vfs_syscalls.c revision 1.329
1 /* $NetBSD: vfs_syscalls.c,v 1.329 2007/10/10 20:42:27 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 1.329 2007/10/10 20:42:27 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 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #endif 79 #include <nfs/nfsproto.h> 80 #ifdef COMPAT_30 81 #include <nfs/nfs.h> 82 #include <nfs/nfs_var.h> 83 #endif 84 85 #if NFSS > 0 86 #include <dev/fssvar.h> 87 #endif 88 89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 90 91 static int change_dir(struct nameidata *, struct lwp *); 92 static int change_flags(struct vnode *, u_long, struct lwp *); 93 static int change_mode(struct vnode *, int, struct lwp *l); 94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 95 static int rename_files(const char *, const char *, struct lwp *, int); 96 97 void checkdirs(struct vnode *); 98 99 int dovfsusermount = 0; 100 101 /* 102 * Virtual File System System Calls 103 */ 104 105 /* 106 * Mount a file system. 107 */ 108 109 #if defined(COMPAT_09) || defined(COMPAT_43) 110 /* 111 * This table is used to maintain compatibility with 4.3BSD 112 * and NetBSD 0.9 mount syscalls. Note, the order is important! 113 * 114 * Do not modify this table. It should only contain filesystems 115 * supported by NetBSD 0.9 and 4.3BSD. 116 */ 117 const char * const mountcompatnames[] = { 118 NULL, /* 0 = MOUNT_NONE */ 119 MOUNT_FFS, /* 1 = MOUNT_UFS */ 120 MOUNT_NFS, /* 2 */ 121 MOUNT_MFS, /* 3 */ 122 MOUNT_MSDOS, /* 4 */ 123 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 124 MOUNT_FDESC, /* 6 */ 125 MOUNT_KERNFS, /* 7 */ 126 NULL, /* 8 = MOUNT_DEVFS */ 127 MOUNT_AFS, /* 9 */ 128 }; 129 const int nmountcompatnames = sizeof(mountcompatnames) / 130 sizeof(mountcompatnames[0]); 131 #endif /* COMPAT_09 || COMPAT_43 */ 132 133 static int 134 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 135 void *data, size_t *data_len) 136 { 137 struct mount *mp; 138 int error = 0, saved_flags; 139 140 mp = vp->v_mount; 141 saved_flags = mp->mnt_flag; 142 143 /* We can operate only on VV_ROOT nodes. */ 144 if ((vp->v_vflag & VV_ROOT) == 0) { 145 error = EINVAL; 146 goto out; 147 } 148 149 /* 150 * We only allow the filesystem to be reloaded if it 151 * is currently mounted read-only. 152 */ 153 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 154 error = EOPNOTSUPP; /* Needs translation */ 155 goto out; 156 } 157 158 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 159 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 160 if (error) 161 goto out; 162 163 if (vfs_busy(mp, LK_NOWAIT, 0)) { 164 error = EPERM; 165 goto out; 166 } 167 168 mp->mnt_flag &= ~MNT_OP_FLAGS; 169 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 170 171 /* 172 * Set the mount level flags. 173 */ 174 if (flags & MNT_RDONLY) 175 mp->mnt_flag |= MNT_RDONLY; 176 else if (mp->mnt_flag & MNT_RDONLY) 177 mp->mnt_iflag |= IMNT_WANTRDWR; 178 mp->mnt_flag &= 179 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 180 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 181 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 182 mp->mnt_flag |= flags & 183 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 184 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 185 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 186 MNT_IGNORE); 187 188 error = VFS_MOUNT(mp, path, data, data_len, l); 189 190 #if defined(COMPAT_30) && defined(NFSSERVER) 191 if (error && data != NULL) { 192 int error2; 193 194 /* Update failed; let's try and see if it was an 195 * export request. */ 196 error2 = nfs_update_exports_30(mp, path, data, l); 197 198 /* Only update error code if the export request was 199 * understood but some problem occurred while 200 * processing it. */ 201 if (error2 != EJUSTRETURN) 202 error = error2; 203 } 204 #endif 205 if (mp->mnt_iflag & IMNT_WANTRDWR) 206 mp->mnt_flag &= ~MNT_RDONLY; 207 if (error) 208 mp->mnt_flag = saved_flags; 209 mp->mnt_flag &= ~MNT_OP_FLAGS; 210 mp->mnt_iflag &= ~IMNT_WANTRDWR; 211 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 212 if (mp->mnt_syncer == NULL) 213 error = vfs_allocate_syncvnode(mp); 214 } else { 215 if (mp->mnt_syncer != NULL) 216 vfs_deallocate_syncvnode(mp); 217 } 218 vfs_unbusy(mp); 219 220 out: 221 return (error); 222 } 223 224 static int 225 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 226 { 227 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 228 int error; 229 230 /* Copy file-system type from userspace. */ 231 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 232 if (error) { 233 #if defined(COMPAT_09) || defined(COMPAT_43) 234 /* 235 * Historically, filesystem types were identified by numbers. 236 * If we get an integer for the filesystem type instead of a 237 * string, we check to see if it matches one of the historic 238 * filesystem types. 239 */ 240 u_long fsindex = (u_long)fstype; 241 if (fsindex >= nmountcompatnames || 242 mountcompatnames[fsindex] == NULL) 243 return ENODEV; 244 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 245 #else 246 return error; 247 #endif 248 } 249 250 #ifdef COMPAT_10 251 /* Accept `ufs' as an alias for `ffs'. */ 252 if (strcmp(fstypename, "ufs") == 0) 253 fstypename[0] = 'f'; 254 #endif 255 256 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 257 return ENODEV; 258 return 0; 259 } 260 261 static int 262 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 263 const char *path, int flags, void *data, size_t *data_len) 264 { 265 struct mount *mp = NULL; 266 struct vnode *vp = *vpp; 267 struct vattr va; 268 int error; 269 270 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 271 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 272 if (error) 273 return error; 274 275 /* Can't make a non-dir a mount-point (from here anyway). */ 276 if (vp->v_type != VDIR) 277 return ENOTDIR; 278 279 /* 280 * If the user is not root, ensure that they own the directory 281 * onto which we are attempting to mount. 282 */ 283 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 284 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 285 (error = kauth_authorize_generic(l->l_cred, 286 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 287 return error; 288 } 289 290 if (flags & MNT_EXPORTED) 291 return EINVAL; 292 293 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 294 return error; 295 296 /* 297 * Check if a file-system is not already mounted on this vnode. 298 */ 299 if (vp->v_mountedhere != NULL) 300 return EBUSY; 301 302 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 303 304 mp->mnt_op = vfsops; 305 306 TAILQ_INIT(&mp->mnt_vnodelist); 307 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 308 simple_lock_init(&mp->mnt_slock); 309 (void)vfs_busy(mp, LK_NOWAIT, 0); 310 311 mp->mnt_op->vfs_refcount++; 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, l); 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 VOP_UNLOCK(vp, 0); 350 checkdirs(vp); 351 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 352 error = vfs_allocate_syncvnode(mp); 353 vfs_unbusy(mp); 354 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 355 error = VFS_START(mp, 0, l); 356 if (error) 357 vrele(vp); 358 *vpp = NULL; 359 return error; 360 } 361 362 static int 363 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 364 void *data, size_t *data_len) 365 { 366 struct mount *mp; 367 int error; 368 369 /* If MNT_GETARGS is specified, it should be the only flag. */ 370 if (flags & ~MNT_GETARGS) 371 return EINVAL; 372 373 mp = vp->v_mount; 374 375 /* XXX: probably some notion of "can see" here if we want isolation. */ 376 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 377 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 378 if (error) 379 return error; 380 381 if ((vp->v_vflag & VV_ROOT) == 0) 382 return EINVAL; 383 384 if (vfs_busy(mp, LK_NOWAIT, 0)) 385 return EPERM; 386 387 mp->mnt_flag &= ~MNT_OP_FLAGS; 388 mp->mnt_flag |= MNT_GETARGS; 389 error = VFS_MOUNT(mp, path, data, data_len, l); 390 mp->mnt_flag &= ~MNT_OP_FLAGS; 391 392 vfs_unbusy(mp); 393 return (error); 394 } 395 396 #ifdef COMPAT_40 397 /* ARGSUSED */ 398 int 399 compat_40_sys_mount(struct lwp *l, void *v, register_t *retval) 400 { 401 struct compat_40_sys_mount_args /* { 402 syscallarg(const char *) type; 403 syscallarg(const char *) path; 404 syscallarg(int) flags; 405 syscallarg(void *) data; 406 } */ *uap = v; 407 register_t dummy; 408 409 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 410 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 411 } 412 #endif 413 414 int 415 sys___mount50(struct lwp *l, void *v, register_t *retval) 416 { 417 struct sys___mount50_args /* { 418 syscallarg(const char *) type; 419 syscallarg(const char *) path; 420 syscallarg(int) flags; 421 syscallarg(void *) data; 422 syscallarg(size_t) data_len; 423 } */ *uap = v; 424 425 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 426 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 427 SCARG(uap, data_len), retval); 428 } 429 430 int 431 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 432 const char *path, int flags, void *data, enum uio_seg data_seg, 433 size_t data_len, register_t *retval) 434 { 435 struct vnode *vp; 436 struct nameidata nd; 437 void *data_buf = data; 438 int error; 439 440 /* 441 * Get vnode to be covered 442 */ 443 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 444 if ((error = namei(&nd)) != 0) 445 return (error); 446 vp = nd.ni_vp; 447 448 /* 449 * A lookup in VFS_MOUNT might result in an attempt to 450 * lock this vnode again, so make the lock recursive. 451 */ 452 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 453 454 if (vfsops == NULL) { 455 if (flags & (MNT_GETARGS | MNT_UPDATE)) 456 vfsops = vp->v_mount->mnt_op; 457 else { 458 /* 'type' is userspace */ 459 error = mount_get_vfsops(type, &vfsops); 460 if (error != 0) 461 goto done; 462 } 463 } 464 465 if (data != NULL && data_seg == UIO_USERSPACE) { 466 if (data_len == 0) { 467 /* No length supplied, use default for filesystem */ 468 data_len = vfsops->vfs_min_mount_data; 469 if (data_len > VFS_MAX_MOUNT_DATA) { 470 /* maybe a force loaded old LKM */ 471 error = EINVAL; 472 goto done; 473 } 474 #ifdef COMPAT_30 475 /* Hopefully a longer buffer won't make copyin() fail */ 476 if (flags & MNT_UPDATE 477 && data_len < sizeof (struct mnt_export_args30)) 478 data_len = sizeof (struct mnt_export_args30); 479 #endif 480 } 481 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 482 483 /* NFS needs the buffer even for mnt_getargs .... */ 484 error = copyin(data, data_buf, data_len); 485 if (error != 0) 486 goto done; 487 } 488 489 if (flags & MNT_GETARGS) { 490 if (data_len == 0) { 491 error = EINVAL; 492 goto done; 493 } 494 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 495 if (error != 0) 496 goto done; 497 if (data_seg == UIO_USERSPACE) 498 error = copyout(data_buf, data, data_len); 499 *retval = data_len; 500 } else if (flags & MNT_UPDATE) { 501 error = mount_update(l, vp, path, flags, data_buf, &data_len); 502 } else { 503 /* Locking is handled internally in mount_domount(). */ 504 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 505 &data_len); 506 } 507 508 done: 509 if (vp) 510 vput(vp); 511 if (data_buf != data) 512 free(data_buf, M_TEMP); 513 return (error); 514 } 515 516 /* 517 * Scan all active processes to see if any of them have a current 518 * or root directory onto which the new filesystem has just been 519 * mounted. If so, replace them with the new mount point. 