vfs_syscalls.c revision 1.274
1 /* $NetBSD: vfs_syscalls.c,v 1.274 2006/10/24 21:53:10 mjf 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 1.274 2006/10/24 21:53:10 mjf 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 #endif 84 #include <nfs/nfsproto.h> 85 #ifdef COMPAT_30 86 #include <nfs/nfs.h> 87 #include <nfs/nfs_var.h> 88 #endif 89 90 #if NFSS > 0 91 #include <dev/fssvar.h> 92 #endif 93 94 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 95 96 static int change_dir(struct nameidata *, struct lwp *); 97 static int change_flags(struct vnode *, u_long, struct lwp *); 98 static int change_mode(struct vnode *, int, struct lwp *l); 99 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 100 static int change_utimes(struct vnode *vp, const struct timeval *, 101 struct lwp *l); 102 static int rename_files(const char *, const char *, struct lwp *, int); 103 104 void checkdirs(struct vnode *); 105 106 int dovfsusermount = 0; 107 108 /* 109 * Virtual File System System Calls 110 */ 111 112 /* 113 * Mount a file system. 114 */ 115 116 #if defined(COMPAT_09) || defined(COMPAT_43) 117 /* 118 * This table is used to maintain compatibility with 4.3BSD 119 * and NetBSD 0.9 mount syscalls. Note, the order is important! 120 * 121 * Do not modify this table. It should only contain filesystems 122 * supported by NetBSD 0.9 and 4.3BSD. 123 */ 124 const char * const mountcompatnames[] = { 125 NULL, /* 0 = MOUNT_NONE */ 126 MOUNT_FFS, /* 1 = MOUNT_UFS */ 127 MOUNT_NFS, /* 2 */ 128 MOUNT_MFS, /* 3 */ 129 MOUNT_MSDOS, /* 4 */ 130 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 131 MOUNT_FDESC, /* 6 */ 132 MOUNT_KERNFS, /* 7 */ 133 NULL, /* 8 = MOUNT_DEVFS */ 134 MOUNT_AFS, /* 9 */ 135 }; 136 const int nmountcompatnames = sizeof(mountcompatnames) / 137 sizeof(mountcompatnames[0]); 138 #endif /* COMPAT_09 || COMPAT_43 */ 139 140 /* ARGSUSED */ 141 int 142 sys_mount(struct lwp *l, void *v, register_t *retval __unused) 143 { 144 struct sys_mount_args /* { 145 syscallarg(const char *) type; 146 syscallarg(const char *) path; 147 syscallarg(int) flags; 148 syscallarg(void *) data; 149 } */ *uap = v; 150 struct vnode *vp; 151 struct mount *mp; 152 int error, flag = 0; 153 char fstypename[MFSNAMELEN]; 154 struct vattr va; 155 struct nameidata nd; 156 struct vfsops *vfs; 157 158 /* 159 * if MNT_GETARGS is specified, it should be only flag. 160 */ 161 162 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 163 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 164 return EINVAL; 165 } 166 167 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 168 (error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 169 &l->l_acflag))) 170 return (error); 171 /* 172 * Get vnode to be covered 173 */ 174 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 175 SCARG(uap, path), l); 176 if ((error = namei(&nd)) != 0) 177 return (error); 178 vp = nd.ni_vp; 179 /* 180 * A lookup in VFS_MOUNT might result in an attempt to 181 * lock this vnode again, so make the lock recursive. 182 */ 183 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 184 185 /* 186 * Do not allow the filesystem to be exported if MNT_NOSHARE 187 * is set. 188 */ 189 if ((SCARG(uap, flags) & MNT_EXPORTED) && 190 (vp->v_mount->mnt_flag & MNT_NOSHARE)) { 191 vput(vp); 192 return (EPERM); 193 } 194 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 195 if ((vp->v_flag & VROOT) == 0) { 196 vput(vp); 197 return (EINVAL); 198 } 199 mp = vp->v_mount; 200 flag = mp->mnt_flag; 201 vfs = mp->mnt_op; 202 /* 203 * We only allow the filesystem to be reloaded if it 204 * is currently mounted read-only. 205 */ 206 if ((SCARG(uap, flags) & MNT_RELOAD) && 207 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 208 vput(vp); 209 return (EOPNOTSUPP); /* Needs translation */ 210 } 211 /* 212 * In "highly secure" mode, don't let the caller do anything 213 * but downgrade a filesystem from read-write to read-only. 214 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 215 */ 216 if (securelevel >= 2 && 217 SCARG(uap, flags) != MNT_GETARGS && 218 SCARG(uap, flags) != 219 (mp->mnt_flag | MNT_RDONLY | 220 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 221 vput(vp); 222 return (EPERM); 223 } 224 mp->mnt_flag |= SCARG(uap, flags) & 225 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 226 /* 227 * Only root, or the user that did the original mount is 228 * permitted to update it. 229 */ 230 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 231 mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred) && 232 (error = kauth_authorize_generic(l->l_cred, 233 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 234 vput(vp); 235 return (error); 236 } 237 /* 238 * Do not allow NFS export by non-root users. For non-root 239 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 240 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 241 */ 242 if (kauth_cred_geteuid(l->l_cred) != 0) { 243 if (SCARG(uap, flags) & MNT_EXPORTED) { 244 vput(vp); 245 return (EPERM); 246 } 247 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 248 if (flag & MNT_NOEXEC) 249 SCARG(uap, flags) |= MNT_NOEXEC; 250 } 251 if (vfs_busy(mp, LK_NOWAIT, 0)) { 252 vput(vp); 253 return (EPERM); 254 } 255 goto update; 256 } else { 257 if (securelevel >= 2) { 258 vput(vp); 259 return (EPERM); 260 } 261 } 262 /* 263 * If the user is not root, ensure that they own the directory 264 * onto which we are attempting to mount. 265 */ 266 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 267 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 268 (error = kauth_authorize_generic(l->l_cred, 269 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0)) { 270 vput(vp); 271 return (error); 272 } 273 /* 274 * Do not allow NFS export by non-root users. For non-root users, 275 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 276 * mount point is already MNT_NOEXEC. 277 */ 278 if (kauth_cred_geteuid(l->l_cred) != 0) { 279 if (SCARG(uap, flags) & MNT_EXPORTED) { 280 vput(vp); 281 return (EPERM); 282 } 283 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 284 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 285 SCARG(uap, flags) |= MNT_NOEXEC; 286 } 287 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 288 vput(vp); 289 return (error); 290 } 291 if (vp->v_type != VDIR) { 292 vput(vp); 293 return (ENOTDIR); 294 } 295 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 296 if (error) { 297 #if defined(COMPAT_09) || defined(COMPAT_43) 298 /* 299 * Historically, filesystem types were identified by numbers. 300 * If we get an integer for the filesystem type instead of a 301 * string, we check to see if it matches one of the historic 302 * filesystem types. 303 */ 304 u_long fsindex = (u_long)SCARG(uap, type); 305 if (fsindex >= nmountcompatnames || 306 mountcompatnames[fsindex] == NULL) { 307 vput(vp); 308 return (ENODEV); 309 } 310 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 311 #else 312 vput(vp); 313 return (error); 314 #endif 315 } 316 #ifdef COMPAT_10 317 /* Accept `ufs' as an alias for `ffs'. */ 318 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 319 strncpy(fstypename, "ffs", MFSNAMELEN); 320 #endif 321 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 322 vput(vp); 323 return (ENODEV); 324 } 325 if (vp->v_mountedhere != NULL) { 326 vput(vp); 327 return (EBUSY); 328 } 329 330 /* 331 * Allocate and initialize the file system. 332 */ 333 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 334 M_MOUNT, M_WAITOK); 335 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 336 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 337 simple_lock_init(&mp->mnt_slock); 338 (void)vfs_busy(mp, LK_NOWAIT, 0); 339 mp->mnt_op = vfs; 340 vfs->vfs_refcount++; 341 mp->mnt_vnodecovered = vp; 342 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 343 mp->mnt_unmounter = NULL; 344 mp->mnt_leaf = mp; 345 346 /* 347 * The underlying file system may refuse the mount for 348 * various reasons. Allow the user to force it to happen. 349 */ 350 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 351 update: 352 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 353 /* 354 * Set the mount level flags. 355 */ 356 if (SCARG(uap, flags) & MNT_RDONLY) 357 mp->mnt_flag |= MNT_RDONLY; 358 else if (mp->mnt_flag & MNT_RDONLY) 359 mp->mnt_iflag |= IMNT_WANTRDWR; 360 mp->mnt_flag &= 361 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 362 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 363 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 364 MNT_NOSHARE); 365 mp->mnt_flag |= SCARG(uap, flags) & 366 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 367 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 368 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 369 MNT_IGNORE | MNT_NOSHARE); 370 } 371 /* 372 * Mount the filesystem. 373 */ 374 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 375 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 376 #if defined(COMPAT_30) && defined(NFSSERVER) 377 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 378 int error2; 379 380 /* Update failed; let's try and see if it was an 381 * export request. */ 382 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 383 SCARG(uap, data), l); 384 385 /* Only update error code if the export request was 386 * understood but some problem occurred while 387 * processing it. */ 388 if (error2 != EJUSTRETURN) 389 error = error2; 390 } 391 #endif 392 if (mp->mnt_iflag & IMNT_WANTRDWR) 393 mp->mnt_flag &= ~MNT_RDONLY; 394 if (error) 395 mp->mnt_flag = flag; 396 mp->mnt_flag &=~ 397 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 398 mp->mnt_iflag &=~ IMNT_WANTRDWR; 399 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 400 if (mp->mnt_syncer == NULL) 401 error = vfs_allocate_syncvnode(mp); 402 } else { 403 if (mp->mnt_syncer != NULL) 404 vfs_deallocate_syncvnode(mp); 405 } 406 vfs_unbusy(mp); 407 VOP_UNLOCK(vp, 0); 408 vrele(vp); 409 return (error); 410 } 411 /* 412 * Put the new filesystem on the mount list after root. 413 */ 414 cache_purge(vp); 415 if (!error) { 416 mp->mnt_flag &=~ 417 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 418 mp->mnt_iflag &=~ IMNT_WANTRDWR; 419 vp->v_mountedhere = mp; 420 simple_lock(&mountlist_slock); 421 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 422 simple_unlock(&mountlist_slock); 423 checkdirs(vp); 424 VOP_UNLOCK(vp, 0); 425 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 426 error = vfs_allocate_syncvnode(mp); 427 vfs_unbusy(mp); 428 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 429 if ((error = VFS_START(mp, 0, l))) 430 vrele(vp); 431 } else { 432 vp->v_mountedhere = (struct mount *)0; 433 vfs->vfs_refcount--; 434 vfs_unbusy(mp); 435 free(mp, M_MOUNT); 436 vput(vp); 437 } 438 return (error); 439 } 440 441 /* 442 * Scan all active processes to see if any of them have a current 443 * or root directory onto which the new filesystem has just been 444 * mounted. If so, replace them with the new mount point. 445 */ 446 void 447 checkdirs(struct vnode *olddp) 448 { 449 struct cwdinfo *cwdi; 450 struct vnode *newdp; 451 struct proc *p; 452 453 if (olddp->v_usecount == 1) 454 return; 455 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 456 panic("mount: lost mount"); 457 proclist_lock_read(); 458 PROCLIST_FOREACH(p, &allproc) { 459 cwdi = p->p_cwdi; 460 if (!cwdi) 461 continue; 462 if (cwdi->cwdi_cdir == olddp) { 463 vrele(cwdi->cwdi_cdir); 464 VREF(newdp); 465 cwdi->cwdi_cdir = newdp; 466 } 467 if (cwdi->cwdi_rdir == olddp) { 468 vrele(cwdi->cwdi_rdir); 469 VREF(newdp); 470 cwdi->cwdi_rdir = newdp; 471 } 472 } 473 proclist_unlock_read(); 474 if (rootvnode == olddp) { 475 vrele(rootvnode); 476 VREF(newdp); 477 rootvnode = newdp; 478 } 479 vput(newdp); 480 } 481 482 /* 483 * Unmount a file system. 