vfs_syscalls.c revision 1.306.2.15
1 /* $NetBSD: vfs_syscalls.c,v 1.306.2.15 2007/10/03 19:11:16 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.306.2.15 2007/10/03 19:11:16 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "fss.h" 46 #include "veriexec.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/kmem.h> 61 #include <sys/dirent.h> 62 #include <sys/sysctl.h> 63 #include <sys/syscallargs.h> 64 #include <sys/vfs_syscalls.h> 65 #include <sys/ktrace.h> 66 #ifdef FILEASSOC 67 #include <sys/fileassoc.h> 68 #endif /* FILEASSOC */ 69 #include <sys/verified_exec.h> 70 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #endif 79 #include <nfs/nfsproto.h> 80 #ifdef COMPAT_30 81 #include <nfs/nfs.h> 82 #include <nfs/nfs_var.h> 83 #endif 84 85 #if NFSS > 0 86 #include <dev/fssvar.h> 87 #endif 88 89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 90 91 static int change_dir(struct nameidata *, struct lwp *); 92 static int change_flags(struct vnode *, u_long, struct lwp *); 93 static int change_mode(struct vnode *, int, struct lwp *l); 94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 95 static int rename_files(const char *, const char *, struct lwp *, int); 96 97 void checkdirs(struct vnode *); 98 99 int dovfsusermount = 0; 100 101 /* 102 * Virtual File System System Calls 103 */ 104 105 /* 106 * Mount a file system. 107 */ 108 109 #if defined(COMPAT_09) || defined(COMPAT_43) 110 /* 111 * This table is used to maintain compatibility with 4.3BSD 112 * and NetBSD 0.9 mount syscalls. Note, the order is important! 113 * 114 * Do not modify this table. It should only contain filesystems 115 * supported by NetBSD 0.9 and 4.3BSD. 116 */ 117 const char * const mountcompatnames[] = { 118 NULL, /* 0 = MOUNT_NONE */ 119 MOUNT_FFS, /* 1 = MOUNT_UFS */ 120 MOUNT_NFS, /* 2 */ 121 MOUNT_MFS, /* 3 */ 122 MOUNT_MSDOS, /* 4 */ 123 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 124 MOUNT_FDESC, /* 6 */ 125 MOUNT_KERNFS, /* 7 */ 126 NULL, /* 8 = MOUNT_DEVFS */ 127 MOUNT_AFS, /* 9 */ 128 }; 129 const int nmountcompatnames = sizeof(mountcompatnames) / 130 sizeof(mountcompatnames[0]); 131 #endif /* COMPAT_09 || COMPAT_43 */ 132 133 static int 134 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 135 void *data, size_t *data_len) 136 { 137 struct mount *mp; 138 int error = 0, saved_flags; 139 140 mp = vp->v_mount; 141 saved_flags = mp->mnt_flag; 142 143 /* We can operate only on VV_ROOT nodes. */ 144 if ((vp->v_vflag & VV_ROOT) == 0) { 145 error = EINVAL; 146 goto out; 147 } 148 149 /* 150 * We only allow the filesystem to be reloaded if it 151 * is currently mounted read-only. 152 */ 153 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 154 error = EOPNOTSUPP; /* Needs translation */ 155 goto out; 156 } 157 158 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 159 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 160 if (error) 161 goto out; 162 163 if (vfs_busy(mp, LK_NOWAIT, 0)) { 164 error = EPERM; 165 goto out; 166 } 167 168 mp->mnt_flag &= ~MNT_OP_FLAGS; 169 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 170 171 /* 172 * Set the mount level flags. 173 */ 174 if (flags & MNT_RDONLY) 175 mp->mnt_flag |= MNT_RDONLY; 176 else if (mp->mnt_flag & MNT_RDONLY) 177 mp->mnt_iflag |= IMNT_WANTRDWR; 178 mp->mnt_flag &= 179 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 180 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 181 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 182 mp->mnt_flag |= flags & 183 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 184 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 185 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 186 MNT_IGNORE); 187 188 error = VFS_MOUNT(mp, path, data, data_len, l); 189 190 #if defined(COMPAT_30) && defined(NFSSERVER) 191 if (error && data != NULL) { 192 int error2; 193 194 /* Update failed; let's try and see if it was an 195 * export request. */ 196 error2 = nfs_update_exports_30(mp, path, data, l); 197 198 /* Only update error code if the export request was 199 * understood but some problem occurred while 200 * processing it. */ 201 if (error2 != EJUSTRETURN) 202 error = error2; 203 } 204 #endif 205 if (mp->mnt_iflag & IMNT_WANTRDWR) 206 mp->mnt_flag &= ~MNT_RDONLY; 207 if (error) 208 mp->mnt_flag = saved_flags; 209 mp->mnt_flag &= ~MNT_OP_FLAGS; 210 mp->mnt_iflag &= ~IMNT_WANTRDWR; 211 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 212 if (mp->mnt_syncer == NULL) 213 error = vfs_allocate_syncvnode(mp); 214 } else { 215 if (mp->mnt_syncer != NULL) 216 vfs_deallocate_syncvnode(mp); 217 } 218 vfs_unbusy(mp); 219 220 out: 221 return (error); 222 } 223 224 static int 225 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 226 { 227 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 228 int error; 229 230 /* Copy file-system type from userspace. */ 231 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 232 if (error) { 233 #if defined(COMPAT_09) || defined(COMPAT_43) 234 /* 235 * Historically, filesystem types were identified by numbers. 236 * If we get an integer for the filesystem type instead of a 237 * string, we check to see if it matches one of the historic 238 * filesystem types. 239 */ 240 u_long fsindex = (u_long)fstype; 241 if (fsindex >= nmountcompatnames || 242 mountcompatnames[fsindex] == NULL) 243 return ENODEV; 244 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 245 #else 246 return error; 247 #endif 248 } 249 250 #ifdef COMPAT_10 251 /* Accept `ufs' as an alias for `ffs'. */ 252 if (strcmp(fstypename, "ufs") == 0) 253 fstypename[0] = 'f'; 254 #endif 255 256 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 257 return ENODEV; 258 return 0; 259 } 260 261 static int 262 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 263 const char *path, int flags, void *data, size_t *data_len) 264 { 265 struct mount *mp = NULL; 266 struct vnode *vp = *vpp; 267 struct vattr va; 268 int error; 269 270 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 271 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 272 if (error) 273 return error; 274 275 /* Can't make a non-dir a mount-point (from here anyway). */ 276 if (vp->v_type != VDIR) 277 return ENOTDIR; 278 279 /* 280 * If the user is not root, ensure that they own the directory 281 * onto which we are attempting to mount. 282 */ 283 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 284 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 285 (error = kauth_authorize_generic(l->l_cred, 286 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 287 return error; 288 } 289 290 if (flags & MNT_EXPORTED) 291 return EINVAL; 292 293 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 294 return error; 295 296 /* 297 * Check if a file-system is not already mounted on this vnode. 298 */ 299 if (vp->v_mountedhere != NULL) 300 return EBUSY; 301 302 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 303 304 mp->mnt_op = vfsops; 305 306 TAILQ_INIT(&mp->mnt_vnodelist); 307 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 308 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, l); 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 VOP_UNLOCK(vp, 0); 349 checkdirs(vp); 350 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 351 error = vfs_allocate_syncvnode(mp); 352 vfs_unbusy(mp); 353 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 354 error = VFS_START(mp, 0, l); 355 if (error) 356 vrele(vp); 357 *vpp = NULL; 358 return error; 359 } 360 361 static int 362 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 363 void *data, size_t *data_len) 364 { 365 struct mount *mp; 366 int error; 367 368 /* If MNT_GETARGS is specified, it should be the only flag. */ 369 if (flags & ~MNT_GETARGS) 370 return EINVAL; 371 372 mp = vp->v_mount; 373 374 /* XXX: probably some notion of "can see" here if we want isolation. */ 375 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 376 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 377 if (error) 378 return error; 379 380 if ((vp->v_vflag & VV_ROOT) == 0) 381 return EINVAL; 382 383 if (vfs_busy(mp, LK_NOWAIT, 0)) 384 return EPERM; 385 386 mp->mnt_flag &= ~MNT_OP_FLAGS; 387 mp->mnt_flag |= MNT_GETARGS; 388 error = VFS_MOUNT(mp, path, data, data_len, l); 389 mp->mnt_flag &= ~MNT_OP_FLAGS; 390 391 vfs_unbusy(mp); 392 return (error); 393 } 394 395 #ifdef COMPAT_40 396 /* ARGSUSED */ 397 int 398 compat_40_sys_mount(struct lwp *l, void *v, register_t *retval) 399 { 400 struct compat_40_sys_mount_args /* { 401 syscallarg(const char *) type; 402 syscallarg(const char *) path; 403 syscallarg(int) flags; 404 syscallarg(void *) data; 405 } */ *uap = v; 406 register_t dummy; 407 408 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 409 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 410 } 411 #endif 412 413 int 414 sys___mount50(struct lwp *l, void *v, register_t *retval) 415 { 416 struct sys___mount50_args /* { 417 syscallarg(const char *) type; 418 syscallarg(const char *) path; 419 syscallarg(int) flags; 420 syscallarg(void *) data; 421 syscallarg(size_t) data_len; 422 } */ *uap = v; 423 424 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 425 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 426 SCARG(uap, data_len), retval); 427 } 428 429 int 430 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 431 const char *path, int flags, void *data, enum uio_seg data_seg, 432 size_t data_len, register_t *retval) 433 { 434 struct vnode *vp; 435 struct nameidata nd; 436 void *data_buf = data; 437 int error; 438 439 /* 440 * Get vnode to be covered 441 */ 442 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 443 if ((error = namei(&nd)) != 0) 444 return (error); 445 vp = nd.ni_vp; 446 447 /* 448 * A lookup in VFS_MOUNT might result in an attempt to 449 * lock this vnode again, so make the lock recursive. 450 */ 451 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 452 453 if (vfsops == NULL) { 454 if (flags & (MNT_GETARGS | MNT_UPDATE)) 455 vfsops = vp->v_mount->mnt_op; 456 else { 457 /* 'type' is userspace */ 458 error = mount_get_vfsops(type, &vfsops); 459 if (error != 0) 460 goto done; 461 } 462 } 463 464 if (data != NULL && data_seg == UIO_USERSPACE) { 465 if (data_len == 0) { 466 /* No length supplied, use default for filesystem */ 467 data_len = vfsops->vfs_min_mount_data; 468 if (data_len > VFS_MAX_MOUNT_DATA) { 469 /* maybe a force loaded old LKM */ 470 error = EINVAL; 471 goto done; 472 } 473 #ifdef COMPAT_30 474 /* Hopefully a longer buffer won't make copyin() fail */ 475 if (flags & MNT_UPDATE 476 && data_len < sizeof (struct mnt_export_args30)) 477 data_len = sizeof (struct mnt_export_args30); 478 #endif 479 } 480 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 481 482 /* NFS needs the buffer even for mnt_getargs .... */ 483 error = copyin(data, data_buf, data_len); 484 if (error != 0) 485 goto done; 486 } 487 488 if (flags & MNT_GETARGS) { 489 if (data_len == 0) { 490 error = EINVAL; 491 goto done; 492 } 493 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 494 if (error != 0) 495 goto done; 496 if (data_seg == UIO_USERSPACE) 497 error = copyout(data_buf, data, data_len); 498 *retval = data_len; 499 } else if (flags & MNT_UPDATE) { 500 error = mount_update(l, vp, path, flags, data_buf, &data_len); 501 } else { 502 /* Locking is handled internally in mount_domount(). */ 503 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 504 &data_len); 505 } 506 507 done: 508 if (vp) 509 vput(vp); 510 if (data_buf != data) 511 free(data_buf, M_TEMP); 512 return (error); 513 } 514 515 /* 516 * Scan all active processes to see if any of them have a current 517 * or root directory onto which the new filesystem has just been 518 * mounted. If so, replace them with the new mount point. 