520 */ 521 void 522 checkdirs(struct vnode *olddp) 523 { 524 struct cwdinfo *cwdi; 525 struct vnode *newdp; 526 struct proc *p; 527 528 if (olddp->v_usecount == 1) 529 return; 530 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 531 panic("mount: lost mount"); 532 mutex_enter(&proclist_lock); 533 PROCLIST_FOREACH(p, &allproc) { 534 cwdi = p->p_cwdi; 535 if (!cwdi) 536 continue; 537 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 538 if (cwdi->cwdi_cdir == olddp) { 539 vrele(cwdi->cwdi_cdir); 540 VREF(newdp); 541 cwdi->cwdi_cdir = newdp; 542 } 543 if (cwdi->cwdi_rdir == olddp) { 544 vrele(cwdi->cwdi_rdir); 545 VREF(newdp); 546 cwdi->cwdi_rdir = newdp; 547 } 548 rw_exit(&cwdi->cwdi_lock); 549 } 550 mutex_exit(&proclist_lock); 551 if (rootvnode == olddp) { 552 vrele(rootvnode); 553 VREF(newdp); 554 rootvnode = newdp; 555 } 556 vput(newdp); 557 } 558 559 /* 560 * Unmount a file system. 561 * 562 * Note: unmount takes a path to the vnode mounted on as argument, 563 * not special file (as before). 564 */ 565 /* ARGSUSED */ 566 int 567 sys_unmount(struct lwp *l, void *v, register_t *retval) 568 { 569 struct sys_unmount_args /* { 570 syscallarg(const char *) path; 571 syscallarg(int) flags; 572 } */ *uap = v; 573 struct vnode *vp; 574 struct mount *mp; 575 int error; 576 struct nameidata nd; 577 578 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 579 SCARG(uap, path), l); 580 if ((error = namei(&nd)) != 0) 581 return (error); 582 vp = nd.ni_vp; 583 mp = vp->v_mount; 584 585 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 586 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 587 if (error) { 588 vput(vp); 589 return (error); 590 } 591 592 /* 593 * Don't allow unmounting the root file system. 594 */ 595 if (mp->mnt_flag & MNT_ROOTFS) { 596 vput(vp); 597 return (EINVAL); 598 } 599 600 /* 601 * Must be the root of the filesystem 602 */ 603 if ((vp->v_vflag & VV_ROOT) == 0) { 604 vput(vp); 605 return (EINVAL); 606 } 607 vput(vp); 608 609 /* 610 * XXX Freeze syncer. Must do this before locking the 611 * mount point. See dounmount() for details. 612 */ 613 mutex_enter(&syncer_mutex); 614 615 if (vfs_busy(mp, 0, 0)) { 616 mutex_exit(&syncer_mutex); 617 return (EBUSY); 618 } 619 620 return (dounmount(mp, SCARG(uap, flags), l)); 621 } 622 623 /* 624 * Do the actual file system unmount. File system is assumed to have been 625 * marked busy by the caller. 626 */ 627 int 628 dounmount(struct mount *mp, int flags, struct lwp *l) 629 { 630 struct vnode *coveredvp; 631 int error; 632 int async; 633 int used_syncer; 634 635 #if NVERIEXEC > 0 636 error = veriexec_unmountchk(mp); 637 if (error) 638 return (error); 639 #endif /* NVERIEXEC > 0 */ 640 641 mutex_enter(&mountlist_lock); 642 vfs_unbusy(mp); 643 used_syncer = (mp->mnt_syncer != NULL); 644 645 /* 646 * XXX Syncer must be frozen when we get here. This should really 647 * be done on a per-mountpoint basis, but especially the softdep 648 * code possibly called from the syncer doesn't exactly work on a 649 * per-mountpoint basis, so the softdep code would become a maze 650 * of vfs_busy() calls. 651 * 652 * The caller of dounmount() must acquire syncer_mutex because 653 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 654 * order, and we must preserve that order to avoid deadlock. 655 * 656 * So, if the file system did not use the syncer, now is 657 * the time to release the syncer_mutex. 658 */ 659 if (used_syncer == 0) 660 mutex_exit(&syncer_mutex); 661 662 mp->mnt_iflag |= IMNT_UNMOUNT; 663 mp->mnt_unmounter = l; 664 mutex_exit(&mountlist_lock); /* XXX */ 665 lockmgr(&mp->mnt_lock, LK_DRAIN, NULL); 666 667 async = mp->mnt_flag & MNT_ASYNC; 668 mp->mnt_flag &= ~MNT_ASYNC; 669 cache_purgevfs(mp); /* remove cache entries for this file sys */ 670 if (mp->mnt_syncer != NULL) 671 vfs_deallocate_syncvnode(mp); 672 error = 0; 673 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 674 #if NFSS > 0 675 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 676 #endif 677 if (error == 0) 678 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 679 } 680 if (error == 0 || (flags & MNT_FORCE)) 681 error = VFS_UNMOUNT(mp, flags, l); 682 if (error) { 683 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 684 (void) vfs_allocate_syncvnode(mp); 685 mutex_enter(&mountlist_lock); 686 mp->mnt_iflag &= ~IMNT_UNMOUNT; 687 mp->mnt_unmounter = NULL; 688 mp->mnt_flag |= async; 689 mutex_exit(&mountlist_lock); /* XXX */ 690 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_REENABLE, 691 NULL); 692 if (used_syncer) 693 mutex_exit(&syncer_mutex); 694 simple_lock(&mp->mnt_slock); 695 while (mp->mnt_wcnt > 0) { 696 wakeup(mp); 697 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 698 0, &mp->mnt_slock); 699 } 700 simple_unlock(&mp->mnt_slock); 701 return (error); 702 } 703 mutex_enter(&mountlist_lock); 704 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 705 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 706 coveredvp->v_mountedhere = NULL; 707 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 708 panic("unmount: dangling vnode"); 709 mp->mnt_iflag |= IMNT_GONE; 710 mutex_exit(&mountlist_lock); 711 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL); 712 if (coveredvp != NULLVP) 713 vrele(coveredvp); 714 if (used_syncer) 715 mutex_exit(&syncer_mutex); 716 simple_lock(&mp->mnt_slock); 717 while (mp->mnt_wcnt > 0) { 718 wakeup(mp); 719 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 720 } 721 simple_unlock(&mp->mnt_slock); 722 vfs_hooks_unmount(mp); 723 vfs_delref(mp->mnt_op); 724 vfs_destroy(mp); 725 return (0); 726 } 727 728 /* 729 * Sync each mounted filesystem. 730 */ 731 #ifdef DEBUG 732 int syncprt = 0; 733 struct ctldebug debug0 = { "syncprt", &syncprt }; 734 #endif 735 736 /* ARGSUSED */ 737 int 738 sys_sync(struct lwp *l, void *v, register_t *retval) 739 { 740 struct mount *mp, *nmp; 741 int asyncflag; 742 743 if (l == NULL) 744 l = &lwp0; 745 746 mutex_enter(&mountlist_lock); 747 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 748 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 749 nmp = mp->mnt_list.cqe_prev; 750 continue; 751 } 752 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 753 asyncflag = mp->mnt_flag & MNT_ASYNC; 754 mp->mnt_flag &= ~MNT_ASYNC; 755 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 756 if (asyncflag) 757 mp->mnt_flag |= MNT_ASYNC; 758 } 759 mutex_enter(&mountlist_lock); 760 nmp = mp->mnt_list.cqe_prev; 761 vfs_unbusy(mp); 762 763 } 764 mutex_exit(&mountlist_lock); 765 #ifdef DEBUG 766 if (syncprt) 767 vfs_bufstats(); 768 #endif /* DEBUG */ 769 return (0); 770 } 771 772 /* 773 * Change filesystem quotas. 774 */ 775 /* ARGSUSED */ 776 int 777 sys_quotactl(struct lwp *l, void *v, register_t *retval) 778 { 779 struct sys_quotactl_args /* { 780 syscallarg(const char *) path; 781 syscallarg(int) cmd; 782 syscallarg(int) uid; 783 syscallarg(void *) arg; 784 } */ *uap = v; 785 struct mount *mp; 786 int error; 787 struct nameidata nd; 788 789 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 790 SCARG(uap, path), l); 791 if ((error = namei(&nd)) != 0) 792 return (error); 793 mp = nd.ni_vp->v_mount; 794 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 795 SCARG(uap, arg), l); 796 vrele(nd.ni_vp); 797 return (error); 798 } 799 800 int 801 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 802 int root) 803 { 804 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 805 int error = 0; 806 807 /* 808 * If MNT_NOWAIT or MNT_LAZY is specified, do not 809 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 810 * overrides MNT_NOWAIT. 811 */ 812 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 813 (flags != MNT_WAIT && flags != 0)) { 814 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 815 goto done; 816 } 817 818 /* Get the filesystem stats now */ 819 memset(sp, 0, sizeof(*sp)); 820 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 821 return error; 822 } 823 824 if (cwdi->cwdi_rdir == NULL) 825 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 826 done: 827 if (cwdi->cwdi_rdir != NULL) { 828 size_t len; 829 char *bp; 830 char *path = PNBUF_GET(); 831 832 bp = path + MAXPATHLEN; 833 *--bp = '\0'; 834 rw_enter(&cwdi->cwdi_lock, RW_READER); 835 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 836 MAXPATHLEN / 2, 0, l); 837 rw_exit(&cwdi->cwdi_lock); 838 if (error) { 839 PNBUF_PUT(path); 840 return error; 841 } 842 len = strlen(bp); 843 /* 844 * for mount points that are below our root, we can see 845 * them, so we fix up the pathname and return them. The 846 * rest we cannot see, so we don't allow viewing the 847 * data. 848 */ 849 if (strncmp(bp, sp->f_mntonname, len) == 0) { 850 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 851 sizeof(sp->f_mntonname)); 852 if (sp->f_mntonname[0] == '\0') 853 (void)strlcpy(sp->f_mntonname, "/", 854 sizeof(sp->f_mntonname)); 855 } else { 856 if (root) 857 (void)strlcpy(sp->f_mntonname, "/", 858 sizeof(sp->f_mntonname)); 859 else 860 error = EPERM; 861 } 862 PNBUF_PUT(path); 863 } 864 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 865 return error; 866 } 867 868 /* 869 * Get filesystem statistics by path. 870 */ 871 int 872 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 873 { 874 struct mount *mp; 875 int error; 876 struct nameidata nd; 877 878 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 879 if ((error = namei(&nd)) != 0) 880 return error; 881 mp = nd.ni_vp->v_mount; 882 error = dostatvfs(mp, sb, l, flags, 1); 883 vrele(nd.ni_vp); 884 return error; 885 } 886 887 /* ARGSUSED */ 888 int 889 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 890 { 891 struct sys_statvfs1_args /* { 892 syscallarg(const char *) path; 893 syscallarg(struct statvfs *) buf; 894 syscallarg(int) flags; 895 } */ *uap = v; 896 struct statvfs *sb; 897 int error; 898 899 sb = STATVFSBUF_GET(); 900 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 901 if (error == 0) 902 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 903 STATVFSBUF_PUT(sb); 904 return error; 905 } 906 907 /* 908 * Get filesystem statistics by fd. 909 */ 910 int 911 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 912 { 913 struct proc *p = l->l_proc; 914 struct file *fp; 915 struct mount *mp; 916 int error; 917 918 /* getvnode() will use the descriptor for us */ 919 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 920 return (error); 921 mp = ((struct vnode *)fp->f_data)->v_mount; 922 error = dostatvfs(mp, sb, l, flags, 1); 923 FILE_UNUSE(fp, l); 924 return error; 925 } 926 927 /* ARGSUSED */ 928 int 929 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 930 { 931 struct sys_fstatvfs1_args /* { 932 syscallarg(int) fd; 933 syscallarg(struct statvfs *) buf; 934 syscallarg(int) flags; 935 } */ *uap = v; 936 struct statvfs *sb; 937 int error; 938 939 sb = STATVFSBUF_GET(); 940 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 941 if (error == 0) 942 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 943 STATVFSBUF_PUT(sb); 944 return error; 945 } 946 947 948 /* 949 * Get statistics on all filesystems. 