vfs_syscalls.c revision 1.237
1 /* $NetBSD: vfs_syscalls.c,v 1.237 2006/02/12 01:32:06 chs 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 1.237 2006/02/12 01:32:06 chs Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_ktrace.h" 45 #include "opt_verified_exec.h" 46 #include "fss.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/dirent.h> 61 #include <sys/sysctl.h> 62 #include <sys/sa.h> 63 #include <sys/syscallargs.h> 64 #ifdef KTRACE 65 #include <sys/ktrace.h> 66 #endif 67 #ifdef VERIFIED_EXEC 68 #include <sys/verified_exec.h> 69 #endif /* VERIFIED_EXEC */ 70 71 #include <miscfs/genfs/genfs.h> 72 #include <miscfs/syncfs/syncfs.h> 73 74 #ifdef COMPAT_30 75 #include "opt_nfsserver.h" 76 #include <nfs/rpcv2.h> 77 #include <nfs/nfsproto.h> 78 #include <nfs/nfs.h> 79 #include <nfs/nfs_var.h> 80 #endif 81 82 #if NFSS > 0 83 #include <dev/fssvar.h> 84 #endif 85 86 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 87 88 static int change_dir(struct nameidata *, struct lwp *); 89 static int change_flags(struct vnode *, u_long, struct lwp *); 90 static int change_mode(struct vnode *, int, struct lwp *l); 91 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 92 static int change_utimes(struct vnode *vp, const struct timeval *, 93 struct lwp *l); 94 static int rename_files(const char *, const char *, struct lwp *, int); 95 96 void checkdirs(struct vnode *); 97 98 int dovfsusermount = 0; 99 100 /* 101 * Virtual File System System Calls 102 */ 103 104 /* 105 * Mount a file system. 106 */ 107 108 #if defined(COMPAT_09) || defined(COMPAT_43) 109 /* 110 * This table is used to maintain compatibility with 4.3BSD 111 * and NetBSD 0.9 mount syscalls. Note, the order is important! 112 * 113 * Do not modify this table. It should only contain filesystems 114 * supported by NetBSD 0.9 and 4.3BSD. 115 */ 116 const char * const mountcompatnames[] = { 117 NULL, /* 0 = MOUNT_NONE */ 118 MOUNT_FFS, /* 1 = MOUNT_UFS */ 119 MOUNT_NFS, /* 2 */ 120 MOUNT_MFS, /* 3 */ 121 MOUNT_MSDOS, /* 4 */ 122 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 123 MOUNT_FDESC, /* 6 */ 124 MOUNT_KERNFS, /* 7 */ 125 NULL, /* 8 = MOUNT_DEVFS */ 126 MOUNT_AFS, /* 9 */ 127 }; 128 const int nmountcompatnames = sizeof(mountcompatnames) / 129 sizeof(mountcompatnames[0]); 130 #endif /* COMPAT_09 || COMPAT_43 */ 131 132 /* ARGSUSED */ 133 int 134 sys_mount(struct lwp *l, void *v, register_t *retval) 135 { 136 struct sys_mount_args /* { 137 syscallarg(const char *) type; 138 syscallarg(const char *) path; 139 syscallarg(int) flags; 140 syscallarg(void *) data; 141 } */ *uap = v; 142 struct proc *p = l->l_proc; 143 struct vnode *vp; 144 struct mount *mp; 145 int error, flag = 0; 146 char fstypename[MFSNAMELEN]; 147 struct vattr va; 148 struct nameidata nd; 149 struct vfsops *vfs; 150 151 /* 152 * if MNT_GETARGS is specified, it should be only flag. 153 */ 154 155 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 156 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 157 return EINVAL; 158 } 159 160 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 161 (error = suser(p->p_ucred, &p->p_acflag))) 162 return (error); 163 /* 164 * Get vnode to be covered 165 */ 166 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 167 SCARG(uap, path), l); 168 if ((error = namei(&nd)) != 0) 169 return (error); 170 vp = nd.ni_vp; 171 /* 172 * A lookup in VFS_MOUNT might result in an attempt to 173 * lock this vnode again, so make the lock recursive. 174 */ 175 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 176 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 177 if ((vp->v_flag & VROOT) == 0) { 178 vput(vp); 179 return (EINVAL); 180 } 181 mp = vp->v_mount; 182 flag = mp->mnt_flag; 183 vfs = mp->mnt_op; 184 /* 185 * We only allow the filesystem to be reloaded if it 186 * is currently mounted read-only. 187 */ 188 if ((SCARG(uap, flags) & MNT_RELOAD) && 189 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 190 vput(vp); 191 return (EOPNOTSUPP); /* Needs translation */ 192 } 193 /* 194 * In "highly secure" mode, don't let the caller do anything 195 * but downgrade a filesystem from read-write to read-only. 196 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 197 */ 198 if (securelevel >= 2 && 199 SCARG(uap, flags) != MNT_GETARGS && 200 SCARG(uap, flags) != 201 (mp->mnt_flag | MNT_RDONLY | 202 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 203 vput(vp); 204 return (EPERM); 205 } 206 mp->mnt_flag |= SCARG(uap, flags) & 207 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 208 /* 209 * Only root, or the user that did the original mount is 210 * permitted to update it. 211 */ 212 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 213 mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 214 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 215 vput(vp); 216 return (error); 217 } 218 /* 219 * Do not allow NFS export by non-root users. For non-root 220 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 221 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 222 */ 223 if (p->p_ucred->cr_uid != 0) { 224 if (SCARG(uap, flags) & MNT_EXPORTED) { 225 vput(vp); 226 return (EPERM); 227 } 228 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 229 if (flag & MNT_NOEXEC) 230 SCARG(uap, flags) |= MNT_NOEXEC; 231 } 232 if (vfs_busy(mp, LK_NOWAIT, 0)) { 233 vput(vp); 234 return (EPERM); 235 } 236 goto update; 237 } else { 238 if (securelevel >= 2) { 239 vput(vp); 240 return (EPERM); 241 } 242 } 243 /* 244 * If the user is not root, ensure that they own the directory 245 * onto which we are attempting to mount. 246 */ 247 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l)) != 0 || 248 (va.va_uid != p->p_ucred->cr_uid && 249 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) { 250 vput(vp); 251 return (error); 252 } 253 /* 254 * Do not allow NFS export by non-root users. For non-root users, 255 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 256 * mount point is already MNT_NOEXEC. 257 */ 258 if (p->p_ucred->cr_uid != 0) { 259 if (SCARG(uap, flags) & MNT_EXPORTED) { 260 vput(vp); 261 return (EPERM); 262 } 263 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 264 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 265 SCARG(uap, flags) |= MNT_NOEXEC; 266 } 267 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, l, 0, 0)) != 0) { 268 vput(vp); 269 return (error); 270 } 271 if (vp->v_type != VDIR) { 272 vput(vp); 273 return (ENOTDIR); 274 } 275 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 276 if (error) { 277 #if defined(COMPAT_09) || defined(COMPAT_43) 278 /* 279 * Historically, filesystem types were identified by numbers. 280 * If we get an integer for the filesystem type instead of a 281 * string, we check to see if it matches one of the historic 282 * filesystem types. 283 */ 284 u_long fsindex = (u_long)SCARG(uap, type); 285 if (fsindex >= nmountcompatnames || 286 mountcompatnames[fsindex] == NULL) { 287 vput(vp); 288 return (ENODEV); 289 } 290 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 291 #else 292 vput(vp); 293 return (error); 294 #endif 295 } 296 #ifdef COMPAT_10 297 /* Accept `ufs' as an alias for `ffs'. */ 298 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 299 strncpy(fstypename, "ffs", MFSNAMELEN); 300 #endif 301 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 302 vput(vp); 303 return (ENODEV); 304 } 305 if (vp->v_mountedhere != NULL) { 306 vput(vp); 307 return (EBUSY); 308 } 309 310 /* 311 * Allocate and initialize the file system. 312 */ 313 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 314 M_MOUNT, M_WAITOK); 315 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 316 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 317 simple_lock_init(&mp->mnt_slock); 318 (void)vfs_busy(mp, LK_NOWAIT, 0); 319 mp->mnt_op = vfs; 320 vfs->vfs_refcount++; 321 mp->mnt_vnodecovered = vp; 322 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 323 mp->mnt_unmounter = NULL; 324 mp->mnt_leaf = mp; 325 326 /* 327 * The underlying file system may refuse the mount for 328 * various reasons. Allow the user to force it to happen. 329 */ 330 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 331 update: 332 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 333 /* 334 * Set the mount level flags. 335 */ 336 if (SCARG(uap, flags) & MNT_RDONLY) 337 mp->mnt_flag |= MNT_RDONLY; 338 else if (mp->mnt_flag & MNT_RDONLY) 339 mp->mnt_iflag |= IMNT_WANTRDWR; 340 mp->mnt_flag &= 341 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 342 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 343 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 344 mp->mnt_flag |= SCARG(uap, flags) & 345 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 346 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 347 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 348 MNT_IGNORE); 349 } 350 /* 351 * Mount the filesystem. 352 */ 353 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 354 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 355 #if defined(COMPAT_30) && defined(NFSSERVER) 356 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 357 int error2; 358 359 /* Update failed; let's try and see if it was an 360 * export request. */ 361 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 362 SCARG(uap, data), l); 363 364 /* Only update error code if the export request was 365 * understood but some problem occurred while 366 * processing it. */ 367 if (error2 != EJUSTRETURN) 368 error = error2; 369 } 370 #endif 371 if (mp->mnt_iflag & IMNT_WANTRDWR) 372 mp->mnt_flag &= ~MNT_RDONLY; 373 if (error) 374 mp->mnt_flag = flag; 375 mp->mnt_flag &=~ 376 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 377 mp->mnt_iflag &=~ IMNT_WANTRDWR; 378 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 379 if (mp->mnt_syncer == NULL) 380 error = vfs_allocate_syncvnode(mp); 381 } else { 382 if (mp->mnt_syncer != NULL) 383 vfs_deallocate_syncvnode(mp); 384 } 385 vfs_unbusy(mp); 386 VOP_UNLOCK(vp, 0); 387 vrele(vp); 388 return (error); 389 } 390 /* 391 * Put the new filesystem on the mount list after root. 392 */ 393 cache_purge(vp); 394 if (!error) { 395 mp->mnt_flag &=~ 396 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 397 mp->mnt_iflag &=~ IMNT_WANTRDWR; 398 vp->v_mountedhere = mp; 399 simple_lock(&mountlist_slock); 400 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 401 simple_unlock(&mountlist_slock); 402 checkdirs(vp); 403 VOP_UNLOCK(vp, 0); 404 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 405 error = vfs_allocate_syncvnode(mp); 406 vfs_unbusy(mp); 407 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 408 if ((error = VFS_START(mp, 0, l))) 409 vrele(vp); 410 } else { 411 vp->v_mountedhere = (struct mount *)0; 412 vfs->vfs_refcount--; 413 vfs_unbusy(mp); 414 free(mp, M_MOUNT); 415 vput(vp); 416 } 417 return (error); 418 } 419 420 /* 421 * Scan all active processes to see if any of them have a current 422 * or root directory onto which the new filesystem has just been 423 * mounted. If so, replace them with the new mount point. 