vfs_syscalls.c revision 1.318
1 /* $NetBSD: vfs_syscalls.c,v 1.318 2007/06/07 10:03:12 hannken 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 1.318 2007/06/07 10:03:12 hannken 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 if (error) { 631 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 632 (void) vfs_allocate_syncvnode(mp); 633 simple_lock(&mountlist_slock); 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 simple_lock(&mountlist_slock); 651 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 652 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 653 coveredvp->v_mountedhere = NULL; 654 mp->mnt_op->vfs_refcount--; 655 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 656 panic("unmount: dangling vnode"); 657 mp->mnt_iflag |= IMNT_GONE; 658 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 659 if (coveredvp != NULLVP) 660 vrele(coveredvp); 661 mount_finispecific(mp); 662 if (used_syncer) 663 mutex_exit(&syncer_mutex); 664 simple_lock(&mp->mnt_slock); 665 while (mp->mnt_wcnt > 0) { 666 wakeup(mp); 667 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 668 } 669 simple_unlock(&mp->mnt_slock); 670 vfs_hooks_unmount(mp); 671 free(mp, M_MOUNT); 672 return (0); 673 } 674 675 /* 676 * Sync each mounted filesystem. 677 */ 678 #ifdef DEBUG 679 int syncprt = 0; 680 struct ctldebug debug0 = { "syncprt", &syncprt }; 681 #endif 682 683 /* ARGSUSED */ 684 int 685 sys_sync(struct lwp *l, void *v, register_t *retval) 686 { 687 struct mount *mp, *nmp; 688 int asyncflag; 689 690 if (l == NULL) 691 l = &lwp0; 692 693 simple_lock(&mountlist_slock); 694 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 695 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 696 nmp = mp->mnt_list.cqe_prev; 697 continue; 698 } 699 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 700 asyncflag = mp->mnt_flag & MNT_ASYNC; 701 mp->mnt_flag &= ~MNT_ASYNC; 702 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 703 if (asyncflag) 704 mp->mnt_flag |= MNT_ASYNC; 705 } 706 simple_lock(&mountlist_slock); 707 nmp = mp->mnt_list.cqe_prev; 708 vfs_unbusy(mp); 709 710 } 711 simple_unlock(&mountlist_slock); 712 #ifdef DEBUG 713 if (syncprt) 714 vfs_bufstats(); 715 #endif /* DEBUG */ 716 return (0); 717 } 718 719 /* 720 * Change filesystem quotas. 721 */ 722 /* ARGSUSED */ 723 int 724 sys_quotactl(struct lwp *l, void *v, register_t *retval) 725 { 726 struct sys_quotactl_args /* { 727 syscallarg(const char *) path; 728 syscallarg(int) cmd; 729 syscallarg(int) uid; 730 syscallarg(void *) arg; 731 } */ *uap = v; 732 struct mount *mp; 733 int error; 734 struct nameidata nd; 735 736 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 737 if ((error = namei(&nd)) != 0) 738 return (error); 739 mp = nd.ni_vp->v_mount; 740 vrele(nd.ni_vp); 741 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 742 SCARG(uap, arg), l); 743 return (error); 744 } 745 746 int 747 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 748 int root) 749 { 750 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 751 int error = 0; 752 753 /* 754 * If MNT_NOWAIT or MNT_LAZY is specified, do not 755 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 756 * overrides MNT_NOWAIT. 757 */ 758 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 759 (flags != MNT_WAIT && flags != 0)) { 760 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 761 goto done; 762 } 763 764 /* Get the filesystem stats now */ 765 memset(sp, 0, sizeof(*sp)); 766 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 767 return error; 768 } 769 770 if (cwdi->cwdi_rdir == NULL) 771 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 772 done: 773 if (cwdi->cwdi_rdir != NULL) { 774 size_t len; 775 char *bp; 776 char *path = PNBUF_GET(); 777 778 bp = path + MAXPATHLEN; 779 *--bp = '\0'; 780 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 781 MAXPATHLEN / 2, 0, l); 782 if (error) { 783 PNBUF_PUT(path); 784 return error; 785 } 786 len = strlen(bp); 787 /* 788 * for mount points that are below our root, we can see 789 * them, so we fix up the pathname and return them. The 790 * rest we cannot see, so we don't allow viewing the 791 * data. 792 */ 793 if (strncmp(bp, sp->f_mntonname, len) == 0) { 794 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 795 sizeof(sp->f_mntonname)); 796 if (sp->f_mntonname[0] == '\0') 797 (void)strlcpy(sp->f_mntonname, "/", 798 sizeof(sp->f_mntonname)); 799 } else { 800 if (root) 801 (void)strlcpy(sp->f_mntonname, "/", 802 sizeof(sp->f_mntonname)); 803 else 804 error = EPERM; 805 } 806 PNBUF_PUT(path); 807 } 808 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 809 return error; 810 } 811 812 /* 813 * Get filesystem statistics by path. 814 */ 815 int 816 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 817 { 818 struct mount *mp; 819 int error; 820 struct nameidata nd; 821 822 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 823 if ((error = namei(&nd)) != 0) 824 return error; 825 mp = nd.ni_vp->v_mount; 826 error = dostatvfs(mp, sb, l, flags, 1); 827 vrele(nd.ni_vp); 828 return error; 829 } 830 831 /* ARGSUSED */ 832 int 833 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 834 { 835 struct sys_statvfs1_args /* { 836 syscallarg(const char *) path; 837 syscallarg(struct statvfs *) buf; 838 syscallarg(int) flags; 839 } */ *uap = v; 840 struct statvfs *sb; 841 int error; 842 843 sb = STATVFSBUF_GET(); 844 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 845 if (error == 0) 846 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 847 STATVFSBUF_PUT(sb); 848 return error; 849 } 850 851 /* 852 * Get filesystem statistics by fd. 853 */ 854 int 855 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 856 { 857 struct proc *p = l->l_proc; 858 struct file *fp; 859 struct mount *mp; 860 int error; 861 862 /* getvnode() will use the descriptor for us */ 863 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 864 return (error); 865 mp = ((struct vnode *)fp->f_data)->v_mount; 866 error = dostatvfs(mp, sb, l, flags, 1); 867 FILE_UNUSE(fp, l); 868 return error; 869 } 870 871 /* ARGSUSED */ 872 int 873 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 874 { 875 struct sys_fstatvfs1_args /* { 876 syscallarg(int) fd; 877 syscallarg(struct statvfs *) buf; 878 syscallarg(int) flags; 879 } */ *uap = v; 880 struct statvfs *sb; 881 int error; 882 883 sb = STATVFSBUF_GET(); 884 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 885 if (error == 0) 886 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 887 STATVFSBUF_PUT(sb); 888 return error; 889 } 890 891 892 /* 893 * Get statistics on all filesystems. 894 */ 895 int 896 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 897 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 898 register_t *retval) 899 { 900 int root = 0; 901 struct proc *p = l->l_proc; 902 struct mount *mp, *nmp; 903 struct statvfs *sb; 904 size_t count, maxcount; 905 int error = 0; 906 907 sb = STATVFSBUF_GET(); 908 maxcount = bufsize / entry_sz; 909 simple_lock(&mountlist_slock); 910 count = 0; 911 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 912 mp = nmp) { 913 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 914 nmp = CIRCLEQ_NEXT(mp, mnt_list); 915 continue; 916 } 917 if (sfsp && count < maxcount) { 918 error = dostatvfs(mp, sb, l, flags, 0); 919 if (error) { 920 simple_lock(&mountlist_slock); 921 nmp = CIRCLEQ_NEXT(mp, mnt_list); 922 vfs_unbusy(mp); 923 continue; 924 } 925 error = copyfn(sb, sfsp, entry_sz); 926 if (error) { 927 vfs_unbusy(mp); 928 goto out; 929 } 930 sfsp = (char *)sfsp + entry_sz; 931 root |= strcmp(sb->f_mntonname, "/") == 0; 932 } 933 count++; 934 simple_lock(&mountlist_slock); 935 nmp = CIRCLEQ_NEXT(mp, mnt_list); 936 vfs_unbusy(mp); 937 } 938 simple_unlock(&mountlist_slock); 939 if (root == 0 && p->p_cwdi->cwdi_rdir) { 940 /* 941 * fake a root entry 942 */ 943 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 944 if (error != 0) 945 goto out; 946 if (sfsp) 947 error = copyfn(sb, sfsp, entry_sz); 948 count++; 949 } 950 if (sfsp && count > maxcount) 951 *retval = maxcount; 952 else 953 *retval = count; 954 out: 955 STATVFSBUF_PUT(sb); 956 return error; 957 } 958 959 int 960 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 961 { 962 struct sys_getvfsstat_args /* { 963 syscallarg(struct statvfs *) buf; 964 syscallarg(size_t) bufsize; 965 syscallarg(int) flags; 966 } */ *uap = v; 967 968 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 969 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 970 } 971 972 /* 973 * Change current working directory to a given file descriptor. 974 */ 975 /* ARGSUSED */ 976 int 977 sys_fchdir(struct lwp *l, void *v, register_t *retval) 978 { 979 struct sys_fchdir_args /* { 980 syscallarg(int) fd; 981 } */ *uap = v; 982 struct proc *p = l->l_proc; 983 struct filedesc *fdp = p->p_fd; 984 struct cwdinfo *cwdi = p->p_cwdi; 985 struct vnode *vp, *tdp; 986 struct mount *mp; 987 struct file *fp; 988 int error; 989 990 /* getvnode() will use the descriptor for us */ 991 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 992 return (error); 993 vp = (struct vnode *)fp->f_data; 994 995 VREF(vp); 996 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 997 if (vp->v_type != VDIR) 998 error = ENOTDIR; 999 else 1000 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1001 if (error) { 1002 vput(vp); 1003 goto out; 1004 } 1005 while ((mp = vp->v_mountedhere) != NULL) { 1006 if (vfs_busy(mp, 0, 0)) 1007 continue; 1008 1009 vput(vp); 1010 error = VFS_ROOT(mp, &tdp); 1011 vfs_unbusy(mp); 1012 if (error) 1013 goto out; 1014 vp = tdp; 1015 } 1016 VOP_UNLOCK(vp, 0); 1017 1018 /* 1019 * Disallow changing to a directory not under the process's 1020 * current root directory (if there is one). 