vfs_syscalls.c revision 1.344
1 /* $NetBSD: vfs_syscalls.c,v 1.344 2008/01/28 14:31:18 dholland 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 1.344 2008/01/28 14:31:18 dholland 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 #include <miscfs/specfs/specdev.h> 75 76 #ifdef COMPAT_30 77 #include "opt_nfsserver.h" 78 #include <nfs/rpcv2.h> 79 #endif 80 #include <nfs/nfsproto.h> 81 #ifdef COMPAT_30 82 #include <nfs/nfs.h> 83 #include <nfs/nfs_var.h> 84 #endif 85 86 #if NFSS > 0 87 #include <dev/fssvar.h> 88 #endif 89 90 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 91 92 static int change_dir(struct nameidata *, struct lwp *); 93 static int change_flags(struct vnode *, u_long, struct lwp *); 94 static int change_mode(struct vnode *, int, struct lwp *l); 95 static int change_owner(struct vnode *, uid_t, gid_t, 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 mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE); 311 (void)vfs_busy(mp, LK_NOWAIT, 0); 312 313 mp->mnt_vnodecovered = vp; 314 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 315 mp->mnt_unmounter = NULL; 316 mount_initspecific(mp); 317 318 /* 319 * The underlying file system may refuse the mount for 320 * various reasons. Allow the user to force it to happen. 321 * 322 * Set the mount level flags. 323 */ 324 mp->mnt_flag = flags & 325 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 326 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 327 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 328 MNT_IGNORE | MNT_RDONLY); 329 330 error = VFS_MOUNT(mp, path, data, data_len); 331 mp->mnt_flag &= ~MNT_OP_FLAGS; 332 333 /* 334 * Put the new filesystem on the mount list after root. 335 */ 336 cache_purge(vp); 337 if (error != 0) { 338 vp->v_mountedhere = NULL; 339 mp->mnt_op->vfs_refcount--; 340 vfs_unbusy(mp); 341 vfs_destroy(mp); 342 return error; 343 } 344 345 mp->mnt_iflag &= ~IMNT_WANTRDWR; 346 vp->v_mountedhere = mp; 347 mutex_enter(&mountlist_lock); 348 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 349 mutex_exit(&mountlist_lock); 350 vn_restorerecurse(vp, recurse); 351 VOP_UNLOCK(vp, 0); 352 checkdirs(vp); 353 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 354 error = vfs_allocate_syncvnode(mp); 355 vfs_unbusy(mp); 356 (void) VFS_STATVFS(mp, &mp->mnt_stat); 357 error = VFS_START(mp, 0); 358 if (error) 359 vrele(vp); 360 *vpp = NULL; 361 return error; 362 } 363 364 static int 365 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 366 void *data, size_t *data_len) 367 { 368 struct mount *mp; 369 int error; 370 371 /* If MNT_GETARGS is specified, it should be the only flag. */ 372 if (flags & ~MNT_GETARGS) 373 return EINVAL; 374 375 mp = vp->v_mount; 376 377 /* XXX: probably some notion of "can see" here if we want isolation. */ 378 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 379 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 380 if (error) 381 return error; 382 383 if ((vp->v_vflag & VV_ROOT) == 0) 384 return EINVAL; 385 386 if (vfs_busy(mp, LK_NOWAIT, 0)) 387 return EPERM; 388 389 mp->mnt_flag &= ~MNT_OP_FLAGS; 390 mp->mnt_flag |= MNT_GETARGS; 391 error = VFS_MOUNT(mp, path, data, data_len); 392 mp->mnt_flag &= ~MNT_OP_FLAGS; 393 394 vfs_unbusy(mp); 395 return (error); 396 } 397 398 #ifdef COMPAT_40 399 /* ARGSUSED */ 400 int 401 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 402 { 403 /* { 404 syscallarg(const char *) type; 405 syscallarg(const char *) path; 406 syscallarg(int) flags; 407 syscallarg(void *) data; 408 } */ 409 register_t dummy; 410 411 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 412 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 413 } 414 #endif 415 416 int 417 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 418 { 419 /* { 420 syscallarg(const char *) type; 421 syscallarg(const char *) path; 422 syscallarg(int) flags; 423 syscallarg(void *) data; 424 syscallarg(size_t) data_len; 425 } */ 426 427 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 428 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 429 SCARG(uap, data_len), retval); 430 } 431 432 int 433 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 434 const char *path, int flags, void *data, enum uio_seg data_seg, 435 size_t data_len, register_t *retval) 436 { 437 struct vnode *vp; 438 struct nameidata nd; 439 void *data_buf = data; 440 u_int recurse; 441 int error; 442 443 /* 444 * Get vnode to be covered 445 */ 446 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 447 if ((error = namei(&nd)) != 0) 448 return (error); 449 vp = nd.ni_vp; 450 451 /* 452 * A lookup in VFS_MOUNT might result in an attempt to 453 * lock this vnode again, so make the lock recursive. 454 */ 455 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 456 recurse = vn_setrecurse(vp); 457 458 if (vfsops == NULL) { 459 if (flags & (MNT_GETARGS | MNT_UPDATE)) 460 vfsops = vp->v_mount->mnt_op; 461 else { 462 /* 'type' is userspace */ 463 error = mount_get_vfsops(type, &vfsops); 464 if (error != 0) 465 goto done; 466 } 467 } 468 469 if (data != NULL && data_seg == UIO_USERSPACE) { 470 if (data_len == 0) { 471 /* No length supplied, use default for filesystem */ 472 data_len = vfsops->vfs_min_mount_data; 473 if (data_len > VFS_MAX_MOUNT_DATA) { 474 /* maybe a force loaded old LKM */ 475 error = EINVAL; 476 goto done; 477 } 478 #ifdef COMPAT_30 479 /* Hopefully a longer buffer won't make copyin() fail */ 480 if (flags & MNT_UPDATE 481 && data_len < sizeof (struct mnt_export_args30)) 482 data_len = sizeof (struct mnt_export_args30); 483 #endif 484 } 485 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 486 487 /* NFS needs the buffer even for mnt_getargs .... */ 488 error = copyin(data, data_buf, data_len); 489 if (error != 0) 490 goto done; 491 } 492 493 if (flags & MNT_GETARGS) { 494 if (data_len == 0) { 495 error = EINVAL; 496 goto done; 497 } 498 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 499 if (error != 0) 500 goto done; 501 if (data_seg == UIO_USERSPACE) 502 error = copyout(data_buf, data, data_len); 503 *retval = data_len; 504 } else if (flags & MNT_UPDATE) { 505 error = mount_update(l, vp, path, flags, data_buf, &data_len); 506 } else { 507 /* Locking is handled internally in mount_domount(). */ 508 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 509 &data_len, recurse); 510 } 511 512 done: 513 if (vp != NULL) { 514 vn_restorerecurse(vp, recurse); 515 vput(vp); 516 } 517 if (data_buf != data) 518 free(data_buf, M_TEMP); 519 return (error); 520 } 521 522 /* 523 * Scan all active processes to see if any of them have a current 524 * or root directory onto which the new filesystem has just been 525 * mounted. If so, replace them with the new mount point. 526 */ 527 void 528 checkdirs(struct vnode *olddp) 529 { 530 struct cwdinfo *cwdi; 531 struct vnode *newdp; 532 struct proc *p; 533 534 if (olddp->v_usecount == 1) 535 return; 536 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 537 panic("mount: lost mount"); 538 mutex_enter(&proclist_lock); 539 PROCLIST_FOREACH(p, &allproc) { 540 cwdi = p->p_cwdi; 541 if (!cwdi) 542 continue; 543 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 544 if (cwdi->cwdi_cdir == olddp) { 545 vrele(cwdi->cwdi_cdir); 546 VREF(newdp); 547 cwdi->cwdi_cdir = newdp; 548 } 549 if (cwdi->cwdi_rdir == olddp) { 550 vrele(cwdi->cwdi_rdir); 551 VREF(newdp); 552 cwdi->cwdi_rdir = newdp; 553 } 554 rw_exit(&cwdi->cwdi_lock); 555 } 556 mutex_exit(&proclist_lock); 557 if (rootvnode == olddp) { 558 vrele(rootvnode); 559 VREF(newdp); 560 rootvnode = newdp; 561 } 562 vput(newdp); 563 } 564 565 /* 566 * Unmount a file system. 567 * 568 * Note: unmount takes a path to the vnode mounted on as argument, 569 * not special file (as before). 570 */ 571 /* ARGSUSED */ 572 int 573 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 574 { 575 /* { 576 syscallarg(const char *) path; 577 syscallarg(int) flags; 578 } */ 579 struct vnode *vp; 580 struct mount *mp; 581 int error; 582 struct nameidata nd; 583 584 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 585 SCARG(uap, path)); 586 if ((error = namei(&nd)) != 0) 587 return (error); 588 vp = nd.ni_vp; 589 mp = vp->v_mount; 590 591 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 592 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 593 if (error) { 594 vput(vp); 595 return (error); 596 } 597 598 /* 599 * Don't allow unmounting the root file system. 600 */ 601 if (mp->mnt_flag & MNT_ROOTFS) { 602 vput(vp); 603 return (EINVAL); 604 } 605 606 /* 607 * Must be the root of the filesystem 608 */ 609 if ((vp->v_vflag & VV_ROOT) == 0) { 610 vput(vp); 611 return (EINVAL); 612 } 613 vput(vp); 614 615 /* 616 * XXX Freeze syncer. Must do this before locking the 617 * mount point. See dounmount() for details. 618 */ 619 mutex_enter(&syncer_mutex); 620 621 if (vfs_busy(mp, 0, 0)) { 622 mutex_exit(&syncer_mutex); 623 return (EBUSY); 624 } 625 626 return (dounmount(mp, SCARG(uap, flags), l)); 627 } 628 629 /* 630 * Do the actual file system unmount. File system is assumed to have been 631 * marked busy by the caller. 632 */ 633 int 634 dounmount(struct mount *mp, int flags, struct lwp *l) 635 { 636 struct vnode *coveredvp; 637 int error; 638 int async; 639 int used_syncer; 640 641 #if NVERIEXEC > 0 642 error = veriexec_unmountchk(mp); 643 if (error) 644 return (error); 645 #endif /* NVERIEXEC > 0 */ 646 647 mutex_enter(&mountlist_lock); 648 vfs_unbusy(mp); 649 used_syncer = (mp->mnt_syncer != NULL); 650 651 /* 652 * XXX Syncer must be frozen when we get here. This should really 653 * be done on a per-mountpoint basis, but especially the softdep 654 * code possibly called from the syncer doesn't exactly work on a 655 * per-mountpoint basis, so the softdep code would become a maze 656 * of vfs_busy() calls. 