vfs_syscalls.c revision 1.306.2.6
1 /* $NetBSD: vfs_syscalls.c,v 1.306.2.6 2007/06/08 14:17:30 ad 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 1.306.2.6 2007/06/08 14:17:30 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/syscallargs.h> 65 #include <sys/vfs_syscalls.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #include <sys/verified_exec.h> 73 #include <sys/kauth.h> 74 75 #include <miscfs/genfs/genfs.h> 76 #include <miscfs/syncfs/syncfs.h> 77 78 #ifdef COMPAT_30 79 #include "opt_nfsserver.h" 80 #include <nfs/rpcv2.h> 81 #endif 82 #include <nfs/nfsproto.h> 83 #ifdef COMPAT_30 84 #include <nfs/nfs.h> 85 #include <nfs/nfs_var.h> 86 #endif 87 88 #if NFSS > 0 89 #include <dev/fssvar.h> 90 #endif 91 92 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 93 94 static int change_dir(struct nameidata *, struct lwp *); 95 static int change_flags(struct vnode *, u_long, struct lwp *); 96 static int change_mode(struct vnode *, int, struct lwp *l); 97 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 98 static int rename_files(const char *, const char *, struct lwp *, int); 99 100 void checkdirs(struct vnode *); 101 102 int dovfsusermount = 0; 103 104 /* 105 * Virtual File System System Calls 106 */ 107 108 /* 109 * Mount a file system. 110 */ 111 112 #if defined(COMPAT_09) || defined(COMPAT_43) 113 /* 114 * This table is used to maintain compatibility with 4.3BSD 115 * and NetBSD 0.9 mount syscalls. Note, the order is important! 116 * 117 * Do not modify this table. It should only contain filesystems 118 * supported by NetBSD 0.9 and 4.3BSD. 119 */ 120 const char * const mountcompatnames[] = { 121 NULL, /* 0 = MOUNT_NONE */ 122 MOUNT_FFS, /* 1 = MOUNT_UFS */ 123 MOUNT_NFS, /* 2 */ 124 MOUNT_MFS, /* 3 */ 125 MOUNT_MSDOS, /* 4 */ 126 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 127 MOUNT_FDESC, /* 6 */ 128 MOUNT_KERNFS, /* 7 */ 129 NULL, /* 8 = MOUNT_DEVFS */ 130 MOUNT_AFS, /* 9 */ 131 }; 132 const int nmountcompatnames = sizeof(mountcompatnames) / 133 sizeof(mountcompatnames[0]); 134 #endif /* COMPAT_09 || COMPAT_43 */ 135 136 static int 137 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 138 void *data, struct nameidata *ndp) 139 { 140 struct mount *mp; 141 int error = 0, saved_flags; 142 143 mp = vp->v_mount; 144 saved_flags = mp->mnt_flag; 145 146 /* We can operate only on VROOT nodes. */ 147 if ((vp->v_flag & VROOT) == 0) { 148 error = EINVAL; 149 goto out; 150 } 151 152 /* 153 * We only allow the filesystem to be reloaded if it 154 * is currently mounted read-only. 155 */ 156 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 157 error = EOPNOTSUPP; /* Needs translation */ 158 goto out; 159 } 160 161 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 162 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 163 if (error) 164 goto out; 165 166 if (vfs_busy(mp, LK_NOWAIT, 0)) { 167 error = EPERM; 168 goto out; 169 } 170 171 mp->mnt_flag &= ~MNT_OP_FLAGS; 172 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 173 174 /* 175 * Set the mount level flags. 176 */ 177 if (flags & MNT_RDONLY) 178 mp->mnt_flag |= MNT_RDONLY; 179 else if (mp->mnt_flag & MNT_RDONLY) 180 mp->mnt_iflag |= IMNT_WANTRDWR; 181 mp->mnt_flag &= 182 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 183 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 184 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 185 mp->mnt_flag |= flags & 186 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 187 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 188 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 189 MNT_IGNORE); 190 191 error = VFS_MOUNT(mp, path, data, ndp, l); 192 193 #if defined(COMPAT_30) && defined(NFSSERVER) 194 if (error) { 195 int error2; 196 197 /* Update failed; let's try and see if it was an 198 * export request. */ 199 error2 = nfs_update_exports_30(mp, path, data, l); 200 201 /* Only update error code if the export request was 202 * understood but some problem occurred while 203 * processing it. */ 204 if (error2 != EJUSTRETURN) 205 error = error2; 206 } 207 #endif 208 if (mp->mnt_iflag & IMNT_WANTRDWR) 209 mp->mnt_flag &= ~MNT_RDONLY; 210 if (error) 211 mp->mnt_flag = saved_flags; 212 mp->mnt_flag &= ~MNT_OP_FLAGS; 213 mp->mnt_iflag &= ~IMNT_WANTRDWR; 214 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 215 if (mp->mnt_syncer == NULL) 216 error = vfs_allocate_syncvnode(mp); 217 } else { 218 if (mp->mnt_syncer != NULL) 219 vfs_deallocate_syncvnode(mp); 220 } 221 vfs_unbusy(mp); 222 223 out: 224 return (error); 225 } 226 227 static int 228 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 229 const char *path, int flags, void *data, struct nameidata *ndp) 230 { 231 struct mount *mp = NULL; 232 struct vattr va; 233 char fstypename[MFSNAMELEN]; 234 int error; 235 236 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 237 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 238 if (error) { 239 vput(vp); 240 goto out; 241 } 242 243 /* Can't make a non-dir a mount-point (from here anyway). */ 244 if (vp->v_type != VDIR) { 245 error = ENOTDIR; 246 vput(vp); 247 goto out; 248 } 249 250 /* 251 * If the user is not root, ensure that they own the directory 252 * onto which we are attempting to mount. 253 */ 254 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 255 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 256 (error = kauth_authorize_generic(l->l_cred, 257 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 258 vput(vp); 259 goto out; 260 } 261 262 if (flags & MNT_EXPORTED) { 263 error = EINVAL; 264 vput(vp); 265 goto out; 266 } 267 268 /* 269 * Copy file-system type from userspace. 270 */ 271 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 272 if (error) { 273 #if defined(COMPAT_09) || defined(COMPAT_43) 274 /* 275 * Historically, filesystem types were identified by numbers. 276 * If we get an integer for the filesystem type instead of a 277 * string, we check to see if it matches one of the historic 278 * filesystem types. 279 */ 280 u_long fsindex = (u_long)fstype; 281 if (fsindex >= nmountcompatnames || 282 mountcompatnames[fsindex] == NULL) { 283 error = ENODEV; 284 vput(vp); 285 goto out; 286 } 287 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 288 #else 289 vput(vp); 290 goto out; 291 #endif 292 } 293 294 #ifdef COMPAT_10 295 /* Accept `ufs' as an alias for `ffs'. */ 296 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 297 strlcpy(fstypename, "ffs", sizeof(fstypename)); 298 #endif 299 300 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 301 vput(vp); 302 goto out; 303 } 304 305 /* 306 * Check if a file-system is not already mounted on this vnode. 307 */ 308 if (vp->v_mountedhere != NULL) { 309 error = EBUSY; 310 vput(vp); 311 goto out; 312 } 313 314 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 315 316 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 317 free(mp, M_MOUNT); 318 error = ENODEV; 319 vput(vp); 320 goto out; 321 } 322 323 TAILQ_INIT(&mp->mnt_vnodelist); 324 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 325 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 326 (void)vfs_busy(mp, LK_NOWAIT, 0); 327 328 mp->mnt_vnodecovered = vp; 329 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 330 mp->mnt_unmounter = NULL; 331 mount_initspecific(mp); 332 333 /* 334 * The underlying file system may refuse the mount for 335 * various reasons. Allow the user to force it to happen. 336 * 337 * Set the mount level flags. 338 */ 339 mp->mnt_flag = flags & 340 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 341 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 342 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 343 MNT_IGNORE | MNT_RDONLY); 344 345 error = VFS_MOUNT(mp, path, data, ndp, l); 346 mp->mnt_flag &= ~MNT_OP_FLAGS; 347 348 /* 349 * Put the new filesystem on the mount list after root. 350 */ 351 cache_purge(vp); 352 if (!error) { 353 mp->mnt_iflag &= ~IMNT_WANTRDWR; 354 vp->v_mountedhere = mp; 355 mutex_enter(&mountlist_lock); 356 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 357 mutex_exit(&mountlist_lock); 358 VOP_UNLOCK(vp, 0); 359 checkdirs(vp); 360 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 361 error = vfs_allocate_syncvnode(mp); 362 vfs_unbusy(mp); 363 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 364 error = VFS_START(mp, 0, l); 365 if (error) 366 vrele(vp); 367 } else { 368 vp->v_mountedhere = NULL; 369 vfs_unbusy(mp); 370 vfs_delref(mp->mnt_op); 371 free(mp, M_MOUNT); 372 vput(vp); 373 } 374 375 out: 376 return (error); 377 } 378 379 static int 380 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 381 void *data, struct nameidata *ndp) 382 { 383 struct mount *mp; 384 int error; 385 386 /* If MNT_GETARGS is specified, it should be the only flag. */ 387 if (flags & ~MNT_GETARGS) { 388 error = EINVAL; 389 goto out; 390 } 391 392 mp = vp->v_mount; 393 394 /* XXX: probably some notion of "can see" here if we want isolation. */ 395 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 396 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 397 if (error) 398 goto out; 399 400 if ((vp->v_flag & VROOT) == 0) { 401 error = EINVAL; 402 goto out; 403 } 404 405 if (vfs_busy(mp, LK_NOWAIT, 0)) { 406 error = EPERM; 407 goto out; 408 } 409 410 mp->mnt_flag &= ~MNT_OP_FLAGS; 411 mp->mnt_flag |= MNT_GETARGS; 412 error = VFS_MOUNT(mp, path, data, ndp, l); 413 mp->mnt_flag &= ~MNT_OP_FLAGS; 414 415 vfs_unbusy(mp); 416 out: 417 return (error); 418 } 419 420 /* ARGSUSED */ 421 int 422 sys_mount(struct lwp *l, void *v, register_t *retval) 423 { 424 struct sys_mount_args /* { 425 syscallarg(const char *) type; 426 syscallarg(const char *) path; 427 syscallarg(int) flags; 428 syscallarg(void *) data; 429 } */ *uap = v; 430 struct vnode *vp; 431 struct nameidata nd; 432 int error; 433 434 /* 435 * Get vnode to be covered 436 */ 437 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 438 SCARG(uap, path), l); 439 if ((error = namei(&nd)) != 0) 440 return (error); 441 vp = nd.ni_vp; 442 443 /* 444 * A lookup in VFS_MOUNT might result in an attempt to 445 * lock this vnode again, so make the lock recursive. 446 */ 447 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 448 449 if (SCARG(uap, flags) & MNT_GETARGS) { 450 error = mount_getargs(l, vp, SCARG(uap, path), 451 SCARG(uap, flags), SCARG(uap, data), &nd); 452 vput(vp); 453 } else if (SCARG(uap, flags) & MNT_UPDATE) { 454 error = mount_update(l, vp, SCARG(uap, path), 455 SCARG(uap, flags), SCARG(uap, data), &nd); 456 vput(vp); 457 } else { 458 /* Locking is handled internally in mount_domount(). */ 459 error = mount_domount(l, vp, SCARG(uap, type), 460 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 461 } 462 463 return (error); 464 } 465 466 /* 467 * Scan all active processes to see if any of them have a current 468 * or root directory onto which the new filesystem has just been 469 * mounted. If so, replace them with the new mount point. 