vfs_syscalls.c revision 1.315
1 /* $NetBSD: vfs_syscalls.c,v 1.315 2007/05/19 22:11:23 christos 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 1.315 2007/05/19 22:11:23 christos 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/syscallargs.h> 65 #include <sys/vfs_syscalls.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #include <sys/verified_exec.h> 73 #include <sys/kauth.h> 74 75 #include <miscfs/genfs/genfs.h> 76 #include <miscfs/syncfs/syncfs.h> 77 78 #ifdef COMPAT_30 79 #include "opt_nfsserver.h" 80 #include <nfs/rpcv2.h> 81 #endif 82 #include <nfs/nfsproto.h> 83 #ifdef COMPAT_30 84 #include <nfs/nfs.h> 85 #include <nfs/nfs_var.h> 86 #endif 87 88 #if NFSS > 0 89 #include <dev/fssvar.h> 90 #endif 91 92 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 93 94 static int change_dir(struct nameidata *, struct lwp *); 95 static int change_flags(struct vnode *, u_long, struct lwp *); 96 static int change_mode(struct vnode *, int, struct lwp *l); 97 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 98 static int rename_files(const char *, const char *, struct lwp *, int); 99 100 void checkdirs(struct vnode *); 101 102 int dovfsusermount = 0; 103 104 /* 105 * Virtual File System System Calls 106 */ 107 108 /* 109 * Mount a file system. 110 */ 111 112 #if defined(COMPAT_09) || defined(COMPAT_43) 113 /* 114 * This table is used to maintain compatibility with 4.3BSD 115 * and NetBSD 0.9 mount syscalls. Note, the order is important! 116 * 117 * Do not modify this table. It should only contain filesystems 118 * supported by NetBSD 0.9 and 4.3BSD. 119 */ 120 const char * const mountcompatnames[] = { 121 NULL, /* 0 = MOUNT_NONE */ 122 MOUNT_FFS, /* 1 = MOUNT_UFS */ 123 MOUNT_NFS, /* 2 */ 124 MOUNT_MFS, /* 3 */ 125 MOUNT_MSDOS, /* 4 */ 126 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 127 MOUNT_FDESC, /* 6 */ 128 MOUNT_KERNFS, /* 7 */ 129 NULL, /* 8 = MOUNT_DEVFS */ 130 MOUNT_AFS, /* 9 */ 131 }; 132 const int nmountcompatnames = sizeof(mountcompatnames) / 133 sizeof(mountcompatnames[0]); 134 #endif /* COMPAT_09 || COMPAT_43 */ 135 136 static int 137 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 138 void *data, struct nameidata *ndp) 139 { 140 struct mount *mp; 141 int error = 0, saved_flags; 142 143 mp = vp->v_mount; 144 saved_flags = mp->mnt_flag; 145 146 /* We can operate only on VROOT nodes. */ 147 if ((vp->v_flag & VROOT) == 0) { 148 error = EINVAL; 149 goto out; 150 } 151 152 /* 153 * We only allow the filesystem to be reloaded if it 154 * is currently mounted read-only. 155 */ 156 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 157 error = EOPNOTSUPP; /* Needs translation */ 158 goto out; 159 } 160 161 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 162 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 163 if (error) 164 goto out; 165 166 if (vfs_busy(mp, LK_NOWAIT, 0)) { 167 error = EPERM; 168 goto out; 169 } 170 171 mp->mnt_flag &= ~MNT_OP_FLAGS; 172 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 173 174 /* 175 * Set the mount level flags. 176 */ 177 if (flags & MNT_RDONLY) 178 mp->mnt_flag |= MNT_RDONLY; 179 else if (mp->mnt_flag & MNT_RDONLY) 180 mp->mnt_iflag |= IMNT_WANTRDWR; 181 mp->mnt_flag &= 182 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 183 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 184 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 185 mp->mnt_flag |= flags & 186 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 187 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 188 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 189 MNT_IGNORE); 190 191 error = VFS_MOUNT(mp, path, data, ndp, l); 192 193 #if defined(COMPAT_30) && defined(NFSSERVER) 194 if (error) { 195 int error2; 196 197 /* Update failed; let's try and see if it was an 198 * export request. */ 199 error2 = nfs_update_exports_30(mp, path, data, l); 200 201 /* Only update error code if the export request was 202 * understood but some problem occurred while 203 * processing it. */ 204 if (error2 != EJUSTRETURN) 205 error = error2; 206 } 207 #endif 208 if (mp->mnt_iflag & IMNT_WANTRDWR) 209 mp->mnt_flag &= ~MNT_RDONLY; 210 if (error) 211 mp->mnt_flag = saved_flags; 212 mp->mnt_flag &= ~MNT_OP_FLAGS; 213 mp->mnt_iflag &= ~IMNT_WANTRDWR; 214 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 215 if (mp->mnt_syncer == NULL) 216 error = vfs_allocate_syncvnode(mp); 217 } else { 218 if (mp->mnt_syncer != NULL) 219 vfs_deallocate_syncvnode(mp); 220 } 221 vfs_unbusy(mp); 222 223 out: 224 return (error); 225 } 226 227 static int 228 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 229 const char *path, int flags, void *data, struct nameidata *ndp) 230 { 231 struct mount *mp = NULL; 232 struct vattr va; 233 char fstypename[MFSNAMELEN]; 234 int error; 235 236 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 237 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 238 if (error) { 239 vput(vp); 240 goto out; 241 } 242 243 /* Can't make a non-dir a mount-point (from here anyway). */ 244 if (vp->v_type != VDIR) { 245 error = ENOTDIR; 246 vput(vp); 247 goto out; 248 } 249 250 /* 251 * If the user is not root, ensure that they own the directory 252 * onto which we are attempting to mount. 253 */ 254 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 255 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 256 (error = kauth_authorize_generic(l->l_cred, 257 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 258 vput(vp); 259 goto out; 260 } 261 262 if (flags & MNT_EXPORTED) { 263 error = EINVAL; 264 vput(vp); 265 goto out; 266 } 267 268 /* 269 * Copy file-system type from userspace. 270 */ 271 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 272 if (error) { 273 #if defined(COMPAT_09) || defined(COMPAT_43) 274 /* 275 * Historically, filesystem types were identified by numbers. 276 * If we get an integer for the filesystem type instead of a 277 * string, we check to see if it matches one of the historic 278 * filesystem types. 279 */ 280 u_long fsindex = (u_long)fstype; 281 if (fsindex >= nmountcompatnames || 282 mountcompatnames[fsindex] == NULL) { 283 error = ENODEV; 284 vput(vp); 285 goto out; 286 } 287 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 288 #else 289 vput(vp); 290 goto out; 291 #endif 292 } 293 294 #ifdef COMPAT_10 295 /* Accept `ufs' as an alias for `ffs'. */ 296 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 297 strlcpy(fstypename, "ffs", sizeof(fstypename)); 298 #endif 299 300 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 301 vput(vp); 302 goto out; 303 } 304 305 /* 306 * Check if a file-system is not already mounted on this vnode. 307 */ 308 if (vp->v_mountedhere != NULL) { 309 error = EBUSY; 310 vput(vp); 311 goto out; 312 } 313 314 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 315 316 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 317 free(mp, M_MOUNT); 318 error = ENODEV; 319 vput(vp); 320 goto out; 321 } 322 323 TAILQ_INIT(&mp->mnt_vnodelist); 324 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 325 simple_lock_init(&mp->mnt_slock); 326 (void)vfs_busy(mp, LK_NOWAIT, 0); 327 328 mp->mnt_op->vfs_refcount++; 329 mp->mnt_vnodecovered = vp; 330 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 331 mp->mnt_unmounter = NULL; 332 mount_initspecific(mp); 333 334 /* 335 * The underlying file system may refuse the mount for 336 * various reasons. Allow the user to force it to happen. 337 * 338 * Set the mount level flags. 339 */ 340 mp->mnt_flag = flags & 341 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 342 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 343 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 344 MNT_IGNORE | MNT_RDONLY); 345 346 error = VFS_MOUNT(mp, path, data, ndp, l); 347 mp->mnt_flag &= ~MNT_OP_FLAGS; 348 349 /* 350 * Put the new filesystem on the mount list after root. 351 */ 352 cache_purge(vp); 353 if (!error) { 354 mp->mnt_iflag &= ~IMNT_WANTRDWR; 355 vp->v_mountedhere = mp; 356 simple_lock(&mountlist_slock); 357 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 358 simple_unlock(&mountlist_slock); 359 VOP_UNLOCK(vp, 0); 360 checkdirs(vp); 361 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 362 error = vfs_allocate_syncvnode(mp); 363 vfs_unbusy(mp); 364 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 365 error = VFS_START(mp, 0, l); 366 if (error) 367 vrele(vp); 368 } else { 369 vp->v_mountedhere = NULL; 370 mp->mnt_op->vfs_refcount--; 371 vfs_unbusy(mp); 372 free(mp, M_MOUNT); 373 vput(vp); 374 } 375 376 out: 377 return (error); 378 } 379 380 static int 381 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 382 void *data, struct nameidata *ndp) 383 { 384 struct mount *mp; 385 int error; 386 387 /* If MNT_GETARGS is specified, it should be the only flag. */ 388 if (flags & ~MNT_GETARGS) { 389 error = EINVAL; 390 goto out; 391 } 392 393 mp = vp->v_mount; 394 395 /* XXX: probably some notion of "can see" here if we want isolation. */ 396 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 397 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 398 if (error) 399 goto out; 400 401 if ((vp->v_flag & VROOT) == 0) { 402 error = EINVAL; 403 goto out; 404 } 405 406 if (vfs_busy(mp, LK_NOWAIT, 0)) { 407 error = EPERM; 408 goto out; 409 } 410 411 mp->mnt_flag &= ~MNT_OP_FLAGS; 412 mp->mnt_flag |= MNT_GETARGS; 413 error = VFS_MOUNT(mp, path, data, ndp, l); 414 mp->mnt_flag &= ~MNT_OP_FLAGS; 415 416 vfs_unbusy(mp); 417 out: 418 return (error); 419 } 420 421 /* ARGSUSED */ 422 int 423 sys_mount(struct lwp *l, void *v, register_t *retval) 424 { 425 struct sys_mount_args /* { 426 syscallarg(const char *) type; 427 syscallarg(const char *) path; 428 syscallarg(int) flags; 429 syscallarg(void *) data; 430 } */ *uap = v; 431 struct vnode *vp; 432 struct nameidata nd; 433 int error; 434 435 /* 436 * Get vnode to be covered 437 */ 438 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 439 SCARG(uap, path), l); 440 if ((error = namei(&nd)) != 0) 441 return (error); 442 vp = nd.ni_vp; 443 444 /* 445 * A lookup in VFS_MOUNT might result in an attempt to 446 * lock this vnode again, so make the lock recursive. 447 */ 448 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 449 450 if (SCARG(uap, flags) & MNT_GETARGS) { 451 error = mount_getargs(l, vp, SCARG(uap, path), 452 SCARG(uap, flags), SCARG(uap, data), &nd); 453 vput(vp); 454 } else if (SCARG(uap, flags) & MNT_UPDATE) { 455 error = mount_update(l, vp, SCARG(uap, path), 456 SCARG(uap, flags), SCARG(uap, data), &nd); 457 vput(vp); 458 } else { 459 /* Locking is handled internally in mount_domount(). */ 460 error = mount_domount(l, vp, SCARG(uap, type), 461 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 462 } 463 464 return (error); 465 } 466 467 /* 468 * Scan all active processes to see if any of them have a current 469 * or root directory onto which the new filesystem has just been 470 * mounted. If so, replace them with the new mount point. 