vfs_syscalls.c revision 1.261
1 /* $NetBSD: vfs_syscalls.c,v 1.261 2006/07/26 09:33:57 dogcow 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 1.261 2006/07/26 09:33:57 dogcow 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/sa.h> 65 #include <sys/syscallargs.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #if NVERIEXEC > 0 73 #include <sys/verified_exec.h> 74 #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 #include <nfs/nfsproto.h> 84 #include <nfs/nfs.h> 85 #include <nfs/nfs_var.h> 86 #endif 87 88 #if NFSS > 0 89 #include <dev/fssvar.h> 90 #endif 91 92 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 93 94 static int change_dir(struct nameidata *, struct lwp *); 95 static int change_flags(struct vnode *, u_long, struct lwp *); 96 static int change_mode(struct vnode *, int, struct lwp *l); 97 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 98 static int change_utimes(struct vnode *vp, const struct timeval *, 99 struct lwp *l); 100 static int rename_files(const char *, const char *, struct lwp *, int); 101 102 void checkdirs(struct vnode *); 103 104 int dovfsusermount = 0; 105 106 /* 107 * Virtual File System System Calls 108 */ 109 110 /* 111 * Mount a file system. 112 */ 113 114 #if defined(COMPAT_09) || defined(COMPAT_43) 115 /* 116 * This table is used to maintain compatibility with 4.3BSD 117 * and NetBSD 0.9 mount syscalls. Note, the order is important! 118 * 119 * Do not modify this table. It should only contain filesystems 120 * supported by NetBSD 0.9 and 4.3BSD. 121 */ 122 const char * const mountcompatnames[] = { 123 NULL, /* 0 = MOUNT_NONE */ 124 MOUNT_FFS, /* 1 = MOUNT_UFS */ 125 MOUNT_NFS, /* 2 */ 126 MOUNT_MFS, /* 3 */ 127 MOUNT_MSDOS, /* 4 */ 128 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 129 MOUNT_FDESC, /* 6 */ 130 MOUNT_KERNFS, /* 7 */ 131 NULL, /* 8 = MOUNT_DEVFS */ 132 MOUNT_AFS, /* 9 */ 133 }; 134 const int nmountcompatnames = sizeof(mountcompatnames) / 135 sizeof(mountcompatnames[0]); 136 #endif /* COMPAT_09 || COMPAT_43 */ 137 138 /* ARGSUSED */ 139 int 140 sys_mount(struct lwp *l, void *v, register_t *retval) 141 { 142 struct sys_mount_args /* { 143 syscallarg(const char *) type; 144 syscallarg(const char *) path; 145 syscallarg(int) flags; 146 syscallarg(void *) data; 147 } */ *uap = v; 148 struct vnode *vp; 149 struct mount *mp; 150 int error, flag = 0; 151 char fstypename[MFSNAMELEN]; 152 struct vattr va; 153 struct nameidata nd; 154 struct vfsops *vfs; 155 156 /* 157 * if MNT_GETARGS is specified, it should be only flag. 158 */ 159 160 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 161 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 162 return EINVAL; 163 } 164 165 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 166 (error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 167 &l->l_acflag))) 168 return (error); 169 /* 170 * Get vnode to be covered 171 */ 172 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 173 SCARG(uap, path), l); 174 if ((error = namei(&nd)) != 0) 175 return (error); 176 vp = nd.ni_vp; 177 /* 178 * A lookup in VFS_MOUNT might result in an attempt to 179 * lock this vnode again, so make the lock recursive. 180 */ 181 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 182 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 183 if ((vp->v_flag & VROOT) == 0) { 184 vput(vp); 185 return (EINVAL); 186 } 187 mp = vp->v_mount; 188 flag = mp->mnt_flag; 189 vfs = mp->mnt_op; 190 /* 191 * We only allow the filesystem to be reloaded if it 192 * is currently mounted read-only. 193 */ 194 if ((SCARG(uap, flags) & MNT_RELOAD) && 195 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 196 vput(vp); 197 return (EOPNOTSUPP); /* Needs translation */ 198 } 199 /* 200 * In "highly secure" mode, don't let the caller do anything 201 * but downgrade a filesystem from read-write to read-only. 202 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 203 */ 204 if (securelevel >= 2 && 205 SCARG(uap, flags) != MNT_GETARGS && 206 SCARG(uap, flags) != 207 (mp->mnt_flag | MNT_RDONLY | 208 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 209 vput(vp); 210 return (EPERM); 211 } 212 mp->mnt_flag |= SCARG(uap, flags) & 213 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 214 /* 215 * Only root, or the user that did the original mount is 216 * permitted to update it. 217 */ 218 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 219 mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred) && 220 (error = kauth_authorize_generic(l->l_cred, 221 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 222 vput(vp); 223 return (error); 224 } 225 /* 226 * Do not allow NFS export by non-root users. For non-root 227 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 228 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 229 */ 230 if (kauth_cred_geteuid(l->l_cred) != 0) { 231 if (SCARG(uap, flags) & MNT_EXPORTED) { 232 vput(vp); 233 return (EPERM); 234 } 235 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 236 if (flag & MNT_NOEXEC) 237 SCARG(uap, flags) |= MNT_NOEXEC; 238 } 239 if (vfs_busy(mp, LK_NOWAIT, 0)) { 240 vput(vp); 241 return (EPERM); 242 } 243 goto update; 244 } else { 245 if (securelevel >= 2) { 246 vput(vp); 247 return (EPERM); 248 } 249 } 250 /* 251 * If the user is not root, ensure that they own the directory 252 * onto which we are attempting to mount. 253 */ 254 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 255 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 256 (error = kauth_authorize_generic(l->l_cred, 257 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0)) { 258 vput(vp); 259 return (error); 260 } 261 /* 262 * Do not allow NFS export by non-root users. For non-root users, 263 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 264 * mount point is already MNT_NOEXEC. 265 */ 266 if (kauth_cred_geteuid(l->l_cred) != 0) { 267 if (SCARG(uap, flags) & MNT_EXPORTED) { 268 vput(vp); 269 return (EPERM); 270 } 271 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 272 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 273 SCARG(uap, flags) |= MNT_NOEXEC; 274 } 275 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 276 vput(vp); 277 return (error); 278 } 279 if (vp->v_type != VDIR) { 280 vput(vp); 281 return (ENOTDIR); 282 } 283 error = copyinstr(SCARG(uap, type), 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)SCARG(uap, type); 293 if (fsindex >= nmountcompatnames || 294 mountcompatnames[fsindex] == NULL) { 295 vput(vp); 296 return (ENODEV); 297 } 298 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 299 #else 300 vput(vp); 301 return (error); 302 #endif 303 } 304 #ifdef COMPAT_10 305 /* Accept `ufs' as an alias for `ffs'. */ 306 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 307 strncpy(fstypename, "ffs", MFSNAMELEN); 308 #endif 309 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 310 vput(vp); 311 return (ENODEV); 312 } 313 if (vp->v_mountedhere != NULL) { 314 vput(vp); 315 return (EBUSY); 316 } 317 318 /* 319 * Allocate and initialize the file system. 320 */ 321 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 322 M_MOUNT, M_WAITOK); 323 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 324 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 325 simple_lock_init(&mp->mnt_slock); 326 (void)vfs_busy(mp, LK_NOWAIT, 0); 327 mp->mnt_op = vfs; 328 vfs->vfs_refcount++; 329 mp->mnt_vnodecovered = vp; 330 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 331 mp->mnt_unmounter = NULL; 332 mp->mnt_leaf = mp; 333 334 /* 335 * The underlying file system may refuse the mount for 336 * various reasons. Allow the user to force it to happen. 337 */ 338 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 339 update: 340 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 341 /* 342 * Set the mount level flags. 343 */ 344 if (SCARG(uap, flags) & MNT_RDONLY) 345 mp->mnt_flag |= MNT_RDONLY; 346 else if (mp->mnt_flag & MNT_RDONLY) 347 mp->mnt_iflag |= IMNT_WANTRDWR; 348 mp->mnt_flag &= 349 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 350 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 351 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 352 mp->mnt_flag |= SCARG(uap, flags) & 353 (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); 357 } 358 /* 359 * Mount the filesystem. 360 */ 361 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 362 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 363 #if defined(COMPAT_30) && defined(NFSSERVER) 364 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 365 int error2; 366 367 /* Update failed; let's try and see if it was an 368 * export request. */ 369 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 370 SCARG(uap, data), l); 371 372 /* Only update error code if the export request was 373 * understood but some problem occurred while 374 * processing it. */ 375 if (error2 != EJUSTRETURN) 376 error = error2; 377 } 378 #endif 379 if (mp->mnt_iflag & IMNT_WANTRDWR) 380 mp->mnt_flag &= ~MNT_RDONLY; 381 if (error) 382 mp->mnt_flag = flag; 383 mp->mnt_flag &=~ 384 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 385 mp->mnt_iflag &=~ IMNT_WANTRDWR; 386 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 387 if (mp->mnt_syncer == NULL) 388 error = vfs_allocate_syncvnode(mp); 389 } else { 390 if (mp->mnt_syncer != NULL) 391 vfs_deallocate_syncvnode(mp); 392 } 393 vfs_unbusy(mp); 394 VOP_UNLOCK(vp, 0); 395 vrele(vp); 396 return (error); 397 } 398 /* 399 * Put the new filesystem on the mount list after root. 400 */ 401 cache_purge(vp); 402 if (!error) { 403 mp->mnt_flag &=~ 404 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 405 mp->mnt_iflag &=~ IMNT_WANTRDWR; 406 vp->v_mountedhere = mp; 407 simple_lock(&mountlist_slock); 408 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 409 simple_unlock(&mountlist_slock); 410 checkdirs(vp); 411 VOP_UNLOCK(vp, 0); 412 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 413 error = vfs_allocate_syncvnode(mp); 414 vfs_unbusy(mp); 415 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 416 if ((error = VFS_START(mp, 0, l))) 417 vrele(vp); 418 } else { 419 vp->v_mountedhere = (struct mount *)0; 420 vfs->vfs_refcount--; 421 vfs_unbusy(mp); 422 free(mp, M_MOUNT); 423 vput(vp); 424 } 425 return (error); 426 } 427 428 /* 429 * Scan all active processes to see if any of them have a current 430 * or root directory onto which the new filesystem has just been 431 * mounted. If so, replace them with the new mount point. 432 */ 433 void 434 checkdirs(struct vnode *olddp) 435 { 436 struct cwdinfo *cwdi; 437 struct vnode *newdp; 438 struct proc *p; 439 440 if (olddp->v_usecount == 1) 441 return; 442 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 443 panic("mount: lost mount"); 444 proclist_lock_read(); 445 PROCLIST_FOREACH(p, &allproc) { 446 cwdi = p->p_cwdi; 447 if (!cwdi) 448 continue; 449 if (cwdi->cwdi_cdir == olddp) { 450 vrele(cwdi->cwdi_cdir); 451 VREF(newdp); 452 cwdi->cwdi_cdir = newdp; 453 } 454 if (cwdi->cwdi_rdir == olddp) { 455 vrele(cwdi->cwdi_rdir); 456 VREF(newdp); 457 cwdi->cwdi_rdir = newdp; 458 } 459 } 460 proclist_unlock_read(); 461 if (rootvnode == olddp) { 462 vrele(rootvnode); 463 VREF(newdp); 464 rootvnode = newdp; 465 } 466 vput(newdp); 467 } 468 469 /* 470 * Unmount a file system. 471 * 472 * Note: unmount takes a path to the vnode mounted on as argument, 473 * not special file (as before). 