vfs_syscalls.c revision 1.135
1 /* $NetBSD: vfs_syscalls.c,v 1.135 1999/05/05 20:01:10 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. All advertising materials mentioning features or use of this software 21 * must display the following acknowledgement: 22 * This product includes software developed by the University of 23 * California, Berkeley and its contributors. 24 * 4. Neither the name of the University nor the names of its contributors 25 * may be used to endorse or promote products derived from this software 26 * without specific prior written permission. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 38 * SUCH DAMAGE. 39 * 40 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 41 */ 42 43 #include "opt_compat_netbsd.h" 44 #include "opt_compat_43.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/namei.h> 49 #include <sys/filedesc.h> 50 #include <sys/kernel.h> 51 #include <sys/file.h> 52 #include <sys/stat.h> 53 #include <sys/vnode.h> 54 #include <sys/mount.h> 55 #include <sys/proc.h> 56 #include <sys/uio.h> 57 #include <sys/malloc.h> 58 #include <sys/dirent.h> 59 60 #include <sys/syscallargs.h> 61 62 #include <vm/vm.h> 63 #include <sys/sysctl.h> 64 65 #include <uvm/uvm_extern.h> 66 67 static int change_dir __P((struct nameidata *, struct proc *)); 68 static int change_mode __P((struct vnode *, int, struct proc *p)); 69 static int change_owner __P((struct vnode *, uid_t, gid_t, struct proc *, 70 int)); 71 static int change_utimes __P((struct vnode *vp, const struct timeval *, 72 struct proc *p)); 73 static int rename_files __P((const char *, const char *, struct proc *, int)); 74 75 void checkdirs __P((struct vnode *)); 76 int dounmount __P((struct mount *, int, struct proc *)); 77 78 /* 79 * Virtual File System System Calls 80 */ 81 82 /* 83 * Mount a file system. 84 */ 85 86 /* 87 * This table is used to maintain compatibility with 4.3BSD 88 * and NetBSD 0.9 mount syscalls. Note, the order is important! 89 * 90 * Also note that not all of these had actual numbers in 4.3BSD 91 * or NetBSD 0.9! 92 */ 93 const char *mountcompatnames[] = { 94 NULL, /* 0 = MOUNT_NONE */ 95 MOUNT_FFS, /* 1 */ 96 MOUNT_NFS, /* 2 */ 97 MOUNT_MFS, /* 3 */ 98 MOUNT_MSDOS, /* 4 */ 99 MOUNT_LFS, /* 5 */ 100 NULL, /* 6 = MOUNT_LOFS */ 101 MOUNT_FDESC, /* 7 */ 102 MOUNT_PORTAL, /* 8 */ 103 MOUNT_NULL, /* 9 */ 104 MOUNT_UMAP, /* 10 */ 105 MOUNT_KERNFS, /* 11 */ 106 MOUNT_PROCFS, /* 12 */ 107 MOUNT_AFS, /* 13 */ 108 MOUNT_CD9660, /* 14 = MOUNT_ISOFS */ 109 MOUNT_UNION, /* 15 */ 110 MOUNT_ADOSFS, /* 16 */ 111 MOUNT_EXT2FS, /* 17 */ 112 MOUNT_CODA, /* 18 */ 113 MOUNT_FILECORE, /* 19 */ 114 }; 115 const int nmountcompatnames = sizeof(mountcompatnames) / 116 sizeof(mountcompatnames[0]); 117 118 /* ARGSUSED */ 119 int 120 sys_mount(p, v, retval) 121 struct proc *p; 122 void *v; 123 register_t *retval; 124 { 125 register struct sys_mount_args /* { 126 syscallarg(const char *) type; 127 syscallarg(const char *) path; 128 syscallarg(int) flags; 129 syscallarg(void *) data; 130 } */ *uap = v; 131 struct vnode *vp; 132 struct mount *mp; 133 int error, flag = 0; 134 char fstypename[MFSNAMELEN]; 135 struct vattr va; 136 struct nameidata nd; 137 struct vfsops *vfs; 138 139 /* 140 * Get vnode to be covered 141 */ 142 NDINIT(&nd, LOOKUP, FOLLOW , UIO_USERSPACE, 143 SCARG(uap, path), p); 144 if ((error = namei(&nd)) != 0) 145 return (error); 146 vp = nd.ni_vp; 147 /* 148 * A lookup in VFS_MOUNT might result in an attempt to 149 * lock this vnode again, so make the lock resursive. 150 */ 151 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 152 if (SCARG(uap, flags) & MNT_UPDATE) { 153 if ((vp->v_flag & VROOT) == 0) { 154 vput(vp); 155 return (EINVAL); 156 } 157 mp = vp->v_mount; 158 flag = mp->mnt_flag; 159 vfs = mp->mnt_op; 160 /* 161 * We only allow the filesystem to be reloaded if it 162 * is currently mounted read-only. 163 */ 164 if ((SCARG(uap, flags) & MNT_RELOAD) && 165 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 166 vput(vp); 167 return (EOPNOTSUPP); /* Needs translation */ 168 } 169 /* 170 * In "highly secure" mode, don't let the caller do anything 171 * but downgrade a filesystem from read-write to read-only. 172 * (see also below; MNT_UPDATE is required.) 173 */ 174 if (securelevel >= 2 && 175 (SCARG(uap, flags) != 176 (mp->mnt_flag | MNT_RDONLY | 177 MNT_RELOAD | MNT_FORCE | MNT_UPDATE))) { 178 vput(vp); 179 return (EPERM); 180 } 181 mp->mnt_flag |= 182 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 183 /* 184 * Only root, or the user that did the original mount is 185 * permitted to update it. 186 */ 187 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 188 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 189 vput(vp); 190 return (error); 191 } 192 /* 193 * Do not allow NFS export by non-root users. For non-root 194 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 195 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 196 */ 197 if (p->p_ucred->cr_uid != 0) { 198 if (SCARG(uap, flags) & MNT_EXPORTED) { 199 vput(vp); 200 return (EPERM); 201 } 202 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 203 if (flag & MNT_NOEXEC) 204 SCARG(uap, flags) |= MNT_NOEXEC; 205 } 206 if (vfs_busy(mp, LK_NOWAIT, 0)) { 207 vput(vp); 208 return (EPERM); 209 } 210 VOP_UNLOCK(vp, 0); 211 goto update; 212 } else { 213 if (securelevel >= 2) 214 return (EPERM); 215 } 216 /* 217 * If the user is not root, ensure that they own the directory 218 * onto which we are attempting to mount. 219 */ 220 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0 || 221 (va.va_uid != p->p_ucred->cr_uid && 222 (error = suser(p->p_ucred, &p->p_acflag)) != 0)) { 223 vput(vp); 224 return (error); 225 } 226 /* 227 * Do not allow NFS export by non-root users. For non-root users, 228 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 229 * mount point is already MNT_NOEXEC. 230 */ 231 if (p->p_ucred->cr_uid != 0) { 232 if (SCARG(uap, flags) & MNT_EXPORTED) { 233 vput(vp); 234 return (EPERM); 235 } 236 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 237 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 238 SCARG(uap, flags) |= MNT_NOEXEC; 239 } 240 if ((error = vinvalbuf(vp, V_SAVE, p->p_ucred, p, 0, 0)) != 0) 241 return (error); 242 if (vp->v_type != VDIR) { 243 vput(vp); 244 return (ENOTDIR); 245 } 246 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 247 if (error) { 248 #if defined(COMPAT_09) || defined(COMPAT_43) 249 /* 250 * Historically filesystem types were identified by number. 251 * If we get an integer for the filesystem type instead of a 252 * string, we check to see if it matches one of the historic 253 * filesystem types. 254 */ 255 u_long fsindex = (u_long)SCARG(uap, type); 256 if (fsindex >= nmountcompatnames || 257 mountcompatnames[fsindex] == NULL) { 258 vput(vp); 259 return (ENODEV); 260 } 261 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 262 #else 263 vput(vp); 264 return (error); 265 #endif 266 } 267 #ifdef COMPAT_10 268 /* Accept `ufs' as an alias for `ffs'. */ 269 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 270 strncpy(fstypename, "ffs", MFSNAMELEN); 271 #endif 272 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 273 vput(vp); 274 return (ENODEV); 275 } 276 if (vp->v_mountedhere != NULL) { 277 vput(vp); 278 return (EBUSY); 279 } 280 281 /* 282 * Allocate and initialize the file system. 283 */ 284 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 285 M_MOUNT, M_WAITOK); 286 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 287 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 288 (void)vfs_busy(mp, LK_NOWAIT, 0); 289 mp->mnt_op = vfs; 290 vfs->vfs_refcount++; 291 mp->mnt_vnodecovered = vp; 292 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 293 update: 294 /* 295 * Set the mount level flags. 296 */ 297 if (SCARG(uap, flags) & MNT_RDONLY) 298 mp->mnt_flag |= MNT_RDONLY; 299 else if (mp->mnt_flag & MNT_RDONLY) 300 mp->mnt_flag |= MNT_WANTRDWR; 301 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 302 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 303 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM); 304 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC | 305 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | 306 MNT_NOCOREDUMP | MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM); 307 /* 308 * Mount the filesystem. 