vfs_syscalls.c revision 1.373
1 /* $NetBSD: vfs_syscalls.c,v 1.373 2008/09/24 10:07:19 ad Exp $ */ 2 3 /*- 4 * Copyright (c) 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 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 1.373 2008/09/24 10:07:19 ad Exp $"); 67 68 #include "opt_compat_netbsd.h" 69 #include "opt_compat_43.h" 70 #include "opt_fileassoc.h" 71 #include "veriexec.h" 72 73 #include <sys/param.h> 74 #include <sys/systm.h> 75 #include <sys/namei.h> 76 #include <sys/filedesc.h> 77 #include <sys/kernel.h> 78 #include <sys/file.h> 79 #include <sys/stat.h> 80 #include <sys/vnode.h> 81 #include <sys/mount.h> 82 #include <sys/proc.h> 83 #include <sys/uio.h> 84 #include <sys/malloc.h> 85 #include <sys/kmem.h> 86 #include <sys/dirent.h> 87 #include <sys/sysctl.h> 88 #include <sys/syscallargs.h> 89 #include <sys/vfs_syscalls.h> 90 #include <sys/ktrace.h> 91 #ifdef FILEASSOC 92 #include <sys/fileassoc.h> 93 #endif /* FILEASSOC */ 94 #include <sys/verified_exec.h> 95 #include <sys/kauth.h> 96 #include <sys/atomic.h> 97 #include <sys/module.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 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 114 115 static int change_dir(struct nameidata *, struct lwp *); 116 static int change_flags(struct vnode *, u_long, struct lwp *); 117 static int change_mode(struct vnode *, int, struct lwp *l); 118 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 119 120 void checkdirs(struct vnode *); 121 122 int dovfsusermount = 0; 123 124 /* 125 * Virtual File System System Calls 126 */ 127 128 /* 129 * Mount a file system. 130 */ 131 132 #if defined(COMPAT_09) || defined(COMPAT_43) 133 /* 134 * This table is used to maintain compatibility with 4.3BSD 135 * and NetBSD 0.9 mount syscalls. Note, the order is important! 136 * 137 * Do not modify this table. It should only contain filesystems 138 * supported by NetBSD 0.9 and 4.3BSD. 139 */ 140 const char * const mountcompatnames[] = { 141 NULL, /* 0 = MOUNT_NONE */ 142 MOUNT_FFS, /* 1 = MOUNT_UFS */ 143 MOUNT_NFS, /* 2 */ 144 MOUNT_MFS, /* 3 */ 145 MOUNT_MSDOS, /* 4 */ 146 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 147 MOUNT_FDESC, /* 6 */ 148 MOUNT_KERNFS, /* 7 */ 149 NULL, /* 8 = MOUNT_DEVFS */ 150 MOUNT_AFS, /* 9 */ 151 }; 152 const int nmountcompatnames = sizeof(mountcompatnames) / 153 sizeof(mountcompatnames[0]); 154 #endif /* COMPAT_09 || COMPAT_43 */ 155 156 static int 157 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 158 void *data, size_t *data_len) 159 { 160 struct mount *mp; 161 int error = 0, saved_flags; 162 163 mp = vp->v_mount; 164 saved_flags = mp->mnt_flag; 165 166 /* We can operate only on VV_ROOT nodes. */ 167 if ((vp->v_vflag & VV_ROOT) == 0) { 168 error = EINVAL; 169 goto out; 170 } 171 172 /* 173 * We only allow the filesystem to be reloaded if it 174 * is currently mounted read-only. Additionally, we 175 * prevent read-write to read-only downgrades. 176 */ 177 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 178 (mp->mnt_flag & MNT_RDONLY) == 0) { 179 error = EOPNOTSUPP; /* Needs translation */ 180 goto out; 181 } 182 183 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 184 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 185 if (error) 186 goto out; 187 188 if (vfs_busy(mp, NULL)) { 189 error = EPERM; 190 goto out; 191 } 192 193 mutex_enter(&mp->mnt_updating); 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 MNT_LOG); 210 mp->mnt_flag |= flags & 211 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 212 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 213 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 214 MNT_LOG | MNT_IGNORE); 215 216 error = VFS_MOUNT(mp, path, data, data_len); 217 218 #if defined(COMPAT_30) && defined(NFSSERVER) 219 if (error && data != NULL) { 220 int error2; 221 222 /* Update failed; let's try and see if it was an 223 * export request. */ 224 error2 = nfs_update_exports_30(mp, path, data, l); 225 226 /* Only update error code if the export request was 227 * understood but some problem occurred while 228 * processing it. */ 229 if (error2 != EJUSTRETURN) 230 error = error2; 231 } 232 #endif 233 if (mp->mnt_iflag & IMNT_WANTRDWR) 234 mp->mnt_flag &= ~MNT_RDONLY; 235 if (error) 236 mp->mnt_flag = saved_flags; 237 mp->mnt_flag &= ~MNT_OP_FLAGS; 238 mp->mnt_iflag &= ~IMNT_WANTRDWR; 239 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 240 if (mp->mnt_syncer == NULL) 241 error = vfs_allocate_syncvnode(mp); 242 } else { 243 if (mp->mnt_syncer != NULL) 244 vfs_deallocate_syncvnode(mp); 245 } 246 mutex_exit(&mp->mnt_updating); 247 vfs_unbusy(mp, false, NULL); 248 249 out: 250 return (error); 251 } 252 253 static int 254 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 255 { 256 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 257 int error; 258 259 /* Copy file-system type from userspace. */ 260 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 261 if (error) { 262 #if defined(COMPAT_09) || defined(COMPAT_43) 263 /* 264 * Historically, filesystem types were identified by numbers. 265 * If we get an integer for the filesystem type instead of a 266 * string, we check to see if it matches one of the historic 267 * filesystem types. 268 */ 269 u_long fsindex = (u_long)fstype; 270 if (fsindex >= nmountcompatnames || 271 mountcompatnames[fsindex] == NULL) 272 return ENODEV; 273 strlcpy(fstypename, mountcompatnames[fsindex], 274 sizeof(fstypename)); 275 #else 276 return error; 277 #endif 278 } 279 280 #ifdef COMPAT_10 281 /* Accept `ufs' as an alias for `ffs'. */ 282 if (strcmp(fstypename, "ufs") == 0) 283 fstypename[0] = 'f'; 284 #endif 285 286 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 287 return 0; 288 289 /* If we can autoload a vfs module, try again */ 290 (void)module_load(fstype, 0, NULL, MODULE_CLASS_VFS, true); 291 292 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 293 return 0; 294 295 return ENODEV; 296 } 297 298 static int 299 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 300 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 301 { 302 struct mount *mp; 303 struct vnode *vp = *vpp; 304 struct vattr va; 305 int error; 306 307 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 308 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 309 if (error) 310 return error; 311 312 /* Can't make a non-dir a mount-point (from here anyway). */ 313 if (vp->v_type != VDIR) 314 return ENOTDIR; 315 316 /* 317 * If the user is not root, ensure that they own the directory 318 * onto which we are attempting to mount. 319 */ 320 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 321 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 322 (error = kauth_authorize_generic(l->l_cred, 323 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 324 return error; 325 } 326 327 if (flags & MNT_EXPORTED) 328 return EINVAL; 329 330 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 331 return error; 332 333 /* 334 * Check if a file-system is not already mounted on this vnode. 335 */ 336 if (vp->v_mountedhere != NULL) 337 return EBUSY; 338 339 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP); 340 if (mp == NULL) 341 return ENOMEM; 342 343 mp->mnt_op = vfsops; 344 mp->mnt_refcnt = 1; 345 346 TAILQ_INIT(&mp->mnt_vnodelist); 347 rw_init(&mp->mnt_unmounting); 348 mutex_init(&mp->mnt_renamelock, MUTEX_DEFAULT, IPL_NONE); 349 mutex_init(&mp->mnt_updating, MUTEX_DEFAULT, IPL_NONE); 350 error = vfs_busy(mp, NULL); 351 KASSERT(error == 0); 352 mutex_enter(&mp->mnt_updating); 353 354 mp->mnt_vnodecovered = vp; 355 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 356 mount_initspecific(mp); 357 358 /* 359 * The underlying file system may refuse the mount for 360 * various reasons. Allow the user to force it to happen. 361 * 362 * Set the mount level flags. 363 */ 364 mp->mnt_flag = flags & 365 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 366 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 367 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 368 MNT_LOG | MNT_IGNORE | MNT_RDONLY); 369 370 error = VFS_MOUNT(mp, path, data, data_len); 371 mp->mnt_flag &= ~MNT_OP_FLAGS; 372 373 /* 374 * Put the new filesystem on the mount list after root. 375 */ 376 cache_purge(vp); 377 if (error != 0) { 378 vp->v_mountedhere = NULL; 379 mutex_exit(&mp->mnt_updating); 380 vfs_unbusy(mp, false, NULL); 381 vfs_destroy(mp); 382 return error; 383 } 384 385 mp->mnt_iflag &= ~IMNT_WANTRDWR; 386 mutex_enter(&mountlist_lock); 387 vp->v_mountedhere = mp; 388 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 389 mutex_exit(&mountlist_lock); 390 vn_restorerecurse(vp, recurse); 391 VOP_UNLOCK(vp, 0); 392 checkdirs(vp); 393 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 394 error = vfs_allocate_syncvnode(mp); 395 /* Hold an additional reference to the mount across VFS_START(). */ 396 mutex_exit(&mp->mnt_updating); 397 vfs_unbusy(mp, true, NULL); 398 (void) VFS_STATVFS(mp, &mp->mnt_stat); 399 error = VFS_START(mp, 0); 400 if (error) 401 vrele(vp); 402 /* Drop reference held for VFS_START(). */ 403 vfs_destroy(mp); 404 *vpp = NULL; 405 return error; 406 } 407 408 static int 409 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 410 void *data, size_t *data_len) 411 { 412 struct mount *mp; 413 int error; 414 415 /* If MNT_GETARGS is specified, it should be the only flag. */ 416 if (flags & ~MNT_GETARGS) 417 return EINVAL; 418 419 mp = vp->v_mount; 420 421 /* XXX: probably some notion of "can see" here if we want isolation. */ 422 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 423 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 424 if (error) 425 return error; 426 427 if ((vp->v_vflag & VV_ROOT) == 0) 428 return EINVAL; 429 430 if (vfs_busy(mp, NULL)) 431 return EPERM; 432 433 mutex_enter(&mp->mnt_updating); 434 mp->mnt_flag &= ~MNT_OP_FLAGS; 435 mp->mnt_flag |= MNT_GETARGS; 436 error = VFS_MOUNT(mp, path, data, data_len); 437 mp->mnt_flag &= ~MNT_OP_FLAGS; 438 mutex_exit(&mp->mnt_updating); 439 440 vfs_unbusy(mp, false, NULL); 441 return (error); 442 } 443 444 #ifdef COMPAT_40 445 /* ARGSUSED */ 446 int 447 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 448 { 449 /* { 450 syscallarg(const char *) type; 451 syscallarg(const char *) path; 452 syscallarg(int) flags; 453 syscallarg(void *) data; 454 } */ 455 register_t dummy; 456 457 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 458 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 459 } 460 #endif 461 462 int 463 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 464 { 465 /* { 466 syscallarg(const char *) type; 467 syscallarg(const char *) path; 468 syscallarg(int) flags; 469 syscallarg(void *) data; 470 syscallarg(size_t) data_len; 471 } */ 472 473 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 474 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 475 SCARG(uap, data_len), retval); 476 } 477 478 int 479 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 480 const char *path, int flags, void *data, enum uio_seg data_seg, 481 size_t data_len, register_t *retval) 482 { 483 struct vnode *vp; 484 struct nameidata nd; 485 void *data_buf = data; 486 u_int recurse; 487 int error; 488 489 /* 490 * Get vnode to be covered 491 */ 492 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 493 if ((error = namei(&nd)) != 0) 494 return (error); 495 vp = nd.ni_vp; 496 497 /* 498 * A lookup in VFS_MOUNT might result in an attempt to 499 * lock this vnode again, so make the lock recursive. 