vfs_syscalls.c revision 1.267.2.4
1 /* $NetBSD: vfs_syscalls.c,v 1.267.2.4 2007/01/30 13:51:42 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 1.267.2.4 2007/01/30 13:51:42 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 #ifdef KTRACE 66 #include <sys/ktrace.h> 67 #endif 68 #ifdef FILEASSOC 69 #include <sys/fileassoc.h> 70 #endif /* FILEASSOC */ 71 #if NVERIEXEC > 0 72 #include <sys/verified_exec.h> 73 #include <sys/syslog.h> 74 #endif /* NVERIEXEC > 0 */ 75 #include <sys/kauth.h> 76 77 #include <miscfs/genfs/genfs.h> 78 #include <miscfs/syncfs/syncfs.h> 79 80 #ifdef COMPAT_30 81 #include "opt_nfsserver.h" 82 #include <nfs/rpcv2.h> 83 #endif 84 #include <nfs/nfsproto.h> 85 #ifdef COMPAT_30 86 #include <nfs/nfs.h> 87 #include <nfs/nfs_var.h> 88 #endif 89 90 #if NFSS > 0 91 #include <dev/fssvar.h> 92 #endif 93 94 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 95 96 static int change_dir(struct nameidata *, struct lwp *); 97 static int change_flags(struct vnode *, u_long, struct lwp *); 98 static int change_mode(struct vnode *, int, struct lwp *l); 99 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 100 static int change_utimes(struct vnode *vp, const struct timeval *, 101 struct lwp *l); 102 static int rename_files(const char *, const char *, struct lwp *, int); 103 104 void checkdirs(struct vnode *); 105 106 static int mount_update(struct lwp *, struct vnode *, const char *, int, 107 void *, struct nameidata *); 108 static int mount_domount(struct lwp *, struct vnode *, const char *, 109 const char *, int, void *, struct nameidata *); 110 static int mount_getargs(struct lwp *, struct vnode *, const char *, int, 111 void *, struct nameidata *); 112 113 int dovfsusermount = 0; 114 115 /* 116 * Virtual File System System Calls 117 */ 118 119 /* 120 * Mount a file system. 121 */ 122 123 #if defined(COMPAT_09) || defined(COMPAT_43) 124 /* 125 * This table is used to maintain compatibility with 4.3BSD 126 * and NetBSD 0.9 mount syscalls. Note, the order is important! 127 * 128 * Do not modify this table. It should only contain filesystems 129 * supported by NetBSD 0.9 and 4.3BSD. 130 */ 131 const char * const mountcompatnames[] = { 132 NULL, /* 0 = MOUNT_NONE */ 133 MOUNT_FFS, /* 1 = MOUNT_UFS */ 134 MOUNT_NFS, /* 2 */ 135 MOUNT_MFS, /* 3 */ 136 MOUNT_MSDOS, /* 4 */ 137 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 138 MOUNT_FDESC, /* 6 */ 139 MOUNT_KERNFS, /* 7 */ 140 NULL, /* 8 = MOUNT_DEVFS */ 141 MOUNT_AFS, /* 9 */ 142 }; 143 const int nmountcompatnames = sizeof(mountcompatnames) / 144 sizeof(mountcompatnames[0]); 145 #endif /* COMPAT_09 || COMPAT_43 */ 146 147 static int 148 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 149 void *data, struct nameidata *ndp) 150 { 151 struct mount *mp; 152 int error = 0, saved_flags; 153 154 mp = vp->v_mount; 155 saved_flags = mp->mnt_flag; 156 157 /* We can't operate on VROOT here. */ 158 if ((vp->v_flag & VROOT) == 0) { 159 vput(vp); 160 error = EINVAL; 161 goto out; 162 } 163 164 /* 165 * We only allow the filesystem to be reloaded if it 166 * is currently mounted read-only. 167 */ 168 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 169 error = EOPNOTSUPP; /* Needs translation */ 170 goto out; 171 } 172 173 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 174 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 175 if (error) 176 goto out; 177 178 if (vfs_busy(mp, LK_NOWAIT, 0)) { 179 error = EPERM; 180 goto out; 181 } 182 183 mp->mnt_flag &= ~MNT_OP_FLAGS; 184 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 185 186 /* 187 * Set the mount level flags. 188 */ 189 if (flags & MNT_RDONLY) 190 mp->mnt_flag |= MNT_RDONLY; 191 else if (mp->mnt_flag & MNT_RDONLY) 192 mp->mnt_iflag |= IMNT_WANTRDWR; 193 mp->mnt_flag &= 194 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 195 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 196 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 197 mp->mnt_flag |= flags & 198 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 199 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 200 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 201 MNT_IGNORE); 202 203 error = VFS_MOUNT(mp, path, data, ndp, l); 204 205 #if defined(COMPAT_30) && defined(NFSSERVER) 206 if (error) { 207 int error2; 208 209 /* Update failed; let's try and see if it was an 210 * export request. */ 211 error2 = nfs_update_exports_30(mp, path, data, l); 212 213 /* Only update error code if the export request was 214 * understood but some problem occurred while 215 * processing it. */ 216 if (error2 != EJUSTRETURN) 217 error = error2; 218 } 219 #endif 220 if (mp->mnt_iflag & IMNT_WANTRDWR) 221 mp->mnt_flag &= ~MNT_RDONLY; 222 if (error) 223 mp->mnt_flag = saved_flags; 224 mp->mnt_flag &= ~MNT_OP_FLAGS; 225 mp->mnt_iflag &= ~IMNT_WANTRDWR; 226 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 227 if (mp->mnt_syncer == NULL) 228 error = vfs_allocate_syncvnode(mp); 229 } else { 230 if (mp->mnt_syncer != NULL) 231 vfs_deallocate_syncvnode(mp); 232 } 233 vfs_unbusy(mp); 234 235 out: 236 return (error); 237 } 238 239 static int 240 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 241 const char *path, int flags, void *data, struct nameidata *ndp) 242 { 243 struct mount *mp = NULL; 244 struct vattr va; 245 char fstypename[MFSNAMELEN]; 246 int error; 247 248 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 249 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 250 if (error) { 251 vput(vp); 252 goto out; 253 } 254 255 /* Can't make a non-dir a mount-point (from here anyway). */ 256 if (vp->v_type != VDIR) { 257 error = ENOTDIR; 258 vput(vp); 259 goto out; 260 } 261 262 /* 263 * If the user is not root, ensure that they own the directory 264 * onto which we are attempting to mount. 265 */ 266 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 267 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 268 (error = kauth_authorize_generic(l->l_cred, 269 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 270 vput(vp); 271 goto out; 272 } 273 274 if (flags & MNT_EXPORTED) { 275 error = EINVAL; 276 vput(vp); 277 goto out; 278 } 279 280 /* 281 * Copy file-system type from userspace. 282 */ 283 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 284 if (error) { 285 #if defined(COMPAT_09) || defined(COMPAT_43) 286 /* 287 * Historically, filesystem types were identified by numbers. 288 * If we get an integer for the filesystem type instead of a 289 * string, we check to see if it matches one of the historic 290 * filesystem types. 291 */ 292 u_long fsindex = (u_long)fstype; 293 if (fsindex >= nmountcompatnames || 294 mountcompatnames[fsindex] == NULL) { 295 error = ENODEV; 296 vput(vp); 297 goto out; 298 } 299 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 300 #else 301 vput(vp); 302 goto out; 303 #endif 304 } 305 306 #ifdef COMPAT_10 307 /* Accept `ufs' as an alias for `ffs'. */ 308 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 309 strlcpy(fstypename, "ffs", sizeof(fstypename)); 310 #endif 311 312 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 313 vput(vp); 314 goto out; 315 } 316 317 /* 318 * Check if a file-system is not already mounted on this vnode. 319 */ 320 if (vp->v_mountedhere != NULL) { 321 error = EBUSY; 322 vput(vp); 323 goto out; 324 } 325 326 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 327 328 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 329 free(mp, M_MOUNT); 330 error = ENODEV; 331 vput(vp); 332 goto out; 333 } 334 335 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 336 simple_lock_init(&mp->mnt_slock); 337 (void)vfs_busy(mp, LK_NOWAIT, 0); 338 339 mp->mnt_op->vfs_refcount++; 340 mp->mnt_vnodecovered = vp; 341 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 342 mp->mnt_unmounter = NULL; 343 mp->mnt_leaf = mp; 344 mount_initspecific(mp); 345 346 /* 347 * The underlying file system may refuse the mount for 348 * various reasons. Allow the user to force it to happen. 349 * 350 * Set the mount level flags. 351 */ 352 mp->mnt_flag = flags & 353 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 354 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 355 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 356 MNT_IGNORE | MNT_RDONLY); 357 358 error = VFS_MOUNT(mp, path, data, ndp, l); 359 mp->mnt_flag &= ~MNT_OP_FLAGS; 360 361 /* 362 * Put the new filesystem on the mount list after root. 363 */ 364 cache_purge(vp); 365 if (!error) { 366 mp->mnt_iflag &= ~IMNT_WANTRDWR; 367 vp->v_mountedhere = mp; 368 simple_lock(&mountlist_slock); 369 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 370 simple_unlock(&mountlist_slock); 371 checkdirs(vp); 372 VOP_UNLOCK(vp, 0); 373 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 374 error = vfs_allocate_syncvnode(mp); 375 vfs_unbusy(mp); 376 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 377 error = VFS_START(mp, 0, l); 378 if (error) 379 vrele(vp); 380 } else { 381 vp->v_mountedhere = NULL; 382 mp->mnt_op->vfs_refcount--; 383 vfs_unbusy(mp); 384 free(mp, M_MOUNT); 385 vput(vp); 386 } 387 388 out: 389 return (error); 390 } 391 392 static int 393 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 394 void *data, struct nameidata *ndp) 395 { 396 struct mount *mp; 397 int error; 398 399 /* If MNT_GETARGS is specified, it should be the only flag. */ 400 if (flags & ~MNT_GETARGS) { 401 error = EINVAL; 402 goto out; 403 } 404 405 mp = vp->v_mount; 406 407 /* XXX: probably some notion of "can see" here if we want isolation. */ 408 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 409 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 410 if (error) 411 goto out; 412 413 if ((vp->v_flag & VROOT) == 0) { 414 error = EINVAL; 415 goto out; 416 } 417 418 if (vfs_busy(mp, LK_NOWAIT, 0)) { 419 error = EPERM; 420 goto out; 421 } 422 423 mp->mnt_flag &= ~MNT_OP_FLAGS; 424 mp->mnt_flag |= MNT_GETARGS; 425 error = VFS_MOUNT(mp, path, data, ndp, l); 426 mp->mnt_flag &= ~MNT_OP_FLAGS; 427 428 vfs_unbusy(mp); 429 out: 430 return (error); 431 } 432 433 /* ARGSUSED */ 434 int 435 sys_mount(struct lwp *l, void *v, register_t *retval) 436 { 437 struct sys_mount_args /* { 438 syscallarg(const char *) type; 439 syscallarg(const char *) path; 440 syscallarg(int) flags; 441 syscallarg(void *) data; 442 } */ *uap = v; 443 struct vnode *vp; 444 struct nameidata nd; 445 int error; 446 447 /* 448 * Get vnode to be covered 449 */ 450 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 451 SCARG(uap, path), l); 452 if ((error = namei(&nd)) != 0) 453 return (error); 454 vp = nd.ni_vp; 455 456 /* 457 * A lookup in VFS_MOUNT might result in an attempt to 458 * lock this vnode again, so make the lock recursive. 459 */ 460 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 461 462 if (SCARG(uap, flags) & MNT_GETARGS) { 463 error = mount_getargs(l, vp, SCARG(uap, path), 464 SCARG(uap, flags), SCARG(uap, data), &nd); 465 vput(vp); 466 } else if (SCARG(uap, flags) & MNT_UPDATE) { 467 error = mount_update(l, vp, SCARG(uap, path), 468 SCARG(uap, flags), SCARG(uap, data), &nd); 469 vput(vp); 470 } else { 471 /* Locking is handled internally in mount_domount(). */ 472 error = mount_domount(l, vp, SCARG(uap, type), 473 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 474 } 475 476 return (error); 477 } 478 479 /* 480 * Scan all active processes to see if any of them have a current 481 * or root directory onto which the new filesystem has just been 482 * mounted. If so, replace them with the new mount point. 