471 */ 472 void 473 checkdirs(struct vnode *olddp) 474 { 475 struct cwdinfo *cwdi; 476 struct vnode *newdp; 477 struct proc *p; 478 479 if (olddp->v_usecount == 1) 480 return; 481 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 482 panic("mount: lost mount"); 483 mutex_enter(&proclist_lock); 484 PROCLIST_FOREACH(p, &allproc) { 485 cwdi = p->p_cwdi; 486 if (!cwdi) 487 continue; 488 if (cwdi->cwdi_cdir == olddp) { 489 vrele(cwdi->cwdi_cdir); 490 VREF(newdp); 491 cwdi->cwdi_cdir = newdp; 492 } 493 if (cwdi->cwdi_rdir == olddp) { 494 vrele(cwdi->cwdi_rdir); 495 VREF(newdp); 496 cwdi->cwdi_rdir = newdp; 497 } 498 } 499 mutex_exit(&proclist_lock); 500 if (rootvnode == olddp) { 501 vrele(rootvnode); 502 VREF(newdp); 503 rootvnode = newdp; 504 } 505 vput(newdp); 506 } 507 508 /* 509 * Unmount a file system. 510 * 511 * Note: unmount takes a path to the vnode mounted on as argument, 512 * not special file (as before). 513 */ 514 /* ARGSUSED */ 515 int 516 sys_unmount(struct lwp *l, void *v, register_t *retval) 517 { 518 struct sys_unmount_args /* { 519 syscallarg(const char *) path; 520 syscallarg(int) flags; 521 } */ *uap = v; 522 struct vnode *vp; 523 struct mount *mp; 524 int error; 525 struct nameidata nd; 526 527 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 528 SCARG(uap, path), l); 529 if ((error = namei(&nd)) != 0) 530 return (error); 531 vp = nd.ni_vp; 532 mp = vp->v_mount; 533 534 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 535 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 536 if (error) { 537 vput(vp); 538 return (error); 539 } 540 541 /* 542 * Don't allow unmounting the root file system. 543 */ 544 if (mp->mnt_flag & MNT_ROOTFS) { 545 vput(vp); 546 return (EINVAL); 547 } 548 549 /* 550 * Must be the root of the filesystem 551 */ 552 if ((vp->v_flag & VROOT) == 0) { 553 vput(vp); 554 return (EINVAL); 555 } 556 vput(vp); 557 558 /* 559 * XXX Freeze syncer. Must do this before locking the 560 * mount point. See dounmount() for details. 561 */ 562 mutex_enter(&syncer_mutex); 563 564 if (vfs_busy(mp, 0, 0)) { 565 mutex_exit(&syncer_mutex); 566 return (EBUSY); 567 } 568 569 return (dounmount(mp, SCARG(uap, flags), l)); 570 } 571 572 /* 573 * Do the actual file system unmount. File system is assumed to have been 574 * marked busy by the caller. 575 */ 576 int 577 dounmount(struct mount *mp, int flags, struct lwp *l) 578 { 579 struct vnode *coveredvp; 580 int error; 581 int async; 582 int used_syncer; 583 584 #if NVERIEXEC > 0 585 error = veriexec_unmountchk(mp); 586 if (error) 587 return (error); 588 #endif /* NVERIEXEC > 0 */ 589 590 simple_lock(&mountlist_slock); 591 vfs_unbusy(mp); 592 used_syncer = (mp->mnt_syncer != NULL); 593 594 /* 595 * XXX Syncer must be frozen when we get here. This should really 596 * be done on a per-mountpoint basis, but especially the softdep 597 * code possibly called from the syncer doesn't exactly work on a 598 * per-mountpoint basis, so the softdep code would become a maze 599 * of vfs_busy() calls. 600 * 601 * The caller of dounmount() must acquire syncer_mutex because 602 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 603 * order, and we must preserve that order to avoid deadlock. 604 * 605 * So, if the file system did not use the syncer, now is 606 * the time to release the syncer_mutex. 607 */ 608 if (used_syncer == 0) 609 mutex_exit(&syncer_mutex); 610 611 mp->mnt_iflag |= IMNT_UNMOUNT; 612 mp->mnt_unmounter = l; 613 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 614 615 async = mp->mnt_flag & MNT_ASYNC; 616 mp->mnt_flag &= ~MNT_ASYNC; 617 cache_purgevfs(mp); /* remove cache entries for this file sys */ 618 if (mp->mnt_syncer != NULL) 619 vfs_deallocate_syncvnode(mp); 620 error = 0; 621 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 622 #if NFSS > 0 623 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 624 #endif 625 if (error == 0) 626 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 627 } 628 if (error == 0 || (flags & MNT_FORCE)) 629 error = VFS_UNMOUNT(mp, flags, l); 630 simple_lock(&mountlist_slock); 631 if (error) { 632 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 633 (void) vfs_allocate_syncvnode(mp); 634 mp->mnt_iflag &= ~IMNT_UNMOUNT; 635 mp->mnt_unmounter = NULL; 636 mp->mnt_flag |= async; 637 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 638 &mountlist_slock); 639 if (used_syncer) 640 mutex_exit(&syncer_mutex); 641 simple_lock(&mp->mnt_slock); 642 while (mp->mnt_wcnt > 0) { 643 wakeup(mp); 644 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 645 0, &mp->mnt_slock); 646 } 647 simple_unlock(&mp->mnt_slock); 648 return (error); 649 } 650 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 651 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 652 coveredvp->v_mountedhere = NULL; 653 mp->mnt_op->vfs_refcount--; 654 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 655 panic("unmount: dangling vnode"); 656 mp->mnt_iflag |= IMNT_GONE; 657 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 658 if (coveredvp != NULLVP) 659 vrele(coveredvp); 660 mount_finispecific(mp); 661 if (used_syncer) 662 mutex_exit(&syncer_mutex); 663 simple_lock(&mp->mnt_slock); 664 while (mp->mnt_wcnt > 0) { 665 wakeup(mp); 666 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 667 } 668 simple_unlock(&mp->mnt_slock); 669 vfs_hooks_unmount(mp); 670 free(mp, M_MOUNT); 671 return (0); 672 } 673 674 /* 675 * Sync each mounted filesystem. 676 */ 677 #ifdef DEBUG 678 int syncprt = 0; 679 struct ctldebug debug0 = { "syncprt", &syncprt }; 680 #endif 681 682 /* ARGSUSED */ 683 int 684 sys_sync(struct lwp *l, void *v, register_t *retval) 685 { 686 struct mount *mp, *nmp; 687 int asyncflag; 688 689 if (l == NULL) 690 l = &lwp0; 691 692 simple_lock(&mountlist_slock); 693 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 694 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 695 nmp = mp->mnt_list.cqe_prev; 696 continue; 697 } 698 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 699 asyncflag = mp->mnt_flag & MNT_ASYNC; 700 mp->mnt_flag &= ~MNT_ASYNC; 701 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 702 if (asyncflag) 703 mp->mnt_flag |= MNT_ASYNC; 704 } 705 simple_lock(&mountlist_slock); 706 nmp = mp->mnt_list.cqe_prev; 707 vfs_unbusy(mp); 708 709 } 710 simple_unlock(&mountlist_slock); 711 #ifdef DEBUG 712 if (syncprt) 713 vfs_bufstats(); 714 #endif /* DEBUG */ 715 return (0); 716 } 717 718 /* 719 * Change filesystem quotas. 720 */ 721 /* ARGSUSED */ 722 int 723 sys_quotactl(struct lwp *l, void *v, register_t *retval) 724 { 725 struct sys_quotactl_args /* { 726 syscallarg(const char *) path; 727 syscallarg(int) cmd; 728 syscallarg(int) uid; 729 syscallarg(void *) arg; 730 } */ *uap = v; 731 struct mount *mp; 732 int error; 733 struct nameidata nd; 734 735 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 736 if ((error = namei(&nd)) != 0) 737 return (error); 738 mp = nd.ni_vp->v_mount; 739 vrele(nd.ni_vp); 740 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 741 SCARG(uap, arg), l); 742 return (error); 743 } 744 745 int 746 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 747 int root) 748 { 749 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 750 int error = 0; 751 752 /* 753 * If MNT_NOWAIT or MNT_LAZY is specified, do not 754 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 755 * overrides MNT_NOWAIT. 756 */ 757 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 758 (flags != MNT_WAIT && flags != 0)) { 759 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 760 goto done; 761 } 762 763 /* Get the filesystem stats now */ 764 memset(sp, 0, sizeof(*sp)); 765 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 766 return error; 767 } 768 769 if (cwdi->cwdi_rdir == NULL) 770 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 771 done: 772 if (cwdi->cwdi_rdir != NULL) { 773 size_t len; 774 char *bp; 775 char *path = PNBUF_GET(); 776 777 bp = path + MAXPATHLEN; 778 *--bp = '\0'; 779 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 780 MAXPATHLEN / 2, 0, l); 781 if (error) { 782 PNBUF_PUT(path); 783 return error; 784 } 785 len = strlen(bp); 786 /* 787 * for mount points that are below our root, we can see 788 * them, so we fix up the pathname and return them. The 789 * rest we cannot see, so we don't allow viewing the 790 * data. 791 */ 792 if (strncmp(bp, sp->f_mntonname, len) == 0) { 793 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 794 sizeof(sp->f_mntonname)); 795 if (sp->f_mntonname[0] == '\0') 796 (void)strlcpy(sp->f_mntonname, "/", 797 sizeof(sp->f_mntonname)); 798 } else { 799 if (root) 800 (void)strlcpy(sp->f_mntonname, "/", 801 sizeof(sp->f_mntonname)); 802 else 803 error = EPERM; 804 } 805 PNBUF_PUT(path); 806 } 807 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 808 return error; 809 } 810 811 /* 812 * Get filesystem statistics by path. 813 */ 814 int 815 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 816 { 817 struct mount *mp; 818 int error; 819 struct nameidata nd; 820 821 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 822 if ((error = namei(&nd)) != 0) 823 return error; 824 mp = nd.ni_vp->v_mount; 825 error = dostatvfs(mp, sb, l, flags, 1); 826 vrele(nd.ni_vp); 827 return error; 828 } 829 830 /* ARGSUSED */ 831 int 832 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 833 { 834 struct sys_statvfs1_args /* { 835 syscallarg(const char *) path; 836 syscallarg(struct statvfs *) buf; 837 syscallarg(int) flags; 838 } */ *uap = v; 839 struct statvfs *sb; 840 int error; 841 842 sb = STATVFSBUF_GET(); 843 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 844 if (error == 0) 845 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 846 STATVFSBUF_PUT(sb); 847 return error; 848 } 849 850 /* 851 * Get filesystem statistics by fd. 852 */ 853 int 854 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 855 { 856 struct proc *p = l->l_proc; 857 struct file *fp; 858 struct mount *mp; 859 int error; 860 861 /* getvnode() will use the descriptor for us */ 862 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 863 return (error); 864 mp = ((struct vnode *)fp->f_data)->v_mount; 865 error = dostatvfs(mp, sb, l, flags, 1); 866 FILE_UNUSE(fp, l); 867 return error; 868 } 869 870 /* ARGSUSED */ 871 int 872 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 873 { 874 struct sys_fstatvfs1_args /* { 875 syscallarg(int) fd; 876 syscallarg(struct statvfs *) buf; 877 syscallarg(int) flags; 878 } */ *uap = v; 879 struct statvfs *sb; 880 int error; 881 882 sb = STATVFSBUF_GET(); 883 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 884 if (error != 0) 885 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 886 STATVFSBUF_PUT(sb); 887 return error; 888 } 889 890 891 /* 892 * Get statistics on all filesystems. 