vfs_syscalls.c revision 1.190.2.5
1 /* $NetBSD: vfs_syscalls.c,v 1.190.2.5 2004/09/21 13:35:17 skrll 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 1.190.2.5 2004/09/21 13:35:17 skrll Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_ktrace.h" 45 #include "fss.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/namei.h> 50 #include <sys/filedesc.h> 51 #include <sys/kernel.h> 52 #include <sys/file.h> 53 #include <sys/stat.h> 54 #include <sys/vnode.h> 55 #include <sys/mount.h> 56 #include <sys/proc.h> 57 #include <sys/uio.h> 58 #include <sys/malloc.h> 59 #include <sys/dirent.h> 60 #include <sys/sysctl.h> 61 #include <sys/sa.h> 62 #include <sys/syscallargs.h> 63 #ifdef KTRACE 64 #include <sys/ktrace.h> 65 #endif 66 67 #include <miscfs/genfs/genfs.h> 68 #include <miscfs/syncfs/syncfs.h> 69 70 #if NFSS > 0 71 #include <dev/fssvar.h> 72 #endif 73 74 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 75 76 static int change_dir(struct nameidata *, struct lwp *); 77 static int change_flags(struct vnode *, u_long, struct lwp *); 78 static int change_mode(struct vnode *, int, struct lwp *l); 79 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 80 static int change_utimes(struct vnode *vp, const struct timeval *, 81 struct lwp *l); 82 static int rename_files(const char *, const char *, struct lwp *, int); 83 84 void checkdirs(struct vnode *); 85 86 int dovfsusermount = 0; 87 88 /* 89 * Virtual File System System Calls 90 */ 91 92 /* 93 * Mount a file system. 94 */ 95 96 #if defined(COMPAT_09) || defined(COMPAT_43) 97 /* 98 * This table is used to maintain compatibility with 4.3BSD 99 * and NetBSD 0.9 mount syscalls. Note, the order is important! 100 * 101 * Do not modify this table. It should only contain filesystems 102 * supported by NetBSD 0.9 and 4.3BSD. 103 */ 104 const char * const mountcompatnames[] = { 105 NULL, /* 0 = MOUNT_NONE */ 106 MOUNT_FFS, /* 1 = MOUNT_UFS */ 107 MOUNT_NFS, /* 2 */ 108 MOUNT_MFS, /* 3 */ 109 MOUNT_MSDOS, /* 4 */ 110 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 111 MOUNT_FDESC, /* 6 */ 112 MOUNT_KERNFS, /* 7 */ 113 NULL, /* 8 = MOUNT_DEVFS */ 114 MOUNT_AFS, /* 9 */ 115 }; 116 const int nmountcompatnames = sizeof(mountcompatnames) / 117 sizeof(mountcompatnames[0]); 118 #endif /* COMPAT_09 || COMPAT_43 */ 119 120 /* ARGSUSED */ 121 int 122 sys_mount(l, v, retval) 123 struct lwp *l; 124 void *v; 125 register_t *retval; 126 { 127 struct sys_mount_args /* { 128 syscallarg(const char *) type; 129 syscallarg(const char *) path; 130 syscallarg(int) flags; 131 syscallarg(void *) data; 132 } */ *uap = v; 133 struct proc *p = l->l_proc; 134 struct vnode *vp; 135 struct mount *mp; 136 int error, flag = 0; 137 char fstypename[MFSNAMELEN]; 138 struct vattr va; 139 struct nameidata nd; 140 struct vfsops *vfs; 141 142 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 143 (error = suser(p->p_ucred, &p->p_acflag))) 144 return (error); 145 /* 146 * Get vnode to be covered 147 */ 148 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 149 SCARG(uap, path), l); 150 if ((error = namei(&nd)) != 0) 151 return (error); 152 vp = nd.ni_vp; 153 /* 154 * A lookup in VFS_MOUNT might result in an attempt to 155 * lock this vnode again, so make the lock recursive. 156 */ 157 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 158 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 159 if ((vp->v_flag & VROOT) == 0) { 160 vput(vp); 161 return (EINVAL); 162 } 163 mp = vp->v_mount; 164 flag = mp->mnt_flag; 165 vfs = mp->mnt_op; 166 /* 167 * We only allow the filesystem to be reloaded if it 168 * is currently mounted read-only. 169 */ 170 if ((SCARG(uap, flags) & MNT_RELOAD) && 171 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 172 vput(vp); 173 return (EOPNOTSUPP); /* Needs translation */ 174 } 175 /* 176 * In "highly secure" mode, don't let the caller do anything 177 * but downgrade a filesystem from read-write to read-only. 178 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 179 */ 180 if (securelevel >= 2 && 181 SCARG(uap, flags) != MNT_GETARGS && 182 SCARG(uap, flags) != 183 (mp->mnt_flag | MNT_RDONLY | 184 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 185 vput(vp); 186 return (EPERM); 187 } 188 mp->mnt_flag |= SCARG(uap, flags) & 189 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 190 /* 191 * Only root, or the user that did the original mount is 192 * permitted to update it. 193 */ 194 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 195 mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 196 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 197 vput(vp); 198 return (error); 199 } 200 /* 201 * Do not allow NFS export by non-root users. For non-root 202 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 203 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 204 */ 205 if (p->p_ucred->cr_uid != 0) { 206 if (SCARG(uap, flags) & MNT_EXPORTED) { 207 vput(vp); 208 return (EPERM); 209 } 210 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 211 if (flag & MNT_NOEXEC) 212 SCARG(uap, flags) |= MNT_NOEXEC; 213 } 214 if (vfs_busy(mp, LK_NOWAIT, 0)) { 215 vput(vp); 216 return (EPERM); 217 } 218 goto update; 219 } else { 220 if (securelevel >= 2) { 221 vput(vp); 222 return (EPERM); 223 } 224 } 225 /* 226 * If the user is not root, ensure that they own the directory 227 * onto which we are attempting to mount. 228 */ 229 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l)) != 0 || 230 (va.va_uid != p->p_ucred->cr_uid && 231 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) { 232 vput(vp); 233 return (error); 234 } 235 /* 236 * Do not allow NFS export by non-root users. For non-root users, 237 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 238 * mount point is already MNT_NOEXEC. 239 */ 240 if (p->p_ucred->cr_uid != 0) { 241 if (SCARG(uap, flags) & MNT_EXPORTED) { 242 vput(vp); 243 return (EPERM); 244 } 245 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 246 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 247 SCARG(uap, flags) |= MNT_NOEXEC; 248 } 249 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, l, 0, 0)) != 0) { 250 vput(vp); 251 return (error); 252 } 253 if (vp->v_type != VDIR) { 254 vput(vp); 255 return (ENOTDIR); 256 } 257 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 258 if (error) { 259 #if defined(COMPAT_09) || defined(COMPAT_43) 260 /* 261 * Historically filesystem types were identified by number. 262 * If we get an integer for the filesystem type instead of a 263 * string, we check to see if it matches one of the historic 264 * filesystem types. 265 */ 266 u_long fsindex = (u_long)SCARG(uap, type); 267 if (fsindex >= nmountcompatnames || 268 mountcompatnames[fsindex] == NULL) { 269 vput(vp); 270 return (ENODEV); 271 } 272 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 273 #else 274 vput(vp); 275 return (error); 276 #endif 277 } 278 #ifdef COMPAT_10 279 /* Accept `ufs' as an alias for `ffs'. */ 280 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 281 strncpy(fstypename, "ffs", MFSNAMELEN); 282 #endif 283 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 284 vput(vp); 285 return (ENODEV); 286 } 287 if (vp->v_mountedhere != NULL) { 288 vput(vp); 289 return (EBUSY); 290 } 291 292 /* 293 * Allocate and initialize the file system. 294 */ 295 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 296 M_MOUNT, M_WAITOK); 297 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 298 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 299 simple_lock_init(&mp->mnt_slock); 300 (void)vfs_busy(mp, LK_NOWAIT, 0); 301 mp->mnt_op = vfs; 302 vfs->vfs_refcount++; 303 mp->mnt_vnodecovered = vp; 304 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 305 mp->mnt_unmounter = NULL; 306 mp->mnt_leaf = mp; 307 308 /* 309 * The underlying file system may refuse the mount for 310 * various reasons. Allow the user to force it to happen. 311 */ 312 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 313 update: 314 /* 315 * Set the mount level flags. 316 */ 317 if (SCARG(uap, flags) & MNT_RDONLY) 318 mp->mnt_flag |= MNT_RDONLY; 319 else if (mp->mnt_flag & MNT_RDONLY) 320 mp->mnt_iflag |= IMNT_WANTRDWR; 321 mp->mnt_flag &= 322 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 323 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 324 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 325 mp->mnt_flag |= SCARG(uap, flags) & 326 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 327 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 328 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 329 MNT_IGNORE); 330 /* 331 * Mount the filesystem. 332 */ 333 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 334 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 335 if (mp->mnt_iflag & IMNT_WANTRDWR) 336 mp->mnt_flag &= ~MNT_RDONLY; 337 if (error || (mp->mnt_flag & MNT_GETARGS)) 338 mp->mnt_flag = flag; 339 mp->mnt_flag &=~ 340 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 341 mp->mnt_iflag &=~ IMNT_WANTRDWR; 342 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 343 if (mp->mnt_syncer == NULL) 344 error = vfs_allocate_syncvnode(mp); 345 } else { 346 if (mp->mnt_syncer != NULL) 347 vfs_deallocate_syncvnode(mp); 348 } 349 vfs_unbusy(mp); 350 VOP_UNLOCK(vp, 0); 351 vrele(vp); 352 return (error); 353 } 354 /* 355 * Put the new filesystem on the mount list after root. 356 */ 357 cache_purge(vp); 358 if (!error) { 359 mp->mnt_flag &=~ 360 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 361 mp->mnt_iflag &=~ IMNT_WANTRDWR; 362 vp->v_mountedhere = mp; 363 simple_lock(&mountlist_slock); 364 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 365 simple_unlock(&mountlist_slock); 366 checkdirs(vp); 367 VOP_UNLOCK(vp, 0); 368 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 369 error = vfs_allocate_syncvnode(mp); 370 vfs_unbusy(mp); 371 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 372 if ((error = VFS_START(mp, 0, l))) 373 vrele(vp); 374 } else { 375 vp->v_mountedhere = (struct mount *)0; 376 vfs->vfs_refcount--; 377 vfs_unbusy(mp); 378 free(mp, M_MOUNT); 379 vput(vp); 380 } 381 return (error); 382 } 383 384 /* 385 * Scan all active processes to see if any of them have a current 386 * or root directory onto which the new filesystem has just been 387 * mounted. If so, replace them with the new mount point. 388 */ 389 void 390 checkdirs(olddp) 391 struct vnode *olddp; 392 { 393 struct cwdinfo *cwdi; 394 struct vnode *newdp; 395 struct proc *p; 396 397 if (olddp->v_usecount == 1) 398 return; 399 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 400 panic("mount: lost mount"); 401 proclist_lock_read(); 402 LIST_FOREACH(p, &allproc, p_list) { 403 cwdi = p->p_cwdi; 404 if (cwdi->cwdi_cdir == olddp) { 405 vrele(cwdi->cwdi_cdir); 406 VREF(newdp); 407 cwdi->cwdi_cdir = newdp; 408 } 409 if (cwdi->cwdi_rdir == olddp) { 410 vrele(cwdi->cwdi_rdir); 411 VREF(newdp); 412 cwdi->cwdi_rdir = newdp; 413 } 414 } 415 proclist_unlock_read(); 416 if (rootvnode == olddp) { 417 vrele(rootvnode); 418 VREF(newdp); 419 rootvnode = newdp; 420 } 421 vput(newdp); 422 } 423 424 /* 425 * Unmount a file system. 