vfs_syscalls.c revision 1.335
1 /* $NetBSD: vfs_syscalls.c,v 1.335 2007/12/20 23:03:13 dsl 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 1.335 2007/12/20 23:03:13 dsl Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "fss.h" 46 #include "veriexec.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/kmem.h> 61 #include <sys/dirent.h> 62 #include <sys/sysctl.h> 63 #include <sys/syscallargs.h> 64 #include <sys/vfs_syscalls.h> 65 #include <sys/ktrace.h> 66 #ifdef FILEASSOC 67 #include <sys/fileassoc.h> 68 #endif /* FILEASSOC */ 69 #include <sys/verified_exec.h> 70 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #endif 79 #include <nfs/nfsproto.h> 80 #ifdef COMPAT_30 81 #include <nfs/nfs.h> 82 #include <nfs/nfs_var.h> 83 #endif 84 85 #if NFSS > 0 86 #include <dev/fssvar.h> 87 #endif 88 89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 90 91 static int change_dir(struct nameidata *, struct lwp *); 92 static int change_flags(struct vnode *, u_long, struct lwp *); 93 static int change_mode(struct vnode *, int, struct lwp *l); 94 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 95 static int rename_files(const char *, const char *, struct lwp *, int); 96 97 void checkdirs(struct vnode *); 98 99 int dovfsusermount = 0; 100 101 /* 102 * Virtual File System System Calls 103 */ 104 105 /* 106 * Mount a file system. 107 */ 108 109 #if defined(COMPAT_09) || defined(COMPAT_43) 110 /* 111 * This table is used to maintain compatibility with 4.3BSD 112 * and NetBSD 0.9 mount syscalls. Note, the order is important! 113 * 114 * Do not modify this table. It should only contain filesystems 115 * supported by NetBSD 0.9 and 4.3BSD. 116 */ 117 const char * const mountcompatnames[] = { 118 NULL, /* 0 = MOUNT_NONE */ 119 MOUNT_FFS, /* 1 = MOUNT_UFS */ 120 MOUNT_NFS, /* 2 */ 121 MOUNT_MFS, /* 3 */ 122 MOUNT_MSDOS, /* 4 */ 123 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 124 MOUNT_FDESC, /* 6 */ 125 MOUNT_KERNFS, /* 7 */ 126 NULL, /* 8 = MOUNT_DEVFS */ 127 MOUNT_AFS, /* 9 */ 128 }; 129 const int nmountcompatnames = sizeof(mountcompatnames) / 130 sizeof(mountcompatnames[0]); 131 #endif /* COMPAT_09 || COMPAT_43 */ 132 133 static int 134 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 135 void *data, size_t *data_len) 136 { 137 struct mount *mp; 138 int error = 0, saved_flags; 139 140 mp = vp->v_mount; 141 saved_flags = mp->mnt_flag; 142 143 /* We can operate only on VV_ROOT nodes. */ 144 if ((vp->v_vflag & VV_ROOT) == 0) { 145 error = EINVAL; 146 goto out; 147 } 148 149 /* 150 * We only allow the filesystem to be reloaded if it 151 * is currently mounted read-only. 152 */ 153 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 154 error = EOPNOTSUPP; /* Needs translation */ 155 goto out; 156 } 157 158 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 159 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 160 if (error) 161 goto out; 162 163 if (vfs_busy(mp, LK_NOWAIT, 0)) { 164 error = EPERM; 165 goto out; 166 } 167 168 mp->mnt_flag &= ~MNT_OP_FLAGS; 169 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 170 171 /* 172 * Set the mount level flags. 173 */ 174 if (flags & MNT_RDONLY) 175 mp->mnt_flag |= MNT_RDONLY; 176 else if (mp->mnt_flag & MNT_RDONLY) 177 mp->mnt_iflag |= IMNT_WANTRDWR; 178 mp->mnt_flag &= 179 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 180 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 181 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 182 mp->mnt_flag |= flags & 183 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 184 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 185 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 186 MNT_IGNORE); 187 188 error = VFS_MOUNT(mp, path, data, data_len); 189 190 #if defined(COMPAT_30) && defined(NFSSERVER) 191 if (error && data != NULL) { 192 int error2; 193 194 /* Update failed; let's try and see if it was an 195 * export request. */ 196 error2 = nfs_update_exports_30(mp, path, data, l); 197 198 /* Only update error code if the export request was 199 * understood but some problem occurred while 200 * processing it. */ 201 if (error2 != EJUSTRETURN) 202 error = error2; 203 } 204 #endif 205 if (mp->mnt_iflag & IMNT_WANTRDWR) 206 mp->mnt_flag &= ~MNT_RDONLY; 207 if (error) 208 mp->mnt_flag = saved_flags; 209 mp->mnt_flag &= ~MNT_OP_FLAGS; 210 mp->mnt_iflag &= ~IMNT_WANTRDWR; 211 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 212 if (mp->mnt_syncer == NULL) 213 error = vfs_allocate_syncvnode(mp); 214 } else { 215 if (mp->mnt_syncer != NULL) 216 vfs_deallocate_syncvnode(mp); 217 } 218 vfs_unbusy(mp); 219 220 out: 221 return (error); 222 } 223 224 static int 225 mount_get_vfsops(const char *fstype, struct vfsops **vfsops) 226 { 227 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 228 int error; 229 230 /* Copy file-system type from userspace. */ 231 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 232 if (error) { 233 #if defined(COMPAT_09) || defined(COMPAT_43) 234 /* 235 * Historically, filesystem types were identified by numbers. 236 * If we get an integer for the filesystem type instead of a 237 * string, we check to see if it matches one of the historic 238 * filesystem types. 239 */ 240 u_long fsindex = (u_long)fstype; 241 if (fsindex >= nmountcompatnames || 242 mountcompatnames[fsindex] == NULL) 243 return ENODEV; 244 strlcpy(fstypename, mountcompatnames[fsindex], 245 sizeof(fstypename)); 246 #else 247 return error; 248 #endif 249 } 250 251 #ifdef COMPAT_10 252 /* Accept `ufs' as an alias for `ffs'. */ 253 if (strcmp(fstypename, "ufs") == 0) 254 fstypename[0] = 'f'; 255 #endif 256 257 if ((*vfsops = vfs_getopsbyname(fstypename)) == NULL) 258 return ENODEV; 259 return 0; 260 } 261 262 static int 263 mount_domount(struct lwp *l, struct vnode **vpp, struct vfsops *vfsops, 264 const char *path, int flags, void *data, size_t *data_len) 265 { 266 struct mount *mp = NULL; 267 struct vnode *vp = *vpp; 268 struct vattr va; 269 int error; 270 271 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 272 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 273 if (error) 274 return error; 275 276 /* Can't make a non-dir a mount-point (from here anyway). */ 277 if (vp->v_type != VDIR) 278 return ENOTDIR; 279 280 /* 281 * If the user is not root, ensure that they own the directory 282 * onto which we are attempting to mount. 283 */ 284 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0 || 285 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 286 (error = kauth_authorize_generic(l->l_cred, 287 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 288 return error; 289 } 290 291 if (flags & MNT_EXPORTED) 292 return EINVAL; 293 294 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) 295 return error; 296 297 /* 298 * Check if a file-system is not already mounted on this vnode. 299 */ 300 if (vp->v_mountedhere != NULL) 301 return EBUSY; 302 303 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 304 305 mp->mnt_op = vfsops; 306 307 TAILQ_INIT(&mp->mnt_vnodelist); 308 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 309 simple_lock_init(&mp->mnt_slock); 310 (void)vfs_busy(mp, LK_NOWAIT, 0); 311 312 mp->mnt_vnodecovered = vp; 313 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 314 mp->mnt_unmounter = NULL; 315 mount_initspecific(mp); 316 317 /* 318 * The underlying file system may refuse the mount for 319 * various reasons. Allow the user to force it to happen. 320 * 321 * Set the mount level flags. 322 */ 323 mp->mnt_flag = flags & 324 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 325 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 326 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 327 MNT_IGNORE | MNT_RDONLY); 328 329 error = VFS_MOUNT(mp, path, data, data_len); 330 mp->mnt_flag &= ~MNT_OP_FLAGS; 331 332 /* 333 * Put the new filesystem on the mount list after root. 334 */ 335 cache_purge(vp); 336 if (error != 0) { 337 vp->v_mountedhere = NULL; 338 mp->mnt_op->vfs_refcount--; 339 vfs_unbusy(mp); 340 vfs_destroy(mp); 341 return error; 342 } 343 344 mp->mnt_iflag &= ~IMNT_WANTRDWR; 345 vp->v_mountedhere = mp; 346 mutex_enter(&mountlist_lock); 347 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 348 mutex_exit(&mountlist_lock); 349 VOP_UNLOCK(vp, 0); 350 checkdirs(vp); 351 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 352 error = vfs_allocate_syncvnode(mp); 353 vfs_unbusy(mp); 354 (void) VFS_STATVFS(mp, &mp->mnt_stat); 355 error = VFS_START(mp, 0); 356 if (error) 357 vrele(vp); 358 *vpp = NULL; 359 return error; 360 } 361 362 static int 363 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 364 void *data, size_t *data_len) 365 { 366 struct mount *mp; 367 int error; 368 369 /* If MNT_GETARGS is specified, it should be the only flag. */ 370 if (flags & ~MNT_GETARGS) 371 return EINVAL; 372 373 mp = vp->v_mount; 374 375 /* XXX: probably some notion of "can see" here if we want isolation. */ 376 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 377 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 378 if (error) 379 return error; 380 381 if ((vp->v_vflag & VV_ROOT) == 0) 382 return EINVAL; 383 384 if (vfs_busy(mp, LK_NOWAIT, 0)) 385 return EPERM; 386 387 mp->mnt_flag &= ~MNT_OP_FLAGS; 388 mp->mnt_flag |= MNT_GETARGS; 389 error = VFS_MOUNT(mp, path, data, data_len); 390 mp->mnt_flag &= ~MNT_OP_FLAGS; 391 392 vfs_unbusy(mp); 393 return (error); 394 } 395 396 #ifdef COMPAT_40 397 /* ARGSUSED */ 398 int 399 compat_40_sys_mount(struct lwp *l, const struct compat_40_sys_mount_args *uap, register_t *retval) 400 { 401 /* { 402 syscallarg(const char *) type; 403 syscallarg(const char *) path; 404 syscallarg(int) flags; 405 syscallarg(void *) data; 406 } */ 407 register_t dummy; 408 409 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 410 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 0, &dummy); 411 } 412 #endif 413 414 int 415 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 416 { 417 /* { 418 syscallarg(const char *) type; 419 syscallarg(const char *) path; 420 syscallarg(int) flags; 421 syscallarg(void *) data; 422 syscallarg(size_t) data_len; 423 } */ 424 425 return do_sys_mount(l, NULL, SCARG(uap, type), SCARG(uap, path), 426 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 427 SCARG(uap, data_len), retval); 428 } 429 430 int 431 do_sys_mount(struct lwp *l, struct vfsops *vfsops, const char *type, 432 const char *path, int flags, void *data, enum uio_seg data_seg, 433 size_t data_len, register_t *retval) 434 { 435 struct vnode *vp; 436 struct nameidata nd; 437 void *data_buf = data; 438 int error; 439 440 /* 441 * Get vnode to be covered 442 */ 443 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 444 if ((error = namei(&nd)) != 0) 445 return (error); 446 vp = nd.ni_vp; 447 448 /* 449 * A lookup in VFS_MOUNT might result in an attempt to 450 * lock this vnode again, so make the lock recursive. 