow changing to a directory not under the process's 1081 * current root directory (if there is one). 1082 */ 1083 cwdi = p->p_cwdi; 1084 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1085 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1086 vrele(vp); 1087 error = EPERM; /* operation not permitted */ 1088 } else { 1089 vrele(cwdi->cwdi_cdir); 1090 cwdi->cwdi_cdir = vp; 1091 } 1092 rw_exit(&cwdi->cwdi_lock); 1093 1094 out: 1095 FILE_UNUSE(fp, l); 1096 return (error); 1097 } 1098 1099 /* 1100 * Change this process's notion of the root directory to a given file 1101 * descriptor. 1102 */ 1103 int 1104 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1105 { 1106 struct sys_fchroot_args *uap = v; 1107 struct proc *p = l->l_proc; 1108 struct filedesc *fdp = p->p_fd; 1109 struct cwdinfo *cwdi; 1110 struct vnode *vp; 1111 struct file *fp; 1112 int error; 1113 1114 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1115 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1116 return error; 1117 /* getvnode() will use the descriptor for us */ 1118 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1119 return error; 1120 vp = (struct vnode *) fp->f_data; 1121 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1122 if (vp->v_type != VDIR) 1123 error = ENOTDIR; 1124 else 1125 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1126 VOP_UNLOCK(vp, 0); 1127 if (error) 1128 goto out; 1129 VREF(vp); 1130 1131 /* 1132 * Prevent escaping from chroot by putting the root under 1133 * the working directory. Silently chdir to / if we aren't 1134 * already there. 1135 */ 1136 cwdi = p->p_cwdi; 1137 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1138 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1139 /* 1140 * XXX would be more failsafe to change directory to a 1141 * deadfs node here instead 1142 */ 1143 vrele(cwdi->cwdi_cdir); 1144 VREF(vp); 1145 cwdi->cwdi_cdir = vp; 1146 } 1147 1148 if (cwdi->cwdi_rdir != NULL) 1149 vrele(cwdi->cwdi_rdir); 1150 cwdi->cwdi_rdir = vp; 1151 rw_exit(&cwdi->cwdi_lock); 1152 1153 out: 1154 FILE_UNUSE(fp, l); 1155 return (error); 1156 } 1157 1158 /* 1159 * Change current working directory (``.''). 1160 */ 1161 /* ARGSUSED */ 1162 int 1163 sys_chdir(struct lwp *l, void *v, register_t *retval) 1164 { 1165 struct sys_chdir_args /* { 1166 syscallarg(const char *) path; 1167 } */ *uap = v; 1168 struct proc *p = l->l_proc; 1169 struct cwdinfo *cwdi; 1170 int error; 1171 struct nameidata nd; 1172 1173 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1174 SCARG(uap, path), l); 1175 if ((error = change_dir(&nd, l)) != 0) 1176 return (error); 1177 cwdi = p->p_cwdi; 1178 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1179 vrele(cwdi->cwdi_cdir); 1180 cwdi->cwdi_cdir = nd.ni_vp; 1181 rw_exit(&cwdi->cwdi_lock); 1182 return (0); 1183 } 1184 1185 /* 1186 * Change notion of root (``/'') directory. 1187 */ 1188 /* ARGSUSED */ 1189 int 1190 sys_chroot(struct lwp *l, void *v, register_t *retval) 1191 { 1192 struct sys_chroot_args /* { 1193 syscallarg(const char *) path; 1194 } */ *uap = v; 1195 struct proc *p = l->l_proc; 1196 struct cwdinfo *cwdi; 1197 struct vnode *vp; 1198 int error; 1199 struct nameidata nd; 1200 1201 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1202 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1203 return (error); 1204 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1205 SCARG(uap, path), l); 1206 if ((error = change_dir(&nd, l)) != 0) 1207 return (error); 1208 1209 cwdi = p->p_cwdi; 1210 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1211 if (cwdi->cwdi_rdir != NULL) 1212 vrele(cwdi->cwdi_rdir); 1213 vp = nd.ni_vp; 1214 cwdi->cwdi_rdir = vp; 1215 1216 /* 1217 * Prevent escaping from chroot by putting the root under 1218 * the working directory. Silently chdir to / if we aren't 1219 * already there. 1220 */ 1221 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1222 /* 1223 * XXX would be more failsafe to change directory to a 1224 * deadfs node here instead 1225 */ 1226 vrele(cwdi->cwdi_cdir); 1227 VREF(vp); 1228 cwdi->cwdi_cdir = vp; 1229 } 1230 rw_exit(&cwdi->cwdi_lock); 1231 1232 return (0); 1233 } 1234 1235 /* 1236 * Common routine for chroot and chdir. 1237 */ 1238 static int 1239 change_dir(struct nameidata *ndp, struct lwp *l) 1240 { 1241 struct vnode *vp; 1242 int error; 1243 1244 if ((error = namei(ndp)) != 0) 1245 return (error); 1246 vp = ndp->ni_vp; 1247 if (vp->v_type != VDIR) 1248 error = ENOTDIR; 1249 else 1250 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1251 1252 if (error) 1253 vput(vp); 1254 else 1255 VOP_UNLOCK(vp, 0); 1256 return (error); 1257 } 1258 1259 /* 1260 * Check permissions, allocate an open file structure, 1261 * and call the device open routine if any. 1262 */ 1263 int 1264 sys_open(struct lwp *l, void *v, register_t *retval) 1265 { 1266 struct sys_open_args /* { 1267 syscallarg(const char *) path; 1268 syscallarg(int) flags; 1269 syscallarg(int) mode; 1270 } */ *uap = v; 1271 struct proc *p = l->l_proc; 1272 struct cwdinfo *cwdi = p->p_cwdi; 1273 struct filedesc *fdp = p->p_fd; 1274 struct file *fp; 1275 struct vnode *vp; 1276 int flags, cmode; 1277 int type, indx, error; 1278 struct flock lf; 1279 struct nameidata nd; 1280 1281 flags = FFLAGS(SCARG(uap, flags)); 1282 if ((flags & (FREAD | FWRITE)) == 0) 1283 return (EINVAL); 1284 /* falloc() will use the file descriptor for us */ 1285 if ((error = falloc(l, &fp, &indx)) != 0) 1286 return (error); 1287 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1288 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1289 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1290 SCARG(uap, path), l); 1291 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1292 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1293 rw_enter(&fdp->fd_lock, RW_WRITER); 1294 FILE_UNUSE(fp, l); 1295 fdp->fd_ofiles[indx] = NULL; 1296 rw_exit(&fdp->fd_lock); 1297 ffree(fp); 1298 if ((error == EDUPFD || error == EMOVEFD) && 1299 l->l_dupfd >= 0 && /* XXX from fdopen */ 1300 (error = 1301 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1302 *retval = indx; 1303 return (0); 1304 } 1305 if (error == ERESTART) 1306 error = EINTR; 1307 fdremove(fdp, indx); 1308 return (error); 1309 } 1310 1311 l->l_dupfd = 0; 1312 vp = nd.ni_vp; 1313 fp->f_flag = flags & FMASK; 1314 fp->f_type = DTYPE_VNODE; 1315 fp->f_ops = &vnops; 1316 fp->f_data = vp; 1317 if (flags & (O_EXLOCK | O_SHLOCK)) { 1318 lf.l_whence = SEEK_SET; 1319 lf.l_start = 0; 1320 lf.l_len = 0; 1321 if (flags & O_EXLOCK) 1322 lf.l_type = F_WRLCK; 1323 else 1324 lf.l_type = F_RDLCK; 1325 type = F_FLOCK; 1326 if ((flags & FNONBLOCK) == 0) 1327 type |= F_WAIT; 1328 VOP_UNLOCK(vp, 0); 1329 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1330 if (error) { 1331 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1332 FILE_UNUSE(fp, l); 1333 ffree(fp); 1334 fdremove(fdp, indx); 1335 return (error); 1336 } 1337 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1338 fp->f_flag |= FHASLOCK; 1339 } 1340 VOP_UNLOCK(vp, 0); 1341 *retval = indx; 1342 FILE_SET_MATURE(fp); 1343 FILE_UNUSE(fp, l); 1344 return (0); 1345 } 1346 1347 static void 1348 vfs__fhfree(fhandle_t *fhp) 1349 { 1350 size_t fhsize; 1351 1352 if (fhp == NULL) { 1353 return; 1354 } 1355 fhsize = FHANDLE_SIZE(fhp); 1356 kmem_free(fhp, fhsize); 1357 } 1358 1359 /* 1360 * vfs_composefh: compose a filehandle. 