426 * 427 * Note: unmount takes a path to the vnode mounted on as argument, 428 * not special file (as before). 429 */ 430 /* ARGSUSED */ 431 int 432 sys_unmount(l, v, retval) 433 struct lwp *l; 434 void *v; 435 register_t *retval; 436 { 437 struct sys_unmount_args /* { 438 syscallarg(const char *) path; 439 syscallarg(int) flags; 440 } */ *uap = v; 441 struct proc *p = l->l_proc; 442 struct vnode *vp; 443 struct mount *mp; 444 int error; 445 struct nameidata nd; 446 447 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 448 SCARG(uap, path), l); 449 if ((error = namei(&nd)) != 0) 450 return (error); 451 vp = nd.ni_vp; 452 mp = vp->v_mount; 453 454 /* 455 * Only root, or the user that did the original mount is 456 * permitted to unmount this filesystem. 457 */ 458 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 459 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 460 vput(vp); 461 return (error); 462 } 463 464 /* 465 * Don't allow unmounting the root file system. 466 */ 467 if (mp->mnt_flag & MNT_ROOTFS) { 468 vput(vp); 469 return (EINVAL); 470 } 471 472 /* 473 * Must be the root of the filesystem 474 */ 475 if ((vp->v_flag & VROOT) == 0) { 476 vput(vp); 477 return (EINVAL); 478 } 479 vput(vp); 480 481 /* 482 * XXX Freeze syncer. Must do this before locking the 483 * mount point. See dounmount() for details. 484 */ 485 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 486 487 if (vfs_busy(mp, 0, 0)) { 488 lockmgr(&syncer_lock, LK_RELEASE, NULL); 489 return (EBUSY); 490 } 491 492 return (dounmount(mp, SCARG(uap, flags), l)); 493 } 494 495 /* 496 * Do the actual file system unmount. File system is assumed to have been 497 * marked busy by the caller. 498 */ 499 int 500 dounmount(mp, flags, l) 501 struct mount *mp; 502 int flags; 503 struct lwp *l; 504 { 505 struct vnode *coveredvp; 506 int error; 507 int async; 508 int used_syncer; 509 510 simple_lock(&mountlist_slock); 511 vfs_unbusy(mp); 512 used_syncer = (mp->mnt_syncer != NULL); 513 514 /* 515 * XXX Syncer must be frozen when we get here. This should really 516 * be done on a per-mountpoint basis, but especially the softdep 517 * code possibly called from the syncer doens't exactly work on a 518 * per-mountpoint basis, so the softdep code would become a maze 519 * of vfs_busy() calls. 520 * 521 * The caller of dounmount() must acquire syncer_lock because 522 * the syncer itself acquires locks in syncer_lock -> vfs_busy 523 * order, and we must preserve that order to avoid deadlock. 524 * 525 * So, if the file system did not use the syncer, now is 526 * the time to release the syncer_lock. 527 */ 528 if (used_syncer == 0) 529 lockmgr(&syncer_lock, LK_RELEASE, NULL); 530 531 mp->mnt_iflag |= IMNT_UNMOUNT; 532 mp->mnt_unmounter = l; 533 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 534 vn_start_write(NULL, &mp, V_WAIT); 535 536 if (mp->mnt_flag & MNT_EXPUBLIC) 537 vfs_setpublicfs(NULL, NULL, NULL); 538 async = mp->mnt_flag & MNT_ASYNC; 539 mp->mnt_flag &= ~MNT_ASYNC; 540 cache_purgevfs(mp); /* remove cache entries for this file sys */ 541 if (mp->mnt_syncer != NULL) 542 vfs_deallocate_syncvnode(mp); 543 error = 0; 544 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 545 #if NFSS > 0 546 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 547 #endif 548 if (error == 0) 549 error = VFS_SYNC(mp, MNT_WAIT, l->l_proc->p_ucred, l); 550 } 551 if (error == 0 || (flags & MNT_FORCE)) 552 error = VFS_UNMOUNT(mp, flags, l); 553 vn_finished_write(mp, 0); 554 simple_lock(&mountlist_slock); 555 if (error) { 556 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 557 (void) vfs_allocate_syncvnode(mp); 558 mp->mnt_iflag &= ~IMNT_UNMOUNT; 559 mp->mnt_unmounter = NULL; 560 mp->mnt_flag |= async; 561 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 562 &mountlist_slock); 563 if (used_syncer) 564 lockmgr(&syncer_lock, LK_RELEASE, NULL); 565 simple_lock(&mp->mnt_slock); 566 while (mp->mnt_wcnt > 0) { 567 wakeup(mp); 568 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 569 0, &mp->mnt_slock); 570 } 571 simple_unlock(&mp->mnt_slock); 572 return (error); 573 } 574 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 575 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 576 coveredvp->v_mountedhere = NULL; 577 vrele(coveredvp); 578 } 579 mp->mnt_op->vfs_refcount--; 580 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 581 panic("unmount: dangling vnode"); 582 mp->mnt_iflag |= IMNT_GONE; 583 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 584 if (used_syncer) 585 lockmgr(&syncer_lock, LK_RELEASE, NULL); 586 simple_lock(&mp->mnt_slock); 587 while (mp->mnt_wcnt > 0) { 588 wakeup(mp); 589 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 590 } 591 simple_unlock(&mp->mnt_slock); 592 free(mp, M_MOUNT); 593 return (0); 594 } 595 596 /* 597 * Sync each mounted filesystem. 598 */ 599 #ifdef DEBUG 600 int syncprt = 0; 601 struct ctldebug debug0 = { "syncprt", &syncprt }; 602 #endif 603 604 /* ARGSUSED */ 605 int 606 sys_sync(l, v, retval) 607 struct lwp *l; 608 void *v; 609 register_t *retval; 610 { 611 struct mount *mp, *nmp; 612 int asyncflag; 613 struct proc *p = l == NULL ? &proc0 : l->l_proc; 614 615 simple_lock(&mountlist_slock); 616 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 617 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 618 nmp = mp->mnt_list.cqe_prev; 619 continue; 620 } 621 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 622 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 623 asyncflag = mp->mnt_flag & MNT_ASYNC; 624 mp->mnt_flag &= ~MNT_ASYNC; 625 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, l); 626 if (asyncflag) 627 mp->mnt_flag |= MNT_ASYNC; 628 vn_finished_write(mp, 0); 629 } 630 simple_lock(&mountlist_slock); 631 nmp = mp->mnt_list.cqe_prev; 632 vfs_unbusy(mp); 633 634 } 635 simple_unlock(&mountlist_slock); 636 #ifdef DEBUG 637 if (syncprt) 638 vfs_bufstats(); 639 #endif /* DEBUG */ 640 return (0); 641 } 642 643 /* 644 * Change filesystem quotas. 645 */ 646 /* ARGSUSED */ 647 int 648 sys_quotactl(l, v, retval) 649 struct lwp *l; 650 void *v; 651 register_t *retval; 652 { 653 struct sys_quotactl_args /* { 654 syscallarg(const char *) path; 655 syscallarg(int) cmd; 656 syscallarg(int) uid; 657 syscallarg(caddr_t) arg; 658 } */ *uap = v; 659 struct mount *mp; 660 int error; 661 struct nameidata nd; 662 663 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 664 if ((error = namei(&nd)) != 0) 665 return (error); 666 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 667 vrele(nd.ni_vp); 668 if (error) 669 return (error); 670 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 671 SCARG(uap, arg), l); 672 vn_finished_write(mp, 0); 673 return (error); 674 } 675 676 int 677 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 678 int root) 679 { 680 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 681 int error = 0; 682 683 /* 684 * If MNT_NOWAIT or MNT_LAZY is specified, do not 685 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 686 * overrides MNT_NOWAIT. 687 */ 688 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 689 (flags != MNT_WAIT && flags != 0)) { 690 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 691 goto done; 692 } 693 694 /* Get the filesystem stats now */ 695 memset(sp, 0, sizeof(*sp)); 696 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 697 return error; 698 } 699 700 if (cwdi->cwdi_rdir == NULL) 701 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 702 done: 703 if (cwdi->cwdi_rdir != NULL) { 704 size_t len; 705 char *bp; 706 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 707 if (!path) 708 return ENOMEM; 709 710 bp = path + MAXPATHLEN; 711 *--bp = '\0'; 712 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 713 MAXPATHLEN / 2, 0, l); 714 if (error) { 715 free(path, M_TEMP); 716 return error; 717 } 718 len = strlen(bp); 719 /* 720 * for mount points that are below our root, we can see 721 * them, so we fix up the pathname and return them. The 722 * rest we cannot see, so we don't allow viewing the 723 * data. 724 */ 725 if (strncmp(bp, sp->f_mntonname, len) == 0) { 726 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 727 sizeof(sp->f_mntonname)); 728 if (sp->f_mntonname[0] == '\0') 729 (void)strlcpy(sp->f_mntonname, "/", 730 sizeof(sp->f_mntonname)); 731 } else { 732 if (root) 733 (void)strlcpy(sp->f_mntonname, "/", 734 sizeof(sp->f_mntonname)); 735 else 736 error = EPERM; 737 } 738 free(path, M_TEMP); 739 } 740 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 741 return error; 742 } 743 744 /* 745 * Get filesystem statistics. 746 */ 747 /* ARGSUSED */ 748 int 749 sys_statvfs1(l, v, retval) 750 struct lwp *l; 751 void *v; 752 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 mount *mp; 760 struct statvfs sbuf; 761 int error; 762 struct nameidata nd; 763 764 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 765 if ((error = namei(&nd)) != 0) 766 return error; 767 mp = nd.ni_vp->v_mount; 768 vrele(nd.ni_vp); 769 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 770 return error; 771 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 772 } 773 774 /* 775 * Get filesystem statistics. 776 */ 777 /* ARGSUSED */ 778 int 779 sys_fstatvfs1(l, v, retval) 780 struct lwp *l; 781 void *v; 782 register_t *retval; 783 { 784 struct sys_fstatvfs1_args /* { 785 syscallarg(int) fd; 786 syscallarg(struct statvfs *) buf; 787 syscallarg(int) flags; 788 } */ *uap = v; 789 struct proc *p = l->l_proc; 790 struct file *fp; 791 struct mount *mp; 792 struct statvfs sbuf; 793 int error; 794 795 /* getvnode() will use the descriptor for us */ 796 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 797 return (error); 798 mp = ((struct vnode *)fp->f_data)->v_mount; 799 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 800 goto out; 801 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 802 out: 803 FILE_UNUSE(fp, l); 804 return error; 805 } 806 807 808 /* 809 * Get statistics on all filesystems. 810 */ 811 int 812 sys_getvfsstat(l, v, retval) 813 struct lwp *l; 814 void *v; 815 register_t *retval; 816 { 817 struct sys_getvfsstat_args /* { 818 syscallarg(struct statvfs *) buf; 819 syscallarg(size_t) bufsize; 820 syscallarg(int) flags; 821 } */ *uap = v; 822 int root = 0; 823 struct proc *p = l->l_proc; 824 struct mount *mp, *nmp; 825 struct statvfs sbuf; 826 struct statvfs *sfsp; 827 size_t count, maxcount; 828 int error = 0; 829 830 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 831 sfsp = SCARG(uap, buf); 832 simple_lock(&mountlist_slock); 833 count = 0; 834 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 835 mp = nmp) { 836 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 837 nmp = CIRCLEQ_NEXT(mp, mnt_list); 838 continue; 839 } 840 if (sfsp && count < maxcount) { 841 error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 0); 842 if (error) { 843 simple_lock(&mountlist_slock); 844 nmp = CIRCLEQ_NEXT(mp, mnt_list); 845 vfs_unbusy(mp); 846 continue; 847 } 848 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 849 if (error) { 850 vfs_unbusy(mp); 851 return (error); 852 } 853 sfsp++; 854 root |= strcmp(sbuf.f_mntonname, "/") == 0; 855 } 856 count++; 857 simple_lock(&mountlist_slock); 858 nmp = CIRCLEQ_NEXT(mp, mnt_list); 859 vfs_unbusy(mp); 860 } 861 simple_unlock(&mountlist_slock); 862 if (root == 0 && p->p_cwdi->cwdi_rdir) { 863 /* 864 * fake a root entry 865 */ 866 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, l, 867 SCARG(uap, flags), 1)) != 0) 868 return error; 869 if (sfsp) 870 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 871 count++; 872 } 873 if (sfsp && count > maxcount) 874 *retval = maxcount; 875 else 876 *retval = count; 877 return error; 878 } 879 880 /* 881 * Change current working directory to a given file descriptor. 