vfs_syscalls.c revision 1.331.2.2
1 /* $NetBSD: vfs_syscalls.c,v 1.331.2.2 2007/12/27 00:46:19 mjf 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 1.331.2.2 2007/12/27 00:46:19 mjf Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "fss.h" 46 #include "veriexec.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/kmem.h> 61 #include <sys/dirent.h> 62 #include <sys/sysctl.h> 63 #include <sys/syscallargs.h> 64 #include <sys/vfs_syscalls.h> 65 #include <sys/ktrace.h> 66 #ifdef FILEASSOC 67 #include <sys/fileassoc.h> 68 #endif /* FILEASSOC */ 69 #include <sys/verified_exec.h> 70 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #endif 79 #include <nfs/nfsproto.h> 80 #ifdef COMPAT_30 81 #include <nfs/nfs.h> 82 #include <nfs/nfs_var.h> 83 #endif 84 85 #if NFSS > 0 86 #include <dev/fssvar.h> 87 #endif 88 89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 90 91 static int change_dir(struct nameidata *, struct lwp *); 92 static int change_flags(struct vnode *, u_long, struct lwp *); 93 static int change_mode(struct vnode *, int, struct lwp *l); 94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 95 96 void checkdirs(struct vnode *); 97 98 int dovfsusermount = 0; 99 100 /* 101 * Virtual File System System Calls 102 */ 103 104 /* 105 * Mount a file system. 106 */ 107 108 #if defined(COMPAT_09) || defined(COMPAT_43) 109 /* 110 * This table is used to maintain compatibility with 4.3BSD 111 * and NetBSD 0.9 mount syscalls. Note, the order is important! 112 * 113 * Do not modify this table. It should only contain filesystems 114 * supported by NetBSD 0.9 and 4.3BSD. 115 */ 116 const char * const mountcompatnames[] = { 117 NULL, /* 0 = MOUNT_NONE */ 118 MOUNT_FFS, /* 1 = MOUNT_UFS */ 119 MOUNT_NFS, /* 2 */ 120 MOUNT_MFS, /* 3 */ 121 MOUNT_MSDOS, /* 4 */ 122 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 123 MOUNT_FDESC, /* 6 */ 124 MOUNT_KERNFS, /* 7 */ 125 NULL, /* 8 = MOUNT_DEVFS */ 126 MOUNT_AFS, /* 9 */ 127 }; 128 const int nmountcompatnames = sizeof(mountcompatnames) / 129 sizeof(mountcompatnames[0]); 130 #endif /* COMPAT_09 || COMPAT_43 */ 131 132 static int 133 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 134 void *data, size_t *data_len) 135 { 136 struct mount *mp; 137 int error = 0, saved_flags; 138 139 mp = vp->v_mount; 140 saved_flags = mp->mnt_flag; 141 142 /* We can operate only on VV_ROOT nodes. */ 143 if ((vp->v_vflag & VV_ROOT) == 0) { 144 error = EINVAL; 145 goto out; 146 } 147 148 /* 149 * We only allow the filesystem to be reloaded if it 150 * is currently mounted read-only. 151 */ 152 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 153 error = EOPNOTSUPP; /* Needs translation */ 154 goto out; 155 } 156 157 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 158 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 159 if (error) 160 goto out; 161 162 if (vfs_busy(mp, LK_NOWAIT, 0)) { 163 error = EPERM; 164 goto out; 165 } 166 167 mp->mnt_flag &= ~MNT_OP_FLAGS; 168 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 169 170 /* 171 * Set the mount level flags. 172 */ 173 if (flags & MNT_RDONLY) 174 mp->mnt_flag |= MNT_RDONLY; 175 else if (mp->mnt_flag & MNT_RDONLY) 176 mp->mnt_iflag |= IMNT_WANTRDWR; 177 mp->mnt_flag &= 178 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 179 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 180 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 181 mp->mnt_flag |= flags & 182 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 183 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 184 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 185 MNT_IGNORE); 186 187 error = VFS_MOUNT(mp, path, data, data_len); 188 189 #if defined(COMPAT_30) && defined(NFSSERVER) 190 if (error && data != NULL) { 191 int error2; 192 193 /* Update failed; let's try and see if it was an 194 * export request. */ 195 error2 = nfs_update_exports_30(mp, path, data, l); 196 197 /* Only update error code if the export request was 198 * understood but some problem occurred while 199 * processing it. */ 200 if (error2 != EJUSTRETURN) 201 error = error2; 202 } 203 #endif 204 if (mp->mnt_iflag & IMNT_WANTRDWR) 205 mp->mnt_flag &= ~MNT_RDONLY; 206 if (error) 207 mp->mnt_flag = saved_flags; 208 mp->mnt_flag &= ~MNT_OP_FLAGS; 209 mp->mnt_iflag &= ~IMNT_WANTRDWR; 210 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 211 if (mp->mnt_syncer == NULL) 212 error = vfs_allocate_syncvnode(mp); 213 } else { 214 if (mp->mnt_syncer != NULL) 215 vfs_deallocate_syncvnode(mp); 216 } 217 vfs_unbusy(mp); 218 219 out: 220 return (error); 221 } 222 223 static int 224 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 225 { 226 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 227 int error; 228 229 /* Copy file-system type from userspace. */ 230 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 231 if (error) { 232 #if defined(COMPAT_09) || defined(COMPAT_43) 233 /* 234 * Historically, filesystem types were identified by numbers. 235 * If we get an integer for the filesystem type instead of a 236 * string, we check to see if it matches one of the historic 237 * filesystem types. 238 */ 239 u_long fsindex = (u_long)fstype; 240 if (fsindex >= nmountcompatnames || 241 mountcompatnames[fsindex] == NULL) 242 return ENODEV; 243 strlcpy(fstypename, mountcompatnames[fsindex], 244 sizeof(fstypename)); 245 #else 246 return error; 247 #endif 248 } 249 250 #ifdef COMPAT_10 251 /* Accept `ufs' as an alias for `ffs'. */ 252 if (strcmp(fstypename, "ufs") == 0) 253 fstypename[0] = 'f'; 254 #endif 255 256 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 257 return ENODEV; 258 return 0; 259 } 260 261 static int 262 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 263 const char *path, int flags, void *data, size_t *data_len) 264 { 265 struct mount *mp = NULL; 266 struct vnode *vp = *vpp; 267 struct vattr va; 268 int error; 269 270 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 271 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 272 if (error) 273 return error; 274 275 /* Can't make a non-dir a mount-point (from here anyway). */ 276 if (vp->v_type != VDIR) 277 return ENOTDIR; 278 279 /* 280 * If the user is not root, ensure that they own the directory 281 * onto which we are attempting to mount. 282 */ 283 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 284 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 285 (error = kauth_authorize_generic(l->l_cred, 286 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 287 return error; 288 } 289 290 if (flags & MNT_EXPORTED) 291 return EINVAL; 292 293 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 294 return error; 295 296 /* 297 * Check if a file-system is not already mounted on this vnode. 298 */ 299 if (vp->v_mountedhere != NULL) 300 return EBUSY; 301 302 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 303 304 mp->mnt_op = vfsops; 305 306 TAILQ_INIT(&mp->mnt_vnodelist); 307 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 308 simple_lock_init(&mp->mnt_slock); 309 (void)vfs_busy(mp, LK_NOWAIT, 0); 310 311 mp->mnt_vnodecovered = vp; 312 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 313 mp->mnt_unmounter = NULL; 314 mount_initspecific(mp); 315 316 /* 317 * The underlying file system may refuse the mount for 318 * various reasons. Allow the user to force it to happen. 319 * 320 * Set the mount level flags. 321 */ 322 mp->mnt_flag = flags & 323 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 324 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 325 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 326 MNT_IGNORE | MNT_RDONLY); 327 328 error = VFS_MOUNT(mp, path, data, data_len); 329 mp->mnt_flag &= ~MNT_OP_FLAGS; 330 331 /* 332 * Put the new filesystem on the mount list after root. 333 */ 334 cache_purge(vp); 335 if (error != 0) { 336 vp->v_mountedhere = NULL; 337 mp->mnt_op->vfs_refcount--; 338 vfs_unbusy(mp); 339 vfs_destroy(mp); 340 return error; 341 } 342 343 mp->mnt_iflag &= ~IMNT_WANTRDWR; 344 vp->v_mountedhere = mp; 345 mutex_enter(&mountlist_lock); 346 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 347 mutex_exit(&mountlist_lock); 348 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); 354 error = VFS_START(mp, 0); 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); 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, const struct compat_40_sys_mount_args *uap, register_t *retval) 399 { 400 /* { 401 syscallarg(const char *) type; 402 syscallarg(const char *) path; 403 syscallarg(int) flags; 404 syscallarg(void *) data; 405 } */ 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, const struct sys___mount50_args *uap, register_t *retval) 415 { 416 /* { 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 } */ 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); 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, const struct sys_unmount_args *uap, register_t *retval) 567 { 568 /* { 569 syscallarg(const char *) path; 570 syscallarg(int) flags; 571 } */ 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)); 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 mutex_exit(&mountlist_lock); /* XXX */ 664 lockmgr(&mp->mnt_lock, LK_DRAIN, NULL); 665 666 async = mp->mnt_flag & MNT_ASYNC; 667 mp->mnt_flag &= ~MNT_ASYNC; 668 cache_purgevfs(mp); /* remove cache entries for this file sys */ 669 if (mp->mnt_syncer != NULL) 670 vfs_deallocate_syncvnode(mp); 671 error = 0; 672 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 673 #if NFSS > 0 674 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 675 #endif 676 if (error == 0) 677 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 678 } 679 if (error == 0 || (flags & MNT_FORCE)) 680 error = VFS_UNMOUNT(mp, flags); 681 if (error) { 682 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 683 (void) vfs_allocate_syncvnode(mp); 684 mutex_enter(&mountlist_lock); 685 mp->mnt_iflag &= ~IMNT_UNMOUNT; 686 mp->mnt_unmounter = NULL; 687 mp->mnt_flag |= async; 688 mutex_exit(&mountlist_lock); /* XXX */ 689 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_REENABLE, 690 NULL); 691 if (used_syncer) 692 mutex_exit(&syncer_mutex); 693 simple_lock(&mp->mnt_slock); 694 while (mp->mnt_wcnt > 0) { 695 wakeup(mp); 696 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 697 0, &mp->mnt_slock); 698 } 699 simple_unlock(&mp->mnt_slock); 700 return (error); 701 } 702 mutex_enter(&mountlist_lock); 703 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 704 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 705 coveredvp->v_mountedhere = NULL; 706 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 707 panic("unmount: dangling vnode"); 708 mp->mnt_iflag |= IMNT_GONE; 709 mutex_exit(&mountlist_lock); 710 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL); 711 if (coveredvp != NULLVP) 712 vrele(coveredvp); 713 if (used_syncer) 714 mutex_exit(&syncer_mutex); 715 simple_lock(&mp->mnt_slock); 716 while (mp->mnt_wcnt > 0) { 717 wakeup(mp); 718 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 719 } 720 simple_unlock(&mp->mnt_slock); 721 vfs_hooks_unmount(mp); 722 vfs_delref(mp->mnt_op); 723 vfs_destroy(mp); 724 return (0); 725 } 726 727 /* 728 * Sync each mounted filesystem. 