vfs_syscalls.c revision 1.291
1 /* $NetBSD: vfs_syscalls.c,v 1.291 2007/01/01 22:00:16 pooka 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 1.291 2007/01/01 22:00:16 pooka Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/sa.h> 65 #include <sys/syscallargs.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #if NVERIEXEC > 0 73 #include <sys/verified_exec.h> 74 #include <sys/syslog.h> 75 #endif /* NVERIEXEC > 0 */ 76 #include <sys/kauth.h> 77 78 #include <miscfs/genfs/genfs.h> 79 #include <miscfs/syncfs/syncfs.h> 80 81 #ifdef COMPAT_30 82 #include "opt_nfsserver.h" 83 #include <nfs/rpcv2.h> 84 #endif 85 #include <nfs/nfsproto.h> 86 #ifdef COMPAT_30 87 #include <nfs/nfs.h> 88 #include <nfs/nfs_var.h> 89 #endif 90 91 #if NFSS > 0 92 #include <dev/fssvar.h> 93 #endif 94 95 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 96 97 static int change_dir(struct nameidata *, struct lwp *); 98 static int change_flags(struct vnode *, u_long, struct lwp *); 99 static int change_mode(struct vnode *, int, struct lwp *l); 100 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 101 static int change_utimes(struct vnode *vp, const struct timeval *, 102 struct lwp *l); 103 static int rename_files(const char *, const char *, struct lwp *, int); 104 105 void checkdirs(struct vnode *); 106 107 static int mount_update(struct lwp *, struct vnode *, const char *, int, 108 void *, struct nameidata *); 109 static int mount_domount(struct lwp *, struct vnode *, const char *, 110 const char *, int, void *, struct nameidata *); 111 static int mount_getargs(struct lwp *, struct vnode *, const char *, int, 112 void *, struct nameidata *); 113 114 int dovfsusermount = 0; 115 116 /* 117 * Virtual File System System Calls 118 */ 119 120 /* 121 * Mount a file system. 122 */ 123 124 #if defined(COMPAT_09) || defined(COMPAT_43) 125 /* 126 * This table is used to maintain compatibility with 4.3BSD 127 * and NetBSD 0.9 mount syscalls. Note, the order is important! 128 * 129 * Do not modify this table. It should only contain filesystems 130 * supported by NetBSD 0.9 and 4.3BSD. 131 */ 132 const char * const mountcompatnames[] = { 133 NULL, /* 0 = MOUNT_NONE */ 134 MOUNT_FFS, /* 1 = MOUNT_UFS */ 135 MOUNT_NFS, /* 2 */ 136 MOUNT_MFS, /* 3 */ 137 MOUNT_MSDOS, /* 4 */ 138 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 139 MOUNT_FDESC, /* 6 */ 140 MOUNT_KERNFS, /* 7 */ 141 NULL, /* 8 = MOUNT_DEVFS */ 142 MOUNT_AFS, /* 9 */ 143 }; 144 const int nmountcompatnames = sizeof(mountcompatnames) / 145 sizeof(mountcompatnames[0]); 146 #endif /* COMPAT_09 || COMPAT_43 */ 147 148 static int 149 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 150 void *data, struct nameidata *ndp) 151 { 152 struct mount *mp; 153 int error = 0, saved_flags; 154 155 mp = vp->v_mount; 156 saved_flags = mp->mnt_flag; 157 158 /* We can't operate on VROOT here. */ 159 if ((vp->v_flag & VROOT) == 0) { 160 vput(vp); 161 error = EINVAL; 162 goto out; 163 } 164 165 /* 166 * We only allow the filesystem to be reloaded if it 167 * is currently mounted read-only. 168 */ 169 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 170 error = EOPNOTSUPP; /* Needs translation */ 171 goto out; 172 } 173 /* 174 * In "highly secure" mode, don't let the caller do anything 175 * but downgrade a filesystem from read-write to read-only. 176 */ 177 if (securelevel >= 2 && 178 flags != 179 (mp->mnt_flag | MNT_RDONLY | MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 180 error = EPERM; 181 goto out; 182 } 183 /* 184 * Only root, or the user that did the original mount is 185 * permitted to update it. 186 */ 187 if (mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred) && 188 (error = kauth_authorize_generic(l->l_cred, 189 KAUTH_GENERIC_ISSUSER, NULL)) != 0) { 190 goto out; 191 } 192 /* 193 * Do not allow NFS export by non-root users. For non-root 194 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 195 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 196 */ 197 if (kauth_cred_geteuid(l->l_cred) != 0) { 198 if (flags & MNT_EXPORTED) { 199 error = EPERM; 200 goto out; 201 } 202 flags |= MNT_NOSUID | MNT_NODEV; 203 if (saved_flags & MNT_NOEXEC) 204 flags |= MNT_NOEXEC; 205 } 206 if (vfs_busy(mp, LK_NOWAIT, 0)) { 207 error = EPERM; 208 goto out; 209 } 210 211 mp->mnt_flag &= ~MNT_OP_FLAGS; 212 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 213 214 /* 215 * Set the mount level flags. 216 */ 217 if (flags & MNT_RDONLY) 218 mp->mnt_flag |= MNT_RDONLY; 219 else if (mp->mnt_flag & MNT_RDONLY) 220 mp->mnt_iflag |= IMNT_WANTRDWR; 221 mp->mnt_flag &= 222 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 223 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 224 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 225 mp->mnt_flag |= flags & 226 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 227 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 228 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 229 MNT_IGNORE); 230 231 error = VFS_MOUNT(mp, path, data, ndp, l); 232 233 #if defined(COMPAT_30) && defined(NFSSERVER) 234 if (error) { 235 int error2; 236 237 /* Update failed; let's try and see if it was an 238 * export request. */ 239 error2 = nfs_update_exports_30(mp, path, data, l); 240 241 /* Only update error code if the export request was 242 * understood but some problem occurred while 243 * processing it. */ 244 if (error2 != EJUSTRETURN) 245 error = error2; 246 } 247 #endif 248 if (mp->mnt_iflag & IMNT_WANTRDWR) 249 mp->mnt_flag &= ~MNT_RDONLY; 250 if (error) 251 mp->mnt_flag = saved_flags; 252 mp->mnt_flag &= ~MNT_OP_FLAGS; 253 mp->mnt_iflag &= ~IMNT_WANTRDWR; 254 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 255 if (mp->mnt_syncer == NULL) 256 error = vfs_allocate_syncvnode(mp); 257 } else { 258 if (mp->mnt_syncer != NULL) 259 vfs_deallocate_syncvnode(mp); 260 } 261 vfs_unbusy(mp); 262 263 out: 264 return (error); 265 } 266 267 static int 268 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 269 const char *path, int flags, void *data, struct nameidata *ndp) 270 { 271 struct mount *mp = NULL; 272 struct vattr va; 273 char fstypename[MFSNAMELEN]; 274 int error; 275 276 /* XXX secmodel stuff. */ 277 if (securelevel >= 2) { 278 error = EPERM; 279 vput(vp); 280 goto out; 281 } 282 283 /* Can't make a non-dir a mount-point (from here anyway). */ 284 if (vp->v_type != VDIR) { 285 error = ENOTDIR; 286 vput(vp); 287 goto out; 288 } 289 290 /* 291 * If the user is not root, ensure that they own the directory 292 * onto which we are attempting to mount. 293 */ 294 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 295 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 296 (error = kauth_authorize_generic(l->l_cred, 297 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 298 vput(vp); 299 goto out; 300 } 301 302 if (flags & MNT_EXPORTED) { 303 error = EINVAL; 304 vput(vp); 305 goto out; 306 } 307 308 /* 309 * For non-root users, silently enforce MNT_NOSUID and MNT_NODEV. 310 * Also propagate MNT_NOEXEC. 311 */ 312 if (kauth_cred_geteuid(l->l_cred) != 0) { 313 flags |= MNT_NOSUID | MNT_NODEV; 314 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 315 flags |= MNT_NOEXEC; 316 } 317 318 /* 319 * Copy file-system type from userspace. 320 */ 321 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 322 if (error) { 323 #if defined(COMPAT_09) || defined(COMPAT_43) 324 /* 325 * Historically, filesystem types were identified by numbers. 326 * If we get an integer for the filesystem type instead of a 327 * string, we check to see if it matches one of the historic 328 * filesystem types. 329 */ 330 u_long fsindex = (u_long)fstype; 331 if (fsindex >= nmountcompatnames || 332 mountcompatnames[fsindex] == NULL) { 333 error = ENODEV; 334 vput(vp); 335 goto out; 336 } 337 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 338 #else 339 vput(vp); 340 goto out; 341 #endif 342 } 343 344 #ifdef COMPAT_10 345 /* Accept `ufs' as an alias for `ffs'. */ 346 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 347 strlcpy(fstypename, "ffs", sizeof(fstypename)); 348 #endif 349 350 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 351 vput(vp); 352 goto out; 353 } 354 355 /* 356 * Check if a file-system is not already mounted on this vnode. 357 */ 358 if (vp->v_mountedhere != NULL) { 359 error = EBUSY; 360 vput(vp); 361 goto out; 362 } 363 364 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 365 366 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 367 free(mp, M_MOUNT); 368 error = ENODEV; 369 vput(vp); 370 goto out; 371 } 372 373 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 374 simple_lock_init(&mp->mnt_slock); 375 (void)vfs_busy(mp, LK_NOWAIT, 0); 376 377 mp->mnt_op->vfs_refcount++; 378 mp->mnt_vnodecovered = vp; 379 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 380 mp->mnt_unmounter = NULL; 381 mp->mnt_leaf = mp; 382 mount_initspecific(mp); 383 384 /* 385 * The underlying file system may refuse the mount for 386 * various reasons. Allow the user to force it to happen. 387 * 388 * Set the mount level flags. 389 */ 390 mp->mnt_flag = flags & 391 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 392 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 393 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 394 MNT_IGNORE | MNT_RDONLY); 395 396 error = VFS_MOUNT(mp, path, data, ndp, l); 397 mp->mnt_flag &= ~MNT_OP_FLAGS; 398 399 /* 400 * Put the new filesystem on the mount list after root. 401 */ 402 cache_purge(vp); 403 if (!error) { 404 mp->mnt_iflag &= ~IMNT_WANTRDWR; 405 vp->v_mountedhere = mp; 406 simple_lock(&mountlist_slock); 407 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 408 simple_unlock(&mountlist_slock); 409 checkdirs(vp); 410 VOP_UNLOCK(vp, 0); 411 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 412 error = vfs_allocate_syncvnode(mp); 413 vfs_unbusy(mp); 414 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 415 error = VFS_START(mp, 0, l); 416 if (error) 417 vrele(vp); 418 } else { 419 vp->v_mountedhere = NULL; 420 mp->mnt_op->vfs_refcount--; 421 vfs_unbusy(mp); 422 free(mp, M_MOUNT); 423 vput(vp); 424 } 425 426 out: 427 return (error); 428 } 429 430 static int 431 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 432 void *data, struct nameidata *ndp) 433 { 434 struct mount *mp; 435 int error; 436 437 /* If MNT_GETARGS is specified, it should be the only flag. */ 438 if (flags & ~MNT_GETARGS) { 439 error = EINVAL; 440 goto out; 441 } 442 443 mp = vp->v_mount; 444 445 if ((vp->v_flag & VROOT) == 0) { 446 error = EINVAL; 447 goto out; 448 } 449 450 if (vfs_busy(mp, LK_NOWAIT, 0)) { 451 error = EPERM; 452 goto out; 453 } 454 455 mp->mnt_flag &= ~MNT_OP_FLAGS; 456 mp->mnt_flag |= MNT_GETARGS; 457 error = VFS_MOUNT(mp, path, data, ndp, l); 458 mp->mnt_flag &= ~MNT_OP_FLAGS; 459 460 vfs_unbusy(mp); 461 out: 462 return (error); 463 } 464 465 /* ARGSUSED */ 466 int 467 sys_mount(struct lwp *l, void *v, register_t *retval) 468 { 469 struct sys_mount_args /* { 470 syscallarg(const char *) type; 471 syscallarg(const char *) path; 472 syscallarg(int) flags; 473 syscallarg(void *) data; 474 } */ *uap = v; 475 struct vnode *vp; 476 struct nameidata nd; 477 int error; 478 479 /* XXX secmodel stuff. */ 480 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 481 (error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 482 &l->l_acflag))) 483 return (error); 484 485 /* 486 * Get vnode to be covered 487 */ 488 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 489 SCARG(uap, path), l); 490 if ((error = namei(&nd)) != 0) 491 return (error); 492 vp = nd.ni_vp; 493 494 /* 495 * A lookup in VFS_MOUNT might result in an attempt to 496 * lock this vnode again, so make the lock recursive. 