vfs_syscalls.c revision 1.287
1 /* $NetBSD: vfs_syscalls.c,v 1.287 2006/12/27 08:55:35 yamt 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 1.287 2006/12/27 08:55:35 yamt 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 * Copy file-system type from userspace. 310 */ 311 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 312 if (error) { 313 #if defined(COMPAT_09) || defined(COMPAT_43) 314 /* 315 * Historically, filesystem types were identified by numbers. 316 * If we get an integer for the filesystem type instead of a 317 * string, we check to see if it matches one of the historic 318 * filesystem types. 319 */ 320 u_long fsindex = (u_long)fstype; 321 if (fsindex >= nmountcompatnames || 322 mountcompatnames[fsindex] == NULL) { 323 error = ENODEV; 324 vput(vp); 325 goto out; 326 } 327 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 328 #else 329 vput(vp); 330 goto out; 331 #endif 332 } 333 334 #ifdef COMPAT_10 335 /* Accept `ufs' as an alias for `ffs'. */ 336 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 337 strlcpy(fstypename, "ffs", sizeof(fstypename)); 338 #endif 339 340 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 341 vput(vp); 342 goto out; 343 } 344 345 /* 346 * Check if a file-system is not already mounted on this vnode. 347 */ 348 if (vp->v_mountedhere != NULL) { 349 error = EBUSY; 350 vput(vp); 351 goto out; 352 } 353 354 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 355 356 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 357 free(mp, M_MOUNT); 358 error = ENODEV; 359 vput(vp); 360 goto out; 361 } 362 363 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 364 simple_lock_init(&mp->mnt_slock); 365 (void)vfs_busy(mp, LK_NOWAIT, 0); 366 367 mp->mnt_op->vfs_refcount++; 368 mp->mnt_vnodecovered = vp; 369 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 370 mp->mnt_unmounter = NULL; 371 mp->mnt_leaf = mp; 372 mount_initspecific(mp); 373 374 /* 375 * The underlying file system may refuse the mount for 376 * various reasons. Allow the user to force it to happen. 377 * 378 * Set the mount level flags. 379 */ 380 mp->mnt_flag = flags & 381 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 382 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 383 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 384 MNT_IGNORE | MNT_RDONLY); 385 386 error = VFS_MOUNT(mp, path, data, ndp, l); 387 mp->mnt_flag &= ~MNT_OP_FLAGS; 388 389 /* 390 * Put the new filesystem on the mount list after root. 391 */ 392 cache_purge(vp); 393 if (!error) { 394 mp->mnt_iflag &= ~IMNT_WANTRDWR; 395 vp->v_mountedhere = mp; 396 simple_lock(&mountlist_slock); 397 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 398 simple_unlock(&mountlist_slock); 399 checkdirs(vp); 400 VOP_UNLOCK(vp, 0); 401 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 402 error = vfs_allocate_syncvnode(mp); 403 vfs_unbusy(mp); 404 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 405 error = VFS_START(mp, 0, l); 406 if (error) 407 vrele(vp); 408 } else { 409 vp->v_mountedhere = NULL; 410 mp->mnt_op->vfs_refcount--; 411 vfs_unbusy(mp); 412 free(mp, M_MOUNT); 413 vput(vp); 414 } 415 416 out: 417 return (error); 418 } 419 420 static int 421 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 422 void *data, struct nameidata *ndp) 423 { 424 struct mount *mp; 425 int error; 426 427 mp = vp->v_mount; 428 429 if ((vp->v_flag & VROOT) == 0) { 430 error = EINVAL; 431 goto out; 432 } 433 434 if (vfs_busy(mp, LK_NOWAIT, 0)) { 435 error = EPERM; 436 goto out; 437 } 438 439 mp->mnt_flag &= ~MNT_OP_FLAGS; 440 mp->mnt_flag |= MNT_GETARGS; 441 error = VFS_MOUNT(mp, path, data, ndp, l); 442 mp->mnt_flag &= ~MNT_OP_FLAGS; 443 444 vfs_unbusy(mp); 445 out: 446 return (error); 447 } 448 449 /* ARGSUSED */ 450 int 451 sys_mount(struct lwp *l, void *v, register_t *retval) 452 { 453 struct sys_mount_args /* { 454 syscallarg(const char *) type; 455 syscallarg(const char *) path; 456 syscallarg(int) flags; 457 syscallarg(void *) data; 458 } */ *uap = v; 459 struct vnode *vp; 460 struct nameidata nd; 461 int error; 462 463 /* 464 * if MNT_GETARGS is specified, it should be only flag. 465 */ 466 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 467 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 468 return EINVAL; 469 } 470 471 /* XXX secmodel stuff. */ 472 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 473 (error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 474 &l->l_acflag))) 475 return (error); 476 477 /* 478 * Get vnode to be covered 479 */ 480 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 481 SCARG(uap, path), l); 482 if ((error = namei(&nd)) != 0) 483 return (error); 484 vp = nd.ni_vp; 485 486 /* 487 * A lookup in VFS_MOUNT might result in an attempt to 488 * lock this vnode again, so make the lock recursive. 489 */ 490 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 491 492 if (SCARG(uap, flags) == MNT_GETARGS) { 493 error = mount_getargs(l, vp, SCARG(uap, path), 494 SCARG(uap, flags), SCARG(uap, data), &nd); 495 vput(vp); 496 } else if (SCARG(uap, flags) & MNT_UPDATE) { 497 error = mount_update(l, vp, SCARG(uap, path), 498 SCARG(uap, flags), SCARG(uap, data), &nd); 499 vput(vp); 500 } else { 501 /* Locking is handled internally in mount_domount(). */ 502 error = mount_domount(l, vp, SCARG(uap, type), 503 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 504 } 505 506 return (error); 507 } 508 509 /* 510 * Scan all active processes to see if any of them have a current 511 * or root directory onto which the new filesystem has just been 512 * mounted. If so, replace them with the new mount point. 513 */ 514 void 515 checkdirs(struct vnode *olddp) 516 { 517 struct cwdinfo *cwdi; 518 struct vnode *newdp; 519 struct proc *p; 520 521 if (olddp->v_usecount == 1) 522 return; 523 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 524 panic("mount: lost mount"); 525 proclist_lock_read(); 526 PROCLIST_FOREACH(p, &allproc) { 527 cwdi = p->p_cwdi; 528 if (!cwdi) 529 continue; 530 if (cwdi->cwdi_cdir == olddp) { 531 vrele(cwdi->cwdi_cdir); 532 VREF(newdp); 533 cwdi->cwdi_cdir = newdp; 534 } 535 if (cwdi->cwdi_rdir == olddp) { 536 vrele(cwdi->cwdi_rdir); 537 VREF(newdp); 538 cwdi->cwdi_rdir = newdp; 539 } 540 } 541 proclist_unlock_read(); 542 if (rootvnode == olddp) { 543 vrele(rootvnode); 544 VREF(newdp); 545 rootvnode = newdp; 546 } 547 vput(newdp); 548 } 549 550 /* 551 * Unmount a file system. 552 * 553 * Note: unmount takes a path to the vnode mounted on as argument, 554 * not special file (as before). 555 */ 556 /* ARGSUSED */ 557 int 558 sys_unmount(struct lwp *l, void *v, register_t *retval) 559 { 560 struct sys_unmount_args /* { 561 syscallarg(const char *) path; 562 syscallarg(int) flags; 563 } */ *uap = v; 564 struct vnode *vp; 565 struct mount *mp; 566 int error; 567 struct nameidata nd; 568 569 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 570 SCARG(uap, path), l); 571 if ((error = namei(&nd)) != 0) 572 return (error); 573 vp = nd.ni_vp; 574 mp = vp->v_mount; 575 576 /* 577 * Only root, or the user that did the original mount is 578 * permitted to unmount this filesystem. 579 */ 580 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred)) && 581 (error = kauth_authorize_generic(l->l_cred, 582 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 583 vput(vp); 584 return (error); 585 } 586 587 /* 588 * Don't allow unmounting the root file system. 589 */ 590 if (mp->mnt_flag & MNT_ROOTFS) { 591 vput(vp); 592 return (EINVAL); 593 } 594 595 /* 596 * Must be the root of the filesystem 597 */ 598 if ((vp->v_flag & VROOT) == 0) { 599 vput(vp); 600 return (EINVAL); 601 } 602 vput(vp); 603 604 /* 605 * XXX Freeze syncer. Must do this before locking the 606 * mount point. See dounmount() for details. 607 */ 608 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 609 610 if (vfs_busy(mp, 0, 0)) { 611 lockmgr(&syncer_lock, LK_RELEASE, NULL); 612 return (EBUSY); 613 } 614 615 return (dounmount(mp, SCARG(uap, flags), l)); 616 } 617 618 /* 619 * Do the actual file system unmount. File system is assumed to have been 620 * marked busy by the caller. 621 */ 622 int 623 dounmount(struct mount *mp, int flags, struct lwp *l) 624 { 625 struct vnode *coveredvp; 626 int error; 627 int async; 628 int used_syncer; 629 630 #if NVERIEXEC > 0 631 error = veriexec_unmountchk(mp); 632 if (error) 633 return (error); 634 #endif /* NVERIEXEC > 0 */ 635 636 simple_lock(&mountlist_slock); 637 vfs_unbusy(mp); 638 used_syncer = (mp->mnt_syncer != NULL); 639 640 /* 641 * XXX Syncer must be frozen when we get here. This should really 642 * be done on a per-mountpoint basis, but especially the softdep 643 * code possibly called from the syncer doesn't exactly work on a 644 * per-mountpoint basis, so the softdep code would become a maze 645 * of vfs_busy() calls. 646 * 647 * The caller of dounmount() must acquire syncer_lock because 648 * the syncer itself acquires locks in syncer_lock -> vfs_busy 649 * order, and we must preserve that order to avoid deadlock. 650 * 651 * So, if the file system did not use the syncer, now is 652 * the time to release the syncer_lock. 653 */ 654 if (used_syncer == 0) 655 lockmgr(&syncer_lock, LK_RELEASE, NULL); 656 657 mp->mnt_iflag |= IMNT_UNMOUNT; 658 mp->mnt_unmounter = l; 659 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 660 vn_start_write(NULL, &mp, V_WAIT); 661 662 async = mp->mnt_flag & MNT_ASYNC; 663 mp->mnt_flag &= ~MNT_ASYNC; 664 cache_purgevfs(mp); /* remove cache entries for this file sys */ 665 if (mp->mnt_syncer != NULL) 666 vfs_deallocate_syncvnode(mp); 667 error = 0; 668 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 669 #if NFSS > 0 670 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 671 #endif 672 if (error == 0) 673 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 674 } 675 if (error == 0 || (flags & MNT_FORCE)) 676 error = VFS_UNMOUNT(mp, flags, l); 677 vn_finished_write(mp, 0); 678 simple_lock(&mountlist_slock); 679 if (error) { 680 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 681 (void) vfs_allocate_syncvnode(mp); 682 mp->mnt_iflag &= ~IMNT_UNMOUNT; 683 mp->mnt_unmounter = NULL; 684 mp->mnt_flag |= async; 685 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 686 &mountlist_slock); 687 if (used_syncer) 688 lockmgr(&syncer_lock, LK_RELEASE, NULL); 689 simple_lock(&mp->mnt_slock); 690 while (mp->mnt_wcnt > 0) { 691 wakeup(mp); 692 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 693 0, &mp->mnt_slock); 694 } 695 simple_unlock(&mp->mnt_slock); 696 return (error); 697 } 698 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 699 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 700 coveredvp->v_mountedhere = NULL; 701 vrele(coveredvp); 702 } 703 mp->mnt_op->vfs_refcount--; 704 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 705 panic("unmount: dangling vnode"); 706 mp->mnt_iflag |= IMNT_GONE; 707 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 708 mount_finispecific(mp); 709 if (used_syncer) 710 lockmgr(&syncer_lock, LK_RELEASE, NULL); 711 simple_lock(&mp->mnt_slock); 712 while (mp->mnt_wcnt > 0) { 713 wakeup(mp); 714 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 715 } 716 simple_unlock(&mp->mnt_slock); 717 vfs_hooks_unmount(mp); 718 free(mp, M_MOUNT); 719 return (0); 720 } 721 722 /* 723 * Sync each mounted filesystem. 