vfs_syscalls.c revision 1.307
1 /* $NetBSD: vfs_syscalls.c,v 1.307 2007/04/01 10:15:01 hannken Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.307 2007/04/01 10:15:01 hannken Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/syscallargs.h> 65 #include <sys/vfs_syscalls.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #if NVERIEXEC > 0 73 #include <sys/verified_exec.h> 74 #include <sys/syslog.h> 75 #endif /* NVERIEXEC > 0 */ 76 #include <sys/kauth.h> 77 78 #include <miscfs/genfs/genfs.h> 79 #include <miscfs/syncfs/syncfs.h> 80 81 #ifdef COMPAT_30 82 #include "opt_nfsserver.h" 83 #include <nfs/rpcv2.h> 84 #endif 85 #include <nfs/nfsproto.h> 86 #ifdef COMPAT_30 87 #include <nfs/nfs.h> 88 #include <nfs/nfs_var.h> 89 #endif 90 91 #if NFSS > 0 92 #include <dev/fssvar.h> 93 #endif 94 95 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 96 97 static int change_dir(struct nameidata *, struct lwp *); 98 static int change_flags(struct vnode *, u_long, struct lwp *); 99 static int change_mode(struct vnode *, int, struct lwp *l); 100 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 101 static int change_utimes(struct vnode *vp, const struct timeval *, 102 struct lwp *l); 103 static int rename_files(const char *, const char *, struct lwp *, int); 104 105 void checkdirs(struct vnode *); 106 107 static int mount_update(struct lwp *, struct vnode *, const char *, int, 108 void *, struct nameidata *); 109 static int mount_domount(struct lwp *, struct vnode *, const char *, 110 const char *, int, void *, struct nameidata *); 111 static int mount_getargs(struct lwp *, struct vnode *, const char *, int, 112 void *, struct nameidata *); 113 114 int dovfsusermount = 0; 115 116 /* 117 * Virtual File System System Calls 118 */ 119 120 /* 121 * Mount a file system. 122 */ 123 124 #if defined(COMPAT_09) || defined(COMPAT_43) 125 /* 126 * This table is used to maintain compatibility with 4.3BSD 127 * and NetBSD 0.9 mount syscalls. Note, the order is important! 128 * 129 * Do not modify this table. It should only contain filesystems 130 * supported by NetBSD 0.9 and 4.3BSD. 131 */ 132 const char * const mountcompatnames[] = { 133 NULL, /* 0 = MOUNT_NONE */ 134 MOUNT_FFS, /* 1 = MOUNT_UFS */ 135 MOUNT_NFS, /* 2 */ 136 MOUNT_MFS, /* 3 */ 137 MOUNT_MSDOS, /* 4 */ 138 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 139 MOUNT_FDESC, /* 6 */ 140 MOUNT_KERNFS, /* 7 */ 141 NULL, /* 8 = MOUNT_DEVFS */ 142 MOUNT_AFS, /* 9 */ 143 }; 144 const int nmountcompatnames = sizeof(mountcompatnames) / 145 sizeof(mountcompatnames[0]); 146 #endif /* COMPAT_09 || COMPAT_43 */ 147 148 static int 149 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 150 void *data, struct nameidata *ndp) 151 { 152 struct mount *mp; 153 int error = 0, saved_flags; 154 155 mp = vp->v_mount; 156 saved_flags = mp->mnt_flag; 157 158 /* We can't operate on VROOT here. */ 159 if ((vp->v_flag & VROOT) == 0) { 160 vput(vp); 161 error = EINVAL; 162 goto out; 163 } 164 165 /* 166 * We only allow the filesystem to be reloaded if it 167 * is currently mounted read-only. 168 */ 169 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 170 error = EOPNOTSUPP; /* Needs translation */ 171 goto out; 172 } 173 174 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 175 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 176 if (error) 177 goto out; 178 179 if (vfs_busy(mp, LK_NOWAIT, 0)) { 180 error = EPERM; 181 goto out; 182 } 183 184 mp->mnt_flag &= ~MNT_OP_FLAGS; 185 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 186 187 /* 188 * Set the mount level flags. 189 */ 190 if (flags & MNT_RDONLY) 191 mp->mnt_flag |= MNT_RDONLY; 192 else if (mp->mnt_flag & MNT_RDONLY) 193 mp->mnt_iflag |= IMNT_WANTRDWR; 194 mp->mnt_flag &= 195 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 196 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 197 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 198 mp->mnt_flag |= flags & 199 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 200 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 201 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 202 MNT_IGNORE); 203 204 error = VFS_MOUNT(mp, path, data, ndp, l); 205 206 #if defined(COMPAT_30) && defined(NFSSERVER) 207 if (error) { 208 int error2; 209 210 /* Update failed; let's try and see if it was an 211 * export request. */ 212 error2 = nfs_update_exports_30(mp, path, data, l); 213 214 /* Only update error code if the export request was 215 * understood but some problem occurred while 216 * processing it. */ 217 if (error2 != EJUSTRETURN) 218 error = error2; 219 } 220 #endif 221 if (mp->mnt_iflag & IMNT_WANTRDWR) 222 mp->mnt_flag &= ~MNT_RDONLY; 223 if (error) 224 mp->mnt_flag = saved_flags; 225 mp->mnt_flag &= ~MNT_OP_FLAGS; 226 mp->mnt_iflag &= ~IMNT_WANTRDWR; 227 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 228 if (mp->mnt_syncer == NULL) 229 error = vfs_allocate_syncvnode(mp); 230 } else { 231 if (mp->mnt_syncer != NULL) 232 vfs_deallocate_syncvnode(mp); 233 } 234 vfs_unbusy(mp); 235 236 out: 237 return (error); 238 } 239 240 static int 241 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 242 const char *path, int flags, void *data, struct nameidata *ndp) 243 { 244 struct mount *mp = NULL; 245 struct vattr va; 246 char fstypename[MFSNAMELEN]; 247 int error; 248 249 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 250 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 251 if (error) { 252 vput(vp); 253 goto out; 254 } 255 256 /* Can't make a non-dir a mount-point (from here anyway). */ 257 if (vp->v_type != VDIR) { 258 error = ENOTDIR; 259 vput(vp); 260 goto out; 261 } 262 263 /* 264 * If the user is not root, ensure that they own the directory 265 * onto which we are attempting to mount. 266 */ 267 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 268 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 269 (error = kauth_authorize_generic(l->l_cred, 270 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 271 vput(vp); 272 goto out; 273 } 274 275 if (flags & MNT_EXPORTED) { 276 error = EINVAL; 277 vput(vp); 278 goto out; 279 } 280 281 /* 282 * Copy file-system type from userspace. 283 */ 284 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 285 if (error) { 286 #if defined(COMPAT_09) || defined(COMPAT_43) 287 /* 288 * Historically, filesystem types were identified by numbers. 289 * If we get an integer for the filesystem type instead of a 290 * string, we check to see if it matches one of the historic 291 * filesystem types. 292 */ 293 u_long fsindex = (u_long)fstype; 294 if (fsindex >= nmountcompatnames || 295 mountcompatnames[fsindex] == NULL) { 296 error = ENODEV; 297 vput(vp); 298 goto out; 299 } 300 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 301 #else 302 vput(vp); 303 goto out; 304 #endif 305 } 306 307 #ifdef COMPAT_10 308 /* Accept `ufs' as an alias for `ffs'. */ 309 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 310 strlcpy(fstypename, "ffs", sizeof(fstypename)); 311 #endif 312 313 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 314 vput(vp); 315 goto out; 316 } 317 318 /* 319 * Check if a file-system is not already mounted on this vnode. 320 */ 321 if (vp->v_mountedhere != NULL) { 322 error = EBUSY; 323 vput(vp); 324 goto out; 325 } 326 327 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 328 329 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 330 free(mp, M_MOUNT); 331 error = ENODEV; 332 vput(vp); 333 goto out; 334 } 335 336 TAILQ_INIT(&mp->mnt_vnodelist); 337 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 338 simple_lock_init(&mp->mnt_slock); 339 (void)vfs_busy(mp, LK_NOWAIT, 0); 340 341 mp->mnt_op->vfs_refcount++; 342 mp->mnt_vnodecovered = vp; 343 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 344 mp->mnt_unmounter = NULL; 345 mp->mnt_leaf = mp; 346 mount_initspecific(mp); 347 348 /* 349 * The underlying file system may refuse the mount for 350 * various reasons. Allow the user to force it to happen. 351 * 352 * Set the mount level flags. 353 */ 354 mp->mnt_flag = flags & 355 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 356 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 357 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 358 MNT_IGNORE | MNT_RDONLY); 359 360 error = VFS_MOUNT(mp, path, data, ndp, l); 361 mp->mnt_flag &= ~MNT_OP_FLAGS; 362 363 /* 364 * Put the new filesystem on the mount list after root. 365 */ 366 cache_purge(vp); 367 if (!error) { 368 mp->mnt_iflag &= ~IMNT_WANTRDWR; 369 vp->v_mountedhere = mp; 370 simple_lock(&mountlist_slock); 371 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 372 simple_unlock(&mountlist_slock); 373 VOP_UNLOCK(vp, 0); 374 checkdirs(vp); 375 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 376 error = vfs_allocate_syncvnode(mp); 377 vfs_unbusy(mp); 378 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 379 error = VFS_START(mp, 0, l); 380 if (error) 381 vrele(vp); 382 } else { 383 vp->v_mountedhere = NULL; 384 mp->mnt_op->vfs_refcount--; 385 vfs_unbusy(mp); 386 free(mp, M_MOUNT); 387 vput(vp); 388 } 389 390 out: 391 return (error); 392 } 393 394 static int 395 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 396 void *data, struct nameidata *ndp) 397 { 398 struct mount *mp; 399 int error; 400 401 /* If MNT_GETARGS is specified, it should be the only flag. */ 402 if (flags & ~MNT_GETARGS) { 403 error = EINVAL; 404 goto out; 405 } 406 407 mp = vp->v_mount; 408 409 /* XXX: probably some notion of "can see" here if we want isolation. */ 410 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 411 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 412 if (error) 413 goto out; 414 415 if ((vp->v_flag & VROOT) == 0) { 416 error = EINVAL; 417 goto out; 418 } 419 420 if (vfs_busy(mp, LK_NOWAIT, 0)) { 421 error = EPERM; 422 goto out; 423 } 424 425 mp->mnt_flag &= ~MNT_OP_FLAGS; 426 mp->mnt_flag |= MNT_GETARGS; 427 error = VFS_MOUNT(mp, path, data, ndp, l); 428 mp->mnt_flag &= ~MNT_OP_FLAGS; 429 430 vfs_unbusy(mp); 431 out: 432 return (error); 433 } 434 435 /* ARGSUSED */ 436 int 437 sys_mount(struct lwp *l, void *v, register_t *retval) 438 { 439 struct sys_mount_args /* { 440 syscallarg(const char *) type; 441 syscallarg(const char *) path; 442 syscallarg(int) flags; 443 syscallarg(void *) data; 444 } */ *uap = v; 445 struct vnode *vp; 446 struct nameidata nd; 447 int error; 448 449 /* 450 * Get vnode to be covered 451 */ 452 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 453 SCARG(uap, path), l); 454 if ((error = namei(&nd)) != 0) 455 return (error); 456 vp = nd.ni_vp; 457 458 /* 459 * A lookup in VFS_MOUNT might result in an attempt to 460 * lock this vnode again, so make the lock recursive. 461 */ 462 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 463 464 if (SCARG(uap, flags) & MNT_GETARGS) { 465 error = mount_getargs(l, vp, SCARG(uap, path), 466 SCARG(uap, flags), SCARG(uap, data), &nd); 467 vput(vp); 468 } else if (SCARG(uap, flags) & MNT_UPDATE) { 469 error = mount_update(l, vp, SCARG(uap, path), 470 SCARG(uap, flags), SCARG(uap, data), &nd); 471 vput(vp); 472 } else { 473 /* Locking is handled internally in mount_domount(). */ 474 error = mount_domount(l, vp, SCARG(uap, type), 475 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 476 } 477 478 return (error); 479 } 480 481 /* 482 * Scan all active processes to see if any of them have a current 483 * or root directory onto which the new filesystem has just been 484 * mounted. If so, replace them with the new mount point. 