vfs_syscalls.c revision 1.306.2.8
1 /* $NetBSD: vfs_syscalls.c,v 1.306.2.8 2007/06/17 21:31:34 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.306.2.8 2007/06/17 21:31:34 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/syscallargs.h> 65 #include <sys/vfs_syscalls.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #include <sys/verified_exec.h> 73 #include <sys/kauth.h> 74 75 #include <miscfs/genfs/genfs.h> 76 #include <miscfs/syncfs/syncfs.h> 77 78 #ifdef COMPAT_30 79 #include "opt_nfsserver.h" 80 #include <nfs/rpcv2.h> 81 #endif 82 #include <nfs/nfsproto.h> 83 #ifdef COMPAT_30 84 #include <nfs/nfs.h> 85 #include <nfs/nfs_var.h> 86 #endif 87 88 #if NFSS > 0 89 #include <dev/fssvar.h> 90 #endif 91 92 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 93 94 static int change_dir(struct nameidata *, struct lwp *); 95 static int change_flags(struct vnode *, u_long, struct lwp *); 96 static int change_mode(struct vnode *, int, struct lwp *l); 97 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 98 static int rename_files(const char *, const char *, struct lwp *, int); 99 100 void checkdirs(struct vnode *); 101 102 int dovfsusermount = 0; 103 104 /* 105 * Virtual File System System Calls 106 */ 107 108 /* 109 * Mount a file system. 110 */ 111 112 #if defined(COMPAT_09) || defined(COMPAT_43) 113 /* 114 * This table is used to maintain compatibility with 4.3BSD 115 * and NetBSD 0.9 mount syscalls. Note, the order is important! 116 * 117 * Do not modify this table. It should only contain filesystems 118 * supported by NetBSD 0.9 and 4.3BSD. 119 */ 120 const char * const mountcompatnames[] = { 121 NULL, /* 0 = MOUNT_NONE */ 122 MOUNT_FFS, /* 1 = MOUNT_UFS */ 123 MOUNT_NFS, /* 2 */ 124 MOUNT_MFS, /* 3 */ 125 MOUNT_MSDOS, /* 4 */ 126 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 127 MOUNT_FDESC, /* 6 */ 128 MOUNT_KERNFS, /* 7 */ 129 NULL, /* 8 = MOUNT_DEVFS */ 130 MOUNT_AFS, /* 9 */ 131 }; 132 const int nmountcompatnames = sizeof(mountcompatnames) / 133 sizeof(mountcompatnames[0]); 134 #endif /* COMPAT_09 || COMPAT_43 */ 135 136 static int 137 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 138 void *data, struct nameidata *ndp) 139 { 140 struct mount *mp; 141 int error = 0, saved_flags; 142 143 mp = vp->v_mount; 144 saved_flags = mp->mnt_flag; 145 146 /* We can operate only on VROOT nodes. */ 147 if ((vp->v_vflag & VV_ROOT) == 0) { 148 error = EINVAL; 149 goto out; 150 } 151 152 /* 153 * We only allow the filesystem to be reloaded if it 154 * is currently mounted read-only. 155 */ 156 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 157 error = EOPNOTSUPP; /* Needs translation */ 158 goto out; 159 } 160 161 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 162 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 163 if (error) 164 goto out; 165 166 if (vfs_busy(mp, LK_NOWAIT, 0)) { 167 error = EPERM; 168 goto out; 169 } 170 171 mp->mnt_flag &= ~MNT_OP_FLAGS; 172 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 173 174 /* 175 * Set the mount level flags. 176 */ 177 if (flags & MNT_RDONLY) 178 mp->mnt_flag |= MNT_RDONLY; 179 else if (mp->mnt_flag & MNT_RDONLY) 180 mp->mnt_iflag |= IMNT_WANTRDWR; 181 mp->mnt_flag &= 182 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 183 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 184 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 185 mp->mnt_flag |= flags & 186 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 187 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 188 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 189 MNT_IGNORE); 190 191 error = VFS_MOUNT(mp, path, data, ndp, l); 192 193 #if defined(COMPAT_30) && defined(NFSSERVER) 194 if (error) { 195 int error2; 196 197 /* Update failed; let's try and see if it was an 198 * export request. */ 199 error2 = nfs_update_exports_30(mp, path, data, l); 200 201 /* Only update error code if the export request was 202 * understood but some problem occurred while 203 * processing it. */ 204 if (error2 != EJUSTRETURN) 205 error = error2; 206 } 207 #endif 208 if (mp->mnt_iflag & IMNT_WANTRDWR) 209 mp->mnt_flag &= ~MNT_RDONLY; 210 if (error) 211 mp->mnt_flag = saved_flags; 212 mp->mnt_flag &= ~MNT_OP_FLAGS; 213 mp->mnt_iflag &= ~IMNT_WANTRDWR; 214 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 215 if (mp->mnt_syncer == NULL) 216 error = vfs_allocate_syncvnode(mp); 217 } else { 218 if (mp->mnt_syncer != NULL) 219 vfs_deallocate_syncvnode(mp); 220 } 221 vfs_unbusy(mp); 222 223 out: 224 return (error); 225 } 226 227 static int 228 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 229 const char *path, int flags, void *data, struct nameidata *ndp) 230 { 231 struct mount *mp = NULL; 232 struct vattr va; 233 char fstypename[MFSNAMELEN]; 234 int error; 235 236 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 237 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 238 if (error) { 239 vput(vp); 240 goto out; 241 } 242 243 /* Can't make a non-dir a mount-point (from here anyway). */ 244 if (vp->v_type != VDIR) { 245 error = ENOTDIR; 246 vput(vp); 247 goto out; 248 } 249 250 /* 251 * If the user is not root, ensure that they own the directory 252 * onto which we are attempting to mount. 253 */ 254 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 255 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 256 (error = kauth_authorize_generic(l->l_cred, 257 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 258 vput(vp); 259 goto out; 260 } 261 262 if (flags & MNT_EXPORTED) { 263 error = EINVAL; 264 vput(vp); 265 goto out; 266 } 267 268 /* 269 * Copy file-system type from userspace. 270 */ 271 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 272 if (error) { 273 #if defined(COMPAT_09) || defined(COMPAT_43) 274 /* 275 * Historically, filesystem types were identified by numbers. 276 * If we get an integer for the filesystem type instead of a 277 * string, we check to see if it matches one of the historic 278 * filesystem types. 279 */ 280 u_long fsindex = (u_long)fstype; 281 if (fsindex >= nmountcompatnames || 282 mountcompatnames[fsindex] == NULL) { 283 error = ENODEV; 284 vput(vp); 285 goto out; 286 } 287 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 288 #else 289 vput(vp); 290 goto out; 291 #endif 292 } 293 294 #ifdef COMPAT_10 295 /* Accept `ufs' as an alias for `ffs'. */ 296 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 297 strlcpy(fstypename, "ffs", sizeof(fstypename)); 298 #endif 299 300 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 301 vput(vp); 302 goto out; 303 } 304 305 /* 306 * Check if a file-system is not already mounted on this vnode. 307 */ 308 if (vp->v_mountedhere != NULL) { 309 error = EBUSY; 310 vput(vp); 311 goto out; 312 } 313 314 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 315 316 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 317 free(mp, M_MOUNT); 318 error = ENODEV; 319 vput(vp); 320 goto out; 321 } 322 323 TAILQ_INIT(&mp->mnt_vnodelist); 324 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 325 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 326 (void)vfs_busy(mp, LK_NOWAIT, 0); 327 328 mp->mnt_vnodecovered = vp; 329 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 330 mp->mnt_unmounter = NULL; 331 mount_initspecific(mp); 332 333 /* 334 * The underlying file system may refuse the mount for 335 * various reasons. Allow the user to force it to happen. 336 * 337 * Set the mount level flags. 338 */ 339 mp->mnt_flag = flags & 340 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 341 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 342 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 343 MNT_IGNORE | MNT_RDONLY); 344 345 error = VFS_MOUNT(mp, path, data, ndp, l); 346 mp->mnt_flag &= ~MNT_OP_FLAGS; 347 348 /* 349 * Put the new filesystem on the mount list after root. 350 */ 351 cache_purge(vp); 352 if (!error) { 353 mp->mnt_iflag &= ~IMNT_WANTRDWR; 354 vp->v_mountedhere = mp; 355 mutex_enter(&mountlist_lock); 356 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 357 mutex_exit(&mountlist_lock); 358 VOP_UNLOCK(vp, 0); 359 checkdirs(vp); 360 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 361 error = vfs_allocate_syncvnode(mp); 362 vfs_unbusy(mp); 363 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 364 error = VFS_START(mp, 0, l); 365 if (error) 366 vrele(vp); 367 } else { 368 vp->v_mountedhere = NULL; 369 vfs_unbusy(mp); 370 vfs_delref(mp->mnt_op); 371 free(mp, M_MOUNT); 372 vput(vp); 373 } 374 375 out: 376 return (error); 377 } 378 379 static int 380 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 381 void *data, struct nameidata *ndp) 382 { 383 struct mount *mp; 384 int error; 385 386 /* If MNT_GETARGS is specified, it should be the only flag. */ 387 if (flags & ~MNT_GETARGS) { 388 error = EINVAL; 389 goto out; 390 } 391 392 mp = vp->v_mount; 393 394 /* XXX: probably some notion of "can see" here if we want isolation. */ 395 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 396 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 397 if (error) 398 goto out; 399 400 if ((vp->v_vflag & VV_ROOT) == 0) { 401 error = EINVAL; 402 goto out; 403 } 404 405 if (vfs_busy(mp, LK_NOWAIT, 0)) { 406 error = EPERM; 407 goto out; 408 } 409 410 mp->mnt_flag &= ~MNT_OP_FLAGS; 411 mp->mnt_flag |= MNT_GETARGS; 412 error = VFS_MOUNT(mp, path, data, ndp, l); 413 mp->mnt_flag &= ~MNT_OP_FLAGS; 414 415 vfs_unbusy(mp); 416 out: 417 return (error); 418 } 419 420 /* ARGSUSED */ 421 int 422 sys_mount(struct lwp *l, void *v, register_t *retval) 423 { 424 struct sys_mount_args /* { 425 syscallarg(const char *) type; 426 syscallarg(const char *) path; 427 syscallarg(int) flags; 428 syscallarg(void *) data; 429 } */ *uap = v; 430 struct vnode *vp; 431 struct nameidata nd; 432 int error; 433 434 /* 435 * Get vnode to be covered 436 */ 437 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, 438 SCARG(uap, path), l); 439 if ((error = namei(&nd)) != 0) 440 return (error); 441 vp = nd.ni_vp; 442 443 /* 444 * A lookup in VFS_MOUNT might result in an attempt to 445 * lock this vnode again, so make the lock recursive. 446 */ 447 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 448 449 if (SCARG(uap, flags) & MNT_GETARGS) { 450 error = mount_getargs(l, vp, SCARG(uap, path), 451 SCARG(uap, flags), SCARG(uap, data), &nd); 452 vput(vp); 453 } else if (SCARG(uap, flags) & MNT_UPDATE) { 454 error = mount_update(l, vp, SCARG(uap, path), 455 SCARG(uap, flags), SCARG(uap, data), &nd); 456 vput(vp); 457 } else { 458 /* Locking is handled internally in mount_domount(). */ 459 error = mount_domount(l, vp, SCARG(uap, type), 460 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 461 } 462 463 return (error); 464 } 465 466 /* 467 * Scan all active processes to see if any of them have a current 468 * or root directory onto which the new filesystem has just been 469 * mounted. If so, replace them with the new mount point. 470 */ 471 void 472 checkdirs(struct vnode *olddp) 473 { 474 struct cwdinfo *cwdi; 475 struct vnode *newdp; 476 struct proc *p; 477 478 if (olddp->v_usecount == 1) 479 return; 480 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 481 panic("mount: lost mount"); 482 mutex_enter(&proclist_lock); 483 PROCLIST_FOREACH(p, &allproc) { 484 cwdi = p->p_cwdi; 485 if (!cwdi) 486 continue; 487 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 488 if (cwdi->cwdi_cdir == olddp) { 489 vrele(cwdi->cwdi_cdir); 490 VREF(newdp); 491 cwdi->cwdi_cdir = newdp; 492 } 493 if (cwdi->cwdi_rdir == olddp) { 494 vrele(cwdi->cwdi_rdir); 495 VREF(newdp); 496 cwdi->cwdi_rdir = newdp; 497 } 498 rw_exit(&cwdi->cwdi_lock); 499 } 500 mutex_exit(&proclist_lock); 501 if (rootvnode == olddp) { 502 vrele(rootvnode); 503 VREF(newdp); 504 rootvnode = newdp; 505 } 506 vput(newdp); 507 } 508 509 /* 510 * Unmount a file system. 