729 */ 730 #ifdef DEBUG 731 int syncprt = 0; 732 struct ctldebug debug0 = { "syncprt", &syncprt }; 733 #endif 734 735 /* ARGSUSED */ 736 int 737 sys_sync(struct lwp *l, const void *v, register_t *retval) 738 { 739 struct mount *mp, *nmp; 740 int asyncflag; 741 742 if (l == NULL) 743 l = &lwp0; 744 745 mutex_enter(&mountlist_lock); 746 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 747 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 748 nmp = mp->mnt_list.cqe_prev; 749 continue; 750 } 751 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 752 asyncflag = mp->mnt_flag & MNT_ASYNC; 753 mp->mnt_flag &= ~MNT_ASYNC; 754 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 755 if (asyncflag) 756 mp->mnt_flag |= MNT_ASYNC; 757 } 758 mutex_enter(&mountlist_lock); 759 nmp = mp->mnt_list.cqe_prev; 760 vfs_unbusy(mp); 761 762 } 763 mutex_exit(&mountlist_lock); 764 #ifdef DEBUG 765 if (syncprt) 766 vfs_bufstats(); 767 #endif /* DEBUG */ 768 return (0); 769 } 770 771 /* 772 * Change filesystem quotas. 773 */ 774 /* ARGSUSED */ 775 int 776 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 777 { 778 /* { 779 syscallarg(const char *) path; 780 syscallarg(int) cmd; 781 syscallarg(int) uid; 782 syscallarg(void *) arg; 783 } */ 784 struct mount *mp; 785 int error; 786 struct nameidata nd; 787 788 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 789 SCARG(uap, path)); 790 if ((error = namei(&nd)) != 0) 791 return (error); 792 mp = nd.ni_vp->v_mount; 793 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 794 SCARG(uap, arg)); 795 vrele(nd.ni_vp); 796 return (error); 797 } 798 799 int 800 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 801 int root) 802 { 803 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 804 int error = 0; 805 806 /* 807 * If MNT_NOWAIT or MNT_LAZY is specified, do not 808 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 809 * overrides MNT_NOWAIT. 810 */ 811 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 812 (flags != MNT_WAIT && flags != 0)) { 813 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 814 goto done; 815 } 816 817 /* Get the filesystem stats now */ 818 memset(sp, 0, sizeof(*sp)); 819 if ((error = VFS_STATVFS(mp, sp)) != 0) { 820 return error; 821 } 822 823 if (cwdi->cwdi_rdir == NULL) 824 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 825 done: 826 if (cwdi->cwdi_rdir != NULL) { 827 size_t len; 828 char *bp; 829 char *path = PNBUF_GET(); 830 831 bp = path + MAXPATHLEN; 832 *--bp = '\0'; 833 rw_enter(&cwdi->cwdi_lock, RW_READER); 834 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 835 MAXPATHLEN / 2, 0, l); 836 rw_exit(&cwdi->cwdi_lock); 837 if (error) { 838 PNBUF_PUT(path); 839 return error; 840 } 841 len = strlen(bp); 842 /* 843 * for mount points that are below our root, we can see 844 * them, so we fix up the pathname and return them. The 845 * rest we cannot see, so we don't allow viewing the 846 * data. 847 */ 848 if (strncmp(bp, sp->f_mntonname, len) == 0) { 849 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 850 sizeof(sp->f_mntonname)); 851 if (sp->f_mntonname[0] == '\0') 852 (void)strlcpy(sp->f_mntonname, "/", 853 sizeof(sp->f_mntonname)); 854 } else { 855 if (root) 856 (void)strlcpy(sp->f_mntonname, "/", 857 sizeof(sp->f_mntonname)); 858 else 859 error = EPERM; 860 } 861 PNBUF_PUT(path); 862 } 863 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 864 return error; 865 } 866 867 /* 868 * Get filesystem statistics by path. 869 */ 870 int 871 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 872 { 873 struct mount *mp; 874 int error; 875 struct nameidata nd; 876 877 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 878 if ((error = namei(&nd)) != 0) 879 return error; 880 mp = nd.ni_vp->v_mount; 881 error = dostatvfs(mp, sb, l, flags, 1); 882 vrele(nd.ni_vp); 883 return error; 884 } 885 886 /* ARGSUSED */ 887 int 888 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 889 { 890 /* { 891 syscallarg(const char *) path; 892 syscallarg(struct statvfs *) buf; 893 syscallarg(int) flags; 894 } */ 895 struct statvfs *sb; 896 int error; 897 898 sb = STATVFSBUF_GET(); 899 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 900 if (error == 0) 901 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 902 STATVFSBUF_PUT(sb); 903 return error; 904 } 905 906 /* 907 * Get filesystem statistics by fd. 908 */ 909 int 910 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 911 { 912 struct proc *p = l->l_proc; 913 struct file *fp; 914 struct mount *mp; 915 int error; 916 917 /* getvnode() will use the descriptor for us */ 918 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 919 return (error); 920 mp = ((struct vnode *)fp->f_data)->v_mount; 921 error = dostatvfs(mp, sb, l, flags, 1); 922 FILE_UNUSE(fp, l); 923 return error; 924 } 925 926 /* ARGSUSED */ 927 int 928 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 929 { 930 /* { 931 syscallarg(int) fd; 932 syscallarg(struct statvfs *) buf; 933 syscallarg(int) flags; 934 } */ 935 struct statvfs *sb; 936 int error; 937 938 sb = STATVFSBUF_GET(); 939 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 940 if (error == 0) 941 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 942 STATVFSBUF_PUT(sb); 943 return error; 944 } 945 946 947 /* 948 * Get statistics on all filesystems. 949 */ 950 int 951 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 952 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 953 register_t *retval) 954 { 955 int root = 0; 956 struct proc *p = l->l_proc; 957 struct mount *mp, *nmp; 958 struct statvfs *sb; 959 size_t count, maxcount; 960 int error = 0; 961 962 sb = STATVFSBUF_GET(); 963 maxcount = bufsize / entry_sz; 964 mutex_enter(&mountlist_lock); 965 count = 0; 966 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 967 mp = nmp) { 968 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 969 nmp = CIRCLEQ_NEXT(mp, mnt_list); 970 continue; 971 } 972 if (sfsp && count < maxcount) { 973 error = dostatvfs(mp, sb, l, flags, 0); 974 if (error) { 975 mutex_enter(&mountlist_lock); 976 nmp = CIRCLEQ_NEXT(mp, mnt_list); 977 vfs_unbusy(mp); 978 continue; 979 } 980 error = copyfn(sb, sfsp, entry_sz); 981 if (error) { 982 vfs_unbusy(mp); 983 goto out; 984 } 985 sfsp = (char *)sfsp + entry_sz; 986 root |= strcmp(sb->f_mntonname, "/") == 0; 987 } 988 count++; 989 mutex_enter(&mountlist_lock); 990 nmp = CIRCLEQ_NEXT(mp, mnt_list); 991 vfs_unbusy(mp); 992 } 993 994 mutex_exit(&mountlist_lock); 995 if (root == 0 && p->p_cwdi->cwdi_rdir) { 996 /* 997 * fake a root entry 998 */ 999 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1000 sb, l, flags, 1); 1001 if (error != 0) 1002 goto out; 1003 if (sfsp) 1004 error = copyfn(sb, sfsp, entry_sz); 1005 count++; 1006 } 1007 if (sfsp && count > maxcount) 1008 *retval = maxcount; 1009 else 1010 *retval = count; 1011 out: 1012 STATVFSBUF_PUT(sb); 1013 return error; 1014 } 1015 1016 int 1017 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1018 { 1019 /* { 1020 syscallarg(struct statvfs *) buf; 1021 syscallarg(size_t) bufsize; 1022 syscallarg(int) flags; 1023 } */ 1024 1025 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1026 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1027 } 1028 1029 /* 1030 * Change current working directory to a given file descriptor. 1031 */ 1032 /* ARGSUSED */ 1033 int 1034 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1035 { 1036 /* { 1037 syscallarg(int) fd; 1038 } */ 1039 struct proc *p = l->l_proc; 1040 struct filedesc *fdp = p->p_fd; 1041 struct cwdinfo *cwdi; 1042 struct vnode *vp, *tdp; 1043 struct mount *mp; 1044 struct file *fp; 1045 int error; 1046 1047 /* getvnode() will use the descriptor for us */ 1048 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1049 return (error); 1050 vp = (struct vnode *)fp->f_data; 1051 1052 VREF(vp); 1053 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1054 if (vp->v_type != VDIR) 1055 error = ENOTDIR; 1056 else 1057 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1058 if (error) { 1059 vput(vp); 1060 goto out; 1061 } 1062 while ((mp = vp->v_mountedhere) != NULL) { 1063 if (vfs_busy(mp, 0, 0)) 1064 continue; 1065 1066 vput(vp); 1067 error = VFS_ROOT(mp, &tdp); 1068 vfs_unbusy(mp); 1069 if (error) 1070 goto out; 1071 vp = tdp; 1072 } 1073 VOP_UNLOCK(vp, 0); 1074 1075 /* 1076 * Disallow changing to a directory not under the process's 1077 * current root directory (if there is one). 1078 */ 1079 cwdi = p->p_cwdi; 1080 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1081 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1082 vrele(vp); 1083 error = EPERM; /* operation not permitted */ 1084 } else { 1085 vrele(cwdi->cwdi_cdir); 1086 cwdi->cwdi_cdir = vp; 1087 } 1088 rw_exit(&cwdi->cwdi_lock); 1089 1090 out: 1091 FILE_UNUSE(fp, l); 1092 return (error); 1093 } 1094 1095 /* 1096 * Change this process's notion of the root directory to a given file 1097 * descriptor. 