vfs_syscalls.c revision 1.223.2.7
1 /* $NetBSD: vfs_syscalls.c,v 1.223.2.7 2008/01/21 09:46:33 yamt 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 1.223.2.7 2008/01/21 09:46:33 yamt 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 96 void checkdirs(struct vnode *); 97 98 int dovfsusermount = 0; 99 100 /* 101 * Virtual File System System Calls 102 */ 103 104 /* 105 * Mount a file system. 106 */ 107 108 #if defined(COMPAT_09) || defined(COMPAT_43) 109 /* 110 * This table is used to maintain compatibility with 4.3BSD 111 * and NetBSD 0.9 mount syscalls. Note, the order is important! 112 * 113 * Do not modify this table. It should only contain filesystems 114 * supported by NetBSD 0.9 and 4.3BSD. 115 */ 116 const char * const mountcompatnames[] = { 117 NULL, /* 0 = MOUNT_NONE */ 118 MOUNT_FFS, /* 1 = MOUNT_UFS */ 119 MOUNT_NFS, /* 2 */ 120 MOUNT_MFS, /* 3 */ 121 MOUNT_MSDOS, /* 4 */ 122 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 123 MOUNT_FDESC, /* 6 */ 124 MOUNT_KERNFS, /* 7 */ 125 NULL, /* 8 = MOUNT_DEVFS */ 126 MOUNT_AFS, /* 9 */ 127 }; 128 const int nmountcompatnames = sizeof(mountcompatnames) / 129 sizeof(mountcompatnames[0]); 130 #endif /* COMPAT_09 || COMPAT_43 */ 131 132 static int 133 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 134 void *data, size_t *data_len) 135 { 136 struct mount *mp; 137 int error = 0, saved_flags; 138 139 mp = vp->v_mount; 140 saved_flags = mp->mnt_flag; 141 142 /* We can operate only on VV_ROOT nodes. */ 143 if ((vp->v_vflag & VV_ROOT) == 0) { 144 error = EINVAL; 145 goto out; 146 } 147 148 /* 149 * We only allow the filesystem to be reloaded if it 150 * is currently mounted read-only. 151 */ 152 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 153 error = EOPNOTSUPP; /* Needs translation */ 154 goto out; 155 } 156 157 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 158 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 159 if (error) 160 goto out; 161 162 if (vfs_busy(mp, LK_NOWAIT, 0)) { 163 error = EPERM; 164 goto out; 165 } 166 167 mp->mnt_flag &= ~MNT_OP_FLAGS; 168 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 169 170 /* 171 * Set the mount level flags. 172 */ 173 if (flags & MNT_RDONLY) 174 mp->mnt_flag |= MNT_RDONLY; 175 else if (mp->mnt_flag & MNT_RDONLY) 176 mp->mnt_iflag |= IMNT_WANTRDWR; 177 mp->mnt_flag &= 178 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 179 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 180 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 181 mp->mnt_flag |= flags & 182 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 183 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 184 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 185 MNT_IGNORE); 186 187 error = VFS_MOUNT(mp, path, data, data_len); 188 189 #if defined(COMPAT_30) && defined(NFSSERVER) 190 if (error && data != NULL) { 191 int error2; 192 193 /* Update failed; let's try and see if it was an 194 * export request. */ 195 error2 = nfs_update_exports_30(mp, path, data, l); 196 197 /* Only update error code if the export request was 198 * understood but some problem occurred while 199 * processing it. */ 200 if (error2 != EJUSTRETURN) 201 error = error2; 202 } 203 #endif 204 if (mp->mnt_iflag & IMNT_WANTRDWR) 205 mp->mnt_flag &= ~MNT_RDONLY; 206 if (error) 207 mp->mnt_flag = saved_flags; 208 mp->mnt_flag &= ~MNT_OP_FLAGS; 209 mp->mnt_iflag &= ~IMNT_WANTRDWR; 210 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 211 if (mp->mnt_syncer == NULL) 212 error = vfs_allocate_syncvnode(mp); 213 } else { 214 if (mp->mnt_syncer != NULL) 215 vfs_deallocate_syncvnode(mp); 216 } 217 vfs_unbusy(mp); 218 219 out: 220 return (error); 221 } 222 223 static int 224 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 225 { 226 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 227 int error; 228 229 /* Copy file-system type from userspace. */ 230 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 231 if (error) { 232 #if defined(COMPAT_09) || defined(COMPAT_43) 233 /* 234 * Historically, filesystem types were identified by numbers. 235 * If we get an integer for the filesystem type instead of a 236 * string, we check to see if it matches one of the historic 237 * filesystem types. 238 */ 239 u_long fsindex = (u_long)fstype; 240 if (fsindex >= nmountcompatnames || 241 mountcompatnames[fsindex] == NULL) 242 return ENODEV; 243 strlcpy(fstypename, mountcompatnames[fsindex], 244 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, u_int recurse) 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)) != 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 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 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); 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 vn_restorerecurse(vp, recurse); 349 VOP_UNLOCK(vp, 0); 350 checkdirs(vp); 351 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 352 error = vfs_allocate_syncvnode(mp); 353 vfs_unbusy(mp); 354 (void) VFS_STATVFS(mp, &mp->mnt_stat); 355 error = VFS_START(mp, 0); 356 if (error) 357 vrele(vp); 358 *vpp = NULL; 359 return error; 360 } 361 362 static int 363 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 364 void *data, size_t *data_len) 365 { 366 struct mount *mp; 367 int error; 368 369 /* If MNT_GETARGS is specified, it should be the only flag. */ 370 if (flags & ~MNT_GETARGS) 371 return EINVAL; 372 373 mp = vp->v_mount; 374 375 /* XXX: probably some notion of "can see" here if we want isolation. */ 376 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 377 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 378 if (error) 379 return error; 380 381 if ((vp->v_vflag & VV_ROOT) == 0) 382 return EINVAL; 383 384 if (vfs_busy(mp, LK_NOWAIT, 0)) 385 return EPERM; 386 387 mp->mnt_flag &= ~MNT_OP_FLAGS; 388 mp->mnt_flag |= MNT_GETARGS; 389 error = VFS_MOUNT(mp, path, data, data_len); 390 mp->mnt_flag &= ~MNT_OP_FLAGS; 391 392 vfs_unbusy(mp); 393 return (error); 394 } 395 396 #ifdef COMPAT_40 397 /* ARGSUSED */ 398 int 399 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 400 { 401 /* { 402 syscallarg(const char *) type; 403 syscallarg(const char *) path; 404 syscallarg(int) flags; 405 syscallarg(void *) data; 406 } */ 407 register_t dummy; 408 409 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 410 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 411 } 412 #endif 413 414 int 415 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 416 { 417 /* { 418 syscallarg(const char *) type; 419 syscallarg(const char *) path; 420 syscallarg(int) flags; 421 syscallarg(void *) data; 422 syscallarg(size_t) data_len; 423 } */ 424 425 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 426 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 427 SCARG(uap, data_len), retval); 428 } 429 430 int 431 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 432 const char *path, int flags, void *data, enum uio_seg data_seg, 433 size_t data_len, register_t *retval) 434 { 435 struct vnode *vp; 436 struct nameidata nd; 437 void *data_buf = data; 438 u_int recurse; 439 int error; 440 441 /* 442 * Get vnode to be covered 443 */ 444 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 445 if ((error = namei(&nd)) != 0) 446 return (error); 447 vp = nd.ni_vp; 448 449 /* 450 * A lookup in VFS_MOUNT might result in an attempt to 451 * lock this vnode again, so make the lock recursive. 452 */ 453 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 454 recurse = vn_setrecurse(vp); 455 456 if (vfsops == NULL) { 457 if (flags & (MNT_GETARGS | MNT_UPDATE)) 458 vfsops = vp->v_mount->mnt_op; 459 else { 460 /* 'type' is userspace */ 461 error = mount_get_vfsops(type, &vfsops); 462 if (error != 0) 463 goto done; 464 } 465 } 466 467 if (data != NULL && data_seg == UIO_USERSPACE) { 468 if (data_len == 0) { 469 /* No length supplied, use default for filesystem */ 470 data_len = vfsops->vfs_min_mount_data; 471 if (data_len > VFS_MAX_MOUNT_DATA) { 472 /* maybe a force loaded old LKM */ 473 error = EINVAL; 474 goto done; 475 } 476 #ifdef COMPAT_30 477 /* Hopefully a longer buffer won't make copyin() fail */ 478 if (flags & MNT_UPDATE 479 && data_len < sizeof (struct mnt_export_args30)) 480 data_len = sizeof (struct mnt_export_args30); 481 #endif 482 } 483 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 484 485 /* NFS needs the buffer even for mnt_getargs .... */ 486 error = copyin(data, data_buf, data_len); 487 if (error != 0) 488 goto done; 489 } 490 491 if (flags & MNT_GETARGS) { 492 if (data_len == 0) { 493 error = EINVAL; 494 goto done; 495 } 496 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 497 if (error != 0) 498 goto done; 499 if (data_seg == UIO_USERSPACE) 500 error = copyout(data_buf, data, data_len); 501 *retval = data_len; 502 } else if (flags & MNT_UPDATE) { 503 error = mount_update(l, vp, path, flags, data_buf, &data_len); 504 } else { 505 /* Locking is handled internally in mount_domount(). */ 506 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 507 &data_len, recurse); 508 } 509 510 done: 511 if (vp != NULL) { 512 vn_restorerecurse(vp, recurse); 513 vput(vp); 514 } 515 if (data_buf != data) 516 free(data_buf, M_TEMP); 517 return (error); 518 } 519 520 /* 521 * Scan all active processes to see if any of them have a current 522 * or root directory onto which the new filesystem has just been 523 * mounted. If so, replace them with the new mount point. 524 */ 525 void 526 checkdirs(struct vnode *olddp) 527 { 528 struct cwdinfo *cwdi; 529 struct vnode *newdp; 530 struct proc *p; 531 532 if (olddp->v_usecount == 1) 533 return; 534 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 535 panic("mount: lost mount"); 536 mutex_enter(&proclist_lock); 537 PROCLIST_FOREACH(p, &allproc) { 538 cwdi = p->p_cwdi; 539 if (!cwdi) 540 continue; 541 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 542 if (cwdi->cwdi_cdir == olddp) { 543 vrele(cwdi->cwdi_cdir); 544 VREF(newdp); 545 cwdi->cwdi_cdir = newdp; 546 } 547 if (cwdi->cwdi_rdir == olddp) { 548 vrele(cwdi->cwdi_rdir); 549 VREF(newdp); 550 cwdi->cwdi_rdir = newdp; 551 } 552 rw_exit(&cwdi->cwdi_lock); 553 } 554 mutex_exit(&proclist_lock); 555 if (rootvnode == olddp) { 556 vrele(rootvnode); 557 VREF(newdp); 558 rootvnode = newdp; 559 } 560 vput(newdp); 561 } 562 563 /* 564 * Unmount a file system. 565 * 566 * Note: unmount takes a path to the vnode mounted on as argument, 567 * not special file (as before). 