470 */ 471 void 472 checkdirs(struct vnode *olddp) 473 { 474 struct cwdinfo *cwdi; 475 struct vnode *newdp; 476 struct proc *p; 477 478 if (olddp->v_usecount == 1) 479 return; 480 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 481 panic("mount: lost mount"); 482 mutex_enter(&proclist_lock); 483 PROCLIST_FOREACH(p, &allproc) { 484 cwdi = p->p_cwdi; 485 if (!cwdi) 486 continue; 487 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 488 if (cwdi->cwdi_cdir == olddp) { 489 vrele(cwdi->cwdi_cdir); 490 VREF(newdp); 491 cwdi->cwdi_cdir = newdp; 492 } 493 if (cwdi->cwdi_rdir == olddp) { 494 vrele(cwdi->cwdi_rdir); 495 VREF(newdp); 496 cwdi->cwdi_rdir = newdp; 497 } 498 rw_exit(&cwdi->cwdi_lock); 499 } 500 mutex_exit(&proclist_lock); 501 if (rootvnode == olddp) { 502 vrele(rootvnode); 503 VREF(newdp); 504 rootvnode = newdp; 505 } 506 vput(newdp); 507 } 508 509 /* 510 * Unmount a file system. 511 * 512 * Note: unmount takes a path to the vnode mounted on as argument, 513 * not special file (as before). 514 */ 515 /* ARGSUSED */ 516 int 517 sys_unmount(struct lwp *l, void *v, register_t *retval) 518 { 519 struct sys_unmount_args /* { 520 syscallarg(const char *) path; 521 syscallarg(int) flags; 522 } */ *uap = v; 523 struct vnode *vp; 524 struct mount *mp; 525 int error; 526 struct nameidata nd; 527 528 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 529 SCARG(uap, path), l); 530 if ((error = namei(&nd)) != 0) 531 return (error); 532 vp = nd.ni_vp; 533 mp = vp->v_mount; 534 535 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 536 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 537 if (error) { 538 vput(vp); 539 return (error); 540 } 541 542 /* 543 * Don't allow unmounting the root file system. 544 */ 545 if (mp->mnt_flag & MNT_ROOTFS) { 546 vput(vp); 547 return (EINVAL); 548 } 549 550 /* 551 * Must be the root of the filesystem 552 */ 553 if ((vp->v_flag & VROOT) == 0) { 554 vput(vp); 555 return (EINVAL); 556 } 557 vput(vp); 558 559 /* 560 * XXX Freeze syncer. Must do this before locking the 561 * mount point. See dounmount() for details. 562 */ 563 mutex_enter(&syncer_mutex); 564 565 if (vfs_busy(mp, 0, 0)) { 566 mutex_exit(&syncer_mutex); 567 return (EBUSY); 568 } 569 570 return (dounmount(mp, SCARG(uap, flags), l)); 571 } 572 573 /* 574 * Do the actual file system unmount. File system is assumed to have been 575 * marked busy by the caller. 576 */ 577 int 578 dounmount(struct mount *mp, int flags, struct lwp *l) 579 { 580 struct vnode *coveredvp; 581 int error; 582 int async; 583 int used_syncer; 584 585 #if NVERIEXEC > 0 586 error = veriexec_unmountchk(mp); 587 if (error) 588 return (error); 589 #endif /* NVERIEXEC > 0 */ 590 591 mutex_enter(&mountlist_lock); 592 vfs_unbusy(mp); 593 used_syncer = (mp->mnt_syncer != NULL); 594 595 /* 596 * XXX Syncer must be frozen when we get here. This should really 597 * be done on a per-mountpoint basis, but especially the softdep 598 * code possibly called from the syncer doesn't exactly work on a 599 * per-mountpoint basis, so the softdep code would become a maze 600 * of vfs_busy() calls. 601 * 602 * The caller of dounmount() must acquire syncer_mutex because 603 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 604 * order, and we must preserve that order to avoid deadlock. 605 * 606 * So, if the file system did not use the syncer, now is 607 * the time to release the syncer_mutex. 608 */ 609 if (used_syncer == 0) 610 mutex_exit(&syncer_mutex); 611 612 mp->mnt_iflag |= IMNT_UNMOUNT; 613 mp->mnt_unmounter = l; 614 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 615 616 async = mp->mnt_flag & MNT_ASYNC; 617 mp->mnt_flag &= ~MNT_ASYNC; 618 cache_purgevfs(mp); /* remove cache entries for this file sys */ 619 if (mp->mnt_syncer != NULL) 620 vfs_deallocate_syncvnode(mp); 621 error = 0; 622 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 623 #if NFSS > 0 624 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 625 #endif 626 if (error == 0) 627 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 628 } 629 if (error == 0 || (flags & MNT_FORCE)) 630 error = VFS_UNMOUNT(mp, flags, l); 631 mutex_enter(&mountlist_lock); 632 if (error) { 633 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 634 (void) vfs_allocate_syncvnode(mp); 635 mp->mnt_iflag &= ~IMNT_UNMOUNT; 636 mp->mnt_unmounter = NULL; 637 mp->mnt_flag |= async; 638 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 639 &mountlist_lock); 640 if (used_syncer) 641 mutex_exit(&syncer_mutex); 642 mutex_enter(&mp->mnt_mutex); 643 while (mp->mnt_wcnt > 0) { 644 wakeup(mp); 645 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 646 0, &mp->mnt_mutex); 647 } 648 mutex_exit(&mp->mnt_mutex); 649 return (error); 650 } 651 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 652 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 653 coveredvp->v_mountedhere = NULL; 654 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 655 panic("unmount: dangling vnode"); 656 mp->mnt_iflag |= IMNT_GONE; 657 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 658 if (coveredvp != NULLVP) 659 vrele(coveredvp); 660 mount_finispecific(mp); 661 if (used_syncer) 662 mutex_exit(&syncer_mutex); 663 mutex_enter(&mp->mnt_mutex); 664 while (mp->mnt_wcnt > 0) { 665 wakeup(mp); 666 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 667 } 668 mutex_exit(&mp->mnt_mutex); 669 vfs_hooks_unmount(mp); 670 mutex_destroy(&mp->mnt_mutex); 671 vfs_delref(mp->mnt_op); 672 free(mp, M_MOUNT); 673 return (0); 674 } 675 676 /* 677 * Sync each mounted filesystem. 678 */ 679 #ifdef DEBUG 680 int syncprt = 0; 681 struct ctldebug debug0 = { "syncprt", &syncprt }; 682 #endif 683 684 /* ARGSUSED */ 685 int 686 sys_sync(struct lwp *l, void *v, register_t *retval) 687 { 688 struct mount *mp, *nmp; 689 int asyncflag; 690 691 if (l == NULL) 692 l = &lwp0; 693 694 mutex_enter(&mountlist_lock); 695 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 696 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 697 nmp = mp->mnt_list.cqe_prev; 698 continue; 699 } 700 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 701 asyncflag = mp->mnt_flag & MNT_ASYNC; 702 mp->mnt_flag &= ~MNT_ASYNC; 703 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 704 if (asyncflag) 705 mp->mnt_flag |= MNT_ASYNC; 706 } 707 mutex_enter(&mountlist_lock); 708 nmp = mp->mnt_list.cqe_prev; 709 vfs_unbusy(mp); 710 711 } 712 mutex_exit(&mountlist_lock); 713 #ifdef DEBUG 714 if (syncprt) 715 vfs_bufstats(); 716 #endif /* DEBUG */ 717 return (0); 718 } 719 720 /* 721 * Change filesystem quotas. 722 */ 723 /* ARGSUSED */ 724 int 725 sys_quotactl(struct lwp *l, void *v, register_t *retval) 726 { 727 struct sys_quotactl_args /* { 728 syscallarg(const char *) path; 729 syscallarg(int) cmd; 730 syscallarg(int) uid; 731 syscallarg(void *) arg; 732 } */ *uap = v; 733 struct mount *mp; 734 int error; 735 struct nameidata nd; 736 737 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 738 if ((error = namei(&nd)) != 0) 739 return (error); 740 mp = nd.ni_vp->v_mount; 741 vrele(nd.ni_vp); 742 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 743 SCARG(uap, arg), l); 744 return (error); 745 } 746 747 int 748 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 749 int root) 750 { 751 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 752 int error = 0; 753 754 /* 755 * If MNT_NOWAIT or MNT_LAZY is specified, do not 756 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 757 * overrides MNT_NOWAIT. 758 */ 759 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 760 (flags != MNT_WAIT && flags != 0)) { 761 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 762 goto done; 763 } 764 765 /* Get the filesystem stats now */ 766 memset(sp, 0, sizeof(*sp)); 767 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 768 return error; 769 } 770 771 if (cwdi->cwdi_rdir == NULL) 772 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 773 done: 774 if (cwdi->cwdi_rdir != NULL) { 775 size_t len; 776 char *bp; 777 char *path = PNBUF_GET(); 778 779 bp = path + MAXPATHLEN; 780 *--bp = '\0'; 781 rw_enter(&cwdi->cwdi_lock, RW_READER); 782 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 783 MAXPATHLEN / 2, 0, l); 784 rw_exit(&cwdi->cwdi_lock); 785 if (error) { 786 PNBUF_PUT(path); 787 return error; 788 } 789 len = strlen(bp); 790 /* 791 * for mount points that are below our root, we can see 792 * them, so we fix up the pathname and return them. The 793 * rest we cannot see, so we don't allow viewing the 794 * data. 795 */ 796 if (strncmp(bp, sp->f_mntonname, len) == 0) { 797 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 798 sizeof(sp->f_mntonname)); 799 if (sp->f_mntonname[0] == '\0') 800 (void)strlcpy(sp->f_mntonname, "/", 801 sizeof(sp->f_mntonname)); 802 } else { 803 if (root) 804 (void)strlcpy(sp->f_mntonname, "/", 805 sizeof(sp->f_mntonname)); 806 else 807 error = EPERM; 808 } 809 PNBUF_PUT(path); 810 } 811 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 812 return error; 813 } 814 815 /* 816 * Get filesystem statistics by path. 817 */ 818 int 819 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 820 { 821 struct mount *mp; 822 int error; 823 struct nameidata nd; 824 825 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 826 if ((error = namei(&nd)) != 0) 827 return error; 828 mp = nd.ni_vp->v_mount; 829 error = dostatvfs(mp, sb, l, flags, 1); 830 vrele(nd.ni_vp); 831 return error; 832 } 833 834 /* ARGSUSED */ 835 int 836 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 837 { 838 struct sys_statvfs1_args /* { 839 syscallarg(const char *) path; 840 syscallarg(struct statvfs *) buf; 841 syscallarg(int) flags; 842 } */ *uap = v; 843 struct statvfs *sb; 844 int error; 845 846 sb = STATVFSBUF_GET(); 847 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 848 if (error == 0) 849 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 850 STATVFSBUF_PUT(sb); 851 return error; 852 } 853 854 /* 855 * Get filesystem statistics by fd. 856 */ 857 int 858 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 859 { 860 struct proc *p = l->l_proc; 861 struct file *fp; 862 struct mount *mp; 863 int error; 864 865 /* getvnode() will use the descriptor for us */ 866 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 867 return (error); 868 mp = ((struct vnode *)fp->f_data)->v_mount; 869 error = dostatvfs(mp, sb, l, flags, 1); 870 FILE_UNUSE(fp, l); 871 return error; 872 } 873 874 /* ARGSUSED */ 875 int 876 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 877 { 878 struct sys_fstatvfs1_args /* { 879 syscallarg(int) fd; 880 syscallarg(struct statvfs *) buf; 881 syscallarg(int) flags; 882 } */ *uap = v; 883 struct statvfs *sb; 884 int error; 885 886 sb = STATVFSBUF_GET(); 887 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 888 if (error == 0) 889 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 890 STATVFSBUF_PUT(sb); 891 return error; 892 } 893 894 895 /* 896 * Get statistics on all filesystems. 