451 */ 452 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 453 454 if (vfsops == NULL) { 455 if (flags & (MNT_GETARGS | MNT_UPDATE)) 456 vfsops = vp->v_mount->mnt_op; 457 else { 458 /* 'type' is userspace */ 459 error = mount_get_vfsops(type, &vfsops); 460 if (error != 0) 461 goto done; 462 } 463 } 464 465 if (data != NULL && data_seg == UIO_USERSPACE) { 466 if (data_len == 0) { 467 /* No length supplied, use default for filesystem */ 468 data_len = vfsops->vfs_min_mount_data; 469 if (data_len > VFS_MAX_MOUNT_DATA) { 470 /* maybe a force loaded old LKM */ 471 error = EINVAL; 472 goto done; 473 } 474 #ifdef COMPAT_30 475 /* Hopefully a longer buffer won't make copyin() fail */ 476 if (flags & MNT_UPDATE 477 && data_len < sizeof (struct mnt_export_args30)) 478 data_len = sizeof (struct mnt_export_args30); 479 #endif 480 } 481 data_buf = malloc(data_len, M_TEMP, M_WAITOK); 482 483 /* NFS needs the buffer even for mnt_getargs .... */ 484 error = copyin(data, data_buf, data_len); 485 if (error != 0) 486 goto done; 487 } 488 489 if (flags & MNT_GETARGS) { 490 if (data_len == 0) { 491 error = EINVAL; 492 goto done; 493 } 494 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 495 if (error != 0) 496 goto done; 497 if (data_seg == UIO_USERSPACE) 498 error = copyout(data_buf, data, data_len); 499 *retval = data_len; 500 } else if (flags & MNT_UPDATE) { 501 error = mount_update(l, vp, path, flags, data_buf, &data_len); 502 } else { 503 /* Locking is handled internally in mount_domount(). */ 504 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 505 &data_len); 506 } 507 508 done: 509 if (vp) 510 vput(vp); 511 if (data_buf != data) 512 free(data_buf, M_TEMP); 513 return (error); 514 } 515 516 /* 517 * Scan all active processes to see if any of them have a current 518 * or root directory onto which the new filesystem has just been 519 * mounted. If so, replace them with the new mount point. 520 */ 521 void 522 checkdirs(struct vnode *olddp) 523 { 524 struct cwdinfo *cwdi; 525 struct vnode *newdp; 526 struct proc *p; 527 528 if (olddp->v_usecount == 1) 529 return; 530 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 531 panic("mount: lost mount"); 532 mutex_enter(&proclist_lock); 533 PROCLIST_FOREACH(p, &allproc) { 534 cwdi = p->p_cwdi; 535 if (!cwdi) 536 continue; 537 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 538 if (cwdi->cwdi_cdir == olddp) { 539 vrele(cwdi->cwdi_cdir); 540 VREF(newdp); 541 cwdi->cwdi_cdir = newdp; 542 } 543 if (cwdi->cwdi_rdir == olddp) { 544 vrele(cwdi->cwdi_rdir); 545 VREF(newdp); 546 cwdi->cwdi_rdir = newdp; 547 } 548 rw_exit(&cwdi->cwdi_lock); 549 } 550 mutex_exit(&proclist_lock); 551 if (rootvnode == olddp) { 552 vrele(rootvnode); 553 VREF(newdp); 554 rootvnode = newdp; 555 } 556 vput(newdp); 557 } 558 559 /* 560 * Unmount a file system. 561 * 562 * Note: unmount takes a path to the vnode mounted on as argument, 563 * not special file (as before). 564 */ 565 /* ARGSUSED */ 566 int 567 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 568 { 569 /* { 570 syscallarg(const char *) path; 571 syscallarg(int) flags; 572 } */ 573 struct vnode *vp; 574 struct mount *mp; 575 int error; 576 struct nameidata nd; 577 578 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 579 SCARG(uap, path)); 580 if ((error = namei(&nd)) != 0) 581 return (error); 582 vp = nd.ni_vp; 583 mp = vp->v_mount; 584 585 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 586 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 587 if (error) { 588 vput(vp); 589 return (error); 590 } 591 592 /* 593 * Don't allow unmounting the root file system. 594 */ 595 if (mp->mnt_flag & MNT_ROOTFS) { 596 vput(vp); 597 return (EINVAL); 598 } 599 600 /* 601 * Must be the root of the filesystem 602 */ 603 if ((vp->v_vflag & VV_ROOT) == 0) { 604 vput(vp); 605 return (EINVAL); 606 } 607 vput(vp); 608 609 /* 610 * XXX Freeze syncer. Must do this before locking the 611 * mount point. See dounmount() for details. 612 */ 613 mutex_enter(&syncer_mutex); 614 615 if (vfs_busy(mp, 0, 0)) { 616 mutex_exit(&syncer_mutex); 617 return (EBUSY); 618 } 619 620 return (dounmount(mp, SCARG(uap, flags), l)); 621 } 622 623 /* 624 * Do the actual file system unmount. File system is assumed to have been 625 * marked busy by the caller. 626 */ 627 int 628 dounmount(struct mount *mp, int flags, struct lwp *l) 629 { 630 struct vnode *coveredvp; 631 int error; 632 int async; 633 int used_syncer; 634 635 #if NVERIEXEC > 0 636 error = veriexec_unmountchk(mp); 637 if (error) 638 return (error); 639 #endif /* NVERIEXEC > 0 */ 640 641 mutex_enter(&mountlist_lock); 642 vfs_unbusy(mp); 643 used_syncer = (mp->mnt_syncer != NULL); 644 645 /* 646 * XXX Syncer must be frozen when we get here. This should really 647 * be done on a per-mountpoint basis, but especially the softdep 648 * code possibly called from the syncer doesn't exactly work on a 649 * per-mountpoint basis, so the softdep code would become a maze 650 * of vfs_busy() calls. 651 * 652 * The caller of dounmount() must acquire syncer_mutex because 653 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 654 * order, and we must preserve that order to avoid deadlock. 655 * 656 * So, if the file system did not use the syncer, now is 657 * the time to release the syncer_mutex. 658 */ 659 if (used_syncer == 0) 660 mutex_exit(&syncer_mutex); 661 662 mp->mnt_iflag |= IMNT_UNMOUNT; 663 mp->mnt_unmounter = l; 664 mutex_exit(&mountlist_lock); /* XXX */ 665 lockmgr(&mp->mnt_lock, LK_DRAIN, NULL); 666 667 async = mp->mnt_flag & MNT_ASYNC; 668 mp->mnt_flag &= ~MNT_ASYNC; 669 cache_purgevfs(mp); /* remove cache entries for this file sys */ 670 if (mp->mnt_syncer != NULL) 671 vfs_deallocate_syncvnode(mp); 672 error = 0; 673 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 674 #if NFSS > 0 675 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 676 #endif 677 if (error == 0) 678 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred); 679 } 680 if (error == 0 || (flags & MNT_FORCE)) 681 error = VFS_UNMOUNT(mp, flags); 682 if (error) { 683 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 684 (void) vfs_allocate_syncvnode(mp); 685 mutex_enter(&mountlist_lock); 686 mp->mnt_iflag &= ~IMNT_UNMOUNT; 687 mp->mnt_unmounter = NULL; 688 mp->mnt_flag |= async; 689 mutex_exit(&mountlist_lock); /* XXX */ 690 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_REENABLE, 691 NULL); 692 if (used_syncer) 693 mutex_exit(&syncer_mutex); 694 simple_lock(&mp->mnt_slock); 695 while (mp->mnt_wcnt > 0) { 696 wakeup(mp); 697 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 698 0, &mp->mnt_slock); 699 } 700 simple_unlock(&mp->mnt_slock); 701 return (error); 702 } 703 mutex_enter(&mountlist_lock); 704 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 705 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 706 coveredvp->v_mountedhere = NULL; 707 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 708 panic("unmount: dangling vnode"); 709 mp->mnt_iflag |= IMNT_GONE; 710 mutex_exit(&mountlist_lock); 711 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL); 712 if (coveredvp != NULLVP) 713 vrele(coveredvp); 714 if (used_syncer) 715 mutex_exit(&syncer_mutex); 716 simple_lock(&mp->mnt_slock); 717 while (mp->mnt_wcnt > 0) { 718 wakeup(mp); 719 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 720 } 721 simple_unlock(&mp->mnt_slock); 722 vfs_hooks_unmount(mp); 723 vfs_delref(mp->mnt_op); 724 vfs_destroy(mp); 725 return (0); 726 } 727 728 /* 729 * Sync each mounted filesystem. 730 */ 731 #ifdef DEBUG 732 int syncprt = 0; 733 struct ctldebug debug0 = { "syncprt", &syncprt }; 734 #endif 735 736 /* ARGSUSED */ 737 int 738 sys_sync(struct lwp *l, const void *v, register_t *retval) 739 { 740 struct mount *mp, *nmp; 741 int asyncflag; 742 743 if (l == NULL) 744 l = &lwp0; 745 746 mutex_enter(&mountlist_lock); 747 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 748 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 749 nmp = mp->mnt_list.cqe_prev; 750 continue; 751 } 752 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 753 asyncflag = mp->mnt_flag & MNT_ASYNC; 754 mp->mnt_flag &= ~MNT_ASYNC; 755 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 756 if (asyncflag) 757 mp->mnt_flag |= MNT_ASYNC; 758 } 759 mutex_enter(&mountlist_lock); 760 nmp = mp->mnt_list.cqe_prev; 761 vfs_unbusy(mp); 762 763 } 764 mutex_exit(&mountlist_lock); 765 #ifdef DEBUG 766 if (syncprt) 767 vfs_bufstats(); 768 #endif /* DEBUG */ 769 return (0); 770 } 771 772 /* 773 * Change filesystem quotas. 774 */ 775 /* ARGSUSED */ 776 int 777 sys_quotactl(struct lwp *l, const struct sys_quotactl_args *uap, register_t *retval) 778 { 779 /* { 780 syscallarg(const char *) path; 781 syscallarg(int) cmd; 782 syscallarg(int) uid; 783 syscallarg(void *) arg; 784 } */ 785 struct mount *mp; 786 int error; 787 struct nameidata nd; 788 789 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 790 SCARG(uap, path)); 791 if ((error = namei(&nd)) != 0) 792 return (error); 793 mp = nd.ni_vp->v_mount; 794 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 795 SCARG(uap, arg)); 796 vrele(nd.ni_vp); 797 return (error); 798 } 799 800 int 801 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 802 int root) 803 { 804 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 805 int error = 0; 806 807 /* 808 * If MNT_NOWAIT or MNT_LAZY is specified, do not 809 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 810 * overrides MNT_NOWAIT. 811 */ 812 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 813 (flags != MNT_WAIT && flags != 0)) { 814 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 815 goto done; 816 } 817 818 /* Get the filesystem stats now */ 819 memset(sp, 0, sizeof(*sp)); 820 if ((error = VFS_STATVFS(mp, sp)) != 0) { 821 return error; 822 } 823 824 if (cwdi->cwdi_rdir == NULL) 825 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 826 done: 827 if (cwdi->cwdi_rdir != NULL) { 828 size_t len; 829 char *bp; 830 char *path = PNBUF_GET(); 831 832 bp = path + MAXPATHLEN; 833 *--bp = '\0'; 834 rw_enter(&cwdi->cwdi_lock, RW_READER); 835 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 836 MAXPATHLEN / 2, 0, l); 837 rw_exit(&cwdi->cwdi_lock); 838 if (error) { 839 PNBUF_PUT(path); 840 return error; 841 } 842 len = strlen(bp); 843 /* 844 * for mount points that are below our root, we can see 845 * them, so we fix up the pathname and return them. The 846 * rest we cannot see, so we don't allow viewing the 847 * data. 848 */ 849 if (strncmp(bp, sp->f_mntonname, len) == 0) { 850 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 851 sizeof(sp->f_mntonname)); 852 if (sp->f_mntonname[0] == '\0') 853 (void)strlcpy(sp->f_mntonname, "/", 854 sizeof(sp->f_mntonname)); 855 } else { 856 if (root) 857 (void)strlcpy(sp->f_mntonname, "/", 858 sizeof(sp->f_mntonname)); 859 else 860 error = EPERM; 861 } 862 PNBUF_PUT(path); 863 } 864 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 865 return error; 866 } 867 868 /* 869 * Get filesystem statistics by path. 870 */ 871 int 872 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 873 { 874 struct mount *mp; 875 int error; 876 struct nameidata nd; 877 878 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path); 879 if ((error = namei(&nd)) != 0) 880 return error; 881 mp = nd.ni_vp->v_mount; 882 error = dostatvfs(mp, sb, l, flags, 1); 883 vrele(nd.ni_vp); 884 return error; 885 } 886 887 /* ARGSUSED */ 888 int 889 sys_statvfs1(struct lwp *l, const struct sys_statvfs1_args *uap, register_t *retval) 890 { 891 /* { 892 syscallarg(const char *) path; 893 syscallarg(struct statvfs *) buf; 894 syscallarg(int) flags; 895 } */ 896 struct statvfs *sb; 897 int error; 898 899 sb = STATVFSBUF_GET(); 900 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 901 if (error == 0) 902 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 903 STATVFSBUF_PUT(sb); 904 return error; 905 } 906 907 /* 908 * Get filesystem statistics by fd. 909 */ 910 int 911 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 912 { 913 struct proc *p = l->l_proc; 914 struct file *fp; 915 struct mount *mp; 916 int error; 917 918 /* getvnode() will use the descriptor for us */ 919 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 920 return (error); 921 mp = ((struct vnode *)fp->f_data)->v_mount; 922 error = dostatvfs(mp, sb, l, flags, 1); 923 FILE_UNUSE(fp, l); 924 return error; 925 } 926 927 /* ARGSUSED */ 928 int 929 sys_fstatvfs1(struct lwp *l, const struct sys_fstatvfs1_args *uap, register_t *retval) 930 { 931 /* { 932 syscallarg(int) fd; 933 syscallarg(struct statvfs *) buf; 934 syscallarg(int) flags; 935 } */ 936 struct statvfs *sb; 937 int error; 938 939 sb = STATVFSBUF_GET(); 940 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 941 if (error == 0) 942 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 943 STATVFSBUF_PUT(sb); 944 return error; 945 } 946 947 948 /* 949 * Get statistics on all filesystems. 