485 */ 486 void 487 checkdirs(struct vnode *olddp) 488 { 489 struct cwdinfo *cwdi; 490 struct vnode *newdp; 491 struct proc *p; 492 493 if (olddp->v_usecount == 1) 494 return; 495 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 496 panic("mount: lost mount"); 497 mutex_enter(&proclist_lock); 498 PROCLIST_FOREACH(p, &allproc) { 499 cwdi = p->p_cwdi; 500 if (!cwdi) 501 continue; 502 if (cwdi->cwdi_cdir == olddp) { 503 vrele(cwdi->cwdi_cdir); 504 VREF(newdp); 505 cwdi->cwdi_cdir = newdp; 506 } 507 if (cwdi->cwdi_rdir == olddp) { 508 vrele(cwdi->cwdi_rdir); 509 VREF(newdp); 510 cwdi->cwdi_rdir = newdp; 511 } 512 } 513 mutex_exit(&proclist_lock); 514 if (rootvnode == olddp) { 515 vrele(rootvnode); 516 VREF(newdp); 517 rootvnode = newdp; 518 } 519 vput(newdp); 520 } 521 522 /* 523 * Unmount a file system. 524 * 525 * Note: unmount takes a path to the vnode mounted on as argument, 526 * not special file (as before). 527 */ 528 /* ARGSUSED */ 529 int 530 sys_unmount(struct lwp *l, void *v, register_t *retval) 531 { 532 struct sys_unmount_args /* { 533 syscallarg(const char *) path; 534 syscallarg(int) flags; 535 } */ *uap = v; 536 struct vnode *vp; 537 struct mount *mp; 538 int error; 539 struct nameidata nd; 540 541 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 542 SCARG(uap, path), l); 543 if ((error = namei(&nd)) != 0) 544 return (error); 545 vp = nd.ni_vp; 546 mp = vp->v_mount; 547 548 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 549 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 550 if (error) { 551 vput(vp); 552 return (error); 553 } 554 555 /* 556 * Don't allow unmounting the root file system. 557 */ 558 if (mp->mnt_flag & MNT_ROOTFS) { 559 vput(vp); 560 return (EINVAL); 561 } 562 563 /* 564 * Must be the root of the filesystem 565 */ 566 if ((vp->v_flag & VROOT) == 0) { 567 vput(vp); 568 return (EINVAL); 569 } 570 vput(vp); 571 572 /* 573 * XXX Freeze syncer. Must do this before locking the 574 * mount point. See dounmount() for details. 575 */ 576 mutex_enter(&syncer_mutex); 577 578 if (vfs_busy(mp, 0, 0)) { 579 mutex_exit(&syncer_mutex); 580 return (EBUSY); 581 } 582 583 return (dounmount(mp, SCARG(uap, flags), l)); 584 } 585 586 /* 587 * Do the actual file system unmount. File system is assumed to have been 588 * marked busy by the caller. 589 */ 590 int 591 dounmount(struct mount *mp, int flags, struct lwp *l) 592 { 593 struct vnode *coveredvp; 594 int error; 595 int async; 596 int used_syncer; 597 598 #if NVERIEXEC > 0 599 error = veriexec_unmountchk(mp); 600 if (error) 601 return (error); 602 #endif /* NVERIEXEC > 0 */ 603 604 simple_lock(&mountlist_slock); 605 vfs_unbusy(mp); 606 used_syncer = (mp->mnt_syncer != NULL); 607 608 /* 609 * XXX Syncer must be frozen when we get here. This should really 610 * be done on a per-mountpoint basis, but especially the softdep 611 * code possibly called from the syncer doesn't exactly work on a 612 * per-mountpoint basis, so the softdep code would become a maze 613 * of vfs_busy() calls. 614 * 615 * The caller of dounmount() must acquire syncer_mutex because 616 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 617 * order, and we must preserve that order to avoid deadlock. 618 * 619 * So, if the file system did not use the syncer, now is 620 * the time to release the syncer_mutex. 621 */ 622 if (used_syncer == 0) 623 mutex_exit(&syncer_mutex); 624 625 mp->mnt_iflag |= IMNT_UNMOUNT; 626 mp->mnt_unmounter = l; 627 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 628 629 async = mp->mnt_flag & MNT_ASYNC; 630 mp->mnt_flag &= ~MNT_ASYNC; 631 cache_purgevfs(mp); /* remove cache entries for this file sys */ 632 if (mp->mnt_syncer != NULL) 633 vfs_deallocate_syncvnode(mp); 634 error = 0; 635 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 636 #if NFSS > 0 637 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 638 #endif 639 if (error == 0) 640 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 641 } 642 if (error == 0 || (flags & MNT_FORCE)) 643 error = VFS_UNMOUNT(mp, flags, l); 644 simple_lock(&mountlist_slock); 645 if (error) { 646 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 647 (void) vfs_allocate_syncvnode(mp); 648 mp->mnt_iflag &= ~IMNT_UNMOUNT; 649 mp->mnt_unmounter = NULL; 650 mp->mnt_flag |= async; 651 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 652 &mountlist_slock); 653 if (used_syncer) 654 mutex_exit(&syncer_mutex); 655 simple_lock(&mp->mnt_slock); 656 while (mp->mnt_wcnt > 0) { 657 wakeup(mp); 658 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 659 0, &mp->mnt_slock); 660 } 661 simple_unlock(&mp->mnt_slock); 662 return (error); 663 } 664 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 665 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 666 coveredvp->v_mountedhere = NULL; 667 mp->mnt_op->vfs_refcount--; 668 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 669 panic("unmount: dangling vnode"); 670 mp->mnt_iflag |= IMNT_GONE; 671 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 672 if (coveredvp != NULLVP) 673 vrele(coveredvp); 674 mount_finispecific(mp); 675 if (used_syncer) 676 mutex_exit(&syncer_mutex); 677 simple_lock(&mp->mnt_slock); 678 while (mp->mnt_wcnt > 0) { 679 wakeup(mp); 680 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 681 } 682 simple_unlock(&mp->mnt_slock); 683 vfs_hooks_unmount(mp); 684 free(mp, M_MOUNT); 685 return (0); 686 } 687 688 /* 689 * Sync each mounted filesystem. 690 */ 691 #ifdef DEBUG 692 int syncprt = 0; 693 struct ctldebug debug0 = { "syncprt", &syncprt }; 694 #endif 695 696 /* ARGSUSED */ 697 int 698 sys_sync(struct lwp *l, void *v, register_t *retval) 699 { 700 struct mount *mp, *nmp; 701 int asyncflag; 702 703 if (l == NULL) 704 l = &lwp0; 705 706 simple_lock(&mountlist_slock); 707 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 708 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 709 nmp = mp->mnt_list.cqe_prev; 710 continue; 711 } 712 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 713 asyncflag = mp->mnt_flag & MNT_ASYNC; 714 mp->mnt_flag &= ~MNT_ASYNC; 715 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 716 if (asyncflag) 717 mp->mnt_flag |= MNT_ASYNC; 718 } 719 simple_lock(&mountlist_slock); 720 nmp = mp->mnt_list.cqe_prev; 721 vfs_unbusy(mp); 722 723 } 724 simple_unlock(&mountlist_slock); 725 #ifdef DEBUG 726 if (syncprt) 727 vfs_bufstats(); 728 #endif /* DEBUG */ 729 return (0); 730 } 731 732 /* 733 * Change filesystem quotas. 734 */ 735 /* ARGSUSED */ 736 int 737 sys_quotactl(struct lwp *l, void *v, register_t *retval) 738 { 739 struct sys_quotactl_args /* { 740 syscallarg(const char *) path; 741 syscallarg(int) cmd; 742 syscallarg(int) uid; 743 syscallarg(void *) arg; 744 } */ *uap = v; 745 struct mount *mp; 746 int error; 747 struct nameidata nd; 748 749 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 750 if ((error = namei(&nd)) != 0) 751 return (error); 752 mp = nd.ni_vp->v_mount; 753 vrele(nd.ni_vp); 754 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 755 SCARG(uap, arg), l); 756 return (error); 757 } 758 759 int 760 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 761 int root) 762 { 763 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 764 int error = 0; 765 766 /* 767 * If MNT_NOWAIT or MNT_LAZY is specified, do not 768 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 769 * overrides MNT_NOWAIT. 770 */ 771 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 772 (flags != MNT_WAIT && flags != 0)) { 773 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 774 goto done; 775 } 776 777 /* Get the filesystem stats now */ 778 memset(sp, 0, sizeof(*sp)); 779 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 780 return error; 781 } 782 783 if (cwdi->cwdi_rdir == NULL) 784 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 785 done: 786 if (cwdi->cwdi_rdir != NULL) { 787 size_t len; 788 char *bp; 789 char *path = PNBUF_GET(); 790 if (!path) 791 return ENOMEM; 792 793 bp = path + MAXPATHLEN; 794 *--bp = '\0'; 795 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 796 MAXPATHLEN / 2, 0, l); 797 if (error) { 798 PNBUF_PUT(path); 799 return error; 800 } 801 len = strlen(bp); 802 /* 803 * for mount points that are below our root, we can see 804 * them, so we fix up the pathname and return them. The 805 * rest we cannot see, so we don't allow viewing the 806 * data. 807 */ 808 if (strncmp(bp, sp->f_mntonname, len) == 0) { 809 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 810 sizeof(sp->f_mntonname)); 811 if (sp->f_mntonname[0] == '\0') 812 (void)strlcpy(sp->f_mntonname, "/", 813 sizeof(sp->f_mntonname)); 814 } else { 815 if (root) 816 (void)strlcpy(sp->f_mntonname, "/", 817 sizeof(sp->f_mntonname)); 818 else 819 error = EPERM; 820 } 821 PNBUF_PUT(path); 822 } 823 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 824 return error; 825 } 826 827 /* 828 * Get filesystem statistics. 829 */ 830 /* ARGSUSED */ 831 int 832 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 833 { 834 struct sys_statvfs1_args /* { 835 syscallarg(const char *) path; 836 syscallarg(struct statvfs *) buf; 837 syscallarg(int) flags; 838 } */ *uap = v; 839 struct mount *mp; 840 struct statvfs *sb; 841 int error; 842 struct nameidata nd; 843 844 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 845 if ((error = namei(&nd)) != 0) 846 return error; 847 mp = nd.ni_vp->v_mount; 848 vrele(nd.ni_vp); 849 sb = STATVFSBUF_GET(); 850 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 851 if (error == 0) { 852 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 853 } 854 STATVFSBUF_PUT(sb); 855 return error; 856 } 857 858 /* 859 * Get filesystem statistics. 860 */ 861 /* ARGSUSED */ 862 int 863 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 864 { 865 struct sys_fstatvfs1_args /* { 866 syscallarg(int) fd; 867 syscallarg(struct statvfs *) buf; 868 syscallarg(int) flags; 869 } */ *uap = v; 870 struct proc *p = l->l_proc; 871 struct file *fp; 872 struct mount *mp; 873 struct statvfs *sb; 874 int error; 875 876 /* getvnode() will use the descriptor for us */ 877 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 878 return (error); 879 mp = ((struct vnode *)fp->f_data)->v_mount; 880 sb = STATVFSBUF_GET(); 881 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 882 goto out; 883 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 884 out: 885 FILE_UNUSE(fp, l); 886 STATVFSBUF_PUT(sb); 887 return error; 888 } 889 890 891 /* 892 * Get statistics on all filesystems. 