vfs_syscalls.c revision 1.203
1 /* $NetBSD: vfs_syscalls.c,v 1.203 2004/02/22 09:56:26 jdolecek 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 1.203 2004/02/22 09:56:26 jdolecek 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 __P((struct nameidata *, struct proc *)); 77 static int change_flags __P((struct vnode *, u_long, struct proc *)); 78 static int change_mode __P((struct vnode *, int, struct proc *p)); 79 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *, 80 int)); 81 static int change_utimes __P((struct vnode *vp, const struct timeval *, 82 struct proc *p)); 83 static int rename_files __P((const char *, const char *, struct proc *, int)); 84 static int dostatfs __P((struct mount *, struct statfs *, struct proc *, int, 85 int)); 86 87 void checkdirs __P((struct vnode *)); 88 89 int dovfsusermount = 0; 90 91 /* 92 * Virtual File System System Calls 93 */ 94 95 /* 96 * Mount a file system. 97 */ 98 99 #if defined(COMPAT_09) || defined(COMPAT_43) 100 /* 101 * This table is used to maintain compatibility with 4.3BSD 102 * and NetBSD 0.9 mount syscalls. Note, the order is important! 103 * 104 * Do not modify this table. It should only contain filesystems 105 * supported by NetBSD 0.9 and 4.3BSD. 106 */ 107 const char * const mountcompatnames[] = { 108 NULL, /* 0 = MOUNT_NONE */ 109 MOUNT_FFS, /* 1 = MOUNT_UFS */ 110 MOUNT_NFS, /* 2 */ 111 MOUNT_MFS, /* 3 */ 112 MOUNT_MSDOS, /* 4 */ 113 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 114 MOUNT_FDESC, /* 6 */ 115 MOUNT_KERNFS, /* 7 */ 116 NULL, /* 8 = MOUNT_DEVFS */ 117 MOUNT_AFS, /* 9 */ 118 }; 119 const int nmountcompatnames = sizeof(mountcompatnames) / 120 sizeof(mountcompatnames[0]); 121 #endif /* COMPAT_09 || COMPAT_43 */ 122 123 /* ARGSUSED */ 124 int 125 sys_mount(l, v, retval) 126 struct lwp *l; 127 void *v; 128 register_t *retval; 129 { 130 struct sys_mount_args /* { 131 syscallarg(const char *) type; 132 syscallarg(const char *) path; 133 syscallarg(int) flags; 134 syscallarg(void *) data; 135 } */ *uap = v; 136 struct proc *p = l->l_proc; 137 struct vnode *vp; 138 struct mount *mp; 139 int error, flag = 0; 140 char fstypename[MFSNAMELEN]; 141 struct vattr va; 142 struct nameidata nd; 143 struct vfsops *vfs; 144 145 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 146 (error = suser(p->p_ucred, &p->p_acflag))) 147 return (error); 148 /* 149 * Get vnode to be covered 150 */ 151 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 152 SCARG(uap, path), p); 153 if ((error = namei(&nd)) != 0) 154 return (error); 155 vp = nd.ni_vp; 156 /* 157 * A lookup in VFS_MOUNT might result in an attempt to 158 * lock this vnode again, so make the lock recursive. 159 */ 160 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 161 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 162 if ((vp->v_flag & VROOT) == 0) { 163 vput(vp); 164 return (EINVAL); 165 } 166 mp = vp->v_mount; 167 flag = mp->mnt_flag; 168 vfs = mp->mnt_op; 169 /* 170 * We only allow the filesystem to be reloaded if it 171 * is currently mounted read-only. 172 */ 173 if ((SCARG(uap, flags) & MNT_RELOAD) && 174 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 175 vput(vp); 176 return (EOPNOTSUPP); /* Needs translation */ 177 } 178 /* 179 * In "highly secure" mode, don't let the caller do anything 180 * but downgrade a filesystem from read-write to read-only. 181 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 182 */ 183 if (securelevel >= 2 && 184 SCARG(uap, flags) != MNT_GETARGS && 185 SCARG(uap, flags) != 186 (mp->mnt_flag | MNT_RDONLY | 187 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 188 vput(vp); 189 return (EPERM); 190 } 191 mp->mnt_flag |= SCARG(uap, flags) & 192 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 193 /* 194 * Only root, or the user that did the original mount is 195 * permitted to update it. 196 */ 197 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 198 mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 199 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 200 vput(vp); 201 return (error); 202 } 203 /* 204 * Do not allow NFS export by non-root users. For non-root 205 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 206 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 207 */ 208 if (p->p_ucred->cr_uid != 0) { 209 if (SCARG(uap, flags) & MNT_EXPORTED) { 210 vput(vp); 211 return (EPERM); 212 } 213 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 214 if (flag & MNT_NOEXEC) 215 SCARG(uap, flags) |= MNT_NOEXEC; 216 } 217 if (vfs_busy(mp, LK_NOWAIT, 0)) { 218 vput(vp); 219 return (EPERM); 220 } 221 goto update; 222 } else { 223 if (securelevel >= 2) { 224 vput(vp); 225 return (EPERM); 226 } 227 } 228 /* 229 * If the user is not root, ensure that they own the directory 230 * onto which we are attempting to mount. 231 */ 232 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 || 233 (va.va_uid != p->p_ucred->cr_uid && 234 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) { 235 vput(vp); 236 return (error); 237 } 238 /* 239 * Do not allow NFS export by non-root users. For non-root users, 240 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 241 * mount point is already MNT_NOEXEC. 242 */ 243 if (p->p_ucred->cr_uid != 0) { 244 if (SCARG(uap, flags) & MNT_EXPORTED) { 245 vput(vp); 246 return (EPERM); 247 } 248 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 249 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 250 SCARG(uap, flags) |= MNT_NOEXEC; 251 } 252 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) { 253 vput(vp); 254 return (error); 255 } 256 if (vp->v_type != VDIR) { 257 vput(vp); 258 return (ENOTDIR); 259 } 260 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 261 if (error) { 262 #if defined(COMPAT_09) || defined(COMPAT_43) 263 /* 264 * Historically filesystem types were identified by number. 265 * If we get an integer for the filesystem type instead of a 266 * string, we check to see if it matches one of the historic 267 * filesystem types. 268 */ 269 u_long fsindex = (u_long)SCARG(uap, type); 270 if (fsindex >= nmountcompatnames || 271 mountcompatnames[fsindex] == NULL) { 272 vput(vp); 273 return (ENODEV); 274 } 275 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 276 #else 277 vput(vp); 278 return (error); 279 #endif 280 } 281 #ifdef COMPAT_10 282 /* Accept `ufs' as an alias for `ffs'. */ 283 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 284 strncpy(fstypename, "ffs", MFSNAMELEN); 285 #endif 286 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 287 vput(vp); 288 return (ENODEV); 289 } 290 if (vp->v_mountedhere != NULL) { 291 vput(vp); 292 return (EBUSY); 293 } 294 295 /* 296 * Allocate and initialize the file system. 297 */ 298 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 299 M_MOUNT, M_WAITOK); 300 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 301 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 302 (void)vfs_busy(mp, LK_NOWAIT, 0); 303 mp->mnt_op = vfs; 304 vfs->vfs_refcount++; 305 mp->mnt_vnodecovered = vp; 306 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 307 mp->mnt_unmounter = NULL; 308 309 /* 310 * The underlying file system may refuse the mount for 311 * various reasons. Allow the user to force it to happen. 312 */ 313 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 314 update: 315 /* 316 * Set the mount level flags. 317 */ 318 if (SCARG(uap, flags) & MNT_RDONLY) 319 mp->mnt_flag |= MNT_RDONLY; 320 else if (mp->mnt_flag & MNT_RDONLY) 321 mp->mnt_iflag |= IMNT_WANTRDWR; 322 mp->mnt_flag &= 323 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 324 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 325 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 326 mp->mnt_flag |= SCARG(uap, flags) & 327 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 328 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 329 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 330 MNT_IGNORE); 331 /* 332 * Mount the filesystem. 333 */ 334 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 335 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 336 if (mp->mnt_iflag & IMNT_WANTRDWR) 337 mp->mnt_flag &= ~MNT_RDONLY; 338 if (error || (mp->mnt_flag & MNT_GETARGS)) 339 mp->mnt_flag = flag; 340 mp->mnt_flag &=~ 341 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 342 mp->mnt_iflag &=~ IMNT_WANTRDWR; 343 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 344 if (mp->mnt_syncer == NULL) 345 error = vfs_allocate_syncvnode(mp); 346 } else { 347 if (mp->mnt_syncer != NULL) 348 vfs_deallocate_syncvnode(mp); 349 } 350 vfs_unbusy(mp); 351 VOP_UNLOCK(vp, 0); 352 vrele(vp); 353 return (error); 354 } 355 /* 356 * Put the new filesystem on the mount list after root. 357 */ 358 cache_purge(vp); 359 if (!error) { 360 mp->mnt_flag &=~ 361 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 362 mp->mnt_iflag &=~ IMNT_WANTRDWR; 363 vp->v_mountedhere = mp; 364 simple_lock(&mountlist_slock); 365 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 366 simple_unlock(&mountlist_slock); 367 checkdirs(vp); 368 VOP_UNLOCK(vp, 0); 369 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 370 error = vfs_allocate_syncvnode(mp); 371 vfs_unbusy(mp); 372 (void) VFS_STATFS(mp, &mp->mnt_stat, p); 373 if ((error = VFS_START(mp, 0, p))) 374 vrele(vp); 375 } else { 376 vp->v_mountedhere = (struct mount *)0; 377 vfs->vfs_refcount--; 378 vfs_unbusy(mp); 379 free(mp, M_MOUNT); 380 vput(vp); 381 } 382 return (error); 383 } 384 385 /* 386 * Scan all active processes to see if any of them have a current 387 * or root directory onto which the new filesystem has just been 388 * mounted. If so, replace them with the new mount point. 389 */ 390 void 391 checkdirs(olddp) 392 struct vnode *olddp; 393 { 394 struct cwdinfo *cwdi; 395 struct vnode *newdp; 396 struct proc *p; 397 398 if (olddp->v_usecount == 1) 399 return; 400 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 401 panic("mount: lost mount"); 402 proclist_lock_read(); 403 LIST_FOREACH(p, &allproc, p_list) { 404 cwdi = p->p_cwdi; 405 if (cwdi->cwdi_cdir == olddp) { 406 vrele(cwdi->cwdi_cdir); 407 VREF(newdp); 408 cwdi->cwdi_cdir = newdp; 409 } 410 if (cwdi->cwdi_rdir == olddp) { 411 vrele(cwdi->cwdi_rdir); 412 VREF(newdp); 413 cwdi->cwdi_rdir = newdp; 414 } 415 } 416 proclist_unlock_read(); 417 if (rootvnode == olddp) { 418 vrele(rootvnode); 419 VREF(newdp); 420 rootvnode = newdp; 421 } 422 vput(newdp); 423 } 424 425 /* 426 * Unmount a file system. 427 * 428 * Note: unmount takes a path to the vnode mounted on as argument, 429 * not special file (as before). 430 */ 431 /* ARGSUSED */ 432 int 433 sys_unmount(l, v, retval) 434 struct lwp *l; 435 void *v; 436 register_t *retval; 437 { 438 struct sys_unmount_args /* { 439 syscallarg(const char *) path; 440 syscallarg(int) flags; 441 } */ *uap = v; 442 struct proc *p = l->l_proc; 443 struct vnode *vp; 444 struct mount *mp; 445 int error; 446 struct nameidata nd; 447 448 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 449 SCARG(uap, path), p); 450 if ((error = namei(&nd)) != 0) 451 return (error); 452 vp = nd.ni_vp; 453 mp = vp->v_mount; 454 455 /* 456 * Only root, or the user that did the original mount is 457 * permitted to unmount this filesystem. 