vfs_syscalls.c revision 1.221
1 /* $NetBSD: vfs_syscalls.c,v 1.221 2005/06/05 23:47:48 thorpej 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 1.221 2005/06/05 23:47:48 thorpej Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_ktrace.h" 45 #include "fss.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/namei.h> 50 #include <sys/filedesc.h> 51 #include <sys/kernel.h> 52 #include <sys/file.h> 53 #include <sys/stat.h> 54 #include <sys/vnode.h> 55 #include <sys/mount.h> 56 #include <sys/proc.h> 57 #include <sys/uio.h> 58 #include <sys/malloc.h> 59 #include <sys/dirent.h> 60 #include <sys/extattr.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 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 number. 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 mp->mnt_flag |= SCARG(uap, flags) & 337 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 338 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 339 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 340 MNT_IGNORE); 341 } 342 /* 343 * Mount the filesystem. 344 */ 345 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 346 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 347 if (mp->mnt_iflag & IMNT_WANTRDWR) 348 mp->mnt_flag &= ~MNT_RDONLY; 349 if (error) 350 mp->mnt_flag = flag; 351 mp->mnt_flag &=~ 352 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 353 mp->mnt_iflag &=~ IMNT_WANTRDWR; 354 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 355 if (mp->mnt_syncer == NULL) 356 error = vfs_allocate_syncvnode(mp); 357 } else { 358 if (mp->mnt_syncer != NULL) 359 vfs_deallocate_syncvnode(mp); 360 } 361 vfs_unbusy(mp); 362 VOP_UNLOCK(vp, 0); 363 vrele(vp); 364 return (error); 365 } 366 /* 367 * Put the new filesystem on the mount list after root. 368 */ 369 cache_purge(vp); 370 if (!error) { 371 mp->mnt_flag &=~ 372 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 373 mp->mnt_iflag &=~ IMNT_WANTRDWR; 374 vp->v_mountedhere = mp; 375 simple_lock(&mountlist_slock); 376 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 377 simple_unlock(&mountlist_slock); 378 checkdirs(vp); 379 VOP_UNLOCK(vp, 0); 380 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 381 error = vfs_allocate_syncvnode(mp); 382 vfs_unbusy(mp); 383 (void) VFS_STATVFS(mp, &mp->mnt_stat, p); 384 if ((error = VFS_START(mp, 0, p))) 385 vrele(vp); 386 } else { 387 vp->v_mountedhere = (struct mount *)0; 388 vfs->vfs_refcount--; 389 vfs_unbusy(mp); 390 free(mp, M_MOUNT); 391 vput(vp); 392 } 393 return (error); 394 } 395 396 /* 397 * Scan all active processes to see if any of them have a current 398 * or root directory onto which the new filesystem has just been 399 * mounted. If so, replace them with the new mount point. 400 */ 401 void 402 checkdirs(struct vnode *olddp) 403 { 404 struct cwdinfo *cwdi; 405 struct vnode *newdp; 406 struct proc *p; 407 408 if (olddp->v_usecount == 1) 409 return; 410 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 411 panic("mount: lost mount"); 412 proclist_lock_read(); 413 PROCLIST_FOREACH(p, &allproc) { 414 cwdi = p->p_cwdi; 415 if (!cwdi) 416 continue; 417 if (cwdi->cwdi_cdir == olddp) { 418 vrele(cwdi->cwdi_cdir); 419 VREF(newdp); 420 cwdi->cwdi_cdir = newdp; 421 } 422 if (cwdi->cwdi_rdir == olddp) { 423 vrele(cwdi->cwdi_rdir); 424 VREF(newdp); 425 cwdi->cwdi_rdir = newdp; 426 } 427 } 428 proclist_unlock_read(); 429 if (rootvnode == olddp) { 430 vrele(rootvnode); 431 VREF(newdp); 432 rootvnode = newdp; 433 } 434 vput(newdp); 435 } 436 437 /* 438 * Unmount a file system. 439 * 440 * Note: unmount takes a path to the vnode mounted on as argument, 441 * not special file (as before). 442 */ 443 /* ARGSUSED */ 444 int 445 sys_unmount(struct lwp *l, void *v, register_t *retval) 446 { 447 struct sys_unmount_args /* { 448 syscallarg(const char *) path; 449 syscallarg(int) flags; 450 } */ *uap = v; 451 struct proc *p = l->l_proc; 452 struct vnode *vp; 453 struct mount *mp; 454 int error; 455 struct nameidata nd; 456 457 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 458 SCARG(uap, path), p); 459 if ((error = namei(&nd)) != 0) 460 return (error); 461 vp = nd.ni_vp; 462 mp = vp->v_mount; 463 464 /* 465 * Only root, or the user that did the original mount is 466 * permitted to unmount this filesystem. 467 */ 468 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 469 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 470 vput(vp); 471 return (error); 472 } 473 474 /* 475 * Don't allow unmounting the root file system. 476 */ 477 if (mp->mnt_flag & MNT_ROOTFS) { 478 vput(vp); 479 return (EINVAL); 480 } 481 482 /* 483 * Must be the root of the filesystem 484 */ 485 if ((vp->v_flag & VROOT) == 0) { 486 vput(vp); 487 return (EINVAL); 488 } 489 vput(vp); 490 491 /* 492 * XXX Freeze syncer. Must do this before locking the 493 * mount point. See dounmount() for details. 494 */ 495 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 496 497 if (vfs_busy(mp, 0, 0)) { 498 lockmgr(&syncer_lock, LK_RELEASE, NULL); 499 return (EBUSY); 500 } 501 502 return (dounmount(mp, SCARG(uap, flags), p)); 503 } 504 505 /* 506 * Do the actual file system unmount. File system is assumed to have been 507 * marked busy by the caller. 508 */ 509 int 510 dounmount(struct mount *mp, int flags, struct proc *p) 511 { 512 struct vnode *coveredvp; 513 int error; 514 int async; 515 int used_syncer; 516 517 simple_lock(&mountlist_slock); 518 vfs_unbusy(mp); 519 used_syncer = (mp->mnt_syncer != NULL); 520 521 /* 522 * XXX Syncer must be frozen when we get here. This should really 523 * be done on a per-mountpoint basis, but especially the softdep 524 * code possibly called from the syncer doens't exactly work on a 525 * per-mountpoint basis, so the softdep code would become a maze 526 * of vfs_busy() calls. 527 * 528 * The caller of dounmount() must acquire syncer_lock because 529 * the syncer itself acquires locks in syncer_lock -> vfs_busy 530 * order, and we must preserve that order to avoid deadlock. 531 * 532 * So, if the file system did not use the syncer, now is 533 * the time to release the syncer_lock. 534 */ 535 if (used_syncer == 0) 536 lockmgr(&syncer_lock, LK_RELEASE, NULL); 537 538 mp->mnt_iflag |= IMNT_UNMOUNT; 539 mp->mnt_unmounter = p; 540 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 541 vn_start_write(NULL, &mp, V_WAIT); 542 543 if (mp->mnt_flag & MNT_EXPUBLIC) 544 vfs_setpublicfs(NULL, NULL, NULL); 545 async = mp->mnt_flag & MNT_ASYNC; 546 mp->mnt_flag &= ~MNT_ASYNC; 547 cache_purgevfs(mp); /* remove cache entries for this file sys */ 548 if (mp->mnt_syncer != NULL) 549 vfs_deallocate_syncvnode(mp); 550 error = 0; 551 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 552 #if NFSS > 0 553 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 554 #endif 555 if (error == 0) 556 error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p); 557 } 558 if (error == 0 || (flags & MNT_FORCE)) 559 error = VFS_UNMOUNT(mp, flags, p); 560 vn_finished_write(mp, 0); 561 simple_lock(&mountlist_slock); 562 if (error) { 563 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 564 (void) vfs_allocate_syncvnode(mp); 565 mp->mnt_iflag &= ~IMNT_UNMOUNT; 566 mp->mnt_unmounter = NULL; 567 mp->mnt_flag |= async; 568 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 569 &mountlist_slock); 570 if (used_syncer) 571 lockmgr(&syncer_lock, LK_RELEASE, NULL); 572 simple_lock(&mp->mnt_slock); 573 while (mp->mnt_wcnt > 0) { 574 wakeup(mp); 575 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 576 0, &mp->mnt_slock); 577 } 578 simple_unlock(&mp->mnt_slock); 579 return (error); 580 } 581 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 582 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 583 coveredvp->v_mountedhere = NULL; 584 vrele(coveredvp); 585 } 586 mp->mnt_op->vfs_refcount--; 587 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 588 panic("unmount: dangling vnode"); 589 mp->mnt_iflag |= IMNT_GONE; 590 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 591 if (used_syncer) 592 lockmgr(&syncer_lock, LK_RELEASE, NULL); 593 simple_lock(&mp->mnt_slock); 594 while (mp->mnt_wcnt > 0) { 595 wakeup(mp); 596 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 597 } 598 simple_unlock(&mp->mnt_slock); 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 memset(&fh, 0, sizeof(fh)); 1195 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1196 error = VFS_VPTOFH(vp, &fh.fh_fid); 1197 vput(vp); 1198 if (error) 1199 return (error); 1200 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1201 return (error); 1202 } 1203 1204 /* 1205 * Open a file given a file handle. 1206 * 1207 * Check permissions, allocate an open file structure, 1208 * and call the device open routine if any. 1209 */ 1210 int 1211 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1212 { 1213 struct sys_fhopen_args /* { 1214 syscallarg(const fhandle_t *) fhp; 1215 syscallarg(int) flags; 1216 } */ *uap = v; 1217 struct proc *p = l->l_proc; 1218 struct filedesc *fdp = p->p_fd; 1219 struct file *fp; 1220 struct vnode *vp = NULL; 1221 struct mount *mp; 1222 struct ucred *cred = p->p_ucred; 1223 int flags; 1224 struct file *nfp; 1225 int type, indx, error=0; 1226 struct flock lf; 1227 struct vattr va; 1228 fhandle_t fh; 1229 1230 /* 1231 * Must be super user 1232 */ 1233 if ((error = suser(p->p_ucred, &p->p_acflag))) 1234 return (error); 1235 1236 flags = FFLAGS(SCARG(uap, flags)); 1237 if ((flags & (FREAD | FWRITE)) == 0) 1238 return (EINVAL); 1239 if ((flags & O_CREAT)) 1240 return (EINVAL); 1241 /* falloc() will use the file descriptor for us */ 1242 if ((error = falloc(p, &nfp, &indx)) != 0) 1243 return (error); 1244 fp = nfp; 1245 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1246 goto bad; 1247 1248 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1249 error = ESTALE; 1250 goto bad; 1251 } 1252 1253 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1254 vp = NULL; /* most likely unnecessary sanity for bad: */ 1255 goto bad; 1256 } 1257 1258 /* Now do an effective vn_open */ 1259 1260 if (vp->v_type == VSOCK) { 1261 error = EOPNOTSUPP; 1262 goto bad; 1263 } 1264 if (flags & FREAD) { 1265 