vfs_syscalls.c revision 1.342
1 /* $NetBSD: vfs_syscalls.c,v 1.342 2008/01/24 17:32:55 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 1.342 2008/01/24 17:32:55 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 VOP_LEASE(vp, cred, LEASE_WRITE); 1632 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1633 VATTR_NULL(&va); 1634 va.va_size = 0; 1635 error = VOP_SETATTR(vp, &va, cred); 1636 if (error) 1637 goto bad; 1638 } 1639 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1640 goto bad; 1641 if (flags & FWRITE) 1642 vp->v_writecount++; 1643 1644 /* done with modified vn_open, now finish what sys_open does. */ 1645 1646 fp->f_flag = flags & FMASK; 1647 fp->f_type = DTYPE_VNODE; 1648 fp->f_ops = &vnops; 1649 fp->f_data = vp; 1650 if (flags & (O_EXLOCK | O_SHLOCK)) { 1651 lf.l_whence = SEEK_SET; 1652 lf.l_start = 0; 1653 lf.l_len = 0; 1654 if (flags & O_EXLOCK) 1655 lf.l_type = F_WRLCK; 1656 else 1657 lf.l_type = F_RDLCK; 1658 type = F_FLOCK; 1659 if ((flags & FNONBLOCK) == 0) 1660 type |= F_WAIT; 1661 VOP_UNLOCK(vp, 0); 1662 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1663 if (error) { 1664 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1665 FILE_UNUSE(fp, l); 1666 ffree(fp); 1667 fdremove(fdp, indx); 1668 return (error); 1669 } 1670 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1671 fp->f_flag |= FHASLOCK; 1672 } 1673 VOP_UNLOCK(vp, 0); 1674 *retval = indx; 1675 FILE_SET_MATURE(fp); 1676 FILE_UNUSE(fp, l); 1677 vfs_copyinfh_free(fh); 1678 return (0); 1679 1680 bad: 1681 FILE_UNUSE(fp, l); 1682 ffree(fp); 1683 fdremove(fdp, indx); 1684 if (vp != NULL) 1685 vput(vp); 1686 vfs_copyinfh_free(fh); 1687 return (error); 1688 } 1689 1690 int 1691 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1692 { 1693 /* { 1694 syscallarg(const void *) fhp; 1695 syscallarg(size_t) fh_size; 1696 syscallarg(int) flags; 1697 } */ 1698 1699 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1700 SCARG(uap, flags), retval); 1701 } 1702 1703 int 1704 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1705 { 1706 int error; 1707 fhandle_t *fh; 1708 struct vnode *vp; 1709 1710 /* 1711 * Must be super user 1712 */ 1713 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1714 0, NULL, NULL, NULL))) 1715 return (error); 1716 1717 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1718 if (error != 0) 1719 return error; 1720 1721 error = vfs_fhtovp(fh, &vp); 1722 vfs_copyinfh_free(fh); 1723 if (error != 0) 1724 return error; 1725 1726 error = vn_stat(vp, sb, l); 1727 vput(vp); 1728 return error; 1729 } 1730 1731 1732 /* ARGSUSED */ 1733 int 1734 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1735 { 1736 /* { 1737 syscallarg(const void *) fhp; 1738 syscallarg(size_t) fh_size; 1739 syscallarg(struct stat *) sb; 1740 } */ 1741 struct stat sb; 1742 int error; 1743 1744 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1745 if (error) 1746 return error; 1747 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1748 } 1749 1750 int 1751 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1752 int flags) 1753 { 1754 fhandle_t *fh; 1755 struct mount *mp; 1756 struct vnode *vp; 1757 int error; 1758 1759 /* 1760 * Must be super user 1761 */ 1762 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1763 0, NULL, NULL, NULL))) 1764 return error; 1765 1766 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1767 if (error != 0) 1768 return error; 1769 1770 error = vfs_fhtovp(fh, &vp); 1771 vfs_copyinfh_free(fh); 1772 if (error != 0) 1773 return error; 1774 1775 mp = vp->v_mount; 1776 error = dostatvfs(mp, sb, l, flags, 1); 1777 vput(vp); 1778 return error; 1779 } 1780 1781 /* ARGSUSED */ 1782 int 1783 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1784 { 1785 /* { 1786 syscallarg(const void *) fhp; 1787 syscallarg(size_t) fh_size; 1788 syscallarg(struct statvfs *) buf; 1789 syscallarg(int) flags; 1790 } */ 1791 struct statvfs *sb = STATVFSBUF_GET(); 1792 int error; 1793 1794 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1795 SCARG(uap, flags)); 1796 if (error == 0) 1797 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1798 STATVFSBUF_PUT(sb); 1799 return error; 1800 } 1801 1802 /* 1803 * Create a special file. 1804 */ 1805 /* ARGSUSED */ 1806 int 1807 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1808 { 1809 /* { 1810 syscallarg(const char *) path; 1811 syscallarg(int) mode; 1812 syscallarg(int) dev; 1813 } */ 1814 struct proc *p = l->l_proc; 1815 struct vnode *vp; 1816 struct vattr vattr; 1817 int error, optype; 1818 struct nameidata nd; 1819 char *path; 1820 const char *cpath; 1821 enum uio_seg seg = UIO_USERSPACE; 1822 1823 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1824 0, NULL, NULL, NULL)) != 0) 1825 return (error); 1826 1827 optype = VOP_MKNOD_DESCOFFSET; 1828 1829 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1830 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1831 1832 if ((error = namei(&nd)) != 0) 1833 goto out; 1834 vp = nd.ni_vp; 1835 if (vp != NULL) 1836 error = EEXIST; 1837 else { 1838 VATTR_NULL(&vattr); 1839 /* We will read cwdi->cwdi_cmask unlocked. */ 1840 vattr.va_mode = 1841 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1842 vattr.va_rdev = SCARG(uap, dev); 1843 1844 switch (SCARG(uap, mode) & S_IFMT) { 1845 case S_IFMT: /* used by badsect to flag bad sectors */ 1846 vattr.va_type = VBAD; 1847 break; 1848 case S_IFCHR: 1849 vattr.va_type = VCHR; 1850 break; 1851 case S_IFBLK: 1852 vattr.va_type = VBLK; 1853 break; 1854 case S_IFWHT: 1855 optype = VOP_WHITEOUT_DESCOFFSET; 1856 break; 1857 case S_IFREG: 1858 #if NVERIEXEC > 0 1859 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1860 O_CREAT); 1861 #endif /* NVERIEXEC > 0 */ 1862 vattr.va_type = VREG; 1863 vattr.va_rdev = VNOVAL; 1864 optype = VOP_CREATE_DESCOFFSET; 1865 break; 1866 default: 1867 error = EINVAL; 1868 break; 1869 } 1870 } 1871 if (!error) { 1872 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1873 switch (optype) { 1874 case VOP_WHITEOUT_DESCOFFSET: 1875 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1876 if (error) 1877 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1878 vput(nd.ni_dvp); 1879 break; 1880 1881 