309 */ 310 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, p); 311 if (mp->mnt_flag & MNT_UPDATE) { 312 vrele(vp); 313 if (mp->mnt_flag & MNT_WANTRDWR) 314 mp->mnt_flag &= ~MNT_RDONLY; 315 mp->mnt_flag &=~ 316 (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_WANTRDWR); 317 if (error) 318 mp->mnt_flag = flag; 319 vfs_unbusy(mp); 320 return (error); 321 } 322 /* 323 * Put the new filesystem on the mount list after root. 324 */ 325 cache_purge(vp); 326 if (!error) { 327 vp->v_mountedhere = mp; 328 simple_lock(&mountlist_slock); 329 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 330 simple_unlock(&mountlist_slock); 331 checkdirs(vp); 332 VOP_UNLOCK(vp, 0); 333 vfs_unbusy(mp); 334 (void) VFS_STATFS(mp, &mp->mnt_stat, p); 335 if ((error = VFS_START(mp, 0, p))) 336 vrele(vp); 337 } else { 338 vp->v_mountedhere = (struct mount *)0; 339 vfs->vfs_refcount--; 340 vfs_unbusy(mp); 341 free((caddr_t)mp, M_MOUNT); 342 vput(vp); 343 } 344 return (error); 345 } 346 347 /* 348 * Scan all active processes to see if any of them have a current 349 * or root directory onto which the new filesystem has just been 350 * mounted. If so, replace them with the new mount point. 351 */ 352 void 353 checkdirs(olddp) 354 struct vnode *olddp; 355 { 356 struct cwdinfo *cwdi; 357 struct vnode *newdp; 358 struct proc *p; 359 360 if (olddp->v_usecount == 1) 361 return; 362 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 363 panic("mount: lost mount"); 364 for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) { 365 cwdi = p->p_cwdi; 366 if (cwdi->cwdi_cdir == olddp) { 367 vrele(cwdi->cwdi_cdir); 368 VREF(newdp); 369 cwdi->cwdi_cdir = newdp; 370 } 371 if (cwdi->cwdi_rdir == olddp) { 372 vrele(cwdi->cwdi_rdir); 373 VREF(newdp); 374 cwdi->cwdi_rdir = newdp; 375 } 376 } 377 if (rootvnode == olddp) { 378 vrele(rootvnode); 379 VREF(newdp); 380 rootvnode = newdp; 381 } 382 vput(newdp); 383 } 384 385 /* 386 * Unmount a file system. 387 * 388 * Note: unmount takes a path to the vnode mounted on as argument, 389 * not special file (as before). 390 */ 391 /* ARGSUSED */ 392 int 393 sys_unmount(p, v, retval) 394 struct proc *p; 395 void *v; 396 register_t *retval; 397 { 398 register struct sys_unmount_args /* { 399 syscallarg(const char *) path; 400 syscallarg(int) flags; 401 } */ *uap = v; 402 register struct vnode *vp; 403 struct mount *mp; 404 int error; 405 struct nameidata nd; 406 407 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 408 SCARG(uap, path), p); 409 if ((error = namei(&nd)) != 0) 410 return (error); 411 vp = nd.ni_vp; 412 mp = vp->v_mount; 413 414 /* 415 * Only root, or the user that did the original mount is 416 * permitted to unmount this filesystem. 417 */ 418 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 419 (error = suser(p->p_ucred, &p->p_acflag)) != 0) { 420 vput(vp); 421 return (error); 422 } 423 424 /* 425 * Don't allow unmounting the root file system. 426 */ 427 if (mp->mnt_flag & MNT_ROOTFS) { 428 vput(vp); 429 return (EINVAL); 430 } 431 432 /* 433 * Must be the root of the filesystem 434 */ 435 if ((vp->v_flag & VROOT) == 0) { 436 vput(vp); 437 return (EINVAL); 438 } 439 vput(vp); 440 441 if (vfs_busy(mp, 0, 0)) 442 return (EBUSY); 443 444 return (dounmount(mp, SCARG(uap, flags), p)); 445 } 446 447 /* 448 * Do the actual file system unmount. File system is assumed to have been 449 * marked busy by the caller. 450 */ 451 int 452 dounmount(mp, flags, p) 453 register struct mount *mp; 454 int flags; 455 struct proc *p; 456 { 457 struct vnode *coveredvp; 458 int error; 459 460 simple_lock(&mountlist_slock); 461 mp->mnt_flag |= MNT_UNMOUNT; 462 vfs_unbusy(mp); 463 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 464 if (mp->mnt_flag & MNT_EXPUBLIC) 465 vfs_setpublicfs(NULL, NULL, NULL); 466 mp->mnt_flag &=~ MNT_ASYNC; 467 cache_purgevfs(mp); /* remove cache entries for this file sys */ 468 if (((mp->mnt_flag & MNT_RDONLY) || 469 (error = VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p)) == 0) || 470 (flags & MNT_FORCE)) 471 error = VFS_UNMOUNT(mp, flags, p); 472 simple_lock(&mountlist_slock); 473 if (error) { 474 mp->mnt_flag &= ~MNT_UNMOUNT; 475 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 476 &mountlist_slock); 477 if (mp->mnt_flag & MNT_MWAIT) 478 wakeup((caddr_t)mp); 479 return (error); 480 } 481 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 482 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 483 coveredvp->v_mountedhere = NULL; 484 vrele(coveredvp); 485 } 486 mp->mnt_op->vfs_refcount--; 487 if (mp->mnt_vnodelist.lh_first != NULL) 488 panic("unmount: dangling vnode"); 489 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 490 if (mp->mnt_flag & MNT_MWAIT) 491 wakeup((caddr_t)mp); 492 free((caddr_t)mp, M_MOUNT); 493 return (0); 494 } 495 496 /* 497 * Sync each mounted filesystem. 498 */ 499 #ifdef DEBUG 500 int syncprt = 0; 501 struct ctldebug debug0 = { "syncprt", &syncprt }; 502 #endif 503 504 /* ARGSUSED */ 505 int 506 sys_sync(p, v, retval) 507 struct proc *p; 508 void *v; 509 register_t *retval; 510 { 511 register struct mount *mp, *nmp; 512 int asyncflag; 513 514 simple_lock(&mountlist_slock); 515 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 516 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 517 nmp = mp->mnt_list.cqe_prev; 518 continue; 519 } 520 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 521 asyncflag = mp->mnt_flag & MNT_ASYNC; 522 mp->mnt_flag &= ~MNT_ASYNC; 523 uvm_vnp_sync(mp); 524 VFS_SYNC(mp, MNT_NOWAIT, p->p_ucred, p); 525 if (asyncflag) 526 mp->mnt_flag |= MNT_ASYNC; 527 } 528 simple_lock(&mountlist_slock); 529 nmp = mp->mnt_list.cqe_prev; 530 vfs_unbusy(mp); 531 532 } 533 simple_unlock(&mountlist_slock); 534 #ifdef DEBUG 535 if (syncprt) 536 vfs_bufstats(); 537 #endif /* DEBUG */ 538 return (0); 539 } 540 541 /* 542 * Change filesystem quotas. 543 */ 544 /* ARGSUSED */ 545 int 546 sys_quotactl(p, v, retval) 547 struct proc *p; 548 void *v; 549 register_t *retval; 550 { 551 register struct sys_quotactl_args /* { 552 syscallarg(const char *) path; 553 syscallarg(int) cmd; 554 syscallarg(int) uid; 555 syscallarg(caddr_t) arg; 556 } */ *uap = v; 557 register struct mount *mp; 558 int error; 559 struct nameidata nd; 560 561 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 562 if ((error = namei(&nd)) != 0) 563 return (error); 564 mp = nd.ni_vp->v_mount; 565 vrele(nd.ni_vp); 566 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 567 SCARG(uap, arg), p)); 568 } 569 570 /* 571 * Get filesystem statistics. 572 */ 573 /* ARGSUSED */ 574 int 575 sys_statfs(p, v, retval) 576 struct proc *p; 577 void *v; 578 register_t *retval; 579 { 580 register struct sys_statfs_args /* { 581 syscallarg(const char *) path; 582 syscallarg(struct statfs *) buf; 583 } */ *uap = v; 584 register struct mount *mp; 585 register struct statfs *sp; 586 int error; 587 struct nameidata nd; 588 589 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 590 if ((error = namei(&nd)) != 0) 591 return (error); 592 mp = nd.ni_vp->v_mount; 593 sp = &mp->mnt_stat; 594 vrele(nd.ni_vp); 595 if ((error = VFS_STATFS(mp, sp, p)) != 0) 596 return (error); 597 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 598 return (copyout(sp, SCARG(uap, buf), sizeof(*sp))); 599 } 600 601 /* 602 * Get filesystem statistics. 603 */ 604 /* ARGSUSED */ 605 int 606 sys_fstatfs(p, v, retval) 607 struct proc *p; 608 void *v; 609 register_t *retval; 610 { 611 register struct sys_fstatfs_args /* { 612 syscallarg(int) fd; 613 syscallarg(struct statfs *) buf; 614 } */ *uap = v; 615 struct file *fp; 616 struct mount *mp; 617 register struct statfs *sp; 618 int error; 619 620 /* getvnode() will use the descriptor for us */ 621 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 622 return (error); 623 mp = ((struct vnode *)fp->f_data)->v_mount; 624 sp = &mp->mnt_stat; 625 if ((error = VFS_STATFS(mp, sp, p)) != 0) 626 goto out; 627 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 628 error = copyout(sp, SCARG(uap, buf), sizeof(*sp)); 629 out: 630 FILE_UNUSE(fp, p); 631 return (error); 632 } 633 634 /* 635 * Get statistics on all filesystems. 636 */ 637 int 638 sys_getfsstat(p, v, retval) 639 struct proc *p; 640 void *v; 641 register_t *retval; 642 { 643 register struct sys_getfsstat_args /* { 644 syscallarg(struct statfs *) buf; 645 syscallarg(long) bufsize; 646 syscallarg(int) flags; 647 } */ *uap = v; 648 register struct mount *mp, *nmp; 649 register struct statfs *sp; 650 caddr_t sfsp; 651 long count, maxcount, error; 652 653 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs); 654 sfsp = (caddr_t)SCARG(uap, buf); 655 simple_lock(&mountlist_slock); 656 count = 0; 657 for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) { 658 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 659 nmp = mp->mnt_list.cqe_next; 660 continue; 661 } 662 if (sfsp && count < maxcount) { 663 sp = &mp->mnt_stat; 664 /* 665 * If MNT_NOWAIT is specified, do not refresh the 666 * fsstat cache. MNT_WAIT overrides MNT_NOWAIT. 