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 1266 goto bad; 1267 } 1268 if (flags & (FWRITE | O_TRUNC)) { 1269 if (vp->v_type == VDIR) { 1270 error = EISDIR; 1271 goto bad; 1272 } 1273 if ((error = vn_writechk(vp)) != 0 || 1274 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 1275 goto bad; 1276 } 1277 if (flags & O_TRUNC) { 1278 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1279 goto bad; 1280 VOP_UNLOCK(vp, 0); /* XXX */ 1281 VOP_LEASE(vp, p, cred, LEASE_WRITE); 1282 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1283 VATTR_NULL(&va); 1284 va.va_size = 0; 1285 error = VOP_SETATTR(vp, &va, cred, p); 1286 vn_finished_write(mp, 0); 1287 if (error) 1288 goto bad; 1289 } 1290 if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) 1291 goto bad; 1292 if (vp->v_type == VREG && 1293 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1294 error = EIO; 1295 goto bad; 1296 } 1297 if (flags & FWRITE) 1298 vp->v_writecount++; 1299 1300 /* done with modified vn_open, now finish what sys_open does. */ 1301 1302 fp->f_flag = flags & FMASK; 1303 fp->f_type = DTYPE_VNODE; 1304 fp->f_ops = &vnops; 1305 fp->f_data = vp; 1306 if (flags & (O_EXLOCK | O_SHLOCK)) { 1307 lf.l_whence = SEEK_SET; 1308 lf.l_start = 0; 1309 lf.l_len = 0; 1310 if (flags & O_EXLOCK) 1311 lf.l_type = F_WRLCK; 1312 else 1313 lf.l_type = F_RDLCK; 1314 type = F_FLOCK; 1315 if ((flags & FNONBLOCK) == 0) 1316 type |= F_WAIT; 1317 VOP_UNLOCK(vp, 0); 1318 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1319 if (error) { 1320 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 1321 FILE_UNUSE(fp, p); 1322 ffree(fp); 1323 fdremove(fdp, indx); 1324 return (error); 1325 } 1326 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1327 fp->f_flag |= FHASLOCK; 1328 } 1329 VOP_UNLOCK(vp, 0); 1330 *retval = indx; 1331 FILE_SET_MATURE(fp); 1332 FILE_UNUSE(fp, p); 1333 return (0); 1334 1335 bad: 1336 FILE_UNUSE(fp, p); 1337 ffree(fp); 1338 fdremove(fdp, indx); 1339 if (vp != NULL) 1340 vput(vp); 1341 return (error); 1342 } 1343 1344 /* ARGSUSED */ 1345 int 1346 sys_fhstat(struct lwp *l, void *v, register_t *retval) 1347 { 1348 struct sys_fhstat_args /* { 1349 syscallarg(const fhandle_t *) fhp; 1350 syscallarg(struct stat *) sb; 1351 } */ *uap = v; 1352 struct proc *p = l->l_proc; 1353 struct stat sb; 1354 int error; 1355 fhandle_t fh; 1356 struct mount *mp; 1357 struct vnode *vp; 1358 1359 /* 1360 * Must be super user 1361 */ 1362 if ((error = suser(p->p_ucred, &p->p_acflag))) 1363 return (error); 1364 1365 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1366 return (error); 1367 1368 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1369 return (ESTALE); 1370 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1371 return (error); 1372 error = vn_stat(vp, &sb, p); 1373 vput(vp); 1374 if (error) 1375 return (error); 1376 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1377 return (error); 1378 } 1379 1380 /* ARGSUSED */ 1381 int 1382 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1383 { 1384 struct sys_fhstatvfs1_args /* { 1385 syscallarg(const fhandle_t *) fhp; 1386 syscallarg(struct statvfs *) buf; 1387 syscallarg(int) flags; 1388 } */ *uap = v; 1389 struct proc *p = l->l_proc; 1390 struct statvfs sbuf; 1391 fhandle_t fh; 1392 struct mount *mp; 1393 struct vnode *vp; 1394 int error; 1395 1396 /* 1397 * Must be super user 1398 */ 1399 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1400 return error; 1401 1402 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1403 return error; 1404 1405 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1406 return ESTALE; 1407 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1408 return error; 1409 1410 mp = vp->v_mount; 1411 if ((error = dostatvfs(mp, &sbuf, p, SCARG(uap, flags), 1)) != 0) { 1412 vput(vp); 1413 return error; 1414 } 1415 vput(vp); 1416 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1417 } 1418 1419 /* 1420 * Create a special file. 1421 */ 1422 /* ARGSUSED */ 1423 int 1424 sys_mknod(struct lwp *l, void *v, register_t *retval) 1425 { 1426 struct sys_mknod_args /* { 1427 syscallarg(const char *) path; 1428 syscallarg(int) mode; 1429 syscallarg(int) dev; 1430 } */ *uap = v; 1431 struct proc *p = l->l_proc; 1432 struct vnode *vp; 1433 struct mount *mp; 1434 struct vattr vattr; 1435 int error; 1436 int whiteout = 0; 1437 struct nameidata nd; 1438 1439 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1440 return (error); 1441 restart: 1442 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1443 if ((error = namei(&nd)) != 0) 1444 return (error); 1445 vp = nd.ni_vp; 1446 if (vp != NULL) 1447 error = EEXIST; 1448 else { 1449 VATTR_NULL(&vattr); 1450 vattr.va_mode = 1451 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1452 vattr.va_rdev = SCARG(uap, dev); 1453 whiteout = 0; 1454 1455 switch (SCARG(uap, mode) & S_IFMT) { 1456 case S_IFMT: /* used by badsect to flag bad sectors */ 1457 vattr.va_type = VBAD; 1458 break; 1459 case S_IFCHR: 1460 vattr.va_type = VCHR; 1461 break; 1462 case S_IFBLK: 1463 vattr.va_type = VBLK; 1464 break; 1465 case S_IFWHT: 1466 whiteout = 1; 1467 break; 1468 default: 1469 error = EINVAL; 1470 break; 1471 } 1472 } 1473 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1474 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1475 if (nd.ni_dvp == vp) 1476 vrele(nd.ni_dvp); 1477 else 1478 vput(nd.ni_dvp); 1479 if (vp) 1480 vrele(vp); 1481 if ((error = vn_start_write(NULL, &mp, 1482 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1483 return (error); 1484 goto restart; 1485 } 1486 if (!error) { 1487 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1488 if (whiteout) { 1489 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1490 if (error) 1491 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1492 vput(nd.ni_dvp); 1493 } else { 1494 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1495 &nd.ni_cnd, &vattr); 1496 if (error == 0) 1497 vput(nd.ni_vp); 1498 } 1499 } else { 1500 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1501 if (nd.ni_dvp == vp) 1502 vrele(nd.ni_dvp); 1503 else 1504 vput(nd.ni_dvp); 1505 if (vp) 1506 vrele(vp); 1507 } 1508 vn_finished_write(mp, 0); 1509 return (error); 1510 } 1511 1512 /* 1513 * Create a named pipe. 1514 */ 1515 /* ARGSUSED */ 1516 int 1517 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1518 { 1519 struct sys_mkfifo_args /* { 1520 syscallarg(const char *) path; 1521 syscallarg(int) mode; 1522 } */ *uap = v; 1523 struct proc *p = l->l_proc; 1524 struct mount *mp; 1525 struct vattr vattr; 1526 int error; 1527 struct nameidata nd; 1528 1529 restart: 1530 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1531 if ((error = namei(&nd)) != 0) 1532 return (error); 1533 if (nd.ni_vp != NULL) { 1534 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1535 if (nd.ni_dvp == nd.ni_vp) 1536 vrele(nd.ni_dvp); 1537 else 1538 vput(nd.ni_dvp); 1539 vrele(nd.ni_vp); 1540 return (EEXIST); 1541 } 1542 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1543 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1544 if (nd.ni_dvp == nd.ni_vp) 1545 vrele(nd.ni_dvp); 1546 else 1547 vput(nd.ni_dvp); 1548 if (nd.ni_vp) 1549 vrele(nd.ni_vp); 1550 if ((error = vn_start_write(NULL, &mp, 1551 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1552 return (error); 1553 goto restart; 1554 } 1555 VATTR_NULL(&vattr); 1556 vattr.va_type = VFIFO; 1557 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1558 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1559 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1560 if (error == 0) 1561 vput(nd.ni_vp); 1562 vn_finished_write(mp, 0); 1563 return (error); 1564 } 1565 1566 /* 1567 * Make a hard file link. 1568 */ 1569 /* ARGSUSED */ 1570 int 1571 sys_link(struct lwp *l, void *v, register_t *retval) 1572 { 1573 struct sys_link_args /* { 1574 syscallarg(const char *) path; 1575 syscallarg(const char *) link; 1576 } */ *uap = v; 1577 struct proc *p = l->l_proc; 1578 struct vnode *vp; 1579 struct mount *mp; 1580 struct nameidata nd; 1581 int error; 1582 1583 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1584 if ((error = namei(&nd)) != 0) 1585 return (error); 1586 vp = nd.ni_vp; 1587 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1588 vrele(vp); 1589 return (error); 1590 } 1591 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1592 if ((error = namei(&nd)) != 0) 1593 goto out; 1594 if (nd.ni_vp) { 1595 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1596 if (nd.ni_dvp == nd.ni_vp) 1597 vrele(nd.ni_dvp); 1598 else 1599 vput(nd.ni_dvp); 1600 vrele(nd.ni_vp); 1601 error = EEXIST; 1602 goto out; 1603 } 1604 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1605 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1606 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1607 out: 1608 vrele(vp); 1609 vn_finished_write(mp, 0); 1610 return (error); 1611 } 1612 1613 /* 1614 * Make a symbolic link. 1615 */ 1616 /* ARGSUSED */ 1617 int 1618 sys_symlink(struct lwp *l, void *v, register_t *retval) 1619 { 1620 struct sys_symlink_args /* { 1621 syscallarg(const char *) path; 1622 syscallarg(const char *) link; 1623 } */ *uap = v; 1624 struct proc *p = l->l_proc; 1625 struct mount *mp; 1626 struct vattr vattr; 1627 char *path; 1628 int error; 1629 struct nameidata nd; 1630 1631 path = PNBUF_GET(); 1632 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1633 if (error) 1634 goto out; 1635 restart: 1636 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1637 if ((error = namei(&nd)) != 0) 1638 goto out; 1639 if (nd.ni_vp) { 1640 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1641 if (nd.ni_dvp == nd.ni_vp) 1642 vrele(nd.ni_dvp); 1643 else 1644 vput(nd.ni_dvp); 1645 vrele(nd.ni_vp); 1646 error = EEXIST; 1647 goto out; 1648 } 1649 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1650 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1651 if (nd.ni_dvp == nd.ni_vp) 1652 vrele(nd.ni_dvp); 1653 else 1654 vput(nd.ni_dvp); 1655 if ((error = vn_start_write(NULL, &mp, 1656 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1657 return (error); 1658 goto restart; 1659 } 1660 VATTR_NULL(&vattr); 1661 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1662 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1663 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1664 if (error == 0) 1665 vput(nd.ni_vp); 1666 vn_finished_write(mp, 0); 1667 out: 1668 PNBUF_PUT(path); 1669 return (error); 1670 } 1671 1672 /* 1673 * Delete a whiteout from the filesystem. 