vfs_syscalls.c revision 1.346
1 /* $NetBSD: vfs_syscalls.c,v 1.346 2008/03/21 21:55:00 ad Exp $ */ 2 3 /*- 4 * Copyright (c) 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the NetBSD 18 * Foundation, Inc. and its contributors. 19 * 4. Neither the name of The NetBSD Foundation nor the names of its 20 * contributors may be used to endorse or promote products derived 21 * from this software without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 24 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 25 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 27 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 33 * POSSIBILITY OF SUCH DAMAGE. 34 */ 35 36 /* 37 * Copyright (c) 1989, 1993 38 * The Regents of the University of California. All rights reserved. 39 * (c) UNIX System Laboratories, Inc. 40 * All or some portions of this file are derived from material licensed 41 * to the University of California by American Telephone and Telegraph 42 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 43 * the permission of UNIX System Laboratories, Inc. 44 * 45 * Redistribution and use in source and binary forms, with or without 46 * modification, are permitted provided that the following conditions 47 * are met: 48 * 1. Redistributions of source code must retain the above copyright 49 * notice, this list of conditions and the following disclaimer. 50 * 2. Redistributions in binary form must reproduce the above copyright 51 * notice, this list of conditions and the following disclaimer in the 52 * documentation and/or other materials provided with the distribution. 53 * 3. Neither the name of the University nor the names of its contributors 54 * may be used to endorse or promote products derived from this software 55 * without specific prior written permission. 56 * 57 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 58 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 59 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 60 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 61 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 62 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 63 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 64 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 65 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 66 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 67 * SUCH DAMAGE. 68 * 69 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.346 2008/03/21 21:55:00 ad Exp $"); 74 75 #include "opt_compat_netbsd.h" 76 #include "opt_compat_43.h" 77 #include "opt_fileassoc.h" 78 #include "fss.h" 79 #include "veriexec.h" 80 81 #include <sys/param.h> 82 #include <sys/systm.h> 83 #include <sys/namei.h> 84 #include <sys/filedesc.h> 85 #include <sys/kernel.h> 86 #include <sys/file.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/proc.h> 91 #include <sys/uio.h> 92 #include <sys/malloc.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/ktrace.h> 99 #ifdef FILEASSOC 100 #include <sys/fileassoc.h> 101 #endif /* FILEASSOC */ 102 #include <sys/verified_exec.h> 103 #include <sys/kauth.h> 104 #include <sys/atomic.h> 105 106 #include <miscfs/genfs/genfs.h> 107 #include <miscfs/syncfs/syncfs.h> 108 #include <miscfs/specfs/specdev.h> 109 110 #ifdef COMPAT_30 111 #include "opt_nfsserver.h" 112 #include <nfs/rpcv2.h> 113 #endif 114 #include <nfs/nfsproto.h> 115 #ifdef COMPAT_30 116 #include <nfs/nfs.h> 117 #include <nfs/nfs_var.h> 118 #endif 119 120 #if NFSS > 0 121 #include <dev/fssvar.h> 122 #endif 123 124 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 125 126 static int change_dir(struct nameidata *, struct lwp *); 127 static int change_flags(struct vnode *, u_long, struct lwp *); 128 static int change_mode(struct vnode *, int, struct lwp *l); 129 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 130 131 void checkdirs(struct vnode *); 132 133 int dovfsusermount = 0; 134 135 /* 136 * Virtual File System System Calls 137 */ 138 139 /* 140 * Mount a file system. 141 */ 142 143 #if defined(COMPAT_09) || defined(COMPAT_43) 144 /* 145 * This table is used to maintain compatibility with 4.3BSD 146 * and NetBSD 0.9 mount syscalls. Note, the order is important! 147 * 148 * Do not modify this table. It should only contain filesystems 149 * supported by NetBSD 0.9 and 4.3BSD. 150 */ 151 const char * const mountcompatnames[] = { 152 NULL, /* 0 = MOUNT_NONE */ 153 MOUNT_FFS, /* 1 = MOUNT_UFS */ 154 MOUNT_NFS, /* 2 */ 155 MOUNT_MFS, /* 3 */ 156 MOUNT_MSDOS, /* 4 */ 157 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 158 MOUNT_FDESC, /* 6 */ 159 MOUNT_KERNFS, /* 7 */ 160 NULL, /* 8 = MOUNT_DEVFS */ 161 MOUNT_AFS, /* 9 */ 162 }; 163 const int nmountcompatnames = sizeof(mountcompatnames) / 164 sizeof(mountcompatnames[0]); 165 #endif /* COMPAT_09 || COMPAT_43 */ 166 167 static int 168 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 169 void *data, size_t *data_len) 170 { 171 struct mount *mp; 172 int error = 0, saved_flags; 173 174 mp = vp->v_mount; 175 saved_flags = mp->mnt_flag; 176 177 /* We can operate only on VV_ROOT nodes. */ 178 if ((vp->v_vflag & VV_ROOT) == 0) { 179 error = EINVAL; 180 goto out; 181 } 182 183 /* 184 * We only allow the filesystem to be reloaded if it 185 * is currently mounted read-only. 186 */ 187 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 188 error = EOPNOTSUPP; /* Needs translation */ 189 goto out; 190 } 191 192 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 193 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 194 if (error) 195 goto out; 196 197 if (vfs_trybusy(mp, RW_WRITER, 0)) { 198 error = EPERM; 199 goto out; 200 } 201 202 mp->mnt_flag &= ~MNT_OP_FLAGS; 203 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 204 205 /* 206 * Set the mount level flags. 207 */ 208 if (flags & MNT_RDONLY) 209 mp->mnt_flag |= MNT_RDONLY; 210 else if (mp->mnt_flag & MNT_RDONLY) 211 mp->mnt_iflag |= IMNT_WANTRDWR; 212 mp->mnt_flag &= 213 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 214 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 215 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 216 mp->mnt_flag |= flags & 217 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 218 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 219 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 220 MNT_IGNORE); 221 222 error = VFS_MOUNT(mp, path, data, data_len); 223 224 #if defined(COMPAT_30) && defined(NFSSERVER) 225 if (error && data != NULL) { 226 int error2; 227 228 /* Update failed; let's try and see if it was an 229 * export request. */ 230 error2 = nfs_update_exports_30(mp, path, data, l); 231 232 /* Only update error code if the export request was 233 * understood but some problem occurred while 234 * processing it. */ 235 if (error2 != EJUSTRETURN) 236 error = error2; 237 } 238 #endif 239 if (mp->mnt_iflag & IMNT_WANTRDWR) 240 mp->mnt_flag &= ~MNT_RDONLY; 241 if (error) 242 mp->mnt_flag = saved_flags; 243 mp->mnt_flag &= ~MNT_OP_FLAGS; 244 mp->mnt_iflag &= ~IMNT_WANTRDWR; 245 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 246 if (mp->mnt_syncer == NULL) 247 error = vfs_allocate_syncvnode(mp); 248 } else { 249 if (mp->mnt_syncer != NULL) 250 vfs_deallocate_syncvnode(mp); 251 } 252 vfs_unbusy(mp, false); 253 254 out: 255 return (error); 256 } 257 258 static int 259 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 260 { 261 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 262 int error; 263 264 /* Copy file-system type from userspace. */ 265 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 266 if (error) { 267 #if defined(COMPAT_09) || defined(COMPAT_43) 268 /* 269 * Historically, filesystem types were identified by numbers. 270 * If we get an integer for the filesystem type instead of a 271 * string, we check to see if it matches one of the historic 272 * filesystem types. 273 */ 274 u_long fsindex = (u_long)fstype; 275 if (fsindex >= nmountcompatnames || 276 mountcompatnames[fsindex] == NULL) 277 return ENODEV; 278 strlcpy(fstypename, mountcompatnames[fsindex], 279 sizeof(fstypename)); 280 #else 281 return error; 282 #endif 283 } 284 285 #ifdef COMPAT_10 286 /* Accept `ufs' as an alias for `ffs'. */ 287 if (strcmp(fstypename, "ufs") == 0) 288 fstypename[0] = 'f'; 289 #endif 290 291 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 292 return ENODEV; 293 return 0; 294 } 295 296 static int 297 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 298 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 299 { 300 struct mount *mp = NULL; 301 struct vnode *vp = *vpp; 302 struct vattr va; 303 int error; 304 305 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 306 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 307 if (error) 308 return error; 309 310 /* Can't make a non-dir a mount-point (from here anyway). */ 311 if (vp->v_type != VDIR) 312 return ENOTDIR; 313 314 /* 315 * If the user is not root, ensure that they own the directory 316 * onto which we are attempting to mount. 317 */ 318 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 319 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 320 (error = kauth_authorize_generic(l->l_cred, 321 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 322 return error; 323 } 324 325 if (flags & MNT_EXPORTED) 326 return EINVAL; 327 328 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 329 return error; 330 331 /* 332 * Check if a file-system is not already mounted on this vnode. 333 */ 334 if (vp->v_mountedhere != NULL) 335 return EBUSY; 336 337 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP); 338 if (mp == NULL) 339 return ENOMEM; 340 341 mp->mnt_op = vfsops; 342 mp->mnt_refcnt = 1; 343 344 TAILQ_INIT(&mp->mnt_vnodelist); 345 rw_init(&mp->mnt_lock); 346 mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE); 347 (void)vfs_busy(mp, RW_WRITER, 0); 348 349 mp->mnt_vnodecovered = vp; 350 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 351 mount_initspecific(mp); 352 353 /* 354 * The underlying file system may refuse the mount for 355 * various reasons. Allow the user to force it to happen. 356 * 357 * Set the mount level flags. 