1 /* 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)vfs_syscalls.c 8.13 (Berkeley) 4/15/94 39 * $FreeBSD: src/sys/kern/vfs_syscalls.c,v 1.151.2.18 2003/04/04 20:35:58 tegge Exp $ 40 * $DragonFly: src/sys/kern/vfs_syscalls.c,v 1.30 2004/03/16 17:53:53 dillon Exp $ 41 */ 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/buf.h> 46 #include <sys/sysent.h> 47 #include <sys/malloc.h> 48 #include <sys/mount.h> 49 #include <sys/sysproto.h> 50 #include <sys/filedesc.h> 51 #include <sys/kernel.h> 52 #include <sys/fcntl.h> 53 #include <sys/file.h> 54 #include <sys/linker.h> 55 #include <sys/stat.h> 56 #include <sys/unistd.h> 57 #include <sys/vnode.h> 58 #include <sys/proc.h> 59 #include <sys/namei.h> 60 #include <sys/dirent.h> 61 #include <sys/extattr.h> 62 #include <sys/kern_syscall.h> 63 64 #include <machine/limits.h> 65 #include <vfs/union/union.h> 66 #include <sys/sysctl.h> 67 #include <vm/vm.h> 68 #include <vm/vm_object.h> 69 #include <vm/vm_zone.h> 70 #include <vm/vm_page.h> 71 72 #include <sys/file2.h> 73 74 static int checkvp_chdir (struct vnode *vn, struct thread *td); 75 static void checkdirs (struct vnode *olddp); 76 static int chroot_refuse_vdir_fds (struct filedesc *fdp); 77 static int getutimes (const struct timeval *, struct timespec *); 78 static int setfown (struct vnode *, uid_t, gid_t); 79 static int setfmode (struct vnode *, int); 80 static int setfflags (struct vnode *, int); 81 static int setutimes (struct vnode *, const struct timespec *, int); 82 static int usermount = 0; /* if 1, non-root can mount fs. */ 83 84 int (*union_dircheckp) (struct thread *, struct vnode **, struct file *); 85 86 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, ""); 87 88 /* 89 * Virtual File System System Calls 90 */ 91 92 /* 93 * Mount a file system. 94 */ 95 /* 96 * mount_args(char *type, char *path, int flags, caddr_t data) 97 */ 98 /* ARGSUSED */ 99 int 100 mount(struct mount_args *uap) 101 { 102 struct thread *td = curthread; 103 struct proc *p = td->td_proc; 104 struct vnode *vp; 105 struct mount *mp; 106 struct vfsconf *vfsp; 107 int error, flag = 0, flag2 = 0; 108 struct vattr va; 109 struct nameidata nd; 110 char fstypename[MFSNAMELEN]; 111 lwkt_tokref vlock; 112 lwkt_tokref ilock; 113 114 if (usermount == 0 && (error = suser(td))) 115 return (error); 116 /* 117 * Do not allow NFS export by non-root users. 118 */ 119 if (SCARG(uap, flags) & MNT_EXPORTED) { 120 error = suser(td); 121 if (error) 122 return (error); 123 } 124 /* 125 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users 126 */ 127 if (suser(td)) 128 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 129 /* 130 * Get vnode to be covered 131 */ 132 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, 133 SCARG(uap, path), td); 134 if ((error = namei(&nd)) != 0) 135 return (error); 136 NDFREE(&nd, NDF_ONLY_PNBUF); 137 vp = nd.ni_vp; 138 if (SCARG(uap, flags) & MNT_UPDATE) { 139 if ((vp->v_flag & VROOT) == 0) { 140 vput(vp); 141 return (EINVAL); 142 } 143 mp = vp->v_mount; 144 flag = mp->mnt_flag; 145 flag2 = mp->mnt_kern_flag; 146 /* 147 * We only allow the filesystem to be reloaded if it 148 * is currently mounted read-only. 149 */ 150 if ((SCARG(uap, flags) & MNT_RELOAD) && 151 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 152 vput(vp); 153 return (EOPNOTSUPP); /* Needs translation */ 154 } 155 /* 156 * Only root, or the user that did the original mount is 157 * permitted to update it. 158 */ 159 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 160 (error = suser(td))) { 161 vput(vp); 162 return (error); 163 } 164 if (vfs_busy(mp, LK_NOWAIT, NULL, td)) { 165 vput(vp); 166 return (EBUSY); 167 } 168 lwkt_gettoken(&vlock, vp->v_interlock); 169 if ((vp->v_flag & VMOUNT) != 0 || 170 vp->v_mountedhere != NULL) { 171 lwkt_reltoken(&vlock); 172 vfs_unbusy(mp, td); 173 vput(vp); 174 return (EBUSY); 175 } 176 vp->v_flag |= VMOUNT; 177 lwkt_reltoken(&vlock); 178 mp->mnt_flag |= 179 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 180 VOP_UNLOCK(vp, NULL, 0, td); 181 goto update; 182 } 183 /* 184 * If the user is not root, ensure that they own the directory 185 * onto which we are attempting to mount. 186 */ 187 if ((error = VOP_GETATTR(vp, &va, td)) || 188 (va.va_uid != p->p_ucred->cr_uid && 189 (error = suser(td)))) { 190 vput(vp); 191 return (error); 192 } 193 if ((error = vinvalbuf(vp, V_SAVE, td, 0, 0)) != 0) { 194 vput(vp); 195 return (error); 196 } 197 if (vp->v_type != VDIR) { 198 vput(vp); 199 return (ENOTDIR); 200 } 201 if ((error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL)) != 0) { 202 vput(vp); 203 return (error); 204 } 205 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 206 if (!strcmp(vfsp->vfc_name, fstypename)) 207 break; 208 if (vfsp == NULL) { 209 linker_file_t lf; 210 211 /* Only load modules for root (very important!) */ 212 if ((error = suser(td)) != 0) { 213 vput(vp); 214 return error; 215 } 216 error = linker_load_file(fstypename, &lf); 217 if (error || lf == NULL) { 218 vput(vp); 219 if (lf == NULL) 220 error = ENODEV; 221 return error; 222 } 223 lf->userrefs++; 224 /* lookup again, see if the VFS was loaded */ 225 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) 226 if (!strcmp(vfsp->vfc_name, fstypename)) 227 break; 228 if (vfsp == NULL) { 229 lf->userrefs--; 230 linker_file_unload(lf); 231 vput(vp); 232 return (ENODEV); 233 } 234 } 235 lwkt_gettoken(&vlock, vp->v_interlock); 236 if ((vp->v_flag & VMOUNT) != 0 || 237 vp->v_mountedhere != NULL) { 238 lwkt_reltoken(&vlock); 239 vput(vp); 240 return (EBUSY); 241 } 242 vp->v_flag |= VMOUNT; 243 lwkt_reltoken(&vlock); 244 245 /* 246 * Allocate and initialize the filesystem. 247 */ 248 mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK); 249 bzero((char *)mp, (u_long)sizeof(struct mount)); 250 TAILQ_INIT(&mp->mnt_nvnodelist); 251 TAILQ_INIT(&mp->mnt_reservedvnlist); 252 mp->mnt_nvnodelistsize = 0; 253 lockinit(&mp->mnt_lock, 0, "vfslock", 0, LK_NOPAUSE); 254 vfs_busy(mp, LK_NOWAIT, NULL, td); 255 mp->mnt_op = vfsp->vfc_vfsops; 256 mp->mnt_vfc = vfsp; 257 vfsp->vfc_refcount++; 258 mp->mnt_stat.f_type = vfsp->vfc_typenum; 259 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 260 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 261 mp->mnt_vnodecovered = vp; 262 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 263 mp->mnt_iosize_max = DFLTPHYS; 264 VOP_UNLOCK(vp, NULL, 0, td); 265 update: 266 /* 267 * Set the mount level flags. 268 */ 269 if (SCARG(uap, flags) & MNT_RDONLY) 270 mp->mnt_flag |= MNT_RDONLY; 271 else if (mp->mnt_flag & MNT_RDONLY) 272 mp->mnt_kern_flag |= MNTK_WANTRDWR; 273 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 274 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME | 275 MNT_NOSYMFOLLOW | MNT_IGNORE | 276 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR); 277 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC | 278 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE | 279 MNT_NOSYMFOLLOW | MNT_IGNORE | 280 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR); 281 /* 282 * Mount the filesystem. 283 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 284 * get. No freeing of cn_pnbuf. 285 */ 286 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, td); 287 if (mp->mnt_flag & MNT_UPDATE) { 288 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 289 mp->mnt_flag &= ~MNT_RDONLY; 290 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE); 291 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 292 if (error) { 293 mp->mnt_flag = flag; 294 mp->mnt_kern_flag = flag2; 295 } 296 vfs_unbusy(mp, td); 297 lwkt_gettoken(&vlock, vp->v_interlock); 298 vp->v_flag &= ~VMOUNT; 299 lwkt_reltoken(&vlock); 300 vrele(vp); 301 return (error); 302 } 303 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 304 /* 305 * Put the new filesystem on the mount list after root. 306 */ 307 cache_purge(vp); 308 if (!error) { 309 lwkt_gettoken(&vlock, vp->v_interlock); 310 vp->v_flag &= ~VMOUNT; 311 vp->v_mountedhere = mp; 312 lwkt_reltoken(&vlock); 313 lwkt_gettoken(&ilock, &mountlist_token); 314 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 315 lwkt_reltoken(&ilock); 316 checkdirs(vp); 317 VOP_UNLOCK(vp, NULL, 0, td); 318 error = vfs_allocate_syncvnode(mp); 319 vfs_unbusy(mp, td); 320 if ((error = VFS_START(mp, 0, td)) != 0) 321 vrele(vp); 322 } else { 323 lwkt_gettoken(&vlock, vp->v_interlock); 324 vp->v_flag &= ~VMOUNT; 325 lwkt_reltoken(&vlock); 326 mp->mnt_vfc->vfc_refcount--; 327 vfs_unbusy(mp, td); 328 free((caddr_t)mp, M_MOUNT); 329 vput(vp); 330 } 331 return (error); 332 } 333 334 /* 335 * Scan all active processes to see if any of them have a current 336 * or root directory onto which the new filesystem has just been 337 * mounted. If so, replace them with the new mount point. 338 */ 339 static void 340 checkdirs(struct vnode *olddp) 341 { 342 struct filedesc *fdp; 343 struct vnode *newdp; 344 struct proc *p; 345 346 if (olddp->v_usecount == 1) 347 return; 348 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 349 panic("mount: lost mount"); 350 FOREACH_PROC_IN_SYSTEM(p) { 351 fdp = p->p_fd; 352 if (fdp->fd_cdir == olddp) { 353 vrele(fdp->fd_cdir); 354 VREF(newdp); 355 fdp->fd_cdir = newdp; 356 } 357 if (fdp->fd_rdir == olddp) { 358 vrele(fdp->fd_rdir); 359 VREF(newdp); 360 fdp->fd_rdir = newdp; 361 } 362 } 363 if (rootvnode == olddp) { 364 vrele(rootvnode); 365 VREF(newdp); 366 rootvnode = newdp; 367 vfs_cache_setroot(rootvnode); 368 } 369 vput(newdp); 370 } 371 372 /* 373 * Unmount a file system. 374 * 375 * Note: unmount takes a path to the vnode mounted on as argument, 376 * not special file (as before). 377 */ 378 /* 379 * umount_args(char *path, int flags) 380 */ 381 /* ARGSUSED */ 382 int 383 unmount(struct unmount_args *uap) 384 { 385 struct thread *td = curthread; 386 struct proc *p = td->td_proc; 387 struct vnode *vp; 388 struct mount *mp; 389 int error; 390 struct nameidata nd; 391 392 KKASSERT(p); 393 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, 394 SCARG(uap, path), td); 395 if ((error = namei(&nd)) != 0) 396 return (error); 397 vp = nd.ni_vp; 398 NDFREE(&nd, NDF_ONLY_PNBUF); 399 mp = vp->v_mount; 400 401 /* 402 * Only root, or the user that did the original mount is 403 * permitted to unmount this filesystem. 