vfs_syscalls.c revision 1.311
1 /* $NetBSD: vfs_syscalls.c,v 1.311 2007/04/30 08:32:14 dsl Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.311 2007/04/30 08:32:14 dsl 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 #if NVERIEXEC > 0 73 #include <sys/verified_exec.h> 74 #include <sys/syslog.h> 75 #endif /* NVERIEXEC > 0 */ 76 #include <sys/kauth.h> 77 78 #include <miscfs/genfs/genfs.h> 79 #include <miscfs/syncfs/syncfs.h> 80 81 #ifdef COMPAT_30 82 #include "opt_nfsserver.h" 83 #include <nfs/rpcv2.h> 84 #endif 85 #include <nfs/nfsproto.h> 86 #ifdef COMPAT_30 87 #include <nfs/nfs.h> 88 #include <nfs/nfs_var.h> 89 #endif 90 91 #if NFSS > 0 92 #include <dev/fssvar.h> 93 #endif 94 95 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 96 97 static int change_dir(struct nameidata *, struct lwp *); 98 static int change_flags(struct vnode *, u_long, struct lwp *); 99 static int change_mode(struct vnode *, int, struct lwp *l); 100 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 101 static int change_utimes(struct vnode *vp, const struct timeval *, 102 struct lwp *l); 103 static int rename_files(const char *, const char *, struct lwp *, int); 104 105 void checkdirs(struct vnode *); 106 107 static int mount_update(struct lwp *, struct vnode *, const char *, int, 108 void *, struct nameidata *); 109 static int mount_domount(struct lwp *, struct vnode *, const char *, 110 const char *, int, void *, struct nameidata *); 111 static int mount_getargs(struct lwp *, struct vnode *, const char *, int, 112 void *, struct nameidata *); 113 114 int dovfsusermount = 0; 115 116 /* 117 * Virtual File System System Calls 118 */ 119 120 /* 121 * Mount a file system. 122 */ 123 124 #if defined(COMPAT_09) || defined(COMPAT_43) 125 /* 126 * This table is used to maintain compatibility with 4.3BSD 127 * and NetBSD 0.9 mount syscalls. Note, the order is important! 128 * 129 * Do not modify this table. It should only contain filesystems 130 * supported by NetBSD 0.9 and 4.3BSD. 131 */ 132 const char * const mountcompatnames[] = { 133 NULL, /* 0 = MOUNT_NONE */ 134 MOUNT_FFS, /* 1 = MOUNT_UFS */ 135 MOUNT_NFS, /* 2 */ 136 MOUNT_MFS, /* 3 */ 137 MOUNT_MSDOS, /* 4 */ 138 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 139 MOUNT_FDESC, /* 6 */ 140 MOUNT_KERNFS, /* 7 */ 141 NULL, /* 8 = MOUNT_DEVFS */ 142 MOUNT_AFS, /* 9 */ 143 }; 144 const int nmountcompatnames = sizeof(mountcompatnames) / 145 sizeof(mountcompatnames[0]); 146 #endif /* COMPAT_09 || COMPAT_43 */ 147 148 static int 149 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 150 void *data, struct nameidata *ndp) 151 { 152 struct mount *mp; 153 int error = 0, saved_flags; 154 155 mp = vp->v_mount; 156 saved_flags = mp->mnt_flag; 157 158 /* We can operate only on VROOT nodes. */ 159 if ((vp->v_flag & VROOT) == 0) { 160 error = EINVAL; 161 goto out; 162 } 163 164 /* 165 * We only allow the filesystem to be reloaded if it 166 * is currently mounted read-only. 167 */ 168 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 169 error = EOPNOTSUPP; /* Needs translation */ 170 goto out; 171 } 172 173 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 174 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 175 if (error) 176 goto out; 177 178 if (vfs_busy(mp, LK_NOWAIT, 0)) { 179 error = EPERM; 180 goto out; 181 } 182 183 mp->mnt_flag &= ~MNT_OP_FLAGS; 184 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 185 186 /* 187 * Set the mount level flags. 188 */ 189 if (flags & MNT_RDONLY) 190 mp->mnt_flag |= MNT_RDONLY; 191 else if (mp->mnt_flag & MNT_RDONLY) 192 mp->mnt_iflag |= IMNT_WANTRDWR; 193 mp->mnt_flag &= 194 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 195 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 196 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 197 mp->mnt_flag |= flags & 198 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 199 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 200 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 201 MNT_IGNORE); 202 203 error = VFS_MOUNT(mp, path, data, ndp, l); 204 205 #if defined(COMPAT_30) && defined(NFSSERVER) 206 if (error) { 207 int error2; 208 209 /* Update failed; let's try and see if it was an 210 * export request. */ 211 error2 = nfs_update_exports_30(mp, path, data, l); 212 213 /* Only update error code if the export request was 214 * understood but some problem occurred while 215 * processing it. */ 216 if (error2 != EJUSTRETURN) 217 error = error2; 218 } 219 #endif 220 if (mp->mnt_iflag & IMNT_WANTRDWR) 221 mp->mnt_flag &= ~MNT_RDONLY; 222 if (error) 223 mp->mnt_flag = saved_flags; 224 mp->mnt_flag &= ~MNT_OP_FLAGS; 225 mp->mnt_iflag &= ~IMNT_WANTRDWR; 226 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 227 if (mp->mnt_syncer == NULL) 228 error = vfs_allocate_syncvnode(mp); 229 } else { 230 if (mp->mnt_syncer != NULL) 231 vfs_deallocate_syncvnode(mp); 232 } 233 vfs_unbusy(mp); 234 235 out: 236 return (error); 237 } 238 239 static int 240 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 241 const char *path, int flags, void *data, struct nameidata *ndp) 242 { 243 struct mount *mp = NULL; 244 struct vattr va; 245 char fstypename[MFSNAMELEN]; 246 int error; 247 248 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 249 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 250 if (error) { 251 vput(vp); 252 goto out; 253 } 254 255 /* Can't make a non-dir a mount-point (from here anyway). */ 256 if (vp->v_type != VDIR) { 257 error = ENOTDIR; 258 vput(vp); 259 goto out; 260 } 261 262 /* 263 * If the user is not root, ensure that they own the directory 264 * onto which we are attempting to mount. 265 */ 266 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 267 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 268 (error = kauth_authorize_generic(l->l_cred, 269 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 270 vput(vp); 271 goto out; 272 } 273 274 if (flags & MNT_EXPORTED) { 275 error = EINVAL; 276 vput(vp); 277 goto out; 278 } 279 280 /* 281 * Copy file-system type from userspace. 282 */ 283 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 284 if (error) { 285 #if defined(COMPAT_09) || defined(COMPAT_43) 286 /* 287 * Historically, filesystem types were identified by numbers. 288 * If we get an integer for the filesystem type instead of a 289 * string, we check to see if it matches one of the historic 290 * filesystem types. 291 */ 292 u_long fsindex = (u_long)fstype; 293 if (fsindex >= nmountcompatnames || 294 mountcompatnames[fsindex] == NULL) { 295 error = ENODEV; 296 vput(vp); 297 goto out; 298 } 299 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 300 #else 301 vput(vp); 302 goto out; 303 #endif 304 } 305 306 #ifdef COMPAT_10 307 /* Accept `ufs' as an alias for `ffs'. */ 308 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 309 strlcpy(fstypename, "ffs", sizeof(fstypename)); 310 #endif 311 312 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 313 vput(vp); 314 goto out; 315 } 316 317 /* 318 * Check if a file-system is not already mounted on this vnode. 319 */ 320 if (vp->v_mountedhere != NULL) { 321 error = EBUSY; 322 vput(vp); 323 goto out; 324 } 325 326 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 327 328 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 329 free(mp, M_MOUNT); 330 error = ENODEV; 331 vput(vp); 332 goto out; 333 } 334 335 TAILQ_INIT(&mp->mnt_vnodelist); 336 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 337 simple_lock_init(&mp->mnt_slock); 338 (void)vfs_busy(mp, LK_NOWAIT, 0); 339 340 mp->mnt_op->vfs_refcount++; 341 mp->mnt_vnodecovered = vp; 342 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 343 mp->mnt_unmounter = NULL; 344 mount_initspecific(mp); 345 346 /* 347 * The underlying file system may refuse the mount for 348 * various reasons. Allow the user to force it to happen. 349 * 350 * Set the mount level flags. 351 */ 352 mp->mnt_flag = flags & 353 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 354 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 355 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 356 MNT_IGNORE | MNT_RDONLY); 357 358 error = VFS_MOUNT(mp, path, data, ndp, l); 359 mp->mnt_flag &= ~MNT_OP_FLAGS; 360 361 /* 362 * Put the new filesystem on the mount list after root. 363 */ 364 cache_purge(vp); 365 if (!error) { 366 mp->mnt_iflag &= ~IMNT_WANTRDWR; 367 vp->v_mountedhere = mp; 368 simple_lock(&mountlist_slock); 369 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 370 simple_unlock(&mountlist_slock); 371 VOP_UNLOCK(vp, 0); 372 checkdirs(vp); 373 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 374 error = vfs_allocate_syncvnode(mp); 375 vfs_unbusy(mp); 376 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 377 error = VFS_START(mp, 0, l); 378 if (error) 379 vrele(vp); 380 } else { 381 vp->v_mountedhere = NULL; 382 mp->mnt_op->vfs_refcount--; 383 vfs_unbusy(mp); 384 free(mp, M_MOUNT); 385 vput(vp); 386 } 387 388 out: 389 return (error); 390 } 391 392 static int 393 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 394 void *data, struct nameidata *ndp) 395 { 396 struct mount *mp; 397 int error; 398 399 /* If MNT_GETARGS is specified, it should be the only flag. */ 400 if (flags & ~MNT_GETARGS) { 401 error = EINVAL; 402 goto out; 403 } 404 405 mp = vp->v_mount; 406 407 /* XXX: probably some notion of "can see" here if we want isolation. */ 408 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 409 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 410 if (error) 411 goto out; 412 413 if ((vp->v_flag & VROOT) == 0) { 414 error = EINVAL; 415 goto out; 416 } 417 418 if (vfs_busy(mp, LK_NOWAIT, 0)) { 419 error = EPERM; 420 goto out; 421 } 422 423 mp->mnt_flag &= ~MNT_OP_FLAGS; 424 mp->mnt_flag |= MNT_GETARGS; 425 error = VFS_MOUNT(mp, path, data, ndp, l); 426 mp->mnt_flag &= ~MNT_OP_FLAGS; 427 428 vfs_unbusy(mp); 429 out: 430 return (error); 431 } 432 433 /* ARGSUSED */ 434 int 435 sys_mount(struct lwp *l, void *v, register_t *retval) 436 { 437 struct sys_mount_args /* { 438 syscallarg(const char *) type; 439 syscallarg(const char *) path; 440 syscallarg(int) flags; 441 syscallarg(void *) data; 442 } */ *uap = v; 443 struct vnode *vp; 444 struct nameidata nd; 445 int error; 446 447 /* 448 * Get vnode to be covered 449 */ 450 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 451 SCARG(uap, path), l); 452 if ((error = namei(&nd)) != 0) 453 return (error); 454 vp = nd.ni_vp; 455 456 /* 457 * A lookup in VFS_MOUNT might result in an attempt to 458 * lock this vnode again, so make the lock recursive. 459 */ 460 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 461 462 if (SCARG(uap, flags) & MNT_GETARGS) { 463 error = mount_getargs(l, vp, SCARG(uap, path), 464 SCARG(uap, flags), SCARG(uap, data), &nd); 465 vput(vp); 466 } else if (SCARG(uap, flags) & MNT_UPDATE) { 467 error = mount_update(l, vp, SCARG(uap, path), 468 SCARG(uap, flags), SCARG(uap, data), &nd); 469 vput(vp); 470 } else { 471 /* Locking is handled internally in mount_domount(). */ 472 error = mount_domount(l, vp, SCARG(uap, type), 473 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 474 } 475 476 return (error); 477 } 478 479 /* 480 * Scan all active processes to see if any of them have a current 481 * or root directory onto which the new filesystem has just been 482 * mounted. If so, replace them with the new mount point. 