vfs_syscalls.c revision 1.343
1 /* $NetBSD: vfs_syscalls.c,v 1.343 2008/01/25 14:32:15 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.343 2008/01/25 14:32:15 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "fss.h" 46 #include "veriexec.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/kmem.h> 61 #include <sys/dirent.h> 62 #include <sys/sysctl.h> 63 #include <sys/syscallargs.h> 64 #include <sys/vfs_syscalls.h> 65 #include <sys/ktrace.h> 66 #ifdef FILEASSOC 67 #include <sys/fileassoc.h> 68 #endif /* FILEASSOC */ 69 #include <sys/verified_exec.h> 70 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 #include <miscfs/specfs/specdev.h> 75 76 #ifdef COMPAT_30 77 #include "opt_nfsserver.h" 78 #include <nfs/rpcv2.h> 79 #endif 80 #include <nfs/nfsproto.h> 81 #ifdef COMPAT_30 82 #include <nfs/nfs.h> 83 #include <nfs/nfs_var.h> 84 #endif 85 86 #if NFSS > 0 87 #include <dev/fssvar.h> 88 #endif 89 90 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 91 92 static int change_dir(struct nameidata *, struct lwp *); 93 static int change_flags(struct vnode *, u_long, struct lwp *); 94 static int change_mode(struct vnode *, int, struct lwp *l); 95 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 96 97 void checkdirs(struct vnode *); 98 99 int dovfsusermount = 0; 100 101 /* 102 * Virtual File System System Calls 103 */ 104 105 /* 106 * Mount a file system. 107 */ 108 109 #if defined(COMPAT_09) || defined(COMPAT_43) 110 /* 111 * This table is used to maintain compatibility with 4.3BSD 112 * and NetBSD 0.9 mount syscalls. Note, the order is important! 113 * 114 * Do not modify this table. It should only contain filesystems 115 * supported by NetBSD 0.9 and 4.3BSD. 116 */ 117 const char * const mountcompatnames[] = { 118 NULL, /* 0 = MOUNT_NONE */ 119 MOUNT_FFS, /* 1 = MOUNT_UFS */ 120 MOUNT_NFS, /* 2 */ 121 MOUNT_MFS, /* 3 */ 122 MOUNT_MSDOS, /* 4 */ 123 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 124 MOUNT_FDESC, /* 6 */ 125 MOUNT_KERNFS, /* 7 */ 126 NULL, /* 8 = MOUNT_DEVFS */ 127 MOUNT_AFS, /* 9 */ 128 }; 129 const int nmountcompatnames = sizeof(mountcompatnames) / 130 sizeof(mountcompatnames[0]); 131 #endif /* COMPAT_09 || COMPAT_43 */ 132 133 static int 134 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 135 void *data, size_t *data_len) 136 { 137 struct mount *mp; 138 int error = 0, saved_flags; 139 140 mp = vp->v_mount; 141 saved_flags = mp->mnt_flag; 142 143 /* We can operate only on VV_ROOT nodes. */ 144 if ((vp->v_vflag & VV_ROOT) == 0) { 145 error = EINVAL; 146 goto out; 147 } 148 149 /* 150 * We only allow the filesystem to be reloaded if it 151 * is currently mounted read-only. 152 */ 153 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 154 error = EOPNOTSUPP; /* Needs translation */ 155 goto out; 156 } 157 158 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 159 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 160 if (error) 161 goto out; 162 163 if (vfs_busy(mp, LK_NOWAIT, 0)) { 164 error = EPERM; 165 goto out; 166 } 167 168 mp->mnt_flag &= ~MNT_OP_FLAGS; 169 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 170 171 /* 172 * Set the mount level flags. 173 */ 174 if (flags & MNT_RDONLY) 175 mp->mnt_flag |= MNT_RDONLY; 176 else if (mp->mnt_flag & MNT_RDONLY) 177 mp->mnt_iflag |= IMNT_WANTRDWR; 178 mp->mnt_flag &= 179 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 180 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 181 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 182 mp->mnt_flag |= flags & 183 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 184 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 185 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 186 MNT_IGNORE); 187 188 error = VFS_MOUNT(mp, path, data, data_len); 189 190 #if defined(COMPAT_30) && defined(NFSSERVER) 191 if (error && data != NULL) { 192 int error2; 193 194 /* Update failed; let's try and see if it was an 195 * export request. */ 196 error2 = nfs_update_exports_30(mp, path, data, l); 197 198 /* Only update error code if the export request was 199 * understood but some problem occurred while 200 * processing it. */ 201 if (error2 != EJUSTRETURN) 202 error = error2; 203 } 204 #endif 205 if (mp->mnt_iflag & IMNT_WANTRDWR) 206 mp->mnt_flag &= ~MNT_RDONLY; 207 if (error) 208 mp->mnt_flag = saved_flags; 209 mp->mnt_flag &= ~MNT_OP_FLAGS; 210 mp->mnt_iflag &= ~IMNT_WANTRDWR; 211 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 212 if (mp->mnt_syncer == NULL) 213 error = vfs_allocate_syncvnode(mp); 214 } else { 215 if (mp->mnt_syncer != NULL) 216 vfs_deallocate_syncvnode(mp); 217 } 218 vfs_unbusy(mp); 219 220 out: 221 return (error); 222 } 223 224 static int 225 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 226 { 227 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 228 int error; 229 230 /* Copy file-system type from userspace. */ 231 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 232 if (error) { 233 #if defined(COMPAT_09) || defined(COMPAT_43) 234 /* 235 * Historically, filesystem types were identified by numbers. 236 * If we get an integer for the filesystem type instead of a 237 * string, we check to see if it matches one of the historic 238 * filesystem types. 239 */ 240 u_long fsindex = (u_long)fstype; 241 if (fsindex >= nmountcompatnames || 242 mountcompatnames[fsindex] == NULL) 243 return ENODEV; 244 strlcpy(fstypename, mountcompatnames[fsindex], 245 sizeof(fstypename)); 246 #else 247 return error; 248 #endif 249 } 250 251 #ifdef COMPAT_10 252 /* Accept `ufs' as an alias for `ffs'. */ 253 if (strcmp(fstypename, "ufs") == 0) 254 fstypename[0] = 'f'; 255 #endif 256 257 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 258 return ENODEV; 259 return 0; 260 } 261 262 static int 263 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 264 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 265 { 266 struct mount *mp = NULL; 267 struct vnode *vp = *vpp; 268 struct vattr va; 269 int error; 270 271 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 272 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 273 if (error) 274 return error; 275 276 /* Can't make a non-dir a mount-point (from here anyway). */ 277 if (vp->v_type != VDIR) 278 return ENOTDIR; 279 280 /* 281 * If the user is not root, ensure that they own the directory 282 * onto which we are attempting to mount. 283 */ 284 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 285 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 286 (error = kauth_authorize_generic(l->l_cred, 287 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 288 return error; 289 } 290 291 if (flags & MNT_EXPORTED) 292 return EINVAL; 293 294 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 295 return error; 296 297 /* 298 * Check if a file-system is not already mounted on this vnode. 299 */ 300 if (vp->v_mountedhere != NULL) 301 return EBUSY; 302 303 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 304 305 mp->mnt_op = vfsops; 306 307 TAILQ_INIT(&mp->mnt_vnodelist); 308 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 309 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 310 (void)vfs_busy(mp, LK_NOWAIT, 0); 311 312 mp->mnt_vnodecovered = vp; 313 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 314 mp->mnt_unmounter = NULL; 315 mount_initspecific(mp); 316 317 /* 318 * The underlying file system may refuse the mount for 319 * various reasons. Allow the user to force it to happen. 320 * 321 * Set the mount level flags. 322 */ 323 mp->mnt_flag = flags & 324 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 325 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 326 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 327 MNT_IGNORE | MNT_RDONLY); 328 329 error = VFS_MOUNT(mp, path, data, data_len); 330 mp->mnt_flag &= ~MNT_OP_FLAGS; 331 332 /* 333 * Put the new filesystem on the mount list after root. 334 */ 335 cache_purge(vp); 336 if (error != 0) { 337 vp->v_mountedhere = NULL; 338 mp->mnt_op->vfs_refcount--; 339 vfs_unbusy(mp); 340 vfs_destroy(mp); 341 return error; 342 } 343 344 mp->mnt_iflag &= ~IMNT_WANTRDWR; 345 vp->v_mountedhere = mp; 346 mutex_enter(&mountlist_lock); 347 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 348 mutex_exit(&mountlist_lock); 349 vn_restorerecurse(vp, recurse); 350 VOP_UNLOCK(vp, 0); 351 checkdirs(vp); 352 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 353 error = vfs_allocate_syncvnode(mp); 354 vfs_unbusy(mp); 355 (void) VFS_STATVFS(mp, &mp->mnt_stat); 356 error = VFS_START(mp, 0); 357 if (error) 358 vrele(vp); 359 *vpp = NULL; 360 return error; 361 } 362 363 static int 364 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 365 void *data, size_t *data_len) 366 { 367 struct mount *mp; 368 int error; 369 370 /* If MNT_GETARGS is specified, it should be the only flag. */ 371 if (flags & ~MNT_GETARGS) 372 return EINVAL; 373 374 mp = vp->v_mount; 375 376 /* XXX: probably some notion of "can see" here if we want isolation. */ 377 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 378 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 379 if (error) 380 return error; 381 382 if ((vp->v_vflag & VV_ROOT) == 0) 383 return EINVAL; 384 385 if (vfs_busy(mp, LK_NOWAIT, 0)) 386 return EPERM; 387 388 mp->mnt_flag &= ~MNT_OP_FLAGS; 389 mp->mnt_flag |= MNT_GETARGS; 390 error = VFS_MOUNT(mp, path, data, data_len); 391 mp->mnt_flag &= ~MNT_OP_FLAGS; 392 393 vfs_unbusy(mp); 394 return (error); 395 } 396 397 #ifdef COMPAT_40 398 /* ARGSUSED */ 399 int 400 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 401 { 402 /* { 403 syscallarg(const char *) type; 404 syscallarg(const char *) path; 405 syscallarg(int) flags; 406 syscallarg(void *) data; 407 } */ 408 register_t dummy; 409 410 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 411 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 412 } 413 #endif 414 415 int 416 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 417 { 418 /* { 419 syscallarg(const char *) type; 420 syscallarg(const char *) path; 421 syscallarg(int) flags; 422 syscallarg(void *) data; 423 syscallarg(size_t) data_len; 424 } */ 425 426 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 427 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 428 SCARG(uap, data_len), retval); 429 } 430 431 int 432 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 433 const char *path, int flags, void *data, enum uio_seg data_seg, 434 size_t data_len, register_t *retval) 435 { 436 struct vnode *vp; 437 struct nameidata nd; 438 void *data_buf = data; 439 u_int recurse; 440 int error; 441 442 /* 443 * Get vnode to be covered 444 */ 445 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 446 if ((error = namei(&nd)) != 0) 447 return (error); 448 vp = nd.ni_vp; 449 450 /* 451 * A lookup in VFS_MOUNT might result in an attempt to 452 * lock this vnode again, so make the lock recursive. 