vfs_syscalls.c revision 1.208
1 /* $NetBSD: vfs_syscalls.c,v 1.208 2004/05/02 12:21:02 pk 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 1.208 2004/05/02 12:21:02 pk 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 proc *); 77 static int change_flags(struct vnode *, u_long, struct proc *); 78 static int change_mode(struct vnode *, int, struct proc *p); 79 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int); 80 static int change_utimes(struct vnode *vp, const struct timeval *, 81 struct proc *p); 82 static int rename_files(const char *, const char *, struct proc *, 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), p); 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, p)) != 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, p, 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; 306 307 /* 308 * The underlying file system may refuse the mount for 309 * various reasons. Allow the user to force it to happen. 310 */ 311 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 312 update: 313 /* 314 * Set the mount level flags. 315 */ 316 if (SCARG(uap, flags) & MNT_RDONLY) 317 mp->mnt_flag |= MNT_RDONLY; 318 else if (mp->mnt_flag & MNT_RDONLY) 319 mp->mnt_iflag |= IMNT_WANTRDWR; 320 mp->mnt_flag &= 321 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 322 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 323 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 324 mp->mnt_flag |= SCARG(uap, flags) & 325 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 326 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 327 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 328 MNT_IGNORE); 329 /* 330 * Mount the filesystem. 331 */ 332 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 333 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 334 if (mp->mnt_iflag & IMNT_WANTRDWR) 335 mp->mnt_flag &= ~MNT_RDONLY; 336 if (error || (mp->mnt_flag & MNT_GETARGS)) 337 mp->mnt_flag = flag; 338 mp->mnt_flag &=~ 339 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 340 mp->mnt_iflag &=~ IMNT_WANTRDWR; 341 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 342 if (mp->mnt_syncer == NULL) 343 error = vfs_allocate_syncvnode(mp); 344 } else { 345 if (mp->mnt_syncer != NULL) 346 vfs_deallocate_syncvnode(mp); 347 } 348 vfs_unbusy(mp); 349 VOP_UNLOCK(vp, 0); 350 vrele(vp); 351 return (error); 352 } 353 /* 354 * Put the new filesystem on the mount list after root. 355 */ 356 cache_purge(vp); 357 if (!error) { 358 mp->mnt_flag &=~ 359 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 360 mp->mnt_iflag &=~ IMNT_WANTRDWR; 361 vp->v_mountedhere = mp; 362 simple_lock(&mountlist_slock); 363 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 364 simple_unlock(&mountlist_slock); 365 checkdirs(vp); 366 VOP_UNLOCK(vp, 0); 367 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 368 error = vfs_allocate_syncvnode(mp); 369 vfs_unbusy(mp); 370 (void) VFS_STATVFS(mp, &mp->mnt_stat, p); 371 if ((error = VFS_START(mp, 0, p))) 372 vrele(vp); 373 } else { 374 vp->v_mountedhere = (struct mount *)0; 375 vfs->vfs_refcount--; 376 vfs_unbusy(mp); 377 free(mp, M_MOUNT); 378 vput(vp); 379 } 380 return (error); 381 } 382 383 /* 384 * Scan all active processes to see if any of them have a current 385 * or root directory onto which the new filesystem has just been 386 * mounted. If so, replace them with the new mount point. 387 */ 388 void 389 checkdirs(olddp) 390 struct vnode *olddp; 391 { 392 struct cwdinfo *cwdi; 393 struct vnode *newdp; 394 struct proc *p; 395 396 if (olddp->v_usecount == 1) 397 return; 398 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 399 panic("mount: lost mount"); 400 proclist_lock_read(); 401 LIST_FOREACH(p, &allproc, p_list) { 402 cwdi = p->p_cwdi; 403 if (cwdi->cwdi_cdir == olddp) { 404 vrele(cwdi->cwdi_cdir); 405 VREF(newdp); 406 cwdi->cwdi_cdir = newdp; 407 } 408 if (cwdi->cwdi_rdir == olddp) { 409 vrele(cwdi->cwdi_rdir); 410 VREF(newdp); 411 cwdi->cwdi_rdir = newdp; 412 } 413 } 414 proclist_unlock_read(); 415 if (rootvnode == olddp) { 416 vrele(rootvnode); 417 VREF(newdp); 418 rootvnode = newdp; 419 } 420 vput(newdp); 421 } 422 423 /* 424 * Unmount a file system. 425 * 426 * Note: unmount takes a path to the vnode mounted on as argument, 427 * not special file (as before). 428 */ 429 /* ARGSUSED */ 430 int 431 sys_unmount(l, v, retval) 432 struct lwp *l; 433 void *v; 434 register_t *retval; 435 { 436 struct sys_unmount_args /* { 437 syscallarg(const char *) path; 438 syscallarg(int) flags; 439 } */ *uap = v; 440 struct proc *p = l->l_proc; 441 struct vnode *vp; 442 struct mount *mp; 443 int error; 444 struct nameidata nd; 445 446 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 447 SCARG(uap, path), p); 448 if ((error = namei(&nd)) != 0) 449 return (error); 450 vp = nd.ni_vp; 451 mp = vp->v_mount; 452 453 /* 454 * Only root, or the user that did the original mount is 455 * permitted to unmount this filesystem. 456 */ 457 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 458 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 459 vput(vp); 460 return (error); 461 } 462 463 /* 464 * Don't allow unmounting the root file system. 465 */ 466 if (mp->mnt_flag & MNT_ROOTFS) { 467 vput(vp); 468 return (EINVAL); 469 } 470 471 /* 472 * Must be the root of the filesystem 473 */ 474 if ((vp->v_flag & VROOT) == 0) { 475 vput(vp); 476 return (EINVAL); 477 } 478 vput(vp); 479 480 /* 481 * XXX Freeze syncer. Must do this before locking the 482 * mount point. See dounmount() for details. 483 */ 484 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 485 486 if (vfs_busy(mp, 0, 0)) { 487 lockmgr(&syncer_lock, LK_RELEASE, NULL); 488 return (EBUSY); 489 } 490 491 return (dounmount(mp, SCARG(uap, flags), p)); 492 } 493 494 /* 495 * Do the actual file system unmount. File system is assumed to have been 496 * marked busy by the caller. 497 */ 498 int 499 dounmount(mp, flags, p) 500 struct mount *mp; 501 int flags; 502 struct proc *p; 503 { 504 struct vnode *coveredvp; 505 int error; 506 int async; 507 int used_syncer; 508 509 simple_lock(&mountlist_slock); 510 vfs_unbusy(mp); 511 used_syncer = (mp->mnt_syncer != NULL); 512 513 /* 514 * XXX Syncer must be frozen when we get here. This should really 515 * be done on a per-mountpoint basis, but especially the softdep 516 * code possibly called from the syncer doens't exactly work on a 517 * per-mountpoint basis, so the softdep code would become a maze 518 * of vfs_busy() calls. 519 * 520 * The caller of dounmount() must acquire syncer_lock because 521 * the syncer itself acquires locks in syncer_lock -> vfs_busy 522 * order, and we must preserve that order to avoid deadlock. 523 * 524 * So, if the file system did not use the syncer, now is 525 * the time to release the syncer_lock. 526 */ 527 if (used_syncer == 0) 528 lockmgr(&syncer_lock, LK_RELEASE, NULL); 529 530 mp->mnt_iflag |= IMNT_UNMOUNT; 531 mp->mnt_unmounter = p; 532 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 533 vn_start_write(NULL, &mp, V_WAIT); 534 535 if (mp->mnt_flag & MNT_EXPUBLIC) 536 vfs_setpublicfs(NULL, NULL, NULL); 537 async = mp->mnt_flag & MNT_ASYNC; 538 mp->mnt_flag &= ~MNT_ASYNC; 539 cache_purgevfs(mp); /* remove cache entries for this file sys */ 540 if (mp->mnt_syncer != NULL) 541 vfs_deallocate_syncvnode(mp); 542 if (((mp->mnt_flag & MNT_RDONLY) || 543 #if NFSS > 0 544 (error = fss_umount_hook(mp, (flags & MNT_FORCE))) == 0 || 545 #endif 546 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) || 547 (flags & MNT_FORCE)) 548 error = VFS_UNMOUNT(mp, flags, p); 549 vn_finished_write(mp, 0); 550 simple_lock(&mountlist_slock); 551 if (error) { 552 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 553 (void) vfs_allocate_syncvnode(mp); 554 mp->mnt_iflag &= ~IMNT_UNMOUNT; 555 mp->mnt_unmounter = NULL; 556 mp->mnt_flag |= async; 557 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 558 &mountlist_slock); 559 if (used_syncer) 560 lockmgr(&syncer_lock, LK_RELEASE, NULL); 561 simple_lock(&mp->mnt_slock); 562 while (mp->mnt_wcnt > 0) { 563 wakeup(mp); 564 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 565 0, &mp->mnt_slock); 566 } 567 simple_unlock(&mp->mnt_slock); 568 return (error); 569 } 570 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 571 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 572 coveredvp->v_mountedhere = NULL; 573 vrele(coveredvp); 574 } 575 mp->mnt_op->vfs_refcount--; 576 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 577 panic("unmount: dangling vnode"); 578 mp->mnt_iflag |= IMNT_GONE; 579 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 580 if (used_syncer) 581 lockmgr(&syncer_lock, LK_RELEASE, NULL); 582 simple_lock(&mp->mnt_slock); 583 while (mp->mnt_wcnt > 0) { 584 wakeup(mp); 585 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 586 } 587 simple_unlock(&mp->mnt_slock); 588 free(mp, M_MOUNT); 589 return (0); 590 } 591 592 /* 593 * Sync each mounted filesystem. 594 */ 595 #ifdef DEBUG 596 int syncprt = 0; 597 struct ctldebug debug0 = { "syncprt", &syncprt }; 598 #endif 599 600 /* ARGSUSED */ 601 int 602 sys_sync(l, v, retval) 603 struct lwp *l; 604 void *v; 605 register_t *retval; 606 { 607 struct mount *mp, *nmp; 608 int asyncflag; 609 struct proc *p = l == NULL ? &proc0 : l->l_proc; 610 611 simple_lock(&mountlist_slock); 612 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 613 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 614 nmp = mp->mnt_list.cqe_prev; 615 continue; 616 } 617 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 618 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 619 asyncflag = mp->mnt_flag & MNT_ASYNC; 620 mp->mnt_flag &= ~MNT_ASYNC; 621 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p); 622 if (asyncflag) 623 mp->mnt_flag |= MNT_ASYNC; 624 vn_finished_write(mp, 0); 625 } 626 simple_lock(&mountlist_slock); 627 nmp = mp->mnt_list.cqe_prev; 628 vfs_unbusy(mp); 629 630 } 631 simple_unlock(&mountlist_slock); 632 #ifdef DEBUG 633 if (syncprt) 634 vfs_bufstats(); 635 #endif /* DEBUG */ 636 return (0); 637 } 638 639 /* 640 * Change filesystem quotas. 