vfs_syscalls.c revision 1.251
1 /* $NetBSD: vfs_syscalls.c,v 1.251 2006/07/14 18:41:40 elad 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 1.251 2006/07/14 18:41:40 elad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_ktrace.h" 45 #include "opt_verified_exec.h" 46 #include "fss.h" 47 #include "opt_fileassoc.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 #ifdef VERIFIED_EXEC 73 #include <sys/verified_exec.h> 74 #endif /* VERIFIED_EXEC */ 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 proc *p = l->l_proc; 149 struct vnode *vp; 150 struct mount *mp; 151 int error, flag = 0; 152 char fstypename[MFSNAMELEN]; 153 struct vattr va; 154 struct nameidata nd; 155 struct vfsops *vfs; 156 157 /* 158 * if MNT_GETARGS is specified, it should be only flag. 159 */ 160 161 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 162 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 163 return EINVAL; 164 } 165 166 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 167 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 168 &p->p_acflag))) 169 return (error); 170 /* 171 * Get vnode to be covered 172 */ 173 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 174 SCARG(uap, path), l); 175 if ((error = namei(&nd)) != 0) 176 return (error); 177 vp = nd.ni_vp; 178 /* 179 * A lookup in VFS_MOUNT might result in an attempt to 180 * lock this vnode again, so make the lock recursive. 181 */ 182 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 183 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 184 if ((vp->v_flag & VROOT) == 0) { 185 vput(vp); 186 return (EINVAL); 187 } 188 mp = vp->v_mount; 189 flag = mp->mnt_flag; 190 vfs = mp->mnt_op; 191 /* 192 * We only allow the filesystem to be reloaded if it 193 * is currently mounted read-only. 194 */ 195 if ((SCARG(uap, flags) & MNT_RELOAD) && 196 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 197 vput(vp); 198 return (EOPNOTSUPP); /* Needs translation */ 199 } 200 /* 201 * In "highly secure" mode, don't let the caller do anything 202 * but downgrade a filesystem from read-write to read-only. 203 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 204 */ 205 if (securelevel >= 2 && 206 SCARG(uap, flags) != MNT_GETARGS && 207 SCARG(uap, flags) != 208 (mp->mnt_flag | MNT_RDONLY | 209 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 210 vput(vp); 211 return (EPERM); 212 } 213 mp->mnt_flag |= SCARG(uap, flags) & 214 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 215 /* 216 * Only root, or the user that did the original mount is 217 * permitted to update it. 218 */ 219 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 220 mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred) && 221 (error = kauth_authorize_generic(p->p_cred, 222 KAUTH_GENERIC_ISSUSER, 223 &p->p_acflag)) != 0) { 224 vput(vp); 225 return (error); 226 } 227 /* 228 * Do not allow NFS export by non-root users. For non-root 229 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 230 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 231 */ 232 if (kauth_cred_geteuid(p->p_cred) != 0) { 233 if (SCARG(uap, flags) & MNT_EXPORTED) { 234 vput(vp); 235 return (EPERM); 236 } 237 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 238 if (flag & MNT_NOEXEC) 239 SCARG(uap, flags) |= MNT_NOEXEC; 240 } 241 if (vfs_busy(mp, LK_NOWAIT, 0)) { 242 vput(vp); 243 return (EPERM); 244 } 245 goto update; 246 } else { 247 if (securelevel >= 2) { 248 vput(vp); 249 return (EPERM); 250 } 251 } 252 /* 253 * If the user is not root, ensure that they own the directory 254 * onto which we are attempting to mount. 255 */ 256 if ((error = VOP_GETATTR(vp, &va, p->p_cred, l)) != 0 || 257 (va.va_uid != kauth_cred_geteuid(p->p_cred) && 258 (error = kauth_authorize_generic(p->p_cred, 259 KAUTH_GENERIC_ISSUSER, 260 &p->p_acflag)) != 0)) { 261 vput(vp); 262 return (error); 263 } 264 /* 265 * Do not allow NFS export by non-root users. For non-root users, 266 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 267 * mount point is already MNT_NOEXEC. 268 */ 269 if (kauth_cred_geteuid(p->p_cred) != 0) { 270 if (SCARG(uap, flags) & MNT_EXPORTED) { 271 vput(vp); 272 return (EPERM); 273 } 274 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 275 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 276 SCARG(uap, flags) |= MNT_NOEXEC; 277 } 278 if ((error = vinvalbuf(vp, V_SAVE, p->p_cred, l, 0, 0)) != 0) { 279 vput(vp); 280 return (error); 281 } 282 if (vp->v_type != VDIR) { 283 vput(vp); 284 return (ENOTDIR); 285 } 286 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 287 if (error) { 288 #if defined(COMPAT_09) || defined(COMPAT_43) 289 /* 290 * Historically, filesystem types were identified by numbers. 291 * If we get an integer for the filesystem type instead of a 292 * string, we check to see if it matches one of the historic 293 * filesystem types. 294 */ 295 u_long fsindex = (u_long)SCARG(uap, type); 296 if (fsindex >= nmountcompatnames || 297 mountcompatnames[fsindex] == NULL) { 298 vput(vp); 299 return (ENODEV); 300 } 301 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 302 #else 303 vput(vp); 304 return (error); 305 #endif 306 } 307 #ifdef COMPAT_10 308 /* Accept `ufs' as an alias for `ffs'. */ 309 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 310 strncpy(fstypename, "ffs", MFSNAMELEN); 311 #endif 312 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 313 vput(vp); 314 return (ENODEV); 315 } 316 if (vp->v_mountedhere != NULL) { 317 vput(vp); 318 return (EBUSY); 319 } 320 321 /* 322 * Allocate and initialize the file system. 323 */ 324 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 325 M_MOUNT, M_WAITOK); 326 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 327 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 328 simple_lock_init(&mp->mnt_slock); 329 (void)vfs_busy(mp, LK_NOWAIT, 0); 330 mp->mnt_op = vfs; 331 vfs->vfs_refcount++; 332 mp->mnt_vnodecovered = vp; 333 mp->mnt_stat.f_owner = kauth_cred_geteuid(p->p_cred); 334 mp->mnt_unmounter = NULL; 335 mp->mnt_leaf = mp; 336 337 /* 338 * The underlying file system may refuse the mount for 339 * various reasons. Allow the user to force it to happen. 340 */ 341 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 342 update: 343 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 344 /* 345 * Set the mount level flags. 346 */ 347 if (SCARG(uap, flags) & MNT_RDONLY) 348 mp->mnt_flag |= MNT_RDONLY; 349 else if (mp->mnt_flag & MNT_RDONLY) 350 mp->mnt_iflag |= IMNT_WANTRDWR; 351 mp->mnt_flag &= 352 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 353 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 354 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 355 mp->mnt_flag |= SCARG(uap, flags) & 356 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 357 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 358 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 359 MNT_IGNORE); 360 } 361 /* 362 * Mount the filesystem. 363 */ 364 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 365 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 366 #if defined(COMPAT_30) && defined(NFSSERVER) 367 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 368 int error2; 369 370 /* Update failed; let's try and see if it was an 371 * export request. */ 372 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 373 SCARG(uap, data), l); 374 375 /* Only update error code if the export request was 376 * understood but some problem occurred while 377 * processing it. */ 378 if (error2 != EJUSTRETURN) 379 error = error2; 380 } 381 #endif 382 if (mp->mnt_iflag & IMNT_WANTRDWR) 383 mp->mnt_flag &= ~MNT_RDONLY; 384 if (error) 385 mp->mnt_flag = flag; 386 mp->mnt_flag &=~ 387 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 388 mp->mnt_iflag &=~ IMNT_WANTRDWR; 389 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 390 if (mp->mnt_syncer == NULL) 391 error = vfs_allocate_syncvnode(mp); 392 } else { 393 if (mp->mnt_syncer != NULL) 394 vfs_deallocate_syncvnode(mp); 395 } 396 vfs_unbusy(mp); 397 VOP_UNLOCK(vp, 0); 398 vrele(vp); 399 return (error); 400 } 401 /* 402 * Put the new filesystem on the mount list after root. 403 */ 404 cache_purge(vp); 405 if (!error) { 406 mp->mnt_flag &=~ 407 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 408 mp->mnt_iflag &=~ IMNT_WANTRDWR; 409 vp->v_mountedhere = mp; 410 simple_lock(&mountlist_slock); 411 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 412 simple_unlock(&mountlist_slock); 413 checkdirs(vp); 414 VOP_UNLOCK(vp, 0); 415 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 416 error = vfs_allocate_syncvnode(mp); 417 vfs_unbusy(mp); 418 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 419 if ((error = VFS_START(mp, 0, l))) 420 vrele(vp); 421 } else { 422 vp->v_mountedhere = (struct mount *)0; 423 vfs->vfs_refcount--; 424 vfs_unbusy(mp); 425 free(mp, M_MOUNT); 426 vput(vp); 427 } 428 return (error); 429 } 430 431 /* 432 * Scan all active processes to see if any of them have a current 433 * or root directory onto which the new filesystem has just been 434 * mounted. If so, replace them with the new mount point. 435 */ 436 void 437 checkdirs(struct vnode *olddp) 438 { 439 struct cwdinfo *cwdi; 440 struct vnode *newdp; 441 struct proc *p; 442 443 if (olddp->v_usecount == 1) 444 return; 445 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 446 panic("mount: lost mount"); 447 proclist_lock_read(); 448 PROCLIST_FOREACH(p, &allproc) { 449 cwdi = p->p_cwdi; 450 if (!cwdi) 451 continue; 452 if (cwdi->cwdi_cdir == olddp) { 453 vrele(cwdi->cwdi_cdir); 454 VREF(newdp); 455 cwdi->cwdi_cdir = newdp; 456 } 457 if (cwdi->cwdi_rdir == olddp) { 458 vrele(cwdi->cwdi_rdir); 459 VREF(newdp); 460 cwdi->cwdi_rdir = newdp; 461 } 462 } 463 proclist_unlock_read(); 464 if (rootvnode == olddp) { 465 vrele(rootvnode); 466 VREF(newdp); 467 rootvnode = newdp; 468 } 469 vput(newdp); 470 } 471 472 /* 473 * Unmount a file system. 474 * 475 * Note: unmount takes a path to the vnode mounted on as argument, 476 * not special file (as before). 477 */ 478 /* ARGSUSED */ 479 int 480 sys_unmount(struct lwp *l, void *v, register_t *retval) 481 { 482 struct sys_unmount_args /* { 483 syscallarg(const char *) path; 484 syscallarg(int) flags; 485 } */ *uap = v; 486 struct proc *p = l->l_proc; 487 struct vnode *vp; 488 struct mount *mp; 489 int error; 490 struct nameidata nd; 491 492 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 493 SCARG(uap, path), l); 494 if ((error = namei(&nd)) != 0) 495 return (error); 496 vp = nd.ni_vp; 497 mp = vp->v_mount; 498 499 /* 500 * Only root, or the user that did the original mount is 501 * permitted to unmount this filesystem. 502 */ 503 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred)) && 504 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 505 &p->p_acflag)) != 0) { 506 vput(vp); 507 return (error); 508 } 509 510 /* 511 * Don't allow unmounting the root file system. 512 */ 513 if (mp->mnt_flag & MNT_ROOTFS) { 514 vput(vp); 515 return (EINVAL); 516 } 517 518 /* 519 * Must be the root of the filesystem 520 */ 521 if ((vp->v_flag & VROOT) == 0) { 522 vput(vp); 523 return (EINVAL); 524 } 525 vput(vp); 526 527 /* 528 * XXX Freeze syncer. Must do this before locking the 529 * mount point. See dounmount() for details. 