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 1.48 2004/11/24 08:37:16 dillon Exp $ 41 */ 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/buf.h> 46 #include <sys/conf.h> 47 #include <sys/sysent.h> 48 #include <sys/malloc.h> 49 #include <sys/mount.h> 50 #include <sys/sysproto.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/fcntl.h> 54 #include <sys/file.h> 55 #include <sys/linker.h> 56 #include <sys/stat.h> 57 #include <sys/unistd.h> 58 #include <sys/vnode.h> 59 #include <sys/proc.h> 60 #include <sys/namei.h> 61 #include <sys/nlookup.h> 62 #include <sys/dirent.h> 63 #include <sys/extattr.h> 64 #include <sys/kern_syscall.h> 65 66 #include <machine/limits.h> 67 #include <vfs/union/union.h> 68 #include <sys/sysctl.h> 69 #include <vm/vm.h> 70 #include <vm/vm_object.h> 71 #include <vm/vm_zone.h> 72 #include <vm/vm_page.h> 73 74 #include <sys/file2.h> 75 76 static int checkvp_chdir (struct vnode *vn, struct thread *td); 77 static void checkdirs (struct vnode *olddp, struct namecache *ncp); 78 static int chroot_refuse_vdir_fds (struct filedesc *fdp); 79 static int getutimes (const struct timeval *, struct timespec *); 80 static int setfown (struct vnode *, uid_t, gid_t); 81 static int setfmode (struct vnode *, int); 82 static int setfflags (struct vnode *, int); 83 static int setutimes (struct vnode *, const struct timespec *, int); 84 static int usermount = 0; /* if 1, non-root can mount fs. */ 85 86 int (*union_dircheckp) (struct thread *, struct vnode **, struct file *); 87 88 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, ""); 89 90 /* 91 * Virtual File System System Calls 92 */ 93 94 /* 95 * Mount a file system. 96 */ 97 /* 98 * mount_args(char *type, char *path, int flags, caddr_t data) 99 */ 100 /* ARGSUSED */ 101 int 102 mount(struct mount_args *uap) 103 { 104 struct thread *td = curthread; 105 struct proc *p = td->td_proc; 106 struct vnode *vp; 107 struct namecache *ncp; 108 struct mount *mp; 109 struct vfsconf *vfsp; 110 int error, flag = 0, flag2 = 0; 111 struct vattr va; 112 struct nlookupdata nd; 113 char fstypename[MFSNAMELEN]; 114 lwkt_tokref ilock; 115 struct nlcomponent nlc; 116 117 KKASSERT(p); 118 if (p->p_ucred->cr_prison != NULL) 119 return (EPERM); 120 if (usermount == 0 && (error = suser(td))) 121 return (error); 122 /* 123 * Do not allow NFS export by non-root users. 124 */ 125 if (SCARG(uap, flags) & MNT_EXPORTED) { 126 error = suser(td); 127 if (error) 128 return (error); 129 } 130 /* 131 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users 132 */ 133 if (suser(td)) 134 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 135 136 /* 137 * Lookup the requested path and extract the ncp and vnode. 138 */ 139 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 140 if (error == 0) { 141 if ((error = nlookup(&nd)) == 0) { 142 if (nd.nl_ncp->nc_vp == NULL) 143 error = ENOENT; 144 } 145 } 146 if (error) { 147 nlookup_done(&nd); 148 return (error); 149 } 150 151 /* 152 * Extract the locked+refd ncp and cleanup the nd structure 153 */ 154 ncp = nd.nl_ncp; 155 nd.nl_ncp = NULL; 156 nlookup_done(&nd); 157 158 /* 159 * now we have the locked ref'd ncp and unreferenced vnode. 160 */ 161 vp = ncp->nc_vp; 162 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0) { 163 cache_put(ncp); 164 return (error); 165 } 166 cache_unlock(ncp); 167 168 /* 169 * Now we have an unlocked ref'd ncp and a locked ref'd vp 170 */ 171 if (SCARG(uap, flags) & MNT_UPDATE) { 172 if ((vp->v_flag & VROOT) == 0) { 173 cache_drop(ncp); 174 vput(vp); 175 return (EINVAL); 176 } 177 mp = vp->v_mount; 178 flag = mp->mnt_flag; 179 flag2 = mp->mnt_kern_flag; 180 /* 181 * We only allow the filesystem to be reloaded if it 182 * is currently mounted read-only. 183 */ 184 if ((SCARG(uap, flags) & MNT_RELOAD) && 185 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 186 cache_drop(ncp); 187 vput(vp); 188 return (EOPNOTSUPP); /* Needs translation */ 189 } 190 /* 191 * Only root, or the user that did the original mount is 192 * permitted to update it. 193 */ 194 if (mp->mnt_stat.f_owner != p->p_ucred->cr_uid && 195 (error = suser(td))) { 196 cache_drop(ncp); 197 vput(vp); 198 return (error); 199 } 200 if (vfs_busy(mp, LK_NOWAIT, NULL, td)) { 201 cache_drop(ncp); 202 vput(vp); 203 return (EBUSY); 204 } 205 if ((vp->v_flag & VMOUNT) != 0 || 206 vp->v_mountedhere != NULL) { 207 cache_drop(ncp); 208 vfs_unbusy(mp, td); 209 vput(vp); 210 return (EBUSY); 211 } 212 vp->v_flag |= VMOUNT; 213 mp->mnt_flag |= 214 SCARG(uap, flags) & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 215 VOP_UNLOCK(vp, 0, td); 216 goto update; 217 } 218 /* 219 * If the user is not root, ensure that they own the directory 220 * onto which we are attempting to mount. 221 */ 222 if ((error = VOP_GETATTR(vp, &va, td)) || 223 (va.va_uid != p->p_ucred->cr_uid && 224 (error = suser(td)))) { 225 cache_drop(ncp); 226 vput(vp); 227 return (error); 228 } 229 if ((error = vinvalbuf(vp, V_SAVE, td, 0, 0)) != 0) { 230 cache_drop(ncp); 231 vput(vp); 232 return (error); 233 } 234 if (vp->v_type != VDIR) { 235 cache_drop(ncp); 236 vput(vp); 237 return (ENOTDIR); 238 } 239 if ((error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL)) != 0) { 240 cache_drop(ncp); 241 vput(vp); 242 return (error); 243 } 244 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { 245 if (!strcmp(vfsp->vfc_name, fstypename)) 246 break; 247 } 248 if (vfsp == NULL) { 249 linker_file_t lf; 250 251 /* Only load modules for root (very important!) */ 252 if ((error = suser(td)) != 0) { 253 cache_drop(ncp); 254 vput(vp); 255 return error; 256 } 257 error = linker_load_file(fstypename, &lf); 258 if (error || lf == NULL) { 259 cache_drop(ncp); 260 vput(vp); 261 if (lf == NULL) 262 error = ENODEV; 263 return error; 264 } 265 lf->userrefs++; 266 /* lookup again, see if the VFS was loaded */ 267 for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) { 268 if (!strcmp(vfsp->vfc_name, fstypename)) 269 break; 270 } 271 if (vfsp == NULL) { 272 lf->userrefs--; 273 linker_file_unload(lf); 274 cache_drop(ncp); 275 vput(vp); 276 return (ENODEV); 277 } 278 } 279 if ((vp->v_flag & VMOUNT) != 0 || 280 vp->v_mountedhere != NULL) { 281 cache_drop(ncp); 282 vput(vp); 283 return (EBUSY); 284 } 285 vp->v_flag |= VMOUNT; 286 287 /* 288 * Allocate and initialize the filesystem. 289 */ 290 mp = malloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK); 291 TAILQ_INIT(&mp->mnt_nvnodelist); 292 TAILQ_INIT(&mp->mnt_reservedvnlist); 293 mp->mnt_nvnodelistsize = 0; 294 lockinit(&mp->mnt_lock, 0, "vfslock", 0, LK_NOPAUSE); 295 vfs_busy(mp, LK_NOWAIT, NULL, td); 296 mp->mnt_op = vfsp->vfc_vfsops; 297 mp->mnt_vfc = vfsp; 298 vfsp->vfc_refcount++; 299 mp->mnt_stat.f_type = vfsp->vfc_typenum; 300 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK; 301 strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN); 302 mp->mnt_vnodecovered = vp; 303 mp->mnt_stat.f_owner = p->p_ucred->cr_uid; 304 mp->mnt_iosize_max = DFLTPHYS; 305 VOP_UNLOCK(vp, 0, td); 306 update: 307 /* 308 * Set the mount level flags. 309 */ 310 if (SCARG(uap, flags) & MNT_RDONLY) 311 mp->mnt_flag |= MNT_RDONLY; 312 else if (mp->mnt_flag & MNT_RDONLY) 313 mp->mnt_kern_flag |= MNTK_WANTRDWR; 314 mp->mnt_flag &=~ (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 315 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOATIME | 316 MNT_NOSYMFOLLOW | MNT_IGNORE | 317 MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR); 318 mp->mnt_flag |= SCARG(uap, flags) & (MNT_NOSUID | MNT_NOEXEC | 319 MNT_NODEV | MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_FORCE | 320 MNT_NOSYMFOLLOW | MNT_IGNORE | 321 MNT_NOATIME | MNT_NOCLUSTERR | MNT_NOCLUSTERW | MNT_SUIDDIR); 322 /* 323 * Mount the filesystem. 324 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they 325 * get. 326 */ 327 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), td); 328 if (mp->mnt_flag & MNT_UPDATE) { 329 if (mp->mnt_kern_flag & MNTK_WANTRDWR) 330 mp->mnt_flag &= ~MNT_RDONLY; 331 mp->mnt_flag &=~ (MNT_UPDATE | MNT_RELOAD | MNT_FORCE); 332 mp->mnt_kern_flag &=~ MNTK_WANTRDWR; 333 if (error) { 334 mp->mnt_flag = flag; 335 mp->mnt_kern_flag = flag2; 336 } 337 vfs_unbusy(mp, td); 338 vp->v_flag &= ~VMOUNT; 339 vrele(vp); 340 cache_drop(ncp); 341 return (error); 342 } 343 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 344 /* 345 * Put the new filesystem on the mount list after root. The mount 346 * point gets its own mnt_ncp which is a special ncp linking the 347 * vnode-under to the root of the new mount. The lookup code 348 * detects the mount point going forward and detects the special 349 * mnt_ncp via NCP_MOUNTPT going backwards. 350 * 351 * It is not necessary to invalidate or purge the vnode underneath 352 * because elements under the mount will be given their own glue 353 * namecache record. 354 */ 355 if (!error) { 356 nlc.nlc_nameptr = ""; 357 nlc.nlc_namelen = 0; 358 mp->mnt_ncp = cache_nlookup(ncp, &nlc); 359 mp->mnt_ncp->nc_flag |= NCF_MOUNTPT; 360 mp->mnt_ncp->nc_mount = mp; 361 cache_drop(ncp); 362 /* XXX get the root of the fs and cache_setvp(mnt_ncp...) */ 363 vp->v_flag &= ~VMOUNT; 364 vp->v_mountedhere = mp; 365 lwkt_gettoken(&ilock, &mountlist_token); 366 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); 367 lwkt_reltoken(&ilock); 368 checkdirs(vp, mp->mnt_ncp); 369 cache_unlock(mp->mnt_ncp); /* leave ref intact */ 370 VOP_UNLOCK(vp, 0, td); 371 error = vfs_allocate_syncvnode(mp); 372 vfs_unbusy(mp, td); 373 if ((error = VFS_START(mp, 0, td)) != 0) 374 vrele(vp); 375 } else { 376 vfs_rm_vnodeops(&mp->mnt_vn_ops); 377 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops); 378 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops); 379 vp->v_flag &= ~VMOUNT; 380 mp->mnt_vfc->vfc_refcount--; 381 vfs_unbusy(mp, td); 382 free(mp, M_MOUNT); 383 cache_drop(ncp); 384 vput(vp); 385 } 386 return (error); 387 } 388 389 /* 390 * Scan all active processes to see if any of them have a current 391 * or root directory onto which the new filesystem has just been 392 * mounted. If so, replace them with the new mount point. 393 * 394 * The passed ncp is ref'd and locked (from the mount code) and 395 * must be associated with the vnode representing the root of the 396 * mount point. 397 */ 398 static void 399 checkdirs(struct vnode *olddp, struct namecache *ncp) 400 { 401 struct filedesc *fdp; 402 struct vnode *newdp; 403 struct mount *mp; 404 struct proc *p; 405 406 if (olddp->v_usecount == 1) 407 return; 408 mp = olddp->v_mountedhere; 409 if (VFS_ROOT(mp, &newdp)) 410 panic("mount: lost mount"); 411 cache_setvp(ncp, newdp); 412 413 if (rootvnode == olddp) { 414 vref(newdp); 415 vfs_cache_setroot(newdp, cache_hold(ncp)); 416 } 417 418 FOREACH_PROC_IN_SYSTEM(p) { 419 fdp = p->p_fd; 420 if (fdp->fd_cdir == olddp) { 421 vrele(fdp->fd_cdir); 422 vref(newdp); 423 fdp->fd_cdir = newdp; 424 cache_drop(fdp->fd_ncdir); 425 fdp->fd_ncdir = cache_hold(ncp); 426 } 427 if (fdp->fd_rdir == olddp) { 428 vrele(fdp->fd_rdir); 429 vref(newdp); 430 fdp->fd_rdir = newdp; 431 cache_drop(fdp->fd_nrdir); 432 fdp->fd_nrdir = cache_hold(ncp); 433 } 434 } 435 vput(newdp); 436 } 437 438 /* 439 * Unmount a file system. 