vfs_syscalls.c revision 1.207
1 /* $NetBSD: vfs_syscalls.c,v 1.207 2004/05/02 11:13:29 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 1.207 2004/05/02 11:13:29 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 (void)vfs_busy(mp, LK_NOWAIT, 0); 300 mp->mnt_op = vfs; 301 vfs->vfs_refcount++; 302 mp->mnt_vnodecovered = vp; 303 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 304 mp->mnt_unmounter = NULL; 305 306 /* 307 * The underlying file system may refuse the mount for 308 * various reasons. Allow the user to force it to happen. 309 */ 310 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 311 update: 312 /* 313 * Set the mount level flags. 314 */ 315 if (SCARG(uap, flags) & MNT_RDONLY) 316 mp->mnt_flag |= MNT_RDONLY; 317 else if (mp->mnt_flag & MNT_RDONLY) 318 mp->mnt_iflag |= IMNT_WANTRDWR; 319 mp->mnt_flag &= 320 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 321 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 322 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 323 mp->mnt_flag |= SCARG(uap, flags) & 324 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 325 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 326 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 327 MNT_IGNORE); 328 /* 329 * Mount the filesystem. 330 */ 331 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 332 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 333 if (mp->mnt_iflag & IMNT_WANTRDWR) 334 mp->mnt_flag &= ~MNT_RDONLY; 335 if (error || (mp->mnt_flag & MNT_GETARGS)) 336 mp->mnt_flag = flag; 337 mp->mnt_flag &=~ 338 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 339 mp->mnt_iflag &=~ IMNT_WANTRDWR; 340 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 341 if (mp->mnt_syncer == NULL) 342 error = vfs_allocate_syncvnode(mp); 343 } else { 344 if (mp->mnt_syncer != NULL) 345 vfs_deallocate_syncvnode(mp); 346 } 347 vfs_unbusy(mp); 348 VOP_UNLOCK(vp, 0); 349 vrele(vp); 350 return (error); 351 } 352 /* 353 * Put the new filesystem on the mount list after root. 354 */ 355 cache_purge(vp); 356 if (!error) { 357 mp->mnt_flag &=~ 358 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 359 mp->mnt_iflag &=~ IMNT_WANTRDWR; 360 vp->v_mountedhere = mp; 361 simple_lock(&mountlist_slock); 362 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 363 simple_unlock(&mountlist_slock); 364 checkdirs(vp); 365 VOP_UNLOCK(vp, 0); 366 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 367 error = vfs_allocate_syncvnode(mp); 368 vfs_unbusy(mp); 369 (void) VFS_STATVFS(mp, &mp->mnt_stat, p); 370 if ((error = VFS_START(mp, 0, p))) 371 vrele(vp); 372 } else { 373 vp->v_mountedhere = (struct mount *)0; 374 vfs->vfs_refcount--; 375 vfs_unbusy(mp); 376 free(mp, M_MOUNT); 377 vput(vp); 378 } 379 return (error); 380 } 381 382 /* 383 * Scan all active processes to see if any of them have a current 384 * or root directory onto which the new filesystem has just been 385 * mounted. If so, replace them with the new mount point. 386 */ 387 void 388 checkdirs(olddp) 389 struct vnode *olddp; 390 { 391 struct cwdinfo *cwdi; 392 struct vnode *newdp; 393 struct proc *p; 394 395 if (olddp->v_usecount == 1) 396 return; 397 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 398 panic("mount: lost mount"); 399 proclist_lock_read(); 400 LIST_FOREACH(p, &allproc, p_list) { 401 cwdi = p->p_cwdi; 402 if (cwdi->cwdi_cdir == olddp) { 403 vrele(cwdi->cwdi_cdir); 404 VREF(newdp); 405 cwdi->cwdi_cdir = newdp; 406 } 407 if (cwdi->cwdi_rdir == olddp) { 408 vrele(cwdi->cwdi_rdir); 409 VREF(newdp); 410 cwdi->cwdi_rdir = newdp; 411 } 412 } 413 proclist_unlock_read(); 414 if (rootvnode == olddp) { 415 vrele(rootvnode); 416 VREF(newdp); 417 rootvnode = newdp; 418 } 419 vput(newdp); 420 } 421 422 /* 423 * Unmount a file system. 424 * 425 * Note: unmount takes a path to the vnode mounted on as argument, 426 * not special file (as before). 427 */ 428 /* ARGSUSED */ 429 int 430 sys_unmount(l, v, retval) 431 struct lwp *l; 432 void *v; 433 register_t *retval; 434 { 435 struct sys_unmount_args /* { 436 syscallarg(const char *) path; 437 syscallarg(int) flags; 438 } */ *uap = v; 439 struct proc *p = l->l_proc; 440 struct vnode *vp; 441 struct mount *mp; 442 int error; 443 struct nameidata nd; 444 445 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 446 SCARG(uap, path), p); 447 if ((error = namei(&nd)) != 0) 448 return (error); 449 vp = nd.ni_vp; 450 mp = vp->v_mount; 451 452 /* 453 * Only root, or the user that did the original mount is 454 * permitted to unmount this filesystem. 455 */ 456 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 457 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 458 vput(vp); 459 return (error); 460 } 461 462 /* 463 * Don't allow unmounting the root file system. 464 */ 465 if (mp->mnt_flag & MNT_ROOTFS) { 466 vput(vp); 467 return (EINVAL); 468 } 469 470 /* 471 * Must be the root of the filesystem 472 */ 473 if ((vp->v_flag & VROOT) == 0) { 474 vput(vp); 475 return (EINVAL); 476 } 477 vput(vp); 478 479 /* 480 * XXX Freeze syncer. Must do this before locking the 481 * mount point. See dounmount() for details. 482 */ 483 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 484 485 if (vfs_busy(mp, 0, 0)) { 486 lockmgr(&syncer_lock, LK_RELEASE, NULL); 487 return (EBUSY); 488 } 489 490 return (dounmount(mp, SCARG(uap, flags), p)); 491 } 492 493 /* 494 * Do the actual file system unmount. File system is assumed to have been 495 * marked busy by the caller. 496 */ 497 int 498 dounmount(mp, flags, p) 499 struct mount *mp; 500 int flags; 501 struct proc *p; 502 { 503 struct vnode *coveredvp; 504 int error; 505 int async; 506 int used_syncer; 507 508 simple_lock(&mountlist_slock); 509 vfs_unbusy(mp); 510 used_syncer = (mp->mnt_syncer != NULL); 511 512 /* 513 * XXX Syncer must be frozen when we get here. This should really 514 * be done on a per-mountpoint basis, but especially the softdep 515 * code possibly called from the syncer doens't exactly work on a 516 * per-mountpoint basis, so the softdep code would become a maze 517 * of vfs_busy() calls. 518 * 519 * The caller of dounmount() must acquire syncer_lock because 520 * the syncer itself acquires locks in syncer_lock -> vfs_busy 521 * order, and we must preserve that order to avoid deadlock. 522 * 523 * So, if the file system did not use the syncer, now is 524 * the time to release the syncer_lock. 525 */ 526 if (used_syncer == 0) 527 lockmgr(&syncer_lock, LK_RELEASE, NULL); 528 529 mp->mnt_iflag |= IMNT_UNMOUNT; 530 mp->mnt_unmounter = p; 531 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 532 vn_start_write(NULL, &mp, V_WAIT); 533 534 if (mp->mnt_flag & MNT_EXPUBLIC) 535 vfs_setpublicfs(NULL, NULL, NULL); 536 async = mp->mnt_flag & MNT_ASYNC; 537 mp->mnt_flag &= ~MNT_ASYNC; 538 cache_purgevfs(mp); /* remove cache entries for this file sys */ 539 if (mp->mnt_syncer != NULL) 540 vfs_deallocate_syncvnode(mp); 541 if (((mp->mnt_flag & MNT_RDONLY) || 542 #if NFSS > 0 543 (error = fss_umount_hook(mp, (flags & MNT_FORCE))) == 0 || 544 #endif 545 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) || 546 (flags & MNT_FORCE)) 547 error = VFS_UNMOUNT(mp, flags, p); 548 vn_finished_write(mp, 0); 549 simple_lock(&mountlist_slock); 550 if (error) { 551 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 552 (void) vfs_allocate_syncvnode(mp); 553 mp->mnt_iflag &= ~IMNT_UNMOUNT; 554 mp->mnt_unmounter = NULL; 555 mp->mnt_flag |= async; 556 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 557 &mountlist_slock); 558 if (used_syncer) 559 lockmgr(&syncer_lock, LK_RELEASE, NULL); 560 while (mp->mnt_wcnt > 0) { 561 wakeup(mp); 562 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 0); 563 } 564 return (error); 565 } 566 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 567 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 568 coveredvp->v_mountedhere = NULL; 569 vrele(coveredvp); 570 } 571 mp->mnt_op->vfs_refcount--; 572 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 573 panic("unmount: dangling vnode"); 574 mp->mnt_iflag |= IMNT_GONE; 575 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 576 if (used_syncer) 577 lockmgr(&syncer_lock, LK_RELEASE, NULL); 578 while(mp->mnt_wcnt > 0) { 579 wakeup(mp); 580 tsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0); 581 } 582 free(mp, M_MOUNT); 583 return (0); 584 } 585 586 /* 587 * Sync each mounted filesystem. 588 */ 589 #ifdef DEBUG 590 int syncprt = 0; 591 struct ctldebug debug0 = { "syncprt", &syncprt }; 592 #endif 593 594 /* ARGSUSED */ 595 int 596 sys_sync(l, v, retval) 597 struct lwp *l; 598 void *v; 599 register_t *retval; 600 { 601 struct mount *mp, *nmp; 602 int asyncflag; 603 struct proc *p = l == NULL ? &proc0 : l->l_proc; 604 605 simple_lock(&mountlist_slock); 606 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 607 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 608 nmp = mp->mnt_list.cqe_prev; 609 continue; 610 } 611 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 612 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 613 asyncflag = mp->mnt_flag & MNT_ASYNC; 614 mp->mnt_flag &= ~MNT_ASYNC; 615 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p); 616 if (asyncflag) 617 mp->mnt_flag |= MNT_ASYNC; 618 vn_finished_write(mp, 0); 619 } 620 simple_lock(&mountlist_slock); 621 nmp = mp->mnt_list.cqe_prev; 622 vfs_unbusy(mp); 623 624 } 625 simple_unlock(&mountlist_slock); 626 #ifdef DEBUG 627 if (syncprt) 628 vfs_bufstats(); 629 #endif /* DEBUG */ 630 return (0); 631 } 632 633 /* 634 * Change filesystem quotas. 635 */ 636 /* ARGSUSED */ 637 int 638 sys_quotactl(l, v, retval) 639 struct lwp *l; 640 void *v; 641 register_t *retval; 642 { 643 struct sys_quotactl_args /* { 644 syscallarg(const char *) path; 645 syscallarg(int) cmd; 646 syscallarg(int) uid; 647 syscallarg(caddr_t) arg; 648 } */ *uap = v; 649 struct proc *p = l->l_proc; 650 struct mount *mp; 651 int error; 652 struct nameidata nd; 653 654 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 655 if ((error = namei(&nd)) != 0) 656 return (error); 657 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 658 vrele(nd.ni_vp); 659 if (error) 660 return (error); 661 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 662 SCARG(uap, arg), p); 663 vn_finished_write(mp, 0); 664 return (error); 665 } 666 667 int 668 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags, 669 int root) 670 { 671 struct cwdinfo *cwdi = p->p_cwdi; 672 int error = 0; 673 674 /* 675 * If MNT_NOWAIT or MNT_LAZY is specified, do not 676 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 677 * overrides MNT_NOWAIT. 