vfs_syscalls.c revision 1.241
1 /* $NetBSD: vfs_syscalls.c,v 1.241 2006/05/10 11:02:29 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 1.241 2006/05/10 11:02:29 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 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 *sb; 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 sb = STATVFSBUF_GET(); 791 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 792 if (error == 0) { 793 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 794 } 795 STATVFSBUF_PUT(sb); 796 return error; 797 } 798 799 /* 800 * Get filesystem statistics. 801 */ 802 /* ARGSUSED */ 803 int 804 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 805 { 806 struct sys_fstatvfs1_args /* { 807 syscallarg(int) fd; 808 syscallarg(struct statvfs *) buf; 809 syscallarg(int) flags; 810 } */ *uap = v; 811 struct proc *p = l->l_proc; 812 struct file *fp; 813 struct mount *mp; 814 struct statvfs *sb; 815 int error; 816 817 /* getvnode() will use the descriptor for us */ 818 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 819 return (error); 820 mp = ((struct vnode *)fp->f_data)->v_mount; 821 sb = STATVFSBUF_GET(); 822 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 823 goto out; 824 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 825 out: 826 FILE_UNUSE(fp, l); 827 STATVFSBUF_PUT(sb); 828 return error; 829 } 830 831 832 /* 833 * Get statistics on all filesystems. 834 */ 835 int 836 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 837 { 838 struct sys_getvfsstat_args /* { 839 syscallarg(struct statvfs *) buf; 840 syscallarg(size_t) bufsize; 841 syscallarg(int) flags; 842 } */ *uap = v; 843 int root = 0; 844 struct proc *p = l->l_proc; 845 struct mount *mp, *nmp; 846 struct statvfs *sb; 847 struct statvfs *sfsp; 848 size_t count, maxcount; 849 int error = 0; 850 851 sb = STATVFSBUF_GET(); 852 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 853 sfsp = SCARG(uap, buf); 854 simple_lock(&mountlist_slock); 855 count = 0; 856 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 857 mp = nmp) { 858 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 859 nmp = CIRCLEQ_NEXT(mp, mnt_list); 860 continue; 861 } 862 if (sfsp && count < maxcount) { 863 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 864 if (error) { 865 simple_lock(&mountlist_slock); 866 nmp = CIRCLEQ_NEXT(mp, mnt_list); 867 vfs_unbusy(mp); 868 continue; 869 } 870 error = copyout(sb, sfsp, sizeof(*sfsp)); 871 if (error) { 872 vfs_unbusy(mp); 873 goto out; 874 } 875 sfsp++; 876 root |= strcmp(sb->f_mntonname, "/") == 0; 877 } 878 count++; 879 simple_lock(&mountlist_slock); 880 nmp = CIRCLEQ_NEXT(mp, mnt_list); 881 vfs_unbusy(mp); 882 } 883 simple_unlock(&mountlist_slock); 884 if (root == 0 && p->p_cwdi->cwdi_rdir) { 885 /* 886 * fake a root entry 887 */ 888 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 889 SCARG(uap, flags), 1)) != 0) 890 goto out; 891 if (sfsp) 892 error = copyout(sb, sfsp, sizeof(*sfsp)); 893 count++; 894 } 895 if (sfsp && count > maxcount) 896 *retval = maxcount; 897 else 898 *retval = count; 899 out: 900 STATVFSBUF_PUT(sb); 901 return error; 902 } 903 904 /* 905 * Change current working directory to a given file descriptor. 906 */ 907 /* ARGSUSED */ 908 int 909 sys_fchdir(struct lwp *l, void *v, register_t *retval) 910 { 911 struct sys_fchdir_args /* { 912 syscallarg(int) fd; 913 } */ *uap = v; 914 struct proc *p = l->l_proc; 915 struct filedesc *fdp = p->p_fd; 916 struct cwdinfo *cwdi = p->p_cwdi; 917 struct vnode *vp, *tdp; 918 struct mount *mp; 919 struct file *fp; 920 int error; 921 922 /* getvnode() will use the descriptor for us */ 923 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 924 return (error); 925 vp = (struct vnode *)fp->f_data; 926 927 VREF(vp); 928 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 929 if (vp->v_type != VDIR) 930 error = ENOTDIR; 931 else 932 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, l); 933 while (!error && (mp = vp->v_mountedhere) != NULL) { 934 if (vfs_busy(mp, 0, 0)) 935 continue; 936 error = VFS_ROOT(mp, &tdp); 937 vfs_unbusy(mp); 938 if (error) 939 break; 940 vput(vp); 941 vp = tdp; 942 } 943 if (error) { 944 vput(vp); 945 goto out; 946 } 947 VOP_UNLOCK(vp, 0); 948 949 /* 950 * Disallow changing to a directory not under the process's 951 * current root directory (if there is one). 952 */ 953 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 954 vrele(vp); 955 error = EPERM; /* operation not permitted */ 956 goto out; 957 } 958 959 vrele(cwdi->cwdi_cdir); 960 cwdi->cwdi_cdir = vp; 961 out: 962 FILE_UNUSE(fp, l); 963 return (error); 964 } 965 966 /* 967 * Change this process's notion of the root directory to a given file 968 * descriptor. 969 */ 970 int 971 sys_fchroot(struct lwp *l, void *v, register_t *retval) 972 { 973 struct sys_fchroot_args *uap = v; 974 struct proc *p = l->l_proc; 975 struct filedesc *fdp = p->p_fd; 976 struct cwdinfo *cwdi = p->p_cwdi; 977 struct vnode *vp; 978 struct file *fp; 979 int error; 980 981 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 982 return error; 983 /* getvnode() will use the descriptor for us */ 984 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 985 return error; 986 vp = (struct vnode *) fp->f_data; 987 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 988 if (vp->v_type != VDIR) 989 error = ENOTDIR; 990 else 991 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, l); 992 VOP_UNLOCK(vp, 0); 993 if (error) 994 goto out; 995 VREF(vp); 996 997 /* 998 * Prevent escaping from chroot by putting the root under 999 * the working directory. Silently chdir to / if we aren't 1000 * already there. 1001 */ 1002 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1003 /* 1004 * XXX would be more failsafe to change directory to a 1005 * deadfs node here instead 1006 */ 1007 vrele(cwdi->cwdi_cdir); 1008 VREF(vp); 1009 cwdi->cwdi_cdir = vp; 1010 } 1011 1012 if (cwdi->cwdi_rdir != NULL) 1013 vrele(cwdi->cwdi_rdir); 1014 cwdi->cwdi_rdir = vp; 1015 out: 1016 FILE_UNUSE(fp, l); 1017 return (error); 1018 } 1019 1020 /* 1021 * Change current working directory (``.''). 1022 */ 1023 /* ARGSUSED */ 1024 int 1025 sys_chdir(struct lwp *l, void *v, register_t *retval) 1026 { 1027 struct sys_chdir_args /* { 1028 syscallarg(const char *) path; 1029 } */ *uap = v; 1030 struct proc *p = l->l_proc; 1031 struct cwdinfo *cwdi = p->p_cwdi; 1032 int error; 1033 struct nameidata nd; 1034 1035 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1036 SCARG(uap, path), l); 1037 if ((error = change_dir(&nd, l)) != 0) 1038 return (error); 1039 vrele(cwdi->cwdi_cdir); 1040 cwdi->cwdi_cdir = nd.ni_vp; 1041 return (0); 1042 } 1043 1044 /* 1045 * Change notion of root (``/'') directory. 1046 */ 1047 /* ARGSUSED */ 1048 int 1049 sys_chroot(struct lwp *l, void *v, register_t *retval) 1050 { 1051 struct sys_chroot_args /* { 1052 syscallarg(const char *) path; 1053 } */ *uap = v; 1054 struct proc *p = l->l_proc; 1055 struct cwdinfo *cwdi = p->p_cwdi; 1056 struct vnode *vp; 1057 int error; 1058 struct nameidata nd; 1059 1060 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1061 return (error); 1062 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1063 SCARG(uap, path), l); 1064 if ((error = change_dir(&nd, l)) != 0) 1065 return (error); 1066 if (cwdi->cwdi_rdir != NULL) 1067 vrele(cwdi->cwdi_rdir); 1068 vp = nd.ni_vp; 1069 cwdi->cwdi_rdir = vp; 1070 1071 /* 1072 * Prevent escaping from chroot by putting the root under 1073 * the working directory. Silently chdir to / if we aren't 1074 * already there. 1075 */ 1076 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1077 /* 1078 * XXX would be more failsafe to change directory to a 1079 * deadfs node here instead 1080 */ 1081 vrele(cwdi->cwdi_cdir); 1082 VREF(vp); 1083 cwdi->cwdi_cdir = vp; 1084 } 1085 1086 return (0); 1087 } 1088 1089 /* 1090 * Common routine for chroot and chdir. 