519 */ 520 void 521 checkdirs(struct vnode *olddp) 522 { 523 struct cwdinfo *cwdi; 524 struct vnode *newdp; 525 struct proc *p; 526 527 if (olddp->v_usecount == 1) 528 return; 529 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 530 panic("mount: lost mount"); 531 mutex_enter(&proclist_lock); 532 PROCLIST_FOREACH(p, &allproc) { 533 cwdi = p->p_cwdi; 534 if (!cwdi) 535 continue; 536 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 537 if (cwdi->cwdi_cdir == olddp) { 538 vrele(cwdi->cwdi_cdir); 539 VREF(newdp); 540 cwdi->cwdi_cdir = newdp; 541 } 542 if (cwdi->cwdi_rdir == olddp) { 543 vrele(cwdi->cwdi_rdir); 544 VREF(newdp); 545 cwdi->cwdi_rdir = newdp; 546 } 547 rw_exit(&cwdi->cwdi_lock); 548 } 549 mutex_exit(&proclist_lock); 550 if (rootvnode == olddp) { 551 vrele(rootvnode); 552 VREF(newdp); 553 rootvnode = newdp; 554 } 555 vput(newdp); 556 } 557 558 /* 559 * Unmount a file system. 560 * 561 * Note: unmount takes a path to the vnode mounted on as argument, 562 * not special file (as before). 563 */ 564 /* ARGSUSED */ 565 int 566 sys_unmount(struct lwp *l, void *v, register_t *retval) 567 { 568 struct sys_unmount_args /* { 569 syscallarg(const char *) path; 570 syscallarg(int) flags; 571 } */ *uap = v; 572 struct vnode *vp; 573 struct mount *mp; 574 int error; 575 struct nameidata nd; 576 577 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 578 SCARG(uap, path), l); 579 if ((error = namei(&nd)) != 0) 580 return (error); 581 vp = nd.ni_vp; 582 mp = vp->v_mount; 583 584 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 585 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 586 if (error) { 587 vput(vp); 588 return (error); 589 } 590 591 /* 592 * Don't allow unmounting the root file system. 593 */ 594 if (mp->mnt_flag & MNT_ROOTFS) { 595 vput(vp); 596 return (EINVAL); 597 } 598 599 /* 600 * Must be the root of the filesystem 601 */ 602 if ((vp->v_vflag & VV_ROOT) == 0) { 603 vput(vp); 604 return (EINVAL); 605 } 606 vput(vp); 607 608 /* 609 * XXX Freeze syncer. Must do this before locking the 610 * mount point. See dounmount() for details. 611 */ 612 mutex_enter(&syncer_mutex); 613 614 if (vfs_busy(mp, 0, 0)) { 615 mutex_exit(&syncer_mutex); 616 return (EBUSY); 617 } 618 619 return (dounmount(mp, SCARG(uap, flags), l)); 620 } 621 622 /* 623 * Do the actual file system unmount. File system is assumed to have been 624 * marked busy by the caller. 625 */ 626 int 627 dounmount(struct mount *mp, int flags, struct lwp *l) 628 { 629 struct vnode *coveredvp; 630 int error; 631 int async; 632 int used_syncer; 633 634 #if NVERIEXEC > 0 635 error = veriexec_unmountchk(mp); 636 if (error) 637 return (error); 638 #endif /* NVERIEXEC > 0 */ 639 640 mutex_enter(&mountlist_lock); 641 vfs_unbusy(mp); 642 used_syncer = (mp->mnt_syncer != NULL); 643 644 /* 645 * XXX Syncer must be frozen when we get here. This should really 646 * be done on a per-mountpoint basis, but especially the softdep 647 * code possibly called from the syncer doesn't exactly work on a 648 * per-mountpoint basis, so the softdep code would become a maze 649 * of vfs_busy() calls. 650 * 651 * The caller of dounmount() must acquire syncer_mutex because 652 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 653 * order, and we must preserve that order to avoid deadlock. 654 * 655 * So, if the file system did not use the syncer, now is 656 * the time to release the syncer_mutex. 657 */ 658 if (used_syncer == 0) 659 mutex_exit(&syncer_mutex); 660 661 mp->mnt_iflag |= IMNT_UNMOUNT; 662 mp->mnt_unmounter = l; 663 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 664 665 async = mp->mnt_flag & MNT_ASYNC; 666 mp->mnt_flag &= ~MNT_ASYNC; 667 cache_purgevfs(mp); /* remove cache entries for this file sys */ 668 if (mp->mnt_syncer != NULL) 669 vfs_deallocate_syncvnode(mp); 670 error = 0; 671 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 672 #if NFSS > 0 673 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 674 #endif 675 if (error == 0) 676 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 677 } 678 if (error == 0 || (flags & MNT_FORCE)) 679 error = VFS_UNMOUNT(mp, flags, l); 680 if (error) { 681 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 682 (void) vfs_allocate_syncvnode(mp); 683 mutex_enter(&mountlist_lock); 684 mp->mnt_iflag &= ~IMNT_UNMOUNT; 685 mp->mnt_unmounter = NULL; 686 mp->mnt_flag |= async; 687 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 688 &mountlist_lock); 689 if (used_syncer) 690 mutex_exit(&syncer_mutex); 691 mutex_enter(&mp->mnt_mutex); 692 while (mp->mnt_wcnt > 0) { 693 wakeup(mp); 694 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 695 0, &mp->mnt_mutex); 696 } 697 mutex_exit(&mp->mnt_mutex); 698 return (error); 699 } 700 vfs_scrubvnlist(mp); 701 mutex_enter(&mountlist_lock); 702 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 703 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 704 coveredvp->v_mountedhere = NULL; 705 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 706 panic("unmount: dangling vnode"); 707 mp->mnt_iflag |= IMNT_GONE; 708 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 709 if (coveredvp != NULLVP) 710 vrele(coveredvp); 711 if (used_syncer) 712 mutex_exit(&syncer_mutex); 713 mutex_enter(&mp->mnt_mutex); 714 while (mp->mnt_wcnt > 0) { 715 wakeup(mp); 716 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 717 } 718 mutex_exit(&mp->mnt_mutex); 719 vfs_hooks_unmount(mp); 720 vfs_delref(mp->mnt_op); 721 vfs_destroy(mp); 722 return (0); 723 } 724 725 /* 726 * Sync each mounted filesystem. 727 */ 728 #ifdef DEBUG 729 int syncprt = 0; 730 struct ctldebug debug0 = { "syncprt", &syncprt }; 731 #endif 732 733 /* ARGSUSED */ 734 int 735 sys_sync(struct lwp *l, void *v, register_t *retval) 736 { 737 struct mount *mp, *nmp; 738 int asyncflag; 739 740 if (l == NULL) 741 l = &lwp0; 742 743 mutex_enter(&mountlist_lock); 744 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 745 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 746 nmp = mp->mnt_list.cqe_prev; 747 continue; 748 } 749 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 750 asyncflag = mp->mnt_flag & MNT_ASYNC; 751 mp->mnt_flag &= ~MNT_ASYNC; 752 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 753 if (asyncflag) 754 mp->mnt_flag |= MNT_ASYNC; 755 } 756 mutex_enter(&mountlist_lock); 757 nmp = mp->mnt_list.cqe_prev; 758 vfs_unbusy(mp); 759 760 } 761 mutex_exit(&mountlist_lock); 762 #ifdef DEBUG 763 if (syncprt) 764 vfs_bufstats(); 765 #endif /* DEBUG */ 766 return (0); 767 } 768 769 /* 770 * Change filesystem quotas. 771 */ 772 /* ARGSUSED */ 773 int 774 sys_quotactl(struct lwp *l, void *v, register_t *retval) 775 { 776 struct sys_quotactl_args /* { 777 syscallarg(const char *) path; 778 syscallarg(int) cmd; 779 syscallarg(int) uid; 780 syscallarg(void *) arg; 781 } */ *uap = v; 782 struct mount *mp; 783 int error; 784 struct nameidata nd; 785 786 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 787 if ((error = namei(&nd)) != 0) 788 return (error); 789 mp = nd.ni_vp->v_mount; 790 vrele(nd.ni_vp); 791 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 792 SCARG(uap, arg), l); 793 return (error); 794 } 795 796 int 797 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 798 int root) 799 { 800 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 801 int error = 0; 802 803 /* 804 * If MNT_NOWAIT or MNT_LAZY is specified, do not 805 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 806 * overrides MNT_NOWAIT. 807 */ 808 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 809 (flags != MNT_WAIT && flags != 0)) { 810 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 811 goto done; 812 } 813 814 /* Get the filesystem stats now */ 815 memset(sp, 0, sizeof(*sp)); 816 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 817 return error; 818 } 819 820 if (cwdi->cwdi_rdir == NULL) 821 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 822 done: 823 if (cwdi->cwdi_rdir != NULL) { 824 size_t len; 825 char *bp; 826 char *path = PNBUF_GET(); 827 828 bp = path + MAXPATHLEN; 829 *--bp = '\0'; 830 rw_enter(&cwdi->cwdi_lock, RW_READER); 831 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 832 MAXPATHLEN / 2, 0, l); 833 rw_exit(&cwdi->cwdi_lock); 834 if (error) { 835 PNBUF_PUT(path); 836 return error; 837 } 838 len = strlen(bp); 839 /* 840 * for mount points that are below our root, we can see 841 * them, so we fix up the pathname and return them. The 842 * rest we cannot see, so we don't allow viewing the 843 * data. 844 */ 845 if (strncmp(bp, sp->f_mntonname, len) == 0) { 846 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 847 sizeof(sp->f_mntonname)); 848 if (sp->f_mntonname[0] == '\0') 849 (void)strlcpy(sp->f_mntonname, "/", 850 sizeof(sp->f_mntonname)); 851 } else { 852 if (root) 853 (void)strlcpy(sp->f_mntonname, "/", 854 sizeof(sp->f_mntonname)); 855 else 856 error = EPERM; 857 } 858 PNBUF_PUT(path); 859 } 860 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 861 return error; 862 } 863 864 /* 865 * Get filesystem statistics by path. 866 */ 867 int 868 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 869 { 870 struct mount *mp; 871 int error; 872 struct nameidata nd; 873 874 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 875 if ((error = namei(&nd)) != 0) 876 return error; 877 mp = nd.ni_vp->v_mount; 878 error = dostatvfs(mp, sb, l, flags, 1); 879 vrele(nd.ni_vp); 880 return error; 881 } 882 883 /* ARGSUSED */ 884 int 885 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 886 { 887 struct sys_statvfs1_args /* { 888 syscallarg(const char *) path; 889 syscallarg(struct statvfs *) buf; 890 syscallarg(int) flags; 891 } */ *uap = v; 892 struct statvfs *sb; 893 int error; 894 895 sb = STATVFSBUF_GET(); 896 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 897 if (error == 0) 898 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 899 STATVFSBUF_PUT(sb); 900 return error; 901 } 902 903 /* 904 * Get filesystem statistics by fd. 905 */ 906 int 907 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 908 { 909 struct proc *p = l->l_proc; 910 struct file *fp; 911 struct mount *mp; 912 int error; 913 914 /* getvnode() will use the descriptor for us */ 915 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 916 return (error); 917 mp = ((struct vnode *)fp->f_data)->v_mount; 918 error = dostatvfs(mp, sb, l, flags, 1); 919 FILE_UNUSE(fp, l); 920 return error; 921 } 922 923 /* ARGSUSED */ 924 int 925 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 926 { 927 struct sys_fstatvfs1_args /* { 928 syscallarg(int) fd; 929 syscallarg(struct statvfs *) buf; 930 syscallarg(int) flags; 931 } */ *uap = v; 932 struct statvfs *sb; 933 int error; 934 935 sb = STATVFSBUF_GET(); 936 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 937 if (error == 0) 938 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 939 STATVFSBUF_PUT(sb); 940 return error; 941 } 942 943 944 /* 945 * Get statistics on all filesystems. 