vfs_syscalls.c revision 1.351
1 /* $NetBSD: vfs_syscalls.c,v 1.351 2008/04/28 20:24:05 martin 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 * 16 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 17 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 18 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 19 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 20 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 21 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 22 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 23 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 24 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 25 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 26 * POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 /* 30 * Copyright (c) 1989, 1993 31 * The Regents of the University of California. All rights reserved. 32 * (c) UNIX System Laboratories, Inc. 33 * All or some portions of this file are derived from material licensed 34 * to the University of California by American Telephone and Telegraph 35 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 36 * the permission of UNIX System Laboratories, Inc. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 63 */ 64 65 #include <sys/cdefs.h> 66 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.351 2008/04/28 20:24:05 martin Exp $"); 67 68 #include "opt_compat_netbsd.h" 69 #include "opt_compat_43.h" 70 #include "opt_fileassoc.h" 71 #include "fss.h" 72 #include "veriexec.h" 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/namei.h> 77 #include <sys/filedesc.h> 78 #include <sys/kernel.h> 79 #include <sys/file.h> 80 #include <sys/stat.h> 81 #include <sys/vnode.h> 82 #include <sys/mount.h> 83 #include <sys/proc.h> 84 #include <sys/uio.h> 85 #include <sys/malloc.h> 86 #include <sys/kmem.h> 87 #include <sys/dirent.h> 88 #include <sys/sysctl.h> 89 #include <sys/syscallargs.h> 90 #include <sys/vfs_syscalls.h> 91 #include <sys/ktrace.h> 92 #ifdef FILEASSOC 93 #include <sys/fileassoc.h> 94 #endif /* FILEASSOC */ 95 #include <sys/verified_exec.h> 96 #include <sys/kauth.h> 97 #include <sys/atomic.h> 98 99 #include <miscfs/genfs/genfs.h> 100 #include <miscfs/syncfs/syncfs.h> 101 #include <miscfs/specfs/specdev.h> 102 103 #ifdef COMPAT_30 104 #include "opt_nfsserver.h" 105 #include <nfs/rpcv2.h> 106 #endif 107 #include <nfs/nfsproto.h> 108 #ifdef COMPAT_30 109 #include <nfs/nfs.h> 110 #include <nfs/nfs_var.h> 111 #endif 112 113 #if NFSS > 0 114 #include <dev/fssvar.h> 115 #endif 116 117 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 118 119 static int change_dir(struct nameidata *, struct lwp *); 120 static int change_flags(struct vnode *, u_long, struct lwp *); 121 static int change_mode(struct vnode *, int, struct lwp *l); 122 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 123 124 void checkdirs(struct vnode *); 125 126 int dovfsusermount = 0; 127 128 /* 129 * Virtual File System System Calls 130 */ 131 132 /* 133 * Mount a file system. 134 */ 135 136 #if defined(COMPAT_09) || defined(COMPAT_43) 137 /* 138 * This table is used to maintain compatibility with 4.3BSD 139 * and NetBSD 0.9 mount syscalls. Note, the order is important! 140 * 141 * Do not modify this table. It should only contain filesystems 142 * supported by NetBSD 0.9 and 4.3BSD. 143 */ 144 const char * const mountcompatnames[] = { 145 NULL, /* 0 = MOUNT_NONE */ 146 MOUNT_FFS, /* 1 = MOUNT_UFS */ 147 MOUNT_NFS, /* 2 */ 148 MOUNT_MFS, /* 3 */ 149 MOUNT_MSDOS, /* 4 */ 150 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 151 MOUNT_FDESC, /* 6 */ 152 MOUNT_KERNFS, /* 7 */ 153 NULL, /* 8 = MOUNT_DEVFS */ 154 MOUNT_AFS, /* 9 */ 155 }; 156 const int nmountcompatnames = sizeof(mountcompatnames) / 157 sizeof(mountcompatnames[0]); 158 #endif /* COMPAT_09 || COMPAT_43 */ 159 160 static int 161 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 162 void *data, size_t *data_len) 163 { 164 struct mount *mp; 165 int error = 0, saved_flags; 166 167 mp = vp->v_mount; 168 saved_flags = mp->mnt_flag; 169 170 /* We can operate only on VV_ROOT nodes. */ 171 if ((vp->v_vflag & VV_ROOT) == 0) { 172 error = EINVAL; 173 goto out; 174 } 175 176 /* 177 * We only allow the filesystem to be reloaded if it 178 * is currently mounted read-only. 179 */ 180 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 181 error = EOPNOTSUPP; /* Needs translation */ 182 goto out; 183 } 184 185 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 186 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 187 if (error) 188 goto out; 189 190 if (vfs_trybusy(mp, RW_WRITER, 0)) { 191 error = EPERM; 192 goto out; 193 } 194 195 mp->mnt_flag &= ~MNT_OP_FLAGS; 196 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 197 198 /* 199 * Set the mount level flags. 200 */ 201 if (flags & MNT_RDONLY) 202 mp->mnt_flag |= MNT_RDONLY; 203 else if (mp->mnt_flag & MNT_RDONLY) 204 mp->mnt_iflag |= IMNT_WANTRDWR; 205 mp->mnt_flag &= 206 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 207 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 208 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 209 mp->mnt_flag |= flags & 210 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 211 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 212 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 213 MNT_IGNORE); 214 215 error = VFS_MOUNT(mp, path, data, data_len); 216 217 #if defined(COMPAT_30) && defined(NFSSERVER) 218 if (error && data != NULL) { 219 int error2; 220 221 /* Update failed; let's try and see if it was an 222 * export request. */ 223 error2 = nfs_update_exports_30(mp, path, data, l); 224 225 /* Only update error code if the export request was 226 * understood but some problem occurred while 227 * processing it. */ 228 if (error2 != EJUSTRETURN) 229 error = error2; 230 } 231 #endif 232 if (mp->mnt_iflag & IMNT_WANTRDWR) 233 mp->mnt_flag &= ~MNT_RDONLY; 234 if (error) 235 mp->mnt_flag = saved_flags; 236 mp->mnt_flag &= ~MNT_OP_FLAGS; 237 mp->mnt_iflag &= ~IMNT_WANTRDWR; 238 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 239 if (mp->mnt_syncer == NULL) 240 error = vfs_allocate_syncvnode(mp); 241 } else { 242 if (mp->mnt_syncer != NULL) 243 vfs_deallocate_syncvnode(mp); 244 } 245 vfs_unbusy(mp, false); 246 247 out: 248 return (error); 249 } 250 251 static int 252 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 253 { 254 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 255 int error; 256 257 /* Copy file-system type from userspace. */ 258 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 259 if (error) { 260 #if defined(COMPAT_09) || defined(COMPAT_43) 261 /* 262 * Historically, filesystem types were identified by numbers. 263 * If we get an integer for the filesystem type instead of a 264 * string, we check to see if it matches one of the historic 265 * filesystem types. 266 */ 267 u_long fsindex = (u_long)fstype; 268 if (fsindex >= nmountcompatnames || 269 mountcompatnames[fsindex] == NULL) 270 return ENODEV; 271 strlcpy(fstypename, mountcompatnames[fsindex], 272 sizeof(fstypename)); 273 #else 274 return error; 275 #endif 276 } 277 278 #ifdef COMPAT_10 279 /* Accept `ufs' as an alias for `ffs'. */ 280 if (strcmp(fstypename, "ufs") == 0) 281 fstypename[0] = 'f'; 282 #endif 283 284 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 285 return ENODEV; 286 return 0; 287 } 288 289 static int 290 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 291 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 292 { 293 struct mount *mp = NULL; 294 struct vnode *vp = *vpp; 295 struct vattr va; 296 int error; 297 298 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 299 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 300 if (error) 301 return error; 302 303 /* Can't make a non-dir a mount-point (from here anyway). */ 304 if (vp->v_type != VDIR) 305 return ENOTDIR; 306 307 /* 308 * If the user is not root, ensure that they own the directory 309 * onto which we are attempting to mount. 310 */ 311 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 312 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 313 (error = kauth_authorize_generic(l->l_cred, 314 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 315 return error; 316 } 317 318 if (flags & MNT_EXPORTED) 319 return EINVAL; 320 321 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 322 return error; 323 324 /* 325 * Check if a file-system is not already mounted on this vnode. 326 */ 327 if (vp->v_mountedhere != NULL) 328 return EBUSY; 329 330 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP); 331 if (mp == NULL) 332 return ENOMEM; 333 334 mp->mnt_op = vfsops; 335 mp->mnt_refcnt = 1; 336 337 TAILQ_INIT(&mp->mnt_vnodelist); 338 rw_init(&mp->mnt_lock); 339 mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE); 340 (void)vfs_busy(mp, RW_WRITER, 0); 341 342 mp->mnt_vnodecovered = vp; 343 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 344 mount_initspecific(mp); 345 346 /* 347 * The underlying file system may refuse the mount for 348 * various reasons. Allow the user to force it to happen. 349 * 350 * Set the mount level flags. 351 */ 352 mp->mnt_flag = flags & 353 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 354 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 355 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 356 MNT_IGNORE | MNT_RDONLY); 357 358 error = VFS_MOUNT(mp, path, data, data_len); 359 mp->mnt_flag &= ~MNT_OP_FLAGS; 360 361 /* 362 * Put the new filesystem on the mount list after root. 363 */ 364 cache_purge(vp); 365 if (error != 0) { 366 vp->v_mountedhere = NULL; 367 mp->mnt_op->vfs_refcount--; 368 vfs_unbusy(mp, false); 369 vfs_destroy(mp); 370 return error; 371 } 372 373 mp->mnt_iflag &= ~IMNT_WANTRDWR; 374 mutex_enter(&mountlist_lock); 375 vp->v_mountedhere = mp; 376 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 377 mutex_exit(&mountlist_lock); 378 vn_restorerecurse(vp, recurse); 379 VOP_UNLOCK(vp, 0); 380 checkdirs(vp); 381 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 382 error = vfs_allocate_syncvnode(mp); 383 /* Hold an additional reference to the mount across VFS_START(). */ 384 vfs_unbusy(mp, true); 385 (void) VFS_STATVFS(mp, &mp->mnt_stat); 386 error = VFS_START(mp, 0); 387 if (error) { 388 vrele(vp); 389 vfs_destroy(mp); 390 } 391 /* Drop reference held for VFS_START(). */ 392 vfs_destroy(mp); 393 *vpp = NULL; 394 return error; 395 } 396 397 static int 398 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 399 void *data, size_t *data_len) 400 { 401 struct mount *mp; 402 int error; 403 404 /* If MNT_GETARGS is specified, it should be the only flag. */ 405 if (flags & ~MNT_GETARGS) 406 return EINVAL; 407 408 mp = vp->v_mount; 409 410 /* XXX: probably some notion of "can see" here if we want isolation. */ 411 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 412 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 413 if (error) 414 return error; 415 416 if ((vp->v_vflag & VV_ROOT) == 0) 417 return EINVAL; 418 419 if (vfs_trybusy(mp, RW_WRITER, NULL)) 420 return EPERM; 421 422 mp->mnt_flag &= ~MNT_OP_FLAGS; 423 mp->mnt_flag |= MNT_GETARGS; 424 error = VFS_MOUNT(mp, path, data, data_len); 425 mp->mnt_flag &= ~MNT_OP_FLAGS; 426 427 vfs_unbusy(mp, false); 428 return (error); 429 } 430 431 #ifdef COMPAT_40 432 /* ARGSUSED */ 433 int 434 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 435 { 436 /* { 437 syscallarg(const char *) type; 438 syscallarg(const char *) path; 439 syscallarg(int) flags; 440 syscallarg(void *) data; 441 } */ 442 register_t dummy; 443 444 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 445 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 446 } 447 #endif 448 449 int 450 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 451 { 452 /* { 453 syscallarg(const char *) type; 454 syscallarg(const char *) path; 455 syscallarg(int) flags; 456 syscallarg(void *) data; 457 syscallarg(size_t) data_len; 458 } */ 459 460 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 461 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 462 SCARG(uap, data_len), retval); 463 } 464 465 int 466 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 467 const char *path, int flags, void *data, enum uio_seg data_seg, 468 size_t data_len, register_t *retval) 469 { 470 struct vnode *vp; 471 struct nameidata nd; 472 void *data_buf = data; 473 u_int recurse; 474 int error; 475 476 /* 477 * Get vnode to be covered 478 */ 479 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 480 if ((error = namei(&nd)) != 0) 481 return (error); 482 vp = nd.ni_vp; 483 484 /* 485 * A lookup in VFS_MOUNT might result in an attempt to 486 * lock this vnode again, so make the lock recursive. 