case VOP_MKNOD_DESCOFFSET: 1882 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1883 &nd.ni_cnd, &vattr); 1884 if (error == 0) 1885 vput(nd.ni_vp); 1886 break; 1887 1888 case VOP_CREATE_DESCOFFSET: 1889 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1890 &nd.ni_cnd, &vattr); 1891 if (error == 0) 1892 vput(nd.ni_vp); 1893 break; 1894 } 1895 } else { 1896 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1897 if (nd.ni_dvp == vp) 1898 vrele(nd.ni_dvp); 1899 else 1900 vput(nd.ni_dvp); 1901 if (vp) 1902 vrele(vp); 1903 } 1904 out: 1905 VERIEXEC_PATH_PUT(path); 1906 return (error); 1907 } 1908 1909 /* 1910 * Create a named pipe. 1911 */ 1912 /* ARGSUSED */ 1913 int 1914 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1915 { 1916 /* { 1917 syscallarg(const char *) path; 1918 syscallarg(int) mode; 1919 } */ 1920 struct proc *p = l->l_proc; 1921 struct vattr vattr; 1922 int error; 1923 struct nameidata nd; 1924 1925 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1926 SCARG(uap, path)); 1927 if ((error = namei(&nd)) != 0) 1928 return (error); 1929 if (nd.ni_vp != NULL) { 1930 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1931 if (nd.ni_dvp == nd.ni_vp) 1932 vrele(nd.ni_dvp); 1933 else 1934 vput(nd.ni_dvp); 1935 vrele(nd.ni_vp); 1936 return (EEXIST); 1937 } 1938 VATTR_NULL(&vattr); 1939 vattr.va_type = VFIFO; 1940 /* We will read cwdi->cwdi_cmask unlocked. */ 1941 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1942 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1943 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1944 if (error == 0) 1945 vput(nd.ni_vp); 1946 return (error); 1947 } 1948 1949 /* 1950 * Make a hard file link. 1951 */ 1952 /* ARGSUSED */ 1953 int 1954 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1955 { 1956 /* { 1957 syscallarg(const char *) path; 1958 syscallarg(const char *) link; 1959 } */ 1960 struct vnode *vp; 1961 struct nameidata nd; 1962 int error; 1963 1964 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1965 SCARG(uap, path)); 1966 if ((error = namei(&nd)) != 0) 1967 return (error); 1968 vp = nd.ni_vp; 1969 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1970 SCARG(uap, link)); 1971 if ((error = namei(&nd)) != 0) 1972 goto out; 1973 if (nd.ni_vp) { 1974 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1975 if (nd.ni_dvp == nd.ni_vp) 1976 vrele(nd.ni_dvp); 1977 else 1978 vput(nd.ni_dvp); 1979 vrele(nd.ni_vp); 1980 error = EEXIST; 1981 goto out; 1982 } 1983 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1984 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 1985 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1986 out: 1987 vrele(vp); 1988 return (error); 1989 } 1990 1991 /* 1992 * Make a symbolic link. 1993 */ 1994 /* ARGSUSED */ 1995 int 1996 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1997 { 1998 /* { 1999 syscallarg(const char *) path; 2000 syscallarg(const char *) link; 2001 } */ 2002 struct proc *p = l->l_proc; 2003 struct vattr vattr; 2004 char *path; 2005 int error; 2006 struct nameidata nd; 2007 2008 path = PNBUF_GET(); 2009 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2010 if (error) 2011 goto out; 2012 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2013 SCARG(uap, link)); 2014 if ((error = namei(&nd)) != 0) 2015 goto out; 2016 if (nd.ni_vp) { 2017 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2018 if (nd.ni_dvp == nd.ni_vp) 2019 vrele(nd.ni_dvp); 2020 else 2021 vput(nd.ni_dvp); 2022 vrele(nd.ni_vp); 2023 error = EEXIST; 2024 goto out; 2025 } 2026 VATTR_NULL(&vattr); 2027 vattr.va_type = VLNK; 2028 /* We will read cwdi->cwdi_cmask unlocked. */ 2029 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2030 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2031 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2032 if (error == 0) 2033 vput(nd.ni_vp); 2034 out: 2035 PNBUF_PUT(path); 2036 return (error); 2037 } 2038 2039 /* 2040 * Delete a whiteout from the filesystem. 2041 */ 2042 /* ARGSUSED */ 2043 int 2044 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2045 { 2046 /* { 2047 syscallarg(const char *) path; 2048 } */ 2049 int error; 2050 struct nameidata nd; 2051 2052 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2053 UIO_USERSPACE, SCARG(uap, path)); 2054 error = namei(&nd); 2055 if (error) 2056 return (error); 2057 2058 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2059 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2060 if (nd.ni_dvp == nd.ni_vp) 2061 vrele(nd.ni_dvp); 2062 else 2063 vput(nd.ni_dvp); 2064 if (nd.ni_vp) 2065 vrele(nd.ni_vp); 2066 return (EEXIST); 2067 } 2068 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2069 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2070 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2071 vput(nd.ni_dvp); 2072 return (error); 2073 } 2074 2075 /* 2076 * Delete a name from the filesystem. 2077 */ 2078 /* ARGSUSED */ 2079 int 2080 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2081 { 2082 /* { 2083 syscallarg(const char *) path; 2084 } */ 2085 2086 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2087 } 2088 2089 int 2090 do_sys_unlink(const char *arg, enum uio_seg seg) 2091 { 2092 struct vnode *vp; 2093 int error; 2094 struct nameidata nd; 2095 kauth_cred_t cred; 2096 char *path; 2097 const char *cpath; 2098 2099 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2100 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2101 2102 if ((error = namei(&nd)) != 0) 2103 goto out; 2104 vp = nd.ni_vp; 2105 2106 /* 2107 * The root of a mounted filesystem cannot be deleted. 2108 */ 2109 if (vp->v_vflag & VV_ROOT) { 2110 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2111 if (nd.ni_dvp == vp) 2112 vrele(nd.ni_dvp); 2113 else 2114 vput(nd.ni_dvp); 2115 vput(vp); 2116 error = EBUSY; 2117 goto out; 2118 } 2119 2120 #if NVERIEXEC > 0 2121 /* Handle remove requests for veriexec entries. */ 2122 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2123 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2124 if (nd.ni_dvp == vp) 2125 vrele(nd.ni_dvp); 2126 else 2127 vput(nd.ni_dvp); 2128 vput(vp); 2129 goto out; 2130 } 2131 #endif /* NVERIEXEC > 0 */ 2132 2133 cred = kauth_cred_get(); 2134 VOP_LEASE(nd.ni_dvp, cred, LEASE_WRITE); 2135 VOP_LEASE(vp, cred, LEASE_WRITE); 2136 #ifdef FILEASSOC 2137 (void)fileassoc_file_delete(vp); 2138 #endif /* FILEASSOC */ 2139 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2140 out: 2141 VERIEXEC_PATH_PUT(path); 2142 return (error); 2143 } 2144 2145 /* 2146 * Reposition read/write file offset. 