1021 */ 1022 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1023 vrele(vp); 1024 error = EPERM; /* operation not permitted */ 1025 goto out; 1026 } 1027 1028 vrele(cwdi->cwdi_cdir); 1029 cwdi->cwdi_cdir = vp; 1030 out: 1031 FILE_UNUSE(fp, l); 1032 return (error); 1033 } 1034 1035 /* 1036 * Change this process's notion of the root directory to a given file 1037 * descriptor. 1038 */ 1039 int 1040 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1041 { 1042 struct sys_fchroot_args *uap = v; 1043 struct proc *p = l->l_proc; 1044 struct filedesc *fdp = p->p_fd; 1045 struct cwdinfo *cwdi = p->p_cwdi; 1046 struct vnode *vp; 1047 struct file *fp; 1048 int error; 1049 1050 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1051 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1052 return error; 1053 /* getvnode() will use the descriptor for us */ 1054 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1055 return error; 1056 vp = (struct vnode *) fp->f_data; 1057 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1058 if (vp->v_type != VDIR) 1059 error = ENOTDIR; 1060 else 1061 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1062 VOP_UNLOCK(vp, 0); 1063 if (error) 1064 goto out; 1065 VREF(vp); 1066 1067 /* 1068 * Prevent escaping from chroot by putting the root under 1069 * the working directory. Silently chdir to / if we aren't 1070 * already there. 1071 */ 1072 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1073 /* 1074 * XXX would be more failsafe to change directory to a 1075 * deadfs node here instead 1076 */ 1077 vrele(cwdi->cwdi_cdir); 1078 VREF(vp); 1079 cwdi->cwdi_cdir = vp; 1080 } 1081 1082 if (cwdi->cwdi_rdir != NULL) 1083 vrele(cwdi->cwdi_rdir); 1084 cwdi->cwdi_rdir = vp; 1085 out: 1086 FILE_UNUSE(fp, l); 1087 return (error); 1088 } 1089 1090 /* 1091 * Change current working directory (``.''). 1092 */ 1093 /* ARGSUSED */ 1094 int 1095 sys_chdir(struct lwp *l, void *v, register_t *retval) 1096 { 1097 struct sys_chdir_args /* { 1098 syscallarg(const char *) path; 1099 } */ *uap = v; 1100 struct proc *p = l->l_proc; 1101 struct cwdinfo *cwdi = p->p_cwdi; 1102 int error; 1103 struct nameidata nd; 1104 1105 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1106 SCARG(uap, path), l); 1107 if ((error = change_dir(&nd, l)) != 0) 1108 return (error); 1109 vrele(cwdi->cwdi_cdir); 1110 cwdi->cwdi_cdir = nd.ni_vp; 1111 return (0); 1112 } 1113 1114 /* 1115 * Change notion of root (``/'') directory. 1116 */ 1117 /* ARGSUSED */ 1118 int 1119 sys_chroot(struct lwp *l, void *v, register_t *retval) 1120 { 1121 struct sys_chroot_args /* { 1122 syscallarg(const char *) path; 1123 } */ *uap = v; 1124 struct proc *p = l->l_proc; 1125 struct cwdinfo *cwdi = p->p_cwdi; 1126 struct vnode *vp; 1127 int error; 1128 struct nameidata nd; 1129 1130 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1131 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1132 return (error); 1133 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1134 SCARG(uap, path), l); 1135 if ((error = change_dir(&nd, l)) != 0) 1136 return (error); 1137 if (cwdi->cwdi_rdir != NULL) 1138 vrele(cwdi->cwdi_rdir); 1139 vp = nd.ni_vp; 1140 cwdi->cwdi_rdir = vp; 1141 1142 /* 1143 * Prevent escaping from chroot by putting the root under 1144 * the working directory. Silently chdir to / if we aren't 1145 * already there. 1146 */ 1147 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1148 /* 1149 * XXX would be more failsafe to change directory to a 1150 * deadfs node here instead 1151 */ 1152 vrele(cwdi->cwdi_cdir); 1153 VREF(vp); 1154 cwdi->cwdi_cdir = vp; 1155 } 1156 1157 return (0); 1158 } 1159 1160 /* 1161 * Common routine for chroot and chdir. 1162 */ 1163 static int 1164 change_dir(struct nameidata *ndp, struct lwp *l) 1165 { 1166 struct vnode *vp; 1167 int error; 1168 1169 if ((error = namei(ndp)) != 0) 1170 return (error); 1171 vp = ndp->ni_vp; 1172 if (vp->v_type != VDIR) 1173 error = ENOTDIR; 1174 else 1175 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1176 1177 if (error) 1178 vput(vp); 1179 else 1180 VOP_UNLOCK(vp, 0); 1181 return (error); 1182 } 1183 1184 /* 1185 * Check permissions, allocate an open file structure, 1186 * and call the device open routine if any. 1187 */ 1188 int 1189 sys_open(struct lwp *l, void *v, register_t *retval) 1190 { 1191 struct sys_open_args /* { 1192 syscallarg(const char *) path; 1193 syscallarg(int) flags; 1194 syscallarg(int) mode; 1195 } */ *uap = v; 1196 struct proc *p = l->l_proc; 1197 struct cwdinfo *cwdi = p->p_cwdi; 1198 struct filedesc *fdp = p->p_fd; 1199 struct file *fp; 1200 struct vnode *vp; 1201 int flags, cmode; 1202 int type, indx, error; 1203 struct flock lf; 1204 struct nameidata nd; 1205 1206 flags = FFLAGS(SCARG(uap, flags)); 1207 if ((flags & (FREAD | FWRITE)) == 0) 1208 return (EINVAL); 1209 /* falloc() will use the file descriptor for us */ 1210 if ((error = falloc(l, &fp, &indx)) != 0) 1211 return (error); 1212 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1213 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1214 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1215 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1216 FILE_UNUSE(fp, l); 1217 fdp->fd_ofiles[indx] = NULL; 1218 ffree(fp); 1219 if ((error == EDUPFD || error == EMOVEFD) && 1220 l->l_dupfd >= 0 && /* XXX from fdopen */ 1221 (error = 1222 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1223 *retval = indx; 1224 return (0); 1225 } 1226 if (error == ERESTART) 1227 error = EINTR; 1228 fdremove(fdp, indx); 1229 return (error); 1230 } 1231 l->l_dupfd = 0; 1232 vp = nd.ni_vp; 1233 fp->f_flag = flags & FMASK; 1234 fp->f_type = DTYPE_VNODE; 1235 fp->f_ops = &vnops; 1236 fp->f_data = vp; 1237 if (flags & (O_EXLOCK | O_SHLOCK)) { 1238 lf.l_whence = SEEK_SET; 1239 lf.l_start = 0; 1240 lf.l_len = 0; 1241 if (flags & O_EXLOCK) 1242 lf.l_type = F_WRLCK; 1243 else 1244 lf.l_type = F_RDLCK; 1245 type = F_FLOCK; 1246 if ((flags & FNONBLOCK) == 0) 1247 type |= F_WAIT; 1248 VOP_UNLOCK(vp, 0); 1249 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1250 if (error) { 1251 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1252 FILE_UNUSE(fp, l); 1253 ffree(fp); 1254 fdremove(fdp, indx); 1255 return (error); 1256 } 1257 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1258 fp->f_flag |= FHASLOCK; 1259 } 1260 VOP_UNLOCK(vp, 0); 1261 *retval = indx; 1262 FILE_SET_MATURE(fp); 1263 FILE_UNUSE(fp, l); 1264 return (0); 1265 } 1266 1267 static void 1268 vfs__fhfree(fhandle_t *fhp) 1269 { 1270 size_t fhsize; 1271 1272 if (fhp == NULL) { 1273 return; 1274 } 1275 fhsize = FHANDLE_SIZE(fhp); 1276 kmem_free(fhp, fhsize); 1277 } 1278 1279 /* 1280 * vfs_composefh: compose a filehandle. 1281 */ 1282 1283 int 1284 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1285 { 1286 struct mount *mp; 1287 struct fid *fidp; 1288 int error; 1289 size_t needfhsize; 1290 size_t fidsize; 1291 1292 mp = vp->v_mount; 1293 fidp = NULL; 1294 if (*fh_size < FHANDLE_SIZE_MIN) { 1295 fidsize = 0; 1296 } else { 1297 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1298 if (fhp != NULL) { 1299 memset(fhp, 0, *fh_size); 1300 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1301 fidp = &fhp->fh_fid; 1302 } 1303 } 1304 error = VFS_VPTOFH(vp, fidp, &fidsize); 1305 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1306 if (error == 0 && *fh_size < needfhsize) { 1307 error = E2BIG; 1308 } 1309 *fh_size = needfhsize; 1310 return error; 1311 } 1312 1313 int 1314 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1315 { 1316 struct mount *mp; 1317 fhandle_t *fhp; 1318 size_t fhsize; 1319 size_t fidsize; 1320 int error; 1321 1322 *fhpp = NULL; 1323 mp = vp->v_mount; 1324 fidsize = 0; 1325 error = VFS_VPTOFH(vp, NULL, &fidsize); 1326 KASSERT(error != 0); 1327 if (error != E2BIG) { 1328 goto out; 1329 } 1330 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1331 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1332 if (fhp == NULL) { 1333 error = ENOMEM; 1334 goto out; 1335 } 1336 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1337 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1338 if (error == 0) { 1339 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1340 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1341 *fhpp = fhp; 1342 } else { 1343 kmem_free(fhp, fhsize); 1344 } 1345 out: 1346 return error; 1347 } 1348 1349 void 1350 vfs_composefh_free(fhandle_t *fhp) 1351 { 1352 1353 vfs__fhfree(fhp); 1354 } 1355 1356 /* 1357 * vfs_fhtovp: lookup a vnode by a filehandle. 