vfs_syscalls.c revision 1.302
1 /* $NetBSD: vfs_syscalls.c,v 1.302 2007/02/18 20:36:36 pooka 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 1.302 2007/02/18 20:36:36 pooka 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 #ifdef KTRACE 66 #include <sys/ktrace.h> 67 #endif 68 #ifdef FILEASSOC 69 #include <sys/fileassoc.h> 70 #endif /* FILEASSOC */ 71 #if NVERIEXEC > 0 72 #include <sys/verified_exec.h> 73 #include <sys/syslog.h> 74 #endif /* NVERIEXEC > 0 */ 75 #include <sys/kauth.h> 76 77 #include <miscfs/genfs/genfs.h> 78 #include <miscfs/syncfs/syncfs.h> 79 80 #ifdef COMPAT_30 81 #include "opt_nfsserver.h" 82 #include <nfs/rpcv2.h> 83 #endif 84 #include <nfs/nfsproto.h> 85 #ifdef COMPAT_30 86 #include <nfs/nfs.h> 87 #include <nfs/nfs_var.h> 88 #endif 89 90 #if NFSS > 0 91 #include <dev/fssvar.h> 92 #endif 93 94 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 95 96 static int change_dir(struct nameidata *, struct lwp *); 97 static int change_flags(struct vnode *, u_long, struct lwp *); 98 static int change_mode(struct vnode *, int, struct lwp *l); 99 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 100 static int change_utimes(struct vnode *vp, const struct timeval *, 101 struct lwp *l); 102 static int rename_files(const char *, const char *, struct lwp *, int); 103 104 void checkdirs(struct vnode *); 105 106 static int mount_update(struct lwp *, struct vnode *, const char *, int, 107 void *, struct nameidata *); 108 static int mount_domount(struct lwp *, struct vnode *, const char *, 109 const char *, int, void *, struct nameidata *); 110 static int mount_getargs(struct lwp *, struct vnode *, const char *, int, 111 void *, struct nameidata *); 112 113 int dovfsusermount = 0; 114 115 /* 116 * Virtual File System System Calls 117 */ 118 119 /* 120 * Mount a file system. 121 */ 122 123 #if defined(COMPAT_09) || defined(COMPAT_43) 124 /* 125 * This table is used to maintain compatibility with 4.3BSD 126 * and NetBSD 0.9 mount syscalls. Note, the order is important! 127 * 128 * Do not modify this table. It should only contain filesystems 129 * supported by NetBSD 0.9 and 4.3BSD. 130 */ 131 const char * const mountcompatnames[] = { 132 NULL, /* 0 = MOUNT_NONE */ 133 MOUNT_FFS, /* 1 = MOUNT_UFS */ 134 MOUNT_NFS, /* 2 */ 135 MOUNT_MFS, /* 3 */ 136 MOUNT_MSDOS, /* 4 */ 137 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 138 MOUNT_FDESC, /* 6 */ 139 MOUNT_KERNFS, /* 7 */ 140 NULL, /* 8 = MOUNT_DEVFS */ 141 MOUNT_AFS, /* 9 */ 142 }; 143 const int nmountcompatnames = sizeof(mountcompatnames) / 144 sizeof(mountcompatnames[0]); 145 #endif /* COMPAT_09 || COMPAT_43 */ 146 147 static int 148 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 149 void *data, struct nameidata *ndp) 150 { 151 struct mount *mp; 152 int error = 0, saved_flags; 153 154 mp = vp->v_mount; 155 saved_flags = mp->mnt_flag; 156 157 /* We can't operate on VROOT here. */ 158 if ((vp->v_flag & VROOT) == 0) { 159 vput(vp); 160 error = EINVAL; 161 goto out; 162 } 163 164 /* 165 * We only allow the filesystem to be reloaded if it 166 * is currently mounted read-only. 167 */ 168 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 169 error = EOPNOTSUPP; /* Needs translation */ 170 goto out; 171 } 172 173 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 174 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 175 if (error) 176 goto out; 177 178 if (vfs_busy(mp, LK_NOWAIT, 0)) { 179 error = EPERM; 180 goto out; 181 } 182 183 mp->mnt_flag &= ~MNT_OP_FLAGS; 184 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 185 186 /* 187 * Set the mount level flags. 188 */ 189 if (flags & MNT_RDONLY) 190 mp->mnt_flag |= MNT_RDONLY; 191 else if (mp->mnt_flag & MNT_RDONLY) 192 mp->mnt_iflag |= IMNT_WANTRDWR; 193 mp->mnt_flag &= 194 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 195 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 196 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 197 mp->mnt_flag |= flags & 198 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 199 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 200 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 201 MNT_IGNORE); 202 203 error = VFS_MOUNT(mp, path, data, ndp, l); 204 205 #if defined(COMPAT_30) && defined(NFSSERVER) 206 if (error) { 207 int error2; 208 209 /* Update failed; let's try and see if it was an 210 * export request. */ 211 error2 = nfs_update_exports_30(mp, path, data, l); 212 213 /* Only update error code if the export request was 214 * understood but some problem occurred while 215 * processing it. */ 216 if (error2 != EJUSTRETURN) 217 error = error2; 218 } 219 #endif 220 if (mp->mnt_iflag & IMNT_WANTRDWR) 221 mp->mnt_flag &= ~MNT_RDONLY; 222 if (error) 223 mp->mnt_flag = saved_flags; 224 mp->mnt_flag &= ~MNT_OP_FLAGS; 225 mp->mnt_iflag &= ~IMNT_WANTRDWR; 226 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 227 if (mp->mnt_syncer == NULL) 228 error = vfs_allocate_syncvnode(mp); 229 } else { 230 if (mp->mnt_syncer != NULL) 231 vfs_deallocate_syncvnode(mp); 232 } 233 vfs_unbusy(mp); 234 235 out: 236 return (error); 237 } 238 239 static int 240 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 241 const char *path, int flags, void *data, struct nameidata *ndp) 242 { 243 struct mount *mp = NULL; 244 struct vattr va; 245 char fstypename[MFSNAMELEN]; 246 int error; 247 248 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 249 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 250 if (error) { 251 vput(vp); 252 goto out; 253 } 254 255 /* Can't make a non-dir a mount-point (from here anyway). */ 256 if (vp->v_type != VDIR) { 257 error = ENOTDIR; 258 vput(vp); 259 goto out; 260 } 261 262 /* 263 * If the user is not root, ensure that they own the directory 264 * onto which we are attempting to mount. 265 */ 266 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 267 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 268 (error = kauth_authorize_generic(l->l_cred, 269 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 270 vput(vp); 271 goto out; 272 } 273 274 if (flags & MNT_EXPORTED) { 275 error = EINVAL; 276 vput(vp); 277 goto out; 278 } 279 280 /* 281 * Copy file-system type from userspace. 282 */ 283 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 284 if (error) { 285 #if defined(COMPAT_09) || defined(COMPAT_43) 286 /* 287 * Historically, filesystem types were identified by numbers. 288 * If we get an integer for the filesystem type instead of a 289 * string, we check to see if it matches one of the historic 290 * filesystem types. 