1 /* $NetBSD: vfs_syscalls.c,v 1.230 2005/08/30 09:37:41 jmmv 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 1.230 2005/08/30 09:37:41 jmmv 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 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/dirent.h> 61 #include <sys/sysctl.h> 62 #include <sys/sa.h> 63 #include <sys/syscallargs.h> 64 #ifdef KTRACE 65 #include <sys/ktrace.h> 66 #endif 67 #ifdef VERIFIED_EXEC 68 #include <sys/verified_exec.h> 69 #endif /* VERIFIED_EXEC */ 70 71 #include <miscfs/genfs/genfs.h> 72 #include <miscfs/syncfs/syncfs.h> 73 74 #if NFSS > 0 75 #include <dev/fssvar.h> 76 #endif 77 78 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 79 80 static int change_dir(struct nameidata *, struct proc *); 81 static int change_flags(struct vnode *, u_long, struct proc *); 82 static int change_mode(struct vnode *, int, struct proc *p); 83 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int); 84 static int change_utimes(struct vnode *vp, const struct timeval *, 85 struct proc *p); 86 static int rename_files(const char *, const char *, struct proc *, int); 87 88 void checkdirs(struct vnode *); 89 90 int dovfsusermount = 0; 91 92 /* 93 * Virtual File System System Calls 94 */ 95 96 /* 97 * Mount a file system. 98 */ 99 100 #if defined(COMPAT_09) || defined(COMPAT_43) 101 /* 102 * This table is used to maintain compatibility with 4.3BSD 103 * and NetBSD 0.9 mount syscalls. Note, the order is important! 104 * 105 * Do not modify this table. It should only contain filesystems 106 * supported by NetBSD 0.9 and 4.3BSD. 107 */ 108 const char * const mountcompatnames[] = { 109 NULL, /* 0 = MOUNT_NONE */ 110 MOUNT_FFS, /* 1 = MOUNT_UFS */ 111 MOUNT_NFS, /* 2 */ 112 MOUNT_MFS, /* 3 */ 113 MOUNT_MSDOS, /* 4 */ 114 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 115 MOUNT_FDESC, /* 6 */ 116 MOUNT_KERNFS, /* 7 */ 117 NULL, /* 8 = MOUNT_DEVFS */ 118 MOUNT_AFS, /* 9 */ 119 }; 120 const int nmountcompatnames = sizeof(mountcompatnames) / 121 sizeof(mountcompatnames[0]); 122 #endif /* COMPAT_09 || COMPAT_43 */ 123 124 /* ARGSUSED */ 125 int 126 sys_mount(struct lwp *l, void *v, register_t *retval) 127 { 128 struct sys_mount_args /* { 129 syscallarg(const char *) type; 130 syscallarg(const char *) path; 131 syscallarg(int) flags; 132 syscallarg(void *) data; 133 } */ *uap = v; 134 struct proc *p = l->l_proc; 135 struct vnode *vp; 136 struct mount *mp; 137 int error, flag = 0; 138 char fstypename[MFSNAMELEN]; 139 struct vattr va; 140 struct nameidata nd; 141 struct vfsops *vfs; 142 143 /* 144 * if MNT_GETARGS is specified, it should be only flag. 145 */ 146 147 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 148 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 149 return EINVAL; 150 } 151 152 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 153 (error = suser(p->p_ucred, &p->p_acflag))) 154 return (error); 155 /* 156 * Get vnode to be covered 157 */ 158 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 159 SCARG(uap, path), p); 160 if ((error = namei(&nd)) != 0) 161 return (error); 162 vp = nd.ni_vp; 163 /* 164 * A lookup in VFS_MOUNT might result in an attempt to 165 * lock this vnode again, so make the lock recursive. 166 */ 167 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 168 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 169 if ((vp->v_flag & VROOT) == 0) { 170 vput(vp); 171 return (EINVAL); 172 } 173 mp = vp->v_mount; 174 flag = mp->mnt_flag; 175 vfs = mp->mnt_op; 176 /* 177 * We only allow the filesystem to be reloaded if it 178 * is currently mounted read-only. 179 */ 180 if ((SCARG(uap, flags) & MNT_RELOAD) && 181 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 182 vput(vp); 183 return (EOPNOTSUPP); /* Needs translation */ 184 } 185 /* 186 * In "highly secure" mode, don't let the caller do anything 187 * but downgrade a filesystem from read-write to read-only. 188 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 189 */ 190 if (securelevel >= 2 && 191 SCARG(uap, flags) != MNT_GETARGS && 192 SCARG(uap, flags) != 193 (mp->mnt_flag | MNT_RDONLY | 194 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 195 vput(vp); 196 return (EPERM); 197 } 198 mp->mnt_flag |= SCARG(uap, flags) & 199 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 200 /* 201 * Only root, or the user that did the original mount is 202 * permitted to update it. 203 */ 204 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 205 mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 206 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 207 vput(vp); 208 return (error); 209 } 210 /* 211 * Do not allow NFS export by non-root users. For non-root 212 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 213 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 214 */ 215 if (p->p_ucred->cr_uid != 0) { 216 if (SCARG(uap, flags) & MNT_EXPORTED) { 217 vput(vp); 218 return (EPERM); 219 } 220 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 221 if (flag & MNT_NOEXEC) 222 SCARG(uap, flags) |= MNT_NOEXEC; 223 } 224 if (vfs_busy(mp, LK_NOWAIT, 0)) { 225 vput(vp); 226 return (EPERM); 227 } 228 goto update; 229 } else { 230 if (securelevel >= 2) { 231 vput(vp); 232 return (EPERM); 233 } 234 } 235 /* 236 * If the user is not root, ensure that they own the directory 237 * onto which we are attempting to mount. 238 */ 239 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 || 240 (va.va_uid != p->p_ucred->cr_uid && 241 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) { 242 vput(vp); 243 return (error); 244 } 245 /* 246 * Do not allow NFS export by non-root users. For non-root users, 247 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 248 * mount point is already MNT_NOEXEC. 249 */ 250 if (p->p_ucred->cr_uid != 0) { 251 if (SCARG(uap, flags) & MNT_EXPORTED) { 252 vput(vp); 253 return (EPERM); 254 } 255 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 256 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 257 SCARG(uap, flags) |= MNT_NOEXEC; 258 } 259 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) { 260 vput(vp); 261 return (error); 262 } 263 if (vp->v_type != VDIR) { 264 vput(vp); 265 return (ENOTDIR); 266 } 267 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 268 if (error) { 269 #if defined(COMPAT_09) || defined(COMPAT_43) 270 /* 271 * Historically, filesystem types were identified by numbers. 272 * If we get an integer for the filesystem type instead of a 273 * string, we check to see if it matches one of the historic 274 * filesystem types. 275 */ 276 u_long fsindex = (u_long)SCARG(uap, type); 277 if (fsindex >= nmountcompatnames || 278 mountcompatnames[fsindex] == NULL) { 279 vput(vp); 280 return (ENODEV); 281 } 282 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 283 #else 284 vput(vp); 285 return (error); 286 #endif 287 } 288 #ifdef COMPAT_10 289 /* Accept `ufs' as an alias for `ffs'. */ 290 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 291 strncpy(fstypename, "ffs", MFSNAMELEN); 292 #endif 293 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 294 vput(vp); 295 return (ENODEV); 296 } 297 if (vp->v_mountedhere != NULL) { 298 vput(vp); 299 return (EBUSY); 300 } 301 302 /* 303 * Allocate and initialize the file system. 304 */ 305 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 306 M_MOUNT, M_WAITOK); 307 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 308 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 309 simple_lock_init(&mp->mnt_slock); 310 (void)vfs_busy(mp, LK_NOWAIT, 0); 311 mp->mnt_op = vfs; 312 vfs->vfs_refcount++; 313 mp->mnt_vnodecovered = vp; 314 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 315 mp->mnt_unmounter = NULL; 316 mp->mnt_leaf = mp; 317 318 /* 319 * The underlying file system may refuse the mount for 320 * various reasons. Allow the user to force it to happen. 321 */ 322 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 323 update: 324 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 325 /* 326 * Set the mount level flags. 327 */ 328 if (SCARG(uap, flags) & MNT_RDONLY) 329 mp->mnt_flag |= MNT_RDONLY; 330 else if (mp->mnt_flag & MNT_RDONLY) 331 mp->mnt_iflag |= IMNT_WANTRDWR; 332 mp->mnt_flag &= 333 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 334 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 335 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 336 MNT_MAGICLINKS); 337 mp->mnt_flag |= SCARG(uap, flags) & 338 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 339 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 340 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 341 MNT_IGNORE | MNT_MAGICLINKS); 342 } 343 /* 344 * Mount the filesystem. 345 */ 346 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 347 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 348 if (mp->mnt_iflag & IMNT_WANTRDWR) 349 mp->mnt_flag &= ~MNT_RDONLY; 350 if (error) 351 mp->mnt_flag = flag; 352 mp->mnt_flag &=~ 353 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 354 mp->mnt_iflag &=~ IMNT_WANTRDWR; 355 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 356 if (mp->mnt_syncer == NULL) 357 error = vfs_allocate_syncvnode(mp); 358 } else { 359 if (mp->mnt_syncer != NULL) 360 vfs_deallocate_syncvnode(mp); 361 } 362 vfs_unbusy(mp); 363 VOP_UNLOCK(vp, 0); 364 vrele(vp); 365 return (error); 366 } 367 /* 368 * Put the new filesystem on the mount list after root. 369 */ 370 cache_purge(vp); 371 if (!error) { 372 mp->mnt_flag &=~ 373 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 374 mp->mnt_iflag &=~ IMNT_WANTRDWR; 375 vp->v_mountedhere = mp; 376 simple_lock(&mountlist_slock); 377 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 378 simple_unlock(&mountlist_slock); 379 checkdirs(vp); 380 VOP_UNLOCK(vp, 0); 381 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 382 error = vfs_allocate_syncvnode(mp); 383 vfs_unbusy(mp); 384 (void) VFS_STATVFS(mp, &mp->mnt_stat, p); 385 if ((error = VFS_START(mp, 0, p))) 386 vrele(vp); 387 } else { 388 vp->v_mountedhere = (struct mount *)0; 389 vfs->vfs_refcount--; 390 vfs_unbusy(mp); 391 free(mp, M_MOUNT); 392 vput(vp); 393 } 394 return (error); 395 } 396 397 /* 398 * Scan all active processes to see if any of them have a current 399 * or root directory onto which the new filesystem has just been 400 * mounted. If so, replace them with the new mount point. 401 */ 402 void 403 checkdirs(struct vnode *olddp) 404 { 405 struct cwdinfo *cwdi; 406 struct vnode *newdp; 407 struct proc *p; 408 409 if (olddp->v_usecount == 1) 410 return; 411 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 412 panic("mount: lost mount"); 413 proclist_lock_read(); 414 PROCLIST_FOREACH(p, &allproc) { 415 cwdi = p->p_cwdi; 416 if (!cwdi) 417 continue; 418 if (cwdi->cwdi_cdir == olddp) { 419 vrele(cwdi->cwdi_cdir); 420 VREF(newdp); 421 cwdi->cwdi_cdir = newdp; 422 } 423 if (cwdi->cwdi_rdir == olddp) { 424 vrele(cwdi->cwdi_rdir); 425 VREF(newdp); 426 cwdi->cwdi_rdir = newdp; 427 } 428 } 429 proclist_unlock_read(); 430 if (rootvnode == olddp) { 431 vrele(rootvnode); 432 VREF(newdp); 433 rootvnode = newdp; 434 } 435 vput(newdp); 436 } 437 438 /* 439 * Unmount a file system. 