950 */ 951 int 952 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 953 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 954 register_t *retval) 955 { 956 int root = 0; 957 struct proc *p = l->l_proc; 958 struct mount *mp, *nmp; 959 struct statvfs *sb; 960 size_t count, maxcount; 961 int error = 0; 962 963 sb = STATVFSBUF_GET(); 964 maxcount = bufsize / entry_sz; 965 mutex_enter(&mountlist_lock); 966 count = 0; 967 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 968 mp = nmp) { 969 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 970 nmp = CIRCLEQ_NEXT(mp, mnt_list); 971 continue; 972 } 973 if (sfsp && count < maxcount) { 974 error = dostatvfs(mp, sb, l, flags, 0); 975 if (error) { 976 mutex_enter(&mountlist_lock); 977 nmp = CIRCLEQ_NEXT(mp, mnt_list); 978 vfs_unbusy(mp); 979 continue; 980 } 981 error = copyfn(sb, sfsp, entry_sz); 982 if (error) { 983 vfs_unbusy(mp); 984 goto out; 985 } 986 sfsp = (char *)sfsp + entry_sz; 987 root |= strcmp(sb->f_mntonname, "/") == 0; 988 } 989 count++; 990 mutex_enter(&mountlist_lock); 991 nmp = CIRCLEQ_NEXT(mp, mnt_list); 992 vfs_unbusy(mp); 993 } 994 995 mutex_exit(&mountlist_lock); 996 if (root == 0 && p->p_cwdi->cwdi_rdir) { 997 /* 998 * fake a root entry 999 */ 1000 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1001 sb, l, flags, 1); 1002 if (error != 0) 1003 goto out; 1004 if (sfsp) 1005 error = copyfn(sb, sfsp, entry_sz); 1006 count++; 1007 } 1008 if (sfsp && count > maxcount) 1009 *retval = maxcount; 1010 else 1011 *retval = count; 1012 out: 1013 STATVFSBUF_PUT(sb); 1014 return error; 1015 } 1016 1017 int 1018 sys_getvfsstat(struct lwp *l, const struct sys_getvfsstat_args *uap, register_t *retval) 1019 { 1020 /* { 1021 syscallarg(struct statvfs *) buf; 1022 syscallarg(size_t) bufsize; 1023 syscallarg(int) flags; 1024 } */ 1025 1026 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1027 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1028 } 1029 1030 /* 1031 * Change current working directory to a given file descriptor. 1032 */ 1033 /* ARGSUSED */ 1034 int 1035 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1036 { 1037 /* { 1038 syscallarg(int) fd; 1039 } */ 1040 struct proc *p = l->l_proc; 1041 struct filedesc *fdp = p->p_fd; 1042 struct cwdinfo *cwdi; 1043 struct vnode *vp, *tdp; 1044 struct mount *mp; 1045 struct file *fp; 1046 int error; 1047 1048 /* getvnode() will use the descriptor for us */ 1049 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1050 return (error); 1051 vp = (struct vnode *)fp->f_data; 1052 1053 VREF(vp); 1054 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1055 if (vp->v_type != VDIR) 1056 error = ENOTDIR; 1057 else 1058 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1059 if (error) { 1060 vput(vp); 1061 goto out; 1062 } 1063 while ((mp = vp->v_mountedhere) != NULL) { 1064 if (vfs_busy(mp, 0, 0)) 1065 continue; 1066 1067 vput(vp); 1068 error = VFS_ROOT(mp, &tdp); 1069 vfs_unbusy(mp); 1070 if (error) 1071 goto out; 1072 vp = tdp; 1073 } 1074 VOP_UNLOCK(vp, 0); 1075 1076 /* 1077 * Disallow changing to a directory not under the process's 1078 * current root directory (if there is one). 1079 */ 1080 cwdi = p->p_cwdi; 1081 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1082 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1083 vrele(vp); 1084 error = EPERM; /* operation not permitted */ 1085 } else { 1086 vrele(cwdi->cwdi_cdir); 1087 cwdi->cwdi_cdir = vp; 1088 } 1089 rw_exit(&cwdi->cwdi_lock); 1090 1091 out: 1092 FILE_UNUSE(fp, l); 1093 return (error); 1094 } 1095 1096 /* 1097 * Change this process's notion of the root directory to a given file 1098 * descriptor. 1099 */ 1100 int 1101 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1102 { 1103 struct proc *p = l->l_proc; 1104 struct filedesc *fdp = p->p_fd; 1105 struct cwdinfo *cwdi; 1106 struct vnode *vp; 1107 struct file *fp; 1108 int error; 1109 1110 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1111 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1112 return error; 1113 /* getvnode() will use the descriptor for us */ 1114 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1115 return error; 1116 vp = (struct vnode *) fp->f_data; 1117 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1118 if (vp->v_type != VDIR) 1119 error = ENOTDIR; 1120 else 1121 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1122 VOP_UNLOCK(vp, 0); 1123 if (error) 1124 goto out; 1125 VREF(vp); 1126 1127 /* 1128 * Prevent escaping from chroot by putting the root under 1129 * the working directory. Silently chdir to / if we aren't 1130 * already there. 1131 */ 1132 cwdi = p->p_cwdi; 1133 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1134 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1135 /* 1136 * XXX would be more failsafe to change directory to a 1137 * deadfs node here instead 1138 */ 1139 vrele(cwdi->cwdi_cdir); 1140 VREF(vp); 1141 cwdi->cwdi_cdir = vp; 1142 } 1143 1144 if (cwdi->cwdi_rdir != NULL) 1145 vrele(cwdi->cwdi_rdir); 1146 cwdi->cwdi_rdir = vp; 1147 rw_exit(&cwdi->cwdi_lock); 1148 1149 out: 1150 FILE_UNUSE(fp, l); 1151 return (error); 1152 } 1153 1154 /* 1155 * Change current working directory (``.''). 1156 */ 1157 /* ARGSUSED */ 1158 int 1159 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1160 { 1161 /* { 1162 syscallarg(const char *) path; 1163 } */ 1164 struct proc *p = l->l_proc; 1165 struct cwdinfo *cwdi; 1166 int error; 1167 struct nameidata nd; 1168 1169 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1170 SCARG(uap, path)); 1171 if ((error = change_dir(&nd, l)) != 0) 1172 return (error); 1173 cwdi = p->p_cwdi; 1174 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1175 vrele(cwdi->cwdi_cdir); 1176 cwdi->cwdi_cdir = nd.ni_vp; 1177 rw_exit(&cwdi->cwdi_lock); 1178 return (0); 1179 } 1180 1181 /* 1182 * Change notion of root (``/'') directory. 1183 */ 1184 /* ARGSUSED */ 1185 int 1186 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1187 { 1188 /* { 1189 syscallarg(const char *) path; 1190 } */ 1191 struct proc *p = l->l_proc; 1192 struct cwdinfo *cwdi; 1193 struct vnode *vp; 1194 int error; 1195 struct nameidata nd; 1196 1197 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1198 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1199 return (error); 1200 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1201 SCARG(uap, path)); 1202 if ((error = change_dir(&nd, l)) != 0) 1203 return (error); 1204 1205 cwdi = p->p_cwdi; 1206 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1207 if (cwdi->cwdi_rdir != NULL) 1208 vrele(cwdi->cwdi_rdir); 1209 vp = nd.ni_vp; 1210 cwdi->cwdi_rdir = vp; 1211 1212 /* 1213 * Prevent escaping from chroot by putting the root under 1214 * the working directory. Silently chdir to / if we aren't 1215 * already there. 1216 */ 1217 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1218 /* 1219 * XXX would be more failsafe to change directory to a 1220 * deadfs node here instead 1221 */ 1222 vrele(cwdi->cwdi_cdir); 1223 VREF(vp); 1224 cwdi->cwdi_cdir = vp; 1225 } 1226 rw_exit(&cwdi->cwdi_lock); 1227 1228 return (0); 1229 } 1230 1231 /* 1232 * Common routine for chroot and chdir. 1233 */ 1234 static int 1235 change_dir(struct nameidata *ndp, struct lwp *l) 1236 { 1237 struct vnode *vp; 1238 int error; 1239 1240 if ((error = namei(ndp)) != 0) 1241 return (error); 1242 vp = ndp->ni_vp; 1243 if (vp->v_type != VDIR) 1244 error = ENOTDIR; 1245 else 1246 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1247 1248 if (error) 1249 vput(vp); 1250 else 1251 VOP_UNLOCK(vp, 0); 1252 return (error); 1253 } 1254 1255 /* 1256 * Check permissions, allocate an open file structure, 1257 * and call the device open routine if any. 1258 */ 1259 int 1260 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1261 { 1262 /* { 1263 syscallarg(const char *) path; 1264 syscallarg(int) flags; 1265 syscallarg(int) mode; 1266 } */ 1267 struct proc *p = l->l_proc; 1268 struct cwdinfo *cwdi = p->p_cwdi; 1269 struct filedesc *fdp = p->p_fd; 1270 struct file *fp; 1271 struct vnode *vp; 1272 int flags, cmode; 1273 int type, indx, error; 1274 struct flock lf; 1275 struct nameidata nd; 1276 1277 flags = FFLAGS(SCARG(uap, flags)); 1278 if ((flags & (FREAD | FWRITE)) == 0) 1279 return (EINVAL); 1280 /* falloc() will use the file descriptor for us */ 1281 if ((error = falloc(l, &fp, &indx)) != 0) 1282 return (error); 1283 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1284 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1285 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1286 SCARG(uap, path)); 1287 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1288 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1289 rw_enter(&fdp->fd_lock, RW_WRITER); 1290 FILE_UNUSE(fp, l); 1291 fdp->fd_ofiles[indx] = NULL; 1292 rw_exit(&fdp->fd_lock); 1293 ffree(fp); 1294 if ((error == EDUPFD || error == EMOVEFD) && 1295 l->l_dupfd >= 0 && /* XXX from fdopen */ 1296 (error = 1297 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1298 *retval = indx; 1299 return (0); 1300 } 1301 if (error == ERESTART) 1302 error = EINTR; 1303 fdremove(fdp, indx); 1304 return (error); 1305 } 1306 1307 l->l_dupfd = 0; 1308 vp = nd.ni_vp; 1309 fp->f_flag = flags & FMASK; 1310 fp->f_type = DTYPE_VNODE; 1311 fp->f_ops = &vnops; 1312 fp->f_data = vp; 1313 if (flags & (O_EXLOCK | O_SHLOCK)) { 1314 lf.l_whence = SEEK_SET; 1315 lf.l_start = 0; 1316 lf.l_len = 0; 1317 if (flags & O_EXLOCK) 1318 lf.l_type = F_WRLCK; 1319 else 1320 lf.l_type = F_RDLCK; 1321 type = F_FLOCK; 1322 if ((flags & FNONBLOCK) == 0) 1323 type |= F_WAIT; 1324 VOP_UNLOCK(vp, 0); 1325 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1326 if (error) { 1327 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1328 FILE_UNUSE(fp, l); 1329 ffree(fp); 1330 fdremove(fdp, indx); 1331 return (error); 1332 } 1333 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1334 fp->f_flag |= FHASLOCK; 1335 } 1336 VOP_UNLOCK(vp, 0); 1337 *retval = indx; 1338 FILE_SET_MATURE(fp); 1339 FILE_UNUSE(fp, l); 1340 return (0); 1341 } 1342 1343 static void 1344 vfs__fhfree(fhandle_t *fhp) 1345 { 1346 size_t fhsize; 1347 1348 if (fhp == NULL) { 1349 return; 1350 } 1351 fhsize = FHANDLE_SIZE(fhp); 1352 kmem_free(fhp, fhsize); 1353 } 1354 1355 /* 1356 * vfs_composefh: compose a filehandle. 