487 */ 488 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 489 recurse = vn_setrecurse(vp); 490 491 if (vfsops == NULL) { 492 if (flags & (MNT_GETARGS | MNT_UPDATE)) 493 vfsops = vp->v_mount->mnt_op; 494 else { 495 /* 'type' is userspace */ 496 error = mount_get_vfsops(type, &vfsops); 497 if (error != 0) 498 goto done; 499 } 500 } 501 502 if (data != NULL && data_seg == UIO_USERSPACE) { 503 if (data_len == 0) { 504 /* No length supplied, use default for filesystem */ 505 data_len = vfsops->vfs_min_mount_data; 506 if (data_len > VFS_MAX_MOUNT_DATA) { 507 /* maybe a force loaded old LKM */ 508 error = EINVAL; 509 goto done; 510 } 511 #ifdef COMPAT_30 512 /* Hopefully a longer buffer won't make copyin() fail */ 513 if (flags & MNT_UPDATE 514 && data_len < sizeof (struct mnt_export_args30)) 515 data_len = sizeof (struct mnt_export_args30); 516 #endif 517 } 518 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 519 520 /* NFS needs the buffer even for mnt_getargs .... */ 521 error = copyin(data, data_buf, data_len); 522 if (error != 0) 523 goto done; 524 } 525 526 if (flags & MNT_GETARGS) { 527 if (data_len == 0) { 528 error = EINVAL; 529 goto done; 530 } 531 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 532 if (error != 0) 533 goto done; 534 if (data_seg == UIO_USERSPACE) 535 error = copyout(data_buf, data, data_len); 536 *retval = data_len; 537 } else if (flags & MNT_UPDATE) { 538 error = mount_update(l, vp, path, flags, data_buf, &data_len); 539 } else { 540 /* Locking is handled internally in mount_domount(). */ 541 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 542 &data_len, recurse); 543 } 544 545 done: 546 if (vp != NULL) { 547 vn_restorerecurse(vp, recurse); 548 vput(vp); 549 } 550 if (data_buf != data) 551 free(data_buf, M_TEMP); 552 return (error); 553 } 554 555 /* 556 * Scan all active processes to see if any of them have a current 557 * or root directory onto which the new filesystem has just been 558 * mounted. If so, replace them with the new mount point. 559 */ 560 void 561 checkdirs(struct vnode *olddp) 562 { 563 struct cwdinfo *cwdi; 564 struct vnode *newdp; 565 struct proc *p; 566 567 if (olddp->v_usecount == 1) 568 return; 569 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 570 panic("mount: lost mount"); 571 mutex_enter(proc_lock); 572 /* XXXAD Should not be acquiring these locks with proc_lock held!! */ 573 PROCLIST_FOREACH(p, &allproc) { 574 cwdi = p->p_cwdi; 575 if (!cwdi) 576 continue; 577 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 578 if (cwdi->cwdi_cdir == olddp) { 579 vrele(cwdi->cwdi_cdir); 580 VREF(newdp); 581 cwdi->cwdi_cdir = newdp; 582 } 583 if (cwdi->cwdi_rdir == olddp) { 584 vrele(cwdi->cwdi_rdir); 585 VREF(newdp); 586 cwdi->cwdi_rdir = newdp; 587 } 588 rw_exit(&cwdi->cwdi_lock); 589 } 590 mutex_exit(proc_lock); 591 if (rootvnode == olddp) { 592 vrele(rootvnode); 593 VREF(newdp); 594 rootvnode = newdp; 595 } 596 vput(newdp); 597 } 598 599 /* 600 * Unmount a file system. 601 * 602 * Note: unmount takes a path to the vnode mounted on as argument, 603 * not special file (as before). 604 */ 605 /* ARGSUSED */ 606 int 607 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 608 { 609 /* { 610 syscallarg(const char *) path; 611 syscallarg(int) flags; 612 } */ 613 struct vnode *vp; 614 struct mount *mp; 615 int error; 616 struct nameidata nd; 617 618 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 619 SCARG(uap, path)); 620 if ((error = namei(&nd)) != 0) 621 return (error); 622 vp = nd.ni_vp; 623 mp = vp->v_mount; 624 VOP_UNLOCK(vp, 0); 625 626 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 627 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 628 if (error) { 629 vrele(vp); 630 return (error); 631 } 632 633 /* 634 * Don't allow unmounting the root file system. 635 */ 636 if (mp->mnt_flag & MNT_ROOTFS) { 637 vrele(vp); 638 return (EINVAL); 639 } 640 641 /* 642 * Must be the root of the filesystem 643 */ 644 if ((vp->v_vflag & VV_ROOT) == 0) { 645 vrele(vp); 646 return (EINVAL); 647 } 648 649 /* 650 * XXX Freeze syncer. Must do this before locking the 651 * mount point. See dounmount() for details. 652 */ 653 mutex_enter(&syncer_mutex); 654 error = vfs_busy(mp, RW_WRITER, NULL); 655 vrele(vp); 656 if (error != 0) { 657 mutex_exit(&syncer_mutex); 658 return (error); 659 } 660 661 return (dounmount(mp, SCARG(uap, flags), l)); 662 } 663 664 /* 665 * Do the actual file system unmount. File system is assumed to have been 666 * marked busy by the caller. 667 */ 668 int 669 dounmount(struct mount *mp, int flags, struct lwp *l) 670 { 671 struct vnode *coveredvp; 672 int error; 673 int async; 674 int used_syncer; 675 676 KASSERT(rw_write_held(&mp->mnt_lock)); 677 678 #if NVERIEXEC > 0 679 error = veriexec_unmountchk(mp); 680 if (error) 681 return (error); 682 #endif /* NVERIEXEC > 0 */ 683 684 used_syncer = (mp->mnt_syncer != NULL); 685 686 /* 687 * XXX Syncer must be frozen when we get here. This should really 688 * be done on a per-mountpoint basis, but especially the softdep 689 * code possibly called from the syncer doesn't exactly work on a 690 * per-mountpoint basis, so the softdep code would become a maze 691 * of vfs_busy() calls. 692 * 693 * The caller of dounmount() must acquire syncer_mutex because 694 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 695 * order, and we must preserve that order to avoid deadlock. 696 * 697 * So, if the file system did not use the syncer, now is 698 * the time to release the syncer_mutex. 699 */ 700 if (used_syncer == 0) 701 mutex_exit(&syncer_mutex); 702 703 mp->mnt_iflag |= IMNT_UNMOUNT; 704 async = mp->mnt_flag & MNT_ASYNC; 705 mp->mnt_flag &= ~MNT_ASYNC; 706 cache_purgevfs(mp); /* remove cache entries for this file sys */ 707 if (mp->mnt_syncer != NULL) 708 vfs_deallocate_syncvnode(mp); 709 error = 0; 710 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 711 #if NFSS > 0 712 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 713 #endif 714 if (error == 0) 715 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 716 } 717 vfs_scrubvnlist(mp); 718 if (error == 0 || (flags & MNT_FORCE)) 719 error = VFS_UNMOUNT(mp, flags); 720 if (error) { 721 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 722 (void) vfs_allocate_syncvnode(mp); 723 mp->mnt_iflag &= ~IMNT_UNMOUNT; 724 mp->mnt_flag |= async; 725 if (used_syncer) 726 mutex_exit(&syncer_mutex); 727 vfs_unbusy(mp, false); 728 return (error); 729 } 730 vfs_scrubvnlist(mp); 731 mutex_enter(&mountlist_lock); 732 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 733 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 734 coveredvp->v_mountedhere = NULL; 735 mutex_exit(&mountlist_lock); 736 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 737 panic("unmount: dangling vnode"); 738 mp->mnt_iflag |= IMNT_GONE; 739 if (used_syncer) 740 mutex_exit(&syncer_mutex); 741 vfs_hooks_unmount(mp); 742 vfs_unbusy(mp, false); 743 vfs_destroy(mp); 744 if (coveredvp != NULLVP) 745 vrele(coveredvp); 746 return (0); 747 } 748 749 /* 750 * Sync each mounted filesystem. 751 */ 752 #ifdef DEBUG 753 int syncprt = 0; 754 struct ctldebug debug0 = { "syncprt", &syncprt }; 755 #endif 756 757 /* ARGSUSED */ 758 int 759 sys_sync(struct lwp *l, const void *v, register_t *retval) 760 { 761 struct mount *mp, *nmp; 762 int asyncflag; 763 764 if (l == NULL) 765 l = &lwp0; 766 767 mutex_enter(&mountlist_lock); 768 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 769 if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) { 770 nmp = mp->mnt_list.cqe_prev; 771 continue; 772 } 773 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 774 asyncflag = mp->mnt_flag & MNT_ASYNC; 775 mp->mnt_flag &= ~MNT_ASYNC; 776 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 777 if (asyncflag) 778 mp->mnt_flag |= MNT_ASYNC; 779 } 780 mutex_enter(&mountlist_lock); 781 nmp = mp->mnt_list.cqe_prev; 782 vfs_unbusy(mp, false); 783 784 } 785 mutex_exit(&mountlist_lock); 786 #ifdef DEBUG 787 if (syncprt) 788 vfs_bufstats(); 789 #endif /* DEBUG */ 790 return (0); 791 } 792 793 /* 794 * Change filesystem quotas. 795 */ 796 /* ARGSUSED */ 797 int 798 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 799 { 800 /* { 801 syscallarg(const char *) path; 802 syscallarg(int) cmd; 803 syscallarg(int) uid; 804 syscallarg(void *) arg; 805 } */ 806 struct mount *mp; 807 int error; 808 struct nameidata nd; 809 810 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 811 SCARG(uap, path)); 812 if ((error = namei(&nd)) != 0) 813 return (error); 814 mp = nd.ni_vp->v_mount; 815 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 816 SCARG(uap, arg)); 817 vrele(nd.ni_vp); 818 return (error); 819 } 820 821 int 822 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 823 int root) 824 { 825 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 826 int error = 0; 827 828 /* 829 * If MNT_NOWAIT or MNT_LAZY is specified, do not 830 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 831 * overrides MNT_NOWAIT. 