1098 */ 1099 int 1100 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1101 { 1102 struct proc *p = l->l_proc; 1103 struct filedesc *fdp = p->p_fd; 1104 struct cwdinfo *cwdi; 1105 struct vnode *vp; 1106 struct file *fp; 1107 int error; 1108 1109 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1110 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1111 return error; 1112 /* getvnode() will use the descriptor for us */ 1113 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1114 return error; 1115 vp = (struct vnode *) fp->f_data; 1116 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1117 if (vp->v_type != VDIR) 1118 error = ENOTDIR; 1119 else 1120 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1121 VOP_UNLOCK(vp, 0); 1122 if (error) 1123 goto out; 1124 VREF(vp); 1125 1126 /* 1127 * Prevent escaping from chroot by putting the root under 1128 * the working directory. Silently chdir to / if we aren't 1129 * already there. 1130 */ 1131 cwdi = p->p_cwdi; 1132 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1133 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1134 /* 1135 * XXX would be more failsafe to change directory to a 1136 * deadfs node here instead 1137 */ 1138 vrele(cwdi->cwdi_cdir); 1139 VREF(vp); 1140 cwdi->cwdi_cdir = vp; 1141 } 1142 1143 if (cwdi->cwdi_rdir != NULL) 1144 vrele(cwdi->cwdi_rdir); 1145 cwdi->cwdi_rdir = vp; 1146 rw_exit(&cwdi->cwdi_lock); 1147 1148 out: 1149 FILE_UNUSE(fp, l); 1150 return (error); 1151 } 1152 1153 /* 1154 * Change current working directory (``.''). 1155 */ 1156 /* ARGSUSED */ 1157 int 1158 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1159 { 1160 /* { 1161 syscallarg(const char *) path; 1162 } */ 1163 struct proc *p = l->l_proc; 1164 struct cwdinfo *cwdi; 1165 int error; 1166 struct nameidata nd; 1167 1168 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1169 SCARG(uap, path)); 1170 if ((error = change_dir(&nd, l)) != 0) 1171 return (error); 1172 cwdi = p->p_cwdi; 1173 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1174 vrele(cwdi->cwdi_cdir); 1175 cwdi->cwdi_cdir = nd.ni_vp; 1176 rw_exit(&cwdi->cwdi_lock); 1177 return (0); 1178 } 1179 1180 /* 1181 * Change notion of root (``/'') directory. 1182 */ 1183 /* ARGSUSED */ 1184 int 1185 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1186 { 1187 /* { 1188 syscallarg(const char *) path; 1189 } */ 1190 struct proc *p = l->l_proc; 1191 struct cwdinfo *cwdi; 1192 struct vnode *vp; 1193 int error; 1194 struct nameidata nd; 1195 1196 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1197 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1198 return (error); 1199 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1200 SCARG(uap, path)); 1201 if ((error = change_dir(&nd, l)) != 0) 1202 return (error); 1203 1204 cwdi = p->p_cwdi; 1205 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1206 if (cwdi->cwdi_rdir != NULL) 1207 vrele(cwdi->cwdi_rdir); 1208 vp = nd.ni_vp; 1209 cwdi->cwdi_rdir = vp; 1210 1211 /* 1212 * Prevent escaping from chroot by putting the root under 1213 * the working directory. Silently chdir to / if we aren't 1214 * already there. 1215 */ 1216 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1217 /* 1218 * XXX would be more failsafe to change directory to a 1219 * deadfs node here instead 1220 */ 1221 vrele(cwdi->cwdi_cdir); 1222 VREF(vp); 1223 cwdi->cwdi_cdir = vp; 1224 } 1225 rw_exit(&cwdi->cwdi_lock); 1226 1227 return (0); 1228 } 1229 1230 /* 1231 * Common routine for chroot and chdir. 1232 */ 1233 static int 1234 change_dir(struct nameidata *ndp, struct lwp *l) 1235 { 1236 struct vnode *vp; 1237 int error; 1238 1239 if ((error = namei(ndp)) != 0) 1240 return (error); 1241 vp = ndp->ni_vp; 1242 if (vp->v_type != VDIR) 1243 error = ENOTDIR; 1244 else 1245 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1246 1247 if (error) 1248 vput(vp); 1249 else 1250 VOP_UNLOCK(vp, 0); 1251 return (error); 1252 } 1253 1254 /* 1255 * Check permissions, allocate an open file structure, 1256 * and call the device open routine if any. 1257 */ 1258 int 1259 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1260 { 1261 /* { 1262 syscallarg(const char *) path; 1263 syscallarg(int) flags; 1264 syscallarg(int) mode; 1265 } */ 1266 struct proc *p = l->l_proc; 1267 struct cwdinfo *cwdi = p->p_cwdi; 1268 struct filedesc *fdp = p->p_fd; 1269 struct file *fp; 1270 struct vnode *vp; 1271 int flags, cmode; 1272 int type, indx, error; 1273 struct flock lf; 1274 struct nameidata nd; 1275 1276 flags = FFLAGS(SCARG(uap, flags)); 1277 if ((flags & (FREAD | FWRITE)) == 0) 1278 return (EINVAL); 1279 /* falloc() will use the file descriptor for us */ 1280 if ((error = falloc(l, &fp, &indx)) != 0) 1281 return (error); 1282 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1283 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1284 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1285 SCARG(uap, path)); 1286 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1287 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1288 rw_enter(&fdp->fd_lock, RW_WRITER); 1289 FILE_UNUSE(fp, l); 1290 fdp->fd_ofiles[indx] = NULL; 1291 rw_exit(&fdp->fd_lock); 1292 ffree(fp); 1293 if ((error == EDUPFD || error == EMOVEFD) && 1294 l->l_dupfd >= 0 && /* XXX from fdopen */ 1295 (error = 1296 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1297 *retval = indx; 1298 return (0); 1299 } 1300 if (error == ERESTART) 1301 error = EINTR; 1302 fdremove(fdp, indx); 1303 return (error); 1304 } 1305 1306 l->l_dupfd = 0; 1307 vp = nd.ni_vp; 1308 fp->f_flag = flags & FMASK; 1309 fp->f_type = DTYPE_VNODE; 1310 fp->f_ops = &vnops; 1311 fp->f_data = vp; 1312 if (flags & (O_EXLOCK | O_SHLOCK)) { 1313 lf.l_whence = SEEK_SET; 1314 lf.l_start = 0; 1315 lf.l_len = 0; 1316 if (flags & O_EXLOCK) 1317 lf.l_type = F_WRLCK; 1318 else 1319 lf.l_type = F_RDLCK; 1320 type = F_FLOCK; 1321 if ((flags & FNONBLOCK) == 0) 1322 type |= F_WAIT; 1323 VOP_UNLOCK(vp, 0); 1324 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1325 if (error) { 1326 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1327 FILE_UNUSE(fp, l); 1328 ffree(fp); 1329 fdremove(fdp, indx); 1330 return (error); 1331 } 1332 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1333 fp->f_flag |= FHASLOCK; 1334 } 1335 VOP_UNLOCK(vp, 0); 1336 *retval = indx; 1337 FILE_SET_MATURE(fp); 1338 FILE_UNUSE(fp, l); 1339 return (0); 1340 } 1341 1342 static void 1343 vfs__fhfree(fhandle_t *fhp) 1344 { 1345 size_t fhsize; 1346 1347 if (fhp == NULL) { 1348 return; 1349 } 1350 fhsize = FHANDLE_SIZE(fhp); 1351 kmem_free(fhp, fhsize); 1352 } 1353 1354 /* 1355 * vfs_composefh: compose a filehandle. 1356 */ 1357 1358 int 1359 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1360 { 1361 struct mount *mp; 1362 struct fid *fidp; 1363 int error; 1364 size_t needfhsize; 1365 size_t fidsize; 1366 1367 mp = vp->v_mount; 1368 fidp = NULL; 1369 if (*fh_size < FHANDLE_SIZE_MIN) { 1370 fidsize = 0; 1371 } else { 1372 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1373 if (fhp != NULL) { 1374 memset(fhp, 0, *fh_size); 1375 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1376 fidp = &fhp->fh_fid; 1377 } 1378 } 1379 error = VFS_VPTOFH(vp, fidp, &fidsize); 1380 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1381 if (error == 0 && *fh_size < needfhsize) { 1382 error = E2BIG; 1383 } 1384 *fh_size = needfhsize; 1385 return error; 1386 } 1387 1388 int 1389 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1390 { 1391 struct mount *mp; 1392 fhandle_t *fhp; 1393 size_t fhsize; 1394 size_t fidsize; 1395 int error; 1396 1397 *fhpp = NULL; 1398 mp = vp->v_mount; 1399 fidsize = 0; 1400 error = VFS_VPTOFH(vp, NULL, &fidsize); 1401 KASSERT(error != 0); 1402 if (error != E2BIG) { 1403 goto out; 1404 } 1405 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1406 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1407 if (fhp == NULL) { 1408 error = ENOMEM; 1409 goto out; 1410 } 1411 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1412 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1413 if (error == 0) { 1414 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1415 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1416 *fhpp = fhp; 1417 } else { 1418 kmem_free(fhp, fhsize); 1419 } 1420 out: 1421 return error; 1422 } 1423 1424 void 1425 vfs_composefh_free(fhandle_t *fhp) 1426 { 1427 1428 vfs__fhfree(fhp); 1429 } 1430 1431 /* 1432 * vfs_fhtovp: lookup a vnode by a filehandle. 1433 */ 1434 1435 int 1436 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1437 { 1438 struct mount *mp; 1439 int error; 1440 1441 *vpp = NULL; 1442 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1443 if (mp == NULL) { 1444 error = ESTALE; 1445 goto out; 1446 } 1447 if (mp->mnt_op->vfs_fhtovp == NULL) { 1448 error = EOPNOTSUPP; 1449 goto out; 1450 } 1451 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1452 out: 1453 return error; 1454 } 1455 1456 /* 1457 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1458 * the needed size. 1459 */ 1460 1461 int 1462 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1463 { 1464 fhandle_t *fhp; 1465 int error; 1466 1467 *fhpp = NULL; 1468 if (fhsize > FHANDLE_SIZE_MAX) { 1469 return EINVAL; 1470 } 1471 if (fhsize < FHANDLE_SIZE_MIN) { 1472 return EINVAL; 1473 } 1474 again: 1475 fhp = kmem_alloc(fhsize, KM_SLEEP); 1476 if (fhp == NULL) { 1477 return ENOMEM; 1478 } 1479 error = copyin(ufhp, fhp, fhsize); 1480 if (error == 0) { 1481 /* XXX this check shouldn't be here */ 1482 if (FHANDLE_SIZE(fhp) == fhsize) { 1483 *fhpp = fhp; 1484 return 0; 1485 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1486 /* 1487 * a kludge for nfsv2 padded handles. 