vfs_syscalls.c revision 1.265
1 /* $NetBSD: vfs_syscalls.c,v 1.265 2006/08/04 16:29:51 yamt 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 1.265 2006/08/04 16:29:51 yamt Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/sa.h> 65 #include <sys/syscallargs.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #if NVERIEXEC > 0 73 #include <sys/verified_exec.h> 74 #endif /* NVERIEXEC > 0 */ 75 #include <sys/kauth.h> 76 77 #include <miscfs/genfs/genfs.h> 78 #include <miscfs/syncfs/syncfs.h> 79 80 #ifdef COMPAT_30 81 #include "opt_nfsserver.h" 82 #include <nfs/rpcv2.h> 83 #include <nfs/nfsproto.h> 84 #include <nfs/nfs.h> 85 #include <nfs/nfs_var.h> 86 #endif 87 88 #if NFSS > 0 89 #include <dev/fssvar.h> 90 #endif 91 92 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 93 94 static int change_dir(struct nameidata *, struct lwp *); 95 static int change_flags(struct vnode *, u_long, struct lwp *); 96 static int change_mode(struct vnode *, int, struct lwp *l); 97 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 98 static int change_utimes(struct vnode *vp, const struct timeval *, 99 struct lwp *l); 100 static int rename_files(const char *, const char *, struct lwp *, int); 101 102 void checkdirs(struct vnode *); 103 104 int dovfsusermount = 0; 105 106 /* 107 * Virtual File System System Calls 108 */ 109 110 /* 111 * Mount a file system. 112 */ 113 114 #if defined(COMPAT_09) || defined(COMPAT_43) 115 /* 116 * This table is used to maintain compatibility with 4.3BSD 117 * and NetBSD 0.9 mount syscalls. Note, the order is important! 118 * 119 * Do not modify this table. It should only contain filesystems 120 * supported by NetBSD 0.9 and 4.3BSD. 121 */ 122 const char * const mountcompatnames[] = { 123 NULL, /* 0 = MOUNT_NONE */ 124 MOUNT_FFS, /* 1 = MOUNT_UFS */ 125 MOUNT_NFS, /* 2 */ 126 MOUNT_MFS, /* 3 */ 127 MOUNT_MSDOS, /* 4 */ 128 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 129 MOUNT_FDESC, /* 6 */ 130 MOUNT_KERNFS, /* 7 */ 131 NULL, /* 8 = MOUNT_DEVFS */ 132 MOUNT_AFS, /* 9 */ 133 }; 134 const int nmountcompatnames = sizeof(mountcompatnames) / 135 sizeof(mountcompatnames[0]); 136 #endif /* COMPAT_09 || COMPAT_43 */ 137 138 /* ARGSUSED */ 139 int 140 sys_mount(struct lwp *l, void *v, register_t *retval) 141 { 142 struct sys_mount_args /* { 143 syscallarg(const char *) type; 144 syscallarg(const char *) path; 145 syscallarg(int) flags; 146 syscallarg(void *) data; 147 } */ *uap = v; 148 struct vnode *vp; 149 struct mount *mp; 150 int error, flag = 0; 151 char fstypename[MFSNAMELEN]; 152 struct vattr va; 153 struct nameidata nd; 154 struct vfsops *vfs; 155 156 /* 157 * if MNT_GETARGS is specified, it should be only flag. 158 */ 159 160 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 161 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 162 return EINVAL; 163 } 164 165 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 166 (error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 167 &l->l_acflag))) 168 return (error); 169 /* 170 * Get vnode to be covered 171 */ 172 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 173 SCARG(uap, path), l); 174 if ((error = namei(&nd)) != 0) 175 return (error); 176 vp = nd.ni_vp; 177 /* 178 * A lookup in VFS_MOUNT might result in an attempt to 179 * lock this vnode again, so make the lock recursive. 180 */ 181 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 182 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 183 if ((vp->v_flag & VROOT) == 0) { 184 vput(vp); 185 return (EINVAL); 186 } 187 mp = vp->v_mount; 188 flag = mp->mnt_flag; 189 vfs = mp->mnt_op; 190 /* 191 * We only allow the filesystem to be reloaded if it 192 * is currently mounted read-only. 193 */ 194 if ((SCARG(uap, flags) & MNT_RELOAD) && 195 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 196 vput(vp); 197 return (EOPNOTSUPP); /* Needs translation */ 198 } 199 /* 200 * In "highly secure" mode, don't let the caller do anything 201 * but downgrade a filesystem from read-write to read-only. 202 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 203 */ 204 if (securelevel >= 2 && 205 SCARG(uap, flags) != MNT_GETARGS && 206 SCARG(uap, flags) != 207 (mp->mnt_flag | MNT_RDONLY | 208 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 209 vput(vp); 210 return (EPERM); 211 } 212 mp->mnt_flag |= SCARG(uap, flags) & 213 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 214 /* 215 * Only root, or the user that did the original mount is 216 * permitted to update it. 217 */ 218 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 219 mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred) && 220 (error = kauth_authorize_generic(l->l_cred, 221 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 222 vput(vp); 223 return (error); 224 } 225 /* 226 * Do not allow NFS export by non-root users. For non-root 227 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 228 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 229 */ 230 if (kauth_cred_geteuid(l->l_cred) != 0) { 231 if (SCARG(uap, flags) & MNT_EXPORTED) { 232 vput(vp); 233 return (EPERM); 234 } 235 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 236 if (flag & MNT_NOEXEC) 237 SCARG(uap, flags) |= MNT_NOEXEC; 238 } 239 if (vfs_busy(mp, LK_NOWAIT, 0)) { 240 vput(vp); 241 return (EPERM); 242 } 243 goto update; 244 } else { 245 if (securelevel >= 2) { 246 vput(vp); 247 return (EPERM); 248 } 249 } 250 /* 251 * If the user is not root, ensure that they own the directory 252 * onto which we are attempting to mount. 253 */ 254 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 255 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 256 (error = kauth_authorize_generic(l->l_cred, 257 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0)) { 258 vput(vp); 259 return (error); 260 } 261 /* 262 * Do not allow NFS export by non-root users. For non-root users, 263 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 264 * mount point is already MNT_NOEXEC. 265 */ 266 if (kauth_cred_geteuid(l->l_cred) != 0) { 267 if (SCARG(uap, flags) & MNT_EXPORTED) { 268 vput(vp); 269 return (EPERM); 270 } 271 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 272 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 273 SCARG(uap, flags) |= MNT_NOEXEC; 274 } 275 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 276 vput(vp); 277 return (error); 278 } 279 if (vp->v_type != VDIR) { 280 vput(vp); 281 return (ENOTDIR); 282 } 283 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 284 if (error) { 285 #if defined(COMPAT_09) || defined(COMPAT_43) 286 /* 287 * Historically, filesystem types were identified by numbers. 288 * If we get an integer for the filesystem type instead of a 289 * string, we check to see if it matches one of the historic 290 * filesystem types. 291 */ 292 u_long fsindex = (u_long)SCARG(uap, type); 293 if (fsindex >= nmountcompatnames || 294 mountcompatnames[fsindex] == NULL) { 295 vput(vp); 296 return (ENODEV); 297 } 298 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 299 #else 300 vput(vp); 301 return (error); 302 #endif 303 } 304 #ifdef COMPAT_10 305 /* Accept `ufs' as an alias for `ffs'. */ 306 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 307 strncpy(fstypename, "ffs", MFSNAMELEN); 308 #endif 309 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 310 vput(vp); 311 return (ENODEV); 312 } 313 if (vp->v_mountedhere != NULL) { 314 vput(vp); 315 return (EBUSY); 316 } 317 318 /* 319 * Allocate and initialize the file system. 320 */ 321 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 322 M_MOUNT, M_WAITOK); 323 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 324 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 325 simple_lock_init(&mp->mnt_slock); 326 (void)vfs_busy(mp, LK_NOWAIT, 0); 327 mp->mnt_op = vfs; 328 vfs->vfs_refcount++; 329 mp->mnt_vnodecovered = vp; 330 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 331 mp->mnt_unmounter = NULL; 332 mp->mnt_leaf = mp; 333 334 /* 335 * The underlying file system may refuse the mount for 336 * various reasons. Allow the user to force it to happen. 337 */ 338 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 339 update: 340 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 341 /* 342 * Set the mount level flags. 343 */ 344 if (SCARG(uap, flags) & MNT_RDONLY) 345 mp->mnt_flag |= MNT_RDONLY; 346 else if (mp->mnt_flag & MNT_RDONLY) 347 mp->mnt_iflag |= IMNT_WANTRDWR; 348 mp->mnt_flag &= 349 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 350 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 351 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 352 mp->mnt_flag |= SCARG(uap, flags) & 353 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 354 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 355 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 356 MNT_IGNORE); 357 } 358 /* 359 * Mount the filesystem. 360 */ 361 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 362 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 363 #if defined(COMPAT_30) && defined(NFSSERVER) 364 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 365 int error2; 366 367 /* Update failed; let's try and see if it was an 368 * export request. */ 369 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 370 SCARG(uap, data), l); 371 372 /* Only update error code if the export request was 373 * understood but some problem occurred while 374 * processing it. */ 375 if (error2 != EJUSTRETURN) 376 error = error2; 377 } 378 #endif 379 if (mp->mnt_iflag & IMNT_WANTRDWR) 380 mp->mnt_flag &= ~MNT_RDONLY; 381 if (error) 382 mp->mnt_flag = flag; 383 mp->mnt_flag &=~ 384 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 385 mp->mnt_iflag &=~ IMNT_WANTRDWR; 386 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 387 if (mp->mnt_syncer == NULL) 388 error = vfs_allocate_syncvnode(mp); 389 } else { 390 if (mp->mnt_syncer != NULL) 391 vfs_deallocate_syncvnode(mp); 392 } 393 vfs_unbusy(mp); 394 VOP_UNLOCK(vp, 0); 395 vrele(vp); 396 return (error); 397 } 398 /* 399 * Put the new filesystem on the mount list after root. 400 */ 401 cache_purge(vp); 402 if (!error) { 403 mp->mnt_flag &=~ 404 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 405 mp->mnt_iflag &=~ IMNT_WANTRDWR; 406 vp->v_mountedhere = mp; 407 simple_lock(&mountlist_slock); 408 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 409 simple_unlock(&mountlist_slock); 410 checkdirs(vp); 411 VOP_UNLOCK(vp, 0); 412 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 413 error = vfs_allocate_syncvnode(mp); 414 vfs_unbusy(mp); 415 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 416 if ((error = VFS_START(mp, 0, l))) 417 vrele(vp); 418 } else { 419 vp->v_mountedhere = (struct mount *)0; 420 vfs->vfs_refcount--; 421 vfs_unbusy(mp); 422 free(mp, M_MOUNT); 423 vput(vp); 424 } 425 return (error); 426 } 427 428 /* 429 * Scan all active processes to see if any of them have a current 430 * or root directory onto which the new filesystem has just been 431 * mounted. If so, replace them with the new mount point. 432 */ 433 void 434 checkdirs(struct vnode *olddp) 435 { 436 struct cwdinfo *cwdi; 437 struct vnode *newdp; 438 struct proc *p; 439 440 if (olddp->v_usecount == 1) 441 return; 442 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 443 panic("mount: lost mount"); 444 proclist_lock_read(); 445 PROCLIST_FOREACH(p, &allproc) { 446 cwdi = p->p_cwdi; 447 if (!cwdi) 448 continue; 449 if (cwdi->cwdi_cdir == olddp) { 450 vrele(cwdi->cwdi_cdir); 451 VREF(newdp); 452 cwdi->cwdi_cdir = newdp; 453 } 454 if (cwdi->cwdi_rdir == olddp) { 455 vrele(cwdi->cwdi_rdir); 456 VREF(newdp); 457 cwdi->cwdi_rdir = newdp; 458 } 459 } 460 proclist_unlock_read(); 461 if (rootvnode == olddp) { 462 vrele(rootvnode); 463 VREF(newdp); 464 rootvnode = newdp; 465 } 466 vput(newdp); 467 } 468 469 /* 470 * Unmount a file system. 