568 */ 569 /* ARGSUSED */ 570 int 571 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 572 { 573 /* { 574 syscallarg(const char *) path; 575 syscallarg(int) flags; 576 } */ 577 struct vnode *vp; 578 struct mount *mp; 579 int error; 580 struct nameidata nd; 581 582 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 583 SCARG(uap, path)); 584 if ((error = namei(&nd)) != 0) 585 return (error); 586 vp = nd.ni_vp; 587 mp = vp->v_mount; 588 589 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 590 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 591 if (error) { 592 vput(vp); 593 return (error); 594 } 595 596 /* 597 * Don't allow unmounting the root file system. 598 */ 599 if (mp->mnt_flag & MNT_ROOTFS) { 600 vput(vp); 601 return (EINVAL); 602 } 603 604 /* 605 * Must be the root of the filesystem 606 */ 607 if ((vp->v_vflag & VV_ROOT) == 0) { 608 vput(vp); 609 return (EINVAL); 610 } 611 vput(vp); 612 613 /* 614 * XXX Freeze syncer. Must do this before locking the 615 * mount point. See dounmount() for details. 616 */ 617 mutex_enter(&syncer_mutex); 618 619 if (vfs_busy(mp, 0, 0)) { 620 mutex_exit(&syncer_mutex); 621 return (EBUSY); 622 } 623 624 return (dounmount(mp, SCARG(uap, flags), l)); 625 } 626 627 /* 628 * Do the actual file system unmount. File system is assumed to have been 629 * marked busy by the caller. 630 */ 631 int 632 dounmount(struct mount *mp, int flags, struct lwp *l) 633 { 634 struct vnode *coveredvp; 635 int error; 636 int async; 637 int used_syncer; 638 639 #if NVERIEXEC > 0 640 error = veriexec_unmountchk(mp); 641 if (error) 642 return (error); 643 #endif /* NVERIEXEC > 0 */ 644 645 mutex_enter(&mountlist_lock); 646 vfs_unbusy(mp); 647 used_syncer = (mp->mnt_syncer != NULL); 648 649 /* 650 * XXX Syncer must be frozen when we get here. This should really 651 * be done on a per-mountpoint basis, but especially the softdep 652 * code possibly called from the syncer doesn't exactly work on a 653 * per-mountpoint basis, so the softdep code would become a maze 654 * of vfs_busy() calls. 655 * 656 * The caller of dounmount() must acquire syncer_mutex because 657 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 658 * order, and we must preserve that order to avoid deadlock. 659 * 660 * So, if the file system did not use the syncer, now is 661 * the time to release the syncer_mutex. 662 */ 663 if (used_syncer == 0) 664 mutex_exit(&syncer_mutex); 665 666 mp->mnt_iflag |= IMNT_UNMOUNT; 667 mp->mnt_unmounter = l; 668 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 669 670 async = mp->mnt_flag & MNT_ASYNC; 671 mp->mnt_flag &= ~MNT_ASYNC; 672 cache_purgevfs(mp); /* remove cache entries for this file sys */ 673 if (mp->mnt_syncer != NULL) 674 vfs_deallocate_syncvnode(mp); 675 error = 0; 676 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 677 #if NFSS > 0 678 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 679 #endif 680 if (error == 0) 681 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 682 } 683 if (error == 0 || (flags & MNT_FORCE)) 684 error = VFS_UNMOUNT(mp, flags); 685 if (error) { 686 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 687 (void) vfs_allocate_syncvnode(mp); 688 mutex_enter(&mountlist_lock); 689 mp->mnt_iflag &= ~IMNT_UNMOUNT; 690 mp->mnt_unmounter = NULL; 691 mp->mnt_flag |= async; 692 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 693 &mountlist_lock); 694 if (used_syncer) 695 mutex_exit(&syncer_mutex); 696 mutex_enter(&mp->mnt_mutex); 697 while (mp->mnt_wcnt > 0) { 698 wakeup(mp); 699 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 700 0, &mp->mnt_mutex); 701 } 702 mutex_exit(&mp->mnt_mutex); 703 return (error); 704 } 705 vfs_scrubvnlist(mp); 706 mutex_enter(&mountlist_lock); 707 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 708 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 709 coveredvp->v_mountedhere = NULL; 710 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 711 panic("unmount: dangling vnode"); 712 mp->mnt_iflag |= IMNT_GONE; 713 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 714 if (coveredvp != NULLVP) 715 vrele(coveredvp); 716 if (used_syncer) 717 mutex_exit(&syncer_mutex); 718 mutex_enter(&mp->mnt_mutex); 719 while (mp->mnt_wcnt > 0) { 720 wakeup(mp); 721 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 722 } 723 mutex_exit(&mp->mnt_mutex); 724 vfs_hooks_unmount(mp); 725 vfs_delref(mp->mnt_op); 726 vfs_destroy(mp); 727 return (0); 728 } 729 730 /* 731 * Sync each mounted filesystem. 732 */ 733 #ifdef DEBUG 734 int syncprt = 0; 735 struct ctldebug debug0 = { "syncprt", &syncprt }; 736 #endif 737 738 /* ARGSUSED */ 739 int 740 sys_sync(struct lwp *l, const void *v, register_t *retval) 741 { 742 struct mount *mp, *nmp; 743 int asyncflag; 744 745 if (l == NULL) 746 l = &lwp0; 747 748 mutex_enter(&mountlist_lock); 749 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 750 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 751 nmp = mp->mnt_list.cqe_prev; 752 continue; 753 } 754 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 755 asyncflag = mp->mnt_flag & MNT_ASYNC; 756 mp->mnt_flag &= ~MNT_ASYNC; 757 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 758 if (asyncflag) 759 mp->mnt_flag |= MNT_ASYNC; 760 } 761 mutex_enter(&mountlist_lock); 762 nmp = mp->mnt_list.cqe_prev; 763 vfs_unbusy(mp); 764 765 } 766 mutex_exit(&mountlist_lock); 767 #ifdef DEBUG 768 if (syncprt) 769 vfs_bufstats(); 770 #endif /* DEBUG */ 771 return (0); 772 } 773 774 /* 775 * Change filesystem quotas. 776 */ 777 /* ARGSUSED */ 778 int 779 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 780 { 781 /* { 782 syscallarg(const char *) path; 783 syscallarg(int) cmd; 784 syscallarg(int) uid; 785 syscallarg(void *) arg; 786 } */ 787 struct mount *mp; 788 int error; 789 struct nameidata nd; 790 791 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 792 SCARG(uap, path)); 793 if ((error = namei(&nd)) != 0) 794 return (error); 795 mp = nd.ni_vp->v_mount; 796 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 797 SCARG(uap, arg)); 798 vrele(nd.ni_vp); 799 return (error); 800 } 801 802 int 803 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 804 int root) 805 { 806 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 807 int error = 0; 808 809 /* 810 * If MNT_NOWAIT or MNT_LAZY is specified, do not 811 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 812 * overrides MNT_NOWAIT. 813 */ 814 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 815 (flags != MNT_WAIT && flags != 0)) { 816 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 817 goto done; 818 } 819 820 /* Get the filesystem stats now */ 821 memset(sp, 0, sizeof(*sp)); 822 if ((error = VFS_STATVFS(mp, sp)) != 0) { 823 return error; 824 } 825 826 if (cwdi->cwdi_rdir == NULL) 827 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 828 done: 829 if (cwdi->cwdi_rdir != NULL) { 830 size_t len; 831 char *bp; 832 char *path = PNBUF_GET(); 833 834 bp = path + MAXPATHLEN; 835 *--bp = '\0'; 836 rw_enter(&cwdi->cwdi_lock, RW_READER); 837 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 838 MAXPATHLEN / 2, 0, l); 839 rw_exit(&cwdi->cwdi_lock); 840 if (error) { 841 PNBUF_PUT(path); 842 return error; 843 } 844 len = strlen(bp); 845 /* 846 * for mount points that are below our root, we can see 847 * them, so we fix up the pathname and return them. The 848 * rest we cannot see, so we don't allow viewing the 849 * data. 850 */ 851 if (strncmp(bp, sp->f_mntonname, len) == 0) { 852 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 853 sizeof(sp->f_mntonname)); 854 if (sp->f_mntonname[0] == '\0') 855 (void)strlcpy(sp->f_mntonname, "/", 856 sizeof(sp->f_mntonname)); 857 } else { 858 if (root) 859 (void)strlcpy(sp->f_mntonname, "/", 860 sizeof(sp->f_mntonname)); 861 else 862 error = EPERM; 863 } 864 PNBUF_PUT(path); 865 } 866 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 867 return error; 868 } 869 870 /* 871 * Get filesystem statistics by path. 872 */ 873 int 874 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 875 { 876 struct mount *mp; 877 int error; 878 struct nameidata nd; 879 880 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 881 if ((error = namei(&nd)) != 0) 882 return error; 883 mp = nd.ni_vp->v_mount; 884 error = dostatvfs(mp, sb, l, flags, 1); 885 vrele(nd.ni_vp); 886 return error; 887 } 888 889 /* ARGSUSED */ 890 int 891 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 892 { 893 /* { 894 syscallarg(const char *) path; 895 syscallarg(struct statvfs *) buf; 896 syscallarg(int) flags; 897 } */ 898 struct statvfs *sb; 899 int error; 900 901 sb = STATVFSBUF_GET(); 902 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 903 if (error == 0) 904 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 905 STATVFSBUF_PUT(sb); 906 return error; 907 } 908 909 /* 910 * Get filesystem statistics by fd. 911 */ 912 int 913 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 914 { 915 struct proc *p = l->l_proc; 916 struct file *fp; 917 struct mount *mp; 918 int error; 919 920 /* getvnode() will use the descriptor for us */ 921 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 922 return (error); 923 mp = ((struct vnode *)fp->f_data)->v_mount; 924 error = dostatvfs(mp, sb, l, flags, 1); 925 FILE_UNUSE(fp, l); 926 return error; 927 } 928 929 /* ARGSUSED */ 930 int 931 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 932 { 933 /* { 934 syscallarg(int) fd; 935 syscallarg(struct statvfs *) buf; 936 syscallarg(int) flags; 937 } */ 938 struct statvfs *sb; 939 int error; 940 941 sb = STATVFSBUF_GET(); 942 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 943 if (error == 0) 944 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 945 STATVFSBUF_PUT(sb); 946 return error; 947 } 948 949 950 /* 951 * Get statistics on all filesystems. 952 */ 953 int 954 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 955 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 956 register_t *retval) 957 { 958 int root = 0; 959 struct proc *p = l->l_proc; 960 struct mount *mp, *nmp; 961 struct statvfs *sb; 962 size_t count, maxcount; 963 int error = 0; 964 965 sb = STATVFSBUF_GET(); 966 maxcount = bufsize / entry_sz; 967 mutex_enter(&mountlist_lock); 968 count = 0; 969 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 970 mp = nmp) { 971 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 972 nmp = CIRCLEQ_NEXT(mp, mnt_list); 973 continue; 974 } 975 if (sfsp && count < maxcount) { 976 error = dostatvfs(mp, sb, l, flags, 0); 977 if (error) { 978 mutex_enter(&mountlist_lock); 979 nmp = CIRCLEQ_NEXT(mp, mnt_list); 980 vfs_unbusy(mp); 981 continue; 982 } 983 error = copyfn(sb, sfsp, entry_sz); 984 if (error) { 985 vfs_unbusy(mp); 986 goto out; 987 } 988 sfsp = (char *)sfsp + entry_sz; 989 root |= strcmp(sb->f_mntonname, "/") == 0; 990 } 991 count++; 992 mutex_enter(&mountlist_lock); 993 nmp = CIRCLEQ_NEXT(mp, mnt_list); 994 vfs_unbusy(mp); 995 } 996 997 mutex_exit(&mountlist_lock); 998 if (root == 0 && p->p_cwdi->cwdi_rdir) { 999 /* 1000 * fake a root entry 1001 */ 1002 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1003 sb, l, flags, 1); 1004 if (error != 0) 1005 goto out; 1006 if (sfsp) 1007 error = copyfn(sb, sfsp, entry_sz); 1008 count++; 1009 } 1010 if (sfsp && count > maxcount) 1011 *retval = maxcount; 1012 else 1013 *retval = count; 1014 out: 1015 STATVFSBUF_PUT(sb); 1016 return error; 1017 } 1018 1019 int 1020 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1021 { 1022 /* { 1023 syscallarg(struct statvfs *) buf; 1024 syscallarg(size_t) bufsize; 1025 syscallarg(int) flags; 1026 } */ 1027 1028 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1029 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1030 } 1031 1032 /* 1033 * Change current working directory to a given file descriptor. 