483 */ 484 void 485 checkdirs(struct vnode *olddp) 486 { 487 struct cwdinfo *cwdi; 488 struct vnode *newdp; 489 struct proc *p; 490 491 if (olddp->v_usecount == 1) 492 return; 493 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 494 panic("mount: lost mount"); 495 rw_enter(&proclist_lock, RW_READER); 496 PROCLIST_FOREACH(p, &allproc) { 497 cwdi = p->p_cwdi; 498 if (!cwdi) 499 continue; 500 if (cwdi->cwdi_cdir == olddp) { 501 vrele(cwdi->cwdi_cdir); 502 VREF(newdp); 503 cwdi->cwdi_cdir = newdp; 504 } 505 if (cwdi->cwdi_rdir == olddp) { 506 vrele(cwdi->cwdi_rdir); 507 VREF(newdp); 508 cwdi->cwdi_rdir = newdp; 509 } 510 } 511 rw_exit(&proclist_lock); 512 if (rootvnode == olddp) { 513 vrele(rootvnode); 514 VREF(newdp); 515 rootvnode = newdp; 516 } 517 vput(newdp); 518 } 519 520 /* 521 * Unmount a file system. 522 * 523 * Note: unmount takes a path to the vnode mounted on as argument, 524 * not special file (as before). 525 */ 526 /* ARGSUSED */ 527 int 528 sys_unmount(struct lwp *l, void *v, register_t *retval) 529 { 530 struct sys_unmount_args /* { 531 syscallarg(const char *) path; 532 syscallarg(int) flags; 533 } */ *uap = v; 534 struct vnode *vp; 535 struct mount *mp; 536 int error; 537 struct nameidata nd; 538 539 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 540 SCARG(uap, path), l); 541 if ((error = namei(&nd)) != 0) 542 return (error); 543 vp = nd.ni_vp; 544 mp = vp->v_mount; 545 546 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 547 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 548 if (error) { 549 vput(vp); 550 return (error); 551 } 552 553 /* 554 * Don't allow unmounting the root file system. 555 */ 556 if (mp->mnt_flag & MNT_ROOTFS) { 557 vput(vp); 558 return (EINVAL); 559 } 560 561 /* 562 * Must be the root of the filesystem 563 */ 564 if ((vp->v_flag & VROOT) == 0) { 565 vput(vp); 566 return (EINVAL); 567 } 568 vput(vp); 569 570 /* 571 * XXX Freeze syncer. Must do this before locking the 572 * mount point. See dounmount() for details. 573 */ 574 mutex_enter(&syncer_mutex); 575 576 if (vfs_busy(mp, 0, 0)) { 577 mutex_exit(&syncer_mutex); 578 return (EBUSY); 579 } 580 581 return (dounmount(mp, SCARG(uap, flags), l)); 582 } 583 584 /* 585 * Do the actual file system unmount. File system is assumed to have been 586 * marked busy by the caller. 587 */ 588 int 589 dounmount(struct mount *mp, int flags, struct lwp *l) 590 { 591 struct vnode *coveredvp; 592 int error; 593 int async; 594 int used_syncer; 595 596 #if NVERIEXEC > 0 597 error = veriexec_unmountchk(mp); 598 if (error) 599 return (error); 600 #endif /* NVERIEXEC > 0 */ 601 602 simple_lock(&mountlist_slock); 603 vfs_unbusy(mp); 604 used_syncer = (mp->mnt_syncer != NULL); 605 606 /* 607 * XXX Syncer must be frozen when we get here. This should really 608 * be done on a per-mountpoint basis, but especially the softdep 609 * code possibly called from the syncer doesn't exactly work on a 610 * per-mountpoint basis, so the softdep code would become a maze 611 * of vfs_busy() calls. 612 * 613 * The caller of dounmount() must acquire syncer_mutex because 614 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 615 * order, and we must preserve that order to avoid deadlock. 616 * 617 * So, if the file system did not use the syncer, now is 618 * the time to release the syncer_mutex. 619 */ 620 if (used_syncer == 0) 621 mutex_exit(&syncer_mutex); 622 623 mp->mnt_iflag |= IMNT_UNMOUNT; 624 mp->mnt_unmounter = l; 625 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 626 vn_start_write(NULL, &mp, V_WAIT); 627 628 async = mp->mnt_flag & MNT_ASYNC; 629 mp->mnt_flag &= ~MNT_ASYNC; 630 cache_purgevfs(mp); /* remove cache entries for this file sys */ 631 if (mp->mnt_syncer != NULL) 632 vfs_deallocate_syncvnode(mp); 633 error = 0; 634 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 635 #if NFSS > 0 636 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 637 #endif 638 if (error == 0) 639 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 640 } 641 if (error == 0 || (flags & MNT_FORCE)) 642 error = VFS_UNMOUNT(mp, flags, l); 643 vn_finished_write(mp, 0); 644 simple_lock(&mountlist_slock); 645 if (error) { 646 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 647 (void) vfs_allocate_syncvnode(mp); 648 mp->mnt_iflag &= ~IMNT_UNMOUNT; 649 mp->mnt_unmounter = NULL; 650 mp->mnt_flag |= async; 651 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 652 &mountlist_slock); 653 if (used_syncer) 654 mutex_exit(&syncer_mutex); 655 simple_lock(&mp->mnt_slock); 656 while (mp->mnt_wcnt > 0) { 657 wakeup(mp); 658 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 659 0, &mp->mnt_slock); 660 } 661 simple_unlock(&mp->mnt_slock); 662 return (error); 663 } 664 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 665 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 666 coveredvp->v_mountedhere = NULL; 667 vrele(coveredvp); 668 } 669 mp->mnt_op->vfs_refcount--; 670 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 671 panic("unmount: dangling vnode"); 672 mp->mnt_iflag |= IMNT_GONE; 673 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 674 mount_finispecific(mp); 675 if (used_syncer) 676 mutex_exit(&syncer_mutex); 677 simple_lock(&mp->mnt_slock); 678 while (mp->mnt_wcnt > 0) { 679 wakeup(mp); 680 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 681 } 682 simple_unlock(&mp->mnt_slock); 683 vfs_hooks_unmount(mp); 684 free(mp, M_MOUNT); 685 return (0); 686 } 687 688 /* 689 * Sync each mounted filesystem. 690 */ 691 #ifdef DEBUG 692 int syncprt = 0; 693 struct ctldebug debug0 = { "syncprt", &syncprt }; 694 #endif 695 696 /* ARGSUSED */ 697 int 698 sys_sync(struct lwp *l, void *v, register_t *retval) 699 { 700 struct mount *mp, *nmp; 701 int asyncflag; 702 703 if (l == NULL) 704 l = &lwp0; 705 706 simple_lock(&mountlist_slock); 707 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 708 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 709 nmp = mp->mnt_list.cqe_prev; 710 continue; 711 } 712 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 713 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 714 asyncflag = mp->mnt_flag & MNT_ASYNC; 715 mp->mnt_flag &= ~MNT_ASYNC; 716 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 717 if (asyncflag) 718 mp->mnt_flag |= MNT_ASYNC; 719 vn_finished_write(mp, 0); 720 } 721 simple_lock(&mountlist_slock); 722 nmp = mp->mnt_list.cqe_prev; 723 vfs_unbusy(mp); 724 725 } 726 simple_unlock(&mountlist_slock); 727 #ifdef DEBUG 728 if (syncprt) 729 vfs_bufstats(); 730 #endif /* DEBUG */ 731 return (0); 732 } 733 734 /* 735 * Change filesystem quotas. 736 */ 737 /* ARGSUSED */ 738 int 739 sys_quotactl(struct lwp *l, void *v, register_t *retval) 740 { 741 struct sys_quotactl_args /* { 742 syscallarg(const char *) path; 743 syscallarg(int) cmd; 744 syscallarg(int) uid; 745 syscallarg(void *) arg; 746 } */ *uap = v; 747 struct mount *mp; 748 int error; 749 struct nameidata nd; 750 751 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 752 if ((error = namei(&nd)) != 0) 753 return (error); 754 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 755 vrele(nd.ni_vp); 756 if (error) 757 return (error); 758 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 759 SCARG(uap, arg), l); 760 vn_finished_write(mp, 0); 761 return (error); 762 } 763 764 int 765 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 766 int root) 767 { 768 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 769 int error = 0; 770 771 /* 772 * If MNT_NOWAIT or MNT_LAZY is specified, do not 773 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 774 * overrides MNT_NOWAIT. 775 */ 776 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 777 (flags != MNT_WAIT && flags != 0)) { 778 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 779 goto done; 780 } 781 782 /* Get the filesystem stats now */ 783 memset(sp, 0, sizeof(*sp)); 784 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 785 return error; 786 } 787 788 if (cwdi->cwdi_rdir == NULL) 789 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 790 done: 791 if (cwdi->cwdi_rdir != NULL) { 792 size_t len; 793 char *bp; 794 char *path = PNBUF_GET(); 795 if (!path) 796 return ENOMEM; 797 798 bp = path + MAXPATHLEN; 799 *--bp = '\0'; 800 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 801 MAXPATHLEN / 2, 0, l); 802 if (error) { 803 PNBUF_PUT(path); 804 return error; 805 } 806 len = strlen(bp); 807 /* 808 * for mount points that are below our root, we can see 809 * them, so we fix up the pathname and return them. The 810 * rest we cannot see, so we don't allow viewing the 811 * data. 812 */ 813 if (strncmp(bp, sp->f_mntonname, len) == 0) { 814 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 815 sizeof(sp->f_mntonname)); 816 if (sp->f_mntonname[0] == '\0') 817 (void)strlcpy(sp->f_mntonname, "/", 818 sizeof(sp->f_mntonname)); 819 } else { 820 if (root) 821 (void)strlcpy(sp->f_mntonname, "/", 822 sizeof(sp->f_mntonname)); 823 else 824 error = EPERM; 825 } 826 PNBUF_PUT(path); 827 } 828 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 829 return error; 830 } 831 832 /* 833 * Get filesystem statistics. 834 */ 835 /* ARGSUSED */ 836 int 837 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 838 { 839 struct sys_statvfs1_args /* { 840 syscallarg(const char *) path; 841 syscallarg(struct statvfs *) buf; 842 syscallarg(int) flags; 843 } */ *uap = v; 844 struct mount *mp; 845 struct statvfs *sb; 846 int error; 847 struct nameidata nd; 848 849 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 850 if ((error = namei(&nd)) != 0) 851 return error; 852 mp = nd.ni_vp->v_mount; 853 vrele(nd.ni_vp); 854 sb = STATVFSBUF_GET(); 855 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 856 if (error == 0) { 857 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 858 } 859 STATVFSBUF_PUT(sb); 860 return error; 861 } 862 863 /* 864 * Get filesystem statistics. 865 */ 866 /* ARGSUSED */ 867 int 868 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 869 { 870 struct sys_fstatvfs1_args /* { 871 syscallarg(int) fd; 872 syscallarg(struct statvfs *) buf; 873 syscallarg(int) flags; 874 } */ *uap = v; 875 struct proc *p = l->l_proc; 876 struct file *fp; 877 struct mount *mp; 878 struct statvfs *sb; 879 int error; 880 881 /* getvnode() will use the descriptor for us */ 882 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 883 return (error); 884 mp = ((struct vnode *)fp->f_data)->v_mount; 885 sb = STATVFSBUF_GET(); 886 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 887 goto out; 888 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 889 out: 890 FILE_UNUSE(fp, l); 891 STATVFSBUF_PUT(sb); 892 return error; 893 } 894 895 896 /* 897 * Get statistics on all filesystems. 898 */ 899 int 900 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 901 { 902 struct sys_getvfsstat_args /* { 903 syscallarg(struct statvfs *) buf; 904 syscallarg(size_t) bufsize; 905 syscallarg(int) flags; 906 } */ *uap = v; 907 int root = 0; 908 struct proc *p = l->l_proc; 909 struct mount *mp, *nmp; 910 struct statvfs *sb; 911 struct statvfs *sfsp; 912 size_t count, maxcount; 913 int error = 0; 914 915 sb = STATVFSBUF_GET(); 916 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 917 sfsp = SCARG(uap, buf); 918 simple_lock(&mountlist_slock); 919 count = 0; 920 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 921 mp = nmp) { 922 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 923 nmp = CIRCLEQ_NEXT(mp, mnt_list); 924 continue; 925 } 926 if (sfsp && count < maxcount) { 927 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 928 if (error) { 929 simple_lock(&mountlist_slock); 930 nmp = CIRCLEQ_NEXT(mp, mnt_list); 931 vfs_unbusy(mp); 932 continue; 933 } 934 error = copyout(sb, sfsp, sizeof(*sfsp)); 935 if (error) { 936 vfs_unbusy(mp); 937 goto out; 938 } 939 sfsp++; 940 root |= strcmp(sb->f_mntonname, "/") == 0; 941 } 942 count++; 943 simple_lock(&mountlist_slock); 944 nmp = CIRCLEQ_NEXT(mp, mnt_list); 945 vfs_unbusy(mp); 946 } 947 simple_unlock(&mountlist_slock); 948 if (root == 0 && p->p_cwdi->cwdi_rdir) { 949 /* 950 * fake a root entry 951 */ 952 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 953 SCARG(uap, flags), 1)) != 0) 954 goto out; 955 if (sfsp) 956 error = copyout(sb, sfsp, sizeof(*sfsp)); 957 count++; 958 } 959 if (sfsp && count > maxcount) 960 *retval = maxcount; 961 else 962 *retval = count; 963 out: 964 STATVFSBUF_PUT(sb); 965 return error; 966 } 967 968 /* 969 * Change current working directory to a given file descriptor. 