484 * 485 * Note: unmount takes a path to the vnode mounted on as argument, 486 * not special file (as before). 487 */ 488 /* ARGSUSED */ 489 int 490 sys_unmount(struct lwp *l, void *v, register_t *retval __unused) 491 { 492 struct sys_unmount_args /* { 493 syscallarg(const char *) path; 494 syscallarg(int) flags; 495 } */ *uap = v; 496 struct vnode *vp; 497 struct mount *mp; 498 int error; 499 struct nameidata nd; 500 501 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 502 SCARG(uap, path), l); 503 if ((error = namei(&nd)) != 0) 504 return (error); 505 vp = nd.ni_vp; 506 mp = vp->v_mount; 507 508 /* 509 * Only root, or the user that did the original mount is 510 * permitted to unmount this filesystem. 511 */ 512 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred)) && 513 (error = kauth_authorize_generic(l->l_cred, 514 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 515 vput(vp); 516 return (error); 517 } 518 519 /* 520 * Don't allow unmounting the root file system. 521 */ 522 if (mp->mnt_flag & MNT_ROOTFS) { 523 vput(vp); 524 return (EINVAL); 525 } 526 527 /* 528 * Must be the root of the filesystem 529 */ 530 if ((vp->v_flag & VROOT) == 0) { 531 vput(vp); 532 return (EINVAL); 533 } 534 vput(vp); 535 536 /* 537 * XXX Freeze syncer. Must do this before locking the 538 * mount point. See dounmount() for details. 539 */ 540 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 541 542 if (vfs_busy(mp, 0, 0)) { 543 lockmgr(&syncer_lock, LK_RELEASE, NULL); 544 return (EBUSY); 545 } 546 547 return (dounmount(mp, SCARG(uap, flags), l)); 548 } 549 550 /* 551 * Do the actual file system unmount. File system is assumed to have been 552 * marked busy by the caller. 553 */ 554 int 555 dounmount(struct mount *mp, int flags, struct lwp *l) 556 { 557 struct vnode *coveredvp; 558 int error; 559 int async; 560 int used_syncer; 561 562 #if NVERIEXEC > 0 563 if (!doing_shutdown) { 564 if (veriexec_strict >= VERIEXEC_LOCKDOWN) { 565 printf("Veriexec: Lockdown mode, preventing unmount of" 566 " \"%s\". (uid=%u)\n", mp->mnt_stat.f_mntonname, 567 kauth_cred_getuid(l->l_cred)); 568 return (EPERM); 569 } 570 571 if (veriexec_strict == VERIEXEC_IPS) { 572 struct veriexec_table_entry *vte; 573 574 /* Check if we have fingerprints on mount. */ 575 vte = fileassoc_tabledata_lookup(mp, veriexec_hook); 576 if ((vte != NULL) && (vte->vte_count > 0)) { 577 printf("Veriexec: IPS mode, preventing unmount" 578 " of \"%s\" with monitored files. " 579 "(uid=%u)\n", mp->mnt_stat.f_mntonname, 580 kauth_cred_getuid(l->l_cred)); 581 return (EPERM); 582 } 583 } 584 } 585 #endif /* NVERIEXEC > 0 */ 586 587 #ifdef FILEASSOC 588 (void)fileassoc_table_delete(mp); 589 #endif /* FILEASSOC */ 590 591 simple_lock(&mountlist_slock); 592 vfs_unbusy(mp); 593 used_syncer = (mp->mnt_syncer != NULL); 594 595 /* 596 * XXX Syncer must be frozen when we get here. This should really 597 * be done on a per-mountpoint basis, but especially the softdep 598 * code possibly called from the syncer doesn't exactly work on a 599 * per-mountpoint basis, so the softdep code would become a maze 600 * of vfs_busy() calls. 601 * 602 * The caller of dounmount() must acquire syncer_lock because 603 * the syncer itself acquires locks in syncer_lock -> vfs_busy 604 * order, and we must preserve that order to avoid deadlock. 605 * 606 * So, if the file system did not use the syncer, now is 607 * the time to release the syncer_lock. 608 */ 609 if (used_syncer == 0) 610 lockmgr(&syncer_lock, LK_RELEASE, NULL); 611 612 mp->mnt_iflag |= IMNT_UNMOUNT; 613 mp->mnt_unmounter = l; 614 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 615 vn_start_write(NULL, &mp, V_WAIT); 616 617 async = mp->mnt_flag & MNT_ASYNC; 618 mp->mnt_flag &= ~MNT_ASYNC; 619 cache_purgevfs(mp); /* remove cache entries for this file sys */ 620 if (mp->mnt_syncer != NULL) 621 vfs_deallocate_syncvnode(mp); 622 error = 0; 623 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 624 #if NFSS > 0 625 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 626 #endif 627 if (error == 0) 628 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 629 } 630 if (error == 0 || (flags & MNT_FORCE)) 631 error = VFS_UNMOUNT(mp, flags, l); 632 vn_finished_write(mp, 0); 633 simple_lock(&mountlist_slock); 634 if (error) { 635 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 636 (void) vfs_allocate_syncvnode(mp); 637 mp->mnt_iflag &= ~IMNT_UNMOUNT; 638 mp->mnt_unmounter = NULL; 639 mp->mnt_flag |= async; 640 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 641 &mountlist_slock); 642 if (used_syncer) 643 lockmgr(&syncer_lock, LK_RELEASE, NULL); 644 simple_lock(&mp->mnt_slock); 645 while (mp->mnt_wcnt > 0) { 646 wakeup(mp); 647 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 648 0, &mp->mnt_slock); 649 } 650 simple_unlock(&mp->mnt_slock); 651 return (error); 652 } 653 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 654 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 655 coveredvp->v_mountedhere = NULL; 656 vrele(coveredvp); 657 } 658 mp->mnt_op->vfs_refcount--; 659 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 660 panic("unmount: dangling vnode"); 661 mp->mnt_iflag |= IMNT_GONE; 662 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 663 if (used_syncer) 664 lockmgr(&syncer_lock, LK_RELEASE, NULL); 665 simple_lock(&mp->mnt_slock); 666 while (mp->mnt_wcnt > 0) { 667 wakeup(mp); 668 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 669 } 670 simple_unlock(&mp->mnt_slock); 671 vfs_hooks_unmount(mp); 672 free(mp, M_MOUNT); 673 return (0); 674 } 675 676 /* 677 * Sync each mounted filesystem. 678 */ 679 #ifdef DEBUG 680 int syncprt = 0; 681 struct ctldebug debug0 = { "syncprt", &syncprt }; 682 #endif 683 684 /* ARGSUSED */ 685 int 686 sys_sync(struct lwp *l, void *v __unused, register_t *retval __unused) 687 { 688 struct mount *mp, *nmp; 689 int asyncflag; 690 691 if (l == NULL) 692 l = &lwp0; 693 694 simple_lock(&mountlist_slock); 695 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 696 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 697 nmp = mp->mnt_list.cqe_prev; 698 continue; 699 } 700 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 701 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 702 asyncflag = mp->mnt_flag & MNT_ASYNC; 703 mp->mnt_flag &= ~MNT_ASYNC; 704 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 705 if (asyncflag) 706 mp->mnt_flag |= MNT_ASYNC; 707 vn_finished_write(mp, 0); 708 } 709 simple_lock(&mountlist_slock); 710 nmp = mp->mnt_list.cqe_prev; 711 vfs_unbusy(mp); 712 713 } 714 simple_unlock(&mountlist_slock); 715 #ifdef DEBUG 716 if (syncprt) 717 vfs_bufstats(); 718 #endif /* DEBUG */ 719 return (0); 720 } 721 722 /* 723 * Change filesystem quotas. 724 */ 725 /* ARGSUSED */ 726 int 727 sys_quotactl(struct lwp *l, void *v, register_t *retval __unused) 728 { 729 struct sys_quotactl_args /* { 730 syscallarg(const char *) path; 731 syscallarg(int) cmd; 732 syscallarg(int) uid; 733 syscallarg(void *) arg; 734 } */ *uap = v; 735 struct mount *mp; 736 int error; 737 struct nameidata nd; 738 739 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 740 if ((error = namei(&nd)) != 0) 741 return (error); 742 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 743 vrele(nd.ni_vp); 744 if (error) 745 return (error); 746 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 747 SCARG(uap, arg), l); 748 vn_finished_write(mp, 0); 749 return (error); 750 } 751 752 int 753 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 754 int root) 755 { 756 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 757 int error = 0; 758 759 /* 760 * If MNT_NOWAIT or MNT_LAZY is specified, do not 761 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 762 * overrides MNT_NOWAIT. 763 */ 764 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 765 (flags != MNT_WAIT && flags != 0)) { 766 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 767 goto done; 768 } 769 770 /* Get the filesystem stats now */ 771 memset(sp, 0, sizeof(*sp)); 772 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 773 return error; 774 } 775 776 if (cwdi->cwdi_rdir == NULL) 777 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 778 done: 779 if (cwdi->cwdi_rdir != NULL) { 780 size_t len; 781 char *bp; 782 char *path = PNBUF_GET(); 783 if (!path) 784 return ENOMEM; 785 786 bp = path + MAXPATHLEN; 787 *--bp = '\0'; 788 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 789 MAXPATHLEN / 2, 0, l); 790 if (error) { 791 PNBUF_PUT(path); 792 return error; 793 } 794 len = strlen(bp); 795 /* 796 * for mount points that are below our root, we can see 797 * them, so we fix up the pathname and return them. The 798 * rest we cannot see, so we don't allow viewing the 799 * data. 800 */ 801 if (strncmp(bp, sp->f_mntonname, len) == 0) { 802 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 803 sizeof(sp->f_mntonname)); 804 if (sp->f_mntonname[0] == '\0') 805 (void)strlcpy(sp->f_mntonname, "/", 806 sizeof(sp->f_mntonname)); 807 } else { 808 if (root) 809 (void)strlcpy(sp->f_mntonname, "/", 810 sizeof(sp->f_mntonname)); 811 else 812 error = EPERM; 813 } 814 PNBUF_PUT(path); 815 } 816 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 817 return error; 818 } 819 820 /* 821 * Get filesystem statistics. 822 */ 823 /* ARGSUSED */ 824 int 825 sys_statvfs1(struct lwp *l, void *v, register_t *retval __unused) 826 { 827 struct sys_statvfs1_args /* { 828 syscallarg(const char *) path; 829 syscallarg(struct statvfs *) buf; 830 syscallarg(int) flags; 831 } */ *uap = v; 832 struct mount *mp; 833 struct statvfs *sb; 834 int error; 835 struct nameidata nd; 836 837 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 838 if ((error = namei(&nd)) != 0) 839 return error; 840 mp = nd.ni_vp->v_mount; 841 vrele(nd.ni_vp); 842 sb = STATVFSBUF_GET(); 843 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 844 if (error == 0) { 845 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 846 } 847 STATVFSBUF_PUT(sb); 848 return error; 849 } 850 851 /* 852 * Get filesystem statistics. 853 */ 854 /* ARGSUSED */ 855 int 856 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval __unused) 857 { 858 struct sys_fstatvfs1_args /* { 859 syscallarg(int) fd; 860 syscallarg(struct statvfs *) buf; 861 syscallarg(int) flags; 862 } */ *uap = v; 863 struct proc *p = l->l_proc; 864 struct file *fp; 865 struct mount *mp; 866 struct statvfs *sb; 867 int error; 868 869 /* getvnode() will use the descriptor for us */ 870 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 871 return (error); 872 mp = ((struct vnode *)fp->f_data)->v_mount; 873 sb = STATVFSBUF_GET(); 874 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 875 goto out; 876 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 877 out: 878 FILE_UNUSE(fp, l); 879 STATVFSBUF_PUT(sb); 880 return error; 881 } 882 883 884 /* 885 * Get statistics on all filesystems. 886 */ 887 int 888 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 889 { 890 struct sys_getvfsstat_args /* { 891 syscallarg(struct statvfs *) buf; 892 syscallarg(size_t) bufsize; 893 syscallarg(int) flags; 894 } */ *uap = v; 895 int root = 0; 896 struct proc *p = l->l_proc; 897 struct mount *mp, *nmp; 898 struct statvfs *sb; 899 struct statvfs *sfsp; 900 size_t count, maxcount; 901 int error = 0; 902 903 sb = STATVFSBUF_GET(); 904 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 905 sfsp = SCARG(uap, buf); 906 simple_lock(&mountlist_slock); 907 count = 0; 908 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 909 mp = nmp) { 910 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 911 nmp = CIRCLEQ_NEXT(mp, mnt_list); 912 continue; 913 } 914 if (sfsp && count < maxcount) { 915 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 916 if (error) { 917 simple_lock(&mountlist_slock); 918 nmp = CIRCLEQ_NEXT(mp, mnt_list); 919 vfs_unbusy(mp); 920 continue; 921 } 922 error = copyout(sb, sfsp, sizeof(*sfsp)); 923 if (error) { 924 vfs_unbusy(mp); 925 goto out; 926 } 927 sfsp++; 928 root |= strcmp(sb->f_mntonname, "/") == 0; 929 } 930 count++; 931 simple_lock(&mountlist_slock); 932 nmp = CIRCLEQ_NEXT(mp, mnt_list); 933 vfs_unbusy(mp); 934 } 935 simple_unlock(&mountlist_slock); 936 if (root == 0 && p->p_cwdi->cwdi_rdir) { 937 /* 938 * fake a root entry 939 */ 940 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 941 SCARG(uap, flags), 1)) != 0) 942 goto out; 943 if (sfsp) 944 error = copyout(sb, sfsp, sizeof(*sfsp)); 945 count++; 946 } 947 if (sfsp && count > maxcount) 948 *retval = maxcount; 949 else 950 *retval = count; 951 out: 952 STATVFSBUF_PUT(sb); 953 return error; 954 } 955 956 /* 957 * Change current working directory to a given file descriptor. 