vfs_syscalls.c revision 1.233
1 /* $NetBSD: vfs_syscalls.c,v 1.233 2005/10/01 06:13:42 yamt 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 1.233 2005/10/01 06:13:42 yamt 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 proc *); 89 static int change_flags(struct vnode *, u_long, struct proc *); 90 static int change_mode(struct vnode *, int, struct proc *p); 91 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int); 92 static int change_utimes(struct vnode *vp, const struct timeval *, 93 struct proc *p); 94 static int rename_files(const char *, const char *, struct proc *, 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), p); 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, p)) != 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, p, 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 MNT_MAGICLINKS); 345 mp->mnt_flag |= SCARG(uap, flags) & 346 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 347 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 348 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 349 MNT_IGNORE | MNT_MAGICLINKS); 350 } 351 /* 352 * Mount the filesystem. 353 */ 354 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 355 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 356 #if defined(COMPAT_30) && defined(NFSSERVER) 357 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 358 int error2; 359 360 /* Update failed; let's try and see if it was an 361 * export request. */ 362 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 363 SCARG(uap, data), p); 364 365 /* Only update error code if the export request was 366 * understood but some problem occurred while 367 * processing it. */ 368 if (error2 != EJUSTRETURN) 369 error = error2; 370 } 371 #endif 372 if (mp->mnt_iflag & IMNT_WANTRDWR) 373 mp->mnt_flag &= ~MNT_RDONLY; 374 if (error) 375 mp->mnt_flag = flag; 376 mp->mnt_flag &=~ 377 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 378 mp->mnt_iflag &=~ IMNT_WANTRDWR; 379 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 380 if (mp->mnt_syncer == NULL) 381 error = vfs_allocate_syncvnode(mp); 382 } else { 383 if (mp->mnt_syncer != NULL) 384 vfs_deallocate_syncvnode(mp); 385 } 386 vfs_unbusy(mp); 387 VOP_UNLOCK(vp, 0); 388 vrele(vp); 389 return (error); 390 } 391 /* 392 * Put the new filesystem on the mount list after root. 393 */ 394 cache_purge(vp); 395 if (!error) { 396 mp->mnt_flag &=~ 397 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 398 mp->mnt_iflag &=~ IMNT_WANTRDWR; 399 vp->v_mountedhere = mp; 400 simple_lock(&mountlist_slock); 401 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 402 simple_unlock(&mountlist_slock); 403 checkdirs(vp); 404 VOP_UNLOCK(vp, 0); 405 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 406 error = vfs_allocate_syncvnode(mp); 407 vfs_unbusy(mp); 408 (void) VFS_STATVFS(mp, &mp->mnt_stat, p); 409 if ((error = VFS_START(mp, 0, p))) 410 vrele(vp); 411 } else { 412 vp->v_mountedhere = (struct mount *)0; 413 vfs->vfs_refcount--; 414 vfs_unbusy(mp); 415 free(mp, M_MOUNT); 416 vput(vp); 417 } 418 return (error); 419 } 420 421 /* 422 * Scan all active processes to see if any of them have a current 423 * or root directory onto which the new filesystem has just been 424 * mounted. If so, replace them with the new mount point. 425 */ 426 void 427 checkdirs(struct vnode *olddp) 428 { 429 struct cwdinfo *cwdi; 430 struct vnode *newdp; 431 struct proc *p; 432 433 if (olddp->v_usecount == 1) 434 return; 435 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 436 panic("mount: lost mount"); 437 proclist_lock_read(); 438 PROCLIST_FOREACH(p, &allproc) { 439 cwdi = p->p_cwdi; 440 if (!cwdi) 441 continue; 442 if (cwdi->cwdi_cdir == olddp) { 443 vrele(cwdi->cwdi_cdir); 444 VREF(newdp); 445 cwdi->cwdi_cdir = newdp; 446 } 447 if (cwdi->cwdi_rdir == olddp) { 448 vrele(cwdi->cwdi_rdir); 449 VREF(newdp); 450 cwdi->cwdi_rdir = newdp; 451 } 452 } 453 proclist_unlock_read(); 454 if (rootvnode == olddp) { 455 vrele(rootvnode); 456 VREF(newdp); 457 rootvnode = newdp; 458 } 459 vput(newdp); 460 } 461 462 /* 463 * Unmount a file system. 464 * 465 * Note: unmount takes a path to the vnode mounted on as argument, 466 * not special file (as before). 467 */ 468 /* ARGSUSED */ 469 int 470 sys_unmount(struct lwp *l, void *v, register_t *retval) 471 { 472 struct sys_unmount_args /* { 473 syscallarg(const char *) path; 474 syscallarg(int) flags; 475 } */ *uap = v; 476 struct proc *p = l->l_proc; 477 struct vnode *vp; 478 struct mount *mp; 479 int error; 480 struct nameidata nd; 481 482 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 483 SCARG(uap, path), p); 484 if ((error = namei(&nd)) != 0) 485 return (error); 486 vp = nd.ni_vp; 487 mp = vp->v_mount; 488 489 /* 490 * Only root, or the user that did the original mount is 491 * permitted to unmount this filesystem. 492 */ 493 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 494 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 495 vput(vp); 496 return (error); 497 } 498 499 /* 500 * Don't allow unmounting the root file system. 501 */ 502 if (mp->mnt_flag & MNT_ROOTFS) { 503 vput(vp); 504 return (EINVAL); 505 } 506 507 /* 508 * Must be the root of the filesystem 509 */ 510 if ((vp->v_flag & VROOT) == 0) { 511 vput(vp); 512 return (EINVAL); 513 } 514 vput(vp); 515 516 /* 517 * XXX Freeze syncer. Must do this before locking the 518 * mount point. See dounmount() for details. 519 */ 520 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 521 522 if (vfs_busy(mp, 0, 0)) { 523 lockmgr(&syncer_lock, LK_RELEASE, NULL); 524 return (EBUSY); 525 } 526 527 return (dounmount(mp, SCARG(uap, flags), p)); 528 } 529 530 /* 531 * Do the actual file system unmount. File system is assumed to have been 532 * marked busy by the caller. 533 */ 534 int 535 dounmount(struct mount *mp, int flags, struct proc *p) 536 { 537 struct vnode *coveredvp; 538 int error; 539 int async; 540 int used_syncer; 541 542 simple_lock(&mountlist_slock); 543 vfs_unbusy(mp); 544 used_syncer = (mp->mnt_syncer != NULL); 545 546 /* 547 * XXX Syncer must be frozen when we get here. This should really 548 * be done on a per-mountpoint basis, but especially the softdep 549 * code possibly called from the syncer doesn't exactly work on a 550 * per-mountpoint basis, so the softdep code would become a maze 551 * of vfs_busy() calls. 552 * 553 * The caller of dounmount() must acquire syncer_lock because 554 * the syncer itself acquires locks in syncer_lock -> vfs_busy 555 * order, and we must preserve that order to avoid deadlock. 556 * 557 * So, if the file system did not use the syncer, now is 558 * the time to release the syncer_lock. 559 */ 560 if (used_syncer == 0) 561 lockmgr(&syncer_lock, LK_RELEASE, NULL); 562 563 mp->mnt_iflag |= IMNT_UNMOUNT; 564 mp->mnt_unmounter = p; 565 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 566 vn_start_write(NULL, &mp, V_WAIT); 567 568 async = mp->mnt_flag & MNT_ASYNC; 569 mp->mnt_flag &= ~MNT_ASYNC; 570 cache_purgevfs(mp); /* remove cache entries for this file sys */ 571 if (mp->mnt_syncer != NULL) 572 vfs_deallocate_syncvnode(mp); 573 error = 0; 574 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 575 #if NFSS > 0 576 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 577 #endif 578 if (error == 0) 579 error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p); 580 } 581 if (error == 0 || (flags & MNT_FORCE)) 582 error = VFS_UNMOUNT(mp, flags, p); 583 vn_finished_write(mp, 0); 584 simple_lock(&mountlist_slock); 585 if (error) { 586 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 587 (void) vfs_allocate_syncvnode(mp); 588 mp->mnt_iflag &= ~IMNT_UNMOUNT; 589 mp->mnt_unmounter = NULL; 590 mp->mnt_flag |= async; 591 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 592 &mountlist_slock); 593 if (used_syncer) 594 lockmgr(&syncer_lock, LK_RELEASE, NULL); 595 simple_lock(&mp->mnt_slock); 596 while (mp->mnt_wcnt > 0) { 597 wakeup(mp); 598 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 599 0, &mp->mnt_slock); 600 } 601 simple_unlock(&mp->mnt_slock); 602 return (error); 603 } 604 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 605 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 606 coveredvp->v_mountedhere = NULL; 607 vrele(coveredvp); 608 } 609 mp->mnt_op->vfs_refcount--; 610 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 611 panic("unmount: dangling vnode"); 612 mp->mnt_iflag |= IMNT_GONE; 613 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 614 if (used_syncer) 615 lockmgr(&syncer_lock, LK_RELEASE, NULL); 616 simple_lock(&mp->mnt_slock); 617 while (mp->mnt_wcnt > 0) { 618 wakeup(mp); 619 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 620 } 621 simple_unlock(&mp->mnt_slock); 622 vfs_hooks_unmount(mp); 623 free(mp, M_MOUNT); 624 return (0); 625 } 626 627 /* 628 * Sync each mounted filesystem. 629 */ 630 #ifdef DEBUG 631 int syncprt = 0; 632 struct ctldebug debug0 = { "syncprt", &syncprt }; 633 #endif 634 635 /* ARGSUSED */ 636 int 637 sys_sync(struct lwp *l, void *v, register_t *retval) 638 { 639 struct mount *mp, *nmp; 640 int asyncflag; 641 struct proc *p = l == NULL ? &proc0 : l->l_proc; 642 643 simple_lock(&mountlist_slock); 644 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 645 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 646 nmp = mp->mnt_list.cqe_prev; 647 continue; 648 } 649 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 650 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 651 asyncflag = mp->mnt_flag & MNT_ASYNC; 652 mp->mnt_flag &= ~MNT_ASYNC; 653 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p); 654 if (asyncflag) 655 mp->mnt_flag |= MNT_ASYNC; 656 vn_finished_write(mp, 0); 657 } 658 simple_lock(&mountlist_slock); 659 nmp = mp->mnt_list.cqe_prev; 660 vfs_unbusy(mp); 661 662 } 663 simple_unlock(&mountlist_slock); 664 #ifdef DEBUG 665 if (syncprt) 666 vfs_bufstats(); 667 #endif /* DEBUG */ 668 return (0); 669 } 670 671 /* 672 * Change filesystem quotas. 