291 */ 292 u_long fsindex = (u_long)fstype; 293 if (fsindex >= nmountcompatnames || 294 mountcompatnames[fsindex] == NULL) { 295 error = ENODEV; 296 vput(vp); 297 goto out; 298 } 299 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 300 #else 301 vput(vp); 302 goto out; 303 #endif 304 } 305 306 #ifdef COMPAT_10 307 /* Accept `ufs' as an alias for `ffs'. */ 308 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 309 strlcpy(fstypename, "ffs", sizeof(fstypename)); 310 #endif 311 312 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 313 vput(vp); 314 goto out; 315 } 316 317 /* 318 * Check if a file-system is not already mounted on this vnode. 319 */ 320 if (vp->v_mountedhere != NULL) { 321 error = EBUSY; 322 vput(vp); 323 goto out; 324 } 325 326 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 327 328 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 329 free(mp, M_MOUNT); 330 error = ENODEV; 331 vput(vp); 332 goto out; 333 } 334 335 TAILQ_INIT(&mp->mnt_vnodelist); 336 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 337 simple_lock_init(&mp->mnt_slock); 338 (void)vfs_busy(mp, LK_NOWAIT, 0); 339 340 mp->mnt_op->vfs_refcount++; 341 mp->mnt_vnodecovered = vp; 342 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 343 mp->mnt_unmounter = NULL; 344 mp->mnt_leaf = mp; 345 mount_initspecific(mp); 346 347 /* 348 * The underlying file system may refuse the mount for 349 * various reasons. Allow the user to force it to happen. 350 * 351 * Set the mount level flags. 352 */ 353 mp->mnt_flag = flags & 354 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 355 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 356 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 357 MNT_IGNORE | MNT_RDONLY); 358 359 error = VFS_MOUNT(mp, path, data, ndp, l); 360 mp->mnt_flag &= ~MNT_OP_FLAGS; 361 362 /* 363 * Put the new filesystem on the mount list after root. 364 */ 365 cache_purge(vp); 366 if (!error) { 367 mp->mnt_iflag &= ~IMNT_WANTRDWR; 368 vp->v_mountedhere = mp; 369 simple_lock(&mountlist_slock); 370 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 371 simple_unlock(&mountlist_slock); 372 VOP_UNLOCK(vp, 0); 373 checkdirs(vp); 374 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 375 error = vfs_allocate_syncvnode(mp); 376 vfs_unbusy(mp); 377 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 378 error = VFS_START(mp, 0, l); 379 if (error) 380 vrele(vp); 381 } else { 382 vp->v_mountedhere = NULL; 383 mp->mnt_op->vfs_refcount--; 384 vfs_unbusy(mp); 385 free(mp, M_MOUNT); 386 vput(vp); 387 } 388 389 out: 390 return (error); 391 } 392 393 static int 394 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 395 void *data, struct nameidata *ndp) 396 { 397 struct mount *mp; 398 int error; 399 400 /* If MNT_GETARGS is specified, it should be the only flag. */ 401 if (flags & ~MNT_GETARGS) { 402 error = EINVAL; 403 goto out; 404 } 405 406 mp = vp->v_mount; 407 408 /* XXX: probably some notion of "can see" here if we want isolation. */ 409 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 410 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 411 if (error) 412 goto out; 413 414 if ((vp->v_flag & VROOT) == 0) { 415 error = EINVAL; 416 goto out; 417 } 418 419 if (vfs_busy(mp, LK_NOWAIT, 0)) { 420 error = EPERM; 421 goto out; 422 } 423 424 mp->mnt_flag &= ~MNT_OP_FLAGS; 425 mp->mnt_flag |= MNT_GETARGS; 426 error = VFS_MOUNT(mp, path, data, ndp, l); 427 mp->mnt_flag &= ~MNT_OP_FLAGS; 428 429 vfs_unbusy(mp); 430 out: 431 return (error); 432 } 433 434 /* ARGSUSED */ 435 int 436 sys_mount(struct lwp *l, void *v, register_t *retval) 437 { 438 struct sys_mount_args /* { 439 syscallarg(const char *) type; 440 syscallarg(const char *) path; 441 syscallarg(int) flags; 442 syscallarg(void *) data; 443 } */ *uap = v; 444 struct vnode *vp; 445 struct nameidata nd; 446 int error; 447 448 /* 449 * Get vnode to be covered 450 */ 451 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 452 SCARG(uap, path), l); 453 if ((error = namei(&nd)) != 0) 454 return (error); 455 vp = nd.ni_vp; 456 457 /* 458 * A lookup in VFS_MOUNT might result in an attempt to 459 * lock this vnode again, so make the lock recursive. 460 */ 461 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 462 463 if (SCARG(uap, flags) & MNT_GETARGS) { 464 error = mount_getargs(l, vp, SCARG(uap, path), 465 SCARG(uap, flags), SCARG(uap, data), &nd); 466 vput(vp); 467 } else if (SCARG(uap, flags) & MNT_UPDATE) { 468 error = mount_update(l, vp, SCARG(uap, path), 469 SCARG(uap, flags), SCARG(uap, data), &nd); 470 vput(vp); 471 } else { 472 /* Locking is handled internally in mount_domount(). */ 473 error = mount_domount(l, vp, SCARG(uap, type), 474 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 475 } 476 477 return (error); 478 } 479 480 /* 481 * Scan all active processes to see if any of them have a current 482 * or root directory onto which the new filesystem has just been 483 * mounted. If so, replace them with the new mount point. 484 */ 485 void 486 checkdirs(struct vnode *olddp) 487 { 488 struct cwdinfo *cwdi; 489 struct vnode *newdp; 490 struct proc *p; 491 492 if (olddp->v_usecount == 1) 493 return; 494 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 495 panic("mount: lost mount"); 496 rw_enter(&proclist_lock, RW_READER); 497 PROCLIST_FOREACH(p, &allproc) { 498 cwdi = p->p_cwdi; 499 if (!cwdi) 500 continue; 501 if (cwdi->cwdi_cdir == olddp) { 502 vrele(cwdi->cwdi_cdir); 503 VREF(newdp); 504 cwdi->cwdi_cdir = newdp; 505 } 506 if (cwdi->cwdi_rdir == olddp) { 507 vrele(cwdi->cwdi_rdir); 508 VREF(newdp); 509 cwdi->cwdi_rdir = newdp; 510 } 511 } 512 rw_exit(&proclist_lock); 513 if (rootvnode == olddp) { 514 vrele(rootvnode); 515 VREF(newdp); 516 rootvnode = newdp; 517 } 518 vput(newdp); 519 } 520 521 /* 522 * Unmount a file system. 523 * 524 * Note: unmount takes a path to the vnode mounted on as argument, 525 * not special file (as before). 526 */ 527 /* ARGSUSED */ 528 int 529 sys_unmount(struct lwp *l, void *v, register_t *retval) 530 { 531 struct sys_unmount_args /* { 532 syscallarg(const char *) path; 533 syscallarg(int) flags; 534 } */ *uap = v; 535 struct vnode *vp; 536 struct mount *mp; 537 int error; 538 struct nameidata nd; 539 540 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 541 SCARG(uap, path), l); 542 if ((error = namei(&nd)) != 0) 543 return (error); 544 vp = nd.ni_vp; 545 mp = vp->v_mount; 546 547 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 548 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 549 if (error) { 550 vput(vp); 551 return (error); 552 } 553 554 /* 555 * Don't allow unmounting the root file system. 556 */ 557 if (mp->mnt_flag & MNT_ROOTFS) { 558 vput(vp); 559 return (EINVAL); 560 } 561 562 /* 563 * Must be the root of the filesystem 564 */ 565 if ((vp->v_flag & VROOT) == 0) { 566 vput(vp); 567 return (EINVAL); 568 } 569 vput(vp); 570 571 /* 572 * XXX Freeze syncer. Must do this before locking the 573 * mount point. See dounmount() for details. 