958 */ 959 /* ARGSUSED */ 960 int 961 sys_fchdir(struct lwp *l, void *v, register_t *retval __unused) 962 { 963 struct sys_fchdir_args /* { 964 syscallarg(int) fd; 965 } */ *uap = v; 966 struct proc *p = l->l_proc; 967 struct filedesc *fdp = p->p_fd; 968 struct cwdinfo *cwdi = p->p_cwdi; 969 struct vnode *vp, *tdp; 970 struct mount *mp; 971 struct file *fp; 972 int error; 973 974 /* getvnode() will use the descriptor for us */ 975 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 976 return (error); 977 vp = (struct vnode *)fp->f_data; 978 979 VREF(vp); 980 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 981 if (vp->v_type != VDIR) 982 error = ENOTDIR; 983 else 984 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 985 while (!error && (mp = vp->v_mountedhere) != NULL) { 986 if (vfs_busy(mp, 0, 0)) 987 continue; 988 error = VFS_ROOT(mp, &tdp); 989 vfs_unbusy(mp); 990 if (error) 991 break; 992 vput(vp); 993 vp = tdp; 994 } 995 if (error) { 996 vput(vp); 997 goto out; 998 } 999 VOP_UNLOCK(vp, 0); 1000 1001 /* 1002 * Disallow changing to a directory not under the process's 1003 * current root directory (if there is one). 1004 */ 1005 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1006 vrele(vp); 1007 error = EPERM; /* operation not permitted */ 1008 goto out; 1009 } 1010 1011 vrele(cwdi->cwdi_cdir); 1012 cwdi->cwdi_cdir = vp; 1013 out: 1014 FILE_UNUSE(fp, l); 1015 return (error); 1016 } 1017 1018 /* 1019 * Change this process's notion of the root directory to a given file 1020 * descriptor. 1021 */ 1022 int 1023 sys_fchroot(struct lwp *l, void *v, register_t *retval __unused) 1024 { 1025 struct sys_fchroot_args *uap = v; 1026 struct proc *p = l->l_proc; 1027 struct filedesc *fdp = p->p_fd; 1028 struct cwdinfo *cwdi = p->p_cwdi; 1029 struct vnode *vp; 1030 struct file *fp; 1031 int error; 1032 1033 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1034 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1035 return error; 1036 /* getvnode() will use the descriptor for us */ 1037 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1038 return error; 1039 vp = (struct vnode *) fp->f_data; 1040 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1041 if (vp->v_type != VDIR) 1042 error = ENOTDIR; 1043 else 1044 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1045 VOP_UNLOCK(vp, 0); 1046 if (error) 1047 goto out; 1048 VREF(vp); 1049 1050 /* 1051 * Prevent escaping from chroot by putting the root under 1052 * the working directory. Silently chdir to / if we aren't 1053 * already there. 1054 */ 1055 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1056 /* 1057 * XXX would be more failsafe to change directory to a 1058 * deadfs node here instead 1059 */ 1060 vrele(cwdi->cwdi_cdir); 1061 VREF(vp); 1062 cwdi->cwdi_cdir = vp; 1063 } 1064 1065 if (cwdi->cwdi_rdir != NULL) 1066 vrele(cwdi->cwdi_rdir); 1067 cwdi->cwdi_rdir = vp; 1068 out: 1069 FILE_UNUSE(fp, l); 1070 return (error); 1071 } 1072 1073 /* 1074 * Change current working directory (``.''). 1075 */ 1076 /* ARGSUSED */ 1077 int 1078 sys_chdir(struct lwp *l, void *v, register_t *retval __unused) 1079 { 1080 struct sys_chdir_args /* { 1081 syscallarg(const char *) path; 1082 } */ *uap = v; 1083 struct proc *p = l->l_proc; 1084 struct cwdinfo *cwdi = p->p_cwdi; 1085 int error; 1086 struct nameidata nd; 1087 1088 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1089 SCARG(uap, path), l); 1090 if ((error = change_dir(&nd, l)) != 0) 1091 return (error); 1092 vrele(cwdi->cwdi_cdir); 1093 cwdi->cwdi_cdir = nd.ni_vp; 1094 return (0); 1095 } 1096 1097 /* 1098 * Change notion of root (``/'') directory. 1099 */ 1100 /* ARGSUSED */ 1101 int 1102 sys_chroot(struct lwp *l, void *v, register_t *retval __unused) 1103 { 1104 struct sys_chroot_args /* { 1105 syscallarg(const char *) path; 1106 } */ *uap = v; 1107 struct proc *p = l->l_proc; 1108 struct cwdinfo *cwdi = p->p_cwdi; 1109 struct vnode *vp; 1110 int error; 1111 struct nameidata nd; 1112 1113 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1114 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1115 return (error); 1116 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1117 SCARG(uap, path), l); 1118 if ((error = change_dir(&nd, l)) != 0) 1119 return (error); 1120 if (cwdi->cwdi_rdir != NULL) 1121 vrele(cwdi->cwdi_rdir); 1122 vp = nd.ni_vp; 1123 cwdi->cwdi_rdir = vp; 1124 1125 /* 1126 * Prevent escaping from chroot by putting the root under 1127 * the working directory. Silently chdir to / if we aren't 1128 * already there. 1129 */ 1130 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1131 /* 1132 * XXX would be more failsafe to change directory to a 1133 * deadfs node here instead 1134 */ 1135 vrele(cwdi->cwdi_cdir); 1136 VREF(vp); 1137 cwdi->cwdi_cdir = vp; 1138 } 1139 1140 return (0); 1141 } 1142 1143 /* 1144 * Common routine for chroot and chdir. 1145 */ 1146 static int 1147 change_dir(struct nameidata *ndp, struct lwp *l) 1148 { 1149 struct vnode *vp; 1150 int error; 1151 1152 if ((error = namei(ndp)) != 0) 1153 return (error); 1154 vp = ndp->ni_vp; 1155 if (vp->v_type != VDIR) 1156 error = ENOTDIR; 1157 else 1158 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1159 1160 if (error) 1161 vput(vp); 1162 else 1163 VOP_UNLOCK(vp, 0); 1164 return (error); 1165 } 1166 1167 /* 1168 * Check permissions, allocate an open file structure, 1169 * and call the device open routine if any. 1170 */ 1171 int 1172 sys_open(struct lwp *l, void *v, register_t *retval) 1173 { 1174 struct sys_open_args /* { 1175 syscallarg(const char *) path; 1176 syscallarg(int) flags; 1177 syscallarg(int) mode; 1178 } */ *uap = v; 1179 struct proc *p = l->l_proc; 1180 struct cwdinfo *cwdi = p->p_cwdi; 1181 struct filedesc *fdp = p->p_fd; 1182 struct file *fp; 1183 struct vnode *vp; 1184 int flags, cmode; 1185 int type, indx, error; 1186 struct flock lf; 1187 struct nameidata nd; 1188 1189 flags = FFLAGS(SCARG(uap, flags)); 1190 if ((flags & (FREAD | FWRITE)) == 0) 1191 return (EINVAL); 1192 /* falloc() will use the file descriptor for us */ 1193 if ((error = falloc(l, &fp, &indx)) != 0) 1194 return (error); 1195 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1196 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1197 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1198 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1199 FILE_UNUSE(fp, l); 1200 fdp->fd_ofiles[indx] = NULL; 1201 ffree(fp); 1202 if ((error == EDUPFD || error == EMOVEFD) && 1203 l->l_dupfd >= 0 && /* XXX from fdopen */ 1204 (error = 1205 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1206 *retval = indx; 1207 return (0); 1208 } 1209 if (error == ERESTART) 1210 error = EINTR; 1211 fdremove(fdp, indx); 1212 return (error); 1213 } 1214 l->l_dupfd = 0; 1215 vp = nd.ni_vp; 1216 fp->f_flag = flags & FMASK; 1217 fp->f_type = DTYPE_VNODE; 1218 fp->f_ops = &vnops; 1219 fp->f_data = vp; 1220 if (flags & (O_EXLOCK | O_SHLOCK)) { 1221 lf.l_whence = SEEK_SET; 1222 lf.l_start = 0; 1223 lf.l_len = 0; 1224 if (flags & O_EXLOCK) 1225 lf.l_type = F_WRLCK; 1226 else 1227 lf.l_type = F_RDLCK; 1228 type = F_FLOCK; 1229 if ((flags & FNONBLOCK) == 0) 1230 type |= F_WAIT; 1231 VOP_UNLOCK(vp, 0); 1232 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1233 if (error) { 1234 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1235 FILE_UNUSE(fp, l); 1236 ffree(fp); 1237 fdremove(fdp, indx); 1238 return (error); 1239 } 1240 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1241 fp->f_flag |= FHASLOCK; 1242 } 1243 VOP_UNLOCK(vp, 0); 1244 *retval = indx; 1245 FILE_SET_MATURE(fp); 1246 FILE_UNUSE(fp, l); 1247 return (0); 1248 } 1249 1250 static void 1251 vfs__fhfree(fhandle_t *fhp) 1252 { 1253 size_t fhsize; 1254 1255 if (fhp == NULL) { 1256 return; 1257 } 1258 fhsize = FHANDLE_SIZE(fhp); 1259 kmem_free(fhp, fhsize); 1260 } 1261 1262 /* 1263 * vfs_composefh: compose a filehandle. 1264 */ 1265 1266 int 1267 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1268 { 1269 struct mount *mp; 1270 struct fid *fidp; 1271 int error; 1272 size_t needfhsize; 1273 size_t fidsize; 1274 1275 mp = vp->v_mount; 1276 if (mp->mnt_op->vfs_vptofh == NULL) { 1277 return EOPNOTSUPP; 1278 } 1279 fidp = NULL; 1280 if (*fh_size < FHANDLE_SIZE_MIN) { 1281 fidsize = 0; 1282 } else { 1283 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1284 if (fhp != NULL) { 1285 memset(fhp, 0, *fh_size); 1286 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1287 fidp = &fhp->fh_fid; 1288 } 1289 } 1290 error = VFS_VPTOFH(vp, fidp, &fidsize); 1291 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1292 if (error == 0 && *fh_size < needfhsize) { 1293 error = E2BIG; 1294 } 1295 *fh_size = needfhsize; 1296 return error; 1297 } 1298 1299 int 1300 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1301 { 1302 struct mount *mp; 1303 fhandle_t *fhp; 1304 size_t fhsize; 1305 size_t fidsize; 1306 int error; 1307 1308 *fhpp = NULL; 1309 mp = vp->v_mount; 1310 if (mp->mnt_op->vfs_vptofh == NULL) { 1311 error = EOPNOTSUPP; 1312 goto out; 1313 } 1314 fidsize = 0; 1315 error = VFS_VPTOFH(vp, NULL, &fidsize); 1316 KASSERT(error != 0); 1317 if (error != E2BIG) { 1318 goto out; 1319 } 1320 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1321 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1322 if (fhp == NULL) { 1323 error = ENOMEM; 1324 goto out; 1325 } 1326 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1327 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1328 if (error == 0) { 1329 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1330 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1331 *fhpp = fhp; 1332 } else { 1333 kmem_free(fhp, fhsize); 1334 } 1335 out: 1336 return error; 1337 } 1338 1339 void 1340 vfs_composefh_free(fhandle_t *fhp) 1341 { 1342 1343 vfs__fhfree(fhp); 1344 } 1345 1346 /* 1347 * vfs_fhtovp: lookup a vnode by a filehandle. 1348 */ 1349 1350 int 1351 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1352 { 1353 struct mount *mp; 1354 int error; 1355 1356 *vpp = NULL; 1357 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1358 if (mp == NULL) { 1359 error = ESTALE; 1360 goto out; 1361 } 1362 if (mp->mnt_op->vfs_fhtovp == NULL) { 1363 error = EOPNOTSUPP; 1364 goto out; 1365 } 1366 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1367 out: 1368 return error; 1369 } 1370 1371 /* 1372 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1373 * the needed size. 1374 */ 1375 1376 int 1377 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1378 { 1379 fhandle_t *fhp; 1380 int error; 1381 1382 *fhpp = NULL; 1383 if (fhsize > FHANDLE_SIZE_MAX) { 1384 return EINVAL; 1385 } 1386 if (fhsize < FHANDLE_SIZE_MIN) { 1387 return EINVAL; 1388 } 1389 again: 1390 fhp = kmem_alloc(fhsize, KM_SLEEP); 1391 if (fhp == NULL) { 1392 return ENOMEM; 1393 } 1394 error = copyin(ufhp, fhp, fhsize); 1395 if (error == 0) { 1396 /* XXX this check shouldn't be here */ 1397 if (FHANDLE_SIZE(fhp) == fhsize) { 1398 *fhpp = fhp; 1399 return 0; 1400 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1401 /* 1402 * a kludge for nfsv2 padded handles. 