641 */ 642 /* ARGSUSED */ 643 int 644 sys_quotactl(l, v, retval) 645 struct lwp *l; 646 void *v; 647 register_t *retval; 648 { 649 struct sys_quotactl_args /* { 650 syscallarg(const char *) path; 651 syscallarg(int) cmd; 652 syscallarg(int) uid; 653 syscallarg(caddr_t) arg; 654 } */ *uap = v; 655 struct proc *p = l->l_proc; 656 struct mount *mp; 657 int error; 658 struct nameidata nd; 659 660 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 661 if ((error = namei(&nd)) != 0) 662 return (error); 663 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 664 vrele(nd.ni_vp); 665 if (error) 666 return (error); 667 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 668 SCARG(uap, arg), p); 669 vn_finished_write(mp, 0); 670 return (error); 671 } 672 673 int 674 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags, 675 int root) 676 { 677 struct cwdinfo *cwdi = p->p_cwdi; 678 int error = 0; 679 680 /* 681 * If MNT_NOWAIT or MNT_LAZY is specified, do not 682 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 683 * overrides MNT_NOWAIT. 684 */ 685 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 686 (flags != MNT_WAIT && flags != 0)) { 687 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 688 goto done; 689 } 690 691 if ((error = VFS_STATVFS(mp, sp, p)) != 0) { 692 return error; 693 } 694 695 if (cwdi->cwdi_rdir == NULL) 696 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 697 done: 698 if (cwdi->cwdi_rdir != NULL) { 699 size_t len; 700 char *bp; 701 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 702 if (!path) 703 return ENOMEM; 704 705 bp = path + MAXPATHLEN; 706 *--bp = '\0'; 707 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 708 MAXPATHLEN / 2, 0, p); 709 if (error) { 710 free(path, M_TEMP); 711 return error; 712 } 713 len = strlen(bp); 714 /* 715 * for mount points that are below our root, we can see 716 * them, so we fix up the pathname and return them. The 717 * rest we cannot see, so we don't allow viewing the 718 * data. 719 */ 720 if (strncmp(bp, sp->f_mntonname, len) == 0) { 721 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 722 sizeof(sp->f_mntonname)); 723 if (sp->f_mntonname[0] == '\0') 724 (void)strlcpy(sp->f_mntonname, "/", 725 sizeof(sp->f_mntonname)); 726 } else { 727 if (root) 728 (void)strlcpy(sp->f_mntonname, "/", 729 sizeof(sp->f_mntonname)); 730 else 731 error = EPERM; 732 } 733 free(path, M_TEMP); 734 } 735 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 736 return error; 737 } 738 739 /* 740 * Get filesystem statistics. 741 */ 742 /* ARGSUSED */ 743 int 744 sys_statvfs1(l, v, retval) 745 struct lwp *l; 746 void *v; 747 register_t *retval; 748 { 749 struct sys_statvfs1_args /* { 750 syscallarg(const char *) path; 751 syscallarg(struct statvfs *) buf; 752 syscallarg(int) flags; 753 } */ *uap = v; 754 struct proc *p = l->l_proc; 755 struct mount *mp; 756 struct statvfs sbuf; 757 int error; 758 struct nameidata nd; 759 760 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 761 if ((error = namei(&nd)) != 0) 762 return error; 763 mp = nd.ni_vp->v_mount; 764 vrele(nd.ni_vp); 765 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 766 return error; 767 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 768 } 769 770 /* 771 * Get filesystem statistics. 772 */ 773 /* ARGSUSED */ 774 int 775 sys_fstatvfs1(l, v, retval) 776 struct lwp *l; 777 void *v; 778 register_t *retval; 779 { 780 struct sys_fstatvfs1_args /* { 781 syscallarg(int) fd; 782 syscallarg(struct statvfs *) buf; 783 syscallarg(int) flags; 784 } */ *uap = v; 785 struct proc *p = l->l_proc; 786 struct file *fp; 787 struct mount *mp; 788 struct statvfs sbuf; 789 int error; 790 791 /* getvnode() will use the descriptor for us */ 792 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 793 return (error); 794 mp = ((struct vnode *)fp->f_data)->v_mount; 795 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 796 goto out; 797 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 798 out: 799 FILE_UNUSE(fp, p); 800 return error; 801 } 802 803 804 /* 805 * Get statistics on all filesystems. 806 */ 807 int 808 sys_getvfsstat(l, v, retval) 809 struct lwp *l; 810 void *v; 811 register_t *retval; 812 { 813 struct sys_getvfsstat_args /* { 814 syscallarg(struct statvfs *) buf; 815 syscallarg(size_t) bufsize; 816 syscallarg(int) flags; 817 } */ *uap = v; 818 int root = 0; 819 struct proc *p = l->l_proc; 820 struct mount *mp, *nmp; 821 struct statvfs sbuf; 822 struct statvfs *sfsp; 823 size_t count, maxcount; 824 int error = 0; 825 826 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 827 sfsp = SCARG(uap, buf); 828 simple_lock(&mountlist_slock); 829 count = 0; 830 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 831 mp = nmp) { 832 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 833 nmp = CIRCLEQ_NEXT(mp, mnt_list); 834 continue; 835 } 836 if (sfsp && count < maxcount) { 837 error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0); 838 if (error) { 839 simple_lock(&mountlist_slock); 840 nmp = CIRCLEQ_NEXT(mp, mnt_list); 841 vfs_unbusy(mp); 842 continue; 843 } 844 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 845 if (error) { 846 vfs_unbusy(mp); 847 return (error); 848 } 849 sfsp++; 850 root |= strcmp(sbuf.f_mntonname, "/") == 0; 851 } 852 count++; 853 simple_lock(&mountlist_slock); 854 nmp = CIRCLEQ_NEXT(mp, mnt_list); 855 vfs_unbusy(mp); 856 } 857 simple_unlock(&mountlist_slock); 858 if (root == 0 && p->p_cwdi->cwdi_rdir) { 859 /* 860 * fake a root entry 861 */ 862 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p, 863 SCARG(uap, flags), 1)) != 0) 864 return error; 865 if (sfsp) 866 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 867 count++; 868 } 869 if (sfsp && count > maxcount) 870 *retval = maxcount; 871 else 872 *retval = count; 873 return error; 874 } 875 876 /* 877 * Change current working directory to a given file descriptor. 878 */ 879 /* ARGSUSED */ 880 int 881 sys_fchdir(l, v, retval) 882 struct lwp *l; 883 void *v; 884 register_t *retval; 885 { 886 struct sys_fchdir_args /* { 887 syscallarg(int) fd; 888 } */ *uap = v; 889 struct proc *p = l->l_proc; 890 struct filedesc *fdp = p->p_fd; 891 struct cwdinfo *cwdi = p->p_cwdi; 892 struct vnode *vp, *tdp; 893 struct mount *mp; 894 struct file *fp; 895 int error; 896 897 /* getvnode() will use the descriptor for us */ 898 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 899 return (error); 900 vp = (struct vnode *)fp->f_data; 901 902 VREF(vp); 903 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 904 if (vp->v_type != VDIR) 905 error = ENOTDIR; 906 else 907 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 908 while (!error && (mp = vp->v_mountedhere) != NULL) { 909 if (vfs_busy(mp, 0, 0)) 910 continue; 911 error = VFS_ROOT(mp, &tdp); 912 vfs_unbusy(mp); 913 if (error) 914 break; 915 vput(vp); 916 vp = tdp; 917 } 918 if (error) { 919 vput(vp); 920 goto out; 921 } 922 VOP_UNLOCK(vp, 0); 923 924 /* 925 * Disallow changing to a directory not under the process's 926 * current root directory (if there is one). 927 */ 928 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) { 929 vrele(vp); 930 error = EPERM; /* operation not permitted */ 931 goto out; 932 } 933 934 vrele(cwdi->cwdi_cdir); 935 cwdi->cwdi_cdir = vp; 936 out: 937 FILE_UNUSE(fp, p); 938 return (error); 939 } 940 941 /* 942 * Change this process's notion of the root directory to a given file descriptor. 943 */ 944 945 int 946 sys_fchroot(l, v, retval) 947 struct lwp *l; 948 void *v; 949 register_t *retval; 950 { 951 struct sys_fchroot_args *uap = v; 952 struct proc *p = l->l_proc; 953 struct filedesc *fdp = p->p_fd; 954 struct cwdinfo *cwdi = p->p_cwdi; 955 struct vnode *vp; 956 struct file *fp; 957 int error; 958 959 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 960 return error; 961 /* getvnode() will use the descriptor for us */ 962 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 963 return error; 964 vp = (struct vnode *) fp->f_data; 965 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 966 if (vp->v_type != VDIR) 967 error = ENOTDIR; 968 else 969 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 970 VOP_UNLOCK(vp, 0); 971 if (error) 972 goto out; 973 VREF(vp); 974 975 /* 976 * Prevent escaping from chroot by putting the root under 977 * the working directory. Silently chdir to / if we aren't 978 * already there. 979 */ 980 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 981 /* 982 * XXX would be more failsafe to change directory to a 983 * deadfs node here instead 984 */ 985 vrele(cwdi->cwdi_cdir); 986 VREF(vp); 987 cwdi->cwdi_cdir = vp; 988 } 989 990 if (cwdi->cwdi_rdir != NULL) 991 vrele(cwdi->cwdi_rdir); 992 cwdi->cwdi_rdir = vp; 993 out: 994 FILE_UNUSE(fp, p); 995 return (error); 996 } 997 998 999 1000 /* 1001 * Change current working directory (``.''). 1002 */ 1003 /* ARGSUSED */ 1004 int 1005 sys_chdir(l, v, retval) 1006 struct lwp *l; 1007 void *v; 1008 register_t *retval; 1009 { 1010 struct sys_chdir_args /* { 1011 syscallarg(const char *) path; 1012 } */ *uap = v; 1013 struct proc *p = l->l_proc; 1014 struct cwdinfo *cwdi = p->p_cwdi; 1015 int error; 1016 struct nameidata nd; 1017 1018 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1019 SCARG(uap, path), p); 1020 if ((error = change_dir(&nd, p)) != 0) 1021 return (error); 1022 vrele(cwdi->cwdi_cdir); 1023 cwdi->cwdi_cdir = nd.ni_vp; 1024 return (0); 1025 } 1026 1027 /* 1028 * Change notion of root (``/'') directory. 1029 */ 1030 /* ARGSUSED */ 1031 int 1032 sys_chroot(l, v, retval) 1033 struct lwp *l; 1034 void *v; 1035 register_t *retval; 1036 { 1037 struct sys_chroot_args /* { 1038 syscallarg(const char *) path; 1039 } */ *uap = v; 1040 struct proc *p = l->l_proc; 1041 struct cwdinfo *cwdi = p->p_cwdi; 1042 struct vnode *vp; 1043 int error; 1044 struct nameidata nd; 1045 1046 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1047 return (error); 1048 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1049 SCARG(uap, path), p); 1050 if ((error = change_dir(&nd, p)) != 0) 1051 return (error); 1052 if (cwdi->cwdi_rdir != NULL) 1053 vrele(cwdi->cwdi_rdir); 1054 vp = nd.ni_vp; 1055 cwdi->cwdi_rdir = vp; 1056 1057 /* 1058 * Prevent escaping from chroot by putting the root under 1059 * the working directory. Silently chdir to / if we aren't 1060 * already there. 