440 * 441 * Note: unmount takes a path to the vnode mounted on as argument, 442 * not special file (as before). 443 */ 444 /* 445 * umount_args(char *path, int flags) 446 */ 447 /* ARGSUSED */ 448 int 449 unmount(struct unmount_args *uap) 450 { 451 struct thread *td = curthread; 452 struct proc *p = td->td_proc; 453 struct vnode *vp; 454 struct mount *mp; 455 int error; 456 struct nlookupdata nd; 457 458 KKASSERT(p); 459 if (p->p_ucred->cr_prison != NULL) 460 return (EPERM); 461 if (usermount == 0 && (error = suser(td))) 462 return (error); 463 464 vp = NULL; 465 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW); 466 if (error == 0) 467 error = nlookup(&nd); 468 if (error == 0) 469 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 470 nlookup_done(&nd); 471 if (error) 472 return (error); 473 474 mp = vp->v_mount; 475 476 /* 477 * Only root, or the user that did the original mount is 478 * permitted to unmount this filesystem. 479 */ 480 if ((mp->mnt_stat.f_owner != p->p_ucred->cr_uid) && 481 (error = suser(td))) { 482 vput(vp); 483 return (error); 484 } 485 486 /* 487 * Don't allow unmounting the root file system. 488 */ 489 if (mp->mnt_flag & MNT_ROOTFS) { 490 vput(vp); 491 return (EINVAL); 492 } 493 494 /* 495 * Must be the root of the filesystem 496 */ 497 if ((vp->v_flag & VROOT) == 0) { 498 vput(vp); 499 return (EINVAL); 500 } 501 vput(vp); 502 return (dounmount(mp, SCARG(uap, flags), td)); 503 } 504 505 /* 506 * Do the actual file system unmount. 507 */ 508 int 509 dounmount(struct mount *mp, int flags, struct thread *td) 510 { 511 struct vnode *coveredvp; 512 int error; 513 int async_flag; 514 lwkt_tokref ilock; 515 516 lwkt_gettoken(&ilock, &mountlist_token); 517 if (mp->mnt_kern_flag & MNTK_UNMOUNT) { 518 lwkt_reltoken(&ilock); 519 return (EBUSY); 520 } 521 mp->mnt_kern_flag |= MNTK_UNMOUNT; 522 /* Allow filesystems to detect that a forced unmount is in progress. */ 523 if (flags & MNT_FORCE) 524 mp->mnt_kern_flag |= MNTK_UNMOUNTF; 525 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK | 526 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &ilock, td); 527 if (error) { 528 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 529 if (mp->mnt_kern_flag & MNTK_MWAIT) 530 wakeup(mp); 531 return (error); 532 } 533 534 if (mp->mnt_flag & MNT_EXPUBLIC) 535 vfs_setpublicfs(NULL, NULL, NULL); 536 537 vfs_msync(mp, MNT_WAIT); 538 async_flag = mp->mnt_flag & MNT_ASYNC; 539 mp->mnt_flag &=~ MNT_ASYNC; 540 cache_purgevfs(mp); /* remove cache entries for this file sys */ 541 if (mp->mnt_syncer != NULL) 542 vrele(mp->mnt_syncer); 543 if (((mp->mnt_flag & MNT_RDONLY) || 544 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) || 545 (flags & MNT_FORCE)) 546 error = VFS_UNMOUNT(mp, flags, td); 547 lwkt_gettokref(&ilock); 548 if (error) { 549 if (mp->mnt_syncer == NULL) 550 vfs_allocate_syncvnode(mp); 551 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF); 552 mp->mnt_flag |= async_flag; 553 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 554 &ilock, td); 555 if (mp->mnt_kern_flag & MNTK_MWAIT) 556 wakeup(mp); 557 return (error); 558 } 559 TAILQ_REMOVE(&mountlist, mp, mnt_list); 560 561 /* 562 * Remove any installed vnode ops here so the individual VFSs don't 563 * have to. 564 */ 565 vfs_rm_vnodeops(&mp->mnt_vn_ops); 566 vfs_rm_vnodeops(&mp->mnt_vn_spec_ops); 567 vfs_rm_vnodeops(&mp->mnt_vn_fifo_ops); 568 569 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 570 coveredvp->v_mountedhere = NULL; 571 vrele(coveredvp); 572 cache_drop(mp->mnt_ncp); 573 mp->mnt_ncp = NULL; 574 } 575 mp->mnt_vfc->vfc_refcount--; 576 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) 577 panic("unmount: dangling vnode"); 578 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &ilock, td); 579 if (mp->mnt_kern_flag & MNTK_MWAIT) 580 wakeup(mp); 581 free(mp, M_MOUNT); 582 return (0); 583 } 584 585 /* 586 * Sync each mounted filesystem. 587 */ 588 589 #ifdef DEBUG 590 static int syncprt = 0; 591 SYSCTL_INT(_debug, OID_AUTO, syncprt, CTLFLAG_RW, &syncprt, 0, ""); 592 #endif /* DEBUG */ 593 594 /* ARGSUSED */ 595 int 596 sync(struct sync_args *uap) 597 { 598 struct thread *td = curthread; 599 struct mount *mp, *nmp; 600 lwkt_tokref ilock; 601 int asyncflag; 602 603 lwkt_gettoken(&ilock, &mountlist_token); 604 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 605 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) { 606 nmp = TAILQ_NEXT(mp, mnt_list); 607 continue; 608 } 609 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 610 asyncflag = mp->mnt_flag & MNT_ASYNC; 611 mp->mnt_flag &= ~MNT_ASYNC; 612 vfs_msync(mp, MNT_NOWAIT); 613 VFS_SYNC(mp, MNT_NOWAIT, td); 614 mp->mnt_flag |= asyncflag; 615 } 616 lwkt_gettokref(&ilock); 617 nmp = TAILQ_NEXT(mp, mnt_list); 618 vfs_unbusy(mp, td); 619 } 620 lwkt_reltoken(&ilock); 621 /* 622 * print out buffer pool stat information on each sync() call. 623 */ 624 #ifdef DEBUG 625 if (syncprt) 626 vfs_bufstats(); 627 #endif /* DEBUG */ 628 return (0); 629 } 630 631 /* XXX PRISON: could be per prison flag */ 632 static int prison_quotas; 633 #if 0 634 SYSCTL_INT(_kern_prison, OID_AUTO, quotas, CTLFLAG_RW, &prison_quotas, 0, ""); 635 #endif 636 637 /* 638 * quotactl_args(char *path, int fcmd, int uid, caddr_t arg) 639 * 640 * Change filesystem quotas. 641 */ 642 /* ARGSUSED */ 643 int 644 quotactl(struct quotactl_args *uap) 645 { 646 struct nlookupdata nd; 647 struct thread *td; 648 struct proc *p; 649 struct mount *mp; 650 int error; 651 652 td = curthread; 653 p = td->td_proc; 654 if (p->p_ucred->cr_prison && !prison_quotas) 655 return (EPERM); 656 657 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW); 658 if (error == 0) 659 error = nlookup(&nd); 660 if (error == 0) { 661 mp = nd.nl_ncp->nc_mount; 662 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 663 SCARG(uap, arg), nd.nl_td); 664 } 665 nlookup_done(&nd); 666 return (error); 667 } 668 669 int 670 kern_statfs(struct nlookupdata *nd, struct statfs *buf) 671 { 672 struct thread *td = curthread; 673 struct mount *mp; 674 struct statfs *sp; 675 int error; 676 677 if ((error = nlookup(nd)) != 0) 678 return (error); 679 mp = nd->nl_ncp->nc_mount; 680 sp = &mp->mnt_stat; 681 if ((error = VFS_STATFS(mp, sp, td)) != 0) 682 return (error); 683 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 684 bcopy(sp, buf, sizeof(*buf)); 685 /* Only root should have access to the fsid's. */ 686 if (suser(td)) 687 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 688 return (0); 689 } 690 691 /* 692 * statfs_args(char *path, struct statfs *buf) 693 * 694 * Get filesystem statistics. 695 */ 696 int 697 statfs(struct statfs_args *uap) 698 { 699 struct nlookupdata nd; 700 struct statfs buf; 701 int error; 702 703 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 704 if (error == 0) 705 error = kern_statfs(&nd, &buf); 706 nlookup_done(&nd); 707 if (error == 0) 708 error = copyout(&buf, uap->buf, sizeof(*uap->buf)); 709 return (error); 710 } 711 712 int 713 kern_fstatfs(int fd, struct statfs *buf) 714 { 715 struct thread *td = curthread; 716 struct proc *p = td->td_proc; 717 struct file *fp; 718 struct mount *mp; 719 struct statfs *sp; 720 int error; 721 722 KKASSERT(p); 723 error = getvnode(p->p_fd, fd, &fp); 724 if (error) 725 return (error); 726 mp = ((struct vnode *)fp->f_data)->v_mount; 727 if (mp == NULL) 728 return (EBADF); 729 sp = &mp->mnt_stat; 730 error = VFS_STATFS(mp, sp, td); 731 if (error) 732 return (error); 733 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 734 bcopy(sp, buf, sizeof(*buf)); 735 /* Only root should have access to the fsid's. */ 736 if (suser(td)) 737 buf->f_fsid.val[0] = buf->f_fsid.val[1] = 0; 738 return (0); 739 } 740 741 /* 742 * fstatfs_args(int fd, struct statfs *buf) 743 * 744 * Get filesystem statistics. 745 */ 746 int 747 fstatfs(struct fstatfs_args *uap) 748 { 749 struct statfs buf; 750 int error; 751 752 error = kern_fstatfs(uap->fd, &buf); 753 754 if (error == 0) 755 error = copyout(&buf, uap->buf, sizeof(*uap->buf)); 756 return (error); 757 } 758 759 /* 760 * getfsstat_args(struct statfs *buf, long bufsize, int flags) 761 * 762 * Get statistics on all filesystems. 763 */ 764 /* ARGSUSED */ 765 int 766 getfsstat(struct getfsstat_args *uap) 767 { 768 struct thread *td = curthread; 769 struct mount *mp, *nmp; 770 struct statfs *sp; 771 caddr_t sfsp; 772 lwkt_tokref ilock; 773 long count, maxcount, error; 774 775 maxcount = SCARG(uap, bufsize) / sizeof(struct statfs); 776 sfsp = (caddr_t)SCARG(uap, buf); 777 count = 0; 778 lwkt_gettoken(&ilock, &mountlist_token); 779 for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) { 780 if (vfs_busy(mp, LK_NOWAIT, &ilock, td)) { 781 nmp = TAILQ_NEXT(mp, mnt_list); 782 continue; 783 } 784 if (sfsp && count < maxcount) { 785 sp = &mp->mnt_stat; 786 /* 787 * If MNT_NOWAIT or MNT_LAZY is specified, do not 788 * refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY 789 * overrides MNT_WAIT. 790 */ 791 if (((SCARG(uap, flags) & (MNT_LAZY|MNT_NOWAIT)) == 0 || 792 (SCARG(uap, flags) & MNT_WAIT)) && 793 (error = VFS_STATFS(mp, sp, td))) { 794 lwkt_gettokref(&ilock); 795 nmp = TAILQ_NEXT(mp, mnt_list); 796 vfs_unbusy(mp, td); 797 continue; 798 } 799 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 800 error = copyout(sp, sfsp, sizeof(*sp)); 801 if (error) { 802 vfs_unbusy(mp, td); 803 return (error); 804 } 805 sfsp += sizeof(*sp); 806 } 807 count++; 808 lwkt_gettokref(&ilock); 809 nmp = TAILQ_NEXT(mp, mnt_list); 810 vfs_unbusy(mp, td); 811 } 812 lwkt_reltoken(&ilock); 813 if (sfsp && count > maxcount) 814 uap->sysmsg_result = maxcount; 815 else 816 uap->sysmsg_result = count; 817 return (0); 818 } 819 820 /* 821 * fchdir_args(int fd) 822 * 823 * Change current working directory to a given file descriptor. 