358 */ 359 mp->mnt_flag = flags & 360 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 361 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 362 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 363 MNT_IGNORE | MNT_RDONLY); 364 365 error = VFS_MOUNT(mp, path, data, data_len); 366 mp->mnt_flag &= ~MNT_OP_FLAGS; 367 368 /* 369 * Put the new filesystem on the mount list after root. 370 */ 371 cache_purge(vp); 372 if (error != 0) { 373 vp->v_mountedhere = NULL; 374 mp->mnt_op->vfs_refcount--; 375 vfs_unbusy(mp, false); 376 vfs_destroy(mp); 377 return error; 378 } 379 380 mp->mnt_iflag &= ~IMNT_WANTRDWR; 381 mutex_enter(&mountlist_lock); 382 vp->v_mountedhere = mp; 383 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 384 mutex_exit(&mountlist_lock); 385 vn_restorerecurse(vp, recurse); 386 VOP_UNLOCK(vp, 0); 387 checkdirs(vp); 388 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 389 error = vfs_allocate_syncvnode(mp); 390 vfs_unbusy(mp, false); 391 (void) VFS_STATVFS(mp, &mp->mnt_stat); 392 error = VFS_START(mp, 0); 393 if (error) { 394 vrele(vp); 395 vfs_destroy(mp); 396 } 397 *vpp = NULL; 398 return error; 399 } 400 401 static int 402 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 403 void *data, size_t *data_len) 404 { 405 struct mount *mp; 406 int error; 407 408 /* If MNT_GETARGS is specified, it should be the only flag. */ 409 if (flags & ~MNT_GETARGS) 410 return EINVAL; 411 412 mp = vp->v_mount; 413 414 /* XXX: probably some notion of "can see" here if we want isolation. */ 415 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 416 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 417 if (error) 418 return error; 419 420 if ((vp->v_vflag & VV_ROOT) == 0) 421 return EINVAL; 422 423 if (vfs_trybusy(mp, RW_WRITER, NULL)) 424 return EPERM; 425 426 mp->mnt_flag &= ~MNT_OP_FLAGS; 427 mp->mnt_flag |= MNT_GETARGS; 428 error = VFS_MOUNT(mp, path, data, data_len); 429 mp->mnt_flag &= ~MNT_OP_FLAGS; 430 431 vfs_unbusy(mp, false); 432 return (error); 433 } 434 435 #ifdef COMPAT_40 436 /* ARGSUSED */ 437 int 438 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 439 { 440 /* { 441 syscallarg(const char *) type; 442 syscallarg(const char *) path; 443 syscallarg(int) flags; 444 syscallarg(void *) data; 445 } */ 446 register_t dummy; 447 448 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 449 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 450 } 451 #endif 452 453 int 454 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 455 { 456 /* { 457 syscallarg(const char *) type; 458 syscallarg(const char *) path; 459 syscallarg(int) flags; 460 syscallarg(void *) data; 461 syscallarg(size_t) data_len; 462 } */ 463 464 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 465 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 466 SCARG(uap, data_len), retval); 467 } 468 469 int 470 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 471 const char *path, int flags, void *data, enum uio_seg data_seg, 472 size_t data_len, register_t *retval) 473 { 474 struct vnode *vp; 475 struct nameidata nd; 476 void *data_buf = data; 477 u_int recurse; 478 int error; 479 480 /* 481 * Get vnode to be covered 482 */ 483 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 484 if ((error = namei(&nd)) != 0) 485 return (error); 486 vp = nd.ni_vp; 487 488 /* 489 * A lookup in VFS_MOUNT might result in an attempt to 490 * lock this vnode again, so make the lock recursive. 491 */ 492 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 493 recurse = vn_setrecurse(vp); 494 495 if (vfsops == NULL) { 496 if (flags & (MNT_GETARGS | MNT_UPDATE)) 497 vfsops = vp->v_mount->mnt_op; 498 else { 499 /* 'type' is userspace */ 500 error = mount_get_vfsops(type, &vfsops); 501 if (error != 0) 502 goto done; 503 } 504 } 505 506 if (data != NULL && data_seg == UIO_USERSPACE) { 507 if (data_len == 0) { 508 /* No length supplied, use default for filesystem */ 509 data_len = vfsops->vfs_min_mount_data; 510 if (data_len > VFS_MAX_MOUNT_DATA) { 511 /* maybe a force loaded old LKM */ 512 error = EINVAL; 513 goto done; 514 } 515 #ifdef COMPAT_30 516 /* Hopefully a longer buffer won't make copyin() fail */ 517 if (flags & MNT_UPDATE 518 && data_len < sizeof (struct mnt_export_args30)) 519 data_len = sizeof (struct mnt_export_args30); 520 #endif 521 } 522 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 523 524 /* NFS needs the buffer even for mnt_getargs .... */ 525 error = copyin(data, data_buf, data_len); 526 if (error != 0) 527 goto done; 528 } 529 530 if (flags & MNT_GETARGS) { 531 if (data_len == 0) { 532 error = EINVAL; 533 goto done; 534 } 535 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 536 if (error != 0) 537 goto done; 538 if (data_seg == UIO_USERSPACE) 539 error = copyout(data_buf, data, data_len); 540 *retval = data_len; 541 } else if (flags & MNT_UPDATE) { 542 error = mount_update(l, vp, path, flags, data_buf, &data_len); 543 } else { 544 /* Locking is handled internally in mount_domount(). */ 545 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 546 &data_len, recurse); 547 } 548 549 done: 550 if (vp != NULL) { 551 vn_restorerecurse(vp, recurse); 552 vput(vp); 553 } 554 if (data_buf != data) 555 free(data_buf, M_TEMP); 556 return (error); 557 } 558 559 /* 560 * Scan all active processes to see if any of them have a current 561 * or root directory onto which the new filesystem has just been 562 * mounted. If so, replace them with the new mount point. 563 */ 564 void 565 checkdirs(struct vnode *olddp) 566 { 567 struct cwdinfo *cwdi; 568 struct vnode *newdp; 569 struct proc *p; 570 571 if (olddp->v_usecount == 1) 572 return; 573 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 574 panic("mount: lost mount"); 575 mutex_enter(&proclist_lock); 576 PROCLIST_FOREACH(p, &allproc) { 577 cwdi = p->p_cwdi; 578 if (!cwdi) 579 continue; 580 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 581 if (cwdi->cwdi_cdir == olddp) { 582 vrele(cwdi->cwdi_cdir); 583 VREF(newdp); 584 cwdi->cwdi_cdir = newdp; 585 } 586 if (cwdi->cwdi_rdir == olddp) { 587 vrele(cwdi->cwdi_rdir); 588 VREF(newdp); 589 cwdi->cwdi_rdir = newdp; 590 } 591 rw_exit(&cwdi->cwdi_lock); 592 } 593 mutex_exit(&proclist_lock); 594 if (rootvnode == olddp) { 595 vrele(rootvnode); 596 VREF(newdp); 597 rootvnode = newdp; 598 } 599 vput(newdp); 600 } 601 602 /* 603 * Unmount a file system. 604 * 605 * Note: unmount takes a path to the vnode mounted on as argument, 606 * not special file (as before). 607 */ 608 /* ARGSUSED */ 609 int 610 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 611 { 612 /* { 613 syscallarg(const char *) path; 614 syscallarg(int) flags; 615 } */ 616 struct vnode *vp; 617 struct mount *mp; 618 int error; 619 struct nameidata nd; 620 621 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 622 SCARG(uap, path)); 623 if ((error = namei(&nd)) != 0) 624 return (error); 625 vp = nd.ni_vp; 626 mp = vp->v_mount; 627 VOP_UNLOCK(vp, 0); 628 629 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 630 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 631 if (error) { 632 vrele(vp); 633 return (error); 634 } 635 636 /* 637 * Don't allow unmounting the root file system. 638 */ 639 if (mp->mnt_flag & MNT_ROOTFS) { 640 vrele(vp); 641 return (EINVAL); 642 } 643 644 /* 645 * Must be the root of the filesystem 646 */ 647 if ((vp->v_vflag & VV_ROOT) == 0) { 648 vrele(vp); 649 return (EINVAL); 650 } 651 652 /* 653 * XXX Freeze syncer. Must do this before locking the 654 * mount point. See dounmount() for details. 655 */ 656 mutex_enter(&syncer_mutex); 657 error = vfs_busy(mp, RW_WRITER, NULL); 658 vrele(vp); 659 if (error != 0) { 660 mutex_exit(&syncer_mutex); 661 return (error); 662 } 663 664 return (dounmount(mp, SCARG(uap, flags), l)); 665 } 666 667 /* 668 * Do the actual file system unmount. File system is assumed to have been 669 * marked busy by the caller. 670 */ 671 int 672 dounmount(struct mount *mp, int flags, struct lwp *l) 673 { 674 struct vnode *coveredvp; 675 int error; 676 int async; 677 int used_syncer; 678 679 KASSERT(rw_write_held(&mp->mnt_lock)); 680 681 #if NVERIEXEC > 0 682 error = veriexec_unmountchk(mp); 683 if (error) 684 return (error); 685 #endif /* NVERIEXEC > 0 */ 686 687 used_syncer = (mp->mnt_syncer != NULL); 688 689 /* 690 * XXX Syncer must be frozen when we get here. This should really 691 * be done on a per-mountpoint basis, but especially the softdep 692 * code possibly called from the syncer doesn't exactly work on a 693 * per-mountpoint basis, so the softdep code would become a maze 694 * of vfs_busy() calls. 695 * 696 * The caller of dounmount() must acquire syncer_mutex because 697 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 698 * order, and we must preserve that order to avoid deadlock. 699 * 700 * So, if the file system did not use the syncer, now is 701 * the time to release the syncer_mutex. 702 */ 703 if (used_syncer == 0) 704 mutex_exit(&syncer_mutex); 705 706 mp->mnt_iflag |= IMNT_UNMOUNT; 707 async = mp->mnt_flag & MNT_ASYNC; 708 mp->mnt_flag &= ~MNT_ASYNC; 709 cache_purgevfs(mp); /* remove cache entries for this file sys */ 710 if (mp->mnt_syncer != NULL) 711 vfs_deallocate_syncvnode(mp); 712 error = 0; 713 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 714 #if NFSS > 0 715 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 716 #endif 717 if (error == 0) 718 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 719 } 720 vfs_scrubvnlist(mp); 721 if (error == 0 || (flags & MNT_FORCE)) 722 error = VFS_UNMOUNT(mp, flags); 723 if (error) { 724 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 725 (void) vfs_allocate_syncvnode(mp); 726 mp->mnt_iflag &= ~IMNT_UNMOUNT; 727 mp->mnt_flag |= async; 728 if (used_syncer) 729 mutex_exit(&syncer_mutex); 730 vfs_unbusy(mp, false); 731 return (error); 732 } 733 vfs_scrubvnlist(mp); 734 mutex_enter(&mountlist_lock); 735 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 736 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 737 coveredvp->v_mountedhere = NULL; 738 mutex_exit(&mountlist_lock); 739 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 740 panic("unmount: dangling vnode"); 741 mp->mnt_iflag |= IMNT_GONE; 742 if (used_syncer) 743 mutex_exit(&syncer_mutex); 744 vfs_hooks_unmount(mp); 745 vfs_unbusy(mp, false); 746 vfs_destroy(mp); 747 if (coveredvp != NULLVP) 748 vrele(coveredvp); 749 return (0); 750 } 751 752 /* 753 * Sync each mounted filesystem. 754 */ 755 #ifdef DEBUG 756 int syncprt = 0; 757 struct ctldebug debug0 = { "syncprt", &syncprt }; 758 #endif 759 760 /* ARGSUSED */ 761 int 762 sys_sync(struct lwp *l, const void *v, register_t *retval) 763 { 764 struct mount *mp, *nmp; 765 int asyncflag; 766 767 if (l == NULL) 768 l = &lwp0; 769 770 mutex_enter(&mountlist_lock); 771 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 772 if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) { 773 nmp = mp->mnt_list.cqe_prev; 774 continue; 775 } 776 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 777 asyncflag = mp->mnt_flag & MNT_ASYNC; 778 mp->mnt_flag &= ~MNT_ASYNC; 779 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 780 if (asyncflag) 781 mp->mnt_flag |= MNT_ASYNC; 782 } 783 mutex_enter(&mountlist_lock); 784 nmp = mp->mnt_list.cqe_prev; 785 vfs_unbusy(mp, false); 786 787 } 788 mutex_exit(&mountlist_lock); 789 #ifdef DEBUG 790 if (syncprt) 791 vfs_bufstats(); 792 #endif /* DEBUG */ 793 return (0); 794 } 795 796 /* 797 * Change filesystem quotas. 798 */ 799 /* ARGSUSED */ 800 int 801 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 802 { 803 /* { 804 syscallarg(const char *) path; 805 syscallarg(int) cmd; 806 syscallarg(int) uid; 807 syscallarg(void *) arg; 808 } */ 809 struct mount *mp; 810 int error; 811 struct nameidata nd; 812 813 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 814 SCARG(uap, path)); 815 if ((error = namei(&nd)) != 0) 816 return (error); 817 mp = nd.ni_vp->v_mount; 818 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 819 SCARG(uap, arg)); 820 vrele(nd.ni_vp); 821 return (error); 822 } 823 824 int 825 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 826 int root) 827 { 828 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 829 int error = 0; 830 831 /* 832 * If MNT_NOWAIT or MNT_LAZY is specified, do not 833 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 834 * overrides MNT_NOWAIT. 