678 */ 679 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 680 (flags != MNT_WAIT && flags != 0)) { 681 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 682 goto done; 683 } 684 685 if ((error = VFS_STATVFS(mp, sp, p)) != 0) { 686 return error; 687 } 688 689 if (cwdi->cwdi_rdir == NULL) 690 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 691 done: 692 if (cwdi->cwdi_rdir != NULL) { 693 size_t len; 694 char *bp; 695 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 696 if (!path) 697 return ENOMEM; 698 699 bp = path + MAXPATHLEN; 700 *--bp = '\0'; 701 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 702 MAXPATHLEN / 2, 0, p); 703 if (error) { 704 free(path, M_TEMP); 705 return error; 706 } 707 len = strlen(bp); 708 /* 709 * for mount points that are below our root, we can see 710 * them, so we fix up the pathname and return them. The 711 * rest we cannot see, so we don't allow viewing the 712 * data. 713 */ 714 if (strncmp(bp, sp->f_mntonname, len) == 0) { 715 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 716 sizeof(sp->f_mntonname)); 717 if (sp->f_mntonname[0] == '\0') 718 (void)strlcpy(sp->f_mntonname, "/", 719 sizeof(sp->f_mntonname)); 720 } else { 721 if (root) 722 (void)strlcpy(sp->f_mntonname, "/", 723 sizeof(sp->f_mntonname)); 724 else 725 error = EPERM; 726 } 727 free(path, M_TEMP); 728 } 729 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 730 return error; 731 } 732 733 /* 734 * Get filesystem statistics. 735 */ 736 /* ARGSUSED */ 737 int 738 sys_statvfs1(l, v, retval) 739 struct lwp *l; 740 void *v; 741 register_t *retval; 742 { 743 struct sys_statvfs1_args /* { 744 syscallarg(const char *) path; 745 syscallarg(struct statvfs *) buf; 746 syscallarg(int) flags; 747 } */ *uap = v; 748 struct proc *p = l->l_proc; 749 struct mount *mp; 750 struct statvfs sbuf; 751 int error; 752 struct nameidata nd; 753 754 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 755 if ((error = namei(&nd)) != 0) 756 return error; 757 mp = nd.ni_vp->v_mount; 758 vrele(nd.ni_vp); 759 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 760 return error; 761 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 762 } 763 764 /* 765 * Get filesystem statistics. 766 */ 767 /* ARGSUSED */ 768 int 769 sys_fstatvfs1(l, v, retval) 770 struct lwp *l; 771 void *v; 772 register_t *retval; 773 { 774 struct sys_fstatvfs1_args /* { 775 syscallarg(int) fd; 776 syscallarg(struct statvfs *) buf; 777 syscallarg(int) flags; 778 } */ *uap = v; 779 struct proc *p = l->l_proc; 780 struct file *fp; 781 struct mount *mp; 782 struct statvfs sbuf; 783 int error; 784 785 /* getvnode() will use the descriptor for us */ 786 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 787 return (error); 788 mp = ((struct vnode *)fp->f_data)->v_mount; 789 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 790 goto out; 791 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 792 out: 793 FILE_UNUSE(fp, p); 794 return error; 795 } 796 797 798 /* 799 * Get statistics on all filesystems. 800 */ 801 int 802 sys_getvfsstat(l, v, retval) 803 struct lwp *l; 804 void *v; 805 register_t *retval; 806 { 807 struct sys_getvfsstat_args /* { 808 syscallarg(struct statvfs *) buf; 809 syscallarg(size_t) bufsize; 810 syscallarg(int) flags; 811 } */ *uap = v; 812 int root = 0; 813 struct proc *p = l->l_proc; 814 struct mount *mp, *nmp; 815 struct statvfs sbuf; 816 struct statvfs *sfsp; 817 size_t count, maxcount; 818 int error = 0; 819 820 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 821 sfsp = SCARG(uap, buf); 822 simple_lock(&mountlist_slock); 823 count = 0; 824 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 825 mp = nmp) { 826 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 827 nmp = CIRCLEQ_NEXT(mp, mnt_list); 828 continue; 829 } 830 if (sfsp && count < maxcount) { 831 error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0); 832 if (error) { 833 simple_lock(&mountlist_slock); 834 nmp = CIRCLEQ_NEXT(mp, mnt_list); 835 vfs_unbusy(mp); 836 continue; 837 } 838 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 839 if (error) { 840 vfs_unbusy(mp); 841 return (error); 842 } 843 sfsp++; 844 root |= strcmp(sbuf.f_mntonname, "/") == 0; 845 } 846 count++; 847 simple_lock(&mountlist_slock); 848 nmp = CIRCLEQ_NEXT(mp, mnt_list); 849 vfs_unbusy(mp); 850 } 851 simple_unlock(&mountlist_slock); 852 if (root == 0 && p->p_cwdi->cwdi_rdir) { 853 /* 854 * fake a root entry 855 */ 856 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p, 857 SCARG(uap, flags), 1)) != 0) 858 return error; 859 if (sfsp) 860 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 861 count++; 862 } 863 if (sfsp && count > maxcount) 864 *retval = maxcount; 865 else 866 *retval = count; 867 return error; 868 } 869 870 /* 871 * Change current working directory to a given file descriptor. 872 */ 873 /* ARGSUSED */ 874 int 875 sys_fchdir(l, v, retval) 876 struct lwp *l; 877 void *v; 878 register_t *retval; 879 { 880 struct sys_fchdir_args /* { 881 syscallarg(int) fd; 882 } */ *uap = v; 883 struct proc *p = l->l_proc; 884 struct filedesc *fdp = p->p_fd; 885 struct cwdinfo *cwdi = p->p_cwdi; 886 struct vnode *vp, *tdp; 887 struct mount *mp; 888 struct file *fp; 889 int error; 890 891 /* getvnode() will use the descriptor for us */ 892 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 893 return (error); 894 vp = (struct vnode *)fp->f_data; 895 896 VREF(vp); 897 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 898 if (vp->v_type != VDIR) 899 error = ENOTDIR; 900 else 901 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 902 while (!error && (mp = vp->v_mountedhere) != NULL) { 903 if (vfs_busy(mp, 0, 0)) 904 continue; 905 error = VFS_ROOT(mp, &tdp); 906 vfs_unbusy(mp); 907 if (error) 908 break; 909 vput(vp); 910 vp = tdp; 911 } 912 if (error) { 913 vput(vp); 914 goto out; 915 } 916 VOP_UNLOCK(vp, 0); 917 918 /* 919 * Disallow changing to a directory not under the process's 920 * current root directory (if there is one). 921 */ 922 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) { 923 vrele(vp); 924 error = EPERM; /* operation not permitted */ 925 goto out; 926 } 927 928 vrele(cwdi->cwdi_cdir); 929 cwdi->cwdi_cdir = vp; 930 out: 931 FILE_UNUSE(fp, p); 932 return (error); 933 } 934 935 /* 936 * Change this process's notion of the root directory to a given file descriptor. 937 */ 938 939 int 940 sys_fchroot(l, v, retval) 941 struct lwp *l; 942 void *v; 943 register_t *retval; 944 { 945 struct sys_fchroot_args *uap = v; 946 struct proc *p = l->l_proc; 947 struct filedesc *fdp = p->p_fd; 948 struct cwdinfo *cwdi = p->p_cwdi; 949 struct vnode *vp; 950 struct file *fp; 951 int error; 952 953 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 954 return error; 955 /* getvnode() will use the descriptor for us */ 956 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 957 return error; 958 vp = (struct vnode *) fp->f_data; 959 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 960 if (vp->v_type != VDIR) 961 error = ENOTDIR; 962 else 963 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 964 VOP_UNLOCK(vp, 0); 965 if (error) 966 goto out; 967 VREF(vp); 968 969 /* 970 * Prevent escaping from chroot by putting the root under 971 * the working directory. Silently chdir to / if we aren't 972 * already there. 973 */ 974 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 975 /* 976 * XXX would be more failsafe to change directory to a 977 * deadfs node here instead 978 */ 979 vrele(cwdi->cwdi_cdir); 980 VREF(vp); 981 cwdi->cwdi_cdir = vp; 982 } 983 984 if (cwdi->cwdi_rdir != NULL) 985 vrele(cwdi->cwdi_rdir); 986 cwdi->cwdi_rdir = vp; 987 out: 988 FILE_UNUSE(fp, p); 989 return (error); 990 } 991 992 993 994 /* 995 * Change current working directory (``.''). 996 */ 997 /* ARGSUSED */ 998 int 999 sys_chdir(l, v, retval) 1000 struct lwp *l; 1001 void *v; 1002 register_t *retval; 1003 { 1004 struct sys_chdir_args /* { 1005 syscallarg(const char *) path; 1006 } */ *uap = v; 1007 struct proc *p = l->l_proc; 1008 struct cwdinfo *cwdi = p->p_cwdi; 1009 int error; 1010 struct nameidata nd; 1011 1012 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1013 SCARG(uap, path), p); 1014 if ((error = change_dir(&nd, p)) != 0) 1015 return (error); 1016 vrele(cwdi->cwdi_cdir); 1017 cwdi->cwdi_cdir = nd.ni_vp; 1018 return (0); 1019 } 1020 1021 /* 1022 * Change notion of root (``/'') directory. 1023 */ 1024 /* ARGSUSED */ 1025 int 1026 sys_chroot(l, v, retval) 1027 struct lwp *l; 1028 void *v; 1029 register_t *retval; 1030 { 1031 struct sys_chroot_args /* { 1032 syscallarg(const char *) path; 1033 } */ *uap = v; 1034 struct proc *p = l->l_proc; 1035 struct cwdinfo *cwdi = p->p_cwdi; 1036 struct vnode *vp; 1037 int error; 1038 struct nameidata nd; 1039 1040 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1041 return (error); 1042 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1043 SCARG(uap, path), p); 1044 if ((error = change_dir(&nd, p)) != 0) 1045 return (error); 1046 if (cwdi->cwdi_rdir != NULL) 1047 vrele(cwdi->cwdi_rdir); 1048 vp = nd.ni_vp; 1049 cwdi->cwdi_rdir = vp; 1050 1051 /* 1052 * Prevent escaping from chroot by putting the root under 1053 * the working directory. Silently chdir to / if we aren't 1054 * already there. 1055 */ 1056 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 1057 /* 1058 * XXX would be more failsafe to change directory to a 1059 * deadfs node here instead 1060 */ 1061 vrele(cwdi->cwdi_cdir); 1062 VREF(vp); 1063 cwdi->cwdi_cdir = vp; 1064 } 1065 1066 return (0); 1067 } 1068 1069 /* 1070 * Common routine for chroot and chdir. 