vfs_syscalls.c revision 1.330
1 /* $NetBSD: vfs_syscalls.c,v 1.330 2007/10/23 16:16:26 pooka 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 1.330 2007/10/23 16:16:26 pooka Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "fss.h" 46 #include "veriexec.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/kmem.h> 61 #include <sys/dirent.h> 62 #include <sys/sysctl.h> 63 #include <sys/syscallargs.h> 64 #include <sys/vfs_syscalls.h> 65 #include <sys/ktrace.h> 66 #ifdef FILEASSOC 67 #include <sys/fileassoc.h> 68 #endif /* FILEASSOC */ 69 #include <sys/verified_exec.h> 70 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #endif 79 #include <nfs/nfsproto.h> 80 #ifdef COMPAT_30 81 #include <nfs/nfs.h> 82 #include <nfs/nfs_var.h> 83 #endif 84 85 #if NFSS > 0 86 #include <dev/fssvar.h> 87 #endif 88 89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 90 91 static int change_dir(struct nameidata *, struct lwp *); 92 static int change_flags(struct vnode *, u_long, struct lwp *); 93 static int change_mode(struct vnode *, int, struct lwp *l); 94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 95 static int rename_files(const char *, const char *, struct lwp *, int); 96 97 void checkdirs(struct vnode *); 98 99 int dovfsusermount = 0; 100 101 /* 102 * Virtual File System System Calls 103 */ 104 105 /* 106 * Mount a file system. 107 */ 108 109 #if defined(COMPAT_09) || defined(COMPAT_43) 110 /* 111 * This table is used to maintain compatibility with 4.3BSD 112 * and NetBSD 0.9 mount syscalls. Note, the order is important! 113 * 114 * Do not modify this table. It should only contain filesystems 115 * supported by NetBSD 0.9 and 4.3BSD. 116 */ 117 const char * const mountcompatnames[] = { 118 NULL, /* 0 = MOUNT_NONE */ 119 MOUNT_FFS, /* 1 = MOUNT_UFS */ 120 MOUNT_NFS, /* 2 */ 121 MOUNT_MFS, /* 3 */ 122 MOUNT_MSDOS, /* 4 */ 123 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 124 MOUNT_FDESC, /* 6 */ 125 MOUNT_KERNFS, /* 7 */ 126 NULL, /* 8 = MOUNT_DEVFS */ 127 MOUNT_AFS, /* 9 */ 128 }; 129 const int nmountcompatnames = sizeof(mountcompatnames) / 130 sizeof(mountcompatnames[0]); 131 #endif /* COMPAT_09 || COMPAT_43 */ 132 133 static int 134 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 135 void *data, size_t *data_len) 136 { 137 struct mount *mp; 138 int error = 0, saved_flags; 139 140 mp = vp->v_mount; 141 saved_flags = mp->mnt_flag; 142 143 /* We can operate only on VV_ROOT nodes. */ 144 if ((vp->v_vflag & VV_ROOT) == 0) { 145 error = EINVAL; 146 goto out; 147 } 148 149 /* 150 * We only allow the filesystem to be reloaded if it 151 * is currently mounted read-only. 152 */ 153 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 154 error = EOPNOTSUPP; /* Needs translation */ 155 goto out; 156 } 157 158 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 159 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 160 if (error) 161 goto out; 162 163 if (vfs_busy(mp, LK_NOWAIT, 0)) { 164 error = EPERM; 165 goto out; 166 } 167 168 mp->mnt_flag &= ~MNT_OP_FLAGS; 169 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 170 171 /* 172 * Set the mount level flags. 173 */ 174 if (flags & MNT_RDONLY) 175 mp->mnt_flag |= MNT_RDONLY; 176 else if (mp->mnt_flag & MNT_RDONLY) 177 mp->mnt_iflag |= IMNT_WANTRDWR; 178 mp->mnt_flag &= 179 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 180 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 181 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 182 mp->mnt_flag |= flags & 183 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 184 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 185 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 186 MNT_IGNORE); 187 188 error = VFS_MOUNT(mp, path, data, data_len, l); 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], sizeof(fstypename)); 245 #else 246 return error; 247 #endif 248 } 249 250 #ifdef COMPAT_10 251 /* Accept `ufs' as an alias for `ffs'. */ 252 if (strcmp(fstypename, "ufs") == 0) 253 fstypename[0] = 'f'; 254 #endif 255 256 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 257 return ENODEV; 258 return 0; 259 } 260 261 static int 262 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 263 const char *path, int flags, void *data, size_t *data_len) 264 { 265 struct mount *mp = NULL; 266 struct vnode *vp = *vpp; 267 struct vattr va; 268 int error; 269 270 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 271 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 272 if (error) 273 return error; 274 275 /* Can't make a non-dir a mount-point (from here anyway). */ 276 if (vp->v_type != VDIR) 277 return ENOTDIR; 278 279 /* 280 * If the user is not root, ensure that they own the directory 281 * onto which we are attempting to mount. 282 */ 283 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 284 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 285 (error = kauth_authorize_generic(l->l_cred, 286 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 287 return error; 288 } 289 290 if (flags & MNT_EXPORTED) 291 return EINVAL; 292 293 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 294 return error; 295 296 /* 297 * Check if a file-system is not already mounted on this vnode. 298 */ 299 if (vp->v_mountedhere != NULL) 300 return EBUSY; 301 302 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 303 304 mp->mnt_op = vfsops; 305 306 TAILQ_INIT(&mp->mnt_vnodelist); 307 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 308 simple_lock_init(&mp->mnt_slock); 309 (void)vfs_busy(mp, LK_NOWAIT, 0); 310 311 mp->mnt_vnodecovered = vp; 312 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 313 mp->mnt_unmounter = NULL; 314 mount_initspecific(mp); 315 316 /* 317 * The underlying file system may refuse the mount for 318 * various reasons. Allow the user to force it to happen. 319 * 320 * Set the mount level flags. 321 */ 322 mp->mnt_flag = flags & 323 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 324 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 325 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 326 MNT_IGNORE | MNT_RDONLY); 327 328 error = VFS_MOUNT(mp, path, data, data_len, l); 329 mp->mnt_flag &= ~MNT_OP_FLAGS; 330 331 /* 332 * Put the new filesystem on the mount list after root. 333 */ 334 cache_purge(vp); 335 if (error != 0) { 336 vp->v_mountedhere = NULL; 337 mp->mnt_op->vfs_refcount--; 338 vfs_unbusy(mp); 339 vfs_destroy(mp); 340 return error; 341 } 342 343 mp->mnt_iflag &= ~IMNT_WANTRDWR; 344 vp->v_mountedhere = mp; 345 mutex_enter(&mountlist_lock); 346 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 347 mutex_exit(&mountlist_lock); 348 VOP_UNLOCK(vp, 0); 349 checkdirs(vp); 350 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 351 error = vfs_allocate_syncvnode(mp); 352 vfs_unbusy(mp); 353 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 354 error = VFS_START(mp, 0, l); 355 if (error) 356 vrele(vp); 357 *vpp = NULL; 358 return error; 359 } 360 361 static int 362 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 363 void *data, size_t *data_len) 364 { 365 struct mount *mp; 366 int error; 367 368 /* If MNT_GETARGS is specified, it should be the only flag. */ 369 if (flags & ~MNT_GETARGS) 370 return EINVAL; 371 372 mp = vp->v_mount; 373 374 /* XXX: probably some notion of "can see" here if we want isolation. */ 375 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 376 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 377 if (error) 378 return error; 379 380 if ((vp->v_vflag & VV_ROOT) == 0) 381 return EINVAL; 382 383 if (vfs_busy(mp, LK_NOWAIT, 0)) 384 return EPERM; 385 386 mp->mnt_flag &= ~MNT_OP_FLAGS; 387 mp->mnt_flag |= MNT_GETARGS; 388 error = VFS_MOUNT(mp, path, data, data_len, l); 389 mp->mnt_flag &= ~MNT_OP_FLAGS; 390 391 vfs_unbusy(mp); 392 return (error); 393 } 394 395 #ifdef COMPAT_40 396 /* ARGSUSED */ 397 int 398 compat_40_sys_mount(struct lwp *l, void *v, register_t *retval) 399 { 400 struct compat_40_sys_mount_args /* { 401 syscallarg(const char *) type; 402 syscallarg(const char *) path; 403 syscallarg(int) flags; 404 syscallarg(void *) data; 405 } */ *uap = v; 406 register_t dummy; 407 408 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 409 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 410 } 411 #endif 412 413 int 414 sys___mount50(struct lwp *l, void *v, register_t *retval) 415 { 416 struct sys___mount50_args /* { 417 syscallarg(const char *) type; 418 syscallarg(const char *) path; 419 syscallarg(int) flags; 420 syscallarg(void *) data; 421 syscallarg(size_t) data_len; 422 } */ *uap = v; 423 424 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 425 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 426 SCARG(uap, data_len), retval); 427 } 428 429 int 430 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 431 const char *path, int flags, void *data, enum uio_seg data_seg, 432 size_t data_len, register_t *retval) 433 { 434 struct vnode *vp; 435 struct nameidata nd; 436 void *data_buf = data; 437 int error; 438 439 /* 440 * Get vnode to be covered 441 */ 442 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 443 if ((error = namei(&nd)) != 0) 444 return (error); 445 vp = nd.ni_vp; 446 447 /* 448 * A lookup in VFS_MOUNT might result in an attempt to 449 * lock this vnode again, so make the lock recursive. 450 */ 451 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 452 453 if (vfsops == NULL) { 454 if (flags & (MNT_GETARGS | MNT_UPDATE)) 455 vfsops = vp->v_mount->mnt_op; 456 else { 457 /* 'type' is userspace */ 458 error = mount_get_vfsops(type, &vfsops); 459 if (error != 0) 460 goto done; 461 } 462 } 463 464 if (data != NULL && data_seg == UIO_USERSPACE) { 465 if (data_len == 0) { 466 /* No length supplied, use default for filesystem */ 467 data_len = vfsops->vfs_min_mount_data; 468 if (data_len > VFS_MAX_MOUNT_DATA) { 469 /* maybe a force loaded old LKM */ 470 error = EINVAL; 471 goto done; 472 } 473 #ifdef COMPAT_30 474 /* Hopefully a longer buffer won't make copyin() fail */ 475 if (flags & MNT_UPDATE 476 && data_len < sizeof (struct mnt_export_args30)) 477 data_len = sizeof (struct mnt_export_args30); 478 #endif 479 } 480 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 481 482 /* NFS needs the buffer even for mnt_getargs .... */ 483 error = copyin(data, data_buf, data_len); 484 if (error != 0) 485 goto done; 486 } 487 488 if (flags & MNT_GETARGS) { 489 if (data_len == 0) { 490 error = EINVAL; 491 goto done; 492 } 493 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 494 if (error != 0) 495 goto done; 496 if (data_seg == UIO_USERSPACE) 497 error = copyout(data_buf, data, data_len); 498 *retval = data_len; 499 } else if (flags & MNT_UPDATE) { 500 error = mount_update(l, vp, path, flags, data_buf, &data_len); 501 } else { 502 /* Locking is handled internally in mount_domount(). */ 503 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 504 &data_len); 505 } 506 507 done: 508 if (vp) 509 vput(vp); 510 if (data_buf != data) 511 free(data_buf, M_TEMP); 512 return (error); 513 } 514 515 /* 516 * Scan all active processes to see if any of them have a current 517 * or root directory onto which the new filesystem has just been 518 * mounted. If so, replace them with the new mount point. 