474 */ 475 /* ARGSUSED */ 476 int 477 sys_unmount(struct lwp *l, void *v, register_t *retval) 478 { 479 struct sys_unmount_args /* { 480 syscallarg(const char *) path; 481 syscallarg(int) flags; 482 } */ *uap = v; 483 struct vnode *vp; 484 struct mount *mp; 485 int error; 486 struct nameidata nd; 487 488 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 489 SCARG(uap, path), l); 490 if ((error = namei(&nd)) != 0) 491 return (error); 492 vp = nd.ni_vp; 493 mp = vp->v_mount; 494 495 /* 496 * Only root, or the user that did the original mount is 497 * permitted to unmount this filesystem. 498 */ 499 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred)) && 500 (error = kauth_authorize_generic(l->l_cred, 501 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 502 vput(vp); 503 return (error); 504 } 505 506 /* 507 * Don't allow unmounting the root file system. 508 */ 509 if (mp->mnt_flag & MNT_ROOTFS) { 510 vput(vp); 511 return (EINVAL); 512 } 513 514 /* 515 * Must be the root of the filesystem 516 */ 517 if ((vp->v_flag & VROOT) == 0) { 518 vput(vp); 519 return (EINVAL); 520 } 521 vput(vp); 522 523 /* 524 * XXX Freeze syncer. Must do this before locking the 525 * mount point. See dounmount() for details. 526 */ 527 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 528 529 if (vfs_busy(mp, 0, 0)) { 530 lockmgr(&syncer_lock, LK_RELEASE, NULL); 531 return (EBUSY); 532 } 533 534 return (dounmount(mp, SCARG(uap, flags), l)); 535 } 536 537 /* 538 * Do the actual file system unmount. File system is assumed to have been 539 * marked busy by the caller. 540 */ 541 int 542 dounmount(struct mount *mp, int flags, struct lwp *l) 543 { 544 struct vnode *coveredvp; 545 int error; 546 int async; 547 int used_syncer; 548 549 #if NVERIEXEC > 0 550 if (!doing_shutdown) { 551 if (veriexec_strict >= VERIEXEC_LOCKDOWN) { 552 printf("Veriexec: Lockdown mode, preventing unmount of" 553 " \"%s\". (uid=%u)\n", mp->mnt_stat.f_mntonname, 554 kauth_cred_getuid(l->l_cred)); 555 return (EPERM); 556 } 557 558 if (veriexec_strict == VERIEXEC_IPS) { 559 struct veriexec_table_entry *vte; 560 561 /* Check if we have fingerprints on mount. */ 562 vte = fileassoc_tabledata_lookup(mp, veriexec_hook); 563 if ((vte != NULL) && (vte->vte_count > 0)) { 564 printf("Veriexec: IPS mode, preventing unmount" 565 " of \"%s\" with monitored files. " 566 "(uid=%u)\n", mp->mnt_stat.f_mntonname, 567 kauth_cred_getuid(l->l_cred)); 568 return (EPERM); 569 } 570 } 571 } 572 #endif /* NVERIEXEC > 0 */ 573 574 #ifdef FILEASSOC 575 (void)fileassoc_table_delete(mp); 576 #endif /* FILEASSOC */ 577 578 simple_lock(&mountlist_slock); 579 vfs_unbusy(mp); 580 used_syncer = (mp->mnt_syncer != NULL); 581 582 /* 583 * XXX Syncer must be frozen when we get here. This should really 584 * be done on a per-mountpoint basis, but especially the softdep 585 * code possibly called from the syncer doesn't exactly work on a 586 * per-mountpoint basis, so the softdep code would become a maze 587 * of vfs_busy() calls. 588 * 589 * The caller of dounmount() must acquire syncer_lock because 590 * the syncer itself acquires locks in syncer_lock -> vfs_busy 591 * order, and we must preserve that order to avoid deadlock. 592 * 593 * So, if the file system did not use the syncer, now is 594 * the time to release the syncer_lock. 595 */ 596 if (used_syncer == 0) 597 lockmgr(&syncer_lock, LK_RELEASE, NULL); 598 599 mp->mnt_iflag |= IMNT_UNMOUNT; 600 mp->mnt_unmounter = l; 601 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 602 vn_start_write(NULL, &mp, V_WAIT); 603 604 async = mp->mnt_flag & MNT_ASYNC; 605 mp->mnt_flag &= ~MNT_ASYNC; 606 cache_purgevfs(mp); /* remove cache entries for this file sys */ 607 if (mp->mnt_syncer != NULL) 608 vfs_deallocate_syncvnode(mp); 609 error = 0; 610 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 611 #if NFSS > 0 612 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 613 #endif 614 if (error == 0) 615 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 616 } 617 if (error == 0 || (flags & MNT_FORCE)) 618 error = VFS_UNMOUNT(mp, flags, l); 619 vn_finished_write(mp, 0); 620 simple_lock(&mountlist_slock); 621 if (error) { 622 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 623 (void) vfs_allocate_syncvnode(mp); 624 mp->mnt_iflag &= ~IMNT_UNMOUNT; 625 mp->mnt_unmounter = NULL; 626 mp->mnt_flag |= async; 627 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 628 &mountlist_slock); 629 if (used_syncer) 630 lockmgr(&syncer_lock, LK_RELEASE, NULL); 631 simple_lock(&mp->mnt_slock); 632 while (mp->mnt_wcnt > 0) { 633 wakeup(mp); 634 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 635 0, &mp->mnt_slock); 636 } 637 simple_unlock(&mp->mnt_slock); 638 return (error); 639 } 640 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 641 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 642 coveredvp->v_mountedhere = NULL; 643 vrele(coveredvp); 644 } 645 mp->mnt_op->vfs_refcount--; 646 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 647 panic("unmount: dangling vnode"); 648 mp->mnt_iflag |= IMNT_GONE; 649 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 650 if (used_syncer) 651 lockmgr(&syncer_lock, LK_RELEASE, NULL); 652 simple_lock(&mp->mnt_slock); 653 while (mp->mnt_wcnt > 0) { 654 wakeup(mp); 655 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 656 } 657 simple_unlock(&mp->mnt_slock); 658 vfs_hooks_unmount(mp); 659 free(mp, M_MOUNT); 660 return (0); 661 } 662 663 /* 664 * Sync each mounted filesystem. 665 */ 666 #ifdef DEBUG 667 int syncprt = 0; 668 struct ctldebug debug0 = { "syncprt", &syncprt }; 669 #endif 670 671 /* ARGSUSED */ 672 int 673 sys_sync(struct lwp *l, void *v, register_t *retval) 674 { 675 struct mount *mp, *nmp; 676 int asyncflag; 677 678 if (l == NULL) 679 l = &lwp0; 680 681 simple_lock(&mountlist_slock); 682 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 683 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 684 nmp = mp->mnt_list.cqe_prev; 685 continue; 686 } 687 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 688 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 689 asyncflag = mp->mnt_flag & MNT_ASYNC; 690 mp->mnt_flag &= ~MNT_ASYNC; 691 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 692 if (asyncflag) 693 mp->mnt_flag |= MNT_ASYNC; 694 vn_finished_write(mp, 0); 695 } 696 simple_lock(&mountlist_slock); 697 nmp = mp->mnt_list.cqe_prev; 698 vfs_unbusy(mp); 699 700 } 701 simple_unlock(&mountlist_slock); 702 #ifdef DEBUG 703 if (syncprt) 704 vfs_bufstats(); 705 #endif /* DEBUG */ 706 return (0); 707 } 708 709 /* 710 * Change filesystem quotas. 711 */ 712 /* ARGSUSED */ 713 int 714 sys_quotactl(struct lwp *l, void *v, register_t *retval) 715 { 716 struct sys_quotactl_args /* { 717 syscallarg(const char *) path; 718 syscallarg(int) cmd; 719 syscallarg(int) uid; 720 syscallarg(caddr_t) arg; 721 } */ *uap = v; 722 struct mount *mp; 723 int error; 724 struct nameidata nd; 725 726 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 727 if ((error = namei(&nd)) != 0) 728 return (error); 729 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 730 vrele(nd.ni_vp); 731 if (error) 732 return (error); 733 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 734 SCARG(uap, arg), l); 735 vn_finished_write(mp, 0); 736 return (error); 737 } 738 739 int 740 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 741 int root) 742 { 743 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 744 int error = 0; 745 746 /* 747 * If MNT_NOWAIT or MNT_LAZY is specified, do not 748 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 749 * overrides MNT_NOWAIT. 750 */ 751 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 752 (flags != MNT_WAIT && flags != 0)) { 753 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 754 goto done; 755 } 756 757 /* Get the filesystem stats now */ 758 memset(sp, 0, sizeof(*sp)); 759 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 760 return error; 761 } 762 763 if (cwdi->cwdi_rdir == NULL) 764 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 765 done: 766 if (cwdi->cwdi_rdir != NULL) { 767 size_t len; 768 char *bp; 769 char *path = PNBUF_GET(); 770 if (!path) 771 return ENOMEM; 772 773 bp = path + MAXPATHLEN; 774 *--bp = '\0'; 775 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 776 MAXPATHLEN / 2, 0, l); 777 if (error) { 778 PNBUF_PUT(path); 779 return error; 780 } 781 len = strlen(bp); 782 /* 783 * for mount points that are below our root, we can see 784 * them, so we fix up the pathname and return them. The 785 * rest we cannot see, so we don't allow viewing the 786 * data. 787 */ 788 if (strncmp(bp, sp->f_mntonname, len) == 0) { 789 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 790 sizeof(sp->f_mntonname)); 791 if (sp->f_mntonname[0] == '\0') 792 (void)strlcpy(sp->f_mntonname, "/", 793 sizeof(sp->f_mntonname)); 794 } else { 795 if (root) 796 (void)strlcpy(sp->f_mntonname, "/", 797 sizeof(sp->f_mntonname)); 798 else 799 error = EPERM; 800 } 801 PNBUF_PUT(path); 802 } 803 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 804 return error; 805 } 806 807 /* 808 * Get filesystem statistics. 809 */ 810 /* ARGSUSED */ 811 int 812 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 813 { 814 struct sys_statvfs1_args /* { 815 syscallarg(const char *) path; 816 syscallarg(struct statvfs *) buf; 817 syscallarg(int) flags; 818 } */ *uap = v; 819 struct mount *mp; 820 struct statvfs *sb; 821 int error; 822 struct nameidata nd; 823 824 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 825 if ((error = namei(&nd)) != 0) 826 return error; 827 mp = nd.ni_vp->v_mount; 828 vrele(nd.ni_vp); 829 sb = STATVFSBUF_GET(); 830 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 831 if (error == 0) { 832 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 833 } 834 STATVFSBUF_PUT(sb); 835 return error; 836 } 837 838 /* 839 * Get filesystem statistics. 840 */ 841 /* ARGSUSED */ 842 int 843 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 844 { 845 struct sys_fstatvfs1_args /* { 846 syscallarg(int) fd; 847 syscallarg(struct statvfs *) buf; 848 syscallarg(int) flags; 849 } */ *uap = v; 850 struct proc *p = l->l_proc; 851 struct file *fp; 852 struct mount *mp; 853 struct statvfs *sb; 854 int error; 855 856 /* getvnode() will use the descriptor for us */ 857 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 858 return (error); 859 mp = ((struct vnode *)fp->f_data)->v_mount; 860 sb = STATVFSBUF_GET(); 861 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 862 goto out; 863 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 864 out: 865 FILE_UNUSE(fp, l); 866 STATVFSBUF_PUT(sb); 867 return error; 868 } 869 870 871 /* 872 * Get statistics on all filesystems. 873 */ 874 int 875 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 876 { 877 struct sys_getvfsstat_args /* { 878 syscallarg(struct statvfs *) buf; 879 syscallarg(size_t) bufsize; 880 syscallarg(int) flags; 881 } */ *uap = v; 882 int root = 0; 883 struct proc *p = l->l_proc; 884 struct mount *mp, *nmp; 885 struct statvfs *sb; 886 struct statvfs *sfsp; 887 size_t count, maxcount; 888 int error = 0; 889 890 sb = STATVFSBUF_GET(); 891 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 892 sfsp = SCARG(uap, buf); 893 simple_lock(&mountlist_slock); 894 count = 0; 895 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 896 mp = nmp) { 897 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 898 nmp = CIRCLEQ_NEXT(mp, mnt_list); 899 continue; 900 } 901 if (sfsp && count < maxcount) { 902 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 903 if (error) { 904 simple_lock(&mountlist_slock); 905 nmp = CIRCLEQ_NEXT(mp, mnt_list); 906 vfs_unbusy(mp); 907 continue; 908 } 909 error = copyout(sb, sfsp, sizeof(*sfsp)); 910 if (error) { 911 vfs_unbusy(mp); 912 goto out; 913 } 914 sfsp++; 915 root |= strcmp(sb->f_mntonname, "/") == 0; 916 } 917 count++; 918 simple_lock(&mountlist_slock); 919 nmp = CIRCLEQ_NEXT(mp, mnt_list); 920 vfs_unbusy(mp); 921 } 922 simple_unlock(&mountlist_slock); 923 if (root == 0 && p->p_cwdi->cwdi_rdir) { 924 /* 925 * fake a root entry 926 */ 927 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 928 SCARG(uap, flags), 1)) != 0) 929 goto out; 930 if (sfsp) 931 error = copyout(sb, sfsp, sizeof(*sfsp)); 932 count++; 933 } 934 if (sfsp && count > maxcount) 935 *retval = maxcount; 936 else 937 *retval = count; 938 out: 939 STATVFSBUF_PUT(sb); 940 return error; 941 } 942 943 /* 944 * Change current working directory to a given file descriptor. 