824 */ 825 /* ARGSUSED */ 826 int 827 fchdir(struct fchdir_args *uap) 828 { 829 struct thread *td = curthread; 830 struct proc *p = td->td_proc; 831 struct filedesc *fdp = p->p_fd; 832 struct vnode *vp, *ovp; 833 struct mount *mp; 834 struct file *fp; 835 struct namecache *ncp, *oncp; 836 struct namecache *nct; 837 int error; 838 839 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 840 return (error); 841 vp = (struct vnode *)fp->f_data; 842 vref(vp); 843 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 844 if (vp->v_type != VDIR || fp->f_ncp == NULL) 845 error = ENOTDIR; 846 else 847 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td); 848 if (error) { 849 vput(vp); 850 return (error); 851 } 852 ncp = cache_hold(fp->f_ncp); 853 while (!error && (mp = vp->v_mountedhere) != NULL) { 854 error = nlookup_mp(mp, &nct); 855 if (error == 0) { 856 cache_unlock(nct); /* leave ref intact */ 857 vput(vp); 858 vp = nct->nc_vp; 859 error = vget(vp, LK_SHARED, td); 860 KKASSERT(error == 0); 861 cache_drop(ncp); 862 ncp = nct; 863 } 864 } 865 if (error == 0) { 866 ovp = fdp->fd_cdir; 867 oncp = fdp->fd_ncdir; 868 VOP_UNLOCK(vp, 0, td); /* leave ref intact */ 869 fdp->fd_cdir = vp; 870 fdp->fd_ncdir = ncp; 871 cache_drop(oncp); 872 vrele(ovp); 873 } else { 874 cache_drop(ncp); 875 vput(vp); 876 } 877 return (error); 878 } 879 880 int 881 kern_chdir(struct nlookupdata *nd) 882 { 883 struct thread *td = curthread; 884 struct proc *p = td->td_proc; 885 struct filedesc *fdp = p->p_fd; 886 struct vnode *vp, *ovp; 887 struct namecache *oncp; 888 int error; 889 890 if ((error = nlookup(nd)) != 0) 891 return (error); 892 if ((vp = nd->nl_ncp->nc_vp) == NULL) 893 return (ENOENT); 894 if ((error = vget(vp, LK_SHARED, td)) != 0) 895 return (error); 896 897 error = checkvp_chdir(vp, td); 898 VOP_UNLOCK(vp, 0, td); 899 if (error == 0) { 900 ovp = fdp->fd_cdir; 901 oncp = fdp->fd_ncdir; 902 cache_unlock(nd->nl_ncp); /* leave reference intact */ 903 fdp->fd_ncdir = nd->nl_ncp; 904 fdp->fd_cdir = vp; 905 cache_drop(oncp); 906 vrele(ovp); 907 nd->nl_ncp = NULL; 908 } else { 909 vrele(vp); 910 } 911 return (error); 912 } 913 914 /* 915 * chdir_args(char *path) 916 * 917 * Change current working directory (``.''). 918 */ 919 int 920 chdir(struct chdir_args *uap) 921 { 922 struct nlookupdata nd; 923 int error; 924 925 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 926 if (error == 0) 927 error = kern_chdir(&nd); 928 nlookup_done(&nd); 929 return (error); 930 } 931 932 /* 933 * Helper function for raised chroot(2) security function: Refuse if 934 * any filedescriptors are open directories. 935 */ 936 static int 937 chroot_refuse_vdir_fds(fdp) 938 struct filedesc *fdp; 939 { 940 struct vnode *vp; 941 struct file *fp; 942 int error; 943 int fd; 944 945 for (fd = 0; fd < fdp->fd_nfiles ; fd++) { 946 error = getvnode(fdp, fd, &fp); 947 if (error) 948 continue; 949 vp = (struct vnode *)fp->f_data; 950 if (vp->v_type != VDIR) 951 continue; 952 return(EPERM); 953 } 954 return (0); 955 } 956 957 /* 958 * This sysctl determines if we will allow a process to chroot(2) if it 959 * has a directory open: 960 * 0: disallowed for all processes. 961 * 1: allowed for processes that were not already chroot(2)'ed. 962 * 2: allowed for all processes. 963 */ 964 965 static int chroot_allow_open_directories = 1; 966 967 SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW, 968 &chroot_allow_open_directories, 0, ""); 969 970 /* 971 * chroot to the specified namecache entry. We obtain the vp from the 972 * namecache data. The passed ncp must be locked and referenced and will 973 * remain locked and referenced on return. 974 */ 975 static int 976 kern_chroot(struct nlookupdata *nd) 977 { 978 struct thread *td = curthread; 979 struct proc *p = td->td_proc; 980 struct filedesc *fdp = p->p_fd; 981 struct namecache *ncp; 982 struct vnode *vp; 983 int error; 984 985 /* 986 * Only root can chroot 987 */ 988 if ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0) 989 return (error); 990 991 if ((error = nlookup(nd)) != 0) 992 return (error); 993 ncp = nd->nl_ncp; 994 995 /* 996 * Disallow open directory descriptors (fchdir() breakouts). 997 */ 998 if (chroot_allow_open_directories == 0 || 999 (chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) { 1000 if ((error = chroot_refuse_vdir_fds(fdp)) != 0) 1001 return (error); 1002 } 1003 if ((vp = ncp->nc_vp) == NULL) 1004 return (ENOENT); 1005 1006 if ((error = vget(vp, LK_SHARED, td)) != 0) 1007 return (error); 1008 1009 /* 1010 * Check the validity of vp as a directory to change to and 1011 * associate it with rdir/jdir. 1012 */ 1013 error = checkvp_chdir(vp, td); 1014 VOP_UNLOCK(vp, 0, td); /* leave reference intact */ 1015 if (error == 0) { 1016 vrele(fdp->fd_rdir); 1017 fdp->fd_rdir = vp; /* reference inherited by fd_rdir */ 1018 cache_drop(fdp->fd_nrdir); 1019 fdp->fd_nrdir = cache_hold(ncp); 1020 if (fdp->fd_jdir == NULL) { 1021 fdp->fd_jdir = vp; 1022 vref(fdp->fd_jdir); 1023 fdp->fd_njdir = cache_hold(ncp); 1024 } 1025 } else { 1026 vrele(vp); 1027 } 1028 return (error); 1029 } 1030 1031 /* 1032 * chroot_args(char *path) 1033 * 1034 * Change notion of root (``/'') directory. 1035 */ 1036 /* ARGSUSED */ 1037 int 1038 chroot(struct chroot_args *uap) 1039 { 1040 struct thread *td = curthread; 1041 struct nlookupdata nd; 1042 int error; 1043 1044 KKASSERT(td->td_proc); 1045 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 1046 if (error == 0) 1047 error = kern_chroot(&nd); 1048 nlookup_done(&nd); 1049 return (error); 1050 } 1051 1052 /* 1053 * Common routine for chroot and chdir. Given a locked, referenced vnode, 1054 * determine whether it is legal to chdir to the vnode. The vnode's state 1055 * is not changed by this call. 1056 */ 1057 int 1058 checkvp_chdir(struct vnode *vp, struct thread *td) 1059 { 1060 int error; 1061 1062 if (vp->v_type != VDIR) 1063 error = ENOTDIR; 1064 else 1065 error = VOP_ACCESS(vp, VEXEC, td->td_proc->p_ucred, td); 1066 return (error); 1067 } 1068 1069 int 1070 kern_open(struct nlookupdata *nd, int oflags, int mode, int *res) 1071 { 1072 struct thread *td = curthread; 1073 struct proc *p = td->td_proc; 1074 struct filedesc *fdp = p->p_fd; 1075 int cmode, flags; 1076 struct file *nfp; 1077 struct file *fp; 1078 struct vnode *vp; 1079 int type, indx, error; 1080 struct flock lf; 1081 1082 if ((oflags & O_ACCMODE) == O_ACCMODE) 1083 return (EINVAL); 1084 flags = FFLAGS(oflags); 1085 error = falloc(p, &nfp, NULL); 1086 if (error) 1087 return (error); 1088 fp = nfp; 1089 cmode = ((mode &~ fdp->fd_cmask) & ALLPERMS) &~ S_ISTXT; 1090 1091 /* 1092 * XXX p_dupfd is a real mess. It allows a device to return a 1093 * file descriptor to be duplicated rather then doing the open 1094 * itself. 1095 */ 1096 p->p_dupfd = -1; 1097 1098 /* 1099 * Call vn_open() to do the lookup and assign the vnode to the 1100 * file pointer. vn_open() does not change the ref count on fp 1101 * and the vnode, on success, will be inherited by the file pointer 1102 * and unlocked. 1103 */ 1104 nd->nl_flags |= NLC_LOCKVP; 1105 error = vn_open(nd, fp, flags, cmode); 1106 nlookup_done(nd); 1107 if (error) { 1108 /* 1109 * handle special fdopen() case. bleh. dupfdopen() is 1110 * responsible for dropping the old contents of ofiles[indx] 1111 * if it succeeds. 1112 * 1113 * Note that if fsetfd() succeeds it will add a ref to fp 1114 * which represents the fd_ofiles[] assignment. We must still 1115 * drop our reference. 1116 */ 1117 if ((error == ENODEV || error == ENXIO) && p->p_dupfd >= 0) { 1118 if (fsetfd(p, fp, &indx) == 0) { 1119 error = dupfdopen(fdp, indx, p->p_dupfd, flags, error); 1120 if (error == 0) { 1121 *res = indx; 1122 fdrop(fp, td); /* our ref */ 1123 return (0); 1124 } 1125 if (fdp->fd_ofiles[indx] == fp) { 1126 fdp->fd_ofiles[indx] = NULL; 1127 fdrop(fp, td); /* fd_ofiles[] ref */ 1128 } 1129 } 1130 } 1131 fdrop(fp, td); /* our ref */ 1132 if (error == ERESTART) 1133 error = EINTR; 1134 return (error); 1135 } 1136 1137 /* 1138 * ref the vnode for ourselves so it can't be ripped out from under 1139 * is. XXX need an ND flag to request that the vnode be returned 1140 * anyway. 1141 */ 1142 vp = (struct vnode *)fp->f_data; 1143 vref(vp); 1144 if ((error = fsetfd(p, fp, &indx)) != 0) { 1145 fdrop(fp, td); 1146 vrele(vp); 1147 return (error); 1148 } 1149 1150 /* 1151 * If no error occurs the vp will have been assigned to the file 1152 * pointer. 1153 */ 1154 p->p_dupfd = 0; 1155 1156 /* 1157 * There should be 2 references on the file, one from the descriptor 1158 * table, and one for us. 1159 * 1160 * Handle the case where someone closed the file (via its file 1161 * descriptor) while we were blocked. The end result should look 1162 * like opening the file succeeded but it was immediately closed. 1163 */ 1164 if (fp->f_count == 1) { 1165 KASSERT(fdp->fd_ofiles[indx] != fp, 1166 ("Open file descriptor lost all refs")); 1167 vrele(vp); 1168 fo_close(fp, td); 1169 fdrop(fp, td); 1170 *res = indx; 1171 return 0; 1172 } 1173 1174 if (flags & (O_EXLOCK | O_SHLOCK)) { 1175 lf.l_whence = SEEK_SET; 1176 lf.l_start = 0; 1177 lf.l_len = 0; 1178 if (flags & O_EXLOCK) 1179 lf.l_type = F_WRLCK; 1180 else 1181 lf.l_type = F_RDLCK; 1182 type = F_FLOCK; 1183 if ((flags & FNONBLOCK) == 0) 1184 type |= F_WAIT; 1185 1186 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) { 1187 /* 1188 * lock request failed. Normally close the descriptor 1189 * but handle the case where someone might have dup()d 1190 * it when we weren't looking. One reference is 1191 * owned by the descriptor array, the other by us. 1192 */ 1193 vrele(vp); 1194 if (fdp->fd_ofiles[indx] == fp) { 1195 fdp->fd_ofiles[indx] = NULL; 1196 fdrop(fp, td); 1197 } 1198 fdrop(fp, td); 1199 return (error); 1200 } 1201 fp->f_flag |= FHASLOCK; 1202 } 1203 /* assert that vn_open created a backing object if one is needed */ 1204 KASSERT(!vn_canvmio(vp) || VOP_GETVOBJECT(vp, NULL) == 0, 1205 ("open: vmio vnode has no backing object after vn_open")); 1206 1207 vrele(vp); 1208 1209 /* 1210 * release our private reference, leaving the one associated with the 1211 * descriptor table intact. 