497 */ 498 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 499 500 if (SCARG(uap, flags) & MNT_GETARGS) { 501 error = mount_getargs(l, vp, SCARG(uap, path), 502 SCARG(uap, flags), SCARG(uap, data), &nd); 503 vput(vp); 504 } else if (SCARG(uap, flags) & MNT_UPDATE) { 505 error = mount_update(l, vp, SCARG(uap, path), 506 SCARG(uap, flags), SCARG(uap, data), &nd); 507 vput(vp); 508 } else { 509 /* Locking is handled internally in mount_domount(). */ 510 error = mount_domount(l, vp, SCARG(uap, type), 511 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 512 } 513 514 return (error); 515 } 516 517 /* 518 * Scan all active processes to see if any of them have a current 519 * or root directory onto which the new filesystem has just been 520 * mounted. If so, replace them with the new mount point. 521 */ 522 void 523 checkdirs(struct vnode *olddp) 524 { 525 struct cwdinfo *cwdi; 526 struct vnode *newdp; 527 struct proc *p; 528 529 if (olddp->v_usecount == 1) 530 return; 531 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 532 panic("mount: lost mount"); 533 proclist_lock_read(); 534 PROCLIST_FOREACH(p, &allproc) { 535 cwdi = p->p_cwdi; 536 if (!cwdi) 537 continue; 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 } 549 proclist_unlock_read(); 550 if (rootvnode == olddp) { 551 vrele(rootvnode); 552 VREF(newdp); 553 rootvnode = newdp; 554 } 555 vput(newdp); 556 } 557 558 /* 559 * Unmount a file system. 560 * 561 * Note: unmount takes a path to the vnode mounted on as argument, 562 * not special file (as before). 563 */ 564 /* ARGSUSED */ 565 int 566 sys_unmount(struct lwp *l, void *v, register_t *retval) 567 { 568 struct sys_unmount_args /* { 569 syscallarg(const char *) path; 570 syscallarg(int) flags; 571 } */ *uap = v; 572 struct vnode *vp; 573 struct mount *mp; 574 int error; 575 struct nameidata nd; 576 577 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 578 SCARG(uap, path), l); 579 if ((error = namei(&nd)) != 0) 580 return (error); 581 vp = nd.ni_vp; 582 mp = vp->v_mount; 583 584 /* 585 * Only root, or the user that did the original mount is 586 * permitted to unmount this filesystem. 587 */ 588 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred)) && 589 (error = kauth_authorize_generic(l->l_cred, 590 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 591 vput(vp); 592 return (error); 593 } 594 595 /* 596 * Don't allow unmounting the root file system. 597 */ 598 if (mp->mnt_flag & MNT_ROOTFS) { 599 vput(vp); 600 return (EINVAL); 601 } 602 603 /* 604 * Must be the root of the filesystem 605 */ 606 if ((vp->v_flag & VROOT) == 0) { 607 vput(vp); 608 return (EINVAL); 609 } 610 vput(vp); 611 612 /* 613 * XXX Freeze syncer. Must do this before locking the 614 * mount point. See dounmount() for details. 615 */ 616 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 617 618 if (vfs_busy(mp, 0, 0)) { 619 lockmgr(&syncer_lock, LK_RELEASE, NULL); 620 return (EBUSY); 621 } 622 623 return (dounmount(mp, SCARG(uap, flags), l)); 624 } 625 626 /* 627 * Do the actual file system unmount. File system is assumed to have been 628 * marked busy by the caller. 629 */ 630 int 631 dounmount(struct mount *mp, int flags, struct lwp *l) 632 { 633 struct vnode *coveredvp; 634 int error; 635 int async; 636 int used_syncer; 637 638 #if NVERIEXEC > 0 639 error = veriexec_unmountchk(mp); 640 if (error) 641 return (error); 642 #endif /* NVERIEXEC > 0 */ 643 644 simple_lock(&mountlist_slock); 645 vfs_unbusy(mp); 646 used_syncer = (mp->mnt_syncer != NULL); 647 648 /* 649 * XXX Syncer must be frozen when we get here. This should really 650 * be done on a per-mountpoint basis, but especially the softdep 651 * code possibly called from the syncer doesn't exactly work on a 652 * per-mountpoint basis, so the softdep code would become a maze 653 * of vfs_busy() calls. 654 * 655 * The caller of dounmount() must acquire syncer_lock because 656 * the syncer itself acquires locks in syncer_lock -> vfs_busy 657 * order, and we must preserve that order to avoid deadlock. 658 * 659 * So, if the file system did not use the syncer, now is 660 * the time to release the syncer_lock. 661 */ 662 if (used_syncer == 0) 663 lockmgr(&syncer_lock, LK_RELEASE, NULL); 664 665 mp->mnt_iflag |= IMNT_UNMOUNT; 666 mp->mnt_unmounter = l; 667 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 668 vn_start_write(NULL, &mp, V_WAIT); 669 670 async = mp->mnt_flag & MNT_ASYNC; 671 mp->mnt_flag &= ~MNT_ASYNC; 672 cache_purgevfs(mp); /* remove cache entries for this file sys */ 673 if (mp->mnt_syncer != NULL) 674 vfs_deallocate_syncvnode(mp); 675 error = 0; 676 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 677 #if NFSS > 0 678 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 679 #endif 680 if (error == 0) 681 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 682 } 683 if (error == 0 || (flags & MNT_FORCE)) 684 error = VFS_UNMOUNT(mp, flags, l); 685 vn_finished_write(mp, 0); 686 simple_lock(&mountlist_slock); 687 if (error) { 688 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 689 (void) vfs_allocate_syncvnode(mp); 690 mp->mnt_iflag &= ~IMNT_UNMOUNT; 691 mp->mnt_unmounter = NULL; 692 mp->mnt_flag |= async; 693 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 694 &mountlist_slock); 695 if (used_syncer) 696 lockmgr(&syncer_lock, LK_RELEASE, NULL); 697 simple_lock(&mp->mnt_slock); 698 while (mp->mnt_wcnt > 0) { 699 wakeup(mp); 700 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 701 0, &mp->mnt_slock); 702 } 703 simple_unlock(&mp->mnt_slock); 704 return (error); 705 } 706 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 707 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 708 coveredvp->v_mountedhere = NULL; 709 vrele(coveredvp); 710 } 711 mp->mnt_op->vfs_refcount--; 712 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 713 panic("unmount: dangling vnode"); 714 mp->mnt_iflag |= IMNT_GONE; 715 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 716 mount_finispecific(mp); 717 if (used_syncer) 718 lockmgr(&syncer_lock, LK_RELEASE, NULL); 719 simple_lock(&mp->mnt_slock); 720 while (mp->mnt_wcnt > 0) { 721 wakeup(mp); 722 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 723 } 724 simple_unlock(&mp->mnt_slock); 725 vfs_hooks_unmount(mp); 726 free(mp, M_MOUNT); 727 return (0); 728 } 729 730 /* 731 * Sync each mounted filesystem. 732 */ 733 #ifdef DEBUG 734 int syncprt = 0; 735 struct ctldebug debug0 = { "syncprt", &syncprt }; 736 #endif 737 738 /* ARGSUSED */ 739 int 740 sys_sync(struct lwp *l, void *v, register_t *retval) 741 { 742 struct mount *mp, *nmp; 743 int asyncflag; 744 745 if (l == NULL) 746 l = &lwp0; 747 748 simple_lock(&mountlist_slock); 749 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 750 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 751 nmp = mp->mnt_list.cqe_prev; 752 continue; 753 } 754 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 755 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 756 asyncflag = mp->mnt_flag & MNT_ASYNC; 757 mp->mnt_flag &= ~MNT_ASYNC; 758 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 759 if (asyncflag) 760 mp->mnt_flag |= MNT_ASYNC; 761 vn_finished_write(mp, 0); 762 } 763 simple_lock(&mountlist_slock); 764 nmp = mp->mnt_list.cqe_prev; 765 vfs_unbusy(mp); 766 767 } 768 simple_unlock(&mountlist_slock); 769 #ifdef DEBUG 770 if (syncprt) 771 vfs_bufstats(); 772 #endif /* DEBUG */ 773 return (0); 774 } 775 776 /* 777 * Change filesystem quotas. 778 */ 779 /* ARGSUSED */ 780 int 781 sys_quotactl(struct lwp *l, void *v, register_t *retval) 782 { 783 struct sys_quotactl_args /* { 784 syscallarg(const char *) path; 785 syscallarg(int) cmd; 786 syscallarg(int) uid; 787 syscallarg(void *) arg; 788 } */ *uap = v; 789 struct mount *mp; 790 int error; 791 struct nameidata nd; 792 793 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 794 if ((error = namei(&nd)) != 0) 795 return (error); 796 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 797 vrele(nd.ni_vp); 798 if (error) 799 return (error); 800 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 801 SCARG(uap, arg), l); 802 vn_finished_write(mp, 0); 803 return (error); 804 } 805 806 int 807 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 808 int root) 809 { 810 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 811 int error = 0; 812 813 /* 814 * If MNT_NOWAIT or MNT_LAZY is specified, do not 815 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 816 * overrides MNT_NOWAIT. 817 */ 818 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 819 (flags != MNT_WAIT && flags != 0)) { 820 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 821 goto done; 822 } 823 824 /* Get the filesystem stats now */ 825 memset(sp, 0, sizeof(*sp)); 826 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 827 return error; 828 } 829 830 if (cwdi->cwdi_rdir == NULL) 831 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 832 done: 833 if (cwdi->cwdi_rdir != NULL) { 834 size_t len; 835 char *bp; 836 char *path = PNBUF_GET(); 837 if (!path) 838 return ENOMEM; 839 840 bp = path + MAXPATHLEN; 841 *--bp = '\0'; 842 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 843 MAXPATHLEN / 2, 0, l); 844 if (error) { 845 PNBUF_PUT(path); 846 return error; 847 } 848 len = strlen(bp); 849 /* 850 * for mount points that are below our root, we can see 851 * them, so we fix up the pathname and return them. The 852 * rest we cannot see, so we don't allow viewing the 853 * data. 854 */ 855 if (strncmp(bp, sp->f_mntonname, len) == 0) { 856 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 857 sizeof(sp->f_mntonname)); 858 if (sp->f_mntonname[0] == '\0') 859 (void)strlcpy(sp->f_mntonname, "/", 860 sizeof(sp->f_mntonname)); 861 } else { 862 if (root) 863 (void)strlcpy(sp->f_mntonname, "/", 864 sizeof(sp->f_mntonname)); 865 else 866 error = EPERM; 867 } 868 PNBUF_PUT(path); 869 } 870 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 871 return error; 872 } 873 874 /* 875 * Get filesystem statistics. 876 */ 877 /* ARGSUSED */ 878 int 879 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 880 { 881 struct sys_statvfs1_args /* { 882 syscallarg(const char *) path; 883 syscallarg(struct statvfs *) buf; 884 syscallarg(int) flags; 885 } */ *uap = v; 886 struct mount *mp; 887 struct statvfs *sb; 888 int error; 889 struct nameidata nd; 890 891 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 892 if ((error = namei(&nd)) != 0) 893 return error; 894 mp = nd.ni_vp->v_mount; 895 vrele(nd.ni_vp); 896 sb = STATVFSBUF_GET(); 897 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 898 if (error == 0) { 899 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 900 } 901 STATVFSBUF_PUT(sb); 902 return error; 903 } 904 905 /* 906 * Get filesystem statistics. 907 */ 908 /* ARGSUSED */ 909 int 910 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 911 { 912 struct sys_fstatvfs1_args /* { 913 syscallarg(int) fd; 914 syscallarg(struct statvfs *) buf; 915 syscallarg(int) flags; 916 } */ *uap = v; 917 struct proc *p = l->l_proc; 918 struct file *fp; 919 struct mount *mp; 920 struct statvfs *sb; 921 int error; 922 923 /* getvnode() will use the descriptor for us */ 924 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 925 return (error); 926 mp = ((struct vnode *)fp->f_data)->v_mount; 927 sb = STATVFSBUF_GET(); 928 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 929 goto out; 930 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 931 out: 932 FILE_UNUSE(fp, l); 933 STATVFSBUF_PUT(sb); 934 return error; 935 } 936 937 938 /* 939 * Get statistics on all filesystems. 