1034 */ 1035 /* ARGSUSED */ 1036 int 1037 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1038 { 1039 /* { 1040 syscallarg(int) fd; 1041 } */ 1042 struct proc *p = l->l_proc; 1043 struct filedesc *fdp = p->p_fd; 1044 struct cwdinfo *cwdi; 1045 struct vnode *vp, *tdp; 1046 struct mount *mp; 1047 struct file *fp; 1048 int error; 1049 1050 /* getvnode() will use the descriptor for us */ 1051 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1052 return (error); 1053 vp = (struct vnode *)fp->f_data; 1054 1055 VREF(vp); 1056 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1057 if (vp->v_type != VDIR) 1058 error = ENOTDIR; 1059 else 1060 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1061 if (error) { 1062 vput(vp); 1063 goto out; 1064 } 1065 while ((mp = vp->v_mountedhere) != NULL) { 1066 if (vfs_busy(mp, 0, 0)) 1067 continue; 1068 1069 vput(vp); 1070 error = VFS_ROOT(mp, &tdp); 1071 vfs_unbusy(mp); 1072 if (error) 1073 goto out; 1074 vp = tdp; 1075 } 1076 VOP_UNLOCK(vp, 0); 1077 1078 /* 1079 * Disallow changing to a directory not under the process's 1080 * current root directory (if there is one). 1081 */ 1082 cwdi = p->p_cwdi; 1083 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1084 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1085 vrele(vp); 1086 error = EPERM; /* operation not permitted */ 1087 } else { 1088 vrele(cwdi->cwdi_cdir); 1089 cwdi->cwdi_cdir = vp; 1090 } 1091 rw_exit(&cwdi->cwdi_lock); 1092 1093 out: 1094 FILE_UNUSE(fp, l); 1095 return (error); 1096 } 1097 1098 /* 1099 * Change this process's notion of the root directory to a given file 1100 * descriptor. 1101 */ 1102 int 1103 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1104 { 1105 struct proc *p = l->l_proc; 1106 struct filedesc *fdp = p->p_fd; 1107 struct cwdinfo *cwdi; 1108 struct vnode *vp; 1109 struct file *fp; 1110 int error; 1111 1112 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1113 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1114 return error; 1115 /* getvnode() will use the descriptor for us */ 1116 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1117 return error; 1118 vp = (struct vnode *) fp->f_data; 1119 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1120 if (vp->v_type != VDIR) 1121 error = ENOTDIR; 1122 else 1123 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1124 VOP_UNLOCK(vp, 0); 1125 if (error) 1126 goto out; 1127 VREF(vp); 1128 1129 /* 1130 * Prevent escaping from chroot by putting the root under 1131 * the working directory. Silently chdir to / if we aren't 1132 * already there. 1133 */ 1134 cwdi = p->p_cwdi; 1135 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1136 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1137 /* 1138 * XXX would be more failsafe to change directory to a 1139 * deadfs node here instead 1140 */ 1141 vrele(cwdi->cwdi_cdir); 1142 VREF(vp); 1143 cwdi->cwdi_cdir = vp; 1144 } 1145 1146 if (cwdi->cwdi_rdir != NULL) 1147 vrele(cwdi->cwdi_rdir); 1148 cwdi->cwdi_rdir = vp; 1149 rw_exit(&cwdi->cwdi_lock); 1150 1151 out: 1152 FILE_UNUSE(fp, l); 1153 return (error); 1154 } 1155 1156 /* 1157 * Change current working directory (``.''). 1158 */ 1159 /* ARGSUSED */ 1160 int 1161 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1162 { 1163 /* { 1164 syscallarg(const char *) path; 1165 } */ 1166 struct proc *p = l->l_proc; 1167 struct cwdinfo *cwdi; 1168 int error; 1169 struct nameidata nd; 1170 1171 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1172 SCARG(uap, path)); 1173 if ((error = change_dir(&nd, l)) != 0) 1174 return (error); 1175 cwdi = p->p_cwdi; 1176 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1177 vrele(cwdi->cwdi_cdir); 1178 cwdi->cwdi_cdir = nd.ni_vp; 1179 rw_exit(&cwdi->cwdi_lock); 1180 return (0); 1181 } 1182 1183 /* 1184 * Change notion of root (``/'') directory. 1185 */ 1186 /* ARGSUSED */ 1187 int 1188 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1189 { 1190 /* { 1191 syscallarg(const char *) path; 1192 } */ 1193 struct proc *p = l->l_proc; 1194 struct cwdinfo *cwdi; 1195 struct vnode *vp; 1196 int error; 1197 struct nameidata nd; 1198 1199 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1200 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1201 return (error); 1202 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1203 SCARG(uap, path)); 1204 if ((error = change_dir(&nd, l)) != 0) 1205 return (error); 1206 1207 cwdi = p->p_cwdi; 1208 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1209 if (cwdi->cwdi_rdir != NULL) 1210 vrele(cwdi->cwdi_rdir); 1211 vp = nd.ni_vp; 1212 cwdi->cwdi_rdir = vp; 1213 1214 /* 1215 * Prevent escaping from chroot by putting the root under 1216 * the working directory. Silently chdir to / if we aren't 1217 * already there. 1218 */ 1219 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1220 /* 1221 * XXX would be more failsafe to change directory to a 1222 * deadfs node here instead 1223 */ 1224 vrele(cwdi->cwdi_cdir); 1225 VREF(vp); 1226 cwdi->cwdi_cdir = vp; 1227 } 1228 rw_exit(&cwdi->cwdi_lock); 1229 1230 return (0); 1231 } 1232 1233 /* 1234 * Common routine for chroot and chdir. 1235 */ 1236 static int 1237 change_dir(struct nameidata *ndp, struct lwp *l) 1238 { 1239 struct vnode *vp; 1240 int error; 1241 1242 if ((error = namei(ndp)) != 0) 1243 return (error); 1244 vp = ndp->ni_vp; 1245 if (vp->v_type != VDIR) 1246 error = ENOTDIR; 1247 else 1248 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1249 1250 if (error) 1251 vput(vp); 1252 else 1253 VOP_UNLOCK(vp, 0); 1254 return (error); 1255 } 1256 1257 /* 1258 * Check permissions, allocate an open file structure, 1259 * and call the device open routine if any. 1260 */ 1261 int 1262 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1263 { 1264 /* { 1265 syscallarg(const char *) path; 1266 syscallarg(int) flags; 1267 syscallarg(int) mode; 1268 } */ 1269 struct proc *p = l->l_proc; 1270 struct cwdinfo *cwdi = p->p_cwdi; 1271 struct filedesc *fdp = p->p_fd; 1272 struct file *fp; 1273 struct vnode *vp; 1274 int flags, cmode; 1275 int type, indx, error; 1276 struct flock lf; 1277 struct nameidata nd; 1278 1279 flags = FFLAGS(SCARG(uap, flags)); 1280 if ((flags & (FREAD | FWRITE)) == 0) 1281 return (EINVAL); 1282 /* falloc() will use the file descriptor for us */ 1283 if ((error = falloc(l, &fp, &indx)) != 0) 1284 return (error); 1285 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1286 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1287 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1288 SCARG(uap, path)); 1289 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1290 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1291 rw_enter(&fdp->fd_lock, RW_WRITER); 1292 FILE_UNUSE(fp, l); 1293 fdp->fd_ofiles[indx] = NULL; 1294 rw_exit(&fdp->fd_lock); 1295 ffree(fp); 1296 if ((error == EDUPFD || error == EMOVEFD) && 1297 l->l_dupfd >= 0 && /* XXX from fdopen */ 1298 (error = 1299 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1300 *retval = indx; 1301 return (0); 1302 } 1303 if (error == ERESTART) 1304 error = EINTR; 1305 fdremove(fdp, indx); 1306 return (error); 1307 } 1308 1309 l->l_dupfd = 0; 1310 vp = nd.ni_vp; 1311 fp->f_flag = flags & FMASK; 1312 fp->f_type = DTYPE_VNODE; 1313 fp->f_ops = &vnops; 1314 fp->f_data = vp; 1315 if (flags & (O_EXLOCK | O_SHLOCK)) { 1316 lf.l_whence = SEEK_SET; 1317 lf.l_start = 0; 1318 lf.l_len = 0; 1319 if (flags & O_EXLOCK) 1320 lf.l_type = F_WRLCK; 1321 else 1322 lf.l_type = F_RDLCK; 1323 type = F_FLOCK; 1324 if ((flags & FNONBLOCK) == 0) 1325 type |= F_WAIT; 1326 VOP_UNLOCK(vp, 0); 1327 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1328 if (error) { 1329 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1330 FILE_UNUSE(fp, l); 1331 ffree(fp); 1332 fdremove(fdp, indx); 1333 return (error); 1334 } 1335 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1336 fp->f_flag |= FHASLOCK; 1337 } 1338 VOP_UNLOCK(vp, 0); 1339 *retval = indx; 1340 FILE_SET_MATURE(fp); 1341 FILE_UNUSE(fp, l); 1342 return (0); 1343 } 1344 1345 static void 1346 vfs__fhfree(fhandle_t *fhp) 1347 { 1348 size_t fhsize; 1349 1350 if (fhp == NULL) { 1351 return; 1352 } 1353 fhsize = FHANDLE_SIZE(fhp); 1354 kmem_free(fhp, fhsize); 1355 } 1356 1357 /* 1358 * vfs_composefh: compose a filehandle. 1359 */ 1360 1361 int 1362 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1363 { 1364 struct mount *mp; 1365 struct fid *fidp; 1366 int error; 1367 size_t needfhsize; 1368 size_t fidsize; 1369 1370 mp = vp->v_mount; 1371 fidp = NULL; 1372 if (*fh_size < FHANDLE_SIZE_MIN) { 1373 fidsize = 0; 1374 } else { 1375 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1376 if (fhp != NULL) { 1377 memset(fhp, 0, *fh_size); 1378 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1379 fidp = &fhp->fh_fid; 1380 } 1381 } 1382 error = VFS_VPTOFH(vp, fidp, &fidsize); 1383 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1384 if (error == 0 && *fh_size < needfhsize) { 1385 error = E2BIG; 1386 } 1387 *fh_size = needfhsize; 1388 return error; 1389 } 1390 1391 int 1392 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1393 { 1394 struct mount *mp; 1395 fhandle_t *fhp; 1396 size_t fhsize; 1397 size_t fidsize; 1398 int error; 1399 1400 *fhpp = NULL; 1401 mp = vp->v_mount; 1402 fidsize = 0; 1403 error = VFS_VPTOFH(vp, NULL, &fidsize); 1404 KASSERT(error != 0); 1405 if (error != E2BIG) { 1406 goto out; 1407 } 1408 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1409 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1410 if (fhp == NULL) { 1411 error = ENOMEM; 1412 goto out; 1413 } 1414 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1415 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1416 if (error == 0) { 1417 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1418 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1419 *fhpp = fhp; 1420 } else { 1421 kmem_free(fhp, fhsize); 1422 } 1423 out: 1424 return error; 1425 } 1426 1427 void 1428 vfs_composefh_free(fhandle_t *fhp) 1429 { 1430 1431 vfs__fhfree(fhp); 1432 } 1433 1434 /* 1435 * vfs_fhtovp: lookup a vnode by a filehandle. 