471 * 472 * Note: unmount takes a path to the vnode mounted on as argument, 473 * not special file (as before). 474 */ 475 /* ARGSUSED */ 476 int 477 sys_unmount(struct lwp *l, void *v, register_t *retval) 478 { 479 struct sys_unmount_args /* { 480 syscallarg(const char *) path; 481 syscallarg(int) flags; 482 } */ *uap = v; 483 struct vnode *vp; 484 struct mount *mp; 485 int error; 486 struct nameidata nd; 487 488 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 489 SCARG(uap, path), l); 490 if ((error = namei(&nd)) != 0) 491 return (error); 492 vp = nd.ni_vp; 493 mp = vp->v_mount; 494 495 /* 496 * Only root, or the user that did the original mount is 497 * permitted to unmount this filesystem. 498 */ 499 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(l->l_cred)) && 500 (error = kauth_authorize_generic(l->l_cred, 501 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) { 502 vput(vp); 503 return (error); 504 } 505 506 /* 507 * Don't allow unmounting the root file system. 508 */ 509 if (mp->mnt_flag & MNT_ROOTFS) { 510 vput(vp); 511 return (EINVAL); 512 } 513 514 /* 515 * Must be the root of the filesystem 516 */ 517 if ((vp->v_flag & VROOT) == 0) { 518 vput(vp); 519 return (EINVAL); 520 } 521 vput(vp); 522 523 /* 524 * XXX Freeze syncer. Must do this before locking the 525 * mount point. See dounmount() for details. 526 */ 527 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 528 529 if (vfs_busy(mp, 0, 0)) { 530 lockmgr(&syncer_lock, LK_RELEASE, NULL); 531 return (EBUSY); 532 } 533 534 return (dounmount(mp, SCARG(uap, flags), l)); 535 } 536 537 /* 538 * Do the actual file system unmount. File system is assumed to have been 539 * marked busy by the caller. 540 */ 541 int 542 dounmount(struct mount *mp, int flags, struct lwp *l) 543 { 544 struct vnode *coveredvp; 545 int error; 546 int async; 547 int used_syncer; 548 549 #if NVERIEXEC > 0 550 if (!doing_shutdown) { 551 if (veriexec_strict >= VERIEXEC_LOCKDOWN) { 552 printf("Veriexec: Lockdown mode, preventing unmount of" 553 " \"%s\". (uid=%u)\n", mp->mnt_stat.f_mntonname, 554 kauth_cred_getuid(l->l_cred)); 555 return (EPERM); 556 } 557 558 if (veriexec_strict == VERIEXEC_IPS) { 559 struct veriexec_table_entry *vte; 560 561 /* Check if we have fingerprints on mount. */ 562 vte = fileassoc_tabledata_lookup(mp, veriexec_hook); 563 if ((vte != NULL) && (vte->vte_count > 0)) { 564 printf("Veriexec: IPS mode, preventing unmount" 565 " of \"%s\" with monitored files. " 566 "(uid=%u)\n", mp->mnt_stat.f_mntonname, 567 kauth_cred_getuid(l->l_cred)); 568 return (EPERM); 569 } 570 } 571 } 572 #endif /* NVERIEXEC > 0 */ 573 574 #ifdef FILEASSOC 575 (void)fileassoc_table_delete(mp); 576 #endif /* FILEASSOC */ 577 578 simple_lock(&mountlist_slock); 579 vfs_unbusy(mp); 580 used_syncer = (mp->mnt_syncer != NULL); 581 582 /* 583 * XXX Syncer must be frozen when we get here. This should really 584 * be done on a per-mountpoint basis, but especially the softdep 585 * code possibly called from the syncer doesn't exactly work on a 586 * per-mountpoint basis, so the softdep code would become a maze 587 * of vfs_busy() calls. 588 * 589 * The caller of dounmount() must acquire syncer_lock because 590 * the syncer itself acquires locks in syncer_lock -> vfs_busy 591 * order, and we must preserve that order to avoid deadlock. 592 * 593 * So, if the file system did not use the syncer, now is 594 * the time to release the syncer_lock. 595 */ 596 if (used_syncer == 0) 597 lockmgr(&syncer_lock, LK_RELEASE, NULL); 598 599 mp->mnt_iflag |= IMNT_UNMOUNT; 600 mp->mnt_unmounter = l; 601 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 602 vn_start_write(NULL, &mp, V_WAIT); 603 604 async = mp->mnt_flag & MNT_ASYNC; 605 mp->mnt_flag &= ~MNT_ASYNC; 606 cache_purgevfs(mp); /* remove cache entries for this file sys */ 607 if (mp->mnt_syncer != NULL) 608 vfs_deallocate_syncvnode(mp); 609 error = 0; 610 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 611 #if NFSS > 0 612 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 613 #endif 614 if (error == 0) 615 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 616 } 617 if (error == 0 || (flags & MNT_FORCE)) 618 error = VFS_UNMOUNT(mp, flags, l); 619 vn_finished_write(mp, 0); 620 simple_lock(&mountlist_slock); 621 if (error) { 622 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 623 (void) vfs_allocate_syncvnode(mp); 624 mp->mnt_iflag &= ~IMNT_UNMOUNT; 625 mp->mnt_unmounter = NULL; 626 mp->mnt_flag |= async; 627 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 628 &mountlist_slock); 629 if (used_syncer) 630 lockmgr(&syncer_lock, LK_RELEASE, NULL); 631 simple_lock(&mp->mnt_slock); 632 while (mp->mnt_wcnt > 0) { 633 wakeup(mp); 634 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 635 0, &mp->mnt_slock); 636 } 637 simple_unlock(&mp->mnt_slock); 638 return (error); 639 } 640 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 641 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 642 coveredvp->v_mountedhere = NULL; 643 vrele(coveredvp); 644 } 645 mp->mnt_op->vfs_refcount--; 646 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 647 panic("unmount: dangling vnode"); 648 mp->mnt_iflag |= IMNT_GONE; 649 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 650 if (used_syncer) 651 lockmgr(&syncer_lock, LK_RELEASE, NULL); 652 simple_lock(&mp->mnt_slock); 653 while (mp->mnt_wcnt > 0) { 654 wakeup(mp); 655 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 656 } 657 simple_unlock(&mp->mnt_slock); 658 vfs_hooks_unmount(mp); 659 free(mp, M_MOUNT); 660 return (0); 661 } 662 663 /* 664 * Sync each mounted filesystem. 665 */ 666 #ifdef DEBUG 667 int syncprt = 0; 668 struct ctldebug debug0 = { "syncprt", &syncprt }; 669 #endif 670 671 /* ARGSUSED */ 672 int 673 sys_sync(struct lwp *l, void *v, register_t *retval) 674 { 675 struct mount *mp, *nmp; 676 int asyncflag; 677 678 if (l == NULL) 679 l = &lwp0; 680 681 simple_lock(&mountlist_slock); 682 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 683 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 684 nmp = mp->mnt_list.cqe_prev; 685 continue; 686 } 687 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 688 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 689 asyncflag = mp->mnt_flag & MNT_ASYNC; 690 mp->mnt_flag &= ~MNT_ASYNC; 691 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 692 if (asyncflag) 693 mp->mnt_flag |= MNT_ASYNC; 694 vn_finished_write(mp, 0); 695 } 696 simple_lock(&mountlist_slock); 697 nmp = mp->mnt_list.cqe_prev; 698 vfs_unbusy(mp); 699 700 } 701 simple_unlock(&mountlist_slock); 702 #ifdef DEBUG 703 if (syncprt) 704 vfs_bufstats(); 705 #endif /* DEBUG */ 706 return (0); 707 } 708 709 /* 710 * Change filesystem quotas. 711 */ 712 /* ARGSUSED */ 713 int 714 sys_quotactl(struct lwp *l, void *v, register_t *retval) 715 { 716 struct sys_quotactl_args /* { 717 syscallarg(const char *) path; 718 syscallarg(int) cmd; 719 syscallarg(int) uid; 720 syscallarg(caddr_t) arg; 721 } */ *uap = v; 722 struct mount *mp; 723 int error; 724 struct nameidata nd; 725 726 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 727 if ((error = namei(&nd)) != 0) 728 return (error); 729 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 730 vrele(nd.ni_vp); 731 if (error) 732 return (error); 733 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 734 SCARG(uap, arg), l); 735 vn_finished_write(mp, 0); 736 return (error); 737 } 738 739 int 740 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 741 int root) 742 { 743 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 744 int error = 0; 745 746 /* 747 * If MNT_NOWAIT or MNT_LAZY is specified, do not 748 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 749 * overrides MNT_NOWAIT. 750 */ 751 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 752 (flags != MNT_WAIT && flags != 0)) { 753 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 754 goto done; 755 } 756 757 /* Get the filesystem stats now */ 758 memset(sp, 0, sizeof(*sp)); 759 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 760 return error; 761 } 762 763 if (cwdi->cwdi_rdir == NULL) 764 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 765 done: 766 if (cwdi->cwdi_rdir != NULL) { 767 size_t len; 768 char *bp; 769 char *path = PNBUF_GET(); 770 if (!path) 771 return ENOMEM; 772 773 bp = path + MAXPATHLEN; 774 *--bp = '\0'; 775 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 776 MAXPATHLEN / 2, 0, l); 777 if (error) { 778 PNBUF_PUT(path); 779 return error; 780 } 781 len = strlen(bp); 782 /* 783 * for mount points that are below our root, we can see 784 * them, so we fix up the pathname and return them. The 785 * rest we cannot see, so we don't allow viewing the 786 * data. 787 */ 788 if (strncmp(bp, sp->f_mntonname, len) == 0) { 789 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 790 sizeof(sp->f_mntonname)); 791 if (sp->f_mntonname[0] == '\0') 792 (void)strlcpy(sp->f_mntonname, "/", 793 sizeof(sp->f_mntonname)); 794 } else { 795 if (root) 796 (void)strlcpy(sp->f_mntonname, "/", 797 sizeof(sp->f_mntonname)); 798 else 799 error = EPERM; 800 } 801 PNBUF_PUT(path); 802 } 803 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 804 return error; 805 } 806 807 /* 808 * Get filesystem statistics. 809 */ 810 /* ARGSUSED */ 811 int 812 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 813 { 814 struct sys_statvfs1_args /* { 815 syscallarg(const char *) path; 816 syscallarg(struct statvfs *) buf; 817 syscallarg(int) flags; 818 } */ *uap = v; 819 struct mount *mp; 820 struct statvfs *sb; 821 int error; 822 struct nameidata nd; 823 824 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 825 if ((error = namei(&nd)) != 0) 826 return error; 827 mp = nd.ni_vp->v_mount; 828 vrele(nd.ni_vp); 829 sb = STATVFSBUF_GET(); 830 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 831 if (error == 0) { 832 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 833 } 834 STATVFSBUF_PUT(sb); 835 return error; 836 } 837 838 /* 839 * Get filesystem statistics. 840 */ 841 /* ARGSUSED */ 842 int 843 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 844 { 845 struct sys_fstatvfs1_args /* { 846 syscallarg(int) fd; 847 syscallarg(struct statvfs *) buf; 848 syscallarg(int) flags; 849 } */ *uap = v; 850 struct proc *p = l->l_proc; 851 struct file *fp; 852 struct mount *mp; 853 struct statvfs *sb; 854 int error; 855 856 /* getvnode() will use the descriptor for us */ 857 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 858 return (error); 859 mp = ((struct vnode *)fp->f_data)->v_mount; 860 sb = STATVFSBUF_GET(); 861 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 862 goto out; 863 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 864 out: 865 FILE_UNUSE(fp, l); 866 STATVFSBUF_PUT(sb); 867 return error; 868 } 869 870 871 /* 872 * Get statistics on all filesystems. 