vfs_syscalls.c revision 1.263
1 /* $NetBSD: vfs_syscalls.c,v 1.263 2006/07/31 16:34:44 martin 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 1.263 2006/07/31 16:34:44 martin 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_alloc: allocate and copyin a filehandle. 1360 */ 1361 1362 int 1363 vfs_copyinfh_alloc(const void *ufhp, fhandle_t **fhpp) 1364 { 1365 fhandle_t tempfh; 1366 size_t fhsize; 1367 int error; 1368 1369 error = copyin(ufhp, &tempfh, sizeof(tempfh)); 1370 if (error) { 1371 return error; 1372 } 1373 fhsize = FHANDLE_SIZE(&tempfh); 1374 return vfs_copyinfh_alloc_size(ufhp, fhsize, fhpp); 1375 } 1376 1377 /* 1378 * vfs_copyinfh_alloc_size: allocate and copyin a filehandle, given 1379 * the needed size. 1380 */ 1381 1382 int 1383 vfs_copyinfh_alloc_size(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1384 { 1385 fhandle_t *fhp; 1386 int error; 1387 1388 *fhpp = NULL; 1389 if (fhsize > FHANDLE_SIZE_MAX) { 1390 return EINVAL; 1391 } 1392 fhp = kmem_alloc(fhsize, KM_SLEEP); 1393 if (fhp == NULL) { 1394 return ENOMEM; 1395 } 1396 error = copyin(ufhp, fhp, fhsize); 1397 if (error == 0) { 1398 if (FHANDLE_SIZE(fhp) >= FHANDLE_SIZE_MIN 1399 || FHANDLE_SIZE(fhp) <= fhsize) { 1400 *fhpp = fhp; 1401 return 0; 1402 } else { 1403 error = EINVAL; 1404 } 1405 } 1406 kmem_free(fhp, fhsize); 1407 return error; 1408 } 1409 1410 void 1411 vfs_copyinfh_free(fhandle_t *fhp) 1412 { 1413 1414 vfs__fhfree(fhp); 1415 } 1416 1417 /* 1418 * Get file handle system call 1419 */ 1420 int 1421 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1422 { 1423 struct sys___getfh30_args /* { 1424 syscallarg(char *) fname; 1425 syscallarg(fhandle_t *) fhp; 1426 syscallarg(size_t *) fh_size; 1427 } */ *uap = v; 1428 struct vnode *vp; 1429 fhandle_t *fh; 1430 int error; 1431 struct nameidata nd; 1432 size_t sz; 1433 size_t usz; 1434 1435 /* 1436 * Must be super user 1437 */ 1438 error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 1439 &l->l_acflag); 1440 if (error) 1441 return (error); 1442 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1443 SCARG(uap, fname), l); 1444 error = namei(&nd); 1445 if (error) 1446 return (error); 1447 vp = nd.ni_vp; 1448 error = vfs_composefh_alloc(vp, &fh); 1449 vput(vp); 1450 if (error != 0) { 1451 goto out; 1452 } 1453 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1454 if (error != 0) { 1455 goto out; 1456 } 1457 sz = FHANDLE_SIZE(fh); 1458 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1459 if (error != 0) { 1460 goto out; 1461 } 1462 if (usz >= sz) { 1463 error = copyout(fh, SCARG(uap, fhp), sz); 1464 } else { 1465 error = E2BIG; 1466 } 1467 out: 1468 vfs_composefh_free(fh); 1469 return (error); 1470 } 1471 1472 /* 1473 * Open a file given a file handle. 1474 * 1475 * Check permissions, allocate an open file structure, 1476 * and call the device open routine if any. 1477 */ 1478 int 1479 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1480 { 1481 struct sys___fhopen40_args /* { 1482 syscallarg(const void *) fhp; 1483 syscallarg(size_t) fh_size; 1484 syscallarg(int) flags; 1485 } */ *uap = v; 1486 struct filedesc *fdp = l->l_proc->p_fd; 1487 struct file *fp; 1488 struct vnode *vp = NULL; 1489 struct mount *mp; 1490 kauth_cred_t cred = l->l_cred; 1491 int flags; 1492 struct file *nfp; 1493 int type, indx, error=0; 1494 struct flock lf; 1495 struct vattr va; 1496 fhandle_t *fh; 1497 1498 /* 1499 * Must be super user 1500 */ 1501 if ((error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 1502 &l->l_acflag))) 1503 return (error); 1504 1505 flags = FFLAGS(SCARG(uap, flags)); 1506 if ((flags & (FREAD | FWRITE)) == 0) 1507 return (EINVAL); 1508 if ((flags & O_CREAT)) 1509 return (EINVAL); 1510 /* falloc() will use the file descriptor for us */ 1511 if ((error = falloc(l, &nfp, &indx)) != 0) 1512 return (error); 1513 fp = nfp; 1514 error = vfs_copyinfh_alloc_size(SCARG(uap, fhp), SCARG(uap, fh_size), 1515 &fh); 1516 if (error != 0) { 1517 goto bad; 1518 } 1519 error = vfs_fhtovp(fh, &vp); 1520 if (error != 0) { 1521 goto bad; 1522 } 1523 1524 /* Now do an effective vn_open */ 1525 1526 if (vp->v_type == VSOCK) { 1527 error = EOPNOTSUPP; 1528 goto bad; 1529 } 1530 if (flags & FREAD) { 1531 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1532 goto bad; 1533 } 1534 if (flags & (FWRITE | O_TRUNC)) { 1535 if (vp->v_type == VDIR) { 1536 error = EISDIR; 1537 goto bad; 1538 } 1539 if ((error = vn_writechk(vp)) != 0 || 1540 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1541 goto bad; 1542 } 1543 if (flags & O_TRUNC) { 1544 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1545 goto bad; 1546 VOP_UNLOCK(vp, 0); /* XXX */ 1547 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1548 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1549 VATTR_NULL(&va); 1550 va.va_size = 0; 1551 error = VOP_SETATTR(vp, &va, cred, l); 1552 vn_finished_write(mp, 0); 1553 if (error) 1554 goto bad; 1555 } 1556 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1557 goto bad; 1558 if (vp->v_type == VREG && 1559 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1560 error = EIO; 1561 goto bad; 1562 } 1563 if (flags & FWRITE) 1564 vp->v_writecount++; 1565 1566 /* done with modified vn_open, now finish what sys_open does. */ 1567 1568 fp->f_flag = flags & FMASK; 1569 fp->f_type = DTYPE_VNODE; 1570 fp->f_ops = &vnops; 1571 fp->f_data = vp; 1572 if (flags & (O_EXLOCK | O_SHLOCK)) { 1573 lf.l_whence = SEEK_SET; 1574 lf.l_start = 0; 1575 lf.l_len = 0; 1576 if (flags & O_EXLOCK) 1577 lf.l_type = F_WRLCK; 1578 else 1579 lf.l_type = F_RDLCK; 1580 type = F_FLOCK; 1581 if ((flags & FNONBLOCK) == 0) 1582 type |= F_WAIT; 1583 VOP_UNLOCK(vp, 0); 1584 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1585 if (error) { 1586 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1587 FILE_UNUSE(fp, l); 1588 ffree(fp); 1589 fdremove(fdp, indx); 1590 return (error); 1591 } 1592 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1593 fp->f_flag |= FHASLOCK; 1594 } 1595 VOP_UNLOCK(vp, 0); 1596 *retval = indx; 1597 FILE_SET_MATURE(fp); 1598 FILE_UNUSE(fp, l); 1599 vfs_copyinfh_free(fh); 1600 return (0); 1601 1602 bad: 1603 FILE_UNUSE(fp, l); 1604 ffree(fp); 1605 fdremove(fdp, indx); 1606 if (vp != NULL) 1607 vput(vp); 1608 vfs_copyinfh_free(fh); 1609 return (error); 1610 } 1611 1612 /* ARGSUSED */ 1613 int 1614 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1615 { 1616 struct sys___fhstat40_args /* { 1617 syscallarg(const void *) fhp; 1618 syscallarg(size_t) fh_size; 1619 syscallarg(struct stat *) sb; 1620 } */ *uap = v; 1621 struct stat sb; 1622 int error; 1623 fhandle_t *fh; 1624 struct vnode *vp; 1625 1626 /* 1627 * Must be super user 1628 */ 1629 if ((error = kauth_authorize_generic(l->l_cred, 1630 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1631 return (error); 1632 1633 error = vfs_copyinfh_alloc_size(SCARG(uap, fhp), SCARG(uap, fh_size), 1634 &fh); 1635 if (error != 0) { 1636 goto bad; 1637 } 1638 error = vfs_fhtovp(fh, &vp); 1639 if (error != 0) { 1640 goto bad; 1641 } 1642 error = vn_stat(vp, &sb, l); 1643 vput(vp); 1644 if (error) { 1645 goto bad; 1646 } 1647 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1648 bad: 1649 vfs_copyinfh_free(fh); 1650 return error; 1651 } 1652 1653 /* ARGSUSED */ 1654 int 1655 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1656 { 1657 struct sys___fhstatvfs140_args /* { 1658 syscallarg(const