483 */ 484 void 485 checkdirs(struct vnode *olddp) 486 { 487 struct cwdinfo *cwdi; 488 struct vnode *newdp; 489 struct proc *p; 490 491 if (olddp->v_usecount == 1) 492 return; 493 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 494 panic("mount: lost mount"); 495 mutex_enter(&proclist_lock); 496 PROCLIST_FOREACH(p, &allproc) { 497 cwdi = p->p_cwdi; 498 if (!cwdi) 499 continue; 500 if (cwdi->cwdi_cdir == olddp) { 501 vrele(cwdi->cwdi_cdir); 502 VREF(newdp); 503 cwdi->cwdi_cdir = newdp; 504 } 505 if (cwdi->cwdi_rdir == olddp) { 506 vrele(cwdi->cwdi_rdir); 507 VREF(newdp); 508 cwdi->cwdi_rdir = newdp; 509 } 510 } 511 mutex_exit(&proclist_lock); 512 if (rootvnode == olddp) { 513 vrele(rootvnode); 514 VREF(newdp); 515 rootvnode = newdp; 516 } 517 vput(newdp); 518 } 519 520 /* 521 * Unmount a file system. 522 * 523 * Note: unmount takes a path to the vnode mounted on as argument, 524 * not special file (as before). 525 */ 526 /* ARGSUSED */ 527 int 528 sys_unmount(struct lwp *l, void *v, register_t *retval) 529 { 530 struct sys_unmount_args /* { 531 syscallarg(const char *) path; 532 syscallarg(int) flags; 533 } */ *uap = v; 534 struct vnode *vp; 535 struct mount *mp; 536 int error; 537 struct nameidata nd; 538 539 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 540 SCARG(uap, path), l); 541 if ((error = namei(&nd)) != 0) 542 return (error); 543 vp = nd.ni_vp; 544 mp = vp->v_mount; 545 546 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 547 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 548 if (error) { 549 vput(vp); 550 return (error); 551 } 552 553 /* 554 * Don't allow unmounting the root file system. 555 */ 556 if (mp->mnt_flag & MNT_ROOTFS) { 557 vput(vp); 558 return (EINVAL); 559 } 560 561 /* 562 * Must be the root of the filesystem 563 */ 564 if ((vp->v_flag & VROOT) == 0) { 565 vput(vp); 566 return (EINVAL); 567 } 568 vput(vp); 569 570 /* 571 * XXX Freeze syncer. Must do this before locking the 572 * mount point. See dounmount() for details. 573 */ 574 mutex_enter(&syncer_mutex); 575 576 if (vfs_busy(mp, 0, 0)) { 577 mutex_exit(&syncer_mutex); 578 return (EBUSY); 579 } 580 581 return (dounmount(mp, SCARG(uap, flags), l)); 582 } 583 584 /* 585 * Do the actual file system unmount. File system is assumed to have been 586 * marked busy by the caller. 587 */ 588 int 589 dounmount(struct mount *mp, int flags, struct lwp *l) 590 { 591 struct vnode *coveredvp; 592 int error; 593 int async; 594 int used_syncer; 595 596 #if NVERIEXEC > 0 597 error = veriexec_unmountchk(mp); 598 if (error) 599 return (error); 600 #endif /* NVERIEXEC > 0 */ 601 602 simple_lock(&mountlist_slock); 603 vfs_unbusy(mp); 604 used_syncer = (mp->mnt_syncer != NULL); 605 606 /* 607 * XXX Syncer must be frozen when we get here. This should really 608 * be done on a per-mountpoint basis, but especially the softdep 609 * code possibly called from the syncer doesn't exactly work on a 610 * per-mountpoint basis, so the softdep code would become a maze 611 * of vfs_busy() calls. 612 * 613 * The caller of dounmount() must acquire syncer_mutex because 614 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 615 * order, and we must preserve that order to avoid deadlock. 616 * 617 * So, if the file system did not use the syncer, now is 618 * the time to release the syncer_mutex. 619 */ 620 if (used_syncer == 0) 621 mutex_exit(&syncer_mutex); 622 623 mp->mnt_iflag |= IMNT_UNMOUNT; 624 mp->mnt_unmounter = l; 625 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 626 627 async = mp->mnt_flag & MNT_ASYNC; 628 mp->mnt_flag &= ~MNT_ASYNC; 629 cache_purgevfs(mp); /* remove cache entries for this file sys */ 630 if (mp->mnt_syncer != NULL) 631 vfs_deallocate_syncvnode(mp); 632 error = 0; 633 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 634 #if NFSS > 0 635 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 636 #endif 637 if (error == 0) 638 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 639 } 640 if (error == 0 || (flags & MNT_FORCE)) 641 error = VFS_UNMOUNT(mp, flags, l); 642 simple_lock(&mountlist_slock); 643 if (error) { 644 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 645 (void) vfs_allocate_syncvnode(mp); 646 mp->mnt_iflag &= ~IMNT_UNMOUNT; 647 mp->mnt_unmounter = NULL; 648 mp->mnt_flag |= async; 649 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 650 &mountlist_slock); 651 if (used_syncer) 652 mutex_exit(&syncer_mutex); 653 simple_lock(&mp->mnt_slock); 654 while (mp->mnt_wcnt > 0) { 655 wakeup(mp); 656 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 657 0, &mp->mnt_slock); 658 } 659 simple_unlock(&mp->mnt_slock); 660 return (error); 661 } 662 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 663 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 664 coveredvp->v_mountedhere = NULL; 665 mp->mnt_op->vfs_refcount--; 666 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 667 panic("unmount: dangling vnode"); 668 mp->mnt_iflag |= IMNT_GONE; 669 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 670 if (coveredvp != NULLVP) 671 vrele(coveredvp); 672 mount_finispecific(mp); 673 if (used_syncer) 674 mutex_exit(&syncer_mutex); 675 simple_lock(&mp->mnt_slock); 676 while (mp->mnt_wcnt > 0) { 677 wakeup(mp); 678 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 679 } 680 simple_unlock(&mp->mnt_slock); 681 vfs_hooks_unmount(mp); 682 free(mp, M_MOUNT); 683 return (0); 684 } 685 686 /* 687 * Sync each mounted filesystem. 688 */ 689 #ifdef DEBUG 690 int syncprt = 0; 691 struct ctldebug debug0 = { "syncprt", &syncprt }; 692 #endif 693 694 /* ARGSUSED */ 695 int 696 sys_sync(struct lwp *l, void *v, register_t *retval) 697 { 698 struct mount *mp, *nmp; 699 int asyncflag; 700 701 if (l == NULL) 702 l = &lwp0; 703 704 simple_lock(&mountlist_slock); 705 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 706 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 707 nmp = mp->mnt_list.cqe_prev; 708 continue; 709 } 710 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 711 asyncflag = mp->mnt_flag & MNT_ASYNC; 712 mp->mnt_flag &= ~MNT_ASYNC; 713 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 714 if (asyncflag) 715 mp->mnt_flag |= MNT_ASYNC; 716 } 717 simple_lock(&mountlist_slock); 718 nmp = mp->mnt_list.cqe_prev; 719 vfs_unbusy(mp); 720 721 } 722 simple_unlock(&mountlist_slock); 723 #ifdef DEBUG 724 if (syncprt) 725 vfs_bufstats(); 726 #endif /* DEBUG */ 727 return (0); 728 } 729 730 /* 731 * Change filesystem quotas. 732 */ 733 /* ARGSUSED */ 734 int 735 sys_quotactl(struct lwp *l, void *v, register_t *retval) 736 { 737 struct sys_quotactl_args /* { 738 syscallarg(const char *) path; 739 syscallarg(int) cmd; 740 syscallarg(int) uid; 741 syscallarg(void *) arg; 742 } */ *uap = v; 743 struct mount *mp; 744 int error; 745 struct nameidata nd; 746 747 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 748 if ((error = namei(&nd)) != 0) 749 return (error); 750 mp = nd.ni_vp->v_mount; 751 vrele(nd.ni_vp); 752 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 753 SCARG(uap, arg), l); 754 return (error); 755 } 756 757 int 758 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 759 int root) 760 { 761 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 762 int error = 0; 763 764 /* 765 * If MNT_NOWAIT or MNT_LAZY is specified, do not 766 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 767 * overrides MNT_NOWAIT. 768 */ 769 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 770 (flags != MNT_WAIT && flags != 0)) { 771 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 772 goto done; 773 } 774 775 /* Get the filesystem stats now */ 776 memset(sp, 0, sizeof(*sp)); 777 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 778 return error; 779 } 780 781 if (cwdi->cwdi_rdir == NULL) 782 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 783 done: 784 if (cwdi->cwdi_rdir != NULL) { 785 size_t len; 786 char *bp; 787 char *path = PNBUF_GET(); 788 if (!path) 789 return ENOMEM; 790 791 bp = path + MAXPATHLEN; 792 *--bp = '\0'; 793 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 794 MAXPATHLEN / 2, 0, l); 795 if (error) { 796 PNBUF_PUT(path); 797 return error; 798 } 799 len = strlen(bp); 800 /* 801 * for mount points that are below our root, we can see 802 * them, so we fix up the pathname and return them. The 803 * rest we cannot see, so we don't allow viewing the 804 * data. 805 */ 806 if (strncmp(bp, sp->f_mntonname, len) == 0) { 807 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 808 sizeof(sp->f_mntonname)); 809 if (sp->f_mntonname[0] == '\0') 810 (void)strlcpy(sp->f_mntonname, "/", 811 sizeof(sp->f_mntonname)); 812 } else { 813 if (root) 814 (void)strlcpy(sp->f_mntonname, "/", 815 sizeof(sp->f_mntonname)); 816 else 817 error = EPERM; 818 } 819 PNBUF_PUT(path); 820 } 821 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 822 return error; 823 } 824 825 /* 826 * Get filesystem statistics by path. 827 */ 828 int 829 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 830 { 831 struct mount *mp; 832 int error; 833 struct nameidata nd; 834 835 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 836 if ((error = namei(&nd)) != 0) 837 return error; 838 mp = nd.ni_vp->v_mount; 839 error = dostatvfs(mp, sb, l, flags, 1); 840 vrele(nd.ni_vp); 841 return error; 842 } 843 844 /* ARGSUSED */ 845 int 846 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 847 { 848 struct sys_statvfs1_args /* { 849 syscallarg(const char *) path; 850 syscallarg(struct statvfs *) buf; 851 syscallarg(int) flags; 852 } */ *uap = v; 853 struct statvfs *sb; 854 int error; 855 856 sb = STATVFSBUF_GET(); 857 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 858 if (error == 0) 859 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 860 STATVFSBUF_PUT(sb); 861 return error; 862 } 863 864 /* 865 * Get filesystem statistics by fd. 866 */ 867 int 868 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 869 { 870 struct proc *p = l->l_proc; 871 struct file *fp; 872 struct mount *mp; 873 int error; 874 875 /* getvnode() will use the descriptor for us */ 876 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 877 return (error); 878 mp = ((struct vnode *)fp->f_data)->v_mount; 879 error = dostatvfs(mp, sb, l, flags, 1); 880 FILE_UNUSE(fp, l); 881 return error; 882 } 883 884 /* ARGSUSED */ 885 int 886 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 887 { 888 struct sys_fstatvfs1_args /* { 889 syscallarg(int) fd; 890 syscallarg(struct statvfs *) buf; 891 syscallarg(int) flags; 892 } */ *uap = v; 893 struct statvfs *sb; 894 int error; 895 896 sb = STATVFSBUF_GET(); 897 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 898 if (error != 0) 899 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 900 STATVFSBUF_PUT(sb); 901 return error; 902 } 903 904 905 /* 906 * Get statistics on all filesystems. 