897 */ 898 int 899 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 900 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 901 register_t *retval) 902 { 903 int root = 0; 904 struct proc *p = l->l_proc; 905 struct mount *mp, *nmp; 906 struct statvfs *sb; 907 size_t count, maxcount; 908 int error = 0; 909 910 sb = STATVFSBUF_GET(); 911 maxcount = bufsize / entry_sz; 912 mutex_enter(&mountlist_lock); 913 count = 0; 914 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 915 mp = nmp) { 916 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 917 nmp = CIRCLEQ_NEXT(mp, mnt_list); 918 continue; 919 } 920 if (sfsp && count < maxcount) { 921 error = dostatvfs(mp, sb, l, flags, 0); 922 if (error) { 923 mutex_enter(&mountlist_lock); 924 nmp = CIRCLEQ_NEXT(mp, mnt_list); 925 vfs_unbusy(mp); 926 continue; 927 } 928 error = copyfn(sb, sfsp, entry_sz); 929 if (error) { 930 vfs_unbusy(mp); 931 goto out; 932 } 933 sfsp = (char *)sfsp + entry_sz; 934 root |= strcmp(sb->f_mntonname, "/") == 0; 935 } 936 count++; 937 mutex_enter(&mountlist_lock); 938 nmp = CIRCLEQ_NEXT(mp, mnt_list); 939 vfs_unbusy(mp); 940 } 941 mutex_exit(&mountlist_lock); 942 if (root == 0 && p->p_cwdi->cwdi_rdir) { 943 /* 944 * fake a root entry 945 */ 946 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 947 if (error != 0) 948 goto out; 949 if (sfsp) 950 error = copyfn(sb, sfsp, entry_sz); 951 count++; 952 } 953 if (sfsp && count > maxcount) 954 *retval = maxcount; 955 else 956 *retval = count; 957 out: 958 STATVFSBUF_PUT(sb); 959 return error; 960 } 961 962 int 963 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 964 { 965 struct sys_getvfsstat_args /* { 966 syscallarg(struct statvfs *) buf; 967 syscallarg(size_t) bufsize; 968 syscallarg(int) flags; 969 } */ *uap = v; 970 971 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 972 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 973 } 974 975 /* 976 * Change current working directory to a given file descriptor. 977 */ 978 /* ARGSUSED */ 979 int 980 sys_fchdir(struct lwp *l, void *v, register_t *retval) 981 { 982 struct sys_fchdir_args /* { 983 syscallarg(int) fd; 984 } */ *uap = v; 985 struct proc *p = l->l_proc; 986 struct filedesc *fdp = p->p_fd; 987 struct cwdinfo *cwdi; 988 struct vnode *vp, *tdp; 989 struct mount *mp; 990 struct file *fp; 991 int error; 992 993 /* getvnode() will use the descriptor for us */ 994 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 995 return (error); 996 vp = (struct vnode *)fp->f_data; 997 998 VREF(vp); 999 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1000 if (vp->v_type != VDIR) 1001 error = ENOTDIR; 1002 else 1003 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1004 if (error) { 1005 vput(vp); 1006 goto out; 1007 } 1008 while ((mp = vp->v_mountedhere) != NULL) { 1009 if (vfs_busy(mp, 0, 0)) 1010 continue; 1011 1012 vput(vp); 1013 error = VFS_ROOT(mp, &tdp); 1014 vfs_unbusy(mp); 1015 if (error) 1016 goto out; 1017 vp = tdp; 1018 } 1019 VOP_UNLOCK(vp, 0); 1020 1021 /* 1022 * Disallow changing to a directory not under the process's 1023 * current root directory (if there is one). 1024 */ 1025 cwdi = p->p_cwdi; 1026 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1027 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1028 vrele(vp); 1029 error = EPERM; /* operation not permitted */ 1030 } else { 1031 vrele(cwdi->cwdi_cdir); 1032 cwdi->cwdi_cdir = vp; 1033 } 1034 rw_exit(&cwdi->cwdi_lock); 1035 1036 out: 1037 FILE_UNUSE(fp, l); 1038 return (error); 1039 } 1040 1041 /* 1042 * Change this process's notion of the root directory to a given file 1043 * descriptor. 1044 */ 1045 int 1046 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1047 { 1048 struct sys_fchroot_args *uap = v; 1049 struct proc *p = l->l_proc; 1050 struct filedesc *fdp = p->p_fd; 1051 struct cwdinfo *cwdi; 1052 struct vnode *vp; 1053 struct file *fp; 1054 int error; 1055 1056 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1057 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1058 return error; 1059 /* getvnode() will use the descriptor for us */ 1060 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1061 return error; 1062 vp = (struct vnode *) fp->f_data; 1063 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1064 if (vp->v_type != VDIR) 1065 error = ENOTDIR; 1066 else 1067 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1068 VOP_UNLOCK(vp, 0); 1069 if (error) 1070 goto out; 1071 VREF(vp); 1072 1073 /* 1074 * Prevent escaping from chroot by putting the root under 1075 * the working directory. Silently chdir to / if we aren't 1076 * already there. 1077 */ 1078 cwdi = p->p_cwdi; 1079 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1080 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1081 /* 1082 * XXX would be more failsafe to change directory to a 1083 * deadfs node here instead 1084 */ 1085 vrele(cwdi->cwdi_cdir); 1086 VREF(vp); 1087 cwdi->cwdi_cdir = vp; 1088 } 1089 1090 if (cwdi->cwdi_rdir != NULL) 1091 vrele(cwdi->cwdi_rdir); 1092 cwdi->cwdi_rdir = vp; 1093 rw_exit(&cwdi->cwdi_lock); 1094 1095 out: 1096 FILE_UNUSE(fp, l); 1097 return (error); 1098 } 1099 1100 /* 1101 * Change current working directory (``.''). 1102 */ 1103 /* ARGSUSED */ 1104 int 1105 sys_chdir(struct lwp *l, void *v, register_t *retval) 1106 { 1107 struct sys_chdir_args /* { 1108 syscallarg(const char *) path; 1109 } */ *uap = v; 1110 struct proc *p = l->l_proc; 1111 struct cwdinfo *cwdi; 1112 int error; 1113 struct nameidata nd; 1114 1115 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1116 SCARG(uap, path), l); 1117 if ((error = change_dir(&nd, l)) != 0) 1118 return (error); 1119 cwdi = p->p_cwdi; 1120 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1121 vrele(cwdi->cwdi_cdir); 1122 cwdi->cwdi_cdir = nd.ni_vp; 1123 rw_exit(&cwdi->cwdi_lock); 1124 return (0); 1125 } 1126 1127 /* 1128 * Change notion of root (``/'') directory. 1129 */ 1130 /* ARGSUSED */ 1131 int 1132 sys_chroot(struct lwp *l, void *v, register_t *retval) 1133 { 1134 struct sys_chroot_args /* { 1135 syscallarg(const char *) path; 1136 } */ *uap = v; 1137 struct proc *p = l->l_proc; 1138 struct cwdinfo *cwdi; 1139 struct vnode *vp; 1140 int error; 1141 struct nameidata nd; 1142 1143 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1144 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1145 return (error); 1146 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1147 SCARG(uap, path), l); 1148 if ((error = change_dir(&nd, l)) != 0) 1149 return (error); 1150 1151 cwdi = p->p_cwdi; 1152 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1153 if (cwdi->cwdi_rdir != NULL) 1154 vrele(cwdi->cwdi_rdir); 1155 vp = nd.ni_vp; 1156 cwdi->cwdi_rdir = vp; 1157 1158 /* 1159 * Prevent escaping from chroot by putting the root under 1160 * the working directory. Silently chdir to / if we aren't 1161 * already there. 1162 */ 1163 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1164 /* 1165 * XXX would be more failsafe to change directory to a 1166 * deadfs node here instead 1167 */ 1168 vrele(cwdi->cwdi_cdir); 1169 VREF(vp); 1170 cwdi->cwdi_cdir = vp; 1171 } 1172 rw_exit(&cwdi->cwdi_lock); 1173 1174 return (0); 1175 } 1176 1177 /* 1178 * Common routine for chroot and chdir. 1179 */ 1180 static int 1181 change_dir(struct nameidata *ndp, struct lwp *l) 1182 { 1183 struct vnode *vp; 1184 int error; 1185 1186 if ((error = namei(ndp)) != 0) 1187 return (error); 1188 vp = ndp->ni_vp; 1189 if (vp->v_type != VDIR) 1190 error = ENOTDIR; 1191 else 1192 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1193 1194 if (error) 1195 vput(vp); 1196 else 1197 VOP_UNLOCK(vp, 0); 1198 return (error); 1199 } 1200 1201 /* 1202 * Check permissions, allocate an open file structure, 1203 * and call the device open routine if any. 1204 */ 1205 int 1206 sys_open(struct lwp *l, void *v, register_t *retval) 1207 { 1208 struct sys_open_args /* { 1209 syscallarg(const char *) path; 1210 syscallarg(int) flags; 1211 syscallarg(int) mode; 1212 } */ *uap = v; 1213 struct proc *p = l->l_proc; 1214 struct cwdinfo *cwdi = p->p_cwdi; 1215 struct filedesc *fdp = p->p_fd; 1216 struct file *fp; 1217 struct vnode *vp; 1218 int flags, cmode; 1219 int type, indx, error; 1220 struct flock lf; 1221 struct nameidata nd; 1222 1223 flags = FFLAGS(SCARG(uap, flags)); 1224 if ((flags & (FREAD | FWRITE)) == 0) 1225 return (EINVAL); 1226 /* falloc() will use the file descriptor for us */ 1227 if ((error = falloc(l, &fp, &indx)) != 0) 1228 return (error); 1229 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1230 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1231 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1232 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1233 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1234 rw_enter(&fdp->fd_lock, RW_WRITER); 1235 FILE_UNUSE(fp, l); 1236 fdp->fd_ofiles[indx] = NULL; 1237 rw_exit(&fdp->fd_lock); 1238 ffree(fp); 1239 if ((error == EDUPFD || error == EMOVEFD) && 1240 l->l_dupfd >= 0 && /* XXX from fdopen */ 1241 (error = 1242 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1243 *retval = indx; 1244 return (0); 1245 } 1246 if (error == ERESTART) 1247 error = EINTR; 1248 fdremove(fdp, indx); 1249 return (error); 1250 } 1251 l->l_dupfd = 0; 1252 vp = nd.ni_vp; 1253 fp->f_flag = flags & FMASK; 1254 fp->f_type = DTYPE_VNODE; 1255 fp->f_ops = &vnops; 1256 fp->f_data = vp; 1257 if (flags & (O_EXLOCK | O_SHLOCK)) { 1258 lf.l_whence = SEEK_SET; 1259 lf.l_start = 0; 1260 lf.l_len = 0; 1261 if (flags & O_EXLOCK) 1262 lf.l_type = F_WRLCK; 1263 else 1264 lf.l_type = F_RDLCK; 1265 type = F_FLOCK; 1266 if ((flags & FNONBLOCK) == 0) 1267 type |= F_WAIT; 1268 VOP_UNLOCK(vp, 0); 1269 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1270 if (error) { 1271 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1272 FILE_UNUSE(fp, l); 1273 ffree(fp); 1274 fdremove(fdp, indx); 1275 return (error); 1276 } 1277 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1278 fp->f_flag |= FHASLOCK; 1279 } 1280 VOP_UNLOCK(vp, 0); 1281 *retval = indx; 1282 FILE_SET_MATURE(fp); 1283 FILE_UNUSE(fp, l); 1284 return (0); 1285 } 1286 1287 static void 1288 vfs__fhfree(fhandle_t *fhp) 1289 { 1290 size_t fhsize; 1291 1292 if (fhp == NULL) { 1293 return; 1294 } 1295 fhsize = FHANDLE_SIZE(fhp); 1296 kmem_free(fhp, fhsize); 1297 } 1298 1299 /* 1300 * vfs_composefh: compose a filehandle. 1301 */ 1302 1303 int 1304 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1305 { 1306 struct mount *mp; 1307 struct fid *fidp; 1308 int error; 1309 size_t needfhsize; 1310 size_t fidsize; 1311 1312 mp = vp->v_mount; 1313 fidp = NULL; 1314 if (*fh_size < FHANDLE_SIZE_MIN) { 1315 fidsize = 0; 1316 } else { 1317 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1318 if (fhp != NULL) { 1319 memset(fhp, 0, *fh_size); 1320 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1321 fidp = &fhp->fh_fid; 1322 } 1323 } 1324 error = VFS_VPTOFH(vp, fidp, &fidsize); 1325 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1326 if (error == 0 && *fh_size < needfhsize) { 1327 error = E2BIG; 1328 } 1329 *fh_size = needfhsize; 1330 return error; 1331 } 1332 1333 int 1334 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1335 { 1336 struct mount *mp; 1337 fhandle_t *fhp; 1338 size_t fhsize; 1339 size_t fidsize; 1340 int error; 1341 1342 *fhpp = NULL; 1343 mp = vp->v_mount; 1344 fidsize = 0; 1345 error = VFS_VPTOFH(vp, NULL, &fidsize); 1346 KASSERT(error != 0); 1347 if (error != E2BIG) { 1348 goto out; 1349 } 1350 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1351 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1352 if (fhp == NULL) { 1353 error = ENOMEM; 1354 goto out; 1355 } 1356 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1357 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1358 if (error == 0) { 1359 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1360 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1361 *fhpp = fhp; 1362 } else { 1363 kmem_free(fhp, fhsize); 1364 } 1365 out: 1366 return error; 1367 } 1368 1369 void 1370 vfs_composefh_free(fhandle_t *fhp) 1371 { 1372 1373 vfs__fhfree(fhp); 1374 } 1375 1376 /* 1377 * vfs_fhtovp: lookup a vnode by a filehandle. 