fhandle_t *) fhp; 1659 syscallarg(size_t) fh_size; 1660 syscallarg(struct statvfs *) buf; 1661 syscallarg(int) flags; 1662 } */ *uap = v; 1663 struct statvfs *sb = NULL; 1664 fhandle_t *fh; 1665 struct mount *mp; 1666 struct vnode *vp; 1667 int error; 1668 1669 /* 1670 * Must be super user 1671 */ 1672 if ((error = kauth_authorize_generic(l->l_cred, 1673 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1674 return error; 1675 1676 error = vfs_copyinfh_alloc_size(SCARG(uap, fhp), SCARG(uap, fh_size), 1677 &fh); 1678 if (error != 0) { 1679 goto out; 1680 } 1681 error = vfs_fhtovp(fh, &vp); 1682 if (error != 0) { 1683 goto out; 1684 } 1685 mp = vp->v_mount; 1686 sb = STATVFSBUF_GET(); 1687 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) { 1688 vput(vp); 1689 goto out; 1690 } 1691 vput(vp); 1692 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1693 out: 1694 if (sb != NULL) { 1695 STATVFSBUF_PUT(sb); 1696 } 1697 vfs_copyinfh_free(fh); 1698 return error; 1699 } 1700 1701 /* 1702 * Create a special file. 1703 */ 1704 /* ARGSUSED */ 1705 int 1706 sys_mknod(struct lwp *l, void *v, register_t *retval) 1707 { 1708 struct sys_mknod_args /* { 1709 syscallarg(const char *) path; 1710 syscallarg(int) mode; 1711 syscallarg(int) dev; 1712 } */ *uap = v; 1713 struct proc *p = l->l_proc; 1714 struct vnode *vp; 1715 struct mount *mp; 1716 struct vattr vattr; 1717 int error; 1718 int whiteout = 0; 1719 struct nameidata nd; 1720 1721 if ((error = kauth_authorize_generic(l->l_cred, 1722 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1723 return (error); 1724 restart: 1725 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1726 if ((error = namei(&nd)) != 0) 1727 return (error); 1728 vp = nd.ni_vp; 1729 if (vp != NULL) 1730 error = EEXIST; 1731 else { 1732 VATTR_NULL(&vattr); 1733 vattr.va_mode = 1734 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1735 vattr.va_rdev = SCARG(uap, dev); 1736 whiteout = 0; 1737 1738 switch (SCARG(uap, mode) & S_IFMT) { 1739 case S_IFMT: /* used by badsect to flag bad sectors */ 1740 vattr.va_type = VBAD; 1741 break; 1742 case S_IFCHR: 1743 vattr.va_type = VCHR; 1744 break; 1745 case S_IFBLK: 1746 vattr.va_type = VBLK; 1747 break; 1748 case S_IFWHT: 1749 whiteout = 1; 1750 break; 1751 default: 1752 error = EINVAL; 1753 break; 1754 } 1755 } 1756 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1757 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1758 if (nd.ni_dvp == vp) 1759 vrele(nd.ni_dvp); 1760 else 1761 vput(nd.ni_dvp); 1762 if (vp) 1763 vrele(vp); 1764 if ((error = vn_start_write(NULL, &mp, 1765 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1766 return (error); 1767 goto restart; 1768 } 1769 if (!error) { 1770 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1771 if (whiteout) { 1772 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1773 if (error) 1774 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1775 vput(nd.ni_dvp); 1776 } else { 1777 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1778 &nd.ni_cnd, &vattr); 1779 if (error == 0) 1780 vput(nd.ni_vp); 1781 } 1782 } else { 1783 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1784 if (nd.ni_dvp == vp) 1785 vrele(nd.ni_dvp); 1786 else 1787 vput(nd.ni_dvp); 1788 if (vp) 1789 vrele(vp); 1790 } 1791 vn_finished_write(mp, 0); 1792 return (error); 1793 } 1794 1795 /* 1796 * Create a named pipe. 1797 */ 1798 /* ARGSUSED */ 1799 int 1800 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1801 { 1802 struct sys_mkfifo_args /* { 1803 syscallarg(const char *) path; 1804 syscallarg(int) mode; 1805 } */ *uap = v; 1806 struct proc *p = l->l_proc; 1807 struct mount *mp; 1808 struct vattr vattr; 1809 int error; 1810 struct nameidata nd; 1811 1812 restart: 1813 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1814 if ((error = namei(&nd)) != 0) 1815 return (error); 1816 if (nd.ni_vp != NULL) { 1817 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1818 if (nd.ni_dvp == nd.ni_vp) 1819 vrele(nd.ni_dvp); 1820 else 1821 vput(nd.ni_dvp); 1822 vrele(nd.ni_vp); 1823 return (EEXIST); 1824 } 1825 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1826 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1827 if (nd.ni_dvp == nd.ni_vp) 1828 vrele(nd.ni_dvp); 1829 else 1830 vput(nd.ni_dvp); 1831 if (nd.ni_vp) 1832 vrele(nd.ni_vp); 1833 if ((error = vn_start_write(NULL, &mp, 1834 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1835 return (error); 1836 goto restart; 1837 } 1838 VATTR_NULL(&vattr); 1839 vattr.va_type = VFIFO; 1840 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1841 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1842 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1843 if (error == 0) 1844 vput(nd.ni_vp); 1845 vn_finished_write(mp, 0); 1846 return (error); 1847 } 1848 1849 /* 1850 * Make a hard file link. 1851 */ 1852 /* ARGSUSED */ 1853 int 1854 sys_link(struct lwp *l, void *v, register_t *retval) 1855 { 1856 struct sys_link_args /* { 1857 syscallarg(const char *) path; 1858 syscallarg(const char *) link; 1859 } */ *uap = v; 1860 struct vnode *vp; 1861 struct mount *mp; 1862 struct nameidata nd; 1863 int error; 1864 1865 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1866 if ((error = namei(&nd)) != 0) 1867 return (error); 1868 vp = nd.ni_vp; 1869 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1870 vrele(vp); 1871 return (error); 1872 } 1873 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1874 if ((error = namei(&nd)) != 0) 1875 goto out; 1876 if (nd.ni_vp) { 1877 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1878 if (nd.ni_dvp == nd.ni_vp) 1879 vrele(nd.ni_dvp); 1880 else 1881 vput(nd.ni_dvp); 1882 vrele(nd.ni_vp); 1883 error = EEXIST; 1884 goto out; 1885 } 1886 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1887 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1888 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1889 out: 1890 vrele(vp); 1891 vn_finished_write(mp, 0); 1892 return (error); 1893 } 1894 1895 /* 1896 * Make a symbolic link. 1897 */ 1898 /* ARGSUSED */ 1899 int 1900 sys_symlink(struct lwp *l, void *v, register_t *retval) 1901 { 1902 struct sys_symlink_args /* { 1903 syscallarg(const char *) path; 1904 syscallarg(const char *) link; 1905 } */ *uap = v; 1906 struct proc *p = l->l_proc; 1907 struct mount *mp; 1908 struct vattr vattr; 1909 char *path; 1910 int error; 1911 struct nameidata nd; 1912 1913 path = PNBUF_GET(); 1914 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1915 if (error) 1916 goto out; 1917 restart: 1918 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1919 if ((error = namei(&nd)) != 0) 1920 goto out; 1921 if (nd.ni_vp) { 1922 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1923 if (nd.ni_dvp == nd.ni_vp) 1924 vrele(nd.ni_dvp); 1925 else 1926 vput(nd.ni_dvp); 1927 vrele(nd.ni_vp); 1928 error = EEXIST; 1929 goto out; 1930 } 1931 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1932 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1933 if (nd.ni_dvp == nd.ni_vp) 1934 vrele(nd.ni_dvp); 1935 else 1936 vput(nd.ni_dvp); 1937 if ((error = vn_start_write(NULL, &mp, 1938 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1939 return (error); 1940 goto restart; 1941 } 1942 VATTR_NULL(&vattr); 1943 vattr.va_type = VLNK; 1944 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1945 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1946 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1947 if (error == 0) 1948 vput(nd.ni_vp); 1949 vn_finished_write(mp, 0); 1950 out: 1951 PNBUF_PUT(path); 1952 return (error); 1953 } 1954 1955 /* 1956 * Delete a whiteout from the filesystem. 