893 */ 894 int 895 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 896 { 897 struct sys_getvfsstat_args /* { 898 syscallarg(struct statvfs *) buf; 899 syscallarg(size_t) bufsize; 900 syscallarg(int) flags; 901 } */ *uap = v; 902 int root = 0; 903 struct proc *p = l->l_proc; 904 struct mount *mp, *nmp; 905 struct statvfs *sb; 906 struct statvfs *sfsp; 907 size_t count, maxcount; 908 int error = 0; 909 910 sb = STATVFSBUF_GET(); 911 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 912 sfsp = SCARG(uap, buf); 913 simple_lock(&mountlist_slock); 914 count = 0; 915 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 916 mp = nmp) { 917 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 918 nmp = CIRCLEQ_NEXT(mp, mnt_list); 919 continue; 920 } 921 if (sfsp && count < maxcount) { 922 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 923 if (error) { 924 simple_lock(&mountlist_slock); 925 nmp = CIRCLEQ_NEXT(mp, mnt_list); 926 vfs_unbusy(mp); 927 continue; 928 } 929 error = copyout(sb, sfsp, sizeof(*sfsp)); 930 if (error) { 931 vfs_unbusy(mp); 932 goto out; 933 } 934 sfsp++; 935 root |= strcmp(sb->f_mntonname, "/") == 0; 936 } 937 count++; 938 simple_lock(&mountlist_slock); 939 nmp = CIRCLEQ_NEXT(mp, mnt_list); 940 vfs_unbusy(mp); 941 } 942 simple_unlock(&mountlist_slock); 943 if (root == 0 && p->p_cwdi->cwdi_rdir) { 944 /* 945 * fake a root entry 946 */ 947 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 948 SCARG(uap, flags), 1)) != 0) 949 goto out; 950 if (sfsp) 951 error = copyout(sb, sfsp, sizeof(*sfsp)); 952 count++; 953 } 954 if (sfsp && count > maxcount) 955 *retval = maxcount; 956 else 957 *retval = count; 958 out: 959 STATVFSBUF_PUT(sb); 960 return error; 961 } 962 963 /* 964 * Change current working directory to a given file descriptor. 965 */ 966 /* ARGSUSED */ 967 int 968 sys_fchdir(struct lwp *l, void *v, register_t *retval) 969 { 970 struct sys_fchdir_args /* { 971 syscallarg(int) fd; 972 } */ *uap = v; 973 struct proc *p = l->l_proc; 974 struct filedesc *fdp = p->p_fd; 975 struct cwdinfo *cwdi = p->p_cwdi; 976 struct vnode *vp, *tdp; 977 struct mount *mp; 978 struct file *fp; 979 int error; 980 981 /* getvnode() will use the descriptor for us */ 982 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 983 return (error); 984 vp = (struct vnode *)fp->f_data; 985 986 VREF(vp); 987 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 988 if (vp->v_type != VDIR) 989 error = ENOTDIR; 990 else 991 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 992 if (error) { 993 vput(vp); 994 goto out; 995 } 996 while ((mp = vp->v_mountedhere) != NULL) { 997 if (vfs_busy(mp, 0, 0)) 998 continue; 999 1000 vput(vp); 1001 error = VFS_ROOT(mp, &tdp); 1002 vfs_unbusy(mp); 1003 if (error) 1004 goto out; 1005 vp = tdp; 1006 } 1007 VOP_UNLOCK(vp, 0); 1008 1009 /* 1010 * Disallow changing to a directory not under the process's 1011 * current root directory (if there is one). 1012 */ 1013 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1014 vrele(vp); 1015 error = EPERM; /* operation not permitted */ 1016 goto out; 1017 } 1018 1019 vrele(cwdi->cwdi_cdir); 1020 cwdi->cwdi_cdir = vp; 1021 out: 1022 FILE_UNUSE(fp, l); 1023 return (error); 1024 } 1025 1026 /* 1027 * Change this process's notion of the root directory to a given file 1028 * descriptor. 1029 */ 1030 int 1031 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1032 { 1033 struct sys_fchroot_args *uap = v; 1034 struct proc *p = l->l_proc; 1035 struct filedesc *fdp = p->p_fd; 1036 struct cwdinfo *cwdi = p->p_cwdi; 1037 struct vnode *vp; 1038 struct file *fp; 1039 int error; 1040 1041 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1042 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1043 return error; 1044 /* getvnode() will use the descriptor for us */ 1045 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1046 return error; 1047 vp = (struct vnode *) fp->f_data; 1048 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1049 if (vp->v_type != VDIR) 1050 error = ENOTDIR; 1051 else 1052 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1053 VOP_UNLOCK(vp, 0); 1054 if (error) 1055 goto out; 1056 VREF(vp); 1057 1058 /* 1059 * Prevent escaping from chroot by putting the root under 1060 * the working directory. Silently chdir to / if we aren't 1061 * already there. 1062 */ 1063 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1064 /* 1065 * XXX would be more failsafe to change directory to a 1066 * deadfs node here instead 1067 */ 1068 vrele(cwdi->cwdi_cdir); 1069 VREF(vp); 1070 cwdi->cwdi_cdir = vp; 1071 } 1072 1073 if (cwdi->cwdi_rdir != NULL) 1074 vrele(cwdi->cwdi_rdir); 1075 cwdi->cwdi_rdir = vp; 1076 out: 1077 FILE_UNUSE(fp, l); 1078 return (error); 1079 } 1080 1081 /* 1082 * Change current working directory (``.''). 1083 */ 1084 /* ARGSUSED */ 1085 int 1086 sys_chdir(struct lwp *l, void *v, register_t *retval) 1087 { 1088 struct sys_chdir_args /* { 1089 syscallarg(const char *) path; 1090 } */ *uap = v; 1091 struct proc *p = l->l_proc; 1092 struct cwdinfo *cwdi = p->p_cwdi; 1093 int error; 1094 struct nameidata nd; 1095 1096 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1097 SCARG(uap, path), l); 1098 if ((error = change_dir(&nd, l)) != 0) 1099 return (error); 1100 vrele(cwdi->cwdi_cdir); 1101 cwdi->cwdi_cdir = nd.ni_vp; 1102 return (0); 1103 } 1104 1105 /* 1106 * Change notion of root (``/'') directory. 1107 */ 1108 /* ARGSUSED */ 1109 int 1110 sys_chroot(struct lwp *l, void *v, register_t *retval) 1111 { 1112 struct sys_chroot_args /* { 1113 syscallarg(const char *) path; 1114 } */ *uap = v; 1115 struct proc *p = l->l_proc; 1116 struct cwdinfo *cwdi = p->p_cwdi; 1117 struct vnode *vp; 1118 int error; 1119 struct nameidata nd; 1120 1121 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1122 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1123 return (error); 1124 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1125 SCARG(uap, path), l); 1126 if ((error = change_dir(&nd, l)) != 0) 1127 return (error); 1128 if (cwdi->cwdi_rdir != NULL) 1129 vrele(cwdi->cwdi_rdir); 1130 vp = nd.ni_vp; 1131 cwdi->cwdi_rdir = vp; 1132 1133 /* 1134 * Prevent escaping from chroot by putting the root under 1135 * the working directory. Silently chdir to / if we aren't 1136 * already there. 1137 */ 1138 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1139 /* 1140 * XXX would be more failsafe to change directory to a 1141 * deadfs node here instead 1142 */ 1143 vrele(cwdi->cwdi_cdir); 1144 VREF(vp); 1145 cwdi->cwdi_cdir = vp; 1146 } 1147 1148 return (0); 1149 } 1150 1151 /* 1152 * Common routine for chroot and chdir. 1153 */ 1154 static int 1155 change_dir(struct nameidata *ndp, struct lwp *l) 1156 { 1157 struct vnode *vp; 1158 int error; 1159 1160 if ((error = namei(ndp)) != 0) 1161 return (error); 1162 vp = ndp->ni_vp; 1163 if (vp->v_type != VDIR) 1164 error = ENOTDIR; 1165 else 1166 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1167 1168 if (error) 1169 vput(vp); 1170 else 1171 VOP_UNLOCK(vp, 0); 1172 return (error); 1173 } 1174 1175 /* 1176 * Check permissions, allocate an open file structure, 1177 * and call the device open routine if any. 1178 */ 1179 int 1180 sys_open(struct lwp *l, void *v, register_t *retval) 1181 { 1182 struct sys_open_args /* { 1183 syscallarg(const char *) path; 1184 syscallarg(int) flags; 1185 syscallarg(int) mode; 1186 } */ *uap = v; 1187 struct proc *p = l->l_proc; 1188 struct cwdinfo *cwdi = p->p_cwdi; 1189 struct filedesc *fdp = p->p_fd; 1190 struct file *fp; 1191 struct vnode *vp; 1192 int flags, cmode; 1193 int type, indx, error; 1194 struct flock lf; 1195 struct nameidata nd; 1196 1197 flags = FFLAGS(SCARG(uap, flags)); 1198 if ((flags & (FREAD | FWRITE)) == 0) 1199 return (EINVAL); 1200 /* falloc() will use the file descriptor for us */ 1201 if ((error = falloc(l, &fp, &indx)) != 0) 1202 return (error); 1203 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1204 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1205 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1206 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1207 FILE_UNUSE(fp, l); 1208 fdp->fd_ofiles[indx] = NULL; 1209 ffree(fp); 1210 if ((error == EDUPFD || error == EMOVEFD) && 1211 l->l_dupfd >= 0 && /* XXX from fdopen */ 1212 (error = 1213 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1214 *retval = indx; 1215 return (0); 1216 } 1217 if (error == ERESTART) 1218 error = EINTR; 1219 fdremove(fdp, indx); 1220 return (error); 1221 } 1222 l->l_dupfd = 0; 1223 vp = nd.ni_vp; 1224 fp->f_flag = flags & FMASK; 1225 fp->f_type = DTYPE_VNODE; 1226 fp->f_ops = &vnops; 1227 fp->f_data = vp; 1228 if (flags & (O_EXLOCK | O_SHLOCK)) { 1229 lf.l_whence = SEEK_SET; 1230 lf.l_start = 0; 1231 lf.l_len = 0; 1232 if (flags & O_EXLOCK) 1233 lf.l_type = F_WRLCK; 1234 else 1235 lf.l_type = F_RDLCK; 1236 type = F_FLOCK; 1237 if ((flags & FNONBLOCK) == 0) 1238 type |= F_WAIT; 1239 VOP_UNLOCK(vp, 0); 1240 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1241 if (error) { 1242 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1243 FILE_UNUSE(fp, l); 1244 ffree(fp); 1245 fdremove(fdp, indx); 1246 return (error); 1247 } 1248 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1249 fp->f_flag |= FHASLOCK; 1250 } 1251 VOP_UNLOCK(vp, 0); 1252 *retval = indx; 1253 FILE_SET_MATURE(fp); 1254 FILE_UNUSE(fp, l); 1255 return (0); 1256 } 1257 1258 static void 1259 vfs__fhfree(fhandle_t *fhp) 1260 { 1261 size_t fhsize; 1262 1263 if (fhp == NULL) { 1264 return; 1265 } 1266 fhsize = FHANDLE_SIZE(fhp); 1267 kmem_free(fhp, fhsize); 1268 } 1269 1270 /* 1271 * vfs_composefh: compose a filehandle. 1272 */ 1273 1274 int 1275 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1276 { 1277 struct mount *mp; 1278 struct fid *fidp; 1279 int error; 1280 size_t needfhsize; 1281 size_t fidsize; 1282 1283 mp = vp->v_mount; 1284 fidp = NULL; 1285 if (*fh_size < FHANDLE_SIZE_MIN) { 1286 fidsize = 0; 1287 } else { 1288 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1289 if (fhp != NULL) { 1290 memset(fhp, 0, *fh_size); 1291 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1292 fidp = &fhp->fh_fid; 1293 } 1294 } 1295 error = VFS_VPTOFH(vp, fidp, &fidsize); 1296 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1297 if (error == 0 && *fh_size < needfhsize) { 1298 error = E2BIG; 1299 } 1300 *fh_size = needfhsize; 1301 return error; 1302 } 1303 1304 int 1305 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1306 { 1307 struct mount *mp; 1308 fhandle_t *fhp; 1309 size_t fhsize; 1310 size_t fidsize; 1311 int error; 1312 1313 *fhpp = NULL; 1314 mp = vp->v_mount; 1315 fidsize = 0; 1316 error = VFS_VPTOFH(vp, NULL, &fidsize); 1317 KASSERT(error != 0); 1318 if (error != E2BIG) { 1319 goto out; 1320 } 1321 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1322 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1323 if (fhp == NULL) { 1324 error = ENOMEM; 1325 goto out; 1326 } 1327 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1328 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1329 if (error == 0) { 1330 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1331 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1332 *fhpp = fhp; 1333 } else { 1334 kmem_free(fhp, fhsize); 1335 } 1336 out: 1337 return error; 1338 } 1339 1340 void 1341 vfs_composefh_free(fhandle_t *fhp) 1342 { 1343 1344 vfs__fhfree(fhp); 1345 } 1346 1347 /* 1348 * vfs_fhtovp: lookup a vnode by a filehandle. 