1357 */ 1358 1359 int 1360 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1361 { 1362 struct mount *mp; 1363 struct fid *fidp; 1364 int error; 1365 size_t needfhsize; 1366 size_t fidsize; 1367 1368 mp = vp->v_mount; 1369 fidp = NULL; 1370 if (*fh_size < FHANDLE_SIZE_MIN) { 1371 fidsize = 0; 1372 } else { 1373 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1374 if (fhp != NULL) { 1375 memset(fhp, 0, *fh_size); 1376 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1377 fidp = &fhp->fh_fid; 1378 } 1379 } 1380 error = VFS_VPTOFH(vp, fidp, &fidsize); 1381 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1382 if (error == 0 && *fh_size < needfhsize) { 1383 error = E2BIG; 1384 } 1385 *fh_size = needfhsize; 1386 return error; 1387 } 1388 1389 int 1390 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1391 { 1392 struct mount *mp; 1393 fhandle_t *fhp; 1394 size_t fhsize; 1395 size_t fidsize; 1396 int error; 1397 1398 *fhpp = NULL; 1399 mp = vp->v_mount; 1400 fidsize = 0; 1401 error = VFS_VPTOFH(vp, NULL, &fidsize); 1402 KASSERT(error != 0); 1403 if (error != E2BIG) { 1404 goto out; 1405 } 1406 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1407 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1408 if (fhp == NULL) { 1409 error = ENOMEM; 1410 goto out; 1411 } 1412 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1413 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1414 if (error == 0) { 1415 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1416 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1417 *fhpp = fhp; 1418 } else { 1419 kmem_free(fhp, fhsize); 1420 } 1421 out: 1422 return error; 1423 } 1424 1425 void 1426 vfs_composefh_free(fhandle_t *fhp) 1427 { 1428 1429 vfs__fhfree(fhp); 1430 } 1431 1432 /* 1433 * vfs_fhtovp: lookup a vnode by a filehandle. 1434 */ 1435 1436 int 1437 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1438 { 1439 struct mount *mp; 1440 int error; 1441 1442 *vpp = NULL; 1443 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1444 if (mp == NULL) { 1445 error = ESTALE; 1446 goto out; 1447 } 1448 if (mp->mnt_op->vfs_fhtovp == NULL) { 1449 error = EOPNOTSUPP; 1450 goto out; 1451 } 1452 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1453 out: 1454 return error; 1455 } 1456 1457 /* 1458 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1459 * the needed size. 1460 */ 1461 1462 int 1463 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1464 { 1465 fhandle_t *fhp; 1466 int error; 1467 1468 *fhpp = NULL; 1469 if (fhsize > FHANDLE_SIZE_MAX) { 1470 return EINVAL; 1471 } 1472 if (fhsize < FHANDLE_SIZE_MIN) { 1473 return EINVAL; 1474 } 1475 again: 1476 fhp = kmem_alloc(fhsize, KM_SLEEP); 1477 if (fhp == NULL) { 1478 return ENOMEM; 1479 } 1480 error = copyin(ufhp, fhp, fhsize); 1481 if (error == 0) { 1482 /* XXX this check shouldn't be here */ 1483 if (FHANDLE_SIZE(fhp) == fhsize) { 1484 *fhpp = fhp; 1485 return 0; 1486 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1487 /* 1488 * a kludge for nfsv2 padded handles. 1489 */ 1490 size_t sz; 1491 1492 sz = FHANDLE_SIZE(fhp); 1493 kmem_free(fhp, fhsize); 1494 fhsize = sz; 1495 goto again; 1496 } else { 1497 /* 1498 * userland told us wrong size. 1499 */ 1500 error = EINVAL; 1501 } 1502 } 1503 kmem_free(fhp, fhsize); 1504 return error; 1505 } 1506 1507 void 1508 vfs_copyinfh_free(fhandle_t *fhp) 1509 { 1510 1511 vfs__fhfree(fhp); 1512 } 1513 1514 /* 1515 * Get file handle system call 1516 */ 1517 int 1518 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 1519 { 1520 /* { 1521 syscallarg(char *) fname; 1522 syscallarg(fhandle_t *) fhp; 1523 syscallarg(size_t *) fh_size; 1524 } */ 1525 struct vnode *vp; 1526 fhandle_t *fh; 1527 int error; 1528 struct nameidata nd; 1529 size_t sz; 1530 size_t usz; 1531 1532 /* 1533 * Must be super user 1534 */ 1535 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1536 0, NULL, NULL, NULL); 1537 if (error) 1538 return (error); 1539 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1540 SCARG(uap, fname)); 1541 error = namei(&nd); 1542 if (error) 1543 return (error); 1544 vp = nd.ni_vp; 1545 error = vfs_composefh_alloc(vp, &fh); 1546 vput(vp); 1547 if (error != 0) { 1548 goto out; 1549 } 1550 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1551 if (error != 0) { 1552 goto out; 1553 } 1554 sz = FHANDLE_SIZE(fh); 1555 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1556 if (error != 0) { 1557 goto out; 1558 } 1559 if (usz >= sz) { 1560 error = copyout(fh, SCARG(uap, fhp), sz); 1561 } else { 1562 error = E2BIG; 1563 } 1564 out: 1565 vfs_composefh_free(fh); 1566 return (error); 1567 } 1568 1569 /* 1570 * Open a file given a file handle. 1571 * 1572 * Check permissions, allocate an open file structure, 1573 * and call the device open routine if any. 1574 */ 1575 1576 int 1577 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1578 register_t *retval) 1579 { 1580 struct filedesc *fdp = l->l_proc->p_fd; 1581 struct file *fp; 1582 struct vnode *vp = NULL; 1583 kauth_cred_t cred = l->l_cred; 1584 struct file *nfp; 1585 int type, indx, error=0; 1586 struct flock lf; 1587 struct vattr va; 1588 fhandle_t *fh; 1589 int flags; 1590 1591 /* 1592 * Must be super user 1593 */ 1594 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1595 0, NULL, NULL, NULL))) 1596 return (error); 1597 1598 flags = FFLAGS(oflags); 1599 if ((flags & (FREAD | FWRITE)) == 0) 1600 return (EINVAL); 1601 if ((flags & O_CREAT)) 1602 return (EINVAL); 1603 /* falloc() will use the file descriptor for us */ 1604 if ((error = falloc(l, &nfp, &indx)) != 0) 1605 return (error); 1606 fp = nfp; 1607 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1608 if (error != 0) { 1609 goto bad; 1610 } 1611 error = vfs_fhtovp(fh, &vp); 1612 if (error != 0) { 1613 goto bad; 1614 } 1615 1616 /* Now do an effective vn_open */ 1617 1618 if (vp->v_type == VSOCK) { 1619 error = EOPNOTSUPP; 1620 goto bad; 1621 } 1622 error = vn_openchk(vp, cred, flags); 1623 if (error != 0) 1624 goto bad; 1625 if (flags & O_TRUNC) { 1626 VOP_UNLOCK(vp, 0); /* XXX */ 1627 VOP_LEASE(vp, cred, LEASE_WRITE); 1628 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1629 VATTR_NULL(&va); 1630 va.va_size = 0; 1631 error = VOP_SETATTR(vp, &va, cred); 1632 if (error) 1633 goto bad; 1634 } 1635 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 1636 goto bad; 1637 if (flags & FWRITE) 1638 vp->v_writecount++; 1639 1640 /* done with modified vn_open, now finish what sys_open does. */ 1641 1642 fp->f_flag = flags & FMASK; 1643 fp->f_type = DTYPE_VNODE; 1644 fp->f_ops = &vnops; 1645 fp->f_data = vp; 1646 if (flags & (O_EXLOCK | O_SHLOCK)) { 1647 lf.l_whence = SEEK_SET; 1648 lf.l_start = 0; 1649 lf.l_len = 0; 1650 if (flags & O_EXLOCK) 1651 lf.l_type = F_WRLCK; 1652 else 1653 lf.l_type = F_RDLCK; 1654 type = F_FLOCK; 1655 if ((flags & FNONBLOCK) == 0) 1656 type |= F_WAIT; 1657 VOP_UNLOCK(vp, 0); 1658 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1659 if (error) { 1660 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1661 FILE_UNUSE(fp, l); 1662 ffree(fp); 1663 fdremove(fdp, indx); 1664 return (error); 1665 } 1666 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1667 fp->f_flag |= FHASLOCK; 1668 } 1669 VOP_UNLOCK(vp, 0); 1670 *retval = indx; 1671 FILE_SET_MATURE(fp); 1672 FILE_UNUSE(fp, l); 1673 vfs_copyinfh_free(fh); 1674 return (0); 1675 1676 bad: 1677 FILE_UNUSE(fp, l); 1678 ffree(fp); 1679 fdremove(fdp, indx); 1680 if (vp != NULL) 1681 vput(vp); 1682 vfs_copyinfh_free(fh); 1683 return (error); 1684 } 1685 1686 int 1687 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 1688 { 1689 /* { 1690 syscallarg(const void *) fhp; 1691 syscallarg(size_t) fh_size; 1692 syscallarg(int) flags; 1693 } */ 1694 1695 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1696 SCARG(uap, flags), retval); 1697 } 1698 1699 int 1700 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1701 { 1702 int error; 1703 fhandle_t *fh; 1704 struct vnode *vp; 1705 1706 /* 1707 * Must be super user 1708 */ 1709 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1710 0, NULL, NULL, NULL))) 1711 return (error); 1712 1713 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1714 if (error != 0) 1715 return error; 1716 1717 error = vfs_fhtovp(fh, &vp); 1718 vfs_copyinfh_free(fh); 1719 if (error != 0) 1720 return error; 1721 1722 error = vn_stat(vp, sb, l); 1723 vput(vp); 1724 return error; 1725 } 1726 1727 1728 /* ARGSUSED */ 1729 int 1730 sys___fhstat40(struct lwp *l, const struct sys___fhstat40_args *uap, register_t *retval) 1731 { 1732 /* { 1733 syscallarg(const void *) fhp; 1734 syscallarg(size_t) fh_size; 1735 syscallarg(struct stat *) sb; 1736 } */ 1737 struct stat sb; 1738 int error; 1739 1740 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1741 if (error) 1742 return error; 1743 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1744 } 1745 1746 int 1747 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1748 int flags) 1749 { 1750 fhandle_t *fh; 1751 struct mount *mp; 1752 struct vnode *vp; 1753 int error; 1754 1755 /* 1756 * Must be super user 1757 */ 1758 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1759 0, NULL, NULL, NULL))) 1760 return error; 1761 1762 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1763 if (error != 0) 1764 return error; 1765 1766 error = vfs_fhtovp(fh, &vp); 1767 vfs_copyinfh_free(fh); 1768 if (error != 0) 1769 return error; 1770 1771 mp = vp->v_mount; 1772 error = dostatvfs(mp, sb, l, flags, 1); 1773 vput(vp); 1774 return error; 1775 } 1776 1777 /* ARGSUSED */ 1778 int 1779 sys___fhstatvfs140(struct lwp *l, const struct sys___fhstatvfs140_args *uap, register_t *retval) 1780 { 1781 /* { 1782 syscallarg(const void *) fhp; 1783 syscallarg(size_t) fh_size; 1784 syscallarg(struct statvfs *) buf; 1785 syscallarg(int) flags; 1786 } */ 1787 struct statvfs *sb = STATVFSBUF_GET(); 1788 int error; 1789 1790 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1791 SCARG(uap, flags)); 1792 if (error == 0) 1793 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1794 STATVFSBUF_PUT(sb); 1795 return error; 1796 } 1797 1798 /* 1799 * Create a special file. 1800 */ 1801 /* ARGSUSED */ 1802 int 1803 sys_mknod(struct lwp *l, const struct sys_mknod_args *uap, register_t *retval) 1804 { 1805 /* { 1806 syscallarg(const char *) path; 1807 syscallarg(int) mode; 1808 syscallarg(int) dev; 1809 } */ 1810 struct proc *p = l->l_proc; 1811 struct vnode *vp; 1812 struct vattr vattr; 1813 int error, optype; 1814 struct nameidata nd; 1815 char *path; 1816 const char *cpath; 1817 enum uio_seg seg = UIO_USERSPACE; 1818 1819 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1820 0, NULL, NULL, NULL)) != 0) 1821 return (error); 1822 1823 optype = VOP_MKNOD_DESCOFFSET; 1824 1825 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1826 