1212 */ 1213 fdrop(fp, td); 1214 *res = indx; 1215 return (0); 1216 } 1217 1218 /* 1219 * open_args(char *path, int flags, int mode) 1220 * 1221 * Check permissions, allocate an open file structure, 1222 * and call the device open routine if any. 1223 */ 1224 int 1225 open(struct open_args *uap) 1226 { 1227 struct nlookupdata nd; 1228 int error; 1229 1230 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 1231 if (error == 0) { 1232 error = kern_open(&nd, uap->flags, 1233 uap->mode, &uap->sysmsg_result); 1234 } 1235 nlookup_done(&nd); 1236 return (error); 1237 } 1238 1239 int 1240 kern_mknod(struct nlookupdata *nd, int mode, int dev) 1241 { 1242 struct namecache *ncp; 1243 struct thread *td = curthread; 1244 struct proc *p = td->td_proc; 1245 struct vnode *vp; 1246 struct vattr vattr; 1247 int error; 1248 int whiteout = 0; 1249 1250 KKASSERT(p); 1251 1252 switch (mode & S_IFMT) { 1253 case S_IFCHR: 1254 case S_IFBLK: 1255 error = suser(td); 1256 break; 1257 default: 1258 error = suser_cred(p->p_ucred, PRISON_ROOT); 1259 break; 1260 } 1261 if (error) 1262 return (error); 1263 1264 bwillwrite(); 1265 nd->nl_flags |= NLC_CREATE; 1266 if ((error = nlookup(nd)) != 0) 1267 return (error); 1268 ncp = nd->nl_ncp; 1269 if (ncp->nc_vp) 1270 return (EEXIST); 1271 1272 VATTR_NULL(&vattr); 1273 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask; 1274 vattr.va_rdev = dev; 1275 whiteout = 0; 1276 1277 switch (mode & S_IFMT) { 1278 case S_IFMT: /* used by badsect to flag bad sectors */ 1279 vattr.va_type = VBAD; 1280 break; 1281 case S_IFCHR: 1282 vattr.va_type = VCHR; 1283 break; 1284 case S_IFBLK: 1285 vattr.va_type = VBLK; 1286 break; 1287 case S_IFWHT: 1288 whiteout = 1; 1289 break; 1290 default: 1291 error = EINVAL; 1292 break; 1293 } 1294 if (error == 0) { 1295 if (whiteout) { 1296 error = VOP_NWHITEOUT(ncp, nd->nl_cred, NAMEI_CREATE); 1297 } else { 1298 vp = NULL; 1299 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr); 1300 if (error == 0) 1301 vput(vp); 1302 } 1303 } 1304 return (error); 1305 } 1306 1307 /* 1308 * mknod_args(char *path, int mode, int dev) 1309 * 1310 * Create a special file. 1311 */ 1312 int 1313 mknod(struct mknod_args *uap) 1314 { 1315 struct nlookupdata nd; 1316 int error; 1317 1318 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 1319 if (error == 0) 1320 error = kern_mknod(&nd, uap->mode, uap->dev); 1321 nlookup_done(&nd); 1322 return (error); 1323 } 1324 1325 int 1326 kern_mkfifo(struct nlookupdata *nd, int mode) 1327 { 1328 struct namecache *ncp; 1329 struct thread *td = curthread; 1330 struct proc *p = td->td_proc; 1331 struct vattr vattr; 1332 struct vnode *vp; 1333 int error; 1334 1335 bwillwrite(); 1336 1337 nd->nl_flags |= NLC_CREATE; 1338 if ((error = nlookup(nd)) != 0) 1339 return (error); 1340 ncp = nd->nl_ncp; 1341 if (ncp->nc_vp) 1342 return (EEXIST); 1343 1344 VATTR_NULL(&vattr); 1345 vattr.va_type = VFIFO; 1346 vattr.va_mode = (mode & ALLPERMS) &~ p->p_fd->fd_cmask; 1347 vp = NULL; 1348 error = VOP_NMKNOD(ncp, &vp, nd->nl_cred, &vattr); 1349 if (error == 0) 1350 vput(vp); 1351 return (error); 1352 } 1353 1354 /* 1355 * mkfifo_args(char *path, int mode) 1356 * 1357 * Create a named pipe. 1358 */ 1359 int 1360 mkfifo(struct mkfifo_args *uap) 1361 { 1362 struct nlookupdata nd; 1363 int error; 1364 1365 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 1366 if (error == 0) 1367 error = kern_mkfifo(&nd, uap->mode); 1368 nlookup_done(&nd); 1369 return (error); 1370 } 1371 1372 int 1373 kern_link(struct nlookupdata *nd, struct nlookupdata *linknd) 1374 { 1375 struct thread *td = curthread; 1376 struct vnode *vp; 1377 int error; 1378 1379 /* 1380 * Lookup the source and obtained a locked vnode. 1381 * 1382 * XXX relookup on vget failure / race ? 1383 */ 1384 bwillwrite(); 1385 if ((error = nlookup(nd)) != 0) 1386 return (error); 1387 vp = nd->nl_ncp->nc_vp; 1388 KKASSERT(vp != NULL); 1389 if (vp->v_type == VDIR) 1390 return (EPERM); /* POSIX */ 1391 if ((error = vget(vp, LK_EXCLUSIVE, td)) != 0) 1392 return (error); 1393 1394 /* 1395 * Unlock the source so we can lookup the target without deadlocking 1396 * (XXX vp is locked already, possible other deadlock?). The target 1397 * must not exist. 1398 */ 1399 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED); 1400 nd->nl_flags &= ~NLC_NCPISLOCKED; 1401 cache_unlock(nd->nl_ncp); 1402 1403 linknd->nl_flags |= NLC_CREATE; 1404 if ((error = nlookup(linknd)) != 0) { 1405 vput(vp); 1406 return (error); 1407 } 1408 if (linknd->nl_ncp->nc_vp) { 1409 vput(vp); 1410 return (EEXIST); 1411 } 1412 1413 /* 1414 * Finally run the new API VOP. 1415 */ 1416 error = VOP_NLINK(linknd->nl_ncp, vp, linknd->nl_cred); 1417 vput(vp); 1418 return (error); 1419 } 1420 1421 /* 1422 * link_args(char *path, char *link) 1423 * 1424 * Make a hard file link. 1425 */ 1426 int 1427 link(struct link_args *uap) 1428 { 1429 struct nlookupdata nd, linknd; 1430 int error; 1431 1432 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 1433 if (error == 0) { 1434 error = nlookup_init(&linknd, uap->link, UIO_USERSPACE, 0); 1435 if (error == 0) 1436 error = kern_link(&nd, &linknd); 1437 nlookup_done(&linknd); 1438 } 1439 nlookup_done(&nd); 1440 return (error); 1441 } 1442 1443 int 1444 kern_symlink(struct nlookupdata *nd, char *path, int mode) 1445 { 1446 struct namecache *ncp; 1447 struct vattr vattr; 1448 struct vnode *vp; 1449 int error; 1450 1451 bwillwrite(); 1452 nd->nl_flags |= NLC_CREATE; 1453 if ((error = nlookup(nd)) != 0) 1454 return (error); 1455 ncp = nd->nl_ncp; 1456 if (ncp->nc_vp) 1457 return (EEXIST); 1458 1459 VATTR_NULL(&vattr); 1460 vattr.va_mode = mode; 1461 error = VOP_NSYMLINK(ncp, &vp, nd->nl_cred, &vattr, path); 1462 if (error == 0) 1463 vput(vp); 1464 return (error); 1465 } 1466 1467 /* 1468 * symlink(char *path, char *link) 1469 * 1470 * Make a symbolic link. 1471 */ 1472 int 1473 symlink(struct symlink_args *uap) 1474 { 1475 struct thread *td = curthread; 1476 struct nlookupdata nd; 1477 char *path; 1478 int error; 1479 int mode; 1480 1481 path = zalloc(namei_zone); 1482 error = copyinstr(uap->path, path, MAXPATHLEN, NULL); 1483 if (error == 0) { 1484 error = nlookup_init(&nd, uap->link, UIO_USERSPACE, 0); 1485 if (error == 0) { 1486 mode = ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask; 1487 error = kern_symlink(&nd, path, mode); 1488 } 1489 nlookup_done(&nd); 1490 } 1491 zfree(namei_zone, path); 1492 return (error); 1493 } 1494 1495 /* 1496 * undelete_args(char *path) 1497 * 1498 * Delete a whiteout from the filesystem. 1499 */ 1500 /* ARGSUSED */ 1501 int 1502 undelete(struct undelete_args *uap) 1503 { 1504 struct nlookupdata nd; 1505 int error; 1506 1507 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0); 1508 bwillwrite(); 1509 nd.nl_flags |= NLC_DELETE; 1510 if (error == 0) 1511 error = nlookup(&nd); 1512 if (error == 0) 1513 error = VOP_NWHITEOUT(nd.nl_ncp, nd.nl_cred, NAMEI_DELETE); 1514 nlookup_done(&nd); 1515 return (error); 1516 } 1517 1518 int 1519 kern_unlink(struct nlookupdata *nd) 1520 { 1521 struct namecache *ncp; 1522 int error; 1523 1524 bwillwrite(); 1525 nd->nl_flags |= NLC_DELETE; 1526 if ((error = nlookup(nd)) != 0) 1527 return (error); 1528 ncp = nd->nl_ncp; 1529 error = VOP_NREMOVE(ncp, nd->nl_cred); 1530 return (error); 1531 } 1532 1533 /* 1534 * unlink_args(char *path) 1535 * 1536 * Delete a name from the filesystem. 1537 */ 1538 int 1539 unlink(struct unlink_args *uap) 1540 { 1541 struct nlookupdata nd; 1542 int error; 1543 1544 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 1545 if (error == 0) 1546 error = kern_unlink(&nd); 1547 nlookup_done(&nd); 1548 return (error); 1549 } 1550 1551 int 1552 kern_lseek(int fd, off_t offset, int whence, off_t *res) 1553 { 1554 struct thread *td = curthread; 1555 struct proc *p = td->td_proc; 1556 struct filedesc *fdp = p->p_fd; 1557 struct file *fp; 1558 struct vattr vattr; 1559 int error; 1560 1561 if (fd >= fdp->fd_nfiles || 1562 (fp = fdp->fd_ofiles[fd]) == NULL) 1563 return (EBADF); 1564 if (fp->f_type != DTYPE_VNODE) 1565 return (ESPIPE); 1566 switch (whence) { 1567 case L_INCR: 1568 fp->f_offset += offset; 1569 break; 1570 case L_XTND: 1571 error=VOP_GETATTR((struct vnode *)fp->f_data, &vattr, td); 1572 if (error) 1573 return (error); 1574 fp->f_offset = offset + vattr.va_size; 1575 break; 1576 case L_SET: 1577 fp->f_offset = offset; 1578 break; 1579 default: 1580 return (EINVAL); 1581 } 1582 *res = fp->f_offset; 1583 return (0); 1584 } 1585 1586 /* 1587 * lseek_args(int fd, int pad, off_t offset, int whence) 1588 * 1589 * Reposition read/write file offset. 1590 */ 1591 int 1592 lseek(struct lseek_args *uap) 1593 { 1594 int error; 1595 1596 error = kern_lseek(uap->fd, uap->offset, uap->whence, 1597 &uap->sysmsg_offset); 1598 1599 return (error); 1600 } 1601 1602 int 1603 kern_access(struct nlookupdata *nd, int aflags) 1604 { 1605 struct thread *td = curthread; 1606 struct vnode *vp; 1607 int error, flags; 1608 1609 if ((error = nlookup(nd)) != 0) 1610 return (error); 1611 retry: 1612 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp); 1613 if (error) 1614 return (error); 1615 1616 /* Flags == 0 means only check for existence. */ 1617 if (aflags) { 1618 flags = 0; 1619 if (aflags & R_OK) 1620 flags |= VREAD; 1621 if (aflags & W_OK) 1622 flags |= VWRITE; 1623 if (aflags & X_OK) 1624 flags |= VEXEC; 1625 if ((flags & VWRITE) == 0 || (error = vn_writechk(vp)) == 0) 1626 error = VOP_ACCESS(vp, flags, nd->nl_cred, td); 1627 1628 /* 1629 * If the file handle is stale we have to re-resolve the 1630 * entry. This is a hack at the moment. 1631 */ 1632 if (error == ESTALE) { 1633 cache_setunresolved(nd->nl_ncp); 1634 error = cache_resolve(nd->nl_ncp, nd->nl_cred); 1635 if (error == 0) { 1636 vput(vp); 1637 vp = NULL; 1638 goto retry; 1639 } 1640 } 1641 } 1642 vput(vp); 1643 return (error); 1644 } 1645 1646 /* 1647 * access_args(char *path, int flags) 1648 * 1649 * Check access permissions. 1650 */ 1651 int 1652 access(struct access_args *uap) 1653 { 1654 struct nlookupdata nd; 1655 int error; 1656 1657 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 1658 if (error == 0) 1659 error = kern_access(&nd, uap->flags); 1660 nlookup_done(&nd); 1661 return (error); 1662 } 1663 1664 int 1665 kern_stat(struct nlookupdata *nd, struct stat *st) 1666 { 1667 int error; 1668 struct vnode *vp; 1669 thread_t td; 1670 1671 if ((error = nlookup(nd)) != 0) 1672 return (error); 1673 again: 1674 if ((vp = nd->nl_ncp->nc_vp) == NULL) 1675 return (ENOENT); 1676 1677 td = curthread; 1678 if ((error = vget(vp, LK_SHARED, td)) != 0) 1679 return (error); 1680 error = vn_stat(vp, st, td); 1681 1682 /* 1683 * If the file handle is stale we have to re-resolve the entry. This 1684 * is a hack at the moment. 1685 */ 1686 if (error == ESTALE) { 1687 cache_setunresolved(nd->nl_ncp); 1688 error = cache_resolve(nd->nl_ncp, nd->nl_cred); 1689 if (error == 0) { 1690 vput(vp); 1691 goto again; 1692 } 1693 } 1694 vput(vp); 1695 return (error); 1696 } 1697 1698 /* 1699 * stat_args(char *path, struct stat *ub) 1700 * 1701 * Get file status; this version follows links. 1702 */ 1703 int 1704 stat(struct stat_args *uap) 1705 { 1706 struct nlookupdata nd; 1707 struct stat st; 1708 int error; 1709 1710 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 1711 if (error == 0) { 1712 error = kern_stat(&nd, &st); 1713 if (error == 0) 1714 error = copyout(&st, uap->ub, sizeof(*uap->ub)); 1715 } 1716 nlookup_done(&nd); 1717 return (error); 1718 } 1719 1720 /* 1721 * lstat_args(char *path, struct stat *ub) 1722 * 1723 * Get file status; this version does not follow links. 