511 * 512 * Note: unmount takes a path to the vnode mounted on as argument, 513 * not special file (as before). 514 */ 515 /* ARGSUSED */ 516 int 517 sys_unmount(struct lwp *l, void *v, register_t *retval) 518 { 519 struct sys_unmount_args /* { 520 syscallarg(const char *) path; 521 syscallarg(int) flags; 522 } */ *uap = v; 523 struct vnode *vp; 524 struct mount *mp; 525 int error; 526 struct nameidata nd; 527 528 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 529 SCARG(uap, path), l); 530 if ((error = namei(&nd)) != 0) 531 return (error); 532 vp = nd.ni_vp; 533 mp = vp->v_mount; 534 535 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 536 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 537 if (error) { 538 vput(vp); 539 return (error); 540 } 541 542 /* 543 * Don't allow unmounting the root file system. 544 */ 545 if (mp->mnt_flag & MNT_ROOTFS) { 546 vput(vp); 547 return (EINVAL); 548 } 549 550 /* 551 * Must be the root of the filesystem 552 */ 553 if ((vp->v_vflag & VV_ROOT) == 0) { 554 vput(vp); 555 return (EINVAL); 556 } 557 vput(vp); 558 559 /* 560 * XXX Freeze syncer. Must do this before locking the 561 * mount point. See dounmount() for details. 562 */ 563 mutex_enter(&syncer_mutex); 564 565 if (vfs_busy(mp, 0, 0)) { 566 mutex_exit(&syncer_mutex); 567 return (EBUSY); 568 } 569 570 return (dounmount(mp, SCARG(uap, flags), l)); 571 } 572 573 /* 574 * Do the actual file system unmount. File system is assumed to have been 575 * marked busy by the caller. 576 */ 577 int 578 dounmount(struct mount *mp, int flags, struct lwp *l) 579 { 580 struct vnode *coveredvp; 581 int error; 582 int async; 583 int used_syncer; 584 585 #if NVERIEXEC > 0 586 error = veriexec_unmountchk(mp); 587 if (error) 588 return (error); 589 #endif /* NVERIEXEC > 0 */ 590 591 mutex_enter(&mountlist_lock); 592 vfs_unbusy(mp); 593 used_syncer = (mp->mnt_syncer != NULL); 594 595 /* 596 * XXX Syncer must be frozen when we get here. This should really 597 * be done on a per-mountpoint basis, but especially the softdep 598 * code possibly called from the syncer doesn't exactly work on a 599 * per-mountpoint basis, so the softdep code would become a maze 600 * of vfs_busy() calls. 601 * 602 * The caller of dounmount() must acquire syncer_mutex because 603 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 604 * order, and we must preserve that order to avoid deadlock. 605 * 606 * So, if the file system did not use the syncer, now is 607 * the time to release the syncer_mutex. 608 */ 609 if (used_syncer == 0) 610 mutex_exit(&syncer_mutex); 611 612 mp->mnt_iflag |= IMNT_UNMOUNT; 613 mp->mnt_unmounter = l; 614 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 615 616 async = mp->mnt_flag & MNT_ASYNC; 617 mp->mnt_flag &= ~MNT_ASYNC; 618 cache_purgevfs(mp); /* remove cache entries for this file sys */ 619 if (mp->mnt_syncer != NULL) 620 vfs_deallocate_syncvnode(mp); 621 error = 0; 622 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 623 #if NFSS > 0 624 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 625 #endif 626 if (error == 0) 627 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 628 } 629 if (error == 0 || (flags & MNT_FORCE)) 630 error = VFS_UNMOUNT(mp, flags, l); 631 if (error) { 632 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 633 (void) vfs_allocate_syncvnode(mp); 634 mutex_enter(&mountlist_lock); 635 mp->mnt_iflag &= ~IMNT_UNMOUNT; 636 mp->mnt_unmounter = NULL; 637 mp->mnt_flag |= async; 638 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 639 &mountlist_lock); 640 if (used_syncer) 641 mutex_exit(&syncer_mutex); 642 mutex_enter(&mp->mnt_mutex); 643 while (mp->mnt_wcnt > 0) { 644 wakeup(mp); 645 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 646 0, &mp->mnt_mutex); 647 } 648 mutex_exit(&mp->mnt_mutex); 649 return (error); 650 } 651 mutex_enter(&mountlist_lock); 652 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 653 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 654 coveredvp->v_mountedhere = NULL; 655 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 656 panic("unmount: dangling vnode"); 657 mp->mnt_iflag |= IMNT_GONE; 658 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 659 if (coveredvp != NULLVP) 660 vrele(coveredvp); 661 mount_finispecific(mp); 662 if (used_syncer) 663 mutex_exit(&syncer_mutex); 664 mutex_enter(&mp->mnt_mutex); 665 while (mp->mnt_wcnt > 0) { 666 wakeup(mp); 667 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 668 } 669 mutex_exit(&mp->mnt_mutex); 670 vfs_hooks_unmount(mp); 671 mutex_destroy(&mp->mnt_mutex); 672 vfs_delref(mp->mnt_op); 673 free(mp, M_MOUNT); 674 return (0); 675 } 676 677 /* 678 * Sync each mounted filesystem. 679 */ 680 #ifdef DEBUG 681 int syncprt = 0; 682 struct ctldebug debug0 = { "syncprt", &syncprt }; 683 #endif 684 685 /* ARGSUSED */ 686 int 687 sys_sync(struct lwp *l, void *v, register_t *retval) 688 { 689 struct mount *mp, *nmp; 690 int asyncflag; 691 692 if (l == NULL) 693 l = &lwp0; 694 695 mutex_enter(&mountlist_lock); 696 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 697 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 698 nmp = mp->mnt_list.cqe_prev; 699 continue; 700 } 701 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 702 asyncflag = mp->mnt_flag & MNT_ASYNC; 703 mp->mnt_flag &= ~MNT_ASYNC; 704 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 705 if (asyncflag) 706 mp->mnt_flag |= MNT_ASYNC; 707 } 708 mutex_enter(&mountlist_lock); 709 nmp = mp->mnt_list.cqe_prev; 710 vfs_unbusy(mp); 711 712 } 713 mutex_exit(&mountlist_lock); 714 #ifdef DEBUG 715 if (syncprt) 716 vfs_bufstats(); 717 #endif /* DEBUG */ 718 return (0); 719 } 720 721 /* 722 * Change filesystem quotas. 723 */ 724 /* ARGSUSED */ 725 int 726 sys_quotactl(struct lwp *l, void *v, register_t *retval) 727 { 728 struct sys_quotactl_args /* { 729 syscallarg(const char *) path; 730 syscallarg(int) cmd; 731 syscallarg(int) uid; 732 syscallarg(void *) arg; 733 } */ *uap = v; 734 struct mount *mp; 735 int error; 736 struct nameidata nd; 737 738 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 739 if ((error = namei(&nd)) != 0) 740 return (error); 741 mp = nd.ni_vp->v_mount; 742 vrele(nd.ni_vp); 743 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 744 SCARG(uap, arg), l); 745 return (error); 746 } 747 748 int 749 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 750 int root) 751 { 752 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 753 int error = 0; 754 755 /* 756 * If MNT_NOWAIT or MNT_LAZY is specified, do not 757 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 758 * overrides MNT_NOWAIT. 759 */ 760 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 761 (flags != MNT_WAIT && flags != 0)) { 762 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 763 goto done; 764 } 765 766 /* Get the filesystem stats now */ 767 memset(sp, 0, sizeof(*sp)); 768 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 769 return error; 770 } 771 772 if (cwdi->cwdi_rdir == NULL) 773 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 774 done: 775 if (cwdi->cwdi_rdir != NULL) { 776 size_t len; 777 char *bp; 778 char *path = PNBUF_GET(); 779 780 bp = path + MAXPATHLEN; 781 *--bp = '\0'; 782 rw_enter(&cwdi->cwdi_lock, RW_READER); 783 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 784 MAXPATHLEN / 2, 0, l); 785 rw_exit(&cwdi->cwdi_lock); 786 if (error) { 787 PNBUF_PUT(path); 788 return error; 789 } 790 len = strlen(bp); 791 /* 792 * for mount points that are below our root, we can see 793 * them, so we fix up the pathname and return them. The 794 * rest we cannot see, so we don't allow viewing the 795 * data. 796 */ 797 if (strncmp(bp, sp->f_mntonname, len) == 0) { 798 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 799 sizeof(sp->f_mntonname)); 800 if (sp->f_mntonname[0] == '\0') 801 (void)strlcpy(sp->f_mntonname, "/", 802 sizeof(sp->f_mntonname)); 803 } else { 804 if (root) 805 (void)strlcpy(sp->f_mntonname, "/", 806 sizeof(sp->f_mntonname)); 807 else 808 error = EPERM; 809 } 810 PNBUF_PUT(path); 811 } 812 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 813 return error; 814 } 815 816 /* 817 * Get filesystem statistics by path. 818 */ 819 int 820 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 821 { 822 struct mount *mp; 823 int error; 824 struct nameidata nd; 825 826 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, path, l); 827 if ((error = namei(&nd)) != 0) 828 return error; 829 mp = nd.ni_vp->v_mount; 830 error = dostatvfs(mp, sb, l, flags, 1); 831 vrele(nd.ni_vp); 832 return error; 833 } 834 835 /* ARGSUSED */ 836 int 837 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 838 { 839 struct sys_statvfs1_args /* { 840 syscallarg(const char *) path; 841 syscallarg(struct statvfs *) buf; 842 syscallarg(int) flags; 843 } */ *uap = v; 844 struct statvfs *sb; 845 int error; 846 847 sb = STATVFSBUF_GET(); 848 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 849 if (error == 0) 850 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 851 STATVFSBUF_PUT(sb); 852 return error; 853 } 854 855 /* 856 * Get filesystem statistics by fd. 857 */ 858 int 859 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 860 { 861 struct proc *p = l->l_proc; 862 struct file *fp; 863 struct mount *mp; 864 int error; 865 866 /* getvnode() will use the descriptor for us */ 867 if ((error = getvnode(p->p_fd, fd, &fp)) != 0) 868 return (error); 869 mp = ((struct vnode *)fp->f_data)->v_mount; 870 error = dostatvfs(mp, sb, l, flags, 1); 871 FILE_UNUSE(fp, l); 872 return error; 873 } 874 875 /* ARGSUSED */ 876 int 877 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 878 { 879 struct sys_fstatvfs1_args /* { 880 syscallarg(int) fd; 881 syscallarg(struct statvfs *) buf; 882 syscallarg(int) flags; 883 } */ *uap = v; 884 struct statvfs *sb; 885 int error; 886 887 sb = STATVFSBUF_GET(); 888 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 889 if (error == 0) 890 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 891 STATVFSBUF_PUT(sb); 892 return error; 893 } 894 895 896 /* 897 * Get statistics on all filesystems. 