1403 */ 1404 size_t sz; 1405 1406 sz = FHANDLE_SIZE(fhp); 1407 kmem_free(fhp, fhsize); 1408 fhsize = sz; 1409 goto again; 1410 } else { 1411 /* 1412 * userland told us wrong size. 1413 */ 1414 error = EINVAL; 1415 } 1416 } 1417 kmem_free(fhp, fhsize); 1418 return error; 1419 } 1420 1421 void 1422 vfs_copyinfh_free(fhandle_t *fhp) 1423 { 1424 1425 vfs__fhfree(fhp); 1426 } 1427 1428 /* 1429 * Get file handle system call 1430 */ 1431 int 1432 sys___getfh30(struct lwp *l, void *v, register_t *retval __unused) 1433 { 1434 struct sys___getfh30_args /* { 1435 syscallarg(char *) fname; 1436 syscallarg(fhandle_t *) fhp; 1437 syscallarg(size_t *) fh_size; 1438 } */ *uap = v; 1439 struct vnode *vp; 1440 fhandle_t *fh; 1441 int error; 1442 struct nameidata nd; 1443 size_t sz; 1444 size_t usz; 1445 1446 /* 1447 * Must be super user 1448 */ 1449 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1450 0, NULL, NULL, NULL); 1451 if (error) 1452 return (error); 1453 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1454 SCARG(uap, fname), l); 1455 error = namei(&nd); 1456 if (error) 1457 return (error); 1458 vp = nd.ni_vp; 1459 error = vfs_composefh_alloc(vp, &fh); 1460 vput(vp); 1461 if (error != 0) { 1462 goto out; 1463 } 1464 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1465 if (error != 0) { 1466 goto out; 1467 } 1468 sz = FHANDLE_SIZE(fh); 1469 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1470 if (error != 0) { 1471 goto out; 1472 } 1473 if (usz >= sz) { 1474 error = copyout(fh, SCARG(uap, fhp), sz); 1475 } else { 1476 error = E2BIG; 1477 } 1478 out: 1479 vfs_composefh_free(fh); 1480 return (error); 1481 } 1482 1483 /* 1484 * Open a file given a file handle. 1485 * 1486 * Check permissions, allocate an open file structure, 1487 * and call the device open routine if any. 1488 */ 1489 1490 int 1491 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1492 register_t *retval) 1493 { 1494 struct filedesc *fdp = l->l_proc->p_fd; 1495 struct file *fp; 1496 struct vnode *vp = NULL; 1497 struct mount *mp; 1498 kauth_cred_t cred = l->l_cred; 1499 struct file *nfp; 1500 int type, indx, error=0; 1501 struct flock lf; 1502 struct vattr va; 1503 fhandle_t *fh; 1504 int flags; 1505 1506 /* 1507 * Must be super user 1508 */ 1509 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1510 0, NULL, NULL, NULL))) 1511 return (error); 1512 1513 flags = FFLAGS(oflags); 1514 if ((flags & (FREAD | FWRITE)) == 0) 1515 return (EINVAL); 1516 if ((flags & O_CREAT)) 1517 return (EINVAL); 1518 /* falloc() will use the file descriptor for us */ 1519 if ((error = falloc(l, &nfp, &indx)) != 0) 1520 return (error); 1521 fp = nfp; 1522 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1523 if (error != 0) { 1524 goto bad; 1525 } 1526 error = vfs_fhtovp(fh, &vp); 1527 if (error != 0) { 1528 goto bad; 1529 } 1530 1531 /* Now do an effective vn_open */ 1532 1533 if (vp->v_type == VSOCK) { 1534 error = EOPNOTSUPP; 1535 goto bad; 1536 } 1537 if (flags & FREAD) { 1538 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1539 goto bad; 1540 } 1541 if (flags & (FWRITE | O_TRUNC)) { 1542 if (vp->v_type == VDIR) { 1543 error = EISDIR; 1544 goto bad; 1545 } 1546 if ((error = vn_writechk(vp)) != 0 || 1547 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1548 goto bad; 1549 } 1550 if (flags & O_TRUNC) { 1551 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1552 goto bad; 1553 VOP_UNLOCK(vp, 0); /* XXX */ 1554 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1555 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1556 VATTR_NULL(&va); 1557 va.va_size = 0; 1558 error = VOP_SETATTR(vp, &va, cred, l); 1559 vn_finished_write(mp, 0); 1560 if (error) 1561 goto bad; 1562 } 1563 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1564 goto bad; 1565 if (vp->v_type == VREG && 1566 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1567 error = EIO; 1568 goto bad; 1569 } 1570 if (flags & FWRITE) 1571 vp->v_writecount++; 1572 1573 /* done with modified vn_open, now finish what sys_open does. */ 1574 1575 fp->f_flag = flags & FMASK; 1576 fp->f_type = DTYPE_VNODE; 1577 fp->f_ops = &vnops; 1578 fp->f_data = vp; 1579 if (flags & (O_EXLOCK | O_SHLOCK)) { 1580 lf.l_whence = SEEK_SET; 1581 lf.l_start = 0; 1582 lf.l_len = 0; 1583 if (flags & O_EXLOCK) 1584 lf.l_type = F_WRLCK; 1585 else 1586 lf.l_type = F_RDLCK; 1587 type = F_FLOCK; 1588 if ((flags & FNONBLOCK) == 0) 1589 type |= F_WAIT; 1590 VOP_UNLOCK(vp, 0); 1591 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1592 if (error) { 1593 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1594 FILE_UNUSE(fp, l); 1595 ffree(fp); 1596 fdremove(fdp, indx); 1597 return (error); 1598 } 1599 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1600 fp->f_flag |= FHASLOCK; 1601 } 1602 VOP_UNLOCK(vp, 0); 1603 *retval = indx; 1604 FILE_SET_MATURE(fp); 1605 FILE_UNUSE(fp, l); 1606 vfs_copyinfh_free(fh); 1607 return (0); 1608 1609 bad: 1610 FILE_UNUSE(fp, l); 1611 ffree(fp); 1612 fdremove(fdp, indx); 1613 if (vp != NULL) 1614 vput(vp); 1615 vfs_copyinfh_free(fh); 1616 return (error); 1617 } 1618 1619 int 1620 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1621 { 1622 struct sys___fhopen40_args /* { 1623 syscallarg(const void *) fhp; 1624 syscallarg(size_t) fh_size; 1625 syscallarg(int) flags; 1626 } */ *uap = v; 1627 1628 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1629 SCARG(uap, flags), retval); 1630 } 1631 1632 int 1633 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1634 register_t *retval __unused) 1635 { 1636 struct stat sb; 1637 int error; 1638 fhandle_t *fh; 1639 struct vnode *vp; 1640 1641 /* 1642 * Must be super user 1643 */ 1644 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1645 0, NULL, NULL, NULL))) 1646 return (error); 1647 1648 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1649 if (error != 0) { 1650 goto bad; 1651 } 1652 error = vfs_fhtovp(fh, &vp); 1653 if (error != 0) { 1654 goto bad; 1655 } 1656 error = vn_stat(vp, &sb, l); 1657 vput(vp); 1658 if (error) { 1659 goto bad; 1660 } 1661 error = copyout(&sb, sbp, sizeof(sb)); 1662 bad: 1663 vfs_copyinfh_free(fh); 1664 return error; 1665 } 1666 1667 1668 /* ARGSUSED */ 1669 int 1670 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1671 { 1672 struct sys___fhstat40_args /* { 1673 syscallarg(const void *) fhp; 1674 syscallarg(size_t) fh_size; 1675 syscallarg(struct stat *) sb; 1676 } */ *uap = v; 1677 1678 return dofhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), SCARG(uap, sb), 1679 retval); 1680 } 1681 1682 int 1683 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1684 int flags, register_t *retval __unused) 1685 { 1686 struct statvfs *sb = NULL; 1687 fhandle_t *fh; 1688 struct mount *mp; 1689 struct vnode *vp; 1690 int error; 1691 1692 /* 1693 * Must be super user 1694 */ 1695 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1696 0, NULL, NULL, NULL))) 1697 return error; 1698 1699 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1700 if (error != 0) { 1701 goto out; 1702 } 1703 error = vfs_fhtovp(fh, &vp); 1704 if (error != 0) { 1705 goto out; 1706 } 1707 mp = vp->v_mount; 1708 sb = STATVFSBUF_GET(); 1709 if ((error = dostatvfs(mp, sb, l, flags, 1)) != 0) { 1710 vput(vp); 1711 goto out; 1712 } 1713 vput(vp); 1714 error = copyout(sb, buf, sizeof(*sb)); 1715 out: 1716 if (sb != NULL) { 1717 STATVFSBUF_PUT(sb); 1718 } 1719 vfs_copyinfh_free(fh); 1720 return error; 1721 } 1722 1723 /* ARGSUSED */ 1724 int 1725 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1726 { 1727 struct sys___fhstatvfs140_args /* { 1728 syscallarg(const void *) fhp; 1729 syscallarg(size_t) fh_size; 1730 syscallarg(struct statvfs *) buf; 1731 syscallarg(int) flags; 1732 } */ *uap = v; 1733 1734 return dofhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1735 SCARG(uap, buf), SCARG(uap, flags), retval); 1736 } 1737 1738 /* 1739 * Create a special file. 1740 */ 1741 /* ARGSUSED */ 1742 int 1743 sys_mknod(struct lwp *l, void *v, register_t *retval __unused) 1744 { 1745 struct sys_mknod_args /* { 1746 syscallarg(const char *) path; 1747 syscallarg(int) mode; 1748 syscallarg(int) dev; 1749 } */ *uap = v; 1750 struct proc *p = l->l_proc; 1751 struct vnode *vp; 1752 struct mount *mp; 1753 struct vattr vattr; 1754 int error; 1755 int whiteout = 0; 1756 struct nameidata nd; 1757 1758 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1759 0, NULL, NULL, NULL)) != 0) 1760 return (error); 1761 restart: 1762 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1763 if ((error = namei(&nd)) != 0) 1764 return (error); 1765 vp = nd.ni_vp; 1766 if (vp != NULL) 1767 error = EEXIST; 1768 else { 1769 VATTR_NULL(&vattr); 1770 vattr.va_mode = 1771 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1772 vattr.va_rdev = SCARG(uap, dev); 1773 whiteout = 0; 1774 1775 switch (SCARG(uap, mode) & S_IFMT) { 1776 case S_IFMT: /* used by badsect to flag bad sectors */ 1777 vattr.va_type = VBAD; 1778 break; 1779 case S_IFCHR: 1780 vattr.va_type = VCHR; 1781 break; 1782 case S_IFBLK: 1783 vattr.va_type = VBLK; 1784 break; 1785 case S_IFWHT: 1786 whiteout = 1; 1787 break; 1788 default: 1789 error = EINVAL; 1790 break; 1791 } 1792 } 1793 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1794 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1795 if (nd.ni_dvp == vp) 1796 vrele(nd.ni_dvp); 1797 else 1798 vput(nd.ni_dvp); 1799 if (vp) 1800 vrele(vp); 1801 if ((error = vn_start_write(NULL, &mp, 1802 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1803 return (error); 1804 goto restart; 1805 } 1806 if (!error) { 1807 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1808 if (whiteout) { 1809 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1810 if (error) 1811 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1812 vput(nd.ni_dvp); 1813 } else { 1814 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1815 &nd.ni_cnd, &vattr); 1816 if (error == 0) 1817 vput(nd.ni_vp); 1818 } 1819 } else { 1820 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1821 if (nd.ni_dvp == vp) 1822 vrele(nd.ni_dvp); 1823 else 1824 vput(nd.ni_dvp); 1825 if (vp) 1826 vrele(vp); 1827 } 1828 vn_finished_write(mp, 0); 1829 return (error); 1830 } 1831 1832 /* 1833 * Create a named pipe. 1834 */ 1835 /* ARGSUSED */ 1836 int 1837 sys_mkfifo(struct lwp *l, void *v, register_t *retval __unused) 1838 { 1839 struct sys_mkfifo_args /* { 1840 syscallarg(const char *) path; 1841 syscallarg(int) mode; 1842 } */ *uap = v; 1843 struct proc *p = l->l_proc; 1844 struct mount *mp; 1845 struct vattr vattr; 1846 int error; 1847 struct nameidata nd; 1848 1849 restart: 1850 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1851 if ((error = namei(&nd)) != 0) 1852 return (error); 1853 if (nd.ni_vp != NULL) { 1854 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1855 if (nd.ni_dvp == nd.ni_vp) 1856 vrele(nd.ni_dvp); 1857 else 1858 vput(nd.ni_dvp); 1859 vrele(nd.ni_vp); 1860 return (EEXIST); 1861 } 1862 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1863 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1864 if (nd.ni_dvp == nd.ni_vp) 1865 vrele(nd.ni_dvp); 1866 else 1867 vput(nd.ni_dvp); 1868 if (nd.ni_vp) 1869 vrele(nd.ni_vp); 1870 if ((error = vn_start_write(NULL, &mp, 1871 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1872 return (error); 1873 goto restart; 1874 } 1875 VATTR_NULL(&vattr); 1876 vattr.va_type = VFIFO; 1877 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1878 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1879 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1880 if (error == 0) 1881 vput(nd.ni_vp); 1882 vn_finished_write(mp, 0); 1883 return (error); 1884 } 1885 1886 /* 1887 * Make a hard file link. 1888 */ 1889 /* ARGSUSED */ 