519 */ 520 void 521 checkdirs(struct vnode *olddp) 522 { 523 struct cwdinfo *cwdi; 524 struct vnode *newdp; 525 struct proc *p; 526 527 if (olddp->v_usecount == 1) 528 return; 529 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 530 panic("mount: lost mount"); 531 mutex_enter(&proclist_lock); 532 PROCLIST_FOREACH(p, &allproc) { 533 cwdi = p->p_cwdi; 534 if (!cwdi) 535 continue; 536 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 537 if (cwdi->cwdi_cdir == olddp) { 538 vrele(cwdi->cwdi_cdir); 539 VREF(newdp); 540 cwdi->cwdi_cdir = newdp; 541 } 542 if (cwdi->cwdi_rdir == olddp) { 543 vrele(cwdi->cwdi_rdir); 544 VREF(newdp); 545 cwdi->cwdi_rdir = newdp; 546 } 547 rw_exit(&cwdi->cwdi_lock); 548 } 549 mutex_exit(&proclist_lock); 550 if (rootvnode == olddp) { 551 vrele(rootvnode); 552 VREF(newdp); 553 rootvnode = newdp; 554 } 555 vput(newdp); 556 } 557 558 /* 559 * Unmount a file system. 560 * 561 * Note: unmount takes a path to the vnode mounted on as argument, 562 * not special file (as before). 563 */ 564 /* ARGSUSED */ 565 int 566 sys_unmount(struct lwp *l, void *v, register_t *retval) 567 { 568 struct sys_unmount_args /* { 569 syscallarg(const char *) path; 570 syscallarg(int) flags; 571 } */ *uap = v; 572 struct vnode *vp; 573 struct mount *mp; 574 int error; 575 struct nameidata nd; 576 577 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 578 SCARG(uap, path), l); 579 if ((error = namei(&nd)) != 0) 580 return (error); 581 vp = nd.ni_vp; 582 mp = vp->v_mount; 583 584 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 585 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 586 if (error) { 587 vput(vp); 588 return (error); 589 } 590 591 /* 592 * Don't allow unmounting the root file system. 593 */ 594 if (mp->mnt_flag & MNT_ROOTFS) { 595 vput(vp); 596 return (EINVAL); 597 } 598 599 /* 600 * Must be the root of the filesystem 601 */ 602 if ((vp->v_vflag & VV_ROOT) == 0) { 603 vput(vp); 604 return (EINVAL); 605 } 606 vput(vp); 607 608 /* 609 * XXX Freeze syncer. Must do this before locking the 610 * mount point. See dounmount() for details. 611 */ 612 mutex_enter(&syncer_mutex); 613 614 if (vfs_busy(mp, 0, 0)) { 615 mutex_exit(&syncer_mutex); 616 return (EBUSY); 617 } 618 619 return (dounmount(mp, SCARG(uap, flags), l)); 620 } 621 622 /* 623 * Do the actual file system unmount. File system is assumed to have been 624 * marked busy by the caller. 625 */ 626 int 627 dounmount(struct mount *mp, int flags, struct lwp *l) 628 { 629 struct vnode *coveredvp; 630 int error; 631 int async; 632 int used_syncer; 633 634 #if NVERIEXEC > 0 635 error = veriexec_unmountchk(mp); 636 if (error) 637 return (error); 638 #endif /* NVERIEXEC > 0 */ 639 640 mutex_enter(&mountlist_lock); 641 vfs_unbusy(mp); 642 used_syncer = (mp->mnt_syncer != NULL); 643 644 /* 645 * XXX Syncer must be frozen when we get here. This should really 646 * be done on a per-mountpoint basis, but especially the softdep 647 * code possibly called from the syncer doesn't exactly work on a 648 * per-mountpoint basis, so the softdep code would become a maze 649 * of vfs_busy() calls. 650 * 651 * The caller of dounmount() must acquire syncer_mutex because 652 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 653 * order, and we must preserve that order to avoid deadlock. 654 * 655 * So, if the file system did not use the syncer, now is 656 * the time to release the syncer_mutex. 657 */ 658 if (used_syncer == 0) 659 mutex_exit(&syncer_mutex); 660 661 mp->mnt_iflag |= IMNT_UNMOUNT; 662 mp->mnt_unmounter = l; 663 mutex_exit(&mountlist_lock); /* XXX */ 664 lockmgr(&mp->mnt_lock, LK_DRAIN, NULL); 665 666 async = mp->mnt_flag & MNT_ASYNC; 667 mp->mnt_flag &= ~MNT_ASYNC; 668 cache_purgevfs(mp); /* remove cache entries for this file sys */ 669 if (mp->mnt_syncer != NULL) 670 vfs_deallocate_syncvnode(mp); 671 error = 0; 672 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 673 #if NFSS > 0 674 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 675 #endif 676 if (error == 0) 677 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 678 } 679 if (error == 0 || (flags & MNT_FORCE)) 680 error = VFS_UNMOUNT(mp, flags, l); 681 if (error) { 682 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 683 (void) vfs_allocate_syncvnode(mp); 684 mutex_enter(&mountlist_lock); 685 mp->mnt_iflag &= ~IMNT_UNMOUNT; 686 mp->mnt_unmounter = NULL; 687 mp->mnt_flag |= async; 688 mutex_exit(&mountlist_lock); /* XXX */ 689 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_REENABLE, 690 NULL); 691 if (used_syncer) 692 mutex_exit(&syncer_mutex); 693 simple_lock(&mp->mnt_slock); 694 while (mp->mnt_wcnt > 0) { 695 wakeup(mp); 696 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 697 0, &mp->mnt_slock); 698 } 699 simple_unlock(&mp->mnt_slock); 700 return (error); 701 } 702 mutex_enter(&mountlist_lock); 703 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 704 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 705 coveredvp->v_mountedhere = NULL; 706 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 707 panic("unmount: dangling vnode"); 708 mp->mnt_iflag |= IMNT_GONE; 709 mutex_exit(&mountlist_lock); 710 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL); 711 if (coveredvp != NULLVP) 712 vrele(coveredvp); 713 if (used_syncer) 714 mutex_exit(&syncer_mutex); 715 simple_lock(&mp->mnt_slock); 716 while (mp->mnt_wcnt > 0) { 717 wakeup(mp); 718 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 719 } 720 simple_unlock(&mp->mnt_slock); 721 vfs_hooks_unmount(mp); 722 vfs_delref(mp->mnt_op); 723 vfs_destroy(mp); 724 return (0); 725 } 726 727 /* 728 * Sync each mounted filesystem. 729 */ 730 #ifdef DEBUG 731 int syncprt = 0; 732 struct ctldebug debug0 = { "syncprt", &syncprt }; 733 #endif 734 735 /* ARGSUSED */ 736 int 737 sys_sync(struct lwp *l, void *v, register_t *retval) 738 { 739 struct mount *mp, *nmp; 740 int asyncflag; 741 742 if (l == NULL) 743 l = &lwp0; 744 745 mutex_enter(&mountlist_lock); 746 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 747 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 748 nmp = mp->mnt_list.cqe_prev; 749 continue; 750 } 751 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 752 asyncflag = mp->mnt_flag & MNT_ASYNC; 753 mp->mnt_flag &= ~MNT_ASYNC; 754 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 755 if (asyncflag) 756 mp->mnt_flag |= MNT_ASYNC; 757 } 758 mutex_enter(&mountlist_lock); 759 nmp = mp->mnt_list.cqe_prev; 760 vfs_unbusy(mp); 761 762 } 763 mutex_exit(&mountlist_lock); 764 #ifdef DEBUG 765 if (syncprt) 766 vfs_bufstats(); 767 #endif /* DEBUG */ 768 return (0); 769 } 770 771 /* 772 * Change filesystem quotas. 773 */ 774 /* ARGSUSED */ 775 int 776 sys_quotactl(struct lwp *l, void *v, register_t *retval) 777 { 778 struct sys_quotactl_args /* { 779 syscallarg(const char *) path; 780 syscallarg(int) cmd; 781 syscallarg(int) uid; 782 syscallarg(void *) arg; 783 } */ *uap = v; 784 struct mount *mp; 785 int error; 786 struct nameidata nd; 787 788 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 789 SCARG(uap, path), l); 790 if ((error = namei(&nd)) != 0) 791 return (error); 792 mp = nd.ni_vp->v_mount; 793 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 794 SCARG(uap, arg), l); 795 vrele(nd.ni_vp); 796 return (error); 797 } 798 799 int 800 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 801 int root) 802 { 803 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 804 int error = 0; 805 806 /* 807 * If MNT_NOWAIT or MNT_LAZY is specified, do not 808 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 809 * overrides MNT_NOWAIT. 810 */ 811 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 812 (flags != MNT_WAIT && flags != 0)) { 813 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 814 goto done; 815 } 816 817 /* Get the filesystem stats now */ 818 memset(sp, 0, sizeof(*sp)); 819 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 820 return error; 821 } 822 823 if (cwdi->cwdi_rdir == NULL) 824 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 825 done: 826 if (cwdi->cwdi_rdir != NULL) { 827 size_t len; 828 char *bp; 829 char *path = PNBUF_GET(); 830 831 bp = path + MAXPATHLEN; 832 *--bp = '\0'; 833 rw_enter(&cwdi->cwdi_lock, RW_READER); 834 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 835 MAXPATHLEN / 2, 0, l); 836 rw_exit(&cwdi->cwdi_lock); 837 if (error) { 838 PNBUF_PUT(path); 839 return error; 840 } 841 len = strlen(bp); 842 /* 843 * for mount points that are below our root, we can see 844 * them, so we fix up the pathname and return them. The 845 * rest we cannot see, so we don't allow viewing the 846 * data. 847 */ 848 if (strncmp(bp, sp->f_mntonname, len) == 0) { 849 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 850 sizeof(sp->f_mntonname)); 851 if (sp->f_mntonname[0] == '\0') 852 (void)strlcpy(sp->f_mntonname, "/", 853 sizeof(sp->f_mntonname)); 854 } else { 855 if (root) 856 (void)strlcpy(sp->f_mntonname, "/", 857 sizeof(sp->f_mntonname)); 858 else 859 error = EPERM; 860 } 861 PNBUF_PUT(path); 862 } 863 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 864 return error; 865 } 866 867 /* 868 * Get filesystem statistics by path. 869 */ 870 int 871 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 872 { 873 struct mount *mp; 874 int error; 875 struct nameidata nd; 876 877 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 878 if ((error = namei(&nd)) != 0) 879 return error; 880 mp = nd.ni_vp->v_mount; 881 error = dostatvfs(mp, sb, l, flags, 1); 882 vrele(nd.ni_vp); 883 return error; 884 } 885 886 /* ARGSUSED */ 887 int 888 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 889 { 890 struct sys_statvfs1_args /* { 891 syscallarg(const char *) path; 892 syscallarg(struct statvfs *) buf; 893 syscallarg(int) flags; 894 } */ *uap = v; 895 struct statvfs *sb; 896 int error; 897 898 sb = STATVFSBUF_GET(); 899 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 900 if (error == 0) 901 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 902 STATVFSBUF_PUT(sb); 903 return error; 904 } 905 906 /* 907 * Get filesystem statistics by fd. 908 */ 909 int 910 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 911 { 912 struct proc *p = l->l_proc; 913 struct file *fp; 914 struct mount *mp; 915 int error; 916 917 /* getvnode() will use the descriptor for us */ 918 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 919 return (error); 920 mp = ((struct vnode *)fp->f_data)->v_mount; 921 error = dostatvfs(mp, sb, l, flags, 1); 922 FILE_UNUSE(fp, l); 923 return error; 924 } 925 926 /* ARGSUSED */ 927 int 928 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 929 { 930 struct sys_fstatvfs1_args /* { 931 syscallarg(int) fd; 932 syscallarg(struct statvfs *) buf; 933 syscallarg(int) flags; 934 } */ *uap = v; 935 struct statvfs *sb; 936 int error; 937 938 sb = STATVFSBUF_GET(); 939 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 940 if (error == 0) 941 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 942 STATVFSBUF_PUT(sb); 943 return error; 944 } 945 946 947 /* 948 * Get statistics on all filesystems. 