907 */ 908 int 909 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 910 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 911 register_t *retval) 912 { 913 int root = 0; 914 struct proc *p = l->l_proc; 915 struct mount *mp, *nmp; 916 struct statvfs *sb; 917 size_t count, maxcount; 918 int error = 0; 919 920 sb = STATVFSBUF_GET(); 921 maxcount = bufsize / entry_sz; 922 simple_lock(&mountlist_slock); 923 count = 0; 924 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 925 mp = nmp) { 926 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 927 nmp = CIRCLEQ_NEXT(mp, mnt_list); 928 continue; 929 } 930 if (sfsp && count < maxcount) { 931 error = dostatvfs(mp, sb, l, flags, 0); 932 if (error) { 933 simple_lock(&mountlist_slock); 934 nmp = CIRCLEQ_NEXT(mp, mnt_list); 935 vfs_unbusy(mp); 936 continue; 937 } 938 error = copyfn(sb, sfsp, entry_sz); 939 if (error) { 940 vfs_unbusy(mp); 941 goto out; 942 } 943 sfsp = (char *)sfsp + entry_sz; 944 root |= strcmp(sb->f_mntonname, "/") == 0; 945 } 946 count++; 947 simple_lock(&mountlist_slock); 948 nmp = CIRCLEQ_NEXT(mp, mnt_list); 949 vfs_unbusy(mp); 950 } 951 simple_unlock(&mountlist_slock); 952 if (root == 0 && p->p_cwdi->cwdi_rdir) { 953 /* 954 * fake a root entry 955 */ 956 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 957 if (error != 0) 958 goto out; 959 if (sfsp) 960 error = copyfn(sb, sfsp, entry_sz); 961 count++; 962 } 963 if (sfsp && count > maxcount) 964 *retval = maxcount; 965 else 966 *retval = count; 967 out: 968 STATVFSBUF_PUT(sb); 969 return error; 970 } 971 972 int 973 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 974 { 975 struct sys_getvfsstat_args /* { 976 syscallarg(struct statvfs *) buf; 977 syscallarg(size_t) bufsize; 978 syscallarg(int) flags; 979 } */ *uap = v; 980 981 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 982 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 983 } 984 985 /* 986 * Change current working directory to a given file descriptor. 987 */ 988 /* ARGSUSED */ 989 int 990 sys_fchdir(struct lwp *l, void *v, register_t *retval) 991 { 992 struct sys_fchdir_args /* { 993 syscallarg(int) fd; 994 } */ *uap = v; 995 struct proc *p = l->l_proc; 996 struct filedesc *fdp = p->p_fd; 997 struct cwdinfo *cwdi = p->p_cwdi; 998 struct vnode *vp, *tdp; 999 struct mount *mp; 1000 struct file *fp; 1001 int error; 1002 1003 /* getvnode() will use the descriptor for us */ 1004 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1005 return (error); 1006 vp = (struct vnode *)fp->f_data; 1007 1008 VREF(vp); 1009 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1010 if (vp->v_type != VDIR) 1011 error = ENOTDIR; 1012 else 1013 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1014 if (error) { 1015 vput(vp); 1016 goto out; 1017 } 1018 while ((mp = vp->v_mountedhere) != NULL) { 1019 if (vfs_busy(mp, 0, 0)) 1020 continue; 1021 1022 vput(vp); 1023 error = VFS_ROOT(mp, &tdp); 1024 vfs_unbusy(mp); 1025 if (error) 1026 goto out; 1027 vp = tdp; 1028 } 1029 VOP_UNLOCK(vp, 0); 1030 1031 /* 1032 * Disallow changing to a directory not under the process's 1033 * current root directory (if there is one). 1034 */ 1035 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1036 vrele(vp); 1037 error = EPERM; /* operation not permitted */ 1038 goto out; 1039 } 1040 1041 vrele(cwdi->cwdi_cdir); 1042 cwdi->cwdi_cdir = vp; 1043 out: 1044 FILE_UNUSE(fp, l); 1045 return (error); 1046 } 1047 1048 /* 1049 * Change this process's notion of the root directory to a given file 1050 * descriptor. 1051 */ 1052 int 1053 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1054 { 1055 struct sys_fchroot_args *uap = v; 1056 struct proc *p = l->l_proc; 1057 struct filedesc *fdp = p->p_fd; 1058 struct cwdinfo *cwdi = p->p_cwdi; 1059 struct vnode *vp; 1060 struct file *fp; 1061 int error; 1062 1063 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1064 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1065 return error; 1066 /* getvnode() will use the descriptor for us */ 1067 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1068 return error; 1069 vp = (struct vnode *) fp->f_data; 1070 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1071 if (vp->v_type != VDIR) 1072 error = ENOTDIR; 1073 else 1074 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1075 VOP_UNLOCK(vp, 0); 1076 if (error) 1077 goto out; 1078 VREF(vp); 1079 1080 /* 1081 * Prevent escaping from chroot by putting the root under 1082 * the working directory. Silently chdir to / if we aren't 1083 * already there. 1084 */ 1085 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1086 /* 1087 * XXX would be more failsafe to change directory to a 1088 * deadfs node here instead 1089 */ 1090 vrele(cwdi->cwdi_cdir); 1091 VREF(vp); 1092 cwdi->cwdi_cdir = vp; 1093 } 1094 1095 if (cwdi->cwdi_rdir != NULL) 1096 vrele(cwdi->cwdi_rdir); 1097 cwdi->cwdi_rdir = vp; 1098 out: 1099 FILE_UNUSE(fp, l); 1100 return (error); 1101 } 1102 1103 /* 1104 * Change current working directory (``.''). 1105 */ 1106 /* ARGSUSED */ 1107 int 1108 sys_chdir(struct lwp *l, void *v, register_t *retval) 1109 { 1110 struct sys_chdir_args /* { 1111 syscallarg(const char *) path; 1112 } */ *uap = v; 1113 struct proc *p = l->l_proc; 1114 struct cwdinfo *cwdi = p->p_cwdi; 1115 int error; 1116 struct nameidata nd; 1117 1118 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1119 SCARG(uap, path), l); 1120 if ((error = change_dir(&nd, l)) != 0) 1121 return (error); 1122 vrele(cwdi->cwdi_cdir); 1123 cwdi->cwdi_cdir = nd.ni_vp; 1124 return (0); 1125 } 1126 1127 /* 1128 * Change notion of root (``/'') directory. 1129 */ 1130 /* ARGSUSED */ 1131 int 1132 sys_chroot(struct lwp *l, void *v, register_t *retval) 1133 { 1134 struct sys_chroot_args /* { 1135 syscallarg(const char *) path; 1136 } */ *uap = v; 1137 struct proc *p = l->l_proc; 1138 struct cwdinfo *cwdi = p->p_cwdi; 1139 struct vnode *vp; 1140 int error; 1141 struct nameidata nd; 1142 1143 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1144 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1145 return (error); 1146 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1147 SCARG(uap, path), l); 1148 if ((error = change_dir(&nd, l)) != 0) 1149 return (error); 1150 if (cwdi->cwdi_rdir != NULL) 1151 vrele(cwdi->cwdi_rdir); 1152 vp = nd.ni_vp; 1153 cwdi->cwdi_rdir = vp; 1154 1155 /* 1156 * Prevent escaping from chroot by putting the root under 1157 * the working directory. Silently chdir to / if we aren't 1158 * already there. 1159 */ 1160 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1161 /* 1162 * XXX would be more failsafe to change directory to a 1163 * deadfs node here instead 1164 */ 1165 vrele(cwdi->cwdi_cdir); 1166 VREF(vp); 1167 cwdi->cwdi_cdir = vp; 1168 } 1169 1170 return (0); 1171 } 1172 1173 /* 1174 * Common routine for chroot and chdir. 1175 */ 1176 static int 1177 change_dir(struct nameidata *ndp, struct lwp *l) 1178 { 1179 struct vnode *vp; 1180 int error; 1181 1182 if ((error = namei(ndp)) != 0) 1183 return (error); 1184 vp = ndp->ni_vp; 1185 if (vp->v_type != VDIR) 1186 error = ENOTDIR; 1187 else 1188 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1189 1190 if (error) 1191 vput(vp); 1192 else 1193 VOP_UNLOCK(vp, 0); 1194 return (error); 1195 } 1196 1197 /* 1198 * Check permissions, allocate an open file structure, 1199 * and call the device open routine if any. 1200 */ 1201 int 1202 sys_open(struct lwp *l, void *v, register_t *retval) 1203 { 1204 struct sys_open_args /* { 1205 syscallarg(const char *) path; 1206 syscallarg(int) flags; 1207 syscallarg(int) mode; 1208 } */ *uap = v; 1209 struct proc *p = l->l_proc; 1210 struct cwdinfo *cwdi = p->p_cwdi; 1211 struct filedesc *fdp = p->p_fd; 1212 struct file *fp; 1213 struct vnode *vp; 1214 int flags, cmode; 1215 int type, indx, error; 1216 struct flock lf; 1217 struct nameidata nd; 1218 1219 flags = FFLAGS(SCARG(uap, flags)); 1220 if ((flags & (FREAD | FWRITE)) == 0) 1221 return (EINVAL); 1222 /* falloc() will use the file descriptor for us */ 1223 if ((error = falloc(l, &fp, &indx)) != 0) 1224 return (error); 1225 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1226 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1227 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1228 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1229 FILE_UNUSE(fp, l); 1230 fdp->fd_ofiles[indx] = NULL; 1231 ffree(fp); 1232 if ((error == EDUPFD || error == EMOVEFD) && 1233 l->l_dupfd >= 0 && /* XXX from fdopen */ 1234 (error = 1235 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1236 *retval = indx; 1237 return (0); 1238 } 1239 if (error == ERESTART) 1240 error = EINTR; 1241 fdremove(fdp, indx); 1242 return (error); 1243 } 1244 l->l_dupfd = 0; 1245 vp = nd.ni_vp; 1246 fp->f_flag = flags & FMASK; 1247 fp->f_type = DTYPE_VNODE; 1248 fp->f_ops = &vnops; 1249 fp->f_data = vp; 1250 if (flags & (O_EXLOCK | O_SHLOCK)) { 1251 lf.l_whence = SEEK_SET; 1252 lf.l_start = 0; 1253 lf.l_len = 0; 1254 if (flags & O_EXLOCK) 1255 lf.l_type = F_WRLCK; 1256 else 1257 lf.l_type = F_RDLCK; 1258 type = F_FLOCK; 1259 if ((flags & FNONBLOCK) == 0) 1260 type |= F_WAIT; 1261 VOP_UNLOCK(vp, 0); 1262 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1263 if (error) { 1264 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1265 FILE_UNUSE(fp, l); 1266 ffree(fp); 1267 fdremove(fdp, indx); 1268 return (error); 1269 } 1270 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1271 fp->f_flag |= FHASLOCK; 1272 } 1273 VOP_UNLOCK(vp, 0); 1274 *retval = indx; 1275 FILE_SET_MATURE(fp); 1276 FILE_UNUSE(fp, l); 1277 return (0); 1278 } 1279 1280 static void 1281 vfs__fhfree(fhandle_t *fhp) 1282 { 1283 size_t fhsize; 1284 1285 if (fhp == NULL) { 1286 return; 1287 } 1288 fhsize = FHANDLE_SIZE(fhp); 1289 kmem_free(fhp, fhsize); 1290 } 1291 1292 /* 1293 * vfs_composefh: compose a filehandle. 1294 */ 1295 1296 int 1297 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1298 { 1299 struct mount *mp; 1300 struct fid *fidp; 1301 int error; 1302 size_t needfhsize; 1303 size_t fidsize; 1304 1305 mp = vp->v_mount; 1306 fidp = NULL; 1307 if (*fh_size < FHANDLE_SIZE_MIN) { 1308 fidsize = 0; 1309 } else { 1310 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1311 if (fhp != NULL) { 1312 memset(fhp, 0, *fh_size); 1313 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1314 fidp = &fhp->fh_fid; 1315 } 1316 } 1317 error = VFS_VPTOFH(vp, fidp, &fidsize); 1318 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1319 if (error == 0 && *fh_size < needfhsize) { 1320 error = E2BIG; 1321 } 1322 *fh_size = needfhsize; 1323 return error; 1324 } 1325 1326 int 1327 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1328 { 1329 struct mount *mp; 1330 fhandle_t *fhp; 1331 size_t fhsize; 1332 size_t fidsize; 1333 int error; 1334 1335 *fhpp = NULL; 1336 mp = vp->v_mount; 1337 fidsize = 0; 1338 error = VFS_VPTOFH(vp, NULL, &fidsize); 1339 KASSERT(error != 0); 1340 if (error != E2BIG) { 1341 goto out; 1342 } 1343 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1344 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1345 if (fhp == NULL) { 1346 error = ENOMEM; 1347 goto out; 1348 } 1349 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1350 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1351 if (error == 0) { 1352 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1353 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1354 *fhpp = fhp; 1355 } else { 1356 kmem_free(fhp, fhsize); 1357 } 1358 out: 1359 return error; 1360 } 1361 1362 void 1363 vfs_composefh_free(fhandle_t *fhp) 1364 { 1365 1366 vfs__fhfree(fhp); 1367 } 1368 1369 /* 1370 * vfs_fhtovp: lookup a vnode by a filehandle. 