1488 */ 1489 size_t sz; 1490 1491 sz = FHANDLE_SIZE(fhp); 1492 kmem_free(fhp, fhsize); 1493 fhsize = sz; 1494 goto again; 1495 } else { 1496 /* 1497 * userland told us wrong size. 1498 */ 1499 error = EINVAL; 1500 } 1501 } 1502 kmem_free(fhp, fhsize); 1503 return error; 1504 } 1505 1506 void 1507 vfs_copyinfh_free(fhandle_t *fhp) 1508 { 1509 1510 vfs__fhfree(fhp); 1511 } 1512 1513 /* 1514 * Get file handle system call 1515 */ 1516 int 1517 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1518 { 1519 /* { 1520 syscallarg(char *) fname; 1521 syscallarg(fhandle_t *) fhp; 1522 syscallarg(size_t *) fh_size; 1523 } */ 1524 struct vnode *vp; 1525 fhandle_t *fh; 1526 int error; 1527 struct nameidata nd; 1528 size_t sz; 1529 size_t usz; 1530 1531 /* 1532 * Must be super user 1533 */ 1534 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1535 0, NULL, NULL, NULL); 1536 if (error) 1537 return (error); 1538 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1539 SCARG(uap, fname)); 1540 error = namei(&nd); 1541 if (error) 1542 return (error); 1543 vp = nd.ni_vp; 1544 error = vfs_composefh_alloc(vp, &fh); 1545 vput(vp); 1546 if (error != 0) { 1547 goto out; 1548 } 1549 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1550 if (error != 0) { 1551 goto out; 1552 } 1553 sz = FHANDLE_SIZE(fh); 1554 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1555 if (error != 0) { 1556 goto out; 1557 } 1558 if (usz >= sz) { 1559 error = copyout(fh, SCARG(uap, fhp), sz); 1560 } else { 1561 error = E2BIG; 1562 } 1563 out: 1564 vfs_composefh_free(fh); 1565 return (error); 1566 } 1567 1568 /* 1569 * Open a file given a file handle. 1570 * 1571 * Check permissions, allocate an open file structure, 1572 * and call the device open routine if any. 1573 */ 1574 1575 int 1576 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1577 register_t *retval) 1578 { 1579 struct filedesc *fdp = l->l_proc->p_fd; 1580 struct file *fp; 1581 struct vnode *vp = NULL; 1582 kauth_cred_t cred = l->l_cred; 1583 struct file *nfp; 1584 int type, indx, error=0; 1585 struct flock lf; 1586 struct vattr va; 1587 fhandle_t *fh; 1588 int flags; 1589 1590 /* 1591 * Must be super user 1592 */ 1593 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1594 0, NULL, NULL, NULL))) 1595 return (error); 1596 1597 flags = FFLAGS(oflags); 1598 if ((flags & (FREAD | FWRITE)) == 0) 1599 return (EINVAL); 1600 if ((flags & O_CREAT)) 1601 return (EINVAL); 1602 /* falloc() will use the file descriptor for us */ 1603 if ((error = falloc(l, &nfp, &indx)) != 0) 1604 return (error); 1605 fp = nfp; 1606 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1607 if (error != 0) { 1608 goto bad; 1609 } 1610 error = vfs_fhtovp(fh, &vp); 1611 if (error != 0) { 1612 goto bad; 1613 } 1614 1615 /* Now do an effective vn_open */ 1616 1617 if (vp->v_type == VSOCK) { 1618 error = EOPNOTSUPP; 1619 goto bad; 1620 } 1621 error = vn_openchk(vp, cred, flags); 1622 if (error != 0) 1623 goto bad; 1624 if (flags & O_TRUNC) { 1625 VOP_UNLOCK(vp, 0); /* XXX */ 1626 VOP_LEASE(vp, cred, LEASE_WRITE); 1627 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1628 VATTR_NULL(&va); 1629 va.va_size = 0; 1630 error = VOP_SETATTR(vp, &va, cred); 1631 if (error) 1632 goto bad; 1633 } 1634 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1635 goto bad; 1636 if (flags & FWRITE) 1637 vp->v_writecount++; 1638 1639 /* done with modified vn_open, now finish what sys_open does. */ 1640 1641 fp->f_flag = flags & FMASK; 1642 fp->f_type = DTYPE_VNODE; 1643 fp->f_ops = &vnops; 1644 fp->f_data = vp; 1645 if (flags & (O_EXLOCK | O_SHLOCK)) { 1646 lf.l_whence = SEEK_SET; 1647 lf.l_start = 0; 1648 lf.l_len = 0; 1649 if (flags & O_EXLOCK) 1650 lf.l_type = F_WRLCK; 1651 else 1652 lf.l_type = F_RDLCK; 1653 type = F_FLOCK; 1654 if ((flags & FNONBLOCK) == 0) 1655 type |= F_WAIT; 1656 VOP_UNLOCK(vp, 0); 1657 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1658 if (error) { 1659 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1660 FILE_UNUSE(fp, l); 1661 ffree(fp); 1662 fdremove(fdp, indx); 1663 return (error); 1664 } 1665 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1666 fp->f_flag |= FHASLOCK; 1667 } 1668 VOP_UNLOCK(vp, 0); 1669 *retval = indx; 1670 FILE_SET_MATURE(fp); 1671 FILE_UNUSE(fp, l); 1672 vfs_copyinfh_free(fh); 1673 return (0); 1674 1675 bad: 1676 FILE_UNUSE(fp, l); 1677 ffree(fp); 1678 fdremove(fdp, indx); 1679 if (vp != NULL) 1680 vput(vp); 1681 vfs_copyinfh_free(fh); 1682 return (error); 1683 } 1684 1685 int 1686 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1687 { 1688 /* { 1689 syscallarg(const void *) fhp; 1690 syscallarg(size_t) fh_size; 1691 syscallarg(int) flags; 1692 } */ 1693 1694 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1695 SCARG(uap, flags), retval); 1696 } 1697 1698 int 1699 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1700 { 1701 int error; 1702 fhandle_t *fh; 1703 struct vnode *vp; 1704 1705 /* 1706 * Must be super user 1707 */ 1708 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1709 0, NULL, NULL, NULL))) 1710 return (error); 1711 1712 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1713 if (error != 0) 1714 return error; 1715 1716 error = vfs_fhtovp(fh, &vp); 1717 vfs_copyinfh_free(fh); 1718 if (error != 0) 1719 return error; 1720 1721 error = vn_stat(vp, sb, l); 1722 vput(vp); 1723 return error; 1724 } 1725 1726 1727 /* ARGSUSED */ 1728 int 1729 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1730 { 1731 /* { 1732 syscallarg(const void *) fhp; 1733 syscallarg(size_t) fh_size; 1734 syscallarg(struct stat *) sb; 1735 } */ 1736 struct stat sb; 1737 int error; 1738 1739 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1740 if (error) 1741 return error; 1742 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1743 } 1744 1745 int 1746 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1747 int flags) 1748 { 1749 fhandle_t *fh; 1750 struct mount *mp; 1751 struct vnode *vp; 1752 int error; 1753 1754 /* 1755 * Must be super user 1756 */ 1757 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1758 0, NULL, NULL, NULL))) 1759 return error; 1760 1761 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1762 if (error != 0) 1763 return error; 1764 1765 error = vfs_fhtovp(fh, &vp); 1766 vfs_copyinfh_free(fh); 1767 if (error != 0) 1768 return error; 1769 1770 mp = vp->v_mount; 1771 error = dostatvfs(mp, sb, l, flags, 1); 1772 vput(vp); 1773 return error; 1774 } 1775 1776 /* ARGSUSED */ 1777 int 1778 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1779 { 1780 /* { 1781 syscallarg(const void *) fhp; 1782 syscallarg(size_t) fh_size; 1783 syscallarg(struct statvfs *) buf; 1784 syscallarg(int) flags; 1785 } */ 1786 struct statvfs *sb = STATVFSBUF_GET(); 1787 int error; 1788 1789 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1790 SCARG(uap, flags)); 1791 if (error == 0) 1792 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1793 STATVFSBUF_PUT(sb); 1794 return error; 1795 } 1796 1797 /* 1798 * Create a special file. 1799 */ 1800 /* ARGSUSED */ 1801 int 1802 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1803 { 1804 /* { 1805 syscallarg(const char *) path; 1806 syscallarg(int) mode; 1807 syscallarg(int) dev; 1808 } */ 1809 struct proc *p = l->l_proc; 1810 struct vnode *vp; 1811 struct vattr vattr; 1812 int error, optype; 1813 struct nameidata nd; 1814 char *path; 1815 const char *cpath; 1816 enum uio_seg seg = UIO_USERSPACE; 1817 1818 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1819 0, NULL, NULL, NULL)) != 0) 1820 return (error); 1821 1822 optype = VOP_MKNOD_DESCOFFSET; 1823 1824 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1825 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1826 1827 if ((error = namei(&nd)) != 0) 1828 goto out; 1829 vp = nd.ni_vp; 1830 if (vp != NULL) 1831 error = EEXIST; 1832 else { 1833 VATTR_NULL(&vattr); 1834 /* We will read cwdi->cwdi_cmask unlocked. */ 1835 vattr.va_mode = 1836 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1837 vattr.va_rdev = SCARG(uap, dev); 1838 1839 switch (SCARG(uap, mode) & S_IFMT) { 1840 case S_IFMT: /* used by badsect to flag bad sectors */ 1841 vattr.va_type = VBAD; 1842 break; 1843 case S_IFCHR: 1844 vattr.va_type = VCHR; 1845 break; 1846 case S_IFBLK: 1847 vattr.va_type = VBLK; 1848 break; 1849 case S_IFWHT: 1850 optype = VOP_WHITEOUT_DESCOFFSET; 1851 break; 1852 case S_IFREG: 1853 #if NVERIEXEC > 0 1854 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1855 O_CREAT); 1856 #endif /* NVERIEXEC > 0 */ 1857 vattr.va_type = VREG; 1858 vattr.va_rdev = VNOVAL; 1859 optype = VOP_CREATE_DESCOFFSET; 1860 break; 1861 default: 1862 error = EINVAL; 1863 break; 1864 } 1865 } 1866 if (!error) { 1867 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1868 switch (optype) { 1869 case VOP_WHITEOUT_DESCOFFSET: 1870 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1871 if (error) 1872 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1873 vput(nd.ni_dvp); 1874 break; 1875 1876 case VOP_MKNOD_DESCOFFSET: 1877 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1878 &nd.ni_cnd, &vattr); 1879 if (error == 0) 1880 vput(nd.ni_vp); 1881 break; 1882 1883 case VOP_CREATE_DESCOFFSET: 1884 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1885 &nd.ni_cnd, &vattr); 1886 if (error == 0) 1887 vput(nd.ni_vp); 1888 break; 1889 } 1890 } else { 1891 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1892 if (nd.ni_dvp == vp) 1893 vrele(nd.ni_dvp); 1894 else 1895 vput(nd.ni_dvp); 1896 if (vp) 1897 vrele(vp); 1898 } 1899 out: 1900 VERIEXEC_PATH_PUT(path); 1901 return (error); 1902 } 1903 1904 /* 1905 * Create a named pipe. 1906 */ 1907 /* ARGSUSED */ 1908 int 1909 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1910 { 1911 /* { 1912 syscallarg(const char *) path; 1913 syscallarg(int) mode; 1914 } */ 1915 struct proc *p = l->l_proc; 1916 struct vattr vattr; 1917 int error; 1918 struct nameidata nd; 1919 1920 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1921 SCARG(uap, path)); 1922 if ((error = namei(&nd)) != 0) 1923 return (error); 1924 if (nd.ni_vp != NULL) { 1925 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1926 if (nd.ni_dvp == nd.ni_vp) 1927 vrele(nd.ni_dvp); 1928 else 1929 vput(nd.ni_dvp); 1930 vrele(nd.ni_vp); 1931 return (EEXIST); 1932 } 1933 VATTR_NULL(&vattr); 1934 vattr.va_type = VFIFO; 1935 /* We will read cwdi->cwdi_cmask unlocked. */ 1936 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1937 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1938 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1939 if (error == 0) 1940 vput(nd.ni_vp); 1941 return (error); 1942 } 1943 1944 /* 1945 * Make a hard file link. 1946 */ 1947 /* ARGSUSED */ 1948 int 1949 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1950 { 1951 /* { 1952 syscallarg(const char *) path; 1953 syscallarg(const char *) link; 1954 } */ 1955 struct vnode *vp; 1956 struct nameidata nd; 1957 int error; 1958 1959 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1960 SCARG(uap, path)); 1961 if ((error = namei(&nd)) != 0) 1962 return (error); 1963 vp = nd.ni_vp; 1964 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1965 SCARG(uap, link)); 1966 if ((error = namei(&nd)) != 0) 1967 goto out; 1968 if (nd.ni_vp) { 1969 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1970 if (nd.ni_dvp == nd.ni_vp) 1971 vrele(nd.ni_dvp); 1972 else 1973 vput(nd.ni_dvp); 1974 vrele(nd.ni_vp); 1975 error = EEXIST; 1976 goto out; 1977 } 1978 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1979 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 1980 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1981 out: 1982 vrele(vp); 1983 return (error); 1984 } 1985 1986 /* 1987 * Make a symbolic link. 1988 */ 1989 /* ARGSUSED */ 1990 int 1991 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1992 { 1993 /* { 1994 syscallarg(const char *) path; 1995 syscallarg(const char *) link; 1996 } */ 1997 struct proc *p = l->l_proc; 1998 struct vattr vattr; 1999 char *path; 2000 int error; 2001 struct nameidata nd; 2002 2003 path = PNBUF_GET(); 2004 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2005 if (error) 2006 goto out; 2007 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2008 SCARG(uap, link)); 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_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, const struct sys_undelete_args *uap, register_t *retval) 2040 { 2041 /* { 2042 syscallarg(const char *) path; 2043 } */ 2044 int error; 2045 struct nameidata nd; 2046 2047 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2048 UIO_USERSPACE, SCARG(uap, path)); 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_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, const struct sys_unlink_args *uap, register_t *retval) 2076 { 2077 /* { 2078 syscallarg(const char *) path; 2079 } */ 2080 2081 return do_sys_unlink(SCARG(uap, path), UIO_USERSPACE); 2082 } 2083 2084 int 2085 do_sys_unlink(const char *arg, enum uio_seg seg) 2086 { 2087 struct vnode *vp; 2088 int error; 2089 struct nameidata nd; 2090 kauth_cred_t cred; 2091 char *path; 2092 const char *cpath; 2093 2094 VERIEXEC_PATH_GET(arg, seg, cpath, path); 2095 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2096 2097 if ((error = namei(&nd)) != 0) 2098 goto out; 2099 vp = nd.ni_vp; 2100 2101 /* 2102 * The root of a mounted filesystem cannot be deleted. 2103 */ 2104 if (vp->v_vflag & VV_ROOT) { 2105 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2106 if (nd.ni_dvp == vp) 2107 vrele(nd.ni_dvp); 2108 else 2109 vput(nd.ni_dvp); 2110 vput(vp); 2111 error = EBUSY; 2112 goto out; 2113 } 2114 2115 #if NVERIEXEC > 0 2116 /* Handle remove requests for veriexec entries. */ 2117 if ((error = veriexec_removechk(curlwp, nd.ni_vp, nd.ni_dirp)) != 0) { 2118 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2119 if (nd.ni_dvp == vp) 2120 vrele(nd.ni_dvp); 2121 else 2122 vput(nd.ni_dvp); 2123 vput(vp); 2124 goto out; 2125 } 2126 #endif /* NVERIEXEC > 0 */ 2127 2128 cred = kauth_cred_get(); 2129 VOP_LEASE(nd.ni_dvp, cred, LEASE_WRITE); 2130 VOP_LEASE(vp, cred, LEASE_WRITE); 2131 #ifdef FILEASSOC 2132 (void)fileassoc_file_delete(vp); 2133 #endif /* FILEASSOC */ 2134 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2135 out: 2136 VERIEXEC_PATH_PUT(path); 2137 return (error); 2138 } 2139 2140 /* 2141 * Reposition read/write file offset. 2142 */ 2143 int 2144 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2145 { 2146 /* { 2147 syscallarg(int) fd; 2148 syscallarg(int) pad; 2149 syscallarg(off_t) offset; 2150 syscallarg(int) whence; 2151 } */ 2152 struct proc *p = l->l_proc; 2153 kauth_cred_t cred = l->l_cred; 2154 struct filedesc *fdp = p->p_fd; 2155 struct file *fp; 2156 struct vnode *vp; 2157 struct vattr vattr; 2158 off_t newoff; 2159 int error; 2160 2161 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2162 return (EBADF); 2163 2164 vp = (struct vnode *)fp->f_data; 2165 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2166 error = ESPIPE; 2167 mutex_exit(&fp->f_lock); 2168 goto out; 2169 } 2170 2171 switch (SCARG(uap, whence)) { 2172 case SEEK_CUR: 2173 newoff = fp->f_offset + SCARG(uap, offset); 2174 FILE_USE(fp); 2175 break; 2176 case SEEK_END: 2177 FILE_USE(fp); 2178 error = VOP_GETATTR(vp, &vattr, cred); 2179 if (error) { 2180 FILE_UNUSE(fp, l); 2181 goto out; 2182 } 2183 newoff = SCARG(uap, offset) + vattr.va_size; 2184 break; 2185 case SEEK_SET: 2186 FILE_USE(fp); 2187 newoff = SCARG(uap, offset); 2188 break; 2189 default: 2190 mutex_exit(&fp->f_lock); 2191 error = EINVAL; 2192 goto out; 2193 } 2194 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2195 mutex_enter(&fp->f_lock); 2196 *(off_t *)retval = fp->f_offset = newoff; 2197 mutex_exit(&fp->f_lock); 2198 } 2199 FILE_UNUSE(fp, l); 2200 out: 2201 return (error); 2202 } 2203 2204 /* 2205 * Positional read system call. 2206 */ 2207 int 2208 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2209 { 2210 /* { 2211 syscallarg(int) fd; 2212 syscallarg(void *) buf; 2213 syscallarg(size_t) nbyte; 2214 syscallarg(off_t) offset; 2215 } */ 2216 struct proc *p = l->l_proc; 2217 struct filedesc *fdp = p->p_fd; 2218 struct file *fp; 2219 struct vnode *vp; 2220 off_t offset; 2221 int error, fd = SCARG(uap, fd); 2222 2223 if ((fp = fd_getfile(fdp, fd)) == NULL) 2224 return (EBADF); 2225 2226 if ((fp->f_flag & FREAD) == 0) { 2227 mutex_exit(&fp->f_lock); 2228 return (EBADF); 2229 } 2230 2231 FILE_USE(fp); 2232 2233 vp = (struct vnode *)fp->f_data; 2234 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2235 error = ESPIPE; 2236 goto out; 2237 } 2238 2239 offset = SCARG(uap, offset); 2240 2241 /* 2242 * XXX This works because no file systems actually 2243 * XXX take any action on the seek operation. 2244 */ 2245 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2246 goto out; 2247 2248 /* dofileread() will unuse the descriptor for us */ 2249 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2250 &offset, 0, retval)); 2251 2252 out: 2253 FILE_UNUSE(fp, l); 2254 return (error); 2255 } 2256 2257 /* 2258 * Positional scatter read system call. 2259 */ 2260 int 2261 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2262 { 2263 /* { 2264 syscallarg(int) fd; 2265 syscallarg(const struct iovec *) iovp; 2266 syscallarg(int) iovcnt; 2267 syscallarg(off_t) offset; 2268 } */ 2269 off_t offset = SCARG(uap, offset); 2270 2271 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2272 SCARG(uap, iovcnt), &offset, 0, retval); 2273 } 2274 2275 /* 2276 * Positional write system call. 2277 */ 2278 int 2279 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2280 { 2281 /* { 2282 syscallarg(int) fd; 2283 syscallarg(const void *) buf; 2284 syscallarg(size_t) nbyte; 2285 syscallarg(off_t) offset; 2286 } */ 2287 struct proc *p = l->l_proc; 2288 struct filedesc *fdp = p->p_fd; 2289 struct file *fp; 2290 struct vnode *vp; 2291 off_t offset; 2292 int error, fd = SCARG(uap, fd); 2293 2294 if ((fp = fd_getfile(fdp, fd)) == NULL) 2295 return (EBADF); 2296 2297 if ((fp->f_flag & FWRITE) == 0) { 2298 mutex_exit(&fp->f_lock); 2299 return (EBADF); 2300 } 2301 2302 FILE_USE(fp); 2303 2304 vp = (struct vnode *)fp->f_data; 2305 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2306 error = ESPIPE; 2307 goto out; 2308 } 2309 2310 offset = SCARG(uap, offset); 2311 2312 /* 2313 * XXX This works because no file systems actually 2314 * XXX take any action on the seek operation. 2315 */ 2316 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2317 goto out; 2318 2319 /* dofilewrite() will unuse the descriptor for us */ 2320 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2321 &offset, 0, retval)); 2322 2323 out: 2324 FILE_UNUSE(fp, l); 2325 return (error); 2326 } 2327 2328 /* 2329 * Positional gather write system call. 2330 */ 2331 int 2332 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2333 { 2334 /* { 2335 syscallarg(int) fd; 2336 syscallarg(const struct iovec *) iovp; 2337 syscallarg(int) iovcnt; 2338 syscallarg(off_t) offset; 2339 } */ 2340 off_t offset = SCARG(uap, offset); 2341 2342 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2343 SCARG(uap, iovcnt), &offset, 0, retval); 2344 } 2345 2346 /* 2347 * Check access permissions. 2348 */ 2349 int 2350 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2351 { 2352 /* { 2353 syscallarg(const char *) path; 2354 syscallarg(int) flags; 2355 } */ 2356 kauth_cred_t cred; 2357 struct vnode *vp; 2358 int error, flags; 2359 struct nameidata nd; 2360 2361 cred = kauth_cred_dup(l->l_cred); 2362 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2363 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2364 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2365 SCARG(uap, path)); 2366 /* Override default credentials */ 2367 nd.ni_cnd.cn_cred = cred; 2368 if ((error = namei(&nd)) != 0) 2369 goto out; 2370 vp = nd.ni_vp; 2371 2372 /* Flags == 0 means only check for existence. */ 2373 if (SCARG(uap, flags)) { 2374 flags = 0; 2375 if (SCARG(uap, flags) & R_OK) 2376 flags |= VREAD; 2377 if (SCARG(uap, flags) & W_OK) 2378 flags |= VWRITE; 2379 if (SCARG(uap, flags) & X_OK) 2380 flags |= VEXEC; 2381 2382 error = VOP_ACCESS(vp, flags, cred); 2383 if (!error && (flags & VWRITE)) 2384 error = vn_writechk(vp); 2385 } 2386 vput(vp); 2387 out: 2388 kauth_cred_free(cred); 2389 return (error); 2390 } 2391 2392 /* 2393 * Common code for all sys_stat functions, including compat versions. 2394 */ 2395 int 2396 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2397 struct stat *sb) 2398 { 2399 int error; 2400 struct nameidata nd; 2401 2402 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2403 UIO_USERSPACE, path); 2404 error = namei(&nd); 2405 if (error != 0) 2406 return error; 2407 error = vn_stat(nd.ni_vp, sb, l); 2408 vput(nd.ni_vp); 2409 return error; 2410 } 2411 2412 /* 2413 * Get file status; this version follows links. 