vfs_syscalls.c revision 1.231
1 /* $NetBSD: vfs_syscalls.c,v 1.231 2005/09/23 12:10:33 jmmv 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 1.231 2005/09/23 12:10:33 jmmv 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 #if NFSS > 0 75 #include <dev/fssvar.h> 76 #endif 77 78 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 79 80 static int change_dir(struct nameidata *, struct proc *); 81 static int change_flags(struct vnode *, u_long, struct proc *); 82 static int change_mode(struct vnode *, int, struct proc *p); 83 static int change_owner(struct vnode *, uid_t, gid_t, struct proc *, int); 84 static int change_utimes(struct vnode *vp, const struct timeval *, 85 struct proc *p); 86 static int rename_files(const char *, const char *, struct proc *, int); 87 88 void checkdirs(struct vnode *); 89 90 int dovfsusermount = 0; 91 92 /* 93 * Virtual File System System Calls 94 */ 95 96 /* 97 * Mount a file system. 98 */ 99 100 #if defined(COMPAT_09) || defined(COMPAT_43) 101 /* 102 * This table is used to maintain compatibility with 4.3BSD 103 * and NetBSD 0.9 mount syscalls. Note, the order is important! 104 * 105 * Do not modify this table. It should only contain filesystems 106 * supported by NetBSD 0.9 and 4.3BSD. 107 */ 108 const char * const mountcompatnames[] = { 109 NULL, /* 0 = MOUNT_NONE */ 110 MOUNT_FFS, /* 1 = MOUNT_UFS */ 111 MOUNT_NFS, /* 2 */ 112 MOUNT_MFS, /* 3 */ 113 MOUNT_MSDOS, /* 4 */ 114 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 115 MOUNT_FDESC, /* 6 */ 116 MOUNT_KERNFS, /* 7 */ 117 NULL, /* 8 = MOUNT_DEVFS */ 118 MOUNT_AFS, /* 9 */ 119 }; 120 const int nmountcompatnames = sizeof(mountcompatnames) / 121 sizeof(mountcompatnames[0]); 122 #endif /* COMPAT_09 || COMPAT_43 */ 123 124 /* ARGSUSED */ 125 int 126 sys_mount(struct lwp *l, void *v, register_t *retval) 127 { 128 struct sys_mount_args /* { 129 syscallarg(const char *) type; 130 syscallarg(const char *) path; 131 syscallarg(int) flags; 132 syscallarg(void *) data; 133 } */ *uap = v; 134 struct proc *p = l->l_proc; 135 struct vnode *vp; 136 struct mount *mp; 137 int error, flag = 0; 138 char fstypename[MFSNAMELEN]; 139 struct vattr va; 140 struct nameidata nd; 141 struct vfsops *vfs; 142 143 /* 144 * if MNT_GETARGS is specified, it should be only flag. 145 */ 146 147 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 148 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 149 return EINVAL; 150 } 151 152 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 153 (error = suser(p->p_ucred, &p->p_acflag))) 154 return (error); 155 /* 156 * Get vnode to be covered 157 */ 158 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 159 SCARG(uap, path), p); 160 if ((error = namei(&nd)) != 0) 161 return (error); 162 vp = nd.ni_vp; 163 /* 164 * A lookup in VFS_MOUNT might result in an attempt to 165 * lock this vnode again, so make the lock recursive. 166 */ 167 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 168 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 169 if ((vp->v_flag & VROOT) == 0) { 170 vput(vp); 171 return (EINVAL); 172 } 173 mp = vp->v_mount; 174 flag = mp->mnt_flag; 175 vfs = mp->mnt_op; 176 /* 177 * We only allow the filesystem to be reloaded if it 178 * is currently mounted read-only. 179 */ 180 if ((SCARG(uap, flags) & MNT_RELOAD) && 181 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 182 vput(vp); 183 return (EOPNOTSUPP); /* Needs translation */ 184 } 185 /* 186 * In "highly secure" mode, don't let the caller do anything 187 * but downgrade a filesystem from read-write to read-only. 188 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 189 */ 190 if (securelevel >= 2 && 191 SCARG(uap, flags) != MNT_GETARGS && 192 SCARG(uap, flags) != 193 (mp->mnt_flag | MNT_RDONLY | 194 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 195 vput(vp); 196 return (EPERM); 197 } 198 mp->mnt_flag |= SCARG(uap, flags) & 199 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 200 /* 201 * Only root, or the user that did the original mount is 202 * permitted to update it. 203 */ 204 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 205 mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 206 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 207 vput(vp); 208 return (error); 209 } 210 /* 211 * Do not allow NFS export by non-root users. For non-root 212 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 213 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 214 */ 215 if (p->p_ucred->cr_uid != 0) { 216 if (SCARG(uap, flags) & MNT_EXPORTED) { 217 vput(vp); 218 return (EPERM); 219 } 220 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 221 if (flag & MNT_NOEXEC) 222 SCARG(uap, flags) |= MNT_NOEXEC; 223 } 224 if (vfs_busy(mp, LK_NOWAIT, 0)) { 225 vput(vp); 226 return (EPERM); 227 } 228 goto update; 229 } else { 230 if (securelevel >= 2) { 231 vput(vp); 232 return (EPERM); 233 } 234 } 235 /* 236 * If the user is not root, ensure that they own the directory 237 * onto which we are attempting to mount. 238 */ 239 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 || 240 (va.va_uid != p->p_ucred->cr_uid && 241 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) { 242 vput(vp); 243 return (error); 244 } 245 /* 246 * Do not allow NFS export by non-root users. For non-root users, 247 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 248 * mount point is already MNT_NOEXEC. 249 */ 250 if (p->p_ucred->cr_uid != 0) { 251 if (SCARG(uap, flags) & MNT_EXPORTED) { 252 vput(vp); 253 return (EPERM); 254 } 255 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 256 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 257 SCARG(uap, flags) |= MNT_NOEXEC; 258 } 259 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) { 260 vput(vp); 261 return (error); 262 } 263 if (vp->v_type != VDIR) { 264 vput(vp); 265 return (ENOTDIR); 266 } 267 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 268 if (error) { 269 #if defined(COMPAT_09) || defined(COMPAT_43) 270 /* 271 * Historically, filesystem types were identified by numbers. 272 * If we get an integer for the filesystem type instead of a 273 * string, we check to see if it matches one of the historic 274 * filesystem types. 275 */ 276 u_long fsindex = (u_long)SCARG(uap, type); 277 if (fsindex >= nmountcompatnames || 278 mountcompatnames[fsindex] == NULL) { 279 vput(vp); 280 return (ENODEV); 281 } 282 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 283 #else 284 vput(vp); 285 return (error); 286 #endif 287 } 288 #ifdef COMPAT_10 289 /* Accept `ufs' as an alias for `ffs'. */ 290 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 291 strncpy(fstypename, "ffs", MFSNAMELEN); 292 #endif 293 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 294 vput(vp); 295 return (ENODEV); 296 } 297 if (vp->v_mountedhere != NULL) { 298 vput(vp); 299 return (EBUSY); 300 } 301 302 /* 303 * Allocate and initialize the file system. 304 */ 305 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 306 M_MOUNT, M_WAITOK); 307 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 308 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 309 simple_lock_init(&mp->mnt_slock); 310 (void)vfs_busy(mp, LK_NOWAIT, 0); 311 mp->mnt_op = vfs; 312 vfs->vfs_refcount++; 313 mp->mnt_vnodecovered = vp; 314 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 315 mp->mnt_unmounter = NULL; 316 mp->mnt_leaf = mp; 317 318 /* 319 * The underlying file system may refuse the mount for 320 * various reasons. Allow the user to force it to happen. 321 */ 322 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 323 update: 324 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 325 /* 326 * Set the mount level flags. 327 */ 328 if (SCARG(uap, flags) & MNT_RDONLY) 329 mp->mnt_flag |= MNT_RDONLY; 330 else if (mp->mnt_flag & MNT_RDONLY) 331 mp->mnt_iflag |= IMNT_WANTRDWR; 332 mp->mnt_flag &= 333 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 334 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 335 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 336 MNT_MAGICLINKS); 337 mp->mnt_flag |= SCARG(uap, flags) & 338 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 339 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 340 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 341 MNT_IGNORE | MNT_MAGICLINKS); 342 } 343 /* 344 * Mount the filesystem. 345 */ 346 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 347 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 348 if (mp->mnt_iflag & IMNT_WANTRDWR) 349 mp->mnt_flag &= ~MNT_RDONLY; 350 if (error) 351 mp->mnt_flag = flag; 352 mp->mnt_flag &=~ 353 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 354 mp->mnt_iflag &=~ IMNT_WANTRDWR; 355 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 356 if (mp->mnt_syncer == NULL) 357 error = vfs_allocate_syncvnode(mp); 358 } else { 359 if (mp->mnt_syncer != NULL) 360 vfs_deallocate_syncvnode(mp); 361 } 362 vfs_unbusy(mp); 363 VOP_UNLOCK(vp, 0); 364 vrele(vp); 365 return (error); 366 } 367 /* 368 * Put the new filesystem on the mount list after root. 369 */ 370 cache_purge(vp); 371 if (!error) { 372 mp->mnt_flag &=~ 373 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 374 mp->mnt_iflag &=~ IMNT_WANTRDWR; 375 vp->v_mountedhere = mp; 376 simple_lock(&mountlist_slock); 377 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 378 simple_unlock(&mountlist_slock); 379 checkdirs(vp); 380 VOP_UNLOCK(vp, 0); 381 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 382 error = vfs_allocate_syncvnode(mp); 383 vfs_unbusy(mp); 384 (void) VFS_STATVFS(mp, &mp->mnt_stat, p); 385 if ((error = VFS_START(mp, 0, p))) 386 vrele(vp); 387 } else { 388 vp->v_mountedhere = (struct mount *)0; 389 vfs->vfs_refcount--; 390 vfs_unbusy(mp); 391 free(mp, M_MOUNT); 392 vput(vp); 393 } 394 return (error); 395 } 396 397 /* 398 * Scan all active processes to see if any of them have a current 399 * or root directory onto which the new filesystem has just been 400 * mounted. If so, replace them with the new mount point. 401 */ 402 void 403 checkdirs(struct vnode *olddp) 404 { 405 struct cwdinfo *cwdi; 406 struct vnode *newdp; 407 struct proc *p; 408 409 if (olddp->v_usecount == 1) 410 return; 411 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 412 panic("mount: lost mount"); 413 proclist_lock_read(); 414 PROCLIST_FOREACH(p, &allproc) { 415 cwdi = p->p_cwdi; 416 if (!cwdi) 417 continue; 418 if (cwdi->cwdi_cdir == olddp) { 419 vrele(cwdi->cwdi_cdir); 420 VREF(newdp); 421 cwdi->cwdi_cdir = newdp; 422 } 423 if (cwdi->cwdi_rdir == olddp) { 424 vrele(cwdi->cwdi_rdir); 425 VREF(newdp); 426 cwdi->cwdi_rdir = newdp; 427 } 428 } 429 proclist_unlock_read(); 430 if (rootvnode == olddp) { 431 vrele(rootvnode); 432 VREF(newdp); 433 rootvnode = newdp; 434 } 435 vput(newdp); 436 } 437 438 /* 439 * Unmount a file system. 440 * 441 * Note: unmount takes a path to the vnode mounted on as argument, 442 * not special file (as before). 