424 */ 425 void 426 checkdirs(struct vnode *olddp) 427 { 428 struct cwdinfo *cwdi; 429 struct vnode *newdp; 430 struct proc *p; 431 432 if (olddp->v_usecount == 1) 433 return; 434 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 435 panic("mount: lost mount"); 436 proclist_lock_read(); 437 PROCLIST_FOREACH(p, &allproc) { 438 cwdi = p->p_cwdi; 439 if (!cwdi) 440 continue; 441 if (cwdi->cwdi_cdir == olddp) { 442 vrele(cwdi->cwdi_cdir); 443 VREF(newdp); 444 cwdi->cwdi_cdir = newdp; 445 } 446 if (cwdi->cwdi_rdir == olddp) { 447 vrele(cwdi->cwdi_rdir); 448 VREF(newdp); 449 cwdi->cwdi_rdir = newdp; 450 } 451 } 452 proclist_unlock_read(); 453 if (rootvnode == olddp) { 454 vrele(rootvnode); 455 VREF(newdp); 456 rootvnode = newdp; 457 } 458 vput(newdp); 459 } 460 461 /* 462 * Unmount a file system. 463 * 464 * Note: unmount takes a path to the vnode mounted on as argument, 465 * not special file (as before). 466 */ 467 /* ARGSUSED */ 468 int 469 sys_unmount(struct lwp *l, void *v, register_t *retval) 470 { 471 struct sys_unmount_args /* { 472 syscallarg(const char *) path; 473 syscallarg(int) flags; 474 } */ *uap = v; 475 struct proc *p = l->l_proc; 476 struct vnode *vp; 477 struct mount *mp; 478 int error; 479 struct nameidata nd; 480 481 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 482 SCARG(uap, path), l); 483 if ((error = namei(&nd)) != 0) 484 return (error); 485 vp = nd.ni_vp; 486 mp = vp->v_mount; 487 488 /* 489 * Only root, or the user that did the original mount is 490 * permitted to unmount this filesystem. 491 */ 492 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 493 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 494 vput(vp); 495 return (error); 496 } 497 498 /* 499 * Don't allow unmounting the root file system. 500 */ 501 if (mp->mnt_flag & MNT_ROOTFS) { 502 vput(vp); 503 return (EINVAL); 504 } 505 506 /* 507 * Must be the root of the filesystem 508 */ 509 if ((vp->v_flag & VROOT) == 0) { 510 vput(vp); 511 return (EINVAL); 512 } 513 vput(vp); 514 515 /* 516 * XXX Freeze syncer. Must do this before locking the 517 * mount point. See dounmount() for details. 518 */ 519 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 520 521 if (vfs_busy(mp, 0, 0)) { 522 lockmgr(&syncer_lock, LK_RELEASE, NULL); 523 return (EBUSY); 524 } 525 526 return (dounmount(mp, SCARG(uap, flags), l)); 527 } 528 529 /* 530 * Do the actual file system unmount. File system is assumed to have been 531 * marked busy by the caller. 532 */ 533 int 534 dounmount(struct mount *mp, int flags, struct lwp *l) 535 { 536 struct vnode *coveredvp; 537 int error; 538 int async; 539 int used_syncer; 540 541 simple_lock(&mountlist_slock); 542 vfs_unbusy(mp); 543 used_syncer = (mp->mnt_syncer != NULL); 544 545 /* 546 * XXX Syncer must be frozen when we get here. This should really 547 * be done on a per-mountpoint basis, but especially the softdep 548 * code possibly called from the syncer doesn't exactly work on a 549 * per-mountpoint basis, so the softdep code would become a maze 550 * of vfs_busy() calls. 551 * 552 * The caller of dounmount() must acquire syncer_lock because 553 * the syncer itself acquires locks in syncer_lock -> vfs_busy 554 * order, and we must preserve that order to avoid deadlock. 555 * 556 * So, if the file system did not use the syncer, now is 557 * the time to release the syncer_lock. 558 */ 559 if (used_syncer == 0) 560 lockmgr(&syncer_lock, LK_RELEASE, NULL); 561 562 mp->mnt_iflag |= IMNT_UNMOUNT; 563 mp->mnt_unmounter = l; 564 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 565 vn_start_write(NULL, &mp, V_WAIT); 566 567 async = mp->mnt_flag & MNT_ASYNC; 568 mp->mnt_flag &= ~MNT_ASYNC; 569 cache_purgevfs(mp); /* remove cache entries for this file sys */ 570 if (mp->mnt_syncer != NULL) 571 vfs_deallocate_syncvnode(mp); 572 error = 0; 573 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 574 #if NFSS > 0 575 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 576 #endif 577 if (error == 0) 578 error = VFS_SYNC(mp, MNT_WAIT, l->l_proc->p_ucred, l); 579 } 580 if (error == 0 || (flags & MNT_FORCE)) 581 error = VFS_UNMOUNT(mp, flags, l); 582 vn_finished_write(mp, 0); 583 simple_lock(&mountlist_slock); 584 if (error) { 585 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 586 (void) vfs_allocate_syncvnode(mp); 587 mp->mnt_iflag &= ~IMNT_UNMOUNT; 588 mp->mnt_unmounter = NULL; 589 mp->mnt_flag |= async; 590 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 591 &mountlist_slock); 592 if (used_syncer) 593 lockmgr(&syncer_lock, LK_RELEASE, NULL); 594 simple_lock(&mp->mnt_slock); 595 while (mp->mnt_wcnt > 0) { 596 wakeup(mp); 597 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 598 0, &mp->mnt_slock); 599 } 600 simple_unlock(&mp->mnt_slock); 601 return (error); 602 } 603 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 604 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 605 coveredvp->v_mountedhere = NULL; 606 vrele(coveredvp); 607 } 608 mp->mnt_op->vfs_refcount--; 609 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 610 panic("unmount: dangling vnode"); 611 mp->mnt_iflag |= IMNT_GONE; 612 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 613 if (used_syncer) 614 lockmgr(&syncer_lock, LK_RELEASE, NULL); 615 simple_lock(&mp->mnt_slock); 616 while (mp->mnt_wcnt > 0) { 617 wakeup(mp); 618 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 619 } 620 simple_unlock(&mp->mnt_slock); 621 vfs_hooks_unmount(mp); 622 free(mp, M_MOUNT); 623 return (0); 624 } 625 626 /* 627 * Sync each mounted filesystem. 628 */ 629 #ifdef DEBUG 630 int syncprt = 0; 631 struct ctldebug debug0 = { "syncprt", &syncprt }; 632 #endif 633 634 /* ARGSUSED */ 635 int 636 sys_sync(struct lwp *l, void *v, register_t *retval) 637 { 638 struct mount *mp, *nmp; 639 int asyncflag; 640 struct proc *p = l == NULL ? &proc0 : l->l_proc; 641 642 simple_lock(&mountlist_slock); 643 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 644 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 645 nmp = mp->mnt_list.cqe_prev; 646 continue; 647 } 648 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 649 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 650 asyncflag = mp->mnt_flag & MNT_ASYNC; 651 mp->mnt_flag &= ~MNT_ASYNC; 652 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, l); 653 if (asyncflag) 654 mp->mnt_flag |= MNT_ASYNC; 655 vn_finished_write(mp, 0); 656 } 657 simple_lock(&mountlist_slock); 658 nmp = mp->mnt_list.cqe_prev; 659 vfs_unbusy(mp); 660 661 } 662 simple_unlock(&mountlist_slock); 663 #ifdef DEBUG 664 if (syncprt) 665 vfs_bufstats(); 666 #endif /* DEBUG */ 667 return (0); 668 } 669 670 /* 671 * Change filesystem quotas. 672 */ 673 /* ARGSUSED */ 674 int 675 sys_quotactl(struct lwp *l, void *v, register_t *retval) 676 { 677 struct sys_quotactl_args /* { 678 syscallarg(const char *) path; 679 syscallarg(int) cmd; 680 syscallarg(int) uid; 681 syscallarg(caddr_t) arg; 682 } */ *uap = v; 683 struct mount *mp; 684 int error; 685 struct nameidata nd; 686 687 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 688 if ((error = namei(&nd)) != 0) 689 return (error); 690 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 691 vrele(nd.ni_vp); 692 if (error) 693 return (error); 694 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 695 SCARG(uap, arg), l); 696 vn_finished_write(mp, 0); 697 return (error); 698 } 699 700 int 701 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 702 int root) 703 { 704 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 705 int error = 0; 706 707 /* 708 * If MNT_NOWAIT or MNT_LAZY is specified, do not 709 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 710 * overrides MNT_NOWAIT. 711 */ 712 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 713 (flags != MNT_WAIT && flags != 0)) { 714 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 715 goto done; 716 } 717 718 /* Get the filesystem stats now */ 719 memset(sp, 0, sizeof(*sp)); 720 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 721 return error; 722 } 723 724 if (cwdi->cwdi_rdir == NULL) 725 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 726 done: 727 if (cwdi->cwdi_rdir != NULL) { 728 size_t len; 729 char *bp; 730 char *path = PNBUF_GET(); 731 if (!path) 732 return ENOMEM; 733 734 bp = path + MAXPATHLEN; 735 *--bp = '\0'; 736 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 737 MAXPATHLEN / 2, 0, l); 738 if (error) { 739 PNBUF_PUT(path); 740 return error; 741 } 742 len = strlen(bp); 743 /* 744 * for mount points that are below our root, we can see 745 * them, so we fix up the pathname and return them. The 746 * rest we cannot see, so we don't allow viewing the 747 * data. 748 */ 749 if (strncmp(bp, sp->f_mntonname, len) == 0) { 750 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 751 sizeof(sp->f_mntonname)); 752 if (sp->f_mntonname[0] == '\0') 753 (void)strlcpy(sp->f_mntonname, "/", 754 sizeof(sp->f_mntonname)); 755 } else { 756 if (root) 757 (void)strlcpy(sp->f_mntonname, "/", 758 sizeof(sp->f_mntonname)); 759 else 760 error = EPERM; 761 } 762 PNBUF_PUT(path); 763 } 764 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 765 return error; 766 } 767 768 /* 769 * Get filesystem statistics. 770 */ 771 /* ARGSUSED */ 772 int 773 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 774 { 775 struct sys_statvfs1_args /* { 776 syscallarg(const char *) path; 777 syscallarg(struct statvfs *) buf; 778 syscallarg(int) flags; 779 } */ *uap = v; 780 struct mount *mp; 781 struct statvfs sbuf; 782 int error; 783 struct nameidata nd; 784 785 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 786 if ((error = namei(&nd)) != 0) 787 return error; 788 mp = nd.ni_vp->v_mount; 789 vrele(nd.ni_vp); 790 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 791 return error; 792 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 793 } 794 795 /* 796 * Get filesystem statistics. 797 */ 798 /* ARGSUSED */ 799 int 800 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 801 { 802 struct sys_fstatvfs1_args /* { 803 syscallarg(int) fd; 804 syscallarg(struct statvfs *) buf; 805 syscallarg(int) flags; 806 } */ *uap = v; 807 struct proc *p = l->l_proc; 808 struct file *fp; 809 struct mount *mp; 810 struct statvfs sbuf; 811 int error; 812 813 /* getvnode() will use the descriptor for us */ 814 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 815 return (error); 816 mp = ((struct vnode *)fp->f_data)->v_mount; 817 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 818 goto out; 819 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 820 out: 821 FILE_UNUSE(fp, l); 822 return error; 823 } 824 825 826 /* 827 * Get statistics on all filesystems. 