832 */ 833 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 834 (flags != MNT_WAIT && flags != 0)) { 835 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 836 goto done; 837 } 838 839 /* Get the filesystem stats now */ 840 memset(sp, 0, sizeof(*sp)); 841 if ((error = VFS_STATVFS(mp, sp)) != 0) { 842 return error; 843 } 844 845 if (cwdi->cwdi_rdir == NULL) 846 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 847 done: 848 if (cwdi->cwdi_rdir != NULL) { 849 size_t len; 850 char *bp; 851 char *path = PNBUF_GET(); 852 853 bp = path + MAXPATHLEN; 854 *--bp = '\0'; 855 rw_enter(&cwdi->cwdi_lock, RW_READER); 856 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 857 MAXPATHLEN / 2, 0, l); 858 rw_exit(&cwdi->cwdi_lock); 859 if (error) { 860 PNBUF_PUT(path); 861 return error; 862 } 863 len = strlen(bp); 864 /* 865 * for mount points that are below our root, we can see 866 * them, so we fix up the pathname and return them. The 867 * rest we cannot see, so we don't allow viewing the 868 * data. 869 */ 870 if (strncmp(bp, sp->f_mntonname, len) == 0) { 871 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 872 sizeof(sp->f_mntonname)); 873 if (sp->f_mntonname[0] == '\0') 874 (void)strlcpy(sp->f_mntonname, "/", 875 sizeof(sp->f_mntonname)); 876 } else { 877 if (root) 878 (void)strlcpy(sp->f_mntonname, "/", 879 sizeof(sp->f_mntonname)); 880 else 881 error = EPERM; 882 } 883 PNBUF_PUT(path); 884 } 885 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 886 return error; 887 } 888 889 /* 890 * Get filesystem statistics by path. 891 */ 892 int 893 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 894 { 895 struct mount *mp; 896 int error; 897 struct nameidata nd; 898 899 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 900 if ((error = namei(&nd)) != 0) 901 return error; 902 mp = nd.ni_vp->v_mount; 903 error = dostatvfs(mp, sb, l, flags, 1); 904 vrele(nd.ni_vp); 905 return error; 906 } 907 908 /* ARGSUSED */ 909 int 910 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 911 { 912 /* { 913 syscallarg(const char *) path; 914 syscallarg(struct statvfs *) buf; 915 syscallarg(int) flags; 916 } */ 917 struct statvfs *sb; 918 int error; 919 920 sb = STATVFSBUF_GET(); 921 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 922 if (error == 0) 923 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 924 STATVFSBUF_PUT(sb); 925 return error; 926 } 927 928 /* 929 * Get filesystem statistics by fd. 930 */ 931 int 932 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 933 { 934 file_t *fp; 935 struct mount *mp; 936 int error; 937 938 /* fd_getvnode() will use the descriptor for us */ 939 if ((error = fd_getvnode(fd, &fp)) != 0) 940 return (error); 941 mp = ((struct vnode *)fp->f_data)->v_mount; 942 error = dostatvfs(mp, sb, curlwp, flags, 1); 943 fd_putfile(fd); 944 return error; 945 } 946 947 /* ARGSUSED */ 948 int 949 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 950 { 951 /* { 952 syscallarg(int) fd; 953 syscallarg(struct statvfs *) buf; 954 syscallarg(int) flags; 955 } */ 956 struct statvfs *sb; 957 int error; 958 959 sb = STATVFSBUF_GET(); 960 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 961 if (error == 0) 962 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 963 STATVFSBUF_PUT(sb); 964 return error; 965 } 966 967 968 /* 969 * Get statistics on all filesystems. 970 */ 971 int 972 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 973 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 974 register_t *retval) 975 { 976 int root = 0; 977 struct proc *p = l->l_proc; 978 struct mount *mp, *nmp; 979 struct statvfs *sb; 980 size_t count, maxcount; 981 int error = 0; 982 983 sb = STATVFSBUF_GET(); 984 maxcount = bufsize / entry_sz; 985 mutex_enter(&mountlist_lock); 986 count = 0; 987 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 988 mp = nmp) { 989 if (vfs_trybusy(mp, RW_READER, &mountlist_lock)) { 990 nmp = CIRCLEQ_NEXT(mp, mnt_list); 991 continue; 992 } 993 if (sfsp && count < maxcount) { 994 error = dostatvfs(mp, sb, l, flags, 0); 995 if (error) { 996 mutex_enter(&mountlist_lock); 997 nmp = CIRCLEQ_NEXT(mp, mnt_list); 998 vfs_unbusy(mp, false); 999 continue; 1000 } 1001 error = copyfn(sb, sfsp, entry_sz); 1002 if (error) { 1003 vfs_unbusy(mp, false); 1004 goto out; 1005 } 1006 sfsp = (char *)sfsp + entry_sz; 1007 root |= strcmp(sb->f_mntonname, "/") == 0; 1008 } 1009 count++; 1010 mutex_enter(&mountlist_lock); 1011 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1012 vfs_unbusy(mp, false); 1013 } 1014 1015 mutex_exit(&mountlist_lock); 1016 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1017 /* 1018 * fake a root entry 1019 */ 1020 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1021 sb, l, flags, 1); 1022 if (error != 0) 1023 goto out; 1024 if (sfsp) 1025 error = copyfn(sb, sfsp, entry_sz); 1026 count++; 1027 } 1028 if (sfsp && count > maxcount) 1029 *retval = maxcount; 1030 else 1031 *retval = count; 1032 out: 1033 STATVFSBUF_PUT(sb); 1034 return error; 1035 } 1036 1037 int 1038 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1039 { 1040 /* { 1041 syscallarg(struct statvfs *) buf; 1042 syscallarg(size_t) bufsize; 1043 syscallarg(int) flags; 1044 } */ 1045 1046 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1047 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1048 } 1049 1050 /* 1051 * Change current working directory to a given file descriptor. 1052 */ 1053 /* ARGSUSED */ 1054 int 1055 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1056 { 1057 /* { 1058 syscallarg(int) fd; 1059 } */ 1060 struct proc *p = l->l_proc; 1061 struct cwdinfo *cwdi; 1062 struct vnode *vp, *tdp; 1063 struct mount *mp; 1064 file_t *fp; 1065 int error, fd; 1066 1067 /* fd_getvnode() will use the descriptor for us */ 1068 fd = SCARG(uap, fd); 1069 if ((error = fd_getvnode(fd, &fp)) != 0) 1070 return (error); 1071 vp = fp->f_data; 1072 1073 VREF(vp); 1074 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1075 if (vp->v_type != VDIR) 1076 error = ENOTDIR; 1077 else 1078 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1079 if (error) { 1080 vput(vp); 1081 goto out; 1082 } 1083 while ((mp = vp->v_mountedhere) != NULL) { 1084 if (vfs_busy(mp, RW_READER, NULL)) 1085 continue; 1086 vput(vp); 1087 error = VFS_ROOT(mp, &tdp); 1088 vfs_unbusy(mp, false); 1089 if (error) 1090 goto out; 1091 vp = tdp; 1092 } 1093 VOP_UNLOCK(vp, 0); 1094 1095 /* 1096 * Disallow changing to a directory not under the process's 1097 * current root directory (if there is one). 1098 */ 1099 cwdi = p->p_cwdi; 1100 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1101 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1102 vrele(vp); 1103 error = EPERM; /* operation not permitted */ 1104 } else { 1105 vrele(cwdi->cwdi_cdir); 1106 cwdi->cwdi_cdir = vp; 1107 } 1108 rw_exit(&cwdi->cwdi_lock); 1109 1110 out: 1111 fd_putfile(fd); 1112 return (error); 1113 } 1114 1115 /* 1116 * Change this process's notion of the root directory to a given file 1117 * descriptor. 1118 */ 1119 int 1120 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1121 { 1122 struct proc *p = l->l_proc; 1123 struct cwdinfo *cwdi; 1124 struct vnode *vp; 1125 file_t *fp; 1126 int error, fd = SCARG(uap, fd); 1127 1128 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1129 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1130 return error; 1131 /* fd_getvnode() will use the descriptor for us */ 1132 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 1133 return error; 1134 vp = fp->f_data; 1135 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1136 if (vp->v_type != VDIR) 1137 error = ENOTDIR; 1138 else 1139 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1140 VOP_UNLOCK(vp, 0); 1141 if (error) 1142 goto out; 1143 VREF(vp); 1144 1145 /* 1146 * Prevent escaping from chroot by putting the root under 1147 * the working directory. Silently chdir to / if we aren't 1148 * already there. 1149 */ 1150 cwdi = p->p_cwdi; 1151 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1152 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1153 /* 1154 * XXX would be more failsafe to change directory to a 1155 * deadfs node here instead 1156 */ 1157 vrele(cwdi->cwdi_cdir); 1158 VREF(vp); 1159 cwdi->cwdi_cdir = vp; 1160 } 1161 1162 if (cwdi->cwdi_rdir != NULL) 1163 vrele(cwdi->cwdi_rdir); 1164 cwdi->cwdi_rdir = vp; 1165 rw_exit(&cwdi->cwdi_lock); 1166 1167 out: 1168 fd_putfile(fd); 1169 return (error); 1170 } 1171 1172 /* 1173 * Change current working directory (``.''). 1174 */ 1175 /* ARGSUSED */ 1176 int 1177 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1178 { 1179 /* { 1180 syscallarg(const char *) path; 1181 } */ 1182 struct proc *p = l->l_proc; 1183 struct cwdinfo *cwdi; 1184 int error; 1185 struct nameidata nd; 1186 1187 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1188 SCARG(uap, path)); 1189 if ((error = change_dir(&nd, l)) != 0) 1190 return (error); 1191 cwdi = p->p_cwdi; 1192 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1193 vrele(cwdi->cwdi_cdir); 1194 cwdi->cwdi_cdir = nd.ni_vp; 1195 rw_exit(&cwdi->cwdi_lock); 1196 return (0); 1197 } 1198 1199 /* 1200 * Change notion of root (``/'') directory. 1201 */ 1202 /* ARGSUSED */ 1203 int 1204 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1205 { 1206 /* { 1207 syscallarg(const char *) path; 1208 } */ 1209 struct proc *p = l->l_proc; 1210 struct cwdinfo *cwdi; 1211 struct vnode *vp; 1212 int error; 1213 struct nameidata nd; 1214 1215 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1216 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1217 return (error); 1218 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1219 SCARG(uap, path)); 1220 if ((error = change_dir(&nd, l)) != 0) 1221 return (error); 1222 1223 cwdi = p->p_cwdi; 1224 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1225 if (cwdi->cwdi_rdir != NULL) 1226 vrele(cwdi->cwdi_rdir); 1227 vp = nd.ni_vp; 1228 cwdi->cwdi_rdir = vp; 1229 1230 /* 1231 * Prevent escaping from chroot by putting the root under 1232 * the working directory. Silently chdir to / if we aren't 1233 * already there. 1234 */ 1235 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1236 /* 1237 * XXX would be more failsafe to change directory to a 1238 * deadfs node here instead 1239 */ 1240 vrele(cwdi->cwdi_cdir); 1241 VREF(vp); 1242 cwdi->cwdi_cdir = vp; 1243 } 1244 rw_exit(&cwdi->cwdi_lock); 1245 1246 return (0); 1247 } 1248 1249 /* 1250 * Common routine for chroot and chdir. 1251 */ 1252 static int 1253 change_dir(struct nameidata *ndp, struct lwp *l) 1254 { 1255 struct vnode *vp; 1256 int error; 1257 1258 if ((error = namei(ndp)) != 0) 1259 return (error); 1260 vp = ndp->ni_vp; 1261 if (vp->v_type != VDIR) 1262 error = ENOTDIR; 1263 else 1264 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1265 1266 if (error) 1267 vput(vp); 1268 else 1269 VOP_UNLOCK(vp, 0); 1270 return (error); 1271 } 1272 1273 /* 1274 * Check permissions, allocate an open file structure, 1275 * and call the device open routine if any. 1276 */ 1277 int 1278 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1279 { 1280 /* { 1281 syscallarg(const char *) path; 1282 syscallarg(int) flags; 1283 syscallarg(int) mode; 1284 } */ 1285 struct proc *p = l->l_proc; 1286 struct cwdinfo *cwdi = p->p_cwdi; 1287 file_t *fp; 1288 struct vnode *vp; 1289 int flags, cmode; 1290 int type, indx, error; 1291 struct flock lf; 1292 struct nameidata nd; 1293 1294 flags = FFLAGS(SCARG(uap, flags)); 1295 if ((flags & (FREAD | FWRITE)) == 0) 1296 return (EINVAL); 1297 if ((error = fd_allocfile(&fp, &indx)) != 0) 1298 return (error); 1299 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1300 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1301 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1302 SCARG(uap, path)); 1303 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1304 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1305 fd_abort(p, fp, indx); 1306 if ((error == EDUPFD || error == EMOVEFD) && 1307 l->l_dupfd >= 0 && /* XXX from fdopen */ 1308 (error = 1309 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1310 *retval = indx; 1311 return (0); 1312 } 1313 if (error == ERESTART) 1314 error = EINTR; 1315 return (error); 1316 } 1317 1318 l->l_dupfd = 0; 1319 vp = nd.ni_vp; 1320 fp->f_flag = flags & FMASK; 1321 fp->f_type = DTYPE_VNODE; 1322 fp->f_ops = &vnops; 1323 fp->f_data = vp; 1324 if (flags & (O_EXLOCK | O_SHLOCK)) { 1325 lf.l_whence = SEEK_SET; 1326 lf.l_start = 0; 1327 lf.l_len = 0; 1328 if (flags & O_EXLOCK) 1329 lf.l_type = F_WRLCK; 1330 else 1331 lf.l_type = F_RDLCK; 1332 type = F_FLOCK; 1333 if ((flags & FNONBLOCK) == 0) 1334 type |= F_WAIT; 1335 VOP_UNLOCK(vp, 0); 1336 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1337 if (error) { 1338 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1339 fd_abort(p, fp, indx); 1340 return (error); 1341 } 1342 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1343 atomic_or_uint(&fp->f_flag, FHASLOCK); 1344 } 1345 VOP_UNLOCK(vp, 0); 1346 *retval = indx; 1347 fd_affix(p, fp, indx); 1348 return (0); 1349 } 1350 1351 static void 1352 vfs__fhfree(fhandle_t *fhp) 1353 { 1354 size_t fhsize; 1355 1356 if (fhp == NULL) { 1357 return; 1358 } 1359 fhsize = FHANDLE_SIZE(fhp); 1360 kmem_free(fhp, fhsize); 1361 } 1362 1363 /* 1364 * vfs_composefh: compose a filehandle. 