1358 */ 1359 1360 int 1361 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1362 { 1363 struct mount *mp; 1364 int error; 1365 1366 *vpp = NULL; 1367 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1368 if (mp == NULL) { 1369 error = ESTALE; 1370 goto out; 1371 } 1372 if (mp->mnt_op->vfs_fhtovp == NULL) { 1373 error = EOPNOTSUPP; 1374 goto out; 1375 } 1376 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1377 out: 1378 return error; 1379 } 1380 1381 /* 1382 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1383 * the needed size. 1384 */ 1385 1386 int 1387 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1388 { 1389 fhandle_t *fhp; 1390 int error; 1391 1392 *fhpp = NULL; 1393 if (fhsize > FHANDLE_SIZE_MAX) { 1394 return EINVAL; 1395 } 1396 if (fhsize < FHANDLE_SIZE_MIN) { 1397 return EINVAL; 1398 } 1399 again: 1400 fhp = kmem_alloc(fhsize, KM_SLEEP); 1401 if (fhp == NULL) { 1402 return ENOMEM; 1403 } 1404 error = copyin(ufhp, fhp, fhsize); 1405 if (error == 0) { 1406 /* XXX this check shouldn't be here */ 1407 if (FHANDLE_SIZE(fhp) == fhsize) { 1408 *fhpp = fhp; 1409 return 0; 1410 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1411 /* 1412 * a kludge for nfsv2 padded handles. 1413 */ 1414 size_t sz; 1415 1416 sz = FHANDLE_SIZE(fhp); 1417 kmem_free(fhp, fhsize); 1418 fhsize = sz; 1419 goto again; 1420 } else { 1421 /* 1422 * userland told us wrong size. 1423 */ 1424 error = EINVAL; 1425 } 1426 } 1427 kmem_free(fhp, fhsize); 1428 return error; 1429 } 1430 1431 void 1432 vfs_copyinfh_free(fhandle_t *fhp) 1433 { 1434 1435 vfs__fhfree(fhp); 1436 } 1437 1438 /* 1439 * Get file handle system call 1440 */ 1441 int 1442 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1443 { 1444 struct sys___getfh30_args /* { 1445 syscallarg(char *) fname; 1446 syscallarg(fhandle_t *) fhp; 1447 syscallarg(size_t *) fh_size; 1448 } */ *uap = v; 1449 struct vnode *vp; 1450 fhandle_t *fh; 1451 int error; 1452 struct nameidata nd; 1453 size_t sz; 1454 size_t usz; 1455 1456 /* 1457 * Must be super user 1458 */ 1459 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1460 0, NULL, NULL, NULL); 1461 if (error) 1462 return (error); 1463 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1464 SCARG(uap, fname), l); 1465 error = namei(&nd); 1466 if (error) 1467 return (error); 1468 vp = nd.ni_vp; 1469 error = vfs_composefh_alloc(vp, &fh); 1470 vput(vp); 1471 if (error != 0) { 1472 goto out; 1473 } 1474 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1475 if (error != 0) { 1476 goto out; 1477 } 1478 sz = FHANDLE_SIZE(fh); 1479 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1480 if (error != 0) { 1481 goto out; 1482 } 1483 if (usz >= sz) { 1484 error = copyout(fh, SCARG(uap, fhp), sz); 1485 } else { 1486 error = E2BIG; 1487 } 1488 out: 1489 vfs_composefh_free(fh); 1490 return (error); 1491 } 1492 1493 /* 1494 * Open a file given a file handle. 1495 * 1496 * Check permissions, allocate an open file structure, 1497 * and call the device open routine if any. 1498 */ 1499 1500 int 1501 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1502 register_t *retval) 1503 { 1504 struct filedesc *fdp = l->l_proc->p_fd; 1505 struct file *fp; 1506 struct vnode *vp = NULL; 1507 kauth_cred_t cred = l->l_cred; 1508 struct file *nfp; 1509 int type, indx, error=0; 1510 struct flock lf; 1511 struct vattr va; 1512 fhandle_t *fh; 1513 int flags; 1514 1515 /* 1516 * Must be super user 1517 */ 1518 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1519 0, NULL, NULL, NULL))) 1520 return (error); 1521 1522 flags = FFLAGS(oflags); 1523 if ((flags & (FREAD | FWRITE)) == 0) 1524 return (EINVAL); 1525 if ((flags & O_CREAT)) 1526 return (EINVAL); 1527 /* falloc() will use the file descriptor for us */ 1528 if ((error = falloc(l, &nfp, &indx)) != 0) 1529 return (error); 1530 fp = nfp; 1531 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1532 if (error != 0) { 1533 goto bad; 1534 } 1535 error = vfs_fhtovp(fh, &vp); 1536 if (error != 0) { 1537 goto bad; 1538 } 1539 1540 /* Now do an effective vn_open */ 1541 1542 if (vp->v_type == VSOCK) { 1543 error = EOPNOTSUPP; 1544 goto bad; 1545 } 1546 if (flags & FREAD) { 1547 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1548 goto bad; 1549 } 1550 if (flags & (FWRITE | O_TRUNC)) { 1551 if (vp->v_type == VDIR) { 1552 error = EISDIR; 1553 goto bad; 1554 } 1555 if ((error = vn_writechk(vp)) != 0 || 1556 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1557 goto bad; 1558 } 1559 if (flags & O_TRUNC) { 1560 VOP_UNLOCK(vp, 0); /* XXX */ 1561 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1562 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1563 VATTR_NULL(&va); 1564 va.va_size = 0; 1565 error = VOP_SETATTR(vp, &va, cred, l); 1566 if (error) 1567 goto bad; 1568 } 1569 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1570 goto bad; 1571 if (vp->v_type == VREG && 1572 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1573 error = EIO; 1574 goto bad; 1575 } 1576 if (flags & FWRITE) 1577 vp->v_writecount++; 1578 1579 /* done with modified vn_open, now finish what sys_open does. */ 1580 1581 fp->f_flag = flags & FMASK; 1582 fp->f_type = DTYPE_VNODE; 1583 fp->f_ops = &vnops; 1584 fp->f_data = vp; 1585 if (flags & (O_EXLOCK | O_SHLOCK)) { 1586 lf.l_whence = SEEK_SET; 1587 lf.l_start = 0; 1588 lf.l_len = 0; 1589 if (flags & O_EXLOCK) 1590 lf.l_type = F_WRLCK; 1591 else 1592 lf.l_type = F_RDLCK; 1593 type = F_FLOCK; 1594 if ((flags & FNONBLOCK) == 0) 1595 type |= F_WAIT; 1596 VOP_UNLOCK(vp, 0); 1597 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1598 if (error) { 1599 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1600 FILE_UNUSE(fp, l); 1601 ffree(fp); 1602 fdremove(fdp, indx); 1603 return (error); 1604 } 1605 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1606 fp->f_flag |= FHASLOCK; 1607 } 1608 VOP_UNLOCK(vp, 0); 1609 *retval = indx; 1610 FILE_SET_MATURE(fp); 1611 FILE_UNUSE(fp, l); 1612 vfs_copyinfh_free(fh); 1613 return (0); 1614 1615 bad: 1616 FILE_UNUSE(fp, l); 1617 ffree(fp); 1618 fdremove(fdp, indx); 1619 if (vp != NULL) 1620 vput(vp); 1621 vfs_copyinfh_free(fh); 1622 return (error); 1623 } 1624 1625 int 1626 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1627 { 1628 struct sys___fhopen40_args /* { 1629 syscallarg(const void *) fhp; 1630 syscallarg(size_t) fh_size; 1631 syscallarg(int) flags; 1632 } */ *uap = v; 1633 1634 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1635 SCARG(uap, flags), retval); 1636 } 1637 1638 int 1639 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1640 { 1641 int error; 1642 fhandle_t *fh; 1643 struct vnode *vp; 1644 1645 /* 1646 * Must be super user 1647 */ 1648 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1649 0, NULL, NULL, NULL))) 1650 return (error); 1651 1652 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1653 if (error != 0) 1654 return error; 1655 1656 error = vfs_fhtovp(fh, &vp); 1657 vfs_copyinfh_free(fh); 1658 if (error != 0) 1659 return error; 1660 1661 error = vn_stat(vp, sb, l); 1662 vput(vp); 1663 return error; 1664 } 1665 1666 1667 /* ARGSUSED */ 1668 int 1669 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1670 { 1671 struct sys___fhstat40_args /* { 1672 syscallarg(const void *) fhp; 1673 syscallarg(size_t) fh_size; 1674 syscallarg(struct stat *) sb; 1675 } */ *uap = v; 1676 struct stat sb; 1677 int error; 1678 1679 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1680 if (error) 1681 return error; 1682 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1683 } 1684 1685 int 1686 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1687 int flags) 1688 { 1689 fhandle_t *fh; 1690 struct mount *mp; 1691 struct vnode *vp; 1692 int error; 1693 1694 /* 1695 * Must be super user 1696 */ 1697 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1698 0, NULL, NULL, NULL))) 1699 return error; 1700 1701 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1702 if (error != 0) 1703 return error; 1704 1705 error = vfs_fhtovp(fh, &vp); 1706 vfs_copyinfh_free(fh); 1707 if (error != 0) 1708 return error; 1709 1710 mp = vp->v_mount; 1711 error = dostatvfs(mp, sb, l, flags, 1); 1712 vput(vp); 1713 return error; 1714 } 1715 1716 /* ARGSUSED */ 1717 int 1718 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1719 { 1720 struct sys___fhstatvfs140_args /* { 1721 syscallarg(const void *) fhp; 1722 syscallarg(size_t) fh_size; 1723 syscallarg(struct statvfs *) buf; 1724 syscallarg(int) flags; 1725 } */ *uap = v; 1726 struct statvfs *sb = STATVFSBUF_GET(); 1727 int error; 1728 1729 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1730 SCARG(uap, flags)); 1731 if (error == 0) 1732 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1733 STATVFSBUF_PUT(sb); 1734 return error; 1735 } 1736 1737 /* 1738 * Create a special file. 1739 */ 1740 /* ARGSUSED */ 1741 int 1742 sys_mknod(struct lwp *l, void *v, register_t *retval) 1743 { 1744 struct sys_mknod_args /* { 1745 syscallarg(const char *) path; 1746 syscallarg(int) mode; 1747 syscallarg(int) dev; 1748 } */ *uap = v; 1749 struct proc *p = l->l_proc; 1750 struct vnode *vp; 1751 struct vattr vattr; 1752 int error, optype; 1753 struct nameidata nd; 1754 char *path; 1755 const char *cpath; 1756 enum uio_seg seg = UIO_USERSPACE; 1757 1758 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1759 0, NULL, NULL, NULL)) != 0) 1760 return (error); 1761 1762 optype = VOP_MKNOD_DESCOFFSET; 1763 1764 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1765 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1766 1767 if ((error = namei(&nd)) != 0) 1768 goto out; 1769 vp = nd.ni_vp; 1770 if (vp != NULL) 1771 