949 */ 950 int 951 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 952 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 953 register_t *retval) 954 { 955 int root = 0; 956 struct proc *p = l->l_proc; 957 struct mount *mp, *nmp; 958 struct statvfs *sb; 959 size_t count, maxcount; 960 int error = 0; 961 962 sb = STATVFSBUF_GET(); 963 maxcount = bufsize / entry_sz; 964 mutex_enter(&mountlist_lock); 965 count = 0; 966 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 967 mp = nmp) { 968 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 969 nmp = CIRCLEQ_NEXT(mp, mnt_list); 970 continue; 971 } 972 if (sfsp && count < maxcount) { 973 error = dostatvfs(mp, sb, l, flags, 0); 974 if (error) { 975 mutex_enter(&mountlist_lock); 976 nmp = CIRCLEQ_NEXT(mp, mnt_list); 977 vfs_unbusy(mp); 978 continue; 979 } 980 error = copyfn(sb, sfsp, entry_sz); 981 if (error) { 982 vfs_unbusy(mp); 983 goto out; 984 } 985 sfsp = (char *)sfsp + entry_sz; 986 root |= strcmp(sb->f_mntonname, "/") == 0; 987 } 988 count++; 989 mutex_enter(&mountlist_lock); 990 nmp = CIRCLEQ_NEXT(mp, mnt_list); 991 vfs_unbusy(mp); 992 } 993 mutex_exit(&mountlist_lock); 994 if (root == 0 && p->p_cwdi->cwdi_rdir) { 995 /* 996 * fake a root entry 997 */ 998 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 999 if (error != 0) 1000 goto out; 1001 if (sfsp) 1002 error = copyfn(sb, sfsp, entry_sz); 1003 count++; 1004 } 1005 if (sfsp && count > maxcount) 1006 *retval = maxcount; 1007 else 1008 *retval = count; 1009 out: 1010 STATVFSBUF_PUT(sb); 1011 return error; 1012 } 1013 1014 int 1015 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 1016 { 1017 struct sys_getvfsstat_args /* { 1018 syscallarg(struct statvfs *) buf; 1019 syscallarg(size_t) bufsize; 1020 syscallarg(int) flags; 1021 } */ *uap = v; 1022 1023 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1024 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1025 } 1026 1027 /* 1028 * Change current working directory to a given file descriptor. 1029 */ 1030 /* ARGSUSED */ 1031 int 1032 sys_fchdir(struct lwp *l, void *v, register_t *retval) 1033 { 1034 struct sys_fchdir_args /* { 1035 syscallarg(int) fd; 1036 } */ *uap = v; 1037 struct proc *p = l->l_proc; 1038 struct filedesc *fdp = p->p_fd; 1039 struct cwdinfo *cwdi; 1040 struct vnode *vp, *tdp; 1041 struct mount *mp; 1042 struct file *fp; 1043 int error; 1044 1045 /* getvnode() will use the descriptor for us */ 1046 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1047 return (error); 1048 vp = (struct vnode *)fp->f_data; 1049 1050 VREF(vp); 1051 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1052 if (vp->v_type != VDIR) 1053 error = ENOTDIR; 1054 else 1055 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1056 if (error) { 1057 vput(vp); 1058 goto out; 1059 } 1060 while ((mp = vp->v_mountedhere) != NULL) { 1061 if (vfs_busy(mp, 0, 0)) 1062 continue; 1063 1064 vput(vp); 1065 error = VFS_ROOT(mp, &tdp); 1066 vfs_unbusy(mp); 1067 if (error) 1068 goto out; 1069 vp = tdp; 1070 } 1071 VOP_UNLOCK(vp, 0); 1072 1073 /* 1074 * Disallow changing to a directory not under the process's 1075 * current root directory (if there is one). 1076 */ 1077 cwdi = p->p_cwdi; 1078 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1079 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1080 vrele(vp); 1081 error = EPERM; /* operation not permitted */ 1082 } else { 1083 vrele(cwdi->cwdi_cdir); 1084 cwdi->cwdi_cdir = vp; 1085 } 1086 rw_exit(&cwdi->cwdi_lock); 1087 1088 out: 1089 FILE_UNUSE(fp, l); 1090 return (error); 1091 } 1092 1093 /* 1094 * Change this process's notion of the root directory to a given file 1095 * descriptor. 1096 */ 1097 int 1098 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1099 { 1100 struct sys_fchroot_args *uap = v; 1101 struct proc *p = l->l_proc; 1102 struct filedesc *fdp = p->p_fd; 1103 struct cwdinfo *cwdi; 1104 struct vnode *vp; 1105 struct file *fp; 1106 int error; 1107 1108 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1109 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1110 return error; 1111 /* getvnode() will use the descriptor for us */ 1112 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1113 return error; 1114 vp = (struct vnode *) fp->f_data; 1115 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1116 if (vp->v_type != VDIR) 1117 error = ENOTDIR; 1118 else 1119 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1120 VOP_UNLOCK(vp, 0); 1121 if (error) 1122 goto out; 1123 VREF(vp); 1124 1125 /* 1126 * Prevent escaping from chroot by putting the root under 1127 * the working directory. Silently chdir to / if we aren't 1128 * already there. 1129 */ 1130 cwdi = p->p_cwdi; 1131 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1132 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1133 /* 1134 * XXX would be more failsafe to change directory to a 1135 * deadfs node here instead 1136 */ 1137 vrele(cwdi->cwdi_cdir); 1138 VREF(vp); 1139 cwdi->cwdi_cdir = vp; 1140 } 1141 1142 if (cwdi->cwdi_rdir != NULL) 1143 vrele(cwdi->cwdi_rdir); 1144 cwdi->cwdi_rdir = vp; 1145 rw_exit(&cwdi->cwdi_lock); 1146 1147 out: 1148 FILE_UNUSE(fp, l); 1149 return (error); 1150 } 1151 1152 /* 1153 * Change current working directory (``.''). 1154 */ 1155 /* ARGSUSED */ 1156 int 1157 sys_chdir(struct lwp *l, void *v, register_t *retval) 1158 { 1159 struct sys_chdir_args /* { 1160 syscallarg(const char *) path; 1161 } */ *uap = v; 1162 struct proc *p = l->l_proc; 1163 struct cwdinfo *cwdi; 1164 int error; 1165 struct nameidata nd; 1166 1167 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1168 SCARG(uap, path), l); 1169 if ((error = change_dir(&nd, l)) != 0) 1170 return (error); 1171 cwdi = p->p_cwdi; 1172 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1173 vrele(cwdi->cwdi_cdir); 1174 cwdi->cwdi_cdir = nd.ni_vp; 1175 rw_exit(&cwdi->cwdi_lock); 1176 return (0); 1177 } 1178 1179 /* 1180 * Change notion of root (``/'') directory. 1181 */ 1182 /* ARGSUSED */ 1183 int 1184 sys_chroot(struct lwp *l, void *v, register_t *retval) 1185 { 1186 struct sys_chroot_args /* { 1187 syscallarg(const char *) path; 1188 } */ *uap = v; 1189 struct proc *p = l->l_proc; 1190 struct cwdinfo *cwdi; 1191 struct vnode *vp; 1192 int error; 1193 struct nameidata nd; 1194 1195 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1196 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1197 return (error); 1198 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1199 SCARG(uap, path), l); 1200 if ((error = change_dir(&nd, l)) != 0) 1201 return (error); 1202 1203 cwdi = p->p_cwdi; 1204 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1205 if (cwdi->cwdi_rdir != NULL) 1206 vrele(cwdi->cwdi_rdir); 1207 vp = nd.ni_vp; 1208 cwdi->cwdi_rdir = vp; 1209 1210 /* 1211 * Prevent escaping from chroot by putting the root under 1212 * the working directory. Silently chdir to / if we aren't 1213 * already there. 1214 */ 1215 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1216 /* 1217 * XXX would be more failsafe to change directory to a 1218 * deadfs node here instead 1219 */ 1220 vrele(cwdi->cwdi_cdir); 1221 VREF(vp); 1222 cwdi->cwdi_cdir = vp; 1223 } 1224 rw_exit(&cwdi->cwdi_lock); 1225 1226 return (0); 1227 } 1228 1229 /* 1230 * Common routine for chroot and chdir. 1231 */ 1232 static int 1233 change_dir(struct nameidata *ndp, struct lwp *l) 1234 { 1235 struct vnode *vp; 1236 int error; 1237 1238 if ((error = namei(ndp)) != 0) 1239 return (error); 1240 vp = ndp->ni_vp; 1241 if (vp->v_type != VDIR) 1242 error = ENOTDIR; 1243 else 1244 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1245 1246 if (error) 1247 vput(vp); 1248 else 1249 VOP_UNLOCK(vp, 0); 1250 return (error); 1251 } 1252 1253 /* 1254 * Check permissions, allocate an open file structure, 1255 * and call the device open routine if any. 1256 */ 1257 int 1258 sys_open(struct lwp *l, void *v, register_t *retval) 1259 { 1260 struct sys_open_args /* { 1261 syscallarg(const char *) path; 1262 syscallarg(int) flags; 1263 syscallarg(int) mode; 1264 } */ *uap = v; 1265 struct proc *p = l->l_proc; 1266 struct cwdinfo *cwdi = p->p_cwdi; 1267 struct filedesc *fdp = p->p_fd; 1268 struct file *fp; 1269 struct vnode *vp; 1270 int flags, cmode; 1271 int type, indx, error; 1272 struct flock lf; 1273 struct nameidata nd; 1274 1275 flags = FFLAGS(SCARG(uap, flags)); 1276 if ((flags & (FREAD | FWRITE)) == 0) 1277 return (EINVAL); 1278 /* falloc() will use the file descriptor for us */ 1279 if ((error = falloc(l, &fp, &indx)) != 0) 1280 return (error); 1281 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1282 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1283 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1284 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1285 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1286 rw_enter(&fdp->fd_lock, RW_WRITER); 1287 FILE_UNUSE(fp, l); 1288 fdp->fd_ofiles[indx] = NULL; 1289 rw_exit(&fdp->fd_lock); 1290 ffree(fp); 1291 if ((error == EDUPFD || error == EMOVEFD) && 1292 l->l_dupfd >= 0 && /* XXX from fdopen */ 1293 (error = 1294 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1295 *retval = indx; 1296 return (0); 1297 } 1298 if (error == ERESTART) 1299 error = EINTR; 1300 fdremove(fdp, indx); 1301 return (error); 1302 } 1303 l->l_dupfd = 0; 1304 vp = nd.ni_vp; 1305 fp->f_flag = flags & FMASK; 1306 fp->f_type = DTYPE_VNODE; 1307 fp->f_ops = &vnops; 1308 fp->f_data = vp; 1309 if (flags & (O_EXLOCK | O_SHLOCK)) { 1310 lf.l_whence = SEEK_SET; 1311 lf.l_start = 0; 1312 lf.l_len = 0; 1313 if (flags & O_EXLOCK) 1314 lf.l_type = F_WRLCK; 1315 else 1316 lf.l_type = F_RDLCK; 1317 type = F_FLOCK; 1318 if ((flags & FNONBLOCK) == 0) 1319 type |= F_WAIT; 1320 VOP_UNLOCK(vp, 0); 1321 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1322 if (error) { 1323 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1324 FILE_UNUSE(fp, l); 1325 ffree(fp); 1326 fdremove(fdp, indx); 1327 return (error); 1328 } 1329 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1330 fp->f_flag |= FHASLOCK; 1331 } 1332 VOP_UNLOCK(vp, 0); 1333 *retval = indx; 1334 FILE_SET_MATURE(fp); 1335 FILE_UNUSE(fp, l); 1336 return (0); 1337 } 1338 1339 static void 1340 vfs__fhfree(fhandle_t *fhp) 1341 { 1342 size_t fhsize; 1343 1344 if (fhp == NULL) { 1345 return; 1346 } 1347 fhsize = FHANDLE_SIZE(fhp); 1348 kmem_free(fhp, fhsize); 1349 } 1350 1351 /* 1352 * vfs_composefh: compose a filehandle. 