1091 */ 1092 static int 1093 change_dir(struct nameidata *ndp, struct lwp *l) 1094 { 1095 struct vnode *vp; 1096 int error; 1097 1098 if ((error = namei(ndp)) != 0) 1099 return (error); 1100 vp = ndp->ni_vp; 1101 if (vp->v_type != VDIR) 1102 error = ENOTDIR; 1103 else 1104 error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_ucred, l); 1105 1106 if (error) 1107 vput(vp); 1108 else 1109 VOP_UNLOCK(vp, 0); 1110 return (error); 1111 } 1112 1113 /* 1114 * Check permissions, allocate an open file structure, 1115 * and call the device open routine if any. 1116 */ 1117 int 1118 sys_open(struct lwp *l, void *v, register_t *retval) 1119 { 1120 struct sys_open_args /* { 1121 syscallarg(const char *) path; 1122 syscallarg(int) flags; 1123 syscallarg(int) mode; 1124 } */ *uap = v; 1125 struct proc *p = l->l_proc; 1126 struct cwdinfo *cwdi = p->p_cwdi; 1127 struct filedesc *fdp = p->p_fd; 1128 struct file *fp; 1129 struct vnode *vp; 1130 int flags, cmode; 1131 int type, indx, error; 1132 struct flock lf; 1133 struct nameidata nd; 1134 1135 flags = FFLAGS(SCARG(uap, flags)); 1136 if ((flags & (FREAD | FWRITE)) == 0) 1137 return (EINVAL); 1138 /* falloc() will use the file descriptor for us */ 1139 if ((error = falloc(p, &fp, &indx)) != 0) 1140 return (error); 1141 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1142 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1143 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1144 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1145 FILE_UNUSE(fp, l); 1146 fdp->fd_ofiles[indx] = NULL; 1147 ffree(fp); 1148 if ((error == EDUPFD || error == EMOVEFD) && 1149 l->l_dupfd >= 0 && /* XXX from fdopen */ 1150 (error = 1151 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1152 *retval = indx; 1153 return (0); 1154 } 1155 if (error == ERESTART) 1156 error = EINTR; 1157 fdremove(fdp, indx); 1158 return (error); 1159 } 1160 l->l_dupfd = 0; 1161 vp = nd.ni_vp; 1162 fp->f_flag = flags & FMASK; 1163 fp->f_type = DTYPE_VNODE; 1164 fp->f_ops = &vnops; 1165 fp->f_data = vp; 1166 if (flags & (O_EXLOCK | O_SHLOCK)) { 1167 lf.l_whence = SEEK_SET; 1168 lf.l_start = 0; 1169 lf.l_len = 0; 1170 if (flags & O_EXLOCK) 1171 lf.l_type = F_WRLCK; 1172 else 1173 lf.l_type = F_RDLCK; 1174 type = F_FLOCK; 1175 if ((flags & FNONBLOCK) == 0) 1176 type |= F_WAIT; 1177 VOP_UNLOCK(vp, 0); 1178 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1179 if (error) { 1180 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1181 FILE_UNUSE(fp, l); 1182 ffree(fp); 1183 fdremove(fdp, indx); 1184 return (error); 1185 } 1186 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1187 fp->f_flag |= FHASLOCK; 1188 } 1189 VOP_UNLOCK(vp, 0); 1190 *retval = indx; 1191 FILE_SET_MATURE(fp); 1192 FILE_UNUSE(fp, l); 1193 return (0); 1194 } 1195 1196 /* 1197 * Get file handle system call 1198 */ 1199 int 1200 sys_getfh(struct lwp *l, void *v, register_t *retval) 1201 { 1202 struct sys_getfh_args /* { 1203 syscallarg(char *) fname; 1204 syscallarg(fhandle_t *) fhp; 1205 } */ *uap = v; 1206 struct proc *p = l->l_proc; 1207 struct vnode *vp; 1208 fhandle_t fh; 1209 int error; 1210 struct nameidata nd; 1211 1212 /* 1213 * Must be super user 1214 */ 1215 error = suser(p->p_ucred, &p->p_acflag); 1216 if (error) 1217 return (error); 1218 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1219 SCARG(uap, fname), l); 1220 error = namei(&nd); 1221 if (error) 1222 return (error); 1223 vp = nd.ni_vp; 1224 if (vp->v_mount->mnt_op->vfs_vptofh == NULL) 1225 return EOPNOTSUPP; 1226 memset(&fh, 0, sizeof(fh)); 1227 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1228 error = VFS_VPTOFH(vp, &fh.fh_fid); 1229 vput(vp); 1230 if (error) 1231 return (error); 1232 KASSERT(fh.fh_fid.fid_len <= _VFS_MAXFIDSZ); 1233 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1234 return (error); 1235 } 1236 1237 /* 1238 * Open a file given a file handle. 1239 * 1240 * Check permissions, allocate an open file structure, 1241 * and call the device open routine if any. 1242 */ 1243 int 1244 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1245 { 1246 struct sys_fhopen_args /* { 1247 syscallarg(const fhandle_t *) fhp; 1248 syscallarg(int) flags; 1249 } */ *uap = v; 1250 struct proc *p = l->l_proc; 1251 struct filedesc *fdp = p->p_fd; 1252 struct file *fp; 1253 struct vnode *vp = NULL; 1254 struct mount *mp; 1255 struct ucred *cred = p->p_ucred; 1256 int flags; 1257 struct file *nfp; 1258 int type, indx, error=0; 1259 struct flock lf; 1260 struct vattr va; 1261 fhandle_t fh; 1262 1263 /* 1264 * Must be super user 1265 */ 1266 if ((error = suser(p->p_ucred, &p->p_acflag))) 1267 return (error); 1268 1269 flags = FFLAGS(SCARG(uap, flags)); 1270 if ((flags & (FREAD | FWRITE)) == 0) 1271 return (EINVAL); 1272 if ((flags & O_CREAT)) 1273 return (EINVAL); 1274 /* falloc() will use the file descriptor for us */ 1275 if ((error = falloc(p, &nfp, &indx)) != 0) 1276 return (error); 1277 fp = nfp; 1278 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1279 goto bad; 1280 1281 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1282 error = ESTALE; 1283 goto bad; 1284 } 1285 1286 if (mp->mnt_op->vfs_fhtovp == NULL) { 1287 error = EOPNOTSUPP; 1288 goto bad; 1289 } 1290 1291 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1292 vp = NULL; /* most likely unnecessary sanity for bad: */ 1293 goto bad; 1294 } 1295 1296 /* Now do an effective vn_open */ 1297 1298 if (vp->v_type == VSOCK) { 1299 error = EOPNOTSUPP; 1300 goto bad; 1301 } 1302 if (flags & FREAD) { 1303 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1304 goto bad; 1305 } 1306 if (flags & (FWRITE | O_TRUNC)) { 1307 if (vp->v_type == VDIR) { 1308 error = EISDIR; 1309 goto bad; 1310 } 1311 if ((error = vn_writechk(vp)) != 0 || 1312 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1313 goto bad; 1314 } 1315 if (flags & O_TRUNC) { 1316 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1317 goto bad; 1318 VOP_UNLOCK(vp, 0); /* XXX */ 1319 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1320 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1321 VATTR_NULL(&va); 1322 va.va_size = 0; 1323 error = VOP_SETATTR(vp, &va, cred, l); 1324 vn_finished_write(mp, 0); 1325 if (error) 1326 goto bad; 1327 } 1328 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1329 goto bad; 1330 if (vp->v_type == VREG && 1331 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1332 error = EIO; 1333 goto bad; 1334 } 1335 if (flags & FWRITE) 1336 vp->v_writecount++; 1337 1338 /* done with modified vn_open, now finish what sys_open does. */ 1339 1340 fp->f_flag = flags & FMASK; 1341 fp->f_type = DTYPE_VNODE; 1342 fp->f_ops = &vnops; 1343 fp->f_data = vp; 1344 if (flags & (O_EXLOCK | O_SHLOCK)) { 1345 lf.l_whence = SEEK_SET; 1346 lf.l_start = 0; 1347 lf.l_len = 0; 1348 if (flags & O_EXLOCK) 1349 lf.l_type = F_WRLCK; 1350 else 1351 lf.l_type = F_RDLCK; 1352 type = F_FLOCK; 1353 if ((flags & FNONBLOCK) == 0) 1354 type |= F_WAIT; 1355 VOP_UNLOCK(vp, 0); 1356 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1357 if (error) { 1358 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1359 FILE_UNUSE(fp, l); 1360 ffree(fp); 1361 fdremove(fdp, indx); 1362 return (error); 1363 } 1364 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1365 fp->f_flag |= FHASLOCK; 1366 } 1367 VOP_UNLOCK(vp, 0); 1368 *retval = indx; 1369 FILE_SET_MATURE(fp); 1370 FILE_UNUSE(fp, l); 1371 return (0); 1372 1373 bad: 1374 FILE_UNUSE(fp, l); 1375 ffree(fp); 1376 fdremove(fdp, indx); 1377 if (vp != NULL) 1378 vput(vp); 1379 return (error); 1380 } 1381 1382 /* ARGSUSED */ 1383 int 1384 sys___fhstat30(struct lwp *l, void *v, register_t *retval) 1385 { 1386 struct sys___fhstat30_args /* { 1387 syscallarg(const fhandle_t *) fhp; 1388 syscallarg(struct stat *) sb; 1389 } */ *uap = v; 1390 struct proc *p = l->l_proc; 1391 struct stat sb; 1392 int error; 1393 fhandle_t fh; 1394 struct mount *mp; 1395 struct vnode *vp; 1396 1397 /* 1398 * Must be super user 1399 */ 1400 if ((error = suser(p->p_ucred, &p->p_acflag))) 1401 return (error); 1402 1403 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1404 return (error); 1405 1406 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1407 return (ESTALE); 1408 if (mp->mnt_op->vfs_fhtovp == NULL) 1409 return EOPNOTSUPP; 1410 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1411 return (error); 1412 error = vn_stat(vp, &sb, l); 1413 vput(vp); 1414 if (error) 1415 return (error); 1416 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1417 return (error); 1418 } 1419 1420 /* ARGSUSED */ 1421 int 1422 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1423 { 1424 struct sys_fhstatvfs1_args /* { 1425 syscallarg(const fhandle_t *) fhp; 1426 syscallarg(struct statvfs *) buf; 1427 syscallarg(int) flags; 1428 } */ *uap = v; 1429 struct proc *p = l->l_proc; 1430 struct statvfs *sb; 1431 fhandle_t fh; 1432 struct mount *mp; 1433 struct vnode *vp; 1434 int error; 1435 1436 /* 1437 * Must be super user 1438 */ 1439 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1440 return error; 1441 1442 