893 */ 894 int 895 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 896 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 897 register_t *retval) 898 { 899 int root = 0; 900 struct proc *p = l->l_proc; 901 struct mount *mp, *nmp; 902 struct statvfs *sb; 903 size_t count, maxcount; 904 int error = 0; 905 906 sb = STATVFSBUF_GET(); 907 maxcount = bufsize / entry_sz; 908 simple_lock(&mountlist_slock); 909 count = 0; 910 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 911 mp = nmp) { 912 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 913 nmp = CIRCLEQ_NEXT(mp, mnt_list); 914 continue; 915 } 916 if (sfsp && count < maxcount) { 917 error = dostatvfs(mp, sb, l, flags, 0); 918 if (error) { 919 simple_lock(&mountlist_slock); 920 nmp = CIRCLEQ_NEXT(mp, mnt_list); 921 vfs_unbusy(mp); 922 continue; 923 } 924 error = copyfn(sb, sfsp, entry_sz); 925 if (error) { 926 vfs_unbusy(mp); 927 goto out; 928 } 929 sfsp = (char *)sfsp + entry_sz; 930 root |= strcmp(sb->f_mntonname, "/") == 0; 931 } 932 count++; 933 simple_lock(&mountlist_slock); 934 nmp = CIRCLEQ_NEXT(mp, mnt_list); 935 vfs_unbusy(mp); 936 } 937 simple_unlock(&mountlist_slock); 938 if (root == 0 && p->p_cwdi->cwdi_rdir) { 939 /* 940 * fake a root entry 941 */ 942 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 943 if (error != 0) 944 goto out; 945 if (sfsp) 946 error = copyfn(sb, sfsp, entry_sz); 947 count++; 948 } 949 if (sfsp && count > maxcount) 950 *retval = maxcount; 951 else 952 *retval = count; 953 out: 954 STATVFSBUF_PUT(sb); 955 return error; 956 } 957 958 int 959 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 960 { 961 struct sys_getvfsstat_args /* { 962 syscallarg(struct statvfs *) buf; 963 syscallarg(size_t) bufsize; 964 syscallarg(int) flags; 965 } */ *uap = v; 966 967 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 968 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 969 } 970 971 /* 972 * Change current working directory to a given file descriptor. 973 */ 974 /* ARGSUSED */ 975 int 976 sys_fchdir(struct lwp *l, void *v, register_t *retval) 977 { 978 struct sys_fchdir_args /* { 979 syscallarg(int) fd; 980 } */ *uap = v; 981 struct proc *p = l->l_proc; 982 struct filedesc *fdp = p->p_fd; 983 struct cwdinfo *cwdi = p->p_cwdi; 984 struct vnode *vp, *tdp; 985 struct mount *mp; 986 struct file *fp; 987 int error; 988 989 /* getvnode() will use the descriptor for us */ 990 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 991 return (error); 992 vp = (struct vnode *)fp->f_data; 993 994 VREF(vp); 995 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 996 if (vp->v_type != VDIR) 997 error = ENOTDIR; 998 else 999 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1000 if (error) { 1001 vput(vp); 1002 goto out; 1003 } 1004 while ((mp = vp->v_mountedhere) != NULL) { 1005 if (vfs_busy(mp, 0, 0)) 1006 continue; 1007 1008 vput(vp); 1009 error = VFS_ROOT(mp, &tdp); 1010 vfs_unbusy(mp); 1011 if (error) 1012 goto out; 1013 vp = tdp; 1014 } 1015 VOP_UNLOCK(vp, 0); 1016 1017 /* 1018 * Disallow changing to a directory not under the process's 1019 * current root directory (if there is one). 1020 */ 1021 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1022 vrele(vp); 1023 error = EPERM; /* operation not permitted */ 1024 goto out; 1025 } 1026 1027 vrele(cwdi->cwdi_cdir); 1028 cwdi->cwdi_cdir = vp; 1029 out: 1030 FILE_UNUSE(fp, l); 1031 return (error); 1032 } 1033 1034 /* 1035 * Change this process's notion of the root directory to a given file 1036 * descriptor. 1037 */ 1038 int 1039 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1040 { 1041 struct sys_fchroot_args *uap = v; 1042 struct proc *p = l->l_proc; 1043 struct filedesc *fdp = p->p_fd; 1044 struct cwdinfo *cwdi = p->p_cwdi; 1045 struct vnode *vp; 1046 struct file *fp; 1047 int error; 1048 1049 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1050 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1051 return error; 1052 /* getvnode() will use the descriptor for us */ 1053 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1054 return error; 1055 vp = (struct vnode *) fp->f_data; 1056 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1057 if (vp->v_type != VDIR) 1058 error = ENOTDIR; 1059 else 1060 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1061 VOP_UNLOCK(vp, 0); 1062 if (error) 1063 goto out; 1064 VREF(vp); 1065 1066 /* 1067 * Prevent escaping from chroot by putting the root under 1068 * the working directory. Silently chdir to / if we aren't 1069 * already there. 1070 */ 1071 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1072 /* 1073 * XXX would be more failsafe to change directory to a 1074 * deadfs node here instead 1075 */ 1076 vrele(cwdi->cwdi_cdir); 1077 VREF(vp); 1078 cwdi->cwdi_cdir = vp; 1079 } 1080 1081 if (cwdi->cwdi_rdir != NULL) 1082 vrele(cwdi->cwdi_rdir); 1083 cwdi->cwdi_rdir = vp; 1084 out: 1085 FILE_UNUSE(fp, l); 1086 return (error); 1087 } 1088 1089 /* 1090 * Change current working directory (``.''). 1091 */ 1092 /* ARGSUSED */ 1093 int 1094 sys_chdir(struct lwp *l, void *v, register_t *retval) 1095 { 1096 struct sys_chdir_args /* { 1097 syscallarg(const char *) path; 1098 } */ *uap = v; 1099 struct proc *p = l->l_proc; 1100 struct cwdinfo *cwdi = p->p_cwdi; 1101 int error; 1102 struct nameidata nd; 1103 1104 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1105 SCARG(uap, path), l); 1106 if ((error = change_dir(&nd, l)) != 0) 1107 return (error); 1108 vrele(cwdi->cwdi_cdir); 1109 cwdi->cwdi_cdir = nd.ni_vp; 1110 return (0); 1111 } 1112 1113 /* 1114 * Change notion of root (``/'') directory. 1115 */ 1116 /* ARGSUSED */ 1117 int 1118 sys_chroot(struct lwp *l, void *v, register_t *retval) 1119 { 1120 struct sys_chroot_args /* { 1121 syscallarg(const char *) path; 1122 } */ *uap = v; 1123 struct proc *p = l->l_proc; 1124 struct cwdinfo *cwdi = p->p_cwdi; 1125 struct vnode *vp; 1126 int error; 1127 struct nameidata nd; 1128 1129 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1130 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1131 return (error); 1132 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1133 SCARG(uap, path), l); 1134 if ((error = change_dir(&nd, l)) != 0) 1135 return (error); 1136 if (cwdi->cwdi_rdir != NULL) 1137 vrele(cwdi->cwdi_rdir); 1138 vp = nd.ni_vp; 1139 cwdi->cwdi_rdir = vp; 1140 1141 /* 1142 * Prevent escaping from chroot by putting the root under 1143 * the working directory. Silently chdir to / if we aren't 1144 * already there. 1145 */ 1146 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1147 /* 1148 * XXX would be more failsafe to change directory to a 1149 * deadfs node here instead 1150 */ 1151 vrele(cwdi->cwdi_cdir); 1152 VREF(vp); 1153 cwdi->cwdi_cdir = vp; 1154 } 1155 1156 return (0); 1157 } 1158 1159 /* 1160 * Common routine for chroot and chdir. 1161 */ 1162 static int 1163 change_dir(struct nameidata *ndp, struct lwp *l) 1164 { 1165 struct vnode *vp; 1166 int error; 1167 1168 if ((error = namei(ndp)) != 0) 1169 return (error); 1170 vp = ndp->ni_vp; 1171 if (vp->v_type != VDIR) 1172 error = ENOTDIR; 1173 else 1174 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1175 1176 if (error) 1177 vput(vp); 1178 else 1179 VOP_UNLOCK(vp, 0); 1180 return (error); 1181 } 1182 1183 /* 1184 * Check permissions, allocate an open file structure, 1185 * and call the device open routine if any. 1186 */ 1187 int 1188 sys_open(struct lwp *l, void *v, register_t *retval) 1189 { 1190 struct sys_open_args /* { 1191 syscallarg(const char *) path; 1192 syscallarg(int) flags; 1193 syscallarg(int) mode; 1194 } */ *uap = v; 1195 struct proc *p = l->l_proc; 1196 struct cwdinfo *cwdi = p->p_cwdi; 1197 struct filedesc *fdp = p->p_fd; 1198 struct file *fp; 1199 struct vnode *vp; 1200 int flags, cmode; 1201 int type, indx, error; 1202 struct flock lf; 1203 struct nameidata nd; 1204 1205 flags = FFLAGS(SCARG(uap, flags)); 1206 if ((flags & (FREAD | FWRITE)) == 0) 1207 return (EINVAL); 1208 /* falloc() will use the file descriptor for us */ 1209 if ((error = falloc(l, &fp, &indx)) != 0) 1210 return (error); 1211 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1212 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1213 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1214 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1215 FILE_UNUSE(fp, l); 1216 fdp->fd_ofiles[indx] = NULL; 1217 ffree(fp); 1218 if ((error == EDUPFD || error == EMOVEFD) && 1219 l->l_dupfd >= 0 && /* XXX from fdopen */ 1220 (error = 1221 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1222 *retval = indx; 1223 return (0); 1224 } 1225 if (error == ERESTART) 1226 error = EINTR; 1227 fdremove(fdp, indx); 1228 return (error); 1229 } 1230 l->l_dupfd = 0; 1231 vp = nd.ni_vp; 1232 fp->f_flag = flags & FMASK; 1233 fp->f_type = DTYPE_VNODE; 1234 fp->f_ops = &vnops; 1235 fp->f_data = vp; 1236 if (flags & (O_EXLOCK | O_SHLOCK)) { 1237 lf.l_whence = SEEK_SET; 1238 lf.l_start = 0; 1239 lf.l_len = 0; 1240 if (flags & O_EXLOCK) 1241 lf.l_type = F_WRLCK; 1242 else 1243 lf.l_type = F_RDLCK; 1244 type = F_FLOCK; 1245 if ((flags & FNONBLOCK) == 0) 1246 type |= F_WAIT; 1247 VOP_UNLOCK(vp, 0); 1248 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1249 if (error) { 1250 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1251 FILE_UNUSE(fp, l); 1252 ffree(fp); 1253 fdremove(fdp, indx); 1254 return (error); 1255 } 1256 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1257 fp->f_flag |= FHASLOCK; 1258 } 1259 VOP_UNLOCK(vp, 0); 1260 *retval = indx; 1261 FILE_SET_MATURE(fp); 1262 FILE_UNUSE(fp, l); 1263 return (0); 1264 } 1265 1266 static void 1267 vfs__fhfree(fhandle_t *fhp) 1268 { 1269 size_t fhsize; 1270 1271 if (fhp == NULL) { 1272 return; 1273 } 1274 fhsize = FHANDLE_SIZE(fhp); 1275 kmem_free(fhp, fhsize); 1276 } 1277 1278 /* 1279 * vfs_composefh: compose a filehandle. 1280 */ 1281 1282 int 1283 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1284 { 1285 struct mount *mp; 1286 struct fid *fidp; 1287 int error; 1288 size_t needfhsize; 1289 size_t fidsize; 1290 1291 mp = vp->v_mount; 1292 fidp = NULL; 1293 if (*fh_size < FHANDLE_SIZE_MIN) { 1294 fidsize = 0; 1295 } else { 1296 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1297 if (fhp != NULL) { 1298 memset(fhp, 0, *fh_size); 1299 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1300 fidp = &fhp->fh_fid; 1301 } 1302 } 1303 error = VFS_VPTOFH(vp, fidp, &fidsize); 1304 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1305 if (error == 0 && *fh_size < needfhsize) { 1306 error = E2BIG; 1307 } 1308 *fh_size = needfhsize; 1309 return error; 1310 } 1311 1312 int 1313 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1314 { 1315 struct mount *mp; 1316 fhandle_t *fhp; 1317 size_t fhsize; 1318 size_t fidsize; 1319 int error; 1320 1321 *fhpp = NULL; 1322 mp = vp->v_mount; 1323 fidsize = 0; 1324 error = VFS_VPTOFH(vp, NULL, &fidsize); 1325 KASSERT(error != 0); 1326 if (error != E2BIG) { 1327 goto out; 1328 } 1329 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1330 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1331 if (fhp == NULL) { 1332 error = ENOMEM; 1333 goto out; 1334 } 1335 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1336 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1337 if (error == 0) { 1338 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1339 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1340 *fhpp = fhp; 1341 } else { 1342 kmem_free(fhp, fhsize); 1343 } 1344 out: 1345 return error; 1346 } 1347 1348 void 1349 vfs_composefh_free(fhandle_t *fhp) 1350 { 1351 1352 vfs__fhfree(fhp); 1353 } 1354 1355 /* 1356 * vfs_fhtovp: lookup a vnode by a filehandle. 