970 */ 971 /* ARGSUSED */ 972 int 973 sys_fchdir(struct lwp *l, void *v, register_t *retval) 974 { 975 struct sys_fchdir_args /* { 976 syscallarg(int) fd; 977 } */ *uap = v; 978 struct proc *p = l->l_proc; 979 struct filedesc *fdp = p->p_fd; 980 struct cwdinfo *cwdi = p->p_cwdi; 981 struct vnode *vp, *tdp; 982 struct mount *mp; 983 struct file *fp; 984 int error; 985 986 /* getvnode() will use the descriptor for us */ 987 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 988 return (error); 989 vp = (struct vnode *)fp->f_data; 990 991 VREF(vp); 992 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 993 if (vp->v_type != VDIR) 994 error = ENOTDIR; 995 else 996 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 997 while (!error && (mp = vp->v_mountedhere) != NULL) { 998 if (vfs_busy(mp, 0, 0)) 999 continue; 1000 error = VFS_ROOT(mp, &tdp); 1001 vfs_unbusy(mp); 1002 if (error) 1003 break; 1004 vput(vp); 1005 vp = tdp; 1006 } 1007 if (error) { 1008 vput(vp); 1009 goto out; 1010 } 1011 VOP_UNLOCK(vp, 0); 1012 1013 /* 1014 * Disallow changing to a directory not under the process's 1015 * current root directory (if there is one). 1016 */ 1017 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1018 vrele(vp); 1019 error = EPERM; /* operation not permitted */ 1020 goto out; 1021 } 1022 1023 vrele(cwdi->cwdi_cdir); 1024 cwdi->cwdi_cdir = vp; 1025 out: 1026 FILE_UNUSE(fp, l); 1027 return (error); 1028 } 1029 1030 /* 1031 * Change this process's notion of the root directory to a given file 1032 * descriptor. 1033 */ 1034 int 1035 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1036 { 1037 struct sys_fchroot_args *uap = v; 1038 struct proc *p = l->l_proc; 1039 struct filedesc *fdp = p->p_fd; 1040 struct cwdinfo *cwdi = p->p_cwdi; 1041 struct vnode *vp; 1042 struct file *fp; 1043 int error; 1044 1045 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1046 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1047 return error; 1048 /* getvnode() will use the descriptor for us */ 1049 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1050 return error; 1051 vp = (struct vnode *) fp->f_data; 1052 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1053 if (vp->v_type != VDIR) 1054 error = ENOTDIR; 1055 else 1056 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1057 VOP_UNLOCK(vp, 0); 1058 if (error) 1059 goto out; 1060 VREF(vp); 1061 1062 /* 1063 * Prevent escaping from chroot by putting the root under 1064 * the working directory. Silently chdir to / if we aren't 1065 * already there. 1066 */ 1067 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1068 /* 1069 * XXX would be more failsafe to change directory to a 1070 * deadfs node here instead 1071 */ 1072 vrele(cwdi->cwdi_cdir); 1073 VREF(vp); 1074 cwdi->cwdi_cdir = vp; 1075 } 1076 1077 if (cwdi->cwdi_rdir != NULL) 1078 vrele(cwdi->cwdi_rdir); 1079 cwdi->cwdi_rdir = vp; 1080 out: 1081 FILE_UNUSE(fp, l); 1082 return (error); 1083 } 1084 1085 /* 1086 * Change current working directory (``.''). 1087 */ 1088 /* ARGSUSED */ 1089 int 1090 sys_chdir(struct lwp *l, void *v, register_t *retval) 1091 { 1092 struct sys_chdir_args /* { 1093 syscallarg(const char *) path; 1094 } */ *uap = v; 1095 struct proc *p = l->l_proc; 1096 struct cwdinfo *cwdi = p->p_cwdi; 1097 int error; 1098 struct nameidata nd; 1099 1100 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1101 SCARG(uap, path), l); 1102 if ((error = change_dir(&nd, l)) != 0) 1103 return (error); 1104 vrele(cwdi->cwdi_cdir); 1105 cwdi->cwdi_cdir = nd.ni_vp; 1106 return (0); 1107 } 1108 1109 /* 1110 * Change notion of root (``/'') directory. 1111 */ 1112 /* ARGSUSED */ 1113 int 1114 sys_chroot(struct lwp *l, void *v, register_t *retval) 1115 { 1116 struct sys_chroot_args /* { 1117 syscallarg(const char *) path; 1118 } */ *uap = v; 1119 struct proc *p = l->l_proc; 1120 struct cwdinfo *cwdi = p->p_cwdi; 1121 struct vnode *vp; 1122 int error; 1123 struct nameidata nd; 1124 1125 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1126 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1127 return (error); 1128 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1129 SCARG(uap, path), l); 1130 if ((error = change_dir(&nd, l)) != 0) 1131 return (error); 1132 if (cwdi->cwdi_rdir != NULL) 1133 vrele(cwdi->cwdi_rdir); 1134 vp = nd.ni_vp; 1135 cwdi->cwdi_rdir = vp; 1136 1137 /* 1138 * Prevent escaping from chroot by putting the root under 1139 * the working directory. Silently chdir to / if we aren't 1140 * already there. 1141 */ 1142 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1143 /* 1144 * XXX would be more failsafe to change directory to a 1145 * deadfs node here instead 1146 */ 1147 vrele(cwdi->cwdi_cdir); 1148 VREF(vp); 1149 cwdi->cwdi_cdir = vp; 1150 } 1151 1152 return (0); 1153 } 1154 1155 /* 1156 * Common routine for chroot and chdir. 1157 */ 1158 static int 1159 change_dir(struct nameidata *ndp, struct lwp *l) 1160 { 1161 struct vnode *vp; 1162 int error; 1163 1164 if ((error = namei(ndp)) != 0) 1165 return (error); 1166 vp = ndp->ni_vp; 1167 if (vp->v_type != VDIR) 1168 error = ENOTDIR; 1169 else 1170 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1171 1172 if (error) 1173 vput(vp); 1174 else 1175 VOP_UNLOCK(vp, 0); 1176 return (error); 1177 } 1178 1179 /* 1180 * Check permissions, allocate an open file structure, 1181 * and call the device open routine if any. 1182 */ 1183 int 1184 sys_open(struct lwp *l, void *v, register_t *retval) 1185 { 1186 struct sys_open_args /* { 1187 syscallarg(const char *) path; 1188 syscallarg(int) flags; 1189 syscallarg(int) mode; 1190 } */ *uap = v; 1191 struct proc *p = l->l_proc; 1192 struct cwdinfo *cwdi = p->p_cwdi; 1193 struct filedesc *fdp = p->p_fd; 1194 struct file *fp; 1195 struct vnode *vp; 1196 int flags, cmode; 1197 int type, indx, error; 1198 struct flock lf; 1199 struct nameidata nd; 1200 1201 flags = FFLAGS(SCARG(uap, flags)); 1202 if ((flags & (FREAD | FWRITE)) == 0) 1203 return (EINVAL); 1204 /* falloc() will use the file descriptor for us */ 1205 if ((error = falloc(l, &fp, &indx)) != 0) 1206 return (error); 1207 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1208 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1209 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1210 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1211 FILE_UNUSE(fp, l); 1212 fdp->fd_ofiles[indx] = NULL; 1213 ffree(fp); 1214 if ((error == EDUPFD || error == EMOVEFD) && 1215 l->l_dupfd >= 0 && /* XXX from fdopen */ 1216 (error = 1217 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1218 *retval = indx; 1219 return (0); 1220 } 1221 if (error == ERESTART) 1222 error = EINTR; 1223 fdremove(fdp, indx); 1224 return (error); 1225 } 1226 l->l_dupfd = 0; 1227 vp = nd.ni_vp; 1228 fp->f_flag = flags & FMASK; 1229 fp->f_type = DTYPE_VNODE; 1230 fp->f_ops = &vnops; 1231 fp->f_data = vp; 1232 if (flags & (O_EXLOCK | O_SHLOCK)) { 1233 lf.l_whence = SEEK_SET; 1234 lf.l_start = 0; 1235 lf.l_len = 0; 1236 if (flags & O_EXLOCK) 1237 lf.l_type = F_WRLCK; 1238 else 1239 lf.l_type = F_RDLCK; 1240 type = F_FLOCK; 1241 if ((flags & FNONBLOCK) == 0) 1242 type |= F_WAIT; 1243 VOP_UNLOCK(vp, 0); 1244 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1245 if (error) { 1246 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1247 FILE_UNUSE(fp, l); 1248 ffree(fp); 1249 fdremove(fdp, indx); 1250 return (error); 1251 } 1252 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1253 fp->f_flag |= FHASLOCK; 1254 } 1255 VOP_UNLOCK(vp, 0); 1256 *retval = indx; 1257 FILE_SET_MATURE(fp); 1258 FILE_UNUSE(fp, l); 1259 return (0); 1260 } 1261 1262 static void 1263 vfs__fhfree(fhandle_t *fhp) 1264 { 1265 size_t fhsize; 1266 1267 if (fhp == NULL) { 1268 return; 1269 } 1270 fhsize = FHANDLE_SIZE(fhp); 1271 kmem_free(fhp, fhsize); 1272 } 1273 1274 /* 1275 * vfs_composefh: compose a filehandle. 1276 */ 1277 1278 int 1279 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1280 { 1281 struct mount *mp; 1282 struct fid *fidp; 1283 int error; 1284 size_t needfhsize; 1285 size_t fidsize; 1286 1287 mp = vp->v_mount; 1288 fidp = NULL; 1289 if (*fh_size < FHANDLE_SIZE_MIN) { 1290 fidsize = 0; 1291 } else { 1292 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1293 if (fhp != NULL) { 1294 memset(fhp, 0, *fh_size); 1295 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1296 fidp = &fhp->fh_fid; 1297 } 1298 } 1299 error = VFS_VPTOFH(vp, fidp, &fidsize); 1300 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1301 if (error == 0 && *fh_size < needfhsize) { 1302 error = E2BIG; 1303 } 1304 *fh_size = needfhsize; 1305 return error; 1306 } 1307 1308 int 1309 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1310 { 1311 struct mount *mp; 1312 fhandle_t *fhp; 1313 size_t fhsize; 1314 size_t fidsize; 1315 int error; 1316 1317 *fhpp = NULL; 1318 mp = vp->v_mount; 1319 fidsize = 0; 1320 error = VFS_VPTOFH(vp, NULL, &fidsize); 1321 KASSERT(error != 0); 1322 if (error != E2BIG) { 1323 goto out; 1324 } 1325 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1326 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1327 if (fhp == NULL) { 1328 error = ENOMEM; 1329 goto out; 1330 } 1331 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1332 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1333 if (error == 0) { 1334 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1335 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1336 *fhpp = fhp; 1337 } else { 1338 kmem_free(fhp, fhsize); 1339 } 1340 out: 1341 return error; 1342 } 1343 1344 void 1345 vfs_composefh_free(fhandle_t *fhp) 1346 { 1347 1348 vfs__fhfree(fhp); 1349 } 1350 1351 /* 1352 * vfs_fhtovp: lookup a vnode by a filehandle. 1353 */ 1354 1355 int 1356 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1357 { 1358 struct mount *mp; 1359 int error; 1360 1361 *vpp = NULL; 1362 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1363 if (mp == NULL) { 1364 error = ESTALE; 1365 goto out; 1366 } 1367 if (mp->mnt_op->vfs_fhtovp == NULL) { 1368 error = EOPNOTSUPP; 1369 goto out; 1370 } 1371 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1372 out: 1373 return error; 1374 } 1375 1376 /* 1377 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1378 * the needed size. 1379 */ 1380 1381 int 1382 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1383 { 1384 fhandle_t *fhp; 1385 int error; 1386 1387 *fhpp = NULL; 1388 if (fhsize > FHANDLE_SIZE_MAX) { 1389 return EINVAL; 1390 } 1391 if (fhsize < FHANDLE_SIZE_MIN) { 1392 return EINVAL; 1393 } 1394 again: 1395 fhp = kmem_alloc(fhsize, KM_SLEEP); 1396 if (fhp == NULL) { 1397 return ENOMEM; 1398 } 1399 error = copyin(ufhp, fhp, fhsize); 1400 if (error == 0) { 1401 /* XXX this check shouldn't be here */ 1402 if (FHANDLE_SIZE(fhp) == fhsize) { 1403 *fhpp = fhp; 1404 return 0; 1405 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1406 /* 1407 * a kludge for nfsv2 padded handles. 