574 */ 575 mutex_enter(&syncer_mutex); 576 577 if (vfs_busy(mp, 0, 0)) { 578 mutex_exit(&syncer_mutex); 579 return (EBUSY); 580 } 581 582 return (dounmount(mp, SCARG(uap, flags), l)); 583 } 584 585 /* 586 * Do the actual file system unmount. File system is assumed to have been 587 * marked busy by the caller. 588 */ 589 int 590 dounmount(struct mount *mp, int flags, struct lwp *l) 591 { 592 struct vnode *coveredvp; 593 int error; 594 int async; 595 int used_syncer; 596 597 #if NVERIEXEC > 0 598 error = veriexec_unmountchk(mp); 599 if (error) 600 return (error); 601 #endif /* NVERIEXEC > 0 */ 602 603 simple_lock(&mountlist_slock); 604 vfs_unbusy(mp); 605 used_syncer = (mp->mnt_syncer != NULL); 606 607 /* 608 * XXX Syncer must be frozen when we get here. This should really 609 * be done on a per-mountpoint basis, but especially the softdep 610 * code possibly called from the syncer doesn't exactly work on a 611 * per-mountpoint basis, so the softdep code would become a maze 612 * of vfs_busy() calls. 613 * 614 * The caller of dounmount() must acquire syncer_mutex because 615 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 616 * order, and we must preserve that order to avoid deadlock. 617 * 618 * So, if the file system did not use the syncer, now is 619 * the time to release the syncer_mutex. 620 */ 621 if (used_syncer == 0) 622 mutex_exit(&syncer_mutex); 623 624 mp->mnt_iflag |= IMNT_UNMOUNT; 625 mp->mnt_unmounter = l; 626 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 627 vn_start_write(NULL, &mp, V_WAIT); 628 629 async = mp->mnt_flag & MNT_ASYNC; 630 mp->mnt_flag &= ~MNT_ASYNC; 631 cache_purgevfs(mp); /* remove cache entries for this file sys */ 632 if (mp->mnt_syncer != NULL) 633 vfs_deallocate_syncvnode(mp); 634 error = 0; 635 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 636 #if NFSS > 0 637 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 638 #endif 639 if (error == 0) 640 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 641 } 642 if (error == 0 || (flags & MNT_FORCE)) 643 error = VFS_UNMOUNT(mp, flags, l); 644 vn_finished_write(mp, 0); 645 simple_lock(&mountlist_slock); 646 if (error) { 647 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 648 (void) vfs_allocate_syncvnode(mp); 649 mp->mnt_iflag &= ~IMNT_UNMOUNT; 650 mp->mnt_unmounter = NULL; 651 mp->mnt_flag |= async; 652 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 653 &mountlist_slock); 654 if (used_syncer) 655 mutex_exit(&syncer_mutex); 656 simple_lock(&mp->mnt_slock); 657 while (mp->mnt_wcnt > 0) { 658 wakeup(mp); 659 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 660 0, &mp->mnt_slock); 661 } 662 simple_unlock(&mp->mnt_slock); 663 return (error); 664 } 665 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 666 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 667 coveredvp->v_mountedhere = NULL; 668 mp->mnt_op->vfs_refcount--; 669 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 670 panic("unmount: dangling vnode"); 671 mp->mnt_iflag |= IMNT_GONE; 672 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 673 if (coveredvp != NULLVP) 674 vrele(coveredvp); 675 mount_finispecific(mp); 676 if (used_syncer) 677 mutex_exit(&syncer_mutex); 678 simple_lock(&mp->mnt_slock); 679 while (mp->mnt_wcnt > 0) { 680 wakeup(mp); 681 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 682 } 683 simple_unlock(&mp->mnt_slock); 684 vfs_hooks_unmount(mp); 685 free(mp, M_MOUNT); 686 return (0); 687 } 688 689 /* 690 * Sync each mounted filesystem. 691 */ 692 #ifdef DEBUG 693 int syncprt = 0; 694 struct ctldebug debug0 = { "syncprt", &syncprt }; 695 #endif 696 697 /* ARGSUSED */ 698 int 699 sys_sync(struct lwp *l, void *v, register_t *retval) 700 { 701 struct mount *mp, *nmp; 702 int asyncflag; 703 704 if (l == NULL) 705 l = &lwp0; 706 707 simple_lock(&mountlist_slock); 708 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 709 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 710 nmp = mp->mnt_list.cqe_prev; 711 continue; 712 } 713 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 714 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 715 asyncflag = mp->mnt_flag & MNT_ASYNC; 716 mp->mnt_flag &= ~MNT_ASYNC; 717 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 718 if (asyncflag) 719 mp->mnt_flag |= MNT_ASYNC; 720 vn_finished_write(mp, 0); 721 } 722 simple_lock(&mountlist_slock); 723 nmp = mp->mnt_list.cqe_prev; 724 vfs_unbusy(mp); 725 726 } 727 simple_unlock(&mountlist_slock); 728 #ifdef DEBUG 729 if (syncprt) 730 vfs_bufstats(); 731 #endif /* DEBUG */ 732 return (0); 733 } 734 735 /* 736 * Change filesystem quotas. 737 */ 738 /* ARGSUSED */ 739 int 740 sys_quotactl(struct lwp *l, void *v, register_t *retval) 741 { 742 struct sys_quotactl_args /* { 743 syscallarg(const char *) path; 744 syscallarg(int) cmd; 745 syscallarg(int) uid; 746 syscallarg(void *) arg; 747 } */ *uap = v; 748 struct mount *mp; 749 int error; 750 struct nameidata nd; 751 752 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 753 if ((error = namei(&nd)) != 0) 754 return (error); 755 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 756 vrele(nd.ni_vp); 757 if (error) 758 return (error); 759 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 760 SCARG(uap, arg), l); 761 vn_finished_write(mp, 0); 762 return (error); 763 } 764 765 int 766 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 767 int root) 768 { 769 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 770 int error = 0; 771 772 /* 773 * If MNT_NOWAIT or MNT_LAZY is specified, do not 774 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 775 * overrides MNT_NOWAIT. 776 */ 777 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 778 (flags != MNT_WAIT && flags != 0)) { 779 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 780 goto done; 781 } 782 783 /* Get the filesystem stats now */ 784 memset(sp, 0, sizeof(*sp)); 785 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 786 return error; 787 } 788 789 if (cwdi->cwdi_rdir == NULL) 790 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 791 done: 792 if (cwdi->cwdi_rdir != NULL) { 793 size_t len; 794 char *bp; 795 char *path = PNBUF_GET(); 796 if (!path) 797 return ENOMEM; 798 799 bp = path + MAXPATHLEN; 800 *--bp = '\0'; 801 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 802 MAXPATHLEN / 2, 0, l); 803 if (error) { 804 PNBUF_PUT(path); 805 return error; 806 } 807 len = strlen(bp); 808 /* 809 * for mount points that are below our root, we can see 810 * them, so we fix up the pathname and return them. The 811 * rest we cannot see, so we don't allow viewing the 812 * data. 813 */ 814 if (strncmp(bp, sp->f_mntonname, len) == 0) { 815 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 816 sizeof(sp->f_mntonname)); 817 if (sp->f_mntonname[0] == '\0') 818 (void)strlcpy(sp->f_mntonname, "/", 819 sizeof(sp->f_mntonname)); 820 } else { 821 if (root) 822 (void)strlcpy(sp->f_mntonname, "/", 823 sizeof(sp->f_mntonname)); 824 else 825 error = EPERM; 826 } 827 PNBUF_PUT(path); 828 } 829 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 830 return error; 831 } 832 833 /* 834 * Get filesystem statistics. 