1357 */ 1358 1359 int 1360 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1361 { 1362 struct mount *mp; 1363 int error; 1364 1365 *vpp = NULL; 1366 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1367 if (mp == NULL) { 1368 error = ESTALE; 1369 goto out; 1370 } 1371 if (mp->mnt_op->vfs_fhtovp == NULL) { 1372 error = EOPNOTSUPP; 1373 goto out; 1374 } 1375 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1376 out: 1377 return error; 1378 } 1379 1380 /* 1381 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1382 * the needed size. 1383 */ 1384 1385 int 1386 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1387 { 1388 fhandle_t *fhp; 1389 int error; 1390 1391 *fhpp = NULL; 1392 if (fhsize > FHANDLE_SIZE_MAX) { 1393 return EINVAL; 1394 } 1395 if (fhsize < FHANDLE_SIZE_MIN) { 1396 return EINVAL; 1397 } 1398 again: 1399 fhp = kmem_alloc(fhsize, KM_SLEEP); 1400 if (fhp == NULL) { 1401 return ENOMEM; 1402 } 1403 error = copyin(ufhp, fhp, fhsize); 1404 if (error == 0) { 1405 /* XXX this check shouldn't be here */ 1406 if (FHANDLE_SIZE(fhp) == fhsize) { 1407 *fhpp = fhp; 1408 return 0; 1409 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1410 /* 1411 * a kludge for nfsv2 padded handles. 1412 */ 1413 size_t sz; 1414 1415 sz = FHANDLE_SIZE(fhp); 1416 kmem_free(fhp, fhsize); 1417 fhsize = sz; 1418 goto again; 1419 } else { 1420 /* 1421 * userland told us wrong size. 1422 */ 1423 error = EINVAL; 1424 } 1425 } 1426 kmem_free(fhp, fhsize); 1427 return error; 1428 } 1429 1430 void 1431 vfs_copyinfh_free(fhandle_t *fhp) 1432 { 1433 1434 vfs__fhfree(fhp); 1435 } 1436 1437 /* 1438 * Get file handle system call 1439 */ 1440 int 1441 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1442 { 1443 struct sys___getfh30_args /* { 1444 syscallarg(char *) fname; 1445 syscallarg(fhandle_t *) fhp; 1446 syscallarg(size_t *) fh_size; 1447 } */ *uap = v; 1448 struct vnode *vp; 1449 fhandle_t *fh; 1450 int error; 1451 struct nameidata nd; 1452 size_t sz; 1453 size_t usz; 1454 1455 /* 1456 * Must be super user 1457 */ 1458 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1459 0, NULL, NULL, NULL); 1460 if (error) 1461 return (error); 1462 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1463 SCARG(uap, fname), l); 1464 error = namei(&nd); 1465 if (error) 1466 return (error); 1467 vp = nd.ni_vp; 1468 error = vfs_composefh_alloc(vp, &fh); 1469 vput(vp); 1470 if (error != 0) { 1471 goto out; 1472 } 1473 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1474 if (error != 0) { 1475 goto out; 1476 } 1477 sz = FHANDLE_SIZE(fh); 1478 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1479 if (error != 0) { 1480 goto out; 1481 } 1482 if (usz >= sz) { 1483 error = copyout(fh, SCARG(uap, fhp), sz); 1484 } else { 1485 error = E2BIG; 1486 } 1487 out: 1488 vfs_composefh_free(fh); 1489 return (error); 1490 } 1491 1492 /* 1493 * Open a file given a file handle. 1494 * 1495 * Check permissions, allocate an open file structure, 1496 * and call the device open routine if any. 1497 */ 1498 1499 int 1500 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1501 register_t *retval) 1502 { 1503 struct filedesc *fdp = l->l_proc->p_fd; 1504 struct file *fp; 1505 struct vnode *vp = NULL; 1506 kauth_cred_t cred = l->l_cred; 1507 struct file *nfp; 1508 int type, indx, error=0; 1509 struct flock lf; 1510 struct vattr va; 1511 fhandle_t *fh; 1512 int flags; 1513 1514 /* 1515 * Must be super user 1516 */ 1517 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1518 0, NULL, NULL, NULL))) 1519 return (error); 1520 1521 flags = FFLAGS(oflags); 1522 if ((flags & (FREAD | FWRITE)) == 0) 1523 return (EINVAL); 1524 if ((flags & O_CREAT)) 1525 return (EINVAL); 1526 /* falloc() will use the file descriptor for us */ 1527 if ((error = falloc(l, &nfp, &indx)) != 0) 1528 return (error); 1529 fp = nfp; 1530 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1531 if (error != 0) { 1532 goto bad; 1533 } 1534 error = vfs_fhtovp(fh, &vp); 1535 if (error != 0) { 1536 goto bad; 1537 } 1538 1539 /* Now do an effective vn_open */ 1540 1541 if (vp->v_type == VSOCK) { 1542 error = EOPNOTSUPP; 1543 goto bad; 1544 } 1545 if (flags & FREAD) { 1546 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1547 goto bad; 1548 } 1549 if (flags & (FWRITE | O_TRUNC)) { 1550 if (vp->v_type == VDIR) { 1551 error = EISDIR; 1552 goto bad; 1553 } 1554 if ((error = vn_writechk(vp)) != 0 || 1555 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1556 goto bad; 1557 } 1558 if (flags & O_TRUNC) { 1559 VOP_UNLOCK(vp, 0); /* XXX */ 1560 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1561 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1562 VATTR_NULL(&va); 1563 va.va_size = 0; 1564 error = VOP_SETATTR(vp, &va, cred, l); 1565 if (error) 1566 goto bad; 1567 } 1568 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1569 goto bad; 1570 if (vp->v_type == VREG && 1571 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1572 error = EIO; 1573 goto bad; 1574 } 1575 if (flags & FWRITE) 1576 vp->v_writecount++; 1577 1578 /* done with modified vn_open, now finish what sys_open does. */ 1579 1580 fp->f_flag = flags & FMASK; 1581 fp->f_type = DTYPE_VNODE; 1582 fp->f_ops = &vnops; 1583 fp->f_data = vp; 1584 if (flags & (O_EXLOCK | O_SHLOCK)) { 1585 lf.l_whence = SEEK_SET; 1586 lf.l_start = 0; 1587 lf.l_len = 0; 1588 if (flags & O_EXLOCK) 1589 lf.l_type = F_WRLCK; 1590 else 1591 lf.l_type = F_RDLCK; 1592 type = F_FLOCK; 1593 if ((flags & FNONBLOCK) == 0) 1594 type |= F_WAIT; 1595 VOP_UNLOCK(vp, 0); 1596 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1597 if (error) { 1598 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1599 FILE_UNUSE(fp, l); 1600 ffree(fp); 1601 fdremove(fdp, indx); 1602 return (error); 1603 } 1604 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1605 fp->f_flag |= FHASLOCK; 1606 } 1607 VOP_UNLOCK(vp, 0); 1608 *retval = indx; 1609 FILE_SET_MATURE(fp); 1610 FILE_UNUSE(fp, l); 1611 vfs_copyinfh_free(fh); 1612 return (0); 1613 1614 bad: 1615 FILE_UNUSE(fp, l); 1616 ffree(fp); 1617 fdremove(fdp, indx); 1618 if (vp != NULL) 1619 vput(vp); 1620 vfs_copyinfh_free(fh); 1621 return (error); 1622 } 1623 1624 int 1625 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1626 { 1627 struct sys___fhopen40_args /* { 1628 syscallarg(const void *) fhp; 1629 syscallarg(size_t) fh_size; 1630 syscallarg(int) flags; 1631 } */ *uap = v; 1632 1633 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1634 SCARG(uap, flags), retval); 1635 } 1636 1637 int 1638 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1639 { 1640 int error; 1641 fhandle_t *fh; 1642 struct vnode *vp; 1643 1644 /* 1645 * Must be super user 1646 */ 1647 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1648 0, NULL, NULL, NULL))) 1649 return (error); 1650 1651 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1652 if (error != 0) 1653 return error; 1654 1655 error = vfs_fhtovp(fh, &vp); 1656 vfs_copyinfh_free(fh); 1657 if (error != 0) 1658 return error; 1659 1660 error = vn_stat(vp, sb, l); 1661 vput(vp); 1662 return error; 1663 } 1664 1665 1666 /* ARGSUSED */ 1667 int 1668 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1669 { 1670 struct sys___fhstat40_args /* { 1671 syscallarg(const void *) fhp; 1672 syscallarg(size_t) fh_size; 1673 syscallarg(struct stat *) sb; 1674 } */ *uap = v; 1675 struct stat sb; 1676 int error; 1677 1678 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1679 if (error) 1680 return error; 1681 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1682 } 1683 1684 int 1685 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1686 int flags) 1687 { 1688 fhandle_t *fh; 1689 struct mount *mp; 1690 struct vnode *vp; 1691 int error; 1692 1693 /* 1694 * Must be super user 1695 */ 1696 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1697 0, NULL, NULL, NULL))) 1698 return error; 1699 1700 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1701 if (error != 0) 1702 return error; 1703 1704 error = vfs_fhtovp(fh, &vp); 1705 vfs_copyinfh_free(fh); 1706 if (error != 0) 1707 return error; 1708 1709 mp = vp->v_mount; 1710 error = dostatvfs(mp, sb, l, flags, 1); 1711 vput(vp); 1712 return error; 1713 } 1714 1715 /* ARGSUSED */ 1716 int 1717 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1718 { 1719 struct sys___fhstatvfs140_args /* { 1720 syscallarg(const void *) fhp; 1721 syscallarg(size_t) fh_size; 1722 syscallarg(struct statvfs *) buf; 1723 syscallarg(int) flags; 1724 } */ *uap = v; 1725 struct statvfs *sb = STATVFSBUF_GET(); 1726 int error; 1727 1728 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1729 SCARG(uap, flags)); 1730 if (error == 0) 1731 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1732 STATVFSBUF_PUT(sb); 1733 return error; 1734 } 1735 1736 /* 1737 * Create a special file. 1738 */ 1739 /* ARGSUSED */ 1740 int 1741 sys_mknod(struct lwp *l, void *v, register_t *retval) 1742 { 1743 struct sys_mknod_args /* { 1744 syscallarg(const char *) path; 1745 syscallarg(int) mode; 1746 syscallarg(int) dev; 1747 } */ *uap = v; 1748 struct proc *p = l->l_proc; 1749 struct vnode *vp; 1750 struct vattr vattr; 1751 int error, optype; 1752 struct nameidata nd; 1753 char *path; 1754 const char *cpath; 1755 enum uio_seg seg = UIO_USERSPACE; 1756 1757 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1758 0, NULL, NULL, NULL)) != 0) 1759 return (error); 1760 1761 optype = VOP_MKNOD_DESCOFFSET; 1762 1763 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1764 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1765 1766 if ((error = namei(&nd)) != 0) 1767 goto out; 1768 vp = nd.ni_vp; 1769 if (vp != NULL) 1770 error = EEXIST; 1771 else { 1772 VATTR_NULL(&vattr); 1773 vattr.va_mode = 1774 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1775 vattr.va_rdev = SCARG(uap, dev); 1776 1777 switch (SCARG(uap, mode) & S_IFMT) { 1778 case S_IFMT: /* used by badsect to flag bad sectors */ 1779 vattr.va_type = VBAD; 1780 break; 1781 case S_IFCHR: 1782 vattr.va_type = VCHR; 1783 break; 1784 case S_IFBLK: 1785 vattr.va_type = VBLK; 1786 break; 1787 case S_IFWHT: 1788 optype = VOP_WHITEOUT_DESCOFFSET; 1789 break; 1790 case S_IFREG: 1791 #if NVERIEXEC > 0 1792 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1793 O_CREAT); 1794 #endif /* NVERIEXEC > 0 */ 1795 vattr.va_type = VREG; 1796 vattr.va_rdev = VNOVAL; 1797 optype = VOP_CREATE_DESCOFFSET; 1798 break; 1799 default: 1800 error = EINVAL; 1801 break; 1802 } 1803 } 1804 if (!error) { 1805 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1806 switch (optype) { 1807 case VOP_WHITEOUT_DESCOFFSET: 1808 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1809 if (error) 1810 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1811 vput(nd.ni_dvp); 1812 break; 1813 1814 case VOP_MKNOD_DESCOFFSET: 1815 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1816 &nd.ni_cnd, &vattr); 1817 if (error == 0) 1818 vput(nd.ni_vp); 1819 break; 1820 1821 case VOP_CREATE_DESCOFFSET: 1822 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1823 &nd.ni_cnd, &vattr); 1824 if (error == 0) 