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1443 return error; 1444 1445 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1446 return ESTALE; 1447 if (mp->mnt_op->vfs_fhtovp == NULL) 1448 return EOPNOTSUPP; 1449 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1450 return error; 1451 1452 mp = vp->v_mount; 1453 sb = STATVFSBUF_GET(); 1454 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) { 1455 vput(vp); 1456 goto out; 1457 } 1458 vput(vp); 1459 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1460 out: 1461 STATVFSBUF_PUT(sb); 1462 return error; 1463 } 1464 1465 /* 1466 * Create a special file. 1467 */ 1468 /* ARGSUSED */ 1469 int 1470 sys_mknod(struct lwp *l, void *v, register_t *retval) 1471 { 1472 struct sys_mknod_args /* { 1473 syscallarg(const char *) path; 1474 syscallarg(int) mode; 1475 syscallarg(int) dev; 1476 } */ *uap = v; 1477 struct proc *p = l->l_proc; 1478 struct vnode *vp; 1479 struct mount *mp; 1480 struct vattr vattr; 1481 int error; 1482 int whiteout = 0; 1483 struct nameidata nd; 1484 1485 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1486 return (error); 1487 restart: 1488 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1489 if ((error = namei(&nd)) != 0) 1490 return (error); 1491 vp = nd.ni_vp; 1492 if (vp != NULL) 1493 error = EEXIST; 1494 else { 1495 VATTR_NULL(&vattr); 1496 vattr.va_mode = 1497 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1498 vattr.va_rdev = SCARG(uap, dev); 1499 whiteout = 0; 1500 1501 switch (SCARG(uap, mode) & S_IFMT) { 1502 case S_IFMT: /* used by badsect to flag bad sectors */ 1503 vattr.va_type = VBAD; 1504 break; 1505 case S_IFCHR: 1506 vattr.va_type = VCHR; 1507 break; 1508 case S_IFBLK: 1509 vattr.va_type = VBLK; 1510 break; 1511 case S_IFWHT: 1512 whiteout = 1; 1513 break; 1514 default: 1515 error = EINVAL; 1516 break; 1517 } 1518 } 1519 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1520 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1521 if (nd.ni_dvp == vp) 1522 vrele(nd.ni_dvp); 1523 else 1524 vput(nd.ni_dvp); 1525 if (vp) 1526 vrele(vp); 1527 if ((error = vn_start_write(NULL, &mp, 1528 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1529 return (error); 1530 goto restart; 1531 } 1532 if (!error) { 1533 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1534 if (whiteout) { 1535 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1536 if (error) 1537 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1538 vput(nd.ni_dvp); 1539 } else { 1540 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1541 &nd.ni_cnd, &vattr); 1542 if (error == 0) 1543 vput(nd.ni_vp); 1544 } 1545 } else { 1546 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1547 if (nd.ni_dvp == vp) 1548 vrele(nd.ni_dvp); 1549 else 1550 vput(nd.ni_dvp); 1551 if (vp) 1552 vrele(vp); 1553 } 1554 vn_finished_write(mp, 0); 1555 return (error); 1556 } 1557 1558 /* 1559 * Create a named pipe. 1560 */ 1561 /* ARGSUSED */ 1562 int 1563 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1564 { 1565 struct sys_mkfifo_args /* { 1566 syscallarg(const char *) path; 1567 syscallarg(int) mode; 1568 } */ *uap = v; 1569 struct proc *p = l->l_proc; 1570 struct mount *mp; 1571 struct vattr vattr; 1572 int error; 1573 struct nameidata nd; 1574 1575 restart: 1576 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1577 if ((error = namei(&nd)) != 0) 1578 return (error); 1579 if (nd.ni_vp != NULL) { 1580 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1581 if (nd.ni_dvp == nd.ni_vp) 1582 vrele(nd.ni_dvp); 1583 else 1584 vput(nd.ni_dvp); 1585 vrele(nd.ni_vp); 1586 return (EEXIST); 1587 } 1588 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1589 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1590 if (nd.ni_dvp == nd.ni_vp) 1591 vrele(nd.ni_dvp); 1592 else 1593 vput(nd.ni_dvp); 1594 if (nd.ni_vp) 1595 vrele(nd.ni_vp); 1596 if ((error = vn_start_write(NULL, &mp, 1597 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1598 return (error); 1599 goto restart; 1600 } 1601 VATTR_NULL(&vattr); 1602 vattr.va_type = VFIFO; 1603 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1604 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1605 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1606 if (error == 0) 1607 vput(nd.ni_vp); 1608 vn_finished_write(mp, 0); 1609 return (error); 1610 } 1611 1612 /* 1613 * Make a hard file link. 1614 */ 1615 /* ARGSUSED */ 1616 int 1617 sys_link(struct lwp *l, void *v, register_t *retval) 1618 { 1619 struct sys_link_args /* { 1620 syscallarg(const char *) path; 1621 syscallarg(const char *) link; 1622 } */ *uap = v; 1623 struct proc *p = l->l_proc; 1624 struct vnode *vp; 1625 struct mount *mp; 1626 struct nameidata nd; 1627 int error; 1628 1629 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1630 if ((error = namei(&nd)) != 0) 1631 return (error); 1632 vp = nd.ni_vp; 1633 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1634 vrele(vp); 1635 return (error); 1636 } 1637 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1638 if ((error = namei(&nd)) != 0) 1639 goto out; 1640 if (nd.ni_vp) { 1641 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1642 if (nd.ni_dvp == nd.ni_vp) 1643 vrele(nd.ni_dvp); 1644 else 1645 vput(nd.ni_dvp); 1646 vrele(nd.ni_vp); 1647 error = EEXIST; 1648 goto out; 1649 } 1650 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1651 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 1652 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1653 out: 1654 vrele(vp); 1655 vn_finished_write(mp, 0); 1656 return (error); 1657 } 1658 1659 /* 1660 * Make a symbolic link. 1661 */ 1662 /* ARGSUSED */ 1663 int 1664 sys_symlink(struct lwp *l, void *v, register_t *retval) 1665 { 1666 struct sys_symlink_args /* { 1667 syscallarg(const char *) path; 1668 syscallarg(const char *) link; 1669 } */ *uap = v; 1670 struct proc *p = l->l_proc; 1671 struct mount *mp; 1672 struct vattr vattr; 1673 char *path; 1674 int error; 1675 struct nameidata nd; 1676 1677 path = PNBUF_GET(); 1678 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1679 if (error) 1680 goto out; 1681 restart: 1682 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1683 if ((error = namei(&nd)) != 0) 1684 goto out; 1685 if (nd.ni_vp) { 1686 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1687 if (nd.ni_dvp == nd.ni_vp) 1688 vrele(nd.ni_dvp); 1689 else 1690 vput(nd.ni_dvp); 1691 vrele(nd.ni_vp); 1692 error = EEXIST; 1693 goto out; 1694 } 1695 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1696 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1697 if (nd.ni_dvp == nd.ni_vp) 1698 vrele(nd.ni_dvp); 1699 else 1700 vput(nd.ni_dvp); 1701 if ((error = vn_start_write(NULL, &mp, 1702 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1703 return (error); 1704 goto restart; 1705 } 1706 VATTR_NULL(&vattr); 1707 vattr.va_type = VLNK; 1708 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1709 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1710 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1711 if (error == 0) 1712 vput(nd.ni_vp); 1713 vn_finished_write(mp, 0); 1714 out: 1715 PNBUF_PUT(path); 1716 return (error); 1717 } 1718 1719 /* 1720 * Delete a whiteout from the filesystem. 1721 */ 1722 /* ARGSUSED */ 1723 int 1724 sys_undelete(struct lwp *l, void *v, register_t *retval) 1725 { 1726 struct sys_undelete_args /* { 1727 syscallarg(const char *) path; 1728 } */ *uap = v; 1729 struct proc *p = l->l_proc; 1730 int error; 1731 struct mount *mp; 1732 struct nameidata nd; 1733 1734 restart: 1735 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1736 SCARG(uap, path), l); 1737 error = namei(&nd); 1738 if (error) 1739 return (error); 1740 1741 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1742 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1743 if (nd.ni_dvp == nd.ni_vp) 1744 vrele(nd.ni_dvp); 1745 else 1746 vput(nd.ni_dvp); 1747 if (nd.ni_vp) 1748 vrele(nd.ni_vp); 1749 return (EEXIST); 1750 } 1751 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1752 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1753 if (nd.ni_dvp == nd.ni_vp) 1754 vrele(nd.ni_dvp); 1755 else 1756 vput(nd.ni_dvp); 1757 if ((error = vn_start_write(NULL, &mp, 1758 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1759 return (error); 1760 goto restart; 1761 } 1762 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1763 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1764 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1765 vput(nd.ni_dvp); 1766 vn_finished_write(mp, 0); 1767 return (error); 1768 } 1769 1770 /* 1771 * Delete a name from the filesystem. 