724 */ 725 #ifdef DEBUG 726 int syncprt = 0; 727 struct ctldebug debug0 = { "syncprt", &syncprt }; 728 #endif 729 730 /* ARGSUSED */ 731 int 732 sys_sync(struct lwp *l, void *v, register_t *retval) 733 { 734 struct mount *mp, *nmp; 735 int asyncflag; 736 737 if (l == NULL) 738 l = &lwp0; 739 740 simple_lock(&mountlist_slock); 741 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 742 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 743 nmp = mp->mnt_list.cqe_prev; 744 continue; 745 } 746 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 747 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 748 asyncflag = mp->mnt_flag & MNT_ASYNC; 749 mp->mnt_flag &= ~MNT_ASYNC; 750 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 751 if (asyncflag) 752 mp->mnt_flag |= MNT_ASYNC; 753 vn_finished_write(mp, 0); 754 } 755 simple_lock(&mountlist_slock); 756 nmp = mp->mnt_list.cqe_prev; 757 vfs_unbusy(mp); 758 759 } 760 simple_unlock(&mountlist_slock); 761 #ifdef DEBUG 762 if (syncprt) 763 vfs_bufstats(); 764 #endif /* DEBUG */ 765 return (0); 766 } 767 768 /* 769 * Change filesystem quotas. 770 */ 771 /* ARGSUSED */ 772 int 773 sys_quotactl(struct lwp *l, void *v, register_t *retval) 774 { 775 struct sys_quotactl_args /* { 776 syscallarg(const char *) path; 777 syscallarg(int) cmd; 778 syscallarg(int) uid; 779 syscallarg(void *) arg; 780 } */ *uap = v; 781 struct mount *mp; 782 int error; 783 struct nameidata nd; 784 785 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 786 if ((error = namei(&nd)) != 0) 787 return (error); 788 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 789 vrele(nd.ni_vp); 790 if (error) 791 return (error); 792 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 793 SCARG(uap, arg), l); 794 vn_finished_write(mp, 0); 795 return (error); 796 } 797 798 int 799 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 800 int root) 801 { 802 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 803 int error = 0; 804 805 /* 806 * If MNT_NOWAIT or MNT_LAZY is specified, do not 807 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 808 * overrides MNT_NOWAIT. 809 */ 810 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 811 (flags != MNT_WAIT && flags != 0)) { 812 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 813 goto done; 814 } 815 816 /* Get the filesystem stats now */ 817 memset(sp, 0, sizeof(*sp)); 818 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 819 return error; 820 } 821 822 if (cwdi->cwdi_rdir == NULL) 823 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 824 done: 825 if (cwdi->cwdi_rdir != NULL) { 826 size_t len; 827 char *bp; 828 char *path = PNBUF_GET(); 829 if (!path) 830 return ENOMEM; 831 832 bp = path + MAXPATHLEN; 833 *--bp = '\0'; 834 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 835 MAXPATHLEN / 2, 0, l); 836 if (error) { 837 PNBUF_PUT(path); 838 return error; 839 } 840 len = strlen(bp); 841 /* 842 * for mount points that are below our root, we can see 843 * them, so we fix up the pathname and return them. The 844 * rest we cannot see, so we don't allow viewing the 845 * data. 846 */ 847 if (strncmp(bp, sp->f_mntonname, len) == 0) { 848 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 849 sizeof(sp->f_mntonname)); 850 if (sp->f_mntonname[0] == '\0') 851 (void)strlcpy(sp->f_mntonname, "/", 852 sizeof(sp->f_mntonname)); 853 } else { 854 if (root) 855 (void)strlcpy(sp->f_mntonname, "/", 856 sizeof(sp->f_mntonname)); 857 else 858 error = EPERM; 859 } 860 PNBUF_PUT(path); 861 } 862 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 863 return error; 864 } 865 866 /* 867 * Get filesystem statistics. 868 */ 869 /* ARGSUSED */ 870 int 871 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 872 { 873 struct sys_statvfs1_args /* { 874 syscallarg(const char *) path; 875 syscallarg(struct statvfs *) buf; 876 syscallarg(int) flags; 877 } */ *uap = v; 878 struct mount *mp; 879 struct statvfs *sb; 880 int error; 881 struct nameidata nd; 882 883 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 884 if ((error = namei(&nd)) != 0) 885 return error; 886 mp = nd.ni_vp->v_mount; 887 vrele(nd.ni_vp); 888 sb = STATVFSBUF_GET(); 889 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 890 if (error == 0) { 891 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 892 } 893 STATVFSBUF_PUT(sb); 894 return error; 895 } 896 897 /* 898 * Get filesystem statistics. 899 */ 900 /* ARGSUSED */ 901 int 902 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 903 { 904 struct sys_fstatvfs1_args /* { 905 syscallarg(int) fd; 906 syscallarg(struct statvfs *) buf; 907 syscallarg(int) flags; 908 } */ *uap = v; 909 struct proc *p = l->l_proc; 910 struct file *fp; 911 struct mount *mp; 912 struct statvfs *sb; 913 int error; 914 915 /* getvnode() will use the descriptor for us */ 916 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 917 return (error); 918 mp = ((struct vnode *)fp->f_data)->v_mount; 919 sb = STATVFSBUF_GET(); 920 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 921 goto out; 922 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 923 out: 924 FILE_UNUSE(fp, l); 925 STATVFSBUF_PUT(sb); 926 return error; 927 } 928 929 930 /* 931 * Get statistics on all filesystems. 932 */ 933 int 934 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 935 { 936 struct sys_getvfsstat_args /* { 937 syscallarg(struct statvfs *) buf; 938 syscallarg(size_t) bufsize; 939 syscallarg(int) flags; 940 } */ *uap = v; 941 int root = 0; 942 struct proc *p = l->l_proc; 943 struct mount *mp, *nmp; 944 struct statvfs *sb; 945 struct statvfs *sfsp; 946 size_t count, maxcount; 947 int error = 0; 948 949 sb = STATVFSBUF_GET(); 950 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 951 sfsp = SCARG(uap, buf); 952 simple_lock(&mountlist_slock); 953 count = 0; 954 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 955 mp = nmp) { 956 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 957 nmp = CIRCLEQ_NEXT(mp, mnt_list); 958 continue; 959 } 960 if (sfsp && count < maxcount) { 961 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 962 if (error) { 963 simple_lock(&mountlist_slock); 964 nmp = CIRCLEQ_NEXT(mp, mnt_list); 965 vfs_unbusy(mp); 966 continue; 967 } 968 error = copyout(sb, sfsp, sizeof(*sfsp)); 969 if (error) { 970 vfs_unbusy(mp); 971 goto out; 972 } 973 sfsp++; 974 root |= strcmp(sb->f_mntonname, "/") == 0; 975 } 976 count++; 977 simple_lock(&mountlist_slock); 978 nmp = CIRCLEQ_NEXT(mp, mnt_list); 979 vfs_unbusy(mp); 980 } 981 simple_unlock(&mountlist_slock); 982 if (root == 0 && p->p_cwdi->cwdi_rdir) { 983 /* 984 * fake a root entry 985 */ 986 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 987 SCARG(uap, flags), 1)) != 0) 988 goto out; 989 if (sfsp) 990 error = copyout(sb, sfsp, sizeof(*sfsp)); 991 count++; 992 } 993 if (sfsp && count > maxcount) 994 *retval = maxcount; 995 else 996 *retval = count; 997 out: 998 STATVFSBUF_PUT(sb); 999 return error; 1000 } 1001 1002 /* 1003 * Change current working directory to a given file descriptor. 1004 */ 1005 /* ARGSUSED */ 1006 int 1007 sys_fchdir(struct lwp *l, void *v, register_t *retval) 1008 { 1009 struct sys_fchdir_args /* { 1010 syscallarg(int) fd; 1011 } */ *uap = v; 1012 struct proc *p = l->l_proc; 1013 struct filedesc *fdp = p->p_fd; 1014 struct cwdinfo *cwdi = p->p_cwdi; 1015 struct vnode *vp, *tdp; 1016 struct mount *mp; 1017 struct file *fp; 1018 int error; 1019 1020 /* getvnode() will use the descriptor for us */ 1021 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1022 return (error); 1023 vp = (struct vnode *)fp->f_data; 1024 1025 VREF(vp); 1026 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1027 if (vp->v_type != VDIR) 1028 error = ENOTDIR; 1029 else 1030 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1031 while (!error && (mp = vp->v_mountedhere) != NULL) { 1032 if (vfs_busy(mp, 0, 0)) 1033 continue; 1034 error = VFS_ROOT(mp, &tdp); 1035 vfs_unbusy(mp); 1036 if (error) 1037 break; 1038 vput(vp); 1039 vp = tdp; 1040 } 1041 if (error) { 1042 vput(vp); 1043 goto out; 1044 } 1045 VOP_UNLOCK(vp, 0); 1046 1047 /* 1048 * Disallow changing to a directory not under the process's 1049 * current root directory (if there is one). 1050 */ 1051 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1052 vrele(vp); 1053 error = EPERM; /* operation not permitted */ 1054 goto out; 1055 } 1056 1057 vrele(cwdi->cwdi_cdir); 1058 cwdi->cwdi_cdir = vp; 1059 out: 1060 FILE_UNUSE(fp, l); 1061 return (error); 1062 } 1063 1064 /* 1065 * Change this process's notion of the root directory to a given file 1066 * descriptor. 1067 */ 1068 int 1069 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1070 { 1071 struct sys_fchroot_args *uap = v; 1072 struct proc *p = l->l_proc; 1073 struct filedesc *fdp = p->p_fd; 1074 struct cwdinfo *cwdi = p->p_cwdi; 1075 struct vnode *vp; 1076 struct file *fp; 1077 int error; 1078 1079 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1080 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1081 return error; 1082 /* getvnode() will use the descriptor for us */ 1083 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1084 return error; 1085 vp = (struct vnode *) fp->f_data; 1086 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1087 if (vp->v_type != VDIR) 1088 error = ENOTDIR; 1089 else 1090 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1091 VOP_UNLOCK(vp, 0); 1092 if (error) 1093 goto out; 1094 VREF(vp); 1095 1096 /* 1097 * Prevent escaping from chroot by putting the root under 1098 * the working directory. Silently chdir to / if we aren't 1099 * already there. 1100 */ 1101 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1102 /* 1103 * XXX would be more failsafe to change directory to a 1104 * deadfs node here instead 1105 */ 1106 vrele(cwdi->cwdi_cdir); 1107 VREF(vp); 1108 cwdi->cwdi_cdir = vp; 1109 } 1110 1111 if (cwdi->cwdi_rdir != NULL) 1112 vrele(cwdi->cwdi_rdir); 1113 cwdi->cwdi_rdir = vp; 1114 out: 1115 FILE_UNUSE(fp, l); 1116 return (error); 1117 } 1118 1119 /* 1120 * Change current working directory (``.''). 1121 */ 1122 /* ARGSUSED */ 1123 int 1124 sys_chdir(struct lwp *l, void *v, register_t *retval) 1125 { 1126 struct sys_chdir_args /* { 1127 syscallarg(const char *) path; 1128 } */ *uap = v; 1129 struct proc *p = l->l_proc; 1130 struct cwdinfo *cwdi = p->p_cwdi; 1131 int error; 1132 struct nameidata nd; 1133 1134 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1135 SCARG(uap, path), l); 1136 if ((error = change_dir(&nd, l)) != 0) 1137 return (error); 1138 vrele(cwdi->cwdi_cdir); 1139 cwdi->cwdi_cdir = nd.ni_vp; 1140 return (0); 1141 } 1142 1143 /* 1144 * Change notion of root (``/'') directory. 