458 */ 459 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 460 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 461 vput(vp); 462 return (error); 463 } 464 465 /* 466 * Don't allow unmounting the root file system. 467 */ 468 if (mp->mnt_flag & MNT_ROOTFS) { 469 vput(vp); 470 return (EINVAL); 471 } 472 473 /* 474 * Must be the root of the filesystem 475 */ 476 if ((vp->v_flag & VROOT) == 0) { 477 vput(vp); 478 return (EINVAL); 479 } 480 vput(vp); 481 482 /* 483 * XXX Freeze syncer. Must do this before locking the 484 * mount point. See dounmount() for details. 485 */ 486 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 487 488 if (vfs_busy(mp, 0, 0)) { 489 lockmgr(&syncer_lock, LK_RELEASE, NULL); 490 return (EBUSY); 491 } 492 493 return (dounmount(mp, SCARG(uap, flags), p)); 494 } 495 496 /* 497 * Do the actual file system unmount. File system is assumed to have been 498 * marked busy by the caller. 499 */ 500 int 501 dounmount(mp, flags, p) 502 struct mount *mp; 503 int flags; 504 struct proc *p; 505 { 506 struct vnode *coveredvp; 507 int error; 508 int async; 509 int used_syncer; 510 511 simple_lock(&mountlist_slock); 512 vfs_unbusy(mp); 513 used_syncer = (mp->mnt_syncer != NULL); 514 515 /* 516 * XXX Syncer must be frozen when we get here. This should really 517 * be done on a per-mountpoint basis, but especially the softdep 518 * code possibly called from the syncer doens't exactly work on a 519 * per-mountpoint basis, so the softdep code would become a maze 520 * of vfs_busy() calls. 521 * 522 * The caller of dounmount() must acquire syncer_lock because 523 * the syncer itself acquires locks in syncer_lock -> vfs_busy 524 * order, and we must preserve that order to avoid deadlock. 525 * 526 * So, if the file system did not use the syncer, now is 527 * the time to release the syncer_lock. 528 */ 529 if (used_syncer == 0) 530 lockmgr(&syncer_lock, LK_RELEASE, NULL); 531 532 mp->mnt_iflag |= IMNT_UNMOUNT; 533 mp->mnt_unmounter = p; 534 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 535 vn_start_write(NULL, &mp, V_WAIT); 536 537 if (mp->mnt_flag & MNT_EXPUBLIC) 538 vfs_setpublicfs(NULL, NULL, NULL); 539 async = mp->mnt_flag & MNT_ASYNC; 540 mp->mnt_flag &= ~MNT_ASYNC; 541 cache_purgevfs(mp); /* remove cache entries for this file sys */ 542 if (mp->mnt_syncer != NULL) 543 vfs_deallocate_syncvnode(mp); 544 if (((mp->mnt_flag & MNT_RDONLY) || 545 #if NFSS > 0 546 (error = fss_umount_hook(mp, (flags & MNT_FORCE))) == 0 || 547 #endif 548 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) || 549 (flags & MNT_FORCE)) 550 error = VFS_UNMOUNT(mp, flags, p); 551 vn_finished_write(mp, 0); 552 simple_lock(&mountlist_slock); 553 if (error) { 554 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 555 (void) vfs_allocate_syncvnode(mp); 556 mp->mnt_iflag &= ~IMNT_UNMOUNT; 557 mp->mnt_unmounter = NULL; 558 mp->mnt_flag |= async; 559 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 560 &mountlist_slock); 561 if (used_syncer) 562 lockmgr(&syncer_lock, LK_RELEASE, NULL); 563 while (mp->mnt_wcnt > 0) { 564 wakeup(mp); 565 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0); 566 } 567 return (error); 568 } 569 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 570 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 571 coveredvp->v_mountedhere = NULL; 572 vrele(coveredvp); 573 } 574 mp->mnt_op->vfs_refcount--; 575 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 576 panic("unmount: dangling vnode"); 577 mp->mnt_iflag |= IMNT_GONE; 578 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 579 if (used_syncer) 580 lockmgr(&syncer_lock, LK_RELEASE, NULL); 581 while(mp->mnt_wcnt > 0) { 582 wakeup(mp); 583 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0); 584 } 585 free(mp, M_MOUNT); 586 return (0); 587 } 588 589 /* 590 * Sync each mounted filesystem. 591 */ 592 #ifdef DEBUG 593 int syncprt = 0; 594 struct ctldebug debug0 = { "syncprt", &syncprt }; 595 #endif 596 597 /* ARGSUSED */ 598 int 599 sys_sync(l, v, retval) 600 struct lwp *l; 601 void *v; 602 register_t *retval; 603 { 604 struct mount *mp, *nmp; 605 int asyncflag; 606 struct proc *p = l == NULL ? &proc0 : l->l_proc; 607 608 simple_lock(&mountlist_slock); 609 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 610 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 611 nmp = mp->mnt_list.cqe_prev; 612 continue; 613 } 614 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 615 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 616 asyncflag = mp->mnt_flag & MNT_ASYNC; 617 mp->mnt_flag &= ~MNT_ASYNC; 618 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p); 619 if (asyncflag) 620 mp->mnt_flag |= MNT_ASYNC; 621 vn_finished_write(mp, 0); 622 } 623 simple_lock(&mountlist_slock); 624 nmp = mp->mnt_list.cqe_prev; 625 vfs_unbusy(mp); 626 627 } 628 simple_unlock(&mountlist_slock); 629 #ifdef DEBUG 630 if (syncprt) 631 vfs_bufstats(); 632 #endif /* DEBUG */ 633 return (0); 634 } 635 636 /* 637 * Change filesystem quotas. 638 */ 639 /* ARGSUSED */ 640 int 641 sys_quotactl(l, v, retval) 642 struct lwp *l; 643 void *v; 644 register_t *retval; 645 { 646 struct sys_quotactl_args /* { 647 syscallarg(const char *) path; 648 syscallarg(int) cmd; 649 syscallarg(int) uid; 650 syscallarg(caddr_t) arg; 651 } */ *uap = v; 652 struct proc *p = l->l_proc; 653 struct mount *mp; 654 int error; 655 struct nameidata nd; 656 657 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 658 if ((error = namei(&nd)) != 0) 659 return (error); 660 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 661 vrele(nd.ni_vp); 662 if (error) 663 return (error); 664 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 665 SCARG(uap, arg), p); 666 vn_finished_write(mp, 0); 667 return (error); 668 } 669 670 static int 671 dostatfs(struct mount *mp, struct statfs *sp, struct proc *p, int flags, 672 int root) 673 { 674 struct cwdinfo *cwdi = p->p_cwdi; 675 int error = 0; 676 677 /* 678 * If MNT_NOWAIT or MNT_LAZY is specified, do not 679 * refresh the fsstat cache. MNT_WAIT or MNT_LAXY 680 * overrides MNT_NOWAIT. 681 */ 682 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 683 (flags != MNT_WAIT && flags != 0)) { 684 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 685 goto done; 686 } 687 688 if ((error = VFS_STATFS(mp, sp, p)) != 0) { 689 return error; 690 } 691 692 if (cwdi->cwdi_rdir == NULL) 693 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 694 done: 695 if (cwdi->cwdi_rdir != NULL) { 696 size_t len; 697 char *bp; 698 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 699 if (!path) 700 return ENOMEM; 701 702 bp = path + MAXPATHLEN; 703 *--bp = '\0'; 704 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 705 MAXPATHLEN / 2, 0, p); 706 if (error) { 707 free(path, M_TEMP); 708 return error; 709 } 710 len = strlen(bp); 711 /* 712 * for mount points that are below our root, we can see 713 * them, so we fix up the pathname and return them. The 714 * rest we cannot see, so we don't allow viewing the 715 * data. 716 */ 717 if (strncmp(bp, sp->f_mntonname, len) == 0) { 718 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 719 sizeof(sp->f_mntonname)); 720 if (sp->f_mntonname[0] == '\0') 721 (void)strlcpy(sp->f_mntonname, "/", 722 sizeof(sp->f_mntonname)); 723 } else { 724 if (root) 725 (void)strlcpy(sp->f_mntonname, "/", 726 sizeof(sp->f_mntonname)); 727 else 728 error = EPERM; 729 } 730 free(path, M_TEMP); 731 } 732 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 733 sp->f_oflags = sp->f_flags & 0xffff; 734 return error; 735 } 736 737 /* 738 * Get filesystem statistics. 739 */ 740 /* ARGSUSED */ 741 int 742 sys_statfs(l, v, retval) 743 struct lwp *l; 744 void *v; 745 register_t *retval; 746 { 747 struct sys_statfs_args /* { 748 syscallarg(const char *) path; 749 syscallarg(struct statfs *) buf; 750 } */ *uap = v; 751 struct proc *p = l->l_proc; 752 struct mount *mp; 753 struct statfs sbuf; 754 int error; 755 struct nameidata nd; 756 757 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 758 if ((error = namei(&nd)) != 0) 759 return error; 760 mp = nd.ni_vp->v_mount; 761 vrele(nd.ni_vp); 762 if ((error = dostatfs(mp, &sbuf, p, 0, 1)) != 0) 763 return error; 764 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 765 } 766 767 /* 768 * Get filesystem statistics. 769 */ 770 /* ARGSUSED */ 771 int 772 sys_fstatfs(l, v, retval) 773 struct lwp *l; 774 void *v; 775 register_t *retval; 776 { 777 struct sys_fstatfs_args /* { 778 syscallarg(int) fd; 779 syscallarg(struct statfs *) buf; 780 } */ *uap = v; 781 struct proc *p = l->l_proc; 782 struct file *fp; 783 struct mount *mp; 784 struct statfs sbuf; 785 int error; 786 787 /* getvnode() will use the descriptor for us */ 788 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 789 return (error); 790 mp = ((struct vnode *)fp->f_data)->v_mount; 791 if ((error = dostatfs(mp, &sbuf, p, 0, 1)) != 0) 792 goto out; 793 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 794 out: 795 FILE_UNUSE(fp, p); 796 return error; 797 } 798 799 800 /* 801 * Get statistics on all filesystems. 802 */ 803 int 804 sys_getfsstat(l, v, retval) 805 struct lwp *l; 806 void *v; 807 register_t *retval; 808 { 809 struct sys_getfsstat_args /* { 810 syscallarg(struct statfs *) buf; 811 syscallarg(long) bufsize; 812 syscallarg(int) flags; 813 } */ *uap = v; 814 int root = 0; 815 struct proc *p = l->l_proc; 816 struct mount *mp, *nmp; 817 struct statfs sbuf; 818 caddr_t sfsp; 819 long count, maxcount, error = 0; 820 821 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs); 822 sfsp = (caddr_t)SCARG(uap, buf); 823 simple_lock(&mountlist_slock); 824 count = 0; 825 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 826 mp = nmp) { 827 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 828 nmp = CIRCLEQ_NEXT(mp, mnt_list); 829 continue; 830 } 831 if (sfsp && count < maxcount) { 832 error = dostatfs(mp, &sbuf, p, SCARG(uap, flags), 0); 833 if (error) { 834 simple_lock(&mountlist_slock); 835 nmp = CIRCLEQ_NEXT(mp, mnt_list); 836 vfs_unbusy(mp); 837 continue; 838 } 839 error = copyout(&sbuf, sfsp, sizeof(sbuf)); 840 if (error) { 841 vfs_unbusy(mp); 842 return (error); 843 } 844 sfsp += sizeof(sbuf); 845 root |= strcmp(sbuf.f_mntonname, "/") == 0; 846 } 847 count++; 848 simple_lock(&mountlist_slock); 849 nmp = CIRCLEQ_NEXT(mp, mnt_list); 850 vfs_unbusy(mp); 851 } 852 simple_unlock(&mountlist_slock); 853 if (root == 0 && p->p_cwdi->cwdi_rdir) { 854 /* 855 * fake a root entry 856 */ 857 if ((error = dostatfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p, 858 SCARG(uap, flags), 1)) != 0) 859 return error; 860 if (sfsp) 861 error = copyout(&sbuf, sfsp, sizeof(sbuf)); 862 count++; 863 } 864 if (sfsp && count > maxcount) 865 *retval = maxcount; 866 else 867 *retval = count; 868 return error; 869 } 870 871 /* 872 * Change current working directory to a given file descriptor. 