530 */ 531 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 532 533 if (vfs_busy(mp, 0, 0)) { 534 lockmgr(&syncer_lock, LK_RELEASE, NULL); 535 return (EBUSY); 536 } 537 538 return (dounmount(mp, SCARG(uap, flags), l)); 539 } 540 541 /* 542 * Do the actual file system unmount. File system is assumed to have been 543 * marked busy by the caller. 544 */ 545 int 546 dounmount(struct mount *mp, int flags, struct lwp *l) 547 { 548 struct vnode *coveredvp; 549 int error; 550 int async; 551 int used_syncer; 552 553 #ifdef VERIFIED_EXEC 554 if (!doing_shutdown) { 555 if (veriexec_strict >= 3) { 556 printf("Veriexec: Lockdown mode, preventing unmount of" 557 " \"%s\". (uid=%u)\n", mp->mnt_stat.f_mntonname, 558 kauth_cred_getuid(l->l_proc->p_cred)); 559 return (EPERM); 560 } 561 562 if (veriexec_strict == 2) { 563 struct veriexec_table_entry *vte; 564 565 /* Check if we have fingerprints on mount. */ 566 vte = fileassoc_tabledata_lookup(mp, veriexec_hook); 567 if ((vte != NULL) && (vte->vte_count > 0)) { 568 printf("Veriexec: IPS mode, preventing unmount" 569 " of \"%s\" with monitored files. " 570 "(uid=%u)\n", mp->mnt_stat.f_mntonname, 571 kauth_cred_getuid(l->l_proc->p_cred)); 572 return (EPERM); 573 } 574 } 575 } 576 #endif /* VERIFIED_EXEC */ 577 578 #ifdef FILEASSOC 579 (void)fileassoc_table_delete(mp); 580 #endif /* FILEASSOC */ 581 582 simple_lock(&mountlist_slock); 583 vfs_unbusy(mp); 584 used_syncer = (mp->mnt_syncer != NULL); 585 586 /* 587 * XXX Syncer must be frozen when we get here. This should really 588 * be done on a per-mountpoint basis, but especially the softdep 589 * code possibly called from the syncer doesn't exactly work on a 590 * per-mountpoint basis, so the softdep code would become a maze 591 * of vfs_busy() calls. 592 * 593 * The caller of dounmount() must acquire syncer_lock because 594 * the syncer itself acquires locks in syncer_lock -> vfs_busy 595 * order, and we must preserve that order to avoid deadlock. 596 * 597 * So, if the file system did not use the syncer, now is 598 * the time to release the syncer_lock. 599 */ 600 if (used_syncer == 0) 601 lockmgr(&syncer_lock, LK_RELEASE, NULL); 602 603 mp->mnt_iflag |= IMNT_UNMOUNT; 604 mp->mnt_unmounter = l; 605 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 606 vn_start_write(NULL, &mp, V_WAIT); 607 608 async = mp->mnt_flag & MNT_ASYNC; 609 mp->mnt_flag &= ~MNT_ASYNC; 610 cache_purgevfs(mp); /* remove cache entries for this file sys */ 611 if (mp->mnt_syncer != NULL) 612 vfs_deallocate_syncvnode(mp); 613 error = 0; 614 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 615 #if NFSS > 0 616 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 617 #endif 618 if (error == 0) 619 error = VFS_SYNC(mp, MNT_WAIT, l->l_proc->p_cred, l); 620 } 621 if (error == 0 || (flags & MNT_FORCE)) 622 error = VFS_UNMOUNT(mp, flags, l); 623 vn_finished_write(mp, 0); 624 simple_lock(&mountlist_slock); 625 if (error) { 626 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 627 (void) vfs_allocate_syncvnode(mp); 628 mp->mnt_iflag &= ~IMNT_UNMOUNT; 629 mp->mnt_unmounter = NULL; 630 mp->mnt_flag |= async; 631 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 632 &mountlist_slock); 633 if (used_syncer) 634 lockmgr(&syncer_lock, LK_RELEASE, NULL); 635 simple_lock(&mp->mnt_slock); 636 while (mp->mnt_wcnt > 0) { 637 wakeup(mp); 638 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 639 0, &mp->mnt_slock); 640 } 641 simple_unlock(&mp->mnt_slock); 642 return (error); 643 } 644 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 645 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 646 coveredvp->v_mountedhere = NULL; 647 vrele(coveredvp); 648 } 649 mp->mnt_op->vfs_refcount--; 650 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 651 panic("unmount: dangling vnode"); 652 mp->mnt_iflag |= IMNT_GONE; 653 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 654 if (used_syncer) 655 lockmgr(&syncer_lock, LK_RELEASE, NULL); 656 simple_lock(&mp->mnt_slock); 657 while (mp->mnt_wcnt > 0) { 658 wakeup(mp); 659 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 660 } 661 simple_unlock(&mp->mnt_slock); 662 vfs_hooks_unmount(mp); 663 free(mp, M_MOUNT); 664 return (0); 665 } 666 667 /* 668 * Sync each mounted filesystem. 669 */ 670 #ifdef DEBUG 671 int syncprt = 0; 672 struct ctldebug debug0 = { "syncprt", &syncprt }; 673 #endif 674 675 /* ARGSUSED */ 676 int 677 sys_sync(struct lwp *l, void *v, register_t *retval) 678 { 679 struct mount *mp, *nmp; 680 int asyncflag; 681 struct proc *p = l == NULL ? &proc0 : l->l_proc; 682 683 simple_lock(&mountlist_slock); 684 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 685 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 686 nmp = mp->mnt_list.cqe_prev; 687 continue; 688 } 689 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 690 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 691 asyncflag = mp->mnt_flag & MNT_ASYNC; 692 mp->mnt_flag &= ~MNT_ASYNC; 693 VFS_SYNC(mp, MNT_NOWAIT, p->p_cred, l); 694 if (asyncflag) 695 mp->mnt_flag |= MNT_ASYNC; 696 vn_finished_write(mp, 0); 697 } 698 simple_lock(&mountlist_slock); 699 nmp = mp->mnt_list.cqe_prev; 700 vfs_unbusy(mp); 701 702 } 703 simple_unlock(&mountlist_slock); 704 #ifdef DEBUG 705 if (syncprt) 706 vfs_bufstats(); 707 #endif /* DEBUG */ 708 return (0); 709 } 710 711 /* 712 * Change filesystem quotas. 713 */ 714 /* ARGSUSED */ 715 int 716 sys_quotactl(struct lwp *l, void *v, register_t *retval) 717 { 718 struct sys_quotactl_args /* { 719 syscallarg(const char *) path; 720 syscallarg(int) cmd; 721 syscallarg(int) uid; 722 syscallarg(caddr_t) arg; 723 } */ *uap = v; 724 struct mount *mp; 725 int error; 726 struct nameidata nd; 727 728 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 729 if ((error = namei(&nd)) != 0) 730 return (error); 731 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 732 vrele(nd.ni_vp); 733 if (error) 734 return (error); 735 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 736 SCARG(uap, arg), l); 737 vn_finished_write(mp, 0); 738 return (error); 739 } 740 741 int 742 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 743 int root) 744 { 745 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 746 int error = 0; 747 748 /* 749 * If MNT_NOWAIT or MNT_LAZY is specified, do not 750 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 751 * overrides MNT_NOWAIT. 752 */ 753 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 754 (flags != MNT_WAIT && flags != 0)) { 755 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 756 goto done; 757 } 758 759 /* Get the filesystem stats now */ 760 memset(sp, 0, sizeof(*sp)); 761 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 762 return error; 763 } 764 765 if (cwdi->cwdi_rdir == NULL) 766 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 767 done: 768 if (cwdi->cwdi_rdir != NULL) { 769 size_t len; 770 char *bp; 771 char *path = PNBUF_GET(); 772 if (!path) 773 return ENOMEM; 774 775 bp = path + MAXPATHLEN; 776 *--bp = '\0'; 777 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 778 MAXPATHLEN / 2, 0, l); 779 if (error) { 780 PNBUF_PUT(path); 781 return error; 782 } 783 len = strlen(bp); 784 /* 785 * for mount points that are below our root, we can see 786 * them, so we fix up the pathname and return them. The 787 * rest we cannot see, so we don't allow viewing the 788 * data. 789 */ 790 if (strncmp(bp, sp->f_mntonname, len) == 0) { 791 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 792 sizeof(sp->f_mntonname)); 793 if (sp->f_mntonname[0] == '\0') 794 (void)strlcpy(sp->f_mntonname, "/", 795 sizeof(sp->f_mntonname)); 796 } else { 797 if (root) 798 (void)strlcpy(sp->f_mntonname, "/", 799 sizeof(sp->f_mntonname)); 800 else 801 error = EPERM; 802 } 803 PNBUF_PUT(path); 804 } 805 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 806 return error; 807 } 808 809 /* 810 * Get filesystem statistics. 811 */ 812 /* ARGSUSED */ 813 int 814 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 815 { 816 struct sys_statvfs1_args /* { 817 syscallarg(const char *) path; 818 syscallarg(struct statvfs *) buf; 819 syscallarg(int) flags; 820 } */ *uap = v; 821 struct mount *mp; 822 struct statvfs *sb; 823 int error; 824 struct nameidata nd; 825 826 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 827 if ((error = namei(&nd)) != 0) 828 return error; 829 mp = nd.ni_vp->v_mount; 830 vrele(nd.ni_vp); 831 sb = STATVFSBUF_GET(); 832 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 833 if (error == 0) { 834 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 835 } 836 STATVFSBUF_PUT(sb); 837 return error; 838 } 839 840 /* 841 * Get filesystem statistics. 842 */ 843 /* ARGSUSED */ 844 int 845 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 846 { 847 struct sys_fstatvfs1_args /* { 848 syscallarg(int) fd; 849 syscallarg(struct statvfs *) buf; 850 syscallarg(int) flags; 851 } */ *uap = v; 852 struct proc *p = l->l_proc; 853 struct file *fp; 854 struct mount *mp; 855 struct statvfs *sb; 856 int error; 857 858 /* getvnode() will use the descriptor for us */ 859 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 860 return (error); 861 mp = ((struct vnode *)fp->f_data)->v_mount; 862 sb = STATVFSBUF_GET(); 863 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 864 goto out; 865 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 866 out: 867 FILE_UNUSE(fp, l); 868 STATVFSBUF_PUT(sb); 869 return error; 870 } 871 872 873 /* 874 * Get statistics on all filesystems. 875 */ 876 int 877 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 878 { 879 struct sys_getvfsstat_args /* { 880 syscallarg(struct statvfs *) buf; 881 syscallarg(size_t) bufsize; 882 syscallarg(int) flags; 883 } */ *uap = v; 884 int root = 0; 885 struct proc *p = l->l_proc; 886 struct mount *mp, *nmp; 887 struct statvfs *sb; 888 struct statvfs *sfsp; 889 size_t count, maxcount; 890 int error = 0; 891 892 sb = STATVFSBUF_GET(); 893 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 894 sfsp = SCARG(uap, buf); 895 simple_lock(&mountlist_slock); 896 count = 0; 897 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 898 mp = nmp) { 899 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 900 nmp = CIRCLEQ_NEXT(mp, mnt_list); 901 continue; 902 } 903 if (sfsp && count < maxcount) { 904 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 905 if (error) { 906 simple_lock(&mountlist_slock); 907 nmp = CIRCLEQ_NEXT(mp, mnt_list); 908 vfs_unbusy(mp); 909 continue; 910 } 911 error = copyout(sb, sfsp, sizeof(*sfsp)); 912 if (error) { 913 vfs_unbusy(mp); 914 goto out; 915 } 916 sfsp++; 917 root |= strcmp(sb->f_mntonname, "/") == 0; 918 } 919 count++; 920 simple_lock(&mountlist_slock); 921 nmp = CIRCLEQ_NEXT(mp, mnt_list); 922 vfs_unbusy(mp); 923 } 924 simple_unlock(&mountlist_slock); 925 if (root == 0 && p->p_cwdi->cwdi_rdir) { 926 /* 927 * fake a root entry 928 */ 929 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 930 SCARG(uap, flags), 1)) != 0) 931 goto out; 932 if (sfsp) 933 error = copyout(sb, sfsp, sizeof(*sfsp)); 934 count++; 935 } 936 if (sfsp && count > maxcount) 937 *retval = maxcount; 938 else 939 *retval = count; 940 out: 941 STATVFSBUF_PUT(sb); 942 return error; 943 } 944 945 /* 946 * Change current working directory to a given file descriptor. 