1371 */ 1372 1373 int 1374 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1375 { 1376 struct mount *mp; 1377 int error; 1378 1379 *vpp = NULL; 1380 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1381 if (mp == NULL) { 1382 error = ESTALE; 1383 goto out; 1384 } 1385 if (mp->mnt_op->vfs_fhtovp == NULL) { 1386 error = EOPNOTSUPP; 1387 goto out; 1388 } 1389 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1390 out: 1391 return error; 1392 } 1393 1394 /* 1395 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1396 * the needed size. 1397 */ 1398 1399 int 1400 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1401 { 1402 fhandle_t *fhp; 1403 int error; 1404 1405 *fhpp = NULL; 1406 if (fhsize > FHANDLE_SIZE_MAX) { 1407 return EINVAL; 1408 } 1409 if (fhsize < FHANDLE_SIZE_MIN) { 1410 return EINVAL; 1411 } 1412 again: 1413 fhp = kmem_alloc(fhsize, KM_SLEEP); 1414 if (fhp == NULL) { 1415 return ENOMEM; 1416 } 1417 error = copyin(ufhp, fhp, fhsize); 1418 if (error == 0) { 1419 /* XXX this check shouldn't be here */ 1420 if (FHANDLE_SIZE(fhp) == fhsize) { 1421 *fhpp = fhp; 1422 return 0; 1423 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1424 /* 1425 * a kludge for nfsv2 padded handles. 1426 */ 1427 size_t sz; 1428 1429 sz = FHANDLE_SIZE(fhp); 1430 kmem_free(fhp, fhsize); 1431 fhsize = sz; 1432 goto again; 1433 } else { 1434 /* 1435 * userland told us wrong size. 1436 */ 1437 error = EINVAL; 1438 } 1439 } 1440 kmem_free(fhp, fhsize); 1441 return error; 1442 } 1443 1444 void 1445 vfs_copyinfh_free(fhandle_t *fhp) 1446 { 1447 1448 vfs__fhfree(fhp); 1449 } 1450 1451 /* 1452 * Get file handle system call 1453 */ 1454 int 1455 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1456 { 1457 struct sys___getfh30_args /* { 1458 syscallarg(char *) fname; 1459 syscallarg(fhandle_t *) fhp; 1460 syscallarg(size_t *) fh_size; 1461 } */ *uap = v; 1462 struct vnode *vp; 1463 fhandle_t *fh; 1464 int error; 1465 struct nameidata nd; 1466 size_t sz; 1467 size_t usz; 1468 1469 /* 1470 * Must be super user 1471 */ 1472 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1473 0, NULL, NULL, NULL); 1474 if (error) 1475 return (error); 1476 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1477 SCARG(uap, fname), l); 1478 error = namei(&nd); 1479 if (error) 1480 return (error); 1481 vp = nd.ni_vp; 1482 error = vfs_composefh_alloc(vp, &fh); 1483 vput(vp); 1484 if (error != 0) { 1485 goto out; 1486 } 1487 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1488 if (error != 0) { 1489 goto out; 1490 } 1491 sz = FHANDLE_SIZE(fh); 1492 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1493 if (error != 0) { 1494 goto out; 1495 } 1496 if (usz >= sz) { 1497 error = copyout(fh, SCARG(uap, fhp), sz); 1498 } else { 1499 error = E2BIG; 1500 } 1501 out: 1502 vfs_composefh_free(fh); 1503 return (error); 1504 } 1505 1506 /* 1507 * Open a file given a file handle. 1508 * 1509 * Check permissions, allocate an open file structure, 1510 * and call the device open routine if any. 1511 */ 1512 1513 int 1514 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1515 register_t *retval) 1516 { 1517 struct filedesc *fdp = l->l_proc->p_fd; 1518 struct file *fp; 1519 struct vnode *vp = NULL; 1520 kauth_cred_t cred = l->l_cred; 1521 struct file *nfp; 1522 int type, indx, error=0; 1523 struct flock lf; 1524 struct vattr va; 1525 fhandle_t *fh; 1526 int flags; 1527 1528 /* 1529 * Must be super user 1530 */ 1531 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1532 0, NULL, NULL, NULL))) 1533 return (error); 1534 1535 flags = FFLAGS(oflags); 1536 if ((flags & (FREAD | FWRITE)) == 0) 1537 return (EINVAL); 1538 if ((flags & O_CREAT)) 1539 return (EINVAL); 1540 /* falloc() will use the file descriptor for us */ 1541 if ((error = falloc(l, &nfp, &indx)) != 0) 1542 return (error); 1543 fp = nfp; 1544 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1545 if (error != 0) { 1546 goto bad; 1547 } 1548 error = vfs_fhtovp(fh, &vp); 1549 if (error != 0) { 1550 goto bad; 1551 } 1552 1553 /* Now do an effective vn_open */ 1554 1555 if (vp->v_type == VSOCK) { 1556 error = EOPNOTSUPP; 1557 goto bad; 1558 } 1559 if (flags & FREAD) { 1560 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1561 goto bad; 1562 } 1563 if (flags & (FWRITE | O_TRUNC)) { 1564 if (vp->v_type == VDIR) { 1565 error = EISDIR; 1566 goto bad; 1567 } 1568 if ((error = vn_writechk(vp)) != 0 || 1569 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1570 goto bad; 1571 } 1572 if (flags & O_TRUNC) { 1573 VOP_UNLOCK(vp, 0); /* XXX */ 1574 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1575 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1576 VATTR_NULL(&va); 1577 va.va_size = 0; 1578 error = VOP_SETATTR(vp, &va, cred, l); 1579 if (error) 1580 goto bad; 1581 } 1582 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1583 goto bad; 1584 if (vp->v_type == VREG && 1585 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1586 error = EIO; 1587 goto bad; 1588 } 1589 if (flags & FWRITE) 1590 vp->v_writecount++; 1591 1592 /* done with modified vn_open, now finish what sys_open does. */ 1593 1594 fp->f_flag = flags & FMASK; 1595 fp->f_type = DTYPE_VNODE; 1596 fp->f_ops = &vnops; 1597 fp->f_data = vp; 1598 if (flags & (O_EXLOCK | O_SHLOCK)) { 1599 lf.l_whence = SEEK_SET; 1600 lf.l_start = 0; 1601 lf.l_len = 0; 1602 if (flags & O_EXLOCK) 1603 lf.l_type = F_WRLCK; 1604 else 1605 lf.l_type = F_RDLCK; 1606 type = F_FLOCK; 1607 if ((flags & FNONBLOCK) == 0) 1608 type |= F_WAIT; 1609 VOP_UNLOCK(vp, 0); 1610 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1611 if (error) { 1612 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1613 FILE_UNUSE(fp, l); 1614 ffree(fp); 1615 fdremove(fdp, indx); 1616 return (error); 1617 } 1618 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1619 fp->f_flag |= FHASLOCK; 1620 } 1621 VOP_UNLOCK(vp, 0); 1622 *retval = indx; 1623 FILE_SET_MATURE(fp); 1624 FILE_UNUSE(fp, l); 1625 vfs_copyinfh_free(fh); 1626 return (0); 1627 1628 bad: 1629 FILE_UNUSE(fp, l); 1630 ffree(fp); 1631 fdremove(fdp, indx); 1632 if (vp != NULL) 1633 vput(vp); 1634 vfs_copyinfh_free(fh); 1635 return (error); 1636 } 1637 1638 int 1639 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1640 { 1641 struct sys___fhopen40_args /* { 1642 syscallarg(const void *) fhp; 1643 syscallarg(size_t) fh_size; 1644 syscallarg(int) flags; 1645 } */ *uap = v; 1646 1647 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1648 SCARG(uap, flags), retval); 1649 } 1650 1651 int 1652 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1653 { 1654 int error; 1655 fhandle_t *fh; 1656 struct vnode *vp; 1657 1658 /* 1659 * Must be super user 1660 */ 1661 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1662 0, NULL, NULL, NULL))) 1663 return (error); 1664 1665 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1666 if (error != 0) 1667 return error; 1668 1669 error = vfs_fhtovp(fh, &vp); 1670 vfs_copyinfh_free(fh); 1671 if (error != 0) 1672 return error; 1673 1674 error = vn_stat(vp, sb, l); 1675 vput(vp); 1676 return error; 1677 } 1678 1679 1680 /* ARGSUSED */ 1681 int 1682 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1683 { 1684 struct sys___fhstat40_args /* { 1685 syscallarg(const void *) fhp; 1686 syscallarg(size_t) fh_size; 1687 syscallarg(struct stat *) sb; 1688 } */ *uap = v; 1689 struct stat sb; 1690 int error; 1691 1692 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1693 if (error) 1694 return error; 1695 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1696 } 1697 1698 int 1699 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1700 int flags) 1701 { 1702 fhandle_t *fh; 1703 struct mount *mp; 1704 struct vnode *vp; 1705 int error; 1706 1707 /* 1708 * Must be super user 1709 */ 1710 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1711 0, NULL, NULL, NULL))) 1712 return error; 1713 1714 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1715 if (error != 0) 1716 return error; 1717 1718 error = vfs_fhtovp(fh, &vp); 1719 vfs_copyinfh_free(fh); 1720 if (error != 0) 1721 return error; 1722 1723 mp = vp->v_mount; 1724 error = dostatvfs(mp, sb, l, flags, 1); 1725 vput(vp); 1726 return error; 1727 } 1728 1729 /* ARGSUSED */ 1730 int 1731 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1732 { 1733 struct sys___fhstatvfs140_args /* { 1734 syscallarg(const void *) fhp; 1735 syscallarg(size_t) fh_size; 1736 syscallarg(struct statvfs *) buf; 1737 syscallarg(int) flags; 1738 } */ *uap = v; 1739 struct statvfs *sb = STATVFSBUF_GET(); 1740 int error; 1741 1742 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1743 SCARG(uap, flags)); 1744 if (error == 0) 1745 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1746 STATVFSBUF_PUT(sb); 1747 return error; 1748 } 1749 1750 /* 1751 * Create a special file. 1752 */ 1753 /* ARGSUSED */ 1754 int 1755 sys_mknod(struct lwp *l, void *v, register_t *retval) 1756 { 1757 struct sys_mknod_args /* { 1758 syscallarg(const char *) path; 1759 syscallarg(int) mode; 1760 syscallarg(int) dev; 1761 } */ *uap = v; 1762 struct proc *p = l->l_proc; 1763 struct vnode *vp; 1764 struct vattr vattr; 1765 int error, optype; 1766 struct nameidata nd; 1767 1768 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1769 0, NULL, NULL, NULL)) != 0) 1770 return (error); 1771 1772 optype = VOP_MKNOD_DESCOFFSET; 1773 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1774 if ((error = namei(&nd)) != 0) 1775 return (error); 1776 vp = nd.ni_vp; 1777 if (vp != NULL) 1778 error = EEXIST; 1779 else { 1780 VATTR_NULL(&vattr); 1781 vattr.va_mode = 1782 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1783 vattr.va_rdev = SCARG(uap, dev); 1784 1785 switch (SCARG(uap, mode) & S_IFMT) { 1786 case S_IFMT: /* used by badsect to flag bad sectors */ 1787 vattr.va_type = VBAD; 1788 break; 1789 case S_IFCHR: 1790 vattr.va_type = VCHR; 1791 break; 1792 case S_IFBLK: 1793 vattr.va_type = VBLK; 1794 break; 1795 case S_IFWHT: 1796 optype = VOP_WHITEOUT_DESCOFFSET; 1797 break; 1798 case S_IFREG: 1799 vattr.va_type = VREG; 1800 vattr.va_rdev = VNOVAL; 1801 optype = VOP_CREATE_DESCOFFSET; 1802 break; 1803 default: 1804 error = EINVAL; 1805 break; 1806 } 1807 } 1808 if (!error) { 1809 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1810 switch (optype) { 1811 case VOP_WHITEOUT_DESCOFFSET: 1812 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1813 if (error) 1814 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1815 vput(nd.ni_dvp); 1816 break; 1817 1818 case VOP_MKNOD_DESCOFFSET: 1819 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1820 &nd.ni_cnd, &vattr); 1821 if (error == 0) 1822 vput(nd.ni_vp); 1823 break; 1824 1825 case VOP_CREATE_DESCOFFSET: 1826 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1827 &nd.ni_cnd, &vattr); 1828 if (error == 0) 1829 vput(nd.ni_vp); 1830 break; 1831 } 1832 } else { 1833 