1145 */ 1146 /* ARGSUSED */ 1147 int 1148 sys_chroot(struct lwp *l, void *v, register_t *retval) 1149 { 1150 struct sys_chroot_args /* { 1151 syscallarg(const char *) path; 1152 } */ *uap = v; 1153 struct proc *p = l->l_proc; 1154 struct cwdinfo *cwdi = p->p_cwdi; 1155 struct vnode *vp; 1156 int error; 1157 struct nameidata nd; 1158 1159 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1160 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1161 return (error); 1162 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1163 SCARG(uap, path), l); 1164 if ((error = change_dir(&nd, l)) != 0) 1165 return (error); 1166 if (cwdi->cwdi_rdir != NULL) 1167 vrele(cwdi->cwdi_rdir); 1168 vp = nd.ni_vp; 1169 cwdi->cwdi_rdir = vp; 1170 1171 /* 1172 * Prevent escaping from chroot by putting the root under 1173 * the working directory. Silently chdir to / if we aren't 1174 * already there. 1175 */ 1176 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1177 /* 1178 * XXX would be more failsafe to change directory to a 1179 * deadfs node here instead 1180 */ 1181 vrele(cwdi->cwdi_cdir); 1182 VREF(vp); 1183 cwdi->cwdi_cdir = vp; 1184 } 1185 1186 return (0); 1187 } 1188 1189 /* 1190 * Common routine for chroot and chdir. 1191 */ 1192 static int 1193 change_dir(struct nameidata *ndp, struct lwp *l) 1194 { 1195 struct vnode *vp; 1196 int error; 1197 1198 if ((error = namei(ndp)) != 0) 1199 return (error); 1200 vp = ndp->ni_vp; 1201 if (vp->v_type != VDIR) 1202 error = ENOTDIR; 1203 else 1204 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1205 1206 if (error) 1207 vput(vp); 1208 else 1209 VOP_UNLOCK(vp, 0); 1210 return (error); 1211 } 1212 1213 /* 1214 * Check permissions, allocate an open file structure, 1215 * and call the device open routine if any. 1216 */ 1217 int 1218 sys_open(struct lwp *l, void *v, register_t *retval) 1219 { 1220 struct sys_open_args /* { 1221 syscallarg(const char *) path; 1222 syscallarg(int) flags; 1223 syscallarg(int) mode; 1224 } */ *uap = v; 1225 struct proc *p = l->l_proc; 1226 struct cwdinfo *cwdi = p->p_cwdi; 1227 struct filedesc *fdp = p->p_fd; 1228 struct file *fp; 1229 struct vnode *vp; 1230 int flags, cmode; 1231 int type, indx, error; 1232 struct flock lf; 1233 struct nameidata nd; 1234 1235 flags = FFLAGS(SCARG(uap, flags)); 1236 if ((flags & (FREAD | FWRITE)) == 0) 1237 return (EINVAL); 1238 /* falloc() will use the file descriptor for us */ 1239 if ((error = falloc(l, &fp, &indx)) != 0) 1240 return (error); 1241 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1242 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1243 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1244 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1245 FILE_UNUSE(fp, l); 1246 fdp->fd_ofiles[indx] = NULL; 1247 ffree(fp); 1248 if ((error == EDUPFD || error == EMOVEFD) && 1249 l->l_dupfd >= 0 && /* XXX from fdopen */ 1250 (error = 1251 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1252 *retval = indx; 1253 return (0); 1254 } 1255 if (error == ERESTART) 1256 error = EINTR; 1257 fdremove(fdp, indx); 1258 return (error); 1259 } 1260 l->l_dupfd = 0; 1261 vp = nd.ni_vp; 1262 fp->f_flag = flags & FMASK; 1263 fp->f_type = DTYPE_VNODE; 1264 fp->f_ops = &vnops; 1265 fp->f_data = vp; 1266 if (flags & (O_EXLOCK | O_SHLOCK)) { 1267 lf.l_whence = SEEK_SET; 1268 lf.l_start = 0; 1269 lf.l_len = 0; 1270 if (flags & O_EXLOCK) 1271 lf.l_type = F_WRLCK; 1272 else 1273 lf.l_type = F_RDLCK; 1274 type = F_FLOCK; 1275 if ((flags & FNONBLOCK) == 0) 1276 type |= F_WAIT; 1277 VOP_UNLOCK(vp, 0); 1278 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1279 if (error) { 1280 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1281 FILE_UNUSE(fp, l); 1282 ffree(fp); 1283 fdremove(fdp, indx); 1284 return (error); 1285 } 1286 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1287 fp->f_flag |= FHASLOCK; 1288 } 1289 VOP_UNLOCK(vp, 0); 1290 *retval = indx; 1291 FILE_SET_MATURE(fp); 1292 FILE_UNUSE(fp, l); 1293 return (0); 1294 } 1295 1296 static void 1297 vfs__fhfree(fhandle_t *fhp) 1298 { 1299 size_t fhsize; 1300 1301 if (fhp == NULL) { 1302 return; 1303 } 1304 fhsize = FHANDLE_SIZE(fhp); 1305 kmem_free(fhp, fhsize); 1306 } 1307 1308 /* 1309 * vfs_composefh: compose a filehandle. 1310 */ 1311 1312 int 1313 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1314 { 1315 struct mount *mp; 1316 struct fid *fidp; 1317 int error; 1318 size_t needfhsize; 1319 size_t fidsize; 1320 1321 mp = vp->v_mount; 1322 fidp = NULL; 1323 if (*fh_size < FHANDLE_SIZE_MIN) { 1324 fidsize = 0; 1325 } else { 1326 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1327 if (fhp != NULL) { 1328 memset(fhp, 0, *fh_size); 1329 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1330 fidp = &fhp->fh_fid; 1331 } 1332 } 1333 error = VFS_VPTOFH(vp, fidp, &fidsize); 1334 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1335 if (error == 0 && *fh_size < needfhsize) { 1336 error = E2BIG; 1337 } 1338 *fh_size = needfhsize; 1339 return error; 1340 } 1341 1342 int 1343 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1344 { 1345 struct mount *mp; 1346 fhandle_t *fhp; 1347 size_t fhsize; 1348 size_t fidsize; 1349 int error; 1350 1351 *fhpp = NULL; 1352 mp = vp->v_mount; 1353 fidsize = 0; 1354 error = VFS_VPTOFH(vp, NULL, &fidsize); 1355 KASSERT(error != 0); 1356 if (error != E2BIG) { 1357 goto out; 1358 } 1359 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1360 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1361 if (fhp == NULL) { 1362 error = ENOMEM; 1363 goto out; 1364 } 1365 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1366 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1367 if (error == 0) { 1368 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1369 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1370 *fhpp = fhp; 1371 } else { 1372 kmem_free(fhp, fhsize); 1373 } 1374 out: 1375 return error; 1376 } 1377 1378 void 1379 vfs_composefh_free(fhandle_t *fhp) 1380 { 1381 1382 vfs__fhfree(fhp); 1383 } 1384 1385 /* 1386 * vfs_fhtovp: lookup a vnode by a filehandle. 1387 */ 1388 1389 int 1390 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1391 { 1392 struct mount *mp; 1393 int error; 1394 1395 *vpp = NULL; 1396 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1397 if (mp == NULL) { 1398 error = ESTALE; 1399 goto out; 1400 } 1401 if (mp->mnt_op->vfs_fhtovp == NULL) { 1402 error = EOPNOTSUPP; 1403 goto out; 1404 } 1405 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1406 out: 1407 return error; 1408 } 1409 1410 /* 1411 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1412 * the needed size. 1413 */ 1414 1415 int 1416 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1417 { 1418 fhandle_t *fhp; 1419 int error; 1420 1421 *fhpp = NULL; 1422 if (fhsize > FHANDLE_SIZE_MAX) { 1423 return EINVAL; 1424 } 1425 if (fhsize < FHANDLE_SIZE_MIN) { 1426 return EINVAL; 1427 } 1428 again: 1429 fhp = kmem_alloc(fhsize, KM_SLEEP); 1430 if (fhp == NULL) { 1431 return ENOMEM; 1432 } 1433 error = copyin(ufhp, fhp, fhsize); 1434 if (error == 0) { 1435 /* XXX this check shouldn't be here */ 1436 if (FHANDLE_SIZE(fhp) == fhsize) { 1437 *fhpp = fhp; 1438 return 0; 1439 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1440 /* 1441 * a kludge for nfsv2 padded handles. 1442 */ 1443 size_t sz; 1444 1445 sz = FHANDLE_SIZE(fhp); 1446 kmem_free(fhp, fhsize); 1447 fhsize = sz; 1448 goto again; 1449 } else { 1450 /* 1451 * userland told us wrong size. 1452 */ 1453 error = EINVAL; 1454 } 1455 } 1456 kmem_free(fhp, fhsize); 1457 return error; 1458 } 1459 1460 void 1461 vfs_copyinfh_free(fhandle_t *fhp) 1462 { 1463 1464 vfs__fhfree(fhp); 1465 } 1466 1467 /* 1468 * Get file handle system call 1469 */ 1470 int 1471 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1472 { 1473 struct sys___getfh30_args /* { 1474 syscallarg(char *) fname; 1475 syscallarg(fhandle_t *) fhp; 1476 syscallarg(size_t *) fh_size; 1477 } */ *uap = v; 1478 struct vnode *vp; 1479 fhandle_t *fh; 1480 int error; 1481 struct nameidata nd; 1482 size_t sz; 1483 size_t usz; 1484 1485 /* 1486 * Must be super user 1487 */ 1488 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1489 0, NULL, NULL, NULL); 1490 if (error) 1491 return (error); 1492 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1493 SCARG(uap, fname), l); 1494 error = namei(&nd); 1495 if (error) 1496 return (error); 1497 vp = nd.ni_vp; 1498 error = vfs_composefh_alloc(vp, &fh); 1499 vput(vp); 1500 if (error != 0) { 1501 goto out; 1502 } 1503 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1504 if (error != 0) { 1505 goto out; 1506 } 1507 sz = FHANDLE_SIZE(fh); 1508 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1509 if (error != 0) { 1510 goto out; 1511 } 1512 if (usz >= sz) { 1513 error = copyout(fh, SCARG(uap, fhp), sz); 1514 } else { 1515 error = E2BIG; 1516 } 1517 out: 1518 vfs_composefh_free(fh); 1519 return (error); 1520 } 1521 1522 /* 1523 * Open a file given a file handle. 1524 * 1525 * Check permissions, allocate an open file structure, 1526 * and call the device open routine if any. 1527 */ 1528 1529 int 1530 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1531 register_t *retval) 1532 { 1533 struct filedesc *fdp = l->l_proc->p_fd; 1534 struct file *fp; 1535 struct vnode *vp = NULL; 1536 struct mount *mp; 1537 kauth_cred_t cred = l->l_cred; 1538 struct file *nfp; 1539 int type, indx, error=0; 1540 struct flock lf; 1541 struct vattr va; 1542 fhandle_t *fh; 1543 int flags; 1544 1545 /* 1546 * Must be super user 1547 */ 1548 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1549 0, NULL, NULL, NULL))) 1550 return (error); 1551 1552 flags = FFLAGS(oflags); 1553 if ((flags & (FREAD | FWRITE)) == 0) 1554 return (EINVAL); 1555 if ((flags & O_CREAT)) 1556 return (EINVAL); 1557 /* falloc() will use the file descriptor for us */ 1558 if ((error = falloc(l, &nfp, &indx)) != 0) 1559 return (error); 1560 fp = nfp; 1561 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1562 if (error != 0) { 1563 goto bad; 1564 } 1565 error = vfs_fhtovp(fh, &vp); 1566 if (error != 0) { 1567 goto bad; 1568 } 1569 1570 /* Now do an effective vn_open */ 1571 1572 if (vp->v_type == VSOCK) { 1573 error = EOPNOTSUPP; 1574 goto bad; 1575 } 1576 if (flags & FREAD) { 1577 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1578 goto bad; 1579 } 1580 if (flags & (FWRITE | O_TRUNC)) { 1581 if (vp->v_type == VDIR) { 1582 error = EISDIR; 1583 goto bad; 1584 } 1585 if ((error = vn_writechk(vp)) != 0 || 1586 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1587 goto bad; 1588 } 1589 if (flags & O_TRUNC) { 1590 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1591 goto bad; 1592 VOP_UNLOCK(vp, 0); /* XXX */ 1593 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1594 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1595 VATTR_NULL(&va); 1596 va.va_size = 0; 1597 error = VOP_SETATTR(vp, &va, cred, l); 1598 vn_finished_write(mp, 0); 1599 if (error) 1600 goto bad; 1601 } 1602 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1603 goto bad; 1604 if (vp->v_type == VREG && 1605 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1606 