873 */ 874 int 875 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 876 { 877 struct sys_getvfsstat_args /* { 878 syscallarg(struct statvfs *) buf; 879 syscallarg(size_t) bufsize; 880 syscallarg(int) flags; 881 } */ *uap = v; 882 int root = 0; 883 struct proc *p = l->l_proc; 884 struct mount *mp, *nmp; 885 struct statvfs *sb; 886 struct statvfs *sfsp; 887 size_t count, maxcount; 888 int error = 0; 889 890 sb = STATVFSBUF_GET(); 891 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 892 sfsp = SCARG(uap, buf); 893 simple_lock(&mountlist_slock); 894 count = 0; 895 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 896 mp = nmp) { 897 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 898 nmp = CIRCLEQ_NEXT(mp, mnt_list); 899 continue; 900 } 901 if (sfsp && count < maxcount) { 902 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 903 if (error) { 904 simple_lock(&mountlist_slock); 905 nmp = CIRCLEQ_NEXT(mp, mnt_list); 906 vfs_unbusy(mp); 907 continue; 908 } 909 error = copyout(sb, sfsp, sizeof(*sfsp)); 910 if (error) { 911 vfs_unbusy(mp); 912 goto out; 913 } 914 sfsp++; 915 root |= strcmp(sb->f_mntonname, "/") == 0; 916 } 917 count++; 918 simple_lock(&mountlist_slock); 919 nmp = CIRCLEQ_NEXT(mp, mnt_list); 920 vfs_unbusy(mp); 921 } 922 simple_unlock(&mountlist_slock); 923 if (root == 0 && p->p_cwdi->cwdi_rdir) { 924 /* 925 * fake a root entry 926 */ 927 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 928 SCARG(uap, flags), 1)) != 0) 929 goto out; 930 if (sfsp) 931 error = copyout(sb, sfsp, sizeof(*sfsp)); 932 count++; 933 } 934 if (sfsp && count > maxcount) 935 *retval = maxcount; 936 else 937 *retval = count; 938 out: 939 STATVFSBUF_PUT(sb); 940 return error; 941 } 942 943 /* 944 * Change current working directory to a given file descriptor. 945 */ 946 /* ARGSUSED */ 947 int 948 sys_fchdir(struct lwp *l, void *v, register_t *retval) 949 { 950 struct sys_fchdir_args /* { 951 syscallarg(int) fd; 952 } */ *uap = v; 953 struct proc *p = l->l_proc; 954 struct filedesc *fdp = p->p_fd; 955 struct cwdinfo *cwdi = p->p_cwdi; 956 struct vnode *vp, *tdp; 957 struct mount *mp; 958 struct file *fp; 959 int error; 960 961 /* getvnode() will use the descriptor for us */ 962 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 963 return (error); 964 vp = (struct vnode *)fp->f_data; 965 966 VREF(vp); 967 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 968 if (vp->v_type != VDIR) 969 error = ENOTDIR; 970 else 971 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 972 while (!error && (mp = vp->v_mountedhere) != NULL) { 973 if (vfs_busy(mp, 0, 0)) 974 continue; 975 error = VFS_ROOT(mp, &tdp); 976 vfs_unbusy(mp); 977 if (error) 978 break; 979 vput(vp); 980 vp = tdp; 981 } 982 if (error) { 983 vput(vp); 984 goto out; 985 } 986 VOP_UNLOCK(vp, 0); 987 988 /* 989 * Disallow changing to a directory not under the process's 990 * current root directory (if there is one). 991 */ 992 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 993 vrele(vp); 994 error = EPERM; /* operation not permitted */ 995 goto out; 996 } 997 998 vrele(cwdi->cwdi_cdir); 999 cwdi->cwdi_cdir = vp; 1000 out: 1001 FILE_UNUSE(fp, l); 1002 return (error); 1003 } 1004 1005 /* 1006 * Change this process's notion of the root directory to a given file 1007 * descriptor. 1008 */ 1009 int 1010 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1011 { 1012 struct sys_fchroot_args *uap = v; 1013 struct proc *p = l->l_proc; 1014 struct filedesc *fdp = p->p_fd; 1015 struct cwdinfo *cwdi = p->p_cwdi; 1016 struct vnode *vp; 1017 struct file *fp; 1018 int error; 1019 1020 if ((error = kauth_authorize_generic(l->l_cred, 1021 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1022 return error; 1023 /* getvnode() will use the descriptor for us */ 1024 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1025 return error; 1026 vp = (struct vnode *) fp->f_data; 1027 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1028 if (vp->v_type != VDIR) 1029 error = ENOTDIR; 1030 else 1031 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1032 VOP_UNLOCK(vp, 0); 1033 if (error) 1034 goto out; 1035 VREF(vp); 1036 1037 /* 1038 * Prevent escaping from chroot by putting the root under 1039 * the working directory. Silently chdir to / if we aren't 1040 * already there. 1041 */ 1042 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1043 /* 1044 * XXX would be more failsafe to change directory to a 1045 * deadfs node here instead 1046 */ 1047 vrele(cwdi->cwdi_cdir); 1048 VREF(vp); 1049 cwdi->cwdi_cdir = vp; 1050 } 1051 1052 if (cwdi->cwdi_rdir != NULL) 1053 vrele(cwdi->cwdi_rdir); 1054 cwdi->cwdi_rdir = vp; 1055 out: 1056 FILE_UNUSE(fp, l); 1057 return (error); 1058 } 1059 1060 /* 1061 * Change current working directory (``.''). 1062 */ 1063 /* ARGSUSED */ 1064 int 1065 sys_chdir(struct lwp *l, void *v, register_t *retval) 1066 { 1067 struct sys_chdir_args /* { 1068 syscallarg(const char *) path; 1069 } */ *uap = v; 1070 struct proc *p = l->l_proc; 1071 struct cwdinfo *cwdi = p->p_cwdi; 1072 int error; 1073 struct nameidata nd; 1074 1075 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1076 SCARG(uap, path), l); 1077 if ((error = change_dir(&nd, l)) != 0) 1078 return (error); 1079 vrele(cwdi->cwdi_cdir); 1080 cwdi->cwdi_cdir = nd.ni_vp; 1081 return (0); 1082 } 1083 1084 /* 1085 * Change notion of root (``/'') directory. 1086 */ 1087 /* ARGSUSED */ 1088 int 1089 sys_chroot(struct lwp *l, void *v, register_t *retval) 1090 { 1091 struct sys_chroot_args /* { 1092 syscallarg(const char *) path; 1093 } */ *uap = v; 1094 struct proc *p = l->l_proc; 1095 struct cwdinfo *cwdi = p->p_cwdi; 1096 struct vnode *vp; 1097 int error; 1098 struct nameidata nd; 1099 1100 if ((error = kauth_authorize_generic(l->l_cred, 1101 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1102 return (error); 1103 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1104 SCARG(uap, path), l); 1105 if ((error = change_dir(&nd, l)) != 0) 1106 return (error); 1107 if (cwdi->cwdi_rdir != NULL) 1108 vrele(cwdi->cwdi_rdir); 1109 vp = nd.ni_vp; 1110 cwdi->cwdi_rdir = vp; 1111 1112 /* 1113 * Prevent escaping from chroot by putting the root under 1114 * the working directory. Silently chdir to / if we aren't 1115 * already there. 1116 */ 1117 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1118 /* 1119 * XXX would be more failsafe to change directory to a 1120 * deadfs node here instead 1121 */ 1122 vrele(cwdi->cwdi_cdir); 1123 VREF(vp); 1124 cwdi->cwdi_cdir = vp; 1125 } 1126 1127 return (0); 1128 } 1129 1130 /* 1131 * Common routine for chroot and chdir. 1132 */ 1133 static int 1134 change_dir(struct nameidata *ndp, struct lwp *l) 1135 { 1136 struct vnode *vp; 1137 int error; 1138 1139 if ((error = namei(ndp)) != 0) 1140 return (error); 1141 vp = ndp->ni_vp; 1142 if (vp->v_type != VDIR) 1143 error = ENOTDIR; 1144 else 1145 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1146 1147 if (error) 1148 vput(vp); 1149 else 1150 VOP_UNLOCK(vp, 0); 1151 return (error); 1152 } 1153 1154 /* 1155 * Check permissions, allocate an open file structure, 1156 * and call the device open routine if any. 1157 */ 1158 int 1159 sys_open(struct lwp *l, void *v, register_t *retval) 1160 { 1161 struct sys_open_args /* { 1162 syscallarg(const char *) path; 1163 syscallarg(int) flags; 1164 syscallarg(int) mode; 1165 } */ *uap = v; 1166 struct proc *p = l->l_proc; 1167 struct cwdinfo *cwdi = p->p_cwdi; 1168 struct filedesc *fdp = p->p_fd; 1169 struct file *fp; 1170 struct vnode *vp; 1171 int flags, cmode; 1172 int type, indx, error; 1173 struct flock lf; 1174 struct nameidata nd; 1175 1176 flags = FFLAGS(SCARG(uap, flags)); 1177 if ((flags & (FREAD | FWRITE)) == 0) 1178 return (EINVAL); 1179 /* falloc() will use the file descriptor for us */ 1180 if ((error = falloc(l, &fp, &indx)) != 0) 1181 return (error); 1182 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1183 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1184 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1185 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1186 FILE_UNUSE(fp, l); 1187 fdp->fd_ofiles[indx] = NULL; 1188 ffree(fp); 1189 if ((error == EDUPFD || error == EMOVEFD) && 1190 l->l_dupfd >= 0 && /* XXX from fdopen */ 1191 (error = 1192 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1193 *retval = indx; 1194 return (0); 1195 } 1196 if (error == ERESTART) 1197 error = EINTR; 1198 fdremove(fdp, indx); 1199 return (error); 1200 } 1201 l->l_dupfd = 0; 1202 vp = nd.ni_vp; 1203 fp->f_flag = flags & FMASK; 1204 fp->f_type = DTYPE_VNODE; 1205 fp->f_ops = &vnops; 1206 fp->f_data = vp; 1207 if (flags & (O_EXLOCK | O_SHLOCK)) { 1208 lf.l_whence = SEEK_SET; 1209 lf.l_start = 0; 1210 lf.l_len = 0; 1211 if (flags & O_EXLOCK) 1212 lf.l_type = F_WRLCK; 1213 else 1214 lf.l_type = F_RDLCK; 1215 type = F_FLOCK; 1216 if ((flags & FNONBLOCK) == 0) 1217 type |= F_WAIT; 1218 VOP_UNLOCK(vp, 0); 1219 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1220 if (error) { 1221 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1222 FILE_UNUSE(fp, l); 1223 ffree(fp); 1224 fdremove(fdp, indx); 1225 return (error); 1226 } 1227 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1228 fp->f_flag |= FHASLOCK; 1229 } 1230 VOP_UNLOCK(vp, 0); 1231 *retval = indx; 1232 FILE_SET_MATURE(fp); 1233 FILE_UNUSE(fp, l); 1234 return (0); 1235 } 1236 1237 static void 1238 vfs__fhfree(fhandle_t *fhp) 1239 { 1240 size_t fhsize; 1241 1242 if (fhp == NULL) { 1243 return; 1244 } 1245 fhsize = FHANDLE_SIZE(fhp); 1246 kmem_free(fhp, fhsize); 1247 } 1248 1249 /* 1250 * vfs_composefh: compose a filehandle. 1251 */ 1252 1253 int 1254 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1255 { 1256 struct mount *mp; 1257 struct fid *fidp; 1258 int error; 1259 size_t needfhsize; 1260 size_t fidsize; 1261 1262 mp = vp->v_mount; 1263 if (mp->mnt_op->vfs_vptofh == NULL) { 1264 return EOPNOTSUPP; 1265 } 1266 fidp = NULL; 1267 if (*fh_size < FHANDLE_SIZE_MIN) { 1268 fidsize = 0; 1269 } else { 1270 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1271 if (fhp != NULL) { 1272 memset(fhp, 0, *fh_size); 1273 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1274 fidp = &fhp->fh_fid; 1275 } 1276 } 1277 error = VFS_VPTOFH(vp, fidp, &fidsize); 1278 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1279 if (error == 0 && *fh_size < needfhsize) { 1280 error = E2BIG; 1281 } 1282 *fh_size = needfhsize; 1283 return error; 1284 } 1285 1286 int 1287 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1288 { 1289 struct mount *mp; 1290 fhandle_t *fhp; 1291 size_t fhsize; 1292 size_t fidsize; 1293 int error; 1294 1295 *fhpp = NULL; 1296 mp = vp->v_mount; 1297 if (mp->mnt_op->vfs_vptofh == NULL) { 1298 error = EOPNOTSUPP; 1299 goto out; 1300 } 1301 fidsize = 0; 1302 error = VFS_VPTOFH(vp, NULL, &fidsize); 1303 KASSERT(error != 0); 1304 if (error != E2BIG) { 1305 goto out; 1306 } 1307 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1308 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1309 if (fhp == NULL) { 1310 error = ENOMEM; 1311 goto out; 1312 } 1313 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1314 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1315 if (error == 0) { 1316 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1317 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1318 *fhpp = fhp; 1319 } else { 1320 kmem_free(fhp, fhsize); 1321 } 1322 out: 1323 return error; 1324 } 1325 1326 void 1327 vfs_composefh_free(fhandle_t *fhp) 1328 { 1329 1330 vfs__fhfree(fhp); 1331 } 1332 1333 /* 1334 * vfs_fhtovp: lookup a vnode by a filehandle. 