835 */ 836 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 837 (flags != MNT_WAIT && flags != 0)) { 838 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 839 goto done; 840 } 841 842 /* Get the filesystem stats now */ 843 memset(sp, 0, sizeof(*sp)); 844 if ((error = VFS_STATVFS(mp, sp)) != 0) { 845 return error; 846 } 847 848 if (cwdi->cwdi_rdir == NULL) 849 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 850 done: 851 if (cwdi->cwdi_rdir != NULL) { 852 size_t len; 853 char *bp; 854 char *path = PNBUF_GET(); 855 856 bp = path + MAXPATHLEN; 857 *--bp = '\0'; 858 rw_enter(&cwdi->cwdi_lock, RW_READER); 859 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 860 MAXPATHLEN / 2, 0, l); 861 rw_exit(&cwdi->cwdi_lock); 862 if (error) { 863 PNBUF_PUT(path); 864 return error; 865 } 866 len = strlen(bp); 867 /* 868 * for mount points that are below our root, we can see 869 * them, so we fix up the pathname and return them. The 870 * rest we cannot see, so we don't allow viewing the 871 * data. 872 */ 873 if (strncmp(bp, sp->f_mntonname, len) == 0) { 874 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 875 sizeof(sp->f_mntonname)); 876 if (sp->f_mntonname[0] == '\0') 877 (void)strlcpy(sp->f_mntonname, "/", 878 sizeof(sp->f_mntonname)); 879 } else { 880 if (root) 881 (void)strlcpy(sp->f_mntonname, "/", 882 sizeof(sp->f_mntonname)); 883 else 884 error = EPERM; 885 } 886 PNBUF_PUT(path); 887 } 888 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 889 return error; 890 } 891 892 /* 893 * Get filesystem statistics by path. 894 */ 895 int 896 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 897 { 898 struct mount *mp; 899 int error; 900 struct nameidata nd; 901 902 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 903 if ((error = namei(&nd)) != 0) 904 return error; 905 mp = nd.ni_vp->v_mount; 906 error = dostatvfs(mp, sb, l, flags, 1); 907 vrele(nd.ni_vp); 908 return error; 909 } 910 911 /* ARGSUSED */ 912 int 913 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 914 { 915 /* { 916 syscallarg(const char *) path; 917 syscallarg(struct statvfs *) buf; 918 syscallarg(int) flags; 919 } */ 920 struct statvfs *sb; 921 int error; 922 923 sb = STATVFSBUF_GET(); 924 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 925 if (error == 0) 926 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 927 STATVFSBUF_PUT(sb); 928 return error; 929 } 930 931 /* 932 * Get filesystem statistics by fd. 933 */ 934 int 935 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 936 { 937 file_t *fp; 938 struct mount *mp; 939 int error; 940 941 /* fd_getvnode() will use the descriptor for us */ 942 if ((error = fd_getvnode(fd, &fp)) != 0) 943 return (error); 944 mp = ((struct vnode *)fp->f_data)->v_mount; 945 error = dostatvfs(mp, sb, curlwp, flags, 1); 946 fd_putfile(fd); 947 return error; 948 } 949 950 /* ARGSUSED */ 951 int 952 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 953 { 954 /* { 955 syscallarg(int) fd; 956 syscallarg(struct statvfs *) buf; 957 syscallarg(int) flags; 958 } */ 959 struct statvfs *sb; 960 int error; 961 962 sb = STATVFSBUF_GET(); 963 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 964 if (error == 0) 965 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 966 STATVFSBUF_PUT(sb); 967 return error; 968 } 969 970 971 /* 972 * Get statistics on all filesystems. 973 */ 974 int 975 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 976 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 977 register_t *retval) 978 { 979 int root = 0; 980 struct proc *p = l->l_proc; 981 struct mount *mp, *nmp; 982 struct statvfs *sb; 983 size_t count, maxcount; 984 int error = 0; 985 986 sb = STATVFSBUF_GET(); 987 maxcount = bufsize / entry_sz; 988 mutex_enter(&mountlist_lock); 989 count = 0; 990 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 991 mp = nmp) { 992 if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) { 993 nmp = CIRCLEQ_NEXT(mp, mnt_list); 994 continue; 995 } 996 if (sfsp && count < maxcount) { 997 error = dostatvfs(mp, sb, l, flags, 0); 998 if (error) { 999 mutex_enter(&mountlist_lock); 1000 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1001 vfs_unbusy(mp, false); 1002 continue; 1003 } 1004 error = copyfn(sb, sfsp, entry_sz); 1005 if (error) { 1006 vfs_unbusy(mp, false); 1007 goto out; 1008 } 1009 sfsp = (char *)sfsp + entry_sz; 1010 root |= strcmp(sb->f_mntonname, "/") == 0; 1011 } 1012 count++; 1013 mutex_enter(&mountlist_lock); 1014 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1015 vfs_unbusy(mp, false); 1016 } 1017 1018 mutex_exit(&mountlist_lock); 1019 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1020 /* 1021 * fake a root entry 1022 */ 1023 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1024 sb, l, flags, 1); 1025 if (error != 0) 1026 goto out; 1027 if (sfsp) 1028 error = copyfn(sb, sfsp, entry_sz); 1029 count++; 1030 } 1031 if (sfsp && count > maxcount) 1032 *retval = maxcount; 1033 else 1034 *retval = count; 1035 out: 1036 STATVFSBUF_PUT(sb); 1037 return error; 1038 } 1039 1040 int 1041 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1042 { 1043 /* { 1044 syscallarg(struct statvfs *) buf; 1045 syscallarg(size_t) bufsize; 1046 syscallarg(int) flags; 1047 } */ 1048 1049 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1050 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1051 } 1052 1053 /* 1054 * Change current working directory to a given file descriptor. 1055 */ 1056 /* ARGSUSED */ 1057 int 1058 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1059 { 1060 /* { 1061 syscallarg(int) fd; 1062 } */ 1063 struct proc *p = l->l_proc; 1064 struct cwdinfo *cwdi; 1065 struct vnode *vp, *tdp; 1066 struct mount *mp; 1067 file_t *fp; 1068 int error, fd; 1069 1070 /* fd_getvnode() will use the descriptor for us */ 1071 fd = SCARG(uap, fd); 1072 if ((error = fd_getvnode(fd, &fp)) != 0) 1073 return (error); 1074 vp = fp->f_data; 1075 1076 VREF(vp); 1077 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1078 if (vp->v_type != VDIR) 1079 error = ENOTDIR; 1080 else 1081 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1082 if (error) { 1083 vput(vp); 1084 goto out; 1085 } 1086 while ((mp = vp->v_mountedhere) != NULL) { 1087 if (vfs_busy(mp, RW_READER, NULL)) 1088 continue; 1089 vput(vp); 1090 error = VFS_ROOT(mp, &tdp); 1091 vfs_unbusy(mp, false); 1092 if (error) 1093 goto out; 1094 vp = tdp; 1095 } 1096 VOP_UNLOCK(vp, 0); 1097 1098 /* 1099 * Disallow changing to a directory not under the process's 1100 * current root directory (if there is one). 1101 */ 1102 cwdi = p->p_cwdi; 1103 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1104 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1105 vrele(vp); 1106 error = EPERM; /* operation not permitted */ 1107 } else { 1108 vrele(cwdi->cwdi_cdir); 1109 cwdi->cwdi_cdir = vp; 1110 } 1111 rw_exit(&cwdi->cwdi_lock); 1112 1113 out: 1114 fd_putfile(fd); 1115 return (error); 1116 } 1117 1118 /* 1119 * Change this process's notion of the root directory to a given file 1120 * descriptor. 1121 */ 1122 int 1123 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1124 { 1125 struct proc *p = l->l_proc; 1126 struct cwdinfo *cwdi; 1127 struct vnode *vp; 1128 file_t *fp; 1129 int error, fd = SCARG(uap, fd); 1130 1131 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1132 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1133 return error; 1134 /* fd_getvnode() will use the descriptor for us */ 1135 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 1136 return error; 1137 vp = fp->f_data; 1138 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1139 if (vp->v_type != VDIR) 1140 error = ENOTDIR; 1141 else 1142 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1143 VOP_UNLOCK(vp, 0); 1144 if (error) 1145 goto out; 1146 VREF(vp); 1147 1148 /* 1149 * Prevent escaping from chroot by putting the root under 1150 * the working directory. Silently chdir to / if we aren't 1151 * already there. 1152 */ 1153 cwdi = p->p_cwdi; 1154 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1155 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1156 /* 1157 * XXX would be more failsafe to change directory to a 1158 * deadfs node here instead 1159 */ 1160 vrele(cwdi->cwdi_cdir); 1161 VREF(vp); 1162 cwdi->cwdi_cdir = vp; 1163 } 1164 1165 if (cwdi->cwdi_rdir != NULL) 1166 vrele(cwdi->cwdi_rdir); 1167 cwdi->cwdi_rdir = vp; 1168 rw_exit(&cwdi->cwdi_lock); 1169 1170 out: 1171 fd_putfile(fd); 1172 return (error); 1173 } 1174 1175 /* 1176 * Change current working directory (``.''). 1177 */ 1178 /* ARGSUSED */ 1179 int 1180 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1181 { 1182 /* { 1183 syscallarg(const char *) path; 1184 } */ 1185 struct proc *p = l->l_proc; 1186 struct cwdinfo *cwdi; 1187 int error; 1188 struct nameidata nd; 1189 1190 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1191 SCARG(uap, path)); 1192 if ((error = change_dir(&nd, l)) != 0) 1193 return (error); 1194 cwdi = p->p_cwdi; 1195 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1196 vrele(cwdi->cwdi_cdir); 1197 cwdi->cwdi_cdir = nd.ni_vp; 1198 rw_exit(&cwdi->cwdi_lock); 1199 return (0); 1200 } 1201 1202 /* 1203 * Change notion of root (``/'') directory. 1204 */ 1205 /* ARGSUSED */ 1206 int 1207 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1208 { 1209 /* { 1210 syscallarg(const char *) path; 1211 } */ 1212 struct proc *p = l->l_proc; 1213 struct cwdinfo *cwdi; 1214 struct vnode *vp; 1215 int error; 1216 struct nameidata nd; 1217 1218 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1219 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1220 return (error); 1221 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1222 SCARG(uap, path)); 1223 if ((error = change_dir(&nd, l)) != 0) 1224 return (error); 1225 1226 cwdi = p->p_cwdi; 1227 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1228 if (cwdi->cwdi_rdir != NULL) 1229 vrele(cwdi->cwdi_rdir); 1230 vp = nd.ni_vp; 1231 cwdi->cwdi_rdir = vp; 1232 1233 /* 1234 * Prevent escaping from chroot by putting the root under 1235 * the working directory. Silently chdir to / if we aren't 1236 * already there. 1237 */ 1238 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1239 /* 1240 * XXX would be more failsafe to change directory to a 1241 * deadfs node here instead 1242 */ 1243 vrele(cwdi->cwdi_cdir); 1244 VREF(vp); 1245 cwdi->cwdi_cdir = vp; 1246 } 1247 rw_exit(&cwdi->cwdi_lock); 1248 1249 return (0); 1250 } 1251 1252 /* 1253 * Common routine for chroot and chdir. 