500 */ 501 if (vfsops == NULL) { 502 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 503 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 504 recurse = vn_setrecurse(vp); 505 vfsops = vp->v_mount->mnt_op; 506 } else { 507 /* 'type' is userspace */ 508 error = mount_get_vfsops(type, &vfsops); 509 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 510 recurse = vn_setrecurse(vp); 511 if (error != 0) 512 goto done; 513 } 514 } else { 515 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 516 recurse = vn_setrecurse(vp); 517 } 518 519 if (data != NULL && data_seg == UIO_USERSPACE) { 520 if (data_len == 0) { 521 /* No length supplied, use default for filesystem */ 522 data_len = vfsops->vfs_min_mount_data; 523 if (data_len > VFS_MAX_MOUNT_DATA) { 524 /* maybe a force loaded old LKM */ 525 error = EINVAL; 526 goto done; 527 } 528 #ifdef COMPAT_30 529 /* Hopefully a longer buffer won't make copyin() fail */ 530 if (flags & MNT_UPDATE 531 && data_len < sizeof (struct mnt_export_args30)) 532 data_len = sizeof (struct mnt_export_args30); 533 #endif 534 } 535 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 536 537 /* NFS needs the buffer even for mnt_getargs .... */ 538 error = copyin(data, data_buf, data_len); 539 if (error != 0) 540 goto done; 541 } 542 543 if (flags & MNT_GETARGS) { 544 if (data_len == 0) { 545 error = EINVAL; 546 goto done; 547 } 548 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 549 if (error != 0) 550 goto done; 551 if (data_seg == UIO_USERSPACE) 552 error = copyout(data_buf, data, data_len); 553 *retval = data_len; 554 } else if (flags & MNT_UPDATE) { 555 error = mount_update(l, vp, path, flags, data_buf, &data_len); 556 } else { 557 /* Locking is handled internally in mount_domount(). */ 558 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 559 &data_len, recurse); 560 } 561 562 done: 563 if (vp != NULL) { 564 vn_restorerecurse(vp, recurse); 565 vput(vp); 566 } 567 if (data_buf != data) 568 free(data_buf, M_TEMP); 569 return (error); 570 } 571 572 /* 573 * Scan all active processes to see if any of them have a current 574 * or root directory onto which the new filesystem has just been 575 * mounted. If so, replace them with the new mount point. 576 */ 577 void 578 checkdirs(struct vnode *olddp) 579 { 580 struct cwdinfo *cwdi; 581 struct vnode *newdp, *rele1, *rele2; 582 struct proc *p; 583 bool retry; 584 585 if (olddp->v_usecount == 1) 586 return; 587 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 588 panic("mount: lost mount"); 589 590 do { 591 retry = false; 592 mutex_enter(proc_lock); 593 PROCLIST_FOREACH(p, &allproc) { 594 if ((p->p_flag & PK_MARKER) != 0) 595 continue; 596 if ((cwdi = p->p_cwdi) == NULL) 597 continue; 598 /* 599 * Can't change to the old directory any more, 600 * so even if we see a stale value it's not a 601 * problem. 602 */ 603 if (cwdi->cwdi_cdir != olddp && 604 cwdi->cwdi_rdir != olddp) 605 continue; 606 retry = true; 607 rele1 = NULL; 608 rele2 = NULL; 609 atomic_inc_uint(&cwdi->cwdi_refcnt); 610 mutex_exit(proc_lock); 611 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 612 if (cwdi->cwdi_cdir == olddp) { 613 rele1 = cwdi->cwdi_cdir; 614 VREF(newdp); 615 cwdi->cwdi_cdir = newdp; 616 } 617 if (cwdi->cwdi_rdir == olddp) { 618 rele2 = cwdi->cwdi_rdir; 619 VREF(newdp); 620 cwdi->cwdi_rdir = newdp; 621 } 622 rw_exit(&cwdi->cwdi_lock); 623 cwdfree(cwdi); 624 if (rele1 != NULL) 625 vrele(rele1); 626 if (rele2 != NULL) 627 vrele(rele2); 628 mutex_enter(proc_lock); 629 break; 630 } 631 mutex_exit(proc_lock); 632 } while (retry); 633 634 if (rootvnode == olddp) { 635 vrele(rootvnode); 636 VREF(newdp); 637 rootvnode = newdp; 638 } 639 vput(newdp); 640 } 641 642 /* 643 * Unmount a file system. 644 * 645 * Note: unmount takes a path to the vnode mounted on as argument, 646 * not special file (as before). 647 */ 648 /* ARGSUSED */ 649 int 650 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 651 { 652 /* { 653 syscallarg(const char *) path; 654 syscallarg(int) flags; 655 } */ 656 struct vnode *vp; 657 struct mount *mp; 658 int error; 659 struct nameidata nd; 660 661 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 662 SCARG(uap, path)); 663 if ((error = namei(&nd)) != 0) 664 return (error); 665 vp = nd.ni_vp; 666 mp = vp->v_mount; 667 atomic_inc_uint(&mp->mnt_refcnt); 668 VOP_UNLOCK(vp, 0); 669 670 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 671 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 672 if (error) { 673 vrele(vp); 674 vfs_destroy(mp); 675 return (error); 676 } 677 678 /* 679 * Don't allow unmounting the root file system. 680 */ 681 if (mp->mnt_flag & MNT_ROOTFS) { 682 vrele(vp); 683 vfs_destroy(mp); 684 return (EINVAL); 685 } 686 687 /* 688 * Must be the root of the filesystem 689 */ 690 if ((vp->v_vflag & VV_ROOT) == 0) { 691 vrele(vp); 692 vfs_destroy(mp); 693 return (EINVAL); 694 } 695 696 vrele(vp); 697 error = dounmount(mp, SCARG(uap, flags), l); 698 return error; 699 } 700 701 /* 702 * Do the actual file system unmount. File system is assumed to have 703 * been locked by the caller. 704 * 705 * => Caller gain reference to the mount, explicility for unmount. 706 * => Reference will be dropped in all cases. 707 */ 708 int 709 dounmount(struct mount *mp, int flags, struct lwp *l) 710 { 711 struct vnode *coveredvp; 712 int error; 713 int async; 714 int used_syncer; 715 716 #if NVERIEXEC > 0 717 error = veriexec_unmountchk(mp); 718 if (error) 719 return (error); 720 #endif /* NVERIEXEC > 0 */ 721 722 /* 723 * XXX Freeze syncer. Must do this before locking the 724 * mount point. See dounmount() for details. 725 */ 726 mutex_enter(&syncer_mutex); 727 rw_enter(&mp->mnt_unmounting, RW_WRITER); 728 if ((mp->mnt_iflag & IMNT_GONE) != 0) { 729 rw_exit(&mp->mnt_unmounting); 730 mutex_exit(&syncer_mutex); 731 vfs_destroy(mp); 732 return ENOENT; 733 } 734 735 used_syncer = (mp->mnt_syncer != NULL); 736 737 /* 738 * XXX Syncer must be frozen when we get here. This should really 739 * be done on a per-mountpoint basis, but especially the softdep 740 * code possibly called from the syncer doesn't exactly work on a 741 * per-mountpoint basis, so the softdep code would become a maze 742 * of vfs_busy() calls. 743 * 744 * The caller of dounmount() must acquire syncer_mutex because 745 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 746 * order, and we must preserve that order to avoid deadlock. 747 * 748 * So, if the file system did not use the syncer, now is 749 * the time to release the syncer_mutex. 750 */ 751 if (used_syncer == 0) 752 mutex_exit(&syncer_mutex); 753 754 mp->mnt_iflag |= IMNT_UNMOUNT; 755 async = mp->mnt_flag & MNT_ASYNC; 756 mp->mnt_flag &= ~MNT_ASYNC; 757 cache_purgevfs(mp); /* remove cache entries for this file sys */ 758 if (mp->mnt_syncer != NULL) 759 vfs_deallocate_syncvnode(mp); 760 error = 0; 761 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 762 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 763 } 764 vfs_scrubvnlist(mp); 765 if (error == 0 || (flags & MNT_FORCE)) 766 error = VFS_UNMOUNT(mp, flags); 767 if (error) { 768 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 769 (void) vfs_allocate_syncvnode(mp); 770 mp->mnt_iflag &= ~IMNT_UNMOUNT; 771 mp->mnt_flag |= async; 772 rw_exit(&mp->mnt_unmounting); 773 if (used_syncer) 774 mutex_exit(&syncer_mutex); 775 return (error); 776 } 777 vfs_scrubvnlist(mp); 778 mutex_enter(&mountlist_lock); 779 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 780 coveredvp->v_mountedhere = NULL; 781 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 782 mp->mnt_iflag |= IMNT_GONE; 783 mutex_exit(&mountlist_lock); 784 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 785 panic("unmount: dangling vnode"); 786 if (used_syncer) 787 mutex_exit(&syncer_mutex); 788 vfs_hooks_unmount(mp); 789 rw_exit(&mp->mnt_unmounting); 790 vfs_destroy(mp); /* caller provided reference */ 791 vfs_destroy(mp); /* from mount(), final nail in coffin */ 792 if (coveredvp != NULLVP) 793 vrele(coveredvp); 794 return (0); 795 } 796 797 /* 798 * Sync each mounted filesystem. 799 */ 800 #ifdef DEBUG 801 int syncprt = 0; 802 struct ctldebug debug0 = { "syncprt", &syncprt }; 803 #endif 804 805 /* ARGSUSED */ 806 int 807 sys_sync(struct lwp *l, const void *v, register_t *retval) 808 { 809 struct mount *mp, *nmp; 810 int asyncflag; 811 812 if (l == NULL) 813 l = &lwp0; 814 815 mutex_enter(&mountlist_lock); 816 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 817 mp = nmp) { 818 if (vfs_busy(mp, &nmp)) { 819 continue; 820 } 821 mutex_enter(&mp->mnt_updating); 822 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 823 asyncflag = mp->mnt_flag & MNT_ASYNC; 824 mp->mnt_flag &= ~MNT_ASYNC; 825 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 826 if (asyncflag) 827 mp->mnt_flag |= MNT_ASYNC; 828 } 829 mutex_exit(&mp->mnt_updating); 830 vfs_unbusy(mp, false, &nmp); 831 } 832 mutex_exit(&mountlist_lock); 833 #ifdef DEBUG 834 if (syncprt) 835 vfs_bufstats(); 836 #endif /* DEBUG */ 837 return (0); 838 } 839 840 /* 841 * Change filesystem quotas. 842 */ 843 /* ARGSUSED */ 844 int 845 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 846 { 847 /* { 848 syscallarg(const char *) path; 849 syscallarg(int) cmd; 850 syscallarg(int) uid; 851 syscallarg(void *) arg; 852 } */ 853 struct mount *mp; 854 int error; 855 struct nameidata nd; 856 857 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 858 SCARG(uap, path)); 859 if ((error = namei(&nd)) != 0) 860 return (error); 861 mp = nd.ni_vp->v_mount; 862 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 863 SCARG(uap, arg)); 864 vrele(nd.ni_vp); 865 return (error); 866 } 867 868 int 869 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 870 int root) 871 { 872 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 873 int error = 0; 874 875 /* 876 * If MNT_NOWAIT or MNT_LAZY is specified, do not 877 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 878 * overrides MNT_NOWAIT. 