2147 */ 2148 int 2149 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2150 { 2151 /* { 2152 syscallarg(int) fd; 2153 syscallarg(int) pad; 2154 syscallarg(off_t) offset; 2155 syscallarg(int) whence; 2156 } */ 2157 struct proc *p = l->l_proc; 2158 kauth_cred_t cred = l->l_cred; 2159 struct filedesc *fdp = p->p_fd; 2160 struct file *fp; 2161 struct vnode *vp; 2162 struct vattr vattr; 2163 off_t newoff; 2164 int error; 2165 2166 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2167 return (EBADF); 2168 2169 vp = (struct vnode *)fp->f_data; 2170 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2171 error = ESPIPE; 2172 FILE_UNLOCK(fp); 2173 goto out; 2174 } 2175 2176 switch (SCARG(uap, whence)) { 2177 case SEEK_CUR: 2178 newoff = fp->f_offset + SCARG(uap, offset); 2179 FILE_USE(fp); 2180 break; 2181 case SEEK_END: 2182 FILE_USE(fp); 2183 error = VOP_GETATTR(vp, &vattr, cred); 2184 if (error) { 2185 FILE_UNUSE(fp, l); 2186 goto out; 2187 } 2188 newoff = SCARG(uap, offset) + vattr.va_size; 2189 break; 2190 case SEEK_SET: 2191 FILE_USE(fp); 2192 newoff = SCARG(uap, offset); 2193 break; 2194 default: 2195 FILE_UNLOCK(fp); 2196 error = EINVAL; 2197 goto out; 2198 } 2199 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2200 FILE_LOCK(fp); 2201 *(off_t *)retval = fp->f_offset = newoff; 2202 FILE_UNLOCK(fp); 2203 } 2204 FILE_UNUSE(fp, l); 2205 out: 2206 return (error); 2207 } 2208 2209 /* 2210 * Positional read system call. 2211 */ 2212 int 2213 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2214 { 2215 /* { 2216 syscallarg(int) fd; 2217 syscallarg(void *) buf; 2218 syscallarg(size_t) nbyte; 2219 syscallarg(off_t) offset; 2220 } */ 2221 struct proc *p = l->l_proc; 2222 struct filedesc *fdp = p->p_fd; 2223 struct file *fp; 2224 struct vnode *vp; 2225 off_t offset; 2226 int error, fd = SCARG(uap, fd); 2227 2228 if ((fp = fd_getfile(fdp, fd)) == NULL) 2229 return (EBADF); 2230 2231 if ((fp->f_flag & FREAD) == 0) { 2232 FILE_UNLOCK(fp); 2233 return (EBADF); 2234 } 2235 2236 FILE_USE(fp); 2237 2238 vp = (struct vnode *)fp->f_data; 2239 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2240 error = ESPIPE; 2241 goto out; 2242 } 2243 2244 offset = SCARG(uap, offset); 2245 2246 /* 2247 * XXX This works because no file systems actually 2248 * XXX take any action on the seek operation. 2249 */ 2250 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2251 goto out; 2252 2253 /* dofileread() will unuse the descriptor for us */ 2254 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2255 &offset, 0, retval)); 2256 2257 out: 2258 FILE_UNUSE(fp, l); 2259 return (error); 2260 } 2261 2262 /* 2263 * Positional scatter read system call. 2264 */ 2265 int 2266 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2267 { 2268 /* { 2269 syscallarg(int) fd; 2270 syscallarg(const struct iovec *) iovp; 2271 syscallarg(int) iovcnt; 2272 syscallarg(off_t) offset; 2273 } */ 2274 off_t offset = SCARG(uap, offset); 2275 2276 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2277 SCARG(uap, iovcnt), &offset, 0, retval); 2278 } 2279 2280 /* 2281 * Positional write system call. 2282 */ 2283 int 2284 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2285 { 2286 /* { 2287 syscallarg(int) fd; 2288 syscallarg(const void *) buf; 2289 syscallarg(size_t) nbyte; 2290 syscallarg(off_t) offset; 2291 } */ 2292 struct proc *p = l->l_proc; 2293 struct filedesc *fdp = p->p_fd; 2294 struct file *fp; 2295 struct vnode *vp; 2296 off_t offset; 2297 int error, fd = SCARG(uap, fd); 2298 2299 if ((fp = fd_getfile(fdp, fd)) == NULL) 2300 return (EBADF); 2301 2302 if ((fp->f_flag & FWRITE) == 0) { 2303 FILE_UNLOCK(fp); 2304 return (EBADF); 2305 } 2306 2307 FILE_USE(fp); 2308 2309 vp = (struct vnode *)fp->f_data; 2310 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2311 error = ESPIPE; 2312 goto out; 2313 } 2314 2315 offset = SCARG(uap, offset); 2316 2317 /* 2318 * XXX This works because no file systems actually 2319 * XXX take any action on the seek operation. 2320 */ 2321 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2322 goto out; 2323 2324 /* dofilewrite() will unuse the descriptor for us */ 2325 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2326 &offset, 0, retval)); 2327 2328 out: 2329 FILE_UNUSE(fp, l); 2330 return (error); 2331 } 2332 2333 /* 2334 * Positional gather write system call. 2335 */ 2336 int 2337 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2338 { 2339 /* { 2340 syscallarg(int) fd; 2341 syscallarg(const struct iovec *) iovp; 2342 syscallarg(int) iovcnt; 2343 syscallarg(off_t) offset; 2344 } */ 2345 off_t offset = SCARG(uap, offset); 2346 2347 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2348 SCARG(uap, iovcnt), &offset, 0, retval); 2349 } 2350 2351 /* 2352 * Check access permissions. 2353 */ 2354 int 2355 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2356 { 2357 /* { 2358 syscallarg(const char *) path; 2359 syscallarg(int) flags; 2360 } */ 2361 kauth_cred_t cred; 2362 struct vnode *vp; 2363 int error, flags; 2364 struct nameidata nd; 2365 2366 cred = kauth_cred_dup(l->l_cred); 2367 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2368 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2369 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2370 SCARG(uap, path)); 2371 /* Override default credentials */ 2372 nd.ni_cnd.cn_cred = cred; 2373 if ((error = namei(&nd)) != 0) 2374 goto out; 2375 vp = nd.ni_vp; 2376 2377 /* Flags == 0 means only check for existence. */ 2378 if (SCARG(uap, flags)) { 2379 flags = 0; 2380 if (SCARG(uap, flags) & R_OK) 2381 flags |= VREAD; 2382 if (SCARG(uap, flags) & W_OK) 2383 flags |= VWRITE; 2384 if (SCARG(uap, flags) & X_OK) 2385 flags |= VEXEC; 2386 2387 error = VOP_ACCESS(vp, flags, cred); 2388 if (!error && (flags & VWRITE)) 2389 error = vn_writechk(vp); 2390 } 2391 vput(vp); 2392 out: 2393 kauth_cred_free(cred); 2394 return (error); 2395 } 2396 2397 /* 2398 * Common code for all sys_stat functions, including compat versions. 