1724 */ 1725 int 1726 lstat(struct lstat_args *uap) 1727 { 1728 struct nlookupdata nd; 1729 struct stat st; 1730 int error; 1731 1732 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 1733 if (error == 0) { 1734 error = kern_stat(&nd, &st); 1735 if (error == 0) 1736 error = copyout(&st, uap->ub, sizeof(*uap->ub)); 1737 } 1738 nlookup_done(&nd); 1739 return (error); 1740 } 1741 1742 void 1743 cvtnstat(sb, nsb) 1744 struct stat *sb; 1745 struct nstat *nsb; 1746 { 1747 nsb->st_dev = sb->st_dev; 1748 nsb->st_ino = sb->st_ino; 1749 nsb->st_mode = sb->st_mode; 1750 nsb->st_nlink = sb->st_nlink; 1751 nsb->st_uid = sb->st_uid; 1752 nsb->st_gid = sb->st_gid; 1753 nsb->st_rdev = sb->st_rdev; 1754 nsb->st_atimespec = sb->st_atimespec; 1755 nsb->st_mtimespec = sb->st_mtimespec; 1756 nsb->st_ctimespec = sb->st_ctimespec; 1757 nsb->st_size = sb->st_size; 1758 nsb->st_blocks = sb->st_blocks; 1759 nsb->st_blksize = sb->st_blksize; 1760 nsb->st_flags = sb->st_flags; 1761 nsb->st_gen = sb->st_gen; 1762 nsb->st_qspare[0] = sb->st_qspare[0]; 1763 nsb->st_qspare[1] = sb->st_qspare[1]; 1764 } 1765 1766 /* 1767 * nstat_args(char *path, struct nstat *ub) 1768 */ 1769 /* ARGSUSED */ 1770 int 1771 nstat(struct nstat_args *uap) 1772 { 1773 struct thread *td = curthread; 1774 struct vnode *vp; 1775 struct stat sb; 1776 struct nstat nsb; 1777 struct nlookupdata nd; 1778 int error; 1779 1780 vp = NULL; 1781 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW); 1782 if (error == 0) 1783 error = nlookup(&nd); 1784 if (error == 0) 1785 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 1786 nlookup_done(&nd); 1787 if (error == 0) { 1788 error = vn_stat(vp, &sb, td); 1789 vput(vp); 1790 if (error == 0) { 1791 cvtnstat(&sb, &nsb); 1792 error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb)); 1793 } 1794 } 1795 return (error); 1796 } 1797 1798 /* 1799 * lstat_args(char *path, struct stat *ub) 1800 * 1801 * Get file status; this version does not follow links. 1802 */ 1803 /* ARGSUSED */ 1804 int 1805 nlstat(struct nlstat_args *uap) 1806 { 1807 struct thread *td = curthread; 1808 struct vnode *vp; 1809 struct stat sb; 1810 struct nstat nsb; 1811 struct nlookupdata nd; 1812 int error; 1813 1814 vp = NULL; 1815 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, 0); 1816 if (error == 0) 1817 error = nlookup(&nd); 1818 if (error == 0) 1819 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 1820 nlookup_done(&nd); 1821 if (error == 0) { 1822 error = vn_stat(vp, &sb, td); 1823 vput(vp); 1824 if (error == 0) { 1825 cvtnstat(&sb, &nsb); 1826 error = copyout(&nsb, SCARG(uap, ub), sizeof(nsb)); 1827 } 1828 } 1829 return (error); 1830 } 1831 1832 /* 1833 * pathconf_Args(char *path, int name) 1834 * 1835 * Get configurable pathname variables. 1836 */ 1837 /* ARGSUSED */ 1838 int 1839 pathconf(struct pathconf_args *uap) 1840 { 1841 struct nlookupdata nd; 1842 struct vnode *vp; 1843 int error; 1844 1845 vp = NULL; 1846 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW); 1847 if (error == 0) 1848 error = nlookup(&nd); 1849 if (error == 0) 1850 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 1851 nlookup_done(&nd); 1852 if (error == 0) { 1853 error = VOP_PATHCONF(vp, SCARG(uap, name), uap->sysmsg_fds); 1854 vput(vp); 1855 } 1856 return (error); 1857 } 1858 1859 /* 1860 * XXX: daver 1861 * kern_readlink isn't properly split yet. There is a copyin burried 1862 * in VOP_READLINK(). 1863 */ 1864 int 1865 kern_readlink(struct nlookupdata *nd, char *buf, int count, int *res) 1866 { 1867 struct thread *td = curthread; 1868 struct proc *p = td->td_proc; 1869 struct vnode *vp; 1870 struct iovec aiov; 1871 struct uio auio; 1872 int error; 1873 1874 if ((error = nlookup(nd)) != 0) 1875 return (error); 1876 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &vp); 1877 if (error) 1878 return (error); 1879 if (vp->v_type != VLNK) { 1880 error = EINVAL; 1881 } else { 1882 aiov.iov_base = buf; 1883 aiov.iov_len = count; 1884 auio.uio_iov = &aiov; 1885 auio.uio_iovcnt = 1; 1886 auio.uio_offset = 0; 1887 auio.uio_rw = UIO_READ; 1888 auio.uio_segflg = UIO_USERSPACE; 1889 auio.uio_td = td; 1890 auio.uio_resid = count; 1891 error = VOP_READLINK(vp, &auio, p->p_ucred); 1892 } 1893 vput(vp); 1894 *res = count - auio.uio_resid; 1895 return (error); 1896 } 1897 1898 /* 1899 * readlink_args(char *path, char *buf, int count) 1900 * 1901 * Return target name of a symbolic link. 1902 */ 1903 int 1904 readlink(struct readlink_args *uap) 1905 { 1906 struct nlookupdata nd; 1907 int error; 1908 1909 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 1910 if (error == 0) { 1911 error = kern_readlink(&nd, uap->buf, uap->count, 1912 &uap->sysmsg_result); 1913 } 1914 nlookup_done(&nd); 1915 return (error); 1916 } 1917 1918 static int 1919 setfflags(struct vnode *vp, int flags) 1920 { 1921 struct thread *td = curthread; 1922 struct proc *p = td->td_proc; 1923 int error; 1924 struct vattr vattr; 1925 1926 /* 1927 * Prevent non-root users from setting flags on devices. When 1928 * a device is reused, users can retain ownership of the device 1929 * if they are allowed to set flags and programs assume that 1930 * chown can't fail when done as root. 1931 */ 1932 if ((vp->v_type == VCHR || vp->v_type == VBLK) && 1933 ((error = suser_cred(p->p_ucred, PRISON_ROOT)) != 0)) 1934 return (error); 1935 1936 /* 1937 * note: vget is required for any operation that might mod the vnode 1938 * so VINACTIVE is properly cleared. 1939 */ 1940 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 1941 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) { 1942 VATTR_NULL(&vattr); 1943 vattr.va_flags = flags; 1944 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 1945 vput(vp); 1946 } 1947 return (error); 1948 } 1949 1950 /* 1951 * chflags(char *path, int flags) 1952 * 1953 * Change flags of a file given a path name. 1954 */ 1955 /* ARGSUSED */ 1956 int 1957 chflags(struct chflags_args *uap) 1958 { 1959 struct nlookupdata nd; 1960 struct vnode *vp; 1961 int error; 1962 1963 vp = NULL; 1964 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW); 1965 /* XXX Add NLC flag indicating modifying operation? */ 1966 if (error == 0) 1967 error = nlookup(&nd); 1968 if (error == 0) 1969 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp); 1970 nlookup_done(&nd); 1971 if (error == 0) { 1972 error = setfflags(vp, SCARG(uap, flags)); 1973 vrele(vp); 1974 } 1975 return (error); 1976 } 1977 1978 /* 1979 * fchflags_args(int fd, int flags) 1980 * 1981 * Change flags of a file given a file descriptor. 1982 */ 1983 /* ARGSUSED */ 1984 int 1985 fchflags(struct fchflags_args *uap) 1986 { 1987 struct thread *td = curthread; 1988 struct proc *p = td->td_proc; 1989 struct file *fp; 1990 int error; 1991 1992 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 1993 return (error); 1994 return setfflags((struct vnode *) fp->f_data, SCARG(uap, flags)); 1995 } 1996 1997 static int 1998 setfmode(struct vnode *vp, int mode) 1999 { 2000 struct thread *td = curthread; 2001 struct proc *p = td->td_proc; 2002 int error; 2003 struct vattr vattr; 2004 2005 /* 2006 * note: vget is required for any operation that might mod the vnode 2007 * so VINACTIVE is properly cleared. 2008 */ 2009 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2010 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) { 2011 VATTR_NULL(&vattr); 2012 vattr.va_mode = mode & ALLPERMS; 2013 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 2014 vput(vp); 2015 } 2016 return error; 2017 } 2018 2019 int 2020 kern_chmod(struct nlookupdata *nd, int mode) 2021 { 2022 struct vnode *vp; 2023 int error; 2024 2025 /* XXX Add NLC flag indicating modifying operation? */ 2026 if ((error = nlookup(nd)) != 0) 2027 return (error); 2028 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0) 2029 return (error); 2030 error = setfmode(vp, mode); 2031 vrele(vp); 2032 return (error); 2033 } 2034 2035 /* 2036 * chmod_args(char *path, int mode) 2037 * 2038 * Change mode of a file given path name. 2039 */ 2040 /* ARGSUSED */ 2041 int 2042 chmod(struct chmod_args *uap) 2043 { 2044 struct nlookupdata nd; 2045 int error; 2046 2047 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 2048 if (error == 0) 2049 error = kern_chmod(&nd, uap->mode); 2050 nlookup_done(&nd); 2051 return (error); 2052 } 2053 2054 /* 2055 * lchmod_args(char *path, int mode) 2056 * 2057 * Change mode of a file given path name (don't follow links.) 2058 */ 2059 /* ARGSUSED */ 2060 int 2061 lchmod(struct lchmod_args *uap) 2062 { 2063 struct nlookupdata nd; 2064 int error; 2065 2066 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 2067 if (error == 0) 2068 error = kern_chmod(&nd, uap->mode); 2069 nlookup_done(&nd); 2070 return (error); 2071 } 2072 2073 /* 2074 * fchmod_args(int fd, int mode) 2075 * 2076 * Change mode of a file given a file descriptor. 2077 */ 2078 /* ARGSUSED */ 2079 int 2080 fchmod(struct fchmod_args *uap) 2081 { 2082 struct thread *td = curthread; 2083 struct proc *p = td->td_proc; 2084 struct file *fp; 2085 int error; 2086 2087 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2088 return (error); 2089 return setfmode((struct vnode *)fp->f_data, SCARG(uap, mode)); 2090 } 2091 2092 static int 2093 setfown(struct vnode *vp, uid_t uid, gid_t gid) 2094 { 2095 struct thread *td = curthread; 2096 struct proc *p = td->td_proc; 2097 int error; 2098 struct vattr vattr; 2099 2100 /* 2101 * note: vget is required for any operation that might mod the vnode 2102 * so VINACTIVE is properly cleared. 2103 */ 2104 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2105 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) { 2106 VATTR_NULL(&vattr); 2107 vattr.va_uid = uid; 2108 vattr.va_gid = gid; 2109 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 2110 vput(vp); 2111 } 2112 return error; 2113 } 2114 2115 int 2116 kern_chown(struct nlookupdata *nd, int uid, int gid) 2117 { 2118 struct vnode *vp; 2119 int error; 2120 2121 /* XXX Add NLC flag indicating modifying operation? */ 2122 if ((error = nlookup(nd)) != 0) 2123 return (error); 2124 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0) 2125 return (error); 2126 error = setfown(vp, uid, gid); 2127 vrele(vp); 2128 return (error); 2129 } 2130 2131 /* 2132 * chown(char *path, int uid, int gid) 2133 * 2134 * Set ownership given a path name. 2135 */ 2136 int 2137 chown(struct chown_args *uap) 2138 { 2139 struct nlookupdata nd; 2140 int error; 2141 2142 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 2143 if (error == 0) 2144 error = kern_chown(&nd, uap->uid, uap->gid); 2145 nlookup_done(&nd); 2146 return (error); 2147 } 2148 2149 /* 2150 * lchown_args(char *path, int uid, int gid) 2151 * 2152 * Set ownership given a path name, do not cross symlinks. 2153 */ 2154 int 2155 lchown(struct lchown_args *uap) 2156 { 2157 struct nlookupdata nd; 2158 int error; 2159 2160 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 2161 if (error == 0) 2162 error = kern_chown(&nd, uap->uid, uap->gid); 2163 nlookup_done(&nd); 2164 return (error); 2165 } 2166 2167 /* 2168 * fchown_args(int fd, int uid, int gid) 2169 * 2170 * Set ownership given a file descriptor. 