879 */ 880 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 881 (flags != MNT_WAIT && flags != 0)) { 882 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 883 goto done; 884 } 885 886 /* Get the filesystem stats now */ 887 memset(sp, 0, sizeof(*sp)); 888 if ((error = VFS_STATVFS(mp, sp)) != 0) { 889 return error; 890 } 891 892 if (cwdi->cwdi_rdir == NULL) 893 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 894 done: 895 if (cwdi->cwdi_rdir != NULL) { 896 size_t len; 897 char *bp; 898 char c; 899 char *path = PNBUF_GET(); 900 901 bp = path + MAXPATHLEN; 902 *--bp = '\0'; 903 rw_enter(&cwdi->cwdi_lock, RW_READER); 904 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 905 MAXPATHLEN / 2, 0, l); 906 rw_exit(&cwdi->cwdi_lock); 907 if (error) { 908 PNBUF_PUT(path); 909 return error; 910 } 911 len = strlen(bp); 912 /* 913 * for mount points that are below our root, we can see 914 * them, so we fix up the pathname and return them. The 915 * rest we cannot see, so we don't allow viewing the 916 * data. 917 */ 918 if (strncmp(bp, sp->f_mntonname, len) == 0 && 919 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 920 (void)strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 921 sizeof(sp->f_mntonname)); 922 if (sp->f_mntonname[0] == '\0') 923 (void)strlcpy(sp->f_mntonname, "/", 924 sizeof(sp->f_mntonname)); 925 } else { 926 if (root) 927 (void)strlcpy(sp->f_mntonname, "/", 928 sizeof(sp->f_mntonname)); 929 else 930 error = EPERM; 931 } 932 PNBUF_PUT(path); 933 } 934 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 935 return error; 936 } 937 938 /* 939 * Get filesystem statistics by path. 940 */ 941 int 942 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 943 { 944 struct mount *mp; 945 int error; 946 struct nameidata nd; 947 948 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 949 if ((error = namei(&nd)) != 0) 950 return error; 951 mp = nd.ni_vp->v_mount; 952 error = dostatvfs(mp, sb, l, flags, 1); 953 vrele(nd.ni_vp); 954 return error; 955 } 956 957 /* ARGSUSED */ 958 int 959 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 960 { 961 /* { 962 syscallarg(const char *) path; 963 syscallarg(struct statvfs *) buf; 964 syscallarg(int) flags; 965 } */ 966 struct statvfs *sb; 967 int error; 968 969 sb = STATVFSBUF_GET(); 970 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 971 if (error == 0) 972 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 973 STATVFSBUF_PUT(sb); 974 return error; 975 } 976 977 /* 978 * Get filesystem statistics by fd. 979 */ 980 int 981 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 982 { 983 file_t *fp; 984 struct mount *mp; 985 int error; 986 987 /* fd_getvnode() will use the descriptor for us */ 988 if ((error = fd_getvnode(fd, &fp)) != 0) 989 return (error); 990 mp = ((struct vnode *)fp->f_data)->v_mount; 991 error = dostatvfs(mp, sb, curlwp, flags, 1); 992 fd_putfile(fd); 993 return error; 994 } 995 996 /* ARGSUSED */ 997 int 998 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 999 { 1000 /* { 1001 syscallarg(int) fd; 1002 syscallarg(struct statvfs *) buf; 1003 syscallarg(int) flags; 1004 } */ 1005 struct statvfs *sb; 1006 int error; 1007 1008 sb = STATVFSBUF_GET(); 1009 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1010 if (error == 0) 1011 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1012 STATVFSBUF_PUT(sb); 1013 return error; 1014 } 1015 1016 1017 /* 1018 * Get statistics on all filesystems. 1019 */ 1020 int 1021 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1022 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1023 register_t *retval) 1024 { 1025 int root = 0; 1026 struct proc *p = l->l_proc; 1027 struct mount *mp, *nmp; 1028 struct statvfs *sb; 1029 size_t count, maxcount; 1030 int error = 0; 1031 1032 sb = STATVFSBUF_GET(); 1033 maxcount = bufsize / entry_sz; 1034 mutex_enter(&mountlist_lock); 1035 count = 0; 1036 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 1037 mp = nmp) { 1038 if (vfs_busy(mp, &nmp)) { 1039 continue; 1040 } 1041 if (sfsp && count < maxcount) { 1042 error = dostatvfs(mp, sb, l, flags, 0); 1043 if (error) { 1044 vfs_unbusy(mp, false, &nmp); 1045 error = 0; 1046 continue; 1047 } 1048 error = copyfn(sb, sfsp, entry_sz); 1049 if (error) { 1050 vfs_unbusy(mp, false, NULL); 1051 goto out; 1052 } 1053 sfsp = (char *)sfsp + entry_sz; 1054 root |= strcmp(sb->f_mntonname, "/") == 0; 1055 } 1056 count++; 1057 vfs_unbusy(mp, false, &nmp); 1058 } 1059 mutex_exit(&mountlist_lock); 1060 1061 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1062 /* 1063 * fake a root entry 1064 */ 1065 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1066 sb, l, flags, 1); 1067 if (error != 0) 1068 goto out; 1069 if (sfsp) { 1070 error = copyfn(sb, sfsp, entry_sz); 1071 if (error != 0) 1072 goto out; 1073 } 1074 count++; 1075 } 1076 if (sfsp && count > maxcount) 1077 *retval = maxcount; 1078 else 1079 *retval = count; 1080 out: 1081 STATVFSBUF_PUT(sb); 1082 return error; 1083 } 1084 1085 int 1086 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1087 { 1088 /* { 1089 syscallarg(struct statvfs *) buf; 1090 syscallarg(size_t) bufsize; 1091 syscallarg(int) flags; 1092 } */ 1093 1094 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1095 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1096 } 1097 1098 /* 1099 * Change current working directory to a given file descriptor. 1100 */ 1101 /* ARGSUSED */ 1102 int 1103 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1104 { 1105 /* { 1106 syscallarg(int) fd; 1107 } */ 1108 struct proc *p = l->l_proc; 1109 struct cwdinfo *cwdi; 1110 struct vnode *vp, *tdp; 1111 struct mount *mp; 1112 file_t *fp; 1113 int error, fd; 1114 1115 /* fd_getvnode() will use the descriptor for us */ 1116 fd = SCARG(uap, fd); 1117 if ((error = fd_getvnode(fd, &fp)) != 0) 1118 return (error); 1119 vp = fp->f_data; 1120 1121 VREF(vp); 1122 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1123 if (vp->v_type != VDIR) 1124 error = ENOTDIR; 1125 else 1126 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1127 if (error) { 1128 vput(vp); 1129 goto out; 1130 } 1131 while ((mp = vp->v_mountedhere) != NULL) { 1132 error = vfs_busy(mp, NULL); 1133 vput(vp); 1134 if (error != 0) 1135 goto out; 1136 error = VFS_ROOT(mp, &tdp); 1137 vfs_unbusy(mp, false, NULL); 1138 if (error) 1139 goto out; 1140 vp = tdp; 1141 } 1142 VOP_UNLOCK(vp, 0); 1143 1144 /* 1145 * Disallow changing to a directory not under the process's 1146 * current root directory (if there is one). 1147 */ 1148 cwdi = p->p_cwdi; 1149 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1150 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1151 vrele(vp); 1152 error = EPERM; /* operation not permitted */ 1153 } else { 1154 vrele(cwdi->cwdi_cdir); 1155 cwdi->cwdi_cdir = vp; 1156 } 1157 rw_exit(&cwdi->cwdi_lock); 1158 1159 out: 1160 fd_putfile(fd); 1161 return (error); 1162 } 1163 1164 /* 1165 * Change this process's notion of the root directory to a given file 1166 * descriptor. 1167 */ 1168 int 1169 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1170 { 1171 struct proc *p = l->l_proc; 1172 struct cwdinfo *cwdi; 1173 struct vnode *vp; 1174 file_t *fp; 1175 int error, fd = SCARG(uap, fd); 1176 1177 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1178 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1179 return error; 1180 /* fd_getvnode() will use the descriptor for us */ 1181 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 1182 return error; 1183 vp = fp->f_data; 1184 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1185 if (vp->v_type != VDIR) 1186 error = ENOTDIR; 1187 else 1188 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1189 VOP_UNLOCK(vp, 0); 1190 if (error) 1191 goto out; 1192 VREF(vp); 1193 1194 /* 1195 * Prevent escaping from chroot by putting the root under 1196 * the working directory. Silently chdir to / if we aren't 1197 * already there. 1198 */ 1199 cwdi = p->p_cwdi; 1200 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1201 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1202 /* 1203 * XXX would be more failsafe to change directory to a 1204 * deadfs node here instead 1205 */ 1206 vrele(cwdi->cwdi_cdir); 1207 VREF(vp); 1208 cwdi->cwdi_cdir = vp; 1209 } 1210 1211 if (cwdi->cwdi_rdir != NULL) 1212 vrele(cwdi->cwdi_rdir); 1213 cwdi->cwdi_rdir = vp; 1214 rw_exit(&cwdi->cwdi_lock); 1215 1216 out: 1217 fd_putfile(fd); 1218 return (error); 1219 } 1220 1221 /* 1222 * Change current working directory (``.''). 1223 */ 1224 /* ARGSUSED */ 1225 int 1226 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1227 { 1228 /* { 1229 syscallarg(const char *) path; 1230 } */ 1231 struct proc *p = l->l_proc; 1232 struct cwdinfo *cwdi; 1233 int error; 1234 struct nameidata nd; 1235 1236 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1237 SCARG(uap, path)); 1238 if ((error = change_dir(&nd, l)) != 0) 1239 return (error); 1240 cwdi = p->p_cwdi; 1241 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1242 vrele(cwdi->cwdi_cdir); 1243 cwdi->cwdi_cdir = nd.ni_vp; 1244 rw_exit(&cwdi->cwdi_lock); 1245 return (0); 1246 } 1247 1248 /* 1249 * Change notion of root (``/'') directory. 1250 */ 1251 /* ARGSUSED */ 1252 int 1253 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1254 { 1255 /* { 1256 syscallarg(const char *) path; 1257 } */ 1258 struct proc *p = l->l_proc; 1259 struct cwdinfo *cwdi; 1260 struct vnode *vp; 1261 int error; 1262 struct nameidata nd; 1263 1264 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1265 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1266 return (error); 1267 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1268 SCARG(uap, path)); 1269 if ((error = change_dir(&nd, l)) != 0) 1270 return (error); 1271 1272 cwdi = p->p_cwdi; 1273 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1274 if (cwdi->cwdi_rdir != NULL) 1275 vrele(cwdi->cwdi_rdir); 1276 vp = nd.ni_vp; 1277 cwdi->cwdi_rdir = vp; 1278 1279 /* 1280 * Prevent escaping from chroot by putting the root under 1281 * the working directory. Silently chdir to / if we aren't 1282 * already there. 1283 */ 1284 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1285 /* 1286 * XXX would be more failsafe to change directory to a 1287 * deadfs node here instead 1288 */ 1289 vrele(cwdi->cwdi_cdir); 1290 VREF(vp); 1291 cwdi->cwdi_cdir = vp; 1292 } 1293 rw_exit(&cwdi->cwdi_lock); 1294 1295 return (0); 1296 } 1297 1298 /* 1299 * Common routine for chroot and chdir. 1300 */ 1301 static int 1302 change_dir(struct nameidata *ndp, struct lwp *l) 1303 { 1304 struct vnode *vp; 1305 int error; 1306 1307 if ((error = namei(ndp)) != 0) 1308 return (error); 1309 vp = ndp->ni_vp; 1310 if (vp->v_type != VDIR) 1311 error = ENOTDIR; 1312 else 1313 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1314 1315 if (error) 1316 vput(vp); 1317 else 1318 VOP_UNLOCK(vp, 0); 1319 return (error); 1320 } 1321 1322 /* 1323 * Check permissions, allocate an open file structure, 1324 * and call the device open routine if any. 1325 */ 1326 int 1327 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1328 { 1329 /* { 1330 syscallarg(const char *) path; 1331 syscallarg(int) flags; 1332 syscallarg(int) mode; 1333 } */ 1334 struct proc *p = l->l_proc; 1335 struct cwdinfo *cwdi = p->p_cwdi; 1336 file_t *fp; 1337 struct vnode *vp; 1338 int flags, cmode; 1339 int type, indx, error; 1340 struct flock lf; 1341 struct nameidata nd; 1342 1343 flags = FFLAGS(SCARG(uap, flags)); 1344 if ((flags & (FREAD | FWRITE)) == 0) 1345 return (EINVAL); 1346 if ((error = fd_allocfile(&fp, &indx)) != 0) 1347 return (error); 1348 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1349 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1350 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1351 SCARG(uap, path)); 1352 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1353 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1354 fd_abort(p, fp, indx); 1355 if ((error == EDUPFD || error == EMOVEFD) && 1356 l->l_dupfd >= 0 && /* XXX from fdopen */ 1357 (error = 1358 fd_dupopen(l->l_dupfd, &indx, flags, error)) == 0) { 1359 *retval = indx; 1360 return (0); 1361 } 1362 if (error == ERESTART) 1363 error = EINTR; 1364 return (error); 1365 } 1366 1367 l->l_dupfd = 0; 1368 vp = nd.ni_vp; 1369 fp->f_flag = flags & FMASK; 1370 fp->f_type = DTYPE_VNODE; 1371 fp->f_ops = &vnops; 1372 fp->f_data = vp; 1373 if (flags & (O_EXLOCK | O_SHLOCK)) { 1374 lf.l_whence = SEEK_SET; 1375 lf.l_start = 0; 1376 lf.l_len = 0; 1377 if (flags & O_EXLOCK) 1378 lf.l_type = F_WRLCK; 1379 else 1380 lf.l_type = F_RDLCK; 1381 type = F_FLOCK; 1382 if ((flags & FNONBLOCK) == 0) 1383 type |= F_WAIT; 1384 VOP_UNLOCK(vp, 0); 1385 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1386 if (error) { 1387 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1388 fd_abort(p, fp, indx); 1389 return (error); 1390 } 1391 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1392 atomic_or_uint(&fp->f_flag, FHASLOCK); 1393 } 1394 VOP_UNLOCK(vp, 0); 1395 *retval = indx; 1396 fd_affix(p, fp, indx); 1397 return (0); 1398 } 1399 1400 static void 1401 vfs__fhfree(fhandle_t *fhp) 1402 { 1403 size_t fhsize; 1404 1405 if (fhp == NULL) { 1406 return; 1407 } 1408 fhsize = FHANDLE_SIZE(fhp); 1409 kmem_free(fhp, fhsize); 1410 } 1411 1412 /* 1413 * vfs_composefh: compose a filehandle. 