673 */ 674 /* ARGSUSED */ 675 int 676 sys_quotactl(struct lwp *l, void *v, register_t *retval) 677 { 678 struct sys_quotactl_args /* { 679 syscallarg(const char *) path; 680 syscallarg(int) cmd; 681 syscallarg(int) uid; 682 syscallarg(caddr_t) arg; 683 } */ *uap = v; 684 struct proc *p = l->l_proc; 685 struct mount *mp; 686 int error; 687 struct nameidata nd; 688 689 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 690 if ((error = namei(&nd)) != 0) 691 return (error); 692 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 693 vrele(nd.ni_vp); 694 if (error) 695 return (error); 696 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 697 SCARG(uap, arg), p); 698 vn_finished_write(mp, 0); 699 return (error); 700 } 701 702 int 703 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags, 704 int root) 705 { 706 struct cwdinfo *cwdi = p->p_cwdi; 707 int error = 0; 708 709 /* 710 * If MNT_NOWAIT or MNT_LAZY is specified, do not 711 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 712 * overrides MNT_NOWAIT. 713 */ 714 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 715 (flags != MNT_WAIT && flags != 0)) { 716 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 717 goto done; 718 } 719 720 /* Get the filesystem stats now */ 721 memset(sp, 0, sizeof(*sp)); 722 if ((error = VFS_STATVFS(mp, sp, p)) != 0) { 723 return error; 724 } 725 726 if (cwdi->cwdi_rdir == NULL) 727 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 728 done: 729 if (cwdi->cwdi_rdir != NULL) { 730 size_t len; 731 char *bp; 732 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 733 if (!path) 734 return ENOMEM; 735 736 bp = path + MAXPATHLEN; 737 *--bp = '\0'; 738 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 739 MAXPATHLEN / 2, 0, p); 740 if (error) { 741 free(path, M_TEMP); 742 return error; 743 } 744 len = strlen(bp); 745 /* 746 * for mount points that are below our root, we can see 747 * them, so we fix up the pathname and return them. The 748 * rest we cannot see, so we don't allow viewing the 749 * data. 750 */ 751 if (strncmp(bp, sp->f_mntonname, len) == 0) { 752 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 753 sizeof(sp->f_mntonname)); 754 if (sp->f_mntonname[0] == '\0') 755 (void)strlcpy(sp->f_mntonname, "/", 756 sizeof(sp->f_mntonname)); 757 } else { 758 if (root) 759 (void)strlcpy(sp->f_mntonname, "/", 760 sizeof(sp->f_mntonname)); 761 else 762 error = EPERM; 763 } 764 free(path, M_TEMP); 765 } 766 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 767 return error; 768 } 769 770 /* 771 * Get filesystem statistics. 772 */ 773 /* ARGSUSED */ 774 int 775 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 776 { 777 struct sys_statvfs1_args /* { 778 syscallarg(const char *) path; 779 syscallarg(struct statvfs *) buf; 780 syscallarg(int) flags; 781 } */ *uap = v; 782 struct proc *p = l->l_proc; 783 struct mount *mp; 784 struct statvfs sbuf; 785 int error; 786 struct nameidata nd; 787 788 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 789 if ((error = namei(&nd)) != 0) 790 return error; 791 mp = nd.ni_vp->v_mount; 792 vrele(nd.ni_vp); 793 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 794 return error; 795 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 796 } 797 798 /* 799 * Get filesystem statistics. 800 */ 801 /* ARGSUSED */ 802 int 803 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 804 { 805 struct sys_fstatvfs1_args /* { 806 syscallarg(int) fd; 807 syscallarg(struct statvfs *) buf; 808 syscallarg(int) flags; 809 } */ *uap = v; 810 struct proc *p = l->l_proc; 811 struct file *fp; 812 struct mount *mp; 813 struct statvfs sbuf; 814 int error; 815 816 /* getvnode() will use the descriptor for us */ 817 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 818 return (error); 819 mp = ((struct vnode *)fp->f_data)->v_mount; 820 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 821 goto out; 822 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 823 out: 824 FILE_UNUSE(fp, p); 825 return error; 826 } 827 828 829 /* 830 * Get statistics on all filesystems. 831 */ 832 int 833 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 834 { 835 struct sys_getvfsstat_args /* { 836 syscallarg(struct statvfs *) buf; 837 syscallarg(size_t) bufsize; 838 syscallarg(int) flags; 839 } */ *uap = v; 840 int root = 0; 841 struct proc *p = l->l_proc; 842 struct mount *mp, *nmp; 843 struct statvfs sbuf; 844 struct statvfs *sfsp; 845 size_t count, maxcount; 846 int error = 0; 847 848 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 849 sfsp = SCARG(uap, buf); 850 simple_lock(&mountlist_slock); 851 count = 0; 852 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 853 mp = nmp) { 854 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 855 nmp = CIRCLEQ_NEXT(mp, mnt_list); 856 continue; 857 } 858 if (sfsp && count < maxcount) { 859 error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0); 860 if (error) { 861 simple_lock(&mountlist_slock); 862 nmp = CIRCLEQ_NEXT(mp, mnt_list); 863 vfs_unbusy(mp); 864 continue; 865 } 866 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 867 if (error) { 868 vfs_unbusy(mp); 869 return (error); 870 } 871 sfsp++; 872 root |= strcmp(sbuf.f_mntonname, "/") == 0; 873 } 874 count++; 875 simple_lock(&mountlist_slock); 876 nmp = CIRCLEQ_NEXT(mp, mnt_list); 877 vfs_unbusy(mp); 878 } 879 simple_unlock(&mountlist_slock); 880 if (root == 0 && p->p_cwdi->cwdi_rdir) { 881 /* 882 * fake a root entry 883 */ 884 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p, 885 SCARG(uap, flags), 1)) != 0) 886 return error; 887 if (sfsp) 888 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 889 count++; 890 } 891 if (sfsp && count > maxcount) 892 *retval = maxcount; 893 else 894 *retval = count; 895 return error; 896 } 897 898 /* 899 * Change current working directory to a given file descriptor. 900 */ 901 /* ARGSUSED */ 902 int 903 sys_fchdir(struct lwp *l, void *v, register_t *retval) 904 { 905 struct sys_fchdir_args /* { 906 syscallarg(int) fd; 907 } */ *uap = v; 908 struct proc *p = l->l_proc; 909 struct filedesc *fdp = p->p_fd; 910 struct cwdinfo *cwdi = p->p_cwdi; 911 struct vnode *vp, *tdp; 912 struct mount *mp; 913 struct file *fp; 914 int error; 915 916 /* getvnode() will use the descriptor for us */ 917 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 918 return (error); 919 vp = (struct vnode *)fp->f_data; 920 921 VREF(vp); 922 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 923 if (vp->v_type != VDIR) 924 error = ENOTDIR; 925 else 926 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 927 while (!error && (mp = vp->v_mountedhere) != NULL) { 928 if (vfs_busy(mp, 0, 0)) 929 continue; 930 error = VFS_ROOT(mp, &tdp); 931 vfs_unbusy(mp); 932 if (error) 933 break; 934 vput(vp); 935 vp = tdp; 936 } 937 if (error) { 938 vput(vp); 939 goto out; 940 } 941 VOP_UNLOCK(vp, 0); 942 943 /* 944 * Disallow changing to a directory not under the process's 945 * current root directory (if there is one). 946 */ 947 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) { 948 vrele(vp); 949 error = EPERM; /* operation not permitted */ 950 goto out; 951 } 952 953 vrele(cwdi->cwdi_cdir); 954 cwdi->cwdi_cdir = vp; 955 out: 956 FILE_UNUSE(fp, p); 957 return (error); 958 } 959 960 /* 961 * Change this process's notion of the root directory to a given file 962 * descriptor. 963 */ 964 int 965 sys_fchroot(struct lwp *l, void *v, register_t *retval) 966 { 967 struct sys_fchroot_args *uap = v; 968 struct proc *p = l->l_proc; 969 struct filedesc *fdp = p->p_fd; 970 struct cwdinfo *cwdi = p->p_cwdi; 971 struct vnode *vp; 972 struct file *fp; 973 int error; 974 975 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 976 return error; 977 /* getvnode() will use the descriptor for us */ 978 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 979 return error; 980 vp = (struct vnode *) fp->f_data; 981 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 982 if (vp->v_type != VDIR) 983 error = ENOTDIR; 984 else 985 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 986 VOP_UNLOCK(vp, 0); 987 if (error) 988 goto out; 989 VREF(vp); 990 991 /* 992 * Prevent escaping from chroot by putting the root under 993 * the working directory. Silently chdir to / if we aren't 994 * already there. 995 */ 996 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 997 /* 998 * XXX would be more failsafe to change directory to a 999 * deadfs node here instead 1000 */ 1001 vrele(cwdi->cwdi_cdir); 1002 VREF(vp); 1003 cwdi->cwdi_cdir = vp; 1004 } 1005 1006 if (cwdi->cwdi_rdir != NULL) 1007 vrele(cwdi->cwdi_rdir); 1008 cwdi->cwdi_rdir = vp; 1009 out: 1010 FILE_UNUSE(fp, p); 1011 return (error); 1012 } 1013 1014 /* 1015 * Change current working directory (``.''). 1016 */ 1017 /* ARGSUSED */ 1018 int 1019 sys_chdir(struct lwp *l, void *v, register_t *retval) 1020 { 1021 struct sys_chdir_args /* { 1022 syscallarg(const char *) path; 1023 } */ *uap = v; 1024 struct proc *p = l->l_proc; 1025 struct cwdinfo *cwdi = p->p_cwdi; 1026 int error; 1027 struct nameidata nd; 1028 1029 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1030 SCARG(uap, path), p); 1031 if ((error = change_dir(&nd, p)) != 0) 1032 return (error); 1033 vrele(cwdi->cwdi_cdir); 1034 cwdi->cwdi_cdir = nd.ni_vp; 1035 return (0); 1036 } 1037 1038 /* 1039 * Change notion of root (``/'') directory. 1040 */ 1041 /* ARGSUSED */ 1042 int 1043 sys_chroot(struct lwp *l, void *v, register_t *retval) 1044 { 1045 struct sys_chroot_args /* { 1046 syscallarg(const char *) path; 1047 } */ *uap = v; 1048 struct proc *p = l->l_proc; 1049 struct cwdinfo *cwdi = p->p_cwdi; 1050 struct vnode *vp; 1051 int error; 1052 struct nameidata nd; 1053 1054 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1055 return (error); 1056 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1057 SCARG(uap, path), p); 1058 if ((error = change_dir(&nd, p)) != 0) 1059 return (error); 1060 if (cwdi->cwdi_rdir != NULL) 1061 vrele(cwdi->cwdi_rdir); 1062 vp = nd.ni_vp; 1063 cwdi->cwdi_rdir = vp; 1064 1065 /* 1066 * Prevent escaping from chroot by putting the root under 1067 * the working directory. Silently chdir to / if we aren't 1068 * already there. 1069 */ 1070 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 1071 /* 1072 * XXX would be more failsafe to change directory to a 1073 * deadfs node here instead 1074 */ 1075 vrele(cwdi->cwdi_cdir); 1076 VREF(vp); 1077 cwdi->cwdi_cdir = vp; 1078 } 1079 1080 return (0); 1081 } 1082 1083 /* 1084 * Common routine for chroot and chdir. 