443 */ 444 /* ARGSUSED */ 445 int 446 sys_unmount(struct lwp *l, void *v, register_t *retval) 447 { 448 struct sys_unmount_args /* { 449 syscallarg(const char *) path; 450 syscallarg(int) flags; 451 } */ *uap = v; 452 struct proc *p = l->l_proc; 453 struct vnode *vp; 454 struct mount *mp; 455 int error; 456 struct nameidata nd; 457 458 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 459 SCARG(uap, path), p); 460 if ((error = namei(&nd)) != 0) 461 return (error); 462 vp = nd.ni_vp; 463 mp = vp->v_mount; 464 465 /* 466 * Only root, or the user that did the original mount is 467 * permitted to unmount this filesystem. 468 */ 469 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 470 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 471 vput(vp); 472 return (error); 473 } 474 475 /* 476 * Don't allow unmounting the root file system. 477 */ 478 if (mp->mnt_flag & MNT_ROOTFS) { 479 vput(vp); 480 return (EINVAL); 481 } 482 483 /* 484 * Must be the root of the filesystem 485 */ 486 if ((vp->v_flag & VROOT) == 0) { 487 vput(vp); 488 return (EINVAL); 489 } 490 vput(vp); 491 492 /* 493 * XXX Freeze syncer. Must do this before locking the 494 * mount point. See dounmount() for details. 495 */ 496 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 497 498 if (vfs_busy(mp, 0, 0)) { 499 lockmgr(&syncer_lock, LK_RELEASE, NULL); 500 return (EBUSY); 501 } 502 503 return (dounmount(mp, SCARG(uap, flags), p)); 504 } 505 506 /* 507 * Do the actual file system unmount. File system is assumed to have been 508 * marked busy by the caller. 509 */ 510 int 511 dounmount(struct mount *mp, int flags, struct proc *p) 512 { 513 struct vnode *coveredvp; 514 int error; 515 int async; 516 int used_syncer; 517 518 simple_lock(&mountlist_slock); 519 vfs_unbusy(mp); 520 used_syncer = (mp->mnt_syncer != NULL); 521 522 /* 523 * XXX Syncer must be frozen when we get here. This should really 524 * be done on a per-mountpoint basis, but especially the softdep 525 * code possibly called from the syncer doesn't exactly work on a 526 * per-mountpoint basis, so the softdep code would become a maze 527 * of vfs_busy() calls. 528 * 529 * The caller of dounmount() must acquire syncer_lock because 530 * the syncer itself acquires locks in syncer_lock -> vfs_busy 531 * order, and we must preserve that order to avoid deadlock. 532 * 533 * So, if the file system did not use the syncer, now is 534 * the time to release the syncer_lock. 535 */ 536 if (used_syncer == 0) 537 lockmgr(&syncer_lock, LK_RELEASE, NULL); 538 539 mp->mnt_iflag |= IMNT_UNMOUNT; 540 mp->mnt_unmounter = p; 541 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 542 vn_start_write(NULL, &mp, V_WAIT); 543 544 async = mp->mnt_flag & MNT_ASYNC; 545 mp->mnt_flag &= ~MNT_ASYNC; 546 cache_purgevfs(mp); /* remove cache entries for this file sys */ 547 if (mp->mnt_syncer != NULL) 548 vfs_deallocate_syncvnode(mp); 549 error = 0; 550 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 551 #if NFSS > 0 552 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 553 #endif 554 if (error == 0) 555 error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p); 556 } 557 if (error == 0 || (flags & MNT_FORCE)) 558 error = VFS_UNMOUNT(mp, flags, p); 559 vn_finished_write(mp, 0); 560 simple_lock(&mountlist_slock); 561 if (error) { 562 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 563 (void) vfs_allocate_syncvnode(mp); 564 mp->mnt_iflag &= ~IMNT_UNMOUNT; 565 mp->mnt_unmounter = NULL; 566 mp->mnt_flag |= async; 567 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 568 &mountlist_slock); 569 if (used_syncer) 570 lockmgr(&syncer_lock, LK_RELEASE, NULL); 571 simple_lock(&mp->mnt_slock); 572 while (mp->mnt_wcnt > 0) { 573 wakeup(mp); 574 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 575 0, &mp->mnt_slock); 576 } 577 simple_unlock(&mp->mnt_slock); 578 return (error); 579 } 580 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 581 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 582 coveredvp->v_mountedhere = NULL; 583 vrele(coveredvp); 584 } 585 mp->mnt_op->vfs_refcount--; 586 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 587 panic("unmount: dangling vnode"); 588 mp->mnt_iflag |= IMNT_GONE; 589 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 590 if (used_syncer) 591 lockmgr(&syncer_lock, LK_RELEASE, NULL); 592 simple_lock(&mp->mnt_slock); 593 while (mp->mnt_wcnt > 0) { 594 wakeup(mp); 595 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 596 } 597 simple_unlock(&mp->mnt_slock); 598 vfs_hooks_unmount(mp); 599 free(mp, M_MOUNT); 600 return (0); 601 } 602 603 /* 604 * Sync each mounted filesystem. 605 */ 606 #ifdef DEBUG 607 int syncprt = 0; 608 struct ctldebug debug0 = { "syncprt", &syncprt }; 609 #endif 610 611 /* ARGSUSED */ 612 int 613 sys_sync(struct lwp *l, void *v, register_t *retval) 614 { 615 struct mount *mp, *nmp; 616 int asyncflag; 617 struct proc *p = l == NULL ? &proc0 : l->l_proc; 618 619 simple_lock(&mountlist_slock); 620 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 621 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 622 nmp = mp->mnt_list.cqe_prev; 623 continue; 624 } 625 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 626 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 627 asyncflag = mp->mnt_flag & MNT_ASYNC; 628 mp->mnt_flag &= ~MNT_ASYNC; 629 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p); 630 if (asyncflag) 631 mp->mnt_flag |= MNT_ASYNC; 632 vn_finished_write(mp, 0); 633 } 634 simple_lock(&mountlist_slock); 635 nmp = mp->mnt_list.cqe_prev; 636 vfs_unbusy(mp); 637 638 } 639 simple_unlock(&mountlist_slock); 640 #ifdef DEBUG 641 if (syncprt) 642 vfs_bufstats(); 643 #endif /* DEBUG */ 644 return (0); 645 } 646 647 /* 648 * Change filesystem quotas. 649 */ 650 /* ARGSUSED */ 651 int 652 sys_quotactl(struct lwp *l, void *v, register_t *retval) 653 { 654 struct sys_quotactl_args /* { 655 syscallarg(const char *) path; 656 syscallarg(int) cmd; 657 syscallarg(int) uid; 658 syscallarg(caddr_t) arg; 659 } */ *uap = v; 660 struct proc *p = l->l_proc; 661 struct mount *mp; 662 int error; 663 struct nameidata nd; 664 665 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 666 if ((error = namei(&nd)) != 0) 667 return (error); 668 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 669 vrele(nd.ni_vp); 670 if (error) 671 return (error); 672 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 673 SCARG(uap, arg), p); 674 vn_finished_write(mp, 0); 675 return (error); 676 } 677 678 int 679 dostatvfs(struct mount *mp, struct statvfs *sp, struct proc *p, int flags, 680 int root) 681 { 682 struct cwdinfo *cwdi = p->p_cwdi; 683 int error = 0; 684 685 /* 686 * If MNT_NOWAIT or MNT_LAZY is specified, do not 687 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 688 * overrides MNT_NOWAIT. 689 */ 690 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 691 (flags != MNT_WAIT && flags != 0)) { 692 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 693 goto done; 694 } 695 696 /* Get the filesystem stats now */ 697 memset(sp, 0, sizeof(*sp)); 698 if ((error = VFS_STATVFS(mp, sp, p)) != 0) { 699 return error; 700 } 701 702 if (cwdi->cwdi_rdir == NULL) 703 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 704 done: 705 if (cwdi->cwdi_rdir != NULL) { 706 size_t len; 707 char *bp; 708 char *path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 709 if (!path) 710 return ENOMEM; 711 712 bp = path + MAXPATHLEN; 713 *--bp = '\0'; 714 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 715 MAXPATHLEN / 2, 0, p); 716 if (error) { 717 free(path, M_TEMP); 718 return error; 719 } 720 len = strlen(bp); 721 /* 722 * for mount points that are below our root, we can see 723 * them, so we fix up the pathname and return them. The 724 * rest we cannot see, so we don't allow viewing the 725 * data. 726 */ 727 if (strncmp(bp, sp->f_mntonname, len) == 0) { 728 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 729 sizeof(sp->f_mntonname)); 730 if (sp->f_mntonname[0] == '\0') 731 (void)strlcpy(sp->f_mntonname, "/", 732 sizeof(sp->f_mntonname)); 733 } else { 734 if (root) 735 (void)strlcpy(sp->f_mntonname, "/", 736 sizeof(sp->f_mntonname)); 737 else 738 error = EPERM; 739 } 740 free(path, M_TEMP); 741 } 742 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 743 return error; 744 } 745 746 /* 747 * Get filesystem statistics. 748 */ 749 /* ARGSUSED */ 750 int 751 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 752 { 753 struct sys_statvfs1_args /* { 754 syscallarg(const char *) path; 755 syscallarg(struct statvfs *) buf; 756 syscallarg(int) flags; 757 } */ *uap = v; 758 struct proc *p = l->l_proc; 759 struct mount *mp; 760 struct statvfs sbuf; 761 int error; 762 struct nameidata nd; 763 764 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 765 if ((error = namei(&nd)) != 0) 766 return error; 767 mp = nd.ni_vp->v_mount; 768 vrele(nd.ni_vp); 769 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 770 return error; 771 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 772 } 773 774 /* 775 * Get filesystem statistics. 776 */ 777 /* ARGSUSED */ 778 int 779 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 780 { 781 struct sys_fstatvfs1_args /* { 782 syscallarg(int) fd; 783 syscallarg(struct statvfs *) buf; 784 syscallarg(int) flags; 785 } */ *uap = v; 786 struct proc *p = l->l_proc; 787 struct file *fp; 788 struct mount *mp; 789 struct statvfs sbuf; 790 int error; 791 792 /* getvnode() will use the descriptor for us */ 793 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 794 return (error); 795 mp = ((struct vnode *)fp->f_data)->v_mount; 796 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) 797 goto out; 798 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 799 out: 800 FILE_UNUSE(fp, p); 801 return error; 802 } 803 804 805 /* 806 * Get statistics on all filesystems. 807 */ 808 int 809 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 810 { 811 struct sys_getvfsstat_args /* { 812 syscallarg(struct statvfs *) buf; 813 syscallarg(size_t) bufsize; 814 syscallarg(int) flags; 815 } */ *uap = v; 816 int root = 0; 817 struct proc *p = l->l_proc; 818 struct mount *mp, *nmp; 819 struct statvfs sbuf; 820 struct statvfs *sfsp; 821 size_t count, maxcount; 822 int error = 0; 823 824 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 825 sfsp = SCARG(uap, buf); 826 simple_lock(&mountlist_slock); 827 count = 0; 828 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 829 mp = nmp) { 830 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 831 nmp = CIRCLEQ_NEXT(mp, mnt_list); 832 continue; 833 } 834 if (sfsp && count < maxcount) { 835 error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 0); 836 if (error) { 837 simple_lock(&mountlist_slock); 838 nmp = CIRCLEQ_NEXT(mp, mnt_list); 839 vfs_unbusy(mp); 840 continue; 841 } 842 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 843 if (error) { 844 vfs_unbusy(mp); 845 return (error); 846 } 847 sfsp++; 848 root |= strcmp(sbuf.f_mntonname, "/") == 0; 849 } 850 count++; 851 simple_lock(&mountlist_slock); 852 nmp = CIRCLEQ_NEXT(mp, mnt_list); 853 vfs_unbusy(mp); 854 } 855 simple_unlock(&mountlist_slock); 856 if (root == 0 && p->p_cwdi->cwdi_rdir) { 857 /* 858 * fake a root entry 859 */ 860 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, p, 861 SCARG(uap, flags), 1)) != 0) 862 return error; 863 if (sfsp) 864 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 865 count++; 866 } 867 if (sfsp && count > maxcount) 868 *retval = maxcount; 869 else 870 *retval = count; 871 return error; 872 } 873 874 /* 875 * Change current working directory to a given file descriptor. 