1349 */ 1350 1351 int 1352 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1353 { 1354 struct mount *mp; 1355 int error; 1356 1357 *vpp = NULL; 1358 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1359 if (mp == NULL) { 1360 error = ESTALE; 1361 goto out; 1362 } 1363 if (mp->mnt_op->vfs_fhtovp == NULL) { 1364 error = EOPNOTSUPP; 1365 goto out; 1366 } 1367 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1368 out: 1369 return error; 1370 } 1371 1372 /* 1373 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1374 * the needed size. 1375 */ 1376 1377 int 1378 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1379 { 1380 fhandle_t *fhp; 1381 int error; 1382 1383 *fhpp = NULL; 1384 if (fhsize > FHANDLE_SIZE_MAX) { 1385 return EINVAL; 1386 } 1387 if (fhsize < FHANDLE_SIZE_MIN) { 1388 return EINVAL; 1389 } 1390 again: 1391 fhp = kmem_alloc(fhsize, KM_SLEEP); 1392 if (fhp == NULL) { 1393 return ENOMEM; 1394 } 1395 error = copyin(ufhp, fhp, fhsize); 1396 if (error == 0) { 1397 /* XXX this check shouldn't be here */ 1398 if (FHANDLE_SIZE(fhp) == fhsize) { 1399 *fhpp = fhp; 1400 return 0; 1401 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1402 /* 1403 * a kludge for nfsv2 padded handles. 1404 */ 1405 size_t sz; 1406 1407 sz = FHANDLE_SIZE(fhp); 1408 kmem_free(fhp, fhsize); 1409 fhsize = sz; 1410 goto again; 1411 } else { 1412 /* 1413 * userland told us wrong size. 1414 */ 1415 error = EINVAL; 1416 } 1417 } 1418 kmem_free(fhp, fhsize); 1419 return error; 1420 } 1421 1422 void 1423 vfs_copyinfh_free(fhandle_t *fhp) 1424 { 1425 1426 vfs__fhfree(fhp); 1427 } 1428 1429 /* 1430 * Get file handle system call 1431 */ 1432 int 1433 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1434 { 1435 struct sys___getfh30_args /* { 1436 syscallarg(char *) fname; 1437 syscallarg(fhandle_t *) fhp; 1438 syscallarg(size_t *) fh_size; 1439 } */ *uap = v; 1440 struct vnode *vp; 1441 fhandle_t *fh; 1442 int error; 1443 struct nameidata nd; 1444 size_t sz; 1445 size_t usz; 1446 1447 /* 1448 * Must be super user 1449 */ 1450 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1451 0, NULL, NULL, NULL); 1452 if (error) 1453 return (error); 1454 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1455 SCARG(uap, fname), l); 1456 error = namei(&nd); 1457 if (error) 1458 return (error); 1459 vp = nd.ni_vp; 1460 error = vfs_composefh_alloc(vp, &fh); 1461 vput(vp); 1462 if (error != 0) { 1463 goto out; 1464 } 1465 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1466 if (error != 0) { 1467 goto out; 1468 } 1469 sz = FHANDLE_SIZE(fh); 1470 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1471 if (error != 0) { 1472 goto out; 1473 } 1474 if (usz >= sz) { 1475 error = copyout(fh, SCARG(uap, fhp), sz); 1476 } else { 1477 error = E2BIG; 1478 } 1479 out: 1480 vfs_composefh_free(fh); 1481 return (error); 1482 } 1483 1484 /* 1485 * Open a file given a file handle. 1486 * 1487 * Check permissions, allocate an open file structure, 1488 * and call the device open routine if any. 1489 */ 1490 1491 int 1492 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1493 register_t *retval) 1494 { 1495 struct filedesc *fdp = l->l_proc->p_fd; 1496 struct file *fp; 1497 struct vnode *vp = NULL; 1498 kauth_cred_t cred = l->l_cred; 1499 struct file *nfp; 1500 int type, indx, error=0; 1501 struct flock lf; 1502 struct vattr va; 1503 fhandle_t *fh; 1504 int flags; 1505 1506 /* 1507 * Must be super user 1508 */ 1509 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1510 0, NULL, NULL, NULL))) 1511 return (error); 1512 1513 flags = FFLAGS(oflags); 1514 if ((flags & (FREAD | FWRITE)) == 0) 1515 return (EINVAL); 1516 if ((flags & O_CREAT)) 1517 return (EINVAL); 1518 /* falloc() will use the file descriptor for us */ 1519 if ((error = falloc(l, &nfp, &indx)) != 0) 1520 return (error); 1521 fp = nfp; 1522 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1523 if (error != 0) { 1524 goto bad; 1525 } 1526 error = vfs_fhtovp(fh, &vp); 1527 if (error != 0) { 1528 goto bad; 1529 } 1530 1531 /* Now do an effective vn_open */ 1532 1533 if (vp->v_type == VSOCK) { 1534 error = EOPNOTSUPP; 1535 goto bad; 1536 } 1537 if (flags & FREAD) { 1538 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1539 goto bad; 1540 } 1541 if (flags & (FWRITE | O_TRUNC)) { 1542 if (vp->v_type == VDIR) { 1543 error = EISDIR; 1544 goto bad; 1545 } 1546 if ((error = vn_writechk(vp)) != 0 || 1547 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1548 goto bad; 1549 } 1550 if (flags & O_TRUNC) { 1551 VOP_UNLOCK(vp, 0); /* XXX */ 1552 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1553 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1554 VATTR_NULL(&va); 1555 va.va_size = 0; 1556 error = VOP_SETATTR(vp, &va, cred, l); 1557 if (error) 1558 goto bad; 1559 } 1560 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1561 goto bad; 1562 if (vp->v_type == VREG && 1563 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1564 error = EIO; 1565 goto bad; 1566 } 1567 if (flags & FWRITE) 1568 vp->v_writecount++; 1569 1570 /* done with modified vn_open, now finish what sys_open does. */ 1571 1572 fp->f_flag = flags & FMASK; 1573 fp->f_type = DTYPE_VNODE; 1574 fp->f_ops = &vnops; 1575 fp->f_data = vp; 1576 if (flags & (O_EXLOCK | O_SHLOCK)) { 1577 lf.l_whence = SEEK_SET; 1578 lf.l_start = 0; 1579 lf.l_len = 0; 1580 if (flags & O_EXLOCK) 1581 lf.l_type = F_WRLCK; 1582 else 1583 lf.l_type = F_RDLCK; 1584 type = F_FLOCK; 1585 if ((flags & FNONBLOCK) == 0) 1586 type |= F_WAIT; 1587 VOP_UNLOCK(vp, 0); 1588 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1589 if (error) { 1590 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1591 FILE_UNUSE(fp, l); 1592 ffree(fp); 1593 fdremove(fdp, indx); 1594 return (error); 1595 } 1596 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1597 fp->f_flag |= FHASLOCK; 1598 } 1599 VOP_UNLOCK(vp, 0); 1600 *retval = indx; 1601 FILE_SET_MATURE(fp); 1602 FILE_UNUSE(fp, l); 1603 vfs_copyinfh_free(fh); 1604 return (0); 1605 1606 bad: 1607 FILE_UNUSE(fp, l); 1608 ffree(fp); 1609 fdremove(fdp, indx); 1610 if (vp != NULL) 1611 vput(vp); 1612 vfs_copyinfh_free(fh); 1613 return (error); 1614 } 1615 1616 int 1617 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1618 { 1619 struct sys___fhopen40_args /* { 1620 syscallarg(const void *) fhp; 1621 syscallarg(size_t) fh_size; 1622 syscallarg(int) flags; 1623 } */ *uap = v; 1624 1625 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1626 SCARG(uap, flags), retval); 1627 } 1628 1629 /* XXX: temp (mar '07) for LKM compat */ 1630 int 1631 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1632 register_t *retval) 1633 { 1634 int error; 1635 struct stat sb; 1636 1637 error = do_fhstat(l, ufhp, fhsize, &sb); 1638 if (error == 0) 1639 error = copyout(&sb, sbp, sizeof(sb)); 1640 return error; 1641 } 1642 1643 int 1644 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1645 { 1646 int error; 1647 fhandle_t *fh; 1648 struct vnode *vp; 1649 1650 /* 1651 * Must be super user 1652 */ 1653 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1654 0, NULL, NULL, NULL))) 1655 return (error); 1656 1657 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1658 if (error != 0) 1659 return error; 1660 1661 error = vfs_fhtovp(fh, &vp); 1662 vfs_copyinfh_free(fh); 1663 if (error != 0) 1664 return error; 1665 1666 error = vn_stat(vp, sb, l); 1667 vput(vp); 1668 return error; 1669 } 1670 1671 1672 /* ARGSUSED */ 1673 int 1674 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1675 { 1676 struct sys___fhstat40_args /* { 1677 syscallarg(const void *) fhp; 1678 syscallarg(size_t) fh_size; 1679 syscallarg(struct stat *) sb; 1680 } */ *uap = v; 1681 struct stat sb; 1682 int error; 1683 1684 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1685 if (error) 1686 return error; 1687 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1688 } 1689 1690 /* XXX: temp (mar '07) for LKM compat */ 1691 int 1692 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1693 int flags, register_t *retval) 1694 { 1695 struct statvfs *sb = STATVFSBUF_GET(); 1696 int error; 1697 1698 error = do_fhstatvfs(l, ufhp, fhsize, sb, flags); 1699 if (error == 0) 1700 error = copyout(sb, buf, sizeof(*sb)); 1701 STATVFSBUF_PUT(sb); 1702 return error; 1703 } 1704 1705 int 1706 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1707 int flags) 1708 { 1709 fhandle_t *fh; 1710 struct mount *mp; 1711 struct vnode *vp; 1712 int error; 1713 1714 /* 1715 * Must be super user 1716 */ 1717 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1718 0, NULL, NULL, NULL))) 1719 return error; 1720 1721 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1722 if (error != 0) 1723 return error; 1724 1725 error = vfs_fhtovp(fh, &vp); 1726 vfs_copyinfh_free(fh); 1727 if (error != 0) 1728 return error; 1729 1730 mp = vp->v_mount; 1731 error = dostatvfs(mp, sb, l, flags, 1); 1732 vput(vp); 1733 return error; 1734 } 1735 1736 /* ARGSUSED */ 1737 int 1738 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1739 { 1740 struct sys___fhstatvfs140_args /* { 1741 syscallarg(const void *) fhp; 1742 syscallarg(size_t) fh_size; 1743 syscallarg(struct statvfs *) buf; 1744 syscallarg(int) flags; 1745 } */ *uap = v; 1746 struct statvfs *sb = STATVFSBUF_GET(); 1747 int error; 1748 1749 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1750 SCARG(uap, flags)); 1751 if (error == 0) 1752 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1753 STATVFSBUF_PUT(sb); 1754 return error; 1755 } 1756 1757 /* 1758 * Create a special file. 1759 */ 1760 /* ARGSUSED */ 1761 int 1762 sys_mknod(struct lwp *l, void *v, register_t *retval) 1763 { 1764 struct sys_mknod_args /* { 1765 syscallarg(const char *) path; 1766 syscallarg(int) mode; 1767 syscallarg(int) dev; 1768 } */ *uap = v; 1769 struct proc *p = l->l_proc; 1770 struct vnode *vp; 1771 struct vattr vattr; 1772 int error, optype; 1773 struct nameidata nd; 1774 1775 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1776 0, NULL, NULL, NULL)) != 0) 1777 return (error); 1778 1779 optype = VOP_MKNOD_DESCOFFSET; 1780 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1781 if ((error = namei(&nd)) != 0) 1782 return (error); 1783 vp = nd.ni_vp; 1784 if (vp != NULL) 1785 error = EEXIST; 1786 else { 1787 VATTR_NULL(&vattr); 1788 vattr.va_mode = 1789 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1790 vattr.va_rdev = SCARG(uap, dev); 1791 1792 switch (SCARG(uap, mode) & S_IFMT) { 1793 case S_IFMT: /* used by badsect to flag bad sectors */ 1794 vattr.va_type = VBAD; 1795 break; 1796 case S_IFCHR: 1797 vattr.va_type = VCHR; 1798 break; 1799 case S_IFBLK: 1800 vattr.va_type = VBLK; 1801 break; 1802 case S_IFWHT: 1803 optype = VOP_WHITEOUT_DESCOFFSET; 1804 break; 1805 case S_IFREG: 1806 vattr.va_type = VREG; 1807 vattr.va_rdev = VNOVAL; 1808 optype = VOP_CREATE_DESCOFFSET; 1809 break; 1810 default: 1811 error = EINVAL; 1812 break; 1813 } 1814 } 1815 if (!error) { 1816 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1817 switch (optype) { 1818 case VOP_WHITEOUT_DESCOFFSET: 1819 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1820 if (error) 1821 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1822 vput(nd.ni_dvp); 1823 break; 1824 1825 case VOP_MKNOD_DESCOFFSET: 1826 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1827 &nd.ni_cnd, &vattr); 1828 if (error == 0) 1829 vput(nd.ni_vp); 1830 break; 1831 1832 case VOP_CREATE_DESCOFFSET: 1833 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1834 &nd.ni_cnd, &vattr); 1835 if (error == 0) 1836 vput(nd.ni_vp); 1837 break; 1838 } 1839 } else { 1840 