1085 */ 1086 static int 1087 change_dir(struct nameidata *ndp, struct proc *p) 1088 { 1089 struct vnode *vp; 1090 int error; 1091 1092 if ((error = namei(ndp)) != 0) 1093 return (error); 1094 vp = ndp->ni_vp; 1095 if (vp->v_type != VDIR) 1096 error = ENOTDIR; 1097 else 1098 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 1099 1100 if (error) 1101 vput(vp); 1102 else 1103 VOP_UNLOCK(vp, 0); 1104 return (error); 1105 } 1106 1107 /* 1108 * Check permissions, allocate an open file structure, 1109 * and call the device open routine if any. 1110 */ 1111 int 1112 sys_open(struct lwp *l, void *v, register_t *retval) 1113 { 1114 struct sys_open_args /* { 1115 syscallarg(const char *) path; 1116 syscallarg(int) flags; 1117 syscallarg(int) mode; 1118 } */ *uap = v; 1119 struct proc *p = l->l_proc; 1120 struct cwdinfo *cwdi = p->p_cwdi; 1121 struct filedesc *fdp = p->p_fd; 1122 struct file *fp; 1123 struct vnode *vp; 1124 int flags, cmode; 1125 int type, indx, error; 1126 struct flock lf; 1127 struct nameidata nd; 1128 1129 flags = FFLAGS(SCARG(uap, flags)); 1130 if ((flags & (FREAD | FWRITE)) == 0) 1131 return (EINVAL); 1132 /* falloc() will use the file descriptor for us */ 1133 if ((error = falloc(p, &fp, &indx)) != 0) 1134 return (error); 1135 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1136 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1137 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1138 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1139 FILE_UNUSE(fp, p); 1140 fdp->fd_ofiles[indx] = NULL; 1141 ffree(fp); 1142 if ((error == EDUPFD || error == EMOVEFD) && 1143 l->l_dupfd >= 0 && /* XXX from fdopen */ 1144 (error = 1145 dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) { 1146 *retval = indx; 1147 return (0); 1148 } 1149 if (error == ERESTART) 1150 error = EINTR; 1151 fdremove(fdp, indx); 1152 return (error); 1153 } 1154 l->l_dupfd = 0; 1155 vp = nd.ni_vp; 1156 fp->f_flag = flags & FMASK; 1157 fp->f_type = DTYPE_VNODE; 1158 fp->f_ops = &vnops; 1159 fp->f_data = vp; 1160 if (flags & (O_EXLOCK | O_SHLOCK)) { 1161 lf.l_whence = SEEK_SET; 1162 lf.l_start = 0; 1163 lf.l_len = 0; 1164 if (flags & O_EXLOCK) 1165 lf.l_type = F_WRLCK; 1166 else 1167 lf.l_type = F_RDLCK; 1168 type = F_FLOCK; 1169 if ((flags & FNONBLOCK) == 0) 1170 type |= F_WAIT; 1171 VOP_UNLOCK(vp, 0); 1172 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1173 if (error) { 1174 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1175 FILE_UNUSE(fp, p); 1176 ffree(fp); 1177 fdremove(fdp, indx); 1178 return (error); 1179 } 1180 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1181 fp->f_flag |= FHASLOCK; 1182 } 1183 VOP_UNLOCK(vp, 0); 1184 *retval = indx; 1185 FILE_SET_MATURE(fp); 1186 FILE_UNUSE(fp, p); 1187 return (0); 1188 } 1189 1190 /* 1191 * Get file handle system call 1192 */ 1193 int 1194 sys_getfh(struct lwp *l, void *v, register_t *retval) 1195 { 1196 struct sys_getfh_args /* { 1197 syscallarg(char *) fname; 1198 syscallarg(fhandle_t *) fhp; 1199 } */ *uap = v; 1200 struct proc *p = l->l_proc; 1201 struct vnode *vp; 1202 fhandle_t fh; 1203 int error; 1204 struct nameidata nd; 1205 1206 /* 1207 * Must be super user 1208 */ 1209 error = suser(p->p_ucred, &p->p_acflag); 1210 if (error) 1211 return (error); 1212 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1213 SCARG(uap, fname), p); 1214 error = namei(&nd); 1215 if (error) 1216 return (error); 1217 vp = nd.ni_vp; 1218 if (vp->v_mount->mnt_op->vfs_vptofh == NULL) 1219 return EOPNOTSUPP; 1220 memset(&fh, 0, sizeof(fh)); 1221 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1222 error = VFS_VPTOFH(vp, &fh.fh_fid); 1223 vput(vp); 1224 if (error) 1225 return (error); 1226 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1227 return (error); 1228 } 1229 1230 /* 1231 * Open a file given a file handle. 1232 * 1233 * Check permissions, allocate an open file structure, 1234 * and call the device open routine if any. 1235 */ 1236 int 1237 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1238 { 1239 struct sys_fhopen_args /* { 1240 syscallarg(const fhandle_t *) fhp; 1241 syscallarg(int) flags; 1242 } */ *uap = v; 1243 struct proc *p = l->l_proc; 1244 struct filedesc *fdp = p->p_fd; 1245 struct file *fp; 1246 struct vnode *vp = NULL; 1247 struct mount *mp; 1248 struct ucred *cred = p->p_ucred; 1249 int flags; 1250 struct file *nfp; 1251 int type, indx, error=0; 1252 struct flock lf; 1253 struct vattr va; 1254 fhandle_t fh; 1255 1256 /* 1257 * Must be super user 1258 */ 1259 if ((error = suser(p->p_ucred, &p->p_acflag))) 1260 return (error); 1261 1262 flags = FFLAGS(SCARG(uap, flags)); 1263 if ((flags & (FREAD | FWRITE)) == 0) 1264 return (EINVAL); 1265 if ((flags & O_CREAT)) 1266 return (EINVAL); 1267 /* falloc() will use the file descriptor for us */ 1268 if ((error = falloc(p, &nfp, &indx)) != 0) 1269 return (error); 1270 fp = nfp; 1271 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1272 goto bad; 1273 1274 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1275 error = ESTALE; 1276 goto bad; 1277 } 1278 1279 if (mp->mnt_op->vfs_fhtovp == NULL) { 1280 error = EOPNOTSUPP; 1281 goto bad; 1282 } 1283 1284 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1285 vp = NULL; /* most likely unnecessary sanity for bad: */ 1286 goto bad; 1287 } 1288 1289 /* Now do an effective vn_open */ 1290 1291 if (vp->v_type == VSOCK) { 1292 error = EOPNOTSUPP; 1293 goto bad; 1294 } 1295 if (flags & FREAD) { 1296 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 1297 goto bad; 1298 } 1299 if (flags & (FWRITE | O_TRUNC)) { 1300 if (vp->v_type == VDIR) { 1301 error = EISDIR; 1302 goto bad; 1303 } 1304 if ((error = vn_writechk(vp)) != 0 || 1305 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 1306 goto bad; 1307 } 1308 if (flags & O_TRUNC) { 1309 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1310 goto bad; 1311 VOP_UNLOCK(vp, 0); /* XXX */ 1312 VOP_LEASE(vp, p, cred, LEASE_WRITE); 1313 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1314 VATTR_NULL(&va); 1315 va.va_size = 0; 1316 error = VOP_SETATTR(vp, &va, cred, p); 1317 vn_finished_write(mp, 0); 1318 if (error) 1319 goto bad; 1320 } 1321 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) 1322 goto bad; 1323 if (vp->v_type == VREG && 1324 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1325 error = EIO; 1326 goto bad; 1327 } 1328 if (flags & FWRITE) 1329 vp->v_writecount++; 1330 1331 /* done with modified vn_open, now finish what sys_open does. */ 1332 1333 fp->f_flag = flags & FMASK; 1334 fp->f_type = DTYPE_VNODE; 1335 fp->f_ops = &vnops; 1336 fp->f_data = vp; 1337 if (flags & (O_EXLOCK | O_SHLOCK)) { 1338 lf.l_whence = SEEK_SET; 1339 lf.l_start = 0; 1340 lf.l_len = 0; 1341 if (flags & O_EXLOCK) 1342 lf.l_type = F_WRLCK; 1343 else 1344 lf.l_type = F_RDLCK; 1345 type = F_FLOCK; 1346 if ((flags & FNONBLOCK) == 0) 1347 type |= F_WAIT; 1348 VOP_UNLOCK(vp, 0); 1349 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1350 if (error) { 1351 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1352 FILE_UNUSE(fp, p); 1353 ffree(fp); 1354 fdremove(fdp, indx); 1355 return (error); 1356 } 1357 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1358 fp->f_flag |= FHASLOCK; 1359 } 1360 VOP_UNLOCK(vp, 0); 1361 *retval = indx; 1362 FILE_SET_MATURE(fp); 1363 FILE_UNUSE(fp, p); 1364 return (0); 1365 1366 bad: 1367 FILE_UNUSE(fp, p); 1368 ffree(fp); 1369 fdremove(fdp, indx); 1370 if (vp != NULL) 1371 vput(vp); 1372 return (error); 1373 } 1374 1375 /* ARGSUSED */ 1376 int 1377 sys_fhstat(struct lwp *l, void *v, register_t *retval) 1378 { 1379 struct sys_fhstat_args /* { 1380 syscallarg(const fhandle_t *) fhp; 1381 syscallarg(struct stat *) sb; 1382 } */ *uap = v; 1383 struct proc *p = l->l_proc; 1384 struct stat sb; 1385 int error; 1386 fhandle_t fh; 1387 struct mount *mp; 1388 struct vnode *vp; 1389 1390 /* 1391 * Must be super user 1392 */ 1393 if ((error = suser(p->p_ucred, &p->p_acflag))) 1394 return (error); 1395 1396 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1397 return (error); 1398 1399 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1400 return (ESTALE); 1401 if (mp->mnt_op->vfs_fhtovp == NULL) 1402 return EOPNOTSUPP; 1403 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1404 return (error); 1405 error = vn_stat(vp, &sb, p); 1406 vput(vp); 1407 if (error) 1408 return (error); 1409 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1410 return (error); 1411 } 1412 1413 /* ARGSUSED */ 1414 int 1415 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1416 { 1417 struct sys_fhstatvfs1_args /* { 1418 syscallarg(const fhandle_t *) fhp; 1419 syscallarg(struct statvfs *) buf; 1420 syscallarg(int) flags; 1421 } */ *uap = v; 1422 struct proc *p = l->l_proc; 1423 struct statvfs sbuf; 1424 fhandle_t fh; 1425 struct mount *mp; 1426 struct vnode *vp; 1427 int error; 1428 1429 /* 1430 * Must be super user 1431 */ 1432 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1433 return error; 1434 1435 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1436 return error; 1437 1438 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1439 return ESTALE; 1440 if (mp->mnt_op->vfs_fhtovp == NULL) 1441 return EOPNOTSUPP; 1442 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1443 return error; 1444 1445 mp = vp->v_mount; 1446 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) { 1447 vput(vp); 1448 return error; 1449 } 1450 vput(vp); 1451 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1452 } 1453 1454 /* 1455 * Create a special file. 1456 */ 1457 /* ARGSUSED */ 1458 int 1459 sys_mknod(struct lwp *l, void *v, register_t *retval) 1460 { 1461 struct sys_mknod_args /* { 1462 syscallarg(const char *) path; 1463 syscallarg(int) mode; 1464 syscallarg(int) dev; 1465 } */ *uap = v; 1466 struct proc *p = l->l_proc; 1467 struct vnode *vp; 1468 struct mount *mp; 1469 struct vattr vattr; 1470 int error; 1471 int whiteout = 0; 1472 struct nameidata nd; 1473 1474 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1475 