1254 */ 1255 static int 1256 change_dir(struct nameidata *ndp, struct lwp *l) 1257 { 1258 struct vnode *vp; 1259 int error; 1260 1261 if ((error = namei(ndp)) != 0) 1262 return (error); 1263 vp = ndp->ni_vp; 1264 if (vp->v_type != VDIR) 1265 error = ENOTDIR; 1266 else 1267 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1268 1269 if (error) 1270 vput(vp); 1271 else 1272 VOP_UNLOCK(vp, 0); 1273 return (error); 1274 } 1275 1276 /* 1277 * Check permissions, allocate an open file structure, 1278 * and call the device open routine if any. 1279 */ 1280 int 1281 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1282 { 1283 /* { 1284 syscallarg(const char *) path; 1285 syscallarg(int) flags; 1286 syscallarg(int) mode; 1287 } */ 1288 struct proc *p = l->l_proc; 1289 struct cwdinfo *cwdi = p->p_cwdi; 1290 file_t *fp; 1291 struct vnode *vp; 1292 int flags, cmode; 1293 int type, indx, error; 1294 struct flock lf; 1295 struct nameidata nd; 1296 1297 flags = FFLAGS(SCARG(uap, flags)); 1298 if ((flags & (FREAD | FWRITE)) == 0) 1299 return (EINVAL); 1300 if ((error = fd_allocfile(&fp, &indx)) != 0) 1301 return (error); 1302 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1303 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1304 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1305 SCARG(uap, path)); 1306 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1307 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1308 fd_abort(p, fp, indx); 1309 if ((error == EDUPFD || error == EMOVEFD) && 1310 l->l_dupfd >= 0 && /* XXX from fdopen */ 1311 (error = 1312 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1313 *retval = indx; 1314 return (0); 1315 } 1316 if (error == ERESTART) 1317 error = EINTR; 1318 return (error); 1319 } 1320 1321 l->l_dupfd = 0; 1322 vp = nd.ni_vp; 1323 fp->f_flag = flags & FMASK; 1324 fp->f_type = DTYPE_VNODE; 1325 fp->f_ops = &vnops; 1326 fp->f_data = vp; 1327 if (flags & (O_EXLOCK | O_SHLOCK)) { 1328 lf.l_whence = SEEK_SET; 1329 lf.l_start = 0; 1330 lf.l_len = 0; 1331 if (flags & O_EXLOCK) 1332 lf.l_type = F_WRLCK; 1333 else 1334 lf.l_type = F_RDLCK; 1335 type = F_FLOCK; 1336 if ((flags & FNONBLOCK) == 0) 1337 type |= F_WAIT; 1338 VOP_UNLOCK(vp, 0); 1339 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1340 if (error) { 1341 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1342 fd_abort(p, fp, indx); 1343 return (error); 1344 } 1345 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1346 atomic_or_uint(&fp->f_flag, FHASLOCK); 1347 } 1348 VOP_UNLOCK(vp, 0); 1349 *retval = indx; 1350 fd_affix(p, fp, indx); 1351 return (0); 1352 } 1353 1354 static void 1355 vfs__fhfree(fhandle_t *fhp) 1356 { 1357 size_t fhsize; 1358 1359 if (fhp == NULL) { 1360 return; 1361 } 1362 fhsize = FHANDLE_SIZE(fhp); 1363 kmem_free(fhp, fhsize); 1364 } 1365 1366 /* 1367 * vfs_composefh: compose a filehandle. 1368 */ 1369 1370 int 1371 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1372 { 1373 struct mount *mp; 1374 struct fid *fidp; 1375 int error; 1376 size_t needfhsize; 1377 size_t fidsize; 1378 1379 mp = vp->v_mount; 1380 fidp = NULL; 1381 if (*fh_size < FHANDLE_SIZE_MIN) { 1382 fidsize = 0; 1383 } else { 1384 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1385 if (fhp != NULL) { 1386 memset(fhp, 0, *fh_size); 1387 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1388 fidp = &fhp->fh_fid; 1389 } 1390 } 1391 error = VFS_VPTOFH(vp, fidp, &fidsize); 1392 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1393 if (error == 0 && *fh_size < needfhsize) { 1394 error = E2BIG; 1395 } 1396 *fh_size = needfhsize; 1397 return error; 1398 } 1399 1400 int 1401 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1402 { 1403 struct mount *mp; 1404 fhandle_t *fhp; 1405 size_t fhsize; 1406 size_t fidsize; 1407 int error; 1408 1409 *fhpp = NULL; 1410 mp = vp->v_mount; 1411 fidsize = 0; 1412 error = VFS_VPTOFH(vp, NULL, &fidsize); 1413 KASSERT(error != 0); 1414 if (error != E2BIG) { 1415 goto out; 1416 } 1417 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1418 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1419 if (fhp == NULL) { 1420 error = ENOMEM; 1421 goto out; 1422 } 1423 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1424 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1425 if (error == 0) { 1426 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1427 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1428 *fhpp = fhp; 1429 } else { 1430 kmem_free(fhp, fhsize); 1431 } 1432 out: 1433 return error; 1434 } 1435 1436 void 1437 vfs_composefh_free(fhandle_t *fhp) 1438 { 1439 1440 vfs__fhfree(fhp); 1441 } 1442 1443 /* 1444 * vfs_fhtovp: lookup a vnode by a filehandle. 1445 */ 1446 1447 int 1448 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1449 { 1450 struct mount *mp; 1451 int error; 1452 1453 *vpp = NULL; 1454 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1455 if (mp == NULL) { 1456 error = ESTALE; 1457 goto out; 1458 } 1459 if (mp->mnt_op->vfs_fhtovp == NULL) { 1460 error = EOPNOTSUPP; 1461 goto out; 1462 } 1463 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1464 out: 1465 return error; 1466 } 1467 1468 /* 1469 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1470 * the needed size. 1471 */ 1472 1473 int 1474 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1475 { 1476 fhandle_t *fhp; 1477 int error; 1478 1479 *fhpp = NULL; 1480 if (fhsize > FHANDLE_SIZE_MAX) { 1481 return EINVAL; 1482 } 1483 if (fhsize < FHANDLE_SIZE_MIN) { 1484 return EINVAL; 1485 } 1486 again: 1487 fhp = kmem_alloc(fhsize, KM_SLEEP); 1488 if (fhp == NULL) { 1489 return ENOMEM; 1490 } 1491 error = copyin(ufhp, fhp, fhsize); 1492 if (error == 0) { 1493 /* XXX this check shouldn't be here */ 1494 if (FHANDLE_SIZE(fhp) == fhsize) { 1495 *fhpp = fhp; 1496 return 0; 1497 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1498 /* 1499 * a kludge for nfsv2 padded handles. 1500 */ 1501 size_t sz; 1502 1503 sz = FHANDLE_SIZE(fhp); 1504 kmem_free(fhp, fhsize); 1505 fhsize = sz; 1506 goto again; 1507 } else { 1508 /* 1509 * userland told us wrong size. 1510 */ 1511 error = EINVAL; 1512 } 1513 } 1514 kmem_free(fhp, fhsize); 1515 return error; 1516 } 1517 1518 void 1519 vfs_copyinfh_free(fhandle_t *fhp) 1520 { 1521 1522 vfs__fhfree(fhp); 1523 } 1524 1525 /* 1526 * Get file handle system call 1527 */ 1528 int 1529 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1530 { 1531 /* { 1532 syscallarg(char *) fname; 1533 syscallarg(fhandle_t *) fhp; 1534 syscallarg(size_t *) fh_size; 1535 } */ 1536 struct vnode *vp; 1537 fhandle_t *fh; 1538 int error; 1539 struct nameidata nd; 1540 size_t sz; 1541 size_t usz; 1542 1543 /* 1544 * Must be super user 1545 */ 1546 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1547 0, NULL, NULL, NULL); 1548 if (error) 1549 return (error); 1550 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1551 SCARG(uap, fname)); 1552 error = namei(&nd); 1553 if (error) 1554 return (error); 1555 vp = nd.ni_vp; 1556 error = vfs_composefh_alloc(vp, &fh); 1557 vput(vp); 1558 if (error != 0) { 1559 goto out; 1560 } 1561 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1562 if (error != 0) { 1563 goto out; 1564 } 1565 sz = FHANDLE_SIZE(fh); 1566 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1567 if (error != 0) { 1568 goto out; 1569 } 1570 if (usz >= sz) { 1571 error = copyout(fh, SCARG(uap, fhp), sz); 1572 } else { 1573 error = E2BIG; 1574 } 1575 out: 1576 vfs_composefh_free(fh); 1577 return (error); 1578 } 1579 1580 /* 1581 * Open a file given a file handle. 1582 * 1583 * Check permissions, allocate an open file structure, 1584 * and call the device open routine if any. 1585 */ 1586 1587 int 1588 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1589 register_t *retval) 1590 { 1591 file_t *fp; 1592 struct vnode *vp = NULL; 1593 kauth_cred_t cred = l->l_cred; 1594 file_t *nfp; 1595 int type, indx, error=0; 1596 struct flock lf; 1597 struct vattr va; 1598 fhandle_t *fh; 1599 int flags; 1600 proc_t *p; 1601 1602 p = curproc; 1603 1604 /* 1605 * Must be super user 1606 */ 1607 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1608 0, NULL, NULL, NULL))) 1609 return (error); 1610 1611 flags = FFLAGS(oflags); 1612 if ((flags & (FREAD | FWRITE)) == 0) 1613 return (EINVAL); 1614 if ((flags & O_CREAT)) 1615 return (EINVAL); 1616 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1617 return (error); 1618 fp = nfp; 1619 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1620 if (error != 0) { 1621 goto bad; 1622 } 1623 error = vfs_fhtovp(fh, &vp); 1624 if (error != 0) { 1625 goto bad; 1626 } 1627 1628 /* Now do an effective vn_open */ 1629 1630 if (vp->v_type == VSOCK) { 1631 error = EOPNOTSUPP; 1632 goto bad; 1633 } 1634 error = vn_openchk(vp, cred, flags); 1635 if (error != 0) 1636 goto bad; 1637 if (flags & O_TRUNC) { 1638 VOP_UNLOCK(vp, 0); /* XXX */ 1639 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1640 VATTR_NULL(&va); 1641 va.va_size = 0; 1642 error = VOP_SETATTR(vp, &va, cred); 1643 if (error) 1644 goto bad; 1645 } 1646 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1647 goto bad; 1648 if (flags & FWRITE) { 1649 mutex_enter(&vp->v_interlock); 1650 vp->v_writecount++; 1651 mutex_exit(&vp->v_interlock); 1652 } 1653 1654 /* done with modified vn_open, now finish what sys_open does. */ 1655 1656 fp->f_flag = flags & FMASK; 1657 fp->f_type = DTYPE_VNODE; 1658 fp->f_ops = &vnops; 1659 fp->f_data = vp; 1660 if (flags & (O_EXLOCK | O_SHLOCK)) { 1661 lf.l_whence = SEEK_SET; 1662 lf.l_start = 0; 1663 lf.l_len = 0; 1664 if (flags & O_EXLOCK) 1665 lf.l_type = F_WRLCK; 1666 else 1667 lf.l_type = F_RDLCK; 1668 type = F_FLOCK; 1669 if ((flags & FNONBLOCK) == 0) 1670 type |= F_WAIT; 1671 VOP_UNLOCK(vp, 0); 1672 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1673 if (error) { 1674 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1675 fd_abort(p, fp, indx); 1676 return (error); 1677 } 1678 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1679 atomic_or_uint(&fp->f_flag, FHASLOCK); 1680 } 1681 VOP_UNLOCK(vp, 0); 1682 *retval = indx; 1683 fd_affix(p, fp, indx); 1684 vfs_copyinfh_free(fh); 1685 return (0); 1686 1687 bad: 1688 fd_abort(p, fp, indx); 1689 if (vp != NULL) 1690 vput(vp); 1691 vfs_copyinfh_free(fh); 1692 return (error); 1693 } 1694 1695 int 1696 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1697 { 1698 /* { 1699 syscallarg(const void *) fhp; 1700 syscallarg(size_t) fh_size; 1701 syscallarg(int) flags; 1702 } */ 1703 1704 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1705 SCARG(uap, flags), retval); 1706 } 1707 1708 int 1709 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1710 { 1711 int error; 1712 fhandle_t *fh; 1713 struct vnode *vp; 1714 1715 /* 1716 * Must be super user 1717 */ 1718 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1719 0, NULL, NULL, NULL))) 1720 return (error); 1721 1722 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1723 if (error != 0) 1724 return error; 1725 1726 error = vfs_fhtovp(fh, &vp); 1727 vfs_copyinfh_free(fh); 1728 if (error != 0) 1729 return error; 1730 1731 error = vn_stat(vp, sb); 1732 vput(vp); 1733 return error; 1734 } 1735 1736 1737 /* ARGSUSED */ 1738 int 1739 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1740 { 1741 /* { 1742 syscallarg(const void *) fhp; 1743 syscallarg(size_t) fh_size; 1744 syscallarg(struct stat *) sb; 1745 } */ 1746 struct stat sb; 1747 int error; 1748 1749 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1750 if (error) 1751 return error; 1752 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1753 } 1754 1755 int 1756 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1757 int flags) 1758 { 1759 fhandle_t *fh; 1760 struct mount *mp; 1761 struct vnode *vp; 1762 int error; 1763 1764 /* 1765 * Must be super user 1766 */ 1767 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1768 0, NULL, NULL, NULL))) 1769 return error; 1770 1771 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1772 if (error != 0) 1773 return error; 1774 1775 error = vfs_fhtovp(fh, &vp); 1776 vfs_copyinfh_free(fh); 1777 if (error != 0) 1778 return error; 1779 1780 mp = vp->v_mount; 1781 error = dostatvfs(mp, sb, l, flags, 1); 1782 vput(vp); 1783 return error; 1784 } 1785 1786 /* ARGSUSED */ 1787 int 1788 