950 */ 951 int 952 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 953 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 954 register_t *retval) 955 { 956 int root = 0; 957 struct proc *p = l->l_proc; 958 struct mount *mp, *nmp; 959 struct statvfs *sb; 960 size_t count, maxcount; 961 int error = 0; 962 963 sb = STATVFSBUF_GET(); 964 maxcount = bufsize / entry_sz; 965 mutex_enter(&mountlist_lock); 966 count = 0; 967 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 968 mp = nmp) { 969 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 970 nmp = CIRCLEQ_NEXT(mp, mnt_list); 971 continue; 972 } 973 if (sfsp && count < maxcount) { 974 error = dostatvfs(mp, sb, l, flags, 0); 975 if (error) { 976 mutex_enter(&mountlist_lock); 977 nmp = CIRCLEQ_NEXT(mp, mnt_list); 978 vfs_unbusy(mp); 979 continue; 980 } 981 error = copyfn(sb, sfsp, entry_sz); 982 if (error) { 983 vfs_unbusy(mp); 984 goto out; 985 } 986 sfsp = (char *)sfsp + entry_sz; 987 root |= strcmp(sb->f_mntonname, "/") == 0; 988 } 989 count++; 990 mutex_enter(&mountlist_lock); 991 nmp = CIRCLEQ_NEXT(mp, mnt_list); 992 vfs_unbusy(mp); 993 } 994 mutex_exit(&mountlist_lock); 995 if (root == 0 && p->p_cwdi->cwdi_rdir) { 996 /* 997 * fake a root entry 998 */ 999 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 1000 if (error != 0) 1001 goto out; 1002 if (sfsp) 1003 error = copyfn(sb, sfsp, entry_sz); 1004 count++; 1005 } 1006 if (sfsp && count > maxcount) 1007 *retval = maxcount; 1008 else 1009 *retval = count; 1010 out: 1011 STATVFSBUF_PUT(sb); 1012 return error; 1013 } 1014 1015 int 1016 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 1017 { 1018 struct sys_getvfsstat_args /* { 1019 syscallarg(struct statvfs *) buf; 1020 syscallarg(size_t) bufsize; 1021 syscallarg(int) flags; 1022 } */ *uap = v; 1023 1024 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1025 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1026 } 1027 1028 /* 1029 * Change current working directory to a given file descriptor. 1030 */ 1031 /* ARGSUSED */ 1032 int 1033 sys_fchdir(struct lwp *l, void *v, register_t *retval) 1034 { 1035 struct sys_fchdir_args /* { 1036 syscallarg(int) fd; 1037 } */ *uap = v; 1038 struct proc *p = l->l_proc; 1039 struct filedesc *fdp = p->p_fd; 1040 struct cwdinfo *cwdi; 1041 struct vnode *vp, *tdp; 1042 struct mount *mp; 1043 struct file *fp; 1044 int error; 1045 1046 /* getvnode() will use the descriptor for us */ 1047 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1048 return (error); 1049 vp = (struct vnode *)fp->f_data; 1050 1051 VREF(vp); 1052 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1053 if (vp->v_type != VDIR) 1054 error = ENOTDIR; 1055 else 1056 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1057 if (error) { 1058 vput(vp); 1059 goto out; 1060 } 1061 while ((mp = vp->v_mountedhere) != NULL) { 1062 if (vfs_busy(mp, 0, 0)) 1063 continue; 1064 1065 vput(vp); 1066 error = VFS_ROOT(mp, &tdp); 1067 vfs_unbusy(mp); 1068 if (error) 1069 goto out; 1070 vp = tdp; 1071 } 1072 VOP_UNLOCK(vp, 0); 1073 1074 /* 1075 * Disallow changing to a directory not under the process's 1076 * current root directory (if there is one). 1077 */ 1078 cwdi = p->p_cwdi; 1079 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1080 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1081 vrele(vp); 1082 error = EPERM; /* operation not permitted */ 1083 } else { 1084 vrele(cwdi->cwdi_cdir); 1085 cwdi->cwdi_cdir = vp; 1086 } 1087 rw_exit(&cwdi->cwdi_lock); 1088 1089 out: 1090 FILE_UNUSE(fp, l); 1091 return (error); 1092 } 1093 1094 /* 1095 * Change this process's notion of the root directory to a given file 1096 * descriptor. 1097 */ 1098 int 1099 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1100 { 1101 struct sys_fchroot_args *uap = v; 1102 struct proc *p = l->l_proc; 1103 struct filedesc *fdp = p->p_fd; 1104 struct cwdinfo *cwdi; 1105 struct vnode *vp; 1106 struct file *fp; 1107 int error; 1108 1109 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1110 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1111 return error; 1112 /* getvnode() will use the descriptor for us */ 1113 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1114 return error; 1115 vp = (struct vnode *) fp->f_data; 1116 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1117 if (vp->v_type != VDIR) 1118 error = ENOTDIR; 1119 else 1120 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1121 VOP_UNLOCK(vp, 0); 1122 if (error) 1123 goto out; 1124 VREF(vp); 1125 1126 /* 1127 * Prevent escaping from chroot by putting the root under 1128 * the working directory. Silently chdir to / if we aren't 1129 * already there. 1130 */ 1131 cwdi = p->p_cwdi; 1132 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1133 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1134 /* 1135 * XXX would be more failsafe to change directory to a 1136 * deadfs node here instead 1137 */ 1138 vrele(cwdi->cwdi_cdir); 1139 VREF(vp); 1140 cwdi->cwdi_cdir = vp; 1141 } 1142 1143 if (cwdi->cwdi_rdir != NULL) 1144 vrele(cwdi->cwdi_rdir); 1145 cwdi->cwdi_rdir = vp; 1146 rw_exit(&cwdi->cwdi_lock); 1147 1148 out: 1149 FILE_UNUSE(fp, l); 1150 return (error); 1151 } 1152 1153 /* 1154 * Change current working directory (``.''). 1155 */ 1156 /* ARGSUSED */ 1157 int 1158 sys_chdir(struct lwp *l, void *v, register_t *retval) 1159 { 1160 struct sys_chdir_args /* { 1161 syscallarg(const char *) path; 1162 } */ *uap = v; 1163 struct proc *p = l->l_proc; 1164 struct cwdinfo *cwdi; 1165 int error; 1166 struct nameidata nd; 1167 1168 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1169 SCARG(uap, path), l); 1170 if ((error = change_dir(&nd, l)) != 0) 1171 return (error); 1172 cwdi = p->p_cwdi; 1173 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1174 vrele(cwdi->cwdi_cdir); 1175 cwdi->cwdi_cdir = nd.ni_vp; 1176 rw_exit(&cwdi->cwdi_lock); 1177 return (0); 1178 } 1179 1180 /* 1181 * Change notion of root (``/'') directory. 1182 */ 1183 /* ARGSUSED */ 1184 int 1185 sys_chroot(struct lwp *l, void *v, register_t *retval) 1186 { 1187 struct sys_chroot_args /* { 1188 syscallarg(const char *) path; 1189 } */ *uap = v; 1190 struct proc *p = l->l_proc; 1191 struct cwdinfo *cwdi; 1192 struct vnode *vp; 1193 int error; 1194 struct nameidata nd; 1195 1196 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1197 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1198 return (error); 1199 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1200 SCARG(uap, path), l); 1201 if ((error = change_dir(&nd, l)) != 0) 1202 return (error); 1203 1204 cwdi = p->p_cwdi; 1205 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1206 if (cwdi->cwdi_rdir != NULL) 1207 vrele(cwdi->cwdi_rdir); 1208 vp = nd.ni_vp; 1209 cwdi->cwdi_rdir = vp; 1210 1211 /* 1212 * Prevent escaping from chroot by putting the root under 1213 * the working directory. Silently chdir to / if we aren't 1214 * already there. 1215 */ 1216 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1217 /* 1218 * XXX would be more failsafe to change directory to a 1219 * deadfs node here instead 1220 */ 1221 vrele(cwdi->cwdi_cdir); 1222 VREF(vp); 1223 cwdi->cwdi_cdir = vp; 1224 } 1225 rw_exit(&cwdi->cwdi_lock); 1226 1227 return (0); 1228 } 1229 1230 /* 1231 * Common routine for chroot and chdir. 1232 */ 1233 static int 1234 change_dir(struct nameidata *ndp, struct lwp *l) 1235 { 1236 struct vnode *vp; 1237 int error; 1238 1239 if ((error = namei(ndp)) != 0) 1240 return (error); 1241 vp = ndp->ni_vp; 1242 if (vp->v_type != VDIR) 1243 error = ENOTDIR; 1244 else 1245 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1246 1247 if (error) 1248 vput(vp); 1249 else 1250 VOP_UNLOCK(vp, 0); 1251 return (error); 1252 } 1253 1254 /* 1255 * Check permissions, allocate an open file structure, 1256 * and call the device open routine if any. 1257 */ 1258 int 1259 sys_open(struct lwp *l, void *v, register_t *retval) 1260 { 1261 struct sys_open_args /* { 1262 syscallarg(const char *) path; 1263 syscallarg(int) flags; 1264 syscallarg(int) mode; 1265 } */ *uap = v; 1266 struct proc *p = l->l_proc; 1267 struct cwdinfo *cwdi = p->p_cwdi; 1268 struct filedesc *fdp = p->p_fd; 1269 struct file *fp; 1270 struct vnode *vp; 1271 int flags, cmode; 1272 int type, indx, error; 1273 struct flock lf; 1274 struct nameidata nd; 1275 1276 flags = FFLAGS(SCARG(uap, flags)); 1277 if ((flags & (FREAD | FWRITE)) == 0) 1278 return (EINVAL); 1279 /* falloc() will use the file descriptor for us */ 1280 if ((error = falloc(l, &fp, &indx)) != 0) 1281 return (error); 1282 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1283 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1284 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1285 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1286 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1287 rw_enter(&fdp->fd_lock, RW_WRITER); 1288 FILE_UNUSE(fp, l); 1289 fdp->fd_ofiles[indx] = NULL; 1290 rw_exit(&fdp->fd_lock); 1291 ffree(fp); 1292 if ((error == EDUPFD || error == EMOVEFD) && 1293 l->l_dupfd >= 0 && /* XXX from fdopen */ 1294 (error = 1295 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1296 *retval = indx; 1297 return (0); 1298 } 1299 if (error == ERESTART) 1300 error = EINTR; 1301 fdremove(fdp, indx); 1302 return (error); 1303 } 1304 l->l_dupfd = 0; 1305 vp = nd.ni_vp; 1306 fp->f_flag = flags & FMASK; 1307 fp->f_type = DTYPE_VNODE; 1308 fp->f_ops = &vnops; 1309 fp->f_data = vp; 1310 if (flags & (O_EXLOCK | O_SHLOCK)) { 1311 lf.l_whence = SEEK_SET; 1312 lf.l_start = 0; 1313 lf.l_len = 0; 1314 if (flags & O_EXLOCK) 1315 lf.l_type = F_WRLCK; 1316 else 1317 lf.l_type = F_RDLCK; 1318 type = F_FLOCK; 1319 if ((flags & FNONBLOCK) == 0) 1320 type |= F_WAIT; 1321 VOP_UNLOCK(vp, 0); 1322 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1323 if (error) { 1324 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1325 FILE_UNUSE(fp, l); 1326 ffree(fp); 1327 fdremove(fdp, indx); 1328 return (error); 1329 } 1330 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1331 fp->f_flag |= FHASLOCK; 1332 } 1333 VOP_UNLOCK(vp, 0); 1334 *retval = indx; 1335 FILE_SET_MATURE(fp); 1336 FILE_UNUSE(fp, l); 1337 return (0); 1338 } 1339 1340 static void 1341 vfs__fhfree(fhandle_t *fhp) 1342 { 1343 size_t fhsize; 1344 1345 if (fhp == NULL) { 1346 return; 1347 } 1348 fhsize = FHANDLE_SIZE(fhp); 1349 kmem_free(fhp, fhsize); 1350 } 1351 1352 /* 1353 * vfs_composefh: compose a filehandle. 