1361 */ 1362 1363 int 1364 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1365 { 1366 struct mount *mp; 1367 struct fid *fidp; 1368 int error; 1369 size_t needfhsize; 1370 size_t fidsize; 1371 1372 mp = vp->v_mount; 1373 fidp = NULL; 1374 if (*fh_size < FHANDLE_SIZE_MIN) { 1375 fidsize = 0; 1376 } else { 1377 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1378 if (fhp != NULL) { 1379 memset(fhp, 0, *fh_size); 1380 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1381 fidp = &fhp->fh_fid; 1382 } 1383 } 1384 error = VFS_VPTOFH(vp, fidp, &fidsize); 1385 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1386 if (error == 0 && *fh_size < needfhsize) { 1387 error = E2BIG; 1388 } 1389 *fh_size = needfhsize; 1390 return error; 1391 } 1392 1393 int 1394 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1395 { 1396 struct mount *mp; 1397 fhandle_t *fhp; 1398 size_t fhsize; 1399 size_t fidsize; 1400 int error; 1401 1402 *fhpp = NULL; 1403 mp = vp->v_mount; 1404 fidsize = 0; 1405 error = VFS_VPTOFH(vp, NULL, &fidsize); 1406 KASSERT(error != 0); 1407 if (error != E2BIG) { 1408 goto out; 1409 } 1410 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1411 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1412 if (fhp == NULL) { 1413 error = ENOMEM; 1414 goto out; 1415 } 1416 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1417 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1418 if (error == 0) { 1419 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1420 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1421 *fhpp = fhp; 1422 } else { 1423 kmem_free(fhp, fhsize); 1424 } 1425 out: 1426 return error; 1427 } 1428 1429 void 1430 vfs_composefh_free(fhandle_t *fhp) 1431 { 1432 1433 vfs__fhfree(fhp); 1434 } 1435 1436 /* 1437 * vfs_fhtovp: lookup a vnode by a filehandle. 1438 */ 1439 1440 int 1441 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1442 { 1443 struct mount *mp; 1444 int error; 1445 1446 *vpp = NULL; 1447 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1448 if (mp == NULL) { 1449 error = ESTALE; 1450 goto out; 1451 } 1452 if (mp->mnt_op->vfs_fhtovp == NULL) { 1453 error = EOPNOTSUPP; 1454 goto out; 1455 } 1456 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1457 out: 1458 return error; 1459 } 1460 1461 /* 1462 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1463 * the needed size. 1464 */ 1465 1466 int 1467 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1468 { 1469 fhandle_t *fhp; 1470 int error; 1471 1472 *fhpp = NULL; 1473 if (fhsize > FHANDLE_SIZE_MAX) { 1474 return EINVAL; 1475 } 1476 if (fhsize < FHANDLE_SIZE_MIN) { 1477 return EINVAL; 1478 } 1479 again: 1480 fhp = kmem_alloc(fhsize, KM_SLEEP); 1481 if (fhp == NULL) { 1482 return ENOMEM; 1483 } 1484 error = copyin(ufhp, fhp, fhsize); 1485 if (error == 0) { 1486 /* XXX this check shouldn't be here */ 1487 if (FHANDLE_SIZE(fhp) == fhsize) { 1488 *fhpp = fhp; 1489 return 0; 1490 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1491 /* 1492 * a kludge for nfsv2 padded handles. 1493 */ 1494 size_t sz; 1495 1496 sz = FHANDLE_SIZE(fhp); 1497 kmem_free(fhp, fhsize); 1498 fhsize = sz; 1499 goto again; 1500 } else { 1501 /* 1502 * userland told us wrong size. 1503 */ 1504 error = EINVAL; 1505 } 1506 } 1507 kmem_free(fhp, fhsize); 1508 return error; 1509 } 1510 1511 void 1512 vfs_copyinfh_free(fhandle_t *fhp) 1513 { 1514 1515 vfs__fhfree(fhp); 1516 } 1517 1518 /* 1519 * Get file handle system call 1520 */ 1521 int 1522 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1523 { 1524 struct sys___getfh30_args /* { 1525 syscallarg(char *) fname; 1526 syscallarg(fhandle_t *) fhp; 1527 syscallarg(size_t *) fh_size; 1528 } */ *uap = v; 1529 struct vnode *vp; 1530 fhandle_t *fh; 1531 int error; 1532 struct nameidata nd; 1533 size_t sz; 1534 size_t usz; 1535 1536 /* 1537 * Must be super user 1538 */ 1539 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1540 0, NULL, NULL, NULL); 1541 if (error) 1542 return (error); 1543 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1544 SCARG(uap, fname), l); 1545 error = namei(&nd); 1546 if (error) 1547 return (error); 1548 vp = nd.ni_vp; 1549 error = vfs_composefh_alloc(vp, &fh); 1550 vput(vp); 1551 if (error != 0) { 1552 goto out; 1553 } 1554 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1555 if (error != 0) { 1556 goto out; 1557 } 1558 sz = FHANDLE_SIZE(fh); 1559 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1560 if (error != 0) { 1561 goto out; 1562 } 1563 if (usz >= sz) { 1564 error = copyout(fh, SCARG(uap, fhp), sz); 1565 } else { 1566 error = E2BIG; 1567 } 1568 out: 1569 vfs_composefh_free(fh); 1570 return (error); 1571 } 1572 1573 /* 1574 * Open a file given a file handle. 1575 * 1576 * Check permissions, allocate an open file structure, 1577 * and call the device open routine if any. 1578 */ 1579 1580 int 1581 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1582 register_t *retval) 1583 { 1584 struct filedesc *fdp = l->l_proc->p_fd; 1585 struct file *fp; 1586 struct vnode *vp = NULL; 1587 kauth_cred_t cred = l->l_cred; 1588 struct file *nfp; 1589 int type, indx, error=0; 1590 struct flock lf; 1591 struct vattr va; 1592 fhandle_t *fh; 1593 int flags; 1594 1595 /* 1596 * Must be super user 1597 */ 1598 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1599 0, NULL, NULL, NULL))) 1600 return (error); 1601 1602 flags = FFLAGS(oflags); 1603 if ((flags & (FREAD | FWRITE)) == 0) 1604 return (EINVAL); 1605 if ((flags & O_CREAT)) 1606 return (EINVAL); 1607 /* falloc() will use the file descriptor for us */ 1608 if ((error = falloc(l, &nfp, &indx)) != 0) 1609 return (error); 1610 fp = nfp; 1611 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1612 if (error != 0) { 1613 goto bad; 1614 } 1615 error = vfs_fhtovp(fh, &vp); 1616 if (error != 0) { 1617 goto bad; 1618 } 1619 1620 /* Now do an effective vn_open */ 1621 1622 if (vp->v_type == VSOCK) { 1623 error = EOPNOTSUPP; 1624 goto bad; 1625 } 1626 error = vn_openchk(vp, cred, flags); 1627 if (error != 0) 1628 goto bad; 1629 if (flags & O_TRUNC) { 1630 VOP_UNLOCK(vp, 0); /* XXX */ 1631 VOP_LEASE(vp, cred, LEASE_WRITE); 1632 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1633 VATTR_NULL(&va); 1634 va.va_size = 0; 1635 error = VOP_SETATTR(vp, &va, cred); 1636 if (error) 1637 goto bad; 1638 } 1639 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1640 goto bad; 1641 if (flags & FWRITE) 1642 vp->v_writecount++; 1643 1644 /* done with modified vn_open, now finish what sys_open does. */ 1645 1646 fp->f_flag = flags & FMASK; 1647 fp->f_type = DTYPE_VNODE; 1648 fp->f_ops = &vnops; 1649 fp->f_data = vp; 1650 if (flags & (O_EXLOCK | O_SHLOCK)) { 1651 lf.l_whence = SEEK_SET; 1652 lf.l_start = 0; 1653 lf.l_len = 0; 1654 if (flags & O_EXLOCK) 1655 lf.l_type = F_WRLCK; 1656 else 1657 lf.l_type = F_RDLCK; 1658 type = F_FLOCK; 1659 if ((flags & FNONBLOCK) == 0) 1660 type |= F_WAIT; 1661 VOP_UNLOCK(vp, 0); 1662 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1663 if (error) { 1664 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1665 FILE_UNUSE(fp, l); 1666 ffree(fp); 1667 fdremove(fdp, indx); 1668 return (error); 1669 } 1670 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1671 fp->f_flag |= FHASLOCK; 1672 } 1673 VOP_UNLOCK(vp, 0); 1674 *retval = indx; 1675 FILE_SET_MATURE(fp); 1676 FILE_UNUSE(fp, l); 1677 vfs_copyinfh_free(fh); 1678 return (0); 1679 1680 bad: 1681 FILE_UNUSE(fp, l); 1682 ffree(fp); 1683 fdremove(fdp, indx); 1684 if (vp != NULL) 1685 vput(vp); 1686 vfs_copyinfh_free(fh); 1687 return (error); 1688 } 1689 1690 int 1691 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1692 { 1693 struct sys___fhopen40_args /* { 1694 syscallarg(const void *) fhp; 1695 syscallarg(size_t) fh_size; 1696 syscallarg(int) flags; 1697 } */ *uap = v; 1698 1699 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1700 SCARG(uap, flags), retval); 1701 } 1702 1703 int 1704 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1705 { 1706 int error; 1707 fhandle_t *fh; 1708 struct vnode *vp; 1709 1710 /* 1711 * Must be super user 1712 */ 1713 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1714 0, NULL, NULL, NULL))) 1715 return (error); 1716 1717 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1718 if (error != 0) 1719 return error; 1720 1721 error = vfs_fhtovp(fh, &vp); 1722 vfs_copyinfh_free(fh); 1723 if (error != 0) 1724 return error; 1725 1726 error = vn_stat(vp, sb, l); 1727 vput(vp); 1728 return error; 1729 } 1730 1731 1732 /* ARGSUSED */ 1733 int 1734 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1735 { 1736 struct sys___fhstat40_args /* { 1737 syscallarg(const void *) fhp; 1738 syscallarg(size_t) fh_size; 1739 syscallarg(struct stat *) sb; 1740 } */ *uap = v; 1741 struct stat sb; 1742 int error; 1743 1744 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1745 if (error) 1746 return error; 1747 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1748 } 1749 1750 int 1751 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1752 int flags) 1753 { 1754 fhandle_t *fh; 1755 struct mount *mp; 1756 struct vnode *vp; 1757 int error; 1758 1759 /* 1760 * Must be super user 1761 */ 1762 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1763 0, NULL, NULL, NULL))) 1764 return error; 1765 1766 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1767 if (error != 0) 1768 return error; 1769 1770 error = vfs_fhtovp(fh, &vp); 1771 vfs_copyinfh_free(fh); 1772 if (error != 0) 1773 return error; 1774 1775 mp = vp->v_mount; 1776 error = dostatvfs(mp, sb, l, flags, 1); 1777 vput(vp); 1778 return error; 1779 } 1780 1781 /* ARGSUSED */ 1782 int 1783 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1784 { 1785 struct sys___fhstatvfs140_args /* { 1786 syscallarg(const void *) fhp; 1787 syscallarg(size_t) fh_size; 1788 syscallarg(struct statvfs *) buf; 1789 syscallarg(int) flags; 1790 } */ *uap = v; 1791 struct statvfs *sb = STATVFSBUF_GET(); 1792 int error; 1793 1794 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1795 SCARG(uap, flags)); 1796 if (error == 0) 1797 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1798 STATVFSBUF_PUT(sb); 1799 return error; 1800 } 1801 1802 /* 1803 * Create a special file. 1804 */ 1805 /* ARGSUSED */ 1806 int 1807 sys_mknod(struct lwp *l, void *v, register_t *retval) 1808 { 1809 struct sys_mknod_args /* { 1810 syscallarg(const char *) path; 1811 syscallarg(int) mode; 1812 syscallarg(int) dev; 1813 } */ *uap = v; 1814 struct proc *p = l->l_proc; 1815 struct vnode *vp; 1816 struct vattr vattr; 1817 int error, optype; 1818 struct nameidata nd; 1819 char *path; 1820 const char *cpath; 1821 enum uio_seg seg = UIO_USERSPACE; 1822 1823 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1824 0, NULL, NULL, NULL)) != 0) 1825 return (error); 1826 1827 optype = VOP_MKNOD_DESCOFFSET; 1828 1829 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1830 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1831 1832 if ((error = namei(&nd)) != 0) 1833 goto out; 1834 vp = nd.ni_vp; 1835 if (vp != NULL) 1836 error = EEXIST; 1837 else { 1838 VATTR_NULL(&vattr); 1839 /* We will read cwdi->cwdi_cmask unlocked. */ 1840 vattr.va_mode = 1841 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1842 vattr.va_rdev = SCARG(uap, dev); 1843 1844 switch (SCARG(uap, mode) & S_IFMT) { 1845 case S_IFMT: /* used by badsect to flag bad sectors */ 1846 vattr.va_type = VBAD; 1847 break; 1848 case S_IFCHR: 1849 vattr.va_type = VCHR; 1850 break; 1851 case S_IFBLK: 1852 vattr.va_type = VBLK; 1853 break; 1854 case S_IFWHT: 1855 optype = VOP_WHITEOUT_DESCOFFSET; 1856 break; 1857 case S_IFREG: 1858 #if NVERIEXEC > 0 1859 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1860 O_CREAT); 1861 #endif /* NVERIEXEC > 0 */ 1862 vattr.va_type = VREG; 1863 vattr.va_rdev = VNOVAL; 1864 optype = VOP_CREATE_DESCOFFSET; 1865 break; 1866 default: 1867 error = EINVAL; 1868 break; 1869 } 1870 } 1871 if (!error) { 1872 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1873 switch (optype) { 1874 case VOP_WHITEOUT_DESCOFFSET: 1875 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1876 if (error) 1877 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1878 vput(nd.ni_dvp); 1879 break; 1880 1881 case VOP_MKNOD_DESCOFFSET: 1882 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1883 &nd.ni_cnd, &vattr); 1884 if (error == 0) 1885 vput(nd.ni_vp); 1886 break; 1887 1888 case VOP_CREATE_DESCOFFSET: 1889 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1890 &nd.ni_cnd, &vattr); 1891 if (error == 0) 1892 vput(nd.ni_vp); 1893 break; 1894 } 1895 } else { 1896 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1897 if (nd.ni_dvp == vp) 1898 vrele(nd.ni_dvp); 1899 else 1900 vput(nd.ni_dvp); 1901 if (vp) 1902 vrele(vp); 1903 } 1904 out: 1905 VERIEXEC_PATH_PUT(path); 1906 return (error); 1907 } 1908 1909 /* 1910 * Create a named pipe. 1911 */ 1912 /* ARGSUSED */ 1913 int 1914 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1915 { 1916 struct sys_mkfifo_args /* { 1917 syscallarg(const char *) path; 1918 syscallarg(int) mode; 1919 } */ *uap = v; 1920 struct proc *p = l->l_proc; 1921 struct vattr vattr; 1922 int error; 1923 struct nameidata nd; 1924 1925 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1926 if ((error = namei(&nd)) != 0) 1927 return (error); 1928 if (nd.ni_vp != NULL) { 1929 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1930 if (nd.ni_dvp == nd.ni_vp) 1931 vrele(nd.ni_dvp); 1932 else 1933 vput(nd.ni_dvp); 1934 vrele(nd.ni_vp); 1935 return (EEXIST); 1936 } 1937 VATTR_NULL(&vattr); 1938 vattr.va_type = VFIFO; 1939 /* We will read cwdi->cwdi_cmask unlocked. */ 1940 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1941 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1942 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1943 if (error == 0) 1944 vput(nd.ni_vp); 1945 return (error); 1946 } 1947 1948 /* 1949 * Make a hard file link. 1950 */ 1951 /* ARGSUSED */ 1952 int 1953 sys_link(struct lwp *l, void *v, register_t *retval) 1954 { 1955 struct sys_link_args /* { 1956 syscallarg(const char *) path; 1957 syscallarg(const char *) link; 1958 } */ *uap = v; 1959 struct vnode *vp; 1960 struct nameidata nd; 1961 int error; 1962 1963 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1964 SCARG(uap, path), l); 1965 if ((error = namei(&nd)) != 0) 1966 return (error); 1967 vp = nd.ni_vp; 1968 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1969 SCARG(uap, link), l); 1970 if ((error = namei(&nd)) != 0) 1971 goto out; 1972 if (nd.ni_vp) { 1973 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1974 if (nd.ni_dvp == nd.ni_vp) 1975 vrele(nd.ni_dvp); 1976 else 1977 vput(nd.ni_dvp); 1978 vrele(nd.ni_vp); 1979 error = EEXIST; 1980 goto out; 1981 } 1982 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1983 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 1984 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1985 out: 1986 vrele(vp); 1987 return (error); 1988 } 1989 1990 /* 1991 * Make a symbolic link. 1992 */ 1993 /* ARGSUSED */ 1994 int 1995 sys_symlink(struct lwp *l, void *v, register_t *retval) 1996 { 1997 struct sys_symlink_args /* { 1998 syscallarg(const char *) path; 1999 syscallarg(const char *) link; 2000 } */ *uap = v; 2001 struct proc *p = l->l_proc; 2002 struct vattr vattr; 2003 char *path; 2004 int error; 2005 struct nameidata nd; 2006 2007 path = PNBUF_GET(); 2008 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2009 if (error) 2010 goto out; 2011 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2012 SCARG(uap, link), l); 2013 if ((error = namei(&nd)) != 0) 2014 goto out; 2015 if (nd.ni_vp) { 2016 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2017 if (nd.ni_dvp == nd.ni_vp) 2018 vrele(nd.ni_dvp); 2019 else 2020 vput(nd.ni_dvp); 2021 vrele(nd.ni_vp); 2022 error = EEXIST; 2023 goto out; 2024 } 2025 VATTR_NULL(&vattr); 2026 vattr.va_type = VLNK; 2027 /* We will read cwdi->cwdi_cmask unlocked. */ 2028 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2029 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2030 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2031 if (error == 0) 2032 vput(nd.ni_vp); 2033 out: 2034 PNBUF_PUT(path); 2035 return (error); 2036 } 2037 2038 /* 2039 * Delete a whiteout from the filesystem. 