1335 */ 1336 1337 int 1338 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1339 { 1340 struct mount *mp; 1341 int error; 1342 1343 *vpp = NULL; 1344 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1345 if (mp == NULL) { 1346 error = ESTALE; 1347 goto out; 1348 } 1349 if (mp->mnt_op->vfs_fhtovp == NULL) { 1350 error = EOPNOTSUPP; 1351 goto out; 1352 } 1353 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1354 out: 1355 return error; 1356 } 1357 1358 /* 1359 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1360 * the needed size. 1361 */ 1362 1363 int 1364 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1365 { 1366 fhandle_t *fhp; 1367 int error; 1368 1369 *fhpp = NULL; 1370 if (fhsize > FHANDLE_SIZE_MAX) { 1371 return EINVAL; 1372 } 1373 if (fhsize < FHANDLE_SIZE_MIN) { 1374 return EINVAL; 1375 } 1376 fhp = kmem_alloc(fhsize, KM_SLEEP); 1377 if (fhp == NULL) { 1378 return ENOMEM; 1379 } 1380 error = copyin(ufhp, fhp, fhsize); 1381 if (error == 0) { 1382 /* XXX this check shouldn't be here */ 1383 if (FHANDLE_SIZE(fhp) == fhsize) { 1384 *fhpp = fhp; 1385 return 0; 1386 } else { 1387 /* 1388 * userland told us wrong size. 1389 */ 1390 error = EINVAL; 1391 } 1392 } 1393 kmem_free(fhp, fhsize); 1394 return error; 1395 } 1396 1397 void 1398 vfs_copyinfh_free(fhandle_t *fhp) 1399 { 1400 1401 vfs__fhfree(fhp); 1402 } 1403 1404 /* 1405 * Get file handle system call 1406 */ 1407 int 1408 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1409 { 1410 struct sys___getfh30_args /* { 1411 syscallarg(char *) fname; 1412 syscallarg(fhandle_t *) fhp; 1413 syscallarg(size_t *) fh_size; 1414 } */ *uap = v; 1415 struct vnode *vp; 1416 fhandle_t *fh; 1417 int error; 1418 struct nameidata nd; 1419 size_t sz; 1420 size_t usz; 1421 1422 /* 1423 * Must be super user 1424 */ 1425 error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 1426 &l->l_acflag); 1427 if (error) 1428 return (error); 1429 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1430 SCARG(uap, fname), l); 1431 error = namei(&nd); 1432 if (error) 1433 return (error); 1434 vp = nd.ni_vp; 1435 error = vfs_composefh_alloc(vp, &fh); 1436 vput(vp); 1437 if (error != 0) { 1438 goto out; 1439 } 1440 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1441 if (error != 0) { 1442 goto out; 1443 } 1444 sz = FHANDLE_SIZE(fh); 1445 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1446 if (error != 0) { 1447 goto out; 1448 } 1449 if (usz >= sz) { 1450 error = copyout(fh, SCARG(uap, fhp), sz); 1451 } else { 1452 error = E2BIG; 1453 } 1454 out: 1455 vfs_composefh_free(fh); 1456 return (error); 1457 } 1458 1459 /* 1460 * Open a file given a file handle. 1461 * 1462 * Check permissions, allocate an open file structure, 1463 * and call the device open routine if any. 1464 */ 1465 1466 int 1467 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1468 register_t *retval) 1469 { 1470 struct filedesc *fdp = l->l_proc->p_fd; 1471 struct file *fp; 1472 struct vnode *vp = NULL; 1473 struct mount *mp; 1474 kauth_cred_t cred = l->l_cred; 1475 struct file *nfp; 1476 int type, indx, error=0; 1477 struct flock lf; 1478 struct vattr va; 1479 fhandle_t *fh; 1480 int flags; 1481 1482 /* 1483 * Must be super user 1484 */ 1485 if ((error = kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 1486 &l->l_acflag))) 1487 return (error); 1488 1489 flags = FFLAGS(oflags); 1490 if ((flags & (FREAD | FWRITE)) == 0) 1491 return (EINVAL); 1492 if ((flags & O_CREAT)) 1493 return (EINVAL); 1494 /* falloc() will use the file descriptor for us */ 1495 if ((error = falloc(l, &nfp, &indx)) != 0) 1496 return (error); 1497 fp = nfp; 1498 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1499 if (error != 0) { 1500 goto bad; 1501 } 1502 error = vfs_fhtovp(fh, &vp); 1503 if (error != 0) { 1504 goto bad; 1505 } 1506 1507 /* Now do an effective vn_open */ 1508 1509 if (vp->v_type == VSOCK) { 1510 error = EOPNOTSUPP; 1511 goto bad; 1512 } 1513 if (flags & FREAD) { 1514 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1515 goto bad; 1516 } 1517 if (flags & (FWRITE | O_TRUNC)) { 1518 if (vp->v_type == VDIR) { 1519 error = EISDIR; 1520 goto bad; 1521 } 1522 if ((error = vn_writechk(vp)) != 0 || 1523 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1524 goto bad; 1525 } 1526 if (flags & O_TRUNC) { 1527 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1528 goto bad; 1529 VOP_UNLOCK(vp, 0); /* XXX */ 1530 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1531 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1532 VATTR_NULL(&va); 1533 va.va_size = 0; 1534 error = VOP_SETATTR(vp, &va, cred, l); 1535 vn_finished_write(mp, 0); 1536 if (error) 1537 goto bad; 1538 } 1539 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1540 goto bad; 1541 if (vp->v_type == VREG && 1542 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1543 error = EIO; 1544 goto bad; 1545 } 1546 if (flags & FWRITE) 1547 vp->v_writecount++; 1548 1549 /* done with modified vn_open, now finish what sys_open does. */ 1550 1551 fp->f_flag = flags & FMASK; 1552 fp->f_type = DTYPE_VNODE; 1553 fp->f_ops = &vnops; 1554 fp->f_data = vp; 1555 if (flags & (O_EXLOCK | O_SHLOCK)) { 1556 lf.l_whence = SEEK_SET; 1557 lf.l_start = 0; 1558 lf.l_len = 0; 1559 if (flags & O_EXLOCK) 1560 lf.l_type = F_WRLCK; 1561 else 1562 lf.l_type = F_RDLCK; 1563 type = F_FLOCK; 1564 if ((flags & FNONBLOCK) == 0) 1565 type |= F_WAIT; 1566 VOP_UNLOCK(vp, 0); 1567 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1568 if (error) { 1569 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1570 FILE_UNUSE(fp, l); 1571 ffree(fp); 1572 fdremove(fdp, indx); 1573 return (error); 1574 } 1575 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1576 fp->f_flag |= FHASLOCK; 1577 } 1578 VOP_UNLOCK(vp, 0); 1579 *retval = indx; 1580 FILE_SET_MATURE(fp); 1581 FILE_UNUSE(fp, l); 1582 vfs_copyinfh_free(fh); 1583 return (0); 1584 1585 bad: 1586 FILE_UNUSE(fp, l); 1587 ffree(fp); 1588 fdremove(fdp, indx); 1589 if (vp != NULL) 1590 vput(vp); 1591 vfs_copyinfh_free(fh); 1592 return (error); 1593 } 1594 1595 int 1596 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1597 { 1598 struct sys___fhopen40_args /* { 1599 syscallarg(const void *) fhp; 1600 syscallarg(size_t) fh_size; 1601 syscallarg(int) flags; 1602 } */ *uap = v; 1603 1604 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1605 SCARG(uap, flags), retval); 1606 } 1607 1608 int 1609 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1610 register_t *retval) 1611 { 1612 struct stat sb; 1613 int error; 1614 fhandle_t *fh; 1615 struct vnode *vp; 1616 1617 /* 1618 * Must be super user 1619 */ 1620 if ((error = kauth_authorize_generic(l->l_cred, 1621 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1622 return (error); 1623 1624 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1625 if (error != 0) { 1626 goto bad; 1627 } 1628 error = vfs_fhtovp(fh, &vp); 1629 if (error != 0) { 1630 goto bad; 1631 } 1632 error = vn_stat(vp, &sb, l); 1633 vput(vp); 1634 if (error) { 1635 goto bad; 1636 } 1637 error = copyout(&sb, sbp, sizeof(sb)); 1638 bad: 1639 vfs_copyinfh_free(fh); 1640 return error; 1641 } 1642 1643 1644 /* ARGSUSED */ 1645 int 1646 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1647 { 1648 struct sys___fhstat40_args /* { 1649 syscallarg(const void *) fhp; 1650 syscallarg(size_t) fh_size; 1651 syscallarg(struct stat *) sb; 1652 } */ *uap = v; 1653 1654 return dofhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), SCARG(uap, sb), 1655 retval); 1656 } 1657 1658 int 1659 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1660 int flags, register_t *retval) 1661 { 1662 struct statvfs *sb = NULL; 1663 fhandle_t *fh; 1664 struct mount *mp; 1665 struct vnode *vp; 1666 int error; 1667 1668 /* 1669 * Must be super user 1670 */ 1671 if ((error = kauth_authorize_generic(l->l_cred, 1672 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1673 return error; 1674 1675 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1676 if (error != 0) { 1677 goto out; 1678 } 1679 error = vfs_fhtovp(fh, &vp); 1680 if (error != 0) { 1681 goto out; 1682 } 1683 mp = vp->v_mount; 1684 sb = STATVFSBUF_GET(); 1685 if ((error = dostatvfs(mp, sb, l, flags, 1)) != 0) { 1686 vput(vp); 1687 goto out; 1688 } 1689 vput(vp); 1690 error = copyout(sb, buf, sizeof(*sb)); 1691 out: 1692 if (sb != NULL) { 1693 STATVFSBUF_PUT(sb); 1694 } 1695 vfs_copyinfh_free(fh); 1696 return error; 1697 } 1698 1699 /* ARGSUSED */ 1700 int 1701 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1702 { 1703 struct sys___fhstatvfs140_args /* { 1704 syscallarg(const fhandle_t *) fhp; 1705 syscallarg(size_t) fh_size; 1706 syscallarg(struct statvfs *) buf; 1707 syscallarg(int) flags; 1708 } */ *uap = v; 1709 1710 return dofhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1711 SCARG(uap, buf), SCARG(uap, flags), retval); 1712 } 1713 1714 /* 1715 * Create a special file. 1716 */ 1717 /* ARGSUSED */ 1718 int 1719 sys_mknod(struct lwp *l, void *v, register_t *retval) 1720 { 1721 struct sys_mknod_args /* { 1722 syscallarg(const char *) path; 1723 syscallarg(int) mode; 1724 syscallarg(int) dev; 1725 } */ *uap = v; 1726 struct proc *p = l->l_proc; 1727 struct vnode *vp; 1728 struct mount *mp; 1729 struct vattr vattr; 1730 int error; 1731 int whiteout = 0; 1732 struct nameidata nd; 1733 1734 if ((error = kauth_authorize_generic(l->l_cred, 1735 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 1736 return (error); 1737 restart: 1738 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1739 if ((error = namei(&nd)) != 0) 1740 return (error); 1741 vp = nd.ni_vp; 1742 if (vp != NULL) 1743 error = EEXIST; 1744 else { 1745 VATTR_NULL(&vattr); 1746 vattr.va_mode = 1747 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1748 vattr.va_rdev = SCARG(uap, dev); 1749 whiteout = 0; 1750 1751 switch (SCARG(uap, mode) & S_IFMT) { 1752 case S_IFMT: /* used by badsect to flag bad sectors */ 1753 vattr.va_type = VBAD; 1754 break; 1755 case S_IFCHR: 1756 vattr.va_type = VCHR; 1757 break; 1758 case S_IFBLK: 1759 vattr.va_type = VBLK; 1760 break; 1761 case S_IFWHT: 1762 whiteout = 1; 1763 break; 1764 default: 1765 error = EINVAL; 1766 break; 1767 } 1768 } 1769 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1770 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1771 if (nd.ni_dvp == vp) 1772 vrele(nd.ni_dvp); 1773 else 1774 vput(nd.ni_dvp); 1775 if (vp) 1776 vrele(vp); 1777 if ((error = vn_start_write(NULL, &mp, 1778 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1779 return (error); 1780 goto restart; 1781 } 1782 if (!error) { 1783 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1784 if (whiteout) { 1785 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1786 if (error) 1787 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1788 vput(nd.ni_dvp); 1789 } else { 1790 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1791 &nd.ni_cnd, &vattr); 1792 if (error == 0) 1793 vput(nd.ni_vp); 1794 } 1795 } else { 1796 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1797 if (nd.ni_dvp == vp) 1798 vrele(nd.ni_dvp); 1799 else 1800 vput(nd.ni_dvp); 1801 if (vp) 1802 vrele(vp); 1803 } 1804 vn_finished_write(mp, 0); 1805 return (error); 1806 } 1807 1808 /* 1809 * Create a named pipe. 1810 */ 1811 /* ARGSUSED */ 1812 int 1813 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1814 { 1815 struct sys_mkfifo_args /* { 1816 syscallarg(const char *) path; 1817 syscallarg(int) mode; 1818 } */ *uap = v; 1819 struct proc *p = l->l_proc; 1820 struct mount *mp; 1821 struct vattr vattr; 1822 int error; 1823 struct nameidata nd; 1824 1825 restart: 1826 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1827 if ((error = namei(&nd)) != 0) 1828 return (error); 1829 if (nd.ni_vp != NULL) { 1830 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1831 if (nd.ni_dvp == nd.ni_vp) 1832 vrele(nd.ni_dvp); 1833 else 1834 vput(nd.ni_dvp); 1835 vrele(nd.ni_vp); 1836 return (EEXIST); 1837 } 1838 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1839 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1840 if (nd.ni_dvp == nd.ni_vp) 1841 vrele(nd.ni_dvp); 1842 else 1843 vput(nd.ni_dvp); 1844 if (nd.ni_vp) 1845 vrele(nd.ni_vp); 1846 if ((error = vn_start_write(NULL, &mp, 1847 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1848 return (error); 1849 goto restart; 1850 } 1851 VATTR_NULL(&vattr); 1852 vattr.va_type = VFIFO; 1853 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1854 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1855 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1856 if (error == 0) 1857 vput(nd.ni_vp); 1858 vn_finished_write(mp, 