1354 */ 1355 1356 int 1357 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1358 { 1359 struct mount *mp; 1360 struct fid *fidp; 1361 int error; 1362 size_t needfhsize; 1363 size_t fidsize; 1364 1365 mp = vp->v_mount; 1366 fidp = NULL; 1367 if (*fh_size < FHANDLE_SIZE_MIN) { 1368 fidsize = 0; 1369 } else { 1370 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1371 if (fhp != NULL) { 1372 memset(fhp, 0, *fh_size); 1373 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1374 fidp = &fhp->fh_fid; 1375 } 1376 } 1377 error = VFS_VPTOFH(vp, fidp, &fidsize); 1378 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1379 if (error == 0 && *fh_size < needfhsize) { 1380 error = E2BIG; 1381 } 1382 *fh_size = needfhsize; 1383 return error; 1384 } 1385 1386 int 1387 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1388 { 1389 struct mount *mp; 1390 fhandle_t *fhp; 1391 size_t fhsize; 1392 size_t fidsize; 1393 int error; 1394 1395 *fhpp = NULL; 1396 mp = vp->v_mount; 1397 fidsize = 0; 1398 error = VFS_VPTOFH(vp, NULL, &fidsize); 1399 KASSERT(error != 0); 1400 if (error != E2BIG) { 1401 goto out; 1402 } 1403 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1404 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1405 if (fhp == NULL) { 1406 error = ENOMEM; 1407 goto out; 1408 } 1409 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1410 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1411 if (error == 0) { 1412 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1413 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1414 *fhpp = fhp; 1415 } else { 1416 kmem_free(fhp, fhsize); 1417 } 1418 out: 1419 return error; 1420 } 1421 1422 void 1423 vfs_composefh_free(fhandle_t *fhp) 1424 { 1425 1426 vfs__fhfree(fhp); 1427 } 1428 1429 /* 1430 * vfs_fhtovp: lookup a vnode by a filehandle. 1431 */ 1432 1433 int 1434 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1435 { 1436 struct mount *mp; 1437 int error; 1438 1439 *vpp = NULL; 1440 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1441 if (mp == NULL) { 1442 error = ESTALE; 1443 goto out; 1444 } 1445 if (mp->mnt_op->vfs_fhtovp == NULL) { 1446 error = EOPNOTSUPP; 1447 goto out; 1448 } 1449 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1450 out: 1451 return error; 1452 } 1453 1454 /* 1455 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1456 * the needed size. 1457 */ 1458 1459 int 1460 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1461 { 1462 fhandle_t *fhp; 1463 int error; 1464 1465 *fhpp = NULL; 1466 if (fhsize > FHANDLE_SIZE_MAX) { 1467 return EINVAL; 1468 } 1469 if (fhsize < FHANDLE_SIZE_MIN) { 1470 return EINVAL; 1471 } 1472 again: 1473 fhp = kmem_alloc(fhsize, KM_SLEEP); 1474 if (fhp == NULL) { 1475 return ENOMEM; 1476 } 1477 error = copyin(ufhp, fhp, fhsize); 1478 if (error == 0) { 1479 /* XXX this check shouldn't be here */ 1480 if (FHANDLE_SIZE(fhp) == fhsize) { 1481 *fhpp = fhp; 1482 return 0; 1483 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1484 /* 1485 * a kludge for nfsv2 padded handles. 1486 */ 1487 size_t sz; 1488 1489 sz = FHANDLE_SIZE(fhp); 1490 kmem_free(fhp, fhsize); 1491 fhsize = sz; 1492 goto again; 1493 } else { 1494 /* 1495 * userland told us wrong size. 1496 */ 1497 error = EINVAL; 1498 } 1499 } 1500 kmem_free(fhp, fhsize); 1501 return error; 1502 } 1503 1504 void 1505 vfs_copyinfh_free(fhandle_t *fhp) 1506 { 1507 1508 vfs__fhfree(fhp); 1509 } 1510 1511 /* 1512 * Get file handle system call 1513 */ 1514 int 1515 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1516 { 1517 struct sys___getfh30_args /* { 1518 syscallarg(char *) fname; 1519 syscallarg(fhandle_t *) fhp; 1520 syscallarg(size_t *) fh_size; 1521 } */ *uap = v; 1522 struct vnode *vp; 1523 fhandle_t *fh; 1524 int error; 1525 struct nameidata nd; 1526 size_t sz; 1527 size_t usz; 1528 1529 /* 1530 * Must be super user 1531 */ 1532 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1533 0, NULL, NULL, NULL); 1534 if (error) 1535 return (error); 1536 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1537 SCARG(uap, fname), l); 1538 error = namei(&nd); 1539 if (error) 1540 return (error); 1541 vp = nd.ni_vp; 1542 error = vfs_composefh_alloc(vp, &fh); 1543 vput(vp); 1544 if (error != 0) { 1545 goto out; 1546 } 1547 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1548 if (error != 0) { 1549 goto out; 1550 } 1551 sz = FHANDLE_SIZE(fh); 1552 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1553 if (error != 0) { 1554 goto out; 1555 } 1556 if (usz >= sz) { 1557 error = copyout(fh, SCARG(uap, fhp), sz); 1558 } else { 1559 error = E2BIG; 1560 } 1561 out: 1562 vfs_composefh_free(fh); 1563 return (error); 1564 } 1565 1566 /* 1567 * Open a file given a file handle. 1568 * 1569 * Check permissions, allocate an open file structure, 1570 * and call the device open routine if any. 1571 */ 1572 1573 int 1574 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1575 register_t *retval) 1576 { 1577 struct filedesc *fdp = l->l_proc->p_fd; 1578 struct file *fp; 1579 struct vnode *vp = NULL; 1580 kauth_cred_t cred = l->l_cred; 1581 struct file *nfp; 1582 int type, indx, error=0; 1583 struct flock lf; 1584 struct vattr va; 1585 fhandle_t *fh; 1586 int flags; 1587 1588 /* 1589 * Must be super user 1590 */ 1591 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1592 0, NULL, NULL, NULL))) 1593 return (error); 1594 1595 flags = FFLAGS(oflags); 1596 if ((flags & (FREAD | FWRITE)) == 0) 1597 return (EINVAL); 1598 if ((flags & O_CREAT)) 1599 return (EINVAL); 1600 /* falloc() will use the file descriptor for us */ 1601 if ((error = falloc(l, &nfp, &indx)) != 0) 1602 return (error); 1603 fp = nfp; 1604 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1605 if (error != 0) { 1606 goto bad; 1607 } 1608 error = vfs_fhtovp(fh, &vp); 1609 if (error != 0) { 1610 goto bad; 1611 } 1612 1613 /* Now do an effective vn_open */ 1614 1615 if (vp->v_type == VSOCK) { 1616 error = EOPNOTSUPP; 1617 goto bad; 1618 } 1619 if (flags & FREAD) { 1620 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1621 goto bad; 1622 } 1623 if (flags & (FWRITE | O_TRUNC)) { 1624 if (vp->v_type == VDIR) { 1625 error = EISDIR; 1626 goto bad; 1627 } 1628 if ((error = vn_writechk(vp)) != 0 || 1629 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1630 goto bad; 1631 } 1632 if (flags & O_TRUNC) { 1633 VOP_UNLOCK(vp, 0); /* XXX */ 1634 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1635 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1636 VATTR_NULL(&va); 1637 va.va_size = 0; 1638 error = VOP_SETATTR(vp, &va, cred, l); 1639 if (error) 1640 goto bad; 1641 } 1642 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1643 goto bad; 1644 if (flags & FWRITE) 1645 vp->v_writecount++; 1646 1647 /* done with modified vn_open, now finish what sys_open does. */ 1648 1649 fp->f_flag = flags & FMASK; 1650 fp->f_type = DTYPE_VNODE; 1651 fp->f_ops = &vnops; 1652 fp->f_data = vp; 1653 if (flags & (O_EXLOCK | O_SHLOCK)) { 1654 lf.l_whence = SEEK_SET; 1655 lf.l_start = 0; 1656 lf.l_len = 0; 1657 if (flags & O_EXLOCK) 1658 lf.l_type = F_WRLCK; 1659 else 1660 lf.l_type = F_RDLCK; 1661 type = F_FLOCK; 1662 if ((flags & FNONBLOCK) == 0) 1663 type |= F_WAIT; 1664 VOP_UNLOCK(vp, 0); 1665 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1666 if (error) { 1667 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1668 FILE_UNUSE(fp, l); 1669 ffree(fp); 1670 fdremove(fdp, indx); 1671 return (error); 1672 } 1673 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1674 fp->f_flag |= FHASLOCK; 1675 } 1676 VOP_UNLOCK(vp, 0); 1677 *retval = indx; 1678 FILE_SET_MATURE(fp); 1679 FILE_UNUSE(fp, l); 1680 vfs_copyinfh_free(fh); 1681 return (0); 1682 1683 bad: 1684 FILE_UNUSE(fp, l); 1685 ffree(fp); 1686 fdremove(fdp, indx); 1687 if (vp != NULL) 1688 vput(vp); 1689 vfs_copyinfh_free(fh); 1690 return (error); 1691 } 1692 1693 int 1694 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1695 { 1696 struct sys___fhopen40_args /* { 1697 syscallarg(const void *) fhp; 1698 syscallarg(size_t) fh_size; 1699 syscallarg(int) flags; 1700 } */ *uap = v; 1701 1702 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1703 SCARG(uap, flags), retval); 1704 } 1705 1706 int 1707 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1708 { 1709 int error; 1710 fhandle_t *fh; 1711 struct vnode *vp; 1712 1713 /* 1714 * Must be super user 1715 */ 1716 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1717 0, NULL, NULL, NULL))) 1718 return (error); 1719 1720 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1721 if (error != 0) 1722 return error; 1723 1724 error = vfs_fhtovp(fh, &vp); 1725 vfs_copyinfh_free(fh); 1726 if (error != 0) 1727 return error; 1728 1729 error = vn_stat(vp, sb, l); 1730 vput(vp); 1731 return error; 1732 } 1733 1734 1735 /* ARGSUSED */ 1736 int 1737 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1738 { 1739 struct sys___fhstat40_args /* { 1740 syscallarg(const void *) fhp; 1741 syscallarg(size_t) fh_size; 1742 syscallarg(struct stat *) sb; 1743 } */ *uap = v; 1744 struct stat sb; 1745 int error; 1746 1747 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1748 if (error) 1749 return error; 1750 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1751 } 1752 1753 int 1754 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1755 int flags) 1756 { 1757 fhandle_t *fh; 1758 struct mount *mp; 1759 struct vnode *vp; 1760 int error; 1761 1762 /* 1763 * Must be super user 1764 */ 1765 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1766 0, NULL, NULL, NULL))) 1767 return error; 1768 1769 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1770 if (error != 0) 1771 return error; 1772 1773 error = vfs_fhtovp(fh, &vp); 1774 vfs_copyinfh_free(fh); 1775 if (error != 0) 1776 return error; 1777 1778 mp = vp->v_mount; 1779 error = dostatvfs(mp, sb, l, flags, 1); 1780 vput(vp); 1781 return error; 1782 } 1783 1784 /* ARGSUSED */ 1785 int 1786 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1787 { 1788 struct sys___fhstatvfs140_args /* { 1789 syscallarg(const void *) fhp; 1790 syscallarg(size_t) fh_size; 1791 syscallarg(struct statvfs *) buf; 1792 syscallarg(int) flags; 1793 } */ *uap = v; 1794 struct statvfs *sb = STATVFSBUF_GET(); 1795 int error; 1796 1797 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1798 SCARG(uap, flags)); 1799 if (error == 0) 1800 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1801 STATVFSBUF_PUT(sb); 1802 return error; 1803 } 1804 1805 /* 1806 * Create a special file. 1807 */ 1808 /* ARGSUSED */ 1809 int 1810 sys_mknod(struct lwp *l, void *v, register_t *retval) 1811 { 1812 struct sys_mknod_args /* { 1813 syscallarg(const char *) path; 1814 syscallarg(int) mode; 1815 syscallarg(int) dev; 1816 } */ *uap = v; 1817 struct proc *p = l->l_proc; 1818 struct vnode *vp; 1819 struct vattr vattr; 1820 int error, optype; 1821 struct nameidata nd; 1822 char *path; 1823 const char *cpath; 1824 enum uio_seg seg = UIO_USERSPACE; 1825 1826 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1827 0, NULL, NULL, NULL)) != 0) 1828 return (error); 1829 1830 optype = VOP_MKNOD_DESCOFFSET; 1831 1832 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1833 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1834 1835 if ((error = namei(&nd)) != 0) 1836 goto out; 1837 vp = nd.ni_vp; 1838 if (vp != NULL) 1839 error = EEXIST; 1840 else { 1841 VATTR_NULL(&vattr); 1842 /* We will read cwdi->cwdi_cmask unlocked. */ 1843 vattr.va_mode = 1844 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1845 vattr.va_rdev = SCARG(uap, dev); 1846 1847 switch (SCARG(uap, mode) & S_IFMT) { 1848 case S_IFMT: /* used