828 */ 829 int 830 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 831 { 832 struct sys_getvfsstat_args /* { 833 syscallarg(struct statvfs *) buf; 834 syscallarg(size_t) bufsize; 835 syscallarg(int) flags; 836 } */ *uap = v; 837 int root = 0; 838 struct proc *p = l->l_proc; 839 struct mount *mp, *nmp; 840 struct statvfs sbuf; 841 struct statvfs *sfsp; 842 size_t count, maxcount; 843 int error = 0; 844 845 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 846 sfsp = SCARG(uap, buf); 847 simple_lock(&mountlist_slock); 848 count = 0; 849 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 850 mp = nmp) { 851 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 852 nmp = CIRCLEQ_NEXT(mp, mnt_list); 853 continue; 854 } 855 if (sfsp && count < maxcount) { 856 error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 0); 857 if (error) { 858 simple_lock(&mountlist_slock); 859 nmp = CIRCLEQ_NEXT(mp, mnt_list); 860 vfs_unbusy(mp); 861 continue; 862 } 863 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 864 if (error) { 865 vfs_unbusy(mp); 866 return (error); 867 } 868 sfsp++; 869 root |= strcmp(sbuf.f_mntonname, "/") == 0; 870 } 871 count++; 872 simple_lock(&mountlist_slock); 873 nmp = CIRCLEQ_NEXT(mp, mnt_list); 874 vfs_unbusy(mp); 875 } 876 simple_unlock(&mountlist_slock); 877 if (root == 0 && p->p_cwdi->cwdi_rdir) { 878 /* 879 * fake a root entry 880 */ 881 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, l, 882 SCARG(uap, flags), 1)) != 0) 883 return error; 884 if (sfsp) 885 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 886 count++; 887 } 888 if (sfsp && count > maxcount) 889 *retval = maxcount; 890 else 891 *retval = count; 892 return error; 893 } 894 895 /* 896 * Change current working directory to a given file descriptor. 897 */ 898 /* ARGSUSED */ 899 int 900 sys_fchdir(struct lwp *l, void *v, register_t *retval) 901 { 902 struct sys_fchdir_args /* { 903 syscallarg(int) fd; 904 } */ *uap = v; 905 struct proc *p = l->l_proc; 906 struct filedesc *fdp = p->p_fd; 907 struct cwdinfo *cwdi = p->p_cwdi; 908 struct vnode *vp, *tdp; 909 struct mount *mp; 910 struct file *fp; 911 int error; 912 913 /* getvnode() will use the descriptor for us */ 914 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 915 return (error); 916 vp = (struct vnode *)fp->f_data; 917 918 VREF(vp); 919 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 920 if (vp->v_type != VDIR) 921 error = ENOTDIR; 922 else 923 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, l); 924 while (!error && (mp = vp->v_mountedhere) != NULL) { 925 if (vfs_busy(mp, 0, 0)) 926 continue; 927 error = VFS_ROOT(mp, &tdp); 928 vfs_unbusy(mp); 929 if (error) 930 break; 931 vput(vp); 932 vp = tdp; 933 } 934 if (error) { 935 vput(vp); 936 goto out; 937 } 938 VOP_UNLOCK(vp, 0); 939 940 /* 941 * Disallow changing to a directory not under the process's 942 * current root directory (if there is one). 943 */ 944 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 945 vrele(vp); 946 error = EPERM; /* operation not permitted */ 947 goto out; 948 } 949 950 vrele(cwdi->cwdi_cdir); 951 cwdi->cwdi_cdir = vp; 952 out: 953 FILE_UNUSE(fp, l); 954 return (error); 955 } 956 957 /* 958 * Change this process's notion of the root directory to a given file 959 * descriptor. 960 */ 961 int 962 sys_fchroot(struct lwp *l, void *v, register_t *retval) 963 { 964 struct sys_fchroot_args *uap = v; 965 struct proc *p = l->l_proc; 966 struct filedesc *fdp = p->p_fd; 967 struct cwdinfo *cwdi = p->p_cwdi; 968 struct vnode *vp; 969 struct file *fp; 970 int error; 971 972 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 973 return error; 974 /* getvnode() will use the descriptor for us */ 975 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 976 return error; 977 vp = (struct vnode *) fp->f_data; 978 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 979 if (vp->v_type != VDIR) 980 error = ENOTDIR; 981 else 982 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, l); 983 VOP_UNLOCK(vp, 0); 984 if (error) 985 goto out; 986 VREF(vp); 987 988 /* 989 * Prevent escaping from chroot by putting the root under 990 * the working directory. Silently chdir to / if we aren't 991 * already there. 992 */ 993 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 994 /* 995 * XXX would be more failsafe to change directory to a 996 * deadfs node here instead 997 */ 998 vrele(cwdi->cwdi_cdir); 999 VREF(vp); 1000 cwdi->cwdi_cdir = vp; 1001 } 1002 1003 if (cwdi->cwdi_rdir != NULL) 1004 vrele(cwdi->cwdi_rdir); 1005 cwdi->cwdi_rdir = vp; 1006 out: 1007 FILE_UNUSE(fp, l); 1008 return (error); 1009 } 1010 1011 /* 1012 * Change current working directory (``.''). 1013 */ 1014 /* ARGSUSED */ 1015 int 1016 sys_chdir(struct lwp *l, void *v, register_t *retval) 1017 { 1018 struct sys_chdir_args /* { 1019 syscallarg(const char *) path; 1020 } */ *uap = v; 1021 struct proc *p = l->l_proc; 1022 struct cwdinfo *cwdi = p->p_cwdi; 1023 int error; 1024 struct nameidata nd; 1025 1026 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1027 SCARG(uap, path), l); 1028 if ((error = change_dir(&nd, l)) != 0) 1029 return (error); 1030 vrele(cwdi->cwdi_cdir); 1031 cwdi->cwdi_cdir = nd.ni_vp; 1032 return (0); 1033 } 1034 1035 /* 1036 * Change notion of root (``/'') directory. 1037 */ 1038 /* ARGSUSED */ 1039 int 1040 sys_chroot(struct lwp *l, void *v, register_t *retval) 1041 { 1042 struct sys_chroot_args /* { 1043 syscallarg(const char *) path; 1044 } */ *uap = v; 1045 struct proc *p = l->l_proc; 1046 struct cwdinfo *cwdi = p->p_cwdi; 1047 struct vnode *vp; 1048 int error; 1049 struct nameidata nd; 1050 1051 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1052 return (error); 1053 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1054 SCARG(uap, path), l); 1055 if ((error = change_dir(&nd, l)) != 0) 1056 return (error); 1057 if (cwdi->cwdi_rdir != NULL) 1058 vrele(cwdi->cwdi_rdir); 1059 vp = nd.ni_vp; 1060 cwdi->cwdi_rdir = vp; 1061 1062 /* 1063 * Prevent escaping from chroot by putting the root under 1064 * the working directory. Silently chdir to / if we aren't 1065 * already there. 1066 */ 1067 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1068 /* 1069 * XXX would be more failsafe to change directory to a 1070 * deadfs node here instead 1071 */ 1072 vrele(cwdi->cwdi_cdir); 1073 VREF(vp); 1074 cwdi->cwdi_cdir = vp; 1075 } 1076 1077 return (0); 1078 } 1079 1080 /* 1081 * Common routine for chroot and chdir. 1082 */ 1083 static int 1084 change_dir(struct nameidata *ndp, struct lwp *l) 1085 { 1086 struct vnode *vp; 1087 int error; 1088 1089 if ((error = namei(ndp)) != 0) 1090 return (error); 1091 vp = ndp->ni_vp; 1092 if (vp->v_type != VDIR) 1093 error = ENOTDIR; 1094 else 1095 error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_ucred, l); 1096 1097 if (error) 1098 vput(vp); 1099 else 1100 VOP_UNLOCK(vp, 0); 1101 return (error); 1102 } 1103 1104 /* 1105 * Check permissions, allocate an open file structure, 1106 * and call the device open routine if any. 1107 */ 1108 int 1109 sys_open(struct lwp *l, void *v, 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), l); 1134 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1135 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1136 FILE_UNUSE(fp, l); 1137 fdp->fd_ofiles[indx] = NULL; 1138 ffree(fp); 1139 if ((error == EDUPFD || error == EMOVEFD) && 1140 l->l_dupfd >= 0 && /* XXX from fdopen */ 1141 (error = 1142 dupfdopen(l, 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, l); 1172 FILE_UNUSE(fp, l); 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, l); 1184 return (0); 1185 } 1186 1187 /* 1188 * Get file handle system call 1189 */ 1190 int 1191 sys_getfh(struct lwp *l, void *v, register_t *retval) 1192 { 1193 struct sys_getfh_args /* { 1194 syscallarg(char *) fname; 1195 syscallarg(fhandle_t *) fhp; 1196 } */ *uap = v; 1197 struct proc *p = l->l_proc; 1198 struct vnode *vp; 1199 fhandle_t fh; 1200 int error; 1201 struct nameidata nd; 1202 1203 /* 1204 * Must be super user 1205 */ 1206 error = suser(p->p_ucred, &p->p_acflag); 1207 if (error) 1208 return (error); 1209 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1210 SCARG(uap, fname), l); 1211 error = namei(&nd); 1212 if (error) 1213 return (error); 1214 vp = nd.ni_vp; 1215 if (vp->v_mount->mnt_op->vfs_vptofh == NULL) 1216 return EOPNOTSUPP; 1217 memset(&fh, 0, sizeof(fh)); 1218 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1219 error = VFS_VPTOFH(vp, &fh.fh_fid); 1220 vput(vp); 1221 if (error) 1222 return (error); 1223 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1224 return (error); 1225 } 1226 1227 /* 1228 * Open a file given a file handle. 1229 * 1230 * Check permissions, allocate an open file structure, 1231 * and call the device open routine if any. 1232 */ 1233 int 1234 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1235 { 1236 struct sys_fhopen_args /* { 1237 syscallarg(const fhandle_t *) fhp; 1238 syscallarg(int) flags; 1239 } */ *uap = v; 1240 struct proc *p = l->l_proc; 1241 struct filedesc *fdp = p->p_fd; 1242 struct file *fp; 1243 struct vnode *vp = NULL; 1244 struct mount *mp; 1245 struct ucred *cred = p->p_ucred; 1246 int flags; 1247 struct file *nfp; 1248 int type, indx, error=0; 1249 struct flock lf; 1250 struct vattr va; 1251 fhandle_t fh; 1252 1253 /* 1254 * Must be super user 1255 */ 1256 if ((error = suser(p->p_ucred, &p->p_acflag))) 1257 return (error); 1258 1259 flags = FFLAGS(SCARG(uap, flags)); 1260 if ((flags & (FREAD | FWRITE)) == 0) 1261 return (EINVAL); 1262 if ((flags & O_CREAT)) 1263 return (EINVAL); 1264 /* falloc() will use the file descriptor for us */ 1265 if ((error = falloc(p, &nfp, &indx)) != 0) 1266 return (error); 1267 fp = nfp; 1268 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1269 goto bad; 1270 1271 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1272 error = ESTALE; 1273 goto bad; 1274 } 1275 1276 if (mp->mnt_op->vfs_fhtovp == NULL) { 1277 error = EOPNOTSUPP; 1278 goto bad; 1279 } 1280 1281 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1282 vp = NULL; /* most likely unnecessary sanity for bad: */ 1283 goto bad; 1284 } 1285 1286 /* Now do an effective vn_open */ 1287 1288 if (vp->v_type == VSOCK) { 1289 error = EOPNOTSUPP; 1290 goto bad; 1291 } 1292 if (flags & FREAD) { 1293 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1294 goto bad; 1295 } 1296 if (flags & (FWRITE | O_TRUNC)) { 1297 if (vp->v_type == VDIR) { 1298 error = EISDIR; 1299 goto bad; 1300 } 1301 if ((error = vn_writechk(vp)) != 0 || 1302 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1303 goto bad; 1304 } 1305 if (flags & O_TRUNC) { 1306 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1307 goto bad; 1308 VOP_UNLOCK(vp, 0); /* XXX */ 1309 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1310 