1061 */ 1062 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 1063 /* 1064 * XXX would be more failsafe to change directory to a 1065 * deadfs node here instead 1066 */ 1067 vrele(cwdi->cwdi_cdir); 1068 VREF(vp); 1069 cwdi->cwdi_cdir = vp; 1070 } 1071 1072 return (0); 1073 } 1074 1075 /* 1076 * Common routine for chroot and chdir. 1077 */ 1078 static int 1079 change_dir(ndp, p) 1080 struct nameidata *ndp; 1081 struct proc *p; 1082 { 1083 struct vnode *vp; 1084 int error; 1085 1086 if ((error = namei(ndp)) != 0) 1087 return (error); 1088 vp = ndp->ni_vp; 1089 if (vp->v_type != VDIR) 1090 error = ENOTDIR; 1091 else 1092 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 1093 1094 if (error) 1095 vput(vp); 1096 else 1097 VOP_UNLOCK(vp, 0); 1098 return (error); 1099 } 1100 1101 /* 1102 * Check permissions, allocate an open file structure, 1103 * and call the device open routine if any. 1104 */ 1105 int 1106 sys_open(l, v, retval) 1107 struct lwp *l; 1108 void *v; 1109 register_t *retval; 1110 { 1111 struct sys_open_args /* { 1112 syscallarg(const char *) path; 1113 syscallarg(int) flags; 1114 syscallarg(int) mode; 1115 } */ *uap = v; 1116 struct proc *p = l->l_proc; 1117 struct cwdinfo *cwdi = p->p_cwdi; 1118 struct filedesc *fdp = p->p_fd; 1119 struct file *fp; 1120 struct vnode *vp; 1121 int flags, cmode; 1122 int type, indx, error; 1123 struct flock lf; 1124 struct nameidata nd; 1125 1126 flags = FFLAGS(SCARG(uap, flags)); 1127 if ((flags & (FREAD | FWRITE)) == 0) 1128 return (EINVAL); 1129 /* falloc() will use the file descriptor for us */ 1130 if ((error = falloc(p, &fp, &indx)) != 0) 1131 return (error); 1132 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1133 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1134 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1135 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1136 FILE_UNUSE(fp, p); 1137 fdp->fd_ofiles[indx] = NULL; 1138 ffree(fp); 1139 if ((error == ENODEV || error == ENXIO) && 1140 l->l_dupfd >= 0 && /* XXX from fdopen */ 1141 (error = 1142 dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) { 1143 *retval = indx; 1144 return (0); 1145 } 1146 if (error == ERESTART) 1147 error = EINTR; 1148 fdremove(fdp, indx); 1149 return (error); 1150 } 1151 l->l_dupfd = 0; 1152 vp = nd.ni_vp; 1153 fp->f_flag = flags & FMASK; 1154 fp->f_type = DTYPE_VNODE; 1155 fp->f_ops = &vnops; 1156 fp->f_data = vp; 1157 if (flags & (O_EXLOCK | O_SHLOCK)) { 1158 lf.l_whence = SEEK_SET; 1159 lf.l_start = 0; 1160 lf.l_len = 0; 1161 if (flags & O_EXLOCK) 1162 lf.l_type = F_WRLCK; 1163 else 1164 lf.l_type = F_RDLCK; 1165 type = F_FLOCK; 1166 if ((flags & FNONBLOCK) == 0) 1167 type |= F_WAIT; 1168 VOP_UNLOCK(vp, 0); 1169 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1170 if (error) { 1171 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1172 FILE_UNUSE(fp, p); 1173 ffree(fp); 1174 fdremove(fdp, indx); 1175 return (error); 1176 } 1177 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1178 fp->f_flag |= FHASLOCK; 1179 } 1180 VOP_UNLOCK(vp, 0); 1181 *retval = indx; 1182 FILE_SET_MATURE(fp); 1183 FILE_UNUSE(fp, p); 1184 return (0); 1185 } 1186 1187 /* 1188 * Get file handle system call 1189 */ 1190 int 1191 sys_getfh(l, v, retval) 1192 struct lwp *l; 1193 void *v; 1194 register_t *retval; 1195 { 1196 struct sys_getfh_args /* { 1197 syscallarg(char *) fname; 1198 syscallarg(fhandle_t *) fhp; 1199 } */ *uap = v; 1200 struct proc *p = l->l_proc; 1201 struct vnode *vp; 1202 fhandle_t fh; 1203 int error; 1204 struct nameidata nd; 1205 1206 /* 1207 * Must be super user 1208 */ 1209 error = suser(p->p_ucred, &p->p_acflag); 1210 if (error) 1211 return (error); 1212 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1213 SCARG(uap, fname), p); 1214 error = namei(&nd); 1215 if (error) 1216 return (error); 1217 vp = nd.ni_vp; 1218 memset(&fh, 0, sizeof(fh)); 1219 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1220 error = VFS_VPTOFH(vp, &fh.fh_fid); 1221 vput(vp); 1222 if (error) 1223 return (error); 1224 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1225 return (error); 1226 } 1227 1228 /* 1229 * Open a file given a file handle. 1230 * 1231 * Check permissions, allocate an open file structure, 1232 * and call the device open routine if any. 1233 */ 1234 int 1235 sys_fhopen(l, v, retval) 1236 struct lwp *l; 1237 void *v; 1238 register_t *retval; 1239 { 1240 struct sys_fhopen_args /* { 1241 syscallarg(const fhandle_t *) fhp; 1242 syscallarg(int) flags; 1243 } */ *uap = v; 1244 struct proc *p = l->l_proc; 1245 struct filedesc *fdp = p->p_fd; 1246 struct file *fp; 1247 struct vnode *vp = NULL; 1248 struct mount *mp; 1249 struct ucred *cred = p->p_ucred; 1250 int flags; 1251 struct file *nfp; 1252 int type, indx, error=0; 1253 struct flock lf; 1254 struct vattr va; 1255 fhandle_t fh; 1256 1257 /* 1258 * Must be super user 1259 */ 1260 if ((error = suser(p->p_ucred, &p->p_acflag))) 1261 return (error); 1262 1263 flags = FFLAGS(SCARG(uap, flags)); 1264 if ((flags & (FREAD | FWRITE)) == 0) 1265 return (EINVAL); 1266 if ((flags & O_CREAT)) 1267 return (EINVAL); 1268 /* falloc() will use the file descriptor for us */ 1269 if ((error = falloc(p, &nfp, &indx)) != 0) 1270 return (error); 1271 fp = nfp; 1272 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1273 goto bad; 1274 1275 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1276 error = ESTALE; 1277 goto bad; 1278 } 1279 1280 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1281 vp = NULL; /* most likely unnecessary sanity for bad: */ 1282 goto bad; 1283 } 1284 1285 /* Now do an effective vn_open */ 1286 1287 if (vp->v_type == VSOCK) { 1288 error = EOPNOTSUPP; 1289 goto bad; 1290 } 1291 if (flags & FREAD) { 1292 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 1293 goto bad; 1294 } 1295 if (flags & (FWRITE | O_TRUNC)) { 1296 if (vp->v_type == VDIR) { 1297 error = EISDIR; 1298 goto bad; 1299 } 1300 if ((error = vn_writechk(vp)) != 0 || 1301 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 1302 goto bad; 1303 } 1304 if (flags & O_TRUNC) { 1305 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1306 goto bad; 1307 VOP_UNLOCK(vp, 0); /* XXX */ 1308 VOP_LEASE(vp, p, cred, LEASE_WRITE); 1309 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1310 VATTR_NULL(&va); 1311 va.va_size = 0; 1312 error = VOP_SETATTR(vp, &va, cred, p); 1313 vn_finished_write(mp, 0); 1314 if (error) 1315 goto bad; 1316 } 1317 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) 1318 goto bad; 1319 if (vp->v_type == VREG && 1320 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1321 error = EIO; 1322 goto bad; 1323 } 1324 if (flags & FWRITE) 1325 vp->v_writecount++; 1326 1327 /* done with modified vn_open, now finish what sys_open does. */ 1328 1329 fp->f_flag = flags & FMASK; 1330 fp->f_type = DTYPE_VNODE; 1331 fp->f_ops = &vnops; 1332 fp->f_data = vp; 1333 if (flags & (O_EXLOCK | O_SHLOCK)) { 1334 lf.l_whence = SEEK_SET; 1335 lf.l_start = 0; 1336 lf.l_len = 0; 1337 if (flags & O_EXLOCK) 1338 lf.l_type = F_WRLCK; 1339 else 1340 lf.l_type = F_RDLCK; 1341 type = F_FLOCK; 1342 if ((flags & FNONBLOCK) == 0) 1343 type |= F_WAIT; 1344 VOP_UNLOCK(vp, 0); 1345 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1346 if (error) { 1347 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1348 FILE_UNUSE(fp, p); 1349 ffree(fp); 1350 fdremove(fdp, indx); 1351 return (error); 1352 } 1353 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1354 fp->f_flag |= FHASLOCK; 1355 } 1356 VOP_UNLOCK(vp, 0); 1357 *retval = indx; 1358 FILE_SET_MATURE(fp); 1359 FILE_UNUSE(fp, p); 1360 return (0); 1361 1362 bad: 1363 FILE_UNUSE(fp, p); 1364 ffree(fp); 1365 fdremove(fdp, indx); 1366 if (vp != NULL) 1367 vput(vp); 1368 return (error); 1369 } 1370 1371 /* ARGSUSED */ 1372 int 1373 sys_fhstat(l, v, retval) 1374 struct lwp *l; 1375 void *v; 1376 register_t *retval; 1377 { 1378 struct sys_fhstat_args /* { 1379 syscallarg(const fhandle_t *) fhp; 1380 syscallarg(struct stat *) sb; 1381 } */ *uap = v; 1382 struct proc *p = l->l_proc; 1383 struct stat sb; 1384 int error; 1385 fhandle_t fh; 1386 struct mount *mp; 1387 struct vnode *vp; 1388 1389 /* 1390 * Must be super user 1391 */ 1392 if ((error = suser(p->p_ucred, &p->p_acflag))) 1393 return (error); 1394 1395 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1396 return (error); 1397 1398 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1399 return (ESTALE); 1400 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1401 return (error); 1402 error = vn_stat(vp, &sb, p); 1403 vput(vp); 1404 if (error) 1405 return (error); 1406 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1407 return (error); 1408 } 1409 1410 /* ARGSUSED */ 1411 int 1412 sys_fhstatvfs1(l, v, retval) 1413 struct lwp *l; 1414 void *v; 1415 register_t *retval; 1416 { 1417 struct sys_fhstatvfs1_args /* 1418 syscallarg(const fhandle_t *) fhp; 1419 syscallarg(struct statvfs *) buf; 1420 syscallarg(int) flags; 1421 } */ *uap = v; 1422 struct proc *p = l->l_proc; 1423 struct statvfs sbuf; 1424 fhandle_t fh; 1425 struct mount *mp; 1426 struct vnode *vp; 1427 int error; 1428 1429 /* 1430 * Must be super user 1431 */ 1432 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1433 return error; 1434 1435 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1436 return error; 1437 1438 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1439 return ESTALE; 1440 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1441 return error; 1442 1443 mp = vp->v_mount; 1444 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) { 1445 vput(vp); 1446 return error; 1447 } 1448 vput(vp); 1449 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1450 } 1451 1452 /* 1453 * Create a special file. 1454 */ 1455 /* ARGSUSED */ 1456 int 1457 sys_mknod(l, v, retval) 1458 struct lwp *l; 1459 void *v; 1460 register_t *retval; 1461 { 1462 struct sys_mknod_args /* { 1463 syscallarg(const char *) path; 1464 syscallarg(int) mode; 1465 syscallarg(int) dev; 1466 } */ *uap = v; 1467 struct proc *p = l->l_proc; 1468 struct vnode *vp; 1469 struct mount *mp; 1470 struct vattr vattr; 1471 int error; 1472 int whiteout = 0; 1473 struct nameidata nd; 1474 1475 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1476 return (error); 1477 restart: 1478 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1479 if ((error = namei(&nd)) != 0) 1480 return (error); 1481 vp = nd.ni_vp; 1482 if (vp != NULL) 1483 error = EEXIST; 1484 else { 1485 VATTR_NULL(&vattr); 1486 vattr.va_mode = 1487 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1488 vattr.va_rdev = SCARG(uap, dev); 1489 whiteout = 0; 1490 1491 switch (SCARG(uap, mode) & S_IFMT) { 1492 case S_IFMT: /* used by badsect to flag bad sectors */ 1493 vattr.va_type = VBAD; 1494 break; 1495 case S_IFCHR: 1496 vattr.va_type = VCHR; 1497 break; 1498 case S_IFBLK: 1499 vattr.va_type = VBLK; 1500 break; 1501 case S_IFWHT: 1502 whiteout = 1; 1503 break; 1504 default: 1505 error = EINVAL; 1506 break; 1507 } 1508 } 1509 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1510 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1511 if (nd.ni_dvp == vp) 1512 vrele(nd.ni_dvp); 1513 else 1514 vput(nd.ni_dvp); 1515 if (vp) 1516 vrele(vp); 1517 if ((error = vn_start_write(NULL, &mp, 1518 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1519 return (error); 1520 goto restart; 1521 } 1522 if (!error) { 1523 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1524 if (whiteout) { 1525 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1526 if (error) 1527 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1528 vput(nd.ni_dvp); 1529 } else { 1530 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1531 &nd.ni_cnd, &vattr); 1532 if (error == 0) 1533 vput(nd.ni_vp); 1534 } 1535 } else { 1536 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1537 if (nd.ni_dvp == vp) 1538 vrele(nd.ni_dvp); 1539 else 1540 vput(nd.ni_dvp); 1541 if (vp) 1542 vrele(vp); 1543 } 1544 vn_finished_write(mp, 0); 1545 return (error); 1546 } 1547 1548 /* 1549 * Create a named pipe. 1550 */ 1551 /* ARGSUSED */ 1552 int 1553 sys_mkfifo(l, v, retval) 1554 struct lwp *l; 1555 void *v; 1556 register_t *retval; 1557 { 1558 struct sys_mkfifo_args /* { 1559 syscallarg(const char *) path; 1560 syscallarg(int) mode; 1561 } */ *uap = v; 1562 struct proc *p = l->l_proc; 1563 struct mount *mp; 1564 struct vattr vattr; 1565 int error; 1566 struct nameidata nd; 1567 1568 restart: 1569 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1570 if ((error = namei(&nd)) != 0) 1571 return (error); 1572 if (nd.ni_vp != NULL) { 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 vrele(nd.ni_vp); 1579 return (EEXIST); 1580 } 1581 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1582 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1583 if (nd.ni_dvp == nd.ni_vp) 1584 vrele(nd.ni_dvp); 1585 else 1586 vput(nd.ni_dvp); 1587 if (nd.ni_vp) 1588 vrele(nd.ni_vp); 1589 if ((error = vn_start_write(NULL, &mp, 1590 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1591 return (error); 1592 goto restart; 1593 } 1594 VATTR_NULL(&vattr); 1595 vattr.va_type = VFIFO; 1596 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1597 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1598 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1599 if (error == 0) 1600 vput(nd.ni_vp); 1601 vn_finished_write(mp, 0); 1602 return (error); 1603 } 1604 1605 /* 1606 * Make a hard file link. 1607 */ 1608 /* ARGSUSED */ 1609 int 1610 sys_link(l, v, retval) 1611 struct lwp *l; 1612 void *v; 1613 register_t *retval; 1614 { 1615 struct sys_link_args /* { 1616 syscallarg(const char *) path; 1617 syscallarg(const char *) link; 1618 } */ *uap = v; 1619 struct proc *p = l->l_proc; 1620 struct vnode *vp; 1621 struct mount *mp; 1622 struct nameidata nd; 1623 int error; 1624 1625 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1626 if ((error = namei(&nd)) != 0) 1627 return (error); 1628 vp = nd.ni_vp; 1629 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1630 vrele(vp); 1631 return (error); 1632 } 1633 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1634 if ((error = namei(&nd)) != 0) 1635 goto out; 1636 if (nd.ni_vp) { 1637 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1638 if (nd.ni_dvp == nd.ni_vp) 1639 vrele(nd.ni_dvp); 1640 else 1641 vput(nd.ni_dvp); 1642 vrele(nd.ni_vp); 1643 error = EEXIST; 1644 goto out; 1645 } 1646 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1647 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1648 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1649 out: 1650 vrele(vp); 1651 vn_finished_write(mp, 0); 1652 return (error); 1653 } 1654 1655 /* 1656 * Make a symbolic link. 1657 */ 1658 /* ARGSUSED */ 1659 int 1660 sys_symlink(l, v, retval) 1661 struct lwp *l; 1662 void *v; 1663 register_t *retval; 1664 { 1665 struct sys_symlink_args /* { 1666 syscallarg(const char *) path; 1667 syscallarg(const char *) link; 1668 } */ *uap = v; 1669 struct proc *p = l->l_proc; 1670 struct mount *mp; 1671 struct vattr vattr; 1672 char *path; 1673 int error; 1674 struct nameidata nd; 1675 1676 path = PNBUF_GET(); 1677 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1678 if (error) 1679 goto out; 1680 restart: 1681 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1682 if ((error = namei(&nd)) != 0) 1683 goto out; 1684 if (nd.ni_vp) { 1685 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1686 if (nd.ni_dvp == nd.ni_vp) 1687 vrele(nd.ni_dvp); 1688 else 1689 vput(nd.ni_dvp); 1690 vrele(nd.ni_vp); 1691 error = EEXIST; 1692 goto out; 1693 } 1694 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1695 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1696 if (nd.ni_dvp == nd.ni_vp) 1697 vrele(nd.ni_dvp); 1698 else 1699 vput(nd.ni_dvp); 1700 if ((error = vn_start_write(NULL, &mp, 1701 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1702 return (error); 1703 goto restart; 1704 } 1705 VATTR_NULL(&vattr); 1706 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1707 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1708 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1709 if (error == 0) 1710 vput(nd.ni_vp); 1711 vn_finished_write(mp, 0); 1712 out: 1713 PNBUF_PUT(path); 1714 return (error); 1715 } 1716 1717 /* 1718 * Delete a whiteout from the filesystem. 1719 */ 1720 /* ARGSUSED */ 1721 int 1722 sys_undelete(l, v, retval) 1723 struct lwp *l; 1724 void *v; 1725 register_t *retval; 1726 { 1727 struct sys_undelete_args /* { 1728 syscallarg(const char *) path; 1729 } */ *uap = v; 1730 struct proc *p = l->l_proc; 1731 int error; 1732 struct mount *mp; 1733 struct nameidata nd; 1734 1735 restart: 1736 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1737 SCARG(uap, path), p); 1738 error = namei(&nd); 1739 if (error) 1740 return (error); 1741 1742 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1743 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1744 if (nd.ni_dvp == nd.ni_vp) 1745 vrele(nd.ni_dvp); 1746 else 1747 vput(nd.ni_dvp); 1748 if (nd.ni_vp) 1749 vrele(nd.ni_vp); 1750 return (EEXIST); 1751 } 1752 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1753 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1754 if (nd.ni_dvp == nd.ni_vp) 1755 vrele(nd.ni_dvp); 1756 else 1757 vput(nd.ni_dvp); 1758 if ((error = vn_start_write(NULL, &mp, 1759 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1760 return (error); 1761 goto restart; 1762 } 1763 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1764 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1765 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1766 vput(nd.ni_dvp); 1767 vn_finished_write(mp, 0); 1768 return (error); 1769 } 1770 1771 /* 1772 * Delete a name from the filesystem. 1773 */ 1774 /* ARGSUSED */ 1775 int 1776 sys_unlink(l, v, retval) 1777 struct lwp *l; 1778 void *v; 1779 register_t *retval; 1780 { 1781 struct sys_unlink_args /* { 1782 syscallarg(const char *) path; 1783 } */ *uap = v; 1784 struct proc *p = l->l_proc; 1785 struct mount *mp; 1786 struct vnode *vp; 1787 int error; 1788 struct nameidata nd; 1789 1790 restart: 1791 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1792 SCARG(uap, path), p); 1793 if ((error = namei(&nd)) != 0) 1794 return (error); 1795 vp = nd.ni_vp; 1796 1797 /* 1798 * The root of a mounted filesystem cannot be deleted. 1799 */ 1800 if (vp->v_flag & VROOT) { 1801 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1802 if (nd.ni_dvp == vp) 1803 vrele(nd.ni_dvp); 1804 else 1805 vput(nd.ni_dvp); 1806 vput(vp); 1807 error = EBUSY; 1808 goto out; 1809 } 1810 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1811 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1812 if (nd.ni_dvp == vp) 1813 vrele(nd.ni_dvp); 1814 else 1815 vput(nd.ni_dvp); 1816 vput(vp); 1817 if ((error = vn_start_write(NULL, &mp, 1818 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1819 return (error); 1820 goto restart; 1821 } 1822 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1823 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1824 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1825 vn_finished_write(mp, 0); 1826 out: 1827 return (error); 1828 } 1829 1830 /* 1831 * Reposition read/write file offset. 