1957 */ 1958 /* ARGSUSED */ 1959 int 1960 sys_undelete(struct lwp *l, void *v, register_t *retval) 1961 { 1962 struct sys_undelete_args /* { 1963 syscallarg(const char *) path; 1964 } */ *uap = v; 1965 int error; 1966 struct mount *mp; 1967 struct nameidata nd; 1968 1969 restart: 1970 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1971 SCARG(uap, path), l); 1972 error = namei(&nd); 1973 if (error) 1974 return (error); 1975 1976 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1977 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1978 if (nd.ni_dvp == nd.ni_vp) 1979 vrele(nd.ni_dvp); 1980 else 1981 vput(nd.ni_dvp); 1982 if (nd.ni_vp) 1983 vrele(nd.ni_vp); 1984 return (EEXIST); 1985 } 1986 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1987 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1988 if (nd.ni_dvp == nd.ni_vp) 1989 vrele(nd.ni_dvp); 1990 else 1991 vput(nd.ni_dvp); 1992 if ((error = vn_start_write(NULL, &mp, 1993 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1994 return (error); 1995 goto restart; 1996 } 1997 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1998 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1999 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2000 vput(nd.ni_dvp); 2001 vn_finished_write(mp, 0); 2002 return (error); 2003 } 2004 2005 /* 2006 * Delete a name from the filesystem. 2007 */ 2008 /* ARGSUSED */ 2009 int 2010 sys_unlink(struct lwp *l, void *v, register_t *retval) 2011 { 2012 struct sys_unlink_args /* { 2013 syscallarg(const char *) path; 2014 } */ *uap = v; 2015 struct mount *mp; 2016 struct vnode *vp; 2017 int error; 2018 struct nameidata nd; 2019 2020 restart: 2021 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2022 SCARG(uap, path), l); 2023 if ((error = namei(&nd)) != 0) 2024 return (error); 2025 vp = nd.ni_vp; 2026 2027 /* 2028 * The root of a mounted filesystem cannot be deleted. 2029 */ 2030 if (vp->v_flag & VROOT) { 2031 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2032 if (nd.ni_dvp == vp) 2033 vrele(nd.ni_dvp); 2034 else 2035 vput(nd.ni_dvp); 2036 vput(vp); 2037 error = EBUSY; 2038 goto out; 2039 } 2040 2041 #if NVERIEXEC > 0 2042 /* Handle remove requests for veriexec entries. */ 2043 if ((error = veriexec_removechk(vp, nd.ni_dirp, l)) != 0) { 2044 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2045 if (nd.ni_dvp == vp) 2046 vrele(nd.ni_dvp); 2047 else 2048 vput(nd.ni_dvp); 2049 vput(vp); 2050 goto out; 2051 } 2052 #endif /* NVERIEXEC > 0 */ 2053 2054 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2055 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2056 if (nd.ni_dvp == vp) 2057 vrele(nd.ni_dvp); 2058 else 2059 vput(nd.ni_dvp); 2060 vput(vp); 2061 if ((error = vn_start_write(NULL, &mp, 2062 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2063 return (error); 2064 goto restart; 2065 } 2066 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2067 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2068 #ifdef FILEASSOC 2069 (void)fileassoc_file_delete(nd.ni_vp); 2070 #endif /* FILEASSOC */ 2071 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2072 vn_finished_write(mp, 0); 2073 out: 2074 return (error); 2075 } 2076 2077 /* 2078 * Reposition read/write file offset. 2079 */ 2080 int 2081 sys_lseek(struct lwp *l, void *v, register_t *retval) 2082 { 2083 struct sys_lseek_args /* { 2084 syscallarg(int) fd; 2085 syscallarg(int) pad; 2086 syscallarg(off_t) offset; 2087 syscallarg(int) whence; 2088 } */ *uap = v; 2089 struct proc *p = l->l_proc; 2090 kauth_cred_t cred = l->l_cred; 2091 struct filedesc *fdp = p->p_fd; 2092 struct file *fp; 2093 struct vnode *vp; 2094 struct vattr vattr; 2095 off_t newoff; 2096 int error; 2097 2098 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2099 return (EBADF); 2100 2101 FILE_USE(fp); 2102 2103 vp = (struct vnode *)fp->f_data; 2104 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2105 error = ESPIPE; 2106 goto out; 2107 } 2108 2109 switch (SCARG(uap, whence)) { 2110 case SEEK_CUR: 2111 newoff = fp->f_offset + SCARG(uap, offset); 2112 break; 2113 case SEEK_END: 2114 error = VOP_GETATTR(vp, &vattr, cred, l); 2115 if (error) 2116 goto out; 2117 newoff = SCARG(uap, offset) + vattr.va_size; 2118 break; 2119 case SEEK_SET: 2120 newoff = SCARG(uap, offset); 2121 break; 2122 default: 2123 error = EINVAL; 2124 goto out; 2125 } 2126 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2127 goto out; 2128 2129 *(off_t *)retval = fp->f_offset = newoff; 2130 out: 2131 FILE_UNUSE(fp, l); 2132 return (error); 2133 } 2134 2135 /* 2136 * Positional read system call. 2137 */ 2138 int 2139 sys_pread(struct lwp *l, void *v, register_t *retval) 2140 { 2141 struct sys_pread_args /* { 2142 syscallarg(int) fd; 2143 syscallarg(void *) buf; 2144 syscallarg(size_t) nbyte; 2145 syscallarg(off_t) offset; 2146 } */ *uap = v; 2147 struct proc *p = l->l_proc; 2148 struct filedesc *fdp = p->p_fd; 2149 struct file *fp; 2150 struct vnode *vp; 2151 off_t offset; 2152 int error, fd = SCARG(uap, fd); 2153 2154 if ((fp = fd_getfile(fdp, fd)) == NULL) 2155 return (EBADF); 2156 2157 if ((fp->f_flag & FREAD) == 0) { 2158 simple_unlock(&fp->f_slock); 2159 return (EBADF); 2160 } 2161 2162 FILE_USE(fp); 2163 2164 vp = (struct vnode *)fp->f_data; 2165 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2166 error = ESPIPE; 2167 goto out; 2168 } 2169 2170 offset = SCARG(uap, offset); 2171 2172 /* 2173 * XXX This works because no file systems actually 2174 * XXX take any action on the seek operation. 2175 */ 2176 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2177 goto out; 2178 2179 /* dofileread() will unuse the descriptor for us */ 2180 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2181 &offset, 0, retval)); 2182 2183 out: 2184 FILE_UNUSE(fp, l); 2185 return (error); 2186 } 2187 2188 /* 2189 * Positional scatter read system call. 2190 */ 2191 int 2192 sys_preadv(struct lwp *l, void *v, register_t *retval) 2193 { 2194 struct sys_preadv_args /* { 2195 syscallarg(int) fd; 2196 syscallarg(const struct iovec *) iovp; 2197 syscallarg(int) iovcnt; 2198 syscallarg(off_t) offset; 2199 } */ *uap = v; 2200 struct proc *p = l->l_proc; 2201 struct filedesc *fdp = p->p_fd; 2202 struct file *fp; 2203 struct vnode *vp; 2204 off_t offset; 2205 int error, fd = SCARG(uap, fd); 2206 2207 if ((fp = fd_getfile(fdp, fd)) == NULL) 2208 return (EBADF); 2209 2210 if ((fp->f_flag & FREAD) == 0) { 2211 simple_unlock(&fp->f_slock); 2212 return (EBADF); 2213 } 2214 2215 FILE_USE(fp); 2216 2217 vp = (struct vnode *)fp->f_data; 2218 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2219 error = ESPIPE; 2220 goto out; 2221 } 2222 2223 offset = SCARG(uap, offset); 2224 2225 /* 2226 * XXX This works because no file systems actually 2227 * XXX take any action on the seek operation. 