876 */ 877 /* ARGSUSED */ 878 int 879 sys_fchdir(struct lwp *l, void *v, register_t *retval) 880 { 881 struct sys_fchdir_args /* { 882 syscallarg(int) fd; 883 } */ *uap = v; 884 struct proc *p = l->l_proc; 885 struct filedesc *fdp = p->p_fd; 886 struct cwdinfo *cwdi = p->p_cwdi; 887 struct vnode *vp, *tdp; 888 struct mount *mp; 889 struct file *fp; 890 int error; 891 892 /* getvnode() will use the descriptor for us */ 893 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 894 return (error); 895 vp = (struct vnode *)fp->f_data; 896 897 VREF(vp); 898 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 899 if (vp->v_type != VDIR) 900 error = ENOTDIR; 901 else 902 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 903 while (!error && (mp = vp->v_mountedhere) != NULL) { 904 if (vfs_busy(mp, 0, 0)) 905 continue; 906 error = VFS_ROOT(mp, &tdp); 907 vfs_unbusy(mp); 908 if (error) 909 break; 910 vput(vp); 911 vp = tdp; 912 } 913 if (error) { 914 vput(vp); 915 goto out; 916 } 917 VOP_UNLOCK(vp, 0); 918 919 /* 920 * Disallow changing to a directory not under the process's 921 * current root directory (if there is one). 922 */ 923 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) { 924 vrele(vp); 925 error = EPERM; /* operation not permitted */ 926 goto out; 927 } 928 929 vrele(cwdi->cwdi_cdir); 930 cwdi->cwdi_cdir = vp; 931 out: 932 FILE_UNUSE(fp, p); 933 return (error); 934 } 935 936 /* 937 * Change this process's notion of the root directory to a given file 938 * descriptor. 939 */ 940 int 941 sys_fchroot(struct lwp *l, void *v, register_t *retval) 942 { 943 struct sys_fchroot_args *uap = v; 944 struct proc *p = l->l_proc; 945 struct filedesc *fdp = p->p_fd; 946 struct cwdinfo *cwdi = p->p_cwdi; 947 struct vnode *vp; 948 struct file *fp; 949 int error; 950 951 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 952 return error; 953 /* getvnode() will use the descriptor for us */ 954 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 955 return error; 956 vp = (struct vnode *) fp->f_data; 957 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 958 if (vp->v_type != VDIR) 959 error = ENOTDIR; 960 else 961 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 962 VOP_UNLOCK(vp, 0); 963 if (error) 964 goto out; 965 VREF(vp); 966 967 /* 968 * Prevent escaping from chroot by putting the root under 969 * the working directory. Silently chdir to / if we aren't 970 * already there. 971 */ 972 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 973 /* 974 * XXX would be more failsafe to change directory to a 975 * deadfs node here instead 976 */ 977 vrele(cwdi->cwdi_cdir); 978 VREF(vp); 979 cwdi->cwdi_cdir = vp; 980 } 981 982 if (cwdi->cwdi_rdir != NULL) 983 vrele(cwdi->cwdi_rdir); 984 cwdi->cwdi_rdir = vp; 985 out: 986 FILE_UNUSE(fp, p); 987 return (error); 988 } 989 990 /* 991 * Change current working directory (``.''). 992 */ 993 /* ARGSUSED */ 994 int 995 sys_chdir(struct lwp *l, void *v, register_t *retval) 996 { 997 struct sys_chdir_args /* { 998 syscallarg(const char *) path; 999 } */ *uap = v; 1000 struct proc *p = l->l_proc; 1001 struct cwdinfo *cwdi = p->p_cwdi; 1002 int error; 1003 struct nameidata nd; 1004 1005 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1006 SCARG(uap, path), p); 1007 if ((error = change_dir(&nd, p)) != 0) 1008 return (error); 1009 vrele(cwdi->cwdi_cdir); 1010 cwdi->cwdi_cdir = nd.ni_vp; 1011 return (0); 1012 } 1013 1014 /* 1015 * Change notion of root (``/'') directory. 1016 */ 1017 /* ARGSUSED */ 1018 int 1019 sys_chroot(struct lwp *l, void *v, register_t *retval) 1020 { 1021 struct sys_chroot_args /* { 1022 syscallarg(const char *) path; 1023 } */ *uap = v; 1024 struct proc *p = l->l_proc; 1025 struct cwdinfo *cwdi = p->p_cwdi; 1026 struct vnode *vp; 1027 int error; 1028 struct nameidata nd; 1029 1030 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1031 return (error); 1032 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1033 SCARG(uap, path), p); 1034 if ((error = change_dir(&nd, p)) != 0) 1035 return (error); 1036 if (cwdi->cwdi_rdir != NULL) 1037 vrele(cwdi->cwdi_rdir); 1038 vp = nd.ni_vp; 1039 cwdi->cwdi_rdir = vp; 1040 1041 /* 1042 * Prevent escaping from chroot by putting the root under 1043 * the working directory. Silently chdir to / if we aren't 1044 * already there. 1045 */ 1046 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 1047 /* 1048 * XXX would be more failsafe to change directory to a 1049 * deadfs node here instead 1050 */ 1051 vrele(cwdi->cwdi_cdir); 1052 VREF(vp); 1053 cwdi->cwdi_cdir = vp; 1054 } 1055 1056 return (0); 1057 } 1058 1059 /* 1060 * Common routine for chroot and chdir. 1061 */ 1062 static int 1063 change_dir(struct nameidata *ndp, struct proc *p) 1064 { 1065 struct vnode *vp; 1066 int error; 1067 1068 if ((error = namei(ndp)) != 0) 1069 return (error); 1070 vp = ndp->ni_vp; 1071 if (vp->v_type != VDIR) 1072 error = ENOTDIR; 1073 else 1074 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 1075 1076 if (error) 1077 vput(vp); 1078 else 1079 VOP_UNLOCK(vp, 0); 1080 return (error); 1081 } 1082 1083 /* 1084 * Check permissions, allocate an open file structure, 1085 * and call the device open routine if any. 1086 */ 1087 int 1088 sys_open(struct lwp *l, void *v, register_t *retval) 1089 { 1090 struct sys_open_args /* { 1091 syscallarg(const char *) path; 1092 syscallarg(int) flags; 1093 syscallarg(int) mode; 1094 } */ *uap = v; 1095 struct proc *p = l->l_proc; 1096 struct cwdinfo *cwdi = p->p_cwdi; 1097 struct filedesc *fdp = p->p_fd; 1098 struct file *fp; 1099 struct vnode *vp; 1100 int flags, cmode; 1101 int type, indx, error; 1102 struct flock lf; 1103 struct nameidata nd; 1104 1105 flags = FFLAGS(SCARG(uap, flags)); 1106 if ((flags & (FREAD | FWRITE)) == 0) 1107 return (EINVAL); 1108 /* falloc() will use the file descriptor for us */ 1109 if ((error = falloc(p, &fp, &indx)) != 0) 1110 return (error); 1111 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1112 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1113 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1114 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1115 FILE_UNUSE(fp, p); 1116 fdp->fd_ofiles[indx] = NULL; 1117 ffree(fp); 1118 if ((error == EDUPFD || error == EMOVEFD) && 1119 l->l_dupfd >= 0 && /* XXX from fdopen */ 1120 (error = 1121 dupfdopen(p, indx, l->l_dupfd, flags, error)) == 0) { 1122 *retval = indx; 1123 return (0); 1124 } 1125 if (error == ERESTART) 1126 error = EINTR; 1127 fdremove(fdp, indx); 1128 return (error); 1129 } 1130 l->l_dupfd = 0; 1131 vp = nd.ni_vp; 1132 fp->f_flag = flags & FMASK; 1133 fp->f_type = DTYPE_VNODE; 1134 fp->f_ops = &vnops; 1135 fp->f_data = vp; 1136 if (flags & (O_EXLOCK | O_SHLOCK)) { 1137 lf.l_whence = SEEK_SET; 1138 lf.l_start = 0; 1139 lf.l_len = 0; 1140 if (flags & O_EXLOCK) 1141 lf.l_type = F_WRLCK; 1142 else 1143 lf.l_type = F_RDLCK; 1144 type = F_FLOCK; 1145 if ((flags & FNONBLOCK) == 0) 1146 type |= F_WAIT; 1147 VOP_UNLOCK(vp, 0); 1148 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1149 if (error) { 1150 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1151 FILE_UNUSE(fp, p); 1152 ffree(fp); 1153 fdremove(fdp, indx); 1154 return (error); 1155 } 1156 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1157 fp->f_flag |= FHASLOCK; 1158 } 1159 VOP_UNLOCK(vp, 0); 1160 *retval = indx; 1161 FILE_SET_MATURE(fp); 1162 FILE_UNUSE(fp, p); 1163 return (0); 1164 } 1165 1166 /* 1167 * Get file handle system call 1168 */ 1169 int 1170 sys_getfh(struct lwp *l, void *v, register_t *retval) 1171 { 1172 struct sys_getfh_args /* { 1173 syscallarg(char *) fname; 1174 syscallarg(fhandle_t *) fhp; 1175 } */ *uap = v; 1176 struct proc *p = l->l_proc; 1177 struct vnode *vp; 1178 fhandle_t fh; 1179 int error; 1180 struct nameidata nd; 1181 1182 /* 1183 * Must be super user 1184 */ 1185 error = suser(p->p_ucred, &p->p_acflag); 1186 if (error) 1187 return (error); 1188 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1189 SCARG(uap, fname), p); 1190 error = namei(&nd); 1191 if (error) 1192 return (error); 1193 vp = nd.ni_vp; 1194 if (vp->v_mount->mnt_op->vfs_vptofh == NULL) 1195 return EOPNOTSUPP; 1196 memset(&fh, 0, sizeof(fh)); 1197 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1198 error = VFS_VPTOFH(vp, &fh.fh_fid); 1199 vput(vp); 1200 if (error) 1201 return (error); 1202 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1203 return (error); 1204 } 1205 1206 /* 1207 * Open a file given a file handle. 1208 * 1209 * Check permissions, allocate an open file structure, 1210 * and call the device open routine if any. 1211 */ 1212 int 1213 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1214 { 1215 struct sys_fhopen_args /* { 1216 syscallarg(const fhandle_t *) fhp; 1217 syscallarg(int) flags; 1218 } */ *uap = v; 1219 struct proc *p = l->l_proc; 1220 struct filedesc *fdp = p->p_fd; 1221 struct file *fp; 1222 struct vnode *vp = NULL; 1223 struct mount *mp; 1224 struct ucred *cred = p->p_ucred; 1225 int flags; 1226 struct file *nfp; 1227 int type, indx, error=0; 1228 struct flock lf; 1229 struct vattr va; 1230 fhandle_t fh; 1231 1232 /* 1233 * Must be super user 1234 */ 1235 if ((error = suser(p->p_ucred, &p->p_acflag))) 1236 return (error); 1237 1238 flags = FFLAGS(SCARG(uap, flags)); 1239 if ((flags & (FREAD | FWRITE)) == 0) 1240 return (EINVAL); 1241 if ((flags & O_CREAT)) 1242 return (EINVAL); 1243 /* falloc() will use the file descriptor for us */ 1244 if ((error = falloc(p, &nfp, &indx)) != 0) 1245 return (error); 1246 fp = nfp; 1247 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1248 goto bad; 1249 1250 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1251 error = ESTALE; 1252 goto bad; 1253 } 1254 1255 if (mp->mnt_op->vfs_fhtovp == NULL) { 1256 error = EOPNOTSUPP; 1257 goto bad; 1258 } 1259 1260 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1261 vp = NULL; /* most likely unnecessary sanity for bad: */ 1262 goto bad; 1263 } 1264 1265 /* Now do an effective vn_open */ 1266 1267 if (vp->v_type == VSOCK) { 1268 error = EOPNOTSUPP; 1269 goto bad; 1270 } 1271 if (flags & FREAD) { 1272 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 1273 goto bad; 1274 } 1275 if (flags & (FWRITE | O_TRUNC)) { 1276 if (vp->v_type == VDIR) { 1277 error = EISDIR; 1278 goto bad; 1279 } 1280 if ((error = vn_writechk(vp)) != 0 || 1281 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 1282 goto bad; 1283 } 1284 if (flags & O_TRUNC) { 1285 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1286 goto bad; 1287 VOP_UNLOCK(vp, 0); /* XXX */ 1288 VOP_LEASE(vp, p, cred, LEASE_WRITE); 1289 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1290 VATTR_NULL(&va); 1291 va.va_size = 0; 1292 error = VOP_SETATTR(vp, &va, cred, p); 1293 vn_finished_write(mp, 0); 1294 if (error) 1295 goto bad; 1296 } 1297 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) 1298 goto bad; 1299 if (vp->v_type == VREG && 1300 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1301 error = EIO; 1302 goto bad; 1303 } 1304 if (flags & FWRITE) 1305 vp->v_writecount++; 1306 1307 /* done with modified vn_open, now finish what sys_open does. */ 1308 1309 fp->f_flag = flags & FMASK; 1310 fp->f_type = DTYPE_VNODE; 1311 fp->f_ops = &vnops; 1312 fp->f_data = vp; 1313 if (flags & (O_EXLOCK | O_SHLOCK)) { 1314 lf.l_whence = SEEK_SET; 1315 lf.l_start = 0; 1316 lf.l_len = 0; 1317 if (flags & O_EXLOCK) 1318 lf.l_type = F_WRLCK; 1319 else 1320 lf.l_type = F_RDLCK; 1321 type = F_FLOCK; 1322 if ((flags & FNONBLOCK) == 0) 1323 type |= F_WAIT; 1324 VOP_UNLOCK(vp, 0); 1325 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1326 if (error) { 1327 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1328 FILE_UNUSE(fp, p); 1329 ffree(fp); 1330 fdremove(fdp, indx); 1331 return (error); 1332 } 1333 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1334 fp->f_flag |= FHASLOCK; 1335 } 1336 VOP_UNLOCK(vp, 0); 1337 *retval = indx; 1338 FILE_SET_MATURE(fp); 1339 FILE_UNUSE(fp, p); 1340 return (0); 1341 1342 bad: 1343 FILE_UNUSE(fp, p); 1344 ffree(fp); 1345 fdremove(fdp, indx); 1346 if (vp != NULL) 1347 vput(vp); 1348 return (error); 1349 } 1350 1351 /* ARGSUSED */ 1352 int 1353 sys_fhstat(struct lwp *l, void *v, register_t *retval) 1354 { 1355 struct sys_fhstat_args /* { 1356 syscallarg(const fhandle_t *) fhp; 1357 syscallarg(struct stat *) sb; 1358 } */ *uap = v; 1359 struct proc *p = l->l_proc; 1360 struct stat sb; 1361 int error; 1362 fhandle_t fh; 1363 struct mount *mp; 1364 struct vnode *vp; 1365 1366 /* 1367 * Must be super user 1368 */ 1369 if ((error = suser(p->p_ucred, &p->p_acflag))) 1370 return (error); 1371 1372 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1373 return (error); 1374 1375 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1376 return (ESTALE); 1377 if (mp->mnt_op->vfs_fhtovp == NULL) 1378 return EOPNOTSUPP; 1379 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1380 return (error); 1381 error = vn_stat(vp, &sb, p); 1382 vput(vp); 1383 if (error) 1384 return (error); 1385 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1386 return (error); 1387 } 1388 1389 /* ARGSUSED */ 1390 int 1391 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1392 { 1393 struct sys_fhstatvfs1_args /* { 1394 syscallarg(const fhandle_t *) fhp; 1395 syscallarg(struct statvfs *) buf; 1396 syscallarg(int) flags; 1397 } */ *uap = v; 1398 struct proc *p = l->l_proc; 1399 struct statvfs sbuf; 1400 fhandle_t fh; 1401 struct mount *mp; 1402 struct vnode *vp; 1403 int error; 1404 1405 /* 1406 * Must be super user 1407 */ 1408 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1409 return error; 1410 1411 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1412 return error; 1413 1414 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1415 return ESTALE; 1416 if (mp->mnt_op->vfs_fhtovp == NULL) 1417 return EOPNOTSUPP; 1418 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1419 return error; 1420 1421 mp = vp->v_mount; 1422 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) { 1423 vput(vp); 1424 return error; 1425 } 1426 vput(vp); 1427 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1428 } 1429 1430 /* 1431 * Create a special file. 1432 */ 1433 /* ARGSUSED */ 1434 int 1435 sys_mknod(struct lwp *l, void *v, register_t *retval) 1436 { 1437 struct sys_mknod_args /* { 1438 syscallarg(const char *) path; 1439 syscallarg(int) mode; 1440 syscallarg(int) dev; 1441 } */ *uap = v; 1442 struct proc *p = l->l_proc; 1443 struct vnode *vp; 1444 struct mount *mp; 1445 struct vattr vattr; 1446 int error; 1447 int whiteout = 0; 1448 struct nameidata nd; 1449 1450 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1451 return (error); 1452 restart: 1453 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1454 if ((error = namei(&nd)) != 0) 1455 return (error); 1456 vp = nd.ni_vp; 1457 if (vp != NULL) 1458 error = EEXIST; 1459 else { 1460 VATTR_NULL(&vattr); 1461 vattr.va_mode = 1462 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1463 vattr.va_rdev = SCARG(uap, dev); 1464 whiteout = 0; 1465 1466 switch (SCARG(uap, mode) & S_IFMT) { 1467 case S_IFMT: /* used by badsect to flag bad sectors */ 1468 vattr.va_type = VBAD; 1469 break; 1470 case S_IFCHR: 1471 vattr.va_type = VCHR; 1472 break; 1473 case S_IFBLK: 1474 vattr.va_type = VBLK; 1475 break; 1476 case S_IFWHT: 1477 whiteout = 1; 1478 break; 1479 default: 1480 error = EINVAL; 1481 break; 1482 } 1483 } 1484 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1485 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1486 if (nd.ni_dvp == vp) 1487 vrele(nd.ni_dvp); 1488 else 1489 vput(nd.ni_dvp); 1490 if (vp) 1491 vrele(vp); 1492 if ((error = vn_start_write(NULL, &mp, 1493 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1494 return (error); 1495 goto restart; 1496 } 1497 if (!error) { 1498 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1499 if (whiteout) { 1500 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1501 if (error) 1502 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1503 vput(nd.ni_dvp); 1504 } else { 1505 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1506 &nd.ni_cnd, &vattr); 1507 if (error == 0) 1508 vput(nd.ni_vp); 1509 } 1510 } else { 1511 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1512 if (nd.ni_dvp == vp) 1513 vrele(nd.ni_dvp); 1514 else 1515 vput(nd.ni_dvp); 1516 if (vp) 1517 vrele(vp); 1518 } 1519 vn_finished_write(mp, 0); 1520 return (error); 1521 } 1522 1523 /* 1524 * Create a named pipe. 1525 */ 1526 /* ARGSUSED */ 1527 int 1528 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1529 { 1530 struct sys_mkfifo_args /* { 1531 syscallarg(const char *) path; 1532 syscallarg(int) mode; 1533 } */ *uap = v; 1534 struct proc *p = l->l_proc; 1535 struct mount *mp; 1536 struct vattr vattr; 1537 int error; 1538 struct nameidata nd; 1539 1540 restart: 1541 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1542 if ((error = namei(&nd)) != 0) 1543 return (error); 1544 if (nd.ni_vp != NULL) { 1545 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1546 if (nd.ni_dvp == nd.ni_vp) 1547 vrele(nd.ni_dvp); 1548 else 1549 vput(nd.ni_dvp); 1550 vrele(nd.ni_vp); 1551 return (EEXIST); 1552 } 1553 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1554 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1555 if (nd.ni_dvp == nd.ni_vp) 1556 vrele(nd.ni_dvp); 1557 else 1558 vput(nd.ni_dvp); 1559 if (nd.ni_vp) 1560 vrele(nd.ni_vp); 1561 if ((error = vn_start_write(NULL, &mp, 1562 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1563 return (error); 1564 goto restart; 1565 } 1566 VATTR_NULL(&vattr); 1567 vattr.va_type = VFIFO; 1568 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1569 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1570 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1571 if (error == 0) 1572 vput(nd.ni_vp); 1573 vn_finished_write(mp, 0); 1574 return (error); 1575 } 1576 1577 /* 1578 * Make a hard file link. 1579 */ 1580 /* ARGSUSED */ 1581 int 1582 sys_link(struct lwp *l, void *v, register_t *retval) 1583 { 1584 struct sys_link_args /* { 1585 syscallarg(const char *) path; 1586 syscallarg(const char *) link; 1587 } */ *uap = v; 1588 struct proc *p = l->l_proc; 1589 struct vnode *vp; 1590 struct mount *mp; 1591 struct nameidata nd; 1592 int error; 1593 1594 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1595 if ((error = namei(&nd)) != 0) 1596 return (error); 1597 vp = nd.ni_vp; 1598 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1599 vrele(vp); 1600 return (error); 1601 } 1602 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1603 if ((error = namei(&nd)) != 0) 1604 goto out; 1605 if (nd.ni_vp) { 1606 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1607 if (nd.ni_dvp == nd.ni_vp) 1608 vrele(nd.ni_dvp); 1609 else 1610 vput(nd.ni_dvp); 1611 vrele(nd.ni_vp); 1612 error = EEXIST; 1613 goto out; 1614 } 1615 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1616 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1617 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1618 out: 1619 vrele(vp); 1620 vn_finished_write(mp, 0); 1621 return (error); 1622 } 1623 1624 /* 1625 * Make a symbolic link. 1626 */ 1627 /* ARGSUSED */ 1628 int 1629 sys_symlink(struct lwp *l, void *v, register_t *retval) 1630 { 1631 struct sys_symlink_args /* { 1632 syscallarg(const char *) path; 1633 syscallarg(const char *) link; 1634 } */ *uap = v; 1635 struct proc *p = l->l_proc; 1636 struct mount *mp; 1637 struct vattr vattr; 1638 char *path; 1639 int error; 1640 struct nameidata nd; 1641 1642 path = PNBUF_GET(); 1643 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1644 if (error) 1645 goto out; 1646 restart: 1647 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1648 if ((error = namei(&nd)) != 0) 1649 goto out; 1650 if (nd.ni_vp) { 1651 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1652 if (nd.ni_dvp == nd.ni_vp) 1653 vrele(nd.ni_dvp); 1654 else 1655 vput(nd.ni_dvp); 1656 vrele(nd.ni_vp); 1657 error = EEXIST; 1658 goto out; 1659 } 1660 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1661 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1662 if (nd.ni_dvp == nd.ni_vp) 1663 vrele(nd.ni_dvp); 1664 else 1665 vput(nd.ni_dvp); 1666 if ((error = vn_start_write(NULL, &mp, 1667 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1668 return (error); 1669 goto restart; 1670 } 1671 VATTR_NULL(&vattr); 1672 vattr.va_type = VLNK; 1673 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1674 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1675 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1676 if (error == 0) 1677 vput(nd.ni_vp); 1678 vn_finished_write(mp, 0); 1679 out: 1680 PNBUF_PUT(path); 1681 return (error); 1682 } 1683 1684 /* 1685 * Delete a whiteout from the filesystem. 