1408 */ 1409 size_t sz; 1410 1411 sz = FHANDLE_SIZE(fhp); 1412 kmem_free(fhp, fhsize); 1413 fhsize = sz; 1414 goto again; 1415 } else { 1416 /* 1417 * userland told us wrong size. 1418 */ 1419 error = EINVAL; 1420 } 1421 } 1422 kmem_free(fhp, fhsize); 1423 return error; 1424 } 1425 1426 void 1427 vfs_copyinfh_free(fhandle_t *fhp) 1428 { 1429 1430 vfs__fhfree(fhp); 1431 } 1432 1433 /* 1434 * Get file handle system call 1435 */ 1436 int 1437 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1438 { 1439 struct sys___getfh30_args /* { 1440 syscallarg(char *) fname; 1441 syscallarg(fhandle_t *) fhp; 1442 syscallarg(size_t *) fh_size; 1443 } */ *uap = v; 1444 struct vnode *vp; 1445 fhandle_t *fh; 1446 int error; 1447 struct nameidata nd; 1448 size_t sz; 1449 size_t usz; 1450 1451 /* 1452 * Must be super user 1453 */ 1454 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1455 0, NULL, NULL, NULL); 1456 if (error) 1457 return (error); 1458 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1459 SCARG(uap, fname), l); 1460 error = namei(&nd); 1461 if (error) 1462 return (error); 1463 vp = nd.ni_vp; 1464 error = vfs_composefh_alloc(vp, &fh); 1465 vput(vp); 1466 if (error != 0) { 1467 goto out; 1468 } 1469 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1470 if (error != 0) { 1471 goto out; 1472 } 1473 sz = FHANDLE_SIZE(fh); 1474 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1475 if (error != 0) { 1476 goto out; 1477 } 1478 if (usz >= sz) { 1479 error = copyout(fh, SCARG(uap, fhp), sz); 1480 } else { 1481 error = E2BIG; 1482 } 1483 out: 1484 vfs_composefh_free(fh); 1485 return (error); 1486 } 1487 1488 /* 1489 * Open a file given a file handle. 1490 * 1491 * Check permissions, allocate an open file structure, 1492 * and call the device open routine if any. 1493 */ 1494 1495 int 1496 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1497 register_t *retval) 1498 { 1499 struct filedesc *fdp = l->l_proc->p_fd; 1500 struct file *fp; 1501 struct vnode *vp = NULL; 1502 struct mount *mp; 1503 kauth_cred_t cred = l->l_cred; 1504 struct file *nfp; 1505 int type, indx, error=0; 1506 struct flock lf; 1507 struct vattr va; 1508 fhandle_t *fh; 1509 int flags; 1510 1511 /* 1512 * Must be super user 1513 */ 1514 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1515 0, NULL, NULL, NULL))) 1516 return (error); 1517 1518 flags = FFLAGS(oflags); 1519 if ((flags & (FREAD | FWRITE)) == 0) 1520 return (EINVAL); 1521 if ((flags & O_CREAT)) 1522 return (EINVAL); 1523 /* falloc() will use the file descriptor for us */ 1524 if ((error = falloc(l, &nfp, &indx)) != 0) 1525 return (error); 1526 fp = nfp; 1527 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1528 if (error != 0) { 1529 goto bad; 1530 } 1531 error = vfs_fhtovp(fh, &vp); 1532 if (error != 0) { 1533 goto bad; 1534 } 1535 1536 /* Now do an effective vn_open */ 1537 1538 if (vp->v_type == VSOCK) { 1539 error = EOPNOTSUPP; 1540 goto bad; 1541 } 1542 if (flags & FREAD) { 1543 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1544 goto bad; 1545 } 1546 if (flags & (FWRITE | O_TRUNC)) { 1547 if (vp->v_type == VDIR) { 1548 error = EISDIR; 1549 goto bad; 1550 } 1551 if ((error = vn_writechk(vp)) != 0 || 1552 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1553 goto bad; 1554 } 1555 if (flags & O_TRUNC) { 1556 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1557 goto bad; 1558 VOP_UNLOCK(vp, 0); /* XXX */ 1559 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1560 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1561 VATTR_NULL(&va); 1562 va.va_size = 0; 1563 error = VOP_SETATTR(vp, &va, cred, l); 1564 vn_finished_write(mp, 0); 1565 if (error) 1566 goto bad; 1567 } 1568 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1569 goto bad; 1570 if (vp->v_type == VREG && 1571 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1572 error = EIO; 1573 goto bad; 1574 } 1575 if (flags & FWRITE) 1576 vp->v_writecount++; 1577 1578 /* done with modified vn_open, now finish what sys_open does. */ 1579 1580 fp->f_flag = flags & FMASK; 1581 fp->f_type = DTYPE_VNODE; 1582 fp->f_ops = &vnops; 1583 fp->f_data = vp; 1584 if (flags & (O_EXLOCK | O_SHLOCK)) { 1585 lf.l_whence = SEEK_SET; 1586 lf.l_start = 0; 1587 lf.l_len = 0; 1588 if (flags & O_EXLOCK) 1589 lf.l_type = F_WRLCK; 1590 else 1591 lf.l_type = F_RDLCK; 1592 type = F_FLOCK; 1593 if ((flags & FNONBLOCK) == 0) 1594 type |= F_WAIT; 1595 VOP_UNLOCK(vp, 0); 1596 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1597 if (error) { 1598 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1599 FILE_UNUSE(fp, l); 1600 ffree(fp); 1601 fdremove(fdp, indx); 1602 return (error); 1603 } 1604 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1605 fp->f_flag |= FHASLOCK; 1606 } 1607 VOP_UNLOCK(vp, 0); 1608 *retval = indx; 1609 FILE_SET_MATURE(fp); 1610 FILE_UNUSE(fp, l); 1611 vfs_copyinfh_free(fh); 1612 return (0); 1613 1614 bad: 1615 FILE_UNUSE(fp, l); 1616 ffree(fp); 1617 fdremove(fdp, indx); 1618 if (vp != NULL) 1619 vput(vp); 1620 vfs_copyinfh_free(fh); 1621 return (error); 1622 } 1623 1624 int 1625 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1626 { 1627 struct sys___fhopen40_args /* { 1628 syscallarg(const void *) fhp; 1629 syscallarg(size_t) fh_size; 1630 syscallarg(int) flags; 1631 } */ *uap = v; 1632 1633 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1634 SCARG(uap, flags), retval); 1635 } 1636 1637 int 1638 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1639 register_t *retval) 1640 { 1641 struct stat sb; 1642 int error; 1643 fhandle_t *fh; 1644 struct vnode *vp; 1645 1646 /* 1647 * Must be super user 1648 */ 1649 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1650 0, NULL, NULL, NULL))) 1651 return (error); 1652 1653 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1654 if (error != 0) { 1655 goto bad; 1656 } 1657 error = vfs_fhtovp(fh, &vp); 1658 if (error != 0) { 1659 goto bad; 1660 } 1661 error = vn_stat(vp, &sb, l); 1662 vput(vp); 1663 if (error) { 1664 goto bad; 1665 } 1666 error = copyout(&sb, sbp, sizeof(sb)); 1667 bad: 1668 vfs_copyinfh_free(fh); 1669 return error; 1670 } 1671 1672 1673 /* ARGSUSED */ 1674 int 1675 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1676 { 1677 struct sys___fhstat40_args /* { 1678 syscallarg(const void *) fhp; 1679 syscallarg(size_t) fh_size; 1680 syscallarg(struct stat *) sb; 1681 } */ *uap = v; 1682 1683 return dofhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), SCARG(uap, sb), 1684 retval); 1685 } 1686 1687 int 1688 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1689 int flags, register_t *retval) 1690 { 1691 struct statvfs *sb = NULL; 1692 fhandle_t *fh; 1693 struct mount *mp; 1694 struct vnode *vp; 1695 int error; 1696 1697 /* 1698 * Must be super user 1699 */ 1700 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1701 0, NULL, NULL, NULL))) 1702 return error; 1703 1704 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1705 if (error != 0) { 1706 goto out; 1707 } 1708 error = vfs_fhtovp(fh, &vp); 1709 if (error != 0) { 1710 goto out; 1711 } 1712 mp = vp->v_mount; 1713 sb = STATVFSBUF_GET(); 1714 if ((error = dostatvfs(mp, sb, l, flags, 1)) != 0) { 1715 vput(vp); 1716 goto out; 1717 } 1718 vput(vp); 1719 error = copyout(sb, buf, sizeof(*sb)); 1720 out: 1721 if (sb != NULL) { 1722 STATVFSBUF_PUT(sb); 1723 } 1724 vfs_copyinfh_free(fh); 1725 return error; 1726 } 1727 1728 /* ARGSUSED */ 1729 int 1730 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1731 { 1732 struct sys___fhstatvfs140_args /* { 1733 syscallarg(const void *) fhp; 1734 syscallarg(size_t) fh_size; 1735 syscallarg(struct statvfs *) buf; 1736 syscallarg(int) flags; 1737 } */ *uap = v; 1738 1739 return dofhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1740 SCARG(uap, buf), SCARG(uap, flags), retval); 1741 } 1742 1743 /* 1744 * Create a special file. 1745 */ 1746 /* ARGSUSED */ 1747 int 1748 sys_mknod(struct lwp *l, void *v, register_t *retval) 1749 { 1750 struct sys_mknod_args /* { 1751 syscallarg(const char *) path; 1752 syscallarg(int) mode; 1753 syscallarg(int) dev; 1754 } */ *uap = v; 1755 struct proc *p = l->l_proc; 1756 struct vnode *vp; 1757 struct mount *mp; 1758 struct vattr vattr; 1759 int error; 1760 int whiteout = 0; 1761 struct nameidata nd; 1762 1763 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1764 0, NULL, NULL, NULL)) != 0) 1765 return (error); 1766 restart: 1767 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1768 if ((error = namei(&nd)) != 0) 1769 return (error); 1770 vp = nd.ni_vp; 1771 if (vp != NULL) 1772 error = EEXIST; 1773 else { 1774 VATTR_NULL(&vattr); 1775 vattr.va_mode = 1776 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1777 vattr.va_rdev = SCARG(uap, dev); 1778 whiteout = 0; 1779 1780 switch (SCARG(uap, mode) & S_IFMT) { 1781 case S_IFMT: /* used by badsect to flag bad sectors */ 1782 vattr.va_type = VBAD; 1783 break; 1784 case S_IFCHR: 1785 vattr.va_type = VCHR; 1786 break; 1787 case S_IFBLK: 1788 vattr.va_type = VBLK; 1789 break; 1790 case S_IFWHT: 1791 whiteout = 1; 1792 break; 1793 default: 1794 error = EINVAL; 1795 break; 1796 } 1797 } 1798 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1799 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1800 if (nd.ni_dvp == vp) 1801 vrele(nd.ni_dvp); 1802 else 1803 vput(nd.ni_dvp); 1804 if (vp) 1805 vrele(vp); 1806 if ((error = vn_start_write(NULL, &mp, 1807 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1808 return (error); 1809 goto restart; 1810 } 1811 if (!error) { 1812 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1813 if (whiteout) { 1814 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1815 if (error) 1816 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1817 vput(nd.ni_dvp); 1818 } else { 1819 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1820 &nd.ni_cnd, &vattr); 1821 if (error == 0) 1822 vput(nd.ni_vp); 1823 } 1824 } else { 1825 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1826 if (nd.ni_dvp == vp) 1827 vrele(nd.ni_dvp); 1828 else 1829 vput(nd.ni_dvp); 1830 if (vp) 1831 vrele(vp); 1832 } 1833 vn_finished_write(mp, 0); 1834 return (error); 1835 } 1836 1837 /* 1838 * Create a named pipe. 1839 */ 1840 /* ARGSUSED */ 1841 int 1842 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1843 { 1844 struct sys_mkfifo_args /* { 1845 syscallarg(const char *) path; 1846 syscallarg(int) mode; 1847 } */ *uap = v; 1848 struct proc *p = l->l_proc; 1849 struct mount *mp; 1850 struct vattr vattr; 1851 int error; 1852 struct nameidata nd; 1853 1854 restart: 1855 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1856 if ((error = namei(&nd)) != 0) 1857 return (error); 1858 if (nd.ni_vp != NULL) { 1859 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1860 if (nd.ni_dvp == nd.ni_vp) 1861 vrele(nd.ni_dvp); 1862 else 1863 vput(nd.ni_dvp); 1864 vrele(nd.ni_vp); 1865 return (EEXIST); 1866 } 1867 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1868 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1869 if (nd.ni_dvp == nd.ni_vp) 1870 vrele(nd.ni_dvp); 1871 else 1872 vput(nd.ni_dvp); 1873 if (nd.ni_vp) 1874 vrele(nd.ni_vp); 1875 if ((error = vn_start_write(NULL, &mp, 1876 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1877 return (error); 1878 goto restart; 1879 } 1880 VATTR_NULL(&vattr); 1881 vattr.va_type = VFIFO; 1882 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1883 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1884 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1885 if (error == 0) 1886 vput(nd.ni_vp); 1887 vn_finished_write(mp, 0); 1888 return (error); 1889 } 1890 1891 /* 1892 * Make a hard file link. 1893 */ 1894 /* ARGSUSED */ 1895 int 1896 sys_link(struct lwp *l, void *v, register_t *retval) 1897 { 1898 struct sys_link_args /* { 1899 syscallarg(const char *) path; 1900 syscallarg(const char *) link; 1901 } */ *uap = v; 1902 struct vnode *vp; 1903 struct mount *mp; 1904 struct nameidata nd; 1905 int error; 1906 1907 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1908 if ((error = namei(&nd)) != 0) 1909 return (error); 1910 vp = nd.ni_vp; 1911 