1365 */ 1366 1367 int 1368 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1369 { 1370 struct mount *mp; 1371 struct fid *fidp; 1372 int error; 1373 size_t needfhsize; 1374 size_t fidsize; 1375 1376 mp = vp->v_mount; 1377 fidp = NULL; 1378 if (*fh_size < FHANDLE_SIZE_MIN) { 1379 fidsize = 0; 1380 } else { 1381 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1382 if (fhp != NULL) { 1383 memset(fhp, 0, *fh_size); 1384 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1385 fidp = &fhp->fh_fid; 1386 } 1387 } 1388 error = VFS_VPTOFH(vp, fidp, &fidsize); 1389 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1390 if (error == 0 && *fh_size < needfhsize) { 1391 error = E2BIG; 1392 } 1393 *fh_size = needfhsize; 1394 return error; 1395 } 1396 1397 int 1398 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1399 { 1400 struct mount *mp; 1401 fhandle_t *fhp; 1402 size_t fhsize; 1403 size_t fidsize; 1404 int error; 1405 1406 *fhpp = NULL; 1407 mp = vp->v_mount; 1408 fidsize = 0; 1409 error = VFS_VPTOFH(vp, NULL, &fidsize); 1410 KASSERT(error != 0); 1411 if (error != E2BIG) { 1412 goto out; 1413 } 1414 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1415 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1416 if (fhp == NULL) { 1417 error = ENOMEM; 1418 goto out; 1419 } 1420 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1421 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1422 if (error == 0) { 1423 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1424 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1425 *fhpp = fhp; 1426 } else { 1427 kmem_free(fhp, fhsize); 1428 } 1429 out: 1430 return error; 1431 } 1432 1433 void 1434 vfs_composefh_free(fhandle_t *fhp) 1435 { 1436 1437 vfs__fhfree(fhp); 1438 } 1439 1440 /* 1441 * vfs_fhtovp: lookup a vnode by a filehandle. 1442 */ 1443 1444 int 1445 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1446 { 1447 struct mount *mp; 1448 int error; 1449 1450 *vpp = NULL; 1451 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1452 if (mp == NULL) { 1453 error = ESTALE; 1454 goto out; 1455 } 1456 if (mp->mnt_op->vfs_fhtovp == NULL) { 1457 error = EOPNOTSUPP; 1458 goto out; 1459 } 1460 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1461 out: 1462 return error; 1463 } 1464 1465 /* 1466 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1467 * the needed size. 1468 */ 1469 1470 int 1471 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1472 { 1473 fhandle_t *fhp; 1474 int error; 1475 1476 *fhpp = NULL; 1477 if (fhsize > FHANDLE_SIZE_MAX) { 1478 return EINVAL; 1479 } 1480 if (fhsize < FHANDLE_SIZE_MIN) { 1481 return EINVAL; 1482 } 1483 again: 1484 fhp = kmem_alloc(fhsize, KM_SLEEP); 1485 if (fhp == NULL) { 1486 return ENOMEM; 1487 } 1488 error = copyin(ufhp, fhp, fhsize); 1489 if (error == 0) { 1490 /* XXX this check shouldn't be here */ 1491 if (FHANDLE_SIZE(fhp) == fhsize) { 1492 *fhpp = fhp; 1493 return 0; 1494 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1495 /* 1496 * a kludge for nfsv2 padded handles. 1497 */ 1498 size_t sz; 1499 1500 sz = FHANDLE_SIZE(fhp); 1501 kmem_free(fhp, fhsize); 1502 fhsize = sz; 1503 goto again; 1504 } else { 1505 /* 1506 * userland told us wrong size. 1507 */ 1508 error = EINVAL; 1509 } 1510 } 1511 kmem_free(fhp, fhsize); 1512 return error; 1513 } 1514 1515 void 1516 vfs_copyinfh_free(fhandle_t *fhp) 1517 { 1518 1519 vfs__fhfree(fhp); 1520 } 1521 1522 /* 1523 * Get file handle system call 1524 */ 1525 int 1526 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1527 { 1528 /* { 1529 syscallarg(char *) fname; 1530 syscallarg(fhandle_t *) fhp; 1531 syscallarg(size_t *) fh_size; 1532 } */ 1533 struct vnode *vp; 1534 fhandle_t *fh; 1535 int error; 1536 struct nameidata nd; 1537 size_t sz; 1538 size_t usz; 1539 1540 /* 1541 * Must be super user 1542 */ 1543 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1544 0, NULL, NULL, NULL); 1545 if (error) 1546 return (error); 1547 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1548 SCARG(uap, fname)); 1549 error = namei(&nd); 1550 if (error) 1551 return (error); 1552 vp = nd.ni_vp; 1553 error = vfs_composefh_alloc(vp, &fh); 1554 vput(vp); 1555 if (error != 0) { 1556 goto out; 1557 } 1558 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1559 if (error != 0) { 1560 goto out; 1561 } 1562 sz = FHANDLE_SIZE(fh); 1563 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1564 if (error != 0) { 1565 goto out; 1566 } 1567 if (usz >= sz) { 1568 error = copyout(fh, SCARG(uap, fhp), sz); 1569 } else { 1570 error = E2BIG; 1571 } 1572 out: 1573 vfs_composefh_free(fh); 1574 return (error); 1575 } 1576 1577 /* 1578 * Open a file given a file handle. 1579 * 1580 * Check permissions, allocate an open file structure, 1581 * and call the device open routine if any. 1582 */ 1583 1584 int 1585 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1586 register_t *retval) 1587 { 1588 file_t *fp; 1589 struct vnode *vp = NULL; 1590 kauth_cred_t cred = l->l_cred; 1591 file_t *nfp; 1592 int type, indx, error=0; 1593 struct flock lf; 1594 struct vattr va; 1595 fhandle_t *fh; 1596 int flags; 1597 proc_t *p; 1598 1599 p = curproc; 1600 1601 /* 1602 * Must be super user 1603 */ 1604 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1605 0, NULL, NULL, NULL))) 1606 return (error); 1607 1608 flags = FFLAGS(oflags); 1609 if ((flags & (FREAD | FWRITE)) == 0) 1610 return (EINVAL); 1611 if ((flags & O_CREAT)) 1612 return (EINVAL); 1613 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1614 return (error); 1615 fp = nfp; 1616 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1617 if (error != 0) { 1618 goto bad; 1619 } 1620 error = vfs_fhtovp(fh, &vp); 1621 if (error != 0) { 1622 goto bad; 1623 } 1624 1625 /* Now do an effective vn_open */ 1626 1627 if (vp->v_type == VSOCK) { 1628 error = EOPNOTSUPP; 1629 goto bad; 1630 } 1631 error = vn_openchk(vp, cred, flags); 1632 if (error != 0) 1633 goto bad; 1634 if (flags & O_TRUNC) { 1635 VOP_UNLOCK(vp, 0); /* XXX */ 1636 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1637 VATTR_NULL(&va); 1638 va.va_size = 0; 1639 error = VOP_SETATTR(vp, &va, cred); 1640 if (error) 1641 goto bad; 1642 } 1643 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1644 goto bad; 1645 if (flags & FWRITE) { 1646 mutex_enter(&vp->v_interlock); 1647 vp->v_writecount++; 1648 mutex_exit(&vp->v_interlock); 1649 } 1650 1651 /* done with modified vn_open, now finish what sys_open does. */ 1652 1653 fp->f_flag = flags & FMASK; 1654 fp->f_type = DTYPE_VNODE; 1655 fp->f_ops = &vnops; 1656 fp->f_data = vp; 1657 if (flags & (O_EXLOCK | O_SHLOCK)) { 1658 lf.l_whence = SEEK_SET; 1659 lf.l_start = 0; 1660 lf.l_len = 0; 1661 if (flags & O_EXLOCK) 1662 lf.l_type = F_WRLCK; 1663 else 1664 lf.l_type = F_RDLCK; 1665 type = F_FLOCK; 1666 if ((flags & FNONBLOCK) == 0) 1667 type |= F_WAIT; 1668 VOP_UNLOCK(vp, 0); 1669 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1670 if (error) { 1671 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1672 fd_abort(p, fp, indx); 1673 return (error); 1674 } 1675 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1676 atomic_or_uint(&fp->f_flag, FHASLOCK); 1677 } 1678 VOP_UNLOCK(vp, 0); 1679 *retval = indx; 1680 fd_affix(p, fp, indx); 1681 vfs_copyinfh_free(fh); 1682 return (0); 1683 1684 bad: 1685 fd_abort(p, fp, indx); 1686 if (vp != NULL) 1687 vput(vp); 1688 vfs_copyinfh_free(fh); 1689 return (error); 1690 } 1691 1692 int 1693 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1694 { 1695 /* { 1696 syscallarg(const void *) fhp; 1697 syscallarg(size_t) fh_size; 1698 syscallarg(int) flags; 1699 } */ 1700 1701 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1702 SCARG(uap, flags), retval); 1703 } 1704 1705 int 1706 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1707 { 1708 int error; 1709 fhandle_t *fh; 1710 struct vnode *vp; 1711 1712 /* 1713 * Must be super user 1714 */ 1715 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1716 0, NULL, NULL, NULL))) 1717 return (error); 1718 1719 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1720 if (error != 0) 1721 return error; 1722 1723 error = vfs_fhtovp(fh, &vp); 1724 vfs_copyinfh_free(fh); 1725 if (error != 0) 1726 return error; 1727 1728 error = vn_stat(vp, sb); 1729 vput(vp); 1730 return error; 1731 } 1732 1733 1734 /* ARGSUSED */ 1735 int 1736 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1737 { 1738 /* { 1739 syscallarg(const void *) fhp; 1740 syscallarg(size_t) fh_size; 1741 syscallarg(struct stat *) sb; 1742 } */ 1743 struct stat sb; 1744 int error; 1745 1746 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1747 if (error) 1748 return error; 1749 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1750 } 1751 1752 int 1753 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1754 int flags) 1755 { 1756 fhandle_t *fh; 1757 struct mount *mp; 1758 struct vnode *vp; 1759 int error; 1760 1761 /* 1762 * Must be super user 1763 */ 1764 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1765 0, NULL, NULL, NULL))) 1766 return error; 1767 1768 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1769 if (error != 0) 1770 return error; 1771 1772 error = vfs_fhtovp(fh, &vp); 1773 vfs_copyinfh_free(fh); 1774 if (error != 0) 1775 return error; 1776 1777 mp = vp->v_mount; 1778 error = dostatvfs(mp, sb, l, flags, 1); 1779 vput(vp); 1780 return error; 1781 } 1782 1783 /* ARGSUSED */ 1784 int 1785 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1786 { 1787 /* { 1788 syscallarg(const void *) fhp; 1789 syscallarg(size_t) fh_size; 1790 syscallarg(struct statvfs *) buf; 1791 syscallarg(int) flags; 1792 } */ 1793 struct statvfs *sb = STATVFSBUF_GET(); 1794 int error; 1795 1796 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1797 SCARG(uap, flags)); 1798 if (error == 0) 1799 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1800 STATVFSBUF_PUT(sb); 1801 return error; 1802 } 1803 1804 /* 1805 * Create a special file. 1806 */ 1807 /* ARGSUSED */ 1808 int 1809 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1810 { 1811 /* { 1812 syscallarg(const char *) path; 1813 syscallarg(int) mode; 1814 syscallarg(int) dev; 