835 */ 836 /* ARGSUSED */ 837 int 838 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 839 { 840 struct sys_statvfs1_args /* { 841 syscallarg(const char *) path; 842 syscallarg(struct statvfs *) buf; 843 syscallarg(int) flags; 844 } */ *uap = v; 845 struct mount *mp; 846 struct statvfs *sb; 847 int error; 848 struct nameidata nd; 849 850 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 851 if ((error = namei(&nd)) != 0) 852 return error; 853 mp = nd.ni_vp->v_mount; 854 vrele(nd.ni_vp); 855 sb = STATVFSBUF_GET(); 856 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 857 if (error == 0) { 858 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 859 } 860 STATVFSBUF_PUT(sb); 861 return error; 862 } 863 864 /* 865 * Get filesystem statistics. 866 */ 867 /* ARGSUSED */ 868 int 869 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 870 { 871 struct sys_fstatvfs1_args /* { 872 syscallarg(int) fd; 873 syscallarg(struct statvfs *) buf; 874 syscallarg(int) flags; 875 } */ *uap = v; 876 struct proc *p = l->l_proc; 877 struct file *fp; 878 struct mount *mp; 879 struct statvfs *sb; 880 int error; 881 882 /* getvnode() will use the descriptor for us */ 883 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 884 return (error); 885 mp = ((struct vnode *)fp->f_data)->v_mount; 886 sb = STATVFSBUF_GET(); 887 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 888 goto out; 889 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 890 out: 891 FILE_UNUSE(fp, l); 892 STATVFSBUF_PUT(sb); 893 return error; 894 } 895 896 897 /* 898 * Get statistics on all filesystems. 899 */ 900 int 901 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 902 { 903 struct sys_getvfsstat_args /* { 904 syscallarg(struct statvfs *) buf; 905 syscallarg(size_t) bufsize; 906 syscallarg(int) flags; 907 } */ *uap = v; 908 int root = 0; 909 struct proc *p = l->l_proc; 910 struct mount *mp, *nmp; 911 struct statvfs *sb; 912 struct statvfs *sfsp; 913 size_t count, maxcount; 914 int error = 0; 915 916 sb = STATVFSBUF_GET(); 917 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 918 sfsp = SCARG(uap, buf); 919 simple_lock(&mountlist_slock); 920 count = 0; 921 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 922 mp = nmp) { 923 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 924 nmp = CIRCLEQ_NEXT(mp, mnt_list); 925 continue; 926 } 927 if (sfsp && count < maxcount) { 928 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 929 if (error) { 930 simple_lock(&mountlist_slock); 931 nmp = CIRCLEQ_NEXT(mp, mnt_list); 932 vfs_unbusy(mp); 933 continue; 934 } 935 error = copyout(sb, sfsp, sizeof(*sfsp)); 936 if (error) { 937 vfs_unbusy(mp); 938 goto out; 939 } 940 sfsp++; 941 root |= strcmp(sb->f_mntonname, "/") == 0; 942 } 943 count++; 944 simple_lock(&mountlist_slock); 945 nmp = CIRCLEQ_NEXT(mp, mnt_list); 946 vfs_unbusy(mp); 947 } 948 simple_unlock(&mountlist_slock); 949 if (root == 0 && p->p_cwdi->cwdi_rdir) { 950 /* 951 * fake a root entry 952 */ 953 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 954 SCARG(uap, flags), 1)) != 0) 955 goto out; 956 if (sfsp) 957 error = copyout(sb, sfsp, sizeof(*sfsp)); 958 count++; 959 } 960 if (sfsp && count > maxcount) 961 *retval = maxcount; 962 else 963 *retval = count; 964 out: 965 STATVFSBUF_PUT(sb); 966 return error; 967 } 968 969 /* 970 * Change current working directory to a given file descriptor. 971 */ 972 /* ARGSUSED */ 973 int 974 sys_fchdir(struct lwp *l, void *v, register_t *retval) 975 { 976 struct sys_fchdir_args /* { 977 syscallarg(int) fd; 978 } */ *uap = v; 979 struct proc *p = l->l_proc; 980 struct filedesc *fdp = p->p_fd; 981 struct cwdinfo *cwdi = p->p_cwdi; 982 struct vnode *vp, *tdp; 983 struct mount *mp; 984 struct file *fp; 985 int error; 986 987 /* getvnode() will use the descriptor for us */ 988 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 989 return (error); 990 vp = (struct vnode *)fp->f_data; 991 992 VREF(vp); 993 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 994 if (vp->v_type != VDIR) 995 error = ENOTDIR; 996 else 997 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 998 while (!error && (mp = vp->v_mountedhere) != NULL) { 999 if (vfs_busy(mp, 0, 0)) 1000 continue; 1001 1002 vput(vp); 1003 error = VFS_ROOT(mp, &tdp); 1004 vfs_unbusy(mp); 1005 if (error) 1006 break; 1007 vp = tdp; 1008 } 1009 if (error) { 1010 goto out; 1011 } 1012 VOP_UNLOCK(vp, 0); 1013 1014 /* 1015 * Disallow changing to a directory not under the process's 1016 * current root directory (if there is one). 1017 */ 1018 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1019 vrele(vp); 1020 error = EPERM; /* operation not permitted */ 1021 goto out; 1022 } 1023 1024 vrele(cwdi->cwdi_cdir); 1025 cwdi->cwdi_cdir = vp; 1026 out: 1027 FILE_UNUSE(fp, l); 1028 return (error); 1029 } 1030 1031 /* 1032 * Change this process's notion of the root directory to a given file 1033 * descriptor. 1034 */ 1035 int 1036 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1037 { 1038 struct sys_fchroot_args *uap = v; 1039 struct proc *p = l->l_proc; 1040 struct filedesc *fdp = p->p_fd; 1041 struct cwdinfo *cwdi = p->p_cwdi; 1042 struct vnode *vp; 1043 struct file *fp; 1044 int error; 1045 1046 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1047 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1048 return error; 1049 /* getvnode() will use the descriptor for us */ 1050 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1051 return error; 1052 vp = (struct vnode *) fp->f_data; 1053 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1054 if (vp->v_type != VDIR) 1055 error = ENOTDIR; 1056 else 1057 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1058 VOP_UNLOCK(vp, 0); 1059 if (error) 1060 goto out; 1061 VREF(vp); 1062 1063 /* 1064 * Prevent escaping from chroot by putting the root under 1065 * the working directory. Silently chdir to / if we aren't 1066 * already there. 