2171 */ 2172 /* ARGSUSED */ 2173 int 2174 fchown(struct fchown_args *uap) 2175 { 2176 struct thread *td = curthread; 2177 struct proc *p = td->td_proc; 2178 struct file *fp; 2179 int error; 2180 2181 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2182 return (error); 2183 return setfown((struct vnode *)fp->f_data, 2184 SCARG(uap, uid), SCARG(uap, gid)); 2185 } 2186 2187 static int 2188 getutimes(const struct timeval *tvp, struct timespec *tsp) 2189 { 2190 struct timeval tv[2]; 2191 2192 if (tvp == NULL) { 2193 microtime(&tv[0]); 2194 TIMEVAL_TO_TIMESPEC(&tv[0], &tsp[0]); 2195 tsp[1] = tsp[0]; 2196 } else { 2197 TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]); 2198 TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]); 2199 } 2200 return 0; 2201 } 2202 2203 static int 2204 setutimes(struct vnode *vp, const struct timespec *ts, int nullflag) 2205 { 2206 struct thread *td = curthread; 2207 struct proc *p = td->td_proc; 2208 int error; 2209 struct vattr vattr; 2210 2211 /* 2212 * note: vget is required for any operation that might mod the vnode 2213 * so VINACTIVE is properly cleared. 2214 */ 2215 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2216 if ((error = vget(vp, LK_EXCLUSIVE, td)) == 0) { 2217 VATTR_NULL(&vattr); 2218 vattr.va_atime = ts[0]; 2219 vattr.va_mtime = ts[1]; 2220 if (nullflag) 2221 vattr.va_vaflags |= VA_UTIMES_NULL; 2222 error = VOP_SETATTR(vp, &vattr, p->p_ucred, td); 2223 vput(vp); 2224 } 2225 return error; 2226 } 2227 2228 int 2229 kern_utimes(struct nlookupdata *nd, struct timeval *tptr) 2230 { 2231 struct timespec ts[2]; 2232 struct vnode *vp; 2233 int error; 2234 2235 if ((error = getutimes(tptr, ts)) != 0) 2236 return (error); 2237 /* XXX Add NLC flag indicating modifying operation? */ 2238 if ((error = nlookup(nd)) != 0) 2239 return (error); 2240 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0) 2241 return (error); 2242 error = setutimes(vp, ts, tptr == NULL); 2243 vrele(vp); 2244 return (error); 2245 } 2246 2247 /* 2248 * utimes_args(char *path, struct timeval *tptr) 2249 * 2250 * Set the access and modification times of a file. 2251 */ 2252 int 2253 utimes(struct utimes_args *uap) 2254 { 2255 struct timeval tv[2]; 2256 struct nlookupdata nd; 2257 int error; 2258 2259 if (uap->tptr) { 2260 error = copyin(uap->tptr, tv, sizeof(tv)); 2261 if (error) 2262 return (error); 2263 } 2264 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 2265 if (error == 0) 2266 error = kern_utimes(&nd, uap->tptr ? tv : NULL); 2267 nlookup_done(&nd); 2268 return (error); 2269 } 2270 2271 /* 2272 * lutimes_args(char *path, struct timeval *tptr) 2273 * 2274 * Set the access and modification times of a file. 2275 */ 2276 int 2277 lutimes(struct lutimes_args *uap) 2278 { 2279 struct timeval tv[2]; 2280 struct nlookupdata nd; 2281 int error; 2282 2283 if (uap->tptr) { 2284 error = copyin(uap->tptr, tv, sizeof(tv)); 2285 if (error) 2286 return (error); 2287 } 2288 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 2289 if (error == 0) 2290 error = kern_utimes(&nd, uap->tptr ? tv : NULL); 2291 nlookup_done(&nd); 2292 return (error); 2293 } 2294 2295 int 2296 kern_futimes(int fd, struct timeval *tptr) 2297 { 2298 struct thread *td = curthread; 2299 struct proc *p = td->td_proc; 2300 struct timespec ts[2]; 2301 struct file *fp; 2302 int error; 2303 2304 error = getutimes(tptr, ts); 2305 if (error) 2306 return (error); 2307 error = getvnode(p->p_fd, fd, &fp); 2308 if (error) 2309 return (error); 2310 error = setutimes((struct vnode *)fp->f_data, ts, tptr == NULL); 2311 return (error); 2312 } 2313 2314 /* 2315 * futimes_args(int fd, struct timeval *tptr) 2316 * 2317 * Set the access and modification times of a file. 2318 */ 2319 int 2320 futimes(struct futimes_args *uap) 2321 { 2322 struct timeval tv[2]; 2323 int error; 2324 2325 if (uap->tptr) { 2326 error = copyin(uap->tptr, tv, sizeof(tv)); 2327 if (error) 2328 return (error); 2329 } 2330 2331 error = kern_futimes(uap->fd, uap->tptr ? tv : NULL); 2332 2333 return (error); 2334 } 2335 2336 int 2337 kern_truncate(struct nlookupdata *nd, off_t length) 2338 { 2339 struct vnode *vp; 2340 struct vattr vattr; 2341 int error; 2342 2343 if (length < 0) 2344 return(EINVAL); 2345 /* XXX Add NLC flag indicating modifying operation? */ 2346 if ((error = nlookup(nd)) != 0) 2347 return (error); 2348 if ((error = cache_vref(nd->nl_ncp, nd->nl_cred, &vp)) != 0) 2349 return (error); 2350 VOP_LEASE(vp, nd->nl_td, nd->nl_cred, LEASE_WRITE); 2351 if ((error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, nd->nl_td)) != 0) { 2352 vrele(vp); 2353 return (error); 2354 } 2355 if (vp->v_type == VDIR) { 2356 error = EISDIR; 2357 } else if ((error = vn_writechk(vp)) == 0 && 2358 (error = VOP_ACCESS(vp, VWRITE, nd->nl_cred, nd->nl_td)) == 0) { 2359 VATTR_NULL(&vattr); 2360 vattr.va_size = length; 2361 error = VOP_SETATTR(vp, &vattr, nd->nl_cred, nd->nl_td); 2362 } 2363 vput(vp); 2364 return (error); 2365 } 2366 2367 /* 2368 * truncate(char *path, int pad, off_t length) 2369 * 2370 * Truncate a file given its path name. 2371 */ 2372 int 2373 truncate(struct truncate_args *uap) 2374 { 2375 struct nlookupdata nd; 2376 int error; 2377 2378 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 2379 if (error == 0) 2380 error = kern_truncate(&nd, uap->length); 2381 nlookup_done(&nd); 2382 return error; 2383 } 2384 2385 int 2386 kern_ftruncate(int fd, off_t length) 2387 { 2388 struct thread *td = curthread; 2389 struct proc *p = td->td_proc; 2390 struct vattr vattr; 2391 struct vnode *vp; 2392 struct file *fp; 2393 int error; 2394 2395 if (length < 0) 2396 return(EINVAL); 2397 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 2398 return (error); 2399 if ((fp->f_flag & FWRITE) == 0) 2400 return (EINVAL); 2401 vp = (struct vnode *)fp->f_data; 2402 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2403 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 2404 if (vp->v_type == VDIR) 2405 error = EISDIR; 2406 else if ((error = vn_writechk(vp)) == 0) { 2407 VATTR_NULL(&vattr); 2408 vattr.va_size = length; 2409 error = VOP_SETATTR(vp, &vattr, fp->f_cred, td); 2410 } 2411 VOP_UNLOCK(vp, 0, td); 2412 return (error); 2413 } 2414 2415 /* 2416 * ftruncate_args(int fd, int pad, off_t length) 2417 * 2418 * Truncate a file given a file descriptor. 2419 */ 2420 int 2421 ftruncate(struct ftruncate_args *uap) 2422 { 2423 int error; 2424 2425 error = kern_ftruncate(uap->fd, uap->length); 2426 2427 return (error); 2428 } 2429 2430 /* 2431 * fsync(int fd) 2432 * 2433 * Sync an open file. 2434 */ 2435 /* ARGSUSED */ 2436 int 2437 fsync(struct fsync_args *uap) 2438 { 2439 struct thread *td = curthread; 2440 struct proc *p = td->td_proc; 2441 struct vnode *vp; 2442 struct file *fp; 2443 vm_object_t obj; 2444 int error; 2445 2446 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2447 return (error); 2448 vp = (struct vnode *)fp->f_data; 2449 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 2450 if (VOP_GETVOBJECT(vp, &obj) == 0) 2451 vm_object_page_clean(obj, 0, 0, 0); 2452 if ((error = VOP_FSYNC(vp, MNT_WAIT, td)) == 0 && 2453 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP) && 2454 bioops.io_fsync) 2455 error = (*bioops.io_fsync)(vp); 2456 VOP_UNLOCK(vp, 0, td); 2457 return (error); 2458 } 2459 2460 int 2461 kern_rename(struct nlookupdata *fromnd, struct nlookupdata *tond) 2462 { 2463 struct namecache *fncpd; 2464 struct namecache *tncpd; 2465 struct namecache *ncp; 2466 struct mount *mp; 2467 int error; 2468 2469 bwillwrite(); 2470 if ((error = nlookup(fromnd)) != 0) 2471 return (error); 2472 if ((fncpd = fromnd->nl_ncp->nc_parent) == NULL) 2473 return (ENOENT); 2474 cache_hold(fncpd); 2475 2476 /* 2477 * unlock the source ncp so we can lookup the target ncp without 2478 * deadlocking. The target may or may not exist so we do not check 2479 * for a target vp like kern_mkdir() and other creation functions do. 2480 * 2481 * The source and target directories are ref'd and rechecked after 2482 * everything is relocked to determine if the source or target file 2483 * has been renamed. 2484 */ 2485 KKASSERT(fromnd->nl_flags & NLC_NCPISLOCKED); 2486 fromnd->nl_flags &= ~NLC_NCPISLOCKED; 2487 cache_unlock(fromnd->nl_ncp); 2488 2489 tond->nl_flags |= NLC_CREATE; 2490 if ((error = nlookup(tond)) != 0) { 2491 cache_drop(fncpd); 2492 return (error); 2493 } 2494 if ((tncpd = tond->nl_ncp->nc_parent) == NULL) { 2495 cache_drop(fncpd); 2496 return (ENOENT); 2497 } 2498 cache_hold(tncpd); 2499 2500 /* 2501 * If the source and target are the same there is nothing to do 2502 */ 2503 if (fromnd->nl_ncp == tond->nl_ncp) { 2504 cache_drop(fncpd); 2505 cache_drop(tncpd); 2506 return (0); 2507 } 2508 2509 /* 2510 * relock the source ncp 2511 */ 2512 if (cache_lock_nonblock(fromnd->nl_ncp) == 0) { 2513 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred); 2514 } else if (fromnd->nl_ncp > tond->nl_ncp) { 2515 cache_lock(fromnd->nl_ncp); 2516 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred); 2517 } else { 2518 cache_unlock(tond->nl_ncp); 2519 cache_lock(fromnd->nl_ncp); 2520 cache_resolve(fromnd->nl_ncp, fromnd->nl_cred); 2521 cache_lock(tond->nl_ncp); 2522 cache_resolve(tond->nl_ncp, tond->nl_cred); 2523 } 2524 fromnd->nl_flags |= NLC_NCPISLOCKED; 2525 2526 /* 2527 * make sure the parent directories linkages are the same 2528 */ 2529 if (fncpd != fromnd->nl_ncp->nc_parent || 2530 tncpd != tond->nl_ncp->nc_parent) { 2531 cache_drop(fncpd); 2532 cache_drop(tncpd); 2533 return (ENOENT); 2534 } 2535 2536 /* 2537 * Both the source and target must be within the same filesystem and 2538 * in the same filesystem as their parent directories within the 2539 * namecache topology. 2540 */ 2541 mp = fncpd->nc_mount; 2542 if (mp != tncpd->nc_mount || mp != fromnd->nl_ncp->nc_mount || 2543 mp != tond->nl_ncp->nc_mount) { 2544 cache_drop(fncpd); 2545 cache_drop(tncpd); 2546 return (EXDEV); 2547 } 2548 2549 /* 2550 * If the target exists and either the source or target is a directory, 2551 * then both must be directories. 2552 */ 2553 if (tond->nl_ncp->nc_vp) { 2554 if (fromnd->nl_ncp->nc_vp->v_type == VDIR) { 2555 if (tond->nl_ncp->nc_vp->v_type != VDIR) 2556 error = ENOTDIR; 2557 } else if (tond->nl_ncp->nc_vp->v_type == VDIR) { 2558 error = EISDIR; 2559 } 2560 } 2561 2562 /* 2563 * You cannot rename a source into itself or a subdirectory of itself. 2564 * We check this by travsersing the target directory upwards looking 2565 * for a match against the source. 2566 */ 2567 if (error == 0) { 2568 for (ncp = tncpd; ncp; ncp = ncp->nc_parent) { 2569 if (fromnd->nl_ncp == ncp) { 2570 error = EINVAL; 2571 break; 2572 } 2573 } 2574 } 2575 2576 cache_drop(fncpd); 2577 cache_drop(tncpd); 2578 if (error) 2579 return (error); 2580 error = VOP_NRENAME(fromnd->nl_ncp, tond->nl_ncp, tond->nl_cred); 2581 return (error); 2582 } 2583 2584 /* 2585 * rename_args(char *from, char *to) 2586 * 2587 * Rename files. Source and destination must either both be directories, 2588 * or both not be directories. If target is a directory, it must be empty. 