2040 */ 2041 /* ARGSUSED */ 2042 int 2043 sys_undelete(struct lwp *l, void *v, register_t *retval) 2044 { 2045 struct sys_undelete_args /* { 2046 syscallarg(const char *) path; 2047 } */ *uap = v; 2048 int error; 2049 struct nameidata nd; 2050 2051 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2052 UIO_USERSPACE, SCARG(uap, path), l); 2053 error = namei(&nd); 2054 if (error) 2055 return (error); 2056 2057 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2058 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2059 if (nd.ni_dvp == nd.ni_vp) 2060 vrele(nd.ni_dvp); 2061 else 2062 vput(nd.ni_dvp); 2063 if (nd.ni_vp) 2064 vrele(nd.ni_vp); 2065 return (EEXIST); 2066 } 2067 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2068 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2069 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2070 vput(nd.ni_dvp); 2071 return (error); 2072 } 2073 2074 /* 2075 * Delete a name from the filesystem. 2076 */ 2077 /* ARGSUSED */ 2078 int 2079 sys_unlink(struct lwp *l, void *v, register_t *retval) 2080 { 2081 struct sys_unlink_args /* { 2082 syscallarg(const char *) path; 2083 } */ *uap = v; 2084 struct vnode *vp; 2085 int error; 2086 struct nameidata nd; 2087 char *path; 2088 const char *cpath; 2089 enum uio_seg seg = UIO_USERSPACE; 2090 2091 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2092 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2093 2094 if ((error = namei(&nd)) != 0) 2095 goto out; 2096 vp = nd.ni_vp; 2097 2098 /* 2099 * The root of a mounted filesystem cannot be deleted. 2100 */ 2101 if (vp->v_vflag & VV_ROOT) { 2102 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2103 if (nd.ni_dvp == vp) 2104 vrele(nd.ni_dvp); 2105 else 2106 vput(nd.ni_dvp); 2107 vput(vp); 2108 error = EBUSY; 2109 goto out; 2110 } 2111 2112 #if NVERIEXEC > 0 2113 /* Handle remove requests for veriexec entries. */ 2114 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2115 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2116 if (nd.ni_dvp == vp) 2117 vrele(nd.ni_dvp); 2118 else 2119 vput(nd.ni_dvp); 2120 vput(vp); 2121 goto out; 2122 } 2123 #endif /* NVERIEXEC > 0 */ 2124 2125 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2126 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2127 #ifdef FILEASSOC 2128 (void)fileassoc_file_delete(vp); 2129 #endif /* FILEASSOC */ 2130 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2131 out: 2132 VERIEXEC_PATH_PUT(path); 2133 return (error); 2134 } 2135 2136 /* 2137 * Reposition read/write file offset. 2138 */ 2139 int 2140 sys_lseek(struct lwp *l, void *v, register_t *retval) 2141 { 2142 struct sys_lseek_args /* { 2143 syscallarg(int) fd; 2144 syscallarg(int) pad; 2145 syscallarg(off_t) offset; 2146 syscallarg(int) whence; 2147 } */ *uap = v; 2148 struct proc *p = l->l_proc; 2149 kauth_cred_t cred = l->l_cred; 2150 struct filedesc *fdp = p->p_fd; 2151 struct file *fp; 2152 struct vnode *vp; 2153 struct vattr vattr; 2154 off_t newoff; 2155 int error; 2156 2157 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2158 return (EBADF); 2159 2160 vp = (struct vnode *)fp->f_data; 2161 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2162 error = ESPIPE; 2163 mutex_exit(&fp->f_lock); 2164 goto out; 2165 } 2166 2167 switch (SCARG(uap, whence)) { 2168 case SEEK_CUR: 2169 newoff = fp->f_offset + SCARG(uap, offset); 2170 FILE_USE(fp); 2171 break; 2172 case SEEK_END: 2173 FILE_USE(fp); 2174 error = VOP_GETATTR(vp, &vattr, cred); 2175 if (error) { 2176 FILE_UNUSE(fp, l); 2177 goto out; 2178 } 2179 newoff = SCARG(uap, offset) + vattr.va_size; 2180 break; 2181 case SEEK_SET: 2182 FILE_USE(fp); 2183 newoff = SCARG(uap, offset); 2184 break; 2185 default: 2186 mutex_exit(&fp->f_lock); 2187 error = EINVAL; 2188 goto out; 2189 } 2190 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2191 mutex_enter(&fp->f_lock); 2192 *(off_t *)retval = fp->f_offset = newoff; 2193 mutex_exit(&fp->f_lock); 2194 } 2195 FILE_UNUSE(fp, l); 2196 out: 2197 return (error); 2198 } 2199 2200 /* 2201 * Positional read system call. 2202 */ 2203 int 2204 sys_pread(struct lwp *l, void *v, register_t *retval) 2205 { 2206 struct sys_pread_args /* { 2207 syscallarg(int) fd; 2208 syscallarg(void *) buf; 2209 syscallarg(size_t) nbyte; 2210 syscallarg(off_t) offset; 2211 } */ *uap = v; 2212 struct proc *p = l->l_proc; 2213 struct filedesc *fdp = p->p_fd; 2214 struct file *fp; 2215 struct vnode *vp; 2216 off_t offset; 2217 int error, fd = SCARG(uap, fd); 2218 2219 if ((fp = fd_getfile(fdp, fd)) == NULL) 2220 return (EBADF); 2221 2222 if ((fp->f_flag & FREAD) == 0) { 2223 mutex_exit(&fp->f_lock); 2224 return (EBADF); 2225 } 2226 2227 FILE_USE(fp); 2228 2229 vp = (struct vnode *)fp->f_data; 2230 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2231 error = ESPIPE; 2232 goto out; 2233 } 2234 2235 offset = SCARG(uap, offset); 2236 2237 /* 2238 * XXX This works because no file systems actually 2239 * XXX take any action on the seek operation. 2240 */ 2241 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2242 goto out; 2243 2244 /* dofileread() will unuse the descriptor for us */ 2245 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2246 &offset, 0, retval)); 2247 2248 out: 2249 FILE_UNUSE(fp, l); 2250 return (error); 2251 } 2252 2253 /* 2254 * Positional scatter read system call. 2255 */ 2256 int 2257 sys_preadv(struct lwp *l, void *v, register_t *retval) 2258 { 2259 struct sys_preadv_args /* { 2260 syscallarg(int) fd; 2261 syscallarg(const struct iovec *) iovp; 2262 syscallarg(int) iovcnt; 2263 syscallarg(off_t) offset; 2264 } */ *uap = v; 2265 2266 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2267 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2268 } 2269 2270 /* 2271 * Positional write system call. 2272 */ 2273 int 2274 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2275 { 2276 struct sys_pwrite_args /* { 2277 syscallarg(int) fd; 2278 syscallarg(const void *) buf; 2279 syscallarg(size_t) nbyte; 2280 syscallarg(off_t) offset; 2281 } */ *uap = v; 2282 struct proc *p = l->l_proc; 2283 struct filedesc *fdp = p->p_fd; 2284 struct file *fp; 2285 struct vnode *vp; 2286 off_t offset; 2287 int error, fd = SCARG(uap, fd); 2288 2289 if ((fp = fd_getfile(fdp, fd)) == NULL) 2290 return (EBADF); 2291 2292 if ((fp->f_flag & FWRITE) == 0) { 2293 mutex_exit(&fp->f_lock); 2294 return (EBADF); 2295 } 2296 2297 FILE_USE(fp); 2298 2299 vp = (struct vnode *)fp->f_data; 2300 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2301 error = ESPIPE; 2302 goto out; 2303 } 2304 2305 offset = SCARG(uap, offset); 2306 2307 /* 2308 * XXX This works because no file systems actually 2309 * XXX take any action on the seek operation. 2310 */ 2311 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2312 goto out; 2313 2314 /* dofilewrite() will unuse the descriptor for us */ 2315 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2316 &offset, 0, retval)); 2317 2318 out: 2319 FILE_UNUSE(fp, l); 2320 return (error); 2321 } 2322 2323 /* 2324 * Positional gather write system call. 2325 */ 2326 int 2327 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2328 { 2329 struct sys_pwritev_args /* { 2330 syscallarg(int) fd; 2331 syscallarg(const struct iovec *) iovp; 2332 syscallarg(int) iovcnt; 2333 syscallarg(off_t) offset; 2334 } */ *uap = v; 2335 2336 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2337 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2338 } 2339 2340 /* 2341 * Check access permissions. 