1436 */ 1437 1438 int 1439 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1440 { 1441 struct mount *mp; 1442 int error; 1443 1444 *vpp = NULL; 1445 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1446 if (mp == NULL) { 1447 error = ESTALE; 1448 goto out; 1449 } 1450 if (mp->mnt_op->vfs_fhtovp == NULL) { 1451 error = EOPNOTSUPP; 1452 goto out; 1453 } 1454 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1455 out: 1456 return error; 1457 } 1458 1459 /* 1460 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1461 * the needed size. 1462 */ 1463 1464 int 1465 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1466 { 1467 fhandle_t *fhp; 1468 int error; 1469 1470 *fhpp = NULL; 1471 if (fhsize > FHANDLE_SIZE_MAX) { 1472 return EINVAL; 1473 } 1474 if (fhsize < FHANDLE_SIZE_MIN) { 1475 return EINVAL; 1476 } 1477 again: 1478 fhp = kmem_alloc(fhsize, KM_SLEEP); 1479 if (fhp == NULL) { 1480 return ENOMEM; 1481 } 1482 error = copyin(ufhp, fhp, fhsize); 1483 if (error == 0) { 1484 /* XXX this check shouldn't be here */ 1485 if (FHANDLE_SIZE(fhp) == fhsize) { 1486 *fhpp = fhp; 1487 return 0; 1488 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1489 /* 1490 * a kludge for nfsv2 padded handles. 1491 */ 1492 size_t sz; 1493 1494 sz = FHANDLE_SIZE(fhp); 1495 kmem_free(fhp, fhsize); 1496 fhsize = sz; 1497 goto again; 1498 } else { 1499 /* 1500 * userland told us wrong size. 1501 */ 1502 error = EINVAL; 1503 } 1504 } 1505 kmem_free(fhp, fhsize); 1506 return error; 1507 } 1508 1509 void 1510 vfs_copyinfh_free(fhandle_t *fhp) 1511 { 1512 1513 vfs__fhfree(fhp); 1514 } 1515 1516 /* 1517 * Get file handle system call 1518 */ 1519 int 1520 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1521 { 1522 /* { 1523 syscallarg(char *) fname; 1524 syscallarg(fhandle_t *) fhp; 1525 syscallarg(size_t *) fh_size; 1526 } */ 1527 struct vnode *vp; 1528 fhandle_t *fh; 1529 int error; 1530 struct nameidata nd; 1531 size_t sz; 1532 size_t usz; 1533 1534 /* 1535 * Must be super user 1536 */ 1537 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1538 0, NULL, NULL, NULL); 1539 if (error) 1540 return (error); 1541 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1542 SCARG(uap, fname)); 1543 error = namei(&nd); 1544 if (error) 1545 return (error); 1546 vp = nd.ni_vp; 1547 error = vfs_composefh_alloc(vp, &fh); 1548 vput(vp); 1549 if (error != 0) { 1550 goto out; 1551 } 1552 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1553 if (error != 0) { 1554 goto out; 1555 } 1556 sz = FHANDLE_SIZE(fh); 1557 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1558 if (error != 0) { 1559 goto out; 1560 } 1561 if (usz >= sz) { 1562 error = copyout(fh, SCARG(uap, fhp), sz); 1563 } else { 1564 error = E2BIG; 1565 } 1566 out: 1567 vfs_composefh_free(fh); 1568 return (error); 1569 } 1570 1571 /* 1572 * Open a file given a file handle. 1573 * 1574 * Check permissions, allocate an open file structure, 1575 * and call the device open routine if any. 1576 */ 1577 1578 int 1579 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1580 register_t *retval) 1581 { 1582 struct filedesc *fdp = l->l_proc->p_fd; 1583 struct file *fp; 1584 struct vnode *vp = NULL; 1585 kauth_cred_t cred = l->l_cred; 1586 struct file *nfp; 1587 int type, indx, error=0; 1588 struct flock lf; 1589 struct vattr va; 1590 fhandle_t *fh; 1591 int flags; 1592 1593 /* 1594 * Must be super user 1595 */ 1596 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1597 0, NULL, NULL, NULL))) 1598 return (error); 1599 1600 flags = FFLAGS(oflags); 1601 if ((flags & (FREAD | FWRITE)) == 0) 1602 return (EINVAL); 1603 if ((flags & O_CREAT)) 1604 return (EINVAL); 1605 /* falloc() will use the file descriptor for us */ 1606 if ((error = falloc(l, &nfp, &indx)) != 0) 1607 return (error); 1608 fp = nfp; 1609 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1610 if (error != 0) { 1611 goto bad; 1612 } 1613 error = vfs_fhtovp(fh, &vp); 1614 if (error != 0) { 1615 goto bad; 1616 } 1617 1618 /* Now do an effective vn_open */ 1619 1620 if (vp->v_type == VSOCK) { 1621 error = EOPNOTSUPP; 1622 goto bad; 1623 } 1624 error = vn_openchk(vp, cred, flags); 1625 if (error != 0) 1626 goto bad; 1627 if (flags & O_TRUNC) { 1628 VOP_UNLOCK(vp, 0); /* XXX */ 1629 VOP_LEASE(vp, cred, LEASE_WRITE); 1630 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1631 VATTR_NULL(&va); 1632 va.va_size = 0; 1633 error = VOP_SETATTR(vp, &va, cred); 1634 if (error) 1635 goto bad; 1636 } 1637 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1638 goto bad; 1639 if (flags & FWRITE) 1640 vp->v_writecount++; 1641 1642 /* done with modified vn_open, now finish what sys_open does. */ 1643 1644 fp->f_flag = flags & FMASK; 1645 fp->f_type = DTYPE_VNODE; 1646 fp->f_ops = &vnops; 1647 fp->f_data = vp; 1648 if (flags & (O_EXLOCK | O_SHLOCK)) { 1649 lf.l_whence = SEEK_SET; 1650 lf.l_start = 0; 1651 lf.l_len = 0; 1652 if (flags & O_EXLOCK) 1653 lf.l_type = F_WRLCK; 1654 else 1655 lf.l_type = F_RDLCK; 1656 type = F_FLOCK; 1657 if ((flags & FNONBLOCK) == 0) 1658 type |= F_WAIT; 1659 VOP_UNLOCK(vp, 0); 1660 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1661 if (error) { 1662 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1663 FILE_UNUSE(fp, l); 1664 ffree(fp); 1665 fdremove(fdp, indx); 1666 return (error); 1667 } 1668 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1669 fp->f_flag |= FHASLOCK; 1670 } 1671 VOP_UNLOCK(vp, 0); 1672 *retval = indx; 1673 FILE_SET_MATURE(fp); 1674 FILE_UNUSE(fp, l); 1675 vfs_copyinfh_free(fh); 1676 return (0); 1677 1678 bad: 1679 FILE_UNUSE(fp, l); 1680 ffree(fp); 1681 fdremove(fdp, indx); 1682 if (vp != NULL) 1683 vput(vp); 1684 vfs_copyinfh_free(fh); 1685 return (error); 1686 } 1687 1688 int 1689 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1690 { 1691 /* { 1692 syscallarg(const void *) fhp; 1693 syscallarg(size_t) fh_size; 1694 syscallarg(int) flags; 1695 } */ 1696 1697 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1698 SCARG(uap, flags), retval); 1699 } 1700 1701 int 1702 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1703 { 1704 int error; 1705 fhandle_t *fh; 1706 struct vnode *vp; 1707 1708 /* 1709 * Must be super user 1710 */ 1711 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1712 0, NULL, NULL, NULL))) 1713 return (error); 1714 1715 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1716 if (error != 0) 1717 return error; 1718 1719 error = vfs_fhtovp(fh, &vp); 1720 vfs_copyinfh_free(fh); 1721 if (error != 0) 1722 return error; 1723 1724 error = vn_stat(vp, sb, l); 1725 vput(vp); 1726 return error; 1727 } 1728 1729 1730 /* ARGSUSED */ 1731 int 1732 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1733 { 1734 /* { 1735 syscallarg(const void *) fhp; 1736 syscallarg(size_t) fh_size; 1737 syscallarg(struct stat *) sb; 1738 } */ 1739 struct stat sb; 1740 int error; 1741 1742 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1743 if (error) 1744 return error; 1745 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1746 } 1747 1748 int 1749 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1750 int flags) 1751 { 1752 fhandle_t *fh; 1753 struct mount *mp; 1754 struct vnode *vp; 1755 int error; 1756 1757 /* 1758 * Must be super user 1759 */ 1760 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1761 0, NULL, NULL, NULL))) 1762 return error; 1763 1764 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1765 if (error != 0) 1766 return error; 1767 1768 error = vfs_fhtovp(fh, &vp); 1769 vfs_copyinfh_free(fh); 1770 if (error != 0) 1771 return error; 1772 1773 mp = vp->v_mount; 1774 error = dostatvfs(mp, sb, l, flags, 1); 1775 vput(vp); 1776 return error; 1777 } 1778 1779 /* ARGSUSED */ 1780 int 1781 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1782 { 1783 /* { 1784 syscallarg(const void *) fhp; 1785 syscallarg(size_t) fh_size; 1786 syscallarg(struct statvfs *) buf; 1787 syscallarg(int) flags; 1788 } */ 1789 struct statvfs *sb = STATVFSBUF_GET(); 1790 int error; 1791 1792 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1793 SCARG(uap, flags)); 1794 if (error == 0) 1795 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1796 STATVFSBUF_PUT(sb); 1797 return error; 1798 } 1799 1800 /* 1801 * Create a special file. 1802 */ 1803 /* ARGSUSED */ 1804 int 1805 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1806 { 1807 /* { 1808 syscallarg(const char *) path; 1809 syscallarg(int) mode; 1810 syscallarg(int) dev; 1811 } */ 1812 struct proc *p = l->l_proc; 1813 struct vnode *vp; 1814 struct vattr vattr; 1815 int error, optype; 1816 struct nameidata nd; 1817 char *path; 1818 const char *cpath; 1819 enum uio_seg seg = UIO_USERSPACE; 1820 1821 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1822 0, NULL, NULL, NULL)) != 0) 1823 return (error); 1824 1825 optype = VOP_MKNOD_DESCOFFSET; 1826 1827 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1828 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1829 1830 if ((error = namei(&nd)) != 0) 1831 goto out; 1832 vp = nd.ni_vp; 1833 if (vp != NULL) 1834 error = EEXIST; 1835 else { 1836 VATTR_NULL(&vattr); 1837 /* We will read cwdi->cwdi_cmask unlocked. */ 1838 vattr.va_mode = 1839 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1840 vattr.va_rdev = SCARG(uap, dev); 1841 1842 switch (SCARG(uap, mode) & S_IFMT) { 1843 case S_IFMT: /* used by badsect to flag bad sectors */ 1844 vattr.va_type = VBAD; 1845 break; 1846 case S_IFCHR: 1847 vattr.va_type = VCHR; 1848 break; 1849 case S_IFBLK: 1850 vattr.va_type = VBLK; 1851 break; 1852 case S_IFWHT: 1853 optype = VOP_WHITEOUT_DESCOFFSET; 1854 break; 1855 case S_IFREG: 1856 #if NVERIEXEC > 0 1857 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1858 O_CREAT); 1859 #endif /* NVERIEXEC > 0 */ 1860 vattr.va_type = VREG; 1861 vattr.va_rdev = VNOVAL; 1862 optype = VOP_CREATE_DESCOFFSET; 1863 break; 1864 default: 1865 error = EINVAL; 1866 break; 1867 } 1868 } 1869 if (!error) { 1870 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1871 switch (optype) { 1872 case VOP_WHITEOUT_DESCOFFSET: 1873 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1874 if (error) 1875 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1876 