1686 */ 1687 /* ARGSUSED */ 1688 int 1689 sys_undelete(struct lwp *l, void *v, register_t *retval) 1690 { 1691 struct sys_undelete_args /* { 1692 syscallarg(const char *) path; 1693 } */ *uap = v; 1694 struct proc *p = l->l_proc; 1695 int error; 1696 struct mount *mp; 1697 struct nameidata nd; 1698 1699 restart: 1700 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1701 SCARG(uap, path), p); 1702 error = namei(&nd); 1703 if (error) 1704 return (error); 1705 1706 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1707 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1708 if (nd.ni_dvp == nd.ni_vp) 1709 vrele(nd.ni_dvp); 1710 else 1711 vput(nd.ni_dvp); 1712 if (nd.ni_vp) 1713 vrele(nd.ni_vp); 1714 return (EEXIST); 1715 } 1716 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1717 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1718 if (nd.ni_dvp == nd.ni_vp) 1719 vrele(nd.ni_dvp); 1720 else 1721 vput(nd.ni_dvp); 1722 if ((error = vn_start_write(NULL, &mp, 1723 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1724 return (error); 1725 goto restart; 1726 } 1727 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1728 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1729 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1730 vput(nd.ni_dvp); 1731 vn_finished_write(mp, 0); 1732 return (error); 1733 } 1734 1735 /* 1736 * Delete a name from the filesystem. 1737 */ 1738 /* ARGSUSED */ 1739 int 1740 sys_unlink(struct lwp *l, void *v, register_t *retval) 1741 { 1742 struct sys_unlink_args /* { 1743 syscallarg(const char *) path; 1744 } */ *uap = v; 1745 struct proc *p = l->l_proc; 1746 struct mount *mp; 1747 struct vnode *vp; 1748 int error; 1749 struct nameidata nd; 1750 1751 restart: 1752 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1753 SCARG(uap, path), p); 1754 if ((error = namei(&nd)) != 0) 1755 return (error); 1756 vp = nd.ni_vp; 1757 1758 /* 1759 * The root of a mounted filesystem cannot be deleted. 1760 */ 1761 if (vp->v_flag & VROOT) { 1762 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1763 if (nd.ni_dvp == vp) 1764 vrele(nd.ni_dvp); 1765 else 1766 vput(nd.ni_dvp); 1767 vput(vp); 1768 error = EBUSY; 1769 goto out; 1770 } 1771 1772 #ifdef VERIFIED_EXEC 1773 /* Handle remove requests for veriexec entries. */ 1774 if ((error = veriexec_removechk(p, vp, nd.ni_dirp)) != 0) { 1775 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1776 if (nd.ni_dvp == vp) 1777 vrele(nd.ni_dvp); 1778 else 1779 vput(nd.ni_dvp); 1780 vput(vp); 1781 goto out; 1782 } 1783 #endif /* VERIFIED_EXEC */ 1784 1785 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 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 if ((error = vn_start_write(NULL, &mp, 1793 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1794 return (error); 1795 goto restart; 1796 } 1797 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1798 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1799 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1800 vn_finished_write(mp, 0); 1801 out: 1802 return (error); 1803 } 1804 1805 /* 1806 * Reposition read/write file offset. 1807 */ 1808 int 1809 sys_lseek(struct lwp *l, void *v, register_t *retval) 1810 { 1811 struct sys_lseek_args /* { 1812 syscallarg(int) fd; 1813 syscallarg(int) pad; 1814 syscallarg(off_t) offset; 1815 syscallarg(int) whence; 1816 } */ *uap = v; 1817 struct proc *p = l->l_proc; 1818 struct ucred *cred = p->p_ucred; 1819 struct filedesc *fdp = p->p_fd; 1820 struct file *fp; 1821 struct vnode *vp; 1822 struct vattr vattr; 1823 off_t newoff; 1824 int error; 1825 1826 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1827 return (EBADF); 1828 1829 FILE_USE(fp); 1830 1831 vp = (struct vnode *)fp->f_data; 1832 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1833 error = ESPIPE; 1834 goto out; 1835 } 1836 1837 switch (SCARG(uap, whence)) { 1838 case SEEK_CUR: 1839 newoff = fp->f_offset + SCARG(uap, offset); 1840 break; 1841 case SEEK_END: 1842 error = VOP_GETATTR(vp, &vattr, cred, p); 1843 if (error) 1844 goto out; 1845 newoff = SCARG(uap, offset) + vattr.va_size; 1846 break; 1847 case SEEK_SET: 1848 newoff = SCARG(uap, offset); 1849 break; 1850 default: 1851 error = EINVAL; 1852 goto out; 1853 } 1854 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1855 goto out; 1856 1857 *(off_t *)retval = fp->f_offset = newoff; 1858 out: 1859 FILE_UNUSE(fp, p); 1860 return (error); 1861 } 1862 1863 /* 1864 * Positional read system call. 1865 */ 1866 int 1867 sys_pread(struct lwp *l, void *v, register_t *retval) 1868 { 1869 struct sys_pread_args /* { 1870 syscallarg(int) fd; 1871 syscallarg(void *) buf; 1872 syscallarg(size_t) nbyte; 1873 syscallarg(off_t) offset; 1874 } */ *uap = v; 1875 struct proc *p = l->l_proc; 1876 struct filedesc *fdp = p->p_fd; 1877 struct file *fp; 1878 struct vnode *vp; 1879 off_t offset; 1880 int error, fd = SCARG(uap, fd); 1881 1882 if ((fp = fd_getfile(fdp, fd)) == NULL) 1883 return (EBADF); 1884 1885 if ((fp->f_flag & FREAD) == 0) { 1886 simple_unlock(&fp->f_slock); 1887 return (EBADF); 1888 } 1889 1890 FILE_USE(fp); 1891 1892 vp = (struct vnode *)fp->f_data; 1893 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1894 error = ESPIPE; 1895 goto out; 1896 } 1897 1898 offset = SCARG(uap, offset); 1899 1900 /* 1901 * XXX This works because no file systems actually 1902 * XXX take any action on the seek operation. 1903 */ 1904 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1905 goto out; 1906 1907 /* dofileread() will unuse the descriptor for us */ 1908 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1909 &offset, 0, retval)); 1910 1911 out: 1912 FILE_UNUSE(fp, p); 1913 return (error); 1914 } 1915 1916 /* 1917 * Positional scatter read system call. 1918 */ 1919 int 1920 sys_preadv(struct lwp *l, void *v, register_t *retval) 1921 { 1922 struct sys_preadv_args /* { 1923 syscallarg(int) fd; 1924 syscallarg(const struct iovec *) iovp; 1925 syscallarg(int) iovcnt; 1926 syscallarg(off_t) offset; 1927 } */ *uap = v; 1928 struct proc *p = l->l_proc; 1929 struct filedesc *fdp = p->p_fd; 1930 struct file *fp; 1931 struct vnode *vp; 1932 off_t offset; 1933 int error, fd = SCARG(uap, fd); 1934 1935 if ((fp = fd_getfile(fdp, fd)) == NULL) 1936 return (EBADF); 1937 1938 if ((fp->f_flag & FREAD) == 0) { 1939 simple_unlock(&fp->f_slock); 1940 return (EBADF); 1941 } 1942 1943 FILE_USE(fp); 1944 1945 vp = (struct vnode *)fp->f_data; 1946 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1947 error = ESPIPE; 1948 goto out; 1949 } 1950 1951 offset = SCARG(uap, offset); 1952 1953 /* 1954 * XXX This works because no file systems actually 1955 * XXX take any action on the seek operation. 1956 */ 1957 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1958 goto out; 1959 1960 /* dofilereadv() will unuse the descriptor for us */ 1961 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1962 &offset, 0, retval)); 1963 1964 out: 1965 FILE_UNUSE(fp, p); 1966 return (error); 1967 } 1968 1969 /* 1970 * Positional write system call. 1971 */ 1972 int 1973 sys_pwrite(struct lwp *l, void *v, register_t *retval) 1974 { 1975 struct sys_pwrite_args /* { 1976 syscallarg(int) fd; 1977 syscallarg(const void *) buf; 1978 syscallarg(size_t) nbyte; 1979 syscallarg(off_t) offset; 1980 } */ *uap = v; 1981 struct proc *p = l->l_proc; 1982 struct filedesc *fdp = p->p_fd; 1983 struct file *fp; 1984 struct vnode *vp; 1985 off_t offset; 1986 int error, fd = SCARG(uap, fd); 1987 1988 if ((fp = fd_getfile(fdp, fd)) == NULL) 1989 return (EBADF); 1990 1991 if ((fp->f_flag & FWRITE) == 0) { 1992 simple_unlock(&fp->f_slock); 1993 return (EBADF); 1994 } 1995 1996 FILE_USE(fp); 1997 1998 vp = (struct vnode *)fp->f_data; 1999 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2000 error = ESPIPE; 2001 goto out; 2002 } 2003 2004 offset = SCARG(uap, offset); 2005 2006 /* 2007 * XXX This works because no file systems actually 2008 * XXX take any action on the seek operation. 2009 */ 2010 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2011 goto out; 2012 2013 /* dofilewrite() will unuse the descriptor for us */ 2014 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2015 &offset, 0, retval)); 2016 2017 out: 2018 FILE_UNUSE(fp, p); 2019 return (error); 2020 } 2021 2022 /* 2023 * Positional gather write system call. 2024 */ 2025 int 2026 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2027 { 2028 struct sys_pwritev_args /* { 2029 syscallarg(int) fd; 2030 syscallarg(const struct iovec *) iovp; 2031 syscallarg(int) iovcnt; 2032 syscallarg(off_t) offset; 2033 } */ *uap = v; 2034 struct proc *p = l->l_proc; 2035 struct filedesc *fdp = p->p_fd; 2036 struct file *fp; 2037 struct vnode *vp; 2038 off_t offset; 2039 int error, fd = SCARG(uap, fd); 2040 2041 if ((fp = fd_getfile(fdp, fd)) == NULL) 2042 return (EBADF); 2043 2044 if ((fp->f_flag & FWRITE) == 0) { 2045 simple_unlock(&fp->f_slock); 2046 return (EBADF); 2047 } 2048 2049 FILE_USE(fp); 2050 2051 vp = (struct vnode *)fp->f_data; 2052 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2053 error = ESPIPE; 2054 goto out; 2055 } 2056 2057 offset = SCARG(uap, offset); 2058 2059 /* 2060 * XXX This works because no file systems actually 2061 * XXX take any action on the seek operation. 2062 */ 2063 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2064 goto out; 2065 2066 /* dofilewritev() will unuse the descriptor for us */ 2067 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2068 &offset, 0, retval)); 2069 2070 out: 2071 FILE_UNUSE(fp, p); 2072 return (error); 2073 } 2074 2075 /* 2076 * Check access permissions. 2077 */ 2078 int 2079 sys_access(struct lwp *l, void *v, register_t *retval) 2080 { 2081 struct sys_access_args /* { 2082 syscallarg(const char *) path; 2083 syscallarg(int) flags; 2084 } */ *uap = v; 2085 struct proc *p = l->l_proc; 2086 struct ucred *cred; 2087 struct vnode *vp; 2088 int error, flags; 2089 struct nameidata nd; 2090 2091 cred = crdup(p->p_ucred); 2092 cred->cr_uid = p->p_cred->p_ruid; 2093 cred->cr_gid = p->p_cred->p_rgid; 2094 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2095 SCARG(uap, path), p); 2096 /* Override default credentials */ 2097 nd.ni_cnd.cn_cred = cred; 2098 if ((error = namei(&nd)) != 0) 2099 goto out; 2100 vp = nd.ni_vp; 2101 2102 /* Flags == 0 means only check for existence. */ 2103 if (SCARG(uap, flags)) { 2104 flags = 0; 2105 if (SCARG(uap, flags) & R_OK) 2106 flags |= VREAD; 2107 if (SCARG(uap, flags) & W_OK) 2108 flags |= VWRITE; 2109 if (SCARG(uap, flags) & X_OK) 2110 flags |= VEXEC; 2111 2112 error = VOP_ACCESS(vp, flags, cred, p); 2113 if (!error && (flags & VWRITE)) 2114 error = vn_writechk(vp); 2115 } 2116 vput(vp); 2117 out: 2118 crfree(cred); 2119 return (error); 2120 } 2121 2122 /* 2123 * Get file status; this version follows links. 2124 */ 2125 /* ARGSUSED */ 2126 int 2127 sys___stat30(struct lwp *l, void *v, register_t *retval) 2128 { 2129 struct sys___stat30_args /* { 2130 syscallarg(const char *) path; 2131 syscallarg(struct stat *) ub; 2132 } */ *uap = v; 2133 struct proc *p = l->l_proc; 2134 struct stat sb; 2135 int error; 2136 struct nameidata nd; 2137 2138 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2139 SCARG(uap, path), p); 2140 if ((error = namei(&nd)) != 0) 2141 return (error); 2142 error = vn_stat(nd.ni_vp, &sb, p); 2143 vput(nd.ni_vp); 2144 if (error) 2145 return (error); 2146 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2147 return (error); 2148 } 2149 2150 /* 2151 * Get file status; this version does not follow links. 