2342 */ 2343 int 2344 sys_access(struct lwp *l, void *v, register_t *retval) 2345 { 2346 struct sys_access_args /* { 2347 syscallarg(const char *) path; 2348 syscallarg(int) flags; 2349 } */ *uap = v; 2350 kauth_cred_t cred; 2351 struct vnode *vp; 2352 int error, flags; 2353 struct nameidata nd; 2354 2355 cred = kauth_cred_dup(l->l_cred); 2356 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2357 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2358 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2359 SCARG(uap, path), l); 2360 /* Override default credentials */ 2361 nd.ni_cnd.cn_cred = cred; 2362 if ((error = namei(&nd)) != 0) 2363 goto out; 2364 vp = nd.ni_vp; 2365 2366 /* Flags == 0 means only check for existence. */ 2367 if (SCARG(uap, flags)) { 2368 flags = 0; 2369 if (SCARG(uap, flags) & R_OK) 2370 flags |= VREAD; 2371 if (SCARG(uap, flags) & W_OK) 2372 flags |= VWRITE; 2373 if (SCARG(uap, flags) & X_OK) 2374 flags |= VEXEC; 2375 2376 error = VOP_ACCESS(vp, flags, cred); 2377 if (!error && (flags & VWRITE)) 2378 error = vn_writechk(vp); 2379 } 2380 vput(vp); 2381 out: 2382 kauth_cred_free(cred); 2383 return (error); 2384 } 2385 2386 /* 2387 * Common code for all sys_stat functions, including compat versions. 2388 */ 2389 int 2390 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2391 struct stat *sb) 2392 { 2393 int error; 2394 struct nameidata nd; 2395 2396 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2397 UIO_USERSPACE, path, l); 2398 error = namei(&nd); 2399 if (error != 0) 2400 return error; 2401 error = vn_stat(nd.ni_vp, sb, l); 2402 vput(nd.ni_vp); 2403 return error; 2404 } 2405 2406 /* 2407 * Get file status; this version follows links. 2408 */ 2409 /* ARGSUSED */ 2410 int 2411 sys___stat30(struct lwp *l, void *v, register_t *retval) 2412 { 2413 struct sys___stat30_args /* { 2414 syscallarg(const char *) path; 2415 syscallarg(struct stat *) ub; 2416 } */ *uap = v; 2417 struct stat sb; 2418 int error; 2419 2420 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2421 if (error) 2422 return error; 2423 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2424 } 2425 2426 /* 2427 * Get file status; this version does not follow links. 2428 */ 2429 /* ARGSUSED */ 2430 int 2431 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2432 { 2433 struct sys___lstat30_args /* { 2434 syscallarg(const char *) path; 2435 syscallarg(struct stat *) ub; 2436 } */ *uap = v; 2437 struct stat sb; 2438 int error; 2439 2440 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2441 if (error) 2442 return error; 2443 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2444 } 2445 2446 /* 2447 * Get configurable pathname variables. 2448 */ 2449 /* ARGSUSED */ 2450 int 2451 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2452 { 2453 struct sys_pathconf_args /* { 2454 syscallarg(const char *) path; 2455 syscallarg(int) name; 2456 } */ *uap = v; 2457 int error; 2458 struct nameidata nd; 2459 2460 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2461 SCARG(uap, path), l); 2462 if ((error = namei(&nd)) != 0) 2463 return (error); 2464 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2465 vput(nd.ni_vp); 2466 return (error); 2467 } 2468 2469 /* 2470 * Return target name of a symbolic link. 2471 */ 2472 /* ARGSUSED */ 2473 int 2474 sys_readlink(struct lwp *l, void *v, register_t *retval) 2475 { 2476 struct sys_readlink_args /* { 2477 syscallarg(const char *) path; 2478 syscallarg(char *) buf; 2479 syscallarg(size_t) count; 2480 } */ *uap = v; 2481 struct vnode *vp; 2482 struct iovec aiov; 2483 struct uio auio; 2484 int error; 2485 struct nameidata nd; 2486 2487 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2488 SCARG(uap, path), l); 2489 if ((error = namei(&nd)) != 0) 2490 return (error); 2491 vp = nd.ni_vp; 2492 if (vp->v_type != VLNK) 2493 error = EINVAL; 2494 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2495 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2496 aiov.iov_base = SCARG(uap, buf); 2497 aiov.iov_len = SCARG(uap, count); 2498 auio.uio_iov = &aiov; 2499 auio.uio_iovcnt = 1; 2500 auio.uio_offset = 0; 2501 auio.uio_rw = UIO_READ; 2502 KASSERT(l == curlwp); 2503 auio.uio_vmspace = l->l_proc->p_vmspace; 2504 auio.uio_resid = SCARG(uap, count); 2505 error = VOP_READLINK(vp, &auio, l->l_cred); 2506 } 2507 vput(vp); 2508 *retval = SCARG(uap, count) - auio.uio_resid; 2509 return (error); 2510 } 2511 2512 /* 2513 * Change flags of a file given a path name. 2514 */ 2515 /* ARGSUSED */ 2516 int 2517 sys_chflags(struct lwp *l, void *v, register_t *retval) 2518 { 2519 struct sys_chflags_args /* { 2520 syscallarg(const char *) path; 2521 syscallarg(u_long) flags; 2522 } */ *uap = v; 2523 struct vnode *vp; 2524 int error; 2525 struct nameidata nd; 2526 2527 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2528 SCARG(uap, path), l); 2529 if ((error = namei(&nd)) != 0) 2530 return (error); 2531 vp = nd.ni_vp; 2532 error = change_flags(vp, SCARG(uap, flags), l); 2533 vput(vp); 2534 return (error); 2535 } 2536 2537 /* 2538 * Change flags of a file given a file descriptor. 2539 */ 2540 /* ARGSUSED */ 2541 int 2542 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2543 { 2544 struct sys_fchflags_args /* { 2545 syscallarg(int) fd; 2546 syscallarg(u_long) flags; 2547 } */ *uap = v; 2548 struct proc *p = l->l_proc; 2549 struct vnode *vp; 2550 struct file *fp; 2551 int error; 2552 2553 /* getvnode() will use the descriptor for us */ 2554 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2555 return (error); 2556 vp = (struct vnode *)fp->f_data; 2557 error = change_flags(vp, SCARG(uap, flags), l); 2558 VOP_UNLOCK(vp, 0); 2559 FILE_UNUSE(fp, l); 2560 return (error); 2561 } 2562 2563 /* 2564 * Change flags of a file given a path name; this version does 2565 * not follow links. 2566 */ 2567 int 2568 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2569 { 2570 struct sys_lchflags_args /* { 2571 syscallarg(const char *) path; 2572 syscallarg(u_long) flags; 2573 } */ *uap = v; 2574 struct vnode *vp; 2575 int error; 2576 struct nameidata nd; 2577 2578 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2579 SCARG(uap, path), l); 2580 if ((error = namei(&nd)) != 0) 2581 return (error); 2582 vp = nd.ni_vp; 2583 error = change_flags(vp, SCARG(uap, flags), l); 2584 vput(vp); 2585 return (error); 2586 } 2587 2588 /* 2589 * Common routine to change flags of a file. 2590 */ 2591 int 2592 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2593 { 2594 struct vattr vattr; 2595 int error; 2596 2597 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 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, void *v, register_t *retval) 2624 { 2625 struct sys_chmod_args /* { 2626 syscallarg(const char *) path; 2627 syscallarg(int) mode; 2628 } */ *uap = v; 2629 int error; 2630 struct nameidata nd; 2631 2632 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2633 SCARG(uap, path), l); 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, void *v, register_t *retval) 2649 { 2650 struct sys_fchmod_args /* { 2651 syscallarg(int) fd; 2652 syscallarg(int) mode; 2653 } */ *uap = v; 2654 struct proc *p = l->l_proc; 2655 struct file *fp; 2656 int error; 2657 2658 /* getvnode() will use the descriptor for us */ 2659 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2660 return (error); 2661 2662 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2663 FILE_UNUSE(fp, l); 2664 return (error); 2665 } 2666 2667 /* 2668 * Change mode of a file given path name; this version does not follow links. 2669 */ 2670 /* ARGSUSED */ 2671 int 2672 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2673 { 2674 struct sys_lchmod_args /* { 2675 syscallarg(const char *) path; 2676 syscallarg(int) mode; 2677 } */ *uap = v; 2678 int error; 2679 struct nameidata nd; 2680 2681 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2682 SCARG(uap, path), l); 2683 if ((error = namei(&nd)) != 0) 2684 return (error); 2685 2686 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2687 2688 vrele(nd.ni_vp); 2689 return (error); 2690 } 2691 2692 /* 2693 * Common routine to set mode given a vnode. 2694 */ 2695 static int 2696 change_mode(struct vnode *vp, int mode, struct lwp *l) 2697 { 2698 struct vattr vattr; 2699 int error; 2700 2701 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2702 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2703 VATTR_NULL(&vattr); 2704 vattr.va_mode = mode & ALLPERMS; 2705 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2706 VOP_UNLOCK(vp, 0); 2707 return (error); 2708 } 2709 2710 /* 2711 * Set ownership given a path name; this version follows links. 