945 */ 946 /* ARGSUSED */ 947 int 948 sys_fchdir(struct lwp *l, void *v, register_t *retval) 949 { 950 struct sys_fchdir_args /* { 951 syscallarg(int) fd; 952 } */ *uap = v; 953 struct proc *p = l->l_proc; 954 struct filedesc *fdp = p->p_fd; 955 struct cwdinfo *cwdi = p->p_cwdi; 956 struct vnode *vp, *tdp; 957 struct mount *mp; 958 struct file *fp; 959 int error; 960 961 /* getvnode() will use the descriptor for us */ 962 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 963 return (error); 964 vp = (struct vnode *)fp->f_data; 965 966 VREF(vp); 967 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 968 if (vp->v_type != VDIR) 969 error = ENOTDIR; 970 else 971 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 972 while (!error && (mp = vp->v_mountedhere) != NULL) { 973 if (vfs_busy(mp, 0, 0)) 974 continue; 975 error = VFS_ROOT(mp, &tdp); 976 vfs_unbusy(mp); 977 if (error) 978 break; 979 vput(vp); 980 vp = tdp; 981 } 982 if (error) { 983 vput(vp); 984 goto out; 985 } 986 VOP_UNLOCK(vp, 0); 987 988 /* 989 * Disallow changing to a directory not under the process's 990 * current root directory (if there is one). 991 */ 992 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 993 vrele(vp); 994 error = EPERM; /* operation not permitted */ 995 goto out; 996 } 997 998 vrele(cwdi->cwdi_cdir); 999 cwdi->cwdi_cdir = vp; 1000 out: 1001 FILE_UNUSE(fp, l); 1002 return (error); 1003 } 1004 1005 /* 1006 * Change this process's notion of the root directory to a given file 1007 * descriptor. 1008 */ 1009 int 1010 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1011 { 1012 struct sys_fchroot_args *uap = v; 1013 struct proc *p = l->l_proc; 1014 struct filedesc *fdp = p->p_fd; 1015 struct cwdinfo *cwdi = p->p_cwdi; 1016 struct vnode *vp; 1017 struct file *fp; 1018 int error; 1019 1020 if ((error = kauth_authorize_generic(l->l_cred, 1021 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1022 return error; 1023 /* getvnode() will use the descriptor for us */ 1024 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1025 return error; 1026 vp = (struct vnode *) fp->f_data; 1027 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1028 if (vp->v_type != VDIR) 1029 error = ENOTDIR; 1030 else 1031 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1032 VOP_UNLOCK(vp, 0); 1033 if (error) 1034 goto out; 1035 VREF(vp); 1036 1037 /* 1038 * Prevent escaping from chroot by putting the root under 1039 * the working directory. Silently chdir to / if we aren't 1040 * already there. 1041 */ 1042 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1043 /* 1044 * XXX would be more failsafe to change directory to a 1045 * deadfs node here instead 1046 */ 1047 vrele(cwdi->cwdi_cdir); 1048 VREF(vp); 1049 cwdi->cwdi_cdir = vp; 1050 } 1051 1052 if (cwdi->cwdi_rdir != NULL) 1053 vrele(cwdi->cwdi_rdir); 1054 cwdi->cwdi_rdir = vp; 1055 out: 1056 FILE_UNUSE(fp, l); 1057 return (error); 1058 } 1059 1060 /* 1061 * Change current working directory (``.''). 1062 */ 1063 /* ARGSUSED */ 1064 int 1065 sys_chdir(struct lwp *l, void *v, register_t *retval) 1066 { 1067 struct sys_chdir_args /* { 1068 syscallarg(const char *) path; 1069 } */ *uap = v; 1070 struct proc *p = l->l_proc; 1071 struct cwdinfo *cwdi = p->p_cwdi; 1072 int error; 1073 struct nameidata nd; 1074 1075 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1076 SCARG(uap, path), l); 1077 if ((error = change_dir(&nd, l)) != 0) 1078 return (error); 1079 vrele(cwdi->cwdi_cdir); 1080 cwdi->cwdi_cdir = nd.ni_vp; 1081 return (0); 1082 } 1083 1084 /* 1085 * Change notion of root (``/'') directory. 1086 */ 1087 /* ARGSUSED */ 1088 int 1089 sys_chroot(struct lwp *l, void *v, register_t *retval) 1090 { 1091 struct sys_chroot_args /* { 1092 syscallarg(const char *) path; 1093 } */ *uap = v; 1094 struct proc *p = l->l_proc; 1095 struct cwdinfo *cwdi = p->p_cwdi; 1096 struct vnode *vp; 1097 int error; 1098 struct nameidata nd; 1099 1100 if ((error = kauth_authorize_generic(l->l_cred, 1101 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1102 return (error); 1103 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1104 SCARG(uap, path), l); 1105 if ((error = change_dir(&nd, l)) != 0) 1106 return (error); 1107 if (cwdi->cwdi_rdir != NULL) 1108 vrele(cwdi->cwdi_rdir); 1109 vp = nd.ni_vp; 1110 cwdi->cwdi_rdir = vp; 1111 1112 /* 1113 * Prevent escaping from chroot by putting the root under 1114 * the working directory. Silently chdir to / if we aren't 1115 * already there. 1116 */ 1117 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1118 /* 1119 * XXX would be more failsafe to change directory to a 1120 * deadfs node here instead 1121 */ 1122 vrele(cwdi->cwdi_cdir); 1123 VREF(vp); 1124 cwdi->cwdi_cdir = vp; 1125 } 1126 1127 return (0); 1128 } 1129 1130 /* 1131 * Common routine for chroot and chdir. 1132 */ 1133 static int 1134 change_dir(struct nameidata *ndp, struct lwp *l) 1135 { 1136 struct vnode *vp; 1137 int error; 1138 1139 if ((error = namei(ndp)) != 0) 1140 return (error); 1141 vp = ndp->ni_vp; 1142 if (vp->v_type != VDIR) 1143 error = ENOTDIR; 1144 else 1145 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1146 1147 if (error) 1148 vput(vp); 1149 else 1150 VOP_UNLOCK(vp, 0); 1151 return (error); 1152 } 1153 1154 /* 1155 * Check permissions, allocate an open file structure, 1156 * and call the device open routine if any. 1157 */ 1158 int 1159 sys_open(struct lwp *l, void *v, register_t *retval) 1160 { 1161 struct sys_open_args /* { 1162 syscallarg(const char *) path; 1163 syscallarg(int) flags; 1164 syscallarg(int) mode; 1165 } */ *uap = v; 1166 struct proc *p = l->l_proc; 1167 struct cwdinfo *cwdi = p->p_cwdi; 1168 struct filedesc *fdp = p->p_fd; 1169 struct file *fp; 1170 struct vnode *vp; 1171 int flags, cmode; 1172 int type, indx, error; 1173 struct flock lf; 1174 struct nameidata nd; 1175 1176 flags = FFLAGS(SCARG(uap, flags)); 1177 if ((flags & (FREAD | FWRITE)) == 0) 1178 return (EINVAL); 1179 /* falloc() will use the file descriptor for us */ 1180 if ((error = falloc(l, &fp, &indx)) != 0) 1181 return (error); 1182 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1183 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1184 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1185 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1186 FILE_UNUSE(fp, l); 1187 fdp->fd_ofiles[indx] = NULL; 1188 ffree(fp); 1189 if ((error == EDUPFD || error == EMOVEFD) && 1190 l->l_dupfd >= 0 && /* XXX from fdopen */ 1191 (error = 1192 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1193 *retval = indx; 1194 return (0); 1195 } 1196 if (error == ERESTART) 1197 error = EINTR; 1198 fdremove(fdp, indx); 1199 return (error); 1200 } 1201 l->l_dupfd = 0; 1202 vp = nd.ni_vp; 1203 fp->f_flag = flags & FMASK; 1204 fp->f_type = DTYPE_VNODE; 1205 fp->f_ops = &vnops; 1206 fp->f_data = vp; 1207 if (flags & (O_EXLOCK | O_SHLOCK)) { 1208 lf.l_whence = SEEK_SET; 1209 lf.l_start = 0; 1210 lf.l_len = 0; 1211 if (flags & O_EXLOCK) 1212 lf.l_type = F_WRLCK; 1213 else 1214 lf.l_type = F_RDLCK; 1215 type = F_FLOCK; 1216 if ((flags & FNONBLOCK) == 0) 1217 type |= F_WAIT; 1218 VOP_UNLOCK(vp, 0); 1219 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1220 if (error) { 1221 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1222 FILE_UNUSE(fp, l); 1223 ffree(fp); 1224 fdremove(fdp, indx); 1225 return (error); 1226 } 1227 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1228 fp->f_flag |= FHASLOCK; 1229 } 1230 VOP_UNLOCK(vp, 0); 1231 *retval = indx; 1232 FILE_SET_MATURE(fp); 1233 FILE_UNUSE(fp, l); 1234 return (0); 1235 } 1236 1237 static void 1238 vfs__fhfree(fhandle_t *fhp) 1239 { 1240 size_t fhsize; 1241 1242 if (fhp == NULL) { 1243 return; 1244 } 1245 fhsize = FHANDLE_SIZE(fhp); 1246 kmem_free(fhp, fhsize); 1247 } 1248 1249 /* 1250 * vfs_composefh: compose a filehandle. 1251 */ 1252 1253 int 1254 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1255 { 1256 struct mount *mp; 1257 struct fid *fidp; 1258 int error; 1259 size_t needfhsize; 1260 size_t fidsize; 1261 1262 mp = vp->v_mount; 1263 if (mp->mnt_op->vfs_vptofh == NULL) { 1264 return EOPNOTSUPP; 1265 } 1266 fidp = NULL; 1267 if (*fh_size < FHANDLE_SIZE_MIN) { 1268 fidsize = 0; 1269 } else { 1270 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1271 if (fhp != NULL) { 1272 memset(fhp, 0, *fh_size); 1273 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1274 fidp = &fhp->fh_fid; 1275 } 1276 } 1277 error = VFS_VPTOFH(vp, fidp, &fidsize); 1278 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1279 if (error == 0 && *fh_size < needfhsize) { 1280 error = E2BIG; 1281 } 1282 *fh_size = needfhsize; 1283 return error; 1284 } 1285 1286 int 1287 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1288 { 1289 struct mount *mp; 1290 fhandle_t *fhp; 1291 size_t fhsize; 1292 size_t fidsize; 1293 int error; 1294 1295 *fhpp = NULL; 1296 mp = vp->v_mount; 1297 if (mp->mnt_op->vfs_vptofh == NULL) { 1298 error = EOPNOTSUPP; 1299 goto out; 1300 } 1301 fidsize = 0; 1302 error = VFS_VPTOFH(vp, NULL, &fidsize); 1303 KASSERT(error != 0); 1304 if (error != E2BIG) { 1305 goto out; 1306 } 1307 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1308 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1309 if (fhp == NULL) { 1310 error = ENOMEM; 1311 goto out; 1312 } 1313 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1314 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1315 if (error == 0) { 1316 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1317 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1318 *fhpp = fhp; 1319 } else { 1320 kmem_free(fhp, fhsize); 1321 } 1322 out: 1323 return error; 1324 } 1325 1326 void 1327 vfs_composefh_free(fhandle_t *fhp) 1328 { 1329 1330 vfs__fhfree(fhp); 1331 } 1332 1333 /* 1334 * vfs_fhtovp: lookup a vnode by a filehandle. 1335 */ 1336 1337 int 1338 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1339 { 1340 struct mount *mp; 1341 int error; 1342 1343 *vpp = NULL; 1344 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1345 if (mp == NULL) { 1346 error = ESTALE; 1347 goto out; 1348 } 1349 if (mp->mnt_op->vfs_fhtovp == NULL) { 1350 error = EOPNOTSUPP; 1351 goto out; 1352 } 1353 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1354 out: 1355 return error; 1356 } 1357 1358 /* 1359 * vfs_copyinfh: copyin a filehandle. 