by badsect to flag bad sectors */ 1849 vattr.va_type = VBAD; 1850 break; 1851 case S_IFCHR: 1852 vattr.va_type = VCHR; 1853 break; 1854 case S_IFBLK: 1855 vattr.va_type = VBLK; 1856 break; 1857 case S_IFWHT: 1858 optype = VOP_WHITEOUT_DESCOFFSET; 1859 break; 1860 case S_IFREG: 1861 #if NVERIEXEC > 0 1862 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1863 O_CREAT); 1864 #endif /* NVERIEXEC > 0 */ 1865 vattr.va_type = VREG; 1866 vattr.va_rdev = VNOVAL; 1867 optype = VOP_CREATE_DESCOFFSET; 1868 break; 1869 default: 1870 error = EINVAL; 1871 break; 1872 } 1873 } 1874 if (!error) { 1875 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1876 switch (optype) { 1877 case VOP_WHITEOUT_DESCOFFSET: 1878 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1879 if (error) 1880 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1881 vput(nd.ni_dvp); 1882 break; 1883 1884 case VOP_MKNOD_DESCOFFSET: 1885 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1886 &nd.ni_cnd, &vattr); 1887 if (error == 0) 1888 vput(nd.ni_vp); 1889 break; 1890 1891 case VOP_CREATE_DESCOFFSET: 1892 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1893 &nd.ni_cnd, &vattr); 1894 if (error == 0) 1895 vput(nd.ni_vp); 1896 break; 1897 } 1898 } else { 1899 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1900 if (nd.ni_dvp == vp) 1901 vrele(nd.ni_dvp); 1902 else 1903 vput(nd.ni_dvp); 1904 if (vp) 1905 vrele(vp); 1906 } 1907 out: 1908 VERIEXEC_PATH_PUT(path); 1909 return (error); 1910 } 1911 1912 /* 1913 * Create a named pipe. 1914 */ 1915 /* ARGSUSED */ 1916 int 1917 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1918 { 1919 struct sys_mkfifo_args /* { 1920 syscallarg(const char *) path; 1921 syscallarg(int) mode; 1922 } */ *uap = v; 1923 struct proc *p = l->l_proc; 1924 struct vattr vattr; 1925 int error; 1926 struct nameidata nd; 1927 1928 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1929 if ((error = namei(&nd)) != 0) 1930 return (error); 1931 if (nd.ni_vp != NULL) { 1932 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1933 if (nd.ni_dvp == nd.ni_vp) 1934 vrele(nd.ni_dvp); 1935 else 1936 vput(nd.ni_dvp); 1937 vrele(nd.ni_vp); 1938 return (EEXIST); 1939 } 1940 VATTR_NULL(&vattr); 1941 vattr.va_type = VFIFO; 1942 /* We will read cwdi->cwdi_cmask unlocked. */ 1943 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1944 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1945 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1946 if (error == 0) 1947 vput(nd.ni_vp); 1948 return (error); 1949 } 1950 1951 /* 1952 * Make a hard file link. 1953 */ 1954 /* ARGSUSED */ 1955 int 1956 sys_link(struct lwp *l, void *v, register_t *retval) 1957 { 1958 struct sys_link_args /* { 1959 syscallarg(const char *) path; 1960 syscallarg(const char *) link; 1961 } */ *uap = v; 1962 struct vnode *vp; 1963 struct nameidata nd; 1964 int error; 1965 1966 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1967 if ((error = namei(&nd)) != 0) 1968 return (error); 1969 vp = nd.ni_vp; 1970 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 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->l_cred, LEASE_WRITE); 1984 VOP_LEASE(vp, l, 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, void *v, register_t *retval) 1997 { 1998 struct sys_symlink_args /* { 1999 syscallarg(const char *) path; 2000 syscallarg(const char *) link; 2001 } */ *uap = v; 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, SCARG(uap, link), l); 2013 if ((error = namei(&nd)) != 0) 2014 goto out; 2015 if (nd.ni_vp) { 2016 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2017 if (nd.ni_dvp == nd.ni_vp) 2018 vrele(nd.ni_dvp); 2019 else 2020 vput(nd.ni_dvp); 2021 vrele(nd.ni_vp); 2022 error = EEXIST; 2023 goto out; 2024 } 2025 VATTR_NULL(&vattr); 2026 vattr.va_type = VLNK; 2027 /* We will read cwdi->cwdi_cmask unlocked. */ 2028 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2029 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2030 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2031 if (error == 0) 2032 vput(nd.ni_vp); 2033 out: 2034 PNBUF_PUT(path); 2035 return (error); 2036 } 2037 2038 /* 2039 * Delete a whiteout from the filesystem. 2040 */ 2041 /* ARGSUSED */ 2042 int 2043 sys_undelete(struct lwp *l, void *v, register_t *retval) 2044 { 2045 struct sys_undelete_args /* { 2046 syscallarg(const char *) path; 2047 } */ *uap = v; 2048 int error; 2049 struct nameidata nd; 2050 2051 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 2052 SCARG(uap, path), l); 2053 error = namei(&nd); 2054 if (error) 2055 return (error); 2056 2057 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2058 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2059 if (nd.ni_dvp == nd.ni_vp) 2060 vrele(nd.ni_dvp); 2061 else 2062 vput(nd.ni_dvp); 2063 if (nd.ni_vp) 2064 vrele(nd.ni_vp); 2065 return (EEXIST); 2066 } 2067 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2068 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2069 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2070 vput(nd.ni_dvp); 2071 return (error); 2072 } 2073 2074 /* 2075 * Delete a name from the filesystem. 2076 */ 2077 /* ARGSUSED */ 2078 int 2079 sys_unlink(struct lwp *l, void *v, register_t *retval) 2080 { 2081 struct sys_unlink_args /* { 2082 syscallarg(const char *) path; 2083 } */ *uap = v; 2084 struct vnode *vp; 2085 int error; 2086 struct nameidata nd; 2087 char *path; 2088 const char *cpath; 2089 enum uio_seg seg = UIO_USERSPACE; 2090 2091 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2092 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2093 2094 if ((error = namei(&nd)) != 0) 2095 goto out; 2096 vp = nd.ni_vp; 2097 2098 /* 2099 * The root of a mounted filesystem cannot be deleted. 2100 */ 2101 if (vp->v_vflag & VV_ROOT) { 2102 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2103 if (nd.ni_dvp == vp) 2104 vrele(nd.ni_dvp); 2105 else 2106 vput(nd.ni_dvp); 2107 vput(vp); 2108 error = EBUSY; 2109 goto out; 2110 } 2111 2112 #if NVERIEXEC > 0 2113 /* Handle remove requests for veriexec entries. */ 2114 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2115 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2116 if (nd.ni_dvp == vp) 2117 vrele(nd.ni_dvp); 2118 else 2119 vput(nd.ni_dvp); 2120 vput(vp); 2121 goto out; 2122 } 2123 #endif /* NVERIEXEC > 0 */ 2124 2125 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2126 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2127 #ifdef FILEASSOC 2128 (void)fileassoc_file_delete(vp); 2129 #endif /* FILEASSOC */ 2130 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2131 out: 2132 VERIEXEC_PATH_PUT(path); 2133 return (error); 2134 } 2135 2136 /* 2137 * Reposition read/write file offset. 2138 */ 2139 int 2140 sys_lseek(struct lwp *l, void *v, register_t *retval) 2141 { 2142 struct sys_lseek_args /* { 2143 syscallarg(int) fd; 2144 syscallarg(int) pad; 2145 syscallarg(off_t) offset; 2146 syscallarg(int) whence; 2147 } */ *uap = v; 2148 struct proc *p = l->l_proc; 2149 kauth_cred_t cred = l->l_cred; 2150 struct filedesc *fdp = p->p_fd; 2151 struct file *fp; 2152 struct vnode *vp; 2153 struct vattr vattr; 2154 off_t newoff; 2155 int error; 2156 2157 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2158 return (EBADF); 2159 2160 vp = (struct vnode *)fp->f_data; 2161 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2162 error = ESPIPE; 2163 mutex_exit(&fp->f_lock); 2164 goto out; 2165 } 2166 2167 switch (SCARG(uap, whence)) { 2168 case SEEK_CUR: 2169 newoff = fp->f_offset + SCARG(uap, offset); 2170 FILE_USE(fp); 2171 break; 2172 case SEEK_END: 2173 FILE_USE(fp); 2174 error = VOP_GETATTR(vp, &vattr, cred, l); 2175 if (error) { 2176 FILE_UNUSE(fp, l); 2177 goto out; 2178 } 2179 newoff = SCARG(uap, offset) + vattr.va_size; 2180 break; 2181 case SEEK_SET: 2182 FILE_USE(fp); 2183 newoff = SCARG(uap, offset); 2184 break; 2185 default: 2186 mutex_exit(&fp->f_lock); 2187 error = EINVAL; 2188 goto out; 2189 } 2190 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2191 mutex_enter(&fp->f_lock); 2192 *(off_t *)retval = fp->f_offset = newoff; 2193 mutex_exit(&fp->f_lock); 2194 } 2195 FILE_UNUSE(fp, l); 2196 out: 2197 return (error); 2198 } 2199 2200 /* 2201 * Positional read system call. 2202 */ 2203 int 2204 sys_pread(struct lwp *l, void *v, register_t *retval) 2205 { 2206 struct sys_pread_args /* { 2207 syscallarg(int) fd; 2208 syscallarg(void *) buf; 2209 syscallarg(size_t) nbyte; 2210 syscallarg(off_t) offset; 2211 } */ *uap = v; 2212 struct proc *p = l->l_proc; 2213 struct filedesc *fdp = p->p_fd; 2214 struct file *fp; 2215 struct vnode *vp; 2216 off_t offset; 2217 int error, fd = SCARG(uap, fd); 2218 2219 if ((fp = fd_getfile(fdp, fd)) == NULL) 2220 return (EBADF); 2221 2222 if ((fp->f_flag & FREAD) == 0) { 2223 mutex_exit(&fp->f_lock); 2224 return (EBADF); 2225 } 2226 2227 FILE_USE(fp); 2228 2229 vp = (struct vnode *)fp->f_data; 2230 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2231 error = ESPIPE; 2232 goto out; 2233 } 2234 2235 offset = SCARG(uap, offset); 2236 2237 /* 2238 * XXX This works because no file systems actually 2239 * XXX take any action on the seek operation. 2240 */ 2241 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2242 goto out; 2243 2244 /* dofileread() will unuse the descriptor for us */ 2245 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2246 &offset, 0, retval)); 2247 2248 out: 2249 FILE_UNUSE(fp, l); 2250 return (error); 2251 } 2252 2253 /* 2254 * Positional scatter read system call. 2255 */ 2256 int 2257 sys_preadv(struct lwp *l, void *v, register_t *retval) 2258 { 2259 struct sys_preadv_args /* { 2260 syscallarg(int) fd; 2261 syscallarg(const struct iovec *) iovp; 2262 syscallarg(int) iovcnt; 2263 syscallarg(off_t) offset; 2264 } */ *uap = v; 2265 2266 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2267 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2268 } 2269 2270 /* 2271 * Positional write system call. 2272 */ 2273 int 2274 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2275 { 2276 struct sys_pwrite_args /* { 2277 syscallarg(int) fd; 2278 syscallarg(const void *) buf; 2279 syscallarg(size_t) nbyte; 2280 syscallarg(off_t) offset; 2281 } */ *uap = v; 2282 struct proc *p = l->l_proc; 2283 struct filedesc *fdp = p->p_fd; 2284 struct file *fp; 2285 struct vnode *vp; 2286 off_t offset; 2287 int error, fd = SCARG(uap, fd); 2288 2289 if ((fp = fd_getfile(fdp, fd)) == NULL) 2290 return (EBADF); 2291 2292 if ((fp->f_flag & FWRITE) == 0) { 2293 mutex_exit(&fp->f_lock); 2294 return (EBADF); 2295 } 2296 2297 FILE_USE(fp); 2298 2299 vp = (struct vnode *)fp->f_data; 2300 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2301 error = ESPIPE; 2302 goto out; 2303 } 2304 2305 offset = SCARG(uap, offset); 2306 2307 /* 2308 * XXX This works because no file systems actually 2309 * XXX take any action on the seek operation. 