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1912 vrele(vp); 1913 return (error); 1914 } 1915 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1916 if ((error = namei(&nd)) != 0) 1917 goto out; 1918 if (nd.ni_vp) { 1919 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1920 if (nd.ni_dvp == nd.ni_vp) 1921 vrele(nd.ni_dvp); 1922 else 1923 vput(nd.ni_dvp); 1924 vrele(nd.ni_vp); 1925 error = EEXIST; 1926 goto out; 1927 } 1928 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1929 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1930 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1931 out: 1932 vrele(vp); 1933 vn_finished_write(mp, 0); 1934 return (error); 1935 } 1936 1937 /* 1938 * Make a symbolic link. 1939 */ 1940 /* ARGSUSED */ 1941 int 1942 sys_symlink(struct lwp *l, void *v, register_t *retval) 1943 { 1944 struct sys_symlink_args /* { 1945 syscallarg(const char *) path; 1946 syscallarg(const char *) link; 1947 } */ *uap = v; 1948 struct proc *p = l->l_proc; 1949 struct mount *mp; 1950 struct vattr vattr; 1951 char *path; 1952 int error; 1953 struct nameidata nd; 1954 1955 path = PNBUF_GET(); 1956 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1957 if (error) 1958 goto out; 1959 restart: 1960 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1961 if ((error = namei(&nd)) != 0) 1962 goto out; 1963 if (nd.ni_vp) { 1964 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1965 if (nd.ni_dvp == nd.ni_vp) 1966 vrele(nd.ni_dvp); 1967 else 1968 vput(nd.ni_dvp); 1969 vrele(nd.ni_vp); 1970 error = EEXIST; 1971 goto out; 1972 } 1973 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1974 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1975 if (nd.ni_dvp == nd.ni_vp) 1976 vrele(nd.ni_dvp); 1977 else 1978 vput(nd.ni_dvp); 1979 if ((error = vn_start_write(NULL, &mp, 1980 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1981 return (error); 1982 goto restart; 1983 } 1984 VATTR_NULL(&vattr); 1985 vattr.va_type = VLNK; 1986 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1987 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1988 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1989 if (error == 0) 1990 vput(nd.ni_vp); 1991 vn_finished_write(mp, 0); 1992 out: 1993 PNBUF_PUT(path); 1994 return (error); 1995 } 1996 1997 /* 1998 * Delete a whiteout from the filesystem. 1999 */ 2000 /* ARGSUSED */ 2001 int 2002 sys_undelete(struct lwp *l, void *v, register_t *retval) 2003 { 2004 struct sys_undelete_args /* { 2005 syscallarg(const char *) path; 2006 } */ *uap = v; 2007 int error; 2008 struct mount *mp; 2009 struct nameidata nd; 2010 2011 restart: 2012 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2013 SCARG(uap, path), l); 2014 error = namei(&nd); 2015 if (error) 2016 return (error); 2017 2018 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2019 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2020 if (nd.ni_dvp == nd.ni_vp) 2021 vrele(nd.ni_dvp); 2022 else 2023 vput(nd.ni_dvp); 2024 if (nd.ni_vp) 2025 vrele(nd.ni_vp); 2026 return (EEXIST); 2027 } 2028 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2029 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2030 if (nd.ni_dvp == nd.ni_vp) 2031 vrele(nd.ni_dvp); 2032 else 2033 vput(nd.ni_dvp); 2034 if ((error = vn_start_write(NULL, &mp, 2035 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2036 return (error); 2037 goto restart; 2038 } 2039 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2040 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2041 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2042 vput(nd.ni_dvp); 2043 vn_finished_write(mp, 0); 2044 return (error); 2045 } 2046 2047 /* 2048 * Delete a name from the filesystem. 2049 */ 2050 /* ARGSUSED */ 2051 int 2052 sys_unlink(struct lwp *l, void *v, register_t *retval) 2053 { 2054 struct sys_unlink_args /* { 2055 syscallarg(const char *) path; 2056 } */ *uap = v; 2057 struct mount *mp; 2058 struct vnode *vp; 2059 int error; 2060 struct nameidata nd; 2061 #if NVERIEXEC > 0 2062 pathname_t pathbuf; 2063 #endif /* NVERIEXEC > 0 */ 2064 2065 restart: 2066 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2067 SCARG(uap, path), l); 2068 if ((error = namei(&nd)) != 0) 2069 return (error); 2070 vp = nd.ni_vp; 2071 2072 /* 2073 * The root of a mounted filesystem cannot be deleted. 2074 */ 2075 if (vp->v_flag & VROOT) { 2076 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2077 if (nd.ni_dvp == vp) 2078 vrele(nd.ni_dvp); 2079 else 2080 vput(nd.ni_dvp); 2081 vput(vp); 2082 error = EBUSY; 2083 goto out; 2084 } 2085 2086 #if NVERIEXEC > 0 2087 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2088 2089 /* Handle remove requests for veriexec entries. */ 2090 if (!error) { 2091 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2092 pathname_put(pathbuf); 2093 } 2094 2095 if (error) { 2096 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2097 if (nd.ni_dvp == vp) 2098 vrele(nd.ni_dvp); 2099 else 2100 vput(nd.ni_dvp); 2101 vput(vp); 2102 goto out; 2103 } 2104 #endif /* NVERIEXEC > 0 */ 2105 2106 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2107 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2108 if (nd.ni_dvp == vp) 2109 vrele(nd.ni_dvp); 2110 else 2111 vput(nd.ni_dvp); 2112 vput(vp); 2113 if ((error = vn_start_write(NULL, &mp, 2114 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2115 return (error); 2116 goto restart; 2117 } 2118 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2119 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2120 #ifdef FILEASSOC 2121 (void)fileassoc_file_delete(vp); 2122 #endif /* FILEASSOC */ 2123 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2124 vn_finished_write(mp, 0); 2125 out: 2126 return (error); 2127 } 2128 2129 /* 2130 * Reposition read/write file offset. 2131 */ 2132 int 2133 sys_lseek(struct lwp *l, void *v, register_t *retval) 2134 { 2135 struct sys_lseek_args /* { 2136 syscallarg(int) fd; 2137 syscallarg(int) pad; 2138 syscallarg(off_t) offset; 2139 syscallarg(int) whence; 2140 } */ *uap = v; 2141 struct proc *p = l->l_proc; 2142 kauth_cred_t cred = l->l_cred; 2143 struct filedesc *fdp = p->p_fd; 2144 struct file *fp; 2145 struct vnode *vp; 2146 struct vattr vattr; 2147 off_t newoff; 2148 int error; 2149 2150 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2151 return (EBADF); 2152 2153 FILE_USE(fp); 2154 2155 vp = (struct vnode *)fp->f_data; 2156 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2157 error = ESPIPE; 2158 goto out; 2159 } 2160 2161 switch (SCARG(uap, whence)) { 2162 case SEEK_CUR: 2163 newoff = fp->f_offset + SCARG(uap, offset); 2164 break; 2165 case SEEK_END: 2166 error = VOP_GETATTR(vp, &vattr, cred, l); 2167 if (error) 2168 goto out; 2169 newoff = SCARG(uap, offset) + vattr.va_size; 2170 break; 2171 case SEEK_SET: 2172 newoff = SCARG(uap, offset); 2173 break; 2174 default: 2175 error = EINVAL; 2176 goto out; 2177 } 2178 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2179 goto out; 2180 2181 *(off_t *)retval = fp->f_offset = newoff; 2182 out: 2183 FILE_UNUSE(fp, l); 2184 return (error); 2185 } 2186 2187 /* 2188 * Positional read system call. 2189 */ 2190 int 2191 sys_pread(struct lwp *l, void *v, register_t *retval) 2192 { 2193 struct sys_pread_args /* { 2194 syscallarg(int) fd; 2195 syscallarg(void *) buf; 2196 syscallarg(size_t) nbyte; 2197 syscallarg(off_t) offset; 2198 } */ *uap = v; 2199 struct proc *p = l->l_proc; 2200 struct filedesc *fdp = p->p_fd; 2201 struct file *fp; 2202 struct vnode *vp; 2203 off_t offset; 2204 int error, fd = SCARG(uap, fd); 2205 2206 if ((fp = fd_getfile(fdp, fd)) == NULL) 2207 return (EBADF); 2208 2209 if ((fp->f_flag & FREAD) == 0) { 2210 simple_unlock(&fp->f_slock); 2211 return (EBADF); 2212 } 2213 2214 FILE_USE(fp); 2215 2216 vp = (struct vnode *)fp->f_data; 2217 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2218 error = ESPIPE; 2219 goto out; 2220 } 2221 2222 offset = SCARG(uap, offset); 2223 2224 /* 2225 * XXX This works because no file systems actually 2226 * XXX take any action on the seek operation. 2227 */ 2228 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2229 goto out; 2230 2231 /* dofileread() will unuse the descriptor for us */ 2232 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2233 &offset, 0, retval)); 2234 2235 out: 2236 FILE_UNUSE(fp, l); 2237 return (error); 2238 } 2239 2240 /* 2241 * Positional scatter read system call. 2242 */ 2243 int 2244 sys_preadv(struct lwp *l, void *v, register_t *retval) 2245 { 2246 struct sys_preadv_args /* { 2247 syscallarg(int) fd; 2248 syscallarg(const struct iovec *) iovp; 2249 syscallarg(int) iovcnt; 2250 syscallarg(off_t) offset; 2251 } */ *uap = v; 2252 struct proc *p = l->l_proc; 2253 struct filedesc *fdp = p->p_fd; 2254 struct file *fp; 2255 struct vnode *vp; 2256 off_t offset; 2257 int error, fd = SCARG(uap, fd); 2258 2259 if ((fp = fd_getfile(fdp, fd)) == NULL) 2260 return (EBADF); 2261 2262 if ((fp->f_flag & FREAD) == 0) { 2263 simple_unlock(&fp->f_slock); 2264 return (EBADF); 2265 } 2266 2267 FILE_USE(fp); 2268 2269 vp = (struct vnode *)fp->f_data; 2270 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2271 error = ESPIPE; 2272 goto out; 2273 } 2274 2275 offset = SCARG(uap, offset); 2276 2277 /* 2278 * XXX This works because no file systems actually 2279 * XXX take any action on the seek operation. 2280 */ 2281 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2282 goto out; 2283 2284 /* dofilereadv() will unuse the descriptor for us */ 2285 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2286 &offset, 0, retval)); 2287 2288 out: 2289 FILE_UNUSE(fp, l); 2290 return (error); 2291 } 2292 2293 /* 2294 * Positional write system call. 2295 */ 2296 int 2297 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2298 { 2299 struct sys_pwrite_args /* { 2300 syscallarg(int) fd; 2301 syscallarg(const void *) buf; 2302 syscallarg(size_t) nbyte; 2303 syscallarg(off_t) offset; 2304 } */ *uap = v; 2305 struct proc *p = l->l_proc; 2306 struct filedesc *fdp = p->p_fd; 2307 struct file *fp; 2308 struct vnode *vp; 2309 off_t offset; 2310 int error, fd = SCARG(uap, fd); 2311 2312 if ((fp = fd_getfile(fdp, fd)) == NULL) 2313 return (EBADF); 2314 2315 if ((fp->f_flag & FWRITE) == 0) { 2316 simple_unlock(&fp->f_slock); 2317 return (EBADF); 2318 } 2319 2320 FILE_USE(fp); 2321 2322 vp = (struct vnode *)fp->f_data; 2323 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2324 error = ESPIPE; 2325 goto out; 2326 } 2327 2328 offset = SCARG(uap, offset); 2329 2330 /* 2331 * XXX This works because no file systems actually 2332 * XXX take any action on the seek operation. 2333 */ 2334 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2335 goto out; 2336 2337 /* dofilewrite() will unuse the descriptor for us */ 2338 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2339 &offset, 0, retval)); 2340 2341 out: 2342 FILE_UNUSE(fp, l); 2343 return (error); 2344 } 2345 2346 /* 2347 * Positional gather write system call. 2348 */ 2349 int 2350 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2351 { 2352 struct sys_pwritev_args /* { 2353 syscallarg(int) fd; 2354 syscallarg(const struct iovec *) iovp; 2355 syscallarg(int) iovcnt; 2356 syscallarg(off_t) offset; 2357 } */ *uap = v; 2358 struct proc *p = l->l_proc; 2359 struct filedesc *fdp = p->p_fd; 2360 struct file *fp; 2361 struct vnode *vp; 2362 off_t offset; 2363 int error, fd = SCARG(uap, fd); 2364 2365 if ((fp = fd_getfile(fdp, fd)) == NULL) 2366 return (EBADF); 2367 2368 if ((fp->f_flag & FWRITE) == 0) { 2369 simple_unlock(&fp->f_slock); 2370 return (EBADF); 2371 } 2372 2373 FILE_USE(fp); 2374 2375 vp = (struct vnode *)fp->f_data; 2376 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2377 error = ESPIPE; 2378 goto out; 2379 } 2380 2381 offset = SCARG(uap, offset); 2382 2383 /* 2384 * XXX This works because no file systems actually 2385 * XXX take any action on the seek operation. 