946 */ 947 int 948 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 949 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 950 register_t *retval) 951 { 952 int root = 0; 953 struct proc *p = l->l_proc; 954 struct mount *mp, *nmp; 955 struct statvfs *sb; 956 size_t count, maxcount; 957 int error = 0; 958 959 sb = STATVFSBUF_GET(); 960 maxcount = bufsize / entry_sz; 961 mutex_enter(&mountlist_lock); 962 count = 0; 963 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 964 mp = nmp) { 965 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 966 nmp = CIRCLEQ_NEXT(mp, mnt_list); 967 continue; 968 } 969 if (sfsp && count < maxcount) { 970 error = dostatvfs(mp, sb, l, flags, 0); 971 if (error) { 972 mutex_enter(&mountlist_lock); 973 nmp = CIRCLEQ_NEXT(mp, mnt_list); 974 vfs_unbusy(mp); 975 continue; 976 } 977 error = copyfn(sb, sfsp, entry_sz); 978 if (error) { 979 vfs_unbusy(mp); 980 goto out; 981 } 982 sfsp = (char *)sfsp + entry_sz; 983 root |= strcmp(sb->f_mntonname, "/") == 0; 984 } 985 count++; 986 mutex_enter(&mountlist_lock); 987 nmp = CIRCLEQ_NEXT(mp, mnt_list); 988 vfs_unbusy(mp); 989 } 990 mutex_exit(&mountlist_lock); 991 if (root == 0 && p->p_cwdi->cwdi_rdir) { 992 /* 993 * fake a root entry 994 */ 995 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 996 if (error != 0) 997 goto out; 998 if (sfsp) 999 error = copyfn(sb, sfsp, entry_sz); 1000 count++; 1001 } 1002 if (sfsp && count > maxcount) 1003 *retval = maxcount; 1004 else 1005 *retval = count; 1006 out: 1007 STATVFSBUF_PUT(sb); 1008 return error; 1009 } 1010 1011 int 1012 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 1013 { 1014 struct sys_getvfsstat_args /* { 1015 syscallarg(struct statvfs *) buf; 1016 syscallarg(size_t) bufsize; 1017 syscallarg(int) flags; 1018 } */ *uap = v; 1019 1020 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1021 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1022 } 1023 1024 /* 1025 * Change current working directory to a given file descriptor. 1026 */ 1027 /* ARGSUSED */ 1028 int 1029 sys_fchdir(struct lwp *l, void *v, register_t *retval) 1030 { 1031 struct sys_fchdir_args /* { 1032 syscallarg(int) fd; 1033 } */ *uap = v; 1034 struct proc *p = l->l_proc; 1035 struct filedesc *fdp = p->p_fd; 1036 struct cwdinfo *cwdi; 1037 struct vnode *vp, *tdp; 1038 struct mount *mp; 1039 struct file *fp; 1040 int error; 1041 1042 /* getvnode() will use the descriptor for us */ 1043 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1044 return (error); 1045 vp = (struct vnode *)fp->f_data; 1046 1047 VREF(vp); 1048 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1049 if (vp->v_type != VDIR) 1050 error = ENOTDIR; 1051 else 1052 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1053 if (error) { 1054 vput(vp); 1055 goto out; 1056 } 1057 while ((mp = vp->v_mountedhere) != NULL) { 1058 if (vfs_busy(mp, 0, 0)) 1059 continue; 1060 1061 vput(vp); 1062 error = VFS_ROOT(mp, &tdp); 1063 vfs_unbusy(mp); 1064 if (error) 1065 goto out; 1066 vp = tdp; 1067 } 1068 VOP_UNLOCK(vp, 0); 1069 1070 /* 1071 * Disallow changing to a directory not under the process's 1072 * current root directory (if there is one). 1073 */ 1074 cwdi = p->p_cwdi; 1075 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1076 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1077 vrele(vp); 1078 error = EPERM; /* operation not permitted */ 1079 } else { 1080 vrele(cwdi->cwdi_cdir); 1081 cwdi->cwdi_cdir = vp; 1082 } 1083 rw_exit(&cwdi->cwdi_lock); 1084 1085 out: 1086 FILE_UNUSE(fp, l); 1087 return (error); 1088 } 1089 1090 /* 1091 * Change this process's notion of the root directory to a given file 1092 * descriptor. 1093 */ 1094 int 1095 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1096 { 1097 struct sys_fchroot_args *uap = v; 1098 struct proc *p = l->l_proc; 1099 struct filedesc *fdp = p->p_fd; 1100 struct cwdinfo *cwdi; 1101 struct vnode *vp; 1102 struct file *fp; 1103 int error; 1104 1105 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1106 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1107 return error; 1108 /* getvnode() will use the descriptor for us */ 1109 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1110 return error; 1111 vp = (struct vnode *) fp->f_data; 1112 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1113 if (vp->v_type != VDIR) 1114 error = ENOTDIR; 1115 else 1116 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1117 VOP_UNLOCK(vp, 0); 1118 if (error) 1119 goto out; 1120 VREF(vp); 1121 1122 /* 1123 * Prevent escaping from chroot by putting the root under 1124 * the working directory. Silently chdir to / if we aren't 1125 * already there. 1126 */ 1127 cwdi = p->p_cwdi; 1128 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1129 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1130 /* 1131 * XXX would be more failsafe to change directory to a 1132 * deadfs node here instead 1133 */ 1134 vrele(cwdi->cwdi_cdir); 1135 VREF(vp); 1136 cwdi->cwdi_cdir = vp; 1137 } 1138 1139 if (cwdi->cwdi_rdir != NULL) 1140 vrele(cwdi->cwdi_rdir); 1141 cwdi->cwdi_rdir = vp; 1142 rw_exit(&cwdi->cwdi_lock); 1143 1144 out: 1145 FILE_UNUSE(fp, l); 1146 return (error); 1147 } 1148 1149 /* 1150 * Change current working directory (``.''). 1151 */ 1152 /* ARGSUSED */ 1153 int 1154 sys_chdir(struct lwp *l, void *v, register_t *retval) 1155 { 1156 struct sys_chdir_args /* { 1157 syscallarg(const char *) path; 1158 } */ *uap = v; 1159 struct proc *p = l->l_proc; 1160 struct cwdinfo *cwdi; 1161 int error; 1162 struct nameidata nd; 1163 1164 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1165 SCARG(uap, path), l); 1166 if ((error = change_dir(&nd, l)) != 0) 1167 return (error); 1168 cwdi = p->p_cwdi; 1169 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1170 vrele(cwdi->cwdi_cdir); 1171 cwdi->cwdi_cdir = nd.ni_vp; 1172 rw_exit(&cwdi->cwdi_lock); 1173 return (0); 1174 } 1175 1176 /* 1177 * Change notion of root (``/'') directory. 1178 */ 1179 /* ARGSUSED */ 1180 int 1181 sys_chroot(struct lwp *l, void *v, register_t *retval) 1182 { 1183 struct sys_chroot_args /* { 1184 syscallarg(const char *) path; 1185 } */ *uap = v; 1186 struct proc *p = l->l_proc; 1187 struct cwdinfo *cwdi; 1188 struct vnode *vp; 1189 int error; 1190 struct nameidata nd; 1191 1192 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1193 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1194 return (error); 1195 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1196 SCARG(uap, path), l); 1197 if ((error = change_dir(&nd, l)) != 0) 1198 return (error); 1199 1200 cwdi = p->p_cwdi; 1201 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1202 if (cwdi->cwdi_rdir != NULL) 1203 vrele(cwdi->cwdi_rdir); 1204 vp = nd.ni_vp; 1205 cwdi->cwdi_rdir = vp; 1206 1207 /* 1208 * Prevent escaping from chroot by putting the root under 1209 * the working directory. Silently chdir to / if we aren't 1210 * already there. 1211 */ 1212 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1213 /* 1214 * XXX would be more failsafe to change directory to a 1215 * deadfs node here instead 1216 */ 1217 vrele(cwdi->cwdi_cdir); 1218 VREF(vp); 1219 cwdi->cwdi_cdir = vp; 1220 } 1221 rw_exit(&cwdi->cwdi_lock); 1222 1223 return (0); 1224 } 1225 1226 /* 1227 * Common routine for chroot and chdir. 1228 */ 1229 static int 1230 change_dir(struct nameidata *ndp, struct lwp *l) 1231 { 1232 struct vnode *vp; 1233 int error; 1234 1235 if ((error = namei(ndp)) != 0) 1236 return (error); 1237 vp = ndp->ni_vp; 1238 if (vp->v_type != VDIR) 1239 error = ENOTDIR; 1240 else 1241 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1242 1243 if (error) 1244 vput(vp); 1245 else 1246 VOP_UNLOCK(vp, 0); 1247 return (error); 1248 } 1249 1250 /* 1251 * Check permissions, allocate an open file structure, 1252 * and call the device open routine if any. 1253 */ 1254 int 1255 sys_open(struct lwp *l, void *v, register_t *retval) 1256 { 1257 struct sys_open_args /* { 1258 syscallarg(const char *) path; 1259 syscallarg(int) flags; 1260 syscallarg(int) mode; 1261 } */ *uap = v; 1262 struct proc *p = l->l_proc; 1263 struct cwdinfo *cwdi = p->p_cwdi; 1264 struct filedesc *fdp = p->p_fd; 1265 struct file *fp; 1266 struct vnode *vp; 1267 int flags, cmode; 1268 int type, indx, error; 1269 struct flock lf; 1270 struct nameidata nd; 1271 1272 flags = FFLAGS(SCARG(uap, flags)); 1273 if ((flags & (FREAD | FWRITE)) == 0) 1274 return (EINVAL); 1275 /* falloc() will use the file descriptor for us */ 1276 if ((error = falloc(l, &fp, &indx)) != 0) 1277 return (error); 1278 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1279 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1280 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1281 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1282 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1283 rw_enter(&fdp->fd_lock, RW_WRITER); 1284 FILE_UNUSE(fp, l); 1285 fdp->fd_ofiles[indx] = NULL; 1286 rw_exit(&fdp->fd_lock); 1287 ffree(fp); 1288 if ((error == EDUPFD || error == EMOVEFD) && 1289 l->l_dupfd >= 0 && /* XXX from fdopen */ 1290 (error = 1291 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1292 *retval = indx; 1293 return (0); 1294 } 1295 if (error == ERESTART) 1296 error = EINTR; 1297 fdremove(fdp, indx); 1298 return (error); 1299 } 1300 l->l_dupfd = 0; 1301 vp = nd.ni_vp; 1302 fp->f_flag = flags & FMASK; 1303 fp->f_type = DTYPE_VNODE; 1304 fp->f_ops = &vnops; 1305 fp->f_data = vp; 1306 if (flags & (O_EXLOCK | O_SHLOCK)) { 1307 lf.l_whence = SEEK_SET; 1308 lf.l_start = 0; 1309 lf.l_len = 0; 1310 if (flags & O_EXLOCK) 1311 lf.l_type = F_WRLCK; 1312 else 1313 lf.l_type = F_RDLCK; 1314 type = F_FLOCK; 1315 if ((flags & FNONBLOCK) == 0) 1316 type |= F_WAIT; 1317 VOP_UNLOCK(vp, 0); 1318 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1319 if (error) { 1320 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1321 FILE_UNUSE(fp, l); 1322 ffree(fp); 1323 fdremove(fdp, indx); 1324 return (error); 1325 } 1326 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1327 fp->f_flag |= FHASLOCK; 1328 } 1329 VOP_UNLOCK(vp, 0); 1330 *retval = indx; 1331 FILE_SET_MATURE(fp); 1332 FILE_UNUSE(fp, l); 1333 return (0); 1334 } 1335 1336 static void 1337 vfs__fhfree(fhandle_t *fhp) 1338 { 1339 size_t fhsize; 1340 1341 if (fhp == NULL) { 1342 return; 1343 } 1344 fhsize = FHANDLE_SIZE(fhp); 1345 kmem_free(fhp, fhsize); 1346 } 1347 1348 /* 1349 * vfs_composefh: compose a filehandle. 