1674 */ 1675 /* ARGSUSED */ 1676 int 1677 sys_undelete(struct lwp *l, void *v, register_t *retval) 1678 { 1679 struct sys_undelete_args /* { 1680 syscallarg(const char *) path; 1681 } */ *uap = v; 1682 struct proc *p = l->l_proc; 1683 int error; 1684 struct mount *mp; 1685 struct nameidata nd; 1686 1687 restart: 1688 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1689 SCARG(uap, path), p); 1690 error = namei(&nd); 1691 if (error) 1692 return (error); 1693 1694 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1695 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1696 if (nd.ni_dvp == nd.ni_vp) 1697 vrele(nd.ni_dvp); 1698 else 1699 vput(nd.ni_dvp); 1700 if (nd.ni_vp) 1701 vrele(nd.ni_vp); 1702 return (EEXIST); 1703 } 1704 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1705 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1706 if (nd.ni_dvp == nd.ni_vp) 1707 vrele(nd.ni_dvp); 1708 else 1709 vput(nd.ni_dvp); 1710 if ((error = vn_start_write(NULL, &mp, 1711 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1712 return (error); 1713 goto restart; 1714 } 1715 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1716 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1717 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1718 vput(nd.ni_dvp); 1719 vn_finished_write(mp, 0); 1720 return (error); 1721 } 1722 1723 /* 1724 * Delete a name from the filesystem. 1725 */ 1726 /* ARGSUSED */ 1727 int 1728 sys_unlink(struct lwp *l, void *v, register_t *retval) 1729 { 1730 struct sys_unlink_args /* { 1731 syscallarg(const char *) path; 1732 } */ *uap = v; 1733 struct proc *p = l->l_proc; 1734 struct mount *mp; 1735 struct vnode *vp; 1736 int error; 1737 struct nameidata nd; 1738 1739 restart: 1740 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1741 SCARG(uap, path), p); 1742 if ((error = namei(&nd)) != 0) 1743 return (error); 1744 vp = nd.ni_vp; 1745 1746 /* 1747 * The root of a mounted filesystem cannot be deleted. 1748 */ 1749 if (vp->v_flag & VROOT) { 1750 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1751 if (nd.ni_dvp == vp) 1752 vrele(nd.ni_dvp); 1753 else 1754 vput(nd.ni_dvp); 1755 vput(vp); 1756 error = EBUSY; 1757 goto out; 1758 } 1759 1760 /* 1761 * Remove the fingerprint from the list if there was one. 1762 */ 1763 #ifdef VERIFIED_EXEC 1764 if ((error = veriexec_removechk(p, vp, nd.ni_dirp)) != 0) { 1765 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1766 if (nd.ni_dvp == vp) 1767 vrele(nd.ni_dvp); 1768 else 1769 vput(nd.ni_dvp); 1770 vput(vp); 1771 goto out; 1772 } 1773 #endif 1774 1775 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1776 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1777 if (nd.ni_dvp == vp) 1778 vrele(nd.ni_dvp); 1779 else 1780 vput(nd.ni_dvp); 1781 vput(vp); 1782 if ((error = vn_start_write(NULL, &mp, 1783 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1784 return (error); 1785 goto restart; 1786 } 1787 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1788 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1789 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1790 vn_finished_write(mp, 0); 1791 out: 1792 return (error); 1793 } 1794 1795 /* 1796 * Reposition read/write file offset. 1797 */ 1798 int 1799 sys_lseek(struct lwp *l, void *v, register_t *retval) 1800 { 1801 struct sys_lseek_args /* { 1802 syscallarg(int) fd; 1803 syscallarg(int) pad; 1804 syscallarg(off_t) offset; 1805 syscallarg(int) whence; 1806 } */ *uap = v; 1807 struct proc *p = l->l_proc; 1808 struct ucred *cred = p->p_ucred; 1809 struct filedesc *fdp = p->p_fd; 1810 struct file *fp; 1811 struct vnode *vp; 1812 struct vattr vattr; 1813 off_t newoff; 1814 int error; 1815 1816 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1817 return (EBADF); 1818 1819 FILE_USE(fp); 1820 1821 vp = (struct vnode *)fp->f_data; 1822 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1823 error = ESPIPE; 1824 goto out; 1825 } 1826 1827 switch (SCARG(uap, whence)) { 1828 case SEEK_CUR: 1829 newoff = fp->f_offset + SCARG(uap, offset); 1830 break; 1831 case SEEK_END: 1832 error = VOP_GETATTR(vp, &vattr, cred, p); 1833 if (error) 1834 goto out; 1835 newoff = SCARG(uap, offset) + vattr.va_size; 1836 break; 1837 case SEEK_SET: 1838 newoff = SCARG(uap, offset); 1839 break; 1840 default: 1841 error = EINVAL; 1842 goto out; 1843 } 1844 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1845 goto out; 1846 1847 *(off_t *)retval = fp->f_offset = newoff; 1848 out: 1849 FILE_UNUSE(fp, p); 1850 return (error); 1851 } 1852 1853 /* 1854 * Positional read system call. 1855 */ 1856 int 1857 sys_pread(struct lwp *l, void *v, register_t *retval) 1858 { 1859 struct sys_pread_args /* { 1860 syscallarg(int) fd; 1861 syscallarg(void *) buf; 1862 syscallarg(size_t) nbyte; 1863 syscallarg(off_t) offset; 1864 } */ *uap = v; 1865 struct proc *p = l->l_proc; 1866 struct filedesc *fdp = p->p_fd; 1867 struct file *fp; 1868 struct vnode *vp; 1869 off_t offset; 1870 int error, fd = SCARG(uap, fd); 1871 1872 if ((fp = fd_getfile(fdp, fd)) == NULL) 1873 return (EBADF); 1874 1875 if ((fp->f_flag & FREAD) == 0) { 1876 simple_unlock(&fp->f_slock); 1877 return (EBADF); 1878 } 1879 1880 FILE_USE(fp); 1881 1882 vp = (struct vnode *)fp->f_data; 1883 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1884 error = ESPIPE; 1885 goto out; 1886 } 1887 1888 offset = SCARG(uap, offset); 1889 1890 /* 1891 * XXX This works because no file systems actually 1892 * XXX take any action on the seek operation. 1893 */ 1894 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1895 goto out; 1896 1897 /* dofileread() will unuse the descriptor for us */ 1898 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1899 &offset, 0, retval)); 1900 1901 out: 1902 FILE_UNUSE(fp, p); 1903 return (error); 1904 } 1905 1906 /* 1907 * Positional scatter read system call. 1908 */ 1909 int 1910 sys_preadv(struct lwp *l, void *v, register_t *retval) 1911 { 1912 struct sys_preadv_args /* { 1913 syscallarg(int) fd; 1914 syscallarg(const struct iovec *) iovp; 1915 syscallarg(int) iovcnt; 1916 syscallarg(off_t) offset; 1917 } */ *uap = v; 1918 struct proc *p = l->l_proc; 1919 struct filedesc *fdp = p->p_fd; 1920 struct file *fp; 1921 struct vnode *vp; 1922 off_t offset; 1923 int error, fd = SCARG(uap, fd); 1924 1925 if ((fp = fd_getfile(fdp, fd)) == NULL) 1926 return (EBADF); 1927 1928 if ((fp->f_flag & FREAD) == 0) { 1929 simple_unlock(&fp->f_slock); 1930 return (EBADF); 1931 } 1932 1933 FILE_USE(fp); 1934 1935 vp = (struct vnode *)fp->f_data; 1936 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1937 error = ESPIPE; 1938 goto out; 1939 } 1940 1941 offset = SCARG(uap, offset); 1942 1943 /* 1944 * XXX This works because no file systems actually 1945 * XXX take any action on the seek operation. 1946 */ 1947 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1948 goto out; 1949 1950 /* dofilereadv() will unuse the descriptor for us */ 1951 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1952 &offset, 0, retval)); 1953 1954 out: 1955 FILE_UNUSE(fp, p); 1956 return (error); 1957 } 1958 1959 /* 1960 * Positional write system call. 1961 */ 1962 int 1963 sys_pwrite(struct lwp *l, void *v, register_t *retval) 1964 { 1965 struct sys_pwrite_args /* { 1966 syscallarg(int) fd; 1967 syscallarg(const void *) buf; 1968 syscallarg(size_t) nbyte; 1969 syscallarg(off_t) offset; 1970 } */ *uap = v; 1971 struct proc *p = l->l_proc; 1972 struct filedesc *fdp = p->p_fd; 1973 struct file *fp; 1974 struct vnode *vp; 1975 off_t offset; 1976 int error, fd = SCARG(uap, fd); 1977 1978 if ((fp = fd_getfile(fdp, fd)) == NULL) 1979 return (EBADF); 1980 1981 if ((fp->f_flag & FWRITE) == 0) { 1982 simple_unlock(&fp->f_slock); 1983 return (EBADF); 1984 } 1985 1986 FILE_USE(fp); 1987 1988 vp = (struct vnode *)fp->f_data; 1989 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1990 error = ESPIPE; 1991 goto out; 1992 } 1993 1994 offset = SCARG(uap, offset); 1995 1996 /* 1997 * XXX This works because no file systems actually 1998 * XXX take any action on the seek operation. 1999 */ 2000 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2001 goto out; 2002 2003 /* dofilewrite() will unuse the descriptor for us */ 2004 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2005 &offset, 0, retval)); 2006 2007 out: 2008 FILE_UNUSE(fp, p); 2009 return (error); 2010 } 2011 2012 /* 2013 * Positional gather write system call. 2014 */ 2015 int 2016 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2017 { 2018 struct sys_pwritev_args /* { 2019 syscallarg(int) fd; 2020 syscallarg(const struct iovec *) iovp; 2021 syscallarg(int) iovcnt; 2022 syscallarg(off_t) offset; 2023 } */ *uap = v; 2024 struct proc *p = l->l_proc; 2025 struct filedesc *fdp = p->p_fd; 2026 struct file *fp; 2027 struct vnode *vp; 2028 off_t offset; 2029 int error, fd = SCARG(uap, fd); 2030 2031 if ((fp = fd_getfile(fdp, fd)) == NULL) 2032 return (EBADF); 2033 2034 if ((fp->f_flag & FWRITE) == 0) { 2035 simple_unlock(&fp->f_slock); 2036 return (EBADF); 2037 } 2038 2039 FILE_USE(fp); 2040 2041 vp = (struct vnode *)fp->f_data; 2042 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2043 error = ESPIPE; 2044 goto out; 2045 } 2046 2047 offset = SCARG(uap, offset); 2048 2049 /* 2050 * XXX This works because no file systems actually 2051 * XXX take any action on the seek operation. 2052 */ 2053 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2054 goto out; 2055 2056 /* dofilewritev() will unuse the descriptor for us */ 2057 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2058 &offset, 0, retval)); 2059 2060 out: 2061 FILE_UNUSE(fp, p); 2062 return (error); 2063 } 2064 2065 /* 2066 * Check access permissions. 