940 */ 941 int 942 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 943 { 944 struct sys_getvfsstat_args /* { 945 syscallarg(struct statvfs *) buf; 946 syscallarg(size_t) bufsize; 947 syscallarg(int) flags; 948 } */ *uap = v; 949 int root = 0; 950 struct proc *p = l->l_proc; 951 struct mount *mp, *nmp; 952 struct statvfs *sb; 953 struct statvfs *sfsp; 954 size_t count, maxcount; 955 int error = 0; 956 957 sb = STATVFSBUF_GET(); 958 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 959 sfsp = SCARG(uap, buf); 960 simple_lock(&mountlist_slock); 961 count = 0; 962 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 963 mp = nmp) { 964 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 965 nmp = CIRCLEQ_NEXT(mp, mnt_list); 966 continue; 967 } 968 if (sfsp && count < maxcount) { 969 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 970 if (error) { 971 simple_lock(&mountlist_slock); 972 nmp = CIRCLEQ_NEXT(mp, mnt_list); 973 vfs_unbusy(mp); 974 continue; 975 } 976 error = copyout(sb, sfsp, sizeof(*sfsp)); 977 if (error) { 978 vfs_unbusy(mp); 979 goto out; 980 } 981 sfsp++; 982 root |= strcmp(sb->f_mntonname, "/") == 0; 983 } 984 count++; 985 simple_lock(&mountlist_slock); 986 nmp = CIRCLEQ_NEXT(mp, mnt_list); 987 vfs_unbusy(mp); 988 } 989 simple_unlock(&mountlist_slock); 990 if (root == 0 && p->p_cwdi->cwdi_rdir) { 991 /* 992 * fake a root entry 993 */ 994 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 995 SCARG(uap, flags), 1)) != 0) 996 goto out; 997 if (sfsp) 998 error = copyout(sb, sfsp, sizeof(*sfsp)); 999 count++; 1000 } 1001 if (sfsp && count > maxcount) 1002 *retval = maxcount; 1003 else 1004 *retval = count; 1005 out: 1006 STATVFSBUF_PUT(sb); 1007 return error; 1008 } 1009 1010 /* 1011 * Change current working directory to a given file descriptor. 1012 */ 1013 /* ARGSUSED */ 1014 int 1015 sys_fchdir(struct lwp *l, void *v, register_t *retval) 1016 { 1017 struct sys_fchdir_args /* { 1018 syscallarg(int) fd; 1019 } */ *uap = v; 1020 struct proc *p = l->l_proc; 1021 struct filedesc *fdp = p->p_fd; 1022 struct cwdinfo *cwdi = p->p_cwdi; 1023 struct vnode *vp, *tdp; 1024 struct mount *mp; 1025 struct file *fp; 1026 int error; 1027 1028 /* getvnode() will use the descriptor for us */ 1029 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1030 return (error); 1031 vp = (struct vnode *)fp->f_data; 1032 1033 VREF(vp); 1034 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1035 if (vp->v_type != VDIR) 1036 error = ENOTDIR; 1037 else 1038 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1039 while (!error && (mp = vp->v_mountedhere) != NULL) { 1040 if (vfs_busy(mp, 0, 0)) 1041 continue; 1042 error = VFS_ROOT(mp, &tdp); 1043 vfs_unbusy(mp); 1044 if (error) 1045 break; 1046 vput(vp); 1047 vp = tdp; 1048 } 1049 if (error) { 1050 vput(vp); 1051 goto out; 1052 } 1053 VOP_UNLOCK(vp, 0); 1054 1055 /* 1056 * Disallow changing to a directory not under the process's 1057 * current root directory (if there is one). 1058 */ 1059 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1060 vrele(vp); 1061 error = EPERM; /* operation not permitted */ 1062 goto out; 1063 } 1064 1065 vrele(cwdi->cwdi_cdir); 1066 cwdi->cwdi_cdir = vp; 1067 out: 1068 FILE_UNUSE(fp, l); 1069 return (error); 1070 } 1071 1072 /* 1073 * Change this process's notion of the root directory to a given file 1074 * descriptor. 1075 */ 1076 int 1077 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1078 { 1079 struct sys_fchroot_args *uap = v; 1080 struct proc *p = l->l_proc; 1081 struct filedesc *fdp = p->p_fd; 1082 struct cwdinfo *cwdi = p->p_cwdi; 1083 struct vnode *vp; 1084 struct file *fp; 1085 int error; 1086 1087 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1088 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1089 return error; 1090 /* getvnode() will use the descriptor for us */ 1091 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1092 return error; 1093 vp = (struct vnode *) fp->f_data; 1094 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1095 if (vp->v_type != VDIR) 1096 error = ENOTDIR; 1097 else 1098 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1099 VOP_UNLOCK(vp, 0); 1100 if (error) 1101 goto out; 1102 VREF(vp); 1103 1104 /* 1105 * Prevent escaping from chroot by putting the root under 1106 * the working directory. Silently chdir to / if we aren't 1107 * already there. 1108 */ 1109 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1110 /* 1111 * XXX would be more failsafe to change directory to a 1112 * deadfs node here instead 1113 */ 1114 vrele(cwdi->cwdi_cdir); 1115 VREF(vp); 1116 cwdi->cwdi_cdir = vp; 1117 } 1118 1119 if (cwdi->cwdi_rdir != NULL) 1120 vrele(cwdi->cwdi_rdir); 1121 cwdi->cwdi_rdir = vp; 1122 out: 1123 FILE_UNUSE(fp, l); 1124 return (error); 1125 } 1126 1127 /* 1128 * Change current working directory (``.''). 1129 */ 1130 /* ARGSUSED */ 1131 int 1132 sys_chdir(struct lwp *l, void *v, register_t *retval) 1133 { 1134 struct sys_chdir_args /* { 1135 syscallarg(const char *) path; 1136 } */ *uap = v; 1137 struct proc *p = l->l_proc; 1138 struct cwdinfo *cwdi = p->p_cwdi; 1139 int error; 1140 struct nameidata nd; 1141 1142 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1143 SCARG(uap, path), l); 1144 if ((error = change_dir(&nd, l)) != 0) 1145 return (error); 1146 vrele(cwdi->cwdi_cdir); 1147 cwdi->cwdi_cdir = nd.ni_vp; 1148 return (0); 1149 } 1150 1151 /* 1152 * Change notion of root (``/'') directory. 1153 */ 1154 /* ARGSUSED */ 1155 int 1156 sys_chroot(struct lwp *l, void *v, register_t *retval) 1157 { 1158 struct sys_chroot_args /* { 1159 syscallarg(const char *) path; 1160 } */ *uap = v; 1161 struct proc *p = l->l_proc; 1162 struct cwdinfo *cwdi = p->p_cwdi; 1163 struct vnode *vp; 1164 int error; 1165 struct nameidata nd; 1166 1167 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1168 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1169 return (error); 1170 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1171 SCARG(uap, path), l); 1172 if ((error = change_dir(&nd, l)) != 0) 1173 return (error); 1174 if (cwdi->cwdi_rdir != NULL) 1175 vrele(cwdi->cwdi_rdir); 1176 vp = nd.ni_vp; 1177 cwdi->cwdi_rdir = vp; 1178 1179 /* 1180 * Prevent escaping from chroot by putting the root under 1181 * the working directory. Silently chdir to / if we aren't 1182 * already there. 1183 */ 1184 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1185 /* 1186 * XXX would be more failsafe to change directory to a 1187 * deadfs node here instead 1188 */ 1189 vrele(cwdi->cwdi_cdir); 1190 VREF(vp); 1191 cwdi->cwdi_cdir = vp; 1192 } 1193 1194 return (0); 1195 } 1196 1197 /* 1198 * Common routine for chroot and chdir. 1199 */ 1200 static int 1201 change_dir(struct nameidata *ndp, struct lwp *l) 1202 { 1203 struct vnode *vp; 1204 int error; 1205 1206 if ((error = namei(ndp)) != 0) 1207 return (error); 1208 vp = ndp->ni_vp; 1209 if (vp->v_type != VDIR) 1210 error = ENOTDIR; 1211 else 1212 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1213 1214 if (error) 1215 vput(vp); 1216 else 1217 VOP_UNLOCK(vp, 0); 1218 return (error); 1219 } 1220 1221 /* 1222 * Check permissions, allocate an open file structure, 1223 * and call the device open routine if any. 1224 */ 1225 int 1226 sys_open(struct lwp *l, void *v, register_t *retval) 1227 { 1228 struct sys_open_args /* { 1229 syscallarg(const char *) path; 1230 syscallarg(int) flags; 1231 syscallarg(int) mode; 1232 } */ *uap = v; 1233 struct proc *p = l->l_proc; 1234 struct cwdinfo *cwdi = p->p_cwdi; 1235 struct filedesc *fdp = p->p_fd; 1236 struct file *fp; 1237 struct vnode *vp; 1238 int flags, cmode; 1239 int type, indx, error; 1240 struct flock lf; 1241 struct nameidata nd; 1242 1243 flags = FFLAGS(SCARG(uap, flags)); 1244 if ((flags & (FREAD | FWRITE)) == 0) 1245 return (EINVAL); 1246 /* falloc() will use the file descriptor for us */ 1247 if ((error = falloc(l, &fp, &indx)) != 0) 1248 return (error); 1249 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1250 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1251 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1252 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1253 FILE_UNUSE(fp, l); 1254 fdp->fd_ofiles[indx] = NULL; 1255 ffree(fp); 1256 if ((error == EDUPFD || error == EMOVEFD) && 1257 l->l_dupfd >= 0 && /* XXX from fdopen */ 1258 (error = 1259 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1260 *retval = indx; 1261 return (0); 1262 } 1263 if (error == ERESTART) 1264 error = EINTR; 1265 fdremove(fdp, indx); 1266 return (error); 1267 } 1268 l->l_dupfd = 0; 1269 vp = nd.ni_vp; 1270 fp->f_flag = flags & FMASK; 1271 fp->f_type = DTYPE_VNODE; 1272 fp->f_ops = &vnops; 1273 fp->f_data = vp; 1274 if (flags & (O_EXLOCK | O_SHLOCK)) { 1275 lf.l_whence = SEEK_SET; 1276 lf.l_start = 0; 1277 lf.l_len = 0; 1278 if (flags & O_EXLOCK) 1279 lf.l_type = F_WRLCK; 1280 else 1281 lf.l_type = F_RDLCK; 1282 type = F_FLOCK; 1283 if ((flags & FNONBLOCK) == 0) 1284 type |= F_WAIT; 1285 VOP_UNLOCK(vp, 0); 1286 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1287 if (error) { 1288 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1289 FILE_UNUSE(fp, l); 1290 ffree(fp); 1291 fdremove(fdp, indx); 1292 return (error); 1293 } 1294 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1295 fp->f_flag |= FHASLOCK; 1296 } 1297 VOP_UNLOCK(vp, 0); 1298 *retval = indx; 1299 FILE_SET_MATURE(fp); 1300 FILE_UNUSE(fp, l); 1301 return (0); 1302 } 1303 1304 static void 1305 vfs__fhfree(fhandle_t *fhp) 1306 { 1307 size_t fhsize; 1308 1309 if (fhp == NULL) { 1310 return; 1311 } 1312 fhsize = FHANDLE_SIZE(fhp); 1313 kmem_free(fhp, fhsize); 1314 } 1315 1316 /* 1317 * vfs_composefh: compose a filehandle. 1318 */ 1319 1320 int 1321 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1322 { 1323 struct mount *mp; 1324 struct fid *fidp; 1325 int error; 1326 size_t needfhsize; 1327 size_t fidsize; 1328 1329 mp = vp->v_mount; 1330 fidp = NULL; 1331 if (*fh_size < FHANDLE_SIZE_MIN) { 1332 fidsize = 0; 1333 } else { 1334 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1335 if (fhp != NULL) { 1336 memset(fhp, 0, *fh_size); 1337 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1338 fidp = &fhp->fh_fid; 1339 } 1340 } 1341 error = VFS_VPTOFH(vp, fidp, &fidsize); 1342 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1343 if (error == 0 && *fh_size < needfhsize) { 1344 error = E2BIG; 1345 } 1346 *fh_size = needfhsize; 1347 return error; 1348 } 1349 1350 int 1351 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1352 { 1353 struct mount *mp; 1354 fhandle_t *fhp; 1355 size_t fhsize; 1356 size_t fidsize; 1357 int error; 1358 1359 *fhpp = NULL; 1360 mp = vp->v_mount; 1361 fidsize = 0; 1362 error = VFS_VPTOFH(vp, NULL, &fidsize); 1363 KASSERT(error != 0); 1364 if (error != E2BIG) { 1365 goto out; 1366 } 1367 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1368 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1369 if (fhp == NULL) { 1370 error = ENOMEM; 1371 goto out; 1372 } 1373 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1374 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1375 if (error == 0) { 1376 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1377 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1378 *fhpp = fhp; 1379 } else { 1380 kmem_free(fhp, fhsize); 1381 } 1382 out: 1383 return error; 1384 } 1385 1386 void 1387 vfs_composefh_free(fhandle_t *fhp) 1388 { 1389 1390 vfs__fhfree(fhp); 1391 } 1392 1393 /* 1394 * vfs_fhtovp: lookup a vnode by a filehandle. 1395 */ 1396 1397 int 1398 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1399 { 1400 struct mount *mp; 1401 int error; 1402 1403 *vpp = NULL; 1404 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1405 if (mp == NULL) { 1406 error = ESTALE; 1407 goto out; 1408 } 1409 if (mp->mnt_op->vfs_fhtovp == NULL) { 1410 error = EOPNOTSUPP; 1411 goto out; 1412 } 1413 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1414 out: 1415 return error; 1416 } 1417 1418 /* 1419 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1420 * the needed size. 1421 */ 1422 1423 int 1424 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1425 { 1426 fhandle_t *fhp; 1427 int error; 1428 1429 *fhpp = NULL; 1430 if (fhsize > FHANDLE_SIZE_MAX) { 1431 return EINVAL; 1432 } 1433 if (fhsize < FHANDLE_SIZE_MIN) { 1434 return EINVAL; 1435 } 1436 again: 1437 fhp = kmem_alloc(fhsize, KM_SLEEP); 1438 if (fhp == NULL) { 1439 return ENOMEM; 1440 } 1441 error = copyin(ufhp, fhp, fhsize); 1442 if (error == 0) { 1443 /* XXX this check shouldn't be here */ 1444 if (FHANDLE_SIZE(fhp) == fhsize) { 1445 *fhpp = fhp; 1446 return 0; 1447 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1448 /* 1449 * a kludge for nfsv2 padded handles. 