2152 */ 2153 /* ARGSUSED */ 2154 int 2155 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2156 { 2157 struct sys___lstat30_args /* { 2158 syscallarg(const char *) path; 2159 syscallarg(struct stat *) ub; 2160 } */ *uap = v; 2161 struct proc *p = l->l_proc; 2162 struct stat sb; 2163 int error; 2164 struct nameidata nd; 2165 2166 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2167 SCARG(uap, path), p); 2168 if ((error = namei(&nd)) != 0) 2169 return (error); 2170 error = vn_stat(nd.ni_vp, &sb, p); 2171 vput(nd.ni_vp); 2172 if (error) 2173 return (error); 2174 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2175 return (error); 2176 } 2177 2178 /* 2179 * Get configurable pathname variables. 2180 */ 2181 /* ARGSUSED */ 2182 int 2183 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2184 { 2185 struct sys_pathconf_args /* { 2186 syscallarg(const char *) path; 2187 syscallarg(int) name; 2188 } */ *uap = v; 2189 struct proc *p = l->l_proc; 2190 int error; 2191 struct nameidata nd; 2192 2193 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2194 SCARG(uap, path), p); 2195 if ((error = namei(&nd)) != 0) 2196 return (error); 2197 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2198 vput(nd.ni_vp); 2199 return (error); 2200 } 2201 2202 /* 2203 * Return target name of a symbolic link. 2204 */ 2205 /* ARGSUSED */ 2206 int 2207 sys_readlink(struct lwp *l, void *v, register_t *retval) 2208 { 2209 struct sys_readlink_args /* { 2210 syscallarg(const char *) path; 2211 syscallarg(char *) buf; 2212 syscallarg(size_t) count; 2213 } */ *uap = v; 2214 struct proc *p = l->l_proc; 2215 struct vnode *vp; 2216 struct iovec aiov; 2217 struct uio auio; 2218 int error; 2219 struct nameidata nd; 2220 2221 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2222 SCARG(uap, path), p); 2223 if ((error = namei(&nd)) != 0) 2224 return (error); 2225 vp = nd.ni_vp; 2226 if (vp->v_type != VLNK) 2227 error = EINVAL; 2228 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2229 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 2230 aiov.iov_base = SCARG(uap, buf); 2231 aiov.iov_len = SCARG(uap, count); 2232 auio.uio_iov = &aiov; 2233 auio.uio_iovcnt = 1; 2234 auio.uio_offset = 0; 2235 auio.uio_rw = UIO_READ; 2236 auio.uio_segflg = UIO_USERSPACE; 2237 auio.uio_procp = p; 2238 auio.uio_resid = SCARG(uap, count); 2239 error = VOP_READLINK(vp, &auio, p->p_ucred); 2240 } 2241 vput(vp); 2242 *retval = SCARG(uap, count) - auio.uio_resid; 2243 return (error); 2244 } 2245 2246 /* 2247 * Change flags of a file given a path name. 2248 */ 2249 /* ARGSUSED */ 2250 int 2251 sys_chflags(struct lwp *l, void *v, register_t *retval) 2252 { 2253 struct sys_chflags_args /* { 2254 syscallarg(const char *) path; 2255 syscallarg(u_long) flags; 2256 } */ *uap = v; 2257 struct proc *p = l->l_proc; 2258 struct vnode *vp; 2259 int error; 2260 struct nameidata nd; 2261 2262 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2263 if ((error = namei(&nd)) != 0) 2264 return (error); 2265 vp = nd.ni_vp; 2266 error = change_flags(vp, SCARG(uap, flags), p); 2267 vput(vp); 2268 return (error); 2269 } 2270 2271 /* 2272 * Change flags of a file given a file descriptor. 2273 */ 2274 /* ARGSUSED */ 2275 int 2276 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2277 { 2278 struct sys_fchflags_args /* { 2279 syscallarg(int) fd; 2280 syscallarg(u_long) flags; 2281 } */ *uap = v; 2282 struct proc *p = l->l_proc; 2283 struct vnode *vp; 2284 struct file *fp; 2285 int error; 2286 2287 /* getvnode() will use the descriptor for us */ 2288 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2289 return (error); 2290 vp = (struct vnode *)fp->f_data; 2291 error = change_flags(vp, SCARG(uap, flags), p); 2292 VOP_UNLOCK(vp, 0); 2293 FILE_UNUSE(fp, p); 2294 return (error); 2295 } 2296 2297 /* 2298 * Change flags of a file given a path name; this version does 2299 * not follow links. 2300 */ 2301 int 2302 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2303 { 2304 struct sys_lchflags_args /* { 2305 syscallarg(const char *) path; 2306 syscallarg(u_long) flags; 2307 } */ *uap = v; 2308 struct proc *p = l->l_proc; 2309 struct vnode *vp; 2310 int error; 2311 struct nameidata nd; 2312 2313 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2314 if ((error = namei(&nd)) != 0) 2315 return (error); 2316 vp = nd.ni_vp; 2317 error = change_flags(vp, SCARG(uap, flags), p); 2318 vput(vp); 2319 return (error); 2320 } 2321 2322 /* 2323 * Common routine to change flags of a file. 2324 */ 2325 int 2326 change_flags(struct vnode *vp, u_long flags, struct proc *p) 2327 { 2328 struct mount *mp; 2329 struct vattr vattr; 2330 int error; 2331 2332 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2333 return (error); 2334 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2335 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2336 /* 2337 * Non-superusers cannot change the flags on devices, even if they 2338 * own them. 2339 */ 2340 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2341 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2342 goto out; 2343 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2344 error = EINVAL; 2345 goto out; 2346 } 2347 } 2348 VATTR_NULL(&vattr); 2349 vattr.va_flags = flags; 2350 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2351 out: 2352 vn_finished_write(mp, 0); 2353 return (error); 2354 } 2355 2356 /* 2357 * Change mode of a file given path name; this version follows links. 2358 */ 2359 /* ARGSUSED */ 2360 int 2361 sys_chmod(struct lwp *l, void *v, register_t *retval) 2362 { 2363 struct sys_chmod_args /* { 2364 syscallarg(const char *) path; 2365 syscallarg(int) mode; 2366 } */ *uap = v; 2367 struct proc *p = l->l_proc; 2368 int error; 2369 struct nameidata nd; 2370 2371 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2372 if ((error = namei(&nd)) != 0) 2373 return (error); 2374 2375 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2376 2377 vrele(nd.ni_vp); 2378 return (error); 2379 } 2380 2381 /* 2382 * Change mode of a file given a file descriptor. 2383 */ 2384 /* ARGSUSED */ 2385 int 2386 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2387 { 2388 struct sys_fchmod_args /* { 2389 syscallarg(int) fd; 2390 syscallarg(int) mode; 2391 } */ *uap = v; 2392 struct proc *p = l->l_proc; 2393 struct file *fp; 2394 int error; 2395 2396 /* getvnode() will use the descriptor for us */ 2397 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2398 return (error); 2399 2400 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 2401 FILE_UNUSE(fp, p); 2402 return (error); 2403 } 2404 2405 /* 2406 * Change mode of a file given path name; this version does not follow links. 2407 */ 2408 /* ARGSUSED */ 2409 int 2410 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2411 { 2412 struct sys_lchmod_args /* { 2413 syscallarg(const char *) path; 2414 syscallarg(int) mode; 2415 } */ *uap = v; 2416 struct proc *p = l->l_proc; 2417 int error; 2418 struct nameidata nd; 2419 2420 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2421 if ((error = namei(&nd)) != 0) 2422 return (error); 2423 2424 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2425 2426 vrele(nd.ni_vp); 2427 return (error); 2428 } 2429 2430 /* 2431 * Common routine to set mode given a vnode. 2432 */ 2433 static int 2434 change_mode(struct vnode *vp, int mode, struct proc *p) 2435 { 2436 struct mount *mp; 2437 struct vattr vattr; 2438 int error; 2439 2440 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2441 return (error); 2442 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2443 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2444 VATTR_NULL(&vattr); 2445 vattr.va_mode = mode & ALLPERMS; 2446 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2447 VOP_UNLOCK(vp, 0); 2448 vn_finished_write(mp, 0); 2449 return (error); 2450 } 2451 2452 /* 2453 * Set ownership given a path name; this version follows links. 2454 */ 2455 /* ARGSUSED */ 2456 int 2457 sys_chown(struct lwp *l, void *v, register_t *retval) 2458 { 2459 struct sys_chown_args /* { 2460 syscallarg(const char *) path; 2461 syscallarg(uid_t) uid; 2462 syscallarg(gid_t) gid; 2463 } */ *uap = v; 2464 struct proc *p = l->l_proc; 2465 int error; 2466 struct nameidata nd; 2467 2468 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2469 if ((error = namei(&nd)) != 0) 2470 return (error); 2471 2472 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2473 2474 vrele(nd.ni_vp); 2475 return (error); 2476 } 2477 2478 /* 2479 * Set ownership given a path name; this version follows links. 2480 * Provides POSIX semantics. 2481 */ 2482 /* ARGSUSED */ 2483 int 2484 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2485 { 2486 struct sys_chown_args /* { 2487 syscallarg(const char *) path; 2488 syscallarg(uid_t) uid; 2489 syscallarg(gid_t) gid; 2490 } */ *uap = v; 2491 struct proc *p = l->l_proc; 2492 int error; 2493 struct nameidata nd; 2494 2495 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2496 if ((error = namei(&nd)) != 0) 2497 return (error); 2498 2499 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2500 2501 vrele(nd.ni_vp); 2502 return (error); 2503 } 2504 2505 /* 2506 * Set ownership given a file descriptor. 2507 */ 2508 /* ARGSUSED */ 2509 int 2510 sys_fchown(struct lwp *l, void *v, register_t *retval) 2511 { 2512 struct sys_fchown_args /* { 2513 syscallarg(int) fd; 2514 syscallarg(uid_t) uid; 2515 syscallarg(gid_t) gid; 2516 } */ *uap = v; 2517 struct proc *p = l->l_proc; 2518 int error; 2519 struct file *fp; 2520 2521 /* getvnode() will use the descriptor for us */ 2522 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2523 return (error); 2524 2525 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2526 SCARG(uap, gid), p, 0); 2527 FILE_UNUSE(fp, p); 2528 return (error); 2529 } 2530 2531 /* 2532 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2533 */ 2534 /* ARGSUSED */ 2535 int 2536 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2537 { 2538 struct sys_fchown_args /* { 2539 syscallarg(int) fd; 2540 syscallarg(uid_t) uid; 2541 syscallarg(gid_t) gid; 2542 } */ *uap = v; 2543 struct proc *p = l->l_proc; 2544 int error; 2545 struct file *fp; 2546 2547 /* getvnode() will use the descriptor for us */ 2548 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2549 return (error); 2550 2551 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2552 SCARG(uap, gid), p, 1); 2553 FILE_UNUSE(fp, p); 2554 return (error); 2555 } 2556 2557 /* 2558 * Set ownership given a path name; this version does not follow links. 2559 */ 2560 /* ARGSUSED */ 2561 int 2562 sys_lchown(struct lwp *l, void *v, register_t *retval) 2563 { 2564 struct sys_lchown_args /* { 2565 syscallarg(const char *) path; 2566 syscallarg(uid_t) uid; 2567 syscallarg(gid_t) gid; 2568 } */ *uap = v; 2569 struct proc *p = l->l_proc; 2570 int error; 2571 struct nameidata nd; 2572 2573 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2574 if ((error = namei(&nd)) != 0) 2575 return (error); 2576 2577 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2578 2579 vrele(nd.ni_vp); 2580 return (error); 2581 } 2582 2583 /* 2584 * Set ownership given a path name; this version does not follow links. 