1353 */ 1354 1355 int 1356 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1357 { 1358 struct mount *mp; 1359 struct fid *fidp; 1360 int error; 1361 size_t needfhsize; 1362 size_t fidsize; 1363 1364 mp = vp->v_mount; 1365 fidp = NULL; 1366 if (*fh_size < FHANDLE_SIZE_MIN) { 1367 fidsize = 0; 1368 } else { 1369 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1370 if (fhp != NULL) { 1371 memset(fhp, 0, *fh_size); 1372 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1373 fidp = &fhp->fh_fid; 1374 } 1375 } 1376 error = VFS_VPTOFH(vp, fidp, &fidsize); 1377 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1378 if (error == 0 && *fh_size < needfhsize) { 1379 error = E2BIG; 1380 } 1381 *fh_size = needfhsize; 1382 return error; 1383 } 1384 1385 int 1386 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1387 { 1388 struct mount *mp; 1389 fhandle_t *fhp; 1390 size_t fhsize; 1391 size_t fidsize; 1392 int error; 1393 1394 *fhpp = NULL; 1395 mp = vp->v_mount; 1396 fidsize = 0; 1397 error = VFS_VPTOFH(vp, NULL, &fidsize); 1398 KASSERT(error != 0); 1399 if (error != E2BIG) { 1400 goto out; 1401 } 1402 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1403 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1404 if (fhp == NULL) { 1405 error = ENOMEM; 1406 goto out; 1407 } 1408 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1409 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1410 if (error == 0) { 1411 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1412 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1413 *fhpp = fhp; 1414 } else { 1415 kmem_free(fhp, fhsize); 1416 } 1417 out: 1418 return error; 1419 } 1420 1421 void 1422 vfs_composefh_free(fhandle_t *fhp) 1423 { 1424 1425 vfs__fhfree(fhp); 1426 } 1427 1428 /* 1429 * vfs_fhtovp: lookup a vnode by a filehandle. 1430 */ 1431 1432 int 1433 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1434 { 1435 struct mount *mp; 1436 int error; 1437 1438 *vpp = NULL; 1439 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1440 if (mp == NULL) { 1441 error = ESTALE; 1442 goto out; 1443 } 1444 if (mp->mnt_op->vfs_fhtovp == NULL) { 1445 error = EOPNOTSUPP; 1446 goto out; 1447 } 1448 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1449 out: 1450 return error; 1451 } 1452 1453 /* 1454 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1455 * the needed size. 1456 */ 1457 1458 int 1459 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1460 { 1461 fhandle_t *fhp; 1462 int error; 1463 1464 *fhpp = NULL; 1465 if (fhsize > FHANDLE_SIZE_MAX) { 1466 return EINVAL; 1467 } 1468 if (fhsize < FHANDLE_SIZE_MIN) { 1469 return EINVAL; 1470 } 1471 again: 1472 fhp = kmem_alloc(fhsize, KM_SLEEP); 1473 if (fhp == NULL) { 1474 return ENOMEM; 1475 } 1476 error = copyin(ufhp, fhp, fhsize); 1477 if (error == 0) { 1478 /* XXX this check shouldn't be here */ 1479 if (FHANDLE_SIZE(fhp) == fhsize) { 1480 *fhpp = fhp; 1481 return 0; 1482 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1483 /* 1484 * a kludge for nfsv2 padded handles. 1485 */ 1486 size_t sz; 1487 1488 sz = FHANDLE_SIZE(fhp); 1489 kmem_free(fhp, fhsize); 1490 fhsize = sz; 1491 goto again; 1492 } else { 1493 /* 1494 * userland told us wrong size. 1495 */ 1496 error = EINVAL; 1497 } 1498 } 1499 kmem_free(fhp, fhsize); 1500 return error; 1501 } 1502 1503 void 1504 vfs_copyinfh_free(fhandle_t *fhp) 1505 { 1506 1507 vfs__fhfree(fhp); 1508 } 1509 1510 /* 1511 * Get file handle system call 1512 */ 1513 int 1514 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1515 { 1516 struct sys___getfh30_args /* { 1517 syscallarg(char *) fname; 1518 syscallarg(fhandle_t *) fhp; 1519 syscallarg(size_t *) fh_size; 1520 } */ *uap = v; 1521 struct vnode *vp; 1522 fhandle_t *fh; 1523 int error; 1524 struct nameidata nd; 1525 size_t sz; 1526 size_t usz; 1527 1528 /* 1529 * Must be super user 1530 */ 1531 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1532 0, NULL, NULL, NULL); 1533 if (error) 1534 return (error); 1535 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1536 SCARG(uap, fname), l); 1537 error = namei(&nd); 1538 if (error) 1539 return (error); 1540 vp = nd.ni_vp; 1541 error = vfs_composefh_alloc(vp, &fh); 1542 vput(vp); 1543 if (error != 0) { 1544 goto out; 1545 } 1546 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1547 if (error != 0) { 1548 goto out; 1549 } 1550 sz = FHANDLE_SIZE(fh); 1551 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1552 if (error != 0) { 1553 goto out; 1554 } 1555 if (usz >= sz) { 1556 error = copyout(fh, SCARG(uap, fhp), sz); 1557 } else { 1558 error = E2BIG; 1559 } 1560 out: 1561 vfs_composefh_free(fh); 1562 return (error); 1563 } 1564 1565 /* 1566 * Open a file given a file handle. 1567 * 1568 * Check permissions, allocate an open file structure, 1569 * and call the device open routine if any. 1570 */ 1571 1572 int 1573 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1574 register_t *retval) 1575 { 1576 struct filedesc *fdp = l->l_proc->p_fd; 1577 struct file *fp; 1578 struct vnode *vp = NULL; 1579 kauth_cred_t cred = l->l_cred; 1580 struct file *nfp; 1581 int type, indx, error=0; 1582 struct flock lf; 1583 struct vattr va; 1584 fhandle_t *fh; 1585 int flags; 1586 1587 /* 1588 * Must be super user 1589 */ 1590 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1591 0, NULL, NULL, NULL))) 1592 return (error); 1593 1594 flags = FFLAGS(oflags); 1595 if ((flags & (FREAD | FWRITE)) == 0) 1596 return (EINVAL); 1597 if ((flags & O_CREAT)) 1598 return (EINVAL); 1599 /* falloc() will use the file descriptor for us */ 1600 if ((error = falloc(l, &nfp, &indx)) != 0) 1601 return (error); 1602 fp = nfp; 1603 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1604 if (error != 0) { 1605 goto bad; 1606 } 1607 error = vfs_fhtovp(fh, &vp); 1608 if (error != 0) { 1609 goto bad; 1610 } 1611 1612 /* Now do an effective vn_open */ 1613 1614 if (vp->v_type == VSOCK) { 1615 error = EOPNOTSUPP; 1616 goto bad; 1617 } 1618 if (flags & FREAD) { 1619 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1620 goto bad; 1621 } 1622 if (flags & (FWRITE | O_TRUNC)) { 1623 if (vp->v_type == VDIR) { 1624 error = EISDIR; 1625 goto bad; 1626 } 1627 if ((error = vn_writechk(vp)) != 0 || 1628 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1629 goto bad; 1630 } 1631 if (flags & O_TRUNC) { 1632 VOP_UNLOCK(vp, 0); /* XXX */ 1633 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1634 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1635 VATTR_NULL(&va); 1636 va.va_size = 0; 1637 error = VOP_SETATTR(vp, &va, cred, l); 1638 if (error) 1639 goto bad; 1640 } 1641 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1642 goto bad; 1643 if (flags & FWRITE) 1644 vp->v_writecount++; 1645 1646 /* done with modified vn_open, now finish what sys_open does. */ 1647 1648 fp->f_flag = flags & FMASK; 1649 fp->f_type = DTYPE_VNODE; 1650 fp->f_ops = &vnops; 1651 fp->f_data = vp; 1652 if (flags & (O_EXLOCK | O_SHLOCK)) { 1653 lf.l_whence = SEEK_SET; 1654 lf.l_start = 0; 1655 lf.l_len = 0; 1656 if (flags & O_EXLOCK) 1657 lf.l_type = F_WRLCK; 1658 else 1659 lf.l_type = F_RDLCK; 1660 type = F_FLOCK; 1661 if ((flags & FNONBLOCK) == 0) 1662 type |= F_WAIT; 1663 VOP_UNLOCK(vp, 0); 1664 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1665 if (error) { 1666 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1667 FILE_UNUSE(fp, l); 1668 ffree(fp); 1669 fdremove(fdp, indx); 1670 return (error); 1671 } 1672 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1673 fp->f_flag |= FHASLOCK; 1674 } 1675 VOP_UNLOCK(vp, 0); 1676 *retval = indx; 1677 FILE_SET_MATURE(fp); 1678 FILE_UNUSE(fp, l); 1679 vfs_copyinfh_free(fh); 1680 return (0); 1681 1682 bad: 1683 FILE_UNUSE(fp, l); 1684 ffree(fp); 1685 fdremove(fdp, indx); 1686 if (vp != NULL) 1687 vput(vp); 1688 vfs_copyinfh_free(fh); 1689 return (error); 1690 } 1691 1692 int 1693 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1694 { 1695 struct sys___fhopen40_args /* { 1696 syscallarg(const void *) fhp; 1697 syscallarg(size_t) fh_size; 1698 syscallarg(int) flags; 1699 } */ *uap = v; 1700 1701 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1702 SCARG(uap, flags), retval); 1703 } 1704 1705 int 1706 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1707 { 1708 int error; 1709 fhandle_t *fh; 1710 struct vnode *vp; 1711 1712 /* 1713 * Must be super user 1714 */ 1715 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1716 0, NULL, NULL, NULL))) 1717 return (error); 1718 1719 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1720 if (error != 0) 1721 return error; 1722 1723 error = vfs_fhtovp(fh, &vp); 1724 vfs_copyinfh_free(fh); 1725 if (error != 0) 1726 return error; 1727 1728 error = vn_stat(vp, sb, l); 1729 vput(vp); 1730 return error; 1731 } 1732 1733 1734 /* ARGSUSED */ 1735 int 1736 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1737 { 1738 struct sys___fhstat40_args /* { 1739 syscallarg(const void *) fhp; 1740 syscallarg(size_t) fh_size; 1741 syscallarg(struct stat *) sb; 1742 } */ *uap = v; 1743 struct stat sb; 1744 int error; 1745 1746 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1747 if (error) 1748 return error; 1749 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1750 } 1751 1752 int 1753 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1754 int flags) 1755 { 1756 fhandle_t *fh; 1757 struct mount *mp; 1758 struct vnode *vp; 1759 int error; 1760 1761 /* 1762 * Must be super user 1763 */ 1764 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1765 0, NULL, NULL, NULL))) 1766 return error; 1767 1768 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1769 if (error != 0) 1770 return error; 1771 1772 error = vfs_fhtovp(fh, &vp); 1773 vfs_copyinfh_free(fh); 1774 if (error != 0) 1775 return error; 1776 1777 mp = vp->v_mount; 1778 error = dostatvfs(mp, sb, l, flags, 1); 1779 vput(vp); 1780 return error; 1781 } 1782 1783 /* ARGSUSED */ 1784 int 1785 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1786 { 1787 struct sys___fhstatvfs140_args /* { 1788 syscallarg(const void *) fhp; 1789 syscallarg(size_t) fh_size; 1790 syscallarg(struct statvfs *) buf; 1791 syscallarg(int) flags; 1792 } */ *uap = v; 1793 struct statvfs *sb = STATVFSBUF_GET(); 1794 int error; 1795 1796 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1797 SCARG(uap, flags)); 1798 if (error == 0) 1799 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1800 STATVFSBUF_PUT(sb); 1801 return error; 1802 } 1803 1804 /* 1805 * Create a special file. 1806 */ 1807 /* ARGSUSED */ 1808 int 1809 sys_mknod(struct lwp *l, void *v, register_t *retval) 1810 { 1811 struct sys_mknod_args /* { 1812 syscallarg(const char *) path; 1813 syscallarg(int) mode; 1814 syscallarg(int) dev; 1815 } */ *uap = v; 1816 struct proc *p = l->l_proc; 1817 struct vnode *vp; 1818 struct vattr vattr; 1819 int error, optype; 1820 struct nameidata nd; 1821 char *path; 1822 const char *cpath; 1823 enum uio_seg seg = UIO_USERSPACE; 1824 1825 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1826 0, NULL, NULL, NULL)) != 0) 1827 return (error); 1828 1829 optype = VOP_MKNOD_DESCOFFSET; 1830 1831 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1832 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1833 1834 if ((error = namei(&nd)) != 0) 1835 goto out; 1836 vp = nd.ni_vp; 1837 if (vp != NULL) 1838 error = EEXIST; 1839 else { 1840 VATTR_NULL(&vattr); 1841 /* We will read cwdi->cwdi_cmask unlocked. */ 1842 vattr.va_mode = 1843 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1844 vattr.va_rdev = SCARG(uap, dev); 1845 1846 switch (SCARG(uap, mode) & S_IFMT) { 1847 case S_IFMT: /* used by badsect to flag bad sectors */ 1848 vattr.va_type = VBAD; 1849 break; 1850 case S_IFCHR: 1851 vattr.va_type = VCHR; 1852 break; 1853 case S_IFBLK: 1854 vattr.va_type = VBLK; 1855 break; 1856 case S_IFWHT: 1857 optype = VOP_WHITEOUT_DESCOFFSET; 1858 break; 1859 case S_IFREG: 1860 #if NVERIEXEC > 0 1861 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1862 O_CREAT); 1863 #endif /* NVERIEXEC > 0 */ 1864 vattr.va_type = VREG; 1865 vattr.va_rdev = VNOVAL; 1866 optype = VOP_CREATE_DESCOFFSET; 1867 break; 1868 default: 1869 error = EINVAL; 1870 break; 1871 } 1872 } 1873 if (!error) { 1874 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1875 switch (optype) { 1876 case VOP_WHITEOUT_DESCOFFSET: 1877 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1878 if (error) 1879 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1880 vput(nd.ni_dvp); 1881 break; 1882 1883 case VOP_MKNOD_DESCOFFSET: 1884 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1885 &nd.ni_cnd, &vattr); 1886 if (error == 0) 1887 vput(nd.ni_vp); 1888 break; 1889 1890 case VOP_CREATE_DESCOFFSET: 1891 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1892 &nd.ni_cnd, &vattr); 1893 if (error == 0) 1894 vput(nd.ni_vp); 1895 break; 1896 } 1897 } else { 1898 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1899 if (nd.ni_dvp == vp) 1900 vrele(nd.ni_dvp); 1901 else 1902 vput(nd.ni_dvp); 1903 if (vp) 1904 vrele(vp); 1905 } 1906 out: 1907 VERIEXEC_PATH_PUT(path); 1908 return (error); 1909 } 1910 1911 /* 1912 * Create a named pipe. 1913 */ 1914 /* ARGSUSED */ 1915 int 1916 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1917 { 1918 struct sys_mkfifo_args /* { 1919 syscallarg(const char *) path; 1920 syscallarg(int) mode; 1921 } */ *uap = v; 1922 struct proc *p = l->l_proc; 1923 struct vattr vattr; 1924 int error; 1925 struct nameidata nd; 1926 1927 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1928 if ((error = namei(&nd)) != 0) 1929 return (error); 1930 if (nd.ni_vp != NULL) { 1931 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1932 if (nd.ni_dvp == nd.ni_vp) 1933 vrele(nd.ni_dvp); 1934 else 1935 vput(nd.ni_dvp); 1936 vrele(nd.ni_vp); 1937 return (EEXIST); 1938 } 1939 VATTR_NULL(&vattr); 1940 vattr.va_type = VFIFO; 1941 /* We will read cwdi->cwdi_cmask unlocked. */ 1942 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1943 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1944 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1945 if (error == 0) 1946 vput(nd.ni_vp); 1947 return (error); 1948 } 1949 1950 /* 1951 * Make a hard file link. 1952 */ 1953 /* ARGSUSED */ 1954 int 1955 sys_link(struct lwp *l, void *v, register_t *retval) 1956 { 1957 struct sys_link_args /* { 1958 syscallarg(const char *) path; 1959 syscallarg(const char *) link; 1960 } */ *uap = v; 1961 struct vnode *vp; 1962 struct nameidata nd; 1963 int error; 1964 1965 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1966 if ((error = namei(&nd)) != 0) 1967 return (error); 1968 vp = nd.ni_vp; 1969 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1970 if ((error = namei(&nd)) != 0) 1971 goto out; 1972 if (nd.ni_vp) { 1973 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1974 if (nd.ni_dvp == nd.ni_vp) 1975 vrele(nd.ni_dvp); 1976 else 1977 vput(nd.ni_dvp); 1978 vrele(nd.ni_vp); 1979 error = EEXIST; 1980 goto out; 1981 } 1982 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1983 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1984 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1985 out: 1986 vrele(vp); 1987 return (error); 1988 } 1989 1990 /* 1991 * Make a symbolic link. 1992 */ 1993 /* ARGSUSED */ 1994 int 1995 sys_symlink(struct lwp *l, void *v, register_t *retval) 1996 { 1997 struct sys_symlink_args /* { 1998 syscallarg(const char *) path; 1999 syscallarg(const char *) link; 2000 } */ *uap = v; 2001 struct proc *p = l->l_proc; 2002 struct vattr vattr; 2003 char *path; 2004 int error; 2005 struct nameidata nd; 2006 2007 path = PNBUF_GET(); 2008 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2009 if (error) 2010 goto out; 2011 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 2012 if ((error = namei(&nd)) != 0) 2013 goto out; 2014 if (nd.ni_vp) { 2015 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2016 if (nd.ni_dvp == nd.ni_vp) 2017 vrele(nd.ni_dvp); 2018 else 2019 vput(nd.ni_dvp); 2020 vrele(nd.ni_vp); 2021 error = EEXIST; 2022 goto out; 2023 } 2024 VATTR_NULL(&vattr); 2025 vattr.va_type = VLNK; 2026 /* We will read cwdi->cwdi_cmask unlocked. */ 2027 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2028 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2029 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2030 if (error == 0) 2031 vput(nd.ni_vp); 2032 out: 2033 PNBUF_PUT(path); 2034 return (error); 2035 } 2036 2037 /* 2038 * Delete a whiteout from the filesystem. 2039 */ 2040 /* ARGSUSED */ 2041 int 2042 sys_undelete(struct lwp *l, void *v, register_t *retval) 2043 { 2044 struct sys_undelete_args /* { 2045 syscallarg(const char *) path; 2046 } */ *uap = v; 2047 int error; 2048 struct nameidata nd; 2049 2050 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 2051 SCARG(uap, path), l); 2052 error = namei(&nd); 2053 if (error) 2054 return (error); 2055 2056 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2057 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2058 if (nd.ni_dvp == nd.ni_vp) 2059 vrele(nd.ni_dvp); 2060 else 2061 vput(nd.ni_dvp); 2062 if (nd.ni_vp) 2063 vrele(nd.ni_vp); 2064 return (EEXIST); 2065 } 2066 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2067 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2068 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2069 vput(nd.ni_dvp); 2070 return (error); 2071 } 2072 2073 /* 2074 * Delete a name from the filesystem. 2075 */ 2076 /* ARGSUSED */ 2077 int 2078 sys_unlink(struct lwp *l, void *v, register_t *retval) 2079 { 2080 struct sys_unlink_args /* { 2081 syscallarg(const char *) path; 2082 } */ *uap = v; 2083 struct vnode *vp; 2084 int error; 2085 struct nameidata nd; 2086 char *path; 2087 const char *cpath; 2088 enum uio_seg seg = UIO_USERSPACE; 2089 2090 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2091 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2092 2093 if ((error = namei(&nd)) != 0) 2094 goto out; 2095 vp = nd.ni_vp; 2096 2097 /* 2098 * The root of a mounted filesystem cannot be deleted. 2099 */ 2100 if (vp->v_vflag & VV_ROOT) { 2101 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2102 if (nd.ni_dvp == vp) 2103 vrele(nd.ni_dvp); 2104 else 2105 vput(nd.ni_dvp); 2106 vput(vp); 2107 error = EBUSY; 2108 goto out; 2109 } 2110 2111 #if NVERIEXEC > 0 2112 /* Handle remove requests for veriexec entries. */ 2113 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2114 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2115 if (nd.ni_dvp == vp) 2116 vrele(nd.ni_dvp); 2117 else 2118 vput(nd.ni_dvp); 2119 vput(vp); 2120 goto out; 2121 } 2122 #endif /* NVERIEXEC > 0 */ 2123 2124 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2125 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2126 #ifdef FILEASSOC 2127 (void)fileassoc_file_delete(vp); 2128 #endif /* FILEASSOC */ 2129 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2130 out: 2131 VERIEXEC_PATH_PUT(path); 2132 return (error); 2133 } 2134 2135 /* 2136 * Reposition read/write file offset. 2137 */ 2138 int 2139 sys_lseek(struct lwp *l, void *v, register_t *retval) 2140 { 2141 struct sys_lseek_args /* { 2142 syscallarg(int) fd; 2143 syscallarg(int) pad; 2144 syscallarg(off_t) offset; 2145 syscallarg(int) whence; 2146 } */ *uap = v; 2147 struct proc *p = l->l_proc; 2148 kauth_cred_t cred = l->l_cred; 2149 struct filedesc *fdp = p->p_fd; 2150 struct file *fp; 2151 struct vnode *vp; 2152 struct vattr vattr; 2153 off_t newoff; 2154 int error; 2155 2156 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2157 return (EBADF); 2158 2159 vp = (struct vnode *)fp->f_data; 2160 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2161 error = ESPIPE; 2162 mutex_exit(&fp->f_lock); 2163 goto out; 2164 } 2165 2166 switch (SCARG(uap, whence)) { 2167 case SEEK_CUR: 2168 newoff = fp->f_offset + SCARG(uap, offset); 2169 FILE_USE(fp); 2170 break; 2171 case SEEK_END: 2172 FILE_USE(fp); 2173 error = VOP_GETATTR(vp, &vattr, cred, l); 2174 if (error) { 2175 FILE_UNUSE(fp, l); 2176 goto out; 2177 } 2178 newoff = SCARG(uap, offset) + vattr.va_size; 2179 break; 2180 case SEEK_SET: 2181 FILE_USE(fp); 2182 newoff = SCARG(uap, offset); 2183 break; 2184 default: 2185 mutex_exit(&fp->f_lock); 2186 error = EINVAL; 2187 goto out; 2188 } 2189 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2190 mutex_enter(&fp->f_lock); 2191 *(off_t *)retval = fp->f_offset = newoff; 2192 mutex_exit(&fp->f_lock); 2193 } 2194 FILE_UNUSE(fp, l); 2195 out: 2196 return (error); 2197 } 2198 2199 /* 2200 * Positional read system call. 2201 */ 2202 int 2203 sys_pread(struct lwp *l, void *v, register_t *retval) 2204 { 2205 struct sys_pread_args /* { 2206 syscallarg(int) fd; 2207 syscallarg(void *) buf; 2208 syscallarg(size_t) nbyte; 2209 syscallarg(off_t) offset; 2210 } */ *uap = v; 2211 struct proc *p = l->l_proc; 2212 struct filedesc *fdp = p->p_fd; 2213 struct file *fp; 2214 struct vnode *vp; 2215 off_t offset; 2216 int error, fd = SCARG(uap, fd); 2217 2218 if ((fp = fd_getfile(fdp, fd)) == NULL) 2219 return (EBADF); 2220 2221 if ((fp->f_flag & FREAD) == 0) { 2222 mutex_exit(&fp->f_lock); 2223 return (EBADF); 2224 } 2225 2226 FILE_USE(fp); 2227 2228 vp = (struct vnode *)fp->f_data; 2229 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2230 error = ESPIPE; 2231 goto out; 2232 } 2233 2234 offset = SCARG(uap, offset); 2235 2236 /* 2237 * XXX This works because no file systems actually 2238 * XXX take any action on the seek operation. 2239 */ 2240 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2241 goto out; 2242 2243 /* dofileread() will unuse the descriptor for us */ 2244 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2245 &offset, 0, retval)); 2246 2247 out: 2248 FILE_UNUSE(fp, l); 2249 return (error); 2250 } 2251 2252 /* 2253 * Positional scatter read system call. 2254 */ 2255 int 2256 sys_preadv(struct lwp *l, void *v, register_t *retval) 2257 { 2258 struct sys_preadv_args /* { 2259 syscallarg(int) fd; 2260 syscallarg(const struct iovec *) iovp; 2261 syscallarg(int) iovcnt; 2262 syscallarg(off_t) offset; 2263 } */ *uap = v; 2264 2265 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2266 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2267 } 2268 2269 /* 2270 * Positional write system call. 2271 */ 2272 int 2273 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2274 { 2275 struct sys_pwrite_args /* { 2276 syscallarg(int) fd; 2277 syscallarg(const void *) buf; 2278 syscallarg(size_t) nbyte; 2279 syscallarg(off_t) offset; 2280 } */ *uap = v; 2281 struct proc *p = l->l_proc; 2282 struct filedesc *fdp = p->p_fd; 2283 struct file *fp; 2284 struct vnode *vp; 2285 off_t offset; 2286 int error, fd = SCARG(uap, fd); 2287 2288 if ((fp = fd_getfile(fdp, fd)) == NULL) 2289 return (EBADF); 2290 2291 if ((fp->f_flag & FWRITE) == 0) { 2292 mutex_exit(&fp->f_lock); 2293 return (EBADF); 2294 } 2295 2296 FILE_USE(fp); 2297 2298 vp = (struct vnode *)fp->f_data; 2299 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2300 error = ESPIPE; 2301 goto out; 2302 } 2303 2304 offset = SCARG(uap, offset); 2305 2306 /* 2307 * XXX This works because no file systems actually 2308 * XXX take any action on the seek operation. 