882 */ 883 /* ARGSUSED */ 884 int 885 sys_fchdir(l, v, retval) 886 struct lwp *l; 887 void *v; 888 register_t *retval; 889 { 890 struct sys_fchdir_args /* { 891 syscallarg(int) fd; 892 } */ *uap = v; 893 struct proc *p = l->l_proc; 894 struct filedesc *fdp = p->p_fd; 895 struct cwdinfo *cwdi = p->p_cwdi; 896 struct vnode *vp, *tdp; 897 struct mount *mp; 898 struct file *fp; 899 int error; 900 901 /* getvnode() will use the descriptor for us */ 902 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 903 return (error); 904 vp = (struct vnode *)fp->f_data; 905 906 VREF(vp); 907 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 908 if (vp->v_type != VDIR) 909 error = ENOTDIR; 910 else 911 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, l); 912 while (!error && (mp = vp->v_mountedhere) != NULL) { 913 if (vfs_busy(mp, 0, 0)) 914 continue; 915 error = VFS_ROOT(mp, &tdp); 916 vfs_unbusy(mp); 917 if (error) 918 break; 919 vput(vp); 920 vp = tdp; 921 } 922 if (error) { 923 vput(vp); 924 goto out; 925 } 926 VOP_UNLOCK(vp, 0); 927 928 /* 929 * Disallow changing to a directory not under the process's 930 * current root directory (if there is one). 931 */ 932 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 933 vrele(vp); 934 error = EPERM; /* operation not permitted */ 935 goto out; 936 } 937 938 vrele(cwdi->cwdi_cdir); 939 cwdi->cwdi_cdir = vp; 940 out: 941 FILE_UNUSE(fp, l); 942 return (error); 943 } 944 945 /* 946 * Change this process's notion of the root directory to a given file descriptor. 947 */ 948 949 int 950 sys_fchroot(l, v, retval) 951 struct lwp *l; 952 void *v; 953 register_t *retval; 954 { 955 struct sys_fchroot_args *uap = v; 956 struct proc *p = l->l_proc; 957 struct filedesc *fdp = p->p_fd; 958 struct cwdinfo *cwdi = p->p_cwdi; 959 struct vnode *vp; 960 struct file *fp; 961 int error; 962 963 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 964 return error; 965 /* getvnode() will use the descriptor for us */ 966 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 967 return error; 968 vp = (struct vnode *) fp->f_data; 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_ucred, l); 974 VOP_UNLOCK(vp, 0); 975 if (error) 976 goto out; 977 VREF(vp); 978 979 /* 980 * Prevent escaping from chroot by putting the root under 981 * the working directory. Silently chdir to / if we aren't 982 * already there. 983 */ 984 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 985 /* 986 * XXX would be more failsafe to change directory to a 987 * deadfs node here instead 988 */ 989 vrele(cwdi->cwdi_cdir); 990 VREF(vp); 991 cwdi->cwdi_cdir = vp; 992 } 993 994 if (cwdi->cwdi_rdir != NULL) 995 vrele(cwdi->cwdi_rdir); 996 cwdi->cwdi_rdir = vp; 997 out: 998 FILE_UNUSE(fp, l); 999 return (error); 1000 } 1001 1002 1003 1004 /* 1005 * Change current working directory (``.''). 1006 */ 1007 /* ARGSUSED */ 1008 int 1009 sys_chdir(l, v, retval) 1010 struct lwp *l; 1011 void *v; 1012 register_t *retval; 1013 { 1014 struct sys_chdir_args /* { 1015 syscallarg(const char *) path; 1016 } */ *uap = v; 1017 struct proc *p = l->l_proc; 1018 struct cwdinfo *cwdi = p->p_cwdi; 1019 int error; 1020 struct nameidata nd; 1021 1022 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1023 SCARG(uap, path), l); 1024 if ((error = change_dir(&nd, l)) != 0) 1025 return (error); 1026 vrele(cwdi->cwdi_cdir); 1027 cwdi->cwdi_cdir = nd.ni_vp; 1028 return (0); 1029 } 1030 1031 /* 1032 * Change notion of root (``/'') directory. 1033 */ 1034 /* ARGSUSED */ 1035 int 1036 sys_chroot(l, v, retval) 1037 struct lwp *l; 1038 void *v; 1039 register_t *retval; 1040 { 1041 struct sys_chroot_args /* { 1042 syscallarg(const char *) path; 1043 } */ *uap = v; 1044 struct proc *p = l->l_proc; 1045 struct cwdinfo *cwdi = p->p_cwdi; 1046 struct vnode *vp; 1047 int error; 1048 struct nameidata nd; 1049 1050 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1051 return (error); 1052 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1053 SCARG(uap, path), l); 1054 if ((error = change_dir(&nd, l)) != 0) 1055 return (error); 1056 if (cwdi->cwdi_rdir != NULL) 1057 vrele(cwdi->cwdi_rdir); 1058 vp = nd.ni_vp; 1059 cwdi->cwdi_rdir = vp; 1060 1061 /* 1062 * Prevent escaping from chroot by putting the root under 1063 * the working directory. Silently chdir to / if we aren't 1064 * already there. 1065 */ 1066 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1067 /* 1068 * XXX would be more failsafe to change directory to a 1069 * deadfs node here instead 1070 */ 1071 vrele(cwdi->cwdi_cdir); 1072 VREF(vp); 1073 cwdi->cwdi_cdir = vp; 1074 } 1075 1076 return (0); 1077 } 1078 1079 /* 1080 * Common routine for chroot and chdir. 1081 */ 1082 static int 1083 change_dir(ndp, l) 1084 struct nameidata *ndp; 1085 struct lwp *l; 1086 { 1087 struct vnode *vp; 1088 int error; 1089 1090 if ((error = namei(ndp)) != 0) 1091 return (error); 1092 vp = ndp->ni_vp; 1093 if (vp->v_type != VDIR) 1094 error = ENOTDIR; 1095 else 1096 error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_ucred, l); 1097 1098 if (error) 1099 vput(vp); 1100 else 1101 VOP_UNLOCK(vp, 0); 1102 return (error); 1103 } 1104 1105 /* 1106 * Check permissions, allocate an open file structure, 1107 * and call the device open routine if any. 1108 */ 1109 int 1110 sys_open(l, v, retval) 1111 struct lwp *l; 1112 void *v; 1113 register_t *retval; 1114 { 1115 struct sys_open_args /* { 1116 syscallarg(const char *) path; 1117 syscallarg(int) flags; 1118 syscallarg(int) mode; 1119 } */ *uap = v; 1120 struct proc *p = l->l_proc; 1121 struct cwdinfo *cwdi = p->p_cwdi; 1122 struct filedesc *fdp = p->p_fd; 1123 struct file *fp; 1124 struct vnode *vp; 1125 int flags, cmode; 1126 int type, indx, error; 1127 struct flock lf; 1128 struct nameidata nd; 1129 1130 flags = FFLAGS(SCARG(uap, flags)); 1131 if ((flags & (FREAD | FWRITE)) == 0) 1132 return (EINVAL); 1133 /* falloc() will use the file descriptor for us */ 1134 if ((error = falloc(p, &fp, &indx)) != 0) 1135 return (error); 1136 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1137 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1138 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1139 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1140 FILE_UNUSE(fp, l); 1141 fdp->fd_ofiles[indx] = NULL; 1142 ffree(fp); 1143 if ((error == ENODEV || error == ENXIO) && 1144 l->l_dupfd >= 0 && /* XXX from fdopen */ 1145 (error = 1146 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1147 *retval = indx; 1148 return (0); 1149 } 1150 if (error == ERESTART) 1151 error = EINTR; 1152 fdremove(fdp, indx); 1153 return (error); 1154 } 1155 l->l_dupfd = 0; 1156 vp = nd.ni_vp; 1157 fp->f_flag = flags & FMASK; 1158 fp->f_type = DTYPE_VNODE; 1159 fp->f_ops = &vnops; 1160 fp->f_data = vp; 1161 if (flags & (O_EXLOCK | O_SHLOCK)) { 1162 lf.l_whence = SEEK_SET; 1163 lf.l_start = 0; 1164 lf.l_len = 0; 1165 if (flags & O_EXLOCK) 1166 lf.l_type = F_WRLCK; 1167 else 1168 lf.l_type = F_RDLCK; 1169 type = F_FLOCK; 1170 if ((flags & FNONBLOCK) == 0) 1171 type |= F_WAIT; 1172 VOP_UNLOCK(vp, 0); 1173 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1174 if (error) { 1175 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1176 FILE_UNUSE(fp, l); 1177 ffree(fp); 1178 fdremove(fdp, indx); 1179 return (error); 1180 } 1181 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1182 fp->f_flag |= FHASLOCK; 1183 } 1184 VOP_UNLOCK(vp, 0); 1185 *retval = indx; 1186 FILE_SET_MATURE(fp); 1187 FILE_UNUSE(fp, l); 1188 return (0); 1189 } 1190 1191 /* 1192 * Get file handle system call 1193 */ 1194 int 1195 sys_getfh(l, v, retval) 1196 struct lwp *l; 1197 void *v; 1198 register_t *retval; 1199 { 1200 struct sys_getfh_args /* { 1201 syscallarg(char *) fname; 1202 syscallarg(fhandle_t *) fhp; 1203 } */ *uap = v; 1204 struct proc *p = l->l_proc; 1205 struct vnode *vp; 1206 fhandle_t fh; 1207 int error; 1208 struct nameidata nd; 1209 1210 /* 1211 * Must be super user 1212 */ 1213 error = suser(p->p_ucred, &p->p_acflag); 1214 if (error) 1215 return (error); 1216 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1217 SCARG(uap, fname), l); 1218 error = namei(&nd); 1219 if (error) 1220 return (error); 1221 vp = nd.ni_vp; 1222 memset(&fh, 0, sizeof(fh)); 1223 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1224 error = VFS_VPTOFH(vp, &fh.fh_fid); 1225 vput(vp); 1226 if (error) 1227 return (error); 1228 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1229 return (error); 1230 } 1231 1232 /* 1233 * Open a file given a file handle. 1234 * 1235 * Check permissions, allocate an open file structure, 1236 * and call the device open routine if any. 1237 */ 1238 int 1239 sys_fhopen(l, v, retval) 1240 struct lwp *l; 1241 void *v; 1242 register_t *retval; 1243 { 1244 struct sys_fhopen_args /* { 1245 syscallarg(const fhandle_t *) fhp; 1246 syscallarg(int) flags; 1247 } */ *uap = v; 1248 struct proc *p = l->l_proc; 1249 struct filedesc *fdp = p->p_fd; 1250 struct file *fp; 1251 struct vnode *vp = NULL; 1252 struct mount *mp; 1253 struct ucred *cred = p->p_ucred; 1254 int flags; 1255 struct file *nfp; 1256 int type, indx, error=0; 1257 struct flock lf; 1258 struct vattr va; 1259 fhandle_t fh; 1260 1261 /* 1262 * Must be super user 1263 */ 1264 if ((error = suser(p->p_ucred, &p->p_acflag))) 1265 return (error); 1266 1267 flags = FFLAGS(SCARG(uap, flags)); 1268 if ((flags & (FREAD | FWRITE)) == 0) 1269 return (EINVAL); 1270 if ((flags & O_CREAT)) 1271 return (EINVAL); 1272 /* falloc() will use the file descriptor for us */ 1273 if ((error = falloc(p, &nfp, &indx)) != 0) 1274 return (error); 1275 fp = nfp; 1276 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1277 goto bad; 1278 1279 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1280 error = ESTALE; 1281 goto bad; 1282 } 1283 1284 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1285 vp = NULL; /* most likely unnecessary sanity for bad: */ 1286 goto bad; 1287 } 1288 1289 /* Now do an effective vn_open */ 1290 1291 if (vp->v_type == VSOCK) { 1292 error = EOPNOTSUPP; 1293 goto bad; 1294 } 1295 if (flags & FREAD) { 1296 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1297 goto bad; 1298 } 1299 if (flags & (FWRITE | O_TRUNC)) { 1300 if (vp->v_type == VDIR) { 1301 error = EISDIR; 1302 goto bad; 1303 } 1304 if ((error = vn_writechk(vp)) != 0 || 1305 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1306 