898 */ 899 int 900 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 901 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 902 register_t *retval) 903 { 904 int root = 0; 905 struct proc *p = l->l_proc; 906 struct mount *mp, *nmp; 907 struct statvfs *sb; 908 size_t count, maxcount; 909 int error = 0; 910 911 sb = STATVFSBUF_GET(); 912 maxcount = bufsize / entry_sz; 913 mutex_enter(&mountlist_lock); 914 count = 0; 915 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 916 mp = nmp) { 917 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 918 nmp = CIRCLEQ_NEXT(mp, mnt_list); 919 continue; 920 } 921 if (sfsp && count < maxcount) { 922 error = dostatvfs(mp, sb, l, flags, 0); 923 if (error) { 924 mutex_enter(&mountlist_lock); 925 nmp = CIRCLEQ_NEXT(mp, mnt_list); 926 vfs_unbusy(mp); 927 continue; 928 } 929 error = copyfn(sb, sfsp, entry_sz); 930 if (error) { 931 vfs_unbusy(mp); 932 goto out; 933 } 934 sfsp = (char *)sfsp + entry_sz; 935 root |= strcmp(sb->f_mntonname, "/") == 0; 936 } 937 count++; 938 mutex_enter(&mountlist_lock); 939 nmp = CIRCLEQ_NEXT(mp, mnt_list); 940 vfs_unbusy(mp); 941 } 942 mutex_exit(&mountlist_lock); 943 if (root == 0 && p->p_cwdi->cwdi_rdir) { 944 /* 945 * fake a root entry 946 */ 947 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, flags, 1); 948 if (error != 0) 949 goto out; 950 if (sfsp) 951 error = copyfn(sb, sfsp, entry_sz); 952 count++; 953 } 954 if (sfsp && count > maxcount) 955 *retval = maxcount; 956 else 957 *retval = count; 958 out: 959 STATVFSBUF_PUT(sb); 960 return error; 961 } 962 963 int 964 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 965 { 966 struct sys_getvfsstat_args /* { 967 syscallarg(struct statvfs *) buf; 968 syscallarg(size_t) bufsize; 969 syscallarg(int) flags; 970 } */ *uap = v; 971 972 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 973 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 974 } 975 976 /* 977 * Change current working directory to a given file descriptor. 978 */ 979 /* ARGSUSED */ 980 int 981 sys_fchdir(struct lwp *l, void *v, register_t *retval) 982 { 983 struct sys_fchdir_args /* { 984 syscallarg(int) fd; 985 } */ *uap = v; 986 struct proc *p = l->l_proc; 987 struct filedesc *fdp = p->p_fd; 988 struct cwdinfo *cwdi; 989 struct vnode *vp, *tdp; 990 struct mount *mp; 991 struct file *fp; 992 int error; 993 994 /* getvnode() will use the descriptor for us */ 995 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 996 return (error); 997 vp = (struct vnode *)fp->f_data; 998 999 VREF(vp); 1000 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1001 if (vp->v_type != VDIR) 1002 error = ENOTDIR; 1003 else 1004 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1005 if (error) { 1006 vput(vp); 1007 goto out; 1008 } 1009 while ((mp = vp->v_mountedhere) != NULL) { 1010 if (vfs_busy(mp, 0, 0)) 1011 continue; 1012 1013 vput(vp); 1014 error = VFS_ROOT(mp, &tdp); 1015 vfs_unbusy(mp); 1016 if (error) 1017 goto out; 1018 vp = tdp; 1019 } 1020 VOP_UNLOCK(vp, 0); 1021 1022 /* 1023 * Disallow changing to a directory not under the process's 1024 * current root directory (if there is one). 1025 */ 1026 cwdi = p->p_cwdi; 1027 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1028 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1029 vrele(vp); 1030 error = EPERM; /* operation not permitted */ 1031 } else { 1032 vrele(cwdi->cwdi_cdir); 1033 cwdi->cwdi_cdir = vp; 1034 } 1035 rw_exit(&cwdi->cwdi_lock); 1036 1037 out: 1038 FILE_UNUSE(fp, l); 1039 return (error); 1040 } 1041 1042 /* 1043 * Change this process's notion of the root directory to a given file 1044 * descriptor. 1045 */ 1046 int 1047 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1048 { 1049 struct sys_fchroot_args *uap = v; 1050 struct proc *p = l->l_proc; 1051 struct filedesc *fdp = p->p_fd; 1052 struct cwdinfo *cwdi; 1053 struct vnode *vp; 1054 struct file *fp; 1055 int error; 1056 1057 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1058 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1059 return error; 1060 /* getvnode() will use the descriptor for us */ 1061 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1062 return error; 1063 vp = (struct vnode *) fp->f_data; 1064 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1065 if (vp->v_type != VDIR) 1066 error = ENOTDIR; 1067 else 1068 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1069 VOP_UNLOCK(vp, 0); 1070 if (error) 1071 goto out; 1072 VREF(vp); 1073 1074 /* 1075 * Prevent escaping from chroot by putting the root under 1076 * the working directory. Silently chdir to / if we aren't 1077 * already there. 1078 */ 1079 cwdi = p->p_cwdi; 1080 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1081 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1082 /* 1083 * XXX would be more failsafe to change directory to a 1084 * deadfs node here instead 1085 */ 1086 vrele(cwdi->cwdi_cdir); 1087 VREF(vp); 1088 cwdi->cwdi_cdir = vp; 1089 } 1090 1091 if (cwdi->cwdi_rdir != NULL) 1092 vrele(cwdi->cwdi_rdir); 1093 cwdi->cwdi_rdir = vp; 1094 rw_exit(&cwdi->cwdi_lock); 1095 1096 out: 1097 FILE_UNUSE(fp, l); 1098 return (error); 1099 } 1100 1101 /* 1102 * Change current working directory (``.''). 1103 */ 1104 /* ARGSUSED */ 1105 int 1106 sys_chdir(struct lwp *l, void *v, register_t *retval) 1107 { 1108 struct sys_chdir_args /* { 1109 syscallarg(const char *) path; 1110 } */ *uap = v; 1111 struct proc *p = l->l_proc; 1112 struct cwdinfo *cwdi; 1113 int error; 1114 struct nameidata nd; 1115 1116 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1117 SCARG(uap, path), l); 1118 if ((error = change_dir(&nd, l)) != 0) 1119 return (error); 1120 cwdi = p->p_cwdi; 1121 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1122 vrele(cwdi->cwdi_cdir); 1123 cwdi->cwdi_cdir = nd.ni_vp; 1124 rw_exit(&cwdi->cwdi_lock); 1125 return (0); 1126 } 1127 1128 /* 1129 * Change notion of root (``/'') directory. 1130 */ 1131 /* ARGSUSED */ 1132 int 1133 sys_chroot(struct lwp *l, void *v, register_t *retval) 1134 { 1135 struct sys_chroot_args /* { 1136 syscallarg(const char *) path; 1137 } */ *uap = v; 1138 struct proc *p = l->l_proc; 1139 struct cwdinfo *cwdi; 1140 struct vnode *vp; 1141 int error; 1142 struct nameidata nd; 1143 1144 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1145 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1146 return (error); 1147 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1148 SCARG(uap, path), l); 1149 if ((error = change_dir(&nd, l)) != 0) 1150 return (error); 1151 1152 cwdi = p->p_cwdi; 1153 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1154 if (cwdi->cwdi_rdir != NULL) 1155 vrele(cwdi->cwdi_rdir); 1156 vp = nd.ni_vp; 1157 cwdi->cwdi_rdir = vp; 1158 1159 /* 1160 * Prevent escaping from chroot by putting the root under 1161 * the working directory. Silently chdir to / if we aren't 1162 * already there. 1163 */ 1164 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1165 /* 1166 * XXX would be more failsafe to change directory to a 1167 * deadfs node here instead 1168 */ 1169 vrele(cwdi->cwdi_cdir); 1170 VREF(vp); 1171 cwdi->cwdi_cdir = vp; 1172 } 1173 rw_exit(&cwdi->cwdi_lock); 1174 1175 return (0); 1176 } 1177 1178 /* 1179 * Common routine for chroot and chdir. 1180 */ 1181 static int 1182 change_dir(struct nameidata *ndp, struct lwp *l) 1183 { 1184 struct vnode *vp; 1185 int error; 1186 1187 if ((error = namei(ndp)) != 0) 1188 return (error); 1189 vp = ndp->ni_vp; 1190 if (vp->v_type != VDIR) 1191 error = ENOTDIR; 1192 else 1193 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1194 1195 if (error) 1196 vput(vp); 1197 else 1198 VOP_UNLOCK(vp, 0); 1199 return (error); 1200 } 1201 1202 /* 1203 * Check permissions, allocate an open file structure, 1204 * and call the device open routine if any. 1205 */ 1206 int 1207 sys_open(struct lwp *l, void *v, register_t *retval) 1208 { 1209 struct sys_open_args /* { 1210 syscallarg(const char *) path; 1211 syscallarg(int) flags; 1212 syscallarg(int) mode; 1213 } */ *uap = v; 1214 struct proc *p = l->l_proc; 1215 struct cwdinfo *cwdi = p->p_cwdi; 1216 struct filedesc *fdp = p->p_fd; 1217 struct file *fp; 1218 struct vnode *vp; 1219 int flags, cmode; 1220 int type, indx, error; 1221 struct flock lf; 1222 struct nameidata nd; 1223 1224 flags = FFLAGS(SCARG(uap, flags)); 1225 if ((flags & (FREAD | FWRITE)) == 0) 1226 return (EINVAL); 1227 /* falloc() will use the file descriptor for us */ 1228 if ((error = falloc(l, &fp, &indx)) != 0) 1229 return (error); 1230 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1231 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1232 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1233 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1234 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1235 rw_enter(&fdp->fd_lock, RW_WRITER); 1236 FILE_UNUSE(fp, l); 1237 fdp->fd_ofiles[indx] = NULL; 1238 rw_exit(&fdp->fd_lock); 1239 ffree(fp); 1240 if ((error == EDUPFD || error == EMOVEFD) && 1241 l->l_dupfd >= 0 && /* XXX from fdopen */ 1242 (error = 1243 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1244 *retval = indx; 1245 return (0); 1246 } 1247 if (error == ERESTART) 1248 error = EINTR; 1249 fdremove(fdp, indx); 1250 return (error); 1251 } 1252 l->l_dupfd = 0; 1253 vp = nd.ni_vp; 1254 fp->f_flag = flags & FMASK; 1255 fp->f_type = DTYPE_VNODE; 1256 fp->f_ops = &vnops; 1257 fp->f_data = vp; 1258 if (flags & (O_EXLOCK | O_SHLOCK)) { 1259 lf.l_whence = SEEK_SET; 1260 lf.l_start = 0; 1261 lf.l_len = 0; 1262 if (flags & O_EXLOCK) 1263 lf.l_type = F_WRLCK; 1264 else 1265 lf.l_type = F_RDLCK; 1266 type = F_FLOCK; 1267 if ((flags & FNONBLOCK) == 0) 1268 type |= F_WAIT; 1269 VOP_UNLOCK(vp, 0); 1270 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1271 if (error) { 1272 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1273 FILE_UNUSE(fp, l); 1274 ffree(fp); 1275 fdremove(fdp, indx); 1276 return (error); 1277 } 1278 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1279 fp->f_flag |= FHASLOCK; 1280 } 1281 VOP_UNLOCK(vp, 0); 1282 *retval = indx; 1283 FILE_SET_MATURE(fp); 1284 FILE_UNUSE(fp, l); 1285 return (0); 1286 } 1287 1288 static void 1289 vfs__fhfree(fhandle_t *fhp) 1290 { 1291 size_t fhsize; 1292 1293 if (fhp == NULL) { 1294 return; 1295 } 1296 fhsize = FHANDLE_SIZE(fhp); 1297 kmem_free(fhp, fhsize); 1298 } 1299 1300 /* 1301 * vfs_composefh: compose a filehandle. 1302 */ 1303 1304 int 1305 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1306 { 1307 struct mount *mp; 1308 struct fid *fidp; 1309 int error; 1310 size_t needfhsize; 1311 size_t fidsize; 1312 1313 mp = vp->v_mount; 1314 fidp = NULL; 1315 if (*fh_size < FHANDLE_SIZE_MIN) { 1316 fidsize = 0; 1317 } else { 1318 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1319 if (fhp != NULL) { 1320 memset(fhp, 0, *fh_size); 1321 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1322 fidp = &fhp->fh_fid; 1323 } 1324 } 1325 error = VFS_VPTOFH(vp, fidp, &fidsize); 1326 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1327 if (error == 0 && *fh_size < needfhsize) { 1328 error = E2BIG; 1329 } 1330 *fh_size = needfhsize; 1331 return error; 1332 } 1333 1334 int 1335 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1336 { 1337 struct mount *mp; 1338 fhandle_t *fhp; 1339 size_t fhsize; 1340 size_t fidsize; 1341 int error; 1342 1343 *fhpp = NULL; 1344 mp = vp->v_mount; 1345 fidsize = 0; 1346 error = VFS_VPTOFH(vp, NULL, &fidsize); 1347 KASSERT(error != 0); 1348 if (error != E2BIG) { 1349 goto out; 1350 } 1351 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1352 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1353 if (fhp == NULL) { 1354 error = ENOMEM; 1355 goto out; 1356 } 1357 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1358 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1359 if (error == 0) { 1360 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1361 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1362 *fhpp = fhp; 1363 } else { 1364 kmem_free(fhp, fhsize); 1365 } 1366 out: 1367 return error; 1368 } 1369 1370 void 1371 vfs_composefh_free(fhandle_t *fhp) 1372 { 1373 1374 vfs__fhfree(fhp); 1375 } 1376 1377 /* 1378 * vfs_fhtovp: lookup a vnode by a filehandle. 