1378 */ 1379 1380 int 1381 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1382 { 1383 struct mount *mp; 1384 int error; 1385 1386 *vpp = NULL; 1387 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1388 if (mp == NULL) { 1389 error = ESTALE; 1390 goto out; 1391 } 1392 if (mp->mnt_op->vfs_fhtovp == NULL) { 1393 error = EOPNOTSUPP; 1394 goto out; 1395 } 1396 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1397 out: 1398 return error; 1399 } 1400 1401 /* 1402 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1403 * the needed size. 1404 */ 1405 1406 int 1407 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1408 { 1409 fhandle_t *fhp; 1410 int error; 1411 1412 *fhpp = NULL; 1413 if (fhsize > FHANDLE_SIZE_MAX) { 1414 return EINVAL; 1415 } 1416 if (fhsize < FHANDLE_SIZE_MIN) { 1417 return EINVAL; 1418 } 1419 again: 1420 fhp = kmem_alloc(fhsize, KM_SLEEP); 1421 if (fhp == NULL) { 1422 return ENOMEM; 1423 } 1424 error = copyin(ufhp, fhp, fhsize); 1425 if (error == 0) { 1426 /* XXX this check shouldn't be here */ 1427 if (FHANDLE_SIZE(fhp) == fhsize) { 1428 *fhpp = fhp; 1429 return 0; 1430 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1431 /* 1432 * a kludge for nfsv2 padded handles. 1433 */ 1434 size_t sz; 1435 1436 sz = FHANDLE_SIZE(fhp); 1437 kmem_free(fhp, fhsize); 1438 fhsize = sz; 1439 goto again; 1440 } else { 1441 /* 1442 * userland told us wrong size. 1443 */ 1444 error = EINVAL; 1445 } 1446 } 1447 kmem_free(fhp, fhsize); 1448 return error; 1449 } 1450 1451 void 1452 vfs_copyinfh_free(fhandle_t *fhp) 1453 { 1454 1455 vfs__fhfree(fhp); 1456 } 1457 1458 /* 1459 * Get file handle system call 1460 */ 1461 int 1462 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1463 { 1464 struct sys___getfh30_args /* { 1465 syscallarg(char *) fname; 1466 syscallarg(fhandle_t *) fhp; 1467 syscallarg(size_t *) fh_size; 1468 } */ *uap = v; 1469 struct vnode *vp; 1470 fhandle_t *fh; 1471 int error; 1472 struct nameidata nd; 1473 size_t sz; 1474 size_t usz; 1475 1476 /* 1477 * Must be super user 1478 */ 1479 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1480 0, NULL, NULL, NULL); 1481 if (error) 1482 return (error); 1483 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1484 SCARG(uap, fname), l); 1485 error = namei(&nd); 1486 if (error) 1487 return (error); 1488 vp = nd.ni_vp; 1489 error = vfs_composefh_alloc(vp, &fh); 1490 vput(vp); 1491 if (error != 0) { 1492 goto out; 1493 } 1494 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1495 if (error != 0) { 1496 goto out; 1497 } 1498 sz = FHANDLE_SIZE(fh); 1499 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1500 if (error != 0) { 1501 goto out; 1502 } 1503 if (usz >= sz) { 1504 error = copyout(fh, SCARG(uap, fhp), sz); 1505 } else { 1506 error = E2BIG; 1507 } 1508 out: 1509 vfs_composefh_free(fh); 1510 return (error); 1511 } 1512 1513 /* 1514 * Open a file given a file handle. 1515 * 1516 * Check permissions, allocate an open file structure, 1517 * and call the device open routine if any. 1518 */ 1519 1520 int 1521 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1522 register_t *retval) 1523 { 1524 struct filedesc *fdp = l->l_proc->p_fd; 1525 struct file *fp; 1526 struct vnode *vp = NULL; 1527 kauth_cred_t cred = l->l_cred; 1528 struct file *nfp; 1529 int type, indx, error=0; 1530 struct flock lf; 1531 struct vattr va; 1532 fhandle_t *fh; 1533 int flags; 1534 1535 /* 1536 * Must be super user 1537 */ 1538 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1539 0, NULL, NULL, NULL))) 1540 return (error); 1541 1542 flags = FFLAGS(oflags); 1543 if ((flags & (FREAD | FWRITE)) == 0) 1544 return (EINVAL); 1545 if ((flags & O_CREAT)) 1546 return (EINVAL); 1547 /* falloc() will use the file descriptor for us */ 1548 if ((error = falloc(l, &nfp, &indx)) != 0) 1549 return (error); 1550 fp = nfp; 1551 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1552 if (error != 0) { 1553 goto bad; 1554 } 1555 error = vfs_fhtovp(fh, &vp); 1556 if (error != 0) { 1557 goto bad; 1558 } 1559 1560 /* Now do an effective vn_open */ 1561 1562 if (vp->v_type == VSOCK) { 1563 error = EOPNOTSUPP; 1564 goto bad; 1565 } 1566 if (flags & FREAD) { 1567 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1568 goto bad; 1569 } 1570 if (flags & (FWRITE | O_TRUNC)) { 1571 if (vp->v_type == VDIR) { 1572 error = EISDIR; 1573 goto bad; 1574 } 1575 if ((error = vn_writechk(vp)) != 0 || 1576 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1577 goto bad; 1578 } 1579 if (flags & O_TRUNC) { 1580 VOP_UNLOCK(vp, 0); /* XXX */ 1581 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1582 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1583 VATTR_NULL(&va); 1584 va.va_size = 0; 1585 error = VOP_SETATTR(vp, &va, cred, l); 1586 if (error) 1587 goto bad; 1588 } 1589 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1590 goto bad; 1591 if (vp->v_type == VREG && 1592 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1593 error = EIO; 1594 goto bad; 1595 } 1596 if (flags & FWRITE) 1597 vp->v_writecount++; 1598 1599 /* done with modified vn_open, now finish what sys_open does. */ 1600 1601 fp->f_flag = flags & FMASK; 1602 fp->f_type = DTYPE_VNODE; 1603 fp->f_ops = &vnops; 1604 fp->f_data = vp; 1605 if (flags & (O_EXLOCK | O_SHLOCK)) { 1606 lf.l_whence = SEEK_SET; 1607 lf.l_start = 0; 1608 lf.l_len = 0; 1609 if (flags & O_EXLOCK) 1610 lf.l_type = F_WRLCK; 1611 else 1612 lf.l_type = F_RDLCK; 1613 type = F_FLOCK; 1614 if ((flags & FNONBLOCK) == 0) 1615 type |= F_WAIT; 1616 VOP_UNLOCK(vp, 0); 1617 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1618 if (error) { 1619 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1620 FILE_UNUSE(fp, l); 1621 ffree(fp); 1622 fdremove(fdp, indx); 1623 return (error); 1624 } 1625 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1626 fp->f_flag |= FHASLOCK; 1627 } 1628 VOP_UNLOCK(vp, 0); 1629 *retval = indx; 1630 FILE_SET_MATURE(fp); 1631 FILE_UNUSE(fp, l); 1632 vfs_copyinfh_free(fh); 1633 return (0); 1634 1635 bad: 1636 FILE_UNUSE(fp, l); 1637 ffree(fp); 1638 fdremove(fdp, indx); 1639 if (vp != NULL) 1640 vput(vp); 1641 vfs_copyinfh_free(fh); 1642 return (error); 1643 } 1644 1645 int 1646 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1647 { 1648 struct sys___fhopen40_args /* { 1649 syscallarg(const void *) fhp; 1650 syscallarg(size_t) fh_size; 1651 syscallarg(int) flags; 1652 } */ *uap = v; 1653 1654 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1655 SCARG(uap, flags), retval); 1656 } 1657 1658 int 1659 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1660 { 1661 int error; 1662 fhandle_t *fh; 1663 struct vnode *vp; 1664 1665 /* 1666 * Must be super user 1667 */ 1668 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1669 0, NULL, NULL, NULL))) 1670 return (error); 1671 1672 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1673 if (error != 0) 1674 return error; 1675 1676 error = vfs_fhtovp(fh, &vp); 1677 vfs_copyinfh_free(fh); 1678 if (error != 0) 1679 return error; 1680 1681 error = vn_stat(vp, sb, l); 1682 vput(vp); 1683 return error; 1684 } 1685 1686 1687 /* ARGSUSED */ 1688 int 1689 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1690 { 1691 struct sys___fhstat40_args /* { 1692 syscallarg(const void *) fhp; 1693 syscallarg(size_t) fh_size; 1694 syscallarg(struct stat *) sb; 1695 } */ *uap = v; 1696 struct stat sb; 1697 int error; 1698 1699 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1700 if (error) 1701 return error; 1702 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 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 char *path; 1775 const char *cpath; 1776 enum uio_seg seg = UIO_USERSPACE; 1777 1778 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1779 0, NULL, NULL, NULL)) != 0) 1780 return (error); 1781 1782 optype = VOP_MKNOD_DESCOFFSET; 1783 1784 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1785 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1786 1787 if ((error = namei(&nd)) != 0) 1788 goto out; 1789 vp = nd.ni_vp; 1790 if (vp != NULL) 1791 error = EEXIST; 1792 else { 1793 VATTR_NULL(&vattr); 1794 /* We will read cwdi->cwdi_cmask unlocked. */ 1795 vattr.va_mode = 1796 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1797 vattr.va_rdev = SCARG(uap, dev); 1798 1799 switch (SCARG(uap, mode) & S_IFMT) { 1800 case S_IFMT: /* used by badsect to flag bad sectors */ 1801 vattr.va_type = VBAD; 1802 break; 1803 case S_IFCHR: 1804 vattr.va_type = VCHR; 1805 break; 1806 case S_IFBLK: 1807 vattr.va_type = VBLK; 1808 break; 1809 case S_IFWHT: 1810 optype = VOP_WHITEOUT_DESCOFFSET; 1811 break; 1812 case S_IFREG: 1813 #if NVERIEXEC > 0 1814 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1815 O_CREAT); 1816 #endif /* NVERIEXEC > 0 */ 1817 vattr.va_type = VREG; 1818 vattr.va_rdev = VNOVAL; 1819 optype = VOP_CREATE_DESCOFFSET; 1820 break; 1821 default: 1822 error = EINVAL; 1823 break; 1824 } 1825 } 1826 if (!error) { 1827 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1828 switch (optype) { 1829 case VOP_WHITEOUT_DESCOFFSET: 1830 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1831 if (error) 1832 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1833 vput(nd.ni_dvp); 1834 break; 1835 1836 case VOP_MKNOD_DESCOFFSET: 1837 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1838 &nd.ni_cnd, &vattr); 1839 if (error == 0) 1840 vput(nd.ni_vp); 1841 break; 1842 1843 case VOP_CREATE_DESCOFFSET: 1844 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1845 &nd.ni_cnd, &vattr); 1846 if (error == 0) 1847 vput(nd.ni_vp); 1848 break; 1849 } 1850 } else { 1851 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1852 if (nd.ni_dvp == vp) 1853 vrele(nd.ni_dvp); 1854 else 1855 vput(nd.ni_dvp); 1856 if (vp) 1857 vrele(vp); 1858 } 1859 out: 1860 VERIEXEC_PATH_PUT(path); 1861 return (error); 1862 } 1863 1864 /* 1865 * Create a named pipe. 1866 */ 1867 /* ARGSUSED */ 1868 int 1869 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1870 { 1871 struct sys_mkfifo_args /* { 1872 syscallarg(const char *) path; 1873 syscallarg(int) mode; 1874 } */ *uap = v; 1875 struct proc *p = l->l_proc; 1876 struct vattr vattr; 1877 int error; 1878 struct nameidata nd; 1879 1880 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1881 if ((error = namei(&nd)) != 0) 1882 return (error); 1883 if (nd.ni_vp != NULL) { 1884 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1885 if (nd.ni_dvp == nd.ni_vp) 1886 vrele(nd.ni_dvp); 1887 else 1888 vput(nd.ni_dvp); 1889 vrele(nd.ni_vp); 1890 return (EEXIST); 1891 } 1892 VATTR_NULL(&vattr); 1893 vattr.va_type = VFIFO; 1894 /* We will read cwdi->cwdi_cmask unlocked. */ 1895 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1896 