2414 */ 2415 /* ARGSUSED */ 2416 int 2417 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2418 { 2419 /* { 2420 syscallarg(const char *) path; 2421 syscallarg(struct stat *) ub; 2422 } */ 2423 struct stat sb; 2424 int error; 2425 2426 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2427 if (error) 2428 return error; 2429 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2430 } 2431 2432 /* 2433 * Get file status; this version does not follow links. 2434 */ 2435 /* ARGSUSED */ 2436 int 2437 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2438 { 2439 /* { 2440 syscallarg(const char *) path; 2441 syscallarg(struct stat *) ub; 2442 } */ 2443 struct stat sb; 2444 int error; 2445 2446 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2447 if (error) 2448 return error; 2449 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2450 } 2451 2452 /* 2453 * Get configurable pathname variables. 2454 */ 2455 /* ARGSUSED */ 2456 int 2457 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2458 { 2459 /* { 2460 syscallarg(const char *) path; 2461 syscallarg(int) name; 2462 } */ 2463 int error; 2464 struct nameidata nd; 2465 2466 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2467 SCARG(uap, path)); 2468 if ((error = namei(&nd)) != 0) 2469 return (error); 2470 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2471 vput(nd.ni_vp); 2472 return (error); 2473 } 2474 2475 /* 2476 * Return target name of a symbolic link. 2477 */ 2478 /* ARGSUSED */ 2479 int 2480 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2481 { 2482 /* { 2483 syscallarg(const char *) path; 2484 syscallarg(char *) buf; 2485 syscallarg(size_t) count; 2486 } */ 2487 struct vnode *vp; 2488 struct iovec aiov; 2489 struct uio auio; 2490 int error; 2491 struct nameidata nd; 2492 2493 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2494 SCARG(uap, path)); 2495 if ((error = namei(&nd)) != 0) 2496 return (error); 2497 vp = nd.ni_vp; 2498 if (vp->v_type != VLNK) 2499 error = EINVAL; 2500 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2501 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2502 aiov.iov_base = SCARG(uap, buf); 2503 aiov.iov_len = SCARG(uap, count); 2504 auio.uio_iov = &aiov; 2505 auio.uio_iovcnt = 1; 2506 auio.uio_offset = 0; 2507 auio.uio_rw = UIO_READ; 2508 KASSERT(l == curlwp); 2509 auio.uio_vmspace = l->l_proc->p_vmspace; 2510 auio.uio_resid = SCARG(uap, count); 2511 error = VOP_READLINK(vp, &auio, l->l_cred); 2512 } 2513 vput(vp); 2514 *retval = SCARG(uap, count) - auio.uio_resid; 2515 return (error); 2516 } 2517 2518 /* 2519 * Change flags of a file given a path name. 2520 */ 2521 /* ARGSUSED */ 2522 int 2523 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2524 { 2525 /* { 2526 syscallarg(const char *) path; 2527 syscallarg(u_long) flags; 2528 } */ 2529 struct vnode *vp; 2530 int error; 2531 struct nameidata nd; 2532 2533 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2534 SCARG(uap, path)); 2535 if ((error = namei(&nd)) != 0) 2536 return (error); 2537 vp = nd.ni_vp; 2538 error = change_flags(vp, SCARG(uap, flags), l); 2539 vput(vp); 2540 return (error); 2541 } 2542 2543 /* 2544 * Change flags of a file given a file descriptor. 2545 */ 2546 /* ARGSUSED */ 2547 int 2548 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2549 { 2550 /* { 2551 syscallarg(int) fd; 2552 syscallarg(u_long) flags; 2553 } */ 2554 struct proc *p = l->l_proc; 2555 struct vnode *vp; 2556 struct file *fp; 2557 int error; 2558 2559 /* getvnode() will use the descriptor for us */ 2560 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2561 return (error); 2562 vp = (struct vnode *)fp->f_data; 2563 error = change_flags(vp, SCARG(uap, flags), l); 2564 VOP_UNLOCK(vp, 0); 2565 FILE_UNUSE(fp, l); 2566 return (error); 2567 } 2568 2569 /* 2570 * Change flags of a file given a path name; this version does 2571 * not follow links. 2572 */ 2573 int 2574 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2575 { 2576 /* { 2577 syscallarg(const char *) path; 2578 syscallarg(u_long) flags; 2579 } */ 2580 struct vnode *vp; 2581 int error; 2582 struct nameidata nd; 2583 2584 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2585 SCARG(uap, path)); 2586 if ((error = namei(&nd)) != 0) 2587 return (error); 2588 vp = nd.ni_vp; 2589 error = change_flags(vp, SCARG(uap, flags), l); 2590 vput(vp); 2591 return (error); 2592 } 2593 2594 /* 2595 * Common routine to change flags of a file. 2596 */ 2597 int 2598 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2599 { 2600 struct vattr vattr; 2601 int error; 2602 2603 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2604 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2605 /* 2606 * Non-superusers cannot change the flags on devices, even if they 2607 * own them. 2608 */ 2609 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2610 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2611 goto out; 2612 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2613 error = EINVAL; 2614 goto out; 2615 } 2616 } 2617 VATTR_NULL(&vattr); 2618 vattr.va_flags = flags; 2619 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2620 out: 2621 return (error); 2622 } 2623 2624 /* 2625 * Change mode of a file given path name; this version follows links. 2626 */ 2627 /* ARGSUSED */ 2628 int 2629 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2630 { 2631 /* { 2632 syscallarg(const char *) path; 2633 syscallarg(int) mode; 2634 } */ 2635 int error; 2636 struct nameidata nd; 2637 2638 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2639 SCARG(uap, path)); 2640 if ((error = namei(&nd)) != 0) 2641 return (error); 2642 2643 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2644 2645 vrele(nd.ni_vp); 2646 return (error); 2647 } 2648 2649 /* 2650 * Change mode of a file given a file descriptor. 2651 */ 2652 /* ARGSUSED */ 2653 int 2654 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2655 { 2656 /* { 2657 syscallarg(int) fd; 2658 syscallarg(int) mode; 2659 } */ 2660 struct proc *p = l->l_proc; 2661 struct file *fp; 2662 int error; 2663 2664 /* getvnode() will use the descriptor for us */ 2665 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2666 return (error); 2667 2668 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2669 FILE_UNUSE(fp, l); 2670 return (error); 2671 } 2672 2673 /* 2674 * Change mode of a file given path name; this version does not follow links. 2675 */ 2676 /* ARGSUSED */ 2677 int 2678 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2679 { 2680 /* { 2681 syscallarg(const char *) path; 2682 syscallarg(int) mode; 2683 } */ 2684 int error; 2685 struct nameidata nd; 2686 2687 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2688 SCARG(uap, path)); 2689 if ((error = namei(&nd)) != 0) 2690 return (error); 2691 2692 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2693 2694 vrele(nd.ni_vp); 2695 return (error); 2696 } 2697 2698 /* 2699 * Common routine to set mode given a vnode. 2700 */ 2701 static int 2702 change_mode(struct vnode *vp, int mode, struct lwp *l) 2703 { 2704 struct vattr vattr; 2705 int error; 2706 2707 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2708 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2709 VATTR_NULL(&vattr); 2710 vattr.va_mode = mode & ALLPERMS; 2711 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2712 VOP_UNLOCK(vp, 0); 2713 return (error); 2714 } 2715 2716 /* 2717 * Set ownership given a path name; this version follows links. 2718 */ 2719 /* ARGSUSED */ 2720 int 2721 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2722 { 2723 /* { 2724 syscallarg(const char *) path; 2725 syscallarg(uid_t) uid; 2726 syscallarg(gid_t) gid; 2727 } */ 2728 int error; 2729 struct nameidata nd; 2730 2731 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2732 SCARG(uap, path)); 2733 if ((error = namei(&nd)) != 0) 2734 return (error); 2735 2736 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2737 2738 vrele(nd.ni_vp); 2739 return (error); 2740 } 2741 2742 /* 2743 * Set ownership given a path name; this version follows links. 2744 * Provides POSIX semantics. 2745 */ 2746 /* ARGSUSED */ 2747 int 2748 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2749 { 2750 /* { 2751 syscallarg(const char *) path; 2752 syscallarg(uid_t) uid; 2753 syscallarg(gid_t) gid; 2754 } */ 2755 int error; 2756 struct nameidata nd; 2757 2758 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2759 SCARG(uap, path)); 2760 if ((error = namei(&nd)) != 0) 2761 return (error); 2762 2763 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2764 2765 vrele(nd.ni_vp); 2766 return (error); 2767 } 2768 2769 /* 2770 * Set ownership given a file descriptor. 2771 */ 2772 /* ARGSUSED */ 2773 int 2774 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2775 { 2776 /* { 2777 syscallarg(int) fd; 2778 syscallarg(uid_t) uid; 2779 syscallarg(gid_t) gid; 2780 } */ 2781 struct proc *p = l->l_proc; 2782 int error; 2783 struct file *fp; 2784 2785 /* getvnode() will use the descriptor for us */ 2786 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2787 return (error); 2788 2789 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2790 SCARG(uap, gid), l, 0); 2791 FILE_UNUSE(fp, l); 2792 return (error); 2793 } 2794 2795 /* 2796 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2797 */ 2798 /* ARGSUSED */ 2799 int 2800 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2801 { 2802 /* { 2803 syscallarg(int) fd; 2804 syscallarg(uid_t) uid; 2805 syscallarg(gid_t) gid; 2806 } */ 2807 struct proc *p = l->l_proc; 2808 int error; 2809 struct file *fp; 2810 2811 /* getvnode() will use the descriptor for us */ 2812 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2813 return (error); 2814 2815 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2816 SCARG(uap, gid), l, 1); 2817 FILE_UNUSE(fp, l); 2818 return (error); 2819 } 2820 2821 /* 2822 * Set ownership given a path name; this version does not follow links. 