1350 */ 1351 1352 int 1353 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1354 { 1355 struct mount *mp; 1356 struct fid *fidp; 1357 int error; 1358 size_t needfhsize; 1359 size_t fidsize; 1360 1361 mp = vp->v_mount; 1362 fidp = NULL; 1363 if (*fh_size < FHANDLE_SIZE_MIN) { 1364 fidsize = 0; 1365 } else { 1366 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1367 if (fhp != NULL) { 1368 memset(fhp, 0, *fh_size); 1369 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1370 fidp = &fhp->fh_fid; 1371 } 1372 } 1373 error = VFS_VPTOFH(vp, fidp, &fidsize); 1374 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1375 if (error == 0 && *fh_size < needfhsize) { 1376 error = E2BIG; 1377 } 1378 *fh_size = needfhsize; 1379 return error; 1380 } 1381 1382 int 1383 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1384 { 1385 struct mount *mp; 1386 fhandle_t *fhp; 1387 size_t fhsize; 1388 size_t fidsize; 1389 int error; 1390 1391 *fhpp = NULL; 1392 mp = vp->v_mount; 1393 fidsize = 0; 1394 error = VFS_VPTOFH(vp, NULL, &fidsize); 1395 KASSERT(error != 0); 1396 if (error != E2BIG) { 1397 goto out; 1398 } 1399 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1400 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1401 if (fhp == NULL) { 1402 error = ENOMEM; 1403 goto out; 1404 } 1405 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1406 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1407 if (error == 0) { 1408 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1409 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1410 *fhpp = fhp; 1411 } else { 1412 kmem_free(fhp, fhsize); 1413 } 1414 out: 1415 return error; 1416 } 1417 1418 void 1419 vfs_composefh_free(fhandle_t *fhp) 1420 { 1421 1422 vfs__fhfree(fhp); 1423 } 1424 1425 /* 1426 * vfs_fhtovp: lookup a vnode by a filehandle. 1427 */ 1428 1429 int 1430 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1431 { 1432 struct mount *mp; 1433 int error; 1434 1435 *vpp = NULL; 1436 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1437 if (mp == NULL) { 1438 error = ESTALE; 1439 goto out; 1440 } 1441 if (mp->mnt_op->vfs_fhtovp == NULL) { 1442 error = EOPNOTSUPP; 1443 goto out; 1444 } 1445 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1446 out: 1447 return error; 1448 } 1449 1450 /* 1451 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1452 * the needed size. 1453 */ 1454 1455 int 1456 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1457 { 1458 fhandle_t *fhp; 1459 int error; 1460 1461 *fhpp = NULL; 1462 if (fhsize > FHANDLE_SIZE_MAX) { 1463 return EINVAL; 1464 } 1465 if (fhsize < FHANDLE_SIZE_MIN) { 1466 return EINVAL; 1467 } 1468 again: 1469 fhp = kmem_alloc(fhsize, KM_SLEEP); 1470 if (fhp == NULL) { 1471 return ENOMEM; 1472 } 1473 error = copyin(ufhp, fhp, fhsize); 1474 if (error == 0) { 1475 /* XXX this check shouldn't be here */ 1476 if (FHANDLE_SIZE(fhp) == fhsize) { 1477 *fhpp = fhp; 1478 return 0; 1479 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1480 /* 1481 * a kludge for nfsv2 padded handles. 1482 */ 1483 size_t sz; 1484 1485 sz = FHANDLE_SIZE(fhp); 1486 kmem_free(fhp, fhsize); 1487 fhsize = sz; 1488 goto again; 1489 } else { 1490 /* 1491 * userland told us wrong size. 1492 */ 1493 error = EINVAL; 1494 } 1495 } 1496 kmem_free(fhp, fhsize); 1497 return error; 1498 } 1499 1500 void 1501 vfs_copyinfh_free(fhandle_t *fhp) 1502 { 1503 1504 vfs__fhfree(fhp); 1505 } 1506 1507 /* 1508 * Get file handle system call 1509 */ 1510 int 1511 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1512 { 1513 struct sys___getfh30_args /* { 1514 syscallarg(char *) fname; 1515 syscallarg(fhandle_t *) fhp; 1516 syscallarg(size_t *) fh_size; 1517 } */ *uap = v; 1518 struct vnode *vp; 1519 fhandle_t *fh; 1520 int error; 1521 struct nameidata nd; 1522 size_t sz; 1523 size_t usz; 1524 1525 /* 1526 * Must be super user 1527 */ 1528 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1529 0, NULL, NULL, NULL); 1530 if (error) 1531 return (error); 1532 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1533 SCARG(uap, fname), l); 1534 error = namei(&nd); 1535 if (error) 1536 return (error); 1537 vp = nd.ni_vp; 1538 error = vfs_composefh_alloc(vp, &fh); 1539 vput(vp); 1540 if (error != 0) { 1541 goto out; 1542 } 1543 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1544 if (error != 0) { 1545 goto out; 1546 } 1547 sz = FHANDLE_SIZE(fh); 1548 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1549 if (error != 0) { 1550 goto out; 1551 } 1552 if (usz >= sz) { 1553 error = copyout(fh, SCARG(uap, fhp), sz); 1554 } else { 1555 error = E2BIG; 1556 } 1557 out: 1558 vfs_composefh_free(fh); 1559 return (error); 1560 } 1561 1562 /* 1563 * Open a file given a file handle. 1564 * 1565 * Check permissions, allocate an open file structure, 1566 * and call the device open routine if any. 1567 */ 1568 1569 int 1570 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1571 register_t *retval) 1572 { 1573 struct filedesc *fdp = l->l_proc->p_fd; 1574 struct file *fp; 1575 struct vnode *vp = NULL; 1576 kauth_cred_t cred = l->l_cred; 1577 struct file *nfp; 1578 int type, indx, error=0; 1579 struct flock lf; 1580 struct vattr va; 1581 fhandle_t *fh; 1582 int flags; 1583 1584 /* 1585 * Must be super user 1586 */ 1587 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1588 0, NULL, NULL, NULL))) 1589 return (error); 1590 1591 flags = FFLAGS(oflags); 1592 if ((flags & (FREAD | FWRITE)) == 0) 1593 return (EINVAL); 1594 if ((flags & O_CREAT)) 1595 return (EINVAL); 1596 /* falloc() will use the file descriptor for us */ 1597 if ((error = falloc(l, &nfp, &indx)) != 0) 1598 return (error); 1599 fp = nfp; 1600 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1601 if (error != 0) { 1602 goto bad; 1603 } 1604 error = vfs_fhtovp(fh, &vp); 1605 if (error != 0) { 1606 goto bad; 1607 } 1608 1609 /* Now do an effective vn_open */ 1610 1611 if (vp->v_type == VSOCK) { 1612 error = EOPNOTSUPP; 1613 goto bad; 1614 } 1615 if (flags & FREAD) { 1616 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1617 goto bad; 1618 } 1619 if (flags & (FWRITE | O_TRUNC)) { 1620 if (vp->v_type == VDIR) { 1621 error = EISDIR; 1622 goto bad; 1623 } 1624 if ((error = vn_writechk(vp)) != 0 || 1625 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1626 goto bad; 1627 } 1628 if (flags & O_TRUNC) { 1629 VOP_UNLOCK(vp, 0); /* XXX */ 1630 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1631 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1632 VATTR_NULL(&va); 1633 va.va_size = 0; 1634 error = VOP_SETATTR(vp, &va, cred, l); 1635 if (error) 1636 goto bad; 1637 } 1638 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1639 goto bad; 1640 if (flags & FWRITE) 1641 vp->v_writecount++; 1642 1643 /* done with modified vn_open, now finish what sys_open does. */ 1644 1645 fp->f_flag = flags & FMASK; 1646 fp->f_type = DTYPE_VNODE; 1647 fp->f_ops = &vnops; 1648 fp->f_data = vp; 1649 if (flags & (O_EXLOCK | O_SHLOCK)) { 1650 lf.l_whence = SEEK_SET; 1651 lf.l_start = 0; 1652 lf.l_len = 0; 1653 if (flags & O_EXLOCK) 1654 lf.l_type = F_WRLCK; 1655 else 1656 lf.l_type = F_RDLCK; 1657 type = F_FLOCK; 1658 if ((flags & FNONBLOCK) == 0) 1659 type |= F_WAIT; 1660 VOP_UNLOCK(vp, 0); 1661 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1662 if (error) { 1663 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1664 FILE_UNUSE(fp, l); 1665 ffree(fp); 1666 fdremove(fdp, indx); 1667 return (error); 1668 } 1669 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1670 fp->f_flag |= FHASLOCK; 1671 } 1672 VOP_UNLOCK(vp, 0); 1673 *retval = indx; 1674 FILE_SET_MATURE(fp); 1675 FILE_UNUSE(fp, l); 1676 vfs_copyinfh_free(fh); 1677 return (0); 1678 1679 bad: 1680 FILE_UNUSE(fp, l); 1681 ffree(fp); 1682 fdremove(fdp, indx); 1683 if (vp != NULL) 1684 vput(vp); 1685 vfs_copyinfh_free(fh); 1686 return (error); 1687 } 1688 1689 int 1690 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1691 { 1692 struct sys___fhopen40_args /* { 1693 syscallarg(const void *) fhp; 1694 syscallarg(size_t) fh_size; 1695 syscallarg(int) flags; 1696 } */ *uap = v; 1697 1698 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1699 SCARG(uap, flags), retval); 1700 } 1701 1702 int 1703 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1704 { 1705 int error; 1706 fhandle_t *fh; 1707 struct vnode *vp; 1708 1709 /* 1710 * Must be super user 1711 */ 1712 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1713 0, NULL, NULL, NULL))) 1714 return (error); 1715 1716 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1717 if (error != 0) 1718 return error; 1719 1720 error = vfs_fhtovp(fh, &vp); 1721 vfs_copyinfh_free(fh); 1722 if (error != 0) 1723 return error; 1724 1725 error = vn_stat(vp, sb, l); 1726 vput(vp); 1727 return error; 1728 } 1729 1730 1731 /* ARGSUSED */ 1732 int 1733 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1734 { 1735 struct sys___fhstat40_args /* { 1736 syscallarg(const void *) fhp; 1737 syscallarg(size_t) fh_size; 1738 syscallarg(struct stat *) sb; 1739 } */ *uap = v; 1740 struct stat sb; 1741 int error; 1742 1743 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1744 if (error) 1745 return error; 1746 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1747 } 1748 1749 int 1750 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1751 int flags) 1752 { 1753 fhandle_t *fh; 1754 struct mount *mp; 1755 struct vnode *vp; 1756 int error; 1757 1758 /* 1759 * Must be super user 1760 */ 1761 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1762 0, NULL, NULL, NULL))) 1763 return error; 1764 1765 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1766 if (error != 0) 1767 return error; 1768 1769 error = vfs_fhtovp(fh, &vp); 1770 vfs_copyinfh_free(fh); 1771 if (error != 0) 1772 return error; 1773 1774 mp = vp->v_mount; 1775 error = dostatvfs(mp, sb, l, flags, 1); 1776 vput(vp); 1777 return error; 1778 } 1779 1780 /* ARGSUSED */ 1781 int 1782 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1783 { 1784 struct sys___fhstatvfs140_args /* { 1785 syscallarg(const void *) fhp; 1786 syscallarg(size_t) fh_size; 1787 syscallarg(struct statvfs *) buf; 1788 syscallarg(int) flags; 1789 } */ *uap = v; 1790 struct statvfs *sb = STATVFSBUF_GET(); 1791 int error; 1792 1793 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1794 SCARG(uap, flags)); 1795 if (error == 0) 1796 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1797 STATVFSBUF_PUT(sb); 1798 return error; 1799 } 1800 1801 /* 1802 * Create a special file. 1803 */ 1804 /* ARGSUSED */ 1805 int 1806 sys_mknod(struct lwp *l, void *v, register_t *retval) 1807 { 1808 struct sys_mknod_args /* { 1809 syscallarg(const char *) path; 1810 syscallarg(int) mode; 1811 syscallarg(int) dev; 1812 } */ *uap = v; 1813 struct proc *p = l->l_proc; 1814 struct vnode *vp; 1815 struct vattr vattr; 1816 int error, optype; 1817 struct nameidata nd; 1818 char *path; 1819 const char *cpath; 1820 enum uio_seg seg = UIO_USERSPACE; 1821 1822 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1823 0, NULL, NULL, NULL)) != 0) 1824 return (error); 1825 1826 optype = VOP_MKNOD_DESCOFFSET; 1827 1828 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1829 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1830 1831 if ((error = namei(&nd)) != 0) 1832 goto out; 1833 vp = nd.ni_vp; 1834 if (vp != NULL) 1835 error = EEXIST; 1836 else { 1837 VATTR_NULL(&vattr); 1838 /* We will read cwdi->cwdi_cmask unlocked. */ 1839 vattr.va_mode = 1840 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1841 vattr.va_rdev = SCARG(uap, dev); 1842 1843 switch (SCARG(uap, mode) & S_IFMT) { 1844 case S_IFMT: /* used by badsect to flag bad sectors */ 1845 vattr.va_type = VBAD; 1846 break; 1847 case S_IFCHR: 1848 vattr.va_type = VCHR; 1849 break; 1850 case S_IFBLK: 1851 vattr.va_type = VBLK; 1852 break; 1853 case S_IFWHT: 1854 optype = VOP_WHITEOUT_DESCOFFSET; 1855 break; 1856 case