goto bad; 1307 } 1308 if (flags & O_TRUNC) { 1309 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1310 goto bad; 1311 VOP_UNLOCK(vp, 0); /* XXX */ 1312 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1313 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1314 VATTR_NULL(&va); 1315 va.va_size = 0; 1316 error = VOP_SETATTR(vp, &va, cred, l); 1317 vn_finished_write(mp, 0); 1318 if (error) 1319 goto bad; 1320 } 1321 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1322 goto bad; 1323 if (vp->v_type == VREG && 1324 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1325 error = EIO; 1326 goto bad; 1327 } 1328 if (flags & FWRITE) 1329 vp->v_writecount++; 1330 1331 /* done with modified vn_open, now finish what sys_open does. */ 1332 1333 fp->f_flag = flags & FMASK; 1334 fp->f_type = DTYPE_VNODE; 1335 fp->f_ops = &vnops; 1336 fp->f_data = vp; 1337 if (flags & (O_EXLOCK | O_SHLOCK)) { 1338 lf.l_whence = SEEK_SET; 1339 lf.l_start = 0; 1340 lf.l_len = 0; 1341 if (flags & O_EXLOCK) 1342 lf.l_type = F_WRLCK; 1343 else 1344 lf.l_type = F_RDLCK; 1345 type = F_FLOCK; 1346 if ((flags & FNONBLOCK) == 0) 1347 type |= F_WAIT; 1348 VOP_UNLOCK(vp, 0); 1349 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1350 if (error) { 1351 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1352 FILE_UNUSE(fp, l); 1353 ffree(fp); 1354 fdremove(fdp, indx); 1355 return (error); 1356 } 1357 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1358 fp->f_flag |= FHASLOCK; 1359 } 1360 VOP_UNLOCK(vp, 0); 1361 *retval = indx; 1362 FILE_SET_MATURE(fp); 1363 FILE_UNUSE(fp, l); 1364 return (0); 1365 1366 bad: 1367 FILE_UNUSE(fp, l); 1368 ffree(fp); 1369 fdremove(fdp, indx); 1370 if (vp != NULL) 1371 vput(vp); 1372 return (error); 1373 } 1374 1375 /* ARGSUSED */ 1376 int 1377 sys_fhstat(l, v, retval) 1378 struct lwp *l; 1379 void *v; 1380 register_t *retval; 1381 { 1382 struct sys_fhstat_args /* { 1383 syscallarg(const fhandle_t *) fhp; 1384 syscallarg(struct stat *) sb; 1385 } */ *uap = v; 1386 struct proc *p = l->l_proc; 1387 struct stat sb; 1388 int error; 1389 fhandle_t fh; 1390 struct mount *mp; 1391 struct vnode *vp; 1392 1393 /* 1394 * Must be super user 1395 */ 1396 if ((error = suser(p->p_ucred, &p->p_acflag))) 1397 return (error); 1398 1399 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1400 return (error); 1401 1402 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1403 return (ESTALE); 1404 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1405 return (error); 1406 error = vn_stat(vp, &sb, l); 1407 vput(vp); 1408 if (error) 1409 return (error); 1410 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1411 return (error); 1412 } 1413 1414 /* ARGSUSED */ 1415 int 1416 sys_fhstatvfs1(l, v, retval) 1417 struct lwp *l; 1418 void *v; 1419 register_t *retval; 1420 { 1421 struct sys_fhstatvfs1_args /* 1422 syscallarg(const fhandle_t *) fhp; 1423 syscallarg(struct statvfs *) buf; 1424 syscallarg(int) flags; 1425 } */ *uap = v; 1426 struct proc *p = l->l_proc; 1427 struct statvfs sbuf; 1428 fhandle_t fh; 1429 struct mount *mp; 1430 struct vnode *vp; 1431 int error; 1432 1433 /* 1434 * Must be super user 1435 */ 1436 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1437 return error; 1438 1439 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1440 return error; 1441 1442 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1443 return ESTALE; 1444 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1445 return error; 1446 1447 mp = vp->v_mount; 1448 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) { 1449 vput(vp); 1450 return error; 1451 } 1452 vput(vp); 1453 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1454 } 1455 1456 /* 1457 * Create a special file. 1458 */ 1459 /* ARGSUSED */ 1460 int 1461 sys_mknod(l, v, retval) 1462 struct lwp *l; 1463 void *v; 1464 register_t *retval; 1465 { 1466 struct sys_mknod_args /* { 1467 syscallarg(const char *) path; 1468 syscallarg(int) mode; 1469 syscallarg(int) dev; 1470 } */ *uap = v; 1471 struct proc *p = l->l_proc; 1472 struct vnode *vp; 1473 struct mount *mp; 1474 struct vattr vattr; 1475 int error; 1476 int whiteout = 0; 1477 struct nameidata nd; 1478 1479 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1480 return (error); 1481 restart: 1482 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1483 if ((error = namei(&nd)) != 0) 1484 return (error); 1485 vp = nd.ni_vp; 1486 if (vp != NULL) 1487 error = EEXIST; 1488 else { 1489 VATTR_NULL(&vattr); 1490 vattr.va_mode = 1491 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1492 vattr.va_rdev = SCARG(uap, dev); 1493 whiteout = 0; 1494 1495 switch (SCARG(uap, mode) & S_IFMT) { 1496 case S_IFMT: /* used by badsect to flag bad sectors */ 1497 vattr.va_type = VBAD; 1498 break; 1499 case S_IFCHR: 1500 vattr.va_type = VCHR; 1501 break; 1502 case S_IFBLK: 1503 vattr.va_type = VBLK; 1504 break; 1505 case S_IFWHT: 1506 whiteout = 1; 1507 break; 1508 default: 1509 error = EINVAL; 1510 break; 1511 } 1512 } 1513 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1514 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1515 if (nd.ni_dvp == vp) 1516 vrele(nd.ni_dvp); 1517 else 1518 vput(nd.ni_dvp); 1519 if (vp) 1520 vrele(vp); 1521 if ((error = vn_start_write(NULL, &mp, 1522 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1523 return (error); 1524 goto restart; 1525 } 1526 if (!error) { 1527 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1528 if (whiteout) { 1529 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1530 if (error) 1531 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1532 vput(nd.ni_dvp); 1533 } else { 1534 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1535 &nd.ni_cnd, &vattr); 1536 if (error == 0) 1537 vput(nd.ni_vp); 1538 } 1539 } else { 1540 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1541 if (nd.ni_dvp == vp) 1542 vrele(nd.ni_dvp); 1543 else 1544 vput(nd.ni_dvp); 1545 if (vp) 1546 vrele(vp); 1547 } 1548 vn_finished_write(mp, 0); 1549 return (error); 1550 } 1551 1552 /* 1553 * Create a named pipe. 1554 */ 1555 /* ARGSUSED */ 1556 int 1557 sys_mkfifo(l, v, retval) 1558 struct lwp *l; 1559 void *v; 1560 register_t *retval; 1561 { 1562 struct sys_mkfifo_args /* { 1563 syscallarg(const char *) path; 1564 syscallarg(int) mode; 1565 } */ *uap = v; 1566 struct proc *p = l->l_proc; 1567 struct mount *mp; 1568 struct vattr vattr; 1569 int error; 1570 struct nameidata nd; 1571 1572 restart: 1573 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1574 if ((error = namei(&nd)) != 0) 1575 return (error); 1576 if (nd.ni_vp != NULL) { 1577 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1578 if (nd.ni_dvp == nd.ni_vp) 1579 vrele(nd.ni_dvp); 1580 else 1581 vput(nd.ni_dvp); 1582 vrele(nd.ni_vp); 1583 return (EEXIST); 1584 } 1585 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1586 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1587 if (nd.ni_dvp == nd.ni_vp) 1588 vrele(nd.ni_dvp); 1589 else 1590 vput(nd.ni_dvp); 1591 if (nd.ni_vp) 1592 vrele(nd.ni_vp); 1593 if ((error = vn_start_write(NULL, &mp, 1594 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1595 return (error); 1596 goto restart; 1597 } 1598 VATTR_NULL(&vattr); 1599 vattr.va_type = VFIFO; 1600 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1601 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1602 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1603 if (error == 0) 1604 vput(nd.ni_vp); 1605 vn_finished_write(mp, 0); 1606 return (error); 1607 } 1608 1609 /* 1610 * Make a hard file link. 1611 */ 1612 /* ARGSUSED */ 1613 int 1614 sys_link(l, v, retval) 1615 struct lwp *l; 1616 void *v; 1617 register_t *retval; 1618 { 1619 struct sys_link_args /* { 1620 syscallarg(const char *) path; 1621 syscallarg(const char *) link; 1622 } */ *uap = v; 1623 struct proc *p = l->l_proc; 1624 struct vnode *vp; 1625 struct mount *mp; 1626 struct nameidata nd; 1627 int error; 1628 1629 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1630 if ((error = namei(&nd)) != 0) 1631 return (error); 1632 vp = nd.ni_vp; 1633 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1634 vrele(vp); 1635 return (error); 1636 } 1637 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 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 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1651 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 1652 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1653 out: 1654 vrele(vp); 1655 vn_finished_write(mp, 0); 1656 return (error); 1657 } 1658 1659 /* 1660 * Make a symbolic link. 1661 */ 1662 /* ARGSUSED */ 1663 int 1664 sys_symlink(l, v, retval) 1665 struct lwp *l; 1666 void *v; 1667 register_t *retval; 1668 { 1669 struct sys_symlink_args /* { 1670 syscallarg(const char *) path; 1671 syscallarg(const char *) link; 1672 } */ *uap = v; 1673 struct proc *p = l->l_proc; 1674 struct mount *mp; 1675 struct vattr vattr; 1676 char *path; 1677 int error; 1678 struct nameidata nd; 1679 1680 path = PNBUF_GET(); 1681 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1682 if (error) 1683 goto out; 1684 restart: 1685 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1686 if ((error = namei(&nd)) != 0) 1687 goto out; 1688 if (nd.ni_vp) { 1689 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1690 if (nd.ni_dvp == nd.ni_vp) 1691 vrele(nd.ni_dvp); 1692 else 1693 vput(nd.ni_dvp); 1694 vrele(nd.ni_vp); 1695 error = EEXIST; 1696 goto out; 1697 } 1698 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1699 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1700 if (nd.ni_dvp == nd.ni_vp) 1701 vrele(nd.ni_dvp); 1702 else 1703 vput(nd.ni_dvp); 1704 if ((error = vn_start_write(NULL, &mp, 1705 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1706 return (error); 1707 goto restart; 1708 } 1709 VATTR_NULL(&vattr); 1710 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1711 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1712 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1713 if (error == 0) 1714 vput(nd.ni_vp); 1715 vn_finished_write(mp, 0); 1716 out: 1717 PNBUF_PUT(path); 1718 return (error); 1719 } 1720 1721 /* 1722 * Delete a whiteout from the filesystem. 1723 */ 1724 /* ARGSUSED */ 1725 int 1726 sys_undelete(l, v, retval) 1727 struct lwp *l; 1728 void *v; 1729 register_t *retval; 1730 { 1731 struct sys_undelete_args /* { 1732 syscallarg(const char *) path; 1733 } */ *uap = v; 1734 struct proc *p = l->l_proc; 1735 int error; 1736 struct mount *mp; 1737 struct nameidata nd; 1738 1739 restart: 1740 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1741 SCARG(uap, path), l); 1742 error = namei(&nd); 1743 if (error) 1744 return (error); 1745 1746 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1747 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1748 if (nd.ni_dvp == nd.ni_vp) 1749 vrele(nd.ni_dvp); 1750 else 1751 vput(nd.ni_dvp); 1752 if (nd.ni_vp) 1753 vrele(nd.ni_vp); 1754 return (EEXIST); 1755 } 1756 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1757 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1758 if (nd.ni_dvp == nd.ni_vp) 1759 vrele(nd.ni_dvp); 1760 else 1761 vput(nd.ni_dvp); 1762 if ((error = vn_start_write(NULL, &mp, 1763 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1764 return (error); 1765 goto restart; 1766 } 1767 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1768 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1769 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1770 vput(nd.ni_dvp); 1771 vn_finished_write(mp, 0); 1772 return (error); 1773 } 1774 1775 /* 1776 * Delete a name from the filesystem. 