return (error); 1476 restart: 1477 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1478 if ((error = namei(&nd)) != 0) 1479 return (error); 1480 vp = nd.ni_vp; 1481 if (vp != NULL) 1482 error = EEXIST; 1483 else { 1484 VATTR_NULL(&vattr); 1485 vattr.va_mode = 1486 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1487 vattr.va_rdev = SCARG(uap, dev); 1488 whiteout = 0; 1489 1490 switch (SCARG(uap, mode) & S_IFMT) { 1491 case S_IFMT: /* used by badsect to flag bad sectors */ 1492 vattr.va_type = VBAD; 1493 break; 1494 case S_IFCHR: 1495 vattr.va_type = VCHR; 1496 break; 1497 case S_IFBLK: 1498 vattr.va_type = VBLK; 1499 break; 1500 case S_IFWHT: 1501 whiteout = 1; 1502 break; 1503 default: 1504 error = EINVAL; 1505 break; 1506 } 1507 } 1508 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1509 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1510 if (nd.ni_dvp == vp) 1511 vrele(nd.ni_dvp); 1512 else 1513 vput(nd.ni_dvp); 1514 if (vp) 1515 vrele(vp); 1516 if ((error = vn_start_write(NULL, &mp, 1517 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1518 return (error); 1519 goto restart; 1520 } 1521 if (!error) { 1522 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1523 if (whiteout) { 1524 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1525 if (error) 1526 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1527 vput(nd.ni_dvp); 1528 } else { 1529 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1530 &nd.ni_cnd, &vattr); 1531 if (error == 0) 1532 vput(nd.ni_vp); 1533 } 1534 } else { 1535 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1536 if (nd.ni_dvp == vp) 1537 vrele(nd.ni_dvp); 1538 else 1539 vput(nd.ni_dvp); 1540 if (vp) 1541 vrele(vp); 1542 } 1543 vn_finished_write(mp, 0); 1544 return (error); 1545 } 1546 1547 /* 1548 * Create a named pipe. 1549 */ 1550 /* ARGSUSED */ 1551 int 1552 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1553 { 1554 struct sys_mkfifo_args /* { 1555 syscallarg(const char *) path; 1556 syscallarg(int) mode; 1557 } */ *uap = v; 1558 struct proc *p = l->l_proc; 1559 struct mount *mp; 1560 struct vattr vattr; 1561 int error; 1562 struct nameidata nd; 1563 1564 restart: 1565 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1566 if ((error = namei(&nd)) != 0) 1567 return (error); 1568 if (nd.ni_vp != NULL) { 1569 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1570 if (nd.ni_dvp == nd.ni_vp) 1571 vrele(nd.ni_dvp); 1572 else 1573 vput(nd.ni_dvp); 1574 vrele(nd.ni_vp); 1575 return (EEXIST); 1576 } 1577 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1578 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1579 if (nd.ni_dvp == nd.ni_vp) 1580 vrele(nd.ni_dvp); 1581 else 1582 vput(nd.ni_dvp); 1583 if (nd.ni_vp) 1584 vrele(nd.ni_vp); 1585 if ((error = vn_start_write(NULL, &mp, 1586 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1587 return (error); 1588 goto restart; 1589 } 1590 VATTR_NULL(&vattr); 1591 vattr.va_type = VFIFO; 1592 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1593 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1594 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1595 if (error == 0) 1596 vput(nd.ni_vp); 1597 vn_finished_write(mp, 0); 1598 return (error); 1599 } 1600 1601 /* 1602 * Make a hard file link. 1603 */ 1604 /* ARGSUSED */ 1605 int 1606 sys_link(struct lwp *l, void *v, register_t *retval) 1607 { 1608 struct sys_link_args /* { 1609 syscallarg(const char *) path; 1610 syscallarg(const char *) link; 1611 } */ *uap = v; 1612 struct proc *p = l->l_proc; 1613 struct vnode *vp; 1614 struct mount *mp; 1615 struct nameidata nd; 1616 int error; 1617 1618 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1619 if ((error = namei(&nd)) != 0) 1620 return (error); 1621 vp = nd.ni_vp; 1622 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1623 vrele(vp); 1624 return (error); 1625 } 1626 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1627 if ((error = namei(&nd)) != 0) 1628 goto out; 1629 if (nd.ni_vp) { 1630 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1631 if (nd.ni_dvp == nd.ni_vp) 1632 vrele(nd.ni_dvp); 1633 else 1634 vput(nd.ni_dvp); 1635 vrele(nd.ni_vp); 1636 error = EEXIST; 1637 goto out; 1638 } 1639 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1640 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1641 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1642 out: 1643 vrele(vp); 1644 vn_finished_write(mp, 0); 1645 return (error); 1646 } 1647 1648 /* 1649 * Make a symbolic link. 1650 */ 1651 /* ARGSUSED */ 1652 int 1653 sys_symlink(struct lwp *l, void *v, register_t *retval) 1654 { 1655 struct sys_symlink_args /* { 1656 syscallarg(const char *) path; 1657 syscallarg(const char *) link; 1658 } */ *uap = v; 1659 struct proc *p = l->l_proc; 1660 struct mount *mp; 1661 struct vattr vattr; 1662 char *path; 1663 int error; 1664 struct nameidata nd; 1665 1666 path = PNBUF_GET(); 1667 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1668 if (error) 1669 goto out; 1670 restart: 1671 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1672 if ((error = namei(&nd)) != 0) 1673 goto out; 1674 if (nd.ni_vp) { 1675 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1676 if (nd.ni_dvp == nd.ni_vp) 1677 vrele(nd.ni_dvp); 1678 else 1679 vput(nd.ni_dvp); 1680 vrele(nd.ni_vp); 1681 error = EEXIST; 1682 goto out; 1683 } 1684 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1685 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1686 if (nd.ni_dvp == nd.ni_vp) 1687 vrele(nd.ni_dvp); 1688 else 1689 vput(nd.ni_dvp); 1690 if ((error = vn_start_write(NULL, &mp, 1691 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1692 return (error); 1693 goto restart; 1694 } 1695 VATTR_NULL(&vattr); 1696 vattr.va_type = VLNK; 1697 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1698 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1699 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1700 if (error == 0) 1701 vput(nd.ni_vp); 1702 vn_finished_write(mp, 0); 1703 out: 1704 PNBUF_PUT(path); 1705 return (error); 1706 } 1707 1708 /* 1709 * Delete a whiteout from the filesystem. 1710 */ 1711 /* ARGSUSED */ 1712 int 1713 sys_undelete(struct lwp *l, void *v, register_t *retval) 1714 { 1715 struct sys_undelete_args /* { 1716 syscallarg(const char *) path; 1717 } */ *uap = v; 1718 struct proc *p = l->l_proc; 1719 int error; 1720 struct mount *mp; 1721 struct nameidata nd; 1722 1723 restart: 1724 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1725 SCARG(uap, path), p); 1726 error = namei(&nd); 1727 if (error) 1728 return (error); 1729 1730 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1731 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1732 if (nd.ni_dvp == nd.ni_vp) 1733 vrele(nd.ni_dvp); 1734 else 1735 vput(nd.ni_dvp); 1736 if (nd.ni_vp) 1737 vrele(nd.ni_vp); 1738 return (EEXIST); 1739 } 1740 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1741 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1742 if (nd.ni_dvp == nd.ni_vp) 1743 vrele(nd.ni_dvp); 1744 else 1745 vput(nd.ni_dvp); 1746 if ((error = vn_start_write(NULL, &mp, 1747 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1748 return (error); 1749 goto restart; 1750 } 1751 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1752 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1753 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1754 vput(nd.ni_dvp); 1755 vn_finished_write(mp, 0); 1756 return (error); 1757 } 1758 1759 /* 1760 * Delete a name from the filesystem. 1761 */ 1762 /* ARGSUSED */ 1763 int 1764 sys_unlink(struct lwp *l, void *v, register_t *retval) 1765 { 1766 struct sys_unlink_args /* { 1767 syscallarg(const char *) path; 1768 } */ *uap = v; 1769 struct proc *p = l->l_proc; 1770 struct mount *mp; 1771 struct vnode *vp; 1772 int error; 1773 struct nameidata nd; 1774 1775 restart: 1776 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1777 SCARG(uap, path), p); 1778 if ((error = namei(&nd)) != 0) 1779 return (error); 1780 vp = nd.ni_vp; 1781 1782 /* 1783 * The root of a mounted filesystem cannot be deleted. 1784 */ 1785 if (vp->v_flag & VROOT) { 1786 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1787 if (nd.ni_dvp == vp) 1788 vrele(nd.ni_dvp); 1789 else 1790 vput(nd.ni_dvp); 1791 vput(vp); 1792 error = EBUSY; 1793 goto out; 1794 } 1795 1796 #ifdef VERIFIED_EXEC 1797 /* Handle remove requests for veriexec entries. */ 1798 if ((error = veriexec_removechk(p, vp, nd.ni_dirp)) != 0) { 1799 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1800 if (nd.ni_dvp == vp) 1801 vrele(nd.ni_dvp); 1802 else 1803 vput(nd.ni_dvp); 1804 vput(vp); 1805 goto out; 1806 } 1807 #endif /* VERIFIED_EXEC */ 1808 1809 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1810 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1811 if (nd.ni_dvp == vp) 1812 vrele(nd.ni_dvp); 1813 else 1814 vput(nd.ni_dvp); 1815 vput(vp); 1816 if ((error = vn_start_write(NULL, &mp, 1817 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1818 return (error); 1819 goto restart; 1820 } 1821 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1822 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1823 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1824 vn_finished_write(mp, 0); 1825 out: 1826 return (error); 1827 } 1828 1829 /* 1830 * Reposition read/write file offset. 