873 */ 874 /* ARGSUSED */ 875 int 876 sys_fchdir(l, v, retval) 877 struct lwp *l; 878 void *v; 879 register_t *retval; 880 { 881 struct sys_fchdir_args /* { 882 syscallarg(int) fd; 883 } */ *uap = v; 884 struct proc *p = l->l_proc; 885 struct filedesc *fdp = p->p_fd; 886 struct cwdinfo *cwdi = p->p_cwdi; 887 struct vnode *vp, *tdp; 888 struct mount *mp; 889 struct file *fp; 890 int error; 891 892 /* getvnode() will use the descriptor for us */ 893 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 894 return (error); 895 vp = (struct vnode *)fp->f_data; 896 897 VREF(vp); 898 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 899 if (vp->v_type != VDIR) 900 error = ENOTDIR; 901 else 902 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 903 while (!error && (mp = vp->v_mountedhere) != NULL) { 904 if (vfs_busy(mp, 0, 0)) 905 continue; 906 error = VFS_ROOT(mp, &tdp); 907 vfs_unbusy(mp); 908 if (error) 909 break; 910 vput(vp); 911 vp = tdp; 912 } 913 if (error) { 914 vput(vp); 915 goto out; 916 } 917 VOP_UNLOCK(vp, 0); 918 919 /* 920 * Disallow changing to a directory not under the process's 921 * current root directory (if there is one). 922 */ 923 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) { 924 vrele(vp); 925 error = EPERM; /* operation not permitted */ 926 goto out; 927 } 928 929 vrele(cwdi->cwdi_cdir); 930 cwdi->cwdi_cdir = vp; 931 out: 932 FILE_UNUSE(fp, p); 933 return (error); 934 } 935 936 /* 937 * Change this process's notion of the root directory to a given file descriptor. 938 */ 939 940 int 941 sys_fchroot(l, v, retval) 942 struct lwp *l; 943 void *v; 944 register_t *retval; 945 { 946 struct sys_fchroot_args *uap = v; 947 struct proc *p = l->l_proc; 948 struct filedesc *fdp = p->p_fd; 949 struct cwdinfo *cwdi = p->p_cwdi; 950 struct vnode *vp; 951 struct file *fp; 952 int error; 953 954 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 955 return error; 956 /* getvnode() will use the descriptor for us */ 957 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 958 return error; 959 vp = (struct vnode *) fp->f_data; 960 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 961 if (vp->v_type != VDIR) 962 error = ENOTDIR; 963 else 964 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 965 VOP_UNLOCK(vp, 0); 966 if (error) 967 goto out; 968 VREF(vp); 969 970 /* 971 * Prevent escaping from chroot by putting the root under 972 * the working directory. Silently chdir to / if we aren't 973 * already there. 974 */ 975 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 976 /* 977 * XXX would be more failsafe to change directory to a 978 * deadfs node here instead 979 */ 980 vrele(cwdi->cwdi_cdir); 981 VREF(vp); 982 cwdi->cwdi_cdir = vp; 983 } 984 985 if (cwdi->cwdi_rdir != NULL) 986 vrele(cwdi->cwdi_rdir); 987 cwdi->cwdi_rdir = vp; 988 out: 989 FILE_UNUSE(fp, p); 990 return (error); 991 } 992 993 994 995 /* 996 * Change current working directory (``.''). 997 */ 998 /* ARGSUSED */ 999 int 1000 sys_chdir(l, v, retval) 1001 struct lwp *l; 1002 void *v; 1003 register_t *retval; 1004 { 1005 struct sys_chdir_args /* { 1006 syscallarg(const char *) path; 1007 } */ *uap = v; 1008 struct proc *p = l->l_proc; 1009 struct cwdinfo *cwdi = p->p_cwdi; 1010 int error; 1011 struct nameidata nd; 1012 1013 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1014 SCARG(uap, path), p); 1015 if ((error = change_dir(&nd, p)) != 0) 1016 return (error); 1017 vrele(cwdi->cwdi_cdir); 1018 cwdi->cwdi_cdir = nd.ni_vp; 1019 return (0); 1020 } 1021 1022 /* 1023 * Change notion of root (``/'') directory. 1024 */ 1025 /* ARGSUSED */ 1026 int 1027 sys_chroot(l, v, retval) 1028 struct lwp *l; 1029 void *v; 1030 register_t *retval; 1031 { 1032 struct sys_chroot_args /* { 1033 syscallarg(const char *) path; 1034 } */ *uap = v; 1035 struct proc *p = l->l_proc; 1036 struct cwdinfo *cwdi = p->p_cwdi; 1037 struct vnode *vp; 1038 int error; 1039 struct nameidata nd; 1040 1041 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1042 return (error); 1043 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1044 SCARG(uap, path), p); 1045 if ((error = change_dir(&nd, p)) != 0) 1046 return (error); 1047 if (cwdi->cwdi_rdir != NULL) 1048 vrele(cwdi->cwdi_rdir); 1049 vp = nd.ni_vp; 1050 cwdi->cwdi_rdir = vp; 1051 1052 /* 1053 * Prevent escaping from chroot by putting the root under 1054 * the working directory. Silently chdir to / if we aren't 1055 * already there. 1056 */ 1057 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 1058 /* 1059 * XXX would be more failsafe to change directory to a 1060 * deadfs node here instead 1061 */ 1062 vrele(cwdi->cwdi_cdir); 1063 VREF(vp); 1064 cwdi->cwdi_cdir = vp; 1065 } 1066 1067 return (0); 1068 } 1069 1070 /* 1071 * Common routine for chroot and chdir. 1072 */ 1073 static int 1074 change_dir(ndp, p) 1075 struct nameidata *ndp; 1076 struct proc *p; 1077 { 1078 struct vnode *vp; 1079 int error; 1080 1081 if ((error = namei(ndp)) != 0) 1082 return (error); 1083 vp = ndp->ni_vp; 1084 if (vp->v_type != VDIR) 1085 error = ENOTDIR; 1086 else 1087 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 1088 1089 if (error) 1090 vput(vp); 1091 else 1092 VOP_UNLOCK(vp, 0); 1093 return (error); 1094 } 1095 1096 /* 1097 * Check permissions, allocate an open file structure, 1098 * and call the device open routine if any. 1099 */ 1100 int 1101 sys_open(l, v, retval) 1102 struct lwp *l; 1103 void *v; 1104 register_t *retval; 1105 { 1106 struct sys_open_args /* { 1107 syscallarg(const char *) path; 1108 syscallarg(int) flags; 1109 syscallarg(int) mode; 1110 } */ *uap = v; 1111 struct proc *p = l->l_proc; 1112 struct cwdinfo *cwdi = p->p_cwdi; 1113 struct filedesc *fdp = p->p_fd; 1114 struct file *fp; 1115 struct vnode *vp; 1116 int flags, cmode; 1117 int type, indx, error; 1118 struct flock lf; 1119 struct nameidata nd; 1120 1121 flags = FFLAGS(SCARG(uap, flags)); 1122 if ((flags & (FREAD | FWRITE)) == 0) 1123 return (EINVAL); 1124 /* falloc() will use the file descriptor for us */ 1125 if ((error = falloc(p, &fp, &indx)) != 0) 1126 return (error); 1127 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1128 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1129 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1130 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1131 FILE_UNUSE(fp, p); 1132 fdp->fd_ofiles[indx] = NULL; 1133 ffree(fp); 1134 if ((error == ENODEV || error == ENXIO) && 1135 l->l_dupfd >= 0 && /* XXX from fdopen */ 1136 (error = 1137 dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) { 1138 *retval = indx; 1139 return (0); 1140 } 1141 if (error == ERESTART) 1142 error = EINTR; 1143 fdremove(fdp, indx); 1144 return (error); 1145 } 1146 l->l_dupfd = 0; 1147 vp = nd.ni_vp; 1148 fp->f_flag = flags & FMASK; 1149 fp->f_type = DTYPE_VNODE; 1150 fp->f_ops = &vnops; 1151 fp->f_data = vp; 1152 if (flags & (O_EXLOCK | O_SHLOCK)) { 1153 lf.l_whence = SEEK_SET; 1154 lf.l_start = 0; 1155 lf.l_len = 0; 1156 if (flags & O_EXLOCK) 1157 lf.l_type = F_WRLCK; 1158 else 1159 lf.l_type = F_RDLCK; 1160 type = F_FLOCK; 1161 if ((flags & FNONBLOCK) == 0) 1162 type |= F_WAIT; 1163 VOP_UNLOCK(vp, 0); 1164 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1165 if (error) { 1166 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1167 FILE_UNUSE(fp, p); 1168 ffree(fp); 1169 fdremove(fdp, indx); 1170 return (error); 1171 } 1172 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1173 fp->f_flag |= FHASLOCK; 1174 } 1175 VOP_UNLOCK(vp, 0); 1176 *retval = indx; 1177 FILE_SET_MATURE(fp); 1178 FILE_UNUSE(fp, p); 1179 return (0); 1180 } 1181 1182 /* 1183 * Get file handle system call 1184 */ 1185 int 1186 sys_getfh(l, v, retval) 1187 struct lwp *l; 1188 void *v; 1189 register_t *retval; 1190 { 1191 struct sys_getfh_args /* { 1192 syscallarg(char *) fname; 1193 syscallarg(fhandle_t *) fhp; 1194 } */ *uap = v; 1195 struct proc *p = l->l_proc; 1196 struct vnode *vp; 1197 fhandle_t fh; 1198 int error; 1199 struct nameidata nd; 1200 1201 /* 1202 * Must be super user 1203 */ 1204 error = suser(p->p_ucred, &p->p_acflag); 1205 if (error) 1206 return (error); 1207 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1208 SCARG(uap, fname), p); 1209 error = namei(&nd); 1210 if (error) 1211 return (error); 1212 vp = nd.ni_vp; 1213 memset(&fh, 0, sizeof(fh)); 1214 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 1215 error = VFS_VPTOFH(vp, &fh.fh_fid); 1216 vput(vp); 1217 if (error) 1218 return (error); 1219 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1220 return (error); 1221 } 1222 1223 /* 1224 * Open a file given a file handle. 1225 * 1226 * Check permissions, allocate an open file structure, 1227 * and call the device open routine if any. 1228 */ 1229 int 1230 sys_fhopen(l, v, retval) 1231 struct lwp *l; 1232 void *v; 1233 register_t *retval; 1234 { 1235 struct sys_fhopen_args /* { 1236 syscallarg(const fhandle_t *) fhp; 1237 syscallarg(int) flags; 1238 } */ *uap = v; 1239 struct proc *p = l->l_proc; 1240 struct filedesc *fdp = p->p_fd; 1241 struct file *fp; 1242 struct vnode *vp = NULL; 1243 struct mount *mp; 1244 struct ucred *cred = p->p_ucred; 1245 int flags; 1246 struct file *nfp; 1247 int type, indx, error=0; 1248 struct flock lf; 1249 struct vattr va; 1250 fhandle_t fh; 1251 1252 /* 1253 * Must be super user 1254 */ 1255 if ((error = suser(p->p_ucred, &p->p_acflag))) 1256 return (error); 1257 1258 flags = FFLAGS(SCARG(uap, flags)); 1259 if ((flags & (FREAD | FWRITE)) == 0) 1260 return (EINVAL); 1261 if ((flags & O_CREAT)) 1262 return (EINVAL); 1263 /* falloc() will use the file descriptor for us */ 1264 if ((error = falloc(p, &nfp, &indx)) != 0) 1265 return (error); 1266 fp = nfp; 1267 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1268 goto bad; 1269 1270 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1271 error = ESTALE; 1272 goto bad; 1273 } 1274 1275 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1276 vp = NULL; /* most likely unnecessary sanity for bad: */ 1277 goto bad; 1278 } 1279 1280 /* Now do an effective vn_open */ 1281 1282 if (vp->v_type == VSOCK) { 1283 error = EOPNOTSUPP; 1284 goto bad; 1285 } 1286 if (flags & FREAD) { 1287 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 1288 goto bad; 1289 } 1290 if (flags & (FWRITE | O_TRUNC)) { 1291 if (vp->v_type == VDIR) { 1292 error = EISDIR; 1293 goto bad; 1294 } 1295 if ((error = vn_writechk(vp)) != 0 || 1296 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 1297 goto bad; 1298 } 1299 if (flags & O_TRUNC) { 1300 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1301 goto bad; 1302 VOP_UNLOCK(vp, 0); /* XXX */ 1303 VOP_LEASE(vp, p, cred, LEASE_WRITE); 1304 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1305 VATTR_NULL(&va); 1306 va.va_size = 0; 1307 error = VOP_SETATTR(vp, &va, cred, p); 1308 vn_finished_write(mp, 0); 1309 if (error) 1310 goto bad; 1311 } 1312 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) 1313 goto bad; 1314 if (vp->v_type == VREG && 1315 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1316 error = EIO; 1317 goto bad; 