1379 */ 1380 1381 int 1382 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1383 { 1384 struct mount *mp; 1385 int error; 1386 1387 *vpp = NULL; 1388 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1389 if (mp == NULL) { 1390 error = ESTALE; 1391 goto out; 1392 } 1393 if (mp->mnt_op->vfs_fhtovp == NULL) { 1394 error = EOPNOTSUPP; 1395 goto out; 1396 } 1397 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1398 out: 1399 return error; 1400 } 1401 1402 /* 1403 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1404 * the needed size. 1405 */ 1406 1407 int 1408 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1409 { 1410 fhandle_t *fhp; 1411 int error; 1412 1413 *fhpp = NULL; 1414 if (fhsize > FHANDLE_SIZE_MAX) { 1415 return EINVAL; 1416 } 1417 if (fhsize < FHANDLE_SIZE_MIN) { 1418 return EINVAL; 1419 } 1420 again: 1421 fhp = kmem_alloc(fhsize, KM_SLEEP); 1422 if (fhp == NULL) { 1423 return ENOMEM; 1424 } 1425 error = copyin(ufhp, fhp, fhsize); 1426 if (error == 0) { 1427 /* XXX this check shouldn't be here */ 1428 if (FHANDLE_SIZE(fhp) == fhsize) { 1429 *fhpp = fhp; 1430 return 0; 1431 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1432 /* 1433 * a kludge for nfsv2 padded handles. 1434 */ 1435 size_t sz; 1436 1437 sz = FHANDLE_SIZE(fhp); 1438 kmem_free(fhp, fhsize); 1439 fhsize = sz; 1440 goto again; 1441 } else { 1442 /* 1443 * userland told us wrong size. 1444 */ 1445 error = EINVAL; 1446 } 1447 } 1448 kmem_free(fhp, fhsize); 1449 return error; 1450 } 1451 1452 void 1453 vfs_copyinfh_free(fhandle_t *fhp) 1454 { 1455 1456 vfs__fhfree(fhp); 1457 } 1458 1459 /* 1460 * Get file handle system call 1461 */ 1462 int 1463 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1464 { 1465 struct sys___getfh30_args /* { 1466 syscallarg(char *) fname; 1467 syscallarg(fhandle_t *) fhp; 1468 syscallarg(size_t *) fh_size; 1469 } */ *uap = v; 1470 struct vnode *vp; 1471 fhandle_t *fh; 1472 int error; 1473 struct nameidata nd; 1474 size_t sz; 1475 size_t usz; 1476 1477 /* 1478 * Must be super user 1479 */ 1480 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1481 0, NULL, NULL, NULL); 1482 if (error) 1483 return (error); 1484 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 1485 SCARG(uap, fname), l); 1486 error = namei(&nd); 1487 if (error) 1488 return (error); 1489 vp = nd.ni_vp; 1490 error = vfs_composefh_alloc(vp, &fh); 1491 vput(vp); 1492 if (error != 0) { 1493 goto out; 1494 } 1495 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1496 if (error != 0) { 1497 goto out; 1498 } 1499 sz = FHANDLE_SIZE(fh); 1500 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1501 if (error != 0) { 1502 goto out; 1503 } 1504 if (usz >= sz) { 1505 error = copyout(fh, SCARG(uap, fhp), sz); 1506 } else { 1507 error = E2BIG; 1508 } 1509 out: 1510 vfs_composefh_free(fh); 1511 return (error); 1512 } 1513 1514 /* 1515 * Open a file given a file handle. 1516 * 1517 * Check permissions, allocate an open file structure, 1518 * and call the device open routine if any. 1519 */ 1520 1521 int 1522 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1523 register_t *retval) 1524 { 1525 struct filedesc *fdp = l->l_proc->p_fd; 1526 struct file *fp; 1527 struct vnode *vp = NULL; 1528 kauth_cred_t cred = l->l_cred; 1529 struct file *nfp; 1530 int type, indx, error=0; 1531 struct flock lf; 1532 struct vattr va; 1533 fhandle_t *fh; 1534 int flags; 1535 1536 /* 1537 * Must be super user 1538 */ 1539 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1540 0, NULL, NULL, NULL))) 1541 return (error); 1542 1543 flags = FFLAGS(oflags); 1544 if ((flags & (FREAD | FWRITE)) == 0) 1545 return (EINVAL); 1546 if ((flags & O_CREAT)) 1547 return (EINVAL); 1548 /* falloc() will use the file descriptor for us */ 1549 if ((error = falloc(l, &nfp, &indx)) != 0) 1550 return (error); 1551 fp = nfp; 1552 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1553 if (error != 0) { 1554 goto bad; 1555 } 1556 error = vfs_fhtovp(fh, &vp); 1557 if (error != 0) { 1558 goto bad; 1559 } 1560 1561 /* Now do an effective vn_open */ 1562 1563 if (vp->v_type == VSOCK) { 1564 error = EOPNOTSUPP; 1565 goto bad; 1566 } 1567 if (flags & FREAD) { 1568 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1569 goto bad; 1570 } 1571 if (flags & (FWRITE | O_TRUNC)) { 1572 if (vp->v_type == VDIR) { 1573 error = EISDIR; 1574 goto bad; 1575 } 1576 if ((error = vn_writechk(vp)) != 0 || 1577 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1578 goto bad; 1579 } 1580 if (flags & O_TRUNC) { 1581 VOP_UNLOCK(vp, 0); /* XXX */ 1582 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1583 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1584 VATTR_NULL(&va); 1585 va.va_size = 0; 1586 error = VOP_SETATTR(vp, &va, cred, l); 1587 if (error) 1588 goto bad; 1589 } 1590 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1591 goto bad; 1592 if (vp->v_type == VREG && 1593 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1594 error = EIO; 1595 goto bad; 1596 } 1597 if (flags & FWRITE) 1598 vp->v_writecount++; 1599 1600 /* done with modified vn_open, now finish what sys_open does. */ 1601 1602 fp->f_flag = flags & FMASK; 1603 fp->f_type = DTYPE_VNODE; 1604 fp->f_ops = &vnops; 1605 fp->f_data = vp; 1606 if (flags & (O_EXLOCK | O_SHLOCK)) { 1607 lf.l_whence = SEEK_SET; 1608 lf.l_start = 0; 1609 lf.l_len = 0; 1610 if (flags & O_EXLOCK) 1611 lf.l_type = F_WRLCK; 1612 else 1613 lf.l_type = F_RDLCK; 1614 type = F_FLOCK; 1615 if ((flags & FNONBLOCK) == 0) 1616 type |= F_WAIT; 1617 VOP_UNLOCK(vp, 0); 1618 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1619 if (error) { 1620 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1621 FILE_UNUSE(fp, l); 1622 ffree(fp); 1623 fdremove(fdp, indx); 1624 return (error); 1625 } 1626 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1627 fp->f_flag |= FHASLOCK; 1628 } 1629 VOP_UNLOCK(vp, 0); 1630 *retval = indx; 1631 FILE_SET_MATURE(fp); 1632 FILE_UNUSE(fp, l); 1633 vfs_copyinfh_free(fh); 1634 return (0); 1635 1636 bad: 1637 FILE_UNUSE(fp, l); 1638 ffree(fp); 1639 fdremove(fdp, indx); 1640 if (vp != NULL) 1641 vput(vp); 1642 vfs_copyinfh_free(fh); 1643 return (error); 1644 } 1645 1646 int 1647 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1648 { 1649 struct sys___fhopen40_args /* { 1650 syscallarg(const void *) fhp; 1651 syscallarg(size_t) fh_size; 1652 syscallarg(int) flags; 1653 } */ *uap = v; 1654 1655 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1656 SCARG(uap, flags), retval); 1657 } 1658 1659 int 1660 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1661 { 1662 int error; 1663 fhandle_t *fh; 1664 struct vnode *vp; 1665 1666 /* 1667 * Must be super user 1668 */ 1669 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1670 0, NULL, NULL, NULL))) 1671 return (error); 1672 1673 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1674 if (error != 0) 1675 return error; 1676 1677 error = vfs_fhtovp(fh, &vp); 1678 vfs_copyinfh_free(fh); 1679 if (error != 0) 1680 return error; 1681 1682 error = vn_stat(vp, sb, l); 1683 vput(vp); 1684 return error; 1685 } 1686 1687 1688 /* ARGSUSED */ 1689 int 1690 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1691 { 1692 struct sys___fhstat40_args /* { 1693 syscallarg(const void *) fhp; 1694 syscallarg(size_t) fh_size; 1695 syscallarg(struct stat *) sb; 1696 } */ *uap = v; 1697 struct stat sb; 1698 int error; 1699 1700 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1701 if (error) 1702 return error; 1703 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1704 } 1705 1706 int 1707 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1708 int flags) 1709 { 1710 fhandle_t *fh; 1711 struct mount *mp; 1712 struct vnode *vp; 1713 int error; 1714 1715 /* 1716 * Must be super user 1717 */ 1718 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1719 0, NULL, NULL, NULL))) 1720 return error; 1721 1722 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1723 if (error != 0) 1724 return error; 1725 1726 error = vfs_fhtovp(fh, &vp); 1727 vfs_copyinfh_free(fh); 1728 if (error != 0) 1729 return error; 1730 1731 mp = vp->v_mount; 1732 error = dostatvfs(mp, sb, l, flags, 1); 1733 vput(vp); 1734 return error; 1735 } 1736 1737 /* ARGSUSED */ 1738 int 1739 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1740 { 1741 struct sys___fhstatvfs140_args /* { 1742 syscallarg(const void *) fhp; 1743 syscallarg(size_t) fh_size; 1744 syscallarg(struct statvfs *) buf; 1745 syscallarg(int) flags; 1746 } */ *uap = v; 1747 struct statvfs *sb = STATVFSBUF_GET(); 1748 int error; 1749 1750 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1751 SCARG(uap, flags)); 1752 if (error == 0) 1753 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1754 STATVFSBUF_PUT(sb); 1755 return error; 1756 } 1757 1758 /* 1759 * Create a special file. 1760 */ 1761 /* ARGSUSED */ 1762 int 1763 sys_mknod(struct lwp *l, void *v, register_t *retval) 1764 { 1765 struct sys_mknod_args /* { 1766 syscallarg(const char *) path; 1767 syscallarg(int) mode; 1768 syscallarg(int) dev; 1769 } */ *uap = v; 1770 struct proc *p = l->l_proc; 1771 struct vnode *vp; 1772 struct vattr vattr; 1773 int error, optype; 1774 struct nameidata nd; 1775 char *path; 1776 const char *cpath; 1777 enum uio_seg seg = UIO_USERSPACE; 1778 1779 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1780 0, NULL, NULL, NULL)) != 0) 1781 return (error); 1782 1783 optype = VOP_MKNOD_DESCOFFSET; 1784 1785 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 1786 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, seg, cpath, l); 1787 1788 if ((error = namei(&nd)) != 0) 1789 goto out; 1790 vp = nd.ni_vp; 1791 if (vp != NULL) 1792 error = EEXIST; 1793 else { 1794 VATTR_NULL(&vattr); 1795 /* We will read cwdi->cwdi_cmask unlocked. */ 1796 vattr.va_mode = 1797 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1798 vattr.va_rdev = SCARG(uap, dev); 1799 1800 switch (SCARG(uap, mode) & S_IFMT) { 1801 case S_IFMT: /* used by badsect to flag bad sectors */ 1802 vattr.va_type = VBAD; 1803 break; 1804 case S_IFCHR: 1805 vattr.va_type = VCHR; 1806 break; 1807 case S_IFBLK: 1808 vattr.va_type = VBLK; 1809 break; 1810 case S_IFWHT: 1811 optype = VOP_WHITEOUT_DESCOFFSET; 1812 break; 1813 case S_IFREG: 1814 #if NVERIEXEC > 0 1815 error = veriexec_openchk(l, nd.ni_vp, nd.ni_dirp, 1816 O_CREAT); 1817 #endif /* NVERIEXEC > 0 */ 1818 vattr.va_type = VREG; 1819 vattr.va_rdev = VNOVAL; 1820 optype = VOP_CREATE_DESCOFFSET; 1821 break; 1822 default: 1823 error = EINVAL; 1824 break; 1825 } 1826 } 1827 if (!error) { 1828 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1829 switch (optype) { 1830 case VOP_WHITEOUT_DESCOFFSET: 1831 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1832 if (error) 1833 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1834 vput(nd.ni_dvp); 1835 break; 1836 1837 case VOP_MKNOD_DESCOFFSET: 1838 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1839 &nd.ni_cnd, &vattr); 1840 if (error == 0) 1841 vput(nd.ni_vp); 1842 break; 1843 1844 case VOP_CREATE_DESCOFFSET: 1845 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1846 &nd.ni_cnd, &vattr); 1847 if (error == 0) 1848 vput(nd.ni_vp); 1849 break; 1850 } 1851 } else { 1852 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1853 if (nd.ni_dvp == vp) 1854 vrele(nd.ni_dvp); 1855 else 1856 vput(nd.ni_dvp); 1857 if (vp) 1858 vrele(vp); 1859 } 1860 out: 1861 VERIEXEC_PATH_PUT(path); 1862 return (error); 1863 } 1864 1865 /* 1866 * Create a named pipe. 1867 */ 1868 /* ARGSUSED */ 1869 int 1870 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1871 { 1872 struct sys_mkfifo_args /* { 1873 syscallarg(const char *) path; 1874 