1360 */ 1361 1362 int 1363 vfs_copyinfh_alloc(const void *ufhp, fhandle_t **fhpp) 1364 { 1365 fhandle_t *fhp; 1366 fhandle_t tempfh; 1367 size_t fhsize; 1368 int error; 1369 1370 *fhpp = NULL; 1371 error = copyin(ufhp, &tempfh, sizeof(tempfh)); 1372 if (error) { 1373 return error; 1374 } 1375 fhsize = FHANDLE_SIZE(&tempfh); 1376 if (fhsize > FHANDLE_SIZE_MAX) { 1377 return EINVAL; 1378 } 1379 fhp = kmem_alloc(fhsize, KM_SLEEP); 1380 if (fhp == NULL) { 1381 return ENOMEM; 1382 } 1383 error = copyin(ufhp, fhp, fhsize); 1384 if (error == 0) { 1385 *fhpp = fhp; 1386 } else { 1387 kmem_free(fhp, fhsize); 1388 } 1389 return error; 1390 } 1391 1392 void 1393 vfs_copyinfh_free(fhandle_t *fhp) 1394 { 1395 1396 vfs__fhfree(fhp); 1397 } 1398 1399 /* 1400 * Get file handle system call 1401 */ 1402 int 1403 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1404 { 1405 struct sys___getfh30_args /* { 1406 syscallarg(char *) fname; 1407 syscallarg(fhandle_t *) fhp; 1408 syscallarg(size_t *) fh_size; 1409 } */ *uap = v; 1410 struct vnode *vp; 1411 fhandle_t *fh; 1412 int error; 1413 struct nameidata nd; 1414 size_t sz; 1415 size_t usz; 1416 1417 /* 1418 * Must be super user 1419 */ 1420 error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 1421 &l->l_acflag); 1422 if (error) 1423 return (error); 1424 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1425 SCARG(uap, fname), l); 1426 error = namei(&nd); 1427 if (error) 1428 return (error); 1429 vp = nd.ni_vp; 1430 error = vfs_composefh_alloc(vp, &fh); 1431 vput(vp); 1432 if (error != 0) { 1433 goto out; 1434 } 1435 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1436 if (error != 0) { 1437 goto out; 1438 } 1439 sz = FHANDLE_SIZE(fh); 1440 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1441 if (error != 0) { 1442 goto out; 1443 } 1444 if (usz >= sz) { 1445 error = copyout(fh, SCARG(uap, fhp), sz); 1446 } else { 1447 error = E2BIG; 1448 } 1449 out: 1450 vfs_composefh_free(fh); 1451 return (error); 1452 } 1453 1454 /* 1455 * Open a file given a file handle. 1456 * 1457 * Check permissions, allocate an open file structure, 1458 * and call the device open routine if any. 1459 */ 1460 int 1461 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1462 { 1463 struct sys_fhopen_args /* { 1464 syscallarg(const fhandle_t *) fhp; 1465 syscallarg(int) flags; 1466 } */ *uap = v; 1467 struct filedesc *fdp = l->l_proc->p_fd; 1468 struct file *fp; 1469 struct vnode *vp = NULL; 1470 struct mount *mp; 1471 kauth_cred_t cred = l->l_cred; 1472 int flags; 1473 struct file *nfp; 1474 int type, indx, error=0; 1475 struct flock lf; 1476 struct vattr va; 1477 fhandle_t *fh; 1478 1479 /* 1480 * Must be super user 1481 */ 1482 if ((error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 1483 &l->l_acflag))) 1484 return (error); 1485 1486 flags = FFLAGS(SCARG(uap, flags)); 1487 if ((flags & (FREAD | FWRITE)) == 0) 1488 return (EINVAL); 1489 if ((flags & O_CREAT)) 1490 return (EINVAL); 1491 /* falloc() will use the file descriptor for us */ 1492 if ((error = falloc(l, &nfp, &indx)) != 0) 1493 return (error); 1494 fp = nfp; 1495 error = vfs_copyinfh_alloc(SCARG(uap, fhp), &fh); 1496 if (error != 0) { 1497 goto bad; 1498 } 1499 error = vfs_fhtovp(fh, &vp); 1500 if (error != 0) { 1501 goto bad; 1502 } 1503 1504 /* Now do an effective vn_open */ 1505 1506 if (vp->v_type == VSOCK) { 1507 error = EOPNOTSUPP; 1508 goto bad; 1509 } 1510 if (flags & FREAD) { 1511 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1512 goto bad; 1513 } 1514 if (flags & (FWRITE | O_TRUNC)) { 1515 if (vp->v_type == VDIR) { 1516 error = EISDIR; 1517 goto bad; 1518 } 1519 if ((error = vn_writechk(vp)) != 0 || 1520 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1521 goto bad; 1522 } 1523 if (flags & O_TRUNC) { 1524 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1525 goto bad; 1526 VOP_UNLOCK(vp, 0); /* XXX */ 1527 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1528 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1529 VATTR_NULL(&va); 1530 va.va_size = 0; 1531 error = VOP_SETATTR(vp, &va, cred, l); 1532 vn_finished_write(mp, 0); 1533 if (error) 1534 goto bad; 1535 } 1536 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1537 goto bad; 1538 if (vp->v_type == VREG && 1539 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1540 error = EIO; 1541 goto bad; 1542 } 1543 if (flags & FWRITE) 1544 vp->v_writecount++; 1545 1546 /* done with modified vn_open, now finish what sys_open does. */ 1547 1548 fp->f_flag = flags & FMASK; 1549 fp->f_type = DTYPE_VNODE; 1550 fp->f_ops = &vnops; 1551 fp->f_data = vp; 1552 if (flags & (O_EXLOCK | O_SHLOCK)) { 1553 lf.l_whence = SEEK_SET; 1554 lf.l_start = 0; 1555 lf.l_len = 0; 1556 if (flags & O_EXLOCK) 1557 lf.l_type = F_WRLCK; 1558 else 1559 lf.l_type = F_RDLCK; 1560 type = F_FLOCK; 1561 if ((flags & FNONBLOCK) == 0) 1562 type |= F_WAIT; 1563 VOP_UNLOCK(vp, 0); 1564 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1565 if (error) { 1566 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1567 FILE_UNUSE(fp, l); 1568 ffree(fp); 1569 fdremove(fdp, indx); 1570 return (error); 1571 } 1572 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1573 fp->f_flag |= FHASLOCK; 1574 } 1575 VOP_UNLOCK(vp, 0); 1576 *retval = indx; 1577 FILE_SET_MATURE(fp); 1578 FILE_UNUSE(fp, l); 1579 vfs_copyinfh_free(fh); 1580 return (0); 1581 1582 bad: 1583 FILE_UNUSE(fp, l); 1584 ffree(fp); 1585 fdremove(fdp, indx); 1586 if (vp != NULL) 1587 vput(vp); 1588 vfs_copyinfh_free(fh); 1589 return (error); 1590 } 1591 1592 /* ARGSUSED */ 1593 int 1594 sys___fhstat30(struct lwp *l, void *v, register_t *retval) 1595 { 1596 struct sys___fhstat30_args /* { 1597 syscallarg(const fhandle_t *) fhp; 1598 syscallarg(struct stat *) sb; 1599 } */ *uap = v; 1600 struct stat sb; 1601 int error; 1602 fhandle_t *fh; 1603 struct vnode *vp; 1604 1605 /* 1606 * Must be super user 1607 */ 1608 if ((error = kauth_authorize_generic(l->l_cred, 1609 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1610 return (error); 1611 1612 error = vfs_copyinfh_alloc(SCARG(uap, fhp), &fh); 1613 if (error != 0) { 1614 goto bad; 1615 } 1616 error = vfs_fhtovp(fh, &vp); 1617 if (error != 0) { 1618 goto bad; 1619 } 1620 error = vn_stat(vp, &sb, l); 1621 vput(vp); 1622 if (error) { 1623 goto bad; 1624 } 1625 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1626 bad: 1627 vfs_copyinfh_free(fh); 1628 return error; 1629 } 1630 1631 /* ARGSUSED */ 1632 int 1633 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1634 { 1635 struct sys_fhstatvfs1_args /* { 1636 syscallarg(const fhandle_t *) fhp; 1637 syscallarg(struct statvfs *) buf; 1638 syscallarg(int) flags; 1639 } */ *uap = v; 1640 struct statvfs *sb = NULL; 1641 fhandle_t *fh; 1642 struct mount *mp; 1643 struct vnode *vp; 1644 int error; 1645 1646 /* 1647 * Must be super user 1648 */ 1649 if ((error = kauth_authorize_generic(l->l_cred, 1650 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1651 return error; 1652 1653 error = vfs_copyinfh_alloc(SCARG(uap, fhp), &fh); 1654 if (error != 0) { 1655 goto out; 1656 } 1657 error = vfs_fhtovp(fh, &vp); 1658 if (error != 0) { 1659 goto out; 1660 } 1661 mp = vp->v_mount; 1662 sb = STATVFSBUF_GET(); 1663 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) { 1664 vput(vp); 1665 goto out; 1666 } 1667 vput(vp); 1668 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1669 out: 1670 if (sb != NULL) { 1671 STATVFSBUF_PUT(sb); 1672 } 1673 vfs_copyinfh_free(fh); 1674 return error; 1675 } 1676 1677 /* 1678 * Create a special file. 1679 */ 1680 /* ARGSUSED */ 1681 int 1682 sys_mknod(struct lwp *l, void *v, register_t *retval) 1683 { 1684 struct sys_mknod_args /* { 1685 syscallarg(const char *) path; 1686 syscallarg(int) mode; 1687 syscallarg(int) dev; 1688 } */ *uap = v; 1689 struct proc *p = l->l_proc; 1690 struct vnode *vp; 1691 struct mount *mp; 1692 struct vattr vattr; 1693 int error; 1694 int whiteout = 0; 1695 struct nameidata nd; 1696 1697 if ((error = kauth_authorize_generic(l->l_cred, 1698 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1699 return (error); 1700 restart: 1701 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1702 if ((error = namei(&nd)) != 0) 1703 return (error); 1704 vp = nd.ni_vp; 1705 if (vp != NULL) 1706 error = EEXIST; 1707 else { 1708 VATTR_NULL(&vattr); 1709 vattr.va_mode = 1710 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1711 vattr.va_rdev = SCARG(uap, dev); 1712 whiteout = 0; 1713 1714 switch (SCARG(uap, mode) & S_IFMT) { 1715 case S_IFMT: /* used by badsect to flag bad sectors */ 1716 vattr.va_type = VBAD; 1717 break; 1718 case S_IFCHR: 1719 vattr.va_type = VCHR; 1720 break; 1721 case S_IFBLK: 1722 vattr.va_type = VBLK; 1723 break; 1724 case S_IFWHT: 1725 whiteout = 1; 1726 break; 1727 default: 1728 error = EINVAL; 1729 break; 1730 } 1731 } 1732 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1733 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1734 if (nd.ni_dvp == vp) 1735 vrele(nd.ni_dvp); 1736 else 1737 vput(nd.ni_dvp); 1738 if (vp) 1739 vrele(vp); 1740 if ((error = vn_start_write(NULL, &mp, 1741 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1742 return (error); 1743 goto restart; 1744 } 1745 if (!error) { 1746 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1747 if (whiteout) { 1748 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1749 if (error) 1750 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1751 vput(nd.ni_dvp); 1752 } else { 1753 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1754 &nd.ni_cnd, &vattr); 1755 if (error == 0) 1756 vput(nd.ni_vp); 1757 } 1758 } else { 1759 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1760 if (nd.ni_dvp == vp) 1761 vrele(nd.ni_dvp); 1762 else 1763 vput(nd.ni_dvp); 1764 if (vp) 1765 vrele(vp); 1766 } 1767 vn_finished_write(mp, 0); 1768 return (error); 1769 } 1770 1771 /* 1772 * Create a named pipe. 1773 */ 1774 /* ARGSUSED */ 1775 int 1776 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1777 { 1778 struct sys_mkfifo_args /* { 1779 syscallarg(const char *) path; 1780 syscallarg(int) mode; 1781 } */ *uap = v; 1782 struct proc *p = l->l_proc; 1783 struct mount *mp; 1784 struct vattr vattr; 1785 int error; 1786 struct nameidata nd; 1787 1788 restart: 1789 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1790 if ((error = namei(&nd)) != 0) 1791 return (error); 1792 if (nd.ni_vp != NULL) { 1793 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1794 if (nd.ni_dvp == nd.ni_vp) 1795 vrele(nd.ni_dvp); 1796 else 1797 vput(nd.ni_dvp); 1798 vrele(nd.ni_vp); 1799 return (EEXIST); 1800 } 1801 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1802 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1803 if (nd.ni_dvp == nd.ni_vp) 1804 vrele(nd.ni_dvp); 1805 else 1806 vput(nd.ni_dvp); 1807 if (nd.ni_vp) 1808 vrele(nd.ni_vp); 1809 if ((error = vn_start_write(NULL, &mp, 1810 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1811 return (error); 1812 goto restart; 1813 } 1814 VATTR_NULL(&vattr); 1815 vattr.va_type = VFIFO; 1816 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1817 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1818 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1819 if (error == 0) 1820 vput(nd.ni_vp); 1821 vn_finished_write(mp, 0); 1822 return (error); 1823 } 1824 1825 /* 1826 * Make a hard file link. 1827 */ 1828 /* ARGSUSED */ 1829 int 1830 sys_link(struct lwp *l, void *v, register_t *retval) 1831 { 1832 struct sys_link_args /* { 1833 syscallarg(const char *) path; 1834 syscallarg(const char *) link; 1835 } */ *uap = v; 1836 struct vnode *vp; 1837 struct mount *mp; 1838 struct nameidata nd; 1839 int error; 1840 1841 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1842 if ((error = namei(&nd)) != 0) 1843 return (error); 1844 vp = nd.ni_vp; 1845 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1846 vrele(vp); 1847 return (error); 1848 } 1849 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1850 if ((error = namei(&nd)) != 0) 1851 goto out; 1852 if (nd.ni_vp) { 1853 