1890 int 1891 sys_link(struct lwp *l, void *v, register_t *retval __unused) 1892 { 1893 struct sys_link_args /* { 1894 syscallarg(const char *) path; 1895 syscallarg(const char *) link; 1896 } */ *uap = v; 1897 struct vnode *vp; 1898 struct mount *mp; 1899 struct nameidata nd; 1900 int error; 1901 1902 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1903 if ((error = namei(&nd)) != 0) 1904 return (error); 1905 vp = nd.ni_vp; 1906 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1907 vrele(vp); 1908 return (error); 1909 } 1910 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1911 if ((error = namei(&nd)) != 0) 1912 goto out; 1913 if (nd.ni_vp) { 1914 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1915 if (nd.ni_dvp == nd.ni_vp) 1916 vrele(nd.ni_dvp); 1917 else 1918 vput(nd.ni_dvp); 1919 vrele(nd.ni_vp); 1920 error = EEXIST; 1921 goto out; 1922 } 1923 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1924 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1925 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1926 out: 1927 vrele(vp); 1928 vn_finished_write(mp, 0); 1929 return (error); 1930 } 1931 1932 /* 1933 * Make a symbolic link. 1934 */ 1935 /* ARGSUSED */ 1936 int 1937 sys_symlink(struct lwp *l, void *v, register_t *retval __unused) 1938 { 1939 struct sys_symlink_args /* { 1940 syscallarg(const char *) path; 1941 syscallarg(const char *) link; 1942 } */ *uap = v; 1943 struct proc *p = l->l_proc; 1944 struct mount *mp; 1945 struct vattr vattr; 1946 char *path; 1947 int error; 1948 struct nameidata nd; 1949 1950 path = PNBUF_GET(); 1951 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1952 if (error) 1953 goto out; 1954 restart: 1955 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1956 if ((error = namei(&nd)) != 0) 1957 goto out; 1958 if (nd.ni_vp) { 1959 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1960 if (nd.ni_dvp == nd.ni_vp) 1961 vrele(nd.ni_dvp); 1962 else 1963 vput(nd.ni_dvp); 1964 vrele(nd.ni_vp); 1965 error = EEXIST; 1966 goto out; 1967 } 1968 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1969 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1970 if (nd.ni_dvp == nd.ni_vp) 1971 vrele(nd.ni_dvp); 1972 else 1973 vput(nd.ni_dvp); 1974 if ((error = vn_start_write(NULL, &mp, 1975 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1976 return (error); 1977 goto restart; 1978 } 1979 VATTR_NULL(&vattr); 1980 vattr.va_type = VLNK; 1981 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1982 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1983 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1984 if (error == 0) 1985 vput(nd.ni_vp); 1986 vn_finished_write(mp, 0); 1987 out: 1988 PNBUF_PUT(path); 1989 return (error); 1990 } 1991 1992 /* 1993 * Delete a whiteout from the filesystem. 1994 */ 1995 /* ARGSUSED */ 1996 int 1997 sys_undelete(struct lwp *l, void *v, register_t *retval __unused) 1998 { 1999 struct sys_undelete_args /* { 2000 syscallarg(const char *) path; 2001 } */ *uap = v; 2002 int error; 2003 struct mount *mp; 2004 struct nameidata nd; 2005 2006 restart: 2007 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2008 SCARG(uap, path), l); 2009 error = namei(&nd); 2010 if (error) 2011 return (error); 2012 2013 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2014 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2015 if (nd.ni_dvp == nd.ni_vp) 2016 vrele(nd.ni_dvp); 2017 else 2018 vput(nd.ni_dvp); 2019 if (nd.ni_vp) 2020 vrele(nd.ni_vp); 2021 return (EEXIST); 2022 } 2023 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2024 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2025 if (nd.ni_dvp == nd.ni_vp) 2026 vrele(nd.ni_dvp); 2027 else 2028 vput(nd.ni_dvp); 2029 if ((error = vn_start_write(NULL, &mp, 2030 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2031 return (error); 2032 goto restart; 2033 } 2034 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2035 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2036 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2037 vput(nd.ni_dvp); 2038 vn_finished_write(mp, 0); 2039 return (error); 2040 } 2041 2042 /* 2043 * Delete a name from the filesystem. 2044 */ 2045 /* ARGSUSED */ 2046 int 2047 sys_unlink(struct lwp *l, void *v, register_t *retval __unused) 2048 { 2049 struct sys_unlink_args /* { 2050 syscallarg(const char *) path; 2051 } */ *uap = v; 2052 struct mount *mp; 2053 struct vnode *vp; 2054 int error; 2055 struct nameidata nd; 2056 2057 restart: 2058 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2059 SCARG(uap, path), l); 2060 if ((error = namei(&nd)) != 0) 2061 return (error); 2062 vp = nd.ni_vp; 2063 2064 /* 2065 * The root of a mounted filesystem cannot be deleted. 2066 */ 2067 if (vp->v_flag & VROOT) { 2068 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2069 if (nd.ni_dvp == vp) 2070 vrele(nd.ni_dvp); 2071 else 2072 vput(nd.ni_dvp); 2073 vput(vp); 2074 error = EBUSY; 2075 goto out; 2076 } 2077 2078 #if NVERIEXEC > 0 2079 /* Handle remove requests for veriexec entries. */ 2080 if ((error = veriexec_removechk(vp, nd.ni_dirp, l)) != 0) { 2081 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2082 if (nd.ni_dvp == vp) 2083 vrele(nd.ni_dvp); 2084 else 2085 vput(nd.ni_dvp); 2086 vput(vp); 2087 goto out; 2088 } 2089 #endif /* NVERIEXEC > 0 */ 2090 2091 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2092 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2093 if (nd.ni_dvp == vp) 2094 vrele(nd.ni_dvp); 2095 else 2096 vput(nd.ni_dvp); 2097 vput(vp); 2098 if ((error = vn_start_write(NULL, &mp, 2099 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2100 return (error); 2101 goto restart; 2102 } 2103 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2104 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2105 #ifdef FILEASSOC 2106 (void)fileassoc_file_delete(nd.ni_vp); 2107 #endif /* FILEASSOC */ 2108 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2109 vn_finished_write(mp, 0); 2110 out: 2111 return (error); 2112 } 2113 2114 /* 2115 * Reposition read/write file offset. 2116 */ 2117 int 2118 sys_lseek(struct lwp *l, void *v, register_t *retval) 2119 { 2120 struct sys_lseek_args /* { 2121 syscallarg(int) fd; 2122 syscallarg(int) pad; 2123 syscallarg(off_t) offset; 2124 syscallarg(int) whence; 2125 } */ *uap = v; 2126 struct proc *p = l->l_proc; 2127 kauth_cred_t cred = l->l_cred; 2128 struct filedesc *fdp = p->p_fd; 2129 struct file *fp; 2130 struct vnode *vp; 2131 struct vattr vattr; 2132 off_t newoff; 2133 int error; 2134 2135 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2136 return (EBADF); 2137 2138 FILE_USE(fp); 2139 2140 vp = (struct vnode *)fp->f_data; 2141 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2142 error = ESPIPE; 2143 goto out; 2144 } 2145 2146 switch (SCARG(uap, whence)) { 2147 case SEEK_CUR: 2148 newoff = fp->f_offset + SCARG(uap, offset); 2149 break; 2150 case SEEK_END: 2151 error = VOP_GETATTR(vp, &vattr, cred, l); 2152 if (error) 2153 goto out; 2154 newoff = SCARG(uap, offset) + vattr.va_size; 2155 break; 2156 case SEEK_SET: 2157 newoff = SCARG(uap, offset); 2158 break; 2159 default: 2160 error = EINVAL; 2161 goto out; 2162 } 2163 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2164 goto out; 2165 2166 *(off_t *)retval = fp->f_offset = newoff; 2167 out: 2168 FILE_UNUSE(fp, l); 2169 return (error); 2170 } 2171 2172 /* 2173 * Positional read system call. 2174 */ 2175 int 2176 sys_pread(struct lwp *l, void *v, register_t *retval) 2177 { 2178 struct sys_pread_args /* { 2179 syscallarg(int) fd; 2180 syscallarg(void *) buf; 2181 syscallarg(size_t) nbyte; 2182 syscallarg(off_t) offset; 2183 } */ *uap = v; 2184 struct proc *p = l->l_proc; 2185 struct filedesc *fdp = p->p_fd; 2186 struct file *fp; 2187 struct vnode *vp; 2188 off_t offset; 2189 int error, fd = SCARG(uap, fd); 2190 2191 if ((fp = fd_getfile(fdp, fd)) == NULL) 2192 return (EBADF); 2193 2194 if ((fp->f_flag & FREAD) == 0) { 2195 simple_unlock(&fp->f_slock); 2196 return (EBADF); 2197 } 2198 2199 FILE_USE(fp); 2200 2201 vp = (struct vnode *)fp->f_data; 2202 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2203 error = ESPIPE; 2204 goto out; 2205 } 2206 2207 offset = SCARG(uap, offset); 2208 2209 /* 2210 * XXX This works because no file systems actually 2211 * XXX take any action on the seek operation. 2212 */ 2213 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2214 goto out; 2215 2216 /* dofileread() will unuse the descriptor for us */ 2217 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2218 &offset, 0, retval)); 2219 2220 out: 2221 FILE_UNUSE(fp, l); 2222 return (error); 2223 } 2224 2225 /* 2226 * Positional scatter read system call. 2227 */ 2228 int 2229 sys_preadv(struct lwp *l, void *v, register_t *retval) 2230 { 2231 struct sys_preadv_args /* { 2232 syscallarg(int) fd; 2233 syscallarg(const struct iovec *) iovp; 2234 syscallarg(int) iovcnt; 2235 syscallarg(off_t) offset; 2236 } */ *uap = v; 2237 struct proc *p = l->l_proc; 2238 struct filedesc *fdp = p->p_fd; 2239 struct file *fp; 2240 struct vnode *vp; 2241 off_t offset; 2242 int error, fd = SCARG(uap, fd); 2243 2244 if ((fp = fd_getfile(fdp, fd)) == NULL) 2245 return (EBADF); 2246 2247 if ((fp->f_flag & FREAD) == 0) { 2248 simple_unlock(&fp->f_slock); 2249 return (EBADF); 2250 } 2251 2252 FILE_USE(fp); 2253 2254 vp = (struct vnode *)fp->f_data; 2255 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2256 error = ESPIPE; 2257 goto out; 2258 } 2259 2260 offset = SCARG(uap, offset); 2261 2262 /* 2263 * XXX This works because no file systems actually 2264 * XXX take any action on the seek operation. 2265 */ 2266 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2267 goto out; 2268 2269 /* dofilereadv() will unuse the descriptor for us */ 2270 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2271 &offset, 0, retval)); 2272 2273 out: 2274 FILE_UNUSE(fp, l); 2275 return (error); 2276 } 2277 2278 /* 2279 * Positional write system call. 2280 */ 2281 int 2282 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2283 { 2284 struct sys_pwrite_args /* { 2285 syscallarg(int) fd; 2286 syscallarg(const void *) buf; 2287 syscallarg(size_t) nbyte; 2288 syscallarg(off_t) offset; 2289 } */ *uap = v; 2290 struct proc *p = l->l_proc; 2291 struct filedesc *fdp = p->p_fd; 2292 struct file *fp; 2293 struct vnode *vp; 2294 off_t offset; 2295 int error, fd = SCARG(uap, fd); 2296 2297 if ((fp = fd_getfile(fdp, fd)) == NULL) 2298 return (EBADF); 2299 2300 if ((fp->f_flag & FWRITE) == 0) { 2301 simple_unlock(&fp->f_slock); 2302 return (EBADF); 2303 } 2304 2305 FILE_USE(fp); 2306 2307 vp = (struct vnode *)fp->f_data; 2308 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2309 error = ESPIPE; 2310 goto out; 2311 } 2312 2313 offset = SCARG(uap, offset); 2314 2315 /* 2316 * XXX This works because no file systems actually 2317 * XXX take any action on the seek operation. 2318 */ 2319 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2320 goto out; 2321 2322 /* dofilewrite() will unuse the descriptor for us */ 2323 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2324 &offset, 0, retval)); 2325 2326 out: 2327 FILE_UNUSE(fp, l); 2328 return (error); 2329 } 2330 2331 /* 2332 * Positional gather write system call. 2333 */ 2334 int 2335 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2336 { 2337 struct sys_pwritev_args /* { 2338 syscallarg(int) fd; 2339 syscallarg(const struct iovec *) iovp; 2340 syscallarg(int) iovcnt; 2341 syscallarg(off_t) offset; 2342 } */ *uap = v; 2343 struct proc *p = l->l_proc; 2344 struct filedesc *fdp = p->p_fd; 2345 struct file *fp; 2346 struct vnode *vp; 2347 off_t offset; 2348 int error, fd = SCARG(uap, fd); 2349 2350 if ((fp = fd_getfile(fdp, fd)) == NULL) 2351 return (EBADF); 2352 2353 if ((fp->f_flag & FWRITE) == 0) { 2354 simple_unlock(&fp->f_slock); 2355 return (EBADF); 2356 } 2357 2358 FILE_USE(fp); 2359 2360 vp = (struct vnode *)fp->f_data; 2361 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2362 error = ESPIPE; 2363 goto out; 2364 } 2365 2366 offset = SCARG(uap, offset); 2367 2368 /* 2369 * XXX This works because no file systems actually 2370 * XXX take any action on the seek operation. 