0); 1859 return (error); 1860 } 1861 1862 /* 1863 * Make a hard file link. 1864 */ 1865 /* ARGSUSED */ 1866 int 1867 sys_link(struct lwp *l, void *v, register_t *retval) 1868 { 1869 struct sys_link_args /* { 1870 syscallarg(const char *) path; 1871 syscallarg(const char *) link; 1872 } */ *uap = v; 1873 struct vnode *vp; 1874 struct mount *mp; 1875 struct nameidata nd; 1876 int error; 1877 1878 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1879 if ((error = namei(&nd)) != 0) 1880 return (error); 1881 vp = nd.ni_vp; 1882 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1883 vrele(vp); 1884 return (error); 1885 } 1886 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1887 if ((error = namei(&nd)) != 0) 1888 goto out; 1889 if (nd.ni_vp) { 1890 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1891 if (nd.ni_dvp == nd.ni_vp) 1892 vrele(nd.ni_dvp); 1893 else 1894 vput(nd.ni_dvp); 1895 vrele(nd.ni_vp); 1896 error = EEXIST; 1897 goto out; 1898 } 1899 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1900 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1901 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1902 out: 1903 vrele(vp); 1904 vn_finished_write(mp, 0); 1905 return (error); 1906 } 1907 1908 /* 1909 * Make a symbolic link. 1910 */ 1911 /* ARGSUSED */ 1912 int 1913 sys_symlink(struct lwp *l, void *v, register_t *retval) 1914 { 1915 struct sys_symlink_args /* { 1916 syscallarg(const char *) path; 1917 syscallarg(const char *) link; 1918 } */ *uap = v; 1919 struct proc *p = l->l_proc; 1920 struct mount *mp; 1921 struct vattr vattr; 1922 char *path; 1923 int error; 1924 struct nameidata nd; 1925 1926 path = PNBUF_GET(); 1927 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1928 if (error) 1929 goto out; 1930 restart: 1931 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1932 if ((error = namei(&nd)) != 0) 1933 goto out; 1934 if (nd.ni_vp) { 1935 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1936 if (nd.ni_dvp == nd.ni_vp) 1937 vrele(nd.ni_dvp); 1938 else 1939 vput(nd.ni_dvp); 1940 vrele(nd.ni_vp); 1941 error = EEXIST; 1942 goto out; 1943 } 1944 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1945 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1946 if (nd.ni_dvp == nd.ni_vp) 1947 vrele(nd.ni_dvp); 1948 else 1949 vput(nd.ni_dvp); 1950 if ((error = vn_start_write(NULL, &mp, 1951 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1952 return (error); 1953 goto restart; 1954 } 1955 VATTR_NULL(&vattr); 1956 vattr.va_type = VLNK; 1957 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1958 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1959 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1960 if (error == 0) 1961 vput(nd.ni_vp); 1962 vn_finished_write(mp, 0); 1963 out: 1964 PNBUF_PUT(path); 1965 return (error); 1966 } 1967 1968 /* 1969 * Delete a whiteout from the filesystem. 1970 */ 1971 /* ARGSUSED */ 1972 int 1973 sys_undelete(struct lwp *l, void *v, register_t *retval) 1974 { 1975 struct sys_undelete_args /* { 1976 syscallarg(const char *) path; 1977 } */ *uap = v; 1978 int error; 1979 struct mount *mp; 1980 struct nameidata nd; 1981 1982 restart: 1983 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1984 SCARG(uap, path), l); 1985 error = namei(&nd); 1986 if (error) 1987 return (error); 1988 1989 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1990 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1991 if (nd.ni_dvp == nd.ni_vp) 1992 vrele(nd.ni_dvp); 1993 else 1994 vput(nd.ni_dvp); 1995 if (nd.ni_vp) 1996 vrele(nd.ni_vp); 1997 return (EEXIST); 1998 } 1999 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2000 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2001 if (nd.ni_dvp == nd.ni_vp) 2002 vrele(nd.ni_dvp); 2003 else 2004 vput(nd.ni_dvp); 2005 if ((error = vn_start_write(NULL, &mp, 2006 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2007 return (error); 2008 goto restart; 2009 } 2010 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2011 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2012 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2013 vput(nd.ni_dvp); 2014 vn_finished_write(mp, 0); 2015 return (error); 2016 } 2017 2018 /* 2019 * Delete a name from the filesystem. 2020 */ 2021 /* ARGSUSED */ 2022 int 2023 sys_unlink(struct lwp *l, void *v, register_t *retval) 2024 { 2025 struct sys_unlink_args /* { 2026 syscallarg(const char *) path; 2027 } */ *uap = v; 2028 struct mount *mp; 2029 struct vnode *vp; 2030 int error; 2031 struct nameidata nd; 2032 2033 restart: 2034 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2035 SCARG(uap, path), l); 2036 if ((error = namei(&nd)) != 0) 2037 return (error); 2038 vp = nd.ni_vp; 2039 2040 /* 2041 * The root of a mounted filesystem cannot be deleted. 2042 */ 2043 if (vp->v_flag & VROOT) { 2044 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2045 if (nd.ni_dvp == vp) 2046 vrele(nd.ni_dvp); 2047 else 2048 vput(nd.ni_dvp); 2049 vput(vp); 2050 error = EBUSY; 2051 goto out; 2052 } 2053 2054 #if NVERIEXEC > 0 2055 /* Handle remove requests for veriexec entries. */ 2056 if ((error = veriexec_removechk(vp, nd.ni_dirp, l)) != 0) { 2057 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2058 if (nd.ni_dvp == vp) 2059 vrele(nd.ni_dvp); 2060 else 2061 vput(nd.ni_dvp); 2062 vput(vp); 2063 goto out; 2064 } 2065 #endif /* NVERIEXEC > 0 */ 2066 2067 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 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 if ((error = vn_start_write(NULL, &mp, 2075 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2076 return (error); 2077 goto restart; 2078 } 2079 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2080 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2081 #ifdef FILEASSOC 2082 (void)fileassoc_file_delete(nd.ni_vp); 2083 #endif /* FILEASSOC */ 2084 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2085 vn_finished_write(mp, 0); 2086 out: 2087 return (error); 2088 } 2089 2090 /* 2091 * Reposition read/write file offset. 2092 */ 2093 int 2094 sys_lseek(struct lwp *l, void *v, register_t *retval) 2095 { 2096 struct sys_lseek_args /* { 2097 syscallarg(int) fd; 2098 syscallarg(int) pad; 2099 syscallarg(off_t) offset; 2100 syscallarg(int) whence; 2101 } */ *uap = v; 2102 struct proc *p = l->l_proc; 2103 kauth_cred_t cred = l->l_cred; 2104 struct filedesc *fdp = p->p_fd; 2105 struct file *fp; 2106 struct vnode *vp; 2107 struct vattr vattr; 2108 off_t newoff; 2109 int error; 2110 2111 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2112 return (EBADF); 2113 2114 FILE_USE(fp); 2115 2116 vp = (struct vnode *)fp->f_data; 2117 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2118 error = ESPIPE; 2119 goto out; 2120 } 2121 2122 switch (SCARG(uap, whence)) { 2123 case SEEK_CUR: 2124 newoff = fp->f_offset + SCARG(uap, offset); 2125 break; 2126 case SEEK_END: 2127 error = VOP_GETATTR(vp, &vattr, cred, l); 2128 if (error) 2129 goto out; 2130 newoff = SCARG(uap, offset) + vattr.va_size; 2131 break; 2132 case SEEK_SET: 2133 newoff = SCARG(uap, offset); 2134 break; 2135 default: 2136 error = EINVAL; 2137 goto out; 2138 } 2139 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2140 goto out; 2141 2142 *(off_t *)retval = fp->f_offset = newoff; 2143 out: 2144 FILE_UNUSE(fp, l); 2145 return (error); 2146 } 2147 2148 /* 2149 * Positional read system call. 2150 */ 2151 int 2152 sys_pread(struct lwp *l, void *v, register_t *retval) 2153 { 2154 struct sys_pread_args /* { 2155 syscallarg(int) fd; 2156 syscallarg(void *) buf; 2157 syscallarg(size_t) nbyte; 2158 syscallarg(off_t) offset; 2159 } */ *uap = v; 2160 struct proc *p = l->l_proc; 2161 struct filedesc *fdp = p->p_fd; 2162 struct file *fp; 2163 struct vnode *vp; 2164 off_t offset; 2165 int error, fd = SCARG(uap, fd); 2166 2167 if ((fp = fd_getfile(fdp, fd)) == NULL) 2168 return (EBADF); 2169 2170 if ((fp->f_flag & FREAD) == 0) { 2171 simple_unlock(&fp->f_slock); 2172 return (EBADF); 2173 } 2174 2175 FILE_USE(fp); 2176 2177 vp = (struct vnode *)fp->f_data; 2178 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2179 error = ESPIPE; 2180 goto out; 2181 } 2182 2183 offset = SCARG(uap, offset); 2184 2185 /* 2186 * XXX This works because no file systems actually 2187 * XXX take any action on the seek operation. 2188 */ 2189 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2190 goto out; 2191 2192 /* dofileread() will unuse the descriptor for us */ 2193 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2194 &offset, 0, retval)); 2195 2196 out: 2197 FILE_UNUSE(fp, l); 2198 return (error); 2199 } 2200 2201 /* 2202 * Positional scatter read system call. 2203 */ 2204 int 2205 sys_preadv(struct lwp *l, void *v, register_t *retval) 2206 { 2207 struct sys_preadv_args /* { 2208 syscallarg(int) fd; 2209 syscallarg(const struct iovec *) iovp; 2210 syscallarg(int) iovcnt; 2211 syscallarg(off_t) offset; 2212 } */ *uap = v; 2213 struct proc *p = l->l_proc; 2214 struct filedesc *fdp = p->p_fd; 2215 struct file *fp; 2216 struct vnode *vp; 2217 off_t offset; 2218 int error, fd = SCARG(uap, fd); 2219 2220 if ((fp = fd_getfile(fdp, fd)) == NULL) 2221 return (EBADF); 2222 2223 if ((fp->f_flag & FREAD) == 0) { 2224 simple_unlock(&fp->f_slock); 2225 return (EBADF); 2226 } 2227 2228 FILE_USE(fp); 2229 2230 vp = (struct vnode *)fp->f_data; 2231 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2232 error = ESPIPE; 2233 goto out; 2234 } 2235 2236 offset = SCARG(uap, offset); 2237 2238 /* 2239 * XXX This works because no file systems actually 2240 * XXX take any action on the seek operation. 2241 */ 2242 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2243 goto out; 2244 2245 /* dofilereadv() will unuse the descriptor for us */ 2246 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2247 &offset, 0, retval)); 2248 2249 out: 2250 FILE_UNUSE(fp, l); 2251 return (error); 2252 } 2253 2254 /* 2255 * Positional write system call. 2256 */ 2257 int 2258 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2259 { 2260 struct sys_pwrite_args /* { 2261 syscallarg(int) fd; 2262 syscallarg(const void *) buf; 2263 syscallarg(size_t) nbyte; 2264 syscallarg(off_t) offset; 2265 } */ *uap = v; 2266 struct proc *p = l->l_proc; 2267 struct filedesc *fdp = p->p_fd; 2268 struct file *fp; 2269 struct vnode *vp; 2270 off_t offset; 2271 int error, fd = SCARG(uap, fd); 2272 2273 if ((fp = fd_getfile(fdp, fd)) == NULL) 2274 return (EBADF); 2275 2276 if ((fp->f_flag & FWRITE) == 0) { 2277 simple_unlock(&fp->f_slock); 2278 return (EBADF); 2279 } 2280 2281 FILE_USE(fp); 2282 2283 vp = (struct vnode *)fp->f_data; 2284 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2285 error = ESPIPE; 2286 goto out; 2287 } 2288 2289 offset = SCARG(uap, offset); 2290 2291 /* 2292 * XXX This works because no file systems actually 2293 * XXX take any action on the seek operation. 2294 */ 2295 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2296 goto out; 2297 2298 /* dofilewrite() will unuse the descriptor for us */ 2299 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2300 &offset, 0, retval)); 2301 2302 out: 2303 FILE_UNUSE(fp, l); 2304 return (error); 2305 } 2306 2307 /* 2308 * Positional gather write system call. 2309 */ 2310 int 2311 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2312 { 2313 struct sys_pwritev_args /* { 2314 syscallarg(int) fd; 2315 syscallarg(const struct iovec *) iovp; 2316 syscallarg(int) iovcnt; 2317 syscallarg(off_t) offset; 2318 } */ *uap = v; 2319 struct proc *p = l->l_proc; 2320 struct filedesc *fdp = p->p_fd; 2321 struct file *fp; 2322 struct vnode *vp; 2323 off_t offset; 2324 int error, fd = SCARG(uap, fd); 2325 2326 if ((fp = fd_getfile(fdp, fd)) == NULL) 2327 return (EBADF); 2328 2329 if ((fp->f_flag & FWRITE) == 0) { 2330 simple_unlock(&fp->f_slock); 2331 return (EBADF); 2332 } 2333 2334 FILE_USE(fp); 2335 2336 vp = (struct vnode *)fp->f_data; 2337 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2338 error = ESPIPE; 2339 goto out; 2340 } 2341 2342 offset = SCARG(uap, offset); 2343 2344 /* 2345 * XXX This works because no file systems actually 2346 * XXX take any action on the seek operation. 