2585 * Provides POSIX/XPG semantics. 2586 */ 2587 /* ARGSUSED */ 2588 int 2589 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2590 { 2591 struct sys_lchown_args /* { 2592 syscallarg(const char *) path; 2593 syscallarg(uid_t) uid; 2594 syscallarg(gid_t) gid; 2595 } */ *uap = v; 2596 struct proc *p = l->l_proc; 2597 int error; 2598 struct nameidata nd; 2599 2600 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2601 if ((error = namei(&nd)) != 0) 2602 return (error); 2603 2604 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2605 2606 vrele(nd.ni_vp); 2607 return (error); 2608 } 2609 2610 /* 2611 * Common routine to set ownership given a vnode. 2612 */ 2613 static int 2614 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct proc *p, 2615 int posix_semantics) 2616 { 2617 struct mount *mp; 2618 struct vattr vattr; 2619 mode_t newmode; 2620 int error; 2621 2622 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2623 return (error); 2624 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2625 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2626 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2627 goto out; 2628 2629 #define CHANGED(x) ((int)(x) != -1) 2630 newmode = vattr.va_mode; 2631 if (posix_semantics) { 2632 /* 2633 * POSIX/XPG semantics: if the caller is not the super-user, 2634 * clear set-user-id and set-group-id bits. Both POSIX and 2635 * the XPG consider the behaviour for calls by the super-user 2636 * implementation-defined; we leave the set-user-id and set- 2637 * group-id settings intact in that case. 2638 */ 2639 if (suser(p->p_ucred, NULL) != 0) 2640 newmode &= ~(S_ISUID | S_ISGID); 2641 } else { 2642 /* 2643 * NetBSD semantics: when changing owner and/or group, 2644 * clear the respective bit(s). 2645 */ 2646 if (CHANGED(uid)) 2647 newmode &= ~S_ISUID; 2648 if (CHANGED(gid)) 2649 newmode &= ~S_ISGID; 2650 } 2651 /* Update va_mode iff altered. */ 2652 if (vattr.va_mode == newmode) 2653 newmode = VNOVAL; 2654 2655 VATTR_NULL(&vattr); 2656 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2657 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2658 vattr.va_mode = newmode; 2659 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2660 #undef CHANGED 2661 2662 out: 2663 VOP_UNLOCK(vp, 0); 2664 vn_finished_write(mp, 0); 2665 return (error); 2666 } 2667 2668 /* 2669 * Set the access and modification times given a path name; this 2670 * version follows links. 2671 */ 2672 /* ARGSUSED */ 2673 int 2674 sys_utimes(struct lwp *l, void *v, register_t *retval) 2675 { 2676 struct sys_utimes_args /* { 2677 syscallarg(const char *) path; 2678 syscallarg(const struct timeval *) tptr; 2679 } */ *uap = v; 2680 struct proc *p = l->l_proc; 2681 int error; 2682 struct nameidata nd; 2683 2684 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2685 if ((error = namei(&nd)) != 0) 2686 return (error); 2687 2688 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2689 2690 vrele(nd.ni_vp); 2691 return (error); 2692 } 2693 2694 /* 2695 * Set the access and modification times given a file descriptor. 2696 */ 2697 /* ARGSUSED */ 2698 int 2699 sys_futimes(struct lwp *l, void *v, register_t *retval) 2700 { 2701 struct sys_futimes_args /* { 2702 syscallarg(int) fd; 2703 syscallarg(const struct timeval *) tptr; 2704 } */ *uap = v; 2705 struct proc *p = l->l_proc; 2706 int error; 2707 struct file *fp; 2708 2709 /* getvnode() will use the descriptor for us */ 2710 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2711 return (error); 2712 2713 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2714 FILE_UNUSE(fp, p); 2715 return (error); 2716 } 2717 2718 /* 2719 * Set the access and modification times given a path name; this 2720 * version does not follow links. 2721 */ 2722 /* ARGSUSED */ 2723 int 2724 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2725 { 2726 struct sys_lutimes_args /* { 2727 syscallarg(const char *) path; 2728 syscallarg(const struct timeval *) tptr; 2729 } */ *uap = v; 2730 struct proc *p = l->l_proc; 2731 int error; 2732 struct nameidata nd; 2733 2734 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2735 if ((error = namei(&nd)) != 0) 2736 return (error); 2737 2738 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2739 2740 vrele(nd.ni_vp); 2741 return (error); 2742 } 2743 2744 /* 2745 * Common routine to set access and modification times given a vnode. 2746 */ 2747 static int 2748 change_utimes(struct vnode *vp, const struct timeval *tptr, struct proc *p) 2749 { 2750 struct timeval tv[2]; 2751 struct mount *mp; 2752 struct vattr vattr; 2753 int error; 2754 2755 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2756 return (error); 2757 VATTR_NULL(&vattr); 2758 if (tptr == NULL) { 2759 microtime(&tv[0]); 2760 tv[1] = tv[0]; 2761 vattr.va_vaflags |= VA_UTIMES_NULL; 2762 } else { 2763 error = copyin(tptr, tv, sizeof(tv)); 2764 if (error) 2765 goto out; 2766 } 2767 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2768 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2769 vattr.va_atime.tv_sec = tv[0].tv_sec; 2770 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2771 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2772 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2773 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2774 VOP_UNLOCK(vp, 0); 2775 out: 2776 vn_finished_write(mp, 0); 2777 return (error); 2778 } 2779 2780 /* 2781 * Truncate a file given its path name. 2782 */ 2783 /* ARGSUSED */ 2784 int 2785 sys_truncate(struct lwp *l, void *v, register_t *retval) 2786 { 2787 struct sys_truncate_args /* { 2788 syscallarg(const char *) path; 2789 syscallarg(int) pad; 2790 syscallarg(off_t) length; 2791 } */ *uap = v; 2792 struct proc *p = l->l_proc; 2793 struct vnode *vp; 2794 struct mount *mp; 2795 struct vattr vattr; 2796 int error; 2797 struct nameidata nd; 2798 2799 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2800 if ((error = namei(&nd)) != 0) 2801 return (error); 2802 vp = nd.ni_vp; 2803 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2804 vrele(vp); 2805 return (error); 2806 } 2807 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2808 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2809 if (vp->v_type == VDIR) 2810 error = EISDIR; 2811 else if ((error = vn_writechk(vp)) == 0 && 2812 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2813 VATTR_NULL(&vattr); 2814 vattr.va_size = SCARG(uap, length); 2815 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2816 } 2817 vput(vp); 2818 vn_finished_write(mp, 0); 2819 return (error); 2820 } 2821 2822 /* 2823 * Truncate a file given a file descriptor. 2824 */ 2825 /* ARGSUSED */ 2826 int 2827 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2828 { 2829 struct sys_ftruncate_args /* { 2830 syscallarg(int) fd; 2831 syscallarg(int) pad; 2832 syscallarg(off_t) length; 2833 } */ *uap = v; 2834 struct proc *p = l->l_proc; 2835 struct mount *mp; 2836 struct vattr vattr; 2837 struct vnode *vp; 2838 struct file *fp; 2839 int error; 2840 2841 /* getvnode() will use the descriptor for us */ 2842 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2843 return (error); 2844 if ((fp->f_flag & FWRITE) == 0) { 2845 error = EINVAL; 2846 goto out; 2847 } 2848 vp = (struct vnode *)fp->f_data; 2849 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2850 FILE_UNUSE(fp, p); 2851 return (error); 2852 } 2853 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2854 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2855 if (vp->v_type == VDIR) 2856 error = EISDIR; 2857 else if ((error = vn_writechk(vp)) == 0) { 2858 VATTR_NULL(&vattr); 2859 vattr.va_size = SCARG(uap, length); 2860 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2861 } 2862 VOP_UNLOCK(vp, 0); 2863 vn_finished_write(mp, 0); 2864 out: 2865 FILE_UNUSE(fp, p); 2866 return (error); 2867 } 2868 2869 /* 2870 * Sync an open file. 2871 */ 2872 /* ARGSUSED */ 2873 int 2874 sys_fsync(struct lwp *l, void *v, register_t *retval) 2875 { 2876 struct sys_fsync_args /* { 2877 syscallarg(int) fd; 2878 } */ *uap = v; 2879 struct proc *p = l->l_proc; 2880 struct vnode *vp; 2881 struct mount *mp; 2882 struct file *fp; 2883 int error; 2884 2885 /* getvnode() will use the descriptor for us */ 2886 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2887 return (error); 2888 vp = (struct vnode *)fp->f_data; 2889 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2890 FILE_UNUSE(fp, p); 2891 return (error); 2892 } 2893 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2894 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p); 2895 if (error == 0 && bioops.io_fsync != NULL && 2896 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2897 (*bioops.io_fsync)(vp, 0); 2898 VOP_UNLOCK(vp, 0); 2899 vn_finished_write(mp, 0); 2900 FILE_UNUSE(fp, p); 2901 return (error); 2902 } 2903 2904 /* 2905 * Sync a range of file data. API modeled after that found in AIX. 2906 * 2907 * FDATASYNC indicates that we need only save enough metadata to be able 2908 * to re-read the written data. Note we duplicate AIX's requirement that 2909 * the file be open for writing. 2910 */ 2911 /* ARGSUSED */ 2912 int 2913 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2914 { 2915 struct sys_fsync_range_args /* { 2916 syscallarg(int) fd; 2917 syscallarg(int) flags; 2918 syscallarg(off_t) start; 2919 syscallarg(off_t) length; 2920 } */ *uap = v; 2921 struct proc *p = l->l_proc; 2922 struct vnode *vp; 2923 struct file *fp; 2924 int flags, nflags; 2925 off_t s, e, len; 2926 int error; 2927 2928 /* getvnode() will use the descriptor for us */ 2929 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2930 return (error); 2931 2932 if ((fp->f_flag & FWRITE) == 0) { 2933 FILE_UNUSE(fp, p); 2934 return (EBADF); 2935 } 2936 2937 flags = SCARG(uap, flags); 2938 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2939 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2940 return (EINVAL); 2941 } 2942 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2943 if (flags & FDATASYNC) 2944 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2945 else 2946 nflags = FSYNC_WAIT; 2947 if (flags & FDISKSYNC) 2948 nflags |= FSYNC_CACHE; 2949 2950 len = SCARG(uap, length); 2951 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2952 if (len) { 2953 s = SCARG(uap, start); 2954 e = s + len; 2955 if (e < s) { 2956 FILE_UNUSE(fp, p); 2957 return (EINVAL); 2958 } 2959 } else { 2960 e = 0; 2961 s = 0; 2962 } 2963 2964 vp = (struct vnode *)fp->f_data; 2965 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2966 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p); 2967 2968 if (error == 0 && bioops.io_fsync != NULL && 2969 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2970 (*bioops.io_fsync)(vp, nflags); 2971 2972 VOP_UNLOCK(vp, 0); 2973 FILE_UNUSE(fp, p); 2974 return (error); 2975 } 2976 2977 /* 2978 * Sync the data of an open file. 2979 */ 2980 /* ARGSUSED */ 2981 int 2982 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 2983 { 2984 struct sys_fdatasync_args /* { 2985 syscallarg(int) fd; 2986 } */ *uap = v; 2987 struct proc *p = l->l_proc; 2988 struct vnode *vp; 2989 struct file *fp; 2990 int error; 2991 2992 /* getvnode() will use the descriptor for us */ 2993 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2994 return (error); 2995 if ((fp->f_flag & FWRITE) == 0) { 2996 FILE_UNUSE(fp, p); 2997 return (EBADF); 2998 } 2999 vp = (struct vnode *)fp->f_data; 3000 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3001 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p); 3002 VOP_UNLOCK(vp, 0); 3003 FILE_UNUSE(fp, p); 3004 return (error); 3005 } 3006 3007 /* 3008 * Rename files, (standard) BSD semantics frontend. 