1318 } 1319 if (flags & FWRITE) 1320 vp->v_writecount++; 1321 1322 /* done with modified vn_open, now finish what sys_open does. */ 1323 1324 fp->f_flag = flags & FMASK; 1325 fp->f_type = DTYPE_VNODE; 1326 fp->f_ops = &vnops; 1327 fp->f_data = vp; 1328 if (flags & (O_EXLOCK | O_SHLOCK)) { 1329 lf.l_whence = SEEK_SET; 1330 lf.l_start = 0; 1331 lf.l_len = 0; 1332 if (flags & O_EXLOCK) 1333 lf.l_type = F_WRLCK; 1334 else 1335 lf.l_type = F_RDLCK; 1336 type = F_FLOCK; 1337 if ((flags & FNONBLOCK) == 0) 1338 type |= F_WAIT; 1339 VOP_UNLOCK(vp, 0); 1340 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1341 if (error) { 1342 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1343 FILE_UNUSE(fp, p); 1344 ffree(fp); 1345 fdremove(fdp, indx); 1346 return (error); 1347 } 1348 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1349 fp->f_flag |= FHASLOCK; 1350 } 1351 VOP_UNLOCK(vp, 0); 1352 *retval = indx; 1353 FILE_SET_MATURE(fp); 1354 FILE_UNUSE(fp, p); 1355 return (0); 1356 1357 bad: 1358 FILE_UNUSE(fp, p); 1359 ffree(fp); 1360 fdremove(fdp, indx); 1361 if (vp != NULL) 1362 vput(vp); 1363 return (error); 1364 } 1365 1366 /* ARGSUSED */ 1367 int 1368 sys_fhstat(l, v, retval) 1369 struct lwp *l; 1370 void *v; 1371 register_t *retval; 1372 { 1373 struct sys_fhstat_args /* { 1374 syscallarg(const fhandle_t *) fhp; 1375 syscallarg(struct stat *) sb; 1376 } */ *uap = v; 1377 struct proc *p = l->l_proc; 1378 struct stat sb; 1379 int error; 1380 fhandle_t fh; 1381 struct mount *mp; 1382 struct vnode *vp; 1383 1384 /* 1385 * Must be super user 1386 */ 1387 if ((error = suser(p->p_ucred, &p->p_acflag))) 1388 return (error); 1389 1390 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1391 return (error); 1392 1393 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1394 return (ESTALE); 1395 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1396 return (error); 1397 error = vn_stat(vp, &sb, p); 1398 vput(vp); 1399 if (error) 1400 return (error); 1401 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1402 return (error); 1403 } 1404 1405 /* ARGSUSED */ 1406 int 1407 sys_fhstatfs(l, v, retval) 1408 struct lwp *l; 1409 void *v; 1410 register_t *retval; 1411 { 1412 struct sys_fhstatfs_args /* 1413 syscallarg(const fhandle_t *) fhp; 1414 syscallarg(struct statfs *) buf; 1415 } */ *uap = v; 1416 struct proc *p = l->l_proc; 1417 struct statfs sbuf; 1418 fhandle_t fh; 1419 struct mount *mp; 1420 struct vnode *vp; 1421 int error; 1422 1423 /* 1424 * Must be super user 1425 */ 1426 if ((error = suser(p->p_ucred, &p->p_acflag))) 1427 return (error); 1428 1429 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1430 return (error); 1431 1432 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1433 return (ESTALE); 1434 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1435 return (error); 1436 mp = vp->v_mount; 1437 vput(vp); 1438 if ((error = VFS_STATFS(mp, &sbuf, p)) != 0) 1439 return (error); 1440 return (copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf))); 1441 } 1442 1443 /* 1444 * Create a special file. 1445 */ 1446 /* ARGSUSED */ 1447 int 1448 sys_mknod(l, v, retval) 1449 struct lwp *l; 1450 void *v; 1451 register_t *retval; 1452 { 1453 struct sys_mknod_args /* { 1454 syscallarg(const char *) path; 1455 syscallarg(int) mode; 1456 syscallarg(int) dev; 1457 } */ *uap = v; 1458 struct proc *p = l->l_proc; 1459 struct vnode *vp; 1460 struct mount *mp; 1461 struct vattr vattr; 1462 int error; 1463 int whiteout = 0; 1464 struct nameidata nd; 1465 1466 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1467 return (error); 1468 restart: 1469 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1470 if ((error = namei(&nd)) != 0) 1471 return (error); 1472 vp = nd.ni_vp; 1473 if (vp != NULL) 1474 error = EEXIST; 1475 else { 1476 VATTR_NULL(&vattr); 1477 vattr.va_mode = 1478 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1479 vattr.va_rdev = SCARG(uap, dev); 1480 whiteout = 0; 1481 1482 switch (SCARG(uap, mode) & S_IFMT) { 1483 case S_IFMT: /* used by badsect to flag bad sectors */ 1484 vattr.va_type = VBAD; 1485 break; 1486 case S_IFCHR: 1487 vattr.va_type = VCHR; 1488 break; 1489 case S_IFBLK: 1490 vattr.va_type = VBLK; 1491 break; 1492 case S_IFWHT: 1493 whiteout = 1; 1494 break; 1495 default: 1496 error = EINVAL; 1497 break; 1498 } 1499 } 1500 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 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 if ((error = vn_start_write(NULL, &mp, 1509 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1510 return (error); 1511 goto restart; 1512 } 1513 if (!error) { 1514 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1515 if (whiteout) { 1516 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1517 if (error) 1518 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1519 vput(nd.ni_dvp); 1520 } else { 1521 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1522 &nd.ni_cnd, &vattr); 1523 if (error == 0) 1524 vput(nd.ni_vp); 1525 } 1526 } else { 1527 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1528 if (nd.ni_dvp == vp) 1529 vrele(nd.ni_dvp); 1530 else 1531 vput(nd.ni_dvp); 1532 if (vp) 1533 vrele(vp); 1534 } 1535 vn_finished_write(mp, 0); 1536 return (error); 1537 } 1538 1539 /* 1540 * Create a named pipe. 1541 */ 1542 /* ARGSUSED */ 1543 int 1544 sys_mkfifo(l, v, retval) 1545 struct lwp *l; 1546 void *v; 1547 register_t *retval; 1548 { 1549 struct sys_mkfifo_args /* { 1550 syscallarg(const char *) path; 1551 syscallarg(int) mode; 1552 } */ *uap = v; 1553 struct proc *p = l->l_proc; 1554 struct mount *mp; 1555 struct vattr vattr; 1556 int error; 1557 struct nameidata nd; 1558 1559 restart: 1560 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1561 if ((error = namei(&nd)) != 0) 1562 return (error); 1563 if (nd.ni_vp != NULL) { 1564 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1565 if (nd.ni_dvp == nd.ni_vp) 1566 vrele(nd.ni_dvp); 1567 else 1568 vput(nd.ni_dvp); 1569 vrele(nd.ni_vp); 1570 return (EEXIST); 1571 } 1572 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1573 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1574 if (nd.ni_dvp == nd.ni_vp) 1575 vrele(nd.ni_dvp); 1576 else 1577 vput(nd.ni_dvp); 1578 if (nd.ni_vp) 1579 vrele(nd.ni_vp); 1580 if ((error = vn_start_write(NULL, &mp, 1581 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1582 return (error); 1583 goto restart; 1584 } 1585 VATTR_NULL(&vattr); 1586 vattr.va_type = VFIFO; 1587 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1588 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1589 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1590 if (error == 0) 1591 vput(nd.ni_vp); 1592 vn_finished_write(mp, 0); 1593 return (error); 1594 } 1595 1596 /* 1597 * Make a hard file link. 1598 */ 1599 /* ARGSUSED */ 1600 int 1601 sys_link(l, v, retval) 1602 struct lwp *l; 1603 void *v; 1604 register_t *retval; 1605 { 1606 struct sys_link_args /* { 1607 syscallarg(const char *) path; 1608 syscallarg(const char *) link; 1609 } */ *uap = v; 1610 struct proc *p = l->l_proc; 1611 struct vnode *vp; 1612 struct mount *mp; 1613 struct nameidata nd; 1614 int error; 1615 1616 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1617 if ((error = namei(&nd)) != 0) 1618 return (error); 1619 vp = nd.ni_vp; 1620 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1621 vrele(vp); 1622 return (error); 1623 } 1624 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1625 if ((error = namei(&nd)) != 0) 1626 goto out; 1627 if (nd.ni_vp) { 1628 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1629 if (nd.ni_dvp == nd.ni_vp) 1630 vrele(nd.ni_dvp); 1631 else 1632 vput(nd.ni_dvp); 1633 vrele(nd.ni_vp); 1634 error = EEXIST; 1635 goto out; 1636 } 1637 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1638 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1639 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1640 out: 1641 vrele(vp); 1642 vn_finished_write(mp, 0); 1643 return (error); 1644 } 1645 1646 /* 1647 * Make a symbolic link. 1648 */ 1649 /* ARGSUSED */ 1650 int 1651 sys_symlink(l, v, retval) 1652 struct lwp *l; 1653 void *v; 1654 register_t *retval; 1655 { 1656 struct sys_symlink_args /* { 1657 syscallarg(const char *) path; 1658 syscallarg(const char *) link; 1659 } */ *uap = v; 1660 struct proc *p = l->l_proc; 1661 struct mount *mp; 1662 struct vattr vattr; 1663 char *path; 1664 int error; 1665 struct nameidata nd; 1666 1667 path = PNBUF_GET(); 1668 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1669 if (error) 1670 goto out; 1671 restart: 1672 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1673 if ((error = namei(&nd)) != 0) 1674 goto out; 1675 if (nd.ni_vp) { 1676 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1677 if (nd.ni_dvp == nd.ni_vp) 1678 vrele(nd.ni_dvp); 1679 else 1680 vput(nd.ni_dvp); 1681 vrele(nd.ni_vp); 1682 error = EEXIST; 1683 goto out; 1684 } 1685 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1686 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1687 if (nd.ni_dvp == nd.ni_vp) 1688 vrele(nd.ni_dvp); 1689 else 1690 vput(nd.ni_dvp); 1691 if ((error = vn_start_write(NULL, &mp, 1692 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1693 return (error); 1694 goto restart; 1695 } 1696 VATTR_NULL(&vattr); 1697 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1698 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1699 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1700 if (error == 0) 1701 vput(nd.ni_vp); 1702 vn_finished_write(mp, 0); 1703 out: 1704 PNBUF_PUT(path); 1705 return (error); 1706 } 1707 1708 /* 1709 * Delete a whiteout from the filesystem. 1710 */ 1711 /* ARGSUSED */ 1712 int 1713 sys_undelete(l, v, retval) 1714 struct lwp *l; 1715 void *v; 1716 register_t *retval; 1717 { 1718 struct sys_undelete_args /* { 1719 syscallarg(const char *) path; 1720 } */ *uap = v; 1721 struct proc *p = l->l_proc; 1722 int error; 1723 struct mount *mp; 1724 struct nameidata nd; 1725 1726 restart: 1727 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1728 SCARG(uap, path), p); 1729 error = namei(&nd); 1730 if (error) 1731 return (error); 1732 1733 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1734 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1735 if (nd.ni_dvp == nd.ni_vp) 1736 vrele(nd.ni_dvp); 1737 else 1738 vput(nd.ni_dvp); 1739 if (nd.ni_vp) 1740 vrele(nd.ni_vp); 1741 return (EEXIST); 1742 } 1743 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1744 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1745 if (nd.ni_dvp == nd.ni_vp) 1746 vrele(nd.ni_dvp); 1747 else 1748 vput(nd.ni_dvp); 1749 if ((error = vn_start_write(NULL, &mp, 1750 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1751 return (error); 1752 goto restart; 1753 } 1754 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1755 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1756 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1757 vput(nd.ni_dvp); 1758 vn_finished_write(mp, 0); 1759 return (error); 1760 } 1761 1762 /* 1763 * Delete a name from the filesystem. 