2347 */ 2348 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2349 goto out; 2350 2351 /* dofilewritev() will unuse the descriptor for us */ 2352 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2353 &offset, 0, retval)); 2354 2355 out: 2356 FILE_UNUSE(fp, l); 2357 return (error); 2358 } 2359 2360 /* 2361 * Check access permissions. 2362 */ 2363 int 2364 sys_access(struct lwp *l, void *v, register_t *retval) 2365 { 2366 struct sys_access_args /* { 2367 syscallarg(const char *) path; 2368 syscallarg(int) flags; 2369 } */ *uap = v; 2370 kauth_cred_t cred; 2371 struct vnode *vp; 2372 int error, flags; 2373 struct nameidata nd; 2374 2375 cred = kauth_cred_dup(l->l_cred); 2376 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2377 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2378 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2379 SCARG(uap, path), l); 2380 /* Override default credentials */ 2381 nd.ni_cnd.cn_cred = cred; 2382 if ((error = namei(&nd)) != 0) 2383 goto out; 2384 vp = nd.ni_vp; 2385 2386 /* Flags == 0 means only check for existence. */ 2387 if (SCARG(uap, flags)) { 2388 flags = 0; 2389 if (SCARG(uap, flags) & R_OK) 2390 flags |= VREAD; 2391 if (SCARG(uap, flags) & W_OK) 2392 flags |= VWRITE; 2393 if (SCARG(uap, flags) & X_OK) 2394 flags |= VEXEC; 2395 2396 error = VOP_ACCESS(vp, flags, cred, l); 2397 if (!error && (flags & VWRITE)) 2398 error = vn_writechk(vp); 2399 } 2400 vput(vp); 2401 out: 2402 kauth_cred_free(cred); 2403 return (error); 2404 } 2405 2406 /* 2407 * Get file status; this version follows links. 2408 */ 2409 /* ARGSUSED */ 2410 int 2411 sys___stat30(struct lwp *l, void *v, register_t *retval) 2412 { 2413 struct sys___stat30_args /* { 2414 syscallarg(const char *) path; 2415 syscallarg(struct stat *) ub; 2416 } */ *uap = v; 2417 struct stat sb; 2418 int error; 2419 struct nameidata nd; 2420 2421 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2422 SCARG(uap, path), l); 2423 if ((error = namei(&nd)) != 0) 2424 return (error); 2425 error = vn_stat(nd.ni_vp, &sb, l); 2426 vput(nd.ni_vp); 2427 if (error) 2428 return (error); 2429 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2430 return (error); 2431 } 2432 2433 /* 2434 * Get file status; this version does not follow links. 2435 */ 2436 /* ARGSUSED */ 2437 int 2438 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2439 { 2440 struct sys___lstat30_args /* { 2441 syscallarg(const char *) path; 2442 syscallarg(struct stat *) ub; 2443 } */ *uap = v; 2444 struct stat sb; 2445 int error; 2446 struct nameidata nd; 2447 2448 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2449 SCARG(uap, path), l); 2450 if ((error = namei(&nd)) != 0) 2451 return (error); 2452 error = vn_stat(nd.ni_vp, &sb, l); 2453 vput(nd.ni_vp); 2454 if (error) 2455 return (error); 2456 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2457 return (error); 2458 } 2459 2460 /* 2461 * Get configurable pathname variables. 2462 */ 2463 /* ARGSUSED */ 2464 int 2465 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2466 { 2467 struct sys_pathconf_args /* { 2468 syscallarg(const char *) path; 2469 syscallarg(int) name; 2470 } */ *uap = v; 2471 int error; 2472 struct nameidata nd; 2473 2474 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2475 SCARG(uap, path), l); 2476 if ((error = namei(&nd)) != 0) 2477 return (error); 2478 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2479 vput(nd.ni_vp); 2480 return (error); 2481 } 2482 2483 /* 2484 * Return target name of a symbolic link. 2485 */ 2486 /* ARGSUSED */ 2487 int 2488 sys_readlink(struct lwp *l, void *v, register_t *retval) 2489 { 2490 struct sys_readlink_args /* { 2491 syscallarg(const char *) path; 2492 syscallarg(char *) buf; 2493 syscallarg(size_t) count; 2494 } */ *uap = v; 2495 struct vnode *vp; 2496 struct iovec aiov; 2497 struct uio auio; 2498 int error; 2499 struct nameidata nd; 2500 2501 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2502 SCARG(uap, path), l); 2503 if ((error = namei(&nd)) != 0) 2504 return (error); 2505 vp = nd.ni_vp; 2506 if (vp->v_type != VLNK) 2507 error = EINVAL; 2508 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2509 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2510 aiov.iov_base = SCARG(uap, buf); 2511 aiov.iov_len = SCARG(uap, count); 2512 auio.uio_iov = &aiov; 2513 auio.uio_iovcnt = 1; 2514 auio.uio_offset = 0; 2515 auio.uio_rw = UIO_READ; 2516 KASSERT(l == curlwp); 2517 auio.uio_vmspace = l->l_proc->p_vmspace; 2518 auio.uio_resid = SCARG(uap, count); 2519 error = VOP_READLINK(vp, &auio, l->l_cred); 2520 } 2521 vput(vp); 2522 *retval = SCARG(uap, count) - auio.uio_resid; 2523 return (error); 2524 } 2525 2526 /* 2527 * Change flags of a file given a path name. 2528 */ 2529 /* ARGSUSED */ 2530 int 2531 sys_chflags(struct lwp *l, void *v, register_t *retval) 2532 { 2533 struct sys_chflags_args /* { 2534 syscallarg(const char *) path; 2535 syscallarg(u_long) flags; 2536 } */ *uap = v; 2537 struct vnode *vp; 2538 int error; 2539 struct nameidata nd; 2540 2541 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2542 if ((error = namei(&nd)) != 0) 2543 return (error); 2544 vp = nd.ni_vp; 2545 error = change_flags(vp, SCARG(uap, flags), l); 2546 vput(vp); 2547 return (error); 2548 } 2549 2550 /* 2551 * Change flags of a file given a file descriptor. 2552 */ 2553 /* ARGSUSED */ 2554 int 2555 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2556 { 2557 struct sys_fchflags_args /* { 2558 syscallarg(int) fd; 2559 syscallarg(u_long) flags; 2560 } */ *uap = v; 2561 struct proc *p = l->l_proc; 2562 struct vnode *vp; 2563 struct file *fp; 2564 int error; 2565 2566 /* getvnode() will use the descriptor for us */ 2567 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2568 return (error); 2569 vp = (struct vnode *)fp->f_data; 2570 error = change_flags(vp, SCARG(uap, flags), l); 2571 VOP_UNLOCK(vp, 0); 2572 FILE_UNUSE(fp, l); 2573 return (error); 2574 } 2575 2576 /* 2577 * Change flags of a file given a path name; this version does 2578 * not follow links. 2579 */ 2580 int 2581 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2582 { 2583 struct sys_lchflags_args /* { 2584 syscallarg(const char *) path; 2585 syscallarg(u_long) flags; 2586 } */ *uap = v; 2587 struct vnode *vp; 2588 int error; 2589 struct nameidata nd; 2590 2591 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2592 if ((error = namei(&nd)) != 0) 2593 return (error); 2594 vp = nd.ni_vp; 2595 error = change_flags(vp, SCARG(uap, flags), l); 2596 vput(vp); 2597 return (error); 2598 } 2599 2600 /* 2601 * Common routine to change flags of a file. 2602 */ 2603 int 2604 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2605 { 2606 struct mount *mp; 2607 struct vattr vattr; 2608 int error; 2609 2610 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2611 return (error); 2612 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2613 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2614 /* 2615 * Non-superusers cannot change the flags on devices, even if they 2616 * own them. 2617 */ 2618 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2619 &l->l_acflag) != 0) { 2620 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2621 goto out; 2622 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2623 error = EINVAL; 2624 goto out; 2625 } 2626 } 2627 VATTR_NULL(&vattr); 2628 vattr.va_flags = flags; 2629 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2630 out: 2631 vn_finished_write(mp, 0); 2632 return (error); 2633 } 2634 2635 /* 2636 * Change mode of a file given path name; this version follows links. 2637 */ 2638 /* ARGSUSED */ 2639 int 2640 sys_chmod(struct lwp *l, void *v, register_t *retval) 2641 { 2642 struct sys_chmod_args /* { 2643 syscallarg(const char *) path; 2644 syscallarg(int) mode; 2645 } */ *uap = v; 2646 int error; 2647 struct nameidata nd; 2648 2649 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2650 if ((error = namei(&nd)) != 0) 2651 return (error); 2652 2653 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2654 2655 vrele(nd.ni_vp); 2656 return (error); 2657 } 2658 2659 /* 2660 * Change mode of a file given a file descriptor. 2661 */ 2662 /* ARGSUSED */ 2663 int 2664 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2665 { 2666 struct sys_fchmod_args /* { 2667 syscallarg(int) fd; 2668 syscallarg(int) mode; 2669 } */ *uap = v; 2670 struct proc *p = l->l_proc; 2671 struct file *fp; 2672 int error; 2673 2674 /* getvnode() will use the descriptor for us */ 2675 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2676 return (error); 2677 2678 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2679 FILE_UNUSE(fp, l); 2680 return (error); 2681 } 2682 2683 /* 2684 * Change mode of a file given path name; this version does not follow links. 2685 */ 2686 /* ARGSUSED */ 2687 int 2688 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2689 { 2690 struct sys_lchmod_args /* { 2691 syscallarg(const char *) path; 2692 syscallarg(int) mode; 2693 } */ *uap = v; 2694 int error; 2695 struct nameidata nd; 2696 2697 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2698 if ((error = namei(&nd)) != 0) 2699 return (error); 2700 2701 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2702 2703 vrele(nd.ni_vp); 2704 return (error); 2705 } 2706 2707 /* 2708 * Common routine to set mode given a vnode. 2709 */ 2710 static int 2711 change_mode(struct vnode *vp, int mode, struct lwp *l) 2712 { 2713 struct mount *mp; 2714 struct vattr vattr; 2715 int error; 2716 2717 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2718 return (error); 2719 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2720 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2721 VATTR_NULL(&vattr); 2722 vattr.va_mode = mode & ALLPERMS; 2723 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2724 VOP_UNLOCK(vp, 0); 2725 vn_finished_write(mp, 0); 2726 return (error); 2727 } 2728 2729 /* 2730 * Set ownership given a path name; this version follows links. 2731 */ 2732 /* ARGSUSED */ 2733 int 2734 sys_chown(struct lwp *l, void *v, register_t *retval) 2735 { 2736 struct sys_chown_args /* { 2737 syscallarg(const char *) path; 2738 syscallarg(uid_t) uid; 2739 syscallarg(gid_t) gid; 2740 } */ *uap = v; 2741 int error; 2742 struct nameidata nd; 2743 2744 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2745 if ((error = namei(&nd)) != 0) 2746 return (error); 2747 2748 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2749 2750 vrele(nd.ni_vp); 2751 return (error); 2752 } 2753 2754 /* 2755 * Set ownership given a path name; this version follows links. 2756 * Provides POSIX semantics. 2757 */ 2758 /* ARGSUSED */ 2759 int 2760 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2761 { 2762 struct sys_chown_args /* { 2763 syscallarg(const char *) path; 2764 syscallarg(uid_t) uid; 2765 syscallarg(gid_t) gid; 2766 } */ *uap = v; 2767 int error; 2768 struct nameidata nd; 2769 2770 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2771 if ((error = namei(&nd)) != 0) 2772 return (error); 2773 2774 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2775 2776 vrele(nd.ni_vp); 2777 return (error); 2778 } 2779 2780 /* 2781 * Set ownership given a file descriptor. 