2371 */ 2372 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2373 goto out; 2374 2375 /* dofilewritev() will unuse the descriptor for us */ 2376 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2377 &offset, 0, retval)); 2378 2379 out: 2380 FILE_UNUSE(fp, l); 2381 return (error); 2382 } 2383 2384 /* 2385 * Check access permissions. 2386 */ 2387 int 2388 sys_access(struct lwp *l, void *v, register_t *retval __unused) 2389 { 2390 struct sys_access_args /* { 2391 syscallarg(const char *) path; 2392 syscallarg(int) flags; 2393 } */ *uap = v; 2394 kauth_cred_t cred; 2395 struct vnode *vp; 2396 int error, flags; 2397 struct nameidata nd; 2398 2399 cred = kauth_cred_dup(l->l_cred); 2400 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2401 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2402 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2403 SCARG(uap, path), l); 2404 /* Override default credentials */ 2405 nd.ni_cnd.cn_cred = cred; 2406 if ((error = namei(&nd)) != 0) 2407 goto out; 2408 vp = nd.ni_vp; 2409 2410 /* Flags == 0 means only check for existence. */ 2411 if (SCARG(uap, flags)) { 2412 flags = 0; 2413 if (SCARG(uap, flags) & R_OK) 2414 flags |= VREAD; 2415 if (SCARG(uap, flags) & W_OK) 2416 flags |= VWRITE; 2417 if (SCARG(uap, flags) & X_OK) 2418 flags |= VEXEC; 2419 2420 error = VOP_ACCESS(vp, flags, cred, l); 2421 if (!error && (flags & VWRITE)) 2422 error = vn_writechk(vp); 2423 } 2424 vput(vp); 2425 out: 2426 kauth_cred_free(cred); 2427 return (error); 2428 } 2429 2430 /* 2431 * Get file status; this version follows links. 2432 */ 2433 /* ARGSUSED */ 2434 int 2435 sys___stat30(struct lwp *l, void *v, register_t *retval __unused) 2436 { 2437 struct sys___stat30_args /* { 2438 syscallarg(const char *) path; 2439 syscallarg(struct stat *) ub; 2440 } */ *uap = v; 2441 struct stat sb; 2442 int error; 2443 struct nameidata nd; 2444 2445 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2446 SCARG(uap, path), l); 2447 if ((error = namei(&nd)) != 0) 2448 return (error); 2449 error = vn_stat(nd.ni_vp, &sb, l); 2450 vput(nd.ni_vp); 2451 if (error) 2452 return (error); 2453 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2454 return (error); 2455 } 2456 2457 /* 2458 * Get file status; this version does not follow links. 2459 */ 2460 /* ARGSUSED */ 2461 int 2462 sys___lstat30(struct lwp *l, void *v, register_t *retval __unused) 2463 { 2464 struct sys___lstat30_args /* { 2465 syscallarg(const char *) path; 2466 syscallarg(struct stat *) ub; 2467 } */ *uap = v; 2468 struct stat sb; 2469 int error; 2470 struct nameidata nd; 2471 2472 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2473 SCARG(uap, path), l); 2474 if ((error = namei(&nd)) != 0) 2475 return (error); 2476 error = vn_stat(nd.ni_vp, &sb, l); 2477 vput(nd.ni_vp); 2478 if (error) 2479 return (error); 2480 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2481 return (error); 2482 } 2483 2484 /* 2485 * Get configurable pathname variables. 2486 */ 2487 /* ARGSUSED */ 2488 int 2489 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2490 { 2491 struct sys_pathconf_args /* { 2492 syscallarg(const char *) path; 2493 syscallarg(int) name; 2494 } */ *uap = v; 2495 int error; 2496 struct nameidata nd; 2497 2498 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2499 SCARG(uap, path), l); 2500 if ((error = namei(&nd)) != 0) 2501 return (error); 2502 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2503 vput(nd.ni_vp); 2504 return (error); 2505 } 2506 2507 /* 2508 * Return target name of a symbolic link. 2509 */ 2510 /* ARGSUSED */ 2511 int 2512 sys_readlink(struct lwp *l, void *v, register_t *retval) 2513 { 2514 struct sys_readlink_args /* { 2515 syscallarg(const char *) path; 2516 syscallarg(char *) buf; 2517 syscallarg(size_t) count; 2518 } */ *uap = v; 2519 struct vnode *vp; 2520 struct iovec aiov; 2521 struct uio auio; 2522 int error; 2523 struct nameidata nd; 2524 2525 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2526 SCARG(uap, path), l); 2527 if ((error = namei(&nd)) != 0) 2528 return (error); 2529 vp = nd.ni_vp; 2530 if (vp->v_type != VLNK) 2531 error = EINVAL; 2532 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2533 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2534 aiov.iov_base = SCARG(uap, buf); 2535 aiov.iov_len = SCARG(uap, count); 2536 auio.uio_iov = &aiov; 2537 auio.uio_iovcnt = 1; 2538 auio.uio_offset = 0; 2539 auio.uio_rw = UIO_READ; 2540 KASSERT(l == curlwp); 2541 auio.uio_vmspace = l->l_proc->p_vmspace; 2542 auio.uio_resid = SCARG(uap, count); 2543 error = VOP_READLINK(vp, &auio, l->l_cred); 2544 } 2545 vput(vp); 2546 *retval = SCARG(uap, count) - auio.uio_resid; 2547 return (error); 2548 } 2549 2550 /* 2551 * Change flags of a file given a path name. 2552 */ 2553 /* ARGSUSED */ 2554 int 2555 sys_chflags(struct lwp *l, void *v, register_t *retval __unused) 2556 { 2557 struct sys_chflags_args /* { 2558 syscallarg(const char *) path; 2559 syscallarg(u_long) flags; 2560 } */ *uap = v; 2561 struct vnode *vp; 2562 int error; 2563 struct nameidata nd; 2564 2565 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2566 if ((error = namei(&nd)) != 0) 2567 return (error); 2568 vp = nd.ni_vp; 2569 error = change_flags(vp, SCARG(uap, flags), l); 2570 vput(vp); 2571 return (error); 2572 } 2573 2574 /* 2575 * Change flags of a file given a file descriptor. 2576 */ 2577 /* ARGSUSED */ 2578 int 2579 sys_fchflags(struct lwp *l, void *v, register_t *retval __unused) 2580 { 2581 struct sys_fchflags_args /* { 2582 syscallarg(int) fd; 2583 syscallarg(u_long) flags; 2584 } */ *uap = v; 2585 struct proc *p = l->l_proc; 2586 struct vnode *vp; 2587 struct file *fp; 2588 int error; 2589 2590 /* getvnode() will use the descriptor for us */ 2591 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2592 return (error); 2593 vp = (struct vnode *)fp->f_data; 2594 error = change_flags(vp, SCARG(uap, flags), l); 2595 VOP_UNLOCK(vp, 0); 2596 FILE_UNUSE(fp, l); 2597 return (error); 2598 } 2599 2600 /* 2601 * Change flags of a file given a path name; this version does 2602 * not follow links. 2603 */ 2604 int 2605 sys_lchflags(struct lwp *l, void *v, register_t *retval __unused) 2606 { 2607 struct sys_lchflags_args /* { 2608 syscallarg(const char *) path; 2609 syscallarg(u_long) flags; 2610 } */ *uap = v; 2611 struct vnode *vp; 2612 int error; 2613 struct nameidata nd; 2614 2615 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2616 if ((error = namei(&nd)) != 0) 2617 return (error); 2618 vp = nd.ni_vp; 2619 error = change_flags(vp, SCARG(uap, flags), l); 2620 vput(vp); 2621 return (error); 2622 } 2623 2624 /* 2625 * Common routine to change flags of a file. 2626 */ 2627 int 2628 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2629 { 2630 struct mount *mp; 2631 struct vattr vattr; 2632 int error; 2633 2634 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2635 return (error); 2636 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2637 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2638 /* 2639 * Non-superusers cannot change the flags on devices, even if they 2640 * own them. 2641 */ 2642 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2643 &l->l_acflag) != 0) { 2644 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2645 goto out; 2646 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2647 error = EINVAL; 2648 goto out; 2649 } 2650 } 2651 VATTR_NULL(&vattr); 2652 vattr.va_flags = flags; 2653 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2654 out: 2655 vn_finished_write(mp, 0); 2656 return (error); 2657 } 2658 2659 /* 2660 * Change mode of a file given path name; this version follows links. 2661 */ 2662 /* ARGSUSED */ 2663 int 2664 sys_chmod(struct lwp *l, void *v, register_t *retval __unused) 2665 { 2666 struct sys_chmod_args /* { 2667 syscallarg(const char *) path; 2668 syscallarg(int) mode; 2669 } */ *uap = v; 2670 int error; 2671 struct nameidata nd; 2672 2673 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2674 if ((error = namei(&nd)) != 0) 2675 return (error); 2676 2677 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2678 2679 vrele(nd.ni_vp); 2680 return (error); 2681 } 2682 2683 /* 2684 * Change mode of a file given a file descriptor. 2685 */ 2686 /* ARGSUSED */ 2687 int 2688 sys_fchmod(struct lwp *l, void *v, register_t *retval __unused) 2689 { 2690 struct sys_fchmod_args /* { 2691 syscallarg(int) fd; 2692 syscallarg(int) mode; 2693 } */ *uap = v; 2694 struct proc *p = l->l_proc; 2695 struct file *fp; 2696 int error; 2697 2698 /* getvnode() will use the descriptor for us */ 2699 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2700 return (error); 2701 2702 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2703 FILE_UNUSE(fp, l); 2704 return (error); 2705 } 2706 2707 /* 2708 * Change mode of a file given path name; this version does not follow links. 2709 */ 2710 /* ARGSUSED */ 2711 int 2712 sys_lchmod(struct lwp *l, void *v, register_t *retval __unused) 2713 { 2714 struct sys_lchmod_args /* { 2715 syscallarg(const char *) path; 2716 syscallarg(int) mode; 2717 } */ *uap = v; 2718 int error; 2719 struct nameidata nd; 2720 2721 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2722 if ((error = namei(&nd)) != 0) 2723 return (error); 2724 2725 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2726 2727 vrele(nd.ni_vp); 2728 return (error); 2729 } 2730 2731 /* 2732 * Common routine to set mode given a vnode. 2733 */ 2734 static int 2735 change_mode(struct vnode *vp, int mode, struct lwp *l) 2736 { 2737 struct mount *mp; 2738 struct vattr vattr; 2739 int error; 2740 2741 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2742 return (error); 2743 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2744 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2745 VATTR_NULL(&vattr); 2746 vattr.va_mode = mode & ALLPERMS; 2747 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2748 VOP_UNLOCK(vp, 0); 2749 vn_finished_write(mp, 0); 2750 return (error); 2751 } 2752 2753 /* 2754 * Set ownership given a path name; this version follows links. 2755 */ 2756 /* ARGSUSED */ 2757 int 2758 sys_chown(struct lwp *l, void *v, register_t *retval __unused) 2759 { 2760 struct sys_chown_args /* { 2761 syscallarg(const char *) path; 2762 syscallarg(uid_t) uid; 2763 syscallarg(gid_t) gid; 2764 } */ *uap = v; 2765 int error; 2766 struct nameidata nd; 2767 2768 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2769 if ((error = namei(&nd)) != 0) 2770 return (error); 2771 2772 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2773 2774 vrele(nd.ni_vp); 2775 return (error); 2776 } 2777 2778 /* 2779 * Set ownership given a path name; this version follows links. 2780 * Provides POSIX semantics. 2781 */ 2782 /* ARGSUSED */ 2783 int 2784 sys___posix_chown(struct lwp *l, void *v, register_t *retval __unused) 2785 { 2786 struct sys_chown_args /* { 2787 syscallarg(const char *) path; 2788 syscallarg(uid_t) uid; 2789 syscallarg(gid_t) gid; 2790 } */ *uap = v; 2791 int error; 2792 struct nameidata nd; 2793 2794 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2795 if ((error = namei(&nd)) != 0) 2796 return (error); 2797 2798 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2799 2800 vrele(nd.ni_vp); 2801 return (error); 2802 } 2803 2804 /* 2805 * Set ownership given a file descriptor. 