error = EIO; 1607 goto bad; 1608 } 1609 if (flags & FWRITE) 1610 vp->v_writecount++; 1611 1612 /* done with modified vn_open, now finish what sys_open does. */ 1613 1614 fp->f_flag = flags & FMASK; 1615 fp->f_type = DTYPE_VNODE; 1616 fp->f_ops = &vnops; 1617 fp->f_data = vp; 1618 if (flags & (O_EXLOCK | O_SHLOCK)) { 1619 lf.l_whence = SEEK_SET; 1620 lf.l_start = 0; 1621 lf.l_len = 0; 1622 if (flags & O_EXLOCK) 1623 lf.l_type = F_WRLCK; 1624 else 1625 lf.l_type = F_RDLCK; 1626 type = F_FLOCK; 1627 if ((flags & FNONBLOCK) == 0) 1628 type |= F_WAIT; 1629 VOP_UNLOCK(vp, 0); 1630 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1631 if (error) { 1632 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1633 FILE_UNUSE(fp, l); 1634 ffree(fp); 1635 fdremove(fdp, indx); 1636 return (error); 1637 } 1638 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1639 fp->f_flag |= FHASLOCK; 1640 } 1641 VOP_UNLOCK(vp, 0); 1642 *retval = indx; 1643 FILE_SET_MATURE(fp); 1644 FILE_UNUSE(fp, l); 1645 vfs_copyinfh_free(fh); 1646 return (0); 1647 1648 bad: 1649 FILE_UNUSE(fp, l); 1650 ffree(fp); 1651 fdremove(fdp, indx); 1652 if (vp != NULL) 1653 vput(vp); 1654 vfs_copyinfh_free(fh); 1655 return (error); 1656 } 1657 1658 int 1659 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1660 { 1661 struct sys___fhopen40_args /* { 1662 syscallarg(const void *) fhp; 1663 syscallarg(size_t) fh_size; 1664 syscallarg(int) flags; 1665 } */ *uap = v; 1666 1667 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1668 SCARG(uap, flags), retval); 1669 } 1670 1671 int 1672 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1673 register_t *retval) 1674 { 1675 struct stat sb; 1676 int error; 1677 fhandle_t *fh; 1678 struct vnode *vp; 1679 1680 /* 1681 * Must be super user 1682 */ 1683 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1684 0, NULL, NULL, NULL))) 1685 return (error); 1686 1687 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1688 if (error != 0) { 1689 goto bad; 1690 } 1691 error = vfs_fhtovp(fh, &vp); 1692 if (error != 0) { 1693 goto bad; 1694 } 1695 error = vn_stat(vp, &sb, l); 1696 vput(vp); 1697 if (error) { 1698 goto bad; 1699 } 1700 error = copyout(&sb, sbp, sizeof(sb)); 1701 bad: 1702 vfs_copyinfh_free(fh); 1703 return error; 1704 } 1705 1706 1707 /* ARGSUSED */ 1708 int 1709 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1710 { 1711 struct sys___fhstat40_args /* { 1712 syscallarg(const void *) fhp; 1713 syscallarg(size_t) fh_size; 1714 syscallarg(struct stat *) sb; 1715 } */ *uap = v; 1716 1717 return dofhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), SCARG(uap, sb), 1718 retval); 1719 } 1720 1721 int 1722 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1723 int flags, register_t *retval) 1724 { 1725 struct statvfs *sb = NULL; 1726 fhandle_t *fh; 1727 struct mount *mp; 1728 struct vnode *vp; 1729 int error; 1730 1731 /* 1732 * Must be super user 1733 */ 1734 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1735 0, NULL, NULL, NULL))) 1736 return error; 1737 1738 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1739 if (error != 0) { 1740 goto out; 1741 } 1742 error = vfs_fhtovp(fh, &vp); 1743 if (error != 0) { 1744 goto out; 1745 } 1746 mp = vp->v_mount; 1747 sb = STATVFSBUF_GET(); 1748 if ((error = dostatvfs(mp, sb, l, flags, 1)) != 0) { 1749 vput(vp); 1750 goto out; 1751 } 1752 vput(vp); 1753 error = copyout(sb, buf, sizeof(*sb)); 1754 out: 1755 if (sb != NULL) { 1756 STATVFSBUF_PUT(sb); 1757 } 1758 vfs_copyinfh_free(fh); 1759 return error; 1760 } 1761 1762 /* ARGSUSED */ 1763 int 1764 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1765 { 1766 struct sys___fhstatvfs140_args /* { 1767 syscallarg(const void *) fhp; 1768 syscallarg(size_t) fh_size; 1769 syscallarg(struct statvfs *) buf; 1770 syscallarg(int) flags; 1771 } */ *uap = v; 1772 1773 return dofhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1774 SCARG(uap, buf), SCARG(uap, flags), retval); 1775 } 1776 1777 /* 1778 * Create a special file. 1779 */ 1780 /* ARGSUSED */ 1781 int 1782 sys_mknod(struct lwp *l, void *v, register_t *retval) 1783 { 1784 struct sys_mknod_args /* { 1785 syscallarg(const char *) path; 1786 syscallarg(int) mode; 1787 syscallarg(int) dev; 1788 } */ *uap = v; 1789 struct proc *p = l->l_proc; 1790 struct vnode *vp; 1791 struct mount *mp; 1792 struct vattr vattr; 1793 int error; 1794 int whiteout = 0; 1795 struct nameidata nd; 1796 1797 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1798 0, NULL, NULL, NULL)) != 0) 1799 return (error); 1800 restart: 1801 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1802 if ((error = namei(&nd)) != 0) 1803 return (error); 1804 vp = nd.ni_vp; 1805 if (vp != NULL) 1806 error = EEXIST; 1807 else { 1808 VATTR_NULL(&vattr); 1809 vattr.va_mode = 1810 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1811 vattr.va_rdev = SCARG(uap, dev); 1812 whiteout = 0; 1813 1814 switch (SCARG(uap, mode) & S_IFMT) { 1815 case S_IFMT: /* used by badsect to flag bad sectors */ 1816 vattr.va_type = VBAD; 1817 break; 1818 case S_IFCHR: 1819 vattr.va_type = VCHR; 1820 break; 1821 case S_IFBLK: 1822 vattr.va_type = VBLK; 1823 break; 1824 case S_IFWHT: 1825 whiteout = 1; 1826 break; 1827 default: 1828 error = EINVAL; 1829 break; 1830 } 1831 } 1832 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1833 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1834 if (nd.ni_dvp == vp) 1835 vrele(nd.ni_dvp); 1836 else 1837 vput(nd.ni_dvp); 1838 if (vp) 1839 vrele(vp); 1840 if ((error = vn_start_write(NULL, &mp, 1841 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1842 return (error); 1843 goto restart; 1844 } 1845 if (!error) { 1846 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1847 if (whiteout) { 1848 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1849 if (error) 1850 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1851 vput(nd.ni_dvp); 1852 } else { 1853 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1854 &nd.ni_cnd, &vattr); 1855 if (error == 0) 1856 vput(nd.ni_vp); 1857 } 1858 } else { 1859 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1860 if (nd.ni_dvp == vp) 1861 vrele(nd.ni_dvp); 1862 else 1863 vput(nd.ni_dvp); 1864 if (vp) 1865 vrele(vp); 1866 } 1867 vn_finished_write(mp, 0); 1868 return (error); 1869 } 1870 1871 /* 1872 * Create a named pipe. 1873 */ 1874 /* ARGSUSED */ 1875 int 1876 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1877 { 1878 struct sys_mkfifo_args /* { 1879 syscallarg(const char *) path; 1880 syscallarg(int) mode; 1881 } */ *uap = v; 1882 struct proc *p = l->l_proc; 1883 struct mount *mp; 1884 struct vattr vattr; 1885 int error; 1886 struct nameidata nd; 1887 1888 restart: 1889 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1890 if ((error = namei(&nd)) != 0) 1891 return (error); 1892 if (nd.ni_vp != NULL) { 1893 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1894 if (nd.ni_dvp == nd.ni_vp) 1895 vrele(nd.ni_dvp); 1896 else 1897 vput(nd.ni_dvp); 1898 vrele(nd.ni_vp); 1899 return (EEXIST); 1900 } 1901 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1902 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1903 if (nd.ni_dvp == nd.ni_vp) 1904 vrele(nd.ni_dvp); 1905 else 1906 vput(nd.ni_dvp); 1907 if (nd.ni_vp) 1908 vrele(nd.ni_vp); 1909 if ((error = vn_start_write(NULL, &mp, 1910 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1911 return (error); 1912 goto restart; 1913 } 1914 VATTR_NULL(&vattr); 1915 vattr.va_type = VFIFO; 1916 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1917 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1918 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1919 if (error == 0) 1920 vput(nd.ni_vp); 1921 vn_finished_write(mp, 0); 1922 return (error); 1923 } 1924 1925 /* 1926 * Make a hard file link. 1927 */ 1928 /* ARGSUSED */ 1929 int 1930 sys_link(struct lwp *l, void *v, register_t *retval) 1931 { 1932 struct sys_link_args /* { 1933 syscallarg(const char *) path; 1934 syscallarg(const char *) link; 1935 } */ *uap = v; 1936 struct vnode *vp; 1937 struct mount *mp; 1938 struct nameidata nd; 1939 int error; 1940 1941 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1942 if ((error = namei(&nd)) != 0) 1943 return (error); 1944 vp = nd.ni_vp; 1945 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1946 vrele(vp); 1947 return (error); 1948 } 1949 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1950 if ((error = namei(&nd)) != 0) 1951 goto out; 1952 if (nd.ni_vp) { 1953 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1954 if (nd.ni_dvp == nd.ni_vp) 1955 vrele(nd.ni_dvp); 1956 else 1957 vput(nd.ni_dvp); 1958 vrele(nd.ni_vp); 1959 error = EEXIST; 1960 goto out; 1961 } 1962 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1963 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1964 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1965 out: 1966 vrele(vp); 1967 vn_finished_write(mp, 0); 1968 return (error); 1969 } 1970 1971 /* 1972 * Make a symbolic link. 1973 */ 1974 /* ARGSUSED */ 1975 int 1976 sys_symlink(struct lwp *l, void *v, register_t *retval) 1977 { 1978 struct sys_symlink_args /* { 1979 syscallarg(const char *) path; 1980 syscallarg(const char *) link; 1981 } */ *uap = v; 1982 struct proc *p = l->l_proc; 1983 struct mount *mp; 1984 struct vattr vattr; 1985 char *path; 1986 int error; 1987 struct nameidata nd; 1988 1989 path = PNBUF_GET(); 1990 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1991 if (error) 1992 goto out; 1993 restart: 1994 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1995 if ((error = namei(&nd)) != 0) 1996 goto out; 1997 if (nd.ni_vp) { 1998 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1999 if (nd.ni_dvp == nd.ni_vp) 2000 vrele(nd.ni_dvp); 2001 else 2002 vput(nd.ni_dvp); 2003 vrele(nd.ni_vp); 2004 error = EEXIST; 2005 goto out; 2006 } 2007 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2008 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2009 if (nd.ni_dvp == nd.ni_vp) 2010 vrele(nd.ni_dvp); 2011 else 2012 vput(nd.ni_dvp); 2013 if ((error = vn_start_write(NULL, &mp, 2014 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2015 return (error); 2016 goto restart; 2017 } 2018 VATTR_NULL(&vattr); 2019 vattr.va_type = VLNK; 2020 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2021 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2022 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2023 if (error == 0) 2024 vput(nd.ni_vp); 2025 vn_finished_write(mp, 0); 2026 out: 2027 PNBUF_PUT(path); 2028 return (error); 2029 } 2030 2031 /* 2032 * Delete a whiteout from the filesystem. 