1067 */ 1068 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1069 /* 1070 * XXX would be more failsafe to change directory to a 1071 * deadfs node here instead 1072 */ 1073 vrele(cwdi->cwdi_cdir); 1074 VREF(vp); 1075 cwdi->cwdi_cdir = vp; 1076 } 1077 1078 if (cwdi->cwdi_rdir != NULL) 1079 vrele(cwdi->cwdi_rdir); 1080 cwdi->cwdi_rdir = vp; 1081 out: 1082 FILE_UNUSE(fp, l); 1083 return (error); 1084 } 1085 1086 /* 1087 * Change current working directory (``.''). 1088 */ 1089 /* ARGSUSED */ 1090 int 1091 sys_chdir(struct lwp *l, void *v, register_t *retval) 1092 { 1093 struct sys_chdir_args /* { 1094 syscallarg(const char *) path; 1095 } */ *uap = v; 1096 struct proc *p = l->l_proc; 1097 struct cwdinfo *cwdi = p->p_cwdi; 1098 int error; 1099 struct nameidata nd; 1100 1101 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1102 SCARG(uap, path), l); 1103 if ((error = change_dir(&nd, l)) != 0) 1104 return (error); 1105 vrele(cwdi->cwdi_cdir); 1106 cwdi->cwdi_cdir = nd.ni_vp; 1107 return (0); 1108 } 1109 1110 /* 1111 * Change notion of root (``/'') directory. 1112 */ 1113 /* ARGSUSED */ 1114 int 1115 sys_chroot(struct lwp *l, void *v, register_t *retval) 1116 { 1117 struct sys_chroot_args /* { 1118 syscallarg(const char *) path; 1119 } */ *uap = v; 1120 struct proc *p = l->l_proc; 1121 struct cwdinfo *cwdi = p->p_cwdi; 1122 struct vnode *vp; 1123 int error; 1124 struct nameidata nd; 1125 1126 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1127 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1128 return (error); 1129 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1130 SCARG(uap, path), l); 1131 if ((error = change_dir(&nd, l)) != 0) 1132 return (error); 1133 if (cwdi->cwdi_rdir != NULL) 1134 vrele(cwdi->cwdi_rdir); 1135 vp = nd.ni_vp; 1136 cwdi->cwdi_rdir = vp; 1137 1138 /* 1139 * Prevent escaping from chroot by putting the root under 1140 * the working directory. Silently chdir to / if we aren't 1141 * already there. 1142 */ 1143 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1144 /* 1145 * XXX would be more failsafe to change directory to a 1146 * deadfs node here instead 1147 */ 1148 vrele(cwdi->cwdi_cdir); 1149 VREF(vp); 1150 cwdi->cwdi_cdir = vp; 1151 } 1152 1153 return (0); 1154 } 1155 1156 /* 1157 * Common routine for chroot and chdir. 1158 */ 1159 static int 1160 change_dir(struct nameidata *ndp, struct lwp *l) 1161 { 1162 struct vnode *vp; 1163 int error; 1164 1165 if ((error = namei(ndp)) != 0) 1166 return (error); 1167 vp = ndp->ni_vp; 1168 if (vp->v_type != VDIR) 1169 error = ENOTDIR; 1170 else 1171 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1172 1173 if (error) 1174 vput(vp); 1175 else 1176 VOP_UNLOCK(vp, 0); 1177 return (error); 1178 } 1179 1180 /* 1181 * Check permissions, allocate an open file structure, 1182 * and call the device open routine if any. 1183 */ 1184 int 1185 sys_open(struct lwp *l, void *v, register_t *retval) 1186 { 1187 struct sys_open_args /* { 1188 syscallarg(const char *) path; 1189 syscallarg(int) flags; 1190 syscallarg(int) mode; 1191 } */ *uap = v; 1192 struct proc *p = l->l_proc; 1193 struct cwdinfo *cwdi = p->p_cwdi; 1194 struct filedesc *fdp = p->p_fd; 1195 struct file *fp; 1196 struct vnode *vp; 1197 int flags, cmode; 1198 int type, indx, error; 1199 struct flock lf; 1200 struct nameidata nd; 1201 1202 flags = FFLAGS(SCARG(uap, flags)); 1203 if ((flags & (FREAD | FWRITE)) == 0) 1204 return (EINVAL); 1205 /* falloc() will use the file descriptor for us */ 1206 if ((error = falloc(l, &fp, &indx)) != 0) 1207 return (error); 1208 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1209 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1210 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1211 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1212 FILE_UNUSE(fp, l); 1213 fdp->fd_ofiles[indx] = NULL; 1214 ffree(fp); 1215 if ((error == EDUPFD || error == EMOVEFD) && 1216 l->l_dupfd >= 0 && /* XXX from fdopen */ 1217 (error = 1218 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1219 *retval = indx; 1220 return (0); 1221 } 1222 if (error == ERESTART) 1223 error = EINTR; 1224 fdremove(fdp, indx); 1225 return (error); 1226 } 1227 l->l_dupfd = 0; 1228 vp = nd.ni_vp; 1229 fp->f_flag = flags & FMASK; 1230 fp->f_type = DTYPE_VNODE; 1231 fp->f_ops = &vnops; 1232 fp->f_data = vp; 1233 if (flags & (O_EXLOCK | O_SHLOCK)) { 1234 lf.l_whence = SEEK_SET; 1235 lf.l_start = 0; 1236 lf.l_len = 0; 1237 if (flags & O_EXLOCK) 1238 lf.l_type = F_WRLCK; 1239 else 1240 lf.l_type = F_RDLCK; 1241 type = F_FLOCK; 1242 if ((flags & FNONBLOCK) == 0) 1243 type |= F_WAIT; 1244 VOP_UNLOCK(vp, 0); 1245 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1246 if (error) { 1247 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1248 FILE_UNUSE(fp, l); 1249 ffree(fp); 1250 fdremove(fdp, indx); 1251 return (error); 1252 } 1253 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1254 fp->f_flag |= FHASLOCK; 1255 } 1256 VOP_UNLOCK(vp, 0); 1257 *retval = indx; 1258 FILE_SET_MATURE(fp); 1259 FILE_UNUSE(fp, l); 1260 return (0); 1261 } 1262 1263 static void 1264 vfs__fhfree(fhandle_t *fhp) 1265 { 1266 size_t fhsize; 1267 1268 if (fhp == NULL) { 1269 return; 1270 } 1271 fhsize = FHANDLE_SIZE(fhp); 1272 kmem_free(fhp, fhsize); 1273 } 1274 1275 /* 1276 * vfs_composefh: compose a filehandle. 1277 */ 1278 1279 int 1280 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1281 { 1282 struct mount *mp; 1283 struct fid *fidp; 1284 int error; 1285 size_t needfhsize; 1286 size_t fidsize; 1287 1288 mp = vp->v_mount; 1289 fidp = NULL; 1290 if (*fh_size < FHANDLE_SIZE_MIN) { 1291 fidsize = 0; 1292 } else { 1293 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1294 if (fhp != NULL) { 1295 memset(fhp, 0, *fh_size); 1296 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1297 fidp = &fhp->fh_fid; 1298 } 1299 } 1300 error = VFS_VPTOFH(vp, fidp, &fidsize); 1301 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1302 if (error == 0 && *fh_size < needfhsize) { 1303 error = E2BIG; 1304 } 1305 *fh_size = needfhsize; 1306 return error; 1307 } 1308 1309 int 1310 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1311 { 1312 struct mount *mp; 1313 fhandle_t *fhp; 1314 size_t fhsize; 1315 size_t fidsize; 1316 int error; 1317 1318 *fhpp = NULL; 1319 mp = vp->v_mount; 1320 fidsize = 0; 1321 error = VFS_VPTOFH(vp, NULL, &fidsize); 1322 KASSERT(error != 0); 1323 if (error != E2BIG) { 1324 goto out; 1325 } 1326 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1327 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1328 if (fhp == NULL) { 1329 error = ENOMEM; 1330 goto out; 1331 } 1332 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1333 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1334 if (error == 0) { 1335 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1336 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1337 *fhpp = fhp; 1338 } else { 1339 kmem_free(fhp, fhsize); 1340 } 1341 out: 1342 return error; 1343 } 1344 1345 void 1346 vfs_composefh_free(fhandle_t *fhp) 1347 { 1348 1349 vfs__fhfree(fhp); 1350 } 1351 1352 /* 1353 * vfs_fhtovp: lookup a vnode by a filehandle. 