1832 */ 1833 int 1834 sys_lseek(l, v, retval) 1835 struct lwp *l; 1836 void *v; 1837 register_t *retval; 1838 { 1839 struct sys_lseek_args /* { 1840 syscallarg(int) fd; 1841 syscallarg(int) pad; 1842 syscallarg(off_t) offset; 1843 syscallarg(int) whence; 1844 } */ *uap = v; 1845 struct proc *p = l->l_proc; 1846 struct ucred *cred = p->p_ucred; 1847 struct filedesc *fdp = p->p_fd; 1848 struct file *fp; 1849 struct vnode *vp; 1850 struct vattr vattr; 1851 off_t newoff; 1852 int error; 1853 1854 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1855 return (EBADF); 1856 1857 FILE_USE(fp); 1858 1859 vp = (struct vnode *)fp->f_data; 1860 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1861 error = ESPIPE; 1862 goto out; 1863 } 1864 1865 switch (SCARG(uap, whence)) { 1866 case SEEK_CUR: 1867 newoff = fp->f_offset + SCARG(uap, offset); 1868 break; 1869 case SEEK_END: 1870 error = VOP_GETATTR(vp, &vattr, cred, p); 1871 if (error) 1872 goto out; 1873 newoff = SCARG(uap, offset) + vattr.va_size; 1874 break; 1875 case SEEK_SET: 1876 newoff = SCARG(uap, offset); 1877 break; 1878 default: 1879 error = EINVAL; 1880 goto out; 1881 } 1882 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1883 goto out; 1884 1885 *(off_t *)retval = fp->f_offset = newoff; 1886 out: 1887 FILE_UNUSE(fp, p); 1888 return (error); 1889 } 1890 1891 /* 1892 * Positional read system call. 1893 */ 1894 int 1895 sys_pread(l, v, retval) 1896 struct lwp *l; 1897 void *v; 1898 register_t *retval; 1899 { 1900 struct sys_pread_args /* { 1901 syscallarg(int) fd; 1902 syscallarg(void *) buf; 1903 syscallarg(size_t) nbyte; 1904 syscallarg(off_t) offset; 1905 } */ *uap = v; 1906 struct proc *p = l->l_proc; 1907 struct filedesc *fdp = p->p_fd; 1908 struct file *fp; 1909 struct vnode *vp; 1910 off_t offset; 1911 int error, fd = SCARG(uap, fd); 1912 1913 if ((fp = fd_getfile(fdp, fd)) == NULL) 1914 return (EBADF); 1915 1916 if ((fp->f_flag & FREAD) == 0) { 1917 simple_unlock(&fp->f_slock); 1918 return (EBADF); 1919 } 1920 1921 FILE_USE(fp); 1922 1923 vp = (struct vnode *)fp->f_data; 1924 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1925 error = ESPIPE; 1926 goto out; 1927 } 1928 1929 offset = SCARG(uap, offset); 1930 1931 /* 1932 * XXX This works because no file systems actually 1933 * XXX take any action on the seek operation. 1934 */ 1935 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1936 goto out; 1937 1938 /* dofileread() will unuse the descriptor for us */ 1939 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1940 &offset, 0, retval)); 1941 1942 out: 1943 FILE_UNUSE(fp, p); 1944 return (error); 1945 } 1946 1947 /* 1948 * Positional scatter read system call. 1949 */ 1950 int 1951 sys_preadv(l, v, retval) 1952 struct lwp *l; 1953 void *v; 1954 register_t *retval; 1955 { 1956 struct sys_preadv_args /* { 1957 syscallarg(int) fd; 1958 syscallarg(const struct iovec *) iovp; 1959 syscallarg(int) iovcnt; 1960 syscallarg(off_t) offset; 1961 } */ *uap = v; 1962 struct proc *p = l->l_proc; 1963 struct filedesc *fdp = p->p_fd; 1964 struct file *fp; 1965 struct vnode *vp; 1966 off_t offset; 1967 int error, fd = SCARG(uap, fd); 1968 1969 if ((fp = fd_getfile(fdp, fd)) == NULL) 1970 return (EBADF); 1971 1972 if ((fp->f_flag & FREAD) == 0) { 1973 simple_unlock(&fp->f_slock); 1974 return (EBADF); 1975 } 1976 1977 FILE_USE(fp); 1978 1979 vp = (struct vnode *)fp->f_data; 1980 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1981 error = ESPIPE; 1982 goto out; 1983 } 1984 1985 offset = SCARG(uap, offset); 1986 1987 /* 1988 * XXX This works because no file systems actually 1989 * XXX take any action on the seek operation. 1990 */ 1991 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1992 goto out; 1993 1994 /* dofilereadv() will unuse the descriptor for us */ 1995 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1996 &offset, 0, retval)); 1997 1998 out: 1999 FILE_UNUSE(fp, p); 2000 return (error); 2001 } 2002 2003 /* 2004 * Positional write system call. 2005 */ 2006 int 2007 sys_pwrite(l, v, retval) 2008 struct lwp *l; 2009 void *v; 2010 register_t *retval; 2011 { 2012 struct sys_pwrite_args /* { 2013 syscallarg(int) fd; 2014 syscallarg(const void *) buf; 2015 syscallarg(size_t) nbyte; 2016 syscallarg(off_t) offset; 2017 } */ *uap = v; 2018 struct proc *p = l->l_proc; 2019 struct filedesc *fdp = p->p_fd; 2020 struct file *fp; 2021 struct vnode *vp; 2022 off_t offset; 2023 int error, fd = SCARG(uap, fd); 2024 2025 if ((fp = fd_getfile(fdp, fd)) == NULL) 2026 return (EBADF); 2027 2028 if ((fp->f_flag & FWRITE) == 0) { 2029 simple_unlock(&fp->f_slock); 2030 return (EBADF); 2031 } 2032 2033 FILE_USE(fp); 2034 2035 vp = (struct vnode *)fp->f_data; 2036 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2037 error = ESPIPE; 2038 goto out; 2039 } 2040 2041 offset = SCARG(uap, offset); 2042 2043 /* 2044 * XXX This works because no file systems actually 2045 * XXX take any action on the seek operation. 2046 */ 2047 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2048 goto out; 2049 2050 /* dofilewrite() will unuse the descriptor for us */ 2051 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2052 &offset, 0, retval)); 2053 2054 out: 2055 FILE_UNUSE(fp, p); 2056 return (error); 2057 } 2058 2059 /* 2060 * Positional gather write system call. 2061 */ 2062 int 2063 sys_pwritev(l, v, retval) 2064 struct lwp *l; 2065 void *v; 2066 register_t *retval; 2067 { 2068 struct sys_pwritev_args /* { 2069 syscallarg(int) fd; 2070 syscallarg(const struct iovec *) iovp; 2071 syscallarg(int) iovcnt; 2072 syscallarg(off_t) offset; 2073 } */ *uap = v; 2074 struct proc *p = l->l_proc; 2075 struct filedesc *fdp = p->p_fd; 2076 struct file *fp; 2077 struct vnode *vp; 2078 off_t offset; 2079 int error, fd = SCARG(uap, fd); 2080 2081 if ((fp = fd_getfile(fdp, fd)) == NULL) 2082 return (EBADF); 2083 2084 if ((fp->f_flag & FWRITE) == 0) { 2085 simple_unlock(&fp->f_slock); 2086 return (EBADF); 2087 } 2088 2089 FILE_USE(fp); 2090 2091 vp = (struct vnode *)fp->f_data; 2092 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2093 error = ESPIPE; 2094 goto out; 2095 } 2096 2097 offset = SCARG(uap, offset); 2098 2099 /* 2100 * XXX This works because no file systems actually 2101 * XXX take any action on the seek operation. 2102 */ 2103 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2104 goto out; 2105 2106 /* dofilewritev() will unuse the descriptor for us */ 2107 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2108 &offset, 0, retval)); 2109 2110 out: 2111 FILE_UNUSE(fp, p); 2112 return (error); 2113 } 2114 2115 /* 2116 * Check access permissions. 2117 */ 2118 int 2119 sys_access(l, v, retval) 2120 struct lwp *l; 2121 void *v; 2122 register_t *retval; 2123 { 2124 struct sys_access_args /* { 2125 syscallarg(const char *) path; 2126 syscallarg(int) flags; 2127 } */ *uap = v; 2128 struct proc *p = l->l_proc; 2129 struct ucred *cred; 2130 struct vnode *vp; 2131 int error, flags; 2132 struct nameidata nd; 2133 2134 cred = crdup(p->p_ucred); 2135 cred->cr_uid = p->p_cred->p_ruid; 2136 cred->cr_gid = p->p_cred->p_rgid; 2137 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2138 SCARG(uap, path), p); 2139 /* Override default credentials */ 2140 nd.ni_cnd.cn_cred = cred; 2141 if ((error = namei(&nd)) != 0) 2142 goto out; 2143 vp = nd.ni_vp; 2144 2145 /* Flags == 0 means only check for existence. */ 2146 if (SCARG(uap, flags)) { 2147 flags = 0; 2148 if (SCARG(uap, flags) & R_OK) 2149 flags |= VREAD; 2150 if (SCARG(uap, flags) & W_OK) 2151 flags |= VWRITE; 2152 if (SCARG(uap, flags) & X_OK) 2153 flags |= VEXEC; 2154 2155 error = VOP_ACCESS(vp, flags, cred, p); 2156 if (!error && (flags & VWRITE)) 2157 error = vn_writechk(vp); 2158 } 2159 vput(vp); 2160 out: 2161 crfree(cred); 2162 return (error); 2163 } 2164 2165 /* 2166 * Get file status; this version follows links. 2167 */ 2168 /* ARGSUSED */ 2169 int 2170 sys___stat13(l, v, retval) 2171 struct lwp *l; 2172 void *v; 2173 register_t *retval; 2174 { 2175 struct sys___stat13_args /* { 2176 syscallarg(const char *) path; 2177 syscallarg(struct stat *) ub; 2178 } */ *uap = v; 2179 struct proc *p = l->l_proc; 2180 struct stat sb; 2181 int error; 2182 struct nameidata nd; 2183 2184 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2185 SCARG(uap, path), p); 2186 if ((error = namei(&nd)) != 0) 2187 return (error); 2188 error = vn_stat(nd.ni_vp, &sb, p); 2189 vput(nd.ni_vp); 2190 if (error) 2191 return (error); 2192 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2193 return (error); 2194 } 2195 2196 /* 2197 * Get file status; this version does not follow links. 2198 */ 2199 /* ARGSUSED */ 2200 int 2201 sys___lstat13(l, v, retval) 2202 struct lwp *l; 2203 void *v; 2204 register_t *retval; 2205 { 2206 struct sys___lstat13_args /* { 2207 syscallarg(const char *) path; 2208 syscallarg(struct stat *) ub; 2209 } */ *uap = v; 2210 struct proc *p = l->l_proc; 2211 struct stat sb; 2212 int error; 2213 struct nameidata nd; 2214 2215 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2216 SCARG(uap, path), p); 2217 if ((error = namei(&nd)) != 0) 2218 return (error); 2219 error = vn_stat(nd.ni_vp, &sb, p); 2220 vput(nd.ni_vp); 2221 if (error) 2222 return (error); 2223 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2224 return (error); 2225 } 2226 2227 /* 2228 * Get configurable pathname variables. 