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1789 { 1790 /* { 1791 syscallarg(const void *) fhp; 1792 syscallarg(size_t) fh_size; 1793 syscallarg(struct statvfs *) buf; 1794 syscallarg(int) flags; 1795 } */ 1796 struct statvfs *sb = STATVFSBUF_GET(); 1797 int error; 1798 1799 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1800 SCARG(uap, flags)); 1801 if (error == 0) 1802 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1803 STATVFSBUF_PUT(sb); 1804 return error; 1805 } 1806 1807 /* 1808 * Create a special file. 1809 */ 1810 /* ARGSUSED */ 1811 int 1812 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1813 { 1814 /* { 1815 syscallarg(const char *) path; 1816 syscallarg(int) mode; 1817 syscallarg(int) dev; 1818 } */ 1819 struct proc *p = l->l_proc; 1820 struct vnode *vp; 1821 struct vattr vattr; 1822 int error, optype; 1823 struct nameidata nd; 1824 char *path; 1825 const char *cpath; 1826 enum uio_seg seg = UIO_USERSPACE; 1827 1828 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1829 0, NULL, NULL, NULL)) != 0) 1830 return (error); 1831 1832 optype = VOP_MKNOD_DESCOFFSET; 1833 1834 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1835 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1836 1837 if ((error = namei(&nd)) != 0) 1838 goto out; 1839 vp = nd.ni_vp; 1840 if (vp != NULL) 1841 error = EEXIST; 1842 else { 1843 VATTR_NULL(&vattr); 1844 /* We will read cwdi->cwdi_cmask unlocked. */ 1845 vattr.va_mode = 1846 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1847 vattr.va_rdev = SCARG(uap, dev); 1848 1849 switch (SCARG(uap, mode) & S_IFMT) { 1850 case S_IFMT: /* used by badsect to flag bad sectors */ 1851 vattr.va_type = VBAD; 1852 break; 1853 case S_IFCHR: 1854 vattr.va_type = VCHR; 1855 break; 1856 case S_IFBLK: 1857 vattr.va_type = VBLK; 1858 break; 1859 case S_IFWHT: 1860 optype = VOP_WHITEOUT_DESCOFFSET; 1861 break; 1862 case S_IFREG: 1863 #if NVERIEXEC > 0 1864 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1865 O_CREAT); 1866 #endif /* NVERIEXEC > 0 */ 1867 vattr.va_type = VREG; 1868 vattr.va_rdev = VNOVAL; 1869 optype = VOP_CREATE_DESCOFFSET; 1870 break; 1871 default: 1872 error = EINVAL; 1873 break; 1874 } 1875 } 1876 if (!error) { 1877 switch (optype) { 1878 case VOP_WHITEOUT_DESCOFFSET: 1879 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1880 if (error) 1881 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1882 vput(nd.ni_dvp); 1883 break; 1884 1885 case VOP_MKNOD_DESCOFFSET: 1886 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1887 &nd.ni_cnd, &vattr); 1888 if (error == 0) 1889 vput(nd.ni_vp); 1890 break; 1891 1892 case VOP_CREATE_DESCOFFSET: 1893 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1894 &nd.ni_cnd, &vattr); 1895 if (error == 0) 1896 vput(nd.ni_vp); 1897 break; 1898 } 1899 } else { 1900 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1901 if (nd.ni_dvp == vp) 1902 vrele(nd.ni_dvp); 1903 else 1904 vput(nd.ni_dvp); 1905 if (vp) 1906 vrele(vp); 1907 } 1908 out: 1909 VERIEXEC_PATH_PUT(path); 1910 return (error); 1911 } 1912 1913 /* 1914 * Create a named pipe. 1915 */ 1916 /* ARGSUSED */ 1917 int 1918 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1919 { 1920 /* { 1921 syscallarg(const char *) path; 1922 syscallarg(int) mode; 1923 } */ 1924 struct proc *p = l->l_proc; 1925 struct vattr vattr; 1926 int error; 1927 struct nameidata nd; 1928 1929 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1930 SCARG(uap, path)); 1931 if ((error = namei(&nd)) != 0) 1932 return (error); 1933 if (nd.ni_vp != NULL) { 1934 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1935 if (nd.ni_dvp == nd.ni_vp) 1936 vrele(nd.ni_dvp); 1937 else 1938 vput(nd.ni_dvp); 1939 vrele(nd.ni_vp); 1940 return (EEXIST); 1941 } 1942 VATTR_NULL(&vattr); 1943 vattr.va_type = VFIFO; 1944 /* We will read cwdi->cwdi_cmask unlocked. */ 1945 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1946 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1947 if (error == 0) 1948 vput(nd.ni_vp); 1949 return (error); 1950 } 1951 1952 /* 1953 * Make a hard file link. 1954 */ 1955 /* ARGSUSED */ 1956 int 1957 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1958 { 1959 /* { 1960 syscallarg(const char *) path; 1961 syscallarg(const char *) link; 1962 } */ 1963 struct vnode *vp; 1964 struct nameidata nd; 1965 int error; 1966 1967 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1968 SCARG(uap, path)); 1969 if ((error = namei(&nd)) != 0) 1970 return (error); 1971 vp = nd.ni_vp; 1972 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1973 SCARG(uap, link)); 1974 if ((error = namei(&nd)) != 0) 1975 goto out; 1976 if (nd.ni_vp) { 1977 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1978 if (nd.ni_dvp == nd.ni_vp) 1979 vrele(nd.ni_dvp); 1980 else 1981 vput(nd.ni_dvp); 1982 vrele(nd.ni_vp); 1983 error = EEXIST; 1984 goto out; 1985 } 1986 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1987 out: 1988 vrele(vp); 1989 return (error); 1990 } 1991 1992 /* 1993 * Make a symbolic link. 1994 */ 1995 /* ARGSUSED */ 1996 int 1997 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1998 { 1999 /* { 2000 syscallarg(const char *) path; 2001 syscallarg(const char *) link; 2002 } */ 2003 struct proc *p = l->l_proc; 2004 struct vattr vattr; 2005 char *path; 2006 int error; 2007 struct nameidata nd; 2008 2009 path = PNBUF_GET(); 2010 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2011 if (error) 2012 goto out; 2013 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2014 SCARG(uap, link)); 2015 if ((error = namei(&nd)) != 0) 2016 goto out; 2017 if (nd.ni_vp) { 2018 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2019 if (nd.ni_dvp == nd.ni_vp) 2020 vrele(nd.ni_dvp); 2021 else 2022 vput(nd.ni_dvp); 2023 vrele(nd.ni_vp); 2024 error = EEXIST; 2025 goto out; 2026 } 2027 VATTR_NULL(&vattr); 2028 vattr.va_type = VLNK; 2029 /* We will read cwdi->cwdi_cmask unlocked. */ 2030 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2031 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2032 if (error == 0) 2033 vput(nd.ni_vp); 2034 out: 2035 PNBUF_PUT(path); 2036 return (error); 2037 } 2038 2039 /* 2040 * Delete a whiteout from the filesystem. 2041 */ 2042 /* ARGSUSED */ 2043 int 2044 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2045 { 2046 /* { 2047 syscallarg(const char *) path; 2048 } */ 2049 int error; 2050 struct nameidata nd; 2051 2052 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2053 UIO_USERSPACE, SCARG(uap, path)); 2054 error = namei(&nd); 2055 if (error) 2056 return (error); 2057 2058 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2059 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2060 if (nd.ni_dvp == nd.ni_vp) 2061 vrele(nd.ni_dvp); 2062 else 2063 vput(nd.ni_dvp); 2064 if (nd.ni_vp) 2065 vrele(nd.ni_vp); 2066 return (EEXIST); 2067 } 2068 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, const struct sys_unlink_args *uap, register_t *retval) 2080 { 2081 /* { 2082 syscallarg(const char *) path; 2083 } */ 2084 2085 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2086 } 2087 2088 int 2089 do_sys_unlink(const char *arg, enum uio_seg seg) 2090 { 2091 struct vnode *vp; 2092 int error; 2093 struct nameidata nd; 2094 kauth_cred_t cred; 2095 char *path; 2096 const char *cpath; 2097 2098 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2099 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2100 2101 if ((error = namei(&nd)) != 0) 2102 goto out; 2103 vp = nd.ni_vp; 2104 2105 /* 2106 * The root of a mounted filesystem cannot be deleted. 2107 */ 2108 if (vp->v_vflag & VV_ROOT) { 2109 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2110 if (nd.ni_dvp == vp) 2111 vrele(nd.ni_dvp); 2112 else 2113 vput(nd.ni_dvp); 2114 vput(vp); 2115 error = EBUSY; 2116 goto out; 2117 } 2118 2119 #if NVERIEXEC > 0 2120 /* Handle remove requests for veriexec entries. */ 2121 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2122 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2123 if (nd.ni_dvp == vp) 2124 vrele(nd.ni_dvp); 2125 else 2126 vput(nd.ni_dvp); 2127 vput(vp); 2128 goto out; 2129 } 2130 #endif /* NVERIEXEC > 0 */ 2131 2132 cred = kauth_cred_get(); 2133 #ifdef FILEASSOC 2134 (void)fileassoc_file_delete(vp); 2135 #endif /* FILEASSOC */ 2136 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2137 out: 2138 VERIEXEC_PATH_PUT(path); 2139 return (error); 2140 } 2141 2142 /* 2143 * Reposition read/write file offset. 2144 */ 2145 int 2146 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2147 { 2148 /* { 2149 syscallarg(int) fd; 2150 syscallarg(int) pad; 2151 syscallarg(off_t) offset; 2152 syscallarg(int) whence; 2153 } */ 2154 kauth_cred_t cred = l->l_cred; 2155 file_t *fp; 2156 struct vnode *vp; 2157 struct vattr vattr; 2158 off_t newoff; 2159 int error, fd; 2160 2161 fd = SCARG(uap, fd); 2162 2163 if ((fp = fd_getfile(fd)) == NULL) 2164 return (EBADF); 2165 2166 vp = fp->f_data; 2167 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2168 error = ESPIPE; 2169 goto out; 2170 } 2171 2172 switch (SCARG(uap, whence)) { 2173 case SEEK_CUR: 2174 newoff = fp->f_offset + SCARG(uap, offset); 2175 break; 2176 case SEEK_END: 2177 error = VOP_GETATTR(vp, &vattr, cred); 2178 if (error) { 2179 goto out; 2180 } 2181 newoff = SCARG(uap, offset) + vattr.va_size; 2182 break; 2183 case SEEK_SET: 2184 newoff = SCARG(uap, offset); 2185 break; 2186 default: 2187 error = EINVAL; 2188 goto out; 2189 } 2190 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2191 *(off_t *)retval = fp->f_offset = newoff; 2192 } 2193 out: 2194 fd_putfile(fd); 2195 return (error); 2196 } 2197 2198 /* 2199 * Positional read system call. 2200 */ 2201 int 2202 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2203 { 2204 /* { 2205 syscallarg(int) fd; 2206 syscallarg(void *) buf; 2207 syscallarg(size_t) nbyte; 2208 syscallarg(off_t) offset; 2209 } */ 2210 file_t *fp; 2211 struct vnode *vp; 2212 off_t offset; 2213 int error, fd = SCARG(uap, fd); 2214 2215 if ((fp = fd_getfile(fd)) == NULL) 2216 return (EBADF); 2217 2218 if ((fp->f_flag & FREAD) == 0) { 2219 fd_putfile(fd); 2220 return (EBADF); 2221 } 2222 2223 vp = fp->f_data; 2224 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2225 error = ESPIPE; 2226 goto out; 2227 } 2228 2229 offset = SCARG(uap, offset); 2230 2231 /* 2232 * XXX This works because no file systems actually 2233 * XXX take any action on the seek operation. 2234 */ 2235 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2236 goto out; 2237 2238 /* dofileread() will unuse the descriptor for us */ 2239 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2240 &offset, 0, retval)); 2241 2242 out: 2243 fd_putfile(fd); 2244 return (error); 2245 } 2246 2247 /* 2248 * Positional scatter read system call. 2249 */ 2250 int 2251 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2252 { 2253 /* { 2254 syscallarg(int) fd; 2255 syscallarg(const struct iovec *) iovp; 2256 syscallarg(int) iovcnt; 2257 syscallarg(off_t) offset; 2258 } */ 2259 off_t offset = SCARG(uap, offset); 2260 2261 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2262 SCARG(uap, iovcnt), &offset, 0, retval); 2263 } 2264 2265 /* 2266 * Positional write system call. 2267 */ 2268 int 2269 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2270 { 2271 /* { 2272 syscallarg(int) fd; 2273 syscallarg(const void *) buf; 2274 syscallarg(size_t) nbyte; 2275 syscallarg(off_t) offset; 2276 } */ 2277 file_t *fp; 2278 struct vnode *vp; 2279 off_t offset; 2280 int error, fd = SCARG(uap, fd); 2281 2282 if ((fp = fd_getfile(fd)) == NULL) 2283 return (EBADF); 2284 2285 if ((fp->f_flag & FWRITE) == 0) { 2286 fd_putfile(fd); 2287 return (EBADF); 2288 } 2289 2290 vp = fp->f_data; 2291 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2292 error = ESPIPE; 2293 goto out; 2294 } 2295 2296 offset = SCARG(uap, offset); 2297 2298 /* 2299 * XXX This works because no file systems actually 2300 * XXX take any action on the seek operation. 