vput(nd.ni_dvp); 1877 break; 1878 1879 case VOP_MKNOD_DESCOFFSET: 1880 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1881 &nd.ni_cnd, &vattr); 1882 if (error == 0) 1883 vput(nd.ni_vp); 1884 break; 1885 1886 case VOP_CREATE_DESCOFFSET: 1887 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1888 &nd.ni_cnd, &vattr); 1889 if (error == 0) 1890 vput(nd.ni_vp); 1891 break; 1892 } 1893 } else { 1894 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1895 if (nd.ni_dvp == vp) 1896 vrele(nd.ni_dvp); 1897 else 1898 vput(nd.ni_dvp); 1899 if (vp) 1900 vrele(vp); 1901 } 1902 out: 1903 VERIEXEC_PATH_PUT(path); 1904 return (error); 1905 } 1906 1907 /* 1908 * Create a named pipe. 1909 */ 1910 /* ARGSUSED */ 1911 int 1912 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1913 { 1914 /* { 1915 syscallarg(const char *) path; 1916 syscallarg(int) mode; 1917 } */ 1918 struct proc *p = l->l_proc; 1919 struct vattr vattr; 1920 int error; 1921 struct nameidata nd; 1922 1923 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1924 SCARG(uap, path)); 1925 if ((error = namei(&nd)) != 0) 1926 return (error); 1927 if (nd.ni_vp != NULL) { 1928 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1929 if (nd.ni_dvp == nd.ni_vp) 1930 vrele(nd.ni_dvp); 1931 else 1932 vput(nd.ni_dvp); 1933 vrele(nd.ni_vp); 1934 return (EEXIST); 1935 } 1936 VATTR_NULL(&vattr); 1937 vattr.va_type = VFIFO; 1938 /* We will read cwdi->cwdi_cmask unlocked. */ 1939 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1940 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1941 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1942 if (error == 0) 1943 vput(nd.ni_vp); 1944 return (error); 1945 } 1946 1947 /* 1948 * Make a hard file link. 1949 */ 1950 /* ARGSUSED */ 1951 int 1952 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1953 { 1954 /* { 1955 syscallarg(const char *) path; 1956 syscallarg(const char *) link; 1957 } */ 1958 struct vnode *vp; 1959 struct nameidata nd; 1960 int error; 1961 1962 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1963 SCARG(uap, path)); 1964 if ((error = namei(&nd)) != 0) 1965 return (error); 1966 vp = nd.ni_vp; 1967 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1968 SCARG(uap, link)); 1969 if ((error = namei(&nd)) != 0) 1970 goto out; 1971 if (nd.ni_vp) { 1972 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1973 if (nd.ni_dvp == nd.ni_vp) 1974 vrele(nd.ni_dvp); 1975 else 1976 vput(nd.ni_dvp); 1977 vrele(nd.ni_vp); 1978 error = EEXIST; 1979 goto out; 1980 } 1981 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1982 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 1983 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1984 out: 1985 vrele(vp); 1986 return (error); 1987 } 1988 1989 /* 1990 * Make a symbolic link. 1991 */ 1992 /* ARGSUSED */ 1993 int 1994 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1995 { 1996 /* { 1997 syscallarg(const char *) path; 1998 syscallarg(const char *) link; 1999 } */ 2000 struct proc *p = l->l_proc; 2001 struct vattr vattr; 2002 char *path; 2003 int error; 2004 struct nameidata nd; 2005 2006 path = PNBUF_GET(); 2007 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2008 if (error) 2009 goto out; 2010 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2011 SCARG(uap, link)); 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_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, const struct sys_undelete_args *uap, register_t *retval) 2043 { 2044 /* { 2045 syscallarg(const char *) path; 2046 } */ 2047 int error; 2048 struct nameidata nd; 2049 2050 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2051 UIO_USERSPACE, SCARG(uap, path)); 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_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, const struct sys_unlink_args *uap, register_t *retval) 2079 { 2080 /* { 2081 syscallarg(const char *) path; 2082 } */ 2083 2084 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2085 } 2086 2087 int 2088 do_sys_unlink(const char *arg, enum uio_seg seg) 2089 { 2090 struct vnode *vp; 2091 int error; 2092 struct nameidata nd; 2093 kauth_cred_t cred; 2094 char *path; 2095 const char *cpath; 2096 2097 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2098 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2099 2100 if ((error = namei(&nd)) != 0) 2101 goto out; 2102 vp = nd.ni_vp; 2103 2104 /* 2105 * The root of a mounted filesystem cannot be deleted. 2106 */ 2107 if (vp->v_vflag & VV_ROOT) { 2108 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2109 if (nd.ni_dvp == vp) 2110 vrele(nd.ni_dvp); 2111 else 2112 vput(nd.ni_dvp); 2113 vput(vp); 2114 error = EBUSY; 2115 goto out; 2116 } 2117 2118 #if NVERIEXEC > 0 2119 /* Handle remove requests for veriexec entries. */ 2120 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2121 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2122 if (nd.ni_dvp == vp) 2123 vrele(nd.ni_dvp); 2124 else 2125 vput(nd.ni_dvp); 2126 vput(vp); 2127 goto out; 2128 } 2129 #endif /* NVERIEXEC > 0 */ 2130 2131 cred = kauth_cred_get(); 2132 VOP_LEASE(nd.ni_dvp, cred, LEASE_WRITE); 2133 VOP_LEASE(vp, cred, LEASE_WRITE); 2134 #ifdef FILEASSOC 2135 (void)fileassoc_file_delete(vp); 2136 #endif /* FILEASSOC */ 2137 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2138 out: 2139 VERIEXEC_PATH_PUT(path); 2140 return (error); 2141 } 2142 2143 /* 2144 * Reposition read/write file offset. 2145 */ 2146 int 2147 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2148 { 2149 /* { 2150 syscallarg(int) fd; 2151 syscallarg(int) pad; 2152 syscallarg(off_t) offset; 2153 syscallarg(int) whence; 2154 } */ 2155 struct proc *p = l->l_proc; 2156 kauth_cred_t cred = l->l_cred; 2157 struct filedesc *fdp = p->p_fd; 2158 struct file *fp; 2159 struct vnode *vp; 2160 struct vattr vattr; 2161 off_t newoff; 2162 int error; 2163 2164 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2165 return (EBADF); 2166 2167 vp = (struct vnode *)fp->f_data; 2168 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2169 error = ESPIPE; 2170 FILE_UNLOCK(fp); 2171 goto out; 2172 } 2173 2174 switch (SCARG(uap, whence)) { 2175 case SEEK_CUR: 2176 newoff = fp->f_offset + SCARG(uap, offset); 2177 FILE_USE(fp); 2178 break; 2179 case SEEK_END: 2180 FILE_USE(fp); 2181 error = VOP_GETATTR(vp, &vattr, cred); 2182 if (error) { 2183 FILE_UNUSE(fp, l); 2184 goto out; 2185 } 2186 newoff = SCARG(uap, offset) + vattr.va_size; 2187 break; 2188 case SEEK_SET: 2189 FILE_USE(fp); 2190 newoff = SCARG(uap, offset); 2191 break; 2192 default: 2193 FILE_UNLOCK(fp); 2194 error = EINVAL; 2195 goto out; 2196 } 2197 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2198 FILE_LOCK(fp); 2199 *(off_t *)retval = fp->f_offset = newoff; 2200 FILE_UNLOCK(fp); 2201 } 2202 FILE_UNUSE(fp, l); 2203 out: 2204 return (error); 2205 } 2206 2207 /* 2208 * Positional read system call. 2209 */ 2210 int 2211 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2212 { 2213 /* { 2214 syscallarg(int) fd; 2215 syscallarg(void *) buf; 2216 syscallarg(size_t) nbyte; 2217 syscallarg(off_t) offset; 2218 } */ 2219 struct proc *p = l->l_proc; 2220 struct filedesc *fdp = p->p_fd; 2221 struct file *fp; 2222 struct vnode *vp; 2223 off_t offset; 2224 int error, fd = SCARG(uap, fd); 2225 2226 if ((fp = fd_getfile(fdp, fd)) == NULL) 2227 return (EBADF); 2228 2229 if ((fp->f_flag & FREAD) == 0) { 2230 FILE_UNLOCK(fp); 2231 return (EBADF); 2232 } 2233 2234 FILE_USE(fp); 2235 2236 vp = (struct vnode *)fp->f_data; 2237 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2238 error = ESPIPE; 2239 goto out; 2240 } 2241 2242 offset = SCARG(uap, offset); 2243 2244 /* 2245 * XXX This works because no file systems actually 2246 * XXX take any action on the seek operation. 2247 */ 2248 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2249 goto out; 2250 2251 /* dofileread() will unuse the descriptor for us */ 2252 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2253 &offset, 0, retval)); 2254 2255 out: 2256 FILE_UNUSE(fp, l); 2257 return (error); 2258 } 2259 2260 /* 2261 * Positional scatter read system call. 2262 */ 2263 int 2264 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2265 { 2266 /* { 2267 syscallarg(int) fd; 2268 syscallarg(const struct iovec *) iovp; 2269 syscallarg(int) iovcnt; 2270 syscallarg(off_t) offset; 2271 } */ 2272 off_t offset = SCARG(uap, offset); 2273 2274 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2275 SCARG(uap, iovcnt), &offset, 0, retval); 2276 } 2277 2278 /* 2279 * Positional write system call. 2280 */ 2281 int 2282 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2283 { 2284 /* { 2285 syscallarg(int) fd; 2286 syscallarg(const void *) buf; 2287 syscallarg(size_t) nbyte; 2288 syscallarg(off_t) offset; 2289 } */ 2290 struct proc *p = l->l_proc; 2291 struct filedesc *fdp = p->p_fd; 2292 struct file *fp; 2293 struct vnode *vp; 2294 off_t offset; 2295 int error, fd = SCARG(uap, fd); 2296 2297 if ((fp = fd_getfile(fdp, fd)) == NULL) 2298 return (EBADF); 2299 2300 if ((fp->f_flag & FWRITE) == 0) { 2301 FILE_UNLOCK(fp); 2302 return (EBADF); 2303 } 2304 2305 FILE_USE(fp); 2306 2307 vp = (struct vnode *)fp->f_data; 2308 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2309 error = ESPIPE; 2310 goto out; 2311 } 2312 2313 offset = SCARG(uap, offset); 2314 2315 /* 2316 * XXX This works because no file systems actually 2317 * XXX take any action on the seek operation. 2318 */ 2319 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2320 goto out; 2321 2322 /* dofilewrite() will unuse the descriptor for us */ 2323 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2324 &offset, 0, retval)); 2325 2326 out: 2327 FILE_UNUSE(fp, l); 2328 return (error); 2329 } 2330 2331 /* 2332 * Positional gather write system call. 2333 */ 2334 int 2335 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2336 { 2337 /* { 2338 syscallarg(int) fd; 2339 syscallarg(const struct iovec *) iovp; 2340 syscallarg(int) iovcnt; 2341 syscallarg(off_t) offset; 2342 } */ 2343 off_t offset = SCARG(uap, offset); 2344 2345 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2346 SCARG(uap, iovcnt), &offset, 0, retval); 2347 } 2348 2349 /* 2350 * Check access permissions. 