667 */ 668 if (((SCARG(uap, flags) & MNT_NOWAIT) == 0 || 669 (SCARG(uap, flags) & MNT_WAIT)) && 670 (error = VFS_STATFS(mp, sp, p)) != 0) { 671 simple_lock(&mountlist_slock); 672 nmp = mp->mnt_list.cqe_next; 673 vfs_unbusy(mp); 674 continue; 675 } 676 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 677 error = copyout(sp, sfsp, sizeof(*sp)); 678 if (error) { 679 vfs_unbusy(mp); 680 return (error); 681 } 682 sfsp += sizeof(*sp); 683 } 684 count++; 685 simple_lock(&mountlist_slock); 686 nmp = mp->mnt_list.cqe_next; 687 vfs_unbusy(mp); 688 } 689 simple_unlock(&mountlist_slock); 690 if (sfsp && count > maxcount) 691 *retval = maxcount; 692 else 693 *retval = count; 694 return (0); 695 } 696 697 /* 698 * Change current working directory to a given file descriptor. 699 */ 700 /* ARGSUSED */ 701 int 702 sys_fchdir(p, v, retval) 703 struct proc *p; 704 void *v; 705 register_t *retval; 706 { 707 struct sys_fchdir_args /* { 708 syscallarg(int) fd; 709 } */ *uap = v; 710 struct filedesc *fdp = p->p_fd; 711 struct cwdinfo *cwdi = p->p_cwdi; 712 struct vnode *vp, *tdp; 713 struct mount *mp; 714 struct file *fp; 715 int error; 716 717 /* getvnode() will use the descriptor for us */ 718 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 719 return (error); 720 vp = (struct vnode *)fp->f_data; 721 722 VREF(vp); 723 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 724 if (vp->v_type != VDIR) 725 error = ENOTDIR; 726 else 727 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 728 while (!error && (mp = vp->v_mountedhere) != NULL) { 729 if (vfs_busy(mp, 0, 0)) 730 continue; 731 error = VFS_ROOT(mp, &tdp); 732 vfs_unbusy(mp); 733 if (error) 734 break; 735 vput(vp); 736 vp = tdp; 737 } 738 if (error) { 739 vput(vp); 740 goto out; 741 } 742 VOP_UNLOCK(vp, 0); 743 744 /* 745 * Disallow changing to a directory not under the process's 746 * current root directory (if there is one). 747 */ 748 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, p)) { 749 vrele(vp); 750 error = EPERM; /* operation not permitted */ 751 goto out; 752 } 753 754 vrele(cwdi->cwdi_cdir); 755 cwdi->cwdi_cdir = vp; 756 out: 757 FILE_UNUSE(fp, p); 758 return (error); 759 } 760 761 /* 762 * Change this process's notion of the root directory to a given file descriptor. 763 */ 764 765 int 766 sys_fchroot(p, v, retval) 767 struct proc *p; 768 void *v; 769 register_t *retval; 770 { 771 struct sys_fchroot_args *uap = v; 772 struct filedesc *fdp = p->p_fd; 773 struct cwdinfo *cwdi = p->p_cwdi; 774 struct vnode *vp; 775 struct file *fp; 776 int error; 777 778 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 779 return error; 780 /* getvnode() will use the descriptor for us */ 781 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 782 return error; 783 vp = (struct vnode *) fp->f_data; 784 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 785 if (vp->v_type != VDIR) 786 error = ENOTDIR; 787 else 788 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 789 VOP_UNLOCK(vp, 0); 790 if (error) 791 goto out; 792 VREF(vp); 793 794 /* 795 * Prevent escaping from chroot by putting the root under 796 * the working directory. Silently chdir to / if we aren't 797 * already there. 798 */ 799 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 800 /* 801 * XXX would be more failsafe to change directory to a 802 * deadfs node here instead 803 */ 804 vrele(cwdi->cwdi_cdir); 805 VREF(vp); 806 cwdi->cwdi_cdir = vp; 807 } 808 809 if (cwdi->cwdi_rdir != NULL) 810 vrele(cwdi->cwdi_rdir); 811 cwdi->cwdi_rdir = vp; 812 out: 813 FILE_UNUSE(fp, p); 814 return (error); 815 } 816 817 818 819 /* 820 * Change current working directory (``.''). 821 */ 822 /* ARGSUSED */ 823 int 824 sys_chdir(p, v, retval) 825 struct proc *p; 826 void *v; 827 register_t *retval; 828 { 829 struct sys_chdir_args /* { 830 syscallarg(const char *) path; 831 } */ *uap = v; 832 struct cwdinfo *cwdi = p->p_cwdi; 833 int error; 834 struct nameidata nd; 835 836 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 837 SCARG(uap, path), p); 838 if ((error = change_dir(&nd, p)) != 0) 839 return (error); 840 vrele(cwdi->cwdi_cdir); 841 cwdi->cwdi_cdir = nd.ni_vp; 842 return (0); 843 } 844 845 /* 846 * Change notion of root (``/'') directory. 847 */ 848 /* ARGSUSED */ 849 int 850 sys_chroot(p, v, retval) 851 struct proc *p; 852 void *v; 853 register_t *retval; 854 { 855 struct sys_chroot_args /* { 856 syscallarg(const char *) path; 857 } */ *uap = v; 858 struct cwdinfo *cwdi = p->p_cwdi; 859 struct vnode *vp; 860 int error; 861 struct nameidata nd; 862 863 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 864 return (error); 865 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 866 SCARG(uap, path), p); 867 if ((error = change_dir(&nd, p)) != 0) 868 return (error); 869 if (cwdi->cwdi_rdir != NULL) 870 vrele(cwdi->cwdi_rdir); 871 vp = nd.ni_vp; 872 cwdi->cwdi_rdir = vp; 873 874 /* 875 * Prevent escaping from chroot by putting the root under 876 * the working directory. Silently chdir to / if we aren't 877 * already there. 878 */ 879 if (!vn_isunder(cwdi->cwdi_cdir, vp, p)) { 880 /* 881 * XXX would be more failsafe to change directory to a 882 * deadfs node here instead 883 */ 884 vrele(cwdi->cwdi_cdir); 885 VREF(vp); 886 cwdi->cwdi_cdir = vp; 887 } 888 889 return (0); 890 } 891 892 /* 893 * Common routine for chroot and chdir. 894 */ 895 static int 896 change_dir(ndp, p) 897 register struct nameidata *ndp; 898 struct proc *p; 899 { 900 struct vnode *vp; 901 int error; 902 903 if ((error = namei(ndp)) != 0) 904 return (error); 905 vp = ndp->ni_vp; 906 if (vp->v_type != VDIR) 907 error = ENOTDIR; 908 else 909 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, p); 910 911 if (error) 912 vput(vp); 913 else 914 VOP_UNLOCK(vp, 0); 915 return (error); 916 } 917 918 /* 919 * Check permissions, allocate an open file structure, 920 * and call the device open routine if any. 921 */ 922 int 923 sys_open(p, v, retval) 924 struct proc *p; 925 void *v; 926 register_t *retval; 927 { 928 register struct sys_open_args /* { 929 syscallarg(const char *) path; 930 syscallarg(int) flags; 931 syscallarg(int) mode; 932 } */ *uap = v; 933 struct cwdinfo *cwdi = p->p_cwdi; 934 struct filedesc *fdp = p->p_fd; 935 struct file *fp; 936 struct vnode *vp; 937 int flags, cmode; 938 int type, indx, error; 939 struct flock lf; 940 struct nameidata nd; 941 extern struct fileops vnops; 942 943 flags = FFLAGS(SCARG(uap, flags)); 944 if ((flags & (FREAD | FWRITE)) == 0) 945 return (EINVAL); 946 /* falloc() will use the file descriptor for us */ 947 if ((error = falloc(p, &fp, &indx)) != 0) 948 return (error); 949 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 950 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 951 p->p_dupfd = -indx - 1; /* XXX check for fdopen */ 952 if ((error = vn_open(&nd, flags, cmode)) != 0) { 953 FILE_UNUSE(fp, p); 954 ffree(fp); 955 if ((error == ENODEV || error == ENXIO) && 956 p->p_dupfd >= 0 && /* XXX from fdopen */ 957 (error = 958 dupfdopen(p, indx, p->p_dupfd, flags, error)) == 0) { 959 *retval = indx; 960 return (0); 961 } 962 if (error == ERESTART) 963 error = EINTR; 964 fdp->fd_ofiles[indx] = NULL; 965 return (error); 966 } 967 p->p_dupfd = 0; 968 vp = nd.ni_vp; 969 fp->f_flag = flags & FMASK; 970 fp->f_type = DTYPE_VNODE; 971 fp->f_ops = &vnops; 972 fp->f_data = (caddr_t)vp; 973 if (flags & (O_EXLOCK | O_SHLOCK)) { 974 lf.l_whence = SEEK_SET; 975 lf.l_start = 0; 976 lf.l_len = 0; 977 if (flags & O_EXLOCK) 978 lf.l_type = F_WRLCK; 979 else 980 lf.l_type = F_RDLCK; 981 type = F_FLOCK; 982 if ((flags & FNONBLOCK) == 0) 983 type |= F_WAIT; 984 VOP_UNLOCK(vp, 0); 985 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type); 986 if (error) { 987 (void) vn_close(vp, fp->f_flag, fp->f_cred, p); 988 FILE_UNUSE(fp, p); 989 ffree(fp); 990 fdp->fd_ofiles[indx] = NULL; 991 return (error); 992 } 993 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 994 fp->f_flag |= FHASLOCK; 995 } 996 VOP_UNLOCK(vp, 0); 997 *retval = indx; 998 FILE_UNUSE(fp, p); 999 return (0); 1000 } 1001 1002 /* 1003 * Create a special file. 1004 */ 1005 /* ARGSUSED */ 1006 int 1007 sys_mknod(p, v, retval) 1008 struct proc *p; 1009 void *v; 1010 register_t *retval; 1011 { 1012 register struct sys_mknod_args /* { 1013 syscallarg(const char *) path; 1014 syscallarg(int) mode; 1015 syscallarg(int) dev; 1016 } */ *uap = v; 1017 register struct vnode *vp; 1018 struct vattr vattr; 1019 int error; 1020 int whiteout = 0; 1021 struct