453 */ 454 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 455 recurse = vn_setrecurse(vp); 456 457 if (vfsops == NULL) { 458 if (flags & (MNT_GETARGS | MNT_UPDATE)) 459 vfsops = vp->v_mount->mnt_op; 460 else { 461 /* 'type' is userspace */ 462 error = mount_get_vfsops(type, &vfsops); 463 if (error != 0) 464 goto done; 465 } 466 } 467 468 if (data != NULL && data_seg == UIO_USERSPACE) { 469 if (data_len == 0) { 470 /* No length supplied, use default for filesystem */ 471 data_len = vfsops->vfs_min_mount_data; 472 if (data_len > VFS_MAX_MOUNT_DATA) { 473 /* maybe a force loaded old LKM */ 474 error = EINVAL; 475 goto done; 476 } 477 #ifdef COMPAT_30 478 /* Hopefully a longer buffer won't make copyin() fail */ 479 if (flags & MNT_UPDATE 480 && data_len < sizeof (struct mnt_export_args30)) 481 data_len = sizeof (struct mnt_export_args30); 482 #endif 483 } 484 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 485 486 /* NFS needs the buffer even for mnt_getargs .... */ 487 error = copyin(data, data_buf, data_len); 488 if (error != 0) 489 goto done; 490 } 491 492 if (flags & MNT_GETARGS) { 493 if (data_len == 0) { 494 error = EINVAL; 495 goto done; 496 } 497 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 498 if (error != 0) 499 goto done; 500 if (data_seg == UIO_USERSPACE) 501 error = copyout(data_buf, data, data_len); 502 *retval = data_len; 503 } else if (flags & MNT_UPDATE) { 504 error = mount_update(l, vp, path, flags, data_buf, &data_len); 505 } else { 506 /* Locking is handled internally in mount_domount(). */ 507 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 508 &data_len, recurse); 509 } 510 511 done: 512 if (vp != NULL) { 513 vn_restorerecurse(vp, recurse); 514 vput(vp); 515 } 516 if (data_buf != data) 517 free(data_buf, M_TEMP); 518 return (error); 519 } 520 521 /* 522 * Scan all active processes to see if any of them have a current 523 * or root directory onto which the new filesystem has just been 524 * mounted. If so, replace them with the new mount point. 525 */ 526 void 527 checkdirs(struct vnode *olddp) 528 { 529 struct cwdinfo *cwdi; 530 struct vnode *newdp; 531 struct proc *p; 532 533 if (olddp->v_usecount == 1) 534 return; 535 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 536 panic("mount: lost mount"); 537 mutex_enter(&proclist_lock); 538 PROCLIST_FOREACH(p, &allproc) { 539 cwdi = p->p_cwdi; 540 if (!cwdi) 541 continue; 542 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 543 if (cwdi->cwdi_cdir == olddp) { 544 vrele(cwdi->cwdi_cdir); 545 VREF(newdp); 546 cwdi->cwdi_cdir = newdp; 547 } 548 if (cwdi->cwdi_rdir == olddp) { 549 vrele(cwdi->cwdi_rdir); 550 VREF(newdp); 551 cwdi->cwdi_rdir = newdp; 552 } 553 rw_exit(&cwdi->cwdi_lock); 554 } 555 mutex_exit(&proclist_lock); 556 if (rootvnode == olddp) { 557 vrele(rootvnode); 558 VREF(newdp); 559 rootvnode = newdp; 560 } 561 vput(newdp); 562 } 563 564 /* 565 * Unmount a file system. 566 * 567 * Note: unmount takes a path to the vnode mounted on as argument, 568 * not special file (as before). 569 */ 570 /* ARGSUSED */ 571 int 572 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 573 { 574 /* { 575 syscallarg(const char *) path; 576 syscallarg(int) flags; 577 } */ 578 struct vnode *vp; 579 struct mount *mp; 580 int error; 581 struct nameidata nd; 582 583 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 584 SCARG(uap, path)); 585 if ((error = namei(&nd)) != 0) 586 return (error); 587 vp = nd.ni_vp; 588 mp = vp->v_mount; 589 590 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 591 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 592 if (error) { 593 vput(vp); 594 return (error); 595 } 596 597 /* 598 * Don't allow unmounting the root file system. 599 */ 600 if (mp->mnt_flag & MNT_ROOTFS) { 601 vput(vp); 602 return (EINVAL); 603 } 604 605 /* 606 * Must be the root of the filesystem 607 */ 608 if ((vp->v_vflag & VV_ROOT) == 0) { 609 vput(vp); 610 return (EINVAL); 611 } 612 vput(vp); 613 614 /* 615 * XXX Freeze syncer. Must do this before locking the 616 * mount point. See dounmount() for details. 617 */ 618 mutex_enter(&syncer_mutex); 619 620 if (vfs_busy(mp, 0, 0)) { 621 mutex_exit(&syncer_mutex); 622 return (EBUSY); 623 } 624 625 return (dounmount(mp, SCARG(uap, flags), l)); 626 } 627 628 /* 629 * Do the actual file system unmount. File system is assumed to have been 630 * marked busy by the caller. 631 */ 632 int 633 dounmount(struct mount *mp, int flags, struct lwp *l) 634 { 635 struct vnode *coveredvp; 636 int error; 637 int async; 638 int used_syncer; 639 640 #if NVERIEXEC > 0 641 error = veriexec_unmountchk(mp); 642 if (error) 643 return (error); 644 #endif /* NVERIEXEC > 0 */ 645 646 mutex_enter(&mountlist_lock); 647 vfs_unbusy(mp); 648 used_syncer = (mp->mnt_syncer != NULL); 649 650 /* 651 * XXX Syncer must be frozen when we get here. This should really 652 * be done on a per-mountpoint basis, but especially the softdep 653 * code possibly called from the syncer doesn't exactly work on a 654 * per-mountpoint basis, so the softdep code would become a maze 655 * of vfs_busy() calls. 656 * 657 * The caller of dounmount() must acquire syncer_mutex because 658 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 659 * order, and we must preserve that order to avoid deadlock. 660 * 661 * So, if the file system did not use the syncer, now is 662 * the time to release the syncer_mutex. 663 */ 664 if (used_syncer == 0) 665 mutex_exit(&syncer_mutex); 666 667 mp->mnt_iflag |= IMNT_UNMOUNT; 668 mp->mnt_unmounter = l; 669 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 670 671 async = mp->mnt_flag & MNT_ASYNC; 672 mp->mnt_flag &= ~MNT_ASYNC; 673 cache_purgevfs(mp); /* remove cache entries for this file sys */ 674 if (mp->mnt_syncer != NULL) 675 vfs_deallocate_syncvnode(mp); 676 error = 0; 677 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 678 #if NFSS > 0 679 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 680 #endif 681 if (error == 0) 682 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 683 } 684 vfs_scrubvnlist(mp); 685 if (error == 0 || (flags & MNT_FORCE)) 686 error = VFS_UNMOUNT(mp, flags); 687 if (error) { 688 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 689 (void) vfs_allocate_syncvnode(mp); 690 mutex_enter(&mountlist_lock); 691 mp->mnt_iflag &= ~IMNT_UNMOUNT; 692 mp->mnt_unmounter = NULL; 693 mp->mnt_flag |= async; 694 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 695 &mountlist_lock); 696 if (used_syncer) 697 mutex_exit(&syncer_mutex); 698 mutex_enter(&mp->mnt_mutex); 699 while (mp->mnt_wcnt > 0) { 700 wakeup(mp); 701 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 702 0, &mp->mnt_mutex); 703 } 704 mutex_exit(&mp->mnt_mutex); 705 return (error); 706 } 707 vfs_scrubvnlist(mp); 708 mutex_enter(&mountlist_lock); 709 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 710 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 711 coveredvp->v_mountedhere = NULL; 712 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 713 panic("unmount: dangling vnode"); 714 mp->mnt_iflag |= IMNT_GONE; 715 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 716 if (coveredvp != NULLVP) 717 vrele(coveredvp); 718 if (used_syncer) 719 mutex_exit(&syncer_mutex); 720 mutex_enter(&mp->mnt_mutex); 721 while (mp->mnt_wcnt > 0) { 722 wakeup(mp); 723 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 724 } 725 mutex_exit(&mp->mnt_mutex); 726 vfs_hooks_unmount(mp); 727 vfs_delref(mp->mnt_op); 728 vfs_destroy(mp); 729 return (0); 730 } 731 732 /* 733 * Sync each mounted filesystem. 734 */ 735 #ifdef DEBUG 736 int syncprt = 0; 737 struct ctldebug debug0 = { "syncprt", &syncprt }; 738 #endif 739 740 /* ARGSUSED */ 741 int 742 sys_sync(struct lwp *l, const void *v, register_t *retval) 743 { 744 struct mount *mp, *nmp; 745 int asyncflag; 746 747 if (l == NULL) 748 l = &lwp0; 749 750 mutex_enter(&mountlist_lock); 751 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 752 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 753 nmp = mp->mnt_list.cqe_prev; 754 continue; 755 } 756 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 757 asyncflag = mp->mnt_flag & MNT_ASYNC; 758 mp->mnt_flag &= ~MNT_ASYNC; 759 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 760 if (asyncflag) 761 mp->mnt_flag |= MNT_ASYNC; 762 } 763 mutex_enter(&mountlist_lock); 764 nmp = mp->mnt_list.cqe_prev; 765 vfs_unbusy(mp); 766 767 } 768 mutex_exit(&mountlist_lock); 769 #ifdef DEBUG 770 if (syncprt) 771 vfs_bufstats(); 772 #endif /* DEBUG */ 773 return (0); 774 } 775 776 /* 777 * Change filesystem quotas. 778 */ 779 /* ARGSUSED */ 780 int 781 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 782 { 783 /* { 784 syscallarg(const char *) path; 785 syscallarg(int) cmd; 786 syscallarg(int) uid; 787 syscallarg(void *) arg; 788 } */ 789 struct mount *mp; 790 int error; 791 struct nameidata nd; 792 793 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 794 SCARG(uap, path)); 795 if ((error = namei(&nd)) != 0) 796 return (error); 797 mp = nd.ni_vp->v_mount; 798 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 799 SCARG(uap, arg)); 800 vrele(nd.ni_vp); 801 return (error); 802 } 803 804 int 805 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 806 int root) 807 { 808 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 809 int error = 0; 810 811 /* 812 * If MNT_NOWAIT or MNT_LAZY is specified, do not 813 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 814 * overrides MNT_NOWAIT. 