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1897 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1898 if (error == 0) 1899 vput(nd.ni_vp); 1900 return (error); 1901 } 1902 1903 /* 1904 * Make a hard file link. 1905 */ 1906 /* ARGSUSED */ 1907 int 1908 sys_link(struct lwp *l, void *v, register_t *retval) 1909 { 1910 struct sys_link_args /* { 1911 syscallarg(const char *) path; 1912 syscallarg(const char *) link; 1913 } */ *uap = v; 1914 struct vnode *vp; 1915 struct nameidata nd; 1916 int error; 1917 1918 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1919 if ((error = namei(&nd)) != 0) 1920 return (error); 1921 vp = nd.ni_vp; 1922 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1923 if ((error = namei(&nd)) != 0) 1924 goto out; 1925 if (nd.ni_vp) { 1926 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1927 if (nd.ni_dvp == nd.ni_vp) 1928 vrele(nd.ni_dvp); 1929 else 1930 vput(nd.ni_dvp); 1931 vrele(nd.ni_vp); 1932 error = EEXIST; 1933 goto out; 1934 } 1935 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1936 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1937 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1938 out: 1939 vrele(vp); 1940 return (error); 1941 } 1942 1943 /* 1944 * Make a symbolic link. 1945 */ 1946 /* ARGSUSED */ 1947 int 1948 sys_symlink(struct lwp *l, void *v, register_t *retval) 1949 { 1950 struct sys_symlink_args /* { 1951 syscallarg(const char *) path; 1952 syscallarg(const char *) link; 1953 } */ *uap = v; 1954 struct proc *p = l->l_proc; 1955 struct vattr vattr; 1956 char *path; 1957 int error; 1958 struct nameidata nd; 1959 1960 path = PNBUF_GET(); 1961 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1962 if (error) 1963 goto out; 1964 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1965 if ((error = namei(&nd)) != 0) 1966 goto out; 1967 if (nd.ni_vp) { 1968 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1969 if (nd.ni_dvp == nd.ni_vp) 1970 vrele(nd.ni_dvp); 1971 else 1972 vput(nd.ni_dvp); 1973 vrele(nd.ni_vp); 1974 error = EEXIST; 1975 goto out; 1976 } 1977 VATTR_NULL(&vattr); 1978 vattr.va_type = VLNK; 1979 /* We will read cwdi->cwdi_cmask unlocked. */ 1980 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1981 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1982 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1983 if (error == 0) 1984 vput(nd.ni_vp); 1985 out: 1986 PNBUF_PUT(path); 1987 return (error); 1988 } 1989 1990 /* 1991 * Delete a whiteout from the filesystem. 1992 */ 1993 /* ARGSUSED */ 1994 int 1995 sys_undelete(struct lwp *l, void *v, register_t *retval) 1996 { 1997 struct sys_undelete_args /* { 1998 syscallarg(const char *) path; 1999 } */ *uap = v; 2000 int error; 2001 struct nameidata nd; 2002 2003 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 2004 SCARG(uap, path), l); 2005 error = namei(&nd); 2006 if (error) 2007 return (error); 2008 2009 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2010 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2011 if (nd.ni_dvp == nd.ni_vp) 2012 vrele(nd.ni_dvp); 2013 else 2014 vput(nd.ni_dvp); 2015 if (nd.ni_vp) 2016 vrele(nd.ni_vp); 2017 return (EEXIST); 2018 } 2019 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2020 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2021 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2022 vput(nd.ni_dvp); 2023 return (error); 2024 } 2025 2026 /* 2027 * Delete a name from the filesystem. 2028 */ 2029 /* ARGSUSED */ 2030 int 2031 sys_unlink(struct lwp *l, void *v, register_t *retval) 2032 { 2033 struct sys_unlink_args /* { 2034 syscallarg(const char *) path; 2035 } */ *uap = v; 2036 struct vnode *vp; 2037 int error; 2038 struct nameidata nd; 2039 char *path; 2040 const char *cpath; 2041 enum uio_seg seg = UIO_USERSPACE; 2042 2043 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2044 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2045 2046 if ((error = namei(&nd)) != 0) 2047 goto out; 2048 vp = nd.ni_vp; 2049 2050 /* 2051 * The root of a mounted filesystem cannot be deleted. 2052 */ 2053 if (vp->v_flag & VROOT) { 2054 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2055 if (nd.ni_dvp == vp) 2056 vrele(nd.ni_dvp); 2057 else 2058 vput(nd.ni_dvp); 2059 vput(vp); 2060 error = EBUSY; 2061 goto out; 2062 } 2063 2064 #if NVERIEXEC > 0 2065 /* Handle remove requests for veriexec entries. */ 2066 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2067 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2068 if (nd.ni_dvp == vp) 2069 vrele(nd.ni_dvp); 2070 else 2071 vput(nd.ni_dvp); 2072 vput(vp); 2073 goto out; 2074 } 2075 #endif /* NVERIEXEC > 0 */ 2076 2077 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2078 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2079 #ifdef FILEASSOC 2080 (void)fileassoc_file_delete(vp); 2081 #endif /* FILEASSOC */ 2082 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2083 out: 2084 VERIEXEC_PATH_PUT(path); 2085 return (error); 2086 } 2087 2088 /* 2089 * Reposition read/write file offset. 2090 */ 2091 int 2092 sys_lseek(struct lwp *l, void *v, register_t *retval) 2093 { 2094 struct sys_lseek_args /* { 2095 syscallarg(int) fd; 2096 syscallarg(int) pad; 2097 syscallarg(off_t) offset; 2098 syscallarg(int) whence; 2099 } */ *uap = v; 2100 struct proc *p = l->l_proc; 2101 kauth_cred_t cred = l->l_cred; 2102 struct filedesc *fdp = p->p_fd; 2103 struct file *fp; 2104 struct vnode *vp; 2105 struct vattr vattr; 2106 off_t newoff; 2107 int error; 2108 2109 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2110 return (EBADF); 2111 2112 vp = (struct vnode *)fp->f_data; 2113 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2114 error = ESPIPE; 2115 mutex_exit(&fp->f_lock); 2116 goto out; 2117 } 2118 2119 switch (SCARG(uap, whence)) { 2120 case SEEK_CUR: 2121 newoff = fp->f_offset + SCARG(uap, offset); 2122 FILE_USE(fp); 2123 break; 2124 case SEEK_END: 2125 FILE_USE(fp); 2126 error = VOP_GETATTR(vp, &vattr, cred, l); 2127 if (error) { 2128 FILE_UNUSE(fp, l); 2129 goto out; 2130 } 2131 newoff = SCARG(uap, offset) + vattr.va_size; 2132 break; 2133 case SEEK_SET: 2134 FILE_USE(fp); 2135 newoff = SCARG(uap, offset); 2136 break; 2137 default: 2138 mutex_exit(&fp->f_lock); 2139 error = EINVAL; 2140 goto out; 2141 } 2142 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2143 mutex_enter(&fp->f_lock); 2144 *(off_t *)retval = fp->f_offset = newoff; 2145 mutex_exit(&fp->f_lock); 2146 } 2147 FILE_UNUSE(fp, l); 2148 out: 2149 return (error); 2150 } 2151 2152 /* 2153 * Positional read system call. 2154 */ 2155 int 2156 sys_pread(struct lwp *l, void *v, register_t *retval) 2157 { 2158 struct sys_pread_args /* { 2159 syscallarg(int) fd; 2160 syscallarg(void *) buf; 2161 syscallarg(size_t) nbyte; 2162 syscallarg(off_t) offset; 2163 } */ *uap = v; 2164 struct proc *p = l->l_proc; 2165 struct filedesc *fdp = p->p_fd; 2166 struct file *fp; 2167 struct vnode *vp; 2168 off_t offset; 2169 int error, fd = SCARG(uap, fd); 2170 2171 if ((fp = fd_getfile(fdp, fd)) == NULL) 2172 return (EBADF); 2173 2174 if ((fp->f_flag & FREAD) == 0) { 2175 mutex_exit(&fp->f_lock); 2176 return (EBADF); 2177 } 2178 2179 FILE_USE(fp); 2180 2181 vp = (struct vnode *)fp->f_data; 2182 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2183 error = ESPIPE; 2184 goto out; 2185 } 2186 2187 offset = SCARG(uap, offset); 2188 2189 /* 2190 * XXX This works because no file systems actually 2191 * XXX take any action on the seek operation. 2192 */ 2193 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2194 goto out; 2195 2196 /* dofileread() will unuse the descriptor for us */ 2197 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2198 &offset, 0, retval)); 2199 2200 out: 2201 FILE_UNUSE(fp, l); 2202 return (error); 2203 } 2204 2205 /* 2206 * Positional scatter read system call. 2207 */ 2208 int 2209 sys_preadv(struct lwp *l, void *v, register_t *retval) 2210 { 2211 struct sys_preadv_args /* { 2212 syscallarg(int) fd; 2213 syscallarg(const struct iovec *) iovp; 2214 syscallarg(int) iovcnt; 2215 syscallarg(off_t) offset; 2216 } */ *uap = v; 2217 struct proc *p = l->l_proc; 2218 struct filedesc *fdp = p->p_fd; 2219 struct file *fp; 2220 struct vnode *vp; 2221 off_t offset; 2222 int error, fd = SCARG(uap, fd); 2223 2224 if ((fp = fd_getfile(fdp, fd)) == NULL) 2225 return (EBADF); 2226 2227 if ((fp->f_flag & FREAD) == 0) { 2228 mutex_exit(&fp->f_lock); 2229 return (EBADF); 2230 } 2231 2232 FILE_USE(fp); 2233 2234 vp = (struct vnode *)fp->f_data; 2235 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2236 error = ESPIPE; 2237 goto out; 2238 } 2239 2240 offset = SCARG(uap, offset); 2241 2242 /* 2243 * XXX This works because no file systems actually 2244 * XXX take any action on the seek operation. 2245 */ 2246 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2247 goto out; 2248 2249 /* dofilereadv() will unuse the descriptor for us */ 2250 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2251 &offset, 0, retval)); 2252 2253 out: 2254 FILE_UNUSE(fp, l); 2255 return (error); 2256 } 2257 2258 /* 2259 * Positional write system call. 2260 */ 2261 int 2262 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2263 { 2264 struct sys_pwrite_args /* { 2265 syscallarg(int) fd; 2266 syscallarg(const void *) buf; 2267 syscallarg(size_t) nbyte; 2268 syscallarg(off_t) offset; 2269 } */ *uap = v; 2270 struct proc *p = l->l_proc; 2271 struct filedesc *fdp = p->p_fd; 2272 struct file *fp; 2273 struct vnode *vp; 2274 off_t offset; 2275 int error, fd = SCARG(uap, fd); 2276 2277 if ((fp = fd_getfile(fdp, fd)) == NULL) 2278 return (EBADF); 2279 2280 if ((fp->f_flag & FWRITE) == 0) { 2281 mutex_exit(&fp->f_lock); 2282 return (EBADF); 2283 } 2284 2285 FILE_USE(fp); 2286 2287 vp = (struct vnode *)fp->f_data; 2288 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2289 error = ESPIPE; 2290 goto out; 2291 } 2292 2293 offset = SCARG(uap, offset); 2294 2295 /* 2296 * XXX This works because no file systems actually 2297 * XXX take any action on the seek operation. 2298 */ 2299 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2300 goto out; 2301 2302 /* dofilewrite() will unuse the descriptor for us */ 2303 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2304 &offset, 0, retval)); 2305 2306 out: 2307 FILE_UNUSE(fp, l); 2308 return (error); 2309 } 2310 2311 /* 2312 * Positional gather write system call. 2313 */ 2314 int 2315 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2316 { 2317 struct sys_pwritev_args /* { 2318 syscallarg(int) fd; 2319 syscallarg(const struct iovec *) iovp; 2320 syscallarg(int) iovcnt; 2321 syscallarg(off_t) offset; 2322 } */ *uap = v; 2323 struct proc *p = l->l_proc; 2324 struct filedesc *fdp = p->p_fd; 2325 struct file *fp; 2326 struct vnode *vp; 2327 off_t offset; 2328 int error, fd = SCARG(uap, fd); 2329 2330 if ((fp = fd_getfile(fdp, fd)) == NULL) 2331 return (EBADF); 2332 2333 if ((fp->f_flag & FWRITE) == 0) { 2334 mutex_exit(&fp->f_lock); 2335 return (EBADF); 2336 } 2337 2338 FILE_USE(fp); 2339 2340 vp = (struct vnode *)fp->f_data; 2341 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2342 error = ESPIPE; 2343 goto out; 2344 } 2345 2346 offset = SCARG(uap, offset); 2347 2348 /* 2349 * XXX This works because no file systems actually 2350 * XXX take any action on the seek operation. 