440 * 441 * Note: unmount takes a path to the vnode mounted on as argument, 442 * not special file (as before). 443 */ 444 /* ARGSUSED */ 445 int 446 sys_unmount(struct lwp *l, void *v, register_t *retval) 447 { 448 struct sys_unmount_args /* { 449 syscallarg(const char *) path; 450 syscallarg(int) flags; 451 } */ *uap = v; 452 struct proc *p = l->l_proc; 453 struct vnode *vp; 454 struct mount *mp; 455 int error; 456 struct nameidata nd; 457 458 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 459 SCARG(uap, path), p); 460 if ((error = namei(&nd)) != 0) 461 return (error); 462 vp = nd.ni_vp; 463 mp = vp->v_mount; 464 465 /* 466 * Only root, or the user that did the original mount is 467 * permitted to unmount this filesystem. 468 */ 469 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 470 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 471 vput(vp); 472 return (error); 473 } 474 475 /* 476 * Don't allow unmounting the root file system. 477 */ 478 if (mp->mnt_flag & MNT_ROOTFS) { 479 vput(vp); 480 return (EINVAL); 481 } 482 483 /* 484 * Must be the root of the filesystem 485 */ 486 if ((vp->v_flag & VROOT) == 0) { 487 vput(vp); 488 return (EINVAL); 489 } 490 vput(vp); 491 492 /* 493 * XXX Freeze syncer. Must do this before locking the 494 * mount point. See dounmount() for details. 495 */ 496 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 497 498 if (vfs_busy(mp, 0, 0)) { 499 lockmgr(&syncer_lock, LK_RELEASE, NULL); 500 return (EBUSY); 501 } 502 503 return (dounmount(mp, SCARG(uap, flags), p)); 504 } 505 506 /* 507 * Do the actual file system unmount. File system is assumed to have been 508 * marked busy by the caller. 509 */ 510 int 511 dounmount(struct mount *mp, int flags, struct proc *p) 512 { 513 struct vnode *coveredvp; 514 int error; 515 int async; 516 int used_syncer; 517 518 simple_lock(&mountlist_slock); 519 vfs_unbusy(mp); 520 used_syncer = (mp->mnt_syncer != NULL); 521 522 /* 523 * XXX Syncer must be frozen when we get here. This should really 524 * be done on a per-mountpoint basis, but especially the softdep 525 * code possibly called from the syncer doesn't exactly work on a 526 * per-mountpoint basis, so the softdep code would become a maze 527 * of vfs_busy() calls. 528 * 529 * The caller of dounmount() must acquire syncer_lock because 530 * the syncer itself acquires locks in syncer_lock -> vfs_busy 531 * order, and we must preserve that order to avoid deadlock. 532 * 533 * So, if the file system did not use the syncer, now is 534 * the time to release the syncer_lock. 535 */ 536 if (used_syncer == 0) 537 lockmgr(&syncer_lock, LK_RELEASE, NULL); 538 539 mp->mnt_iflag |= IMNT_UNMOUNT; 540 mp->mnt_unmounter = p; 541 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 542 vn_start_write(NULL, &mp, V_WAIT); 543 544 if (mp->mnt_flag & MNT_EXPUBLIC) 545 vfs_setpublicfs(NULL, NULL, NULL); 546 async = mp->mnt_flag & MNT_ASYNC; 547 mp->mnt_flag &= ~MNT_ASYNC; 548 cache_purgevfs(mp); /* remove cache entries for this file sys */ 549 if (mp->mnt_syncer != NULL) 550 vfs_deallocate_syncvnode(mp); 551 error = 0; 552 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 553 #if NFSS > 0 554 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 555 #endif 556 if (error == 0) 557 error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p); 558 } 559 if (error == 0 || (flags & MNT_FORCE)) 560 error = VFS_UNMOUNT(mp, flags, p); 561 vn_finished_write(mp, 0); 562 simple_lock(&mountlist_slock); 563 if (error) { 564 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 565 (void) vfs_allocate_syncvnode(mp); 566 mp->mnt_iflag &= ~IMNT_UNMOUNT; 567 mp->mnt_unmounter = NULL; 568 mp->mnt_flag |= async; 569 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 570 &mountlist_slock); 571 if (used_syncer) 572 lockmgr(&syncer_lock, LK_RELEASE, NULL); 573 simple_lock(&mp->mnt_slock); 574 while (mp->mnt_wcnt > 0) { 575 wakeup(mp); 576 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 577 0, &mp->mnt_slock); 578 } 579 simple_unlock(&mp->mnt_slock); 580 return (error); 581 } 582 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 583 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 584 coveredvp->v_mountedhere = NULL; 585 vrele(coveredvp); 586 } 587 mp->mnt_op->vfs_refcount--; 588 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 589 panic("unmount: dangling vnode"); 590 mp->mnt_iflag |= IMNT_GONE; 591 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 592 if (used_syncer) 593 lockmgr(&syncer_lock, LK_RELEASE, NULL); 594 simple_lock(&mp->mnt_slock); 595 while (mp->mnt_wcnt > 0) { 596 wakeup(mp); 597 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 598 } 599 simple_unlock(&mp->mnt_slock); 600 free(mp, M_MOUNT); 601 return (0); 602 } 603 604 /* 605 * Sync each mounted filesystem. 606 */ 607 #ifdef DEBUG 608 int syncprt = 0; 609 struct ctldebug debug0 = { "syncprt", &syncprt }; 610 #endif 611 612 /* ARGSUSED */ 613 int 614 sys_sync(struct lwp *l, void *v, register_t *retval) 615 { 616 struct mount *mp, *nmp; 617 int asyncflag; 618 struct proc *p = l == NULL ? &proc0 : l->l_proc; 619 620 simple_lock(&mountlist_slock); 621 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 622 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 623 nmp = mp->mnt_list.cqe_prev; 624 continue; 625 } 626 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 627 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 628 asyncflag = mp->mnt_flag & MNT_ASYNC; 629 mp->mnt_flag &= ~MNT_ASYNC; 630 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p); 631 if (asyncflag) 632 mp->mnt_flag |= MNT_ASYNC; 633 vn_finished_write(mp, 0); 634 } 635 simple_lock(&mountlist_slock); 636 nmp = mp->mnt_list.cqe_prev; 637 vfs_unbusy(mp); 638 639 } 640 simple_unlock(&mountlist_slock); 641 #ifdef DEBUG 642 if (syncprt) 643 vfs_bufstats(); 644 #endif /* DEBUG */ 645 return (0); 646 } 647 648 /* 649 * Change filesystem quotas. 650 */ 651 /* ARGSUSED */ 652 int 653 sys_quotactl(struct lwp *l, void *v, register_t *retval) 654 { 655 struct sys_quotactl_args /* { 656 syscallarg(const char *) path; 657 syscallarg(int) cmd; 658 syscallarg(int) uid; 659 syscallarg(caddr_t) arg; 660 } */ *uap = v; 661 struct proc *p = l->l_proc; 662 struct mount *mp; 663 int error; 664 struct nameidata nd; 665 666 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 667 if ((error = namei(&nd)) != 0) 668 return (error); 669 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 670 vrele(nd.ni_vp); 671 if (error) 672 return (error); 673 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 674 SCARG(uap, arg), p); 675 vn_finished_write(mp, 0); 676 return (error); 677 } 678 679 int 680 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags, 681 int root) 682 { 683 struct cwdinfo *cwdi = p->p_cwdi; 684 int error = 0; 685 686 /* 687 * If MNT_NOWAIT or MNT_LAZY is specified, do not 688 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 689 * overrides MNT_NOWAIT. 690 */ 691 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 692 (flags != MNT_WAIT && flags != 0)) { 693 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 694 goto done; 695 } 696 697 /* Get the filesystem stats now */ 698 memset(sp, 0, sizeof(*sp)); 699 if ((error = VFS_STATVFS(mp, sp, p)) != 0) { 700 return error; 701 } 702 703 if (cwdi->cwdi_rdir == NULL) 704 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 705 done: 706 if (cwdi->cwdi_rdir != NULL) { 707 size_t len; 708 char *bp; 709 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 710 if (!path) 711 return ENOMEM; 712 713 bp = path + MAXPATHLEN; 714 *--bp = '\0'; 715 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 716 MAXPATHLEN / 2, 0, p); 717 if (error) { 718 free(path, M_TEMP); 719 return error; 720 } 721 len = strlen(bp); 722 /* 723 * for mount points that are below our root, we can see 724 * them, so we fix up the pathname and return them. The 725 * rest we cannot see, so we don't allow viewing the 726 * data. 727 */ 728 if (strncmp(bp, sp->f_mntonname, len) == 0) { 729 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 730 sizeof(sp->f_mntonname)); 731 if (sp->f_mntonname[0] == '\0') 732 (void)strlcpy(sp->f_mntonname, "/", 733 sizeof(sp->f_mntonname)); 734 } else { 735 if (root) 736 (void)strlcpy(sp->f_mntonname, "/", 737 sizeof(sp->f_mntonname)); 738 else 739 error = EPERM; 740 } 741 free(path, M_TEMP); 742 } 743 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 744 return error; 745 } 746 747 /* 748 * Get filesystem statistics. 749 */ 750 /* ARGSUSED */ 751 int 752 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 753 { 754 struct sys_statvfs1_args /* { 755 syscallarg(const char *) path; 756 syscallarg(struct statvfs *) buf; 757 syscallarg(int) flags; 758 } */ *uap = v; 759 struct proc *p = l->l_proc; 760 struct mount *mp; 761 struct statvfs sbuf; 762 int error; 763 struct nameidata nd; 764 765 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 766 if ((error = namei(&nd)) != 0) 767 return error; 768 mp = nd.ni_vp->v_mount; 769 vrele(nd.ni_vp); 770 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 771 return error; 772 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 773 } 774 775 /* 776 * Get filesystem statistics. 777 */ 778 /* ARGSUSED */ 779 int 780 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 781 { 782 struct sys_fstatvfs1_args /* { 783 syscallarg(int) fd; 784 syscallarg(struct statvfs *) buf; 785 syscallarg(int) flags; 786 } */ *uap = v; 787 struct proc *p = l->l_proc; 788 struct file *fp; 789 struct mount *mp; 790 struct statvfs sbuf; 791 int error; 792 793 /* getvnode() will use the descriptor for us */ 794 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 795 return (error); 796 mp = ((struct vnode *)fp->f_data)->v_mount; 797 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 798 goto out; 799 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 800 out: 801 FILE_UNUSE(fp, p); 802 return error; 803 } 804 805 806 /* 807 * Get statistics on all filesystems. 