1815 } */ 1816 struct proc *p = l->l_proc; 1817 struct vnode *vp; 1818 struct vattr vattr; 1819 int error, optype; 1820 struct nameidata nd; 1821 char *path; 1822 const char *cpath; 1823 enum uio_seg seg = UIO_USERSPACE; 1824 1825 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1826 0, NULL, NULL, NULL)) != 0) 1827 return (error); 1828 1829 optype = VOP_MKNOD_DESCOFFSET; 1830 1831 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1832 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1833 1834 if ((error = namei(&nd)) != 0) 1835 goto out; 1836 vp = nd.ni_vp; 1837 if (vp != NULL) 1838 error = EEXIST; 1839 else { 1840 VATTR_NULL(&vattr); 1841 /* We will read cwdi->cwdi_cmask unlocked. */ 1842 vattr.va_mode = 1843 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1844 vattr.va_rdev = SCARG(uap, dev); 1845 1846 switch (SCARG(uap, mode) & S_IFMT) { 1847 case S_IFMT: /* used by badsect to flag bad sectors */ 1848 vattr.va_type = VBAD; 1849 break; 1850 case S_IFCHR: 1851 vattr.va_type = VCHR; 1852 break; 1853 case S_IFBLK: 1854 vattr.va_type = VBLK; 1855 break; 1856 case S_IFWHT: 1857 optype = VOP_WHITEOUT_DESCOFFSET; 1858 break; 1859 case S_IFREG: 1860 #if NVERIEXEC > 0 1861 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1862 O_CREAT); 1863 #endif /* NVERIEXEC > 0 */ 1864 vattr.va_type = VREG; 1865 vattr.va_rdev = VNOVAL; 1866 optype = VOP_CREATE_DESCOFFSET; 1867 break; 1868 default: 1869 error = EINVAL; 1870 break; 1871 } 1872 } 1873 if (!error) { 1874 switch (optype) { 1875 case VOP_WHITEOUT_DESCOFFSET: 1876 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1877 if (error) 1878 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1879 vput(nd.ni_dvp); 1880 break; 1881 1882 case VOP_MKNOD_DESCOFFSET: 1883 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1884 &nd.ni_cnd, &vattr); 1885 if (error == 0) 1886 vput(nd.ni_vp); 1887 break; 1888 1889 case VOP_CREATE_DESCOFFSET: 1890 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1891 &nd.ni_cnd, &vattr); 1892 if (error == 0) 1893 vput(nd.ni_vp); 1894 break; 1895 } 1896 } else { 1897 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1898 if (nd.ni_dvp == vp) 1899 vrele(nd.ni_dvp); 1900 else 1901 vput(nd.ni_dvp); 1902 if (vp) 1903 vrele(vp); 1904 } 1905 out: 1906 VERIEXEC_PATH_PUT(path); 1907 return (error); 1908 } 1909 1910 /* 1911 * Create a named pipe. 1912 */ 1913 /* ARGSUSED */ 1914 int 1915 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1916 { 1917 /* { 1918 syscallarg(const char *) path; 1919 syscallarg(int) mode; 1920 } */ 1921 struct proc *p = l->l_proc; 1922 struct vattr vattr; 1923 int error; 1924 struct nameidata nd; 1925 1926 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1927 SCARG(uap, path)); 1928 if ((error = namei(&nd)) != 0) 1929 return (error); 1930 if (nd.ni_vp != NULL) { 1931 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1932 if (nd.ni_dvp == nd.ni_vp) 1933 vrele(nd.ni_dvp); 1934 else 1935 vput(nd.ni_dvp); 1936 vrele(nd.ni_vp); 1937 return (EEXIST); 1938 } 1939 VATTR_NULL(&vattr); 1940 vattr.va_type = VFIFO; 1941 /* We will read cwdi->cwdi_cmask unlocked. */ 1942 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1943 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1944 if (error == 0) 1945 vput(nd.ni_vp); 1946 return (error); 1947 } 1948 1949 /* 1950 * Make a hard file link. 1951 */ 1952 /* ARGSUSED */ 1953 int 1954 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1955 { 1956 /* { 1957 syscallarg(const char *) path; 1958 syscallarg(const char *) link; 1959 } */ 1960 struct vnode *vp; 1961 struct nameidata nd; 1962 int error; 1963 1964 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1965 SCARG(uap, path)); 1966 if ((error = namei(&nd)) != 0) 1967 return (error); 1968 vp = nd.ni_vp; 1969 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1970 SCARG(uap, link)); 1971 if ((error = namei(&nd)) != 0) 1972 goto out; 1973 if (nd.ni_vp) { 1974 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1975 if (nd.ni_dvp == nd.ni_vp) 1976 vrele(nd.ni_dvp); 1977 else 1978 vput(nd.ni_dvp); 1979 vrele(nd.ni_vp); 1980 error = EEXIST; 1981 goto out; 1982 } 1983 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1984 out: 1985 vrele(vp); 1986 return (error); 1987 } 1988 1989 /* 1990 * Make a symbolic link. 1991 */ 1992 /* ARGSUSED */ 1993 int 1994 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1995 { 1996 /* { 1997 syscallarg(const char *) path; 1998 syscallarg(const char *) link; 1999 } */ 2000 struct proc *p = l->l_proc; 2001 struct vattr vattr; 2002 char *path; 2003 int error; 2004 struct nameidata nd; 2005 2006 path = PNBUF_GET(); 2007 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2008 if (error) 2009 goto out; 2010 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2011 SCARG(uap, link)); 2012 if ((error = namei(&nd)) != 0) 2013 goto out; 2014 if (nd.ni_vp) { 2015 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2016 if (nd.ni_dvp == nd.ni_vp) 2017 vrele(nd.ni_dvp); 2018 else 2019 vput(nd.ni_dvp); 2020 vrele(nd.ni_vp); 2021 error = EEXIST; 2022 goto out; 2023 } 2024 VATTR_NULL(&vattr); 2025 vattr.va_type = VLNK; 2026 /* We will read cwdi->cwdi_cmask unlocked. */ 2027 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2028 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2029 if (error == 0) 2030 vput(nd.ni_vp); 2031 out: 2032 PNBUF_PUT(path); 2033 return (error); 2034 } 2035 2036 /* 2037 * Delete a whiteout from the filesystem. 2038 */ 2039 /* ARGSUSED */ 2040 int 2041 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2042 { 2043 /* { 2044 syscallarg(const char *) path; 2045 } */ 2046 int error; 2047 struct nameidata nd; 2048 2049 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2050 UIO_USERSPACE, SCARG(uap, path)); 2051 error = namei(&nd); 2052 if (error) 2053 return (error); 2054 2055 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2056 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2057 if (nd.ni_dvp == nd.ni_vp) 2058 vrele(nd.ni_dvp); 2059 else 2060 vput(nd.ni_dvp); 2061 if (nd.ni_vp) 2062 vrele(nd.ni_vp); 2063 return (EEXIST); 2064 } 2065 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2066 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2067 vput(nd.ni_dvp); 2068 return (error); 2069 } 2070 2071 /* 2072 * Delete a name from the filesystem. 2073 */ 2074 /* ARGSUSED */ 2075 int 2076 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2077 { 2078 /* { 2079 syscallarg(const char *) path; 2080 } */ 2081 2082 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2083 } 2084 2085 int 2086 do_sys_unlink(const char *arg, enum uio_seg seg) 2087 { 2088 struct vnode *vp; 2089 int error; 2090 struct nameidata nd; 2091 kauth_cred_t cred; 2092 char *path; 2093 const char *cpath; 2094 2095 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2096 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2097 2098 if ((error = namei(&nd)) != 0) 2099 goto out; 2100 vp = nd.ni_vp; 2101 2102 /* 2103 * The root of a mounted filesystem cannot be deleted. 2104 */ 2105 if (vp->v_vflag & VV_ROOT) { 2106 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2107 if (nd.ni_dvp == vp) 2108 vrele(nd.ni_dvp); 2109 else 2110 vput(nd.ni_dvp); 2111 vput(vp); 2112 error = EBUSY; 2113 goto out; 2114 } 2115 2116 #if NVERIEXEC > 0 2117 /* Handle remove requests for veriexec entries. */ 2118 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2119 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2120 if (nd.ni_dvp == vp) 2121 vrele(nd.ni_dvp); 2122 else 2123 vput(nd.ni_dvp); 2124 vput(vp); 2125 goto out; 2126 } 2127 #endif /* NVERIEXEC > 0 */ 2128 2129 cred = kauth_cred_get(); 2130 #ifdef FILEASSOC 2131 (void)fileassoc_file_delete(vp); 2132 #endif /* FILEASSOC */ 2133 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2134 out: 2135 VERIEXEC_PATH_PUT(path); 2136 return (error); 2137 } 2138 2139 /* 2140 * Reposition read/write file offset. 2141 */ 2142 int 2143 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2144 { 2145 /* { 2146 syscallarg(int) fd; 2147 syscallarg(int) pad; 2148 syscallarg(off_t) offset; 2149 syscallarg(int) whence; 2150 } */ 2151 kauth_cred_t cred = l->l_cred; 2152 file_t *fp; 2153 struct vnode *vp; 2154 struct vattr vattr; 2155 off_t newoff; 2156 int error, fd; 2157 2158 fd = SCARG(uap, fd); 2159 2160 if ((fp = fd_getfile(fd)) == NULL) 2161 return (EBADF); 2162 2163 vp = fp->f_data; 2164 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2165 error = ESPIPE; 2166 goto out; 2167 } 2168 2169 switch (SCARG(uap, whence)) { 2170 case SEEK_CUR: 2171 newoff = fp->f_offset + SCARG(uap, offset); 2172 break; 2173 case SEEK_END: 2174 error = VOP_GETATTR(vp, &vattr, cred); 2175 if (error) { 2176 goto out; 2177 } 2178 newoff = SCARG(uap, offset) + vattr.va_size; 2179 break; 2180 case SEEK_SET: 2181 newoff = SCARG(uap, offset); 2182 break; 2183 default: 2184 error = EINVAL; 2185 goto out; 2186 } 2187 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2188 *(off_t *)retval = fp->f_offset = newoff; 2189 } 2190 out: 2191 fd_putfile(fd); 2192 return (error); 2193 } 2194 2195 /* 2196 * Positional read system call. 2197 */ 2198 int 2199 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2200 { 2201 /* { 2202 syscallarg(int) fd; 2203 syscallarg(void *) buf; 2204 syscallarg(size_t) nbyte; 2205 syscallarg(off_t) offset; 2206 } */ 2207 file_t *fp; 2208 struct vnode *vp; 2209 off_t offset; 2210 int error, fd = SCARG(uap, fd); 2211 2212 if ((fp = fd_getfile(fd)) == NULL) 2213 return (EBADF); 2214 2215 if ((fp->f_flag & FREAD) == 0) { 2216 fd_putfile(fd); 2217 return (EBADF); 2218 } 2219 2220 vp = fp->f_data; 2221 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2222 error = ESPIPE; 2223 goto out; 2224 } 2225 2226 offset = SCARG(uap, offset); 2227 2228 /* 2229 * XXX This works because no file systems actually 2230 * XXX take any action on the seek operation. 2231 */ 2232 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2233 goto out; 2234 2235 /* dofileread() will unuse the descriptor for us */ 2236 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2237 &offset, 0, retval)); 2238 2239 out: 2240 fd_putfile(fd); 2241 return (error); 2242 } 2243 2244 /* 2245 * Positional scatter read system call. 2246 */ 2247 int 2248 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2249 { 2250 /* { 2251 syscallarg(int) fd; 2252 syscallarg(const struct iovec *) iovp; 2253 syscallarg(int) iovcnt; 2254 syscallarg(off_t) offset; 2255 } */ 2256 off_t offset = SCARG(uap, offset); 2257 2258 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2259 SCARG(uap, iovcnt), &offset, 0, retval); 2260 } 2261 2262 /* 2263 * Positional write system call. 2264 */ 2265 int 2266 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2267 { 2268 /* { 2269 syscallarg(int) fd; 2270 syscallarg(const void *) buf; 2271 syscallarg(size_t) nbyte; 2272 syscallarg(off_t) offset; 2273 } */ 2274 file_t *fp; 2275 struct vnode *vp; 2276 off_t offset; 2277 int error, fd = SCARG(uap, fd); 2278 2279 if ((fp = fd_getfile(fd)) == NULL) 2280 return (EBADF); 2281 2282 if ((fp->f_flag & FWRITE) == 0) { 2283 fd_putfile(fd); 2284 return (EBADF); 2285 } 2286 2287 vp = fp->f_data; 2288 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2289 error = ESPIPE; 2290 goto out; 2291 } 2292 2293 offset = SCARG(uap, offset); 2294 2295 /* 2296 * XXX This works because no file systems actually 2297 * XXX take any action on the seek operation. 