2033 */ 2034 /* ARGSUSED */ 2035 int 2036 sys_undelete(struct lwp *l, void *v, register_t *retval) 2037 { 2038 struct sys_undelete_args /* { 2039 syscallarg(const char *) path; 2040 } */ *uap = v; 2041 int error; 2042 struct mount *mp; 2043 struct nameidata nd; 2044 2045 restart: 2046 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2047 SCARG(uap, path), l); 2048 error = namei(&nd); 2049 if (error) 2050 return (error); 2051 2052 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2053 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2054 if (nd.ni_dvp == nd.ni_vp) 2055 vrele(nd.ni_dvp); 2056 else 2057 vput(nd.ni_dvp); 2058 if (nd.ni_vp) 2059 vrele(nd.ni_vp); 2060 return (EEXIST); 2061 } 2062 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2063 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2064 if (nd.ni_dvp == nd.ni_vp) 2065 vrele(nd.ni_dvp); 2066 else 2067 vput(nd.ni_dvp); 2068 if ((error = vn_start_write(NULL, &mp, 2069 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2070 return (error); 2071 goto restart; 2072 } 2073 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2074 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2075 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2076 vput(nd.ni_dvp); 2077 vn_finished_write(mp, 0); 2078 return (error); 2079 } 2080 2081 /* 2082 * Delete a name from the filesystem. 2083 */ 2084 /* ARGSUSED */ 2085 int 2086 sys_unlink(struct lwp *l, void *v, register_t *retval) 2087 { 2088 struct sys_unlink_args /* { 2089 syscallarg(const char *) path; 2090 } */ *uap = v; 2091 struct mount *mp; 2092 struct vnode *vp; 2093 int error; 2094 struct nameidata nd; 2095 #if NVERIEXEC > 0 2096 pathname_t pathbuf; 2097 #endif /* NVERIEXEC > 0 */ 2098 2099 restart: 2100 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2101 SCARG(uap, path), l); 2102 if ((error = namei(&nd)) != 0) 2103 return (error); 2104 vp = nd.ni_vp; 2105 2106 /* 2107 * The root of a mounted filesystem cannot be deleted. 2108 */ 2109 if (vp->v_flag & VROOT) { 2110 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2111 if (nd.ni_dvp == vp) 2112 vrele(nd.ni_dvp); 2113 else 2114 vput(nd.ni_dvp); 2115 vput(vp); 2116 error = EBUSY; 2117 goto out; 2118 } 2119 2120 #if NVERIEXEC > 0 2121 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2122 2123 /* Handle remove requests for veriexec entries. */ 2124 if (!error) { 2125 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2126 pathname_put(pathbuf); 2127 } 2128 2129 if (error) { 2130 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2131 if (nd.ni_dvp == vp) 2132 vrele(nd.ni_dvp); 2133 else 2134 vput(nd.ni_dvp); 2135 vput(vp); 2136 goto out; 2137 } 2138 #endif /* NVERIEXEC > 0 */ 2139 2140 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2141 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2142 if (nd.ni_dvp == vp) 2143 vrele(nd.ni_dvp); 2144 else 2145 vput(nd.ni_dvp); 2146 vput(vp); 2147 if ((error = vn_start_write(NULL, &mp, 2148 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2149 return (error); 2150 goto restart; 2151 } 2152 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2153 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2154 #ifdef FILEASSOC 2155 (void)fileassoc_file_delete(vp); 2156 #endif /* FILEASSOC */ 2157 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2158 vn_finished_write(mp, 0); 2159 out: 2160 return (error); 2161 } 2162 2163 /* 2164 * Reposition read/write file offset. 2165 */ 2166 int 2167 sys_lseek(struct lwp *l, void *v, register_t *retval) 2168 { 2169 struct sys_lseek_args /* { 2170 syscallarg(int) fd; 2171 syscallarg(int) pad; 2172 syscallarg(off_t) offset; 2173 syscallarg(int) whence; 2174 } */ *uap = v; 2175 struct proc *p = l->l_proc; 2176 kauth_cred_t cred = l->l_cred; 2177 struct filedesc *fdp = p->p_fd; 2178 struct file *fp; 2179 struct vnode *vp; 2180 struct vattr vattr; 2181 off_t newoff; 2182 int error; 2183 2184 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2185 return (EBADF); 2186 2187 FILE_USE(fp); 2188 2189 vp = (struct vnode *)fp->f_data; 2190 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2191 error = ESPIPE; 2192 goto out; 2193 } 2194 2195 switch (SCARG(uap, whence)) { 2196 case SEEK_CUR: 2197 newoff = fp->f_offset + SCARG(uap, offset); 2198 break; 2199 case SEEK_END: 2200 error = VOP_GETATTR(vp, &vattr, cred, l); 2201 if (error) 2202 goto out; 2203 newoff = SCARG(uap, offset) + vattr.va_size; 2204 break; 2205 case SEEK_SET: 2206 newoff = SCARG(uap, offset); 2207 break; 2208 default: 2209 error = EINVAL; 2210 goto out; 2211 } 2212 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2213 goto out; 2214 2215 *(off_t *)retval = fp->f_offset = newoff; 2216 out: 2217 FILE_UNUSE(fp, l); 2218 return (error); 2219 } 2220 2221 /* 2222 * Positional read system call. 2223 */ 2224 int 2225 sys_pread(struct lwp *l, void *v, register_t *retval) 2226 { 2227 struct sys_pread_args /* { 2228 syscallarg(int) fd; 2229 syscallarg(void *) buf; 2230 syscallarg(size_t) nbyte; 2231 syscallarg(off_t) offset; 2232 } */ *uap = v; 2233 struct proc *p = l->l_proc; 2234 struct filedesc *fdp = p->p_fd; 2235 struct file *fp; 2236 struct vnode *vp; 2237 off_t offset; 2238 int error, fd = SCARG(uap, fd); 2239 2240 if ((fp = fd_getfile(fdp, fd)) == NULL) 2241 return (EBADF); 2242 2243 if ((fp->f_flag & FREAD) == 0) { 2244 simple_unlock(&fp->f_slock); 2245 return (EBADF); 2246 } 2247 2248 FILE_USE(fp); 2249 2250 vp = (struct vnode *)fp->f_data; 2251 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2252 error = ESPIPE; 2253 goto out; 2254 } 2255 2256 offset = SCARG(uap, offset); 2257 2258 /* 2259 * XXX This works because no file systems actually 2260 * XXX take any action on the seek operation. 2261 */ 2262 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2263 goto out; 2264 2265 /* dofileread() will unuse the descriptor for us */ 2266 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2267 &offset, 0, retval)); 2268 2269 out: 2270 FILE_UNUSE(fp, l); 2271 return (error); 2272 } 2273 2274 /* 2275 * Positional scatter read system call. 2276 */ 2277 int 2278 sys_preadv(struct lwp *l, void *v, register_t *retval) 2279 { 2280 struct sys_preadv_args /* { 2281 syscallarg(int) fd; 2282 syscallarg(const struct iovec *) iovp; 2283 syscallarg(int) iovcnt; 2284 syscallarg(off_t) offset; 2285 } */ *uap = v; 2286 struct proc *p = l->l_proc; 2287 struct filedesc *fdp = p->p_fd; 2288 struct file *fp; 2289 struct vnode *vp; 2290 off_t offset; 2291 int error, fd = SCARG(uap, fd); 2292 2293 if ((fp = fd_getfile(fdp, fd)) == NULL) 2294 return (EBADF); 2295 2296 if ((fp->f_flag & FREAD) == 0) { 2297 simple_unlock(&fp->f_slock); 2298 return (EBADF); 2299 } 2300 2301 FILE_USE(fp); 2302 2303 vp = (struct vnode *)fp->f_data; 2304 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2305 error = ESPIPE; 2306 goto out; 2307 } 2308 2309 offset = SCARG(uap, offset); 2310 2311 /* 2312 * XXX This works because no file systems actually 2313 * XXX take any action on the seek operation. 2314 */ 2315 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2316 goto out; 2317 2318 /* dofilereadv() will unuse the descriptor for us */ 2319 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2320 &offset, 0, retval)); 2321 2322 out: 2323 FILE_UNUSE(fp, l); 2324 return (error); 2325 } 2326 2327 /* 2328 * Positional write system call. 2329 */ 2330 int 2331 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2332 { 2333 struct sys_pwrite_args /* { 2334 syscallarg(int) fd; 2335 syscallarg(const void *) buf; 2336 syscallarg(size_t) nbyte; 2337 syscallarg(off_t) offset; 2338 } */ *uap = v; 2339 struct proc *p = l->l_proc; 2340 struct filedesc *fdp = p->p_fd; 2341 struct file *fp; 2342 struct vnode *vp; 2343 off_t offset; 2344 int error, fd = SCARG(uap, fd); 2345 2346 if ((fp = fd_getfile(fdp, fd)) == NULL) 2347 return (EBADF); 2348 2349 if ((fp->f_flag & FWRITE) == 0) { 2350 simple_unlock(&fp->f_slock); 2351 return (EBADF); 2352 } 2353 2354 FILE_USE(fp); 2355 2356 vp = (struct vnode *)fp->f_data; 2357 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2358 error = ESPIPE; 2359 goto out; 2360 } 2361 2362 offset = SCARG(uap, offset); 2363 2364 /* 2365 * XXX This works because no file systems actually 2366 * XXX take any action on the seek operation. 2367 */ 2368 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2369 goto out; 2370 2371 /* dofilewrite() will unuse the descriptor for us */ 2372 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2373 &offset, 0, retval)); 2374 2375 out: 2376 FILE_UNUSE(fp, l); 2377 return (error); 2378 } 2379 2380 /* 2381 * Positional gather write system call. 2382 */ 2383 int 2384 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2385 { 2386 struct sys_pwritev_args /* { 2387 syscallarg(int) fd; 2388 syscallarg(const struct iovec *) iovp; 2389 syscallarg(int) iovcnt; 2390 syscallarg(off_t) offset; 2391 } */ *uap = v; 2392 struct proc *p = l->l_proc; 2393 struct filedesc *fdp = p->p_fd; 2394 struct file *fp; 2395 struct vnode *vp; 2396 off_t offset; 2397 int error, fd = SCARG(uap, fd); 2398 2399 if ((fp = fd_getfile(fdp, fd)) == NULL) 2400 return (EBADF); 2401 2402 if ((fp->f_flag & FWRITE) == 0) { 2403 simple_unlock(&fp->f_slock); 2404 return (EBADF); 2405 } 2406 2407 FILE_USE(fp); 2408 2409 vp = (struct vnode *)fp->f_data; 2410 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2411 error = ESPIPE; 2412 goto out; 2413 } 2414 2415 offset = SCARG(uap, offset); 2416 2417 /* 2418 * XXX This works because no file systems actually 2419 * XXX take any action on the seek operation. 2420 */ 2421 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2422 goto out; 2423 2424 /* dofilewritev() will unuse the descriptor for us */ 2425 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2426 &offset, 0, retval)); 2427 2428 out: 2429 FILE_UNUSE(fp, l); 2430 return (error); 2431 } 2432 2433 /* 2434 * Check access permissions. 2435 */ 2436 int 2437 sys_access(struct lwp *l, void *v, register_t *retval) 2438 { 2439 struct sys_access_args /* { 2440 syscallarg(const char *) path; 2441 syscallarg(int) flags; 2442 } */ *uap = v; 2443 kauth_cred_t cred; 2444 struct vnode *vp; 2445 int error, flags; 2446 struct nameidata nd; 2447 2448 cred = kauth_cred_dup(l->l_cred); 2449 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2450 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2451 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2452 SCARG(uap, path), l); 2453 /* Override default credentials */ 2454 nd.ni_cnd.cn_cred = cred; 2455 if ((error = namei(&nd)) != 0) 2456 goto out; 2457 vp = nd.ni_vp; 2458 2459 /* Flags == 0 means only check for existence. */ 2460 if (SCARG(uap, flags)) { 2461 flags = 0; 2462 if (SCARG(uap, flags) & R_OK) 2463 flags |= VREAD; 2464 if (SCARG(uap, flags) & W_OK) 2465 flags |= VWRITE; 2466 if (SCARG(uap, flags) & X_OK) 2467 flags |= VEXEC; 2468 2469 error = VOP_ACCESS(vp, flags, cred, l); 2470 if (!error && (flags & VWRITE)) 2471 error = vn_writechk(vp); 2472 } 2473 vput(vp); 2474 out: 2475 kauth_cred_free(cred); 2476 return (error); 2477 } 2478 2479 /* 2480 * Get file status; this version follows links. 