2067 */ 2068 int 2069 sys_access(struct lwp *l, void *v, register_t *retval) 2070 { 2071 struct sys_access_args /* { 2072 syscallarg(const char *) path; 2073 syscallarg(int) flags; 2074 } */ *uap = v; 2075 struct proc *p = l->l_proc; 2076 struct ucred *cred; 2077 struct vnode *vp; 2078 int error, flags; 2079 struct nameidata nd; 2080 2081 cred = crdup(p->p_ucred); 2082 cred->cr_uid = p->p_cred->p_ruid; 2083 cred->cr_gid = p->p_cred->p_rgid; 2084 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2085 SCARG(uap, path), p); 2086 /* Override default credentials */ 2087 nd.ni_cnd.cn_cred = cred; 2088 if ((error = namei(&nd)) != 0) 2089 goto out; 2090 vp = nd.ni_vp; 2091 2092 /* Flags == 0 means only check for existence. */ 2093 if (SCARG(uap, flags)) { 2094 flags = 0; 2095 if (SCARG(uap, flags) & R_OK) 2096 flags |= VREAD; 2097 if (SCARG(uap, flags) & W_OK) 2098 flags |= VWRITE; 2099 if (SCARG(uap, flags) & X_OK) 2100 flags |= VEXEC; 2101 2102 error = VOP_ACCESS(vp, flags, cred, p); 2103 if (!error && (flags & VWRITE)) 2104 error = vn_writechk(vp); 2105 } 2106 vput(vp); 2107 out: 2108 crfree(cred); 2109 return (error); 2110 } 2111 2112 /* 2113 * Get file status; this version follows links. 2114 */ 2115 /* ARGSUSED */ 2116 int 2117 sys___stat13(struct lwp *l, void *v, register_t *retval) 2118 { 2119 struct sys___stat13_args /* { 2120 syscallarg(const char *) path; 2121 syscallarg(struct stat *) ub; 2122 } */ *uap = v; 2123 struct proc *p = l->l_proc; 2124 struct stat sb; 2125 int error; 2126 struct nameidata nd; 2127 2128 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2129 SCARG(uap, path), p); 2130 if ((error = namei(&nd)) != 0) 2131 return (error); 2132 error = vn_stat(nd.ni_vp, &sb, p); 2133 vput(nd.ni_vp); 2134 if (error) 2135 return (error); 2136 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2137 return (error); 2138 } 2139 2140 /* 2141 * Get file status; this version does not follow links. 2142 */ 2143 /* ARGSUSED */ 2144 int 2145 sys___lstat13(struct lwp *l, void *v, register_t *retval) 2146 { 2147 struct sys___lstat13_args /* { 2148 syscallarg(const char *) path; 2149 syscallarg(struct stat *) ub; 2150 } */ *uap = v; 2151 struct proc *p = l->l_proc; 2152 struct stat sb; 2153 int error; 2154 struct nameidata nd; 2155 2156 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2157 SCARG(uap, path), p); 2158 if ((error = namei(&nd)) != 0) 2159 return (error); 2160 error = vn_stat(nd.ni_vp, &sb, p); 2161 vput(nd.ni_vp); 2162 if (error) 2163 return (error); 2164 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2165 return (error); 2166 } 2167 2168 /* 2169 * Get configurable pathname variables. 2170 */ 2171 /* ARGSUSED */ 2172 int 2173 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2174 { 2175 struct sys_pathconf_args /* { 2176 syscallarg(const char *) path; 2177 syscallarg(int) name; 2178 } */ *uap = v; 2179 struct proc *p = l->l_proc; 2180 int error; 2181 struct nameidata nd; 2182 2183 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2184 SCARG(uap, path), p); 2185 if ((error = namei(&nd)) != 0) 2186 return (error); 2187 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2188 vput(nd.ni_vp); 2189 return (error); 2190 } 2191 2192 /* 2193 * Return target name of a symbolic link. 2194 */ 2195 /* ARGSUSED */ 2196 int 2197 sys_readlink(struct lwp *l, void *v, register_t *retval) 2198 { 2199 struct sys_readlink_args /* { 2200 syscallarg(const char *) path; 2201 syscallarg(char *) buf; 2202 syscallarg(size_t) count; 2203 } */ *uap = v; 2204 struct proc *p = l->l_proc; 2205 struct vnode *vp; 2206 struct iovec aiov; 2207 struct uio auio; 2208 int error; 2209 struct nameidata nd; 2210 2211 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2212 SCARG(uap, path), p); 2213 if ((error = namei(&nd)) != 0) 2214 return (error); 2215 vp = nd.ni_vp; 2216 if (vp->v_type != VLNK) 2217 error = EINVAL; 2218 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2219 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 2220 aiov.iov_base = SCARG(uap, buf); 2221 aiov.iov_len = SCARG(uap, count); 2222 auio.uio_iov = &aiov; 2223 auio.uio_iovcnt = 1; 2224 auio.uio_offset = 0; 2225 auio.uio_rw = UIO_READ; 2226 auio.uio_segflg = UIO_USERSPACE; 2227 auio.uio_procp = p; 2228 auio.uio_resid = SCARG(uap, count); 2229 error = VOP_READLINK(vp, &auio, p->p_ucred); 2230 } 2231 vput(vp); 2232 *retval = SCARG(uap, count) - auio.uio_resid; 2233 return (error); 2234 } 2235 2236 /* 2237 * Change flags of a file given a path name. 2238 */ 2239 /* ARGSUSED */ 2240 int 2241 sys_chflags(struct lwp *l, void *v, register_t *retval) 2242 { 2243 struct sys_chflags_args /* { 2244 syscallarg(const char *) path; 2245 syscallarg(u_long) flags; 2246 } */ *uap = v; 2247 struct proc *p = l->l_proc; 2248 struct vnode *vp; 2249 int error; 2250 struct nameidata nd; 2251 2252 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2253 if ((error = namei(&nd)) != 0) 2254 return (error); 2255 vp = nd.ni_vp; 2256 error = change_flags(vp, SCARG(uap, flags), p); 2257 vput(vp); 2258 return (error); 2259 } 2260 2261 /* 2262 * Change flags of a file given a file descriptor. 2263 */ 2264 /* ARGSUSED */ 2265 int 2266 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2267 { 2268 struct sys_fchflags_args /* { 2269 syscallarg(int) fd; 2270 syscallarg(u_long) flags; 2271 } */ *uap = v; 2272 struct proc *p = l->l_proc; 2273 struct vnode *vp; 2274 struct file *fp; 2275 int error; 2276 2277 /* getvnode() will use the descriptor for us */ 2278 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2279 return (error); 2280 vp = (struct vnode *)fp->f_data; 2281 error = change_flags(vp, SCARG(uap, flags), p); 2282 VOP_UNLOCK(vp, 0); 2283 FILE_UNUSE(fp, p); 2284 return (error); 2285 } 2286 2287 /* 2288 * Change flags of a file given a path name; this version does 2289 * not follow links. 2290 */ 2291 int 2292 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2293 { 2294 struct sys_lchflags_args /* { 2295 syscallarg(const char *) path; 2296 syscallarg(u_long) flags; 2297 } */ *uap = v; 2298 struct proc *p = l->l_proc; 2299 struct vnode *vp; 2300 int error; 2301 struct nameidata nd; 2302 2303 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2304 if ((error = namei(&nd)) != 0) 2305 return (error); 2306 vp = nd.ni_vp; 2307 error = change_flags(vp, SCARG(uap, flags), p); 2308 vput(vp); 2309 return (error); 2310 } 2311 2312 /* 2313 * Common routine to change flags of a file. 2314 */ 2315 int 2316 change_flags(struct vnode *vp, u_long flags, struct proc *p) 2317 { 2318 struct mount *mp; 2319 struct vattr vattr; 2320 int error; 2321 2322 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2323 return (error); 2324 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2325 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2326 /* 2327 * Non-superusers cannot change the flags on devices, even if they 2328 * own them. 2329 */ 2330 if (suser(p->p_ucred, &p->p_acflag) != 0) { 2331 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2332 goto out; 2333 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2334 error = EINVAL; 2335 goto out; 2336 } 2337 } 2338 VATTR_NULL(&vattr); 2339 vattr.va_flags = flags; 2340 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2341 out: 2342 vn_finished_write(mp, 0); 2343 return (error); 2344 } 2345 2346 /* 2347 * Change mode of a file given path name; this version follows links. 2348 */ 2349 /* ARGSUSED */ 2350 int 2351 sys_chmod(struct lwp *l, void *v, register_t *retval) 2352 { 2353 struct sys_chmod_args /* { 2354 syscallarg(const char *) path; 2355 syscallarg(int) mode; 2356 } */ *uap = v; 2357 struct proc *p = l->l_proc; 2358 int error; 2359 struct nameidata nd; 2360 2361 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2362 if ((error = namei(&nd)) != 0) 2363 return (error); 2364 2365 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2366 2367 vrele(nd.ni_vp); 2368 return (error); 2369 } 2370 2371 /* 2372 * Change mode of a file given a file descriptor. 2373 */ 2374 /* ARGSUSED */ 2375 int 2376 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2377 { 2378 struct sys_fchmod_args /* { 2379 syscallarg(int) fd; 2380 syscallarg(int) mode; 2381 } */ *uap = v; 2382 struct proc *p = l->l_proc; 2383 struct file *fp; 2384 int error; 2385 2386 /* getvnode() will use the descriptor for us */ 2387 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2388 return (error); 2389 2390 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 2391 FILE_UNUSE(fp, p); 2392 return (error); 2393 } 2394 2395 /* 2396 * Change mode of a file given path name; this version does not follow links. 2397 */ 2398 /* ARGSUSED */ 2399 int 2400 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2401 { 2402 struct sys_lchmod_args /* { 2403 syscallarg(const char *) path; 2404 syscallarg(int) mode; 2405 } */ *uap = v; 2406 struct proc *p = l->l_proc; 2407 int error; 2408 struct nameidata nd; 2409 2410 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2411 if ((error = namei(&nd)) != 0) 2412 return (error); 2413 2414 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 2415 2416 vrele(nd.ni_vp); 2417 return (error); 2418 } 2419 2420 /* 2421 * Common routine to set mode given a vnode. 2422 */ 2423 static int 2424 change_mode(struct vnode *vp, int mode, struct proc *p) 2425 { 2426 struct mount *mp; 2427 struct vattr vattr; 2428 int error; 2429 2430 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2431 return (error); 2432 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2433 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2434 VATTR_NULL(&vattr); 2435 vattr.va_mode = mode & ALLPERMS; 2436 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2437 VOP_UNLOCK(vp, 0); 2438 vn_finished_write(mp, 0); 2439 return (error); 2440 } 2441 2442 /* 2443 * Set ownership given a path name; this version follows links. 2444 */ 2445 /* ARGSUSED */ 2446 int 2447 sys_chown(struct lwp *l, void *v, register_t *retval) 2448 { 2449 struct sys_chown_args /* { 2450 syscallarg(const char *) path; 2451 syscallarg(uid_t) uid; 2452 syscallarg(gid_t) gid; 2453 } */ *uap = v; 2454 struct proc *p = l->l_proc; 2455 int error; 2456 struct nameidata nd; 2457 2458 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2459 if ((error = namei(&nd)) != 0) 2460 return (error); 2461 2462 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2463 2464 vrele(nd.ni_vp); 2465 return (error); 2466 } 2467 2468 /* 2469 * Set ownership given a path name; this version follows links. 