S_IFREG: 1857 #if NVERIEXEC > 0 1858 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1859 O_CREAT); 1860 #endif /* NVERIEXEC > 0 */ 1861 vattr.va_type = VREG; 1862 vattr.va_rdev = VNOVAL; 1863 optype = VOP_CREATE_DESCOFFSET; 1864 break; 1865 default: 1866 error = EINVAL; 1867 break; 1868 } 1869 } 1870 if (!error) { 1871 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1872 switch (optype) { 1873 case VOP_WHITEOUT_DESCOFFSET: 1874 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1875 if (error) 1876 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1877 vput(nd.ni_dvp); 1878 break; 1879 1880 case VOP_MKNOD_DESCOFFSET: 1881 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1882 &nd.ni_cnd, &vattr); 1883 if (error == 0) 1884 vput(nd.ni_vp); 1885 break; 1886 1887 case VOP_CREATE_DESCOFFSET: 1888 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1889 &nd.ni_cnd, &vattr); 1890 if (error == 0) 1891 vput(nd.ni_vp); 1892 break; 1893 } 1894 } else { 1895 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1896 if (nd.ni_dvp == vp) 1897 vrele(nd.ni_dvp); 1898 else 1899 vput(nd.ni_dvp); 1900 if (vp) 1901 vrele(vp); 1902 } 1903 out: 1904 VERIEXEC_PATH_PUT(path); 1905 return (error); 1906 } 1907 1908 /* 1909 * Create a named pipe. 1910 */ 1911 /* ARGSUSED */ 1912 int 1913 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1914 { 1915 struct sys_mkfifo_args /* { 1916 syscallarg(const char *) path; 1917 syscallarg(int) mode; 1918 } */ *uap = v; 1919 struct proc *p = l->l_proc; 1920 struct vattr vattr; 1921 int error; 1922 struct nameidata nd; 1923 1924 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1925 if ((error = namei(&nd)) != 0) 1926 return (error); 1927 if (nd.ni_vp != NULL) { 1928 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1929 if (nd.ni_dvp == nd.ni_vp) 1930 vrele(nd.ni_dvp); 1931 else 1932 vput(nd.ni_dvp); 1933 vrele(nd.ni_vp); 1934 return (EEXIST); 1935 } 1936 VATTR_NULL(&vattr); 1937 vattr.va_type = VFIFO; 1938 /* We will read cwdi->cwdi_cmask unlocked. */ 1939 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1940 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1941 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1942 if (error == 0) 1943 vput(nd.ni_vp); 1944 return (error); 1945 } 1946 1947 /* 1948 * Make a hard file link. 1949 */ 1950 /* ARGSUSED */ 1951 int 1952 sys_link(struct lwp *l, void *v, register_t *retval) 1953 { 1954 struct sys_link_args /* { 1955 syscallarg(const char *) path; 1956 syscallarg(const char *) link; 1957 } */ *uap = v; 1958 struct vnode *vp; 1959 struct nameidata nd; 1960 int error; 1961 1962 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1963 if ((error = namei(&nd)) != 0) 1964 return (error); 1965 vp = nd.ni_vp; 1966 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1967 if ((error = namei(&nd)) != 0) 1968 goto out; 1969 if (nd.ni_vp) { 1970 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1971 if (nd.ni_dvp == nd.ni_vp) 1972 vrele(nd.ni_dvp); 1973 else 1974 vput(nd.ni_dvp); 1975 vrele(nd.ni_vp); 1976 error = EEXIST; 1977 goto out; 1978 } 1979 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1980 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1981 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1982 out: 1983 vrele(vp); 1984 return (error); 1985 } 1986 1987 /* 1988 * Make a symbolic link. 1989 */ 1990 /* ARGSUSED */ 1991 int 1992 sys_symlink(struct lwp *l, void *v, register_t *retval) 1993 { 1994 struct sys_symlink_args /* { 1995 syscallarg(const char *) path; 1996 syscallarg(const char *) link; 1997 } */ *uap = v; 1998 struct proc *p = l->l_proc; 1999 struct vattr vattr; 2000 char *path; 2001 int error; 2002 struct nameidata nd; 2003 2004 path = PNBUF_GET(); 2005 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2006 if (error) 2007 goto out; 2008 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 2009 if ((error = namei(&nd)) != 0) 2010 goto out; 2011 if (nd.ni_vp) { 2012 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2013 if (nd.ni_dvp == nd.ni_vp) 2014 vrele(nd.ni_dvp); 2015 else 2016 vput(nd.ni_dvp); 2017 vrele(nd.ni_vp); 2018 error = EEXIST; 2019 goto out; 2020 } 2021 VATTR_NULL(&vattr); 2022 vattr.va_type = VLNK; 2023 /* We will read cwdi->cwdi_cmask unlocked. */ 2024 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2025 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2026 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2027 if (error == 0) 2028 vput(nd.ni_vp); 2029 out: 2030 PNBUF_PUT(path); 2031 return (error); 2032 } 2033 2034 /* 2035 * Delete a whiteout from the filesystem. 2036 */ 2037 /* ARGSUSED */ 2038 int 2039 sys_undelete(struct lwp *l, void *v, register_t *retval) 2040 { 2041 struct sys_undelete_args /* { 2042 syscallarg(const char *) path; 2043 } */ *uap = v; 2044 int error; 2045 struct nameidata nd; 2046 2047 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 2048 SCARG(uap, path), l); 2049 error = namei(&nd); 2050 if (error) 2051 return (error); 2052 2053 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2054 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2055 if (nd.ni_dvp == nd.ni_vp) 2056 vrele(nd.ni_dvp); 2057 else 2058 vput(nd.ni_dvp); 2059 if (nd.ni_vp) 2060 vrele(nd.ni_vp); 2061 return (EEXIST); 2062 } 2063 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2064 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2065 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2066 vput(nd.ni_dvp); 2067 return (error); 2068 } 2069 2070 /* 2071 * Delete a name from the filesystem. 2072 */ 2073 /* ARGSUSED */ 2074 int 2075 sys_unlink(struct lwp *l, void *v, register_t *retval) 2076 { 2077 struct sys_unlink_args /* { 2078 syscallarg(const char *) path; 2079 } */ *uap = v; 2080 struct vnode *vp; 2081 int error; 2082 struct nameidata nd; 2083 char *path; 2084 const char *cpath; 2085 enum uio_seg seg = UIO_USERSPACE; 2086 2087 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2088 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2089 2090 if ((error = namei(&nd)) != 0) 2091 goto out; 2092 vp = nd.ni_vp; 2093 2094 /* 2095 * The root of a mounted filesystem cannot be deleted. 2096 */ 2097 if (vp->v_vflag & VV_ROOT) { 2098 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2099 if (nd.ni_dvp == vp) 2100 vrele(nd.ni_dvp); 2101 else 2102 vput(nd.ni_dvp); 2103 vput(vp); 2104 error = EBUSY; 2105 goto out; 2106 } 2107 2108 #if NVERIEXEC > 0 2109 /* Handle remove requests for veriexec entries. */ 2110 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2111 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2112 if (nd.ni_dvp == vp) 2113 vrele(nd.ni_dvp); 2114 else 2115 vput(nd.ni_dvp); 2116 vput(vp); 2117 goto out; 2118 } 2119 #endif /* NVERIEXEC > 0 */ 2120 2121 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2122 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2123 #ifdef FILEASSOC 2124 (void)fileassoc_file_delete(vp); 2125 #endif /* FILEASSOC */ 2126 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2127 out: 2128 VERIEXEC_PATH_PUT(path); 2129 return (error); 2130 } 2131 2132 /* 2133 * Reposition read/write file offset. 2134 */ 2135 int 2136 sys_lseek(struct lwp *l, void *v, register_t *retval) 2137 { 2138 struct sys_lseek_args /* { 2139 syscallarg(int) fd; 2140 syscallarg(int) pad; 2141 syscallarg(off_t) offset; 2142 syscallarg(int) whence; 2143 } */ *uap = v; 2144 struct proc *p = l->l_proc; 2145 kauth_cred_t cred = l->l_cred; 2146 struct filedesc *fdp = p->p_fd; 2147 struct file *fp; 2148 struct vnode *vp; 2149 struct vattr vattr; 2150 off_t newoff; 2151 int error; 2152 2153 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2154 return (EBADF); 2155 2156 vp = (struct vnode *)fp->f_data; 2157 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2158 error = ESPIPE; 2159 mutex_exit(&fp->f_lock); 2160 goto out; 2161 } 2162 2163 switch (SCARG(uap, whence)) { 2164 case SEEK_CUR: 2165 newoff = fp->f_offset + SCARG(uap, offset); 2166 FILE_USE(fp); 2167 break; 2168 case SEEK_END: 2169 FILE_USE(fp); 2170 error = VOP_GETATTR(vp, &vattr, cred, l); 2171 if (error) { 2172 FILE_UNUSE(fp, l); 2173 goto out; 2174 } 2175 newoff = SCARG(uap, offset) + vattr.va_size; 2176 break; 2177 case SEEK_SET: 2178 FILE_USE(fp); 2179 newoff = SCARG(uap, offset); 2180 break; 2181 default: 2182 mutex_exit(&fp->f_lock); 2183 error = EINVAL; 2184 goto out; 2185 } 2186 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2187 mutex_enter(&fp->f_lock); 2188 *(off_t *)retval = fp->f_offset = newoff; 2189 mutex_exit(&fp->f_lock); 2190 } 2191 FILE_UNUSE(fp, l); 2192 out: 2193 return (error); 2194 } 2195 2196 /* 2197 * Positional read system call. 2198 */ 2199 int 2200 sys_pread(struct lwp *l, void *v, register_t *retval) 2201 { 2202 struct sys_pread_args /* { 2203 syscallarg(int) fd; 2204 syscallarg(void *) buf; 2205 syscallarg(size_t) nbyte; 2206 syscallarg(off_t) offset; 2207 } */ *uap = v; 2208 struct proc *p = l->l_proc; 2209 struct filedesc *fdp = p->p_fd; 2210 struct file *fp; 2211 struct vnode *vp; 2212 off_t offset; 2213 int error, fd = SCARG(uap, fd); 2214 2215 if ((fp = fd_getfile(fdp, fd)) == NULL) 2216 return (EBADF); 2217 2218 if ((fp->f_flag & FREAD) == 0) { 2219 mutex_exit(&fp->f_lock); 2220 return (EBADF); 2221 } 2222 2223 FILE_USE(fp); 2224 2225 vp = (struct vnode *)fp->f_data; 2226 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2227 error = ESPIPE; 2228 goto out; 2229 } 2230 2231 offset = SCARG(uap, offset); 2232 2233 /* 2234 * XXX This works because no file systems actually 2235 * XXX take any action on the seek operation. 2236 */ 2237 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2238 goto out; 2239 2240 /* dofileread() will unuse the descriptor for us */ 2241 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2242 &offset, 0, retval)); 2243 2244 out: 2245 FILE_UNUSE(fp, l); 2246 return (error); 2247 } 2248 2249 /* 2250 * Positional scatter read system call. 2251 */ 2252 int 2253 sys_preadv(struct lwp *l, void *v, register_t *retval) 2254 { 2255 struct sys_preadv_args /* { 2256 syscallarg(int) fd; 2257 syscallarg(const struct iovec *) iovp; 2258 syscallarg(int) iovcnt; 2259 syscallarg(off_t) offset; 2260 } */ *uap = v; 2261 2262 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2263 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2264 } 2265 2266 /* 2267 * Positional write system call. 2268 */ 2269 int 2270 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2271 { 2272 struct sys_pwrite_args /* { 2273 syscallarg(int) fd; 2274 syscallarg(const void *) buf; 2275 syscallarg(size_t) nbyte; 2276 syscallarg(off_t) offset; 2277 } */ *uap = v; 2278 struct proc *p = l->l_proc; 2279 struct filedesc *fdp = p->p_fd; 2280 struct file *fp; 2281 struct vnode *vp; 2282 off_t offset; 2283 int error, fd = SCARG(uap, fd); 2284 2285 if ((fp = fd_getfile(fdp, fd)) == NULL) 2286 return (EBADF); 2287 2288 if ((fp->f_flag & FWRITE) == 0) { 2289 mutex_exit(&fp->f_lock); 2290 return (EBADF); 2291 } 2292 2293 FILE_USE(fp); 2294 2295 vp = (struct vnode *)fp->f_data; 2296 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2297 error = ESPIPE; 2298 goto out; 2299 } 2300 2301 offset = SCARG(uap, offset); 2302 2303 /* 2304 * XXX This works because no file systems actually 2305 * XXX take any action on the seek operation. 