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1311 VATTR_NULL(&va); 1312 va.va_size = 0; 1313 error = VOP_SETATTR(vp, &va, cred, l); 1314 vn_finished_write(mp, 0); 1315 if (error) 1316 goto bad; 1317 } 1318 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1319 goto bad; 1320 if (vp->v_type == VREG && 1321 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1322 error = EIO; 1323 goto bad; 1324 } 1325 if (flags & FWRITE) 1326 vp->v_writecount++; 1327 1328 /* done with modified vn_open, now finish what sys_open does. */ 1329 1330 fp->f_flag = flags & FMASK; 1331 fp->f_type = DTYPE_VNODE; 1332 fp->f_ops = &vnops; 1333 fp->f_data = vp; 1334 if (flags & (O_EXLOCK | O_SHLOCK)) { 1335 lf.l_whence = SEEK_SET; 1336 lf.l_start = 0; 1337 lf.l_len = 0; 1338 if (flags & O_EXLOCK) 1339 lf.l_type = F_WRLCK; 1340 else 1341 lf.l_type = F_RDLCK; 1342 type = F_FLOCK; 1343 if ((flags & FNONBLOCK) == 0) 1344 type |= F_WAIT; 1345 VOP_UNLOCK(vp, 0); 1346 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1347 if (error) { 1348 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1349 FILE_UNUSE(fp, l); 1350 ffree(fp); 1351 fdremove(fdp, indx); 1352 return (error); 1353 } 1354 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1355 fp->f_flag |= FHASLOCK; 1356 } 1357 VOP_UNLOCK(vp, 0); 1358 *retval = indx; 1359 FILE_SET_MATURE(fp); 1360 FILE_UNUSE(fp, l); 1361 return (0); 1362 1363 bad: 1364 FILE_UNUSE(fp, l); 1365 ffree(fp); 1366 fdremove(fdp, indx); 1367 if (vp != NULL) 1368 vput(vp); 1369 return (error); 1370 } 1371 1372 /* ARGSUSED */ 1373 int 1374 sys_fhstat(struct lwp *l, void *v, register_t *retval) 1375 { 1376 struct sys_fhstat_args /* { 1377 syscallarg(const fhandle_t *) fhp; 1378 syscallarg(struct stat *) sb; 1379 } */ *uap = v; 1380 struct proc *p = l->l_proc; 1381 struct stat sb; 1382 int error; 1383 fhandle_t fh; 1384 struct mount *mp; 1385 struct vnode *vp; 1386 1387 /* 1388 * Must be super user 1389 */ 1390 if ((error = suser(p->p_ucred, &p->p_acflag))) 1391 return (error); 1392 1393 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1394 return (error); 1395 1396 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1397 return (ESTALE); 1398 if (mp->mnt_op->vfs_fhtovp == NULL) 1399 return EOPNOTSUPP; 1400 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1401 return (error); 1402 error = vn_stat(vp, &sb, l); 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(struct lwp *l, void *v, register_t *retval) 1413 { 1414 struct sys_fhstatvfs1_args /* { 1415 syscallarg(const fhandle_t *) fhp; 1416 syscallarg(struct statvfs *) buf; 1417 syscallarg(int) flags; 1418 } */ *uap = v; 1419 struct proc *p = l->l_proc; 1420 struct statvfs sbuf; 1421 fhandle_t fh; 1422 struct mount *mp; 1423 struct vnode *vp; 1424 int error; 1425 1426 /* 1427 * Must be super user 1428 */ 1429 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1430 return error; 1431 1432 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1433 return error; 1434 1435 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1436 return ESTALE; 1437 if (mp->mnt_op->vfs_fhtovp == NULL) 1438 return EOPNOTSUPP; 1439 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1440 return error; 1441 1442 mp = vp->v_mount; 1443 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) { 1444 vput(vp); 1445 return error; 1446 } 1447 vput(vp); 1448 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1449 } 1450 1451 /* 1452 * Create a special file. 1453 */ 1454 /* ARGSUSED */ 1455 int 1456 sys_mknod(struct lwp *l, void *v, register_t *retval) 1457 { 1458 struct sys_mknod_args /* { 1459 syscallarg(const char *) path; 1460 syscallarg(int) mode; 1461 syscallarg(int) dev; 1462 } */ *uap = v; 1463 struct proc *p = l->l_proc; 1464 struct vnode *vp; 1465 struct mount *mp; 1466 struct vattr vattr; 1467 int error; 1468 int whiteout = 0; 1469 struct nameidata nd; 1470 1471 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1472 return (error); 1473 restart: 1474 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1475 if ((error = namei(&nd)) != 0) 1476 return (error); 1477 vp = nd.ni_vp; 1478 if (vp != NULL) 1479 error = EEXIST; 1480 else { 1481 VATTR_NULL(&vattr); 1482 vattr.va_mode = 1483 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1484 vattr.va_rdev = SCARG(uap, dev); 1485 whiteout = 0; 1486 1487 switch (SCARG(uap, mode) & S_IFMT) { 1488 case S_IFMT: /* used by badsect to flag bad sectors */ 1489 vattr.va_type = VBAD; 1490 break; 1491 case S_IFCHR: 1492 vattr.va_type = VCHR; 1493 break; 1494 case S_IFBLK: 1495 vattr.va_type = VBLK; 1496 break; 1497 case S_IFWHT: 1498 whiteout = 1; 1499 break; 1500 default: 1501 error = EINVAL; 1502 break; 1503 } 1504 } 1505 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1506 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1507 if (nd.ni_dvp == vp) 1508 vrele(nd.ni_dvp); 1509 else 1510 vput(nd.ni_dvp); 1511 if (vp) 1512 vrele(vp); 1513 if ((error = vn_start_write(NULL, &mp, 1514 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1515 return (error); 1516 goto restart; 1517 } 1518 if (!error) { 1519 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1520 if (whiteout) { 1521 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1522 if (error) 1523 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1524 vput(nd.ni_dvp); 1525 } else { 1526 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1527 &nd.ni_cnd, &vattr); 1528 if (error == 0) 1529 vput(nd.ni_vp); 1530 } 1531 } else { 1532 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1533 if (nd.ni_dvp == vp) 1534 vrele(nd.ni_dvp); 1535 else 1536 vput(nd.ni_dvp); 1537 if (vp) 1538 vrele(vp); 1539 } 1540 vn_finished_write(mp, 0); 1541 return (error); 1542 } 1543 1544 /* 1545 * Create a named pipe. 1546 */ 1547 /* ARGSUSED */ 1548 int 1549 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1550 { 1551 struct sys_mkfifo_args /* { 1552 syscallarg(const char *) path; 1553 syscallarg(int) mode; 1554 } */ *uap = v; 1555 struct proc *p = l->l_proc; 1556 struct mount *mp; 1557 struct vattr vattr; 1558 int error; 1559 struct nameidata nd; 1560 1561 restart: 1562 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1563 if ((error = namei(&nd)) != 0) 1564 return (error); 1565 if (nd.ni_vp != NULL) { 1566 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1567 if (nd.ni_dvp == nd.ni_vp) 1568 vrele(nd.ni_dvp); 1569 else 1570 vput(nd.ni_dvp); 1571 vrele(nd.ni_vp); 1572 return (EEXIST); 1573 } 1574 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1575 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1576 if (nd.ni_dvp == nd.ni_vp) 1577 vrele(nd.ni_dvp); 1578 else 1579 vput(nd.ni_dvp); 1580 if (nd.ni_vp) 1581 vrele(nd.ni_vp); 1582 if ((error = vn_start_write(NULL, &mp, 1583 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1584 return (error); 1585 goto restart; 1586 } 1587 VATTR_NULL(&vattr); 1588 vattr.va_type = VFIFO; 1589 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1590 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1591 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1592 if (error == 0) 1593 vput(nd.ni_vp); 1594 vn_finished_write(mp, 0); 1595 return (error); 1596 } 1597 1598 /* 1599 * Make a hard file link. 1600 */ 1601 /* ARGSUSED */ 1602 int 1603 sys_link(struct lwp *l, void *v, register_t *retval) 1604 { 1605 struct sys_link_args /* { 1606 syscallarg(const char *) path; 1607 syscallarg(const char *) link; 1608 } */ *uap = v; 1609 struct proc *p = l->l_proc; 1610 struct vnode *vp; 1611 struct mount *mp; 1612 struct nameidata nd; 1613 int error; 1614 1615 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1616 if ((error = namei(&nd)) != 0) 1617 return (error); 1618 vp = nd.ni_vp; 1619 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1620 vrele(vp); 1621 return (error); 1622 } 1623 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1624 if ((error = namei(&nd)) != 0) 1625 goto out; 1626 if (nd.ni_vp) { 1627 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1628 if (nd.ni_dvp == nd.ni_vp) 1629 vrele(nd.ni_dvp); 1630 else 1631 vput(nd.ni_dvp); 1632 vrele(nd.ni_vp); 1633 error = EEXIST; 1634 goto out; 1635 } 1636 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1637 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 1638 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1639 out: 1640 vrele(vp); 1641 vn_finished_write(mp, 0); 1642 return (error); 1643 } 1644 1645 /* 1646 * Make a symbolic link. 1647 */ 1648 /* ARGSUSED */ 1649 int 1650 sys_symlink(struct lwp *l, void *v, register_t *retval) 1651 { 1652 struct sys_symlink_args /* { 1653 syscallarg(const char *) path; 1654 syscallarg(const char *) link; 1655 } */ *uap = v; 1656 struct proc *p = l->l_proc; 1657 struct mount *mp; 1658 struct vattr vattr; 1659 char *path; 1660 int error; 1661 struct nameidata nd; 1662 1663 path = PNBUF_GET(); 1664 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1665 if (error) 1666 goto out; 1667 restart: 1668 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1669 if ((error = namei(&nd)) != 0) 1670 goto out; 1671 if (nd.ni_vp) { 1672 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1673 if (nd.ni_dvp == nd.ni_vp) 1674 vrele(nd.ni_dvp); 1675 else 1676 vput(nd.ni_dvp); 1677 vrele(nd.ni_vp); 1678 error = EEXIST; 1679 goto out; 1680 } 1681 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1682 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1683 if (nd.ni_dvp == nd.ni_vp) 1684 vrele(nd.ni_dvp); 1685 else 1686 vput(nd.ni_dvp); 1687 if ((error = vn_start_write(NULL, &mp, 1688 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1689 return (error); 1690 goto restart; 1691 } 1692 VATTR_NULL(&vattr); 1693 vattr.va_type = VLNK; 1694 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1695 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1696 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1697 if (error == 0) 1698 vput(nd.ni_vp); 1699 vn_finished_write(mp, 0); 1700 out: 1701 PNBUF_PUT(path); 1702 return (error); 1703 } 1704 1705 /* 1706 * Delete a whiteout from the filesystem. 1707 */ 1708 /* ARGSUSED */ 1709 int 1710 sys_undelete(struct lwp *l, void *v, register_t *retval) 1711 { 1712 struct sys_undelete_args /* { 1713 syscallarg(const char *) path; 1714 } */ *uap = v; 1715 struct proc *p = l->l_proc; 1716 int error; 1717 struct mount *mp; 1718 struct nameidata nd; 1719 1720 restart: 1721 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1722 SCARG(uap, path), l); 1723 error = namei(&nd); 1724 if (error) 1725 return (error); 1726 1727 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1728 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1729 if (nd.ni_dvp == nd.ni_vp) 1730 vrele(nd.ni_dvp); 1731 else 1732 vput(nd.ni_dvp); 1733 if (nd.ni_vp) 1734 vrele(nd.ni_vp); 1735 return (EEXIST); 1736 } 1737 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1738 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1739 if (nd.ni_dvp == nd.ni_vp) 1740 vrele(nd.ni_dvp); 1741 else 1742 vput(nd.ni_dvp); 1743 if ((error = vn_start_write(NULL, &mp, 1744 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1745 return (error); 1746 goto restart; 1747 } 1748 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1749 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1750 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1751 vput(nd.ni_dvp); 1752 vn_finished_write(mp, 0); 1753 return (error); 1754 } 1755 1756 /* 1757 * Delete a name from the filesystem. 