404 */ 405 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 406 (error = suser(td))) { 407 vput(vp); 408 return (error); 409 } 410 411 /* 412 * Don't allow unmounting the root file system. 413 */ 414 if (mp->mnt_flag & MNT_ROOTFS) { 415 vput(vp); 416 return (EINVAL); 417 } 418 419 /* 420 * Must be the root of the filesystem 421 */ 422 if ((vp->v_flag & VROOT) == 0) { 423 vput(vp); 424 return (EINVAL); 425 } 426 vput(vp); 427 return (dounmount(mp, SCARG(uap, flags), td)); 428 } 429 430 /* 431 * Do the actual file system unmount. 432 */ 433 int 434 dounmount(struct mount *mp, int flags, struct thread *td) 435 { 436 struct vnode *coveredvp; 437 int error; 438 int async_flag; 439 lwkt_tokref ilock; 440 441 lwkt_gettoken(&ilock, &mountlist_token); 442 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 443 lwkt_reltoken(&ilock); 444 return (EBUSY); 445 } 446 mp->mnt_kern_flag |= MNTK_UNMOUNT; 447 /* Allow filesystems to detect that a forced unmount is in progress. */ 448 if (flags & MNT_FORCE) 449 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 450 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 451 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &ilock, td); 452 if (error) { 453 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 454 if (mp->mnt_kern_flag & MNTK_MWAIT) 455 wakeup((caddr_t)mp); 456 return (error); 457 } 458 459 if (mp->mnt_flag & MNT_EXPUBLIC) 460 vfs_setpublicfs(NULL, NULL, NULL); 461 462 vfs_msync(mp, MNT_WAIT); 463 async_flag = mp->mnt_flag & MNT_ASYNC; 464 mp->mnt_flag &=~ MNT_ASYNC; 465 cache_purgevfs(mp); /* remove cache entries for this file sys */ 466 if (mp->mnt_syncer != NULL) 467 vrele(mp->mnt_syncer); 468 if (((mp->mnt_flag & MNT_RDONLY) || 469 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) || 470 (flags & MNT_FORCE)) 471 error = VFS_UNMOUNT(mp, flags, td); 472 lwkt_gettokref(&ilock); 473 if (error) { 474 if (mp->mnt_syncer == NULL) 475 vfs_allocate_syncvnode(mp); 476 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 477 mp->mnt_flag |= async_flag; 478 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 479 &ilock, td); 480 if (mp->mnt_kern_flag & MNTK_MWAIT) 481 wakeup((caddr_t)mp); 482 return (error); 483 } 484 TAILQ_REMOVE(&mountlist, mp, mnt_list); 485 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 486 coveredvp->v_mountedhere = NULL; 487 vrele(coveredvp); 488 } 489 mp->mnt_vfc->vfc_refcount--; 490 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 491 panic("unmount: dangling vnode"); 492 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &ilock, td); 493 if (mp->mnt_kern_flag & MNTK_MWAIT) 494 wakeup((caddr_t)mp); 495 free((caddr_t)mp, M_MOUNT); 496 return (0); 497 } 498 499 /* 500 * Sync each mounted filesystem. 501 */ 502 503 #ifdef DEBUG 504 static int syncprt = 0; 505 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, ""); 506 #endif 507 508 /* ARGSUSED */ 509 int 510 sync(struct sync_args *uap) 511 { 512 struct thread *td = curthread; 513 struct mount *mp, *nmp; 514 lwkt_tokref ilock; 515 int asyncflag; 516 517 lwkt_gettoken(&ilock, &mountlist_token); 518 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 519 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) { 520 nmp = TAILQ_NEXT(mp, mnt_list); 521 continue; 522 } 523 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 524 asyncflag = mp->mnt_flag & MNT_ASYNC; 525 mp->mnt_flag &= ~MNT_ASYNC; 526 vfs_msync(mp, MNT_NOWAIT); 527 VFS_SYNC(mp, MNT_NOWAIT, td); 528 mp->mnt_flag |= asyncflag; 529 } 530 lwkt_gettokref(&ilock); 531 nmp = TAILQ_NEXT(mp, mnt_list); 532 vfs_unbusy(mp, td); 533 } 534 lwkt_reltoken(&ilock); 535 #if 0 536 /* 537 * XXX don't call vfs_bufstats() yet because that routine 538 * was not imported in the Lite2 merge. 539 */ 540 #ifdef DIAGNOSTIC 541 if (syncprt) 542 vfs_bufstats(); 543 #endif /* DIAGNOSTIC */ 544 #endif 545 return (0); 546 } 547 548 /* XXX PRISON: could be per prison flag */ 549 static int prison_quotas; 550 #if 0 551 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, ""); 552 #endif 553 554 /* 555 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg) 556 * 557 * Change filesystem quotas. 558 */ 559 /* ARGSUSED */ 560 int 561 quotactl(struct quotactl_args *uap) 562 { 563 struct thread *td = curthread; 564 struct proc *p = td->td_proc; 565 struct mount *mp; 566 int error; 567 struct nameidata nd; 568 569 KKASSERT(p); 570 if (p->p_ucred->cr_prison && !prison_quotas) 571 return (EPERM); 572 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, 573 SCARG(uap, path), td); 574 if ((error = namei(&nd)) != 0) 575 return (error); 576 mp = nd.ni_vp->v_mount; 577 NDFREE(&nd, NDF_ONLY_PNBUF); 578 vrele(nd.ni_vp); 579 return (VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 580 SCARG(uap, arg), td)); 581 } 582 583 int 584 kern_statfs(struct nameidata *nd, struct statfs *buf) 585 { 586 struct thread *td = curthread; 587 struct mount *mp; 588 struct statfs *sp; 589 int error; 590 591 error = namei(nd); 592 if (error) 593 return (error); 594 mp = nd->ni_vp->v_mount; 595 sp = &mp->mnt_stat; 596 NDFREE(nd, NDF_ONLY_PNBUF); 597 vrele(nd->ni_vp); 598 error = VFS_STATFS(mp, sp, td); 599 if (error) 600 return (error); 601 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 602 bcopy(sp, buf, sizeof(*buf)); 603 /* Only root should have access to the fsid's. */ 604 if (suser(td)) 605 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 606 return (0); 607 } 608 609 /* 610 * statfs_args(char *path, struct statfs *buf) 611 * 612 * Get filesystem statistics. 613 */ 614 int 615 statfs(struct statfs_args *uap) 616 { 617 struct thread *td = curthread; 618 struct nameidata nd; 619 struct statfs buf; 620 int error; 621 622 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, uap->path, td); 623 624 error = kern_statfs(&nd, &buf); 625 626 if (error == 0) 627 error = copyout(&buf, uap->buf, sizeof(*uap->buf)); 628 return (error); 629 } 630 631 int 632 kern_fstatfs(int fd, struct statfs *buf) 633 { 634 struct thread *td = curthread; 635 struct proc *p = td->td_proc; 636 struct file *fp; 637 struct mount *mp; 638 struct statfs *sp; 639 int error; 640 641 KKASSERT(p); 642 error = getvnode(p->p_fd, fd, &fp); 643 if (error) 644 return (error); 645 mp = ((struct vnode *)fp->f_data)->v_mount; 646 if (mp == NULL) 647 return (EBADF); 648 sp = &mp->mnt_stat; 649 error = VFS_STATFS(mp, sp, td); 650 if (error) 651 return (error); 652 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 653 bcopy(sp, buf, sizeof(*buf)); 654 /* Only root should have access to the fsid's. */ 655 if (suser(td)) 656 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 657 return (0); 658 } 659 660 /* 661 * fstatfs_args(int fd, struct statfs *buf) 662 * 663 * Get filesystem statistics. 664 */ 665 int 666 fstatfs(struct fstatfs_args *uap) 667 { 668 struct statfs buf; 669 int error; 670 671 error = kern_fstatfs(uap->fd, &buf); 672 673 if (error == 0) 674 error = copyout(&buf, uap->buf, sizeof(*uap->buf)); 675 return (error); 676 } 677 678 /* 679 * getfsstat_args(struct statfs *buf, long bufsize, int flags) 680 * 681 * Get statistics on all filesystems. 682 */ 683 /* ARGSUSED */ 684 int 685 getfsstat(struct getfsstat_args *uap) 686 { 687 struct thread *td = curthread; 688 struct mount *mp, *nmp; 689 struct statfs *sp; 690 caddr_t sfsp; 691 lwkt_tokref ilock; 692 long count, maxcount, error; 693 694 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs); 695 sfsp = (caddr_t)SCARG(uap, buf); 696 count = 0; 697 lwkt_gettoken(&ilock, &mountlist_token); 698 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 699 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) { 700 nmp = TAILQ_NEXT(mp, mnt_list); 701 continue; 702 } 703 if (sfsp && count < maxcount) { 704 sp = &mp->mnt_stat; 705 /* 706 * If MNT_NOWAIT or MNT_LAZY is specified, do not 707 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY 708 * overrides MNT_WAIT. 709 */ 710 if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 || 711 (SCARG(uap, flags) & MNT_WAIT)) && 712 (error = VFS_STATFS(mp, sp, td))) { 713 lwkt_gettokref(&ilock); 714 nmp = TAILQ_NEXT(mp, mnt_list); 715 vfs_unbusy(mp, td); 716 continue; 717 } 718 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 719 error = copyout((caddr_t)sp, sfsp, sizeof(*sp)); 720 if (error) { 721 vfs_unbusy(mp, td); 722 return (error); 723 } 724 sfsp += sizeof(*sp); 725 } 726 count++; 727 lwkt_gettokref(&ilock); 728 nmp = TAILQ_NEXT(mp, mnt_list); 729 vfs_unbusy(mp, td); 730 } 731 lwkt_reltoken(&ilock); 732 if (sfsp && count > maxcount) 733 uap->sysmsg_result = maxcount; 734 else 735 uap->sysmsg_result = count; 736 return (0); 737 } 738 739 /* 740 * fchdir_args(int fd) 741 * 742 * Change current working directory to a given file descriptor. 743 */ 744 /* ARGSUSED */ 745 int 746 fchdir(struct fchdir_args *uap) 747 { 748 struct thread *td = curthread; 749 struct proc *p = td->td_proc; 750 struct filedesc *fdp = p->p_fd; 751 struct vnode *vp, *tdp; 752 struct mount *mp; 753 struct file *fp; 754 int error; 755 756 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 757 return (error); 758 vp = (struct vnode *)fp->f_data; 759 VREF(vp); 760 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 761 if (vp->v_type != VDIR) 762 error = ENOTDIR; 763 else 764 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td); 765 while (!error && (mp = vp->v_mountedhere) != NULL) { 766 if (vfs_busy(mp, 0, NULL, td)) 767 continue; 768 error = VFS_ROOT(mp, &tdp); 769 vfs_unbusy(mp, td); 770 if (error) 771 break; 772 vput(vp); 773 vp = tdp; 774 } 775 if (error) { 776 vput(vp); 777 return (error); 778 } 779 VOP_UNLOCK(vp, NULL, 0, td); 780 vrele(fdp->fd_cdir); 781 fdp->fd_cdir = vp; 782 return (0); 783 } 784 785 int 786 kern_chdir(struct nameidata *nd) 787 { 788 struct thread *td = curthread; 789 struct proc *p = td->td_proc; 790 struct filedesc *fdp = p->p_fd; 791 int error; 792 793 if ((error = namei(nd)) != 0) 794 return (error); 795 if ((error = checkvp_chdir(nd->ni_vp, td)) == 0) { 796 vrele(fdp->fd_cdir); 797 fdp->fd_cdir = nd->ni_vp; 798 VREF(fdp->fd_cdir); 799 } 800 NDFREE(nd, ~(NDF_NO_FREE_PNBUF | NDF_NO_VP_PUT)); 801 return (error); 802 } 803 804 /* 805 * chdir_args(char *path) 806 * 807 * Change current working directory (``.''). 