error = EEXIST; 1772 else { 1773 VATTR_NULL(&vattr); 1774 vattr.va_mode = 1775 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1776 vattr.va_rdev = SCARG(uap, dev); 1777 1778 switch (SCARG(uap, mode) & S_IFMT) { 1779 case S_IFMT: /* used by badsect to flag bad sectors */ 1780 vattr.va_type = VBAD; 1781 break; 1782 case S_IFCHR: 1783 vattr.va_type = VCHR; 1784 break; 1785 case S_IFBLK: 1786 vattr.va_type = VBLK; 1787 break; 1788 case S_IFWHT: 1789 optype = VOP_WHITEOUT_DESCOFFSET; 1790 break; 1791 case S_IFREG: 1792 #if NVERIEXEC > 0 1793 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1794 O_CREAT); 1795 #endif /* NVERIEXEC > 0 */ 1796 vattr.va_type = VREG; 1797 vattr.va_rdev = VNOVAL; 1798 optype = VOP_CREATE_DESCOFFSET; 1799 break; 1800 default: 1801 error = EINVAL; 1802 break; 1803 } 1804 } 1805 if (!error) { 1806 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1807 switch (optype) { 1808 case VOP_WHITEOUT_DESCOFFSET: 1809 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1810 if (error) 1811 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1812 vput(nd.ni_dvp); 1813 break; 1814 1815 case VOP_MKNOD_DESCOFFSET: 1816 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1817 &nd.ni_cnd, &vattr); 1818 if (error == 0) 1819 vput(nd.ni_vp); 1820 break; 1821 1822 case VOP_CREATE_DESCOFFSET: 1823 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1824 &nd.ni_cnd, &vattr); 1825 if (error == 0) 1826 vput(nd.ni_vp); 1827 break; 1828 } 1829 } else { 1830 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1831 if (nd.ni_dvp == vp) 1832 vrele(nd.ni_dvp); 1833 else 1834 vput(nd.ni_dvp); 1835 if (vp) 1836 vrele(vp); 1837 } 1838 out: 1839 VERIEXEC_PATH_PUT(path); 1840 return (error); 1841 } 1842 1843 /* 1844 * Create a named pipe. 1845 */ 1846 /* ARGSUSED */ 1847 int 1848 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1849 { 1850 struct sys_mkfifo_args /* { 1851 syscallarg(const char *) path; 1852 syscallarg(int) mode; 1853 } */ *uap = v; 1854 struct proc *p = l->l_proc; 1855 struct vattr vattr; 1856 int error; 1857 struct nameidata nd; 1858 1859 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1860 if ((error = namei(&nd)) != 0) 1861 return (error); 1862 if (nd.ni_vp != NULL) { 1863 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1864 if (nd.ni_dvp == nd.ni_vp) 1865 vrele(nd.ni_dvp); 1866 else 1867 vput(nd.ni_dvp); 1868 vrele(nd.ni_vp); 1869 return (EEXIST); 1870 } 1871 VATTR_NULL(&vattr); 1872 vattr.va_type = VFIFO; 1873 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1874 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1875 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1876 if (error == 0) 1877 vput(nd.ni_vp); 1878 return (error); 1879 } 1880 1881 /* 1882 * Make a hard file link. 1883 */ 1884 /* ARGSUSED */ 1885 int 1886 sys_link(struct lwp *l, void *v, register_t *retval) 1887 { 1888 struct sys_link_args /* { 1889 syscallarg(const char *) path; 1890 syscallarg(const char *) link; 1891 } */ *uap = v; 1892 struct vnode *vp; 1893 struct nameidata nd; 1894 int error; 1895 1896 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1897 if ((error = namei(&nd)) != 0) 1898 return (error); 1899 vp = nd.ni_vp; 1900 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1901 if ((error = namei(&nd)) != 0) 1902 goto out; 1903 if (nd.ni_vp) { 1904 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1905 if (nd.ni_dvp == nd.ni_vp) 1906 vrele(nd.ni_dvp); 1907 else 1908 vput(nd.ni_dvp); 1909 vrele(nd.ni_vp); 1910 error = EEXIST; 1911 goto out; 1912 } 1913 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1914 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1915 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1916 out: 1917 vrele(vp); 1918 return (error); 1919 } 1920 1921 /* 1922 * Make a symbolic link. 1923 */ 1924 /* ARGSUSED */ 1925 int 1926 sys_symlink(struct lwp *l, void *v, register_t *retval) 1927 { 1928 struct sys_symlink_args /* { 1929 syscallarg(const char *) path; 1930 syscallarg(const char *) link; 1931 } */ *uap = v; 1932 struct proc *p = l->l_proc; 1933 struct vattr vattr; 1934 char *path; 1935 int error; 1936 struct nameidata nd; 1937 1938 path = PNBUF_GET(); 1939 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1940 if (error) 1941 goto out; 1942 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1943 if ((error = namei(&nd)) != 0) 1944 goto out; 1945 if (nd.ni_vp) { 1946 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1947 if (nd.ni_dvp == nd.ni_vp) 1948 vrele(nd.ni_dvp); 1949 else 1950 vput(nd.ni_dvp); 1951 vrele(nd.ni_vp); 1952 error = EEXIST; 1953 goto out; 1954 } 1955 VATTR_NULL(&vattr); 1956 vattr.va_type = VLNK; 1957 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1958 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1959 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1960 if (error == 0) 1961 vput(nd.ni_vp); 1962 out: 1963 PNBUF_PUT(path); 1964 return (error); 1965 } 1966 1967 /* 1968 * Delete a whiteout from the filesystem. 1969 */ 1970 /* ARGSUSED */ 1971 int 1972 sys_undelete(struct lwp *l, void *v, register_t *retval) 1973 { 1974 struct sys_undelete_args /* { 1975 syscallarg(const char *) path; 1976 } */ *uap = v; 1977 int error; 1978 struct nameidata nd; 1979 1980 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 1981 SCARG(uap, path), l); 1982 error = namei(&nd); 1983 if (error) 1984 return (error); 1985 1986 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1987 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1988 if (nd.ni_dvp == nd.ni_vp) 1989 vrele(nd.ni_dvp); 1990 else 1991 vput(nd.ni_dvp); 1992 if (nd.ni_vp) 1993 vrele(nd.ni_vp); 1994 return (EEXIST); 1995 } 1996 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1997 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1998 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1999 vput(nd.ni_dvp); 2000 return (error); 2001 } 2002 2003 /* 2004 * Delete a name from the filesystem. 2005 */ 2006 /* ARGSUSED */ 2007 int 2008 sys_unlink(struct lwp *l, void *v, register_t *retval) 2009 { 2010 struct sys_unlink_args /* { 2011 syscallarg(const char *) path; 2012 } */ *uap = v; 2013 struct vnode *vp; 2014 int error; 2015 struct nameidata nd; 2016 char *path; 2017 const char *cpath; 2018 enum uio_seg seg = UIO_USERSPACE; 2019 2020 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2021 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2022 2023 if ((error = namei(&nd)) != 0) 2024 goto out; 2025 vp = nd.ni_vp; 2026 2027 /* 2028 * The root of a mounted filesystem cannot be deleted. 2029 */ 2030 if (vp->v_flag & VROOT) { 2031 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2032 if (nd.ni_dvp == vp) 2033 vrele(nd.ni_dvp); 2034 else 2035 vput(nd.ni_dvp); 2036 vput(vp); 2037 error = EBUSY; 2038 goto out; 2039 } 2040 2041 #if NVERIEXEC > 0 2042 /* Handle remove requests for veriexec entries. */ 2043 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2044 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2045 if (nd.ni_dvp == vp) 2046 vrele(nd.ni_dvp); 2047 else 2048 vput(nd.ni_dvp); 2049 vput(vp); 2050 goto out; 2051 } 2052 #endif /* NVERIEXEC > 0 */ 2053 2054 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2055 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2056 #ifdef FILEASSOC 2057 (void)fileassoc_file_delete(vp); 2058 #endif /* FILEASSOC */ 2059 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2060 out: 2061 VERIEXEC_PATH_PUT(path); 2062 return (error); 2063 } 2064 2065 /* 2066 * Reposition read/write file offset. 2067 */ 2068 int 2069 sys_lseek(struct lwp *l, void *v, register_t *retval) 2070 { 2071 struct sys_lseek_args /* { 2072 syscallarg(int) fd; 2073 syscallarg(int) pad; 2074 syscallarg(off_t) offset; 2075 syscallarg(int) whence; 2076 } */ *uap = v; 2077 struct proc *p = l->l_proc; 2078 kauth_cred_t cred = l->l_cred; 2079 struct filedesc *fdp = p->p_fd; 2080 struct file *fp; 2081 struct vnode *vp; 2082 struct vattr vattr; 2083 off_t newoff; 2084 int error; 2085 2086 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2087 return (EBADF); 2088 2089 FILE_USE(fp); 2090 2091 vp = (struct vnode *)fp->f_data; 2092 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2093 error = ESPIPE; 2094 goto out; 2095 } 2096 2097 switch (SCARG(uap, whence)) { 2098 case SEEK_CUR: 2099 newoff = fp->f_offset + SCARG(uap, offset); 2100 break; 2101 case SEEK_END: 2102 error = VOP_GETATTR(vp, &vattr, cred, l); 2103 if (error) 2104 goto out; 2105 newoff = SCARG(uap, offset) + vattr.va_size; 2106 break; 2107 case SEEK_SET: 2108 newoff = SCARG(uap, offset); 2109 break; 2110 default: 2111 error = EINVAL; 2112 goto out; 2113 } 2114 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2115 goto out; 2116 2117 *(off_t *)retval = fp->f_offset = newoff; 2118 out: 2119 FILE_UNUSE(fp, l); 2120 return (error); 2121 } 2122 2123 /* 2124 * Positional read system call. 2125 */ 2126 int 2127 sys_pread(struct lwp *l, void *v, register_t *retval) 2128 { 2129 struct sys_pread_args /* { 2130 syscallarg(int) fd; 2131 syscallarg(void *) buf; 2132 syscallarg(size_t) nbyte; 2133 syscallarg(off_t) offset; 2134 } */ *uap = v; 2135 struct proc *p = l->l_proc; 2136 struct filedesc *fdp = p->p_fd; 2137 struct file *fp; 2138 struct vnode *vp; 2139 off_t offset; 2140 int error, fd = SCARG(uap, fd); 2141 2142 if ((fp = fd_getfile(fdp, fd)) == NULL) 2143 return (EBADF); 2144 2145 if ((fp->f_flag & FREAD) == 0) { 2146 simple_unlock(&fp->f_slock); 2147 return (EBADF); 2148 } 2149 2150 FILE_USE(fp); 2151 2152 vp = (struct vnode *)fp->f_data; 2153 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2154 error = ESPIPE; 2155 goto out; 2156 } 2157 2158 offset = SCARG(uap, offset); 2159 2160 /* 2161 * XXX This works because no file systems actually 2162 * XXX take any action on the seek operation. 