1758 */ 1759 /* ARGSUSED */ 1760 int 1761 sys_unlink(struct lwp *l, void *v, register_t *retval) 1762 { 1763 struct sys_unlink_args /* { 1764 syscallarg(const char *) path; 1765 } */ *uap = v; 1766 struct proc *p = l->l_proc; 1767 struct mount *mp; 1768 struct vnode *vp; 1769 int error; 1770 struct nameidata nd; 1771 1772 restart: 1773 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1774 SCARG(uap, path), l); 1775 if ((error = namei(&nd)) != 0) 1776 return (error); 1777 vp = nd.ni_vp; 1778 1779 /* 1780 * The root of a mounted filesystem cannot be deleted. 1781 */ 1782 if (vp->v_flag & VROOT) { 1783 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1784 if (nd.ni_dvp == vp) 1785 vrele(nd.ni_dvp); 1786 else 1787 vput(nd.ni_dvp); 1788 vput(vp); 1789 error = EBUSY; 1790 goto out; 1791 } 1792 1793 #ifdef VERIFIED_EXEC 1794 /* Handle remove requests for veriexec entries. */ 1795 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) { 1796 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1797 if (nd.ni_dvp == vp) 1798 vrele(nd.ni_dvp); 1799 else 1800 vput(nd.ni_dvp); 1801 vput(vp); 1802 goto out; 1803 } 1804 #endif /* VERIFIED_EXEC */ 1805 1806 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1807 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1808 if (nd.ni_dvp == vp) 1809 vrele(nd.ni_dvp); 1810 else 1811 vput(nd.ni_dvp); 1812 vput(vp); 1813 if ((error = vn_start_write(NULL, &mp, 1814 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1815 return (error); 1816 goto restart; 1817 } 1818 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1819 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 1820 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1821 vn_finished_write(mp, 0); 1822 out: 1823 return (error); 1824 } 1825 1826 /* 1827 * Reposition read/write file offset. 1828 */ 1829 int 1830 sys_lseek(struct lwp *l, void *v, register_t *retval) 1831 { 1832 struct sys_lseek_args /* { 1833 syscallarg(int) fd; 1834 syscallarg(int) pad; 1835 syscallarg(off_t) offset; 1836 syscallarg(int) whence; 1837 } */ *uap = v; 1838 struct proc *p = l->l_proc; 1839 struct ucred *cred = p->p_ucred; 1840 struct filedesc *fdp = p->p_fd; 1841 struct file *fp; 1842 struct vnode *vp; 1843 struct vattr vattr; 1844 off_t newoff; 1845 int error; 1846 1847 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1848 return (EBADF); 1849 1850 FILE_USE(fp); 1851 1852 vp = (struct vnode *)fp->f_data; 1853 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1854 error = ESPIPE; 1855 goto out; 1856 } 1857 1858 switch (SCARG(uap, whence)) { 1859 case SEEK_CUR: 1860 newoff = fp->f_offset + SCARG(uap, offset); 1861 break; 1862 case SEEK_END: 1863 error = VOP_GETATTR(vp, &vattr, cred, l); 1864 if (error) 1865 goto out; 1866 newoff = SCARG(uap, offset) + vattr.va_size; 1867 break; 1868 case SEEK_SET: 1869 newoff = SCARG(uap, offset); 1870 break; 1871 default: 1872 error = EINVAL; 1873 goto out; 1874 } 1875 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1876 goto out; 1877 1878 *(off_t *)retval = fp->f_offset = newoff; 1879 out: 1880 FILE_UNUSE(fp, l); 1881 return (error); 1882 } 1883 1884 /* 1885 * Positional read system call. 1886 */ 1887 int 1888 sys_pread(struct lwp *l, void *v, register_t *retval) 1889 { 1890 struct sys_pread_args /* { 1891 syscallarg(int) fd; 1892 syscallarg(void *) buf; 1893 syscallarg(size_t) nbyte; 1894 syscallarg(off_t) offset; 1895 } */ *uap = v; 1896 struct proc *p = l->l_proc; 1897 struct filedesc *fdp = p->p_fd; 1898 struct file *fp; 1899 struct vnode *vp; 1900 off_t offset; 1901 int error, fd = SCARG(uap, fd); 1902 1903 if ((fp = fd_getfile(fdp, fd)) == NULL) 1904 return (EBADF); 1905 1906 if ((fp->f_flag & FREAD) == 0) { 1907 simple_unlock(&fp->f_slock); 1908 return (EBADF); 1909 } 1910 1911 FILE_USE(fp); 1912 1913 vp = (struct vnode *)fp->f_data; 1914 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1915 error = ESPIPE; 1916 goto out; 1917 } 1918 1919 offset = SCARG(uap, offset); 1920 1921 /* 1922 * XXX This works because no file systems actually 1923 * XXX take any action on the seek operation. 1924 */ 1925 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1926 goto out; 1927 1928 /* dofileread() will unuse the descriptor for us */ 1929 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1930 &offset, 0, retval)); 1931 1932 out: 1933 FILE_UNUSE(fp, l); 1934 return (error); 1935 } 1936 1937 /* 1938 * Positional scatter read system call. 1939 */ 1940 int 1941 sys_preadv(struct lwp *l, void *v, register_t *retval) 1942 { 1943 struct sys_preadv_args /* { 1944 syscallarg(int) fd; 1945 syscallarg(const struct iovec *) iovp; 1946 syscallarg(int) iovcnt; 1947 syscallarg(off_t) offset; 1948 } */ *uap = v; 1949 struct proc *p = l->l_proc; 1950 struct filedesc *fdp = p->p_fd; 1951 struct file *fp; 1952 struct vnode *vp; 1953 off_t offset; 1954 int error, fd = SCARG(uap, fd); 1955 1956 if ((fp = fd_getfile(fdp, fd)) == NULL) 1957 return (EBADF); 1958 1959 if ((fp->f_flag & FREAD) == 0) { 1960 simple_unlock(&fp->f_slock); 1961 return (EBADF); 1962 } 1963 1964 FILE_USE(fp); 1965 1966 vp = (struct vnode *)fp->f_data; 1967 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1968 error = ESPIPE; 1969 goto out; 1970 } 1971 1972 offset = SCARG(uap, offset); 1973 1974 /* 1975 * XXX This works because no file systems actually 1976 * XXX take any action on the seek operation. 1977 */ 1978 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1979 goto out; 1980 1981 /* dofilereadv() will unuse the descriptor for us */ 1982 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1983 &offset, 0, retval)); 1984 1985 out: 1986 FILE_UNUSE(fp, l); 1987 return (error); 1988 } 1989 1990 /* 1991 * Positional write system call. 1992 */ 1993 int 1994 sys_pwrite(struct lwp *l, void *v, register_t *retval) 1995 { 1996 struct sys_pwrite_args /* { 1997 syscallarg(int) fd; 1998 syscallarg(const void *) buf; 1999 syscallarg(size_t) nbyte; 2000 syscallarg(off_t) offset; 2001 } */ *uap = v; 2002 struct proc *p = l->l_proc; 2003 struct filedesc *fdp = p->p_fd; 2004 struct file *fp; 2005 struct vnode *vp; 2006 off_t offset; 2007 int error, fd = SCARG(uap, fd); 2008 2009 if ((fp = fd_getfile(fdp, fd)) == NULL) 2010 return (EBADF); 2011 2012 if ((fp->f_flag & FWRITE) == 0) { 2013 simple_unlock(&fp->f_slock); 2014 return (EBADF); 2015 } 2016 2017 FILE_USE(fp); 2018 2019 vp = (struct vnode *)fp->f_data; 2020 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2021 error = ESPIPE; 2022 goto out; 2023 } 2024 2025 offset = SCARG(uap, offset); 2026 2027 /* 2028 * XXX This works because no file systems actually 2029 * XXX take any action on the seek operation. 2030 */ 2031 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2032 goto out; 2033 2034 /* dofilewrite() will unuse the descriptor for us */ 2035 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2036 &offset, 0, retval)); 2037 2038 out: 2039 FILE_UNUSE(fp, l); 2040 return (error); 2041 } 2042 2043 /* 2044 * Positional gather write system call. 2045 */ 2046 int 2047 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2048 { 2049 struct sys_pwritev_args /* { 2050 syscallarg(int) fd; 2051 syscallarg(const struct iovec *) iovp; 2052 syscallarg(int) iovcnt; 2053 syscallarg(off_t) offset; 2054 } */ *uap = v; 2055 struct proc *p = l->l_proc; 2056 struct filedesc *fdp = p->p_fd; 2057 struct file *fp; 2058 struct vnode *vp; 2059 off_t offset; 2060 int error, fd = SCARG(uap, fd); 2061 2062 if ((fp = fd_getfile(fdp, fd)) == NULL) 2063 return (EBADF); 2064 2065 if ((fp->f_flag & FWRITE) == 0) { 2066 simple_unlock(&fp->f_slock); 2067 return (EBADF); 2068 } 2069 2070 FILE_USE(fp); 2071 2072 vp = (struct vnode *)fp->f_data; 2073 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2074 error = ESPIPE; 2075 goto out; 2076 } 2077 2078 offset = SCARG(uap, offset); 2079 2080 /* 2081 * XXX This works because no file systems actually 2082 * XXX take any action on the seek operation. 2083 */ 2084 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2085 goto out; 2086 2087 /* dofilewritev() will unuse the descriptor for us */ 2088 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2089 &offset, 0, retval)); 2090 2091 out: 2092 FILE_UNUSE(fp, l); 2093 return (error); 2094 } 2095 2096 /* 2097 * Check access permissions. 2098 */ 2099 int 2100 sys_access(struct lwp *l, void *v, register_t *retval) 2101 { 2102 struct sys_access_args /* { 2103 syscallarg(const char *) path; 2104 syscallarg(int) flags; 2105 } */ *uap = v; 2106 struct proc *p = l->l_proc; 2107 struct ucred *cred; 2108 struct vnode *vp; 2109 int error, flags; 2110 struct nameidata nd; 2111 2112 cred = crdup(p->p_ucred); 2113 cred->cr_uid = p->p_cred->p_ruid; 2114 cred->cr_gid = p->p_cred->p_rgid; 2115 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2116 SCARG(uap, path), l); 2117 /* Override default credentials */ 2118 nd.ni_cnd.cn_cred = cred; 2119 if ((error = namei(&nd)) != 0) 2120 goto out; 2121 vp = nd.ni_vp; 2122 2123 /* Flags == 0 means only check for existence. */ 2124 if (SCARG(uap, flags)) { 2125 flags = 0; 2126 if (SCARG(uap, flags) & R_OK) 2127 flags |= VREAD; 2128 if (SCARG(uap, flags) & W_OK) 2129 flags |= VWRITE; 2130 if (SCARG(uap, flags) & X_OK) 2131 flags |= VEXEC; 2132 2133 error = VOP_ACCESS(vp, flags, cred, l); 2134 if (!error && (flags & VWRITE)) 2135 error = vn_writechk(vp); 2136 } 2137 vput(vp); 2138 out: 2139 crfree(cred); 2140 return (error); 2141 } 2142 2143 /* 2144 * Get file status; this version follows links. 2145 */ 2146 /* ARGSUSED */ 2147 int 2148 sys___stat30(struct lwp *l, void *v, register_t *retval) 2149 { 2150 struct sys___stat30_args /* { 2151 syscallarg(const char *) path; 2152 syscallarg(struct stat *) ub; 2153 } */ *uap = v; 2154 struct stat sb; 2155 int error; 2156 struct nameidata nd; 2157 2158 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2159 SCARG(uap, path), l); 2160 if ((error = namei(&nd)) != 0) 2161 return (error); 2162 error = vn_stat(nd.ni_vp, &sb, l); 2163 vput(nd.ni_vp); 2164 if (error) 2165 return (error); 2166 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2167 return (error); 2168 } 2169 2170 /* 2171 * Get file status; this version does not follow links. 