vfs_syscalls.c revision 1.340
1 /* $NetBSD: vfs_syscalls.c,v 1.340 2008/01/09 08:18:12 elad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.340 2008/01/09 08:18:12 elad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "fss.h" 46 #include "veriexec.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/kmem.h> 61 #include <sys/dirent.h> 62 #include <sys/sysctl.h> 63 #include <sys/syscallargs.h> 64 #include <sys/vfs_syscalls.h> 65 #include <sys/ktrace.h> 66 #ifdef FILEASSOC 67 #include <sys/fileassoc.h> 68 #endif /* FILEASSOC */ 69 #include <sys/verified_exec.h> 70 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #endif 79 #include <nfs/nfsproto.h> 80 #ifdef COMPAT_30 81 #include <nfs/nfs.h> 82 #include <nfs/nfs_var.h> 83 #endif 84 85 #if NFSS > 0 86 #include <dev/fssvar.h> 87 #endif 88 89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 90 91 static int change_dir(struct nameidata *, struct lwp *); 92 static int change_flags(struct vnode *, u_long, struct lwp *); 93 static int change_mode(struct vnode *, int, struct lwp *l); 94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 95 96 void checkdirs(struct vnode *); 97 98 int dovfsusermount = 0; 99 100 /* 101 * Virtual File System System Calls 102 */ 103 104 /* 105 * Mount a file system. 106 */ 107 108 #if defined(COMPAT_09) || defined(COMPAT_43) 109 /* 110 * This table is used to maintain compatibility with 4.3BSD 111 * and NetBSD 0.9 mount syscalls. Note, the order is important! 112 * 113 * Do not modify this table. It should only contain filesystems 114 * supported by NetBSD 0.9 and 4.3BSD. 115 */ 116 const char * const mountcompatnames[] = { 117 NULL, /* 0 = MOUNT_NONE */ 118 MOUNT_FFS, /* 1 = MOUNT_UFS */ 119 MOUNT_NFS, /* 2 */ 120 MOUNT_MFS, /* 3 */ 121 MOUNT_MSDOS, /* 4 */ 122 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 123 MOUNT_FDESC, /* 6 */ 124 MOUNT_KERNFS, /* 7 */ 125 NULL, /* 8 = MOUNT_DEVFS */ 126 MOUNT_AFS, /* 9 */ 127 }; 128 const int nmountcompatnames = sizeof(mountcompatnames) / 129 sizeof(mountcompatnames[0]); 130 #endif /* COMPAT_09 || COMPAT_43 */ 131 132 static int 133 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 134 void *data, size_t *data_len) 135 { 136 struct mount *mp; 137 int error = 0, saved_flags; 138 139 mp = vp->v_mount; 140 saved_flags = mp->mnt_flag; 141 142 /* We can operate only on VV_ROOT nodes. */ 143 if ((vp->v_vflag & VV_ROOT) == 0) { 144 error = EINVAL; 145 goto out; 146 } 147 148 /* 149 * We only allow the filesystem to be reloaded if it 150 * is currently mounted read-only. 151 */ 152 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 153 error = EOPNOTSUPP; /* Needs translation */ 154 goto out; 155 } 156 157 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 158 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 159 if (error) 160 goto out; 161 162 if (vfs_busy(mp, LK_NOWAIT, 0)) { 163 error = EPERM; 164 goto out; 165 } 166 167 mp->mnt_flag &= ~MNT_OP_FLAGS; 168 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 169 170 /* 171 * Set the mount level flags. 172 */ 173 if (flags & MNT_RDONLY) 174 mp->mnt_flag |= MNT_RDONLY; 175 else if (mp->mnt_flag & MNT_RDONLY) 176 mp->mnt_iflag |= IMNT_WANTRDWR; 177 mp->mnt_flag &= 178 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 179 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 180 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 181 mp->mnt_flag |= flags & 182 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 183 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 184 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 185 MNT_IGNORE); 186 187 error = VFS_MOUNT(mp, path, data, data_len); 188 189 #if defined(COMPAT_30) && defined(NFSSERVER) 190 if (error && data != NULL) { 191 int error2; 192 193 /* Update failed; let's try and see if it was an 194 * export request. */ 195 error2 = nfs_update_exports_30(mp, path, data, l); 196 197 /* Only update error code if the export request was 198 * understood but some problem occurred while 199 * processing it. */ 200 if (error2 != EJUSTRETURN) 201 error = error2; 202 } 203 #endif 204 if (mp->mnt_iflag & IMNT_WANTRDWR) 205 mp->mnt_flag &= ~MNT_RDONLY; 206 if (error) 207 mp->mnt_flag = saved_flags; 208 mp->mnt_flag &= ~MNT_OP_FLAGS; 209 mp->mnt_iflag &= ~IMNT_WANTRDWR; 210 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 211 if (mp->mnt_syncer == NULL) 212 error = vfs_allocate_syncvnode(mp); 213 } else { 214 if (mp->mnt_syncer != NULL) 215 vfs_deallocate_syncvnode(mp); 216 } 217 vfs_unbusy(mp); 218 219 out: 220 return (error); 221 } 222 223 static int 224 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 225 { 226 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 227 int error; 228 229 /* Copy file-system type from userspace. */ 230 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 231 if (error) { 232 #if defined(COMPAT_09) || defined(COMPAT_43) 233 /* 234 * Historically, filesystem types were identified by numbers. 235 * If we get an integer for the filesystem type instead of a 236 * string, we check to see if it matches one of the historic 237 * filesystem types. 238 */ 239 u_long fsindex = (u_long)fstype; 240 if (fsindex >= nmountcompatnames || 241 mountcompatnames[fsindex] == NULL) 242 return ENODEV; 243 strlcpy(fstypename, mountcompatnames[fsindex], 244 sizeof(fstypename)); 245 #else 246 return error; 247 #endif 248 } 249 250 #ifdef COMPAT_10 251 /* Accept `ufs' as an alias for `ffs'. */ 252 if (strcmp(fstypename, "ufs") == 0) 253 fstypename[0] = 'f'; 254 #endif 255 256 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 257 return ENODEV; 258 return 0; 259 } 260 261 static int 262 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 263 const char *path, int flags, void *data, size_t *data_len, u_int recurse) 264 { 265 struct mount *mp = NULL; 266 struct vnode *vp = *vpp; 267 struct vattr va; 268 int error; 269 270 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 271 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 272 if (error) 273 return error; 274 275 /* Can't make a non-dir a mount-point (from here anyway). */ 276 if (vp->v_type != VDIR) 277 return ENOTDIR; 278 279 /* 280 * If the user is not root, ensure that they own the directory 281 * onto which we are attempting to mount. 282 */ 283 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 284 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 285 (error = kauth_authorize_generic(l->l_cred, 286 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 287 return error; 288 } 289 290 if (flags & MNT_EXPORTED) 291 return EINVAL; 292 293 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 294 return error; 295 296 /* 297 * Check if a file-system is not already mounted on this vnode. 298 */ 299 if (vp->v_mountedhere != NULL) 300 return EBUSY; 301 302 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 303 304 mp->mnt_op = vfsops; 305 306 TAILQ_INIT(&mp->mnt_vnodelist); 307 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 308 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 309 (void)vfs_busy(mp, LK_NOWAIT, 0); 310 311 mp->mnt_vnodecovered = vp; 312 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 313 mp->mnt_unmounter = NULL; 314 mount_initspecific(mp); 315 316 /* 317 * The underlying file system may refuse the mount for 318 * various reasons. Allow the user to force it to happen. 319 * 320 * Set the mount level flags. 321 */ 322 mp->mnt_flag = flags & 323 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 324 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 325 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 326 MNT_IGNORE | MNT_RDONLY); 327 328 error = VFS_MOUNT(mp, path, data, data_len); 329 mp->mnt_flag &= ~MNT_OP_FLAGS; 330 331 /* 332 * Put the new filesystem on the mount list after root. 333 */ 334 cache_purge(vp); 335 if (error != 0) { 336 vp->v_mountedhere = NULL; 337 mp->mnt_op->vfs_refcount--; 338 vfs_unbusy(mp); 339 vfs_destroy(mp); 340 return error; 341 } 342 343 mp->mnt_iflag &= ~IMNT_WANTRDWR; 344 vp->v_mountedhere = mp; 345 mutex_enter(&mountlist_lock); 346 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 347 mutex_exit(&mountlist_lock); 348 vn_restorerecurse(vp, recurse); 349 VOP_UNLOCK(vp, 0); 350 checkdirs(vp); 351 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 352 error = vfs_allocate_syncvnode(mp); 353 vfs_unbusy(mp); 354 (void) VFS_STATVFS(mp, &mp->mnt_stat); 355 error = VFS_START(mp, 0); 356 if (error) 357 vrele(vp); 358 *vpp = NULL; 359 return error; 360 } 361 362 static int 363 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 364 void *data, size_t *data_len) 365 { 366 struct mount *mp; 367 int error; 368 369 /* If MNT_GETARGS is specified, it should be the only flag. */ 370 if (flags & ~MNT_GETARGS) 371 return EINVAL; 372 373 mp = vp->v_mount; 374 375 /* XXX: probably some notion of "can see" here if we want isolation. */ 376 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 377 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 378 if (error) 379 return error; 380 381 if ((vp->v_vflag & VV_ROOT) == 0) 382 return EINVAL; 383 384 if (vfs_busy(mp, LK_NOWAIT, 0)) 385 return EPERM; 386 387 mp->mnt_flag &= ~MNT_OP_FLAGS; 388 mp->mnt_flag |= MNT_GETARGS; 389 error = VFS_MOUNT(mp, path, data, data_len); 390 mp->mnt_flag &= ~MNT_OP_FLAGS; 391 392 vfs_unbusy(mp); 393 return (error); 394 } 395 396 #ifdef COMPAT_40 397 /* ARGSUSED */ 398 int 399 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 400 { 401 /* { 402 syscallarg(const char *) type; 403 syscallarg(const char *) path; 404 syscallarg(int) flags; 405 syscallarg(void *) data; 406 } */ 407 register_t dummy; 408 409 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 410 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 411 } 412 #endif 413 414 int 415 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 416 { 417 /* { 418 syscallarg(const char *) type; 419 syscallarg(const char *) path; 420 syscallarg(int) flags; 421 syscallarg(void *) data; 422 syscallarg(size_t) data_len; 423 } */ 424 425 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 426 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 427 SCARG(uap, data_len), retval); 428 } 429 430 int 431 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 432 const char *path, int flags, void *data, enum uio_seg data_seg, 433 size_t data_len, register_t *retval) 434 { 435 struct vnode *vp; 436 struct nameidata nd; 437 void *data_buf = data; 438 u_int recurse; 439 int error; 440 441 /* 442 * Get vnode to be covered 443 */ 444 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 445 if ((error = namei(&nd)) != 0) 446 return (error); 447 vp = nd.ni_vp; 448 449 /* 450 * A lookup in VFS_MOUNT might result in an attempt to 451 * lock this vnode again, so make the lock recursive. 