1825 vput(nd.ni_vp); 1826 break; 1827 } 1828 } else { 1829 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1830 if (nd.ni_dvp == vp) 1831 vrele(nd.ni_dvp); 1832 else 1833 vput(nd.ni_dvp); 1834 if (vp) 1835 vrele(vp); 1836 } 1837 out: 1838 VERIEXEC_PATH_PUT(path); 1839 return (error); 1840 } 1841 1842 /* 1843 * Create a named pipe. 1844 */ 1845 /* ARGSUSED */ 1846 int 1847 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1848 { 1849 struct sys_mkfifo_args /* { 1850 syscallarg(const char *) path; 1851 syscallarg(int) mode; 1852 } */ *uap = v; 1853 struct proc *p = l->l_proc; 1854 struct vattr vattr; 1855 int error; 1856 struct nameidata nd; 1857 1858 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1859 if ((error = namei(&nd)) != 0) 1860 return (error); 1861 if (nd.ni_vp != NULL) { 1862 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1863 if (nd.ni_dvp == nd.ni_vp) 1864 vrele(nd.ni_dvp); 1865 else 1866 vput(nd.ni_dvp); 1867 vrele(nd.ni_vp); 1868 return (EEXIST); 1869 } 1870 VATTR_NULL(&vattr); 1871 vattr.va_type = VFIFO; 1872 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1873 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1874 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1875 if (error == 0) 1876 vput(nd.ni_vp); 1877 return (error); 1878 } 1879 1880 /* 1881 * Make a hard file link. 1882 */ 1883 /* ARGSUSED */ 1884 int 1885 sys_link(struct lwp *l, void *v, register_t *retval) 1886 { 1887 struct sys_link_args /* { 1888 syscallarg(const char *) path; 1889 syscallarg(const char *) link; 1890 } */ *uap = v; 1891 struct vnode *vp; 1892 struct nameidata nd; 1893 int error; 1894 1895 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1896 if ((error = namei(&nd)) != 0) 1897 return (error); 1898 vp = nd.ni_vp; 1899 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1900 if ((error = namei(&nd)) != 0) 1901 goto out; 1902 if (nd.ni_vp) { 1903 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1904 if (nd.ni_dvp == nd.ni_vp) 1905 vrele(nd.ni_dvp); 1906 else 1907 vput(nd.ni_dvp); 1908 vrele(nd.ni_vp); 1909 error = EEXIST; 1910 goto out; 1911 } 1912 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1913 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1914 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1915 out: 1916 vrele(vp); 1917 return (error); 1918 } 1919 1920 /* 1921 * Make a symbolic link. 1922 */ 1923 /* ARGSUSED */ 1924 int 1925 sys_symlink(struct lwp *l, void *v, register_t *retval) 1926 { 1927 struct sys_symlink_args /* { 1928 syscallarg(const char *) path; 1929 syscallarg(const char *) link; 1930 } */ *uap = v; 1931 struct proc *p = l->l_proc; 1932 struct vattr vattr; 1933 char *path; 1934 int error; 1935 struct nameidata nd; 1936 1937 path = PNBUF_GET(); 1938 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1939 if (error) 1940 goto out; 1941 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1942 if ((error = namei(&nd)) != 0) 1943 goto out; 1944 if (nd.ni_vp) { 1945 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1946 if (nd.ni_dvp == nd.ni_vp) 1947 vrele(nd.ni_dvp); 1948 else 1949 vput(nd.ni_dvp); 1950 vrele(nd.ni_vp); 1951 error = EEXIST; 1952 goto out; 1953 } 1954 VATTR_NULL(&vattr); 1955 vattr.va_type = VLNK; 1956 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1957 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1958 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1959 if (error == 0) 1960 vput(nd.ni_vp); 1961 out: 1962 PNBUF_PUT(path); 1963 return (error); 1964 } 1965 1966 /* 1967 * Delete a whiteout from the filesystem. 1968 */ 1969 /* ARGSUSED */ 1970 int 1971 sys_undelete(struct lwp *l, void *v, register_t *retval) 1972 { 1973 struct sys_undelete_args /* { 1974 syscallarg(const char *) path; 1975 } */ *uap = v; 1976 int error; 1977 struct nameidata nd; 1978 1979 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 1980 SCARG(uap, path), l); 1981 error = namei(&nd); 1982 if (error) 1983 return (error); 1984 1985 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1986 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1987 if (nd.ni_dvp == nd.ni_vp) 1988 vrele(nd.ni_dvp); 1989 else 1990 vput(nd.ni_dvp); 1991 if (nd.ni_vp) 1992 vrele(nd.ni_vp); 1993 return (EEXIST); 1994 } 1995 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1996 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1997 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1998 vput(nd.ni_dvp); 1999 return (error); 2000 } 2001 2002 /* 2003 * Delete a name from the filesystem. 2004 */ 2005 /* ARGSUSED */ 2006 int 2007 sys_unlink(struct lwp *l, void *v, register_t *retval) 2008 { 2009 struct sys_unlink_args /* { 2010 syscallarg(const char *) path; 2011 } */ *uap = v; 2012 struct vnode *vp; 2013 int error; 2014 struct nameidata nd; 2015 char *path; 2016 const char *cpath; 2017 enum uio_seg seg = UIO_USERSPACE; 2018 2019 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2020 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2021 2022 if ((error = namei(&nd)) != 0) 2023 goto out; 2024 vp = nd.ni_vp; 2025 2026 /* 2027 * The root of a mounted filesystem cannot be deleted. 2028 */ 2029 if (vp->v_flag & VROOT) { 2030 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2031 if (nd.ni_dvp == vp) 2032 vrele(nd.ni_dvp); 2033 else 2034 vput(nd.ni_dvp); 2035 vput(vp); 2036 error = EBUSY; 2037 goto out; 2038 } 2039 2040 #if NVERIEXEC > 0 2041 /* Handle remove requests for veriexec entries. */ 2042 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2043 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2044 if (nd.ni_dvp == vp) 2045 vrele(nd.ni_dvp); 2046 else 2047 vput(nd.ni_dvp); 2048 vput(vp); 2049 goto out; 2050 } 2051 #endif /* NVERIEXEC > 0 */ 2052 2053 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2054 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2055 #ifdef FILEASSOC 2056 (void)fileassoc_file_delete(vp); 2057 #endif /* FILEASSOC */ 2058 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2059 out: 2060 VERIEXEC_PATH_PUT(path); 2061 return (error); 2062 } 2063 2064 /* 2065 * Reposition read/write file offset. 2066 */ 2067 int 2068 sys_lseek(struct lwp *l, void *v, register_t *retval) 2069 { 2070 struct sys_lseek_args /* { 2071 syscallarg(int) fd; 2072 syscallarg(int) pad; 2073 syscallarg(off_t) offset; 2074 syscallarg(int) whence; 2075 } */ *uap = v; 2076 struct proc *p = l->l_proc; 2077 kauth_cred_t cred = l->l_cred; 2078 struct filedesc *fdp = p->p_fd; 2079 struct file *fp; 2080 struct vnode *vp; 2081 struct vattr vattr; 2082 off_t newoff; 2083 int error; 2084 2085 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2086 return (EBADF); 2087 2088 FILE_USE(fp); 2089 2090 vp = (struct vnode *)fp->f_data; 2091 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2092 error = ESPIPE; 2093 goto out; 2094 } 2095 2096 switch (SCARG(uap, whence)) { 2097 case SEEK_CUR: 2098 newoff = fp->f_offset + SCARG(uap, offset); 2099 break; 2100 case SEEK_END: 2101 error = VOP_GETATTR(vp, &vattr, cred, l); 2102 if (error) 2103 goto out; 2104 newoff = SCARG(uap, offset) + vattr.va_size; 2105 break; 2106 case SEEK_SET: 2107 newoff = SCARG(uap, offset); 2108 break; 2109 default: 2110 error = EINVAL; 2111 goto out; 2112 } 2113 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2114 goto out; 2115 2116 *(off_t *)retval = fp->f_offset = newoff; 2117 out: 2118 FILE_UNUSE(fp, l); 2119 return (error); 2120 } 2121 2122 /* 2123 * Positional read system call. 2124 */ 2125 int 2126 sys_pread(struct lwp *l, void *v, register_t *retval) 2127 { 2128 struct sys_pread_args /* { 2129 syscallarg(int) fd; 2130 syscallarg(void *) buf; 2131 syscallarg(size_t) nbyte; 2132 syscallarg(off_t) offset; 2133 } */ *uap = v; 2134 struct proc *p = l->l_proc; 2135 struct filedesc *fdp = p->p_fd; 2136 struct file *fp; 2137 struct vnode *vp; 2138 off_t offset; 2139 int error, fd = SCARG(uap, fd); 2140 2141 if ((fp = fd_getfile(fdp, fd)) == NULL) 2142 return (EBADF); 2143 2144 if ((fp->f_flag & FREAD) == 0) { 2145 simple_unlock(&fp->f_slock); 2146 return (EBADF); 2147 } 2148 2149 FILE_USE(fp); 2150 2151 vp = (struct vnode *)fp->f_data; 2152 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2153 error = ESPIPE; 2154 goto out; 2155 } 2156 2157 offset = SCARG(uap, offset); 2158 2159 /* 2160 * XXX This works because no file systems actually 2161 * XXX take any action on the seek operation. 2162 */ 2163 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2164 goto out; 2165 2166 /* dofileread() will unuse the descriptor for us */ 2167 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2168 &offset, 0, retval)); 2169 2170 out: 2171 FILE_UNUSE(fp, l); 2172 return (error); 2173 } 2174 2175 /* 2176 * Positional scatter read system call. 2177 */ 2178 int 2179 sys_preadv(struct lwp *l, void *v, register_t *retval) 2180 { 2181 struct sys_preadv_args /* { 2182 syscallarg(int) fd; 2183 syscallarg(const struct iovec *) iovp; 2184 syscallarg(int) iovcnt; 2185 syscallarg(off_t) offset; 2186 } */ *uap = v; 2187 struct proc *p = l->l_proc; 2188 struct filedesc *fdp = p->p_fd; 2189 struct file *fp; 2190 struct vnode *vp; 2191 off_t offset; 2192 int error, fd = SCARG(uap, fd); 2193 2194 if ((fp = fd_getfile(fdp, fd)) == NULL) 2195 return (EBADF); 2196 2197 if ((fp->f_flag & FREAD) == 0) { 2198 simple_unlock(&fp->f_slock); 2199 return (EBADF); 2200 } 2201 2202 FILE_USE(fp); 2203 2204 vp = (struct vnode *)fp->f_data; 2205 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2206 error = ESPIPE; 2207 goto out; 2208 } 2209 2210 offset = SCARG(uap, offset); 2211 2212 /* 2213 * XXX This works because no file systems actually 2214 * XXX take any action on the seek operation. 2215 */ 2216 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2217 goto out; 2218 2219 /* dofilereadv() will unuse the descriptor for us */ 2220 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2221 &offset, 0, retval)); 2222 2223 out: 2224 FILE_UNUSE(fp, l); 2225 return (error); 2226 } 2227 2228 /* 2229 * Positional write system call. 2230 */ 2231 int 2232 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2233 { 2234 struct sys_pwrite_args /* { 2235 syscallarg(int) fd; 2236 syscallarg(const void *) buf; 2237 syscallarg(size_t) nbyte; 2238 syscallarg(off_t) offset; 2239 } */ *uap = v; 2240 struct proc *p = l->l_proc; 2241 struct filedesc *fdp = p->p_fd; 2242 struct file *fp; 2243 struct vnode *vp; 2244 off_t offset; 2245 int error, fd = SCARG(uap, fd); 2246 2247 if ((fp = fd_getfile(fdp, fd)) == NULL) 2248 return (EBADF); 2249 2250 if ((fp->f_flag & FWRITE) == 0) { 2251 simple_unlock(&fp->f_slock); 2252 return (EBADF); 2253 } 2254 2255 FILE_USE(fp); 2256 2257 vp = (struct vnode *)fp->f_data; 2258 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2259 error = ESPIPE; 2260 goto out; 2261 } 2262 2263 offset = SCARG(uap, offset); 2264 2265 /* 2266 * XXX This works because no file systems actually 2267 * XXX take any action on the seek operation. 