2351 */ 2352 int 2353 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2354 { 2355 /* { 2356 syscallarg(const char *) path; 2357 syscallarg(int) flags; 2358 } */ 2359 kauth_cred_t cred; 2360 struct vnode *vp; 2361 int error, flags; 2362 struct nameidata nd; 2363 2364 cred = kauth_cred_dup(l->l_cred); 2365 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2366 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2367 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2368 SCARG(uap, path)); 2369 /* Override default credentials */ 2370 nd.ni_cnd.cn_cred = cred; 2371 if ((error = namei(&nd)) != 0) 2372 goto out; 2373 vp = nd.ni_vp; 2374 2375 /* Flags == 0 means only check for existence. */ 2376 if (SCARG(uap, flags)) { 2377 flags = 0; 2378 if (SCARG(uap, flags) & R_OK) 2379 flags |= VREAD; 2380 if (SCARG(uap, flags) & W_OK) 2381 flags |= VWRITE; 2382 if (SCARG(uap, flags) & X_OK) 2383 flags |= VEXEC; 2384 2385 error = VOP_ACCESS(vp, flags, cred); 2386 if (!error && (flags & VWRITE)) 2387 error = vn_writechk(vp); 2388 } 2389 vput(vp); 2390 out: 2391 kauth_cred_free(cred); 2392 return (error); 2393 } 2394 2395 /* 2396 * Common code for all sys_stat functions, including compat versions. 2397 */ 2398 int 2399 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2400 struct stat *sb) 2401 { 2402 int error; 2403 struct nameidata nd; 2404 2405 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2406 UIO_USERSPACE, path); 2407 error = namei(&nd); 2408 if (error != 0) 2409 return error; 2410 error = vn_stat(nd.ni_vp, sb, l); 2411 vput(nd.ni_vp); 2412 return error; 2413 } 2414 2415 /* 2416 * Get file status; this version follows links. 2417 */ 2418 /* ARGSUSED */ 2419 int 2420 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2421 { 2422 /* { 2423 syscallarg(const char *) path; 2424 syscallarg(struct stat *) ub; 2425 } */ 2426 struct stat sb; 2427 int error; 2428 2429 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2430 if (error) 2431 return error; 2432 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2433 } 2434 2435 /* 2436 * Get file status; this version does not follow links. 2437 */ 2438 /* ARGSUSED */ 2439 int 2440 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2441 { 2442 /* { 2443 syscallarg(const char *) path; 2444 syscallarg(struct stat *) ub; 2445 } */ 2446 struct stat sb; 2447 int error; 2448 2449 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2450 if (error) 2451 return error; 2452 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2453 } 2454 2455 /* 2456 * Get configurable pathname variables. 2457 */ 2458 /* ARGSUSED */ 2459 int 2460 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2461 { 2462 /* { 2463 syscallarg(const char *) path; 2464 syscallarg(int) name; 2465 } */ 2466 int error; 2467 struct nameidata nd; 2468 2469 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2470 SCARG(uap, path)); 2471 if ((error = namei(&nd)) != 0) 2472 return (error); 2473 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2474 vput(nd.ni_vp); 2475 return (error); 2476 } 2477 2478 /* 2479 * Return target name of a symbolic link. 2480 */ 2481 /* ARGSUSED */ 2482 int 2483 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2484 { 2485 /* { 2486 syscallarg(const char *) path; 2487 syscallarg(char *) buf; 2488 syscallarg(size_t) count; 2489 } */ 2490 struct vnode *vp; 2491 struct iovec aiov; 2492 struct uio auio; 2493 int error; 2494 struct nameidata nd; 2495 2496 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2497 SCARG(uap, path)); 2498 if ((error = namei(&nd)) != 0) 2499 return (error); 2500 vp = nd.ni_vp; 2501 if (vp->v_type != VLNK) 2502 error = EINVAL; 2503 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2504 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2505 aiov.iov_base = SCARG(uap, buf); 2506 aiov.iov_len = SCARG(uap, count); 2507 auio.uio_iov = &aiov; 2508 auio.uio_iovcnt = 1; 2509 auio.uio_offset = 0; 2510 auio.uio_rw = UIO_READ; 2511 KASSERT(l == curlwp); 2512 auio.uio_vmspace = l->l_proc->p_vmspace; 2513 auio.uio_resid = SCARG(uap, count); 2514 error = VOP_READLINK(vp, &auio, l->l_cred); 2515 } 2516 vput(vp); 2517 *retval = SCARG(uap, count) - auio.uio_resid; 2518 return (error); 2519 } 2520 2521 /* 2522 * Change flags of a file given a path name. 2523 */ 2524 /* ARGSUSED */ 2525 int 2526 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2527 { 2528 /* { 2529 syscallarg(const char *) path; 2530 syscallarg(u_long) flags; 2531 } */ 2532 struct vnode *vp; 2533 int error; 2534 struct nameidata nd; 2535 2536 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2537 SCARG(uap, path)); 2538 if ((error = namei(&nd)) != 0) 2539 return (error); 2540 vp = nd.ni_vp; 2541 error = change_flags(vp, SCARG(uap, flags), l); 2542 vput(vp); 2543 return (error); 2544 } 2545 2546 /* 2547 * Change flags of a file given a file descriptor. 2548 */ 2549 /* ARGSUSED */ 2550 int 2551 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2552 { 2553 /* { 2554 syscallarg(int) fd; 2555 syscallarg(u_long) flags; 2556 } */ 2557 struct proc *p = l->l_proc; 2558 struct vnode *vp; 2559 struct file *fp; 2560 int error; 2561 2562 /* getvnode() will use the descriptor for us */ 2563 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2564 return (error); 2565 vp = (struct vnode *)fp->f_data; 2566 error = change_flags(vp, SCARG(uap, flags), l); 2567 VOP_UNLOCK(vp, 0); 2568 FILE_UNUSE(fp, l); 2569 return (error); 2570 } 2571 2572 /* 2573 * Change flags of a file given a path name; this version does 2574 * not follow links. 2575 */ 2576 int 2577 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2578 { 2579 /* { 2580 syscallarg(const char *) path; 2581 syscallarg(u_long) flags; 2582 } */ 2583 struct vnode *vp; 2584 int error; 2585 struct nameidata nd; 2586 2587 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2588 SCARG(uap, path)); 2589 if ((error = namei(&nd)) != 0) 2590 return (error); 2591 vp = nd.ni_vp; 2592 error = change_flags(vp, SCARG(uap, flags), l); 2593 vput(vp); 2594 return (error); 2595 } 2596 2597 /* 2598 * Common routine to change flags of a file. 2599 */ 2600 int 2601 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2602 { 2603 struct vattr vattr; 2604 int error; 2605 2606 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2607 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2608 /* 2609 * Non-superusers cannot change the flags on devices, even if they 2610 * own them. 2611 */ 2612 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2613 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2614 goto out; 2615 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2616 error = EINVAL; 2617 goto out; 2618 } 2619 } 2620 VATTR_NULL(&vattr); 2621 vattr.va_flags = flags; 2622 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2623 out: 2624 return (error); 2625 } 2626 2627 /* 2628 * Change mode of a file given path name; this version follows links. 2629 */ 2630 /* ARGSUSED */ 2631 int 2632 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2633 { 2634 /* { 2635 syscallarg(const char *) path; 2636 syscallarg(int) mode; 2637 } */ 2638 int error; 2639 struct nameidata nd; 2640 2641 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2642 SCARG(uap, path)); 2643 if ((error = namei(&nd)) != 0) 2644 return (error); 2645 2646 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2647 2648 vrele(nd.ni_vp); 2649 return (error); 2650 } 2651 2652 /* 2653 * Change mode of a file given a file descriptor. 2654 */ 2655 /* ARGSUSED */ 2656 int 2657 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2658 { 2659 /* { 2660 syscallarg(int) fd; 2661 syscallarg(int) mode; 2662 } */ 2663 struct proc *p = l->l_proc; 2664 struct file *fp; 2665 int error; 2666 2667 /* getvnode() will use the descriptor for us */ 2668 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2669 return (error); 2670 2671 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2672 FILE_UNUSE(fp, l); 2673 return (error); 2674 } 2675 2676 /* 2677 * Change mode of a file given path name; this version does not follow links. 2678 */ 2679 /* ARGSUSED */ 2680 int 2681 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2682 { 2683 /* { 2684 syscallarg(const char *) path; 2685 syscallarg(int) mode; 2686 } */ 2687 int error; 2688 struct nameidata nd; 2689 2690 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2691 SCARG(uap, path)); 2692 if ((error = namei(&nd)) != 0) 2693 return (error); 2694 2695 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2696 2697 vrele(nd.ni_vp); 2698 return (error); 2699 } 2700 2701 /* 2702 * Common routine to set mode given a vnode. 2703 */ 2704 static int 2705 change_mode(struct vnode *vp, int mode, struct lwp *l) 2706 { 2707 struct vattr vattr; 2708 int error; 2709 2710 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2711 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2712 VATTR_NULL(&vattr); 2713 vattr.va_mode = mode & ALLPERMS; 2714 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2715 VOP_UNLOCK(vp, 0); 2716 return (error); 2717 } 2718 2719 /* 2720 * Set ownership given a path name; this version follows links. 2721 */ 2722 /* ARGSUSED */ 2723 int 2724 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2725 { 2726 /* { 2727 syscallarg(const char *) path; 2728 syscallarg(uid_t) uid; 2729 syscallarg(gid_t) gid; 2730 } */ 2731 int error; 2732 struct nameidata nd; 2733 2734 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2735 SCARG(uap, path)); 2736 if ((error = namei(&nd)) != 0) 2737 return (error); 2738 2739 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2740 2741 vrele(nd.ni_vp); 2742 return (error); 2743 } 2744 2745 /* 2746 * Set ownership given a path name; this version follows links. 2747 * Provides POSIX semantics. 2748 */ 2749 /* ARGSUSED */ 2750 int 2751 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2752 { 2753 /* { 2754 syscallarg(const char *) path; 2755 syscallarg(uid_t) uid; 2756 syscallarg(gid_t) gid; 2757 } */ 2758 int error; 2759 struct nameidata nd; 2760 2761 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2762 SCARG(uap, path)); 2763 if ((error = namei(&nd)) != 0) 2764 return (error); 2765 2766 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2767 2768 vrele(nd.ni_vp); 2769 return (error); 2770 } 2771 2772 /* 2773 * Set ownership given a file descriptor. 2774 */ 2775 /* ARGSUSED */ 2776 int 2777 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2778 { 2779 /* { 2780 syscallarg(int) fd; 2781 syscallarg(uid_t) uid; 2782 syscallarg(gid_t) gid; 2783 } */ 2784 struct proc *p = l->l_proc; 2785 int error; 2786 struct file *fp; 2787 2788 /* getvnode() will use the descriptor for us */ 2789 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2790 return (error); 2791 2792 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2793 SCARG(uap, gid), l, 0); 2794 FILE_UNUSE(fp, l); 2795 return (error); 2796 } 2797 2798 /* 2799 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2800 */ 2801 /* ARGSUSED */ 2802 int 2803 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2804 { 2805 /* { 2806 syscallarg(int) fd; 2807 syscallarg(uid_t) uid; 2808 syscallarg(gid_t) gid; 2809 } */ 2810 struct proc *p = l->l_proc; 2811 int error; 2812 struct file *fp; 2813 2814 /* getvnode() will use the descriptor for us */ 2815 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2816 return (error); 2817 2818 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2819 SCARG(uap, gid), l, 1); 2820 FILE_UNUSE(fp, l); 2821 return (error); 2822 } 2823 2824 /* 2825 * Set ownership given a path name; this version does not follow links. 