2712 */ 2713 /* ARGSUSED */ 2714 int 2715 sys_chown(struct lwp *l, void *v, register_t *retval) 2716 { 2717 struct sys_chown_args /* { 2718 syscallarg(const char *) path; 2719 syscallarg(uid_t) uid; 2720 syscallarg(gid_t) gid; 2721 } */ *uap = v; 2722 int error; 2723 struct nameidata nd; 2724 2725 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2726 SCARG(uap, path), l); 2727 if ((error = namei(&nd)) != 0) 2728 return (error); 2729 2730 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2731 2732 vrele(nd.ni_vp); 2733 return (error); 2734 } 2735 2736 /* 2737 * Set ownership given a path name; this version follows links. 2738 * Provides POSIX semantics. 2739 */ 2740 /* ARGSUSED */ 2741 int 2742 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2743 { 2744 struct sys_chown_args /* { 2745 syscallarg(const char *) path; 2746 syscallarg(uid_t) uid; 2747 syscallarg(gid_t) gid; 2748 } */ *uap = v; 2749 int error; 2750 struct nameidata nd; 2751 2752 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2753 SCARG(uap, path), l); 2754 if ((error = namei(&nd)) != 0) 2755 return (error); 2756 2757 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2758 2759 vrele(nd.ni_vp); 2760 return (error); 2761 } 2762 2763 /* 2764 * Set ownership given a file descriptor. 2765 */ 2766 /* ARGSUSED */ 2767 int 2768 sys_fchown(struct lwp *l, void *v, register_t *retval) 2769 { 2770 struct sys_fchown_args /* { 2771 syscallarg(int) fd; 2772 syscallarg(uid_t) uid; 2773 syscallarg(gid_t) gid; 2774 } */ *uap = v; 2775 struct proc *p = l->l_proc; 2776 int error; 2777 struct file *fp; 2778 2779 /* getvnode() will use the descriptor for us */ 2780 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2781 return (error); 2782 2783 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2784 SCARG(uap, gid), l, 0); 2785 FILE_UNUSE(fp, l); 2786 return (error); 2787 } 2788 2789 /* 2790 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2791 */ 2792 /* ARGSUSED */ 2793 int 2794 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2795 { 2796 struct sys_fchown_args /* { 2797 syscallarg(int) fd; 2798 syscallarg(uid_t) uid; 2799 syscallarg(gid_t) gid; 2800 } */ *uap = v; 2801 struct proc *p = l->l_proc; 2802 int error; 2803 struct file *fp; 2804 2805 /* getvnode() will use the descriptor for us */ 2806 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2807 return (error); 2808 2809 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2810 SCARG(uap, gid), l, 1); 2811 FILE_UNUSE(fp, l); 2812 return (error); 2813 } 2814 2815 /* 2816 * Set ownership given a path name; this version does not follow links. 2817 */ 2818 /* ARGSUSED */ 2819 int 2820 sys_lchown(struct lwp *l, void *v, register_t *retval) 2821 { 2822 struct sys_lchown_args /* { 2823 syscallarg(const char *) path; 2824 syscallarg(uid_t) uid; 2825 syscallarg(gid_t) gid; 2826 } */ *uap = v; 2827 int error; 2828 struct nameidata nd; 2829 2830 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2831 SCARG(uap, path), l); 2832 if ((error = namei(&nd)) != 0) 2833 return (error); 2834 2835 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2836 2837 vrele(nd.ni_vp); 2838 return (error); 2839 } 2840 2841 /* 2842 * Set ownership given a path name; this version does not follow links. 2843 * Provides POSIX/XPG semantics. 2844 */ 2845 /* ARGSUSED */ 2846 int 2847 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2848 { 2849 struct sys_lchown_args /* { 2850 syscallarg(const char *) path; 2851 syscallarg(uid_t) uid; 2852 syscallarg(gid_t) gid; 2853 } */ *uap = v; 2854 int error; 2855 struct nameidata nd; 2856 2857 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2858 SCARG(uap, path), l); 2859 if ((error = namei(&nd)) != 0) 2860 return (error); 2861 2862 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2863 2864 vrele(nd.ni_vp); 2865 return (error); 2866 } 2867 2868 /* 2869 * Common routine to set ownership given a vnode. 2870 */ 2871 static int 2872 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2873 int posix_semantics) 2874 { 2875 struct vattr vattr; 2876 mode_t newmode; 2877 int error; 2878 2879 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2880 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2881 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2882 goto out; 2883 2884 #define CHANGED(x) ((int)(x) != -1) 2885 newmode = vattr.va_mode; 2886 if (posix_semantics) { 2887 /* 2888 * POSIX/XPG semantics: if the caller is not the super-user, 2889 * clear set-user-id and set-group-id bits. Both POSIX and 2890 * the XPG consider the behaviour for calls by the super-user 2891 * implementation-defined; we leave the set-user-id and set- 2892 * group-id settings intact in that case. 2893 */ 2894 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2895 NULL) != 0) 2896 newmode &= ~(S_ISUID | S_ISGID); 2897 } else { 2898 /* 2899 * NetBSD semantics: when changing owner and/or group, 2900 * clear the respective bit(s). 2901 */ 2902 if (CHANGED(uid)) 2903 newmode &= ~S_ISUID; 2904 if (CHANGED(gid)) 2905 newmode &= ~S_ISGID; 2906 } 2907 /* Update va_mode iff altered. */ 2908 if (vattr.va_mode == newmode) 2909 newmode = VNOVAL; 2910 2911 VATTR_NULL(&vattr); 2912 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2913 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2914 vattr.va_mode = newmode; 2915 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2916 #undef CHANGED 2917 2918 out: 2919 VOP_UNLOCK(vp, 0); 2920 return (error); 2921 } 2922 2923 /* 2924 * Set the access and modification times given a path name; this 2925 * version follows links. 2926 */ 2927 /* ARGSUSED */ 2928 int 2929 sys_utimes(struct lwp *l, void *v, register_t *retval) 2930 { 2931 struct sys_utimes_args /* { 2932 syscallarg(const char *) path; 2933 syscallarg(const struct timeval *) tptr; 2934 } */ *uap = v; 2935 2936 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2937 SCARG(uap, tptr), UIO_USERSPACE); 2938 } 2939 2940 /* 2941 * Set the access and modification times given a file descriptor. 2942 */ 2943 /* ARGSUSED */ 2944 int 2945 sys_futimes(struct lwp *l, void *v, register_t *retval) 2946 { 2947 struct sys_futimes_args /* { 2948 syscallarg(int) fd; 2949 syscallarg(const struct timeval *) tptr; 2950 } */ *uap = v; 2951 int error; 2952 struct file *fp; 2953 2954 /* getvnode() will use the descriptor for us */ 2955 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2956 return (error); 2957 2958 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2959 SCARG(uap, tptr), UIO_USERSPACE); 2960 2961 FILE_UNUSE(fp, l); 2962 return (error); 2963 } 2964 2965 /* 2966 * Set the access and modification times given a path name; this 2967 * version does not follow links. 2968 */ 2969 int 2970 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2971 { 2972 struct sys_lutimes_args /* { 2973 syscallarg(const char *) path; 2974 syscallarg(const struct timeval *) tptr; 2975 } */ *uap = v; 2976 2977 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2978 SCARG(uap, tptr), UIO_USERSPACE); 2979 } 2980 2981 /* 2982 * Common routine to set access and modification times given a vnode. 