2782 */ 2783 /* ARGSUSED */ 2784 int 2785 sys_fchown(struct lwp *l, void *v, register_t *retval) 2786 { 2787 struct sys_fchown_args /* { 2788 syscallarg(int) fd; 2789 syscallarg(uid_t) uid; 2790 syscallarg(gid_t) gid; 2791 } */ *uap = v; 2792 struct proc *p = l->l_proc; 2793 int error; 2794 struct file *fp; 2795 2796 /* getvnode() will use the descriptor for us */ 2797 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2798 return (error); 2799 2800 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2801 SCARG(uap, gid), l, 0); 2802 FILE_UNUSE(fp, l); 2803 return (error); 2804 } 2805 2806 /* 2807 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2808 */ 2809 /* ARGSUSED */ 2810 int 2811 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2812 { 2813 struct sys_fchown_args /* { 2814 syscallarg(int) fd; 2815 syscallarg(uid_t) uid; 2816 syscallarg(gid_t) gid; 2817 } */ *uap = v; 2818 struct proc *p = l->l_proc; 2819 int error; 2820 struct file *fp; 2821 2822 /* getvnode() will use the descriptor for us */ 2823 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2824 return (error); 2825 2826 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2827 SCARG(uap, gid), l, 1); 2828 FILE_UNUSE(fp, l); 2829 return (error); 2830 } 2831 2832 /* 2833 * Set ownership given a path name; this version does not follow links. 2834 */ 2835 /* ARGSUSED */ 2836 int 2837 sys_lchown(struct lwp *l, void *v, register_t *retval) 2838 { 2839 struct sys_lchown_args /* { 2840 syscallarg(const char *) path; 2841 syscallarg(uid_t) uid; 2842 syscallarg(gid_t) gid; 2843 } */ *uap = v; 2844 int error; 2845 struct nameidata nd; 2846 2847 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2848 if ((error = namei(&nd)) != 0) 2849 return (error); 2850 2851 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2852 2853 vrele(nd.ni_vp); 2854 return (error); 2855 } 2856 2857 /* 2858 * Set ownership given a path name; this version does not follow links. 2859 * Provides POSIX/XPG semantics. 2860 */ 2861 /* ARGSUSED */ 2862 int 2863 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2864 { 2865 struct sys_lchown_args /* { 2866 syscallarg(const char *) path; 2867 syscallarg(uid_t) uid; 2868 syscallarg(gid_t) gid; 2869 } */ *uap = v; 2870 int error; 2871 struct nameidata nd; 2872 2873 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2874 if ((error = namei(&nd)) != 0) 2875 return (error); 2876 2877 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2878 2879 vrele(nd.ni_vp); 2880 return (error); 2881 } 2882 2883 /* 2884 * Common routine to set ownership given a vnode. 2885 */ 2886 static int 2887 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2888 int posix_semantics) 2889 { 2890 struct mount *mp; 2891 struct vattr vattr; 2892 mode_t newmode; 2893 int error; 2894 2895 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2896 return (error); 2897 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2898 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2899 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2900 goto out; 2901 2902 #define CHANGED(x) ((int)(x) != -1) 2903 newmode = vattr.va_mode; 2904 if (posix_semantics) { 2905 /* 2906 * POSIX/XPG semantics: if the caller is not the super-user, 2907 * clear set-user-id and set-group-id bits. Both POSIX and 2908 * the XPG consider the behaviour for calls by the super-user 2909 * implementation-defined; we leave the set-user-id and set- 2910 * group-id settings intact in that case. 2911 */ 2912 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2913 NULL) != 0) 2914 newmode &= ~(S_ISUID | S_ISGID); 2915 } else { 2916 /* 2917 * NetBSD semantics: when changing owner and/or group, 2918 * clear the respective bit(s). 2919 */ 2920 if (CHANGED(uid)) 2921 newmode &= ~S_ISUID; 2922 if (CHANGED(gid)) 2923 newmode &= ~S_ISGID; 2924 } 2925 /* Update va_mode iff altered. */ 2926 if (vattr.va_mode == newmode) 2927 newmode = VNOVAL; 2928 2929 VATTR_NULL(&vattr); 2930 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2931 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2932 vattr.va_mode = newmode; 2933 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2934 #undef CHANGED 2935 2936 out: 2937 VOP_UNLOCK(vp, 0); 2938 vn_finished_write(mp, 0); 2939 return (error); 2940 } 2941 2942 /* 2943 * Set the access and modification times given a path name; this 2944 * version follows links. 2945 */ 2946 /* ARGSUSED */ 2947 int 2948 sys_utimes(struct lwp *l, void *v, register_t *retval) 2949 { 2950 struct sys_utimes_args /* { 2951 syscallarg(const char *) path; 2952 syscallarg(const struct timeval *) tptr; 2953 } */ *uap = v; 2954 int error; 2955 struct nameidata nd; 2956 2957 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2958 if ((error = namei(&nd)) != 0) 2959 return (error); 2960 2961 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2962 2963 vrele(nd.ni_vp); 2964 return (error); 2965 } 2966 2967 /* 2968 * Set the access and modification times given a file descriptor. 2969 */ 2970 /* ARGSUSED */ 2971 int 2972 sys_futimes(struct lwp *l, void *v, register_t *retval) 2973 { 2974 struct sys_futimes_args /* { 2975 syscallarg(int) fd; 2976 syscallarg(const struct timeval *) tptr; 2977 } */ *uap = v; 2978 struct proc *p = l->l_proc; 2979 int error; 2980 struct file *fp; 2981 2982 /* getvnode() will use the descriptor for us */ 2983 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2984 return (error); 2985 2986 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2987 FILE_UNUSE(fp, l); 2988 return (error); 2989 } 2990 2991 /* 2992 * Set the access and modification times given a path name; this 2993 * version does not follow links. 2994 */ 2995 /* ARGSUSED */ 2996 int 2997 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2998 { 2999 struct sys_lutimes_args /* { 3000 syscallarg(const char *) path; 3001 syscallarg(const struct timeval *) tptr; 3002 } */ *uap = v; 3003 int error; 3004 struct nameidata nd; 3005 3006 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3007 if ((error = namei(&nd)) != 0) 3008 return (error); 3009 3010 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3011 3012 vrele(nd.ni_vp); 3013 return (error); 3014 } 3015 3016 /* 3017 * Common routine to set access and modification times given a vnode. 3018 */ 3019 static int 3020 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3021 { 3022 struct mount *mp; 3023 struct vattr vattr; 3024 int error; 3025 3026 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3027 return (error); 3028 VATTR_NULL(&vattr); 3029 if (tptr == NULL) { 3030 nanotime(&vattr.va_atime); 3031 vattr.va_mtime = vattr.va_atime; 3032 vattr.va_vaflags |= VA_UTIMES_NULL; 3033 } else { 3034 struct timeval tv[2]; 3035 3036 error = copyin(tptr, tv, sizeof(tv)); 3037 if (error) 3038 goto out; 3039 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3040 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3041 } 3042 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3043 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3044 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3045 VOP_UNLOCK(vp, 0); 3046 out: 3047 vn_finished_write(mp, 0); 3048 return (error); 3049 } 3050 3051 /* 3052 * Truncate a file given its path name. 3053 */ 3054 /* ARGSUSED */ 3055 int 3056 sys_truncate(struct lwp *l, void *v, register_t *retval) 3057 { 3058 struct sys_truncate_args /* { 3059 syscallarg(const char *) path; 3060 syscallarg(int) pad; 3061 syscallarg(off_t) length; 3062 } */ *uap = v; 3063 struct vnode *vp; 3064 struct mount *mp; 3065 struct vattr vattr; 3066 int error; 3067 struct nameidata nd; 3068 3069 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3070 if ((error = namei(&nd)) != 0) 3071 return (error); 3072 vp = nd.ni_vp; 3073 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3074 vrele(vp); 3075 return (error); 3076 } 3077 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3078 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3079 if (vp->v_type == VDIR) 3080 error = EISDIR; 3081 else if ((error = vn_writechk(vp)) == 0 && 3082 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3083 VATTR_NULL(&vattr); 3084 vattr.va_size = SCARG(uap, length); 3085 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3086 } 3087 vput(vp); 3088 vn_finished_write(mp, 0); 3089 return (error); 3090 } 3091 3092 /* 3093 * Truncate a file given a file descriptor. 3094 */ 3095 /* ARGSUSED */ 3096 int 3097 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3098 { 3099 struct sys_ftruncate_args /* { 3100 syscallarg(int) fd; 3101 syscallarg(int) pad; 3102 syscallarg(off_t) length; 3103 } */ *uap = v; 3104 struct proc *p = l->l_proc; 3105 struct mount *mp; 3106 struct vattr vattr; 3107 struct vnode *vp; 3108 struct file *fp; 3109 int error; 3110 3111 /* getvnode() will use the descriptor for us */ 3112 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3113 return (error); 3114 if ((fp->f_flag & FWRITE) == 0) { 3115 error = EINVAL; 3116 goto out; 3117 } 3118 vp = (struct vnode *)fp->f_data; 3119 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3120 FILE_UNUSE(fp, l); 3121 return (error); 3122 } 3123 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3124 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3125 if (vp->v_type == VDIR) 3126 error = EISDIR; 3127 else if ((error = vn_writechk(vp)) == 0) { 3128 VATTR_NULL(&vattr); 3129 vattr.va_size = SCARG(uap, length); 3130 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3131 } 3132 VOP_UNLOCK(vp, 0); 3133 vn_finished_write(mp, 0); 3134 out: 3135 FILE_UNUSE(fp, l); 3136 return (error); 3137 } 3138 3139 /* 3140 * Sync an open file. 3141 */ 3142 /* ARGSUSED */ 3143 int 3144 sys_fsync(struct lwp *l, void *v, register_t *retval) 3145 { 3146 struct sys_fsync_args /* { 3147 syscallarg(int) fd; 3148 } */ *uap = v; 3149 struct proc *p = l->l_proc; 3150 struct vnode *vp; 3151 struct mount *mp; 3152 struct file *fp; 3153 int error; 3154 3155 /* getvnode() will use the descriptor for us */ 3156 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3157 return (error); 3158 vp = (struct vnode *)fp->f_data; 3159 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3160 FILE_UNUSE(fp, l); 3161 return (error); 3162 } 3163 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3164 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3165 if (error == 0 && bioops.io_fsync != NULL && 3166 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3167 (*bioops.io_fsync)(vp, 0); 3168 VOP_UNLOCK(vp, 0); 3169 vn_finished_write(mp, 0); 3170 FILE_UNUSE(fp, l); 3171 return (error); 3172 } 3173 3174 /* 3175 * Sync a range of file data. API modeled after that found in AIX. 3176 * 3177 * FDATASYNC indicates that we need only save enough metadata to be able 3178 * to re-read the written data. Note we duplicate AIX's requirement that 3179 * the file be open for writing. 