2823 */ 2824 /* ARGSUSED */ 2825 int 2826 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2827 { 2828 /* { 2829 syscallarg(const char *) path; 2830 syscallarg(uid_t) uid; 2831 syscallarg(gid_t) gid; 2832 } */ 2833 int error; 2834 struct nameidata nd; 2835 2836 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2837 SCARG(uap, path)); 2838 if ((error = namei(&nd)) != 0) 2839 return (error); 2840 2841 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2842 2843 vrele(nd.ni_vp); 2844 return (error); 2845 } 2846 2847 /* 2848 * Set ownership given a path name; this version does not follow links. 2849 * Provides POSIX/XPG semantics. 2850 */ 2851 /* ARGSUSED */ 2852 int 2853 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2854 { 2855 /* { 2856 syscallarg(const char *) path; 2857 syscallarg(uid_t) uid; 2858 syscallarg(gid_t) gid; 2859 } */ 2860 int error; 2861 struct nameidata nd; 2862 2863 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2864 SCARG(uap, path)); 2865 if ((error = namei(&nd)) != 0) 2866 return (error); 2867 2868 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2869 2870 vrele(nd.ni_vp); 2871 return (error); 2872 } 2873 2874 /* 2875 * Common routine to set ownership given a vnode. 2876 */ 2877 static int 2878 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2879 int posix_semantics) 2880 { 2881 struct vattr vattr; 2882 mode_t newmode; 2883 int error; 2884 2885 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2886 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2887 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2888 goto out; 2889 2890 #define CHANGED(x) ((int)(x) != -1) 2891 newmode = vattr.va_mode; 2892 if (posix_semantics) { 2893 /* 2894 * POSIX/XPG semantics: if the caller is not the super-user, 2895 * clear set-user-id and set-group-id bits. Both POSIX and 2896 * the XPG consider the behaviour for calls by the super-user 2897 * implementation-defined; we leave the set-user-id and set- 2898 * group-id settings intact in that case. 2899 */ 2900 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2901 NULL) != 0) 2902 newmode &= ~(S_ISUID | S_ISGID); 2903 } else { 2904 /* 2905 * NetBSD semantics: when changing owner and/or group, 2906 * clear the respective bit(s). 2907 */ 2908 if (CHANGED(uid)) 2909 newmode &= ~S_ISUID; 2910 if (CHANGED(gid)) 2911 newmode &= ~S_ISGID; 2912 } 2913 /* Update va_mode iff altered. */ 2914 if (vattr.va_mode == newmode) 2915 newmode = VNOVAL; 2916 2917 VATTR_NULL(&vattr); 2918 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2919 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2920 vattr.va_mode = newmode; 2921 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2922 #undef CHANGED 2923 2924 out: 2925 VOP_UNLOCK(vp, 0); 2926 return (error); 2927 } 2928 2929 /* 2930 * Set the access and modification times given a path name; this 2931 * version follows links. 2932 */ 2933 /* ARGSUSED */ 2934 int 2935 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2936 { 2937 /* { 2938 syscallarg(const char *) path; 2939 syscallarg(const struct timeval *) tptr; 2940 } */ 2941 2942 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2943 SCARG(uap, tptr), UIO_USERSPACE); 2944 } 2945 2946 /* 2947 * Set the access and modification times given a file descriptor. 2948 */ 2949 /* ARGSUSED */ 2950 int 2951 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2952 { 2953 /* { 2954 syscallarg(int) fd; 2955 syscallarg(const struct timeval *) tptr; 2956 } */ 2957 int error; 2958 struct file *fp; 2959 2960 /* getvnode() will use the descriptor for us */ 2961 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2962 return (error); 2963 2964 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2965 SCARG(uap, tptr), UIO_USERSPACE); 2966 2967 FILE_UNUSE(fp, l); 2968 return (error); 2969 } 2970 2971 /* 2972 * Set the access and modification times given a path name; this 2973 * version does not follow links. 2974 */ 2975 int 2976 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2977 { 2978 /* { 2979 syscallarg(const char *) path; 2980 syscallarg(const struct timeval *) tptr; 2981 } */ 2982 2983 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2984 SCARG(uap, tptr), UIO_USERSPACE); 2985 } 2986 2987 /* 2988 * Common routine to set access and modification times given a vnode. 2989 */ 2990 int 2991 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2992 const struct timeval *tptr, enum uio_seg seg) 2993 { 2994 struct vattr vattr; 2995 struct nameidata nd; 2996 int error; 2997 2998 VATTR_NULL(&vattr); 2999 if (tptr == NULL) { 3000 nanotime(&vattr.va_atime); 3001 vattr.va_mtime = vattr.va_atime; 3002 vattr.va_vaflags |= VA_UTIMES_NULL; 3003 } else { 3004 struct timeval tv[2]; 3005 3006 if (seg != UIO_SYSSPACE) { 3007 error = copyin(tptr, &tv, sizeof (tv)); 3008 if (error != 0) 3009 return error; 3010 tptr = tv; 3011 } 3012 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3013 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3014 } 3015 3016 if (vp == NULL) { 3017 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3018 if ((error = namei(&nd)) != 0) 3019 return (error); 3020 vp = nd.ni_vp; 3021 } else 3022 nd.ni_vp = NULL; 3023 3024 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3025 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3026 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3027 VOP_UNLOCK(vp, 0); 3028 3029 if (nd.ni_vp != NULL) 3030 vrele(nd.ni_vp); 3031 3032 return (error); 3033 } 3034 3035 /* 3036 * Truncate a file given its path name. 3037 */ 3038 /* ARGSUSED */ 3039 int 3040 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3041 { 3042 /* { 3043 syscallarg(const char *) path; 3044 syscallarg(int) pad; 3045 syscallarg(off_t) length; 3046 } */ 3047 struct vnode *vp; 3048 struct vattr vattr; 3049 int error; 3050 struct nameidata nd; 3051 3052 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3053 SCARG(uap, path)); 3054 if ((error = namei(&nd)) != 0) 3055 return (error); 3056 vp = nd.ni_vp; 3057 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3058 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3059 if (vp->v_type == VDIR) 3060 error = EISDIR; 3061 else if ((error = vn_writechk(vp)) == 0 && 3062 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3063 VATTR_NULL(&vattr); 3064 vattr.va_size = SCARG(uap, length); 3065 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3066 } 3067 vput(vp); 3068 return (error); 3069 } 3070 3071 /* 3072 * Truncate a file given a file descriptor. 3073 */ 3074 /* ARGSUSED */ 3075 int 3076 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3077 { 3078 /* { 3079 syscallarg(int) fd; 3080 syscallarg(int) pad; 3081 syscallarg(off_t) length; 3082 } */ 3083 struct proc *p = l->l_proc; 3084 struct vattr vattr; 3085 struct vnode *vp; 3086 struct file *fp; 3087 int error; 3088 3089 /* getvnode() will use the descriptor for us */ 3090 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3091 return (error); 3092 if ((fp->f_flag & FWRITE) == 0) { 3093 error = EINVAL; 3094 goto out; 3095 } 3096 vp = (struct vnode *)fp->f_data; 3097 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3098 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3099 if (vp->v_type == VDIR) 3100 error = EISDIR; 3101 else if ((error = vn_writechk(vp)) == 0) { 3102 VATTR_NULL(&vattr); 3103 vattr.va_size = SCARG(uap, length); 3104 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3105 } 3106 VOP_UNLOCK(vp, 0); 3107 out: 3108 FILE_UNUSE(fp, l); 3109 return (error); 3110 } 3111 3112 /* 3113 * Sync an open file. 3114 */ 3115 /* ARGSUSED */ 3116 int 3117 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3118 { 3119 /* { 3120 syscallarg(int) fd; 3121 } */ 3122 struct proc *p = l->l_proc; 3123 struct vnode *vp; 3124 struct file *fp; 3125 int error; 3126 3127 /* getvnode() will use the descriptor for us */ 3128 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3129 return (error); 3130 vp = (struct vnode *)fp->f_data; 3131 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3132 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3133 if (error == 0 && bioopsp != NULL && 3134 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3135 (*bioopsp->io_fsync)(vp, 0); 3136 VOP_UNLOCK(vp, 0); 3137 FILE_UNUSE(fp, l); 3138 return (error); 3139 } 3140 3141 /* 3142 * Sync a range of file data. API modeled after that found in AIX. 3143 * 3144 * FDATASYNC indicates that we need only save enough metadata to be able 3145 * to re-read the written data. Note we duplicate AIX's requirement that 3146 * the file be open for writing. 3147 */ 3148 /* ARGSUSED */ 3149 int 3150 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3151 { 3152 /* { 3153 syscallarg(int) fd; 3154 syscallarg(int) flags; 3155 syscallarg(off_t) start; 3156 syscallarg(off_t) length; 3157 } */ 3158 struct proc *p = l->l_proc; 3159 struct vnode *vp; 3160 struct file *fp; 3161 int flags, nflags; 3162 off_t s, e, len; 3163 int error; 3164 3165 /* getvnode() will use the descriptor for us */ 3166 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3167 return (error); 3168 3169 if ((fp->f_flag & FWRITE) == 0) { 3170 error = EBADF; 3171 goto out; 3172 } 3173 3174 flags = SCARG(uap, flags); 3175 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3176 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3177 error = EINVAL; 3178 goto out; 3179 } 3180 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3181 if (flags & FDATASYNC) 3182 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3183 else 3184 nflags = FSYNC_WAIT; 3185 if (flags & FDISKSYNC) 3186 nflags |= FSYNC_CACHE; 3187 3188 len = SCARG(uap, length); 3189 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3190 if (len) { 3191 s = SCARG(uap, start); 3192 e = s + len; 3193 if (e < s) { 3194 error = EINVAL; 3195 goto out; 3196 } 3197 } else { 3198 e = 0; 3199 s = 0; 3200 } 3201 3202 vp = (struct vnode *)fp->f_data; 3203 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3204 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3205 3206 if (error == 0 && bioopsp != NULL && 3207 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3208 (*bioopsp->io_fsync)(vp, nflags); 3209 3210 VOP_UNLOCK(vp, 0); 3211 out: 3212 FILE_UNUSE(fp, l); 3213 return (error); 3214 } 3215 3216 /* 3217 * Sync the data of an open file. 3218 */ 3219 /* ARGSUSED */ 3220 int 3221 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3222 { 3223 /* { 3224 syscallarg(int) fd; 3225 } */ 3226 struct proc *p = l->l_proc; 3227 struct vnode *vp; 3228 struct file *fp; 3229 int error; 3230 3231 /* getvnode() will use the descriptor for us */ 3232 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3233 return (error); 3234 if ((fp->f_flag & FWRITE) == 0) { 3235 FILE_UNUSE(fp, l); 3236 return (EBADF); 3237 } 3238 vp = (struct vnode *)fp->f_data; 3239 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3240 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3241 VOP_UNLOCK(vp, 0); 3242 FILE_UNUSE(fp, l); 3243 return (error); 3244 } 3245 3246 /* 3247 * Rename files, (standard) BSD semantics frontend. 