2301 */ 2302 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2303 goto out; 2304 2305 /* dofilewrite() will unuse the descriptor for us */ 2306 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2307 &offset, 0, retval)); 2308 2309 out: 2310 fd_putfile(fd); 2311 return (error); 2312 } 2313 2314 /* 2315 * Positional gather write system call. 2316 */ 2317 int 2318 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2319 { 2320 /* { 2321 syscallarg(int) fd; 2322 syscallarg(const struct iovec *) iovp; 2323 syscallarg(int) iovcnt; 2324 syscallarg(off_t) offset; 2325 } */ 2326 off_t offset = SCARG(uap, offset); 2327 2328 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2329 SCARG(uap, iovcnt), &offset, 0, retval); 2330 } 2331 2332 /* 2333 * Check access permissions. 2334 */ 2335 int 2336 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2337 { 2338 /* { 2339 syscallarg(const char *) path; 2340 syscallarg(int) flags; 2341 } */ 2342 kauth_cred_t cred; 2343 struct vnode *vp; 2344 int error, flags; 2345 struct nameidata nd; 2346 2347 cred = kauth_cred_dup(l->l_cred); 2348 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2349 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2350 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2351 SCARG(uap, path)); 2352 /* Override default credentials */ 2353 nd.ni_cnd.cn_cred = cred; 2354 if ((error = namei(&nd)) != 0) 2355 goto out; 2356 vp = nd.ni_vp; 2357 2358 /* Flags == 0 means only check for existence. */ 2359 if (SCARG(uap, flags)) { 2360 flags = 0; 2361 if (SCARG(uap, flags) & R_OK) 2362 flags |= VREAD; 2363 if (SCARG(uap, flags) & W_OK) 2364 flags |= VWRITE; 2365 if (SCARG(uap, flags) & X_OK) 2366 flags |= VEXEC; 2367 2368 error = VOP_ACCESS(vp, flags, cred); 2369 if (!error && (flags & VWRITE)) 2370 error = vn_writechk(vp); 2371 } 2372 vput(vp); 2373 out: 2374 kauth_cred_free(cred); 2375 return (error); 2376 } 2377 2378 /* 2379 * Common code for all sys_stat functions, including compat versions. 2380 */ 2381 int 2382 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb) 2383 { 2384 int error; 2385 struct nameidata nd; 2386 2387 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2388 UIO_USERSPACE, path); 2389 error = namei(&nd); 2390 if (error != 0) 2391 return error; 2392 error = vn_stat(nd.ni_vp, sb); 2393 vput(nd.ni_vp); 2394 return error; 2395 } 2396 2397 /* 2398 * Get file status; this version follows links. 2399 */ 2400 /* ARGSUSED */ 2401 int 2402 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2403 { 2404 /* { 2405 syscallarg(const char *) path; 2406 syscallarg(struct stat *) ub; 2407 } */ 2408 struct stat sb; 2409 int error; 2410 2411 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2412 if (error) 2413 return error; 2414 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2415 } 2416 2417 /* 2418 * Get file status; this version does not follow links. 2419 */ 2420 /* ARGSUSED */ 2421 int 2422 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2423 { 2424 /* { 2425 syscallarg(const char *) path; 2426 syscallarg(struct stat *) ub; 2427 } */ 2428 struct stat sb; 2429 int error; 2430 2431 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2432 if (error) 2433 return error; 2434 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2435 } 2436 2437 /* 2438 * Get configurable pathname variables. 2439 */ 2440 /* ARGSUSED */ 2441 int 2442 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2443 { 2444 /* { 2445 syscallarg(const char *) path; 2446 syscallarg(int) name; 2447 } */ 2448 int error; 2449 struct nameidata nd; 2450 2451 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2452 SCARG(uap, path)); 2453 if ((error = namei(&nd)) != 0) 2454 return (error); 2455 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2456 vput(nd.ni_vp); 2457 return (error); 2458 } 2459 2460 /* 2461 * Return target name of a symbolic link. 2462 */ 2463 /* ARGSUSED */ 2464 int 2465 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2466 { 2467 /* { 2468 syscallarg(const char *) path; 2469 syscallarg(char *) buf; 2470 syscallarg(size_t) count; 2471 } */ 2472 struct vnode *vp; 2473 struct iovec aiov; 2474 struct uio auio; 2475 int error; 2476 struct nameidata nd; 2477 2478 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2479 SCARG(uap, path)); 2480 if ((error = namei(&nd)) != 0) 2481 return (error); 2482 vp = nd.ni_vp; 2483 if (vp->v_type != VLNK) 2484 error = EINVAL; 2485 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2486 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2487 aiov.iov_base = SCARG(uap, buf); 2488 aiov.iov_len = SCARG(uap, count); 2489 auio.uio_iov = &aiov; 2490 auio.uio_iovcnt = 1; 2491 auio.uio_offset = 0; 2492 auio.uio_rw = UIO_READ; 2493 KASSERT(l == curlwp); 2494 auio.uio_vmspace = l->l_proc->p_vmspace; 2495 auio.uio_resid = SCARG(uap, count); 2496 error = VOP_READLINK(vp, &auio, l->l_cred); 2497 } 2498 vput(vp); 2499 *retval = SCARG(uap, count) - auio.uio_resid; 2500 return (error); 2501 } 2502 2503 /* 2504 * Change flags of a file given a path name. 2505 */ 2506 /* ARGSUSED */ 2507 int 2508 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2509 { 2510 /* { 2511 syscallarg(const char *) path; 2512 syscallarg(u_long) flags; 2513 } */ 2514 struct vnode *vp; 2515 int error; 2516 struct nameidata nd; 2517 2518 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2519 SCARG(uap, path)); 2520 if ((error = namei(&nd)) != 0) 2521 return (error); 2522 vp = nd.ni_vp; 2523 error = change_flags(vp, SCARG(uap, flags), l); 2524 vput(vp); 2525 return (error); 2526 } 2527 2528 /* 2529 * Change flags of a file given a file descriptor. 2530 */ 2531 /* ARGSUSED */ 2532 int 2533 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2534 { 2535 /* { 2536 syscallarg(int) fd; 2537 syscallarg(u_long) flags; 2538 } */ 2539 struct vnode *vp; 2540 file_t *fp; 2541 int error; 2542 2543 /* fd_getvnode() will use the descriptor for us */ 2544 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2545 return (error); 2546 vp = fp->f_data; 2547 error = change_flags(vp, SCARG(uap, flags), l); 2548 VOP_UNLOCK(vp, 0); 2549 fd_putfile(SCARG(uap, fd)); 2550 return (error); 2551 } 2552 2553 /* 2554 * Change flags of a file given a path name; this version does 2555 * not follow links. 2556 */ 2557 int 2558 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2559 { 2560 /* { 2561 syscallarg(const char *) path; 2562 syscallarg(u_long) flags; 2563 } */ 2564 struct vnode *vp; 2565 int error; 2566 struct nameidata nd; 2567 2568 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2569 SCARG(uap, path)); 2570 if ((error = namei(&nd)) != 0) 2571 return (error); 2572 vp = nd.ni_vp; 2573 error = change_flags(vp, SCARG(uap, flags), l); 2574 vput(vp); 2575 return (error); 2576 } 2577 2578 /* 2579 * Common routine to change flags of a file. 2580 */ 2581 int 2582 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2583 { 2584 struct vattr vattr; 2585 int error; 2586 2587 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2588 /* 2589 * Non-superusers cannot change the flags on devices, even if they 2590 * own them. 2591 */ 2592 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2593 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2594 goto out; 2595 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2596 error = EINVAL; 2597 goto out; 2598 } 2599 } 2600 VATTR_NULL(&vattr); 2601 vattr.va_flags = flags; 2602 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2603 out: 2604 return (error); 2605 } 2606 2607 /* 2608 * Change mode of a file given path name; this version follows links. 2609 */ 2610 /* ARGSUSED */ 2611 int 2612 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2613 { 2614 /* { 2615 syscallarg(const char *) path; 2616 syscallarg(int) mode; 2617 } */ 2618 int error; 2619 struct nameidata nd; 2620 2621 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2622 SCARG(uap, path)); 2623 if ((error = namei(&nd)) != 0) 2624 return (error); 2625 2626 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2627 2628 vrele(nd.ni_vp); 2629 return (error); 2630 } 2631 2632 /* 2633 * Change mode of a file given a file descriptor. 2634 */ 2635 /* ARGSUSED */ 2636 int 2637 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2638 { 2639 /* { 2640 syscallarg(int) fd; 2641 syscallarg(int) mode; 2642 } */ 2643 file_t *fp; 2644 int error; 2645 2646 /* fd_getvnode() will use the descriptor for us */ 2647 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2648 return (error); 2649 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2650 fd_putfile(SCARG(uap, fd)); 2651 return (error); 2652 } 2653 2654 /* 2655 * Change mode of a file given path name; this version does not follow links. 2656 */ 2657 /* ARGSUSED */ 2658 int 2659 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2660 { 2661 /* { 2662 syscallarg(const char *) path; 2663 syscallarg(int) mode; 2664 } */ 2665 int error; 2666 struct nameidata nd; 2667 2668 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2669 SCARG(uap, path)); 2670 if ((error = namei(&nd)) != 0) 2671 return (error); 2672 2673 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2674 2675 vrele(nd.ni_vp); 2676 return (error); 2677 } 2678 2679 /* 2680 * Common routine to set mode given a vnode. 2681 */ 2682 static int 2683 change_mode(struct vnode *vp, int mode, struct lwp *l) 2684 { 2685 struct vattr vattr; 2686 int error; 2687 2688 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2689 VATTR_NULL(&vattr); 2690 vattr.va_mode = mode & ALLPERMS; 2691 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2692 VOP_UNLOCK(vp, 0); 2693 return (error); 2694 } 2695 2696 /* 2697 * Set ownership given a path name; this version follows links. 2698 */ 2699 /* ARGSUSED */ 2700 int 2701 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2702 { 2703 /* { 2704 syscallarg(const char *) path; 2705 syscallarg(uid_t) uid; 2706 syscallarg(gid_t) gid; 2707 } */ 2708 int error; 2709 struct nameidata nd; 2710 2711 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2712 SCARG(uap, path)); 2713 if ((error = namei(&nd)) != 0) 2714 return (error); 2715 2716 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2717 2718 vrele(nd.ni_vp); 2719 return (error); 2720 } 2721 2722 /* 2723 * Set ownership given a path name; this version follows links. 2724 * Provides POSIX semantics. 2725 */ 2726 /* ARGSUSED */ 2727 int 2728 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2729 { 2730 /* { 2731 syscallarg(const char *) path; 2732 syscallarg(uid_t) uid; 2733 syscallarg(gid_t) gid; 2734 } */ 2735 int error; 2736 struct nameidata nd; 2737 2738 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2739 SCARG(uap, path)); 2740 if ((error = namei(&nd)) != 0) 2741 return (error); 2742 2743 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2744 2745 vrele(nd.ni_vp); 2746 return (error); 2747 } 2748 2749 /* 2750 * Set ownership given a file descriptor. 