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1854 if (nd.ni_dvp == nd.ni_vp) 1855 vrele(nd.ni_dvp); 1856 else 1857 vput(nd.ni_dvp); 1858 vrele(nd.ni_vp); 1859 error = EEXIST; 1860 goto out; 1861 } 1862 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1863 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1864 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1865 out: 1866 vrele(vp); 1867 vn_finished_write(mp, 0); 1868 return (error); 1869 } 1870 1871 /* 1872 * Make a symbolic link. 1873 */ 1874 /* ARGSUSED */ 1875 int 1876 sys_symlink(struct lwp *l, void *v, register_t *retval) 1877 { 1878 struct sys_symlink_args /* { 1879 syscallarg(const char *) path; 1880 syscallarg(const char *) link; 1881 } */ *uap = v; 1882 struct proc *p = l->l_proc; 1883 struct mount *mp; 1884 struct vattr vattr; 1885 char *path; 1886 int error; 1887 struct nameidata nd; 1888 1889 path = PNBUF_GET(); 1890 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1891 if (error) 1892 goto out; 1893 restart: 1894 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1895 if ((error = namei(&nd)) != 0) 1896 goto out; 1897 if (nd.ni_vp) { 1898 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1899 if (nd.ni_dvp == nd.ni_vp) 1900 vrele(nd.ni_dvp); 1901 else 1902 vput(nd.ni_dvp); 1903 vrele(nd.ni_vp); 1904 error = EEXIST; 1905 goto out; 1906 } 1907 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1908 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1909 if (nd.ni_dvp == nd.ni_vp) 1910 vrele(nd.ni_dvp); 1911 else 1912 vput(nd.ni_dvp); 1913 if ((error = vn_start_write(NULL, &mp, 1914 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1915 return (error); 1916 goto restart; 1917 } 1918 VATTR_NULL(&vattr); 1919 vattr.va_type = VLNK; 1920 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1921 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1922 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1923 if (error == 0) 1924 vput(nd.ni_vp); 1925 vn_finished_write(mp, 0); 1926 out: 1927 PNBUF_PUT(path); 1928 return (error); 1929 } 1930 1931 /* 1932 * Delete a whiteout from the filesystem. 1933 */ 1934 /* ARGSUSED */ 1935 int 1936 sys_undelete(struct lwp *l, void *v, register_t *retval) 1937 { 1938 struct sys_undelete_args /* { 1939 syscallarg(const char *) path; 1940 } */ *uap = v; 1941 int error; 1942 struct mount *mp; 1943 struct nameidata nd; 1944 1945 restart: 1946 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1947 SCARG(uap, path), l); 1948 error = namei(&nd); 1949 if (error) 1950 return (error); 1951 1952 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1953 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1954 if (nd.ni_dvp == nd.ni_vp) 1955 vrele(nd.ni_dvp); 1956 else 1957 vput(nd.ni_dvp); 1958 if (nd.ni_vp) 1959 vrele(nd.ni_vp); 1960 return (EEXIST); 1961 } 1962 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1963 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1964 if (nd.ni_dvp == nd.ni_vp) 1965 vrele(nd.ni_dvp); 1966 else 1967 vput(nd.ni_dvp); 1968 if ((error = vn_start_write(NULL, &mp, 1969 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1970 return (error); 1971 goto restart; 1972 } 1973 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1974 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1975 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1976 vput(nd.ni_dvp); 1977 vn_finished_write(mp, 0); 1978 return (error); 1979 } 1980 1981 /* 1982 * Delete a name from the filesystem. 1983 */ 1984 /* ARGSUSED */ 1985 int 1986 sys_unlink(struct lwp *l, void *v, register_t *retval) 1987 { 1988 struct sys_unlink_args /* { 1989 syscallarg(const char *) path; 1990 } */ *uap = v; 1991 struct mount *mp; 1992 struct vnode *vp; 1993 int error; 1994 struct nameidata nd; 1995 1996 restart: 1997 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1998 SCARG(uap, path), l); 1999 if ((error = namei(&nd)) != 0) 2000 return (error); 2001 vp = nd.ni_vp; 2002 2003 /* 2004 * The root of a mounted filesystem cannot be deleted. 2005 */ 2006 if (vp->v_flag & VROOT) { 2007 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2008 if (nd.ni_dvp == vp) 2009 vrele(nd.ni_dvp); 2010 else 2011 vput(nd.ni_dvp); 2012 vput(vp); 2013 error = EBUSY; 2014 goto out; 2015 } 2016 2017 #if NVERIEXEC > 0 2018 /* Handle remove requests for veriexec entries. */ 2019 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) { 2020 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2021 if (nd.ni_dvp == vp) 2022 vrele(nd.ni_dvp); 2023 else 2024 vput(nd.ni_dvp); 2025 vput(vp); 2026 goto out; 2027 } 2028 #endif /* NVERIEXEC > 0 */ 2029 2030 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2031 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2032 if (nd.ni_dvp == vp) 2033 vrele(nd.ni_dvp); 2034 else 2035 vput(nd.ni_dvp); 2036 vput(vp); 2037 if ((error = vn_start_write(NULL, &mp, 2038 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2039 return (error); 2040 goto restart; 2041 } 2042 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2043 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2044 #ifdef FILEASSOC 2045 (void)fileassoc_file_delete(nd.ni_vp); 2046 #endif /* FILEASSOC */ 2047 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2048 vn_finished_write(mp, 0); 2049 out: 2050 return (error); 2051 } 2052 2053 /* 2054 * Reposition read/write file offset. 2055 */ 2056 int 2057 sys_lseek(struct lwp *l, void *v, register_t *retval) 2058 { 2059 struct sys_lseek_args /* { 2060 syscallarg(int) fd; 2061 syscallarg(int) pad; 2062 syscallarg(off_t) offset; 2063 syscallarg(int) whence; 2064 } */ *uap = v; 2065 struct proc *p = l->l_proc; 2066 kauth_cred_t cred = l->l_cred; 2067 struct filedesc *fdp = p->p_fd; 2068 struct file *fp; 2069 struct vnode *vp; 2070 struct vattr vattr; 2071 off_t newoff; 2072 int error; 2073 2074 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2075 return (EBADF); 2076 2077 FILE_USE(fp); 2078 2079 vp = (struct vnode *)fp->f_data; 2080 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2081 error = ESPIPE; 2082 goto out; 2083 } 2084 2085 switch (SCARG(uap, whence)) { 2086 case SEEK_CUR: 2087 newoff = fp->f_offset + SCARG(uap, offset); 2088 break; 2089 case SEEK_END: 2090 error = VOP_GETATTR(vp, &vattr, cred, l); 2091 if (error) 2092 goto out; 2093 newoff = SCARG(uap, offset) + vattr.va_size; 2094 break; 2095 case SEEK_SET: 2096 newoff = SCARG(uap, offset); 2097 break; 2098 default: 2099 error = EINVAL; 2100 goto out; 2101 } 2102 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2103 goto out; 2104 2105 *(off_t *)retval = fp->f_offset = newoff; 2106 out: 2107 FILE_UNUSE(fp, l); 2108 return (error); 2109 } 2110 2111 /* 2112 * Positional read system call. 2113 */ 2114 int 2115 sys_pread(struct lwp *l, void *v, register_t *retval) 2116 { 2117 struct sys_pread_args /* { 2118 syscallarg(int) fd; 2119 syscallarg(void *) buf; 2120 syscallarg(size_t) nbyte; 2121 syscallarg(off_t) offset; 2122 } */ *uap = v; 2123 struct proc *p = l->l_proc; 2124 struct filedesc *fdp = p->p_fd; 2125 struct file *fp; 2126 struct vnode *vp; 2127 off_t offset; 2128 int error, fd = SCARG(uap, fd); 2129 2130 if ((fp = fd_getfile(fdp, fd)) == NULL) 2131 return (EBADF); 2132 2133 if ((fp->f_flag & FREAD) == 0) { 2134 simple_unlock(&fp->f_slock); 2135 return (EBADF); 2136 } 2137 2138 FILE_USE(fp); 2139 2140 vp = (struct vnode *)fp->f_data; 2141 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2142 error = ESPIPE; 2143 goto out; 2144 } 2145 2146 offset = SCARG(uap, offset); 2147 2148 /* 2149 * XXX This works because no file systems actually 2150 * XXX take any action on the seek operation. 2151 */ 2152 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2153 goto out; 2154 2155 /* dofileread() will unuse the descriptor for us */ 2156 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2157 &offset, 0, retval)); 2158 2159 out: 2160 FILE_UNUSE(fp, l); 2161 return (error); 2162 } 2163 2164 /* 2165 * Positional scatter read system call. 2166 */ 2167 int 2168 sys_preadv(struct lwp *l, void *v, register_t *retval) 2169 { 2170 struct sys_preadv_args /* { 2171 syscallarg(int) fd; 2172 syscallarg(const struct iovec *) iovp; 2173 syscallarg(int) iovcnt; 2174 syscallarg(off_t) offset; 2175 } */ *uap = v; 2176 struct proc *p = l->l_proc; 2177 struct filedesc *fdp = p->p_fd; 2178 struct file *fp; 2179 struct vnode *vp; 2180 off_t offset; 2181 int error, fd = SCARG(uap, fd); 2182 2183 if ((fp = fd_getfile(fdp, fd)) == NULL) 2184 return (EBADF); 2185 2186 if ((fp->f_flag & FREAD) == 0) { 2187 simple_unlock(&fp->f_slock); 2188 return (EBADF); 2189 } 2190 2191 FILE_USE(fp); 2192 2193 vp = (struct vnode *)fp->f_data; 2194 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2195 error = ESPIPE; 2196 goto out; 2197 } 2198 2199 offset = SCARG(uap, offset); 2200 2201 /* 2202 * XXX This works because no file systems actually 2203 * XXX take any action on the seek operation. 2204 */ 2205 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2206 goto out; 2207 2208 /* dofilereadv() will unuse the descriptor for us */ 2209 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2210 &offset, 0, retval)); 2211 2212 out: 2213 FILE_UNUSE(fp, l); 2214 return (error); 2215 } 2216 2217 /* 2218 * Positional write system call. 2219 */ 2220 int 2221 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2222 { 2223 struct sys_pwrite_args /* { 2224 syscallarg(int) fd; 2225 syscallarg(const void *) buf; 2226 syscallarg(size_t) nbyte; 2227 syscallarg(off_t) offset; 2228 } */ *uap = v; 2229 struct proc *p = l->l_proc; 2230 struct filedesc *fdp = p->p_fd; 2231 struct file *fp; 2232 struct vnode *vp; 2233 off_t offset; 2234 int error, fd = SCARG(uap, fd); 2235 2236 if ((fp = fd_getfile(fdp, fd)) == NULL) 2237 return (EBADF); 2238 2239 if ((fp->f_flag & FWRITE) == 0) { 2240 simple_unlock(&fp->f_slock); 2241 return (EBADF); 2242 } 2243 2244 FILE_USE(fp); 2245 2246 vp = (struct vnode *)fp->f_data; 2247 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2248 error = ESPIPE; 2249 goto out; 2250 } 2251 2252 offset = SCARG(uap, offset); 2253 2254 /* 2255 * XXX This works because no file systems actually 2256 * XXX take any action on the seek operation. 2257 */ 2258 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2259 goto out; 2260 2261 /* dofilewrite() will unuse the descriptor for us */ 2262 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2263 &offset, 0, retval)); 2264 2265 out: 2266 FILE_UNUSE(fp, l); 2267 return (error); 2268 } 2269 2270 /* 2271 * Positional gather write system call. 2272 */ 2273 int 2274 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2275 { 2276 struct sys_pwritev_args /* { 2277 syscallarg(int) fd; 2278 syscallarg(const struct iovec *) iovp; 2279 syscallarg(int) iovcnt; 2280 syscallarg(off_t) offset; 2281 } */ *uap = v; 2282 struct proc *p = l->l_proc; 2283 struct filedesc *fdp = p->p_fd; 2284 struct file *fp; 2285 struct vnode *vp; 2286 off_t offset; 2287 int error, fd = SCARG(uap, fd); 2288 2289 if ((fp = fd_getfile(fdp, fd)) == NULL) 2290 return (EBADF); 2291 2292 if ((fp->f_flag & FWRITE) == 0) { 2293 simple_unlock(&fp->f_slock); 2294 return (EBADF); 2295 } 2296 2297 FILE_USE(fp); 2298 2299 vp = (struct vnode *)fp->f_data; 2300 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2301 error = ESPIPE; 2302 goto out; 2303 } 2304 2305 offset = SCARG(uap, offset); 2306 2307 /* 2308 * XXX This works because no file systems actually 2309 * XXX take any action on the seek operation. 