2983 */ 2984 int 2985 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2986 const struct timeval *tptr, enum uio_seg seg) 2987 { 2988 struct vattr vattr; 2989 struct nameidata nd; 2990 int error; 2991 2992 VATTR_NULL(&vattr); 2993 if (tptr == NULL) { 2994 nanotime(&vattr.va_atime); 2995 vattr.va_mtime = vattr.va_atime; 2996 vattr.va_vaflags |= VA_UTIMES_NULL; 2997 } else { 2998 struct timeval tv[2]; 2999 3000 if (seg != UIO_SYSSPACE) { 3001 error = copyin(tptr, &tv, sizeof (tv)); 3002 if (error != 0) 3003 return error; 3004 tptr = tv; 3005 } 3006 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3007 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3008 } 3009 3010 if (vp == NULL) { 3011 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 3012 if ((error = namei(&nd)) != 0) 3013 return (error); 3014 vp = nd.ni_vp; 3015 } else 3016 nd.ni_vp = NULL; 3017 3018 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3019 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3020 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3021 VOP_UNLOCK(vp, 0); 3022 3023 if (nd.ni_vp != NULL) 3024 vrele(nd.ni_vp); 3025 3026 return (error); 3027 } 3028 3029 /* 3030 * Truncate a file given its path name. 3031 */ 3032 /* ARGSUSED */ 3033 int 3034 sys_truncate(struct lwp *l, void *v, register_t *retval) 3035 { 3036 struct sys_truncate_args /* { 3037 syscallarg(const char *) path; 3038 syscallarg(int) pad; 3039 syscallarg(off_t) length; 3040 } */ *uap = v; 3041 struct vnode *vp; 3042 struct vattr vattr; 3043 int error; 3044 struct nameidata nd; 3045 3046 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3047 SCARG(uap, path), l); 3048 if ((error = namei(&nd)) != 0) 3049 return (error); 3050 vp = nd.ni_vp; 3051 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3052 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3053 if (vp->v_type == VDIR) 3054 error = EISDIR; 3055 else if ((error = vn_writechk(vp)) == 0 && 3056 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3057 VATTR_NULL(&vattr); 3058 vattr.va_size = SCARG(uap, length); 3059 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3060 } 3061 vput(vp); 3062 return (error); 3063 } 3064 3065 /* 3066 * Truncate a file given a file descriptor. 3067 */ 3068 /* ARGSUSED */ 3069 int 3070 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3071 { 3072 struct sys_ftruncate_args /* { 3073 syscallarg(int) fd; 3074 syscallarg(int) pad; 3075 syscallarg(off_t) length; 3076 } */ *uap = v; 3077 struct proc *p = l->l_proc; 3078 struct vattr vattr; 3079 struct vnode *vp; 3080 struct file *fp; 3081 int error; 3082 3083 /* getvnode() will use the descriptor for us */ 3084 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3085 return (error); 3086 if ((fp->f_flag & FWRITE) == 0) { 3087 error = EINVAL; 3088 goto out; 3089 } 3090 vp = (struct vnode *)fp->f_data; 3091 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3092 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3093 if (vp->v_type == VDIR) 3094 error = EISDIR; 3095 else if ((error = vn_writechk(vp)) == 0) { 3096 VATTR_NULL(&vattr); 3097 vattr.va_size = SCARG(uap, length); 3098 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3099 } 3100 VOP_UNLOCK(vp, 0); 3101 out: 3102 FILE_UNUSE(fp, l); 3103 return (error); 3104 } 3105 3106 /* 3107 * Sync an open file. 3108 */ 3109 /* ARGSUSED */ 3110 int 3111 sys_fsync(struct lwp *l, void *v, register_t *retval) 3112 { 3113 struct sys_fsync_args /* { 3114 syscallarg(int) fd; 3115 } */ *uap = v; 3116 struct proc *p = l->l_proc; 3117 struct vnode *vp; 3118 struct file *fp; 3119 int error; 3120 3121 /* getvnode() will use the descriptor for us */ 3122 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3123 return (error); 3124 vp = (struct vnode *)fp->f_data; 3125 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3126 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3127 if (error == 0 && bioopsp != NULL && 3128 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3129 (*bioopsp->io_fsync)(vp, 0); 3130 VOP_UNLOCK(vp, 0); 3131 FILE_UNUSE(fp, l); 3132 return (error); 3133 } 3134 3135 /* 3136 * Sync a range of file data. API modeled after that found in AIX. 3137 * 3138 * FDATASYNC indicates that we need only save enough metadata to be able 3139 * to re-read the written data. Note we duplicate AIX's requirement that 3140 * the file be open for writing. 3141 */ 3142 /* ARGSUSED */ 3143 int 3144 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3145 { 3146 struct sys_fsync_range_args /* { 3147 syscallarg(int) fd; 3148 syscallarg(int) flags; 3149 syscallarg(off_t) start; 3150 syscallarg(off_t) length; 3151 } */ *uap = v; 3152 struct proc *p = l->l_proc; 3153 struct vnode *vp; 3154 struct file *fp; 3155 int flags, nflags; 3156 off_t s, e, len; 3157 int error; 3158 3159 /* getvnode() will use the descriptor for us */ 3160 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3161 return (error); 3162 3163 if ((fp->f_flag & FWRITE) == 0) { 3164 error = EBADF; 3165 goto out; 3166 } 3167 3168 flags = SCARG(uap, flags); 3169 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3170 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3171 error = EINVAL; 3172 goto out; 3173 } 3174 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3175 if (flags & FDATASYNC) 3176 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3177 else 3178 nflags = FSYNC_WAIT; 3179 if (flags & FDISKSYNC) 3180 nflags |= FSYNC_CACHE; 3181 3182 len = SCARG(uap, length); 3183 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3184 if (len) { 3185 s = SCARG(uap, start); 3186 e = s + len; 3187 if (e < s) { 3188 error = EINVAL; 3189 goto out; 3190 } 3191 } else { 3192 e = 0; 3193 s = 0; 3194 } 3195 3196 vp = (struct vnode *)fp->f_data; 3197 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3198 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3199 3200 if (error == 0 && bioopsp != NULL && 3201 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3202 (*bioopsp->io_fsync)(vp, nflags); 3203 3204 VOP_UNLOCK(vp, 0); 3205 out: 3206 FILE_UNUSE(fp, l); 3207 return (error); 3208 } 3209 3210 /* 3211 * Sync the data of an open file. 3212 */ 3213 /* ARGSUSED */ 3214 int 3215 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3216 { 3217 struct sys_fdatasync_args /* { 3218 syscallarg(int) fd; 3219 } */ *uap = v; 3220 struct proc *p = l->l_proc; 3221 struct vnode *vp; 3222 struct file *fp; 3223 int error; 3224 3225 /* getvnode() will use the descriptor for us */ 3226 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3227 return (error); 3228 if ((fp->f_flag & FWRITE) == 0) { 3229 FILE_UNUSE(fp, l); 3230 return (EBADF); 3231 } 3232 vp = (struct vnode *)fp->f_data; 3233 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3234 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3235 VOP_UNLOCK(vp, 0); 3236 FILE_UNUSE(fp, l); 3237 return (error); 3238 } 3239 3240 /* 3241 * Rename files, (standard) BSD semantics frontend. 3242 */ 3243 /* ARGSUSED */ 3244 int 3245 sys_rename(struct lwp *l, void *v, register_t *retval) 3246 { 3247 struct sys_rename_args /* { 3248 syscallarg(const char *) from; 3249 syscallarg(const char *) to; 3250 } */ *uap = v; 3251 3252 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3253 } 3254 3255 /* 3256 * Rename files, POSIX semantics frontend. 3257 */ 3258 /* ARGSUSED */ 3259 int 3260 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3261 { 3262 struct sys___posix_rename_args /* { 3263 syscallarg(const char *) from; 3264 syscallarg(const char *) to; 3265 } */ *uap = v; 3266 3267 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3268 } 3269 3270 /* 3271 * Rename files. Source and destination must either both be directories, 3272 * or both not be directories. If target is a directory, it must be empty. 3273 * If `from' and `to' refer to the same object, the value of the `retain' 3274 * argument is used to determine whether `from' will be 3275 * 3276 * (retain == 0) deleted unless `from' and `to' refer to the same 3277 * object in the file system's name space (BSD). 3278 * (retain == 1) always retained (POSIX). 3279 */ 3280 static int 3281 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3282 { 3283 struct vnode *tvp, *fvp, *tdvp; 3284 struct nameidata fromnd, tond; 3285 struct proc *p; 3286 int error; 3287 3288 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3289 UIO_USERSPACE, from, l); 3290 if ((error = namei(&fromnd)) != 0) 3291 return (error); 3292 if (fromnd.ni_dvp != fromnd.ni_vp) 3293 VOP_UNLOCK(fromnd.ni_dvp, 0); 3294 fvp = fromnd.ni_vp; 3295 NDINIT(&tond, RENAME, 3296 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3297 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3298 UIO_USERSPACE, to, l); 3299 if ((error = namei(&tond)) != 0) { 3300 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3301 vrele(fromnd.ni_dvp); 3302 vrele(fvp); 3303 goto out1; 3304 } 3305 tdvp = tond.ni_dvp; 3306 tvp = tond.ni_vp; 3307 3308 if (tvp != NULL) { 3309 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3310 error = ENOTDIR; 3311 goto out; 3312 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3313 error = EISDIR; 3314 goto out; 3315 } 3316 } 3317 3318 if (fvp == tdvp) 3319 error = EINVAL; 3320 3321 /* 3322 * Source and destination refer to the same object. 