452 */ 453 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 454 recurse = vn_setrecurse(vp); 455 456 if (vfsops == NULL) { 457 if (flags & (MNT_GETARGS | MNT_UPDATE)) 458 vfsops = vp->v_mount->mnt_op; 459 else { 460 /* 'type' is userspace */ 461 error = mount_get_vfsops(type, &vfsops); 462 if (error != 0) 463 goto done; 464 } 465 } 466 467 if (data != NULL && data_seg == UIO_USERSPACE) { 468 if (data_len == 0) { 469 /* No length supplied, use default for filesystem */ 470 data_len = vfsops->vfs_min_mount_data; 471 if (data_len > VFS_MAX_MOUNT_DATA) { 472 /* maybe a force loaded old LKM */ 473 error = EINVAL; 474 goto done; 475 } 476 #ifdef COMPAT_30 477 /* Hopefully a longer buffer won't make copyin() fail */ 478 if (flags & MNT_UPDATE 479 && data_len < sizeof (struct mnt_export_args30)) 480 data_len = sizeof (struct mnt_export_args30); 481 #endif 482 } 483 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 484 485 /* NFS needs the buffer even for mnt_getargs .... */ 486 error = copyin(data, data_buf, data_len); 487 if (error != 0) 488 goto done; 489 } 490 491 if (flags & MNT_GETARGS) { 492 if (data_len == 0) { 493 error = EINVAL; 494 goto done; 495 } 496 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 497 if (error != 0) 498 goto done; 499 if (data_seg == UIO_USERSPACE) 500 error = copyout(data_buf, data, data_len); 501 *retval = data_len; 502 } else if (flags & MNT_UPDATE) { 503 error = mount_update(l, vp, path, flags, data_buf, &data_len); 504 } else { 505 /* Locking is handled internally in mount_domount(). */ 506 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 507 &data_len, recurse); 508 } 509 510 done: 511 if (vp != NULL) { 512 vn_restorerecurse(vp, recurse); 513 vput(vp); 514 } 515 if (data_buf != data) 516 free(data_buf, M_TEMP); 517 return (error); 518 } 519 520 /* 521 * Scan all active processes to see if any of them have a current 522 * or root directory onto which the new filesystem has just been 523 * mounted. If so, replace them with the new mount point. 524 */ 525 void 526 checkdirs(struct vnode *olddp) 527 { 528 struct cwdinfo *cwdi; 529 struct vnode *newdp; 530 struct proc *p; 531 532 if (olddp->v_usecount == 1) 533 return; 534 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 535 panic("mount: lost mount"); 536 mutex_enter(&proclist_lock); 537 PROCLIST_FOREACH(p, &allproc) { 538 cwdi = p->p_cwdi; 539 if (!cwdi) 540 continue; 541 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 542 if (cwdi->cwdi_cdir == olddp) { 543 vrele(cwdi->cwdi_cdir); 544 VREF(newdp); 545 cwdi->cwdi_cdir = newdp; 546 } 547 if (cwdi->cwdi_rdir == olddp) { 548 vrele(cwdi->cwdi_rdir); 549 VREF(newdp); 550 cwdi->cwdi_rdir = newdp; 551 } 552 rw_exit(&cwdi->cwdi_lock); 553 } 554 mutex_exit(&proclist_lock); 555 if (rootvnode == olddp) { 556 vrele(rootvnode); 557 VREF(newdp); 558 rootvnode = newdp; 559 } 560 vput(newdp); 561 } 562 563 /* 564 * Unmount a file system. 565 * 566 * Note: unmount takes a path to the vnode mounted on as argument, 567 * not special file (as before). 568 */ 569 /* ARGSUSED */ 570 int 571 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 572 { 573 /* { 574 syscallarg(const char *) path; 575 syscallarg(int) flags; 576 } */ 577 struct vnode *vp; 578 struct mount *mp; 579 int error; 580 struct nameidata nd; 581 582 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 583 SCARG(uap, path)); 584 if ((error = namei(&nd)) != 0) 585 return (error); 586 vp = nd.ni_vp; 587 mp = vp->v_mount; 588 589 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 590 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 591 if (error) { 592 vput(vp); 593 return (error); 594 } 595 596 /* 597 * Don't allow unmounting the root file system. 598 */ 599 if (mp->mnt_flag & MNT_ROOTFS) { 600 vput(vp); 601 return (EINVAL); 602 } 603 604 /* 605 * Must be the root of the filesystem 606 */ 607 if ((vp->v_vflag & VV_ROOT) == 0) { 608 vput(vp); 609 return (EINVAL); 610 } 611 vput(vp); 612 613 /* 614 * XXX Freeze syncer. Must do this before locking the 615 * mount point. See dounmount() for details. 616 */ 617 mutex_enter(&syncer_mutex); 618 619 if (vfs_busy(mp, 0, 0)) { 620 mutex_exit(&syncer_mutex); 621 return (EBUSY); 622 } 623 624 return (dounmount(mp, SCARG(uap, flags), l)); 625 } 626 627 /* 628 * Do the actual file system unmount. File system is assumed to have been 629 * marked busy by the caller. 630 */ 631 int 632 dounmount(struct mount *mp, int flags, struct lwp *l) 633 { 634 struct vnode *coveredvp; 635 int error; 636 int async; 637 int used_syncer; 638 639 #if NVERIEXEC > 0 640 error = veriexec_unmountchk(mp); 641 if (error) 642 return (error); 643 #endif /* NVERIEXEC > 0 */ 644 645 mutex_enter(&mountlist_lock); 646 vfs_unbusy(mp); 647 used_syncer = (mp->mnt_syncer != NULL); 648 649 /* 650 * XXX Syncer must be frozen when we get here. This should really 651 * be done on a per-mountpoint basis, but especially the softdep 652 * code possibly called from the syncer doesn't exactly work on a 653 * per-mountpoint basis, so the softdep code would become a maze 654 * of vfs_busy() calls. 655 * 656 * The caller of dounmount() must acquire syncer_mutex because 657 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 658 * order, and we must preserve that order to avoid deadlock. 659 * 660 * So, if the file system did not use the syncer, now is 661 * the time to release the syncer_mutex. 662 */ 663 if (used_syncer == 0) 664 mutex_exit(&syncer_mutex); 665 666 mp->mnt_iflag |= IMNT_UNMOUNT; 667 mp->mnt_unmounter = l; 668 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 669 670 async = mp->mnt_flag & MNT_ASYNC; 671 mp->mnt_flag &= ~MNT_ASYNC; 672 cache_purgevfs(mp); /* remove cache entries for this file sys */ 673 if (mp->mnt_syncer != NULL) 674 vfs_deallocate_syncvnode(mp); 675 error = 0; 676 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 677 #if NFSS > 0 678 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 679 #endif 680 if (error == 0) 681 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 682 } 683 if (error == 0 || (flags & MNT_FORCE)) 684 error = VFS_UNMOUNT(mp, flags); 685 if (error) { 686 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 687 (void) vfs_allocate_syncvnode(mp); 688 mutex_enter(&mountlist_lock); 689 mp->mnt_iflag &= ~IMNT_UNMOUNT; 690 mp->mnt_unmounter = NULL; 691 mp->mnt_flag |= async; 692 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 693 &mountlist_lock); 694 if (used_syncer) 695 mutex_exit(&syncer_mutex); 696 mutex_enter(&mp->mnt_mutex); 697 while (mp->mnt_wcnt > 0) { 698 wakeup(mp); 699 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 700 0, &mp->mnt_mutex); 701 } 702 mutex_exit(&mp->mnt_mutex); 703 return (error); 704 } 705 vfs_scrubvnlist(mp); 706 mutex_enter(&mountlist_lock); 707 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 708 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 709 coveredvp->v_mountedhere = NULL; 710 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 711 panic("unmount: dangling vnode"); 712 mp->mnt_iflag |= IMNT_GONE; 713 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 714 if (coveredvp != NULLVP) 715 vrele(coveredvp); 716 if (used_syncer) 717 mutex_exit(&syncer_mutex); 718 mutex_enter(&mp->mnt_mutex); 719 while (mp->mnt_wcnt > 0) { 720 wakeup(mp); 721 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 722 } 723 mutex_exit(&mp->mnt_mutex); 724 vfs_hooks_unmount(mp); 725 vfs_delref(mp->mnt_op); 726 vfs_destroy(mp); 727 return (0); 728 } 729 730 /* 731 * Sync each mounted filesystem. 732 */ 733 #ifdef DEBUG 734 int syncprt = 0; 735 struct ctldebug debug0 = { "syncprt", &syncprt }; 736 #endif 737 738 /* ARGSUSED */ 739 int 740 sys_sync(struct lwp *l, const void *v, register_t *retval) 741 { 742 struct mount *mp, *nmp; 743 int asyncflag; 744 745 if (l == NULL) 746 l = &lwp0; 747 748 mutex_enter(&mountlist_lock); 749 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 750 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 751 nmp = mp->mnt_list.cqe_prev; 752 continue; 753 } 754 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 755 asyncflag = mp->mnt_flag & MNT_ASYNC; 756 mp->mnt_flag &= ~MNT_ASYNC; 757 /* XXXSMP hack, sync is slow. */ 758 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 759 KERNEL_LOCK(1, NULL); 760 } 761 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 762 if ((mp->mnt_iflag & IMNT_MPSAFE) == 0) { 763 KERNEL_UNLOCK_ONE(NULL); 764 } 765 if (asyncflag) 766 mp->mnt_flag |= MNT_ASYNC; 767 } 768 mutex_enter(&mountlist_lock); 769 nmp = mp->mnt_list.cqe_prev; 770 vfs_unbusy(mp); 771 772 } 773 mutex_exit(&mountlist_lock); 774 #ifdef DEBUG 775 if (syncprt) 776 vfs_bufstats(); 777 #endif /* DEBUG */ 778 return (0); 779 } 780 781 /* 782 * Change filesystem quotas. 783 */ 784 /* ARGSUSED */ 785 int 786 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 787 { 788 /* { 789 syscallarg(const char *) path; 790 syscallarg(int) cmd; 791 syscallarg(int) uid; 792 syscallarg(void *) arg; 793 } */ 794 struct mount *mp; 795 int error; 796 struct nameidata nd; 797 798 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 799 SCARG(uap, path)); 800 if ((error = namei(&nd)) != 0) 801 return (error); 802 mp = nd.ni_vp->v_mount; 803 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 804 SCARG(uap, arg)); 805 vrele(nd.ni_vp); 806 return (error); 807 } 808 809 int 810 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 811 int root) 812 { 813 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 814 int error = 0; 815 816 /* 817 * If MNT_NOWAIT or MNT_LAZY is specified, do not 818 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 819 * overrides MNT_NOWAIT. 820 */ 821 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 822 (flags != MNT_WAIT && flags != 0)) { 823 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 824 goto done; 825 } 826 827 /* Get the filesystem stats now */ 828 memset(sp, 0, sizeof(*sp)); 829 if ((error = VFS_STATVFS(mp, sp)) != 0) { 830 return error; 831 } 832 833 if (cwdi->cwdi_rdir == NULL) 834 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 835 done: 836 if (cwdi->cwdi_rdir != NULL) { 837 size_t len; 838 char *bp; 839 char *path = PNBUF_GET(); 840 841 bp = path + MAXPATHLEN; 842 *--bp = '\0'; 843 rw_enter(&cwdi->cwdi_lock, RW_READER); 844 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 845 MAXPATHLEN / 2, 0, l); 846 rw_exit(&cwdi->cwdi_lock); 847 if (error) { 848 PNBUF_PUT(path); 849 return error; 850 } 851 len = strlen(bp); 852 /* 853 * for mount points that are below our root, we can see 854 * them, so we fix up the pathname and return them. The 855 * rest we cannot see, so we don't allow viewing the 856 * data. 857 */ 858 if (strncmp(bp, sp->f_mntonname, len) == 0) { 859 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 860 sizeof(sp->f_mntonname)); 861 if (sp->f_mntonname[0] == '\0') 862 (void)strlcpy(sp->f_mntonname, "/", 863 sizeof(sp->f_mntonname)); 864 } else { 865 if (root) 866 (void)strlcpy(sp->f_mntonname, "/", 867 sizeof(sp->f_mntonname)); 868 else 869 error = EPERM; 870 } 871 PNBUF_PUT(path); 872 } 873 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 874 return error; 875 } 876 877 /* 878 * Get filesystem statistics by path. 879 */ 880 int 881 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 882 { 883 struct mount *mp; 884 int error; 885 struct nameidata nd; 886 887 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 888 if ((error = namei(&nd)) != 0) 889 return error; 890 mp = nd.ni_vp->v_mount; 891 error = dostatvfs(mp, sb, l, flags, 1); 892 vrele(nd.ni_vp); 893 return error; 894 } 895 896 /* ARGSUSED */ 897 int 898 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 899 { 900 /* { 901 syscallarg(const char *) path; 902 syscallarg(struct statvfs *) buf; 903 syscallarg(int) flags; 904 } */ 905 struct statvfs *sb; 906 int error; 907 908 sb = STATVFSBUF_GET(); 909 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 910 if (error == 0) 911 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 912 STATVFSBUF_PUT(sb); 913 return error; 914 } 915 916 /* 917 * Get filesystem statistics by fd. 918 */ 919 int 920 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 921 { 922 struct proc *p = l->l_proc; 923 struct file *fp; 924 struct mount *mp; 925 int error; 926 927 /* getvnode() will use the descriptor for us */ 928 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 929 return (error); 930 mp = ((struct vnode *)fp->f_data)->v_mount; 931 error = dostatvfs(mp, sb, l, flags, 1); 932 FILE_UNUSE(fp, l); 933 return error; 934 } 935 936 /* ARGSUSED */ 937 int 938 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 939 { 940 /* { 941 syscallarg(int) fd; 942 syscallarg(struct statvfs *) buf; 943 syscallarg(int) flags; 944 } */ 945 struct statvfs *sb; 946 int error; 947 948 sb = STATVFSBUF_GET(); 949 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 950 if (error == 0) 951 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 952 STATVFSBUF_PUT(sb); 953 return error; 954 } 955 956 957 /* 958 * Get statistics on all filesystems. 