1777 */ 1778 /* ARGSUSED */ 1779 int 1780 sys_unlink(l, v, retval) 1781 struct lwp *l; 1782 void *v; 1783 register_t *retval; 1784 { 1785 struct sys_unlink_args /* { 1786 syscallarg(const char *) path; 1787 } */ *uap = v; 1788 struct proc *p = l->l_proc; 1789 struct mount *mp; 1790 struct vnode *vp; 1791 int error; 1792 struct nameidata nd; 1793 1794 restart: 1795 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1796 SCARG(uap, path), l); 1797 if ((error = namei(&nd)) != 0) 1798 return (error); 1799 vp = nd.ni_vp; 1800 1801 /* 1802 * The root of a mounted filesystem cannot be deleted. 1803 */ 1804 if (vp->v_flag & VROOT) { 1805 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1806 if (nd.ni_dvp == vp) 1807 vrele(nd.ni_dvp); 1808 else 1809 vput(nd.ni_dvp); 1810 vput(vp); 1811 error = EBUSY; 1812 goto out; 1813 } 1814 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1815 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1816 if (nd.ni_dvp == vp) 1817 vrele(nd.ni_dvp); 1818 else 1819 vput(nd.ni_dvp); 1820 vput(vp); 1821 if ((error = vn_start_write(NULL, &mp, 1822 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1823 return (error); 1824 goto restart; 1825 } 1826 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1827 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 1828 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1829 vn_finished_write(mp, 0); 1830 out: 1831 return (error); 1832 } 1833 1834 /* 1835 * Reposition read/write file offset. 1836 */ 1837 int 1838 sys_lseek(l, v, retval) 1839 struct lwp *l; 1840 void *v; 1841 register_t *retval; 1842 { 1843 struct sys_lseek_args /* { 1844 syscallarg(int) fd; 1845 syscallarg(int) pad; 1846 syscallarg(off_t) offset; 1847 syscallarg(int) whence; 1848 } */ *uap = v; 1849 struct proc *p = l->l_proc; 1850 struct ucred *cred = p->p_ucred; 1851 struct filedesc *fdp = p->p_fd; 1852 struct file *fp; 1853 struct vnode *vp; 1854 struct vattr vattr; 1855 off_t newoff; 1856 int error; 1857 1858 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1859 return (EBADF); 1860 1861 FILE_USE(fp); 1862 1863 vp = (struct vnode *)fp->f_data; 1864 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1865 error = ESPIPE; 1866 goto out; 1867 } 1868 1869 switch (SCARG(uap, whence)) { 1870 case SEEK_CUR: 1871 newoff = fp->f_offset + SCARG(uap, offset); 1872 break; 1873 case SEEK_END: 1874 error = VOP_GETATTR(vp, &vattr, cred, l); 1875 if (error) 1876 goto out; 1877 newoff = SCARG(uap, offset) + vattr.va_size; 1878 break; 1879 case SEEK_SET: 1880 newoff = SCARG(uap, offset); 1881 break; 1882 default: 1883 error = EINVAL; 1884 goto out; 1885 } 1886 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1887 goto out; 1888 1889 *(off_t *)retval = fp->f_offset = newoff; 1890 out: 1891 FILE_UNUSE(fp, l); 1892 return (error); 1893 } 1894 1895 /* 1896 * Positional read system call. 1897 */ 1898 int 1899 sys_pread(l, v, retval) 1900 struct lwp *l; 1901 void *v; 1902 register_t *retval; 1903 { 1904 struct sys_pread_args /* { 1905 syscallarg(int) fd; 1906 syscallarg(void *) buf; 1907 syscallarg(size_t) nbyte; 1908 syscallarg(off_t) offset; 1909 } */ *uap = v; 1910 struct proc *p = l->l_proc; 1911 struct filedesc *fdp = p->p_fd; 1912 struct file *fp; 1913 struct vnode *vp; 1914 off_t offset; 1915 int error, fd = SCARG(uap, fd); 1916 1917 if ((fp = fd_getfile(fdp, fd)) == NULL) 1918 return (EBADF); 1919 1920 if ((fp->f_flag & FREAD) == 0) { 1921 simple_unlock(&fp->f_slock); 1922 return (EBADF); 1923 } 1924 1925 FILE_USE(fp); 1926 1927 vp = (struct vnode *)fp->f_data; 1928 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1929 error = ESPIPE; 1930 goto out; 1931 } 1932 1933 offset = SCARG(uap, offset); 1934 1935 /* 1936 * XXX This works because no file systems actually 1937 * XXX take any action on the seek operation. 1938 */ 1939 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1940 goto out; 1941 1942 /* dofileread() will unuse the descriptor for us */ 1943 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1944 &offset, 0, retval)); 1945 1946 out: 1947 FILE_UNUSE(fp, l); 1948 return (error); 1949 } 1950 1951 /* 1952 * Positional scatter read system call. 1953 */ 1954 int 1955 sys_preadv(l, v, retval) 1956 struct lwp *l; 1957 void *v; 1958 register_t *retval; 1959 { 1960 struct sys_preadv_args /* { 1961 syscallarg(int) fd; 1962 syscallarg(const struct iovec *) iovp; 1963 syscallarg(int) iovcnt; 1964 syscallarg(off_t) offset; 1965 } */ *uap = v; 1966 struct proc *p = l->l_proc; 1967 struct filedesc *fdp = p->p_fd; 1968 struct file *fp; 1969 struct vnode *vp; 1970 off_t offset; 1971 int error, fd = SCARG(uap, fd); 1972 1973 if ((fp = fd_getfile(fdp, fd)) == NULL) 1974 return (EBADF); 1975 1976 if ((fp->f_flag & FREAD) == 0) { 1977 simple_unlock(&fp->f_slock); 1978 return (EBADF); 1979 } 1980 1981 FILE_USE(fp); 1982 1983 vp = (struct vnode *)fp->f_data; 1984 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1985 error = ESPIPE; 1986 goto out; 1987 } 1988 1989 offset = SCARG(uap, offset); 1990 1991 /* 1992 * XXX This works because no file systems actually 1993 * XXX take any action on the seek operation. 1994 */ 1995 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1996 goto out; 1997 1998 /* dofilereadv() will unuse the descriptor for us */ 1999 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2000 &offset, 0, retval)); 2001 2002 out: 2003 FILE_UNUSE(fp, l); 2004 return (error); 2005 } 2006 2007 /* 2008 * Positional write system call. 2009 */ 2010 int 2011 sys_pwrite(l, v, retval) 2012 struct lwp *l; 2013 void *v; 2014 register_t *retval; 2015 { 2016 struct sys_pwrite_args /* { 2017 syscallarg(int) fd; 2018 syscallarg(const void *) buf; 2019 syscallarg(size_t) nbyte; 2020 syscallarg(off_t) offset; 2021 } */ *uap = v; 2022 struct proc *p = l->l_proc; 2023 struct filedesc *fdp = p->p_fd; 2024 struct file *fp; 2025 struct vnode *vp; 2026 off_t offset; 2027 int error, fd = SCARG(uap, fd); 2028 2029 if ((fp = fd_getfile(fdp, fd)) == NULL) 2030 return (EBADF); 2031 2032 if ((fp->f_flag & FWRITE) == 0) { 2033 simple_unlock(&fp->f_slock); 2034 return (EBADF); 2035 } 2036 2037 FILE_USE(fp); 2038 2039 vp = (struct vnode *)fp->f_data; 2040 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2041 error = ESPIPE; 2042 goto out; 2043 } 2044 2045 offset = SCARG(uap, offset); 2046 2047 /* 2048 * XXX This works because no file systems actually 2049 * XXX take any action on the seek operation. 2050 */ 2051 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2052 goto out; 2053 2054 /* dofilewrite() will unuse the descriptor for us */ 2055 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2056 &offset, 0, retval)); 2057 2058 out: 2059 FILE_UNUSE(fp, l); 2060 return (error); 2061 } 2062 2063 /* 2064 * Positional gather write system call. 2065 */ 2066 int 2067 sys_pwritev(l, v, retval) 2068 struct lwp *l; 2069 void *v; 2070 register_t *retval; 2071 { 2072 struct sys_pwritev_args /* { 2073 syscallarg(int) fd; 2074 syscallarg(const struct iovec *) iovp; 2075 syscallarg(int) iovcnt; 2076 syscallarg(off_t) offset; 2077 } */ *uap = v; 2078 struct proc *p = l->l_proc; 2079 struct filedesc *fdp = p->p_fd; 2080 struct file *fp; 2081 struct vnode *vp; 2082 off_t offset; 2083 int error, fd = SCARG(uap, fd); 2084 2085 if ((fp = fd_getfile(fdp, fd)) == NULL) 2086 return (EBADF); 2087 2088 if ((fp->f_flag & FWRITE) == 0) { 2089 simple_unlock(&fp->f_slock); 2090 return (EBADF); 2091 } 2092 2093 FILE_USE(fp); 2094 2095 vp = (struct vnode *)fp->f_data; 2096 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2097 error = ESPIPE; 2098 goto out; 2099 } 2100 2101 offset = SCARG(uap, offset); 2102 2103 /* 2104 * XXX This works because no file systems actually 2105 * XXX take any action on the seek operation. 2106 */ 2107 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2108 goto out; 2109 2110 /* dofilewritev() will unuse the descriptor for us */ 2111 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2112 &offset, 0, retval)); 2113 2114 out: 2115 FILE_UNUSE(fp, l); 2116 return (error); 2117 } 2118 2119 /* 2120 * Check access permissions. 2121 */ 2122 int 2123 sys_access(l, v, retval) 2124 struct lwp *l; 2125 void *v; 2126 register_t *retval; 2127 { 2128 struct sys_access_args /* { 2129 syscallarg(const char *) path; 2130 syscallarg(int) flags; 2131 } */ *uap = v; 2132 struct proc *p = l->l_proc; 2133 struct ucred *cred; 2134 struct vnode *vp; 2135 int error, flags; 2136 struct nameidata nd; 2137 2138 cred = crdup(p->p_ucred); 2139 cred->cr_uid = p->p_cred->p_ruid; 2140 cred->cr_gid = p->p_cred->p_rgid; 2141 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2142 SCARG(uap, path), l); 2143 /* Override default credentials */ 2144 nd.ni_cnd.cn_cred = cred; 2145 if ((error = namei(&nd)) != 0) 2146 goto out; 2147 vp = nd.ni_vp; 2148 2149 /* Flags == 0 means only check for existence. */ 2150 if (SCARG(uap, flags)) { 2151 flags = 0; 2152 if (SCARG(uap, flags) & R_OK) 2153 flags |= VREAD; 2154 if (SCARG(uap, flags) & W_OK) 2155 flags |= VWRITE; 2156 if (SCARG(uap, flags) & X_OK) 2157 flags |= VEXEC; 2158 2159 error = VOP_ACCESS(vp, flags, cred, l); 2160 if (!error && (flags & VWRITE)) 2161 error = vn_writechk(vp); 2162 } 2163 vput(vp); 2164 out: 2165 crfree(cred); 2166 return (error); 2167 } 2168 2169 /* 2170 * Get file status; this version follows links. 2171 */ 2172 /* ARGSUSED */ 2173 int 2174 sys___stat13(l, v, retval) 2175 struct lwp *l; 2176 void *v; 2177 register_t *retval; 2178 { 2179 struct sys___stat13_args /* { 2180 syscallarg(const char *) path; 2181 syscallarg(struct stat *) ub; 2182 } */ *uap = v; 2183 struct stat sb; 2184 int error; 2185 struct nameidata nd; 2186 2187 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2188 SCARG(uap, path), l); 2189 if ((error = namei(&nd)) != 0) 2190 return (error); 2191 error = vn_stat(nd.ni_vp, &sb, l); 2192 vput(nd.ni_vp); 2193 if (error) 2194 return (error); 2195 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2196 return (error); 2197 } 2198 2199 /* 2200 * Get file status; this version does not follow links. 2201 */ 2202 /* ARGSUSED */ 2203 int 2204 sys___lstat13(l, v, retval) 2205 struct lwp *l; 2206 void *v; 2207 register_t *retval; 2208 { 2209 struct sys___lstat13_args /* { 2210 syscallarg(const char *) path; 2211 syscallarg(struct stat *) ub; 2212 } */ *uap = v; 2213 struct stat sb; 2214 int error; 2215 struct nameidata nd; 2216 2217 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2218 SCARG(uap, path), l); 2219 if ((error = namei(&nd)) != 0) 2220 return (error); 2221 error = vn_stat(nd.ni_vp, &sb, l); 2222 vput(nd.ni_vp); 2223 if (error) 2224 return (error); 2225 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2226 return (error); 2227 } 2228 2229 /* 2230 * Get configurable pathname variables. 