1071 */ 1072 static int 1073 change_dir(ndp, p) 1074 struct nameidata *ndp; 1075 struct proc *p; 1076 { 1077 struct vnode *vp; 1078 int error; 1079 1080 if ((error = namei(ndp)) != 0) 1081 return (error); 1082 vp = ndp->ni_vp; 1083 if (vp->v_type != VDIR) 1084 error = ENOTDIR; 1085 else 1086 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 1087 1088 if (error) 1089 vput(vp); 1090 else 1091 VOP_UNLOCK(vp, 0); 1092 return (error); 1093 } 1094 1095 /* 1096 * Check permissions, allocate an open file structure, 1097 * and call the device open routine if any. 1098 */ 1099 int 1100 sys_open(l, v, retval) 1101 struct lwp *l; 1102 void *v; 1103 register_t *retval; 1104 { 1105 struct sys_open_args /* { 1106 syscallarg(const char *) path; 1107 syscallarg(int) flags; 1108 syscallarg(int) mode; 1109 } */ *uap = v; 1110 struct proc *p = l->l_proc; 1111 struct cwdinfo *cwdi = p->p_cwdi; 1112 struct filedesc *fdp = p->p_fd; 1113 struct file *fp; 1114 struct vnode *vp; 1115 int flags, cmode; 1116 int type, indx, error; 1117 struct flock lf; 1118 struct nameidata nd; 1119 1120 flags = FFLAGS(SCARG(uap, flags)); 1121 if ((flags & (FREAD | FWRITE)) == 0) 1122 return (EINVAL); 1123 /* falloc() will use the file descriptor for us */ 1124 if ((error = falloc(p, &fp, &indx)) != 0) 1125 return (error); 1126 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1127 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1128 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1129 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1130 FILE_UNUSE(fp, p); 1131 fdp->fd_ofiles[indx] = NULL; 1132 ffree(fp); 1133 if ((error == ENODEV || error == ENXIO) && 1134 l->l_dupfd >= 0 && /* XXX from fdopen */ 1135 (error = 1136 dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) { 1137 *retval = indx; 1138 return (0); 1139 } 1140 if (error == ERESTART) 1141 error = EINTR; 1142 fdremove(fdp, indx); 1143 return (error); 1144 } 1145 l->l_dupfd = 0; 1146 vp = nd.ni_vp; 1147 fp->f_flag = flags & FMASK; 1148 fp->f_type = DTYPE_VNODE; 1149 fp->f_ops = &vnops; 1150 fp->f_data = vp; 1151 if (flags & (O_EXLOCK | O_SHLOCK)) { 1152 lf.l_whence = SEEK_SET; 1153 lf.l_start = 0; 1154 lf.l_len = 0; 1155 if (flags & O_EXLOCK) 1156 lf.l_type = F_WRLCK; 1157 else 1158 lf.l_type = F_RDLCK; 1159 type = F_FLOCK; 1160 if ((flags & FNONBLOCK) == 0) 1161 type |= F_WAIT; 1162 VOP_UNLOCK(vp, 0); 1163 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1164 if (error) { 1165 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1166 FILE_UNUSE(fp, p); 1167 ffree(fp); 1168 fdremove(fdp, indx); 1169 return (error); 1170 } 1171 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1172 fp->f_flag |= FHASLOCK; 1173 } 1174 VOP_UNLOCK(vp, 0); 1175 *retval = indx; 1176 FILE_SET_MATURE(fp); 1177 FILE_UNUSE(fp, p); 1178 return (0); 1179 } 1180 1181 /* 1182 * Get file handle system call 1183 */ 1184 int 1185 sys_getfh(l, v, retval) 1186 struct lwp *l; 1187 void *v; 1188 register_t *retval; 1189 { 1190 struct sys_getfh_args /* { 1191 syscallarg(char *) fname; 1192 syscallarg(fhandle_t *) fhp; 1193 } */ *uap = v; 1194 struct proc *p = l->l_proc; 1195 struct vnode *vp; 1196 fhandle_t fh; 1197 int error; 1198 struct nameidata nd; 1199 1200 /* 1201 * Must be super user 1202 */ 1203 error = suser(p->p_ucred, &p->p_acflag); 1204 if (error) 1205 return (error); 1206 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1207 SCARG(uap, fname), p); 1208 error = namei(&nd); 1209 if (error) 1210 return (error); 1211 vp = nd.ni_vp; 1212 memset(&fh, 0, sizeof(fh)); 1213 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1214 error = VFS_VPTOFH(vp, &fh.fh_fid); 1215 vput(vp); 1216 if (error) 1217 return (error); 1218 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1219 return (error); 1220 } 1221 1222 /* 1223 * Open a file given a file handle. 1224 * 1225 * Check permissions, allocate an open file structure, 1226 * and call the device open routine if any. 1227 */ 1228 int 1229 sys_fhopen(l, v, retval) 1230 struct lwp *l; 1231 void *v; 1232 register_t *retval; 1233 { 1234 struct sys_fhopen_args /* { 1235 syscallarg(const fhandle_t *) fhp; 1236 syscallarg(int) flags; 1237 } */ *uap = v; 1238 struct proc *p = l->l_proc; 1239 struct filedesc *fdp = p->p_fd; 1240 struct file *fp; 1241 struct vnode *vp = NULL; 1242 struct mount *mp; 1243 struct ucred *cred = p->p_ucred; 1244 int flags; 1245 struct file *nfp; 1246 int type, indx, error=0; 1247 struct flock lf; 1248 struct vattr va; 1249 fhandle_t fh; 1250 1251 /* 1252 * Must be super user 1253 */ 1254 if ((error = suser(p->p_ucred, &p->p_acflag))) 1255 return (error); 1256 1257 flags = FFLAGS(SCARG(uap, flags)); 1258 if ((flags & (FREAD | FWRITE)) == 0) 1259 return (EINVAL); 1260 if ((flags & O_CREAT)) 1261 return (EINVAL); 1262 /* falloc() will use the file descriptor for us */ 1263 if ((error = falloc(p, &nfp, &indx)) != 0) 1264 return (error); 1265 fp = nfp; 1266 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1267 goto bad; 1268 1269 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1270 error = ESTALE; 1271 goto bad; 1272 } 1273 1274 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1275 vp = NULL; /* most likely unnecessary sanity for bad: */ 1276 goto bad; 1277 } 1278 1279 /* Now do an effective vn_open */ 1280 1281 if (vp->v_type == VSOCK) { 1282 error = EOPNOTSUPP; 1283 goto bad; 1284 } 1285 if (flags & FREAD) { 1286 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 1287 goto bad; 1288 } 1289 if (flags & (FWRITE | O_TRUNC)) { 1290 if (vp->v_type == VDIR) { 1291 error = EISDIR; 1292 goto bad; 1293 } 1294 if ((error = vn_writechk(vp)) != 0 || 1295 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 1296 goto bad; 1297 } 1298 if (flags & O_TRUNC) { 1299 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1300 goto bad; 1301 VOP_UNLOCK(vp, 0); /* XXX */ 1302 VOP_LEASE(vp, p, cred, LEASE_WRITE); 1303 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1304 VATTR_NULL(&va); 1305 va.va_size = 0; 1306 error = VOP_SETATTR(vp, &va, cred, p); 1307 vn_finished_write(mp, 0); 1308 if (error) 1309 goto bad; 1310 } 1311 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) 1312 goto bad; 1313 if (vp->v_type == VREG && 1314 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1315 error = EIO; 1316 goto bad; 1317 } 1318 if (flags & FWRITE) 1319 vp->v_writecount++; 1320 1321 /* done with modified vn_open, now finish what sys_open does. */ 1322 1323 fp->f_flag = flags & FMASK; 1324 fp->f_type = DTYPE_VNODE; 1325 fp->f_ops = &vnops; 1326 fp->f_data = vp; 1327 if (flags & (O_EXLOCK | O_SHLOCK)) { 1328 lf.l_whence = SEEK_SET; 1329 lf.l_start = 0; 1330 lf.l_len = 0; 1331 if (flags & O_EXLOCK) 1332 lf.l_type = F_WRLCK; 1333 else 1334 lf.l_type = F_RDLCK; 1335 type = F_FLOCK; 1336 if ((flags & FNONBLOCK) == 0) 1337 type |= F_WAIT; 1338 VOP_UNLOCK(vp, 0); 1339 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1340 if (error) { 1341 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1342 FILE_UNUSE(fp, p); 1343 ffree(fp); 1344 fdremove(fdp, indx); 1345 return (error); 1346 } 1347 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1348 fp->f_flag |= FHASLOCK; 1349 } 1350 VOP_UNLOCK(vp, 0); 1351 *retval = indx; 1352 FILE_SET_MATURE(fp); 1353 FILE_UNUSE(fp, p); 1354 return (0); 1355 1356 bad: 1357 FILE_UNUSE(fp, p); 1358 ffree(fp); 1359 fdremove(fdp, indx); 1360 if (vp != NULL) 1361 vput(vp); 1362 return (error); 1363 } 1364 1365 /* ARGSUSED */ 1366 int 1367 sys_fhstat(l, v, retval) 1368 struct lwp *l; 1369 void *v; 1370 register_t *retval; 1371 { 1372 struct sys_fhstat_args /* { 1373 syscallarg(const fhandle_t *) fhp; 1374 syscallarg(struct stat *) sb; 1375 } */ *uap = v; 1376 struct proc *p = l->l_proc; 1377 struct stat sb; 1378 int error; 1379 fhandle_t fh; 1380 struct mount *mp; 1381 struct vnode *vp; 1382 1383 /* 1384 * Must be super user 1385 */ 1386 if ((error = suser(p->p_ucred, &p->p_acflag))) 1387 return (error); 1388 1389 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1390 return (error); 1391 1392 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1393 return (ESTALE); 1394 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1395 return (error); 1396 error = vn_stat(vp, &sb, p); 1397 vput(vp); 1398 if (error) 1399 return (error); 1400 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1401 return (error); 1402 } 1403 1404 /* ARGSUSED */ 1405 int 1406 sys_fhstatvfs1(l, v, retval) 1407 struct lwp *l; 1408 void *v; 1409 register_t *retval; 1410 { 1411 struct sys_fhstatvfs1_args /* 1412 syscallarg(const fhandle_t *) fhp; 1413 syscallarg(struct statvfs *) buf; 1414 syscallarg(int) flags; 1415 } */ *uap = v; 1416 struct proc *p = l->l_proc; 1417 struct statvfs 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)) != 0) 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 1437 mp = vp->v_mount; 1438 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) { 1439 vput(vp); 1440 return error; 1441 } 1442 vput(vp); 1443 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1444 } 1445 1446 /* 1447 * Create a special file. 1448 */ 1449 /* ARGSUSED */ 1450 int 1451 sys_mknod(l, v, retval) 1452 struct lwp *l; 1453 void *v; 1454 register_t *retval; 1455 { 1456 struct sys_mknod_args /* { 1457 syscallarg(const char *) path; 1458 syscallarg(int) mode; 1459 syscallarg(int) dev; 1460 } */ *uap = v; 1461 struct proc *p = l->l_proc; 1462 struct vnode *vp; 1463 struct mount *mp; 1464 struct vattr vattr; 1465 int error; 1466 int whiteout = 0; 1467 struct nameidata nd; 1468 1469 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1470 return (error); 1471 restart: 1472 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1473 if ((error = namei(&nd)) != 0) 1474 return (error); 1475 vp = nd.ni_vp; 1476 if (vp != NULL) 1477 error = EEXIST; 1478 else { 1479 VATTR_NULL(&vattr); 1480 vattr.va_mode = 1481 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1482 vattr.va_rdev = SCARG(uap, dev); 1483 whiteout = 0; 1484 1485 switch (SCARG(uap, mode) & S_IFMT) { 1486 case S_IFMT: /* used by badsect to flag bad sectors */ 1487 vattr.va_type = VBAD; 1488 break; 1489 case S_IFCHR: 1490 vattr.va_type = VCHR; 1491 break; 1492 case S_IFBLK: 1493 vattr.va_type = VBLK; 1494 break; 1495 case S_IFWHT: 1496 whiteout = 1; 1497 break; 1498 default: 1499 error = EINVAL; 1500 break; 1501 } 1502 } 1503 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1504 