syscallarg(int) mode; 1875 } */ *uap = v; 1876 struct proc *p = l->l_proc; 1877 struct vattr vattr; 1878 int error; 1879 struct nameidata nd; 1880 1881 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1882 if ((error = namei(&nd)) != 0) 1883 return (error); 1884 if (nd.ni_vp != NULL) { 1885 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1886 if (nd.ni_dvp == nd.ni_vp) 1887 vrele(nd.ni_dvp); 1888 else 1889 vput(nd.ni_dvp); 1890 vrele(nd.ni_vp); 1891 return (EEXIST); 1892 } 1893 VATTR_NULL(&vattr); 1894 vattr.va_type = VFIFO; 1895 /* We will read cwdi->cwdi_cmask unlocked. */ 1896 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1897 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1898 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1899 if (error == 0) 1900 vput(nd.ni_vp); 1901 return (error); 1902 } 1903 1904 /* 1905 * Make a hard file link. 1906 */ 1907 /* ARGSUSED */ 1908 int 1909 sys_link(struct lwp *l, void *v, register_t *retval) 1910 { 1911 struct sys_link_args /* { 1912 syscallarg(const char *) path; 1913 syscallarg(const char *) link; 1914 } */ *uap = v; 1915 struct vnode *vp; 1916 struct nameidata nd; 1917 int error; 1918 1919 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 1920 if ((error = namei(&nd)) != 0) 1921 return (error); 1922 vp = nd.ni_vp; 1923 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1924 if ((error = namei(&nd)) != 0) 1925 goto out; 1926 if (nd.ni_vp) { 1927 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1928 if (nd.ni_dvp == nd.ni_vp) 1929 vrele(nd.ni_dvp); 1930 else 1931 vput(nd.ni_dvp); 1932 vrele(nd.ni_vp); 1933 error = EEXIST; 1934 goto out; 1935 } 1936 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1937 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1938 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1939 out: 1940 vrele(vp); 1941 return (error); 1942 } 1943 1944 /* 1945 * Make a symbolic link. 1946 */ 1947 /* ARGSUSED */ 1948 int 1949 sys_symlink(struct lwp *l, void *v, register_t *retval) 1950 { 1951 struct sys_symlink_args /* { 1952 syscallarg(const char *) path; 1953 syscallarg(const char *) link; 1954 } */ *uap = v; 1955 struct proc *p = l->l_proc; 1956 struct vattr vattr; 1957 char *path; 1958 int error; 1959 struct nameidata nd; 1960 1961 path = PNBUF_GET(); 1962 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1963 if (error) 1964 goto out; 1965 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, link), l); 1966 if ((error = namei(&nd)) != 0) 1967 goto out; 1968 if (nd.ni_vp) { 1969 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1970 if (nd.ni_dvp == nd.ni_vp) 1971 vrele(nd.ni_dvp); 1972 else 1973 vput(nd.ni_dvp); 1974 vrele(nd.ni_vp); 1975 error = EEXIST; 1976 goto out; 1977 } 1978 VATTR_NULL(&vattr); 1979 vattr.va_type = VLNK; 1980 /* We will read cwdi->cwdi_cmask unlocked. */ 1981 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1982 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1983 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1984 if (error == 0) 1985 vput(nd.ni_vp); 1986 out: 1987 PNBUF_PUT(path); 1988 return (error); 1989 } 1990 1991 /* 1992 * Delete a whiteout from the filesystem. 1993 */ 1994 /* ARGSUSED */ 1995 int 1996 sys_undelete(struct lwp *l, void *v, register_t *retval) 1997 { 1998 struct sys_undelete_args /* { 1999 syscallarg(const char *) path; 2000 } */ *uap = v; 2001 int error; 2002 struct nameidata nd; 2003 2004 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, UIO_USERSPACE, 2005 SCARG(uap, path), l); 2006 error = namei(&nd); 2007 if (error) 2008 return (error); 2009 2010 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2011 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2012 if (nd.ni_dvp == nd.ni_vp) 2013 vrele(nd.ni_dvp); 2014 else 2015 vput(nd.ni_dvp); 2016 if (nd.ni_vp) 2017 vrele(nd.ni_vp); 2018 return (EEXIST); 2019 } 2020 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2021 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2022 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2023 vput(nd.ni_dvp); 2024 return (error); 2025 } 2026 2027 /* 2028 * Delete a name from the filesystem. 2029 */ 2030 /* ARGSUSED */ 2031 int 2032 sys_unlink(struct lwp *l, void *v, register_t *retval) 2033 { 2034 struct sys_unlink_args /* { 2035 syscallarg(const char *) path; 2036 } */ *uap = v; 2037 struct vnode *vp; 2038 int error; 2039 struct nameidata nd; 2040 char *path; 2041 const char *cpath; 2042 enum uio_seg seg = UIO_USERSPACE; 2043 2044 VERIEXEC_PATH_GET(SCARG(uap, path), seg, cpath, path); 2045 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, seg, cpath, l); 2046 2047 if ((error = namei(&nd)) != 0) 2048 goto out; 2049 vp = nd.ni_vp; 2050 2051 /* 2052 * The root of a mounted filesystem cannot be deleted. 2053 */ 2054 if (vp->v_vflag & VV_ROOT) { 2055 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2056 if (nd.ni_dvp == vp) 2057 vrele(nd.ni_dvp); 2058 else 2059 vput(nd.ni_dvp); 2060 vput(vp); 2061 error = EBUSY; 2062 goto out; 2063 } 2064 2065 #if NVERIEXEC > 0 2066 /* Handle remove requests for veriexec entries. */ 2067 if ((error = veriexec_removechk(l, nd.ni_vp, nd.ni_dirp)) != 0) { 2068 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2069 if (nd.ni_dvp == vp) 2070 vrele(nd.ni_dvp); 2071 else 2072 vput(nd.ni_dvp); 2073 vput(vp); 2074 goto out; 2075 } 2076 #endif /* NVERIEXEC > 0 */ 2077 2078 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2079 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2080 #ifdef FILEASSOC 2081 (void)fileassoc_file_delete(vp); 2082 #endif /* FILEASSOC */ 2083 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2084 out: 2085 VERIEXEC_PATH_PUT(path); 2086 return (error); 2087 } 2088 2089 /* 2090 * Reposition read/write file offset. 2091 */ 2092 int 2093 sys_lseek(struct lwp *l, void *v, register_t *retval) 2094 { 2095 struct sys_lseek_args /* { 2096 syscallarg(int) fd; 2097 syscallarg(int) pad; 2098 syscallarg(off_t) offset; 2099 syscallarg(int) whence; 2100 } */ *uap = v; 2101 struct proc *p = l->l_proc; 2102 kauth_cred_t cred = l->l_cred; 2103 struct filedesc *fdp = p->p_fd; 2104 struct file *fp; 2105 struct vnode *vp; 2106 struct vattr vattr; 2107 off_t newoff; 2108 int error; 2109 2110 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2111 return (EBADF); 2112 2113 vp = (struct vnode *)fp->f_data; 2114 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2115 error = ESPIPE; 2116 mutex_exit(&fp->f_lock); 2117 goto out; 2118 } 2119 2120 switch (SCARG(uap, whence)) { 2121 case SEEK_CUR: 2122 newoff = fp->f_offset + SCARG(uap, offset); 2123 FILE_USE(fp); 2124 break; 2125 case SEEK_END: 2126 FILE_USE(fp); 2127 error = VOP_GETATTR(vp, &vattr, cred, l); 2128 if (error) { 2129 FILE_UNUSE(fp, l); 2130 goto out; 2131 } 2132 newoff = SCARG(uap, offset) + vattr.va_size; 2133 break; 2134 case SEEK_SET: 2135 FILE_USE(fp); 2136 newoff = SCARG(uap, offset); 2137 break; 2138 default: 2139 mutex_exit(&fp->f_lock); 2140 error = EINVAL; 2141 goto out; 2142 } 2143 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2144 mutex_enter(&fp->f_lock); 2145 *(off_t *)retval = fp->f_offset = newoff; 2146 mutex_exit(&fp->f_lock); 2147 } 2148 FILE_UNUSE(fp, l); 2149 out: 2150 return (error); 2151 } 2152 2153 /* 2154 * Positional read system call. 2155 */ 2156 int 2157 sys_pread(struct lwp *l, void *v, register_t *retval) 2158 { 2159 struct sys_pread_args /* { 2160 syscallarg(int) fd; 2161 syscallarg(void *) buf; 2162 syscallarg(size_t) nbyte; 2163 syscallarg(off_t) offset; 2164 } */ *uap = v; 2165 struct proc *p = l->l_proc; 2166 struct filedesc *fdp = p->p_fd; 2167 struct file *fp; 2168 struct vnode *vp; 2169 off_t offset; 2170 int error, fd = SCARG(uap, fd); 2171 2172 if ((fp = fd_getfile(fdp, fd)) == NULL) 2173 return (EBADF); 2174 2175 if ((fp->f_flag & FREAD) == 0) { 2176 mutex_exit(&fp->f_lock); 2177 return (EBADF); 2178 } 2179 2180 FILE_USE(fp); 2181 2182 vp = (struct vnode *)fp->f_data; 2183 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2184 error = ESPIPE; 2185 goto out; 2186 } 2187 2188 offset = SCARG(uap, offset); 2189 2190 /* 2191 * XXX This works because no file systems actually 2192 * XXX take any action on the seek operation. 2193 */ 2194 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2195 goto out; 2196 2197 /* dofileread() will unuse the descriptor for us */ 2198 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2199 &offset, 0, retval)); 2200 2201 out: 2202 FILE_UNUSE(fp, l); 2203 return (error); 2204 } 2205 2206 /* 2207 * Positional scatter read system call. 2208 */ 2209 int 2210 sys_preadv(struct lwp *l, void *v, register_t *retval) 2211 { 2212 struct sys_preadv_args /* { 2213 syscallarg(int) fd; 2214 syscallarg(const struct iovec *) iovp; 2215 syscallarg(int) iovcnt; 2216 syscallarg(off_t) offset; 2217 } */ *uap = v; 2218 struct proc *p = l->l_proc; 2219 struct filedesc *fdp = p->p_fd; 2220 struct file *fp; 2221 struct vnode *vp; 2222 off_t offset; 2223 int error, fd = SCARG(uap, fd); 2224 2225 if ((fp = fd_getfile(fdp, fd)) == NULL) 2226 return (EBADF); 2227 2228 if ((fp->f_flag & FREAD) == 0) { 2229 mutex_exit(&fp->f_lock); 2230 return (EBADF); 2231 } 2232 2233 FILE_USE(fp); 2234 2235 vp = (struct vnode *)fp->f_data; 2236 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2237 error = ESPIPE; 2238 goto out; 2239 } 2240 2241 offset = SCARG(uap, offset); 2242 2243 /* 2244 * XXX This works because no file systems actually 2245 * XXX take any action on the seek operation. 2246 */ 2247 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2248 goto out; 2249 2250 /* dofilereadv() will unuse the descriptor for us */ 2251 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2252 &offset, 0, retval)); 2253 2254 out: 2255 FILE_UNUSE(fp, l); 2256 return (error); 2257 } 2258 2259 /* 2260 * Positional write system call. 2261 */ 2262 int 2263 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2264 { 2265 struct sys_pwrite_args /* { 2266 syscallarg(int) fd; 2267 syscallarg(const void *) buf; 2268 syscallarg(size_t) nbyte; 2269 syscallarg(off_t) offset; 2270 } */ *uap = v; 2271 struct proc *p = l->l_proc; 2272 struct filedesc *fdp = p->p_fd; 2273 struct file *fp; 2274 struct vnode *vp; 2275 off_t offset; 2276 int error, fd = SCARG(uap, fd); 2277 2278 if ((fp = fd_getfile(fdp, fd)) == NULL) 2279 return (EBADF); 2280 2281 if ((fp->f_flag & FWRITE) == 0) { 2282 mutex_exit(&fp->f_lock); 2283 return (EBADF); 2284 } 2285 2286 FILE_USE(fp); 2287 2288 vp = (struct vnode *)fp->f_data; 2289 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2290 error = ESPIPE; 2291 goto out; 2292 } 2293 2294 offset = SCARG(uap, offset); 2295 2296 /* 2297 * XXX This works because no file systems actually 2298 * XXX take any action on the seek operation. 2299 */ 2300 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2301 goto out; 2302 2303 /* dofilewrite() will unuse the descriptor for us */ 2304 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2305 &offset, 0, retval)); 2306 2307 out: 2308 FILE_UNUSE(fp, l); 2309 return (error); 2310 } 2311 2312 /* 2313 * Positional gather write system call. 2314 */ 2315 int 2316 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2317 { 2318 struct sys_pwritev_args /* { 2319 syscallarg(int) fd; 2320 syscallarg(const struct iovec *) iovp; 2321 syscallarg(int) iovcnt; 2322 syscallarg(off_t) offset; 2323 } */ *uap = v; 2324 struct proc *p = l->l_proc; 2325 struct filedesc *fdp = p->p_fd; 2326 struct file *fp; 2327 struct vnode *vp; 2328 off_t offset; 2329 int error, fd = SCARG(uap, fd); 2330 2331 if ((fp = fd_getfile(fdp, fd)) == NULL) 2332 return (EBADF); 2333 2334 if ((fp->f_flag & FWRITE) == 0) { 2335 mutex_exit(&fp->f_lock); 2336 return (EBADF); 2337 } 2338 2339 FILE_USE(fp); 2340 2341 vp = (struct vnode *)fp->f_data; 2342 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2343 error = ESPIPE; 2344 goto out; 2345 } 2346 2347 offset = SCARG(uap, offset); 2348 2349 /* 2350 * XXX This works because no file systems actually 2351 * XXX take any action on the seek operation. 