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath); 1827 1828 if ((error = namei(&nd)) != 0) 1829 goto out; 1830 vp = nd.ni_vp; 1831 if (vp != NULL) 1832 error = EEXIST; 1833 else { 1834 VATTR_NULL(&vattr); 1835 /* We will read cwdi->cwdi_cmask unlocked. */ 1836 vattr.va_mode = 1837 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1838 vattr.va_rdev = SCARG(uap, dev); 1839 1840 switch (SCARG(uap, mode) & S_IFMT) { 1841 case S_IFMT: /* used by badsect to flag bad sectors */ 1842 vattr.va_type = VBAD; 1843 break; 1844 case S_IFCHR: 1845 vattr.va_type = VCHR; 1846 break; 1847 case S_IFBLK: 1848 vattr.va_type = VBLK; 1849 break; 1850 case S_IFWHT: 1851 optype = VOP_WHITEOUT_DESCOFFSET; 1852 break; 1853 case S_IFREG: 1854 #if NVERIEXEC > 0 1855 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1856 O_CREAT); 1857 #endif /* NVERIEXEC > 0 */ 1858 vattr.va_type = VREG; 1859 vattr.va_rdev = VNOVAL; 1860 optype = VOP_CREATE_DESCOFFSET; 1861 break; 1862 default: 1863 error = EINVAL; 1864 break; 1865 } 1866 } 1867 if (!error) { 1868 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1869 switch (optype) { 1870 case VOP_WHITEOUT_DESCOFFSET: 1871 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1872 if (error) 1873 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1874 vput(nd.ni_dvp); 1875 break; 1876 1877 case VOP_MKNOD_DESCOFFSET: 1878 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1879 &nd.ni_cnd, &vattr); 1880 if (error == 0) 1881 vput(nd.ni_vp); 1882 break; 1883 1884 case VOP_CREATE_DESCOFFSET: 1885 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1886 &nd.ni_cnd, &vattr); 1887 if (error == 0) 1888 vput(nd.ni_vp); 1889 break; 1890 } 1891 } else { 1892 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1893 if (nd.ni_dvp == vp) 1894 vrele(nd.ni_dvp); 1895 else 1896 vput(nd.ni_dvp); 1897 if (vp) 1898 vrele(vp); 1899 } 1900 out: 1901 VERIEXEC_PATH_PUT(path); 1902 return (error); 1903 } 1904 1905 /* 1906 * Create a named pipe. 1907 */ 1908 /* ARGSUSED */ 1909 int 1910 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 1911 { 1912 /* { 1913 syscallarg(const char *) path; 1914 syscallarg(int) mode; 1915 } */ 1916 struct proc *p = l->l_proc; 1917 struct vattr vattr; 1918 int error; 1919 struct nameidata nd; 1920 1921 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1922 SCARG(uap, path)); 1923 if ((error = namei(&nd)) != 0) 1924 return (error); 1925 if (nd.ni_vp != NULL) { 1926 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1927 if (nd.ni_dvp == nd.ni_vp) 1928 vrele(nd.ni_dvp); 1929 else 1930 vput(nd.ni_dvp); 1931 vrele(nd.ni_vp); 1932 return (EEXIST); 1933 } 1934 VATTR_NULL(&vattr); 1935 vattr.va_type = VFIFO; 1936 /* We will read cwdi->cwdi_cmask unlocked. */ 1937 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1938 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1939 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1940 if (error == 0) 1941 vput(nd.ni_vp); 1942 return (error); 1943 } 1944 1945 /* 1946 * Make a hard file link. 1947 */ 1948 /* ARGSUSED */ 1949 int 1950 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 1951 { 1952 /* { 1953 syscallarg(const char *) path; 1954 syscallarg(const char *) link; 1955 } */ 1956 struct vnode *vp; 1957 struct nameidata nd; 1958 int error; 1959 1960 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 1961 SCARG(uap, path)); 1962 if ((error = namei(&nd)) != 0) 1963 return (error); 1964 vp = nd.ni_vp; 1965 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 1966 SCARG(uap, link)); 1967 if ((error = namei(&nd)) != 0) 1968 goto out; 1969 if (nd.ni_vp) { 1970 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1971 if (nd.ni_dvp == nd.ni_vp) 1972 vrele(nd.ni_dvp); 1973 else 1974 vput(nd.ni_dvp); 1975 vrele(nd.ni_vp); 1976 error = EEXIST; 1977 goto out; 1978 } 1979 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 1980 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 1981 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1982 out: 1983 vrele(vp); 1984 return (error); 1985 } 1986 1987 /* 1988 * Make a symbolic link. 1989 */ 1990 /* ARGSUSED */ 1991 int 1992 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 1993 { 1994 /* { 1995 syscallarg(const char *) path; 1996 syscallarg(const char *) link; 1997 } */ 1998 struct proc *p = l->l_proc; 1999 struct vattr vattr; 2000 char *path; 2001 int error; 2002 struct nameidata nd; 2003 2004 path = PNBUF_GET(); 2005 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2006 if (error) 2007 goto out; 2008 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, 2009 SCARG(uap, link)); 2010 if ((error = namei(&nd)) != 0) 2011 goto out; 2012 if (nd.ni_vp) { 2013 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2014 if (nd.ni_dvp == nd.ni_vp) 2015 vrele(nd.ni_dvp); 2016 else 2017 vput(nd.ni_dvp); 2018 vrele(nd.ni_vp); 2019 error = EEXIST; 2020 goto out; 2021 } 2022 VATTR_NULL(&vattr); 2023 vattr.va_type = VLNK; 2024 /* We will read cwdi->cwdi_cmask unlocked. */ 2025 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2026 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2027 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2028 if (error == 0) 2029 vput(nd.ni_vp); 2030 out: 2031 PNBUF_PUT(path); 2032 return (error); 2033 } 2034 2035 /* 2036 * Delete a whiteout from the filesystem. 2037 */ 2038 /* ARGSUSED */ 2039 int 2040 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2041 { 2042 /* { 2043 syscallarg(const char *) path; 2044 } */ 2045 int error; 2046 struct nameidata nd; 2047 2048 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, 2049 UIO_USERSPACE, SCARG(uap, path)); 2050 error = namei(&nd); 2051 if (error) 2052 return (error); 2053 2054 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2055 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2056 if (nd.ni_dvp == nd.ni_vp) 2057 vrele(nd.ni_dvp); 2058 else 2059 vput(nd.ni_dvp); 2060 if (nd.ni_vp) 2061 vrele(nd.ni_vp); 2062 return (EEXIST); 2063 } 2064 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2065 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2066 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2067 vput(nd.ni_dvp); 2068 return (error); 2069 } 2070 2071 /* 2072 * Delete a name from the filesystem. 2073 */ 2074 /* ARGSUSED */ 2075 int 2076 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2077 { 2078 /* { 2079 syscallarg(const char *) path; 2080 } */ 2081 struct vnode *vp; 2082 int error; 2083 struct nameidata nd; 2084 char *path; 2085 const char *cpath; 2086 enum uio_seg seg = UIO_USERSPACE; 2087 2088 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2089 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath); 2090 2091 if ((error = namei(&nd)) != 0) 2092 goto out; 2093 vp = nd.ni_vp; 2094 2095 /* 2096 * The root of a mounted filesystem cannot be deleted. 2097 */ 2098 if (vp->v_vflag & VV_ROOT) { 2099 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2100 if (nd.ni_dvp == vp) 2101 vrele(nd.ni_dvp); 2102 else 2103 vput(nd.ni_dvp); 2104 vput(vp); 2105 error = EBUSY; 2106 goto out; 2107 } 2108 2109 #if NVERIEXEC > 0 2110 /* Handle remove requests for veriexec entries. */ 2111 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2112 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2113 if (nd.ni_dvp == vp) 2114 vrele(nd.ni_dvp); 2115 else 2116 vput(nd.ni_dvp); 2117 vput(vp); 2118 goto out; 2119 } 2120 #endif /* NVERIEXEC > 0 */ 2121 2122 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 2123 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2124 #ifdef FILEASSOC 2125 (void)fileassoc_file_delete(vp); 2126 #endif /* FILEASSOC */ 2127 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2128 out: 2129 VERIEXEC_PATH_PUT(path); 2130 return (error); 2131 } 2132 2133 /* 2134 * Reposition read/write file offset. 2135 */ 2136 int 2137 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2138 { 2139 /* { 2140 syscallarg(int) fd; 2141 syscallarg(int) pad; 2142 syscallarg(off_t) offset; 2143 syscallarg(int) whence; 2144 } */ 2145 struct proc *p = l->l_proc; 2146 kauth_cred_t cred = l->l_cred; 2147 struct filedesc *fdp = p->p_fd; 2148 struct file *fp; 2149 struct vnode *vp; 2150 struct vattr vattr; 2151 off_t newoff; 2152 int error; 2153 2154 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2155 return (EBADF); 2156 2157 vp = (struct vnode *)fp->f_data; 2158 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2159 error = ESPIPE; 2160 mutex_exit(&fp->f_lock); 2161 goto out; 2162 } 2163 2164 switch (SCARG(uap, whence)) { 2165 case SEEK_CUR: 2166 newoff = fp->f_offset + SCARG(uap, offset); 2167 FILE_USE(fp); 2168 break; 2169 case SEEK_END: 2170 FILE_USE(fp); 2171 error = VOP_GETATTR(vp, &vattr, cred); 2172 if (error) { 2173 FILE_UNUSE(fp, l); 2174 goto out; 2175 } 2176 newoff = SCARG(uap, offset) + vattr.va_size; 2177 break; 2178 case SEEK_SET: 2179 FILE_USE(fp); 2180 newoff = SCARG(uap, offset); 2181 break; 2182 default: 2183 mutex_exit(&fp->f_lock); 2184 error = EINVAL; 2185 goto out; 2186 } 2187 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2188 mutex_enter(&fp->f_lock); 2189 *(off_t *)retval = fp->f_offset = newoff; 2190 mutex_exit(&fp->f_lock); 2191 } 2192 FILE_UNUSE(fp, l); 2193 out: 2194 return (error); 2195 } 2196 2197 /* 2198 * Positional read system call. 2199 */ 2200 int 2201 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2202 { 2203 /* { 2204 syscallarg(int) fd; 2205 syscallarg(void *) buf; 2206 syscallarg(size_t) nbyte; 2207 syscallarg(off_t) offset; 2208 } */ 2209 struct proc *p = l->l_proc; 2210 struct filedesc *fdp = p->p_fd; 2211 struct file *fp; 2212 struct vnode *vp; 2213 off_t offset; 2214 int error, fd = SCARG(uap, fd); 2215 2216 if ((fp = fd_getfile(fdp, fd)) == NULL) 2217 return (EBADF); 2218 2219 if ((fp->f_flag & FREAD) == 0) { 2220 mutex_exit(&fp->f_lock); 2221 return (EBADF); 2222 } 2223 2224 FILE_USE(fp); 2225 2226 vp = (struct vnode *)fp->f_data; 2227 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2228 error = ESPIPE; 2229 goto out; 2230 } 2231 2232 offset = SCARG(uap, offset); 2233 2234 /* 2235 * XXX This works because no file systems actually 2236 * XXX take any action on the seek operation. 2237 */ 2238 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2239 goto out; 2240 2241 /* dofileread() will unuse the descriptor for us */ 2242 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2243 &offset, 0, retval)); 2244 2245 out: 2246 FILE_UNUSE(fp, l); 2247 return (error); 2248 } 2249 2250 /* 2251 * Positional scatter read system call. 2252 */ 2253 int 2254 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 2255 { 2256 /* { 2257 syscallarg(int) fd; 2258 syscallarg(const struct iovec *) iovp; 2259 syscallarg(int) iovcnt; 2260 syscallarg(off_t) offset; 2261 } */ 2262 off_t offset = SCARG(uap, offset); 2263 2264 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 2265 SCARG(uap, iovcnt), &offset, 0, retval); 2266 } 2267 2268 /* 2269 * Positional write system call. 2270 */ 2271 int 2272 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 2273 { 2274 /* { 2275 syscallarg(int) fd; 2276 syscallarg(const void *) buf; 2277 syscallarg(size_t) nbyte; 2278 syscallarg(off_t) offset; 2279 } */ 2280 struct proc *p = l->l_proc; 2281 struct filedesc *fdp = p->p_fd; 2282 struct file *fp; 2283 struct vnode *vp; 2284 off_t offset; 2285 int error, fd = SCARG(uap, fd); 2286 2287 if ((fp = fd_getfile(fdp, fd)) == NULL) 2288 return (EBADF); 2289 2290 if ((fp->f_flag & FWRITE) == 0) { 2291 mutex_exit(&fp->f_lock); 2292 return (EBADF); 2293 } 2294 2295 FILE_USE(fp); 2296 2297 vp = (struct vnode *)fp->f_data; 2298 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2299 error = ESPIPE; 2300 goto out; 2301 } 2302 2303 offset = SCARG(uap, offset); 2304 2305 /* 2306 * XXX This works because no file systems actually 2307 * XXX take any action on the seek operation. 