808 */ 809 int 810 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 811 { 812 struct sys_getvfsstat_args /* { 813 syscallarg(struct statvfs *) buf; 814 syscallarg(size_t) bufsize; 815 syscallarg(int) flags; 816 } */ *uap = v; 817 int root = 0; 818 struct proc *p = l->l_proc; 819 struct mount *mp, *nmp; 820 struct statvfs sbuf; 821 struct statvfs *sfsp; 822 size_t count, maxcount; 823 int error = 0; 824 825 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 826 sfsp = SCARG(uap, buf); 827 simple_lock(&mountlist_slock); 828 count = 0; 829 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 830 mp = nmp) { 831 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 832 nmp = CIRCLEQ_NEXT(mp, mnt_list); 833 continue; 834 } 835 if (sfsp && count < maxcount) { 836 error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0); 837 if (error) { 838 simple_lock(&mountlist_slock); 839 nmp = CIRCLEQ_NEXT(mp, mnt_list); 840 vfs_unbusy(mp); 841 continue; 842 } 843 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 844 if (error) { 845 vfs_unbusy(mp); 846 return (error); 847 } 848 sfsp++; 849 root |= strcmp(sbuf.f_mntonname, "/") == 0; 850 } 851 count++; 852 simple_lock(&mountlist_slock); 853 nmp = CIRCLEQ_NEXT(mp, mnt_list); 854 vfs_unbusy(mp); 855 } 856 simple_unlock(&mountlist_slock); 857 if (root == 0 && p->p_cwdi->cwdi_rdir) { 858 /* 859 * fake a root entry 860 */ 861 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p, 862 SCARG(uap, flags), 1)) != 0) 863 return error; 864 if (sfsp) 865 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 866 count++; 867 } 868 if (sfsp && count > maxcount) 869 *retval = maxcount; 870 else 871 *retval = count; 872 return error; 873 } 874 875 /* 876 * Change current working directory to a given file descriptor. 877 */ 878 /* ARGSUSED */ 879 int 880 sys_fchdir(struct lwp *l, void *v, register_t *retval) 881 { 882 struct sys_fchdir_args /* { 883 syscallarg(int) fd; 884 } */ *uap = v; 885 struct proc *p = l->l_proc; 886 struct filedesc *fdp = p->p_fd; 887 struct cwdinfo *cwdi = p->p_cwdi; 888 struct vnode *vp, *tdp; 889 struct mount *mp; 890 struct file *fp; 891 int error; 892 893 /* getvnode() will use the descriptor for us */ 894 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 895 return (error); 896 vp = (struct vnode *)fp->f_data; 897 898 VREF(vp); 899 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 900 if (vp->v_type != VDIR) 901 error = ENOTDIR; 902 else 903 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 904 while (!error && (mp = vp->v_mountedhere) != NULL) { 905 if (vfs_busy(mp, 0, 0)) 906 continue; 907 error = VFS_ROOT(mp, &tdp); 908 vfs_unbusy(mp); 909 if (error) 910 break; 911 vput(vp); 912 vp = tdp; 913 } 914 if (error) { 915 vput(vp); 916 goto out; 917 } 918 VOP_UNLOCK(vp, 0); 919 920 /* 921 * Disallow changing to a directory not under the process's 922 * current root directory (if there is one). 923 */ 924 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) { 925 vrele(vp); 926 error = EPERM; /* operation not permitted */ 927 goto out; 928 } 929 930 vrele(cwdi->cwdi_cdir); 931 cwdi->cwdi_cdir = vp; 932 out: 933 FILE_UNUSE(fp, p); 934 return (error); 935 } 936 937 /* 938 * Change this process's notion of the root directory to a given file 939 * descriptor. 940 */ 941 int 942 sys_fchroot(struct lwp *l, void *v, register_t *retval) 943 { 944 struct sys_fchroot_args *uap = v; 945 struct proc *p = l->l_proc; 946 struct filedesc *fdp = p->p_fd; 947 struct cwdinfo *cwdi = p->p_cwdi; 948 struct vnode *vp; 949 struct file *fp; 950 int error; 951 952 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 953 return error; 954 /* getvnode() will use the descriptor for us */ 955 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 956 return error; 957 vp = (struct vnode *) fp->f_data; 958 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 959 if (vp->v_type != VDIR) 960 error = ENOTDIR; 961 else 962 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 963 VOP_UNLOCK(vp, 0); 964 if (error) 965 goto out; 966 VREF(vp); 967 968 /* 969 * Prevent escaping from chroot by putting the root under 970 * the working directory. Silently chdir to / if we aren't 971 * already there. 972 */ 973 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 974 /* 975 * XXX would be more failsafe to change directory to a 976 * deadfs node here instead 977 */ 978 vrele(cwdi->cwdi_cdir); 979 VREF(vp); 980 cwdi->cwdi_cdir = vp; 981 } 982 983 if (cwdi->cwdi_rdir != NULL) 984 vrele(cwdi->cwdi_rdir); 985 cwdi->cwdi_rdir = vp; 986 out: 987 FILE_UNUSE(fp, p); 988 return (error); 989 } 990 991 /* 992 * Change current working directory (``.''). 993 */ 994 /* ARGSUSED */ 995 int 996 sys_chdir(struct lwp *l, void *v, register_t *retval) 997 { 998 struct sys_chdir_args /* { 999 syscallarg(const char *) path; 1000 } */ *uap = v; 1001 struct proc *p = l->l_proc; 1002 struct cwdinfo *cwdi = p->p_cwdi; 1003 int error; 1004 struct nameidata nd; 1005 1006 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1007 SCARG(uap, path), p); 1008 if ((error = change_dir(&nd, p)) != 0) 1009 return (error); 1010 vrele(cwdi->cwdi_cdir); 1011 cwdi->cwdi_cdir = nd.ni_vp; 1012 return (0); 1013 } 1014 1015 /* 1016 * Change notion of root (``/'') directory. 1017 */ 1018 /* ARGSUSED */ 1019 int 1020 sys_chroot(struct lwp *l, void *v, register_t *retval) 1021 { 1022 struct sys_chroot_args /* { 1023 syscallarg(const char *) path; 1024 } */ *uap = v; 1025 struct proc *p = l->l_proc; 1026 struct cwdinfo *cwdi = p->p_cwdi; 1027 struct vnode *vp; 1028 int error; 1029 struct nameidata nd; 1030 1031 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1032 return (error); 1033 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1034 SCARG(uap, path), p); 1035 if ((error = change_dir(&nd, p)) != 0) 1036 return (error); 1037 if (cwdi->cwdi_rdir != NULL) 1038 vrele(cwdi->cwdi_rdir); 1039 vp = nd.ni_vp; 1040 cwdi->cwdi_rdir = vp; 1041 1042 /* 1043 * Prevent escaping from chroot by putting the root under 1044 * the working directory. Silently chdir to / if we aren't 1045 * already there. 1046 */ 1047 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 1048 /* 1049 * XXX would be more failsafe to change directory to a 1050 * deadfs node here instead 1051 */ 1052 vrele(cwdi->cwdi_cdir); 1053 VREF(vp); 1054 cwdi->cwdi_cdir = vp; 1055 } 1056 1057 return (0); 1058 } 1059 1060 /* 1061 * Common routine for chroot and chdir. 1062 */ 1063 static int 1064 change_dir(struct nameidata *ndp, struct proc *p) 1065 { 1066 struct vnode *vp; 1067 int error; 1068 1069 if ((error = namei(ndp)) != 0) 1070 return (error); 1071 vp = ndp->ni_vp; 1072 if (vp->v_type != VDIR) 1073 error = ENOTDIR; 1074 else 1075 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 1076 1077 if (error) 1078 vput(vp); 1079 else 1080 VOP_UNLOCK(vp, 0); 1081 return (error); 1082 } 1083 1084 /* 1085 * Check permissions, allocate an open file structure, 1086 * and call the device open routine if any. 1087 */ 1088 int 1089 sys_open(struct lwp *l, void *v, register_t *retval) 1090 { 1091 struct sys_open_args /* { 1092 syscallarg(const char *) path; 1093 syscallarg(int) flags; 1094 syscallarg(int) mode; 1095 } */ *uap = v; 1096 struct proc *p = l->l_proc; 1097 struct cwdinfo *cwdi = p->p_cwdi; 1098 struct filedesc *fdp = p->p_fd; 1099 struct file *fp; 1100 struct vnode *vp; 1101 int flags, cmode; 1102 int type, indx, error; 1103 struct flock lf; 1104 struct nameidata nd; 1105 1106 flags = FFLAGS(SCARG(uap, flags)); 1107 if ((flags & (FREAD | FWRITE)) == 0) 1108 return (EINVAL); 1109 /* falloc() will use the file descriptor for us */ 1110 if ((error = falloc(p, &fp, &indx)) != 0) 1111 return (error); 1112 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1113 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1114 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1115 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1116 FILE_UNUSE(fp, p); 1117 fdp->fd_ofiles[indx] = NULL; 1118 ffree(fp); 1119 if ((error == EDUPFD || error == EMOVEFD) && 1120 l->l_dupfd >= 0 && /* XXX from fdopen */ 1121 (error = 1122 dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) { 1123 *retval = indx; 1124 return (0); 1125 } 1126 if (error == ERESTART) 1127 error = EINTR; 1128 fdremove(fdp, indx); 1129 return (error); 1130 } 1131 l->l_dupfd = 0; 1132 vp = nd.ni_vp; 1133 fp->f_flag = flags & FMASK; 1134 fp->f_type = DTYPE_VNODE; 1135 fp->f_ops = &vnops; 1136 fp->f_data = vp; 1137 if (flags & (O_EXLOCK | O_SHLOCK)) { 1138 lf.l_whence = SEEK_SET; 1139 lf.l_start = 0; 1140 lf.l_len = 0; 1141 if (flags & O_EXLOCK) 1142 lf.l_type = F_WRLCK; 1143 else 1144 lf.l_type = F_RDLCK; 1145 type = F_FLOCK; 1146 if ((flags & FNONBLOCK) == 0) 1147 type |= F_WAIT; 1148 VOP_UNLOCK(vp, 0); 1149 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1150 if (error) { 1151 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1152 FILE_UNUSE(fp, p); 1153 ffree(fp); 1154 fdremove(fdp, indx); 1155 return (error); 1156 } 1157 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1158 fp->f_flag |= FHASLOCK; 1159 } 1160 VOP_UNLOCK(vp, 0); 1161 *retval = indx; 1162 FILE_SET_MATURE(fp); 1163 FILE_UNUSE(fp, p); 1164 return (0); 1165 } 1166 1167 /* 1168 * Get file handle system call 1169 */ 1170 int 1171 sys_getfh(struct lwp *l, void *v, register_t *retval) 1172 { 1173 struct sys_getfh_args /* { 1174 syscallarg(char *) fname; 1175 syscallarg(fhandle_t *) fhp; 1176 } */ *uap = v; 1177 struct proc *p = l->l_proc; 1178 struct vnode *vp; 1179 fhandle_t fh; 1180 int error; 1181 struct nameidata nd; 1182 1183 /* 1184 * Must be super user 1185 */ 1186 error = suser(p->p_ucred, &p->p_acflag); 1187 if (error) 1188 return (error); 1189 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1190 SCARG(uap, fname), p); 1191 error = namei(&nd); 1192 if (error) 1193 return (error); 1194 vp = nd.ni_vp; 1195 memset(&fh, 0, sizeof(fh)); 1196 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1197 error = VFS_VPTOFH(vp, &fh.fh_fid); 1198 vput(vp); 1199 if (error) 1200 return (error); 1201 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1202 return (error); 1203 } 1204 1205 /* 1206 * Open a file given a file handle. 1207 * 1208 * Check permissions, allocate an open file structure, 1209 * and call the device open routine if any. 1210 */ 1211 int 1212 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1213 { 1214 struct sys_fhopen_args /* { 1215 syscallarg(const fhandle_t *) fhp; 1216 syscallarg(int) flags; 1217 } */ *uap = v; 1218 struct proc *p = l->l_proc; 1219 struct filedesc *fdp = p->p_fd; 1220 struct file *fp; 1221 struct vnode *vp = NULL; 1222 struct mount *mp; 1223 struct ucred *cred = p->p_ucred; 1224 int flags; 1225 struct file *nfp; 1226 int type, indx, error=0; 1227 struct flock lf; 1228 struct vattr va; 1229 fhandle_t fh; 1230 1231 /* 1232 * Must be super user 1233 */ 1234 if ((error = suser(p->p_ucred, &p->p_acflag))) 1235 return (error); 1236 1237 flags = FFLAGS(SCARG(uap, flags)); 1238 if ((flags & (FREAD | FWRITE)) == 0) 1239 return (EINVAL); 1240 if ((flags & O_CREAT)) 1241 return (EINVAL); 1242 /* falloc() will use the file descriptor for us */ 1243 if ((error = falloc(p, &nfp, &indx)) != 0) 1244 return (error); 1245 fp = nfp; 1246 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1247 goto bad; 1248 1249 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1250 error = ESTALE; 1251 goto bad; 1252 } 1253 1254 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1255 vp = NULL; /* most likely unnecessary sanity for bad: */ 1256 goto bad; 1257 } 1258 1259 /* Now do an effective vn_open */ 1260 1261 if (vp->v_type == VSOCK) { 1262 error = EOPNOTSUPP; 1263 goto bad; 1264 } 1265 if (flags & FREAD) { 1266 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 1267 goto bad; 1268 } 1269 if (flags & (FWRITE | O_TRUNC)) { 1270 if (vp->v_type == VDIR) { 1271 