2231 */ 2232 /* ARGSUSED */ 2233 int 2234 sys_pathconf(l, v, retval) 2235 struct lwp *l; 2236 void *v; 2237 register_t *retval; 2238 { 2239 struct sys_pathconf_args /* { 2240 syscallarg(const char *) path; 2241 syscallarg(int) name; 2242 } */ *uap = v; 2243 int error; 2244 struct nameidata nd; 2245 2246 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2247 SCARG(uap, path), l); 2248 if ((error = namei(&nd)) != 0) 2249 return (error); 2250 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2251 vput(nd.ni_vp); 2252 return (error); 2253 } 2254 2255 /* 2256 * Return target name of a symbolic link. 2257 */ 2258 /* ARGSUSED */ 2259 int 2260 sys_readlink(l, v, retval) 2261 struct lwp *l; 2262 void *v; 2263 register_t *retval; 2264 { 2265 struct sys_readlink_args /* { 2266 syscallarg(const char *) path; 2267 syscallarg(char *) buf; 2268 syscallarg(size_t) count; 2269 } */ *uap = v; 2270 struct proc *p = l->l_proc; 2271 struct vnode *vp; 2272 struct iovec aiov; 2273 struct uio auio; 2274 int error; 2275 struct nameidata nd; 2276 2277 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2278 SCARG(uap, path), l); 2279 if ((error = namei(&nd)) != 0) 2280 return (error); 2281 vp = nd.ni_vp; 2282 if (vp->v_type != VLNK) 2283 error = EINVAL; 2284 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2285 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, l)) == 0) { 2286 aiov.iov_base = SCARG(uap, buf); 2287 aiov.iov_len = SCARG(uap, count); 2288 auio.uio_iov = &aiov; 2289 auio.uio_iovcnt = 1; 2290 auio.uio_offset = 0; 2291 auio.uio_rw = UIO_READ; 2292 auio.uio_segflg = UIO_USERSPACE; 2293 auio.uio_lwp = l; 2294 auio.uio_resid = SCARG(uap, count); 2295 error = VOP_READLINK(vp, &auio, p->p_ucred); 2296 } 2297 vput(vp); 2298 *retval = SCARG(uap, count) - auio.uio_resid; 2299 return (error); 2300 } 2301 2302 /* 2303 * Change flags of a file given a path name. 2304 */ 2305 /* ARGSUSED */ 2306 int 2307 sys_chflags(l, v, retval) 2308 struct lwp *l; 2309 void *v; 2310 register_t *retval; 2311 { 2312 struct sys_chflags_args /* { 2313 syscallarg(const char *) path; 2314 syscallarg(u_long) flags; 2315 } */ *uap = v; 2316 struct vnode *vp; 2317 int error; 2318 struct nameidata nd; 2319 2320 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2321 if ((error = namei(&nd)) != 0) 2322 return (error); 2323 vp = nd.ni_vp; 2324 error = change_flags(vp, SCARG(uap, flags), l); 2325 vput(vp); 2326 return (error); 2327 } 2328 2329 /* 2330 * Change flags of a file given a file descriptor. 2331 */ 2332 /* ARGSUSED */ 2333 int 2334 sys_fchflags(l, v, retval) 2335 struct lwp *l; 2336 void *v; 2337 register_t *retval; 2338 { 2339 struct sys_fchflags_args /* { 2340 syscallarg(int) fd; 2341 syscallarg(u_long) flags; 2342 } */ *uap = v; 2343 struct proc *p = l->l_proc; 2344 struct vnode *vp; 2345 struct file *fp; 2346 int error; 2347 2348 /* getvnode() will use the descriptor for us */ 2349 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2350 return (error); 2351 vp = (struct vnode *)fp->f_data; 2352 error = change_flags(vp, SCARG(uap, flags), l); 2353 VOP_UNLOCK(vp, 0); 2354 FILE_UNUSE(fp, l); 2355 return (error); 2356 } 2357 2358 /* 2359 * Change flags of a file given a path name; this version does 2360 * not follow links. 2361 */ 2362 int 2363 sys_lchflags(l, v, retval) 2364 struct lwp *l; 2365 void *v; 2366 register_t *retval; 2367 { 2368 struct sys_lchflags_args /* { 2369 syscallarg(const char *) path; 2370 syscallarg(u_long) flags; 2371 } */ *uap = v; 2372 struct vnode *vp; 2373 int error; 2374 struct nameidata nd; 2375 2376 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2377 if ((error = namei(&nd)) != 0) 2378 return (error); 2379 vp = nd.ni_vp; 2380 error = change_flags(vp, SCARG(uap, flags), l); 2381 vput(vp); 2382 return (error); 2383 } 2384 2385 /* 2386 * Common routine to change flags of a file. 2387 */ 2388 int 2389 change_flags(vp, flags, l) 2390 struct vnode *vp; 2391 u_long flags; 2392 struct lwp *l; 2393 { 2394 struct proc *p = l->l_proc; 2395 struct mount *mp; 2396 struct vattr vattr; 2397 int error; 2398 2399 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2400 return (error); 2401 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2402 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2403 /* 2404 * Non-superusers cannot change the flags on devices, even if they 2405 * own them. 2406 */ 2407 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2408 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 2409 goto out; 2410 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2411 error = EINVAL; 2412 goto out; 2413 } 2414 } 2415 VATTR_NULL(&vattr); 2416 vattr.va_flags = flags; 2417 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2418 out: 2419 vn_finished_write(mp, 0); 2420 return (error); 2421 } 2422 2423 /* 2424 * Change mode of a file given path name; this version follows links. 2425 */ 2426 /* ARGSUSED */ 2427 int 2428 sys_chmod(l, v, retval) 2429 struct lwp *l; 2430 void *v; 2431 register_t *retval; 2432 { 2433 struct sys_chmod_args /* { 2434 syscallarg(const char *) path; 2435 syscallarg(int) mode; 2436 } */ *uap = v; 2437 int error; 2438 struct nameidata nd; 2439 2440 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2441 if ((error = namei(&nd)) != 0) 2442 return (error); 2443 2444 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2445 2446 vrele(nd.ni_vp); 2447 return (error); 2448 } 2449 2450 /* 2451 * Change mode of a file given a file descriptor. 2452 */ 2453 /* ARGSUSED */ 2454 int 2455 sys_fchmod(l, v, retval) 2456 struct lwp *l; 2457 void *v; 2458 register_t *retval; 2459 { 2460 struct sys_fchmod_args /* { 2461 syscallarg(int) fd; 2462 syscallarg(int) mode; 2463 } */ *uap = v; 2464 struct proc *p = l->l_proc; 2465 struct file *fp; 2466 int error; 2467 2468 /* getvnode() will use the descriptor for us */ 2469 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2470 return (error); 2471 2472 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2473 FILE_UNUSE(fp, l); 2474 return (error); 2475 } 2476 2477 /* 2478 * Change mode of a file given path name; this version does not follow links. 2479 */ 2480 /* ARGSUSED */ 2481 int 2482 sys_lchmod(l, v, retval) 2483 struct lwp *l; 2484 void *v; 2485 register_t *retval; 2486 { 2487 struct sys_lchmod_args /* { 2488 syscallarg(const char *) path; 2489 syscallarg(int) mode; 2490 } */ *uap = v; 2491 int error; 2492 struct nameidata nd; 2493 2494 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2495 if ((error = namei(&nd)) != 0) 2496 return (error); 2497 2498 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2499 2500 vrele(nd.ni_vp); 2501 return (error); 2502 } 2503 2504 /* 2505 * Common routine to set mode given a vnode. 2506 */ 2507 static int 2508 change_mode(vp, mode, l) 2509 struct vnode *vp; 2510 int mode; 2511 struct lwp *l; 2512 { 2513 struct proc *p = l->l_proc; 2514 struct mount *mp; 2515 struct vattr vattr; 2516 int error; 2517 2518 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2519 return (error); 2520 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2521 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2522 VATTR_NULL(&vattr); 2523 vattr.va_mode = mode & ALLPERMS; 2524 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2525 VOP_UNLOCK(vp, 0); 2526 vn_finished_write(mp, 0); 2527 return (error); 2528 } 2529 2530 /* 2531 * Set ownership given a path name; this version follows links. 2532 */ 2533 /* ARGSUSED */ 2534 int 2535 sys_chown(l, v, retval) 2536 struct lwp *l; 2537 void *v; 2538 register_t *retval; 2539 { 2540 struct sys_chown_args /* { 2541 syscallarg(const char *) path; 2542 syscallarg(uid_t) uid; 2543 syscallarg(gid_t) gid; 2544 } */ *uap = v; 2545 int error; 2546 struct nameidata nd; 2547 2548 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2549 if ((error = namei(&nd)) != 0) 2550 return (error); 2551 2552 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2553 2554 vrele(nd.ni_vp); 2555 return (error); 2556 } 2557 2558 /* 2559 * Set ownership given a path name; this version follows links. 2560 * Provides POSIX semantics. 2561 */ 2562 /* ARGSUSED */ 2563 int 2564 sys___posix_chown(l, v, retval) 2565 struct lwp *l; 2566 void *v; 2567 register_t *retval; 2568 { 2569 struct sys_chown_args /* { 2570 syscallarg(const char *) path; 2571 syscallarg(uid_t) uid; 2572 syscallarg(gid_t) gid; 2573 } */ *uap = v; 2574 int error; 2575 struct nameidata nd; 2576 2577 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2578 if ((error = namei(&nd)) != 0) 2579 return (error); 2580 2581 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2582 2583 vrele(nd.ni_vp); 2584 return (error); 2585 } 2586 2587 /* 2588 * Set ownership given a file descriptor. 2589 */ 2590 /* ARGSUSED */ 2591 int 2592 sys_fchown(l, v, retval) 2593 struct lwp *l; 2594 void *v; 2595 register_t *retval; 2596 { 2597 struct sys_fchown_args /* { 2598 syscallarg(int) fd; 2599 syscallarg(uid_t) uid; 2600 syscallarg(gid_t) gid; 2601 } */ *uap = v; 2602 struct proc *p = l->l_proc; 2603 int error; 2604 struct file *fp; 2605 2606 /* getvnode() will use the descriptor for us */ 2607 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2608 return (error); 2609 2610 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2611 SCARG(uap, gid), l, 0); 2612 FILE_UNUSE(fp, l); 2613 return (error); 2614 } 2615 2616 /* 2617 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2618 */ 2619 /* ARGSUSED */ 2620 int 2621 sys___posix_fchown(l, v, retval) 2622 struct lwp *l; 2623 void *v; 2624 register_t *retval; 2625 { 2626 struct sys_fchown_args /* { 2627 syscallarg(int) fd; 2628 syscallarg(uid_t) uid; 2629 syscallarg(gid_t) gid; 2630 } */ *uap = v; 2631 struct proc *p = l->l_proc; 2632 int error; 2633 struct file *fp; 2634 2635 /* getvnode() will use the descriptor for us */ 2636 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2637 return (error); 2638 2639 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2640 SCARG(uap, gid), l, 1); 2641 FILE_UNUSE(fp, l); 2642 return (error); 2643 } 2644 2645 /* 2646 * Set ownership given a path name; this version does not follow links. 2647 */ 2648 /* ARGSUSED */ 2649 int 2650 sys_lchown(l, v, retval) 2651 struct lwp *l; 2652 void *v; 2653 register_t *retval; 2654 { 2655 struct sys_lchown_args /* { 2656 syscallarg(const char *) path; 2657 syscallarg(uid_t) uid; 2658 syscallarg(gid_t) gid; 2659 } */ *uap = v; 2660 int error; 2661 struct nameidata nd; 2662 2663 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2664 if ((error = namei(&nd)) != 0) 2665 return (error); 2666 2667 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2668 2669 vrele(nd.ni_vp); 2670 return (error); 2671 } 2672 2673 /* 2674 * Set ownership given a path name; this version does not follow links. 2675 * Provides POSIX/XPG semantics. 