2172 */ 2173 /* ARGSUSED */ 2174 int 2175 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2176 { 2177 struct sys___lstat30_args /* { 2178 syscallarg(const char *) path; 2179 syscallarg(struct stat *) ub; 2180 } */ *uap = v; 2181 struct stat sb; 2182 int error; 2183 struct nameidata nd; 2184 2185 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2186 SCARG(uap, path), l); 2187 if ((error = namei(&nd)) != 0) 2188 return (error); 2189 error = vn_stat(nd.ni_vp, &sb, l); 2190 vput(nd.ni_vp); 2191 if (error) 2192 return (error); 2193 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2194 return (error); 2195 } 2196 2197 /* 2198 * Get configurable pathname variables. 2199 */ 2200 /* ARGSUSED */ 2201 int 2202 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2203 { 2204 struct sys_pathconf_args /* { 2205 syscallarg(const char *) path; 2206 syscallarg(int) name; 2207 } */ *uap = v; 2208 int error; 2209 struct nameidata nd; 2210 2211 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2212 SCARG(uap, path), l); 2213 if ((error = namei(&nd)) != 0) 2214 return (error); 2215 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2216 vput(nd.ni_vp); 2217 return (error); 2218 } 2219 2220 /* 2221 * Return target name of a symbolic link. 2222 */ 2223 /* ARGSUSED */ 2224 int 2225 sys_readlink(struct lwp *l, void *v, register_t *retval) 2226 { 2227 struct sys_readlink_args /* { 2228 syscallarg(const char *) path; 2229 syscallarg(char *) buf; 2230 syscallarg(size_t) count; 2231 } */ *uap = v; 2232 struct proc *p = l->l_proc; 2233 struct vnode *vp; 2234 struct iovec aiov; 2235 struct uio auio; 2236 int error; 2237 struct nameidata nd; 2238 2239 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2240 SCARG(uap, path), l); 2241 if ((error = namei(&nd)) != 0) 2242 return (error); 2243 vp = nd.ni_vp; 2244 if (vp->v_type != VLNK) 2245 error = EINVAL; 2246 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2247 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, l)) == 0) { 2248 aiov.iov_base = SCARG(uap, buf); 2249 aiov.iov_len = SCARG(uap, count); 2250 auio.uio_iov = &aiov; 2251 auio.uio_iovcnt = 1; 2252 auio.uio_offset = 0; 2253 auio.uio_rw = UIO_READ; 2254 auio.uio_segflg = UIO_USERSPACE; 2255 auio.uio_lwp = l; 2256 auio.uio_resid = SCARG(uap, count); 2257 error = VOP_READLINK(vp, &auio, p->p_ucred); 2258 } 2259 vput(vp); 2260 *retval = SCARG(uap, count) - auio.uio_resid; 2261 return (error); 2262 } 2263 2264 /* 2265 * Change flags of a file given a path name. 2266 */ 2267 /* ARGSUSED */ 2268 int 2269 sys_chflags(struct lwp *l, void *v, register_t *retval) 2270 { 2271 struct sys_chflags_args /* { 2272 syscallarg(const char *) path; 2273 syscallarg(u_long) flags; 2274 } */ *uap = v; 2275 struct vnode *vp; 2276 int error; 2277 struct nameidata nd; 2278 2279 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2280 if ((error = namei(&nd)) != 0) 2281 return (error); 2282 vp = nd.ni_vp; 2283 error = change_flags(vp, SCARG(uap, flags), l); 2284 vput(vp); 2285 return (error); 2286 } 2287 2288 /* 2289 * Change flags of a file given a file descriptor. 2290 */ 2291 /* ARGSUSED */ 2292 int 2293 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2294 { 2295 struct sys_fchflags_args /* { 2296 syscallarg(int) fd; 2297 syscallarg(u_long) flags; 2298 } */ *uap = v; 2299 struct proc *p = l->l_proc; 2300 struct vnode *vp; 2301 struct file *fp; 2302 int error; 2303 2304 /* getvnode() will use the descriptor for us */ 2305 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2306 return (error); 2307 vp = (struct vnode *)fp->f_data; 2308 error = change_flags(vp, SCARG(uap, flags), l); 2309 VOP_UNLOCK(vp, 0); 2310 FILE_UNUSE(fp, l); 2311 return (error); 2312 } 2313 2314 /* 2315 * Change flags of a file given a path name; this version does 2316 * not follow links. 2317 */ 2318 int 2319 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2320 { 2321 struct sys_lchflags_args /* { 2322 syscallarg(const char *) path; 2323 syscallarg(u_long) flags; 2324 } */ *uap = v; 2325 struct vnode *vp; 2326 int error; 2327 struct nameidata nd; 2328 2329 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2330 if ((error = namei(&nd)) != 0) 2331 return (error); 2332 vp = nd.ni_vp; 2333 error = change_flags(vp, SCARG(uap, flags), l); 2334 vput(vp); 2335 return (error); 2336 } 2337 2338 /* 2339 * Common routine to change flags of a file. 2340 */ 2341 int 2342 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2343 { 2344 struct proc *p = l->l_proc; 2345 struct mount *mp; 2346 struct vattr vattr; 2347 int error; 2348 2349 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2350 return (error); 2351 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2352 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2353 /* 2354 * Non-superusers cannot change the flags on devices, even if they 2355 * own them. 2356 */ 2357 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2358 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 2359 goto out; 2360 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2361 error = EINVAL; 2362 goto out; 2363 } 2364 } 2365 VATTR_NULL(&vattr); 2366 vattr.va_flags = flags; 2367 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2368 out: 2369 vn_finished_write(mp, 0); 2370 return (error); 2371 } 2372 2373 /* 2374 * Change mode of a file given path name; this version follows links. 2375 */ 2376 /* ARGSUSED */ 2377 int 2378 sys_chmod(struct lwp *l, void *v, register_t *retval) 2379 { 2380 struct sys_chmod_args /* { 2381 syscallarg(const char *) path; 2382 syscallarg(int) mode; 2383 } */ *uap = v; 2384 int error; 2385 struct nameidata nd; 2386 2387 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2388 if ((error = namei(&nd)) != 0) 2389 return (error); 2390 2391 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2392 2393 vrele(nd.ni_vp); 2394 return (error); 2395 } 2396 2397 /* 2398 * Change mode of a file given a file descriptor. 2399 */ 2400 /* ARGSUSED */ 2401 int 2402 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2403 { 2404 struct sys_fchmod_args /* { 2405 syscallarg(int) fd; 2406 syscallarg(int) mode; 2407 } */ *uap = v; 2408 struct proc *p = l->l_proc; 2409 struct file *fp; 2410 int error; 2411 2412 /* getvnode() will use the descriptor for us */ 2413 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2414 return (error); 2415 2416 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2417 FILE_UNUSE(fp, l); 2418 return (error); 2419 } 2420 2421 /* 2422 * Change mode of a file given path name; this version does not follow links. 2423 */ 2424 /* ARGSUSED */ 2425 int 2426 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2427 { 2428 struct sys_lchmod_args /* { 2429 syscallarg(const char *) path; 2430 syscallarg(int) mode; 2431 } */ *uap = v; 2432 int error; 2433 struct nameidata nd; 2434 2435 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2436 if ((error = namei(&nd)) != 0) 2437 return (error); 2438 2439 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2440 2441 vrele(nd.ni_vp); 2442 return (error); 2443 } 2444 2445 /* 2446 * Common routine to set mode given a vnode. 2447 */ 2448 static int 2449 change_mode(struct vnode *vp, int mode, struct lwp *l) 2450 { 2451 struct proc *p = l->l_proc; 2452 struct mount *mp; 2453 struct vattr vattr; 2454 int error; 2455 2456 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2457 return (error); 2458 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2459 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2460 VATTR_NULL(&vattr); 2461 vattr.va_mode = mode & ALLPERMS; 2462 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2463 VOP_UNLOCK(vp, 0); 2464 vn_finished_write(mp, 0); 2465 return (error); 2466 } 2467 2468 /* 2469 * Set ownership given a path name; this version follows links. 2470 */ 2471 /* ARGSUSED */ 2472 int 2473 sys_chown(struct lwp *l, void *v, register_t *retval) 2474 { 2475 struct sys_chown_args /* { 2476 syscallarg(const char *) path; 2477 syscallarg(uid_t) uid; 2478 syscallarg(gid_t) gid; 2479 } */ *uap = v; 2480 int error; 2481 struct nameidata nd; 2482 2483 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2484 if ((error = namei(&nd)) != 0) 2485 return (error); 2486 2487 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2488 2489 vrele(nd.ni_vp); 2490 return (error); 2491 } 2492 2493 /* 2494 * Set ownership given a path name; this version follows links. 2495 * Provides POSIX semantics. 2496 */ 2497 /* ARGSUSED */ 2498 int 2499 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2500 { 2501 struct sys_chown_args /* { 2502 syscallarg(const char *) path; 2503 syscallarg(uid_t) uid; 2504 syscallarg(gid_t) gid; 2505 } */ *uap = v; 2506 int error; 2507 struct nameidata nd; 2508 2509 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2510 if ((error = namei(&nd)) != 0) 2511 return (error); 2512 2513 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2514 2515 vrele(nd.ni_vp); 2516 return (error); 2517 } 2518 2519 /* 2520 * Set ownership given a file descriptor. 2521 */ 2522 /* ARGSUSED */ 2523 int 2524 sys_fchown(struct lwp *l, void *v, register_t *retval) 2525 { 2526 struct sys_fchown_args /* { 2527 syscallarg(int) fd; 2528 syscallarg(uid_t) uid; 2529 syscallarg(gid_t) gid; 2530 } */ *uap = v; 2531 struct proc *p = l->l_proc; 2532 int error; 2533 struct file *fp; 2534 2535 /* getvnode() will use the descriptor for us */ 2536 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2537 return (error); 2538 2539 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2540 SCARG(uap, gid), l, 0); 2541 FILE_UNUSE(fp, l); 2542 return (error); 2543 } 2544 2545 /* 2546 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2547 */ 2548 /* ARGSUSED */ 2549 int 2550 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2551 { 2552 struct sys_fchown_args /* { 2553 syscallarg(int) fd; 2554 syscallarg(uid_t) uid; 2555 syscallarg(gid_t) gid; 2556 } */ *uap = v; 2557 struct proc *p = l->l_proc; 2558 int error; 2559 struct file *fp; 2560 2561 /* getvnode() will use the descriptor for us */ 2562 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2563 return (error); 2564 2565 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2566 SCARG(uap, gid), l, 1); 2567 FILE_UNUSE(fp, l); 2568 return (error); 2569 } 2570 2571 /* 2572 * Set ownership given a path name; this version does not follow links. 2573 */ 2574 /* ARGSUSED */ 2575 int 2576 sys_lchown(struct lwp *l, void *v, register_t *retval) 2577 { 2578 struct sys_lchown_args /* { 2579 syscallarg(const char *) path; 2580 syscallarg(uid_t) uid; 2581 syscallarg(gid_t) gid; 2582 } */ *uap = v; 2583 int error; 2584 struct nameidata nd; 2585 2586 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2587 if ((error = namei(&nd)) != 0) 2588 return (error); 2589 2590 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2591 2592 vrele(nd.ni_vp); 2593 return (error); 2594 } 2595 2596 /* 2597 * Set ownership given a path name; this version does not follow links. 