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1505 if (nd.ni_dvp == vp) 1506 vrele(nd.ni_dvp); 1507 else 1508 vput(nd.ni_dvp); 1509 if (vp) 1510 vrele(vp); 1511 if ((error = vn_start_write(NULL, &mp, 1512 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1513 return (error); 1514 goto restart; 1515 } 1516 if (!error) { 1517 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1518 if (whiteout) { 1519 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1520 if (error) 1521 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1522 vput(nd.ni_dvp); 1523 } else { 1524 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1525 &nd.ni_cnd, &vattr); 1526 if (error == 0) 1527 vput(nd.ni_vp); 1528 } 1529 } else { 1530 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1531 if (nd.ni_dvp == vp) 1532 vrele(nd.ni_dvp); 1533 else 1534 vput(nd.ni_dvp); 1535 if (vp) 1536 vrele(vp); 1537 } 1538 vn_finished_write(mp, 0); 1539 return (error); 1540 } 1541 1542 /* 1543 * Create a named pipe. 1544 */ 1545 /* ARGSUSED */ 1546 int 1547 sys_mkfifo(l, v, retval) 1548 struct lwp *l; 1549 void *v; 1550 register_t *retval; 1551 { 1552 struct sys_mkfifo_args /* { 1553 syscallarg(const char *) path; 1554 syscallarg(int) mode; 1555 } */ *uap = v; 1556 struct proc *p = l->l_proc; 1557 struct mount *mp; 1558 struct vattr vattr; 1559 int error; 1560 struct nameidata nd; 1561 1562 restart: 1563 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1564 if ((error = namei(&nd)) != 0) 1565 return (error); 1566 if (nd.ni_vp != NULL) { 1567 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1568 if (nd.ni_dvp == nd.ni_vp) 1569 vrele(nd.ni_dvp); 1570 else 1571 vput(nd.ni_dvp); 1572 vrele(nd.ni_vp); 1573 return (EEXIST); 1574 } 1575 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1576 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1577 if (nd.ni_dvp == nd.ni_vp) 1578 vrele(nd.ni_dvp); 1579 else 1580 vput(nd.ni_dvp); 1581 if (nd.ni_vp) 1582 vrele(nd.ni_vp); 1583 if ((error = vn_start_write(NULL, &mp, 1584 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1585 return (error); 1586 goto restart; 1587 } 1588 VATTR_NULL(&vattr); 1589 vattr.va_type = VFIFO; 1590 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1591 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1592 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1593 if (error == 0) 1594 vput(nd.ni_vp); 1595 vn_finished_write(mp, 0); 1596 return (error); 1597 } 1598 1599 /* 1600 * Make a hard file link. 1601 */ 1602 /* ARGSUSED */ 1603 int 1604 sys_link(l, v, retval) 1605 struct lwp *l; 1606 void *v; 1607 register_t *retval; 1608 { 1609 struct sys_link_args /* { 1610 syscallarg(const char *) path; 1611 syscallarg(const char *) link; 1612 } */ *uap = v; 1613 struct proc *p = l->l_proc; 1614 struct vnode *vp; 1615 struct mount *mp; 1616 struct nameidata nd; 1617 int error; 1618 1619 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1620 if ((error = namei(&nd)) != 0) 1621 return (error); 1622 vp = nd.ni_vp; 1623 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1624 vrele(vp); 1625 return (error); 1626 } 1627 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1628 if ((error = namei(&nd)) != 0) 1629 goto out; 1630 if (nd.ni_vp) { 1631 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1632 if (nd.ni_dvp == nd.ni_vp) 1633 vrele(nd.ni_dvp); 1634 else 1635 vput(nd.ni_dvp); 1636 vrele(nd.ni_vp); 1637 error = EEXIST; 1638 goto out; 1639 } 1640 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1641 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1642 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1643 out: 1644 vrele(vp); 1645 vn_finished_write(mp, 0); 1646 return (error); 1647 } 1648 1649 /* 1650 * Make a symbolic link. 1651 */ 1652 /* ARGSUSED */ 1653 int 1654 sys_symlink(l, v, retval) 1655 struct lwp *l; 1656 void *v; 1657 register_t *retval; 1658 { 1659 struct sys_symlink_args /* { 1660 syscallarg(const char *) path; 1661 syscallarg(const char *) link; 1662 } */ *uap = v; 1663 struct proc *p = l->l_proc; 1664 struct mount *mp; 1665 struct vattr vattr; 1666 char *path; 1667 int error; 1668 struct nameidata nd; 1669 1670 path = PNBUF_GET(); 1671 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1672 if (error) 1673 goto out; 1674 restart: 1675 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1676 if ((error = namei(&nd)) != 0) 1677 goto out; 1678 if (nd.ni_vp) { 1679 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1680 if (nd.ni_dvp == nd.ni_vp) 1681 vrele(nd.ni_dvp); 1682 else 1683 vput(nd.ni_dvp); 1684 vrele(nd.ni_vp); 1685 error = EEXIST; 1686 goto out; 1687 } 1688 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1689 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1690 if (nd.ni_dvp == nd.ni_vp) 1691 vrele(nd.ni_dvp); 1692 else 1693 vput(nd.ni_dvp); 1694 if ((error = vn_start_write(NULL, &mp, 1695 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1696 return (error); 1697 goto restart; 1698 } 1699 VATTR_NULL(&vattr); 1700 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1701 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1702 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1703 if (error == 0) 1704 vput(nd.ni_vp); 1705 vn_finished_write(mp, 0); 1706 out: 1707 PNBUF_PUT(path); 1708 return (error); 1709 } 1710 1711 /* 1712 * Delete a whiteout from the filesystem. 1713 */ 1714 /* ARGSUSED */ 1715 int 1716 sys_undelete(l, v, retval) 1717 struct lwp *l; 1718 void *v; 1719 register_t *retval; 1720 { 1721 struct sys_undelete_args /* { 1722 syscallarg(const char *) path; 1723 } */ *uap = v; 1724 struct proc *p = l->l_proc; 1725 int error; 1726 struct mount *mp; 1727 struct nameidata nd; 1728 1729 restart: 1730 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1731 SCARG(uap, path), p); 1732 error = namei(&nd); 1733 if (error) 1734 return (error); 1735 1736 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1737 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1738 if (nd.ni_dvp == nd.ni_vp) 1739 vrele(nd.ni_dvp); 1740 else 1741 vput(nd.ni_dvp); 1742 if (nd.ni_vp) 1743 vrele(nd.ni_vp); 1744 return (EEXIST); 1745 } 1746 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1747 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1748 if (nd.ni_dvp == nd.ni_vp) 1749 vrele(nd.ni_dvp); 1750 else 1751 vput(nd.ni_dvp); 1752 if ((error = vn_start_write(NULL, &mp, 1753 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1754 return (error); 1755 goto restart; 1756 } 1757 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1758 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1759 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1760 vput(nd.ni_dvp); 1761 vn_finished_write(mp, 0); 1762 return (error); 1763 } 1764 1765 /* 1766 * Delete a name from the filesystem. 1767 */ 1768 /* ARGSUSED */ 1769 int 1770 sys_unlink(l, v, retval) 1771 struct lwp *l; 1772 void *v; 1773 register_t *retval; 1774 { 1775 struct sys_unlink_args /* { 1776 syscallarg(const char *) path; 1777 } */ *uap = v; 1778 struct proc *p = l->l_proc; 1779 struct mount *mp; 1780 struct vnode *vp; 1781 int error; 1782 struct nameidata nd; 1783 1784 restart: 1785 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1786 SCARG(uap, path), p); 1787 if ((error = namei(&nd)) != 0) 1788 return (error); 1789 vp = nd.ni_vp; 1790 1791 /* 1792 * The root of a mounted filesystem cannot be deleted. 1793 */ 1794 if (vp->v_flag & VROOT) { 1795 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1796 if (nd.ni_dvp == vp) 1797 vrele(nd.ni_dvp); 1798 else 1799 vput(nd.ni_dvp); 1800 vput(vp); 1801 error = EBUSY; 1802 goto out; 1803 } 1804 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1805 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1806 if (nd.ni_dvp == vp) 1807 vrele(nd.ni_dvp); 1808 else 1809 vput(nd.ni_dvp); 1810 vput(vp); 1811 if ((error = vn_start_write(NULL, &mp, 1812 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1813 return (error); 1814 goto restart; 1815 } 1816 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1817 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1818 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1819 vn_finished_write(mp, 0); 1820 out: 1821 return (error); 1822 } 1823 1824 /* 1825 * Reposition read/write file offset. 1826 */ 1827 int 1828 sys_lseek(l, v, retval) 1829 struct lwp *l; 1830 void *v; 1831 register_t *retval; 1832 { 1833 struct sys_lseek_args /* { 1834 syscallarg(int) fd; 1835 syscallarg(int) pad; 1836 syscallarg(off_t) offset; 1837 syscallarg(int) whence; 1838 } */ *uap = v; 1839 struct proc *p = l->l_proc; 1840 struct ucred *cred = p->p_ucred; 1841 struct filedesc *fdp = p->p_fd; 1842 struct file *fp; 1843 struct vnode *vp; 1844 struct vattr vattr; 1845 off_t newoff; 1846 int error; 1847 1848 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1849 return (EBADF); 1850 1851 FILE_USE(fp); 1852 1853 vp = (struct vnode *)fp->f_data; 1854 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1855 error = ESPIPE; 1856 goto out; 1857 } 1858 1859 switch (SCARG(uap, whence)) { 1860 case SEEK_CUR: 1861 newoff = fp->f_offset + SCARG(uap, offset); 1862 break; 1863 case SEEK_END: 1864 error = VOP_GETATTR(vp, &vattr, cred, p); 1865 if (error) 1866 goto out; 1867 newoff = SCARG(uap, offset) + vattr.va_size; 1868 break; 1869 case SEEK_SET: 1870 newoff = SCARG(uap, offset); 1871 break; 1872 default: 1873 error = EINVAL; 1874 goto out; 1875 } 1876 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1877 goto out; 1878 1879 *(off_t *)retval = fp->f_offset = newoff; 1880 out: 1881 FILE_UNUSE(fp, p); 1882 return (error); 1883 } 1884 1885 /* 1886 * Positional read system call. 1887 */ 1888 int 1889 sys_pread(l, v, retval) 1890 struct lwp *l; 1891 void *v; 1892 register_t *retval; 1893 { 1894 struct sys_pread_args /* { 1895 syscallarg(int) fd; 1896 syscallarg(void *) buf; 1897 syscallarg(size_t) nbyte; 1898 syscallarg(off_t) offset; 1899 } */ *uap = v; 1900 struct proc *p = l->l_proc; 1901 struct filedesc *fdp = p->p_fd; 1902 struct file *fp; 1903 struct vnode *vp; 1904 off_t offset; 1905 int error, fd = SCARG(uap, fd); 1906 1907 if ((fp = fd_getfile(fdp, fd)) == NULL) 1908 return (EBADF); 1909 1910 if ((fp->f_flag & FREAD) == 0) { 1911 simple_unlock(&fp->f_slock); 1912 return (EBADF); 1913 } 1914 1915 FILE_USE(fp); 1916 1917 vp = (struct vnode *)fp->f_data; 1918 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1919 error = ESPIPE; 1920 goto out; 1921 } 1922 1923 offset = SCARG(uap, offset); 1924 1925 /* 1926 * XXX This works because no file systems actually 1927 * XXX take any action on the seek operation. 