815 */ 816 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 817 (flags != MNT_WAIT && flags != 0)) { 818 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 819 goto done; 820 } 821 822 /* Get the filesystem stats now */ 823 memset(sp, 0, sizeof(*sp)); 824 if ((error = VFS_STATVFS(mp, sp)) != 0) { 825 return error; 826 } 827 828 if (cwdi->cwdi_rdir == NULL) 829 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 830 done: 831 if (cwdi->cwdi_rdir != NULL) { 832 size_t len; 833 char *bp; 834 char *path = PNBUF_GET(); 835 836 bp = path + MAXPATHLEN; 837 *--bp = '\0'; 838 rw_enter(&cwdi->cwdi_lock, RW_READER); 839 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 840 MAXPATHLEN / 2, 0, l); 841 rw_exit(&cwdi->cwdi_lock); 842 if (error) { 843 PNBUF_PUT(path); 844 return error; 845 } 846 len = strlen(bp); 847 /* 848 * for mount points that are below our root, we can see 849 * them, so we fix up the pathname and return them. The 850 * rest we cannot see, so we don't allow viewing the 851 * data. 852 */ 853 if (strncmp(bp, sp->f_mntonname, len) == 0) { 854 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 855 sizeof(sp->f_mntonname)); 856 if (sp->f_mntonname[0] == '\0') 857 (void)strlcpy(sp->f_mntonname, "/", 858 sizeof(sp->f_mntonname)); 859 } else { 860 if (root) 861 (void)strlcpy(sp->f_mntonname, "/", 862 sizeof(sp->f_mntonname)); 863 else 864 error = EPERM; 865 } 866 PNBUF_PUT(path); 867 } 868 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 869 return error; 870 } 871 872 /* 873 * Get filesystem statistics by path. 874 */ 875 int 876 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 877 { 878 struct mount *mp; 879 int error; 880 struct nameidata nd; 881 882 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 883 if ((error = namei(&nd)) != 0) 884 return error; 885 mp = nd.ni_vp->v_mount; 886 error = dostatvfs(mp, sb, l, flags, 1); 887 vrele(nd.ni_vp); 888 return error; 889 } 890 891 /* ARGSUSED */ 892 int 893 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 894 { 895 /* { 896 syscallarg(const char *) path; 897 syscallarg(struct statvfs *) buf; 898 syscallarg(int) flags; 899 } */ 900 struct statvfs *sb; 901 int error; 902 903 sb = STATVFSBUF_GET(); 904 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 905 if (error == 0) 906 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 907 STATVFSBUF_PUT(sb); 908 return error; 909 } 910 911 /* 912 * Get filesystem statistics by fd. 913 */ 914 int 915 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 916 { 917 struct proc *p = l->l_proc; 918 struct file *fp; 919 struct mount *mp; 920 int error; 921 922 /* getvnode() will use the descriptor for us */ 923 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 924 return (error); 925 mp = ((struct vnode *)fp->f_data)->v_mount; 926 error = dostatvfs(mp, sb, l, flags, 1); 927 FILE_UNUSE(fp, l); 928 return error; 929 } 930 931 /* ARGSUSED */ 932 int 933 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 934 { 935 /* { 936 syscallarg(int) fd; 937 syscallarg(struct statvfs *) buf; 938 syscallarg(int) flags; 939 } */ 940 struct statvfs *sb; 941 int error; 942 943 sb = STATVFSBUF_GET(); 944 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 945 if (error == 0) 946 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 947 STATVFSBUF_PUT(sb); 948 return error; 949 } 950 951 952 /* 953 * Get statistics on all filesystems. 954 */ 955 int 956 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 957 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 958 register_t *retval) 959 { 960 int root = 0; 961 struct proc *p = l->l_proc; 962 struct mount *mp, *nmp; 963 struct statvfs *sb; 964 size_t count, maxcount; 965 int error = 0; 966 967 sb = STATVFSBUF_GET(); 968 maxcount = bufsize / entry_sz; 969 mutex_enter(&mountlist_lock); 970 count = 0; 971 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 972 mp = nmp) { 973 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 974 nmp = CIRCLEQ_NEXT(mp, mnt_list); 975 continue; 976 } 977 if (sfsp && count < maxcount) { 978 error = dostatvfs(mp, sb, l, flags, 0); 979 if (error) { 980 mutex_enter(&mountlist_lock); 981 nmp = CIRCLEQ_NEXT(mp, mnt_list); 982 vfs_unbusy(mp); 983 continue; 984 } 985 error = copyfn(sb, sfsp, entry_sz); 986 if (error) { 987 vfs_unbusy(mp); 988 goto out; 989 } 990 sfsp = (char *)sfsp + entry_sz; 991 root |= strcmp(sb->f_mntonname, "/") == 0; 992 } 993 count++; 994 mutex_enter(&mountlist_lock); 995 nmp = CIRCLEQ_NEXT(mp, mnt_list); 996 vfs_unbusy(mp); 997 } 998 999 mutex_exit(&mountlist_lock); 1000 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1001 /* 1002 * fake a root entry 1003 */ 1004 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1005 sb, l, flags, 1); 1006 if (error != 0) 1007 goto out; 1008 if (sfsp) 1009 error = copyfn(sb, sfsp, entry_sz); 1010 count++; 1011 } 1012 if (sfsp && count > maxcount) 1013 *retval = maxcount; 1014 else 1015 *retval = count; 1016 out: 1017 STATVFSBUF_PUT(sb); 1018 return error; 1019 } 1020 1021 int 1022 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1023 { 1024 /* { 1025 syscallarg(struct statvfs *) buf; 1026 syscallarg(size_t) bufsize; 1027 syscallarg(int) flags; 1028 } */ 1029 1030 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1031 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1032 } 1033 1034 /* 1035 * Change current working directory to a given file descriptor. 1036 */ 1037 /* ARGSUSED */ 1038 int 1039 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1040 { 1041 /* { 1042 syscallarg(int) fd; 1043 } */ 1044 struct proc *p = l->l_proc; 1045 struct filedesc *fdp = p->p_fd; 1046 struct cwdinfo *cwdi; 1047 struct vnode *vp, *tdp; 1048 struct mount *mp; 1049 struct file *fp; 1050 int error; 1051 1052 /* getvnode() will use the descriptor for us */ 1053 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1054 return (error); 1055 vp = (struct vnode *)fp->f_data; 1056 1057 VREF(vp); 1058 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1059 if (vp->v_type != VDIR) 1060 error = ENOTDIR; 1061 else 1062 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1063 if (error) { 1064 vput(vp); 1065 goto out; 1066 } 1067 while ((mp = vp->v_mountedhere) != NULL) { 1068 if (vfs_busy(mp, 0, 0)) 1069 continue; 1070 1071 vput(vp); 1072 error = VFS_ROOT(mp, &tdp); 1073 vfs_unbusy(mp); 1074 if (error) 1075 goto out; 1076 vp = tdp; 1077 } 1078 VOP_UNLOCK(vp, 0); 1079 1080 /* 1081 * Disallow changing to a directory not under the process's 1082 * current root directory (if there is one). 1083 */ 1084 cwdi = p->p_cwdi; 1085 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1086 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1087 vrele(vp); 1088 error = EPERM; /* operation not permitted */ 1089 } else { 1090 vrele(cwdi->cwdi_cdir); 1091 cwdi->cwdi_cdir = vp; 1092 } 1093 rw_exit(&cwdi->cwdi_lock); 1094 1095 out: 1096 FILE_UNUSE(fp, l); 1097 return (error); 1098 } 1099 1100 /* 1101 * Change this process's notion of the root directory to a given file 1102 * descriptor. 1103 */ 1104 int 1105 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1106 { 1107 struct proc *p = l->l_proc; 1108 struct filedesc *fdp = p->p_fd; 1109 struct cwdinfo *cwdi; 1110 struct vnode *vp; 1111 struct file *fp; 1112 int error; 1113 1114 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1115 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1116 return error; 1117 /* getvnode() will use the descriptor for us */ 1118 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1119 return error; 1120 vp = (struct vnode *) fp->f_data; 1121 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1122 if (vp->v_type != VDIR) 1123 error = ENOTDIR; 1124 else 1125 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1126 VOP_UNLOCK(vp, 0); 1127 if (error) 1128 goto out; 1129 VREF(vp); 1130 1131 /* 1132 * Prevent escaping from chroot by putting the root under 1133 * the working directory. Silently chdir to / if we aren't 1134 * already there. 1135 */ 1136 cwdi = p->p_cwdi; 1137 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1138 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1139 /* 1140 * XXX would be more failsafe to change directory to a 1141 * deadfs node here instead 1142 */ 1143 vrele(cwdi->cwdi_cdir); 1144 VREF(vp); 1145 cwdi->cwdi_cdir = vp; 1146 } 1147 1148 if (cwdi->cwdi_rdir != NULL) 1149 vrele(cwdi->cwdi_rdir); 1150 cwdi->cwdi_rdir = vp; 1151 rw_exit(&cwdi->cwdi_lock); 1152 1153 out: 1154 FILE_UNUSE(fp, l); 1155 return (error); 1156 } 1157 1158 /* 1159 * Change current working directory (``.''). 1160 */ 1161 /* ARGSUSED */ 1162 int 1163 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1164 { 1165 /* { 1166 syscallarg(const char *) path; 1167 } */ 1168 struct proc *p = l->l_proc; 1169 struct cwdinfo *cwdi; 1170 int error; 1171 struct nameidata nd; 1172 1173 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1174 SCARG(uap, path)); 1175 if ((error = change_dir(&nd, l)) != 0) 1176 return (error); 1177 cwdi = p->p_cwdi; 1178 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1179 vrele(cwdi->cwdi_cdir); 1180 cwdi->cwdi_cdir = nd.ni_vp; 1181 rw_exit(&cwdi->cwdi_lock); 1182 return (0); 1183 } 1184 1185 /* 1186 * Change notion of root (``/'') directory. 1187 */ 1188 /* ARGSUSED */ 1189 int 1190 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1191 { 1192 /* { 1193 syscallarg(const char *) path; 1194 } */ 1195 struct proc *p = l->l_proc; 1196 struct cwdinfo *cwdi; 1197 struct vnode *vp; 1198 int error; 1199 struct nameidata nd; 1200 1201 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1202 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1203 return (error); 1204 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1205 SCARG(uap, path)); 1206 if ((error = change_dir(&nd, l)) != 0) 1207 return (error); 1208 1209 cwdi = p->p_cwdi; 1210 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1211 if (cwdi->cwdi_rdir != NULL) 1212 vrele(cwdi->cwdi_rdir); 1213 vp = nd.ni_vp; 1214 cwdi->cwdi_rdir = vp; 1215 1216 /* 1217 * Prevent escaping from chroot by putting the root under 1218 * the working directory. Silently chdir to / if we aren't 1219 * already there. 1220 */ 1221 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1222 /* 1223 * XXX would be more failsafe to change directory to a 1224 * deadfs node here instead 1225 */ 1226 vrele(cwdi->cwdi_cdir); 1227 VREF(vp); 1228 cwdi->cwdi_cdir = vp; 1229 } 1230 rw_exit(&cwdi->cwdi_lock); 1231 1232 return (0); 1233 } 1234 1235 /* 1236 * Common routine for chroot and chdir. 1237 */ 1238 static int 1239 change_dir(struct nameidata *ndp, struct lwp *l) 1240 { 1241 struct vnode *vp; 1242 int error; 1243 1244 if ((error = namei(ndp)) != 0) 1245 return (error); 1246 vp = ndp->ni_vp; 1247 if (vp->v_type != VDIR) 1248 error = ENOTDIR; 1249 else 1250 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1251 1252 if (error) 1253 vput(vp); 1254 else 1255 VOP_UNLOCK(vp, 0); 1256 return (error); 1257 } 1258 1259 /* 1260 * Check permissions, allocate an open file structure, 1261 * and call the device open routine if any. 1262 */ 1263 int 1264 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1265 { 1266 /* { 1267 syscallarg(const char *) path; 1268 syscallarg(int) flags; 1269 syscallarg(int) mode; 1270 } */ 1271 struct proc *p = l->l_proc; 1272 struct cwdinfo *cwdi = p->p_cwdi; 1273 struct filedesc *fdp = p->p_fd; 1274 struct file *fp; 1275 struct vnode *vp; 1276 int flags, cmode; 1277 int type, indx, error; 1278 struct flock lf; 1279 struct nameidata nd; 1280 1281 flags = FFLAGS(SCARG(uap, flags)); 1282 if ((flags & (FREAD | FWRITE)) == 0) 1283 return (EINVAL); 1284 /* falloc() will use the file descriptor for us */ 1285 if ((error = falloc(l, &fp, &indx)) != 0) 1286 return (error); 1287 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1288 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1289 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1290 SCARG(uap, path)); 1291 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1292 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1293 rw_enter(&fdp->fd_lock, RW_WRITER); 1294 FILE_UNUSE(fp, l); 1295 fdp->fd_ofiles[indx] = NULL; 1296 rw_exit(&fdp->fd_lock); 1297 ffree(fp); 1298 if ((error == EDUPFD || error == EMOVEFD) && 1299 l->l_dupfd >= 0 && /* XXX from fdopen */ 1300 (error = 1301 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1302 *retval = indx; 1303 return (0); 1304 } 1305 if (error == ERESTART) 1306 error = EINTR; 1307 fdremove(fdp, indx); 1308 return (error); 1309 } 1310 1311 l->l_dupfd = 0; 1312 vp = nd.ni_vp; 1313 fp->f_flag = flags & FMASK; 1314 fp->f_type = DTYPE_VNODE; 1315 fp->f_ops = &vnops; 1316 fp->f_data = vp; 1317 if (flags & (O_EXLOCK | O_SHLOCK)) { 1318 lf.l_whence = SEEK_SET; 1319 lf.l_start = 0; 1320 lf.l_len = 0; 1321 if (flags & O_EXLOCK) 1322 lf.l_type = F_WRLCK; 1323 else 1324 lf.l_type = F_RDLCK; 1325 type = F_FLOCK; 1326 if ((flags & FNONBLOCK) == 0) 1327 type |= F_WAIT; 1328 VOP_UNLOCK(vp, 0); 1329 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1330 if (error) { 1331 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1332 FILE_UNUSE(fp, l); 1333 ffree(fp); 1334 fdremove(fdp, indx); 1335 return (error); 1336 } 1337 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1338 fp->f_flag |= FHASLOCK; 1339 } 1340 VOP_UNLOCK(vp, 0); 1341 *retval = indx; 1342 FILE_SET_MATURE(fp); 1343 FILE_UNUSE(fp, l); 1344 return (0); 1345 } 1346 1347 static void 1348 vfs__fhfree(fhandle_t *fhp) 1349 { 1350 size_t fhsize; 1351 1352 if (fhp == NULL) { 1353 return; 1354 } 1355 fhsize = FHANDLE_SIZE(fhp); 1356 kmem_free(fhp, fhsize); 1357 } 1358 1359 /* 1360 * vfs_composefh: compose a filehandle. 