2268 */ 2269 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2270 goto out; 2271 2272 /* dofilewrite() will unuse the descriptor for us */ 2273 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2274 &offset, 0, retval)); 2275 2276 out: 2277 FILE_UNUSE(fp, l); 2278 return (error); 2279 } 2280 2281 /* 2282 * Positional gather write system call. 2283 */ 2284 int 2285 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2286 { 2287 struct sys_pwritev_args /* { 2288 syscallarg(int) fd; 2289 syscallarg(const struct iovec *) iovp; 2290 syscallarg(int) iovcnt; 2291 syscallarg(off_t) offset; 2292 } */ *uap = v; 2293 struct proc *p = l->l_proc; 2294 struct filedesc *fdp = p->p_fd; 2295 struct file *fp; 2296 struct vnode *vp; 2297 off_t offset; 2298 int error, fd = SCARG(uap, fd); 2299 2300 if ((fp = fd_getfile(fdp, fd)) == NULL) 2301 return (EBADF); 2302 2303 if ((fp->f_flag & FWRITE) == 0) { 2304 simple_unlock(&fp->f_slock); 2305 return (EBADF); 2306 } 2307 2308 FILE_USE(fp); 2309 2310 vp = (struct vnode *)fp->f_data; 2311 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2312 error = ESPIPE; 2313 goto out; 2314 } 2315 2316 offset = SCARG(uap, offset); 2317 2318 /* 2319 * XXX This works because no file systems actually 2320 * XXX take any action on the seek operation. 2321 */ 2322 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2323 goto out; 2324 2325 /* dofilewritev() will unuse the descriptor for us */ 2326 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2327 &offset, 0, retval)); 2328 2329 out: 2330 FILE_UNUSE(fp, l); 2331 return (error); 2332 } 2333 2334 /* 2335 * Check access permissions. 2336 */ 2337 int 2338 sys_access(struct lwp *l, void *v, register_t *retval) 2339 { 2340 struct sys_access_args /* { 2341 syscallarg(const char *) path; 2342 syscallarg(int) flags; 2343 } */ *uap = v; 2344 kauth_cred_t cred; 2345 struct vnode *vp; 2346 int error, flags; 2347 struct nameidata nd; 2348 2349 cred = kauth_cred_dup(l->l_cred); 2350 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2351 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2352 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2353 SCARG(uap, path), l); 2354 /* Override default credentials */ 2355 nd.ni_cnd.cn_cred = cred; 2356 if ((error = namei(&nd)) != 0) 2357 goto out; 2358 vp = nd.ni_vp; 2359 2360 /* Flags == 0 means only check for existence. */ 2361 if (SCARG(uap, flags)) { 2362 flags = 0; 2363 if (SCARG(uap, flags) & R_OK) 2364 flags |= VREAD; 2365 if (SCARG(uap, flags) & W_OK) 2366 flags |= VWRITE; 2367 if (SCARG(uap, flags) & X_OK) 2368 flags |= VEXEC; 2369 2370 error = VOP_ACCESS(vp, flags, cred, l); 2371 if (!error && (flags & VWRITE)) 2372 error = vn_writechk(vp); 2373 } 2374 vput(vp); 2375 out: 2376 kauth_cred_free(cred); 2377 return (error); 2378 } 2379 2380 /* 2381 * Common code for all sys_stat functions, including compat versions. 2382 */ 2383 int 2384 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2385 struct stat *sb) 2386 { 2387 int error; 2388 struct nameidata nd; 2389 2390 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2391 error = namei(&nd); 2392 if (error != 0) 2393 return error; 2394 error = vn_stat(nd.ni_vp, sb, l); 2395 vput(nd.ni_vp); 2396 return error; 2397 } 2398 2399 /* 2400 * Get file status; this version follows links. 2401 */ 2402 /* ARGSUSED */ 2403 int 2404 sys___stat30(struct lwp *l, void *v, register_t *retval) 2405 { 2406 struct sys___stat30_args /* { 2407 syscallarg(const char *) path; 2408 syscallarg(struct stat *) ub; 2409 } */ *uap = v; 2410 struct stat sb; 2411 int error; 2412 2413 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2414 if (error) 2415 return error; 2416 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2417 } 2418 2419 /* 2420 * Get file status; this version does not follow links. 2421 */ 2422 /* ARGSUSED */ 2423 int 2424 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2425 { 2426 struct sys___lstat30_args /* { 2427 syscallarg(const char *) path; 2428 syscallarg(struct stat *) ub; 2429 } */ *uap = v; 2430 struct stat sb; 2431 int error; 2432 2433 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2434 if (error) 2435 return error; 2436 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2437 } 2438 2439 /* 2440 * Get configurable pathname variables. 2441 */ 2442 /* ARGSUSED */ 2443 int 2444 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2445 { 2446 struct sys_pathconf_args /* { 2447 syscallarg(const char *) path; 2448 syscallarg(int) name; 2449 } */ *uap = v; 2450 int error; 2451 struct nameidata nd; 2452 2453 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2454 SCARG(uap, path), l); 2455 if ((error = namei(&nd)) != 0) 2456 return (error); 2457 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2458 vput(nd.ni_vp); 2459 return (error); 2460 } 2461 2462 /* 2463 * Return target name of a symbolic link. 2464 */ 2465 /* ARGSUSED */ 2466 int 2467 sys_readlink(struct lwp *l, void *v, register_t *retval) 2468 { 2469 struct sys_readlink_args /* { 2470 syscallarg(const char *) path; 2471 syscallarg(char *) buf; 2472 syscallarg(size_t) count; 2473 } */ *uap = v; 2474 struct vnode *vp; 2475 struct iovec aiov; 2476 struct uio auio; 2477 int error; 2478 struct nameidata nd; 2479 2480 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2481 SCARG(uap, path), l); 2482 if ((error = namei(&nd)) != 0) 2483 return (error); 2484 vp = nd.ni_vp; 2485 if (vp->v_type != VLNK) 2486 error = EINVAL; 2487 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2488 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2489 aiov.iov_base = SCARG(uap, buf); 2490 aiov.iov_len = SCARG(uap, count); 2491 auio.uio_iov = &aiov; 2492 auio.uio_iovcnt = 1; 2493 auio.uio_offset = 0; 2494 auio.uio_rw = UIO_READ; 2495 KASSERT(l == curlwp); 2496 auio.uio_vmspace = l->l_proc->p_vmspace; 2497 auio.uio_resid = SCARG(uap, count); 2498 error = VOP_READLINK(vp, &auio, l->l_cred); 2499 } 2500 vput(vp); 2501 *retval = SCARG(uap, count) - auio.uio_resid; 2502 return (error); 2503 } 2504 2505 /* 2506 * Change flags of a file given a path name. 2507 */ 2508 /* ARGSUSED */ 2509 int 2510 sys_chflags(struct lwp *l, void *v, register_t *retval) 2511 { 2512 struct sys_chflags_args /* { 2513 syscallarg(const char *) path; 2514 syscallarg(u_long) flags; 2515 } */ *uap = v; 2516 struct vnode *vp; 2517 int error; 2518 struct nameidata nd; 2519 2520 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2521 if ((error = namei(&nd)) != 0) 2522 return (error); 2523 vp = nd.ni_vp; 2524 error = change_flags(vp, SCARG(uap, flags), l); 2525 vput(vp); 2526 return (error); 2527 } 2528 2529 /* 2530 * Change flags of a file given a file descriptor. 2531 */ 2532 /* ARGSUSED */ 2533 int 2534 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2535 { 2536 struct sys_fchflags_args /* { 2537 syscallarg(int) fd; 2538 syscallarg(u_long) flags; 2539 } */ *uap = v; 2540 struct proc *p = l->l_proc; 2541 struct vnode *vp; 2542 struct file *fp; 2543 int error; 2544 2545 /* getvnode() will use the descriptor for us */ 2546 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2547 return (error); 2548 vp = (struct vnode *)fp->f_data; 2549 error = change_flags(vp, SCARG(uap, flags), l); 2550 VOP_UNLOCK(vp, 0); 2551 FILE_UNUSE(fp, l); 2552 return (error); 2553 } 2554 2555 /* 2556 * Change flags of a file given a path name; this version does 2557 * not follow links. 2558 */ 2559 int 2560 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2561 { 2562 struct sys_lchflags_args /* { 2563 syscallarg(const char *) path; 2564 syscallarg(u_long) flags; 2565 } */ *uap = v; 2566 struct vnode *vp; 2567 int error; 2568 struct nameidata nd; 2569 2570 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2571 if ((error = namei(&nd)) != 0) 2572 return (error); 2573 vp = nd.ni_vp; 2574 error = change_flags(vp, SCARG(uap, flags), l); 2575 vput(vp); 2576 return (error); 2577 } 2578 2579 /* 2580 * Common routine to change flags of a file. 2581 */ 2582 int 2583 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2584 { 2585 struct vattr vattr; 2586 int error; 2587 2588 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2589 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2590 /* 2591 * Non-superusers cannot change the flags on devices, even if they 2592 * own them. 2593 */ 2594 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2595 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2596 goto out; 2597 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2598 error = EINVAL; 2599 goto out; 2600 } 2601 } 2602 VATTR_NULL(&vattr); 2603 vattr.va_flags = flags; 2604 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2605 out: 2606 return (error); 2607 } 2608 2609 /* 2610 * Change mode of a file given path name; this version follows links. 2611 */ 2612 /* ARGSUSED */ 2613 int 2614 sys_chmod(struct lwp *l, void *v, register_t *retval) 2615 { 2616 struct sys_chmod_args /* { 2617 syscallarg(const char *) path; 2618 syscallarg(int) mode; 2619 } */ *uap = v; 2620 int error; 2621 struct nameidata nd; 2622 2623 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2624 if ((error = namei(&nd)) != 0) 2625 return (error); 2626 2627 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2628 2629 vrele(nd.ni_vp); 2630 return (error); 2631 } 2632 2633 /* 2634 * Change mode of a file given a file descriptor. 2635 */ 2636 /* ARGSUSED */ 2637 int 2638 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2639 { 2640 struct sys_fchmod_args /* { 2641 syscallarg(int) fd; 2642 syscallarg(int) mode; 2643 } */ *uap = v; 2644 struct proc *p = l->l_proc; 2645 struct file *fp; 2646 int error; 2647 2648 /* getvnode() will use the descriptor for us */ 2649 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2650 return (error); 2651 2652 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2653 FILE_UNUSE(fp, l); 2654 return (error); 2655 } 2656 2657 /* 2658 * Change mode of a file given path name; this version does not follow links. 2659 */ 2660 /* ARGSUSED */ 2661 int 2662 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2663 { 2664 struct sys_lchmod_args /* { 2665 syscallarg(const char *) path; 2666 syscallarg(int) mode; 2667 } */ *uap = v; 2668 int error; 2669 struct nameidata nd; 2670 2671 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2672 if ((error = namei(&nd)) != 0) 2673 return (error); 2674 2675 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2676 2677 vrele(nd.ni_vp); 2678 return (error); 2679 } 2680 2681 /* 2682 * Common routine to set mode given a vnode. 2683 */ 2684 static int 2685 change_mode(struct vnode *vp, int mode, struct lwp *l) 2686 { 2687 struct vattr vattr; 2688 int error; 2689 2690 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2691 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2692 VATTR_NULL(&vattr); 2693 vattr.va_mode = mode & ALLPERMS; 2694 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2695 VOP_UNLOCK(vp, 0); 2696 return (error); 2697 } 2698 2699 /* 2700 * Set ownership given a path name; this version follows links. 