2481 */ 2482 /* ARGSUSED */ 2483 int 2484 sys___stat30(struct lwp *l, void *v, register_t *retval) 2485 { 2486 struct sys___stat30_args /* { 2487 syscallarg(const char *) path; 2488 syscallarg(struct stat *) ub; 2489 } */ *uap = v; 2490 struct stat sb; 2491 int error; 2492 struct nameidata nd; 2493 2494 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2495 SCARG(uap, path), l); 2496 if ((error = namei(&nd)) != 0) 2497 return (error); 2498 error = vn_stat(nd.ni_vp, &sb, l); 2499 vput(nd.ni_vp); 2500 if (error) 2501 return (error); 2502 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2503 return (error); 2504 } 2505 2506 /* 2507 * Get file status; this version does not follow links. 2508 */ 2509 /* ARGSUSED */ 2510 int 2511 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2512 { 2513 struct sys___lstat30_args /* { 2514 syscallarg(const char *) path; 2515 syscallarg(struct stat *) ub; 2516 } */ *uap = v; 2517 struct stat sb; 2518 int error; 2519 struct nameidata nd; 2520 2521 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2522 SCARG(uap, path), l); 2523 if ((error = namei(&nd)) != 0) 2524 return (error); 2525 error = vn_stat(nd.ni_vp, &sb, l); 2526 vput(nd.ni_vp); 2527 if (error) 2528 return (error); 2529 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2530 return (error); 2531 } 2532 2533 /* 2534 * Get configurable pathname variables. 2535 */ 2536 /* ARGSUSED */ 2537 int 2538 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2539 { 2540 struct sys_pathconf_args /* { 2541 syscallarg(const char *) path; 2542 syscallarg(int) name; 2543 } */ *uap = v; 2544 int error; 2545 struct nameidata nd; 2546 2547 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2548 SCARG(uap, path), l); 2549 if ((error = namei(&nd)) != 0) 2550 return (error); 2551 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2552 vput(nd.ni_vp); 2553 return (error); 2554 } 2555 2556 /* 2557 * Return target name of a symbolic link. 2558 */ 2559 /* ARGSUSED */ 2560 int 2561 sys_readlink(struct lwp *l, void *v, register_t *retval) 2562 { 2563 struct sys_readlink_args /* { 2564 syscallarg(const char *) path; 2565 syscallarg(char *) buf; 2566 syscallarg(size_t) count; 2567 } */ *uap = v; 2568 struct vnode *vp; 2569 struct iovec aiov; 2570 struct uio auio; 2571 int error; 2572 struct nameidata nd; 2573 2574 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2575 SCARG(uap, path), l); 2576 if ((error = namei(&nd)) != 0) 2577 return (error); 2578 vp = nd.ni_vp; 2579 if (vp->v_type != VLNK) 2580 error = EINVAL; 2581 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2582 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2583 aiov.iov_base = SCARG(uap, buf); 2584 aiov.iov_len = SCARG(uap, count); 2585 auio.uio_iov = &aiov; 2586 auio.uio_iovcnt = 1; 2587 auio.uio_offset = 0; 2588 auio.uio_rw = UIO_READ; 2589 KASSERT(l == curlwp); 2590 auio.uio_vmspace = l->l_proc->p_vmspace; 2591 auio.uio_resid = SCARG(uap, count); 2592 error = VOP_READLINK(vp, &auio, l->l_cred); 2593 } 2594 vput(vp); 2595 *retval = SCARG(uap, count) - auio.uio_resid; 2596 return (error); 2597 } 2598 2599 /* 2600 * Change flags of a file given a path name. 2601 */ 2602 /* ARGSUSED */ 2603 int 2604 sys_chflags(struct lwp *l, void *v, register_t *retval) 2605 { 2606 struct sys_chflags_args /* { 2607 syscallarg(const char *) path; 2608 syscallarg(u_long) flags; 2609 } */ *uap = v; 2610 struct vnode *vp; 2611 int error; 2612 struct nameidata nd; 2613 2614 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2615 if ((error = namei(&nd)) != 0) 2616 return (error); 2617 vp = nd.ni_vp; 2618 error = change_flags(vp, SCARG(uap, flags), l); 2619 vput(vp); 2620 return (error); 2621 } 2622 2623 /* 2624 * Change flags of a file given a file descriptor. 2625 */ 2626 /* ARGSUSED */ 2627 int 2628 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2629 { 2630 struct sys_fchflags_args /* { 2631 syscallarg(int) fd; 2632 syscallarg(u_long) flags; 2633 } */ *uap = v; 2634 struct proc *p = l->l_proc; 2635 struct vnode *vp; 2636 struct file *fp; 2637 int error; 2638 2639 /* getvnode() will use the descriptor for us */ 2640 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2641 return (error); 2642 vp = (struct vnode *)fp->f_data; 2643 error = change_flags(vp, SCARG(uap, flags), l); 2644 VOP_UNLOCK(vp, 0); 2645 FILE_UNUSE(fp, l); 2646 return (error); 2647 } 2648 2649 /* 2650 * Change flags of a file given a path name; this version does 2651 * not follow links. 2652 */ 2653 int 2654 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2655 { 2656 struct sys_lchflags_args /* { 2657 syscallarg(const char *) path; 2658 syscallarg(u_long) flags; 2659 } */ *uap = v; 2660 struct vnode *vp; 2661 int error; 2662 struct nameidata nd; 2663 2664 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2665 if ((error = namei(&nd)) != 0) 2666 return (error); 2667 vp = nd.ni_vp; 2668 error = change_flags(vp, SCARG(uap, flags), l); 2669 vput(vp); 2670 return (error); 2671 } 2672 2673 /* 2674 * Common routine to change flags of a file. 2675 */ 2676 int 2677 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2678 { 2679 struct mount *mp; 2680 struct vattr vattr; 2681 int error; 2682 2683 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2684 return (error); 2685 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2686 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2687 /* 2688 * Non-superusers cannot change the flags on devices, even if they 2689 * own them. 2690 */ 2691 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2692 &l->l_acflag) != 0) { 2693 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2694 goto out; 2695 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2696 error = EINVAL; 2697 goto out; 2698 } 2699 } 2700 VATTR_NULL(&vattr); 2701 vattr.va_flags = flags; 2702 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2703 out: 2704 vn_finished_write(mp, 0); 2705 return (error); 2706 } 2707 2708 /* 2709 * Change mode of a file given path name; this version follows links. 2710 */ 2711 /* ARGSUSED */ 2712 int 2713 sys_chmod(struct lwp *l, void *v, register_t *retval) 2714 { 2715 struct sys_chmod_args /* { 2716 syscallarg(const char *) path; 2717 syscallarg(int) mode; 2718 } */ *uap = v; 2719 int error; 2720 struct nameidata nd; 2721 2722 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2723 if ((error = namei(&nd)) != 0) 2724 return (error); 2725 2726 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2727 2728 vrele(nd.ni_vp); 2729 return (error); 2730 } 2731 2732 /* 2733 * Change mode of a file given a file descriptor. 2734 */ 2735 /* ARGSUSED */ 2736 int 2737 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2738 { 2739 struct sys_fchmod_args /* { 2740 syscallarg(int) fd; 2741 syscallarg(int) mode; 2742 } */ *uap = v; 2743 struct proc *p = l->l_proc; 2744 struct file *fp; 2745 int error; 2746 2747 /* getvnode() will use the descriptor for us */ 2748 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2749 return (error); 2750 2751 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2752 FILE_UNUSE(fp, l); 2753 return (error); 2754 } 2755 2756 /* 2757 * Change mode of a file given path name; this version does not follow links. 2758 */ 2759 /* ARGSUSED */ 2760 int 2761 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2762 { 2763 struct sys_lchmod_args /* { 2764 syscallarg(const char *) path; 2765 syscallarg(int) mode; 2766 } */ *uap = v; 2767 int error; 2768 struct nameidata nd; 2769 2770 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2771 if ((error = namei(&nd)) != 0) 2772 return (error); 2773 2774 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2775 2776 vrele(nd.ni_vp); 2777 return (error); 2778 } 2779 2780 /* 2781 * Common routine to set mode given a vnode. 2782 */ 2783 static int 2784 change_mode(struct vnode *vp, int mode, struct lwp *l) 2785 { 2786 struct mount *mp; 2787 struct vattr vattr; 2788 int error; 2789 2790 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2791 return (error); 2792 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2793 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2794 VATTR_NULL(&vattr); 2795 vattr.va_mode = mode & ALLPERMS; 2796 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2797 VOP_UNLOCK(vp, 0); 2798 vn_finished_write(mp, 0); 2799 return (error); 2800 } 2801 2802 /* 2803 * Set ownership given a path name; this version follows links. 2804 */ 2805 /* ARGSUSED */ 2806 int 2807 sys_chown(struct lwp *l, void *v, register_t *retval) 2808 { 2809 struct sys_chown_args /* { 2810 syscallarg(const char *) path; 2811 syscallarg(uid_t) uid; 2812 syscallarg(gid_t) gid; 2813 } */ *uap = v; 2814 int error; 2815 struct nameidata nd; 2816 2817 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2818 if ((error = namei(&nd)) != 0) 2819 return (error); 2820 2821 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2822 2823 vrele(nd.ni_vp); 2824 return (error); 2825 } 2826 2827 /* 2828 * Set ownership given a path name; this version follows links. 2829 * Provides POSIX semantics. 2830 */ 2831 /* ARGSUSED */ 2832 int 2833 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2834 { 2835 struct sys_chown_args /* { 2836 syscallarg(const char *) path; 2837 syscallarg(uid_t) uid; 2838 syscallarg(gid_t) gid; 2839 } */ *uap = v; 2840 int error; 2841 struct nameidata nd; 2842 2843 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2844 if ((error = namei(&nd)) != 0) 2845 return (error); 2846 2847 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2848 2849 vrele(nd.ni_vp); 2850 return (error); 2851 } 2852 2853 /* 2854 * Set ownership given a file descriptor. 2855 */ 2856 /* ARGSUSED */ 2857 int 2858 sys_fchown(struct lwp *l, void *v, register_t *retval) 2859 { 2860 struct sys_fchown_args /* { 2861 syscallarg(int) fd; 2862 syscallarg(uid_t) uid; 2863 syscallarg(gid_t) gid; 2864 } */ *uap = v; 2865 struct proc *p = l->l_proc; 2866 int error; 2867 struct file *fp; 2868 2869 /* getvnode() will use the descriptor for us */ 2870 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2871 return (error); 2872 2873 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2874 SCARG(uap, gid), l, 0); 2875 FILE_UNUSE(fp, l); 2876 return (error); 2877 } 2878 2879 /* 2880 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2881 */ 2882 /* ARGSUSED */ 2883 int 2884 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2885 { 2886 struct sys_fchown_args /* { 2887 syscallarg(int) fd; 2888 syscallarg(uid_t) uid; 2889 syscallarg(gid_t) gid; 2890 } */ *uap = v; 2891 struct proc *p = l->l_proc; 2892 int error; 2893 struct file *fp; 2894 2895 /* getvnode() will use the descriptor for us */ 2896 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2897 return (error); 2898 2899 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2900 SCARG(uap, gid), l, 1); 2901 FILE_UNUSE(fp, l); 2902 return (error); 2903 } 2904 2905 /* 2906 * Set ownership given a path name; this version does not follow links. 2907 */ 2908 /* ARGSUSED */ 2909 int 2910 sys_lchown(struct lwp *l, void *v, register_t *retval) 2911 { 2912 struct sys_lchown_args /* { 2913 syscallarg(const char *) path; 2914 syscallarg(uid_t) uid; 2915 syscallarg(gid_t) gid; 2916 } */ *uap = v; 2917 int error; 2918 struct nameidata nd; 2919 2920 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2921 if ((error = namei(&nd)) != 0) 2922 return (error); 2923 2924 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2925 2926 vrele(nd.ni_vp); 2927 return (error); 2928 } 2929 2930 /* 2931 * Set ownership given a path name; this version does not follow links. 