2228 */ 2229 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2230 goto out; 2231 2232 /* dofilereadv() will unuse the descriptor for us */ 2233 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2234 &offset, 0, retval)); 2235 2236 out: 2237 FILE_UNUSE(fp, l); 2238 return (error); 2239 } 2240 2241 /* 2242 * Positional write system call. 2243 */ 2244 int 2245 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2246 { 2247 struct sys_pwrite_args /* { 2248 syscallarg(int) fd; 2249 syscallarg(const void *) buf; 2250 syscallarg(size_t) nbyte; 2251 syscallarg(off_t) offset; 2252 } */ *uap = v; 2253 struct proc *p = l->l_proc; 2254 struct filedesc *fdp = p->p_fd; 2255 struct file *fp; 2256 struct vnode *vp; 2257 off_t offset; 2258 int error, fd = SCARG(uap, fd); 2259 2260 if ((fp = fd_getfile(fdp, fd)) == NULL) 2261 return (EBADF); 2262 2263 if ((fp->f_flag & FWRITE) == 0) { 2264 simple_unlock(&fp->f_slock); 2265 return (EBADF); 2266 } 2267 2268 FILE_USE(fp); 2269 2270 vp = (struct vnode *)fp->f_data; 2271 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2272 error = ESPIPE; 2273 goto out; 2274 } 2275 2276 offset = SCARG(uap, offset); 2277 2278 /* 2279 * XXX This works because no file systems actually 2280 * XXX take any action on the seek operation. 2281 */ 2282 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2283 goto out; 2284 2285 /* dofilewrite() will unuse the descriptor for us */ 2286 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2287 &offset, 0, retval)); 2288 2289 out: 2290 FILE_UNUSE(fp, l); 2291 return (error); 2292 } 2293 2294 /* 2295 * Positional gather write system call. 2296 */ 2297 int 2298 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2299 { 2300 struct sys_pwritev_args /* { 2301 syscallarg(int) fd; 2302 syscallarg(const struct iovec *) iovp; 2303 syscallarg(int) iovcnt; 2304 syscallarg(off_t) offset; 2305 } */ *uap = v; 2306 struct proc *p = l->l_proc; 2307 struct filedesc *fdp = p->p_fd; 2308 struct file *fp; 2309 struct vnode *vp; 2310 off_t offset; 2311 int error, fd = SCARG(uap, fd); 2312 2313 if ((fp = fd_getfile(fdp, fd)) == NULL) 2314 return (EBADF); 2315 2316 if ((fp->f_flag & FWRITE) == 0) { 2317 simple_unlock(&fp->f_slock); 2318 return (EBADF); 2319 } 2320 2321 FILE_USE(fp); 2322 2323 vp = (struct vnode *)fp->f_data; 2324 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2325 error = ESPIPE; 2326 goto out; 2327 } 2328 2329 offset = SCARG(uap, offset); 2330 2331 /* 2332 * XXX This works because no file systems actually 2333 * XXX take any action on the seek operation. 2334 */ 2335 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2336 goto out; 2337 2338 /* dofilewritev() will unuse the descriptor for us */ 2339 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2340 &offset, 0, retval)); 2341 2342 out: 2343 FILE_UNUSE(fp, l); 2344 return (error); 2345 } 2346 2347 /* 2348 * Check access permissions. 2349 */ 2350 int 2351 sys_access(struct lwp *l, void *v, register_t *retval) 2352 { 2353 struct sys_access_args /* { 2354 syscallarg(const char *) path; 2355 syscallarg(int) flags; 2356 } */ *uap = v; 2357 kauth_cred_t cred; 2358 struct vnode *vp; 2359 int error, flags; 2360 struct nameidata nd; 2361 2362 cred = kauth_cred_dup(l->l_cred); 2363 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2364 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2365 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2366 SCARG(uap, path), l); 2367 /* Override default credentials */ 2368 nd.ni_cnd.cn_cred = cred; 2369 if ((error = namei(&nd)) != 0) 2370 goto out; 2371 vp = nd.ni_vp; 2372 2373 /* Flags == 0 means only check for existence. */ 2374 if (SCARG(uap, flags)) { 2375 flags = 0; 2376 if (SCARG(uap, flags) & R_OK) 2377 flags |= VREAD; 2378 if (SCARG(uap, flags) & W_OK) 2379 flags |= VWRITE; 2380 if (SCARG(uap, flags) & X_OK) 2381 flags |= VEXEC; 2382 2383 error = VOP_ACCESS(vp, flags, cred, l); 2384 if (!error && (flags & VWRITE)) 2385 error = vn_writechk(vp); 2386 } 2387 vput(vp); 2388 out: 2389 kauth_cred_free(cred); 2390 return (error); 2391 } 2392 2393 /* 2394 * Get file status; this version follows links. 2395 */ 2396 /* ARGSUSED */ 2397 int 2398 sys___stat30(struct lwp *l, void *v, register_t *retval) 2399 { 2400 struct sys___stat30_args /* { 2401 syscallarg(const char *) path; 2402 syscallarg(struct stat *) ub; 2403 } */ *uap = v; 2404 struct stat sb; 2405 int error; 2406 struct nameidata nd; 2407 2408 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2409 SCARG(uap, path), l); 2410 if ((error = namei(&nd)) != 0) 2411 return (error); 2412 error = vn_stat(nd.ni_vp, &sb, l); 2413 vput(nd.ni_vp); 2414 if (error) 2415 return (error); 2416 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2417 return (error); 2418 } 2419 2420 /* 2421 * Get file status; this version does not follow links. 2422 */ 2423 /* ARGSUSED */ 2424 int 2425 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2426 { 2427 struct sys___lstat30_args /* { 2428 syscallarg(const char *) path; 2429 syscallarg(struct stat *) ub; 2430 } */ *uap = v; 2431 struct stat sb; 2432 int error; 2433 struct nameidata nd; 2434 2435 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2436 SCARG(uap, path), l); 2437 if ((error = namei(&nd)) != 0) 2438 return (error); 2439 error = vn_stat(nd.ni_vp, &sb, l); 2440 vput(nd.ni_vp); 2441 if (error) 2442 return (error); 2443 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2444 return (error); 2445 } 2446 2447 /* 2448 * Get configurable pathname variables. 2449 */ 2450 /* ARGSUSED */ 2451 int 2452 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2453 { 2454 struct sys_pathconf_args /* { 2455 syscallarg(const char *) path; 2456 syscallarg(int) name; 2457 } */ *uap = v; 2458 int error; 2459 struct nameidata nd; 2460 2461 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2462 SCARG(uap, path), l); 2463 if ((error = namei(&nd)) != 0) 2464 return (error); 2465 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2466 vput(nd.ni_vp); 2467 return (error); 2468 } 2469 2470 /* 2471 * Return target name of a symbolic link. 2472 */ 2473 /* ARGSUSED */ 2474 int 2475 sys_readlink(struct lwp *l, void *v, register_t *retval) 2476 { 2477 struct sys_readlink_args /* { 2478 syscallarg(const char *) path; 2479 syscallarg(char *) buf; 2480 syscallarg(size_t) count; 2481 } */ *uap = v; 2482 struct vnode *vp; 2483 struct iovec aiov; 2484 struct uio auio; 2485 int error; 2486 struct nameidata nd; 2487 2488 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2489 SCARG(uap, path), l); 2490 if ((error = namei(&nd)) != 0) 2491 return (error); 2492 vp = nd.ni_vp; 2493 if (vp->v_type != VLNK) 2494 error = EINVAL; 2495 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2496 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2497 aiov.iov_base = SCARG(uap, buf); 2498 aiov.iov_len = SCARG(uap, count); 2499 auio.uio_iov = &aiov; 2500 auio.uio_iovcnt = 1; 2501 auio.uio_offset = 0; 2502 auio.uio_rw = UIO_READ; 2503 KASSERT(l == curlwp); 2504 auio.uio_vmspace = l->l_proc->p_vmspace; 2505 auio.uio_resid = SCARG(uap, count); 2506 error = VOP_READLINK(vp, &auio, l->l_cred); 2507 } 2508 vput(vp); 2509 *retval = SCARG(uap, count) - auio.uio_resid; 2510 return (error); 2511 } 2512 2513 /* 2514 * Change flags of a file given a path name. 2515 */ 2516 /* ARGSUSED */ 2517 int 2518 sys_chflags(struct lwp *l, void *v, register_t *retval) 2519 { 2520 struct sys_chflags_args /* { 2521 syscallarg(const char *) path; 2522 syscallarg(u_long) flags; 2523 } */ *uap = v; 2524 struct vnode *vp; 2525 int error; 2526 struct nameidata nd; 2527 2528 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2529 if ((error = namei(&nd)) != 0) 2530 return (error); 2531 vp = nd.ni_vp; 2532 error = change_flags(vp, SCARG(uap, flags), l); 2533 vput(vp); 2534 return (error); 2535 } 2536 2537 /* 2538 * Change flags of a file given a file descriptor. 