2351 */ 2352 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2353 goto out; 2354 2355 /* dofilewritev() will unuse the descriptor for us */ 2356 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2357 &offset, 0, retval)); 2358 2359 out: 2360 FILE_UNUSE(fp, l); 2361 return (error); 2362 } 2363 2364 /* 2365 * Check access permissions. 2366 */ 2367 int 2368 sys_access(struct lwp *l, void *v, register_t *retval) 2369 { 2370 struct sys_access_args /* { 2371 syscallarg(const char *) path; 2372 syscallarg(int) flags; 2373 } */ *uap = v; 2374 kauth_cred_t cred; 2375 struct vnode *vp; 2376 int error, flags; 2377 struct nameidata nd; 2378 2379 cred = kauth_cred_dup(l->l_cred); 2380 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2381 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2382 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2383 SCARG(uap, path), l); 2384 /* Override default credentials */ 2385 nd.ni_cnd.cn_cred = cred; 2386 if ((error = namei(&nd)) != 0) 2387 goto out; 2388 vp = nd.ni_vp; 2389 2390 /* Flags == 0 means only check for existence. */ 2391 if (SCARG(uap, flags)) { 2392 flags = 0; 2393 if (SCARG(uap, flags) & R_OK) 2394 flags |= VREAD; 2395 if (SCARG(uap, flags) & W_OK) 2396 flags |= VWRITE; 2397 if (SCARG(uap, flags) & X_OK) 2398 flags |= VEXEC; 2399 2400 error = VOP_ACCESS(vp, flags, cred, l); 2401 if (!error && (flags & VWRITE)) 2402 error = vn_writechk(vp); 2403 } 2404 vput(vp); 2405 out: 2406 kauth_cred_free(cred); 2407 return (error); 2408 } 2409 2410 /* 2411 * Common code for all sys_stat functions, including compat versions. 2412 */ 2413 int 2414 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2415 struct stat *sb) 2416 { 2417 int error; 2418 struct nameidata nd; 2419 2420 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2421 error = namei(&nd); 2422 if (error != 0) 2423 return error; 2424 error = vn_stat(nd.ni_vp, sb, l); 2425 vput(nd.ni_vp); 2426 return error; 2427 } 2428 2429 /* 2430 * Get file status; this version follows links. 2431 */ 2432 /* ARGSUSED */ 2433 int 2434 sys___stat30(struct lwp *l, void *v, register_t *retval) 2435 { 2436 struct sys___stat30_args /* { 2437 syscallarg(const char *) path; 2438 syscallarg(struct stat *) ub; 2439 } */ *uap = v; 2440 struct stat sb; 2441 int error; 2442 2443 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2444 if (error) 2445 return error; 2446 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2447 } 2448 2449 /* 2450 * Get file status; this version does not follow links. 2451 */ 2452 /* ARGSUSED */ 2453 int 2454 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2455 { 2456 struct sys___lstat30_args /* { 2457 syscallarg(const char *) path; 2458 syscallarg(struct stat *) ub; 2459 } */ *uap = v; 2460 struct stat sb; 2461 int error; 2462 2463 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2464 if (error) 2465 return error; 2466 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2467 } 2468 2469 /* 2470 * Get configurable pathname variables. 2471 */ 2472 /* ARGSUSED */ 2473 int 2474 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2475 { 2476 struct sys_pathconf_args /* { 2477 syscallarg(const char *) path; 2478 syscallarg(int) name; 2479 } */ *uap = v; 2480 int error; 2481 struct nameidata nd; 2482 2483 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2484 SCARG(uap, path), l); 2485 if ((error = namei(&nd)) != 0) 2486 return (error); 2487 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2488 vput(nd.ni_vp); 2489 return (error); 2490 } 2491 2492 /* 2493 * Return target name of a symbolic link. 2494 */ 2495 /* ARGSUSED */ 2496 int 2497 sys_readlink(struct lwp *l, void *v, register_t *retval) 2498 { 2499 struct sys_readlink_args /* { 2500 syscallarg(const char *) path; 2501 syscallarg(char *) buf; 2502 syscallarg(size_t) count; 2503 } */ *uap = v; 2504 struct vnode *vp; 2505 struct iovec aiov; 2506 struct uio auio; 2507 int error; 2508 struct nameidata nd; 2509 2510 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2511 SCARG(uap, path), l); 2512 if ((error = namei(&nd)) != 0) 2513 return (error); 2514 vp = nd.ni_vp; 2515 if (vp->v_type != VLNK) 2516 error = EINVAL; 2517 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2518 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2519 aiov.iov_base = SCARG(uap, buf); 2520 aiov.iov_len = SCARG(uap, count); 2521 auio.uio_iov = &aiov; 2522 auio.uio_iovcnt = 1; 2523 auio.uio_offset = 0; 2524 auio.uio_rw = UIO_READ; 2525 KASSERT(l == curlwp); 2526 auio.uio_vmspace = l->l_proc->p_vmspace; 2527 auio.uio_resid = SCARG(uap, count); 2528 error = VOP_READLINK(vp, &auio, l->l_cred); 2529 } 2530 vput(vp); 2531 *retval = SCARG(uap, count) - auio.uio_resid; 2532 return (error); 2533 } 2534 2535 /* 2536 * Change flags of a file given a path name. 2537 */ 2538 /* ARGSUSED */ 2539 int 2540 sys_chflags(struct lwp *l, void *v, register_t *retval) 2541 { 2542 struct sys_chflags_args /* { 2543 syscallarg(const char *) path; 2544 syscallarg(u_long) flags; 2545 } */ *uap = v; 2546 struct vnode *vp; 2547 int error; 2548 struct nameidata nd; 2549 2550 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2551 if ((error = namei(&nd)) != 0) 2552 return (error); 2553 vp = nd.ni_vp; 2554 error = change_flags(vp, SCARG(uap, flags), l); 2555 vput(vp); 2556 return (error); 2557 } 2558 2559 /* 2560 * Change flags of a file given a file descriptor. 2561 */ 2562 /* ARGSUSED */ 2563 int 2564 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2565 { 2566 struct sys_fchflags_args /* { 2567 syscallarg(int) fd; 2568 syscallarg(u_long) flags; 2569 } */ *uap = v; 2570 struct proc *p = l->l_proc; 2571 struct vnode *vp; 2572 struct file *fp; 2573 int error; 2574 2575 /* getvnode() will use the descriptor for us */ 2576 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2577 return (error); 2578 vp = (struct vnode *)fp->f_data; 2579 error = change_flags(vp, SCARG(uap, flags), l); 2580 VOP_UNLOCK(vp, 0); 2581 FILE_UNUSE(fp, l); 2582 return (error); 2583 } 2584 2585 /* 2586 * Change flags of a file given a path name; this version does 2587 * not follow links. 2588 */ 2589 int 2590 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2591 { 2592 struct sys_lchflags_args /* { 2593 syscallarg(const char *) path; 2594 syscallarg(u_long) flags; 2595 } */ *uap = v; 2596 struct vnode *vp; 2597 int error; 2598 struct nameidata nd; 2599 2600 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2601 if ((error = namei(&nd)) != 0) 2602 return (error); 2603 vp = nd.ni_vp; 2604 error = change_flags(vp, SCARG(uap, flags), l); 2605 vput(vp); 2606 return (error); 2607 } 2608 2609 /* 2610 * Common routine to change flags of a file. 2611 */ 2612 int 2613 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2614 { 2615 struct vattr vattr; 2616 int error; 2617 2618 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2619 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2620 /* 2621 * Non-superusers cannot change the flags on devices, even if they 2622 * own them. 2623 */ 2624 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2625 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2626 goto out; 2627 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2628 error = EINVAL; 2629 goto out; 2630 } 2631 } 2632 VATTR_NULL(&vattr); 2633 vattr.va_flags = flags; 2634 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2635 out: 2636 return (error); 2637 } 2638 2639 /* 2640 * Change mode of a file given path name; this version follows links. 2641 */ 2642 /* ARGSUSED */ 2643 int 2644 sys_chmod(struct lwp *l, void *v, register_t *retval) 2645 { 2646 struct sys_chmod_args /* { 2647 syscallarg(const char *) path; 2648 syscallarg(int) mode; 2649 } */ *uap = v; 2650 int error; 2651 struct nameidata nd; 2652 2653 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2654 if ((error = namei(&nd)) != 0) 2655 return (error); 2656 2657 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2658 2659 vrele(nd.ni_vp); 2660 return (error); 2661 } 2662 2663 /* 2664 * Change mode of a file given a file descriptor. 2665 */ 2666 /* ARGSUSED */ 2667 int 2668 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2669 { 2670 struct sys_fchmod_args /* { 2671 syscallarg(int) fd; 2672 syscallarg(int) mode; 2673 } */ *uap = v; 2674 struct proc *p = l->l_proc; 2675 struct file *fp; 2676 int error; 2677 2678 /* getvnode() will use the descriptor for us */ 2679 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2680 return (error); 2681 2682 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2683 FILE_UNUSE(fp, l); 2684 return (error); 2685 } 2686 2687 /* 2688 * Change mode of a file given path name; this version does not follow links. 2689 */ 2690 /* ARGSUSED */ 2691 int 2692 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2693 { 2694 struct sys_lchmod_args /* { 2695 syscallarg(const char *) path; 2696 syscallarg(int) mode; 2697 } */ *uap = v; 2698 int error; 2699 struct nameidata nd; 2700 2701 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2702 if ((error = namei(&nd)) != 0) 2703 return (error); 2704 2705 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2706 2707 vrele(nd.ni_vp); 2708 return (error); 2709 } 2710 2711 /* 2712 * Common routine to set mode given a vnode. 2713 */ 2714 static int 2715 change_mode(struct vnode *vp, int mode, struct lwp *l) 2716 { 2717 struct vattr vattr; 2718 int error; 2719 2720 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2721 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2722 VATTR_NULL(&vattr); 2723 vattr.va_mode = mode & ALLPERMS; 2724 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2725 VOP_UNLOCK(vp, 0); 2726 return (error); 2727 } 2728 2729 /* 2730 * Set ownership given a path name; this version follows links. 2731 */ 2732 /* ARGSUSED */ 2733 int 2734 sys_chown(struct lwp *l, void *v, register_t *retval) 2735 { 2736 struct sys_chown_args /* { 2737 syscallarg(const char *) path; 2738 syscallarg(uid_t) uid; 2739 syscallarg(gid_t) gid; 2740 } */ *uap = v; 2741 int error; 2742 struct nameidata nd; 2743 2744 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2745 if ((error = namei(&nd)) != 0) 2746 return (error); 2747 2748 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2749 2750 vrele(nd.ni_vp); 2751 return (error); 2752 } 2753 2754 /* 2755 * Set ownership given a path name; this version follows links. 2756 * Provides POSIX semantics. 2757 */ 2758 /* ARGSUSED */ 2759 int 2760 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2761 { 2762 struct sys_chown_args /* { 2763 syscallarg(const char *) path; 2764 syscallarg(uid_t) uid; 2765 syscallarg(gid_t) gid; 2766 } */ *uap = v; 2767 int error; 2768 struct nameidata nd; 2769 2770 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2771 if ((error = namei(&nd)) != 0) 2772 return (error); 2773 2774 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2775 2776 vrele(nd.ni_vp); 2777 return (error); 2778 } 2779 2780 /* 2781 * Set ownership given a file descriptor. 