959 */ 960 int 961 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 962 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 963 register_t *retval) 964 { 965 int root = 0; 966 struct proc *p = l->l_proc; 967 struct mount *mp, *nmp; 968 struct statvfs *sb; 969 size_t count, maxcount; 970 int error = 0; 971 972 sb = STATVFSBUF_GET(); 973 maxcount = bufsize / entry_sz; 974 mutex_enter(&mountlist_lock); 975 count = 0; 976 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 977 mp = nmp) { 978 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 979 nmp = CIRCLEQ_NEXT(mp, mnt_list); 980 continue; 981 } 982 if (sfsp && count < maxcount) { 983 error = dostatvfs(mp, sb, l, flags, 0); 984 if (error) { 985 mutex_enter(&mountlist_lock); 986 nmp = CIRCLEQ_NEXT(mp, mnt_list); 987 vfs_unbusy(mp); 988 continue; 989 } 990 error = copyfn(sb, sfsp, entry_sz); 991 if (error) { 992 vfs_unbusy(mp); 993 goto out; 994 } 995 sfsp = (char *)sfsp + entry_sz; 996 root |= strcmp(sb->f_mntonname, "/") == 0; 997 } 998 count++; 999 mutex_enter(&mountlist_lock); 1000 nmp = CIRCLEQ_NEXT(mp, mnt_list); 1001 vfs_unbusy(mp); 1002 } 1003 1004 mutex_exit(&mountlist_lock); 1005 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1006 /* 1007 * fake a root entry 1008 */ 1009 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1010 sb, l, flags, 1); 1011 if (error != 0) 1012 goto out; 1013 if (sfsp) 1014 error = copyfn(sb, sfsp, entry_sz); 1015 count++; 1016 } 1017 if (sfsp && count > maxcount) 1018 *retval = maxcount; 1019 else 1020 *retval = count; 1021 out: 1022 STATVFSBUF_PUT(sb); 1023 return error; 1024 } 1025 1026 int 1027 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1028 { 1029 /* { 1030 syscallarg(struct statvfs *) buf; 1031 syscallarg(size_t) bufsize; 1032 syscallarg(int) flags; 1033 } */ 1034 1035 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1036 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1037 } 1038 1039 /* 1040 * Change current working directory to a given file descriptor. 1041 */ 1042 /* ARGSUSED */ 1043 int 1044 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1045 { 1046 /* { 1047 syscallarg(int) fd; 1048 } */ 1049 struct proc *p = l->l_proc; 1050 struct filedesc *fdp = p->p_fd; 1051 struct cwdinfo *cwdi; 1052 struct vnode *vp, *tdp; 1053 struct mount *mp; 1054 struct file *fp; 1055 int error; 1056 1057 /* getvnode() will use the descriptor for us */ 1058 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1059 return (error); 1060 vp = (struct vnode *)fp->f_data; 1061 1062 VREF(vp); 1063 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1064 if (vp->v_type != VDIR) 1065 error = ENOTDIR; 1066 else 1067 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1068 if (error) { 1069 vput(vp); 1070 goto out; 1071 } 1072 while ((mp = vp->v_mountedhere) != NULL) { 1073 if (vfs_busy(mp, 0, 0)) 1074 continue; 1075 1076 vput(vp); 1077 error = VFS_ROOT(mp, &tdp); 1078 vfs_unbusy(mp); 1079 if (error) 1080 goto out; 1081 vp = tdp; 1082 } 1083 VOP_UNLOCK(vp, 0); 1084 1085 /* 1086 * Disallow changing to a directory not under the process's 1087 * current root directory (if there is one). 1088 */ 1089 cwdi = p->p_cwdi; 1090 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1091 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1092 vrele(vp); 1093 error = EPERM; /* operation not permitted */ 1094 } else { 1095 vrele(cwdi->cwdi_cdir); 1096 cwdi->cwdi_cdir = vp; 1097 } 1098 rw_exit(&cwdi->cwdi_lock); 1099 1100 out: 1101 FILE_UNUSE(fp, l); 1102 return (error); 1103 } 1104 1105 /* 1106 * Change this process's notion of the root directory to a given file 1107 * descriptor. 1108 */ 1109 int 1110 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1111 { 1112 struct proc *p = l->l_proc; 1113 struct filedesc *fdp = p->p_fd; 1114 struct cwdinfo *cwdi; 1115 struct vnode *vp; 1116 struct file *fp; 1117 int error; 1118 1119 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1120 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1121 return error; 1122 /* getvnode() will use the descriptor for us */ 1123 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1124 return error; 1125 vp = (struct vnode *) fp->f_data; 1126 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1127 if (vp->v_type != VDIR) 1128 error = ENOTDIR; 1129 else 1130 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1131 VOP_UNLOCK(vp, 0); 1132 if (error) 1133 goto out; 1134 VREF(vp); 1135 1136 /* 1137 * Prevent escaping from chroot by putting the root under 1138 * the working directory. Silently chdir to / if we aren't 1139 * already there. 1140 */ 1141 cwdi = p->p_cwdi; 1142 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1143 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1144 /* 1145 * XXX would be more failsafe to change directory to a 1146 * deadfs node here instead 1147 */ 1148 vrele(cwdi->cwdi_cdir); 1149 VREF(vp); 1150 cwdi->cwdi_cdir = vp; 1151 } 1152 1153 if (cwdi->cwdi_rdir != NULL) 1154 vrele(cwdi->cwdi_rdir); 1155 cwdi->cwdi_rdir = vp; 1156 rw_exit(&cwdi->cwdi_lock); 1157 1158 out: 1159 FILE_UNUSE(fp, l); 1160 return (error); 1161 } 1162 1163 /* 1164 * Change current working directory (``.''). 1165 */ 1166 /* ARGSUSED */ 1167 int 1168 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1169 { 1170 /* { 1171 syscallarg(const char *) path; 1172 } */ 1173 struct proc *p = l->l_proc; 1174 struct cwdinfo *cwdi; 1175 int error; 1176 struct nameidata nd; 1177 1178 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1179 SCARG(uap, path)); 1180 if ((error = change_dir(&nd, l)) != 0) 1181 return (error); 1182 cwdi = p->p_cwdi; 1183 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1184 vrele(cwdi->cwdi_cdir); 1185 cwdi->cwdi_cdir = nd.ni_vp; 1186 rw_exit(&cwdi->cwdi_lock); 1187 return (0); 1188 } 1189 1190 /* 1191 * Change notion of root (``/'') directory. 1192 */ 1193 /* ARGSUSED */ 1194 int 1195 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1196 { 1197 /* { 1198 syscallarg(const char *) path; 1199 } */ 1200 struct proc *p = l->l_proc; 1201 struct cwdinfo *cwdi; 1202 struct vnode *vp; 1203 int error; 1204 struct nameidata nd; 1205 1206 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1207 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1208 return (error); 1209 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1210 SCARG(uap, path)); 1211 if ((error = change_dir(&nd, l)) != 0) 1212 return (error); 1213 1214 cwdi = p->p_cwdi; 1215 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1216 if (cwdi->cwdi_rdir != NULL) 1217 vrele(cwdi->cwdi_rdir); 1218 vp = nd.ni_vp; 1219 cwdi->cwdi_rdir = vp; 1220 1221 /* 1222 * Prevent escaping from chroot by putting the root under 1223 * the working directory. Silently chdir to / if we aren't 1224 * already there. 1225 */ 1226 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1227 /* 1228 * XXX would be more failsafe to change directory to a 1229 * deadfs node here instead 1230 */ 1231 vrele(cwdi->cwdi_cdir); 1232 VREF(vp); 1233 cwdi->cwdi_cdir = vp; 1234 } 1235 rw_exit(&cwdi->cwdi_lock); 1236 1237 return (0); 1238 } 1239 1240 /* 1241 * Common routine for chroot and chdir. 1242 */ 1243 static int 1244 change_dir(struct nameidata *ndp, struct lwp *l) 1245 { 1246 struct vnode *vp; 1247 int error; 1248 1249 if ((error = namei(ndp)) != 0) 1250 return (error); 1251 vp = ndp->ni_vp; 1252 if (vp->v_type != VDIR) 1253 error = ENOTDIR; 1254 else 1255 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1256 1257 if (error) 1258 vput(vp); 1259 else 1260 VOP_UNLOCK(vp, 0); 1261 return (error); 1262 } 1263 1264 /* 1265 * Check permissions, allocate an open file structure, 1266 * and call the device open routine if any. 1267 */ 1268 int 1269 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1270 { 1271 /* { 1272 syscallarg(const char *) path; 1273 syscallarg(int) flags; 1274 syscallarg(int) mode; 1275 } */ 1276 struct proc *p = l->l_proc; 1277 struct cwdinfo *cwdi = p->p_cwdi; 1278 struct filedesc *fdp = p->p_fd; 1279 struct file *fp; 1280 struct vnode *vp; 1281 int flags, cmode; 1282 int type, indx, error; 1283 struct flock lf; 1284 struct nameidata nd; 1285 1286 flags = FFLAGS(SCARG(uap, flags)); 1287 if ((flags & (FREAD | FWRITE)) == 0) 1288 return (EINVAL); 1289 /* falloc() will use the file descriptor for us */ 1290 if ((error = falloc(l, &fp, &indx)) != 0) 1291 return (error); 1292 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1293 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1294 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1295 SCARG(uap, path)); 1296 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1297 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1298 rw_enter(&fdp->fd_lock, RW_WRITER); 1299 FILE_UNUSE(fp, l); 1300 fdp->fd_ofiles[indx] = NULL; 1301 rw_exit(&fdp->fd_lock); 1302 ffree(fp); 1303 if ((error == EDUPFD || error == EMOVEFD) && 1304 l->l_dupfd >= 0 && /* XXX from fdopen */ 1305 (error = 1306 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1307 *retval = indx; 1308 return (0); 1309 } 1310 if (error == ERESTART) 1311 error = EINTR; 1312 fdremove(fdp, indx); 1313 return (error); 1314 } 1315 1316 l->l_dupfd = 0; 1317 vp = nd.ni_vp; 1318 fp->f_flag = flags & FMASK; 1319 fp->f_type = DTYPE_VNODE; 1320 fp->f_ops = &vnops; 1321 fp->f_data = vp; 1322 if (flags & (O_EXLOCK | O_SHLOCK)) { 1323 lf.l_whence = SEEK_SET; 1324 lf.l_start = 0; 1325 lf.l_len = 0; 1326 if (flags & O_EXLOCK) 1327 lf.l_type = F_WRLCK; 1328 else 1329 lf.l_type = F_RDLCK; 1330 type = F_FLOCK; 1331 if ((flags & FNONBLOCK) == 0) 1332 type |= F_WAIT; 1333 VOP_UNLOCK(vp, 0); 1334 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1335 if (error) { 1336 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1337 FILE_UNUSE(fp, l); 1338 ffree(fp); 1339 fdremove(fdp, indx); 1340 return (error); 1341 } 1342 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1343 fp->f_flag |= FHASLOCK; 1344 } 1345 VOP_UNLOCK(vp, 0); 1346 *retval = indx; 1347 FILE_SET_MATURE(fp); 1348 FILE_UNUSE(fp, l); 1349 return (0); 1350 } 1351 1352 static void 1353 vfs__fhfree(fhandle_t *fhp) 1354 { 1355 size_t fhsize; 1356 1357 if (fhp == NULL) { 1358 return; 1359 } 1360 fhsize = FHANDLE_SIZE(fhp); 1361 kmem_free(fhp, fhsize); 1362 } 1363 1364 /* 1365 * vfs_composefh: compose a filehandle. 