2310 */ 2311 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2312 goto out; 2313 2314 /* dofilewritev() will unuse the descriptor for us */ 2315 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2316 &offset, 0, retval)); 2317 2318 out: 2319 FILE_UNUSE(fp, l); 2320 return (error); 2321 } 2322 2323 /* 2324 * Check access permissions. 2325 */ 2326 int 2327 sys_access(struct lwp *l, void *v, register_t *retval) 2328 { 2329 struct sys_access_args /* { 2330 syscallarg(const char *) path; 2331 syscallarg(int) flags; 2332 } */ *uap = v; 2333 kauth_cred_t cred; 2334 struct vnode *vp; 2335 int error, flags; 2336 struct nameidata nd; 2337 2338 cred = kauth_cred_dup(l->l_cred); 2339 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2340 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2341 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2342 SCARG(uap, path), l); 2343 /* Override default credentials */ 2344 nd.ni_cnd.cn_cred = cred; 2345 if ((error = namei(&nd)) != 0) 2346 goto out; 2347 vp = nd.ni_vp; 2348 2349 /* Flags == 0 means only check for existence. */ 2350 if (SCARG(uap, flags)) { 2351 flags = 0; 2352 if (SCARG(uap, flags) & R_OK) 2353 flags |= VREAD; 2354 if (SCARG(uap, flags) & W_OK) 2355 flags |= VWRITE; 2356 if (SCARG(uap, flags) & X_OK) 2357 flags |= VEXEC; 2358 2359 error = VOP_ACCESS(vp, flags, cred, l); 2360 if (!error && (flags & VWRITE)) 2361 error = vn_writechk(vp); 2362 } 2363 vput(vp); 2364 out: 2365 kauth_cred_free(cred); 2366 return (error); 2367 } 2368 2369 /* 2370 * Get file status; this version follows links. 2371 */ 2372 /* ARGSUSED */ 2373 int 2374 sys___stat30(struct lwp *l, void *v, register_t *retval) 2375 { 2376 struct sys___stat30_args /* { 2377 syscallarg(const char *) path; 2378 syscallarg(struct stat *) ub; 2379 } */ *uap = v; 2380 struct stat sb; 2381 int error; 2382 struct nameidata nd; 2383 2384 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2385 SCARG(uap, path), l); 2386 if ((error = namei(&nd)) != 0) 2387 return (error); 2388 error = vn_stat(nd.ni_vp, &sb, l); 2389 vput(nd.ni_vp); 2390 if (error) 2391 return (error); 2392 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2393 return (error); 2394 } 2395 2396 /* 2397 * Get file status; this version does not follow links. 2398 */ 2399 /* ARGSUSED */ 2400 int 2401 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2402 { 2403 struct sys___lstat30_args /* { 2404 syscallarg(const char *) path; 2405 syscallarg(struct stat *) ub; 2406 } */ *uap = v; 2407 struct stat sb; 2408 int error; 2409 struct nameidata nd; 2410 2411 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2412 SCARG(uap, path), l); 2413 if ((error = namei(&nd)) != 0) 2414 return (error); 2415 error = vn_stat(nd.ni_vp, &sb, l); 2416 vput(nd.ni_vp); 2417 if (error) 2418 return (error); 2419 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2420 return (error); 2421 } 2422 2423 /* 2424 * Get configurable pathname variables. 2425 */ 2426 /* ARGSUSED */ 2427 int 2428 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2429 { 2430 struct sys_pathconf_args /* { 2431 syscallarg(const char *) path; 2432 syscallarg(int) name; 2433 } */ *uap = v; 2434 int error; 2435 struct nameidata nd; 2436 2437 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2438 SCARG(uap, path), l); 2439 if ((error = namei(&nd)) != 0) 2440 return (error); 2441 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2442 vput(nd.ni_vp); 2443 return (error); 2444 } 2445 2446 /* 2447 * Return target name of a symbolic link. 2448 */ 2449 /* ARGSUSED */ 2450 int 2451 sys_readlink(struct lwp *l, void *v, register_t *retval) 2452 { 2453 struct sys_readlink_args /* { 2454 syscallarg(const char *) path; 2455 syscallarg(char *) buf; 2456 syscallarg(size_t) count; 2457 } */ *uap = v; 2458 struct vnode *vp; 2459 struct iovec aiov; 2460 struct uio auio; 2461 int error; 2462 struct nameidata nd; 2463 2464 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2465 SCARG(uap, path), l); 2466 if ((error = namei(&nd)) != 0) 2467 return (error); 2468 vp = nd.ni_vp; 2469 if (vp->v_type != VLNK) 2470 error = EINVAL; 2471 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2472 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2473 aiov.iov_base = SCARG(uap, buf); 2474 aiov.iov_len = SCARG(uap, count); 2475 auio.uio_iov = &aiov; 2476 auio.uio_iovcnt = 1; 2477 auio.uio_offset = 0; 2478 auio.uio_rw = UIO_READ; 2479 KASSERT(l == curlwp); 2480 auio.uio_vmspace = l->l_proc->p_vmspace; 2481 auio.uio_resid = SCARG(uap, count); 2482 error = VOP_READLINK(vp, &auio, l->l_cred); 2483 } 2484 vput(vp); 2485 *retval = SCARG(uap, count) - auio.uio_resid; 2486 return (error); 2487 } 2488 2489 /* 2490 * Change flags of a file given a path name. 2491 */ 2492 /* ARGSUSED */ 2493 int 2494 sys_chflags(struct lwp *l, void *v, register_t *retval) 2495 { 2496 struct sys_chflags_args /* { 2497 syscallarg(const char *) path; 2498 syscallarg(u_long) flags; 2499 } */ *uap = v; 2500 struct vnode *vp; 2501 int error; 2502 struct nameidata nd; 2503 2504 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2505 if ((error = namei(&nd)) != 0) 2506 return (error); 2507 vp = nd.ni_vp; 2508 error = change_flags(vp, SCARG(uap, flags), l); 2509 vput(vp); 2510 return (error); 2511 } 2512 2513 /* 2514 * Change flags of a file given a file descriptor. 2515 */ 2516 /* ARGSUSED */ 2517 int 2518 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2519 { 2520 struct sys_fchflags_args /* { 2521 syscallarg(int) fd; 2522 syscallarg(u_long) flags; 2523 } */ *uap = v; 2524 struct proc *p = l->l_proc; 2525 struct vnode *vp; 2526 struct file *fp; 2527 int error; 2528 2529 /* getvnode() will use the descriptor for us */ 2530 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2531 return (error); 2532 vp = (struct vnode *)fp->f_data; 2533 error = change_flags(vp, SCARG(uap, flags), l); 2534 VOP_UNLOCK(vp, 0); 2535 FILE_UNUSE(fp, l); 2536 return (error); 2537 } 2538 2539 /* 2540 * Change flags of a file given a path name; this version does 2541 * not follow links. 2542 */ 2543 int 2544 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2545 { 2546 struct sys_lchflags_args /* { 2547 syscallarg(const char *) path; 2548 syscallarg(u_long) flags; 2549 } */ *uap = v; 2550 struct vnode *vp; 2551 int error; 2552 struct nameidata nd; 2553 2554 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2555 if ((error = namei(&nd)) != 0) 2556 return (error); 2557 vp = nd.ni_vp; 2558 error = change_flags(vp, SCARG(uap, flags), l); 2559 vput(vp); 2560 return (error); 2561 } 2562 2563 /* 2564 * Common routine to change flags of a file. 2565 */ 2566 int 2567 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2568 { 2569 struct mount *mp; 2570 struct vattr vattr; 2571 int error; 2572 2573 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2574 return (error); 2575 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2576 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2577 /* 2578 * Non-superusers cannot change the flags on devices, even if they 2579 * own them. 2580 */ 2581 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2582 &l->l_acflag) != 0) { 2583 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2584 goto out; 2585 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2586 error = EINVAL; 2587 goto out; 2588 } 2589 } 2590 VATTR_NULL(&vattr); 2591 vattr.va_flags = flags; 2592 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2593 out: 2594 vn_finished_write(mp, 0); 2595 return (error); 2596 } 2597 2598 /* 2599 * Change mode of a file given path name; this version follows links. 2600 */ 2601 /* ARGSUSED */ 2602 int 2603 sys_chmod(struct lwp *l, void *v, register_t *retval) 2604 { 2605 struct sys_chmod_args /* { 2606 syscallarg(const char *) path; 2607 syscallarg(int) mode; 2608 } */ *uap = v; 2609 int error; 2610 struct nameidata nd; 2611 2612 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2613 if ((error = namei(&nd)) != 0) 2614 return (error); 2615 2616 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2617 2618 vrele(nd.ni_vp); 2619 return (error); 2620 } 2621 2622 /* 2623 * Change mode of a file given a file descriptor. 2624 */ 2625 /* ARGSUSED */ 2626 int 2627 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2628 { 2629 struct sys_fchmod_args /* { 2630 syscallarg(int) fd; 2631 syscallarg(int) mode; 2632 } */ *uap = v; 2633 struct proc *p = l->l_proc; 2634 struct file *fp; 2635 int error; 2636 2637 /* getvnode() will use the descriptor for us */ 2638 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2639 return (error); 2640 2641 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2642 FILE_UNUSE(fp, l); 2643 return (error); 2644 } 2645 2646 /* 2647 * Change mode of a file given path name; this version does not follow links. 2648 */ 2649 /* ARGSUSED */ 2650 int 2651 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2652 { 2653 struct sys_lchmod_args /* { 2654 syscallarg(const char *) path; 2655 syscallarg(int) mode; 2656 } */ *uap = v; 2657 int error; 2658 struct nameidata nd; 2659 2660 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2661 if ((error = namei(&nd)) != 0) 2662 return (error); 2663 2664 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2665 2666 vrele(nd.ni_vp); 2667 return (error); 2668 } 2669 2670 /* 2671 * Common routine to set mode given a vnode. 2672 */ 2673 static int 2674 change_mode(struct vnode *vp, int mode, struct lwp *l) 2675 { 2676 struct mount *mp; 2677 struct vattr vattr; 2678 int error; 2679 2680 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2681 return (error); 2682 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2683 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2684 VATTR_NULL(&vattr); 2685 vattr.va_mode = mode & ALLPERMS; 2686 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2687 VOP_UNLOCK(vp, 0); 2688 vn_finished_write(mp, 0); 2689 return (error); 2690 } 2691 2692 /* 2693 * Set ownership given a path name; this version follows links. 2694 */ 2695 /* ARGSUSED */ 2696 int 2697 sys_chown(struct lwp *l, void *v, register_t *retval) 2698 { 2699 struct sys_chown_args /* { 2700 syscallarg(const char *) path; 2701 syscallarg(uid_t) uid; 2702 syscallarg(gid_t) gid; 2703 } */ *uap = v; 2704 int error; 2705 struct nameidata nd; 2706 2707 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2708 if ((error = namei(&nd)) != 0) 2709 return (error); 2710 2711 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2712 2713 vrele(nd.ni_vp); 2714 return (error); 2715 } 2716 2717 /* 2718 * Set ownership given a path name; this version follows links. 2719 * Provides POSIX semantics. 2720 */ 2721 /* ARGSUSED */ 2722 int 2723 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2724 { 2725 struct sys_chown_args /* { 2726 syscallarg(const char *) path; 2727 syscallarg(uid_t) uid; 2728 syscallarg(gid_t) gid; 2729 } */ *uap = v; 2730 int error; 2731 struct nameidata nd; 2732 2733 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2734 if ((error = namei(&nd)) != 0) 2735 return (error); 2736 2737 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2738 2739 vrele(nd.ni_vp); 2740 return (error); 2741 } 2742 2743 /* 2744 * Set ownership given a file descriptor. 