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1834 if (nd.ni_dvp == vp) 1835 vrele(nd.ni_dvp); 1836 else 1837 vput(nd.ni_dvp); 1838 if (vp) 1839 vrele(vp); 1840 } 1841 return (error); 1842 } 1843 1844 /* 1845 * Create a named pipe. 1846 */ 1847 /* ARGSUSED */ 1848 int 1849 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1850 { 1851 struct sys_mkfifo_args /* { 1852 syscallarg(const char *) path; 1853 syscallarg(int) mode; 1854 } */ *uap = v; 1855 struct proc *p = l->l_proc; 1856 struct vattr vattr; 1857 int error; 1858 struct nameidata nd; 1859 1860 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1861 if ((error = namei(&nd)) != 0) 1862 return (error); 1863 if (nd.ni_vp != NULL) { 1864 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1865 if (nd.ni_dvp == nd.ni_vp) 1866 vrele(nd.ni_dvp); 1867 else 1868 vput(nd.ni_dvp); 1869 vrele(nd.ni_vp); 1870 return (EEXIST); 1871 } 1872 VATTR_NULL(&vattr); 1873 vattr.va_type = VFIFO; 1874 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1875 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1876 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1877 if (error == 0) 1878 vput(nd.ni_vp); 1879 return (error); 1880 } 1881 1882 /* 1883 * Make a hard file link. 1884 */ 1885 /* ARGSUSED */ 1886 int 1887 sys_link(struct lwp *l, void *v, register_t *retval) 1888 { 1889 struct sys_link_args /* { 1890 syscallarg(const char *) path; 1891 syscallarg(const char *) link; 1892 } */ *uap = v; 1893 struct vnode *vp; 1894 struct nameidata nd; 1895 int error; 1896 1897 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1898 if ((error = namei(&nd)) != 0) 1899 return (error); 1900 vp = nd.ni_vp; 1901 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1902 if ((error = namei(&nd)) != 0) 1903 goto out; 1904 if (nd.ni_vp) { 1905 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1906 if (nd.ni_dvp == nd.ni_vp) 1907 vrele(nd.ni_dvp); 1908 else 1909 vput(nd.ni_dvp); 1910 vrele(nd.ni_vp); 1911 error = EEXIST; 1912 goto out; 1913 } 1914 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1915 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1916 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1917 out: 1918 vrele(vp); 1919 return (error); 1920 } 1921 1922 /* 1923 * Make a symbolic link. 1924 */ 1925 /* ARGSUSED */ 1926 int 1927 sys_symlink(struct lwp *l, void *v, register_t *retval) 1928 { 1929 struct sys_symlink_args /* { 1930 syscallarg(const char *) path; 1931 syscallarg(const char *) link; 1932 } */ *uap = v; 1933 struct proc *p = l->l_proc; 1934 struct vattr vattr; 1935 char *path; 1936 int error; 1937 struct nameidata nd; 1938 1939 path = PNBUF_GET(); 1940 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1941 if (error) 1942 goto out; 1943 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1944 if ((error = namei(&nd)) != 0) 1945 goto out; 1946 if (nd.ni_vp) { 1947 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1948 if (nd.ni_dvp == nd.ni_vp) 1949 vrele(nd.ni_dvp); 1950 else 1951 vput(nd.ni_dvp); 1952 vrele(nd.ni_vp); 1953 error = EEXIST; 1954 goto out; 1955 } 1956 VATTR_NULL(&vattr); 1957 vattr.va_type = VLNK; 1958 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1959 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1960 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1961 if (error == 0) 1962 vput(nd.ni_vp); 1963 out: 1964 PNBUF_PUT(path); 1965 return (error); 1966 } 1967 1968 /* 1969 * Delete a whiteout from the filesystem. 1970 */ 1971 /* ARGSUSED */ 1972 int 1973 sys_undelete(struct lwp *l, void *v, register_t *retval) 1974 { 1975 struct sys_undelete_args /* { 1976 syscallarg(const char *) path; 1977 } */ *uap = v; 1978 int error; 1979 struct nameidata nd; 1980 1981 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 1982 SCARG(uap, path), l); 1983 error = namei(&nd); 1984 if (error) 1985 return (error); 1986 1987 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1988 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1989 if (nd.ni_dvp == nd.ni_vp) 1990 vrele(nd.ni_dvp); 1991 else 1992 vput(nd.ni_dvp); 1993 if (nd.ni_vp) 1994 vrele(nd.ni_vp); 1995 return (EEXIST); 1996 } 1997 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1998 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1999 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2000 vput(nd.ni_dvp); 2001 return (error); 2002 } 2003 2004 /* 2005 * Delete a name from the filesystem. 2006 */ 2007 /* ARGSUSED */ 2008 int 2009 sys_unlink(struct lwp *l, void *v, register_t *retval) 2010 { 2011 struct sys_unlink_args /* { 2012 syscallarg(const char *) path; 2013 } */ *uap = v; 2014 struct vnode *vp; 2015 int error; 2016 struct nameidata nd; 2017 #if NVERIEXEC > 0 2018 pathname_t pathbuf = NULL; 2019 #endif /* NVERIEXEC > 0 */ 2020 2021 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2022 SCARG(uap, path), l); 2023 if ((error = namei(&nd)) != 0) 2024 return (error); 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 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2043 2044 /* Handle remove requests for veriexec entries. */ 2045 if (!error) { 2046 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2047 pathname_put(pathbuf); 2048 } 2049 2050 if (error) { 2051 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2052 if (nd.ni_dvp == vp) 2053 vrele(nd.ni_dvp); 2054 else 2055 vput(nd.ni_dvp); 2056 vput(vp); 2057 goto out; 2058 } 2059 #endif /* NVERIEXEC > 0 */ 2060 2061 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2062 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2063 #ifdef FILEASSOC 2064 (void)fileassoc_file_delete(vp); 2065 #endif /* FILEASSOC */ 2066 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2067 out: 2068 return (error); 2069 } 2070 2071 /* 2072 * Reposition read/write file offset. 2073 */ 2074 int 2075 sys_lseek(struct lwp *l, void *v, register_t *retval) 2076 { 2077 struct sys_lseek_args /* { 2078 syscallarg(int) fd; 2079 syscallarg(int) pad; 2080 syscallarg(off_t) offset; 2081 syscallarg(int) whence; 2082 } */ *uap = v; 2083 struct proc *p = l->l_proc; 2084 kauth_cred_t cred = l->l_cred; 2085 struct filedesc *fdp = p->p_fd; 2086 struct file *fp; 2087 struct vnode *vp; 2088 struct vattr vattr; 2089 off_t newoff; 2090 int error; 2091 2092 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2093 return (EBADF); 2094 2095 FILE_USE(fp); 2096 2097 vp = (struct vnode *)fp->f_data; 2098 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2099 error = ESPIPE; 2100 goto out; 2101 } 2102 2103 switch (SCARG(uap, whence)) { 2104 case SEEK_CUR: 2105 newoff = fp->f_offset + SCARG(uap, offset); 2106 break; 2107 case SEEK_END: 2108 error = VOP_GETATTR(vp, &vattr, cred, l); 2109 if (error) 2110 goto out; 2111 newoff = SCARG(uap, offset) + vattr.va_size; 2112 break; 2113 case SEEK_SET: 2114 newoff = SCARG(uap, offset); 2115 break; 2116 default: 2117 error = EINVAL; 2118 goto out; 2119 } 2120 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2121 goto out; 2122 2123 *(off_t *)retval = fp->f_offset = newoff; 2124 out: 2125 FILE_UNUSE(fp, l); 2126 return (error); 2127 } 2128 2129 /* 2130 * Positional read system call. 2131 */ 2132 int 2133 sys_pread(struct lwp *l, void *v, register_t *retval) 2134 { 2135 struct sys_pread_args /* { 2136 syscallarg(int) fd; 2137 syscallarg(void *) buf; 2138 syscallarg(size_t) nbyte; 2139 syscallarg(off_t) offset; 2140 } */ *uap = v; 2141 struct proc *p = l->l_proc; 2142 struct filedesc *fdp = p->p_fd; 2143 struct file *fp; 2144 struct vnode *vp; 2145 off_t offset; 2146 int error, fd = SCARG(uap, fd); 2147 2148 if ((fp = fd_getfile(fdp, fd)) == NULL) 2149 return (EBADF); 2150 2151 if ((fp->f_flag & FREAD) == 0) { 2152 simple_unlock(&fp->f_slock); 2153 return (EBADF); 2154 } 2155 2156 FILE_USE(fp); 2157 2158 vp = (struct vnode *)fp->f_data; 2159 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2160 error = ESPIPE; 2161 goto out; 2162 } 2163 2164 offset = SCARG(uap, offset); 2165 2166 /* 2167 * XXX This works because no file systems actually 2168 * XXX take any action on the seek operation. 2169 */ 2170 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2171 goto out; 2172 2173 /* dofileread() will unuse the descriptor for us */ 2174 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2175 &offset, 0, retval)); 2176 2177 out: 2178 FILE_UNUSE(fp, l); 2179 return (error); 2180 } 2181 2182 /* 2183 * Positional scatter read system call. 2184 */ 2185 int 2186 sys_preadv(struct lwp *l, void *v, register_t *retval) 2187 { 2188 struct sys_preadv_args /* { 2189 syscallarg(int) fd; 2190 syscallarg(const struct iovec *) iovp; 2191 syscallarg(int) iovcnt; 2192 syscallarg(off_t) offset; 2193 } */ *uap = v; 2194 struct proc *p = l->l_proc; 2195 struct filedesc *fdp = p->p_fd; 2196 struct file *fp; 2197 struct vnode *vp; 2198 off_t offset; 2199 int error, fd = SCARG(uap, fd); 2200 2201 if ((fp = fd_getfile(fdp, fd)) == NULL) 2202 return (EBADF); 2203 2204 if ((fp->f_flag & FREAD) == 0) { 2205 simple_unlock(&fp->f_slock); 2206 return (EBADF); 2207 } 2208 2209 FILE_USE(fp); 2210 2211 vp = (struct vnode *)fp->f_data; 2212 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2213 error = ESPIPE; 2214 goto out; 2215 } 2216 2217 offset = SCARG(uap, offset); 2218 2219 /* 2220 * XXX This works because no file systems actually 2221 * XXX take any action on the seek operation. 2222 */ 2223 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2224 goto out; 2225 2226 /* dofilereadv() will unuse the descriptor for us */ 2227 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2228 &offset, 0, retval)); 2229 2230 out: 2231 FILE_UNUSE(fp, l); 2232 return (error); 2233 } 2234 2235 /* 2236 * Positional write system call. 2237 */ 2238 int 2239 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2240 { 2241 struct sys_pwrite_args /* { 2242 syscallarg(int) fd; 2243 syscallarg(const void *) buf; 2244 syscallarg(size_t) nbyte; 2245 syscallarg(off_t) offset; 2246 } */ *uap = v; 2247 struct proc *p = l->l_proc; 2248 struct filedesc *fdp = p->p_fd; 2249 struct file *fp; 2250 struct vnode *vp; 2251 off_t offset; 2252 int error, fd = SCARG(uap, fd); 2253 2254 if ((fp = fd_getfile(fdp, fd)) == NULL) 2255 return (EBADF); 2256 2257 if ((fp->f_flag & FWRITE) == 0) { 2258 simple_unlock(&fp->f_slock); 2259 return (EBADF); 2260 } 2261 2262 FILE_USE(fp); 2263 2264 vp = (struct vnode *)fp->f_data; 2265 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2266 error = ESPIPE; 2267 goto out; 2268 } 2269 2270 offset = SCARG(uap, offset); 2271 2272 /* 2273 * XXX This works because no file systems actually 2274 * XXX take any action on the seek operation. 