657 * 658 * The caller of dounmount() must acquire syncer_mutex because 659 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 660 * order, and we must preserve that order to avoid deadlock. 661 * 662 * So, if the file system did not use the syncer, now is 663 * the time to release the syncer_mutex. 664 */ 665 if (used_syncer == 0) 666 mutex_exit(&syncer_mutex); 667 668 mp->mnt_iflag |= IMNT_UNMOUNT; 669 mp->mnt_unmounter = l; 670 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 671 672 async = mp->mnt_flag & MNT_ASYNC; 673 mp->mnt_flag &= ~MNT_ASYNC; 674 cache_purgevfs(mp); /* remove cache entries for this file sys */ 675 if (mp->mnt_syncer != NULL) 676 vfs_deallocate_syncvnode(mp); 677 error = 0; 678 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 679 #if NFSS > 0 680 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 681 #endif 682 if (error == 0) 683 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 684 } 685 vfs_scrubvnlist(mp); 686 if (error == 0 || (flags & MNT_FORCE)) 687 error = VFS_UNMOUNT(mp, flags); 688 if (error) { 689 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 690 (void) vfs_allocate_syncvnode(mp); 691 mutex_enter(&mountlist_lock); 692 mp->mnt_iflag &= ~IMNT_UNMOUNT; 693 mp->mnt_unmounter = NULL; 694 mp->mnt_flag |= async; 695 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 696 &mountlist_lock); 697 if (used_syncer) 698 mutex_exit(&syncer_mutex); 699 mutex_enter(&mp->mnt_mutex); 700 while (mp->mnt_wcnt > 0) { 701 wakeup(mp); 702 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 703 0, &mp->mnt_mutex); 704 } 705 mutex_exit(&mp->mnt_mutex); 706 return (error); 707 } 708 vfs_scrubvnlist(mp); 709 mutex_enter(&mountlist_lock); 710 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 711 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 712 coveredvp->v_mountedhere = NULL; 713 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 714 panic("unmount: dangling vnode"); 715 mp->mnt_iflag |= IMNT_GONE; 716 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 717 if (coveredvp != NULLVP) 718 vrele(coveredvp); 719 if (used_syncer) 720 mutex_exit(&syncer_mutex); 721 mutex_enter(&mp->mnt_mutex); 722 while (mp->mnt_wcnt > 0) { 723 wakeup(mp); 724 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 725 } 726 mutex_exit(&mp->mnt_mutex); 727 vfs_hooks_unmount(mp); 728 vfs_delref(mp->mnt_op); 729 vfs_destroy(mp); 730 return (0); 731 } 732 733 /* 734 * Sync each mounted filesystem. 735 */ 736 #ifdef DEBUG 737 int syncprt = 0; 738 struct ctldebug debug0 = { "syncprt", &syncprt }; 739 #endif 740 741 /* ARGSUSED */ 742 int 743 sys_sync(struct lwp *l, const void *v, register_t *retval) 744 { 745 struct mount *mp, *nmp; 746 int asyncflag; 747 748 if (l == NULL) 749 l = &lwp0; 750 751 mutex_enter(&mountlist_lock); 752 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 753 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 754 nmp = mp->mnt_list.cqe_prev; 755 continue; 756 } 757 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 758 asyncflag = mp->mnt_flag & MNT_ASYNC; 759 mp->mnt_flag &= ~MNT_ASYNC; 760 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 761 if (asyncflag) 762 mp->mnt_flag |= MNT_ASYNC; 763 } 764 mutex_enter(&mountlist_lock); 765 nmp = mp->mnt_list.cqe_prev; 766 vfs_unbusy(mp); 767 768 } 769 mutex_exit(&mountlist_lock); 770 #ifdef DEBUG 771 if (syncprt) 772 vfs_bufstats(); 773 #endif /* DEBUG */ 774 return (0); 775 } 776 777 /* 778 * Change filesystem quotas. 779 */ 780 /* ARGSUSED */ 781 int 782 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 783 { 784 /* { 785 syscallarg(const char *) path; 786 syscallarg(int) cmd; 787 syscallarg(int) uid; 788 syscallarg(void *) arg; 789 } */ 790 struct mount *mp; 791 int error; 792 struct nameidata nd; 793 794 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 795 SCARG(uap, path)); 796 if ((error = namei(&nd)) != 0) 797 return (error); 798 mp = nd.ni_vp->v_mount; 799 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 800 SCARG(uap, arg)); 801 vrele(nd.ni_vp); 802 return (error); 803 } 804 805 int 806 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 807 int root) 808 { 809 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 810 int error = 0; 811 812 /* 813 * If MNT_NOWAIT or MNT_LAZY is specified, do not 814 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 815 * overrides MNT_NOWAIT. 816 */ 817 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 818 (flags != MNT_WAIT && flags != 0)) { 819 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 820 goto done; 821 } 822 823 /* Get the filesystem stats now */ 824 memset(sp, 0, sizeof(*sp)); 825 if ((error = VFS_STATVFS(mp, sp)) != 0) { 826 return error; 827 } 828 829 if (cwdi->cwdi_rdir == NULL) 830 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 831 done: 832 if (cwdi->cwdi_rdir != NULL) { 833 size_t len; 834 char *bp; 835 char *path = PNBUF_GET(); 836 837 bp = path + MAXPATHLEN; 838 *--bp = '\0'; 839 rw_enter(&cwdi->cwdi_lock, RW_READER); 840 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 841 MAXPATHLEN / 2, 0, l); 842 rw_exit(&cwdi->cwdi_lock); 843 if (error) { 844 PNBUF_PUT(path); 845 return error; 846 } 847 len = strlen(bp); 848 /* 849 * for mount points that are below our root, we can see 850 * them, so we fix up the pathname and return them. The 851 * rest we cannot see, so we don't allow viewing the 852 * data. 853 */ 854 if (strncmp(bp, sp->f_mntonname, len) == 0) { 855 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 856 sizeof(sp->f_mntonname)); 857 if (sp->f_mntonname[0] == '\0') 858 (void)strlcpy(sp->f_mntonname, "/", 859 sizeof(sp->f_mntonname)); 860 } else { 861 if (root) 862 (void)strlcpy(sp->f_mntonname, "/", 863 sizeof(sp->f_mntonname)); 864 else 865 error = EPERM; 866 } 867 PNBUF_PUT(path); 868 } 869 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 870 return error; 871 } 872 873 /* 874 * Get filesystem statistics by path. 875 */ 876 int 877 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 878 { 879 struct mount *mp; 880 int error; 881 struct nameidata nd; 882 883 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 884 if ((error = namei(&nd)) != 0) 885 return error; 886 mp = nd.ni_vp->v_mount; 887 error = dostatvfs(mp, sb, l, flags, 1); 888 vrele(nd.ni_vp); 889 return error; 890 } 891 892 /* ARGSUSED */ 893 int 894 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 895 { 896 /* { 897 syscallarg(const char *) path; 898 syscallarg(struct statvfs *) buf; 899 syscallarg(int) flags; 900 } */ 901 struct statvfs *sb; 902 int error; 903 904 sb = STATVFSBUF_GET(); 905 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 906 if (error == 0) 907 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 908 STATVFSBUF_PUT(sb); 909 return error; 910 } 911 912 /* 913 * Get filesystem statistics by fd. 914 */ 915 int 916 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 917 { 918 struct proc *p = l->l_proc; 919 struct file *fp; 920 struct mount *mp; 921 int error; 922 923 /* getvnode() will use the descriptor for us */ 924 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 925 return (error); 926 mp = ((struct vnode *)fp->f_data)->v_mount; 927 error = dostatvfs(mp, sb, l, flags, 1); 928 FILE_UNUSE(fp, l); 929 return error; 930 } 931 932 /* ARGSUSED */ 933 int 934 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 935 { 936 /* { 937 syscallarg(int) fd; 938 syscallarg(struct statvfs *) buf; 939 syscallarg(int) flags; 940 } */ 941 struct statvfs *sb; 942 int error; 943 944 sb = STATVFSBUF_GET(); 945 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 946 if (error == 0) 947 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 948 STATVFSBUF_PUT(sb); 949 return error; 950 } 951 952 953 /* 954 * Get statistics on all filesystems. 955 */ 956 int 957 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 958 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 959 register_t *retval) 960 { 961 int root = 0; 962 struct proc *p = l->l_proc; 963 struct mount *mp, *nmp; 964 struct statvfs *sb; 965 size_t count, maxcount; 966 int error = 0; 967 968 sb = STATVFSBUF_GET(); 969 maxcount = bufsize / entry_sz; 970 mutex_enter(&mountlist_lock); 971 count = 0; 972 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 973 mp = nmp) { 974 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 975 nmp = CIRCLEQ_NEXT(mp, mnt_list); 976 continue; 977 } 978 if (sfsp && count < maxcount) { 979 error = dostatvfs(mp, sb, l, flags, 0); 980 if (error) { 981 mutex_enter(&mountlist_lock); 982 nmp = CIRCLEQ_NEXT(mp, mnt_list); 983 vfs_unbusy(mp); 984 continue; 985 } 986 error = copyfn(sb, sfsp, entry_sz); 987 if (error) { 988 vfs_unbusy(mp); 989 goto out; 990 } 991 sfsp = (char *)sfsp + entry_sz; 992 root |= strcmp(sb->f_mntonname, "/") == 0; 993 } 994 count++; 995 mutex_enter(&mountlist_lock); 996 nmp = CIRCLEQ_NEXT(mp, mnt_list); 997 vfs_unbusy(mp); 998 } 999 1000 mutex_exit(&mountlist_lock); 1001 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1002 /* 1003 * fake a root entry 1004 */ 1005 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1006 sb, l, flags, 1); 1007 if (error != 0) 1008 goto out; 1009 if (sfsp) 1010 error = copyfn(sb, sfsp, entry_sz); 1011 count++; 1012 } 1013 if (sfsp && count > maxcount) 1014 *retval = maxcount; 1015 else 1016 *retval = count; 1017 out: 1018 STATVFSBUF_PUT(sb); 1019 return error; 1020 } 1021 1022 int 1023 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1024 { 1025 /* { 1026 syscallarg(struct statvfs *) buf; 1027 syscallarg(size_t) bufsize; 1028 syscallarg(int) flags; 1029 } */ 1030 1031 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1032 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1033 } 1034 1035 /* 1036 * Change current working directory to a given file descriptor. 