1361 */ 1362 1363 int 1364 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1365 { 1366 struct mount *mp; 1367 struct fid *fidp; 1368 int error; 1369 size_t needfhsize; 1370 size_t fidsize; 1371 1372 mp = vp->v_mount; 1373 fidp = NULL; 1374 if (*fh_size < FHANDLE_SIZE_MIN) { 1375 fidsize = 0; 1376 } else { 1377 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1378 if (fhp != NULL) { 1379 memset(fhp, 0, *fh_size); 1380 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1381 fidp = &fhp->fh_fid; 1382 } 1383 } 1384 error = VFS_VPTOFH(vp, fidp, &fidsize); 1385 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1386 if (error == 0 && *fh_size < needfhsize) { 1387 error = E2BIG; 1388 } 1389 *fh_size = needfhsize; 1390 return error; 1391 } 1392 1393 int 1394 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1395 { 1396 struct mount *mp; 1397 fhandle_t *fhp; 1398 size_t fhsize; 1399 size_t fidsize; 1400 int error; 1401 1402 *fhpp = NULL; 1403 mp = vp->v_mount; 1404 fidsize = 0; 1405 error = VFS_VPTOFH(vp, NULL, &fidsize); 1406 KASSERT(error != 0); 1407 if (error != E2BIG) { 1408 goto out; 1409 } 1410 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1411 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1412 if (fhp == NULL) { 1413 error = ENOMEM; 1414 goto out; 1415 } 1416 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1417 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1418 if (error == 0) { 1419 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1420 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1421 *fhpp = fhp; 1422 } else { 1423 kmem_free(fhp, fhsize); 1424 } 1425 out: 1426 return error; 1427 } 1428 1429 void 1430 vfs_composefh_free(fhandle_t *fhp) 1431 { 1432 1433 vfs__fhfree(fhp); 1434 } 1435 1436 /* 1437 * vfs_fhtovp: lookup a vnode by a filehandle. 1438 */ 1439 1440 int 1441 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1442 { 1443 struct mount *mp; 1444 int error; 1445 1446 *vpp = NULL; 1447 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1448 if (mp == NULL) { 1449 error = ESTALE; 1450 goto out; 1451 } 1452 if (mp->mnt_op->vfs_fhtovp == NULL) { 1453 error = EOPNOTSUPP; 1454 goto out; 1455 } 1456 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1457 out: 1458 return error; 1459 } 1460 1461 /* 1462 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1463 * the needed size. 1464 */ 1465 1466 int 1467 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1468 { 1469 fhandle_t *fhp; 1470 int error; 1471 1472 *fhpp = NULL; 1473 if (fhsize > FHANDLE_SIZE_MAX) { 1474 return EINVAL; 1475 } 1476 if (fhsize < FHANDLE_SIZE_MIN) { 1477 return EINVAL; 1478 } 1479 again: 1480 fhp = kmem_alloc(fhsize, KM_SLEEP); 1481 if (fhp == NULL) { 1482 return ENOMEM; 1483 } 1484 error = copyin(ufhp, fhp, fhsize); 1485 if (error == 0) { 1486 /* XXX this check shouldn't be here */ 1487 if (FHANDLE_SIZE(fhp) == fhsize) { 1488 *fhpp = fhp; 1489 return 0; 1490 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1491 /* 1492 * a kludge for nfsv2 padded handles. 1493 */ 1494 size_t sz; 1495 1496 sz = FHANDLE_SIZE(fhp); 1497 kmem_free(fhp, fhsize); 1498 fhsize = sz; 1499 goto again; 1500 } else { 1501 /* 1502 * userland told us wrong size. 1503 */ 1504 error = EINVAL; 1505 } 1506 } 1507 kmem_free(fhp, fhsize); 1508 return error; 1509 } 1510 1511 void 1512 vfs_copyinfh_free(fhandle_t *fhp) 1513 { 1514 1515 vfs__fhfree(fhp); 1516 } 1517 1518 /* 1519 * Get file handle system call 1520 */ 1521 int 1522 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1523 { 1524 /* { 1525 syscallarg(char *) fname; 1526 syscallarg(fhandle_t *) fhp; 1527 syscallarg(size_t *) fh_size; 1528 } */ 1529 struct vnode *vp; 1530 fhandle_t *fh; 1531 int error; 1532 struct nameidata nd; 1533 size_t sz; 1534 size_t usz; 1535 1536 /* 1537 * Must be super user 1538 */ 1539 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1540 0, NULL, NULL, NULL); 1541 if (error) 1542 return (error); 1543 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1544 SCARG(uap, fname)); 1545 error = namei(&nd); 1546 if (error) 1547 return (error); 1548 vp = nd.ni_vp; 1549 error = vfs_composefh_alloc(vp, &fh); 1550 vput(vp); 1551 if (error != 0) { 1552 goto out; 1553 } 1554 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1555 if (error != 0) { 1556 goto out; 1557 } 1558 sz = FHANDLE_SIZE(fh); 1559 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1560 if (error != 0) { 1561 goto out; 1562 } 1563 if (usz >= sz) { 1564 error = copyout(fh, SCARG(uap, fhp), sz); 1565 } else { 1566 error = E2BIG; 1567 } 1568 out: 1569 vfs_composefh_free(fh); 1570 return (error); 1571 } 1572 1573 /* 1574 * Open a file given a file handle. 1575 * 1576 * Check permissions, allocate an open file structure, 1577 * and call the device open routine if any. 1578 */ 1579 1580 int 1581 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1582 register_t *retval) 1583 { 1584 struct filedesc *fdp = l->l_proc->p_fd; 1585 struct file *fp; 1586 struct vnode *vp = NULL; 1587 kauth_cred_t cred = l->l_cred; 1588 struct file *nfp; 1589 int type, indx, error=0; 1590 struct flock lf; 1591 struct vattr va; 1592 fhandle_t *fh; 1593 int flags; 1594 1595 /* 1596 * Must be super user 1597 */ 1598 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1599 0, NULL, NULL, NULL))) 1600 return (error); 1601 1602 flags = FFLAGS(oflags); 1603 if ((flags & (FREAD | FWRITE)) == 0) 1604 return (EINVAL); 1605 if ((flags & O_CREAT)) 1606 return (EINVAL); 1607 /* falloc() will use the file descriptor for us */ 1608 if ((error = falloc(l, &nfp, &indx)) != 0) 1609 return (error); 1610 fp = nfp; 1611 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1612 if (error != 0) { 1613 goto bad; 1614 } 1615 error = vfs_fhtovp(fh, &vp); 1616 if (error != 0) { 1617 goto bad; 1618 } 1619 1620 /* Now do an effective vn_open */ 1621 1622 if (vp->v_type == VSOCK) { 1623 error = EOPNOTSUPP; 1624 goto bad; 1625 } 1626 error = vn_openchk(vp, cred, flags); 1627 if (error != 0) 1628 goto bad; 1629 if (flags & O_TRUNC) { 1630 VOP_UNLOCK(vp, 0); /* XXX */ 1631 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1632 VATTR_NULL(&va); 1633 va.va_size = 0; 1634 error = VOP_SETATTR(vp, &va, cred); 1635 if (error) 1636 goto bad; 1637 } 1638 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1639 goto bad; 1640 if (flags & FWRITE) 1641 vp->v_writecount++; 1642 1643 /* done with modified vn_open, now finish what sys_open does. */ 1644 1645 fp->f_flag = flags & FMASK; 1646 fp->f_type = DTYPE_VNODE; 1647 fp->f_ops = &vnops; 1648 fp->f_data = vp; 1649 if (flags & (O_EXLOCK | O_SHLOCK)) { 1650 lf.l_whence = SEEK_SET; 1651 lf.l_start = 0; 1652 lf.l_len = 0; 1653 if (flags & O_EXLOCK) 1654 lf.l_type = F_WRLCK; 1655 else 1656 lf.l_type = F_RDLCK; 1657 type = F_FLOCK; 1658 if ((flags & FNONBLOCK) == 0) 1659 type |= F_WAIT; 1660 VOP_UNLOCK(vp, 0); 1661 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1662 if (error) { 1663 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1664 FILE_UNUSE(fp, l); 1665 ffree(fp); 1666 fdremove(fdp, indx); 1667 return (error); 1668 } 1669 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1670 fp->f_flag |= FHASLOCK; 1671 } 1672 VOP_UNLOCK(vp, 0); 1673 *retval = indx; 1674 FILE_SET_MATURE(fp); 1675 FILE_UNUSE(fp, l); 1676 vfs_copyinfh_free(fh); 1677 return (0); 1678 1679 bad: 1680 FILE_UNUSE(fp, l); 1681 ffree(fp); 1682 fdremove(fdp, indx); 1683 if (vp != NULL) 1684 vput(vp); 1685 vfs_copyinfh_free(fh); 1686 return (error); 1687 } 1688 1689 int 1690 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1691 { 1692 /* { 1693 syscallarg(const void *) fhp; 1694 syscallarg(size_t) fh_size; 1695 syscallarg(int) flags; 1696 } */ 1697 1698 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1699 SCARG(uap, flags), retval); 1700 } 1701 1702 int 1703 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1704 { 1705 int error; 1706 fhandle_t *fh; 1707 struct vnode *vp; 1708 1709 /* 1710 * Must be super user 1711 */ 1712 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1713 0, NULL, NULL, NULL))) 1714 return (error); 1715 1716 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1717 if (error != 0) 1718 return error; 1719 1720 error = vfs_fhtovp(fh, &vp); 1721 vfs_copyinfh_free(fh); 1722 if (error != 0) 1723 return error; 1724 1725 error = vn_stat(vp, sb, l); 1726 vput(vp); 1727 return error; 1728 } 1729 1730 1731 /* ARGSUSED */ 1732 int 1733 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1734 { 1735 /* { 1736 syscallarg(const void *) fhp; 1737 syscallarg(size_t) fh_size; 1738 syscallarg(struct stat *) sb; 1739 } */ 1740 struct stat sb; 1741 int error; 1742 1743 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1744 if (error) 1745 return error; 1746 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1747 } 1748 1749 int 1750 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1751 int flags) 1752 { 1753 fhandle_t *fh; 1754 struct mount *mp; 1755 struct vnode *vp; 1756 int error; 1757 1758 /* 1759 * Must be super user 1760 */ 1761 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1762 0, NULL, NULL, NULL))) 1763 return error; 1764 1765 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1766 if (error != 0) 1767 return error; 1768 1769 error = vfs_fhtovp(fh, &vp); 1770 vfs_copyinfh_free(fh); 1771 if (error != 0) 1772 return error; 1773 1774 mp = vp->v_mount; 1775 error = dostatvfs(mp, sb, l, flags, 1); 1776 vput(vp); 1777 return error; 1778 } 1779 1780 /* ARGSUSED */ 1781 int 1782 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1783 { 1784 /* { 1785 syscallarg(const void *) fhp; 1786 syscallarg(size_t) fh_size; 1787 syscallarg(struct statvfs *) buf; 1788 syscallarg(int) flags; 1789 } */ 1790 struct statvfs *sb = STATVFSBUF_GET(); 1791 int error; 1792 1793 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1794 SCARG(uap, flags)); 1795 if (error == 0) 1796 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1797 STATVFSBUF_PUT(sb); 1798 return error; 1799 } 1800 1801 /* 1802 * Create a special file. 1803 */ 1804 /* ARGSUSED */ 1805 int 1806 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1807 { 1808 /* { 1809 syscallarg(const char *) path; 1810 syscallarg(int) mode; 1811 syscallarg(int) dev; 1812 } */ 1813 struct proc *p = l->l_proc; 1814 struct vnode *vp; 1815 struct vattr vattr; 1816 int error, optype; 1817 struct nameidata nd; 1818 char *path; 1819 const char *cpath; 1820 enum uio_seg seg = UIO_USERSPACE; 1821 1822 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1823 0, NULL, NULL, NULL)) != 0) 1824 return (error); 1825 1826 optype = VOP_MKNOD_DESCOFFSET; 1827 1828 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1829 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1830 1831 if ((error = namei(&nd)) != 0) 1832 goto out; 1833 vp = nd.ni_vp; 1834 if (vp != NULL) 1835 error = EEXIST; 1836 else { 1837 VATTR_NULL(&vattr); 1838 /* We will read cwdi->cwdi_cmask unlocked. */ 1839 vattr.va_mode = 1840 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1841 vattr.va_rdev = SCARG(uap, dev); 1842 1843 switch (SCARG(uap, mode) & S_IFMT) { 1844 case S_IFMT: /* used by badsect to flag bad sectors */ 1845 vattr.va_type = VBAD; 1846 break; 1847 case S_IFCHR: 1848 vattr.va_type = VCHR; 1849 break; 1850 case S_IFBLK: 1851 vattr.va_type = VBLK; 1852 break; 1853 case S_IFWHT: 1854 optype = VOP_WHITEOUT_DESCOFFSET; 1855 break; 1856 case S_IFREG: 1857 #if NVERIEXEC > 0 1858 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1859 O_CREAT); 1860 #endif /* NVERIEXEC > 0 */ 1861 vattr.va_type = VREG; 1862 vattr.va_rdev = VNOVAL; 1863 optype = VOP_CREATE_DESCOFFSET; 1864 break; 1865 default: 1866 error = EINVAL; 1867 break; 1868 } 1869 } 1870 if (!error) { 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 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1941 if (error == 0) 1942 vput(nd.ni_vp); 1943 return (error); 1944 } 1945 1946 /* 1947 * Make a hard file link. 1948 */ 1949 /* ARGSUSED */ 1950 int 1951 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1952 { 1953 /* { 1954 syscallarg(const char *) path; 1955 syscallarg(const char *) link; 1956 } */ 1957 struct vnode *vp; 1958 struct nameidata nd; 1959 int error; 1960 1961 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1962 SCARG(uap, path)); 1963 if ((error = namei(&nd)) != 0) 1964 return (error); 1965 vp = nd.ni_vp; 1966 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1967 SCARG(uap, link)); 1968 if ((error = namei(&nd)) != 0) 1969 goto out; 1970 if (nd.ni_vp) { 1971 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1972 if (nd.ni_dvp == nd.ni_vp) 1973 vrele(nd.ni_dvp); 1974 else 1975 vput(nd.ni_dvp); 1976 vrele(nd.ni_vp); 1977 error = EEXIST; 1978 goto out; 1979 } 1980 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1981 out: 1982 vrele(vp); 1983 return (error); 1984 } 1985 1986 /* 1987 * Make a symbolic link. 1988 */ 1989 /* ARGSUSED */ 1990 int 1991 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1992 { 1993 /* { 1994 syscallarg(const char *) path; 1995 syscallarg(const char *) link; 1996 } */ 1997 struct proc *p = l->l_proc; 1998 struct vattr vattr; 1999 char *path; 2000 int error; 2001 struct nameidata nd; 2002 2003 path = PNBUF_GET(); 2004 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2005 if (error) 2006 goto out; 2007 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2008 SCARG(uap, link)); 2009 if ((error = namei(&nd)) != 0) 2010 goto out; 2011 if (nd.ni_vp) { 2012 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2013 if (nd.ni_dvp == nd.ni_vp) 2014 vrele(nd.ni_dvp); 2015 else 2016 vput(nd.ni_dvp); 2017 vrele(nd.ni_vp); 2018 error = EEXIST; 2019 goto out; 2020 } 2021 VATTR_NULL(&vattr); 2022 vattr.va_type = VLNK; 2023 /* We will read cwdi->cwdi_cmask unlocked. */ 2024 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2025 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2026 if (error == 0) 2027 vput(nd.ni_vp); 2028 out: 2029 PNBUF_PUT(path); 2030 return (error); 2031 } 2032 2033 /* 2034 * Delete a whiteout from the filesystem. 2035 */ 2036 /* ARGSUSED */ 2037 int 2038 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2039 { 2040 /* { 2041 syscallarg(const char *) path; 2042 } */ 2043 int error; 2044 struct nameidata nd; 2045 2046 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2047 UIO_USERSPACE, SCARG(uap, path)); 2048 error = namei(&nd); 2049 if (error) 2050 return (error); 2051 2052 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2053 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2054 if (nd.ni_dvp == nd.ni_vp) 2055 vrele(nd.ni_dvp); 2056 else 2057 vput(nd.ni_dvp); 2058 if (nd.ni_vp) 2059 vrele(nd.ni_vp); 2060 return (EEXIST); 2061 } 2062 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2063 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2064 vput(nd.ni_dvp); 2065 return (error); 2066 } 2067 2068 /* 2069 * Delete a name from the filesystem. 2070 */ 2071 /* ARGSUSED */ 2072 int 2073 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2074 { 2075 /* { 2076 syscallarg(const char *) path; 2077 } */ 2078 2079 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2080 } 2081 2082 int 2083 do_sys_unlink(const char *arg, enum uio_seg seg) 2084 { 2085 struct vnode *vp; 2086 int error; 2087 struct nameidata nd; 2088 kauth_cred_t cred; 2089 char *path; 2090 const char *cpath; 2091 2092 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2093 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2094 2095 if ((error = namei(&nd)) != 0) 2096 goto out; 2097 vp = nd.ni_vp; 2098 2099 /* 2100 * The root of a mounted filesystem cannot be deleted. 2101 */ 2102 if (vp->v_vflag & VV_ROOT) { 2103 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2104 if (nd.ni_dvp == vp) 2105 vrele(nd.ni_dvp); 2106 else 2107 vput(nd.ni_dvp); 2108 vput(vp); 2109 error = EBUSY; 2110 goto out; 2111 } 2112 2113 #if NVERIEXEC > 0 2114 /* Handle remove requests for veriexec entries. */ 2115 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2116 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2117 if (nd.ni_dvp == vp) 2118 vrele(nd.ni_dvp); 2119 else 2120 vput(nd.ni_dvp); 2121 vput(vp); 2122 goto out; 2123 } 2124 #endif /* NVERIEXEC > 0 */ 2125 2126 cred = kauth_cred_get(); 2127 #ifdef FILEASSOC 2128 (void)fileassoc_file_delete(vp); 2129 #endif /* FILEASSOC */ 2130 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2131 out: 2132 VERIEXEC_PATH_PUT(path); 2133 return (error); 2134 } 2135 2136 /* 2137 * Reposition read/write file offset. 2138 */ 2139 int 2140 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2141 { 2142 /* { 2143 syscallarg(int) fd; 2144 syscallarg(int) pad; 2145 syscallarg(off_t) offset; 2146 syscallarg(int) whence; 2147 } */ 2148 struct proc *p = l->l_proc; 2149 kauth_cred_t cred = l->l_cred; 2150 struct filedesc *fdp = p->p_fd; 2151 struct file *fp; 2152 struct vnode *vp; 2153 struct vattr vattr; 2154 off_t newoff; 2155 int error; 2156 2157 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2158 return (EBADF); 2159 2160 vp = (struct vnode *)fp->f_data; 2161 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2162 error = ESPIPE; 2163 FILE_UNLOCK(fp); 2164 goto out; 2165 } 2166 2167 switch (SCARG(uap, whence)) { 2168 case SEEK_CUR: 2169 newoff = fp->f_offset + SCARG(uap, offset); 2170 FILE_USE(fp); 2171 break; 2172 case SEEK_END: 2173 FILE_USE(fp); 2174 error = VOP_GETATTR(vp, &vattr, cred); 2175 if (error) { 2176 FILE_UNUSE(fp, l); 2177 goto out; 2178 } 2179 newoff = SCARG(uap, offset) + vattr.va_size; 2180 break; 2181 case SEEK_SET: 2182 FILE_USE(fp); 2183 newoff = SCARG(uap, offset); 2184 break; 2185 default: 2186 FILE_UNLOCK(fp); 2187 error = EINVAL; 2188 goto out; 2189 } 2190 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2191 FILE_LOCK(fp); 2192 *(off_t *)retval = fp->f_offset = newoff; 2193 FILE_UNLOCK(fp); 2194 } 2195 FILE_UNUSE(fp, l); 2196 out: 2197 return (error); 2198 } 2199 2200 /* 2201 * Positional read system call. 2202 */ 2203 int 2204 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2205 { 2206 /* { 2207 syscallarg(int) fd; 2208 syscallarg(void *) buf; 2209 syscallarg(size_t) nbyte; 2210 syscallarg(off_t) offset; 2211 } */ 2212 struct proc *p = l->l_proc; 2213 struct filedesc *fdp = p->p_fd; 2214 struct file *fp; 2215 struct vnode *vp; 2216 off_t offset; 2217 int error, fd = SCARG(uap, fd); 2218 2219 if ((fp = fd_getfile(fdp, fd)) == NULL) 2220 return (EBADF); 2221 2222 if ((fp->f_flag & FREAD) == 0) { 2223 FILE_UNLOCK(fp); 2224 return (EBADF); 2225 } 2226 2227 FILE_USE(fp); 2228 2229 vp = (struct vnode *)fp->f_data; 2230 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2231 error = ESPIPE; 2232 goto out; 2233 } 2234 2235 offset = SCARG(uap, offset); 2236 2237 /* 2238 * XXX This works because no file systems actually 2239 * XXX take any action on the seek operation. 2240 */ 2241 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2242 goto out; 2243 2244 /* dofileread() will unuse the descriptor for us */ 2245 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2246 &offset, 0, retval)); 2247 2248 out: 2249 FILE_UNUSE(fp, l); 2250 return (error); 2251 } 2252 2253 /* 2254 * Positional scatter read system call. 2255 */ 2256 int 2257 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2258 { 2259 /* { 2260 syscallarg(int) fd; 2261 syscallarg(const struct iovec *) iovp; 2262 syscallarg(int) iovcnt; 2263 syscallarg(off_t) offset; 2264 } */ 2265 off_t offset = SCARG(uap, offset); 2266 2267 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2268 SCARG(uap, iovcnt), &offset, 0, retval); 2269 } 2270 2271 /* 2272 * Positional write system call. 2273 */ 2274 int 2275 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2276 { 2277 /* { 2278 syscallarg(int) fd; 2279 syscallarg(const void *) buf; 2280 syscallarg(size_t) nbyte; 2281 syscallarg(off_t) offset; 2282 } */ 2283 struct proc *p = l->l_proc; 2284 struct filedesc *fdp = p->p_fd; 2285 struct file *fp; 2286 struct vnode *vp; 2287 off_t offset; 2288 int error, fd = SCARG(uap, fd); 2289 2290 if ((fp = fd_getfile(fdp, fd)) == NULL) 2291 return (EBADF); 2292 2293 if ((fp->f_flag & FWRITE) == 0) { 2294 FILE_UNLOCK(fp); 2295 return (EBADF); 2296 } 2297 2298 FILE_USE(fp); 2299 2300 vp = (struct vnode *)fp->f_data; 2301 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2302 error = ESPIPE; 2303 goto out; 2304 } 2305 2306 offset = SCARG(uap, offset); 2307 2308 /* 2309 * XXX This works because no file systems actually 2310 * XXX take any action on the seek operation. 