2589 */ 2590 int 2591 rename(struct rename_args *uap) 2592 { 2593 struct nlookupdata fromnd, tond; 2594 int error; 2595 2596 error = nlookup_init(&fromnd, uap->from, UIO_USERSPACE, 0); 2597 if (error == 0) { 2598 error = nlookup_init(&tond, uap->to, UIO_USERSPACE, 0); 2599 if (error == 0) 2600 error = kern_rename(&fromnd, &tond); 2601 nlookup_done(&tond); 2602 } 2603 nlookup_done(&fromnd); 2604 return (error); 2605 } 2606 2607 int 2608 kern_mkdir(struct nlookupdata *nd, int mode) 2609 { 2610 struct thread *td = curthread; 2611 struct proc *p = td->td_proc; 2612 struct namecache *ncp; 2613 struct vnode *vp; 2614 struct vattr vattr; 2615 int error; 2616 2617 bwillwrite(); 2618 nd->nl_flags |= NLC_WILLBEDIR | NLC_CREATE; 2619 if ((error = nlookup(nd)) != 0) 2620 return (error); 2621 2622 ncp = nd->nl_ncp; 2623 if (ncp->nc_vp) 2624 return (EEXIST); 2625 2626 VATTR_NULL(&vattr); 2627 vattr.va_type = VDIR; 2628 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_fd->fd_cmask; 2629 2630 vp = NULL; 2631 error = VOP_NMKDIR(ncp, &vp, p->p_ucred, &vattr); 2632 if (error == 0) 2633 vput(vp); 2634 return (error); 2635 } 2636 2637 /* 2638 * mkdir_args(char *path, int mode) 2639 * 2640 * Make a directory file. 2641 */ 2642 /* ARGSUSED */ 2643 int 2644 mkdir(struct mkdir_args *uap) 2645 { 2646 struct nlookupdata nd; 2647 int error; 2648 2649 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 2650 if (error == 0) 2651 error = kern_mkdir(&nd, uap->mode); 2652 nlookup_done(&nd); 2653 return (error); 2654 } 2655 2656 int 2657 kern_rmdir(struct nlookupdata *nd) 2658 { 2659 struct namecache *ncp; 2660 int error; 2661 2662 bwillwrite(); 2663 nd->nl_flags |= NLC_DELETE; 2664 if ((error = nlookup(nd)) != 0) 2665 return (error); 2666 2667 ncp = nd->nl_ncp; 2668 error = VOP_NRMDIR(ncp, nd->nl_cred); 2669 return (error); 2670 } 2671 2672 /* 2673 * rmdir_args(char *path) 2674 * 2675 * Remove a directory file. 2676 */ 2677 /* ARGSUSED */ 2678 int 2679 rmdir(struct rmdir_args *uap) 2680 { 2681 struct nlookupdata nd; 2682 int error; 2683 2684 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 0); 2685 if (error == 0) 2686 error = kern_rmdir(&nd); 2687 nlookup_done(&nd); 2688 return (error); 2689 } 2690 2691 int 2692 kern_getdirentries(int fd, char *buf, u_int count, long *basep, int *res) 2693 { 2694 struct thread *td = curthread; 2695 struct proc *p = td->td_proc; 2696 struct vnode *vp; 2697 struct file *fp; 2698 struct uio auio; 2699 struct iovec aiov; 2700 long loff; 2701 int error, eofflag; 2702 2703 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 2704 return (error); 2705 if ((fp->f_flag & FREAD) == 0) 2706 return (EBADF); 2707 vp = (struct vnode *)fp->f_data; 2708 unionread: 2709 if (vp->v_type != VDIR) 2710 return (EINVAL); 2711 aiov.iov_base = buf; 2712 aiov.iov_len = count; 2713 auio.uio_iov = &aiov; 2714 auio.uio_iovcnt = 1; 2715 auio.uio_rw = UIO_READ; 2716 auio.uio_segflg = UIO_USERSPACE; 2717 auio.uio_td = td; 2718 auio.uio_resid = count; 2719 /* vn_lock(vp, LK_SHARED | LK_RETRY, td); */ 2720 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 2721 loff = auio.uio_offset = fp->f_offset; 2722 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL, NULL); 2723 fp->f_offset = auio.uio_offset; 2724 VOP_UNLOCK(vp, 0, td); 2725 if (error) 2726 return (error); 2727 if (count == auio.uio_resid) { 2728 if (union_dircheckp) { 2729 error = union_dircheckp(td, &vp, fp); 2730 if (error == -1) 2731 goto unionread; 2732 if (error) 2733 return (error); 2734 } 2735 if ((vp->v_flag & VROOT) && 2736 (vp->v_mount->mnt_flag & MNT_UNION)) { 2737 struct vnode *tvp = vp; 2738 vp = vp->v_mount->mnt_vnodecovered; 2739 vref(vp); 2740 fp->f_data = (caddr_t)vp; 2741 fp->f_offset = 0; 2742 vrele(tvp); 2743 goto unionread; 2744 } 2745 } 2746 if (basep) { 2747 *basep = loff; 2748 } 2749 *res = count - auio.uio_resid; 2750 return (error); 2751 } 2752 2753 /* 2754 * getdirentries_args(int fd, char *buf, u_int conut, long *basep) 2755 * 2756 * Read a block of directory entries in a file system independent format. 2757 */ 2758 int 2759 getdirentries(struct getdirentries_args *uap) 2760 { 2761 long base; 2762 int error; 2763 2764 error = kern_getdirentries(uap->fd, uap->buf, uap->count, &base, 2765 &uap->sysmsg_result); 2766 2767 if (error == 0) 2768 error = copyout(&base, uap->basep, sizeof(*uap->basep)); 2769 return (error); 2770 } 2771 2772 /* 2773 * getdents_args(int fd, char *buf, size_t count) 2774 */ 2775 int 2776 getdents(struct getdents_args *uap) 2777 { 2778 int error; 2779 2780 error = kern_getdirentries(uap->fd, uap->buf, uap->count, NULL, 2781 &uap->sysmsg_result); 2782 2783 return (error); 2784 } 2785 2786 /* 2787 * umask(int newmask) 2788 * 2789 * Set the mode mask for creation of filesystem nodes. 2790 * 2791 * MP SAFE 2792 */ 2793 int 2794 umask(struct umask_args *uap) 2795 { 2796 struct thread *td = curthread; 2797 struct proc *p = td->td_proc; 2798 struct filedesc *fdp; 2799 2800 fdp = p->p_fd; 2801 uap->sysmsg_result = fdp->fd_cmask; 2802 fdp->fd_cmask = SCARG(uap, newmask) & ALLPERMS; 2803 return (0); 2804 } 2805 2806 /* 2807 * revoke(char *path) 2808 * 2809 * Void all references to file by ripping underlying filesystem 2810 * away from vnode. 2811 */ 2812 /* ARGSUSED */ 2813 int 2814 revoke(struct revoke_args *uap) 2815 { 2816 struct thread *td = curthread; 2817 struct nlookupdata nd; 2818 struct vattr vattr; 2819 struct vnode *vp; 2820 struct ucred *cred; 2821 int error; 2822 2823 vp = NULL; 2824 error = nlookup_init(&nd, SCARG(uap, path), UIO_USERSPACE, NLC_FOLLOW); 2825 if (error == 0) 2826 error = nlookup(&nd); 2827 if (error == 0) 2828 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp); 2829 cred = crhold(nd.nl_cred); 2830 nlookup_done(&nd); 2831 if (error == 0) { 2832 if (vp->v_type != VCHR && vp->v_type != VBLK) 2833 error = EINVAL; 2834 if (error == 0) 2835 error = VOP_GETATTR(vp, &vattr, td); 2836 if (error == 0 && cred->cr_uid != vattr.va_uid) 2837 error = suser_cred(cred, PRISON_ROOT); 2838 if (error == 0 && count_udev(vp->v_udev) > 0) { 2839 if ((error = vx_lock(vp)) == 0) { 2840 VOP_REVOKE(vp, REVOKEALL); 2841 vx_unlock(vp); 2842 } 2843 } 2844 vrele(vp); 2845 } 2846 crfree(cred); 2847 return (error); 2848 } 2849 2850 /* 2851 * Convert a user file descriptor to a kernel file entry. 2852 */ 2853 int 2854 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 2855 { 2856 struct file *fp; 2857 2858 if ((u_int)fd >= fdp->fd_nfiles || 2859 (fp = fdp->fd_ofiles[fd]) == NULL) 2860 return (EBADF); 2861 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_FIFO) 2862 return (EINVAL); 2863 *fpp = fp; 2864 return (0); 2865 } 2866 /* 2867 * getfh_args(char *fname, fhandle_t *fhp) 2868 * 2869 * Get (NFS) file handle 2870 */ 2871 int 2872 getfh(struct getfh_args *uap) 2873 { 2874 struct thread *td = curthread; 2875 struct nlookupdata nd; 2876 fhandle_t fh; 2877 struct vnode *vp; 2878 int error; 2879 2880 /* 2881 * Must be super user 2882 */ 2883 if ((error = suser(td)) != 0) 2884 return (error); 2885 2886 vp = NULL; 2887 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW); 2888 if (error == 0) 2889 error = nlookup(&nd); 2890 if (error == 0) 2891 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 2892 nlookup_done(&nd); 2893 if (error == 0) { 2894 bzero(&fh, sizeof(fh)); 2895 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid; 2896 error = VFS_VPTOFH(vp, &fh.fh_fid); 2897 vput(vp); 2898 if (error == 0) 2899 error = copyout(&fh, uap->fhp, sizeof(fh)); 2900 } 2901 return (error); 2902 } 2903 2904 /* 2905 * fhopen_args(const struct fhandle *u_fhp, int flags) 2906 * 2907 * syscall for the rpc.lockd to use to translate a NFS file handle into 2908 * an open descriptor. 2909 * 2910 * warning: do not remove the suser() call or this becomes one giant 2911 * security hole. 2912 */ 2913 int 2914 fhopen(struct fhopen_args *uap) 2915 { 2916 struct thread *td = curthread; 2917 struct proc *p = td->td_proc; 2918 struct mount *mp; 2919 struct vnode *vp; 2920 struct fhandle fhp; 2921 struct vattr vat; 2922 struct vattr *vap = &vat; 2923 struct flock lf; 2924 struct filedesc *fdp = p->p_fd; 2925 int fmode, mode, error, type; 2926 struct file *nfp; 2927 struct file *fp; 2928 int indx; 2929 2930 /* 2931 * Must be super user 2932 */ 2933 error = suser(td); 2934 if (error) 2935 return (error); 2936 2937 fmode = FFLAGS(SCARG(uap, flags)); 2938 /* why not allow a non-read/write open for our lockd? */ 2939 if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT)) 2940 return (EINVAL); 2941 error = copyin(SCARG(uap,u_fhp), &fhp, sizeof(fhp)); 2942 if (error) 2943 return(error); 2944 /* find the mount point */ 2945 mp = vfs_getvfs(&fhp.fh_fsid); 2946 if (mp == NULL) 2947 return (ESTALE); 2948 /* now give me my vnode, it gets returned to me locked */ 2949 error = VFS_FHTOVP(mp, &fhp.fh_fid, &vp); 2950 if (error) 2951 return (error); 2952 /* 2953 * from now on we have to make sure not 2954 * to forget about the vnode 2955 * any error that causes an abort must vput(vp) 2956 * just set error = err and 'goto bad;'. 2957 */ 2958 2959 /* 2960 * from vn_open 2961 */ 2962 if (vp->v_type == VLNK) { 2963 error = EMLINK; 2964 goto bad; 2965 } 2966 if (vp->v_type == VSOCK) { 2967 error = EOPNOTSUPP; 2968 goto bad; 2969 } 2970 mode = 0; 2971 if (fmode & (FWRITE | O_TRUNC)) { 2972 if (vp->v_type == VDIR) { 2973 error = EISDIR; 2974 goto bad; 2975 } 2976 error = vn_writechk(vp); 2977 if (error) 2978 goto bad; 2979 mode |= VWRITE; 2980 } 2981 if (fmode & FREAD) 2982 mode |= VREAD; 2983 if (mode) { 2984 error = VOP_ACCESS(vp, mode, p->p_ucred, td); 2985 if (error) 2986 goto bad; 2987 } 2988 if (fmode & O_TRUNC) { 2989 VOP_UNLOCK(vp, 0, td); /* XXX */ 2990 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 2991 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); /* XXX */ 2992 VATTR_NULL(vap); 2993 vap->va_size = 0; 2994 error = VOP_SETATTR(vp, vap, p->p_ucred, td); 2995 if (error) 2996 goto bad; 2997 } 2998 2999 /* 3000 * VOP_OPEN needs the file pointer so it can potentially override 3001 * it. 3002 * 3003 * WARNING! no f_ncp will be associated when fhopen()ing a directory. 3004 * XXX 3005 */ 3006 if ((error = falloc(p, &nfp, NULL)) != 0) 3007 goto bad; 3008 fp = nfp; 3009 3010 fp->f_data = (caddr_t)vp; 3011 fp->f_flag = fmode & FMASK; 3012 fp->f_ops = &vnode_fileops; 3013 fp->f_type = DTYPE_VNODE; 3014 3015 error = VOP_OPEN(vp, fmode, p->p_ucred, fp, td); 3016 if (error) { 3017 /* 3018 * setting f_ops this way prevents VOP_CLOSE from being 3019 * called or fdrop() releasing the vp from v_data. Since 3020 * the VOP_OPEN failed we don't want to VOP_CLOSE. 