2470 * Provides POSIX semantics. 2471 */ 2472 /* ARGSUSED */ 2473 int 2474 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2475 { 2476 struct sys_chown_args /* { 2477 syscallarg(const char *) path; 2478 syscallarg(uid_t) uid; 2479 syscallarg(gid_t) gid; 2480 } */ *uap = v; 2481 struct proc *p = l->l_proc; 2482 int error; 2483 struct nameidata nd; 2484 2485 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2486 if ((error = namei(&nd)) != 0) 2487 return (error); 2488 2489 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2490 2491 vrele(nd.ni_vp); 2492 return (error); 2493 } 2494 2495 /* 2496 * Set ownership given a file descriptor. 2497 */ 2498 /* ARGSUSED */ 2499 int 2500 sys_fchown(struct lwp *l, void *v, register_t *retval) 2501 { 2502 struct sys_fchown_args /* { 2503 syscallarg(int) fd; 2504 syscallarg(uid_t) uid; 2505 syscallarg(gid_t) gid; 2506 } */ *uap = v; 2507 struct proc *p = l->l_proc; 2508 int error; 2509 struct file *fp; 2510 2511 /* getvnode() will use the descriptor for us */ 2512 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2513 return (error); 2514 2515 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2516 SCARG(uap, gid), p, 0); 2517 FILE_UNUSE(fp, p); 2518 return (error); 2519 } 2520 2521 /* 2522 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2523 */ 2524 /* ARGSUSED */ 2525 int 2526 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2527 { 2528 struct sys_fchown_args /* { 2529 syscallarg(int) fd; 2530 syscallarg(uid_t) uid; 2531 syscallarg(gid_t) gid; 2532 } */ *uap = v; 2533 struct proc *p = l->l_proc; 2534 int error; 2535 struct file *fp; 2536 2537 /* getvnode() will use the descriptor for us */ 2538 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2539 return (error); 2540 2541 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2542 SCARG(uap, gid), p, 1); 2543 FILE_UNUSE(fp, p); 2544 return (error); 2545 } 2546 2547 /* 2548 * Set ownership given a path name; this version does not follow links. 2549 */ 2550 /* ARGSUSED */ 2551 int 2552 sys_lchown(struct lwp *l, void *v, register_t *retval) 2553 { 2554 struct sys_lchown_args /* { 2555 syscallarg(const char *) path; 2556 syscallarg(uid_t) uid; 2557 syscallarg(gid_t) gid; 2558 } */ *uap = v; 2559 struct proc *p = l->l_proc; 2560 int error; 2561 struct nameidata nd; 2562 2563 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2564 if ((error = namei(&nd)) != 0) 2565 return (error); 2566 2567 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2568 2569 vrele(nd.ni_vp); 2570 return (error); 2571 } 2572 2573 /* 2574 * Set ownership given a path name; this version does not follow links. 2575 * Provides POSIX/XPG semantics. 2576 */ 2577 /* ARGSUSED */ 2578 int 2579 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2580 { 2581 struct sys_lchown_args /* { 2582 syscallarg(const char *) path; 2583 syscallarg(uid_t) uid; 2584 syscallarg(gid_t) gid; 2585 } */ *uap = v; 2586 struct proc *p = l->l_proc; 2587 int error; 2588 struct nameidata nd; 2589 2590 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2591 if ((error = namei(&nd)) != 0) 2592 return (error); 2593 2594 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2595 2596 vrele(nd.ni_vp); 2597 return (error); 2598 } 2599 2600 /* 2601 * Common routine to set ownership given a vnode. 2602 */ 2603 static int 2604 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct proc *p, 2605 int posix_semantics) 2606 { 2607 struct mount *mp; 2608 struct vattr vattr; 2609 mode_t newmode; 2610 int error; 2611 2612 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2613 return (error); 2614 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2615 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2616 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2617 goto out; 2618 2619 #define CHANGED(x) ((int)(x) != -1) 2620 newmode = vattr.va_mode; 2621 if (posix_semantics) { 2622 /* 2623 * POSIX/XPG semantics: if the caller is not the super-user, 2624 * clear set-user-id and set-group-id bits. Both POSIX and 2625 * the XPG consider the behaviour for calls by the super-user 2626 * implementation-defined; we leave the set-user-id and set- 2627 * group-id settings intact in that case. 2628 */ 2629 if (suser(p->p_ucred, NULL) != 0) 2630 newmode &= ~(S_ISUID | S_ISGID); 2631 } else { 2632 /* 2633 * NetBSD semantics: when changing owner and/or group, 2634 * clear the respective bit(s). 2635 */ 2636 if (CHANGED(uid)) 2637 newmode &= ~S_ISUID; 2638 if (CHANGED(gid)) 2639 newmode &= ~S_ISGID; 2640 } 2641 /* Update va_mode iff altered. */ 2642 if (vattr.va_mode == newmode) 2643 newmode = VNOVAL; 2644 2645 VATTR_NULL(&vattr); 2646 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2647 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2648 vattr.va_mode = newmode; 2649 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2650 #undef CHANGED 2651 2652 out: 2653 VOP_UNLOCK(vp, 0); 2654 vn_finished_write(mp, 0); 2655 return (error); 2656 } 2657 2658 /* 2659 * Set the access and modification times given a path name; this 2660 * version follows links. 2661 */ 2662 /* ARGSUSED */ 2663 int 2664 sys_utimes(struct lwp *l, void *v, register_t *retval) 2665 { 2666 struct sys_utimes_args /* { 2667 syscallarg(const char *) path; 2668 syscallarg(const struct timeval *) tptr; 2669 } */ *uap = v; 2670 struct proc *p = l->l_proc; 2671 int error; 2672 struct nameidata nd; 2673 2674 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2675 if ((error = namei(&nd)) != 0) 2676 return (error); 2677 2678 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2679 2680 vrele(nd.ni_vp); 2681 return (error); 2682 } 2683 2684 /* 2685 * Set the access and modification times given a file descriptor. 2686 */ 2687 /* ARGSUSED */ 2688 int 2689 sys_futimes(struct lwp *l, void *v, register_t *retval) 2690 { 2691 struct sys_futimes_args /* { 2692 syscallarg(int) fd; 2693 syscallarg(const struct timeval *) tptr; 2694 } */ *uap = v; 2695 struct proc *p = l->l_proc; 2696 int error; 2697 struct file *fp; 2698 2699 /* getvnode() will use the descriptor for us */ 2700 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2701 return (error); 2702 2703 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2704 FILE_UNUSE(fp, p); 2705 return (error); 2706 } 2707 2708 /* 2709 * Set the access and modification times given a path name; this 2710 * version does not follow links. 2711 */ 2712 /* ARGSUSED */ 2713 int 2714 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2715 { 2716 struct sys_lutimes_args /* { 2717 syscallarg(const char *) path; 2718 syscallarg(const struct timeval *) tptr; 2719 } */ *uap = v; 2720 struct proc *p = l->l_proc; 2721 int error; 2722 struct nameidata nd; 2723 2724 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2725 if ((error = namei(&nd)) != 0) 2726 return (error); 2727 2728 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2729 2730 vrele(nd.ni_vp); 2731 return (error); 2732 } 2733 2734 /* 2735 * Common routine to set access and modification times given a vnode. 2736 */ 2737 static int 2738 change_utimes(struct vnode *vp, const struct timeval *tptr, struct proc *p) 2739 { 2740 struct timeval tv[2]; 2741 struct mount *mp; 2742 struct vattr vattr; 2743 int error; 2744 2745 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2746 return (error); 2747 VATTR_NULL(&vattr); 2748 if (tptr == NULL) { 2749 microtime(&tv[0]); 2750 tv[1] = tv[0]; 2751 vattr.va_vaflags |= VA_UTIMES_NULL; 2752 } else { 2753 error = copyin(tptr, tv, sizeof(tv)); 2754 if (error) 2755 goto out; 2756 } 2757 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2758 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2759 vattr.va_atime.tv_sec = tv[0].tv_sec; 2760 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2761 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2762 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2763 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2764 VOP_UNLOCK(vp, 0); 2765 out: 2766 vn_finished_write(mp, 0); 2767 return (error); 2768 } 2769 2770 /* 2771 * Truncate a file given its path name. 2772 */ 2773 /* ARGSUSED */ 2774 int 2775 sys_truncate(struct lwp *l, void *v, register_t *retval) 2776 { 2777 struct sys_truncate_args /* { 2778 syscallarg(const char *) path; 2779 syscallarg(int) pad; 2780 syscallarg(off_t) length; 2781 } */ *uap = v; 2782 struct proc *p = l->l_proc; 2783 struct vnode *vp; 2784 struct mount *mp; 2785 struct vattr vattr; 2786 int error; 2787 struct nameidata nd; 2788 2789 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2790 if ((error = namei(&nd)) != 0) 2791 return (error); 2792 vp = nd.ni_vp; 2793 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2794 vrele(vp); 2795 return (error); 2796 } 2797 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2798 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2799 if (vp->v_type == VDIR) 2800 error = EISDIR; 2801 else if ((error = vn_writechk(vp)) == 0 && 2802 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2803 VATTR_NULL(&vattr); 2804 vattr.va_size = SCARG(uap, length); 2805 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2806 } 2807 vput(vp); 2808 vn_finished_write(mp, 0); 2809 return (error); 2810 } 2811 2812 /* 2813 * Truncate a file given a file descriptor. 