1450 */ 1451 size_t sz; 1452 1453 sz = FHANDLE_SIZE(fhp); 1454 kmem_free(fhp, fhsize); 1455 fhsize = sz; 1456 goto again; 1457 } else { 1458 /* 1459 * userland told us wrong size. 1460 */ 1461 error = EINVAL; 1462 } 1463 } 1464 kmem_free(fhp, fhsize); 1465 return error; 1466 } 1467 1468 void 1469 vfs_copyinfh_free(fhandle_t *fhp) 1470 { 1471 1472 vfs__fhfree(fhp); 1473 } 1474 1475 /* 1476 * Get file handle system call 1477 */ 1478 int 1479 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1480 { 1481 struct sys___getfh30_args /* { 1482 syscallarg(char *) fname; 1483 syscallarg(fhandle_t *) fhp; 1484 syscallarg(size_t *) fh_size; 1485 } */ *uap = v; 1486 struct vnode *vp; 1487 fhandle_t *fh; 1488 int error; 1489 struct nameidata nd; 1490 size_t sz; 1491 size_t usz; 1492 1493 /* 1494 * Must be super user 1495 */ 1496 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1497 0, NULL, NULL, NULL); 1498 if (error) 1499 return (error); 1500 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1501 SCARG(uap, fname), l); 1502 error = namei(&nd); 1503 if (error) 1504 return (error); 1505 vp = nd.ni_vp; 1506 error = vfs_composefh_alloc(vp, &fh); 1507 vput(vp); 1508 if (error != 0) { 1509 goto out; 1510 } 1511 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1512 if (error != 0) { 1513 goto out; 1514 } 1515 sz = FHANDLE_SIZE(fh); 1516 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1517 if (error != 0) { 1518 goto out; 1519 } 1520 if (usz >= sz) { 1521 error = copyout(fh, SCARG(uap, fhp), sz); 1522 } else { 1523 error = E2BIG; 1524 } 1525 out: 1526 vfs_composefh_free(fh); 1527 return (error); 1528 } 1529 1530 /* 1531 * Open a file given a file handle. 1532 * 1533 * Check permissions, allocate an open file structure, 1534 * and call the device open routine if any. 1535 */ 1536 1537 int 1538 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1539 register_t *retval) 1540 { 1541 struct filedesc *fdp = l->l_proc->p_fd; 1542 struct file *fp; 1543 struct vnode *vp = NULL; 1544 struct mount *mp; 1545 kauth_cred_t cred = l->l_cred; 1546 struct file *nfp; 1547 int type, indx, error=0; 1548 struct flock lf; 1549 struct vattr va; 1550 fhandle_t *fh; 1551 int flags; 1552 1553 /* 1554 * Must be super user 1555 */ 1556 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1557 0, NULL, NULL, NULL))) 1558 return (error); 1559 1560 flags = FFLAGS(oflags); 1561 if ((flags & (FREAD | FWRITE)) == 0) 1562 return (EINVAL); 1563 if ((flags & O_CREAT)) 1564 return (EINVAL); 1565 /* falloc() will use the file descriptor for us */ 1566 if ((error = falloc(l, &nfp, &indx)) != 0) 1567 return (error); 1568 fp = nfp; 1569 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1570 if (error != 0) { 1571 goto bad; 1572 } 1573 error = vfs_fhtovp(fh, &vp); 1574 if (error != 0) { 1575 goto bad; 1576 } 1577 1578 /* Now do an effective vn_open */ 1579 1580 if (vp->v_type == VSOCK) { 1581 error = EOPNOTSUPP; 1582 goto bad; 1583 } 1584 if (flags & FREAD) { 1585 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1586 goto bad; 1587 } 1588 if (flags & (FWRITE | O_TRUNC)) { 1589 if (vp->v_type == VDIR) { 1590 error = EISDIR; 1591 goto bad; 1592 } 1593 if ((error = vn_writechk(vp)) != 0 || 1594 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1595 goto bad; 1596 } 1597 if (flags & O_TRUNC) { 1598 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1599 goto bad; 1600 VOP_UNLOCK(vp, 0); /* XXX */ 1601 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1602 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1603 VATTR_NULL(&va); 1604 va.va_size = 0; 1605 error = VOP_SETATTR(vp, &va, cred, l); 1606 vn_finished_write(mp, 0); 1607 if (error) 1608 goto bad; 1609 } 1610 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1611 goto bad; 1612 if (vp->v_type == VREG && 1613 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1614 error = EIO; 1615 goto bad; 1616 } 1617 if (flags & FWRITE) 1618 vp->v_writecount++; 1619 1620 /* done with modified vn_open, now finish what sys_open does. */ 1621 1622 fp->f_flag = flags & FMASK; 1623 fp->f_type = DTYPE_VNODE; 1624 fp->f_ops = &vnops; 1625 fp->f_data = vp; 1626 if (flags & (O_EXLOCK | O_SHLOCK)) { 1627 lf.l_whence = SEEK_SET; 1628 lf.l_start = 0; 1629 lf.l_len = 0; 1630 if (flags & O_EXLOCK) 1631 lf.l_type = F_WRLCK; 1632 else 1633 lf.l_type = F_RDLCK; 1634 type = F_FLOCK; 1635 if ((flags & FNONBLOCK) == 0) 1636 type |= F_WAIT; 1637 VOP_UNLOCK(vp, 0); 1638 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1639 if (error) { 1640 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1641 FILE_UNUSE(fp, l); 1642 ffree(fp); 1643 fdremove(fdp, indx); 1644 return (error); 1645 } 1646 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1647 fp->f_flag |= FHASLOCK; 1648 } 1649 VOP_UNLOCK(vp, 0); 1650 *retval = indx; 1651 FILE_SET_MATURE(fp); 1652 FILE_UNUSE(fp, l); 1653 vfs_copyinfh_free(fh); 1654 return (0); 1655 1656 bad: 1657 FILE_UNUSE(fp, l); 1658 ffree(fp); 1659 fdremove(fdp, indx); 1660 if (vp != NULL) 1661 vput(vp); 1662 vfs_copyinfh_free(fh); 1663 return (error); 1664 } 1665 1666 int 1667 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1668 { 1669 struct sys___fhopen40_args /* { 1670 syscallarg(const void *) fhp; 1671 syscallarg(size_t) fh_size; 1672 syscallarg(int) flags; 1673 } */ *uap = v; 1674 1675 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1676 SCARG(uap, flags), retval); 1677 } 1678 1679 int 1680 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1681 register_t *retval) 1682 { 1683 struct stat sb; 1684 int error; 1685 fhandle_t *fh; 1686 struct vnode *vp; 1687 1688 /* 1689 * Must be super user 1690 */ 1691 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1692 0, NULL, NULL, NULL))) 1693 return (error); 1694 1695 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1696 if (error != 0) { 1697 goto bad; 1698 } 1699 error = vfs_fhtovp(fh, &vp); 1700 if (error != 0) { 1701 goto bad; 1702 } 1703 error = vn_stat(vp, &sb, l); 1704 vput(vp); 1705 if (error) { 1706 goto bad; 1707 } 1708 error = copyout(&sb, sbp, sizeof(sb)); 1709 bad: 1710 vfs_copyinfh_free(fh); 1711 return error; 1712 } 1713 1714 1715 /* ARGSUSED */ 1716 int 1717 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1718 { 1719 struct sys___fhstat40_args /* { 1720 syscallarg(const void *) fhp; 1721 syscallarg(size_t) fh_size; 1722 syscallarg(struct stat *) sb; 1723 } */ *uap = v; 1724 1725 return dofhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), SCARG(uap, sb), 1726 retval); 1727 } 1728 1729 int 1730 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1731 int flags, register_t *retval) 1732 { 1733 struct statvfs *sb = NULL; 1734 fhandle_t *fh; 1735 struct mount *mp; 1736 struct vnode *vp; 1737 int error; 1738 1739 /* 1740 * Must be super user 1741 */ 1742 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1743 0, NULL, NULL, NULL))) 1744 return error; 1745 1746 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1747 if (error != 0) { 1748 goto out; 1749 } 1750 error = vfs_fhtovp(fh, &vp); 1751 if (error != 0) { 1752 goto out; 1753 } 1754 mp = vp->v_mount; 1755 sb = STATVFSBUF_GET(); 1756 if ((error = dostatvfs(mp, sb, l, flags, 1)) != 0) { 1757 vput(vp); 1758 goto out; 1759 } 1760 vput(vp); 1761 error = copyout(sb, buf, sizeof(*sb)); 1762 out: 1763 if (sb != NULL) { 1764 STATVFSBUF_PUT(sb); 1765 } 1766 vfs_copyinfh_free(fh); 1767 return error; 1768 } 1769 1770 /* ARGSUSED */ 1771 int 1772 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1773 { 1774 struct sys___fhstatvfs140_args /* { 1775 syscallarg(const void *) fhp; 1776 syscallarg(size_t) fh_size; 1777 syscallarg(struct statvfs *) buf; 1778 syscallarg(int) flags; 1779 } */ *uap = v; 1780 1781 return dofhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1782 SCARG(uap, buf), SCARG(uap, flags), retval); 1783 } 1784 1785 /* 1786 * Create a special file. 1787 */ 1788 /* ARGSUSED */ 1789 int 1790 sys_mknod(struct lwp *l, void *v, register_t *retval) 1791 { 1792 struct sys_mknod_args /* { 1793 syscallarg(const char *) path; 1794 syscallarg(int) mode; 1795 syscallarg(int) dev; 1796 } */ *uap = v; 1797 struct proc *p = l->l_proc; 1798 struct vnode *vp; 1799 struct mount *mp; 1800 struct vattr vattr; 1801 int error; 1802 int whiteout = 0; 1803 struct nameidata nd; 1804 1805 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1806 0, NULL, NULL, NULL)) != 0) 1807 return (error); 1808 restart: 1809 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1810 if ((error = namei(&nd)) != 0) 1811 return (error); 1812 vp = nd.ni_vp; 1813 if (vp != NULL) 1814 error = EEXIST; 1815 else { 1816 VATTR_NULL(&vattr); 1817 vattr.va_mode = 1818 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1819 vattr.va_rdev = SCARG(uap, dev); 1820 whiteout = 0; 1821 1822 switch (SCARG(uap, mode) & S_IFMT) { 1823 case S_IFMT: /* used by badsect to flag bad sectors */ 1824 vattr.va_type = VBAD; 1825 break; 1826 case S_IFCHR: 1827 vattr.va_type = VCHR; 1828 break; 1829 case S_IFBLK: 1830 vattr.va_type = VBLK; 1831 break; 1832 case S_IFWHT: 1833 whiteout = 1; 1834 break; 1835 default: 1836 error = EINVAL; 1837 break; 1838 } 1839 } 1840 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1841 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1842 if (nd.ni_dvp == vp) 1843 vrele(nd.ni_dvp); 1844 else 1845 vput(nd.ni_dvp); 1846 if (vp) 1847 vrele(vp); 1848 if ((error = vn_start_write(NULL, &mp, 1849 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1850 return (error); 1851 goto restart; 1852 } 1853 if (!error) { 1854 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1855 if (whiteout) { 1856 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1857 if (error) 1858 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1859 vput(nd.ni_dvp); 1860 } else { 1861 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1862 &nd.ni_cnd, &vattr); 1863 if (error == 0) 1864 vput(nd.ni_vp); 1865 } 1866 } else { 1867 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1868 if (nd.ni_dvp == vp) 1869 vrele(nd.ni_dvp); 1870 else 1871 vput(nd.ni_dvp); 1872 if (vp) 1873 vrele(vp); 1874 } 1875 vn_finished_write(mp, 0); 1876 return (error); 1877 } 1878 1879 /* 1880 * Create a named pipe. 1881 */ 1882 /* ARGSUSED */ 1883 int 1884 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1885 { 1886 struct sys_mkfifo_args /* { 1887 syscallarg(const char *) path; 1888 syscallarg(int) mode; 1889 } */ *uap = v; 1890 struct proc *p = l->l_proc; 1891 struct mount *mp; 1892 struct vattr vattr; 1893 int error; 1894 struct nameidata nd; 1895 1896 restart: 1897 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1898 if ((error = namei(&nd)) != 0) 1899 return (error); 1900 if (nd.ni_vp != NULL) { 1901 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1902 if (nd.ni_dvp == nd.ni_vp) 1903 vrele(nd.ni_dvp); 1904 else 1905 vput(nd.ni_dvp); 1906 vrele(nd.ni_vp); 1907 return (EEXIST); 1908 } 1909 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1910 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1911 if (nd.ni_dvp == nd.ni_vp) 1912 vrele(nd.ni_dvp); 1913 else 1914 vput(nd.ni_dvp); 1915 if (nd.ni_vp) 1916 vrele(nd.ni_vp); 1917 if ((error = vn_start_write(NULL, &mp, 1918 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1919 return (error); 1920 goto restart; 1921 } 1922 VATTR_NULL(&vattr); 1923 vattr.va_type = VFIFO; 1924 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1925 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1926 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1927 if (error == 0) 1928 vput(nd.ni_vp); 1929 vn_finished_write(mp, 0); 1930 return (error); 