2298 */ 2299 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2300 goto out; 2301 2302 /* dofilewrite() will unuse the descriptor for us */ 2303 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2304 &offset, 0, retval)); 2305 2306 out: 2307 fd_putfile(fd); 2308 return (error); 2309 } 2310 2311 /* 2312 * Positional gather write system call. 2313 */ 2314 int 2315 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2316 { 2317 /* { 2318 syscallarg(int) fd; 2319 syscallarg(const struct iovec *) iovp; 2320 syscallarg(int) iovcnt; 2321 syscallarg(off_t) offset; 2322 } */ 2323 off_t offset = SCARG(uap, offset); 2324 2325 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2326 SCARG(uap, iovcnt), &offset, 0, retval); 2327 } 2328 2329 /* 2330 * Check access permissions. 2331 */ 2332 int 2333 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2334 { 2335 /* { 2336 syscallarg(const char *) path; 2337 syscallarg(int) flags; 2338 } */ 2339 kauth_cred_t cred; 2340 struct vnode *vp; 2341 int error, flags; 2342 struct nameidata nd; 2343 2344 cred = kauth_cred_dup(l->l_cred); 2345 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2346 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2347 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2348 SCARG(uap, path)); 2349 /* Override default credentials */ 2350 nd.ni_cnd.cn_cred = cred; 2351 if ((error = namei(&nd)) != 0) 2352 goto out; 2353 vp = nd.ni_vp; 2354 2355 /* Flags == 0 means only check for existence. */ 2356 if (SCARG(uap, flags)) { 2357 flags = 0; 2358 if (SCARG(uap, flags) & R_OK) 2359 flags |= VREAD; 2360 if (SCARG(uap, flags) & W_OK) 2361 flags |= VWRITE; 2362 if (SCARG(uap, flags) & X_OK) 2363 flags |= VEXEC; 2364 2365 error = VOP_ACCESS(vp, flags, cred); 2366 if (!error && (flags & VWRITE)) 2367 error = vn_writechk(vp); 2368 } 2369 vput(vp); 2370 out: 2371 kauth_cred_free(cred); 2372 return (error); 2373 } 2374 2375 /* 2376 * Common code for all sys_stat functions, including compat versions. 2377 */ 2378 int 2379 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb) 2380 { 2381 int error; 2382 struct nameidata nd; 2383 2384 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2385 UIO_USERSPACE, path); 2386 error = namei(&nd); 2387 if (error != 0) 2388 return error; 2389 error = vn_stat(nd.ni_vp, sb); 2390 vput(nd.ni_vp); 2391 return error; 2392 } 2393 2394 /* 2395 * Get file status; this version follows links. 2396 */ 2397 /* ARGSUSED */ 2398 int 2399 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2400 { 2401 /* { 2402 syscallarg(const char *) path; 2403 syscallarg(struct stat *) ub; 2404 } */ 2405 struct stat sb; 2406 int error; 2407 2408 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2409 if (error) 2410 return error; 2411 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2412 } 2413 2414 /* 2415 * Get file status; this version does not follow links. 2416 */ 2417 /* ARGSUSED */ 2418 int 2419 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2420 { 2421 /* { 2422 syscallarg(const char *) path; 2423 syscallarg(struct stat *) ub; 2424 } */ 2425 struct stat sb; 2426 int error; 2427 2428 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2429 if (error) 2430 return error; 2431 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2432 } 2433 2434 /* 2435 * Get configurable pathname variables. 2436 */ 2437 /* ARGSUSED */ 2438 int 2439 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2440 { 2441 /* { 2442 syscallarg(const char *) path; 2443 syscallarg(int) name; 2444 } */ 2445 int error; 2446 struct nameidata nd; 2447 2448 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2449 SCARG(uap, path)); 2450 if ((error = namei(&nd)) != 0) 2451 return (error); 2452 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2453 vput(nd.ni_vp); 2454 return (error); 2455 } 2456 2457 /* 2458 * Return target name of a symbolic link. 2459 */ 2460 /* ARGSUSED */ 2461 int 2462 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2463 { 2464 /* { 2465 syscallarg(const char *) path; 2466 syscallarg(char *) buf; 2467 syscallarg(size_t) count; 2468 } */ 2469 struct vnode *vp; 2470 struct iovec aiov; 2471 struct uio auio; 2472 int error; 2473 struct nameidata nd; 2474 2475 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2476 SCARG(uap, path)); 2477 if ((error = namei(&nd)) != 0) 2478 return (error); 2479 vp = nd.ni_vp; 2480 if (vp->v_type != VLNK) 2481 error = EINVAL; 2482 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2483 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2484 aiov.iov_base = SCARG(uap, buf); 2485 aiov.iov_len = SCARG(uap, count); 2486 auio.uio_iov = &aiov; 2487 auio.uio_iovcnt = 1; 2488 auio.uio_offset = 0; 2489 auio.uio_rw = UIO_READ; 2490 KASSERT(l == curlwp); 2491 auio.uio_vmspace = l->l_proc->p_vmspace; 2492 auio.uio_resid = SCARG(uap, count); 2493 error = VOP_READLINK(vp, &auio, l->l_cred); 2494 } 2495 vput(vp); 2496 *retval = SCARG(uap, count) - auio.uio_resid; 2497 return (error); 2498 } 2499 2500 /* 2501 * Change flags of a file given a path name. 2502 */ 2503 /* ARGSUSED */ 2504 int 2505 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2506 { 2507 /* { 2508 syscallarg(const char *) path; 2509 syscallarg(u_long) flags; 2510 } */ 2511 struct vnode *vp; 2512 int error; 2513 struct nameidata nd; 2514 2515 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2516 SCARG(uap, path)); 2517 if ((error = namei(&nd)) != 0) 2518 return (error); 2519 vp = nd.ni_vp; 2520 error = change_flags(vp, SCARG(uap, flags), l); 2521 vput(vp); 2522 return (error); 2523 } 2524 2525 /* 2526 * Change flags of a file given a file descriptor. 2527 */ 2528 /* ARGSUSED */ 2529 int 2530 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2531 { 2532 /* { 2533 syscallarg(int) fd; 2534 syscallarg(u_long) flags; 2535 } */ 2536 struct vnode *vp; 2537 file_t *fp; 2538 int error; 2539 2540 /* fd_getvnode() will use the descriptor for us */ 2541 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2542 return (error); 2543 vp = fp->f_data; 2544 error = change_flags(vp, SCARG(uap, flags), l); 2545 VOP_UNLOCK(vp, 0); 2546 fd_putfile(SCARG(uap, fd)); 2547 return (error); 2548 } 2549 2550 /* 2551 * Change flags of a file given a path name; this version does 2552 * not follow links. 2553 */ 2554 int 2555 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2556 { 2557 /* { 2558 syscallarg(const char *) path; 2559 syscallarg(u_long) flags; 2560 } */ 2561 struct vnode *vp; 2562 int error; 2563 struct nameidata nd; 2564 2565 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2566 SCARG(uap, path)); 2567 if ((error = namei(&nd)) != 0) 2568 return (error); 2569 vp = nd.ni_vp; 2570 error = change_flags(vp, SCARG(uap, flags), l); 2571 vput(vp); 2572 return (error); 2573 } 2574 2575 /* 2576 * Common routine to change flags of a file. 2577 */ 2578 int 2579 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2580 { 2581 struct vattr vattr; 2582 int error; 2583 2584 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2585 /* 2586 * Non-superusers cannot change the flags on devices, even if they 2587 * own them. 2588 */ 2589 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2590 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2591 goto out; 2592 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2593 error = EINVAL; 2594 goto out; 2595 } 2596 } 2597 VATTR_NULL(&vattr); 2598 vattr.va_flags = flags; 2599 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2600 out: 2601 return (error); 2602 } 2603 2604 /* 2605 * Change mode of a file given path name; this version follows links. 2606 */ 2607 /* ARGSUSED */ 2608 int 2609 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2610 { 2611 /* { 2612 syscallarg(const char *) path; 2613 syscallarg(int) mode; 2614 } */ 2615 int error; 2616 struct nameidata nd; 2617 2618 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2619 SCARG(uap, path)); 2620 if ((error = namei(&nd)) != 0) 2621 return (error); 2622 2623 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2624 2625 vrele(nd.ni_vp); 2626 return (error); 2627 } 2628 2629 /* 2630 * Change mode of a file given a file descriptor. 2631 */ 2632 /* ARGSUSED */ 2633 int 2634 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2635 { 2636 /* { 2637 syscallarg(int) fd; 2638 syscallarg(int) mode; 2639 } */ 2640 file_t *fp; 2641 int error; 2642 2643 /* fd_getvnode() will use the descriptor for us */ 2644 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2645 return (error); 2646 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2647 fd_putfile(SCARG(uap, fd)); 2648 return (error); 2649 } 2650 2651 /* 2652 * Change mode of a file given path name; this version does not follow links. 2653 */ 2654 /* ARGSUSED */ 2655 int 2656 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2657 { 2658 /* { 2659 syscallarg(const char *) path; 2660 syscallarg(int) mode; 2661 } */ 2662 int error; 2663 struct nameidata nd; 2664 2665 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2666 SCARG(uap, path)); 2667 if ((error = namei(&nd)) != 0) 2668 return (error); 2669 2670 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2671 2672 vrele(nd.ni_vp); 2673 return (error); 2674 } 2675 2676 /* 2677 * Common routine to set mode given a vnode. 2678 */ 2679 static int 2680 change_mode(struct vnode *vp, int mode, struct lwp *l) 2681 { 2682 struct vattr vattr; 2683 int error; 2684 2685 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2686 VATTR_NULL(&vattr); 2687 vattr.va_mode = mode & ALLPERMS; 2688 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2689 VOP_UNLOCK(vp, 0); 2690 return (error); 2691 } 2692 2693 /* 2694 * Set ownership given a path name; this version follows links. 2695 */ 2696 /* ARGSUSED */ 2697 int 2698 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2699 { 2700 /* { 2701 syscallarg(const char *) path; 2702 syscallarg(uid_t) uid; 2703 syscallarg(gid_t) gid; 2704 } */ 2705 int error; 2706 struct nameidata nd; 2707 2708 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2709 SCARG(uap, path)); 2710 if ((error = namei(&nd)) != 0) 2711 return (error); 2712 2713 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2714 2715 vrele(nd.ni_vp); 2716 return (error); 2717 } 2718 2719 /* 2720 * Set ownership given a path name; this version follows links. 2721 * Provides POSIX semantics. 2722 */ 2723 /* ARGSUSED */ 2724 int 2725 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2726 { 2727 /* { 2728 syscallarg(const char *) path; 2729 syscallarg(uid_t) uid; 2730 syscallarg(gid_t) gid; 2731 } */ 2732 int error; 2733 struct nameidata nd; 2734 2735 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2736 SCARG(uap, path)); 2737 if ((error = namei(&nd)) != 0) 2738 return (error); 2739 2740 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2741 2742 vrele(nd.ni_vp); 2743 return (error); 2744 } 2745 2746 /* 2747 * Set ownership given a file descriptor. 