2814 */ 2815 /* ARGSUSED */ 2816 int 2817 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2818 { 2819 struct sys_ftruncate_args /* { 2820 syscallarg(int) fd; 2821 syscallarg(int) pad; 2822 syscallarg(off_t) length; 2823 } */ *uap = v; 2824 struct proc *p = l->l_proc; 2825 struct mount *mp; 2826 struct vattr vattr; 2827 struct vnode *vp; 2828 struct file *fp; 2829 int error; 2830 2831 /* getvnode() will use the descriptor for us */ 2832 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2833 return (error); 2834 if ((fp->f_flag & FWRITE) == 0) { 2835 error = EINVAL; 2836 goto out; 2837 } 2838 vp = (struct vnode *)fp->f_data; 2839 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2840 FILE_UNUSE(fp, p); 2841 return (error); 2842 } 2843 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2844 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2845 if (vp->v_type == VDIR) 2846 error = EISDIR; 2847 else if ((error = vn_writechk(vp)) == 0) { 2848 VATTR_NULL(&vattr); 2849 vattr.va_size = SCARG(uap, length); 2850 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2851 } 2852 VOP_UNLOCK(vp, 0); 2853 vn_finished_write(mp, 0); 2854 out: 2855 FILE_UNUSE(fp, p); 2856 return (error); 2857 } 2858 2859 /* 2860 * Sync an open file. 2861 */ 2862 /* ARGSUSED */ 2863 int 2864 sys_fsync(struct lwp *l, void *v, register_t *retval) 2865 { 2866 struct sys_fsync_args /* { 2867 syscallarg(int) fd; 2868 } */ *uap = v; 2869 struct proc *p = l->l_proc; 2870 struct vnode *vp; 2871 struct mount *mp; 2872 struct file *fp; 2873 int error; 2874 2875 /* getvnode() will use the descriptor for us */ 2876 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2877 return (error); 2878 vp = (struct vnode *)fp->f_data; 2879 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2880 FILE_UNUSE(fp, p); 2881 return (error); 2882 } 2883 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2884 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, p); 2885 if (error == 0 && bioops.io_fsync != NULL && 2886 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2887 (*bioops.io_fsync)(vp, 0); 2888 VOP_UNLOCK(vp, 0); 2889 vn_finished_write(mp, 0); 2890 FILE_UNUSE(fp, p); 2891 return (error); 2892 } 2893 2894 /* 2895 * Sync a range of file data. API modeled after that found in AIX. 2896 * 2897 * FDATASYNC indicates that we need only save enough metadata to be able 2898 * to re-read the written data. Note we duplicate AIX's requirement that 2899 * the file be open for writing. 2900 */ 2901 /* ARGSUSED */ 2902 int 2903 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2904 { 2905 struct sys_fsync_range_args /* { 2906 syscallarg(int) fd; 2907 syscallarg(int) flags; 2908 syscallarg(off_t) start; 2909 syscallarg(int) length; 2910 } */ *uap = v; 2911 struct proc *p = l->l_proc; 2912 struct vnode *vp; 2913 struct file *fp; 2914 int flags, nflags; 2915 off_t s, e, len; 2916 int error; 2917 2918 /* getvnode() will use the descriptor for us */ 2919 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2920 return (error); 2921 2922 if ((fp->f_flag & FWRITE) == 0) { 2923 FILE_UNUSE(fp, p); 2924 return (EBADF); 2925 } 2926 2927 flags = SCARG(uap, flags); 2928 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2929 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2930 return (EINVAL); 2931 } 2932 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2933 if (flags & FDATASYNC) 2934 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2935 else 2936 nflags = FSYNC_WAIT; 2937 if (flags & FDISKSYNC) 2938 nflags |= FSYNC_CACHE; 2939 2940 len = SCARG(uap, length); 2941 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2942 if (len) { 2943 s = SCARG(uap, start); 2944 e = s + len; 2945 if (e < s) { 2946 FILE_UNUSE(fp, p); 2947 return (EINVAL); 2948 } 2949 } else { 2950 e = 0; 2951 s = 0; 2952 } 2953 2954 vp = (struct vnode *)fp->f_data; 2955 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2956 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, p); 2957 2958 if (error == 0 && bioops.io_fsync != NULL && 2959 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2960 (*bioops.io_fsync)(vp, nflags); 2961 2962 VOP_UNLOCK(vp, 0); 2963 FILE_UNUSE(fp, p); 2964 return (error); 2965 } 2966 2967 /* 2968 * Sync the data of an open file. 2969 */ 2970 /* ARGSUSED */ 2971 int 2972 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 2973 { 2974 struct sys_fdatasync_args /* { 2975 syscallarg(int) fd; 2976 } */ *uap = v; 2977 struct proc *p = l->l_proc; 2978 struct vnode *vp; 2979 struct file *fp; 2980 int error; 2981 2982 /* getvnode() will use the descriptor for us */ 2983 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2984 return (error); 2985 if ((fp->f_flag & FWRITE) == 0) { 2986 FILE_UNUSE(fp, p); 2987 return (EBADF); 2988 } 2989 vp = (struct vnode *)fp->f_data; 2990 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2991 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, p); 2992 VOP_UNLOCK(vp, 0); 2993 FILE_UNUSE(fp, p); 2994 return (error); 2995 } 2996 2997 /* 2998 * Rename files, (standard) BSD semantics frontend. 2999 */ 3000 /* ARGSUSED */ 3001 int 3002 sys_rename(struct lwp *l, void *v, register_t *retval) 3003 { 3004 struct sys_rename_args /* { 3005 syscallarg(const char *) from; 3006 syscallarg(const char *) to; 3007 } */ *uap = v; 3008 struct proc *p = l->l_proc; 3009 3010 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 3011 } 3012 3013 /* 3014 * Rename files, POSIX semantics frontend. 3015 */ 3016 /* ARGSUSED */ 3017 int 3018 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3019 { 3020 struct sys___posix_rename_args /* { 3021 syscallarg(const char *) from; 3022 syscallarg(const char *) to; 3023 } */ *uap = v; 3024 struct proc *p = l->l_proc; 3025 3026 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 3027 } 3028 3029 /* 3030 * Rename files. Source and destination must either both be directories, 3031 * or both not be directories. If target is a directory, it must be empty. 3032 * If `from' and `to' refer to the same object, the value of the `retain' 3033 * argument is used to determine whether `from' will be 3034 * 3035 * (retain == 0) deleted unless `from' and `to' refer to the same 3036 * object in the file system's name space (BSD). 3037 * (retain == 1) always retained (POSIX). 3038 */ 3039 static int 3040 rename_files(const char *from, const char *to, struct proc *p, int retain) 3041 { 3042 struct mount *mp = NULL; 3043 struct vnode *tvp, *fvp, *tdvp; 3044 struct nameidata fromnd, tond; 3045 int error; 3046 3047 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3048 from, p); 3049 if ((error = namei(&fromnd)) != 0) 3050 return (error); 3051 fvp = fromnd.ni_vp; 3052 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3053 if (error != 0) { 3054 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3055 vrele(fromnd.ni_dvp); 3056 vrele(fvp); 3057 if (fromnd.ni_startdir) 3058 vrele(fromnd.ni_startdir); 3059 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3060 return (error); 3061 } 3062 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3063 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, p); 3064 if ((error = namei(&tond)) != 0) { 3065 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3066 vrele(fromnd.ni_dvp); 3067 vrele(fvp); 3068 goto out1; 3069 } 3070 tdvp = tond.ni_dvp; 3071 tvp = tond.ni_vp; 3072 3073 if (tvp != NULL) { 3074 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3075 error = ENOTDIR; 3076 goto out; 3077 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3078 error = EISDIR; 3079 goto out; 3080 } 3081 } 3082 3083 if (fvp == tdvp) 3084 error = EINVAL; 3085 3086 /* 3087 * Source and destination refer to the same object. 3088 */ 3089 if (fvp == tvp) { 3090 if (retain) 3091 error = -1; 3092 else if (fromnd.ni_dvp == tdvp && 3093 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3094 !memcmp(fromnd.ni_cnd.cn_nameptr, 3095 tond.ni_cnd.cn_nameptr, 3096 fromnd.ni_cnd.cn_namelen)) 3097 error = -1; 3098 } 3099 3100 out: 3101 if (!error) { 3102 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 3103 if (fromnd.ni_dvp != tdvp) 3104 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3105 if (tvp) { 3106 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 3107 } 3108 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3109 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3110 } else { 3111 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3112 if (tdvp == tvp) 3113 vrele(tdvp); 3114 else 3115 vput(tdvp); 3116 if (tvp) 3117 vput(tvp); 3118 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3119 vrele(fromnd.ni_dvp); 3120 vrele(fvp); 3121 } 3122 vrele(tond.ni_startdir); 3123 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3124 out1: 3125 vn_finished_write(mp, 0); 3126 if (fromnd.ni_startdir) 3127 vrele(fromnd.ni_startdir); 3128 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3129 return (error == -1 ? 0 : error); 3130 } 3131 3132 /* 3133 * Make a directory file. 3134 */ 3135 /* ARGSUSED */ 3136 int 3137 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3138 { 3139 struct sys_mkdir_args /* { 3140 syscallarg(const char *) path; 3141 syscallarg(int) mode; 3142 } */ *uap = v; 3143 struct proc *p = l->l_proc; 3144 struct mount *mp; 3145 struct vnode *vp; 3146 struct vattr vattr; 3147 int error; 3148 struct nameidata nd; 3149 3150 restart: 3151 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3152 SCARG(uap, path), p); 3153 if ((error = namei(&nd)) != 0) 3154 return (error); 3155 vp = nd.ni_vp; 3156 if (vp != NULL) { 3157 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3158 if (nd.ni_dvp == vp) 3159 vrele(nd.ni_dvp); 3160 else 3161 vput(nd.ni_dvp); 3162 vrele(vp); 3163 return (EEXIST); 3164 } 3165 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3166 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3167 if (nd.ni_dvp == vp) 3168 vrele(nd.ni_dvp); 3169 else 3170 vput(nd.ni_dvp); 3171 if ((error = vn_start_write(NULL, &mp, 3172 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3173 return (error); 3174 goto restart; 3175 } 3176 VATTR_NULL(&vattr); 3177 vattr.va_type = VDIR; 3178 vattr.va_mode = 3179 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3180 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3181 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3182 if (!error) 3183 vput(nd.ni_vp); 3184 vn_finished_write(mp, 0); 3185 return (error); 3186 } 3187 3188 /* 3189 * Remove a directory file. 3190 */ 3191 /* ARGSUSED */ 3192 int 3193 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3194 { 3195 struct sys_rmdir_args /* { 3196 syscallarg(const char *) path; 3197 } */ *uap = v; 3198 struct proc *p = l->l_proc; 3199 struct mount *mp; 3200 struct vnode *vp; 3201 int error; 3202 struct nameidata nd; 3203 3204 restart: 3205 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3206 SCARG(uap, path), p); 3207 if ((error = namei(&nd)) != 0) 3208 return (error); 3209 vp = nd.ni_vp; 3210 if (vp->v_type != VDIR) { 3211 error = ENOTDIR; 3212 goto out; 3213 } 3214 /* 3215 * No rmdir "." please. 