1037 */ 1038 /* ARGSUSED */ 1039 int 1040 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1041 { 1042 /* { 1043 syscallarg(int) fd; 1044 } */ 1045 struct proc *p = l->l_proc; 1046 struct filedesc *fdp = p->p_fd; 1047 struct cwdinfo *cwdi; 1048 struct vnode *vp, *tdp; 1049 struct mount *mp; 1050 struct file *fp; 1051 int error; 1052 1053 /* getvnode() will use the descriptor for us */ 1054 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1055 return (error); 1056 vp = (struct vnode *)fp->f_data; 1057 1058 VREF(vp); 1059 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1060 if (vp->v_type != VDIR) 1061 error = ENOTDIR; 1062 else 1063 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1064 if (error) { 1065 vput(vp); 1066 goto out; 1067 } 1068 while ((mp = vp->v_mountedhere) != NULL) { 1069 if (vfs_busy(mp, 0, 0)) 1070 continue; 1071 1072 vput(vp); 1073 error = VFS_ROOT(mp, &tdp); 1074 vfs_unbusy(mp); 1075 if (error) 1076 goto out; 1077 vp = tdp; 1078 } 1079 VOP_UNLOCK(vp, 0); 1080 1081 /* 1082 * Disallow changing to a directory not under the process's 1083 * current root directory (if there is one). 1084 */ 1085 cwdi = p->p_cwdi; 1086 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1087 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1088 vrele(vp); 1089 error = EPERM; /* operation not permitted */ 1090 } else { 1091 vrele(cwdi->cwdi_cdir); 1092 cwdi->cwdi_cdir = vp; 1093 } 1094 rw_exit(&cwdi->cwdi_lock); 1095 1096 out: 1097 FILE_UNUSE(fp, l); 1098 return (error); 1099 } 1100 1101 /* 1102 * Change this process's notion of the root directory to a given file 1103 * descriptor. 1104 */ 1105 int 1106 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1107 { 1108 struct proc *p = l->l_proc; 1109 struct filedesc *fdp = p->p_fd; 1110 struct cwdinfo *cwdi; 1111 struct vnode *vp; 1112 struct file *fp; 1113 int error; 1114 1115 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1116 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1117 return error; 1118 /* getvnode() will use the descriptor for us */ 1119 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1120 return error; 1121 vp = (struct vnode *) fp->f_data; 1122 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1123 if (vp->v_type != VDIR) 1124 error = ENOTDIR; 1125 else 1126 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1127 VOP_UNLOCK(vp, 0); 1128 if (error) 1129 goto out; 1130 VREF(vp); 1131 1132 /* 1133 * Prevent escaping from chroot by putting the root under 1134 * the working directory. Silently chdir to / if we aren't 1135 * already there. 1136 */ 1137 cwdi = p->p_cwdi; 1138 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1139 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1140 /* 1141 * XXX would be more failsafe to change directory to a 1142 * deadfs node here instead 1143 */ 1144 vrele(cwdi->cwdi_cdir); 1145 VREF(vp); 1146 cwdi->cwdi_cdir = vp; 1147 } 1148 1149 if (cwdi->cwdi_rdir != NULL) 1150 vrele(cwdi->cwdi_rdir); 1151 cwdi->cwdi_rdir = vp; 1152 rw_exit(&cwdi->cwdi_lock); 1153 1154 out: 1155 FILE_UNUSE(fp, l); 1156 return (error); 1157 } 1158 1159 /* 1160 * Change current working directory (``.''). 1161 */ 1162 /* ARGSUSED */ 1163 int 1164 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1165 { 1166 /* { 1167 syscallarg(const char *) path; 1168 } */ 1169 struct proc *p = l->l_proc; 1170 struct cwdinfo *cwdi; 1171 int error; 1172 struct nameidata nd; 1173 1174 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1175 SCARG(uap, path)); 1176 if ((error = change_dir(&nd, l)) != 0) 1177 return (error); 1178 cwdi = p->p_cwdi; 1179 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1180 vrele(cwdi->cwdi_cdir); 1181 cwdi->cwdi_cdir = nd.ni_vp; 1182 rw_exit(&cwdi->cwdi_lock); 1183 return (0); 1184 } 1185 1186 /* 1187 * Change notion of root (``/'') directory. 1188 */ 1189 /* ARGSUSED */ 1190 int 1191 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1192 { 1193 /* { 1194 syscallarg(const char *) path; 1195 } */ 1196 struct proc *p = l->l_proc; 1197 struct cwdinfo *cwdi; 1198 struct vnode *vp; 1199 int error; 1200 struct nameidata nd; 1201 1202 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1203 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1204 return (error); 1205 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1206 SCARG(uap, path)); 1207 if ((error = change_dir(&nd, l)) != 0) 1208 return (error); 1209 1210 cwdi = p->p_cwdi; 1211 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1212 if (cwdi->cwdi_rdir != NULL) 1213 vrele(cwdi->cwdi_rdir); 1214 vp = nd.ni_vp; 1215 cwdi->cwdi_rdir = vp; 1216 1217 /* 1218 * Prevent escaping from chroot by putting the root under 1219 * the working directory. Silently chdir to / if we aren't 1220 * already there. 1221 */ 1222 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1223 /* 1224 * XXX would be more failsafe to change directory to a 1225 * deadfs node here instead 1226 */ 1227 vrele(cwdi->cwdi_cdir); 1228 VREF(vp); 1229 cwdi->cwdi_cdir = vp; 1230 } 1231 rw_exit(&cwdi->cwdi_lock); 1232 1233 return (0); 1234 } 1235 1236 /* 1237 * Common routine for chroot and chdir. 1238 */ 1239 static int 1240 change_dir(struct nameidata *ndp, struct lwp *l) 1241 { 1242 struct vnode *vp; 1243 int error; 1244 1245 if ((error = namei(ndp)) != 0) 1246 return (error); 1247 vp = ndp->ni_vp; 1248 if (vp->v_type != VDIR) 1249 error = ENOTDIR; 1250 else 1251 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1252 1253 if (error) 1254 vput(vp); 1255 else 1256 VOP_UNLOCK(vp, 0); 1257 return (error); 1258 } 1259 1260 /* 1261 * Check permissions, allocate an open file structure, 1262 * and call the device open routine if any. 1263 */ 1264 int 1265 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1266 { 1267 /* { 1268 syscallarg(const char *) path; 1269 syscallarg(int) flags; 1270 syscallarg(int) mode; 1271 } */ 1272 struct proc *p = l->l_proc; 1273 struct cwdinfo *cwdi = p->p_cwdi; 1274 struct filedesc *fdp = p->p_fd; 1275 struct file *fp; 1276 struct vnode *vp; 1277 int flags, cmode; 1278 int type, indx, error; 1279 struct flock lf; 1280 struct nameidata nd; 1281 1282 flags = FFLAGS(SCARG(uap, flags)); 1283 if ((flags & (FREAD | FWRITE)) == 0) 1284 return (EINVAL); 1285 /* falloc() will use the file descriptor for us */ 1286 if ((error = falloc(l, &fp, &indx)) != 0) 1287 return (error); 1288 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1289 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1290 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1291 SCARG(uap, path)); 1292 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1293 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1294 rw_enter(&fdp->fd_lock, RW_WRITER); 1295 FILE_UNUSE(fp, l); 1296 fdp->fd_ofiles[indx] = NULL; 1297 rw_exit(&fdp->fd_lock); 1298 ffree(fp); 1299 if ((error == EDUPFD || error == EMOVEFD) && 1300 l->l_dupfd >= 0 && /* XXX from fdopen */ 1301 (error = 1302 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1303 *retval = indx; 1304 return (0); 1305 } 1306 if (error == ERESTART) 1307 error = EINTR; 1308 fdremove(fdp, indx); 1309 return (error); 1310 } 1311 1312 l->l_dupfd = 0; 1313 vp = nd.ni_vp; 1314 fp->f_flag = flags & FMASK; 1315 fp->f_type = DTYPE_VNODE; 1316 fp->f_ops = &vnops; 1317 fp->f_data = vp; 1318 if (flags & (O_EXLOCK | O_SHLOCK)) { 1319 lf.l_whence = SEEK_SET; 1320 lf.l_start = 0; 1321 lf.l_len = 0; 1322 if (flags & O_EXLOCK) 1323 lf.l_type = F_WRLCK; 1324 else 1325 lf.l_type = F_RDLCK; 1326 type = F_FLOCK; 1327 if ((flags & FNONBLOCK) == 0) 1328 type |= F_WAIT; 1329 VOP_UNLOCK(vp, 0); 1330 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1331 if (error) { 1332 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1333 FILE_UNUSE(fp, l); 1334 ffree(fp); 1335 fdremove(fdp, indx); 1336 return (error); 1337 } 1338 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1339 fp->f_flag |= FHASLOCK; 1340 } 1341 VOP_UNLOCK(vp, 0); 1342 *retval = indx; 1343 FILE_SET_MATURE(fp); 1344 FILE_UNUSE(fp, l); 1345 return (0); 1346 } 1347 1348 static void 1349 vfs__fhfree(fhandle_t *fhp) 1350 { 1351 size_t fhsize; 1352 1353 if (fhp == NULL) { 1354 return; 1355 } 1356 fhsize = FHANDLE_SIZE(fhp); 1357 kmem_free(fhp, fhsize); 1358 } 1359 1360 /* 1361 * vfs_composefh: compose a filehandle. 