3180 */ 3181 /* ARGSUSED */ 3182 int 3183 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3184 { 3185 struct sys_fsync_range_args /* { 3186 syscallarg(int) fd; 3187 syscallarg(int) flags; 3188 syscallarg(off_t) start; 3189 syscallarg(off_t) length; 3190 } */ *uap = v; 3191 struct proc *p = l->l_proc; 3192 struct vnode *vp; 3193 struct file *fp; 3194 int flags, nflags; 3195 off_t s, e, len; 3196 int error; 3197 3198 /* getvnode() will use the descriptor for us */ 3199 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3200 return (error); 3201 3202 if ((fp->f_flag & FWRITE) == 0) { 3203 FILE_UNUSE(fp, l); 3204 return (EBADF); 3205 } 3206 3207 flags = SCARG(uap, flags); 3208 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3209 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3210 return (EINVAL); 3211 } 3212 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3213 if (flags & FDATASYNC) 3214 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3215 else 3216 nflags = FSYNC_WAIT; 3217 if (flags & FDISKSYNC) 3218 nflags |= FSYNC_CACHE; 3219 3220 len = SCARG(uap, length); 3221 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3222 if (len) { 3223 s = SCARG(uap, start); 3224 e = s + len; 3225 if (e < s) { 3226 FILE_UNUSE(fp, l); 3227 return (EINVAL); 3228 } 3229 } else { 3230 e = 0; 3231 s = 0; 3232 } 3233 3234 vp = (struct vnode *)fp->f_data; 3235 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3236 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3237 3238 if (error == 0 && bioops.io_fsync != NULL && 3239 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3240 (*bioops.io_fsync)(vp, nflags); 3241 3242 VOP_UNLOCK(vp, 0); 3243 FILE_UNUSE(fp, l); 3244 return (error); 3245 } 3246 3247 /* 3248 * Sync the data of an open file. 3249 */ 3250 /* ARGSUSED */ 3251 int 3252 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3253 { 3254 struct sys_fdatasync_args /* { 3255 syscallarg(int) fd; 3256 } */ *uap = v; 3257 struct proc *p = l->l_proc; 3258 struct vnode *vp; 3259 struct file *fp; 3260 int error; 3261 3262 /* getvnode() will use the descriptor for us */ 3263 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3264 return (error); 3265 if ((fp->f_flag & FWRITE) == 0) { 3266 FILE_UNUSE(fp, l); 3267 return (EBADF); 3268 } 3269 vp = (struct vnode *)fp->f_data; 3270 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3271 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3272 VOP_UNLOCK(vp, 0); 3273 FILE_UNUSE(fp, l); 3274 return (error); 3275 } 3276 3277 /* 3278 * Rename files, (standard) BSD semantics frontend. 3279 */ 3280 /* ARGSUSED */ 3281 int 3282 sys_rename(struct lwp *l, void *v, register_t *retval) 3283 { 3284 struct sys_rename_args /* { 3285 syscallarg(const char *) from; 3286 syscallarg(const char *) to; 3287 } */ *uap = v; 3288 3289 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3290 } 3291 3292 /* 3293 * Rename files, POSIX semantics frontend. 3294 */ 3295 /* ARGSUSED */ 3296 int 3297 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3298 { 3299 struct sys___posix_rename_args /* { 3300 syscallarg(const char *) from; 3301 syscallarg(const char *) to; 3302 } */ *uap = v; 3303 3304 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3305 } 3306 3307 /* 3308 * Rename files. Source and destination must either both be directories, 3309 * or both not be directories. If target is a directory, it must be empty. 3310 * If `from' and `to' refer to the same object, the value of the `retain' 3311 * argument is used to determine whether `from' will be 3312 * 3313 * (retain == 0) deleted unless `from' and `to' refer to the same 3314 * object in the file system's name space (BSD). 3315 * (retain == 1) always retained (POSIX). 3316 */ 3317 static int 3318 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3319 { 3320 struct mount *mp = NULL; 3321 struct vnode *tvp, *fvp, *tdvp; 3322 struct nameidata fromnd, tond; 3323 struct proc *p; 3324 int error; 3325 3326 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3327 from, l); 3328 if ((error = namei(&fromnd)) != 0) 3329 return (error); 3330 fvp = fromnd.ni_vp; 3331 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3332 if (error != 0) { 3333 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3334 vrele(fromnd.ni_dvp); 3335 vrele(fvp); 3336 if (fromnd.ni_startdir) 3337 vrele(fromnd.ni_startdir); 3338 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3339 return (error); 3340 } 3341 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3342 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3343 if ((error = namei(&tond)) != 0) { 3344 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3345 vrele(fromnd.ni_dvp); 3346 vrele(fvp); 3347 goto out1; 3348 } 3349 tdvp = tond.ni_dvp; 3350 tvp = tond.ni_vp; 3351 3352 if (tvp != NULL) { 3353 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3354 error = ENOTDIR; 3355 goto out; 3356 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3357 error = EISDIR; 3358 goto out; 3359 } 3360 } 3361 3362 if (fvp == tdvp) 3363 error = EINVAL; 3364 3365 /* 3366 * Source and destination refer to the same object. 3367 */ 3368 if (fvp == tvp) { 3369 if (retain) 3370 error = -1; 3371 else if (fromnd.ni_dvp == tdvp && 3372 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3373 !memcmp(fromnd.ni_cnd.cn_nameptr, 3374 tond.ni_cnd.cn_nameptr, 3375 fromnd.ni_cnd.cn_namelen)) 3376 error = -1; 3377 } 3378 3379 #if NVERIEXEC > 0 3380 if (!error) 3381 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tvp, 3382 tond.ni_dirp, l); 3383 #endif /* NVERIEXEC > 0 */ 3384 3385 out: 3386 p = l->l_proc; 3387 if (!error) { 3388 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3389 if (fromnd.ni_dvp != tdvp) 3390 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3391 if (tvp) { 3392 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3393 } 3394 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3395 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3396 } else { 3397 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3398 if (tdvp == tvp) 3399 vrele(tdvp); 3400 else 3401 vput(tdvp); 3402 if (tvp) 3403 vput(tvp); 3404 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3405 vrele(fromnd.ni_dvp); 3406 vrele(fvp); 3407 } 3408 vrele(tond.ni_startdir); 3409 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3410 out1: 3411 vn_finished_write(mp, 0); 3412 if (fromnd.ni_startdir) 3413 vrele(fromnd.ni_startdir); 3414 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3415 return (error == -1 ? 0 : error); 3416 } 3417 3418 /* 3419 * Make a directory file. 3420 */ 3421 /* ARGSUSED */ 3422 int 3423 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3424 { 3425 struct sys_mkdir_args /* { 3426 syscallarg(const char *) path; 3427 syscallarg(int) mode; 3428 } */ *uap = v; 3429 struct proc *p = l->l_proc; 3430 struct mount *mp; 3431 struct vnode *vp; 3432 struct vattr vattr; 3433 int error; 3434 struct nameidata nd; 3435 3436 restart: 3437 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3438 SCARG(uap, path), l); 3439 if ((error = namei(&nd)) != 0) 3440 return (error); 3441 vp = nd.ni_vp; 3442 if (vp != NULL) { 3443 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3444 if (nd.ni_dvp == vp) 3445 vrele(nd.ni_dvp); 3446 else 3447 vput(nd.ni_dvp); 3448 vrele(vp); 3449 return (EEXIST); 3450 } 3451 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3452 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3453 if (nd.ni_dvp == vp) 3454 vrele(nd.ni_dvp); 3455 else 3456 vput(nd.ni_dvp); 3457 if ((error = vn_start_write(NULL, &mp, 3458 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3459 return (error); 3460 goto restart; 3461 } 3462 VATTR_NULL(&vattr); 3463 vattr.va_type = VDIR; 3464 vattr.va_mode = 3465 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3466 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3467 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3468 if (!error) 3469 vput(nd.ni_vp); 3470 vn_finished_write(mp, 0); 3471 return (error); 3472 } 3473 3474 /* 3475 * Remove a directory file. 3476 */ 3477 /* ARGSUSED */ 3478 int 3479 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3480 { 3481 struct sys_rmdir_args /* { 3482 syscallarg(const char *) path; 3483 } */ *uap = v; 3484 struct mount *mp; 3485 struct vnode *vp; 3486 int error; 3487 struct nameidata nd; 3488 3489 restart: 3490 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3491 SCARG(uap, path), l); 3492 if ((error = namei(&nd)) != 0) 3493 return (error); 3494 vp = nd.ni_vp; 3495 if (vp->v_type != VDIR) { 3496 error = ENOTDIR; 3497 goto out; 3498 } 3499 /* 3500 * No rmdir "." please. 3501 */ 3502 if (nd.ni_dvp == vp) { 3503 error = EINVAL; 3504 goto out; 3505 } 3506 /* 3507 * The root of a mounted filesystem cannot be deleted. 3508 */ 3509 if (vp->v_flag & VROOT) { 3510 error = EBUSY; 3511 goto out; 3512 } 3513 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3514 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3515 if (nd.ni_dvp == vp) 3516 vrele(nd.ni_dvp); 3517 else 3518 vput(nd.ni_dvp); 3519 vput(vp); 3520 if ((error = vn_start_write(NULL, &mp, 3521 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3522 return (error); 3523 goto restart; 3524 } 3525 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3526 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3527 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3528 vn_finished_write(mp, 0); 3529 return (error); 3530 3531 out: 3532 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3533 if (nd.ni_dvp == vp) 3534 vrele(nd.ni_dvp); 3535 else 3536 vput(nd.ni_dvp); 3537 vput(vp); 3538 return (error); 3539 } 3540 3541 /* 3542 * Read a block of directory entries in a file system independent format. 3543 */ 3544 int 3545 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3546 { 3547 struct sys___getdents30_args /* { 3548 syscallarg(int) fd; 3549 syscallarg(char *) buf; 3550 syscallarg(size_t) count; 3551 } */ *uap = v; 3552 struct proc *p = l->l_proc; 3553 struct file *fp; 3554 int error, done; 3555 3556 /* getvnode() will use the descriptor for us */ 3557 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3558 return (error); 3559 if ((fp->f_flag & FREAD) == 0) { 3560 error = EBADF; 3561 goto out; 3562 } 3563 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3564 SCARG(uap, count), &done, l, 0, 0); 3565 #ifdef KTRACE 3566 if (!error && KTRPOINT(p, KTR_GENIO)) { 3567 struct iovec iov; 3568 iov.iov_base = SCARG(uap, buf); 3569 iov.iov_len = done; 3570 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3571 } 3572 #endif 3573 *retval = done; 3574 out: 3575 FILE_UNUSE(fp, l); 3576 return (error); 3577 } 3578 3579 /* 3580 * Set the mode mask for creation of filesystem nodes. 3581 */ 3582 int 3583 sys_umask(struct lwp *l, void *v, register_t *retval) 3584 { 3585 struct sys_umask_args /* { 3586 syscallarg(mode_t) newmask; 3587 } */ *uap = v; 3588 struct proc *p = l->l_proc; 3589 struct cwdinfo *cwdi; 3590 3591 cwdi = p->p_cwdi; 3592 *retval = cwdi->cwdi_cmask; 3593 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3594 return (0); 3595 } 3596 3597 /* 3598 * Void all references to file by ripping underlying filesystem 3599 * away from vnode. 3600 */ 3601 /* ARGSUSED */ 3602 int 3603 sys_revoke(struct lwp *l, void *v, register_t *retval) 3604 { 3605 struct sys_revoke_args /* { 3606 syscallarg(const char *) path; 3607 } */ *uap = v; 3608 struct mount *mp; 3609 struct vnode *vp; 3610 struct vattr vattr; 3611 int error; 3612 struct nameidata nd; 3613 3614 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3615 if ((error = namei(&nd)) != 0) 3616 return (error); 3617 vp = nd.ni_vp; 3618 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3619 goto out; 3620 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3621 (error = kauth_authorize_generic(l->l_cred, 3622 KAUTH_GENERIC_ISSUSER, &l->l_acflag)) != 0) 3623 goto out; 3624 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3625 goto out; 3626 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3627 VOP_REVOKE(vp, REVOKEALL); 3628 vn_finished_write(mp, 0); 3629 out: 3630 vrele(vp); 3631 return (error); 3632 } 3633 3634 /* 3635 * Convert a user file descriptor to a kernel file entry. 3636 */ 3637 int 3638 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3639 { 3640 struct vnode *vp; 3641 struct file *fp; 3642 3643 if ((fp = fd_getfile(fdp, fd)) == NULL) 3644 return (EBADF); 3645 3646 FILE_USE(fp); 3647 3648 if (fp->f_type != DTYPE_VNODE) { 3649 FILE_UNUSE(fp, NULL); 3650 return (EINVAL); 3651 } 3652 3653 vp = (struct vnode *)fp->f_data; 3654 if (vp->v_type == VBAD) { 3655 FILE_UNUSE(fp, NULL); 3656 return (EBADF); 3657 } 3658 3659 *fpp = fp; 3660 return (0); 3661 } 3662
Indexes created Sun Sep 28 07:09:56 GMT 2025