1931 } 1932 1933 /* 1934 * Make a hard file link. 1935 */ 1936 /* ARGSUSED */ 1937 int 1938 sys_link(struct lwp *l, void *v, register_t *retval) 1939 { 1940 struct sys_link_args /* { 1941 syscallarg(const char *) path; 1942 syscallarg(const char *) link; 1943 } */ *uap = v; 1944 struct vnode *vp; 1945 struct mount *mp; 1946 struct nameidata nd; 1947 int error; 1948 1949 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1950 if ((error = namei(&nd)) != 0) 1951 return (error); 1952 vp = nd.ni_vp; 1953 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1954 vrele(vp); 1955 return (error); 1956 } 1957 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1958 if ((error = namei(&nd)) != 0) 1959 goto out; 1960 if (nd.ni_vp) { 1961 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1962 if (nd.ni_dvp == nd.ni_vp) 1963 vrele(nd.ni_dvp); 1964 else 1965 vput(nd.ni_dvp); 1966 vrele(nd.ni_vp); 1967 error = EEXIST; 1968 goto out; 1969 } 1970 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1971 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1972 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1973 out: 1974 vrele(vp); 1975 vn_finished_write(mp, 0); 1976 return (error); 1977 } 1978 1979 /* 1980 * Make a symbolic link. 1981 */ 1982 /* ARGSUSED */ 1983 int 1984 sys_symlink(struct lwp *l, void *v, register_t *retval) 1985 { 1986 struct sys_symlink_args /* { 1987 syscallarg(const char *) path; 1988 syscallarg(const char *) link; 1989 } */ *uap = v; 1990 struct proc *p = l->l_proc; 1991 struct mount *mp; 1992 struct vattr vattr; 1993 char *path; 1994 int error; 1995 struct nameidata nd; 1996 1997 path = PNBUF_GET(); 1998 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1999 if (error) 2000 goto out; 2001 restart: 2002 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 2003 if ((error = namei(&nd)) != 0) 2004 goto out; 2005 if (nd.ni_vp) { 2006 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2007 if (nd.ni_dvp == nd.ni_vp) 2008 vrele(nd.ni_dvp); 2009 else 2010 vput(nd.ni_dvp); 2011 vrele(nd.ni_vp); 2012 error = EEXIST; 2013 goto out; 2014 } 2015 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2016 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2017 if (nd.ni_dvp == nd.ni_vp) 2018 vrele(nd.ni_dvp); 2019 else 2020 vput(nd.ni_dvp); 2021 if ((error = vn_start_write(NULL, &mp, 2022 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2023 return (error); 2024 goto restart; 2025 } 2026 VATTR_NULL(&vattr); 2027 vattr.va_type = VLNK; 2028 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2029 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2030 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2031 if (error == 0) 2032 vput(nd.ni_vp); 2033 vn_finished_write(mp, 0); 2034 out: 2035 PNBUF_PUT(path); 2036 return (error); 2037 } 2038 2039 /* 2040 * Delete a whiteout from the filesystem. 2041 */ 2042 /* ARGSUSED */ 2043 int 2044 sys_undelete(struct lwp *l, void *v, register_t *retval) 2045 { 2046 struct sys_undelete_args /* { 2047 syscallarg(const char *) path; 2048 } */ *uap = v; 2049 int error; 2050 struct mount *mp; 2051 struct nameidata nd; 2052 2053 restart: 2054 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2055 SCARG(uap, path), l); 2056 error = namei(&nd); 2057 if (error) 2058 return (error); 2059 2060 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2061 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2062 if (nd.ni_dvp == nd.ni_vp) 2063 vrele(nd.ni_dvp); 2064 else 2065 vput(nd.ni_dvp); 2066 if (nd.ni_vp) 2067 vrele(nd.ni_vp); 2068 return (EEXIST); 2069 } 2070 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2071 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2072 if (nd.ni_dvp == nd.ni_vp) 2073 vrele(nd.ni_dvp); 2074 else 2075 vput(nd.ni_dvp); 2076 if ((error = vn_start_write(NULL, &mp, 2077 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2078 return (error); 2079 goto restart; 2080 } 2081 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2082 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2083 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2084 vput(nd.ni_dvp); 2085 vn_finished_write(mp, 0); 2086 return (error); 2087 } 2088 2089 /* 2090 * Delete a name from the filesystem. 2091 */ 2092 /* ARGSUSED */ 2093 int 2094 sys_unlink(struct lwp *l, void *v, register_t *retval) 2095 { 2096 struct sys_unlink_args /* { 2097 syscallarg(const char *) path; 2098 } */ *uap = v; 2099 struct mount *mp; 2100 struct vnode *vp; 2101 int error; 2102 struct nameidata nd; 2103 #if NVERIEXEC > 0 2104 pathname_t pathbuf; 2105 #endif /* NVERIEXEC > 0 */ 2106 2107 restart: 2108 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2109 SCARG(uap, path), l); 2110 if ((error = namei(&nd)) != 0) 2111 return (error); 2112 vp = nd.ni_vp; 2113 2114 /* 2115 * The root of a mounted filesystem cannot be deleted. 2116 */ 2117 if (vp->v_flag & VROOT) { 2118 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2119 if (nd.ni_dvp == vp) 2120 vrele(nd.ni_dvp); 2121 else 2122 vput(nd.ni_dvp); 2123 vput(vp); 2124 error = EBUSY; 2125 goto out; 2126 } 2127 2128 #if NVERIEXEC > 0 2129 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2130 2131 /* Handle remove requests for veriexec entries. */ 2132 if (!error) { 2133 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2134 pathname_put(pathbuf); 2135 } 2136 2137 if (error) { 2138 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2139 if (nd.ni_dvp == vp) 2140 vrele(nd.ni_dvp); 2141 else 2142 vput(nd.ni_dvp); 2143 vput(vp); 2144 goto out; 2145 } 2146 #endif /* NVERIEXEC > 0 */ 2147 2148 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2149 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2150 if (nd.ni_dvp == vp) 2151 vrele(nd.ni_dvp); 2152 else 2153 vput(nd.ni_dvp); 2154 vput(vp); 2155 if ((error = vn_start_write(NULL, &mp, 2156 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2157 return (error); 2158 goto restart; 2159 } 2160 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2161 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2162 #ifdef FILEASSOC 2163 (void)fileassoc_file_delete(vp); 2164 #endif /* FILEASSOC */ 2165 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2166 vn_finished_write(mp, 0); 2167 out: 2168 return (error); 2169 } 2170 2171 /* 2172 * Reposition read/write file offset. 2173 */ 2174 int 2175 sys_lseek(struct lwp *l, void *v, register_t *retval) 2176 { 2177 struct sys_lseek_args /* { 2178 syscallarg(int) fd; 2179 syscallarg(int) pad; 2180 syscallarg(off_t) offset; 2181 syscallarg(int) whence; 2182 } */ *uap = v; 2183 struct proc *p = l->l_proc; 2184 kauth_cred_t cred = l->l_cred; 2185 struct filedesc *fdp = p->p_fd; 2186 struct file *fp; 2187 struct vnode *vp; 2188 struct vattr vattr; 2189 off_t newoff; 2190 int error; 2191 2192 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2193 return (EBADF); 2194 2195 FILE_USE(fp); 2196 2197 vp = (struct vnode *)fp->f_data; 2198 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2199 error = ESPIPE; 2200 goto out; 2201 } 2202 2203 switch (SCARG(uap, whence)) { 2204 case SEEK_CUR: 2205 newoff = fp->f_offset + SCARG(uap, offset); 2206 break; 2207 case SEEK_END: 2208 error = VOP_GETATTR(vp, &vattr, cred, l); 2209 if (error) 2210 goto out; 2211 newoff = SCARG(uap, offset) + vattr.va_size; 2212 break; 2213 case SEEK_SET: 2214 newoff = SCARG(uap, offset); 2215 break; 2216 default: 2217 error = EINVAL; 2218 goto out; 2219 } 2220 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2221 goto out; 2222 2223 *(off_t *)retval = fp->f_offset = newoff; 2224 out: 2225 FILE_UNUSE(fp, l); 2226 return (error); 2227 } 2228 2229 /* 2230 * Positional read system call. 2231 */ 2232 int 2233 sys_pread(struct lwp *l, void *v, register_t *retval) 2234 { 2235 struct sys_pread_args /* { 2236 syscallarg(int) fd; 2237 syscallarg(void *) buf; 2238 syscallarg(size_t) nbyte; 2239 syscallarg(off_t) offset; 2240 } */ *uap = v; 2241 struct proc *p = l->l_proc; 2242 struct filedesc *fdp = p->p_fd; 2243 struct file *fp; 2244 struct vnode *vp; 2245 off_t offset; 2246 int error, fd = SCARG(uap, fd); 2247 2248 if ((fp = fd_getfile(fdp, fd)) == NULL) 2249 return (EBADF); 2250 2251 if ((fp->f_flag & FREAD) == 0) { 2252 simple_unlock(&fp->f_slock); 2253 return (EBADF); 2254 } 2255 2256 FILE_USE(fp); 2257 2258 vp = (struct vnode *)fp->f_data; 2259 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2260 error = ESPIPE; 2261 goto out; 2262 } 2263 2264 offset = SCARG(uap, offset); 2265 2266 /* 2267 * XXX This works because no file systems actually 2268 * XXX take any action on the seek operation. 2269 */ 2270 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2271 goto out; 2272 2273 /* dofileread() will unuse the descriptor for us */ 2274 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2275 &offset, 0, retval)); 2276 2277 out: 2278 FILE_UNUSE(fp, l); 2279 return (error); 2280 } 2281 2282 /* 2283 * Positional scatter read system call. 2284 */ 2285 int 2286 sys_preadv(struct lwp *l, void *v, register_t *retval) 2287 { 2288 struct sys_preadv_args /* { 2289 syscallarg(int) fd; 2290 syscallarg(const struct iovec *) iovp; 2291 syscallarg(int) iovcnt; 2292 syscallarg(off_t) offset; 2293 } */ *uap = v; 2294 struct proc *p = l->l_proc; 2295 struct filedesc *fdp = p->p_fd; 2296 struct file *fp; 2297 struct vnode *vp; 2298 off_t offset; 2299 int error, fd = SCARG(uap, fd); 2300 2301 if ((fp = fd_getfile(fdp, fd)) == NULL) 2302 return (EBADF); 2303 2304 if ((fp->f_flag & FREAD) == 0) { 2305 simple_unlock(&fp->f_slock); 2306 return (EBADF); 2307 } 2308 2309 FILE_USE(fp); 2310 2311 vp = (struct vnode *)fp->f_data; 2312 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2313 error = ESPIPE; 2314 goto out; 2315 } 2316 2317 offset = SCARG(uap, offset); 2318 2319 /* 2320 * XXX This works because no file systems actually 2321 * XXX take any action on the seek operation. 2322 */ 2323 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2324 goto out; 2325 2326 /* dofilereadv() will unuse the descriptor for us */ 2327 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2328 &offset, 0, retval)); 2329 2330 out: 2331 FILE_UNUSE(fp, l); 2332 return (error); 2333 } 2334 2335 /* 2336 * Positional write system call. 2337 */ 2338 int 2339 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2340 { 2341 struct sys_pwrite_args /* { 2342 syscallarg(int) fd; 2343 syscallarg(const void *) buf; 2344 syscallarg(size_t) nbyte; 2345 syscallarg(off_t) offset; 2346 } */ *uap = v; 2347 struct proc *p = l->l_proc; 2348 struct filedesc *fdp = p->p_fd; 2349 struct file *fp; 2350 struct vnode *vp; 2351 off_t offset; 2352 int error, fd = SCARG(uap, fd); 2353 2354 if ((fp = fd_getfile(fdp, fd)) == NULL) 2355 return (EBADF); 2356 2357 if ((fp->f_flag & FWRITE) == 0) { 2358 simple_unlock(&fp->f_slock); 2359 return (EBADF); 2360 } 2361 2362 FILE_USE(fp); 2363 2364 vp = (struct vnode *)fp->f_data; 2365 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2366 error = ESPIPE; 2367 goto out; 2368 } 2369 2370 offset = SCARG(uap, offset); 2371 2372 /* 2373 * XXX This works because no file systems actually 2374 * XXX take any action on the seek operation. 2375 */ 2376 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2377 goto out; 2378 2379 /* dofilewrite() will unuse the descriptor for us */ 2380 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2381 &offset, 0, retval)); 2382 2383 out: 2384 FILE_UNUSE(fp, l); 2385 return (error); 2386 } 2387 2388 /* 2389 * Positional gather write system call. 2390 */ 2391 int 2392 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2393 { 2394 struct sys_pwritev_args /* { 2395 syscallarg(int) fd; 2396 syscallarg(const struct iovec *) iovp; 2397 syscallarg(int) iovcnt; 2398 syscallarg(off_t) offset; 2399 } */ *uap = v; 2400 struct proc *p = l->l_proc; 2401 struct filedesc *fdp = p->p_fd; 2402 struct file *fp; 2403 struct vnode *vp; 2404 off_t offset; 2405 int error, fd = SCARG(uap, fd); 2406 2407 if ((fp = fd_getfile(fdp, fd)) == NULL) 2408 return (EBADF); 2409 2410 if ((fp->f_flag & FWRITE) == 0) { 2411 simple_unlock(&fp->f_slock); 2412 return (EBADF); 2413 } 2414 2415 FILE_USE(fp); 2416 2417 vp = (struct vnode *)fp->f_data; 2418 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2419 error = ESPIPE; 2420 goto out; 2421 } 2422 2423 offset = SCARG(uap, offset); 2424 2425 /* 2426 * XXX This works because no file systems actually 2427 * XXX take any action on the seek operation. 