808 */ 809 int 810 chdir(struct chdir_args *uap) 811 { 812 struct thread *td = curthread; 813 struct nameidata nd; 814 int error; 815 816 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, 817 uap->path, td); 818 819 error = kern_chdir(&nd); 820 821 return (error); 822 } 823 824 /* 825 * Helper function for raised chroot(2) security function: Refuse if 826 * any filedescriptors are open directories. 827 */ 828 static int 829 chroot_refuse_vdir_fds(fdp) 830 struct filedesc *fdp; 831 { 832 struct vnode *vp; 833 struct file *fp; 834 int error; 835 int fd; 836 837 for (fd = 0; fd < fdp->fd_nfiles ; fd++) { 838 error = getvnode(fdp, fd, &fp); 839 if (error) 840 continue; 841 vp = (struct vnode *)fp->f_data; 842 if (vp->v_type != VDIR) 843 continue; 844 return(EPERM); 845 } 846 return (0); 847 } 848 849 /* 850 * This sysctl determines if we will allow a process to chroot(2) if it 851 * has a directory open: 852 * 0: disallowed for all processes. 853 * 1: allowed for processes that were not already chroot(2)'ed. 854 * 2: allowed for all processes. 855 */ 856 857 static int chroot_allow_open_directories = 1; 858 859 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW, 860 &chroot_allow_open_directories, 0, ""); 861 862 /* 863 * Chroot to the specified vnode. vp must be locked and referenced on 864 * call, and will be left locked and referenced on return. This routine 865 * may acquire additional refs on the vnode when associating it with 866 * the process's root and/or jail dirs. 867 */ 868 int 869 kern_chroot(struct vnode *vp) 870 { 871 struct thread *td = curthread; 872 struct proc *p = td->td_proc; 873 struct filedesc *fdp = p->p_fd; 874 int error; 875 876 /* 877 * Only root can chroot 878 */ 879 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0) 880 return (error); 881 882 /* 883 * Disallow open directory descriptors (fchdir() breakouts). 884 */ 885 if (chroot_allow_open_directories == 0 || 886 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) { 887 if ((error = chroot_refuse_vdir_fds(fdp)) != 0) 888 return (error); 889 } 890 891 /* 892 * Check the validity of vp as a directory to change to and 893 * associate it with rdir/jdir. 894 */ 895 if ((error = checkvp_chdir(vp, td)) == 0) { 896 vrele(fdp->fd_rdir); 897 fdp->fd_rdir = vp; 898 VREF(fdp->fd_rdir); 899 if (fdp->fd_jdir == NULL) { 900 fdp->fd_jdir = vp; 901 VREF(fdp->fd_jdir); 902 } 903 } 904 return (error); 905 } 906 907 /* 908 * chroot_args(char *path) 909 * 910 * Change notion of root (``/'') directory. 911 */ 912 /* ARGSUSED */ 913 int 914 chroot(struct chroot_args *uap) 915 { 916 struct thread *td = curthread; 917 struct nameidata nd; 918 int error; 919 920 KKASSERT(td->td_proc); 921 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, 922 SCARG(uap, path), td); 923 if ((error = namei(&nd)) == 0) { 924 error = kern_chroot(nd.ni_vp); 925 NDFREE(&nd, ~(NDF_NO_FREE_PNBUF | NDF_NO_VP_PUT)); 926 } 927 return (error); 928 } 929 930 /* 931 * Common routine for chroot and chdir. Given a locked, referenced vnode, 932 * determine whether it is legal to chdir to the vnode. The vnode's state 933 * is not changed by this call. 934 */ 935 int 936 checkvp_chdir(struct vnode *vp, struct thread *td) 937 { 938 int error; 939 940 if (vp->v_type != VDIR) 941 error = ENOTDIR; 942 else 943 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td); 944 return (error); 945 } 946 947 int 948 kern_open(struct nameidata *nd, int oflags, int mode, int *res) 949 { 950 struct thread *td = curthread; 951 struct proc *p = td->td_proc; 952 struct filedesc *fdp = p->p_fd; 953 struct file *fp; 954 struct vnode *vp; 955 int cmode, flags; 956 struct file *nfp; 957 int type, indx, error; 958 struct flock lf; 959 960 if ((oflags & O_ACCMODE) == O_ACCMODE) 961 return (EINVAL); 962 flags = FFLAGS(oflags); 963 error = falloc(p, &nfp, &indx); 964 if (error) 965 return (error); 966 fp = nfp; 967 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT; 968 p->p_dupfd = -indx - 1; /* XXX check for fdopen */ 969 /* 970 * Bump the ref count to prevent another process from closing 971 * the descriptor while we are blocked in vn_open() 972 */ 973 fhold(fp); 974 error = vn_open(nd, flags, cmode); 975 if (error) { 976 /* 977 * release our own reference 978 */ 979 fdrop(fp, td); 980 981 /* 982 * handle special fdopen() case. bleh. dupfdopen() is 983 * responsible for dropping the old contents of ofiles[indx] 984 * if it succeeds. 985 */ 986 if ((error == ENODEV || error == ENXIO) && 987 p->p_dupfd >= 0 && /* XXX from fdopen */ 988 (error = 989 dupfdopen(fdp, indx, p->p_dupfd, flags, error)) == 0) { 990 *res = indx; 991 return (0); 992 } 993 /* 994 * Clean up the descriptor, but only if another thread hadn't 995 * replaced or closed it. 996 */ 997 if (fdp->fd_ofiles[indx] == fp) { 998 fdp->fd_ofiles[indx] = NULL; 999 fdrop(fp, td); 1000 } 1001 1002 if (error == ERESTART) 1003 error = EINTR; 1004 return (error); 1005 } 1006 p->p_dupfd = 0; 1007 NDFREE(nd, NDF_ONLY_PNBUF); 1008 vp = nd->ni_vp; 1009 1010 /* 1011 * There should be 2 references on the file, one from the descriptor 1012 * table, and one for us. 1013 * 1014 * Handle the case where someone closed the file (via its file 1015 * descriptor) while we were blocked. The end result should look 1016 * like opening the file succeeded but it was immediately closed. 1017 */ 1018 if (fp->f_count == 1) { 1019 KASSERT(fdp->fd_ofiles[indx] != fp, 1020 ("Open file descriptor lost all refs")); 1021 VOP_UNLOCK(vp, NULL, 0, td); 1022 vn_close(vp, flags & FMASK, td); 1023 fdrop(fp, td); 1024 *res = indx; 1025 return 0; 1026 } 1027 1028 fp->f_data = (caddr_t)vp; 1029 fp->f_flag = flags & FMASK; 1030 fp->f_ops = &vnops; 1031 fp->f_type = (vp->v_type == VFIFO ? DTYPE_FIFO : DTYPE_VNODE); 1032 if (flags & (O_EXLOCK | O_SHLOCK)) { 1033 lf.l_whence = SEEK_SET; 1034 lf.l_start = 0; 1035 lf.l_len = 0; 1036 if (flags & O_EXLOCK) 1037 lf.l_type = F_WRLCK; 1038 else 1039 lf.l_type = F_RDLCK; 1040 type = F_FLOCK; 1041 if ((flags & FNONBLOCK) == 0) 1042 type |= F_WAIT; 1043 VOP_UNLOCK(vp, NULL, 0, td); 1044 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) { 1045 /* 1046 * lock request failed. Normally close the descriptor 1047 * but handle the case where someone might have dup()d 1048 * it when we weren't looking. One reference is 1049 * owned by the descriptor array, the other by us. 1050 */ 1051 if (fdp->fd_ofiles[indx] == fp) { 1052 fdp->fd_ofiles[indx] = NULL; 1053 fdrop(fp, td); 1054 } 1055 fdrop(fp, td); 1056 return (error); 1057 } 1058 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 1059 fp->f_flag |= FHASLOCK; 1060 } 1061 /* assert that vn_open created a backing object if one is needed */ 1062 KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0, 1063 ("open: vmio vnode has no backing object after vn_open")); 1064 VOP_UNLOCK(vp, NULL, 0, td); 1065 1066 /* 1067 * release our private reference, leaving the one associated with the 1068 * descriptor table intact. 1069 */ 1070 fdrop(fp, td); 1071 *res = indx; 1072 return (0); 1073 } 1074 1075 /* 1076 * open_args(char *path, int flags, int mode) 1077 * 1078 * Check permissions, allocate an open file structure, 1079 * and call the device open routine if any. 1080 */ 1081 int 1082 open(struct open_args *uap) 1083 { 1084 struct thread *td = curthread; 1085 struct nameidata nd; 1086 int error; 1087 1088 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, uap->path, td); 1089 1090 error = kern_open(&nd, uap->flags, uap->mode, &uap->sysmsg_result); 1091 1092 return (error); 1093 } 1094 1095 int 1096 kern_mknod(struct nameidata *nd, int mode, int dev) 1097 { 1098 struct thread *td = curthread; 1099 struct proc *p = td->td_proc; 1100 struct vnode *vp; 1101 struct vattr vattr; 1102 int error; 1103 int whiteout = 0; 1104 1105 KKASSERT(p); 1106 1107 switch (mode & S_IFMT) { 1108 case S_IFCHR: 1109 case S_IFBLK: 1110 error = suser(td); 1111 break; 1112 default: 1113 error = suser_cred(p->p_ucred, PRISON_ROOT); 1114 break; 1115 } 1116 if (error) 1117 return (error); 1118 bwillwrite(); 1119 error = namei(nd); 1120 if (error) 1121 return (error); 1122 vp = nd->ni_vp; 1123 if (vp != NULL) 1124 error = EEXIST; 1125 else { 1126 VATTR_NULL(&vattr); 1127 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask; 1128 vattr.va_rdev = dev; 1129 whiteout = 0; 1130 1131 switch (mode & S_IFMT) { 1132 case S_IFMT: /* used by badsect to flag bad sectors */ 1133 vattr.va_type = VBAD; 1134 break; 1135 case S_IFCHR: 1136 vattr.va_type = VCHR; 1137 break; 1138 case S_IFBLK: 1139 vattr.va_type = VBLK; 1140 break; 1141 case S_IFWHT: 1142 whiteout = 1; 1143 break; 1144 default: 1145 error = EINVAL; 1146 break; 1147 } 1148 } 1149 if (error == 0) { 1150 VOP_LEASE(nd->ni_dvp, td, p->p_ucred, LEASE_WRITE); 1151 if (whiteout) 1152 error = VOP_WHITEOUT(nd->ni_dvp, NCPNULL, 1153 &nd->ni_cnd, NAMEI_CREATE); 1154 else { 1155 error = VOP_MKNOD(nd->ni_dvp, NCPNULL, &nd->ni_vp, 1156 &nd->ni_cnd, &vattr); 1157 if (error == 0) 1158 vput(nd->ni_vp); 1159 } 1160 NDFREE(nd, NDF_ONLY_PNBUF); 1161 vput(nd->ni_dvp); 1162 } else { 1163 NDFREE(nd, NDF_ONLY_PNBUF); 1164 if (nd->ni_dvp == vp) 1165 vrele(nd->ni_dvp); 1166 else 1167 vput(nd->ni_dvp); 1168 if (vp) 1169 vrele(vp); 1170 } 1171 ASSERT_VOP_UNLOCKED(nd->ni_dvp, "mknod"); 1172 ASSERT_VOP_UNLOCKED(nd->ni_vp, "mknod"); 1173 return (error); 1174 } 1175 1176 /* 1177 * mknod_args(char *path, int mode, int dev) 1178 * 1179 * Create a special file. 1180 */ 1181 int 1182 mknod(struct mknod_args *uap) 1183 { 1184 struct thread *td = curthread; 1185 struct nameidata nd; 1186 int error; 1187 1188 NDINIT(&nd, NAMEI_CREATE, CNP_LOCKPARENT, UIO_USERSPACE, uap->path, 1189 td); 1190 1191 error = kern_mknod(&nd, uap->mode, uap->dev); 1192 1193 return (error); 1194 } 1195 1196 int 1197 kern_mkfifo(struct nameidata *nd, int mode) 1198 { 1199 struct thread *td = curthread; 1200 struct proc *p = td->td_proc; 1201 struct vattr vattr; 1202 int error; 1203 1204 bwillwrite(); 1205 