2399 */ 2400 int 2401 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2402 struct stat *sb) 2403 { 2404 int error; 2405 struct nameidata nd; 2406 2407 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2408 UIO_USERSPACE, path); 2409 error = namei(&nd); 2410 if (error != 0) 2411 return error; 2412 error = vn_stat(nd.ni_vp, sb, l); 2413 vput(nd.ni_vp); 2414 return error; 2415 } 2416 2417 /* 2418 * Get file status; this version follows links. 2419 */ 2420 /* ARGSUSED */ 2421 int 2422 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2423 { 2424 /* { 2425 syscallarg(const char *) path; 2426 syscallarg(struct stat *) ub; 2427 } */ 2428 struct stat sb; 2429 int error; 2430 2431 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2432 if (error) 2433 return error; 2434 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2435 } 2436 2437 /* 2438 * Get file status; this version does not follow links. 2439 */ 2440 /* ARGSUSED */ 2441 int 2442 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2443 { 2444 /* { 2445 syscallarg(const char *) path; 2446 syscallarg(struct stat *) ub; 2447 } */ 2448 struct stat sb; 2449 int error; 2450 2451 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2452 if (error) 2453 return error; 2454 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2455 } 2456 2457 /* 2458 * Get configurable pathname variables. 2459 */ 2460 /* ARGSUSED */ 2461 int 2462 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2463 { 2464 /* { 2465 syscallarg(const char *) path; 2466 syscallarg(int) name; 2467 } */ 2468 int error; 2469 struct nameidata nd; 2470 2471 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2472 SCARG(uap, path)); 2473 if ((error = namei(&nd)) != 0) 2474 return (error); 2475 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2476 vput(nd.ni_vp); 2477 return (error); 2478 } 2479 2480 /* 2481 * Return target name of a symbolic link. 2482 */ 2483 /* ARGSUSED */ 2484 int 2485 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2486 { 2487 /* { 2488 syscallarg(const char *) path; 2489 syscallarg(char *) buf; 2490 syscallarg(size_t) count; 2491 } */ 2492 struct vnode *vp; 2493 struct iovec aiov; 2494 struct uio auio; 2495 int error; 2496 struct nameidata nd; 2497 2498 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2499 SCARG(uap, path)); 2500 if ((error = namei(&nd)) != 0) 2501 return (error); 2502 vp = nd.ni_vp; 2503 if (vp->v_type != VLNK) 2504 error = EINVAL; 2505 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2506 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2507 aiov.iov_base = SCARG(uap, buf); 2508 aiov.iov_len = SCARG(uap, count); 2509 auio.uio_iov = &aiov; 2510 auio.uio_iovcnt = 1; 2511 auio.uio_offset = 0; 2512 auio.uio_rw = UIO_READ; 2513 KASSERT(l == curlwp); 2514 auio.uio_vmspace = l->l_proc->p_vmspace; 2515 auio.uio_resid = SCARG(uap, count); 2516 error = VOP_READLINK(vp, &auio, l->l_cred); 2517 } 2518 vput(vp); 2519 *retval = SCARG(uap, count) - auio.uio_resid; 2520 return (error); 2521 } 2522 2523 /* 2524 * Change flags of a file given a path name. 2525 */ 2526 /* ARGSUSED */ 2527 int 2528 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2529 { 2530 /* { 2531 syscallarg(const char *) path; 2532 syscallarg(u_long) flags; 2533 } */ 2534 struct vnode *vp; 2535 int error; 2536 struct nameidata nd; 2537 2538 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2539 SCARG(uap, path)); 2540 if ((error = namei(&nd)) != 0) 2541 return (error); 2542 vp = nd.ni_vp; 2543 error = change_flags(vp, SCARG(uap, flags), l); 2544 vput(vp); 2545 return (error); 2546 } 2547 2548 /* 2549 * Change flags of a file given a file descriptor. 2550 */ 2551 /* ARGSUSED */ 2552 int 2553 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2554 { 2555 /* { 2556 syscallarg(int) fd; 2557 syscallarg(u_long) flags; 2558 } */ 2559 struct proc *p = l->l_proc; 2560 struct vnode *vp; 2561 struct file *fp; 2562 int error; 2563 2564 /* getvnode() will use the descriptor for us */ 2565 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2566 return (error); 2567 vp = (struct vnode *)fp->f_data; 2568 error = change_flags(vp, SCARG(uap, flags), l); 2569 VOP_UNLOCK(vp, 0); 2570 FILE_UNUSE(fp, l); 2571 return (error); 2572 } 2573 2574 /* 2575 * Change flags of a file given a path name; this version does 2576 * not follow links. 2577 */ 2578 int 2579 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2580 { 2581 /* { 2582 syscallarg(const char *) path; 2583 syscallarg(u_long) flags; 2584 } */ 2585 struct vnode *vp; 2586 int error; 2587 struct nameidata nd; 2588 2589 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2590 SCARG(uap, path)); 2591 if ((error = namei(&nd)) != 0) 2592 return (error); 2593 vp = nd.ni_vp; 2594 error = change_flags(vp, SCARG(uap, flags), l); 2595 vput(vp); 2596 return (error); 2597 } 2598 2599 /* 2600 * Common routine to change flags of a file. 2601 */ 2602 int 2603 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2604 { 2605 struct vattr vattr; 2606 int error; 2607 2608 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2609 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2610 /* 2611 * Non-superusers cannot change the flags on devices, even if they 2612 * own them. 2613 */ 2614 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2615 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2616 goto out; 2617 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2618 error = EINVAL; 2619 goto out; 2620 } 2621 } 2622 VATTR_NULL(&vattr); 2623 vattr.va_flags = flags; 2624 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2625 out: 2626 return (error); 2627 } 2628 2629 /* 2630 * Change mode of a file given path name; this version follows links. 2631 */ 2632 /* ARGSUSED */ 2633 int 2634 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2635 { 2636 /* { 2637 syscallarg(const char *) path; 2638 syscallarg(int) mode; 2639 } */ 2640 int error; 2641 struct nameidata nd; 2642 2643 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2644 SCARG(uap, path)); 2645 if ((error = namei(&nd)) != 0) 2646 return (error); 2647 2648 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2649 2650 vrele(nd.ni_vp); 2651 return (error); 2652 } 2653 2654 /* 2655 * Change mode of a file given a file descriptor. 