947 */ 948 /* ARGSUSED */ 949 int 950 sys_fchdir(struct lwp *l, void *v, register_t *retval) 951 { 952 struct sys_fchdir_args /* { 953 syscallarg(int) fd; 954 } */ *uap = v; 955 struct proc *p = l->l_proc; 956 struct filedesc *fdp = p->p_fd; 957 struct cwdinfo *cwdi = p->p_cwdi; 958 struct vnode *vp, *tdp; 959 struct mount *mp; 960 struct file *fp; 961 int error; 962 963 /* getvnode() will use the descriptor for us */ 964 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 965 return (error); 966 vp = (struct vnode *)fp->f_data; 967 968 VREF(vp); 969 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 970 if (vp->v_type != VDIR) 971 error = ENOTDIR; 972 else 973 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 974 while (!error && (mp = vp->v_mountedhere) != NULL) { 975 if (vfs_busy(mp, 0, 0)) 976 continue; 977 error = VFS_ROOT(mp, &tdp); 978 vfs_unbusy(mp); 979 if (error) 980 break; 981 vput(vp); 982 vp = tdp; 983 } 984 if (error) { 985 vput(vp); 986 goto out; 987 } 988 VOP_UNLOCK(vp, 0); 989 990 /* 991 * Disallow changing to a directory not under the process's 992 * current root directory (if there is one). 993 */ 994 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 995 vrele(vp); 996 error = EPERM; /* operation not permitted */ 997 goto out; 998 } 999 1000 vrele(cwdi->cwdi_cdir); 1001 cwdi->cwdi_cdir = vp; 1002 out: 1003 FILE_UNUSE(fp, l); 1004 return (error); 1005 } 1006 1007 /* 1008 * Change this process's notion of the root directory to a given file 1009 * descriptor. 1010 */ 1011 int 1012 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1013 { 1014 struct sys_fchroot_args *uap = v; 1015 struct proc *p = l->l_proc; 1016 struct filedesc *fdp = p->p_fd; 1017 struct cwdinfo *cwdi = p->p_cwdi; 1018 struct vnode *vp; 1019 struct file *fp; 1020 int error; 1021 1022 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1023 &p->p_acflag)) != 0) 1024 return error; 1025 /* getvnode() will use the descriptor for us */ 1026 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1027 return error; 1028 vp = (struct vnode *) fp->f_data; 1029 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1030 if (vp->v_type != VDIR) 1031 error = ENOTDIR; 1032 else 1033 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 1034 VOP_UNLOCK(vp, 0); 1035 if (error) 1036 goto out; 1037 VREF(vp); 1038 1039 /* 1040 * Prevent escaping from chroot by putting the root under 1041 * the working directory. Silently chdir to / if we aren't 1042 * already there. 1043 */ 1044 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1045 /* 1046 * XXX would be more failsafe to change directory to a 1047 * deadfs node here instead 1048 */ 1049 vrele(cwdi->cwdi_cdir); 1050 VREF(vp); 1051 cwdi->cwdi_cdir = vp; 1052 } 1053 1054 if (cwdi->cwdi_rdir != NULL) 1055 vrele(cwdi->cwdi_rdir); 1056 cwdi->cwdi_rdir = vp; 1057 out: 1058 FILE_UNUSE(fp, l); 1059 return (error); 1060 } 1061 1062 /* 1063 * Change current working directory (``.''). 1064 */ 1065 /* ARGSUSED */ 1066 int 1067 sys_chdir(struct lwp *l, void *v, register_t *retval) 1068 { 1069 struct sys_chdir_args /* { 1070 syscallarg(const char *) path; 1071 } */ *uap = v; 1072 struct proc *p = l->l_proc; 1073 struct cwdinfo *cwdi = p->p_cwdi; 1074 int error; 1075 struct nameidata nd; 1076 1077 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1078 SCARG(uap, path), l); 1079 if ((error = change_dir(&nd, l)) != 0) 1080 return (error); 1081 vrele(cwdi->cwdi_cdir); 1082 cwdi->cwdi_cdir = nd.ni_vp; 1083 return (0); 1084 } 1085 1086 /* 1087 * Change notion of root (``/'') directory. 1088 */ 1089 /* ARGSUSED */ 1090 int 1091 sys_chroot(struct lwp *l, void *v, register_t *retval) 1092 { 1093 struct sys_chroot_args /* { 1094 syscallarg(const char *) path; 1095 } */ *uap = v; 1096 struct proc *p = l->l_proc; 1097 struct cwdinfo *cwdi = p->p_cwdi; 1098 struct vnode *vp; 1099 int error; 1100 struct nameidata nd; 1101 1102 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1103 &p->p_acflag)) != 0) 1104 return (error); 1105 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1106 SCARG(uap, path), l); 1107 if ((error = change_dir(&nd, l)) != 0) 1108 return (error); 1109 if (cwdi->cwdi_rdir != NULL) 1110 vrele(cwdi->cwdi_rdir); 1111 vp = nd.ni_vp; 1112 cwdi->cwdi_rdir = vp; 1113 1114 /* 1115 * Prevent escaping from chroot by putting the root under 1116 * the working directory. Silently chdir to / if we aren't 1117 * already there. 1118 */ 1119 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1120 /* 1121 * XXX would be more failsafe to change directory to a 1122 * deadfs node here instead 1123 */ 1124 vrele(cwdi->cwdi_cdir); 1125 VREF(vp); 1126 cwdi->cwdi_cdir = vp; 1127 } 1128 1129 return (0); 1130 } 1131 1132 /* 1133 * Common routine for chroot and chdir. 1134 */ 1135 static int 1136 change_dir(struct nameidata *ndp, struct lwp *l) 1137 { 1138 struct vnode *vp; 1139 int error; 1140 1141 if ((error = namei(ndp)) != 0) 1142 return (error); 1143 vp = ndp->ni_vp; 1144 if (vp->v_type != VDIR) 1145 error = ENOTDIR; 1146 else 1147 error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_cred, l); 1148 1149 if (error) 1150 vput(vp); 1151 else 1152 VOP_UNLOCK(vp, 0); 1153 return (error); 1154 } 1155 1156 /* 1157 * Check permissions, allocate an open file structure, 1158 * and call the device open routine if any. 1159 */ 1160 int 1161 sys_open(struct lwp *l, void *v, register_t *retval) 1162 { 1163 struct sys_open_args /* { 1164 syscallarg(const char *) path; 1165 syscallarg(int) flags; 1166 syscallarg(int) mode; 1167 } */ *uap = v; 1168 struct proc *p = l->l_proc; 1169 struct cwdinfo *cwdi = p->p_cwdi; 1170 struct filedesc *fdp = p->p_fd; 1171 struct file *fp; 1172 struct vnode *vp; 1173 int flags, cmode; 1174 int type, indx, error; 1175 struct flock lf; 1176 struct nameidata nd; 1177 1178 flags = FFLAGS(SCARG(uap, flags)); 1179 if ((flags & (FREAD | FWRITE)) == 0) 1180 return (EINVAL); 1181 /* falloc() will use the file descriptor for us */ 1182 if ((error = falloc(p, &fp, &indx)) != 0) 1183 return (error); 1184 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1185 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1186 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1187 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1188 FILE_UNUSE(fp, l); 1189 fdp->fd_ofiles[indx] = NULL; 1190 ffree(fp); 1191 if ((error == EDUPFD || error == EMOVEFD) && 1192 l->l_dupfd >= 0 && /* XXX from fdopen */ 1193 (error = 1194 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1195 *retval = indx; 1196 return (0); 1197 } 1198 if (error == ERESTART) 1199 error = EINTR; 1200 fdremove(fdp, indx); 1201 return (error); 1202 } 1203 l->l_dupfd = 0; 1204 vp = nd.ni_vp; 1205 fp->f_flag = flags & FMASK; 1206 fp->f_type = DTYPE_VNODE; 1207 fp->f_ops = &vnops; 1208 fp->f_data = vp; 1209 if (flags & (O_EXLOCK | O_SHLOCK)) { 1210 lf.l_whence = SEEK_SET; 1211 lf.l_start = 0; 1212 lf.l_len = 0; 1213 if (flags & O_EXLOCK) 1214 lf.l_type = F_WRLCK; 1215 else 1216 lf.l_type = F_RDLCK; 1217 type = F_FLOCK; 1218 if ((flags & FNONBLOCK) == 0) 1219 type |= F_WAIT; 1220 VOP_UNLOCK(vp, 0); 1221 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1222 if (error) { 1223 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1224 FILE_UNUSE(fp, l); 1225 ffree(fp); 1226 fdremove(fdp, indx); 1227 return (error); 1228 } 1229 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1230 fp->f_flag |= FHASLOCK; 1231 } 1232 VOP_UNLOCK(vp, 0); 1233 *retval = indx; 1234 FILE_SET_MATURE(fp); 1235 FILE_UNUSE(fp, l); 1236 return (0); 1237 } 1238 1239 static void 1240 vfs__fhfree(fhandle_t *fhp) 1241 { 1242 size_t fhsize; 1243 1244 if (fhp == NULL) { 1245 return; 1246 } 1247 fhsize = FHANDLE_SIZE(fhp); 1248 kmem_free(fhp, fhsize); 1249 } 1250 1251 /* 1252 * vfs_composefh: compose a filehandle. 1253 */ 1254 1255 int 1256 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1257 { 1258 struct mount *mp; 1259 struct fid *fidp; 1260 int error; 1261 size_t needfhsize; 1262 size_t fidsize; 1263 1264 mp = vp->v_mount; 1265 if (mp->mnt_op->vfs_vptofh == NULL) { 1266 return EOPNOTSUPP; 1267 } 1268 fidp = NULL; 1269 if (*fh_size <= FHANDLE_SIZE_MIN) { 1270 fidsize = 0; 1271 } else { 1272 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1273 if (fhp != NULL) { 1274 memset(fhp, 0, *fh_size); 1275 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1276 fidp = &fhp->fh_fid; 1277 } 1278 } 1279 error = VFS_VPTOFH(vp, fidp, &fidsize); 1280 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1281 if (error == 0 && *fh_size < needfhsize) { 1282 error = E2BIG; 1283 } 1284 *fh_size = needfhsize; 1285 return error; 1286 } 1287 1288 int 1289 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1290 { 1291 struct mount *mp; 1292 fhandle_t *fhp; 1293 size_t fhsize; 1294 size_t fidsize; 1295 int error; 1296 1297 *fhpp = NULL; 1298 mp = vp->v_mount; 1299 if (mp->mnt_op->vfs_vptofh == NULL) { 1300 error = EOPNOTSUPP; 1301 goto out; 1302 } 1303 fhsize = 0; 1304 error = VFS_VPTOFH(vp, NULL, &fidsize); 1305 KASSERT(error != 0); 1306 if (error != E2BIG) { 1307 goto out; 1308 } 1309 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1310 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1311 if (fhp == NULL) { 1312 error = ENOMEM; 1313 goto out; 1314 } 1315 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1316 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1317 if (error == 0) { 1318 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1319 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1320 *fhpp = fhp; 1321 } else { 1322 kmem_free(fhp, fhsize); 1323 } 1324 out: 1325 return error; 1326 } 1327 1328 void 1329 vfs_composefh_free(fhandle_t *fhp) 1330 { 1331 1332 vfs__fhfree(fhp); 1333 } 1334 1335 /* 1336 * vfs_fhtovp: lookup a vnode by a filehandle. 1337 */ 1338 1339 int 1340 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1341 { 1342 struct mount *mp; 1343 int error; 1344 1345 *vpp = NULL; 1346 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1347 if (mp == NULL) { 1348 error = ESTALE; 1349 goto out; 1350 } 1351 if (mp->mnt_op->vfs_fhtovp == NULL) { 1352 error = EOPNOTSUPP; 1353 goto out; 1354 } 1355 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1356 out: 1357 return error; 1358 } 1359 1360 /* 1361 * vfs_copyinfh: copyin a filehandle. 1362 */ 1363 1364 int 1365 vfs_copyinfh_alloc(const void *ufhp, fhandle_t **fhpp) 1366 { 1367 fhandle_t *fhp; 1368 fhandle_t tempfh; 1369 size_t fhsize; 1370 int error; 1371 1372 *fhpp = NULL; 1373 error = copyin(ufhp, &tempfh, sizeof(tempfh)); 1374 if (error) { 1375 return error; 1376 } 1377 fhsize = FHANDLE_SIZE(&tempfh); 1378 if (fhsize > FHANDLE_SIZE_MAX) { 1379 return EINVAL; 1380 } 1381 fhp = kmem_alloc(fhsize, KM_SLEEP); 1382 if (fhp == NULL) { 1383 return ENOMEM; 1384 } 1385 error = copyin(ufhp, fhp, fhsize); 1386 if (error == 0) { 1387 *fhpp = fhp; 1388 } else { 1389 kmem_free(fhp, fhsize); 1390 } 1391 return error; 1392 } 1393 1394 void 1395 vfs_copyinfh_free(fhandle_t *fhp) 1396 { 1397 1398 vfs__fhfree(fhp); 1399 } 1400 1401 /* 1402 * Get file handle system call 1403 */ 1404 int 1405 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1406 { 1407 struct sys___getfh30_args /* { 1408 syscallarg(char *) fname; 1409 syscallarg(fhandle_t *) fhp; 1410 syscallarg(size_t *) fh_size; 1411 } */ *uap = v; 1412 struct proc *p = l->l_proc; 1413 struct vnode *vp; 1414 fhandle_t *fh; 1415 int error; 1416 struct nameidata nd; 1417 size_t sz; 1418 size_t usz; 1419 1420 /* 1421 * Must be super user 1422 */ 1423 error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1424 &p->p_acflag); 1425 if (error) 1426 return (error); 1427 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1428 SCARG(uap, fname), l); 1429 error = namei(&nd); 1430 if (error) 1431 return (error); 1432 vp = nd.ni_vp; 1433 error = vfs_composefh_alloc(vp, &fh); 1434 vput(vp); 1435 if (error != 0) { 1436 goto out; 1437 } 1438 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1439 if (error != 0) { 1440 goto out; 1441 } 1442 sz = FHANDLE_SIZE(fh); 1443 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1444 if (error != 0) { 1445 goto out; 1446 } 1447 if (usz >= sz) { 1448 error = copyout(fh, SCARG(uap, fhp), sz); 1449 } else { 1450 error = E2BIG; 1451 } 1452 out: 1453 vfs_composefh_free(fh); 1454 return (error); 1455 } 1456 1457 /* 1458 * Open a file given a file handle. 