2428 */ 2429 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2430 goto out; 2431 2432 /* dofilewritev() will unuse the descriptor for us */ 2433 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2434 &offset, 0, retval)); 2435 2436 out: 2437 FILE_UNUSE(fp, l); 2438 return (error); 2439 } 2440 2441 /* 2442 * Check access permissions. 2443 */ 2444 int 2445 sys_access(struct lwp *l, void *v, register_t *retval) 2446 { 2447 struct sys_access_args /* { 2448 syscallarg(const char *) path; 2449 syscallarg(int) flags; 2450 } */ *uap = v; 2451 kauth_cred_t cred; 2452 struct vnode *vp; 2453 int error, flags; 2454 struct nameidata nd; 2455 2456 cred = kauth_cred_dup(l->l_cred); 2457 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2458 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2459 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2460 SCARG(uap, path), l); 2461 /* Override default credentials */ 2462 nd.ni_cnd.cn_cred = cred; 2463 if ((error = namei(&nd)) != 0) 2464 goto out; 2465 vp = nd.ni_vp; 2466 2467 /* Flags == 0 means only check for existence. */ 2468 if (SCARG(uap, flags)) { 2469 flags = 0; 2470 if (SCARG(uap, flags) & R_OK) 2471 flags |= VREAD; 2472 if (SCARG(uap, flags) & W_OK) 2473 flags |= VWRITE; 2474 if (SCARG(uap, flags) & X_OK) 2475 flags |= VEXEC; 2476 2477 error = VOP_ACCESS(vp, flags, cred, l); 2478 if (!error && (flags & VWRITE)) 2479 error = vn_writechk(vp); 2480 } 2481 vput(vp); 2482 out: 2483 kauth_cred_free(cred); 2484 return (error); 2485 } 2486 2487 /* 2488 * Get file status; this version follows links. 2489 */ 2490 /* ARGSUSED */ 2491 int 2492 sys___stat30(struct lwp *l, void *v, register_t *retval) 2493 { 2494 struct sys___stat30_args /* { 2495 syscallarg(const char *) path; 2496 syscallarg(struct stat *) ub; 2497 } */ *uap = v; 2498 struct stat sb; 2499 int error; 2500 struct nameidata nd; 2501 2502 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2503 SCARG(uap, path), l); 2504 if ((error = namei(&nd)) != 0) 2505 return (error); 2506 error = vn_stat(nd.ni_vp, &sb, l); 2507 vput(nd.ni_vp); 2508 if (error) 2509 return (error); 2510 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2511 return (error); 2512 } 2513 2514 /* 2515 * Get file status; this version does not follow links. 2516 */ 2517 /* ARGSUSED */ 2518 int 2519 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2520 { 2521 struct sys___lstat30_args /* { 2522 syscallarg(const char *) path; 2523 syscallarg(struct stat *) ub; 2524 } */ *uap = v; 2525 struct stat sb; 2526 int error; 2527 struct nameidata nd; 2528 2529 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2530 SCARG(uap, path), l); 2531 if ((error = namei(&nd)) != 0) 2532 return (error); 2533 error = vn_stat(nd.ni_vp, &sb, l); 2534 vput(nd.ni_vp); 2535 if (error) 2536 return (error); 2537 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2538 return (error); 2539 } 2540 2541 /* 2542 * Get configurable pathname variables. 2543 */ 2544 /* ARGSUSED */ 2545 int 2546 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2547 { 2548 struct sys_pathconf_args /* { 2549 syscallarg(const char *) path; 2550 syscallarg(int) name; 2551 } */ *uap = v; 2552 int error; 2553 struct nameidata nd; 2554 2555 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2556 SCARG(uap, path), l); 2557 if ((error = namei(&nd)) != 0) 2558 return (error); 2559 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2560 vput(nd.ni_vp); 2561 return (error); 2562 } 2563 2564 /* 2565 * Return target name of a symbolic link. 2566 */ 2567 /* ARGSUSED */ 2568 int 2569 sys_readlink(struct lwp *l, void *v, register_t *retval) 2570 { 2571 struct sys_readlink_args /* { 2572 syscallarg(const char *) path; 2573 syscallarg(char *) buf; 2574 syscallarg(size_t) count; 2575 } */ *uap = v; 2576 struct vnode *vp; 2577 struct iovec aiov; 2578 struct uio auio; 2579 int error; 2580 struct nameidata nd; 2581 2582 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2583 SCARG(uap, path), l); 2584 if ((error = namei(&nd)) != 0) 2585 return (error); 2586 vp = nd.ni_vp; 2587 if (vp->v_type != VLNK) 2588 error = EINVAL; 2589 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2590 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2591 aiov.iov_base = SCARG(uap, buf); 2592 aiov.iov_len = SCARG(uap, count); 2593 auio.uio_iov = &aiov; 2594 auio.uio_iovcnt = 1; 2595 auio.uio_offset = 0; 2596 auio.uio_rw = UIO_READ; 2597 KASSERT(l == curlwp); 2598 auio.uio_vmspace = l->l_proc->p_vmspace; 2599 auio.uio_resid = SCARG(uap, count); 2600 error = VOP_READLINK(vp, &auio, l->l_cred); 2601 } 2602 vput(vp); 2603 *retval = SCARG(uap, count) - auio.uio_resid; 2604 return (error); 2605 } 2606 2607 /* 2608 * Change flags of a file given a path name. 2609 */ 2610 /* ARGSUSED */ 2611 int 2612 sys_chflags(struct lwp *l, void *v, register_t *retval) 2613 { 2614 struct sys_chflags_args /* { 2615 syscallarg(const char *) path; 2616 syscallarg(u_long) flags; 2617 } */ *uap = v; 2618 struct vnode *vp; 2619 int error; 2620 struct nameidata nd; 2621 2622 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2623 if ((error = namei(&nd)) != 0) 2624 return (error); 2625 vp = nd.ni_vp; 2626 error = change_flags(vp, SCARG(uap, flags), l); 2627 vput(vp); 2628 return (error); 2629 } 2630 2631 /* 2632 * Change flags of a file given a file descriptor. 2633 */ 2634 /* ARGSUSED */ 2635 int 2636 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2637 { 2638 struct sys_fchflags_args /* { 2639 syscallarg(int) fd; 2640 syscallarg(u_long) flags; 2641 } */ *uap = v; 2642 struct proc *p = l->l_proc; 2643 struct vnode *vp; 2644 struct file *fp; 2645 int error; 2646 2647 /* getvnode() will use the descriptor for us */ 2648 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2649 return (error); 2650 vp = (struct vnode *)fp->f_data; 2651 error = change_flags(vp, SCARG(uap, flags), l); 2652 VOP_UNLOCK(vp, 0); 2653 FILE_UNUSE(fp, l); 2654 return (error); 2655 } 2656 2657 /* 2658 * Change flags of a file given a path name; this version does 2659 * not follow links. 2660 */ 2661 int 2662 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2663 { 2664 struct sys_lchflags_args /* { 2665 syscallarg(const char *) path; 2666 syscallarg(u_long) flags; 2667 } */ *uap = v; 2668 struct vnode *vp; 2669 int error; 2670 struct nameidata nd; 2671 2672 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2673 if ((error = namei(&nd)) != 0) 2674 return (error); 2675 vp = nd.ni_vp; 2676 error = change_flags(vp, SCARG(uap, flags), l); 2677 vput(vp); 2678 return (error); 2679 } 2680 2681 /* 2682 * Common routine to change flags of a file. 2683 */ 2684 int 2685 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2686 { 2687 struct mount *mp; 2688 struct vattr vattr; 2689 int error; 2690 2691 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2692 return (error); 2693 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2694 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2695 /* 2696 * Non-superusers cannot change the flags on devices, even if they 2697 * own them. 2698 */ 2699 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2700 &l->l_acflag) != 0) { 2701 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2702 goto out; 2703 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2704 error = EINVAL; 2705 goto out; 2706 } 2707 } 2708 VATTR_NULL(&vattr); 2709 vattr.va_flags = flags; 2710 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2711 out: 2712 vn_finished_write(mp, 0); 2713 return (error); 2714 } 2715 2716 /* 2717 * Change mode of a file given path name; this version follows links. 2718 */ 2719 /* ARGSUSED */ 2720 int 2721 sys_chmod(struct lwp *l, void *v, register_t *retval) 2722 { 2723 struct sys_chmod_args /* { 2724 syscallarg(const char *) path; 2725 syscallarg(int) mode; 2726 } */ *uap = v; 2727 int error; 2728 struct nameidata nd; 2729 2730 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2731 if ((error = namei(&nd)) != 0) 2732 return (error); 2733 2734 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2735 2736 vrele(nd.ni_vp); 2737 return (error); 2738 } 2739 2740 /* 2741 * Change mode of a file given a file descriptor. 2742 */ 2743 /* ARGSUSED */ 2744 int 2745 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2746 { 2747 struct sys_fchmod_args /* { 2748 syscallarg(int) fd; 2749 syscallarg(int) mode; 2750 } */ *uap = v; 2751 struct proc *p = l->l_proc; 2752 struct file *fp; 2753 int error; 2754 2755 /* getvnode() will use the descriptor for us */ 2756 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2757 return (error); 2758 2759 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2760 FILE_UNUSE(fp, l); 2761 return (error); 2762 } 2763 2764 /* 2765 * Change mode of a file given path name; this version does not follow links. 2766 */ 2767 /* ARGSUSED */ 2768 int 2769 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2770 { 2771 struct sys_lchmod_args /* { 2772 syscallarg(const char *) path; 2773 syscallarg(int) mode; 2774 } */ *uap = v; 2775 int error; 2776 struct nameidata nd; 2777 2778 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2779 if ((error = namei(&nd)) != 0) 2780 return (error); 2781 2782 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2783 2784 vrele(nd.ni_vp); 2785 return (error); 2786 } 2787 2788 /* 2789 * Common routine to set mode given a vnode. 2790 */ 2791 static int 2792 change_mode(struct vnode *vp, int mode, struct lwp *l) 2793 { 2794 struct mount *mp; 2795 struct vattr vattr; 2796 int error; 2797 2798 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2799 return (error); 2800 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2801 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2802 VATTR_NULL(&vattr); 2803 vattr.va_mode = mode & ALLPERMS; 2804 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2805 VOP_UNLOCK(vp, 0); 2806 vn_finished_write(mp, 0); 2807 return (error); 2808 } 2809 2810 /* 2811 * Set ownership given a path name; this version follows links. 2812 */ 2813 /* ARGSUSED */ 2814 int 2815 sys_chown(struct lwp *l, void *v, register_t *retval) 2816 { 2817 struct sys_chown_args /* { 2818 syscallarg(const char *) path; 2819 syscallarg(uid_t) uid; 2820 syscallarg(gid_t) gid; 2821 } */ *uap = v; 2822 int error; 2823 struct nameidata nd; 2824 2825 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2826 if ((error = namei(&nd)) != 0) 2827 return (error); 2828 2829 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2830 2831 vrele(nd.ni_vp); 2832 return (error); 2833 } 2834 2835 /* 2836 * Set ownership given a path name; this version follows links. 2837 * Provides POSIX semantics. 2838 */ 2839 /* ARGSUSED */ 2840 int 2841 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2842 { 2843 struct sys_chown_args /* { 2844 syscallarg(const char *) path; 2845 syscallarg(uid_t) uid; 2846 syscallarg(gid_t) gid; 2847 } */ *uap = v; 2848 int error; 2849 struct nameidata nd; 2850 2851 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2852 if ((error = namei(&nd)) != 0) 2853 return (error); 2854 2855 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2856 2857 vrele(nd.ni_vp); 2858 return (error); 2859 } 2860 2861 /* 2862 * Set ownership given a file descriptor. 