error = EISDIR; 1272 goto bad; 1273 } 1274 if ((error = vn_writechk(vp)) != 0 || 1275 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 1276 goto bad; 1277 } 1278 if (flags & O_TRUNC) { 1279 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1280 goto bad; 1281 VOP_UNLOCK(vp, 0); /* XXX */ 1282 VOP_LEASE(vp, p, cred, LEASE_WRITE); 1283 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1284 VATTR_NULL(&va); 1285 va.va_size = 0; 1286 error = VOP_SETATTR(vp, &va, cred, p); 1287 vn_finished_write(mp, 0); 1288 if (error) 1289 goto bad; 1290 } 1291 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) 1292 goto bad; 1293 if (vp->v_type == VREG && 1294 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1295 error = EIO; 1296 goto bad; 1297 } 1298 if (flags & FWRITE) 1299 vp->v_writecount++; 1300 1301 /* done with modified vn_open, now finish what sys_open does. */ 1302 1303 fp->f_flag = flags & FMASK; 1304 fp->f_type = DTYPE_VNODE; 1305 fp->f_ops = &vnops; 1306 fp->f_data = vp; 1307 if (flags & (O_EXLOCK | O_SHLOCK)) { 1308 lf.l_whence = SEEK_SET; 1309 lf.l_start = 0; 1310 lf.l_len = 0; 1311 if (flags & O_EXLOCK) 1312 lf.l_type = F_WRLCK; 1313 else 1314 lf.l_type = F_RDLCK; 1315 type = F_FLOCK; 1316 if ((flags & FNONBLOCK) == 0) 1317 type |= F_WAIT; 1318 VOP_UNLOCK(vp, 0); 1319 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1320 if (error) { 1321 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1322 FILE_UNUSE(fp, p); 1323 ffree(fp); 1324 fdremove(fdp, indx); 1325 return (error); 1326 } 1327 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1328 fp->f_flag |= FHASLOCK; 1329 } 1330 VOP_UNLOCK(vp, 0); 1331 *retval = indx; 1332 FILE_SET_MATURE(fp); 1333 FILE_UNUSE(fp, p); 1334 return (0); 1335 1336 bad: 1337 FILE_UNUSE(fp, p); 1338 ffree(fp); 1339 fdremove(fdp, indx); 1340 if (vp != NULL) 1341 vput(vp); 1342 return (error); 1343 } 1344 1345 /* ARGSUSED */ 1346 int 1347 sys_fhstat(struct lwp *l, void *v, register_t *retval) 1348 { 1349 struct sys_fhstat_args /* { 1350 syscallarg(const fhandle_t *) fhp; 1351 syscallarg(struct stat *) sb; 1352 } */ *uap = v; 1353 struct proc *p = l->l_proc; 1354 struct stat sb; 1355 int error; 1356 fhandle_t fh; 1357 struct mount *mp; 1358 struct vnode *vp; 1359 1360 /* 1361 * Must be super user 1362 */ 1363 if ((error = suser(p->p_ucred, &p->p_acflag))) 1364 return (error); 1365 1366 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1367 return (error); 1368 1369 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1370 return (ESTALE); 1371 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1372 return (error); 1373 error = vn_stat(vp, &sb, p); 1374 vput(vp); 1375 if (error) 1376 return (error); 1377 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1378 return (error); 1379 } 1380 1381 /* ARGSUSED */ 1382 int 1383 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1384 { 1385 struct sys_fhstatvfs1_args /* { 1386 syscallarg(const fhandle_t *) fhp; 1387 syscallarg(struct statvfs *) buf; 1388 syscallarg(int) flags; 1389 } */ *uap = v; 1390 struct proc *p = l->l_proc; 1391 struct statvfs sbuf; 1392 fhandle_t fh; 1393 struct mount *mp; 1394 struct vnode *vp; 1395 int error; 1396 1397 /* 1398 * Must be super user 1399 */ 1400 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1401 return error; 1402 1403 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1404 return error; 1405 1406 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1407 return ESTALE; 1408 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1409 return error; 1410 1411 mp = vp->v_mount; 1412 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) { 1413 vput(vp); 1414 return error; 1415 } 1416 vput(vp); 1417 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1418 } 1419 1420 /* 1421 * Create a special file. 1422 */ 1423 /* ARGSUSED */ 1424 int 1425 sys_mknod(struct lwp *l, void *v, register_t *retval) 1426 { 1427 struct sys_mknod_args /* { 1428 syscallarg(const char *) path; 1429 syscallarg(int) mode; 1430 syscallarg(int) dev; 1431 } */ *uap = v; 1432 struct proc *p = l->l_proc; 1433 struct vnode *vp; 1434 struct mount *mp; 1435 struct vattr vattr; 1436 int error; 1437 int whiteout = 0; 1438 struct nameidata nd; 1439 1440 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1441 return (error); 1442 restart: 1443 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1444 if ((error = namei(&nd)) != 0) 1445 return (error); 1446 vp = nd.ni_vp; 1447 if (vp != NULL) 1448 error = EEXIST; 1449 else { 1450 VATTR_NULL(&vattr); 1451 vattr.va_mode = 1452 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1453 vattr.va_rdev = SCARG(uap, dev); 1454 whiteout = 0; 1455 1456 switch (SCARG(uap, mode) & S_IFMT) { 1457 case S_IFMT: /* used by badsect to flag bad sectors */ 1458 vattr.va_type = VBAD; 1459 break; 1460 case S_IFCHR: 1461 vattr.va_type = VCHR; 1462 break; 1463 case S_IFBLK: 1464 vattr.va_type = VBLK; 1465 break; 1466 case S_IFWHT: 1467 whiteout = 1; 1468 break; 1469 default: 1470 error = EINVAL; 1471 break; 1472 } 1473 } 1474 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1475 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1476 if (nd.ni_dvp == vp) 1477 vrele(nd.ni_dvp); 1478 else 1479 vput(nd.ni_dvp); 1480 if (vp) 1481 vrele(vp); 1482 if ((error = vn_start_write(NULL, &mp, 1483 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1484 return (error); 1485 goto restart; 1486 } 1487 if (!error) { 1488 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1489 if (whiteout) { 1490 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1491 if (error) 1492 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1493 vput(nd.ni_dvp); 1494 } else { 1495 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1496 &nd.ni_cnd, &vattr); 1497 if (error == 0) 1498 vput(nd.ni_vp); 1499 } 1500 } else { 1501 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1502 if (nd.ni_dvp == vp) 1503 vrele(nd.ni_dvp); 1504 else 1505 vput(nd.ni_dvp); 1506 if (vp) 1507 vrele(vp); 1508 } 1509 vn_finished_write(mp, 0); 1510 return (error); 1511 } 1512 1513 /* 1514 * Create a named pipe. 1515 */ 1516 /* ARGSUSED */ 1517 int 1518 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1519 { 1520 struct sys_mkfifo_args /* { 1521 syscallarg(const char *) path; 1522 syscallarg(int) mode; 1523 } */ *uap = v; 1524 struct proc *p = l->l_proc; 1525 struct mount *mp; 1526 struct vattr vattr; 1527 int error; 1528 struct nameidata nd; 1529 1530 restart: 1531 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1532 if ((error = namei(&nd)) != 0) 1533 return (error); 1534 if (nd.ni_vp != NULL) { 1535 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1536 if (nd.ni_dvp == nd.ni_vp) 1537 vrele(nd.ni_dvp); 1538 else 1539 vput(nd.ni_dvp); 1540 vrele(nd.ni_vp); 1541 return (EEXIST); 1542 } 1543 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1544 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1545 if (nd.ni_dvp == nd.ni_vp) 1546 vrele(nd.ni_dvp); 1547 else 1548 vput(nd.ni_dvp); 1549 if (nd.ni_vp) 1550 vrele(nd.ni_vp); 1551 if ((error = vn_start_write(NULL, &mp, 1552 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1553 return (error); 1554 goto restart; 1555 } 1556 VATTR_NULL(&vattr); 1557 vattr.va_type = VFIFO; 1558 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1559 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1560 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1561 if (error == 0) 1562 vput(nd.ni_vp); 1563 vn_finished_write(mp, 0); 1564 return (error); 1565 } 1566 1567 /* 1568 * Make a hard file link. 1569 */ 1570 /* ARGSUSED */ 1571 int 1572 sys_link(struct lwp *l, void *v, register_t *retval) 1573 { 1574 struct sys_link_args /* { 1575 syscallarg(const char *) path; 1576 syscallarg(const char *) link; 1577 } */ *uap = v; 1578 struct proc *p = l->l_proc; 1579 struct vnode *vp; 1580 struct mount *mp; 1581 struct nameidata nd; 1582 int error; 1583 1584 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1585 if ((error = namei(&nd)) != 0) 1586 return (error); 1587 vp = nd.ni_vp; 1588 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1589 vrele(vp); 1590 return (error); 1591 } 1592 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1593 if ((error = namei(&nd)) != 0) 1594 goto out; 1595 if (nd.ni_vp) { 1596 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1597 if (nd.ni_dvp == nd.ni_vp) 1598 vrele(nd.ni_dvp); 1599 else 1600 vput(nd.ni_dvp); 1601 vrele(nd.ni_vp); 1602 error = EEXIST; 1603 goto out; 1604 } 1605 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1606 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1607 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1608 out: 1609 vrele(vp); 1610 vn_finished_write(mp, 0); 1611 return (error); 1612 } 1613 1614 /* 1615 * Make a symbolic link. 1616 */ 1617 /* ARGSUSED */ 1618 int 1619 sys_symlink(struct lwp *l, void *v, register_t *retval) 1620 { 1621 struct sys_symlink_args /* { 1622 syscallarg(const char *) path; 1623 syscallarg(const char *) link; 1624 } */ *uap = v; 1625 struct proc *p = l->l_proc; 1626 struct mount *mp; 1627 struct vattr vattr; 1628 char *path; 1629 int error; 1630 struct nameidata nd; 1631 1632 path = PNBUF_GET(); 1633 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1634 if (error) 1635 goto out; 1636 restart: 1637 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1638 if ((error = namei(&nd)) != 0) 1639 goto out; 1640 if (nd.ni_vp) { 1641 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1642 if (nd.ni_dvp == nd.ni_vp) 1643 vrele(nd.ni_dvp); 1644 else 1645 vput(nd.ni_dvp); 1646 vrele(nd.ni_vp); 1647 error = EEXIST; 1648 goto out; 1649 } 1650 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1651 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1652 if (nd.ni_dvp == nd.ni_vp) 1653 vrele(nd.ni_dvp); 1654 else 1655 vput(nd.ni_dvp); 1656 if ((error = vn_start_write(NULL, &mp, 1657 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1658 return (error); 1659 goto restart; 1660 } 1661 VATTR_NULL(&vattr); 1662 vattr.va_type = VLNK; 1663 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1664 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1665 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1666 if (error == 0) 1667 vput(nd.ni_vp); 1668 vn_finished_write(mp, 0); 1669 out: 1670 PNBUF_PUT(path); 1671 return (error); 1672 } 1673 1674 /* 1675 * Delete a whiteout from the filesystem. 