2229 */ 2230 /* ARGSUSED */ 2231 int 2232 sys_pathconf(l, v, retval) 2233 struct lwp *l; 2234 void *v; 2235 register_t *retval; 2236 { 2237 struct sys_pathconf_args /* { 2238 syscallarg(const char *) path; 2239 syscallarg(int) name; 2240 } */ *uap = v; 2241 struct proc *p = l->l_proc; 2242 int error; 2243 struct nameidata nd; 2244 2245 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2246 SCARG(uap, path), p); 2247 if ((error = namei(&nd)) != 0) 2248 return (error); 2249 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2250 vput(nd.ni_vp); 2251 return (error); 2252 } 2253 2254 /* 2255 * Return target name of a symbolic link. 2256 */ 2257 /* ARGSUSED */ 2258 int 2259 sys_readlink(l, v, retval) 2260 struct lwp *l; 2261 void *v; 2262 register_t *retval; 2263 { 2264 struct sys_readlink_args /* { 2265 syscallarg(const char *) path; 2266 syscallarg(char *) buf; 2267 syscallarg(size_t) count; 2268 } */ *uap = v; 2269 struct proc *p = l->l_proc; 2270 struct vnode *vp; 2271 struct iovec aiov; 2272 struct uio auio; 2273 int error; 2274 struct nameidata nd; 2275 2276 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2277 SCARG(uap, path), p); 2278 if ((error = namei(&nd)) != 0) 2279 return (error); 2280 vp = nd.ni_vp; 2281 if (vp->v_type != VLNK) 2282 error = EINVAL; 2283 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2284 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 2285 aiov.iov_base = SCARG(uap, buf); 2286 aiov.iov_len = SCARG(uap, count); 2287 auio.uio_iov = &aiov; 2288 auio.uio_iovcnt = 1; 2289 auio.uio_offset = 0; 2290 auio.uio_rw = UIO_READ; 2291 auio.uio_segflg = UIO_USERSPACE; 2292 auio.uio_procp = p; 2293 auio.uio_resid = SCARG(uap, count); 2294 error = VOP_READLINK(vp, &auio, p->p_ucred); 2295 } 2296 vput(vp); 2297 *retval = SCARG(uap, count) - auio.uio_resid; 2298 return (error); 2299 } 2300 2301 /* 2302 * Change flags of a file given a path name. 2303 */ 2304 /* ARGSUSED */ 2305 int 2306 sys_chflags(l, v, retval) 2307 struct lwp *l; 2308 void *v; 2309 register_t *retval; 2310 { 2311 struct sys_chflags_args /* { 2312 syscallarg(const char *) path; 2313 syscallarg(u_long) flags; 2314 } */ *uap = v; 2315 struct proc *p = l->l_proc; 2316 struct vnode *vp; 2317 int error; 2318 struct nameidata nd; 2319 2320 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2321 if ((error = namei(&nd)) != 0) 2322 return (error); 2323 vp = nd.ni_vp; 2324 error = change_flags(vp, SCARG(uap, flags), p); 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), p); 2353 VOP_UNLOCK(vp, 0); 2354 FILE_UNUSE(fp, p); 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 proc *p = l->l_proc; 2373 struct vnode *vp; 2374 int error; 2375 struct nameidata nd; 2376 2377 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2378 if ((error = namei(&nd)) != 0) 2379 return (error); 2380 vp = nd.ni_vp; 2381 error = change_flags(vp, SCARG(uap, flags), p); 2382 vput(vp); 2383 return (error); 2384 } 2385 2386 /* 2387 * Common routine to change flags of a file. 2388 */ 2389 int 2390 change_flags(vp, flags, p) 2391 struct vnode *vp; 2392 u_long flags; 2393 struct proc *p; 2394 { 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, p, 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, p)) != 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, p); 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 struct proc *p = l->l_proc; 2438 int error; 2439 struct nameidata nd; 2440 2441 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2442 if ((error = namei(&nd)) != 0) 2443 return (error); 2444 2445 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2446 2447 vrele(nd.ni_vp); 2448 return (error); 2449 } 2450 2451 /* 2452 * Change mode of a file given a file descriptor. 2453 */ 2454 /* ARGSUSED */ 2455 int 2456 sys_fchmod(l, v, retval) 2457 struct lwp *l; 2458 void *v; 2459 register_t *retval; 2460 { 2461 struct sys_fchmod_args /* { 2462 syscallarg(int) fd; 2463 syscallarg(int) mode; 2464 } */ *uap = v; 2465 struct proc *p = l->l_proc; 2466 struct file *fp; 2467 int error; 2468 2469 /* getvnode() will use the descriptor for us */ 2470 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2471 return (error); 2472 2473 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 2474 FILE_UNUSE(fp, p); 2475 return (error); 2476 } 2477 2478 /* 2479 * Change mode of a file given path name; this version does not follow links. 2480 */ 2481 /* ARGSUSED */ 2482 int 2483 sys_lchmod(l, v, retval) 2484 struct lwp *l; 2485 void *v; 2486 register_t *retval; 2487 { 2488 struct sys_lchmod_args /* { 2489 syscallarg(const char *) path; 2490 syscallarg(int) mode; 2491 } */ *uap = v; 2492 struct proc *p = l->l_proc; 2493 int error; 2494 struct nameidata nd; 2495 2496 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2497 if ((error = namei(&nd)) != 0) 2498 return (error); 2499 2500 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2501 2502 vrele(nd.ni_vp); 2503 return (error); 2504 } 2505 2506 /* 2507 * Common routine to set mode given a vnode. 2508 */ 2509 static int 2510 change_mode(vp, mode, p) 2511 struct vnode *vp; 2512 int mode; 2513 struct proc *p; 2514 { 2515 struct mount *mp; 2516 struct vattr vattr; 2517 int error; 2518 2519 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2520 return (error); 2521 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2522 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2523 VATTR_NULL(&vattr); 2524 vattr.va_mode = mode & ALLPERMS; 2525 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2526 VOP_UNLOCK(vp, 0); 2527 vn_finished_write(mp, 0); 2528 return (error); 2529 } 2530 2531 /* 2532 * Set ownership given a path name; this version follows links. 2533 */ 2534 /* ARGSUSED */ 2535 int 2536 sys_chown(l, v, retval) 2537 struct lwp *l; 2538 void *v; 2539 register_t *retval; 2540 { 2541 struct sys_chown_args /* { 2542 syscallarg(const char *) path; 2543 syscallarg(uid_t) uid; 2544 syscallarg(gid_t) gid; 2545 } */ *uap = v; 2546 struct proc *p = l->l_proc; 2547 int error; 2548 struct nameidata nd; 2549 2550 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2551 if ((error = namei(&nd)) != 0) 2552 return (error); 2553 2554 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2555 2556 vrele(nd.ni_vp); 2557 return (error); 2558 } 2559 2560 /* 2561 * Set ownership given a path name; this version follows links. 2562 * Provides POSIX semantics. 2563 */ 2564 /* ARGSUSED */ 2565 int 2566 sys___posix_chown(l, v, retval) 2567 struct lwp *l; 2568 void *v; 2569 register_t *retval; 2570 { 2571 struct sys_chown_args /* { 2572 syscallarg(const char *) path; 2573 syscallarg(uid_t) uid; 2574 syscallarg(gid_t) gid; 2575 } */ *uap = v; 2576 struct proc *p = l->l_proc; 2577 int error; 2578 struct nameidata nd; 2579 2580 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2581 if ((error = namei(&nd)) != 0) 2582 return (error); 2583 2584 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2585 2586 vrele(nd.ni_vp); 2587 return (error); 2588 } 2589 2590 /* 2591 * Set ownership given a file descriptor. 2592 */ 2593 /* ARGSUSED */ 2594 int 2595 sys_fchown(l, v, retval) 2596 struct lwp *l; 2597 void *v; 2598 register_t *retval; 2599 { 2600 struct sys_fchown_args /* { 2601 syscallarg(int) fd; 2602 syscallarg(uid_t) uid; 2603 syscallarg(gid_t) gid; 2604 } */ *uap = v; 2605 struct proc *p = l->l_proc; 2606 int error; 2607 struct file *fp; 2608 2609 /* getvnode() will use the descriptor for us */ 2610 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2611 return (error); 2612 2613 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2614 SCARG(uap, gid), p, 0); 2615 FILE_UNUSE(fp, p); 2616 return (error); 2617 } 2618 2619 /* 2620 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2621 */ 2622 /* ARGSUSED */ 2623 int 2624 sys___posix_fchown(l, v, retval) 2625 struct lwp *l; 2626 void *v; 2627 register_t *retval; 2628 { 2629 struct sys_fchown_args /* { 2630 syscallarg(int) fd; 2631 syscallarg(uid_t) uid; 2632 syscallarg(gid_t) gid; 2633 } */ *uap = v; 2634 struct proc *p = l->l_proc; 2635 int error; 2636 struct file *fp; 2637 2638 /* getvnode() will use the descriptor for us */ 2639 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2640 return (error); 2641 2642 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2643 SCARG(uap, gid), p, 1); 2644 FILE_UNUSE(fp, p); 2645 return (error); 2646 } 2647 2648 /* 2649 * Set ownership given a path name; this version does not follow links. 2650 */ 2651 /* ARGSUSED */ 2652 int 2653 sys_lchown(l, v, retval) 2654 struct lwp *l; 2655 void *v; 2656 register_t *retval; 2657 { 2658 struct sys_lchown_args /* { 2659 syscallarg(const char *) path; 2660 syscallarg(uid_t) uid; 2661 syscallarg(gid_t) gid; 2662 } */ *uap = v; 2663 struct proc *p = l->l_proc; 2664 int error; 2665 struct nameidata nd; 2666 2667 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2668 if ((error = namei(&nd)) != 0) 2669 return (error); 2670 2671 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2672 2673 vrele(nd.ni_vp); 2674 return (error); 2675 } 2676 2677 /* 2678 * Set ownership given a path name; this version does not follow links. 2679 * Provides POSIX/XPG semantics. 2680 */ 2681 /* ARGSUSED */ 2682 int 2683 sys___posix_lchown(l, v, retval) 2684 struct lwp *l; 2685 void *v; 2686 register_t *retval; 2687 { 2688 struct sys_lchown_args /* { 2689 syscallarg(const char *) path; 2690 syscallarg(uid_t) uid; 2691 syscallarg(gid_t) gid; 2692 } */ *uap = v; 2693 struct proc *p = l->l_proc; 2694 int error; 2695 struct nameidata nd; 2696 2697 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2698 if ((error = namei(&nd)) != 0) 2699 return (error); 2700 2701 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2702 2703 vrele(nd.ni_vp); 2704 return (error); 2705 } 2706 2707 /* 2708 * Common routine to set ownership given a vnode. 2709 */ 2710 static int 2711 change_owner(vp, uid, gid, p, posix_semantics) 2712 struct vnode *vp; 2713 uid_t uid; 2714 gid_t gid; 2715 struct proc *p; 2716 int posix_semantics; 2717 { 2718 struct mount *mp; 2719 struct vattr vattr; 2720 mode_t newmode; 2721 int error; 2722 2723 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2724 return (error); 2725 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2726 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2727 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2728 goto out; 2729 2730 #define CHANGED(x) ((int)(x) != -1) 2731 newmode = vattr.va_mode; 2732 if (posix_semantics) { 2733 /* 2734 * POSIX/XPG semantics: if the caller is not the super-user, 2735 * clear set-user-id and set-group-id bits. Both POSIX and 2736 * the XPG consider the behaviour for calls by the super-user 2737 * implementation-defined; we leave the set-user-id and set- 2738 * group-id settings intact in that case. 2739 */ 2740 if (suser(p->p_ucred, NULL) != 0) 2741 newmode &= ~(S_ISUID | S_ISGID); 2742 } else { 2743 /* 2744 * NetBSD semantics: when changing owner and/or group, 2745 * clear the respective bit(s). 2746 */ 2747 if (CHANGED(uid)) 2748 newmode &= ~S_ISUID; 2749 if (CHANGED(gid)) 2750 newmode &= ~S_ISGID; 2751 } 2752 /* Update va_mode iff altered. */ 2753 if (vattr.va_mode == newmode) 2754 newmode = VNOVAL; 2755 2756 VATTR_NULL(&vattr); 2757 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2758 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2759 vattr.va_mode = newmode; 2760 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2761 #undef CHANGED 2762 2763 out: 2764 VOP_UNLOCK(vp, 0); 2765 vn_finished_write(mp, 0); 2766 return (error); 2767 } 2768 2769 /* 2770 * Set the access and modification times given a path name; this 2771 * version follows links. 