1764 */ 1765 /* ARGSUSED */ 1766 int 1767 sys_unlink(l, v, retval) 1768 struct lwp *l; 1769 void *v; 1770 register_t *retval; 1771 { 1772 struct sys_unlink_args /* { 1773 syscallarg(const char *) path; 1774 } */ *uap = v; 1775 struct proc *p = l->l_proc; 1776 struct mount *mp; 1777 struct vnode *vp; 1778 int error; 1779 struct nameidata nd; 1780 1781 restart: 1782 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1783 SCARG(uap, path), p); 1784 if ((error = namei(&nd)) != 0) 1785 return (error); 1786 vp = nd.ni_vp; 1787 1788 /* 1789 * The root of a mounted filesystem cannot be deleted. 1790 */ 1791 if (vp->v_flag & VROOT) { 1792 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1793 if (nd.ni_dvp == vp) 1794 vrele(nd.ni_dvp); 1795 else 1796 vput(nd.ni_dvp); 1797 vput(vp); 1798 error = EBUSY; 1799 goto out; 1800 } 1801 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1802 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1803 if (nd.ni_dvp == vp) 1804 vrele(nd.ni_dvp); 1805 else 1806 vput(nd.ni_dvp); 1807 vput(vp); 1808 if ((error = vn_start_write(NULL, &mp, 1809 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1810 return (error); 1811 goto restart; 1812 } 1813 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1814 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1815 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1816 vn_finished_write(mp, 0); 1817 out: 1818 return (error); 1819 } 1820 1821 /* 1822 * Reposition read/write file offset. 1823 */ 1824 int 1825 sys_lseek(l, v, retval) 1826 struct lwp *l; 1827 void *v; 1828 register_t *retval; 1829 { 1830 struct sys_lseek_args /* { 1831 syscallarg(int) fd; 1832 syscallarg(int) pad; 1833 syscallarg(off_t) offset; 1834 syscallarg(int) whence; 1835 } */ *uap = v; 1836 struct proc *p = l->l_proc; 1837 struct ucred *cred = p->p_ucred; 1838 struct filedesc *fdp = p->p_fd; 1839 struct file *fp; 1840 struct vnode *vp; 1841 struct vattr vattr; 1842 off_t newoff; 1843 int error; 1844 1845 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1846 return (EBADF); 1847 1848 FILE_USE(fp); 1849 1850 vp = (struct vnode *)fp->f_data; 1851 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1852 error = ESPIPE; 1853 goto out; 1854 } 1855 1856 switch (SCARG(uap, whence)) { 1857 case SEEK_CUR: 1858 newoff = fp->f_offset + SCARG(uap, offset); 1859 break; 1860 case SEEK_END: 1861 error = VOP_GETATTR(vp, &vattr, cred, p); 1862 if (error) 1863 goto out; 1864 newoff = SCARG(uap, offset) + vattr.va_size; 1865 break; 1866 case SEEK_SET: 1867 newoff = SCARG(uap, offset); 1868 break; 1869 default: 1870 error = EINVAL; 1871 goto out; 1872 } 1873 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1874 goto out; 1875 1876 *(off_t *)retval = fp->f_offset = newoff; 1877 out: 1878 FILE_UNUSE(fp, p); 1879 return (error); 1880 } 1881 1882 /* 1883 * Positional read system call. 1884 */ 1885 int 1886 sys_pread(l, v, retval) 1887 struct lwp *l; 1888 void *v; 1889 register_t *retval; 1890 { 1891 struct sys_pread_args /* { 1892 syscallarg(int) fd; 1893 syscallarg(void *) buf; 1894 syscallarg(size_t) nbyte; 1895 syscallarg(off_t) offset; 1896 } */ *uap = v; 1897 struct proc *p = l->l_proc; 1898 struct filedesc *fdp = p->p_fd; 1899 struct file *fp; 1900 struct vnode *vp; 1901 off_t offset; 1902 int error, fd = SCARG(uap, fd); 1903 1904 if ((fp = fd_getfile(fdp, fd)) == NULL) 1905 return (EBADF); 1906 1907 if ((fp->f_flag & FREAD) == 0) { 1908 simple_unlock(&fp->f_slock); 1909 return (EBADF); 1910 } 1911 1912 FILE_USE(fp); 1913 1914 vp = (struct vnode *)fp->f_data; 1915 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1916 error = ESPIPE; 1917 goto out; 1918 } 1919 1920 offset = SCARG(uap, offset); 1921 1922 /* 1923 * XXX This works because no file systems actually 1924 * XXX take any action on the seek operation. 1925 */ 1926 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1927 goto out; 1928 1929 /* dofileread() will unuse the descriptor for us */ 1930 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1931 &offset, 0, retval)); 1932 1933 out: 1934 FILE_UNUSE(fp, p); 1935 return (error); 1936 } 1937 1938 /* 1939 * Positional scatter read system call. 1940 */ 1941 int 1942 sys_preadv(l, v, retval) 1943 struct lwp *l; 1944 void *v; 1945 register_t *retval; 1946 { 1947 struct sys_preadv_args /* { 1948 syscallarg(int) fd; 1949 syscallarg(const struct iovec *) iovp; 1950 syscallarg(int) iovcnt; 1951 syscallarg(off_t) offset; 1952 } */ *uap = v; 1953 struct proc *p = l->l_proc; 1954 struct filedesc *fdp = p->p_fd; 1955 struct file *fp; 1956 struct vnode *vp; 1957 off_t offset; 1958 int error, fd = SCARG(uap, fd); 1959 1960 if ((fp = fd_getfile(fdp, fd)) == NULL) 1961 return (EBADF); 1962 1963 if ((fp->f_flag & FREAD) == 0) { 1964 simple_unlock(&fp->f_slock); 1965 return (EBADF); 1966 } 1967 1968 FILE_USE(fp); 1969 1970 vp = (struct vnode *)fp->f_data; 1971 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1972 error = ESPIPE; 1973 goto out; 1974 } 1975 1976 offset = SCARG(uap, offset); 1977 1978 /* 1979 * XXX This works because no file systems actually 1980 * XXX take any action on the seek operation. 1981 */ 1982 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1983 goto out; 1984 1985 /* dofilereadv() will unuse the descriptor for us */ 1986 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1987 &offset, 0, retval)); 1988 1989 out: 1990 FILE_UNUSE(fp, p); 1991 return (error); 1992 } 1993 1994 /* 1995 * Positional write system call. 1996 */ 1997 int 1998 sys_pwrite(l, v, retval) 1999 struct lwp *l; 2000 void *v; 2001 register_t *retval; 2002 { 2003 struct sys_pwrite_args /* { 2004 syscallarg(int) fd; 2005 syscallarg(const void *) buf; 2006 syscallarg(size_t) nbyte; 2007 syscallarg(off_t) offset; 2008 } */ *uap = v; 2009 struct proc *p = l->l_proc; 2010 struct filedesc *fdp = p->p_fd; 2011 struct file *fp; 2012 struct vnode *vp; 2013 off_t offset; 2014 int error, fd = SCARG(uap, fd); 2015 2016 if ((fp = fd_getfile(fdp, fd)) == NULL) 2017 return (EBADF); 2018 2019 if ((fp->f_flag & FWRITE) == 0) { 2020 simple_unlock(&fp->f_slock); 2021 return (EBADF); 2022 } 2023 2024 FILE_USE(fp); 2025 2026 vp = (struct vnode *)fp->f_data; 2027 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2028 error = ESPIPE; 2029 goto out; 2030 } 2031 2032 offset = SCARG(uap, offset); 2033 2034 /* 2035 * XXX This works because no file systems actually 2036 * XXX take any action on the seek operation. 2037 */ 2038 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2039 goto out; 2040 2041 /* dofilewrite() will unuse the descriptor for us */ 2042 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2043 &offset, 0, retval)); 2044 2045 out: 2046 FILE_UNUSE(fp, p); 2047 return (error); 2048 } 2049 2050 /* 2051 * Positional gather write system call. 2052 */ 2053 int 2054 sys_pwritev(l, v, retval) 2055 struct lwp *l; 2056 void *v; 2057 register_t *retval; 2058 { 2059 struct sys_pwritev_args /* { 2060 syscallarg(int) fd; 2061 syscallarg(const struct iovec *) iovp; 2062 syscallarg(int) iovcnt; 2063 syscallarg(off_t) offset; 2064 } */ *uap = v; 2065 struct proc *p = l->l_proc; 2066 struct filedesc *fdp = p->p_fd; 2067 struct file *fp; 2068 struct vnode *vp; 2069 off_t offset; 2070 int error, fd = SCARG(uap, fd); 2071 2072 if ((fp = fd_getfile(fdp, fd)) == NULL) 2073 return (EBADF); 2074 2075 if ((fp->f_flag & FWRITE) == 0) { 2076 simple_unlock(&fp->f_slock); 2077 return (EBADF); 2078 } 2079 2080 FILE_USE(fp); 2081 2082 vp = (struct vnode *)fp->f_data; 2083 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2084 error = ESPIPE; 2085 goto out; 2086 } 2087 2088 offset = SCARG(uap, offset); 2089 2090 /* 2091 * XXX This works because no file systems actually 2092 * XXX take any action on the seek operation. 2093 */ 2094 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2095 goto out; 2096 2097 /* dofilewritev() will unuse the descriptor for us */ 2098 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2099 &offset, 0, retval)); 2100 2101 out: 2102 FILE_UNUSE(fp, p); 2103 return (error); 2104 } 2105 2106 /* 2107 * Check access permissions. 2108 */ 2109 int 2110 sys_access(l, v, retval) 2111 struct lwp *l; 2112 void *v; 2113 register_t *retval; 2114 { 2115 struct sys_access_args /* { 2116 syscallarg(const char *) path; 2117 syscallarg(int) flags; 2118 } */ *uap = v; 2119 struct proc *p = l->l_proc; 2120 struct ucred *cred = crget(); 2121 struct vnode *vp; 2122 int error, flags; 2123 struct nameidata nd; 2124 2125 (void)memcpy(cred, p->p_ucred, sizeof(*cred)); 2126 cred->cr_ref = 1; 2127 cred->cr_uid = p->p_cred->p_ruid; 2128 cred->cr_gid = p->p_cred->p_rgid; 2129 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2130 SCARG(uap, path), p); 2131 /* Override default credentials */ 2132 nd.ni_cnd.cn_cred = cred; 2133 if ((error = namei(&nd)) != 0) 2134 goto out; 2135 vp = nd.ni_vp; 2136 2137 /* Flags == 0 means only check for existence. */ 2138 if (SCARG(uap, flags)) { 2139 flags = 0; 2140 if (SCARG(uap, flags) & R_OK) 2141 flags |= VREAD; 2142 if (SCARG(uap, flags) & W_OK) 2143 flags |= VWRITE; 2144 if (SCARG(uap, flags) & X_OK) 2145 flags |= VEXEC; 2146 2147 error = VOP_ACCESS(vp, flags, cred, p); 2148 if (!error && (flags & VWRITE)) 2149 error = vn_writechk(vp); 2150 } 2151 vput(vp); 2152 out: 2153 crfree(cred); 2154 return (error); 2155 } 2156 2157 /* 2158 * Get file status; this version follows links. 2159 */ 2160 /* ARGSUSED */ 2161 int 2162 sys___stat13(l, v, retval) 2163 struct lwp *l; 2164 void *v; 2165 register_t *retval; 2166 { 2167 struct sys___stat13_args /* { 2168 syscallarg(const char *) path; 2169 syscallarg(struct stat *) ub; 2170 } */ *uap = v; 2171 struct proc *p = l->l_proc; 2172 struct stat sb; 2173 int error; 2174 struct nameidata nd; 2175 2176 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2177 SCARG(uap, path), p); 2178 if ((error = namei(&nd)) != 0) 2179 return (error); 2180 error = vn_stat(nd.ni_vp, &sb, p); 2181 vput(nd.ni_vp); 2182 if (error) 2183 return (error); 2184 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2185 return (error); 2186 } 2187 2188 /* 2189 * Get file status; this version does not follow links. 2190 */ 2191 /* ARGSUSED */ 2192 int 2193 sys___lstat13(l, v, retval) 2194 struct lwp *l; 2195 void *v; 2196 register_t *retval; 2197 { 2198 struct sys___lstat13_args /* { 2199 syscallarg(const char *) path; 2200 syscallarg(struct stat *) ub; 2201 } */ *uap = v; 2202 struct proc *p = l->l_proc; 2203 struct stat sb; 2204 int error; 2205 struct nameidata nd; 2206 2207 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2208 SCARG(uap, path), p); 2209 if ((error = namei(&nd)) != 0) 2210 return (error); 2211 error = vn_stat(nd.ni_vp, &sb, p); 2212 vput(nd.ni_vp); 2213 if (error) 2214 return (error); 2215 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2216 return (error); 2217 } 2218 2219 /* 2220 * Get configurable pathname variables. 