3216 */ 3217 if (nd.ni_dvp == vp) { 3218 error = EINVAL; 3219 goto out; 3220 } 3221 /* 3222 * The root of a mounted filesystem cannot be deleted. 3223 */ 3224 if (vp->v_flag & VROOT) { 3225 error = EBUSY; 3226 goto out; 3227 } 3228 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3229 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3230 if (nd.ni_dvp == vp) 3231 vrele(nd.ni_dvp); 3232 else 3233 vput(nd.ni_dvp); 3234 vput(vp); 3235 if ((error = vn_start_write(NULL, &mp, 3236 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3237 return (error); 3238 goto restart; 3239 } 3240 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 3241 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3242 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3243 vn_finished_write(mp, 0); 3244 return (error); 3245 3246 out: 3247 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3248 if (nd.ni_dvp == vp) 3249 vrele(nd.ni_dvp); 3250 else 3251 vput(nd.ni_dvp); 3252 vput(vp); 3253 return (error); 3254 } 3255 3256 /* 3257 * Read a block of directory entries in a file system independent format. 3258 */ 3259 int 3260 sys_getdents(struct lwp *l, void *v, register_t *retval) 3261 { 3262 struct sys_getdents_args /* { 3263 syscallarg(int) fd; 3264 syscallarg(char *) buf; 3265 syscallarg(size_t) count; 3266 } */ *uap = v; 3267 struct proc *p = l->l_proc; 3268 struct file *fp; 3269 int error, done; 3270 3271 /* getvnode() will use the descriptor for us */ 3272 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3273 return (error); 3274 if ((fp->f_flag & FREAD) == 0) { 3275 error = EBADF; 3276 goto out; 3277 } 3278 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3279 SCARG(uap, count), &done, p, 0, 0); 3280 #ifdef KTRACE 3281 if (!error && KTRPOINT(p, KTR_GENIO)) { 3282 struct iovec iov; 3283 iov.iov_base = SCARG(uap, buf); 3284 iov.iov_len = done; 3285 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3286 } 3287 #endif 3288 *retval = done; 3289 out: 3290 FILE_UNUSE(fp, p); 3291 return (error); 3292 } 3293 3294 /* 3295 * Set the mode mask for creation of filesystem nodes. 3296 */ 3297 int 3298 sys_umask(struct lwp *l, void *v, register_t *retval) 3299 { 3300 struct sys_umask_args /* { 3301 syscallarg(mode_t) newmask; 3302 } */ *uap = v; 3303 struct proc *p = l->l_proc; 3304 struct cwdinfo *cwdi; 3305 3306 cwdi = p->p_cwdi; 3307 *retval = cwdi->cwdi_cmask; 3308 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3309 return (0); 3310 } 3311 3312 /* 3313 * Void all references to file by ripping underlying filesystem 3314 * away from vnode. 3315 */ 3316 /* ARGSUSED */ 3317 int 3318 sys_revoke(struct lwp *l, void *v, register_t *retval) 3319 { 3320 struct sys_revoke_args /* { 3321 syscallarg(const char *) path; 3322 } */ *uap = v; 3323 struct proc *p = l->l_proc; 3324 struct mount *mp; 3325 struct vnode *vp; 3326 struct vattr vattr; 3327 int error; 3328 struct nameidata nd; 3329 3330 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3331 if ((error = namei(&nd)) != 0) 3332 return (error); 3333 vp = nd.ni_vp; 3334 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 3335 goto out; 3336 if (p->p_ucred->cr_uid != vattr.va_uid && 3337 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 3338 goto out; 3339 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3340 goto out; 3341 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3342 VOP_REVOKE(vp, REVOKEALL); 3343 vn_finished_write(mp, 0); 3344 out: 3345 vrele(vp); 3346 return (error); 3347 } 3348 3349 /* 3350 * Convert a user file descriptor to a kernel file entry. 3351 */ 3352 int 3353 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3354 { 3355 struct vnode *vp; 3356 struct file *fp; 3357 3358 if ((fp = fd_getfile(fdp, fd)) == NULL) 3359 return (EBADF); 3360 3361 FILE_USE(fp); 3362 3363 if (fp->f_type != DTYPE_VNODE) { 3364 FILE_UNUSE(fp, NULL); 3365 return (EINVAL); 3366 } 3367 3368 vp = (struct vnode *)fp->f_data; 3369 if (vp->v_type == VBAD) { 3370 FILE_UNUSE(fp, NULL); 3371 return (EBADF); 3372 } 3373 3374 *fpp = fp; 3375 return (0); 3376 } 3377 3378 /* 3379 * Push extended attribute configuration information into the VFS. 3380 * 3381 * NOTE: Not all file systems that support extended attributes will 3382 * require the use of this system call. 3383 */ 3384 int 3385 sys_extattrctl(struct lwp *l, void *v, register_t *retval) 3386 { 3387 struct sys_extattrctl_args /* { 3388 syscallarg(const char *) path; 3389 syscallarg(int) cmd; 3390 syscallarg(const char *) filename; 3391 syscallarg(int) attrnamespace; 3392 syscallarg(const char *) attrname; 3393 } */ *uap = v; 3394 struct proc *p = l->l_proc; 3395 struct vnode *vp; 3396 struct nameidata nd; 3397 struct mount *mp; 3398 char attrname[EXTATTR_MAXNAMELEN]; 3399 int error; 3400 3401 if (SCARG(uap, attrname) != NULL) { 3402 error = copyinstr(SCARG(uap, attrname), attrname, 3403 sizeof(attrname), NULL); 3404 if (error) 3405 return (error); 3406 } 3407 3408 vp = NULL; 3409 if (SCARG(uap, filename) != NULL) { 3410 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 3411 SCARG(uap, filename), p); 3412 error = namei(&nd); 3413 if (error) 3414 return (error); 3415 vp = nd.ni_vp; 3416 } 3417 3418 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3419 error = namei(&nd); 3420 if (error) { 3421 if (vp != NULL) 3422 vput(vp); 3423 return (error); 3424 } 3425 3426 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 3427 if (error) { 3428 if (vp != NULL) 3429 vput(vp); 3430 return (error); 3431 } 3432 3433 error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), vp, 3434 SCARG(uap, attrnamespace), 3435 SCARG(uap, attrname) != NULL ? attrname : NULL, p); 3436 3437 vn_finished_write(mp, 0); 3438 3439 if (vp != NULL) 3440 vrele(vp); 3441 3442 return (error); 3443 } 3444 3445 /* 3446 * Set a named extended attribute on a file or directory. 3447 */ 3448 static int 3449 extattr_set_vp(struct vnode *vp, int attrnamespace, const char *attrname, 3450 const void *data, size_t nbytes, struct proc *p, register_t *retval) 3451 { 3452 struct mount *mp; 3453 struct uio auio; 3454 struct iovec aiov; 3455 ssize_t cnt; 3456 int error; 3457 3458 error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH); 3459 if (error) 3460 return (error); 3461 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3462 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3463 3464 aiov.iov_base = __UNCONST(data); /* XXXUNCONST kills const */ 3465 aiov.iov_len = nbytes; 3466 auio.uio_iov = &aiov; 3467 auio.uio_iovcnt = 1; 3468 auio.uio_offset = 0; 3469 if (nbytes > INT_MAX) { 3470 error = EINVAL; 3471 goto done; 3472 } 3473 auio.uio_resid = nbytes; 3474 auio.uio_rw = UIO_WRITE; 3475 auio.uio_segflg = UIO_USERSPACE; 3476 auio.uio_procp = p; 3477 cnt = nbytes; 3478 3479 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, 3480 p->p_ucred, p); 3481 cnt -= auio.uio_resid; 3482 retval[0] = cnt; 3483 3484 done: 3485 VOP_UNLOCK(vp, 0); 3486 vn_finished_write(mp, 0); 3487 return (error); 3488 } 3489 3490 int 3491 sys_extattr_set_fd(struct lwp *l, void *v, register_t *retval) 3492 { 3493 struct sys_extattr_set_fd_args /* { 3494 syscallarg(int) fd; 3495 syscallarg(int) attrnamespace; 3496 syscallarg(const char *) attrname; 3497 syscallarg(const void *) data; 3498 syscallarg(size_t) nbytes; 3499 } */ *uap = v; 3500 struct proc *p = l->l_proc; 3501 struct file *fp; 3502 struct vnode *vp; 3503 char attrname[EXTATTR_MAXNAMELEN]; 3504 int error; 3505 3506 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3507 NULL); 3508 if (error) 3509 return (error); 3510 3511 error = getvnode(p->p_fd, SCARG(uap, fd), &fp); 3512 if (error) 3513 return (error); 3514 vp = (struct vnode *) fp->f_data; 3515 3516 error = extattr_set_vp(vp, SCARG(uap, attrnamespace), attrname, 3517 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3518 3519 FILE_UNUSE(fp, p); 3520 return (error); 3521 } 3522 3523 int 3524 sys_extattr_set_file(struct lwp *l, void *v, register_t *retval) 3525 { 3526 struct sys_extattr_set_file_args /* { 3527 syscallarg(const char *) path; 3528 syscallarg(int) attrnamespace; 3529 syscallarg(const char *) attrname; 3530 syscallarg(const void *) data; 3531 syscallarg(size_t) nbytes; 3532 } */ *uap = v; 3533 struct proc *p = l->l_proc; 3534 struct nameidata nd; 3535 char attrname[EXTATTR_MAXNAMELEN]; 3536 int error; 3537 3538 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3539 NULL); 3540 if (error) 3541 return (error); 3542 3543 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3544 error = namei(&nd); 3545 if (error) 3546 return (error); 3547 3548 error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname, 3549 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3550 3551 vrele(nd.ni_vp); 3552 return (error); 3553 } 3554 3555 int 3556 sys_extattr_set_link(struct lwp *l, void *v, register_t *retval) 3557 { 3558 struct sys_extattr_set_link_args /* { 3559 syscallarg(const char *) path; 3560 syscallarg(int) attrnamespace; 3561 syscallarg(const char *) attrname; 3562 syscallarg(const void *) data; 3563 syscallarg(size_t) nbytes; 3564 } */ *uap = v; 3565 struct proc *p = l->l_proc; 3566 struct nameidata nd; 3567 char attrname[EXTATTR_MAXNAMELEN]; 3568 int error; 3569 3570 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3571 NULL); 3572 if (error) 3573 return (error); 3574 3575 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3576 error = namei(&nd); 3577 if (error) 3578 return (error); 3579 3580 error = extattr_set_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname, 3581 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3582 3583 vrele(nd.ni_vp); 3584 return (error); 3585 } 3586 3587 /* 3588 * Get a named extended attribute on a file or directory. 