1459 * 1460 * Check permissions, allocate an open file structure, 1461 * and call the device open routine if any. 1462 */ 1463 int 1464 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1465 { 1466 struct sys_fhopen_args /* { 1467 syscallarg(const fhandle_t *) fhp; 1468 syscallarg(int) flags; 1469 } */ *uap = v; 1470 struct proc *p = l->l_proc; 1471 struct filedesc *fdp = p->p_fd; 1472 struct file *fp; 1473 struct vnode *vp = NULL; 1474 struct mount *mp; 1475 kauth_cred_t cred = p->p_cred; 1476 int flags; 1477 struct file *nfp; 1478 int type, indx, error=0; 1479 struct flock lf; 1480 struct vattr va; 1481 fhandle_t *fh; 1482 1483 /* 1484 * Must be super user 1485 */ 1486 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1487 &p->p_acflag))) 1488 return (error); 1489 1490 flags = FFLAGS(SCARG(uap, flags)); 1491 if ((flags & (FREAD | FWRITE)) == 0) 1492 return (EINVAL); 1493 if ((flags & O_CREAT)) 1494 return (EINVAL); 1495 /* falloc() will use the file descriptor for us */ 1496 if ((error = falloc(p, &nfp, &indx)) != 0) 1497 return (error); 1498 fp = nfp; 1499 error = vfs_copyinfh_alloc(SCARG(uap, fhp), &fh); 1500 if (error != 0) { 1501 goto bad; 1502 } 1503 error = vfs_fhtovp(fh, &vp); 1504 if (error != 0) { 1505 goto bad; 1506 } 1507 1508 /* Now do an effective vn_open */ 1509 1510 if (vp->v_type == VSOCK) { 1511 error = EOPNOTSUPP; 1512 goto bad; 1513 } 1514 if (flags & FREAD) { 1515 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1516 goto bad; 1517 } 1518 if (flags & (FWRITE | O_TRUNC)) { 1519 if (vp->v_type == VDIR) { 1520 error = EISDIR; 1521 goto bad; 1522 } 1523 if ((error = vn_writechk(vp)) != 0 || 1524 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1525 goto bad; 1526 } 1527 if (flags & O_TRUNC) { 1528 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1529 goto bad; 1530 VOP_UNLOCK(vp, 0); /* XXX */ 1531 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1532 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1533 VATTR_NULL(&va); 1534 va.va_size = 0; 1535 error = VOP_SETATTR(vp, &va, cred, l); 1536 vn_finished_write(mp, 0); 1537 if (error) 1538 goto bad; 1539 } 1540 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1541 goto bad; 1542 if (vp->v_type == VREG && 1543 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1544 error = EIO; 1545 goto bad; 1546 } 1547 if (flags & FWRITE) 1548 vp->v_writecount++; 1549 1550 /* done with modified vn_open, now finish what sys_open does. */ 1551 1552 fp->f_flag = flags & FMASK; 1553 fp->f_type = DTYPE_VNODE; 1554 fp->f_ops = &vnops; 1555 fp->f_data = vp; 1556 if (flags & (O_EXLOCK | O_SHLOCK)) { 1557 lf.l_whence = SEEK_SET; 1558 lf.l_start = 0; 1559 lf.l_len = 0; 1560 if (flags & O_EXLOCK) 1561 lf.l_type = F_WRLCK; 1562 else 1563 lf.l_type = F_RDLCK; 1564 type = F_FLOCK; 1565 if ((flags & FNONBLOCK) == 0) 1566 type |= F_WAIT; 1567 VOP_UNLOCK(vp, 0); 1568 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1569 if (error) { 1570 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1571 FILE_UNUSE(fp, l); 1572 ffree(fp); 1573 fdremove(fdp, indx); 1574 return (error); 1575 } 1576 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1577 fp->f_flag |= FHASLOCK; 1578 } 1579 VOP_UNLOCK(vp, 0); 1580 *retval = indx; 1581 FILE_SET_MATURE(fp); 1582 FILE_UNUSE(fp, l); 1583 vfs_copyinfh_free(fh); 1584 return (0); 1585 1586 bad: 1587 FILE_UNUSE(fp, l); 1588 ffree(fp); 1589 fdremove(fdp, indx); 1590 if (vp != NULL) 1591 vput(vp); 1592 vfs_copyinfh_free(fh); 1593 return (error); 1594 } 1595 1596 /* ARGSUSED */ 1597 int 1598 sys___fhstat30(struct lwp *l, void *v, register_t *retval) 1599 { 1600 struct sys___fhstat30_args /* { 1601 syscallarg(const fhandle_t *) fhp; 1602 syscallarg(struct stat *) sb; 1603 } */ *uap = v; 1604 struct proc *p = l->l_proc; 1605 struct stat sb; 1606 int error; 1607 fhandle_t *fh; 1608 struct vnode *vp; 1609 1610 /* 1611 * Must be super user 1612 */ 1613 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1614 &p->p_acflag))) 1615 return (error); 1616 1617 error = vfs_copyinfh_alloc(SCARG(uap, fhp), &fh); 1618 if (error != 0) { 1619 goto bad; 1620 } 1621 error = vfs_fhtovp(fh, &vp); 1622 if (error != 0) { 1623 goto bad; 1624 } 1625 error = vn_stat(vp, &sb, l); 1626 vput(vp); 1627 if (error) { 1628 goto bad; 1629 } 1630 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1631 bad: 1632 vfs_copyinfh_free(fh); 1633 return error; 1634 } 1635 1636 /* ARGSUSED */ 1637 int 1638 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1639 { 1640 struct sys_fhstatvfs1_args /* { 1641 syscallarg(const fhandle_t *) fhp; 1642 syscallarg(struct statvfs *) buf; 1643 syscallarg(int) flags; 1644 } */ *uap = v; 1645 struct proc *p = l->l_proc; 1646 struct statvfs *sb = NULL; 1647 fhandle_t *fh; 1648 struct mount *mp; 1649 struct vnode *vp; 1650 int error; 1651 1652 /* 1653 * Must be super user 1654 */ 1655 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1656 &p->p_acflag)) != 0) 1657 return error; 1658 1659 error = vfs_copyinfh_alloc(SCARG(uap, fhp), &fh); 1660 if (error != 0) { 1661 goto out; 1662 } 1663 error = vfs_fhtovp(fh, &vp); 1664 if (error != 0) { 1665 goto out; 1666 } 1667 mp = vp->v_mount; 1668 sb = STATVFSBUF_GET(); 1669 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) { 1670 vput(vp); 1671 goto out; 1672 } 1673 vput(vp); 1674 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1675 out: 1676 if (sb != NULL) { 1677 STATVFSBUF_PUT(sb); 1678 } 1679 vfs_copyinfh_free(fh); 1680 return error; 1681 } 1682 1683 /* 1684 * Create a special file. 1685 */ 1686 /* ARGSUSED */ 1687 int 1688 sys_mknod(struct lwp *l, void *v, register_t *retval) 1689 { 1690 struct sys_mknod_args /* { 1691 syscallarg(const char *) path; 1692 syscallarg(int) mode; 1693 syscallarg(int) dev; 1694 } */ *uap = v; 1695 struct proc *p = l->l_proc; 1696 struct vnode *vp; 1697 struct mount *mp; 1698 struct vattr vattr; 1699 int error; 1700 int whiteout = 0; 1701 struct nameidata nd; 1702 1703 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1704 &p->p_acflag)) != 0) 1705 return (error); 1706 restart: 1707 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1708 if ((error = namei(&nd)) != 0) 1709 return (error); 1710 vp = nd.ni_vp; 1711 if (vp != NULL) 1712 error = EEXIST; 1713 else { 1714 VATTR_NULL(&vattr); 1715 vattr.va_mode = 1716 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1717 vattr.va_rdev = SCARG(uap, dev); 1718 whiteout = 0; 1719 1720 switch (SCARG(uap, mode) & S_IFMT) { 1721 case S_IFMT: /* used by badsect to flag bad sectors */ 1722 vattr.va_type = VBAD; 1723 break; 1724 case S_IFCHR: 1725 vattr.va_type = VCHR; 1726 break; 1727 case S_IFBLK: 1728 vattr.va_type = VBLK; 1729 break; 1730 case S_IFWHT: 1731 whiteout = 1; 1732 break; 1733 default: 1734 error = EINVAL; 1735 break; 1736 } 1737 } 1738 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1739 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1740 if (nd.ni_dvp == vp) 1741 vrele(nd.ni_dvp); 1742 else 1743 vput(nd.ni_dvp); 1744 if (vp) 1745 vrele(vp); 1746 if ((error = vn_start_write(NULL, &mp, 1747 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1748 return (error); 1749 goto restart; 1750 } 1751 if (!error) { 1752 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1753 if (whiteout) { 1754 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1755 if (error) 1756 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1757 vput(nd.ni_dvp); 1758 } else { 1759 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1760 &nd.ni_cnd, &vattr); 1761 if (error == 0) 1762 vput(nd.ni_vp); 1763 } 1764 } else { 1765 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1766 if (nd.ni_dvp == vp) 1767 vrele(nd.ni_dvp); 1768 else 1769 vput(nd.ni_dvp); 1770 if (vp) 1771 vrele(vp); 1772 } 1773 vn_finished_write(mp, 0); 1774 return (error); 1775 } 1776 1777 /* 1778 * Create a named pipe. 1779 */ 1780 /* ARGSUSED */ 1781 int 1782 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1783 { 1784 struct sys_mkfifo_args /* { 1785 syscallarg(const char *) path; 1786 syscallarg(int) mode; 1787 } */ *uap = v; 1788 struct proc *p = l->l_proc; 1789 struct mount *mp; 1790 struct vattr vattr; 1791 int error; 1792 struct nameidata nd; 1793 1794 restart: 1795 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1796 if ((error = namei(&nd)) != 0) 1797 return (error); 1798 if (nd.ni_vp != NULL) { 1799 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1800 if (nd.ni_dvp == nd.ni_vp) 1801 vrele(nd.ni_dvp); 1802 else 1803 vput(nd.ni_dvp); 1804 vrele(nd.ni_vp); 1805 return (EEXIST); 1806 } 1807 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1808 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1809 if (nd.ni_dvp == nd.ni_vp) 1810 vrele(nd.ni_dvp); 1811 else 1812 vput(nd.ni_dvp); 1813 if (nd.ni_vp) 1814 vrele(nd.ni_vp); 1815 if ((error = vn_start_write(NULL, &mp, 1816 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1817 return (error); 1818 goto restart; 1819 } 1820 VATTR_NULL(&vattr); 1821 vattr.va_type = VFIFO; 1822 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1823 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1824 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1825 if (error == 0) 1826 vput(nd.ni_vp); 1827 vn_finished_write(mp, 0); 1828 return (error); 1829 } 1830 1831 /* 1832 * Make a hard file link. 1833 */ 1834 /* ARGSUSED */ 1835 int 1836 sys_link(struct lwp *l, void *v, register_t *retval) 1837 { 1838 struct sys_link_args /* { 1839 syscallarg(const char *) path; 1840 syscallarg(const char *) link; 1841 } */ *uap = v; 1842 struct proc *p = l->l_proc; 1843 struct vnode *vp; 1844 struct mount *mp; 1845 struct nameidata nd; 1846 int error; 1847 1848 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1849 if ((error = namei(&nd)) != 0) 1850 return (error); 1851 vp = nd.ni_vp; 1852 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1853 vrele(vp); 1854 return (error); 1855 } 1856 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1857 if ((error = namei(&nd)) != 0) 1858 goto out; 1859 if (nd.ni_vp) { 1860 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1861 if (nd.ni_dvp == nd.ni_vp) 1862 vrele(nd.ni_dvp); 1863 else 1864 vput(nd.ni_dvp); 1865 vrele(nd.ni_vp); 1866 error = EEXIST; 1867 goto out; 1868 } 1869 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1870 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1871 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1872 out: 1873 vrele(vp); 1874 vn_finished_write(mp, 0); 1875 return (error); 1876 } 1877 1878 /* 1879 * Make a symbolic link. 1880 */ 1881 /* ARGSUSED */ 1882 int 1883 sys_symlink(struct lwp *l, void *v, register_t *retval) 1884 { 1885 struct sys_symlink_args /* { 1886 syscallarg(const char *) path; 1887 syscallarg(const char *) link; 1888 } */ *uap = v; 1889 struct proc *p = l->l_proc; 1890 struct mount *mp; 1891 struct vattr vattr; 1892 char *path; 1893 int error; 1894 struct nameidata nd; 1895 1896 path = PNBUF_GET(); 1897 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1898 if (error) 1899 goto out; 1900 restart: 1901 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1902 if ((error = namei(&nd)) != 0) 1903 goto out; 1904 if (nd.ni_vp) { 1905 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1906 if (nd.ni_dvp == nd.ni_vp) 1907 vrele(nd.ni_dvp); 1908 else 1909 vput(nd.ni_dvp); 1910 vrele(nd.ni_vp); 1911 error = EEXIST; 1912 goto out; 1913 } 1914 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1915 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1916 if (nd.ni_dvp == nd.ni_vp) 1917 vrele(nd.ni_dvp); 1918 else 1919 vput(nd.ni_dvp); 1920 if ((error = vn_start_write(NULL, &mp, 1921 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1922 return (error); 1923 goto restart; 1924 } 1925 VATTR_NULL(&vattr); 1926 vattr.va_type = VLNK; 1927 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1928 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1929 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1930 if (error == 0) 1931 vput(nd.ni_vp); 1932 vn_finished_write(mp, 0); 1933 out: 1934 PNBUF_PUT(path); 1935 return (error); 1936 } 1937 1938 /* 1939 * Delete a whiteout from the filesystem. 