1414 */ 1415 1416 int 1417 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1418 { 1419 struct mount *mp; 1420 struct fid *fidp; 1421 int error; 1422 size_t needfhsize; 1423 size_t fidsize; 1424 1425 mp = vp->v_mount; 1426 fidp = NULL; 1427 if (*fh_size < FHANDLE_SIZE_MIN) { 1428 fidsize = 0; 1429 } else { 1430 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1431 if (fhp != NULL) { 1432 memset(fhp, 0, *fh_size); 1433 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1434 fidp = &fhp->fh_fid; 1435 } 1436 } 1437 error = VFS_VPTOFH(vp, fidp, &fidsize); 1438 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1439 if (error == 0 && *fh_size < needfhsize) { 1440 error = E2BIG; 1441 } 1442 *fh_size = needfhsize; 1443 return error; 1444 } 1445 1446 int 1447 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1448 { 1449 struct mount *mp; 1450 fhandle_t *fhp; 1451 size_t fhsize; 1452 size_t fidsize; 1453 int error; 1454 1455 *fhpp = NULL; 1456 mp = vp->v_mount; 1457 fidsize = 0; 1458 error = VFS_VPTOFH(vp, NULL, &fidsize); 1459 KASSERT(error != 0); 1460 if (error != E2BIG) { 1461 goto out; 1462 } 1463 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1464 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1465 if (fhp == NULL) { 1466 error = ENOMEM; 1467 goto out; 1468 } 1469 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1470 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1471 if (error == 0) { 1472 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1473 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1474 *fhpp = fhp; 1475 } else { 1476 kmem_free(fhp, fhsize); 1477 } 1478 out: 1479 return error; 1480 } 1481 1482 void 1483 vfs_composefh_free(fhandle_t *fhp) 1484 { 1485 1486 vfs__fhfree(fhp); 1487 } 1488 1489 /* 1490 * vfs_fhtovp: lookup a vnode by a filehandle. 1491 */ 1492 1493 int 1494 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1495 { 1496 struct mount *mp; 1497 int error; 1498 1499 *vpp = NULL; 1500 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1501 if (mp == NULL) { 1502 error = ESTALE; 1503 goto out; 1504 } 1505 if (mp->mnt_op->vfs_fhtovp == NULL) { 1506 error = EOPNOTSUPP; 1507 goto out; 1508 } 1509 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1510 out: 1511 return error; 1512 } 1513 1514 /* 1515 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1516 * the needed size. 1517 */ 1518 1519 int 1520 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1521 { 1522 fhandle_t *fhp; 1523 int error; 1524 1525 *fhpp = NULL; 1526 if (fhsize > FHANDLE_SIZE_MAX) { 1527 return EINVAL; 1528 } 1529 if (fhsize < FHANDLE_SIZE_MIN) { 1530 return EINVAL; 1531 } 1532 again: 1533 fhp = kmem_alloc(fhsize, KM_SLEEP); 1534 if (fhp == NULL) { 1535 return ENOMEM; 1536 } 1537 error = copyin(ufhp, fhp, fhsize); 1538 if (error == 0) { 1539 /* XXX this check shouldn't be here */ 1540 if (FHANDLE_SIZE(fhp) == fhsize) { 1541 *fhpp = fhp; 1542 return 0; 1543 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1544 /* 1545 * a kludge for nfsv2 padded handles. 1546 */ 1547 size_t sz; 1548 1549 sz = FHANDLE_SIZE(fhp); 1550 kmem_free(fhp, fhsize); 1551 fhsize = sz; 1552 goto again; 1553 } else { 1554 /* 1555 * userland told us wrong size. 1556 */ 1557 error = EINVAL; 1558 } 1559 } 1560 kmem_free(fhp, fhsize); 1561 return error; 1562 } 1563 1564 void 1565 vfs_copyinfh_free(fhandle_t *fhp) 1566 { 1567 1568 vfs__fhfree(fhp); 1569 } 1570 1571 /* 1572 * Get file handle system call 1573 */ 1574 int 1575 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1576 { 1577 /* { 1578 syscallarg(char *) fname; 1579 syscallarg(fhandle_t *) fhp; 1580 syscallarg(size_t *) fh_size; 1581 } */ 1582 struct vnode *vp; 1583 fhandle_t *fh; 1584 int error; 1585 struct nameidata nd; 1586 size_t sz; 1587 size_t usz; 1588 1589 /* 1590 * Must be super user 1591 */ 1592 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1593 0, NULL, NULL, NULL); 1594 if (error) 1595 return (error); 1596 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1597 SCARG(uap, fname)); 1598 error = namei(&nd); 1599 if (error) 1600 return (error); 1601 vp = nd.ni_vp; 1602 error = vfs_composefh_alloc(vp, &fh); 1603 vput(vp); 1604 if (error != 0) { 1605 goto out; 1606 } 1607 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1608 if (error != 0) { 1609 goto out; 1610 } 1611 sz = FHANDLE_SIZE(fh); 1612 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1613 if (error != 0) { 1614 goto out; 1615 } 1616 if (usz >= sz) { 1617 error = copyout(fh, SCARG(uap, fhp), sz); 1618 } else { 1619 error = E2BIG; 1620 } 1621 out: 1622 vfs_composefh_free(fh); 1623 return (error); 1624 } 1625 1626 /* 1627 * Open a file given a file handle. 1628 * 1629 * Check permissions, allocate an open file structure, 1630 * and call the device open routine if any. 1631 */ 1632 1633 int 1634 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1635 register_t *retval) 1636 { 1637 file_t *fp; 1638 struct vnode *vp = NULL; 1639 kauth_cred_t cred = l->l_cred; 1640 file_t *nfp; 1641 int type, indx, error=0; 1642 struct flock lf; 1643 struct vattr va; 1644 fhandle_t *fh; 1645 int flags; 1646 proc_t *p; 1647 1648 p = curproc; 1649 1650 /* 1651 * Must be super user 1652 */ 1653 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1654 0, NULL, NULL, NULL))) 1655 return (error); 1656 1657 flags = FFLAGS(oflags); 1658 if ((flags & (FREAD | FWRITE)) == 0) 1659 return (EINVAL); 1660 if ((flags & O_CREAT)) 1661 return (EINVAL); 1662 if ((error = fd_allocfile(&nfp, &indx)) != 0) 1663 return (error); 1664 fp = nfp; 1665 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1666 if (error != 0) { 1667 goto bad; 1668 } 1669 error = vfs_fhtovp(fh, &vp); 1670 if (error != 0) { 1671 goto bad; 1672 } 1673 1674 /* Now do an effective vn_open */ 1675 1676 if (vp->v_type == VSOCK) { 1677 error = EOPNOTSUPP; 1678 goto bad; 1679 } 1680 error = vn_openchk(vp, cred, flags); 1681 if (error != 0) 1682 goto bad; 1683 if (flags & O_TRUNC) { 1684 VOP_UNLOCK(vp, 0); /* XXX */ 1685 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1686 VATTR_NULL(&va); 1687 va.va_size = 0; 1688 error = VOP_SETATTR(vp, &va, cred); 1689 if (error) 1690 goto bad; 1691 } 1692 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1693 goto bad; 1694 if (flags & FWRITE) { 1695 mutex_enter(&vp->v_interlock); 1696 vp->v_writecount++; 1697 mutex_exit(&vp->v_interlock); 1698 } 1699 1700 /* done with modified vn_open, now finish what sys_open does. */ 1701 1702 fp->f_flag = flags & FMASK; 1703 fp->f_type = DTYPE_VNODE; 1704 fp->f_ops = &vnops; 1705 fp->f_data = vp; 1706 if (flags & (O_EXLOCK | O_SHLOCK)) { 1707 lf.l_whence = SEEK_SET; 1708 lf.l_start = 0; 1709 lf.l_len = 0; 1710 if (flags & O_EXLOCK) 1711 lf.l_type = F_WRLCK; 1712 else 1713 lf.l_type = F_RDLCK; 1714 type = F_FLOCK; 1715 if ((flags & FNONBLOCK) == 0) 1716 type |= F_WAIT; 1717 VOP_UNLOCK(vp, 0); 1718 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1719 if (error) { 1720 (void) vn_close(vp, fp->f_flag, fp->f_cred); 1721 fd_abort(p, fp, indx); 1722 return (error); 1723 } 1724 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1725 atomic_or_uint(&fp->f_flag, FHASLOCK); 1726 } 1727 VOP_UNLOCK(vp, 0); 1728 *retval = indx; 1729 fd_affix(p, fp, indx); 1730 vfs_copyinfh_free(fh); 1731 return (0); 1732 1733 bad: 1734 fd_abort(p, fp, indx); 1735 if (vp != NULL) 1736 vput(vp); 1737 vfs_copyinfh_free(fh); 1738 return (error); 1739 } 1740 1741 int 1742 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1743 { 1744 /* { 1745 syscallarg(const void *) fhp; 1746 syscallarg(size_t) fh_size; 1747 syscallarg(int) flags; 1748 } */ 1749 1750 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1751 SCARG(uap, flags), retval); 1752 } 1753 1754 int 1755 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1756 { 1757 int error; 1758 fhandle_t *fh; 1759 struct vnode *vp; 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 error = vn_stat(vp, sb); 1778 vput(vp); 1779 return error; 1780 } 1781 1782 1783 /* ARGSUSED */ 1784 int 1785 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1786 { 1787 /* { 1788 syscallarg(const void *) fhp; 1789 syscallarg(size_t) fh_size; 1790 syscallarg(struct stat *) sb; 1791 } */ 1792 struct stat sb; 1793 int error; 1794 1795 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1796 if (error) 1797 return error; 1798 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1799 } 1800 1801 int 1802 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1803 int flags) 1804 { 1805 fhandle_t *fh; 1806 struct mount *mp; 1807 struct vnode *vp; 1808 int error; 1809 1810 /* 1811 * Must be super user 1812 */ 1813 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1814 0, NULL, NULL, NULL))) 1815 return error; 1816 1817 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1818 if (error != 0) 1819 return error; 1820 1821 error = vfs_fhtovp(fh, &vp); 1822 vfs_copyinfh_free(fh); 1823 if (error != 0) 1824 return error; 1825 1826 mp = vp->v_mount; 1827 error = dostatvfs(mp, sb, l, flags, 1); 1828 vput(vp); 1829 return error; 1830 } 1831 1832 /* ARGSUSED */ 1833 int 1834 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1835 { 1836 /* { 1837 syscallarg(const void *) fhp; 1838 syscallarg(size_t) fh_size; 1839 syscallarg(struct statvfs *) buf; 1840 syscallarg(int) flags; 1841 } */ 1842 struct statvfs *sb = STATVFSBUF_GET(); 1843 int error; 1844 1845 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1846 SCARG(uap, flags)); 1847 if (error == 0) 1848 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1849 STATVFSBUF_PUT(sb); 1850 return error; 1851 } 1852 1853 /* 1854 * Create a special file. 1855 */ 1856 /* ARGSUSED */ 1857 int 1858 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1859 { 1860 /* { 1861 syscallarg(const char *) path; 1862 syscallarg(int) mode; 1863 syscallarg(int) dev; 1864 } */ 1865 struct proc *p = l->l_proc; 1866 struct vnode *vp; 1867 struct vattr vattr; 1868 int error, optype; 1869 struct nameidata nd; 1870 char *path; 