1366 */ 1367 1368 int 1369 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1370 { 1371 struct mount *mp; 1372 struct fid *fidp; 1373 int error; 1374 size_t needfhsize; 1375 size_t fidsize; 1376 1377 mp = vp->v_mount; 1378 fidp = NULL; 1379 if (*fh_size < FHANDLE_SIZE_MIN) { 1380 fidsize = 0; 1381 } else { 1382 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1383 if (fhp != NULL) { 1384 memset(fhp, 0, *fh_size); 1385 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1386 fidp = &fhp->fh_fid; 1387 } 1388 } 1389 error = VFS_VPTOFH(vp, fidp, &fidsize); 1390 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1391 if (error == 0 && *fh_size < needfhsize) { 1392 error = E2BIG; 1393 } 1394 *fh_size = needfhsize; 1395 return error; 1396 } 1397 1398 int 1399 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1400 { 1401 struct mount *mp; 1402 fhandle_t *fhp; 1403 size_t fhsize; 1404 size_t fidsize; 1405 int error; 1406 1407 *fhpp = NULL; 1408 mp = vp->v_mount; 1409 fidsize = 0; 1410 error = VFS_VPTOFH(vp, NULL, &fidsize); 1411 KASSERT(error != 0); 1412 if (error != E2BIG) { 1413 goto out; 1414 } 1415 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1416 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1417 if (fhp == NULL) { 1418 error = ENOMEM; 1419 goto out; 1420 } 1421 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1422 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1423 if (error == 0) { 1424 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1425 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1426 *fhpp = fhp; 1427 } else { 1428 kmem_free(fhp, fhsize); 1429 } 1430 out: 1431 return error; 1432 } 1433 1434 void 1435 vfs_composefh_free(fhandle_t *fhp) 1436 { 1437 1438 vfs__fhfree(fhp); 1439 } 1440 1441 /* 1442 * vfs_fhtovp: lookup a vnode by a filehandle. 1443 */ 1444 1445 int 1446 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1447 { 1448 struct mount *mp; 1449 int error; 1450 1451 *vpp = NULL; 1452 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1453 if (mp == NULL) { 1454 error = ESTALE; 1455 goto out; 1456 } 1457 if (mp->mnt_op->vfs_fhtovp == NULL) { 1458 error = EOPNOTSUPP; 1459 goto out; 1460 } 1461 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1462 out: 1463 return error; 1464 } 1465 1466 /* 1467 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1468 * the needed size. 1469 */ 1470 1471 int 1472 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1473 { 1474 fhandle_t *fhp; 1475 int error; 1476 1477 *fhpp = NULL; 1478 if (fhsize > FHANDLE_SIZE_MAX) { 1479 return EINVAL; 1480 } 1481 if (fhsize < FHANDLE_SIZE_MIN) { 1482 return EINVAL; 1483 } 1484 again: 1485 fhp = kmem_alloc(fhsize, KM_SLEEP); 1486 if (fhp == NULL) { 1487 return ENOMEM; 1488 } 1489 error = copyin(ufhp, fhp, fhsize); 1490 if (error == 0) { 1491 /* XXX this check shouldn't be here */ 1492 if (FHANDLE_SIZE(fhp) == fhsize) { 1493 *fhpp = fhp; 1494 return 0; 1495 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1496 /* 1497 * a kludge for nfsv2 padded handles. 1498 */ 1499 size_t sz; 1500 1501 sz = FHANDLE_SIZE(fhp); 1502 kmem_free(fhp, fhsize); 1503 fhsize = sz; 1504 goto again; 1505 } else { 1506 /* 1507 * userland told us wrong size. 1508 */ 1509 error = EINVAL; 1510 } 1511 } 1512 kmem_free(fhp, fhsize); 1513 return error; 1514 } 1515 1516 void 1517 vfs_copyinfh_free(fhandle_t *fhp) 1518 { 1519 1520 vfs__fhfree(fhp); 1521 } 1522 1523 /* 1524 * Get file handle system call 1525 */ 1526 int 1527 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1528 { 1529 /* { 1530 syscallarg(char *) fname; 1531 syscallarg(fhandle_t *) fhp; 1532 syscallarg(size_t *) fh_size; 1533 } */ 1534 struct vnode *vp; 1535 fhandle_t *fh; 1536 int error; 1537 struct nameidata nd; 1538 size_t sz; 1539 size_t usz; 1540 1541 /* 1542 * Must be super user 1543 */ 1544 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1545 0, NULL, NULL, NULL); 1546 if (error) 1547 return (error); 1548 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1549 SCARG(uap, fname)); 1550 error = namei(&nd); 1551 if (error) 1552 return (error); 1553 vp = nd.ni_vp; 1554 error = vfs_composefh_alloc(vp, &fh); 1555 vput(vp); 1556 if (error != 0) { 1557 goto out; 1558 } 1559 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1560 if (error != 0) { 1561 goto out; 1562 } 1563 sz = FHANDLE_SIZE(fh); 1564 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1565 if (error != 0) { 1566 goto out; 1567 } 1568 if (usz >= sz) { 1569 error = copyout(fh, SCARG(uap, fhp), sz); 1570 } else { 1571 error = E2BIG; 1572 } 1573 out: 1574 vfs_composefh_free(fh); 1575 return (error); 1576 } 1577 1578 /* 1579 * Open a file given a file handle. 1580 * 1581 * Check permissions, allocate an open file structure, 1582 * and call the device open routine if any. 1583 */ 1584 1585 int 1586 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1587 register_t *retval) 1588 { 1589 struct filedesc *fdp = l->l_proc->p_fd; 1590 struct file *fp; 1591 struct vnode *vp = NULL; 1592 kauth_cred_t cred = l->l_cred; 1593 struct file *nfp; 1594 int type, indx, error=0; 1595 struct flock lf; 1596 struct vattr va; 1597 fhandle_t *fh; 1598 int flags; 1599 1600 /* 1601 * Must be super user 1602 */ 1603 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1604 0, NULL, NULL, NULL))) 1605 return (error); 1606 1607 flags = FFLAGS(oflags); 1608 if ((flags & (FREAD | FWRITE)) == 0) 1609 return (EINVAL); 1610 if ((flags & O_CREAT)) 1611 return (EINVAL); 1612 /* falloc() will use the file descriptor for us */ 1613 if ((error = falloc(l, &nfp, &indx)) != 0) 1614 return (error); 1615 fp = nfp; 1616 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1617 if (error != 0) { 1618 goto bad; 1619 } 1620 error = vfs_fhtovp(fh, &vp); 1621 if (error != 0) { 1622 goto bad; 1623 } 1624 1625 /* Now do an effective vn_open */ 1626 1627 if (vp->v_type == VSOCK) { 1628 error = EOPNOTSUPP; 1629 goto bad; 1630 } 1631 error = vn_openchk(vp, cred, flags); 1632 if (error != 0) 1633 goto bad; 1634 if (flags & O_TRUNC) { 1635 VOP_UNLOCK(vp, 0); /* XXX */ 1636 VOP_LEASE(vp, cred, LEASE_WRITE); 1637 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1638 VATTR_NULL(&va); 1639 va.va_size = 0; 1640 error = VOP_SETATTR(vp, &va, cred); 1641 if (error) 1642 goto bad; 1643 } 1644 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1645 goto bad; 1646 if (flags & FWRITE) 1647 vp->v_writecount++; 1648 1649 /* done with modified vn_open, now finish what sys_open does. */ 1650 1651 fp->f_flag = flags & FMASK; 1652 fp->f_type = DTYPE_VNODE; 1653 fp->f_ops = &vnops; 1654 fp->f_data = vp; 1655 if (flags & (O_EXLOCK | O_SHLOCK)) { 1656 lf.l_whence = SEEK_SET; 1657 lf.l_start = 0; 1658 lf.l_len = 0; 1659 if (flags & O_EXLOCK) 1660 lf.l_type = F_WRLCK; 1661 else 1662 lf.l_type = F_RDLCK; 1663 type = F_FLOCK; 1664 if ((flags & FNONBLOCK) == 0) 1665 type |= F_WAIT; 1666 VOP_UNLOCK(vp, 0); 1667 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1668 if (error) { 1669 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1670 FILE_UNUSE(fp, l); 1671 ffree(fp); 1672 fdremove(fdp, indx); 1673 return (error); 1674 } 1675 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1676 fp->f_flag |= FHASLOCK; 1677 } 1678 VOP_UNLOCK(vp, 0); 1679 *retval = indx; 1680 FILE_SET_MATURE(fp); 1681 FILE_UNUSE(fp, l); 1682 vfs_copyinfh_free(fh); 1683 return (0); 1684 1685 bad: 1686 FILE_UNUSE(fp, l); 1687 ffree(fp); 1688 fdremove(fdp, indx); 1689 if (vp != NULL) 1690 vput(vp); 1691 vfs_copyinfh_free(fh); 1692 return (error); 1693 } 1694 1695 int 1696 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1697 { 1698 /* { 1699 syscallarg(const void *) fhp; 1700 syscallarg(size_t) fh_size; 1701 syscallarg(int) flags; 1702 } */ 1703 1704 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1705 SCARG(uap, flags), retval); 1706 } 1707 1708 int 1709 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1710 { 1711 int error; 1712 fhandle_t *fh; 1713 struct vnode *vp; 1714 1715 /* 1716 * Must be super user 1717 */ 1718 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1719 0, NULL, NULL, NULL))) 1720 return (error); 1721 1722 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1723 if (error != 0) 1724 return error; 1725 1726 error = vfs_fhtovp(fh, &vp); 1727 vfs_copyinfh_free(fh); 1728 if (error != 0) 1729 return error; 1730 1731 error = vn_stat(vp, sb, l); 1732 vput(vp); 1733 return error; 1734 } 1735 1736 1737 /* ARGSUSED */ 1738 int 1739 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1740 { 1741 /* { 1742 syscallarg(const void *) fhp; 1743 syscallarg(size_t) fh_size; 1744 syscallarg(struct stat *) sb; 1745 } */ 1746 struct stat sb; 1747 int error; 1748 1749 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1750 if (error) 1751 return error; 1752 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1753 } 1754 1755 int 1756 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1757 int flags) 1758 { 1759 fhandle_t *fh; 1760 struct mount *mp; 1761 struct vnode *vp; 1762 int error; 1763 1764 /* 1765 * Must be super user 1766 */ 1767 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1768 0, NULL, NULL, NULL))) 1769 return error; 1770 1771 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1772 if (error != 0) 1773 return error; 1774 1775 error = vfs_fhtovp(fh, &vp); 1776 vfs_copyinfh_free(fh); 1777 if (error != 0) 1778 return error; 1779 1780 mp = vp->v_mount; 1781 error = dostatvfs(mp, sb, l, flags, 1); 1782 vput(vp); 1783 return error; 1784 } 1785 1786 /* ARGSUSED */ 1787 int 1788 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1789 { 1790 /* { 1791 syscallarg(const void *) fhp; 1792 syscallarg(size_t) fh_size; 1793 syscallarg(struct statvfs *) buf; 1794 syscallarg(int) flags; 1795 } */ 1796 struct statvfs *sb = STATVFSBUF_GET(); 1797 int error; 1798 1799 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1800 SCARG(uap, flags)); 1801 if (error == 0) 1802 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1803 STATVFSBUF_PUT(sb); 1804 return error; 1805 } 1806 1807 /* 1808 * Create a special file. 1809 */ 1810 /* ARGSUSED */ 1811 int 1812 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1813 { 1814 /* { 1815 syscallarg(const char *) path; 1816 syscallarg(int) mode; 1817 syscallarg(int) dev; 1818 } */ 1819 struct proc *p = l->l_proc; 1820 struct vnode *vp; 1821 struct vattr vattr; 1822 int error, optype; 1823 struct nameidata nd; 1824 char *path; 1825 const char *cpath; 1826 enum uio_seg seg = UIO_USERSPACE; 1827 1828 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1829 0, NULL, NULL, NULL)) != 0) 1830 return (error); 1831 1832 optype = VOP_MKNOD_DESCOFFSET; 1833 1834 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1835 