2221 */ 2222 /* ARGSUSED */ 2223 int 2224 sys_pathconf(l, v, retval) 2225 struct lwp *l; 2226 void *v; 2227 register_t *retval; 2228 { 2229 struct sys_pathconf_args /* { 2230 syscallarg(const char *) path; 2231 syscallarg(int) name; 2232 } */ *uap = v; 2233 struct proc *p = l->l_proc; 2234 int error; 2235 struct nameidata nd; 2236 2237 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2238 SCARG(uap, path), p); 2239 if ((error = namei(&nd)) != 0) 2240 return (error); 2241 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2242 vput(nd.ni_vp); 2243 return (error); 2244 } 2245 2246 /* 2247 * Return target name of a symbolic link. 2248 */ 2249 /* ARGSUSED */ 2250 int 2251 sys_readlink(l, v, retval) 2252 struct lwp *l; 2253 void *v; 2254 register_t *retval; 2255 { 2256 struct sys_readlink_args /* { 2257 syscallarg(const char *) path; 2258 syscallarg(char *) buf; 2259 syscallarg(size_t) count; 2260 } */ *uap = v; 2261 struct proc *p = l->l_proc; 2262 struct vnode *vp; 2263 struct iovec aiov; 2264 struct uio auio; 2265 int error; 2266 struct nameidata nd; 2267 2268 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2269 SCARG(uap, path), p); 2270 if ((error = namei(&nd)) != 0) 2271 return (error); 2272 vp = nd.ni_vp; 2273 if (vp->v_type != VLNK) 2274 error = EINVAL; 2275 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2276 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 2277 aiov.iov_base = SCARG(uap, buf); 2278 aiov.iov_len = SCARG(uap, count); 2279 auio.uio_iov = &aiov; 2280 auio.uio_iovcnt = 1; 2281 auio.uio_offset = 0; 2282 auio.uio_rw = UIO_READ; 2283 auio.uio_segflg = UIO_USERSPACE; 2284 auio.uio_procp = p; 2285 auio.uio_resid = SCARG(uap, count); 2286 error = VOP_READLINK(vp, &auio, p->p_ucred); 2287 } 2288 vput(vp); 2289 *retval = SCARG(uap, count) - auio.uio_resid; 2290 return (error); 2291 } 2292 2293 /* 2294 * Change flags of a file given a path name. 2295 */ 2296 /* ARGSUSED */ 2297 int 2298 sys_chflags(l, v, retval) 2299 struct lwp *l; 2300 void *v; 2301 register_t *retval; 2302 { 2303 struct sys_chflags_args /* { 2304 syscallarg(const char *) path; 2305 syscallarg(u_long) flags; 2306 } */ *uap = v; 2307 struct proc *p = l->l_proc; 2308 struct vnode *vp; 2309 int error; 2310 struct nameidata nd; 2311 2312 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2313 if ((error = namei(&nd)) != 0) 2314 return (error); 2315 vp = nd.ni_vp; 2316 error = change_flags(vp, SCARG(uap, flags), p); 2317 vput(vp); 2318 return (error); 2319 } 2320 2321 /* 2322 * Change flags of a file given a file descriptor. 2323 */ 2324 /* ARGSUSED */ 2325 int 2326 sys_fchflags(l, v, retval) 2327 struct lwp *l; 2328 void *v; 2329 register_t *retval; 2330 { 2331 struct sys_fchflags_args /* { 2332 syscallarg(int) fd; 2333 syscallarg(u_long) flags; 2334 } */ *uap = v; 2335 struct proc *p = l->l_proc; 2336 struct vnode *vp; 2337 struct file *fp; 2338 int error; 2339 2340 /* getvnode() will use the descriptor for us */ 2341 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2342 return (error); 2343 vp = (struct vnode *)fp->f_data; 2344 error = change_flags(vp, SCARG(uap, flags), p); 2345 VOP_UNLOCK(vp, 0); 2346 FILE_UNUSE(fp, p); 2347 return (error); 2348 } 2349 2350 /* 2351 * Change flags of a file given a path name; this version does 2352 * not follow links. 2353 */ 2354 int 2355 sys_lchflags(l, v, retval) 2356 struct lwp *l; 2357 void *v; 2358 register_t *retval; 2359 { 2360 struct sys_lchflags_args /* { 2361 syscallarg(const char *) path; 2362 syscallarg(u_long) flags; 2363 } */ *uap = v; 2364 struct proc *p = l->l_proc; 2365 struct vnode *vp; 2366 int error; 2367 struct nameidata nd; 2368 2369 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2370 if ((error = namei(&nd)) != 0) 2371 return (error); 2372 vp = nd.ni_vp; 2373 error = change_flags(vp, SCARG(uap, flags), p); 2374 vput(vp); 2375 return (error); 2376 } 2377 2378 /* 2379 * Common routine to change flags of a file. 2380 */ 2381 int 2382 change_flags(vp, flags, p) 2383 struct vnode *vp; 2384 u_long flags; 2385 struct proc *p; 2386 { 2387 struct mount *mp; 2388 struct vattr vattr; 2389 int error; 2390 2391 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2392 return (error); 2393 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2394 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2395 /* 2396 * Non-superusers cannot change the flags on devices, even if they 2397 * own them. 2398 */ 2399 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2400 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2401 goto out; 2402 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2403 error = EINVAL; 2404 goto out; 2405 } 2406 } 2407 VATTR_NULL(&vattr); 2408 vattr.va_flags = flags; 2409 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2410 out: 2411 vn_finished_write(mp, 0); 2412 return (error); 2413 } 2414 2415 /* 2416 * Change mode of a file given path name; this version follows links. 2417 */ 2418 /* ARGSUSED */ 2419 int 2420 sys_chmod(l, v, retval) 2421 struct lwp *l; 2422 void *v; 2423 register_t *retval; 2424 { 2425 struct sys_chmod_args /* { 2426 syscallarg(const char *) path; 2427 syscallarg(int) mode; 2428 } */ *uap = v; 2429 struct proc *p = l->l_proc; 2430 int error; 2431 struct nameidata nd; 2432 2433 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2434 if ((error = namei(&nd)) != 0) 2435 return (error); 2436 2437 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2438 2439 vrele(nd.ni_vp); 2440 return (error); 2441 } 2442 2443 /* 2444 * Change mode of a file given a file descriptor. 2445 */ 2446 /* ARGSUSED */ 2447 int 2448 sys_fchmod(l, v, retval) 2449 struct lwp *l; 2450 void *v; 2451 register_t *retval; 2452 { 2453 struct sys_fchmod_args /* { 2454 syscallarg(int) fd; 2455 syscallarg(int) mode; 2456 } */ *uap = v; 2457 struct proc *p = l->l_proc; 2458 struct file *fp; 2459 int error; 2460 2461 /* getvnode() will use the descriptor for us */ 2462 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2463 return (error); 2464 2465 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 2466 FILE_UNUSE(fp, p); 2467 return (error); 2468 } 2469 2470 /* 2471 * Change mode of a file given path name; this version does not follow links. 2472 */ 2473 /* ARGSUSED */ 2474 int 2475 sys_lchmod(l, v, retval) 2476 struct lwp *l; 2477 void *v; 2478 register_t *retval; 2479 { 2480 struct sys_lchmod_args /* { 2481 syscallarg(const char *) path; 2482 syscallarg(int) mode; 2483 } */ *uap = v; 2484 struct proc *p = l->l_proc; 2485 int error; 2486 struct nameidata nd; 2487 2488 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2489 if ((error = namei(&nd)) != 0) 2490 return (error); 2491 2492 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2493 2494 vrele(nd.ni_vp); 2495 return (error); 2496 } 2497 2498 /* 2499 * Common routine to set mode given a vnode. 2500 */ 2501 static int 2502 change_mode(vp, mode, p) 2503 struct vnode *vp; 2504 int mode; 2505 struct proc *p; 2506 { 2507 struct mount *mp; 2508 struct vattr vattr; 2509 int error; 2510 2511 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2512 return (error); 2513 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2514 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2515 VATTR_NULL(&vattr); 2516 vattr.va_mode = mode & ALLPERMS; 2517 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2518 VOP_UNLOCK(vp, 0); 2519 vn_finished_write(mp, 0); 2520 return (error); 2521 } 2522 2523 /* 2524 * Set ownership given a path name; this version follows links. 2525 */ 2526 /* ARGSUSED */ 2527 int 2528 sys_chown(l, v, retval) 2529 struct lwp *l; 2530 void *v; 2531 register_t *retval; 2532 { 2533 struct sys_chown_args /* { 2534 syscallarg(const char *) path; 2535 syscallarg(uid_t) uid; 2536 syscallarg(gid_t) gid; 2537 } */ *uap = v; 2538 struct proc *p = l->l_proc; 2539 int error; 2540 struct nameidata nd; 2541 2542 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2543 if ((error = namei(&nd)) != 0) 2544 return (error); 2545 2546 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2547 2548 vrele(nd.ni_vp); 2549 return (error); 2550 } 2551 2552 /* 2553 * Set ownership given a path name; this version follows links. 2554 * Provides POSIX semantics. 2555 */ 2556 /* ARGSUSED */ 2557 int 2558 sys___posix_chown(l, v, retval) 2559 struct lwp *l; 2560 void *v; 2561 register_t *retval; 2562 { 2563 struct sys_chown_args /* { 2564 syscallarg(const char *) path; 2565 syscallarg(uid_t) uid; 2566 syscallarg(gid_t) gid; 2567 } */ *uap = v; 2568 struct proc *p = l->l_proc; 2569 int error; 2570 struct nameidata nd; 2571 2572 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2573 if ((error = namei(&nd)) != 0) 2574 return (error); 2575 2576 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2577 2578 vrele(nd.ni_vp); 2579 return (error); 2580 } 2581 2582 /* 2583 * Set ownership given a file descriptor. 2584 */ 2585 /* ARGSUSED */ 2586 int 2587 sys_fchown(l, v, retval) 2588 struct lwp *l; 2589 void *v; 2590 register_t *retval; 2591 { 2592 struct sys_fchown_args /* { 2593 syscallarg(int) fd; 2594 syscallarg(uid_t) uid; 2595 syscallarg(gid_t) gid; 2596 } */ *uap = v; 2597 struct proc *p = l->l_proc; 2598 int error; 2599 struct file *fp; 2600 2601 /* getvnode() will use the descriptor for us */ 2602 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2603 return (error); 2604 2605 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2606 SCARG(uap, gid), p, 0); 2607 FILE_UNUSE(fp, p); 2608 return (error); 2609 } 2610 2611 /* 2612 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2613 */ 2614 /* ARGSUSED */ 2615 int 2616 sys___posix_fchown(l, v, retval) 2617 struct lwp *l; 2618 void *v; 2619 register_t *retval; 2620 { 2621 struct sys_fchown_args /* { 2622 syscallarg(int) fd; 2623 syscallarg(uid_t) uid; 2624 syscallarg(gid_t) gid; 2625 } */ *uap = v; 2626 struct proc *p = l->l_proc; 2627 int error; 2628 struct file *fp; 2629 2630 /* getvnode() will use the descriptor for us */ 2631 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2632 return (error); 2633 2634 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2635 SCARG(uap, gid), p, 1); 2636 FILE_UNUSE(fp, p); 2637 return (error); 2638 } 2639 2640 /* 2641 * Set ownership given a path name; this version does not follow links. 2642 */ 2643 /* ARGSUSED */ 2644 int 2645 sys_lchown(l, v, retval) 2646 struct lwp *l; 2647 void *v; 2648 register_t *retval; 2649 { 2650 struct sys_lchown_args /* { 2651 syscallarg(const char *) path; 2652 syscallarg(uid_t) uid; 2653 syscallarg(gid_t) gid; 2654 } */ *uap = v; 2655 struct proc *p = l->l_proc; 2656 int error; 2657 struct nameidata nd; 2658 2659 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2660 if ((error = namei(&nd)) != 0) 2661 return (error); 2662 2663 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2664 2665 vrele(nd.ni_vp); 2666 return (error); 2667 } 2668 2669 /* 2670 * Set ownership given a path name; this version does not follow links. 2671 * Provides POSIX/XPG semantics. 