3323 */ 3324 if (fvp == tvp) { 3325 if (retain) 3326 error = -1; 3327 else if (fromnd.ni_dvp == tdvp && 3328 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3329 !memcmp(fromnd.ni_cnd.cn_nameptr, 3330 tond.ni_cnd.cn_nameptr, 3331 fromnd.ni_cnd.cn_namelen)) 3332 error = -1; 3333 } 3334 3335 #if NVERIEXEC > 0 3336 if (!error) { 3337 char *f1, *f2; 3338 3339 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3340 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3341 3342 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3343 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3344 3345 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3346 3347 free(f1, M_TEMP); 3348 free(f2, M_TEMP); 3349 } 3350 #endif /* NVERIEXEC > 0 */ 3351 3352 out: 3353 p = l->l_proc; 3354 if (!error) { 3355 VOP_LEASE(tdvp, l->l_cred, LEASE_WRITE); 3356 if (fromnd.ni_dvp != tdvp) 3357 VOP_LEASE(fromnd.ni_dvp, l->l_cred, LEASE_WRITE); 3358 if (tvp) { 3359 VOP_LEASE(tvp, l->l_cred, LEASE_WRITE); 3360 } 3361 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3362 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3363 } else { 3364 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3365 if (tdvp == tvp) 3366 vrele(tdvp); 3367 else 3368 vput(tdvp); 3369 if (tvp) 3370 vput(tvp); 3371 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3372 vrele(fromnd.ni_dvp); 3373 vrele(fvp); 3374 } 3375 vrele(tond.ni_startdir); 3376 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3377 out1: 3378 if (fromnd.ni_startdir) 3379 vrele(fromnd.ni_startdir); 3380 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3381 return (error == -1 ? 0 : error); 3382 } 3383 3384 /* 3385 * Make a directory file. 3386 */ 3387 /* ARGSUSED */ 3388 int 3389 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3390 { 3391 struct sys_mkdir_args /* { 3392 syscallarg(const char *) path; 3393 syscallarg(int) mode; 3394 } */ *uap = v; 3395 struct proc *p = l->l_proc; 3396 struct vnode *vp; 3397 struct vattr vattr; 3398 int error; 3399 struct nameidata nd; 3400 3401 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3402 SCARG(uap, path), l); 3403 if ((error = namei(&nd)) != 0) 3404 return (error); 3405 vp = nd.ni_vp; 3406 if (vp != NULL) { 3407 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3408 if (nd.ni_dvp == vp) 3409 vrele(nd.ni_dvp); 3410 else 3411 vput(nd.ni_dvp); 3412 vrele(vp); 3413 return (EEXIST); 3414 } 3415 VATTR_NULL(&vattr); 3416 vattr.va_type = VDIR; 3417 /* We will read cwdi->cwdi_cmask unlocked. */ 3418 vattr.va_mode = 3419 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3420 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3421 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3422 if (!error) 3423 vput(nd.ni_vp); 3424 return (error); 3425 } 3426 3427 /* 3428 * Remove a directory file. 3429 */ 3430 /* ARGSUSED */ 3431 int 3432 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3433 { 3434 struct sys_rmdir_args /* { 3435 syscallarg(const char *) path; 3436 } */ *uap = v; 3437 struct vnode *vp; 3438 int error; 3439 struct nameidata nd; 3440 3441 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3442 SCARG(uap, path), l); 3443 if ((error = namei(&nd)) != 0) 3444 return (error); 3445 vp = nd.ni_vp; 3446 if (vp->v_type != VDIR) { 3447 error = ENOTDIR; 3448 goto out; 3449 } 3450 /* 3451 * No rmdir "." please. 3452 */ 3453 if (nd.ni_dvp == vp) { 3454 error = EINVAL; 3455 goto out; 3456 } 3457 /* 3458 * The root of a mounted filesystem cannot be deleted. 3459 */ 3460 if (vp->v_vflag & VV_ROOT) { 3461 error = EBUSY; 3462 goto out; 3463 } 3464 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3465 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3466 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3467 return (error); 3468 3469 out: 3470 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3471 if (nd.ni_dvp == vp) 3472 vrele(nd.ni_dvp); 3473 else 3474 vput(nd.ni_dvp); 3475 vput(vp); 3476 return (error); 3477 } 3478 3479 /* 3480 * Read a block of directory entries in a file system independent format. 3481 */ 3482 int 3483 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3484 { 3485 struct sys___getdents30_args /* { 3486 syscallarg(int) fd; 3487 syscallarg(char *) buf; 3488 syscallarg(size_t) count; 3489 } */ *uap = v; 3490 struct proc *p = l->l_proc; 3491 struct file *fp; 3492 int error, done; 3493 3494 /* getvnode() will use the descriptor for us */ 3495 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3496 return (error); 3497 if ((fp->f_flag & FREAD) == 0) { 3498 error = EBADF; 3499 goto out; 3500 } 3501 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3502 SCARG(uap, count), &done, l, 0, 0); 3503 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3504 *retval = done; 3505 out: 3506 FILE_UNUSE(fp, l); 3507 return (error); 3508 } 3509 3510 /* 3511 * Set the mode mask for creation of filesystem nodes. 3512 */ 3513 int 3514 sys_umask(struct lwp *l, void *v, register_t *retval) 3515 { 3516 struct sys_umask_args /* { 3517 syscallarg(mode_t) newmask; 3518 } */ *uap = v; 3519 struct proc *p = l->l_proc; 3520 struct cwdinfo *cwdi; 3521 3522 /* 3523 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3524 * important is that we serialize changes to the mask. The 3525 * rw_exit() will issue a write memory barrier on our behalf, 3526 * and force the changes out to other CPUs (as it must use an 3527 * atomic operation, draining the local CPU's store buffers). 3528 */ 3529 cwdi = p->p_cwdi; 3530 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3531 *retval = cwdi->cwdi_cmask; 3532 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3533 rw_exit(&cwdi->cwdi_lock); 3534 3535 return (0); 3536 } 3537 3538 /* 3539 * Void all references to file by ripping underlying filesystem 3540 * away from vnode. 3541 */ 3542 /* ARGSUSED */ 3543 int 3544 sys_revoke(struct lwp *l, void *v, register_t *retval) 3545 { 3546 struct sys_revoke_args /* { 3547 syscallarg(const char *) path; 3548 } */ *uap = v; 3549 struct vnode *vp; 3550 struct vattr vattr; 3551 int error; 3552 bool revoke; 3553 struct nameidata nd; 3554 3555 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3556 SCARG(uap, path), l); 3557 if ((error = namei(&nd)) != 0) 3558 return (error); 3559 vp = nd.ni_vp; 3560 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3561 goto out; 3562 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3563 (error = kauth_authorize_generic(l->l_cred, 3564 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3565 goto out; 3566 mutex_enter(&vp->v_interlock); 3567 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED|VI_LAYER))); 3568 mutex_exit(&vp->v_interlock); 3569 if (revoke) 3570 VOP_REVOKE(vp, REVOKEALL); 3571 out: 3572 vrele(vp); 3573 return (error); 3574 } 3575 3576 /* 3577 * Convert a user file descriptor to a kernel file entry. 3578 */ 3579 int 3580 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3581 { 3582 struct vnode *vp; 3583 struct file *fp; 3584 3585 if ((fp = fd_getfile(fdp, fd)) == NULL) 3586 return (EBADF); 3587 3588 FILE_USE(fp); 3589 3590 if (fp->f_type != DTYPE_VNODE) { 3591 FILE_UNUSE(fp, NULL); 3592 return (EINVAL); 3593 } 3594 3595 vp = (struct vnode *)fp->f_data; 3596 if (vp->v_type == VBAD) { 3597 FILE_UNUSE(fp, NULL); 3598 return (EBADF); 3599 } 3600 3601 *fpp = fp; 3602 return (0); 3603 } 3604
Indexes created Fri Oct 03 02:09:57 GMT 2025