2386 */ 2387 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2388 goto out; 2389 2390 /* dofilewritev() will unuse the descriptor for us */ 2391 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2392 &offset, 0, retval)); 2393 2394 out: 2395 FILE_UNUSE(fp, l); 2396 return (error); 2397 } 2398 2399 /* 2400 * Check access permissions. 2401 */ 2402 int 2403 sys_access(struct lwp *l, void *v, register_t *retval) 2404 { 2405 struct sys_access_args /* { 2406 syscallarg(const char *) path; 2407 syscallarg(int) flags; 2408 } */ *uap = v; 2409 kauth_cred_t cred; 2410 struct vnode *vp; 2411 int error, flags; 2412 struct nameidata nd; 2413 2414 cred = kauth_cred_dup(l->l_cred); 2415 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2416 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2417 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2418 SCARG(uap, path), l); 2419 /* Override default credentials */ 2420 nd.ni_cnd.cn_cred = cred; 2421 if ((error = namei(&nd)) != 0) 2422 goto out; 2423 vp = nd.ni_vp; 2424 2425 /* Flags == 0 means only check for existence. */ 2426 if (SCARG(uap, flags)) { 2427 flags = 0; 2428 if (SCARG(uap, flags) & R_OK) 2429 flags |= VREAD; 2430 if (SCARG(uap, flags) & W_OK) 2431 flags |= VWRITE; 2432 if (SCARG(uap, flags) & X_OK) 2433 flags |= VEXEC; 2434 2435 error = VOP_ACCESS(vp, flags, cred, l); 2436 if (!error && (flags & VWRITE)) 2437 error = vn_writechk(vp); 2438 } 2439 vput(vp); 2440 out: 2441 kauth_cred_free(cred); 2442 return (error); 2443 } 2444 2445 /* 2446 * Get file status; this version follows links. 2447 */ 2448 /* ARGSUSED */ 2449 int 2450 sys___stat30(struct lwp *l, void *v, register_t *retval) 2451 { 2452 struct sys___stat30_args /* { 2453 syscallarg(const char *) path; 2454 syscallarg(struct stat *) ub; 2455 } */ *uap = v; 2456 struct stat sb; 2457 int error; 2458 struct nameidata nd; 2459 2460 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2461 SCARG(uap, path), l); 2462 if ((error = namei(&nd)) != 0) 2463 return (error); 2464 error = vn_stat(nd.ni_vp, &sb, l); 2465 vput(nd.ni_vp); 2466 if (error) 2467 return (error); 2468 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2469 return (error); 2470 } 2471 2472 /* 2473 * Get file status; this version does not follow links. 2474 */ 2475 /* ARGSUSED */ 2476 int 2477 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2478 { 2479 struct sys___lstat30_args /* { 2480 syscallarg(const char *) path; 2481 syscallarg(struct stat *) ub; 2482 } */ *uap = v; 2483 struct stat sb; 2484 int error; 2485 struct nameidata nd; 2486 2487 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2488 SCARG(uap, path), l); 2489 if ((error = namei(&nd)) != 0) 2490 return (error); 2491 error = vn_stat(nd.ni_vp, &sb, l); 2492 vput(nd.ni_vp); 2493 if (error) 2494 return (error); 2495 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2496 return (error); 2497 } 2498 2499 /* 2500 * Get configurable pathname variables. 2501 */ 2502 /* ARGSUSED */ 2503 int 2504 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2505 { 2506 struct sys_pathconf_args /* { 2507 syscallarg(const char *) path; 2508 syscallarg(int) name; 2509 } */ *uap = v; 2510 int error; 2511 struct nameidata nd; 2512 2513 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2514 SCARG(uap, path), l); 2515 if ((error = namei(&nd)) != 0) 2516 return (error); 2517 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2518 vput(nd.ni_vp); 2519 return (error); 2520 } 2521 2522 /* 2523 * Return target name of a symbolic link. 2524 */ 2525 /* ARGSUSED */ 2526 int 2527 sys_readlink(struct lwp *l, void *v, register_t *retval) 2528 { 2529 struct sys_readlink_args /* { 2530 syscallarg(const char *) path; 2531 syscallarg(char *) buf; 2532 syscallarg(size_t) count; 2533 } */ *uap = v; 2534 struct vnode *vp; 2535 struct iovec aiov; 2536 struct uio auio; 2537 int error; 2538 struct nameidata nd; 2539 2540 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2541 SCARG(uap, path), l); 2542 if ((error = namei(&nd)) != 0) 2543 return (error); 2544 vp = nd.ni_vp; 2545 if (vp->v_type != VLNK) 2546 error = EINVAL; 2547 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2548 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2549 aiov.iov_base = SCARG(uap, buf); 2550 aiov.iov_len = SCARG(uap, count); 2551 auio.uio_iov = &aiov; 2552 auio.uio_iovcnt = 1; 2553 auio.uio_offset = 0; 2554 auio.uio_rw = UIO_READ; 2555 KASSERT(l == curlwp); 2556 auio.uio_vmspace = l->l_proc->p_vmspace; 2557 auio.uio_resid = SCARG(uap, count); 2558 error = VOP_READLINK(vp, &auio, l->l_cred); 2559 } 2560 vput(vp); 2561 *retval = SCARG(uap, count) - auio.uio_resid; 2562 return (error); 2563 } 2564 2565 /* 2566 * Change flags of a file given a path name. 2567 */ 2568 /* ARGSUSED */ 2569 int 2570 sys_chflags(struct lwp *l, void *v, register_t *retval) 2571 { 2572 struct sys_chflags_args /* { 2573 syscallarg(const char *) path; 2574 syscallarg(u_long) flags; 2575 } */ *uap = v; 2576 struct vnode *vp; 2577 int error; 2578 struct nameidata nd; 2579 2580 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2581 if ((error = namei(&nd)) != 0) 2582 return (error); 2583 vp = nd.ni_vp; 2584 error = change_flags(vp, SCARG(uap, flags), l); 2585 vput(vp); 2586 return (error); 2587 } 2588 2589 /* 2590 * Change flags of a file given a file descriptor. 2591 */ 2592 /* ARGSUSED */ 2593 int 2594 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2595 { 2596 struct sys_fchflags_args /* { 2597 syscallarg(int) fd; 2598 syscallarg(u_long) flags; 2599 } */ *uap = v; 2600 struct proc *p = l->l_proc; 2601 struct vnode *vp; 2602 struct file *fp; 2603 int error; 2604 2605 /* getvnode() will use the descriptor for us */ 2606 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2607 return (error); 2608 vp = (struct vnode *)fp->f_data; 2609 error = change_flags(vp, SCARG(uap, flags), l); 2610 VOP_UNLOCK(vp, 0); 2611 FILE_UNUSE(fp, l); 2612 return (error); 2613 } 2614 2615 /* 2616 * Change flags of a file given a path name; this version does 2617 * not follow links. 2618 */ 2619 int 2620 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2621 { 2622 struct sys_lchflags_args /* { 2623 syscallarg(const char *) path; 2624 syscallarg(u_long) flags; 2625 } */ *uap = v; 2626 struct vnode *vp; 2627 int error; 2628 struct nameidata nd; 2629 2630 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2631 if ((error = namei(&nd)) != 0) 2632 return (error); 2633 vp = nd.ni_vp; 2634 error = change_flags(vp, SCARG(uap, flags), l); 2635 vput(vp); 2636 return (error); 2637 } 2638 2639 /* 2640 * Common routine to change flags of a file. 2641 */ 2642 int 2643 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2644 { 2645 struct mount *mp; 2646 struct vattr vattr; 2647 int error; 2648 2649 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2650 return (error); 2651 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2652 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2653 /* 2654 * Non-superusers cannot change the flags on devices, even if they 2655 * own them. 2656 */ 2657 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2658 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2659 goto out; 2660 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2661 error = EINVAL; 2662 goto out; 2663 } 2664 } 2665 VATTR_NULL(&vattr); 2666 vattr.va_flags = flags; 2667 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2668 out: 2669 vn_finished_write(mp, 0); 2670 return (error); 2671 } 2672 2673 /* 2674 * Change mode of a file given path name; this version follows links. 2675 */ 2676 /* ARGSUSED */ 2677 int 2678 sys_chmod(struct lwp *l, void *v, register_t *retval) 2679 { 2680 struct sys_chmod_args /* { 2681 syscallarg(const char *) path; 2682 syscallarg(int) mode; 2683 } */ *uap = v; 2684 int error; 2685 struct nameidata nd; 2686 2687 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2688 if ((error = namei(&nd)) != 0) 2689 return (error); 2690 2691 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2692 2693 vrele(nd.ni_vp); 2694 return (error); 2695 } 2696 2697 /* 2698 * Change mode of a file given a file descriptor. 2699 */ 2700 /* ARGSUSED */ 2701 int 2702 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2703 { 2704 struct sys_fchmod_args /* { 2705 syscallarg(int) fd; 2706 syscallarg(int) mode; 2707 } */ *uap = v; 2708 struct proc *p = l->l_proc; 2709 struct file *fp; 2710 int error; 2711 2712 /* getvnode() will use the descriptor for us */ 2713 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2714 return (error); 2715 2716 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2717 FILE_UNUSE(fp, l); 2718 return (error); 2719 } 2720 2721 /* 2722 * Change mode of a file given path name; this version does not follow links. 2723 */ 2724 /* ARGSUSED */ 2725 int 2726 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2727 { 2728 struct sys_lchmod_args /* { 2729 syscallarg(const char *) path; 2730 syscallarg(int) mode; 2731 } */ *uap = v; 2732 int error; 2733 struct nameidata nd; 2734 2735 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2736 if ((error = namei(&nd)) != 0) 2737 return (error); 2738 2739 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2740 2741 vrele(nd.ni_vp); 2742 return (error); 2743 } 2744 2745 /* 2746 * Common routine to set mode given a vnode. 2747 */ 2748 static int 2749 change_mode(struct vnode *vp, int mode, struct lwp *l) 2750 { 2751 struct mount *mp; 2752 struct vattr vattr; 2753 int error; 2754 2755 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2756 return (error); 2757 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2758 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2759 VATTR_NULL(&vattr); 2760 vattr.va_mode = mode & ALLPERMS; 2761 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2762 VOP_UNLOCK(vp, 0); 2763 vn_finished_write(mp, 0); 2764 return (error); 2765 } 2766 2767 /* 2768 * Set ownership given a path name; this version follows links. 2769 */ 2770 /* ARGSUSED */ 2771 int 2772 sys_chown(struct lwp *l, void *v, register_t *retval) 2773 { 2774 struct sys_chown_args /* { 2775 syscallarg(const char *) path; 2776 syscallarg(uid_t) uid; 2777 syscallarg(gid_t) gid; 2778 } */ *uap = v; 2779 int error; 2780 struct nameidata nd; 2781 2782 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2783 if ((error = namei(&nd)) != 0) 2784 return (error); 2785 2786 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2787 2788 vrele(nd.ni_vp); 2789 return (error); 2790 } 2791 2792 /* 2793 * Set ownership given a path name; this version follows links. 2794 * Provides POSIX semantics. 2795 */ 2796 /* ARGSUSED */ 2797 int 2798 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2799 { 2800 struct sys_chown_args /* { 2801 syscallarg(const char *) path; 2802 syscallarg(uid_t) uid; 2803 syscallarg(gid_t) gid; 2804 } */ *uap = v; 2805 int error; 2806 struct nameidata nd; 2807 2808 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2809 if ((error = namei(&nd)) != 0) 2810 return (error); 2811 2812 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2813 2814 vrele(nd.ni_vp); 2815 return (error); 2816 } 2817 2818 /* 2819 * Set ownership given a file descriptor. 