2751 */ 2752 /* ARGSUSED */ 2753 int 2754 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2755 { 2756 /* { 2757 syscallarg(int) fd; 2758 syscallarg(uid_t) uid; 2759 syscallarg(gid_t) gid; 2760 } */ 2761 int error; 2762 file_t *fp; 2763 2764 /* fd_getvnode() will use the descriptor for us */ 2765 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2766 return (error); 2767 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2768 l, 0); 2769 fd_putfile(SCARG(uap, fd)); 2770 return (error); 2771 } 2772 2773 /* 2774 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2775 */ 2776 /* ARGSUSED */ 2777 int 2778 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2779 { 2780 /* { 2781 syscallarg(int) fd; 2782 syscallarg(uid_t) uid; 2783 syscallarg(gid_t) gid; 2784 } */ 2785 int error; 2786 file_t *fp; 2787 2788 /* fd_getvnode() will use the descriptor for us */ 2789 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2790 return (error); 2791 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2792 l, 1); 2793 fd_putfile(SCARG(uap, fd)); 2794 return (error); 2795 } 2796 2797 /* 2798 * Set ownership given a path name; this version does not follow links. 2799 */ 2800 /* ARGSUSED */ 2801 int 2802 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2803 { 2804 /* { 2805 syscallarg(const char *) path; 2806 syscallarg(uid_t) uid; 2807 syscallarg(gid_t) gid; 2808 } */ 2809 int error; 2810 struct nameidata nd; 2811 2812 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2813 SCARG(uap, path)); 2814 if ((error = namei(&nd)) != 0) 2815 return (error); 2816 2817 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2818 2819 vrele(nd.ni_vp); 2820 return (error); 2821 } 2822 2823 /* 2824 * Set ownership given a path name; this version does not follow links. 2825 * Provides POSIX/XPG semantics. 2826 */ 2827 /* ARGSUSED */ 2828 int 2829 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2830 { 2831 /* { 2832 syscallarg(const char *) path; 2833 syscallarg(uid_t) uid; 2834 syscallarg(gid_t) gid; 2835 } */ 2836 int error; 2837 struct nameidata nd; 2838 2839 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2840 SCARG(uap, path)); 2841 if ((error = namei(&nd)) != 0) 2842 return (error); 2843 2844 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2845 2846 vrele(nd.ni_vp); 2847 return (error); 2848 } 2849 2850 /* 2851 * Common routine to set ownership given a vnode. 2852 */ 2853 static int 2854 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2855 int posix_semantics) 2856 { 2857 struct vattr vattr; 2858 mode_t newmode; 2859 int error; 2860 2861 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2862 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2863 goto out; 2864 2865 #define CHANGED(x) ((int)(x) != -1) 2866 newmode = vattr.va_mode; 2867 if (posix_semantics) { 2868 /* 2869 * POSIX/XPG semantics: if the caller is not the super-user, 2870 * clear set-user-id and set-group-id bits. Both POSIX and 2871 * the XPG consider the behaviour for calls by the super-user 2872 * implementation-defined; we leave the set-user-id and set- 2873 * group-id settings intact in that case. 2874 */ 2875 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2876 NULL) != 0) 2877 newmode &= ~(S_ISUID | S_ISGID); 2878 } else { 2879 /* 2880 * NetBSD semantics: when changing owner and/or group, 2881 * clear the respective bit(s). 2882 */ 2883 if (CHANGED(uid)) 2884 newmode &= ~S_ISUID; 2885 if (CHANGED(gid)) 2886 newmode &= ~S_ISGID; 2887 } 2888 /* Update va_mode iff altered. */ 2889 if (vattr.va_mode == newmode) 2890 newmode = VNOVAL; 2891 2892 VATTR_NULL(&vattr); 2893 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2894 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2895 vattr.va_mode = newmode; 2896 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2897 #undef CHANGED 2898 2899 out: 2900 VOP_UNLOCK(vp, 0); 2901 return (error); 2902 } 2903 2904 /* 2905 * Set the access and modification times given a path name; this 2906 * version follows links. 2907 */ 2908 /* ARGSUSED */ 2909 int 2910 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2911 { 2912 /* { 2913 syscallarg(const char *) path; 2914 syscallarg(const struct timeval *) tptr; 2915 } */ 2916 2917 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2918 SCARG(uap, tptr), UIO_USERSPACE); 2919 } 2920 2921 /* 2922 * Set the access and modification times given a file descriptor. 2923 */ 2924 /* ARGSUSED */ 2925 int 2926 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2927 { 2928 /* { 2929 syscallarg(int) fd; 2930 syscallarg(const struct timeval *) tptr; 2931 } */ 2932 int error; 2933 file_t *fp; 2934 2935 /* fd_getvnode() will use the descriptor for us */ 2936 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2937 return (error); 2938 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2939 UIO_USERSPACE); 2940 fd_putfile(SCARG(uap, fd)); 2941 return (error); 2942 } 2943 2944 /* 2945 * Set the access and modification times given a path name; this 2946 * version does not follow links. 2947 */ 2948 int 2949 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2950 { 2951 /* { 2952 syscallarg(const char *) path; 2953 syscallarg(const struct timeval *) tptr; 2954 } */ 2955 2956 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2957 SCARG(uap, tptr), UIO_USERSPACE); 2958 } 2959 2960 /* 2961 * Common routine to set access and modification times given a vnode. 2962 */ 2963 int 2964 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2965 const struct timeval *tptr, enum uio_seg seg) 2966 { 2967 struct vattr vattr; 2968 struct nameidata nd; 2969 int error; 2970 2971 VATTR_NULL(&vattr); 2972 if (tptr == NULL) { 2973 nanotime(&vattr.va_atime); 2974 vattr.va_mtime = vattr.va_atime; 2975 vattr.va_vaflags |= VA_UTIMES_NULL; 2976 } else { 2977 struct timeval tv[2]; 2978 2979 if (seg != UIO_SYSSPACE) { 2980 error = copyin(tptr, &tv, sizeof (tv)); 2981 if (error != 0) 2982 return error; 2983 tptr = tv; 2984 } 2985 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 2986 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 2987 } 2988 2989 if (vp == NULL) { 2990 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 2991 if ((error = namei(&nd)) != 0) 2992 return (error); 2993 vp = nd.ni_vp; 2994 } else 2995 nd.ni_vp = NULL; 2996 2997 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2998 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2999 VOP_UNLOCK(vp, 0); 3000 3001 if (nd.ni_vp != NULL) 3002 vrele(nd.ni_vp); 3003 3004 return (error); 3005 } 3006 3007 /* 3008 * Truncate a file given its path name. 3009 */ 3010 /* ARGSUSED */ 3011 int 3012 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3013 { 3014 /* { 3015 syscallarg(const char *) path; 3016 syscallarg(int) pad; 3017 syscallarg(off_t) length; 3018 } */ 3019 struct vnode *vp; 3020 struct vattr vattr; 3021 int error; 3022 struct nameidata nd; 3023 3024 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3025 SCARG(uap, path)); 3026 if ((error = namei(&nd)) != 0) 3027 return (error); 3028 vp = nd.ni_vp; 3029 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3030 if (vp->v_type == VDIR) 3031 error = EISDIR; 3032 else if ((error = vn_writechk(vp)) == 0 && 3033 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3034 VATTR_NULL(&vattr); 3035 vattr.va_size = SCARG(uap, length); 3036 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3037 } 3038 vput(vp); 3039 return (error); 3040 } 3041 3042 /* 3043 * Truncate a file given a file descriptor. 3044 */ 3045 /* ARGSUSED */ 3046 int 3047 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3048 { 3049 /* { 3050 syscallarg(int) fd; 3051 syscallarg(int) pad; 3052 syscallarg(off_t) length; 3053 } */ 3054 struct vattr vattr; 3055 struct vnode *vp; 3056 file_t *fp; 3057 int error; 3058 3059 /* fd_getvnode() will use the descriptor for us */ 3060 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3061 return (error); 3062 if ((fp->f_flag & FWRITE) == 0) { 3063 error = EINVAL; 3064 goto out; 3065 } 3066 vp = fp->f_data; 3067 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3068 if (vp->v_type == VDIR) 3069 error = EISDIR; 3070 else if ((error = vn_writechk(vp)) == 0) { 3071 VATTR_NULL(&vattr); 3072 vattr.va_size = SCARG(uap, length); 3073 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3074 } 3075 VOP_UNLOCK(vp, 0); 3076 out: 3077 fd_putfile(SCARG(uap, fd)); 3078 return (error); 3079 } 3080 3081 /* 3082 * Sync an open file. 3083 */ 3084 /* ARGSUSED */ 3085 int 3086 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3087 { 3088 /* { 3089 syscallarg(int) fd; 3090 } */ 3091 struct vnode *vp; 3092 file_t *fp; 3093 int error; 3094 3095 /* fd_getvnode() will use the descriptor for us */ 3096 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3097 return (error); 3098 vp = fp->f_data; 3099 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3100 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3101 if (error == 0 && bioopsp != NULL && 3102 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3103 (*bioopsp->io_fsync)(vp, 0); 3104 VOP_UNLOCK(vp, 0); 3105 fd_putfile(SCARG(uap, fd)); 3106 return (error); 3107 } 3108 3109 /* 3110 * Sync a range of file data. API modeled after that found in AIX. 3111 * 3112 * FDATASYNC indicates that we need only save enough metadata to be able 3113 * to re-read the written data. Note we duplicate AIX's requirement that 3114 * the file be open for writing. 3115 */ 3116 /* ARGSUSED */ 3117 int 3118 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3119 { 3120 /* { 3121 syscallarg(int) fd; 3122 syscallarg(int) flags; 3123 syscallarg(off_t) start; 3124 syscallarg(off_t) length; 3125 } */ 3126 struct vnode *vp; 3127 file_t *fp; 3128 int flags, nflags; 3129 off_t s, e, len; 3130 int error; 3131 3132 /* fd_getvnode() will use the descriptor for us */ 3133 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3134 return (error); 3135 3136 if ((fp->f_flag & FWRITE) == 0) { 3137 error = EBADF; 3138 goto out; 3139 } 3140 3141 flags = SCARG(uap, flags); 3142 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3143 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3144 error = EINVAL; 3145 goto out; 3146 } 3147 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3148 if (flags & FDATASYNC) 3149 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3150 else 3151 nflags = FSYNC_WAIT; 3152 if (flags & FDISKSYNC) 3153 nflags |= FSYNC_CACHE; 3154 3155 len = SCARG(uap, length); 3156 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3157 if (len) { 3158 s = SCARG(uap, start); 3159 e = s + len; 3160 if (e < s) { 3161 error = EINVAL; 3162 goto out; 3163 } 3164 } else { 3165 e = 0; 3166 s = 0; 3167 } 3168 3169 vp = fp->f_data; 3170 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3171 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3172 3173 if (error == 0 && bioopsp != NULL && 3174 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3175 (*bioopsp->io_fsync)(vp, nflags); 3176 3177 VOP_UNLOCK(vp, 0); 3178 out: 3179 fd_putfile(SCARG(uap, fd)); 3180 return (error); 3181 } 3182 3183 /* 3184 * Sync the data of an open file. 3185 */ 3186 /* ARGSUSED */ 3187 int 3188 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3189 { 3190 /* { 3191 syscallarg(int) fd; 3192 } */ 3193 struct vnode *vp; 3194 file_t *fp; 3195 int error; 3196 3197 /* fd_getvnode() will use the descriptor for us */ 3198 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3199 return (error); 3200 if ((fp->f_flag & FWRITE) == 0) { 3201 fd_putfile(SCARG(uap, fd)); 3202 return (EBADF); 3203 } 3204 vp = fp->f_data; 3205 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3206 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3207 VOP_UNLOCK(vp, 0); 3208 fd_putfile(SCARG(uap, fd)); 3209 return (error); 3210 } 3211 3212 /* 3213 * Rename files, (standard) BSD semantics frontend. 