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1841 if (nd.ni_dvp == vp) 1842 vrele(nd.ni_dvp); 1843 else 1844 vput(nd.ni_dvp); 1845 if (vp) 1846 vrele(vp); 1847 } 1848 return (error); 1849 } 1850 1851 /* 1852 * Create a named pipe. 1853 */ 1854 /* ARGSUSED */ 1855 int 1856 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1857 { 1858 struct sys_mkfifo_args /* { 1859 syscallarg(const char *) path; 1860 syscallarg(int) mode; 1861 } */ *uap = v; 1862 struct proc *p = l->l_proc; 1863 struct vattr vattr; 1864 int error; 1865 struct nameidata nd; 1866 1867 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1868 if ((error = namei(&nd)) != 0) 1869 return (error); 1870 if (nd.ni_vp != NULL) { 1871 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1872 if (nd.ni_dvp == nd.ni_vp) 1873 vrele(nd.ni_dvp); 1874 else 1875 vput(nd.ni_dvp); 1876 vrele(nd.ni_vp); 1877 return (EEXIST); 1878 } 1879 VATTR_NULL(&vattr); 1880 vattr.va_type = VFIFO; 1881 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1882 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1883 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1884 if (error == 0) 1885 vput(nd.ni_vp); 1886 return (error); 1887 } 1888 1889 /* 1890 * Make a hard file link. 1891 */ 1892 /* ARGSUSED */ 1893 int 1894 sys_link(struct lwp *l, void *v, register_t *retval) 1895 { 1896 struct sys_link_args /* { 1897 syscallarg(const char *) path; 1898 syscallarg(const char *) link; 1899 } */ *uap = v; 1900 struct vnode *vp; 1901 struct nameidata nd; 1902 int error; 1903 1904 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1905 if ((error = namei(&nd)) != 0) 1906 return (error); 1907 vp = nd.ni_vp; 1908 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1909 if ((error = namei(&nd)) != 0) 1910 goto out; 1911 if (nd.ni_vp) { 1912 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1913 if (nd.ni_dvp == nd.ni_vp) 1914 vrele(nd.ni_dvp); 1915 else 1916 vput(nd.ni_dvp); 1917 vrele(nd.ni_vp); 1918 error = EEXIST; 1919 goto out; 1920 } 1921 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1922 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1923 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1924 out: 1925 vrele(vp); 1926 return (error); 1927 } 1928 1929 /* 1930 * Make a symbolic link. 1931 */ 1932 /* ARGSUSED */ 1933 int 1934 sys_symlink(struct lwp *l, void *v, register_t *retval) 1935 { 1936 struct sys_symlink_args /* { 1937 syscallarg(const char *) path; 1938 syscallarg(const char *) link; 1939 } */ *uap = v; 1940 struct proc *p = l->l_proc; 1941 struct vattr vattr; 1942 char *path; 1943 int error; 1944 struct nameidata nd; 1945 1946 path = PNBUF_GET(); 1947 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1948 if (error) 1949 goto out; 1950 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1951 if ((error = namei(&nd)) != 0) 1952 goto out; 1953 if (nd.ni_vp) { 1954 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1955 if (nd.ni_dvp == nd.ni_vp) 1956 vrele(nd.ni_dvp); 1957 else 1958 vput(nd.ni_dvp); 1959 vrele(nd.ni_vp); 1960 error = EEXIST; 1961 goto out; 1962 } 1963 VATTR_NULL(&vattr); 1964 vattr.va_type = VLNK; 1965 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1966 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1967 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1968 if (error == 0) 1969 vput(nd.ni_vp); 1970 out: 1971 PNBUF_PUT(path); 1972 return (error); 1973 } 1974 1975 /* 1976 * Delete a whiteout from the filesystem. 1977 */ 1978 /* ARGSUSED */ 1979 int 1980 sys_undelete(struct lwp *l, void *v, register_t *retval) 1981 { 1982 struct sys_undelete_args /* { 1983 syscallarg(const char *) path; 1984 } */ *uap = v; 1985 int error; 1986 struct nameidata nd; 1987 1988 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1989 SCARG(uap, path), l); 1990 error = namei(&nd); 1991 if (error) 1992 return (error); 1993 1994 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1995 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1996 if (nd.ni_dvp == nd.ni_vp) 1997 vrele(nd.ni_dvp); 1998 else 1999 vput(nd.ni_dvp); 2000 if (nd.ni_vp) 2001 vrele(nd.ni_vp); 2002 return (EEXIST); 2003 } 2004 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2005 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2006 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2007 vput(nd.ni_dvp); 2008 return (error); 2009 } 2010 2011 /* 2012 * Delete a name from the filesystem. 2013 */ 2014 /* ARGSUSED */ 2015 int 2016 sys_unlink(struct lwp *l, void *v, register_t *retval) 2017 { 2018 struct sys_unlink_args /* { 2019 syscallarg(const char *) path; 2020 } */ *uap = v; 2021 struct vnode *vp; 2022 int error; 2023 struct nameidata nd; 2024 #if NVERIEXEC > 0 2025 pathname_t pathbuf = NULL; 2026 #endif /* NVERIEXEC > 0 */ 2027 2028 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2029 SCARG(uap, path), l); 2030 if ((error = namei(&nd)) != 0) 2031 return (error); 2032 vp = nd.ni_vp; 2033 2034 /* 2035 * The root of a mounted filesystem cannot be deleted. 2036 */ 2037 if (vp->v_flag & VROOT) { 2038 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2039 if (nd.ni_dvp == vp) 2040 vrele(nd.ni_dvp); 2041 else 2042 vput(nd.ni_dvp); 2043 vput(vp); 2044 error = EBUSY; 2045 goto out; 2046 } 2047 2048 #if NVERIEXEC > 0 2049 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2050 2051 /* Handle remove requests for veriexec entries. */ 2052 if (!error) { 2053 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2054 pathname_put(pathbuf); 2055 } 2056 2057 if (error) { 2058 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2059 if (nd.ni_dvp == vp) 2060 vrele(nd.ni_dvp); 2061 else 2062 vput(nd.ni_dvp); 2063 vput(vp); 2064 goto out; 2065 } 2066 #endif /* NVERIEXEC > 0 */ 2067 2068 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2069 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2070 #ifdef FILEASSOC 2071 (void)fileassoc_file_delete(vp); 2072 #endif /* FILEASSOC */ 2073 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2074 out: 2075 return (error); 2076 } 2077 2078 /* 2079 * Reposition read/write file offset. 2080 */ 2081 int 2082 sys_lseek(struct lwp *l, void *v, register_t *retval) 2083 { 2084 struct sys_lseek_args /* { 2085 syscallarg(int) fd; 2086 syscallarg(int) pad; 2087 syscallarg(off_t) offset; 2088 syscallarg(int) whence; 2089 } */ *uap = v; 2090 struct proc *p = l->l_proc; 2091 kauth_cred_t cred = l->l_cred; 2092 struct filedesc *fdp = p->p_fd; 2093 struct file *fp; 2094 struct vnode *vp; 2095 struct vattr vattr; 2096 off_t newoff; 2097 int error; 2098 2099 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2100 return (EBADF); 2101 2102 FILE_USE(fp); 2103 2104 vp = (struct vnode *)fp->f_data; 2105 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2106 error = ESPIPE; 2107 goto out; 2108 } 2109 2110 switch (SCARG(uap, whence)) { 2111 case SEEK_CUR: 2112 newoff = fp->f_offset + SCARG(uap, offset); 2113 break; 2114 case SEEK_END: 2115 error = VOP_GETATTR(vp, &vattr, cred, l); 2116 if (error) 2117 goto out; 2118 newoff = SCARG(uap, offset) + vattr.va_size; 2119 break; 2120 case SEEK_SET: 2121 newoff = SCARG(uap, offset); 2122 break; 2123 default: 2124 error = EINVAL; 2125 goto out; 2126 } 2127 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2128 goto out; 2129 2130 *(off_t *)retval = fp->f_offset = newoff; 2131 out: 2132 FILE_UNUSE(fp, l); 2133 return (error); 2134 } 2135 2136 /* 2137 * Positional read system call. 2138 */ 2139 int 2140 sys_pread(struct lwp *l, void *v, register_t *retval) 2141 { 2142 struct sys_pread_args /* { 2143 syscallarg(int) fd; 2144 syscallarg(void *) buf; 2145 syscallarg(size_t) nbyte; 2146 syscallarg(off_t) offset; 2147 } */ *uap = v; 2148 struct proc *p = l->l_proc; 2149 struct filedesc *fdp = p->p_fd; 2150 struct file *fp; 2151 struct vnode *vp; 2152 off_t offset; 2153 int error, fd = SCARG(uap, fd); 2154 2155 if ((fp = fd_getfile(fdp, fd)) == NULL) 2156 return (EBADF); 2157 2158 if ((fp->f_flag & FREAD) == 0) { 2159 simple_unlock(&fp->f_slock); 2160 return (EBADF); 2161 } 2162 2163 FILE_USE(fp); 2164 2165 vp = (struct vnode *)fp->f_data; 2166 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2167 error = ESPIPE; 2168 goto out; 2169 } 2170 2171 offset = SCARG(uap, offset); 2172 2173 /* 2174 * XXX This works because no file systems actually 2175 * XXX take any action on the seek operation. 2176 */ 2177 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2178 goto out; 2179 2180 /* dofileread() will unuse the descriptor for us */ 2181 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2182 &offset, 0, retval)); 2183 2184 out: 2185 FILE_UNUSE(fp, l); 2186 return (error); 2187 } 2188 2189 /* 2190 * Positional scatter read system call. 2191 */ 2192 int 2193 sys_preadv(struct lwp *l, void *v, register_t *retval) 2194 { 2195 struct sys_preadv_args /* { 2196 syscallarg(int) fd; 2197 syscallarg(const struct iovec *) iovp; 2198 syscallarg(int) iovcnt; 2199 syscallarg(off_t) offset; 2200 } */ *uap = v; 2201 struct proc *p = l->l_proc; 2202 struct filedesc *fdp = p->p_fd; 2203 struct file *fp; 2204 struct vnode *vp; 2205 off_t offset; 2206 int error, fd = SCARG(uap, fd); 2207 2208 if ((fp = fd_getfile(fdp, fd)) == NULL) 2209 return (EBADF); 2210 2211 if ((fp->f_flag & FREAD) == 0) { 2212 simple_unlock(&fp->f_slock); 2213 return (EBADF); 2214 } 2215 2216 FILE_USE(fp); 2217 2218 vp = (struct vnode *)fp->f_data; 2219 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2220 error = ESPIPE; 2221 goto out; 2222 } 2223 2224 offset = SCARG(uap, offset); 2225 2226 /* 2227 * XXX This works because no file systems actually 2228 * XXX take any action on the seek operation. 2229 */ 2230 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2231 goto out; 2232 2233 /* dofilereadv() will unuse the descriptor for us */ 2234 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2235 &offset, 0, retval)); 2236 2237 out: 2238 FILE_UNUSE(fp, l); 2239 return (error); 2240 } 2241 2242 /* 2243 * Positional write system call. 2244 */ 2245 int 2246 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2247 { 2248 struct sys_pwrite_args /* { 2249 syscallarg(int) fd; 2250 syscallarg(const void *) buf; 2251 syscallarg(size_t) nbyte; 2252 syscallarg(off_t) offset; 2253 } */ *uap = v; 2254 struct proc *p = l->l_proc; 2255 struct filedesc *fdp = p->p_fd; 2256 struct file *fp; 2257 struct vnode *vp; 2258 off_t offset; 2259 int error, fd = SCARG(uap, fd); 2260 2261 if ((fp = fd_getfile(fdp, fd)) == NULL) 2262 return (EBADF); 2263 2264 if ((fp->f_flag & FWRITE) == 0) { 2265 simple_unlock(&fp->f_slock); 2266 return (EBADF); 2267 } 2268 2269 FILE_USE(fp); 2270 2271 vp = (struct vnode *)fp->f_data; 2272 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2273 error = ESPIPE; 2274 goto out; 2275 } 2276 2277 offset = SCARG(uap, offset); 2278 2279 /* 2280 * XXX This works because no file systems actually 2281 * XXX take any action on the seek operation. 