2782 */ 2783 /* ARGSUSED */ 2784 int 2785 sys_fchown(struct lwp *l, void *v, register_t *retval) 2786 { 2787 struct sys_fchown_args /* { 2788 syscallarg(int) fd; 2789 syscallarg(uid_t) uid; 2790 syscallarg(gid_t) gid; 2791 } */ *uap = v; 2792 struct proc *p = l->l_proc; 2793 int error; 2794 struct file *fp; 2795 2796 /* getvnode() will use the descriptor for us */ 2797 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2798 return (error); 2799 2800 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2801 SCARG(uap, gid), l, 0); 2802 FILE_UNUSE(fp, l); 2803 return (error); 2804 } 2805 2806 /* 2807 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2808 */ 2809 /* ARGSUSED */ 2810 int 2811 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2812 { 2813 struct sys_fchown_args /* { 2814 syscallarg(int) fd; 2815 syscallarg(uid_t) uid; 2816 syscallarg(gid_t) gid; 2817 } */ *uap = v; 2818 struct proc *p = l->l_proc; 2819 int error; 2820 struct file *fp; 2821 2822 /* getvnode() will use the descriptor for us */ 2823 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2824 return (error); 2825 2826 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2827 SCARG(uap, gid), l, 1); 2828 FILE_UNUSE(fp, l); 2829 return (error); 2830 } 2831 2832 /* 2833 * Set ownership given a path name; this version does not follow links. 2834 */ 2835 /* ARGSUSED */ 2836 int 2837 sys_lchown(struct lwp *l, void *v, register_t *retval) 2838 { 2839 struct sys_lchown_args /* { 2840 syscallarg(const char *) path; 2841 syscallarg(uid_t) uid; 2842 syscallarg(gid_t) gid; 2843 } */ *uap = v; 2844 int error; 2845 struct nameidata nd; 2846 2847 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2848 if ((error = namei(&nd)) != 0) 2849 return (error); 2850 2851 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2852 2853 vrele(nd.ni_vp); 2854 return (error); 2855 } 2856 2857 /* 2858 * Set ownership given a path name; this version does not follow links. 2859 * Provides POSIX/XPG semantics. 2860 */ 2861 /* ARGSUSED */ 2862 int 2863 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2864 { 2865 struct sys_lchown_args /* { 2866 syscallarg(const char *) path; 2867 syscallarg(uid_t) uid; 2868 syscallarg(gid_t) gid; 2869 } */ *uap = v; 2870 int error; 2871 struct nameidata nd; 2872 2873 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2874 if ((error = namei(&nd)) != 0) 2875 return (error); 2876 2877 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2878 2879 vrele(nd.ni_vp); 2880 return (error); 2881 } 2882 2883 /* 2884 * Common routine to set ownership given a vnode. 2885 */ 2886 static int 2887 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2888 int posix_semantics) 2889 { 2890 struct vattr vattr; 2891 mode_t newmode; 2892 int error; 2893 2894 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2895 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2896 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2897 goto out; 2898 2899 #define CHANGED(x) ((int)(x) != -1) 2900 newmode = vattr.va_mode; 2901 if (posix_semantics) { 2902 /* 2903 * POSIX/XPG semantics: if the caller is not the super-user, 2904 * clear set-user-id and set-group-id bits. Both POSIX and 2905 * the XPG consider the behaviour for calls by the super-user 2906 * implementation-defined; we leave the set-user-id and set- 2907 * group-id settings intact in that case. 2908 */ 2909 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2910 NULL) != 0) 2911 newmode &= ~(S_ISUID | S_ISGID); 2912 } else { 2913 /* 2914 * NetBSD semantics: when changing owner and/or group, 2915 * clear the respective bit(s). 2916 */ 2917 if (CHANGED(uid)) 2918 newmode &= ~S_ISUID; 2919 if (CHANGED(gid)) 2920 newmode &= ~S_ISGID; 2921 } 2922 /* Update va_mode iff altered. */ 2923 if (vattr.va_mode == newmode) 2924 newmode = VNOVAL; 2925 2926 VATTR_NULL(&vattr); 2927 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2928 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2929 vattr.va_mode = newmode; 2930 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2931 #undef CHANGED 2932 2933 out: 2934 VOP_UNLOCK(vp, 0); 2935 return (error); 2936 } 2937 2938 /* 2939 * Set the access and modification times given a path name; this 2940 * version follows links. 2941 */ 2942 /* ARGSUSED */ 2943 int 2944 sys_utimes(struct lwp *l, void *v, register_t *retval) 2945 { 2946 struct sys_utimes_args /* { 2947 syscallarg(const char *) path; 2948 syscallarg(const struct timeval *) tptr; 2949 } */ *uap = v; 2950 2951 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2952 SCARG(uap, tptr), UIO_USERSPACE); 2953 } 2954 2955 /* 2956 * Set the access and modification times given a file descriptor. 2957 */ 2958 /* ARGSUSED */ 2959 int 2960 sys_futimes(struct lwp *l, void *v, register_t *retval) 2961 { 2962 struct sys_futimes_args /* { 2963 syscallarg(int) fd; 2964 syscallarg(const struct timeval *) tptr; 2965 } */ *uap = v; 2966 int error; 2967 struct file *fp; 2968 2969 /* getvnode() will use the descriptor for us */ 2970 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2971 return (error); 2972 2973 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2974 SCARG(uap, tptr), UIO_USERSPACE); 2975 2976 FILE_UNUSE(fp, l); 2977 return (error); 2978 } 2979 2980 /* 2981 * Set the access and modification times given a path name; this 2982 * version does not follow links. 2983 */ 2984 int 2985 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2986 { 2987 struct sys_lutimes_args /* { 2988 syscallarg(const char *) path; 2989 syscallarg(const struct timeval *) tptr; 2990 } */ *uap = v; 2991 2992 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2993 SCARG(uap, tptr), UIO_USERSPACE); 2994 } 2995 2996 /* 2997 * Common routine to set access and modification times given a vnode. 2998 */ 2999 int 3000 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3001 const struct timeval *tptr, enum uio_seg seg) 3002 { 3003 struct vattr vattr; 3004 struct nameidata nd; 3005 int error; 3006 3007 VATTR_NULL(&vattr); 3008 if (tptr == NULL) { 3009 nanotime(&vattr.va_atime); 3010 vattr.va_mtime = vattr.va_atime; 3011 vattr.va_vaflags |= VA_UTIMES_NULL; 3012 } else { 3013 struct timeval tv[2]; 3014 3015 if (seg != UIO_SYSSPACE) { 3016 error = copyin(tptr, &tv, sizeof (tv)); 3017 if (error != 0) 3018 return error; 3019 tptr = tv; 3020 } 3021 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3022 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3023 } 3024 3025 if (vp == NULL) { 3026 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 3027 if ((error = namei(&nd)) != 0) 3028 return (error); 3029 vp = nd.ni_vp; 3030 } else 3031 nd.ni_vp = NULL; 3032 3033 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3034 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3035 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3036 VOP_UNLOCK(vp, 0); 3037 3038 if (nd.ni_vp != NULL) 3039 vrele(nd.ni_vp); 3040 3041 return (error); 3042 } 3043 3044 /* 3045 * Truncate a file given its path name. 3046 */ 3047 /* ARGSUSED */ 3048 int 3049 sys_truncate(struct lwp *l, void *v, register_t *retval) 3050 { 3051 struct sys_truncate_args /* { 3052 syscallarg(const char *) path; 3053 syscallarg(int) pad; 3054 syscallarg(off_t) length; 3055 } */ *uap = v; 3056 struct vnode *vp; 3057 struct vattr vattr; 3058 int error; 3059 struct nameidata nd; 3060 3061 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3062 if ((error = namei(&nd)) != 0) 3063 return (error); 3064 vp = nd.ni_vp; 3065 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3066 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3067 if (vp->v_type == VDIR) 3068 error = EISDIR; 3069 else if ((error = vn_writechk(vp)) == 0 && 3070 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3071 VATTR_NULL(&vattr); 3072 vattr.va_size = SCARG(uap, length); 3073 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3074 } 3075 vput(vp); 3076 return (error); 3077 } 3078 3079 /* 3080 * Truncate a file given a file descriptor. 3081 */ 3082 /* ARGSUSED */ 3083 int 3084 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3085 { 3086 struct sys_ftruncate_args /* { 3087 syscallarg(int) fd; 3088 syscallarg(int) pad; 3089 syscallarg(off_t) length; 3090 } */ *uap = v; 3091 struct proc *p = l->l_proc; 3092 struct vattr vattr; 3093 struct vnode *vp; 3094 struct file *fp; 3095 int error; 3096 3097 /* getvnode() will use the descriptor for us */ 3098 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3099 return (error); 3100 if ((fp->f_flag & FWRITE) == 0) { 3101 error = EINVAL; 3102 goto out; 3103 } 3104 vp = (struct vnode *)fp->f_data; 3105 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3106 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3107 if (vp->v_type == VDIR) 3108 error = EISDIR; 3109 else if ((error = vn_writechk(vp)) == 0) { 3110 VATTR_NULL(&vattr); 3111 vattr.va_size = SCARG(uap, length); 3112 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3113 } 3114 VOP_UNLOCK(vp, 0); 3115 out: 3116 FILE_UNUSE(fp, l); 3117 return (error); 3118 } 3119 3120 /* 3121 * Sync an open file. 3122 */ 3123 /* ARGSUSED */ 3124 int 3125 sys_fsync(struct lwp *l, void *v, register_t *retval) 3126 { 3127 struct sys_fsync_args /* { 3128 syscallarg(int) fd; 3129 } */ *uap = v; 3130 struct proc *p = l->l_proc; 3131 struct vnode *vp; 3132 struct file *fp; 3133 int error; 3134 3135 /* getvnode() will use the descriptor for us */ 3136 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3137 return (error); 3138 vp = (struct vnode *)fp->f_data; 3139 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3140 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3141 if (error == 0 && bioops.io_fsync != NULL && 3142 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) { 3143 KERNEL_LOCK(1, l); 3144 (*bioops.io_fsync)(vp, 0); 3145 KERNEL_UNLOCK_ONE(l); 3146 } 3147 VOP_UNLOCK(vp, 0); 3148 FILE_UNUSE(fp, l); 3149 return (error); 3150 } 3151 3152 /* 3153 * Sync a range of file data. API modeled after that found in AIX. 3154 * 3155 * FDATASYNC indicates that we need only save enough metadata to be able 3156 * to re-read the written data. Note we duplicate AIX's requirement that 3157 * the file be open for writing. 