1772 */ 1773 /* ARGSUSED */ 1774 int 1775 sys_unlink(struct lwp *l, void *v, register_t *retval) 1776 { 1777 struct sys_unlink_args /* { 1778 syscallarg(const char *) path; 1779 } */ *uap = v; 1780 struct proc *p = l->l_proc; 1781 struct mount *mp; 1782 struct vnode *vp; 1783 int error; 1784 struct nameidata nd; 1785 1786 restart: 1787 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1788 SCARG(uap, path), l); 1789 if ((error = namei(&nd)) != 0) 1790 return (error); 1791 vp = nd.ni_vp; 1792 1793 /* 1794 * The root of a mounted filesystem cannot be deleted. 1795 */ 1796 if (vp->v_flag & VROOT) { 1797 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1798 if (nd.ni_dvp == vp) 1799 vrele(nd.ni_dvp); 1800 else 1801 vput(nd.ni_dvp); 1802 vput(vp); 1803 error = EBUSY; 1804 goto out; 1805 } 1806 1807 #ifdef VERIFIED_EXEC 1808 /* Handle remove requests for veriexec entries. */ 1809 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 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 goto out; 1817 } 1818 #endif /* VERIFIED_EXEC */ 1819 1820 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1821 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1822 if (nd.ni_dvp == vp) 1823 vrele(nd.ni_dvp); 1824 else 1825 vput(nd.ni_dvp); 1826 vput(vp); 1827 if ((error = vn_start_write(NULL, &mp, 1828 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1829 return (error); 1830 goto restart; 1831 } 1832 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 1833 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 1834 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1835 vn_finished_write(mp, 0); 1836 out: 1837 return (error); 1838 } 1839 1840 /* 1841 * Reposition read/write file offset. 1842 */ 1843 int 1844 sys_lseek(struct lwp *l, void *v, register_t *retval) 1845 { 1846 struct sys_lseek_args /* { 1847 syscallarg(int) fd; 1848 syscallarg(int) pad; 1849 syscallarg(off_t) offset; 1850 syscallarg(int) whence; 1851 } */ *uap = v; 1852 struct proc *p = l->l_proc; 1853 struct ucred *cred = p->p_ucred; 1854 struct filedesc *fdp = p->p_fd; 1855 struct file *fp; 1856 struct vnode *vp; 1857 struct vattr vattr; 1858 off_t newoff; 1859 int error; 1860 1861 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1862 return (EBADF); 1863 1864 FILE_USE(fp); 1865 1866 vp = (struct vnode *)fp->f_data; 1867 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1868 error = ESPIPE; 1869 goto out; 1870 } 1871 1872 switch (SCARG(uap, whence)) { 1873 case SEEK_CUR: 1874 newoff = fp->f_offset + SCARG(uap, offset); 1875 break; 1876 case SEEK_END: 1877 error = VOP_GETATTR(vp, &vattr, cred, l); 1878 if (error) 1879 goto out; 1880 newoff = SCARG(uap, offset) + vattr.va_size; 1881 break; 1882 case SEEK_SET: 1883 newoff = SCARG(uap, offset); 1884 break; 1885 default: 1886 error = EINVAL; 1887 goto out; 1888 } 1889 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1890 goto out; 1891 1892 *(off_t *)retval = fp->f_offset = newoff; 1893 out: 1894 FILE_UNUSE(fp, l); 1895 return (error); 1896 } 1897 1898 /* 1899 * Positional read system call. 1900 */ 1901 int 1902 sys_pread(struct lwp *l, void *v, register_t *retval) 1903 { 1904 struct sys_pread_args /* { 1905 syscallarg(int) fd; 1906 syscallarg(void *) buf; 1907 syscallarg(size_t) nbyte; 1908 syscallarg(off_t) offset; 1909 } */ *uap = v; 1910 struct proc *p = l->l_proc; 1911 struct filedesc *fdp = p->p_fd; 1912 struct file *fp; 1913 struct vnode *vp; 1914 off_t offset; 1915 int error, fd = SCARG(uap, fd); 1916 1917 if ((fp = fd_getfile(fdp, fd)) == NULL) 1918 return (EBADF); 1919 1920 if ((fp->f_flag & FREAD) == 0) { 1921 simple_unlock(&fp->f_slock); 1922 return (EBADF); 1923 } 1924 1925 FILE_USE(fp); 1926 1927 vp = (struct vnode *)fp->f_data; 1928 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1929 error = ESPIPE; 1930 goto out; 1931 } 1932 1933 offset = SCARG(uap, offset); 1934 1935 /* 1936 * XXX This works because no file systems actually 1937 * XXX take any action on the seek operation. 1938 */ 1939 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1940 goto out; 1941 1942 /* dofileread() will unuse the descriptor for us */ 1943 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1944 &offset, 0, retval)); 1945 1946 out: 1947 FILE_UNUSE(fp, l); 1948 return (error); 1949 } 1950 1951 /* 1952 * Positional scatter read system call. 1953 */ 1954 int 1955 sys_preadv(struct lwp *l, void *v, register_t *retval) 1956 { 1957 struct sys_preadv_args /* { 1958 syscallarg(int) fd; 1959 syscallarg(const struct iovec *) iovp; 1960 syscallarg(int) iovcnt; 1961 syscallarg(off_t) offset; 1962 } */ *uap = v; 1963 struct proc *p = l->l_proc; 1964 struct filedesc *fdp = p->p_fd; 1965 struct file *fp; 1966 struct vnode *vp; 1967 off_t offset; 1968 int error, fd = SCARG(uap, fd); 1969 1970 if ((fp = fd_getfile(fdp, fd)) == NULL) 1971 return (EBADF); 1972 1973 if ((fp->f_flag & FREAD) == 0) { 1974 simple_unlock(&fp->f_slock); 1975 return (EBADF); 1976 } 1977 1978 FILE_USE(fp); 1979 1980 vp = (struct vnode *)fp->f_data; 1981 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1982 error = ESPIPE; 1983 goto out; 1984 } 1985 1986 offset = SCARG(uap, offset); 1987 1988 /* 1989 * XXX This works because no file systems actually 1990 * XXX take any action on the seek operation. 1991 */ 1992 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1993 goto out; 1994 1995 /* dofilereadv() will unuse the descriptor for us */ 1996 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1997 &offset, 0, retval)); 1998 1999 out: 2000 FILE_UNUSE(fp, l); 2001 return (error); 2002 } 2003 2004 /* 2005 * Positional write system call. 2006 */ 2007 int 2008 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2009 { 2010 struct sys_pwrite_args /* { 2011 syscallarg(int) fd; 2012 syscallarg(const void *) buf; 2013 syscallarg(size_t) nbyte; 2014 syscallarg(off_t) offset; 2015 } */ *uap = v; 2016 struct proc *p = l->l_proc; 2017 struct filedesc *fdp = p->p_fd; 2018 struct file *fp; 2019 struct vnode *vp; 2020 off_t offset; 2021 int error, fd = SCARG(uap, fd); 2022 2023 if ((fp = fd_getfile(fdp, fd)) == NULL) 2024 return (EBADF); 2025 2026 if ((fp->f_flag & FWRITE) == 0) { 2027 simple_unlock(&fp->f_slock); 2028 return (EBADF); 2029 } 2030 2031 FILE_USE(fp); 2032 2033 vp = (struct vnode *)fp->f_data; 2034 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2035 error = ESPIPE; 2036 goto out; 2037 } 2038 2039 offset = SCARG(uap, offset); 2040 2041 /* 2042 * XXX This works because no file systems actually 2043 * XXX take any action on the seek operation. 2044 */ 2045 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2046 goto out; 2047 2048 /* dofilewrite() will unuse the descriptor for us */ 2049 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2050 &offset, 0, retval)); 2051 2052 out: 2053 FILE_UNUSE(fp, l); 2054 return (error); 2055 } 2056 2057 /* 2058 * Positional gather write system call. 2059 */ 2060 int 2061 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2062 { 2063 struct sys_pwritev_args /* { 2064 syscallarg(int) fd; 2065 syscallarg(const struct iovec *) iovp; 2066 syscallarg(int) iovcnt; 2067 syscallarg(off_t) offset; 2068 } */ *uap = v; 2069 struct proc *p = l->l_proc; 2070 struct filedesc *fdp = p->p_fd; 2071 struct file *fp; 2072 struct vnode *vp; 2073 off_t offset; 2074 int error, fd = SCARG(uap, fd); 2075 2076 if ((fp = fd_getfile(fdp, fd)) == NULL) 2077 return (EBADF); 2078 2079 if ((fp->f_flag & FWRITE) == 0) { 2080 simple_unlock(&fp->f_slock); 2081 return (EBADF); 2082 } 2083 2084 FILE_USE(fp); 2085 2086 vp = (struct vnode *)fp->f_data; 2087 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2088 error = ESPIPE; 2089 goto out; 2090 } 2091 2092 offset = SCARG(uap, offset); 2093 2094 /* 2095 * XXX This works because no file systems actually 2096 * XXX take any action on the seek operation. 2097 */ 2098 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2099 goto out; 2100 2101 /* dofilewritev() will unuse the descriptor for us */ 2102 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2103 &offset, 0, retval)); 2104 2105 out: 2106 FILE_UNUSE(fp, l); 2107 return (error); 2108 } 2109 2110 /* 2111 * Check access permissions. 