nameidata nd; 1022 1023 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1024 return (error); 1025 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1026 if ((error = namei(&nd)) != 0) 1027 return (error); 1028 vp = nd.ni_vp; 1029 if (vp != NULL) 1030 error = EEXIST; 1031 else { 1032 VATTR_NULL(&vattr); 1033 vattr.va_mode = 1034 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1035 vattr.va_rdev = SCARG(uap, dev); 1036 whiteout = 0; 1037 1038 switch (SCARG(uap, mode) & S_IFMT) { 1039 case S_IFMT: /* used by badsect to flag bad sectors */ 1040 vattr.va_type = VBAD; 1041 break; 1042 case S_IFCHR: 1043 vattr.va_type = VCHR; 1044 break; 1045 case S_IFBLK: 1046 vattr.va_type = VBLK; 1047 break; 1048 case S_IFWHT: 1049 whiteout = 1; 1050 break; 1051 default: 1052 error = EINVAL; 1053 break; 1054 } 1055 } 1056 if (!error) { 1057 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1058 if (whiteout) { 1059 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1060 if (error) 1061 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1062 vput(nd.ni_dvp); 1063 } else { 1064 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1065 &nd.ni_cnd, &vattr); 1066 } 1067 } else { 1068 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1069 if (nd.ni_dvp == vp) 1070 vrele(nd.ni_dvp); 1071 else 1072 vput(nd.ni_dvp); 1073 if (vp) 1074 vrele(vp); 1075 } 1076 return (error); 1077 } 1078 1079 /* 1080 * Create a named pipe. 1081 */ 1082 /* ARGSUSED */ 1083 int 1084 sys_mkfifo(p, v, retval) 1085 struct proc *p; 1086 void *v; 1087 register_t *retval; 1088 { 1089 register struct sys_mkfifo_args /* { 1090 syscallarg(const char *) path; 1091 syscallarg(int) mode; 1092 } */ *uap = v; 1093 struct vattr vattr; 1094 int error; 1095 struct nameidata nd; 1096 1097 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 1098 if ((error = namei(&nd)) != 0) 1099 return (error); 1100 if (nd.ni_vp != NULL) { 1101 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1102 if (nd.ni_dvp == nd.ni_vp) 1103 vrele(nd.ni_dvp); 1104 else 1105 vput(nd.ni_dvp); 1106 vrele(nd.ni_vp); 1107 return (EEXIST); 1108 } 1109 VATTR_NULL(&vattr); 1110 vattr.va_type = VFIFO; 1111 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1112 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1113 return (VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr)); 1114 } 1115 1116 /* 1117 * Make a hard file link. 1118 */ 1119 /* ARGSUSED */ 1120 int 1121 sys_link(p, v, retval) 1122 struct proc *p; 1123 void *v; 1124 register_t *retval; 1125 { 1126 register struct sys_link_args /* { 1127 syscallarg(const char *) path; 1128 syscallarg(const char *) link; 1129 } */ *uap = v; 1130 register struct vnode *vp; 1131 struct nameidata nd; 1132 int error; 1133 1134 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1135 if ((error = namei(&nd)) != 0) 1136 return (error); 1137 vp = nd.ni_vp; 1138 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1139 if ((error = namei(&nd)) != 0) 1140 goto out; 1141 if (nd.ni_vp) { 1142 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1143 if (nd.ni_dvp == nd.ni_vp) 1144 vrele(nd.ni_dvp); 1145 else 1146 vput(nd.ni_dvp); 1147 vrele(nd.ni_vp); 1148 error = EEXIST; 1149 goto out; 1150 } 1151 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1152 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1153 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1154 out: 1155 vrele(vp); 1156 return (error); 1157 } 1158 1159 /* 1160 * Make a symbolic link. 1161 */ 1162 /* ARGSUSED */ 1163 int 1164 sys_symlink(p, v, retval) 1165 struct proc *p; 1166 void *v; 1167 register_t *retval; 1168 { 1169 register struct sys_symlink_args /* { 1170 syscallarg(const char *) path; 1171 syscallarg(const char *) link; 1172 } */ *uap = v; 1173 struct vattr vattr; 1174 char *path; 1175 int error; 1176 struct nameidata nd; 1177 1178 MALLOC(path, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); 1179 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1180 if (error) 1181 goto out; 1182 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), p); 1183 if ((error = namei(&nd)) != 0) 1184 goto out; 1185 if (nd.ni_vp) { 1186 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1187 if (nd.ni_dvp == nd.ni_vp) 1188 vrele(nd.ni_dvp); 1189 else 1190 vput(nd.ni_dvp); 1191 vrele(nd.ni_vp); 1192 error = EEXIST; 1193 goto out; 1194 } 1195 VATTR_NULL(&vattr); 1196 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1197 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1198 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1199 out: 1200 FREE(path, M_NAMEI); 1201 return (error); 1202 } 1203 1204 /* 1205 * Delete a whiteout from the filesystem. 1206 */ 1207 /* ARGSUSED */ 1208 int 1209 sys_undelete(p, v, retval) 1210 struct proc *p; 1211 void *v; 1212 register_t *retval; 1213 { 1214 register struct sys_undelete_args /* { 1215 syscallarg(const char *) path; 1216 } */ *uap = v; 1217 int error; 1218 struct nameidata nd; 1219 1220 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1221 SCARG(uap, path), p); 1222 error = namei(&nd); 1223 if (error) 1224 return (error); 1225 1226 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1227 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1228 if (nd.ni_dvp == nd.ni_vp) 1229 vrele(nd.ni_dvp); 1230 else 1231 vput(nd.ni_dvp); 1232 if (nd.ni_vp) 1233 vrele(nd.ni_vp); 1234 return (EEXIST); 1235 } 1236 1237 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1238 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1239 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1240 vput(nd.ni_dvp); 1241 return (error); 1242 } 1243 1244 /* 1245 * Delete a name from the filesystem. 1246 */ 1247 /* ARGSUSED */ 1248 int 1249 sys_unlink(p, v, retval) 1250 struct proc *p; 1251 void *v; 1252 register_t *retval; 1253 { 1254 struct sys_unlink_args /* { 1255 syscallarg(const char *) path; 1256 } */ *uap = v; 1257 register struct vnode *vp; 1258 int error; 1259 struct nameidata nd; 1260 1261 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1262 SCARG(uap, path), p); 1263 if ((error = namei(&nd)) != 0) 1264 return (error); 1265 vp = nd.ni_vp; 1266 1267 /* 1268 * The root of a mounted filesystem cannot be deleted. 1269 */ 1270 if (vp->v_flag & VROOT) { 1271 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1272 if (nd.ni_dvp == vp) 1273 vrele(nd.ni_dvp); 1274 else 1275 vput(nd.ni_dvp); 1276 vput(vp); 1277 error = EBUSY; 1278 goto out; 1279 } 1280 1281 (void)uvm_vnp_uncache(vp); 1282 1283 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 1284 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1285 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1286 out: 1287 return (error); 1288 } 1289 1290 /* 1291 * Reposition read/write file offset. 1292 */ 1293 int 1294 sys_lseek(p, v, retval) 1295 struct proc *p; 1296 void *v; 1297 register_t *retval; 1298 { 1299 register struct sys_lseek_args /* { 1300 syscallarg(int) fd; 1301 syscallarg(int) pad; 1302 syscallarg(off_t) offset; 1303 syscallarg(int) whence; 1304 } */ *uap = v; 1305 struct ucred *cred = p->p_ucred; 1306 register struct filedesc *fdp = p->p_fd; 1307 register struct file *fp; 1308 struct vnode *vp; 1309 struct vattr vattr; 1310 register off_t newoff; 1311 int error; 1312 1313 if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles || 1314 (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL || 1315 (fp->f_iflags & FIF_WANTCLOSE) != 0) 1316 return (EBADF); 1317 1318 FILE_USE(fp); 1319 1320 vp = (struct vnode *)fp->f_data; 1321 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1322 error = ESPIPE; 1323 goto out; 1324 } 1325 1326 switch (SCARG(uap, whence)) { 1327 case SEEK_CUR: 1328 newoff = fp->f_offset + SCARG(uap, offset); 1329 break; 1330 case SEEK_END: 1331 error = VOP_GETATTR(vp, &vattr, cred, p); 1332 if (error) 1333 goto out; 1334 newoff = SCARG(uap, offset) + vattr.va_size; 1335 break; 1336 case SEEK_SET: 1337 newoff = SCARG(uap, offset); 1338 break; 1339 default: 1340 error = EINVAL; 1341 goto out; 1342 } 1343 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1344 goto out; 1345 1346 *(off_t *)retval = fp->f_offset = newoff; 1347 out: 1348 FILE_UNUSE(fp, p); 1349 return (error); 1350 } 1351 1352 /* 1353 * Positional read system call. 1354 */ 1355 int 1356 sys_pread(p, v, retval) 1357 struct proc *p; 1358 void *v; 1359 register_t *retval; 1360 { 1361 struct sys_pread_args /* { 1362 syscallarg(int) fd; 1363 syscallarg(void *) buf; 1364 syscallarg(size_t) nbyte; 1365 syscallarg(off_t) offset; 1366 } */ *uap = v; 1367 struct filedesc *fdp = p->p_fd; 1368 struct file *fp; 1369 struct vnode *vp; 1370 off_t offset; 1371 int error, fd = SCARG(uap, fd); 1372 1373 if ((u_int)fd >= fdp->fd_nfiles || 1374 (fp = fdp->fd_ofiles[fd]) == NULL || 1375 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 1376 (fp->f_flag & FREAD) == 0) 1377 return (EBADF); 1378 1379 FILE_USE(fp); 1380 1381 vp = (struct vnode *)fp->f_data; 1382 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1383 error = ESPIPE; 1384 goto out; 1385 } 1386 1387 offset = SCARG(uap, offset); 1388 1389 /* 1390 * XXX This works because no file systems actually 1391 * XXX take any action on the seek operation. 