2656 */ 2657 /* ARGSUSED */ 2658 int 2659 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2660 { 2661 /* { 2662 syscallarg(int) fd; 2663 syscallarg(int) mode; 2664 } */ 2665 struct proc *p = l->l_proc; 2666 struct file *fp; 2667 int error; 2668 2669 /* getvnode() will use the descriptor for us */ 2670 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2671 return (error); 2672 2673 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2674 FILE_UNUSE(fp, l); 2675 return (error); 2676 } 2677 2678 /* 2679 * Change mode of a file given path name; this version does not follow links. 2680 */ 2681 /* ARGSUSED */ 2682 int 2683 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2684 { 2685 /* { 2686 syscallarg(const char *) path; 2687 syscallarg(int) mode; 2688 } */ 2689 int error; 2690 struct nameidata nd; 2691 2692 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2693 SCARG(uap, path)); 2694 if ((error = namei(&nd)) != 0) 2695 return (error); 2696 2697 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2698 2699 vrele(nd.ni_vp); 2700 return (error); 2701 } 2702 2703 /* 2704 * Common routine to set mode given a vnode. 2705 */ 2706 static int 2707 change_mode(struct vnode *vp, int mode, struct lwp *l) 2708 { 2709 struct vattr vattr; 2710 int error; 2711 2712 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2713 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2714 VATTR_NULL(&vattr); 2715 vattr.va_mode = mode & ALLPERMS; 2716 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2717 VOP_UNLOCK(vp, 0); 2718 return (error); 2719 } 2720 2721 /* 2722 * Set ownership given a path name; this version follows links. 2723 */ 2724 /* ARGSUSED */ 2725 int 2726 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2727 { 2728 /* { 2729 syscallarg(const char *) path; 2730 syscallarg(uid_t) uid; 2731 syscallarg(gid_t) gid; 2732 } */ 2733 int error; 2734 struct nameidata nd; 2735 2736 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2737 SCARG(uap, path)); 2738 if ((error = namei(&nd)) != 0) 2739 return (error); 2740 2741 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2742 2743 vrele(nd.ni_vp); 2744 return (error); 2745 } 2746 2747 /* 2748 * Set ownership given a path name; this version follows links. 2749 * Provides POSIX semantics. 2750 */ 2751 /* ARGSUSED */ 2752 int 2753 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2754 { 2755 /* { 2756 syscallarg(const char *) path; 2757 syscallarg(uid_t) uid; 2758 syscallarg(gid_t) gid; 2759 } */ 2760 int error; 2761 struct nameidata nd; 2762 2763 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2764 SCARG(uap, path)); 2765 if ((error = namei(&nd)) != 0) 2766 return (error); 2767 2768 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2769 2770 vrele(nd.ni_vp); 2771 return (error); 2772 } 2773 2774 /* 2775 * Set ownership given a file descriptor. 2776 */ 2777 /* ARGSUSED */ 2778 int 2779 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2780 { 2781 /* { 2782 syscallarg(int) fd; 2783 syscallarg(uid_t) uid; 2784 syscallarg(gid_t) gid; 2785 } */ 2786 struct proc *p = l->l_proc; 2787 int error; 2788 struct file *fp; 2789 2790 /* getvnode() will use the descriptor for us */ 2791 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2792 return (error); 2793 2794 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2795 SCARG(uap, gid), l, 0); 2796 FILE_UNUSE(fp, l); 2797 return (error); 2798 } 2799 2800 /* 2801 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2802 */ 2803 /* ARGSUSED */ 2804 int 2805 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2806 { 2807 /* { 2808 syscallarg(int) fd; 2809 syscallarg(uid_t) uid; 2810 syscallarg(gid_t) gid; 2811 } */ 2812 struct proc *p = l->l_proc; 2813 int error; 2814 struct file *fp; 2815 2816 /* getvnode() will use the descriptor for us */ 2817 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2818 return (error); 2819 2820 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2821 SCARG(uap, gid), l, 1); 2822 FILE_UNUSE(fp, l); 2823 return (error); 2824 } 2825 2826 /* 2827 * Set ownership given a path name; this version does not follow links. 2828 */ 2829 /* ARGSUSED */ 2830 int 2831 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2832 { 2833 /* { 2834 syscallarg(const char *) path; 2835 syscallarg(uid_t) uid; 2836 syscallarg(gid_t) gid; 2837 } */ 2838 int error; 2839 struct nameidata nd; 2840 2841 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2842 SCARG(uap, path)); 2843 if ((error = namei(&nd)) != 0) 2844 return (error); 2845 2846 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2847 2848 vrele(nd.ni_vp); 2849 return (error); 2850 } 2851 2852 /* 2853 * Set ownership given a path name; this version does not follow links. 2854 * Provides POSIX/XPG semantics. 2855 */ 2856 /* ARGSUSED */ 2857 int 2858 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2859 { 2860 /* { 2861 syscallarg(const char *) path; 2862 syscallarg(uid_t) uid; 2863 syscallarg(gid_t) gid; 2864 } */ 2865 int error; 2866 struct nameidata nd; 2867 2868 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2869 SCARG(uap, path)); 2870 if ((error = namei(&nd)) != 0) 2871 return (error); 2872 2873 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2874 2875 vrele(nd.ni_vp); 2876 return (error); 2877 } 2878 2879 /* 2880 * Common routine to set ownership given a vnode. 2881 */ 2882 static int 2883 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2884 int posix_semantics) 2885 { 2886 struct vattr vattr; 2887 mode_t newmode; 2888 int error; 2889 2890 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2891 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2892 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2893 goto out; 2894 2895 #define CHANGED(x) ((int)(x) != -1) 2896 newmode = vattr.va_mode; 2897 if (posix_semantics) { 2898 /* 2899 * POSIX/XPG semantics: if the caller is not the super-user, 2900 * clear set-user-id and set-group-id bits. Both POSIX and 2901 * the XPG consider the behaviour for calls by the super-user 2902 * implementation-defined; we leave the set-user-id and set- 2903 * group-id settings intact in that case. 2904 */ 2905 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2906 NULL) != 0) 2907 newmode &= ~(S_ISUID | S_ISGID); 2908 } else { 2909 /* 2910 * NetBSD semantics: when changing owner and/or group, 2911 * clear the respective bit(s). 