error = namei(nd); 1206 if (error) 1207 return (error); 1208 if (nd->ni_vp != NULL) { 1209 NDFREE(nd, NDF_ONLY_PNBUF); 1210 if (nd->ni_dvp == nd->ni_vp) 1211 vrele(nd->ni_dvp); 1212 else 1213 vput(nd->ni_dvp); 1214 vrele(nd->ni_vp); 1215 return (EEXIST); 1216 } 1217 VATTR_NULL(&vattr); 1218 vattr.va_type = VFIFO; 1219 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask; 1220 VOP_LEASE(nd->ni_dvp, td, p->p_ucred, LEASE_WRITE); 1221 error = VOP_MKNOD(nd->ni_dvp, NCPNULL, &nd->ni_vp, &nd->ni_cnd, &vattr); 1222 if (error == 0) 1223 vput(nd->ni_vp); 1224 NDFREE(nd, NDF_ONLY_PNBUF); 1225 vput(nd->ni_dvp); 1226 return (error); 1227 } 1228 1229 /* 1230 * mkfifo_args(char *path, int mode) 1231 * 1232 * Create a named pipe. 1233 */ 1234 int 1235 mkfifo(struct mkfifo_args *uap) 1236 { 1237 struct thread *td = curthread; 1238 struct nameidata nd; 1239 int error; 1240 1241 NDINIT(&nd, NAMEI_CREATE, CNP_LOCKPARENT, UIO_USERSPACE, uap->path, 1242 td); 1243 1244 error = kern_mkfifo(&nd, uap->mode); 1245 1246 return (error); 1247 } 1248 1249 int 1250 kern_link(struct nameidata *nd, struct nameidata *linknd) 1251 { 1252 struct thread *td = curthread; 1253 struct proc *p = td->td_proc; 1254 struct vnode *vp; 1255 int error; 1256 1257 bwillwrite(); 1258 error = namei(nd); 1259 if (error) 1260 return (error); 1261 NDFREE(nd, NDF_ONLY_PNBUF); 1262 vp = nd->ni_vp; 1263 if (vp->v_type == VDIR) 1264 error = EPERM; /* POSIX */ 1265 else { 1266 error = namei(linknd); 1267 if (error == 0) { 1268 if (linknd->ni_vp != NULL) { 1269 if (linknd->ni_vp) 1270 vrele(linknd->ni_vp); 1271 error = EEXIST; 1272 } else { 1273 VOP_LEASE(linknd->ni_dvp, td, p->p_ucred, 1274 LEASE_WRITE); 1275 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 1276 error = VOP_LINK(linknd->ni_dvp, NCPNULL, vp, 1277 &linknd->ni_cnd); 1278 } 1279 NDFREE(linknd, NDF_ONLY_PNBUF); 1280 if (linknd->ni_dvp == linknd->ni_vp) 1281 vrele(linknd->ni_dvp); 1282 else 1283 vput(linknd->ni_dvp); 1284 ASSERT_VOP_UNLOCKED(linknd->ni_dvp, "link"); 1285 ASSERT_VOP_UNLOCKED(linknd->ni_vp, "link"); 1286 } 1287 } 1288 vrele(vp); 1289 return (error); 1290 } 1291 1292 /* 1293 * link_args(char *path, char *link) 1294 * 1295 * Make a hard file link. 1296 */ 1297 int 1298 link(struct link_args *uap) 1299 { 1300 struct thread *td = curthread; 1301 struct nameidata nd, linknd; 1302 int error; 1303 1304 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_NOOBJ, UIO_USERSPACE, 1305 uap->path, td); 1306 NDINIT(&linknd, NAMEI_CREATE, CNP_LOCKPARENT | CNP_NOOBJ, 1307 UIO_USERSPACE, uap->link, td); 1308 1309 error = kern_link(&nd, &linknd); 1310 1311 return (error); 1312 } 1313 1314 int 1315 kern_symlink(char *path, struct nameidata *nd) 1316 { 1317 struct thread *td = curthread; 1318 struct proc *p = td->td_proc; 1319 struct vattr vattr; 1320 int error; 1321 1322 bwillwrite(); 1323 error = namei(nd); 1324 if (error) 1325 return (error); 1326 if (nd->ni_vp) { 1327 NDFREE(nd, NDF_ONLY_PNBUF); 1328 if (nd->ni_dvp == nd->ni_vp) 1329 vrele(nd->ni_dvp); 1330 else 1331 vput(nd->ni_dvp); 1332 vrele(nd->ni_vp); 1333 return (EEXIST); 1334 } 1335 VATTR_NULL(&vattr); 1336 vattr.va_mode = ACCESSPERMS &~ p->p_fd->fd_cmask; 1337 VOP_LEASE(nd->ni_dvp, td, p->p_ucred, LEASE_WRITE); 1338 error = VOP_SYMLINK(nd->ni_dvp, NCPNULL, &nd->ni_vp, &nd->ni_cnd, 1339 &vattr, path); 1340 NDFREE(nd, NDF_ONLY_PNBUF); 1341 if (error == 0) 1342 vput(nd->ni_vp); 1343 vput(nd->ni_dvp); 1344 ASSERT_VOP_UNLOCKED(nd->ni_dvp, "symlink"); 1345 ASSERT_VOP_UNLOCKED(nd->ni_vp, "symlink"); 1346 1347 return (error); 1348 } 1349 1350 /* 1351 * symlink(char *path, char *link) 1352 * 1353 * Make a symbolic link. 1354 */ 1355 int 1356 symlink(struct symlink_args *uap) 1357 { 1358 struct thread *td = curthread; 1359 struct nameidata nd; 1360 char *path; 1361 int error; 1362 1363 path = zalloc(namei_zone); 1364 error = copyinstr(uap->path, path, MAXPATHLEN, NULL); 1365 if (error == 0) { 1366 NDINIT(&nd, NAMEI_CREATE, CNP_LOCKPARENT | CNP_NOOBJ, 1367 UIO_USERSPACE, uap->link, td); 1368 error = kern_symlink(path, &nd); 1369 } 1370 zfree(namei_zone, path); 1371 return (error); 1372 } 1373 1374 /* 1375 * undelete_args(char *path) 1376 * 1377 * Delete a whiteout from the filesystem. 1378 */ 1379 /* ARGSUSED */ 1380 int 1381 undelete(struct undelete_args *uap) 1382 { 1383 struct thread *td = curthread; 1384 struct proc *p = td->td_proc; 1385 int error; 1386 struct nameidata nd; 1387 1388 bwillwrite(); 1389 NDINIT(&nd, NAMEI_DELETE, CNP_LOCKPARENT | CNP_DOWHITEOUT, UIO_USERSPACE, 1390 SCARG(uap, path), td); 1391 error = namei(&nd); 1392 if (error) 1393 return (error); 1394 1395 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & CNP_ISWHITEOUT)) { 1396 NDFREE(&nd, NDF_ONLY_PNBUF); 1397 if (nd.ni_dvp == nd.ni_vp) 1398 vrele(nd.ni_dvp); 1399 else 1400 vput(nd.ni_dvp); 1401 if (nd.ni_vp) 1402 vrele(nd.ni_vp); 1403 return (EEXIST); 1404 } 1405 1406 VOP_LEASE(nd.ni_dvp, td, p->p_ucred, LEASE_WRITE); 1407 error = VOP_WHITEOUT(nd.ni_dvp, NCPNULL, &nd.ni_cnd, NAMEI_DELETE); 1408 NDFREE(&nd, NDF_ONLY_PNBUF); 1409 vput(nd.ni_dvp); 1410 ASSERT_VOP_UNLOCKED(nd.ni_dvp, "undelete"); 1411 ASSERT_VOP_UNLOCKED(nd.ni_vp, "undelete"); 1412 return (error); 1413 } 1414 1415 int 1416 kern_unlink(struct nameidata *nd) 1417 { 1418 struct thread *td = curthread; 1419 struct proc *p = td->td_proc; 1420 struct vnode *vp; 1421 int error; 1422 1423 bwillwrite(); 1424 error = namei(nd); 1425 if (error) 1426 return (error); 1427 vp = nd->ni_vp; 1428 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 1429 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 1430 1431 if (vp->v_type == VDIR) 1432 error = EPERM; /* POSIX */ 1433 else { 1434 /* 1435 * The root of a mounted filesystem cannot be deleted. 1436 * 1437 * XXX: can this only be a VDIR case? 1438 */ 1439 if (vp->v_flag & VROOT) 1440 error = EBUSY; 1441 } 1442 1443 if (error == 0) { 1444 VOP_LEASE(nd->ni_dvp, td, p->p_ucred, LEASE_WRITE); 1445 error = VOP_REMOVE(nd->ni_dvp, NCPNULL, vp, &nd->ni_cnd); 1446 } 1447 NDFREE(nd, NDF_ONLY_PNBUF); 1448 if (nd->ni_dvp == vp) 1449 vrele(nd->ni_dvp); 1450 else 1451 vput(nd->ni_dvp); 1452 if (vp != NULLVP) 1453 vput(vp); 1454 ASSERT_VOP_UNLOCKED(nd->ni_dvp, "unlink"); 1455 ASSERT_VOP_UNLOCKED(nd->ni_vp, "unlink"); 1456 return (error); 1457 } 1458 1459 /* 1460 * unlink_args(char *path) 1461 * 1462 * Delete a name from the filesystem. 1463 */ 1464 int 1465 unlink(struct unlink_args *uap) 1466 { 1467 struct thread *td = curthread; 1468 struct nameidata nd; 1469 int error; 1470 1471 NDINIT(&nd, NAMEI_DELETE, CNP_LOCKPARENT, UIO_USERSPACE, uap->path, 1472 td); 1473 1474 error = kern_unlink(&nd); 1475 1476 return (error); 1477 } 1478 1479 int 1480 kern_lseek(int fd, off_t offset, int whence, off_t *res) 1481 { 1482 struct thread *td = curthread; 1483 struct proc *p = td->td_proc; 1484 struct filedesc *fdp = p->p_fd; 1485 struct file *fp; 1486 struct vattr vattr; 1487 int error; 1488 1489 if (fd >= fdp->fd_nfiles || 1490 (fp = fdp->fd_ofiles[fd]) == NULL) 1491 return (EBADF); 1492 if (fp->f_type != DTYPE_VNODE) 1493 return (ESPIPE); 1494 switch (whence) { 1495 case L_INCR: 1496 fp->f_offset += offset; 1497 break; 1498 case L_XTND: 1499 error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, td); 1500 if (error) 1501 return (error); 1502 fp->f_offset = offset + vattr.va_size; 1503 break; 1504 case L_SET: 1505 fp->f_offset = offset; 1506 break; 1507 default: 1508 return (EINVAL); 1509 } 1510 *res = fp->f_offset; 1511 return (0); 1512 } 1513 1514 /* 1515 * lseek_args(int fd, int pad, off_t offset, int whence) 1516 * 1517 * Reposition read/write file offset. 1518 */ 1519 int 1520 lseek(struct lseek_args *uap) 1521 { 1522 int error; 1523 1524 error = kern_lseek(uap->fd, uap->offset, uap->whence, 1525 &uap->sysmsg_offset); 1526 1527 return (error); 1528 } 1529 1530 int 1531 kern_access(struct nameidata *nd, int aflags) 1532 { 1533 struct thread *td = curthread; 1534 struct proc *p = td->td_proc; 1535 struct ucred *cred, *tmpcred; 1536 struct vnode *vp; 1537 int error, flags; 1538 1539 cred = p->p_ucred; 1540 /* 1541 * Create and modify a temporary credential instead of one that 1542 * is potentially shared. This could also mess up socket 1543 * buffer accounting which can run in an interrupt context. 1544 */ 1545 tmpcred = crdup(cred); 1546 tmpcred->cr_uid = p->p_ucred->cr_ruid; 1547 tmpcred->cr_groups[0] = p->p_ucred->cr_rgid; 1548 p->p_ucred = tmpcred; 1549 nd->ni_cnd.cn_cred = tmpcred; 1550 error = namei(nd); 1551 if (error) 1552 goto out1; 1553 vp = nd->ni_vp; 1554 1555 /* Flags == 0 means only check for existence. */ 1556 if (aflags) { 1557 flags = 0; 1558 if (aflags & R_OK) 1559 flags |= VREAD; 1560 if (aflags & W_OK) 1561 flags |= VWRITE; 1562 if (aflags & X_OK) 1563 flags |= VEXEC; 1564 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 1565 error = VOP_ACCESS(vp, flags, tmpcred, td); 1566 } 1567 NDFREE(nd, NDF_ONLY_PNBUF); 1568 vput(vp); 1569 out1: 1570 p->p_ucred = cred; 1571 crfree(tmpcred); 1572 return (error); 1573 } 1574 1575 /* 1576 * access_args(char *path, int flags) 1577 * 1578 * Check access permissions. 1579 */ 1580 int 1581 access(struct access_args *uap) 1582 { 1583 struct thread *td = curthread; 1584 struct nameidata nd; 1585 int error; 1586 1587 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF | CNP_NOOBJ, 1588 UIO_USERSPACE, uap->path, td); 1589 1590 error = kern_access(&nd, uap->flags); 1591 1592 return (error); 1593 } 1594 1595 int 1596 kern_stat(struct nameidata *nd, struct stat *st) 1597 { 1598 struct thread *td = curthread; 1599 int error; 1600 1601 error = namei(nd); 1602 if (error) 1603 return (error); 1604 error = vn_stat(nd->ni_vp, st, td); 1605 NDFREE(nd, NDF_ONLY_PNBUF); 1606 vput(nd->ni_vp); 1607 return (error); 1608 } 1609 1610 /* 1611 * stat_args(char *path, struct stat *ub) 1612 * 1613 * Get file status; this version follows links. 