2772 */ 2773 /* ARGSUSED */ 2774 int 2775 sys_utimes(l, v, retval) 2776 struct lwp *l; 2777 void *v; 2778 register_t *retval; 2779 { 2780 struct sys_utimes_args /* { 2781 syscallarg(const char *) path; 2782 syscallarg(const struct timeval *) tptr; 2783 } */ *uap = v; 2784 struct proc *p = l->l_proc; 2785 int error; 2786 struct nameidata nd; 2787 2788 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2789 if ((error = namei(&nd)) != 0) 2790 return (error); 2791 2792 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2793 2794 vrele(nd.ni_vp); 2795 return (error); 2796 } 2797 2798 /* 2799 * Set the access and modification times given a file descriptor. 2800 */ 2801 /* ARGSUSED */ 2802 int 2803 sys_futimes(l, v, retval) 2804 struct lwp *l; 2805 void *v; 2806 register_t *retval; 2807 { 2808 struct sys_futimes_args /* { 2809 syscallarg(int) fd; 2810 syscallarg(const struct timeval *) tptr; 2811 } */ *uap = v; 2812 struct proc *p = l->l_proc; 2813 int error; 2814 struct file *fp; 2815 2816 /* getvnode() will use the descriptor for us */ 2817 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2818 return (error); 2819 2820 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2821 FILE_UNUSE(fp, p); 2822 return (error); 2823 } 2824 2825 /* 2826 * Set the access and modification times given a path name; this 2827 * version does not follow links. 2828 */ 2829 /* ARGSUSED */ 2830 int 2831 sys_lutimes(l, v, retval) 2832 struct lwp *l; 2833 void *v; 2834 register_t *retval; 2835 { 2836 struct sys_lutimes_args /* { 2837 syscallarg(const char *) path; 2838 syscallarg(const struct timeval *) tptr; 2839 } */ *uap = v; 2840 struct proc *p = l->l_proc; 2841 int error; 2842 struct nameidata nd; 2843 2844 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2845 if ((error = namei(&nd)) != 0) 2846 return (error); 2847 2848 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2849 2850 vrele(nd.ni_vp); 2851 return (error); 2852 } 2853 2854 /* 2855 * Common routine to set access and modification times given a vnode. 2856 */ 2857 static int 2858 change_utimes(vp, tptr, p) 2859 struct vnode *vp; 2860 const struct timeval *tptr; 2861 struct proc *p; 2862 { 2863 struct timeval tv[2]; 2864 struct mount *mp; 2865 struct vattr vattr; 2866 int error; 2867 2868 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2869 return (error); 2870 VATTR_NULL(&vattr); 2871 if (tptr == NULL) { 2872 microtime(&tv[0]); 2873 tv[1] = tv[0]; 2874 vattr.va_vaflags |= VA_UTIMES_NULL; 2875 } else { 2876 error = copyin(tptr, tv, sizeof(tv)); 2877 if (error) 2878 goto out; 2879 } 2880 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2881 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2882 vattr.va_atime.tv_sec = tv[0].tv_sec; 2883 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2884 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2885 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2886 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2887 VOP_UNLOCK(vp, 0); 2888 out: 2889 vn_finished_write(mp, 0); 2890 return (error); 2891 } 2892 2893 /* 2894 * Truncate a file given its path name. 2895 */ 2896 /* ARGSUSED */ 2897 int 2898 sys_truncate(l, v, retval) 2899 struct lwp *l; 2900 void *v; 2901 register_t *retval; 2902 { 2903 struct sys_truncate_args /* { 2904 syscallarg(const char *) path; 2905 syscallarg(int) pad; 2906 syscallarg(off_t) length; 2907 } */ *uap = v; 2908 struct proc *p = l->l_proc; 2909 struct vnode *vp; 2910 struct mount *mp; 2911 struct vattr vattr; 2912 int error; 2913 struct nameidata nd; 2914 2915 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2916 if ((error = namei(&nd)) != 0) 2917 return (error); 2918 vp = nd.ni_vp; 2919 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2920 vrele(vp); 2921 return (error); 2922 } 2923 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2924 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2925 if (vp->v_type == VDIR) 2926 error = EISDIR; 2927 else if ((error = vn_writechk(vp)) == 0 && 2928 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2929 VATTR_NULL(&vattr); 2930 vattr.va_size = SCARG(uap, length); 2931 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2932 } 2933 vput(vp); 2934 vn_finished_write(mp, 0); 2935 return (error); 2936 } 2937 2938 /* 2939 * Truncate a file given a file descriptor. 2940 */ 2941 /* ARGSUSED */ 2942 int 2943 sys_ftruncate(l, v, retval) 2944 struct lwp *l; 2945 void *v; 2946 register_t *retval; 2947 { 2948 struct sys_ftruncate_args /* { 2949 syscallarg(int) fd; 2950 syscallarg(int) pad; 2951 syscallarg(off_t) length; 2952 } */ *uap = v; 2953 struct proc *p = l->l_proc; 2954 struct mount *mp; 2955 struct vattr vattr; 2956 struct vnode *vp; 2957 struct file *fp; 2958 int error; 2959 2960 /* getvnode() will use the descriptor for us */ 2961 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2962 return (error); 2963 if ((fp->f_flag & FWRITE) == 0) { 2964 error = EINVAL; 2965 goto out; 2966 } 2967 vp = (struct vnode *)fp->f_data; 2968 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2969 FILE_UNUSE(fp, p); 2970 return (error); 2971 } 2972 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2973 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2974 if (vp->v_type == VDIR) 2975 error = EISDIR; 2976 else if ((error = vn_writechk(vp)) == 0) { 2977 VATTR_NULL(&vattr); 2978 vattr.va_size = SCARG(uap, length); 2979 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2980 } 2981 VOP_UNLOCK(vp, 0); 2982 vn_finished_write(mp, 0); 2983 out: 2984 FILE_UNUSE(fp, p); 2985 return (error); 2986 } 2987 2988 /* 2989 * Sync an open file. 2990 */ 2991 /* ARGSUSED */ 2992 int 2993 sys_fsync(l, v, retval) 2994 struct lwp *l; 2995 void *v; 2996 register_t *retval; 2997 { 2998 struct sys_fsync_args /* { 2999 syscallarg(int) fd; 3000 } */ *uap = v; 3001 struct proc *p = l->l_proc; 3002 struct vnode *vp; 3003 struct mount *mp; 3004 struct file *fp; 3005 int error; 3006 3007 /* getvnode() will use the descriptor for us */ 3008 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3009 return (error); 3010 vp = (struct vnode *)fp->f_data; 3011 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3012 FILE_UNUSE(fp, p); 3013 return (error); 3014 } 3015 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3016 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p); 3017 if (error == 0 && bioops.io_fsync != NULL && 3018 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3019 (*bioops.io_fsync)(vp); 3020 VOP_UNLOCK(vp, 0); 3021 vn_finished_write(mp, 0); 3022 FILE_UNUSE(fp, p); 3023 return (error); 3024 } 3025 3026 /* 3027 * Sync a range of file data. API modeled after that found in AIX. 3028 * 3029 * FDATASYNC indicates that we need only save enough metadata to be able 3030 * to re-read the written data. Note we duplicate AIX's requirement that 3031 * the file be open for writing. 3032 */ 3033 /* ARGSUSED */ 3034 int 3035 sys_fsync_range(l, v, retval) 3036 struct lwp *l; 3037 void *v; 3038 register_t *retval; 3039 { 3040 struct sys_fsync_range_args /* { 3041 syscallarg(int) fd; 3042 syscallarg(int) flags; 3043 syscallarg(off_t) start; 3044 syscallarg(int) length; 3045 } */ *uap = v; 3046 struct proc *p = l->l_proc; 3047 struct vnode *vp; 3048 struct file *fp; 3049 int flags, nflags; 3050 off_t s, e, len; 3051 int error; 3052 3053 /* getvnode() will use the descriptor for us */ 3054 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3055 return (error); 3056 3057 if ((fp->f_flag & FWRITE) == 0) { 3058 FILE_UNUSE(fp, p); 3059 return (EBADF); 3060 } 3061 3062 flags = SCARG(uap, flags); 3063 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3064 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3065 return (EINVAL); 3066 } 3067 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3068 if (flags & FDATASYNC) 3069 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3070 else 3071 nflags = FSYNC_WAIT; 3072 3073 len = SCARG(uap, length); 3074 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3075 if (len) { 3076 s = SCARG(uap, start); 3077 e = s + len; 3078 if (e < s) { 3079 FILE_UNUSE(fp, p); 3080 return (EINVAL); 3081 } 3082 } else { 3083 e = 0; 3084 s = 0; 3085 } 3086 3087 vp = (struct vnode *)fp->f_data; 3088 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3089 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p); 3090 3091 if (error == 0 && bioops.io_fsync != NULL && 3092 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3093 (*bioops.io_fsync)(vp); 3094 3095 VOP_UNLOCK(vp, 0); 3096 FILE_UNUSE(fp, p); 3097 return (error); 3098 } 3099 3100 /* 3101 * Sync the data of an open file. 3102 */ 3103 /* ARGSUSED */ 3104 int 3105 sys_fdatasync(l, v, retval) 3106 struct lwp *l; 3107 void *v; 3108 register_t *retval; 3109 { 3110 struct sys_fdatasync_args /* { 3111 syscallarg(int) fd; 3112 } */ *uap = v; 3113 struct proc *p = l->l_proc; 3114 struct vnode *vp; 3115 struct file *fp; 3116 int error; 3117 3118 /* getvnode() will use the descriptor for us */ 3119 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3120 return (error); 3121 if ((fp->f_flag & FWRITE) == 0) { 3122 FILE_UNUSE(fp, p); 3123 return (EBADF); 3124 } 3125 vp = (struct vnode *)fp->f_data; 3126 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3127 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p); 3128 VOP_UNLOCK(vp, 0); 3129 FILE_UNUSE(fp, p); 3130 return (error); 3131 } 3132 3133 /* 3134 * Rename files, (standard) BSD semantics frontend. 3135 */ 3136 /* ARGSUSED */ 3137 int 3138 sys_rename(l, v, retval) 3139 struct lwp *l; 3140 void *v; 3141 register_t *retval; 3142 { 3143 struct sys_rename_args /* { 3144 syscallarg(const char *) from; 3145 syscallarg(const char *) to; 3146 } */ *uap = v; 3147 struct proc *p = l->l_proc; 3148 3149 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 3150 } 3151 3152 /* 3153 * Rename files, POSIX semantics frontend. 