1354 */ 1355 1356 int 1357 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1358 { 1359 struct mount *mp; 1360 int error; 1361 1362 *vpp = NULL; 1363 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1364 if (mp == NULL) { 1365 error = ESTALE; 1366 goto out; 1367 } 1368 if (mp->mnt_op->vfs_fhtovp == NULL) { 1369 error = EOPNOTSUPP; 1370 goto out; 1371 } 1372 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1373 out: 1374 return error; 1375 } 1376 1377 /* 1378 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1379 * the needed size. 1380 */ 1381 1382 int 1383 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1384 { 1385 fhandle_t *fhp; 1386 int error; 1387 1388 *fhpp = NULL; 1389 if (fhsize > FHANDLE_SIZE_MAX) { 1390 return EINVAL; 1391 } 1392 if (fhsize < FHANDLE_SIZE_MIN) { 1393 return EINVAL; 1394 } 1395 again: 1396 fhp = kmem_alloc(fhsize, KM_SLEEP); 1397 if (fhp == NULL) { 1398 return ENOMEM; 1399 } 1400 error = copyin(ufhp, fhp, fhsize); 1401 if (error == 0) { 1402 /* XXX this check shouldn't be here */ 1403 if (FHANDLE_SIZE(fhp) == fhsize) { 1404 *fhpp = fhp; 1405 return 0; 1406 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1407 /* 1408 * a kludge for nfsv2 padded handles. 1409 */ 1410 size_t sz; 1411 1412 sz = FHANDLE_SIZE(fhp); 1413 kmem_free(fhp, fhsize); 1414 fhsize = sz; 1415 goto again; 1416 } else { 1417 /* 1418 * userland told us wrong size. 1419 */ 1420 error = EINVAL; 1421 } 1422 } 1423 kmem_free(fhp, fhsize); 1424 return error; 1425 } 1426 1427 void 1428 vfs_copyinfh_free(fhandle_t *fhp) 1429 { 1430 1431 vfs__fhfree(fhp); 1432 } 1433 1434 /* 1435 * Get file handle system call 1436 */ 1437 int 1438 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1439 { 1440 struct sys___getfh30_args /* { 1441 syscallarg(char *) fname; 1442 syscallarg(fhandle_t *) fhp; 1443 syscallarg(size_t *) fh_size; 1444 } */ *uap = v; 1445 struct vnode *vp; 1446 fhandle_t *fh; 1447 int error; 1448 struct nameidata nd; 1449 size_t sz; 1450 size_t usz; 1451 1452 /* 1453 * Must be super user 1454 */ 1455 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1456 0, NULL, NULL, NULL); 1457 if (error) 1458 return (error); 1459 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1460 SCARG(uap, fname), l); 1461 error = namei(&nd); 1462 if (error) 1463 return (error); 1464 vp = nd.ni_vp; 1465 error = vfs_composefh_alloc(vp, &fh); 1466 vput(vp); 1467 if (error != 0) { 1468 goto out; 1469 } 1470 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1471 if (error != 0) { 1472 goto out; 1473 } 1474 sz = FHANDLE_SIZE(fh); 1475 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1476 if (error != 0) { 1477 goto out; 1478 } 1479 if (usz >= sz) { 1480 error = copyout(fh, SCARG(uap, fhp), sz); 1481 } else { 1482 error = E2BIG; 1483 } 1484 out: 1485 vfs_composefh_free(fh); 1486 return (error); 1487 } 1488 1489 /* 1490 * Open a file given a file handle. 1491 * 1492 * Check permissions, allocate an open file structure, 1493 * and call the device open routine if any. 1494 */ 1495 1496 int 1497 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1498 register_t *retval) 1499 { 1500 struct filedesc *fdp = l->l_proc->p_fd; 1501 struct file *fp; 1502 struct vnode *vp = NULL; 1503 struct mount *mp; 1504 kauth_cred_t cred = l->l_cred; 1505 struct file *nfp; 1506 int type, indx, error=0; 1507 struct flock lf; 1508 struct vattr va; 1509 fhandle_t *fh; 1510 int flags; 1511 1512 /* 1513 * Must be super user 1514 */ 1515 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1516 0, NULL, NULL, NULL))) 1517 return (error); 1518 1519 flags = FFLAGS(oflags); 1520 if ((flags & (FREAD | FWRITE)) == 0) 1521 return (EINVAL); 1522 if ((flags & O_CREAT)) 1523 return (EINVAL); 1524 /* falloc() will use the file descriptor for us */ 1525 if ((error = falloc(l, &nfp, &indx)) != 0) 1526 return (error); 1527 fp = nfp; 1528 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1529 if (error != 0) { 1530 goto bad; 1531 } 1532 error = vfs_fhtovp(fh, &vp); 1533 if (error != 0) { 1534 goto bad; 1535 } 1536 1537 /* Now do an effective vn_open */ 1538 1539 if (vp->v_type == VSOCK) { 1540 error = EOPNOTSUPP; 1541 goto bad; 1542 } 1543 if (flags & FREAD) { 1544 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1545 goto bad; 1546 } 1547 if (flags & (FWRITE | O_TRUNC)) { 1548 if (vp->v_type == VDIR) { 1549 error = EISDIR; 1550 goto bad; 1551 } 1552 if ((error = vn_writechk(vp)) != 0 || 1553 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1554 goto bad; 1555 } 1556 if (flags & O_TRUNC) { 1557 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1558 goto bad; 1559 VOP_UNLOCK(vp, 0); /* XXX */ 1560 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1561 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1562 VATTR_NULL(&va); 1563 va.va_size = 0; 1564 error = VOP_SETATTR(vp, &va, cred, l); 1565 vn_finished_write(mp, 0); 1566 if (error) 1567 goto bad; 1568 } 1569 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1570 goto bad; 1571 if (vp->v_type == VREG && 1572 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1573 error = EIO; 1574 goto bad; 1575 } 1576 if (flags & FWRITE) 1577 vp->v_writecount++; 1578 1579 /* done with modified vn_open, now finish what sys_open does. */ 1580 1581 fp->f_flag = flags & FMASK; 1582 fp->f_type = DTYPE_VNODE; 1583 fp->f_ops = &vnops; 1584 fp->f_data = vp; 1585 if (flags & (O_EXLOCK | O_SHLOCK)) { 1586 lf.l_whence = SEEK_SET; 1587 lf.l_start = 0; 1588 lf.l_len = 0; 1589 if (flags & O_EXLOCK) 1590 lf.l_type = F_WRLCK; 1591 else 1592 lf.l_type = F_RDLCK; 1593 type = F_FLOCK; 1594 if ((flags & FNONBLOCK) == 0) 1595 type |= F_WAIT; 1596 VOP_UNLOCK(vp, 0); 1597 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1598 if (error) { 1599 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1600 FILE_UNUSE(fp, l); 1601 ffree(fp); 1602 fdremove(fdp, indx); 1603 return (error); 1604 } 1605 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1606 fp->f_flag |= FHASLOCK; 1607 } 1608 VOP_UNLOCK(vp, 0); 1609 *retval = indx; 1610 FILE_SET_MATURE(fp); 1611 FILE_UNUSE(fp, l); 1612 vfs_copyinfh_free(fh); 1613 return (0); 1614 1615 bad: 1616 FILE_UNUSE(fp, l); 1617 ffree(fp); 1618 fdremove(fdp, indx); 1619 if (vp != NULL) 1620 vput(vp); 1621 vfs_copyinfh_free(fh); 1622 return (error); 1623 } 1624 1625 int 1626 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1627 { 1628 struct sys___fhopen40_args /* { 