2310 */ 2311 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2312 goto out; 2313 2314 /* dofilewrite() will unuse the descriptor for us */ 2315 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2316 &offset, 0, retval)); 2317 2318 out: 2319 FILE_UNUSE(fp, l); 2320 return (error); 2321 } 2322 2323 /* 2324 * Positional gather write system call. 2325 */ 2326 int 2327 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2328 { 2329 struct sys_pwritev_args /* { 2330 syscallarg(int) fd; 2331 syscallarg(const struct iovec *) iovp; 2332 syscallarg(int) iovcnt; 2333 syscallarg(off_t) offset; 2334 } */ *uap = v; 2335 2336 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2337 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2338 } 2339 2340 /* 2341 * Check access permissions. 2342 */ 2343 int 2344 sys_access(struct lwp *l, void *v, register_t *retval) 2345 { 2346 struct sys_access_args /* { 2347 syscallarg(const char *) path; 2348 syscallarg(int) flags; 2349 } */ *uap = v; 2350 kauth_cred_t cred; 2351 struct vnode *vp; 2352 int error, flags; 2353 struct nameidata nd; 2354 2355 cred = kauth_cred_dup(l->l_cred); 2356 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2357 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2358 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2359 SCARG(uap, path), l); 2360 /* Override default credentials */ 2361 nd.ni_cnd.cn_cred = cred; 2362 if ((error = namei(&nd)) != 0) 2363 goto out; 2364 vp = nd.ni_vp; 2365 2366 /* Flags == 0 means only check for existence. */ 2367 if (SCARG(uap, flags)) { 2368 flags = 0; 2369 if (SCARG(uap, flags) & R_OK) 2370 flags |= VREAD; 2371 if (SCARG(uap, flags) & W_OK) 2372 flags |= VWRITE; 2373 if (SCARG(uap, flags) & X_OK) 2374 flags |= VEXEC; 2375 2376 error = VOP_ACCESS(vp, flags, cred, l); 2377 if (!error && (flags & VWRITE)) 2378 error = vn_writechk(vp); 2379 } 2380 vput(vp); 2381 out: 2382 kauth_cred_free(cred); 2383 return (error); 2384 } 2385 2386 /* 2387 * Common code for all sys_stat functions, including compat versions. 2388 */ 2389 int 2390 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2391 struct stat *sb) 2392 { 2393 int error; 2394 struct nameidata nd; 2395 2396 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2397 error = namei(&nd); 2398 if (error != 0) 2399 return error; 2400 error = vn_stat(nd.ni_vp, sb, l); 2401 vput(nd.ni_vp); 2402 return error; 2403 } 2404 2405 /* 2406 * Get file status; this version follows links. 2407 */ 2408 /* ARGSUSED */ 2409 int 2410 sys___stat30(struct lwp *l, void *v, register_t *retval) 2411 { 2412 struct sys___stat30_args /* { 2413 syscallarg(const char *) path; 2414 syscallarg(struct stat *) ub; 2415 } */ *uap = v; 2416 struct stat sb; 2417 int error; 2418 2419 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2420 if (error) 2421 return error; 2422 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2423 } 2424 2425 /* 2426 * Get file status; this version does not follow links. 2427 */ 2428 /* ARGSUSED */ 2429 int 2430 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2431 { 2432 struct sys___lstat30_args /* { 2433 syscallarg(const char *) path; 2434 syscallarg(struct stat *) ub; 2435 } */ *uap = v; 2436 struct stat sb; 2437 int error; 2438 2439 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2440 if (error) 2441 return error; 2442 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2443 } 2444 2445 /* 2446 * Get configurable pathname variables. 2447 */ 2448 /* ARGSUSED */ 2449 int 2450 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2451 { 2452 struct sys_pathconf_args /* { 2453 syscallarg(const char *) path; 2454 syscallarg(int) name; 2455 } */ *uap = v; 2456 int error; 2457 struct nameidata nd; 2458 2459 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2460 SCARG(uap, path), l); 2461 if ((error = namei(&nd)) != 0) 2462 return (error); 2463 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2464 vput(nd.ni_vp); 2465 return (error); 2466 } 2467 2468 /* 2469 * Return target name of a symbolic link. 2470 */ 2471 /* ARGSUSED */ 2472 int 2473 sys_readlink(struct lwp *l, void *v, register_t *retval) 2474 { 2475 struct sys_readlink_args /* { 2476 syscallarg(const char *) path; 2477 syscallarg(char *) buf; 2478 syscallarg(size_t) count; 2479 } */ *uap = v; 2480 struct vnode *vp; 2481 struct iovec aiov; 2482 struct uio auio; 2483 int error; 2484 struct nameidata nd; 2485 2486 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2487 SCARG(uap, path), l); 2488 if ((error = namei(&nd)) != 0) 2489 return (error); 2490 vp = nd.ni_vp; 2491 if (vp->v_type != VLNK) 2492 error = EINVAL; 2493 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2494 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2495 aiov.iov_base = SCARG(uap, buf); 2496 aiov.iov_len = SCARG(uap, count); 2497 auio.uio_iov = &aiov; 2498 auio.uio_iovcnt = 1; 2499 auio.uio_offset = 0; 2500 auio.uio_rw = UIO_READ; 2501 KASSERT(l == curlwp); 2502 auio.uio_vmspace = l->l_proc->p_vmspace; 2503 auio.uio_resid = SCARG(uap, count); 2504 error = VOP_READLINK(vp, &auio, l->l_cred); 2505 } 2506 vput(vp); 2507 *retval = SCARG(uap, count) - auio.uio_resid; 2508 return (error); 2509 } 2510 2511 /* 2512 * Change flags of a file given a path name. 2513 */ 2514 /* ARGSUSED */ 2515 int 2516 sys_chflags(struct lwp *l, void *v, register_t *retval) 2517 { 2518 struct sys_chflags_args /* { 2519 syscallarg(const char *) path; 2520 syscallarg(u_long) flags; 2521 } */ *uap = v; 2522 struct vnode *vp; 2523 int error; 2524 struct nameidata nd; 2525 2526 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2527 if ((error = namei(&nd)) != 0) 2528 return (error); 2529 vp = nd.ni_vp; 2530 error = change_flags(vp, SCARG(uap, flags), l); 2531 vput(vp); 2532 return (error); 2533 } 2534 2535 /* 2536 * Change flags of a file given a file descriptor. 2537 */ 2538 /* ARGSUSED */ 2539 int 2540 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2541 { 2542 struct sys_fchflags_args /* { 2543 syscallarg(int) fd; 2544 syscallarg(u_long) flags; 2545 } */ *uap = v; 2546 struct proc *p = l->l_proc; 2547 struct vnode *vp; 2548 struct file *fp; 2549 int error; 2550 2551 /* getvnode() will use the descriptor for us */ 2552 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2553 return (error); 2554 vp = (struct vnode *)fp->f_data; 2555 error = change_flags(vp, SCARG(uap, flags), l); 2556 VOP_UNLOCK(vp, 0); 2557 FILE_UNUSE(fp, l); 2558 return (error); 2559 } 2560 2561 /* 2562 * Change flags of a file given a path name; this version does 2563 * not follow links. 2564 */ 2565 int 2566 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2567 { 2568 struct sys_lchflags_args /* { 2569 syscallarg(const char *) path; 2570 syscallarg(u_long) flags; 2571 } */ *uap = v; 2572 struct vnode *vp; 2573 int error; 2574 struct nameidata nd; 2575 2576 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2577 if ((error = namei(&nd)) != 0) 2578 return (error); 2579 vp = nd.ni_vp; 2580 error = change_flags(vp, SCARG(uap, flags), l); 2581 vput(vp); 2582 return (error); 2583 } 2584 2585 /* 2586 * Common routine to change flags of a file. 2587 */ 2588 int 2589 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2590 { 2591 struct vattr vattr; 2592 int error; 2593 2594 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2595 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2596 /* 2597 * Non-superusers cannot change the flags on devices, even if they 2598 * own them. 2599 */ 2600 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2601 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2602 goto out; 2603 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2604 error = EINVAL; 2605 goto out; 2606 } 2607 } 2608 VATTR_NULL(&vattr); 2609 vattr.va_flags = flags; 2610 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2611 out: 2612 return (error); 2613 } 2614 2615 /* 2616 * Change mode of a file given path name; this version follows links. 2617 */ 2618 /* ARGSUSED */ 2619 int 2620 sys_chmod(struct lwp *l, void *v, register_t *retval) 2621 { 2622 struct sys_chmod_args /* { 2623 syscallarg(const char *) path; 2624 syscallarg(int) mode; 2625 } */ *uap = v; 2626 int error; 2627 struct nameidata nd; 2628 2629 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2630 if ((error = namei(&nd)) != 0) 2631 return (error); 2632 2633 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2634 2635 vrele(nd.ni_vp); 2636 return (error); 2637 } 2638 2639 /* 2640 * Change mode of a file given a file descriptor. 2641 */ 2642 /* ARGSUSED */ 2643 int 2644 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2645 { 2646 struct sys_fchmod_args /* { 2647 syscallarg(int) fd; 2648 syscallarg(int) mode; 2649 } */ *uap = v; 2650 struct proc *p = l->l_proc; 2651 struct file *fp; 2652 int error; 2653 2654 /* getvnode() will use the descriptor for us */ 2655 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2656 return (error); 2657 2658 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2659 FILE_UNUSE(fp, l); 2660 return (error); 2661 } 2662 2663 /* 2664 * Change mode of a file given path name; this version does not follow links. 2665 */ 2666 /* ARGSUSED */ 2667 int 2668 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2669 { 2670 struct sys_lchmod_args /* { 2671 syscallarg(const char *) path; 2672 syscallarg(int) mode; 2673 } */ *uap = v; 2674 int error; 2675 struct nameidata nd; 2676 2677 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2678 if ((error = namei(&nd)) != 0) 2679 return (error); 2680 2681 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2682 2683 vrele(nd.ni_vp); 2684 return (error); 2685 } 2686 2687 /* 2688 * Common routine to set mode given a vnode. 2689 */ 2690 static int 2691 change_mode(struct vnode *vp, int mode, struct lwp *l) 2692 { 2693 struct vattr vattr; 2694 int error; 2695 2696 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2697 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2698 VATTR_NULL(&vattr); 2699 vattr.va_mode = mode & ALLPERMS; 2700 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2701 VOP_UNLOCK(vp, 0); 2702 return (error); 2703 } 2704 2705 /* 2706 * Set ownership given a path name; this version follows links. 2707 */ 2708 /* ARGSUSED */ 2709 int 2710 sys_chown(struct lwp *l, void *v, register_t *retval) 2711 { 2712 struct sys_chown_args /* { 2713 syscallarg(const char *) path; 2714 syscallarg(uid_t) uid; 2715 syscallarg(gid_t) gid; 2716 } */ *uap = v; 2717 int error; 2718 struct nameidata nd; 2719 2720 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2721 if ((error = namei(&nd)) != 0) 2722 return (error); 2723 2724 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2725 2726 vrele(nd.ni_vp); 2727 return (error); 2728 } 2729 2730 /* 2731 * Set ownership given a path name; this version follows links. 2732 * Provides POSIX semantics. 2733 */ 2734 /* ARGSUSED */ 2735 int 2736 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2737 { 2738 struct sys_chown_args /* { 2739 syscallarg(const char *) path; 2740 syscallarg(uid_t) uid; 2741 syscallarg(gid_t) gid; 2742 } */ *uap = v; 2743 int error; 2744 struct nameidata nd; 2745 2746 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2747 if ((error = namei(&nd)) != 0) 2748 return (error); 2749 2750 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2751 2752 vrele(nd.ni_vp); 2753 return (error); 2754 } 2755 2756 /* 2757 * Set ownership given a file descriptor. 