3154 */ 3155 /* ARGSUSED */ 3156 int 3157 sys___posix_rename(l, v, retval) 3158 struct lwp *l; 3159 void *v; 3160 register_t *retval; 3161 { 3162 struct sys___posix_rename_args /* { 3163 syscallarg(const char *) from; 3164 syscallarg(const char *) to; 3165 } */ *uap = v; 3166 struct proc *p = l->l_proc; 3167 3168 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 3169 } 3170 3171 /* 3172 * Rename files. Source and destination must either both be directories, 3173 * or both not be directories. If target is a directory, it must be empty. 3174 * If `from' and `to' refer to the same object, the value of the `retain' 3175 * argument is used to determine whether `from' will be 3176 * 3177 * (retain == 0) deleted unless `from' and `to' refer to the same 3178 * object in the file system's name space (BSD). 3179 * (retain == 1) always retained (POSIX). 3180 */ 3181 static int 3182 rename_files(from, to, p, retain) 3183 const char *from, *to; 3184 struct proc *p; 3185 int retain; 3186 { 3187 struct mount *mp = NULL; 3188 struct vnode *tvp, *fvp, *tdvp; 3189 struct nameidata fromnd, tond; 3190 int error; 3191 3192 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3193 from, p); 3194 if ((error = namei(&fromnd)) != 0) 3195 return (error); 3196 fvp = fromnd.ni_vp; 3197 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3198 if (error != 0) { 3199 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3200 vrele(fromnd.ni_dvp); 3201 vrele(fvp); 3202 if (fromnd.ni_startdir) 3203 vrele(fromnd.ni_startdir); 3204 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3205 return (error); 3206 } 3207 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3208 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p); 3209 if ((error = namei(&tond)) != 0) { 3210 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3211 vrele(fromnd.ni_dvp); 3212 vrele(fvp); 3213 goto out1; 3214 } 3215 tdvp = tond.ni_dvp; 3216 tvp = tond.ni_vp; 3217 3218 if (tvp != NULL) { 3219 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3220 error = ENOTDIR; 3221 goto out; 3222 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3223 error = EISDIR; 3224 goto out; 3225 } 3226 } 3227 3228 if (fvp == tdvp) 3229 error = EINVAL; 3230 3231 /* 3232 * Source and destination refer to the same object. 3233 */ 3234 if (fvp == tvp) { 3235 if (retain) 3236 error = -1; 3237 else if (fromnd.ni_dvp == tdvp && 3238 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3239 !memcmp(fromnd.ni_cnd.cn_nameptr, 3240 tond.ni_cnd.cn_nameptr, 3241 fromnd.ni_cnd.cn_namelen)) 3242 error = -1; 3243 } 3244 3245 out: 3246 if (!error) { 3247 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 3248 if (fromnd.ni_dvp != tdvp) 3249 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3250 if (tvp) { 3251 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 3252 } 3253 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3254 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3255 } else { 3256 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3257 if (tdvp == tvp) 3258 vrele(tdvp); 3259 else 3260 vput(tdvp); 3261 if (tvp) 3262 vput(tvp); 3263 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3264 vrele(fromnd.ni_dvp); 3265 vrele(fvp); 3266 } 3267 vrele(tond.ni_startdir); 3268 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3269 out1: 3270 vn_finished_write(mp, 0); 3271 if (fromnd.ni_startdir) 3272 vrele(fromnd.ni_startdir); 3273 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3274 return (error == -1 ? 0 : error); 3275 } 3276 3277 /* 3278 * Make a directory file. 3279 */ 3280 /* ARGSUSED */ 3281 int 3282 sys_mkdir(l, v, retval) 3283 struct lwp *l; 3284 void *v; 3285 register_t *retval; 3286 { 3287 struct sys_mkdir_args /* { 3288 syscallarg(const char *) path; 3289 syscallarg(int) mode; 3290 } */ *uap = v; 3291 struct proc *p = l->l_proc; 3292 struct mount *mp; 3293 struct vnode *vp; 3294 struct vattr vattr; 3295 int error; 3296 struct nameidata nd; 3297 3298 restart: 3299 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3300 SCARG(uap, path), p); 3301 if ((error = namei(&nd)) != 0) 3302 return (error); 3303 vp = nd.ni_vp; 3304 if (vp != NULL) { 3305 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3306 if (nd.ni_dvp == vp) 3307 vrele(nd.ni_dvp); 3308 else 3309 vput(nd.ni_dvp); 3310 vrele(vp); 3311 return (EEXIST); 3312 } 3313 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3314 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3315 if (nd.ni_dvp == vp) 3316 vrele(nd.ni_dvp); 3317 else 3318 vput(nd.ni_dvp); 3319 if ((error = vn_start_write(NULL, &mp, 3320 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3321 return (error); 3322 goto restart; 3323 } 3324 VATTR_NULL(&vattr); 3325 vattr.va_type = VDIR; 3326 vattr.va_mode = 3327 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3328 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3329 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3330 if (!error) 3331 vput(nd.ni_vp); 3332 vn_finished_write(mp, 0); 3333 return (error); 3334 } 3335 3336 /* 3337 * Remove a directory file. 3338 */ 3339 /* ARGSUSED */ 3340 int 3341 sys_rmdir(l, v, retval) 3342 struct lwp *l; 3343 void *v; 3344 register_t *retval; 3345 { 3346 struct sys_rmdir_args /* { 3347 syscallarg(const char *) path; 3348 } */ *uap = v; 3349 struct proc *p = l->l_proc; 3350 struct mount *mp; 3351 struct vnode *vp; 3352 int error; 3353 struct nameidata nd; 3354 3355 restart: 3356 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3357 SCARG(uap, path), p); 3358 if ((error = namei(&nd)) != 0) 3359 return (error); 3360 vp = nd.ni_vp; 3361 if (vp->v_type != VDIR) { 3362 error = ENOTDIR; 3363 goto out; 3364 } 3365 /* 3366 * No rmdir "." please. 3367 */ 3368 if (nd.ni_dvp == vp) { 3369 error = EINVAL; 3370 goto out; 3371 } 3372 /* 3373 * The root of a mounted filesystem cannot be deleted. 3374 */ 3375 if (vp->v_flag & VROOT) { 3376 error = EBUSY; 3377 goto out; 3378 } 3379 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3380 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3381 if (nd.ni_dvp == vp) 3382 vrele(nd.ni_dvp); 3383 else 3384 vput(nd.ni_dvp); 3385 vput(vp); 3386 if ((error = vn_start_write(NULL, &mp, 3387 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3388 return (error); 3389 goto restart; 3390 } 3391 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3392 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3393 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3394 vn_finished_write(mp, 0); 3395 return (error); 3396 3397 out: 3398 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3399 if (nd.ni_dvp == vp) 3400 vrele(nd.ni_dvp); 3401 else 3402 vput(nd.ni_dvp); 3403 vput(vp); 3404 return (error); 3405 } 3406 3407 /* 3408 * Read a block of directory entries in a file system independent format. 3409 */ 3410 int 3411 sys_getdents(l, v, retval) 3412 struct lwp *l; 3413 void *v; 3414 register_t *retval; 3415 { 3416 struct sys_getdents_args /* { 3417 syscallarg(int) fd; 3418 syscallarg(char *) buf; 3419 syscallarg(size_t) count; 3420 } */ *uap = v; 3421 struct proc *p = l->l_proc; 3422 struct file *fp; 3423 int error, done; 3424 3425 /* getvnode() will use the descriptor for us */ 3426 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3427 return (error); 3428 if ((fp->f_flag & FREAD) == 0) { 3429 error = EBADF; 3430 goto out; 3431 } 3432 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3433 SCARG(uap, count), &done, p, 0, 0); 3434 #ifdef KTRACE 3435 if (!error && KTRPOINT(p, KTR_GENIO)) { 3436 struct iovec iov; 3437 iov.iov_base = SCARG(uap, buf); 3438 iov.iov_len = done; 3439 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3440 } 3441 #endif 3442 *retval = done; 3443 out: 3444 FILE_UNUSE(fp, p); 3445 return (error); 3446 } 3447 3448 /* 3449 * Set the mode mask for creation of filesystem nodes. 3450 */ 3451 int 3452 sys_umask(l, v, retval) 3453 struct lwp *l; 3454 void *v; 3455 register_t *retval; 3456 { 3457 struct sys_umask_args /* { 3458 syscallarg(mode_t) newmask; 3459 } */ *uap = v; 3460 struct proc *p = l->l_proc; 3461 struct cwdinfo *cwdi; 3462 3463 cwdi = p->p_cwdi; 3464 *retval = cwdi->cwdi_cmask; 3465 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3466 return (0); 3467 } 3468 3469 /* 3470 * Void all references to file by ripping underlying filesystem 3471 * away from vnode. 3472 */ 3473 /* ARGSUSED */ 3474 int 3475 sys_revoke(l, v, retval) 3476 struct lwp *l; 3477 void *v; 3478 register_t *retval; 3479 { 3480 struct sys_revoke_args /* { 3481 syscallarg(const char *) path; 3482 } */ *uap = v; 3483 struct proc *p = l->l_proc; 3484 struct mount *mp; 3485 struct vnode *vp; 3486 struct vattr vattr; 3487 int error; 3488 struct nameidata nd; 3489 3490 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3491 if ((error = namei(&nd)) != 0) 3492 return (error); 3493 vp = nd.ni_vp; 3494 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 3495 goto out; 3496 if (p->p_ucred->cr_uid != vattr.va_uid && 3497 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3498 goto out; 3499 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3500 goto out; 3501 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3502 VOP_REVOKE(vp, REVOKEALL); 3503 vn_finished_write(mp, 0); 3504 out: 3505 vrele(vp); 3506 return (error); 3507 } 3508 3509 /* 3510 * Convert a user file descriptor to a kernel file entry. 3511 */ 3512 int 3513 getvnode(fdp, fd, fpp) 3514 struct filedesc *fdp; 3515 int fd; 3516 struct file **fpp; 3517 { 3518 struct vnode *vp; 3519 struct file *fp; 3520 3521 if ((fp = fd_getfile(fdp, fd)) == NULL) 3522 return (EBADF); 3523 3524 FILE_USE(fp); 3525 3526 if (fp->f_type != DTYPE_VNODE) { 3527 FILE_UNUSE(fp, NULL); 3528 return (EINVAL); 3529 } 3530 3531 vp = (struct vnode *)fp->f_data; 3532 if (vp->v_type == VBAD) { 3533 FILE_UNUSE(fp, NULL); 3534 return (EBADF); 3535 } 3536 3537 *fpp = fp; 3538 return (0); 3539 } 3540
Indexes created Thu Sep 25 16:09:42 GMT 2025