1362 */ 1363 1364 int 1365 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1366 { 1367 struct mount *mp; 1368 struct fid *fidp; 1369 int error; 1370 size_t needfhsize; 1371 size_t fidsize; 1372 1373 mp = vp->v_mount; 1374 fidp = NULL; 1375 if (*fh_size < FHANDLE_SIZE_MIN) { 1376 fidsize = 0; 1377 } else { 1378 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1379 if (fhp != NULL) { 1380 memset(fhp, 0, *fh_size); 1381 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1382 fidp = &fhp->fh_fid; 1383 } 1384 } 1385 error = VFS_VPTOFH(vp, fidp, &fidsize); 1386 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1387 if (error == 0 && *fh_size < needfhsize) { 1388 error = E2BIG; 1389 } 1390 *fh_size = needfhsize; 1391 return error; 1392 } 1393 1394 int 1395 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1396 { 1397 struct mount *mp; 1398 fhandle_t *fhp; 1399 size_t fhsize; 1400 size_t fidsize; 1401 int error; 1402 1403 *fhpp = NULL; 1404 mp = vp->v_mount; 1405 fidsize = 0; 1406 error = VFS_VPTOFH(vp, NULL, &fidsize); 1407 KASSERT(error != 0); 1408 if (error != E2BIG) { 1409 goto out; 1410 } 1411 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1412 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1413 if (fhp == NULL) { 1414 error = ENOMEM; 1415 goto out; 1416 } 1417 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1418 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1419 if (error == 0) { 1420 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1421 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1422 *fhpp = fhp; 1423 } else { 1424 kmem_free(fhp, fhsize); 1425 } 1426 out: 1427 return error; 1428 } 1429 1430 void 1431 vfs_composefh_free(fhandle_t *fhp) 1432 { 1433 1434 vfs__fhfree(fhp); 1435 } 1436 1437 /* 1438 * vfs_fhtovp: lookup a vnode by a filehandle. 1439 */ 1440 1441 int 1442 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1443 { 1444 struct mount *mp; 1445 int error; 1446 1447 *vpp = NULL; 1448 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1449 if (mp == NULL) { 1450 error = ESTALE; 1451 goto out; 1452 } 1453 if (mp->mnt_op->vfs_fhtovp == NULL) { 1454 error = EOPNOTSUPP; 1455 goto out; 1456 } 1457 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1458 out: 1459 return error; 1460 } 1461 1462 /* 1463 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1464 * the needed size. 1465 */ 1466 1467 int 1468 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1469 { 1470 fhandle_t *fhp; 1471 int error; 1472 1473 *fhpp = NULL; 1474 if (fhsize > FHANDLE_SIZE_MAX) { 1475 return EINVAL; 1476 } 1477 if (fhsize < FHANDLE_SIZE_MIN) { 1478 return EINVAL; 1479 } 1480 again: 1481 fhp = kmem_alloc(fhsize, KM_SLEEP); 1482 if (fhp == NULL) { 1483 return ENOMEM; 1484 } 1485 error = copyin(ufhp, fhp, fhsize); 1486 if (error == 0) { 1487 /* XXX this check shouldn't be here */ 1488 if (FHANDLE_SIZE(fhp) == fhsize) { 1489 *fhpp = fhp; 1490 return 0; 1491 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1492 /* 1493 * a kludge for nfsv2 padded handles. 1494 */ 1495 size_t sz; 1496 1497 sz = FHANDLE_SIZE(fhp); 1498 kmem_free(fhp, fhsize); 1499 fhsize = sz; 1500 goto again; 1501 } else { 1502 /* 1503 * userland told us wrong size. 1504 */ 1505 error = EINVAL; 1506 } 1507 } 1508 kmem_free(fhp, fhsize); 1509 return error; 1510 } 1511 1512 void 1513 vfs_copyinfh_free(fhandle_t *fhp) 1514 { 1515 1516 vfs__fhfree(fhp); 1517 } 1518 1519 /* 1520 * Get file handle system call 1521 */ 1522 int 1523 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1524 { 1525 /* { 1526 syscallarg(char *) fname; 1527 syscallarg(fhandle_t *) fhp; 1528 syscallarg(size_t *) fh_size; 1529 } */ 1530 struct vnode *vp; 1531 fhandle_t *fh; 1532 int error; 1533 struct nameidata nd; 1534 size_t sz; 1535 size_t usz; 1536 1537 /* 1538 * Must be super user 1539 */ 1540 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1541 0, NULL, NULL, NULL); 1542 if (error) 1543 return (error); 1544 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1545 SCARG(uap, fname)); 1546 error = namei(&nd); 1547 if (error) 1548 return (error); 1549 vp = nd.ni_vp; 1550 error = vfs_composefh_alloc(vp, &fh); 1551 vput(vp); 1552 if (error != 0) { 1553 goto out; 1554 } 1555 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1556 if (error != 0) { 1557 goto out; 1558 } 1559 sz = FHANDLE_SIZE(fh); 1560 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1561 if (error != 0) { 1562 goto out; 1563 } 1564 if (usz >= sz) { 1565 error = copyout(fh, SCARG(uap, fhp), sz); 1566 } else { 1567 error = E2BIG; 1568 } 1569 out: 1570 vfs_composefh_free(fh); 1571 return (error); 1572 } 1573 1574 /* 1575 * Open a file given a file handle. 1576 * 1577 * Check permissions, allocate an open file structure, 1578 * and call the device open routine if any. 1579 */ 1580 1581 int 1582 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1583 register_t *retval) 1584 { 1585 struct filedesc *fdp = l->l_proc->p_fd; 1586 struct file *fp; 1587 struct vnode *vp = NULL; 1588 kauth_cred_t cred = l->l_cred; 1589 struct file *nfp; 1590 int type, indx, error=0; 1591 struct flock lf; 1592 struct vattr va; 1593 fhandle_t *fh; 1594 int flags; 1595 1596 /* 1597 * Must be super user 1598 */ 1599 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1600 0, NULL, NULL, NULL))) 1601 return (error); 1602 1603 flags = FFLAGS(oflags); 1604 if ((flags & (FREAD | FWRITE)) == 0) 1605 return (EINVAL); 1606 if ((flags & O_CREAT)) 1607 return (EINVAL); 1608 /* falloc() will use the file descriptor for us */ 1609 if ((error = falloc(l, &nfp, &indx)) != 0) 1610 return (error); 1611 fp = nfp; 1612 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1613 if (error != 0) { 1614 goto bad; 1615 } 1616 error = vfs_fhtovp(fh, &vp); 1617 if (error != 0) { 1618 goto bad; 1619 } 1620 1621 /* Now do an effective vn_open */ 1622 1623 if (vp->v_type == VSOCK) { 1624 error = EOPNOTSUPP; 1625 goto bad; 1626 } 1627 error = vn_openchk(vp, cred, flags); 1628 if (error != 0) 1629 goto bad; 1630 if (flags & O_TRUNC) { 1631 VOP_UNLOCK(vp, 0); /* XXX */ 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, const struct sys___fhopen40_args *uap, register_t *retval) 1692 { 1693 /* { 1694 syscallarg(const void *) fhp; 1695 syscallarg(size_t) fh_size; 1696 syscallarg(int) flags; 1697 } */ 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, const struct sys___fhstat40_args *uap, register_t *retval) 1735 { 1736 /* { 1737 syscallarg(const void *) fhp; 1738 syscallarg(size_t) fh_size; 1739 syscallarg(struct stat *) sb; 1740 } */ 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, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1784 { 1785 /* { 1786 syscallarg(const void *) fhp; 1787 syscallarg(size_t) fh_size; 1788 syscallarg(struct statvfs *) buf; 1789 syscallarg(int) flags; 1790 } */ 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, const struct sys_mknod_args *uap, register_t *retval) 1808 { 1809 /* { 1810 syscallarg(const char *) path; 1811 syscallarg(int) mode; 1812 syscallarg(int) dev; 1813 } */ 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); 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 switch (optype) { 1873 case VOP_WHITEOUT_DESCOFFSET: 1874 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1875 if (error) 1876 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1877 vput(nd.ni_dvp); 1878 break; 1879 1880 case VOP_MKNOD_DESCOFFSET: 1881 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1882 &nd.ni_cnd, &vattr); 1883 if (error == 0) 1884 vput(nd.ni_vp); 1885 break; 1886 1887 case VOP_CREATE_DESCOFFSET: 1888 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1889 &nd.ni_cnd, &vattr); 1890 if (error == 0) 1891 vput(nd.ni_vp); 1892 break; 1893 } 1894 } else { 1895 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1896 if (nd.ni_dvp == vp) 1897 vrele(nd.ni_dvp); 1898 else 1899 vput(nd.ni_dvp); 1900 if (vp) 1901 vrele(vp); 1902 } 1903 out: 1904 VERIEXEC_PATH_PUT(path); 1905 return (error); 1906 } 1907 1908 /* 1909 * Create a named pipe. 1910 */ 1911 /* ARGSUSED */ 1912 int 1913 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1914 { 1915 /* { 1916 syscallarg(const char *) path; 1917 syscallarg(int) mode; 1918 } */ 1919 struct proc *p = l->l_proc; 1920 struct vattr vattr; 1921 int error; 1922 struct nameidata nd; 1923 1924 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1925 SCARG(uap, path)); 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 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1942 if (error == 0) 1943 vput(nd.ni_vp); 1944 return (error); 1945 } 1946 1947 /* 1948 * Make a hard file link. 1949 */ 1950 /* ARGSUSED */ 1951 int 1952 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1953 { 1954 /* { 1955 syscallarg(const char *) path; 1956 syscallarg(const char *) link; 1957 } */ 1958 struct vnode *vp; 1959 struct nameidata nd; 1960 int error; 1961 1962 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1963 SCARG(uap, path)); 1964 if ((error = namei(&nd)) != 0) 1965 return (error); 1966 vp = nd.ni_vp; 1967 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1968 SCARG(uap, link)); 1969 if ((error = namei(&nd)) != 0) 1970 goto out; 1971 if (nd.ni_vp) { 1972 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1973 if (nd.ni_dvp == nd.ni_vp) 1974 vrele(nd.ni_dvp); 1975 else 1976 vput(nd.ni_dvp); 1977 vrele(nd.ni_vp); 1978 error = EEXIST; 1979 goto out; 1980 } 1981 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1982 out: 1983 vrele(vp); 1984 return (error); 1985 } 1986 1987 /* 1988 * Make a symbolic link. 1989 */ 1990 /* ARGSUSED */ 1991 int 1992 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1993 { 1994 /* { 1995 syscallarg(const char *) path; 1996 syscallarg(const char *) link; 1997 } */ 1998 struct proc *p = l->l_proc; 1999 struct vattr vattr; 2000 char *path; 2001 int error; 2002 struct nameidata nd; 2003 2004 path = PNBUF_GET(); 2005 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2006 if (error) 2007 goto out; 2008 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2009 SCARG(uap, link)); 2010 if ((error = namei(&nd)) != 0) 2011 goto out; 2012 if (nd.ni_vp) { 2013 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2014 if (nd.ni_dvp == nd.ni_vp) 2015 vrele(nd.ni_dvp); 2016 else 2017 vput(nd.ni_dvp); 2018 vrele(nd.ni_vp); 2019 error = EEXIST; 2020 goto out; 2021 } 2022 VATTR_NULL(&vattr); 2023 vattr.va_type = VLNK; 2024 /* We will read cwdi->cwdi_cmask unlocked. */ 2025 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2026 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2027 if (error == 0) 2028 vput(nd.ni_vp); 2029 out: 2030 PNBUF_PUT(path); 2031 return (error); 2032 } 2033 2034 /* 2035 * Delete a whiteout from the filesystem. 