2748 */ 2749 /* ARGSUSED */ 2750 int 2751 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2752 { 2753 /* { 2754 syscallarg(int) fd; 2755 syscallarg(uid_t) uid; 2756 syscallarg(gid_t) gid; 2757 } */ 2758 int error; 2759 file_t *fp; 2760 2761 /* fd_getvnode() will use the descriptor for us */ 2762 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2763 return (error); 2764 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2765 l, 0); 2766 fd_putfile(SCARG(uap, fd)); 2767 return (error); 2768 } 2769 2770 /* 2771 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2772 */ 2773 /* ARGSUSED */ 2774 int 2775 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2776 { 2777 /* { 2778 syscallarg(int) fd; 2779 syscallarg(uid_t) uid; 2780 syscallarg(gid_t) gid; 2781 } */ 2782 int error; 2783 file_t *fp; 2784 2785 /* fd_getvnode() will use the descriptor for us */ 2786 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2787 return (error); 2788 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2789 l, 1); 2790 fd_putfile(SCARG(uap, fd)); 2791 return (error); 2792 } 2793 2794 /* 2795 * Set ownership given a path name; this version does not follow links. 2796 */ 2797 /* ARGSUSED */ 2798 int 2799 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2800 { 2801 /* { 2802 syscallarg(const char *) path; 2803 syscallarg(uid_t) uid; 2804 syscallarg(gid_t) gid; 2805 } */ 2806 int error; 2807 struct nameidata nd; 2808 2809 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2810 SCARG(uap, path)); 2811 if ((error = namei(&nd)) != 0) 2812 return (error); 2813 2814 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2815 2816 vrele(nd.ni_vp); 2817 return (error); 2818 } 2819 2820 /* 2821 * Set ownership given a path name; this version does not follow links. 2822 * Provides POSIX/XPG semantics. 2823 */ 2824 /* ARGSUSED */ 2825 int 2826 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2827 { 2828 /* { 2829 syscallarg(const char *) path; 2830 syscallarg(uid_t) uid; 2831 syscallarg(gid_t) gid; 2832 } */ 2833 int error; 2834 struct nameidata nd; 2835 2836 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2837 SCARG(uap, path)); 2838 if ((error = namei(&nd)) != 0) 2839 return (error); 2840 2841 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2842 2843 vrele(nd.ni_vp); 2844 return (error); 2845 } 2846 2847 /* 2848 * Common routine to set ownership given a vnode. 2849 */ 2850 static int 2851 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2852 int posix_semantics) 2853 { 2854 struct vattr vattr; 2855 mode_t newmode; 2856 int error; 2857 2858 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2859 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2860 goto out; 2861 2862 #define CHANGED(x) ((int)(x) != -1) 2863 newmode = vattr.va_mode; 2864 if (posix_semantics) { 2865 /* 2866 * POSIX/XPG semantics: if the caller is not the super-user, 2867 * clear set-user-id and set-group-id bits. Both POSIX and 2868 * the XPG consider the behaviour for calls by the super-user 2869 * implementation-defined; we leave the set-user-id and set- 2870 * group-id settings intact in that case. 2871 */ 2872 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2873 NULL) != 0) 2874 newmode &= ~(S_ISUID | S_ISGID); 2875 } else { 2876 /* 2877 * NetBSD semantics: when changing owner and/or group, 2878 * clear the respective bit(s). 2879 */ 2880 if (CHANGED(uid)) 2881 newmode &= ~S_ISUID; 2882 if (CHANGED(gid)) 2883 newmode &= ~S_ISGID; 2884 } 2885 /* Update va_mode iff altered. */ 2886 if (vattr.va_mode == newmode) 2887 newmode = VNOVAL; 2888 2889 VATTR_NULL(&vattr); 2890 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2891 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2892 vattr.va_mode = newmode; 2893 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2894 #undef CHANGED 2895 2896 out: 2897 VOP_UNLOCK(vp, 0); 2898 return (error); 2899 } 2900 2901 /* 2902 * Set the access and modification times given a path name; this 2903 * version follows links. 2904 */ 2905 /* ARGSUSED */ 2906 int 2907 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2908 { 2909 /* { 2910 syscallarg(const char *) path; 2911 syscallarg(const struct timeval *) tptr; 2912 } */ 2913 2914 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2915 SCARG(uap, tptr), UIO_USERSPACE); 2916 } 2917 2918 /* 2919 * Set the access and modification times given a file descriptor. 2920 */ 2921 /* ARGSUSED */ 2922 int 2923 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2924 { 2925 /* { 2926 syscallarg(int) fd; 2927 syscallarg(const struct timeval *) tptr; 2928 } */ 2929 int error; 2930 file_t *fp; 2931 2932 /* fd_getvnode() will use the descriptor for us */ 2933 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2934 return (error); 2935 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2936 UIO_USERSPACE); 2937 fd_putfile(SCARG(uap, fd)); 2938 return (error); 2939 } 2940 2941 /* 2942 * Set the access and modification times given a path name; this 2943 * version does not follow links. 2944 */ 2945 int 2946 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2947 { 2948 /* { 2949 syscallarg(const char *) path; 2950 syscallarg(const struct timeval *) tptr; 2951 } */ 2952 2953 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2954 SCARG(uap, tptr), UIO_USERSPACE); 2955 } 2956 2957 /* 2958 * Common routine to set access and modification times given a vnode. 2959 */ 2960 int 2961 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2962 const struct timeval *tptr, enum uio_seg seg) 2963 { 2964 struct vattr vattr; 2965 struct nameidata nd; 2966 int error; 2967 2968 VATTR_NULL(&vattr); 2969 if (tptr == NULL) { 2970 nanotime(&vattr.va_atime); 2971 vattr.va_mtime = vattr.va_atime; 2972 vattr.va_vaflags |= VA_UTIMES_NULL; 2973 } else { 2974 struct timeval tv[2]; 2975 2976 if (seg != UIO_SYSSPACE) { 2977 error = copyin(tptr, &tv, sizeof (tv)); 2978 if (error != 0) 2979 return error; 2980 tptr = tv; 2981 } 2982 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 2983 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 2984 } 2985 2986 if (vp == NULL) { 2987 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 2988 if ((error = namei(&nd)) != 0) 2989 return (error); 2990 vp = nd.ni_vp; 2991 } else 2992 nd.ni_vp = NULL; 2993 2994 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2995 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2996 VOP_UNLOCK(vp, 0); 2997 2998 if (nd.ni_vp != NULL) 2999 vrele(nd.ni_vp); 3000 3001 return (error); 3002 } 3003 3004 /* 3005 * Truncate a file given its path name. 3006 */ 3007 /* ARGSUSED */ 3008 int 3009 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3010 { 3011 /* { 3012 syscallarg(const char *) path; 3013 syscallarg(int) pad; 3014 syscallarg(off_t) length; 3015 } */ 3016 struct vnode *vp; 3017 struct vattr vattr; 3018 int error; 3019 struct nameidata nd; 3020 3021 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3022 SCARG(uap, path)); 3023 if ((error = namei(&nd)) != 0) 3024 return (error); 3025 vp = nd.ni_vp; 3026 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3027 if (vp->v_type == VDIR) 3028 error = EISDIR; 3029 else if ((error = vn_writechk(vp)) == 0 && 3030 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3031 VATTR_NULL(&vattr); 3032 vattr.va_size = SCARG(uap, length); 3033 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3034 } 3035 vput(vp); 3036 return (error); 3037 } 3038 3039 /* 3040 * Truncate a file given a file descriptor. 3041 */ 3042 /* ARGSUSED */ 3043 int 3044 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3045 { 3046 /* { 3047 syscallarg(int) fd; 3048 syscallarg(int) pad; 3049 syscallarg(off_t) length; 3050 } */ 3051 struct vattr vattr; 3052 struct vnode *vp; 3053 file_t *fp; 3054 int error; 3055 3056 /* fd_getvnode() will use the descriptor for us */ 3057 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3058 return (error); 3059 if ((fp->f_flag & FWRITE) == 0) { 3060 error = EINVAL; 3061 goto out; 3062 } 3063 vp = fp->f_data; 3064 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3065 if (vp->v_type == VDIR) 3066 error = EISDIR; 3067 else if ((error = vn_writechk(vp)) == 0) { 3068 VATTR_NULL(&vattr); 3069 vattr.va_size = SCARG(uap, length); 3070 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3071 } 3072 VOP_UNLOCK(vp, 0); 3073 out: 3074 fd_putfile(SCARG(uap, fd)); 3075 return (error); 3076 } 3077 3078 /* 3079 * Sync an open file. 3080 */ 3081 /* ARGSUSED */ 3082 int 3083 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3084 { 3085 /* { 3086 syscallarg(int) fd; 3087 } */ 3088 struct vnode *vp; 3089 file_t *fp; 3090 int error; 3091 3092 /* fd_getvnode() will use the descriptor for us */ 3093 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3094 return (error); 3095 vp = fp->f_data; 3096 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3097 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3098 if (error == 0 && bioopsp != NULL && 3099 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3100 (*bioopsp->io_fsync)(vp, 0); 3101 VOP_UNLOCK(vp, 0); 3102 fd_putfile(SCARG(uap, fd)); 3103 return (error); 3104 } 3105 3106 /* 3107 * Sync a range of file data. API modeled after that found in AIX. 3108 * 3109 * FDATASYNC indicates that we need only save enough metadata to be able 3110 * to re-read the written data. Note we duplicate AIX's requirement that 3111 * the file be open for writing. 3112 */ 3113 /* ARGSUSED */ 3114 int 3115 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3116 { 3117 /* { 3118 syscallarg(int) fd; 3119 syscallarg(int) flags; 3120 syscallarg(off_t) start; 3121 syscallarg(off_t) length; 3122 } */ 3123 struct vnode *vp; 3124 file_t *fp; 3125 int flags, nflags; 3126 off_t s, e, len; 3127 int error; 3128 3129 /* fd_getvnode() will use the descriptor for us */ 3130 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3131 return (error); 3132 3133 if ((fp->f_flag & FWRITE) == 0) { 3134 error = EBADF; 3135 goto out; 3136 } 3137 3138 flags = SCARG(uap, flags); 3139 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3140 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3141 error = EINVAL; 3142 goto out; 3143 } 3144 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3145 if (flags & FDATASYNC) 3146 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3147 else 3148 nflags = FSYNC_WAIT; 3149 if (flags & FDISKSYNC) 3150 nflags |= FSYNC_CACHE; 3151 3152 len = SCARG(uap, length); 3153 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3154 if (len) { 3155 s = SCARG(uap, start); 3156 e = s + len; 3157 if (e < s) { 3158 error = EINVAL; 3159 goto out; 3160 } 3161 } else { 3162 e = 0; 3163 s = 0; 3164 } 3165 3166 vp = fp->f_data; 3167 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3168 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3169 3170 if (error == 0 && bioopsp != NULL && 3171 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3172 (*bioopsp->io_fsync)(vp, nflags); 3173 3174 VOP_UNLOCK(vp, 0); 3175 out: 3176 fd_putfile(SCARG(uap, fd)); 3177 return (error); 3178 } 3179 3180 /* 3181 * Sync the data of an open file. 3182 */ 3183 /* ARGSUSED */ 3184 int 3185 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3186 { 3187 /* { 3188 syscallarg(int) fd; 3189 } */ 3190 struct vnode *vp; 3191 file_t *fp; 3192 int error; 3193 3194 /* fd_getvnode() will use the descriptor for us */ 3195 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3196 return (error); 3197 if ((fp->f_flag & FWRITE) == 0) { 3198 fd_putfile(SCARG(uap, fd)); 3199 return (EBADF); 3200 } 3201 vp = fp->f_data; 3202 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3203 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3204 VOP_UNLOCK(vp, 0); 3205 fd_putfile(SCARG(uap, fd)); 3206 return (error); 3207 } 3208 3209 /* 3210 * Rename files, (standard) BSD semantics frontend. 