1940 */ 1941 /* ARGSUSED */ 1942 int 1943 sys_undelete(struct lwp *l, void *v, register_t *retval) 1944 { 1945 struct sys_undelete_args /* { 1946 syscallarg(const char *) path; 1947 } */ *uap = v; 1948 struct proc *p = l->l_proc; 1949 int error; 1950 struct mount *mp; 1951 struct nameidata nd; 1952 1953 restart: 1954 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1955 SCARG(uap, path), l); 1956 error = namei(&nd); 1957 if (error) 1958 return (error); 1959 1960 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1961 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1962 if (nd.ni_dvp == nd.ni_vp) 1963 vrele(nd.ni_dvp); 1964 else 1965 vput(nd.ni_dvp); 1966 if (nd.ni_vp) 1967 vrele(nd.ni_vp); 1968 return (EEXIST); 1969 } 1970 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1971 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1972 if (nd.ni_dvp == nd.ni_vp) 1973 vrele(nd.ni_dvp); 1974 else 1975 vput(nd.ni_dvp); 1976 if ((error = vn_start_write(NULL, &mp, 1977 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1978 return (error); 1979 goto restart; 1980 } 1981 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1982 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1983 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1984 vput(nd.ni_dvp); 1985 vn_finished_write(mp, 0); 1986 return (error); 1987 } 1988 1989 /* 1990 * Delete a name from the filesystem. 1991 */ 1992 /* ARGSUSED */ 1993 int 1994 sys_unlink(struct lwp *l, void *v, register_t *retval) 1995 { 1996 struct sys_unlink_args /* { 1997 syscallarg(const char *) path; 1998 } */ *uap = v; 1999 struct proc *p = l->l_proc; 2000 struct mount *mp; 2001 struct vnode *vp; 2002 int error; 2003 struct nameidata nd; 2004 2005 restart: 2006 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2007 SCARG(uap, path), l); 2008 if ((error = namei(&nd)) != 0) 2009 return (error); 2010 vp = nd.ni_vp; 2011 2012 /* 2013 * The root of a mounted filesystem cannot be deleted. 2014 */ 2015 if (vp->v_flag & VROOT) { 2016 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2017 if (nd.ni_dvp == vp) 2018 vrele(nd.ni_dvp); 2019 else 2020 vput(nd.ni_dvp); 2021 vput(vp); 2022 error = EBUSY; 2023 goto out; 2024 } 2025 2026 #ifdef VERIFIED_EXEC 2027 /* Handle remove requests for veriexec entries. */ 2028 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) { 2029 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2030 if (nd.ni_dvp == vp) 2031 vrele(nd.ni_dvp); 2032 else 2033 vput(nd.ni_dvp); 2034 vput(vp); 2035 goto out; 2036 } 2037 #endif /* VERIFIED_EXEC */ 2038 2039 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2040 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2041 if (nd.ni_dvp == vp) 2042 vrele(nd.ni_dvp); 2043 else 2044 vput(nd.ni_dvp); 2045 vput(vp); 2046 if ((error = vn_start_write(NULL, &mp, 2047 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2048 return (error); 2049 goto restart; 2050 } 2051 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 2052 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2053 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2054 vn_finished_write(mp, 0); 2055 #ifdef FILEASSOC 2056 if (!error) 2057 (void)fileassoc_file_delete(nd.ni_vp); 2058 #endif /* FILEASSOC */ 2059 out: 2060 return (error); 2061 } 2062 2063 /* 2064 * Reposition read/write file offset. 2065 */ 2066 int 2067 sys_lseek(struct lwp *l, void *v, register_t *retval) 2068 { 2069 struct sys_lseek_args /* { 2070 syscallarg(int) fd; 2071 syscallarg(int) pad; 2072 syscallarg(off_t) offset; 2073 syscallarg(int) whence; 2074 } */ *uap = v; 2075 struct proc *p = l->l_proc; 2076 kauth_cred_t cred = p->p_cred; 2077 struct filedesc *fdp = p->p_fd; 2078 struct file *fp; 2079 struct vnode *vp; 2080 struct vattr vattr; 2081 off_t newoff; 2082 int error; 2083 2084 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2085 return (EBADF); 2086 2087 FILE_USE(fp); 2088 2089 vp = (struct vnode *)fp->f_data; 2090 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2091 error = ESPIPE; 2092 goto out; 2093 } 2094 2095 switch (SCARG(uap, whence)) { 2096 case SEEK_CUR: 2097 newoff = fp->f_offset + SCARG(uap, offset); 2098 break; 2099 case SEEK_END: 2100 error = VOP_GETATTR(vp, &vattr, cred, l); 2101 if (error) 2102 goto out; 2103 newoff = SCARG(uap, offset) + vattr.va_size; 2104 break; 2105 case SEEK_SET: 2106 newoff = SCARG(uap, offset); 2107 break; 2108 default: 2109 error = EINVAL; 2110 goto out; 2111 } 2112 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2113 goto out; 2114 2115 *(off_t *)retval = fp->f_offset = newoff; 2116 out: 2117 FILE_UNUSE(fp, l); 2118 return (error); 2119 } 2120 2121 /* 2122 * Positional read system call. 2123 */ 2124 int 2125 sys_pread(struct lwp *l, void *v, register_t *retval) 2126 { 2127 struct sys_pread_args /* { 2128 syscallarg(int) fd; 2129 syscallarg(void *) buf; 2130 syscallarg(size_t) nbyte; 2131 syscallarg(off_t) offset; 2132 } */ *uap = v; 2133 struct proc *p = l->l_proc; 2134 struct filedesc *fdp = p->p_fd; 2135 struct file *fp; 2136 struct vnode *vp; 2137 off_t offset; 2138 int error, fd = SCARG(uap, fd); 2139 2140 if ((fp = fd_getfile(fdp, fd)) == NULL) 2141 return (EBADF); 2142 2143 if ((fp->f_flag & FREAD) == 0) { 2144 simple_unlock(&fp->f_slock); 2145 return (EBADF); 2146 } 2147 2148 FILE_USE(fp); 2149 2150 vp = (struct vnode *)fp->f_data; 2151 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2152 error = ESPIPE; 2153 goto out; 2154 } 2155 2156 offset = SCARG(uap, offset); 2157 2158 /* 2159 * XXX This works because no file systems actually 2160 * XXX take any action on the seek operation. 2161 */ 2162 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2163 goto out; 2164 2165 /* dofileread() will unuse the descriptor for us */ 2166 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2167 &offset, 0, retval)); 2168 2169 out: 2170 FILE_UNUSE(fp, l); 2171 return (error); 2172 } 2173 2174 /* 2175 * Positional scatter read system call. 2176 */ 2177 int 2178 sys_preadv(struct lwp *l, void *v, register_t *retval) 2179 { 2180 struct sys_preadv_args /* { 2181 syscallarg(int) fd; 2182 syscallarg(const struct iovec *) iovp; 2183 syscallarg(int) iovcnt; 2184 syscallarg(off_t) offset; 2185 } */ *uap = v; 2186 struct proc *p = l->l_proc; 2187 struct filedesc *fdp = p->p_fd; 2188 struct file *fp; 2189 struct vnode *vp; 2190 off_t offset; 2191 int error, fd = SCARG(uap, fd); 2192 2193 if ((fp = fd_getfile(fdp, fd)) == NULL) 2194 return (EBADF); 2195 2196 if ((fp->f_flag & FREAD) == 0) { 2197 simple_unlock(&fp->f_slock); 2198 return (EBADF); 2199 } 2200 2201 FILE_USE(fp); 2202 2203 vp = (struct vnode *)fp->f_data; 2204 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2205 error = ESPIPE; 2206 goto out; 2207 } 2208 2209 offset = SCARG(uap, offset); 2210 2211 /* 2212 * XXX This works because no file systems actually 2213 * XXX take any action on the seek operation. 2214 */ 2215 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2216 goto out; 2217 2218 /* dofilereadv() will unuse the descriptor for us */ 2219 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2220 &offset, 0, retval)); 2221 2222 out: 2223 FILE_UNUSE(fp, l); 2224 return (error); 2225 } 2226 2227 /* 2228 * Positional write system call. 2229 */ 2230 int 2231 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2232 { 2233 struct sys_pwrite_args /* { 2234 syscallarg(int) fd; 2235 syscallarg(const void *) buf; 2236 syscallarg(size_t) nbyte; 2237 syscallarg(off_t) offset; 2238 } */ *uap = v; 2239 struct proc *p = l->l_proc; 2240 struct filedesc *fdp = p->p_fd; 2241 struct file *fp; 2242 struct vnode *vp; 2243 off_t offset; 2244 int error, fd = SCARG(uap, fd); 2245 2246 if ((fp = fd_getfile(fdp, fd)) == NULL) 2247 return (EBADF); 2248 2249 if ((fp->f_flag & FWRITE) == 0) { 2250 simple_unlock(&fp->f_slock); 2251 return (EBADF); 2252 } 2253 2254 FILE_USE(fp); 2255 2256 vp = (struct vnode *)fp->f_data; 2257 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2258 error = ESPIPE; 2259 goto out; 2260 } 2261 2262 offset = SCARG(uap, offset); 2263 2264 /* 2265 * XXX This works because no file systems actually 2266 * XXX take any action on the seek operation. 2267 */ 2268 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2269 goto out; 2270 2271 /* dofilewrite() will unuse the descriptor for us */ 2272 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2273 &offset, 0, retval)); 2274 2275 out: 2276 FILE_UNUSE(fp, l); 2277 return (error); 2278 } 2279 2280 /* 2281 * Positional gather write system call. 2282 */ 2283 int 2284 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2285 { 2286 struct sys_pwritev_args /* { 2287 syscallarg(int) fd; 2288 syscallarg(const struct iovec *) iovp; 2289 syscallarg(int) iovcnt; 2290 syscallarg(off_t) offset; 2291 } */ *uap = v; 2292 struct proc *p = l->l_proc; 2293 struct filedesc *fdp = p->p_fd; 2294 struct file *fp; 2295 struct vnode *vp; 2296 off_t offset; 2297 int error, fd = SCARG(uap, fd); 2298 2299 if ((fp = fd_getfile(fdp, fd)) == NULL) 2300 return (EBADF); 2301 2302 if ((fp->f_flag & FWRITE) == 0) { 2303 simple_unlock(&fp->f_slock); 2304 return (EBADF); 2305 } 2306 2307 FILE_USE(fp); 2308 2309 vp = (struct vnode *)fp->f_data; 2310 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2311 error = ESPIPE; 2312 goto out; 2313 } 2314 2315 offset = SCARG(uap, offset); 2316 2317 /* 2318 * XXX This works because no file systems actually 2319 * XXX take any action on the seek operation. 2320 */ 2321 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2322 goto out; 2323 2324 /* dofilewritev() will unuse the descriptor for us */ 2325 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2326 &offset, 0, retval)); 2327 2328 out: 2329 FILE_UNUSE(fp, l); 2330 return (error); 2331 } 2332 2333 /* 2334 * Check access permissions. 