1614 */ 1615 int 1616 stat(struct stat_args *uap) 1617 { 1618 struct thread *td = curthread; 1619 struct nameidata nd; 1620 struct stat st; 1621 int error; 1622 1623 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF | CNP_NOOBJ, 1624 UIO_USERSPACE, uap->path, td); 1625 1626 error = kern_stat(&nd, &st); 1627 1628 if (error == 0) 1629 error = copyout(&st, uap->ub, sizeof(*uap->ub)); 1630 return (error); 1631 } 1632 1633 /* 1634 * lstat_args(char *path, struct stat *ub) 1635 * 1636 * Get file status; this version does not follow links. 1637 */ 1638 int 1639 lstat(struct lstat_args *uap) 1640 { 1641 struct thread *td = curthread; 1642 struct nameidata nd; 1643 struct stat st; 1644 int error; 1645 1646 NDINIT(&nd, NAMEI_LOOKUP, CNP_LOCKLEAF | CNP_NOOBJ, 1647 UIO_USERSPACE, SCARG(uap, path), td); 1648 1649 error = kern_stat(&nd, &st); 1650 1651 if (error == 0) 1652 error = copyout(&st, uap->ub, sizeof(*uap->ub)); 1653 return (error); 1654 } 1655 1656 void 1657 cvtnstat(sb, nsb) 1658 struct stat *sb; 1659 struct nstat *nsb; 1660 { 1661 nsb->st_dev = sb->st_dev; 1662 nsb->st_ino = sb->st_ino; 1663 nsb->st_mode = sb->st_mode; 1664 nsb->st_nlink = sb->st_nlink; 1665 nsb->st_uid = sb->st_uid; 1666 nsb->st_gid = sb->st_gid; 1667 nsb->st_rdev = sb->st_rdev; 1668 nsb->st_atimespec = sb->st_atimespec; 1669 nsb->st_mtimespec = sb->st_mtimespec; 1670 nsb->st_ctimespec = sb->st_ctimespec; 1671 nsb->st_size = sb->st_size; 1672 nsb->st_blocks = sb->st_blocks; 1673 nsb->st_blksize = sb->st_blksize; 1674 nsb->st_flags = sb->st_flags; 1675 nsb->st_gen = sb->st_gen; 1676 nsb->st_qspare[0] = sb->st_qspare[0]; 1677 nsb->st_qspare[1] = sb->st_qspare[1]; 1678 } 1679 1680 /* 1681 * nstat_args(char *path, struct nstat *ub) 1682 */ 1683 /* ARGSUSED */ 1684 int 1685 nstat(struct nstat_args *uap) 1686 { 1687 struct thread *td = curthread; 1688 struct stat sb; 1689 struct nstat nsb; 1690 int error; 1691 struct nameidata nd; 1692 1693 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF | CNP_NOOBJ, 1694 UIO_USERSPACE, SCARG(uap, path), td); 1695 if ((error = namei(&nd)) != 0) 1696 return (error); 1697 NDFREE(&nd, NDF_ONLY_PNBUF); 1698 error = vn_stat(nd.ni_vp, &sb, td); 1699 vput(nd.ni_vp); 1700 if (error) 1701 return (error); 1702 cvtnstat(&sb, &nsb); 1703 error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb)); 1704 return (error); 1705 } 1706 1707 /* 1708 * lstat_args(char *path, struct stat *ub) 1709 * 1710 * Get file status; this version does not follow links. 1711 */ 1712 /* ARGSUSED */ 1713 int 1714 nlstat(struct nlstat_args *uap) 1715 { 1716 struct thread *td = curthread; 1717 int error; 1718 struct vnode *vp; 1719 struct stat sb; 1720 struct nstat nsb; 1721 struct nameidata nd; 1722 1723 NDINIT(&nd, NAMEI_LOOKUP, CNP_LOCKLEAF | CNP_NOOBJ, 1724 UIO_USERSPACE, SCARG(uap, path), td); 1725 if ((error = namei(&nd)) != 0) 1726 return (error); 1727 vp = nd.ni_vp; 1728 NDFREE(&nd, NDF_ONLY_PNBUF); 1729 error = vn_stat(vp, &sb, td); 1730 vput(vp); 1731 if (error) 1732 return (error); 1733 cvtnstat(&sb, &nsb); 1734 error = copyout((caddr_t)&nsb, (caddr_t)SCARG(uap, ub), sizeof (nsb)); 1735 return (error); 1736 } 1737 1738 /* 1739 * pathconf_Args(char *path, int name) 1740 * 1741 * Get configurable pathname variables. 1742 */ 1743 /* ARGSUSED */ 1744 int 1745 pathconf(struct pathconf_args *uap) 1746 { 1747 struct thread *td = curthread; 1748 int error; 1749 struct nameidata nd; 1750 1751 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF | CNP_NOOBJ, 1752 UIO_USERSPACE, SCARG(uap, path), td); 1753 if ((error = namei(&nd)) != 0) 1754 return (error); 1755 NDFREE(&nd, NDF_ONLY_PNBUF); 1756 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), uap->sysmsg_fds); 1757 vput(nd.ni_vp); 1758 return (error); 1759 } 1760 1761 /* 1762 * XXX: daver 1763 * kern_readlink isn't properly split yet. There is a copyin burried 1764 * in VOP_READLINK(). 1765 */ 1766 int 1767 kern_readlink(struct nameidata *nd, char *buf, int count, int *res) 1768 { 1769 struct thread *td = curthread; 1770 struct proc *p = td->td_proc; 1771 struct vnode *vp; 1772 struct iovec aiov; 1773 struct uio auio; 1774 int error; 1775 1776 error = namei(nd); 1777 if (error) 1778 return (error); 1779 NDFREE(nd, NDF_ONLY_PNBUF); 1780 vp = nd->ni_vp; 1781 if (vp->v_type != VLNK) 1782 error = EINVAL; 1783 else { 1784 aiov.iov_base = buf; 1785 aiov.iov_len = count; 1786 auio.uio_iov = &aiov; 1787 auio.uio_iovcnt = 1; 1788 auio.uio_offset = 0; 1789 auio.uio_rw = UIO_READ; 1790 auio.uio_segflg = UIO_USERSPACE; 1791 auio.uio_td = td; 1792 auio.uio_resid = count; 1793 error = VOP_READLINK(vp, &auio, p->p_ucred); 1794 } 1795 vput(vp); 1796 *res = count - auio.uio_resid; 1797 return (error); 1798 } 1799 1800 /* 1801 * readlink_args(char *path, char *buf, int count) 1802 * 1803 * Return target name of a symbolic link. 1804 */ 1805 int 1806 readlink(struct readlink_args *uap) 1807 { 1808 struct thread *td = curthread; 1809 struct nameidata nd; 1810 int error; 1811 1812 NDINIT(&nd, NAMEI_LOOKUP, CNP_LOCKLEAF | CNP_NOOBJ, UIO_USERSPACE, 1813 uap->path, td); 1814 1815 error = kern_readlink(&nd, uap->buf, uap->count, 1816 &uap->sysmsg_result); 1817 1818 return (error); 1819 } 1820 1821 static int 1822 setfflags(struct vnode *vp, int flags) 1823 { 1824 struct thread *td = curthread; 1825 struct proc *p = td->td_proc; 1826 int error; 1827 struct vattr vattr; 1828 1829 /* 1830 * Prevent non-root users from setting flags on devices. When 1831 * a device is reused, users can retain ownership of the device 1832 * if they are allowed to set flags and programs assume that 1833 * chown can't fail when done as root. 1834 */ 1835 if ((vp->v_type == VCHR || vp->v_type == VBLK) && 1836 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)) 1837 return (error); 1838 1839 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 1840 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 1841 VATTR_NULL(&vattr); 1842 vattr.va_flags = flags; 1843 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 1844 VOP_UNLOCK(vp, NULL, 0, td); 1845 return (error); 1846 } 1847 1848 /* 1849 * chflags(char *path, int flags) 1850 * 1851 * Change flags of a file given a path name. 1852 */ 1853 /* ARGSUSED */ 1854 int 1855 chflags(struct chflags_args *uap) 1856 { 1857 struct thread *td = curthread; 1858 int error; 1859 struct nameidata nd; 1860 1861 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, 1862 SCARG(uap, path), td); 1863 if ((error = namei(&nd)) != 0) 1864 return (error); 1865 NDFREE(&nd, NDF_ONLY_PNBUF); 1866 error = setfflags(nd.ni_vp, SCARG(uap, flags)); 1867 vrele(nd.ni_vp); 1868 return error; 1869 } 1870 1871 /* 1872 * fchflags_args(int fd, int flags) 1873 * 1874 * Change flags of a file given a file descriptor. 1875 */ 1876 /* ARGSUSED */ 1877 int 1878 fchflags(struct fchflags_args *uap) 1879 { 1880 struct thread *td = curthread; 1881 struct proc *p = td->td_proc; 1882 struct file *fp; 1883 int error; 1884 1885 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 1886 return (error); 1887 return setfflags((struct vnode *) fp->f_data, SCARG(uap, flags)); 1888 } 1889 1890 static int 1891 setfmode(struct vnode *vp, int mode) 1892 { 1893 struct thread *td = curthread; 1894 struct proc *p = td->td_proc; 1895 int error; 1896 struct vattr vattr; 1897 1898 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 1899 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 1900 VATTR_NULL(&vattr); 1901 vattr.va_mode = mode & ALLPERMS; 1902 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 1903 VOP_UNLOCK(vp, NULL, 0, td); 1904 return error; 1905 } 1906 1907 int 1908 kern_chmod(struct nameidata *nd, int mode) 1909 { 1910 int error; 1911 1912 error = namei(nd); 1913 if (error) 1914 return (error); 1915 NDFREE(nd, NDF_ONLY_PNBUF); 1916 error = setfmode(nd->ni_vp, mode); 1917 vrele(nd->ni_vp); 1918 return error; 1919 } 1920 1921 /* 1922 * chmod_args(char *path, int mode) 1923 * 1924 * Change mode of a file given path name. 1925 */ 1926 /* ARGSUSED */ 1927 int 1928 chmod(struct chmod_args *uap) 1929 { 1930 struct thread *td = curthread; 1931 struct nameidata nd; 1932 int error; 1933 1934 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, uap->path, td); 1935 1936 error = kern_chmod(&nd, uap->mode); 1937 1938 return (error); 1939 } 1940 1941 /* 1942 * lchmod_args(char *path, int mode) 1943 * 1944 * Change mode of a file given path name (don't follow links.) 1945 */ 1946 /* ARGSUSED */ 1947 int 1948 lchmod(struct lchmod_args *uap) 1949 { 1950 struct thread *td = curthread; 1951 int error; 1952 struct nameidata nd; 1953 1954 NDINIT(&nd, NAMEI_LOOKUP, 0, UIO_USERSPACE, SCARG(uap, path), td); 1955 if ((error = namei(&nd)) != 0) 1956 return (error); 1957 NDFREE(&nd, NDF_ONLY_PNBUF); 1958 error = setfmode(nd.ni_vp, SCARG(uap, mode)); 1959 vrele(nd.ni_vp); 1960 return error; 1961 } 1962 1963 /* 1964 * fchmod_args(int fd, int mode) 1965 * 1966 * Change mode of a file given a file descriptor. 1967 */ 1968 /* ARGSUSED */ 1969 int 1970 fchmod(struct fchmod_args *uap) 1971 { 1972 struct thread *td = curthread; 1973 struct proc *p = td->td_proc; 1974 struct file *fp; 1975 int error; 1976 1977 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 1978 return (error); 1979 return setfmode((struct vnode *)fp->f_data, SCARG(uap, mode)); 1980 } 1981 1982 static int 1983 setfown(struct vnode *vp, uid_t uid, gid_t gid) 1984 { 1985 struct thread *td = curthread; 1986 struct proc *p = td->td_proc; 1987 int error; 1988 struct vattr vattr; 1989 1990 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 1991 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 1992 VATTR_NULL(&vattr); 1993 vattr.va_uid = uid; 1994 vattr.va_gid = gid; 1995 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 1996 VOP_UNLOCK(vp, NULL, 0, td); 1997 return error; 1998 } 1999 2000 int 2001 kern_chown(struct nameidata *nd, int uid, int gid) 2002 { 2003 int error; 2004 2005 error = namei(nd); 2006 if (error) 2007 return (error); 2008 NDFREE(nd, NDF_ONLY_PNBUF); 2009 error = setfown(nd->ni_vp, uid, gid); 2010 vrele(nd->ni_vp); 2011 return (error); 2012 } 2013 2014 /* 2015 * chown(char *path, int uid, int gid) 2016 * 2017 * Set ownership given a path name. 2018 */ 2019 int 2020 chown(struct chown_args *uap) 2021 { 2022 struct thread *td = curthread; 2023 struct nameidata nd; 2024 int error; 2025 2026 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, uap->path, td); 2027 2028 error = kern_chown(&nd, uap->uid, uap->gid); 2029 2030 return (error); 2031 } 2032 2033 /* 2034 * lchown_args(char *path, int uid, int gid) 2035 * 2036 * Set ownership given a path name, do not cross symlinks. 