1676 */ 1677 /* ARGSUSED */ 1678 int 1679 sys_undelete(struct lwp *l, void *v, register_t *retval) 1680 { 1681 struct sys_undelete_args /* { 1682 syscallarg(const char *) path; 1683 } */ *uap = v; 1684 struct proc *p = l->l_proc; 1685 int error; 1686 struct mount *mp; 1687 struct nameidata nd; 1688 1689 restart: 1690 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1691 SCARG(uap, path), p); 1692 error = namei(&nd); 1693 if (error) 1694 return (error); 1695 1696 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1697 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1698 if (nd.ni_dvp == nd.ni_vp) 1699 vrele(nd.ni_dvp); 1700 else 1701 vput(nd.ni_dvp); 1702 if (nd.ni_vp) 1703 vrele(nd.ni_vp); 1704 return (EEXIST); 1705 } 1706 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1707 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1708 if (nd.ni_dvp == nd.ni_vp) 1709 vrele(nd.ni_dvp); 1710 else 1711 vput(nd.ni_dvp); 1712 if ((error = vn_start_write(NULL, &mp, 1713 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1714 return (error); 1715 goto restart; 1716 } 1717 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1718 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1719 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1720 vput(nd.ni_dvp); 1721 vn_finished_write(mp, 0); 1722 return (error); 1723 } 1724 1725 /* 1726 * Delete a name from the filesystem. 1727 */ 1728 /* ARGSUSED */ 1729 int 1730 sys_unlink(struct lwp *l, void *v, register_t *retval) 1731 { 1732 struct sys_unlink_args /* { 1733 syscallarg(const char *) path; 1734 } */ *uap = v; 1735 struct proc *p = l->l_proc; 1736 struct mount *mp; 1737 struct vnode *vp; 1738 int error; 1739 struct nameidata nd; 1740 1741 restart: 1742 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1743 SCARG(uap, path), p); 1744 if ((error = namei(&nd)) != 0) 1745 return (error); 1746 vp = nd.ni_vp; 1747 1748 /* 1749 * The root of a mounted filesystem cannot be deleted. 1750 */ 1751 if (vp->v_flag & VROOT) { 1752 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1753 if (nd.ni_dvp == vp) 1754 vrele(nd.ni_dvp); 1755 else 1756 vput(nd.ni_dvp); 1757 vput(vp); 1758 error = EBUSY; 1759 goto out; 1760 } 1761 1762 #ifdef VERIFIED_EXEC 1763 /* Handle remove requests for veriexec entries. */ 1764 if ((error = veriexec_removechk(p, vp, nd.ni_dirp)) != 0) { 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 vput(vp); 1771 goto out; 1772 } 1773 #endif /* VERIFIED_EXEC */ 1774 1775 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1776 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1777 if (nd.ni_dvp == vp) 1778 vrele(nd.ni_dvp); 1779 else 1780 vput(nd.ni_dvp); 1781 vput(vp); 1782 if ((error = vn_start_write(NULL, &mp, 1783 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1784 return (error); 1785 goto restart; 1786 } 1787 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1788 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1789 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1790 vn_finished_write(mp, 0); 1791 out: 1792 return (error); 1793 } 1794 1795 /* 1796 * Reposition read/write file offset. 1797 */ 1798 int 1799 sys_lseek(struct lwp *l, void *v, register_t *retval) 1800 { 1801 struct sys_lseek_args /* { 1802 syscallarg(int) fd; 1803 syscallarg(int) pad; 1804 syscallarg(off_t) offset; 1805 syscallarg(int) whence; 1806 } */ *uap = v; 1807 struct proc *p = l->l_proc; 1808 struct ucred *cred = p->p_ucred; 1809 struct filedesc *fdp = p->p_fd; 1810 struct file *fp; 1811 struct vnode *vp; 1812 struct vattr vattr; 1813 off_t newoff; 1814 int error; 1815 1816 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1817 return (EBADF); 1818 1819 FILE_USE(fp); 1820 1821 vp = (struct vnode *)fp->f_data; 1822 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1823 error = ESPIPE; 1824 goto out; 1825 } 1826 1827 switch (SCARG(uap, whence)) { 1828 case SEEK_CUR: 1829 newoff = fp->f_offset + SCARG(uap, offset); 1830 break; 1831 case SEEK_END: 1832 error = VOP_GETATTR(vp, &vattr, cred, p); 1833 if (error) 1834 goto out; 1835 newoff = SCARG(uap, offset) + vattr.va_size; 1836 break; 1837 case SEEK_SET: 1838 newoff = SCARG(uap, offset); 1839 break; 1840 default: 1841 error = EINVAL; 1842 goto out; 1843 } 1844 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1845 goto out; 1846 1847 *(off_t *)retval = fp->f_offset = newoff; 1848 out: 1849 FILE_UNUSE(fp, p); 1850 return (error); 1851 } 1852 1853 /* 1854 * Positional read system call. 1855 */ 1856 int 1857 sys_pread(struct lwp *l, void *v, register_t *retval) 1858 { 1859 struct sys_pread_args /* { 1860 syscallarg(int) fd; 1861 syscallarg(void *) buf; 1862 syscallarg(size_t) nbyte; 1863 syscallarg(off_t) offset; 1864 } */ *uap = v; 1865 struct proc *p = l->l_proc; 1866 struct filedesc *fdp = p->p_fd; 1867 struct file *fp; 1868 struct vnode *vp; 1869 off_t offset; 1870 int error, fd = SCARG(uap, fd); 1871 1872 if ((fp = fd_getfile(fdp, fd)) == NULL) 1873 return (EBADF); 1874 1875 if ((fp->f_flag & FREAD) == 0) { 1876 simple_unlock(&fp->f_slock); 1877 return (EBADF); 1878 } 1879 1880 FILE_USE(fp); 1881 1882 vp = (struct vnode *)fp->f_data; 1883 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1884 error = ESPIPE; 1885 goto out; 1886 } 1887 1888 offset = SCARG(uap, offset); 1889 1890 /* 1891 * XXX This works because no file systems actually 1892 * XXX take any action on the seek operation. 1893 */ 1894 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1895 goto out; 1896 1897 /* dofileread() will unuse the descriptor for us */ 1898 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1899 &offset, 0, retval)); 1900 1901 out: 1902 FILE_UNUSE(fp, p); 1903 return (error); 1904 } 1905 1906 /* 1907 * Positional scatter read system call. 1908 */ 1909 int 1910 sys_preadv(struct lwp *l, void *v, register_t *retval) 1911 { 1912 struct sys_preadv_args /* { 1913 syscallarg(int) fd; 1914 syscallarg(const struct iovec *) iovp; 1915 syscallarg(int) iovcnt; 1916 syscallarg(off_t) offset; 1917 } */ *uap = v; 1918 struct proc *p = l->l_proc; 1919 struct filedesc *fdp = p->p_fd; 1920 struct file *fp; 1921 struct vnode *vp; 1922 off_t offset; 1923 int error, fd = SCARG(uap, fd); 1924 1925 if ((fp = fd_getfile(fdp, fd)) == NULL) 1926 return (EBADF); 1927 1928 if ((fp->f_flag & FREAD) == 0) { 1929 simple_unlock(&fp->f_slock); 1930 return (EBADF); 1931 } 1932 1933 FILE_USE(fp); 1934 1935 vp = (struct vnode *)fp->f_data; 1936 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1937 error = ESPIPE; 1938 goto out; 1939 } 1940 1941 offset = SCARG(uap, offset); 1942 1943 /* 1944 * XXX This works because no file systems actually 1945 * XXX take any action on the seek operation. 1946 */ 1947 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1948 goto out; 1949 1950 /* dofilereadv() will unuse the descriptor for us */ 1951 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1952 &offset, 0, retval)); 1953 1954 out: 1955 FILE_UNUSE(fp, p); 1956 return (error); 1957 } 1958 1959 /* 1960 * Positional write system call. 1961 */ 1962 int 1963 sys_pwrite(struct lwp *l, void *v, register_t *retval) 1964 { 1965 struct sys_pwrite_args /* { 1966 syscallarg(int) fd; 1967 syscallarg(const void *) buf; 1968 syscallarg(size_t) nbyte; 1969 syscallarg(off_t) offset; 1970 } */ *uap = v; 1971 struct proc *p = l->l_proc; 1972 struct filedesc *fdp = p->p_fd; 1973 struct file *fp; 1974 struct vnode *vp; 1975 off_t offset; 1976 int error, fd = SCARG(uap, fd); 1977 1978 if ((fp = fd_getfile(fdp, fd)) == NULL) 1979 return (EBADF); 1980 1981 if ((fp->f_flag & FWRITE) == 0) { 1982 simple_unlock(&fp->f_slock); 1983 return (EBADF); 1984 } 1985 1986 FILE_USE(fp); 1987 1988 vp = (struct vnode *)fp->f_data; 1989 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1990 error = ESPIPE; 1991 goto out; 1992 } 1993 1994 offset = SCARG(uap, offset); 1995 1996 /* 1997 * XXX This works because no file systems actually 1998 * XXX take any action on the seek operation. 1999 */ 2000 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2001 goto out; 2002 2003 /* dofilewrite() will unuse the descriptor for us */ 2004 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2005 &offset, 0, retval)); 2006 2007 out: 2008 FILE_UNUSE(fp, p); 2009 return (error); 2010 } 2011 2012 /* 2013 * Positional gather write system call. 2014 */ 2015 int 2016 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2017 { 2018 struct sys_pwritev_args /* { 2019 syscallarg(int) fd; 2020 syscallarg(const struct iovec *) iovp; 2021 syscallarg(int) iovcnt; 2022 syscallarg(off_t) offset; 2023 } */ *uap = v; 2024 struct proc *p = l->l_proc; 2025 struct filedesc *fdp = p->p_fd; 2026 struct file *fp; 2027 struct vnode *vp; 2028 off_t offset; 2029 int error, fd = SCARG(uap, fd); 2030 2031 if ((fp = fd_getfile(fdp, fd)) == NULL) 2032 return (EBADF); 2033 2034 if ((fp->f_flag & FWRITE) == 0) { 2035 simple_unlock(&fp->f_slock); 2036 return (EBADF); 2037 } 2038 2039 FILE_USE(fp); 2040 2041 vp = (struct vnode *)fp->f_data; 2042 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2043 error = ESPIPE; 2044 goto out; 2045 } 2046 2047 offset = SCARG(uap, offset); 2048 2049 /* 2050 * XXX This works because no file systems actually 2051 * XXX take any action on the seek operation. 2052 */ 2053 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2054 goto out; 2055 2056 /* dofilewritev() will unuse the descriptor for us */ 2057 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2058 &offset, 0, retval)); 2059 2060 out: 2061 FILE_UNUSE(fp, p); 2062 return (error); 2063 } 2064 2065 /* 2066 * Check access permissions. 2067 */ 2068 int 2069 sys_access(struct lwp *l, void *v, register_t *retval) 2070 { 2071 struct sys_access_args /* { 2072 syscallarg(const char *) path; 2073 syscallarg(int) flags; 2074 } */ *uap = v; 2075 struct proc *p = l->l_proc; 2076 struct ucred *cred; 2077 struct vnode *vp; 2078 int error, flags; 2079 struct nameidata nd; 2080 2081 cred = crdup(p->p_ucred); 2082 cred->cr_uid = p->p_cred->p_ruid; 2083 cred->cr_gid = p->p_cred->p_rgid; 2084 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2085 SCARG(uap, path), p); 2086 /* Override default credentials */ 2087 nd.ni_cnd.cn_cred = cred; 2088 if ((error = namei(&nd)) != 0) 2089 goto out; 2090 vp = nd.ni_vp; 2091 2092 /* Flags == 0 means only check for existence. */ 2093 if (SCARG(uap, flags)) { 2094 flags = 0; 2095 if (SCARG(uap, flags) & R_OK) 2096 flags |= VREAD; 2097 if (SCARG(uap, flags) & W_OK) 2098 flags |= VWRITE; 2099 if (SCARG(uap, flags) & X_OK) 2100 flags |= VEXEC; 2101 2102 error = VOP_ACCESS(vp, flags, cred, p); 2103 if (!error && (flags & VWRITE)) 2104 error = vn_writechk(vp); 2105 } 2106 vput(vp); 2107 out: 2108 crfree(cred); 2109 return (error); 2110 } 2111 2112 /* 2113 * Get file status; this version follows links. 2114 */ 2115 /* ARGSUSED */ 2116 int 2117 sys___stat30(struct lwp *l, void *v, register_t *retval) 2118 { 2119 struct sys___stat30_args /* { 2120 syscallarg(const char *) path; 2121 syscallarg(struct stat *) ub; 2122 } */ *uap = v; 2123 struct proc *p = l->l_proc; 2124 struct stat sb; 2125 int error; 2126 struct nameidata nd; 2127 2128 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2129 SCARG(uap, path), p); 2130 if ((error = namei(&nd)) != 0) 2131 return (error); 2132 error = vn_stat(nd.ni_vp, &sb, p); 2133 vput(nd.ni_vp); 2134 if (error) 2135 return (error); 2136 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2137 return (error); 2138 } 2139 2140 /* 2141 * Get file status; this version does not follow links. 