1871 const char *cpath; 1872 enum uio_seg seg = UIO_USERSPACE; 1873 1874 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1875 0, NULL, NULL, NULL)) != 0) 1876 return (error); 1877 1878 optype = VOP_MKNOD_DESCOFFSET; 1879 1880 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1881 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1882 1883 if ((error = namei(&nd)) != 0) 1884 goto out; 1885 vp = nd.ni_vp; 1886 if (vp != NULL) 1887 error = EEXIST; 1888 else { 1889 VATTR_NULL(&vattr); 1890 /* We will read cwdi->cwdi_cmask unlocked. */ 1891 vattr.va_mode = 1892 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1893 vattr.va_rdev = SCARG(uap, dev); 1894 1895 switch (SCARG(uap, mode) & S_IFMT) { 1896 case S_IFMT: /* used by badsect to flag bad sectors */ 1897 vattr.va_type = VBAD; 1898 break; 1899 case S_IFCHR: 1900 vattr.va_type = VCHR; 1901 break; 1902 case S_IFBLK: 1903 vattr.va_type = VBLK; 1904 break; 1905 case S_IFWHT: 1906 optype = VOP_WHITEOUT_DESCOFFSET; 1907 break; 1908 case S_IFREG: 1909 #if NVERIEXEC > 0 1910 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1911 O_CREAT); 1912 #endif /* NVERIEXEC > 0 */ 1913 vattr.va_type = VREG; 1914 vattr.va_rdev = VNOVAL; 1915 optype = VOP_CREATE_DESCOFFSET; 1916 break; 1917 default: 1918 error = EINVAL; 1919 break; 1920 } 1921 } 1922 if (!error) { 1923 switch (optype) { 1924 case VOP_WHITEOUT_DESCOFFSET: 1925 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1926 if (error) 1927 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1928 vput(nd.ni_dvp); 1929 break; 1930 1931 case VOP_MKNOD_DESCOFFSET: 1932 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1933 &nd.ni_cnd, &vattr); 1934 if (error == 0) 1935 vput(nd.ni_vp); 1936 break; 1937 1938 case VOP_CREATE_DESCOFFSET: 1939 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1940 &nd.ni_cnd, &vattr); 1941 if (error == 0) 1942 vput(nd.ni_vp); 1943 break; 1944 } 1945 } else { 1946 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1947 if (nd.ni_dvp == vp) 1948 vrele(nd.ni_dvp); 1949 else 1950 vput(nd.ni_dvp); 1951 if (vp) 1952 vrele(vp); 1953 } 1954 out: 1955 VERIEXEC_PATH_PUT(path); 1956 return (error); 1957 } 1958 1959 /* 1960 * Create a named pipe. 1961 */ 1962 /* ARGSUSED */ 1963 int 1964 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1965 { 1966 /* { 1967 syscallarg(const char *) path; 1968 syscallarg(int) mode; 1969 } */ 1970 struct proc *p = l->l_proc; 1971 struct vattr vattr; 1972 int error; 1973 struct nameidata nd; 1974 1975 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1976 SCARG(uap, path)); 1977 if ((error = namei(&nd)) != 0) 1978 return (error); 1979 if (nd.ni_vp != NULL) { 1980 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1981 if (nd.ni_dvp == nd.ni_vp) 1982 vrele(nd.ni_dvp); 1983 else 1984 vput(nd.ni_dvp); 1985 vrele(nd.ni_vp); 1986 return (EEXIST); 1987 } 1988 VATTR_NULL(&vattr); 1989 vattr.va_type = VFIFO; 1990 /* We will read cwdi->cwdi_cmask unlocked. */ 1991 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1992 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1993 if (error == 0) 1994 vput(nd.ni_vp); 1995 return (error); 1996 } 1997 1998 /* 1999 * Make a hard file link. 2000 */ 2001 /* ARGSUSED */ 2002 int 2003 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2004 { 2005 /* { 2006 syscallarg(const char *) path; 2007 syscallarg(const char *) link; 2008 } */ 2009 struct vnode *vp; 2010 struct nameidata nd; 2011 int error; 2012 2013 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2014 SCARG(uap, path)); 2015 if ((error = namei(&nd)) != 0) 2016 return (error); 2017 vp = nd.ni_vp; 2018 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2019 SCARG(uap, link)); 2020 if ((error = namei(&nd)) != 0) 2021 goto out; 2022 if (nd.ni_vp) { 2023 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2024 if (nd.ni_dvp == nd.ni_vp) 2025 vrele(nd.ni_dvp); 2026 else 2027 vput(nd.ni_dvp); 2028 vrele(nd.ni_vp); 2029 error = EEXIST; 2030 goto out; 2031 } 2032 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2033 out: 2034 vrele(vp); 2035 return (error); 2036 } 2037 2038 /* 2039 * Make a symbolic link. 2040 */ 2041 /* ARGSUSED */ 2042 int 2043 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2044 { 2045 /* { 2046 syscallarg(const char *) path; 2047 syscallarg(const char *) link; 2048 } */ 2049 struct proc *p = l->l_proc; 2050 struct vattr vattr; 2051 char *path; 2052 int error; 2053 struct nameidata nd; 2054 2055 path = PNBUF_GET(); 2056 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2057 if (error) 2058 goto out; 2059 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2060 SCARG(uap, link)); 2061 if ((error = namei(&nd)) != 0) 2062 goto out; 2063 if (nd.ni_vp) { 2064 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2065 if (nd.ni_dvp == nd.ni_vp) 2066 vrele(nd.ni_dvp); 2067 else 2068 vput(nd.ni_dvp); 2069 vrele(nd.ni_vp); 2070 error = EEXIST; 2071 goto out; 2072 } 2073 VATTR_NULL(&vattr); 2074 vattr.va_type = VLNK; 2075 /* We will read cwdi->cwdi_cmask unlocked. */ 2076 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2077 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2078 if (error == 0) 2079 vput(nd.ni_vp); 2080 out: 2081 PNBUF_PUT(path); 2082 return (error); 2083 } 2084 2085 /* 2086 * Delete a whiteout from the filesystem. 2087 */ 2088 /* ARGSUSED */ 2089 int 2090 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2091 { 2092 /* { 2093 syscallarg(const char *) path; 2094 } */ 2095 int error; 2096 struct nameidata nd; 2097 2098 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2099 UIO_USERSPACE, SCARG(uap, path)); 2100 error = namei(&nd); 2101 if (error) 2102 return (error); 2103 2104 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2105 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2106 if (nd.ni_dvp == nd.ni_vp) 2107 vrele(nd.ni_dvp); 2108 else 2109 vput(nd.ni_dvp); 2110 if (nd.ni_vp) 2111 vrele(nd.ni_vp); 2112 return (EEXIST); 2113 } 2114 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2115 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2116 vput(nd.ni_dvp); 2117 return (error); 2118 } 2119 2120 /* 2121 * Delete a name from the filesystem. 2122 */ 2123 /* ARGSUSED */ 2124 int 2125 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2126 { 2127 /* { 2128 syscallarg(const char *) path; 2129 } */ 2130 2131 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2132 } 2133 2134 int 2135 do_sys_unlink(const char *arg, enum uio_seg seg) 2136 { 2137 struct vnode *vp; 2138 int error; 2139 struct nameidata nd; 2140 kauth_cred_t cred; 2141 char *path; 2142 const char *cpath; 2143 2144 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2145 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2146 2147 if ((error = namei(&nd)) != 0) 2148 goto out; 2149 vp = nd.ni_vp; 2150 2151 /* 2152 * The root of a mounted filesystem cannot be deleted. 2153 */ 2154 if (vp->v_vflag & VV_ROOT) { 2155 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2156 if (nd.ni_dvp == vp) 2157 vrele(nd.ni_dvp); 2158 else 2159 vput(nd.ni_dvp); 2160 vput(vp); 2161 error = EBUSY; 2162 goto out; 2163 } 2164 2165 #if NVERIEXEC > 0 2166 /* Handle remove requests for veriexec entries. */ 2167 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2168 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2169 if (nd.ni_dvp == vp) 2170 vrele(nd.ni_dvp); 2171 else 2172 vput(nd.ni_dvp); 2173 vput(vp); 2174 goto out; 2175 } 2176 #endif /* NVERIEXEC > 0 */ 2177 2178 cred = kauth_cred_get(); 2179 #ifdef FILEASSOC 2180 (void)fileassoc_file_delete(vp); 2181 #endif /* FILEASSOC */ 2182 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2183 out: 2184 VERIEXEC_PATH_PUT(path); 2185 return (error); 2186 } 2187 2188 /* 2189 * Reposition read/write file offset. 2190 */ 2191 int 2192 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2193 { 2194 /* { 2195 syscallarg(int) fd; 2196 syscallarg(int) pad; 2197 syscallarg(off_t) offset; 2198 syscallarg(int) whence; 2199 } */ 2200 kauth_cred_t cred = l->l_cred; 2201 file_t *fp; 2202 struct vnode *vp; 2203 struct vattr vattr; 2204 off_t newoff; 2205 int error, fd; 2206 2207 fd = SCARG(uap, fd); 2208 2209 if ((fp = fd_getfile(fd)) == NULL) 2210 return (EBADF); 2211 2212 vp = fp->f_data; 2213 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2214 error = ESPIPE; 2215 goto out; 2216 } 2217 2218 switch (SCARG(uap, whence)) { 2219 case SEEK_CUR: 2220 newoff = fp->f_offset + SCARG(uap, offset); 2221 break; 2222 case SEEK_END: 2223 error = VOP_GETATTR(vp, &vattr, cred); 2224 if (error) { 2225 goto out; 2226 } 2227 newoff = SCARG(uap, offset) + vattr.va_size; 2228 break; 2229 case SEEK_SET: 2230 newoff = SCARG(uap, offset); 2231 break; 2232 default: 2233 error = EINVAL; 2234 goto out; 2235 } 2236 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2237 *(off_t *)retval = fp->f_offset = newoff; 2238 } 2239 out: 2240 fd_putfile(fd); 2241 return (error); 2242 } 2243 2244 /* 2245 * Positional read system call. 2246 */ 2247 int 2248 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2249 { 2250 /* { 2251 syscallarg(int) fd; 2252 syscallarg(void *) buf; 2253 syscallarg(size_t) nbyte; 2254 syscallarg(off_t) offset; 2255 } */ 2256 file_t *fp; 2257 struct vnode *vp; 2258 off_t offset; 2259 int error, fd = SCARG(uap, fd); 2260 2261 if ((fp = fd_getfile(fd)) == NULL) 2262 return (EBADF); 2263 2264 if ((fp->f_flag & FREAD) == 0) { 2265 fd_putfile(fd); 2266 return (EBADF); 2267 } 2268 2269 vp = fp->f_data; 2270 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2271 error = ESPIPE; 2272 goto out; 2273 } 2274 2275 offset = SCARG(uap, offset); 2276 2277 /* 2278 * XXX This works because no file systems actually 2279 * XXX take any action on the seek operation. 2280 */ 2281 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2282 goto out; 2283 2284 /* dofileread() will unuse the descriptor for us */ 2285 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2286 &offset, 0, retval)); 2287 2288 out: 2289 fd_putfile(fd); 2290 return (error); 2291 } 2292 2293 /* 2294 * Positional scatter read system call. 2295 */ 2296 int 2297 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2298 { 2299 /* { 2300 syscallarg(int) fd; 2301 syscallarg(const struct iovec *) iovp; 2302 syscallarg(int) iovcnt; 2303 syscallarg(off_t) offset; 2304 } */ 2305 off_t offset = SCARG(uap, offset); 2306 2307 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2308 SCARG(uap, iovcnt), &offset, 0, retval); 2309 } 2310 2311 /* 2312 * Positional write system call. 2313 */ 2314 int 2315 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2316 { 2317 /* { 2318 syscallarg(int) fd; 2319 syscallarg(const void *) buf; 2320 syscallarg(size_t) nbyte; 2321 syscallarg(off_t) offset; 2322 } */ 2323 file_t *fp; 2324 struct vnode *vp; 2325 off_t offset; 2326 int error, fd = SCARG(uap, fd); 2327 2328 if ((fp = fd_getfile(fd)) == NULL) 2329 return (EBADF); 2330 2331 if ((fp->f_flag & FWRITE) == 0) { 2332 fd_putfile(fd); 2333 return (EBADF); 2334 } 2335 2336 vp = fp->f_data; 2337 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2338 error = ESPIPE; 2339 goto out; 2340 } 2341 2342 offset = SCARG(uap, offset); 2343 2344 /* 2345 * XXX This works because no file systems actually 2346 * XXX take any action on the seek operation. 