2036 */ 2037 /* ARGSUSED */ 2038 int 2039 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2040 { 2041 /* { 2042 syscallarg(const char *) path; 2043 } */ 2044 int error; 2045 struct nameidata nd; 2046 2047 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2048 UIO_USERSPACE, SCARG(uap, path)); 2049 error = namei(&nd); 2050 if (error) 2051 return (error); 2052 2053 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2054 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2055 if (nd.ni_dvp == nd.ni_vp) 2056 vrele(nd.ni_dvp); 2057 else 2058 vput(nd.ni_dvp); 2059 if (nd.ni_vp) 2060 vrele(nd.ni_vp); 2061 return (EEXIST); 2062 } 2063 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2064 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2065 vput(nd.ni_dvp); 2066 return (error); 2067 } 2068 2069 /* 2070 * Delete a name from the filesystem. 2071 */ 2072 /* ARGSUSED */ 2073 int 2074 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2075 { 2076 /* { 2077 syscallarg(const char *) path; 2078 } */ 2079 2080 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2081 } 2082 2083 int 2084 do_sys_unlink(const char *arg, enum uio_seg seg) 2085 { 2086 struct vnode *vp; 2087 int error; 2088 struct nameidata nd; 2089 kauth_cred_t cred; 2090 char *path; 2091 const char *cpath; 2092 2093 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2094 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2095 2096 if ((error = namei(&nd)) != 0) 2097 goto out; 2098 vp = nd.ni_vp; 2099 2100 /* 2101 * The root of a mounted filesystem cannot be deleted. 2102 */ 2103 if (vp->v_vflag & VV_ROOT) { 2104 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2105 if (nd.ni_dvp == vp) 2106 vrele(nd.ni_dvp); 2107 else 2108 vput(nd.ni_dvp); 2109 vput(vp); 2110 error = EBUSY; 2111 goto out; 2112 } 2113 2114 #if NVERIEXEC > 0 2115 /* Handle remove requests for veriexec entries. */ 2116 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2117 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2118 if (nd.ni_dvp == vp) 2119 vrele(nd.ni_dvp); 2120 else 2121 vput(nd.ni_dvp); 2122 vput(vp); 2123 goto out; 2124 } 2125 #endif /* NVERIEXEC > 0 */ 2126 2127 cred = kauth_cred_get(); 2128 #ifdef FILEASSOC 2129 (void)fileassoc_file_delete(vp); 2130 #endif /* FILEASSOC */ 2131 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2132 out: 2133 VERIEXEC_PATH_PUT(path); 2134 return (error); 2135 } 2136 2137 /* 2138 * Reposition read/write file offset. 2139 */ 2140 int 2141 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2142 { 2143 /* { 2144 syscallarg(int) fd; 2145 syscallarg(int) pad; 2146 syscallarg(off_t) offset; 2147 syscallarg(int) whence; 2148 } */ 2149 struct proc *p = l->l_proc; 2150 kauth_cred_t cred = l->l_cred; 2151 struct filedesc *fdp = p->p_fd; 2152 struct file *fp; 2153 struct vnode *vp; 2154 struct vattr vattr; 2155 off_t newoff; 2156 int error; 2157 2158 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2159 return (EBADF); 2160 2161 vp = (struct vnode *)fp->f_data; 2162 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2163 error = ESPIPE; 2164 FILE_UNLOCK(fp); 2165 goto out; 2166 } 2167 2168 switch (SCARG(uap, whence)) { 2169 case SEEK_CUR: 2170 newoff = fp->f_offset + SCARG(uap, offset); 2171 FILE_USE(fp); 2172 break; 2173 case SEEK_END: 2174 FILE_USE(fp); 2175 error = VOP_GETATTR(vp, &vattr, cred); 2176 if (error) { 2177 FILE_UNUSE(fp, l); 2178 goto out; 2179 } 2180 newoff = SCARG(uap, offset) + vattr.va_size; 2181 break; 2182 case SEEK_SET: 2183 FILE_USE(fp); 2184 newoff = SCARG(uap, offset); 2185 break; 2186 default: 2187 FILE_UNLOCK(fp); 2188 error = EINVAL; 2189 goto out; 2190 } 2191 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2192 FILE_LOCK(fp); 2193 *(off_t *)retval = fp->f_offset = newoff; 2194 FILE_UNLOCK(fp); 2195 } 2196 FILE_UNUSE(fp, l); 2197 out: 2198 return (error); 2199 } 2200 2201 /* 2202 * Positional read system call. 2203 */ 2204 int 2205 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2206 { 2207 /* { 2208 syscallarg(int) fd; 2209 syscallarg(void *) buf; 2210 syscallarg(size_t) nbyte; 2211 syscallarg(off_t) offset; 2212 } */ 2213 struct proc *p = l->l_proc; 2214 struct filedesc *fdp = p->p_fd; 2215 struct file *fp; 2216 struct vnode *vp; 2217 off_t offset; 2218 int error, fd = SCARG(uap, fd); 2219 2220 if ((fp = fd_getfile(fdp, fd)) == NULL) 2221 return (EBADF); 2222 2223 if ((fp->f_flag & FREAD) == 0) { 2224 FILE_UNLOCK(fp); 2225 return (EBADF); 2226 } 2227 2228 FILE_USE(fp); 2229 2230 vp = (struct vnode *)fp->f_data; 2231 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2232 error = ESPIPE; 2233 goto out; 2234 } 2235 2236 offset = SCARG(uap, offset); 2237 2238 /* 2239 * XXX This works because no file systems actually 2240 * XXX take any action on the seek operation. 2241 */ 2242 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2243 goto out; 2244 2245 /* dofileread() will unuse the descriptor for us */ 2246 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2247 &offset, 0, retval)); 2248 2249 out: 2250 FILE_UNUSE(fp, l); 2251 return (error); 2252 } 2253 2254 /* 2255 * Positional scatter read system call. 2256 */ 2257 int 2258 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2259 { 2260 /* { 2261 syscallarg(int) fd; 2262 syscallarg(const struct iovec *) iovp; 2263 syscallarg(int) iovcnt; 2264 syscallarg(off_t) offset; 2265 } */ 2266 off_t offset = SCARG(uap, offset); 2267 2268 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2269 SCARG(uap, iovcnt), &offset, 0, retval); 2270 } 2271 2272 /* 2273 * Positional write system call. 2274 */ 2275 int 2276 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2277 { 2278 /* { 2279 syscallarg(int) fd; 2280 syscallarg(const void *) buf; 2281 syscallarg(size_t) nbyte; 2282 syscallarg(off_t) offset; 2283 } */ 2284 struct proc *p = l->l_proc; 2285 struct filedesc *fdp = p->p_fd; 2286 struct file *fp; 2287 struct vnode *vp; 2288 off_t offset; 2289 int error, fd = SCARG(uap, fd); 2290 2291 if ((fp = fd_getfile(fdp, fd)) == NULL) 2292 return (EBADF); 2293 2294 if ((fp->f_flag & FWRITE) == 0) { 2295 FILE_UNLOCK(fp); 2296 return (EBADF); 2297 } 2298 2299 FILE_USE(fp); 2300 2301 vp = (struct vnode *)fp->f_data; 2302 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2303 error = ESPIPE; 2304 goto out; 2305 } 2306 2307 offset = SCARG(uap, offset); 2308 2309 /* 2310 * XXX This works because no file systems actually 2311 * XXX take any action on the seek operation. 2312 */ 2313 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2314 goto out; 2315 2316 /* dofilewrite() will unuse the descriptor for us */ 2317 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2318 &offset, 0, retval)); 2319 2320 out: 2321 FILE_UNUSE(fp, l); 2322 return (error); 2323 } 2324 2325 /* 2326 * Positional gather write system call. 2327 */ 2328 int 2329 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2330 { 2331 /* { 2332 syscallarg(int) fd; 2333 syscallarg(const struct iovec *) iovp; 2334 syscallarg(int) iovcnt; 2335 syscallarg(off_t) offset; 2336 } */ 2337 off_t offset = SCARG(uap, offset); 2338 2339 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2340 SCARG(uap, iovcnt), &offset, 0, retval); 2341 } 2342 2343 /* 2344 * Check access permissions. 2345 */ 2346 int 2347 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2348 { 2349 /* { 2350 syscallarg(const char *) path; 2351 syscallarg(int) flags; 2352 } */ 2353 kauth_cred_t cred; 2354 struct vnode *vp; 2355 int error, flags; 2356 struct nameidata nd; 2357 2358 cred = kauth_cred_dup(l->l_cred); 2359 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2360 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2361 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2362 SCARG(uap, path)); 2363 /* Override default credentials */ 2364 nd.ni_cnd.cn_cred = cred; 2365 if ((error = namei(&nd)) != 0) 2366 goto out; 2367 vp = nd.ni_vp; 2368 2369 /* Flags == 0 means only check for existence. */ 2370 if (SCARG(uap, flags)) { 2371 flags = 0; 2372 if (SCARG(uap, flags) & R_OK) 2373 flags |= VREAD; 2374 if (SCARG(uap, flags) & W_OK) 2375 flags |= VWRITE; 2376 if (SCARG(uap, flags) & X_OK) 2377 flags |= VEXEC; 2378 2379 error = VOP_ACCESS(vp, flags, cred); 2380 if (!error && (flags & VWRITE)) 2381 error = vn_writechk(vp); 2382 } 2383 vput(vp); 2384 out: 2385 kauth_cred_free(cred); 2386 return (error); 2387 } 2388 2389 /* 2390 * Common code for all sys_stat functions, including compat versions. 2391 */ 2392 int 2393 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2394 struct stat *sb) 2395 { 2396 int error; 2397 struct nameidata nd; 2398 2399 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2400 UIO_USERSPACE, path); 2401 error = namei(&nd); 2402 if (error != 0) 2403 return error; 2404 error = vn_stat(nd.ni_vp, sb, l); 2405 vput(nd.ni_vp); 2406 return error; 2407 } 2408 2409 /* 2410 * Get file status; this version follows links. 