2912 */ 2913 if (CHANGED(uid)) 2914 newmode &= ~S_ISUID; 2915 if (CHANGED(gid)) 2916 newmode &= ~S_ISGID; 2917 } 2918 /* Update va_mode iff altered. */ 2919 if (vattr.va_mode == newmode) 2920 newmode = VNOVAL; 2921 2922 VATTR_NULL(&vattr); 2923 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2924 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2925 vattr.va_mode = newmode; 2926 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2927 #undef CHANGED 2928 2929 out: 2930 VOP_UNLOCK(vp, 0); 2931 return (error); 2932 } 2933 2934 /* 2935 * Set the access and modification times given a path name; this 2936 * version follows links. 2937 */ 2938 /* ARGSUSED */ 2939 int 2940 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2941 { 2942 /* { 2943 syscallarg(const char *) path; 2944 syscallarg(const struct timeval *) tptr; 2945 } */ 2946 2947 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2948 SCARG(uap, tptr), UIO_USERSPACE); 2949 } 2950 2951 /* 2952 * Set the access and modification times given a file descriptor. 2953 */ 2954 /* ARGSUSED */ 2955 int 2956 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2957 { 2958 /* { 2959 syscallarg(int) fd; 2960 syscallarg(const struct timeval *) tptr; 2961 } */ 2962 int error; 2963 struct file *fp; 2964 2965 /* getvnode() will use the descriptor for us */ 2966 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2967 return (error); 2968 2969 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2970 SCARG(uap, tptr), UIO_USERSPACE); 2971 2972 FILE_UNUSE(fp, l); 2973 return (error); 2974 } 2975 2976 /* 2977 * Set the access and modification times given a path name; this 2978 * version does not follow links. 2979 */ 2980 int 2981 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2982 { 2983 /* { 2984 syscallarg(const char *) path; 2985 syscallarg(const struct timeval *) tptr; 2986 } */ 2987 2988 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2989 SCARG(uap, tptr), UIO_USERSPACE); 2990 } 2991 2992 /* 2993 * Common routine to set access and modification times given a vnode. 2994 */ 2995 int 2996 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2997 const struct timeval *tptr, enum uio_seg seg) 2998 { 2999 struct vattr vattr; 3000 struct nameidata nd; 3001 int error; 3002 3003 VATTR_NULL(&vattr); 3004 if (tptr == NULL) { 3005 nanotime(&vattr.va_atime); 3006 vattr.va_mtime = vattr.va_atime; 3007 vattr.va_vaflags |= VA_UTIMES_NULL; 3008 } else { 3009 struct timeval tv[2]; 3010 3011 if (seg != UIO_SYSSPACE) { 3012 error = copyin(tptr, &tv, sizeof (tv)); 3013 if (error != 0) 3014 return error; 3015 tptr = tv; 3016 } 3017 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3018 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3019 } 3020 3021 if (vp == NULL) { 3022 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3023 if ((error = namei(&nd)) != 0) 3024 return (error); 3025 vp = nd.ni_vp; 3026 } else 3027 nd.ni_vp = NULL; 3028 3029 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3030 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3031 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3032 VOP_UNLOCK(vp, 0); 3033 3034 if (nd.ni_vp != NULL) 3035 vrele(nd.ni_vp); 3036 3037 return (error); 3038 } 3039 3040 /* 3041 * Truncate a file given its path name. 3042 */ 3043 /* ARGSUSED */ 3044 int 3045 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3046 { 3047 /* { 3048 syscallarg(const char *) path; 3049 syscallarg(int) pad; 3050 syscallarg(off_t) length; 3051 } */ 3052 struct vnode *vp; 3053 struct vattr vattr; 3054 int error; 3055 struct nameidata nd; 3056 3057 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3058 SCARG(uap, path)); 3059 if ((error = namei(&nd)) != 0) 3060 return (error); 3061 vp = nd.ni_vp; 3062 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3063 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3064 if (vp->v_type == VDIR) 3065 error = EISDIR; 3066 else if ((error = vn_writechk(vp)) == 0 && 3067 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3068 VATTR_NULL(&vattr); 3069 vattr.va_size = SCARG(uap, length); 3070 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3071 } 3072 vput(vp); 3073 return (error); 3074 } 3075 3076 /* 3077 * Truncate a file given a file descriptor. 3078 */ 3079 /* ARGSUSED */ 3080 int 3081 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3082 { 3083 /* { 3084 syscallarg(int) fd; 3085 syscallarg(int) pad; 3086 syscallarg(off_t) length; 3087 } */ 3088 struct proc *p = l->l_proc; 3089 struct vattr vattr; 3090 struct vnode *vp; 3091 struct file *fp; 3092 int error; 3093 3094 /* getvnode() will use the descriptor for us */ 3095 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3096 return (error); 3097 if ((fp->f_flag & FWRITE) == 0) { 3098 error = EINVAL; 3099 goto out; 3100 } 3101 vp = (struct vnode *)fp->f_data; 3102 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3103 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3104 if (vp->v_type == VDIR) 3105 error = EISDIR; 3106 else if ((error = vn_writechk(vp)) == 0) { 3107 VATTR_NULL(&vattr); 3108 vattr.va_size = SCARG(uap, length); 3109 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3110 } 3111 VOP_UNLOCK(vp, 0); 3112 out: 3113 FILE_UNUSE(fp, l); 3114 return (error); 3115 } 3116 3117 /* 3118 * Sync an open file. 3119 */ 3120 /* ARGSUSED */ 3121 int 3122 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3123 { 3124 /* { 3125 syscallarg(int) fd; 3126 } */ 3127 struct proc *p = l->l_proc; 3128 struct vnode *vp; 3129 struct file *fp; 3130 int error; 3131 3132 /* getvnode() will use the descriptor for us */ 3133 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3134 return (error); 3135 vp = (struct vnode *)fp->f_data; 3136 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3137 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3138 if (error == 0 && bioopsp != NULL && 3139 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3140 (*bioopsp->io_fsync)(vp, 0); 3141 VOP_UNLOCK(vp, 0); 3142 FILE_UNUSE(fp, l); 3143 return (error); 3144 } 3145 3146 /* 3147 * Sync a range of file data. API modeled after that found in AIX. 3148 * 3149 * FDATASYNC indicates that we need only save enough metadata to be able 3150 * to re-read the written data. Note we duplicate AIX's requirement that 3151 * the file be open for writing. 