2932 * Provides POSIX/XPG semantics. 2933 */ 2934 /* ARGSUSED */ 2935 int 2936 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2937 { 2938 struct sys_lchown_args /* { 2939 syscallarg(const char *) path; 2940 syscallarg(uid_t) uid; 2941 syscallarg(gid_t) gid; 2942 } */ *uap = v; 2943 int error; 2944 struct nameidata nd; 2945 2946 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2947 if ((error = namei(&nd)) != 0) 2948 return (error); 2949 2950 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2951 2952 vrele(nd.ni_vp); 2953 return (error); 2954 } 2955 2956 /* 2957 * Common routine to set ownership given a vnode. 2958 */ 2959 static int 2960 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2961 int posix_semantics) 2962 { 2963 struct mount *mp; 2964 struct vattr vattr; 2965 mode_t newmode; 2966 int error; 2967 2968 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2969 return (error); 2970 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2971 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2972 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2973 goto out; 2974 2975 #define CHANGED(x) ((int)(x) != -1) 2976 newmode = vattr.va_mode; 2977 if (posix_semantics) { 2978 /* 2979 * POSIX/XPG semantics: if the caller is not the super-user, 2980 * clear set-user-id and set-group-id bits. Both POSIX and 2981 * the XPG consider the behaviour for calls by the super-user 2982 * implementation-defined; we leave the set-user-id and set- 2983 * group-id settings intact in that case. 2984 */ 2985 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2986 NULL) != 0) 2987 newmode &= ~(S_ISUID | S_ISGID); 2988 } else { 2989 /* 2990 * NetBSD semantics: when changing owner and/or group, 2991 * clear the respective bit(s). 2992 */ 2993 if (CHANGED(uid)) 2994 newmode &= ~S_ISUID; 2995 if (CHANGED(gid)) 2996 newmode &= ~S_ISGID; 2997 } 2998 /* Update va_mode iff altered. */ 2999 if (vattr.va_mode == newmode) 3000 newmode = VNOVAL; 3001 3002 VATTR_NULL(&vattr); 3003 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3004 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3005 vattr.va_mode = newmode; 3006 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3007 #undef CHANGED 3008 3009 out: 3010 VOP_UNLOCK(vp, 0); 3011 vn_finished_write(mp, 0); 3012 return (error); 3013 } 3014 3015 /* 3016 * Set the access and modification times given a path name; this 3017 * version follows links. 3018 */ 3019 /* ARGSUSED */ 3020 int 3021 sys_utimes(struct lwp *l, void *v, register_t *retval) 3022 { 3023 struct sys_utimes_args /* { 3024 syscallarg(const char *) path; 3025 syscallarg(const struct timeval *) tptr; 3026 } */ *uap = v; 3027 int error; 3028 struct nameidata nd; 3029 3030 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3031 if ((error = namei(&nd)) != 0) 3032 return (error); 3033 3034 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3035 3036 vrele(nd.ni_vp); 3037 return (error); 3038 } 3039 3040 /* 3041 * Set the access and modification times given a file descriptor. 3042 */ 3043 /* ARGSUSED */ 3044 int 3045 sys_futimes(struct lwp *l, void *v, register_t *retval) 3046 { 3047 struct sys_futimes_args /* { 3048 syscallarg(int) fd; 3049 syscallarg(const struct timeval *) tptr; 3050 } */ *uap = v; 3051 struct proc *p = l->l_proc; 3052 int error; 3053 struct file *fp; 3054 3055 /* getvnode() will use the descriptor for us */ 3056 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3057 return (error); 3058 3059 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3060 FILE_UNUSE(fp, l); 3061 return (error); 3062 } 3063 3064 /* 3065 * Set the access and modification times given a path name; this 3066 * version does not follow links. 3067 */ 3068 /* ARGSUSED */ 3069 int 3070 sys_lutimes(struct lwp *l, void *v, register_t *retval) 3071 { 3072 struct sys_lutimes_args /* { 3073 syscallarg(const char *) path; 3074 syscallarg(const struct timeval *) tptr; 3075 } */ *uap = v; 3076 int error; 3077 struct nameidata nd; 3078 3079 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3080 if ((error = namei(&nd)) != 0) 3081 return (error); 3082 3083 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3084 3085 vrele(nd.ni_vp); 3086 return (error); 3087 } 3088 3089 /* 3090 * Common routine to set access and modification times given a vnode. 3091 */ 3092 static int 3093 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3094 { 3095 struct mount *mp; 3096 struct vattr vattr; 3097 int error; 3098 3099 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3100 return (error); 3101 VATTR_NULL(&vattr); 3102 if (tptr == NULL) { 3103 nanotime(&vattr.va_atime); 3104 vattr.va_mtime = vattr.va_atime; 3105 vattr.va_vaflags |= VA_UTIMES_NULL; 3106 } else { 3107 struct timeval tv[2]; 3108 3109 error = copyin(tptr, tv, sizeof(tv)); 3110 if (error) 3111 goto out; 3112 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3113 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3114 } 3115 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3116 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3117 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3118 VOP_UNLOCK(vp, 0); 3119 out: 3120 vn_finished_write(mp, 0); 3121 return (error); 3122 } 3123 3124 /* 3125 * Truncate a file given its path name. 3126 */ 3127 /* ARGSUSED */ 3128 int 3129 sys_truncate(struct lwp *l, void *v, register_t *retval) 3130 { 3131 struct sys_truncate_args /* { 3132 syscallarg(const char *) path; 3133 syscallarg(int) pad; 3134 syscallarg(off_t) length; 3135 } */ *uap = v; 3136 struct vnode *vp; 3137 struct mount *mp; 3138 struct vattr vattr; 3139 int error; 3140 struct nameidata nd; 3141 3142 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3143 if ((error = namei(&nd)) != 0) 3144 return (error); 3145 vp = nd.ni_vp; 3146 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3147 vrele(vp); 3148 return (error); 3149 } 3150 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3151 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3152 if (vp->v_type == VDIR) 3153 error = EISDIR; 3154 else if ((error = vn_writechk(vp)) == 0 && 3155 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3156 VATTR_NULL(&vattr); 3157 vattr.va_size = SCARG(uap, length); 3158 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3159 } 3160 vput(vp); 3161 vn_finished_write(mp, 0); 3162 return (error); 3163 } 3164 3165 /* 3166 * Truncate a file given a file descriptor. 3167 */ 3168 /* ARGSUSED */ 3169 int 3170 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3171 { 3172 struct sys_ftruncate_args /* { 3173 syscallarg(int) fd; 3174 syscallarg(int) pad; 3175 syscallarg(off_t) length; 3176 } */ *uap = v; 3177 struct proc *p = l->l_proc; 3178 struct mount *mp; 3179 struct vattr vattr; 3180 struct vnode *vp; 3181 struct file *fp; 3182 int error; 3183 3184 /* getvnode() will use the descriptor for us */ 3185 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3186 return (error); 3187 if ((fp->f_flag & FWRITE) == 0) { 3188 error = EINVAL; 3189 goto out; 3190 } 3191 vp = (struct vnode *)fp->f_data; 3192 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3193 FILE_UNUSE(fp, l); 3194 return (error); 3195 } 3196 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3197 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3198 if (vp->v_type == VDIR) 3199 error = EISDIR; 3200 else if ((error = vn_writechk(vp)) == 0) { 3201 VATTR_NULL(&vattr); 3202 vattr.va_size = SCARG(uap, length); 3203 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3204 } 3205 VOP_UNLOCK(vp, 0); 3206 vn_finished_write(mp, 0); 3207 out: 3208 FILE_UNUSE(fp, l); 3209 return (error); 3210 } 3211 3212 /* 3213 * Sync an open file. 3214 */ 3215 /* ARGSUSED */ 3216 int 3217 sys_fsync(struct lwp *l, void *v, register_t *retval) 3218 { 3219 struct sys_fsync_args /* { 3220 syscallarg(int) fd; 3221 } */ *uap = v; 3222 struct proc *p = l->l_proc; 3223 struct vnode *vp; 3224 struct mount *mp; 3225 struct file *fp; 3226 int error; 3227 3228 /* getvnode() will use the descriptor for us */ 3229 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3230 return (error); 3231 vp = (struct vnode *)fp->f_data; 3232 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3233 FILE_UNUSE(fp, l); 3234 return (error); 3235 } 3236 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3237 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3238 if (error == 0 && bioops.io_fsync != NULL && 3239 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3240 (*bioops.io_fsync)(vp, 0); 3241 VOP_UNLOCK(vp, 0); 3242 vn_finished_write(mp, 0); 3243 FILE_UNUSE(fp, l); 3244 return (error); 3245 } 3246 3247 /* 3248 * Sync a range of file data. API modeled after that found in AIX. 3249 * 3250 * FDATASYNC indicates that we need only save enough metadata to be able 3251 * to re-read the written data. Note we duplicate AIX's requirement that 3252 * the file be open for writing. 3253 */ 3254 /* ARGSUSED */ 3255 int 3256 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3257 { 3258 struct sys_fsync_range_args /* { 3259 syscallarg(int) fd; 3260 syscallarg(int) flags; 3261 syscallarg(off_t) start; 3262 syscallarg(off_t) length; 3263 } */ *uap = v; 3264 struct proc *p = l->l_proc; 3265 struct vnode *vp; 3266 struct file *fp; 3267 int flags, nflags; 3268 off_t s, e, len; 3269 int error; 3270 3271 /* getvnode() will use the descriptor for us */ 3272 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3273 return (error); 3274 3275 if ((fp->f_flag & FWRITE) == 0) { 3276 FILE_UNUSE(fp, l); 3277 return (EBADF); 3278 } 3279 3280 flags = SCARG(uap, flags); 3281 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3282 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3283 return (EINVAL); 3284 } 3285 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3286 if (flags & FDATASYNC) 3287 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3288 else 3289 nflags = FSYNC_WAIT; 3290 if (flags & FDISKSYNC) 3291 nflags |= FSYNC_CACHE; 3292 3293 len = SCARG(uap, length); 3294 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3295 if (len) { 3296 s = SCARG(uap, start); 3297 e = s + len; 3298 if (e < s) { 3299 FILE_UNUSE(fp, l); 3300 return (EINVAL); 3301 } 3302 } else { 3303 e = 0; 3304 s = 0; 3305 } 3306 3307 vp = (struct vnode *)fp->f_data; 3308 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3309 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3310 3311 if (error == 0 && bioops.io_fsync != NULL && 3312 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3313 (*bioops.io_fsync)(vp, nflags); 3314 3315 VOP_UNLOCK(vp, 0); 3316 FILE_UNUSE(fp, l); 3317 return (error); 3318 } 3319 3320 /* 3321 * Sync the data of an open file. 3322 */ 3323 /* ARGSUSED */ 3324 int 3325 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3326 { 3327 struct sys_fdatasync_args /* { 3328 syscallarg(int) fd; 3329 } */ *uap = v; 3330 struct proc *p = l->l_proc; 3331 struct vnode *vp; 3332 struct file *fp; 3333 int error; 3334 3335 /* getvnode() will use the descriptor for us */ 3336 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3337 return (error); 3338 if ((fp->f_flag & FWRITE) == 0) { 3339 FILE_UNUSE(fp, l); 3340 return (EBADF); 3341 } 3342 vp = (struct vnode *)fp->f_data; 3343 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3344 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3345 VOP_UNLOCK(vp, 0); 3346 FILE_UNUSE(fp, l); 3347 return (error); 3348 } 3349 3350 /* 3351 * Rename files, (standard) BSD semantics frontend. 