1629 syscallarg(const void *) fhp; 1630 syscallarg(size_t) fh_size; 1631 syscallarg(int) flags; 1632 } */ *uap = v; 1633 1634 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1635 SCARG(uap, flags), retval); 1636 } 1637 1638 int 1639 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1640 register_t *retval) 1641 { 1642 struct stat sb; 1643 int error; 1644 fhandle_t *fh; 1645 struct vnode *vp; 1646 1647 /* 1648 * Must be super user 1649 */ 1650 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1651 0, NULL, NULL, NULL))) 1652 return (error); 1653 1654 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1655 if (error != 0) { 1656 goto bad; 1657 } 1658 error = vfs_fhtovp(fh, &vp); 1659 if (error != 0) { 1660 goto bad; 1661 } 1662 error = vn_stat(vp, &sb, l); 1663 vput(vp); 1664 if (error) { 1665 goto bad; 1666 } 1667 error = copyout(&sb, sbp, sizeof(sb)); 1668 bad: 1669 vfs_copyinfh_free(fh); 1670 return error; 1671 } 1672 1673 1674 /* ARGSUSED */ 1675 int 1676 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1677 { 1678 struct sys___fhstat40_args /* { 1679 syscallarg(const void *) fhp; 1680 syscallarg(size_t) fh_size; 1681 syscallarg(struct stat *) sb; 1682 } */ *uap = v; 1683 1684 return dofhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), SCARG(uap, sb), 1685 retval); 1686 } 1687 1688 int 1689 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1690 int flags, register_t *retval) 1691 { 1692 struct statvfs *sb = NULL; 1693 fhandle_t *fh; 1694 struct mount *mp; 1695 struct vnode *vp; 1696 int error; 1697 1698 /* 1699 * Must be super user 1700 */ 1701 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1702 0, NULL, NULL, NULL))) 1703 return error; 1704 1705 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1706 if (error != 0) { 1707 goto out; 1708 } 1709 error = vfs_fhtovp(fh, &vp); 1710 if (error != 0) { 1711 goto out; 1712 } 1713 mp = vp->v_mount; 1714 sb = STATVFSBUF_GET(); 1715 if ((error = dostatvfs(mp, sb, l, flags, 1)) != 0) { 1716 vput(vp); 1717 goto out; 1718 } 1719 vput(vp); 1720 error = copyout(sb, buf, sizeof(*sb)); 1721 out: 1722 if (sb != NULL) { 1723 STATVFSBUF_PUT(sb); 1724 } 1725 vfs_copyinfh_free(fh); 1726 return error; 1727 } 1728 1729 /* ARGSUSED */ 1730 int 1731 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1732 { 1733 struct sys___fhstatvfs140_args /* { 1734 syscallarg(const void *) fhp; 1735 syscallarg(size_t) fh_size; 1736 syscallarg(struct statvfs *) buf; 1737 syscallarg(int) flags; 1738 } */ *uap = v; 1739 1740 return dofhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1741 SCARG(uap, buf), SCARG(uap, flags), retval); 1742 } 1743 1744 /* 1745 * Create a special file. 1746 */ 1747 /* ARGSUSED */ 1748 int 1749 sys_mknod(struct lwp *l, void *v, register_t *retval) 1750 { 1751 struct sys_mknod_args /* { 1752 syscallarg(const char *) path; 1753 syscallarg(int) mode; 1754 syscallarg(int) dev; 1755 } */ *uap = v; 1756 struct proc *p = l->l_proc; 1757 struct vnode *vp; 1758 struct mount *mp; 1759 struct vattr vattr; 1760 int error, optype; 1761 struct nameidata nd; 1762 1763 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1764 0, NULL, NULL, NULL)) != 0) 1765 return (error); 1766 1767 optype = VOP_MKNOD_DESCOFFSET; 1768 restart: 1769 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1770 if ((error = namei(&nd)) != 0) 1771 return (error); 1772 vp = nd.ni_vp; 1773 if (vp != NULL) 1774 error = EEXIST; 1775 else { 1776 VATTR_NULL(&vattr); 1777 vattr.va_mode = 1778 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1779 vattr.va_rdev = SCARG(uap, dev); 1780 1781 switch (SCARG(uap, mode) & S_IFMT) { 1782 case S_IFMT: /* used by badsect to flag bad sectors */ 1783 vattr.va_type = VBAD; 1784 break; 1785 case S_IFCHR: 1786 vattr.va_type = VCHR; 1787 break; 1788 case S_IFBLK: 1789 vattr.va_type = VBLK; 1790 break; 1791 case S_IFWHT: 1792 optype = VOP_WHITEOUT_DESCOFFSET; 1793 break; 1794 case S_IFREG: 1795 vattr.va_type = VREG; 1796 vattr.va_rdev = VNOVAL; 1797 optype = VOP_CREATE_DESCOFFSET; 1798 break; 1799 default: 1800 error = EINVAL; 1801 break; 1802 } 1803 } 1804 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1805 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1806 if (nd.ni_dvp == vp) 1807 vrele(nd.ni_dvp); 1808 else 1809 vput(nd.ni_dvp); 1810 if (vp) 1811 vrele(vp); 1812 if ((error = vn_start_write(NULL, &mp, 1813 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1814 return (error); 1815 goto restart; 1816 } 1817 if (!error) { 1818 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1819 switch (optype) { 1820 case VOP_WHITEOUT_DESCOFFSET: 1821 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1822 if (error) 1823 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1824 vput(nd.ni_dvp); 1825 break; 1826 1827 case VOP_MKNOD_DESCOFFSET: 1828 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1829 &nd.ni_cnd, &vattr); 1830 if (error == 0) 1831 vput(nd.ni_vp); 1832 break; 1833 1834 case VOP_CREATE_DESCOFFSET: 1835 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1836 &nd.ni_cnd, &vattr); 1837 if (error == 0) 1838 vput(nd.ni_vp); 1839 break; 1840 } 1841 } else { 1842 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1843 if (nd.ni_dvp == vp) 1844 vrele(nd.ni_dvp); 1845 else 1846 vput(nd.ni_dvp); 1847 if (vp) 1848 vrele(vp); 1849 } 1850 vn_finished_write(mp, 0); 1851 return (error); 1852 } 1853 1854 /* 1855 * Create a named pipe. 1856 */ 1857 /* ARGSUSED */ 1858 int 1859 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1860 { 1861 struct sys_mkfifo_args /* { 1862 syscallarg(const char *) path; 1863 syscallarg(int) mode; 1864 } */ *uap = v; 1865 struct proc *p = l->l_proc; 1866 struct mount *mp; 1867 struct vattr vattr; 1868 int error; 1869 struct nameidata nd; 1870 1871 restart: 1872 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1873 if ((error = namei(&nd)) != 0) 1874 return (error); 1875 if (nd.ni_vp != NULL) { 1876 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1877 if (nd.ni_dvp == nd.ni_vp) 1878 vrele(nd.ni_dvp); 1879 else 1880 vput(nd.ni_dvp); 1881 vrele(nd.ni_vp); 1882 return (EEXIST); 1883 } 1884 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 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 if (nd.ni_vp) 1891 vrele(nd.ni_vp); 1892 if ((error = vn_start_write(NULL, &mp, 1893 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1894 return (error); 1895 goto restart; 1896 } 1897 VATTR_NULL(&vattr); 1898 vattr.va_type = VFIFO; 1899 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1900 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1901 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1902 if (error == 0) 1903 vput(nd.ni_vp); 1904 