3211 */ 3212 /* ARGSUSED */ 3213 int 3214 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3215 { 3216 /* { 3217 syscallarg(const char *) from; 3218 syscallarg(const char *) to; 3219 } */ 3220 3221 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3222 } 3223 3224 /* 3225 * Rename files, POSIX semantics frontend. 3226 */ 3227 /* ARGSUSED */ 3228 int 3229 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3230 { 3231 /* { 3232 syscallarg(const char *) from; 3233 syscallarg(const char *) to; 3234 } */ 3235 3236 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3237 } 3238 3239 /* 3240 * Rename files. Source and destination must either both be directories, 3241 * or both not be directories. If target is a directory, it must be empty. 3242 * If `from' and `to' refer to the same object, the value of the `retain' 3243 * argument is used to determine whether `from' will be 3244 * 3245 * (retain == 0) deleted unless `from' and `to' refer to the same 3246 * object in the file system's name space (BSD). 3247 * (retain == 1) always retained (POSIX). 3248 */ 3249 int 3250 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3251 { 3252 struct vnode *tvp, *fvp, *tdvp; 3253 struct nameidata fromnd, tond; 3254 struct mount *fs; 3255 struct lwp *l = curlwp; 3256 struct proc *p; 3257 uint32_t saveflag; 3258 int error; 3259 3260 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3261 seg, from); 3262 if ((error = namei(&fromnd)) != 0) 3263 return (error); 3264 if (fromnd.ni_dvp != fromnd.ni_vp) 3265 VOP_UNLOCK(fromnd.ni_dvp, 0); 3266 fvp = fromnd.ni_vp; 3267 3268 fs = fvp->v_mount; 3269 error = VFS_RENAMELOCK_ENTER(fs); 3270 if (error) { 3271 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3272 vrele(fromnd.ni_dvp); 3273 vrele(fvp); 3274 goto out1; 3275 } 3276 3277 /* 3278 * close, partially, yet another race - ideally we should only 3279 * go as far as getting fromnd.ni_dvp before getting the per-fs 3280 * lock, and then continue to get fromnd.ni_vp, but we can't do 3281 * that with namei as it stands. 3282 * 3283 * This still won't prevent rmdir from nuking fromnd.ni_vp 3284 * under us. The real fix is to get the locks in the right 3285 * order and do the lookups in the right places, but that's a 3286 * major rototill. 3287 * 3288 * Preserve the SAVESTART in cn_flags, because who knows what 3289 * might happen if we don't. 3290 * 3291 * Note: this logic (as well as this whole function) is cloned 3292 * in nfs_serv.c. Proceed accordingly. 3293 */ 3294 vrele(fvp); 3295 if ((fromnd.ni_cnd.cn_namelen == 1 && 3296 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3297 (fromnd.ni_cnd.cn_namelen == 2 && 3298 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3299 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3300 error = EINVAL; 3301 VFS_RENAMELOCK_EXIT(fs); 3302 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3303 vrele(fromnd.ni_dvp); 3304 goto out1; 3305 } 3306 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3307 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3308 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3309 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3310 fromnd.ni_cnd.cn_flags |= saveflag; 3311 if (error) { 3312 VOP_UNLOCK(fromnd.ni_dvp, 0); 3313 VFS_RENAMELOCK_EXIT(fs); 3314 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3315 vrele(fromnd.ni_dvp); 3316 goto out1; 3317 } 3318 VOP_UNLOCK(fromnd.ni_vp, 0); 3319 if (fromnd.ni_dvp != fromnd.ni_vp) 3320 VOP_UNLOCK(fromnd.ni_dvp, 0); 3321 fvp = fromnd.ni_vp; 3322 3323 NDINIT(&tond, RENAME, 3324 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3325 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3326 seg, to); 3327 if ((error = namei(&tond)) != 0) { 3328 VFS_RENAMELOCK_EXIT(fs); 3329 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3330 vrele(fromnd.ni_dvp); 3331 vrele(fvp); 3332 goto out1; 3333 } 3334 tdvp = tond.ni_dvp; 3335 tvp = tond.ni_vp; 3336 3337 if (tvp != NULL) { 3338 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3339 error = ENOTDIR; 3340 goto out; 3341 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3342 error = EISDIR; 3343 goto out; 3344 } 3345 } 3346 3347 if (fvp == tdvp) 3348 error = EINVAL; 3349 3350 /* 3351 * Source and destination refer to the same object. 3352 */ 3353 if (fvp == tvp) { 3354 if (retain) 3355 error = -1; 3356 else if (fromnd.ni_dvp == tdvp && 3357 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3358 !memcmp(fromnd.ni_cnd.cn_nameptr, 3359 tond.ni_cnd.cn_nameptr, 3360 fromnd.ni_cnd.cn_namelen)) 3361 error = -1; 3362 } 3363 3364 #if NVERIEXEC > 0 3365 if (!error) { 3366 char *f1, *f2; 3367 3368 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3369 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3370 3371 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3372 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3373 3374 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3375 3376 free(f1, M_TEMP); 3377 free(f2, M_TEMP); 3378 } 3379 #endif /* NVERIEXEC > 0 */ 3380 3381 out: 3382 p = l->l_proc; 3383 if (!error) { 3384 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3385 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3386 VFS_RENAMELOCK_EXIT(fs); 3387 } else { 3388 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3389 if (tdvp == tvp) 3390 vrele(tdvp); 3391 else 3392 vput(tdvp); 3393 if (tvp) 3394 vput(tvp); 3395 VFS_RENAMELOCK_EXIT(fs); 3396 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3397 vrele(fromnd.ni_dvp); 3398 vrele(fvp); 3399 } 3400 vrele(tond.ni_startdir); 3401 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3402 out1: 3403 if (fromnd.ni_startdir) 3404 vrele(fromnd.ni_startdir); 3405 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3406 return (error == -1 ? 0 : error); 3407 } 3408 3409 /* 3410 * Make a directory file. 3411 */ 3412 /* ARGSUSED */ 3413 int 3414 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3415 { 3416 /* { 3417 syscallarg(const char *) path; 3418 syscallarg(int) mode; 3419 } */ 3420 struct proc *p = l->l_proc; 3421 struct vnode *vp; 3422 struct vattr vattr; 3423 int error; 3424 struct nameidata nd; 3425 3426 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3427 SCARG(uap, path)); 3428 if ((error = namei(&nd)) != 0) 3429 return (error); 3430 vp = nd.ni_vp; 3431 if (vp != NULL) { 3432 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3433 if (nd.ni_dvp == vp) 3434 vrele(nd.ni_dvp); 3435 else 3436 vput(nd.ni_dvp); 3437 vrele(vp); 3438 return (EEXIST); 3439 } 3440 VATTR_NULL(&vattr); 3441 vattr.va_type = VDIR; 3442 /* We will read cwdi->cwdi_cmask unlocked. */ 3443 vattr.va_mode = 3444 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3445 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3446 if (!error) 3447 vput(nd.ni_vp); 3448 return (error); 3449 } 3450 3451 /* 3452 * Remove a directory file. 3453 */ 3454 /* ARGSUSED */ 3455 int 3456 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3457 { 3458 /* { 3459 syscallarg(const char *) path; 3460 } */ 3461 struct vnode *vp; 3462 int error; 3463 struct nameidata nd; 3464 3465 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3466 SCARG(uap, path)); 3467 if ((error = namei(&nd)) != 0) 3468 return (error); 3469 vp = nd.ni_vp; 3470 if (vp->v_type != VDIR) { 3471 error = ENOTDIR; 3472 goto out; 3473 } 3474 /* 3475 * No rmdir "." please. 3476 */ 3477 if (nd.ni_dvp == vp) { 3478 error = EINVAL; 3479 goto out; 3480 } 3481 /* 3482 * The root of a mounted filesystem cannot be deleted. 3483 */ 3484 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3485 error = EBUSY; 3486 goto out; 3487 } 3488 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3489 return (error); 3490 3491 out: 3492 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3493 if (nd.ni_dvp == vp) 3494 vrele(nd.ni_dvp); 3495 else 3496 vput(nd.ni_dvp); 3497 vput(vp); 3498 return (error); 3499 } 3500 3501 /* 3502 * Read a block of directory entries in a file system independent format. 3503 */ 3504 int 3505 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3506 { 3507 /* { 3508 syscallarg(int) fd; 3509 syscallarg(char *) buf; 3510 syscallarg(size_t) count; 3511 } */ 3512 file_t *fp; 3513 int error, done; 3514 3515 /* fd_getvnode() will use the descriptor for us */ 3516 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3517 return (error); 3518 if ((fp->f_flag & FREAD) == 0) { 3519 error = EBADF; 3520 goto out; 3521 } 3522 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3523 SCARG(uap, count), &done, l, 0, 0); 3524 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3525 *retval = done; 3526 out: 3527 fd_putfile(SCARG(uap, fd)); 3528 return (error); 3529 } 3530 3531 /* 3532 * Set the mode mask for creation of filesystem nodes. 3533 */ 3534 int 3535 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3536 { 3537 /* { 3538 syscallarg(mode_t) newmask; 3539 } */ 3540 struct proc *p = l->l_proc; 3541 struct cwdinfo *cwdi; 3542 3543 /* 3544 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3545 * important is that we serialize changes to the mask. The 3546 * rw_exit() will issue a write memory barrier on our behalf, 3547 * and force the changes out to other CPUs (as it must use an 3548 * atomic operation, draining the local CPU's store buffers). 3549 */ 3550 cwdi = p->p_cwdi; 3551 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3552 *retval = cwdi->cwdi_cmask; 3553 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3554 rw_exit(&cwdi->cwdi_lock); 3555 3556 return (0); 3557 } 3558 3559 int 3560 dorevoke(struct vnode *vp, kauth_cred_t cred) 3561 { 3562 struct vattr vattr; 3563 int error; 3564 3565 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3566 return error; 3567 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3568 (error = kauth_authorize_generic(cred, 3569 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3570 VOP_REVOKE(vp, REVOKEALL); 3571 return (error); 3572 } 3573 3574 /* 3575 * Void all references to file by ripping underlying filesystem 3576 * away from vnode. 3577 */ 3578 /* ARGSUSED */ 3579 int 3580 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3581 { 3582 /* { 3583 syscallarg(const char *) path; 3584 } */ 3585 struct vnode *vp; 3586 int error; 3587 struct nameidata nd; 3588 3589 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3590 SCARG(uap, path)); 3591 if ((error = namei(&nd)) != 0) 3592 return (error); 3593 vp = nd.ni_vp; 3594 error = dorevoke(vp, l->l_cred); 3595 vrele(vp); 3596 return (error); 3597 } 3598 3599 /* 3600 * Convert a user file descriptor to a kernel file entry. 3601 */ 3602 int 3603 getvnode(int fd, file_t **fpp) 3604 { 3605 struct vnode *vp; 3606 file_t *fp; 3607 3608 if ((fp = fd_getfile(fd)) == NULL) 3609 return (EBADF); 3610 3611 if (fp->f_type != DTYPE_VNODE) { 3612 fd_putfile(fd); 3613 return (EINVAL); 3614 } 3615 3616 vp = fp->f_data; 3617 if (vp->v_type == VBAD) { 3618 fd_putfile(fd); 3619 return (EBADF); 3620 } 3621 3622 *fpp = fp; 3623 return (0); 3624 } 3625
Indexes created Fri Oct 03 02:09:57 GMT 2025