2863 */ 2864 /* ARGSUSED */ 2865 int 2866 sys_fchown(struct lwp *l, void *v, register_t *retval) 2867 { 2868 struct sys_fchown_args /* { 2869 syscallarg(int) fd; 2870 syscallarg(uid_t) uid; 2871 syscallarg(gid_t) gid; 2872 } */ *uap = v; 2873 struct proc *p = l->l_proc; 2874 int error; 2875 struct file *fp; 2876 2877 /* getvnode() will use the descriptor for us */ 2878 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2879 return (error); 2880 2881 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2882 SCARG(uap, gid), l, 0); 2883 FILE_UNUSE(fp, l); 2884 return (error); 2885 } 2886 2887 /* 2888 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2889 */ 2890 /* ARGSUSED */ 2891 int 2892 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2893 { 2894 struct sys_fchown_args /* { 2895 syscallarg(int) fd; 2896 syscallarg(uid_t) uid; 2897 syscallarg(gid_t) gid; 2898 } */ *uap = v; 2899 struct proc *p = l->l_proc; 2900 int error; 2901 struct file *fp; 2902 2903 /* getvnode() will use the descriptor for us */ 2904 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2905 return (error); 2906 2907 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2908 SCARG(uap, gid), l, 1); 2909 FILE_UNUSE(fp, l); 2910 return (error); 2911 } 2912 2913 /* 2914 * Set ownership given a path name; this version does not follow links. 2915 */ 2916 /* ARGSUSED */ 2917 int 2918 sys_lchown(struct lwp *l, void *v, register_t *retval) 2919 { 2920 struct sys_lchown_args /* { 2921 syscallarg(const char *) path; 2922 syscallarg(uid_t) uid; 2923 syscallarg(gid_t) gid; 2924 } */ *uap = v; 2925 int error; 2926 struct nameidata nd; 2927 2928 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2929 if ((error = namei(&nd)) != 0) 2930 return (error); 2931 2932 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2933 2934 vrele(nd.ni_vp); 2935 return (error); 2936 } 2937 2938 /* 2939 * Set ownership given a path name; this version does not follow links. 2940 * Provides POSIX/XPG semantics. 2941 */ 2942 /* ARGSUSED */ 2943 int 2944 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2945 { 2946 struct sys_lchown_args /* { 2947 syscallarg(const char *) path; 2948 syscallarg(uid_t) uid; 2949 syscallarg(gid_t) gid; 2950 } */ *uap = v; 2951 int error; 2952 struct nameidata nd; 2953 2954 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2955 if ((error = namei(&nd)) != 0) 2956 return (error); 2957 2958 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2959 2960 vrele(nd.ni_vp); 2961 return (error); 2962 } 2963 2964 /* 2965 * Common routine to set ownership given a vnode. 2966 */ 2967 static int 2968 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2969 int posix_semantics) 2970 { 2971 struct mount *mp; 2972 struct vattr vattr; 2973 mode_t newmode; 2974 int error; 2975 2976 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2977 return (error); 2978 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2979 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2980 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2981 goto out; 2982 2983 #define CHANGED(x) ((int)(x) != -1) 2984 newmode = vattr.va_mode; 2985 if (posix_semantics) { 2986 /* 2987 * POSIX/XPG semantics: if the caller is not the super-user, 2988 * clear set-user-id and set-group-id bits. Both POSIX and 2989 * the XPG consider the behaviour for calls by the super-user 2990 * implementation-defined; we leave the set-user-id and set- 2991 * group-id settings intact in that case. 2992 */ 2993 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2994 NULL) != 0) 2995 newmode &= ~(S_ISUID | S_ISGID); 2996 } else { 2997 /* 2998 * NetBSD semantics: when changing owner and/or group, 2999 * clear the respective bit(s). 3000 */ 3001 if (CHANGED(uid)) 3002 newmode &= ~S_ISUID; 3003 if (CHANGED(gid)) 3004 newmode &= ~S_ISGID; 3005 } 3006 /* Update va_mode iff altered. */ 3007 if (vattr.va_mode == newmode) 3008 newmode = VNOVAL; 3009 3010 VATTR_NULL(&vattr); 3011 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3012 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3013 vattr.va_mode = newmode; 3014 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3015 #undef CHANGED 3016 3017 out: 3018 VOP_UNLOCK(vp, 0); 3019 vn_finished_write(mp, 0); 3020 return (error); 3021 } 3022 3023 /* 3024 * Set the access and modification times given a path name; this 3025 * version follows links. 3026 */ 3027 /* ARGSUSED */ 3028 int 3029 sys_utimes(struct lwp *l, void *v, register_t *retval) 3030 { 3031 struct sys_utimes_args /* { 3032 syscallarg(const char *) path; 3033 syscallarg(const struct timeval *) tptr; 3034 } */ *uap = v; 3035 int error; 3036 struct nameidata nd; 3037 3038 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3039 if ((error = namei(&nd)) != 0) 3040 return (error); 3041 3042 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3043 3044 vrele(nd.ni_vp); 3045 return (error); 3046 } 3047 3048 /* 3049 * Set the access and modification times given a file descriptor. 3050 */ 3051 /* ARGSUSED */ 3052 int 3053 sys_futimes(struct lwp *l, void *v, register_t *retval) 3054 { 3055 struct sys_futimes_args /* { 3056 syscallarg(int) fd; 3057 syscallarg(const struct timeval *) tptr; 3058 } */ *uap = v; 3059 struct proc *p = l->l_proc; 3060 int error; 3061 struct file *fp; 3062 3063 /* getvnode() will use the descriptor for us */ 3064 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3065 return (error); 3066 3067 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3068 FILE_UNUSE(fp, l); 3069 return (error); 3070 } 3071 3072 /* 3073 * Set the access and modification times given a path name; this 3074 * version does not follow links. 3075 */ 3076 /* ARGSUSED */ 3077 int 3078 sys_lutimes(struct lwp *l, void *v, register_t *retval) 3079 { 3080 struct sys_lutimes_args /* { 3081 syscallarg(const char *) path; 3082 syscallarg(const struct timeval *) tptr; 3083 } */ *uap = v; 3084 int error; 3085 struct nameidata nd; 3086 3087 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3088 if ((error = namei(&nd)) != 0) 3089 return (error); 3090 3091 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3092 3093 vrele(nd.ni_vp); 3094 return (error); 3095 } 3096 3097 /* 3098 * Common routine to set access and modification times given a vnode. 3099 */ 3100 static int 3101 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3102 { 3103 struct mount *mp; 3104 struct vattr vattr; 3105 int error; 3106 3107 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3108 return (error); 3109 VATTR_NULL(&vattr); 3110 if (tptr == NULL) { 3111 nanotime(&vattr.va_atime); 3112 vattr.va_mtime = vattr.va_atime; 3113 vattr.va_vaflags |= VA_UTIMES_NULL; 3114 } else { 3115 struct timeval tv[2]; 3116 3117 error = copyin(tptr, tv, sizeof(tv)); 3118 if (error) 3119 goto out; 3120 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3121 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3122 } 3123 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3124 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3125 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3126 VOP_UNLOCK(vp, 0); 3127 out: 3128 vn_finished_write(mp, 0); 3129 return (error); 3130 } 3131 3132 /* 3133 * Truncate a file given its path name. 3134 */ 3135 /* ARGSUSED */ 3136 int 3137 sys_truncate(struct lwp *l, void *v, register_t *retval) 3138 { 3139 struct sys_truncate_args /* { 3140 syscallarg(const char *) path; 3141 syscallarg(int) pad; 3142 syscallarg(off_t) length; 3143 } */ *uap = v; 3144 struct vnode *vp; 3145 struct mount *mp; 3146 struct vattr vattr; 3147 int error; 3148 struct nameidata nd; 3149 3150 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3151 if ((error = namei(&nd)) != 0) 3152 return (error); 3153 vp = nd.ni_vp; 3154 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3155 vrele(vp); 3156 return (error); 3157 } 3158 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3159 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3160 if (vp->v_type == VDIR) 3161 error = EISDIR; 3162 else if ((error = vn_writechk(vp)) == 0 && 3163 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3164 VATTR_NULL(&vattr); 3165 vattr.va_size = SCARG(uap, length); 3166 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3167 } 3168 vput(vp); 3169 vn_finished_write(mp, 0); 3170 return (error); 3171 } 3172 3173 /* 3174 * Truncate a file given a file descriptor. 3175 */ 3176 /* ARGSUSED */ 3177 int 3178 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3179 { 3180 struct sys_ftruncate_args /* { 3181 syscallarg(int) fd; 3182 syscallarg(int) pad; 3183 syscallarg(off_t) length; 3184 } */ *uap = v; 3185 struct proc *p = l->l_proc; 3186 struct mount *mp; 3187 struct vattr vattr; 3188 struct vnode *vp; 3189 struct file *fp; 3190 int error; 3191 3192 /* getvnode() will use the descriptor for us */ 3193 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3194 return (error); 3195 if ((fp->f_flag & FWRITE) == 0) { 3196 error = EINVAL; 3197 goto out; 3198 } 3199 vp = (struct vnode *)fp->f_data; 3200 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3201 FILE_UNUSE(fp, l); 3202 return (error); 3203 } 3204 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3205 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3206 if (vp->v_type == VDIR) 3207 error = EISDIR; 3208 else if ((error = vn_writechk(vp)) == 0) { 3209 VATTR_NULL(&vattr); 3210 vattr.va_size = SCARG(uap, length); 3211 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3212 } 3213 VOP_UNLOCK(vp, 0); 3214 vn_finished_write(mp, 0); 3215 out: 3216 FILE_UNUSE(fp, l); 3217 return (error); 3218 } 3219 3220 /* 3221 * Sync an open file. 3222 */ 3223 /* ARGSUSED */ 3224 int 3225 sys_fsync(struct lwp *l, void *v, register_t *retval) 3226 { 3227 struct sys_fsync_args /* { 3228 syscallarg(int) fd; 3229 } */ *uap = v; 3230 struct proc *p = l->l_proc; 3231 struct vnode *vp; 3232 struct mount *mp; 3233 struct file *fp; 3234 int error; 3235 3236 /* getvnode() will use the descriptor for us */ 3237 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3238 return (error); 3239 vp = (struct vnode *)fp->f_data; 3240 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3241 FILE_UNUSE(fp, l); 3242 return (error); 3243 } 3244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3245 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3246 if (error == 0 && bioops.io_fsync != NULL && 3247 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3248 (*bioops.io_fsync)(vp, 0); 3249 VOP_UNLOCK(vp, 0); 3250 vn_finished_write(mp, 0); 3251 FILE_UNUSE(fp, l); 3252 return (error); 3253 } 3254 3255 /* 3256 * Sync a range of file data. API modeled after that found in AIX. 3257 * 3258 * FDATASYNC indicates that we need only save enough metadata to be able 3259 * to re-read the written data. Note we duplicate AIX's requirement that 3260 * the file be open for writing. 