3352 */ 3353 /* ARGSUSED */ 3354 int 3355 sys_rename(struct lwp *l, void *v, register_t *retval) 3356 { 3357 struct sys_rename_args /* { 3358 syscallarg(const char *) from; 3359 syscallarg(const char *) to; 3360 } */ *uap = v; 3361 3362 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3363 } 3364 3365 /* 3366 * Rename files, POSIX semantics frontend. 3367 */ 3368 /* ARGSUSED */ 3369 int 3370 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3371 { 3372 struct sys___posix_rename_args /* { 3373 syscallarg(const char *) from; 3374 syscallarg(const char *) to; 3375 } */ *uap = v; 3376 3377 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3378 } 3379 3380 /* 3381 * Rename files. Source and destination must either both be directories, 3382 * or both not be directories. If target is a directory, it must be empty. 3383 * If `from' and `to' refer to the same object, the value of the `retain' 3384 * argument is used to determine whether `from' will be 3385 * 3386 * (retain == 0) deleted unless `from' and `to' refer to the same 3387 * object in the file system's name space (BSD). 3388 * (retain == 1) always retained (POSIX). 3389 */ 3390 static int 3391 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3392 { 3393 struct mount *mp = NULL; 3394 struct vnode *tvp, *fvp, *tdvp; 3395 struct nameidata fromnd, tond; 3396 struct proc *p; 3397 int error; 3398 3399 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3400 from, l); 3401 if ((error = namei(&fromnd)) != 0) 3402 return (error); 3403 VOP_UNLOCK(fromnd.ni_dvp, 0); 3404 fvp = fromnd.ni_vp; 3405 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3406 if (error != 0) { 3407 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3408 vrele(fromnd.ni_dvp); 3409 vrele(fvp); 3410 if (fromnd.ni_startdir) 3411 vrele(fromnd.ni_startdir); 3412 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3413 return (error); 3414 } 3415 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3416 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3417 if ((error = namei(&tond)) != 0) { 3418 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3419 vrele(fromnd.ni_dvp); 3420 vrele(fvp); 3421 goto out1; 3422 } 3423 tdvp = tond.ni_dvp; 3424 tvp = tond.ni_vp; 3425 3426 if (tvp != NULL) { 3427 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3428 error = ENOTDIR; 3429 goto out; 3430 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3431 error = EISDIR; 3432 goto out; 3433 } 3434 } 3435 3436 if (fvp == tdvp) 3437 error = EINVAL; 3438 3439 /* 3440 * Source and destination refer to the same object. 3441 */ 3442 if (fvp == tvp) { 3443 if (retain) 3444 error = -1; 3445 else if (fromnd.ni_dvp == tdvp && 3446 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3447 !memcmp(fromnd.ni_cnd.cn_nameptr, 3448 tond.ni_cnd.cn_nameptr, 3449 fromnd.ni_cnd.cn_namelen)) 3450 error = -1; 3451 } 3452 3453 #if NVERIEXEC > 0 3454 if (!error) { 3455 pathname_t frompath, topath; 3456 3457 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3458 &frompath); 3459 if (!error) 3460 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3461 &topath); 3462 if (!error) 3463 error = veriexec_renamechk(fvp, pathname_path(frompath), 3464 tvp, pathname_path(topath), l); 3465 3466 pathname_put(frompath); 3467 pathname_put(topath); 3468 } 3469 #endif /* NVERIEXEC > 0 */ 3470 3471 out: 3472 p = l->l_proc; 3473 if (!error) { 3474 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3475 if (fromnd.ni_dvp != tdvp) 3476 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3477 if (tvp) { 3478 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3479 } 3480 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3481 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3482 } else { 3483 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3484 if (tdvp == tvp) 3485 vrele(tdvp); 3486 else 3487 vput(tdvp); 3488 if (tvp) 3489 vput(tvp); 3490 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3491 vrele(fromnd.ni_dvp); 3492 vrele(fvp); 3493 } 3494 vrele(tond.ni_startdir); 3495 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3496 out1: 3497 vn_finished_write(mp, 0); 3498 if (fromnd.ni_startdir) 3499 vrele(fromnd.ni_startdir); 3500 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3501 return (error == -1 ? 0 : error); 3502 } 3503 3504 /* 3505 * Make a directory file. 3506 */ 3507 /* ARGSUSED */ 3508 int 3509 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3510 { 3511 struct sys_mkdir_args /* { 3512 syscallarg(const char *) path; 3513 syscallarg(int) mode; 3514 } */ *uap = v; 3515 struct proc *p = l->l_proc; 3516 struct mount *mp; 3517 struct vnode *vp; 3518 struct vattr vattr; 3519 int error; 3520 struct nameidata nd; 3521 3522 restart: 3523 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3524 SCARG(uap, path), l); 3525 if ((error = namei(&nd)) != 0) 3526 return (error); 3527 vp = nd.ni_vp; 3528 if (vp != NULL) { 3529 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3530 if (nd.ni_dvp == vp) 3531 vrele(nd.ni_dvp); 3532 else 3533 vput(nd.ni_dvp); 3534 vrele(vp); 3535 return (EEXIST); 3536 } 3537 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 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 if ((error = vn_start_write(NULL, &mp, 3544 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3545 return (error); 3546 goto restart; 3547 } 3548 VATTR_NULL(&vattr); 3549 vattr.va_type = VDIR; 3550 vattr.va_mode = 3551 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3552 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3553 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3554 if (!error) 3555 vput(nd.ni_vp); 3556 vn_finished_write(mp, 0); 3557 return (error); 3558 } 3559 3560 /* 3561 * Remove a directory file. 3562 */ 3563 /* ARGSUSED */ 3564 int 3565 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3566 { 3567 struct sys_rmdir_args /* { 3568 syscallarg(const char *) path; 3569 } */ *uap = v; 3570 struct mount *mp; 3571 struct vnode *vp; 3572 int error; 3573 struct nameidata nd; 3574 3575 restart: 3576 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3577 SCARG(uap, path), l); 3578 if ((error = namei(&nd)) != 0) 3579 return (error); 3580 vp = nd.ni_vp; 3581 if (vp->v_type != VDIR) { 3582 error = ENOTDIR; 3583 goto out; 3584 } 3585 /* 3586 * No rmdir "." please. 3587 */ 3588 if (nd.ni_dvp == vp) { 3589 error = EINVAL; 3590 goto out; 3591 } 3592 /* 3593 * The root of a mounted filesystem cannot be deleted. 3594 */ 3595 if (vp->v_flag & VROOT) { 3596 error = EBUSY; 3597 goto out; 3598 } 3599 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3600 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3601 if (nd.ni_dvp == vp) 3602 vrele(nd.ni_dvp); 3603 else 3604 vput(nd.ni_dvp); 3605 vput(vp); 3606 if ((error = vn_start_write(NULL, &mp, 3607 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3608 return (error); 3609 goto restart; 3610 } 3611 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3612 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3613 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3614 vn_finished_write(mp, 0); 3615 return (error); 3616 3617 out: 3618 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3619 if (nd.ni_dvp == vp) 3620 vrele(nd.ni_dvp); 3621 else 3622 vput(nd.ni_dvp); 3623 vput(vp); 3624 return (error); 3625 } 3626 3627 /* 3628 * Read a block of directory entries in a file system independent format. 3629 */ 3630 int 3631 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3632 { 3633 struct sys___getdents30_args /* { 3634 syscallarg(int) fd; 3635 syscallarg(char *) buf; 3636 syscallarg(size_t) count; 3637 } */ *uap = v; 3638 struct proc *p = l->l_proc; 3639 struct file *fp; 3640 int error, done; 3641 3642 /* getvnode() will use the descriptor for us */ 3643 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3644 return (error); 3645 if ((fp->f_flag & FREAD) == 0) { 3646 error = EBADF; 3647 goto out; 3648 } 3649 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3650 SCARG(uap, count), &done, l, 0, 0); 3651 #ifdef KTRACE 3652 if (!error && KTRPOINT(p, KTR_GENIO)) { 3653 struct iovec iov; 3654 iov.iov_base = SCARG(uap, buf); 3655 iov.iov_len = done; 3656 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3657 } 3658 #endif 3659 *retval = done; 3660 out: 3661 FILE_UNUSE(fp, l); 3662 return (error); 3663 } 3664 3665 /* 3666 * Set the mode mask for creation of filesystem nodes. 3667 */ 3668 int 3669 sys_umask(struct lwp *l, void *v, register_t *retval) 3670 { 3671 struct sys_umask_args /* { 3672 syscallarg(mode_t) newmask; 3673 } */ *uap = v; 3674 struct proc *p = l->l_proc; 3675 struct cwdinfo *cwdi; 3676 3677 cwdi = p->p_cwdi; 3678 *retval = cwdi->cwdi_cmask; 3679 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3680 return (0); 3681 } 3682 3683 /* 3684 * Void all references to file by ripping underlying filesystem 3685 * away from vnode. 3686 */ 3687 /* ARGSUSED */ 3688 int 3689 sys_revoke(struct lwp *l, void *v, register_t *retval) 3690 { 3691 struct sys_revoke_args /* { 3692 syscallarg(const char *) path; 3693 } */ *uap = v; 3694 struct mount *mp; 3695 struct vnode *vp; 3696 struct vattr vattr; 3697 int error; 3698 struct nameidata nd; 3699 3700 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3701 if ((error = namei(&nd)) != 0) 3702 return (error); 3703 vp = nd.ni_vp; 3704 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3705 goto out; 3706 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3707 (error = kauth_authorize_generic(l->l_cred, 3708 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 3709 goto out; 3710 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3711 goto out; 3712 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3713 VOP_REVOKE(vp, REVOKEALL); 3714 vn_finished_write(mp, 0); 3715 out: 3716 vrele(vp); 3717 return (error); 3718 } 3719 3720 /* 3721 * Convert a user file descriptor to a kernel file entry. 3722 */ 3723 int 3724 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3725 { 3726 struct vnode *vp; 3727 struct file *fp; 3728 3729 if ((fp = fd_getfile(fdp, fd)) == NULL) 3730 return (EBADF); 3731 3732 FILE_USE(fp); 3733 3734 if (fp->f_type != DTYPE_VNODE) { 3735 FILE_UNUSE(fp, NULL); 3736 return (EINVAL); 3737 } 3738 3739 vp = (struct vnode *)fp->f_data; 3740 if (vp->v_type == VBAD) { 3741 FILE_UNUSE(fp, NULL); 3742 return (EBADF); 3743 } 3744 3745 *fpp = fp; 3746 return (0); 3747 } 3748
Indexes created Fri Sep 26 20:09:58 GMT 2025