1831 */ 1832 int 1833 sys_lseek(struct lwp *l, void *v, register_t *retval) 1834 { 1835 struct sys_lseek_args /* { 1836 syscallarg(int) fd; 1837 syscallarg(int) pad; 1838 syscallarg(off_t) offset; 1839 syscallarg(int) whence; 1840 } */ *uap = v; 1841 struct proc *p = l->l_proc; 1842 struct ucred *cred = p->p_ucred; 1843 struct filedesc *fdp = p->p_fd; 1844 struct file *fp; 1845 struct vnode *vp; 1846 struct vattr vattr; 1847 off_t newoff; 1848 int error; 1849 1850 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1851 return (EBADF); 1852 1853 FILE_USE(fp); 1854 1855 vp = (struct vnode *)fp->f_data; 1856 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1857 error = ESPIPE; 1858 goto out; 1859 } 1860 1861 switch (SCARG(uap, whence)) { 1862 case SEEK_CUR: 1863 newoff = fp->f_offset + SCARG(uap, offset); 1864 break; 1865 case SEEK_END: 1866 error = VOP_GETATTR(vp, &vattr, cred, p); 1867 if (error) 1868 goto out; 1869 newoff = SCARG(uap, offset) + vattr.va_size; 1870 break; 1871 case SEEK_SET: 1872 newoff = SCARG(uap, offset); 1873 break; 1874 default: 1875 error = EINVAL; 1876 goto out; 1877 } 1878 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1879 goto out; 1880 1881 *(off_t *)retval = fp->f_offset = newoff; 1882 out: 1883 FILE_UNUSE(fp, p); 1884 return (error); 1885 } 1886 1887 /* 1888 * Positional read system call. 1889 */ 1890 int 1891 sys_pread(struct lwp *l, void *v, register_t *retval) 1892 { 1893 struct sys_pread_args /* { 1894 syscallarg(int) fd; 1895 syscallarg(void *) buf; 1896 syscallarg(size_t) nbyte; 1897 syscallarg(off_t) offset; 1898 } */ *uap = v; 1899 struct proc *p = l->l_proc; 1900 struct filedesc *fdp = p->p_fd; 1901 struct file *fp; 1902 struct vnode *vp; 1903 off_t offset; 1904 int error, fd = SCARG(uap, fd); 1905 1906 if ((fp = fd_getfile(fdp, fd)) == NULL) 1907 return (EBADF); 1908 1909 if ((fp->f_flag & FREAD) == 0) { 1910 simple_unlock(&fp->f_slock); 1911 return (EBADF); 1912 } 1913 1914 FILE_USE(fp); 1915 1916 vp = (struct vnode *)fp->f_data; 1917 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1918 error = ESPIPE; 1919 goto out; 1920 } 1921 1922 offset = SCARG(uap, offset); 1923 1924 /* 1925 * XXX This works because no file systems actually 1926 * XXX take any action on the seek operation. 1927 */ 1928 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1929 goto out; 1930 1931 /* dofileread() will unuse the descriptor for us */ 1932 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1933 &offset, 0, retval)); 1934 1935 out: 1936 FILE_UNUSE(fp, p); 1937 return (error); 1938 } 1939 1940 /* 1941 * Positional scatter read system call. 1942 */ 1943 int 1944 sys_preadv(struct lwp *l, void *v, register_t *retval) 1945 { 1946 struct sys_preadv_args /* { 1947 syscallarg(int) fd; 1948 syscallarg(const struct iovec *) iovp; 1949 syscallarg(int) iovcnt; 1950 syscallarg(off_t) offset; 1951 } */ *uap = v; 1952 struct proc *p = l->l_proc; 1953 struct filedesc *fdp = p->p_fd; 1954 struct file *fp; 1955 struct vnode *vp; 1956 off_t offset; 1957 int error, fd = SCARG(uap, fd); 1958 1959 if ((fp = fd_getfile(fdp, fd)) == NULL) 1960 return (EBADF); 1961 1962 if ((fp->f_flag & FREAD) == 0) { 1963 simple_unlock(&fp->f_slock); 1964 return (EBADF); 1965 } 1966 1967 FILE_USE(fp); 1968 1969 vp = (struct vnode *)fp->f_data; 1970 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1971 error = ESPIPE; 1972 goto out; 1973 } 1974 1975 offset = SCARG(uap, offset); 1976 1977 /* 1978 * XXX This works because no file systems actually 1979 * XXX take any action on the seek operation. 1980 */ 1981 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1982 goto out; 1983 1984 /* dofilereadv() will unuse the descriptor for us */ 1985 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1986 &offset, 0, retval)); 1987 1988 out: 1989 FILE_UNUSE(fp, p); 1990 return (error); 1991 } 1992 1993 /* 1994 * Positional write system call. 1995 */ 1996 int 1997 sys_pwrite(struct lwp *l, void *v, register_t *retval) 1998 { 1999 struct sys_pwrite_args /* { 2000 syscallarg(int) fd; 2001 syscallarg(const void *) buf; 2002 syscallarg(size_t) nbyte; 2003 syscallarg(off_t) offset; 2004 } */ *uap = v; 2005 struct proc *p = l->l_proc; 2006 struct filedesc *fdp = p->p_fd; 2007 struct file *fp; 2008 struct vnode *vp; 2009 off_t offset; 2010 int error, fd = SCARG(uap, fd); 2011 2012 if ((fp = fd_getfile(fdp, fd)) == NULL) 2013 return (EBADF); 2014 2015 if ((fp->f_flag & FWRITE) == 0) { 2016 simple_unlock(&fp->f_slock); 2017 return (EBADF); 2018 } 2019 2020 FILE_USE(fp); 2021 2022 vp = (struct vnode *)fp->f_data; 2023 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2024 error = ESPIPE; 2025 goto out; 2026 } 2027 2028 offset = SCARG(uap, offset); 2029 2030 /* 2031 * XXX This works because no file systems actually 2032 * XXX take any action on the seek operation. 2033 */ 2034 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2035 goto out; 2036 2037 /* dofilewrite() will unuse the descriptor for us */ 2038 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2039 &offset, 0, retval)); 2040 2041 out: 2042 FILE_UNUSE(fp, p); 2043 return (error); 2044 } 2045 2046 /* 2047 * Positional gather write system call. 2048 */ 2049 int 2050 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2051 { 2052 struct sys_pwritev_args /* { 2053 syscallarg(int) fd; 2054 syscallarg(const struct iovec *) iovp; 2055 syscallarg(int) iovcnt; 2056 syscallarg(off_t) offset; 2057 } */ *uap = v; 2058 struct proc *p = l->l_proc; 2059 struct filedesc *fdp = p->p_fd; 2060 struct file *fp; 2061 struct vnode *vp; 2062 off_t offset; 2063 int error, fd = SCARG(uap, fd); 2064 2065 if ((fp = fd_getfile(fdp, fd)) == NULL) 2066 return (EBADF); 2067 2068 if ((fp->f_flag & FWRITE) == 0) { 2069 simple_unlock(&fp->f_slock); 2070 return (EBADF); 2071 } 2072 2073 FILE_USE(fp); 2074 2075 vp = (struct vnode *)fp->f_data; 2076 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2077 error = ESPIPE; 2078 goto out; 2079 } 2080 2081 offset = SCARG(uap, offset); 2082 2083 /* 2084 * XXX This works because no file systems actually 2085 * XXX take any action on the seek operation. 2086 */ 2087 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2088 goto out; 2089 2090 /* dofilewritev() will unuse the descriptor for us */ 2091 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2092 &offset, 0, retval)); 2093 2094 out: 2095 FILE_UNUSE(fp, p); 2096 return (error); 2097 } 2098 2099 /* 2100 * Check access permissions. 2101 */ 2102 int 2103 sys_access(struct lwp *l, void *v, register_t *retval) 2104 { 2105 struct sys_access_args /* { 2106 syscallarg(const char *) path; 2107 syscallarg(int) flags; 2108 } */ *uap = v; 2109 struct proc *p = l->l_proc; 2110 struct ucred *cred; 2111 struct vnode *vp; 2112 int error, flags; 2113 struct nameidata nd; 2114 2115 cred = crdup(p->p_ucred); 2116 cred->cr_uid = p->p_cred->p_ruid; 2117 cred->cr_gid = p->p_cred->p_rgid; 2118 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2119 SCARG(uap, path), p); 2120 /* Override default credentials */ 2121 nd.ni_cnd.cn_cred = cred; 2122 if ((error = namei(&nd)) != 0) 2123 goto out; 2124 vp = nd.ni_vp; 2125 2126 /* Flags == 0 means only check for existence. */ 2127 if (SCARG(uap, flags)) { 2128 flags = 0; 2129 if (SCARG(uap, flags) & R_OK) 2130 flags |= VREAD; 2131 if (SCARG(uap, flags) & W_OK) 2132 flags |= VWRITE; 2133 if (SCARG(uap, flags) & X_OK) 2134 flags |= VEXEC; 2135 2136 error = VOP_ACCESS(vp, flags, cred, p); 2137 if (!error && (flags & VWRITE)) 2138 error = vn_writechk(vp); 2139 } 2140 vput(vp); 2141 out: 2142 crfree(cred); 2143 return (error); 2144 } 2145 2146 /* 2147 * Get file status; this version follows links. 2148 */ 2149 /* ARGSUSED */ 2150 int 2151 sys___stat30(struct lwp *l, void *v, register_t *retval) 2152 { 2153 struct sys___stat30_args /* { 2154 syscallarg(const char *) path; 2155 syscallarg(struct stat *) ub; 2156 } */ *uap = v; 2157 struct proc *p = l->l_proc; 2158 struct stat sb; 2159 int error; 2160 struct nameidata nd; 2161 2162 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2163 SCARG(uap, path), p); 2164 if ((error = namei(&nd)) != 0) 2165 return (error); 2166 error = vn_stat(nd.ni_vp, &sb, p); 2167 vput(nd.ni_vp); 2168 if (error) 2169 return (error); 2170 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2171 return (error); 2172 } 2173 2174 /* 2175 * Get file status; this version does not follow links. 2176 */ 2177 /* ARGSUSED */ 2178 int 2179 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2180 { 2181 struct sys___lstat30_args /* { 2182 syscallarg(const char *) path; 2183 syscallarg(struct stat *) ub; 2184 } */ *uap = v; 2185 struct proc *p = l->l_proc; 2186 struct stat sb; 2187 int error; 2188 struct nameidata nd; 2189 2190 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2191 SCARG(uap, path), p); 2192 if ((error = namei(&nd)) != 0) 2193 return (error); 2194 error = vn_stat(nd.ni_vp, &sb, p); 2195 vput(nd.ni_vp); 2196 if (error) 2197 return (error); 2198 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2199 return (error); 2200 } 2201 2202 /* 2203 * Get configurable pathname variables. 2204 */ 2205 /* ARGSUSED */ 2206 int 2207 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2208 { 2209 struct sys_pathconf_args /* { 2210 syscallarg(const char *) path; 2211 syscallarg(int) name; 2212 } */ *uap = v; 2213 struct proc *p = l->l_proc; 2214 int error; 2215 struct nameidata nd; 2216 2217 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2218 SCARG(uap, path), p); 2219 if ((error = namei(&nd)) != 0) 2220 return (error); 2221 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2222 vput(nd.ni_vp); 2223 return (error); 2224 } 2225 2226 /* 2227 * Return target name of a symbolic link. 2228 */ 2229 /* ARGSUSED */ 2230 int 2231 sys_readlink(struct lwp *l, void *v, register_t *retval) 2232 { 2233 struct sys_readlink_args /* { 2234 syscallarg(const char *) path; 2235 syscallarg(char *) buf; 2236 syscallarg(size_t) count; 2237 } */ *uap = v; 2238 struct proc *p = l->l_proc; 2239 struct vnode *vp; 2240 struct iovec aiov; 2241 struct uio auio; 2242 int error; 2243 struct nameidata nd; 2244 2245 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2246 SCARG(uap, path), p); 2247 if ((error = namei(&nd)) != 0) 2248 return (error); 2249 vp = nd.ni_vp; 2250 if (vp->v_type != VLNK) 2251 error = EINVAL; 2252 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2253 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 2254 aiov.iov_base = SCARG(uap, buf); 2255 aiov.iov_len = SCARG(uap, count); 2256 auio.uio_iov = &aiov; 2257 auio.uio_iovcnt = 1; 2258 auio.uio_offset = 0; 2259 auio.uio_rw = UIO_READ; 2260 auio.uio_segflg = UIO_USERSPACE; 2261 auio.uio_procp = p; 2262 auio.uio_resid = SCARG(uap, count); 2263 error = VOP_READLINK(vp, &auio, p->p_ucred); 2264 } 2265 vput(vp); 2266 *retval = SCARG(uap, count) - auio.uio_resid; 2267 return (error); 2268 } 2269 2270 /* 2271 * Change flags of a file given a path name. 2272 */ 2273 /* ARGSUSED */ 2274 int 2275 sys_chflags(struct lwp *l, void *v, register_t *retval) 2276 { 2277 struct sys_chflags_args /* { 2278 syscallarg(const char *) path; 2279 syscallarg(u_long) flags; 2280 } */ *uap = v; 2281 struct proc *p = l->l_proc; 2282 struct vnode *vp; 2283 int error; 2284 struct nameidata nd; 2285 2286 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2287 if ((error = namei(&nd)) != 0) 2288 return (error); 2289 vp = nd.ni_vp; 2290 error = change_flags(vp, SCARG(uap, flags), p); 2291 vput(vp); 2292 return (error); 2293 } 2294 2295 /* 2296 * Change flags of a file given a file descriptor. 