2335 */ 2336 int 2337 sys_access(struct lwp *l, void *v, register_t *retval) 2338 { 2339 struct sys_access_args /* { 2340 syscallarg(const char *) path; 2341 syscallarg(int) flags; 2342 } */ *uap = v; 2343 struct proc *p = l->l_proc; 2344 kauth_cred_t cred; 2345 struct vnode *vp; 2346 int error, flags; 2347 struct nameidata nd; 2348 2349 cred = kauth_cred_dup(p->p_cred); 2350 kauth_cred_seteuid(cred, kauth_cred_getuid(p->p_cred)); 2351 kauth_cred_setegid(cred, kauth_cred_getgid(p->p_cred)); 2352 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2353 SCARG(uap, path), l); 2354 /* Override default credentials */ 2355 nd.ni_cnd.cn_cred = cred; 2356 if ((error = namei(&nd)) != 0) 2357 goto out; 2358 vp = nd.ni_vp; 2359 2360 /* Flags == 0 means only check for existence. */ 2361 if (SCARG(uap, flags)) { 2362 flags = 0; 2363 if (SCARG(uap, flags) & R_OK) 2364 flags |= VREAD; 2365 if (SCARG(uap, flags) & W_OK) 2366 flags |= VWRITE; 2367 if (SCARG(uap, flags) & X_OK) 2368 flags |= VEXEC; 2369 2370 error = VOP_ACCESS(vp, flags, cred, l); 2371 if (!error && (flags & VWRITE)) 2372 error = vn_writechk(vp); 2373 } 2374 vput(vp); 2375 out: 2376 kauth_cred_free(cred); 2377 return (error); 2378 } 2379 2380 /* 2381 * Get file status; this version follows links. 2382 */ 2383 /* ARGSUSED */ 2384 int 2385 sys___stat30(struct lwp *l, void *v, register_t *retval) 2386 { 2387 struct sys___stat30_args /* { 2388 syscallarg(const char *) path; 2389 syscallarg(struct stat *) ub; 2390 } */ *uap = v; 2391 struct stat sb; 2392 int error; 2393 struct nameidata nd; 2394 2395 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2396 SCARG(uap, path), l); 2397 if ((error = namei(&nd)) != 0) 2398 return (error); 2399 error = vn_stat(nd.ni_vp, &sb, l); 2400 vput(nd.ni_vp); 2401 if (error) 2402 return (error); 2403 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2404 return (error); 2405 } 2406 2407 /* 2408 * Get file status; this version does not follow links. 2409 */ 2410 /* ARGSUSED */ 2411 int 2412 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2413 { 2414 struct sys___lstat30_args /* { 2415 syscallarg(const char *) path; 2416 syscallarg(struct stat *) ub; 2417 } */ *uap = v; 2418 struct stat sb; 2419 int error; 2420 struct nameidata nd; 2421 2422 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2423 SCARG(uap, path), l); 2424 if ((error = namei(&nd)) != 0) 2425 return (error); 2426 error = vn_stat(nd.ni_vp, &sb, l); 2427 vput(nd.ni_vp); 2428 if (error) 2429 return (error); 2430 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2431 return (error); 2432 } 2433 2434 /* 2435 * Get configurable pathname variables. 2436 */ 2437 /* ARGSUSED */ 2438 int 2439 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2440 { 2441 struct sys_pathconf_args /* { 2442 syscallarg(const char *) path; 2443 syscallarg(int) name; 2444 } */ *uap = v; 2445 int error; 2446 struct nameidata nd; 2447 2448 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2449 SCARG(uap, path), l); 2450 if ((error = namei(&nd)) != 0) 2451 return (error); 2452 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2453 vput(nd.ni_vp); 2454 return (error); 2455 } 2456 2457 /* 2458 * Return target name of a symbolic link. 2459 */ 2460 /* ARGSUSED */ 2461 int 2462 sys_readlink(struct lwp *l, void *v, register_t *retval) 2463 { 2464 struct sys_readlink_args /* { 2465 syscallarg(const char *) path; 2466 syscallarg(char *) buf; 2467 syscallarg(size_t) count; 2468 } */ *uap = v; 2469 struct proc *p = l->l_proc; 2470 struct vnode *vp; 2471 struct iovec aiov; 2472 struct uio auio; 2473 int error; 2474 struct nameidata nd; 2475 2476 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2477 SCARG(uap, path), l); 2478 if ((error = namei(&nd)) != 0) 2479 return (error); 2480 vp = nd.ni_vp; 2481 if (vp->v_type != VLNK) 2482 error = EINVAL; 2483 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2484 (error = VOP_ACCESS(vp, VREAD, p->p_cred, l)) == 0) { 2485 aiov.iov_base = SCARG(uap, buf); 2486 aiov.iov_len = SCARG(uap, count); 2487 auio.uio_iov = &aiov; 2488 auio.uio_iovcnt = 1; 2489 auio.uio_offset = 0; 2490 auio.uio_rw = UIO_READ; 2491 KASSERT(l == curlwp); 2492 auio.uio_vmspace = l->l_proc->p_vmspace; 2493 auio.uio_resid = SCARG(uap, count); 2494 error = VOP_READLINK(vp, &auio, p->p_cred); 2495 } 2496 vput(vp); 2497 *retval = SCARG(uap, count) - auio.uio_resid; 2498 return (error); 2499 } 2500 2501 /* 2502 * Change flags of a file given a path name. 2503 */ 2504 /* ARGSUSED */ 2505 int 2506 sys_chflags(struct lwp *l, void *v, register_t *retval) 2507 { 2508 struct sys_chflags_args /* { 2509 syscallarg(const char *) path; 2510 syscallarg(u_long) flags; 2511 } */ *uap = v; 2512 struct vnode *vp; 2513 int error; 2514 struct nameidata nd; 2515 2516 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2517 if ((error = namei(&nd)) != 0) 2518 return (error); 2519 vp = nd.ni_vp; 2520 error = change_flags(vp, SCARG(uap, flags), l); 2521 vput(vp); 2522 return (error); 2523 } 2524 2525 /* 2526 * Change flags of a file given a file descriptor. 2527 */ 2528 /* ARGSUSED */ 2529 int 2530 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2531 { 2532 struct sys_fchflags_args /* { 2533 syscallarg(int) fd; 2534 syscallarg(u_long) flags; 2535 } */ *uap = v; 2536 struct proc *p = l->l_proc; 2537 struct vnode *vp; 2538 struct file *fp; 2539 int error; 2540 2541 /* getvnode() will use the descriptor for us */ 2542 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2543 return (error); 2544 vp = (struct vnode *)fp->f_data; 2545 error = change_flags(vp, SCARG(uap, flags), l); 2546 VOP_UNLOCK(vp, 0); 2547 FILE_UNUSE(fp, l); 2548 return (error); 2549 } 2550 2551 /* 2552 * Change flags of a file given a path name; this version does 2553 * not follow links. 2554 */ 2555 int 2556 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2557 { 2558 struct sys_lchflags_args /* { 2559 syscallarg(const char *) path; 2560 syscallarg(u_long) flags; 2561 } */ *uap = v; 2562 struct vnode *vp; 2563 int error; 2564 struct nameidata nd; 2565 2566 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2567 if ((error = namei(&nd)) != 0) 2568 return (error); 2569 vp = nd.ni_vp; 2570 error = change_flags(vp, SCARG(uap, flags), l); 2571 vput(vp); 2572 return (error); 2573 } 2574 2575 /* 2576 * Common routine to change flags of a file. 2577 */ 2578 int 2579 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2580 { 2581 struct proc *p = l->l_proc; 2582 struct mount *mp; 2583 struct vattr vattr; 2584 int error; 2585 2586 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2587 return (error); 2588 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2589 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2590 /* 2591 * Non-superusers cannot change the flags on devices, even if they 2592 * own them. 2593 */ 2594 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2595 &p->p_acflag) != 0) { 2596 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2597 goto out; 2598 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2599 error = EINVAL; 2600 goto out; 2601 } 2602 } 2603 VATTR_NULL(&vattr); 2604 vattr.va_flags = flags; 2605 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2606 out: 2607 vn_finished_write(mp, 0); 2608 return (error); 2609 } 2610 2611 /* 2612 * Change mode of a file given path name; this version follows links. 2613 */ 2614 /* ARGSUSED */ 2615 int 2616 sys_chmod(struct lwp *l, void *v, register_t *retval) 2617 { 2618 struct sys_chmod_args /* { 2619 syscallarg(const char *) path; 2620 syscallarg(int) mode; 2621 } */ *uap = v; 2622 int error; 2623 struct nameidata nd; 2624 2625 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2626 if ((error = namei(&nd)) != 0) 2627 return (error); 2628 2629 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2630 2631 vrele(nd.ni_vp); 2632 return (error); 2633 } 2634 2635 /* 2636 * Change mode of a file given a file descriptor. 2637 */ 2638 /* ARGSUSED */ 2639 int 2640 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2641 { 2642 struct sys_fchmod_args /* { 2643 syscallarg(int) fd; 2644 syscallarg(int) mode; 2645 } */ *uap = v; 2646 struct proc *p = l->l_proc; 2647 struct file *fp; 2648 int error; 2649 2650 /* getvnode() will use the descriptor for us */ 2651 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2652 return (error); 2653 2654 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2655 FILE_UNUSE(fp, l); 2656 return (error); 2657 } 2658 2659 /* 2660 * Change mode of a file given path name; this version does not follow links. 2661 */ 2662 /* ARGSUSED */ 2663 int 2664 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2665 { 2666 struct sys_lchmod_args /* { 2667 syscallarg(const char *) path; 2668 syscallarg(int) mode; 2669 } */ *uap = v; 2670 int error; 2671 struct nameidata nd; 2672 2673 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2674 if ((error = namei(&nd)) != 0) 2675 return (error); 2676 2677 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2678 2679 vrele(nd.ni_vp); 2680 return (error); 2681 } 2682 2683 /* 2684 * Common routine to set mode given a vnode. 2685 */ 2686 static int 2687 change_mode(struct vnode *vp, int mode, struct lwp *l) 2688 { 2689 struct proc *p = l->l_proc; 2690 struct mount *mp; 2691 struct vattr vattr; 2692 int error; 2693 2694 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2695 return (error); 2696 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2697 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2698 VATTR_NULL(&vattr); 2699 vattr.va_mode = mode & ALLPERMS; 2700 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2701 VOP_UNLOCK(vp, 0); 2702 vn_finished_write(mp, 0); 2703 return (error); 2704 } 2705 2706 /* 2707 * Set ownership given a path name; this version follows links. 2708 */ 2709 /* ARGSUSED */ 2710 int 2711 sys_chown(struct lwp *l, void *v, register_t *retval) 2712 { 2713 struct sys_chown_args /* { 2714 syscallarg(const char *) path; 2715 syscallarg(uid_t) uid; 2716 syscallarg(gid_t) gid; 2717 } */ *uap = v; 2718 int error; 2719 struct nameidata nd; 2720 2721 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2722 if ((error = namei(&nd)) != 0) 2723 return (error); 2724 2725 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2726 2727 vrele(nd.ni_vp); 2728 return (error); 2729 } 2730 2731 /* 2732 * Set ownership given a path name; this version follows links. 2733 * Provides POSIX semantics. 2734 */ 2735 /* ARGSUSED */ 2736 int 2737 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2738 { 2739 struct sys_chown_args /* { 2740 syscallarg(const char *) path; 2741 syscallarg(uid_t) uid; 2742 syscallarg(gid_t) gid; 2743 } */ *uap = v; 2744 int error; 2745 struct nameidata nd; 2746 2747 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2748 if ((error = namei(&nd)) != 0) 2749 return (error); 2750 2751 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2752 2753 vrele(nd.ni_vp); 2754 return (error); 2755 } 2756 2757 /* 2758 * Set ownership given a file descriptor. 2759 */ 2760 /* ARGSUSED */ 2761 int 2762 sys_fchown(struct lwp *l, void *v, register_t *retval) 2763 { 2764 struct sys_fchown_args /* { 2765 syscallarg(int) fd; 2766 syscallarg(uid_t) uid; 2767 syscallarg(gid_t) gid; 2768 } */ *uap = v; 2769 struct proc *p = l->l_proc; 2770 int error; 2771 struct file *fp; 2772 2773 /* getvnode() will use the descriptor for us */ 2774 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2775 return (error); 2776 2777 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2778 SCARG(uap, gid), l, 0); 2779 FILE_UNUSE(fp, l); 2780 return (error); 2781 } 2782 2783 /* 2784 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2785 */ 2786 /* ARGSUSED */ 2787 int 2788 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2789 { 2790 struct sys_fchown_args /* { 2791 syscallarg(int) fd; 2792 syscallarg(uid_t) uid; 2793 syscallarg(gid_t) gid; 2794 } */ *uap = v; 2795 struct proc *p = l->l_proc; 2796 int error; 2797 struct file *fp; 2798 2799 /* getvnode() will use the descriptor for us */ 2800 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2801 return (error); 2802 2803 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2804 SCARG(uap, gid), l, 1); 2805 FILE_UNUSE(fp, l); 2806 return (error); 2807 } 2808 2809 /* 2810 * Set ownership given a path name; this version does not follow links. 