2282 */ 2283 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2284 goto out; 2285 2286 /* dofilewrite() will unuse the descriptor for us */ 2287 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2288 &offset, 0, retval)); 2289 2290 out: 2291 FILE_UNUSE(fp, l); 2292 return (error); 2293 } 2294 2295 /* 2296 * Positional gather write system call. 2297 */ 2298 int 2299 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2300 { 2301 struct sys_pwritev_args /* { 2302 syscallarg(int) fd; 2303 syscallarg(const struct iovec *) iovp; 2304 syscallarg(int) iovcnt; 2305 syscallarg(off_t) offset; 2306 } */ *uap = v; 2307 struct proc *p = l->l_proc; 2308 struct filedesc *fdp = p->p_fd; 2309 struct file *fp; 2310 struct vnode *vp; 2311 off_t offset; 2312 int error, fd = SCARG(uap, fd); 2313 2314 if ((fp = fd_getfile(fdp, fd)) == NULL) 2315 return (EBADF); 2316 2317 if ((fp->f_flag & FWRITE) == 0) { 2318 simple_unlock(&fp->f_slock); 2319 return (EBADF); 2320 } 2321 2322 FILE_USE(fp); 2323 2324 vp = (struct vnode *)fp->f_data; 2325 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2326 error = ESPIPE; 2327 goto out; 2328 } 2329 2330 offset = SCARG(uap, offset); 2331 2332 /* 2333 * XXX This works because no file systems actually 2334 * XXX take any action on the seek operation. 2335 */ 2336 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2337 goto out; 2338 2339 /* dofilewritev() will unuse the descriptor for us */ 2340 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2341 &offset, 0, retval)); 2342 2343 out: 2344 FILE_UNUSE(fp, l); 2345 return (error); 2346 } 2347 2348 /* 2349 * Check access permissions. 2350 */ 2351 int 2352 sys_access(struct lwp *l, void *v, register_t *retval) 2353 { 2354 struct sys_access_args /* { 2355 syscallarg(const char *) path; 2356 syscallarg(int) flags; 2357 } */ *uap = v; 2358 kauth_cred_t cred; 2359 struct vnode *vp; 2360 int error, flags; 2361 struct nameidata nd; 2362 2363 cred = kauth_cred_dup(l->l_cred); 2364 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2365 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2366 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2367 SCARG(uap, path), l); 2368 /* Override default credentials */ 2369 nd.ni_cnd.cn_cred = cred; 2370 if ((error = namei(&nd)) != 0) 2371 goto out; 2372 vp = nd.ni_vp; 2373 2374 /* Flags == 0 means only check for existence. */ 2375 if (SCARG(uap, flags)) { 2376 flags = 0; 2377 if (SCARG(uap, flags) & R_OK) 2378 flags |= VREAD; 2379 if (SCARG(uap, flags) & W_OK) 2380 flags |= VWRITE; 2381 if (SCARG(uap, flags) & X_OK) 2382 flags |= VEXEC; 2383 2384 error = VOP_ACCESS(vp, flags, cred, l); 2385 if (!error && (flags & VWRITE)) 2386 error = vn_writechk(vp); 2387 } 2388 vput(vp); 2389 out: 2390 kauth_cred_free(cred); 2391 return (error); 2392 } 2393 2394 /* 2395 * Common code for all sys_stat functions, including compat versions. 2396 */ 2397 int 2398 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2399 struct stat *sb) 2400 { 2401 int error; 2402 struct nameidata nd; 2403 2404 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF, UIO_USERSPACE, path, l); 2405 error = namei(&nd); 2406 if (error != 0) 2407 return error; 2408 error = vn_stat(nd.ni_vp, sb, l); 2409 vput(nd.ni_vp); 2410 return error; 2411 } 2412 2413 /* 2414 * Get file status; this version follows links. 2415 */ 2416 /* ARGSUSED */ 2417 int 2418 sys___stat30(struct lwp *l, void *v, register_t *retval) 2419 { 2420 struct sys___stat30_args /* { 2421 syscallarg(const char *) path; 2422 syscallarg(struct stat *) ub; 2423 } */ *uap = v; 2424 struct stat sb; 2425 int error; 2426 2427 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2428 if (error) 2429 return error; 2430 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2431 } 2432 2433 /* 2434 * Get file status; this version does not follow links. 2435 */ 2436 /* ARGSUSED */ 2437 int 2438 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2439 { 2440 struct sys___lstat30_args /* { 2441 syscallarg(const char *) path; 2442 syscallarg(struct stat *) ub; 2443 } */ *uap = v; 2444 struct stat sb; 2445 int error; 2446 2447 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2448 if (error) 2449 return error; 2450 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2451 } 2452 2453 /* 2454 * Get configurable pathname variables. 2455 */ 2456 /* ARGSUSED */ 2457 int 2458 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2459 { 2460 struct sys_pathconf_args /* { 2461 syscallarg(const char *) path; 2462 syscallarg(int) name; 2463 } */ *uap = v; 2464 int error; 2465 struct nameidata nd; 2466 2467 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2468 SCARG(uap, path), l); 2469 if ((error = namei(&nd)) != 0) 2470 return (error); 2471 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2472 vput(nd.ni_vp); 2473 return (error); 2474 } 2475 2476 /* 2477 * Return target name of a symbolic link. 2478 */ 2479 /* ARGSUSED */ 2480 int 2481 sys_readlink(struct lwp *l, void *v, register_t *retval) 2482 { 2483 struct sys_readlink_args /* { 2484 syscallarg(const char *) path; 2485 syscallarg(char *) buf; 2486 syscallarg(size_t) count; 2487 } */ *uap = v; 2488 struct vnode *vp; 2489 struct iovec aiov; 2490 struct uio auio; 2491 int error; 2492 struct nameidata nd; 2493 2494 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2495 SCARG(uap, path), l); 2496 if ((error = namei(&nd)) != 0) 2497 return (error); 2498 vp = nd.ni_vp; 2499 if (vp->v_type != VLNK) 2500 error = EINVAL; 2501 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2502 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2503 aiov.iov_base = SCARG(uap, buf); 2504 aiov.iov_len = SCARG(uap, count); 2505 auio.uio_iov = &aiov; 2506 auio.uio_iovcnt = 1; 2507 auio.uio_offset = 0; 2508 auio.uio_rw = UIO_READ; 2509 KASSERT(l == curlwp); 2510 auio.uio_vmspace = l->l_proc->p_vmspace; 2511 auio.uio_resid = SCARG(uap, count); 2512 error = VOP_READLINK(vp, &auio, l->l_cred); 2513 } 2514 vput(vp); 2515 *retval = SCARG(uap, count) - auio.uio_resid; 2516 return (error); 2517 } 2518 2519 /* 2520 * Change flags of a file given a path name. 2521 */ 2522 /* ARGSUSED */ 2523 int 2524 sys_chflags(struct lwp *l, void *v, register_t *retval) 2525 { 2526 struct sys_chflags_args /* { 2527 syscallarg(const char *) path; 2528 syscallarg(u_long) flags; 2529 } */ *uap = v; 2530 struct vnode *vp; 2531 int error; 2532 struct nameidata nd; 2533 2534 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2535 if ((error = namei(&nd)) != 0) 2536 return (error); 2537 vp = nd.ni_vp; 2538 error = change_flags(vp, SCARG(uap, flags), l); 2539 vput(vp); 2540 return (error); 2541 } 2542 2543 /* 2544 * Change flags of a file given a file descriptor. 2545 */ 2546 /* ARGSUSED */ 2547 int 2548 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2549 { 2550 struct sys_fchflags_args /* { 2551 syscallarg(int) fd; 2552 syscallarg(u_long) flags; 2553 } */ *uap = v; 2554 struct proc *p = l->l_proc; 2555 struct vnode *vp; 2556 struct file *fp; 2557 int error; 2558 2559 /* getvnode() will use the descriptor for us */ 2560 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2561 return (error); 2562 vp = (struct vnode *)fp->f_data; 2563 error = change_flags(vp, SCARG(uap, flags), l); 2564 VOP_UNLOCK(vp, 0); 2565 FILE_UNUSE(fp, l); 2566 return (error); 2567 } 2568 2569 /* 2570 * Change flags of a file given a path name; this version does 2571 * not follow links. 2572 */ 2573 int 2574 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2575 { 2576 struct sys_lchflags_args /* { 2577 syscallarg(const char *) path; 2578 syscallarg(u_long) flags; 2579 } */ *uap = v; 2580 struct vnode *vp; 2581 int error; 2582 struct nameidata nd; 2583 2584 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2585 if ((error = namei(&nd)) != 0) 2586 return (error); 2587 vp = nd.ni_vp; 2588 error = change_flags(vp, SCARG(uap, flags), l); 2589 vput(vp); 2590 return (error); 2591 } 2592 2593 /* 2594 * Common routine to change flags of a file. 2595 */ 2596 int 2597 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2598 { 2599 struct vattr vattr; 2600 int error; 2601 2602 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2603 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2604 /* 2605 * Non-superusers cannot change the flags on devices, even if they 2606 * own them. 2607 */ 2608 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2609 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2610 goto out; 2611 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2612 error = EINVAL; 2613 goto out; 2614 } 2615 } 2616 VATTR_NULL(&vattr); 2617 vattr.va_flags = flags; 2618 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2619 out: 2620 return (error); 2621 } 2622 2623 /* 2624 * Change mode of a file given path name; this version follows links. 2625 */ 2626 /* ARGSUSED */ 2627 int 2628 sys_chmod(struct lwp *l, void *v, register_t *retval) 2629 { 2630 struct sys_chmod_args /* { 2631 syscallarg(const char *) path; 2632 syscallarg(int) mode; 2633 } */ *uap = v; 2634 int error; 2635 struct nameidata nd; 2636 2637 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2638 if ((error = namei(&nd)) != 0) 2639 return (error); 2640 2641 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2642 2643 vrele(nd.ni_vp); 2644 return (error); 2645 } 2646 2647 /* 2648 * Change mode of a file given a file descriptor. 2649 */ 2650 /* ARGSUSED */ 2651 int 2652 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2653 { 2654 struct sys_fchmod_args /* { 2655 syscallarg(int) fd; 2656 syscallarg(int) mode; 2657 } */ *uap = v; 2658 struct proc *p = l->l_proc; 2659 struct file *fp; 2660 int error; 2661 2662 /* getvnode() will use the descriptor for us */ 2663 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2664 return (error); 2665 2666 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2667 FILE_UNUSE(fp, l); 2668 return (error); 2669 } 2670 2671 /* 2672 * Change mode of a file given path name; this version does not follow links. 2673 */ 2674 /* ARGSUSED */ 2675 int 2676 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2677 { 2678 struct sys_lchmod_args /* { 2679 syscallarg(const char *) path; 2680 syscallarg(int) mode; 2681 } */ *uap = v; 2682 int error; 2683 struct nameidata nd; 2684 2685 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2686 if ((error = namei(&nd)) != 0) 2687 return (error); 2688 2689 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2690 2691 vrele(nd.ni_vp); 2692 return (error); 2693 } 2694 2695 /* 2696 * Common routine to set mode given a vnode. 2697 */ 2698 static int 2699 change_mode(struct vnode *vp, int mode, struct lwp *l) 2700 { 2701 struct vattr vattr; 2702 int error; 2703 2704 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2705 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2706 VATTR_NULL(&vattr); 2707 vattr.va_mode = mode & ALLPERMS; 2708 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2709 VOP_UNLOCK(vp, 0); 2710 return (error); 2711 } 2712 2713 /* 2714 * Set ownership given a path name; this version follows links. 2715 */ 2716 /* ARGSUSED */ 2717 int 2718 sys_chown(struct lwp *l, void *v, register_t *retval) 2719 { 2720 struct sys_chown_args /* { 2721 syscallarg(const char *) path; 2722 syscallarg(uid_t) uid; 2723 syscallarg(gid_t) gid; 2724 } */ *uap = v; 2725 int error; 2726 struct nameidata nd; 2727 2728 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2729 if ((error = namei(&nd)) != 0) 2730 return (error); 2731 2732 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2733 2734 vrele(nd.ni_vp); 2735 return (error); 2736 } 2737 2738 /* 2739 * Set ownership given a path name; this version follows links. 