2306 */ 2307 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2308 goto out; 2309 2310 /* dofilewrite() will unuse the descriptor for us */ 2311 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2312 &offset, 0, retval)); 2313 2314 out: 2315 FILE_UNUSE(fp, l); 2316 return (error); 2317 } 2318 2319 /* 2320 * Positional gather write system call. 2321 */ 2322 int 2323 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2324 { 2325 struct sys_pwritev_args /* { 2326 syscallarg(int) fd; 2327 syscallarg(const struct iovec *) iovp; 2328 syscallarg(int) iovcnt; 2329 syscallarg(off_t) offset; 2330 } */ *uap = v; 2331 2332 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2333 SCARG(uap, iovcnt), &SCARG(uap, offset), 0, retval); 2334 } 2335 2336 /* 2337 * Check access permissions. 2338 */ 2339 int 2340 sys_access(struct lwp *l, void *v, register_t *retval) 2341 { 2342 struct sys_access_args /* { 2343 syscallarg(const char *) path; 2344 syscallarg(int) flags; 2345 } */ *uap = v; 2346 kauth_cred_t cred; 2347 struct vnode *vp; 2348 int error, flags; 2349 struct nameidata nd; 2350 2351 cred = kauth_cred_dup(l->l_cred); 2352 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2353 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2354 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2355 SCARG(uap, path), l); 2356 /* Override default credentials */ 2357 nd.ni_cnd.cn_cred = cred; 2358 if ((error = namei(&nd)) != 0) 2359 goto out; 2360 vp = nd.ni_vp; 2361 2362 /* Flags == 0 means only check for existence. */ 2363 if (SCARG(uap, flags)) { 2364 flags = 0; 2365 if (SCARG(uap, flags) & R_OK) 2366 flags |= VREAD; 2367 if (SCARG(uap, flags) & W_OK) 2368 flags |= VWRITE; 2369 if (SCARG(uap, flags) & X_OK) 2370 flags |= VEXEC; 2371 2372 error = VOP_ACCESS(vp, flags, cred, l); 2373 if (!error && (flags & VWRITE)) 2374 error = vn_writechk(vp); 2375 } 2376 vput(vp); 2377 out: 2378 kauth_cred_free(cred); 2379 return (error); 2380 } 2381 2382 /* 2383 * Common code for all sys_stat functions, including compat versions. 2384 */ 2385 int 2386 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2387 struct stat *sb) 2388 { 2389 int error; 2390 struct nameidata nd; 2391 2392 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2393 error = namei(&nd); 2394 if (error != 0) 2395 return error; 2396 error = vn_stat(nd.ni_vp, sb, l); 2397 vput(nd.ni_vp); 2398 return error; 2399 } 2400 2401 /* 2402 * Get file status; this version follows links. 2403 */ 2404 /* ARGSUSED */ 2405 int 2406 sys___stat30(struct lwp *l, void *v, register_t *retval) 2407 { 2408 struct sys___stat30_args /* { 2409 syscallarg(const char *) path; 2410 syscallarg(struct stat *) ub; 2411 } */ *uap = v; 2412 struct stat sb; 2413 int error; 2414 2415 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2416 if (error) 2417 return error; 2418 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2419 } 2420 2421 /* 2422 * Get file status; this version does not follow links. 2423 */ 2424 /* ARGSUSED */ 2425 int 2426 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2427 { 2428 struct sys___lstat30_args /* { 2429 syscallarg(const char *) path; 2430 syscallarg(struct stat *) ub; 2431 } */ *uap = v; 2432 struct stat sb; 2433 int error; 2434 2435 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2436 if (error) 2437 return error; 2438 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2439 } 2440 2441 /* 2442 * Get configurable pathname variables. 2443 */ 2444 /* ARGSUSED */ 2445 int 2446 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2447 { 2448 struct sys_pathconf_args /* { 2449 syscallarg(const char *) path; 2450 syscallarg(int) name; 2451 } */ *uap = v; 2452 int error; 2453 struct nameidata nd; 2454 2455 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2456 SCARG(uap, path), l); 2457 if ((error = namei(&nd)) != 0) 2458 return (error); 2459 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2460 vput(nd.ni_vp); 2461 return (error); 2462 } 2463 2464 /* 2465 * Return target name of a symbolic link. 2466 */ 2467 /* ARGSUSED */ 2468 int 2469 sys_readlink(struct lwp *l, void *v, register_t *retval) 2470 { 2471 struct sys_readlink_args /* { 2472 syscallarg(const char *) path; 2473 syscallarg(char *) buf; 2474 syscallarg(size_t) count; 2475 } */ *uap = v; 2476 struct vnode *vp; 2477 struct iovec aiov; 2478 struct uio auio; 2479 int error; 2480 struct nameidata nd; 2481 2482 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2483 SCARG(uap, path), l); 2484 if ((error = namei(&nd)) != 0) 2485 return (error); 2486 vp = nd.ni_vp; 2487 if (vp->v_type != VLNK) 2488 error = EINVAL; 2489 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2490 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2491 aiov.iov_base = SCARG(uap, buf); 2492 aiov.iov_len = SCARG(uap, count); 2493 auio.uio_iov = &aiov; 2494 auio.uio_iovcnt = 1; 2495 auio.uio_offset = 0; 2496 auio.uio_rw = UIO_READ; 2497 KASSERT(l == curlwp); 2498 auio.uio_vmspace = l->l_proc->p_vmspace; 2499 auio.uio_resid = SCARG(uap, count); 2500 error = VOP_READLINK(vp, &auio, l->l_cred); 2501 } 2502 vput(vp); 2503 *retval = SCARG(uap, count) - auio.uio_resid; 2504 return (error); 2505 } 2506 2507 /* 2508 * Change flags of a file given a path name. 2509 */ 2510 /* ARGSUSED */ 2511 int 2512 sys_chflags(struct lwp *l, void *v, register_t *retval) 2513 { 2514 struct sys_chflags_args /* { 2515 syscallarg(const char *) path; 2516 syscallarg(u_long) flags; 2517 } */ *uap = v; 2518 struct vnode *vp; 2519 int error; 2520 struct nameidata nd; 2521 2522 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2523 if ((error = namei(&nd)) != 0) 2524 return (error); 2525 vp = nd.ni_vp; 2526 error = change_flags(vp, SCARG(uap, flags), l); 2527 vput(vp); 2528 return (error); 2529 } 2530 2531 /* 2532 * Change flags of a file given a file descriptor. 2533 */ 2534 /* ARGSUSED */ 2535 int 2536 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2537 { 2538 struct sys_fchflags_args /* { 2539 syscallarg(int) fd; 2540 syscallarg(u_long) flags; 2541 } */ *uap = v; 2542 struct proc *p = l->l_proc; 2543 struct vnode *vp; 2544 struct file *fp; 2545 int error; 2546 2547 /* getvnode() will use the descriptor for us */ 2548 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2549 return (error); 2550 vp = (struct vnode *)fp->f_data; 2551 error = change_flags(vp, SCARG(uap, flags), l); 2552 VOP_UNLOCK(vp, 0); 2553 FILE_UNUSE(fp, l); 2554 return (error); 2555 } 2556 2557 /* 2558 * Change flags of a file given a path name; this version does 2559 * not follow links. 2560 */ 2561 int 2562 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2563 { 2564 struct sys_lchflags_args /* { 2565 syscallarg(const char *) path; 2566 syscallarg(u_long) flags; 2567 } */ *uap = v; 2568 struct vnode *vp; 2569 int error; 2570 struct nameidata nd; 2571 2572 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2573 if ((error = namei(&nd)) != 0) 2574 return (error); 2575 vp = nd.ni_vp; 2576 error = change_flags(vp, SCARG(uap, flags), l); 2577 vput(vp); 2578 return (error); 2579 } 2580 2581 /* 2582 * Common routine to change flags of a file. 2583 */ 2584 int 2585 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2586 { 2587 struct vattr vattr; 2588 int error; 2589 2590 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2591 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2592 /* 2593 * Non-superusers cannot change the flags on devices, even if they 2594 * own them. 2595 */ 2596 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2597 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2598 goto out; 2599 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2600 error = EINVAL; 2601 goto out; 2602 } 2603 } 2604 VATTR_NULL(&vattr); 2605 vattr.va_flags = flags; 2606 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2607 out: 2608 return (error); 2609 } 2610 2611 /* 2612 * Change mode of a file given path name; this version follows links. 2613 */ 2614 /* ARGSUSED */ 2615 int 2616 sys_chmod(struct lwp *l, void *v, register_t *retval) 2617 { 2618 struct sys_chmod_args /* { 2619 syscallarg(const char *) path; 2620 syscallarg(int) mode; 2621 } */ *uap = v; 2622 int error; 2623 struct nameidata nd; 2624 2625 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2626 if ((error = namei(&nd)) != 0) 2627 return (error); 2628 2629 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2630 2631 vrele(nd.ni_vp); 2632 return (error); 2633 } 2634 2635 /* 2636 * Change mode of a file given a file descriptor. 2637 */ 2638 /* ARGSUSED */ 2639 int 2640 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2641 { 2642 struct sys_fchmod_args /* { 2643 syscallarg(int) fd; 2644 syscallarg(int) mode; 2645 } */ *uap = v; 2646 struct proc *p = l->l_proc; 2647 struct file *fp; 2648 int error; 2649 2650 /* getvnode() will use the descriptor for us */ 2651 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2652 return (error); 2653 2654 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2655 FILE_UNUSE(fp, l); 2656 return (error); 2657 } 2658 2659 /* 2660 * Change mode of a file given path name; this version does not follow links. 2661 */ 2662 /* ARGSUSED */ 2663 int 2664 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2665 { 2666 struct sys_lchmod_args /* { 2667 syscallarg(const char *) path; 2668 syscallarg(int) mode; 2669 } */ *uap = v; 2670 int error; 2671 struct nameidata nd; 2672 2673 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2674 if ((error = namei(&nd)) != 0) 2675 return (error); 2676 2677 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2678 2679 vrele(nd.ni_vp); 2680 return (error); 2681 } 2682 2683 /* 2684 * Common routine to set mode given a vnode. 2685 */ 2686 static int 2687 change_mode(struct vnode *vp, int mode, struct lwp *l) 2688 { 2689 struct vattr vattr; 2690 int error; 2691 2692 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2693 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2694 VATTR_NULL(&vattr); 2695 vattr.va_mode = mode & ALLPERMS; 2696 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2697 VOP_UNLOCK(vp, 0); 2698 return (error); 2699 } 2700 2701 /* 2702 * Set ownership given a path name; this version follows links. 2703 */ 2704 /* ARGSUSED */ 2705 int 2706 sys_chown(struct lwp *l, void *v, register_t *retval) 2707 { 2708 struct sys_chown_args /* { 2709 syscallarg(const char *) path; 2710 syscallarg(uid_t) uid; 2711 syscallarg(gid_t) gid; 2712 } */ *uap = v; 2713 int error; 2714 struct nameidata nd; 2715 2716 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2717 if ((error = namei(&nd)) != 0) 2718 return (error); 2719 2720 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2721 2722 vrele(nd.ni_vp); 2723 return (error); 2724 } 2725 2726 /* 2727 * Set ownership given a path name; this version follows links. 2728 * Provides POSIX semantics. 2729 */ 2730 /* ARGSUSED */ 2731 int 2732 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2733 { 2734 struct sys_chown_args /* { 2735 syscallarg(const char *) path; 2736 syscallarg(uid_t) uid; 2737 syscallarg(gid_t) gid; 2738 } */ *uap = v; 2739 int error; 2740 struct nameidata nd; 2741 2742 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2743 if ((error = namei(&nd)) != 0) 2744 return (error); 2745 2746 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2747 2748 vrele(nd.ni_vp); 2749 return (error); 2750 } 2751 2752 /* 2753 * Set ownership given a file descriptor. 