2311 */ 2312 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2313 goto out; 2314 2315 /* dofilewrite() will unuse the descriptor for us */ 2316 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2317 &offset, 0, retval)); 2318 2319 out: 2320 FILE_UNUSE(fp, l); 2321 return (error); 2322 } 2323 2324 /* 2325 * Positional gather write system call. 2326 */ 2327 int 2328 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2329 { 2330 /* { 2331 syscallarg(int) fd; 2332 syscallarg(const struct iovec *) iovp; 2333 syscallarg(int) iovcnt; 2334 syscallarg(off_t) offset; 2335 } */ 2336 off_t offset = SCARG(uap, offset); 2337 2338 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2339 SCARG(uap, iovcnt), &offset, 0, retval); 2340 } 2341 2342 /* 2343 * Check access permissions. 2344 */ 2345 int 2346 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2347 { 2348 /* { 2349 syscallarg(const char *) path; 2350 syscallarg(int) flags; 2351 } */ 2352 kauth_cred_t cred; 2353 struct vnode *vp; 2354 int error, flags; 2355 struct nameidata nd; 2356 2357 cred = kauth_cred_dup(l->l_cred); 2358 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2359 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2360 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2361 SCARG(uap, path)); 2362 /* Override default credentials */ 2363 nd.ni_cnd.cn_cred = cred; 2364 if ((error = namei(&nd)) != 0) 2365 goto out; 2366 vp = nd.ni_vp; 2367 2368 /* Flags == 0 means only check for existence. */ 2369 if (SCARG(uap, flags)) { 2370 flags = 0; 2371 if (SCARG(uap, flags) & R_OK) 2372 flags |= VREAD; 2373 if (SCARG(uap, flags) & W_OK) 2374 flags |= VWRITE; 2375 if (SCARG(uap, flags) & X_OK) 2376 flags |= VEXEC; 2377 2378 error = VOP_ACCESS(vp, flags, cred); 2379 if (!error && (flags & VWRITE)) 2380 error = vn_writechk(vp); 2381 } 2382 vput(vp); 2383 out: 2384 kauth_cred_free(cred); 2385 return (error); 2386 } 2387 2388 /* 2389 * Common code for all sys_stat functions, including compat versions. 2390 */ 2391 int 2392 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2393 struct stat *sb) 2394 { 2395 int error; 2396 struct nameidata nd; 2397 2398 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2399 UIO_USERSPACE, path); 2400 error = namei(&nd); 2401 if (error != 0) 2402 return error; 2403 error = vn_stat(nd.ni_vp, sb, l); 2404 vput(nd.ni_vp); 2405 return error; 2406 } 2407 2408 /* 2409 * Get file status; this version follows links. 2410 */ 2411 /* ARGSUSED */ 2412 int 2413 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2414 { 2415 /* { 2416 syscallarg(const char *) path; 2417 syscallarg(struct stat *) ub; 2418 } */ 2419 struct stat sb; 2420 int error; 2421 2422 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2423 if (error) 2424 return error; 2425 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2426 } 2427 2428 /* 2429 * Get file status; this version does not follow links. 2430 */ 2431 /* ARGSUSED */ 2432 int 2433 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2434 { 2435 /* { 2436 syscallarg(const char *) path; 2437 syscallarg(struct stat *) ub; 2438 } */ 2439 struct stat sb; 2440 int error; 2441 2442 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2443 if (error) 2444 return error; 2445 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2446 } 2447 2448 /* 2449 * Get configurable pathname variables. 2450 */ 2451 /* ARGSUSED */ 2452 int 2453 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2454 { 2455 /* { 2456 syscallarg(const char *) path; 2457 syscallarg(int) name; 2458 } */ 2459 int error; 2460 struct nameidata nd; 2461 2462 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2463 SCARG(uap, path)); 2464 if ((error = namei(&nd)) != 0) 2465 return (error); 2466 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2467 vput(nd.ni_vp); 2468 return (error); 2469 } 2470 2471 /* 2472 * Return target name of a symbolic link. 2473 */ 2474 /* ARGSUSED */ 2475 int 2476 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2477 { 2478 /* { 2479 syscallarg(const char *) path; 2480 syscallarg(char *) buf; 2481 syscallarg(size_t) count; 2482 } */ 2483 struct vnode *vp; 2484 struct iovec aiov; 2485 struct uio auio; 2486 int error; 2487 struct nameidata nd; 2488 2489 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2490 SCARG(uap, path)); 2491 if ((error = namei(&nd)) != 0) 2492 return (error); 2493 vp = nd.ni_vp; 2494 if (vp->v_type != VLNK) 2495 error = EINVAL; 2496 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2497 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2498 aiov.iov_base = SCARG(uap, buf); 2499 aiov.iov_len = SCARG(uap, count); 2500 auio.uio_iov = &aiov; 2501 auio.uio_iovcnt = 1; 2502 auio.uio_offset = 0; 2503 auio.uio_rw = UIO_READ; 2504 KASSERT(l == curlwp); 2505 auio.uio_vmspace = l->l_proc->p_vmspace; 2506 auio.uio_resid = SCARG(uap, count); 2507 error = VOP_READLINK(vp, &auio, l->l_cred); 2508 } 2509 vput(vp); 2510 *retval = SCARG(uap, count) - auio.uio_resid; 2511 return (error); 2512 } 2513 2514 /* 2515 * Change flags of a file given a path name. 2516 */ 2517 /* ARGSUSED */ 2518 int 2519 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2520 { 2521 /* { 2522 syscallarg(const char *) path; 2523 syscallarg(u_long) flags; 2524 } */ 2525 struct vnode *vp; 2526 int error; 2527 struct nameidata nd; 2528 2529 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2530 SCARG(uap, path)); 2531 if ((error = namei(&nd)) != 0) 2532 return (error); 2533 vp = nd.ni_vp; 2534 error = change_flags(vp, SCARG(uap, flags), l); 2535 vput(vp); 2536 return (error); 2537 } 2538 2539 /* 2540 * Change flags of a file given a file descriptor. 2541 */ 2542 /* ARGSUSED */ 2543 int 2544 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2545 { 2546 /* { 2547 syscallarg(int) fd; 2548 syscallarg(u_long) flags; 2549 } */ 2550 struct proc *p = l->l_proc; 2551 struct vnode *vp; 2552 struct file *fp; 2553 int error; 2554 2555 /* getvnode() will use the descriptor for us */ 2556 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2557 return (error); 2558 vp = (struct vnode *)fp->f_data; 2559 error = change_flags(vp, SCARG(uap, flags), l); 2560 VOP_UNLOCK(vp, 0); 2561 FILE_UNUSE(fp, l); 2562 return (error); 2563 } 2564 2565 /* 2566 * Change flags of a file given a path name; this version does 2567 * not follow links. 2568 */ 2569 int 2570 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2571 { 2572 /* { 2573 syscallarg(const char *) path; 2574 syscallarg(u_long) flags; 2575 } */ 2576 struct vnode *vp; 2577 int error; 2578 struct nameidata nd; 2579 2580 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2581 SCARG(uap, path)); 2582 if ((error = namei(&nd)) != 0) 2583 return (error); 2584 vp = nd.ni_vp; 2585 error = change_flags(vp, SCARG(uap, flags), l); 2586 vput(vp); 2587 return (error); 2588 } 2589 2590 /* 2591 * Common routine to change flags of a file. 2592 */ 2593 int 2594 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2595 { 2596 struct vattr vattr; 2597 int error; 2598 2599 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2600 /* 2601 * Non-superusers cannot change the flags on devices, even if they 2602 * own them. 2603 */ 2604 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2605 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2606 goto out; 2607 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2608 error = EINVAL; 2609 goto out; 2610 } 2611 } 2612 VATTR_NULL(&vattr); 2613 vattr.va_flags = flags; 2614 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2615 out: 2616 return (error); 2617 } 2618 2619 /* 2620 * Change mode of a file given path name; this version follows links. 2621 */ 2622 /* ARGSUSED */ 2623 int 2624 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2625 { 2626 /* { 2627 syscallarg(const char *) path; 2628 syscallarg(int) mode; 2629 } */ 2630 int error; 2631 struct nameidata nd; 2632 2633 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2634 SCARG(uap, path)); 2635 if ((error = namei(&nd)) != 0) 2636 return (error); 2637 2638 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2639 2640 vrele(nd.ni_vp); 2641 return (error); 2642 } 2643 2644 /* 2645 * Change mode of a file given a file descriptor. 2646 */ 2647 /* ARGSUSED */ 2648 int 2649 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2650 { 2651 /* { 2652 syscallarg(int) fd; 2653 syscallarg(int) mode; 2654 } */ 2655 struct proc *p = l->l_proc; 2656 struct file *fp; 2657 int error; 2658 2659 /* getvnode() will use the descriptor for us */ 2660 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2661 return (error); 2662 2663 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2664 FILE_UNUSE(fp, l); 2665 return (error); 2666 } 2667 2668 /* 2669 * Change mode of a file given path name; this version does not follow links. 2670 */ 2671 /* ARGSUSED */ 2672 int 2673 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2674 { 2675 /* { 2676 syscallarg(const char *) path; 2677 syscallarg(int) mode; 2678 } */ 2679 int error; 2680 struct nameidata nd; 2681 2682 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2683 SCARG(uap, path)); 2684 if ((error = namei(&nd)) != 0) 2685 return (error); 2686 2687 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2688 2689 vrele(nd.ni_vp); 2690 return (error); 2691 } 2692 2693 /* 2694 * Common routine to set mode given a vnode. 2695 */ 2696 static int 2697 change_mode(struct vnode *vp, int mode, struct lwp *l) 2698 { 2699 struct vattr vattr; 2700 int error; 2701 2702 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2703 VATTR_NULL(&vattr); 2704 vattr.va_mode = mode & ALLPERMS; 2705 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2706 VOP_UNLOCK(vp, 0); 2707 return (error); 2708 } 2709 2710 /* 2711 * Set ownership given a path name; this version follows links. 2712 */ 2713 /* ARGSUSED */ 2714 int 2715 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2716 { 2717 /* { 2718 syscallarg(const char *) path; 2719 syscallarg(uid_t) uid; 2720 syscallarg(gid_t) gid; 2721 } */ 2722 int error; 2723 struct nameidata nd; 2724 2725 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2726 SCARG(uap, path)); 2727 if ((error = namei(&nd)) != 0) 2728 return (error); 2729 2730 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2731 2732 vrele(nd.ni_vp); 2733 return (error); 2734 } 2735 2736 /* 2737 * Set ownership given a path name; this version follows links. 2738 * Provides POSIX semantics. 