3152 */ 3153 /* ARGSUSED */ 3154 int 3155 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3156 { 3157 /* { 3158 syscallarg(int) fd; 3159 syscallarg(int) flags; 3160 syscallarg(off_t) start; 3161 syscallarg(off_t) length; 3162 } */ 3163 struct proc *p = l->l_proc; 3164 struct vnode *vp; 3165 struct file *fp; 3166 int flags, nflags; 3167 off_t s, e, len; 3168 int error; 3169 3170 /* getvnode() will use the descriptor for us */ 3171 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3172 return (error); 3173 3174 if ((fp->f_flag & FWRITE) == 0) { 3175 error = EBADF; 3176 goto out; 3177 } 3178 3179 flags = SCARG(uap, flags); 3180 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3181 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3182 error = EINVAL; 3183 goto out; 3184 } 3185 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3186 if (flags & FDATASYNC) 3187 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3188 else 3189 nflags = FSYNC_WAIT; 3190 if (flags & FDISKSYNC) 3191 nflags |= FSYNC_CACHE; 3192 3193 len = SCARG(uap, length); 3194 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3195 if (len) { 3196 s = SCARG(uap, start); 3197 e = s + len; 3198 if (e < s) { 3199 error = EINVAL; 3200 goto out; 3201 } 3202 } else { 3203 e = 0; 3204 s = 0; 3205 } 3206 3207 vp = (struct vnode *)fp->f_data; 3208 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3209 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3210 3211 if (error == 0 && bioopsp != NULL && 3212 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3213 (*bioopsp->io_fsync)(vp, nflags); 3214 3215 VOP_UNLOCK(vp, 0); 3216 out: 3217 FILE_UNUSE(fp, l); 3218 return (error); 3219 } 3220 3221 /* 3222 * Sync the data of an open file. 3223 */ 3224 /* ARGSUSED */ 3225 int 3226 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3227 { 3228 /* { 3229 syscallarg(int) fd; 3230 } */ 3231 struct proc *p = l->l_proc; 3232 struct vnode *vp; 3233 struct file *fp; 3234 int error; 3235 3236 /* getvnode() will use the descriptor for us */ 3237 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3238 return (error); 3239 if ((fp->f_flag & FWRITE) == 0) { 3240 FILE_UNUSE(fp, l); 3241 return (EBADF); 3242 } 3243 vp = (struct vnode *)fp->f_data; 3244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3245 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3246 VOP_UNLOCK(vp, 0); 3247 FILE_UNUSE(fp, l); 3248 return (error); 3249 } 3250 3251 /* 3252 * Rename files, (standard) BSD semantics frontend. 3253 */ 3254 /* ARGSUSED */ 3255 int 3256 sys_rename(struct lwp *l, const struct sys_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, 0)); 3264 } 3265 3266 /* 3267 * Rename files, POSIX semantics frontend. 3268 */ 3269 /* ARGSUSED */ 3270 int 3271 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3272 { 3273 /* { 3274 syscallarg(const char *) from; 3275 syscallarg(const char *) to; 3276 } */ 3277 3278 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3279 } 3280 3281 /* 3282 * Rename files. Source and destination must either both be directories, 3283 * or both not be directories. If target is a directory, it must be empty. 3284 * If `from' and `to' refer to the same object, the value of the `retain' 3285 * argument is used to determine whether `from' will be 3286 * 3287 * (retain == 0) deleted unless `from' and `to' refer to the same 3288 * object in the file system's name space (BSD). 3289 * (retain == 1) always retained (POSIX). 3290 */ 3291 int 3292 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3293 { 3294 struct vnode *tvp, *fvp, *tdvp; 3295 struct nameidata fromnd, tond; 3296 struct lwp *l = curlwp; 3297 struct proc *p; 3298 int error; 3299 3300 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3301 seg, from); 3302 if ((error = namei(&fromnd)) != 0) 3303 return (error); 3304 if (fromnd.ni_dvp != fromnd.ni_vp) 3305 VOP_UNLOCK(fromnd.ni_dvp, 0); 3306 fvp = fromnd.ni_vp; 3307 NDINIT(&tond, RENAME, 3308 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3309 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3310 seg, to); 3311 if ((error = namei(&tond)) != 0) { 3312 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3313 vrele(fromnd.ni_dvp); 3314 vrele(fvp); 3315 goto out1; 3316 } 3317 tdvp = tond.ni_dvp; 3318 tvp = tond.ni_vp; 3319 3320 if (tvp != NULL) { 3321 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3322 error = ENOTDIR; 3323 goto out; 3324 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3325 error = EISDIR; 3326 goto out; 3327 } 3328 } 3329 3330 if (fvp == tdvp) 3331 error = EINVAL; 3332 3333 /* 3334 * Source and destination refer to the same object. 3335 */ 3336 if (fvp == tvp) { 3337 if (retain) 3338 error = -1; 3339 else if (fromnd.ni_dvp == tdvp && 3340 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3341 !memcmp(fromnd.ni_cnd.cn_nameptr, 3342 tond.ni_cnd.cn_nameptr, 3343 fromnd.ni_cnd.cn_namelen)) 3344 error = -1; 3345 } 3346 3347 #if NVERIEXEC > 0 3348 if (!error) { 3349 char *f1, *f2; 3350 3351 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3352 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3353 3354 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3355 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3356 3357 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3358 3359 free(f1, M_TEMP); 3360 free(f2, M_TEMP); 3361 } 3362 #endif /* NVERIEXEC > 0 */ 3363 3364 out: 3365 p = l->l_proc; 3366 if (!error) { 3367 VOP_LEASE(tdvp, l->l_cred, LEASE_WRITE); 3368 if (fromnd.ni_dvp != tdvp) 3369 VOP_LEASE(fromnd.ni_dvp, l->l_cred, LEASE_WRITE); 3370 if (tvp) { 3371 VOP_LEASE(tvp, l->l_cred, LEASE_WRITE); 3372 } 3373 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3374 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3375 } else { 3376 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3377 if (tdvp == tvp) 3378 vrele(tdvp); 3379 else 3380 vput(tdvp); 3381 if (tvp) 3382 vput(tvp); 3383 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3384 vrele(fromnd.ni_dvp); 3385 vrele(fvp); 3386 } 3387 vrele(tond.ni_startdir); 3388 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3389 out1: 3390 if (fromnd.ni_startdir) 3391 vrele(fromnd.ni_startdir); 3392 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3393 return (error == -1 ? 