2806 */ 2807 /* ARGSUSED */ 2808 int 2809 sys_fchown(struct lwp *l, void *v, register_t *retval __unused) 2810 { 2811 struct sys_fchown_args /* { 2812 syscallarg(int) fd; 2813 syscallarg(uid_t) uid; 2814 syscallarg(gid_t) gid; 2815 } */ *uap = v; 2816 struct proc *p = l->l_proc; 2817 int error; 2818 struct file *fp; 2819 2820 /* getvnode() will use the descriptor for us */ 2821 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2822 return (error); 2823 2824 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2825 SCARG(uap, gid), l, 0); 2826 FILE_UNUSE(fp, l); 2827 return (error); 2828 } 2829 2830 /* 2831 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2832 */ 2833 /* ARGSUSED */ 2834 int 2835 sys___posix_fchown(struct lwp *l, void *v, register_t *retval __unused) 2836 { 2837 struct sys_fchown_args /* { 2838 syscallarg(int) fd; 2839 syscallarg(uid_t) uid; 2840 syscallarg(gid_t) gid; 2841 } */ *uap = v; 2842 struct proc *p = l->l_proc; 2843 int error; 2844 struct file *fp; 2845 2846 /* getvnode() will use the descriptor for us */ 2847 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2848 return (error); 2849 2850 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2851 SCARG(uap, gid), l, 1); 2852 FILE_UNUSE(fp, l); 2853 return (error); 2854 } 2855 2856 /* 2857 * Set ownership given a path name; this version does not follow links. 2858 */ 2859 /* ARGSUSED */ 2860 int 2861 sys_lchown(struct lwp *l, void *v, register_t *retval __unused) 2862 { 2863 struct sys_lchown_args /* { 2864 syscallarg(const char *) path; 2865 syscallarg(uid_t) uid; 2866 syscallarg(gid_t) gid; 2867 } */ *uap = v; 2868 int error; 2869 struct nameidata nd; 2870 2871 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2872 if ((error = namei(&nd)) != 0) 2873 return (error); 2874 2875 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2876 2877 vrele(nd.ni_vp); 2878 return (error); 2879 } 2880 2881 /* 2882 * Set ownership given a path name; this version does not follow links. 2883 * Provides POSIX/XPG semantics. 2884 */ 2885 /* ARGSUSED */ 2886 int 2887 sys___posix_lchown(struct lwp *l, void *v, register_t *retval __unused) 2888 { 2889 struct sys_lchown_args /* { 2890 syscallarg(const char *) path; 2891 syscallarg(uid_t) uid; 2892 syscallarg(gid_t) gid; 2893 } */ *uap = v; 2894 int error; 2895 struct nameidata nd; 2896 2897 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2898 if ((error = namei(&nd)) != 0) 2899 return (error); 2900 2901 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2902 2903 vrele(nd.ni_vp); 2904 return (error); 2905 } 2906 2907 /* 2908 * Common routine to set ownership given a vnode. 2909 */ 2910 static int 2911 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2912 int posix_semantics) 2913 { 2914 struct mount *mp; 2915 struct vattr vattr; 2916 mode_t newmode; 2917 int error; 2918 2919 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2920 return (error); 2921 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2922 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2923 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2924 goto out; 2925 2926 #define CHANGED(x) ((int)(x) != -1) 2927 newmode = vattr.va_mode; 2928 if (posix_semantics) { 2929 /* 2930 * POSIX/XPG semantics: if the caller is not the super-user, 2931 * clear set-user-id and set-group-id bits. Both POSIX and 2932 * the XPG consider the behaviour for calls by the super-user 2933 * implementation-defined; we leave the set-user-id and set- 2934 * group-id settings intact in that case. 2935 */ 2936 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2937 NULL) != 0) 2938 newmode &= ~(S_ISUID | S_ISGID); 2939 } else { 2940 /* 2941 * NetBSD semantics: when changing owner and/or group, 2942 * clear the respective bit(s). 2943 */ 2944 if (CHANGED(uid)) 2945 newmode &= ~S_ISUID; 2946 if (CHANGED(gid)) 2947 newmode &= ~S_ISGID; 2948 } 2949 /* Update va_mode iff altered. */ 2950 if (vattr.va_mode == newmode) 2951 newmode = VNOVAL; 2952 2953 VATTR_NULL(&vattr); 2954 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2955 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2956 vattr.va_mode = newmode; 2957 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2958 #undef CHANGED 2959 2960 out: 2961 VOP_UNLOCK(vp, 0); 2962 vn_finished_write(mp, 0); 2963 return (error); 2964 } 2965 2966 /* 2967 * Set the access and modification times given a path name; this 2968 * version follows links. 2969 */ 2970 /* ARGSUSED */ 2971 int 2972 sys_utimes(struct lwp *l, void *v, register_t *retval __unused) 2973 { 2974 struct sys_utimes_args /* { 2975 syscallarg(const char *) path; 2976 syscallarg(const struct timeval *) tptr; 2977 } */ *uap = v; 2978 int error; 2979 struct nameidata nd; 2980 2981 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2982 if ((error = namei(&nd)) != 0) 2983 return (error); 2984 2985 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2986 2987 vrele(nd.ni_vp); 2988 return (error); 2989 } 2990 2991 /* 2992 * Set the access and modification times given a file descriptor. 2993 */ 2994 /* ARGSUSED */ 2995 int 2996 sys_futimes(struct lwp *l, void *v, register_t *retval __unused) 2997 { 2998 struct sys_futimes_args /* { 2999 syscallarg(int) fd; 3000 syscallarg(const struct timeval *) tptr; 3001 } */ *uap = v; 3002 struct proc *p = l->l_proc; 3003 int error; 3004 struct file *fp; 3005 3006 /* getvnode() will use the descriptor for us */ 3007 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3008 return (error); 3009 3010 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3011 FILE_UNUSE(fp, l); 3012 return (error); 3013 } 3014 3015 /* 3016 * Set the access and modification times given a path name; this 3017 * version does not follow links. 3018 */ 3019 /* ARGSUSED */ 3020 int 3021 sys_lutimes(struct lwp *l, void *v, register_t *retval __unused) 3022 { 3023 struct sys_lutimes_args /* { 3024 syscallarg(const char *) path; 3025 syscallarg(const struct timeval *) tptr; 3026 } */ *uap = v; 3027 int error; 3028 struct nameidata nd; 3029 3030 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3031 if ((error = namei(&nd)) != 0) 3032 return (error); 3033 3034 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3035 3036 vrele(nd.ni_vp); 3037 return (error); 3038 } 3039 3040 /* 3041 * Common routine to set access and modification times given a vnode. 3042 */ 3043 static int 3044 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3045 { 3046 struct mount *mp; 3047 struct vattr vattr; 3048 int error; 3049 3050 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3051 return (error); 3052 VATTR_NULL(&vattr); 3053 if (tptr == NULL) { 3054 nanotime(&vattr.va_atime); 3055 vattr.va_mtime = vattr.va_atime; 3056 vattr.va_vaflags |= VA_UTIMES_NULL; 3057 } else { 3058 struct timeval tv[2]; 3059 3060 error = copyin(tptr, tv, sizeof(tv)); 3061 if (error) 3062 goto out; 3063 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3064 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3065 } 3066 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3067 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3068 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3069 VOP_UNLOCK(vp, 0); 3070 out: 3071 vn_finished_write(mp, 0); 3072 return (error); 3073 } 3074 3075 /* 3076 * Truncate a file given its path name. 3077 */ 3078 /* ARGSUSED */ 3079 int 3080 sys_truncate(struct lwp *l, void *v, register_t *retval __unused) 3081 { 3082 struct sys_truncate_args /* { 3083 syscallarg(const char *) path; 3084 syscallarg(int) pad; 3085 syscallarg(off_t) length; 3086 } */ *uap = v; 3087 struct vnode *vp; 3088 struct mount *mp; 3089 struct vattr vattr; 3090 int error; 3091 struct nameidata nd; 3092 3093 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3094 if ((error = namei(&nd)) != 0) 3095 return (error); 3096 vp = nd.ni_vp; 3097 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3098 vrele(vp); 3099 return (error); 3100 } 3101 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3102 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3103 if (vp->v_type == VDIR) 3104 error = EISDIR; 3105 else if ((error = vn_writechk(vp)) == 0 && 3106 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3107 VATTR_NULL(&vattr); 3108 vattr.va_size = SCARG(uap, length); 3109 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3110 } 3111 vput(vp); 3112 vn_finished_write(mp, 0); 3113 return (error); 3114 } 3115 3116 /* 3117 * Truncate a file given a file descriptor. 3118 */ 3119 /* ARGSUSED */ 3120 int 3121 sys_ftruncate(struct lwp *l, void *v, register_t *retval __unused) 3122 { 3123 struct sys_ftruncate_args /* { 3124 syscallarg(int) fd; 3125 syscallarg(int) pad; 3126 syscallarg(off_t) length; 3127 } */ *uap = v; 3128 struct proc *p = l->l_proc; 3129 struct mount *mp; 3130 struct vattr vattr; 3131 struct vnode *vp; 3132 struct file *fp; 3133 int error; 3134 3135 /* getvnode() will use the descriptor for us */ 3136 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3137 return (error); 3138 if ((fp->f_flag & FWRITE) == 0) { 3139 error = EINVAL; 3140 goto out; 3141 } 3142 vp = (struct vnode *)fp->f_data; 3143 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3144 FILE_UNUSE(fp, l); 3145 return (error); 3146 } 3147 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3148 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3149 if (vp->v_type == VDIR) 3150 error = EISDIR; 3151 else if ((error = vn_writechk(vp)) == 0) { 3152 VATTR_NULL(&vattr); 3153 vattr.va_size = SCARG(uap, length); 3154 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3155 } 3156 VOP_UNLOCK(vp, 0); 3157 vn_finished_write(mp, 0); 3158 out: 3159 FILE_UNUSE(fp, l); 3160 return (error); 3161 } 3162 3163 /* 3164 * Sync an open file. 3165 */ 3166 /* ARGSUSED */ 3167 int 3168 sys_fsync(struct lwp *l, void *v, register_t *retval __unused) 3169 { 3170 struct sys_fsync_args /* { 3171 syscallarg(int) fd; 3172 } */ *uap = v; 3173 struct proc *p = l->l_proc; 3174 struct vnode *vp; 3175 struct mount *mp; 3176 struct file *fp; 3177 int error; 3178 3179 /* getvnode() will use the descriptor for us */ 3180 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3181 return (error); 3182 vp = (struct vnode *)fp->f_data; 3183 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3184 FILE_UNUSE(fp, l); 3185 return (error); 3186 } 3187 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3188 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3189 if (error == 0 && bioops.io_fsync != NULL && 3190 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3191 (*bioops.io_fsync)(vp, 0); 3192 VOP_UNLOCK(vp, 0); 3193 vn_finished_write(mp, 0); 3194 FILE_UNUSE(fp, l); 3195 return (error); 3196 } 3197 3198 /* 3199 * Sync a range of file data. API modeled after that found in AIX. 3200 * 3201 * FDATASYNC indicates that we need only save enough metadata to be able 3202 * to re-read the written data. Note we duplicate AIX's requirement that 3203 * the file be open for writing. 