2163 */ 2164 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2165 goto out; 2166 2167 /* dofileread() will unuse the descriptor for us */ 2168 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2169 &offset, 0, retval)); 2170 2171 out: 2172 FILE_UNUSE(fp, l); 2173 return (error); 2174 } 2175 2176 /* 2177 * Positional scatter read system call. 2178 */ 2179 int 2180 sys_preadv(struct lwp *l, void *v, register_t *retval) 2181 { 2182 struct sys_preadv_args /* { 2183 syscallarg(int) fd; 2184 syscallarg(const struct iovec *) iovp; 2185 syscallarg(int) iovcnt; 2186 syscallarg(off_t) offset; 2187 } */ *uap = v; 2188 struct proc *p = l->l_proc; 2189 struct filedesc *fdp = p->p_fd; 2190 struct file *fp; 2191 struct vnode *vp; 2192 off_t offset; 2193 int error, fd = SCARG(uap, fd); 2194 2195 if ((fp = fd_getfile(fdp, fd)) == NULL) 2196 return (EBADF); 2197 2198 if ((fp->f_flag & FREAD) == 0) { 2199 simple_unlock(&fp->f_slock); 2200 return (EBADF); 2201 } 2202 2203 FILE_USE(fp); 2204 2205 vp = (struct vnode *)fp->f_data; 2206 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2207 error = ESPIPE; 2208 goto out; 2209 } 2210 2211 offset = SCARG(uap, offset); 2212 2213 /* 2214 * XXX This works because no file systems actually 2215 * XXX take any action on the seek operation. 2216 */ 2217 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2218 goto out; 2219 2220 /* dofilereadv() will unuse the descriptor for us */ 2221 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2222 &offset, 0, retval)); 2223 2224 out: 2225 FILE_UNUSE(fp, l); 2226 return (error); 2227 } 2228 2229 /* 2230 * Positional write system call. 2231 */ 2232 int 2233 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2234 { 2235 struct sys_pwrite_args /* { 2236 syscallarg(int) fd; 2237 syscallarg(const void *) buf; 2238 syscallarg(size_t) nbyte; 2239 syscallarg(off_t) offset; 2240 } */ *uap = v; 2241 struct proc *p = l->l_proc; 2242 struct filedesc *fdp = p->p_fd; 2243 struct file *fp; 2244 struct vnode *vp; 2245 off_t offset; 2246 int error, fd = SCARG(uap, fd); 2247 2248 if ((fp = fd_getfile(fdp, fd)) == NULL) 2249 return (EBADF); 2250 2251 if ((fp->f_flag & FWRITE) == 0) { 2252 simple_unlock(&fp->f_slock); 2253 return (EBADF); 2254 } 2255 2256 FILE_USE(fp); 2257 2258 vp = (struct vnode *)fp->f_data; 2259 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2260 error = ESPIPE; 2261 goto out; 2262 } 2263 2264 offset = SCARG(uap, offset); 2265 2266 /* 2267 * XXX This works because no file systems actually 2268 * XXX take any action on the seek operation. 2269 */ 2270 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2271 goto out; 2272 2273 /* dofilewrite() will unuse the descriptor for us */ 2274 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2275 &offset, 0, retval)); 2276 2277 out: 2278 FILE_UNUSE(fp, l); 2279 return (error); 2280 } 2281 2282 /* 2283 * Positional gather write system call. 2284 */ 2285 int 2286 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2287 { 2288 struct sys_pwritev_args /* { 2289 syscallarg(int) fd; 2290 syscallarg(const struct iovec *) iovp; 2291 syscallarg(int) iovcnt; 2292 syscallarg(off_t) offset; 2293 } */ *uap = v; 2294 struct proc *p = l->l_proc; 2295 struct filedesc *fdp = p->p_fd; 2296 struct file *fp; 2297 struct vnode *vp; 2298 off_t offset; 2299 int error, fd = SCARG(uap, fd); 2300 2301 if ((fp = fd_getfile(fdp, fd)) == NULL) 2302 return (EBADF); 2303 2304 if ((fp->f_flag & FWRITE) == 0) { 2305 simple_unlock(&fp->f_slock); 2306 return (EBADF); 2307 } 2308 2309 FILE_USE(fp); 2310 2311 vp = (struct vnode *)fp->f_data; 2312 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2313 error = ESPIPE; 2314 goto out; 2315 } 2316 2317 offset = SCARG(uap, offset); 2318 2319 /* 2320 * XXX This works because no file systems actually 2321 * XXX take any action on the seek operation. 2322 */ 2323 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2324 goto out; 2325 2326 /* dofilewritev() will unuse the descriptor for us */ 2327 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2328 &offset, 0, retval)); 2329 2330 out: 2331 FILE_UNUSE(fp, l); 2332 return (error); 2333 } 2334 2335 /* 2336 * Check access permissions. 2337 */ 2338 int 2339 sys_access(struct lwp *l, void *v, register_t *retval) 2340 { 2341 struct sys_access_args /* { 2342 syscallarg(const char *) path; 2343 syscallarg(int) flags; 2344 } */ *uap = v; 2345 kauth_cred_t cred; 2346 struct vnode *vp; 2347 int error, flags; 2348 struct nameidata nd; 2349 2350 cred = kauth_cred_dup(l->l_cred); 2351 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2352 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2353 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2354 SCARG(uap, path), l); 2355 /* Override default credentials */ 2356 nd.ni_cnd.cn_cred = cred; 2357 if ((error = namei(&nd)) != 0) 2358 goto out; 2359 vp = nd.ni_vp; 2360 2361 /* Flags == 0 means only check for existence. */ 2362 if (SCARG(uap, flags)) { 2363 flags = 0; 2364 if (SCARG(uap, flags) & R_OK) 2365 flags |= VREAD; 2366 if (SCARG(uap, flags) & W_OK) 2367 flags |= VWRITE; 2368 if (SCARG(uap, flags) & X_OK) 2369 flags |= VEXEC; 2370 2371 error = VOP_ACCESS(vp, flags, cred, l); 2372 if (!error && (flags & VWRITE)) 2373 error = vn_writechk(vp); 2374 } 2375 vput(vp); 2376 out: 2377 kauth_cred_free(cred); 2378 return (error); 2379 } 2380 2381 /* 2382 * Common code for all sys_stat functions, including compat versions. 2383 */ 2384 int 2385 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2386 struct stat *sb) 2387 { 2388 int error; 2389 struct nameidata nd; 2390 2391 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2392 error = namei(&nd); 2393 if (error != 0) 2394 return error; 2395 error = vn_stat(nd.ni_vp, sb, l); 2396 vput(nd.ni_vp); 2397 return error; 2398 } 2399 2400 /* 2401 * Get file status; this version follows links. 2402 */ 2403 /* ARGSUSED */ 2404 int 2405 sys___stat30(struct lwp *l, void *v, register_t *retval) 2406 { 2407 struct sys___stat30_args /* { 2408 syscallarg(const char *) path; 2409 syscallarg(struct stat *) ub; 2410 } */ *uap = v; 2411 struct stat sb; 2412 int error; 2413 2414 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2415 if (error) 2416 return error; 2417 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2418 } 2419 2420 /* 2421 * Get file status; this version does not follow links. 2422 */ 2423 /* ARGSUSED */ 2424 int 2425 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2426 { 2427 struct sys___lstat30_args /* { 2428 syscallarg(const char *) path; 2429 syscallarg(struct stat *) ub; 2430 } */ *uap = v; 2431 struct stat sb; 2432 int error; 2433 2434 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2435 if (error) 2436 return error; 2437 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2438 } 2439 2440 /* 2441 * Get configurable pathname variables. 2442 */ 2443 /* ARGSUSED */ 2444 int 2445 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2446 { 2447 struct sys_pathconf_args /* { 2448 syscallarg(const char *) path; 2449 syscallarg(int) name; 2450 } */ *uap = v; 2451 int error; 2452 struct nameidata nd; 2453 2454 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2455 SCARG(uap, path), l); 2456 if ((error = namei(&nd)) != 0) 2457 return (error); 2458 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2459 vput(nd.ni_vp); 2460 return (error); 2461 } 2462 2463 /* 2464 * Return target name of a symbolic link. 2465 */ 2466 /* ARGSUSED */ 2467 int 2468 sys_readlink(struct lwp *l, void *v, register_t *retval) 2469 { 2470 struct sys_readlink_args /* { 2471 syscallarg(const char *) path; 2472 syscallarg(char *) buf; 2473 syscallarg(size_t) count; 2474 } */ *uap = v; 2475 struct vnode *vp; 2476 struct iovec aiov; 2477 struct uio auio; 2478 int error; 2479 struct nameidata nd; 2480 2481 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2482 SCARG(uap, path), l); 2483 if ((error = namei(&nd)) != 0) 2484 return (error); 2485 vp = nd.ni_vp; 2486 if (vp->v_type != VLNK) 2487 error = EINVAL; 2488 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2489 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2490 aiov.iov_base = SCARG(uap, buf); 2491 aiov.iov_len = SCARG(uap, count); 2492 auio.uio_iov = &aiov; 2493 auio.uio_iovcnt = 1; 2494 auio.uio_offset = 0; 2495 auio.uio_rw = UIO_READ; 2496 KASSERT(l == curlwp); 2497 auio.uio_vmspace = l->l_proc->p_vmspace; 2498 auio.uio_resid = SCARG(uap, count); 2499 error = VOP_READLINK(vp, &auio, l->l_cred); 2500 } 2501 vput(vp); 2502 *retval = SCARG(uap, count) - auio.uio_resid; 2503 return (error); 2504 } 2505 2506 /* 2507 * Change flags of a file given a path name. 2508 */ 2509 /* ARGSUSED */ 2510 int 2511 sys_chflags(struct lwp *l, void *v, register_t *retval) 2512 { 2513 struct sys_chflags_args /* { 2514 syscallarg(const char *) path; 2515 syscallarg(u_long) flags; 2516 } */ *uap = v; 2517 struct vnode *vp; 2518 int error; 2519 struct nameidata nd; 2520 2521 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2522 if ((error = namei(&nd)) != 0) 2523 return (error); 2524 vp = nd.ni_vp; 2525 error = change_flags(vp, SCARG(uap, flags), l); 2526 vput(vp); 2527 return (error); 2528 } 2529 2530 /* 2531 * Change flags of a file given a file descriptor. 