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1836 1837 if ((error = namei(&nd)) != 0) 1838 goto out; 1839 vp = nd.ni_vp; 1840 if (vp != NULL) 1841 error = EEXIST; 1842 else { 1843 VATTR_NULL(&vattr); 1844 /* We will read cwdi->cwdi_cmask unlocked. */ 1845 vattr.va_mode = 1846 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1847 vattr.va_rdev = SCARG(uap, dev); 1848 1849 switch (SCARG(uap, mode) & S_IFMT) { 1850 case S_IFMT: /* used by badsect to flag bad sectors */ 1851 vattr.va_type = VBAD; 1852 break; 1853 case S_IFCHR: 1854 vattr.va_type = VCHR; 1855 break; 1856 case S_IFBLK: 1857 vattr.va_type = VBLK; 1858 break; 1859 case S_IFWHT: 1860 optype = VOP_WHITEOUT_DESCOFFSET; 1861 break; 1862 case S_IFREG: 1863 #if NVERIEXEC > 0 1864 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1865 O_CREAT); 1866 #endif /* NVERIEXEC > 0 */ 1867 vattr.va_type = VREG; 1868 vattr.va_rdev = VNOVAL; 1869 optype = VOP_CREATE_DESCOFFSET; 1870 break; 1871 default: 1872 error = EINVAL; 1873 break; 1874 } 1875 } 1876 if (!error) { 1877 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1878 switch (optype) { 1879 case VOP_WHITEOUT_DESCOFFSET: 1880 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1881 if (error) 1882 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1883 vput(nd.ni_dvp); 1884 break; 1885 1886 case VOP_MKNOD_DESCOFFSET: 1887 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1888 &nd.ni_cnd, &vattr); 1889 if (error == 0) 1890 vput(nd.ni_vp); 1891 break; 1892 1893 case VOP_CREATE_DESCOFFSET: 1894 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1895 &nd.ni_cnd, &vattr); 1896 if (error == 0) 1897 vput(nd.ni_vp); 1898 break; 1899 } 1900 } else { 1901 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1902 if (nd.ni_dvp == vp) 1903 vrele(nd.ni_dvp); 1904 else 1905 vput(nd.ni_dvp); 1906 if (vp) 1907 vrele(vp); 1908 } 1909 out: 1910 VERIEXEC_PATH_PUT(path); 1911 return (error); 1912 } 1913 1914 /* 1915 * Create a named pipe. 1916 */ 1917 /* ARGSUSED */ 1918 int 1919 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1920 { 1921 /* { 1922 syscallarg(const char *) path; 1923 syscallarg(int) mode; 1924 } */ 1925 struct proc *p = l->l_proc; 1926 struct vattr vattr; 1927 int error; 1928 struct nameidata nd; 1929 1930 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1931 SCARG(uap, path)); 1932 if ((error = namei(&nd)) != 0) 1933 return (error); 1934 if (nd.ni_vp != NULL) { 1935 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1936 if (nd.ni_dvp == nd.ni_vp) 1937 vrele(nd.ni_dvp); 1938 else 1939 vput(nd.ni_dvp); 1940 vrele(nd.ni_vp); 1941 return (EEXIST); 1942 } 1943 VATTR_NULL(&vattr); 1944 vattr.va_type = VFIFO; 1945 /* We will read cwdi->cwdi_cmask unlocked. */ 1946 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1947 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1948 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1949 if (error == 0) 1950 vput(nd.ni_vp); 1951 return (error); 1952 } 1953 1954 /* 1955 * Make a hard file link. 1956 */ 1957 /* ARGSUSED */ 1958 int 1959 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1960 { 1961 /* { 1962 syscallarg(const char *) path; 1963 syscallarg(const char *) link; 1964 } */ 1965 struct vnode *vp; 1966 struct nameidata nd; 1967 int error; 1968 1969 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1970 SCARG(uap, path)); 1971 if ((error = namei(&nd)) != 0) 1972 return (error); 1973 vp = nd.ni_vp; 1974 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1975 SCARG(uap, link)); 1976 if ((error = namei(&nd)) != 0) 1977 goto out; 1978 if (nd.ni_vp) { 1979 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1980 if (nd.ni_dvp == nd.ni_vp) 1981 vrele(nd.ni_dvp); 1982 else 1983 vput(nd.ni_dvp); 1984 vrele(nd.ni_vp); 1985 error = EEXIST; 1986 goto out; 1987 } 1988 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1989 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 1990 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1991 out: 1992 vrele(vp); 1993 return (error); 1994 } 1995 1996 /* 1997 * Make a symbolic link. 1998 */ 1999 /* ARGSUSED */ 2000 int 2001 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2002 { 2003 /* { 2004 syscallarg(const char *) path; 2005 syscallarg(const char *) link; 2006 } */ 2007 struct proc *p = l->l_proc; 2008 struct vattr vattr; 2009 char *path; 2010 int error; 2011 struct nameidata nd; 2012 2013 path = PNBUF_GET(); 2014 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2015 if (error) 2016 goto out; 2017 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2018 SCARG(uap, link)); 2019 if ((error = namei(&nd)) != 0) 2020 goto out; 2021 if (nd.ni_vp) { 2022 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2023 if (nd.ni_dvp == nd.ni_vp) 2024 vrele(nd.ni_dvp); 2025 else 2026 vput(nd.ni_dvp); 2027 vrele(nd.ni_vp); 2028 error = EEXIST; 2029 goto out; 2030 } 2031 VATTR_NULL(&vattr); 2032 vattr.va_type = VLNK; 2033 /* We will read cwdi->cwdi_cmask unlocked. */ 2034 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2035 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2036 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2037 if (error == 0) 2038 vput(nd.ni_vp); 2039 out: 2040 PNBUF_PUT(path); 2041 return (error); 2042 } 2043 2044 /* 2045 * Delete a whiteout from the filesystem. 2046 */ 2047 /* ARGSUSED */ 2048 int 2049 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2050 { 2051 /* { 2052 syscallarg(const char *) path; 2053 } */ 2054 int error; 2055 struct nameidata nd; 2056 2057 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2058 UIO_USERSPACE, SCARG(uap, path)); 2059 error = namei(&nd); 2060 if (error) 2061 return (error); 2062 2063 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 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 if (nd.ni_vp) 2070 vrele(nd.ni_vp); 2071 return (EEXIST); 2072 } 2073 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2074 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2075 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2076 vput(nd.ni_dvp); 2077 return (error); 2078 } 2079 2080 /* 2081 * Delete a name from the filesystem. 2082 */ 2083 /* ARGSUSED */ 2084 int 2085 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2086 { 2087 /* { 2088 syscallarg(const char *) path; 2089 } */ 2090 2091 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2092 } 2093 2094 int 2095 do_sys_unlink(const char *arg, enum uio_seg seg) 2096 { 2097 struct vnode *vp; 2098 int error; 2099 struct nameidata nd; 2100 kauth_cred_t cred; 2101 char *path; 2102 const char *cpath; 2103 2104 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2105 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2106 2107 if ((error = namei(&nd)) != 0) 2108 goto out; 2109 vp = nd.ni_vp; 2110 2111 /* 2112 * The root of a mounted filesystem cannot be deleted. 2113 */ 2114 if (vp->v_vflag & VV_ROOT) { 2115 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2116 if (nd.ni_dvp == vp) 2117 vrele(nd.ni_dvp); 2118 else 2119 vput(nd.ni_dvp); 2120 vput(vp); 2121 error = EBUSY; 2122 goto out; 2123 } 2124 2125 #if NVERIEXEC > 0 2126 /* Handle remove requests for veriexec entries. */ 2127 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2128 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2129 if (nd.ni_dvp == vp) 2130 vrele(nd.ni_dvp); 2131 else 2132 vput(nd.ni_dvp); 2133 vput(vp); 2134 goto out; 2135 } 2136 #endif /* NVERIEXEC > 0 */ 2137 2138 cred = kauth_cred_get(); 2139 VOP_LEASE(nd.ni_dvp, cred, LEASE_WRITE); 2140 VOP_LEASE(vp, cred, LEASE_WRITE); 2141 #ifdef FILEASSOC 2142 (void)fileassoc_file_delete(vp); 2143 #endif /* FILEASSOC */ 2144 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2145 out: 2146 VERIEXEC_PATH_PUT(path); 2147 return (error); 2148 } 2149 2150 /* 2151 * Reposition read/write file offset. 2152 */ 2153 int 2154 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2155 { 2156 /* { 2157 syscallarg(int) fd; 2158 syscallarg(int) pad; 2159 syscallarg(off_t) offset; 2160 syscallarg(int) whence; 2161 } */ 2162 struct proc *p = l->l_proc; 2163 kauth_cred_t cred = l->l_cred; 2164 struct filedesc *fdp = p->p_fd; 2165 struct file *fp; 2166 struct vnode *vp; 2167 struct vattr vattr; 2168 off_t newoff; 2169 int error; 2170 2171 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2172 return (EBADF); 2173 2174 vp = (struct vnode *)fp->f_data; 2175 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2176 error = ESPIPE; 2177 FILE_UNLOCK(fp); 2178 goto out; 2179 } 2180 2181 switch (SCARG(uap, whence)) { 2182 case SEEK_CUR: 2183 newoff = fp->f_offset + SCARG(uap, offset); 2184 FILE_USE(fp); 2185 break; 2186 case SEEK_END: 2187 FILE_USE(fp); 2188 error = VOP_GETATTR(vp, &vattr, cred); 2189 if (error) { 2190 FILE_UNUSE(fp, l); 2191 goto out; 2192 } 2193 newoff = SCARG(uap, offset) + vattr.va_size; 2194 break; 2195 case SEEK_SET: 2196 FILE_USE(fp); 2197 newoff = SCARG(uap, offset); 2198 break; 2199 default: 2200 FILE_UNLOCK(fp); 2201 error = EINVAL; 2202 goto out; 2203 } 2204 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2205 FILE_LOCK(fp); 2206 *(off_t *)retval = fp->f_offset = newoff; 2207 FILE_UNLOCK(fp); 2208 } 2209 FILE_UNUSE(fp, l); 2210 out: 2211 return (error); 2212 } 2213 2214 /* 2215 * Positional read system call. 2216 */ 2217 int 2218 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2219 { 2220 /* { 2221 syscallarg(int) fd; 2222 syscallarg(void *) buf; 2223 syscallarg(size_t) nbyte; 2224 syscallarg(off_t) offset; 2225 } */ 2226 struct proc *p = l->l_proc; 2227 struct filedesc *fdp = p->p_fd; 2228 struct file *fp; 2229 struct vnode *vp; 2230 off_t offset; 2231 int error, fd = SCARG(uap, fd); 2232 2233 if ((fp = fd_getfile(fdp, fd)) == NULL) 2234 return (EBADF); 2235 2236 if ((fp->f_flag & FREAD) == 0) { 2237 FILE_UNLOCK(fp); 2238 return (EBADF); 2239 } 2240 2241 FILE_USE(fp); 2242 2243 vp = (struct vnode *)fp->f_data; 2244 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2245 error = ESPIPE; 2246 goto out; 2247 } 2248 2249 offset = SCARG(uap, offset); 2250 2251 /* 2252 * XXX This works because no file systems actually 2253 * XXX take any action on the seek operation. 2254 */ 2255 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2256 goto out; 2257 2258 /* dofileread() will unuse the descriptor for us */ 2259 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2260 &offset, 0, retval)); 2261 2262 out: 2263 FILE_UNUSE(fp, l); 2264 return (error); 2265 } 2266 2267 /* 2268 * Positional scatter read system call. 2269 */ 2270 int 2271 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2272 { 2273 /* { 2274 syscallarg(int) fd; 2275 syscallarg(const struct iovec *) iovp; 2276 syscallarg(int) iovcnt; 2277 syscallarg(off_t) offset; 2278 } */ 2279 off_t offset = SCARG(uap, offset); 2280 2281 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2282 SCARG(uap, iovcnt), &offset, 0, retval); 2283 } 2284 2285 /* 2286 * Positional write system call. 2287 */ 2288 int 2289 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2290 { 2291 /* { 2292 syscallarg(int) fd; 2293 syscallarg(const void *) buf; 2294 syscallarg(size_t) nbyte; 2295 syscallarg(off_t) offset; 2296 } */ 2297 struct proc *p = l->l_proc; 2298 struct filedesc *fdp = p->p_fd; 2299 struct file *fp; 2300 struct vnode *vp; 2301 off_t offset; 2302 int error, fd = SCARG(uap, fd); 2303 2304 if ((fp = fd_getfile(fdp, fd)) == NULL) 2305 return (EBADF); 2306 2307 if ((fp->f_flag & FWRITE) == 0) { 2308 FILE_UNLOCK(fp); 2309 return (EBADF); 2310 } 2311 2312 FILE_USE(fp); 2313 2314 vp = (struct vnode *)fp->f_data; 2315 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2316 error = ESPIPE; 2317 goto out; 2318 } 2319 2320 offset = SCARG(uap, offset); 2321 2322 /* 2323 * XXX This works because no file systems actually 2324 * XXX take any action on the seek operation. 