2811 */ 2812 /* ARGSUSED */ 2813 int 2814 sys_lchown(struct lwp *l, void *v, register_t *retval) 2815 { 2816 struct sys_lchown_args /* { 2817 syscallarg(const char *) path; 2818 syscallarg(uid_t) uid; 2819 syscallarg(gid_t) gid; 2820 } */ *uap = v; 2821 int error; 2822 struct nameidata nd; 2823 2824 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2825 if ((error = namei(&nd)) != 0) 2826 return (error); 2827 2828 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2829 2830 vrele(nd.ni_vp); 2831 return (error); 2832 } 2833 2834 /* 2835 * Set ownership given a path name; this version does not follow links. 2836 * Provides POSIX/XPG semantics. 2837 */ 2838 /* ARGSUSED */ 2839 int 2840 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2841 { 2842 struct sys_lchown_args /* { 2843 syscallarg(const char *) path; 2844 syscallarg(uid_t) uid; 2845 syscallarg(gid_t) gid; 2846 } */ *uap = v; 2847 int error; 2848 struct nameidata nd; 2849 2850 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2851 if ((error = namei(&nd)) != 0) 2852 return (error); 2853 2854 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2855 2856 vrele(nd.ni_vp); 2857 return (error); 2858 } 2859 2860 /* 2861 * Common routine to set ownership given a vnode. 2862 */ 2863 static int 2864 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2865 int posix_semantics) 2866 { 2867 struct proc *p = l->l_proc; 2868 struct mount *mp; 2869 struct vattr vattr; 2870 mode_t newmode; 2871 int error; 2872 2873 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2874 return (error); 2875 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2876 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2877 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2878 goto out; 2879 2880 #define CHANGED(x) ((int)(x) != -1) 2881 newmode = vattr.va_mode; 2882 if (posix_semantics) { 2883 /* 2884 * POSIX/XPG semantics: if the caller is not the super-user, 2885 * clear set-user-id and set-group-id bits. Both POSIX and 2886 * the XPG consider the behaviour for calls by the super-user 2887 * implementation-defined; we leave the set-user-id and set- 2888 * group-id settings intact in that case. 2889 */ 2890 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2891 NULL) != 0) 2892 newmode &= ~(S_ISUID | S_ISGID); 2893 } else { 2894 /* 2895 * NetBSD semantics: when changing owner and/or group, 2896 * clear the respective bit(s). 2897 */ 2898 if (CHANGED(uid)) 2899 newmode &= ~S_ISUID; 2900 if (CHANGED(gid)) 2901 newmode &= ~S_ISGID; 2902 } 2903 /* Update va_mode iff altered. */ 2904 if (vattr.va_mode == newmode) 2905 newmode = VNOVAL; 2906 2907 VATTR_NULL(&vattr); 2908 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2909 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2910 vattr.va_mode = newmode; 2911 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2912 #undef CHANGED 2913 2914 out: 2915 VOP_UNLOCK(vp, 0); 2916 vn_finished_write(mp, 0); 2917 return (error); 2918 } 2919 2920 /* 2921 * Set the access and modification times given a path name; this 2922 * version follows links. 2923 */ 2924 /* ARGSUSED */ 2925 int 2926 sys_utimes(struct lwp *l, void *v, register_t *retval) 2927 { 2928 struct sys_utimes_args /* { 2929 syscallarg(const char *) path; 2930 syscallarg(const struct timeval *) tptr; 2931 } */ *uap = v; 2932 int error; 2933 struct nameidata nd; 2934 2935 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2936 if ((error = namei(&nd)) != 0) 2937 return (error); 2938 2939 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2940 2941 vrele(nd.ni_vp); 2942 return (error); 2943 } 2944 2945 /* 2946 * Set the access and modification times given a file descriptor. 2947 */ 2948 /* ARGSUSED */ 2949 int 2950 sys_futimes(struct lwp *l, void *v, register_t *retval) 2951 { 2952 struct sys_futimes_args /* { 2953 syscallarg(int) fd; 2954 syscallarg(const struct timeval *) tptr; 2955 } */ *uap = v; 2956 struct proc *p = l->l_proc; 2957 int error; 2958 struct file *fp; 2959 2960 /* getvnode() will use the descriptor for us */ 2961 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2962 return (error); 2963 2964 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2965 FILE_UNUSE(fp, l); 2966 return (error); 2967 } 2968 2969 /* 2970 * Set the access and modification times given a path name; this 2971 * version does not follow links. 2972 */ 2973 /* ARGSUSED */ 2974 int 2975 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2976 { 2977 struct sys_lutimes_args /* { 2978 syscallarg(const char *) path; 2979 syscallarg(const struct timeval *) tptr; 2980 } */ *uap = v; 2981 int error; 2982 struct nameidata nd; 2983 2984 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2985 if ((error = namei(&nd)) != 0) 2986 return (error); 2987 2988 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2989 2990 vrele(nd.ni_vp); 2991 return (error); 2992 } 2993 2994 /* 2995 * Common routine to set access and modification times given a vnode. 2996 */ 2997 static int 2998 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2999 { 3000 struct proc *p = l->l_proc; 3001 struct mount *mp; 3002 struct vattr vattr; 3003 int error; 3004 3005 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3006 return (error); 3007 VATTR_NULL(&vattr); 3008 if (tptr == NULL) { 3009 nanotime(&vattr.va_atime); 3010 vattr.va_mtime = vattr.va_atime; 3011 vattr.va_vaflags |= VA_UTIMES_NULL; 3012 } else { 3013 struct timeval tv[2]; 3014 3015 error = copyin(tptr, tv, sizeof(tv)); 3016 if (error) 3017 goto out; 3018 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3019 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3020 } 3021 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3022 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3023 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 3024 VOP_UNLOCK(vp, 0); 3025 out: 3026 vn_finished_write(mp, 0); 3027 return (error); 3028 } 3029 3030 /* 3031 * Truncate a file given its path name. 3032 */ 3033 /* ARGSUSED */ 3034 int 3035 sys_truncate(struct lwp *l, void *v, register_t *retval) 3036 { 3037 struct sys_truncate_args /* { 3038 syscallarg(const char *) path; 3039 syscallarg(int) pad; 3040 syscallarg(off_t) length; 3041 } */ *uap = v; 3042 struct proc *p = l->l_proc; 3043 struct vnode *vp; 3044 struct mount *mp; 3045 struct vattr vattr; 3046 int error; 3047 struct nameidata nd; 3048 3049 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3050 if ((error = namei(&nd)) != 0) 3051 return (error); 3052 vp = nd.ni_vp; 3053 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3054 vrele(vp); 3055 return (error); 3056 } 3057 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3058 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3059 if (vp->v_type == VDIR) 3060 error = EISDIR; 3061 else if ((error = vn_writechk(vp)) == 0 && 3062 (error = VOP_ACCESS(vp, VWRITE, p->p_cred, l)) == 0) { 3063 VATTR_NULL(&vattr); 3064 vattr.va_size = SCARG(uap, length); 3065 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 3066 } 3067 vput(vp); 3068 vn_finished_write(mp, 0); 3069 return (error); 3070 } 3071 3072 /* 3073 * Truncate a file given a file descriptor. 3074 */ 3075 /* ARGSUSED */ 3076 int 3077 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3078 { 3079 struct sys_ftruncate_args /* { 3080 syscallarg(int) fd; 3081 syscallarg(int) pad; 3082 syscallarg(off_t) length; 3083 } */ *uap = v; 3084 struct proc *p = l->l_proc; 3085 struct mount *mp; 3086 struct vattr vattr; 3087 struct vnode *vp; 3088 struct file *fp; 3089 int error; 3090 3091 /* getvnode() will use the descriptor for us */ 3092 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3093 return (error); 3094 if ((fp->f_flag & FWRITE) == 0) { 3095 error = EINVAL; 3096 goto out; 3097 } 3098 vp = (struct vnode *)fp->f_data; 3099 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3100 FILE_UNUSE(fp, l); 3101 return (error); 3102 } 3103 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3104 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3105 if (vp->v_type == VDIR) 3106 error = EISDIR; 3107 else if ((error = vn_writechk(vp)) == 0) { 3108 VATTR_NULL(&vattr); 3109 vattr.va_size = SCARG(uap, length); 3110 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3111 } 3112 VOP_UNLOCK(vp, 0); 3113 vn_finished_write(mp, 0); 3114 out: 3115 FILE_UNUSE(fp, l); 3116 return (error); 3117 } 3118 3119 /* 3120 * Sync an open file. 3121 */ 3122 /* ARGSUSED */ 3123 int 3124 sys_fsync(struct lwp *l, void *v, register_t *retval) 3125 { 3126 struct sys_fsync_args /* { 3127 syscallarg(int) fd; 3128 } */ *uap = v; 3129 struct proc *p = l->l_proc; 3130 struct vnode *vp; 3131 struct mount *mp; 3132 struct file *fp; 3133 int error; 3134 3135 /* getvnode() will use the descriptor for us */ 3136 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3137 return (error); 3138 vp = (struct vnode *)fp->f_data; 3139 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3140 FILE_UNUSE(fp, l); 3141 return (error); 3142 } 3143 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3144 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3145 if (error == 0 && bioops.io_fsync != NULL && 3146 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3147 (*bioops.io_fsync)(vp, 0); 3148 VOP_UNLOCK(vp, 0); 3149 vn_finished_write(mp, 0); 3150 FILE_UNUSE(fp, l); 3151 return (error); 3152 } 3153 3154 /* 3155 * Sync a range of file data. API modeled after that found in AIX. 3156 * 3157 * FDATASYNC indicates that we need only save enough metadata to be able 3158 * to re-read the written data. Note we duplicate AIX's requirement that 3159 * the file be open for writing. 3160 */ 3161 /* ARGSUSED */ 3162 int 3163 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3164 { 3165 struct sys_fsync_range_args /* { 3166 syscallarg(int) fd; 3167 syscallarg(int) flags; 3168 syscallarg(off_t) start; 3169 syscallarg(off_t) length; 3170 } */ *uap = v; 3171 struct proc *p = l->l_proc; 3172 struct vnode *vp; 3173 struct file *fp; 3174 int flags, nflags; 3175 off_t s, e, len; 3176 int error; 3177 3178 /* getvnode() will use the descriptor for us */ 3179 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3180 return (error); 3181 3182 if ((fp->f_flag & FWRITE) == 0) { 3183 FILE_UNUSE(fp, l); 3184 return (EBADF); 3185 } 3186 3187 flags = SCARG(uap, flags); 3188 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3189 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3190 return (EINVAL); 3191 } 3192 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3193 if (flags & FDATASYNC) 3194 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3195 else 3196 nflags = FSYNC_WAIT; 3197 if (flags & FDISKSYNC) 3198 nflags |= FSYNC_CACHE; 3199 3200 len = SCARG(uap, length); 3201 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3202 if (len) { 3203 s = SCARG(uap, start); 3204 e = s + len; 3205 if (e < s) { 3206 FILE_UNUSE(fp, l); 3207 return (EINVAL); 3208 } 3209 } else { 3210 e = 0; 3211 s = 0; 3212 } 3213 3214 vp = (struct vnode *)fp->f_data; 3215 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3216 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3217 3218 if (error == 0 && bioops.io_fsync != NULL && 3219 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3220 (*bioops.io_fsync)(vp, nflags); 3221 3222 VOP_UNLOCK(vp, 0); 3223 FILE_UNUSE(fp, l); 3224 return (error); 3225 } 3226 3227 /* 3228 * Sync the data of an open file. 3229 */ 3230 /* ARGSUSED */ 3231 int 3232 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3233 { 3234 struct sys_fdatasync_args /* { 3235 syscallarg(int) fd; 3236 } */ *uap = v; 3237 struct proc *p = l->l_proc; 3238 struct vnode *vp; 3239 struct file *fp; 3240 int error; 3241 3242 /* getvnode() will use the descriptor for us */ 3243 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3244 return (error); 3245 if ((fp->f_flag & FWRITE) == 0) { 3246 FILE_UNUSE(fp, l); 3247 return (EBADF); 3248 } 3249 vp = (struct vnode *)fp->f_data; 3250 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3251 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3252 VOP_UNLOCK(vp, 0); 3253 FILE_UNUSE(fp, l); 3254 return (error); 3255 } 3256 3257 /* 3258 * Rename files, (standard) BSD semantics frontend. 