2739 */ 2740 /* ARGSUSED */ 2741 int 2742 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2743 { 2744 /* { 2745 syscallarg(const char *) path; 2746 syscallarg(uid_t) uid; 2747 syscallarg(gid_t) gid; 2748 } */ 2749 int error; 2750 struct nameidata nd; 2751 2752 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2753 SCARG(uap, path)); 2754 if ((error = namei(&nd)) != 0) 2755 return (error); 2756 2757 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2758 2759 vrele(nd.ni_vp); 2760 return (error); 2761 } 2762 2763 /* 2764 * Set ownership given a file descriptor. 2765 */ 2766 /* ARGSUSED */ 2767 int 2768 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2769 { 2770 /* { 2771 syscallarg(int) fd; 2772 syscallarg(uid_t) uid; 2773 syscallarg(gid_t) gid; 2774 } */ 2775 struct proc *p = l->l_proc; 2776 int error; 2777 struct file *fp; 2778 2779 /* getvnode() will use the descriptor for us */ 2780 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2781 return (error); 2782 2783 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2784 SCARG(uap, gid), l, 0); 2785 FILE_UNUSE(fp, l); 2786 return (error); 2787 } 2788 2789 /* 2790 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2791 */ 2792 /* ARGSUSED */ 2793 int 2794 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2795 { 2796 /* { 2797 syscallarg(int) fd; 2798 syscallarg(uid_t) uid; 2799 syscallarg(gid_t) gid; 2800 } */ 2801 struct proc *p = l->l_proc; 2802 int error; 2803 struct file *fp; 2804 2805 /* getvnode() will use the descriptor for us */ 2806 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2807 return (error); 2808 2809 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2810 SCARG(uap, gid), l, 1); 2811 FILE_UNUSE(fp, l); 2812 return (error); 2813 } 2814 2815 /* 2816 * Set ownership given a path name; this version does not follow links. 2817 */ 2818 /* ARGSUSED */ 2819 int 2820 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2821 { 2822 /* { 2823 syscallarg(const char *) path; 2824 syscallarg(uid_t) uid; 2825 syscallarg(gid_t) gid; 2826 } */ 2827 int error; 2828 struct nameidata nd; 2829 2830 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2831 SCARG(uap, path)); 2832 if ((error = namei(&nd)) != 0) 2833 return (error); 2834 2835 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2836 2837 vrele(nd.ni_vp); 2838 return (error); 2839 } 2840 2841 /* 2842 * Set ownership given a path name; this version does not follow links. 2843 * Provides POSIX/XPG semantics. 2844 */ 2845 /* ARGSUSED */ 2846 int 2847 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2848 { 2849 /* { 2850 syscallarg(const char *) path; 2851 syscallarg(uid_t) uid; 2852 syscallarg(gid_t) gid; 2853 } */ 2854 int error; 2855 struct nameidata nd; 2856 2857 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2858 SCARG(uap, path)); 2859 if ((error = namei(&nd)) != 0) 2860 return (error); 2861 2862 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2863 2864 vrele(nd.ni_vp); 2865 return (error); 2866 } 2867 2868 /* 2869 * Common routine to set ownership given a vnode. 2870 */ 2871 static int 2872 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2873 int posix_semantics) 2874 { 2875 struct vattr vattr; 2876 mode_t newmode; 2877 int error; 2878 2879 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2880 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2881 goto out; 2882 2883 #define CHANGED(x) ((int)(x) != -1) 2884 newmode = vattr.va_mode; 2885 if (posix_semantics) { 2886 /* 2887 * POSIX/XPG semantics: if the caller is not the super-user, 2888 * clear set-user-id and set-group-id bits. Both POSIX and 2889 * the XPG consider the behaviour for calls by the super-user 2890 * implementation-defined; we leave the set-user-id and set- 2891 * group-id settings intact in that case. 2892 */ 2893 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2894 NULL) != 0) 2895 newmode &= ~(S_ISUID | S_ISGID); 2896 } else { 2897 /* 2898 * NetBSD semantics: when changing owner and/or group, 2899 * clear the respective bit(s). 2900 */ 2901 if (CHANGED(uid)) 2902 newmode &= ~S_ISUID; 2903 if (CHANGED(gid)) 2904 newmode &= ~S_ISGID; 2905 } 2906 /* Update va_mode iff altered. */ 2907 if (vattr.va_mode == newmode) 2908 newmode = VNOVAL; 2909 2910 VATTR_NULL(&vattr); 2911 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2912 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2913 vattr.va_mode = newmode; 2914 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2915 #undef CHANGED 2916 2917 out: 2918 VOP_UNLOCK(vp, 0); 2919 return (error); 2920 } 2921 2922 /* 2923 * Set the access and modification times given a path name; this 2924 * version follows links. 2925 */ 2926 /* ARGSUSED */ 2927 int 2928 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2929 { 2930 /* { 2931 syscallarg(const char *) path; 2932 syscallarg(const struct timeval *) tptr; 2933 } */ 2934 2935 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2936 SCARG(uap, tptr), UIO_USERSPACE); 2937 } 2938 2939 /* 2940 * Set the access and modification times given a file descriptor. 2941 */ 2942 /* ARGSUSED */ 2943 int 2944 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2945 { 2946 /* { 2947 syscallarg(int) fd; 2948 syscallarg(const struct timeval *) tptr; 2949 } */ 2950 int error; 2951 struct file *fp; 2952 2953 /* getvnode() will use the descriptor for us */ 2954 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2955 return (error); 2956 2957 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2958 SCARG(uap, tptr), UIO_USERSPACE); 2959 2960 FILE_UNUSE(fp, l); 2961 return (error); 2962 } 2963 2964 /* 2965 * Set the access and modification times given a path name; this 2966 * version does not follow links. 2967 */ 2968 int 2969 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2970 { 2971 /* { 2972 syscallarg(const char *) path; 2973 syscallarg(const struct timeval *) tptr; 2974 } */ 2975 2976 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2977 SCARG(uap, tptr), UIO_USERSPACE); 2978 } 2979 2980 /* 2981 * Common routine to set access and modification times given a vnode. 2982 */ 2983 int 2984 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2985 const struct timeval *tptr, enum uio_seg seg) 2986 { 2987 struct vattr vattr; 2988 struct nameidata nd; 2989 int error; 2990 2991 VATTR_NULL(&vattr); 2992 if (tptr == NULL) { 2993 nanotime(&vattr.va_atime); 2994 vattr.va_mtime = vattr.va_atime; 2995 vattr.va_vaflags |= VA_UTIMES_NULL; 2996 } else { 2997 struct timeval tv[2]; 2998 2999 if (seg != UIO_SYSSPACE) { 3000 error = copyin(tptr, &tv, sizeof (tv)); 3001 if (error != 0) 3002 return error; 3003 tptr = tv; 3004 } 3005 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3006 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3007 } 3008 3009 if (vp == NULL) { 3010 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3011 if ((error = namei(&nd)) != 0) 3012 return (error); 3013 vp = nd.ni_vp; 3014 } else 3015 nd.ni_vp = NULL; 3016 3017 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3018 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3019 VOP_UNLOCK(vp, 0); 3020 3021 if (nd.ni_vp != NULL) 3022 vrele(nd.ni_vp); 3023 3024 return (error); 3025 } 3026 3027 /* 3028 * Truncate a file given its path name. 3029 */ 3030 /* ARGSUSED */ 3031 int 3032 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3033 { 3034 /* { 3035 syscallarg(const char *) path; 3036 syscallarg(int) pad; 3037 syscallarg(off_t) length; 3038 } */ 3039 struct vnode *vp; 3040 struct vattr vattr; 3041 int error; 3042 struct nameidata nd; 3043 3044 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3045 SCARG(uap, path)); 3046 if ((error = namei(&nd)) != 0) 3047 return (error); 3048 vp = nd.ni_vp; 3049 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3050 if (vp->v_type == VDIR) 3051 error = EISDIR; 3052 else if ((error = vn_writechk(vp)) == 0 && 3053 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3054 VATTR_NULL(&vattr); 3055 vattr.va_size = SCARG(uap, length); 3056 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3057 } 3058 vput(vp); 3059 return (error); 3060 } 3061 3062 /* 3063 * Truncate a file given a file descriptor. 3064 */ 3065 /* ARGSUSED */ 3066 int 3067 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3068 { 3069 /* { 3070 syscallarg(int) fd; 3071 syscallarg(int) pad; 3072 syscallarg(off_t) length; 3073 } */ 3074 struct proc *p = l->l_proc; 3075 struct vattr vattr; 3076 struct vnode *vp; 3077 struct file *fp; 3078 int error; 3079 3080 /* getvnode() will use the descriptor for us */ 3081 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3082 return (error); 3083 if ((fp->f_flag & FWRITE) == 0) { 3084 error = EINVAL; 3085 goto out; 3086 } 3087 vp = (struct vnode *)fp->f_data; 3088 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3089 if (vp->v_type == VDIR) 3090 error = EISDIR; 3091 else if ((error = vn_writechk(vp)) == 0) { 3092 VATTR_NULL(&vattr); 3093 vattr.va_size = SCARG(uap, length); 3094 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3095 } 3096 VOP_UNLOCK(vp, 0); 3097 out: 3098 FILE_UNUSE(fp, l); 3099 return (error); 3100 } 3101 3102 /* 3103 * Sync an open file. 3104 */ 3105 /* ARGSUSED */ 3106 int 3107 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3108 { 3109 /* { 3110 syscallarg(int) fd; 3111 } */ 3112 struct proc *p = l->l_proc; 3113 struct vnode *vp; 3114 struct file *fp; 3115 int error; 3116 3117 /* getvnode() will use the descriptor for us */ 3118 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3119 return (error); 3120 vp = (struct vnode *)fp->f_data; 3121 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3122 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3123 if (error == 0 && bioopsp != NULL && 3124 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3125 (*bioopsp->io_fsync)(vp, 0); 3126 VOP_UNLOCK(vp, 0); 3127 FILE_UNUSE(fp, l); 3128 return (error); 3129 } 3130 3131 /* 3132 * Sync a range of file data. API modeled after that found in AIX. 3133 * 3134 * FDATASYNC indicates that we need only save enough metadata to be able 3135 * to re-read the written data. Note we duplicate AIX's requirement that 3136 * the file be open for writing. 