2275 */ 2276 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2277 goto out; 2278 2279 /* dofilewrite() will unuse the descriptor for us */ 2280 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2281 &offset, 0, retval)); 2282 2283 out: 2284 FILE_UNUSE(fp, l); 2285 return (error); 2286 } 2287 2288 /* 2289 * Positional gather write system call. 2290 */ 2291 int 2292 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2293 { 2294 struct sys_pwritev_args /* { 2295 syscallarg(int) fd; 2296 syscallarg(const struct iovec *) iovp; 2297 syscallarg(int) iovcnt; 2298 syscallarg(off_t) offset; 2299 } */ *uap = v; 2300 struct proc *p = l->l_proc; 2301 struct filedesc *fdp = p->p_fd; 2302 struct file *fp; 2303 struct vnode *vp; 2304 off_t offset; 2305 int error, fd = SCARG(uap, fd); 2306 2307 if ((fp = fd_getfile(fdp, fd)) == NULL) 2308 return (EBADF); 2309 2310 if ((fp->f_flag & FWRITE) == 0) { 2311 simple_unlock(&fp->f_slock); 2312 return (EBADF); 2313 } 2314 2315 FILE_USE(fp); 2316 2317 vp = (struct vnode *)fp->f_data; 2318 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2319 error = ESPIPE; 2320 goto out; 2321 } 2322 2323 offset = SCARG(uap, offset); 2324 2325 /* 2326 * XXX This works because no file systems actually 2327 * XXX take any action on the seek operation. 2328 */ 2329 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2330 goto out; 2331 2332 /* dofilewritev() will unuse the descriptor for us */ 2333 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2334 &offset, 0, retval)); 2335 2336 out: 2337 FILE_UNUSE(fp, l); 2338 return (error); 2339 } 2340 2341 /* 2342 * Check access permissions. 2343 */ 2344 int 2345 sys_access(struct lwp *l, void *v, register_t *retval) 2346 { 2347 struct sys_access_args /* { 2348 syscallarg(const char *) path; 2349 syscallarg(int) flags; 2350 } */ *uap = v; 2351 kauth_cred_t cred; 2352 struct vnode *vp; 2353 int error, flags; 2354 struct nameidata nd; 2355 2356 cred = kauth_cred_dup(l->l_cred); 2357 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2358 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2359 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2360 SCARG(uap, path), l); 2361 /* Override default credentials */ 2362 nd.ni_cnd.cn_cred = cred; 2363 if ((error = namei(&nd)) != 0) 2364 goto out; 2365 vp = nd.ni_vp; 2366 2367 /* Flags == 0 means only check for existence. */ 2368 if (SCARG(uap, flags)) { 2369 flags = 0; 2370 if (SCARG(uap, flags) & R_OK) 2371 flags |= VREAD; 2372 if (SCARG(uap, flags) & W_OK) 2373 flags |= VWRITE; 2374 if (SCARG(uap, flags) & X_OK) 2375 flags |= VEXEC; 2376 2377 error = VOP_ACCESS(vp, flags, cred, l); 2378 if (!error && (flags & VWRITE)) 2379 error = vn_writechk(vp); 2380 } 2381 vput(vp); 2382 out: 2383 kauth_cred_free(cred); 2384 return (error); 2385 } 2386 2387 /* 2388 * Common code for all sys_stat functions, including compat versions. 2389 */ 2390 int 2391 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2392 struct stat *sb) 2393 { 2394 int error; 2395 struct nameidata nd; 2396 2397 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2398 error = namei(&nd); 2399 if (error != 0) 2400 return error; 2401 error = vn_stat(nd.ni_vp, sb, l); 2402 vput(nd.ni_vp); 2403 return error; 2404 } 2405 2406 /* 2407 * Get file status; this version follows links. 2408 */ 2409 /* ARGSUSED */ 2410 int 2411 sys___stat30(struct lwp *l, void *v, register_t *retval) 2412 { 2413 struct sys___stat30_args /* { 2414 syscallarg(const char *) path; 2415 syscallarg(struct stat *) ub; 2416 } */ *uap = v; 2417 struct stat sb; 2418 int error; 2419 2420 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2421 if (error) 2422 return error; 2423 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2424 } 2425 2426 /* 2427 * Get file status; this version does not follow links. 2428 */ 2429 /* ARGSUSED */ 2430 int 2431 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2432 { 2433 struct sys___lstat30_args /* { 2434 syscallarg(const char *) path; 2435 syscallarg(struct stat *) ub; 2436 } */ *uap = v; 2437 struct stat sb; 2438 int error; 2439 2440 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2441 if (error) 2442 return error; 2443 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2444 } 2445 2446 /* 2447 * Get configurable pathname variables. 2448 */ 2449 /* ARGSUSED */ 2450 int 2451 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2452 { 2453 struct sys_pathconf_args /* { 2454 syscallarg(const char *) path; 2455 syscallarg(int) name; 2456 } */ *uap = v; 2457 int error; 2458 struct nameidata nd; 2459 2460 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2461 SCARG(uap, path), l); 2462 if ((error = namei(&nd)) != 0) 2463 return (error); 2464 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2465 vput(nd.ni_vp); 2466 return (error); 2467 } 2468 2469 /* 2470 * Return target name of a symbolic link. 2471 */ 2472 /* ARGSUSED */ 2473 int 2474 sys_readlink(struct lwp *l, void *v, register_t *retval) 2475 { 2476 struct sys_readlink_args /* { 2477 syscallarg(const char *) path; 2478 syscallarg(char *) buf; 2479 syscallarg(size_t) count; 2480 } */ *uap = v; 2481 struct vnode *vp; 2482 struct iovec aiov; 2483 struct uio auio; 2484 int error; 2485 struct nameidata nd; 2486 2487 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2488 SCARG(uap, path), l); 2489 if ((error = namei(&nd)) != 0) 2490 return (error); 2491 vp = nd.ni_vp; 2492 if (vp->v_type != VLNK) 2493 error = EINVAL; 2494 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2495 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2496 aiov.iov_base = SCARG(uap, buf); 2497 aiov.iov_len = SCARG(uap, count); 2498 auio.uio_iov = &aiov; 2499 auio.uio_iovcnt = 1; 2500 auio.uio_offset = 0; 2501 auio.uio_rw = UIO_READ; 2502 KASSERT(l == curlwp); 2503 auio.uio_vmspace = l->l_proc->p_vmspace; 2504 auio.uio_resid = SCARG(uap, count); 2505 error = VOP_READLINK(vp, &auio, l->l_cred); 2506 } 2507 vput(vp); 2508 *retval = SCARG(uap, count) - auio.uio_resid; 2509 return (error); 2510 } 2511 2512 /* 2513 * Change flags of a file given a path name. 2514 */ 2515 /* ARGSUSED */ 2516 int 2517 sys_chflags(struct lwp *l, void *v, register_t *retval) 2518 { 2519 struct sys_chflags_args /* { 2520 syscallarg(const char *) path; 2521 syscallarg(u_long) flags; 2522 } */ *uap = v; 2523 struct vnode *vp; 2524 int error; 2525 struct nameidata nd; 2526 2527 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2528 if ((error = namei(&nd)) != 0) 2529 return (error); 2530 vp = nd.ni_vp; 2531 error = change_flags(vp, SCARG(uap, flags), l); 2532 vput(vp); 2533 return (error); 2534 } 2535 2536 /* 2537 * Change flags of a file given a file descriptor. 2538 */ 2539 /* ARGSUSED */ 2540 int 2541 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2542 { 2543 struct sys_fchflags_args /* { 2544 syscallarg(int) fd; 2545 syscallarg(u_long) flags; 2546 } */ *uap = v; 2547 struct proc *p = l->l_proc; 2548 struct vnode *vp; 2549 struct file *fp; 2550 int error; 2551 2552 /* getvnode() will use the descriptor for us */ 2553 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2554 return (error); 2555 vp = (struct vnode *)fp->f_data; 2556 error = change_flags(vp, SCARG(uap, flags), l); 2557 VOP_UNLOCK(vp, 0); 2558 FILE_UNUSE(fp, l); 2559 return (error); 2560 } 2561 2562 /* 2563 * Change flags of a file given a path name; this version does 2564 * not follow links. 2565 */ 2566 int 2567 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2568 { 2569 struct sys_lchflags_args /* { 2570 syscallarg(const char *) path; 2571 syscallarg(u_long) flags; 2572 } */ *uap = v; 2573 struct vnode *vp; 2574 int error; 2575 struct nameidata nd; 2576 2577 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2578 if ((error = namei(&nd)) != 0) 2579 return (error); 2580 vp = nd.ni_vp; 2581 error = change_flags(vp, SCARG(uap, flags), l); 2582 vput(vp); 2583 return (error); 2584 } 2585 2586 /* 2587 * Common routine to change flags of a file. 2588 */ 2589 int 2590 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2591 { 2592 struct vattr vattr; 2593 int error; 2594 2595 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2596 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2597 /* 2598 * Non-superusers cannot change the flags on devices, even if they 2599 * own them. 2600 */ 2601 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2602 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2603 goto out; 2604 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2605 error = EINVAL; 2606 goto out; 2607 } 2608 } 2609 VATTR_NULL(&vattr); 2610 vattr.va_flags = flags; 2611 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2612 out: 2613 return (error); 2614 } 2615 2616 /* 2617 * Change mode of a file given path name; this version follows links. 2618 */ 2619 /* ARGSUSED */ 2620 int 2621 sys_chmod(struct lwp *l, void *v, register_t *retval) 2622 { 2623 struct sys_chmod_args /* { 2624 syscallarg(const char *) path; 2625 syscallarg(int) mode; 2626 } */ *uap = v; 2627 int error; 2628 struct nameidata nd; 2629 2630 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2631 if ((error = namei(&nd)) != 0) 2632 return (error); 2633 2634 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2635 2636 vrele(nd.ni_vp); 2637 return (error); 2638 } 2639 2640 /* 2641 * Change mode of a file given a file descriptor. 2642 */ 2643 /* ARGSUSED */ 2644 int 2645 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2646 { 2647 struct sys_fchmod_args /* { 2648 syscallarg(int) fd; 2649 syscallarg(int) mode; 2650 } */ *uap = v; 2651 struct proc *p = l->l_proc; 2652 struct file *fp; 2653 int error; 2654 2655 /* getvnode() will use the descriptor for us */ 2656 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2657 return (error); 2658 2659 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2660 FILE_UNUSE(fp, l); 2661 return (error); 2662 } 2663 2664 /* 2665 * Change mode of a file given path name; this version does not follow links. 2666 */ 2667 /* ARGSUSED */ 2668 int 2669 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2670 { 2671 struct sys_lchmod_args /* { 2672 syscallarg(const char *) path; 2673 syscallarg(int) mode; 2674 } */ *uap = v; 2675 int error; 2676 struct nameidata nd; 2677 2678 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2679 if ((error = namei(&nd)) != 0) 2680 return (error); 2681 2682 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2683 2684 vrele(nd.ni_vp); 2685 return (error); 2686 } 2687 2688 /* 2689 * Common routine to set mode given a vnode. 2690 */ 2691 static int 2692 change_mode(struct vnode *vp, int mode, struct lwp *l) 2693 { 2694 struct vattr vattr; 2695 int error; 2696 2697 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2698 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2699 VATTR_NULL(&vattr); 2700 vattr.va_mode = mode & ALLPERMS; 2701 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2702 VOP_UNLOCK(vp, 0); 2703 return (error); 2704 } 2705 2706 /* 2707 * Set ownership given a path name; this version follows links. 2708 */ 2709 /* ARGSUSED */ 2710 int 2711 sys_chown(struct lwp *l, void *v, register_t *retval) 2712 { 2713 struct sys_chown_args /* { 2714 syscallarg(const char *) path; 2715 syscallarg(uid_t) uid; 2716 syscallarg(gid_t) gid; 2717 } */ *uap = v; 2718 int error; 2719 struct nameidata nd; 2720 2721 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2722 if ((error = namei(&nd)) != 0) 2723 return (error); 2724 2725 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2726 2727 vrele(nd.ni_vp); 2728 return (error); 2729 } 2730 2731 /* 2732 * Set ownership given a path name; this version follows links. 2733 * Provides POSIX semantics. 