2142 */ 2143 /* ARGSUSED */ 2144 int 2145 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2146 { 2147 struct sys___lstat30_args /* { 2148 syscallarg(const char *) path; 2149 syscallarg(struct stat *) ub; 2150 } */ *uap = v; 2151 struct proc *p = l->l_proc; 2152 struct stat sb; 2153 int error; 2154 struct nameidata nd; 2155 2156 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2157 SCARG(uap, path), p); 2158 if ((error = namei(&nd)) != 0) 2159 return (error); 2160 error = vn_stat(nd.ni_vp, &sb, p); 2161 vput(nd.ni_vp); 2162 if (error) 2163 return (error); 2164 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2165 return (error); 2166 } 2167 2168 /* 2169 * Get configurable pathname variables. 2170 */ 2171 /* ARGSUSED */ 2172 int 2173 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2174 { 2175 struct sys_pathconf_args /* { 2176 syscallarg(const char *) path; 2177 syscallarg(int) name; 2178 } */ *uap = v; 2179 struct proc *p = l->l_proc; 2180 int error; 2181 struct nameidata nd; 2182 2183 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2184 SCARG(uap, path), p); 2185 if ((error = namei(&nd)) != 0) 2186 return (error); 2187 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2188 vput(nd.ni_vp); 2189 return (error); 2190 } 2191 2192 /* 2193 * Return target name of a symbolic link. 2194 */ 2195 /* ARGSUSED */ 2196 int 2197 sys_readlink(struct lwp *l, void *v, register_t *retval) 2198 { 2199 struct sys_readlink_args /* { 2200 syscallarg(const char *) path; 2201 syscallarg(char *) buf; 2202 syscallarg(size_t) count; 2203 } */ *uap = v; 2204 struct proc *p = l->l_proc; 2205 struct vnode *vp; 2206 struct iovec aiov; 2207 struct uio auio; 2208 int error; 2209 struct nameidata nd; 2210 2211 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2212 SCARG(uap, path), p); 2213 if ((error = namei(&nd)) != 0) 2214 return (error); 2215 vp = nd.ni_vp; 2216 if (vp->v_type != VLNK) 2217 error = EINVAL; 2218 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2219 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 2220 aiov.iov_base = SCARG(uap, buf); 2221 aiov.iov_len = SCARG(uap, count); 2222 auio.uio_iov = &aiov; 2223 auio.uio_iovcnt = 1; 2224 auio.uio_offset = 0; 2225 auio.uio_rw = UIO_READ; 2226 auio.uio_segflg = UIO_USERSPACE; 2227 auio.uio_procp = p; 2228 auio.uio_resid = SCARG(uap, count); 2229 error = VOP_READLINK(vp, &auio, p->p_ucred); 2230 } 2231 vput(vp); 2232 *retval = SCARG(uap, count) - auio.uio_resid; 2233 return (error); 2234 } 2235 2236 /* 2237 * Change flags of a file given a path name. 2238 */ 2239 /* ARGSUSED */ 2240 int 2241 sys_chflags(struct lwp *l, void *v, register_t *retval) 2242 { 2243 struct sys_chflags_args /* { 2244 syscallarg(const char *) path; 2245 syscallarg(u_long) flags; 2246 } */ *uap = v; 2247 struct proc *p = l->l_proc; 2248 struct vnode *vp; 2249 int error; 2250 struct nameidata nd; 2251 2252 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2253 if ((error = namei(&nd)) != 0) 2254 return (error); 2255 vp = nd.ni_vp; 2256 error = change_flags(vp, SCARG(uap, flags), p); 2257 vput(vp); 2258 return (error); 2259 } 2260 2261 /* 2262 * Change flags of a file given a file descriptor. 2263 */ 2264 /* ARGSUSED */ 2265 int 2266 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2267 { 2268 struct sys_fchflags_args /* { 2269 syscallarg(int) fd; 2270 syscallarg(u_long) flags; 2271 } */ *uap = v; 2272 struct proc *p = l->l_proc; 2273 struct vnode *vp; 2274 struct file *fp; 2275 int error; 2276 2277 /* getvnode() will use the descriptor for us */ 2278 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2279 return (error); 2280 vp = (struct vnode *)fp->f_data; 2281 error = change_flags(vp, SCARG(uap, flags), p); 2282 VOP_UNLOCK(vp, 0); 2283 FILE_UNUSE(fp, p); 2284 return (error); 2285 } 2286 2287 /* 2288 * Change flags of a file given a path name; this version does 2289 * not follow links. 2290 */ 2291 int 2292 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2293 { 2294 struct sys_lchflags_args /* { 2295 syscallarg(const char *) path; 2296 syscallarg(u_long) flags; 2297 } */ *uap = v; 2298 struct proc *p = l->l_proc; 2299 struct vnode *vp; 2300 int error; 2301 struct nameidata nd; 2302 2303 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2304 if ((error = namei(&nd)) != 0) 2305 return (error); 2306 vp = nd.ni_vp; 2307 error = change_flags(vp, SCARG(uap, flags), p); 2308 vput(vp); 2309 return (error); 2310 } 2311 2312 /* 2313 * Common routine to change flags of a file. 2314 */ 2315 int 2316 change_flags(struct vnode *vp, u_long flags, struct proc *p) 2317 { 2318 struct mount *mp; 2319 struct vattr vattr; 2320 int error; 2321 2322 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2323 return (error); 2324 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2325 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2326 /* 2327 * Non-superusers cannot change the flags on devices, even if they 2328 * own them. 2329 */ 2330 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2331 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2332 goto out; 2333 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2334 error = EINVAL; 2335 goto out; 2336 } 2337 } 2338 VATTR_NULL(&vattr); 2339 vattr.va_flags = flags; 2340 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2341 out: 2342 vn_finished_write(mp, 0); 2343 return (error); 2344 } 2345 2346 /* 2347 * Change mode of a file given path name; this version follows links. 2348 */ 2349 /* ARGSUSED */ 2350 int 2351 sys_chmod(struct lwp *l, void *v, register_t *retval) 2352 { 2353 struct sys_chmod_args /* { 2354 syscallarg(const char *) path; 2355 syscallarg(int) mode; 2356 } */ *uap = v; 2357 struct proc *p = l->l_proc; 2358 int error; 2359 struct nameidata nd; 2360 2361 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2362 if ((error = namei(&nd)) != 0) 2363 return (error); 2364 2365 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2366 2367 vrele(nd.ni_vp); 2368 return (error); 2369 } 2370 2371 /* 2372 * Change mode of a file given a file descriptor. 2373 */ 2374 /* ARGSUSED */ 2375 int 2376 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2377 { 2378 struct sys_fchmod_args /* { 2379 syscallarg(int) fd; 2380 syscallarg(int) mode; 2381 } */ *uap = v; 2382 struct proc *p = l->l_proc; 2383 struct file *fp; 2384 int error; 2385 2386 /* getvnode() will use the descriptor for us */ 2387 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2388 return (error); 2389 2390 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 2391 FILE_UNUSE(fp, p); 2392 return (error); 2393 } 2394 2395 /* 2396 * Change mode of a file given path name; this version does not follow links. 2397 */ 2398 /* ARGSUSED */ 2399 int 2400 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2401 { 2402 struct sys_lchmod_args /* { 2403 syscallarg(const char *) path; 2404 syscallarg(int) mode; 2405 } */ *uap = v; 2406 struct proc *p = l->l_proc; 2407 int error; 2408 struct nameidata nd; 2409 2410 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2411 if ((error = namei(&nd)) != 0) 2412 return (error); 2413 2414 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2415 2416 vrele(nd.ni_vp); 2417 return (error); 2418 } 2419 2420 /* 2421 * Common routine to set mode given a vnode. 2422 */ 2423 static int 2424 change_mode(struct vnode *vp, int mode, struct proc *p) 2425 { 2426 struct mount *mp; 2427 struct vattr vattr; 2428 int error; 2429 2430 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2431 return (error); 2432 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2433 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2434 VATTR_NULL(&vattr); 2435 vattr.va_mode = mode & ALLPERMS; 2436 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2437 VOP_UNLOCK(vp, 0); 2438 vn_finished_write(mp, 0); 2439 return (error); 2440 } 2441 2442 /* 2443 * Set ownership given a path name; this version follows links. 2444 */ 2445 /* ARGSUSED */ 2446 int 2447 sys_chown(struct lwp *l, void *v, register_t *retval) 2448 { 2449 struct sys_chown_args /* { 2450 syscallarg(const char *) path; 2451 syscallarg(uid_t) uid; 2452 syscallarg(gid_t) gid; 2453 } */ *uap = v; 2454 struct proc *p = l->l_proc; 2455 int error; 2456 struct nameidata nd; 2457 2458 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2459 if ((error = namei(&nd)) != 0) 2460 return (error); 2461 2462 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2463 2464 vrele(nd.ni_vp); 2465 return (error); 2466 } 2467 2468 /* 2469 * Set ownership given a path name; this version follows links. 2470 * Provides POSIX semantics. 2471 */ 2472 /* ARGSUSED */ 2473 int 2474 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2475 { 2476 struct sys_chown_args /* { 2477 syscallarg(const char *) path; 2478 syscallarg(uid_t) uid; 2479 syscallarg(gid_t) gid; 2480 } */ *uap = v; 2481 struct proc *p = l->l_proc; 2482 int error; 2483 struct nameidata nd; 2484 2485 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2486 if ((error = namei(&nd)) != 0) 2487 return (error); 2488 2489 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2490 2491 vrele(nd.ni_vp); 2492 return (error); 2493 } 2494 2495 /* 2496 * Set ownership given a file descriptor. 2497 */ 2498 /* ARGSUSED */ 2499 int 2500 sys_fchown(struct lwp *l, void *v, register_t *retval) 2501 { 2502 struct sys_fchown_args /* { 2503 syscallarg(int) fd; 2504 syscallarg(uid_t) uid; 2505 syscallarg(gid_t) gid; 2506 } */ *uap = v; 2507 struct proc *p = l->l_proc; 2508 int error; 2509 struct file *fp; 2510 2511 /* getvnode() will use the descriptor for us */ 2512 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2513 return (error); 2514 2515 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2516 SCARG(uap, gid), p, 0); 2517 FILE_UNUSE(fp, p); 2518 return (error); 2519 } 2520 2521 /* 2522 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2523 */ 2524 /* ARGSUSED */ 2525 int 2526 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2527 { 2528 struct sys_fchown_args /* { 2529 syscallarg(int) fd; 2530 syscallarg(uid_t) uid; 2531 syscallarg(gid_t) gid; 2532 } */ *uap = v; 2533 struct proc *p = l->l_proc; 2534 int error; 2535 struct file *fp; 2536 2537 /* getvnode() will use the descriptor for us */ 2538 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2539 return (error); 2540 2541 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2542 SCARG(uap, gid), p, 1); 2543 FILE_UNUSE(fp, p); 2544 return (error); 2545 } 2546 2547 /* 2548 * Set ownership given a path name; this version does not follow links. 2549 */ 2550 /* ARGSUSED */ 2551 int 2552 sys_lchown(struct lwp *l, void *v, register_t *retval) 2553 { 2554 struct sys_lchown_args /* { 2555 syscallarg(const char *) path; 2556 syscallarg(uid_t) uid; 2557 syscallarg(gid_t) gid; 2558 } */ *uap = v; 2559 struct proc *p = l->l_proc; 2560 int error; 2561 struct nameidata nd; 2562 2563 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2564 if ((error = namei(&nd)) != 0) 2565 return (error); 2566 2567 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2568 2569 vrele(nd.ni_vp); 2570 return (error); 2571 } 2572 2573 /* 2574 * Set ownership given a path name; this version does not follow links. 