3261 */ 3262 /* ARGSUSED */ 3263 int 3264 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3265 { 3266 struct sys_fsync_range_args /* { 3267 syscallarg(int) fd; 3268 syscallarg(int) flags; 3269 syscallarg(off_t) start; 3270 syscallarg(off_t) length; 3271 } */ *uap = v; 3272 struct proc *p = l->l_proc; 3273 struct vnode *vp; 3274 struct file *fp; 3275 int flags, nflags; 3276 off_t s, e, len; 3277 int error; 3278 3279 /* getvnode() will use the descriptor for us */ 3280 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3281 return (error); 3282 3283 if ((fp->f_flag & FWRITE) == 0) { 3284 FILE_UNUSE(fp, l); 3285 return (EBADF); 3286 } 3287 3288 flags = SCARG(uap, flags); 3289 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3290 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3291 return (EINVAL); 3292 } 3293 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3294 if (flags & FDATASYNC) 3295 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3296 else 3297 nflags = FSYNC_WAIT; 3298 if (flags & FDISKSYNC) 3299 nflags |= FSYNC_CACHE; 3300 3301 len = SCARG(uap, length); 3302 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3303 if (len) { 3304 s = SCARG(uap, start); 3305 e = s + len; 3306 if (e < s) { 3307 FILE_UNUSE(fp, l); 3308 return (EINVAL); 3309 } 3310 } else { 3311 e = 0; 3312 s = 0; 3313 } 3314 3315 vp = (struct vnode *)fp->f_data; 3316 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3317 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3318 3319 if (error == 0 && bioops.io_fsync != NULL && 3320 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3321 (*bioops.io_fsync)(vp, nflags); 3322 3323 VOP_UNLOCK(vp, 0); 3324 FILE_UNUSE(fp, l); 3325 return (error); 3326 } 3327 3328 /* 3329 * Sync the data of an open file. 3330 */ 3331 /* ARGSUSED */ 3332 int 3333 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3334 { 3335 struct sys_fdatasync_args /* { 3336 syscallarg(int) fd; 3337 } */ *uap = v; 3338 struct proc *p = l->l_proc; 3339 struct vnode *vp; 3340 struct file *fp; 3341 int error; 3342 3343 /* getvnode() will use the descriptor for us */ 3344 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3345 return (error); 3346 if ((fp->f_flag & FWRITE) == 0) { 3347 FILE_UNUSE(fp, l); 3348 return (EBADF); 3349 } 3350 vp = (struct vnode *)fp->f_data; 3351 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3352 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3353 VOP_UNLOCK(vp, 0); 3354 FILE_UNUSE(fp, l); 3355 return (error); 3356 } 3357 3358 /* 3359 * Rename files, (standard) BSD semantics frontend. 3360 */ 3361 /* ARGSUSED */ 3362 int 3363 sys_rename(struct lwp *l, void *v, register_t *retval) 3364 { 3365 struct sys_rename_args /* { 3366 syscallarg(const char *) from; 3367 syscallarg(const char *) to; 3368 } */ *uap = v; 3369 3370 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3371 } 3372 3373 /* 3374 * Rename files, POSIX semantics frontend. 3375 */ 3376 /* ARGSUSED */ 3377 int 3378 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3379 { 3380 struct sys___posix_rename_args /* { 3381 syscallarg(const char *) from; 3382 syscallarg(const char *) to; 3383 } */ *uap = v; 3384 3385 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3386 } 3387 3388 /* 3389 * Rename files. Source and destination must either both be directories, 3390 * or both not be directories. If target is a directory, it must be empty. 3391 * If `from' and `to' refer to the same object, the value of the `retain' 3392 * argument is used to determine whether `from' will be 3393 * 3394 * (retain == 0) deleted unless `from' and `to' refer to the same 3395 * object in the file system's name space (BSD). 3396 * (retain == 1) always retained (POSIX). 3397 */ 3398 static int 3399 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3400 { 3401 struct mount *mp = NULL; 3402 struct vnode *tvp, *fvp, *tdvp; 3403 struct nameidata fromnd, tond; 3404 struct proc *p; 3405 int error; 3406 3407 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3408 from, l); 3409 if ((error = namei(&fromnd)) != 0) 3410 return (error); 3411 if (fromnd.ni_dvp != fromnd.ni_vp) 3412 VOP_UNLOCK(fromnd.ni_dvp, 0); 3413 fvp = fromnd.ni_vp; 3414 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3415 if (error != 0) { 3416 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3417 vrele(fromnd.ni_dvp); 3418 vrele(fvp); 3419 if (fromnd.ni_startdir) 3420 vrele(fromnd.ni_startdir); 3421 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3422 return (error); 3423 } 3424 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3425 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3426 if ((error = namei(&tond)) != 0) { 3427 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3428 vrele(fromnd.ni_dvp); 3429 vrele(fvp); 3430 goto out1; 3431 } 3432 tdvp = tond.ni_dvp; 3433 tvp = tond.ni_vp; 3434 3435 if (tvp != NULL) { 3436 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3437 error = ENOTDIR; 3438 goto out; 3439 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3440 error = EISDIR; 3441 goto out; 3442 } 3443 } 3444 3445 if (fvp == tdvp) 3446 error = EINVAL; 3447 3448 /* 3449 * Source and destination refer to the same object. 3450 */ 3451 if (fvp == tvp) { 3452 if (retain) 3453 error = -1; 3454 else if (fromnd.ni_dvp == tdvp && 3455 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3456 !memcmp(fromnd.ni_cnd.cn_nameptr, 3457 tond.ni_cnd.cn_nameptr, 3458 fromnd.ni_cnd.cn_namelen)) 3459 error = -1; 3460 } 3461 3462 #if NVERIEXEC > 0 3463 if (!error) { 3464 pathname_t frompath, topath; 3465 3466 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3467 &frompath); 3468 if (!error) 3469 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3470 &topath); 3471 if (!error) 3472 error = veriexec_renamechk(fvp, pathname_path(frompath), 3473 tvp, pathname_path(topath), l); 3474 3475 pathname_put(frompath); 3476 pathname_put(topath); 3477 } 3478 #endif /* NVERIEXEC > 0 */ 3479 3480 out: 3481 p = l->l_proc; 3482 if (!error) { 3483 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3484 if (fromnd.ni_dvp != tdvp) 3485 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3486 if (tvp) { 3487 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3488 } 3489 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3490 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3491 } else { 3492 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3493 if (tdvp == tvp) 3494 vrele(tdvp); 3495 else 3496 vput(tdvp); 3497 if (tvp) 3498 vput(tvp); 3499 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3500 vrele(fromnd.ni_dvp); 3501 vrele(fvp); 3502 } 3503 vrele(tond.ni_startdir); 3504 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3505 out1: 3506 vn_finished_write(mp, 0); 3507 if (fromnd.ni_startdir) 3508 vrele(fromnd.ni_startdir); 3509 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3510 return (error == -1 ? 0 : error); 3511 } 3512 3513 /* 3514 * Make a directory file. 3515 */ 3516 /* ARGSUSED */ 3517 int 3518 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3519 { 3520 struct sys_mkdir_args /* { 3521 syscallarg(const char *) path; 3522 syscallarg(int) mode; 3523 } */ *uap = v; 3524 struct proc *p = l->l_proc; 3525 struct mount *mp; 3526 struct vnode *vp; 3527 struct vattr vattr; 3528 int error; 3529 struct nameidata nd; 3530 3531 restart: 3532 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3533 SCARG(uap, path), l); 3534 if ((error = namei(&nd)) != 0) 3535 return (error); 3536 vp = nd.ni_vp; 3537 if (vp != NULL) { 3538 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3539 if (nd.ni_dvp == vp) 3540 vrele(nd.ni_dvp); 3541 else 3542 vput(nd.ni_dvp); 3543 vrele(vp); 3544 return (EEXIST); 3545 } 3546 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3547 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3548 if (nd.ni_dvp == vp) 3549 vrele(nd.ni_dvp); 3550 else 3551 vput(nd.ni_dvp); 3552 if ((error = vn_start_write(NULL, &mp, 3553 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3554 return (error); 3555 goto restart; 3556 } 3557 VATTR_NULL(&vattr); 3558 vattr.va_type = VDIR; 3559 vattr.va_mode = 3560 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3561 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3562 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3563 if (!error) 3564 vput(nd.ni_vp); 3565 vn_finished_write(mp, 0); 3566 return (error); 3567 } 3568 3569 /* 3570 * Remove a directory file. 3571 */ 3572 /* ARGSUSED */ 3573 int 3574 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3575 { 3576 struct sys_rmdir_args /* { 3577 syscallarg(const char *) path; 3578 } */ *uap = v; 3579 struct mount *mp; 3580 struct vnode *vp; 3581 int error; 3582 struct nameidata nd; 3583 3584 restart: 3585 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3586 SCARG(uap, path), l); 3587 if ((error = namei(&nd)) != 0) 3588 return (error); 3589 vp = nd.ni_vp; 3590 if (vp->v_type != VDIR) { 3591 error = ENOTDIR; 3592 goto out; 3593 } 3594 /* 3595 * No rmdir "." please. 3596 */ 3597 if (nd.ni_dvp == vp) { 3598 error = EINVAL; 3599 goto out; 3600 } 3601 /* 3602 * The root of a mounted filesystem cannot be deleted. 3603 */ 3604 if (vp->v_flag & VROOT) { 3605 error = EBUSY; 3606 goto out; 3607 } 3608 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3609 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3610 if (nd.ni_dvp == vp) 3611 vrele(nd.ni_dvp); 3612 else 3613 vput(nd.ni_dvp); 3614 vput(vp); 3615 if ((error = vn_start_write(NULL, &mp, 3616 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3617 return (error); 3618 goto restart; 3619 } 3620 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3621 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3622 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3623 vn_finished_write(mp, 0); 3624 return (error); 3625 3626 out: 3627 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3628 if (nd.ni_dvp == vp) 3629 vrele(nd.ni_dvp); 3630 else 3631 vput(nd.ni_dvp); 3632 vput(vp); 3633 return (error); 3634 } 3635 3636 /* 3637 * Read a block of directory entries in a file system independent format. 3638 */ 3639 int 3640 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3641 { 3642 struct sys___getdents30_args /* { 3643 syscallarg(int) fd; 3644 syscallarg(char *) buf; 3645 syscallarg(size_t) count; 3646 } */ *uap = v; 3647 struct proc *p = l->l_proc; 3648 struct file *fp; 3649 int error, done; 3650 3651 /* getvnode() will use the descriptor for us */ 3652 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3653 return (error); 3654 if ((fp->f_flag & FREAD) == 0) { 3655 error = EBADF; 3656 goto out; 3657 } 3658 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3659 SCARG(uap, count), &done, l, 0, 0); 3660 #ifdef KTRACE 3661 if (!error && KTRPOINT(p, KTR_GENIO)) { 3662 struct iovec iov; 3663 iov.iov_base = SCARG(uap, buf); 3664 iov.iov_len = done; 3665 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3666 } 3667 #endif 3668 *retval = done; 3669 out: 3670 FILE_UNUSE(fp, l); 3671 return (error); 3672 } 3673 3674 /* 3675 * Set the mode mask for creation of filesystem nodes. 3676 */ 3677 int 3678 sys_umask(struct lwp *l, void *v, register_t *retval) 3679 { 3680 struct sys_umask_args /* { 3681 syscallarg(mode_t) newmask; 3682 } */ *uap = v; 3683 struct proc *p = l->l_proc; 3684 struct cwdinfo *cwdi; 3685 3686 cwdi = p->p_cwdi; 3687 *retval = cwdi->cwdi_cmask; 3688 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3689 return (0); 3690 } 3691 3692 /* 3693 * Void all references to file by ripping underlying filesystem 3694 * away from vnode. 3695 */ 3696 /* ARGSUSED */ 3697 int 3698 sys_revoke(struct lwp *l, void *v, register_t *retval) 3699 { 3700 struct sys_revoke_args /* { 3701 syscallarg(const char *) path; 3702 } */ *uap = v; 3703 struct mount *mp; 3704 struct vnode *vp; 3705 struct vattr vattr; 3706 int error; 3707 struct nameidata nd; 3708 3709 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3710 if ((error = namei(&nd)) != 0) 3711 return (error); 3712 vp = nd.ni_vp; 3713 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3714 goto out; 3715 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3716 (error = kauth_authorize_generic(l->l_cred, 3717 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 3718 goto out; 3719 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3720 goto out; 3721 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3722 VOP_REVOKE(vp, REVOKEALL); 3723 vn_finished_write(mp, 0); 3724 out: 3725 vrele(vp); 3726 return (error); 3727 } 3728 3729 /* 3730 * Convert a user file descriptor to a kernel file entry. 3731 */ 3732 int 3733 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3734 { 3735 struct vnode *vp; 3736 struct file *fp; 3737 3738 if ((fp = fd_getfile(fdp, fd)) == NULL) 3739 return (EBADF); 3740 3741 FILE_USE(fp); 3742 3743 if (fp->f_type != DTYPE_VNODE) { 3744 FILE_UNUSE(fp, NULL); 3745 return (EINVAL); 3746 } 3747 3748 vp = (struct vnode *)fp->f_data; 3749 if (vp->v_type == VBAD) { 3750 FILE_UNUSE(fp, NULL); 3751 return (EBADF); 3752 } 3753 3754 *fpp = fp; 3755 return (0); 3756 } 3757
Indexes created Fri Sep 26 20:09:58 GMT 2025