2740 * Provides POSIX semantics. 2741 */ 2742 /* ARGSUSED */ 2743 int 2744 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2745 { 2746 struct sys_chown_args /* { 2747 syscallarg(const char *) path; 2748 syscallarg(uid_t) uid; 2749 syscallarg(gid_t) gid; 2750 } */ *uap = v; 2751 int error; 2752 struct nameidata nd; 2753 2754 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2755 if ((error = namei(&nd)) != 0) 2756 return (error); 2757 2758 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2759 2760 vrele(nd.ni_vp); 2761 return (error); 2762 } 2763 2764 /* 2765 * Set ownership given a file descriptor. 2766 */ 2767 /* ARGSUSED */ 2768 int 2769 sys_fchown(struct lwp *l, void *v, register_t *retval) 2770 { 2771 struct sys_fchown_args /* { 2772 syscallarg(int) fd; 2773 syscallarg(uid_t) uid; 2774 syscallarg(gid_t) gid; 2775 } */ *uap = v; 2776 struct proc *p = l->l_proc; 2777 int error; 2778 struct file *fp; 2779 2780 /* getvnode() will use the descriptor for us */ 2781 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2782 return (error); 2783 2784 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2785 SCARG(uap, gid), l, 0); 2786 FILE_UNUSE(fp, l); 2787 return (error); 2788 } 2789 2790 /* 2791 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2792 */ 2793 /* ARGSUSED */ 2794 int 2795 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2796 { 2797 struct sys_fchown_args /* { 2798 syscallarg(int) fd; 2799 syscallarg(uid_t) uid; 2800 syscallarg(gid_t) gid; 2801 } */ *uap = v; 2802 struct proc *p = l->l_proc; 2803 int error; 2804 struct file *fp; 2805 2806 /* getvnode() will use the descriptor for us */ 2807 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2808 return (error); 2809 2810 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2811 SCARG(uap, gid), l, 1); 2812 FILE_UNUSE(fp, l); 2813 return (error); 2814 } 2815 2816 /* 2817 * Set ownership given a path name; this version does not follow links. 2818 */ 2819 /* ARGSUSED */ 2820 int 2821 sys_lchown(struct lwp *l, void *v, register_t *retval) 2822 { 2823 struct sys_lchown_args /* { 2824 syscallarg(const char *) path; 2825 syscallarg(uid_t) uid; 2826 syscallarg(gid_t) gid; 2827 } */ *uap = v; 2828 int error; 2829 struct nameidata nd; 2830 2831 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2832 if ((error = namei(&nd)) != 0) 2833 return (error); 2834 2835 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2836 2837 vrele(nd.ni_vp); 2838 return (error); 2839 } 2840 2841 /* 2842 * Set ownership given a path name; this version does not follow links. 2843 * Provides POSIX/XPG semantics. 2844 */ 2845 /* ARGSUSED */ 2846 int 2847 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2848 { 2849 struct sys_lchown_args /* { 2850 syscallarg(const char *) path; 2851 syscallarg(uid_t) uid; 2852 syscallarg(gid_t) gid; 2853 } */ *uap = v; 2854 int error; 2855 struct nameidata nd; 2856 2857 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2858 if ((error = namei(&nd)) != 0) 2859 return (error); 2860 2861 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2862 2863 vrele(nd.ni_vp); 2864 return (error); 2865 } 2866 2867 /* 2868 * Common routine to set ownership given a vnode. 2869 */ 2870 static int 2871 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2872 int posix_semantics) 2873 { 2874 struct vattr vattr; 2875 mode_t newmode; 2876 int error; 2877 2878 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2879 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2880 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2881 goto out; 2882 2883 #define CHANGED(x) ((int)(x) != -1) 2884 newmode = vattr.va_mode; 2885 if (posix_semantics) { 2886 /* 2887 * POSIX/XPG semantics: if the caller is not the super-user, 2888 * clear set-user-id and set-group-id bits. Both POSIX and 2889 * the XPG consider the behaviour for calls by the super-user 2890 * implementation-defined; we leave the set-user-id and set- 2891 * group-id settings intact in that case. 2892 */ 2893 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2894 NULL) != 0) 2895 newmode &= ~(S_ISUID | S_ISGID); 2896 } else { 2897 /* 2898 * NetBSD semantics: when changing owner and/or group, 2899 * clear the respective bit(s). 2900 */ 2901 if (CHANGED(uid)) 2902 newmode &= ~S_ISUID; 2903 if (CHANGED(gid)) 2904 newmode &= ~S_ISGID; 2905 } 2906 /* Update va_mode iff altered. */ 2907 if (vattr.va_mode == newmode) 2908 newmode = VNOVAL; 2909 2910 VATTR_NULL(&vattr); 2911 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2912 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2913 vattr.va_mode = newmode; 2914 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2915 #undef CHANGED 2916 2917 out: 2918 VOP_UNLOCK(vp, 0); 2919 return (error); 2920 } 2921 2922 /* 2923 * Set the access and modification times given a path name; this 2924 * version follows links. 2925 */ 2926 /* ARGSUSED */ 2927 int 2928 sys_utimes(struct lwp *l, void *v, register_t *retval) 2929 { 2930 struct sys_utimes_args /* { 2931 syscallarg(const char *) path; 2932 syscallarg(const struct timeval *) tptr; 2933 } */ *uap = v; 2934 int error; 2935 struct nameidata nd; 2936 2937 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2938 if ((error = namei(&nd)) != 0) 2939 return (error); 2940 2941 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2942 2943 vrele(nd.ni_vp); 2944 return (error); 2945 } 2946 2947 /* 2948 * Set the access and modification times given a file descriptor. 2949 */ 2950 /* ARGSUSED */ 2951 int 2952 sys_futimes(struct lwp *l, void *v, register_t *retval) 2953 { 2954 struct sys_futimes_args /* { 2955 syscallarg(int) fd; 2956 syscallarg(const struct timeval *) tptr; 2957 } */ *uap = v; 2958 struct proc *p = l->l_proc; 2959 int error; 2960 struct file *fp; 2961 2962 /* getvnode() will use the descriptor for us */ 2963 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2964 return (error); 2965 2966 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2967 FILE_UNUSE(fp, l); 2968 return (error); 2969 } 2970 2971 /* 2972 * Set the access and modification times given a path name; this 2973 * version does not follow links. 2974 */ 2975 /* ARGSUSED */ 2976 int 2977 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2978 { 2979 struct sys_lutimes_args /* { 2980 syscallarg(const char *) path; 2981 syscallarg(const struct timeval *) tptr; 2982 } */ *uap = v; 2983 int error; 2984 struct nameidata nd; 2985 2986 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2987 if ((error = namei(&nd)) != 0) 2988 return (error); 2989 2990 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2991 2992 vrele(nd.ni_vp); 2993 return (error); 2994 } 2995 2996 /* 2997 * Common routine to set access and modification times given a vnode. 2998 */ 2999 static int 3000 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3001 { 3002 struct vattr vattr; 3003 int error; 3004 3005 VATTR_NULL(&vattr); 3006 if (tptr == NULL) { 3007 nanotime(&vattr.va_atime); 3008 vattr.va_mtime = vattr.va_atime; 3009 vattr.va_vaflags |= VA_UTIMES_NULL; 3010 } else { 3011 struct timeval tv[2]; 3012 3013 error = copyin(tptr, tv, sizeof(tv)); 3014 if (error) 3015 goto out; 3016 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3017 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3018 } 3019 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3020 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3021 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3022 VOP_UNLOCK(vp, 0); 3023 out: 3024 return (error); 3025 } 3026 3027 /* 3028 * Truncate a file given its path name. 3029 */ 3030 /* ARGSUSED */ 3031 int 3032 sys_truncate(struct lwp *l, void *v, register_t *retval) 3033 { 3034 struct sys_truncate_args /* { 3035 syscallarg(const char *) path; 3036 syscallarg(int) pad; 3037 syscallarg(off_t) length; 3038 } */ *uap = v; 3039 struct vnode *vp; 3040 struct vattr vattr; 3041 int error; 3042 struct nameidata nd; 3043 3044 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3045 if ((error = namei(&nd)) != 0) 3046 return (error); 3047 vp = nd.ni_vp; 3048 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3049 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3050 if (vp->v_type == VDIR) 3051 error = EISDIR; 3052 else if ((error = vn_writechk(vp)) == 0 && 3053 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3054 VATTR_NULL(&vattr); 3055 vattr.va_size = SCARG(uap, length); 3056 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3057 } 3058 vput(vp); 3059 return (error); 3060 } 3061 3062 /* 3063 * Truncate a file given a file descriptor. 3064 */ 3065 /* ARGSUSED */ 3066 int 3067 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3068 { 3069 struct sys_ftruncate_args /* { 3070 syscallarg(int) fd; 3071 syscallarg(int) pad; 3072 syscallarg(off_t) length; 3073 } */ *uap = v; 3074 struct proc *p = l->l_proc; 3075 struct vattr vattr; 3076 struct vnode *vp; 3077 struct file *fp; 3078 int error; 3079 3080 /* getvnode() will use the descriptor for us */ 3081 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3082 return (error); 3083 if ((fp->f_flag & FWRITE) == 0) { 3084 error = EINVAL; 3085 goto out; 3086 } 3087 vp = (struct vnode *)fp->f_data; 3088 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3089 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3090 if (vp->v_type == VDIR) 3091 error = EISDIR; 3092 else if ((error = vn_writechk(vp)) == 0) { 3093 VATTR_NULL(&vattr); 3094 vattr.va_size = SCARG(uap, length); 3095 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3096 } 3097 VOP_UNLOCK(vp, 0); 3098 out: 3099 FILE_UNUSE(fp, l); 3100 return (error); 3101 } 3102 3103 /* 3104 * Sync an open file. 3105 */ 3106 /* ARGSUSED */ 3107 int 3108 sys_fsync(struct lwp *l, void *v, register_t *retval) 3109 { 3110 struct sys_fsync_args /* { 3111 syscallarg(int) fd; 3112 } */ *uap = v; 3113 struct proc *p = l->l_proc; 3114 struct vnode *vp; 3115 struct file *fp; 3116 int error; 3117 3118 /* getvnode() will use the descriptor for us */ 3119 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3120 return (error); 3121 vp = (struct vnode *)fp->f_data; 3122 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3123 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3124 if (error == 0 && bioops.io_fsync != NULL && 3125 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3126 (*bioops.io_fsync)(vp, 0); 3127 VOP_UNLOCK(vp, 0); 3128 FILE_UNUSE(fp, l); 3129 return (error); 3130 } 3131 3132 /* 3133 * Sync a range of file data. API modeled after that found in AIX. 3134 * 3135 * FDATASYNC indicates that we need only save enough metadata to be able 3136 * to re-read the written data. Note we duplicate AIX's requirement that 3137 * the file be open for writing. 