2352 */ 2353 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2354 goto out; 2355 2356 /* dofilewritev() will unuse the descriptor for us */ 2357 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2358 &offset, 0, retval)); 2359 2360 out: 2361 FILE_UNUSE(fp, l); 2362 return (error); 2363 } 2364 2365 /* 2366 * Check access permissions. 2367 */ 2368 int 2369 sys_access(struct lwp *l, void *v, register_t *retval) 2370 { 2371 struct sys_access_args /* { 2372 syscallarg(const char *) path; 2373 syscallarg(int) flags; 2374 } */ *uap = v; 2375 kauth_cred_t cred; 2376 struct vnode *vp; 2377 int error, flags; 2378 struct nameidata nd; 2379 2380 cred = kauth_cred_dup(l->l_cred); 2381 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2382 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2383 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2384 SCARG(uap, path), l); 2385 /* Override default credentials */ 2386 nd.ni_cnd.cn_cred = cred; 2387 if ((error = namei(&nd)) != 0) 2388 goto out; 2389 vp = nd.ni_vp; 2390 2391 /* Flags == 0 means only check for existence. */ 2392 if (SCARG(uap, flags)) { 2393 flags = 0; 2394 if (SCARG(uap, flags) & R_OK) 2395 flags |= VREAD; 2396 if (SCARG(uap, flags) & W_OK) 2397 flags |= VWRITE; 2398 if (SCARG(uap, flags) & X_OK) 2399 flags |= VEXEC; 2400 2401 error = VOP_ACCESS(vp, flags, cred, l); 2402 if (!error && (flags & VWRITE)) 2403 error = vn_writechk(vp); 2404 } 2405 vput(vp); 2406 out: 2407 kauth_cred_free(cred); 2408 return (error); 2409 } 2410 2411 /* 2412 * Common code for all sys_stat functions, including compat versions. 2413 */ 2414 int 2415 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2416 struct stat *sb) 2417 { 2418 int error; 2419 struct nameidata nd; 2420 2421 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF | TRYEMULROOT , UIO_USERSPACE, path, l); 2422 error = namei(&nd); 2423 if (error != 0) 2424 return error; 2425 error = vn_stat(nd.ni_vp, sb, l); 2426 vput(nd.ni_vp); 2427 return error; 2428 } 2429 2430 /* 2431 * Get file status; this version follows links. 2432 */ 2433 /* ARGSUSED */ 2434 int 2435 sys___stat30(struct lwp *l, void *v, register_t *retval) 2436 { 2437 struct sys___stat30_args /* { 2438 syscallarg(const char *) path; 2439 syscallarg(struct stat *) ub; 2440 } */ *uap = v; 2441 struct stat sb; 2442 int error; 2443 2444 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2445 if (error) 2446 return error; 2447 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2448 } 2449 2450 /* 2451 * Get file status; this version does not follow links. 2452 */ 2453 /* ARGSUSED */ 2454 int 2455 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2456 { 2457 struct sys___lstat30_args /* { 2458 syscallarg(const char *) path; 2459 syscallarg(struct stat *) ub; 2460 } */ *uap = v; 2461 struct stat sb; 2462 int error; 2463 2464 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2465 if (error) 2466 return error; 2467 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2468 } 2469 2470 /* 2471 * Get configurable pathname variables. 2472 */ 2473 /* ARGSUSED */ 2474 int 2475 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2476 { 2477 struct sys_pathconf_args /* { 2478 syscallarg(const char *) path; 2479 syscallarg(int) name; 2480 } */ *uap = v; 2481 int error; 2482 struct nameidata nd; 2483 2484 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2485 SCARG(uap, path), l); 2486 if ((error = namei(&nd)) != 0) 2487 return (error); 2488 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2489 vput(nd.ni_vp); 2490 return (error); 2491 } 2492 2493 /* 2494 * Return target name of a symbolic link. 2495 */ 2496 /* ARGSUSED */ 2497 int 2498 sys_readlink(struct lwp *l, void *v, register_t *retval) 2499 { 2500 struct sys_readlink_args /* { 2501 syscallarg(const char *) path; 2502 syscallarg(char *) buf; 2503 syscallarg(size_t) count; 2504 } */ *uap = v; 2505 struct vnode *vp; 2506 struct iovec aiov; 2507 struct uio auio; 2508 int error; 2509 struct nameidata nd; 2510 2511 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 2512 SCARG(uap, path), l); 2513 if ((error = namei(&nd)) != 0) 2514 return (error); 2515 vp = nd.ni_vp; 2516 if (vp->v_type != VLNK) 2517 error = EINVAL; 2518 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2519 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2520 aiov.iov_base = SCARG(uap, buf); 2521 aiov.iov_len = SCARG(uap, count); 2522 auio.uio_iov = &aiov; 2523 auio.uio_iovcnt = 1; 2524 auio.uio_offset = 0; 2525 auio.uio_rw = UIO_READ; 2526 KASSERT(l == curlwp); 2527 auio.uio_vmspace = l->l_proc->p_vmspace; 2528 auio.uio_resid = SCARG(uap, count); 2529 error = VOP_READLINK(vp, &auio, l->l_cred); 2530 } 2531 vput(vp); 2532 *retval = SCARG(uap, count) - auio.uio_resid; 2533 return (error); 2534 } 2535 2536 /* 2537 * Change flags of a file given a path name. 2538 */ 2539 /* ARGSUSED */ 2540 int 2541 sys_chflags(struct lwp *l, void *v, register_t *retval) 2542 { 2543 struct sys_chflags_args /* { 2544 syscallarg(const char *) path; 2545 syscallarg(u_long) flags; 2546 } */ *uap = v; 2547 struct vnode *vp; 2548 int error; 2549 struct nameidata nd; 2550 2551 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2552 if ((error = namei(&nd)) != 0) 2553 return (error); 2554 vp = nd.ni_vp; 2555 error = change_flags(vp, SCARG(uap, flags), l); 2556 vput(vp); 2557 return (error); 2558 } 2559 2560 /* 2561 * Change flags of a file given a file descriptor. 2562 */ 2563 /* ARGSUSED */ 2564 int 2565 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2566 { 2567 struct sys_fchflags_args /* { 2568 syscallarg(int) fd; 2569 syscallarg(u_long) flags; 2570 } */ *uap = v; 2571 struct proc *p = l->l_proc; 2572 struct vnode *vp; 2573 struct file *fp; 2574 int error; 2575 2576 /* getvnode() will use the descriptor for us */ 2577 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2578 return (error); 2579 vp = (struct vnode *)fp->f_data; 2580 error = change_flags(vp, SCARG(uap, flags), l); 2581 VOP_UNLOCK(vp, 0); 2582 FILE_UNUSE(fp, l); 2583 return (error); 2584 } 2585 2586 /* 2587 * Change flags of a file given a path name; this version does 2588 * not follow links. 2589 */ 2590 int 2591 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2592 { 2593 struct sys_lchflags_args /* { 2594 syscallarg(const char *) path; 2595 syscallarg(u_long) flags; 2596 } */ *uap = v; 2597 struct vnode *vp; 2598 int error; 2599 struct nameidata nd; 2600 2601 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2602 if ((error = namei(&nd)) != 0) 2603 return (error); 2604 vp = nd.ni_vp; 2605 error = change_flags(vp, SCARG(uap, flags), l); 2606 vput(vp); 2607 return (error); 2608 } 2609 2610 /* 2611 * Common routine to change flags of a file. 2612 */ 2613 int 2614 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2615 { 2616 struct vattr vattr; 2617 int error; 2618 2619 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2620 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2621 /* 2622 * Non-superusers cannot change the flags on devices, even if they 2623 * own them. 2624 */ 2625 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2626 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2627 goto out; 2628 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2629 error = EINVAL; 2630 goto out; 2631 } 2632 } 2633 VATTR_NULL(&vattr); 2634 vattr.va_flags = flags; 2635 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2636 out: 2637 return (error); 2638 } 2639 2640 /* 2641 * Change mode of a file given path name; this version follows links. 2642 */ 2643 /* ARGSUSED */ 2644 int 2645 sys_chmod(struct lwp *l, void *v, register_t *retval) 2646 { 2647 struct sys_chmod_args /* { 2648 syscallarg(const char *) path; 2649 syscallarg(int) mode; 2650 } */ *uap = v; 2651 int error; 2652 struct nameidata nd; 2653 2654 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2655 if ((error = namei(&nd)) != 0) 2656 return (error); 2657 2658 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2659 2660 vrele(nd.ni_vp); 2661 return (error); 2662 } 2663 2664 /* 2665 * Change mode of a file given a file descriptor. 2666 */ 2667 /* ARGSUSED */ 2668 int 2669 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2670 { 2671 struct sys_fchmod_args /* { 2672 syscallarg(int) fd; 2673 syscallarg(int) mode; 2674 } */ *uap = v; 2675 struct proc *p = l->l_proc; 2676 struct file *fp; 2677 int error; 2678 2679 /* getvnode() will use the descriptor for us */ 2680 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2681 return (error); 2682 2683 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2684 FILE_UNUSE(fp, l); 2685 return (error); 2686 } 2687 2688 /* 2689 * Change mode of a file given path name; this version does not follow links. 2690 */ 2691 /* ARGSUSED */ 2692 int 2693 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2694 { 2695 struct sys_lchmod_args /* { 2696 syscallarg(const char *) path; 2697 syscallarg(int) mode; 2698 } */ *uap = v; 2699 int error; 2700 struct nameidata nd; 2701 2702 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2703 if ((error = namei(&nd)) != 0) 2704 return (error); 2705 2706 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2707 2708 vrele(nd.ni_vp); 2709 return (error); 2710 } 2711 2712 /* 2713 * Common routine to set mode given a vnode. 2714 */ 2715 static int 2716 change_mode(struct vnode *vp, int mode, struct lwp *l) 2717 { 2718 struct vattr vattr; 2719 int error; 2720 2721 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2722 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2723 VATTR_NULL(&vattr); 2724 vattr.va_mode = mode & ALLPERMS; 2725 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2726 VOP_UNLOCK(vp, 0); 2727 return (error); 2728 } 2729 2730 /* 2731 * Set ownership given a path name; this version follows links. 2732 */ 2733 /* ARGSUSED */ 2734 int 2735 sys_chown(struct lwp *l, void *v, register_t *retval) 2736 { 2737 struct sys_chown_args /* { 2738 syscallarg(const char *) path; 2739 syscallarg(uid_t) uid; 2740 syscallarg(gid_t) gid; 2741 } */ *uap = v; 2742 int error; 2743 struct nameidata nd; 2744 2745 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2746 if ((error = namei(&nd)) != 0) 2747 return (error); 2748 2749 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2750 2751 vrele(nd.ni_vp); 2752 return (error); 2753 } 2754 2755 /* 2756 * Set ownership given a path name; this version follows links. 2757 * Provides POSIX semantics. 2758 */ 2759 /* ARGSUSED */ 2760 int 2761 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2762 { 2763 struct sys_chown_args /* { 2764 syscallarg(const char *) path; 2765 syscallarg(uid_t) uid; 2766 syscallarg(gid_t) gid; 2767 } */ *uap = v; 2768 int error; 2769 struct nameidata nd; 2770 2771 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2772 if ((error = namei(&nd)) != 0) 2773 return (error); 2774 2775 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2776 2777 vrele(nd.ni_vp); 2778 return (error); 2779 } 2780 2781 /* 2782 * Set ownership given a file descriptor. 