2309 */ 2310 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2311 goto out; 2312 2313 /* dofilewrite() will unuse the descriptor for us */ 2314 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2315 &offset, 0, retval)); 2316 2317 out: 2318 FILE_UNUSE(fp, l); 2319 return (error); 2320 } 2321 2322 /* 2323 * Positional gather write system call. 2324 */ 2325 int 2326 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2327 { 2328 struct sys_pwritev_args /* { 2329 syscallarg(int) fd; 2330 syscallarg(const struct iovec *) iovp; 2331 syscallarg(int) iovcnt; 2332 syscallarg(off_t) offset; 2333 } */ *uap = v; 2334 2335 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2336 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2337 } 2338 2339 /* 2340 * Check access permissions. 2341 */ 2342 int 2343 sys_access(struct lwp *l, void *v, register_t *retval) 2344 { 2345 struct sys_access_args /* { 2346 syscallarg(const char *) path; 2347 syscallarg(int) flags; 2348 } */ *uap = v; 2349 kauth_cred_t cred; 2350 struct vnode *vp; 2351 int error, flags; 2352 struct nameidata nd; 2353 2354 cred = kauth_cred_dup(l->l_cred); 2355 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2356 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2357 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2358 SCARG(uap, path), l); 2359 /* Override default credentials */ 2360 nd.ni_cnd.cn_cred = cred; 2361 if ((error = namei(&nd)) != 0) 2362 goto out; 2363 vp = nd.ni_vp; 2364 2365 /* Flags == 0 means only check for existence. */ 2366 if (SCARG(uap, flags)) { 2367 flags = 0; 2368 if (SCARG(uap, flags) & R_OK) 2369 flags |= VREAD; 2370 if (SCARG(uap, flags) & W_OK) 2371 flags |= VWRITE; 2372 if (SCARG(uap, flags) & X_OK) 2373 flags |= VEXEC; 2374 2375 error = VOP_ACCESS(vp, flags, cred, l); 2376 if (!error && (flags & VWRITE)) 2377 error = vn_writechk(vp); 2378 } 2379 vput(vp); 2380 out: 2381 kauth_cred_free(cred); 2382 return (error); 2383 } 2384 2385 /* 2386 * Common code for all sys_stat functions, including compat versions. 2387 */ 2388 int 2389 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2390 struct stat *sb) 2391 { 2392 int error; 2393 struct nameidata nd; 2394 2395 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2396 error = namei(&nd); 2397 if (error != 0) 2398 return error; 2399 error = vn_stat(nd.ni_vp, sb, l); 2400 vput(nd.ni_vp); 2401 return error; 2402 } 2403 2404 /* 2405 * Get file status; this version follows links. 2406 */ 2407 /* ARGSUSED */ 2408 int 2409 sys___stat30(struct lwp *l, void *v, register_t *retval) 2410 { 2411 struct sys___stat30_args /* { 2412 syscallarg(const char *) path; 2413 syscallarg(struct stat *) ub; 2414 } */ *uap = v; 2415 struct stat sb; 2416 int error; 2417 2418 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2419 if (error) 2420 return error; 2421 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2422 } 2423 2424 /* 2425 * Get file status; this version does not follow links. 2426 */ 2427 /* ARGSUSED */ 2428 int 2429 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2430 { 2431 struct sys___lstat30_args /* { 2432 syscallarg(const char *) path; 2433 syscallarg(struct stat *) ub; 2434 } */ *uap = v; 2435 struct stat sb; 2436 int error; 2437 2438 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2439 if (error) 2440 return error; 2441 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2442 } 2443 2444 /* 2445 * Get configurable pathname variables. 2446 */ 2447 /* ARGSUSED */ 2448 int 2449 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2450 { 2451 struct sys_pathconf_args /* { 2452 syscallarg(const char *) path; 2453 syscallarg(int) name; 2454 } */ *uap = v; 2455 int error; 2456 struct nameidata nd; 2457 2458 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2459 SCARG(uap, path), l); 2460 if ((error = namei(&nd)) != 0) 2461 return (error); 2462 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2463 vput(nd.ni_vp); 2464 return (error); 2465 } 2466 2467 /* 2468 * Return target name of a symbolic link. 2469 */ 2470 /* ARGSUSED */ 2471 int 2472 sys_readlink(struct lwp *l, void *v, register_t *retval) 2473 { 2474 struct sys_readlink_args /* { 2475 syscallarg(const char *) path; 2476 syscallarg(char *) buf; 2477 syscallarg(size_t) count; 2478 } */ *uap = v; 2479 struct vnode *vp; 2480 struct iovec aiov; 2481 struct uio auio; 2482 int error; 2483 struct nameidata nd; 2484 2485 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2486 SCARG(uap, path), l); 2487 if ((error = namei(&nd)) != 0) 2488 return (error); 2489 vp = nd.ni_vp; 2490 if (vp->v_type != VLNK) 2491 error = EINVAL; 2492 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2493 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2494 aiov.iov_base = SCARG(uap, buf); 2495 aiov.iov_len = SCARG(uap, count); 2496 auio.uio_iov = &aiov; 2497 auio.uio_iovcnt = 1; 2498 auio.uio_offset = 0; 2499 auio.uio_rw = UIO_READ; 2500 KASSERT(l == curlwp); 2501 auio.uio_vmspace = l->l_proc->p_vmspace; 2502 auio.uio_resid = SCARG(uap, count); 2503 error = VOP_READLINK(vp, &auio, l->l_cred); 2504 } 2505 vput(vp); 2506 *retval = SCARG(uap, count) - auio.uio_resid; 2507 return (error); 2508 } 2509 2510 /* 2511 * Change flags of a file given a path name. 2512 */ 2513 /* ARGSUSED */ 2514 int 2515 sys_chflags(struct lwp *l, void *v, register_t *retval) 2516 { 2517 struct sys_chflags_args /* { 2518 syscallarg(const char *) path; 2519 syscallarg(u_long) flags; 2520 } */ *uap = v; 2521 struct vnode *vp; 2522 int error; 2523 struct nameidata nd; 2524 2525 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2526 if ((error = namei(&nd)) != 0) 2527 return (error); 2528 vp = nd.ni_vp; 2529 error = change_flags(vp, SCARG(uap, flags), l); 2530 vput(vp); 2531 return (error); 2532 } 2533 2534 /* 2535 * Change flags of a file given a file descriptor. 2536 */ 2537 /* ARGSUSED */ 2538 int 2539 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2540 { 2541 struct sys_fchflags_args /* { 2542 syscallarg(int) fd; 2543 syscallarg(u_long) flags; 2544 } */ *uap = v; 2545 struct proc *p = l->l_proc; 2546 struct vnode *vp; 2547 struct file *fp; 2548 int error; 2549 2550 /* getvnode() will use the descriptor for us */ 2551 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2552 return (error); 2553 vp = (struct vnode *)fp->f_data; 2554 error = change_flags(vp, SCARG(uap, flags), l); 2555 VOP_UNLOCK(vp, 0); 2556 FILE_UNUSE(fp, l); 2557 return (error); 2558 } 2559 2560 /* 2561 * Change flags of a file given a path name; this version does 2562 * not follow links. 2563 */ 2564 int 2565 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2566 { 2567 struct sys_lchflags_args /* { 2568 syscallarg(const char *) path; 2569 syscallarg(u_long) flags; 2570 } */ *uap = v; 2571 struct vnode *vp; 2572 int error; 2573 struct nameidata nd; 2574 2575 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2576 if ((error = namei(&nd)) != 0) 2577 return (error); 2578 vp = nd.ni_vp; 2579 error = change_flags(vp, SCARG(uap, flags), l); 2580 vput(vp); 2581 return (error); 2582 } 2583 2584 /* 2585 * Common routine to change flags of a file. 2586 */ 2587 int 2588 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2589 { 2590 struct vattr vattr; 2591 int error; 2592 2593 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2594 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2595 /* 2596 * Non-superusers cannot change the flags on devices, even if they 2597 * own them. 2598 */ 2599 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2600 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2601 goto out; 2602 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2603 error = EINVAL; 2604 goto out; 2605 } 2606 } 2607 VATTR_NULL(&vattr); 2608 vattr.va_flags = flags; 2609 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2610 out: 2611 return (error); 2612 } 2613 2614 /* 2615 * Change mode of a file given path name; this version follows links. 2616 */ 2617 /* ARGSUSED */ 2618 int 2619 sys_chmod(struct lwp *l, void *v, register_t *retval) 2620 { 2621 struct sys_chmod_args /* { 2622 syscallarg(const char *) path; 2623 syscallarg(int) mode; 2624 } */ *uap = v; 2625 int error; 2626 struct nameidata nd; 2627 2628 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2629 if ((error = namei(&nd)) != 0) 2630 return (error); 2631 2632 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2633 2634 vrele(nd.ni_vp); 2635 return (error); 2636 } 2637 2638 /* 2639 * Change mode of a file given a file descriptor. 2640 */ 2641 /* ARGSUSED */ 2642 int 2643 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2644 { 2645 struct sys_fchmod_args /* { 2646 syscallarg(int) fd; 2647 syscallarg(int) mode; 2648 } */ *uap = v; 2649 struct proc *p = l->l_proc; 2650 struct file *fp; 2651 int error; 2652 2653 /* getvnode() will use the descriptor for us */ 2654 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2655 return (error); 2656 2657 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2658 FILE_UNUSE(fp, l); 2659 return (error); 2660 } 2661 2662 /* 2663 * Change mode of a file given path name; this version does not follow links. 2664 */ 2665 /* ARGSUSED */ 2666 int 2667 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2668 { 2669 struct sys_lchmod_args /* { 2670 syscallarg(const char *) path; 2671 syscallarg(int) mode; 2672 } */ *uap = v; 2673 int error; 2674 struct nameidata nd; 2675 2676 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2677 if ((error = namei(&nd)) != 0) 2678 return (error); 2679 2680 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2681 2682 vrele(nd.ni_vp); 2683 return (error); 2684 } 2685 2686 /* 2687 * Common routine to set mode given a vnode. 2688 */ 2689 static int 2690 change_mode(struct vnode *vp, int mode, struct lwp *l) 2691 { 2692 struct vattr vattr; 2693 int error; 2694 2695 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2696 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2697 VATTR_NULL(&vattr); 2698 vattr.va_mode = mode & ALLPERMS; 2699 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2700 VOP_UNLOCK(vp, 0); 2701 return (error); 2702 } 2703 2704 /* 2705 * Set ownership given a path name; this version follows links. 2706 */ 2707 /* ARGSUSED */ 2708 int 2709 sys_chown(struct lwp *l, void *v, register_t *retval) 2710 { 2711 struct sys_chown_args /* { 2712 syscallarg(const char *) path; 2713 syscallarg(uid_t) uid; 2714 syscallarg(gid_t) gid; 2715 } */ *uap = v; 2716 int error; 2717 struct nameidata nd; 2718 2719 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2720 if ((error = namei(&nd)) != 0) 2721 return (error); 2722 2723 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2724 2725 vrele(nd.ni_vp); 2726 return (error); 2727 } 2728 2729 /* 2730 * Set ownership given a path name; this version follows links. 2731 * Provides POSIX semantics. 2732 */ 2733 /* ARGSUSED */ 2734 int 2735 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2736 { 2737 struct sys_chown_args /* { 2738 syscallarg(const char *) path; 2739 syscallarg(uid_t) uid; 2740 syscallarg(gid_t) gid; 2741 } */ *uap = v; 2742 int error; 2743 struct nameidata nd; 2744 2745 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2746 if ((error = namei(&nd)) != 0) 2747 return (error); 2748 2749 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2750 2751 vrele(nd.ni_vp); 2752 return (error); 2753 } 2754 2755 /* 2756 * Set ownership given a file descriptor. 