2297 */ 2298 /* ARGSUSED */ 2299 int 2300 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2301 { 2302 struct sys_fchflags_args /* { 2303 syscallarg(int) fd; 2304 syscallarg(u_long) flags; 2305 } */ *uap = v; 2306 struct proc *p = l->l_proc; 2307 struct vnode *vp; 2308 struct file *fp; 2309 int error; 2310 2311 /* getvnode() will use the descriptor for us */ 2312 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2313 return (error); 2314 vp = (struct vnode *)fp->f_data; 2315 error = change_flags(vp, SCARG(uap, flags), p); 2316 VOP_UNLOCK(vp, 0); 2317 FILE_UNUSE(fp, p); 2318 return (error); 2319 } 2320 2321 /* 2322 * Change flags of a file given a path name; this version does 2323 * not follow links. 2324 */ 2325 int 2326 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2327 { 2328 struct sys_lchflags_args /* { 2329 syscallarg(const char *) path; 2330 syscallarg(u_long) flags; 2331 } */ *uap = v; 2332 struct proc *p = l->l_proc; 2333 struct vnode *vp; 2334 int error; 2335 struct nameidata nd; 2336 2337 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2338 if ((error = namei(&nd)) != 0) 2339 return (error); 2340 vp = nd.ni_vp; 2341 error = change_flags(vp, SCARG(uap, flags), p); 2342 vput(vp); 2343 return (error); 2344 } 2345 2346 /* 2347 * Common routine to change flags of a file. 2348 */ 2349 int 2350 change_flags(struct vnode *vp, u_long flags, struct proc *p) 2351 { 2352 struct mount *mp; 2353 struct vattr vattr; 2354 int error; 2355 2356 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2357 return (error); 2358 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2359 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2360 /* 2361 * Non-superusers cannot change the flags on devices, even if they 2362 * own them. 2363 */ 2364 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2365 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2366 goto out; 2367 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2368 error = EINVAL; 2369 goto out; 2370 } 2371 } 2372 VATTR_NULL(&vattr); 2373 vattr.va_flags = flags; 2374 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2375 out: 2376 vn_finished_write(mp, 0); 2377 return (error); 2378 } 2379 2380 /* 2381 * Change mode of a file given path name; this version follows links. 2382 */ 2383 /* ARGSUSED */ 2384 int 2385 sys_chmod(struct lwp *l, void *v, register_t *retval) 2386 { 2387 struct sys_chmod_args /* { 2388 syscallarg(const char *) path; 2389 syscallarg(int) mode; 2390 } */ *uap = v; 2391 struct proc *p = l->l_proc; 2392 int error; 2393 struct nameidata nd; 2394 2395 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2396 if ((error = namei(&nd)) != 0) 2397 return (error); 2398 2399 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2400 2401 vrele(nd.ni_vp); 2402 return (error); 2403 } 2404 2405 /* 2406 * Change mode of a file given a file descriptor. 2407 */ 2408 /* ARGSUSED */ 2409 int 2410 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2411 { 2412 struct sys_fchmod_args /* { 2413 syscallarg(int) fd; 2414 syscallarg(int) mode; 2415 } */ *uap = v; 2416 struct proc *p = l->l_proc; 2417 struct file *fp; 2418 int error; 2419 2420 /* getvnode() will use the descriptor for us */ 2421 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2422 return (error); 2423 2424 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 2425 FILE_UNUSE(fp, p); 2426 return (error); 2427 } 2428 2429 /* 2430 * Change mode of a file given path name; this version does not follow links. 2431 */ 2432 /* ARGSUSED */ 2433 int 2434 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2435 { 2436 struct sys_lchmod_args /* { 2437 syscallarg(const char *) path; 2438 syscallarg(int) mode; 2439 } */ *uap = v; 2440 struct proc *p = l->l_proc; 2441 int error; 2442 struct nameidata nd; 2443 2444 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2445 if ((error = namei(&nd)) != 0) 2446 return (error); 2447 2448 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2449 2450 vrele(nd.ni_vp); 2451 return (error); 2452 } 2453 2454 /* 2455 * Common routine to set mode given a vnode. 2456 */ 2457 static int 2458 change_mode(struct vnode *vp, int mode, struct proc *p) 2459 { 2460 struct mount *mp; 2461 struct vattr vattr; 2462 int error; 2463 2464 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2465 return (error); 2466 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2467 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2468 VATTR_NULL(&vattr); 2469 vattr.va_mode = mode & ALLPERMS; 2470 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2471 VOP_UNLOCK(vp, 0); 2472 vn_finished_write(mp, 0); 2473 return (error); 2474 } 2475 2476 /* 2477 * Set ownership given a path name; this version follows links. 2478 */ 2479 /* ARGSUSED */ 2480 int 2481 sys_chown(struct lwp *l, void *v, register_t *retval) 2482 { 2483 struct sys_chown_args /* { 2484 syscallarg(const char *) path; 2485 syscallarg(uid_t) uid; 2486 syscallarg(gid_t) gid; 2487 } */ *uap = v; 2488 struct proc *p = l->l_proc; 2489 int error; 2490 struct nameidata nd; 2491 2492 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2493 if ((error = namei(&nd)) != 0) 2494 return (error); 2495 2496 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2497 2498 vrele(nd.ni_vp); 2499 return (error); 2500 } 2501 2502 /* 2503 * Set ownership given a path name; this version follows links. 2504 * Provides POSIX semantics. 2505 */ 2506 /* ARGSUSED */ 2507 int 2508 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2509 { 2510 struct sys_chown_args /* { 2511 syscallarg(const char *) path; 2512 syscallarg(uid_t) uid; 2513 syscallarg(gid_t) gid; 2514 } */ *uap = v; 2515 struct proc *p = l->l_proc; 2516 int error; 2517 struct nameidata nd; 2518 2519 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2520 if ((error = namei(&nd)) != 0) 2521 return (error); 2522 2523 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2524 2525 vrele(nd.ni_vp); 2526 return (error); 2527 } 2528 2529 /* 2530 * Set ownership given a file descriptor. 2531 */ 2532 /* ARGSUSED */ 2533 int 2534 sys_fchown(struct lwp *l, void *v, register_t *retval) 2535 { 2536 struct sys_fchown_args /* { 2537 syscallarg(int) fd; 2538 syscallarg(uid_t) uid; 2539 syscallarg(gid_t) gid; 2540 } */ *uap = v; 2541 struct proc *p = l->l_proc; 2542 int error; 2543 struct file *fp; 2544 2545 /* getvnode() will use the descriptor for us */ 2546 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2547 return (error); 2548 2549 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2550 SCARG(uap, gid), p, 0); 2551 FILE_UNUSE(fp, p); 2552 return (error); 2553 } 2554 2555 /* 2556 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2557 */ 2558 /* ARGSUSED */ 2559 int 2560 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2561 { 2562 struct sys_fchown_args /* { 2563 syscallarg(int) fd; 2564 syscallarg(uid_t) uid; 2565 syscallarg(gid_t) gid; 2566 } */ *uap = v; 2567 struct proc *p = l->l_proc; 2568 int error; 2569 struct file *fp; 2570 2571 /* getvnode() will use the descriptor for us */ 2572 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2573 return (error); 2574 2575 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2576 SCARG(uap, gid), p, 1); 2577 FILE_UNUSE(fp, p); 2578 return (error); 2579 } 2580 2581 /* 2582 * Set ownership given a path name; this version does not follow links. 2583 */ 2584 /* ARGSUSED */ 2585 int 2586 sys_lchown(struct lwp *l, void *v, register_t *retval) 2587 { 2588 struct sys_lchown_args /* { 2589 syscallarg(const char *) path; 2590 syscallarg(uid_t) uid; 2591 syscallarg(gid_t) gid; 2592 } */ *uap = v; 2593 struct proc *p = l->l_proc; 2594 int error; 2595 struct nameidata nd; 2596 2597 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2598 if ((error = namei(&nd)) != 0) 2599 return (error); 2600 2601 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2602 2603 vrele(nd.ni_vp); 2604 return (error); 2605 } 2606 2607 /* 2608 * Set ownership given a path name; this version does not follow links. 2609 * Provides POSIX/XPG semantics. 2610 */ 2611 /* ARGSUSED */ 2612 int 2613 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2614 { 2615 struct sys_lchown_args /* { 2616 syscallarg(const char *) path; 2617 syscallarg(uid_t) uid; 2618 syscallarg(gid_t) gid; 2619 } */ *uap = v; 2620 struct proc *p = l->l_proc; 2621 int error; 2622 struct nameidata nd; 2623 2624 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2625 if ((error = namei(&nd)) != 0) 2626 return (error); 2627 2628 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2629 2630 vrele(nd.ni_vp); 2631 return (error); 2632 } 2633 2634 /* 2635 * Common routine to set ownership given a vnode. 2636 */ 2637 static int 2638 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct proc *p, 2639 int posix_semantics) 2640 { 2641 struct mount *mp; 2642 struct vattr vattr; 2643 mode_t newmode; 2644 int error; 2645 2646 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2647 return (error); 2648 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2649 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2650 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2651 goto out; 2652 2653 #define CHANGED(x) ((int)(x) != -1) 2654 newmode = vattr.va_mode; 2655 if (posix_semantics) { 2656 /* 2657 * POSIX/XPG semantics: if the caller is not the super-user, 2658 * clear set-user-id and set-group-id bits. Both POSIX and 2659 * the XPG consider the behaviour for calls by the super-user 2660 * implementation-defined; we leave the set-user-id and set- 2661 * group-id settings intact in that case. 2662 */ 2663 if (suser(p->p_ucred, NULL) != 0) 2664 newmode &= ~(S_ISUID | S_ISGID); 2665 } else { 2666 /* 2667 * NetBSD semantics: when changing owner and/or group, 2668 * clear the respective bit(s). 