2532 */ 2533 /* ARGSUSED */ 2534 int 2535 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2536 { 2537 struct sys_fchflags_args /* { 2538 syscallarg(int) fd; 2539 syscallarg(u_long) flags; 2540 } */ *uap = v; 2541 struct proc *p = l->l_proc; 2542 struct vnode *vp; 2543 struct file *fp; 2544 int error; 2545 2546 /* getvnode() will use the descriptor for us */ 2547 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2548 return (error); 2549 vp = (struct vnode *)fp->f_data; 2550 error = change_flags(vp, SCARG(uap, flags), l); 2551 VOP_UNLOCK(vp, 0); 2552 FILE_UNUSE(fp, l); 2553 return (error); 2554 } 2555 2556 /* 2557 * Change flags of a file given a path name; this version does 2558 * not follow links. 2559 */ 2560 int 2561 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2562 { 2563 struct sys_lchflags_args /* { 2564 syscallarg(const char *) path; 2565 syscallarg(u_long) flags; 2566 } */ *uap = v; 2567 struct vnode *vp; 2568 int error; 2569 struct nameidata nd; 2570 2571 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2572 if ((error = namei(&nd)) != 0) 2573 return (error); 2574 vp = nd.ni_vp; 2575 error = change_flags(vp, SCARG(uap, flags), l); 2576 vput(vp); 2577 return (error); 2578 } 2579 2580 /* 2581 * Common routine to change flags of a file. 2582 */ 2583 int 2584 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2585 { 2586 struct vattr vattr; 2587 int error; 2588 2589 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2590 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2591 /* 2592 * Non-superusers cannot change the flags on devices, even if they 2593 * own them. 2594 */ 2595 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2596 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2597 goto out; 2598 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2599 error = EINVAL; 2600 goto out; 2601 } 2602 } 2603 VATTR_NULL(&vattr); 2604 vattr.va_flags = flags; 2605 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2606 out: 2607 return (error); 2608 } 2609 2610 /* 2611 * Change mode of a file given path name; this version follows links. 2612 */ 2613 /* ARGSUSED */ 2614 int 2615 sys_chmod(struct lwp *l, void *v, register_t *retval) 2616 { 2617 struct sys_chmod_args /* { 2618 syscallarg(const char *) path; 2619 syscallarg(int) mode; 2620 } */ *uap = v; 2621 int error; 2622 struct nameidata nd; 2623 2624 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2625 if ((error = namei(&nd)) != 0) 2626 return (error); 2627 2628 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2629 2630 vrele(nd.ni_vp); 2631 return (error); 2632 } 2633 2634 /* 2635 * Change mode of a file given a file descriptor. 2636 */ 2637 /* ARGSUSED */ 2638 int 2639 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2640 { 2641 struct sys_fchmod_args /* { 2642 syscallarg(int) fd; 2643 syscallarg(int) mode; 2644 } */ *uap = v; 2645 struct proc *p = l->l_proc; 2646 struct file *fp; 2647 int error; 2648 2649 /* getvnode() will use the descriptor for us */ 2650 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2651 return (error); 2652 2653 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2654 FILE_UNUSE(fp, l); 2655 return (error); 2656 } 2657 2658 /* 2659 * Change mode of a file given path name; this version does not follow links. 2660 */ 2661 /* ARGSUSED */ 2662 int 2663 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2664 { 2665 struct sys_lchmod_args /* { 2666 syscallarg(const char *) path; 2667 syscallarg(int) mode; 2668 } */ *uap = v; 2669 int error; 2670 struct nameidata nd; 2671 2672 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2673 if ((error = namei(&nd)) != 0) 2674 return (error); 2675 2676 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2677 2678 vrele(nd.ni_vp); 2679 return (error); 2680 } 2681 2682 /* 2683 * Common routine to set mode given a vnode. 2684 */ 2685 static int 2686 change_mode(struct vnode *vp, int mode, struct lwp *l) 2687 { 2688 struct vattr vattr; 2689 int error; 2690 2691 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2692 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2693 VATTR_NULL(&vattr); 2694 vattr.va_mode = mode & ALLPERMS; 2695 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2696 VOP_UNLOCK(vp, 0); 2697 return (error); 2698 } 2699 2700 /* 2701 * Set ownership given a path name; this version follows links. 2702 */ 2703 /* ARGSUSED */ 2704 int 2705 sys_chown(struct lwp *l, void *v, register_t *retval) 2706 { 2707 struct sys_chown_args /* { 2708 syscallarg(const char *) path; 2709 syscallarg(uid_t) uid; 2710 syscallarg(gid_t) gid; 2711 } */ *uap = v; 2712 int error; 2713 struct nameidata nd; 2714 2715 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2716 if ((error = namei(&nd)) != 0) 2717 return (error); 2718 2719 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2720 2721 vrele(nd.ni_vp); 2722 return (error); 2723 } 2724 2725 /* 2726 * Set ownership given a path name; this version follows links. 2727 * Provides POSIX semantics. 2728 */ 2729 /* ARGSUSED */ 2730 int 2731 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2732 { 2733 struct sys_chown_args /* { 2734 syscallarg(const char *) path; 2735 syscallarg(uid_t) uid; 2736 syscallarg(gid_t) gid; 2737 } */ *uap = v; 2738 int error; 2739 struct nameidata nd; 2740 2741 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2742 if ((error = namei(&nd)) != 0) 2743 return (error); 2744 2745 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2746 2747 vrele(nd.ni_vp); 2748 return (error); 2749 } 2750 2751 /* 2752 * Set ownership given a file descriptor. 2753 */ 2754 /* ARGSUSED */ 2755 int 2756 sys_fchown(struct lwp *l, void *v, register_t *retval) 2757 { 2758 struct sys_fchown_args /* { 2759 syscallarg(int) fd; 2760 syscallarg(uid_t) uid; 2761 syscallarg(gid_t) gid; 2762 } */ *uap = v; 2763 struct proc *p = l->l_proc; 2764 int error; 2765 struct file *fp; 2766 2767 /* getvnode() will use the descriptor for us */ 2768 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2769 return (error); 2770 2771 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2772 SCARG(uap, gid), l, 0); 2773 FILE_UNUSE(fp, l); 2774 return (error); 2775 } 2776 2777 /* 2778 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2779 */ 2780 /* ARGSUSED */ 2781 int 2782 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2783 { 2784 struct sys_fchown_args /* { 2785 syscallarg(int) fd; 2786 syscallarg(uid_t) uid; 2787 syscallarg(gid_t) gid; 2788 } */ *uap = v; 2789 struct proc *p = l->l_proc; 2790 int error; 2791 struct file *fp; 2792 2793 /* getvnode() will use the descriptor for us */ 2794 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2795 return (error); 2796 2797 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2798 SCARG(uap, gid), l, 1); 2799 FILE_UNUSE(fp, l); 2800 return (error); 2801 } 2802 2803 /* 2804 * Set ownership given a path name; this version does not follow links. 2805 */ 2806 /* ARGSUSED */ 2807 int 2808 sys_lchown(struct lwp *l, void *v, register_t *retval) 2809 { 2810 struct sys_lchown_args /* { 2811 syscallarg(const char *) path; 2812 syscallarg(uid_t) uid; 2813 syscallarg(gid_t) gid; 2814 } */ *uap = v; 2815 int error; 2816 struct nameidata nd; 2817 2818 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2819 if ((error = namei(&nd)) != 0) 2820 return (error); 2821 2822 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2823 2824 vrele(nd.ni_vp); 2825 return (error); 2826 } 2827 2828 /* 2829 * Set ownership given a path name; this version does not follow links. 2830 * Provides POSIX/XPG semantics. 2831 */ 2832 /* ARGSUSED */ 2833 int 2834 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2835 { 2836 struct sys_lchown_args /* { 2837 syscallarg(const char *) path; 2838 syscallarg(uid_t) uid; 2839 syscallarg(gid_t) gid; 2840 } */ *uap = v; 2841 int error; 2842 struct nameidata nd; 2843 2844 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2845 if ((error = namei(&nd)) != 0) 2846 return (error); 2847 2848 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2849 2850 vrele(nd.ni_vp); 2851 return (error); 2852 } 2853 2854 /* 2855 * Common routine to set ownership given a vnode. 2856 */ 2857 static int 2858 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2859 int posix_semantics) 2860 { 2861 struct vattr vattr; 2862 mode_t newmode; 2863 int error; 2864 2865 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2866 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2867 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2868 goto out; 2869 2870 #define CHANGED(x) ((int)(x) != -1) 2871 newmode = vattr.va_mode; 2872 if (posix_semantics) { 2873 /* 2874 * POSIX/XPG semantics: if the caller is not the super-user, 2875 * clear set-user-id and set-group-id bits. Both POSIX and 2876 * the XPG consider the behaviour for calls by the super-user 2877 * implementation-defined; we leave the set-user-id and set- 2878 * group-id settings intact in that case. 2879 */ 2880 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2881 NULL) != 0) 2882 newmode &= ~(S_ISUID | S_ISGID); 2883 } else { 2884 /* 2885 * NetBSD semantics: when changing owner and/or group, 2886 * clear the respective bit(s). 