2826 */ 2827 /* ARGSUSED */ 2828 int 2829 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2830 { 2831 /* { 2832 syscallarg(const char *) path; 2833 syscallarg(uid_t) uid; 2834 syscallarg(gid_t) gid; 2835 } */ 2836 int error; 2837 struct nameidata nd; 2838 2839 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2840 SCARG(uap, path)); 2841 if ((error = namei(&nd)) != 0) 2842 return (error); 2843 2844 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2845 2846 vrele(nd.ni_vp); 2847 return (error); 2848 } 2849 2850 /* 2851 * Set ownership given a path name; this version does not follow links. 2852 * Provides POSIX/XPG semantics. 2853 */ 2854 /* ARGSUSED */ 2855 int 2856 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2857 { 2858 /* { 2859 syscallarg(const char *) path; 2860 syscallarg(uid_t) uid; 2861 syscallarg(gid_t) gid; 2862 } */ 2863 int error; 2864 struct nameidata nd; 2865 2866 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2867 SCARG(uap, path)); 2868 if ((error = namei(&nd)) != 0) 2869 return (error); 2870 2871 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2872 2873 vrele(nd.ni_vp); 2874 return (error); 2875 } 2876 2877 /* 2878 * Common routine to set ownership given a vnode. 2879 */ 2880 static int 2881 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2882 int posix_semantics) 2883 { 2884 struct vattr vattr; 2885 mode_t newmode; 2886 int error; 2887 2888 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2889 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2890 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2891 goto out; 2892 2893 #define CHANGED(x) ((int)(x) != -1) 2894 newmode = vattr.va_mode; 2895 if (posix_semantics) { 2896 /* 2897 * POSIX/XPG semantics: if the caller is not the super-user, 2898 * clear set-user-id and set-group-id bits. Both POSIX and 2899 * the XPG consider the behaviour for calls by the super-user 2900 * implementation-defined; we leave the set-user-id and set- 2901 * group-id settings intact in that case. 2902 */ 2903 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2904 NULL) != 0) 2905 newmode &= ~(S_ISUID | S_ISGID); 2906 } else { 2907 /* 2908 * NetBSD semantics: when changing owner and/or group, 2909 * clear the respective bit(s). 2910 */ 2911 if (CHANGED(uid)) 2912 newmode &= ~S_ISUID; 2913 if (CHANGED(gid)) 2914 newmode &= ~S_ISGID; 2915 } 2916 /* Update va_mode iff altered. */ 2917 if (vattr.va_mode == newmode) 2918 newmode = VNOVAL; 2919 2920 VATTR_NULL(&vattr); 2921 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2922 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2923 vattr.va_mode = newmode; 2924 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2925 #undef CHANGED 2926 2927 out: 2928 VOP_UNLOCK(vp, 0); 2929 return (error); 2930 } 2931 2932 /* 2933 * Set the access and modification times given a path name; this 2934 * version follows links. 2935 */ 2936 /* ARGSUSED */ 2937 int 2938 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2939 { 2940 /* { 2941 syscallarg(const char *) path; 2942 syscallarg(const struct timeval *) tptr; 2943 } */ 2944 2945 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2946 SCARG(uap, tptr), UIO_USERSPACE); 2947 } 2948 2949 /* 2950 * Set the access and modification times given a file descriptor. 2951 */ 2952 /* ARGSUSED */ 2953 int 2954 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2955 { 2956 /* { 2957 syscallarg(int) fd; 2958 syscallarg(const struct timeval *) tptr; 2959 } */ 2960 int error; 2961 struct file *fp; 2962 2963 /* getvnode() will use the descriptor for us */ 2964 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2965 return (error); 2966 2967 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2968 SCARG(uap, tptr), UIO_USERSPACE); 2969 2970 FILE_UNUSE(fp, l); 2971 return (error); 2972 } 2973 2974 /* 2975 * Set the access and modification times given a path name; this 2976 * version does not follow links. 2977 */ 2978 int 2979 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2980 { 2981 /* { 2982 syscallarg(const char *) path; 2983 syscallarg(const struct timeval *) tptr; 2984 } */ 2985 2986 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2987 SCARG(uap, tptr), UIO_USERSPACE); 2988 } 2989 2990 /* 2991 * Common routine to set access and modification times given a vnode. 2992 */ 2993 int 2994 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2995 const struct timeval *tptr, enum uio_seg seg) 2996 { 2997 struct vattr vattr; 2998 struct nameidata nd; 2999 int error; 3000 3001 VATTR_NULL(&vattr); 3002 if (tptr == NULL) { 3003 nanotime(&vattr.va_atime); 3004 vattr.va_mtime = vattr.va_atime; 3005 vattr.va_vaflags |= VA_UTIMES_NULL; 3006 } else { 3007 struct timeval tv[2]; 3008 3009 if (seg != UIO_SYSSPACE) { 3010 error = copyin(tptr, &tv, sizeof (tv)); 3011 if (error != 0) 3012 return error; 3013 tptr = tv; 3014 } 3015 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3016 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3017 } 3018 3019 if (vp == NULL) { 3020 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3021 if ((error = namei(&nd)) != 0) 3022 return (error); 3023 vp = nd.ni_vp; 3024 } else 3025 nd.ni_vp = NULL; 3026 3027 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3028 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3029 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3030 VOP_UNLOCK(vp, 0); 3031 3032 if (nd.ni_vp != NULL) 3033 vrele(nd.ni_vp); 3034 3035 return (error); 3036 } 3037 3038 /* 3039 * Truncate a file given its path name. 3040 */ 3041 /* ARGSUSED */ 3042 int 3043 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3044 { 3045 /* { 3046 syscallarg(const char *) path; 3047 syscallarg(int) pad; 3048 syscallarg(off_t) length; 3049 } */ 3050 struct vnode *vp; 3051 struct vattr vattr; 3052 int error; 3053 struct nameidata nd; 3054 3055 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3056 SCARG(uap, path)); 3057 if ((error = namei(&nd)) != 0) 3058 return (error); 3059 vp = nd.ni_vp; 3060 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3061 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3062 if (vp->v_type == VDIR) 3063 error = EISDIR; 3064 else if ((error = vn_writechk(vp)) == 0 && 3065 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3066 VATTR_NULL(&vattr); 3067 vattr.va_size = SCARG(uap, length); 3068 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3069 } 3070 vput(vp); 3071 return (error); 3072 } 3073 3074 /* 3075 * Truncate a file given a file descriptor. 3076 */ 3077 /* ARGSUSED */ 3078 int 3079 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3080 { 3081 /* { 3082 syscallarg(int) fd; 3083 syscallarg(int) pad; 3084 syscallarg(off_t) length; 3085 } */ 3086 struct proc *p = l->l_proc; 3087 struct vattr vattr; 3088 struct vnode *vp; 3089 struct file *fp; 3090 int error; 3091 3092 /* getvnode() will use the descriptor for us */ 3093 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3094 return (error); 3095 if ((fp->f_flag & FWRITE) == 0) { 3096 error = EINVAL; 3097 goto out; 3098 } 3099 vp = (struct vnode *)fp->f_data; 3100 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3101 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3102 if (vp->v_type == VDIR) 3103 error = EISDIR; 3104 else if ((error = vn_writechk(vp)) == 0) { 3105 VATTR_NULL(&vattr); 3106 vattr.va_size = SCARG(uap, length); 3107 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3108 } 3109 VOP_UNLOCK(vp, 0); 3110 out: 3111 FILE_UNUSE(fp, l); 3112 return (error); 3113 } 3114 3115 /* 3116 * Sync an open file. 3117 */ 3118 /* ARGSUSED */ 3119 int 3120 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3121 { 3122 /* { 3123 syscallarg(int) fd; 3124 } */ 3125 struct proc *p = l->l_proc; 3126 struct vnode *vp; 3127 struct file *fp; 3128 int error; 3129 3130 /* getvnode() will use the descriptor for us */ 3131 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3132 return (error); 3133 vp = (struct vnode *)fp->f_data; 3134 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3135 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3136 if (error == 0 && bioopsp != NULL && 3137 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3138 (*bioopsp->io_fsync)(vp, 0); 3139 VOP_UNLOCK(vp, 0); 3140 FILE_UNUSE(fp, l); 3141 return (error); 3142 } 3143 3144 /* 3145 * Sync a range of file data. API modeled after that found in AIX. 3146 * 3147 * FDATASYNC indicates that we need only save enough metadata to be able 3148 * to re-read the written data. Note we duplicate AIX's requirement that 3149 * the file be open for writing. 3150 */ 3151 /* ARGSUSED */ 3152 int 3153 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3154 { 3155 /* { 3156 syscallarg(int) fd; 3157 syscallarg(int) flags; 3158 syscallarg(off_t) start; 3159 syscallarg(off_t) length; 3160 } */ 3161 struct proc *p = l->l_proc; 3162 struct vnode *vp; 3163 struct file *fp; 3164 int flags, nflags; 3165 off_t s, e, len; 3166 int error; 3167 3168 /* getvnode() will use the descriptor for us */ 3169 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3170 return (error); 3171 3172 if ((fp->f_flag & FWRITE) == 0) { 3173 error = EBADF; 3174 goto out; 3175 } 3176 3177 flags = SCARG(uap, flags); 3178 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3179 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3180 error = EINVAL; 3181 goto out; 3182 } 3183 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3184 if (flags & FDATASYNC) 3185 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3186 else 3187 nflags = FSYNC_WAIT; 3188 if (flags & FDISKSYNC) 3189 nflags |= FSYNC_CACHE; 3190 3191 len = SCARG(uap, length); 3192 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3193 if (len) { 3194 s = SCARG(uap, start); 3195 e = s + len; 3196 if (e < s) { 3197 error = EINVAL; 3198 goto out; 3199 } 3200 } else { 3201 e = 0; 3202 s = 0; 3203 } 3204 3205 vp = (struct vnode *)fp->f_data; 3206 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3207 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3208 3209 if (error == 0 && bioopsp != NULL && 3210 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3211 (*bioopsp->io_fsync)(vp, nflags); 3212 3213 VOP_UNLOCK(vp, 0); 3214 out: 3215 FILE_UNUSE(fp, l); 3216 return (error); 3217 } 3218 3219 /* 3220 * Sync the data of an open file. 3221 */ 3222 /* ARGSUSED */ 3223 int 3224 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3225 { 3226 /* { 3227 syscallarg(int) fd; 3228 } */ 3229 struct proc *p = l->l_proc; 3230 struct vnode *vp; 3231 struct file *fp; 3232 int error; 3233 3234 /* getvnode() will use the descriptor for us */ 3235 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3236 return (error); 3237 if ((fp->f_flag & FWRITE) == 0) { 3238 FILE_UNUSE(fp, l); 3239 return (EBADF); 3240 } 3241 vp = (struct vnode *)fp->f_data; 3242 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3243 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3244 VOP_UNLOCK(vp, 0); 3245 FILE_UNUSE(fp, l); 3246 return (error); 3247 } 3248 3249 /* 3250 * Rename files, (standard) BSD semantics frontend. 