2757 */ 2758 /* ARGSUSED */ 2759 int 2760 sys_fchown(struct lwp *l, void *v, register_t *retval) 2761 { 2762 struct sys_fchown_args /* { 2763 syscallarg(int) fd; 2764 syscallarg(uid_t) uid; 2765 syscallarg(gid_t) gid; 2766 } */ *uap = v; 2767 struct proc *p = l->l_proc; 2768 int error; 2769 struct file *fp; 2770 2771 /* getvnode() will use the descriptor for us */ 2772 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2773 return (error); 2774 2775 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2776 SCARG(uap, gid), l, 0); 2777 FILE_UNUSE(fp, l); 2778 return (error); 2779 } 2780 2781 /* 2782 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2783 */ 2784 /* ARGSUSED */ 2785 int 2786 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2787 { 2788 struct sys_fchown_args /* { 2789 syscallarg(int) fd; 2790 syscallarg(uid_t) uid; 2791 syscallarg(gid_t) gid; 2792 } */ *uap = v; 2793 struct proc *p = l->l_proc; 2794 int error; 2795 struct file *fp; 2796 2797 /* getvnode() will use the descriptor for us */ 2798 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2799 return (error); 2800 2801 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2802 SCARG(uap, gid), l, 1); 2803 FILE_UNUSE(fp, l); 2804 return (error); 2805 } 2806 2807 /* 2808 * Set ownership given a path name; this version does not follow links. 2809 */ 2810 /* ARGSUSED */ 2811 int 2812 sys_lchown(struct lwp *l, void *v, register_t *retval) 2813 { 2814 struct sys_lchown_args /* { 2815 syscallarg(const char *) path; 2816 syscallarg(uid_t) uid; 2817 syscallarg(gid_t) gid; 2818 } */ *uap = v; 2819 int error; 2820 struct nameidata nd; 2821 2822 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2823 if ((error = namei(&nd)) != 0) 2824 return (error); 2825 2826 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2827 2828 vrele(nd.ni_vp); 2829 return (error); 2830 } 2831 2832 /* 2833 * Set ownership given a path name; this version does not follow links. 2834 * Provides POSIX/XPG semantics. 2835 */ 2836 /* ARGSUSED */ 2837 int 2838 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2839 { 2840 struct sys_lchown_args /* { 2841 syscallarg(const char *) path; 2842 syscallarg(uid_t) uid; 2843 syscallarg(gid_t) gid; 2844 } */ *uap = v; 2845 int error; 2846 struct nameidata nd; 2847 2848 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2849 if ((error = namei(&nd)) != 0) 2850 return (error); 2851 2852 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2853 2854 vrele(nd.ni_vp); 2855 return (error); 2856 } 2857 2858 /* 2859 * Common routine to set ownership given a vnode. 2860 */ 2861 static int 2862 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2863 int posix_semantics) 2864 { 2865 struct vattr vattr; 2866 mode_t newmode; 2867 int error; 2868 2869 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2870 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2871 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2872 goto out; 2873 2874 #define CHANGED(x) ((int)(x) != -1) 2875 newmode = vattr.va_mode; 2876 if (posix_semantics) { 2877 /* 2878 * POSIX/XPG semantics: if the caller is not the super-user, 2879 * clear set-user-id and set-group-id bits. Both POSIX and 2880 * the XPG consider the behaviour for calls by the super-user 2881 * implementation-defined; we leave the set-user-id and set- 2882 * group-id settings intact in that case. 2883 */ 2884 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2885 NULL) != 0) 2886 newmode &= ~(S_ISUID | S_ISGID); 2887 } else { 2888 /* 2889 * NetBSD semantics: when changing owner and/or group, 2890 * clear the respective bit(s). 2891 */ 2892 if (CHANGED(uid)) 2893 newmode &= ~S_ISUID; 2894 if (CHANGED(gid)) 2895 newmode &= ~S_ISGID; 2896 } 2897 /* Update va_mode iff altered. */ 2898 if (vattr.va_mode == newmode) 2899 newmode = VNOVAL; 2900 2901 VATTR_NULL(&vattr); 2902 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2903 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2904 vattr.va_mode = newmode; 2905 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2906 #undef CHANGED 2907 2908 out: 2909 VOP_UNLOCK(vp, 0); 2910 return (error); 2911 } 2912 2913 /* 2914 * Set the access and modification times given a path name; this 2915 * version follows links. 2916 */ 2917 /* ARGSUSED */ 2918 int 2919 sys_utimes(struct lwp *l, void *v, register_t *retval) 2920 { 2921 struct sys_utimes_args /* { 2922 syscallarg(const char *) path; 2923 syscallarg(const struct timeval *) tptr; 2924 } */ *uap = v; 2925 2926 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2927 SCARG(uap, tptr), UIO_USERSPACE); 2928 } 2929 2930 /* 2931 * Set the access and modification times given a file descriptor. 2932 */ 2933 /* ARGSUSED */ 2934 int 2935 sys_futimes(struct lwp *l, void *v, register_t *retval) 2936 { 2937 struct sys_futimes_args /* { 2938 syscallarg(int) fd; 2939 syscallarg(const struct timeval *) tptr; 2940 } */ *uap = v; 2941 int error; 2942 struct file *fp; 2943 2944 /* getvnode() will use the descriptor for us */ 2945 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2946 return (error); 2947 2948 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2949 SCARG(uap, tptr), UIO_USERSPACE); 2950 2951 FILE_UNUSE(fp, l); 2952 return (error); 2953 } 2954 2955 /* 2956 * Set the access and modification times given a path name; this 2957 * version does not follow links. 2958 */ 2959 int 2960 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2961 { 2962 struct sys_lutimes_args /* { 2963 syscallarg(const char *) path; 2964 syscallarg(const struct timeval *) tptr; 2965 } */ *uap = v; 2966 2967 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2968 SCARG(uap, tptr), UIO_USERSPACE); 2969 } 2970 2971 /* 2972 * Common routine to set access and modification times given a vnode. 2973 */ 2974 int 2975 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2976 const struct timeval *tptr, enum uio_seg seg) 2977 { 2978 struct vattr vattr; 2979 struct nameidata nd; 2980 int error; 2981 2982 VATTR_NULL(&vattr); 2983 if (tptr == NULL) { 2984 nanotime(&vattr.va_atime); 2985 vattr.va_mtime = vattr.va_atime; 2986 vattr.va_vaflags |= VA_UTIMES_NULL; 2987 } else { 2988 struct timeval tv[2]; 2989 2990 if (seg != UIO_SYSSPACE) { 2991 error = copyin(tptr, &tv, sizeof (tv)); 2992 if (error != 0) 2993 return error; 2994 tptr = tv; 2995 } 2996 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 2997 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 2998 } 2999 3000 if (vp == NULL) { 3001 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 3002 if ((error = namei(&nd)) != 0) 3003 return (error); 3004 vp = nd.ni_vp; 3005 } else 3006 nd.ni_vp = NULL; 3007 3008 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3009 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3010 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3011 VOP_UNLOCK(vp, 0); 3012 3013 if (nd.ni_vp != NULL) 3014 vrele(nd.ni_vp); 3015 3016 return (error); 3017 } 3018 3019 /* 3020 * Truncate a file given its path name. 3021 */ 3022 /* ARGSUSED */ 3023 int 3024 sys_truncate(struct lwp *l, void *v, register_t *retval) 3025 { 3026 struct sys_truncate_args /* { 3027 syscallarg(const char *) path; 3028 syscallarg(int) pad; 3029 syscallarg(off_t) length; 3030 } */ *uap = v; 3031 struct vnode *vp; 3032 struct vattr vattr; 3033 int error; 3034 struct nameidata nd; 3035 3036 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3037 if ((error = namei(&nd)) != 0) 3038 return (error); 3039 vp = nd.ni_vp; 3040 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3041 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3042 if (vp->v_type == VDIR) 3043 error = EISDIR; 3044 else if ((error = vn_writechk(vp)) == 0 && 3045 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3046 VATTR_NULL(&vattr); 3047 vattr.va_size = SCARG(uap, length); 3048 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3049 } 3050 vput(vp); 3051 return (error); 3052 } 3053 3054 /* 3055 * Truncate a file given a file descriptor. 3056 */ 3057 /* ARGSUSED */ 3058 int 3059 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3060 { 3061 struct sys_ftruncate_args /* { 3062 syscallarg(int) fd; 3063 syscallarg(int) pad; 3064 syscallarg(off_t) length; 3065 } */ *uap = v; 3066 struct proc *p = l->l_proc; 3067 struct vattr vattr; 3068 struct vnode *vp; 3069 struct file *fp; 3070 int error; 3071 3072 /* getvnode() will use the descriptor for us */ 3073 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3074 return (error); 3075 if ((fp->f_flag & FWRITE) == 0) { 3076 error = EINVAL; 3077 goto out; 3078 } 3079 vp = (struct vnode *)fp->f_data; 3080 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3081 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3082 if (vp->v_type == VDIR) 3083 error = EISDIR; 3084 else if ((error = vn_writechk(vp)) == 0) { 3085 VATTR_NULL(&vattr); 3086 vattr.va_size = SCARG(uap, length); 3087 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3088 } 3089 VOP_UNLOCK(vp, 0); 3090 out: 3091 FILE_UNUSE(fp, l); 3092 return (error); 3093 } 3094 3095 /* 3096 * Sync an open file. 3097 */ 3098 /* ARGSUSED */ 3099 int 3100 sys_fsync(struct lwp *l, void *v, register_t *retval) 3101 { 3102 struct sys_fsync_args /* { 3103 syscallarg(int) fd; 3104 } */ *uap = v; 3105 struct proc *p = l->l_proc; 3106 struct vnode *vp; 3107 struct file *fp; 3108 int error; 3109 3110 /* getvnode() will use the descriptor for us */ 3111 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3112 return (error); 3113 vp = (struct vnode *)fp->f_data; 3114 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3115 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3116 if (error == 0 && bioopsp != NULL && 3117 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3118 (*bioopsp->io_fsync)(vp, 0); 3119 VOP_UNLOCK(vp, 0); 3120 FILE_UNUSE(fp, l); 3121 return (error); 3122 } 3123 3124 /* 3125 * Sync a range of file data. API modeled after that found in AIX. 3126 * 3127 * FDATASYNC indicates that we need only save enough metadata to be able 3128 * to re-read the written data. Note we duplicate AIX's requirement that 3129 * the file be open for writing. 