2783 */ 2784 /* ARGSUSED */ 2785 int 2786 sys_fchown(struct lwp *l, void *v, register_t *retval) 2787 { 2788 struct sys_fchown_args /* { 2789 syscallarg(int) fd; 2790 syscallarg(uid_t) uid; 2791 syscallarg(gid_t) gid; 2792 } */ *uap = v; 2793 struct proc *p = l->l_proc; 2794 int error; 2795 struct file *fp; 2796 2797 /* getvnode() will use the descriptor for us */ 2798 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2799 return (error); 2800 2801 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2802 SCARG(uap, gid), l, 0); 2803 FILE_UNUSE(fp, l); 2804 return (error); 2805 } 2806 2807 /* 2808 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2809 */ 2810 /* ARGSUSED */ 2811 int 2812 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2813 { 2814 struct sys_fchown_args /* { 2815 syscallarg(int) fd; 2816 syscallarg(uid_t) uid; 2817 syscallarg(gid_t) gid; 2818 } */ *uap = v; 2819 struct proc *p = l->l_proc; 2820 int error; 2821 struct file *fp; 2822 2823 /* getvnode() will use the descriptor for us */ 2824 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2825 return (error); 2826 2827 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2828 SCARG(uap, gid), l, 1); 2829 FILE_UNUSE(fp, l); 2830 return (error); 2831 } 2832 2833 /* 2834 * Set ownership given a path name; this version does not follow links. 2835 */ 2836 /* ARGSUSED */ 2837 int 2838 sys_lchown(struct lwp *l, void *v, register_t *retval) 2839 { 2840 struct sys_lchown_args /* { 2841 syscallarg(const char *) path; 2842 syscallarg(uid_t) uid; 2843 syscallarg(gid_t) gid; 2844 } */ *uap = v; 2845 int error; 2846 struct nameidata nd; 2847 2848 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2849 if ((error = namei(&nd)) != 0) 2850 return (error); 2851 2852 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2853 2854 vrele(nd.ni_vp); 2855 return (error); 2856 } 2857 2858 /* 2859 * Set ownership given a path name; this version does not follow links. 2860 * Provides POSIX/XPG semantics. 2861 */ 2862 /* ARGSUSED */ 2863 int 2864 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2865 { 2866 struct sys_lchown_args /* { 2867 syscallarg(const char *) path; 2868 syscallarg(uid_t) uid; 2869 syscallarg(gid_t) gid; 2870 } */ *uap = v; 2871 int error; 2872 struct nameidata nd; 2873 2874 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 2875 if ((error = namei(&nd)) != 0) 2876 return (error); 2877 2878 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2879 2880 vrele(nd.ni_vp); 2881 return (error); 2882 } 2883 2884 /* 2885 * Common routine to set ownership given a vnode. 2886 */ 2887 static int 2888 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2889 int posix_semantics) 2890 { 2891 struct vattr vattr; 2892 mode_t newmode; 2893 int error; 2894 2895 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2896 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2897 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2898 goto out; 2899 2900 #define CHANGED(x) ((int)(x) != -1) 2901 newmode = vattr.va_mode; 2902 if (posix_semantics) { 2903 /* 2904 * POSIX/XPG semantics: if the caller is not the super-user, 2905 * clear set-user-id and set-group-id bits. Both POSIX and 2906 * the XPG consider the behaviour for calls by the super-user 2907 * implementation-defined; we leave the set-user-id and set- 2908 * group-id settings intact in that case. 2909 */ 2910 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2911 NULL) != 0) 2912 newmode &= ~(S_ISUID | S_ISGID); 2913 } else { 2914 /* 2915 * NetBSD semantics: when changing owner and/or group, 2916 * clear the respective bit(s). 2917 */ 2918 if (CHANGED(uid)) 2919 newmode &= ~S_ISUID; 2920 if (CHANGED(gid)) 2921 newmode &= ~S_ISGID; 2922 } 2923 /* Update va_mode iff altered. */ 2924 if (vattr.va_mode == newmode) 2925 newmode = VNOVAL; 2926 2927 VATTR_NULL(&vattr); 2928 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2929 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2930 vattr.va_mode = newmode; 2931 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2932 #undef CHANGED 2933 2934 out: 2935 VOP_UNLOCK(vp, 0); 2936 return (error); 2937 } 2938 2939 /* 2940 * Set the access and modification times given a path name; this 2941 * version follows links. 2942 */ 2943 /* ARGSUSED */ 2944 int 2945 sys_utimes(struct lwp *l, void *v, register_t *retval) 2946 { 2947 struct sys_utimes_args /* { 2948 syscallarg(const char *) path; 2949 syscallarg(const struct timeval *) tptr; 2950 } */ *uap = v; 2951 2952 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 2953 SCARG(uap, tptr), UIO_USERSPACE); 2954 } 2955 2956 /* 2957 * Set the access and modification times given a file descriptor. 2958 */ 2959 /* ARGSUSED */ 2960 int 2961 sys_futimes(struct lwp *l, void *v, register_t *retval) 2962 { 2963 struct sys_futimes_args /* { 2964 syscallarg(int) fd; 2965 syscallarg(const struct timeval *) tptr; 2966 } */ *uap = v; 2967 int error; 2968 struct file *fp; 2969 2970 /* getvnode() will use the descriptor for us */ 2971 if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0) 2972 return (error); 2973 2974 error = do_sys_utimes(l, fp->f_data, NULL, 0, 2975 SCARG(uap, tptr), UIO_USERSPACE); 2976 2977 FILE_UNUSE(fp, l); 2978 return (error); 2979 } 2980 2981 /* 2982 * Set the access and modification times given a path name; this 2983 * version does not follow links. 2984 */ 2985 int 2986 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2987 { 2988 struct sys_lutimes_args /* { 2989 syscallarg(const char *) path; 2990 syscallarg(const struct timeval *) tptr; 2991 } */ *uap = v; 2992 2993 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 2994 SCARG(uap, tptr), UIO_USERSPACE); 2995 } 2996 2997 /* 2998 * Common routine to set access and modification times given a vnode. 2999 */ 3000 int 3001 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 3002 const struct timeval *tptr, enum uio_seg seg) 3003 { 3004 struct vattr vattr; 3005 struct nameidata nd; 3006 int error; 3007 3008 VATTR_NULL(&vattr); 3009 if (tptr == NULL) { 3010 nanotime(&vattr.va_atime); 3011 vattr.va_mtime = vattr.va_atime; 3012 vattr.va_vaflags |= VA_UTIMES_NULL; 3013 } else { 3014 struct timeval tv[2]; 3015 3016 if (seg != UIO_SYSSPACE) { 3017 error = copyin(tptr, &tv, sizeof (tv)); 3018 if (error != 0) 3019 return error; 3020 tptr = tv; 3021 } 3022 TIMEVAL_TO_TIMESPEC(tptr, &vattr.va_atime); 3023 TIMEVAL_TO_TIMESPEC(tptr + 1, &vattr.va_mtime); 3024 } 3025 3026 if (vp == NULL) { 3027 NDINIT(&nd, LOOKUP, flag | TRYEMULROOT, UIO_USERSPACE, path, l); 3028 if ((error = namei(&nd)) != 0) 3029 return (error); 3030 vp = nd.ni_vp; 3031 } else 3032 nd.ni_vp = NULL; 3033 3034 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3035 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3036 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3037 VOP_UNLOCK(vp, 0); 3038 3039 if (nd.ni_vp != NULL) 3040 vrele(nd.ni_vp); 3041 3042 return (error); 3043 } 3044 3045 /* 3046 * Truncate a file given its path name. 3047 */ 3048 /* ARGSUSED */ 3049 int 3050 sys_truncate(struct lwp *l, void *v, register_t *retval) 3051 { 3052 struct sys_truncate_args /* { 3053 syscallarg(const char *) path; 3054 syscallarg(int) pad; 3055 syscallarg(off_t) length; 3056 } */ *uap = v; 3057 struct vnode *vp; 3058 struct vattr vattr; 3059 int error; 3060 struct nameidata nd; 3061 3062 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3063 if ((error = namei(&nd)) != 0) 3064 return (error); 3065 vp = nd.ni_vp; 3066 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3067 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3068 if (vp->v_type == VDIR) 3069 error = EISDIR; 3070 else if ((error = vn_writechk(vp)) == 0 && 3071 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3072 VATTR_NULL(&vattr); 3073 vattr.va_size = SCARG(uap, length); 3074 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3075 } 3076 vput(vp); 3077 return (error); 3078 } 3079 3080 /* 3081 * Truncate a file given a file descriptor. 3082 */ 3083 /* ARGSUSED */ 3084 int 3085 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3086 { 3087 struct sys_ftruncate_args /* { 3088 syscallarg(int) fd; 3089 syscallarg(int) pad; 3090 syscallarg(off_t) length; 3091 } */ *uap = v; 3092 struct proc *p = l->l_proc; 3093 struct vattr vattr; 3094 struct vnode *vp; 3095 struct file *fp; 3096 int error; 3097 3098 /* getvnode() will use the descriptor for us */ 3099 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3100 return (error); 3101 if ((fp->f_flag & FWRITE) == 0) { 3102 error = EINVAL; 3103 goto out; 3104 } 3105 vp = (struct vnode *)fp->f_data; 3106 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3107 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3108 if (vp->v_type == VDIR) 3109 error = EISDIR; 3110 else if ((error = vn_writechk(vp)) == 0) { 3111 VATTR_NULL(&vattr); 3112 vattr.va_size = SCARG(uap, length); 3113 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3114 } 3115 VOP_UNLOCK(vp, 0); 3116 out: 3117 FILE_UNUSE(fp, l); 3118 return (error); 3119 } 3120 3121 /* 3122 * Sync an open file. 3123 */ 3124 /* ARGSUSED */ 3125 int 3126 sys_fsync(struct lwp *l, void *v, register_t *retval) 3127 { 3128 struct sys_fsync_args /* { 3129 syscallarg(int) fd; 3130 } */ *uap = v; 3131 struct proc *p = l->l_proc; 3132 struct vnode *vp; 3133 struct file *fp; 3134 int error; 3135 3136 /* getvnode() will use the descriptor for us */ 3137 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3138 return (error); 3139 vp = (struct vnode *)fp->f_data; 3140 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3141 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3142 if (error == 0 && bioops.io_fsync != NULL && 3143 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) { 3144 KERNEL_LOCK(1, l); 3145 (*bioops.io_fsync)(vp, 0); 3146 KERNEL_UNLOCK_ONE(l); 3147 } 3148 VOP_UNLOCK(vp, 0); 3149 FILE_UNUSE(fp, l); 3150 return (error); 3151 } 3152 3153 /* 3154 * Sync a range of file data. API modeled after that found in AIX. 3155 * 3156 * FDATASYNC indicates that we need only save enough metadata to be able 3157 * to re-read the written data. Note we duplicate AIX's requirement that 3158 * the file be open for writing. 