2037 */ 2038 int 2039 lchown(struct lchown_args *uap) 2040 { 2041 struct thread *td = curthread; 2042 int error; 2043 struct nameidata nd; 2044 2045 NDINIT(&nd, NAMEI_LOOKUP, 0, UIO_USERSPACE, uap->path, td); 2046 2047 error = kern_chown(&nd, uap->uid, uap->gid); 2048 2049 return (error); 2050 } 2051 2052 /* 2053 * fchown_args(int fd, int uid, int gid) 2054 * 2055 * Set ownership given a file descriptor. 2056 */ 2057 /* ARGSUSED */ 2058 int 2059 fchown(struct fchown_args *uap) 2060 { 2061 struct thread *td = curthread; 2062 struct proc *p = td->td_proc; 2063 struct file *fp; 2064 int error; 2065 2066 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2067 return (error); 2068 return setfown((struct vnode *)fp->f_data, 2069 SCARG(uap, uid), SCARG(uap, gid)); 2070 } 2071 2072 static int 2073 getutimes(const struct timeval *tvp, struct timespec *tsp) 2074 { 2075 struct timeval tv[2]; 2076 2077 if (tvp == NULL) { 2078 microtime(&tv[0]); 2079 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]); 2080 tsp[1] = tsp[0]; 2081 } else { 2082 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 2083 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 2084 } 2085 return 0; 2086 } 2087 2088 static int 2089 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag) 2090 { 2091 struct thread *td = curthread; 2092 struct proc *p = td->td_proc; 2093 int error; 2094 struct vattr vattr; 2095 2096 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2097 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 2098 VATTR_NULL(&vattr); 2099 vattr.va_atime = ts[0]; 2100 vattr.va_mtime = ts[1]; 2101 if (nullflag) 2102 vattr.va_vaflags |= VA_UTIMES_NULL; 2103 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 2104 VOP_UNLOCK(vp, NULL, 0, td); 2105 return error; 2106 } 2107 2108 int 2109 kern_utimes(struct nameidata *nd, struct timeval *tptr) 2110 { 2111 struct timespec ts[2]; 2112 int error; 2113 2114 error = getutimes(tptr, ts); 2115 if (error) 2116 return (error); 2117 error = namei(nd); 2118 if (error) 2119 return (error); 2120 NDFREE(nd, NDF_ONLY_PNBUF); 2121 error = setutimes(nd->ni_vp, ts, tptr == NULL); 2122 vrele(nd->ni_vp); 2123 return (error); 2124 } 2125 2126 /* 2127 * utimes_args(char *path, struct timeval *tptr) 2128 * 2129 * Set the access and modification times of a file. 2130 */ 2131 int 2132 utimes(struct utimes_args *uap) 2133 { 2134 struct thread *td = curthread; 2135 struct timeval tv[2]; 2136 struct nameidata nd; 2137 int error; 2138 2139 if (uap->tptr) { 2140 error = copyin(uap->tptr, tv, sizeof(tv)); 2141 if (error) 2142 return (error); 2143 } 2144 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, uap->path, td); 2145 2146 error = kern_utimes(&nd, uap->tptr ? tv : NULL); 2147 2148 return (error); 2149 } 2150 2151 /* 2152 * lutimes_args(char *path, struct timeval *tptr) 2153 * 2154 * Set the access and modification times of a file. 2155 */ 2156 int 2157 lutimes(struct lutimes_args *uap) 2158 { 2159 struct thread *td = curthread; 2160 struct timeval tv[2]; 2161 struct nameidata nd; 2162 int error; 2163 2164 if (uap->tptr) { 2165 error = copyin(uap->tptr, tv, sizeof(tv)); 2166 if (error) 2167 return (error); 2168 } 2169 NDINIT(&nd, NAMEI_LOOKUP, 0, UIO_USERSPACE, uap->path, td); 2170 2171 error = kern_utimes(&nd, uap->tptr ? tv : NULL); 2172 2173 return (error); 2174 } 2175 2176 int 2177 kern_futimes(int fd, struct timeval *tptr) 2178 { 2179 struct thread *td = curthread; 2180 struct proc *p = td->td_proc; 2181 struct timespec ts[2]; 2182 struct file *fp; 2183 int error; 2184 2185 error = getutimes(tptr, ts); 2186 if (error) 2187 return (error); 2188 error = getvnode(p->p_fd, fd, &fp); 2189 if (error) 2190 return (error); 2191 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL); 2192 return (error); 2193 } 2194 2195 /* 2196 * futimes_args(int fd, struct timeval *tptr) 2197 * 2198 * Set the access and modification times of a file. 2199 */ 2200 int 2201 futimes(struct futimes_args *uap) 2202 { 2203 struct timeval tv[2]; 2204 int error; 2205 2206 if (uap->tptr) { 2207 error = copyin(uap->tptr, tv, sizeof(tv)); 2208 if (error) 2209 return (error); 2210 } 2211 2212 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL); 2213 2214 return (error); 2215 } 2216 2217 int 2218 kern_truncate(struct nameidata* nd, off_t length) 2219 { 2220 struct thread *td = curthread; 2221 struct proc *p = td->td_proc; 2222 struct vnode *vp; 2223 struct vattr vattr; 2224 int error; 2225 2226 if (length < 0) 2227 return(EINVAL); 2228 if ((error = namei(nd)) != 0) 2229 return (error); 2230 vp = nd->ni_vp; 2231 NDFREE(nd, NDF_ONLY_PNBUF); 2232 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2233 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 2234 if (vp->v_type == VDIR) 2235 error = EISDIR; 2236 else if ((error = vn_writechk(vp)) == 0 && 2237 (error = VOP_ACCESS(vp, VWRITE, p->p_ucred, td)) == 0) { 2238 VATTR_NULL(&vattr); 2239 vattr.va_size = length; 2240 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 2241 } 2242 vput(vp); 2243 return (error); 2244 } 2245 2246 /* 2247 * truncate(char *path, int pad, off_t length) 2248 * 2249 * Truncate a file given its path name. 2250 */ 2251 int 2252 truncate(struct truncate_args *uap) 2253 { 2254 struct thread *td = curthread; 2255 struct nameidata nd; 2256 int error; 2257 2258 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, uap->path, td); 2259 2260 error = kern_truncate(&nd, uap->length); 2261 2262 return error; 2263 } 2264 2265 int 2266 kern_ftruncate(int fd, off_t length) 2267 { 2268 struct thread *td = curthread; 2269 struct proc *p = td->td_proc; 2270 struct vattr vattr; 2271 struct vnode *vp; 2272 struct file *fp; 2273 int error; 2274 2275 if (length < 0) 2276 return(EINVAL); 2277 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 2278 return (error); 2279 if ((fp->f_flag & FWRITE) == 0) 2280 return (EINVAL); 2281 vp = (struct vnode *)fp->f_data; 2282 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2283 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 2284 if (vp->v_type == VDIR) 2285 error = EISDIR; 2286 else if ((error = vn_writechk(vp)) == 0) { 2287 VATTR_NULL(&vattr); 2288 vattr.va_size = length; 2289 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td); 2290 } 2291 VOP_UNLOCK(vp, NULL, 0, td); 2292 return (error); 2293 } 2294 2295 /* 2296 * ftruncate_args(int fd, int pad, off_t length) 2297 * 2298 * Truncate a file given a file descriptor. 2299 */ 2300 int 2301 ftruncate(struct ftruncate_args *uap) 2302 { 2303 int error; 2304 2305 error = kern_ftruncate(uap->fd, uap->length); 2306 2307 return (error); 2308 } 2309 2310 /* 2311 * fsync(int fd) 2312 * 2313 * Sync an open file. 2314 */ 2315 /* ARGSUSED */ 2316 int 2317 fsync(struct fsync_args *uap) 2318 { 2319 struct thread *td = curthread; 2320 struct proc *p = td->td_proc; 2321 struct vnode *vp; 2322 struct file *fp; 2323 vm_object_t obj; 2324 int error; 2325 2326 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2327 return (error); 2328 vp = (struct vnode *)fp->f_data; 2329 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 2330 if (VOP_GETVOBJECT(vp, &obj) == 0) 2331 vm_object_page_clean(obj, 0, 0, 0); 2332 if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) == 0 && 2333 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) && 2334 bioops.io_fsync) 2335 error = (*bioops.io_fsync)(vp); 2336 VOP_UNLOCK(vp, NULL, 0, td); 2337 return (error); 2338 } 2339 2340 int 2341 kern_rename(struct nameidata *fromnd, struct nameidata *tond) 2342 { 2343 struct thread *td = curthread; 2344 struct proc *p = td->td_proc; 2345 struct vnode *tvp, *fvp, *tdvp; 2346 int error; 2347 2348 bwillwrite(); 2349 error = namei(fromnd); 2350 if (error) 2351 return (error); 2352 fvp = fromnd->ni_vp; 2353 if (fromnd->ni_vp->v_type == VDIR) 2354 tond->ni_cnd.cn_flags |= CNP_WILLBEDIR; 2355 error = namei(tond); 2356 if (error) { 2357 /* Translate error code for rename("dir1", "dir2/."). */ 2358 if (error == EISDIR && fvp->v_type == VDIR) 2359 error = EINVAL; 2360 NDFREE(fromnd, NDF_ONLY_PNBUF); 2361 vrele(fromnd->ni_dvp); 2362 vrele(fvp); 2363 goto out1; 2364 } 2365 tdvp = tond->ni_dvp; 2366 tvp = tond->ni_vp; 2367 if (tvp != NULL) { 2368 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 2369 error = ENOTDIR; 2370 goto out; 2371 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 2372 error = EISDIR; 2373 goto out; 2374 } 2375 } 2376 if (fvp == tdvp) 2377 error = EINVAL; 2378 /* 2379 * If the source is the same as the destination (that is, if they 2380 * are links to the same vnode), then there is nothing to do. 2381 */ 2382 if (fvp == tvp) 2383 error = -1; 2384 out: 2385 if (!error) { 2386 VOP_LEASE(tdvp, td, p->p_ucred, LEASE_WRITE); 2387 if (fromnd->ni_dvp != tdvp) { 2388 VOP_LEASE(fromnd->ni_dvp, td, p->p_ucred, LEASE_WRITE); 2389 } 2390 if (tvp) { 2391 VOP_LEASE(tvp, td, p->p_ucred, LEASE_WRITE); 2392 } 2393 error = VOP_RENAME(fromnd->ni_dvp, NCPNULL, fromnd->ni_vp, 2394 &fromnd->ni_cnd, tond->ni_dvp, NCPNULL, tond->ni_vp, 2395 &tond->ni_cnd); 2396 NDFREE(fromnd, NDF_ONLY_PNBUF); 2397 NDFREE(tond, NDF_ONLY_PNBUF); 2398 } else { 2399 NDFREE(fromnd, NDF_ONLY_PNBUF); 2400 NDFREE(tond, NDF_ONLY_PNBUF); 2401 if (tdvp == tvp) 2402 vrele(tdvp); 2403 else 2404 vput(tdvp); 2405 if (tvp) 2406 vput(tvp); 2407 vrele(fromnd->ni_dvp); 2408 vrele(fvp); 2409 } 2410 vrele(tond->ni_startdir); 2411 ASSERT_VOP_UNLOCKED(fromnd->ni_dvp, "rename"); 2412 ASSERT_VOP_UNLOCKED(fromnd->ni_vp, "rename"); 2413 ASSERT_VOP_UNLOCKED(tond->ni_dvp, "rename"); 2414 ASSERT_VOP_UNLOCKED(tond->ni_vp, "rename"); 2415 out1: 2416 if (fromnd->ni_startdir) 2417 vrele(fromnd->ni_startdir); 2418 if (error == -1) 2419 return (0); 2420 return (error); 2421 } 2422 2423 /* 2424 * rename_args(char *from, char *to) 2425 * 2426 * Rename files. Source and destination must either both be directories, 2427 * or both not be directories. If target is a directory, it must be empty. 