3138 */ 3139 /* ARGSUSED */ 3140 int 3141 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3142 { 3143 struct sys_fsync_range_args /* { 3144 syscallarg(int) fd; 3145 syscallarg(int) flags; 3146 syscallarg(off_t) start; 3147 syscallarg(off_t) length; 3148 } */ *uap = v; 3149 struct proc *p = l->l_proc; 3150 struct vnode *vp; 3151 struct file *fp; 3152 int flags, nflags; 3153 off_t s, e, len; 3154 int error; 3155 3156 /* getvnode() will use the descriptor for us */ 3157 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3158 return (error); 3159 3160 if ((fp->f_flag & FWRITE) == 0) { 3161 error = EBADF; 3162 goto out; 3163 } 3164 3165 flags = SCARG(uap, flags); 3166 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3167 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3168 error = EINVAL; 3169 goto out; 3170 } 3171 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3172 if (flags & FDATASYNC) 3173 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3174 else 3175 nflags = FSYNC_WAIT; 3176 if (flags & FDISKSYNC) 3177 nflags |= FSYNC_CACHE; 3178 3179 len = SCARG(uap, length); 3180 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3181 if (len) { 3182 s = SCARG(uap, start); 3183 e = s + len; 3184 if (e < s) { 3185 error = EINVAL; 3186 goto out; 3187 } 3188 } else { 3189 e = 0; 3190 s = 0; 3191 } 3192 3193 vp = (struct vnode *)fp->f_data; 3194 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3195 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3196 3197 if (error == 0 && bioops.io_fsync != NULL && 3198 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3199 (*bioops.io_fsync)(vp, nflags); 3200 3201 VOP_UNLOCK(vp, 0); 3202 out: 3203 FILE_UNUSE(fp, l); 3204 return (error); 3205 } 3206 3207 /* 3208 * Sync the data of an open file. 3209 */ 3210 /* ARGSUSED */ 3211 int 3212 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3213 { 3214 struct sys_fdatasync_args /* { 3215 syscallarg(int) fd; 3216 } */ *uap = v; 3217 struct proc *p = l->l_proc; 3218 struct vnode *vp; 3219 struct file *fp; 3220 int error; 3221 3222 /* getvnode() will use the descriptor for us */ 3223 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3224 return (error); 3225 if ((fp->f_flag & FWRITE) == 0) { 3226 FILE_UNUSE(fp, l); 3227 return (EBADF); 3228 } 3229 vp = (struct vnode *)fp->f_data; 3230 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3231 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3232 VOP_UNLOCK(vp, 0); 3233 FILE_UNUSE(fp, l); 3234 return (error); 3235 } 3236 3237 /* 3238 * Rename files, (standard) BSD semantics frontend. 3239 */ 3240 /* ARGSUSED */ 3241 int 3242 sys_rename(struct lwp *l, void *v, register_t *retval) 3243 { 3244 struct sys_rename_args /* { 3245 syscallarg(const char *) from; 3246 syscallarg(const char *) to; 3247 } */ *uap = v; 3248 3249 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3250 } 3251 3252 /* 3253 * Rename files, POSIX semantics frontend. 3254 */ 3255 /* ARGSUSED */ 3256 int 3257 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3258 { 3259 struct sys___posix_rename_args /* { 3260 syscallarg(const char *) from; 3261 syscallarg(const char *) to; 3262 } */ *uap = v; 3263 3264 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3265 } 3266 3267 /* 3268 * Rename files. Source and destination must either both be directories, 3269 * or both not be directories. If target is a directory, it must be empty. 3270 * If `from' and `to' refer to the same object, the value of the `retain' 3271 * argument is used to determine whether `from' will be 3272 * 3273 * (retain == 0) deleted unless `from' and `to' refer to the same 3274 * object in the file system's name space (BSD). 3275 * (retain == 1) always retained (POSIX). 3276 */ 3277 static int 3278 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3279 { 3280 struct vnode *tvp, *fvp, *tdvp; 3281 struct nameidata fromnd, tond; 3282 struct proc *p; 3283 int error; 3284 3285 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3286 from, l); 3287 if ((error = namei(&fromnd)) != 0) 3288 return (error); 3289 if (fromnd.ni_dvp != fromnd.ni_vp) 3290 VOP_UNLOCK(fromnd.ni_dvp, 0); 3291 fvp = fromnd.ni_vp; 3292 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3293 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3294 if ((error = namei(&tond)) != 0) { 3295 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3296 vrele(fromnd.ni_dvp); 3297 vrele(fvp); 3298 goto out1; 3299 } 3300 tdvp = tond.ni_dvp; 3301 tvp = tond.ni_vp; 3302 3303 if (tvp != NULL) { 3304 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3305 error = ENOTDIR; 3306 goto out; 3307 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3308 error = EISDIR; 3309 goto out; 3310 } 3311 } 3312 3313 if (fvp == tdvp) 3314 error = EINVAL; 3315 3316 /* 3317 * Source and destination refer to the same object. 3318 */ 3319 if (fvp == tvp) { 3320 if (retain) 3321 error = -1; 3322 else if (fromnd.ni_dvp == tdvp && 3323 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3324 !memcmp(fromnd.ni_cnd.cn_nameptr, 3325 tond.ni_cnd.cn_nameptr, 3326 fromnd.ni_cnd.cn_namelen)) 3327 error = -1; 3328 } 3329 3330 #if NVERIEXEC > 0 3331 if (!error) { 3332 pathname_t frompath = NULL, topath = NULL; 3333 3334 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3335 &frompath); 3336 if (!error) 3337 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3338 &topath); 3339 if (!error) 3340 error = veriexec_renamechk(fvp, pathname_path(frompath), 3341 tvp, pathname_path(topath), l); 3342 3343 pathname_put(frompath); 3344 pathname_put(topath); 3345 } 3346 #endif /* NVERIEXEC > 0 */ 3347 3348 out: 3349 p = l->l_proc; 3350 if (!error) { 3351 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3352 if (fromnd.ni_dvp != tdvp) 3353 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3354 if (tvp) { 3355 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3356 } 3357 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3358 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3359 } else { 3360 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3361 if (tdvp == tvp) 3362 vrele(tdvp); 3363 else 3364 vput(tdvp); 3365 if (tvp) 3366 vput(tvp); 3367 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3368 vrele(fromnd.ni_dvp); 3369 vrele(fvp); 3370 } 3371 vrele(tond.ni_startdir); 3372 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3373 out1: 3374 if (fromnd.ni_startdir) 3375 vrele(fromnd.ni_startdir); 3376 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3377 return (error == -1 ? 0 : error); 3378 } 3379 3380 /* 3381 * Make a directory file. 3382 */ 3383 /* ARGSUSED */ 3384 int 3385 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3386 { 3387 struct sys_mkdir_args /* { 3388 syscallarg(const char *) path; 3389 syscallarg(int) mode; 3390 } */ *uap = v; 3391 struct proc *p = l->l_proc; 3392 struct vnode *vp; 3393 struct vattr vattr; 3394 int error; 3395 struct nameidata nd; 3396 3397 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3398 SCARG(uap, path), l); 3399 if ((error = namei(&nd)) != 0) 3400 return (error); 3401 vp = nd.ni_vp; 3402 if (vp != NULL) { 3403 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3404 if (nd.ni_dvp == vp) 3405 vrele(nd.ni_dvp); 3406 else 3407 vput(nd.ni_dvp); 3408 vrele(vp); 3409 return (EEXIST); 3410 } 3411 VATTR_NULL(&vattr); 3412 vattr.va_type = VDIR; 3413 vattr.va_mode = 3414 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3415 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3416 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3417 if (!error) 3418 vput(nd.ni_vp); 3419 return (error); 3420 } 3421 3422 /* 3423 * Remove a directory file. 3424 */ 3425 /* ARGSUSED */ 3426 int 3427 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3428 { 3429 struct sys_rmdir_args /* { 3430 syscallarg(const char *) path; 3431 } */ *uap = v; 3432 struct vnode *vp; 3433 int error; 3434 struct nameidata nd; 3435 3436 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3437 SCARG(uap, path), l); 3438 if ((error = namei(&nd)) != 0) 3439 return (error); 3440 vp = nd.ni_vp; 3441 if (vp->v_type != VDIR) { 3442 error = ENOTDIR; 3443 goto out; 3444 } 3445 /* 3446 * No rmdir "." please. 3447 */ 3448 if (nd.ni_dvp == vp) { 3449 error = EINVAL; 3450 goto out; 3451 } 3452 /* 3453 * The root of a mounted filesystem cannot be deleted. 3454 */ 3455 if (vp->v_flag & VROOT) { 3456 error = EBUSY; 3457 goto out; 3458 } 3459 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3460 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3461 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3462 return (error); 3463 3464 out: 3465 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3466 if (nd.ni_dvp == vp) 3467 vrele(nd.ni_dvp); 3468 else 3469 vput(nd.ni_dvp); 3470 vput(vp); 3471 return (error); 3472 } 3473 3474 /* 3475 * Read a block of directory entries in a file system independent format. 3476 */ 3477 int 3478 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3479 { 3480 struct sys___getdents30_args /* { 3481 syscallarg(int) fd; 3482 syscallarg(char *) buf; 3483 syscallarg(size_t) count; 3484 } */ *uap = v; 3485 struct proc *p = l->l_proc; 3486 struct file *fp; 3487 int error, done; 3488 3489 /* getvnode() will use the descriptor for us */ 3490 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3491 return (error); 3492 if ((fp->f_flag & FREAD) == 0) { 3493 error = EBADF; 3494 goto out; 3495 } 3496 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3497 SCARG(uap, count), &done, l, 0, 0); 3498 #ifdef KTRACE 3499 if (!error && KTRPOINT(p, KTR_GENIO)) { 3500 struct iovec iov; 3501 iov.iov_base = SCARG(uap, buf); 3502 iov.iov_len = done; 3503 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3504 } 3505 #endif 3506 *retval = done; 3507 out: 3508 FILE_UNUSE(fp, l); 3509 return (error); 3510 } 3511 3512 /* 3513 * Set the mode mask for creation of filesystem nodes. 3514 */ 3515 int 3516 sys_umask(struct lwp *l, void *v, register_t *retval) 3517 { 3518 struct sys_umask_args /* { 3519 syscallarg(mode_t) newmask; 3520 } */ *uap = v; 3521 struct proc *p = l->l_proc; 3522 struct cwdinfo *cwdi; 3523 3524 cwdi = p->p_cwdi; 3525 *retval = cwdi->cwdi_cmask; 3526 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3527 return (0); 3528 } 3529 3530 /* 3531 * Void all references to file by ripping underlying filesystem 3532 * away from vnode. 3533 */ 3534 /* ARGSUSED */ 3535 int 3536 sys_revoke(struct lwp *l, void *v, register_t *retval) 3537 { 3538 struct sys_revoke_args /* { 3539 syscallarg(const char *) path; 3540 } */ *uap = v; 3541 struct vnode *vp; 3542 struct vattr vattr; 3543 int error; 3544 struct nameidata nd; 3545 3546 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3547 if ((error = namei(&nd)) != 0) 3548 return (error); 3549 vp = nd.ni_vp; 3550 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3551 goto out; 3552 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3553 (error = kauth_authorize_generic(l->l_cred, 3554 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3555 goto out; 3556 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3557 VOP_REVOKE(vp, REVOKEALL); 3558 out: 3559 vrele(vp); 3560 return (error); 3561 } 3562 3563 /* 3564 * Convert a user file descriptor to a kernel file entry. 3565 */ 3566 int 3567 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3568 { 3569 struct vnode *vp; 3570 struct file *fp; 3571 3572 if ((fp = fd_getfile(fdp, fd)) == NULL) 3573 return (EBADF); 3574 3575 FILE_USE(fp); 3576 3577 if (fp->f_type != DTYPE_VNODE) { 3578 FILE_UNUSE(fp, NULL); 3579 return (EINVAL); 3580 } 3581 3582 vp = (struct vnode *)fp->f_data; 3583 if (vp->v_type == VBAD) { 3584 FILE_UNUSE(fp, NULL); 3585 return (EBADF); 3586 } 3587 3588 *fpp = fp; 3589 return (0); 3590 } 3591
Indexes created Thu Sep 25 16:09:42 GMT 2025