3159 */ 3160 /* ARGSUSED */ 3161 int 3162 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3163 { 3164 struct sys_fsync_range_args /* { 3165 syscallarg(int) fd; 3166 syscallarg(int) flags; 3167 syscallarg(off_t) start; 3168 syscallarg(off_t) length; 3169 } */ *uap = v; 3170 struct proc *p = l->l_proc; 3171 struct vnode *vp; 3172 struct file *fp; 3173 int flags, nflags; 3174 off_t s, e, len; 3175 int error; 3176 3177 /* getvnode() will use the descriptor for us */ 3178 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3179 return (error); 3180 3181 if ((fp->f_flag & FWRITE) == 0) { 3182 error = EBADF; 3183 goto out; 3184 } 3185 3186 flags = SCARG(uap, flags); 3187 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3188 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3189 error = EINVAL; 3190 goto out; 3191 } 3192 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3193 if (flags & FDATASYNC) 3194 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3195 else 3196 nflags = FSYNC_WAIT; 3197 if (flags & FDISKSYNC) 3198 nflags |= FSYNC_CACHE; 3199 3200 len = SCARG(uap, length); 3201 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3202 if (len) { 3203 s = SCARG(uap, start); 3204 e = s + len; 3205 if (e < s) { 3206 error = EINVAL; 3207 goto out; 3208 } 3209 } else { 3210 e = 0; 3211 s = 0; 3212 } 3213 3214 vp = (struct vnode *)fp->f_data; 3215 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3216 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3217 3218 if (error == 0 && bioops.io_fsync != NULL && 3219 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) { 3220 KERNEL_LOCK(1, l); 3221 (*bioops.io_fsync)(vp, nflags); 3222 KERNEL_UNLOCK_ONE(l); 3223 } 3224 3225 VOP_UNLOCK(vp, 0); 3226 out: 3227 FILE_UNUSE(fp, l); 3228 return (error); 3229 } 3230 3231 /* 3232 * Sync the data of an open file. 3233 */ 3234 /* ARGSUSED */ 3235 int 3236 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3237 { 3238 struct sys_fdatasync_args /* { 3239 syscallarg(int) fd; 3240 } */ *uap = v; 3241 struct proc *p = l->l_proc; 3242 struct vnode *vp; 3243 struct file *fp; 3244 int error; 3245 3246 /* getvnode() will use the descriptor for us */ 3247 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3248 return (error); 3249 if ((fp->f_flag & FWRITE) == 0) { 3250 FILE_UNUSE(fp, l); 3251 return (EBADF); 3252 } 3253 vp = (struct vnode *)fp->f_data; 3254 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3255 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3256 VOP_UNLOCK(vp, 0); 3257 FILE_UNUSE(fp, l); 3258 return (error); 3259 } 3260 3261 /* 3262 * Rename files, (standard) BSD semantics frontend. 3263 */ 3264 /* ARGSUSED */ 3265 int 3266 sys_rename(struct lwp *l, void *v, register_t *retval) 3267 { 3268 struct sys_rename_args /* { 3269 syscallarg(const char *) from; 3270 syscallarg(const char *) to; 3271 } */ *uap = v; 3272 3273 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3274 } 3275 3276 /* 3277 * Rename files, POSIX semantics frontend. 3278 */ 3279 /* ARGSUSED */ 3280 int 3281 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3282 { 3283 struct sys___posix_rename_args /* { 3284 syscallarg(const char *) from; 3285 syscallarg(const char *) to; 3286 } */ *uap = v; 3287 3288 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3289 } 3290 3291 /* 3292 * Rename files. Source and destination must either both be directories, 3293 * or both not be directories. If target is a directory, it must be empty. 3294 * If `from' and `to' refer to the same object, the value of the `retain' 3295 * argument is used to determine whether `from' will be 3296 * 3297 * (retain == 0) deleted unless `from' and `to' refer to the same 3298 * object in the file system's name space (BSD). 3299 * (retain == 1) always retained (POSIX). 3300 */ 3301 static int 3302 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3303 { 3304 struct vnode *tvp, *fvp, *tdvp; 3305 struct nameidata fromnd, tond; 3306 struct proc *p; 3307 int error; 3308 3309 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART | TRYEMULROOT, UIO_USERSPACE, 3310 from, l); 3311 if ((error = namei(&fromnd)) != 0) 3312 return (error); 3313 if (fromnd.ni_dvp != fromnd.ni_vp) 3314 VOP_UNLOCK(fromnd.ni_dvp, 0); 3315 fvp = fromnd.ni_vp; 3316 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | TRYEMULROOT | 3317 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3318 if ((error = namei(&tond)) != 0) { 3319 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3320 vrele(fromnd.ni_dvp); 3321 vrele(fvp); 3322 goto out1; 3323 } 3324 tdvp = tond.ni_dvp; 3325 tvp = tond.ni_vp; 3326 3327 if (tvp != NULL) { 3328 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3329 error = ENOTDIR; 3330 goto out; 3331 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3332 error = EISDIR; 3333 goto out; 3334 } 3335 } 3336 3337 if (fvp == tdvp) 3338 error = EINVAL; 3339 3340 /* 3341 * Source and destination refer to the same object. 3342 */ 3343 if (fvp == tvp) { 3344 if (retain) 3345 error = -1; 3346 else if (fromnd.ni_dvp == tdvp && 3347 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3348 !memcmp(fromnd.ni_cnd.cn_nameptr, 3349 tond.ni_cnd.cn_nameptr, 3350 fromnd.ni_cnd.cn_namelen)) 3351 error = -1; 3352 } 3353 3354 #if NVERIEXEC > 0 3355 if (!error) { 3356 char *f1, *f2; 3357 3358 f1 = malloc(fromnd.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3359 strlcpy(f1, fromnd.ni_cnd.cn_nameptr, fromnd.ni_cnd.cn_namelen); 3360 3361 f2 = malloc(tond.ni_cnd.cn_namelen + 1, M_TEMP, M_WAITOK); 3362 strlcpy(f2, tond.ni_cnd.cn_nameptr, tond.ni_cnd.cn_namelen); 3363 3364 error = veriexec_renamechk(l, fvp, f1, tvp, f2); 3365 3366 free(f1, M_TEMP); 3367 free(f2, M_TEMP); 3368 } 3369 #endif /* NVERIEXEC > 0 */ 3370 3371 out: 3372 p = l->l_proc; 3373 if (!error) { 3374 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3375 if (fromnd.ni_dvp != tdvp) 3376 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3377 if (tvp) { 3378 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3379 } 3380 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3381 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3382 } else { 3383 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3384 if (tdvp == tvp) 3385 vrele(tdvp); 3386 else 3387 vput(tdvp); 3388 if (tvp) 3389 vput(tvp); 3390 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3391 vrele(fromnd.ni_dvp); 3392 vrele(fvp); 3393 } 3394 vrele(tond.ni_startdir); 3395 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3396 out1: 3397 if (fromnd.ni_startdir) 3398 vrele(fromnd.ni_startdir); 3399 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3400 return (error == -1 ? 0 : error); 3401 } 3402 3403 /* 3404 * Make a directory file. 3405 */ 3406 /* ARGSUSED */ 3407 int 3408 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3409 { 3410 struct sys_mkdir_args /* { 3411 syscallarg(const char *) path; 3412 syscallarg(int) mode; 3413 } */ *uap = v; 3414 struct proc *p = l->l_proc; 3415 struct vnode *vp; 3416 struct vattr vattr; 3417 int error; 3418 struct nameidata nd; 3419 3420 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, UIO_USERSPACE, 3421 SCARG(uap, path), l); 3422 if ((error = namei(&nd)) != 0) 3423 return (error); 3424 vp = nd.ni_vp; 3425 if (vp != NULL) { 3426 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3427 if (nd.ni_dvp == vp) 3428 vrele(nd.ni_dvp); 3429 else 3430 vput(nd.ni_dvp); 3431 vrele(vp); 3432 return (EEXIST); 3433 } 3434 VATTR_NULL(&vattr); 3435 vattr.va_type = VDIR; 3436 /* We will read cwdi->cwdi_cmask unlocked. */ 3437 vattr.va_mode = 3438 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3439 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3440 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3441 if (!error) 3442 vput(nd.ni_vp); 3443 return (error); 3444 } 3445 3446 /* 3447 * Remove a directory file. 3448 */ 3449 /* ARGSUSED */ 3450 int 3451 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3452 { 3453 struct sys_rmdir_args /* { 3454 syscallarg(const char *) path; 3455 } */ *uap = v; 3456 struct vnode *vp; 3457 int error; 3458 struct nameidata nd; 3459 3460 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, UIO_USERSPACE, 3461 SCARG(uap, path), l); 3462 if ((error = namei(&nd)) != 0) 3463 return (error); 3464 vp = nd.ni_vp; 3465 if (vp->v_type != VDIR) { 3466 error = ENOTDIR; 3467 goto out; 3468 } 3469 /* 3470 * No rmdir "." please. 3471 */ 3472 if (nd.ni_dvp == vp) { 3473 error = EINVAL; 3474 goto out; 3475 } 3476 /* 3477 * The root of a mounted filesystem cannot be deleted. 3478 */ 3479 if (vp->v_vflag & VV_ROOT) { 3480 error = EBUSY; 3481 goto out; 3482 } 3483 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3484 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3485 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3486 return (error); 3487 3488 out: 3489 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3490 if (nd.ni_dvp == vp) 3491 vrele(nd.ni_dvp); 3492 else 3493 vput(nd.ni_dvp); 3494 vput(vp); 3495 return (error); 3496 } 3497 3498 /* 3499 * Read a block of directory entries in a file system independent format. 3500 */ 3501 int 3502 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3503 { 3504 struct sys___getdents30_args /* { 3505 syscallarg(int) fd; 3506 syscallarg(char *) buf; 3507 syscallarg(size_t) count; 3508 } */ *uap = v; 3509 struct proc *p = l->l_proc; 3510 struct file *fp; 3511 int error, done; 3512 3513 /* getvnode() will use the descriptor for us */ 3514 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3515 return (error); 3516 if ((fp->f_flag & FREAD) == 0) { 3517 error = EBADF; 3518 goto out; 3519 } 3520 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3521 SCARG(uap, count), &done, l, 0, 0); 3522 #ifdef KTRACE 3523 if (!error && KTRPOINT(p, KTR_GENIO)) { 3524 struct iovec iov; 3525 iov.iov_base = SCARG(uap, buf); 3526 iov.iov_len = done; 3527 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3528 } 3529 #endif 3530 *retval = done; 3531 out: 3532 FILE_UNUSE(fp, l); 3533 return (error); 3534 } 3535 3536 /* 3537 * Set the mode mask for creation of filesystem nodes. 3538 */ 3539 int 3540 sys_umask(struct lwp *l, void *v, register_t *retval) 3541 { 3542 struct sys_umask_args /* { 3543 syscallarg(mode_t) newmask; 3544 } */ *uap = v; 3545 struct proc *p = l->l_proc; 3546 struct cwdinfo *cwdi; 3547 3548 /* 3549 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3550 * important is that we serialize changes to the mask. The 3551 * rw_exit() will issue a write memory barrier on our behalf, 3552 * and force the changes out to other CPUs (as it must use an 3553 * atomic operation, draining the local CPU's store buffers). 3554 */ 3555 cwdi = p->p_cwdi; 3556 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3557 *retval = cwdi->cwdi_cmask; 3558 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3559 rw_exit(&cwdi->cwdi_lock); 3560 3561 return (0); 3562 } 3563 3564 /* 3565 * Void all references to file by ripping underlying filesystem 3566 * away from vnode. 3567 */ 3568 /* ARGSUSED */ 3569 int 3570 sys_revoke(struct lwp *l, void *v, register_t *retval) 3571 { 3572 struct sys_revoke_args /* { 3573 syscallarg(const char *) path; 3574 } */ *uap = v; 3575 struct vnode *vp; 3576 struct vattr vattr; 3577 int error; 3578 bool revoke; 3579 struct nameidata nd; 3580 3581 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), l); 3582 if ((error = namei(&nd)) != 0) 3583 return (error); 3584 vp = nd.ni_vp; 3585 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3586 goto out; 3587 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3588 (error = kauth_authorize_generic(l->l_cred, 3589 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3590 goto out; 3591 mutex_enter(&vp->v_interlock); 3592 revoke = (vp->v_usecount > 1 || (vp->v_iflag & (VI_ALIASED | VI_LAYER))); 3593 mutex_exit(&vp->v_interlock); 3594 if (revoke) 3595 VOP_REVOKE(vp, REVOKEALL); 3596 out: 3597 vrele(vp); 3598 return (error); 3599 } 3600 3601 /* 3602 * Convert a user file descriptor to a kernel file entry. 3603 */ 3604 int 3605 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3606 { 3607 struct vnode *vp; 3608 struct file *fp; 3609 3610 if ((fp = fd_getfile(fdp, fd)) == NULL) 3611 return (EBADF); 3612 3613 FILE_USE(fp); 3614 3615 if (fp->f_type != DTYPE_VNODE) { 3616 FILE_UNUSE(fp, NULL); 3617 return (EINVAL); 3618 } 3619 3620 vp = (struct vnode *)fp->f_data; 3621 if (vp->v_type == VBAD) { 3622 FILE_UNUSE(fp, NULL); 3623 return (EBADF); 3624 } 3625 3626 *fpp = fp; 3627 return (0); 3628 } 3629
Indexes created Fri Oct 03 02:09:57 GMT 2025