2575 * Provides POSIX/XPG semantics. 2576 */ 2577 /* ARGSUSED */ 2578 int 2579 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2580 { 2581 struct sys_lchown_args /* { 2582 syscallarg(const char *) path; 2583 syscallarg(uid_t) uid; 2584 syscallarg(gid_t) gid; 2585 } */ *uap = v; 2586 struct proc *p = l->l_proc; 2587 int error; 2588 struct nameidata nd; 2589 2590 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2591 if ((error = namei(&nd)) != 0) 2592 return (error); 2593 2594 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2595 2596 vrele(nd.ni_vp); 2597 return (error); 2598 } 2599 2600 /* 2601 * Common routine to set ownership given a vnode. 2602 */ 2603 static int 2604 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct proc *p, 2605 int posix_semantics) 2606 { 2607 struct mount *mp; 2608 struct vattr vattr; 2609 mode_t newmode; 2610 int error; 2611 2612 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2613 return (error); 2614 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2615 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2616 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2617 goto out; 2618 2619 #define CHANGED(x) ((int)(x) != -1) 2620 newmode = vattr.va_mode; 2621 if (posix_semantics) { 2622 /* 2623 * POSIX/XPG semantics: if the caller is not the super-user, 2624 * clear set-user-id and set-group-id bits. Both POSIX and 2625 * the XPG consider the behaviour for calls by the super-user 2626 * implementation-defined; we leave the set-user-id and set- 2627 * group-id settings intact in that case. 2628 */ 2629 if (suser(p->p_ucred, NULL) != 0) 2630 newmode &= ~(S_ISUID | S_ISGID); 2631 } else { 2632 /* 2633 * NetBSD semantics: when changing owner and/or group, 2634 * clear the respective bit(s). 2635 */ 2636 if (CHANGED(uid)) 2637 newmode &= ~S_ISUID; 2638 if (CHANGED(gid)) 2639 newmode &= ~S_ISGID; 2640 } 2641 /* Update va_mode iff altered. */ 2642 if (vattr.va_mode == newmode) 2643 newmode = VNOVAL; 2644 2645 VATTR_NULL(&vattr); 2646 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2647 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2648 vattr.va_mode = newmode; 2649 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2650 #undef CHANGED 2651 2652 out: 2653 VOP_UNLOCK(vp, 0); 2654 vn_finished_write(mp, 0); 2655 return (error); 2656 } 2657 2658 /* 2659 * Set the access and modification times given a path name; this 2660 * version follows links. 2661 */ 2662 /* ARGSUSED */ 2663 int 2664 sys_utimes(struct lwp *l, void *v, register_t *retval) 2665 { 2666 struct sys_utimes_args /* { 2667 syscallarg(const char *) path; 2668 syscallarg(const struct timeval *) tptr; 2669 } */ *uap = v; 2670 struct proc *p = l->l_proc; 2671 int error; 2672 struct nameidata nd; 2673 2674 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2675 if ((error = namei(&nd)) != 0) 2676 return (error); 2677 2678 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2679 2680 vrele(nd.ni_vp); 2681 return (error); 2682 } 2683 2684 /* 2685 * Set the access and modification times given a file descriptor. 2686 */ 2687 /* ARGSUSED */ 2688 int 2689 sys_futimes(struct lwp *l, void *v, register_t *retval) 2690 { 2691 struct sys_futimes_args /* { 2692 syscallarg(int) fd; 2693 syscallarg(const struct timeval *) tptr; 2694 } */ *uap = v; 2695 struct proc *p = l->l_proc; 2696 int error; 2697 struct file *fp; 2698 2699 /* getvnode() will use the descriptor for us */ 2700 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2701 return (error); 2702 2703 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2704 FILE_UNUSE(fp, p); 2705 return (error); 2706 } 2707 2708 /* 2709 * Set the access and modification times given a path name; this 2710 * version does not follow links. 2711 */ 2712 /* ARGSUSED */ 2713 int 2714 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2715 { 2716 struct sys_lutimes_args /* { 2717 syscallarg(const char *) path; 2718 syscallarg(const struct timeval *) tptr; 2719 } */ *uap = v; 2720 struct proc *p = l->l_proc; 2721 int error; 2722 struct nameidata nd; 2723 2724 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2725 if ((error = namei(&nd)) != 0) 2726 return (error); 2727 2728 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2729 2730 vrele(nd.ni_vp); 2731 return (error); 2732 } 2733 2734 /* 2735 * Common routine to set access and modification times given a vnode. 2736 */ 2737 static int 2738 change_utimes(struct vnode *vp, const struct timeval *tptr, struct proc *p) 2739 { 2740 struct timeval tv[2]; 2741 struct mount *mp; 2742 struct vattr vattr; 2743 int error; 2744 2745 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2746 return (error); 2747 VATTR_NULL(&vattr); 2748 if (tptr == NULL) { 2749 microtime(&tv[0]); 2750 tv[1] = tv[0]; 2751 vattr.va_vaflags |= VA_UTIMES_NULL; 2752 } else { 2753 error = copyin(tptr, tv, sizeof(tv)); 2754 if (error) 2755 goto out; 2756 } 2757 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2758 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2759 vattr.va_atime.tv_sec = tv[0].tv_sec; 2760 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2761 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2762 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2763 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2764 VOP_UNLOCK(vp, 0); 2765 out: 2766 vn_finished_write(mp, 0); 2767 return (error); 2768 } 2769 2770 /* 2771 * Truncate a file given its path name. 2772 */ 2773 /* ARGSUSED */ 2774 int 2775 sys_truncate(struct lwp *l, void *v, register_t *retval) 2776 { 2777 struct sys_truncate_args /* { 2778 syscallarg(const char *) path; 2779 syscallarg(int) pad; 2780 syscallarg(off_t) length; 2781 } */ *uap = v; 2782 struct proc *p = l->l_proc; 2783 struct vnode *vp; 2784 struct mount *mp; 2785 struct vattr vattr; 2786 int error; 2787 struct nameidata nd; 2788 2789 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2790 if ((error = namei(&nd)) != 0) 2791 return (error); 2792 vp = nd.ni_vp; 2793 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2794 vrele(vp); 2795 return (error); 2796 } 2797 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2798 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2799 if (vp->v_type == VDIR) 2800 error = EISDIR; 2801 else if ((error = vn_writechk(vp)) == 0 && 2802 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2803 VATTR_NULL(&vattr); 2804 vattr.va_size = SCARG(uap, length); 2805 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2806 } 2807 vput(vp); 2808 vn_finished_write(mp, 0); 2809 return (error); 2810 } 2811 2812 /* 2813 * Truncate a file given a file descriptor. 2814 */ 2815 /* ARGSUSED */ 2816 int 2817 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2818 { 2819 struct sys_ftruncate_args /* { 2820 syscallarg(int) fd; 2821 syscallarg(int) pad; 2822 syscallarg(off_t) length; 2823 } */ *uap = v; 2824 struct proc *p = l->l_proc; 2825 struct mount *mp; 2826 struct vattr vattr; 2827 struct vnode *vp; 2828 struct file *fp; 2829 int error; 2830 2831 /* getvnode() will use the descriptor for us */ 2832 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2833 return (error); 2834 if ((fp->f_flag & FWRITE) == 0) { 2835 error = EINVAL; 2836 goto out; 2837 } 2838 vp = (struct vnode *)fp->f_data; 2839 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2840 FILE_UNUSE(fp, p); 2841 return (error); 2842 } 2843 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2844 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2845 if (vp->v_type == VDIR) 2846 error = EISDIR; 2847 else if ((error = vn_writechk(vp)) == 0) { 2848 VATTR_NULL(&vattr); 2849 vattr.va_size = SCARG(uap, length); 2850 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2851 } 2852 VOP_UNLOCK(vp, 0); 2853 vn_finished_write(mp, 0); 2854 out: 2855 FILE_UNUSE(fp, p); 2856 return (error); 2857 } 2858 2859 /* 2860 * Sync an open file. 2861 */ 2862 /* ARGSUSED */ 2863 int 2864 sys_fsync(struct lwp *l, void *v, register_t *retval) 2865 { 2866 struct sys_fsync_args /* { 2867 syscallarg(int) fd; 2868 } */ *uap = v; 2869 struct proc *p = l->l_proc; 2870 struct vnode *vp; 2871 struct mount *mp; 2872 struct file *fp; 2873 int error; 2874 2875 /* getvnode() will use the descriptor for us */ 2876 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2877 return (error); 2878 vp = (struct vnode *)fp->f_data; 2879 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2880 FILE_UNUSE(fp, p); 2881 return (error); 2882 } 2883 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2884 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p); 2885 if (error == 0 && bioops.io_fsync != NULL && 2886 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2887 (*bioops.io_fsync)(vp, 0); 2888 VOP_UNLOCK(vp, 0); 2889 vn_finished_write(mp, 0); 2890 FILE_UNUSE(fp, p); 2891 return (error); 2892 } 2893 2894 /* 2895 * Sync a range of file data. API modeled after that found in AIX. 2896 * 2897 * FDATASYNC indicates that we need only save enough metadata to be able 2898 * to re-read the written data. Note we duplicate AIX's requirement that 2899 * the file be open for writing. 2900 */ 2901 /* ARGSUSED */ 2902 int 2903 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2904 { 2905 struct sys_fsync_range_args /* { 2906 syscallarg(int) fd; 2907 syscallarg(int) flags; 2908 syscallarg(off_t) start; 2909 syscallarg(off_t) length; 2910 } */ *uap = v; 2911 struct proc *p = l->l_proc; 2912 struct vnode *vp; 2913 struct file *fp; 2914 int flags, nflags; 2915 off_t s, e, len; 2916 int error; 2917 2918 /* getvnode() will use the descriptor for us */ 2919 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2920 return (error); 2921 2922 if ((fp->f_flag & FWRITE) == 0) { 2923 FILE_UNUSE(fp, p); 2924 return (EBADF); 2925 } 2926 2927 flags = SCARG(uap, flags); 2928 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2929 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2930 return (EINVAL); 2931 } 2932 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2933 if (flags & FDATASYNC) 2934 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2935 else 2936 nflags = FSYNC_WAIT; 2937 if (flags & FDISKSYNC) 2938 nflags |= FSYNC_CACHE; 2939 2940 len = SCARG(uap, length); 2941 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2942 if (len) { 2943 s = SCARG(uap, start); 2944 e = s + len; 2945 if (e < s) { 2946 FILE_UNUSE(fp, p); 2947 return (EINVAL); 2948 } 2949 } else { 2950 e = 0; 2951 s = 0; 2952 } 2953 2954 vp = (struct vnode *)fp->f_data; 2955 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2956 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p); 2957 2958 if (error == 0 && bioops.io_fsync != NULL && 2959 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2960 (*bioops.io_fsync)(vp, nflags); 2961 2962 VOP_UNLOCK(vp, 0); 2963 FILE_UNUSE(fp, p); 2964 return (error); 2965 } 2966 2967 /* 2968 * Sync the data of an open file. 2969 */ 2970 /* ARGSUSED */ 2971 int 2972 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 2973 { 2974 struct sys_fdatasync_args /* { 2975 syscallarg(int) fd; 2976 } */ *uap = v; 2977 struct proc *p = l->l_proc; 2978 struct vnode *vp; 2979 struct file *fp; 2980 int error; 2981 2982 /* getvnode() will use the descriptor for us */ 2983 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2984 return (error); 2985 if ((fp->f_flag & FWRITE) == 0) { 2986 FILE_UNUSE(fp, p); 2987 return (EBADF); 2988 } 2989 vp = (struct vnode *)fp->f_data; 2990 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2991 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p); 2992 VOP_UNLOCK(vp, 0); 2993 FILE_UNUSE(fp, p); 2994 return (error); 2995 } 2996 2997 /* 2998 * Rename files, (standard) BSD semantics frontend. 