2347 */ 2348 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2349 goto out; 2350 2351 /* dofilewrite() will unuse the descriptor for us */ 2352 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2353 &offset, 0, retval)); 2354 2355 out: 2356 fd_putfile(fd); 2357 return (error); 2358 } 2359 2360 /* 2361 * Positional gather write system call. 2362 */ 2363 int 2364 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2365 { 2366 /* { 2367 syscallarg(int) fd; 2368 syscallarg(const struct iovec *) iovp; 2369 syscallarg(int) iovcnt; 2370 syscallarg(off_t) offset; 2371 } */ 2372 off_t offset = SCARG(uap, offset); 2373 2374 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2375 SCARG(uap, iovcnt), &offset, 0, retval); 2376 } 2377 2378 /* 2379 * Check access permissions. 2380 */ 2381 int 2382 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2383 { 2384 /* { 2385 syscallarg(const char *) path; 2386 syscallarg(int) flags; 2387 } */ 2388 kauth_cred_t cred; 2389 struct vnode *vp; 2390 int error, flags; 2391 struct nameidata nd; 2392 2393 cred = kauth_cred_dup(l->l_cred); 2394 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2395 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2396 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2397 SCARG(uap, path)); 2398 /* Override default credentials */ 2399 nd.ni_cnd.cn_cred = cred; 2400 if ((error = namei(&nd)) != 0) 2401 goto out; 2402 vp = nd.ni_vp; 2403 2404 /* Flags == 0 means only check for existence. */ 2405 if (SCARG(uap, flags)) { 2406 flags = 0; 2407 if (SCARG(uap, flags) & R_OK) 2408 flags |= VREAD; 2409 if (SCARG(uap, flags) & W_OK) 2410 flags |= VWRITE; 2411 if (SCARG(uap, flags) & X_OK) 2412 flags |= VEXEC; 2413 2414 error = VOP_ACCESS(vp, flags, cred); 2415 if (!error && (flags & VWRITE)) 2416 error = vn_writechk(vp); 2417 } 2418 vput(vp); 2419 out: 2420 kauth_cred_free(cred); 2421 return (error); 2422 } 2423 2424 /* 2425 * Common code for all sys_stat functions, including compat versions. 2426 */ 2427 int 2428 do_sys_stat(const char *path, unsigned int nd_flags, struct stat *sb) 2429 { 2430 int error; 2431 struct nameidata nd; 2432 2433 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2434 UIO_USERSPACE, path); 2435 error = namei(&nd); 2436 if (error != 0) 2437 return error; 2438 error = vn_stat(nd.ni_vp, sb); 2439 vput(nd.ni_vp); 2440 return error; 2441 } 2442 2443 /* 2444 * Get file status; this version follows links. 2445 */ 2446 /* ARGSUSED */ 2447 int 2448 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2449 { 2450 /* { 2451 syscallarg(const char *) path; 2452 syscallarg(struct stat *) ub; 2453 } */ 2454 struct stat sb; 2455 int error; 2456 2457 error = do_sys_stat(SCARG(uap, path), FOLLOW, &sb); 2458 if (error) 2459 return error; 2460 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2461 } 2462 2463 /* 2464 * Get file status; this version does not follow links. 2465 */ 2466 /* ARGSUSED */ 2467 int 2468 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2469 { 2470 /* { 2471 syscallarg(const char *) path; 2472 syscallarg(struct stat *) ub; 2473 } */ 2474 struct stat sb; 2475 int error; 2476 2477 error = do_sys_stat(SCARG(uap, path), NOFOLLOW, &sb); 2478 if (error) 2479 return error; 2480 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2481 } 2482 2483 /* 2484 * Get configurable pathname variables. 2485 */ 2486 /* ARGSUSED */ 2487 int 2488 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2489 { 2490 /* { 2491 syscallarg(const char *) path; 2492 syscallarg(int) name; 2493 } */ 2494 int error; 2495 struct nameidata nd; 2496 2497 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2498 SCARG(uap, path)); 2499 if ((error = namei(&nd)) != 0) 2500 return (error); 2501 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2502 vput(nd.ni_vp); 2503 return (error); 2504 } 2505 2506 /* 2507 * Return target name of a symbolic link. 2508 */ 2509 /* ARGSUSED */ 2510 int 2511 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2512 { 2513 /* { 2514 syscallarg(const char *) path; 2515 syscallarg(char *) buf; 2516 syscallarg(size_t) count; 2517 } */ 2518 struct vnode *vp; 2519 struct iovec aiov; 2520 struct uio auio; 2521 int error; 2522 struct nameidata nd; 2523 2524 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2525 SCARG(uap, path)); 2526 if ((error = namei(&nd)) != 0) 2527 return (error); 2528 vp = nd.ni_vp; 2529 if (vp->v_type != VLNK) 2530 error = EINVAL; 2531 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2532 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2533 aiov.iov_base = SCARG(uap, buf); 2534 aiov.iov_len = SCARG(uap, count); 2535 auio.uio_iov = &aiov; 2536 auio.uio_iovcnt = 1; 2537 auio.uio_offset = 0; 2538 auio.uio_rw = UIO_READ; 2539 KASSERT(l == curlwp); 2540 auio.uio_vmspace = l->l_proc->p_vmspace; 2541 auio.uio_resid = SCARG(uap, count); 2542 error = VOP_READLINK(vp, &auio, l->l_cred); 2543 } 2544 vput(vp); 2545 *retval = SCARG(uap, count) - auio.uio_resid; 2546 return (error); 2547 } 2548 2549 /* 2550 * Change flags of a file given a path name. 2551 */ 2552 /* ARGSUSED */ 2553 int 2554 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2555 { 2556 /* { 2557 syscallarg(const char *) path; 2558 syscallarg(u_long) flags; 2559 } */ 2560 struct vnode *vp; 2561 int error; 2562 struct nameidata nd; 2563 2564 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2565 SCARG(uap, path)); 2566 if ((error = namei(&nd)) != 0) 2567 return (error); 2568 vp = nd.ni_vp; 2569 error = change_flags(vp, SCARG(uap, flags), l); 2570 vput(vp); 2571 return (error); 2572 } 2573 2574 /* 2575 * Change flags of a file given a file descriptor. 2576 */ 2577 /* ARGSUSED */ 2578 int 2579 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2580 { 2581 /* { 2582 syscallarg(int) fd; 2583 syscallarg(u_long) flags; 2584 } */ 2585 struct vnode *vp; 2586 file_t *fp; 2587 int error; 2588 2589 /* fd_getvnode() will use the descriptor for us */ 2590 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2591 return (error); 2592 vp = fp->f_data; 2593 error = change_flags(vp, SCARG(uap, flags), l); 2594 VOP_UNLOCK(vp, 0); 2595 fd_putfile(SCARG(uap, fd)); 2596 return (error); 2597 } 2598 2599 /* 2600 * Change flags of a file given a path name; this version does 2601 * not follow links. 2602 */ 2603 int 2604 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2605 { 2606 /* { 2607 syscallarg(const char *) path; 2608 syscallarg(u_long) flags; 2609 } */ 2610 struct vnode *vp; 2611 int error; 2612 struct nameidata nd; 2613 2614 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2615 SCARG(uap, path)); 2616 if ((error = namei(&nd)) != 0) 2617 return (error); 2618 vp = nd.ni_vp; 2619 error = change_flags(vp, SCARG(uap, flags), l); 2620 vput(vp); 2621 return (error); 2622 } 2623 2624 /* 2625 * Common routine to change flags of a file. 2626 */ 2627 int 2628 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2629 { 2630 struct vattr vattr; 2631 int error; 2632 2633 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2634 /* 2635 * Non-superusers cannot change the flags on devices, even if they 2636 * own them. 2637 */ 2638 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2639 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2640 goto out; 2641 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2642 error = EINVAL; 2643 goto out; 2644 } 2645 } 2646 VATTR_NULL(&vattr); 2647 vattr.va_flags = flags; 2648 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2649 out: 2650 return (error); 2651 } 2652 2653 /* 2654 * Change mode of a file given path name; this version follows links. 2655 */ 2656 /* ARGSUSED */ 2657 int 2658 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2659 { 2660 /* { 2661 syscallarg(const char *) path; 2662 syscallarg(int) mode; 2663 } */ 2664 int error; 2665 struct nameidata nd; 2666 2667 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2668 SCARG(uap, path)); 2669 if ((error = namei(&nd)) != 0) 2670 return (error); 2671 2672 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2673 2674 vrele(nd.ni_vp); 2675 return (error); 2676 } 2677 2678 /* 2679 * Change mode of a file given a file descriptor. 2680 */ 2681 /* ARGSUSED */ 2682 int 2683 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2684 { 2685 /* { 2686 syscallarg(int) fd; 2687 syscallarg(int) mode; 2688 } */ 2689 file_t *fp; 2690 int error; 2691 2692 /* fd_getvnode() will use the descriptor for us */ 2693 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2694 return (error); 2695 error = change_mode(fp->f_data, SCARG(uap, mode), l); 2696 fd_putfile(SCARG(uap, fd)); 2697 return (error); 2698 } 2699 2700 /* 2701 * Change mode of a file given path name; this version does not follow links. 2702 */ 2703 /* ARGSUSED */ 2704 int 2705 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2706 { 2707 /* { 2708 syscallarg(const char *) path; 2709 syscallarg(int) mode; 2710 } */ 2711 int error; 2712 struct nameidata nd; 2713 2714 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2715 SCARG(uap, path)); 2716 if ((error = namei(&nd)) != 0) 2717 return (error); 2718 2719 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2720 2721 vrele(nd.ni_vp); 2722 return (error); 2723 } 2724 2725 /* 2726 * Common routine to set mode given a vnode. 2727 */ 2728 static int 2729 change_mode(struct vnode *vp, int mode, struct lwp *l) 2730 { 2731 struct vattr vattr; 2732 int error; 2733 2734 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2735 VATTR_NULL(&vattr); 2736 vattr.va_mode = mode & ALLPERMS; 2737 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2738 VOP_UNLOCK(vp, 0); 2739 return (error); 2740 } 2741 2742 /* 2743 * Set ownership given a path name; this version follows links. 2744 */ 2745 /* ARGSUSED */ 2746 int 2747 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2748 { 2749 /* { 2750 syscallarg(const char *) path; 2751 syscallarg(uid_t) uid; 2752 syscallarg(gid_t) gid; 2753 } */ 2754 int error; 2755 struct nameidata nd; 2756 2757 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2758 SCARG(uap, path)); 2759 if ((error = namei(&nd)) != 0) 2760 return (error); 2761 2762 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2763 2764 vrele(nd.ni_vp); 2765 return (error); 2766 } 2767 2768 /* 2769 * Set ownership given a path name; this version follows links. 2770 * Provides POSIX semantics. 2771 */ 2772 /* ARGSUSED */ 2773 int 2774 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2775 { 2776 /* { 2777 syscallarg(const char *) path; 2778 syscallarg(uid_t) uid; 2779 syscallarg(gid_t) gid; 2780 } */ 2781 int error; 2782 struct nameidata nd; 2783 2784 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2785 SCARG(uap, path)); 2786 if ((error = namei(&nd)) != 0) 2787 return (error); 2788 2789 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2790 2791 vrele(nd.ni_vp); 2792 return (error); 2793 } 2794 2795 /* 2796 * Set ownership given a file descriptor. 