1392 */ 1393 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1394 goto out; 1395 1396 /* dofileread() will unuse the descriptor for us */ 1397 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1398 &offset, 0, retval)); 1399 1400 out: 1401 FILE_UNUSE(fp, p); 1402 return (error); 1403 } 1404 1405 /* 1406 * Positional scatter read system call. 1407 */ 1408 int 1409 sys_preadv(p, v, retval) 1410 struct proc *p; 1411 void *v; 1412 register_t *retval; 1413 { 1414 struct sys_preadv_args /* { 1415 syscallarg(int) fd; 1416 syscallarg(const struct iovec *) iovp; 1417 syscallarg(int) iovcnt; 1418 syscallarg(off_t) offset; 1419 } */ *uap = v; 1420 struct filedesc *fdp = p->p_fd; 1421 struct file *fp; 1422 struct vnode *vp; 1423 off_t offset; 1424 int error, fd = SCARG(uap, fd); 1425 1426 if ((u_int)fd >= fdp->fd_nfiles || 1427 (fp = fdp->fd_ofiles[fd]) == NULL || 1428 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 1429 (fp->f_flag & FREAD) == 0) 1430 return (EBADF); 1431 1432 FILE_USE(fp); 1433 1434 vp = (struct vnode *)fp->f_data; 1435 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1436 error = ESPIPE; 1437 goto out; 1438 } 1439 1440 offset = SCARG(uap, offset); 1441 1442 /* 1443 * XXX This works because no file systems actually 1444 * XXX take any action on the seek operation. 1445 */ 1446 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1447 goto out; 1448 1449 /* dofilereadv() will unuse the descriptor for us */ 1450 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1451 &offset, 0, retval)); 1452 1453 out: 1454 FILE_UNUSE(fp, p); 1455 return (error); 1456 } 1457 1458 /* 1459 * Positional write system call. 1460 */ 1461 int 1462 sys_pwrite(p, v, retval) 1463 struct proc *p; 1464 void *v; 1465 register_t *retval; 1466 { 1467 struct sys_pwrite_args /* { 1468 syscallarg(int) fd; 1469 syscallarg(const void *) buf; 1470 syscallarg(size_t) nbyte; 1471 syscallarg(off_t) offset; 1472 } */ *uap = v; 1473 struct filedesc *fdp = p->p_fd; 1474 struct file *fp; 1475 struct vnode *vp; 1476 off_t offset; 1477 int error, fd = SCARG(uap, fd); 1478 1479 if ((u_int)fd >= fdp->fd_nfiles || 1480 (fp = fdp->fd_ofiles[fd]) == NULL || 1481 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 1482 (fp->f_flag & FWRITE) == 0) 1483 return (EBADF); 1484 1485 FILE_USE(fp); 1486 1487 vp = (struct vnode *)fp->f_data; 1488 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1489 error = ESPIPE; 1490 goto out; 1491 } 1492 1493 offset = SCARG(uap, offset); 1494 1495 /* 1496 * XXX This works because no file systems actually 1497 * XXX take any action on the seek operation. 1498 */ 1499 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1500 goto out; 1501 1502 /* dofilewrite() will unuse the descriptor for us */ 1503 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1504 &offset, 0, retval)); 1505 1506 out: 1507 FILE_UNUSE(fp, p); 1508 return (error); 1509 } 1510 1511 /* 1512 * Positional gather write system call. 1513 */ 1514 int 1515 sys_pwritev(p, v, retval) 1516 struct proc *p; 1517 void *v; 1518 register_t *retval; 1519 { 1520 struct sys_pwritev_args /* { 1521 syscallarg(int) fd; 1522 syscallarg(const struct iovec *) iovp; 1523 syscallarg(int) iovcnt; 1524 syscallarg(off_t) offset; 1525 } */ *uap = v; 1526 struct filedesc *fdp = p->p_fd; 1527 struct file *fp; 1528 struct vnode *vp; 1529 off_t offset; 1530 int error, fd = SCARG(uap, fd); 1531 1532 if ((u_int)fd >= fdp->fd_nfiles || 1533 (fp = fdp->fd_ofiles[fd]) == NULL || 1534 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 1535 (fp->f_flag & FWRITE) == 0) 1536 return (EBADF); 1537 1538 FILE_USE(fp); 1539 1540 vp = (struct vnode *)fp->f_data; 1541 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1542 error = ESPIPE; 1543 goto out; 1544 } 1545 1546 offset = SCARG(uap, offset); 1547 1548 /* 1549 * XXX This works because no file systems actually 1550 * XXX take any action on the seek operation. 1551 */ 1552 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1553 goto out; 1554 1555 /* dofilewritev() will unuse the descriptor for us */ 1556 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1557 &offset, 0, retval)); 1558 1559 out: 1560 FILE_UNUSE(fp, p); 1561 return (error); 1562 } 1563 1564 /* 1565 * Check access permissions. 1566 */ 1567 int 1568 sys_access(p, v, retval) 1569 struct proc *p; 1570 void *v; 1571 register_t *retval; 1572 { 1573 register struct sys_access_args /* { 1574 syscallarg(const char *) path; 1575 syscallarg(int) flags; 1576 } */ *uap = v; 1577 register struct ucred *cred = p->p_ucred; 1578 register struct vnode *vp; 1579 int error, flags, t_gid, t_uid; 1580 struct nameidata nd; 1581 1582 t_uid = cred->cr_uid; 1583 t_gid = cred->cr_gid; 1584 cred->cr_uid = p->p_cred->p_ruid; 1585 cred->cr_gid = p->p_cred->p_rgid; 1586 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1587 SCARG(uap, path), p); 1588 if ((error = namei(&nd)) != 0) 1589 goto out1; 1590 vp = nd.ni_vp; 1591 1592 /* Flags == 0 means only check for existence. */ 1593 if (SCARG(uap, flags)) { 1594 flags = 0; 1595 if (SCARG(uap, flags) & R_OK) 1596 flags |= VREAD; 1597 if (SCARG(uap, flags) & W_OK) 1598 flags |= VWRITE; 1599 if (SCARG(uap, flags) & X_OK) 1600 flags |= VEXEC; 1601 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 1602 error = VOP_ACCESS(vp, flags, cred, p); 1603 } 1604 vput(vp); 1605 out1: 1606 cred->cr_uid = t_uid; 1607 cred->cr_gid = t_gid; 1608 return (error); 1609 } 1610 1611 /* 1612 * Get file status; this version follows links. 1613 */ 1614 /* ARGSUSED */ 1615 int 1616 sys___stat13(p, v, retval) 1617 struct proc *p; 1618 void *v; 1619 register_t *retval; 1620 { 1621 register struct sys___stat13_args /* { 1622 syscallarg(const char *) path; 1623 syscallarg(struct stat *) ub; 1624 } */ *uap = v; 1625 struct stat sb; 1626 int error; 1627 struct nameidata nd; 1628 1629 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1630 SCARG(uap, path), p); 1631 if ((error = namei(&nd)) != 0) 1632 return (error); 1633 error = vn_stat(nd.ni_vp, &sb, p); 1634 vput(nd.ni_vp); 1635 if (error) 1636 return (error); 1637 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 1638 return (error); 1639 } 1640 1641 /* 1642 * Get file status; this version does not follow links. 1643 */ 1644 /* ARGSUSED */ 1645 int 1646 sys___lstat13(p, v, retval) 1647 struct proc *p; 1648 void *v; 1649 register_t *retval; 1650 { 1651 register struct sys___lstat13_args /* { 1652 syscallarg(const char *) path; 1653 syscallarg(struct stat *) ub; 1654 } */ *uap = v; 1655 struct stat sb; 1656 int error; 1657 struct nameidata nd; 1658 1659 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 1660 SCARG(uap, path), p); 1661 if ((error = namei(&nd)) != 0) 1662 return (error); 1663 error = vn_stat(nd.ni_vp, &sb, p); 1664 vput(nd.ni_vp); 1665 if (error) 1666 return (error); 1667 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 1668 return (error); 1669 } 1670 1671 /* 1672 * Get configurable pathname variables. 