1928 */ 1929 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1930 goto out; 1931 1932 /* dofileread() will unuse the descriptor for us */ 1933 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1934 &offset, 0, retval)); 1935 1936 out: 1937 FILE_UNUSE(fp, p); 1938 return (error); 1939 } 1940 1941 /* 1942 * Positional scatter read system call. 1943 */ 1944 int 1945 sys_preadv(l, v, retval) 1946 struct lwp *l; 1947 void *v; 1948 register_t *retval; 1949 { 1950 struct sys_preadv_args /* { 1951 syscallarg(int) fd; 1952 syscallarg(const struct iovec *) iovp; 1953 syscallarg(int) iovcnt; 1954 syscallarg(off_t) offset; 1955 } */ *uap = v; 1956 struct proc *p = l->l_proc; 1957 struct filedesc *fdp = p->p_fd; 1958 struct file *fp; 1959 struct vnode *vp; 1960 off_t offset; 1961 int error, fd = SCARG(uap, fd); 1962 1963 if ((fp = fd_getfile(fdp, fd)) == NULL) 1964 return (EBADF); 1965 1966 if ((fp->f_flag & FREAD) == 0) { 1967 simple_unlock(&fp->f_slock); 1968 return (EBADF); 1969 } 1970 1971 FILE_USE(fp); 1972 1973 vp = (struct vnode *)fp->f_data; 1974 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1975 error = ESPIPE; 1976 goto out; 1977 } 1978 1979 offset = SCARG(uap, offset); 1980 1981 /* 1982 * XXX This works because no file systems actually 1983 * XXX take any action on the seek operation. 1984 */ 1985 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1986 goto out; 1987 1988 /* dofilereadv() will unuse the descriptor for us */ 1989 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1990 &offset, 0, retval)); 1991 1992 out: 1993 FILE_UNUSE(fp, p); 1994 return (error); 1995 } 1996 1997 /* 1998 * Positional write system call. 1999 */ 2000 int 2001 sys_pwrite(l, v, retval) 2002 struct lwp *l; 2003 void *v; 2004 register_t *retval; 2005 { 2006 struct sys_pwrite_args /* { 2007 syscallarg(int) fd; 2008 syscallarg(const void *) buf; 2009 syscallarg(size_t) nbyte; 2010 syscallarg(off_t) offset; 2011 } */ *uap = v; 2012 struct proc *p = l->l_proc; 2013 struct filedesc *fdp = p->p_fd; 2014 struct file *fp; 2015 struct vnode *vp; 2016 off_t offset; 2017 int error, fd = SCARG(uap, fd); 2018 2019 if ((fp = fd_getfile(fdp, fd)) == NULL) 2020 return (EBADF); 2021 2022 if ((fp->f_flag & FWRITE) == 0) { 2023 simple_unlock(&fp->f_slock); 2024 return (EBADF); 2025 } 2026 2027 FILE_USE(fp); 2028 2029 vp = (struct vnode *)fp->f_data; 2030 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2031 error = ESPIPE; 2032 goto out; 2033 } 2034 2035 offset = SCARG(uap, offset); 2036 2037 /* 2038 * XXX This works because no file systems actually 2039 * XXX take any action on the seek operation. 2040 */ 2041 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2042 goto out; 2043 2044 /* dofilewrite() will unuse the descriptor for us */ 2045 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2046 &offset, 0, retval)); 2047 2048 out: 2049 FILE_UNUSE(fp, p); 2050 return (error); 2051 } 2052 2053 /* 2054 * Positional gather write system call. 2055 */ 2056 int 2057 sys_pwritev(l, v, retval) 2058 struct lwp *l; 2059 void *v; 2060 register_t *retval; 2061 { 2062 struct sys_pwritev_args /* { 2063 syscallarg(int) fd; 2064 syscallarg(const struct iovec *) iovp; 2065 syscallarg(int) iovcnt; 2066 syscallarg(off_t) offset; 2067 } */ *uap = v; 2068 struct proc *p = l->l_proc; 2069 struct filedesc *fdp = p->p_fd; 2070 struct file *fp; 2071 struct vnode *vp; 2072 off_t offset; 2073 int error, fd = SCARG(uap, fd); 2074 2075 if ((fp = fd_getfile(fdp, fd)) == NULL) 2076 return (EBADF); 2077 2078 if ((fp->f_flag & FWRITE) == 0) { 2079 simple_unlock(&fp->f_slock); 2080 return (EBADF); 2081 } 2082 2083 FILE_USE(fp); 2084 2085 vp = (struct vnode *)fp->f_data; 2086 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2087 error = ESPIPE; 2088 goto out; 2089 } 2090 2091 offset = SCARG(uap, offset); 2092 2093 /* 2094 * XXX This works because no file systems actually 2095 * XXX take any action on the seek operation. 2096 */ 2097 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2098 goto out; 2099 2100 /* dofilewritev() will unuse the descriptor for us */ 2101 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2102 &offset, 0, retval)); 2103 2104 out: 2105 FILE_UNUSE(fp, p); 2106 return (error); 2107 } 2108 2109 /* 2110 * Check access permissions. 2111 */ 2112 int 2113 sys_access(l, v, retval) 2114 struct lwp *l; 2115 void *v; 2116 register_t *retval; 2117 { 2118 struct sys_access_args /* { 2119 syscallarg(const char *) path; 2120 syscallarg(int) flags; 2121 } */ *uap = v; 2122 struct proc *p = l->l_proc; 2123 struct ucred *cred; 2124 struct vnode *vp; 2125 int error, flags; 2126 struct nameidata nd; 2127 2128 cred = crdup(p->p_ucred); 2129 cred->cr_uid = p->p_cred->p_ruid; 2130 cred->cr_gid = p->p_cred->p_rgid; 2131 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2132 SCARG(uap, path), p); 2133 /* Override default credentials */ 2134 nd.ni_cnd.cn_cred = cred; 2135 if ((error = namei(&nd)) != 0) 2136 goto out; 2137 vp = nd.ni_vp; 2138 2139 /* Flags == 0 means only check for existence. */ 2140 if (SCARG(uap, flags)) { 2141 flags = 0; 2142 if (SCARG(uap, flags) & R_OK) 2143 flags |= VREAD; 2144 if (SCARG(uap, flags) & W_OK) 2145 flags |= VWRITE; 2146 if (SCARG(uap, flags) & X_OK) 2147 flags |= VEXEC; 2148 2149 error = VOP_ACCESS(vp, flags, cred, p); 2150 if (!error && (flags & VWRITE)) 2151 error = vn_writechk(vp); 2152 } 2153 vput(vp); 2154 out: 2155 crfree(cred); 2156 return (error); 2157 } 2158 2159 /* 2160 * Get file status; this version follows links. 2161 */ 2162 /* ARGSUSED */ 2163 int 2164 sys___stat13(l, v, retval) 2165 struct lwp *l; 2166 void *v; 2167 register_t *retval; 2168 { 2169 struct sys___stat13_args /* { 2170 syscallarg(const char *) path; 2171 syscallarg(struct stat *) ub; 2172 } */ *uap = v; 2173 struct proc *p = l->l_proc; 2174 struct stat sb; 2175 int error; 2176 struct nameidata nd; 2177 2178 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2179 SCARG(uap, path), p); 2180 if ((error = namei(&nd)) != 0) 2181 return (error); 2182 error = vn_stat(nd.ni_vp, &sb, p); 2183 vput(nd.ni_vp); 2184 if (error) 2185 return (error); 2186 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2187 return (error); 2188 } 2189 2190 /* 2191 * Get file status; this version does not follow links. 2192 */ 2193 /* ARGSUSED */ 2194 int 2195 sys___lstat13(l, v, retval) 2196 struct lwp *l; 2197 void *v; 2198 register_t *retval; 2199 { 2200 struct sys___lstat13_args /* { 2201 syscallarg(const char *) path; 2202 syscallarg(struct stat *) ub; 2203 } */ *uap = v; 2204 struct proc *p = l->l_proc; 2205 struct stat sb; 2206 int error; 2207 struct nameidata nd; 2208 2209 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2210 SCARG(uap, path), p); 2211 if ((error = namei(&nd)) != 0) 2212 return (error); 2213 error = vn_stat(nd.ni_vp, &sb, p); 2214 vput(nd.ni_vp); 2215 if (error) 2216 return (error); 2217 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2218 return (error); 2219 } 2220 2221 /* 2222 * Get configurable pathname variables. 2223 */ 2224 /* ARGSUSED */ 2225 int 2226 sys_pathconf(l, v, retval) 2227 struct lwp *l; 2228 void *v; 2229 register_t *retval; 2230 { 2231 struct sys_pathconf_args /* { 2232 syscallarg(const char *) path; 2233 syscallarg(int) name; 2234 } */ *uap = v; 2235 struct proc *p = l->l_proc; 2236 int error; 2237 struct nameidata nd; 2238 2239 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2240 SCARG(uap, path), p); 2241 if ((error = namei(&nd)) != 0) 2242 return (error); 2243 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2244 vput(nd.ni_vp); 2245 return (error); 2246 } 2247 2248 /* 2249 * Return target name of a symbolic link. 2250 */ 2251 /* ARGSUSED */ 2252 int 2253 sys_readlink(l, v, retval) 2254 struct lwp *l; 2255 void *v; 2256 register_t *retval; 2257 { 2258 struct sys_readlink_args /* { 2259 syscallarg(const char *) path; 2260 syscallarg(char *) buf; 2261 syscallarg(size_t) count; 2262 } */ *uap = v; 2263 struct proc *p = l->l_proc; 2264 struct vnode *vp; 2265 struct iovec aiov; 2266 struct uio auio; 2267 int error; 2268 struct nameidata nd; 2269 2270 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2271 SCARG(uap, path), p); 2272 if ((error = namei(&nd)) != 0) 2273 return (error); 2274 vp = nd.ni_vp; 2275 if (vp->v_type != VLNK) 2276 error = EINVAL; 2277 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2278 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 2279 aiov.iov_base = SCARG(uap, buf); 2280 aiov.iov_len = SCARG(uap, count); 2281 auio.uio_iov = &aiov; 2282 auio.uio_iovcnt = 1; 2283 auio.uio_offset = 0; 2284 auio.uio_rw = UIO_READ; 2285 auio.uio_segflg = UIO_USERSPACE; 2286 auio.uio_procp = p; 2287 auio.uio_resid = SCARG(uap, count); 2288 error = VOP_READLINK(vp, &auio, p->p_ucred); 2289 } 2290 vput(vp); 2291 *retval = SCARG(uap, count) - auio.uio_resid; 2292 return (error); 2293 } 2294 2295 /* 2296 * Change flags of a file given a path name. 2297 */ 2298 /* ARGSUSED */ 2299 int 2300 sys_chflags(l, v, retval) 2301 struct lwp *l; 2302 void *v; 2303 register_t *retval; 2304 { 2305 struct sys_chflags_args /* { 2306 syscallarg(const char *) path; 2307 syscallarg(u_long) flags; 2308 } */ *uap = v; 2309 struct proc *p = l->l_proc; 2310 struct vnode *vp; 2311 int error; 2312 struct nameidata nd; 2313 2314 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2315 if ((error = namei(&nd)) != 0) 2316 return (error); 2317 vp = nd.ni_vp; 2318 error = change_flags(vp, SCARG(uap, flags), p); 2319 vput(vp); 2320 return (error); 2321 } 2322 2323 /* 2324 * Change flags of a file given a file descriptor. 