2308 */ 2309 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2310 goto out; 2311 2312 /* dofilewrite() will unuse the descriptor for us */ 2313 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2314 &offset, 0, retval)); 2315 2316 out: 2317 FILE_UNUSE(fp, l); 2318 return (error); 2319 } 2320 2321 /* 2322 * Positional gather write system call. 2323 */ 2324 int 2325 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 2326 { 2327 /* { 2328 syscallarg(int) fd; 2329 syscallarg(const struct iovec *) iovp; 2330 syscallarg(int) iovcnt; 2331 syscallarg(off_t) offset; 2332 } */ 2333 off_t offset = SCARG(uap, offset); 2334 2335 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 2336 SCARG(uap, iovcnt), &offset, 0, retval); 2337 } 2338 2339 /* 2340 * Check access permissions. 2341 */ 2342 int 2343 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 2344 { 2345 /* { 2346 syscallarg(const char *) path; 2347 syscallarg(int) flags; 2348 } */ 2349 kauth_cred_t cred; 2350 struct vnode *vp; 2351 int error, flags; 2352 struct nameidata nd; 2353 2354 cred = kauth_cred_dup(l->l_cred); 2355 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2356 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2357 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2358 SCARG(uap, path)); 2359 /* Override default credentials */ 2360 nd.ni_cnd.cn_cred = cred; 2361 if ((error = namei(&nd)) != 0) 2362 goto out; 2363 vp = nd.ni_vp; 2364 2365 /* Flags == 0 means only check for existence. */ 2366 if (SCARG(uap, flags)) { 2367 flags = 0; 2368 if (SCARG(uap, flags) & R_OK) 2369 flags |= VREAD; 2370 if (SCARG(uap, flags) & W_OK) 2371 flags |= VWRITE; 2372 if (SCARG(uap, flags) & X_OK) 2373 flags |= VEXEC; 2374 2375 error = VOP_ACCESS(vp, flags, cred); 2376 if (!error && (flags & VWRITE)) 2377 error = vn_writechk(vp); 2378 } 2379 vput(vp); 2380 out: 2381 kauth_cred_free(cred); 2382 return (error); 2383 } 2384 2385 /* 2386 * Common code for all sys_stat functions, including compat versions. 2387 */ 2388 int 2389 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2390 struct stat *sb) 2391 { 2392 int error; 2393 struct nameidata nd; 2394 2395 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT, 2396 UIO_USERSPACE, path); 2397 error = namei(&nd); 2398 if (error != 0) 2399 return error; 2400 error = vn_stat(nd.ni_vp, sb, l); 2401 vput(nd.ni_vp); 2402 return error; 2403 } 2404 2405 /* 2406 * Get file status; this version follows links. 2407 */ 2408 /* ARGSUSED */ 2409 int 2410 sys___stat30(struct lwp *l, const struct sys___stat30_args *uap, register_t *retval) 2411 { 2412 /* { 2413 syscallarg(const char *) path; 2414 syscallarg(struct stat *) ub; 2415 } */ 2416 struct stat sb; 2417 int error; 2418 2419 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2420 if (error) 2421 return error; 2422 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2423 } 2424 2425 /* 2426 * Get file status; this version does not follow links. 2427 */ 2428 /* ARGSUSED */ 2429 int 2430 sys___lstat30(struct lwp *l, const struct sys___lstat30_args *uap, register_t *retval) 2431 { 2432 /* { 2433 syscallarg(const char *) path; 2434 syscallarg(struct stat *) ub; 2435 } */ 2436 struct stat sb; 2437 int error; 2438 2439 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2440 if (error) 2441 return error; 2442 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2443 } 2444 2445 /* 2446 * Get configurable pathname variables. 2447 */ 2448 /* ARGSUSED */ 2449 int 2450 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, register_t *retval) 2451 { 2452 /* { 2453 syscallarg(const char *) path; 2454 syscallarg(int) name; 2455 } */ 2456 int error; 2457 struct nameidata nd; 2458 2459 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2460 SCARG(uap, path)); 2461 if ((error = namei(&nd)) != 0) 2462 return (error); 2463 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2464 vput(nd.ni_vp); 2465 return (error); 2466 } 2467 2468 /* 2469 * Return target name of a symbolic link. 2470 */ 2471 /* ARGSUSED */ 2472 int 2473 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, register_t *retval) 2474 { 2475 /* { 2476 syscallarg(const char *) path; 2477 syscallarg(char *) buf; 2478 syscallarg(size_t) count; 2479 } */ 2480 struct vnode *vp; 2481 struct iovec aiov; 2482 struct uio auio; 2483 int error; 2484 struct nameidata nd; 2485 2486 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2487 SCARG(uap, path)); 2488 if ((error = namei(&nd)) != 0) 2489 return (error); 2490 vp = nd.ni_vp; 2491 if (vp->v_type != VLNK) 2492 error = EINVAL; 2493 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2494 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 2495 aiov.iov_base = SCARG(uap, buf); 2496 aiov.iov_len = SCARG(uap, count); 2497 auio.uio_iov = &aiov; 2498 auio.uio_iovcnt = 1; 2499 auio.uio_offset = 0; 2500 auio.uio_rw = UIO_READ; 2501 KASSERT(l == curlwp); 2502 auio.uio_vmspace = l->l_proc->p_vmspace; 2503 auio.uio_resid = SCARG(uap, count); 2504 error = VOP_READLINK(vp, &auio, l->l_cred); 2505 } 2506 vput(vp); 2507 *retval = SCARG(uap, count) - auio.uio_resid; 2508 return (error); 2509 } 2510 2511 /* 2512 * Change flags of a file given a path name. 2513 */ 2514 /* ARGSUSED */ 2515 int 2516 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 2517 { 2518 /* { 2519 syscallarg(const char *) path; 2520 syscallarg(u_long) flags; 2521 } */ 2522 struct vnode *vp; 2523 int error; 2524 struct nameidata nd; 2525 2526 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2527 SCARG(uap, path)); 2528 if ((error = namei(&nd)) != 0) 2529 return (error); 2530 vp = nd.ni_vp; 2531 error = change_flags(vp, SCARG(uap, flags), l); 2532 vput(vp); 2533 return (error); 2534 } 2535 2536 /* 2537 * Change flags of a file given a file descriptor. 2538 */ 2539 /* ARGSUSED */ 2540 int 2541 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 2542 { 2543 /* { 2544 syscallarg(int) fd; 2545 syscallarg(u_long) flags; 2546 } */ 2547 struct proc *p = l->l_proc; 2548 struct vnode *vp; 2549 struct file *fp; 2550 int error; 2551 2552 /* getvnode() will use the descriptor for us */ 2553 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2554 return (error); 2555 vp = (struct vnode *)fp->f_data; 2556 error = change_flags(vp, SCARG(uap, flags), l); 2557 VOP_UNLOCK(vp, 0); 2558 FILE_UNUSE(fp, l); 2559 return (error); 2560 } 2561 2562 /* 2563 * Change flags of a file given a path name; this version does 2564 * not follow links. 2565 */ 2566 int 2567 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 2568 { 2569 /* { 2570 syscallarg(const char *) path; 2571 syscallarg(u_long) flags; 2572 } */ 2573 struct vnode *vp; 2574 int error; 2575 struct nameidata nd; 2576 2577 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2578 SCARG(uap, path)); 2579 if ((error = namei(&nd)) != 0) 2580 return (error); 2581 vp = nd.ni_vp; 2582 error = change_flags(vp, SCARG(uap, flags), l); 2583 vput(vp); 2584 return (error); 2585 } 2586 2587 /* 2588 * Common routine to change flags of a file. 2589 */ 2590 int 2591 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2592 { 2593 struct vattr vattr; 2594 int error; 2595 2596 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2597 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2598 /* 2599 * Non-superusers cannot change the flags on devices, even if they 2600 * own them. 2601 */ 2602 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2603 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2604 goto out; 2605 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2606 error = EINVAL; 2607 goto out; 2608 } 2609 } 2610 VATTR_NULL(&vattr); 2611 vattr.va_flags = flags; 2612 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2613 out: 2614 return (error); 2615 } 2616 2617 /* 2618 * Change mode of a file given path name; this version follows links. 2619 */ 2620 /* ARGSUSED */ 2621 int 2622 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 2623 { 2624 /* { 2625 syscallarg(const char *) path; 2626 syscallarg(int) mode; 2627 } */ 2628 int error; 2629 struct nameidata nd; 2630 2631 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2632 SCARG(uap, path)); 2633 if ((error = namei(&nd)) != 0) 2634 return (error); 2635 2636 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2637 2638 vrele(nd.ni_vp); 2639 return (error); 2640 } 2641 2642 /* 2643 * Change mode of a file given a file descriptor. 2644 */ 2645 /* ARGSUSED */ 2646 int 2647 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 2648 { 2649 /* { 2650 syscallarg(int) fd; 2651 syscallarg(int) mode; 2652 } */ 2653 struct proc *p = l->l_proc; 2654 struct file *fp; 2655 int error; 2656 2657 /* getvnode() will use the descriptor for us */ 2658 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2659 return (error); 2660 2661 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2662 FILE_UNUSE(fp, l); 2663 return (error); 2664 } 2665 2666 /* 2667 * Change mode of a file given path name; this version does not follow links. 2668 */ 2669 /* ARGSUSED */ 2670 int 2671 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 2672 { 2673 /* { 2674 syscallarg(const char *) path; 2675 syscallarg(int) mode; 2676 } */ 2677 int error; 2678 struct nameidata nd; 2679 2680 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2681 SCARG(uap, path)); 2682 if ((error = namei(&nd)) != 0) 2683 return (error); 2684 2685 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2686 2687 vrele(nd.ni_vp); 2688 return (error); 2689 } 2690 2691 /* 2692 * Common routine to set mode given a vnode. 2693 */ 2694 static int 2695 change_mode(struct vnode *vp, int mode, struct lwp *l) 2696 { 2697 struct vattr vattr; 2698 int error; 2699 2700 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2701 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2702 VATTR_NULL(&vattr); 2703 vattr.va_mode = mode & ALLPERMS; 2704 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2705 VOP_UNLOCK(vp, 0); 2706 return (error); 2707 } 2708 2709 /* 2710 * Set ownership given a path name; this version follows links. 2711 */ 2712 /* ARGSUSED */ 2713 int 2714 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 2715 { 2716 /* { 2717 syscallarg(const char *) path; 2718 syscallarg(uid_t) uid; 2719 syscallarg(gid_t) gid; 2720 } */ 2721 int error; 2722 struct nameidata nd; 2723 2724 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2725 SCARG(uap, path)); 2726 if ((error = namei(&nd)) != 0) 2727 return (error); 2728 2729 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2730 2731 vrele(nd.ni_vp); 2732 return (error); 2733 } 2734 2735 /* 2736 * Set ownership given a path name; this version follows links. 