3009 */ 3010 /* ARGSUSED */ 3011 int 3012 sys_rename(struct lwp *l, void *v, register_t *retval) 3013 { 3014 struct sys_rename_args /* { 3015 syscallarg(const char *) from; 3016 syscallarg(const char *) to; 3017 } */ *uap = v; 3018 struct proc *p = l->l_proc; 3019 3020 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 3021 } 3022 3023 /* 3024 * Rename files, POSIX semantics frontend. 3025 */ 3026 /* ARGSUSED */ 3027 int 3028 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3029 { 3030 struct sys___posix_rename_args /* { 3031 syscallarg(const char *) from; 3032 syscallarg(const char *) to; 3033 } */ *uap = v; 3034 struct proc *p = l->l_proc; 3035 3036 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 3037 } 3038 3039 /* 3040 * Rename files. Source and destination must either both be directories, 3041 * or both not be directories. If target is a directory, it must be empty. 3042 * If `from' and `to' refer to the same object, the value of the `retain' 3043 * argument is used to determine whether `from' will be 3044 * 3045 * (retain == 0) deleted unless `from' and `to' refer to the same 3046 * object in the file system's name space (BSD). 3047 * (retain == 1) always retained (POSIX). 3048 */ 3049 static int 3050 rename_files(const char *from, const char *to, struct proc *p, int retain) 3051 { 3052 struct mount *mp = NULL; 3053 struct vnode *tvp, *fvp, *tdvp; 3054 struct nameidata fromnd, tond; 3055 int error; 3056 3057 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3058 from, p); 3059 if ((error = namei(&fromnd)) != 0) 3060 return (error); 3061 fvp = fromnd.ni_vp; 3062 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3063 if (error != 0) { 3064 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3065 vrele(fromnd.ni_dvp); 3066 vrele(fvp); 3067 if (fromnd.ni_startdir) 3068 vrele(fromnd.ni_startdir); 3069 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3070 return (error); 3071 } 3072 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3073 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p); 3074 if ((error = namei(&tond)) != 0) { 3075 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3076 vrele(fromnd.ni_dvp); 3077 vrele(fvp); 3078 goto out1; 3079 } 3080 tdvp = tond.ni_dvp; 3081 tvp = tond.ni_vp; 3082 3083 if (tvp != NULL) { 3084 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3085 error = ENOTDIR; 3086 goto out; 3087 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3088 error = EISDIR; 3089 goto out; 3090 } 3091 } 3092 3093 if (fvp == tdvp) 3094 error = EINVAL; 3095 3096 /* 3097 * Source and destination refer to the same object. 3098 */ 3099 if (fvp == tvp) { 3100 if (retain) 3101 error = -1; 3102 else if (fromnd.ni_dvp == tdvp && 3103 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3104 !memcmp(fromnd.ni_cnd.cn_nameptr, 3105 tond.ni_cnd.cn_nameptr, 3106 fromnd.ni_cnd.cn_namelen)) 3107 error = -1; 3108 } 3109 3110 #ifdef VERIFIED_EXEC 3111 if (!error) 3112 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp); 3113 #endif /* VERIFIED_EXEC */ 3114 3115 out: 3116 if (!error) { 3117 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 3118 if (fromnd.ni_dvp != tdvp) 3119 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3120 if (tvp) { 3121 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 3122 } 3123 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3124 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3125 } else { 3126 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3127 if (tdvp == tvp) 3128 vrele(tdvp); 3129 else 3130 vput(tdvp); 3131 if (tvp) 3132 vput(tvp); 3133 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3134 vrele(fromnd.ni_dvp); 3135 vrele(fvp); 3136 } 3137 vrele(tond.ni_startdir); 3138 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3139 out1: 3140 vn_finished_write(mp, 0); 3141 if (fromnd.ni_startdir) 3142 vrele(fromnd.ni_startdir); 3143 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3144 return (error == -1 ? 0 : error); 3145 } 3146 3147 /* 3148 * Make a directory file. 3149 */ 3150 /* ARGSUSED */ 3151 int 3152 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3153 { 3154 struct sys_mkdir_args /* { 3155 syscallarg(const char *) path; 3156 syscallarg(int) mode; 3157 } */ *uap = v; 3158 struct proc *p = l->l_proc; 3159 struct mount *mp; 3160 struct vnode *vp; 3161 struct vattr vattr; 3162 int error; 3163 struct nameidata nd; 3164 3165 restart: 3166 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3167 SCARG(uap, path), p); 3168 if ((error = namei(&nd)) != 0) 3169 return (error); 3170 vp = nd.ni_vp; 3171 if (vp != NULL) { 3172 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3173 if (nd.ni_dvp == vp) 3174 vrele(nd.ni_dvp); 3175 else 3176 vput(nd.ni_dvp); 3177 vrele(vp); 3178 return (EEXIST); 3179 } 3180 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3181 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3182 if (nd.ni_dvp == vp) 3183 vrele(nd.ni_dvp); 3184 else 3185 vput(nd.ni_dvp); 3186 if ((error = vn_start_write(NULL, &mp, 3187 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3188 return (error); 3189 goto restart; 3190 } 3191 VATTR_NULL(&vattr); 3192 vattr.va_type = VDIR; 3193 vattr.va_mode = 3194 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3195 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3196 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3197 if (!error) 3198 vput(nd.ni_vp); 3199 vn_finished_write(mp, 0); 3200 return (error); 3201 } 3202 3203 /* 3204 * Remove a directory file. 3205 */ 3206 /* ARGSUSED */ 3207 int 3208 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3209 { 3210 struct sys_rmdir_args /* { 3211 syscallarg(const char *) path; 3212 } */ *uap = v; 3213 struct proc *p = l->l_proc; 3214 struct mount *mp; 3215 struct vnode *vp; 3216 int error; 3217 struct nameidata nd; 3218 3219 restart: 3220 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3221 SCARG(uap, path), p); 3222 if ((error = namei(&nd)) != 0) 3223 return (error); 3224 vp = nd.ni_vp; 3225 if (vp->v_type != VDIR) { 3226 error = ENOTDIR; 3227 goto out; 3228 } 3229 /* 3230 * No rmdir "." please. 3231 */ 3232 if (nd.ni_dvp == vp) { 3233 error = EINVAL; 3234 goto out; 3235 } 3236 /* 3237 * The root of a mounted filesystem cannot be deleted. 3238 */ 3239 if (vp->v_flag & VROOT) { 3240 error = EBUSY; 3241 goto out; 3242 } 3243 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3244 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3245 if (nd.ni_dvp == vp) 3246 vrele(nd.ni_dvp); 3247 else 3248 vput(nd.ni_dvp); 3249 vput(vp); 3250 if ((error = vn_start_write(NULL, &mp, 3251 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3252 return (error); 3253 goto restart; 3254 } 3255 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3256 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3257 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3258 vn_finished_write(mp, 0); 3259 return (error); 3260 3261 out: 3262 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3263 if (nd.ni_dvp == vp) 3264 vrele(nd.ni_dvp); 3265 else 3266 vput(nd.ni_dvp); 3267 vput(vp); 3268 return (error); 3269 } 3270 3271 /* 3272 * Read a block of directory entries in a file system independent format. 3273 */ 3274 int 3275 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3276 { 3277 struct sys___getdents30_args /* { 3278 syscallarg(int) fd; 3279 syscallarg(char *) buf; 3280 syscallarg(size_t) count; 3281 } */ *uap = v; 3282 struct proc *p = l->l_proc; 3283 struct file *fp; 3284 int error, done; 3285 3286 /* getvnode() will use the descriptor for us */ 3287 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3288 return (error); 3289 if ((fp->f_flag & FREAD) == 0) { 3290 error = EBADF; 3291 goto out; 3292 } 3293 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3294 SCARG(uap, count), &done, p, 0, 0); 3295 #ifdef KTRACE 3296 if (!error && KTRPOINT(p, KTR_GENIO)) { 3297 struct iovec iov; 3298 iov.iov_base = SCARG(uap, buf); 3299 iov.iov_len = done; 3300 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3301 } 3302 #endif 3303 *retval = done; 3304 out: 3305 FILE_UNUSE(fp, p); 3306 return (error); 3307 } 3308 3309 /* 3310 * Set the mode mask for creation of filesystem nodes. 3311 */ 3312 int 3313 sys_umask(struct lwp *l, void *v, register_t *retval) 3314 { 3315 struct sys_umask_args /* { 3316 syscallarg(mode_t) newmask; 3317 } */ *uap = v; 3318 struct proc *p = l->l_proc; 3319 struct cwdinfo *cwdi; 3320 3321 cwdi = p->p_cwdi; 3322 *retval = cwdi->cwdi_cmask; 3323 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3324 return (0); 3325 } 3326 3327 /* 3328 * Void all references to file by ripping underlying filesystem 3329 * away from vnode. 3330 */ 3331 /* ARGSUSED */ 3332 int 3333 sys_revoke(struct lwp *l, void *v, register_t *retval) 3334 { 3335 struct sys_revoke_args /* { 3336 syscallarg(const char *) path; 3337 } */ *uap = v; 3338 struct proc *p = l->l_proc; 3339 struct mount *mp; 3340 struct vnode *vp; 3341 struct vattr vattr; 3342 int error; 3343 struct nameidata nd; 3344 3345 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3346 if ((error = namei(&nd)) != 0) 3347 return (error); 3348 vp = nd.ni_vp; 3349 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 3350 goto out; 3351 if (p->p_ucred->cr_uid != vattr.va_uid && 3352 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3353 goto out; 3354 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3355 goto out; 3356 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3357 VOP_REVOKE(vp, REVOKEALL); 3358 vn_finished_write(mp, 0); 3359 out: 3360 vrele(vp); 3361 return (error); 3362 } 3363 3364 /* 3365 * Convert a user file descriptor to a kernel file entry. 3366 */ 3367 int 3368 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3369 { 3370 struct vnode *vp; 3371 struct file *fp; 3372 3373 if ((fp = fd_getfile(fdp, fd)) == NULL) 3374 return (EBADF); 3375 3376 FILE_USE(fp); 3377 3378 if (fp->f_type != DTYPE_VNODE) { 3379 FILE_UNUSE(fp, NULL); 3380 return (EINVAL); 3381 } 3382 3383 vp = (struct vnode *)fp->f_data; 3384 if (vp->v_type == VBAD) { 3385 FILE_UNUSE(fp, NULL); 3386 return (EBADF); 3387 } 3388 3389 *fpp = fp; 3390 return (0); 3391 } 3392
Indexes created Thu Sep 25 16:09:42 GMT 2025