1673 */ 1674 /* ARGSUSED */ 1675 int 1676 sys_pathconf(p, v, retval) 1677 struct proc *p; 1678 void *v; 1679 register_t *retval; 1680 { 1681 register struct sys_pathconf_args /* { 1682 syscallarg(const char *) path; 1683 syscallarg(int) name; 1684 } */ *uap = v; 1685 int error; 1686 struct nameidata nd; 1687 1688 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1689 SCARG(uap, path), p); 1690 if ((error = namei(&nd)) != 0) 1691 return (error); 1692 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 1693 vput(nd.ni_vp); 1694 return (error); 1695 } 1696 1697 /* 1698 * Return target name of a symbolic link. 1699 */ 1700 /* ARGSUSED */ 1701 int 1702 sys_readlink(p, v, retval) 1703 struct proc *p; 1704 void *v; 1705 register_t *retval; 1706 { 1707 register struct sys_readlink_args /* { 1708 syscallarg(const char *) path; 1709 syscallarg(char *) buf; 1710 syscallarg(size_t) count; 1711 } */ *uap = v; 1712 register struct vnode *vp; 1713 struct iovec aiov; 1714 struct uio auio; 1715 int error; 1716 struct nameidata nd; 1717 1718 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 1719 SCARG(uap, path), p); 1720 if ((error = namei(&nd)) != 0) 1721 return (error); 1722 vp = nd.ni_vp; 1723 if (vp->v_type != VLNK) 1724 error = EINVAL; 1725 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 1726 (error = VOP_ACCESS(vp, VREAD, p->p_ucred, p)) == 0) { 1727 aiov.iov_base = SCARG(uap, buf); 1728 aiov.iov_len = SCARG(uap, count); 1729 auio.uio_iov = &aiov; 1730 auio.uio_iovcnt = 1; 1731 auio.uio_offset = 0; 1732 auio.uio_rw = UIO_READ; 1733 auio.uio_segflg = UIO_USERSPACE; 1734 auio.uio_procp = p; 1735 auio.uio_resid = SCARG(uap, count); 1736 error = VOP_READLINK(vp, &auio, p->p_ucred); 1737 } 1738 vput(vp); 1739 *retval = SCARG(uap, count) - auio.uio_resid; 1740 return (error); 1741 } 1742 1743 /* 1744 * Change flags of a file given a path name. 1745 */ 1746 /* ARGSUSED */ 1747 int 1748 sys_chflags(p, v, retval) 1749 struct proc *p; 1750 void *v; 1751 register_t *retval; 1752 { 1753 register struct sys_chflags_args /* { 1754 syscallarg(const char *) path; 1755 syscallarg(u_long) flags; 1756 } */ *uap = v; 1757 register struct vnode *vp; 1758 struct vattr vattr; 1759 int error; 1760 struct nameidata nd; 1761 1762 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1763 if ((error = namei(&nd)) != 0) 1764 return (error); 1765 vp = nd.ni_vp; 1766 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1767 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1768 VATTR_NULL(&vattr); 1769 vattr.va_flags = SCARG(uap, flags); 1770 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 1771 vput(vp); 1772 return (error); 1773 } 1774 1775 /* 1776 * Change flags of a file given a file descriptor. 1777 */ 1778 /* ARGSUSED */ 1779 int 1780 sys_fchflags(p, v, retval) 1781 struct proc *p; 1782 void *v; 1783 register_t *retval; 1784 { 1785 register struct sys_fchflags_args /* { 1786 syscallarg(int) fd; 1787 syscallarg(u_long) flags; 1788 } */ *uap = v; 1789 struct vattr vattr; 1790 struct vnode *vp; 1791 struct file *fp; 1792 int error; 1793 1794 /* getvnode() will use the descriptor for us */ 1795 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 1796 return (error); 1797 vp = (struct vnode *)fp->f_data; 1798 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1799 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1800 VATTR_NULL(&vattr); 1801 vattr.va_flags = SCARG(uap, flags); 1802 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 1803 VOP_UNLOCK(vp, 0); 1804 FILE_UNUSE(fp, p); 1805 return (error); 1806 } 1807 1808 /* 1809 * Change mode of a file given path name; this version follows links. 1810 */ 1811 /* ARGSUSED */ 1812 int 1813 sys_chmod(p, v, retval) 1814 struct proc *p; 1815 void *v; 1816 register_t *retval; 1817 { 1818 register struct sys_chmod_args /* { 1819 syscallarg(const char *) path; 1820 syscallarg(int) mode; 1821 } */ *uap = v; 1822 int error; 1823 struct nameidata nd; 1824 1825 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1826 if ((error = namei(&nd)) != 0) 1827 return (error); 1828 1829 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 1830 1831 vrele(nd.ni_vp); 1832 return (error); 1833 } 1834 1835 /* 1836 * Change mode of a file given a file descriptor. 1837 */ 1838 /* ARGSUSED */ 1839 int 1840 sys_fchmod(p, v, retval) 1841 struct proc *p; 1842 void *v; 1843 register_t *retval; 1844 { 1845 register struct sys_fchmod_args /* { 1846 syscallarg(int) fd; 1847 syscallarg(int) mode; 1848 } */ *uap = v; 1849 struct file *fp; 1850 int error; 1851 1852 /* getvnode() will use the descriptor for us */ 1853 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 1854 return (error); 1855 1856 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), p); 1857 FILE_UNUSE(fp, p); 1858 return (error); 1859 } 1860 1861 /* 1862 * Change mode of a file given path name; this version does not follow links. 1863 */ 1864 /* ARGSUSED */ 1865 int 1866 sys_lchmod(p, v, retval) 1867 struct proc *p; 1868 void *v; 1869 register_t *retval; 1870 { 1871 register struct sys_lchmod_args /* { 1872 syscallarg(const char *) path; 1873 syscallarg(int) mode; 1874 } */ *uap = v; 1875 int error; 1876 struct nameidata nd; 1877 1878 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1879 if ((error = namei(&nd)) != 0) 1880 return (error); 1881 1882 error = change_mode(nd.ni_vp, SCARG(uap, mode), p); 1883 1884 vrele(nd.ni_vp); 1885 return (error); 1886 } 1887 1888 /* 1889 * Common routine to set mode given a vnode. 1890 */ 1891 static int 1892 change_mode(vp, mode, p) 1893 struct vnode *vp; 1894 int mode; 1895 struct proc *p; 1896 { 1897 struct vattr vattr; 1898 int error; 1899 1900 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 1901 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1902 VATTR_NULL(&vattr); 1903 vattr.va_mode = mode & ALLPERMS; 1904 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 1905 VOP_UNLOCK(vp, 0); 1906 return (error); 1907 } 1908 1909 /* 1910 * Set ownership given a path name; this version follows links. 1911 */ 1912 /* ARGSUSED */ 1913 int 1914 sys_chown(p, v, retval) 1915 struct proc *p; 1916 void *v; 1917 register_t *retval; 1918 { 1919 register struct sys_chown_args /* { 1920 syscallarg(const char *) path; 1921 syscallarg(uid_t) uid; 1922 syscallarg(gid_t) gid; 1923 } */ *uap = v; 1924 int error; 1925 struct nameidata nd; 1926 1927 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1928 if ((error = namei(&nd)) != 0) 1929 return (error); 1930 1931 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 1932 1933 vrele(nd.ni_vp); 1934 return (error); 1935 } 1936 1937 /* 1938 * Set ownership given a path name; this version follows links. 1939 * Provides POSIX semantics. 1940 */ 1941 /* ARGSUSED */ 1942 int 1943 sys___posix_chown(p, v, retval) 1944 struct proc *p; 1945 void *v; 1946 register_t *retval; 1947 { 1948 register struct sys_chown_args /* { 1949 syscallarg(const char *) path; 1950 syscallarg(uid_t) uid; 1951 syscallarg(gid_t) gid; 1952 } */ *uap = v; 1953 int error; 1954 struct nameidata nd; 1955 1956 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 1957 if ((error = namei(&nd)) != 0) 1958 return (error); 1959 1960 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 1961 1962 vrele(nd.ni_vp); 1963 return (error); 1964 } 1965 1966 /* 1967 * Set ownership given a file descriptor. 1968 */ 1969 /* ARGSUSED */ 1970 int 1971 sys_fchown(p, v, retval) 1972 struct proc *p; 1973 void *v; 1974 register_t *retval; 1975 { 1976 register struct sys_fchown_args /* { 1977 syscallarg(int) fd; 1978 syscallarg(uid_t) uid; 1979 syscallarg(gid_t) gid; 1980 } */ *uap = v; 1981 int error; 1982 struct file *fp; 1983 1984 /* getvnode() will use the descriptor for us */ 1985 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 1986 return (error); 1987 1988 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 1989 SCARG(uap, gid), p, 0); 1990 FILE_UNUSE(fp, p); 1991 return (error); 1992 } 1993 1994 /* 1995 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 1996 */ 1997 /* ARGSUSED */ 1998 int 1999 sys___posix_fchown(p, v, retval) 2000 struct proc *p; 2001 void *v; 2002 register_t *retval; 2003 { 2004 register struct sys_fchown_args /* { 2005 syscallarg(int) fd; 2006 syscallarg(uid_t) uid; 2007 syscallarg(gid_t) gid; 2008 } */ *uap = v; 2009 int error; 2010 struct file *fp; 2011 2012 /* getvnode() will use the descriptor for us */ 2013 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2014 return (error); 2015 2016 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2017 SCARG(uap, gid), p, 1); 2018 FILE_UNUSE(fp, p); 2019 return (error); 2020 } 2021 2022 /* 2023 * Set ownership given a path name; this version does not follow links. 2024 */ 2025 /* ARGSUSED */ 2026 int 2027 sys_lchown(p, v, retval) 2028 struct proc *p; 2029 void *v; 2030 register_t *retval; 2031 { 2032 register struct sys_lchown_args /* { 2033 syscallarg(const char *) path; 2034 syscallarg(uid_t) uid; 2035 syscallarg(gid_t) gid; 2036 } */ *uap = v; 2037 int error; 2038 struct nameidata nd; 2039 2040 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2041 if ((error = namei(&nd)) != 0) 2042 return (error); 2043 2044 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 0); 2045 2046 vrele(nd.ni_vp); 2047 return (error); 2048 } 2049 2050 /* 2051 * Set ownership given a path name; this version does not follow links. 2052 * Provides POSIX/XPG semantics. 