2112 */ 2113 int 2114 sys_access(struct lwp *l, void *v, register_t *retval) 2115 { 2116 struct sys_access_args /* { 2117 syscallarg(const char *) path; 2118 syscallarg(int) flags; 2119 } */ *uap = v; 2120 struct proc *p = l->l_proc; 2121 struct ucred *cred; 2122 struct vnode *vp; 2123 int error, flags; 2124 struct nameidata nd; 2125 2126 cred = crdup(p->p_ucred); 2127 cred->cr_uid = p->p_cred->p_ruid; 2128 cred->cr_gid = p->p_cred->p_rgid; 2129 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2130 SCARG(uap, path), l); 2131 /* Override default credentials */ 2132 nd.ni_cnd.cn_cred = cred; 2133 if ((error = namei(&nd)) != 0) 2134 goto out; 2135 vp = nd.ni_vp; 2136 2137 /* Flags == 0 means only check for existence. */ 2138 if (SCARG(uap, flags)) { 2139 flags = 0; 2140 if (SCARG(uap, flags) & R_OK) 2141 flags |= VREAD; 2142 if (SCARG(uap, flags) & W_OK) 2143 flags |= VWRITE; 2144 if (SCARG(uap, flags) & X_OK) 2145 flags |= VEXEC; 2146 2147 error = VOP_ACCESS(vp, flags, cred, l); 2148 if (!error && (flags & VWRITE)) 2149 error = vn_writechk(vp); 2150 } 2151 vput(vp); 2152 out: 2153 crfree(cred); 2154 return (error); 2155 } 2156 2157 /* 2158 * Get file status; this version follows links. 2159 */ 2160 /* ARGSUSED */ 2161 int 2162 sys___stat30(struct lwp *l, void *v, register_t *retval) 2163 { 2164 struct sys___stat30_args /* { 2165 syscallarg(const char *) path; 2166 syscallarg(struct stat *) ub; 2167 } */ *uap = v; 2168 struct stat sb; 2169 int error; 2170 struct nameidata nd; 2171 2172 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2173 SCARG(uap, path), l); 2174 if ((error = namei(&nd)) != 0) 2175 return (error); 2176 error = vn_stat(nd.ni_vp, &sb, l); 2177 vput(nd.ni_vp); 2178 if (error) 2179 return (error); 2180 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2181 return (error); 2182 } 2183 2184 /* 2185 * Get file status; this version does not follow links. 2186 */ 2187 /* ARGSUSED */ 2188 int 2189 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2190 { 2191 struct sys___lstat30_args /* { 2192 syscallarg(const char *) path; 2193 syscallarg(struct stat *) ub; 2194 } */ *uap = v; 2195 struct stat sb; 2196 int error; 2197 struct nameidata nd; 2198 2199 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2200 SCARG(uap, path), l); 2201 if ((error = namei(&nd)) != 0) 2202 return (error); 2203 error = vn_stat(nd.ni_vp, &sb, l); 2204 vput(nd.ni_vp); 2205 if (error) 2206 return (error); 2207 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2208 return (error); 2209 } 2210 2211 /* 2212 * Get configurable pathname variables. 2213 */ 2214 /* ARGSUSED */ 2215 int 2216 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2217 { 2218 struct sys_pathconf_args /* { 2219 syscallarg(const char *) path; 2220 syscallarg(int) name; 2221 } */ *uap = v; 2222 int error; 2223 struct nameidata nd; 2224 2225 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2226 SCARG(uap, path), l); 2227 if ((error = namei(&nd)) != 0) 2228 return (error); 2229 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2230 vput(nd.ni_vp); 2231 return (error); 2232 } 2233 2234 /* 2235 * Return target name of a symbolic link. 2236 */ 2237 /* ARGSUSED */ 2238 int 2239 sys_readlink(struct lwp *l, void *v, register_t *retval) 2240 { 2241 struct sys_readlink_args /* { 2242 syscallarg(const char *) path; 2243 syscallarg(char *) buf; 2244 syscallarg(size_t) count; 2245 } */ *uap = v; 2246 struct proc *p = l->l_proc; 2247 struct vnode *vp; 2248 struct iovec aiov; 2249 struct uio auio; 2250 int error; 2251 struct nameidata nd; 2252 2253 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2254 SCARG(uap, path), l); 2255 if ((error = namei(&nd)) != 0) 2256 return (error); 2257 vp = nd.ni_vp; 2258 if (vp->v_type != VLNK) 2259 error = EINVAL; 2260 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2261 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, l)) == 0) { 2262 aiov.iov_base = SCARG(uap, buf); 2263 aiov.iov_len = SCARG(uap, count); 2264 auio.uio_iov = &aiov; 2265 auio.uio_iovcnt = 1; 2266 auio.uio_offset = 0; 2267 auio.uio_rw = UIO_READ; 2268 KASSERT(l == curlwp); 2269 auio.uio_vmspace = l->l_proc->p_vmspace; 2270 auio.uio_resid = SCARG(uap, count); 2271 error = VOP_READLINK(vp, &auio, p->p_ucred); 2272 } 2273 vput(vp); 2274 *retval = SCARG(uap, count) - auio.uio_resid; 2275 return (error); 2276 } 2277 2278 /* 2279 * Change flags of a file given a path name. 2280 */ 2281 /* ARGSUSED */ 2282 int 2283 sys_chflags(struct lwp *l, void *v, register_t *retval) 2284 { 2285 struct sys_chflags_args /* { 2286 syscallarg(const char *) path; 2287 syscallarg(u_long) flags; 2288 } */ *uap = v; 2289 struct vnode *vp; 2290 int error; 2291 struct nameidata nd; 2292 2293 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2294 if ((error = namei(&nd)) != 0) 2295 return (error); 2296 vp = nd.ni_vp; 2297 error = change_flags(vp, SCARG(uap, flags), l); 2298 vput(vp); 2299 return (error); 2300 } 2301 2302 /* 2303 * Change flags of a file given a file descriptor. 2304 */ 2305 /* ARGSUSED */ 2306 int 2307 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2308 { 2309 struct sys_fchflags_args /* { 2310 syscallarg(int) fd; 2311 syscallarg(u_long) flags; 2312 } */ *uap = v; 2313 struct proc *p = l->l_proc; 2314 struct vnode *vp; 2315 struct file *fp; 2316 int error; 2317 2318 /* getvnode() will use the descriptor for us */ 2319 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2320 return (error); 2321 vp = (struct vnode *)fp->f_data; 2322 error = change_flags(vp, SCARG(uap, flags), l); 2323 VOP_UNLOCK(vp, 0); 2324 FILE_UNUSE(fp, l); 2325 return (error); 2326 } 2327 2328 /* 2329 * Change flags of a file given a path name; this version does 2330 * not follow links. 2331 */ 2332 int 2333 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2334 { 2335 struct sys_lchflags_args /* { 2336 syscallarg(const char *) path; 2337 syscallarg(u_long) flags; 2338 } */ *uap = v; 2339 struct vnode *vp; 2340 int error; 2341 struct nameidata nd; 2342 2343 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2344 if ((error = namei(&nd)) != 0) 2345 return (error); 2346 vp = nd.ni_vp; 2347 error = change_flags(vp, SCARG(uap, flags), l); 2348 vput(vp); 2349 return (error); 2350 } 2351 2352 /* 2353 * Common routine to change flags of a file. 2354 */ 2355 int 2356 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2357 { 2358 struct proc *p = l->l_proc; 2359 struct mount *mp; 2360 struct vattr vattr; 2361 int error; 2362 2363 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2364 return (error); 2365 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2366 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2367 /* 2368 * Non-superusers cannot change the flags on devices, even if they 2369 * own them. 2370 */ 2371 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2372 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 2373 goto out; 2374 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2375 error = EINVAL; 2376 goto out; 2377 } 2378 } 2379 VATTR_NULL(&vattr); 2380 vattr.va_flags = flags; 2381 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2382 out: 2383 vn_finished_write(mp, 0); 2384 return (error); 2385 } 2386 2387 /* 2388 * Change mode of a file given path name; this version follows links. 2389 */ 2390 /* ARGSUSED */ 2391 int 2392 sys_chmod(struct lwp *l, void *v, register_t *retval) 2393 { 2394 struct sys_chmod_args /* { 2395 syscallarg(const char *) path; 2396 syscallarg(int) mode; 2397 } */ *uap = v; 2398 int error; 2399 struct nameidata nd; 2400 2401 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2402 if ((error = namei(&nd)) != 0) 2403 return (error); 2404 2405 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2406 2407 vrele(nd.ni_vp); 2408 return (error); 2409 } 2410 2411 /* 2412 * Change mode of a file given a file descriptor. 2413 */ 2414 /* ARGSUSED */ 2415 int 2416 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2417 { 2418 struct sys_fchmod_args /* { 2419 syscallarg(int) fd; 2420 syscallarg(int) mode; 2421 } */ *uap = v; 2422 struct proc *p = l->l_proc; 2423 struct file *fp; 2424 int error; 2425 2426 /* getvnode() will use the descriptor for us */ 2427 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2428 return (error); 2429 2430 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2431 FILE_UNUSE(fp, l); 2432 return (error); 2433 } 2434 2435 /* 2436 * Change mode of a file given path name; this version does not follow links. 