2325 */ 2326 /* ARGSUSED */ 2327 int 2328 sys_fchflags(l, v, retval) 2329 struct lwp *l; 2330 void *v; 2331 register_t *retval; 2332 { 2333 struct sys_fchflags_args /* { 2334 syscallarg(int) fd; 2335 syscallarg(u_long) flags; 2336 } */ *uap = v; 2337 struct proc *p = l->l_proc; 2338 struct vnode *vp; 2339 struct file *fp; 2340 int error; 2341 2342 /* getvnode() will use the descriptor for us */ 2343 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2344 return (error); 2345 vp = (struct vnode *)fp->f_data; 2346 error = change_flags(vp, SCARG(uap, flags), p); 2347 VOP_UNLOCK(vp, 0); 2348 FILE_UNUSE(fp, p); 2349 return (error); 2350 } 2351 2352 /* 2353 * Change flags of a file given a path name; this version does 2354 * not follow links. 2355 */ 2356 int 2357 sys_lchflags(l, v, retval) 2358 struct lwp *l; 2359 void *v; 2360 register_t *retval; 2361 { 2362 struct sys_lchflags_args /* { 2363 syscallarg(const char *) path; 2364 syscallarg(u_long) flags; 2365 } */ *uap = v; 2366 struct proc *p = l->l_proc; 2367 struct vnode *vp; 2368 int error; 2369 struct nameidata nd; 2370 2371 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2372 if ((error = namei(&nd)) != 0) 2373 return (error); 2374 vp = nd.ni_vp; 2375 error = change_flags(vp, SCARG(uap, flags), p); 2376 vput(vp); 2377 return (error); 2378 } 2379 2380 /* 2381 * Common routine to change flags of a file. 2382 */ 2383 int 2384 change_flags(vp, flags, p) 2385 struct vnode *vp; 2386 u_long flags; 2387 struct proc *p; 2388 { 2389 struct mount *mp; 2390 struct vattr vattr; 2391 int error; 2392 2393 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2394 return (error); 2395 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2396 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2397 /* 2398 * Non-superusers cannot change the flags on devices, even if they 2399 * own them. 2400 */ 2401 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2402 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2403 goto out; 2404 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2405 error = EINVAL; 2406 goto out; 2407 } 2408 } 2409 VATTR_NULL(&vattr); 2410 vattr.va_flags = flags; 2411 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2412 out: 2413 vn_finished_write(mp, 0); 2414 return (error); 2415 } 2416 2417 /* 2418 * Change mode of a file given path name; this version follows links. 2419 */ 2420 /* ARGSUSED */ 2421 int 2422 sys_chmod(l, v, retval) 2423 struct lwp *l; 2424 void *v; 2425 register_t *retval; 2426 { 2427 struct sys_chmod_args /* { 2428 syscallarg(const char *) path; 2429 syscallarg(int) mode; 2430 } */ *uap = v; 2431 struct proc *p = l->l_proc; 2432 int error; 2433 struct nameidata nd; 2434 2435 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2436 if ((error = namei(&nd)) != 0) 2437 return (error); 2438 2439 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2440 2441 vrele(nd.ni_vp); 2442 return (error); 2443 } 2444 2445 /* 2446 * Change mode of a file given a file descriptor. 2447 */ 2448 /* ARGSUSED */ 2449 int 2450 sys_fchmod(l, v, retval) 2451 struct lwp *l; 2452 void *v; 2453 register_t *retval; 2454 { 2455 struct sys_fchmod_args /* { 2456 syscallarg(int) fd; 2457 syscallarg(int) mode; 2458 } */ *uap = v; 2459 struct proc *p = l->l_proc; 2460 struct file *fp; 2461 int error; 2462 2463 /* getvnode() will use the descriptor for us */ 2464 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2465 return (error); 2466 2467 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 2468 FILE_UNUSE(fp, p); 2469 return (error); 2470 } 2471 2472 /* 2473 * Change mode of a file given path name; this version does not follow links. 2474 */ 2475 /* ARGSUSED */ 2476 int 2477 sys_lchmod(l, v, retval) 2478 struct lwp *l; 2479 void *v; 2480 register_t *retval; 2481 { 2482 struct sys_lchmod_args /* { 2483 syscallarg(const char *) path; 2484 syscallarg(int) mode; 2485 } */ *uap = v; 2486 struct proc *p = l->l_proc; 2487 int error; 2488 struct nameidata nd; 2489 2490 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2491 if ((error = namei(&nd)) != 0) 2492 return (error); 2493 2494 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2495 2496 vrele(nd.ni_vp); 2497 return (error); 2498 } 2499 2500 /* 2501 * Common routine to set mode given a vnode. 2502 */ 2503 static int 2504 change_mode(vp, mode, p) 2505 struct vnode *vp; 2506 int mode; 2507 struct proc *p; 2508 { 2509 struct mount *mp; 2510 struct vattr vattr; 2511 int error; 2512 2513 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2514 return (error); 2515 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2516 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2517 VATTR_NULL(&vattr); 2518 vattr.va_mode = mode & ALLPERMS; 2519 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2520 VOP_UNLOCK(vp, 0); 2521 vn_finished_write(mp, 0); 2522 return (error); 2523 } 2524 2525 /* 2526 * Set ownership given a path name; this version follows links. 2527 */ 2528 /* ARGSUSED */ 2529 int 2530 sys_chown(l, v, retval) 2531 struct lwp *l; 2532 void *v; 2533 register_t *retval; 2534 { 2535 struct sys_chown_args /* { 2536 syscallarg(const char *) path; 2537 syscallarg(uid_t) uid; 2538 syscallarg(gid_t) gid; 2539 } */ *uap = v; 2540 struct proc *p = l->l_proc; 2541 int error; 2542 struct nameidata nd; 2543 2544 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2545 if ((error = namei(&nd)) != 0) 2546 return (error); 2547 2548 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2549 2550 vrele(nd.ni_vp); 2551 return (error); 2552 } 2553 2554 /* 2555 * Set ownership given a path name; this version follows links. 2556 * Provides POSIX semantics. 2557 */ 2558 /* ARGSUSED */ 2559 int 2560 sys___posix_chown(l, v, retval) 2561 struct lwp *l; 2562 void *v; 2563 register_t *retval; 2564 { 2565 struct sys_chown_args /* { 2566 syscallarg(const char *) path; 2567 syscallarg(uid_t) uid; 2568 syscallarg(gid_t) gid; 2569 } */ *uap = v; 2570 struct proc *p = l->l_proc; 2571 int error; 2572 struct nameidata nd; 2573 2574 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2575 if ((error = namei(&nd)) != 0) 2576 return (error); 2577 2578 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2579 2580 vrele(nd.ni_vp); 2581 return (error); 2582 } 2583 2584 /* 2585 * Set ownership given a file descriptor. 2586 */ 2587 /* ARGSUSED */ 2588 int 2589 sys_fchown(l, v, retval) 2590 struct lwp *l; 2591 void *v; 2592 register_t *retval; 2593 { 2594 struct sys_fchown_args /* { 2595 syscallarg(int) fd; 2596 syscallarg(uid_t) uid; 2597 syscallarg(gid_t) gid; 2598 } */ *uap = v; 2599 struct proc *p = l->l_proc; 2600 int error; 2601 struct file *fp; 2602 2603 /* getvnode() will use the descriptor for us */ 2604 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2605 return (error); 2606 2607 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2608 SCARG(uap, gid), p, 0); 2609 FILE_UNUSE(fp, p); 2610 return (error); 2611 } 2612 2613 /* 2614 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2615 */ 2616 /* ARGSUSED */ 2617 int 2618 sys___posix_fchown(l, v, retval) 2619 struct lwp *l; 2620 void *v; 2621 register_t *retval; 2622 { 2623 struct sys_fchown_args /* { 2624 syscallarg(int) fd; 2625 syscallarg(uid_t) uid; 2626 syscallarg(gid_t) gid; 2627 } */ *uap = v; 2628 struct proc *p = l->l_proc; 2629 int error; 2630 struct file *fp; 2631 2632 /* getvnode() will use the descriptor for us */ 2633 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2634 return (error); 2635 2636 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2637 SCARG(uap, gid), p, 1); 2638 FILE_UNUSE(fp, p); 2639 return (error); 2640 } 2641 2642 /* 2643 * Set ownership given a path name; this version does not follow links. 2644 */ 2645 /* ARGSUSED */ 2646 int 2647 sys_lchown(l, v, retval) 2648 struct lwp *l; 2649 void *v; 2650 register_t *retval; 2651 { 2652 struct sys_lchown_args /* { 2653 syscallarg(const char *) path; 2654 syscallarg(uid_t) uid; 2655 syscallarg(gid_t) gid; 2656 } */ *uap = v; 2657 struct proc *p = l->l_proc; 2658 int error; 2659 struct nameidata nd; 2660 2661 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2662 if ((error = namei(&nd)) != 0) 2663 return (error); 2664 2665 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2666 2667 vrele(nd.ni_vp); 2668 return (error); 2669 } 2670 2671 /* 2672 * Set ownership given a path name; this version does not follow links. 2673 * Provides POSIX/XPG semantics. 2674 */ 2675 /* ARGSUSED */ 2676 int 2677 sys___posix_lchown(l, v, retval) 2678 struct lwp *l; 2679 void *v; 2680 register_t *retval; 2681 { 2682 struct sys_lchown_args /* { 2683 syscallarg(const char *) path; 2684 syscallarg(uid_t) uid; 2685 syscallarg(gid_t) gid; 2686 } */ *uap = v; 2687 struct proc *p = l->l_proc; 2688 int error; 2689 struct nameidata nd; 2690 2691 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2692 if ((error = namei(&nd)) != 0) 2693 return (error); 2694 2695 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2696 2697 vrele(nd.ni_vp); 2698 return (error); 2699 } 2700 2701 /* 2702 * Common routine to set ownership given a vnode. 2703 */ 2704 static int 2705 change_owner(vp, uid, gid, p, posix_semantics) 2706 struct vnode *vp; 2707 uid_t uid; 2708 gid_t gid; 2709 struct proc *p; 2710 int posix_semantics; 2711 { 2712 struct mount *mp; 2713 struct vattr vattr; 2714 mode_t newmode; 2715 int error; 2716 2717 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2718 return (error); 2719 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2720 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2721 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2722 goto out; 2723 2724 #define CHANGED(x) ((int)(x) != -1) 2725 newmode = vattr.va_mode; 2726 if (posix_semantics) { 2727 /* 2728 * POSIX/XPG semantics: if the caller is not the super-user, 2729 * clear set-user-id and set-group-id bits. Both POSIX and 2730 * the XPG consider the behaviour for calls by the super-user 2731 * implementation-defined; we leave the set-user-id and set- 2732 * group-id settings intact in that case. 2733 */ 2734 if (suser(p->p_ucred, NULL) != 0) 2735 newmode &= ~(S_ISUID | S_ISGID); 2736 } else { 2737 /* 2738 * NetBSD semantics: when changing owner and/or group, 2739 * clear the respective bit(s). 