2754 */ 2755 /* ARGSUSED */ 2756 int 2757 sys_fchown(struct lwp *l, void *v, register_t *retval) 2758 { 2759 struct sys_fchown_args /* { 2760 syscallarg(int) fd; 2761 syscallarg(uid_t) uid; 2762 syscallarg(gid_t) gid; 2763 } */ *uap = v; 2764 struct proc *p = l->l_proc; 2765 int error; 2766 struct file *fp; 2767 2768 /* getvnode() will use the descriptor for us */ 2769 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2770 return (error); 2771 2772 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2773 SCARG(uap, gid), l, 0); 2774 FILE_UNUSE(fp, l); 2775 return (error); 2776 } 2777 2778 /* 2779 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2780 */ 2781 /* ARGSUSED */ 2782 int 2783 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2784 { 2785 struct sys_fchown_args /* { 2786 syscallarg(int) fd; 2787 syscallarg(uid_t) uid; 2788 syscallarg(gid_t) gid; 2789 } */ *uap = v; 2790 struct proc *p = l->l_proc; 2791 int error; 2792 struct file *fp; 2793 2794 /* getvnode() will use the descriptor for us */ 2795 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2796 return (error); 2797 2798 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2799 SCARG(uap, gid), l, 1); 2800 FILE_UNUSE(fp, l); 2801 return (error); 2802 } 2803 2804 /* 2805 * Set ownership given a path name; this version does not follow links. 2806 */ 2807 /* ARGSUSED */ 2808 int 2809 sys_lchown(struct lwp *l, void *v, register_t *retval) 2810 { 2811 struct sys_lchown_args /* { 2812 syscallarg(const char *) path; 2813 syscallarg(uid_t) uid; 2814 syscallarg(gid_t) gid; 2815 } */ *uap = v; 2816 int error; 2817 struct nameidata nd; 2818 2819 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2820 if ((error = namei(&nd)) != 0) 2821 return (error); 2822 2823 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2824 2825 vrele(nd.ni_vp); 2826 return (error); 2827 } 2828 2829 /* 2830 * Set ownership given a path name; this version does not follow links. 2831 * Provides POSIX/XPG semantics. 2832 */ 2833 /* ARGSUSED */ 2834 int 2835 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2836 { 2837 struct sys_lchown_args /* { 2838 syscallarg(const char *) path; 2839 syscallarg(uid_t) uid; 2840 syscallarg(gid_t) gid; 2841 } */ *uap = v; 2842 int error; 2843 struct nameidata nd; 2844 2845 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2846 if ((error = namei(&nd)) != 0) 2847 return (error); 2848 2849 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2850 2851 vrele(nd.ni_vp); 2852 return (error); 2853 } 2854 2855 /* 2856 * Common routine to set ownership given a vnode. 2857 */ 2858 static int 2859 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2860 int posix_semantics) 2861 { 2862 struct vattr vattr; 2863 mode_t newmode; 2864 int error; 2865 2866 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2867 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2868 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2869 goto out; 2870 2871 #define CHANGED(x) ((int)(x) != -1) 2872 newmode = vattr.va_mode; 2873 if (posix_semantics) { 2874 /* 2875 * POSIX/XPG semantics: if the caller is not the super-user, 2876 * clear set-user-id and set-group-id bits. Both POSIX and 2877 * the XPG consider the behaviour for calls by the super-user 2878 * implementation-defined; we leave the set-user-id and set- 2879 * group-id settings intact in that case. 2880 */ 2881 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2882 NULL) != 0) 2883 newmode &= ~(S_ISUID | S_ISGID); 2884 } else { 2885 /* 2886 * NetBSD semantics: when changing owner and/or group, 2887 * clear the respective bit(s). 2888 */ 2889 if (CHANGED(uid)) 2890 newmode &= ~S_ISUID; 2891 if (CHANGED(gid)) 2892 newmode &= ~S_ISGID; 2893 } 2894 /* Update va_mode iff altered. */ 2895 if (vattr.va_mode == newmode) 2896 newmode = VNOVAL; 2897 2898 VATTR_NULL(&vattr); 2899 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2900 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2901 vattr.va_mode = newmode; 2902 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2903 #undef CHANGED 2904 2905 out: 2906 VOP_UNLOCK(vp, 0); 2907 return (error); 2908 } 2909 2910 /* 2911 * Set the access and modification times given a path name; this 2912 * version follows links. 2913 */ 2914 /* ARGSUSED */ 2915 int 2916 sys_utimes(struct lwp *l, void *v, register_t *retval) 2917 { 2918 struct sys_utimes_args /* { 2919 syscallarg(const char *) path; 2920 syscallarg(const struct timeval *) tptr; 2921 } */ *uap = v; 2922 2923 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2924 SCARG(uap, tptr), UIO_USERSPACE); 2925 } 2926 2927 /* 2928 * Set the access and modification times given a file descriptor. 2929 */ 2930 /* ARGSUSED */ 2931 int 2932 sys_futimes(struct lwp *l, void *v, register_t *retval) 2933 { 2934 struct sys_futimes_args /* { 2935 syscallarg(int) fd; 2936 syscallarg(const struct timeval *) tptr; 2937 } */ *uap = v; 2938 int error; 2939 struct file *fp; 2940 2941 /* getvnode() will use the descriptor for us */ 2942 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2943 return (error); 2944 2945 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2946 SCARG(uap, tptr), UIO_USERSPACE); 2947 2948 FILE_UNUSE(fp, l); 2949 return (error); 2950 } 2951 2952 /* 2953 * Set the access and modification times given a path name; this 2954 * version does not follow links. 2955 */ 2956 int 2957 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2958 { 2959 struct sys_lutimes_args /* { 2960 syscallarg(const char *) path; 2961 syscallarg(const struct timeval *) tptr; 2962 } */ *uap = v; 2963 2964 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2965 SCARG(uap, tptr), UIO_USERSPACE); 2966 } 2967 2968 /* 2969 * Common routine to set access and modification times given a vnode. 2970 */ 2971 int 2972 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2973 const struct timeval *tptr, enum uio_seg seg) 2974 { 2975 struct vattr vattr; 2976 struct nameidata nd; 2977 int error; 2978 2979 VATTR_NULL(&vattr); 2980 if (tptr == NULL) { 2981 nanotime(&vattr.va_atime); 2982 vattr.va_mtime = vattr.va_atime; 2983 vattr.va_vaflags |= VA_UTIMES_NULL; 2984 } else { 2985 struct timeval tv[2]; 2986 2987 if (seg != UIO_SYSSPACE) { 2988 error = copyin(tptr, &tv, sizeof (tv)); 2989 if (error != 0) 2990 return error; 2991 tptr = tv; 2992 } 2993 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 2994 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 2995 } 2996 2997 if (vp == NULL) { 2998 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 2999 if ((error = namei(&nd)) != 0) 3000 return (error); 3001 vp = nd.ni_vp; 3002 } else 3003 nd.ni_vp = NULL; 3004 3005 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3006 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3007 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3008 VOP_UNLOCK(vp, 0); 3009 3010 if (nd.ni_vp != NULL) 3011 vrele(nd.ni_vp); 3012 3013 return (error); 3014 } 3015 3016 /* 3017 * Truncate a file given its path name. 3018 */ 3019 /* ARGSUSED */ 3020 int 3021 sys_truncate(struct lwp *l, void *v, register_t *retval) 3022 { 3023 struct sys_truncate_args /* { 3024 syscallarg(const char *) path; 3025 syscallarg(int) pad; 3026 syscallarg(off_t) length; 3027 } */ *uap = v; 3028 struct vnode *vp; 3029 struct vattr vattr; 3030 int error; 3031 struct nameidata nd; 3032 3033 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3034 if ((error = namei(&nd)) != 0) 3035 return (error); 3036 vp = nd.ni_vp; 3037 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3038 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3039 if (vp->v_type == VDIR) 3040 error = EISDIR; 3041 else if ((error = vn_writechk(vp)) == 0 && 3042 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3043 VATTR_NULL(&vattr); 3044 vattr.va_size = SCARG(uap, length); 3045 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3046 } 3047 vput(vp); 3048 return (error); 3049 } 3050 3051 /* 3052 * Truncate a file given a file descriptor. 3053 */ 3054 /* ARGSUSED */ 3055 int 3056 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3057 { 3058 struct sys_ftruncate_args /* { 3059 syscallarg(int) fd; 3060 syscallarg(int) pad; 3061 syscallarg(off_t) length; 3062 } */ *uap = v; 3063 struct proc *p = l->l_proc; 3064 struct vattr vattr; 3065 struct vnode *vp; 3066 struct file *fp; 3067 int error; 3068 3069 /* getvnode() will use the descriptor for us */ 3070 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3071 return (error); 3072 if ((fp->f_flag & FWRITE) == 0) { 3073 error = EINVAL; 3074 goto out; 3075 } 3076 vp = (struct vnode *)fp->f_data; 3077 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3078 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3079 if (vp->v_type == VDIR) 3080 error = EISDIR; 3081 else if ((error = vn_writechk(vp)) == 0) { 3082 VATTR_NULL(&vattr); 3083 vattr.va_size = SCARG(uap, length); 3084 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3085 } 3086 VOP_UNLOCK(vp, 0); 3087 out: 3088 FILE_UNUSE(fp, l); 3089 return (error); 3090 } 3091 3092 /* 3093 * Sync an open file. 3094 */ 3095 /* ARGSUSED */ 3096 int 3097 sys_fsync(struct lwp *l, void *v, register_t *retval) 3098 { 3099 struct sys_fsync_args /* { 3100 syscallarg(int) fd; 3101 } */ *uap = v; 3102 struct proc *p = l->l_proc; 3103 struct vnode *vp; 3104 struct file *fp; 3105 int error; 3106 3107 /* getvnode() will use the descriptor for us */ 3108 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3109 return (error); 3110 vp = (struct vnode *)fp->f_data; 3111 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3112 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3113 if (error == 0 && bioops != NULL && 3114 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3115 (*bioops->io_fsync)(vp, 0); 3116 VOP_UNLOCK(vp, 0); 3117 FILE_UNUSE(fp, l); 3118 return (error); 3119 } 3120 3121 /* 3122 * Sync a range of file data. API modeled after that found in AIX. 3123 * 3124 * FDATASYNC indicates that we need only save enough metadata to be able 3125 * to re-read the written data. Note we duplicate AIX's requirement that 3126 * the file be open for writing. 