3021 */ 3022 fp->f_ops = &badfileops; 3023 fp->f_data = NULL; 3024 fdrop(fp, td); 3025 goto bad; 3026 } 3027 if (fmode & FWRITE) 3028 vp->v_writecount++; 3029 3030 /* 3031 * The fp now owns a reference on the vnode. We still have our own 3032 * ref+lock. 3033 */ 3034 vref(vp); 3035 3036 /* 3037 * Make sure that a VM object is created for VMIO support. If this 3038 * fails just fdrop() normally to clean up. 3039 */ 3040 if (vn_canvmio(vp) == TRUE) { 3041 if ((error = vfs_object_create(vp, td)) != 0) { 3042 fdrop(fp, td); 3043 goto bad; 3044 } 3045 } 3046 3047 /* 3048 * The open was successful, associate it with a file descriptor. 3049 */ 3050 if ((error = fsetfd(p, fp, &indx)) != 0) { 3051 if (fmode & FWRITE) 3052 vp->v_writecount--; 3053 fdrop(fp, td); 3054 goto bad; 3055 } 3056 3057 if (fmode & (O_EXLOCK | O_SHLOCK)) { 3058 lf.l_whence = SEEK_SET; 3059 lf.l_start = 0; 3060 lf.l_len = 0; 3061 if (fmode & O_EXLOCK) 3062 lf.l_type = F_WRLCK; 3063 else 3064 lf.l_type = F_RDLCK; 3065 type = F_FLOCK; 3066 if ((fmode & FNONBLOCK) == 0) 3067 type |= F_WAIT; 3068 VOP_UNLOCK(vp, 0, td); 3069 if ((error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, type)) != 0) { 3070 /* 3071 * lock request failed. Normally close the descriptor 3072 * but handle the case where someone might have dup()d 3073 * or close()d it when we weren't looking. 3074 */ 3075 if (fdp->fd_ofiles[indx] == fp) { 3076 fdp->fd_ofiles[indx] = NULL; 3077 fdrop(fp, td); 3078 } 3079 3080 /* 3081 * release our private reference. 3082 */ 3083 fdrop(fp, td); 3084 vrele(vp); 3085 return (error); 3086 } 3087 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 3088 fp->f_flag |= FHASLOCK; 3089 } 3090 if ((vp->v_type == VREG) && (VOP_GETVOBJECT(vp, NULL) != 0)) 3091 vfs_object_create(vp, td); 3092 3093 vput(vp); 3094 fdrop(fp, td); 3095 uap->sysmsg_result = indx; 3096 return (0); 3097 3098 bad: 3099 vput(vp); 3100 return (error); 3101 } 3102 3103 /* 3104 * fhstat_args(struct fhandle *u_fhp, struct stat *sb) 3105 */ 3106 int 3107 fhstat(struct fhstat_args *uap) 3108 { 3109 struct thread *td = curthread; 3110 struct stat sb; 3111 fhandle_t fh; 3112 struct mount *mp; 3113 struct vnode *vp; 3114 int error; 3115 3116 /* 3117 * Must be super user 3118 */ 3119 error = suser(td); 3120 if (error) 3121 return (error); 3122 3123 error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t)); 3124 if (error) 3125 return (error); 3126 3127 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 3128 return (ESTALE); 3129 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 3130 return (error); 3131 error = vn_stat(vp, &sb, td); 3132 vput(vp); 3133 if (error) 3134 return (error); 3135 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 3136 return (error); 3137 } 3138 3139 /* 3140 * fhstatfs_args(struct fhandle *u_fhp, struct statfs *buf) 3141 */ 3142 int 3143 fhstatfs(struct fhstatfs_args *uap) 3144 { 3145 struct thread *td = curthread; 3146 struct statfs *sp; 3147 struct mount *mp; 3148 struct vnode *vp; 3149 struct statfs sb; 3150 fhandle_t fh; 3151 int error; 3152 3153 /* 3154 * Must be super user 3155 */ 3156 if ((error = suser(td))) 3157 return (error); 3158 3159 if ((error = copyin(SCARG(uap, u_fhp), &fh, sizeof(fhandle_t))) != 0) 3160 return (error); 3161 3162 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 3163 return (ESTALE); 3164 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 3165 return (error); 3166 mp = vp->v_mount; 3167 sp = &mp->mnt_stat; 3168 vput(vp); 3169 if ((error = VFS_STATFS(mp, sp, td)) != 0) 3170 return (error); 3171 sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK; 3172 if (suser(td)) { 3173 bcopy(sp, &sb, sizeof(sb)); 3174 sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0; 3175 sp = &sb; 3176 } 3177 return (copyout(sp, SCARG(uap, buf), sizeof(*sp))); 3178 } 3179 3180 /* 3181 * Syscall to push extended attribute configuration information into the 3182 * VFS. Accepts a path, which it converts to a mountpoint, as well as 3183 * a command (int cmd), and attribute name and misc data. For now, the 3184 * attribute name is left in userspace for consumption by the VFS_op. 3185 * It will probably be changed to be copied into sysspace by the 3186 * syscall in the future, once issues with various consumers of the 3187 * attribute code have raised their hands. 3188 * 3189 * Currently this is used only by UFS Extended Attributes. 3190 */ 3191 int 3192 extattrctl(struct extattrctl_args *uap) 3193 { 3194 struct nlookupdata nd; 3195 struct mount *mp; 3196 struct vnode *vp; 3197 int error; 3198 3199 vp = NULL; 3200 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 3201 if (error == 0) 3202 error = nlookup(&nd); 3203 if (error == 0) { 3204 mp = nd.nl_ncp->nc_mount; 3205 error = VFS_EXTATTRCTL(mp, SCARG(uap, cmd), 3206 SCARG(uap, attrname), SCARG(uap, arg), 3207 nd.nl_td); 3208 } 3209 nlookup_done(&nd); 3210 return (error); 3211 } 3212 3213 /* 3214 * Syscall to set a named extended attribute on a file or directory. 3215 * Accepts attribute name, and a uio structure pointing to the data to set. 3216 * The uio is consumed in the style of writev(). The real work happens 3217 * in VOP_SETEXTATTR(). 3218 */ 3219 int 3220 extattr_set_file(struct extattr_set_file_args *uap) 3221 { 3222 char attrname[EXTATTR_MAXNAMELEN]; 3223 struct iovec aiov[UIO_SMALLIOV]; 3224 struct iovec *needfree; 3225 struct nlookupdata nd; 3226 struct iovec *iov; 3227 struct vnode *vp; 3228 struct uio auio; 3229 u_int iovlen; 3230 u_int cnt; 3231 int error; 3232 int i; 3233 3234 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN); 3235 if (error) 3236 return (error); 3237 3238 vp = NULL; 3239 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 3240 if (error == 0) 3241 error = nlookup(&nd); 3242 if (error == 0) 3243 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 3244 if (error) { 3245 nlookup_done(&nd); 3246 return (error); 3247 } 3248 3249 needfree = NULL; 3250 iovlen = uap->iovcnt * sizeof(struct iovec); 3251 if (uap->iovcnt > UIO_SMALLIOV) { 3252 if (uap->iovcnt > UIO_MAXIOV) { 3253 error = EINVAL; 3254 goto done; 3255 } 3256 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 3257 needfree = iov; 3258 } else { 3259 iov = aiov; 3260 } 3261 auio.uio_iov = iov; 3262 auio.uio_iovcnt = uap->iovcnt; 3263 auio.uio_rw = UIO_WRITE; 3264 auio.uio_segflg = UIO_USERSPACE; 3265 auio.uio_td = nd.nl_td; 3266 auio.uio_offset = 0; 3267 if ((error = copyin(uap->iovp, iov, iovlen))) 3268 goto done; 3269 auio.uio_resid = 0; 3270 for (i = 0; i < uap->iovcnt; i++) { 3271 if (iov->iov_len > INT_MAX - auio.uio_resid) { 3272 error = EINVAL; 3273 goto done; 3274 } 3275 auio.uio_resid += iov->iov_len; 3276 iov++; 3277 } 3278 cnt = auio.uio_resid; 3279 error = VOP_SETEXTATTR(vp, attrname, &auio, nd.nl_cred, nd.nl_td); 3280 cnt -= auio.uio_resid; 3281 uap->sysmsg_result = cnt; 3282 done: 3283 vput(vp); 3284 nlookup_done(&nd); 3285 if (needfree) 3286 FREE(needfree, M_IOV); 3287 return (error); 3288 } 3289 3290 /* 3291 * Syscall to get a named extended attribute on a file or directory. 3292 * Accepts attribute name, and a uio structure pointing to a buffer for the 3293 * data. The uio is consumed in the style of readv(). The real work 3294 * happens in VOP_GETEXTATTR(); 3295 */ 3296 int 3297 extattr_get_file(struct extattr_get_file_args *uap) 3298 { 3299 char attrname[EXTATTR_MAXNAMELEN]; 3300 struct iovec aiov[UIO_SMALLIOV]; 3301 struct iovec *needfree; 3302 struct nlookupdata nd; 3303 struct iovec *iov; 3304 struct vnode *vp; 3305 struct uio auio; 3306 u_int iovlen; 3307 u_int cnt; 3308 int error; 3309 int i; 3310 3311 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN); 3312 if (error) 3313 return (error); 3314 3315 vp = NULL; 3316 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 3317 if (error == 0) 3318 error = nlookup(&nd); 3319 if (error == 0) 3320 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 3321 if (error) { 3322 nlookup_done(&nd); 3323 return (error); 3324 } 3325 3326 iovlen = uap->iovcnt * sizeof (struct iovec); 3327 needfree = NULL; 3328 if (uap->iovcnt > UIO_SMALLIOV) { 3329 if (uap->iovcnt > UIO_MAXIOV) { 3330 error = EINVAL; 3331 goto done; 3332 } 3333 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 3334 needfree = iov; 3335 } else { 3336 iov = aiov; 3337 } 3338 auio.uio_iov = iov; 3339 auio.uio_iovcnt = uap->iovcnt; 3340 auio.uio_rw = UIO_READ; 3341 auio.uio_segflg = UIO_USERSPACE; 3342 auio.uio_td = nd.nl_td; 3343 auio.uio_offset = 0; 3344 if ((error = copyin(uap->iovp, iov, iovlen))) 3345 goto done; 3346 auio.uio_resid = 0; 3347 for (i = 0; i < uap->iovcnt; i++) { 3348 if (iov->iov_len > INT_MAX - auio.uio_resid) { 3349 error = EINVAL; 3350 goto done; 3351 } 3352 auio.uio_resid += iov->iov_len; 3353 iov++; 3354 } 3355 cnt = auio.uio_resid; 3356 error = VOP_GETEXTATTR(vp, attrname, &auio, nd.nl_cred, nd.nl_td); 3357 cnt -= auio.uio_resid; 3358 uap->sysmsg_result = cnt; 3359 done: 3360 vput(vp); 3361 nlookup_done(&nd); 3362 if (needfree) 3363 FREE(needfree, M_IOV); 3364 return(error); 3365 } 3366 3367 /* 3368 * Syscall to delete a named extended attribute from a file or directory. 3369 * Accepts attribute name. The real work happens in VOP_SETEXTATTR(). 3370 */ 3371 int 3372 extattr_delete_file(struct extattr_delete_file_args *uap) 3373 { 3374 char attrname[EXTATTR_MAXNAMELEN]; 3375 struct nlookupdata nd; 3376 struct vnode *vp; 3377 int error; 3378 3379 error = copyin(SCARG(uap, attrname), attrname, EXTATTR_MAXNAMELEN); 3380 if (error) 3381 return(error); 3382 3383 vp = NULL; 3384 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, NLC_FOLLOW); 3385 if (error == 0) 3386 error = nlookup(&nd); 3387 if (error == 0) 3388 error = cache_vget(nd.nl_ncp, nd.nl_cred, LK_EXCLUSIVE, &vp); 3389 if (error) { 3390 nlookup_done(&nd); 3391 return (error); 3392 } 3393 3394 error = VOP_SETEXTATTR(vp, attrname, NULL, nd.nl_cred, nd.nl_td); 3395 vput(vp); 3396 nlookup_done(&nd); 3397 return(error); 3398 } 3399 3400 /* 3401 * print out statistics from the current status of the buffer pool 3402 * this can be toggeled by the system control option debug.syncprt 3403 */ 3404 #ifdef DEBUG 3405 void 3406 vfs_bufstats(void) 3407 { 3408 int s, i, j, count; 3409 struct buf *bp; 3410 struct bqueues *dp; 3411 int counts[(MAXBSIZE / PAGE_SIZE) + 1]; 3412 static char *bname[3] = { "LOCKED", "LRU", "AGE" }; 3413 3414 for (dp = bufqueues, i = 0; dp < &bufqueues[3]; dp++, i++) { 3415 count = 0; 3416 for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++) 3417 counts[j] = 0; 3418 s = splbio(); 3419 TAILQ_FOREACH(bp, dp, b_freelist) { 3420 counts[bp->b_bufsize/PAGE_SIZE]++; 3421 count++; 3422 } 3423 splx(s); 3424 printf("%s: total-%d", bname[i], count); 3425 for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++) 3426 if (counts[j] != 0) 3427 printf(", %d-%d", j * PAGE_SIZE, counts[j]); 3428 printf("\n"); 3429 } 3430 } 3431 #endif 3432