2598 * Provides POSIX/XPG semantics. 2599 */ 2600 /* ARGSUSED */ 2601 int 2602 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2603 { 2604 struct sys_lchown_args /* { 2605 syscallarg(const char *) path; 2606 syscallarg(uid_t) uid; 2607 syscallarg(gid_t) gid; 2608 } */ *uap = v; 2609 int error; 2610 struct nameidata nd; 2611 2612 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2613 if ((error = namei(&nd)) != 0) 2614 return (error); 2615 2616 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2617 2618 vrele(nd.ni_vp); 2619 return (error); 2620 } 2621 2622 /* 2623 * Common routine to set ownership given a vnode. 2624 */ 2625 static int 2626 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2627 int posix_semantics) 2628 { 2629 struct proc *p = l->l_proc; 2630 struct mount *mp; 2631 struct vattr vattr; 2632 mode_t newmode; 2633 int error; 2634 2635 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2636 return (error); 2637 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2638 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2639 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 2640 goto out; 2641 2642 #define CHANGED(x) ((int)(x) != -1) 2643 newmode = vattr.va_mode; 2644 if (posix_semantics) { 2645 /* 2646 * POSIX/XPG semantics: if the caller is not the super-user, 2647 * clear set-user-id and set-group-id bits. Both POSIX and 2648 * the XPG consider the behaviour for calls by the super-user 2649 * implementation-defined; we leave the set-user-id and set- 2650 * group-id settings intact in that case. 2651 */ 2652 if (suser(p->p_ucred, NULL) != 0) 2653 newmode &= ~(S_ISUID | S_ISGID); 2654 } else { 2655 /* 2656 * NetBSD semantics: when changing owner and/or group, 2657 * clear the respective bit(s). 2658 */ 2659 if (CHANGED(uid)) 2660 newmode &= ~S_ISUID; 2661 if (CHANGED(gid)) 2662 newmode &= ~S_ISGID; 2663 } 2664 /* Update va_mode iff altered. */ 2665 if (vattr.va_mode == newmode) 2666 newmode = VNOVAL; 2667 2668 VATTR_NULL(&vattr); 2669 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2670 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2671 vattr.va_mode = newmode; 2672 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2673 #undef CHANGED 2674 2675 out: 2676 VOP_UNLOCK(vp, 0); 2677 vn_finished_write(mp, 0); 2678 return (error); 2679 } 2680 2681 /* 2682 * Set the access and modification times given a path name; this 2683 * version follows links. 2684 */ 2685 /* ARGSUSED */ 2686 int 2687 sys_utimes(struct lwp *l, void *v, register_t *retval) 2688 { 2689 struct sys_utimes_args /* { 2690 syscallarg(const char *) path; 2691 syscallarg(const struct timeval *) tptr; 2692 } */ *uap = v; 2693 int error; 2694 struct nameidata nd; 2695 2696 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2697 if ((error = namei(&nd)) != 0) 2698 return (error); 2699 2700 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2701 2702 vrele(nd.ni_vp); 2703 return (error); 2704 } 2705 2706 /* 2707 * Set the access and modification times given a file descriptor. 2708 */ 2709 /* ARGSUSED */ 2710 int 2711 sys_futimes(struct lwp *l, void *v, register_t *retval) 2712 { 2713 struct sys_futimes_args /* { 2714 syscallarg(int) fd; 2715 syscallarg(const struct timeval *) tptr; 2716 } */ *uap = v; 2717 struct proc *p = l->l_proc; 2718 int error; 2719 struct file *fp; 2720 2721 /* getvnode() will use the descriptor for us */ 2722 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2723 return (error); 2724 2725 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2726 FILE_UNUSE(fp, l); 2727 return (error); 2728 } 2729 2730 /* 2731 * Set the access and modification times given a path name; this 2732 * version does not follow links. 2733 */ 2734 /* ARGSUSED */ 2735 int 2736 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2737 { 2738 struct sys_lutimes_args /* { 2739 syscallarg(const char *) path; 2740 syscallarg(const struct timeval *) tptr; 2741 } */ *uap = v; 2742 int error; 2743 struct nameidata nd; 2744 2745 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2746 if ((error = namei(&nd)) != 0) 2747 return (error); 2748 2749 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2750 2751 vrele(nd.ni_vp); 2752 return (error); 2753 } 2754 2755 /* 2756 * Common routine to set access and modification times given a vnode. 2757 */ 2758 static int 2759 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2760 { 2761 struct proc *p = l->l_proc; 2762 struct mount *mp; 2763 struct vattr vattr; 2764 int error; 2765 2766 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2767 return (error); 2768 VATTR_NULL(&vattr); 2769 if (tptr == NULL) { 2770 nanotime(&vattr.va_atime); 2771 vattr.va_mtime = vattr.va_atime; 2772 vattr.va_vaflags |= VA_UTIMES_NULL; 2773 } else { 2774 struct timeval tv[2]; 2775 2776 error = copyin(tptr, tv, sizeof(tv)); 2777 if (error) 2778 goto out; 2779 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 2780 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 2781 } 2782 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2783 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2784 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2785 VOP_UNLOCK(vp, 0); 2786 out: 2787 vn_finished_write(mp, 0); 2788 return (error); 2789 } 2790 2791 /* 2792 * Truncate a file given its path name. 2793 */ 2794 /* ARGSUSED */ 2795 int 2796 sys_truncate(struct lwp *l, void *v, register_t *retval) 2797 { 2798 struct sys_truncate_args /* { 2799 syscallarg(const char *) path; 2800 syscallarg(int) pad; 2801 syscallarg(off_t) length; 2802 } */ *uap = v; 2803 struct proc *p = l->l_proc; 2804 struct vnode *vp; 2805 struct mount *mp; 2806 struct vattr vattr; 2807 int error; 2808 struct nameidata nd; 2809 2810 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2811 if ((error = namei(&nd)) != 0) 2812 return (error); 2813 vp = nd.ni_vp; 2814 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2815 vrele(vp); 2816 return (error); 2817 } 2818 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2819 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2820 if (vp->v_type == VDIR) 2821 error = EISDIR; 2822 else if ((error = vn_writechk(vp)) == 0 && 2823 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, l)) == 0) { 2824 VATTR_NULL(&vattr); 2825 vattr.va_size = SCARG(uap, length); 2826 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2827 } 2828 vput(vp); 2829 vn_finished_write(mp, 0); 2830 return (error); 2831 } 2832 2833 /* 2834 * Truncate a file given a file descriptor. 2835 */ 2836 /* ARGSUSED */ 2837 int 2838 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2839 { 2840 struct sys_ftruncate_args /* { 2841 syscallarg(int) fd; 2842 syscallarg(int) pad; 2843 syscallarg(off_t) length; 2844 } */ *uap = v; 2845 struct proc *p = l->l_proc; 2846 struct mount *mp; 2847 struct vattr vattr; 2848 struct vnode *vp; 2849 struct file *fp; 2850 int error; 2851 2852 /* getvnode() will use the descriptor for us */ 2853 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2854 return (error); 2855 if ((fp->f_flag & FWRITE) == 0) { 2856 error = EINVAL; 2857 goto out; 2858 } 2859 vp = (struct vnode *)fp->f_data; 2860 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2861 FILE_UNUSE(fp, l); 2862 return (error); 2863 } 2864 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2865 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2866 if (vp->v_type == VDIR) 2867 error = EISDIR; 2868 else if ((error = vn_writechk(vp)) == 0) { 2869 VATTR_NULL(&vattr); 2870 vattr.va_size = SCARG(uap, length); 2871 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 2872 } 2873 VOP_UNLOCK(vp, 0); 2874 vn_finished_write(mp, 0); 2875 out: 2876 FILE_UNUSE(fp, l); 2877 return (error); 2878 } 2879 2880 /* 2881 * Sync an open file. 2882 */ 2883 /* ARGSUSED */ 2884 int 2885 sys_fsync(struct lwp *l, void *v, register_t *retval) 2886 { 2887 struct sys_fsync_args /* { 2888 syscallarg(int) fd; 2889 } */ *uap = v; 2890 struct proc *p = l->l_proc; 2891 struct vnode *vp; 2892 struct mount *mp; 2893 struct file *fp; 2894 int error; 2895 2896 /* getvnode() will use the descriptor for us */ 2897 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2898 return (error); 2899 vp = (struct vnode *)fp->f_data; 2900 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2901 FILE_UNUSE(fp, l); 2902 return (error); 2903 } 2904 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2905 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 2906 if (error == 0 && bioops.io_fsync != NULL && 2907 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2908 (*bioops.io_fsync)(vp, 0); 2909 VOP_UNLOCK(vp, 0); 2910 vn_finished_write(mp, 0); 2911 FILE_UNUSE(fp, l); 2912 return (error); 2913 } 2914 2915 /* 2916 * Sync a range of file data. API modeled after that found in AIX. 2917 * 2918 * FDATASYNC indicates that we need only save enough metadata to be able 2919 * to re-read the written data. Note we duplicate AIX's requirement that 2920 * the file be open for writing. 2921 */ 2922 /* ARGSUSED */ 2923 int 2924 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2925 { 2926 struct sys_fsync_range_args /* { 2927 syscallarg(int) fd; 2928 syscallarg(int) flags; 2929 syscallarg(off_t) start; 2930 syscallarg(off_t) length; 2931 } */ *uap = v; 2932 struct proc *p = l->l_proc; 2933 struct vnode *vp; 2934 struct file *fp; 2935 int flags, nflags; 2936 off_t s, e, len; 2937 int error; 2938 2939 /* getvnode() will use the descriptor for us */ 2940 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2941 return (error); 2942 2943 if ((fp->f_flag & FWRITE) == 0) { 2944 FILE_UNUSE(fp, l); 2945 return (EBADF); 2946 } 2947 2948 flags = SCARG(uap, flags); 2949 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2950 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2951 return (EINVAL); 2952 } 2953 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2954 if (flags & FDATASYNC) 2955 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2956 else 2957 nflags = FSYNC_WAIT; 2958 if (flags & FDISKSYNC) 2959 nflags |= FSYNC_CACHE; 2960 2961 len = SCARG(uap, length); 2962 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2963 if (len) { 2964 s = SCARG(uap, start); 2965 e = s + len; 2966 if (e < s) { 2967 FILE_UNUSE(fp, l); 2968 return (EINVAL); 2969 } 2970 } else { 2971 e = 0; 2972 s = 0; 2973 } 2974 2975 vp = (struct vnode *)fp->f_data; 2976 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2977 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 2978 2979 if (error == 0 && bioops.io_fsync != NULL && 2980 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2981 (*bioops.io_fsync)(vp, nflags); 2982 2983 VOP_UNLOCK(vp, 0); 2984 FILE_UNUSE(fp, l); 2985 return (error); 2986 } 2987 2988 /* 2989 * Sync the data of an open file. 2990 */ 2991 /* ARGSUSED */ 2992 int 2993 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 2994 { 2995 struct sys_fdatasync_args /* { 2996 syscallarg(int) fd; 2997 } */ *uap = v; 2998 struct proc *p = l->l_proc; 2999 struct vnode *vp; 3000 struct file *fp; 3001 int error; 3002 3003 /* getvnode() will use the descriptor for us */ 3004 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3005 return (error); 3006 if ((fp->f_flag & FWRITE) == 0) { 3007 FILE_UNUSE(fp, l); 3008 return (EBADF); 3009 } 3010 vp = (struct vnode *)fp->f_data; 3011 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3012 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3013 VOP_UNLOCK(vp, 0); 3014 FILE_UNUSE(fp, l); 3015 return (error); 3016 } 3017 3018 /* 3019 * Rename files, (standard) BSD semantics frontend. 