2053 */ 2054 /* ARGSUSED */ 2055 int 2056 sys___posix_lchown(p, v, retval) 2057 struct proc *p; 2058 void *v; 2059 register_t *retval; 2060 { 2061 register struct sys_lchown_args /* { 2062 syscallarg(const char *) path; 2063 syscallarg(uid_t) uid; 2064 syscallarg(gid_t) gid; 2065 } */ *uap = v; 2066 int error; 2067 struct nameidata nd; 2068 2069 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2070 if ((error = namei(&nd)) != 0) 2071 return (error); 2072 2073 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), p, 1); 2074 2075 vrele(nd.ni_vp); 2076 return (error); 2077 } 2078 2079 /* 2080 * Common routine to set ownership given a vnode. 2081 */ 2082 static int 2083 change_owner(vp, uid, gid, p, posix_semantics) 2084 register struct vnode *vp; 2085 uid_t uid; 2086 gid_t gid; 2087 struct proc *p; 2088 int posix_semantics; 2089 { 2090 struct vattr vattr; 2091 mode_t newmode; 2092 int error; 2093 2094 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2095 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2096 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2097 goto out; 2098 2099 #define CHANGED(x) ((x) != -1) 2100 newmode = vattr.va_mode; 2101 if (posix_semantics) { 2102 /* 2103 * POSIX/XPG semantics: if the caller is not the super-user, 2104 * clear set-user-id and set-group-id bits. Both POSIX and 2105 * the XPG consider the behaviour for calls by the super-user 2106 * implementation-defined; we leave the set-user-id and set- 2107 * group-id settings intact in that case. 2108 */ 2109 if (suser(p->p_ucred, NULL) != 0) 2110 newmode &= ~(S_ISUID | S_ISGID); 2111 } else { 2112 /* 2113 * NetBSD semantics: when changing owner and/or group, 2114 * clear the respective bit(s). 2115 */ 2116 if (CHANGED(uid)) 2117 newmode &= ~S_ISUID; 2118 if (CHANGED(gid)) 2119 newmode &= ~S_ISGID; 2120 } 2121 /* Update va_mode iff altered. */ 2122 if (vattr.va_mode == newmode) 2123 newmode = VNOVAL; 2124 2125 VATTR_NULL(&vattr); 2126 vattr.va_uid = CHANGED(uid) ? uid : VNOVAL; 2127 vattr.va_gid = CHANGED(gid) ? gid : VNOVAL; 2128 vattr.va_mode = newmode; 2129 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2130 #undef CHANGED 2131 2132 out: 2133 VOP_UNLOCK(vp, 0); 2134 return (error); 2135 } 2136 2137 /* 2138 * Set the access and modification times given a path name; this 2139 * version follows links. 2140 */ 2141 /* ARGSUSED */ 2142 int 2143 sys_utimes(p, v, retval) 2144 struct proc *p; 2145 void *v; 2146 register_t *retval; 2147 { 2148 register struct sys_utimes_args /* { 2149 syscallarg(const char *) path; 2150 syscallarg(const struct timeval *) tptr; 2151 } */ *uap = v; 2152 int error; 2153 struct nameidata nd; 2154 2155 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2156 if ((error = namei(&nd)) != 0) 2157 return (error); 2158 2159 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2160 2161 vrele(nd.ni_vp); 2162 return (error); 2163 } 2164 2165 /* 2166 * Set the access and modification times given a file descriptor. 2167 */ 2168 /* ARGSUSED */ 2169 int 2170 sys_futimes(p, v, retval) 2171 struct proc *p; 2172 void *v; 2173 register_t *retval; 2174 { 2175 register struct sys_futimes_args /* { 2176 syscallarg(int) fd; 2177 syscallarg(const struct timeval *) tptr; 2178 } */ *uap = v; 2179 int error; 2180 struct file *fp; 2181 2182 /* getvnode() will use the descriptor for us */ 2183 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2184 return (error); 2185 2186 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), p); 2187 FILE_UNUSE(fp, p); 2188 return (error); 2189 } 2190 2191 /* 2192 * Set the access and modification times given a path name; this 2193 * version does not follow links. 2194 */ 2195 /* ARGSUSED */ 2196 int 2197 sys_lutimes(p, v, retval) 2198 struct proc *p; 2199 void *v; 2200 register_t *retval; 2201 { 2202 register struct sys_lutimes_args /* { 2203 syscallarg(const char *) path; 2204 syscallarg(const struct timeval *) tptr; 2205 } */ *uap = v; 2206 int error; 2207 struct nameidata nd; 2208 2209 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2210 if ((error = namei(&nd)) != 0) 2211 return (error); 2212 2213 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), p); 2214 2215 vrele(nd.ni_vp); 2216 return (error); 2217 } 2218 2219 /* 2220 * Common routine to set access and modification times given a vnode. 2221 */ 2222 static int 2223 change_utimes(vp, tptr, p) 2224 struct vnode *vp; 2225 const struct timeval *tptr; 2226 struct proc *p; 2227 { 2228 struct timeval tv[2]; 2229 struct vattr vattr; 2230 int error; 2231 2232 VATTR_NULL(&vattr); 2233 if (tptr == NULL) { 2234 microtime(&tv[0]); 2235 tv[1] = tv[0]; 2236 vattr.va_vaflags |= VA_UTIMES_NULL; 2237 } else { 2238 error = copyin(tptr, tv, sizeof(tv)); 2239 if (error) 2240 return (error); 2241 } 2242 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2243 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2244 vattr.va_atime.tv_sec = tv[0].tv_sec; 2245 vattr.va_atime.tv_nsec = tv[0].tv_usec * 1000; 2246 vattr.va_mtime.tv_sec = tv[1].tv_sec; 2247 vattr.va_mtime.tv_nsec = tv[1].tv_usec * 1000; 2248 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2249 VOP_UNLOCK(vp, 0); 2250 return (error); 2251 } 2252 2253 /* 2254 * Truncate a file given its path name. 2255 */ 2256 /* ARGSUSED */ 2257 int 2258 sys_truncate(p, v, retval) 2259 struct proc *p; 2260 void *v; 2261 register_t *retval; 2262 { 2263 register struct sys_truncate_args /* { 2264 syscallarg(const char *) path; 2265 syscallarg(int) pad; 2266 syscallarg(off_t) length; 2267 } */ *uap = v; 2268 register struct vnode *vp; 2269 struct vattr vattr; 2270 int error; 2271 struct nameidata nd; 2272 2273 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2274 if ((error = namei(&nd)) != 0) 2275 return (error); 2276 vp = nd.ni_vp; 2277 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2278 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2279 if (vp->v_type == VDIR) 2280 error = EISDIR; 2281 else if ((error = vn_writechk(vp)) == 0 && 2282 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p)) == 0) { 2283 VATTR_NULL(&vattr); 2284 vattr.va_size = SCARG(uap, length); 2285 error = VOP_SETATTR(vp, &vattr, p->p_ucred, p); 2286 } 2287 vput(vp); 2288 return (error); 2289 } 2290 2291 /* 2292 * Truncate a file given a file descriptor. 2293 */ 2294 /* ARGSUSED */ 2295 int 2296 sys_ftruncate(p, v, retval) 2297 struct proc *p; 2298 void *v; 2299 register_t *retval; 2300 { 2301 register struct sys_ftruncate_args /* { 2302 syscallarg(int) fd; 2303 syscallarg(int) pad; 2304 syscallarg(off_t) length; 2305 } */ *uap = v; 2306 struct vattr vattr; 2307 struct vnode *vp; 2308 struct file *fp; 2309 int error; 2310 2311 /* getvnode() will use the descriptor for us */ 2312 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2313 return (error); 2314 if ((fp->f_flag & FWRITE) == 0) { 2315 error = EINVAL; 2316 goto out; 2317 } 2318 vp = (struct vnode *)fp->f_data; 2319 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2320 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2321 if (vp->v_type == VDIR) 2322 error = EISDIR; 2323 else if ((error = vn_writechk(vp)) == 0) { 2324 VATTR_NULL(&vattr); 2325 vattr.va_size = SCARG(uap, length); 2326 error = VOP_SETATTR(vp, &vattr, fp->f_cred, p); 2327 } 2328 VOP_UNLOCK(vp, 0); 2329 out: 2330 FILE_UNUSE(fp, p); 2331 return (error); 2332 } 2333 2334 /* 2335 * Sync an open file. 2336 */ 2337 /* ARGSUSED */ 2338 int 2339 sys_fsync(p, v, retval) 2340 struct proc *p; 2341 void *v; 2342 register_t *retval; 2343 { 2344 struct sys_fsync_args /* { 2345 syscallarg(int) fd; 2346 } */ *uap = v; 2347 register struct vnode *vp; 2348 struct file *fp; 2349 int error; 2350 2351 /* getvnode() will use the descriptor for us */ 2352 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2353 return (error); 2354 vp = (struct vnode *)fp->f_data; 2355 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2356 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, p); 2357 VOP_UNLOCK(vp, 0); 2358 FILE_UNUSE(fp, p); 2359 return (error); 2360 } 2361 2362 /* 2363 * Sync the data of an open file. 2364 */ 2365 /* ARGSUSED */ 2366 int 2367 sys_fdatasync(p, v, retval) 2368 struct proc *p; 2369 void *v; 2370 register_t *retval; 2371 { 2372 struct sys_fdatasync_args /* { 2373 syscallarg(int) fd; 2374 } */ *uap = v; 2375 struct vnode *vp; 2376 struct file *fp; 2377 int error; 2378 2379 /* getvnode() will use the descriptor for us */ 2380 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2381 return (error); 2382 vp = (struct vnode *)fp->f_data; 2383 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2384 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, p); 2385 VOP_UNLOCK(vp, 0); 2386 FILE_UNUSE(fp, p); 2387 return (error); 2388 } 2389 2390 /* 2391 * Rename files, (standard) BSD semantics frontend. 