2820 */ 2821 /* ARGSUSED */ 2822 int 2823 sys_fchown(struct lwp *l, void *v, register_t *retval) 2824 { 2825 struct sys_fchown_args /* { 2826 syscallarg(int) fd; 2827 syscallarg(uid_t) uid; 2828 syscallarg(gid_t) gid; 2829 } */ *uap = v; 2830 struct proc *p = l->l_proc; 2831 int error; 2832 struct file *fp; 2833 2834 /* getvnode() will use the descriptor for us */ 2835 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2836 return (error); 2837 2838 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2839 SCARG(uap, gid), l, 0); 2840 FILE_UNUSE(fp, l); 2841 return (error); 2842 } 2843 2844 /* 2845 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2846 */ 2847 /* ARGSUSED */ 2848 int 2849 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2850 { 2851 struct sys_fchown_args /* { 2852 syscallarg(int) fd; 2853 syscallarg(uid_t) uid; 2854 syscallarg(gid_t) gid; 2855 } */ *uap = v; 2856 struct proc *p = l->l_proc; 2857 int error; 2858 struct file *fp; 2859 2860 /* getvnode() will use the descriptor for us */ 2861 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2862 return (error); 2863 2864 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2865 SCARG(uap, gid), l, 1); 2866 FILE_UNUSE(fp, l); 2867 return (error); 2868 } 2869 2870 /* 2871 * Set ownership given a path name; this version does not follow links. 2872 */ 2873 /* ARGSUSED */ 2874 int 2875 sys_lchown(struct lwp *l, void *v, register_t *retval) 2876 { 2877 struct sys_lchown_args /* { 2878 syscallarg(const char *) path; 2879 syscallarg(uid_t) uid; 2880 syscallarg(gid_t) gid; 2881 } */ *uap = v; 2882 int error; 2883 struct nameidata nd; 2884 2885 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2886 if ((error = namei(&nd)) != 0) 2887 return (error); 2888 2889 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2890 2891 vrele(nd.ni_vp); 2892 return (error); 2893 } 2894 2895 /* 2896 * Set ownership given a path name; this version does not follow links. 2897 * Provides POSIX/XPG semantics. 2898 */ 2899 /* ARGSUSED */ 2900 int 2901 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2902 { 2903 struct sys_lchown_args /* { 2904 syscallarg(const char *) path; 2905 syscallarg(uid_t) uid; 2906 syscallarg(gid_t) gid; 2907 } */ *uap = v; 2908 int error; 2909 struct nameidata nd; 2910 2911 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2912 if ((error = namei(&nd)) != 0) 2913 return (error); 2914 2915 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2916 2917 vrele(nd.ni_vp); 2918 return (error); 2919 } 2920 2921 /* 2922 * Common routine to set ownership given a vnode. 2923 */ 2924 static int 2925 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2926 int posix_semantics) 2927 { 2928 struct mount *mp; 2929 struct vattr vattr; 2930 mode_t newmode; 2931 int error; 2932 2933 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2934 return (error); 2935 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2936 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2937 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2938 goto out; 2939 2940 #define CHANGED(x) ((int)(x) != -1) 2941 newmode = vattr.va_mode; 2942 if (posix_semantics) { 2943 /* 2944 * POSIX/XPG semantics: if the caller is not the super-user, 2945 * clear set-user-id and set-group-id bits. Both POSIX and 2946 * the XPG consider the behaviour for calls by the super-user 2947 * implementation-defined; we leave the set-user-id and set- 2948 * group-id settings intact in that case. 2949 */ 2950 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2951 NULL) != 0) 2952 newmode &= ~(S_ISUID | S_ISGID); 2953 } else { 2954 /* 2955 * NetBSD semantics: when changing owner and/or group, 2956 * clear the respective bit(s). 2957 */ 2958 if (CHANGED(uid)) 2959 newmode &= ~S_ISUID; 2960 if (CHANGED(gid)) 2961 newmode &= ~S_ISGID; 2962 } 2963 /* Update va_mode iff altered. */ 2964 if (vattr.va_mode == newmode) 2965 newmode = VNOVAL; 2966 2967 VATTR_NULL(&vattr); 2968 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2969 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2970 vattr.va_mode = newmode; 2971 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2972 #undef CHANGED 2973 2974 out: 2975 VOP_UNLOCK(vp, 0); 2976 vn_finished_write(mp, 0); 2977 return (error); 2978 } 2979 2980 /* 2981 * Set the access and modification times given a path name; this 2982 * version follows links. 2983 */ 2984 /* ARGSUSED */ 2985 int 2986 sys_utimes(struct lwp *l, void *v, register_t *retval) 2987 { 2988 struct sys_utimes_args /* { 2989 syscallarg(const char *) path; 2990 syscallarg(const struct timeval *) tptr; 2991 } */ *uap = v; 2992 int error; 2993 struct nameidata nd; 2994 2995 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2996 if ((error = namei(&nd)) != 0) 2997 return (error); 2998 2999 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3000 3001 vrele(nd.ni_vp); 3002 return (error); 3003 } 3004 3005 /* 3006 * Set the access and modification times given a file descriptor. 3007 */ 3008 /* ARGSUSED */ 3009 int 3010 sys_futimes(struct lwp *l, void *v, register_t *retval) 3011 { 3012 struct sys_futimes_args /* { 3013 syscallarg(int) fd; 3014 syscallarg(const struct timeval *) tptr; 3015 } */ *uap = v; 3016 struct proc *p = l->l_proc; 3017 int error; 3018 struct file *fp; 3019 3020 /* getvnode() will use the descriptor for us */ 3021 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3022 return (error); 3023 3024 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3025 FILE_UNUSE(fp, l); 3026 return (error); 3027 } 3028 3029 /* 3030 * Set the access and modification times given a path name; this 3031 * version does not follow links. 3032 */ 3033 /* ARGSUSED */ 3034 int 3035 sys_lutimes(struct lwp *l, void *v, register_t *retval) 3036 { 3037 struct sys_lutimes_args /* { 3038 syscallarg(const char *) path; 3039 syscallarg(const struct timeval *) tptr; 3040 } */ *uap = v; 3041 int error; 3042 struct nameidata nd; 3043 3044 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3045 if ((error = namei(&nd)) != 0) 3046 return (error); 3047 3048 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3049 3050 vrele(nd.ni_vp); 3051 return (error); 3052 } 3053 3054 /* 3055 * Common routine to set access and modification times given a vnode. 3056 */ 3057 static int 3058 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3059 { 3060 struct mount *mp; 3061 struct vattr vattr; 3062 int error; 3063 3064 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3065 return (error); 3066 VATTR_NULL(&vattr); 3067 if (tptr == NULL) { 3068 nanotime(&vattr.va_atime); 3069 vattr.va_mtime = vattr.va_atime; 3070 vattr.va_vaflags |= VA_UTIMES_NULL; 3071 } else { 3072 struct timeval tv[2]; 3073 3074 error = copyin(tptr, tv, sizeof(tv)); 3075 if (error) 3076 goto out; 3077 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3078 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3079 } 3080 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3081 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3082 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3083 VOP_UNLOCK(vp, 0); 3084 out: 3085 vn_finished_write(mp, 0); 3086 return (error); 3087 } 3088 3089 /* 3090 * Truncate a file given its path name. 3091 */ 3092 /* ARGSUSED */ 3093 int 3094 sys_truncate(struct lwp *l, void *v, register_t *retval) 3095 { 3096 struct sys_truncate_args /* { 3097 syscallarg(const char *) path; 3098 syscallarg(int) pad; 3099 syscallarg(off_t) length; 3100 } */ *uap = v; 3101 struct vnode *vp; 3102 struct mount *mp; 3103 struct vattr vattr; 3104 int error; 3105 struct nameidata nd; 3106 3107 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3108 if ((error = namei(&nd)) != 0) 3109 return (error); 3110 vp = nd.ni_vp; 3111 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3112 vrele(vp); 3113 return (error); 3114 } 3115 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3116 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3117 if (vp->v_type == VDIR) 3118 error = EISDIR; 3119 else if ((error = vn_writechk(vp)) == 0 && 3120 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3121 VATTR_NULL(&vattr); 3122 vattr.va_size = SCARG(uap, length); 3123 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3124 } 3125 vput(vp); 3126 vn_finished_write(mp, 0); 3127 return (error); 3128 } 3129 3130 /* 3131 * Truncate a file given a file descriptor. 3132 */ 3133 /* ARGSUSED */ 3134 int 3135 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3136 { 3137 struct sys_ftruncate_args /* { 3138 syscallarg(int) fd; 3139 syscallarg(int) pad; 3140 syscallarg(off_t) length; 3141 } */ *uap = v; 3142 struct proc *p = l->l_proc; 3143 struct mount *mp; 3144 struct vattr vattr; 3145 struct vnode *vp; 3146 struct file *fp; 3147 int error; 3148 3149 /* getvnode() will use the descriptor for us */ 3150 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3151 return (error); 3152 if ((fp->f_flag & FWRITE) == 0) { 3153 error = EINVAL; 3154 goto out; 3155 } 3156 vp = (struct vnode *)fp->f_data; 3157 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3158 FILE_UNUSE(fp, l); 3159 return (error); 3160 } 3161 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3162 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3163 if (vp->v_type == VDIR) 3164 error = EISDIR; 3165 else if ((error = vn_writechk(vp)) == 0) { 3166 VATTR_NULL(&vattr); 3167 vattr.va_size = SCARG(uap, length); 3168 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3169 } 3170 VOP_UNLOCK(vp, 0); 3171 vn_finished_write(mp, 0); 3172 out: 3173 FILE_UNUSE(fp, l); 3174 return (error); 3175 } 3176 3177 /* 3178 * Sync an open file. 3179 */ 3180 /* ARGSUSED */ 3181 int 3182 sys_fsync(struct lwp *l, void *v, register_t *retval) 3183 { 3184 struct sys_fsync_args /* { 3185 syscallarg(int) fd; 3186 } */ *uap = v; 3187 struct proc *p = l->l_proc; 3188 struct vnode *vp; 3189 struct mount *mp; 3190 struct file *fp; 3191 int error; 3192 3193 /* getvnode() will use the descriptor for us */ 3194 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3195 return (error); 3196 vp = (struct vnode *)fp->f_data; 3197 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3198 FILE_UNUSE(fp, l); 3199 return (error); 3200 } 3201 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3202 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3203 if (error == 0 && bioops.io_fsync != NULL && 3204 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3205 (*bioops.io_fsync)(vp, 0); 3206 VOP_UNLOCK(vp, 0); 3207 vn_finished_write(mp, 0); 3208 FILE_UNUSE(fp, l); 3209 return (error); 3210 } 3211 3212 /* 3213 * Sync a range of file data. API modeled after that found in AIX. 3214 * 3215 * FDATASYNC indicates that we need only save enough metadata to be able 3216 * to re-read the written data. Note we duplicate AIX's requirement that 3217 * the file be open for writing. 3218 */ 3219 /* ARGSUSED */ 3220 int 3221 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3222 { 3223 struct sys_fsync_range_args /* { 3224 syscallarg(int) fd; 3225 syscallarg(int) flags; 3226 syscallarg(off_t) start; 3227 syscallarg(off_t) length; 3228 } */ *uap = v; 3229 struct proc *p = l->l_proc; 3230 struct vnode *vp; 3231 struct file *fp; 3232 int flags, nflags; 3233 off_t s, e, len; 3234 int error; 3235 3236 /* getvnode() will use the descriptor for us */ 3237 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3238 return (error); 3239 3240 if ((fp->f_flag & FWRITE) == 0) { 3241 error = EBADF; 3242 goto out; 3243 } 3244 3245 flags = SCARG(uap, flags); 3246 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3247 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3248 error = EINVAL; 3249 goto out; 3250 } 3251 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3252 if (flags & FDATASYNC) 3253 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3254 else 3255 nflags = FSYNC_WAIT; 3256 if (flags & FDISKSYNC) 3257 nflags |= FSYNC_CACHE; 3258 3259 len = SCARG(uap, length); 3260 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3261 if (len) { 3262 s = SCARG(uap, start); 3263 e = s + len; 3264 if (e < s) { 3265 error = EINVAL; 3266 goto out; 3267 } 3268 } else { 3269 e = 0; 3270 s = 0; 3271 } 3272 3273 vp = (struct vnode *)fp->f_data; 3274 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3275 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3276 3277 if (error == 0 && bioops.io_fsync != NULL && 3278 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3279 (*bioops.io_fsync)(vp, nflags); 3280 3281 VOP_UNLOCK(vp, 0); 3282 out: 3283 FILE_UNUSE(fp, l); 3284 return (error); 3285 } 3286 3287 /* 3288 * Sync the data of an open file. 3289 */ 3290 /* ARGSUSED */ 3291 int 3292 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3293 { 3294 struct sys_fdatasync_args /* { 3295 syscallarg(int) fd; 3296 } */ *uap = v; 3297 struct proc *p = l->l_proc; 3298 struct vnode *vp; 3299 struct file *fp; 3300 int error; 3301 3302 /* getvnode() will use the descriptor for us */ 3303 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3304 return (error); 3305 if ((fp->f_flag & FWRITE) == 0) { 3306 FILE_UNUSE(fp, l); 3307 return (EBADF); 3308 } 3309 vp = (struct vnode *)fp->f_data; 3310 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3311 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3312 VOP_UNLOCK(vp, 0); 3313 FILE_UNUSE(fp, l); 3314 return (error); 3315 } 3316 3317 /* 3318 * Rename files, (standard) BSD semantics frontend. 