3204 */ 3205 /* ARGSUSED */ 3206 int 3207 sys_fsync_range(struct lwp *l, void *v, register_t *retval __unused) 3208 { 3209 struct sys_fsync_range_args /* { 3210 syscallarg(int) fd; 3211 syscallarg(int) flags; 3212 syscallarg(off_t) start; 3213 syscallarg(off_t) length; 3214 } */ *uap = v; 3215 struct proc *p = l->l_proc; 3216 struct vnode *vp; 3217 struct file *fp; 3218 int flags, nflags; 3219 off_t s, e, len; 3220 int error; 3221 3222 /* getvnode() will use the descriptor for us */ 3223 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3224 return (error); 3225 3226 if ((fp->f_flag & FWRITE) == 0) { 3227 FILE_UNUSE(fp, l); 3228 return (EBADF); 3229 } 3230 3231 flags = SCARG(uap, flags); 3232 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3233 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3234 return (EINVAL); 3235 } 3236 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3237 if (flags & FDATASYNC) 3238 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3239 else 3240 nflags = FSYNC_WAIT; 3241 if (flags & FDISKSYNC) 3242 nflags |= FSYNC_CACHE; 3243 3244 len = SCARG(uap, length); 3245 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3246 if (len) { 3247 s = SCARG(uap, start); 3248 e = s + len; 3249 if (e < s) { 3250 FILE_UNUSE(fp, l); 3251 return (EINVAL); 3252 } 3253 } else { 3254 e = 0; 3255 s = 0; 3256 } 3257 3258 vp = (struct vnode *)fp->f_data; 3259 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3260 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3261 3262 if (error == 0 && bioops.io_fsync != NULL && 3263 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3264 (*bioops.io_fsync)(vp, nflags); 3265 3266 VOP_UNLOCK(vp, 0); 3267 FILE_UNUSE(fp, l); 3268 return (error); 3269 } 3270 3271 /* 3272 * Sync the data of an open file. 3273 */ 3274 /* ARGSUSED */ 3275 int 3276 sys_fdatasync(struct lwp *l, void *v, register_t *retval __unused) 3277 { 3278 struct sys_fdatasync_args /* { 3279 syscallarg(int) fd; 3280 } */ *uap = v; 3281 struct proc *p = l->l_proc; 3282 struct vnode *vp; 3283 struct file *fp; 3284 int error; 3285 3286 /* getvnode() will use the descriptor for us */ 3287 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3288 return (error); 3289 if ((fp->f_flag & FWRITE) == 0) { 3290 FILE_UNUSE(fp, l); 3291 return (EBADF); 3292 } 3293 vp = (struct vnode *)fp->f_data; 3294 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3295 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3296 VOP_UNLOCK(vp, 0); 3297 FILE_UNUSE(fp, l); 3298 return (error); 3299 } 3300 3301 /* 3302 * Rename files, (standard) BSD semantics frontend. 3303 */ 3304 /* ARGSUSED */ 3305 int 3306 sys_rename(struct lwp *l, void *v, register_t *retval __unused) 3307 { 3308 struct sys_rename_args /* { 3309 syscallarg(const char *) from; 3310 syscallarg(const char *) to; 3311 } */ *uap = v; 3312 3313 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3314 } 3315 3316 /* 3317 * Rename files, POSIX semantics frontend. 3318 */ 3319 /* ARGSUSED */ 3320 int 3321 sys___posix_rename(struct lwp *l, void *v, register_t *retval __unused) 3322 { 3323 struct sys___posix_rename_args /* { 3324 syscallarg(const char *) from; 3325 syscallarg(const char *) to; 3326 } */ *uap = v; 3327 3328 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3329 } 3330 3331 /* 3332 * Rename files. Source and destination must either both be directories, 3333 * or both not be directories. If target is a directory, it must be empty. 3334 * If `from' and `to' refer to the same object, the value of the `retain' 3335 * argument is used to determine whether `from' will be 3336 * 3337 * (retain == 0) deleted unless `from' and `to' refer to the same 3338 * object in the file system's name space (BSD). 3339 * (retain == 1) always retained (POSIX). 3340 */ 3341 static int 3342 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3343 { 3344 struct mount *mp = NULL; 3345 struct vnode *tvp, *fvp, *tdvp; 3346 struct nameidata fromnd, tond; 3347 struct proc *p; 3348 int error; 3349 3350 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3351 from, l); 3352 if ((error = namei(&fromnd)) != 0) 3353 return (error); 3354 fvp = fromnd.ni_vp; 3355 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3356 if (error != 0) { 3357 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3358 vrele(fromnd.ni_dvp); 3359 vrele(fvp); 3360 if (fromnd.ni_startdir) 3361 vrele(fromnd.ni_startdir); 3362 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3363 return (error); 3364 } 3365 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3366 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3367 if ((error = namei(&tond)) != 0) { 3368 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3369 vrele(fromnd.ni_dvp); 3370 vrele(fvp); 3371 goto out1; 3372 } 3373 tdvp = tond.ni_dvp; 3374 tvp = tond.ni_vp; 3375 3376 if (tvp != NULL) { 3377 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3378 error = ENOTDIR; 3379 goto out; 3380 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3381 error = EISDIR; 3382 goto out; 3383 } 3384 } 3385 3386 if (fvp == tdvp) 3387 error = EINVAL; 3388 3389 /* 3390 * Source and destination refer to the same object. 3391 */ 3392 if (fvp == tvp) { 3393 if (retain) 3394 error = -1; 3395 else if (fromnd.ni_dvp == tdvp && 3396 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3397 !memcmp(fromnd.ni_cnd.cn_nameptr, 3398 tond.ni_cnd.cn_nameptr, 3399 fromnd.ni_cnd.cn_namelen)) 3400 error = -1; 3401 } 3402 3403 #if NVERIEXEC > 0 3404 if (!error) 3405 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tvp, 3406 tond.ni_dirp, l); 3407 #endif /* NVERIEXEC > 0 */ 3408 3409 out: 3410 p = l->l_proc; 3411 if (!error) { 3412 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3413 if (fromnd.ni_dvp != tdvp) 3414 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3415 if (tvp) { 3416 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3417 } 3418 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3419 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3420 } else { 3421 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3422 if (tdvp == tvp) 3423 vrele(tdvp); 3424 else 3425 vput(tdvp); 3426 if (tvp) 3427 vput(tvp); 3428 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3429 vrele(fromnd.ni_dvp); 3430 vrele(fvp); 3431 } 3432 vrele(tond.ni_startdir); 3433 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3434 out1: 3435 vn_finished_write(mp, 0); 3436 if (fromnd.ni_startdir) 3437 vrele(fromnd.ni_startdir); 3438 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3439 return (error == -1 ? 0 : error); 3440 } 3441 3442 /* 3443 * Make a directory file. 3444 */ 3445 /* ARGSUSED */ 3446 int 3447 sys_mkdir(struct lwp *l, void *v, register_t *retval __unused) 3448 { 3449 struct sys_mkdir_args /* { 3450 syscallarg(const char *) path; 3451 syscallarg(int) mode; 3452 } */ *uap = v; 3453 struct proc *p = l->l_proc; 3454 struct mount *mp; 3455 struct vnode *vp; 3456 struct vattr vattr; 3457 int error; 3458 struct nameidata nd; 3459 3460 restart: 3461 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3462 SCARG(uap, path), l); 3463 if ((error = namei(&nd)) != 0) 3464 return (error); 3465 vp = nd.ni_vp; 3466 if (vp != NULL) { 3467 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3468 if (nd.ni_dvp == vp) 3469 vrele(nd.ni_dvp); 3470 else 3471 vput(nd.ni_dvp); 3472 vrele(vp); 3473 return (EEXIST); 3474 } 3475 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3476 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3477 if (nd.ni_dvp == vp) 3478 vrele(nd.ni_dvp); 3479 else 3480 vput(nd.ni_dvp); 3481 if ((error = vn_start_write(NULL, &mp, 3482 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3483 return (error); 3484 goto restart; 3485 } 3486 VATTR_NULL(&vattr); 3487 vattr.va_type = VDIR; 3488 vattr.va_mode = 3489 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3490 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3491 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3492 if (!error) 3493 vput(nd.ni_vp); 3494 vn_finished_write(mp, 0); 3495 return (error); 3496 } 3497 3498 /* 3499 * Remove a directory file. 3500 */ 3501 /* ARGSUSED */ 3502 int 3503 sys_rmdir(struct lwp *l, void *v, register_t *retval __unused) 3504 { 3505 struct sys_rmdir_args /* { 3506 syscallarg(const char *) path; 3507 } */ *uap = v; 3508 struct mount *mp; 3509 struct vnode *vp; 3510 int error; 3511 struct nameidata nd; 3512 3513 restart: 3514 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3515 SCARG(uap, path), l); 3516 if ((error = namei(&nd)) != 0) 3517 return (error); 3518 vp = nd.ni_vp; 3519 if (vp->v_type != VDIR) { 3520 error = ENOTDIR; 3521 goto out; 3522 } 3523 /* 3524 * No rmdir "." please. 3525 */ 3526 if (nd.ni_dvp == vp) { 3527 error = EINVAL; 3528 goto out; 3529 } 3530 /* 3531 * The root of a mounted filesystem cannot be deleted. 3532 */ 3533 if (vp->v_flag & VROOT) { 3534 error = EBUSY; 3535 goto out; 3536 } 3537 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3538 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3539 if (nd.ni_dvp == vp) 3540 vrele(nd.ni_dvp); 3541 else 3542 vput(nd.ni_dvp); 3543 vput(vp); 3544 if ((error = vn_start_write(NULL, &mp, 3545 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3546 return (error); 3547 goto restart; 3548 } 3549 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3550 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3551 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3552 vn_finished_write(mp, 0); 3553 return (error); 3554 3555 out: 3556 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3557 if (nd.ni_dvp == vp) 3558 vrele(nd.ni_dvp); 3559 else 3560 vput(nd.ni_dvp); 3561 vput(vp); 3562 return (error); 3563 } 3564 3565 /* 3566 * Read a block of directory entries in a file system independent format. 3567 */ 3568 int 3569 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3570 { 3571 struct sys___getdents30_args /* { 3572 syscallarg(int) fd; 3573 syscallarg(char *) buf; 3574 syscallarg(size_t) count; 3575 } */ *uap = v; 3576 struct proc *p = l->l_proc; 3577 struct file *fp; 3578 int error, done; 3579 3580 /* getvnode() will use the descriptor for us */ 3581 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3582 return (error); 3583 if ((fp->f_flag & FREAD) == 0) { 3584 error = EBADF; 3585 goto out; 3586 } 3587 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3588 SCARG(uap, count), &done, l, 0, 0); 3589 #ifdef KTRACE 3590 if (!error && KTRPOINT(p, KTR_GENIO)) { 3591 struct iovec iov; 3592 iov.iov_base = SCARG(uap, buf); 3593 iov.iov_len = done; 3594 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3595 } 3596 #endif 3597 *retval = done; 3598 out: 3599 FILE_UNUSE(fp, l); 3600 return (error); 3601 } 3602 3603 /* 3604 * Set the mode mask for creation of filesystem nodes. 3605 */ 3606 int 3607 sys_umask(struct lwp *l, void *v, register_t *retval) 3608 { 3609 struct sys_umask_args /* { 3610 syscallarg(mode_t) newmask; 3611 } */ *uap = v; 3612 struct proc *p = l->l_proc; 3613 struct cwdinfo *cwdi; 3614 3615 cwdi = p->p_cwdi; 3616 *retval = cwdi->cwdi_cmask; 3617 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3618 return (0); 3619 } 3620 3621 /* 3622 * Void all references to file by ripping underlying filesystem 3623 * away from vnode. 3624 */ 3625 /* ARGSUSED */ 3626 int 3627 sys_revoke(struct lwp *l, void *v, register_t *retval __unused) 3628 { 3629 struct sys_revoke_args /* { 3630 syscallarg(const char *) path; 3631 } */ *uap = v; 3632 struct mount *mp; 3633 struct vnode *vp; 3634 struct vattr vattr; 3635 int error; 3636 struct nameidata nd; 3637 3638 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3639 if ((error = namei(&nd)) != 0) 3640 return (error); 3641 vp = nd.ni_vp; 3642 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3643 goto out; 3644 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3645 (error = kauth_authorize_generic(l->l_cred, 3646 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 3647 goto out; 3648 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3649 goto out; 3650 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3651 VOP_REVOKE(vp, REVOKEALL); 3652 vn_finished_write(mp, 0); 3653 out: 3654 vrele(vp); 3655 return (error); 3656 } 3657 3658 /* 3659 * Convert a user file descriptor to a kernel file entry. 3660 */ 3661 int 3662 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3663 { 3664 struct vnode *vp; 3665 struct file *fp; 3666 3667 if ((fp = fd_getfile(fdp, fd)) == NULL) 3668 return (EBADF); 3669 3670 FILE_USE(fp); 3671 3672 if (fp->f_type != DTYPE_VNODE) { 3673 FILE_UNUSE(fp, NULL); 3674 return (EINVAL); 3675 } 3676 3677 vp = (struct vnode *)fp->f_data; 3678 if (vp->v_type == VBAD) { 3679 FILE_UNUSE(fp, NULL); 3680 return (EBADF); 3681 } 3682 3683 *fpp = fp; 3684 return (0); 3685 } 3686
Indexes created Sat Sep 27 14:09:57 GMT 2025