3137 */ 3138 /* ARGSUSED */ 3139 int 3140 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3141 { 3142 /* { 3143 syscallarg(int) fd; 3144 syscallarg(int) flags; 3145 syscallarg(off_t) start; 3146 syscallarg(off_t) length; 3147 } */ 3148 struct proc *p = l->l_proc; 3149 struct vnode *vp; 3150 struct file *fp; 3151 int flags, nflags; 3152 off_t s, e, len; 3153 int error; 3154 3155 /* getvnode() will use the descriptor for us */ 3156 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3157 return (error); 3158 3159 if ((fp->f_flag & FWRITE) == 0) { 3160 error = EBADF; 3161 goto out; 3162 } 3163 3164 flags = SCARG(uap, flags); 3165 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3166 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3167 error = EINVAL; 3168 goto out; 3169 } 3170 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3171 if (flags & FDATASYNC) 3172 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3173 else 3174 nflags = FSYNC_WAIT; 3175 if (flags & FDISKSYNC) 3176 nflags |= FSYNC_CACHE; 3177 3178 len = SCARG(uap, length); 3179 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3180 if (len) { 3181 s = SCARG(uap, start); 3182 e = s + len; 3183 if (e < s) { 3184 error = EINVAL; 3185 goto out; 3186 } 3187 } else { 3188 e = 0; 3189 s = 0; 3190 } 3191 3192 vp = (struct vnode *)fp->f_data; 3193 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3194 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3195 3196 if (error == 0 && bioopsp != NULL && 3197 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3198 (*bioopsp->io_fsync)(vp, nflags); 3199 3200 VOP_UNLOCK(vp, 0); 3201 out: 3202 FILE_UNUSE(fp, l); 3203 return (error); 3204 } 3205 3206 /* 3207 * Sync the data of an open file. 3208 */ 3209 /* ARGSUSED */ 3210 int 3211 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3212 { 3213 /* { 3214 syscallarg(int) fd; 3215 } */ 3216 struct proc *p = l->l_proc; 3217 struct vnode *vp; 3218 struct file *fp; 3219 int error; 3220 3221 /* getvnode() will use the descriptor for us */ 3222 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3223 return (error); 3224 if ((fp->f_flag & FWRITE) == 0) { 3225 FILE_UNUSE(fp, l); 3226 return (EBADF); 3227 } 3228 vp = (struct vnode *)fp->f_data; 3229 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3230 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3231 VOP_UNLOCK(vp, 0); 3232 FILE_UNUSE(fp, l); 3233 return (error); 3234 } 3235 3236 /* 3237 * Rename files, (standard) BSD semantics frontend. 3238 */ 3239 /* ARGSUSED */ 3240 int 3241 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3242 { 3243 /* { 3244 syscallarg(const char *) from; 3245 syscallarg(const char *) to; 3246 } */ 3247 3248 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3249 } 3250 3251 /* 3252 * Rename files, POSIX semantics frontend. 3253 */ 3254 /* ARGSUSED */ 3255 int 3256 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3257 { 3258 /* { 3259 syscallarg(const char *) from; 3260 syscallarg(const char *) to; 3261 } */ 3262 3263 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3264 } 3265 3266 /* 3267 * Rename files. Source and destination must either both be directories, 3268 * or both not be directories. If target is a directory, it must be empty. 3269 * If `from' and `to' refer to the same object, the value of the `retain' 3270 * argument is used to determine whether `from' will be 3271 * 3272 * (retain == 0) deleted unless `from' and `to' refer to the same 3273 * object in the file system's name space (BSD). 3274 * (retain == 1) always retained (POSIX). 3275 */ 3276 int 3277 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3278 { 3279 struct vnode *tvp, *fvp, *tdvp; 3280 struct nameidata fromnd, tond; 3281 struct lwp *l = curlwp; 3282 struct proc *p; 3283 int error; 3284 3285 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3286 seg, from); 3287 if ((error = namei(&fromnd)) != 0) 3288 return (error); 3289 if (fromnd.ni_dvp != fromnd.ni_vp) 3290 VOP_UNLOCK(fromnd.ni_dvp, 0); 3291 fvp = fromnd.ni_vp; 3292 NDINIT(&tond, RENAME, 3293 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3294 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3295 seg, to); 3296 if ((error = namei(&tond)) != 0) { 3297 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3298 vrele(fromnd.ni_dvp); 3299 vrele(fvp); 3300 goto out1; 3301 } 3302 tdvp = tond.ni_dvp; 3303 tvp = tond.ni_vp; 3304 3305 if (tvp != NULL) { 3306 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3307 error = ENOTDIR; 3308 goto out; 3309 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3310 error = EISDIR; 3311 goto out; 3312 } 3313 } 3314 3315 if (fvp == tdvp) 3316 error = EINVAL; 3317 3318 /* 3319 * Source and destination refer to the same object. 3320 */ 3321 if (fvp == tvp) { 3322 if (retain) 3323 error = -1; 3324 else if (fromnd.ni_dvp == tdvp && 3325 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3326 !memcmp(fromnd.ni_cnd.cn_nameptr, 3327 tond.ni_cnd.cn_nameptr, 3328 fromnd.ni_cnd.cn_namelen)) 3329 error = -1; 3330 } 3331 3332 #if NVERIEXEC > 0 3333 if (!error) { 3334 char *f1, *f2; 3335 3336 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3337 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3338 3339 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3340 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3341 3342 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3343 3344 free(f1, M_TEMP); 3345 free(f2, M_TEMP); 3346 } 3347 #endif /* NVERIEXEC > 0 */ 3348 3349 out: 3350 p = l->l_proc; 3351 if (!error) { 3352 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3353 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3354 } else { 3355 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3356 if (tdvp == tvp) 3357 vrele(tdvp); 3358 else 3359 vput(tdvp); 3360 if (tvp) 3361 vput(tvp); 3362 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3363 vrele(fromnd.ni_dvp); 3364 vrele(fvp); 3365 } 3366 vrele(tond.ni_startdir); 3367 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3368 out1: 3369 if (fromnd.ni_startdir) 3370 vrele(fromnd.ni_startdir); 3371 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3372 return (error == -1 ? 0 : error); 3373 } 3374 3375 /* 3376 * Make a directory file. 3377 */ 3378 /* ARGSUSED */ 3379 int 3380 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3381 { 3382 /* { 3383 syscallarg(const char *) path; 3384 syscallarg(int) mode; 3385 } */ 3386 struct proc *p = l->l_proc; 3387 struct vnode *vp; 3388 struct vattr vattr; 3389 int error; 3390 struct nameidata nd; 3391 3392 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3393 SCARG(uap, path)); 3394 if ((error = namei(&nd)) != 0) 3395 return (error); 3396 vp = nd.ni_vp; 3397 if (vp != NULL) { 3398 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3399 if (nd.ni_dvp == vp) 3400 vrele(nd.ni_dvp); 3401 else 3402 vput(nd.ni_dvp); 3403 vrele(vp); 3404 return (EEXIST); 3405 } 3406 VATTR_NULL(&vattr); 3407 vattr.va_type = VDIR; 3408 /* We will read cwdi->cwdi_cmask unlocked. */ 3409 vattr.va_mode = 3410 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3411 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3412 if (!error) 3413 vput(nd.ni_vp); 3414 return (error); 3415 } 3416 3417 /* 3418 * Remove a directory file. 3419 */ 3420 /* ARGSUSED */ 3421 int 3422 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3423 { 3424 /* { 3425 syscallarg(const char *) path; 3426 } */ 3427 struct vnode *vp; 3428 int error; 3429 struct nameidata nd; 3430 3431 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3432 SCARG(uap, path)); 3433 if ((error = namei(&nd)) != 0) 3434 return (error); 3435 vp = nd.ni_vp; 3436 if (vp->v_type != VDIR) { 3437 error = ENOTDIR; 3438 goto out; 3439 } 3440 /* 3441 * No rmdir "." please. 3442 */ 3443 if (nd.ni_dvp == vp) { 3444 error = EINVAL; 3445 goto out; 3446 } 3447 /* 3448 * The root of a mounted filesystem cannot be deleted. 3449 */ 3450 if (vp->v_vflag & VV_ROOT) { 3451 error = EBUSY; 3452 goto out; 3453 } 3454 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3455 return (error); 3456 3457 out: 3458 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3459 if (nd.ni_dvp == vp) 3460 vrele(nd.ni_dvp); 3461 else 3462 vput(nd.ni_dvp); 3463 vput(vp); 3464 return (error); 3465 } 3466 3467 /* 3468 * Read a block of directory entries in a file system independent format. 3469 */ 3470 int 3471 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3472 { 3473 /* { 3474 syscallarg(int) fd; 3475 syscallarg(char *) buf; 3476 syscallarg(size_t) count; 3477 } */ 3478 struct proc *p = l->l_proc; 3479 struct file *fp; 3480 int error, done; 3481 3482 /* getvnode() will use the descriptor for us */ 3483 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3484 return (error); 3485 if ((fp->f_flag & FREAD) == 0) { 3486 error = EBADF; 3487 goto out; 3488 } 3489 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3490 SCARG(uap, count), &done, l, 0, 0); 3491 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3492 *retval = done; 3493 out: 3494 FILE_UNUSE(fp, l); 3495 return (error); 3496 } 3497 3498 /* 3499 * Set the mode mask for creation of filesystem nodes. 3500 */ 3501 int 3502 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3503 { 3504 /* { 3505 syscallarg(mode_t) newmask; 3506 } */ 3507 struct proc *p = l->l_proc; 3508 struct cwdinfo *cwdi; 3509 3510 /* 3511 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3512 * important is that we serialize changes to the mask. The 3513 * rw_exit() will issue a write memory barrier on our behalf, 3514 * and force the changes out to other CPUs (as it must use an 3515 * atomic operation, draining the local CPU's store buffers). 3516 */ 3517 cwdi = p->p_cwdi; 3518 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3519 *retval = cwdi->cwdi_cmask; 3520 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3521 rw_exit(&cwdi->cwdi_lock); 3522 3523 return (0); 3524 } 3525 3526 int 3527 dorevoke(struct vnode *vp, kauth_cred_t cred) 3528 { 3529 struct vattr vattr; 3530 int error; 3531 3532 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3533 return error; 3534 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3535 (error = kauth_authorize_generic(cred, 3536 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3537 VOP_REVOKE(vp, REVOKEALL); 3538 return (error); 3539 } 3540 3541 /* 3542 * Void all references to file by ripping underlying filesystem 3543 * away from vnode. 3544 */ 3545 /* ARGSUSED */ 3546 int 3547 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3548 { 3549 /* { 3550 syscallarg(const char *) path; 3551 } */ 3552 struct vnode *vp; 3553 int error; 3554 struct nameidata nd; 3555 3556 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3557 SCARG(uap, path)); 3558 if ((error = namei(&nd)) != 0) 3559 return (error); 3560 vp = nd.ni_vp; 3561 error = dorevoke(vp, l->l_cred); 3562 vrele(vp); 3563 return (error); 3564 } 3565 3566 /* 3567 * Convert a user file descriptor to a kernel file entry. 3568 */ 3569 int 3570 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3571 { 3572 struct vnode *vp; 3573 struct file *fp; 3574 3575 if ((fp = fd_getfile(fdp, fd)) == NULL) 3576 return (EBADF); 3577 3578 FILE_USE(fp); 3579 3580 if (fp->f_type != DTYPE_VNODE) { 3581 FILE_UNUSE(fp, NULL); 3582 return (EINVAL); 3583 } 3584 3585 vp = (struct vnode *)fp->f_data; 3586 if (vp->v_type == VBAD) { 3587 FILE_UNUSE(fp, NULL); 3588 return (EBADF); 3589 } 3590 3591 *fpp = fp; 3592 return (0); 3593 } 3594
Indexes created Fri Oct 03 02:09:57 GMT 2025