2701 */ 2702 /* ARGSUSED */ 2703 int 2704 sys_chown(struct lwp *l, void *v, register_t *retval) 2705 { 2706 struct sys_chown_args /* { 2707 syscallarg(const char *) path; 2708 syscallarg(uid_t) uid; 2709 syscallarg(gid_t) gid; 2710 } */ *uap = v; 2711 int error; 2712 struct nameidata nd; 2713 2714 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2715 if ((error = namei(&nd)) != 0) 2716 return (error); 2717 2718 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2719 2720 vrele(nd.ni_vp); 2721 return (error); 2722 } 2723 2724 /* 2725 * Set ownership given a path name; this version follows links. 2726 * Provides POSIX semantics. 2727 */ 2728 /* ARGSUSED */ 2729 int 2730 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2731 { 2732 struct sys_chown_args /* { 2733 syscallarg(const char *) path; 2734 syscallarg(uid_t) uid; 2735 syscallarg(gid_t) gid; 2736 } */ *uap = v; 2737 int error; 2738 struct nameidata nd; 2739 2740 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2741 if ((error = namei(&nd)) != 0) 2742 return (error); 2743 2744 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2745 2746 vrele(nd.ni_vp); 2747 return (error); 2748 } 2749 2750 /* 2751 * Set ownership given a file descriptor. 2752 */ 2753 /* ARGSUSED */ 2754 int 2755 sys_fchown(struct lwp *l, void *v, register_t *retval) 2756 { 2757 struct sys_fchown_args /* { 2758 syscallarg(int) fd; 2759 syscallarg(uid_t) uid; 2760 syscallarg(gid_t) gid; 2761 } */ *uap = v; 2762 struct proc *p = l->l_proc; 2763 int error; 2764 struct file *fp; 2765 2766 /* getvnode() will use the descriptor for us */ 2767 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2768 return (error); 2769 2770 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2771 SCARG(uap, gid), l, 0); 2772 FILE_UNUSE(fp, l); 2773 return (error); 2774 } 2775 2776 /* 2777 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2778 */ 2779 /* ARGSUSED */ 2780 int 2781 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2782 { 2783 struct sys_fchown_args /* { 2784 syscallarg(int) fd; 2785 syscallarg(uid_t) uid; 2786 syscallarg(gid_t) gid; 2787 } */ *uap = v; 2788 struct proc *p = l->l_proc; 2789 int error; 2790 struct file *fp; 2791 2792 /* getvnode() will use the descriptor for us */ 2793 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2794 return (error); 2795 2796 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2797 SCARG(uap, gid), l, 1); 2798 FILE_UNUSE(fp, l); 2799 return (error); 2800 } 2801 2802 /* 2803 * Set ownership given a path name; this version does not follow links. 2804 */ 2805 /* ARGSUSED */ 2806 int 2807 sys_lchown(struct lwp *l, void *v, register_t *retval) 2808 { 2809 struct sys_lchown_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, NOFOLLOW | TRYEMULROOT, 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 does not follow links. 2829 * Provides POSIX/XPG semantics. 2830 */ 2831 /* ARGSUSED */ 2832 int 2833 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2834 { 2835 struct sys_lchown_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, NOFOLLOW | TRYEMULROOT, 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 * Common routine to set ownership given a vnode. 2855 */ 2856 static int 2857 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2858 int posix_semantics) 2859 { 2860 struct vattr vattr; 2861 mode_t newmode; 2862 int error; 2863 2864 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2865 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2866 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2867 goto out; 2868 2869 #define CHANGED(x) ((int)(x) != -1) 2870 newmode = vattr.va_mode; 2871 if (posix_semantics) { 2872 /* 2873 * POSIX/XPG semantics: if the caller is not the super-user, 2874 * clear set-user-id and set-group-id bits. Both POSIX and 2875 * the XPG consider the behaviour for calls by the super-user 2876 * implementation-defined; we leave the set-user-id and set- 2877 * group-id settings intact in that case. 2878 */ 2879 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2880 NULL) != 0) 2881 newmode &= ~(S_ISUID | S_ISGID); 2882 } else { 2883 /* 2884 * NetBSD semantics: when changing owner and/or group, 2885 * clear the respective bit(s). 2886 */ 2887 if (CHANGED(uid)) 2888 newmode &= ~S_ISUID; 2889 if (CHANGED(gid)) 2890 newmode &= ~S_ISGID; 2891 } 2892 /* Update va_mode iff altered. */ 2893 if (vattr.va_mode == newmode) 2894 newmode = VNOVAL; 2895 2896 VATTR_NULL(&vattr); 2897 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2898 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2899 vattr.va_mode = newmode; 2900 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2901 #undef CHANGED 2902 2903 out: 2904 VOP_UNLOCK(vp, 0); 2905 return (error); 2906 } 2907 2908 /* 2909 * Set the access and modification times given a path name; this 2910 * version follows links. 2911 */ 2912 /* ARGSUSED */ 2913 int 2914 sys_utimes(struct lwp *l, void *v, register_t *retval) 2915 { 2916 struct sys_utimes_args /* { 2917 syscallarg(const char *) path; 2918 syscallarg(const struct timeval *) tptr; 2919 } */ *uap = v; 2920 2921 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2922 SCARG(uap, tptr), UIO_USERSPACE); 2923 } 2924 2925 /* 2926 * Set the access and modification times given a file descriptor. 2927 */ 2928 /* ARGSUSED */ 2929 int 2930 sys_futimes(struct lwp *l, void *v, register_t *retval) 2931 { 2932 struct sys_futimes_args /* { 2933 syscallarg(int) fd; 2934 syscallarg(const struct timeval *) tptr; 2935 } */ *uap = v; 2936 int error; 2937 struct file *fp; 2938 2939 /* getvnode() will use the descriptor for us */ 2940 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2941 return (error); 2942 2943 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2944 SCARG(uap, tptr), UIO_USERSPACE); 2945 2946 FILE_UNUSE(fp, l); 2947 return (error); 2948 } 2949 2950 /* 2951 * Set the access and modification times given a path name; this 2952 * version does not follow links. 2953 */ 2954 int 2955 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2956 { 2957 struct sys_lutimes_args /* { 2958 syscallarg(const char *) path; 2959 syscallarg(const struct timeval *) tptr; 2960 } */ *uap = v; 2961 2962 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2963 SCARG(uap, tptr), UIO_USERSPACE); 2964 } 2965 2966 /* 2967 * Common routine to set access and modification times given a vnode. 2968 */ 2969 int 2970 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2971 const struct timeval *tptr, enum uio_seg seg) 2972 { 2973 struct vattr vattr; 2974 struct nameidata nd; 2975 int error; 2976 2977 VATTR_NULL(&vattr); 2978 if (tptr == NULL) { 2979 nanotime(&vattr.va_atime); 2980 vattr.va_mtime = vattr.va_atime; 2981 vattr.va_vaflags |= VA_UTIMES_NULL; 2982 } else { 2983 struct timeval tv[2]; 2984 2985 if (seg != UIO_SYSSPACE) { 2986 error = copyin(tptr, &tv, sizeof (tv)); 2987 if (error != 0) 2988 return error; 2989 tptr = tv; 2990 } 2991 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 2992 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 2993 } 2994 2995 if (vp == NULL) { 2996 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 2997 if ((error = namei(&nd)) != 0) 2998 return (error); 2999 vp = nd.ni_vp; 3000 } else 3001 nd.ni_vp = NULL; 3002 3003 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3004 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3005 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3006 VOP_UNLOCK(vp, 0); 3007 3008 if (nd.ni_vp != NULL) 3009 vrele(nd.ni_vp); 3010 3011 return (error); 3012 } 3013 3014 /* 3015 * Truncate a file given its path name. 3016 */ 3017 /* ARGSUSED */ 3018 int 3019 sys_truncate(struct lwp *l, void *v, register_t *retval) 3020 { 3021 struct sys_truncate_args /* { 3022 syscallarg(const char *) path; 3023 syscallarg(int) pad; 3024 syscallarg(off_t) length; 3025 } */ *uap = v; 3026 struct vnode *vp; 3027 struct vattr vattr; 3028 int error; 3029 struct nameidata nd; 3030 3031 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3032 if ((error = namei(&nd)) != 0) 3033 return (error); 3034 vp = nd.ni_vp; 3035 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3036 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3037 if (vp->v_type == VDIR) 3038 error = EISDIR; 3039 else if ((error = vn_writechk(vp)) == 0 && 3040 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3041 VATTR_NULL(&vattr); 3042 vattr.va_size = SCARG(uap, length); 3043 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3044 } 3045 vput(vp); 3046 return (error); 3047 } 3048 3049 /* 3050 * Truncate a file given a file descriptor. 3051 */ 3052 /* ARGSUSED */ 3053 int 3054 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3055 { 3056 struct sys_ftruncate_args /* { 3057 syscallarg(int) fd; 3058 syscallarg(int) pad; 3059 syscallarg(off_t) length; 3060 } */ *uap = v; 3061 struct proc *p = l->l_proc; 3062 struct vattr vattr; 3063 struct vnode *vp; 3064 struct file *fp; 3065 int error; 3066 3067 /* getvnode() will use the descriptor for us */ 3068 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3069 return (error); 3070 if ((fp->f_flag & FWRITE) == 0) { 3071 error = EINVAL; 3072 goto out; 3073 } 3074 vp = (struct vnode *)fp->f_data; 3075 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3076 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3077 if (vp->v_type == VDIR) 3078 error = EISDIR; 3079 else if ((error = vn_writechk(vp)) == 0) { 3080 VATTR_NULL(&vattr); 3081 vattr.va_size = SCARG(uap, length); 3082 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3083 } 3084 VOP_UNLOCK(vp, 0); 3085 out: 3086 FILE_UNUSE(fp, l); 3087 return (error); 3088 } 3089 3090 /* 3091 * Sync an open file. 3092 */ 3093 /* ARGSUSED */ 3094 int 3095 sys_fsync(struct lwp *l, void *v, register_t *retval) 3096 { 3097 struct sys_fsync_args /* { 3098 syscallarg(int) fd; 3099 } */ *uap = v; 3100 struct proc *p = l->l_proc; 3101 struct vnode *vp; 3102 struct file *fp; 3103 int error; 3104 3105 /* getvnode() will use the descriptor for us */ 3106 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3107 return (error); 3108 vp = (struct vnode *)fp->f_data; 3109 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3110 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3111 if (error == 0 && bioops.io_fsync != NULL && 3112 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3113 (*bioops.io_fsync)(vp, 0); 3114 VOP_UNLOCK(vp, 0); 3115 FILE_UNUSE(fp, l); 3116 return (error); 3117 } 3118 3119 /* 3120 * Sync a range of file data. API modeled after that found in AIX. 3121 * 3122 * FDATASYNC indicates that we need only save enough metadata to be able 3123 * to re-read the written data. Note we duplicate AIX's requirement that 3124 * the file be open for writing. 