2411 */ 2412 /* ARGSUSED */ 2413 int 2414 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2415 { 2416 /* { 2417 syscallarg(const char *) path; 2418 syscallarg(struct stat *) ub; 2419 } */ 2420 struct stat sb; 2421 int error; 2422 2423 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2424 if (error) 2425 return error; 2426 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2427 } 2428 2429 /* 2430 * Get file status; this version does not follow links. 2431 */ 2432 /* ARGSUSED */ 2433 int 2434 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2435 { 2436 /* { 2437 syscallarg(const char *) path; 2438 syscallarg(struct stat *) ub; 2439 } */ 2440 struct stat sb; 2441 int error; 2442 2443 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2444 if (error) 2445 return error; 2446 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2447 } 2448 2449 /* 2450 * Get configurable pathname variables. 2451 */ 2452 /* ARGSUSED */ 2453 int 2454 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2455 { 2456 /* { 2457 syscallarg(const char *) path; 2458 syscallarg(int) name; 2459 } */ 2460 int error; 2461 struct nameidata nd; 2462 2463 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2464 SCARG(uap, path)); 2465 if ((error = namei(&nd)) != 0) 2466 return (error); 2467 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2468 vput(nd.ni_vp); 2469 return (error); 2470 } 2471 2472 /* 2473 * Return target name of a symbolic link. 2474 */ 2475 /* ARGSUSED */ 2476 int 2477 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2478 { 2479 /* { 2480 syscallarg(const char *) path; 2481 syscallarg(char *) buf; 2482 syscallarg(size_t) count; 2483 } */ 2484 struct vnode *vp; 2485 struct iovec aiov; 2486 struct uio auio; 2487 int error; 2488 struct nameidata nd; 2489 2490 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2491 SCARG(uap, path)); 2492 if ((error = namei(&nd)) != 0) 2493 return (error); 2494 vp = nd.ni_vp; 2495 if (vp->v_type != VLNK) 2496 error = EINVAL; 2497 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2498 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2499 aiov.iov_base = SCARG(uap, buf); 2500 aiov.iov_len = SCARG(uap, count); 2501 auio.uio_iov = &aiov; 2502 auio.uio_iovcnt = 1; 2503 auio.uio_offset = 0; 2504 auio.uio_rw = UIO_READ; 2505 KASSERT(l == curlwp); 2506 auio.uio_vmspace = l->l_proc->p_vmspace; 2507 auio.uio_resid = SCARG(uap, count); 2508 error = VOP_READLINK(vp, &auio, l->l_cred); 2509 } 2510 vput(vp); 2511 *retval = SCARG(uap, count) - auio.uio_resid; 2512 return (error); 2513 } 2514 2515 /* 2516 * Change flags of a file given a path name. 2517 */ 2518 /* ARGSUSED */ 2519 int 2520 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2521 { 2522 /* { 2523 syscallarg(const char *) path; 2524 syscallarg(u_long) flags; 2525 } */ 2526 struct vnode *vp; 2527 int error; 2528 struct nameidata nd; 2529 2530 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2531 SCARG(uap, path)); 2532 if ((error = namei(&nd)) != 0) 2533 return (error); 2534 vp = nd.ni_vp; 2535 error = change_flags(vp, SCARG(uap, flags), l); 2536 vput(vp); 2537 return (error); 2538 } 2539 2540 /* 2541 * Change flags of a file given a file descriptor. 2542 */ 2543 /* ARGSUSED */ 2544 int 2545 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2546 { 2547 /* { 2548 syscallarg(int) fd; 2549 syscallarg(u_long) flags; 2550 } */ 2551 struct proc *p = l->l_proc; 2552 struct vnode *vp; 2553 struct file *fp; 2554 int error; 2555 2556 /* getvnode() will use the descriptor for us */ 2557 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2558 return (error); 2559 vp = (struct vnode *)fp->f_data; 2560 error = change_flags(vp, SCARG(uap, flags), l); 2561 VOP_UNLOCK(vp, 0); 2562 FILE_UNUSE(fp, l); 2563 return (error); 2564 } 2565 2566 /* 2567 * Change flags of a file given a path name; this version does 2568 * not follow links. 2569 */ 2570 int 2571 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2572 { 2573 /* { 2574 syscallarg(const char *) path; 2575 syscallarg(u_long) flags; 2576 } */ 2577 struct vnode *vp; 2578 int error; 2579 struct nameidata nd; 2580 2581 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2582 SCARG(uap, path)); 2583 if ((error = namei(&nd)) != 0) 2584 return (error); 2585 vp = nd.ni_vp; 2586 error = change_flags(vp, SCARG(uap, flags), l); 2587 vput(vp); 2588 return (error); 2589 } 2590 2591 /* 2592 * Common routine to change flags of a file. 2593 */ 2594 int 2595 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2596 { 2597 struct vattr vattr; 2598 int error; 2599 2600 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2601 /* 2602 * Non-superusers cannot change the flags on devices, even if they 2603 * own them. 2604 */ 2605 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2606 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2607 goto out; 2608 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2609 error = EINVAL; 2610 goto out; 2611 } 2612 } 2613 VATTR_NULL(&vattr); 2614 vattr.va_flags = flags; 2615 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2616 out: 2617 return (error); 2618 } 2619 2620 /* 2621 * Change mode of a file given path name; this version follows links. 2622 */ 2623 /* ARGSUSED */ 2624 int 2625 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2626 { 2627 /* { 2628 syscallarg(const char *) path; 2629 syscallarg(int) mode; 2630 } */ 2631 int error; 2632 struct nameidata nd; 2633 2634 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2635 SCARG(uap, path)); 2636 if ((error = namei(&nd)) != 0) 2637 return (error); 2638 2639 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2640 2641 vrele(nd.ni_vp); 2642 return (error); 2643 } 2644 2645 /* 2646 * Change mode of a file given a file descriptor. 2647 */ 2648 /* ARGSUSED */ 2649 int 2650 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2651 { 2652 /* { 2653 syscallarg(int) fd; 2654 syscallarg(int) mode; 2655 } */ 2656 struct proc *p = l->l_proc; 2657 struct file *fp; 2658 int error; 2659 2660 /* getvnode() will use the descriptor for us */ 2661 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2662 return (error); 2663 2664 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2665 FILE_UNUSE(fp, l); 2666 return (error); 2667 } 2668 2669 /* 2670 * Change mode of a file given path name; this version does not follow links. 2671 */ 2672 /* ARGSUSED */ 2673 int 2674 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2675 { 2676 /* { 2677 syscallarg(const char *) path; 2678 syscallarg(int) mode; 2679 } */ 2680 int error; 2681 struct nameidata nd; 2682 2683 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2684 SCARG(uap, path)); 2685 if ((error = namei(&nd)) != 0) 2686 return (error); 2687 2688 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2689 2690 vrele(nd.ni_vp); 2691 return (error); 2692 } 2693 2694 /* 2695 * Common routine to set mode given a vnode. 2696 */ 2697 static int 2698 change_mode(struct vnode *vp, int mode, struct lwp *l) 2699 { 2700 struct vattr vattr; 2701 int error; 2702 2703 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2704 VATTR_NULL(&vattr); 2705 vattr.va_mode = mode & ALLPERMS; 2706 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2707 VOP_UNLOCK(vp, 0); 2708 return (error); 2709 } 2710 2711 /* 2712 * Set ownership given a path name; this version follows links. 2713 */ 2714 /* ARGSUSED */ 2715 int 2716 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2717 { 2718 /* { 2719 syscallarg(const char *) path; 2720 syscallarg(uid_t) uid; 2721 syscallarg(gid_t) gid; 2722 } */ 2723 int error; 2724 struct nameidata nd; 2725 2726 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2727 SCARG(uap, path)); 2728 if ((error = namei(&nd)) != 0) 2729 return (error); 2730 2731 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2732 2733 vrele(nd.ni_vp); 2734 return (error); 2735 } 2736 2737 /* 2738 * Set ownership given a path name; this version follows links. 2739 * Provides POSIX semantics. 2740 */ 2741 /* ARGSUSED */ 2742 int 2743 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2744 { 2745 /* { 2746 syscallarg(const char *) path; 2747 syscallarg(uid_t) uid; 2748 syscallarg(gid_t) gid; 2749 } */ 2750 int error; 2751 struct nameidata nd; 2752 2753 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2754 SCARG(uap, path)); 2755 if ((error = namei(&nd)) != 0) 2756 return (error); 2757 2758 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2759 2760 vrele(nd.ni_vp); 2761 return (error); 2762 } 2763 2764 /* 2765 * Set ownership given a file descriptor. 2766 */ 2767 /* ARGSUSED */ 2768 int 2769 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2770 { 2771 /* { 2772 syscallarg(int) fd; 2773 syscallarg(uid_t) uid; 2774 syscallarg(gid_t) gid; 2775 } */ 2776 struct proc *p = l->l_proc; 2777 int error; 2778 struct file *fp; 2779 2780 /* getvnode() will use the descriptor for us */ 2781 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2782 return (error); 2783 2784 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2785 SCARG(uap, gid), l, 0); 2786 FILE_UNUSE(fp, l); 2787 return (error); 2788 } 2789 2790 /* 2791 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2792 */ 2793 /* ARGSUSED */ 2794 int 2795 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2796 { 2797 /* { 2798 syscallarg(int) fd; 2799 syscallarg(uid_t) uid; 2800 syscallarg(gid_t) gid; 2801 } */ 2802 struct proc *p = l->l_proc; 2803 int error; 2804 struct file *fp; 2805 2806 /* getvnode() will use the descriptor for us */ 2807 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2808 return (error); 2809 2810 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2811 SCARG(uap, gid), l, 1); 2812 FILE_UNUSE(fp, l); 2813 return (error); 2814 } 2815 2816 /* 2817 * Set ownership given a path name; this version does not follow links. 