2539 */ 2540 /* ARGSUSED */ 2541 int 2542 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2543 { 2544 struct sys_fchflags_args /* { 2545 syscallarg(int) fd; 2546 syscallarg(u_long) flags; 2547 } */ *uap = v; 2548 struct proc *p = l->l_proc; 2549 struct vnode *vp; 2550 struct file *fp; 2551 int error; 2552 2553 /* getvnode() will use the descriptor for us */ 2554 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2555 return (error); 2556 vp = (struct vnode *)fp->f_data; 2557 error = change_flags(vp, SCARG(uap, flags), l); 2558 VOP_UNLOCK(vp, 0); 2559 FILE_UNUSE(fp, l); 2560 return (error); 2561 } 2562 2563 /* 2564 * Change flags of a file given a path name; this version does 2565 * not follow links. 2566 */ 2567 int 2568 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2569 { 2570 struct sys_lchflags_args /* { 2571 syscallarg(const char *) path; 2572 syscallarg(u_long) flags; 2573 } */ *uap = v; 2574 struct vnode *vp; 2575 int error; 2576 struct nameidata nd; 2577 2578 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2579 if ((error = namei(&nd)) != 0) 2580 return (error); 2581 vp = nd.ni_vp; 2582 error = change_flags(vp, SCARG(uap, flags), l); 2583 vput(vp); 2584 return (error); 2585 } 2586 2587 /* 2588 * Common routine to change flags of a file. 2589 */ 2590 int 2591 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2592 { 2593 struct mount *mp; 2594 struct vattr vattr; 2595 int error; 2596 2597 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2598 return (error); 2599 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2600 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2601 /* 2602 * Non-superusers cannot change the flags on devices, even if they 2603 * own them. 2604 */ 2605 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2606 &l->l_acflag) != 0) { 2607 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2608 goto out; 2609 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2610 error = EINVAL; 2611 goto out; 2612 } 2613 } 2614 VATTR_NULL(&vattr); 2615 vattr.va_flags = flags; 2616 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2617 out: 2618 vn_finished_write(mp, 0); 2619 return (error); 2620 } 2621 2622 /* 2623 * Change mode of a file given path name; this version follows links. 2624 */ 2625 /* ARGSUSED */ 2626 int 2627 sys_chmod(struct lwp *l, void *v, register_t *retval) 2628 { 2629 struct sys_chmod_args /* { 2630 syscallarg(const char *) path; 2631 syscallarg(int) mode; 2632 } */ *uap = v; 2633 int error; 2634 struct nameidata nd; 2635 2636 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2637 if ((error = namei(&nd)) != 0) 2638 return (error); 2639 2640 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2641 2642 vrele(nd.ni_vp); 2643 return (error); 2644 } 2645 2646 /* 2647 * Change mode of a file given a file descriptor. 2648 */ 2649 /* ARGSUSED */ 2650 int 2651 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2652 { 2653 struct sys_fchmod_args /* { 2654 syscallarg(int) fd; 2655 syscallarg(int) mode; 2656 } */ *uap = v; 2657 struct proc *p = l->l_proc; 2658 struct file *fp; 2659 int error; 2660 2661 /* getvnode() will use the descriptor for us */ 2662 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2663 return (error); 2664 2665 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2666 FILE_UNUSE(fp, l); 2667 return (error); 2668 } 2669 2670 /* 2671 * Change mode of a file given path name; this version does not follow links. 2672 */ 2673 /* ARGSUSED */ 2674 int 2675 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2676 { 2677 struct sys_lchmod_args /* { 2678 syscallarg(const char *) path; 2679 syscallarg(int) mode; 2680 } */ *uap = v; 2681 int error; 2682 struct nameidata nd; 2683 2684 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2685 if ((error = namei(&nd)) != 0) 2686 return (error); 2687 2688 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2689 2690 vrele(nd.ni_vp); 2691 return (error); 2692 } 2693 2694 /* 2695 * Common routine to set mode given a vnode. 2696 */ 2697 static int 2698 change_mode(struct vnode *vp, int mode, struct lwp *l) 2699 { 2700 struct mount *mp; 2701 struct vattr vattr; 2702 int error; 2703 2704 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2705 return (error); 2706 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2707 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2708 VATTR_NULL(&vattr); 2709 vattr.va_mode = mode & ALLPERMS; 2710 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2711 VOP_UNLOCK(vp, 0); 2712 vn_finished_write(mp, 0); 2713 return (error); 2714 } 2715 2716 /* 2717 * Set ownership given a path name; this version follows links. 2718 */ 2719 /* ARGSUSED */ 2720 int 2721 sys_chown(struct lwp *l, void *v, register_t *retval) 2722 { 2723 struct sys_chown_args /* { 2724 syscallarg(const char *) path; 2725 syscallarg(uid_t) uid; 2726 syscallarg(gid_t) gid; 2727 } */ *uap = v; 2728 int error; 2729 struct nameidata nd; 2730 2731 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2732 if ((error = namei(&nd)) != 0) 2733 return (error); 2734 2735 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2736 2737 vrele(nd.ni_vp); 2738 return (error); 2739 } 2740 2741 /* 2742 * Set ownership given a path name; this version follows links. 2743 * Provides POSIX semantics. 2744 */ 2745 /* ARGSUSED */ 2746 int 2747 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2748 { 2749 struct sys_chown_args /* { 2750 syscallarg(const char *) path; 2751 syscallarg(uid_t) uid; 2752 syscallarg(gid_t) gid; 2753 } */ *uap = v; 2754 int error; 2755 struct nameidata nd; 2756 2757 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2758 if ((error = namei(&nd)) != 0) 2759 return (error); 2760 2761 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2762 2763 vrele(nd.ni_vp); 2764 return (error); 2765 } 2766 2767 /* 2768 * Set ownership given a file descriptor. 2769 */ 2770 /* ARGSUSED */ 2771 int 2772 sys_fchown(struct lwp *l, void *v, register_t *retval) 2773 { 2774 struct sys_fchown_args /* { 2775 syscallarg(int) fd; 2776 syscallarg(uid_t) uid; 2777 syscallarg(gid_t) gid; 2778 } */ *uap = v; 2779 struct proc *p = l->l_proc; 2780 int error; 2781 struct file *fp; 2782 2783 /* getvnode() will use the descriptor for us */ 2784 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2785 return (error); 2786 2787 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2788 SCARG(uap, gid), l, 0); 2789 FILE_UNUSE(fp, l); 2790 return (error); 2791 } 2792 2793 /* 2794 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2795 */ 2796 /* ARGSUSED */ 2797 int 2798 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2799 { 2800 struct sys_fchown_args /* { 2801 syscallarg(int) fd; 2802 syscallarg(uid_t) uid; 2803 syscallarg(gid_t) gid; 2804 } */ *uap = v; 2805 struct proc *p = l->l_proc; 2806 int error; 2807 struct file *fp; 2808 2809 /* getvnode() will use the descriptor for us */ 2810 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2811 return (error); 2812 2813 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2814 SCARG(uap, gid), l, 1); 2815 FILE_UNUSE(fp, l); 2816 return (error); 2817 } 2818 2819 /* 2820 * Set ownership given a path name; this version does not follow links. 