2797 */ 2798 /* ARGSUSED */ 2799 int 2800 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2801 { 2802 /* { 2803 syscallarg(int) fd; 2804 syscallarg(uid_t) uid; 2805 syscallarg(gid_t) gid; 2806 } */ 2807 int error; 2808 file_t *fp; 2809 2810 /* fd_getvnode() will use the descriptor for us */ 2811 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2812 return (error); 2813 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2814 l, 0); 2815 fd_putfile(SCARG(uap, fd)); 2816 return (error); 2817 } 2818 2819 /* 2820 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2821 */ 2822 /* ARGSUSED */ 2823 int 2824 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2825 { 2826 /* { 2827 syscallarg(int) fd; 2828 syscallarg(uid_t) uid; 2829 syscallarg(gid_t) gid; 2830 } */ 2831 int error; 2832 file_t *fp; 2833 2834 /* fd_getvnode() will use the descriptor for us */ 2835 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2836 return (error); 2837 error = change_owner(fp->f_data, SCARG(uap, uid), SCARG(uap, gid), 2838 l, 1); 2839 fd_putfile(SCARG(uap, fd)); 2840 return (error); 2841 } 2842 2843 /* 2844 * Set ownership given a path name; this version does not follow links. 2845 */ 2846 /* ARGSUSED */ 2847 int 2848 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2849 { 2850 /* { 2851 syscallarg(const char *) path; 2852 syscallarg(uid_t) uid; 2853 syscallarg(gid_t) gid; 2854 } */ 2855 int error; 2856 struct nameidata nd; 2857 2858 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2859 SCARG(uap, path)); 2860 if ((error = namei(&nd)) != 0) 2861 return (error); 2862 2863 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2864 2865 vrele(nd.ni_vp); 2866 return (error); 2867 } 2868 2869 /* 2870 * Set ownership given a path name; this version does not follow links. 2871 * Provides POSIX/XPG semantics. 2872 */ 2873 /* ARGSUSED */ 2874 int 2875 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2876 { 2877 /* { 2878 syscallarg(const char *) path; 2879 syscallarg(uid_t) uid; 2880 syscallarg(gid_t) gid; 2881 } */ 2882 int error; 2883 struct nameidata nd; 2884 2885 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2886 SCARG(uap, path)); 2887 if ((error = namei(&nd)) != 0) 2888 return (error); 2889 2890 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2891 2892 vrele(nd.ni_vp); 2893 return (error); 2894 } 2895 2896 /* 2897 * Common routine to set ownership given a vnode. 2898 */ 2899 static int 2900 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2901 int posix_semantics) 2902 { 2903 struct vattr vattr; 2904 mode_t newmode; 2905 int error; 2906 2907 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2908 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2909 goto out; 2910 2911 #define CHANGED(x) ((int)(x) != -1) 2912 newmode = vattr.va_mode; 2913 if (posix_semantics) { 2914 /* 2915 * POSIX/XPG semantics: if the caller is not the super-user, 2916 * clear set-user-id and set-group-id bits. Both POSIX and 2917 * the XPG consider the behaviour for calls by the super-user 2918 * implementation-defined; we leave the set-user-id and set- 2919 * group-id settings intact in that case. 2920 */ 2921 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2922 NULL) != 0) 2923 newmode &= ~(S_ISUID | S_ISGID); 2924 } else { 2925 /* 2926 * NetBSD semantics: when changing owner and/or group, 2927 * clear the respective bit(s). 2928 */ 2929 if (CHANGED(uid)) 2930 newmode &= ~S_ISUID; 2931 if (CHANGED(gid)) 2932 newmode &= ~S_ISGID; 2933 } 2934 /* Update va_mode iff altered. */ 2935 if (vattr.va_mode == newmode) 2936 newmode = VNOVAL; 2937 2938 VATTR_NULL(&vattr); 2939 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2940 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2941 vattr.va_mode = newmode; 2942 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2943 #undef CHANGED 2944 2945 out: 2946 VOP_UNLOCK(vp, 0); 2947 return (error); 2948 } 2949 2950 /* 2951 * Set the access and modification times given a path name; this 2952 * version follows links. 2953 */ 2954 /* ARGSUSED */ 2955 int 2956 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2957 { 2958 /* { 2959 syscallarg(const char *) path; 2960 syscallarg(const struct timeval *) tptr; 2961 } */ 2962 2963 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2964 SCARG(uap, tptr), UIO_USERSPACE); 2965 } 2966 2967 /* 2968 * Set the access and modification times given a file descriptor. 2969 */ 2970 /* ARGSUSED */ 2971 int 2972 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2973 { 2974 /* { 2975 syscallarg(int) fd; 2976 syscallarg(const struct timeval *) tptr; 2977 } */ 2978 int error; 2979 file_t *fp; 2980 2981 /* fd_getvnode() will use the descriptor for us */ 2982 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 2983 return (error); 2984 error = do_sys_utimes(l, fp->f_data, NULL, 0, SCARG(uap, tptr), 2985 UIO_USERSPACE); 2986 fd_putfile(SCARG(uap, fd)); 2987 return (error); 2988 } 2989 2990 /* 2991 * Set the access and modification times given a path name; this 2992 * version does not follow links. 2993 */ 2994 int 2995 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2996 { 2997 /* { 2998 syscallarg(const char *) path; 2999 syscallarg(const struct timeval *) tptr; 3000 } */ 3001 3002 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3003 SCARG(uap, tptr), UIO_USERSPACE); 3004 } 3005 3006 /* 3007 * Common routine to set access and modification times given a vnode. 3008 */ 3009 int 3010 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3011 const struct timeval *tptr, enum uio_seg seg) 3012 { 3013 struct vattr vattr; 3014 struct nameidata nd; 3015 int error; 3016 bool vanull, setbirthtime; 3017 struct timespec ts[2]; 3018 3019 if (tptr == NULL) { 3020 vanull = true; 3021 nanotime(&ts[0]); 3022 ts[1] = ts[0]; 3023 } else { 3024 struct timeval tv[2]; 3025 3026 vanull = false; 3027 if (seg != UIO_SYSSPACE) { 3028 error = copyin(tptr, &tv, sizeof (tv)); 3029 if (error != 0) 3030 return error; 3031 tptr = tv; 3032 } 3033 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 3034 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 3035 } 3036 3037 if (vp == NULL) { 3038 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3039 if ((error = namei(&nd)) != 0) 3040 return error; 3041 vp = nd.ni_vp; 3042 } else 3043 nd.ni_vp = NULL; 3044 3045 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3046 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 3047 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 3048 VATTR_NULL(&vattr); 3049 vattr.va_atime = ts[0]; 3050 vattr.va_mtime = ts[1]; 3051 if (setbirthtime) 3052 vattr.va_birthtime = ts[1]; 3053 if (vanull) 3054 vattr.va_flags |= VA_UTIMES_NULL; 3055 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3056 VOP_UNLOCK(vp, 0); 3057 3058 if (nd.ni_vp != NULL) 3059 vrele(nd.ni_vp); 3060 3061 return error; 3062 } 3063 3064 /* 3065 * Truncate a file given its path name. 3066 */ 3067 /* ARGSUSED */ 3068 int 3069 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3070 { 3071 /* { 3072 syscallarg(const char *) path; 3073 syscallarg(int) pad; 3074 syscallarg(off_t) length; 3075 } */ 3076 struct vnode *vp; 3077 struct vattr vattr; 3078 int error; 3079 struct nameidata nd; 3080 3081 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3082 SCARG(uap, path)); 3083 if ((error = namei(&nd)) != 0) 3084 return (error); 3085 vp = nd.ni_vp; 3086 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3087 if (vp->v_type == VDIR) 3088 error = EISDIR; 3089 else if ((error = vn_writechk(vp)) == 0 && 3090 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3091 VATTR_NULL(&vattr); 3092 vattr.va_size = SCARG(uap, length); 3093 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3094 } 3095 vput(vp); 3096 return (error); 3097 } 3098 3099 /* 3100 * Truncate a file given a file descriptor. 3101 */ 3102 /* ARGSUSED */ 3103 int 3104 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3105 { 3106 /* { 3107 syscallarg(int) fd; 3108 syscallarg(int) pad; 3109 syscallarg(off_t) length; 3110 } */ 3111 struct vattr vattr; 3112 struct vnode *vp; 3113 file_t *fp; 3114 int error; 3115 3116 /* fd_getvnode() will use the descriptor for us */ 3117 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3118 return (error); 3119 if ((fp->f_flag & FWRITE) == 0) { 3120 error = EINVAL; 3121 goto out; 3122 } 3123 vp = fp->f_data; 3124 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3125 if (vp->v_type == VDIR) 3126 error = EISDIR; 3127 else if ((error = vn_writechk(vp)) == 0) { 3128 VATTR_NULL(&vattr); 3129 vattr.va_size = SCARG(uap, length); 3130 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3131 } 3132 VOP_UNLOCK(vp, 0); 3133 out: 3134 fd_putfile(SCARG(uap, fd)); 3135 return (error); 3136 } 3137 3138 /* 3139 * Sync an open file. 3140 */ 3141 /* ARGSUSED */ 3142 int 3143 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3144 { 3145 /* { 3146 syscallarg(int) fd; 3147 } */ 3148 struct vnode *vp; 3149 file_t *fp; 3150 int error; 3151 3152 /* fd_getvnode() will use the descriptor for us */ 3153 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3154 return (error); 3155 vp = fp->f_data; 3156 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3157 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3158 if (error == 0 && bioopsp != NULL && 3159 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3160 (*bioopsp->io_fsync)(vp, 0); 3161 VOP_UNLOCK(vp, 0); 3162 fd_putfile(SCARG(uap, fd)); 3163 return (error); 3164 } 3165 3166 /* 3167 * Sync a range of file data. API modeled after that found in AIX. 3168 * 3169 * FDATASYNC indicates that we need only save enough metadata to be able 3170 * to re-read the written data. Note we duplicate AIX's requirement that 3171 * the file be open for writing. 3172 */ 3173 /* ARGSUSED */ 3174 int 3175 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3176 { 3177 /* { 3178 syscallarg(int) fd; 3179 syscallarg(int) flags; 3180 syscallarg(off_t) start; 3181 syscallarg(off_t) length; 3182 } */ 3183 struct vnode *vp; 3184 file_t *fp; 3185 int flags, nflags; 3186 off_t s, e, len; 3187 int error; 3188 3189 /* fd_getvnode() will use the descriptor for us */ 3190 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3191 return (error); 3192 3193 if ((fp->f_flag & FWRITE) == 0) { 3194 error = EBADF; 3195 goto out; 3196 } 3197 3198 flags = SCARG(uap, flags); 3199 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3200 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3201 error = EINVAL; 3202 goto out; 3203 } 3204 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3205 if (flags & FDATASYNC) 3206 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3207 else 3208 nflags = FSYNC_WAIT; 3209 if (flags & FDISKSYNC) 3210 nflags |= FSYNC_CACHE; 3211 3212 len = SCARG(uap, length); 3213 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3214 if (len) { 3215 s = SCARG(uap, start); 3216 e = s + len; 3217 if (e < s) { 3218 error = EINVAL; 3219 goto out; 3220 } 3221 } else { 3222 e = 0; 3223 s = 0; 3224 } 3225 3226 vp = fp->f_data; 3227 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3228 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3229 3230 if (error == 0 && bioopsp != NULL && 3231 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3232 (*bioopsp->io_fsync)(vp, nflags); 3233 3234 VOP_UNLOCK(vp, 0); 3235 out: 3236 fd_putfile(SCARG(uap, fd)); 3237 return (error); 3238 } 3239 3240 /* 3241 * Sync the data of an open file. 3242 */ 3243 /* ARGSUSED */ 3244 int 3245 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3246 { 3247 /* { 3248 syscallarg(int) fd; 3249 } */ 3250 struct vnode *vp; 3251 file_t *fp; 3252 int error; 3253 3254 /* fd_getvnode() will use the descriptor for us */ 3255 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3256 return (error); 3257 if ((fp->f_flag & FWRITE) == 0) { 3258 fd_putfile(SCARG(uap, fd)); 3259 return (EBADF); 3260 } 3261 vp = fp->f_data; 3262 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3263 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3264 VOP_UNLOCK(vp, 0); 3265 fd_putfile(SCARG(uap, fd)); 3266 return (error); 3267 } 3268 3269 /* 3270 * Rename files, (standard) BSD semantics frontend. 