2737 * Provides POSIX semantics. 2738 */ 2739 /* ARGSUSED */ 2740 int 2741 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 2742 { 2743 /* { 2744 syscallarg(const char *) path; 2745 syscallarg(uid_t) uid; 2746 syscallarg(gid_t) gid; 2747 } */ 2748 int error; 2749 struct nameidata nd; 2750 2751 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 2752 SCARG(uap, path)); 2753 if ((error = namei(&nd)) != 0) 2754 return (error); 2755 2756 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2757 2758 vrele(nd.ni_vp); 2759 return (error); 2760 } 2761 2762 /* 2763 * Set ownership given a file descriptor. 2764 */ 2765 /* ARGSUSED */ 2766 int 2767 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 2768 { 2769 /* { 2770 syscallarg(int) fd; 2771 syscallarg(uid_t) uid; 2772 syscallarg(gid_t) gid; 2773 } */ 2774 struct proc *p = l->l_proc; 2775 int error; 2776 struct file *fp; 2777 2778 /* getvnode() will use the descriptor for us */ 2779 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2780 return (error); 2781 2782 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2783 SCARG(uap, gid), l, 0); 2784 FILE_UNUSE(fp, l); 2785 return (error); 2786 } 2787 2788 /* 2789 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2790 */ 2791 /* ARGSUSED */ 2792 int 2793 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 2794 { 2795 /* { 2796 syscallarg(int) fd; 2797 syscallarg(uid_t) uid; 2798 syscallarg(gid_t) gid; 2799 } */ 2800 struct proc *p = l->l_proc; 2801 int error; 2802 struct file *fp; 2803 2804 /* getvnode() will use the descriptor for us */ 2805 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2806 return (error); 2807 2808 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2809 SCARG(uap, gid), l, 1); 2810 FILE_UNUSE(fp, l); 2811 return (error); 2812 } 2813 2814 /* 2815 * Set ownership given a path name; this version does not follow links. 2816 */ 2817 /* ARGSUSED */ 2818 int 2819 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 2820 { 2821 /* { 2822 syscallarg(const char *) path; 2823 syscallarg(uid_t) uid; 2824 syscallarg(gid_t) gid; 2825 } */ 2826 int error; 2827 struct nameidata nd; 2828 2829 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2830 SCARG(uap, path)); 2831 if ((error = namei(&nd)) != 0) 2832 return (error); 2833 2834 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2835 2836 vrele(nd.ni_vp); 2837 return (error); 2838 } 2839 2840 /* 2841 * Set ownership given a path name; this version does not follow links. 2842 * Provides POSIX/XPG semantics. 2843 */ 2844 /* ARGSUSED */ 2845 int 2846 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 2847 { 2848 /* { 2849 syscallarg(const char *) path; 2850 syscallarg(uid_t) uid; 2851 syscallarg(gid_t) gid; 2852 } */ 2853 int error; 2854 struct nameidata nd; 2855 2856 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 2857 SCARG(uap, path)); 2858 if ((error = namei(&nd)) != 0) 2859 return (error); 2860 2861 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2862 2863 vrele(nd.ni_vp); 2864 return (error); 2865 } 2866 2867 /* 2868 * Common routine to set ownership given a vnode. 2869 */ 2870 static int 2871 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2872 int posix_semantics) 2873 { 2874 struct vattr vattr; 2875 mode_t newmode; 2876 int error; 2877 2878 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 2879 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2880 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 2881 goto out; 2882 2883 #define CHANGED(x) ((int)(x) != -1) 2884 newmode = vattr.va_mode; 2885 if (posix_semantics) { 2886 /* 2887 * POSIX/XPG semantics: if the caller is not the super-user, 2888 * clear set-user-id and set-group-id bits. Both POSIX and 2889 * the XPG consider the behaviour for calls by the super-user 2890 * implementation-defined; we leave the set-user-id and set- 2891 * group-id settings intact in that case. 2892 */ 2893 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2894 NULL) != 0) 2895 newmode &= ~(S_ISUID | S_ISGID); 2896 } else { 2897 /* 2898 * NetBSD semantics: when changing owner and/or group, 2899 * clear the respective bit(s). 2900 */ 2901 if (CHANGED(uid)) 2902 newmode &= ~S_ISUID; 2903 if (CHANGED(gid)) 2904 newmode &= ~S_ISGID; 2905 } 2906 /* Update va_mode iff altered. */ 2907 if (vattr.va_mode == newmode) 2908 newmode = VNOVAL; 2909 2910 VATTR_NULL(&vattr); 2911 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2912 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2913 vattr.va_mode = newmode; 2914 error = VOP_SETATTR(vp, &vattr, l->l_cred); 2915 #undef CHANGED 2916 2917 out: 2918 VOP_UNLOCK(vp, 0); 2919 return (error); 2920 } 2921 2922 /* 2923 * Set the access and modification times given a path name; this 2924 * version follows links. 2925 */ 2926 /* ARGSUSED */ 2927 int 2928 sys_utimes(struct lwp *l, const struct sys_utimes_args *uap, register_t *retval) 2929 { 2930 /* { 2931 syscallarg(const char *) path; 2932 syscallarg(const struct timeval *) tptr; 2933 } */ 2934 2935 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2936 SCARG(uap, tptr), UIO_USERSPACE); 2937 } 2938 2939 /* 2940 * Set the access and modification times given a file descriptor. 2941 */ 2942 /* ARGSUSED */ 2943 int 2944 sys_futimes(struct lwp *l, const struct sys_futimes_args *uap, register_t *retval) 2945 { 2946 /* { 2947 syscallarg(int) fd; 2948 syscallarg(const struct timeval *) tptr; 2949 } */ 2950 int error; 2951 struct file *fp; 2952 2953 /* getvnode() will use the descriptor for us */ 2954 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2955 return (error); 2956 2957 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2958 SCARG(uap, tptr), UIO_USERSPACE); 2959 2960 FILE_UNUSE(fp, l); 2961 return (error); 2962 } 2963 2964 /* 2965 * Set the access and modification times given a path name; this 2966 * version does not follow links. 2967 */ 2968 int 2969 sys_lutimes(struct lwp *l, const struct sys_lutimes_args *uap, register_t *retval) 2970 { 2971 /* { 2972 syscallarg(const char *) path; 2973 syscallarg(const struct timeval *) tptr; 2974 } */ 2975 2976 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2977 SCARG(uap, tptr), UIO_USERSPACE); 2978 } 2979 2980 /* 2981 * Common routine to set access and modification times given a vnode. 2982 */ 2983 int 2984 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2985 const struct timeval *tptr, enum uio_seg seg) 2986 { 2987 struct vattr vattr; 2988 struct nameidata nd; 2989 int error; 2990 2991 VATTR_NULL(&vattr); 2992 if (tptr == NULL) { 2993 nanotime(&vattr.va_atime); 2994 vattr.va_mtime = vattr.va_atime; 2995 vattr.va_vaflags |= VA_UTIMES_NULL; 2996 } else { 2997 struct timeval tv[2]; 2998 2999 if (seg != UIO_SYSSPACE) { 3000 error = copyin(tptr, &tv, sizeof (tv)); 3001 if (error != 0) 3002 return error; 3003 tptr = tv; 3004 } 3005 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3006 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3007 } 3008 3009 if (vp == NULL) { 3010 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path); 3011 if ((error = namei(&nd)) != 0) 3012 return (error); 3013 vp = nd.ni_vp; 3014 } else 3015 nd.ni_vp = NULL; 3016 3017 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3018 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3019 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3020 VOP_UNLOCK(vp, 0); 3021 3022 if (nd.ni_vp != NULL) 3023 vrele(nd.ni_vp); 3024 3025 return (error); 3026 } 3027 3028 /* 3029 * Truncate a file given its path name. 3030 */ 3031 /* ARGSUSED */ 3032 int 3033 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 3034 { 3035 /* { 3036 syscallarg(const char *) path; 3037 syscallarg(int) pad; 3038 syscallarg(off_t) length; 3039 } */ 3040 struct vnode *vp; 3041 struct vattr vattr; 3042 int error; 3043 struct nameidata nd; 3044 3045 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3046 SCARG(uap, path)); 3047 if ((error = namei(&nd)) != 0) 3048 return (error); 3049 vp = nd.ni_vp; 3050 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3051 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3052 if (vp->v_type == VDIR) 3053 error = EISDIR; 3054 else if ((error = vn_writechk(vp)) == 0 && 3055 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 3056 VATTR_NULL(&vattr); 3057 vattr.va_size = SCARG(uap, length); 3058 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3059 } 3060 vput(vp); 3061 return (error); 3062 } 3063 3064 /* 3065 * Truncate a file given a file descriptor. 3066 */ 3067 /* ARGSUSED */ 3068 int 3069 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 3070 { 3071 /* { 3072 syscallarg(int) fd; 3073 syscallarg(int) pad; 3074 syscallarg(off_t) length; 3075 } */ 3076 struct proc *p = l->l_proc; 3077 struct vattr vattr; 3078 struct vnode *vp; 3079 struct file *fp; 3080 int error; 3081 3082 /* getvnode() will use the descriptor for us */ 3083 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3084 return (error); 3085 if ((fp->f_flag & FWRITE) == 0) { 3086 error = EINVAL; 3087 goto out; 3088 } 3089 vp = (struct vnode *)fp->f_data; 3090 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3091 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3092 if (vp->v_type == VDIR) 3093 error = EISDIR; 3094 else if ((error = vn_writechk(vp)) == 0) { 3095 VATTR_NULL(&vattr); 3096 vattr.va_size = SCARG(uap, length); 3097 error = VOP_SETATTR(vp, &vattr, fp->f_cred); 3098 } 3099 VOP_UNLOCK(vp, 0); 3100 out: 3101 FILE_UNUSE(fp, l); 3102 return (error); 3103 } 3104 3105 /* 3106 * Sync an open file. 3107 */ 3108 /* ARGSUSED */ 3109 int 3110 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 3111 { 3112 /* { 3113 syscallarg(int) fd; 3114 } */ 3115 struct proc *p = l->l_proc; 3116 struct vnode *vp; 3117 struct file *fp; 3118 int error; 3119 3120 /* getvnode() will use the descriptor for us */ 3121 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3122 return (error); 3123 vp = (struct vnode *)fp->f_data; 3124 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3125 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 3126 if (error == 0 && bioopsp != NULL && 3127 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3128 (*bioopsp->io_fsync)(vp, 0); 3129 VOP_UNLOCK(vp, 0); 3130 FILE_UNUSE(fp, l); 3131 return (error); 3132 } 3133 3134 /* 3135 * Sync a range of file data. API modeled after that found in AIX. 3136 * 3137 * FDATASYNC indicates that we need only save enough metadata to be able 3138 * to re-read the written data. Note we duplicate AIX's requirement that 3139 * the file be open for writing. 