2672 */ 2673 /* ARGSUSED */ 2674 int 2675 sys___posix_lchown(l, v, retval) 2676 struct lwp *l; 2677 void *v; 2678 register_t *retval; 2679 { 2680 struct sys_lchown_args /* { 2681 syscallarg(const char *) path; 2682 syscallarg(uid_t) uid; 2683 syscallarg(gid_t) gid; 2684 } */ *uap = v; 2685 struct proc *p = l->l_proc; 2686 int error; 2687 struct nameidata nd; 2688 2689 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2690 if ((error = namei(&nd)) != 0) 2691 return (error); 2692 2693 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2694 2695 vrele(nd.ni_vp); 2696 return (error); 2697 } 2698 2699 /* 2700 * Common routine to set ownership given a vnode. 2701 */ 2702 static int 2703 change_owner(vp, uid, gid, p, posix_semantics) 2704 struct vnode *vp; 2705 uid_t uid; 2706 gid_t gid; 2707 struct proc *p; 2708 int posix_semantics; 2709 { 2710 struct mount *mp; 2711 struct vattr vattr; 2712 mode_t newmode; 2713 int error; 2714 2715 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2716 return (error); 2717 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2718 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2719 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2720 goto out; 2721 2722 #define CHANGED(x) ((int)(x) != -1) 2723 newmode = vattr.va_mode; 2724 if (posix_semantics) { 2725 /* 2726 * POSIX/XPG semantics: if the caller is not the super-user, 2727 * clear set-user-id and set-group-id bits. Both POSIX and 2728 * the XPG consider the behaviour for calls by the super-user 2729 * implementation-defined; we leave the set-user-id and set- 2730 * group-id settings intact in that case. 2731 */ 2732 if (suser(p->p_ucred, NULL) != 0) 2733 newmode &= ~(S_ISUID | S_ISGID); 2734 } else { 2735 /* 2736 * NetBSD semantics: when changing owner and/or group, 2737 * clear the respective bit(s). 2738 */ 2739 if (CHANGED(uid)) 2740 newmode &= ~S_ISUID; 2741 if (CHANGED(gid)) 2742 newmode &= ~S_ISGID; 2743 } 2744 /* Update va_mode iff altered. */ 2745 if (vattr.va_mode == newmode) 2746 newmode = VNOVAL; 2747 2748 VATTR_NULL(&vattr); 2749 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2750 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2751 vattr.va_mode = newmode; 2752 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2753 #undef CHANGED 2754 2755 out: 2756 VOP_UNLOCK(vp, 0); 2757 vn_finished_write(mp, 0); 2758 return (error); 2759 } 2760 2761 /* 2762 * Set the access and modification times given a path name; this 2763 * version follows links. 2764 */ 2765 /* ARGSUSED */ 2766 int 2767 sys_utimes(l, v, retval) 2768 struct lwp *l; 2769 void *v; 2770 register_t *retval; 2771 { 2772 struct sys_utimes_args /* { 2773 syscallarg(const char *) path; 2774 syscallarg(const struct timeval *) tptr; 2775 } */ *uap = v; 2776 struct proc *p = l->l_proc; 2777 int error; 2778 struct nameidata nd; 2779 2780 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2781 if ((error = namei(&nd)) != 0) 2782 return (error); 2783 2784 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2785 2786 vrele(nd.ni_vp); 2787 return (error); 2788 } 2789 2790 /* 2791 * Set the access and modification times given a file descriptor. 2792 */ 2793 /* ARGSUSED */ 2794 int 2795 sys_futimes(l, v, retval) 2796 struct lwp *l; 2797 void *v; 2798 register_t *retval; 2799 { 2800 struct sys_futimes_args /* { 2801 syscallarg(int) fd; 2802 syscallarg(const struct timeval *) tptr; 2803 } */ *uap = v; 2804 struct proc *p = l->l_proc; 2805 int error; 2806 struct file *fp; 2807 2808 /* getvnode() will use the descriptor for us */ 2809 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2810 return (error); 2811 2812 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2813 FILE_UNUSE(fp, p); 2814 return (error); 2815 } 2816 2817 /* 2818 * Set the access and modification times given a path name; this 2819 * version does not follow links. 2820 */ 2821 /* ARGSUSED */ 2822 int 2823 sys_lutimes(l, v, retval) 2824 struct lwp *l; 2825 void *v; 2826 register_t *retval; 2827 { 2828 struct sys_lutimes_args /* { 2829 syscallarg(const char *) path; 2830 syscallarg(const struct timeval *) tptr; 2831 } */ *uap = v; 2832 struct proc *p = l->l_proc; 2833 int error; 2834 struct nameidata nd; 2835 2836 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2837 if ((error = namei(&nd)) != 0) 2838 return (error); 2839 2840 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2841 2842 vrele(nd.ni_vp); 2843 return (error); 2844 } 2845 2846 /* 2847 * Common routine to set access and modification times given a vnode. 2848 */ 2849 static int 2850 change_utimes(vp, tptr, p) 2851 struct vnode *vp; 2852 const struct timeval *tptr; 2853 struct proc *p; 2854 { 2855 struct timeval tv[2]; 2856 struct mount *mp; 2857 struct vattr vattr; 2858 int error; 2859 2860 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2861 return (error); 2862 VATTR_NULL(&vattr); 2863 if (tptr == NULL) { 2864 microtime(&tv[0]); 2865 tv[1] = tv[0]; 2866 vattr.va_vaflags |= VA_UTIMES_NULL; 2867 } else { 2868 error = copyin(tptr, tv, sizeof(tv)); 2869 if (error) 2870 goto out; 2871 } 2872 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2873 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2874 vattr.va_atime.tv_sec = tv[0].tv_sec; 2875 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2876 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2877 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2878 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2879 VOP_UNLOCK(vp, 0); 2880 out: 2881 vn_finished_write(mp, 0); 2882 return (error); 2883 } 2884 2885 /* 2886 * Truncate a file given its path name. 2887 */ 2888 /* ARGSUSED */ 2889 int 2890 sys_truncate(l, v, retval) 2891 struct lwp *l; 2892 void *v; 2893 register_t *retval; 2894 { 2895 struct sys_truncate_args /* { 2896 syscallarg(const char *) path; 2897 syscallarg(int) pad; 2898 syscallarg(off_t) length; 2899 } */ *uap = v; 2900 struct proc *p = l->l_proc; 2901 struct vnode *vp; 2902 struct mount *mp; 2903 struct vattr vattr; 2904 int error; 2905 struct nameidata nd; 2906 2907 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2908 if ((error = namei(&nd)) != 0) 2909 return (error); 2910 vp = nd.ni_vp; 2911 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2912 vrele(vp); 2913 return (error); 2914 } 2915 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2916 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2917 if (vp->v_type == VDIR) 2918 error = EISDIR; 2919 else if ((error = vn_writechk(vp)) == 0 && 2920 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2921 VATTR_NULL(&vattr); 2922 vattr.va_size = SCARG(uap, length); 2923 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2924 } 2925 vput(vp); 2926 vn_finished_write(mp, 0); 2927 return (error); 2928 } 2929 2930 /* 2931 * Truncate a file given a file descriptor. 2932 */ 2933 /* ARGSUSED */ 2934 int 2935 sys_ftruncate(l, v, retval) 2936 struct lwp *l; 2937 void *v; 2938 register_t *retval; 2939 { 2940 struct sys_ftruncate_args /* { 2941 syscallarg(int) fd; 2942 syscallarg(int) pad; 2943 syscallarg(off_t) length; 2944 } */ *uap = v; 2945 struct proc *p = l->l_proc; 2946 struct mount *mp; 2947 struct vattr vattr; 2948 struct vnode *vp; 2949 struct file *fp; 2950 int error; 2951 2952 /* getvnode() will use the descriptor for us */ 2953 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2954 return (error); 2955 if ((fp->f_flag & FWRITE) == 0) { 2956 error = EINVAL; 2957 goto out; 2958 } 2959 vp = (struct vnode *)fp->f_data; 2960 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2961 FILE_UNUSE(fp, p); 2962 return (error); 2963 } 2964 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2965 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2966 if (vp->v_type == VDIR) 2967 error = EISDIR; 2968 else if ((error = vn_writechk(vp)) == 0) { 2969 VATTR_NULL(&vattr); 2970 vattr.va_size = SCARG(uap, length); 2971 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2972 } 2973 VOP_UNLOCK(vp, 0); 2974 vn_finished_write(mp, 0); 2975 out: 2976 FILE_UNUSE(fp, p); 2977 return (error); 2978 } 2979 2980 /* 2981 * Sync an open file. 2982 */ 2983 /* ARGSUSED */ 2984 int 2985 sys_fsync(l, v, retval) 2986 struct lwp *l; 2987 void *v; 2988 register_t *retval; 2989 { 2990 struct sys_fsync_args /* { 2991 syscallarg(int) fd; 2992 } */ *uap = v; 2993 struct proc *p = l->l_proc; 2994 struct vnode *vp; 2995 struct mount *mp; 2996 struct file *fp; 2997 int error; 2998 2999 /* getvnode() will use the descriptor for us */ 3000 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3001 return (error); 3002 vp = (struct vnode *)fp->f_data; 3003 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3004 FILE_UNUSE(fp, p); 3005 return (error); 3006 } 3007 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3008 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p); 3009 if (error == 0 && bioops.io_fsync != NULL && 3010 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3011 (*bioops.io_fsync)(vp); 3012 VOP_UNLOCK(vp, 0); 3013 vn_finished_write(mp, 0); 3014 FILE_UNUSE(fp, p); 3015 return (error); 3016 } 3017 3018 /* 3019 * Sync a range of file data. API modeled after that found in AIX. 3020 * 3021 * FDATASYNC indicates that we need only save enough metadata to be able 3022 * to re-read the written data. Note we duplicate AIX's requirement that 3023 * the file be open for writing. 3024 */ 3025 /* ARGSUSED */ 3026 int 3027 sys_fsync_range(l, v, retval) 3028 struct lwp *l; 3029 void *v; 3030 register_t *retval; 3031 { 3032 struct sys_fsync_range_args /* { 3033 syscallarg(int) fd; 3034 syscallarg(int) flags; 3035 syscallarg(off_t) start; 3036 syscallarg(int) length; 3037 } */ *uap = v; 3038 struct proc *p = l->l_proc; 3039 struct vnode *vp; 3040 struct file *fp; 3041 int flags, nflags; 3042 off_t s, e, len; 3043 int error; 3044 3045 /* getvnode() will use the descriptor for us */ 3046 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3047 return (error); 3048 3049 if ((fp->f_flag & FWRITE) == 0) { 3050 FILE_UNUSE(fp, p); 3051 return (EBADF); 3052 } 3053 3054 flags = SCARG(uap, flags); 3055 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3056 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3057 return (EINVAL); 3058 } 3059 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3060 if (flags & FDATASYNC) 3061 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3062 else 3063 nflags = FSYNC_WAIT; 3064 3065 len = SCARG(uap, length); 3066 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3067 if (len) { 3068 s = SCARG(uap, start); 3069 e = s + len; 3070 if (e < s) { 3071 FILE_UNUSE(fp, p); 3072 return (EINVAL); 3073 } 3074 } else { 3075 e = 0; 3076 s = 0; 3077 } 3078 3079 vp = (struct vnode *)fp->f_data; 3080 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3081 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p); 3082 3083 if (error == 0 && bioops.io_fsync != NULL && 3084 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3085 (*bioops.io_fsync)(vp); 3086 3087 VOP_UNLOCK(vp, 0); 3088 FILE_UNUSE(fp, p); 3089 return (error); 3090 } 3091 3092 /* 3093 * Sync the data of an open file. 3094 */ 3095 /* ARGSUSED */ 3096 int 3097 sys_fdatasync(l, v, retval) 3098 struct lwp *l; 3099 void *v; 3100 register_t *retval; 3101 { 3102 struct sys_fdatasync_args /* { 3103 syscallarg(int) fd; 3104 } */ *uap = v; 3105 struct proc *p = l->l_proc; 3106 struct vnode *vp; 3107 struct file *fp; 3108 int error; 3109 3110 /* getvnode() will use the descriptor for us */ 3111 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3112 return (error); 3113 if ((fp->f_flag & FWRITE) == 0) { 3114 FILE_UNUSE(fp, p); 3115 return (EBADF); 3116 } 3117 vp = (struct vnode *)fp->f_data; 3118 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3119 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p); 3120 VOP_UNLOCK(vp, 0); 3121 FILE_UNUSE(fp, p); 3122 return (error); 3123 } 3124 3125 /* 3126 * Rename files, (standard) BSD semantics frontend. 