2676 */ 2677 /* ARGSUSED */ 2678 int 2679 sys___posix_lchown(l, v, retval) 2680 struct lwp *l; 2681 void *v; 2682 register_t *retval; 2683 { 2684 struct sys_lchown_args /* { 2685 syscallarg(const char *) path; 2686 syscallarg(uid_t) uid; 2687 syscallarg(gid_t) gid; 2688 } */ *uap = v; 2689 int error; 2690 struct nameidata nd; 2691 2692 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2693 if ((error = namei(&nd)) != 0) 2694 return (error); 2695 2696 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2697 2698 vrele(nd.ni_vp); 2699 return (error); 2700 } 2701 2702 /* 2703 * Common routine to set ownership given a vnode. 2704 */ 2705 static int 2706 change_owner(vp, uid, gid, l, posix_semantics) 2707 struct vnode *vp; 2708 uid_t uid; 2709 gid_t gid; 2710 struct lwp *l; 2711 int posix_semantics; 2712 { 2713 struct proc *p = l->l_proc; 2714 struct mount *mp; 2715 struct vattr vattr; 2716 mode_t newmode; 2717 int error; 2718 2719 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2720 return (error); 2721 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2722 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2723 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 2724 goto out; 2725 2726 #define CHANGED(x) ((int)(x) != -1) 2727 newmode = vattr.va_mode; 2728 if (posix_semantics) { 2729 /* 2730 * POSIX/XPG semantics: if the caller is not the super-user, 2731 * clear set-user-id and set-group-id bits. Both POSIX and 2732 * the XPG consider the behaviour for calls by the super-user 2733 * implementation-defined; we leave the set-user-id and set- 2734 * group-id settings intact in that case. 2735 */ 2736 if (suser(p->p_ucred, NULL) != 0) 2737 newmode &= ~(S_ISUID | S_ISGID); 2738 } else { 2739 /* 2740 * NetBSD semantics: when changing owner and/or group, 2741 * clear the respective bit(s). 2742 */ 2743 if (CHANGED(uid)) 2744 newmode &= ~S_ISUID; 2745 if (CHANGED(gid)) 2746 newmode &= ~S_ISGID; 2747 } 2748 /* Update va_mode iff altered. */ 2749 if (vattr.va_mode == newmode) 2750 newmode = VNOVAL; 2751 2752 VATTR_NULL(&vattr); 2753 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2754 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2755 vattr.va_mode = newmode; 2756 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2757 #undef CHANGED 2758 2759 out: 2760 VOP_UNLOCK(vp, 0); 2761 vn_finished_write(mp, 0); 2762 return (error); 2763 } 2764 2765 /* 2766 * Set the access and modification times given a path name; this 2767 * version follows links. 2768 */ 2769 /* ARGSUSED */ 2770 int 2771 sys_utimes(l, v, retval) 2772 struct lwp *l; 2773 void *v; 2774 register_t *retval; 2775 { 2776 struct sys_utimes_args /* { 2777 syscallarg(const char *) path; 2778 syscallarg(const struct timeval *) tptr; 2779 } */ *uap = v; 2780 int error; 2781 struct nameidata nd; 2782 2783 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2784 if ((error = namei(&nd)) != 0) 2785 return (error); 2786 2787 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2788 2789 vrele(nd.ni_vp); 2790 return (error); 2791 } 2792 2793 /* 2794 * Set the access and modification times given a file descriptor. 2795 */ 2796 /* ARGSUSED */ 2797 int 2798 sys_futimes(l, v, retval) 2799 struct lwp *l; 2800 void *v; 2801 register_t *retval; 2802 { 2803 struct sys_futimes_args /* { 2804 syscallarg(int) fd; 2805 syscallarg(const struct timeval *) tptr; 2806 } */ *uap = v; 2807 struct proc *p = l->l_proc; 2808 int error; 2809 struct file *fp; 2810 2811 /* getvnode() will use the descriptor for us */ 2812 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2813 return (error); 2814 2815 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2816 FILE_UNUSE(fp, l); 2817 return (error); 2818 } 2819 2820 /* 2821 * Set the access and modification times given a path name; this 2822 * version does not follow links. 2823 */ 2824 /* ARGSUSED */ 2825 int 2826 sys_lutimes(l, v, retval) 2827 struct lwp *l; 2828 void *v; 2829 register_t *retval; 2830 { 2831 struct sys_lutimes_args /* { 2832 syscallarg(const char *) path; 2833 syscallarg(const struct timeval *) tptr; 2834 } */ *uap = v; 2835 int error; 2836 struct nameidata nd; 2837 2838 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2839 if ((error = namei(&nd)) != 0) 2840 return (error); 2841 2842 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2843 2844 vrele(nd.ni_vp); 2845 return (error); 2846 } 2847 2848 /* 2849 * Common routine to set access and modification times given a vnode. 2850 */ 2851 static int 2852 change_utimes(vp, tptr, l) 2853 struct vnode *vp; 2854 const struct timeval *tptr; 2855 struct lwp *l; 2856 { 2857 struct proc *p = l->l_proc; 2858 struct timeval tv[2]; 2859 struct mount *mp; 2860 struct vattr vattr; 2861 int error; 2862 2863 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2864 return (error); 2865 VATTR_NULL(&vattr); 2866 if (tptr == NULL) { 2867 microtime(&tv[0]); 2868 tv[1] = tv[0]; 2869 vattr.va_vaflags |= VA_UTIMES_NULL; 2870 } else { 2871 error = copyin(tptr, tv, sizeof(tv)); 2872 if (error) 2873 goto out; 2874 } 2875 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2876 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2877 vattr.va_atime.tv_sec = tv[0].tv_sec; 2878 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2879 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2880 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2881 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2882 VOP_UNLOCK(vp, 0); 2883 out: 2884 vn_finished_write(mp, 0); 2885 return (error); 2886 } 2887 2888 /* 2889 * Truncate a file given its path name. 2890 */ 2891 /* ARGSUSED */ 2892 int 2893 sys_truncate(l, v, retval) 2894 struct lwp *l; 2895 void *v; 2896 register_t *retval; 2897 { 2898 struct sys_truncate_args /* { 2899 syscallarg(const char *) path; 2900 syscallarg(int) pad; 2901 syscallarg(off_t) length; 2902 } */ *uap = v; 2903 struct proc *p = l->l_proc; 2904 struct vnode *vp; 2905 struct mount *mp; 2906 struct vattr vattr; 2907 int error; 2908 struct nameidata nd; 2909 2910 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2911 if ((error = namei(&nd)) != 0) 2912 return (error); 2913 vp = nd.ni_vp; 2914 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2915 vrele(vp); 2916 return (error); 2917 } 2918 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2919 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2920 if (vp->v_type == VDIR) 2921 error = EISDIR; 2922 else if ((error = vn_writechk(vp)) == 0 && 2923 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, l)) == 0) { 2924 VATTR_NULL(&vattr); 2925 vattr.va_size = SCARG(uap, length); 2926 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2927 } 2928 vput(vp); 2929 vn_finished_write(mp, 0); 2930 return (error); 2931 } 2932 2933 /* 2934 * Truncate a file given a file descriptor. 2935 */ 2936 /* ARGSUSED */ 2937 int 2938 sys_ftruncate(l, v, retval) 2939 struct lwp *l; 2940 void *v; 2941 register_t *retval; 2942 { 2943 struct sys_ftruncate_args /* { 2944 syscallarg(int) fd; 2945 syscallarg(int) pad; 2946 syscallarg(off_t) length; 2947 } */ *uap = v; 2948 struct proc *p = l->l_proc; 2949 struct mount *mp; 2950 struct vattr vattr; 2951 struct vnode *vp; 2952 struct file *fp; 2953 int error; 2954 2955 /* getvnode() will use the descriptor for us */ 2956 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2957 return (error); 2958 if ((fp->f_flag & FWRITE) == 0) { 2959 error = EINVAL; 2960 goto out; 2961 } 2962 vp = (struct vnode *)fp->f_data; 2963 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2964 FILE_UNUSE(fp, l); 2965 return (error); 2966 } 2967 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2968 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2969 if (vp->v_type == VDIR) 2970 error = EISDIR; 2971 else if ((error = vn_writechk(vp)) == 0) { 2972 VATTR_NULL(&vattr); 2973 vattr.va_size = SCARG(uap, length); 2974 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 2975 } 2976 VOP_UNLOCK(vp, 0); 2977 vn_finished_write(mp, 0); 2978 out: 2979 FILE_UNUSE(fp, l); 2980 return (error); 2981 } 2982 2983 /* 2984 * Sync an open file. 2985 */ 2986 /* ARGSUSED */ 2987 int 2988 sys_fsync(l, v, retval) 2989 struct lwp *l; 2990 void *v; 2991 register_t *retval; 2992 { 2993 struct sys_fsync_args /* { 2994 syscallarg(int) fd; 2995 } */ *uap = v; 2996 struct proc *p = l->l_proc; 2997 struct vnode *vp; 2998 struct mount *mp; 2999 struct file *fp; 3000 int error; 3001 3002 /* getvnode() will use the descriptor for us */ 3003 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3004 return (error); 3005 vp = (struct vnode *)fp->f_data; 3006 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3007 FILE_UNUSE(fp, l); 3008 return (error); 3009 } 3010 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3011 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3012 if (error == 0 && bioops.io_fsync != NULL && 3013 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3014 (*bioops.io_fsync)(vp); 3015 VOP_UNLOCK(vp, 0); 3016 vn_finished_write(mp, 0); 3017 FILE_UNUSE(fp, l); 3018 return (error); 3019 } 3020 3021 /* 3022 * Sync a range of file data. API modeled after that found in AIX. 3023 * 3024 * FDATASYNC indicates that we need only save enough metadata to be able 3025 * to re-read the written data. Note we duplicate AIX's requirement that 3026 * the file be open for writing. 3027 */ 3028 /* ARGSUSED */ 3029 int 3030 sys_fsync_range(l, v, retval) 3031 struct lwp *l; 3032 void *v; 3033 register_t *retval; 3034 { 3035 struct sys_fsync_range_args /* { 3036 syscallarg(int) fd; 3037 syscallarg(int) flags; 3038 syscallarg(off_t) start; 3039 syscallarg(int) length; 3040 } */ *uap = v; 3041 struct proc *p = l->l_proc; 3042 struct vnode *vp; 3043 struct file *fp; 3044 int flags, nflags; 3045 off_t s, e, len; 3046 int error; 3047 3048 /* getvnode() will use the descriptor for us */ 3049 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3050 return (error); 3051 3052 if ((fp->f_flag & FWRITE) == 0) { 3053 FILE_UNUSE(fp, l); 3054 return (EBADF); 3055 } 3056 3057 flags = SCARG(uap, flags); 3058 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3059 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3060 return (EINVAL); 3061 } 3062 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3063 if (flags & FDATASYNC) 3064 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3065 else 3066 nflags = FSYNC_WAIT; 3067 3068 len = SCARG(uap, length); 3069 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3070 if (len) { 3071 s = SCARG(uap, start); 3072 e = s + len; 3073 if (e < s) { 3074 FILE_UNUSE(fp, l); 3075 return (EINVAL); 3076 } 3077 } else { 3078 e = 0; 3079 s = 0; 3080 } 3081 3082 vp = (struct vnode *)fp->f_data; 3083 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3084 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3085 3086 if (error == 0 && bioops.io_fsync != NULL && 3087 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3088 (*bioops.io_fsync)(vp); 3089 3090 VOP_UNLOCK(vp, 0); 3091 FILE_UNUSE(fp, l); 3092 return (error); 3093 } 3094 3095 /* 3096 * Sync the data of an open file. 3097 */ 3098 /* ARGSUSED */ 3099 int 3100 sys_fdatasync(l, v, retval) 3101 struct lwp *l; 3102 void *v; 3103 register_t *retval; 3104 { 3105 struct sys_fdatasync_args /* { 3106 syscallarg(int) fd; 3107 } */ *uap = v; 3108 struct proc *p = l->l_proc; 3109 struct vnode *vp; 3110 struct file *fp; 3111 int error; 3112 3113 /* getvnode() will use the descriptor for us */ 3114 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3115 return (error); 3116 if ((fp->f_flag & FWRITE) == 0) { 3117 FILE_UNUSE(fp, l); 3118 return (EBADF); 3119 } 3120 vp = (struct vnode *)fp->f_data; 3121 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3122 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3123 VOP_UNLOCK(vp, 0); 3124 FILE_UNUSE(fp, l); 3125 return (error); 3126 } 3127 3128 /* 3129 * Rename files, (standard) BSD semantics frontend. 