2745 */ 2746 /* ARGSUSED */ 2747 int 2748 sys_fchown(struct lwp *l, void *v, register_t *retval) 2749 { 2750 struct sys_fchown_args /* { 2751 syscallarg(int) fd; 2752 syscallarg(uid_t) uid; 2753 syscallarg(gid_t) gid; 2754 } */ *uap = v; 2755 struct proc *p = l->l_proc; 2756 int error; 2757 struct file *fp; 2758 2759 /* getvnode() will use the descriptor for us */ 2760 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2761 return (error); 2762 2763 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2764 SCARG(uap, gid), l, 0); 2765 FILE_UNUSE(fp, l); 2766 return (error); 2767 } 2768 2769 /* 2770 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2771 */ 2772 /* ARGSUSED */ 2773 int 2774 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2775 { 2776 struct sys_fchown_args /* { 2777 syscallarg(int) fd; 2778 syscallarg(uid_t) uid; 2779 syscallarg(gid_t) gid; 2780 } */ *uap = v; 2781 struct proc *p = l->l_proc; 2782 int error; 2783 struct file *fp; 2784 2785 /* getvnode() will use the descriptor for us */ 2786 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2787 return (error); 2788 2789 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2790 SCARG(uap, gid), l, 1); 2791 FILE_UNUSE(fp, l); 2792 return (error); 2793 } 2794 2795 /* 2796 * Set ownership given a path name; this version does not follow links. 2797 */ 2798 /* ARGSUSED */ 2799 int 2800 sys_lchown(struct lwp *l, void *v, register_t *retval) 2801 { 2802 struct sys_lchown_args /* { 2803 syscallarg(const char *) path; 2804 syscallarg(uid_t) uid; 2805 syscallarg(gid_t) gid; 2806 } */ *uap = v; 2807 int error; 2808 struct nameidata nd; 2809 2810 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2811 if ((error = namei(&nd)) != 0) 2812 return (error); 2813 2814 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2815 2816 vrele(nd.ni_vp); 2817 return (error); 2818 } 2819 2820 /* 2821 * Set ownership given a path name; this version does not follow links. 2822 * Provides POSIX/XPG semantics. 2823 */ 2824 /* ARGSUSED */ 2825 int 2826 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2827 { 2828 struct sys_lchown_args /* { 2829 syscallarg(const char *) path; 2830 syscallarg(uid_t) uid; 2831 syscallarg(gid_t) gid; 2832 } */ *uap = v; 2833 int error; 2834 struct nameidata nd; 2835 2836 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2837 if ((error = namei(&nd)) != 0) 2838 return (error); 2839 2840 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2841 2842 vrele(nd.ni_vp); 2843 return (error); 2844 } 2845 2846 /* 2847 * Common routine to set ownership given a vnode. 2848 */ 2849 static int 2850 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2851 int posix_semantics) 2852 { 2853 struct mount *mp; 2854 struct vattr vattr; 2855 mode_t newmode; 2856 int error; 2857 2858 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2859 return (error); 2860 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2861 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2862 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2863 goto out; 2864 2865 #define CHANGED(x) ((int)(x) != -1) 2866 newmode = vattr.va_mode; 2867 if (posix_semantics) { 2868 /* 2869 * POSIX/XPG semantics: if the caller is not the super-user, 2870 * clear set-user-id and set-group-id bits. Both POSIX and 2871 * the XPG consider the behaviour for calls by the super-user 2872 * implementation-defined; we leave the set-user-id and set- 2873 * group-id settings intact in that case. 2874 */ 2875 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2876 NULL) != 0) 2877 newmode &= ~(S_ISUID | S_ISGID); 2878 } else { 2879 /* 2880 * NetBSD semantics: when changing owner and/or group, 2881 * clear the respective bit(s). 2882 */ 2883 if (CHANGED(uid)) 2884 newmode &= ~S_ISUID; 2885 if (CHANGED(gid)) 2886 newmode &= ~S_ISGID; 2887 } 2888 /* Update va_mode iff altered. */ 2889 if (vattr.va_mode == newmode) 2890 newmode = VNOVAL; 2891 2892 VATTR_NULL(&vattr); 2893 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2894 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2895 vattr.va_mode = newmode; 2896 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2897 #undef CHANGED 2898 2899 out: 2900 VOP_UNLOCK(vp, 0); 2901 vn_finished_write(mp, 0); 2902 return (error); 2903 } 2904 2905 /* 2906 * Set the access and modification times given a path name; this 2907 * version follows links. 2908 */ 2909 /* ARGSUSED */ 2910 int 2911 sys_utimes(struct lwp *l, void *v, register_t *retval) 2912 { 2913 struct sys_utimes_args /* { 2914 syscallarg(const char *) path; 2915 syscallarg(const struct timeval *) tptr; 2916 } */ *uap = v; 2917 int error; 2918 struct nameidata nd; 2919 2920 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2921 if ((error = namei(&nd)) != 0) 2922 return (error); 2923 2924 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2925 2926 vrele(nd.ni_vp); 2927 return (error); 2928 } 2929 2930 /* 2931 * Set the access and modification times given a file descriptor. 2932 */ 2933 /* ARGSUSED */ 2934 int 2935 sys_futimes(struct lwp *l, void *v, register_t *retval) 2936 { 2937 struct sys_futimes_args /* { 2938 syscallarg(int) fd; 2939 syscallarg(const struct timeval *) tptr; 2940 } */ *uap = v; 2941 struct proc *p = l->l_proc; 2942 int error; 2943 struct file *fp; 2944 2945 /* getvnode() will use the descriptor for us */ 2946 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2947 return (error); 2948 2949 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2950 FILE_UNUSE(fp, l); 2951 return (error); 2952 } 2953 2954 /* 2955 * Set the access and modification times given a path name; this 2956 * version does not follow links. 2957 */ 2958 /* ARGSUSED */ 2959 int 2960 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2961 { 2962 struct sys_lutimes_args /* { 2963 syscallarg(const char *) path; 2964 syscallarg(const struct timeval *) tptr; 2965 } */ *uap = v; 2966 int error; 2967 struct nameidata nd; 2968 2969 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2970 if ((error = namei(&nd)) != 0) 2971 return (error); 2972 2973 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2974 2975 vrele(nd.ni_vp); 2976 return (error); 2977 } 2978 2979 /* 2980 * Common routine to set access and modification times given a vnode. 2981 */ 2982 static int 2983 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2984 { 2985 struct mount *mp; 2986 struct vattr vattr; 2987 int error; 2988 2989 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2990 return (error); 2991 VATTR_NULL(&vattr); 2992 if (tptr == NULL) { 2993 nanotime(&vattr.va_atime); 2994 vattr.va_mtime = vattr.va_atime; 2995 vattr.va_vaflags |= VA_UTIMES_NULL; 2996 } else { 2997 struct timeval tv[2]; 2998 2999 error = copyin(tptr, tv, sizeof(tv)); 3000 if (error) 3001 goto out; 3002 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3003 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3004 } 3005 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3006 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3007 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3008 VOP_UNLOCK(vp, 0); 3009 out: 3010 vn_finished_write(mp, 0); 3011 return (error); 3012 } 3013 3014 /* 3015 * Truncate a file given its path name. 3016 */ 3017 /* ARGSUSED */ 3018 int 3019 sys_truncate(struct lwp *l, void *v, register_t *retval) 3020 { 3021 struct sys_truncate_args /* { 3022 syscallarg(const char *) path; 3023 syscallarg(int) pad; 3024 syscallarg(off_t) length; 3025 } */ *uap = v; 3026 struct vnode *vp; 3027 struct mount *mp; 3028 struct vattr vattr; 3029 int error; 3030 struct nameidata nd; 3031 3032 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3033 if ((error = namei(&nd)) != 0) 3034 return (error); 3035 vp = nd.ni_vp; 3036 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3037 vrele(vp); 3038 return (error); 3039 } 3040 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3041 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3042 if (vp->v_type == VDIR) 3043 error = EISDIR; 3044 else if ((error = vn_writechk(vp)) == 0 && 3045 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3046 VATTR_NULL(&vattr); 3047 vattr.va_size = SCARG(uap, length); 3048 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3049 } 3050 vput(vp); 3051 vn_finished_write(mp, 0); 3052 return (error); 3053 } 3054 3055 /* 3056 * Truncate a file given a file descriptor. 3057 */ 3058 /* ARGSUSED */ 3059 int 3060 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3061 { 3062 struct sys_ftruncate_args /* { 3063 syscallarg(int) fd; 3064 syscallarg(int) pad; 3065 syscallarg(off_t) length; 3066 } */ *uap = v; 3067 struct proc *p = l->l_proc; 3068 struct mount *mp; 3069 struct vattr vattr; 3070 struct vnode *vp; 3071 struct file *fp; 3072 int error; 3073 3074 /* getvnode() will use the descriptor for us */ 3075 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3076 return (error); 3077 if ((fp->f_flag & FWRITE) == 0) { 3078 error = EINVAL; 3079 goto out; 3080 } 3081 vp = (struct vnode *)fp->f_data; 3082 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3083 FILE_UNUSE(fp, l); 3084 return (error); 3085 } 3086 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3087 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3088 if (vp->v_type == VDIR) 3089 error = EISDIR; 3090 else if ((error = vn_writechk(vp)) == 0) { 3091 VATTR_NULL(&vattr); 3092 vattr.va_size = SCARG(uap, length); 3093 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3094 } 3095 VOP_UNLOCK(vp, 0); 3096 vn_finished_write(mp, 0); 3097 out: 3098 FILE_UNUSE(fp, l); 3099 return (error); 3100 } 3101 3102 /* 3103 * Sync an open file. 3104 */ 3105 /* ARGSUSED */ 3106 int 3107 sys_fsync(struct lwp *l, void *v, register_t *retval) 3108 { 3109 struct sys_fsync_args /* { 3110 syscallarg(int) fd; 3111 } */ *uap = v; 3112 struct proc *p = l->l_proc; 3113 struct vnode *vp; 3114 struct mount *mp; 3115 struct file *fp; 3116 int error; 3117 3118 /* getvnode() will use the descriptor for us */ 3119 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3120 return (error); 3121 vp = (struct vnode *)fp->f_data; 3122 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3123 FILE_UNUSE(fp, l); 3124 return (error); 3125 } 3126 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3127 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3128 if (error == 0 && bioops.io_fsync != NULL && 3129 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3130 (*bioops.io_fsync)(vp, 0); 3131 VOP_UNLOCK(vp, 0); 3132 vn_finished_write(mp, 0); 3133 FILE_UNUSE(fp, l); 3134 return (error); 3135 } 3136 3137 /* 3138 * Sync a range of file data. API modeled after that found in AIX. 3139 * 3140 * FDATASYNC indicates that we need only save enough metadata to be able 3141 * to re-read the written data. Note we duplicate AIX's requirement that 3142 * the file be open for writing. 