2437 */ 2438 /* ARGSUSED */ 2439 int 2440 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2441 { 2442 struct sys_lchmod_args /* { 2443 syscallarg(const char *) path; 2444 syscallarg(int) mode; 2445 } */ *uap = v; 2446 int error; 2447 struct nameidata nd; 2448 2449 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2450 if ((error = namei(&nd)) != 0) 2451 return (error); 2452 2453 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2454 2455 vrele(nd.ni_vp); 2456 return (error); 2457 } 2458 2459 /* 2460 * Common routine to set mode given a vnode. 2461 */ 2462 static int 2463 change_mode(struct vnode *vp, int mode, struct lwp *l) 2464 { 2465 struct proc *p = l->l_proc; 2466 struct mount *mp; 2467 struct vattr vattr; 2468 int error; 2469 2470 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2471 return (error); 2472 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2473 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2474 VATTR_NULL(&vattr); 2475 vattr.va_mode = mode & ALLPERMS; 2476 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2477 VOP_UNLOCK(vp, 0); 2478 vn_finished_write(mp, 0); 2479 return (error); 2480 } 2481 2482 /* 2483 * Set ownership given a path name; this version follows links. 2484 */ 2485 /* ARGSUSED */ 2486 int 2487 sys_chown(struct lwp *l, void *v, register_t *retval) 2488 { 2489 struct sys_chown_args /* { 2490 syscallarg(const char *) path; 2491 syscallarg(uid_t) uid; 2492 syscallarg(gid_t) gid; 2493 } */ *uap = v; 2494 int error; 2495 struct nameidata nd; 2496 2497 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2498 if ((error = namei(&nd)) != 0) 2499 return (error); 2500 2501 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2502 2503 vrele(nd.ni_vp); 2504 return (error); 2505 } 2506 2507 /* 2508 * Set ownership given a path name; this version follows links. 2509 * Provides POSIX semantics. 2510 */ 2511 /* ARGSUSED */ 2512 int 2513 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2514 { 2515 struct sys_chown_args /* { 2516 syscallarg(const char *) path; 2517 syscallarg(uid_t) uid; 2518 syscallarg(gid_t) gid; 2519 } */ *uap = v; 2520 int error; 2521 struct nameidata nd; 2522 2523 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2524 if ((error = namei(&nd)) != 0) 2525 return (error); 2526 2527 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2528 2529 vrele(nd.ni_vp); 2530 return (error); 2531 } 2532 2533 /* 2534 * Set ownership given a file descriptor. 2535 */ 2536 /* ARGSUSED */ 2537 int 2538 sys_fchown(struct lwp *l, void *v, register_t *retval) 2539 { 2540 struct sys_fchown_args /* { 2541 syscallarg(int) fd; 2542 syscallarg(uid_t) uid; 2543 syscallarg(gid_t) gid; 2544 } */ *uap = v; 2545 struct proc *p = l->l_proc; 2546 int error; 2547 struct file *fp; 2548 2549 /* getvnode() will use the descriptor for us */ 2550 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2551 return (error); 2552 2553 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2554 SCARG(uap, gid), l, 0); 2555 FILE_UNUSE(fp, l); 2556 return (error); 2557 } 2558 2559 /* 2560 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2561 */ 2562 /* ARGSUSED */ 2563 int 2564 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2565 { 2566 struct sys_fchown_args /* { 2567 syscallarg(int) fd; 2568 syscallarg(uid_t) uid; 2569 syscallarg(gid_t) gid; 2570 } */ *uap = v; 2571 struct proc *p = l->l_proc; 2572 int error; 2573 struct file *fp; 2574 2575 /* getvnode() will use the descriptor for us */ 2576 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2577 return (error); 2578 2579 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2580 SCARG(uap, gid), l, 1); 2581 FILE_UNUSE(fp, l); 2582 return (error); 2583 } 2584 2585 /* 2586 * Set ownership given a path name; this version does not follow links. 2587 */ 2588 /* ARGSUSED */ 2589 int 2590 sys_lchown(struct lwp *l, void *v, register_t *retval) 2591 { 2592 struct sys_lchown_args /* { 2593 syscallarg(const char *) path; 2594 syscallarg(uid_t) uid; 2595 syscallarg(gid_t) gid; 2596 } */ *uap = v; 2597 int error; 2598 struct nameidata nd; 2599 2600 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2601 if ((error = namei(&nd)) != 0) 2602 return (error); 2603 2604 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2605 2606 vrele(nd.ni_vp); 2607 return (error); 2608 } 2609 2610 /* 2611 * Set ownership given a path name; this version does not follow links. 2612 * Provides POSIX/XPG semantics. 2613 */ 2614 /* ARGSUSED */ 2615 int 2616 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2617 { 2618 struct sys_lchown_args /* { 2619 syscallarg(const char *) path; 2620 syscallarg(uid_t) uid; 2621 syscallarg(gid_t) gid; 2622 } */ *uap = v; 2623 int error; 2624 struct nameidata nd; 2625 2626 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2627 if ((error = namei(&nd)) != 0) 2628 return (error); 2629 2630 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2631 2632 vrele(nd.ni_vp); 2633 return (error); 2634 } 2635 2636 /* 2637 * Common routine to set ownership given a vnode. 2638 */ 2639 static int 2640 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2641 int posix_semantics) 2642 { 2643 struct proc *p = l->l_proc; 2644 struct mount *mp; 2645 struct vattr vattr; 2646 mode_t newmode; 2647 int error; 2648 2649 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2650 return (error); 2651 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2652 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2653 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 2654 goto out; 2655 2656 #define CHANGED(x) ((int)(x) != -1) 2657 newmode = vattr.va_mode; 2658 if (posix_semantics) { 2659 /* 2660 * POSIX/XPG semantics: if the caller is not the super-user, 2661 * clear set-user-id and set-group-id bits. Both POSIX and 2662 * the XPG consider the behaviour for calls by the super-user 2663 * implementation-defined; we leave the set-user-id and set- 2664 * group-id settings intact in that case. 2665 */ 2666 if (suser(p->p_ucred, NULL) != 0) 2667 newmode &= ~(S_ISUID | S_ISGID); 2668 } else { 2669 /* 2670 * NetBSD semantics: when changing owner and/or group, 2671 * clear the respective bit(s). 2672 */ 2673 if (CHANGED(uid)) 2674 newmode &= ~S_ISUID; 2675 if (CHANGED(gid)) 2676 newmode &= ~S_ISGID; 2677 } 2678 /* Update va_mode iff altered. */ 2679 if (vattr.va_mode == newmode) 2680 newmode = VNOVAL; 2681 2682 VATTR_NULL(&vattr); 2683 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2684 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2685 vattr.va_mode = newmode; 2686 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2687 #undef CHANGED 2688 2689 out: 2690 VOP_UNLOCK(vp, 0); 2691 vn_finished_write(mp, 0); 2692 return (error); 2693 } 2694 2695 /* 2696 * Set the access and modification times given a path name; this 2697 * version follows links. 2698 */ 2699 /* ARGSUSED */ 2700 int 2701 sys_utimes(struct lwp *l, void *v, register_t *retval) 2702 { 2703 struct sys_utimes_args /* { 2704 syscallarg(const char *) path; 2705 syscallarg(const struct timeval *) tptr; 2706 } */ *uap = v; 2707 int error; 2708 struct nameidata nd; 2709 2710 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2711 if ((error = namei(&nd)) != 0) 2712 return (error); 2713 2714 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2715 2716 vrele(nd.ni_vp); 2717 return (error); 2718 } 2719 2720 /* 2721 * Set the access and modification times given a file descriptor. 2722 */ 2723 /* ARGSUSED */ 2724 int 2725 sys_futimes(struct lwp *l, void *v, register_t *retval) 2726 { 2727 struct sys_futimes_args /* { 2728 syscallarg(int) fd; 2729 syscallarg(const struct timeval *) tptr; 2730 } */ *uap = v; 2731 struct proc *p = l->l_proc; 2732 int error; 2733 struct file *fp; 2734 2735 /* getvnode() will use the descriptor for us */ 2736 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2737 return (error); 2738 2739 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2740 FILE_UNUSE(fp, l); 2741 return (error); 2742 } 2743 2744 /* 2745 * Set the access and modification times given a path name; this 2746 * version does not follow links. 2747 */ 2748 /* ARGSUSED */ 2749 int 2750 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2751 { 2752 struct sys_lutimes_args /* { 2753 syscallarg(const char *) path; 2754 syscallarg(const struct timeval *) tptr; 2755 } */ *uap = v; 2756 int error; 2757 struct nameidata nd; 2758 2759 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2760 if ((error = namei(&nd)) != 0) 2761 return (error); 2762 2763 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2764 2765 vrele(nd.ni_vp); 2766 return (error); 2767 } 2768 2769 /* 2770 * Common routine to set access and modification times given a vnode. 