3214 */ 3215 /* ARGSUSED */ 3216 int 3217 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3218 { 3219 /* { 3220 syscallarg(const char *) from; 3221 syscallarg(const char *) to; 3222 } */ 3223 3224 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3225 } 3226 3227 /* 3228 * Rename files, POSIX semantics frontend. 3229 */ 3230 /* ARGSUSED */ 3231 int 3232 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3233 { 3234 /* { 3235 syscallarg(const char *) from; 3236 syscallarg(const char *) to; 3237 } */ 3238 3239 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3240 } 3241 3242 /* 3243 * Rename files. Source and destination must either both be directories, 3244 * or both not be directories. If target is a directory, it must be empty. 3245 * If `from' and `to' refer to the same object, the value of the `retain' 3246 * argument is used to determine whether `from' will be 3247 * 3248 * (retain == 0) deleted unless `from' and `to' refer to the same 3249 * object in the file system's name space (BSD). 3250 * (retain == 1) always retained (POSIX). 3251 */ 3252 int 3253 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3254 { 3255 struct vnode *tvp, *fvp, *tdvp; 3256 struct nameidata fromnd, tond; 3257 struct mount *fs; 3258 struct lwp *l = curlwp; 3259 struct proc *p; 3260 uint32_t saveflag; 3261 int error; 3262 3263 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3264 seg, from); 3265 if ((error = namei(&fromnd)) != 0) 3266 return (error); 3267 if (fromnd.ni_dvp != fromnd.ni_vp) 3268 VOP_UNLOCK(fromnd.ni_dvp, 0); 3269 fvp = fromnd.ni_vp; 3270 3271 fs = fvp->v_mount; 3272 error = VFS_RENAMELOCK_ENTER(fs); 3273 if (error) { 3274 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3275 vrele(fromnd.ni_dvp); 3276 vrele(fvp); 3277 goto out1; 3278 } 3279 3280 /* 3281 * close, partially, yet another race - ideally we should only 3282 * go as far as getting fromnd.ni_dvp before getting the per-fs 3283 * lock, and then continue to get fromnd.ni_vp, but we can't do 3284 * that with namei as it stands. 3285 * 3286 * This still won't prevent rmdir from nuking fromnd.ni_vp 3287 * under us. The real fix is to get the locks in the right 3288 * order and do the lookups in the right places, but that's a 3289 * major rototill. 3290 * 3291 * Preserve the SAVESTART in cn_flags, because who knows what 3292 * might happen if we don't. 3293 * 3294 * Note: this logic (as well as this whole function) is cloned 3295 * in nfs_serv.c. Proceed accordingly. 3296 */ 3297 vrele(fvp); 3298 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3299 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3300 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3301 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3302 fromnd.ni_cnd.cn_flags |= saveflag; 3303 if (error) { 3304 VOP_UNLOCK(fromnd.ni_dvp, 0); 3305 VFS_RENAMELOCK_EXIT(fs); 3306 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3307 vrele(fromnd.ni_dvp); 3308 goto out1; 3309 } 3310 VOP_UNLOCK(fromnd.ni_vp, 0); 3311 if (fromnd.ni_dvp != fromnd.ni_vp) 3312 VOP_UNLOCK(fromnd.ni_dvp, 0); 3313 fvp = fromnd.ni_vp; 3314 3315 NDINIT(&tond, RENAME, 3316 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3317 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3318 seg, to); 3319 if ((error = namei(&tond)) != 0) { 3320 VFS_RENAMELOCK_EXIT(fs); 3321 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3322 vrele(fromnd.ni_dvp); 3323 vrele(fvp); 3324 goto out1; 3325 } 3326 tdvp = tond.ni_dvp; 3327 tvp = tond.ni_vp; 3328 3329 if (tvp != NULL) { 3330 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3331 error = ENOTDIR; 3332 goto out; 3333 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3334 error = EISDIR; 3335 goto out; 3336 } 3337 } 3338 3339 if (fvp == tdvp) 3340 error = EINVAL; 3341 3342 /* 3343 * Source and destination refer to the same object. 3344 */ 3345 if (fvp == tvp) { 3346 if (retain) 3347 error = -1; 3348 else if (fromnd.ni_dvp == tdvp && 3349 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3350 !memcmp(fromnd.ni_cnd.cn_nameptr, 3351 tond.ni_cnd.cn_nameptr, 3352 fromnd.ni_cnd.cn_namelen)) 3353 error = -1; 3354 } 3355 3356 #if NVERIEXEC > 0 3357 if (!error) { 3358 char *f1, *f2; 3359 3360 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3361 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3362 3363 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3364 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3365 3366 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3367 3368 free(f1, M_TEMP); 3369 free(f2, M_TEMP); 3370 } 3371 #endif /* NVERIEXEC > 0 */ 3372 3373 out: 3374 p = l->l_proc; 3375 if (!error) { 3376 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3377 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3378 VFS_RENAMELOCK_EXIT(fs); 3379 } else { 3380 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3381 if (tdvp == tvp) 3382 vrele(tdvp); 3383 else 3384 vput(tdvp); 3385 if (tvp) 3386 vput(tvp); 3387 VFS_RENAMELOCK_EXIT(fs); 3388 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3389 vrele(fromnd.ni_dvp); 3390 vrele(fvp); 3391 } 3392 vrele(tond.ni_startdir); 3393 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3394 out1: 3395 if (fromnd.ni_startdir) 3396 vrele(fromnd.ni_startdir); 3397 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3398 return (error == -1 ? 0 : error); 3399 } 3400 3401 /* 3402 * Make a directory file. 3403 */ 3404 /* ARGSUSED */ 3405 int 3406 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3407 { 3408 /* { 3409 syscallarg(const char *) path; 3410 syscallarg(int) mode; 3411 } */ 3412 struct proc *p = l->l_proc; 3413 struct vnode *vp; 3414 struct vattr vattr; 3415 int error; 3416 struct nameidata nd; 3417 3418 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3419 SCARG(uap, path)); 3420 if ((error = namei(&nd)) != 0) 3421 return (error); 3422 vp = nd.ni_vp; 3423 if (vp != NULL) { 3424 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3425 if (nd.ni_dvp == vp) 3426 vrele(nd.ni_dvp); 3427 else 3428 vput(nd.ni_dvp); 3429 vrele(vp); 3430 return (EEXIST); 3431 } 3432 VATTR_NULL(&vattr); 3433 vattr.va_type = VDIR; 3434 /* We will read cwdi->cwdi_cmask unlocked. */ 3435 vattr.va_mode = 3436 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3437 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3438 if (!error) 3439 vput(nd.ni_vp); 3440 return (error); 3441 } 3442 3443 /* 3444 * Remove a directory file. 3445 */ 3446 /* ARGSUSED */ 3447 int 3448 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3449 { 3450 /* { 3451 syscallarg(const char *) path; 3452 } */ 3453 struct vnode *vp; 3454 int error; 3455 struct nameidata nd; 3456 3457 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3458 SCARG(uap, path)); 3459 if ((error = namei(&nd)) != 0) 3460 return (error); 3461 vp = nd.ni_vp; 3462 if (vp->v_type != VDIR) { 3463 error = ENOTDIR; 3464 goto out; 3465 } 3466 /* 3467 * No rmdir "." please. 3468 */ 3469 if (nd.ni_dvp == vp) { 3470 error = EINVAL; 3471 goto out; 3472 } 3473 /* 3474 * The root of a mounted filesystem cannot be deleted. 3475 */ 3476 if (vp->v_vflag & VV_ROOT) { 3477 error = EBUSY; 3478 goto out; 3479 } 3480 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3481 return (error); 3482 3483 out: 3484 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3485 if (nd.ni_dvp == vp) 3486 vrele(nd.ni_dvp); 3487 else 3488 vput(nd.ni_dvp); 3489 vput(vp); 3490 return (error); 3491 } 3492 3493 /* 3494 * Read a block of directory entries in a file system independent format. 3495 */ 3496 int 3497 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3498 { 3499 /* { 3500 syscallarg(int) fd; 3501 syscallarg(char *) buf; 3502 syscallarg(size_t) count; 3503 } */ 3504 file_t *fp; 3505 int error, done; 3506 3507 /* fd_getvnode() will use the descriptor for us */ 3508 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3509 return (error); 3510 if ((fp->f_flag & FREAD) == 0) { 3511 error = EBADF; 3512 goto out; 3513 } 3514 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3515 SCARG(uap, count), &done, l, 0, 0); 3516 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3517 *retval = done; 3518 out: 3519 fd_putfile(SCARG(uap, fd)); 3520 return (error); 3521 } 3522 3523 /* 3524 * Set the mode mask for creation of filesystem nodes. 3525 */ 3526 int 3527 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3528 { 3529 /* { 3530 syscallarg(mode_t) newmask; 3531 } */ 3532 struct proc *p = l->l_proc; 3533 struct cwdinfo *cwdi; 3534 3535 /* 3536 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3537 * important is that we serialize changes to the mask. The 3538 * rw_exit() will issue a write memory barrier on our behalf, 3539 * and force the changes out to other CPUs (as it must use an 3540 * atomic operation, draining the local CPU's store buffers). 3541 */ 3542 cwdi = p->p_cwdi; 3543 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3544 *retval = cwdi->cwdi_cmask; 3545 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3546 rw_exit(&cwdi->cwdi_lock); 3547 3548 return (0); 3549 } 3550 3551 int 3552 dorevoke(struct vnode *vp, kauth_cred_t cred) 3553 { 3554 struct vattr vattr; 3555 int error; 3556 3557 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3558 return error; 3559 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3560 (error = kauth_authorize_generic(cred, 3561 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3562 VOP_REVOKE(vp, REVOKEALL); 3563 return (error); 3564 } 3565 3566 /* 3567 * Void all references to file by ripping underlying filesystem 3568 * away from vnode. 3569 */ 3570 /* ARGSUSED */ 3571 int 3572 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3573 { 3574 /* { 3575 syscallarg(const char *) path; 3576 } */ 3577 struct vnode *vp; 3578 int error; 3579 struct nameidata nd; 3580 3581 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3582 SCARG(uap, path)); 3583 if ((error = namei(&nd)) != 0) 3584 return (error); 3585 vp = nd.ni_vp; 3586 error = dorevoke(vp, l->l_cred); 3587 vrele(vp); 3588 return (error); 3589 } 3590 3591 /* 3592 * Convert a user file descriptor to a kernel file entry. 3593 */ 3594 int 3595 getvnode(int fd, file_t **fpp) 3596 { 3597 struct vnode *vp; 3598 file_t *fp; 3599 3600 if ((fp = fd_getfile(fd)) == NULL) 3601 return (EBADF); 3602 3603 if (fp->f_type != DTYPE_VNODE) { 3604 fd_putfile(fd); 3605 return (EINVAL); 3606 } 3607 3608 vp = fp->f_data; 3609 if (vp->v_type == VBAD) { 3610 fd_putfile(fd); 3611 return (EBADF); 3612 } 3613 3614 *fpp = fp; 3615 return (0); 3616 } 3617
Indexes created Fri Sep 26 20:09:58 GMT 2025