3125 */ 3126 /* ARGSUSED */ 3127 int 3128 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3129 { 3130 struct sys_fsync_range_args /* { 3131 syscallarg(int) fd; 3132 syscallarg(int) flags; 3133 syscallarg(off_t) start; 3134 syscallarg(off_t) length; 3135 } */ *uap = v; 3136 struct proc *p = l->l_proc; 3137 struct vnode *vp; 3138 struct file *fp; 3139 int flags, nflags; 3140 off_t s, e, len; 3141 int error; 3142 3143 /* getvnode() will use the descriptor for us */ 3144 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3145 return (error); 3146 3147 if ((fp->f_flag & FWRITE) == 0) { 3148 error = EBADF; 3149 goto out; 3150 } 3151 3152 flags = SCARG(uap, flags); 3153 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3154 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3155 error = EINVAL; 3156 goto out; 3157 } 3158 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3159 if (flags & FDATASYNC) 3160 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3161 else 3162 nflags = FSYNC_WAIT; 3163 if (flags & FDISKSYNC) 3164 nflags |= FSYNC_CACHE; 3165 3166 len = SCARG(uap, length); 3167 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3168 if (len) { 3169 s = SCARG(uap, start); 3170 e = s + len; 3171 if (e < s) { 3172 error = EINVAL; 3173 goto out; 3174 } 3175 } else { 3176 e = 0; 3177 s = 0; 3178 } 3179 3180 vp = (struct vnode *)fp->f_data; 3181 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3182 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3183 3184 if (error == 0 && bioops.io_fsync != NULL && 3185 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3186 (*bioops.io_fsync)(vp, nflags); 3187 3188 VOP_UNLOCK(vp, 0); 3189 out: 3190 FILE_UNUSE(fp, l); 3191 return (error); 3192 } 3193 3194 /* 3195 * Sync the data of an open file. 3196 */ 3197 /* ARGSUSED */ 3198 int 3199 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3200 { 3201 struct sys_fdatasync_args /* { 3202 syscallarg(int) fd; 3203 } */ *uap = v; 3204 struct proc *p = l->l_proc; 3205 struct vnode *vp; 3206 struct file *fp; 3207 int error; 3208 3209 /* getvnode() will use the descriptor for us */ 3210 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3211 return (error); 3212 if ((fp->f_flag & FWRITE) == 0) { 3213 FILE_UNUSE(fp, l); 3214 return (EBADF); 3215 } 3216 vp = (struct vnode *)fp->f_data; 3217 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3218 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3219 VOP_UNLOCK(vp, 0); 3220 FILE_UNUSE(fp, l); 3221 return (error); 3222 } 3223 3224 /* 3225 * Rename files, (standard) BSD semantics frontend. 3226 */ 3227 /* ARGSUSED */ 3228 int 3229 sys_rename(struct lwp *l, void *v, register_t *retval) 3230 { 3231 struct sys_rename_args /* { 3232 syscallarg(const char *) from; 3233 syscallarg(const char *) to; 3234 } */ *uap = v; 3235 3236 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3237 } 3238 3239 /* 3240 * Rename files, POSIX semantics frontend. 3241 */ 3242 /* ARGSUSED */ 3243 int 3244 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3245 { 3246 struct sys___posix_rename_args /* { 3247 syscallarg(const char *) from; 3248 syscallarg(const char *) to; 3249 } */ *uap = v; 3250 3251 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3252 } 3253 3254 /* 3255 * Rename files. Source and destination must either both be directories, 3256 * or both not be directories. If target is a directory, it must be empty. 3257 * If `from' and `to' refer to the same object, the value of the `retain' 3258 * argument is used to determine whether `from' will be 3259 * 3260 * (retain == 0) deleted unless `from' and `to' refer to the same 3261 * object in the file system's name space (BSD). 3262 * (retain == 1) always retained (POSIX). 3263 */ 3264 static int 3265 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3266 { 3267 struct vnode *tvp, *fvp, *tdvp; 3268 struct nameidata fromnd, tond; 3269 struct proc *p; 3270 int error; 3271 3272 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3273 from, l); 3274 if ((error = namei(&fromnd)) != 0) 3275 return (error); 3276 if (fromnd.ni_dvp != fromnd.ni_vp) 3277 VOP_UNLOCK(fromnd.ni_dvp, 0); 3278 fvp = fromnd.ni_vp; 3279 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3280 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3281 if ((error = namei(&tond)) != 0) { 3282 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3283 vrele(fromnd.ni_dvp); 3284 vrele(fvp); 3285 goto out1; 3286 } 3287 tdvp = tond.ni_dvp; 3288 tvp = tond.ni_vp; 3289 3290 if (tvp != NULL) { 3291 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3292 error = ENOTDIR; 3293 goto out; 3294 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3295 error = EISDIR; 3296 goto out; 3297 } 3298 } 3299 3300 if (fvp == tdvp) 3301 error = EINVAL; 3302 3303 /* 3304 * Source and destination refer to the same object. 3305 */ 3306 if (fvp == tvp) { 3307 if (retain) 3308 error = -1; 3309 else if (fromnd.ni_dvp == tdvp && 3310 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3311 !memcmp(fromnd.ni_cnd.cn_nameptr, 3312 tond.ni_cnd.cn_nameptr, 3313 fromnd.ni_cnd.cn_namelen)) 3314 error = -1; 3315 } 3316 3317 #if NVERIEXEC > 0 3318 if (!error) { 3319 char *f1, *f2; 3320 3321 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3322 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3323 3324 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3325 strlcpy(f1, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3326 3327 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3328 3329 free(f1, M_TEMP); 3330 free(f2, M_TEMP); 3331 } 3332 #endif /* NVERIEXEC > 0 */ 3333 3334 out: 3335 p = l->l_proc; 3336 if (!error) { 3337 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3338 if (fromnd.ni_dvp != tdvp) 3339 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3340 if (tvp) { 3341 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3342 } 3343 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3344 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3345 } else { 3346 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3347 if (tdvp == tvp) 3348 vrele(tdvp); 3349 else 3350 vput(tdvp); 3351 if (tvp) 3352 vput(tvp); 3353 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3354 vrele(fromnd.ni_dvp); 3355 vrele(fvp); 3356 } 3357 vrele(tond.ni_startdir); 3358 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3359 out1: 3360 if (fromnd.ni_startdir) 3361 vrele(fromnd.ni_startdir); 3362 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3363 return (error == -1 ? 0 : error); 3364 } 3365 3366 /* 3367 * Make a directory file. 3368 */ 3369 /* ARGSUSED */ 3370 int 3371 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3372 { 3373 struct sys_mkdir_args /* { 3374 syscallarg(const char *) path; 3375 syscallarg(int) mode; 3376 } */ *uap = v; 3377 struct proc *p = l->l_proc; 3378 struct vnode *vp; 3379 struct vattr vattr; 3380 int error; 3381 struct nameidata nd; 3382 3383 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3384 SCARG(uap, path), l); 3385 if ((error = namei(&nd)) != 0) 3386 return (error); 3387 vp = nd.ni_vp; 3388 if (vp != NULL) { 3389 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3390 if (nd.ni_dvp == vp) 3391 vrele(nd.ni_dvp); 3392 else 3393 vput(nd.ni_dvp); 3394 vrele(vp); 3395 return (EEXIST); 3396 } 3397 VATTR_NULL(&vattr); 3398 vattr.va_type = VDIR; 3399 vattr.va_mode = 3400 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3401 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3402 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3403 if (!error) 3404 vput(nd.ni_vp); 3405 return (error); 3406 } 3407 3408 /* 3409 * Remove a directory file. 3410 */ 3411 /* ARGSUSED */ 3412 int 3413 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3414 { 3415 struct sys_rmdir_args /* { 3416 syscallarg(const char *) path; 3417 } */ *uap = v; 3418 struct vnode *vp; 3419 int error; 3420 struct nameidata nd; 3421 3422 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3423 SCARG(uap, path), l); 3424 if ((error = namei(&nd)) != 0) 3425 return (error); 3426 vp = nd.ni_vp; 3427 if (vp->v_type != VDIR) { 3428 error = ENOTDIR; 3429 goto out; 3430 } 3431 /* 3432 * No rmdir "." please. 3433 */ 3434 if (nd.ni_dvp == vp) { 3435 error = EINVAL; 3436 goto out; 3437 } 3438 /* 3439 * The root of a mounted filesystem cannot be deleted. 3440 */ 3441 if (vp->v_flag & VROOT) { 3442 error = EBUSY; 3443 goto out; 3444 } 3445 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3446 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3447 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3448 return (error); 3449 3450 out: 3451 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3452 if (nd.ni_dvp == vp) 3453 vrele(nd.ni_dvp); 3454 else 3455 vput(nd.ni_dvp); 3456 vput(vp); 3457 return (error); 3458 } 3459 3460 /* 3461 * Read a block of directory entries in a file system independent format. 3462 */ 3463 int 3464 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3465 { 3466 struct sys___getdents30_args /* { 3467 syscallarg(int) fd; 3468 syscallarg(char *) buf; 3469 syscallarg(size_t) count; 3470 } */ *uap = v; 3471 struct proc *p = l->l_proc; 3472 struct file *fp; 3473 int error, done; 3474 3475 /* getvnode() will use the descriptor for us */ 3476 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3477 return (error); 3478 if ((fp->f_flag & FREAD) == 0) { 3479 error = EBADF; 3480 goto out; 3481 } 3482 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3483 SCARG(uap, count), &done, l, 0, 0); 3484 #ifdef KTRACE 3485 if (!error && KTRPOINT(p, KTR_GENIO)) { 3486 struct iovec iov; 3487 iov.iov_base = SCARG(uap, buf); 3488 iov.iov_len = done; 3489 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3490 } 3491 #endif 3492 *retval = done; 3493 out: 3494 FILE_UNUSE(fp, l); 3495 return (error); 3496 } 3497 3498 /* 3499 * Set the mode mask for creation of filesystem nodes. 3500 */ 3501 int 3502 sys_umask(struct lwp *l, void *v, register_t *retval) 3503 { 3504 struct sys_umask_args /* { 3505 syscallarg(mode_t) newmask; 3506 } */ *uap = v; 3507 struct proc *p = l->l_proc; 3508 struct cwdinfo *cwdi; 3509 3510 cwdi = p->p_cwdi; 3511 *retval = cwdi->cwdi_cmask; 3512 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3513 return (0); 3514 } 3515 3516 /* 3517 * Void all references to file by ripping underlying filesystem 3518 * away from vnode. 3519 */ 3520 /* ARGSUSED */ 3521 int 3522 sys_revoke(struct lwp *l, void *v, register_t *retval) 3523 { 3524 struct sys_revoke_args /* { 3525 syscallarg(const char *) path; 3526 } */ *uap = v; 3527 struct vnode *vp; 3528 struct vattr vattr; 3529 int error; 3530 struct nameidata nd; 3531 3532 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3533 if ((error = namei(&nd)) != 0) 3534 return (error); 3535 vp = nd.ni_vp; 3536 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3537 goto out; 3538 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3539 (error = kauth_authorize_generic(l->l_cred, 3540 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3541 goto out; 3542 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3543 VOP_REVOKE(vp, REVOKEALL); 3544 out: 3545 vrele(vp); 3546 return (error); 3547 } 3548 3549 /* 3550 * Convert a user file descriptor to a kernel file entry. 3551 */ 3552 int 3553 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3554 { 3555 struct vnode *vp; 3556 struct file *fp; 3557 3558 if ((fp = fd_getfile(fdp, fd)) == NULL) 3559 return (EBADF); 3560 3561 FILE_USE(fp); 3562 3563 if (fp->f_type != DTYPE_VNODE) { 3564 FILE_UNUSE(fp, NULL); 3565 return (EINVAL); 3566 } 3567 3568 vp = (struct vnode *)fp->f_data; 3569 if (vp->v_type == VBAD) { 3570 FILE_UNUSE(fp, NULL); 3571 return (EBADF); 3572 } 3573 3574 *fpp = fp; 3575 return (0); 3576 } 3577
Indexes created Thu Sep 25 15:09:46 GMT 2025