2740 */ 2741 if (CHANGED(uid)) 2742 newmode &= ~S_ISUID; 2743 if (CHANGED(gid)) 2744 newmode &= ~S_ISGID; 2745 } 2746 /* Update va_mode iff altered. */ 2747 if (vattr.va_mode == newmode) 2748 newmode = VNOVAL; 2749 2750 VATTR_NULL(&vattr); 2751 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2752 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2753 vattr.va_mode = newmode; 2754 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2755 #undef CHANGED 2756 2757 out: 2758 VOP_UNLOCK(vp, 0); 2759 vn_finished_write(mp, 0); 2760 return (error); 2761 } 2762 2763 /* 2764 * Set the access and modification times given a path name; this 2765 * version follows links. 2766 */ 2767 /* ARGSUSED */ 2768 int 2769 sys_utimes(l, v, retval) 2770 struct lwp *l; 2771 void *v; 2772 register_t *retval; 2773 { 2774 struct sys_utimes_args /* { 2775 syscallarg(const char *) path; 2776 syscallarg(const struct timeval *) tptr; 2777 } */ *uap = v; 2778 struct proc *p = l->l_proc; 2779 int error; 2780 struct nameidata nd; 2781 2782 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2783 if ((error = namei(&nd)) != 0) 2784 return (error); 2785 2786 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2787 2788 vrele(nd.ni_vp); 2789 return (error); 2790 } 2791 2792 /* 2793 * Set the access and modification times given a file descriptor. 2794 */ 2795 /* ARGSUSED */ 2796 int 2797 sys_futimes(l, v, retval) 2798 struct lwp *l; 2799 void *v; 2800 register_t *retval; 2801 { 2802 struct sys_futimes_args /* { 2803 syscallarg(int) fd; 2804 syscallarg(const struct timeval *) tptr; 2805 } */ *uap = v; 2806 struct proc *p = l->l_proc; 2807 int error; 2808 struct file *fp; 2809 2810 /* getvnode() will use the descriptor for us */ 2811 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2812 return (error); 2813 2814 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2815 FILE_UNUSE(fp, p); 2816 return (error); 2817 } 2818 2819 /* 2820 * Set the access and modification times given a path name; this 2821 * version does not follow links. 2822 */ 2823 /* ARGSUSED */ 2824 int 2825 sys_lutimes(l, v, retval) 2826 struct lwp *l; 2827 void *v; 2828 register_t *retval; 2829 { 2830 struct sys_lutimes_args /* { 2831 syscallarg(const char *) path; 2832 syscallarg(const struct timeval *) tptr; 2833 } */ *uap = v; 2834 struct proc *p = l->l_proc; 2835 int error; 2836 struct nameidata nd; 2837 2838 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2839 if ((error = namei(&nd)) != 0) 2840 return (error); 2841 2842 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2843 2844 vrele(nd.ni_vp); 2845 return (error); 2846 } 2847 2848 /* 2849 * Common routine to set access and modification times given a vnode. 2850 */ 2851 static int 2852 change_utimes(vp, tptr, p) 2853 struct vnode *vp; 2854 const struct timeval *tptr; 2855 struct proc *p; 2856 { 2857 struct timeval tv[2]; 2858 struct mount *mp; 2859 struct vattr vattr; 2860 int error; 2861 2862 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2863 return (error); 2864 VATTR_NULL(&vattr); 2865 if (tptr == NULL) { 2866 microtime(&tv[0]); 2867 tv[1] = tv[0]; 2868 vattr.va_vaflags |= VA_UTIMES_NULL; 2869 } else { 2870 error = copyin(tptr, tv, sizeof(tv)); 2871 if (error) 2872 goto out; 2873 } 2874 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2875 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2876 vattr.va_atime.tv_sec = tv[0].tv_sec; 2877 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2878 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2879 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2880 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2881 VOP_UNLOCK(vp, 0); 2882 out: 2883 vn_finished_write(mp, 0); 2884 return (error); 2885 } 2886 2887 /* 2888 * Truncate a file given its path name. 2889 */ 2890 /* ARGSUSED */ 2891 int 2892 sys_truncate(l, v, retval) 2893 struct lwp *l; 2894 void *v; 2895 register_t *retval; 2896 { 2897 struct sys_truncate_args /* { 2898 syscallarg(const char *) path; 2899 syscallarg(int) pad; 2900 syscallarg(off_t) length; 2901 } */ *uap = v; 2902 struct proc *p = l->l_proc; 2903 struct vnode *vp; 2904 struct mount *mp; 2905 struct vattr vattr; 2906 int error; 2907 struct nameidata nd; 2908 2909 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2910 if ((error = namei(&nd)) != 0) 2911 return (error); 2912 vp = nd.ni_vp; 2913 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2914 vrele(vp); 2915 return (error); 2916 } 2917 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2918 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2919 if (vp->v_type == VDIR) 2920 error = EISDIR; 2921 else if ((error = vn_writechk(vp)) == 0 && 2922 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2923 VATTR_NULL(&vattr); 2924 vattr.va_size = SCARG(uap, length); 2925 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2926 } 2927 vput(vp); 2928 vn_finished_write(mp, 0); 2929 return (error); 2930 } 2931 2932 /* 2933 * Truncate a file given a file descriptor. 2934 */ 2935 /* ARGSUSED */ 2936 int 2937 sys_ftruncate(l, v, retval) 2938 struct lwp *l; 2939 void *v; 2940 register_t *retval; 2941 { 2942 struct sys_ftruncate_args /* { 2943 syscallarg(int) fd; 2944 syscallarg(int) pad; 2945 syscallarg(off_t) length; 2946 } */ *uap = v; 2947 struct proc *p = l->l_proc; 2948 struct mount *mp; 2949 struct vattr vattr; 2950 struct vnode *vp; 2951 struct file *fp; 2952 int error; 2953 2954 /* getvnode() will use the descriptor for us */ 2955 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2956 return (error); 2957 if ((fp->f_flag & FWRITE) == 0) { 2958 error = EINVAL; 2959 goto out; 2960 } 2961 vp = (struct vnode *)fp->f_data; 2962 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2963 FILE_UNUSE(fp, p); 2964 return (error); 2965 } 2966 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2967 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2968 if (vp->v_type == VDIR) 2969 error = EISDIR; 2970 else if ((error = vn_writechk(vp)) == 0) { 2971 VATTR_NULL(&vattr); 2972 vattr.va_size = SCARG(uap, length); 2973 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2974 } 2975 VOP_UNLOCK(vp, 0); 2976 vn_finished_write(mp, 0); 2977 out: 2978 FILE_UNUSE(fp, p); 2979 return (error); 2980 } 2981 2982 /* 2983 * Sync an open file. 2984 */ 2985 /* ARGSUSED */ 2986 int 2987 sys_fsync(l, v, retval) 2988 struct lwp *l; 2989 void *v; 2990 register_t *retval; 2991 { 2992 struct sys_fsync_args /* { 2993 syscallarg(int) fd; 2994 } */ *uap = v; 2995 struct proc *p = l->l_proc; 2996 struct vnode *vp; 2997 struct mount *mp; 2998 struct file *fp; 2999 int error; 3000 3001 /* getvnode() will use the descriptor for us */ 3002 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3003 return (error); 3004 vp = (struct vnode *)fp->f_data; 3005 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3006 FILE_UNUSE(fp, p); 3007 return (error); 3008 } 3009 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3010 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p); 3011 if (error == 0 && bioops.io_fsync != NULL && 3012 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3013 (*bioops.io_fsync)(vp); 3014 VOP_UNLOCK(vp, 0); 3015 vn_finished_write(mp, 0); 3016 FILE_UNUSE(fp, p); 3017 return (error); 3018 } 3019 3020 /* 3021 * Sync a range of file data. API modeled after that found in AIX. 3022 * 3023 * FDATASYNC indicates that we need only save enough metadata to be able 3024 * to re-read the written data. Note we duplicate AIX's requirement that 3025 * the file be open for writing. 3026 */ 3027 /* ARGSUSED */ 3028 int 3029 sys_fsync_range(l, v, retval) 3030 struct lwp *l; 3031 void *v; 3032 register_t *retval; 3033 { 3034 struct sys_fsync_range_args /* { 3035 syscallarg(int) fd; 3036 syscallarg(int) flags; 3037 syscallarg(off_t) start; 3038 syscallarg(int) length; 3039 } */ *uap = v; 3040 struct proc *p = l->l_proc; 3041 struct vnode *vp; 3042 struct file *fp; 3043 int flags, nflags; 3044 off_t s, e, len; 3045 int error; 3046 3047 /* getvnode() will use the descriptor for us */ 3048 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3049 return (error); 3050 3051 if ((fp->f_flag & FWRITE) == 0) { 3052 FILE_UNUSE(fp, p); 3053 return (EBADF); 3054 } 3055 3056 flags = SCARG(uap, flags); 3057 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3058 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3059 return (EINVAL); 3060 } 3061 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3062 if (flags & FDATASYNC) 3063 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3064 else 3065 nflags = FSYNC_WAIT; 3066 3067 len = SCARG(uap, length); 3068 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3069 if (len) { 3070 s = SCARG(uap, start); 3071 e = s + len; 3072 if (e < s) { 3073 FILE_UNUSE(fp, p); 3074 return (EINVAL); 3075 } 3076 } else { 3077 e = 0; 3078 s = 0; 3079 } 3080 3081 vp = (struct vnode *)fp->f_data; 3082 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3083 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p); 3084 3085 if (error == 0 && bioops.io_fsync != NULL && 3086 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3087 (*bioops.io_fsync)(vp); 3088 3089 VOP_UNLOCK(vp, 0); 3090 FILE_UNUSE(fp, p); 3091 return (error); 3092 } 3093 3094 /* 3095 * Sync the data of an open file. 3096 */ 3097 /* ARGSUSED */ 3098 int 3099 sys_fdatasync(l, v, retval) 3100 struct lwp *l; 3101 void *v; 3102 register_t *retval; 3103 { 3104 struct sys_fdatasync_args /* { 3105 syscallarg(int) fd; 3106 } */ *uap = v; 3107 struct proc *p = l->l_proc; 3108 struct vnode *vp; 3109 struct file *fp; 3110 int error; 3111 3112 /* getvnode() will use the descriptor for us */ 3113 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3114 return (error); 3115 if ((fp->f_flag & FWRITE) == 0) { 3116 FILE_UNUSE(fp, p); 3117 return (EBADF); 3118 } 3119 vp = (struct vnode *)fp->f_data; 3120 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3121 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p); 3122 VOP_UNLOCK(vp, 0); 3123 FILE_UNUSE(fp, p); 3124 return (error); 3125 } 3126 3127 /* 3128 * Rename files, (standard) BSD semantics frontend. 