0 : error); 3394 } 3395 3396 /* 3397 * Make a directory file. 3398 */ 3399 /* ARGSUSED */ 3400 int 3401 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3402 { 3403 /* { 3404 syscallarg(const char *) path; 3405 syscallarg(int) mode; 3406 } */ 3407 struct proc *p = l->l_proc; 3408 struct vnode *vp; 3409 struct vattr vattr; 3410 int error; 3411 struct nameidata nd; 3412 3413 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3414 SCARG(uap, path)); 3415 if ((error = namei(&nd)) != 0) 3416 return (error); 3417 vp = nd.ni_vp; 3418 if (vp != NULL) { 3419 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3420 if (nd.ni_dvp == vp) 3421 vrele(nd.ni_dvp); 3422 else 3423 vput(nd.ni_dvp); 3424 vrele(vp); 3425 return (EEXIST); 3426 } 3427 VATTR_NULL(&vattr); 3428 vattr.va_type = VDIR; 3429 /* We will read cwdi->cwdi_cmask unlocked. */ 3430 vattr.va_mode = 3431 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3432 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3433 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3434 if (!error) 3435 vput(nd.ni_vp); 3436 return (error); 3437 } 3438 3439 /* 3440 * Remove a directory file. 3441 */ 3442 /* ARGSUSED */ 3443 int 3444 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3445 { 3446 /* { 3447 syscallarg(const char *) path; 3448 } */ 3449 struct vnode *vp; 3450 int error; 3451 struct nameidata nd; 3452 3453 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3454 SCARG(uap, path)); 3455 if ((error = namei(&nd)) != 0) 3456 return (error); 3457 vp = nd.ni_vp; 3458 if (vp->v_type != VDIR) { 3459 error = ENOTDIR; 3460 goto out; 3461 } 3462 /* 3463 * No rmdir "." please. 3464 */ 3465 if (nd.ni_dvp == vp) { 3466 error = EINVAL; 3467 goto out; 3468 } 3469 /* 3470 * The root of a mounted filesystem cannot be deleted. 3471 */ 3472 if (vp->v_vflag & VV_ROOT) { 3473 error = EBUSY; 3474 goto out; 3475 } 3476 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3477 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3478 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3479 return (error); 3480 3481 out: 3482 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3483 if (nd.ni_dvp == vp) 3484 vrele(nd.ni_dvp); 3485 else 3486 vput(nd.ni_dvp); 3487 vput(vp); 3488 return (error); 3489 } 3490 3491 /* 3492 * Read a block of directory entries in a file system independent format. 3493 */ 3494 int 3495 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3496 { 3497 /* { 3498 syscallarg(int) fd; 3499 syscallarg(char *) buf; 3500 syscallarg(size_t) count; 3501 } */ 3502 struct proc *p = l->l_proc; 3503 struct file *fp; 3504 int error, done; 3505 3506 /* getvnode() will use the descriptor for us */ 3507 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3508 return (error); 3509 if ((fp->f_flag & FREAD) == 0) { 3510 error = EBADF; 3511 goto out; 3512 } 3513 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3514 SCARG(uap, count), &done, l, 0, 0); 3515 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3516 *retval = done; 3517 out: 3518 FILE_UNUSE(fp, l); 3519 return (error); 3520 } 3521 3522 /* 3523 * Set the mode mask for creation of filesystem nodes. 3524 */ 3525 int 3526 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3527 { 3528 /* { 3529 syscallarg(mode_t) newmask; 3530 } */ 3531 struct proc *p = l->l_proc; 3532 struct cwdinfo *cwdi; 3533 3534 /* 3535 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3536 * important is that we serialize changes to the mask. The 3537 * rw_exit() will issue a write memory barrier on our behalf, 3538 * and force the changes out to other CPUs (as it must use an 3539 * atomic operation, draining the local CPU's store buffers). 3540 */ 3541 cwdi = p->p_cwdi; 3542 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3543 *retval = cwdi->cwdi_cmask; 3544 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3545 rw_exit(&cwdi->cwdi_lock); 3546 3547 return (0); 3548 } 3549 3550 int 3551 dorevoke(struct vnode *vp, kauth_cred_t cred) 3552 { 3553 struct vattr vattr; 3554 int error; 3555 3556 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3557 return error; 3558 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3559 (error = kauth_authorize_generic(cred, 3560 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3561 VOP_REVOKE(vp, REVOKEALL); 3562 return (error); 3563 } 3564 3565 /* 3566 * Void all references to file by ripping underlying filesystem 3567 * away from vnode. 3568 */ 3569 /* ARGSUSED */ 3570 int 3571 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3572 { 3573 /* { 3574 syscallarg(const char *) path; 3575 } */ 3576 struct vnode *vp; 3577 int error; 3578 struct nameidata nd; 3579 3580 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3581 SCARG(uap, path)); 3582 if ((error = namei(&nd)) != 0) 3583 return (error); 3584 vp = nd.ni_vp; 3585 error = dorevoke(vp, l->l_cred); 3586 vrele(vp); 3587 return (error); 3588 } 3589 3590 /* 3591 * Convert a user file descriptor to a kernel file entry. 3592 */ 3593 int 3594 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3595 { 3596 struct vnode *vp; 3597 struct file *fp; 3598 3599 if ((fp = fd_getfile(fdp, fd)) == NULL) 3600 return (EBADF); 3601 3602 FILE_USE(fp); 3603 3604 if (fp->f_type != DTYPE_VNODE) { 3605 FILE_UNUSE(fp, NULL); 3606 return (EINVAL); 3607 } 3608 3609 vp = (struct vnode *)fp->f_data; 3610 if (vp->v_type == VBAD) { 3611 FILE_UNUSE(fp, NULL); 3612 return (EBADF); 3613 } 3614 3615 *fpp = fp; 3616 return (0); 3617 } 3618
Indexes created Fri Sep 26 20:09:58 GMT 2025