3271 */ 3272 /* ARGSUSED */ 3273 int 3274 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3275 { 3276 /* { 3277 syscallarg(const char *) from; 3278 syscallarg(const char *) to; 3279 } */ 3280 3281 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3282 } 3283 3284 /* 3285 * Rename files, POSIX semantics frontend. 3286 */ 3287 /* ARGSUSED */ 3288 int 3289 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3290 { 3291 /* { 3292 syscallarg(const char *) from; 3293 syscallarg(const char *) to; 3294 } */ 3295 3296 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3297 } 3298 3299 /* 3300 * Rename files. Source and destination must either both be directories, 3301 * or both not be directories. If target is a directory, it must be empty. 3302 * If `from' and `to' refer to the same object, the value of the `retain' 3303 * argument is used to determine whether `from' will be 3304 * 3305 * (retain == 0) deleted unless `from' and `to' refer to the same 3306 * object in the file system's name space (BSD). 3307 * (retain == 1) always retained (POSIX). 3308 */ 3309 int 3310 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3311 { 3312 struct vnode *tvp, *fvp, *tdvp; 3313 struct nameidata fromnd, tond; 3314 struct mount *fs; 3315 struct lwp *l = curlwp; 3316 struct proc *p; 3317 uint32_t saveflag; 3318 int error; 3319 3320 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3321 seg, from); 3322 if ((error = namei(&fromnd)) != 0) 3323 return (error); 3324 if (fromnd.ni_dvp != fromnd.ni_vp) 3325 VOP_UNLOCK(fromnd.ni_dvp, 0); 3326 fvp = fromnd.ni_vp; 3327 3328 fs = fvp->v_mount; 3329 error = VFS_RENAMELOCK_ENTER(fs); 3330 if (error) { 3331 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3332 vrele(fromnd.ni_dvp); 3333 vrele(fvp); 3334 goto out1; 3335 } 3336 3337 /* 3338 * close, partially, yet another race - ideally we should only 3339 * go as far as getting fromnd.ni_dvp before getting the per-fs 3340 * lock, and then continue to get fromnd.ni_vp, but we can't do 3341 * that with namei as it stands. 3342 * 3343 * This still won't prevent rmdir from nuking fromnd.ni_vp 3344 * under us. The real fix is to get the locks in the right 3345 * order and do the lookups in the right places, but that's a 3346 * major rototill. 3347 * 3348 * Preserve the SAVESTART in cn_flags, because who knows what 3349 * might happen if we don't. 3350 * 3351 * Note: this logic (as well as this whole function) is cloned 3352 * in nfs_serv.c. Proceed accordingly. 3353 */ 3354 vrele(fvp); 3355 if ((fromnd.ni_cnd.cn_namelen == 1 && 3356 fromnd.ni_cnd.cn_nameptr[0] == '.') || 3357 (fromnd.ni_cnd.cn_namelen == 2 && 3358 fromnd.ni_cnd.cn_nameptr[0] == '.' && 3359 fromnd.ni_cnd.cn_nameptr[1] == '.')) { 3360 error = EINVAL; 3361 VFS_RENAMELOCK_EXIT(fs); 3362 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3363 vrele(fromnd.ni_dvp); 3364 goto out1; 3365 } 3366 saveflag = fromnd.ni_cnd.cn_flags & SAVESTART; 3367 fromnd.ni_cnd.cn_flags &= ~SAVESTART; 3368 vn_lock(fromnd.ni_dvp, LK_EXCLUSIVE | LK_RETRY); 3369 error = relookup(fromnd.ni_dvp, &fromnd.ni_vp, &fromnd.ni_cnd); 3370 fromnd.ni_cnd.cn_flags |= saveflag; 3371 if (error) { 3372 VOP_UNLOCK(fromnd.ni_dvp, 0); 3373 VFS_RENAMELOCK_EXIT(fs); 3374 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3375 vrele(fromnd.ni_dvp); 3376 goto out1; 3377 } 3378 VOP_UNLOCK(fromnd.ni_vp, 0); 3379 if (fromnd.ni_dvp != fromnd.ni_vp) 3380 VOP_UNLOCK(fromnd.ni_dvp, 0); 3381 fvp = fromnd.ni_vp; 3382 3383 NDINIT(&tond, RENAME, 3384 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3385 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3386 seg, to); 3387 if ((error = namei(&tond)) != 0) { 3388 VFS_RENAMELOCK_EXIT(fs); 3389 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3390 vrele(fromnd.ni_dvp); 3391 vrele(fvp); 3392 goto out1; 3393 } 3394 tdvp = tond.ni_dvp; 3395 tvp = tond.ni_vp; 3396 3397 if (tvp != NULL) { 3398 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3399 error = ENOTDIR; 3400 goto out; 3401 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3402 error = EISDIR; 3403 goto out; 3404 } 3405 } 3406 3407 if (fvp == tdvp) 3408 error = EINVAL; 3409 3410 /* 3411 * Source and destination refer to the same object. 3412 */ 3413 if (fvp == tvp) { 3414 if (retain) 3415 error = -1; 3416 else if (fromnd.ni_dvp == tdvp && 3417 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3418 !memcmp(fromnd.ni_cnd.cn_nameptr, 3419 tond.ni_cnd.cn_nameptr, 3420 fromnd.ni_cnd.cn_namelen)) 3421 error = -1; 3422 } 3423 3424 #if NVERIEXEC > 0 3425 if (!error) { 3426 char *f1, *f2; 3427 3428 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3429 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3430 3431 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3432 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3433 3434 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3435 3436 free(f1, M_TEMP); 3437 free(f2, M_TEMP); 3438 } 3439 #endif /* NVERIEXEC > 0 */ 3440 3441 out: 3442 p = l->l_proc; 3443 if (!error) { 3444 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3445 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3446 VFS_RENAMELOCK_EXIT(fs); 3447 } else { 3448 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3449 if (tdvp == tvp) 3450 vrele(tdvp); 3451 else 3452 vput(tdvp); 3453 if (tvp) 3454 vput(tvp); 3455 VFS_RENAMELOCK_EXIT(fs); 3456 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3457 vrele(fromnd.ni_dvp); 3458 vrele(fvp); 3459 } 3460 vrele(tond.ni_startdir); 3461 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3462 out1: 3463 if (fromnd.ni_startdir) 3464 vrele(fromnd.ni_startdir); 3465 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3466 return (error == -1 ? 0 : error); 3467 } 3468 3469 /* 3470 * Make a directory file. 3471 */ 3472 /* ARGSUSED */ 3473 int 3474 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3475 { 3476 /* { 3477 syscallarg(const char *) path; 3478 syscallarg(int) mode; 3479 } */ 3480 struct proc *p = l->l_proc; 3481 struct vnode *vp; 3482 struct vattr vattr; 3483 int error; 3484 struct nameidata nd; 3485 3486 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3487 SCARG(uap, path)); 3488 if ((error = namei(&nd)) != 0) 3489 return (error); 3490 vp = nd.ni_vp; 3491 if (vp != NULL) { 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 vrele(vp); 3498 return (EEXIST); 3499 } 3500 VATTR_NULL(&vattr); 3501 vattr.va_type = VDIR; 3502 /* We will read cwdi->cwdi_cmask unlocked. */ 3503 vattr.va_mode = 3504 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3505 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3506 if (!error) 3507 vput(nd.ni_vp); 3508 return (error); 3509 } 3510 3511 /* 3512 * Remove a directory file. 3513 */ 3514 /* ARGSUSED */ 3515 int 3516 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3517 { 3518 /* { 3519 syscallarg(const char *) path; 3520 } */ 3521 struct vnode *vp; 3522 int error; 3523 struct nameidata nd; 3524 3525 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3526 SCARG(uap, path)); 3527 if ((error = namei(&nd)) != 0) 3528 return (error); 3529 vp = nd.ni_vp; 3530 if (vp->v_type != VDIR) { 3531 error = ENOTDIR; 3532 goto out; 3533 } 3534 /* 3535 * No rmdir "." please. 3536 */ 3537 if (nd.ni_dvp == vp) { 3538 error = EINVAL; 3539 goto out; 3540 } 3541 /* 3542 * The root of a mounted filesystem cannot be deleted. 3543 */ 3544 if ((vp->v_vflag & VV_ROOT) != 0 || vp->v_mountedhere != NULL) { 3545 error = EBUSY; 3546 goto out; 3547 } 3548 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3549 return (error); 3550 3551 out: 3552 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3553 if (nd.ni_dvp == vp) 3554 vrele(nd.ni_dvp); 3555 else 3556 vput(nd.ni_dvp); 3557 vput(vp); 3558 return (error); 3559 } 3560 3561 /* 3562 * Read a block of directory entries in a file system independent format. 3563 */ 3564 int 3565 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3566 { 3567 /* { 3568 syscallarg(int) fd; 3569 syscallarg(char *) buf; 3570 syscallarg(size_t) count; 3571 } */ 3572 file_t *fp; 3573 int error, done; 3574 3575 /* fd_getvnode() will use the descriptor for us */ 3576 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3577 return (error); 3578 if ((fp->f_flag & FREAD) == 0) { 3579 error = EBADF; 3580 goto out; 3581 } 3582 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3583 SCARG(uap, count), &done, l, 0, 0); 3584 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3585 *retval = done; 3586 out: 3587 fd_putfile(SCARG(uap, fd)); 3588 return (error); 3589 } 3590 3591 /* 3592 * Set the mode mask for creation of filesystem nodes. 3593 */ 3594 int 3595 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3596 { 3597 /* { 3598 syscallarg(mode_t) newmask; 3599 } */ 3600 struct proc *p = l->l_proc; 3601 struct cwdinfo *cwdi; 3602 3603 /* 3604 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3605 * important is that we serialize changes to the mask. The 3606 * rw_exit() will issue a write memory barrier on our behalf, 3607 * and force the changes out to other CPUs (as it must use an 3608 * atomic operation, draining the local CPU's store buffers). 3609 */ 3610 cwdi = p->p_cwdi; 3611 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3612 *retval = cwdi->cwdi_cmask; 3613 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3614 rw_exit(&cwdi->cwdi_lock); 3615 3616 return (0); 3617 } 3618 3619 int 3620 dorevoke(struct vnode *vp, kauth_cred_t cred) 3621 { 3622 struct vattr vattr; 3623 int error; 3624 3625 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3626 return error; 3627 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3628 (error = kauth_authorize_generic(cred, 3629 KAUTH_GENERIC_ISSUSER, NULL)) == 0) 3630 VOP_REVOKE(vp, REVOKEALL); 3631 return (error); 3632 } 3633 3634 /* 3635 * Void all references to file by ripping underlying filesystem 3636 * away from vnode. 3637 */ 3638 /* ARGSUSED */ 3639 int 3640 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3641 { 3642 /* { 3643 syscallarg(const char *) path; 3644 } */ 3645 struct vnode *vp; 3646 int error; 3647 struct nameidata nd; 3648 3649 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3650 SCARG(uap, path)); 3651 if ((error = namei(&nd)) != 0) 3652 return (error); 3653 vp = nd.ni_vp; 3654 error = dorevoke(vp, l->l_cred); 3655 vrele(vp); 3656 return (error); 3657 } 3658
Indexes created Thu Sep 25 16:09:42 GMT 2025