3129 */ 3130 /* ARGSUSED */ 3131 int 3132 sys_rename(l, v, retval) 3133 struct lwp *l; 3134 void *v; 3135 register_t *retval; 3136 { 3137 struct sys_rename_args /* { 3138 syscallarg(const char *) from; 3139 syscallarg(const char *) to; 3140 } */ *uap = v; 3141 struct proc *p = l->l_proc; 3142 3143 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 3144 } 3145 3146 /* 3147 * Rename files, POSIX semantics frontend. 3148 */ 3149 /* ARGSUSED */ 3150 int 3151 sys___posix_rename(l, v, retval) 3152 struct lwp *l; 3153 void *v; 3154 register_t *retval; 3155 { 3156 struct sys___posix_rename_args /* { 3157 syscallarg(const char *) from; 3158 syscallarg(const char *) to; 3159 } */ *uap = v; 3160 struct proc *p = l->l_proc; 3161 3162 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 3163 } 3164 3165 /* 3166 * Rename files. Source and destination must either both be directories, 3167 * or both not be directories. If target is a directory, it must be empty. 3168 * If `from' and `to' refer to the same object, the value of the `retain' 3169 * argument is used to determine whether `from' will be 3170 * 3171 * (retain == 0) deleted unless `from' and `to' refer to the same 3172 * object in the file system's name space (BSD). 3173 * (retain == 1) always retained (POSIX). 3174 */ 3175 static int 3176 rename_files(from, to, p, retain) 3177 const char *from, *to; 3178 struct proc *p; 3179 int retain; 3180 { 3181 struct mount *mp = NULL; 3182 struct vnode *tvp, *fvp, *tdvp; 3183 struct nameidata fromnd, tond; 3184 int error; 3185 3186 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3187 from, p); 3188 if ((error = namei(&fromnd)) != 0) 3189 return (error); 3190 fvp = fromnd.ni_vp; 3191 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3192 if (error != 0) { 3193 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3194 vrele(fromnd.ni_dvp); 3195 vrele(fvp); 3196 if (fromnd.ni_startdir) 3197 vrele(fromnd.ni_startdir); 3198 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3199 return (error); 3200 } 3201 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3202 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p); 3203 if ((error = namei(&tond)) != 0) { 3204 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3205 vrele(fromnd.ni_dvp); 3206 vrele(fvp); 3207 goto out1; 3208 } 3209 tdvp = tond.ni_dvp; 3210 tvp = tond.ni_vp; 3211 3212 if (tvp != NULL) { 3213 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3214 error = ENOTDIR; 3215 goto out; 3216 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3217 error = EISDIR; 3218 goto out; 3219 } 3220 } 3221 3222 if (fvp == tdvp) 3223 error = EINVAL; 3224 3225 /* 3226 * Source and destination refer to the same object. 3227 */ 3228 if (fvp == tvp) { 3229 if (retain) 3230 error = -1; 3231 else if (fromnd.ni_dvp == tdvp && 3232 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3233 !memcmp(fromnd.ni_cnd.cn_nameptr, 3234 tond.ni_cnd.cn_nameptr, 3235 fromnd.ni_cnd.cn_namelen)) 3236 error = -1; 3237 } 3238 3239 out: 3240 if (!error) { 3241 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 3242 if (fromnd.ni_dvp != tdvp) 3243 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3244 if (tvp) { 3245 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 3246 } 3247 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3248 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3249 } else { 3250 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3251 if (tdvp == tvp) 3252 vrele(tdvp); 3253 else 3254 vput(tdvp); 3255 if (tvp) 3256 vput(tvp); 3257 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3258 vrele(fromnd.ni_dvp); 3259 vrele(fvp); 3260 } 3261 vrele(tond.ni_startdir); 3262 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3263 out1: 3264 vn_finished_write(mp, 0); 3265 if (fromnd.ni_startdir) 3266 vrele(fromnd.ni_startdir); 3267 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3268 return (error == -1 ? 0 : error); 3269 } 3270 3271 /* 3272 * Make a directory file. 3273 */ 3274 /* ARGSUSED */ 3275 int 3276 sys_mkdir(l, v, retval) 3277 struct lwp *l; 3278 void *v; 3279 register_t *retval; 3280 { 3281 struct sys_mkdir_args /* { 3282 syscallarg(const char *) path; 3283 syscallarg(int) mode; 3284 } */ *uap = v; 3285 struct proc *p = l->l_proc; 3286 struct mount *mp; 3287 struct vnode *vp; 3288 struct vattr vattr; 3289 int error; 3290 struct nameidata nd; 3291 3292 restart: 3293 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3294 SCARG(uap, path), p); 3295 if ((error = namei(&nd)) != 0) 3296 return (error); 3297 vp = nd.ni_vp; 3298 if (vp != NULL) { 3299 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3300 if (nd.ni_dvp == vp) 3301 vrele(nd.ni_dvp); 3302 else 3303 vput(nd.ni_dvp); 3304 vrele(vp); 3305 return (EEXIST); 3306 } 3307 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3308 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3309 if (nd.ni_dvp == vp) 3310 vrele(nd.ni_dvp); 3311 else 3312 vput(nd.ni_dvp); 3313 if ((error = vn_start_write(NULL, &mp, 3314 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3315 return (error); 3316 goto restart; 3317 } 3318 VATTR_NULL(&vattr); 3319 vattr.va_type = VDIR; 3320 vattr.va_mode = 3321 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3322 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3323 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3324 if (!error) 3325 vput(nd.ni_vp); 3326 vn_finished_write(mp, 0); 3327 return (error); 3328 } 3329 3330 /* 3331 * Remove a directory file. 3332 */ 3333 /* ARGSUSED */ 3334 int 3335 sys_rmdir(l, v, retval) 3336 struct lwp *l; 3337 void *v; 3338 register_t *retval; 3339 { 3340 struct sys_rmdir_args /* { 3341 syscallarg(const char *) path; 3342 } */ *uap = v; 3343 struct proc *p = l->l_proc; 3344 struct mount *mp; 3345 struct vnode *vp; 3346 int error; 3347 struct nameidata nd; 3348 3349 restart: 3350 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3351 SCARG(uap, path), p); 3352 if ((error = namei(&nd)) != 0) 3353 return (error); 3354 vp = nd.ni_vp; 3355 if (vp->v_type != VDIR) { 3356 error = ENOTDIR; 3357 goto out; 3358 } 3359 /* 3360 * No rmdir "." please. 3361 */ 3362 if (nd.ni_dvp == vp) { 3363 error = EINVAL; 3364 goto out; 3365 } 3366 /* 3367 * The root of a mounted filesystem cannot be deleted. 3368 */ 3369 if (vp->v_flag & VROOT) { 3370 error = EBUSY; 3371 goto out; 3372 } 3373 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3374 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3375 if (nd.ni_dvp == vp) 3376 vrele(nd.ni_dvp); 3377 else 3378 vput(nd.ni_dvp); 3379 vput(vp); 3380 if ((error = vn_start_write(NULL, &mp, 3381 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3382 return (error); 3383 goto restart; 3384 } 3385 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3386 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3387 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3388 vn_finished_write(mp, 0); 3389 return (error); 3390 3391 out: 3392 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3393 if (nd.ni_dvp == vp) 3394 vrele(nd.ni_dvp); 3395 else 3396 vput(nd.ni_dvp); 3397 vput(vp); 3398 return (error); 3399 } 3400 3401 /* 3402 * Read a block of directory entries in a file system independent format. 3403 */ 3404 int 3405 sys_getdents(l, v, retval) 3406 struct lwp *l; 3407 void *v; 3408 register_t *retval; 3409 { 3410 struct sys_getdents_args /* { 3411 syscallarg(int) fd; 3412 syscallarg(char *) buf; 3413 syscallarg(size_t) count; 3414 } */ *uap = v; 3415 struct proc *p = l->l_proc; 3416 struct file *fp; 3417 int error, done; 3418 3419 /* getvnode() will use the descriptor for us */ 3420 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3421 return (error); 3422 if ((fp->f_flag & FREAD) == 0) { 3423 error = EBADF; 3424 goto out; 3425 } 3426 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3427 SCARG(uap, count), &done, p, 0, 0); 3428 #ifdef KTRACE 3429 if (!error && KTRPOINT(p, KTR_GENIO)) { 3430 struct iovec iov; 3431 iov.iov_base = SCARG(uap, buf); 3432 iov.iov_len = done; 3433 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3434 } 3435 #endif 3436 *retval = done; 3437 out: 3438 FILE_UNUSE(fp, p); 3439 return (error); 3440 } 3441 3442 /* 3443 * Set the mode mask for creation of filesystem nodes. 3444 */ 3445 int 3446 sys_umask(l, v, retval) 3447 struct lwp *l; 3448 void *v; 3449 register_t *retval; 3450 { 3451 struct sys_umask_args /* { 3452 syscallarg(mode_t) newmask; 3453 } */ *uap = v; 3454 struct proc *p = l->l_proc; 3455 struct cwdinfo *cwdi; 3456 3457 cwdi = p->p_cwdi; 3458 *retval = cwdi->cwdi_cmask; 3459 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3460 return (0); 3461 } 3462 3463 /* 3464 * Void all references to file by ripping underlying filesystem 3465 * away from vnode. 3466 */ 3467 /* ARGSUSED */ 3468 int 3469 sys_revoke(l, v, retval) 3470 struct lwp *l; 3471 void *v; 3472 register_t *retval; 3473 { 3474 struct sys_revoke_args /* { 3475 syscallarg(const char *) path; 3476 } */ *uap = v; 3477 struct proc *p = l->l_proc; 3478 struct mount *mp; 3479 struct vnode *vp; 3480 struct vattr vattr; 3481 int error; 3482 struct nameidata nd; 3483 3484 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3485 if ((error = namei(&nd)) != 0) 3486 return (error); 3487 vp = nd.ni_vp; 3488 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 3489 goto out; 3490 if (p->p_ucred->cr_uid != vattr.va_uid && 3491 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3492 goto out; 3493 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3494 goto out; 3495 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3496 VOP_REVOKE(vp, REVOKEALL); 3497 vn_finished_write(mp, 0); 3498 out: 3499 vrele(vp); 3500 return (error); 3501 } 3502 3503 /* 3504 * Convert a user file descriptor to a kernel file entry. 3505 */ 3506 int 3507 getvnode(fdp, fd, fpp) 3508 struct filedesc *fdp; 3509 int fd; 3510 struct file **fpp; 3511 { 3512 struct vnode *vp; 3513 struct file *fp; 3514 3515 if ((fp = fd_getfile(fdp, fd)) == NULL) 3516 return (EBADF); 3517 3518 FILE_USE(fp); 3519 3520 if (fp->f_type != DTYPE_VNODE) { 3521 FILE_UNUSE(fp, NULL); 3522 return (EINVAL); 3523 } 3524 3525 vp = (struct vnode *)fp->f_data; 3526 if (vp->v_type == VBAD) { 3527 FILE_UNUSE(fp, NULL); 3528 return (EBADF); 3529 } 3530 3531 *fpp = fp; 3532 return (0); 3533 } 3534
Indexes created Fri Sep 26 08:10:20 GMT 2025