3251 */ 3252 /* ARGSUSED */ 3253 int 3254 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3255 { 3256 /* { 3257 syscallarg(const char *) from; 3258 syscallarg(const char *) to; 3259 } */ 3260 3261 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3262 } 3263 3264 /* 3265 * Rename files, POSIX semantics frontend. 3266 */ 3267 /* ARGSUSED */ 3268 int 3269 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3270 { 3271 /* { 3272 syscallarg(const char *) from; 3273 syscallarg(const char *) to; 3274 } */ 3275 3276 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3277 } 3278 3279 /* 3280 * Rename files. Source and destination must either both be directories, 3281 * or both not be directories. If target is a directory, it must be empty. 3282 * If `from' and `to' refer to the same object, the value of the `retain' 3283 * argument is used to determine whether `from' will be 3284 * 3285 * (retain == 0) deleted unless `from' and `to' refer to the same 3286 * object in the file system's name space (BSD). 3287 * (retain == 1) always retained (POSIX). 3288 */ 3289 int 3290 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3291 { 3292 struct vnode *tvp, *fvp, *tdvp; 3293 struct nameidata fromnd, tond; 3294 struct lwp *l = curlwp; 3295 struct proc *p; 3296 int error; 3297 3298 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3299 seg, from); 3300 if ((error = namei(&fromnd)) != 0) 3301 return (error); 3302 if (fromnd.ni_dvp != fromnd.ni_vp) 3303 VOP_UNLOCK(fromnd.ni_dvp, 0); 3304 fvp = fromnd.ni_vp; 3305 NDINIT(&tond, RENAME, 3306 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3307 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3308 seg, to); 3309 if ((error = namei(&tond)) != 0) { 3310 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3311 vrele(fromnd.ni_dvp); 3312 vrele(fvp); 3313 goto out1; 3314 } 3315 tdvp = tond.ni_dvp; 3316 tvp = tond.ni_vp; 3317 3318 if (tvp != NULL) { 3319 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3320 error = ENOTDIR; 3321 goto out; 3322 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3323 error = EISDIR; 3324 goto out; 3325 } 3326 } 3327 3328 if (fvp == tdvp) 3329 error = EINVAL; 3330 3331 /* 3332 * Source and destination refer to the same object. 3333 */ 3334 if (fvp == tvp) { 3335 if (retain) 3336 error = -1; 3337 else if (fromnd.ni_dvp == tdvp && 3338 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3339 !memcmp(fromnd.ni_cnd.cn_nameptr, 3340 tond.ni_cnd.cn_nameptr, 3341 fromnd.ni_cnd.cn_namelen)) 3342 error = -1; 3343 } 3344 3345 #if NVERIEXEC > 0 3346 if (!error) { 3347 char *f1, *f2; 3348 3349 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3350 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3351 3352 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3353 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3354 3355 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3356 3357 free(f1, M_TEMP); 3358 free(f2, M_TEMP); 3359 } 3360 #endif /* NVERIEXEC > 0 */ 3361 3362 out: 3363 p = l->l_proc; 3364 if (!error) { 3365 VOP_LEASE(tdvp, l->l_cred, LEASE_WRITE); 3366 if (fromnd.ni_dvp != tdvp) 3367 VOP_LEASE(fromnd.ni_dvp, l->l_cred, LEASE_WRITE); 3368 if (tvp) { 3369 VOP_LEASE(tvp, l->l_cred, LEASE_WRITE); 3370 } 3371 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3372 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3373 } else { 3374 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3375 if (tdvp == tvp) 3376 vrele(tdvp); 3377 else 3378 vput(tdvp); 3379 if (tvp) 3380 vput(tvp); 3381 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3382 vrele(fromnd.ni_dvp); 3383 vrele(fvp); 3384 } 3385 vrele(tond.ni_startdir); 3386 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3387 out1: 3388 if (fromnd.ni_startdir) 3389 vrele(fromnd.ni_startdir); 3390 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3391 return (error == -1 ? 0 : error); 3392 } 3393 3394 /* 3395 * Make a directory file. 3396 */ 3397 /* ARGSUSED */ 3398 int 3399 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3400 { 3401 /* { 3402 syscallarg(const char *) path; 3403 syscallarg(int) mode; 3404 } */ 3405 struct proc *p = l->l_proc; 3406 struct vnode *vp; 3407 struct vattr vattr; 3408 int error; 3409 struct nameidata nd; 3410 3411 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3412 SCARG(uap, path)); 3413 if ((error = namei(&nd)) != 0) 3414 return (error); 3415 vp = nd.ni_vp; 3416 if (vp != NULL) { 3417 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3418 if (nd.ni_dvp == vp) 3419 vrele(nd.ni_dvp); 3420 else 3421 vput(nd.ni_dvp); 3422 vrele(vp); 3423 return (EEXIST); 3424 } 3425 VATTR_NULL(&vattr); 3426 vattr.va_type = VDIR; 3427 /* We will read cwdi->cwdi_cmask unlocked. */ 3428 vattr.va_mode = 3429 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3430 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3431 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3432 if (!error) 3433 vput(nd.ni_vp); 3434 return (error); 3435 } 3436 3437 /* 3438 * Remove a directory file. 3439 */ 3440 /* ARGSUSED */ 3441 int 3442 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3443 { 3444 /* { 3445 syscallarg(const char *) path; 3446 } */ 3447 struct vnode *vp; 3448 int error; 3449 struct nameidata nd; 3450 3451 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3452 SCARG(uap, path)); 3453 if ((error = namei(&nd)) != 0) 3454 return (error); 3455 vp = nd.ni_vp; 3456 if (vp->v_type != VDIR) { 3457 error = ENOTDIR; 3458 goto out; 3459 } 3460 /* 3461 * No rmdir "." please. 3462 */ 3463 if (nd.ni_dvp == vp) { 3464 error = EINVAL; 3465 goto out; 3466 } 3467 /* 3468 * The root of a mounted filesystem cannot be deleted. 3469 */ 3470 if (vp->v_vflag & VV_ROOT) { 3471 error = EBUSY; 3472 goto out; 3473 } 3474 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3475 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3476 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3477 return (error); 3478 3479 out: 3480 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3481 if (nd.ni_dvp == vp) 3482 vrele(nd.ni_dvp); 3483 else 3484 vput(nd.ni_dvp); 3485 vput(vp); 3486 return (error); 3487 } 3488 3489 /* 3490 * Read a block of directory entries in a file system independent format. 3491 */ 3492 int 3493 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3494 { 3495 /* { 3496 syscallarg(int) fd; 3497 syscallarg(char *) buf; 3498 syscallarg(size_t) count; 3499 } */ 3500 struct proc *p = l->l_proc; 3501 struct file *fp; 3502 int error, done; 3503 3504 /* getvnode() will use the descriptor for us */ 3505 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3506 return (error); 3507 if ((fp->f_flag & FREAD) == 0) { 3508 error = EBADF; 3509 goto out; 3510 } 3511 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3512 SCARG(uap, count), &done, l, 0, 0); 3513 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3514 *retval = done; 3515 out: 3516 FILE_UNUSE(fp, l); 3517 return (error); 3518 } 3519 3520 /* 3521 * Set the mode mask for creation of filesystem nodes. 3522 */ 3523 int 3524 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3525 { 3526 /* { 3527 syscallarg(mode_t) newmask; 3528 } */ 3529 struct proc *p = l->l_proc; 3530 struct cwdinfo *cwdi; 3531 3532 /* 3533 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3534 * important is that we serialize changes to the mask. The 3535 * rw_exit() will issue a write memory barrier on our behalf, 3536 * and force the changes out to other CPUs (as it must use an 3537 * atomic operation, draining the local CPU's store buffers). 3538 */ 3539 cwdi = p->p_cwdi; 3540 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3541 *retval = cwdi->cwdi_cmask; 3542 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3543 rw_exit(&cwdi->cwdi_lock); 3544 3545 return (0); 3546 } 3547 3548 int 3549 dorevoke(struct vnode *vp, kauth_cred_t cred) 3550 { 3551 struct vattr vattr; 3552 int error; 3553 bool revoke; 3554 3555 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3556 goto out; 3557 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3558 (error = kauth_authorize_generic(cred, 3559 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3560 goto out; 3561 mutex_enter(&vp->v_interlock); 3562 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED|VI_LAYER))); 3563 mutex_exit(&vp->v_interlock); 3564 if (revoke) 3565 VOP_REVOKE(vp, REVOKEALL); 3566 3567 out: 3568 return (error); 3569 } 3570 3571 /* 3572 * Void all references to file by ripping underlying filesystem 3573 * away from vnode. 3574 */ 3575 /* ARGSUSED */ 3576 int 3577 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3578 { 3579 /* { 3580 syscallarg(const char *) path; 3581 } */ 3582 struct vnode *vp; 3583 int error; 3584 struct nameidata nd; 3585 3586 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3587 SCARG(uap, path)); 3588 if ((error = namei(&nd)) != 0) 3589 return (error); 3590 vp = nd.ni_vp; 3591 error = dorevoke(vp, l->l_cred); 3592 vrele(vp); 3593 return (error); 3594 } 3595 3596 /* 3597 * Convert a user file descriptor to a kernel file entry. 3598 */ 3599 int 3600 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3601 { 3602 struct vnode *vp; 3603 struct file *fp; 3604 3605 if ((fp = fd_getfile(fdp, fd)) == NULL) 3606 return (EBADF); 3607 3608 FILE_USE(fp); 3609 3610 if (fp->f_type != DTYPE_VNODE) { 3611 FILE_UNUSE(fp, NULL); 3612 return (EINVAL); 3613 } 3614 3615 vp = (struct vnode *)fp->f_data; 3616 if (vp->v_type == VBAD) { 3617 FILE_UNUSE(fp, NULL); 3618 return (EBADF); 3619 } 3620 3621 *fpp = fp; 3622 return (0); 3623 } 3624
Indexes created Fri Oct 03 02:09:57 GMT 2025