2821 */ 2822 /* ARGSUSED */ 2823 int 2824 sys_lchown(struct lwp *l, void *v, register_t *retval) 2825 { 2826 struct sys_lchown_args /* { 2827 syscallarg(const char *) path; 2828 syscallarg(uid_t) uid; 2829 syscallarg(gid_t) gid; 2830 } */ *uap = v; 2831 int error; 2832 struct nameidata nd; 2833 2834 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2835 if ((error = namei(&nd)) != 0) 2836 return (error); 2837 2838 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2839 2840 vrele(nd.ni_vp); 2841 return (error); 2842 } 2843 2844 /* 2845 * Set ownership given a path name; this version does not follow links. 2846 * Provides POSIX/XPG semantics. 2847 */ 2848 /* ARGSUSED */ 2849 int 2850 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2851 { 2852 struct sys_lchown_args /* { 2853 syscallarg(const char *) path; 2854 syscallarg(uid_t) uid; 2855 syscallarg(gid_t) gid; 2856 } */ *uap = v; 2857 int error; 2858 struct nameidata nd; 2859 2860 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2861 if ((error = namei(&nd)) != 0) 2862 return (error); 2863 2864 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2865 2866 vrele(nd.ni_vp); 2867 return (error); 2868 } 2869 2870 /* 2871 * Common routine to set ownership given a vnode. 2872 */ 2873 static int 2874 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2875 int posix_semantics) 2876 { 2877 struct mount *mp; 2878 struct vattr vattr; 2879 mode_t newmode; 2880 int error; 2881 2882 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2883 return (error); 2884 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2885 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2886 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2887 goto out; 2888 2889 #define CHANGED(x) ((int)(x) != -1) 2890 newmode = vattr.va_mode; 2891 if (posix_semantics) { 2892 /* 2893 * POSIX/XPG semantics: if the caller is not the super-user, 2894 * clear set-user-id and set-group-id bits. Both POSIX and 2895 * the XPG consider the behaviour for calls by the super-user 2896 * implementation-defined; we leave the set-user-id and set- 2897 * group-id settings intact in that case. 2898 */ 2899 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2900 NULL) != 0) 2901 newmode &= ~(S_ISUID | S_ISGID); 2902 } else { 2903 /* 2904 * NetBSD semantics: when changing owner and/or group, 2905 * clear the respective bit(s). 2906 */ 2907 if (CHANGED(uid)) 2908 newmode &= ~S_ISUID; 2909 if (CHANGED(gid)) 2910 newmode &= ~S_ISGID; 2911 } 2912 /* Update va_mode iff altered. */ 2913 if (vattr.va_mode == newmode) 2914 newmode = VNOVAL; 2915 2916 VATTR_NULL(&vattr); 2917 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2918 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2919 vattr.va_mode = newmode; 2920 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2921 #undef CHANGED 2922 2923 out: 2924 VOP_UNLOCK(vp, 0); 2925 vn_finished_write(mp, 0); 2926 return (error); 2927 } 2928 2929 /* 2930 * Set the access and modification times given a path name; this 2931 * version follows links. 2932 */ 2933 /* ARGSUSED */ 2934 int 2935 sys_utimes(struct lwp *l, void *v, register_t *retval) 2936 { 2937 struct sys_utimes_args /* { 2938 syscallarg(const char *) path; 2939 syscallarg(const struct timeval *) tptr; 2940 } */ *uap = v; 2941 int error; 2942 struct nameidata nd; 2943 2944 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2945 if ((error = namei(&nd)) != 0) 2946 return (error); 2947 2948 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2949 2950 vrele(nd.ni_vp); 2951 return (error); 2952 } 2953 2954 /* 2955 * Set the access and modification times given a file descriptor. 2956 */ 2957 /* ARGSUSED */ 2958 int 2959 sys_futimes(struct lwp *l, void *v, register_t *retval) 2960 { 2961 struct sys_futimes_args /* { 2962 syscallarg(int) fd; 2963 syscallarg(const struct timeval *) tptr; 2964 } */ *uap = v; 2965 struct proc *p = l->l_proc; 2966 int error; 2967 struct file *fp; 2968 2969 /* getvnode() will use the descriptor for us */ 2970 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2971 return (error); 2972 2973 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2974 FILE_UNUSE(fp, l); 2975 return (error); 2976 } 2977 2978 /* 2979 * Set the access and modification times given a path name; this 2980 * version does not follow links. 2981 */ 2982 /* ARGSUSED */ 2983 int 2984 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2985 { 2986 struct sys_lutimes_args /* { 2987 syscallarg(const char *) path; 2988 syscallarg(const struct timeval *) tptr; 2989 } */ *uap = v; 2990 int error; 2991 struct nameidata nd; 2992 2993 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2994 if ((error = namei(&nd)) != 0) 2995 return (error); 2996 2997 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2998 2999 vrele(nd.ni_vp); 3000 return (error); 3001 } 3002 3003 /* 3004 * Common routine to set access and modification times given a vnode. 3005 */ 3006 static int 3007 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3008 { 3009 struct mount *mp; 3010 struct vattr vattr; 3011 int error; 3012 3013 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3014 return (error); 3015 VATTR_NULL(&vattr); 3016 if (tptr == NULL) { 3017 nanotime(&vattr.va_atime); 3018 vattr.va_mtime = vattr.va_atime; 3019 vattr.va_vaflags |= VA_UTIMES_NULL; 3020 } else { 3021 struct timeval tv[2]; 3022 3023 error = copyin(tptr, tv, sizeof(tv)); 3024 if (error) 3025 goto out; 3026 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3027 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3028 } 3029 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3030 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3031 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3032 VOP_UNLOCK(vp, 0); 3033 out: 3034 vn_finished_write(mp, 0); 3035 return (error); 3036 } 3037 3038 /* 3039 * Truncate a file given its path name. 3040 */ 3041 /* ARGSUSED */ 3042 int 3043 sys_truncate(struct lwp *l, void *v, register_t *retval) 3044 { 3045 struct sys_truncate_args /* { 3046 syscallarg(const char *) path; 3047 syscallarg(int) pad; 3048 syscallarg(off_t) length; 3049 } */ *uap = v; 3050 struct vnode *vp; 3051 struct mount *mp; 3052 struct vattr vattr; 3053 int error; 3054 struct nameidata nd; 3055 3056 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3057 if ((error = namei(&nd)) != 0) 3058 return (error); 3059 vp = nd.ni_vp; 3060 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3061 vrele(vp); 3062 return (error); 3063 } 3064 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3065 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3066 if (vp->v_type == VDIR) 3067 error = EISDIR; 3068 else if ((error = vn_writechk(vp)) == 0 && 3069 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3070 VATTR_NULL(&vattr); 3071 vattr.va_size = SCARG(uap, length); 3072 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3073 } 3074 vput(vp); 3075 vn_finished_write(mp, 0); 3076 return (error); 3077 } 3078 3079 /* 3080 * Truncate a file given a file descriptor. 