2887 */ 2888 if (CHANGED(uid)) 2889 newmode &= ~S_ISUID; 2890 if (CHANGED(gid)) 2891 newmode &= ~S_ISGID; 2892 } 2893 /* Update va_mode iff altered. */ 2894 if (vattr.va_mode == newmode) 2895 newmode = VNOVAL; 2896 2897 VATTR_NULL(&vattr); 2898 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2899 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2900 vattr.va_mode = newmode; 2901 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2902 #undef CHANGED 2903 2904 out: 2905 VOP_UNLOCK(vp, 0); 2906 return (error); 2907 } 2908 2909 /* 2910 * Set the access and modification times given a path name; this 2911 * version follows links. 2912 */ 2913 /* ARGSUSED */ 2914 int 2915 sys_utimes(struct lwp *l, void *v, register_t *retval) 2916 { 2917 struct sys_utimes_args /* { 2918 syscallarg(const char *) path; 2919 syscallarg(const struct timeval *) tptr; 2920 } */ *uap = v; 2921 2922 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2923 SCARG(uap, tptr), UIO_USERSPACE); 2924 } 2925 2926 /* 2927 * Set the access and modification times given a file descriptor. 2928 */ 2929 /* ARGSUSED */ 2930 int 2931 sys_futimes(struct lwp *l, void *v, register_t *retval) 2932 { 2933 struct sys_futimes_args /* { 2934 syscallarg(int) fd; 2935 syscallarg(const struct timeval *) tptr; 2936 } */ *uap = v; 2937 int error; 2938 struct file *fp; 2939 2940 /* getvnode() will use the descriptor for us */ 2941 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2942 return (error); 2943 2944 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2945 SCARG(uap, tptr), UIO_USERSPACE); 2946 2947 FILE_UNUSE(fp, l); 2948 return (error); 2949 } 2950 2951 /* 2952 * Set the access and modification times given a path name; this 2953 * version does not follow links. 2954 */ 2955 int 2956 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2957 { 2958 struct sys_lutimes_args /* { 2959 syscallarg(const char *) path; 2960 syscallarg(const struct timeval *) tptr; 2961 } */ *uap = v; 2962 2963 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2964 SCARG(uap, tptr), UIO_USERSPACE); 2965 } 2966 2967 /* 2968 * Common routine to set access and modification times given a vnode. 2969 */ 2970 int 2971 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 2972 const struct timeval *tptr, enum uio_seg seg) 2973 { 2974 struct vattr vattr; 2975 struct nameidata nd; 2976 int error; 2977 2978 VATTR_NULL(&vattr); 2979 if (tptr == NULL) { 2980 nanotime(&vattr.va_atime); 2981 vattr.va_mtime = vattr.va_atime; 2982 vattr.va_vaflags |= VA_UTIMES_NULL; 2983 } else { 2984 struct timeval tv[2]; 2985 2986 if (seg != UIO_SYSSPACE) { 2987 error = copyin(tptr, &tv, sizeof (tv)); 2988 if (error != 0) 2989 return error; 2990 tptr = tv; 2991 } 2992 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 2993 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 2994 } 2995 2996 if (vp == NULL) { 2997 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 2998 if ((error = namei(&nd)) != 0) 2999 return (error); 3000 vp = nd.ni_vp; 3001 } else 3002 nd.ni_vp = NULL; 3003 3004 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3005 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3006 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3007 VOP_UNLOCK(vp, 0); 3008 3009 if (nd.ni_vp != NULL) 3010 vrele(nd.ni_vp); 3011 3012 return (error); 3013 } 3014 3015 /* 3016 * Truncate a file given its path name. 3017 */ 3018 /* ARGSUSED */ 3019 int 3020 sys_truncate(struct lwp *l, void *v, register_t *retval) 3021 { 3022 struct sys_truncate_args /* { 3023 syscallarg(const char *) path; 3024 syscallarg(int) pad; 3025 syscallarg(off_t) length; 3026 } */ *uap = v; 3027 struct vnode *vp; 3028 struct vattr vattr; 3029 int error; 3030 struct nameidata nd; 3031 3032 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3033 if ((error = namei(&nd)) != 0) 3034 return (error); 3035 vp = nd.ni_vp; 3036 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3037 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3038 if (vp->v_type == VDIR) 3039 error = EISDIR; 3040 else if ((error = vn_writechk(vp)) == 0 && 3041 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3042 VATTR_NULL(&vattr); 3043 vattr.va_size = SCARG(uap, length); 3044 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3045 } 3046 vput(vp); 3047 return (error); 3048 } 3049 3050 /* 3051 * Truncate a file given a file descriptor. 3052 */ 3053 /* ARGSUSED */ 3054 int 3055 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3056 { 3057 struct sys_ftruncate_args /* { 3058 syscallarg(int) fd; 3059 syscallarg(int) pad; 3060 syscallarg(off_t) length; 3061 } */ *uap = v; 3062 struct proc *p = l->l_proc; 3063 struct vattr vattr; 3064 struct vnode *vp; 3065 struct file *fp; 3066 int error; 3067 3068 /* getvnode() will use the descriptor for us */ 3069 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3070 return (error); 3071 if ((fp->f_flag & FWRITE) == 0) { 3072 error = EINVAL; 3073 goto out; 3074 } 3075 vp = (struct vnode *)fp->f_data; 3076 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3077 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3078 if (vp->v_type == VDIR) 3079 error = EISDIR; 3080 else if ((error = vn_writechk(vp)) == 0) { 3081 VATTR_NULL(&vattr); 3082 vattr.va_size = SCARG(uap, length); 3083 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3084 } 3085 VOP_UNLOCK(vp, 0); 3086 out: 3087 FILE_UNUSE(fp, l); 3088 return (error); 3089 } 3090 3091 /* 3092 * Sync an open file. 3093 */ 3094 /* ARGSUSED */ 3095 int 3096 sys_fsync(struct lwp *l, void *v, register_t *retval) 3097 { 3098 struct sys_fsync_args /* { 3099 syscallarg(int) fd; 3100 } */ *uap = v; 3101 struct proc *p = l->l_proc; 3102 struct vnode *vp; 3103 struct file *fp; 3104 int error; 3105 3106 /* getvnode() will use the descriptor for us */ 3107 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3108 return (error); 3109 vp = (struct vnode *)fp->f_data; 3110 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3111 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3112 if (error == 0 && bioops.io_fsync != NULL && 3113 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3114 (*bioops.io_fsync)(vp, 0); 3115 VOP_UNLOCK(vp, 0); 3116 FILE_UNUSE(fp, l); 3117 return (error); 3118 } 3119 3120 /* 3121 * Sync a range of file data. API modeled after that found in AIX. 3122 * 3123 * FDATASYNC indicates that we need only save enough metadata to be able 3124 * to re-read the written data. Note we duplicate AIX's requirement that 3125 * the file be open for writing. 3126 */ 3127 /* ARGSUSED */ 3128 int 3129 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3130 { 3131 struct sys_fsync_range_args /* { 3132 syscallarg(int) fd; 3133 syscallarg(int) flags; 3134 syscallarg(off_t) start; 3135 syscallarg(off_t) length; 3136 } */ *uap = v; 3137 struct proc *p = l->l_proc; 3138 struct vnode *vp; 3139 struct file *fp; 3140 int flags, nflags; 3141 off_t s, e, len; 3142 int error; 3143 3144 /* getvnode() will use the descriptor for us */ 3145 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3146 return (error); 3147 3148 if ((fp->f_flag & FWRITE) == 0) { 3149 error = EBADF; 3150 goto out; 3151 } 3152 3153 flags = SCARG(uap, flags); 3154 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3155 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3156 error = EINVAL; 3157 goto out; 3158 } 3159 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3160 if (flags & FDATASYNC) 3161 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3162 else 3163 nflags = FSYNC_WAIT; 3164 if (flags & FDISKSYNC) 3165 nflags |= FSYNC_CACHE; 3166 3167 len = SCARG(uap, length); 3168 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3169 if (len) { 3170 s = SCARG(uap, start); 3171 e = s + len; 3172 if (e < s) { 3173 error = EINVAL; 3174 goto out; 3175 } 3176 } else { 3177 e = 0; 3178 s = 0; 3179 } 3180 3181 vp = (struct vnode *)fp->f_data; 3182 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3183 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3184 3185 if (error == 0 && bioops.io_fsync != NULL && 3186 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3187 (*bioops.io_fsync)(vp, nflags); 3188 3189 VOP_UNLOCK(vp, 0); 3190 out: 3191 FILE_UNUSE(fp, l); 3192 return (error); 3193 } 3194 3195 /* 3196 * Sync the data of an open file. 3197 */ 3198 /* ARGSUSED */ 3199 int 3200 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3201 { 3202 struct sys_fdatasync_args /* { 3203 syscallarg(int) fd; 3204 } */ *uap = v; 3205 struct proc *p = l->l_proc; 3206 struct vnode *vp; 3207 struct file *fp; 3208 int error; 3209 3210 /* getvnode() will use the descriptor for us */ 3211 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3212 return (error); 3213 if ((fp->f_flag & FWRITE) == 0) { 3214 FILE_UNUSE(fp, l); 3215 return (EBADF); 3216 } 3217 vp = (struct vnode *)fp->f_data; 3218 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3219 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3220 VOP_UNLOCK(vp, 0); 3221 FILE_UNUSE(fp, l); 3222 return (error); 3223 } 3224 3225 /* 3226 * Rename files, (standard) BSD semantics frontend. 