2818 */ 2819 /* ARGSUSED */ 2820 int 2821 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2822 { 2823 /* { 2824 syscallarg(const char *) path; 2825 syscallarg(uid_t) uid; 2826 syscallarg(gid_t) gid; 2827 } */ 2828 int error; 2829 struct nameidata nd; 2830 2831 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2832 SCARG(uap, path)); 2833 if ((error = namei(&nd)) != 0) 2834 return (error); 2835 2836 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2837 2838 vrele(nd.ni_vp); 2839 return (error); 2840 } 2841 2842 /* 2843 * Set ownership given a path name; this version does not follow links. 2844 * Provides POSIX/XPG semantics. 2845 */ 2846 /* ARGSUSED */ 2847 int 2848 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2849 { 2850 /* { 2851 syscallarg(const char *) path; 2852 syscallarg(uid_t) uid; 2853 syscallarg(gid_t) gid; 2854 } */ 2855 int error; 2856 struct nameidata nd; 2857 2858 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2859 SCARG(uap, path)); 2860 if ((error = namei(&nd)) != 0) 2861 return (error); 2862 2863 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2864 2865 vrele(nd.ni_vp); 2866 return (error); 2867 } 2868 2869 /* 2870 * Common routine to set ownership given a vnode. 2871 */ 2872 static int 2873 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2874 int posix_semantics) 2875 { 2876 struct vattr vattr; 2877 mode_t newmode; 2878 int error; 2879 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, const struct sys_utimes_args *uap, register_t *retval) 2930 { 2931 /* { 2932 syscallarg(const char *) path; 2933 syscallarg(const struct timeval *) tptr; 2934 } */ 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, const struct sys_futimes_args *uap, register_t *retval) 2946 { 2947 /* { 2948 syscallarg(int) fd; 2949 syscallarg(const struct timeval *) tptr; 2950 } */ 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, const struct sys_lutimes_args *uap, register_t *retval) 2971 { 2972 /* { 2973 syscallarg(const char *) path; 2974 syscallarg(const struct timeval *) tptr; 2975 } */ 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); 3012 if ((error = namei(&nd)) != 0) 3013 return (error); 3014 vp = nd.ni_vp; 3015 } else 3016 nd.ni_vp = NULL; 3017 3018 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3019 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3020 VOP_UNLOCK(vp, 0); 3021 3022 if (nd.ni_vp != NULL) 3023 vrele(nd.ni_vp); 3024 3025 return (error); 3026 } 3027 3028 /* 3029 * Truncate a file given its path name. 3030 */ 3031 /* ARGSUSED */ 3032 int 3033 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3034 { 3035 /* { 3036 syscallarg(const char *) path; 3037 syscallarg(int) pad; 3038 syscallarg(off_t) length; 3039 } */ 3040 struct vnode *vp; 3041 struct vattr vattr; 3042 int error; 3043 struct nameidata nd; 3044 3045 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3046 SCARG(uap, path)); 3047 if ((error = namei(&nd)) != 0) 3048 return (error); 3049 vp = nd.ni_vp; 3050 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3051 if (vp->v_type == VDIR) 3052 error = EISDIR; 3053 else if ((error = vn_writechk(vp)) == 0 && 3054 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3055 VATTR_NULL(&vattr); 3056 vattr.va_size = SCARG(uap, length); 3057 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3058 } 3059 vput(vp); 3060 return (error); 3061 } 3062 3063 /* 3064 * Truncate a file given a file descriptor. 3065 */ 3066 /* ARGSUSED */ 3067 int 3068 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3069 { 3070 /* { 3071 syscallarg(int) fd; 3072 syscallarg(int) pad; 3073 syscallarg(off_t) length; 3074 } */ 3075 struct proc *p = l->l_proc; 3076 struct vattr vattr; 3077 struct vnode *vp; 3078 struct file *fp; 3079 int error; 3080 3081 /* getvnode() will use the descriptor for us */ 3082 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3083 return (error); 3084 if ((fp->f_flag & FWRITE) == 0) { 3085 error = EINVAL; 3086 goto out; 3087 } 3088 vp = (struct vnode *)fp->f_data; 3089 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3090 if (vp->v_type == VDIR) 3091 error = EISDIR; 3092 else if ((error = vn_writechk(vp)) == 0) { 3093 VATTR_NULL(&vattr); 3094 vattr.va_size = SCARG(uap, length); 3095 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3096 } 3097 VOP_UNLOCK(vp, 0); 3098 out: 3099 FILE_UNUSE(fp, l); 3100 return (error); 3101 } 3102 3103 /* 3104 * Sync an open file. 3105 */ 3106 /* ARGSUSED */ 3107 int 3108 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3109 { 3110 /* { 3111 syscallarg(int) fd; 3112 } */ 3113 struct proc *p = l->l_proc; 3114 struct vnode *vp; 3115 struct file *fp; 3116 int error; 3117 3118 /* getvnode() will use the descriptor for us */ 3119 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3120 return (error); 3121 vp = (struct vnode *)fp->f_data; 3122 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3123 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3124 if (error == 0 && bioopsp != NULL && 3125 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3126 (*bioopsp->io_fsync)(vp, 0); 3127 VOP_UNLOCK(vp, 0); 3128 FILE_UNUSE(fp, l); 3129 return (error); 3130 } 3131 3132 /* 3133 * Sync a range of file data. API modeled after that found in AIX. 3134 * 3135 * FDATASYNC indicates that we need only save enough metadata to be able 3136 * to re-read the written data. Note we duplicate AIX's requirement that 3137 * the file be open for writing. 3138 */ 3139 /* ARGSUSED */ 3140 int 3141 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3142 { 3143 /* { 3144 syscallarg(int) fd; 3145 syscallarg(int) flags; 3146 syscallarg(off_t) start; 3147 syscallarg(off_t) length; 3148 } */ 3149 struct proc *p = l->l_proc; 3150 struct vnode *vp; 3151 struct file *fp; 3152 int flags, nflags; 3153 off_t s, e, len; 3154 int error; 3155 3156 /* getvnode() will use the descriptor for us */ 3157 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3158 return (error); 3159 3160 if ((fp->f_flag & FWRITE) == 0) { 3161 error = EBADF; 3162 goto out; 3163 } 3164 3165 flags = SCARG(uap, flags); 3166 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3167 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3168 error = EINVAL; 3169 goto out; 3170 } 3171 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3172 if (flags & FDATASYNC) 3173 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3174 else 3175 nflags = FSYNC_WAIT; 3176 if (flags & FDISKSYNC) 3177 nflags |= FSYNC_CACHE; 3178 3179 len = SCARG(uap, length); 3180 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3181 if (len) { 3182 s = SCARG(uap, start); 3183 e = s + len; 3184 if (e < s) { 3185 error = EINVAL; 3186 goto out; 3187 } 3188 } else { 3189 e = 0; 3190 s = 0; 3191 } 3192 3193 vp = (struct vnode *)fp->f_data; 3194 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3195 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3196 3197 if (error == 0 && bioopsp != NULL && 3198 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3199 (*bioopsp->io_fsync)(vp, nflags); 3200 3201 VOP_UNLOCK(vp, 0); 3202 out: 3203 FILE_UNUSE(fp, l); 3204 return (error); 3205 } 3206 3207 /* 3208 * Sync the data of an open file. 3209 */ 3210 /* ARGSUSED */ 3211 int 3212 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3213 { 3214 /* { 3215 syscallarg(int) fd; 3216 } */ 3217 struct proc *p = l->l_proc; 3218 struct vnode *vp; 3219 struct file *fp; 3220 int error; 3221 3222 /* getvnode() will use the descriptor for us */ 3223 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3224 return (error); 3225 if ((fp->f_flag & FWRITE) == 0) { 3226 FILE_UNUSE(fp, l); 3227 return (EBADF); 3228 } 3229 vp = (struct vnode *)fp->f_data; 3230 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3231 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3232 VOP_UNLOCK(vp, 0); 3233 FILE_UNUSE(fp, l); 3234 return (error); 3235 } 3236 3237 /* 3238 * Rename files, (standard) BSD semantics frontend. 3239 */ 3240 /* ARGSUSED */ 3241 int 3242 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3243 { 3244 /* { 3245 syscallarg(const char *) from; 3246 syscallarg(const char *) to; 3247 } */ 3248 3249 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3250 } 3251 3252 /* 3253 * Rename files, POSIX semantics frontend. 3254 */ 3255 /* ARGSUSED */ 3256 int 3257 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3258 { 3259 /* { 3260 syscallarg(const char *) from; 3261 syscallarg(const char *) to; 3262 } */ 3263 3264 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3265 } 3266 3267 /* 3268 * Rename files. Source and destination must either both be directories, 3269 * or both not be directories. If target is a directory, it must be empty. 3270 * If `from' and `to' refer to the same object, the value of the `retain' 3271 * argument is used to determine whether `from' will be 3272 * 3273 * (retain == 0) deleted unless `from' and `to' refer to the same 3274 * object in the file system's name space (BSD). 3275 * (retain == 1) always retained (POSIX). 