2325 */ 2326 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2327 goto out; 2328 2329 /* dofilewrite() will unuse the descriptor for us */ 2330 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2331 &offset, 0, retval)); 2332 2333 out: 2334 FILE_UNUSE(fp, l); 2335 return (error); 2336 } 2337 2338 /* 2339 * Positional gather write system call. 2340 */ 2341 int 2342 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2343 { 2344 /* { 2345 syscallarg(int) fd; 2346 syscallarg(const struct iovec *) iovp; 2347 syscallarg(int) iovcnt; 2348 syscallarg(off_t) offset; 2349 } */ 2350 off_t offset = SCARG(uap, offset); 2351 2352 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2353 SCARG(uap, iovcnt), &offset, 0, retval); 2354 } 2355 2356 /* 2357 * Check access permissions. 2358 */ 2359 int 2360 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2361 { 2362 /* { 2363 syscallarg(const char *) path; 2364 syscallarg(int) flags; 2365 } */ 2366 kauth_cred_t cred; 2367 struct vnode *vp; 2368 int error, flags; 2369 struct nameidata nd; 2370 2371 cred = kauth_cred_dup(l->l_cred); 2372 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2373 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2374 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2375 SCARG(uap, path)); 2376 /* Override default credentials */ 2377 nd.ni_cnd.cn_cred = cred; 2378 if ((error = namei(&nd)) != 0) 2379 goto out; 2380 vp = nd.ni_vp; 2381 2382 /* Flags == 0 means only check for existence. */ 2383 if (SCARG(uap, flags)) { 2384 flags = 0; 2385 if (SCARG(uap, flags) & R_OK) 2386 flags |= VREAD; 2387 if (SCARG(uap, flags) & W_OK) 2388 flags |= VWRITE; 2389 if (SCARG(uap, flags) & X_OK) 2390 flags |= VEXEC; 2391 2392 error = VOP_ACCESS(vp, flags, cred); 2393 if (!error && (flags & VWRITE)) 2394 error = vn_writechk(vp); 2395 } 2396 vput(vp); 2397 out: 2398 kauth_cred_free(cred); 2399 return (error); 2400 } 2401 2402 /* 2403 * Common code for all sys_stat functions, including compat versions. 2404 */ 2405 int 2406 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2407 struct stat *sb) 2408 { 2409 int error; 2410 struct nameidata nd; 2411 2412 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2413 UIO_USERSPACE, path); 2414 error = namei(&nd); 2415 if (error != 0) 2416 return error; 2417 error = vn_stat(nd.ni_vp, sb, l); 2418 vput(nd.ni_vp); 2419 return error; 2420 } 2421 2422 /* 2423 * Get file status; this version follows links. 2424 */ 2425 /* ARGSUSED */ 2426 int 2427 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2428 { 2429 /* { 2430 syscallarg(const char *) path; 2431 syscallarg(struct stat *) ub; 2432 } */ 2433 struct stat sb; 2434 int error; 2435 2436 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2437 if (error) 2438 return error; 2439 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2440 } 2441 2442 /* 2443 * Get file status; this version does not follow links. 2444 */ 2445 /* ARGSUSED */ 2446 int 2447 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2448 { 2449 /* { 2450 syscallarg(const char *) path; 2451 syscallarg(struct stat *) ub; 2452 } */ 2453 struct stat sb; 2454 int error; 2455 2456 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2457 if (error) 2458 return error; 2459 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2460 } 2461 2462 /* 2463 * Get configurable pathname variables. 2464 */ 2465 /* ARGSUSED */ 2466 int 2467 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2468 { 2469 /* { 2470 syscallarg(const char *) path; 2471 syscallarg(int) name; 2472 } */ 2473 int error; 2474 struct nameidata nd; 2475 2476 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2477 SCARG(uap, path)); 2478 if ((error = namei(&nd)) != 0) 2479 return (error); 2480 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2481 vput(nd.ni_vp); 2482 return (error); 2483 } 2484 2485 /* 2486 * Return target name of a symbolic link. 2487 */ 2488 /* ARGSUSED */ 2489 int 2490 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2491 { 2492 /* { 2493 syscallarg(const char *) path; 2494 syscallarg(char *) buf; 2495 syscallarg(size_t) count; 2496 } */ 2497 struct vnode *vp; 2498 struct iovec aiov; 2499 struct uio auio; 2500 int error; 2501 struct nameidata nd; 2502 2503 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2504 SCARG(uap, path)); 2505 if ((error = namei(&nd)) != 0) 2506 return (error); 2507 vp = nd.ni_vp; 2508 if (vp->v_type != VLNK) 2509 error = EINVAL; 2510 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2511 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2512 aiov.iov_base = SCARG(uap, buf); 2513 aiov.iov_len = SCARG(uap, count); 2514 auio.uio_iov = &aiov; 2515 auio.uio_iovcnt = 1; 2516 auio.uio_offset = 0; 2517 auio.uio_rw = UIO_READ; 2518 KASSERT(l == curlwp); 2519 auio.uio_vmspace = l->l_proc->p_vmspace; 2520 auio.uio_resid = SCARG(uap, count); 2521 error = VOP_READLINK(vp, &auio, l->l_cred); 2522 } 2523 vput(vp); 2524 *retval = SCARG(uap, count) - auio.uio_resid; 2525 return (error); 2526 } 2527 2528 /* 2529 * Change flags of a file given a path name. 2530 */ 2531 /* ARGSUSED */ 2532 int 2533 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2534 { 2535 /* { 2536 syscallarg(const char *) path; 2537 syscallarg(u_long) flags; 2538 } */ 2539 struct vnode *vp; 2540 int error; 2541 struct nameidata nd; 2542 2543 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2544 SCARG(uap, path)); 2545 if ((error = namei(&nd)) != 0) 2546 return (error); 2547 vp = nd.ni_vp; 2548 error = change_flags(vp, SCARG(uap, flags), l); 2549 vput(vp); 2550 return (error); 2551 } 2552 2553 /* 2554 * Change flags of a file given a file descriptor. 2555 */ 2556 /* ARGSUSED */ 2557 int 2558 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2559 { 2560 /* { 2561 syscallarg(int) fd; 2562 syscallarg(u_long) flags; 2563 } */ 2564 struct proc *p = l->l_proc; 2565 struct vnode *vp; 2566 struct file *fp; 2567 int error; 2568 2569 /* getvnode() will use the descriptor for us */ 2570 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2571 return (error); 2572 vp = (struct vnode *)fp->f_data; 2573 error = change_flags(vp, SCARG(uap, flags), l); 2574 VOP_UNLOCK(vp, 0); 2575 FILE_UNUSE(fp, l); 2576 return (error); 2577 } 2578 2579 /* 2580 * Change flags of a file given a path name; this version does 2581 * not follow links. 2582 */ 2583 int 2584 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2585 { 2586 /* { 2587 syscallarg(const char *) path; 2588 syscallarg(u_long) flags; 2589 } */ 2590 struct vnode *vp; 2591 int error; 2592 struct nameidata nd; 2593 2594 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2595 SCARG(uap, path)); 2596 if ((error = namei(&nd)) != 0) 2597 return (error); 2598 vp = nd.ni_vp; 2599 error = change_flags(vp, SCARG(uap, flags), l); 2600 vput(vp); 2601 return (error); 2602 } 2603 2604 /* 2605 * Common routine to change flags of a file. 2606 */ 2607 int 2608 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2609 { 2610 struct vattr vattr; 2611 int error; 2612 2613 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2614 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2615 /* 2616 * Non-superusers cannot change the flags on devices, even if they 2617 * own them. 2618 */ 2619 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2620 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2621 goto out; 2622 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2623 error = EINVAL; 2624 goto out; 2625 } 2626 } 2627 VATTR_NULL(&vattr); 2628 vattr.va_flags = flags; 2629 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2630 out: 2631 return (error); 2632 } 2633 2634 /* 2635 * Change mode of a file given path name; this version follows links. 2636 */ 2637 /* ARGSUSED */ 2638 int 2639 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2640 { 2641 /* { 2642 syscallarg(const char *) path; 2643 syscallarg(int) mode; 2644 } */ 2645 int error; 2646 struct nameidata nd; 2647 2648 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2649 SCARG(uap, path)); 2650 if ((error = namei(&nd)) != 0) 2651 return (error); 2652 2653 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2654 2655 vrele(nd.ni_vp); 2656 return (error); 2657 } 2658 2659 /* 2660 * Change mode of a file given a file descriptor. 2661 */ 2662 /* ARGSUSED */ 2663 int 2664 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2665 { 2666 /* { 2667 syscallarg(int) fd; 2668 syscallarg(int) mode; 2669 } */ 2670 struct proc *p = l->l_proc; 2671 struct file *fp; 2672 int error; 2673 2674 /* getvnode() will use the descriptor for us */ 2675 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2676 return (error); 2677 2678 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2679 FILE_UNUSE(fp, l); 2680 return (error); 2681 } 2682 2683 /* 2684 * Change mode of a file given path name; this version does not follow links. 2685 */ 2686 /* ARGSUSED */ 2687 int 2688 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2689 { 2690 /* { 2691 syscallarg(const char *) path; 2692 syscallarg(int) mode; 2693 } */ 2694 int error; 2695 struct nameidata nd; 2696 2697 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2698 SCARG(uap, path)); 2699 if ((error = namei(&nd)) != 0) 2700 return (error); 2701 2702 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2703 2704 vrele(nd.ni_vp); 2705 return (error); 2706 } 2707 2708 /* 2709 * Common routine to set mode given a vnode. 2710 */ 2711 static int 2712 change_mode(struct vnode *vp, int mode, struct lwp *l) 2713 { 2714 struct vattr vattr; 2715 int error; 2716 2717 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2718 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2719 VATTR_NULL(&vattr); 2720 vattr.va_mode = mode & ALLPERMS; 2721 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2722 VOP_UNLOCK(vp, 0); 2723 return (error); 2724 } 2725 2726 /* 2727 * Set ownership given a path name; this version follows links. 2728 */ 2729 /* ARGSUSED */ 2730 int 2731 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2732 { 2733 /* { 2734 syscallarg(const char *) path; 2735 syscallarg(uid_t) uid; 2736 syscallarg(gid_t) gid; 2737 } */ 2738 int error; 2739 struct nameidata nd; 2740 2741 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2742 SCARG(uap, path)); 2743 if ((error = namei(&nd)) != 0) 2744 return (error); 2745 2746 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2747 2748 vrele(nd.ni_vp); 2749 return (error); 2750 } 2751 2752 /* 2753 * Set ownership given a path name; this version follows links. 2754 * Provides POSIX semantics. 