3140 */ 3141 /* ARGSUSED */ 3142 int 3143 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 3144 { 3145 /* { 3146 syscallarg(int) fd; 3147 syscallarg(int) flags; 3148 syscallarg(off_t) start; 3149 syscallarg(off_t) length; 3150 } */ 3151 struct proc *p = l->l_proc; 3152 struct vnode *vp; 3153 struct file *fp; 3154 int flags, nflags; 3155 off_t s, e, len; 3156 int error; 3157 3158 /* getvnode() will use the descriptor for us */ 3159 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3160 return (error); 3161 3162 if ((fp->f_flag & FWRITE) == 0) { 3163 error = EBADF; 3164 goto out; 3165 } 3166 3167 flags = SCARG(uap, flags); 3168 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3169 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3170 error = EINVAL; 3171 goto out; 3172 } 3173 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3174 if (flags & FDATASYNC) 3175 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3176 else 3177 nflags = FSYNC_WAIT; 3178 if (flags & FDISKSYNC) 3179 nflags |= FSYNC_CACHE; 3180 3181 len = SCARG(uap, length); 3182 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3183 if (len) { 3184 s = SCARG(uap, start); 3185 e = s + len; 3186 if (e < s) { 3187 error = EINVAL; 3188 goto out; 3189 } 3190 } else { 3191 e = 0; 3192 s = 0; 3193 } 3194 3195 vp = (struct vnode *)fp->f_data; 3196 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3197 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 3198 3199 if (error == 0 && bioopsp != NULL && 3200 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3201 (*bioopsp->io_fsync)(vp, nflags); 3202 3203 VOP_UNLOCK(vp, 0); 3204 out: 3205 FILE_UNUSE(fp, l); 3206 return (error); 3207 } 3208 3209 /* 3210 * Sync the data of an open file. 3211 */ 3212 /* ARGSUSED */ 3213 int 3214 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 3215 { 3216 /* { 3217 syscallarg(int) fd; 3218 } */ 3219 struct proc *p = l->l_proc; 3220 struct vnode *vp; 3221 struct file *fp; 3222 int error; 3223 3224 /* getvnode() will use the descriptor for us */ 3225 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3226 return (error); 3227 if ((fp->f_flag & FWRITE) == 0) { 3228 FILE_UNUSE(fp, l); 3229 return (EBADF); 3230 } 3231 vp = (struct vnode *)fp->f_data; 3232 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3233 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 3234 VOP_UNLOCK(vp, 0); 3235 FILE_UNUSE(fp, l); 3236 return (error); 3237 } 3238 3239 /* 3240 * Rename files, (standard) BSD semantics frontend. 3241 */ 3242 /* ARGSUSED */ 3243 int 3244 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 3245 { 3246 /* { 3247 syscallarg(const char *) from; 3248 syscallarg(const char *) to; 3249 } */ 3250 3251 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3252 } 3253 3254 /* 3255 * Rename files, POSIX semantics frontend. 3256 */ 3257 /* ARGSUSED */ 3258 int 3259 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 3260 { 3261 /* { 3262 syscallarg(const char *) from; 3263 syscallarg(const char *) to; 3264 } */ 3265 3266 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3267 } 3268 3269 /* 3270 * Rename files. Source and destination must either both be directories, 3271 * or both not be directories. If target is a directory, it must be empty. 3272 * If `from' and `to' refer to the same object, the value of the `retain' 3273 * argument is used to determine whether `from' will be 3274 * 3275 * (retain == 0) deleted unless `from' and `to' refer to the same 3276 * object in the file system's name space (BSD). 3277 * (retain == 1) always retained (POSIX). 3278 */ 3279 static int 3280 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3281 { 3282 struct vnode *tvp, *fvp, *tdvp; 3283 struct nameidata fromnd, tond; 3284 struct proc *p; 3285 int error; 3286 3287 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, 3288 UIO_USERSPACE, from); 3289 if ((error = namei(&fromnd)) != 0) 3290 return (error); 3291 if (fromnd.ni_dvp != fromnd.ni_vp) 3292 VOP_UNLOCK(fromnd.ni_dvp, 0); 3293 fvp = fromnd.ni_vp; 3294 NDINIT(&tond, RENAME, 3295 LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT 3296 | (fvp->v_type == VDIR ? CREATEDIR : 0), 3297 UIO_USERSPACE, to); 3298 if ((error = namei(&tond)) != 0) { 3299 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3300 vrele(fromnd.ni_dvp); 3301 vrele(fvp); 3302 goto out1; 3303 } 3304 tdvp = tond.ni_dvp; 3305 tvp = tond.ni_vp; 3306 3307 if (tvp != NULL) { 3308 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3309 error = ENOTDIR; 3310 goto out; 3311 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3312 error = EISDIR; 3313 goto out; 3314 } 3315 } 3316 3317 if (fvp == tdvp) 3318 error = EINVAL; 3319 3320 /* 3321 * Source and destination refer to the same object. 3322 */ 3323 if (fvp == tvp) { 3324 if (retain) 3325 error = -1; 3326 else if (fromnd.ni_dvp == tdvp && 3327 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3328 !memcmp(fromnd.ni_cnd.cn_nameptr, 3329 tond.ni_cnd.cn_nameptr, 3330 fromnd.ni_cnd.cn_namelen)) 3331 error = -1; 3332 } 3333 3334 #if NVERIEXEC > 0 3335 if (!error) { 3336 char *f1, *f2; 3337 3338 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3339 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3340 3341 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3342 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3343 3344 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3345 3346 free(f1, M_TEMP); 3347 free(f2, M_TEMP); 3348 } 3349 #endif /* NVERIEXEC > 0 */ 3350 3351 out: 3352 p = l->l_proc; 3353 if (!error) { 3354 VOP_LEASE(tdvp, l->l_cred, LEASE_WRITE); 3355 if (fromnd.ni_dvp != tdvp) 3356 VOP_LEASE(fromnd.ni_dvp, l->l_cred, LEASE_WRITE); 3357 if (tvp) { 3358 VOP_LEASE(tvp, l->l_cred, LEASE_WRITE); 3359 } 3360 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3361 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3362 } else { 3363 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3364 if (tdvp == tvp) 3365 vrele(tdvp); 3366 else 3367 vput(tdvp); 3368 if (tvp) 3369 vput(tvp); 3370 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3371 vrele(fromnd.ni_dvp); 3372 vrele(fvp); 3373 } 3374 vrele(tond.ni_startdir); 3375 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3376 out1: 3377 if (fromnd.ni_startdir) 3378 vrele(fromnd.ni_startdir); 3379 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3380 return (error == -1 ? 0 : error); 3381 } 3382 3383 /* 3384 * Make a directory file. 3385 */ 3386 /* ARGSUSED */ 3387 int 3388 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 3389 { 3390 /* { 3391 syscallarg(const char *) path; 3392 syscallarg(int) mode; 3393 } */ 3394 struct proc *p = l->l_proc; 3395 struct vnode *vp; 3396 struct vattr vattr; 3397 int error; 3398 struct nameidata nd; 3399 3400 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3401 SCARG(uap, path)); 3402 if ((error = namei(&nd)) != 0) 3403 return (error); 3404 vp = nd.ni_vp; 3405 if (vp != NULL) { 3406 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3407 if (nd.ni_dvp == vp) 3408 vrele(nd.ni_dvp); 3409 else 3410 vput(nd.ni_dvp); 3411 vrele(vp); 3412 return (EEXIST); 3413 } 3414 VATTR_NULL(&vattr); 3415 vattr.va_type = VDIR; 3416 /* We will read cwdi->cwdi_cmask unlocked. */ 3417 vattr.va_mode = 3418 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3419 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3420 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3421 if (!error) 3422 vput(nd.ni_vp); 3423 return (error); 3424 } 3425 3426 /* 3427 * Remove a directory file. 3428 */ 3429 /* ARGSUSED */ 3430 int 3431 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 3432 { 3433 /* { 3434 syscallarg(const char *) path; 3435 } */ 3436 struct vnode *vp; 3437 int error; 3438 struct nameidata nd; 3439 3440 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3441 SCARG(uap, path)); 3442 if ((error = namei(&nd)) != 0) 3443 return (error); 3444 vp = nd.ni_vp; 3445 if (vp->v_type != VDIR) { 3446 error = ENOTDIR; 3447 goto out; 3448 } 3449 /* 3450 * No rmdir "." please. 3451 */ 3452 if (nd.ni_dvp == vp) { 3453 error = EINVAL; 3454 goto out; 3455 } 3456 /* 3457 * The root of a mounted filesystem cannot be deleted. 3458 */ 3459 if (vp->v_vflag & VV_ROOT) { 3460 error = EBUSY; 3461 goto out; 3462 } 3463 VOP_LEASE(nd.ni_dvp, l->l_cred, LEASE_WRITE); 3464 VOP_LEASE(vp, l->l_cred, LEASE_WRITE); 3465 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3466 return (error); 3467 3468 out: 3469 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3470 if (nd.ni_dvp == vp) 3471 vrele(nd.ni_dvp); 3472 else 3473 vput(nd.ni_dvp); 3474 vput(vp); 3475 return (error); 3476 } 3477 3478 /* 3479 * Read a block of directory entries in a file system independent format. 3480 */ 3481 int 3482 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 3483 { 3484 /* { 3485 syscallarg(int) fd; 3486 syscallarg(char *) buf; 3487 syscallarg(size_t) count; 3488 } */ 3489 struct proc *p = l->l_proc; 3490 struct file *fp; 3491 int error, done; 3492 3493 /* getvnode() will use the descriptor for us */ 3494 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3495 return (error); 3496 if ((fp->f_flag & FREAD) == 0) { 3497 error = EBADF; 3498 goto out; 3499 } 3500 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3501 SCARG(uap, count), &done, l, 0, 0); 3502 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 3503 *retval = done; 3504 out: 3505 FILE_UNUSE(fp, l); 3506 return (error); 3507 } 3508 3509 /* 3510 * Set the mode mask for creation of filesystem nodes. 3511 */ 3512 int 3513 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 3514 { 3515 /* { 3516 syscallarg(mode_t) newmask; 3517 } */ 3518 struct proc *p = l->l_proc; 3519 struct cwdinfo *cwdi; 3520 3521 /* 3522 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3523 * important is that we serialize changes to the mask. The 3524 * rw_exit() will issue a write memory barrier on our behalf, 3525 * and force the changes out to other CPUs (as it must use an 3526 * atomic operation, draining the local CPU's store buffers). 3527 */ 3528 cwdi = p->p_cwdi; 3529 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3530 *retval = cwdi->cwdi_cmask; 3531 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3532 rw_exit(&cwdi->cwdi_lock); 3533 3534 return (0); 3535 } 3536 3537 /* 3538 * Void all references to file by ripping underlying filesystem 3539 * away from vnode. 3540 */ 3541 /* ARGSUSED */ 3542 int 3543 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 3544 { 3545 /* { 3546 syscallarg(const char *) path; 3547 } */ 3548 struct vnode *vp; 3549 struct vattr vattr; 3550 int error; 3551 bool revoke; 3552 struct nameidata nd; 3553 3554 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 3555 SCARG(uap, path)); 3556 if ((error = namei(&nd)) != 0) 3557 return (error); 3558 vp = nd.ni_vp; 3559 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3560 goto out; 3561 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3562 (error = kauth_authorize_generic(l->l_cred, 3563 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3564 goto out; 3565 simple_lock(&vp->v_interlock); 3566 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED|VI_LAYER))); 3567 simple_unlock(&vp->v_interlock); 3568 if (revoke) 3569 VOP_REVOKE(vp, REVOKEALL); 3570 out: 3571 vrele(vp); 3572 return (error); 3573 } 3574 3575 /* 3576 * Convert a user file descriptor to a kernel file entry. 3577 */ 3578 int 3579 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3580 { 3581 struct vnode *vp; 3582 struct file *fp; 3583 3584 if ((fp = fd_getfile(fdp, fd)) == NULL) 3585 return (EBADF); 3586 3587 FILE_USE(fp); 3588 3589 if (fp->f_type != DTYPE_VNODE) { 3590 FILE_UNUSE(fp, NULL); 3591 return (EINVAL); 3592 } 3593 3594 vp = (struct vnode *)fp->f_data; 3595 if (vp->v_type == VBAD) { 3596 FILE_UNUSE(fp, NULL); 3597 return (EBADF); 3598 } 3599 3600 *fpp = fp; 3601 return (0); 3602 } 3603
Indexes created Thu Sep 25 16:09:42 GMT 2025