2734 */ 2735 /* ARGSUSED */ 2736 int 2737 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2738 { 2739 struct sys_chown_args /* { 2740 syscallarg(const char *) path; 2741 syscallarg(uid_t) uid; 2742 syscallarg(gid_t) gid; 2743 } */ *uap = v; 2744 int error; 2745 struct nameidata nd; 2746 2747 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2748 if ((error = namei(&nd)) != 0) 2749 return (error); 2750 2751 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2752 2753 vrele(nd.ni_vp); 2754 return (error); 2755 } 2756 2757 /* 2758 * Set ownership given a file descriptor. 2759 */ 2760 /* ARGSUSED */ 2761 int 2762 sys_fchown(struct lwp *l, void *v, register_t *retval) 2763 { 2764 struct sys_fchown_args /* { 2765 syscallarg(int) fd; 2766 syscallarg(uid_t) uid; 2767 syscallarg(gid_t) gid; 2768 } */ *uap = v; 2769 struct proc *p = l->l_proc; 2770 int error; 2771 struct file *fp; 2772 2773 /* getvnode() will use the descriptor for us */ 2774 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2775 return (error); 2776 2777 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2778 SCARG(uap, gid), l, 0); 2779 FILE_UNUSE(fp, l); 2780 return (error); 2781 } 2782 2783 /* 2784 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2785 */ 2786 /* ARGSUSED */ 2787 int 2788 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2789 { 2790 struct sys_fchown_args /* { 2791 syscallarg(int) fd; 2792 syscallarg(uid_t) uid; 2793 syscallarg(gid_t) gid; 2794 } */ *uap = v; 2795 struct proc *p = l->l_proc; 2796 int error; 2797 struct file *fp; 2798 2799 /* getvnode() will use the descriptor for us */ 2800 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2801 return (error); 2802 2803 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2804 SCARG(uap, gid), l, 1); 2805 FILE_UNUSE(fp, l); 2806 return (error); 2807 } 2808 2809 /* 2810 * Set ownership given a path name; this version does not follow links. 2811 */ 2812 /* ARGSUSED */ 2813 int 2814 sys_lchown(struct lwp *l, void *v, register_t *retval) 2815 { 2816 struct sys_lchown_args /* { 2817 syscallarg(const char *) path; 2818 syscallarg(uid_t) uid; 2819 syscallarg(gid_t) gid; 2820 } */ *uap = v; 2821 int error; 2822 struct nameidata nd; 2823 2824 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2825 if ((error = namei(&nd)) != 0) 2826 return (error); 2827 2828 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2829 2830 vrele(nd.ni_vp); 2831 return (error); 2832 } 2833 2834 /* 2835 * Set ownership given a path name; this version does not follow links. 2836 * Provides POSIX/XPG semantics. 2837 */ 2838 /* ARGSUSED */ 2839 int 2840 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2841 { 2842 struct sys_lchown_args /* { 2843 syscallarg(const char *) path; 2844 syscallarg(uid_t) uid; 2845 syscallarg(gid_t) gid; 2846 } */ *uap = v; 2847 int error; 2848 struct nameidata nd; 2849 2850 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2851 if ((error = namei(&nd)) != 0) 2852 return (error); 2853 2854 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2855 2856 vrele(nd.ni_vp); 2857 return (error); 2858 } 2859 2860 /* 2861 * Common routine to set ownership given a vnode. 2862 */ 2863 static int 2864 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2865 int posix_semantics) 2866 { 2867 struct vattr vattr; 2868 mode_t newmode; 2869 int error; 2870 2871 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2872 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2873 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2874 goto out; 2875 2876 #define CHANGED(x) ((int)(x) != -1) 2877 newmode = vattr.va_mode; 2878 if (posix_semantics) { 2879 /* 2880 * POSIX/XPG semantics: if the caller is not the super-user, 2881 * clear set-user-id and set-group-id bits. Both POSIX and 2882 * the XPG consider the behaviour for calls by the super-user 2883 * implementation-defined; we leave the set-user-id and set- 2884 * group-id settings intact in that case. 2885 */ 2886 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2887 NULL) != 0) 2888 newmode &= ~(S_ISUID | S_ISGID); 2889 } else { 2890 /* 2891 * NetBSD semantics: when changing owner and/or group, 2892 * clear the respective bit(s). 2893 */ 2894 if (CHANGED(uid)) 2895 newmode &= ~S_ISUID; 2896 if (CHANGED(gid)) 2897 newmode &= ~S_ISGID; 2898 } 2899 /* Update va_mode iff altered. */ 2900 if (vattr.va_mode == newmode) 2901 newmode = VNOVAL; 2902 2903 VATTR_NULL(&vattr); 2904 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2905 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2906 vattr.va_mode = newmode; 2907 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2908 #undef CHANGED 2909 2910 out: 2911 VOP_UNLOCK(vp, 0); 2912 return (error); 2913 } 2914 2915 /* 2916 * Set the access and modification times given a path name; this 2917 * version follows links. 2918 */ 2919 /* ARGSUSED */ 2920 int 2921 sys_utimes(struct lwp *l, void *v, register_t *retval) 2922 { 2923 struct sys_utimes_args /* { 2924 syscallarg(const char *) path; 2925 syscallarg(const struct timeval *) tptr; 2926 } */ *uap = v; 2927 int error; 2928 struct nameidata nd; 2929 2930 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2931 if ((error = namei(&nd)) != 0) 2932 return (error); 2933 2934 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2935 2936 vrele(nd.ni_vp); 2937 return (error); 2938 } 2939 2940 /* 2941 * Set the access and modification times given a file descriptor. 2942 */ 2943 /* ARGSUSED */ 2944 int 2945 sys_futimes(struct lwp *l, void *v, register_t *retval) 2946 { 2947 struct sys_futimes_args /* { 2948 syscallarg(int) fd; 2949 syscallarg(const struct timeval *) tptr; 2950 } */ *uap = v; 2951 struct proc *p = l->l_proc; 2952 int error; 2953 struct file *fp; 2954 2955 /* getvnode() will use the descriptor for us */ 2956 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2957 return (error); 2958 2959 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2960 FILE_UNUSE(fp, l); 2961 return (error); 2962 } 2963 2964 /* 2965 * Set the access and modification times given a path name; this 2966 * version does not follow links. 2967 */ 2968 /* ARGSUSED */ 2969 int 2970 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2971 { 2972 struct sys_lutimes_args /* { 2973 syscallarg(const char *) path; 2974 syscallarg(const struct timeval *) tptr; 2975 } */ *uap = v; 2976 int error; 2977 struct nameidata nd; 2978 2979 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2980 if ((error = namei(&nd)) != 0) 2981 return (error); 2982 2983 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2984 2985 vrele(nd.ni_vp); 2986 return (error); 2987 } 2988 2989 /* 2990 * Common routine to set access and modification times given a vnode. 2991 */ 2992 static int 2993 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2994 { 2995 struct vattr vattr; 2996 int error; 2997 2998 VATTR_NULL(&vattr); 2999 if (tptr == NULL) { 3000 nanotime(&vattr.va_atime); 3001 vattr.va_mtime = vattr.va_atime; 3002 vattr.va_vaflags |= VA_UTIMES_NULL; 3003 } else { 3004 struct timeval tv[2]; 3005 3006 error = copyin(tptr, tv, sizeof(tv)); 3007 if (error) 3008 goto out; 3009 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3010 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3011 } 3012 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3013 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3014 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3015 VOP_UNLOCK(vp, 0); 3016 out: 3017 return (error); 3018 } 3019 3020 /* 3021 * Truncate a file given its path name. 3022 */ 3023 /* ARGSUSED */ 3024 int 3025 sys_truncate(struct lwp *l, void *v, register_t *retval) 3026 { 3027 struct sys_truncate_args /* { 3028 syscallarg(const char *) path; 3029 syscallarg(int) pad; 3030 syscallarg(off_t) length; 3031 } */ *uap = v; 3032 struct vnode *vp; 3033 struct vattr vattr; 3034 int error; 3035 struct nameidata nd; 3036 3037 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3038 if ((error = namei(&nd)) != 0) 3039 return (error); 3040 vp = nd.ni_vp; 3041 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3042 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3043 if (vp->v_type == VDIR) 3044 error = EISDIR; 3045 else if ((error = vn_writechk(vp)) == 0 && 3046 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3047 VATTR_NULL(&vattr); 3048 vattr.va_size = SCARG(uap, length); 3049 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3050 } 3051 vput(vp); 3052 return (error); 3053 } 3054 3055 /* 3056 * Truncate a file given a file descriptor. 3057 */ 3058 /* ARGSUSED */ 3059 int 3060 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3061 { 3062 struct sys_ftruncate_args /* { 3063 syscallarg(int) fd; 3064 syscallarg(int) pad; 3065 syscallarg(off_t) length; 3066 } */ *uap = v; 3067 struct proc *p = l->l_proc; 3068 struct vattr vattr; 3069 struct vnode *vp; 3070 struct file *fp; 3071 int error; 3072 3073 /* getvnode() will use the descriptor for us */ 3074 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3075 return (error); 3076 if ((fp->f_flag & FWRITE) == 0) { 3077 error = EINVAL; 3078 goto out; 3079 } 3080 vp = (struct vnode *)fp->f_data; 3081 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3082 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3083 if (vp->v_type == VDIR) 3084 error = EISDIR; 3085 else if ((error = vn_writechk(vp)) == 0) { 3086 VATTR_NULL(&vattr); 3087 vattr.va_size = SCARG(uap, length); 3088 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3089 } 3090 VOP_UNLOCK(vp, 0); 3091 out: 3092 FILE_UNUSE(fp, l); 3093 return (error); 3094 } 3095 3096 /* 3097 * Sync an open file. 3098 */ 3099 /* ARGSUSED */ 3100 int 3101 sys_fsync(struct lwp *l, void *v, register_t *retval) 3102 { 3103 struct sys_fsync_args /* { 3104 syscallarg(int) fd; 3105 } */ *uap = v; 3106 struct proc *p = l->l_proc; 3107 struct vnode *vp; 3108 struct file *fp; 3109 int error; 3110 3111 /* getvnode() will use the descriptor for us */ 3112 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3113 return (error); 3114 vp = (struct vnode *)fp->f_data; 3115 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3116 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3117 if (error == 0 && bioops.io_fsync != NULL && 3118 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3119 (*bioops.io_fsync)(vp, 0); 3120 VOP_UNLOCK(vp, 0); 3121 FILE_UNUSE(fp, l); 3122 return (error); 3123 } 3124 3125 /* 3126 * Sync a range of file data. API modeled after that found in AIX. 3127 * 3128 * FDATASYNC indicates that we need only save enough metadata to be able 3129 * to re-read the written data. Note we duplicate AIX's requirement that 3130 * the file be open for writing. 