3127 */ 3128 /* ARGSUSED */ 3129 int 3130 sys_rename(l, v, retval) 3131 struct lwp *l; 3132 void *v; 3133 register_t *retval; 3134 { 3135 struct sys_rename_args /* { 3136 syscallarg(const char *) from; 3137 syscallarg(const char *) to; 3138 } */ *uap = v; 3139 struct proc *p = l->l_proc; 3140 3141 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 3142 } 3143 3144 /* 3145 * Rename files, POSIX semantics frontend. 3146 */ 3147 /* ARGSUSED */ 3148 int 3149 sys___posix_rename(l, v, retval) 3150 struct lwp *l; 3151 void *v; 3152 register_t *retval; 3153 { 3154 struct sys___posix_rename_args /* { 3155 syscallarg(const char *) from; 3156 syscallarg(const char *) to; 3157 } */ *uap = v; 3158 struct proc *p = l->l_proc; 3159 3160 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 3161 } 3162 3163 /* 3164 * Rename files. Source and destination must either both be directories, 3165 * or both not be directories. If target is a directory, it must be empty. 3166 * If `from' and `to' refer to the same object, the value of the `retain' 3167 * argument is used to determine whether `from' will be 3168 * 3169 * (retain == 0) deleted unless `from' and `to' refer to the same 3170 * object in the file system's name space (BSD). 3171 * (retain == 1) always retained (POSIX). 3172 */ 3173 static int 3174 rename_files(from, to, p, retain) 3175 const char *from, *to; 3176 struct proc *p; 3177 int retain; 3178 { 3179 struct mount *mp = NULL; 3180 struct vnode *tvp, *fvp, *tdvp; 3181 struct nameidata fromnd, tond; 3182 int error; 3183 3184 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3185 from, p); 3186 if ((error = namei(&fromnd)) != 0) 3187 return (error); 3188 fvp = fromnd.ni_vp; 3189 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3190 if (error != 0) { 3191 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3192 vrele(fromnd.ni_dvp); 3193 vrele(fvp); 3194 if (fromnd.ni_startdir) 3195 vrele(fromnd.ni_startdir); 3196 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3197 return (error); 3198 } 3199 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3200 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p); 3201 if ((error = namei(&tond)) != 0) { 3202 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3203 vrele(fromnd.ni_dvp); 3204 vrele(fvp); 3205 goto out1; 3206 } 3207 tdvp = tond.ni_dvp; 3208 tvp = tond.ni_vp; 3209 3210 if (tvp != NULL) { 3211 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3212 error = ENOTDIR; 3213 goto out; 3214 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3215 error = EISDIR; 3216 goto out; 3217 } 3218 } 3219 3220 if (fvp == tdvp) 3221 error = EINVAL; 3222 3223 /* 3224 * Source and destination refer to the same object. 3225 */ 3226 if (fvp == tvp) { 3227 if (retain) 3228 error = -1; 3229 else if (fromnd.ni_dvp == tdvp && 3230 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3231 !memcmp(fromnd.ni_cnd.cn_nameptr, 3232 tond.ni_cnd.cn_nameptr, 3233 fromnd.ni_cnd.cn_namelen)) 3234 error = -1; 3235 } 3236 3237 out: 3238 if (!error) { 3239 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 3240 if (fromnd.ni_dvp != tdvp) 3241 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3242 if (tvp) { 3243 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 3244 } 3245 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3246 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3247 } else { 3248 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3249 if (tdvp == tvp) 3250 vrele(tdvp); 3251 else 3252 vput(tdvp); 3253 if (tvp) 3254 vput(tvp); 3255 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3256 vrele(fromnd.ni_dvp); 3257 vrele(fvp); 3258 } 3259 vrele(tond.ni_startdir); 3260 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3261 out1: 3262 vn_finished_write(mp, 0); 3263 if (fromnd.ni_startdir) 3264 vrele(fromnd.ni_startdir); 3265 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3266 return (error == -1 ? 0 : error); 3267 } 3268 3269 /* 3270 * Make a directory file. 3271 */ 3272 /* ARGSUSED */ 3273 int 3274 sys_mkdir(l, v, retval) 3275 struct lwp *l; 3276 void *v; 3277 register_t *retval; 3278 { 3279 struct sys_mkdir_args /* { 3280 syscallarg(const char *) path; 3281 syscallarg(int) mode; 3282 } */ *uap = v; 3283 struct proc *p = l->l_proc; 3284 struct mount *mp; 3285 struct vnode *vp; 3286 struct vattr vattr; 3287 int error; 3288 struct nameidata nd; 3289 3290 restart: 3291 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3292 SCARG(uap, path), p); 3293 if ((error = namei(&nd)) != 0) 3294 return (error); 3295 vp = nd.ni_vp; 3296 if (vp != NULL) { 3297 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3298 if (nd.ni_dvp == vp) 3299 vrele(nd.ni_dvp); 3300 else 3301 vput(nd.ni_dvp); 3302 vrele(vp); 3303 return (EEXIST); 3304 } 3305 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3306 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3307 if (nd.ni_dvp == vp) 3308 vrele(nd.ni_dvp); 3309 else 3310 vput(nd.ni_dvp); 3311 if ((error = vn_start_write(NULL, &mp, 3312 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3313 return (error); 3314 goto restart; 3315 } 3316 VATTR_NULL(&vattr); 3317 vattr.va_type = VDIR; 3318 vattr.va_mode = 3319 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3320 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3321 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3322 if (!error) 3323 vput(nd.ni_vp); 3324 vn_finished_write(mp, 0); 3325 return (error); 3326 } 3327 3328 /* 3329 * Remove a directory file. 3330 */ 3331 /* ARGSUSED */ 3332 int 3333 sys_rmdir(l, v, retval) 3334 struct lwp *l; 3335 void *v; 3336 register_t *retval; 3337 { 3338 struct sys_rmdir_args /* { 3339 syscallarg(const char *) path; 3340 } */ *uap = v; 3341 struct proc *p = l->l_proc; 3342 struct mount *mp; 3343 struct vnode *vp; 3344 int error; 3345 struct nameidata nd; 3346 3347 restart: 3348 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3349 SCARG(uap, path), p); 3350 if ((error = namei(&nd)) != 0) 3351 return (error); 3352 vp = nd.ni_vp; 3353 if (vp->v_type != VDIR) { 3354 error = ENOTDIR; 3355 goto out; 3356 } 3357 /* 3358 * No rmdir "." please. 3359 */ 3360 if (nd.ni_dvp == vp) { 3361 error = EINVAL; 3362 goto out; 3363 } 3364 /* 3365 * The root of a mounted filesystem cannot be deleted. 3366 */ 3367 if (vp->v_flag & VROOT) { 3368 error = EBUSY; 3369 goto out; 3370 } 3371 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3372 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3373 if (nd.ni_dvp == vp) 3374 vrele(nd.ni_dvp); 3375 else 3376 vput(nd.ni_dvp); 3377 vput(vp); 3378 if ((error = vn_start_write(NULL, &mp, 3379 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3380 return (error); 3381 goto restart; 3382 } 3383 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3384 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3385 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3386 vn_finished_write(mp, 0); 3387 return (error); 3388 3389 out: 3390 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3391 if (nd.ni_dvp == vp) 3392 vrele(nd.ni_dvp); 3393 else 3394 vput(nd.ni_dvp); 3395 vput(vp); 3396 return (error); 3397 } 3398 3399 /* 3400 * Read a block of directory entries in a file system independent format. 3401 */ 3402 int 3403 sys_getdents(l, v, retval) 3404 struct lwp *l; 3405 void *v; 3406 register_t *retval; 3407 { 3408 struct sys_getdents_args /* { 3409 syscallarg(int) fd; 3410 syscallarg(char *) buf; 3411 syscallarg(size_t) count; 3412 } */ *uap = v; 3413 struct proc *p = l->l_proc; 3414 struct file *fp; 3415 int error, done; 3416 3417 /* getvnode() will use the descriptor for us */ 3418 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3419 return (error); 3420 if ((fp->f_flag & FREAD) == 0) { 3421 error = EBADF; 3422 goto out; 3423 } 3424 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3425 SCARG(uap, count), &done, p, 0, 0); 3426 #ifdef KTRACE 3427 if (!error && KTRPOINT(p, KTR_GENIO)) { 3428 struct iovec iov; 3429 iov.iov_base = SCARG(uap, buf); 3430 iov.iov_len = done; 3431 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3432 } 3433 #endif 3434 *retval = done; 3435 out: 3436 FILE_UNUSE(fp, p); 3437 return (error); 3438 } 3439 3440 /* 3441 * Set the mode mask for creation of filesystem nodes. 3442 */ 3443 int 3444 sys_umask(l, v, retval) 3445 struct lwp *l; 3446 void *v; 3447 register_t *retval; 3448 { 3449 struct sys_umask_args /* { 3450 syscallarg(mode_t) newmask; 3451 } */ *uap = v; 3452 struct proc *p = l->l_proc; 3453 struct cwdinfo *cwdi; 3454 3455 cwdi = p->p_cwdi; 3456 *retval = cwdi->cwdi_cmask; 3457 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3458 return (0); 3459 } 3460 3461 /* 3462 * Void all references to file by ripping underlying filesystem 3463 * away from vnode. 3464 */ 3465 /* ARGSUSED */ 3466 int 3467 sys_revoke(l, v, retval) 3468 struct lwp *l; 3469 void *v; 3470 register_t *retval; 3471 { 3472 struct sys_revoke_args /* { 3473 syscallarg(const char *) path; 3474 } */ *uap = v; 3475 struct proc *p = l->l_proc; 3476 struct mount *mp; 3477 struct vnode *vp; 3478 struct vattr vattr; 3479 int error; 3480 struct nameidata nd; 3481 3482 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3483 if ((error = namei(&nd)) != 0) 3484 return (error); 3485 vp = nd.ni_vp; 3486 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 3487 goto out; 3488 if (p->p_ucred->cr_uid != vattr.va_uid && 3489 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3490 goto out; 3491 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3492 goto out; 3493 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3494 VOP_REVOKE(vp, REVOKEALL); 3495 vn_finished_write(mp, 0); 3496 out: 3497 vrele(vp); 3498 return (error); 3499 } 3500 3501 /* 3502 * Convert a user file descriptor to a kernel file entry. 3503 */ 3504 int 3505 getvnode(fdp, fd, fpp) 3506 struct filedesc *fdp; 3507 int fd; 3508 struct file **fpp; 3509 { 3510 struct vnode *vp; 3511 struct file *fp; 3512 3513 if ((fp = fd_getfile(fdp, fd)) == NULL) 3514 return (EBADF); 3515 3516 FILE_USE(fp); 3517 3518 if (fp->f_type != DTYPE_VNODE) { 3519 FILE_UNUSE(fp, NULL); 3520 return (EINVAL); 3521 } 3522 3523 vp = (struct vnode *)fp->f_data; 3524 if (vp->v_type == VBAD) { 3525 FILE_UNUSE(fp, NULL); 3526 return (EBADF); 3527 } 3528 3529 *fpp = fp; 3530 return (0); 3531 } 3532
Indexes created Tue Sep 23 02:09:52 GMT 2025