2392 */ 2393 /* ARGSUSED */ 2394 int 2395 sys_rename(p, v, retval) 2396 struct proc *p; 2397 void *v; 2398 register_t *retval; 2399 { 2400 register struct sys_rename_args /* { 2401 syscallarg(const char *) from; 2402 syscallarg(const char *) to; 2403 } */ *uap = v; 2404 2405 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 0)); 2406 } 2407 2408 /* 2409 * Rename files, POSIX semantics frontend. 2410 */ 2411 /* ARGSUSED */ 2412 int 2413 sys___posix_rename(p, v, retval) 2414 struct proc *p; 2415 void *v; 2416 register_t *retval; 2417 { 2418 register struct sys___posix_rename_args /* { 2419 syscallarg(const char *) from; 2420 syscallarg(const char *) to; 2421 } */ *uap = v; 2422 2423 return (rename_files(SCARG(uap, from), SCARG(uap, to), p, 1)); 2424 } 2425 2426 /* 2427 * Rename files. Source and destination must either both be directories, 2428 * or both not be directories. If target is a directory, it must be empty. 2429 * If `from' and `to' refer to the same object, the value of the `retain' 2430 * argument is used to determine whether `from' will be 2431 * 2432 * (retain == 0) deleted unless `from' and `to' refer to the same 2433 * object in the file system's name space (BSD). 2434 * (retain == 1) always retained (POSIX). 2435 */ 2436 static int 2437 rename_files(from, to, p, retain) 2438 const char *from, *to; 2439 struct proc *p; 2440 int retain; 2441 { 2442 register struct vnode *tvp, *fvp, *tdvp; 2443 struct nameidata fromnd, tond; 2444 int error; 2445 2446 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 2447 from, p); 2448 if ((error = namei(&fromnd)) != 0) 2449 return (error); 2450 fvp = fromnd.ni_vp; 2451 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART, 2452 UIO_USERSPACE, to, p); 2453 if ((error = namei(&tond)) != 0) { 2454 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2455 vrele(fromnd.ni_dvp); 2456 vrele(fvp); 2457 goto out1; 2458 } 2459 tdvp = tond.ni_dvp; 2460 tvp = tond.ni_vp; 2461 2462 if (tvp != NULL) { 2463 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 2464 error = ENOTDIR; 2465 goto out; 2466 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 2467 error = EISDIR; 2468 goto out; 2469 } 2470 } 2471 2472 if (fvp == tdvp) 2473 error = EINVAL; 2474 2475 /* 2476 * Source and destination refer to the same object. 2477 */ 2478 if (fvp == tvp) { 2479 if (retain) 2480 error = -1; 2481 else if (fromnd.ni_dvp == tdvp && 2482 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 2483 !memcmp(fromnd.ni_cnd.cn_nameptr, 2484 tond.ni_cnd.cn_nameptr, 2485 fromnd.ni_cnd.cn_namelen)) 2486 error = -1; 2487 } 2488 2489 out: 2490 if (!error) { 2491 VOP_LEASE(tdvp, p, p->p_ucred, LEASE_WRITE); 2492 if (fromnd.ni_dvp != tdvp) 2493 VOP_LEASE(fromnd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 2494 if (tvp) { 2495 (void)uvm_vnp_uncache(tvp); 2496 VOP_LEASE(tvp, p, p->p_ucred, LEASE_WRITE); 2497 } 2498 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 2499 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 2500 } else { 2501 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 2502 if (tdvp == tvp) 2503 vrele(tdvp); 2504 else 2505 vput(tdvp); 2506 if (tvp) 2507 vput(tvp); 2508 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 2509 vrele(fromnd.ni_dvp); 2510 vrele(fvp); 2511 } 2512 vrele(tond.ni_startdir); 2513 FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI); 2514 out1: 2515 if (fromnd.ni_startdir) 2516 vrele(fromnd.ni_startdir); 2517 FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI); 2518 return (error == -1 ? 0 : error); 2519 } 2520 2521 /* 2522 * Make a directory file. 2523 */ 2524 /* ARGSUSED */ 2525 int 2526 sys_mkdir(p, v, retval) 2527 struct proc *p; 2528 void *v; 2529 register_t *retval; 2530 { 2531 register struct sys_mkdir_args /* { 2532 syscallarg(const char *) path; 2533 syscallarg(int) mode; 2534 } */ *uap = v; 2535 register struct vnode *vp; 2536 struct vattr vattr; 2537 int error; 2538 struct nameidata nd; 2539 2540 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), p); 2541 if ((error = namei(&nd)) != 0) 2542 return (error); 2543 vp = nd.ni_vp; 2544 if (vp != NULL) { 2545 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2546 if (nd.ni_dvp == vp) 2547 vrele(nd.ni_dvp); 2548 else 2549 vput(nd.ni_dvp); 2550 vrele(vp); 2551 return (EEXIST); 2552 } 2553 VATTR_NULL(&vattr); 2554 vattr.va_type = VDIR; 2555 vattr.va_mode = 2556 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 2557 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 2558 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2559 if (!error) 2560 vput(nd.ni_vp); 2561 return (error); 2562 } 2563 2564 /* 2565 * Remove a directory file. 2566 */ 2567 /* ARGSUSED */ 2568 int 2569 sys_rmdir(p, v, retval) 2570 struct proc *p; 2571 void *v; 2572 register_t *retval; 2573 { 2574 struct sys_rmdir_args /* { 2575 syscallarg(const char *) path; 2576 } */ *uap = v; 2577 register struct vnode *vp; 2578 int error; 2579 struct nameidata nd; 2580 2581 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2582 SCARG(uap, path), p); 2583 if ((error = namei(&nd)) != 0) 2584 return (error); 2585 vp = nd.ni_vp; 2586 if (vp->v_type != VDIR) { 2587 error = ENOTDIR; 2588 goto out; 2589 } 2590 /* 2591 * No rmdir "." please. 2592 */ 2593 if (nd.ni_dvp == vp) { 2594 error = EINVAL; 2595 goto out; 2596 } 2597 /* 2598 * The root of a mounted filesystem cannot be deleted. 2599 */ 2600 if (vp->v_flag & VROOT) 2601 error = EBUSY; 2602 out: 2603 if (!error) { 2604 VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE); 2605 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 2606 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2607 } else { 2608 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2609 if (nd.ni_dvp == vp) 2610 vrele(nd.ni_dvp); 2611 else 2612 vput(nd.ni_dvp); 2613 vput(vp); 2614 } 2615 return (error); 2616 } 2617 2618 /* 2619 * Read a block of directory entries in a file system independent format. 2620 */ 2621 int 2622 sys_getdents(p, v, retval) 2623 struct proc *p; 2624 void *v; 2625 register_t *retval; 2626 { 2627 register struct sys_getdents_args /* { 2628 syscallarg(int) fd; 2629 syscallarg(char *) buf; 2630 syscallarg(size_t) count; 2631 } */ *uap = v; 2632 struct file *fp; 2633 int error, done; 2634 2635 /* getvnode() will use the descriptor for us */ 2636 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2637 return (error); 2638 if ((fp->f_flag & FREAD) == 0) { 2639 error = EBADF; 2640 goto out; 2641 } 2642 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 2643 SCARG(uap, count), &done, p, 0, 0); 2644 *retval = done; 2645 out: 2646 FILE_UNUSE(fp, p); 2647 return (error); 2648 } 2649 2650 /* 2651 * Set the mode mask for creation of filesystem nodes. 2652 */ 2653 int 2654 sys_umask(p, v, retval) 2655 struct proc *p; 2656 void *v; 2657 register_t *retval; 2658 { 2659 struct sys_umask_args /* { 2660 syscallarg(mode_t) newmask; 2661 } */ *uap = v; 2662 struct cwdinfo *cwdi; 2663 2664 cwdi = p->p_cwdi; 2665 *retval = cwdi->cwdi_cmask; 2666 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 2667 return (0); 2668 } 2669 2670 /* 2671 * Void all references to file by ripping underlying filesystem 2672 * away from vnode. 2673 */ 2674 /* ARGSUSED */ 2675 int 2676 sys_revoke(p, v, retval) 2677 struct proc *p; 2678 void *v; 2679 register_t *retval; 2680 { 2681 register struct sys_revoke_args /* { 2682 syscallarg(const char *) path; 2683 } */ *uap = v; 2684 register struct vnode *vp; 2685 struct vattr vattr; 2686 int error; 2687 struct nameidata nd; 2688 2689 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), p); 2690 if ((error = namei(&nd)) != 0) 2691 return (error); 2692 vp = nd.ni_vp; 2693 if ((error = VOP_GETATTR(vp, &vattr, p->p_ucred, p)) != 0) 2694 goto out; 2695 if (p->p_ucred->cr_uid != vattr.va_uid && 2696 (error = suser(p->p_ucred, &p->p_acflag)) != 0) 2697 goto out; 2698 if (vp->v_usecount > 1 || (vp->v_flag & VALIASED)) 2699 VOP_REVOKE(vp, REVOKEALL); 2700 out: 2701 vrele(vp); 2702 return (error); 2703 } 2704 2705 /* 2706 * Convert a user file descriptor to a kernel file entry. 2707 */ 2708 int 2709 getvnode(fdp, fd, fpp) 2710 struct filedesc *fdp; 2711 int fd; 2712 struct file **fpp; 2713 { 2714 struct vnode *vp; 2715 struct file *fp; 2716 2717 if ((u_int)fd >= fdp->fd_nfiles || 2718 (fp = fdp->fd_ofiles[fd]) == NULL || 2719 (fp->f_iflags & FIF_WANTCLOSE) != 0) 2720 return (EBADF); 2721 2722 FILE_USE(fp); 2723 2724 if (fp->f_type != DTYPE_VNODE) { 2725 FILE_UNUSE(fp, NULL); 2726 return (EINVAL); 2727 } 2728 2729 vp = (struct vnode *)fp->f_data; 2730 if (vp->v_type == VBAD) { 2731 FILE_UNUSE(fp, NULL); 2732 return (EBADF); 2733 } 2734 2735 *fpp = fp; 2736 return (0); 2737 } 2738
Indexes created Wed Sep 24 15:09:51 GMT 2025