2755 */ 2756 /* ARGSUSED */ 2757 int 2758 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2759 { 2760 /* { 2761 syscallarg(const char *) path; 2762 syscallarg(uid_t) uid; 2763 syscallarg(gid_t) gid; 2764 } */ 2765 int error; 2766 struct nameidata nd; 2767 2768 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2769 SCARG(uap, path)); 2770 if ((error = namei(&nd)) != 0) 2771 return (error); 2772 2773 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2774 2775 vrele(nd.ni_vp); 2776 return (error); 2777 } 2778 2779 /* 2780 * Set ownership given a file descriptor. 2781 */ 2782 /* ARGSUSED */ 2783 int 2784 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2785 { 2786 /* { 2787 syscallarg(int) fd; 2788 syscallarg(uid_t) uid; 2789 syscallarg(gid_t) gid; 2790 } */ 2791 struct proc *p = l->l_proc; 2792 int error; 2793 struct file *fp; 2794 2795 /* getvnode() will use the descriptor for us */ 2796 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2797 return (error); 2798 2799 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2800 SCARG(uap, gid), l, 0); 2801 FILE_UNUSE(fp, l); 2802 return (error); 2803 } 2804 2805 /* 2806 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2807 */ 2808 /* ARGSUSED */ 2809 int 2810 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2811 { 2812 /* { 2813 syscallarg(int) fd; 2814 syscallarg(uid_t) uid; 2815 syscallarg(gid_t) gid; 2816 } */ 2817 struct proc *p = l->l_proc; 2818 int error; 2819 struct file *fp; 2820 2821 /* getvnode() will use the descriptor for us */ 2822 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2823 return (error); 2824 2825 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2826 SCARG(uap, gid), l, 1); 2827 FILE_UNUSE(fp, l); 2828 return (error); 2829 } 2830 2831 /* 2832 * Set ownership given a path name; this version does not follow links. 2833 */ 2834 /* ARGSUSED */ 2835 int 2836 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2837 { 2838 /* { 2839 syscallarg(const char *) path; 2840 syscallarg(uid_t) uid; 2841 syscallarg(gid_t) gid; 2842 } */ 2843 int error; 2844 struct nameidata nd; 2845 2846 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2847 SCARG(uap, path)); 2848 if ((error = namei(&nd)) != 0) 2849 return (error); 2850 2851 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2852 2853 vrele(nd.ni_vp); 2854 return (error); 2855 } 2856 2857 /* 2858 * Set ownership given a path name; this version does not follow links. 2859 * Provides POSIX/XPG semantics. 2860 */ 2861 /* ARGSUSED */ 2862 int 2863 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2864 { 2865 /* { 2866 syscallarg(const char *) path; 2867 syscallarg(uid_t) uid; 2868 syscallarg(gid_t) gid; 2869 } */ 2870 int error; 2871 struct nameidata nd; 2872 2873 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2874 SCARG(uap, path)); 2875 if ((error = namei(&nd)) != 0) 2876 return (error); 2877 2878 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2879 2880 vrele(nd.ni_vp); 2881 return (error); 2882 } 2883 2884 /* 2885 * Common routine to set ownership given a vnode. 2886 */ 2887 static int 2888 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2889 int posix_semantics) 2890 { 2891 struct vattr vattr; 2892 mode_t newmode; 2893 int error; 2894 2895 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2896 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2897 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2898 goto out; 2899 2900 #define CHANGED(x) ((int)(x) != -1) 2901 newmode = vattr.va_mode; 2902 if (posix_semantics) { 2903 /* 2904 * POSIX/XPG semantics: if the caller is not the super-user, 2905 * clear set-user-id and set-group-id bits. Both POSIX and 2906 * the XPG consider the behaviour for calls by the super-user 2907 * implementation-defined; we leave the set-user-id and set- 2908 * group-id settings intact in that case. 2909 */ 2910 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2911 NULL) != 0) 2912 newmode &= ~(S_ISUID | S_ISGID); 2913 } else { 2914 /* 2915 * NetBSD semantics: when changing owner and/or group, 2916 * clear the respective bit(s). 2917 */ 2918 if (CHANGED(uid)) 2919 newmode &= ~S_ISUID; 2920 if (CHANGED(gid)) 2921 newmode &= ~S_ISGID; 2922 } 2923 /* Update va_mode iff altered. */ 2924 if (vattr.va_mode == newmode) 2925 newmode = VNOVAL; 2926 2927 VATTR_NULL(&vattr); 2928 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2929 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2930 vattr.va_mode = newmode; 2931 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2932 #undef CHANGED 2933 2934 out: 2935 VOP_UNLOCK(vp, 0); 2936 return (error); 2937 } 2938 2939 /* 2940 * Set the access and modification times given a path name; this 2941 * version follows links. 2942 */ 2943 /* ARGSUSED */ 2944 int 2945 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2946 { 2947 /* { 2948 syscallarg(const char *) path; 2949 syscallarg(const struct timeval *) tptr; 2950 } */ 2951 2952 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2953 SCARG(uap, tptr), UIO_USERSPACE); 2954 } 2955 2956 /* 2957 * Set the access and modification times given a file descriptor. 2958 */ 2959 /* ARGSUSED */ 2960 int 2961 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2962 { 2963 /* { 2964 syscallarg(int) fd; 2965 syscallarg(const struct timeval *) tptr; 2966 } */ 2967 int error; 2968 struct file *fp; 2969 2970 /* getvnode() will use the descriptor for us */ 2971 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2972 return (error); 2973 2974 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2975 SCARG(uap, tptr), UIO_USERSPACE); 2976 2977 FILE_UNUSE(fp, l); 2978 return (error); 2979 } 2980 2981 /* 2982 * Set the access and modification times given a path name; this 2983 * version does not follow links. 2984 */ 2985 int 2986 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2987 { 2988 /* { 2989 syscallarg(const char *) path; 2990 syscallarg(const struct timeval *) tptr; 2991 } */ 2992 2993 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2994 SCARG(uap, tptr), UIO_USERSPACE); 2995 } 2996 2997 /* 2998 * Common routine to set access and modification times given a vnode. 2999 */ 3000 int 3001 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3002 const struct timeval *tptr, enum uio_seg seg) 3003 { 3004 struct vattr vattr; 3005 struct nameidata nd; 3006 int error; 3007 3008 VATTR_NULL(&vattr); 3009 if (tptr == NULL) { 3010 nanotime(&vattr.va_atime); 3011 vattr.va_mtime = vattr.va_atime; 3012 vattr.va_vaflags |= VA_UTIMES_NULL; 3013 } else { 3014 struct timeval tv[2]; 3015 3016 if (seg != UIO_SYSSPACE) { 3017 error = copyin(tptr, &tv, sizeof (tv)); 3018 if (error != 0) 3019 return error; 3020 tptr = tv; 3021 } 3022 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3023 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3024 } 3025 3026 if (vp == NULL) { 3027 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3028 if ((error = namei(&nd)) != 0) 3029 return (error); 3030 vp = nd.ni_vp; 3031 } else 3032 nd.ni_vp = NULL; 3033 3034 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3035 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3036 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3037 VOP_UNLOCK(vp, 0); 3038 3039 if (nd.ni_vp != NULL) 3040 vrele(nd.ni_vp); 3041 3042 return (error); 3043 } 3044 3045 /* 3046 * Truncate a file given its path name. 3047 */ 3048 /* ARGSUSED */ 3049 int 3050 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3051 { 3052 /* { 3053 syscallarg(const char *) path; 3054 syscallarg(int) pad; 3055 syscallarg(off_t) length; 3056 } */ 3057 struct vnode *vp; 3058 struct vattr vattr; 3059 int error; 3060 struct nameidata nd; 3061 3062 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3063 SCARG(uap, path)); 3064 if ((error = namei(&nd)) != 0) 3065 return (error); 3066 vp = nd.ni_vp; 3067 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3068 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3069 if (vp->v_type == VDIR) 3070 error = EISDIR; 3071 else if ((error = vn_writechk(vp)) == 0 && 3072 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3073 VATTR_NULL(&vattr); 3074 vattr.va_size = SCARG(uap, length); 3075 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3076 } 3077 vput(vp); 3078 return (error); 3079 } 3080 3081 /* 3082 * Truncate a file given a file descriptor. 3083 */ 3084 /* ARGSUSED */ 3085 int 3086 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3087 { 3088 /* { 3089 syscallarg(int) fd; 3090 syscallarg(int) pad; 3091 syscallarg(off_t) length; 3092 } */ 3093 struct proc *p = l->l_proc; 3094 struct vattr vattr; 3095 struct vnode *vp; 3096 struct file *fp; 3097 int error; 3098 3099 /* getvnode() will use the descriptor for us */ 3100 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3101 return (error); 3102 if ((fp->f_flag & FWRITE) == 0) { 3103 error = EINVAL; 3104 goto out; 3105 } 3106 vp = (struct vnode *)fp->f_data; 3107 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3108 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3109 if (vp->v_type == VDIR) 3110 error = EISDIR; 3111 else if ((error = vn_writechk(vp)) == 0) { 3112 VATTR_NULL(&vattr); 3113 vattr.va_size = SCARG(uap, length); 3114 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3115 } 3116 VOP_UNLOCK(vp, 0); 3117 out: 3118 FILE_UNUSE(fp, l); 3119 return (error); 3120 } 3121 3122 /* 3123 * Sync an open file. 3124 */ 3125 /* ARGSUSED */ 3126 int 3127 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3128 { 3129 /* { 3130 syscallarg(int) fd; 3131 } */ 3132 struct proc *p = l->l_proc; 3133 struct vnode *vp; 3134 struct file *fp; 3135 int error; 3136 3137 /* getvnode() will use the descriptor for us */ 3138 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3139 return (error); 3140 vp = (struct vnode *)fp->f_data; 3141 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3142 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3143 if (error == 0 && bioopsp != NULL && 3144 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3145 (*bioopsp->io_fsync)(vp, 0); 3146 VOP_UNLOCK(vp, 0); 3147 FILE_UNUSE(fp, l); 3148 return (error); 3149 } 3150 3151 /* 3152 * Sync a range of file data. API modeled after that found in AIX. 3153 * 3154 * FDATASYNC indicates that we need only save enough metadata to be able 3155 * to re-read the written data. Note we duplicate AIX's requirement that 3156 * the file be open for writing. 