3319 */ 3320 /* ARGSUSED */ 3321 int 3322 sys_rename(struct lwp *l, void *v, register_t *retval) 3323 { 3324 struct sys_rename_args /* { 3325 syscallarg(const char *) from; 3326 syscallarg(const char *) to; 3327 } */ *uap = v; 3328 3329 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3330 } 3331 3332 /* 3333 * Rename files, POSIX semantics frontend. 3334 */ 3335 /* ARGSUSED */ 3336 int 3337 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3338 { 3339 struct sys___posix_rename_args /* { 3340 syscallarg(const char *) from; 3341 syscallarg(const char *) to; 3342 } */ *uap = v; 3343 3344 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3345 } 3346 3347 /* 3348 * Rename files. Source and destination must either both be directories, 3349 * or both not be directories. If target is a directory, it must be empty. 3350 * If `from' and `to' refer to the same object, the value of the `retain' 3351 * argument is used to determine whether `from' will be 3352 * 3353 * (retain == 0) deleted unless `from' and `to' refer to the same 3354 * object in the file system's name space (BSD). 3355 * (retain == 1) always retained (POSIX). 3356 */ 3357 static int 3358 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3359 { 3360 struct mount *mp = NULL; 3361 struct vnode *tvp, *fvp, *tdvp; 3362 struct nameidata fromnd, tond; 3363 struct proc *p; 3364 int error; 3365 3366 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3367 from, l); 3368 if ((error = namei(&fromnd)) != 0) 3369 return (error); 3370 if (fromnd.ni_dvp != fromnd.ni_vp) 3371 VOP_UNLOCK(fromnd.ni_dvp, 0); 3372 fvp = fromnd.ni_vp; 3373 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3374 if (error != 0) { 3375 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3376 vrele(fromnd.ni_dvp); 3377 vrele(fvp); 3378 if (fromnd.ni_startdir) 3379 vrele(fromnd.ni_startdir); 3380 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3381 return (error); 3382 } 3383 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3384 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3385 if ((error = namei(&tond)) != 0) { 3386 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3387 vrele(fromnd.ni_dvp); 3388 vrele(fvp); 3389 goto out1; 3390 } 3391 tdvp = tond.ni_dvp; 3392 tvp = tond.ni_vp; 3393 3394 if (tvp != NULL) { 3395 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3396 error = ENOTDIR; 3397 goto out; 3398 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3399 error = EISDIR; 3400 goto out; 3401 } 3402 } 3403 3404 if (fvp == tdvp) 3405 error = EINVAL; 3406 3407 /* 3408 * Source and destination refer to the same object. 3409 */ 3410 if (fvp == tvp) { 3411 if (retain) 3412 error = -1; 3413 else if (fromnd.ni_dvp == tdvp && 3414 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3415 !memcmp(fromnd.ni_cnd.cn_nameptr, 3416 tond.ni_cnd.cn_nameptr, 3417 fromnd.ni_cnd.cn_namelen)) 3418 error = -1; 3419 } 3420 3421 #if NVERIEXEC > 0 3422 if (!error) { 3423 pathname_t frompath, topath; 3424 3425 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3426 &frompath); 3427 if (!error) 3428 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3429 &topath); 3430 if (!error) 3431 error = veriexec_renamechk(fvp, pathname_path(frompath), 3432 tvp, pathname_path(topath), l); 3433 3434 pathname_put(frompath); 3435 pathname_put(topath); 3436 } 3437 #endif /* NVERIEXEC > 0 */ 3438 3439 out: 3440 p = l->l_proc; 3441 if (!error) { 3442 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3443 if (fromnd.ni_dvp != tdvp) 3444 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3445 if (tvp) { 3446 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3447 } 3448 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3449 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3450 } else { 3451 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3452 if (tdvp == tvp) 3453 vrele(tdvp); 3454 else 3455 vput(tdvp); 3456 if (tvp) 3457 vput(tvp); 3458 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3459 vrele(fromnd.ni_dvp); 3460 vrele(fvp); 3461 } 3462 vrele(tond.ni_startdir); 3463 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3464 out1: 3465 vn_finished_write(mp, 0); 3466 if (fromnd.ni_startdir) 3467 vrele(fromnd.ni_startdir); 3468 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3469 return (error == -1 ? 0 : error); 3470 } 3471 3472 /* 3473 * Make a directory file. 3474 */ 3475 /* ARGSUSED */ 3476 int 3477 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3478 { 3479 struct sys_mkdir_args /* { 3480 syscallarg(const char *) path; 3481 syscallarg(int) mode; 3482 } */ *uap = v; 3483 struct proc *p = l->l_proc; 3484 struct mount *mp; 3485 struct vnode *vp; 3486 struct vattr vattr; 3487 int error; 3488 struct nameidata nd; 3489 3490 restart: 3491 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3492 SCARG(uap, path), l); 3493 if ((error = namei(&nd)) != 0) 3494 return (error); 3495 vp = nd.ni_vp; 3496 if (vp != NULL) { 3497 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3498 if (nd.ni_dvp == vp) 3499 vrele(nd.ni_dvp); 3500 else 3501 vput(nd.ni_dvp); 3502 vrele(vp); 3503 return (EEXIST); 3504 } 3505 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3507 if (nd.ni_dvp == vp) 3508 vrele(nd.ni_dvp); 3509 else 3510 vput(nd.ni_dvp); 3511 if ((error = vn_start_write(NULL, &mp, 3512 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3513 return (error); 3514 goto restart; 3515 } 3516 VATTR_NULL(&vattr); 3517 vattr.va_type = VDIR; 3518 vattr.va_mode = 3519 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3520 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3521 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3522 if (!error) 3523 vput(nd.ni_vp); 3524 vn_finished_write(mp, 0); 3525 return (error); 3526 } 3527 3528 /* 3529 * Remove a directory file. 3530 */ 3531 /* ARGSUSED */ 3532 int 3533 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3534 { 3535 struct sys_rmdir_args /* { 3536 syscallarg(const char *) path; 3537 } */ *uap = v; 3538 struct mount *mp; 3539 struct vnode *vp; 3540 int error; 3541 struct nameidata nd; 3542 3543 restart: 3544 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3545 SCARG(uap, path), l); 3546 if ((error = namei(&nd)) != 0) 3547 return (error); 3548 vp = nd.ni_vp; 3549 if (vp->v_type != VDIR) { 3550 error = ENOTDIR; 3551 goto out; 3552 } 3553 /* 3554 * No rmdir "." please. 3555 */ 3556 if (nd.ni_dvp == vp) { 3557 error = EINVAL; 3558 goto out; 3559 } 3560 /* 3561 * The root of a mounted filesystem cannot be deleted. 3562 */ 3563 if (vp->v_flag & VROOT) { 3564 error = EBUSY; 3565 goto out; 3566 } 3567 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3568 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3569 if (nd.ni_dvp == vp) 3570 vrele(nd.ni_dvp); 3571 else 3572 vput(nd.ni_dvp); 3573 vput(vp); 3574 if ((error = vn_start_write(NULL, &mp, 3575 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3576 return (error); 3577 goto restart; 3578 } 3579 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3580 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3581 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3582 vn_finished_write(mp, 0); 3583 return (error); 3584 3585 out: 3586 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3587 if (nd.ni_dvp == vp) 3588 vrele(nd.ni_dvp); 3589 else 3590 vput(nd.ni_dvp); 3591 vput(vp); 3592 return (error); 3593 } 3594 3595 /* 3596 * Read a block of directory entries in a file system independent format. 3597 */ 3598 int 3599 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3600 { 3601 struct sys___getdents30_args /* { 3602 syscallarg(int) fd; 3603 syscallarg(char *) buf; 3604 syscallarg(size_t) count; 3605 } */ *uap = v; 3606 struct proc *p = l->l_proc; 3607 struct file *fp; 3608 int error, done; 3609 3610 /* getvnode() will use the descriptor for us */ 3611 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3612 return (error); 3613 if ((fp->f_flag & FREAD) == 0) { 3614 error = EBADF; 3615 goto out; 3616 } 3617 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3618 SCARG(uap, count), &done, l, 0, 0); 3619 #ifdef KTRACE 3620 if (!error && KTRPOINT(p, KTR_GENIO)) { 3621 struct iovec iov; 3622 iov.iov_base = SCARG(uap, buf); 3623 iov.iov_len = done; 3624 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3625 } 3626 #endif 3627 *retval = done; 3628 out: 3629 FILE_UNUSE(fp, l); 3630 return (error); 3631 } 3632 3633 /* 3634 * Set the mode mask for creation of filesystem nodes. 3635 */ 3636 int 3637 sys_umask(struct lwp *l, void *v, register_t *retval) 3638 { 3639 struct sys_umask_args /* { 3640 syscallarg(mode_t) newmask; 3641 } */ *uap = v; 3642 struct proc *p = l->l_proc; 3643 struct cwdinfo *cwdi; 3644 3645 cwdi = p->p_cwdi; 3646 *retval = cwdi->cwdi_cmask; 3647 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3648 return (0); 3649 } 3650 3651 /* 3652 * Void all references to file by ripping underlying filesystem 3653 * away from vnode. 3654 */ 3655 /* ARGSUSED */ 3656 int 3657 sys_revoke(struct lwp *l, void *v, register_t *retval) 3658 { 3659 struct sys_revoke_args /* { 3660 syscallarg(const char *) path; 3661 } */ *uap = v; 3662 struct mount *mp; 3663 struct vnode *vp; 3664 struct vattr vattr; 3665 int error; 3666 struct nameidata nd; 3667 3668 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3669 if ((error = namei(&nd)) != 0) 3670 return (error); 3671 vp = nd.ni_vp; 3672 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3673 goto out; 3674 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3675 (error = kauth_authorize_generic(l->l_cred, 3676 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3677 goto out; 3678 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3679 goto out; 3680 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3681 VOP_REVOKE(vp, REVOKEALL); 3682 vn_finished_write(mp, 0); 3683 out: 3684 vrele(vp); 3685 return (error); 3686 } 3687 3688 /* 3689 * Convert a user file descriptor to a kernel file entry. 3690 */ 3691 int 3692 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3693 { 3694 struct vnode *vp; 3695 struct file *fp; 3696 3697 if ((fp = fd_getfile(fdp, fd)) == NULL) 3698 return (EBADF); 3699 3700 FILE_USE(fp); 3701 3702 if (fp->f_type != DTYPE_VNODE) { 3703 FILE_UNUSE(fp, NULL); 3704 return (EINVAL); 3705 } 3706 3707 vp = (struct vnode *)fp->f_data; 3708 if (vp->v_type == VBAD) { 3709 FILE_UNUSE(fp, NULL); 3710 return (EBADF); 3711 } 3712 3713 *fpp = fp; 3714 return (0); 3715 } 3716
Indexes created Fri Oct 03 02:09:57 GMT 2025