3227 */ 3228 /* ARGSUSED */ 3229 int 3230 sys_rename(struct lwp *l, void *v, register_t *retval) 3231 { 3232 struct sys_rename_args /* { 3233 syscallarg(const char *) from; 3234 syscallarg(const char *) to; 3235 } */ *uap = v; 3236 3237 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3238 } 3239 3240 /* 3241 * Rename files, POSIX semantics frontend. 3242 */ 3243 /* ARGSUSED */ 3244 int 3245 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3246 { 3247 struct sys___posix_rename_args /* { 3248 syscallarg(const char *) from; 3249 syscallarg(const char *) to; 3250 } */ *uap = v; 3251 3252 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3253 } 3254 3255 /* 3256 * Rename files. Source and destination must either both be directories, 3257 * or both not be directories. If target is a directory, it must be empty. 3258 * If `from' and `to' refer to the same object, the value of the `retain' 3259 * argument is used to determine whether `from' will be 3260 * 3261 * (retain == 0) deleted unless `from' and `to' refer to the same 3262 * object in the file system's name space (BSD). 3263 * (retain == 1) always retained (POSIX). 3264 */ 3265 static int 3266 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3267 { 3268 struct vnode *tvp, *fvp, *tdvp; 3269 struct nameidata fromnd, tond; 3270 struct proc *p; 3271 int error; 3272 3273 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3274 from, l); 3275 if ((error = namei(&fromnd)) != 0) 3276 return (error); 3277 if (fromnd.ni_dvp != fromnd.ni_vp) 3278 VOP_UNLOCK(fromnd.ni_dvp, 0); 3279 fvp = fromnd.ni_vp; 3280 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3281 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3282 if ((error = namei(&tond)) != 0) { 3283 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3284 vrele(fromnd.ni_dvp); 3285 vrele(fvp); 3286 goto out1; 3287 } 3288 tdvp = tond.ni_dvp; 3289 tvp = tond.ni_vp; 3290 3291 if (tvp != NULL) { 3292 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3293 error = ENOTDIR; 3294 goto out; 3295 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3296 error = EISDIR; 3297 goto out; 3298 } 3299 } 3300 3301 if (fvp == tdvp) 3302 error = EINVAL; 3303 3304 /* 3305 * Source and destination refer to the same object. 3306 */ 3307 if (fvp == tvp) { 3308 if (retain) 3309 error = -1; 3310 else if (fromnd.ni_dvp == tdvp && 3311 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3312 !memcmp(fromnd.ni_cnd.cn_nameptr, 3313 tond.ni_cnd.cn_nameptr, 3314 fromnd.ni_cnd.cn_namelen)) 3315 error = -1; 3316 } 3317 3318 #if NVERIEXEC > 0 3319 if (!error) { 3320 char *f1, *f2; 3321 3322 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3323 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3324 3325 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3326 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3327 3328 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3329 3330 free(f1, M_TEMP); 3331 free(f2, M_TEMP); 3332 } 3333 #endif /* NVERIEXEC > 0 */ 3334 3335 out: 3336 p = l->l_proc; 3337 if (!error) { 3338 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3339 if (fromnd.ni_dvp != tdvp) 3340 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3341 if (tvp) { 3342 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3343 } 3344 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3345 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3346 } else { 3347 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3348 if (tdvp == tvp) 3349 vrele(tdvp); 3350 else 3351 vput(tdvp); 3352 if (tvp) 3353 vput(tvp); 3354 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3355 vrele(fromnd.ni_dvp); 3356 vrele(fvp); 3357 } 3358 vrele(tond.ni_startdir); 3359 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3360 out1: 3361 if (fromnd.ni_startdir) 3362 vrele(fromnd.ni_startdir); 3363 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3364 return (error == -1 ? 0 : error); 3365 } 3366 3367 /* 3368 * Make a directory file. 3369 */ 3370 /* ARGSUSED */ 3371 int 3372 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3373 { 3374 struct sys_mkdir_args /* { 3375 syscallarg(const char *) path; 3376 syscallarg(int) mode; 3377 } */ *uap = v; 3378 struct proc *p = l->l_proc; 3379 struct vnode *vp; 3380 struct vattr vattr; 3381 int error; 3382 struct nameidata nd; 3383 3384 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3385 SCARG(uap, path), l); 3386 if ((error = namei(&nd)) != 0) 3387 return (error); 3388 vp = nd.ni_vp; 3389 if (vp != NULL) { 3390 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3391 if (nd.ni_dvp == vp) 3392 vrele(nd.ni_dvp); 3393 else 3394 vput(nd.ni_dvp); 3395 vrele(vp); 3396 return (EEXIST); 3397 } 3398 VATTR_NULL(&vattr); 3399 vattr.va_type = VDIR; 3400 vattr.va_mode = 3401 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3402 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3403 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3404 if (!error) 3405 vput(nd.ni_vp); 3406 return (error); 3407 } 3408 3409 /* 3410 * Remove a directory file. 3411 */ 3412 /* ARGSUSED */ 3413 int 3414 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3415 { 3416 struct sys_rmdir_args /* { 3417 syscallarg(const char *) path; 3418 } */ *uap = v; 3419 struct vnode *vp; 3420 int error; 3421 struct nameidata nd; 3422 3423 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3424 SCARG(uap, path), l); 3425 if ((error = namei(&nd)) != 0) 3426 return (error); 3427 vp = nd.ni_vp; 3428 if (vp->v_type != VDIR) { 3429 error = ENOTDIR; 3430 goto out; 3431 } 3432 /* 3433 * No rmdir "." please. 3434 */ 3435 if (nd.ni_dvp == vp) { 3436 error = EINVAL; 3437 goto out; 3438 } 3439 /* 3440 * The root of a mounted filesystem cannot be deleted. 3441 */ 3442 if (vp->v_flag & VROOT) { 3443 error = EBUSY; 3444 goto out; 3445 } 3446 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3447 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3448 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3449 return (error); 3450 3451 out: 3452 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3453 if (nd.ni_dvp == vp) 3454 vrele(nd.ni_dvp); 3455 else 3456 vput(nd.ni_dvp); 3457 vput(vp); 3458 return (error); 3459 } 3460 3461 /* 3462 * Read a block of directory entries in a file system independent format. 3463 */ 3464 int 3465 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3466 { 3467 struct sys___getdents30_args /* { 3468 syscallarg(int) fd; 3469 syscallarg(char *) buf; 3470 syscallarg(size_t) count; 3471 } */ *uap = v; 3472 struct proc *p = l->l_proc; 3473 struct file *fp; 3474 int error, done; 3475 3476 /* getvnode() will use the descriptor for us */ 3477 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3478 return (error); 3479 if ((fp->f_flag & FREAD) == 0) { 3480 error = EBADF; 3481 goto out; 3482 } 3483 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3484 SCARG(uap, count), &done, l, 0, 0); 3485 #ifdef KTRACE 3486 if (!error && KTRPOINT(p, KTR_GENIO)) { 3487 struct iovec iov; 3488 iov.iov_base = SCARG(uap, buf); 3489 iov.iov_len = done; 3490 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3491 } 3492 #endif 3493 *retval = done; 3494 out: 3495 FILE_UNUSE(fp, l); 3496 return (error); 3497 } 3498 3499 /* 3500 * Set the mode mask for creation of filesystem nodes. 3501 */ 3502 int 3503 sys_umask(struct lwp *l, void *v, register_t *retval) 3504 { 3505 struct sys_umask_args /* { 3506 syscallarg(mode_t) newmask; 3507 } */ *uap = v; 3508 struct proc *p = l->l_proc; 3509 struct cwdinfo *cwdi; 3510 3511 cwdi = p->p_cwdi; 3512 *retval = cwdi->cwdi_cmask; 3513 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3514 return (0); 3515 } 3516 3517 /* 3518 * Void all references to file by ripping underlying filesystem 3519 * away from vnode. 3520 */ 3521 /* ARGSUSED */ 3522 int 3523 sys_revoke(struct lwp *l, void *v, register_t *retval) 3524 { 3525 struct sys_revoke_args /* { 3526 syscallarg(const char *) path; 3527 } */ *uap = v; 3528 struct vnode *vp; 3529 struct vattr vattr; 3530 int error; 3531 struct nameidata nd; 3532 3533 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3534 if ((error = namei(&nd)) != 0) 3535 return (error); 3536 vp = nd.ni_vp; 3537 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3538 goto out; 3539 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3540 (error = kauth_authorize_generic(l->l_cred, 3541 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3542 goto out; 3543 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3544 VOP_REVOKE(vp, REVOKEALL); 3545 out: 3546 vrele(vp); 3547 return (error); 3548 } 3549 3550 /* 3551 * Convert a user file descriptor to a kernel file entry. 3552 */ 3553 int 3554 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3555 { 3556 struct vnode *vp; 3557 struct file *fp; 3558 3559 if ((fp = fd_getfile(fdp, fd)) == NULL) 3560 return (EBADF); 3561 3562 FILE_USE(fp); 3563 3564 if (fp->f_type != DTYPE_VNODE) { 3565 FILE_UNUSE(fp, NULL); 3566 return (EINVAL); 3567 } 3568 3569 vp = (struct vnode *)fp->f_data; 3570 if (vp->v_type == VBAD) { 3571 FILE_UNUSE(fp, NULL); 3572 return (EBADF); 3573 } 3574 3575 *fpp = fp; 3576 return (0); 3577 } 3578
Indexes created Thu Sep 25 02:09:54 GMT 2025