3259 */ 3260 /* ARGSUSED */ 3261 int 3262 sys_rename(struct lwp *l, void *v, register_t *retval) 3263 { 3264 struct sys_rename_args /* { 3265 syscallarg(const char *) from; 3266 syscallarg(const char *) to; 3267 } */ *uap = v; 3268 3269 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3270 } 3271 3272 /* 3273 * Rename files, POSIX semantics frontend. 3274 */ 3275 /* ARGSUSED */ 3276 int 3277 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3278 { 3279 struct sys___posix_rename_args /* { 3280 syscallarg(const char *) from; 3281 syscallarg(const char *) to; 3282 } */ *uap = v; 3283 3284 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3285 } 3286 3287 /* 3288 * Rename files. Source and destination must either both be directories, 3289 * or both not be directories. If target is a directory, it must be empty. 3290 * If `from' and `to' refer to the same object, the value of the `retain' 3291 * argument is used to determine whether `from' will be 3292 * 3293 * (retain == 0) deleted unless `from' and `to' refer to the same 3294 * object in the file system's name space (BSD). 3295 * (retain == 1) always retained (POSIX). 3296 */ 3297 static int 3298 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3299 { 3300 struct mount *mp = NULL; 3301 struct vnode *tvp, *fvp, *tdvp; 3302 struct nameidata fromnd, tond; 3303 struct proc *p; 3304 int error; 3305 3306 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3307 from, l); 3308 if ((error = namei(&fromnd)) != 0) 3309 return (error); 3310 fvp = fromnd.ni_vp; 3311 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3312 if (error != 0) { 3313 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3314 vrele(fromnd.ni_dvp); 3315 vrele(fvp); 3316 if (fromnd.ni_startdir) 3317 vrele(fromnd.ni_startdir); 3318 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3319 return (error); 3320 } 3321 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3322 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3323 if ((error = namei(&tond)) != 0) { 3324 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3325 vrele(fromnd.ni_dvp); 3326 vrele(fvp); 3327 goto out1; 3328 } 3329 tdvp = tond.ni_dvp; 3330 tvp = tond.ni_vp; 3331 3332 if (tvp != NULL) { 3333 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3334 error = ENOTDIR; 3335 goto out; 3336 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3337 error = EISDIR; 3338 goto out; 3339 } 3340 } 3341 3342 if (fvp == tdvp) 3343 error = EINVAL; 3344 3345 /* 3346 * Source and destination refer to the same object. 3347 */ 3348 if (fvp == tvp) { 3349 if (retain) 3350 error = -1; 3351 else if (fromnd.ni_dvp == tdvp && 3352 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3353 !memcmp(fromnd.ni_cnd.cn_nameptr, 3354 tond.ni_cnd.cn_nameptr, 3355 fromnd.ni_cnd.cn_namelen)) 3356 error = -1; 3357 } 3358 3359 #ifdef VERIFIED_EXEC 3360 if (!error) 3361 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l); 3362 #endif /* VERIFIED_EXEC */ 3363 3364 out: 3365 p = l->l_proc; 3366 if (!error) { 3367 VOP_LEASE(tdvp, l, p->p_cred, LEASE_WRITE); 3368 if (fromnd.ni_dvp != tdvp) 3369 VOP_LEASE(fromnd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3370 if (tvp) { 3371 VOP_LEASE(tvp, l, p->p_cred, LEASE_WRITE); 3372 } 3373 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3374 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3375 } else { 3376 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3377 if (tdvp == tvp) 3378 vrele(tdvp); 3379 else 3380 vput(tdvp); 3381 if (tvp) 3382 vput(tvp); 3383 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3384 vrele(fromnd.ni_dvp); 3385 vrele(fvp); 3386 } 3387 vrele(tond.ni_startdir); 3388 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3389 out1: 3390 vn_finished_write(mp, 0); 3391 if (fromnd.ni_startdir) 3392 vrele(fromnd.ni_startdir); 3393 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3394 return (error == -1 ? 0 : error); 3395 } 3396 3397 /* 3398 * Make a directory file. 3399 */ 3400 /* ARGSUSED */ 3401 int 3402 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3403 { 3404 struct sys_mkdir_args /* { 3405 syscallarg(const char *) path; 3406 syscallarg(int) mode; 3407 } */ *uap = v; 3408 struct proc *p = l->l_proc; 3409 struct mount *mp; 3410 struct vnode *vp; 3411 struct vattr vattr; 3412 int error; 3413 struct nameidata nd; 3414 3415 restart: 3416 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3417 SCARG(uap, path), l); 3418 if ((error = namei(&nd)) != 0) 3419 return (error); 3420 vp = nd.ni_vp; 3421 if (vp != NULL) { 3422 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3423 if (nd.ni_dvp == vp) 3424 vrele(nd.ni_dvp); 3425 else 3426 vput(nd.ni_dvp); 3427 vrele(vp); 3428 return (EEXIST); 3429 } 3430 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3431 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3432 if (nd.ni_dvp == vp) 3433 vrele(nd.ni_dvp); 3434 else 3435 vput(nd.ni_dvp); 3436 if ((error = vn_start_write(NULL, &mp, 3437 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3438 return (error); 3439 goto restart; 3440 } 3441 VATTR_NULL(&vattr); 3442 vattr.va_type = VDIR; 3443 vattr.va_mode = 3444 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3445 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3446 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3447 if (!error) 3448 vput(nd.ni_vp); 3449 vn_finished_write(mp, 0); 3450 return (error); 3451 } 3452 3453 /* 3454 * Remove a directory file. 3455 */ 3456 /* ARGSUSED */ 3457 int 3458 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3459 { 3460 struct sys_rmdir_args /* { 3461 syscallarg(const char *) path; 3462 } */ *uap = v; 3463 struct proc *p = l->l_proc; 3464 struct mount *mp; 3465 struct vnode *vp; 3466 int error; 3467 struct nameidata nd; 3468 3469 restart: 3470 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3471 SCARG(uap, path), l); 3472 if ((error = namei(&nd)) != 0) 3473 return (error); 3474 vp = nd.ni_vp; 3475 if (vp->v_type != VDIR) { 3476 error = ENOTDIR; 3477 goto out; 3478 } 3479 /* 3480 * No rmdir "." please. 3481 */ 3482 if (nd.ni_dvp == vp) { 3483 error = EINVAL; 3484 goto out; 3485 } 3486 /* 3487 * The root of a mounted filesystem cannot be deleted. 3488 */ 3489 if (vp->v_flag & VROOT) { 3490 error = EBUSY; 3491 goto out; 3492 } 3493 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3494 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3495 if (nd.ni_dvp == vp) 3496 vrele(nd.ni_dvp); 3497 else 3498 vput(nd.ni_dvp); 3499 vput(vp); 3500 if ((error = vn_start_write(NULL, &mp, 3501 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3502 return (error); 3503 goto restart; 3504 } 3505 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3506 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3507 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3508 vn_finished_write(mp, 0); 3509 return (error); 3510 3511 out: 3512 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3513 if (nd.ni_dvp == vp) 3514 vrele(nd.ni_dvp); 3515 else 3516 vput(nd.ni_dvp); 3517 vput(vp); 3518 return (error); 3519 } 3520 3521 /* 3522 * Read a block of directory entries in a file system independent format. 3523 */ 3524 int 3525 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3526 { 3527 struct sys___getdents30_args /* { 3528 syscallarg(int) fd; 3529 syscallarg(char *) buf; 3530 syscallarg(size_t) count; 3531 } */ *uap = v; 3532 struct proc *p = l->l_proc; 3533 struct file *fp; 3534 int error, done; 3535 3536 /* getvnode() will use the descriptor for us */ 3537 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3538 return (error); 3539 if ((fp->f_flag & FREAD) == 0) { 3540 error = EBADF; 3541 goto out; 3542 } 3543 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3544 SCARG(uap, count), &done, l, 0, 0); 3545 #ifdef KTRACE 3546 if (!error && KTRPOINT(p, KTR_GENIO)) { 3547 struct iovec iov; 3548 iov.iov_base = SCARG(uap, buf); 3549 iov.iov_len = done; 3550 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3551 } 3552 #endif 3553 *retval = done; 3554 out: 3555 FILE_UNUSE(fp, l); 3556 return (error); 3557 } 3558 3559 /* 3560 * Set the mode mask for creation of filesystem nodes. 3561 */ 3562 int 3563 sys_umask(struct lwp *l, void *v, register_t *retval) 3564 { 3565 struct sys_umask_args /* { 3566 syscallarg(mode_t) newmask; 3567 } */ *uap = v; 3568 struct proc *p = l->l_proc; 3569 struct cwdinfo *cwdi; 3570 3571 cwdi = p->p_cwdi; 3572 *retval = cwdi->cwdi_cmask; 3573 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3574 return (0); 3575 } 3576 3577 /* 3578 * Void all references to file by ripping underlying filesystem 3579 * away from vnode. 3580 */ 3581 /* ARGSUSED */ 3582 int 3583 sys_revoke(struct lwp *l, void *v, register_t *retval) 3584 { 3585 struct sys_revoke_args /* { 3586 syscallarg(const char *) path; 3587 } */ *uap = v; 3588 struct proc *p = l->l_proc; 3589 struct mount *mp; 3590 struct vnode *vp; 3591 struct vattr vattr; 3592 int error; 3593 struct nameidata nd; 3594 3595 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3596 if ((error = namei(&nd)) != 0) 3597 return (error); 3598 vp = nd.ni_vp; 3599 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 3600 goto out; 3601 if (kauth_cred_geteuid(p->p_cred) != vattr.va_uid && 3602 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 3603 &p->p_acflag)) != 0) 3604 goto out; 3605 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3606 goto out; 3607 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3608 VOP_REVOKE(vp, REVOKEALL); 3609 vn_finished_write(mp, 0); 3610 out: 3611 vrele(vp); 3612 return (error); 3613 } 3614 3615 /* 3616 * Convert a user file descriptor to a kernel file entry. 3617 */ 3618 int 3619 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3620 { 3621 struct vnode *vp; 3622 struct file *fp; 3623 3624 if ((fp = fd_getfile(fdp, fd)) == NULL) 3625 return (EBADF); 3626 3627 FILE_USE(fp); 3628 3629 if (fp->f_type != DTYPE_VNODE) { 3630 FILE_UNUSE(fp, NULL); 3631 return (EINVAL); 3632 } 3633 3634 vp = (struct vnode *)fp->f_data; 3635 if (vp->v_type == VBAD) { 3636 FILE_UNUSE(fp, NULL); 3637 return (EBADF); 3638 } 3639 3640 *fpp = fp; 3641 return (0); 3642 } 3643
Indexes created Thu Sep 25 02:09:54 GMT 2025