3589 */ 3590 static int 3591 extattr_get_vp(struct vnode *vp, int attrnamespace, const char *attrname, 3592 void *data, size_t nbytes, struct proc *p, register_t *retval) 3593 { 3594 struct uio auio, *auiop; 3595 struct iovec aiov; 3596 ssize_t cnt; 3597 size_t size, *sizep; 3598 int error; 3599 3600 VOP_LEASE(vp, p, p->p_ucred, LEASE_READ); 3601 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3602 3603 /* 3604 * Slightly unusual semantics: if the user provides a NULL data 3605 * pointer, they don't want to receive the data, just the maximum 3606 * read length. 3607 */ 3608 auiop = NULL; 3609 sizep = NULL; 3610 cnt = 0; 3611 if (data != NULL) { 3612 aiov.iov_base = data; 3613 aiov.iov_len = nbytes; 3614 auio.uio_iov = &aiov; 3615 auio.uio_offset = 0; 3616 if (nbytes > INT_MAX) { 3617 error = EINVAL; 3618 goto done; 3619 } 3620 auio.uio_resid = nbytes; 3621 auio.uio_rw = UIO_READ; 3622 auio.uio_segflg = UIO_USERSPACE; 3623 auio.uio_procp = p; 3624 auiop = &auio; 3625 cnt = nbytes; 3626 } else 3627 sizep = &size; 3628 3629 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, auiop, sizep, 3630 p->p_ucred, p); 3631 3632 if (auiop != NULL) { 3633 cnt -= auio.uio_resid; 3634 retval[0] = cnt; 3635 } else 3636 retval[0] = size; 3637 3638 done: 3639 VOP_UNLOCK(vp, 0); 3640 return (error); 3641 } 3642 3643 int 3644 sys_extattr_get_fd(struct lwp *l, void *v, register_t *retval) 3645 { 3646 struct sys_extattr_get_fd_args /* { 3647 syscallarg(int) fd; 3648 syscallarg(int) attrnamespace; 3649 syscallarg(const char *) attrname; 3650 syscallarg(void *) data; 3651 syscallarg(size_t) nbytes; 3652 } */ *uap = v; 3653 struct proc *p = l->l_proc; 3654 struct file *fp; 3655 struct vnode *vp; 3656 char attrname[EXTATTR_MAXNAMELEN]; 3657 int error; 3658 3659 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3660 NULL); 3661 if (error) 3662 return (error); 3663 3664 error = getvnode(p->p_fd, SCARG(uap, fd), &fp); 3665 if (error) 3666 return (error); 3667 vp = (struct vnode *) fp->f_data; 3668 3669 error = extattr_get_vp(vp, SCARG(uap, attrnamespace), attrname, 3670 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3671 3672 FILE_UNUSE(fp, p); 3673 return (error); 3674 } 3675 3676 int 3677 sys_extattr_get_file(struct lwp *l, void *v, register_t *retval) 3678 { 3679 struct sys_extattr_get_file_args /* { 3680 syscallarg(const char *) path; 3681 syscallarg(int) attrnamespace; 3682 syscallarg(const char *) attrname; 3683 syscallarg(void *) data; 3684 syscallarg(size_t) nbytes; 3685 } */ *uap = v; 3686 struct proc *p = l->l_proc; 3687 struct nameidata nd; 3688 char attrname[EXTATTR_MAXNAMELEN]; 3689 int error; 3690 3691 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3692 NULL); 3693 if (error) 3694 return (error); 3695 3696 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3697 error = namei(&nd); 3698 if (error) 3699 return (error); 3700 3701 error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname, 3702 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3703 3704 vrele(nd.ni_vp); 3705 return (error); 3706 } 3707 3708 int 3709 sys_extattr_get_link(struct lwp *l, void *v, register_t *retval) 3710 { 3711 struct sys_extattr_get_link_args /* { 3712 syscallarg(const char *) path; 3713 syscallarg(int) attrnamespace; 3714 syscallarg(const char *) attrname; 3715 syscallarg(void *) data; 3716 syscallarg(size_t) nbytes; 3717 } */ *uap = v; 3718 struct proc *p = l->l_proc; 3719 struct nameidata nd; 3720 char attrname[EXTATTR_MAXNAMELEN]; 3721 int error; 3722 3723 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3724 NULL); 3725 if (error) 3726 return (error); 3727 3728 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3729 error = namei(&nd); 3730 if (error) 3731 return (error); 3732 3733 error = extattr_get_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname, 3734 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3735 3736 vrele(nd.ni_vp); 3737 return (error); 3738 } 3739 3740 /* 3741 * Delete a named extended attribute on a file or directory. 3742 */ 3743 static int 3744 extattr_delete_vp(struct vnode *vp, int attrnamespace, const char *attrname, 3745 struct proc *p) 3746 { 3747 struct mount *mp; 3748 int error; 3749 3750 error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH); 3751 if (error) 3752 return (error); 3753 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 3754 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3755 3756 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, p->p_ucred, p); 3757 if (error == EOPNOTSUPP) 3758 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 3759 p->p_ucred, p); 3760 3761 VOP_UNLOCK(vp, 0); 3762 vn_finished_write(mp, 0); 3763 return (error); 3764 } 3765 3766 int 3767 sys_extattr_delete_fd(struct lwp *l, void *v, register_t *retval) 3768 { 3769 struct sys_extattr_delete_fd_args /* { 3770 syscallarg(int) fd; 3771 syscallarg(int) attrnamespace; 3772 syscallarg(const char *) attrname; 3773 } */ *uap = v; 3774 struct proc *p = l->l_proc; 3775 struct file *fp; 3776 struct vnode *vp; 3777 char attrname[EXTATTR_MAXNAMELEN]; 3778 int error; 3779 3780 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3781 NULL); 3782 if (error) 3783 return (error); 3784 3785 error = getvnode(p->p_fd, SCARG(uap, fd), &fp); 3786 if (error) 3787 return (error); 3788 vp = (struct vnode *) fp->f_data; 3789 3790 error = extattr_delete_vp(vp, SCARG(uap, attrnamespace), attrname, p); 3791 3792 FILE_UNUSE(fp, p); 3793 return (error); 3794 } 3795 3796 int 3797 sys_extattr_delete_file(struct lwp *l, void *v, register_t *retval) 3798 { 3799 struct sys_extattr_delete_file_args /* { 3800 syscallarg(const char *) path; 3801 syscallarg(int) attrnamespace; 3802 syscallarg(const char *) attrname; 3803 } */ *uap = v; 3804 struct proc *p = l->l_proc; 3805 struct nameidata nd; 3806 char attrname[EXTATTR_MAXNAMELEN]; 3807 int error; 3808 3809 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3810 NULL); 3811 if (error) 3812 return (error); 3813 3814 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3815 error = namei(&nd); 3816 if (error) 3817 return (error); 3818 3819 error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname, 3820 p); 3821 3822 vrele(nd.ni_vp); 3823 return (error); 3824 } 3825 3826 int 3827 sys_extattr_delete_link(struct lwp *l, void *v, register_t *retval) 3828 { 3829 struct sys_extattr_delete_link_args /* { 3830 syscallarg(const char *) path; 3831 syscallarg(int) attrnamespace; 3832 syscallarg(const char *) attrname; 3833 } */ *uap = v; 3834 struct proc *p = l->l_proc; 3835 struct nameidata nd; 3836 char attrname[EXTATTR_MAXNAMELEN]; 3837 int error; 3838 3839 error = copyinstr(SCARG(uap, attrname), attrname, sizeof(attrname), 3840 NULL); 3841 if (error) 3842 return (error); 3843 3844 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3845 error = namei(&nd); 3846 if (error) 3847 return (error); 3848 3849 error = extattr_delete_vp(nd.ni_vp, SCARG(uap, attrnamespace), attrname, 3850 p); 3851 3852 vrele(nd.ni_vp); 3853 return (error); 3854 } 3855 3856 /* 3857 * Retrieve a list of extended attributes on a file or directory. 3858 */ 3859 static int 3860 extattr_list_vp(struct vnode *vp, int attrnamespace, void *data, size_t nbytes, 3861 struct proc *p, register_t *retval) 3862 { 3863 struct uio auio, *auiop; 3864 size_t size, *sizep; 3865 struct iovec aiov; 3866 ssize_t cnt; 3867 int error; 3868 3869 VOP_LEASE(vp, p, p->p_ucred, LEASE_READ); 3870 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3871 3872 auiop = NULL; 3873 sizep = NULL; 3874 cnt = 0; 3875 if (data != NULL) { 3876 aiov.iov_base = data; 3877 aiov.iov_len = nbytes; 3878 auio.uio_iov = &aiov; 3879 auio.uio_offset = 0; 3880 if (nbytes > INT_MAX) { 3881 error = EINVAL; 3882 goto done; 3883 } 3884 auio.uio_resid = nbytes; 3885 auio.uio_rw = UIO_READ; 3886 auio.uio_segflg = UIO_USERSPACE; 3887 auio.uio_procp = p; 3888 auiop = &auio; 3889 cnt = nbytes; 3890 } else 3891 sizep = &size; 3892 3893 error = VOP_LISTEXTATTR(vp, attrnamespace, auiop, sizep, 3894 p->p_ucred, p); 3895 3896 if (auiop != NULL) { 3897 cnt -= auio.uio_resid; 3898 retval[0] = cnt; 3899 } else 3900 retval[0] = size; 3901 3902 done: 3903 VOP_UNLOCK(vp, 0); 3904 return (error); 3905 } 3906 3907 int 3908 sys_extattr_list_fd(struct lwp *l, void *v, register_t *retval) 3909 { 3910 struct sys_extattr_list_fd_args /* { 3911 syscallarg(int) fd; 3912 syscallarg(int) attrnamespace; 3913 syscallarg(void *) data; 3914 syscallarg(size_t) nbytes; 3915 } */ *uap = v; 3916 struct proc *p = l->l_proc; 3917 struct file *fp; 3918 struct vnode *vp; 3919 int error; 3920 3921 error = getvnode(p->p_fd, SCARG(uap, fd), &fp); 3922 if (error) 3923 return (error); 3924 vp = (struct vnode *) fp->f_data; 3925 3926 error = extattr_list_vp(vp, SCARG(uap, attrnamespace), 3927 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3928 3929 FILE_UNUSE(fp, p); 3930 return (error); 3931 } 3932 3933 int 3934 sys_extattr_list_file(struct lwp *l, void *v, register_t *retval) 3935 { 3936 struct sys_extattr_list_file_args /* { 3937 syscallarg(const char *) path; 3938 syscallarg(int) attrnamespace; 3939 syscallarg(void *) data; 3940 syscallarg(size_t) nbytes; 3941 } */ *uap = v; 3942 struct proc *p = l->l_proc; 3943 struct nameidata nd; 3944 int error; 3945 3946 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3947 error = namei(&nd); 3948 if (error) 3949 return (error); 3950 3951 error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace), 3952 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3953 3954 vrele(nd.ni_vp); 3955 return (error); 3956 } 3957 3958 int 3959 sys_extattr_list_link(struct lwp *l, void *v, register_t *retval) 3960 { 3961 struct sys_extattr_list_link_args /* { 3962 syscallarg(const char *) path; 3963 syscallarg(int) attrnamespace; 3964 syscallarg(void *) data; 3965 syscallarg(size_t) nbytes; 3966 } */ *uap = v; 3967 struct proc *p = l->l_proc; 3968 struct nameidata nd; 3969 int error; 3970 3971 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 3972 error = namei(&nd); 3973 if (error) 3974 return (error); 3975 3976 error = extattr_list_vp(nd.ni_vp, SCARG(uap, attrnamespace), 3977 SCARG(uap, data), SCARG(uap, nbytes), p, retval); 3978 3979 vrele(nd.ni_vp); 3980 return (error); 3981 } 3982
Indexes created Mon Sep 22 21:09:42 GMT 2025