3276 */ 3277 int 3278 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3279 { 3280 struct vnode *tvp, *fvp, *tdvp; 3281 struct nameidata fromnd, tond; 3282 struct mount *fs; 3283 struct lwp *l = curlwp; 3284 struct proc *p; 3285 uint32_t saveflag; 3286 int error; 3287 3288 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3289 seg, from); 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 3296 fs = fvp->v_mount; 3297 error = VFS_RENAMELOCK_ENTER(fs); 3298 if (error) { 3299 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3300 vrele(fromnd.ni_dvp); 3301 vrele(fvp); 3302 goto out1; 3303 } 3304 3305 /* 3306 * close, partially, yet another race - ideally we should only 3307 * go as far as getting fromnd.ni_dvp before getting the per-fs 3308 * lock, and then continue to get fromnd.ni_vp, but we can't do 3309 * that with namei as it stands. 3310 * 3311 * This still won't prevent rmdir from nuking fromnd.ni_vp 3312 * under us. The real fix is to get the locks in the right 3313 * order and do the lookups in the right places, but that's a 3314 * major rototill. 3315 * 3316 * Preserve the SAVESTART in cn_flags, because who knows what 3317 * might happen if we don't. 3318 * 3319 * Note: this logic (as well as this whole function) is cloned 3320 * in nfs_serv.c. Proceed accordingly. 3321 */ 3322 vrele(fvp); 3323 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3324 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3325 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3326 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3327 fromnd.ni_cnd.cn_flags |= saveflag; 3328 if (error) { 3329 VOP_UNLOCK(fromnd.ni_dvp, 0); 3330 VFS_RENAMELOCK_EXIT(fs); 3331 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3332 vrele(fromnd.ni_dvp); 3333 goto out1; 3334 } 3335 VOP_UNLOCK(fromnd.ni_vp, 0); 3336 if (fromnd.ni_dvp != fromnd.ni_vp) 3337 VOP_UNLOCK(fromnd.ni_dvp, 0); 3338 fvp = fromnd.ni_vp; 3339 3340 NDINIT(&tond, RENAME, 3341 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3342 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3343 seg, to); 3344 if ((error = namei(&tond)) != 0) { 3345 VFS_RENAMELOCK_EXIT(fs); 3346 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3347 vrele(fromnd.ni_dvp); 3348 vrele(fvp); 3349 goto out1; 3350 } 3351 tdvp = tond.ni_dvp; 3352 tvp = tond.ni_vp; 3353 3354 if (tvp != NULL) { 3355 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3356 error = ENOTDIR; 3357 goto out; 3358 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3359 error = EISDIR; 3360 goto out; 3361 } 3362 } 3363 3364 if (fvp == tdvp) 3365 error = EINVAL; 3366 3367 /* 3368 * Source and destination refer to the same object. 3369 */ 3370 if (fvp == tvp) { 3371 if (retain) 3372 error = -1; 3373 else if (fromnd.ni_dvp == tdvp && 3374 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3375 !memcmp(fromnd.ni_cnd.cn_nameptr, 3376 tond.ni_cnd.cn_nameptr, 3377 fromnd.ni_cnd.cn_namelen)) 3378 error = -1; 3379 } 3380 3381 #if NVERIEXEC > 0 3382 if (!error) { 3383 char *f1, *f2; 3384 3385 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3386 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3387 3388 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3389 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3390 3391 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3392 3393 free(f1, M_TEMP); 3394 free(f2, M_TEMP); 3395 } 3396 #endif /* NVERIEXEC > 0 */ 3397 3398 out: 3399 p = l->l_proc; 3400 if (!error) { 3401 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3402 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3403 VFS_RENAMELOCK_EXIT(fs); 3404 } else { 3405 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3406 if (tdvp == tvp) 3407 vrele(tdvp); 3408 else 3409 vput(tdvp); 3410 if (tvp) 3411 vput(tvp); 3412 VFS_RENAMELOCK_EXIT(fs); 3413 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3414 vrele(fromnd.ni_dvp); 3415 vrele(fvp); 3416 } 3417 vrele(tond.ni_startdir); 3418 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3419 out1: 3420 if (fromnd.ni_startdir) 3421 vrele(fromnd.ni_startdir); 3422 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3423 return (error == -1 ? 0 : error); 3424 } 3425 3426 /* 3427 * Make a directory file. 3428 */ 3429 /* ARGSUSED */ 3430 int 3431 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3432 { 3433 /* { 3434 syscallarg(const char *) path; 3435 syscallarg(int) mode; 3436 } */ 3437 struct proc *p = l->l_proc; 3438 struct vnode *vp; 3439 struct vattr vattr; 3440 int error; 3441 struct nameidata nd; 3442 3443 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3444 SCARG(uap, path)); 3445 if ((error = namei(&nd)) != 0) 3446 return (error); 3447 vp = nd.ni_vp; 3448 if (vp != NULL) { 3449 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3450 if (nd.ni_dvp == vp) 3451 vrele(nd.ni_dvp); 3452 else 3453 vput(nd.ni_dvp); 3454 vrele(vp); 3455 return (EEXIST); 3456 } 3457 VATTR_NULL(&vattr); 3458 vattr.va_type = VDIR; 3459 /* We will read cwdi->cwdi_cmask unlocked. */ 3460 vattr.va_mode = 3461 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3462 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3463 if (!error) 3464 vput(nd.ni_vp); 3465 return (error); 3466 } 3467 3468 /* 3469 * Remove a directory file. 3470 */ 3471 /* ARGSUSED */ 3472 int 3473 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3474 { 3475 /* { 3476 syscallarg(const char *) path; 3477 } */ 3478 struct vnode *vp; 3479 int error; 3480 struct nameidata nd; 3481 3482 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3483 SCARG(uap, path)); 3484 if ((error = namei(&nd)) != 0) 3485 return (error); 3486 vp = nd.ni_vp; 3487 if (vp->v_type != VDIR) { 3488 error = ENOTDIR; 3489 goto out; 3490 } 3491 /* 3492 * No rmdir "." please. 3493 */ 3494 if (nd.ni_dvp == vp) { 3495 error = EINVAL; 3496 goto out; 3497 } 3498 /* 3499 * The root of a mounted filesystem cannot be deleted. 3500 */ 3501 if (vp->v_vflag & VV_ROOT) { 3502 error = EBUSY; 3503 goto out; 3504 } 3505 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3506 return (error); 3507 3508 out: 3509 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3510 if (nd.ni_dvp == vp) 3511 vrele(nd.ni_dvp); 3512 else 3513 vput(nd.ni_dvp); 3514 vput(vp); 3515 return (error); 3516 } 3517 3518 /* 3519 * Read a block of directory entries in a file system independent format. 3520 */ 3521 int 3522 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3523 { 3524 /* { 3525 syscallarg(int) fd; 3526 syscallarg(char *) buf; 3527 syscallarg(size_t) count; 3528 } */ 3529 struct proc *p = l->l_proc; 3530 struct file *fp; 3531 int error, done; 3532 3533 /* getvnode() will use the descriptor for us */ 3534 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3535 return (error); 3536 if ((fp->f_flag & FREAD) == 0) { 3537 error = EBADF; 3538 goto out; 3539 } 3540 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3541 SCARG(uap, count), &done, l, 0, 0); 3542 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3543 *retval = done; 3544 out: 3545 FILE_UNUSE(fp, l); 3546 return (error); 3547 } 3548 3549 /* 3550 * Set the mode mask for creation of filesystem nodes. 3551 */ 3552 int 3553 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3554 { 3555 /* { 3556 syscallarg(mode_t) newmask; 3557 } */ 3558 struct proc *p = l->l_proc; 3559 struct cwdinfo *cwdi; 3560 3561 /* 3562 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3563 * important is that we serialize changes to the mask. The 3564 * rw_exit() will issue a write memory barrier on our behalf, 3565 * and force the changes out to other CPUs (as it must use an 3566 * atomic operation, draining the local CPU's store buffers). 3567 */ 3568 cwdi = p->p_cwdi; 3569 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3570 *retval = cwdi->cwdi_cmask; 3571 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3572 rw_exit(&cwdi->cwdi_lock); 3573 3574 return (0); 3575 } 3576 3577 int 3578 dorevoke(struct vnode *vp, kauth_cred_t cred) 3579 { 3580 struct vattr vattr; 3581 int error; 3582 3583 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3584 return error; 3585 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3586 (error = kauth_authorize_generic(cred, 3587 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3588 VOP_REVOKE(vp, REVOKEALL); 3589 return (error); 3590 } 3591 3592 /* 3593 * Void all references to file by ripping underlying filesystem 3594 * away from vnode. 3595 */ 3596 /* ARGSUSED */ 3597 int 3598 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3599 { 3600 /* { 3601 syscallarg(const char *) path; 3602 } */ 3603 struct vnode *vp; 3604 int error; 3605 struct nameidata nd; 3606 3607 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3608 SCARG(uap, path)); 3609 if ((error = namei(&nd)) != 0) 3610 return (error); 3611 vp = nd.ni_vp; 3612 error = dorevoke(vp, l->l_cred); 3613 vrele(vp); 3614 return (error); 3615 } 3616 3617 /* 3618 * Convert a user file descriptor to a kernel file entry. 3619 */ 3620 int 3621 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3622 { 3623 struct vnode *vp; 3624 struct file *fp; 3625 3626 if ((fp = fd_getfile(fdp, fd)) == NULL) 3627 return (EBADF); 3628 3629 FILE_USE(fp); 3630 3631 if (fp->f_type != DTYPE_VNODE) { 3632 FILE_UNUSE(fp, NULL); 3633 return (EINVAL); 3634 } 3635 3636 vp = (struct vnode *)fp->f_data; 3637 if (vp->v_type == VBAD) { 3638 FILE_UNUSE(fp, NULL); 3639 return (EBADF); 3640 } 3641 3642 *fpp = fp; 3643 return (0); 3644 } 3645
Indexes created Thu Sep 25 16:09:42 GMT 2025