3143 */ 3144 /* ARGSUSED */ 3145 int 3146 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3147 { 3148 struct sys_fsync_range_args /* { 3149 syscallarg(int) fd; 3150 syscallarg(int) flags; 3151 syscallarg(off_t) start; 3152 syscallarg(off_t) length; 3153 } */ *uap = v; 3154 struct proc *p = l->l_proc; 3155 struct vnode *vp; 3156 struct file *fp; 3157 int flags, nflags; 3158 off_t s, e, len; 3159 int error; 3160 3161 /* getvnode() will use the descriptor for us */ 3162 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3163 return (error); 3164 3165 if ((fp->f_flag & FWRITE) == 0) { 3166 FILE_UNUSE(fp, l); 3167 return (EBADF); 3168 } 3169 3170 flags = SCARG(uap, flags); 3171 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3172 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3173 return (EINVAL); 3174 } 3175 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3176 if (flags & FDATASYNC) 3177 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3178 else 3179 nflags = FSYNC_WAIT; 3180 if (flags & FDISKSYNC) 3181 nflags |= FSYNC_CACHE; 3182 3183 len = SCARG(uap, length); 3184 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3185 if (len) { 3186 s = SCARG(uap, start); 3187 e = s + len; 3188 if (e < s) { 3189 FILE_UNUSE(fp, l); 3190 return (EINVAL); 3191 } 3192 } else { 3193 e = 0; 3194 s = 0; 3195 } 3196 3197 vp = (struct vnode *)fp->f_data; 3198 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3199 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3200 3201 if (error == 0 && bioops.io_fsync != NULL && 3202 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3203 (*bioops.io_fsync)(vp, nflags); 3204 3205 VOP_UNLOCK(vp, 0); 3206 FILE_UNUSE(fp, l); 3207 return (error); 3208 } 3209 3210 /* 3211 * Sync the data of an open file. 3212 */ 3213 /* ARGSUSED */ 3214 int 3215 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3216 { 3217 struct sys_fdatasync_args /* { 3218 syscallarg(int) fd; 3219 } */ *uap = v; 3220 struct proc *p = l->l_proc; 3221 struct vnode *vp; 3222 struct file *fp; 3223 int error; 3224 3225 /* getvnode() will use the descriptor for us */ 3226 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3227 return (error); 3228 if ((fp->f_flag & FWRITE) == 0) { 3229 FILE_UNUSE(fp, l); 3230 return (EBADF); 3231 } 3232 vp = (struct vnode *)fp->f_data; 3233 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3234 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3235 VOP_UNLOCK(vp, 0); 3236 FILE_UNUSE(fp, l); 3237 return (error); 3238 } 3239 3240 /* 3241 * Rename files, (standard) BSD semantics frontend. 3242 */ 3243 /* ARGSUSED */ 3244 int 3245 sys_rename(struct lwp *l, void *v, register_t *retval) 3246 { 3247 struct sys_rename_args /* { 3248 syscallarg(const char *) from; 3249 syscallarg(const char *) to; 3250 } */ *uap = v; 3251 3252 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3253 } 3254 3255 /* 3256 * Rename files, POSIX semantics frontend. 3257 */ 3258 /* ARGSUSED */ 3259 int 3260 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3261 { 3262 struct sys___posix_rename_args /* { 3263 syscallarg(const char *) from; 3264 syscallarg(const char *) to; 3265 } */ *uap = v; 3266 3267 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3268 } 3269 3270 /* 3271 * Rename files. Source and destination must either both be directories, 3272 * or both not be directories. If target is a directory, it must be empty. 3273 * If `from' and `to' refer to the same object, the value of the `retain' 3274 * argument is used to determine whether `from' will be 3275 * 3276 * (retain == 0) deleted unless `from' and `to' refer to the same 3277 * object in the file system's name space (BSD). 3278 * (retain == 1) always retained (POSIX). 3279 */ 3280 static int 3281 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3282 { 3283 struct mount *mp = NULL; 3284 struct vnode *tvp, *fvp, *tdvp; 3285 struct nameidata fromnd, tond; 3286 struct proc *p; 3287 int error; 3288 3289 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3290 from, l); 3291 if ((error = namei(&fromnd)) != 0) 3292 return (error); 3293 fvp = fromnd.ni_vp; 3294 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3295 if (error != 0) { 3296 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3297 vrele(fromnd.ni_dvp); 3298 vrele(fvp); 3299 if (fromnd.ni_startdir) 3300 vrele(fromnd.ni_startdir); 3301 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3302 return (error); 3303 } 3304 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3305 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3306 if ((error = namei(&tond)) != 0) { 3307 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3308 vrele(fromnd.ni_dvp); 3309 vrele(fvp); 3310 goto out1; 3311 } 3312 tdvp = tond.ni_dvp; 3313 tvp = tond.ni_vp; 3314 3315 if (tvp != NULL) { 3316 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3317 error = ENOTDIR; 3318 goto out; 3319 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3320 error = EISDIR; 3321 goto out; 3322 } 3323 } 3324 3325 if (fvp == tdvp) 3326 error = EINVAL; 3327 3328 /* 3329 * Source and destination refer to the same object. 3330 */ 3331 if (fvp == tvp) { 3332 if (retain) 3333 error = -1; 3334 else if (fromnd.ni_dvp == tdvp && 3335 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3336 !memcmp(fromnd.ni_cnd.cn_nameptr, 3337 tond.ni_cnd.cn_nameptr, 3338 fromnd.ni_cnd.cn_namelen)) 3339 error = -1; 3340 } 3341 3342 #if NVERIEXEC > 0 3343 if (!error) 3344 error = veriexec_renamechk(fvp, tvp, fromnd.ni_dirp, 3345 tond.ni_dirp, l); 3346 #endif /* NVERIEXEC > 0 */ 3347 3348 out: 3349 p = l->l_proc; 3350 if (!error) { 3351 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3352 if (fromnd.ni_dvp != tdvp) 3353 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3354 if (tvp) { 3355 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3356 } 3357 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3358 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3359 } else { 3360 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3361 if (tdvp == tvp) 3362 vrele(tdvp); 3363 else 3364 vput(tdvp); 3365 if (tvp) 3366 vput(tvp); 3367 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3368 vrele(fromnd.ni_dvp); 3369 vrele(fvp); 3370 } 3371 vrele(tond.ni_startdir); 3372 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3373 out1: 3374 vn_finished_write(mp, 0); 3375 if (fromnd.ni_startdir) 3376 vrele(fromnd.ni_startdir); 3377 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3378 return (error == -1 ? 0 : error); 3379 } 3380 3381 /* 3382 * Make a directory file. 3383 */ 3384 /* ARGSUSED */ 3385 int 3386 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3387 { 3388 struct sys_mkdir_args /* { 3389 syscallarg(const char *) path; 3390 syscallarg(int) mode; 3391 } */ *uap = v; 3392 struct proc *p = l->l_proc; 3393 struct mount *mp; 3394 struct vnode *vp; 3395 struct vattr vattr; 3396 int error; 3397 struct nameidata nd; 3398 3399 restart: 3400 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3401 SCARG(uap, path), l); 3402 if ((error = namei(&nd)) != 0) 3403 return (error); 3404 vp = nd.ni_vp; 3405 if (vp != NULL) { 3406 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3407 if (nd.ni_dvp == vp) 3408 vrele(nd.ni_dvp); 3409 else 3410 vput(nd.ni_dvp); 3411 vrele(vp); 3412 return (EEXIST); 3413 } 3414 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3415 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3416 if (nd.ni_dvp == vp) 3417 vrele(nd.ni_dvp); 3418 else 3419 vput(nd.ni_dvp); 3420 if ((error = vn_start_write(NULL, &mp, 3421 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3422 return (error); 3423 goto restart; 3424 } 3425 VATTR_NULL(&vattr); 3426 vattr.va_type = VDIR; 3427 vattr.va_mode = 3428 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3429 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3430 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3431 if (!error) 3432 vput(nd.ni_vp); 3433 vn_finished_write(mp, 0); 3434 return (error); 3435 } 3436 3437 /* 3438 * Remove a directory file. 3439 */ 3440 /* ARGSUSED */ 3441 int 3442 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3443 { 3444 struct sys_rmdir_args /* { 3445 syscallarg(const char *) path; 3446 } */ *uap = v; 3447 struct mount *mp; 3448 struct vnode *vp; 3449 int error; 3450 struct nameidata nd; 3451 3452 restart: 3453 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3454 SCARG(uap, path), l); 3455 if ((error = namei(&nd)) != 0) 3456 return (error); 3457 vp = nd.ni_vp; 3458 if (vp->v_type != VDIR) { 3459 error = ENOTDIR; 3460 goto out; 3461 } 3462 /* 3463 * No rmdir "." please. 3464 */ 3465 if (nd.ni_dvp == vp) { 3466 error = EINVAL; 3467 goto out; 3468 } 3469 /* 3470 * The root of a mounted filesystem cannot be deleted. 3471 */ 3472 if (vp->v_flag & VROOT) { 3473 error = EBUSY; 3474 goto out; 3475 } 3476 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3477 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3478 if (nd.ni_dvp == vp) 3479 vrele(nd.ni_dvp); 3480 else 3481 vput(nd.ni_dvp); 3482 vput(vp); 3483 if ((error = vn_start_write(NULL, &mp, 3484 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3485 return (error); 3486 goto restart; 3487 } 3488 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3489 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3490 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3491 vn_finished_write(mp, 0); 3492 return (error); 3493 3494 out: 3495 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3496 if (nd.ni_dvp == vp) 3497 vrele(nd.ni_dvp); 3498 else 3499 vput(nd.ni_dvp); 3500 vput(vp); 3501 return (error); 3502 } 3503 3504 /* 3505 * Read a block of directory entries in a file system independent format. 3506 */ 3507 int 3508 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3509 { 3510 struct sys___getdents30_args /* { 3511 syscallarg(int) fd; 3512 syscallarg(char *) buf; 3513 syscallarg(size_t) count; 3514 } */ *uap = v; 3515 struct proc *p = l->l_proc; 3516 struct file *fp; 3517 int error, done; 3518 3519 /* getvnode() will use the descriptor for us */ 3520 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3521 return (error); 3522 if ((fp->f_flag & FREAD) == 0) { 3523 error = EBADF; 3524 goto out; 3525 } 3526 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3527 SCARG(uap, count), &done, l, 0, 0); 3528 #ifdef KTRACE 3529 if (!error && KTRPOINT(p, KTR_GENIO)) { 3530 struct iovec iov; 3531 iov.iov_base = SCARG(uap, buf); 3532 iov.iov_len = done; 3533 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3534 } 3535 #endif 3536 *retval = done; 3537 out: 3538 FILE_UNUSE(fp, l); 3539 return (error); 3540 } 3541 3542 /* 3543 * Set the mode mask for creation of filesystem nodes. 3544 */ 3545 int 3546 sys_umask(struct lwp *l, void *v, register_t *retval) 3547 { 3548 struct sys_umask_args /* { 3549 syscallarg(mode_t) newmask; 3550 } */ *uap = v; 3551 struct proc *p = l->l_proc; 3552 struct cwdinfo *cwdi; 3553 3554 cwdi = p->p_cwdi; 3555 *retval = cwdi->cwdi_cmask; 3556 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3557 return (0); 3558 } 3559 3560 /* 3561 * Void all references to file by ripping underlying filesystem 3562 * away from vnode. 3563 */ 3564 /* ARGSUSED */ 3565 int 3566 sys_revoke(struct lwp *l, void *v, register_t *retval) 3567 { 3568 struct sys_revoke_args /* { 3569 syscallarg(const char *) path; 3570 } */ *uap = v; 3571 struct mount *mp; 3572 struct vnode *vp; 3573 struct vattr vattr; 3574 int error; 3575 struct nameidata nd; 3576 3577 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3578 if ((error = namei(&nd)) != 0) 3579 return (error); 3580 vp = nd.ni_vp; 3581 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3582 goto out; 3583 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3584 (error = kauth_authorize_generic(l->l_cred, 3585 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 3586 goto out; 3587 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3588 goto out; 3589 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3590 VOP_REVOKE(vp, REVOKEALL); 3591 vn_finished_write(mp, 0); 3592 out: 3593 vrele(vp); 3594 return (error); 3595 } 3596 3597 /* 3598 * Convert a user file descriptor to a kernel file entry. 3599 */ 3600 int 3601 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3602 { 3603 struct vnode *vp; 3604 struct file *fp; 3605 3606 if ((fp = fd_getfile(fdp, fd)) == NULL) 3607 return (EBADF); 3608 3609 FILE_USE(fp); 3610 3611 if (fp->f_type != DTYPE_VNODE) { 3612 FILE_UNUSE(fp, NULL); 3613 return (EINVAL); 3614 } 3615 3616 vp = (struct vnode *)fp->f_data; 3617 if (vp->v_type == VBAD) { 3618 FILE_UNUSE(fp, NULL); 3619 return (EBADF); 3620 } 3621 3622 *fpp = fp; 3623 return (0); 3624 } 3625
Indexes created Fri Sep 26 08:10:20 GMT 2025