3127 */ 3128 /* ARGSUSED */ 3129 int 3130 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3131 { 3132 struct sys_fsync_range_args /* { 3133 syscallarg(int) fd; 3134 syscallarg(int) flags; 3135 syscallarg(off_t) start; 3136 syscallarg(off_t) length; 3137 } */ *uap = v; 3138 struct proc *p = l->l_proc; 3139 struct vnode *vp; 3140 struct file *fp; 3141 int flags, nflags; 3142 off_t s, e, len; 3143 int error; 3144 3145 /* getvnode() will use the descriptor for us */ 3146 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3147 return (error); 3148 3149 if ((fp->f_flag & FWRITE) == 0) { 3150 error = EBADF; 3151 goto out; 3152 } 3153 3154 flags = SCARG(uap, flags); 3155 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3156 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3157 error = EINVAL; 3158 goto out; 3159 } 3160 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3161 if (flags & FDATASYNC) 3162 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3163 else 3164 nflags = FSYNC_WAIT; 3165 if (flags & FDISKSYNC) 3166 nflags |= FSYNC_CACHE; 3167 3168 len = SCARG(uap, length); 3169 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3170 if (len) { 3171 s = SCARG(uap, start); 3172 e = s + len; 3173 if (e < s) { 3174 error = EINVAL; 3175 goto out; 3176 } 3177 } else { 3178 e = 0; 3179 s = 0; 3180 } 3181 3182 vp = (struct vnode *)fp->f_data; 3183 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3184 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3185 3186 if (error == 0 && bioops != NULL && 3187 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3188 (*bioops->io_fsync)(vp, nflags); 3189 3190 VOP_UNLOCK(vp, 0); 3191 out: 3192 FILE_UNUSE(fp, l); 3193 return (error); 3194 } 3195 3196 /* 3197 * Sync the data of an open file. 3198 */ 3199 /* ARGSUSED */ 3200 int 3201 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3202 { 3203 struct sys_fdatasync_args /* { 3204 syscallarg(int) fd; 3205 } */ *uap = v; 3206 struct proc *p = l->l_proc; 3207 struct vnode *vp; 3208 struct file *fp; 3209 int error; 3210 3211 /* getvnode() will use the descriptor for us */ 3212 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3213 return (error); 3214 if ((fp->f_flag & FWRITE) == 0) { 3215 FILE_UNUSE(fp, l); 3216 return (EBADF); 3217 } 3218 vp = (struct vnode *)fp->f_data; 3219 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3220 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3221 VOP_UNLOCK(vp, 0); 3222 FILE_UNUSE(fp, l); 3223 return (error); 3224 } 3225 3226 /* 3227 * Rename files, (standard) BSD semantics frontend. 3228 */ 3229 /* ARGSUSED */ 3230 int 3231 sys_rename(struct lwp *l, void *v, register_t *retval) 3232 { 3233 struct sys_rename_args /* { 3234 syscallarg(const char *) from; 3235 syscallarg(const char *) to; 3236 } */ *uap = v; 3237 3238 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3239 } 3240 3241 /* 3242 * Rename files, POSIX semantics frontend. 3243 */ 3244 /* ARGSUSED */ 3245 int 3246 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3247 { 3248 struct sys___posix_rename_args /* { 3249 syscallarg(const char *) from; 3250 syscallarg(const char *) to; 3251 } */ *uap = v; 3252 3253 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3254 } 3255 3256 /* 3257 * Rename files. Source and destination must either both be directories, 3258 * or both not be directories. If target is a directory, it must be empty. 3259 * If `from' and `to' refer to the same object, the value of the `retain' 3260 * argument is used to determine whether `from' will be 3261 * 3262 * (retain == 0) deleted unless `from' and `to' refer to the same 3263 * object in the file system's name space (BSD). 3264 * (retain == 1) always retained (POSIX). 3265 */ 3266 static int 3267 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3268 { 3269 struct vnode *tvp, *fvp, *tdvp; 3270 struct nameidata fromnd, tond; 3271 struct proc *p; 3272 int error; 3273 3274 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3275 from, l); 3276 if ((error = namei(&fromnd)) != 0) 3277 return (error); 3278 if (fromnd.ni_dvp != fromnd.ni_vp) 3279 VOP_UNLOCK(fromnd.ni_dvp, 0); 3280 fvp = fromnd.ni_vp; 3281 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3282 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3283 if ((error = namei(&tond)) != 0) { 3284 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3285 vrele(fromnd.ni_dvp); 3286 vrele(fvp); 3287 goto out1; 3288 } 3289 tdvp = tond.ni_dvp; 3290 tvp = tond.ni_vp; 3291 3292 if (tvp != NULL) { 3293 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3294 error = ENOTDIR; 3295 goto out; 3296 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3297 error = EISDIR; 3298 goto out; 3299 } 3300 } 3301 3302 if (fvp == tdvp) 3303 error = EINVAL; 3304 3305 /* 3306 * Source and destination refer to the same object. 3307 */ 3308 if (fvp == tvp) { 3309 if (retain) 3310 error = -1; 3311 else if (fromnd.ni_dvp == tdvp && 3312 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3313 !memcmp(fromnd.ni_cnd.cn_nameptr, 3314 tond.ni_cnd.cn_nameptr, 3315 fromnd.ni_cnd.cn_namelen)) 3316 error = -1; 3317 } 3318 3319 #if NVERIEXEC > 0 3320 if (!error) { 3321 char *f1, *f2; 3322 3323 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3324 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3325 3326 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3327 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3328 3329 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3330 3331 free(f1, M_TEMP); 3332 free(f2, M_TEMP); 3333 } 3334 #endif /* NVERIEXEC > 0 */ 3335 3336 out: 3337 p = l->l_proc; 3338 if (!error) { 3339 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3340 if (fromnd.ni_dvp != tdvp) 3341 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3342 if (tvp) { 3343 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3344 } 3345 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3346 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3347 } else { 3348 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3349 if (tdvp == tvp) 3350 vrele(tdvp); 3351 else 3352 vput(tdvp); 3353 if (tvp) 3354 vput(tvp); 3355 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3356 vrele(fromnd.ni_dvp); 3357 vrele(fvp); 3358 } 3359 vrele(tond.ni_startdir); 3360 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3361 out1: 3362 if (fromnd.ni_startdir) 3363 vrele(fromnd.ni_startdir); 3364 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3365 return (error == -1 ? 0 : error); 3366 } 3367 3368 /* 3369 * Make a directory file. 3370 */ 3371 /* ARGSUSED */ 3372 int 3373 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3374 { 3375 struct sys_mkdir_args /* { 3376 syscallarg(const char *) path; 3377 syscallarg(int) mode; 3378 } */ *uap = v; 3379 struct proc *p = l->l_proc; 3380 struct vnode *vp; 3381 struct vattr vattr; 3382 int error; 3383 struct nameidata nd; 3384 3385 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3386 SCARG(uap, path), l); 3387 if ((error = namei(&nd)) != 0) 3388 return (error); 3389 vp = nd.ni_vp; 3390 if (vp != NULL) { 3391 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3392 if (nd.ni_dvp == vp) 3393 vrele(nd.ni_dvp); 3394 else 3395 vput(nd.ni_dvp); 3396 vrele(vp); 3397 return (EEXIST); 3398 } 3399 VATTR_NULL(&vattr); 3400 vattr.va_type = VDIR; 3401 /* We will read cwdi->cwdi_cmask unlocked. */ 3402 vattr.va_mode = 3403 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3404 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3405 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3406 if (!error) 3407 vput(nd.ni_vp); 3408 return (error); 3409 } 3410 3411 /* 3412 * Remove a directory file. 3413 */ 3414 /* ARGSUSED */ 3415 int 3416 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3417 { 3418 struct sys_rmdir_args /* { 3419 syscallarg(const char *) path; 3420 } */ *uap = v; 3421 struct vnode *vp; 3422 int error; 3423 struct nameidata nd; 3424 3425 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3426 SCARG(uap, path), l); 3427 if ((error = namei(&nd)) != 0) 3428 return (error); 3429 vp = nd.ni_vp; 3430 if (vp->v_type != VDIR) { 3431 error = ENOTDIR; 3432 goto out; 3433 } 3434 /* 3435 * No rmdir "." please. 3436 */ 3437 if (nd.ni_dvp == vp) { 3438 error = EINVAL; 3439 goto out; 3440 } 3441 /* 3442 * The root of a mounted filesystem cannot be deleted. 3443 */ 3444 if (vp->v_vflag & VV_ROOT) { 3445 error = EBUSY; 3446 goto out; 3447 } 3448 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3449 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3450 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3451 return (error); 3452 3453 out: 3454 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3455 if (nd.ni_dvp == vp) 3456 vrele(nd.ni_dvp); 3457 else 3458 vput(nd.ni_dvp); 3459 vput(vp); 3460 return (error); 3461 } 3462 3463 /* 3464 * Read a block of directory entries in a file system independent format. 3465 */ 3466 int 3467 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3468 { 3469 struct sys___getdents30_args /* { 3470 syscallarg(int) fd; 3471 syscallarg(char *) buf; 3472 syscallarg(size_t) count; 3473 } */ *uap = v; 3474 struct proc *p = l->l_proc; 3475 struct file *fp; 3476 int error, done; 3477 3478 /* getvnode() will use the descriptor for us */ 3479 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3480 return (error); 3481 if ((fp->f_flag & FREAD) == 0) { 3482 error = EBADF; 3483 goto out; 3484 } 3485 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3486 SCARG(uap, count), &done, l, 0, 0); 3487 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3488 *retval = done; 3489 out: 3490 FILE_UNUSE(fp, l); 3491 return (error); 3492 } 3493 3494 /* 3495 * Set the mode mask for creation of filesystem nodes. 3496 */ 3497 int 3498 sys_umask(struct lwp *l, void *v, register_t *retval) 3499 { 3500 struct sys_umask_args /* { 3501 syscallarg(mode_t) newmask; 3502 } */ *uap = v; 3503 struct proc *p = l->l_proc; 3504 struct cwdinfo *cwdi; 3505 3506 /* 3507 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3508 * important is that we serialize changes to the mask. The 3509 * rw_exit() will issue a write memory barrier on our behalf, 3510 * and force the changes out to other CPUs (as it must use an 3511 * atomic operation, draining the local CPU's store buffers). 3512 */ 3513 cwdi = p->p_cwdi; 3514 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3515 *retval = cwdi->cwdi_cmask; 3516 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3517 rw_exit(&cwdi->cwdi_lock); 3518 3519 return (0); 3520 } 3521 3522 /* 3523 * Void all references to file by ripping underlying filesystem 3524 * away from vnode. 3525 */ 3526 /* ARGSUSED */ 3527 int 3528 sys_revoke(struct lwp *l, void *v, register_t *retval) 3529 { 3530 struct sys_revoke_args /* { 3531 syscallarg(const char *) path; 3532 } */ *uap = v; 3533 struct vnode *vp; 3534 struct vattr vattr; 3535 int error; 3536 bool revoke; 3537 struct nameidata nd; 3538 3539 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3540 if ((error = namei(&nd)) != 0) 3541 return (error); 3542 vp = nd.ni_vp; 3543 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3544 goto out; 3545 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3546 (error = kauth_authorize_generic(l->l_cred, 3547 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3548 goto out; 3549 mutex_enter(&vp->v_interlock); 3550 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED | VI_LAYER))); 3551 mutex_exit(&vp->v_interlock); 3552 if (revoke) 3553 VOP_REVOKE(vp, REVOKEALL); 3554 out: 3555 vrele(vp); 3556 return (error); 3557 } 3558 3559 /* 3560 * Convert a user file descriptor to a kernel file entry. 3561 */ 3562 int 3563 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3564 { 3565 struct vnode *vp; 3566 struct file *fp; 3567 3568 if ((fp = fd_getfile(fdp, fd)) == NULL) 3569 return (EBADF); 3570 3571 FILE_USE(fp); 3572 3573 if (fp->f_type != DTYPE_VNODE) { 3574 FILE_UNUSE(fp, NULL); 3575 return (EINVAL); 3576 } 3577 3578 vp = (struct vnode *)fp->f_data; 3579 if (vp->v_type == VBAD) { 3580 FILE_UNUSE(fp, NULL); 3581 return (EBADF); 3582 } 3583 3584 *fpp = fp; 3585 return (0); 3586 } 3587
Indexes created Fri Oct 03 17:10:05 GMT 2025