3158 */ 3159 /* ARGSUSED */ 3160 int 3161 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3162 { 3163 struct sys_fsync_range_args /* { 3164 syscallarg(int) fd; 3165 syscallarg(int) flags; 3166 syscallarg(off_t) start; 3167 syscallarg(off_t) length; 3168 } */ *uap = v; 3169 struct proc *p = l->l_proc; 3170 struct vnode *vp; 3171 struct file *fp; 3172 int flags, nflags; 3173 off_t s, e, len; 3174 int error; 3175 3176 /* getvnode() will use the descriptor for us */ 3177 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3178 return (error); 3179 3180 if ((fp->f_flag & FWRITE) == 0) { 3181 error = EBADF; 3182 goto out; 3183 } 3184 3185 flags = SCARG(uap, flags); 3186 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3187 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3188 error = EINVAL; 3189 goto out; 3190 } 3191 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3192 if (flags & FDATASYNC) 3193 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3194 else 3195 nflags = FSYNC_WAIT; 3196 if (flags & FDISKSYNC) 3197 nflags |= FSYNC_CACHE; 3198 3199 len = SCARG(uap, length); 3200 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3201 if (len) { 3202 s = SCARG(uap, start); 3203 e = s + len; 3204 if (e < s) { 3205 error = EINVAL; 3206 goto out; 3207 } 3208 } else { 3209 e = 0; 3210 s = 0; 3211 } 3212 3213 vp = (struct vnode *)fp->f_data; 3214 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3215 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3216 3217 if (error == 0 && bioops.io_fsync != NULL && 3218 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) { 3219 KERNEL_LOCK(1, l); 3220 (*bioops.io_fsync)(vp, nflags); 3221 KERNEL_UNLOCK_ONE(l); 3222 } 3223 3224 VOP_UNLOCK(vp, 0); 3225 out: 3226 FILE_UNUSE(fp, l); 3227 return (error); 3228 } 3229 3230 /* 3231 * Sync the data of an open file. 3232 */ 3233 /* ARGSUSED */ 3234 int 3235 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3236 { 3237 struct sys_fdatasync_args /* { 3238 syscallarg(int) fd; 3239 } */ *uap = v; 3240 struct proc *p = l->l_proc; 3241 struct vnode *vp; 3242 struct file *fp; 3243 int error; 3244 3245 /* getvnode() will use the descriptor for us */ 3246 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3247 return (error); 3248 if ((fp->f_flag & FWRITE) == 0) { 3249 FILE_UNUSE(fp, l); 3250 return (EBADF); 3251 } 3252 vp = (struct vnode *)fp->f_data; 3253 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3254 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3255 VOP_UNLOCK(vp, 0); 3256 FILE_UNUSE(fp, l); 3257 return (error); 3258 } 3259 3260 /* 3261 * Rename files, (standard) BSD semantics frontend. 3262 */ 3263 /* ARGSUSED */ 3264 int 3265 sys_rename(struct lwp *l, void *v, register_t *retval) 3266 { 3267 struct sys_rename_args /* { 3268 syscallarg(const char *) from; 3269 syscallarg(const char *) to; 3270 } */ *uap = v; 3271 3272 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3273 } 3274 3275 /* 3276 * Rename files, POSIX semantics frontend. 3277 */ 3278 /* ARGSUSED */ 3279 int 3280 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3281 { 3282 struct sys___posix_rename_args /* { 3283 syscallarg(const char *) from; 3284 syscallarg(const char *) to; 3285 } */ *uap = v; 3286 3287 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3288 } 3289 3290 /* 3291 * Rename files. Source and destination must either both be directories, 3292 * or both not be directories. If target is a directory, it must be empty. 3293 * If `from' and `to' refer to the same object, the value of the `retain' 3294 * argument is used to determine whether `from' will be 3295 * 3296 * (retain == 0) deleted unless `from' and `to' refer to the same 3297 * object in the file system's name space (BSD). 3298 * (retain == 1) always retained (POSIX). 3299 */ 3300 static int 3301 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3302 { 3303 struct vnode *tvp, *fvp, *tdvp; 3304 struct nameidata fromnd, tond; 3305 struct proc *p; 3306 int error; 3307 3308 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3309 from, l); 3310 if ((error = namei(&fromnd)) != 0) 3311 return (error); 3312 if (fromnd.ni_dvp != fromnd.ni_vp) 3313 VOP_UNLOCK(fromnd.ni_dvp, 0); 3314 fvp = fromnd.ni_vp; 3315 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3316 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3317 if ((error = namei(&tond)) != 0) { 3318 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3319 vrele(fromnd.ni_dvp); 3320 vrele(fvp); 3321 goto out1; 3322 } 3323 tdvp = tond.ni_dvp; 3324 tvp = tond.ni_vp; 3325 3326 if (tvp != NULL) { 3327 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3328 error = ENOTDIR; 3329 goto out; 3330 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3331 error = EISDIR; 3332 goto out; 3333 } 3334 } 3335 3336 if (fvp == tdvp) 3337 error = EINVAL; 3338 3339 /* 3340 * Source and destination refer to the same object. 3341 */ 3342 if (fvp == tvp) { 3343 if (retain) 3344 error = -1; 3345 else if (fromnd.ni_dvp == tdvp && 3346 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3347 !memcmp(fromnd.ni_cnd.cn_nameptr, 3348 tond.ni_cnd.cn_nameptr, 3349 fromnd.ni_cnd.cn_namelen)) 3350 error = -1; 3351 } 3352 3353 #if NVERIEXEC > 0 3354 if (!error) { 3355 char *f1, *f2; 3356 3357 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3358 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3359 3360 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3361 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3362 3363 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3364 3365 free(f1, M_TEMP); 3366 free(f2, M_TEMP); 3367 } 3368 #endif /* NVERIEXEC > 0 */ 3369 3370 out: 3371 p = l->l_proc; 3372 if (!error) { 3373 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3374 if (fromnd.ni_dvp != tdvp) 3375 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3376 if (tvp) { 3377 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3378 } 3379 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3380 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3381 } else { 3382 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3383 if (tdvp == tvp) 3384 vrele(tdvp); 3385 else 3386 vput(tdvp); 3387 if (tvp) 3388 vput(tvp); 3389 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3390 vrele(fromnd.ni_dvp); 3391 vrele(fvp); 3392 } 3393 vrele(tond.ni_startdir); 3394 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3395 out1: 3396 if (fromnd.ni_startdir) 3397 vrele(fromnd.ni_startdir); 3398 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3399 return (error == -1 ? 0 : error); 3400 } 3401 3402 /* 3403 * Make a directory file. 3404 */ 3405 /* ARGSUSED */ 3406 int 3407 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3408 { 3409 struct sys_mkdir_args /* { 3410 syscallarg(const char *) path; 3411 syscallarg(int) mode; 3412 } */ *uap = v; 3413 struct proc *p = l->l_proc; 3414 struct vnode *vp; 3415 struct vattr vattr; 3416 int error; 3417 struct nameidata nd; 3418 3419 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3420 SCARG(uap, path), l); 3421 if ((error = namei(&nd)) != 0) 3422 return (error); 3423 vp = nd.ni_vp; 3424 if (vp != NULL) { 3425 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3426 if (nd.ni_dvp == vp) 3427 vrele(nd.ni_dvp); 3428 else 3429 vput(nd.ni_dvp); 3430 vrele(vp); 3431 return (EEXIST); 3432 } 3433 VATTR_NULL(&vattr); 3434 vattr.va_type = VDIR; 3435 /* We will read cwdi->cwdi_cmask unlocked. */ 3436 vattr.va_mode = 3437 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3438 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3439 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3440 if (!error) 3441 vput(nd.ni_vp); 3442 return (error); 3443 } 3444 3445 /* 3446 * Remove a directory file. 3447 */ 3448 /* ARGSUSED */ 3449 int 3450 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3451 { 3452 struct sys_rmdir_args /* { 3453 syscallarg(const char *) path; 3454 } */ *uap = v; 3455 struct vnode *vp; 3456 int error; 3457 struct nameidata nd; 3458 3459 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3460 SCARG(uap, path), l); 3461 if ((error = namei(&nd)) != 0) 3462 return (error); 3463 vp = nd.ni_vp; 3464 if (vp->v_type != VDIR) { 3465 error = ENOTDIR; 3466 goto out; 3467 } 3468 /* 3469 * No rmdir "." please. 3470 */ 3471 if (nd.ni_dvp == vp) { 3472 error = EINVAL; 3473 goto out; 3474 } 3475 /* 3476 * The root of a mounted filesystem cannot be deleted. 3477 */ 3478 if (vp->v_flag & VROOT) { 3479 error = EBUSY; 3480 goto out; 3481 } 3482 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3483 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3484 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3485 return (error); 3486 3487 out: 3488 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3489 if (nd.ni_dvp == vp) 3490 vrele(nd.ni_dvp); 3491 else 3492 vput(nd.ni_dvp); 3493 vput(vp); 3494 return (error); 3495 } 3496 3497 /* 3498 * Read a block of directory entries in a file system independent format. 3499 */ 3500 int 3501 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3502 { 3503 struct sys___getdents30_args /* { 3504 syscallarg(int) fd; 3505 syscallarg(char *) buf; 3506 syscallarg(size_t) count; 3507 } */ *uap = v; 3508 struct proc *p = l->l_proc; 3509 struct file *fp; 3510 int error, done; 3511 3512 /* getvnode() will use the descriptor for us */ 3513 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3514 return (error); 3515 if ((fp->f_flag & FREAD) == 0) { 3516 error = EBADF; 3517 goto out; 3518 } 3519 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3520 SCARG(uap, count), &done, l, 0, 0); 3521 #ifdef KTRACE 3522 if (!error && KTRPOINT(p, KTR_GENIO)) { 3523 struct iovec iov; 3524 iov.iov_base = SCARG(uap, buf); 3525 iov.iov_len = done; 3526 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3527 } 3528 #endif 3529 *retval = done; 3530 out: 3531 FILE_UNUSE(fp, l); 3532 return (error); 3533 } 3534 3535 /* 3536 * Set the mode mask for creation of filesystem nodes. 3537 */ 3538 int 3539 sys_umask(struct lwp *l, void *v, register_t *retval) 3540 { 3541 struct sys_umask_args /* { 3542 syscallarg(mode_t) newmask; 3543 } */ *uap = v; 3544 struct proc *p = l->l_proc; 3545 struct cwdinfo *cwdi; 3546 3547 /* 3548 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3549 * important is that we serialize changes to the mask. The 3550 * rw_exit() will issue a write memory barrier on our behalf, 3551 * and force the changes out to other CPUs (as it must use an 3552 * atomic operation, draining the local CPU's store buffers). 3553 */ 3554 cwdi = p->p_cwdi; 3555 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3556 *retval = cwdi->cwdi_cmask; 3557 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3558 rw_exit(&cwdi->cwdi_lock); 3559 3560 return (0); 3561 } 3562 3563 /* 3564 * Void all references to file by ripping underlying filesystem 3565 * away from vnode. 3566 */ 3567 /* ARGSUSED */ 3568 int 3569 sys_revoke(struct lwp *l, void *v, register_t *retval) 3570 { 3571 struct sys_revoke_args /* { 3572 syscallarg(const char *) path; 3573 } */ *uap = v; 3574 struct vnode *vp; 3575 struct vattr vattr; 3576 int error; 3577 struct nameidata nd; 3578 3579 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3580 if ((error = namei(&nd)) != 0) 3581 return (error); 3582 vp = nd.ni_vp; 3583 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3584 goto out; 3585 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3586 (error = kauth_authorize_generic(l->l_cred, 3587 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3588 goto out; 3589 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3590 VOP_REVOKE(vp, REVOKEALL); 3591 out: 3592 vrele(vp); 3593 return (error); 3594 } 3595 3596 /* 3597 * Convert a user file descriptor to a kernel file entry. 3598 */ 3599 int 3600 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3601 { 3602 struct vnode *vp; 3603 struct file *fp; 3604 3605 if ((fp = fd_getfile(fdp, fd)) == NULL) 3606 return (EBADF); 3607 3608 FILE_USE(fp); 3609 3610 if (fp->f_type != DTYPE_VNODE) { 3611 FILE_UNUSE(fp, NULL); 3612 return (EINVAL); 3613 } 3614 3615 vp = (struct vnode *)fp->f_data; 3616 if (vp->v_type == VBAD) { 3617 FILE_UNUSE(fp, NULL); 3618 return (EBADF); 3619 } 3620 3621 *fpp = fp; 3622 return (0); 3623 } 3624
Indexes created Fri Oct 03 02:09:57 GMT 2025