2771 */ 2772 static int 2773 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2774 { 2775 struct proc *p = l->l_proc; 2776 struct mount *mp; 2777 struct vattr vattr; 2778 int error; 2779 2780 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2781 return (error); 2782 VATTR_NULL(&vattr); 2783 if (tptr == NULL) { 2784 nanotime(&vattr.va_atime); 2785 vattr.va_mtime = vattr.va_atime; 2786 vattr.va_vaflags |= VA_UTIMES_NULL; 2787 } else { 2788 struct timeval tv[2]; 2789 2790 error = copyin(tptr, tv, sizeof(tv)); 2791 if (error) 2792 goto out; 2793 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 2794 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 2795 } 2796 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2797 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2798 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2799 VOP_UNLOCK(vp, 0); 2800 out: 2801 vn_finished_write(mp, 0); 2802 return (error); 2803 } 2804 2805 /* 2806 * Truncate a file given its path name. 2807 */ 2808 /* ARGSUSED */ 2809 int 2810 sys_truncate(struct lwp *l, void *v, register_t *retval) 2811 { 2812 struct sys_truncate_args /* { 2813 syscallarg(const char *) path; 2814 syscallarg(int) pad; 2815 syscallarg(off_t) length; 2816 } */ *uap = v; 2817 struct proc *p = l->l_proc; 2818 struct vnode *vp; 2819 struct mount *mp; 2820 struct vattr vattr; 2821 int error; 2822 struct nameidata nd; 2823 2824 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2825 if ((error = namei(&nd)) != 0) 2826 return (error); 2827 vp = nd.ni_vp; 2828 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2829 vrele(vp); 2830 return (error); 2831 } 2832 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2833 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2834 if (vp->v_type == VDIR) 2835 error = EISDIR; 2836 else if ((error = vn_writechk(vp)) == 0 && 2837 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, l)) == 0) { 2838 VATTR_NULL(&vattr); 2839 vattr.va_size = SCARG(uap, length); 2840 error = VOP_SETATTR(vp, &vattr, p->p_ucred, l); 2841 } 2842 vput(vp); 2843 vn_finished_write(mp, 0); 2844 return (error); 2845 } 2846 2847 /* 2848 * Truncate a file given a file descriptor. 2849 */ 2850 /* ARGSUSED */ 2851 int 2852 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2853 { 2854 struct sys_ftruncate_args /* { 2855 syscallarg(int) fd; 2856 syscallarg(int) pad; 2857 syscallarg(off_t) length; 2858 } */ *uap = v; 2859 struct proc *p = l->l_proc; 2860 struct mount *mp; 2861 struct vattr vattr; 2862 struct vnode *vp; 2863 struct file *fp; 2864 int error; 2865 2866 /* getvnode() will use the descriptor for us */ 2867 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2868 return (error); 2869 if ((fp->f_flag & FWRITE) == 0) { 2870 error = EINVAL; 2871 goto out; 2872 } 2873 vp = (struct vnode *)fp->f_data; 2874 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2875 FILE_UNUSE(fp, l); 2876 return (error); 2877 } 2878 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 2879 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2880 if (vp->v_type == VDIR) 2881 error = EISDIR; 2882 else if ((error = vn_writechk(vp)) == 0) { 2883 VATTR_NULL(&vattr); 2884 vattr.va_size = SCARG(uap, length); 2885 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 2886 } 2887 VOP_UNLOCK(vp, 0); 2888 vn_finished_write(mp, 0); 2889 out: 2890 FILE_UNUSE(fp, l); 2891 return (error); 2892 } 2893 2894 /* 2895 * Sync an open file. 2896 */ 2897 /* ARGSUSED */ 2898 int 2899 sys_fsync(struct lwp *l, void *v, register_t *retval) 2900 { 2901 struct sys_fsync_args /* { 2902 syscallarg(int) fd; 2903 } */ *uap = v; 2904 struct proc *p = l->l_proc; 2905 struct vnode *vp; 2906 struct mount *mp; 2907 struct file *fp; 2908 int error; 2909 2910 /* getvnode() will use the descriptor for us */ 2911 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2912 return (error); 2913 vp = (struct vnode *)fp->f_data; 2914 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2915 FILE_UNUSE(fp, l); 2916 return (error); 2917 } 2918 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2919 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 2920 if (error == 0 && bioops.io_fsync != NULL && 2921 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2922 (*bioops.io_fsync)(vp, 0); 2923 VOP_UNLOCK(vp, 0); 2924 vn_finished_write(mp, 0); 2925 FILE_UNUSE(fp, l); 2926 return (error); 2927 } 2928 2929 /* 2930 * Sync a range of file data. API modeled after that found in AIX. 2931 * 2932 * FDATASYNC indicates that we need only save enough metadata to be able 2933 * to re-read the written data. Note we duplicate AIX's requirement that 2934 * the file be open for writing. 2935 */ 2936 /* ARGSUSED */ 2937 int 2938 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2939 { 2940 struct sys_fsync_range_args /* { 2941 syscallarg(int) fd; 2942 syscallarg(int) flags; 2943 syscallarg(off_t) start; 2944 syscallarg(off_t) length; 2945 } */ *uap = v; 2946 struct proc *p = l->l_proc; 2947 struct vnode *vp; 2948 struct file *fp; 2949 int flags, nflags; 2950 off_t s, e, len; 2951 int error; 2952 2953 /* getvnode() will use the descriptor for us */ 2954 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2955 return (error); 2956 2957 if ((fp->f_flag & FWRITE) == 0) { 2958 FILE_UNUSE(fp, l); 2959 return (EBADF); 2960 } 2961 2962 flags = SCARG(uap, flags); 2963 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2964 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2965 return (EINVAL); 2966 } 2967 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2968 if (flags & FDATASYNC) 2969 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2970 else 2971 nflags = FSYNC_WAIT; 2972 if (flags & FDISKSYNC) 2973 nflags |= FSYNC_CACHE; 2974 2975 len = SCARG(uap, length); 2976 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2977 if (len) { 2978 s = SCARG(uap, start); 2979 e = s + len; 2980 if (e < s) { 2981 FILE_UNUSE(fp, l); 2982 return (EINVAL); 2983 } 2984 } else { 2985 e = 0; 2986 s = 0; 2987 } 2988 2989 vp = (struct vnode *)fp->f_data; 2990 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2991 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 2992 2993 if (error == 0 && bioops.io_fsync != NULL && 2994 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2995 (*bioops.io_fsync)(vp, nflags); 2996 2997 VOP_UNLOCK(vp, 0); 2998 FILE_UNUSE(fp, l); 2999 return (error); 3000 } 3001 3002 /* 3003 * Sync the data of an open file. 3004 */ 3005 /* ARGSUSED */ 3006 int 3007 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3008 { 3009 struct sys_fdatasync_args /* { 3010 syscallarg(int) fd; 3011 } */ *uap = v; 3012 struct proc *p = l->l_proc; 3013 struct vnode *vp; 3014 struct file *fp; 3015 int error; 3016 3017 /* getvnode() will use the descriptor for us */ 3018 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3019 return (error); 3020 if ((fp->f_flag & FWRITE) == 0) { 3021 FILE_UNUSE(fp, l); 3022 return (EBADF); 3023 } 3024 vp = (struct vnode *)fp->f_data; 3025 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3026 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3027 VOP_UNLOCK(vp, 0); 3028 FILE_UNUSE(fp, l); 3029 return (error); 3030 } 3031 3032 /* 3033 * Rename files, (standard) BSD semantics frontend. 3034 */ 3035 /* ARGSUSED */ 3036 int 3037 sys_rename(struct lwp *l, void *v, register_t *retval) 3038 { 3039 struct sys_rename_args /* { 3040 syscallarg(const char *) from; 3041 syscallarg(const char *) to; 3042 } */ *uap = v; 3043 3044 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3045 } 3046 3047 /* 3048 * Rename files, POSIX semantics frontend. 