2999 */ 3000 /* ARGSUSED */ 3001 int 3002 sys_rename(struct lwp *l, void *v, register_t *retval) 3003 { 3004 struct sys_rename_args /* { 3005 syscallarg(const char *) from; 3006 syscallarg(const char *) to; 3007 } */ *uap = v; 3008 struct proc *p = l->l_proc; 3009 3010 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 3011 } 3012 3013 /* 3014 * Rename files, POSIX semantics frontend. 3015 */ 3016 /* ARGSUSED */ 3017 int 3018 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3019 { 3020 struct sys___posix_rename_args /* { 3021 syscallarg(const char *) from; 3022 syscallarg(const char *) to; 3023 } */ *uap = v; 3024 struct proc *p = l->l_proc; 3025 3026 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 3027 } 3028 3029 /* 3030 * Rename files. Source and destination must either both be directories, 3031 * or both not be directories. If target is a directory, it must be empty. 3032 * If `from' and `to' refer to the same object, the value of the `retain' 3033 * argument is used to determine whether `from' will be 3034 * 3035 * (retain == 0) deleted unless `from' and `to' refer to the same 3036 * object in the file system's name space (BSD). 3037 * (retain == 1) always retained (POSIX). 3038 */ 3039 static int 3040 rename_files(const char *from, const char *to, struct proc *p, int retain) 3041 { 3042 struct mount *mp = NULL; 3043 struct vnode *tvp, *fvp, *tdvp; 3044 struct nameidata fromnd, tond; 3045 int error; 3046 3047 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3048 from, p); 3049 if ((error = namei(&fromnd)) != 0) 3050 return (error); 3051 fvp = fromnd.ni_vp; 3052 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3053 if (error != 0) { 3054 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3055 vrele(fromnd.ni_dvp); 3056 vrele(fvp); 3057 if (fromnd.ni_startdir) 3058 vrele(fromnd.ni_startdir); 3059 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3060 return (error); 3061 } 3062 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3063 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p); 3064 if ((error = namei(&tond)) != 0) { 3065 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3066 vrele(fromnd.ni_dvp); 3067 vrele(fvp); 3068 goto out1; 3069 } 3070 tdvp = tond.ni_dvp; 3071 tvp = tond.ni_vp; 3072 3073 if (tvp != NULL) { 3074 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3075 error = ENOTDIR; 3076 goto out; 3077 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3078 error = EISDIR; 3079 goto out; 3080 } 3081 } 3082 3083 if (fvp == tdvp) 3084 error = EINVAL; 3085 3086 /* 3087 * Source and destination refer to the same object. 3088 */ 3089 if (fvp == tvp) { 3090 if (retain) 3091 error = -1; 3092 else if (fromnd.ni_dvp == tdvp && 3093 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3094 !memcmp(fromnd.ni_cnd.cn_nameptr, 3095 tond.ni_cnd.cn_nameptr, 3096 fromnd.ni_cnd.cn_namelen)) 3097 error = -1; 3098 } 3099 3100 #ifdef VERIFIED_EXEC 3101 if (!error) 3102 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp); 3103 #endif /* VERIFIED_EXEC */ 3104 3105 out: 3106 if (!error) { 3107 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 3108 if (fromnd.ni_dvp != tdvp) 3109 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3110 if (tvp) { 3111 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 3112 } 3113 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3114 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3115 } else { 3116 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3117 if (tdvp == tvp) 3118 vrele(tdvp); 3119 else 3120 vput(tdvp); 3121 if (tvp) 3122 vput(tvp); 3123 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3124 vrele(fromnd.ni_dvp); 3125 vrele(fvp); 3126 } 3127 vrele(tond.ni_startdir); 3128 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3129 out1: 3130 vn_finished_write(mp, 0); 3131 if (fromnd.ni_startdir) 3132 vrele(fromnd.ni_startdir); 3133 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3134 return (error == -1 ? 0 : error); 3135 } 3136 3137 /* 3138 * Make a directory file. 3139 */ 3140 /* ARGSUSED */ 3141 int 3142 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3143 { 3144 struct sys_mkdir_args /* { 3145 syscallarg(const char *) path; 3146 syscallarg(int) mode; 3147 } */ *uap = v; 3148 struct proc *p = l->l_proc; 3149 struct mount *mp; 3150 struct vnode *vp; 3151 struct vattr vattr; 3152 int error; 3153 struct nameidata nd; 3154 3155 restart: 3156 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3157 SCARG(uap, path), p); 3158 if ((error = namei(&nd)) != 0) 3159 return (error); 3160 vp = nd.ni_vp; 3161 if (vp != NULL) { 3162 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3163 if (nd.ni_dvp == vp) 3164 vrele(nd.ni_dvp); 3165 else 3166 vput(nd.ni_dvp); 3167 vrele(vp); 3168 return (EEXIST); 3169 } 3170 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3171 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3172 if (nd.ni_dvp == vp) 3173 vrele(nd.ni_dvp); 3174 else 3175 vput(nd.ni_dvp); 3176 if ((error = vn_start_write(NULL, &mp, 3177 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3178 return (error); 3179 goto restart; 3180 } 3181 VATTR_NULL(&vattr); 3182 vattr.va_type = VDIR; 3183 vattr.va_mode = 3184 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3185 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3186 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3187 if (!error) 3188 vput(nd.ni_vp); 3189 vn_finished_write(mp, 0); 3190 return (error); 3191 } 3192 3193 /* 3194 * Remove a directory file. 3195 */ 3196 /* ARGSUSED */ 3197 int 3198 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3199 { 3200 struct sys_rmdir_args /* { 3201 syscallarg(const char *) path; 3202 } */ *uap = v; 3203 struct proc *p = l->l_proc; 3204 struct mount *mp; 3205 struct vnode *vp; 3206 int error; 3207 struct nameidata nd; 3208 3209 restart: 3210 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3211 SCARG(uap, path), p); 3212 if ((error = namei(&nd)) != 0) 3213 return (error); 3214 vp = nd.ni_vp; 3215 if (vp->v_type != VDIR) { 3216 error = ENOTDIR; 3217 goto out; 3218 } 3219 /* 3220 * No rmdir "." please. 3221 */ 3222 if (nd.ni_dvp == vp) { 3223 error = EINVAL; 3224 goto out; 3225 } 3226 /* 3227 * The root of a mounted filesystem cannot be deleted. 3228 */ 3229 if (vp->v_flag & VROOT) { 3230 error = EBUSY; 3231 goto out; 3232 } 3233 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3234 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3235 if (nd.ni_dvp == vp) 3236 vrele(nd.ni_dvp); 3237 else 3238 vput(nd.ni_dvp); 3239 vput(vp); 3240 if ((error = vn_start_write(NULL, &mp, 3241 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3242 return (error); 3243 goto restart; 3244 } 3245 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3246 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3247 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3248 vn_finished_write(mp, 0); 3249 return (error); 3250 3251 out: 3252 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3253 if (nd.ni_dvp == vp) 3254 vrele(nd.ni_dvp); 3255 else 3256 vput(nd.ni_dvp); 3257 vput(vp); 3258 return (error); 3259 } 3260 3261 /* 3262 * Read a block of directory entries in a file system independent format. 3263 */ 3264 int 3265 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3266 { 3267 struct sys___getdents30_args /* { 3268 syscallarg(int) fd; 3269 syscallarg(char *) buf; 3270 syscallarg(size_t) count; 3271 } */ *uap = v; 3272 struct proc *p = l->l_proc; 3273 struct file *fp; 3274 int error, done; 3275 3276 /* getvnode() will use the descriptor for us */ 3277 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3278 return (error); 3279 if ((fp->f_flag & FREAD) == 0) { 3280 error = EBADF; 3281 goto out; 3282 } 3283 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3284 SCARG(uap, count), &done, p, 0, 0); 3285 #ifdef KTRACE 3286 if (!error && KTRPOINT(p, KTR_GENIO)) { 3287 struct iovec iov; 3288 iov.iov_base = SCARG(uap, buf); 3289 iov.iov_len = done; 3290 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3291 } 3292 #endif 3293 *retval = done; 3294 out: 3295 FILE_UNUSE(fp, p); 3296 return (error); 3297 } 3298 3299 /* 3300 * Set the mode mask for creation of filesystem nodes. 3301 */ 3302 int 3303 sys_umask(struct lwp *l, void *v, register_t *retval) 3304 { 3305 struct sys_umask_args /* { 3306 syscallarg(mode_t) newmask; 3307 } */ *uap = v; 3308 struct proc *p = l->l_proc; 3309 struct cwdinfo *cwdi; 3310 3311 cwdi = p->p_cwdi; 3312 *retval = cwdi->cwdi_cmask; 3313 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3314 return (0); 3315 } 3316 3317 /* 3318 * Void all references to file by ripping underlying filesystem 3319 * away from vnode. 3320 */ 3321 /* ARGSUSED */ 3322 int 3323 sys_revoke(struct lwp *l, void *v, register_t *retval) 3324 { 3325 struct sys_revoke_args /* { 3326 syscallarg(const char *) path; 3327 } */ *uap = v; 3328 struct proc *p = l->l_proc; 3329 struct mount *mp; 3330 struct vnode *vp; 3331 struct vattr vattr; 3332 int error; 3333 struct nameidata nd; 3334 3335 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3336 if ((error = namei(&nd)) != 0) 3337 return (error); 3338 vp = nd.ni_vp; 3339 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 3340 goto out; 3341 if (p->p_ucred->cr_uid != vattr.va_uid && 3342 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3343 goto out; 3344 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3345 goto out; 3346 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3347 VOP_REVOKE(vp, REVOKEALL); 3348 vn_finished_write(mp, 0); 3349 out: 3350 vrele(vp); 3351 return (error); 3352 } 3353 3354 /* 3355 * Convert a user file descriptor to a kernel file entry. 3356 */ 3357 int 3358 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3359 { 3360 struct vnode *vp; 3361 struct file *fp; 3362 3363 if ((fp = fd_getfile(fdp, fd)) == NULL) 3364 return (EBADF); 3365 3366 FILE_USE(fp); 3367 3368 if (fp->f_type != DTYPE_VNODE) { 3369 FILE_UNUSE(fp, NULL); 3370 return (EINVAL); 3371 } 3372 3373 vp = (struct vnode *)fp->f_data; 3374 if (vp->v_type == VBAD) { 3375 FILE_UNUSE(fp, NULL); 3376 return (EBADF); 3377 } 3378 3379 *fpp = fp; 3380 return (0); 3381 } 3382