2669 */ 2670 if (CHANGED(uid)) 2671 newmode &= ~S_ISUID; 2672 if (CHANGED(gid)) 2673 newmode &= ~S_ISGID; 2674 } 2675 /* Update va_mode iff altered. */ 2676 if (vattr.va_mode == newmode) 2677 newmode = VNOVAL; 2678 2679 VATTR_NULL(&vattr); 2680 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2681 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2682 vattr.va_mode = newmode; 2683 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2684 #undef CHANGED 2685 2686 out: 2687 VOP_UNLOCK(vp, 0); 2688 vn_finished_write(mp, 0); 2689 return (error); 2690 } 2691 2692 /* 2693 * Set the access and modification times given a path name; this 2694 * version follows links. 2695 */ 2696 /* ARGSUSED */ 2697 int 2698 sys_utimes(struct lwp *l, void *v, register_t *retval) 2699 { 2700 struct sys_utimes_args /* { 2701 syscallarg(const char *) path; 2702 syscallarg(const struct timeval *) tptr; 2703 } */ *uap = v; 2704 struct proc *p = l->l_proc; 2705 int error; 2706 struct nameidata nd; 2707 2708 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2709 if ((error = namei(&nd)) != 0) 2710 return (error); 2711 2712 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2713 2714 vrele(nd.ni_vp); 2715 return (error); 2716 } 2717 2718 /* 2719 * Set the access and modification times given a file descriptor. 2720 */ 2721 /* ARGSUSED */ 2722 int 2723 sys_futimes(struct lwp *l, void *v, register_t *retval) 2724 { 2725 struct sys_futimes_args /* { 2726 syscallarg(int) fd; 2727 syscallarg(const struct timeval *) tptr; 2728 } */ *uap = v; 2729 struct proc *p = l->l_proc; 2730 int error; 2731 struct file *fp; 2732 2733 /* getvnode() will use the descriptor for us */ 2734 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2735 return (error); 2736 2737 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2738 FILE_UNUSE(fp, p); 2739 return (error); 2740 } 2741 2742 /* 2743 * Set the access and modification times given a path name; this 2744 * version does not follow links. 2745 */ 2746 /* ARGSUSED */ 2747 int 2748 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2749 { 2750 struct sys_lutimes_args /* { 2751 syscallarg(const char *) path; 2752 syscallarg(const struct timeval *) tptr; 2753 } */ *uap = v; 2754 struct proc *p = l->l_proc; 2755 int error; 2756 struct nameidata nd; 2757 2758 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2759 if ((error = namei(&nd)) != 0) 2760 return (error); 2761 2762 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2763 2764 vrele(nd.ni_vp); 2765 return (error); 2766 } 2767 2768 /* 2769 * Common routine to set access and modification times given a vnode. 2770 */ 2771 static int 2772 change_utimes(struct vnode *vp, const struct timeval *tptr, struct proc *p) 2773 { 2774 struct mount *mp; 2775 struct vattr vattr; 2776 int error; 2777 2778 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2779 return (error); 2780 VATTR_NULL(&vattr); 2781 if (tptr == NULL) { 2782 nanotime(&vattr.va_atime); 2783 vattr.va_mtime = vattr.va_atime; 2784 vattr.va_vaflags |= VA_UTIMES_NULL; 2785 } else { 2786 struct timeval tv[2]; 2787 2788 error = copyin(tptr, tv, sizeof(tv)); 2789 if (error) 2790 goto out; 2791 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 2792 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 2793 } 2794 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2795 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2796 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2797 VOP_UNLOCK(vp, 0); 2798 out: 2799 vn_finished_write(mp, 0); 2800 return (error); 2801 } 2802 2803 /* 2804 * Truncate a file given its path name. 2805 */ 2806 /* ARGSUSED */ 2807 int 2808 sys_truncate(struct lwp *l, void *v, register_t *retval) 2809 { 2810 struct sys_truncate_args /* { 2811 syscallarg(const char *) path; 2812 syscallarg(int) pad; 2813 syscallarg(off_t) length; 2814 } */ *uap = v; 2815 struct proc *p = l->l_proc; 2816 struct vnode *vp; 2817 struct mount *mp; 2818 struct vattr vattr; 2819 int error; 2820 struct nameidata nd; 2821 2822 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2823 if ((error = namei(&nd)) != 0) 2824 return (error); 2825 vp = nd.ni_vp; 2826 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2827 vrele(vp); 2828 return (error); 2829 } 2830 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2831 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2832 if (vp->v_type == VDIR) 2833 error = EISDIR; 2834 else if ((error = vn_writechk(vp)) == 0 && 2835 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2836 VATTR_NULL(&vattr); 2837 vattr.va_size = SCARG(uap, length); 2838 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2839 } 2840 vput(vp); 2841 vn_finished_write(mp, 0); 2842 return (error); 2843 } 2844 2845 /* 2846 * Truncate a file given a file descriptor. 2847 */ 2848 /* ARGSUSED */ 2849 int 2850 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2851 { 2852 struct sys_ftruncate_args /* { 2853 syscallarg(int) fd; 2854 syscallarg(int) pad; 2855 syscallarg(off_t) length; 2856 } */ *uap = v; 2857 struct proc *p = l->l_proc; 2858 struct mount *mp; 2859 struct vattr vattr; 2860 struct vnode *vp; 2861 struct file *fp; 2862 int error; 2863 2864 /* getvnode() will use the descriptor for us */ 2865 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2866 return (error); 2867 if ((fp->f_flag & FWRITE) == 0) { 2868 error = EINVAL; 2869 goto out; 2870 } 2871 vp = (struct vnode *)fp->f_data; 2872 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2873 FILE_UNUSE(fp, p); 2874 return (error); 2875 } 2876 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2877 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2878 if (vp->v_type == VDIR) 2879 error = EISDIR; 2880 else if ((error = vn_writechk(vp)) == 0) { 2881 VATTR_NULL(&vattr); 2882 vattr.va_size = SCARG(uap, length); 2883 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2884 } 2885 VOP_UNLOCK(vp, 0); 2886 vn_finished_write(mp, 0); 2887 out: 2888 FILE_UNUSE(fp, p); 2889 return (error); 2890 } 2891 2892 /* 2893 * Sync an open file. 2894 */ 2895 /* ARGSUSED */ 2896 int 2897 sys_fsync(struct lwp *l, void *v, register_t *retval) 2898 { 2899 struct sys_fsync_args /* { 2900 syscallarg(int) fd; 2901 } */ *uap = v; 2902 struct proc *p = l->l_proc; 2903 struct vnode *vp; 2904 struct mount *mp; 2905 struct file *fp; 2906 int error; 2907 2908 /* getvnode() will use the descriptor for us */ 2909 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2910 return (error); 2911 vp = (struct vnode *)fp->f_data; 2912 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2913 FILE_UNUSE(fp, p); 2914 return (error); 2915 } 2916 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2917 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p); 2918 if (error == 0 && bioops.io_fsync != NULL && 2919 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2920 (*bioops.io_fsync)(vp, 0); 2921 VOP_UNLOCK(vp, 0); 2922 vn_finished_write(mp, 0); 2923 FILE_UNUSE(fp, p); 2924 return (error); 2925 } 2926 2927 /* 2928 * Sync a range of file data. API modeled after that found in AIX. 2929 * 2930 * FDATASYNC indicates that we need only save enough metadata to be able 2931 * to re-read the written data. Note we duplicate AIX's requirement that 2932 * the file be open for writing. 2933 */ 2934 /* ARGSUSED */ 2935 int 2936 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2937 { 2938 struct sys_fsync_range_args /* { 2939 syscallarg(int) fd; 2940 syscallarg(int) flags; 2941 syscallarg(off_t) start; 2942 syscallarg(off_t) length; 2943 } */ *uap = v; 2944 struct proc *p = l->l_proc; 2945 struct vnode *vp; 2946 struct file *fp; 2947 int flags, nflags; 2948 off_t s, e, len; 2949 int error; 2950 2951 /* getvnode() will use the descriptor for us */ 2952 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2953 return (error); 2954 2955 if ((fp->f_flag & FWRITE) == 0) { 2956 FILE_UNUSE(fp, p); 2957 return (EBADF); 2958 } 2959 2960 flags = SCARG(uap, flags); 2961 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2962 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2963 return (EINVAL); 2964 } 2965 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2966 if (flags & FDATASYNC) 2967 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2968 else 2969 nflags = FSYNC_WAIT; 2970 if (flags & FDISKSYNC) 2971 nflags |= FSYNC_CACHE; 2972 2973 len = SCARG(uap, length); 2974 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2975 if (len) { 2976 s = SCARG(uap, start); 2977 e = s + len; 2978 if (e < s) { 2979 FILE_UNUSE(fp, p); 2980 return (EINVAL); 2981 } 2982 } else { 2983 e = 0; 2984 s = 0; 2985 } 2986 2987 vp = (struct vnode *)fp->f_data; 2988 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2989 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p); 2990 2991 if (error == 0 && bioops.io_fsync != NULL && 2992 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2993 (*bioops.io_fsync)(vp, nflags); 2994 2995 VOP_UNLOCK(vp, 0); 2996 FILE_UNUSE(fp, p); 2997 return (error); 2998 } 2999 3000 /* 3001 * Sync the data of an open file. 3002 */ 3003 /* ARGSUSED */ 3004 int 3005 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3006 { 3007 struct sys_fdatasync_args /* { 3008 syscallarg(int) fd; 3009 } */ *uap = v; 3010 struct proc *p = l->l_proc; 3011 struct vnode *vp; 3012 struct file *fp; 3013 int error; 3014 3015 /* getvnode() will use the descriptor for us */ 3016 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3017 return (error); 3018 if ((fp->f_flag & FWRITE) == 0) { 3019 FILE_UNUSE(fp, p); 3020 return (EBADF); 3021 } 3022 vp = (struct vnode *)fp->f_data; 3023 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3024 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p); 3025 VOP_UNLOCK(vp, 0); 3026 FILE_UNUSE(fp, p); 3027 return (error); 3028 } 3029 3030 /* 3031 * Rename files, (standard) BSD semantics frontend. 