3049 */ 3050 /* ARGSUSED */ 3051 int 3052 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3053 { 3054 struct sys___posix_rename_args /* { 3055 syscallarg(const char *) from; 3056 syscallarg(const char *) to; 3057 } */ *uap = v; 3058 3059 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 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 lwp *l, int retain) 3074 { 3075 struct mount *mp = NULL; 3076 struct vnode *tvp, *fvp, *tdvp; 3077 struct nameidata fromnd, tond; 3078 struct proc *p; 3079 int error; 3080 3081 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3082 from, l); 3083 if ((error = namei(&fromnd)) != 0) 3084 return (error); 3085 fvp = fromnd.ni_vp; 3086 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3087 if (error != 0) { 3088 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3089 vrele(fromnd.ni_dvp); 3090 vrele(fvp); 3091 if (fromnd.ni_startdir) 3092 vrele(fromnd.ni_startdir); 3093 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3094 return (error); 3095 } 3096 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3097 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3098 if ((error = namei(&tond)) != 0) { 3099 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3100 vrele(fromnd.ni_dvp); 3101 vrele(fvp); 3102 goto out1; 3103 } 3104 tdvp = tond.ni_dvp; 3105 tvp = tond.ni_vp; 3106 3107 if (tvp != NULL) { 3108 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3109 error = ENOTDIR; 3110 goto out; 3111 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3112 error = EISDIR; 3113 goto out; 3114 } 3115 } 3116 3117 if (fvp == tdvp) 3118 error = EINVAL; 3119 3120 /* 3121 * Source and destination refer to the same object. 3122 */ 3123 if (fvp == tvp) { 3124 if (retain) 3125 error = -1; 3126 else if (fromnd.ni_dvp == tdvp && 3127 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3128 !memcmp(fromnd.ni_cnd.cn_nameptr, 3129 tond.ni_cnd.cn_nameptr, 3130 fromnd.ni_cnd.cn_namelen)) 3131 error = -1; 3132 } 3133 3134 #ifdef VERIFIED_EXEC 3135 if (!error) 3136 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l); 3137 #endif /* VERIFIED_EXEC */ 3138 3139 out: 3140 p = l->l_proc; 3141 if (!error) { 3142 VOP_LEASE(tdvp, l, p->p_ucred, LEASE_WRITE); 3143 if (fromnd.ni_dvp != tdvp) 3144 VOP_LEASE(fromnd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3145 if (tvp) { 3146 VOP_LEASE(tvp, l, p->p_ucred, LEASE_WRITE); 3147 } 3148 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3149 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3150 } else { 3151 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3152 if (tdvp == tvp) 3153 vrele(tdvp); 3154 else 3155 vput(tdvp); 3156 if (tvp) 3157 vput(tvp); 3158 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3159 vrele(fromnd.ni_dvp); 3160 vrele(fvp); 3161 } 3162 vrele(tond.ni_startdir); 3163 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3164 out1: 3165 vn_finished_write(mp, 0); 3166 if (fromnd.ni_startdir) 3167 vrele(fromnd.ni_startdir); 3168 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3169 return (error == -1 ? 0 : error); 3170 } 3171 3172 /* 3173 * Make a directory file. 3174 */ 3175 /* ARGSUSED */ 3176 int 3177 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3178 { 3179 struct sys_mkdir_args /* { 3180 syscallarg(const char *) path; 3181 syscallarg(int) mode; 3182 } */ *uap = v; 3183 struct proc *p = l->l_proc; 3184 struct mount *mp; 3185 struct vnode *vp; 3186 struct vattr vattr; 3187 int error; 3188 struct nameidata nd; 3189 3190 restart: 3191 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3192 SCARG(uap, path), l); 3193 if ((error = namei(&nd)) != 0) 3194 return (error); 3195 vp = nd.ni_vp; 3196 if (vp != NULL) { 3197 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3198 if (nd.ni_dvp == vp) 3199 vrele(nd.ni_dvp); 3200 else 3201 vput(nd.ni_dvp); 3202 vrele(vp); 3203 return (EEXIST); 3204 } 3205 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3206 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3207 if (nd.ni_dvp == vp) 3208 vrele(nd.ni_dvp); 3209 else 3210 vput(nd.ni_dvp); 3211 if ((error = vn_start_write(NULL, &mp, 3212 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3213 return (error); 3214 goto restart; 3215 } 3216 VATTR_NULL(&vattr); 3217 vattr.va_type = VDIR; 3218 vattr.va_mode = 3219 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3220 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3221 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3222 if (!error) 3223 vput(nd.ni_vp); 3224 vn_finished_write(mp, 0); 3225 return (error); 3226 } 3227 3228 /* 3229 * Remove a directory file. 3230 */ 3231 /* ARGSUSED */ 3232 int 3233 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3234 { 3235 struct sys_rmdir_args /* { 3236 syscallarg(const char *) path; 3237 } */ *uap = v; 3238 struct proc *p = l->l_proc; 3239 struct mount *mp; 3240 struct vnode *vp; 3241 int error; 3242 struct nameidata nd; 3243 3244 restart: 3245 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3246 SCARG(uap, path), l); 3247 if ((error = namei(&nd)) != 0) 3248 return (error); 3249 vp = nd.ni_vp; 3250 if (vp->v_type != VDIR) { 3251 error = ENOTDIR; 3252 goto out; 3253 } 3254 /* 3255 * No rmdir "." please. 3256 */ 3257 if (nd.ni_dvp == vp) { 3258 error = EINVAL; 3259 goto out; 3260 } 3261 /* 3262 * The root of a mounted filesystem cannot be deleted. 3263 */ 3264 if (vp->v_flag & VROOT) { 3265 error = EBUSY; 3266 goto out; 3267 } 3268 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3269 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3270 if (nd.ni_dvp == vp) 3271 vrele(nd.ni_dvp); 3272 else 3273 vput(nd.ni_dvp); 3274 vput(vp); 3275 if ((error = vn_start_write(NULL, &mp, 3276 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3277 return (error); 3278 goto restart; 3279 } 3280 VOP_LEASE(nd.ni_dvp, l, p->p_ucred, LEASE_WRITE); 3281 VOP_LEASE(vp, l, p->p_ucred, LEASE_WRITE); 3282 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3283 vn_finished_write(mp, 0); 3284 return (error); 3285 3286 out: 3287 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3288 if (nd.ni_dvp == vp) 3289 vrele(nd.ni_dvp); 3290 else 3291 vput(nd.ni_dvp); 3292 vput(vp); 3293 return (error); 3294 } 3295 3296 /* 3297 * Read a block of directory entries in a file system independent format. 3298 */ 3299 int 3300 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3301 { 3302 struct sys___getdents30_args /* { 3303 syscallarg(int) fd; 3304 syscallarg(char *) buf; 3305 syscallarg(size_t) count; 3306 } */ *uap = v; 3307 struct proc *p = l->l_proc; 3308 struct file *fp; 3309 int error, done; 3310 3311 /* getvnode() will use the descriptor for us */ 3312 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3313 return (error); 3314 if ((fp->f_flag & FREAD) == 0) { 3315 error = EBADF; 3316 goto out; 3317 } 3318 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3319 SCARG(uap, count), &done, l, 0, 0); 3320 #ifdef KTRACE 3321 if (!error && KTRPOINT(p, KTR_GENIO)) { 3322 struct iovec iov; 3323 iov.iov_base = SCARG(uap, buf); 3324 iov.iov_len = done; 3325 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3326 } 3327 #endif 3328 *retval = done; 3329 out: 3330 FILE_UNUSE(fp, l); 3331 return (error); 3332 } 3333 3334 /* 3335 * Set the mode mask for creation of filesystem nodes. 3336 */ 3337 int 3338 sys_umask(struct lwp *l, void *v, register_t *retval) 3339 { 3340 struct sys_umask_args /* { 3341 syscallarg(mode_t) newmask; 3342 } */ *uap = v; 3343 struct proc *p = l->l_proc; 3344 struct cwdinfo *cwdi; 3345 3346 cwdi = p->p_cwdi; 3347 *retval = cwdi->cwdi_cmask; 3348 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3349 return (0); 3350 } 3351 3352 /* 3353 * Void all references to file by ripping underlying filesystem 3354 * away from vnode. 3355 */ 3356 /* ARGSUSED */ 3357 int 3358 sys_revoke(struct lwp *l, void *v, register_t *retval) 3359 { 3360 struct sys_revoke_args /* { 3361 syscallarg(const char *) path; 3362 } */ *uap = v; 3363 struct proc *p = l->l_proc; 3364 struct mount *mp; 3365 struct vnode *vp; 3366 struct vattr vattr; 3367 int error; 3368 struct nameidata nd; 3369 3370 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3371 if ((error = namei(&nd)) != 0) 3372 return (error); 3373 vp = nd.ni_vp; 3374 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, l)) != 0) 3375 goto out; 3376 if (p->p_ucred->cr_uid != vattr.va_uid && 3377 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3378 goto out; 3379 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3380 goto out; 3381 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3382 VOP_REVOKE(vp, REVOKEALL); 3383 vn_finished_write(mp, 0); 3384 out: 3385 vrele(vp); 3386 return (error); 3387 } 3388 3389 /* 3390 * Convert a user file descriptor to a kernel file entry. 3391 */ 3392 int 3393 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3394 { 3395 struct vnode *vp; 3396 struct file *fp; 3397 3398 if ((fp = fd_getfile(fdp, fd)) == NULL) 3399 return (EBADF); 3400 3401 FILE_USE(fp); 3402 3403 if (fp->f_type != DTYPE_VNODE) { 3404 FILE_UNUSE(fp, NULL); 3405 return (EINVAL); 3406 } 3407 3408 vp = (struct vnode *)fp->f_data; 3409 if (vp->v_type == VBAD) { 3410 FILE_UNUSE(fp, NULL); 3411 return (EBADF); 3412 } 3413 3414 *fpp = fp; 3415 return (0); 3416 } 3417
Indexes created Sat Sep 27 14:09:57 GMT 2025