3081 */ 3082 /* ARGSUSED */ 3083 int 3084 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3085 { 3086 struct sys_ftruncate_args /* { 3087 syscallarg(int) fd; 3088 syscallarg(int) pad; 3089 syscallarg(off_t) length; 3090 } */ *uap = v; 3091 struct proc *p = l->l_proc; 3092 struct mount *mp; 3093 struct vattr vattr; 3094 struct vnode *vp; 3095 struct file *fp; 3096 int error; 3097 3098 /* getvnode() will use the descriptor for us */ 3099 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3100 return (error); 3101 if ((fp->f_flag & FWRITE) == 0) { 3102 error = EINVAL; 3103 goto out; 3104 } 3105 vp = (struct vnode *)fp->f_data; 3106 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3107 FILE_UNUSE(fp, l); 3108 return (error); 3109 } 3110 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3111 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3112 if (vp->v_type == VDIR) 3113 error = EISDIR; 3114 else if ((error = vn_writechk(vp)) == 0) { 3115 VATTR_NULL(&vattr); 3116 vattr.va_size = SCARG(uap, length); 3117 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3118 } 3119 VOP_UNLOCK(vp, 0); 3120 vn_finished_write(mp, 0); 3121 out: 3122 FILE_UNUSE(fp, l); 3123 return (error); 3124 } 3125 3126 /* 3127 * Sync an open file. 3128 */ 3129 /* ARGSUSED */ 3130 int 3131 sys_fsync(struct lwp *l, void *v, register_t *retval) 3132 { 3133 struct sys_fsync_args /* { 3134 syscallarg(int) fd; 3135 } */ *uap = v; 3136 struct proc *p = l->l_proc; 3137 struct vnode *vp; 3138 struct mount *mp; 3139 struct file *fp; 3140 int error; 3141 3142 /* getvnode() will use the descriptor for us */ 3143 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3144 return (error); 3145 vp = (struct vnode *)fp->f_data; 3146 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3147 FILE_UNUSE(fp, l); 3148 return (error); 3149 } 3150 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3151 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3152 if (error == 0 && bioops.io_fsync != NULL && 3153 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3154 (*bioops.io_fsync)(vp, 0); 3155 VOP_UNLOCK(vp, 0); 3156 vn_finished_write(mp, 0); 3157 FILE_UNUSE(fp, l); 3158 return (error); 3159 } 3160 3161 /* 3162 * Sync a range of file data. API modeled after that found in AIX. 3163 * 3164 * FDATASYNC indicates that we need only save enough metadata to be able 3165 * to re-read the written data. Note we duplicate AIX's requirement that 3166 * the file be open for writing. 3167 */ 3168 /* ARGSUSED */ 3169 int 3170 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3171 { 3172 struct sys_fsync_range_args /* { 3173 syscallarg(int) fd; 3174 syscallarg(int) flags; 3175 syscallarg(off_t) start; 3176 syscallarg(off_t) length; 3177 } */ *uap = v; 3178 struct proc *p = l->l_proc; 3179 struct vnode *vp; 3180 struct file *fp; 3181 int flags, nflags; 3182 off_t s, e, len; 3183 int error; 3184 3185 /* getvnode() will use the descriptor for us */ 3186 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3187 return (error); 3188 3189 if ((fp->f_flag & FWRITE) == 0) { 3190 FILE_UNUSE(fp, l); 3191 return (EBADF); 3192 } 3193 3194 flags = SCARG(uap, flags); 3195 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3196 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3197 return (EINVAL); 3198 } 3199 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3200 if (flags & FDATASYNC) 3201 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3202 else 3203 nflags = FSYNC_WAIT; 3204 if (flags & FDISKSYNC) 3205 nflags |= FSYNC_CACHE; 3206 3207 len = SCARG(uap, length); 3208 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3209 if (len) { 3210 s = SCARG(uap, start); 3211 e = s + len; 3212 if (e < s) { 3213 FILE_UNUSE(fp, l); 3214 return (EINVAL); 3215 } 3216 } else { 3217 e = 0; 3218 s = 0; 3219 } 3220 3221 vp = (struct vnode *)fp->f_data; 3222 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3223 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3224 3225 if (error == 0 && bioops.io_fsync != NULL && 3226 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3227 (*bioops.io_fsync)(vp, nflags); 3228 3229 VOP_UNLOCK(vp, 0); 3230 FILE_UNUSE(fp, l); 3231 return (error); 3232 } 3233 3234 /* 3235 * Sync the data of an open file. 3236 */ 3237 /* ARGSUSED */ 3238 int 3239 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3240 { 3241 struct sys_fdatasync_args /* { 3242 syscallarg(int) fd; 3243 } */ *uap = v; 3244 struct proc *p = l->l_proc; 3245 struct vnode *vp; 3246 struct file *fp; 3247 int error; 3248 3249 /* getvnode() will use the descriptor for us */ 3250 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3251 return (error); 3252 if ((fp->f_flag & FWRITE) == 0) { 3253 FILE_UNUSE(fp, l); 3254 return (EBADF); 3255 } 3256 vp = (struct vnode *)fp->f_data; 3257 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3258 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3259 VOP_UNLOCK(vp, 0); 3260 FILE_UNUSE(fp, l); 3261 return (error); 3262 } 3263 3264 /* 3265 * Rename files, (standard) BSD semantics frontend. 3266 */ 3267 /* ARGSUSED */ 3268 int 3269 sys_rename(struct lwp *l, void *v, register_t *retval) 3270 { 3271 struct sys_rename_args /* { 3272 syscallarg(const char *) from; 3273 syscallarg(const char *) to; 3274 } */ *uap = v; 3275 3276 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3277 } 3278 3279 /* 3280 * Rename files, POSIX semantics frontend. 3281 */ 3282 /* ARGSUSED */ 3283 int 3284 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3285 { 3286 struct sys___posix_rename_args /* { 3287 syscallarg(const char *) from; 3288 syscallarg(const char *) to; 3289 } */ *uap = v; 3290 3291 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3292 } 3293 3294 /* 3295 * Rename files. Source and destination must either both be directories, 3296 * or both not be directories. If target is a directory, it must be empty. 3297 * If `from' and `to' refer to the same object, the value of the `retain' 3298 * argument is used to determine whether `from' will be 3299 * 3300 * (retain == 0) deleted unless `from' and `to' refer to the same 3301 * object in the file system's name space (BSD). 3302 * (retain == 1) always retained (POSIX). 3303 */ 3304 static int 3305 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3306 { 3307 struct mount *mp = NULL; 3308 struct vnode *tvp, *fvp, *tdvp; 3309 struct nameidata fromnd, tond; 3310 struct proc *p; 3311 int error; 3312 3313 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3314 from, l); 3315 if ((error = namei(&fromnd)) != 0) 3316 return (error); 3317 fvp = fromnd.ni_vp; 3318 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3319 if (error != 0) { 3320 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3321 vrele(fromnd.ni_dvp); 3322 vrele(fvp); 3323 if (fromnd.ni_startdir) 3324 vrele(fromnd.ni_startdir); 3325 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3326 return (error); 3327 } 3328 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3329 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3330 if ((error = namei(&tond)) != 0) { 3331 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3332 vrele(fromnd.ni_dvp); 3333 vrele(fvp); 3334 goto out1; 3335 } 3336 tdvp = tond.ni_dvp; 3337 tvp = tond.ni_vp; 3338 3339 if (tvp != NULL) { 3340 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3341 error = ENOTDIR; 3342 goto out; 3343 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3344 error = EISDIR; 3345 goto out; 3346 } 3347 } 3348 3349 if (fvp == tdvp) 3350 error = EINVAL; 3351 3352 /* 3353 * Source and destination refer to the same object. 