2758 */ 2759 /* ARGSUSED */ 2760 int 2761 sys_fchown(struct lwp *l, void *v, register_t *retval) 2762 { 2763 struct sys_fchown_args /* { 2764 syscallarg(int) fd; 2765 syscallarg(uid_t) uid; 2766 syscallarg(gid_t) gid; 2767 } */ *uap = v; 2768 struct proc *p = l->l_proc; 2769 int error; 2770 struct file *fp; 2771 2772 /* getvnode() will use the descriptor for us */ 2773 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2774 return (error); 2775 2776 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2777 SCARG(uap, gid), l, 0); 2778 FILE_UNUSE(fp, l); 2779 return (error); 2780 } 2781 2782 /* 2783 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2784 */ 2785 /* ARGSUSED */ 2786 int 2787 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2788 { 2789 struct sys_fchown_args /* { 2790 syscallarg(int) fd; 2791 syscallarg(uid_t) uid; 2792 syscallarg(gid_t) gid; 2793 } */ *uap = v; 2794 struct proc *p = l->l_proc; 2795 int error; 2796 struct file *fp; 2797 2798 /* getvnode() will use the descriptor for us */ 2799 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2800 return (error); 2801 2802 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2803 SCARG(uap, gid), l, 1); 2804 FILE_UNUSE(fp, l); 2805 return (error); 2806 } 2807 2808 /* 2809 * Set ownership given a path name; this version does not follow links. 2810 */ 2811 /* ARGSUSED */ 2812 int 2813 sys_lchown(struct lwp *l, void *v, register_t *retval) 2814 { 2815 struct sys_lchown_args /* { 2816 syscallarg(const char *) path; 2817 syscallarg(uid_t) uid; 2818 syscallarg(gid_t) gid; 2819 } */ *uap = v; 2820 int error; 2821 struct nameidata nd; 2822 2823 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2824 if ((error = namei(&nd)) != 0) 2825 return (error); 2826 2827 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2828 2829 vrele(nd.ni_vp); 2830 return (error); 2831 } 2832 2833 /* 2834 * Set ownership given a path name; this version does not follow links. 2835 * Provides POSIX/XPG semantics. 2836 */ 2837 /* ARGSUSED */ 2838 int 2839 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2840 { 2841 struct sys_lchown_args /* { 2842 syscallarg(const char *) path; 2843 syscallarg(uid_t) uid; 2844 syscallarg(gid_t) gid; 2845 } */ *uap = v; 2846 int error; 2847 struct nameidata nd; 2848 2849 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2850 if ((error = namei(&nd)) != 0) 2851 return (error); 2852 2853 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2854 2855 vrele(nd.ni_vp); 2856 return (error); 2857 } 2858 2859 /* 2860 * Common routine to set ownership given a vnode. 2861 */ 2862 static int 2863 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2864 int posix_semantics) 2865 { 2866 struct vattr vattr; 2867 mode_t newmode; 2868 int error; 2869 2870 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2871 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2872 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2873 goto out; 2874 2875 #define CHANGED(x) ((int)(x) != -1) 2876 newmode = vattr.va_mode; 2877 if (posix_semantics) { 2878 /* 2879 * POSIX/XPG semantics: if the caller is not the super-user, 2880 * clear set-user-id and set-group-id bits. Both POSIX and 2881 * the XPG consider the behaviour for calls by the super-user 2882 * implementation-defined; we leave the set-user-id and set- 2883 * group-id settings intact in that case. 2884 */ 2885 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2886 NULL) != 0) 2887 newmode &= ~(S_ISUID | S_ISGID); 2888 } else { 2889 /* 2890 * NetBSD semantics: when changing owner and/or group, 2891 * clear the respective bit(s). 2892 */ 2893 if (CHANGED(uid)) 2894 newmode &= ~S_ISUID; 2895 if (CHANGED(gid)) 2896 newmode &= ~S_ISGID; 2897 } 2898 /* Update va_mode iff altered. */ 2899 if (vattr.va_mode == newmode) 2900 newmode = VNOVAL; 2901 2902 VATTR_NULL(&vattr); 2903 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2904 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2905 vattr.va_mode = newmode; 2906 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2907 #undef CHANGED 2908 2909 out: 2910 VOP_UNLOCK(vp, 0); 2911 return (error); 2912 } 2913 2914 /* 2915 * Set the access and modification times given a path name; this 2916 * version follows links. 2917 */ 2918 /* ARGSUSED */ 2919 int 2920 sys_utimes(struct lwp *l, void *v, register_t *retval) 2921 { 2922 struct sys_utimes_args /* { 2923 syscallarg(const char *) path; 2924 syscallarg(const struct timeval *) tptr; 2925 } */ *uap = v; 2926 2927 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2928 SCARG(uap, tptr), UIO_USERSPACE); 2929 } 2930 2931 /* 2932 * Set the access and modification times given a file descriptor. 2933 */ 2934 /* ARGSUSED */ 2935 int 2936 sys_futimes(struct lwp *l, void *v, register_t *retval) 2937 { 2938 struct sys_futimes_args /* { 2939 syscallarg(int) fd; 2940 syscallarg(const struct timeval *) tptr; 2941 } */ *uap = v; 2942 int error; 2943 struct file *fp; 2944 2945 /* getvnode() will use the descriptor for us */ 2946 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2947 return (error); 2948 2949 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2950 SCARG(uap, tptr), UIO_USERSPACE); 2951 2952 FILE_UNUSE(fp, l); 2953 return (error); 2954 } 2955 2956 /* 2957 * Set the access and modification times given a path name; this 2958 * version does not follow links. 2959 */ 2960 int 2961 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2962 { 2963 struct sys_lutimes_args /* { 2964 syscallarg(const char *) path; 2965 syscallarg(const struct timeval *) tptr; 2966 } */ *uap = v; 2967 2968 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2969 SCARG(uap, tptr), UIO_USERSPACE); 2970 } 2971 2972 /* 2973 * Common routine to set access and modification times given a vnode. 2974 */ 2975 int 2976 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2977 const struct timeval *tptr, enum uio_seg seg) 2978 { 2979 struct vattr vattr; 2980 struct nameidata nd; 2981 int error; 2982 2983 VATTR_NULL(&vattr); 2984 if (tptr == NULL) { 2985 nanotime(&vattr.va_atime); 2986 vattr.va_mtime = vattr.va_atime; 2987 vattr.va_vaflags |= VA_UTIMES_NULL; 2988 } else { 2989 struct timeval tv[2]; 2990 2991 if (seg != UIO_SYSSPACE) { 2992 error = copyin(tptr, &tv, sizeof (tv)); 2993 if (error != 0) 2994 return error; 2995 tptr = tv; 2996 } 2997 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 2998 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 2999 } 3000 3001 if (vp == NULL) { 3002 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 3003 if ((error = namei(&nd)) != 0) 3004 return (error); 3005 vp = nd.ni_vp; 3006 } else 3007 nd.ni_vp = NULL; 3008 3009 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3010 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3011 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3012 VOP_UNLOCK(vp, 0); 3013 3014 if (nd.ni_vp != NULL) 3015 vrele(nd.ni_vp); 3016 3017 return (error); 3018 } 3019 3020 /* 3021 * Truncate a file given its path name. 3022 */ 3023 /* ARGSUSED */ 3024 int 3025 sys_truncate(struct lwp *l, void *v, register_t *retval) 3026 { 3027 struct sys_truncate_args /* { 3028 syscallarg(const char *) path; 3029 syscallarg(int) pad; 3030 syscallarg(off_t) length; 3031 } */ *uap = v; 3032 struct vnode *vp; 3033 struct vattr vattr; 3034 int error; 3035 struct nameidata nd; 3036 3037 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3038 if ((error = namei(&nd)) != 0) 3039 return (error); 3040 vp = nd.ni_vp; 3041 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3042 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3043 if (vp->v_type == VDIR) 3044 error = EISDIR; 3045 else if ((error = vn_writechk(vp)) == 0 && 3046 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3047 VATTR_NULL(&vattr); 3048 vattr.va_size = SCARG(uap, length); 3049 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3050 } 3051 vput(vp); 3052 return (error); 3053 } 3054 3055 /* 3056 * Truncate a file given a file descriptor. 3057 */ 3058 /* ARGSUSED */ 3059 int 3060 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3061 { 3062 struct sys_ftruncate_args /* { 3063 syscallarg(int) fd; 3064 syscallarg(int) pad; 3065 syscallarg(off_t) length; 3066 } */ *uap = v; 3067 struct proc *p = l->l_proc; 3068 struct vattr vattr; 3069 struct vnode *vp; 3070 struct file *fp; 3071 int error; 3072 3073 /* getvnode() will use the descriptor for us */ 3074 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3075 return (error); 3076 if ((fp->f_flag & FWRITE) == 0) { 3077 error = EINVAL; 3078 goto out; 3079 } 3080 vp = (struct vnode *)fp->f_data; 3081 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3082 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3083 if (vp->v_type == VDIR) 3084 error = EISDIR; 3085 else if ((error = vn_writechk(vp)) == 0) { 3086 VATTR_NULL(&vattr); 3087 vattr.va_size = SCARG(uap, length); 3088 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3089 } 3090 VOP_UNLOCK(vp, 0); 3091 out: 3092 FILE_UNUSE(fp, l); 3093 return (error); 3094 } 3095 3096 /* 3097 * Sync an open file. 3098 */ 3099 /* ARGSUSED */ 3100 int 3101 sys_fsync(struct lwp *l, void *v, register_t *retval) 3102 { 3103 struct sys_fsync_args /* { 3104 syscallarg(int) fd; 3105 } */ *uap = v; 3106 struct proc *p = l->l_proc; 3107 struct vnode *vp; 3108 struct file *fp; 3109 int error; 3110 3111 /* getvnode() will use the descriptor for us */ 3112 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3113 return (error); 3114 vp = (struct vnode *)fp->f_data; 3115 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3116 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3117 if (error == 0 && bioopsp != NULL && 3118 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3119 (*bioopsp->io_fsync)(vp, 0); 3120 VOP_UNLOCK(vp, 0); 3121 FILE_UNUSE(fp, l); 3122 return (error); 3123 } 3124 3125 /* 3126 * Sync a range of file data. API modeled after that found in AIX. 3127 * 3128 * FDATASYNC indicates that we need only save enough metadata to be able 3129 * to re-read the written data. Note we duplicate AIX's requirement that 3130 * the file be open for writing. 