3131 */ 3132 /* ARGSUSED */ 3133 int 3134 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3135 { 3136 struct sys_fsync_range_args /* { 3137 syscallarg(int) fd; 3138 syscallarg(int) flags; 3139 syscallarg(off_t) start; 3140 syscallarg(off_t) length; 3141 } */ *uap = v; 3142 struct proc *p = l->l_proc; 3143 struct vnode *vp; 3144 struct file *fp; 3145 int flags, nflags; 3146 off_t s, e, len; 3147 int error; 3148 3149 /* getvnode() will use the descriptor for us */ 3150 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3151 return (error); 3152 3153 if ((fp->f_flag & FWRITE) == 0) { 3154 error = EBADF; 3155 goto out; 3156 } 3157 3158 flags = SCARG(uap, flags); 3159 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3160 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3161 error = EINVAL; 3162 goto out; 3163 } 3164 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3165 if (flags & FDATASYNC) 3166 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3167 else 3168 nflags = FSYNC_WAIT; 3169 if (flags & FDISKSYNC) 3170 nflags |= FSYNC_CACHE; 3171 3172 len = SCARG(uap, length); 3173 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3174 if (len) { 3175 s = SCARG(uap, start); 3176 e = s + len; 3177 if (e < s) { 3178 error = EINVAL; 3179 goto out; 3180 } 3181 } else { 3182 e = 0; 3183 s = 0; 3184 } 3185 3186 vp = (struct vnode *)fp->f_data; 3187 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3188 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3189 3190 if (error == 0 && bioops.io_fsync != NULL && 3191 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3192 (*bioops.io_fsync)(vp, nflags); 3193 3194 VOP_UNLOCK(vp, 0); 3195 out: 3196 FILE_UNUSE(fp, l); 3197 return (error); 3198 } 3199 3200 /* 3201 * Sync the data of an open file. 3202 */ 3203 /* ARGSUSED */ 3204 int 3205 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3206 { 3207 struct sys_fdatasync_args /* { 3208 syscallarg(int) fd; 3209 } */ *uap = v; 3210 struct proc *p = l->l_proc; 3211 struct vnode *vp; 3212 struct file *fp; 3213 int error; 3214 3215 /* getvnode() will use the descriptor for us */ 3216 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3217 return (error); 3218 if ((fp->f_flag & FWRITE) == 0) { 3219 FILE_UNUSE(fp, l); 3220 return (EBADF); 3221 } 3222 vp = (struct vnode *)fp->f_data; 3223 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3224 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3225 VOP_UNLOCK(vp, 0); 3226 FILE_UNUSE(fp, l); 3227 return (error); 3228 } 3229 3230 /* 3231 * Rename files, (standard) BSD semantics frontend. 3232 */ 3233 /* ARGSUSED */ 3234 int 3235 sys_rename(struct lwp *l, void *v, register_t *retval) 3236 { 3237 struct sys_rename_args /* { 3238 syscallarg(const char *) from; 3239 syscallarg(const char *) to; 3240 } */ *uap = v; 3241 3242 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3243 } 3244 3245 /* 3246 * Rename files, POSIX semantics frontend. 3247 */ 3248 /* ARGSUSED */ 3249 int 3250 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3251 { 3252 struct sys___posix_rename_args /* { 3253 syscallarg(const char *) from; 3254 syscallarg(const char *) to; 3255 } */ *uap = v; 3256 3257 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3258 } 3259 3260 /* 3261 * Rename files. Source and destination must either both be directories, 3262 * or both not be directories. If target is a directory, it must be empty. 3263 * If `from' and `to' refer to the same object, the value of the `retain' 3264 * argument is used to determine whether `from' will be 3265 * 3266 * (retain == 0) deleted unless `from' and `to' refer to the same 3267 * object in the file system's name space (BSD). 3268 * (retain == 1) always retained (POSIX). 3269 */ 3270 static int 3271 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3272 { 3273 struct vnode *tvp, *fvp, *tdvp; 3274 struct nameidata fromnd, tond; 3275 struct proc *p; 3276 int error; 3277 3278 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3279 from, l); 3280 if ((error = namei(&fromnd)) != 0) 3281 return (error); 3282 if (fromnd.ni_dvp != fromnd.ni_vp) 3283 VOP_UNLOCK(fromnd.ni_dvp, 0); 3284 fvp = fromnd.ni_vp; 3285 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3286 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3287 if ((error = namei(&tond)) != 0) { 3288 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3289 vrele(fromnd.ni_dvp); 3290 vrele(fvp); 3291 goto out1; 3292 } 3293 tdvp = tond.ni_dvp; 3294 tvp = tond.ni_vp; 3295 3296 if (tvp != NULL) { 3297 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3298 error = ENOTDIR; 3299 goto out; 3300 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3301 error = EISDIR; 3302 goto out; 3303 } 3304 } 3305 3306 if (fvp == tdvp) 3307 error = EINVAL; 3308 3309 /* 3310 * Source and destination refer to the same object. 3311 */ 3312 if (fvp == tvp) { 3313 if (retain) 3314 error = -1; 3315 else if (fromnd.ni_dvp == tdvp && 3316 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3317 !memcmp(fromnd.ni_cnd.cn_nameptr, 3318 tond.ni_cnd.cn_nameptr, 3319 fromnd.ni_cnd.cn_namelen)) 3320 error = -1; 3321 } 3322 3323 #if NVERIEXEC > 0 3324 if (!error) { 3325 pathname_t frompath = NULL, topath = NULL; 3326 3327 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3328 &frompath); 3329 if (!error) 3330 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3331 &topath); 3332 if (!error) 3333 error = veriexec_renamechk(fvp, pathname_path(frompath), 3334 tvp, pathname_path(topath), l); 3335 3336 pathname_put(frompath); 3337 pathname_put(topath); 3338 } 3339 #endif /* NVERIEXEC > 0 */ 3340 3341 out: 3342 p = l->l_proc; 3343 if (!error) { 3344 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3345 if (fromnd.ni_dvp != tdvp) 3346 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3347 if (tvp) { 3348 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3349 } 3350 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3351 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3352 } else { 3353 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3354 if (tdvp == tvp) 3355 vrele(tdvp); 3356 else 3357 vput(tdvp); 3358 if (tvp) 3359 vput(tvp); 3360 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3361 vrele(fromnd.ni_dvp); 3362 vrele(fvp); 3363 } 3364 vrele(tond.ni_startdir); 3365 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3366 out1: 3367 if (fromnd.ni_startdir) 3368 vrele(fromnd.ni_startdir); 3369 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3370 return (error == -1 ? 0 : error); 3371 } 3372 3373 /* 3374 * Make a directory file. 3375 */ 3376 /* ARGSUSED */ 3377 int 3378 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3379 { 3380 struct sys_mkdir_args /* { 3381 syscallarg(const char *) path; 3382 syscallarg(int) mode; 3383 } */ *uap = v; 3384 struct proc *p = l->l_proc; 3385 struct vnode *vp; 3386 struct vattr vattr; 3387 int error; 3388 struct nameidata nd; 3389 3390 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3391 SCARG(uap, path), l); 3392 if ((error = namei(&nd)) != 0) 3393 return (error); 3394 vp = nd.ni_vp; 3395 if (vp != NULL) { 3396 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3397 if (nd.ni_dvp == vp) 3398 vrele(nd.ni_dvp); 3399 else 3400 vput(nd.ni_dvp); 3401 vrele(vp); 3402 return (EEXIST); 3403 } 3404 VATTR_NULL(&vattr); 3405 vattr.va_type = VDIR; 3406 vattr.va_mode = 3407 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3408 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3409 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3410 if (!error) 3411 vput(nd.ni_vp); 3412 return (error); 3413 } 3414 3415 /* 3416 * Remove a directory file. 3417 */ 3418 /* ARGSUSED */ 3419 int 3420 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3421 { 3422 struct sys_rmdir_args /* { 3423 syscallarg(const char *) path; 3424 } */ *uap = v; 3425 struct vnode *vp; 3426 int error; 3427 struct nameidata nd; 3428 3429 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3430 SCARG(uap, path), l); 3431 if ((error = namei(&nd)) != 0) 3432 return (error); 3433 vp = nd.ni_vp; 3434 if (vp->v_type != VDIR) { 3435 error = ENOTDIR; 3436 goto out; 3437 } 3438 /* 3439 * No rmdir "." please. 3440 */ 3441 if (nd.ni_dvp == vp) { 3442 error = EINVAL; 3443 goto out; 3444 } 3445 /* 3446 * The root of a mounted filesystem cannot be deleted. 3447 */ 3448 if (vp->v_flag & VROOT) { 3449 error = EBUSY; 3450 goto out; 3451 } 3452 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3453 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3454 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3455 return (error); 3456 3457 out: 3458 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3459 if (nd.ni_dvp == vp) 3460 vrele(nd.ni_dvp); 3461 else 3462 vput(nd.ni_dvp); 3463 vput(vp); 3464 return (error); 3465 } 3466 3467 /* 3468 * Read a block of directory entries in a file system independent format. 3469 */ 3470 int 3471 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3472 { 3473 struct sys___getdents30_args /* { 3474 syscallarg(int) fd; 3475 syscallarg(char *) buf; 3476 syscallarg(size_t) count; 3477 } */ *uap = v; 3478 struct proc *p = l->l_proc; 3479 struct file *fp; 3480 int error, done; 3481 3482 /* getvnode() will use the descriptor for us */ 3483 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3484 return (error); 3485 if ((fp->f_flag & FREAD) == 0) { 3486 error = EBADF; 3487 goto out; 3488 } 3489 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3490 SCARG(uap, count), &done, l, 0, 0); 3491 #ifdef KTRACE 3492 if (!error && KTRPOINT(p, KTR_GENIO)) { 3493 struct iovec iov; 3494 iov.iov_base = SCARG(uap, buf); 3495 iov.iov_len = done; 3496 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3497 } 3498 #endif 3499 *retval = done; 3500 out: 3501 FILE_UNUSE(fp, l); 3502 return (error); 3503 } 3504 3505 /* 3506 * Set the mode mask for creation of filesystem nodes. 3507 */ 3508 int 3509 sys_umask(struct lwp *l, void *v, register_t *retval) 3510 { 3511 struct sys_umask_args /* { 3512 syscallarg(mode_t) newmask; 3513 } */ *uap = v; 3514 struct proc *p = l->l_proc; 3515 struct cwdinfo *cwdi; 3516 3517 cwdi = p->p_cwdi; 3518 *retval = cwdi->cwdi_cmask; 3519 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3520 return (0); 3521 } 3522 3523 /* 3524 * Void all references to file by ripping underlying filesystem 3525 * away from vnode. 3526 */ 3527 /* ARGSUSED */ 3528 int 3529 sys_revoke(struct lwp *l, void *v, register_t *retval) 3530 { 3531 struct sys_revoke_args /* { 3532 syscallarg(const char *) path; 3533 } */ *uap = v; 3534 struct vnode *vp; 3535 struct vattr vattr; 3536 int error; 3537 struct nameidata nd; 3538 3539 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3540 if ((error = namei(&nd)) != 0) 3541 return (error); 3542 vp = nd.ni_vp; 3543 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3544 goto out; 3545 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3546 (error = kauth_authorize_generic(l->l_cred, 3547 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3548 goto out; 3549 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3550 VOP_REVOKE(vp, REVOKEALL); 3551 out: 3552 vrele(vp); 3553 return (error); 3554 } 3555 3556 /* 3557 * Convert a user file descriptor to a kernel file entry. 3558 */ 3559 int 3560 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3561 { 3562 struct vnode *vp; 3563 struct file *fp; 3564 3565 if ((fp = fd_getfile(fdp, fd)) == NULL) 3566 return (EBADF); 3567 3568 FILE_USE(fp); 3569 3570 if (fp->f_type != DTYPE_VNODE) { 3571 FILE_UNUSE(fp, NULL); 3572 return (EINVAL); 3573 } 3574 3575 vp = (struct vnode *)fp->f_data; 3576 if (vp->v_type == VBAD) { 3577 FILE_UNUSE(fp, NULL); 3578 return (EBADF); 3579 } 3580 3581 *fpp = fp; 3582 return (0); 3583 } 3584
Indexes created Sun Sep 28 16:09:52 GMT 2025