3020 */ 3021 /* ARGSUSED */ 3022 int 3023 sys_rename(struct lwp *l, void *v, register_t *retval) 3024 { 3025 struct sys_rename_args /* { 3026 syscallarg(const char *) from; 3027 syscallarg(const char *) to; 3028 } */ *uap = v; 3029 3030 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3031 } 3032 3033 /* 3034 * Rename files, POSIX semantics frontend. 3035 */ 3036 /* ARGSUSED */ 3037 int 3038 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3039 { 3040 struct sys___posix_rename_args /* { 3041 syscallarg(const char *) from; 3042 syscallarg(const char *) to; 3043 } */ *uap = v; 3044 3045 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3046 } 3047 3048 /* 3049 * Rename files. Source and destination must either both be directories, 3050 * or both not be directories. If target is a directory, it must be empty. 3051 * If `from' and `to' refer to the same object, the value of the `retain' 3052 * argument is used to determine whether `from' will be 3053 * 3054 * (retain == 0) deleted unless `from' and `to' refer to the same 3055 * object in the file system's name space (BSD). 3056 * (retain == 1) always retained (POSIX). 3057 */ 3058 static int 3059 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3060 { 3061 struct mount *mp = NULL; 3062 struct vnode *tvp, *fvp, *tdvp; 3063 struct nameidata fromnd, tond; 3064 struct proc *p; 3065 int error; 3066 3067 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3068 from, l); 3069 if ((error = namei(&fromnd)) != 0) 3070 return (error); 3071 fvp = fromnd.ni_vp; 3072 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3073 if (error != 0) { 3074 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3075 vrele(fromnd.ni_dvp); 3076 vrele(fvp); 3077 if (fromnd.ni_startdir) 3078 vrele(fromnd.ni_startdir); 3079 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3080 return (error); 3081 } 3082 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3083 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3084 if ((error = namei(&tond)) != 0) { 3085 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3086 vrele(fromnd.ni_dvp); 3087 vrele(fvp); 3088 goto out1; 3089 } 3090 tdvp = tond.ni_dvp; 3091 tvp = tond.ni_vp; 3092 3093 if (tvp != NULL) { 3094 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3095 error = ENOTDIR; 3096 goto out; 3097 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3098 error = EISDIR; 3099 goto out; 3100 } 3101 } 3102 3103 if (fvp == tdvp) 3104 error = EINVAL; 3105 3106 /* 3107 * Source and destination refer to the same object. 3108 */ 3109 if (fvp == tvp) { 3110 if (retain) 3111 error = -1; 3112 else if (fromnd.ni_dvp == tdvp && 3113 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3114 !memcmp(fromnd.ni_cnd.cn_nameptr, 3115 tond.ni_cnd.cn_nameptr, 3116 fromnd.ni_cnd.cn_namelen)) 3117 error = -1; 3118 } 3119 3120 #ifdef VERIFIED_EXEC 3121 if (!error) 3122 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l); 3123 #endif /* VERIFIED_EXEC */ 3124 3125 out: 3126 p = l->l_proc; 3127 if (!error) { 3128 VOP_LEASE(tdvp, l, p->p_ucred, LEASE_WRITE); 3129 if (fromnd.ni_dvp != tdvp) 3130 VOP_LEASE(fromnd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3131 if (tvp) { 3132 VOP_LEASE(tvp, l, p->p_ucred, LEASE_WRITE); 3133 } 3134 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3135 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3136 } else { 3137 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3138 if (tdvp == tvp) 3139 vrele(tdvp); 3140 else 3141 vput(tdvp); 3142 if (tvp) 3143 vput(tvp); 3144 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3145 vrele(fromnd.ni_dvp); 3146 vrele(fvp); 3147 } 3148 vrele(tond.ni_startdir); 3149 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3150 out1: 3151 vn_finished_write(mp, 0); 3152 if (fromnd.ni_startdir) 3153 vrele(fromnd.ni_startdir); 3154 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3155 return (error == -1 ? 0 : error); 3156 } 3157 3158 /* 3159 * Make a directory file. 3160 */ 3161 /* ARGSUSED */ 3162 int 3163 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3164 { 3165 struct sys_mkdir_args /* { 3166 syscallarg(const char *) path; 3167 syscallarg(int) mode; 3168 } */ *uap = v; 3169 struct proc *p = l->l_proc; 3170 struct mount *mp; 3171 struct vnode *vp; 3172 struct vattr vattr; 3173 int error; 3174 struct nameidata nd; 3175 3176 restart: 3177 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3178 SCARG(uap, path), l); 3179 if ((error = namei(&nd)) != 0) 3180 return (error); 3181 vp = nd.ni_vp; 3182 if (vp != NULL) { 3183 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3184 if (nd.ni_dvp == vp) 3185 vrele(nd.ni_dvp); 3186 else 3187 vput(nd.ni_dvp); 3188 vrele(vp); 3189 return (EEXIST); 3190 } 3191 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3192 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3193 if (nd.ni_dvp == vp) 3194 vrele(nd.ni_dvp); 3195 else 3196 vput(nd.ni_dvp); 3197 if ((error = vn_start_write(NULL, &mp, 3198 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3199 return (error); 3200 goto restart; 3201 } 3202 VATTR_NULL(&vattr); 3203 vattr.va_type = VDIR; 3204 vattr.va_mode = 3205 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3206 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3207 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3208 if (!error) 3209 vput(nd.ni_vp); 3210 vn_finished_write(mp, 0); 3211 return (error); 3212 } 3213 3214 /* 3215 * Remove a directory file. 3216 */ 3217 /* ARGSUSED */ 3218 int 3219 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3220 { 3221 struct sys_rmdir_args /* { 3222 syscallarg(const char *) path; 3223 } */ *uap = v; 3224 struct proc *p = l->l_proc; 3225 struct mount *mp; 3226 struct vnode *vp; 3227 int error; 3228 struct nameidata nd; 3229 3230 restart: 3231 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3232 SCARG(uap, path), l); 3233 if ((error = namei(&nd)) != 0) 3234 return (error); 3235 vp = nd.ni_vp; 3236 if (vp->v_type != VDIR) { 3237 error = ENOTDIR; 3238 goto out; 3239 } 3240 /* 3241 * No rmdir "." please. 3242 */ 3243 if (nd.ni_dvp == vp) { 3244 error = EINVAL; 3245 goto out; 3246 } 3247 /* 3248 * The root of a mounted filesystem cannot be deleted. 3249 */ 3250 if (vp->v_flag & VROOT) { 3251 error = EBUSY; 3252 goto out; 3253 } 3254 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3255 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3256 if (nd.ni_dvp == vp) 3257 vrele(nd.ni_dvp); 3258 else 3259 vput(nd.ni_dvp); 3260 vput(vp); 3261 if ((error = vn_start_write(NULL, &mp, 3262 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3263 return (error); 3264 goto restart; 3265 } 3266 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3267 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 3268 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3269 vn_finished_write(mp, 0); 3270 return (error); 3271 3272 out: 3273 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3274 if (nd.ni_dvp == vp) 3275 vrele(nd.ni_dvp); 3276 else 3277 vput(nd.ni_dvp); 3278 vput(vp); 3279 return (error); 3280 } 3281 3282 /* 3283 * Read a block of directory entries in a file system independent format. 3284 */ 3285 int 3286 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3287 { 3288 struct sys___getdents30_args /* { 3289 syscallarg(int) fd; 3290 syscallarg(char *) buf; 3291 syscallarg(size_t) count; 3292 } */ *uap = v; 3293 struct proc *p = l->l_proc; 3294 struct file *fp; 3295 int error, done; 3296 3297 /* getvnode() will use the descriptor for us */ 3298 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3299 return (error); 3300 if ((fp->f_flag & FREAD) == 0) { 3301 error = EBADF; 3302 goto out; 3303 } 3304 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3305 SCARG(uap, count), &done, l, 0, 0); 3306 #ifdef KTRACE 3307 if (!error && KTRPOINT(p, KTR_GENIO)) { 3308 struct iovec iov; 3309 iov.iov_base = SCARG(uap, buf); 3310 iov.iov_len = done; 3311 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3312 } 3313 #endif 3314 *retval = done; 3315 out: 3316 FILE_UNUSE(fp, l); 3317 return (error); 3318 } 3319 3320 /* 3321 * Set the mode mask for creation of filesystem nodes. 3322 */ 3323 int 3324 sys_umask(struct lwp *l, void *v, register_t *retval) 3325 { 3326 struct sys_umask_args /* { 3327 syscallarg(mode_t) newmask; 3328 } */ *uap = v; 3329 struct proc *p = l->l_proc; 3330 struct cwdinfo *cwdi; 3331 3332 cwdi = p->p_cwdi; 3333 *retval = cwdi->cwdi_cmask; 3334 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3335 return (0); 3336 } 3337 3338 /* 3339 * Void all references to file by ripping underlying filesystem 3340 * away from vnode. 3341 */ 3342 /* ARGSUSED */ 3343 int 3344 sys_revoke(struct lwp *l, void *v, register_t *retval) 3345 { 3346 struct sys_revoke_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 struct vattr vattr; 3353 int error; 3354 struct nameidata nd; 3355 3356 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3357 if ((error = namei(&nd)) != 0) 3358 return (error); 3359 vp = nd.ni_vp; 3360 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 3361 goto out; 3362 if (p->p_ucred->cr_uid != vattr.va_uid && 3363 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3364 goto out; 3365 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3366 goto out; 3367 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3368 VOP_REVOKE(vp, REVOKEALL); 3369 vn_finished_write(mp, 0); 3370 out: 3371 vrele(vp); 3372 return (error); 3373 } 3374 3375 /* 3376 * Convert a user file descriptor to a kernel file entry. 3377 */ 3378 int 3379 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3380 { 3381 struct vnode *vp; 3382 struct file *fp; 3383 3384 if ((fp = fd_getfile(fdp, fd)) == NULL) 3385 return (EBADF); 3386 3387 FILE_USE(fp); 3388 3389 if (fp->f_type != DTYPE_VNODE) { 3390 FILE_UNUSE(fp, NULL); 3391 return (EINVAL); 3392 } 3393 3394 vp = (struct vnode *)fp->f_data; 3395 if (vp->v_type == VBAD) { 3396 FILE_UNUSE(fp, NULL); 3397 return (EBADF); 3398 } 3399 3400 *fpp = fp; 3401 return (0); 3402 } 3403
Indexes created Thu Sep 25 16:09:42 GMT 2025