3131 */ 3132 /* ARGSUSED */ 3133 int 3134 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3135 { 3136 struct sys_fsync_range_args /* { 3137 syscallarg(int) fd; 3138 syscallarg(int) flags; 3139 syscallarg(off_t) start; 3140 syscallarg(off_t) length; 3141 } */ *uap = v; 3142 struct proc *p = l->l_proc; 3143 struct vnode *vp; 3144 struct file *fp; 3145 int flags, nflags; 3146 off_t s, e, len; 3147 int error; 3148 3149 /* getvnode() will use the descriptor for us */ 3150 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3151 return (error); 3152 3153 if ((fp->f_flag & FWRITE) == 0) { 3154 error = EBADF; 3155 goto out; 3156 } 3157 3158 flags = SCARG(uap, flags); 3159 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3160 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3161 error = EINVAL; 3162 goto out; 3163 } 3164 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3165 if (flags & FDATASYNC) 3166 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3167 else 3168 nflags = FSYNC_WAIT; 3169 if (flags & FDISKSYNC) 3170 nflags |= FSYNC_CACHE; 3171 3172 len = SCARG(uap, length); 3173 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3174 if (len) { 3175 s = SCARG(uap, start); 3176 e = s + len; 3177 if (e < s) { 3178 error = EINVAL; 3179 goto out; 3180 } 3181 } else { 3182 e = 0; 3183 s = 0; 3184 } 3185 3186 vp = (struct vnode *)fp->f_data; 3187 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3188 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3189 3190 if (error == 0 && bioopsp != NULL && 3191 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3192 (*bioopsp->io_fsync)(vp, nflags); 3193 3194 VOP_UNLOCK(vp, 0); 3195 out: 3196 FILE_UNUSE(fp, l); 3197 return (error); 3198 } 3199 3200 /* 3201 * Sync the data of an open file. 3202 */ 3203 /* ARGSUSED */ 3204 int 3205 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3206 { 3207 struct sys_fdatasync_args /* { 3208 syscallarg(int) fd; 3209 } */ *uap = v; 3210 struct proc *p = l->l_proc; 3211 struct vnode *vp; 3212 struct file *fp; 3213 int error; 3214 3215 /* getvnode() will use the descriptor for us */ 3216 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3217 return (error); 3218 if ((fp->f_flag & FWRITE) == 0) { 3219 FILE_UNUSE(fp, l); 3220 return (EBADF); 3221 } 3222 vp = (struct vnode *)fp->f_data; 3223 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3224 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3225 VOP_UNLOCK(vp, 0); 3226 FILE_UNUSE(fp, l); 3227 return (error); 3228 } 3229 3230 /* 3231 * Rename files, (standard) BSD semantics frontend. 3232 */ 3233 /* ARGSUSED */ 3234 int 3235 sys_rename(struct lwp *l, void *v, register_t *retval) 3236 { 3237 struct sys_rename_args /* { 3238 syscallarg(const char *) from; 3239 syscallarg(const char *) to; 3240 } */ *uap = v; 3241 3242 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3243 } 3244 3245 /* 3246 * Rename files, POSIX semantics frontend. 3247 */ 3248 /* ARGSUSED */ 3249 int 3250 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3251 { 3252 struct sys___posix_rename_args /* { 3253 syscallarg(const char *) from; 3254 syscallarg(const char *) to; 3255 } */ *uap = v; 3256 3257 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3258 } 3259 3260 /* 3261 * Rename files. Source and destination must either both be directories, 3262 * or both not be directories. If target is a directory, it must be empty. 3263 * If `from' and `to' refer to the same object, the value of the `retain' 3264 * argument is used to determine whether `from' will be 3265 * 3266 * (retain == 0) deleted unless `from' and `to' refer to the same 3267 * object in the file system's name space (BSD). 3268 * (retain == 1) always retained (POSIX). 3269 */ 3270 static int 3271 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3272 { 3273 struct vnode *tvp, *fvp, *tdvp; 3274 struct nameidata fromnd, tond; 3275 struct proc *p; 3276 int error; 3277 3278 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3279 from, l); 3280 if ((error = namei(&fromnd)) != 0) 3281 return (error); 3282 if (fromnd.ni_dvp != fromnd.ni_vp) 3283 VOP_UNLOCK(fromnd.ni_dvp, 0); 3284 fvp = fromnd.ni_vp; 3285 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3286 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3287 if ((error = namei(&tond)) != 0) { 3288 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3289 vrele(fromnd.ni_dvp); 3290 vrele(fvp); 3291 goto out1; 3292 } 3293 tdvp = tond.ni_dvp; 3294 tvp = tond.ni_vp; 3295 3296 if (tvp != NULL) { 3297 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3298 error = ENOTDIR; 3299 goto out; 3300 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3301 error = EISDIR; 3302 goto out; 3303 } 3304 } 3305 3306 if (fvp == tdvp) 3307 error = EINVAL; 3308 3309 /* 3310 * Source and destination refer to the same object. 3311 */ 3312 if (fvp == tvp) { 3313 if (retain) 3314 error = -1; 3315 else if (fromnd.ni_dvp == tdvp && 3316 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3317 !memcmp(fromnd.ni_cnd.cn_nameptr, 3318 tond.ni_cnd.cn_nameptr, 3319 fromnd.ni_cnd.cn_namelen)) 3320 error = -1; 3321 } 3322 3323 #if NVERIEXEC > 0 3324 if (!error) { 3325 char *f1, *f2; 3326 3327 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3328 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3329 3330 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3331 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3332 3333 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3334 3335 free(f1, M_TEMP); 3336 free(f2, M_TEMP); 3337 } 3338 #endif /* NVERIEXEC > 0 */ 3339 3340 out: 3341 p = l->l_proc; 3342 if (!error) { 3343 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3344 if (fromnd.ni_dvp != tdvp) 3345 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3346 if (tvp) { 3347 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3348 } 3349 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3350 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3351 } else { 3352 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3353 if (tdvp == tvp) 3354 vrele(tdvp); 3355 else 3356 vput(tdvp); 3357 if (tvp) 3358 vput(tvp); 3359 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3360 vrele(fromnd.ni_dvp); 3361 vrele(fvp); 3362 } 3363 vrele(tond.ni_startdir); 3364 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3365 out1: 3366 if (fromnd.ni_startdir) 3367 vrele(fromnd.ni_startdir); 3368 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3369 return (error == -1 ? 0 : error); 3370 } 3371 3372 /* 3373 * Make a directory file. 3374 */ 3375 /* ARGSUSED */ 3376 int 3377 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3378 { 3379 struct sys_mkdir_args /* { 3380 syscallarg(const char *) path; 3381 syscallarg(int) mode; 3382 } */ *uap = v; 3383 struct proc *p = l->l_proc; 3384 struct vnode *vp; 3385 struct vattr vattr; 3386 int error; 3387 struct nameidata nd; 3388 3389 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3390 SCARG(uap, path), l); 3391 if ((error = namei(&nd)) != 0) 3392 return (error); 3393 vp = nd.ni_vp; 3394 if (vp != NULL) { 3395 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3396 if (nd.ni_dvp == vp) 3397 vrele(nd.ni_dvp); 3398 else 3399 vput(nd.ni_dvp); 3400 vrele(vp); 3401 return (EEXIST); 3402 } 3403 VATTR_NULL(&vattr); 3404 vattr.va_type = VDIR; 3405 /* We will read cwdi->cwdi_cmask unlocked. */ 3406 vattr.va_mode = 3407 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3408 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3409 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3410 if (!error) 3411 vput(nd.ni_vp); 3412 return (error); 3413 } 3414 3415 /* 3416 * Remove a directory file. 3417 */ 3418 /* ARGSUSED */ 3419 int 3420 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3421 { 3422 struct sys_rmdir_args /* { 3423 syscallarg(const char *) path; 3424 } */ *uap = v; 3425 struct vnode *vp; 3426 int error; 3427 struct nameidata nd; 3428 3429 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3430 SCARG(uap, path), l); 3431 if ((error = namei(&nd)) != 0) 3432 return (error); 3433 vp = nd.ni_vp; 3434 if (vp->v_type != VDIR) { 3435 error = ENOTDIR; 3436 goto out; 3437 } 3438 /* 3439 * No rmdir "." please. 3440 */ 3441 if (nd.ni_dvp == vp) { 3442 error = EINVAL; 3443 goto out; 3444 } 3445 /* 3446 * The root of a mounted filesystem cannot be deleted. 3447 */ 3448 if (vp->v_vflag & VV_ROOT) { 3449 error = EBUSY; 3450 goto out; 3451 } 3452 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3453 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3454 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3455 return (error); 3456 3457 out: 3458 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3459 if (nd.ni_dvp == vp) 3460 vrele(nd.ni_dvp); 3461 else 3462 vput(nd.ni_dvp); 3463 vput(vp); 3464 return (error); 3465 } 3466 3467 /* 3468 * Read a block of directory entries in a file system independent format. 3469 */ 3470 int 3471 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3472 { 3473 struct sys___getdents30_args /* { 3474 syscallarg(int) fd; 3475 syscallarg(char *) buf; 3476 syscallarg(size_t) count; 3477 } */ *uap = v; 3478 struct proc *p = l->l_proc; 3479 struct file *fp; 3480 int error, done; 3481 3482 /* getvnode() will use the descriptor for us */ 3483 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3484 return (error); 3485 if ((fp->f_flag & FREAD) == 0) { 3486 error = EBADF; 3487 goto out; 3488 } 3489 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3490 SCARG(uap, count), &done, l, 0, 0); 3491 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3492 *retval = done; 3493 out: 3494 FILE_UNUSE(fp, l); 3495 return (error); 3496 } 3497 3498 /* 3499 * Set the mode mask for creation of filesystem nodes. 3500 */ 3501 int 3502 sys_umask(struct lwp *l, void *v, register_t *retval) 3503 { 3504 struct sys_umask_args /* { 3505 syscallarg(mode_t) newmask; 3506 } */ *uap = v; 3507 struct proc *p = l->l_proc; 3508 struct cwdinfo *cwdi; 3509 3510 /* 3511 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3512 * important is that we serialize changes to the mask. The 3513 * rw_exit() will issue a write memory barrier on our behalf, 3514 * and force the changes out to other CPUs (as it must use an 3515 * atomic operation, draining the local CPU's store buffers). 3516 */ 3517 cwdi = p->p_cwdi; 3518 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3519 *retval = cwdi->cwdi_cmask; 3520 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3521 rw_exit(&cwdi->cwdi_lock); 3522 3523 return (0); 3524 } 3525 3526 /* 3527 * Void all references to file by ripping underlying filesystem 3528 * away from vnode. 3529 */ 3530 /* ARGSUSED */ 3531 int 3532 sys_revoke(struct lwp *l, void *v, register_t *retval) 3533 { 3534 struct sys_revoke_args /* { 3535 syscallarg(const char *) path; 3536 } */ *uap = v; 3537 struct vnode *vp; 3538 struct vattr vattr; 3539 int error; 3540 bool revoke; 3541 struct nameidata nd; 3542 3543 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3544 if ((error = namei(&nd)) != 0) 3545 return (error); 3546 vp = nd.ni_vp; 3547 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3548 goto out; 3549 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3550 (error = kauth_authorize_generic(l->l_cred, 3551 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3552 goto out; 3553 simple_lock(&vp->v_interlock); 3554 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED | VI_LAYER))); 3555 simple_unlock(&vp->v_interlock); 3556 if (revoke) 3557 VOP_REVOKE(vp, REVOKEALL); 3558 out: 3559 vrele(vp); 3560 return (error); 3561 } 3562 3563 /* 3564 * Convert a user file descriptor to a kernel file entry. 3565 */ 3566 int 3567 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3568 { 3569 struct vnode *vp; 3570 struct file *fp; 3571 3572 if ((fp = fd_getfile(fdp, fd)) == NULL) 3573 return (EBADF); 3574 3575 FILE_USE(fp); 3576 3577 if (fp->f_type != DTYPE_VNODE) { 3578 FILE_UNUSE(fp, NULL); 3579 return (EINVAL); 3580 } 3581 3582 vp = (struct vnode *)fp->f_data; 3583 if (vp->v_type == VBAD) { 3584 FILE_UNUSE(fp, NULL); 3585 return (EBADF); 3586 } 3587 3588 *fpp = fp; 3589 return (0); 3590 } 3591
Indexes created Sat Sep 27 14:09:57 GMT 2025