3130 */ 3131 /* ARGSUSED */ 3132 int 3133 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3134 { 3135 struct sys_fsync_range_args /* { 3136 syscallarg(int) fd; 3137 syscallarg(int) flags; 3138 syscallarg(off_t) start; 3139 syscallarg(off_t) length; 3140 } */ *uap = v; 3141 struct proc *p = l->l_proc; 3142 struct vnode *vp; 3143 struct file *fp; 3144 int flags, nflags; 3145 off_t s, e, len; 3146 int error; 3147 3148 /* getvnode() will use the descriptor for us */ 3149 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3150 return (error); 3151 3152 if ((fp->f_flag & FWRITE) == 0) { 3153 error = EBADF; 3154 goto out; 3155 } 3156 3157 flags = SCARG(uap, flags); 3158 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3159 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3160 error = EINVAL; 3161 goto out; 3162 } 3163 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3164 if (flags & FDATASYNC) 3165 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3166 else 3167 nflags = FSYNC_WAIT; 3168 if (flags & FDISKSYNC) 3169 nflags |= FSYNC_CACHE; 3170 3171 len = SCARG(uap, length); 3172 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3173 if (len) { 3174 s = SCARG(uap, start); 3175 e = s + len; 3176 if (e < s) { 3177 error = EINVAL; 3178 goto out; 3179 } 3180 } else { 3181 e = 0; 3182 s = 0; 3183 } 3184 3185 vp = (struct vnode *)fp->f_data; 3186 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3187 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3188 3189 if (error == 0 && bioopsp != NULL && 3190 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3191 (*bioopsp->io_fsync)(vp, nflags); 3192 3193 VOP_UNLOCK(vp, 0); 3194 out: 3195 FILE_UNUSE(fp, l); 3196 return (error); 3197 } 3198 3199 /* 3200 * Sync the data of an open file. 3201 */ 3202 /* ARGSUSED */ 3203 int 3204 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3205 { 3206 struct sys_fdatasync_args /* { 3207 syscallarg(int) fd; 3208 } */ *uap = v; 3209 struct proc *p = l->l_proc; 3210 struct vnode *vp; 3211 struct file *fp; 3212 int error; 3213 3214 /* getvnode() will use the descriptor for us */ 3215 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3216 return (error); 3217 if ((fp->f_flag & FWRITE) == 0) { 3218 FILE_UNUSE(fp, l); 3219 return (EBADF); 3220 } 3221 vp = (struct vnode *)fp->f_data; 3222 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3223 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3224 VOP_UNLOCK(vp, 0); 3225 FILE_UNUSE(fp, l); 3226 return (error); 3227 } 3228 3229 /* 3230 * Rename files, (standard) BSD semantics frontend. 3231 */ 3232 /* ARGSUSED */ 3233 int 3234 sys_rename(struct lwp *l, void *v, register_t *retval) 3235 { 3236 struct sys_rename_args /* { 3237 syscallarg(const char *) from; 3238 syscallarg(const char *) to; 3239 } */ *uap = v; 3240 3241 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3242 } 3243 3244 /* 3245 * Rename files, POSIX semantics frontend. 3246 */ 3247 /* ARGSUSED */ 3248 int 3249 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3250 { 3251 struct sys___posix_rename_args /* { 3252 syscallarg(const char *) from; 3253 syscallarg(const char *) to; 3254 } */ *uap = v; 3255 3256 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3257 } 3258 3259 /* 3260 * Rename files. Source and destination must either both be directories, 3261 * or both not be directories. If target is a directory, it must be empty. 3262 * If `from' and `to' refer to the same object, the value of the `retain' 3263 * argument is used to determine whether `from' will be 3264 * 3265 * (retain == 0) deleted unless `from' and `to' refer to the same 3266 * object in the file system's name space (BSD). 3267 * (retain == 1) always retained (POSIX). 3268 */ 3269 static int 3270 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3271 { 3272 struct vnode *tvp, *fvp, *tdvp; 3273 struct nameidata fromnd, tond; 3274 struct proc *p; 3275 int error; 3276 3277 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3278 from, l); 3279 if ((error = namei(&fromnd)) != 0) 3280 return (error); 3281 if (fromnd.ni_dvp != fromnd.ni_vp) 3282 VOP_UNLOCK(fromnd.ni_dvp, 0); 3283 fvp = fromnd.ni_vp; 3284 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3285 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3286 if ((error = namei(&tond)) != 0) { 3287 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3288 vrele(fromnd.ni_dvp); 3289 vrele(fvp); 3290 goto out1; 3291 } 3292 tdvp = tond.ni_dvp; 3293 tvp = tond.ni_vp; 3294 3295 if (tvp != NULL) { 3296 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3297 error = ENOTDIR; 3298 goto out; 3299 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3300 error = EISDIR; 3301 goto out; 3302 } 3303 } 3304 3305 if (fvp == tdvp) 3306 error = EINVAL; 3307 3308 /* 3309 * Source and destination refer to the same object. 3310 */ 3311 if (fvp == tvp) { 3312 if (retain) 3313 error = -1; 3314 else if (fromnd.ni_dvp == tdvp && 3315 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3316 !memcmp(fromnd.ni_cnd.cn_nameptr, 3317 tond.ni_cnd.cn_nameptr, 3318 fromnd.ni_cnd.cn_namelen)) 3319 error = -1; 3320 } 3321 3322 #if NVERIEXEC > 0 3323 if (!error) { 3324 char *f1, *f2; 3325 3326 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3327 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3328 3329 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3330 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3331 3332 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3333 3334 free(f1, M_TEMP); 3335 free(f2, M_TEMP); 3336 } 3337 #endif /* NVERIEXEC > 0 */ 3338 3339 out: 3340 p = l->l_proc; 3341 if (!error) { 3342 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3343 if (fromnd.ni_dvp != tdvp) 3344 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3345 if (tvp) { 3346 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3347 } 3348 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3349 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3350 } else { 3351 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3352 if (tdvp == tvp) 3353 vrele(tdvp); 3354 else 3355 vput(tdvp); 3356 if (tvp) 3357 vput(tvp); 3358 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3359 vrele(fromnd.ni_dvp); 3360 vrele(fvp); 3361 } 3362 vrele(tond.ni_startdir); 3363 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3364 out1: 3365 if (fromnd.ni_startdir) 3366 vrele(fromnd.ni_startdir); 3367 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3368 return (error == -1 ? 0 : error); 3369 } 3370 3371 /* 3372 * Make a directory file. 3373 */ 3374 /* ARGSUSED */ 3375 int 3376 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3377 { 3378 struct sys_mkdir_args /* { 3379 syscallarg(const char *) path; 3380 syscallarg(int) mode; 3381 } */ *uap = v; 3382 struct proc *p = l->l_proc; 3383 struct vnode *vp; 3384 struct vattr vattr; 3385 int error; 3386 struct nameidata nd; 3387 3388 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3389 SCARG(uap, path), l); 3390 if ((error = namei(&nd)) != 0) 3391 return (error); 3392 vp = nd.ni_vp; 3393 if (vp != NULL) { 3394 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3395 if (nd.ni_dvp == vp) 3396 vrele(nd.ni_dvp); 3397 else 3398 vput(nd.ni_dvp); 3399 vrele(vp); 3400 return (EEXIST); 3401 } 3402 VATTR_NULL(&vattr); 3403 vattr.va_type = VDIR; 3404 /* We will read cwdi->cwdi_cmask unlocked. */ 3405 vattr.va_mode = 3406 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3407 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3408 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3409 if (!error) 3410 vput(nd.ni_vp); 3411 return (error); 3412 } 3413 3414 /* 3415 * Remove a directory file. 3416 */ 3417 /* ARGSUSED */ 3418 int 3419 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3420 { 3421 struct sys_rmdir_args /* { 3422 syscallarg(const char *) path; 3423 } */ *uap = v; 3424 struct vnode *vp; 3425 int error; 3426 struct nameidata nd; 3427 3428 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3429 SCARG(uap, path), l); 3430 if ((error = namei(&nd)) != 0) 3431 return (error); 3432 vp = nd.ni_vp; 3433 if (vp->v_type != VDIR) { 3434 error = ENOTDIR; 3435 goto out; 3436 } 3437 /* 3438 * No rmdir "." please. 3439 */ 3440 if (nd.ni_dvp == vp) { 3441 error = EINVAL; 3442 goto out; 3443 } 3444 /* 3445 * The root of a mounted filesystem cannot be deleted. 3446 */ 3447 if (vp->v_vflag & VV_ROOT) { 3448 error = EBUSY; 3449 goto out; 3450 } 3451 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3452 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3453 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3454 return (error); 3455 3456 out: 3457 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3458 if (nd.ni_dvp == vp) 3459 vrele(nd.ni_dvp); 3460 else 3461 vput(nd.ni_dvp); 3462 vput(vp); 3463 return (error); 3464 } 3465 3466 /* 3467 * Read a block of directory entries in a file system independent format. 3468 */ 3469 int 3470 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3471 { 3472 struct sys___getdents30_args /* { 3473 syscallarg(int) fd; 3474 syscallarg(char *) buf; 3475 syscallarg(size_t) count; 3476 } */ *uap = v; 3477 struct proc *p = l->l_proc; 3478 struct file *fp; 3479 int error, done; 3480 3481 /* getvnode() will use the descriptor for us */ 3482 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3483 return (error); 3484 if ((fp->f_flag & FREAD) == 0) { 3485 error = EBADF; 3486 goto out; 3487 } 3488 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3489 SCARG(uap, count), &done, l, 0, 0); 3490 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3491 *retval = done; 3492 out: 3493 FILE_UNUSE(fp, l); 3494 return (error); 3495 } 3496 3497 /* 3498 * Set the mode mask for creation of filesystem nodes. 3499 */ 3500 int 3501 sys_umask(struct lwp *l, void *v, register_t *retval) 3502 { 3503 struct sys_umask_args /* { 3504 syscallarg(mode_t) newmask; 3505 } */ *uap = v; 3506 struct proc *p = l->l_proc; 3507 struct cwdinfo *cwdi; 3508 3509 /* 3510 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3511 * important is that we serialize changes to the mask. The 3512 * rw_exit() will issue a write memory barrier on our behalf, 3513 * and force the changes out to other CPUs (as it must use an 3514 * atomic operation, draining the local CPU's store buffers). 3515 */ 3516 cwdi = p->p_cwdi; 3517 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3518 *retval = cwdi->cwdi_cmask; 3519 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3520 rw_exit(&cwdi->cwdi_lock); 3521 3522 return (0); 3523 } 3524 3525 /* 3526 * Void all references to file by ripping underlying filesystem 3527 * away from vnode. 3528 */ 3529 /* ARGSUSED */ 3530 int 3531 sys_revoke(struct lwp *l, void *v, register_t *retval) 3532 { 3533 struct sys_revoke_args /* { 3534 syscallarg(const char *) path; 3535 } */ *uap = v; 3536 struct vnode *vp; 3537 struct vattr vattr; 3538 int error; 3539 bool revoke; 3540 struct nameidata nd; 3541 3542 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3543 if ((error = namei(&nd)) != 0) 3544 return (error); 3545 vp = nd.ni_vp; 3546 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3547 goto out; 3548 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3549 (error = kauth_authorize_generic(l->l_cred, 3550 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3551 goto out; 3552 simple_lock(&vp->v_interlock); 3553 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED | VI_LAYER))); 3554 simple_unlock(&vp->v_interlock); 3555 if (revoke) 3556 VOP_REVOKE(vp, REVOKEALL); 3557 out: 3558 vrele(vp); 3559 return (error); 3560 } 3561 3562 /* 3563 * Convert a user file descriptor to a kernel file entry. 3564 */ 3565 int 3566 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3567 { 3568 struct vnode *vp; 3569 struct file *fp; 3570 3571 if ((fp = fd_getfile(fdp, fd)) == NULL) 3572 return (EBADF); 3573 3574 FILE_USE(fp); 3575 3576 if (fp->f_type != DTYPE_VNODE) { 3577 FILE_UNUSE(fp, NULL); 3578 return (EINVAL); 3579 } 3580 3581 vp = (struct vnode *)fp->f_data; 3582 if (vp->v_type == VBAD) { 3583 FILE_UNUSE(fp, NULL); 3584 return (EBADF); 3585 } 3586 3587 *fpp = fp; 3588 return (0); 3589 } 3590
Indexes created Fri Oct 03 02:09:57 GMT 2025