3354 */ 3355 if (fvp == tvp) { 3356 if (retain) 3357 error = -1; 3358 else if (fromnd.ni_dvp == tdvp && 3359 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3360 !memcmp(fromnd.ni_cnd.cn_nameptr, 3361 tond.ni_cnd.cn_nameptr, 3362 fromnd.ni_cnd.cn_namelen)) 3363 error = -1; 3364 } 3365 3366 #if NVERIEXEC > 0 3367 if (!error) 3368 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tvp, 3369 tond.ni_dirp, l); 3370 #endif /* NVERIEXEC > 0 */ 3371 3372 out: 3373 p = l->l_proc; 3374 if (!error) { 3375 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3376 if (fromnd.ni_dvp != tdvp) 3377 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3378 if (tvp) { 3379 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3380 } 3381 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3382 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3383 } else { 3384 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3385 if (tdvp == tvp) 3386 vrele(tdvp); 3387 else 3388 vput(tdvp); 3389 if (tvp) 3390 vput(tvp); 3391 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3392 vrele(fromnd.ni_dvp); 3393 vrele(fvp); 3394 } 3395 vrele(tond.ni_startdir); 3396 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3397 out1: 3398 vn_finished_write(mp, 0); 3399 if (fromnd.ni_startdir) 3400 vrele(fromnd.ni_startdir); 3401 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3402 return (error == -1 ? 0 : error); 3403 } 3404 3405 /* 3406 * Make a directory file. 3407 */ 3408 /* ARGSUSED */ 3409 int 3410 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3411 { 3412 struct sys_mkdir_args /* { 3413 syscallarg(const char *) path; 3414 syscallarg(int) mode; 3415 } */ *uap = v; 3416 struct proc *p = l->l_proc; 3417 struct mount *mp; 3418 struct vnode *vp; 3419 struct vattr vattr; 3420 int error; 3421 struct nameidata nd; 3422 3423 restart: 3424 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3425 SCARG(uap, path), l); 3426 if ((error = namei(&nd)) != 0) 3427 return (error); 3428 vp = nd.ni_vp; 3429 if (vp != NULL) { 3430 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3431 if (nd.ni_dvp == vp) 3432 vrele(nd.ni_dvp); 3433 else 3434 vput(nd.ni_dvp); 3435 vrele(vp); 3436 return (EEXIST); 3437 } 3438 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3439 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3440 if (nd.ni_dvp == vp) 3441 vrele(nd.ni_dvp); 3442 else 3443 vput(nd.ni_dvp); 3444 if ((error = vn_start_write(NULL, &mp, 3445 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3446 return (error); 3447 goto restart; 3448 } 3449 VATTR_NULL(&vattr); 3450 vattr.va_type = VDIR; 3451 vattr.va_mode = 3452 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3453 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3454 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3455 if (!error) 3456 vput(nd.ni_vp); 3457 vn_finished_write(mp, 0); 3458 return (error); 3459 } 3460 3461 /* 3462 * Remove a directory file. 3463 */ 3464 /* ARGSUSED */ 3465 int 3466 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3467 { 3468 struct sys_rmdir_args /* { 3469 syscallarg(const char *) path; 3470 } */ *uap = v; 3471 struct mount *mp; 3472 struct vnode *vp; 3473 int error; 3474 struct nameidata nd; 3475 3476 restart: 3477 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3478 SCARG(uap, path), l); 3479 if ((error = namei(&nd)) != 0) 3480 return (error); 3481 vp = nd.ni_vp; 3482 if (vp->v_type != VDIR) { 3483 error = ENOTDIR; 3484 goto out; 3485 } 3486 /* 3487 * No rmdir "." please. 3488 */ 3489 if (nd.ni_dvp == vp) { 3490 error = EINVAL; 3491 goto out; 3492 } 3493 /* 3494 * The root of a mounted filesystem cannot be deleted. 3495 */ 3496 if (vp->v_flag & VROOT) { 3497 error = EBUSY; 3498 goto out; 3499 } 3500 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3501 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3502 if (nd.ni_dvp == vp) 3503 vrele(nd.ni_dvp); 3504 else 3505 vput(nd.ni_dvp); 3506 vput(vp); 3507 if ((error = vn_start_write(NULL, &mp, 3508 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3509 return (error); 3510 goto restart; 3511 } 3512 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3513 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3514 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3515 vn_finished_write(mp, 0); 3516 return (error); 3517 3518 out: 3519 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3520 if (nd.ni_dvp == vp) 3521 vrele(nd.ni_dvp); 3522 else 3523 vput(nd.ni_dvp); 3524 vput(vp); 3525 return (error); 3526 } 3527 3528 /* 3529 * Read a block of directory entries in a file system independent format. 3530 */ 3531 int 3532 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3533 { 3534 struct sys___getdents30_args /* { 3535 syscallarg(int) fd; 3536 syscallarg(char *) buf; 3537 syscallarg(size_t) count; 3538 } */ *uap = v; 3539 struct proc *p = l->l_proc; 3540 struct file *fp; 3541 int error, done; 3542 3543 /* getvnode() will use the descriptor for us */ 3544 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3545 return (error); 3546 if ((fp->f_flag & FREAD) == 0) { 3547 error = EBADF; 3548 goto out; 3549 } 3550 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3551 SCARG(uap, count), &done, l, 0, 0); 3552 #ifdef KTRACE 3553 if (!error && KTRPOINT(p, KTR_GENIO)) { 3554 struct iovec iov; 3555 iov.iov_base = SCARG(uap, buf); 3556 iov.iov_len = done; 3557 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3558 } 3559 #endif 3560 *retval = done; 3561 out: 3562 FILE_UNUSE(fp, l); 3563 return (error); 3564 } 3565 3566 /* 3567 * Set the mode mask for creation of filesystem nodes. 3568 */ 3569 int 3570 sys_umask(struct lwp *l, void *v, register_t *retval) 3571 { 3572 struct sys_umask_args /* { 3573 syscallarg(mode_t) newmask; 3574 } */ *uap = v; 3575 struct proc *p = l->l_proc; 3576 struct cwdinfo *cwdi; 3577 3578 cwdi = p->p_cwdi; 3579 *retval = cwdi->cwdi_cmask; 3580 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3581 return (0); 3582 } 3583 3584 /* 3585 * Void all references to file by ripping underlying filesystem 3586 * away from vnode. 3587 */ 3588 /* ARGSUSED */ 3589 int 3590 sys_revoke(struct lwp *l, void *v, register_t *retval) 3591 { 3592 struct sys_revoke_args /* { 3593 syscallarg(const char *) path; 3594 } */ *uap = v; 3595 struct mount *mp; 3596 struct vnode *vp; 3597 struct vattr vattr; 3598 int error; 3599 struct nameidata nd; 3600 3601 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3602 if ((error = namei(&nd)) != 0) 3603 return (error); 3604 vp = nd.ni_vp; 3605 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3606 goto out; 3607 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3608 (error = kauth_authorize_generic(l->l_cred, 3609 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 3610 goto out; 3611 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3612 goto out; 3613 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3614 VOP_REVOKE(vp, REVOKEALL); 3615 vn_finished_write(mp, 0); 3616 out: 3617 vrele(vp); 3618 return (error); 3619 } 3620 3621 /* 3622 * Convert a user file descriptor to a kernel file entry. 3623 */ 3624 int 3625 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3626 { 3627 struct vnode *vp; 3628 struct file *fp; 3629 3630 if ((fp = fd_getfile(fdp, fd)) == NULL) 3631 return (EBADF); 3632 3633 FILE_USE(fp); 3634 3635 if (fp->f_type != DTYPE_VNODE) { 3636 FILE_UNUSE(fp, NULL); 3637 return (EINVAL); 3638 } 3639 3640 vp = (struct vnode *)fp->f_data; 3641 if (vp->v_type == VBAD) { 3642 FILE_UNUSE(fp, NULL); 3643 return (EBADF); 3644 } 3645 3646 *fpp = fp; 3647 return (0); 3648 } 3649
Indexes created Sat Sep 27 14:09:57 GMT 2025