2428 */ 2429 int 2430 rename(struct rename_args *uap) 2431 { 2432 struct thread *td = curthread; 2433 struct nameidata fromnd, tond; 2434 int error; 2435 2436 NDINIT(&fromnd, NAMEI_DELETE, CNP_WANTPARENT | CNP_SAVESTART, 2437 UIO_USERSPACE, uap->from, td); 2438 NDINIT(&tond, NAMEI_RENAME, 2439 CNP_LOCKPARENT | CNP_LOCKLEAF | CNP_NOCACHE | 2440 CNP_SAVESTART | CNP_NOOBJ, 2441 UIO_USERSPACE, uap->to, td); 2442 2443 error = kern_rename(&fromnd, &tond); 2444 2445 return (error); 2446 } 2447 2448 int 2449 kern_mkdir(struct nameidata *nd, int mode) 2450 { 2451 struct thread *td = curthread; 2452 struct proc *p = td->td_proc; 2453 struct vnode *vp; 2454 struct vattr vattr; 2455 int error; 2456 2457 bwillwrite(); 2458 nd->ni_cnd.cn_flags |= CNP_WILLBEDIR; 2459 error = namei(nd); 2460 if (error) 2461 return (error); 2462 vp = nd->ni_vp; 2463 if (vp) { 2464 NDFREE(nd, NDF_ONLY_PNBUF); 2465 if (nd->ni_dvp == vp) 2466 vrele(nd->ni_dvp); 2467 else 2468 vput(nd->ni_dvp); 2469 vrele(vp); 2470 return (EEXIST); 2471 } 2472 VATTR_NULL(&vattr); 2473 vattr.va_type = VDIR; 2474 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask; 2475 VOP_LEASE(nd->ni_dvp, td, p->p_ucred, LEASE_WRITE); 2476 error = VOP_MKDIR(nd->ni_dvp, NCPNULL, &nd->ni_vp, &nd->ni_cnd, 2477 &vattr); 2478 NDFREE(nd, NDF_ONLY_PNBUF); 2479 vput(nd->ni_dvp); 2480 if (error == 0) 2481 vput(nd->ni_vp); 2482 ASSERT_VOP_UNLOCKED(nd->ni_dvp, "mkdir"); 2483 ASSERT_VOP_UNLOCKED(nd->ni_vp, "mkdir"); 2484 return (error); 2485 } 2486 2487 /* 2488 * mkdir_args(char *path, int mode) 2489 * 2490 * Make a directory file. 2491 */ 2492 /* ARGSUSED */ 2493 int 2494 mkdir(struct mkdir_args *uap) 2495 { 2496 struct thread *td = curthread; 2497 struct nameidata nd; 2498 int error; 2499 2500 NDINIT(&nd, NAMEI_CREATE, CNP_LOCKPARENT, UIO_USERSPACE, uap->path, 2501 td); 2502 2503 error = kern_mkdir(&nd, uap->mode); 2504 2505 return (error); 2506 } 2507 2508 int 2509 kern_rmdir(struct nameidata *nd) 2510 { 2511 struct thread *td = curthread; 2512 struct proc *p = td->td_proc; 2513 struct vnode *vp; 2514 int error; 2515 2516 bwillwrite(); 2517 error = namei(nd); 2518 if (error) 2519 return (error); 2520 vp = nd->ni_vp; 2521 if (vp->v_type != VDIR) { 2522 error = ENOTDIR; 2523 goto out; 2524 } 2525 /* 2526 * No rmdir "." please. 2527 */ 2528 if (nd->ni_dvp == vp) { 2529 error = EINVAL; 2530 goto out; 2531 } 2532 /* 2533 * The root of a mounted filesystem cannot be deleted. 2534 */ 2535 if (vp->v_flag & VROOT) 2536 error = EBUSY; 2537 else { 2538 VOP_LEASE(nd->ni_dvp, td, p->p_ucred, LEASE_WRITE); 2539 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2540 error = VOP_RMDIR(nd->ni_dvp, NCPNULL, nd->ni_vp, 2541 &nd->ni_cnd); 2542 } 2543 out: 2544 NDFREE(nd, NDF_ONLY_PNBUF); 2545 if (nd->ni_dvp == vp) 2546 vrele(nd->ni_dvp); 2547 else 2548 vput(nd->ni_dvp); 2549 if (vp != NULLVP) 2550 vput(vp); 2551 ASSERT_VOP_UNLOCKED(nd->ni_dvp, "rmdir"); 2552 ASSERT_VOP_UNLOCKED(nd->ni_vp, "rmdir"); 2553 return (error); 2554 } 2555 2556 /* 2557 * rmdir_args(char *path) 2558 * 2559 * Remove a directory file. 2560 */ 2561 /* ARGSUSED */ 2562 int 2563 rmdir(struct rmdir_args *uap) 2564 { 2565 struct thread *td = curthread; 2566 struct nameidata nd; 2567 int error; 2568 2569 NDINIT(&nd, NAMEI_DELETE, CNP_LOCKPARENT | CNP_LOCKLEAF, 2570 UIO_USERSPACE, uap->path, td); 2571 2572 error = kern_rmdir(&nd); 2573 2574 return (error); 2575 } 2576 2577 int 2578 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res) 2579 { 2580 struct thread *td = curthread; 2581 struct proc *p = td->td_proc; 2582 struct vnode *vp; 2583 struct file *fp; 2584 struct uio auio; 2585 struct iovec aiov; 2586 long loff; 2587 int error, eofflag; 2588 2589 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 2590 return (error); 2591 if ((fp->f_flag & FREAD) == 0) 2592 return (EBADF); 2593 vp = (struct vnode *)fp->f_data; 2594 unionread: 2595 if (vp->v_type != VDIR) 2596 return (EINVAL); 2597 aiov.iov_base = buf; 2598 aiov.iov_len = count; 2599 auio.uio_iov = &aiov; 2600 auio.uio_iovcnt = 1; 2601 auio.uio_rw = UIO_READ; 2602 auio.uio_segflg = UIO_USERSPACE; 2603 auio.uio_td = td; 2604 auio.uio_resid = count; 2605 /* vn_lock(vp, NULL, LK_SHARED | LK_RETRY, td); */ 2606 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 2607 loff = auio.uio_offset = fp->f_offset; 2608 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL); 2609 fp->f_offset = auio.uio_offset; 2610 VOP_UNLOCK(vp, NULL, 0, td); 2611 if (error) 2612 return (error); 2613 if (count == auio.uio_resid) { 2614 if (union_dircheckp) { 2615 error = union_dircheckp(td, &vp, fp); 2616 if (error == -1) 2617 goto unionread; 2618 if (error) 2619 return (error); 2620 } 2621 if ((vp->v_flag & VROOT) && 2622 (vp->v_mount->mnt_flag & MNT_UNION)) { 2623 struct vnode *tvp = vp; 2624 vp = vp->v_mount->mnt_vnodecovered; 2625 VREF(vp); 2626 fp->f_data = (caddr_t) vp; 2627 fp->f_offset = 0; 2628 vrele(tvp); 2629 goto unionread; 2630 } 2631 } 2632 if (basep) { 2633 *basep = loff; 2634 } 2635 *res = count - auio.uio_resid; 2636 return (error); 2637 } 2638 2639 /* 2640 * getdirentries_args(int fd, char *buf, u_int conut, long *basep) 2641 * 2642 * Read a block of directory entries in a file system independent format. 2643 */ 2644 int 2645 getdirentries(struct getdirentries_args *uap) 2646 { 2647 long base; 2648 int error; 2649 2650 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base, 2651 &uap->sysmsg_result); 2652 2653 if (error == 0) 2654 error = copyout(&base, uap->basep, sizeof(*uap->basep)); 2655 return (error); 2656 } 2657 2658 /* 2659 * getdents_args(int fd, char *buf, size_t count) 2660 */ 2661 int 2662 getdents(struct getdents_args *uap) 2663 { 2664 int error; 2665 2666 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL, 2667 &uap->sysmsg_result); 2668 2669 return (error); 2670 } 2671 2672 /* 2673 * umask(int newmask) 2674 * 2675 * Set the mode mask for creation of filesystem nodes. 2676 * 2677 * MP SAFE 2678 */ 2679 int 2680 umask(struct umask_args *uap) 2681 { 2682 struct thread *td = curthread; 2683 struct proc *p = td->td_proc; 2684 struct filedesc *fdp; 2685 2686 fdp = p->p_fd; 2687 uap->sysmsg_result = fdp->fd_cmask; 2688 fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS; 2689 return (0); 2690 } 2691 2692 /* 2693 * revoke(char *path) 2694 * 2695 * Void all references to file by ripping underlying filesystem 2696 * away from vnode. 2697 */ 2698 /* ARGSUSED */ 2699 int 2700 revoke(struct revoke_args *uap) 2701 { 2702 struct thread *td = curthread; 2703 struct proc *p = td->td_proc; 2704 struct vnode *vp; 2705 struct vattr vattr; 2706 int error; 2707 struct nameidata nd; 2708 2709 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, SCARG(uap, path), td); 2710 if ((error = namei(&nd)) != 0) 2711 return (error); 2712 vp = nd.ni_vp; 2713 NDFREE(&nd, NDF_ONLY_PNBUF); 2714 if (vp->v_type != VCHR && vp->v_type != VBLK) { 2715 error = EINVAL; 2716 goto out; 2717 } 2718 if ((error = VOP_GETATTR(vp, &vattr, td)) != 0) 2719 goto out; 2720 if (p->p_ucred->cr_uid != vattr.va_uid && 2721 (error = suser_cred(p->p_ucred, PRISON_ROOT))) 2722 goto out; 2723 if (vcount(vp) > 1) 2724 VOP_REVOKE(vp, REVOKEALL); 2725 out: 2726 vrele(vp); 2727 return (error); 2728 } 2729 2730 /* 2731 * Convert a user file descriptor to a kernel file entry. 2732 */ 2733 int 2734 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 2735 { 2736 struct file *fp; 2737 2738 if ((u_int)fd >= fdp->fd_nfiles || 2739 (fp = fdp->fd_ofiles[fd]) == NULL) 2740 return (EBADF); 2741 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO) 2742 return (EINVAL); 2743 *fpp = fp; 2744 return (0); 2745 } 2746 /* 2747 * getfh_args(char *fname, fhandle_t *fhp) 2748 * 2749 * Get (NFS) file handle 2750 */ 2751 int 2752 getfh(struct getfh_args *uap) 2753 { 2754 struct thread *td = curthread; 2755 struct nameidata nd; 2756 fhandle_t fh; 2757 struct vnode *vp; 2758 int error; 2759 2760 /* 2761 * Must be super user 2762 */ 2763 error = suser(td); 2764 if (error) 2765 return (error); 2766 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, uap->fname, td); 2767 error = namei(&nd); 2768 if (error) 2769 return (error); 2770 NDFREE(&nd, NDF_ONLY_PNBUF); 2771 vp = nd.ni_vp; 2772 bzero(&fh, sizeof(fh)); 2773 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 2774 error = VFS_VPTOFH(vp, &fh.fh_fid); 2775 vput(vp); 2776 if (error) 2777 return (error); 2778 error = copyout(&fh, uap->fhp, sizeof (fh)); 2779 return (error); 2780 } 2781 2782 /* 2783 * fhopen_args(const struct fhandle *u_fhp, int flags) 2784 * 2785 * syscall for the rpc.lockd to use to translate a NFS file handle into 2786 * an open descriptor. 2787 * 2788 * warning: do not remove the suser() call or this becomes one giant 2789 * security hole. 2790 */ 2791 int 2792 fhopen(struct fhopen_args *uap) 2793 { 2794 struct thread *td = curthread; 2795 struct proc *p = td->td_proc; 2796 struct mount *mp; 2797 struct vnode *vp; 2798 struct fhandle fhp; 2799 struct vattr vat; 2800 struct vattr *vap = &vat; 2801 struct flock lf; 2802 struct file *fp; 2803 struct filedesc *fdp = p->p_fd; 2804 int fmode, mode, error, type; 2805 struct file *nfp; 2806 int indx; 2807 2808 /* 2809 * Must be super user 2810 */ 2811 error = suser(td); 2812 if (error) 2813 return (error); 2814 2815 fmode = FFLAGS(SCARG(uap, flags)); 2816 /* why not allow a non-read/write open for our lockd? */ 2817 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 2818 return (EINVAL); 2819 error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp)); 2820 if (error) 2821 return(error); 2822 /* find the mount point */ 2823 mp = vfs_getvfs(&fhp.fh_fsid); 2824 if (mp == NULL) 2825 return (ESTALE); 2826 /* now give me my vnode, it gets returned to me locked */ 2827 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp); 2828 if (error) 2829 return (error); 2830 /* 2831 * from now on we have to make sure not 2832 * to forget about the vnode 2833 * any error that causes an abort must vput(vp) 2834 * just set error = err and 'goto bad;'. 