3157 */ 3158 /* ARGSUSED */ 3159 int 3160 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3161 { 3162 /* { 3163 syscallarg(int) fd; 3164 syscallarg(int) flags; 3165 syscallarg(off_t) start; 3166 syscallarg(off_t) length; 3167 } */ 3168 struct proc *p = l->l_proc; 3169 struct vnode *vp; 3170 struct file *fp; 3171 int flags, nflags; 3172 off_t s, e, len; 3173 int error; 3174 3175 /* getvnode() will use the descriptor for us */ 3176 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3177 return (error); 3178 3179 if ((fp->f_flag & FWRITE) == 0) { 3180 error = EBADF; 3181 goto out; 3182 } 3183 3184 flags = SCARG(uap, flags); 3185 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3186 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3187 error = EINVAL; 3188 goto out; 3189 } 3190 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3191 if (flags & FDATASYNC) 3192 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3193 else 3194 nflags = FSYNC_WAIT; 3195 if (flags & FDISKSYNC) 3196 nflags |= FSYNC_CACHE; 3197 3198 len = SCARG(uap, length); 3199 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3200 if (len) { 3201 s = SCARG(uap, start); 3202 e = s + len; 3203 if (e < s) { 3204 error = EINVAL; 3205 goto out; 3206 } 3207 } else { 3208 e = 0; 3209 s = 0; 3210 } 3211 3212 vp = (struct vnode *)fp->f_data; 3213 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3214 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3215 3216 if (error == 0 && bioopsp != NULL && 3217 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3218 (*bioopsp->io_fsync)(vp, nflags); 3219 3220 VOP_UNLOCK(vp, 0); 3221 out: 3222 FILE_UNUSE(fp, l); 3223 return (error); 3224 } 3225 3226 /* 3227 * Sync the data of an open file. 3228 */ 3229 /* ARGSUSED */ 3230 int 3231 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3232 { 3233 /* { 3234 syscallarg(int) fd; 3235 } */ 3236 struct proc *p = l->l_proc; 3237 struct vnode *vp; 3238 struct file *fp; 3239 int error; 3240 3241 /* getvnode() will use the descriptor for us */ 3242 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3243 return (error); 3244 if ((fp->f_flag & FWRITE) == 0) { 3245 FILE_UNUSE(fp, l); 3246 return (EBADF); 3247 } 3248 vp = (struct vnode *)fp->f_data; 3249 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3250 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3251 VOP_UNLOCK(vp, 0); 3252 FILE_UNUSE(fp, l); 3253 return (error); 3254 } 3255 3256 /* 3257 * Rename files, (standard) BSD semantics frontend. 3258 */ 3259 /* ARGSUSED */ 3260 int 3261 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3262 { 3263 /* { 3264 syscallarg(const char *) from; 3265 syscallarg(const char *) to; 3266 } */ 3267 3268 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3269 } 3270 3271 /* 3272 * Rename files, POSIX semantics frontend. 3273 */ 3274 /* ARGSUSED */ 3275 int 3276 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3277 { 3278 /* { 3279 syscallarg(const char *) from; 3280 syscallarg(const char *) to; 3281 } */ 3282 3283 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3284 } 3285 3286 /* 3287 * Rename files. Source and destination must either both be directories, 3288 * or both not be directories. If target is a directory, it must be empty. 3289 * If `from' and `to' refer to the same object, the value of the `retain' 3290 * argument is used to determine whether `from' will be 3291 * 3292 * (retain == 0) deleted unless `from' and `to' refer to the same 3293 * object in the file system's name space (BSD). 3294 * (retain == 1) always retained (POSIX). 3295 */ 3296 int 3297 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3298 { 3299 struct vnode *tvp, *fvp, *tdvp; 3300 struct nameidata fromnd, tond; 3301 struct lwp *l = curlwp; 3302 struct proc *p; 3303 int error; 3304 3305 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3306 seg, from); 3307 if ((error = namei(&fromnd)) != 0) 3308 return (error); 3309 if (fromnd.ni_dvp != fromnd.ni_vp) 3310 VOP_UNLOCK(fromnd.ni_dvp, 0); 3311 fvp = fromnd.ni_vp; 3312 NDINIT(&tond, RENAME, 3313 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3314 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3315 seg, to); 3316 if ((error = namei(&tond)) != 0) { 3317 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3318 vrele(fromnd.ni_dvp); 3319 vrele(fvp); 3320 goto out1; 3321 } 3322 tdvp = tond.ni_dvp; 3323 tvp = tond.ni_vp; 3324 3325 if (tvp != NULL) { 3326 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3327 error = ENOTDIR; 3328 goto out; 3329 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3330 error = EISDIR; 3331 goto out; 3332 } 3333 } 3334 3335 if (fvp == tdvp) 3336 error = EINVAL; 3337 3338 /* 3339 * Source and destination refer to the same object. 3340 */ 3341 if (fvp == tvp) { 3342 if (retain) 3343 error = -1; 3344 else if (fromnd.ni_dvp == tdvp && 3345 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3346 !memcmp(fromnd.ni_cnd.cn_nameptr, 3347 tond.ni_cnd.cn_nameptr, 3348 fromnd.ni_cnd.cn_namelen)) 3349 error = -1; 3350 } 3351 3352 #if NVERIEXEC > 0 3353 if (!error) { 3354 char *f1, *f2; 3355 3356 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3357 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3358 3359 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3360 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3361 3362 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3363 3364 free(f1, M_TEMP); 3365 free(f2, M_TEMP); 3366 } 3367 #endif /* NVERIEXEC > 0 */ 3368 3369 out: 3370 p = l->l_proc; 3371 if (!error) { 3372 VOP_LEASE(tdvp, l->l_cred, LEASE_WRITE); 3373 if (fromnd.ni_dvp != tdvp) 3374 VOP_LEASE(fromnd.ni_dvp, l->l_cred, LEASE_WRITE); 3375 if (tvp) { 3376 VOP_LEASE(tvp, l->l_cred, LEASE_WRITE); 3377 } 3378 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3379 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3380 } else { 3381 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3382 if (tdvp == tvp) 3383 vrele(tdvp); 3384 else 3385 vput(tdvp); 3386 if (tvp) 3387 vput(tvp); 3388 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3389 vrele(fromnd.ni_dvp); 3390 vrele(fvp); 3391 } 3392 vrele(tond.ni_startdir); 3393 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3394 out1: 3395 if (fromnd.ni_startdir) 3396 vrele(fromnd.ni_startdir); 3397 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3398 return (error == -1 ? 0 : error); 3399 } 3400 3401 /* 3402 * Make a directory file. 3403 */ 3404 /* ARGSUSED */ 3405 int 3406 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3407 { 3408 /* { 3409 syscallarg(const char *) path; 3410 syscallarg(int) mode; 3411 } */ 3412 struct proc *p = l->l_proc; 3413 struct vnode *vp; 3414 struct vattr vattr; 3415 int error; 3416 struct nameidata nd; 3417 3418 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3419 SCARG(uap, path)); 3420 if ((error = namei(&nd)) != 0) 3421 return (error); 3422 vp = nd.ni_vp; 3423 if (vp != NULL) { 3424 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3425 if (nd.ni_dvp == vp) 3426 vrele(nd.ni_dvp); 3427 else 3428 vput(nd.ni_dvp); 3429 vrele(vp); 3430 return (EEXIST); 3431 } 3432 VATTR_NULL(&vattr); 3433 vattr.va_type = VDIR; 3434 /* We will read cwdi->cwdi_cmask unlocked. */ 3435 vattr.va_mode = 3436 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3437 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3438 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3439 if (!error) 3440 vput(nd.ni_vp); 3441 return (error); 3442 } 3443 3444 /* 3445 * Remove a directory file. 3446 */ 3447 /* ARGSUSED */ 3448 int 3449 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3450 { 3451 /* { 3452 syscallarg(const char *) path; 3453 } */ 3454 struct vnode *vp; 3455 int error; 3456 struct nameidata nd; 3457 3458 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3459 SCARG(uap, path)); 3460 if ((error = namei(&nd)) != 0) 3461 return (error); 3462 vp = nd.ni_vp; 3463 if (vp->v_type != VDIR) { 3464 error = ENOTDIR; 3465 goto out; 3466 } 3467 /* 3468 * No rmdir "." please. 3469 */ 3470 if (nd.ni_dvp == vp) { 3471 error = EINVAL; 3472 goto out; 3473 } 3474 /* 3475 * The root of a mounted filesystem cannot be deleted. 3476 */ 3477 if (vp->v_vflag & VV_ROOT) { 3478 error = EBUSY; 3479 goto out; 3480 } 3481 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3482 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3483 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3484 return (error); 3485 3486 out: 3487 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3488 if (nd.ni_dvp == vp) 3489 vrele(nd.ni_dvp); 3490 else 3491 vput(nd.ni_dvp); 3492 vput(vp); 3493 return (error); 3494 } 3495 3496 /* 3497 * Read a block of directory entries in a file system independent format. 3498 */ 3499 int 3500 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3501 { 3502 /* { 3503 syscallarg(int) fd; 3504 syscallarg(char *) buf; 3505 syscallarg(size_t) count; 3506 } */ 3507 struct proc *p = l->l_proc; 3508 struct file *fp; 3509 int error, done; 3510 3511 /* getvnode() will use the descriptor for us */ 3512 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3513 return (error); 3514 if ((fp->f_flag & FREAD) == 0) { 3515 error = EBADF; 3516 goto out; 3517 } 3518 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3519 SCARG(uap, count), &done, l, 0, 0); 3520 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3521 *retval = done; 3522 out: 3523 FILE_UNUSE(fp, l); 3524 return (error); 3525 } 3526 3527 /* 3528 * Set the mode mask for creation of filesystem nodes. 3529 */ 3530 int 3531 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3532 { 3533 /* { 3534 syscallarg(mode_t) newmask; 3535 } */ 3536 struct proc *p = l->l_proc; 3537 struct cwdinfo *cwdi; 3538 3539 /* 3540 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3541 * important is that we serialize changes to the mask. The 3542 * rw_exit() will issue a write memory barrier on our behalf, 3543 * and force the changes out to other CPUs (as it must use an 3544 * atomic operation, draining the local CPU's store buffers). 3545 */ 3546 cwdi = p->p_cwdi; 3547 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3548 *retval = cwdi->cwdi_cmask; 3549 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3550 rw_exit(&cwdi->cwdi_lock); 3551 3552 return (0); 3553 } 3554 3555 int 3556 dorevoke(struct vnode *vp, kauth_cred_t cred) 3557 { 3558 struct vattr vattr; 3559 int error; 3560 bool revoke; 3561 3562 if ((error = VOP_GETATTR(vp, &vattr, cred)) != 0) 3563 goto out; 3564 if (kauth_cred_geteuid(cred) != vattr.va_uid && 3565 (error = kauth_authorize_generic(cred, 3566 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3567 goto out; 3568 mutex_enter(&vp->v_interlock); 3569 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED|VI_LAYER))); 3570 mutex_exit(&vp->v_interlock); 3571 if (revoke) 3572 VOP_REVOKE(vp, REVOKEALL); 3573 3574 out: 3575 return (error); 3576 } 3577 3578 /* 3579 * Void all references to file by ripping underlying filesystem 3580 * away from vnode. 3581 */ 3582 /* ARGSUSED */ 3583 int 3584 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3585 { 3586 /* { 3587 syscallarg(const char *) path; 3588 } */ 3589 struct vnode *vp; 3590 int error; 3591 struct nameidata nd; 3592 3593 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3594 SCARG(uap, path)); 3595 if ((error = namei(&nd)) != 0) 3596 return (error); 3597 vp = nd.ni_vp; 3598 error = dorevoke(vp, l->l_cred); 3599 vrele(vp); 3600 return (error); 3601 } 3602 3603 /* 3604 * Convert a user file descriptor to a kernel file entry. 3605 */ 3606 int 3607 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3608 { 3609 struct vnode *vp; 3610 struct file *fp; 3611 3612 if ((fp = fd_getfile(fdp, fd)) == NULL) 3613 return (EBADF); 3614 3615 FILE_USE(fp); 3616 3617 if (fp->f_type != DTYPE_VNODE) { 3618 FILE_UNUSE(fp, NULL); 3619 return (EINVAL); 3620 } 3621 3622 vp = (struct vnode *)fp->f_data; 3623 if (vp->v_type == VBAD) { 3624 FILE_UNUSE(fp, NULL); 3625 return (EBADF); 3626 } 3627 3628 *fpp = fp; 3629 return (0); 3630 } 3631
Indexes created Thu Sep 25 16:09:42 GMT 2025