3032 */ 3033 /* ARGSUSED */ 3034 int 3035 sys_rename(struct lwp *l, void *v, register_t *retval) 3036 { 3037 struct sys_rename_args /* { 3038 syscallarg(const char *) from; 3039 syscallarg(const char *) to; 3040 } */ *uap = v; 3041 struct proc *p = l->l_proc; 3042 3043 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 3044 } 3045 3046 /* 3047 * Rename files, POSIX semantics frontend. 3048 */ 3049 /* ARGSUSED */ 3050 int 3051 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3052 { 3053 struct sys___posix_rename_args /* { 3054 syscallarg(const char *) from; 3055 syscallarg(const char *) to; 3056 } */ *uap = v; 3057 struct proc *p = l->l_proc; 3058 3059 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 3060 } 3061 3062 /* 3063 * Rename files. Source and destination must either both be directories, 3064 * or both not be directories. If target is a directory, it must be empty. 3065 * If `from' and `to' refer to the same object, the value of the `retain' 3066 * argument is used to determine whether `from' will be 3067 * 3068 * (retain == 0) deleted unless `from' and `to' refer to the same 3069 * object in the file system's name space (BSD). 3070 * (retain == 1) always retained (POSIX). 3071 */ 3072 static int 3073 rename_files(const char *from, const char *to, struct proc *p, int retain) 3074 { 3075 struct mount *mp = NULL; 3076 struct vnode *tvp, *fvp, *tdvp; 3077 struct nameidata fromnd, tond; 3078 int error; 3079 3080 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3081 from, p); 3082 if ((error = namei(&fromnd)) != 0) 3083 return (error); 3084 fvp = fromnd.ni_vp; 3085 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3086 if (error != 0) { 3087 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3088 vrele(fromnd.ni_dvp); 3089 vrele(fvp); 3090 if (fromnd.ni_startdir) 3091 vrele(fromnd.ni_startdir); 3092 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3093 return (error); 3094 } 3095 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3096 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p); 3097 if ((error = namei(&tond)) != 0) { 3098 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3099 vrele(fromnd.ni_dvp); 3100 vrele(fvp); 3101 goto out1; 3102 } 3103 tdvp = tond.ni_dvp; 3104 tvp = tond.ni_vp; 3105 3106 if (tvp != NULL) { 3107 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3108 error = ENOTDIR; 3109 goto out; 3110 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3111 error = EISDIR; 3112 goto out; 3113 } 3114 } 3115 3116 if (fvp == tdvp) 3117 error = EINVAL; 3118 3119 /* 3120 * Source and destination refer to the same object. 3121 */ 3122 if (fvp == tvp) { 3123 if (retain) 3124 error = -1; 3125 else if (fromnd.ni_dvp == tdvp && 3126 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3127 !memcmp(fromnd.ni_cnd.cn_nameptr, 3128 tond.ni_cnd.cn_nameptr, 3129 fromnd.ni_cnd.cn_namelen)) 3130 error = -1; 3131 } 3132 3133 #ifdef VERIFIED_EXEC 3134 if (!error) 3135 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp); 3136 #endif /* VERIFIED_EXEC */ 3137 3138 out: 3139 if (!error) { 3140 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 3141 if (fromnd.ni_dvp != tdvp) 3142 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3143 if (tvp) { 3144 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 3145 } 3146 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3147 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3148 } else { 3149 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3150 if (tdvp == tvp) 3151 vrele(tdvp); 3152 else 3153 vput(tdvp); 3154 if (tvp) 3155 vput(tvp); 3156 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3157 vrele(fromnd.ni_dvp); 3158 vrele(fvp); 3159 } 3160 vrele(tond.ni_startdir); 3161 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3162 out1: 3163 vn_finished_write(mp, 0); 3164 if (fromnd.ni_startdir) 3165 vrele(fromnd.ni_startdir); 3166 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3167 return (error == -1 ? 0 : error); 3168 } 3169 3170 /* 3171 * Make a directory file. 3172 */ 3173 /* ARGSUSED */ 3174 int 3175 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3176 { 3177 struct sys_mkdir_args /* { 3178 syscallarg(const char *) path; 3179 syscallarg(int) mode; 3180 } */ *uap = v; 3181 struct proc *p = l->l_proc; 3182 struct mount *mp; 3183 struct vnode *vp; 3184 struct vattr vattr; 3185 int error; 3186 struct nameidata nd; 3187 3188 restart: 3189 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3190 SCARG(uap, path), p); 3191 if ((error = namei(&nd)) != 0) 3192 return (error); 3193 vp = nd.ni_vp; 3194 if (vp != NULL) { 3195 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3196 if (nd.ni_dvp == vp) 3197 vrele(nd.ni_dvp); 3198 else 3199 vput(nd.ni_dvp); 3200 vrele(vp); 3201 return (EEXIST); 3202 } 3203 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3204 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3205 if (nd.ni_dvp == vp) 3206 vrele(nd.ni_dvp); 3207 else 3208 vput(nd.ni_dvp); 3209 if ((error = vn_start_write(NULL, &mp, 3210 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3211 return (error); 3212 goto restart; 3213 } 3214 VATTR_NULL(&vattr); 3215 vattr.va_type = VDIR; 3216 vattr.va_mode = 3217 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3218 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3219 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3220 if (!error) 3221 vput(nd.ni_vp); 3222 vn_finished_write(mp, 0); 3223 return (error); 3224 } 3225 3226 /* 3227 * Remove a directory file. 3228 */ 3229 /* ARGSUSED */ 3230 int 3231 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3232 { 3233 struct sys_rmdir_args /* { 3234 syscallarg(const char *) path; 3235 } */ *uap = v; 3236 struct proc *p = l->l_proc; 3237 struct mount *mp; 3238 struct vnode *vp; 3239 int error; 3240 struct nameidata nd; 3241 3242 restart: 3243 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3244 SCARG(uap, path), p); 3245 if ((error = namei(&nd)) != 0) 3246 return (error); 3247 vp = nd.ni_vp; 3248 if (vp->v_type != VDIR) { 3249 error = ENOTDIR; 3250 goto out; 3251 } 3252 /* 3253 * No rmdir "." please. 3254 */ 3255 if (nd.ni_dvp == vp) { 3256 error = EINVAL; 3257 goto out; 3258 } 3259 /* 3260 * The root of a mounted filesystem cannot be deleted. 3261 */ 3262 if (vp->v_flag & VROOT) { 3263 error = EBUSY; 3264 goto out; 3265 } 3266 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3267 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3268 if (nd.ni_dvp == vp) 3269 vrele(nd.ni_dvp); 3270 else 3271 vput(nd.ni_dvp); 3272 vput(vp); 3273 if ((error = vn_start_write(NULL, &mp, 3274 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3275 return (error); 3276 goto restart; 3277 } 3278 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3279 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3280 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3281 vn_finished_write(mp, 0); 3282 return (error); 3283 3284 out: 3285 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3286 if (nd.ni_dvp == vp) 3287 vrele(nd.ni_dvp); 3288 else 3289 vput(nd.ni_dvp); 3290 vput(vp); 3291 return (error); 3292 } 3293 3294 /* 3295 * Read a block of directory entries in a file system independent format. 3296 */ 3297 int 3298 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3299 { 3300 struct sys___getdents30_args /* { 3301 syscallarg(int) fd; 3302 syscallarg(char *) buf; 3303 syscallarg(size_t) count; 3304 } */ *uap = v; 3305 struct proc *p = l->l_proc; 3306 struct file *fp; 3307 int error, done; 3308 3309 /* getvnode() will use the descriptor for us */ 3310 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3311 return (error); 3312 if ((fp->f_flag & FREAD) == 0) { 3313 error = EBADF; 3314 goto out; 3315 } 3316 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3317 SCARG(uap, count), &done, p, 0, 0); 3318 #ifdef KTRACE 3319 if (!error && KTRPOINT(p, KTR_GENIO)) { 3320 struct iovec iov; 3321 iov.iov_base = SCARG(uap, buf); 3322 iov.iov_len = done; 3323 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3324 } 3325 #endif 3326 *retval = done; 3327 out: 3328 FILE_UNUSE(fp, p); 3329 return (error); 3330 } 3331 3332 /* 3333 * Set the mode mask for creation of filesystem nodes. 3334 */ 3335 int 3336 sys_umask(struct lwp *l, void *v, register_t *retval) 3337 { 3338 struct sys_umask_args /* { 3339 syscallarg(mode_t) newmask; 3340 } */ *uap = v; 3341 struct proc *p = l->l_proc; 3342 struct cwdinfo *cwdi; 3343 3344 cwdi = p->p_cwdi; 3345 *retval = cwdi->cwdi_cmask; 3346 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3347 return (0); 3348 } 3349 3350 /* 3351 * Void all references to file by ripping underlying filesystem 3352 * away from vnode. 3353 */ 3354 /* ARGSUSED */ 3355 int 3356 sys_revoke(struct lwp *l, void *v, register_t *retval) 3357 { 3358 struct sys_revoke_args /* { 3359 syscallarg(const char *) path; 3360 } */ *uap = v; 3361 struct proc *p = l->l_proc; 3362 struct mount *mp; 3363 struct vnode *vp; 3364 struct vattr vattr; 3365 int error; 3366 struct nameidata nd; 3367 3368 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3369 if ((error = namei(&nd)) != 0) 3370 return (error); 3371 vp = nd.ni_vp; 3372 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 3373 goto out; 3374 if (p->p_ucred->cr_uid != vattr.va_uid && 3375 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3376 goto out; 3377 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3378 goto out; 3379 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3380 VOP_REVOKE(vp, REVOKEALL); 3381 vn_finished_write(mp, 0); 3382 out: 3383 vrele(vp); 3384 return (error); 3385 } 3386 3387 /* 3388 * Convert a user file descriptor to a kernel file entry. 3389 */ 3390 int 3391 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3392 { 3393 struct vnode *vp; 3394 struct file *fp; 3395 3396 if ((fp = fd_getfile(fdp, fd)) == NULL) 3397 return (EBADF); 3398 3399 FILE_USE(fp); 3400 3401 if (fp->f_type != DTYPE_VNODE) { 3402 FILE_UNUSE(fp, NULL); 3403 return (EINVAL); 3404 } 3405 3406 vp = (struct vnode *)fp->f_data; 3407 if (vp->v_type == VBAD) { 3408 FILE_UNUSE(fp, NULL); 3409 return (EBADF); 3410 } 3411 3412 *fpp = fp; 3413 return (0); 3414 } 3415
Indexes created Fri Sep 26 20:09:58 GMT 2025