3130 */ 3131 /* ARGSUSED */ 3132 int 3133 sys_rename(l, v, retval) 3134 struct lwp *l; 3135 void *v; 3136 register_t *retval; 3137 { 3138 struct sys_rename_args /* { 3139 syscallarg(const char *) from; 3140 syscallarg(const char *) to; 3141 } */ *uap = v; 3142 3143 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3144 } 3145 3146 /* 3147 * Rename files, POSIX semantics frontend. 3148 */ 3149 /* ARGSUSED */ 3150 int 3151 sys___posix_rename(l, v, retval) 3152 struct lwp *l; 3153 void *v; 3154 register_t *retval; 3155 { 3156 struct sys___posix_rename_args /* { 3157 syscallarg(const char *) from; 3158 syscallarg(const char *) to; 3159 } */ *uap = v; 3160 3161 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3162 } 3163 3164 /* 3165 * Rename files. Source and destination must either both be directories, 3166 * or both not be directories. If target is a directory, it must be empty. 3167 * If `from' and `to' refer to the same object, the value of the `retain' 3168 * argument is used to determine whether `from' will be 3169 * 3170 * (retain == 0) deleted unless `from' and `to' refer to the same 3171 * object in the file system's name space (BSD). 3172 * (retain == 1) always retained (POSIX). 3173 */ 3174 static int 3175 rename_files(from, to, l, retain) 3176 const char *from, *to; 3177 struct lwp *l; 3178 int retain; 3179 { 3180 struct mount *mp = NULL; 3181 struct vnode *tvp, *fvp, *tdvp; 3182 struct nameidata fromnd, tond; 3183 struct proc *p; 3184 int error; 3185 3186 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3187 from, l); 3188 if ((error = namei(&fromnd)) != 0) 3189 return (error); 3190 fvp = fromnd.ni_vp; 3191 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3192 if (error != 0) { 3193 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3194 vrele(fromnd.ni_dvp); 3195 vrele(fvp); 3196 if (fromnd.ni_startdir) 3197 vrele(fromnd.ni_startdir); 3198 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3199 return (error); 3200 } 3201 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3202 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3203 if ((error = namei(&tond)) != 0) { 3204 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3205 vrele(fromnd.ni_dvp); 3206 vrele(fvp); 3207 goto out1; 3208 } 3209 tdvp = tond.ni_dvp; 3210 tvp = tond.ni_vp; 3211 3212 if (tvp != NULL) { 3213 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3214 error = ENOTDIR; 3215 goto out; 3216 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3217 error = EISDIR; 3218 goto out; 3219 } 3220 } 3221 3222 if (fvp == tdvp) 3223 error = EINVAL; 3224 3225 /* 3226 * Source and destination refer to the same object. 3227 */ 3228 if (fvp == tvp) { 3229 if (retain) 3230 error = -1; 3231 else if (fromnd.ni_dvp == tdvp && 3232 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3233 !memcmp(fromnd.ni_cnd.cn_nameptr, 3234 tond.ni_cnd.cn_nameptr, 3235 fromnd.ni_cnd.cn_namelen)) 3236 error = -1; 3237 } 3238 3239 out: 3240 p = l->l_proc; 3241 if (!error) { 3242 VOP_LEASE(tdvp, l, p->p_ucred, LEASE_WRITE); 3243 if (fromnd.ni_dvp != tdvp) 3244 VOP_LEASE(fromnd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3245 if (tvp) { 3246 VOP_LEASE(tvp, l, p->p_ucred, LEASE_WRITE); 3247 } 3248 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3249 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3250 } else { 3251 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3252 if (tdvp == tvp) 3253 vrele(tdvp); 3254 else 3255 vput(tdvp); 3256 if (tvp) 3257 vput(tvp); 3258 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3259 vrele(fromnd.ni_dvp); 3260 vrele(fvp); 3261 } 3262 vrele(tond.ni_startdir); 3263 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3264 out1: 3265 vn_finished_write(mp, 0); 3266 if (fromnd.ni_startdir) 3267 vrele(fromnd.ni_startdir); 3268 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3269 return (error == -1 ? 0 : error); 3270 } 3271 3272 /* 3273 * Make a directory file. 3274 */ 3275 /* ARGSUSED */ 3276 int 3277 sys_mkdir(l, v, retval) 3278 struct lwp *l; 3279 void *v; 3280 register_t *retval; 3281 { 3282 struct sys_mkdir_args /* { 3283 syscallarg(const char *) path; 3284 syscallarg(int) mode; 3285 } */ *uap = v; 3286 struct proc *p = l->l_proc; 3287 struct mount *mp; 3288 struct vnode *vp; 3289 struct vattr vattr; 3290 int error; 3291 struct nameidata nd; 3292 3293 restart: 3294 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3295 SCARG(uap, path), l); 3296 if ((error = namei(&nd)) != 0) 3297 return (error); 3298 vp = nd.ni_vp; 3299 if (vp != NULL) { 3300 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3301 if (nd.ni_dvp == vp) 3302 vrele(nd.ni_dvp); 3303 else 3304 vput(nd.ni_dvp); 3305 vrele(vp); 3306 return (EEXIST); 3307 } 3308 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3309 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3310 if (nd.ni_dvp == vp) 3311 vrele(nd.ni_dvp); 3312 else 3313 vput(nd.ni_dvp); 3314 if ((error = vn_start_write(NULL, &mp, 3315 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3316 return (error); 3317 goto restart; 3318 } 3319 VATTR_NULL(&vattr); 3320 vattr.va_type = VDIR; 3321 vattr.va_mode = 3322 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3323 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3324 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3325 if (!error) 3326 vput(nd.ni_vp); 3327 vn_finished_write(mp, 0); 3328 return (error); 3329 } 3330 3331 /* 3332 * Remove a directory file. 3333 */ 3334 /* ARGSUSED */ 3335 int 3336 sys_rmdir(l, v, retval) 3337 struct lwp *l; 3338 void *v; 3339 register_t *retval; 3340 { 3341 struct sys_rmdir_args /* { 3342 syscallarg(const char *) path; 3343 } */ *uap = v; 3344 struct proc *p = l->l_proc; 3345 struct mount *mp; 3346 struct vnode *vp; 3347 int error; 3348 struct nameidata nd; 3349 3350 restart: 3351 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3352 SCARG(uap, path), l); 3353 if ((error = namei(&nd)) != 0) 3354 return (error); 3355 vp = nd.ni_vp; 3356 if (vp->v_type != VDIR) { 3357 error = ENOTDIR; 3358 goto out; 3359 } 3360 /* 3361 * No rmdir "." please. 3362 */ 3363 if (nd.ni_dvp == vp) { 3364 error = EINVAL; 3365 goto out; 3366 } 3367 /* 3368 * The root of a mounted filesystem cannot be deleted. 3369 */ 3370 if (vp->v_flag & VROOT) { 3371 error = EBUSY; 3372 goto out; 3373 } 3374 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3375 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3376 if (nd.ni_dvp == vp) 3377 vrele(nd.ni_dvp); 3378 else 3379 vput(nd.ni_dvp); 3380 vput(vp); 3381 if ((error = vn_start_write(NULL, &mp, 3382 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3383 return (error); 3384 goto restart; 3385 } 3386 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3387 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 3388 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3389 vn_finished_write(mp, 0); 3390 return (error); 3391 3392 out: 3393 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3394 if (nd.ni_dvp == vp) 3395 vrele(nd.ni_dvp); 3396 else 3397 vput(nd.ni_dvp); 3398 vput(vp); 3399 return (error); 3400 } 3401 3402 /* 3403 * Read a block of directory entries in a file system independent format. 3404 */ 3405 int 3406 sys_getdents(l, v, retval) 3407 struct lwp *l; 3408 void *v; 3409 register_t *retval; 3410 { 3411 struct sys_getdents_args /* { 3412 syscallarg(int) fd; 3413 syscallarg(char *) buf; 3414 syscallarg(size_t) count; 3415 } */ *uap = v; 3416 struct proc *p = l->l_proc; 3417 struct file *fp; 3418 int error, done; 3419 3420 /* getvnode() will use the descriptor for us */ 3421 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3422 return (error); 3423 if ((fp->f_flag & FREAD) == 0) { 3424 error = EBADF; 3425 goto out; 3426 } 3427 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3428 SCARG(uap, count), &done, l, 0, 0); 3429 #ifdef KTRACE 3430 if (!error && KTRPOINT(p, KTR_GENIO)) { 3431 struct iovec iov; 3432 iov.iov_base = SCARG(uap, buf); 3433 iov.iov_len = done; 3434 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3435 } 3436 #endif 3437 *retval = done; 3438 out: 3439 FILE_UNUSE(fp, l); 3440 return (error); 3441 } 3442 3443 /* 3444 * Set the mode mask for creation of filesystem nodes. 3445 */ 3446 int 3447 sys_umask(l, v, retval) 3448 struct lwp *l; 3449 void *v; 3450 register_t *retval; 3451 { 3452 struct sys_umask_args /* { 3453 syscallarg(mode_t) newmask; 3454 } */ *uap = v; 3455 struct proc *p = l->l_proc; 3456 struct cwdinfo *cwdi; 3457 3458 cwdi = p->p_cwdi; 3459 *retval = cwdi->cwdi_cmask; 3460 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3461 return (0); 3462 } 3463 3464 /* 3465 * Void all references to file by ripping underlying filesystem 3466 * away from vnode. 3467 */ 3468 /* ARGSUSED */ 3469 int 3470 sys_revoke(l, v, retval) 3471 struct lwp *l; 3472 void *v; 3473 register_t *retval; 3474 { 3475 struct sys_revoke_args /* { 3476 syscallarg(const char *) path; 3477 } */ *uap = v; 3478 struct proc *p = l->l_proc; 3479 struct mount *mp; 3480 struct vnode *vp; 3481 struct vattr vattr; 3482 int error; 3483 struct nameidata nd; 3484 3485 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3486 if ((error = namei(&nd)) != 0) 3487 return (error); 3488 vp = nd.ni_vp; 3489 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 3490 goto out; 3491 if (p->p_ucred->cr_uid != vattr.va_uid && 3492 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3493 goto out; 3494 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3495 goto out; 3496 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3497 VOP_REVOKE(vp, REVOKEALL); 3498 vn_finished_write(mp, 0); 3499 out: 3500 vrele(vp); 3501 return (error); 3502 } 3503 3504 /* 3505 * Convert a user file descriptor to a kernel file entry. 3506 */ 3507 int 3508 getvnode(fdp, fd, fpp) 3509 struct filedesc *fdp; 3510 int fd; 3511 struct file **fpp; 3512 { 3513 struct vnode *vp; 3514 struct file *fp; 3515 3516 if ((fp = fd_getfile(fdp, fd)) == NULL) 3517 return (EBADF); 3518 3519 FILE_USE(fp); 3520 3521 if (fp->f_type != DTYPE_VNODE) { 3522 FILE_UNUSE(fp, NULL); 3523 return (EINVAL); 3524 } 3525 3526 vp = (struct vnode *)fp->f_data; 3527 if (vp->v_type == VBAD) { 3528 FILE_UNUSE(fp, NULL); 3529 return (EBADF); 3530 } 3531 3532 *fpp = fp; 3533 return (0); 3534 } 3535
Indexes created Fri Oct 03 02:09:57 GMT 2025