2835 */ 2836 2837 /* 2838 * from vn_open 2839 */ 2840 if (vp->v_type == VLNK) { 2841 error = EMLINK; 2842 goto bad; 2843 } 2844 if (vp->v_type == VSOCK) { 2845 error = EOPNOTSUPP; 2846 goto bad; 2847 } 2848 mode = 0; 2849 if (fmode & (FWRITE | O_TRUNC)) { 2850 if (vp->v_type == VDIR) { 2851 error = EISDIR; 2852 goto bad; 2853 } 2854 error = vn_writechk(vp); 2855 if (error) 2856 goto bad; 2857 mode |= VWRITE; 2858 } 2859 if (fmode & FREAD) 2860 mode |= VREAD; 2861 if (mode) { 2862 error = VOP_ACCESS(vp, mode, p->p_ucred, td); 2863 if (error) 2864 goto bad; 2865 } 2866 if (fmode & O_TRUNC) { 2867 VOP_UNLOCK(vp, NULL, 0, td); /* XXX */ 2868 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2869 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); /* XXX */ 2870 VATTR_NULL(vap); 2871 vap->va_size = 0; 2872 error = VOP_SETATTR(vp, vap, p->p_ucred, td); 2873 if (error) 2874 goto bad; 2875 } 2876 error = VOP_OPEN(vp, fmode, p->p_ucred, td); 2877 if (error) 2878 goto bad; 2879 /* 2880 * Make sure that a VM object is created for VMIO support. 2881 */ 2882 if (vn_canvmio(vp) == TRUE) { 2883 if ((error = vfs_object_create(vp, td)) != 0) 2884 goto bad; 2885 } 2886 if (fmode & FWRITE) 2887 vp->v_writecount++; 2888 2889 /* 2890 * end of vn_open code 2891 */ 2892 2893 if ((error = falloc(p, &nfp, &indx)) != 0) { 2894 if (fmode & FWRITE) 2895 vp->v_writecount--; 2896 goto bad; 2897 } 2898 fp = nfp; 2899 2900 /* 2901 * hold an extra reference to avoid having fp ripped out 2902 * from under us while we block in the lock op. 2903 */ 2904 fhold(fp); 2905 nfp->f_data = (caddr_t)vp; 2906 nfp->f_flag = fmode & FMASK; 2907 nfp->f_ops = &vnops; 2908 nfp->f_type = DTYPE_VNODE; 2909 if (fmode & (O_EXLOCK | O_SHLOCK)) { 2910 lf.l_whence = SEEK_SET; 2911 lf.l_start = 0; 2912 lf.l_len = 0; 2913 if (fmode & O_EXLOCK) 2914 lf.l_type = F_WRLCK; 2915 else 2916 lf.l_type = F_RDLCK; 2917 type = F_FLOCK; 2918 if ((fmode & FNONBLOCK) == 0) 2919 type |= F_WAIT; 2920 VOP_UNLOCK(vp, NULL, 0, td); 2921 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) { 2922 /* 2923 * lock request failed. Normally close the descriptor 2924 * but handle the case where someone might have dup()d 2925 * or close()d it when we weren't looking. 2926 */ 2927 if (fdp->fd_ofiles[indx] == fp) { 2928 fdp->fd_ofiles[indx] = NULL; 2929 fdrop(fp, td); 2930 } 2931 2932 /* 2933 * release our private reference. 2934 */ 2935 fdrop(fp, td); 2936 return (error); 2937 } 2938 vn_lock(vp, NULL, LK_EXCLUSIVE | LK_RETRY, td); 2939 fp->f_flag |= FHASLOCK; 2940 } 2941 if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0)) 2942 vfs_object_create(vp, td); 2943 2944 VOP_UNLOCK(vp, NULL, 0, td); 2945 fdrop(fp, td); 2946 uap->sysmsg_result = indx; 2947 return (0); 2948 2949 bad: 2950 vput(vp); 2951 return (error); 2952 } 2953 2954 /* 2955 * fhstat_args(struct fhandle *u_fhp, struct stat *sb) 2956 */ 2957 int 2958 fhstat(struct fhstat_args *uap) 2959 { 2960 struct thread *td = curthread; 2961 struct stat sb; 2962 fhandle_t fh; 2963 struct mount *mp; 2964 struct vnode *vp; 2965 int error; 2966 2967 /* 2968 * Must be super user 2969 */ 2970 error = suser(td); 2971 if (error) 2972 return (error); 2973 2974 error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t)); 2975 if (error) 2976 return (error); 2977 2978 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 2979 return (ESTALE); 2980 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 2981 return (error); 2982 error = vn_stat(vp, &sb, td); 2983 vput(vp); 2984 if (error) 2985 return (error); 2986 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2987 return (error); 2988 } 2989 2990 /* 2991 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf) 2992 */ 2993 int 2994 fhstatfs(struct fhstatfs_args *uap) 2995 { 2996 struct thread *td = curthread; 2997 struct statfs *sp; 2998 struct mount *mp; 2999 struct vnode *vp; 3000 struct statfs sb; 3001 fhandle_t fh; 3002 int error; 3003 3004 /* 3005 * Must be super user 3006 */ 3007 if ((error = suser(td))) 3008 return (error); 3009 3010 if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0) 3011 return (error); 3012 3013 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 3014 return (ESTALE); 3015 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 3016 return (error); 3017 mp = vp->v_mount; 3018 sp = &mp->mnt_stat; 3019 vput(vp); 3020 if ((error = VFS_STATFS(mp, sp, td)) != 0) 3021 return (error); 3022 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 3023 if (suser(td)) { 3024 bcopy((caddr_t)sp, (caddr_t)&sb, sizeof(sb)); 3025 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0; 3026 sp = &sb; 3027 } 3028 return (copyout(sp, SCARG(uap, buf), sizeof(*sp))); 3029 } 3030 3031 /* 3032 * Syscall to push extended attribute configuration information into the 3033 * VFS. Accepts a path, which it converts to a mountpoint, as well as 3034 * a command (int cmd), and attribute name and misc data. For now, the 3035 * attribute name is left in userspace for consumption by the VFS_op. 3036 * It will probably be changed to be copied into sysspace by the 3037 * syscall in the future, once issues with various consumers of the 3038 * attribute code have raised their hands. 3039 * 3040 * Currently this is used only by UFS Extended Attributes. 3041 */ 3042 int 3043 extattrctl(struct extattrctl_args *uap) 3044 { 3045 struct thread *td = curthread; 3046 struct nameidata nd; 3047 struct mount *mp; 3048 int error; 3049 3050 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW, UIO_USERSPACE, SCARG(uap, path), td); 3051 if ((error = namei(&nd)) != 0) 3052 return (error); 3053 mp = nd.ni_vp->v_mount; 3054 NDFREE(&nd, 0); 3055 return (VFS_EXTATTRCTL(mp, SCARG(uap, cmd), SCARG(uap, attrname), 3056 SCARG(uap, arg), td)); 3057 } 3058 3059 /* 3060 * Syscall to set a named extended attribute on a file or directory. 3061 * Accepts attribute name, and a uio structure pointing to the data to set. 3062 * The uio is consumed in the style of writev(). The real work happens 3063 * in VOP_SETEXTATTR(). 3064 */ 3065 int 3066 extattr_set_file(struct extattr_set_file_args *uap) 3067 { 3068 struct thread *td = curthread; 3069 struct proc *p = td->td_proc; 3070 struct nameidata nd; 3071 struct uio auio; 3072 struct iovec *iov, *needfree = NULL, aiov[UIO_SMALLIOV]; 3073 char attrname[EXTATTR_MAXNAMELEN]; 3074 u_int iovlen, cnt; 3075 int error, i; 3076 3077 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN); 3078 if (error) 3079 return (error); 3080 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, 3081 SCARG(uap, path), td); 3082 if ((error = namei(&nd)) != 0) 3083 return(error); 3084 iovlen = uap->iovcnt * sizeof(struct iovec); 3085 if (uap->iovcnt > UIO_SMALLIOV) { 3086 if (uap->iovcnt > UIO_MAXIOV) { 3087 error = EINVAL; 3088 goto done; 3089 } 3090 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 3091 needfree = iov; 3092 } else 3093 iov = aiov; 3094 auio.uio_iov = iov; 3095 auio.uio_iovcnt = uap->iovcnt; 3096 auio.uio_rw = UIO_WRITE; 3097 auio.uio_segflg = UIO_USERSPACE; 3098 auio.uio_td = td; 3099 auio.uio_offset = 0; 3100 if ((error = copyin((caddr_t)uap->iovp, (caddr_t)iov, iovlen))) 3101 goto done; 3102 auio.uio_resid = 0; 3103 for (i = 0; i < uap->iovcnt; i++) { 3104 if (iov->iov_len > INT_MAX - auio.uio_resid) { 3105 error = EINVAL; 3106 goto done; 3107 } 3108 auio.uio_resid += iov->iov_len; 3109 iov++; 3110 } 3111 cnt = auio.uio_resid; 3112 error = VOP_SETEXTATTR(nd.ni_vp, attrname, &auio, p->p_ucred, td); 3113 cnt -= auio.uio_resid; 3114 uap->sysmsg_result = cnt; 3115 done: 3116 if (needfree) 3117 FREE(needfree, M_IOV); 3118 NDFREE(&nd, 0); 3119 return (error); 3120 } 3121 3122 /* 3123 * Syscall to get a named extended attribute on a file or directory. 3124 * Accepts attribute name, and a uio structure pointing to a buffer for the 3125 * data. The uio is consumed in the style of readv(). The real work 3126 * happens in VOP_GETEXTATTR(); 3127 */ 3128 int 3129 extattr_get_file(struct extattr_get_file_args *uap) 3130 { 3131 struct thread *td = curthread; 3132 struct proc *p = td->td_proc; 3133 struct nameidata nd; 3134 struct uio auio; 3135 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 3136 char attrname[EXTATTR_MAXNAMELEN]; 3137 u_int iovlen, cnt; 3138 int error, i; 3139 3140 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN); 3141 if (error) 3142 return (error); 3143 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, 3144 SCARG(uap, path), td); 3145 if ((error = namei(&nd)) != 0) 3146 return (error); 3147 iovlen = uap->iovcnt * sizeof (struct iovec); 3148 if (uap->iovcnt > UIO_SMALLIOV) { 3149 if (uap->iovcnt > UIO_MAXIOV) { 3150 NDFREE(&nd, 0); 3151 return (EINVAL); 3152 } 3153 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 3154 needfree = iov; 3155 } else { 3156 iov = aiov; 3157 needfree = NULL; 3158 } 3159 auio.uio_iov = iov; 3160 auio.uio_iovcnt = uap->iovcnt; 3161 auio.uio_rw = UIO_READ; 3162 auio.uio_segflg = UIO_USERSPACE; 3163 auio.uio_td = td; 3164 auio.uio_offset = 0; 3165 if ((error = copyin((caddr_t)uap->iovp, (caddr_t)iov, iovlen))) 3166 goto done; 3167 auio.uio_resid = 0; 3168 for (i = 0; i < uap->iovcnt; i++) { 3169 if (iov->iov_len > INT_MAX - auio.uio_resid) { 3170 error = EINVAL; 3171 goto done; 3172 } 3173 auio.uio_resid += iov->iov_len; 3174 iov++; 3175 } 3176 cnt = auio.uio_resid; 3177 error = VOP_GETEXTATTR(nd.ni_vp, attrname, &auio, p->p_ucred, td); 3178 cnt -= auio.uio_resid; 3179 uap->sysmsg_result = cnt; 3180 done: 3181 if (needfree) 3182 FREE(needfree, M_IOV); 3183 NDFREE(&nd, 0); 3184 return(error); 3185 } 3186 3187 /* 3188 * Syscall to delete a named extended attribute from a file or directory. 3189 * Accepts attribute name. The real work happens in VOP_SETEXTATTR(). 3190 */ 3191 int 3192 extattr_delete_file(struct extattr_delete_file_args *uap) 3193 { 3194 struct thread *td = curthread; 3195 struct proc *p = td->td_proc; 3196 struct nameidata nd; 3197 char attrname[EXTATTR_MAXNAMELEN]; 3198 int error; 3199 3200 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN); 3201 if (error) 3202 return(error); 3203 NDINIT(&nd, NAMEI_LOOKUP, CNP_FOLLOW | CNP_LOCKLEAF, UIO_USERSPACE, 3204 SCARG(uap, path), td); 3205 if ((error = namei(&nd)) != 0) 3206 return(error); 3207 error = VOP_SETEXTATTR(nd.ni_vp, attrname, NULL, p->p_ucred, td); 3208 NDFREE(&nd, 0); 3209 return(error); 3210 } 3211