3248 */ 3249 /* ARGSUSED */ 3250 int 3251 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3252 { 3253 /* { 3254 syscallarg(const char *) from; 3255 syscallarg(const char *) to; 3256 } */ 3257 3258 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 0)); 3259 } 3260 3261 /* 3262 * Rename files, POSIX semantics frontend. 3263 */ 3264 /* ARGSUSED */ 3265 int 3266 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3267 { 3268 /* { 3269 syscallarg(const char *) from; 3270 syscallarg(const char *) to; 3271 } */ 3272 3273 return (do_sys_rename(SCARG(uap, from), SCARG(uap, to), UIO_USERSPACE, 1)); 3274 } 3275 3276 /* 3277 * Rename files. Source and destination must either both be directories, 3278 * or both not be directories. If target is a directory, it must be empty. 3279 * If `from' and `to' refer to the same object, the value of the `retain' 3280 * argument is used to determine whether `from' will be 3281 * 3282 * (retain == 0) deleted unless `from' and `to' refer to the same 3283 * object in the file system's name space (BSD). 3284 * (retain == 1) always retained (POSIX). 3285 */ 3286 int 3287 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 3288 { 3289 struct vnode *tvp, *fvp, *tdvp; 3290 struct nameidata fromnd, tond; 3291 struct lwp *l = curlwp; 3292 struct proc *p; 3293 int error; 3294 3295 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3296 seg, from); 3297 if ((error = namei(&fromnd)) != 0) 3298 return (error); 3299 if (fromnd.ni_dvp != fromnd.ni_vp) 3300 VOP_UNLOCK(fromnd.ni_dvp, 0); 3301 fvp = fromnd.ni_vp; 3302 NDINIT(&tond, RENAME, 3303 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3304 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3305 seg, to); 3306 if ((error = namei(&tond)) != 0) { 3307 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3308 vrele(fromnd.ni_dvp); 3309 vrele(fvp); 3310 goto out1; 3311 } 3312 tdvp = tond.ni_dvp; 3313 tvp = tond.ni_vp; 3314 3315 if (tvp != NULL) { 3316 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3317 error = ENOTDIR; 3318 goto out; 3319 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3320 error = EISDIR; 3321 goto out; 3322 } 3323 } 3324 3325 if (fvp == tdvp) 3326 error = EINVAL; 3327 3328 /* 3329 * Source and destination refer to the same object. 3330 */ 3331 if (fvp == tvp) { 3332 if (retain) 3333 error = -1; 3334 else if (fromnd.ni_dvp == tdvp && 3335 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3336 !memcmp(fromnd.ni_cnd.cn_nameptr, 3337 tond.ni_cnd.cn_nameptr, 3338 fromnd.ni_cnd.cn_namelen)) 3339 error = -1; 3340 } 3341 3342 #if NVERIEXEC > 0 3343 if (!error) { 3344 char *f1, *f2; 3345 3346 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3347 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3348 3349 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3350 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3351 3352 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3353 3354 free(f1, M_TEMP); 3355 free(f2, M_TEMP); 3356 } 3357 #endif /* NVERIEXEC > 0 */ 3358 3359 out: 3360 p = l->l_proc; 3361 if (!error) { 3362 VOP_LEASE(tdvp, l->l_cred, LEASE_WRITE); 3363 if (fromnd.ni_dvp != tdvp) 3364 VOP_LEASE(fromnd.ni_dvp, l->l_cred, LEASE_WRITE); 3365 if (tvp) { 3366 VOP_LEASE(tvp, l->l_cred, LEASE_WRITE); 3367 } 3368 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3369 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3370 } else { 3371 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3372 if (tdvp == tvp) 3373 vrele(tdvp); 3374 else 3375 vput(tdvp); 3376 if (tvp) 3377 vput(tvp); 3378 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3379 vrele(fromnd.ni_dvp); 3380 vrele(fvp); 3381 } 3382 vrele(tond.ni_startdir); 3383 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3384 out1: 3385 if (fromnd.ni_startdir) 3386 vrele(fromnd.ni_startdir); 3387 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3388 return (error == -1 ? 0 : error); 3389 } 3390 3391 /* 3392 * Make a directory file. 3393 */ 3394 /* ARGSUSED */ 3395 int 3396 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3397 { 3398 /* { 3399 syscallarg(const char *) path; 3400 syscallarg(int) mode; 3401 } */ 3402 struct proc *p = l->l_proc; 3403 struct vnode *vp; 3404 struct vattr vattr; 3405 int error; 3406 struct nameidata nd; 3407 3408 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3409 SCARG(uap, path)); 3410 if ((error = namei(&nd)) != 0) 3411 return (error); 3412 vp = nd.ni_vp; 3413 if (vp != NULL) { 3414 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3415 if (nd.ni_dvp == vp) 3416 vrele(nd.ni_dvp); 3417 else 3418 vput(nd.ni_dvp); 3419 vrele(vp); 3420 return (EEXIST); 3421 } 3422 VATTR_NULL(&vattr); 3423 vattr.va_type = VDIR; 3424 /* We will read cwdi->cwdi_cmask unlocked. */ 3425 vattr.va_mode = 3426 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3427 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3428 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3429 if (!error) 3430 vput(nd.ni_vp); 3431 return (error); 3432 } 3433 3434 /* 3435 * Remove a directory file. 3436 */ 3437 /* ARGSUSED */ 3438 int 3439 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3440 { 3441 /* { 3442 syscallarg(const char *) path; 3443 } */ 3444 struct vnode *vp; 3445 int error; 3446 struct nameidata nd; 3447 3448 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3449 SCARG(uap, path)); 3450 if ((error = namei(&nd)) != 0) 3451 return (error); 3452 vp = nd.ni_vp; 3453 if (vp->v_type != VDIR) { 3454 error = ENOTDIR; 3455 goto out; 3456 } 3457 /* 3458 * No rmdir "." please. 3459 */ 3460 if (nd.ni_dvp == vp) { 3461 error = EINVAL; 3462 goto out; 3463 } 3464 /* 3465 * The root of a mounted filesystem cannot be deleted. 3466 */ 3467 if (vp->v_vflag & VV_ROOT) { 3468 error = EBUSY; 3469 goto out; 3470 } 3471 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3472 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3473 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3474 return (error); 3475 3476 out: 3477 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3478 if (nd.ni_dvp == vp) 3479 vrele(nd.ni_dvp); 3480 else 3481 vput(nd.ni_dvp); 3482 vput(vp); 3483 return (error); 3484 } 3485 3486 /* 3487 * Read a block of directory entries in a file system independent format. 3488 */ 3489 int 3490 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3491 { 3492 /* { 3493 syscallarg(int) fd; 3494 syscallarg(char *) buf; 3495 syscallarg(size_t) count; 3496 } */ 3497 struct proc *p = l->l_proc; 3498 struct file *fp; 3499 int error, done; 3500 3501 /* getvnode() will use the descriptor for us */ 3502 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3503 return (error); 3504 if ((fp->f_flag & FREAD) == 0) { 3505 error = EBADF; 3506 goto out; 3507 } 3508 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3509 SCARG(uap, count), &done, l, 0, 0); 3510 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3511 *retval = done; 3512 out: 3513 FILE_UNUSE(fp, l); 3514 return (error); 3515 } 3516 3517 /* 3518 * Set the mode mask for creation of filesystem nodes. 3519 */ 3520 int 3521 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3522 { 3523 /* { 3524 syscallarg(mode_t) newmask; 3525 } */ 3526 struct proc *p = l->l_proc; 3527 struct cwdinfo *cwdi; 3528 3529 /* 3530 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3531 * important is that we serialize changes to the mask. The 3532 * rw_exit() will issue a write memory barrier on our behalf, 3533 * and force the changes out to other CPUs (as it must use an 3534 * atomic operation, draining the local CPU's store buffers). 3535 */ 3536 cwdi = p->p_cwdi; 3537 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3538 *retval = cwdi->cwdi_cmask; 3539 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3540 rw_exit(&cwdi->cwdi_lock); 3541 3542 return (0); 3543 } 3544 3545 /* 3546 * Void all references to file by ripping underlying filesystem 3547 * away from vnode. 3548 */ 3549 /* ARGSUSED */ 3550 int 3551 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3552 { 3553 /* { 3554 syscallarg(const char *) path; 3555 } */ 3556 struct vnode *vp; 3557 struct vattr vattr; 3558 int error; 3559 bool revoke; 3560 struct nameidata nd; 3561 3562 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3563 SCARG(uap, path)); 3564 if ((error = namei(&nd)) != 0) 3565 return (error); 3566 vp = nd.ni_vp; 3567 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3568 goto out; 3569 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3570 (error = kauth_authorize_generic(l->l_cred, 3571 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3572 goto out; 3573 simple_lock(&vp->v_interlock); 3574 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED|VI_LAYER))); 3575 simple_unlock(&vp->v_interlock); 3576 if (revoke) 3577 VOP_REVOKE(vp, REVOKEALL); 3578 out: 3579 vrele(vp); 3580 return (error); 3581 } 3582 3583 /* 3584 * Convert a user file descriptor to a kernel file entry. 3585 */ 3586 int 3587 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3588 { 3589 struct vnode *vp; 3590 struct file *fp; 3591 3592 if ((fp = fd_getfile(fdp, fd)) == NULL) 3593 return (EBADF); 3594 3595 FILE_USE(fp); 3596 3597 if (fp->f_type != DTYPE_VNODE) { 3598 FILE_UNUSE(fp, NULL); 3599 return (EINVAL); 3600 } 3601 3602 vp = (struct vnode *)fp->f_data; 3603 if (vp->v_type == VBAD) { 3604 FILE_UNUSE(fp, NULL); 3605 return (EBADF); 3606 } 3607 3608 *fpp = fp; 3609 return (0); 3610 } 3611
Indexes created Fri Oct 03 02:09:57 GMT 2025