vn_finished_write(mp, 0); 1905 return (error); 1906 } 1907 1908 /* 1909 * Make a hard file link. 1910 */ 1911 /* ARGSUSED */ 1912 int 1913 sys_link(struct lwp *l, void *v, register_t *retval) 1914 { 1915 struct sys_link_args /* { 1916 syscallarg(const char *) path; 1917 syscallarg(const char *) link; 1918 } */ *uap = v; 1919 struct vnode *vp; 1920 struct mount *mp; 1921 struct nameidata nd; 1922 int error; 1923 1924 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1925 if ((error = namei(&nd)) != 0) 1926 return (error); 1927 vp = nd.ni_vp; 1928 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1929 vrele(vp); 1930 return (error); 1931 } 1932 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1933 if ((error = namei(&nd)) != 0) 1934 goto out; 1935 if (nd.ni_vp) { 1936 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1937 if (nd.ni_dvp == nd.ni_vp) 1938 vrele(nd.ni_dvp); 1939 else 1940 vput(nd.ni_dvp); 1941 vrele(nd.ni_vp); 1942 error = EEXIST; 1943 goto out; 1944 } 1945 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1946 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1947 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1948 out: 1949 vrele(vp); 1950 vn_finished_write(mp, 0); 1951 return (error); 1952 } 1953 1954 /* 1955 * Make a symbolic link. 1956 */ 1957 /* ARGSUSED */ 1958 int 1959 sys_symlink(struct lwp *l, void *v, register_t *retval) 1960 { 1961 struct sys_symlink_args /* { 1962 syscallarg(const char *) path; 1963 syscallarg(const char *) link; 1964 } */ *uap = v; 1965 struct proc *p = l->l_proc; 1966 struct mount *mp; 1967 struct vattr vattr; 1968 char *path; 1969 int error; 1970 struct nameidata nd; 1971 1972 path = PNBUF_GET(); 1973 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1974 if (error) 1975 goto out; 1976 restart: 1977 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1978 if ((error = namei(&nd)) != 0) 1979 goto out; 1980 if (nd.ni_vp) { 1981 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1982 if (nd.ni_dvp == nd.ni_vp) 1983 vrele(nd.ni_dvp); 1984 else 1985 vput(nd.ni_dvp); 1986 vrele(nd.ni_vp); 1987 error = EEXIST; 1988 goto out; 1989 } 1990 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1991 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1992 if (nd.ni_dvp == nd.ni_vp) 1993 vrele(nd.ni_dvp); 1994 else 1995 vput(nd.ni_dvp); 1996 if ((error = vn_start_write(NULL, &mp, 1997 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1998 return (error); 1999 goto restart; 2000 } 2001 VATTR_NULL(&vattr); 2002 vattr.va_type = VLNK; 2003 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2004 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2005 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2006 if (error == 0) 2007 vput(nd.ni_vp); 2008 vn_finished_write(mp, 0); 2009 out: 2010 PNBUF_PUT(path); 2011 return (error); 2012 } 2013 2014 /* 2015 * Delete a whiteout from the filesystem. 2016 */ 2017 /* ARGSUSED */ 2018 int 2019 sys_undelete(struct lwp *l, void *v, register_t *retval) 2020 { 2021 struct sys_undelete_args /* { 2022 syscallarg(const char *) path; 2023 } */ *uap = v; 2024 int error; 2025 struct mount *mp; 2026 struct nameidata nd; 2027 2028 restart: 2029 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2030 SCARG(uap, path), l); 2031 error = namei(&nd); 2032 if (error) 2033 return (error); 2034 2035 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2036 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2037 if (nd.ni_dvp == nd.ni_vp) 2038 vrele(nd.ni_dvp); 2039 else 2040 vput(nd.ni_dvp); 2041 if (nd.ni_vp) 2042 vrele(nd.ni_vp); 2043 return (EEXIST); 2044 } 2045 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2046 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2047 if (nd.ni_dvp == nd.ni_vp) 2048 vrele(nd.ni_dvp); 2049 else 2050 vput(nd.ni_dvp); 2051 if ((error = vn_start_write(NULL, &mp, 2052 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2053 return (error); 2054 goto restart; 2055 } 2056 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2057 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2058 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2059 vput(nd.ni_dvp); 2060 vn_finished_write(mp, 0); 2061 return (error); 2062 } 2063 2064 /* 2065 * Delete a name from the filesystem. 2066 */ 2067 /* ARGSUSED */ 2068 int 2069 sys_unlink(struct lwp *l, void *v, register_t *retval) 2070 { 2071 struct sys_unlink_args /* { 2072 syscallarg(const char *) path; 2073 } */ *uap = v; 2074 struct mount *mp; 2075 struct vnode *vp; 2076 int error; 2077 struct nameidata nd; 2078 #if NVERIEXEC > 0 2079 pathname_t pathbuf = NULL; 2080 #endif /* NVERIEXEC > 0 */ 2081 2082 restart: 2083 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2084 SCARG(uap, path), l); 2085 if ((error = namei(&nd)) != 0) 2086 return (error); 2087 vp = nd.ni_vp; 2088 2089 /* 2090 * The root of a mounted filesystem cannot be deleted. 2091 */ 2092 if (vp->v_flag & VROOT) { 2093 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2094 if (nd.ni_dvp == vp) 2095 vrele(nd.ni_dvp); 2096 else 2097 vput(nd.ni_dvp); 2098 vput(vp); 2099 error = EBUSY; 2100 goto out; 2101 } 2102 2103 #if NVERIEXEC > 0 2104 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2105 2106 /* Handle remove requests for veriexec entries. */ 2107 if (!error) { 2108 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2109 pathname_put(pathbuf); 2110 } 2111 2112 if (error) { 2113 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2114 if (nd.ni_dvp == vp) 2115 vrele(nd.ni_dvp); 2116 else 2117 vput(nd.ni_dvp); 2118 vput(vp); 2119 goto out; 2120 } 2121 #endif /* NVERIEXEC > 0 */ 2122 2123 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2124 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2125 if (nd.ni_dvp == vp) 2126 vrele(nd.ni_dvp); 2127 else 2128 vput(nd.ni_dvp); 2129 vput(vp); 2130 if ((error = vn_start_write(NULL, &mp, 2131 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2132 return (error); 2133 goto restart; 2134 } 2135 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2136 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2137 #ifdef FILEASSOC 2138 (void)fileassoc_file_delete(vp); 2139 #endif /* FILEASSOC */ 2140 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2141 vn_finished_write(mp, 0); 2142 out: 2143 return (error); 2144 } 2145 2146 /* 2147 * Reposition read/write file offset. 2148 */ 2149 int 2150 sys_lseek(struct lwp *l, void *v, register_t *retval) 2151 { 2152 struct sys_lseek_args /* { 2153 syscallarg(int) fd; 2154 syscallarg(int) pad; 2155 syscallarg(off_t) offset; 2156 syscallarg(int) whence; 2157 } */ *uap = v; 2158 struct proc *p = l->l_proc; 2159 kauth_cred_t cred = l->l_cred; 2160 struct filedesc *fdp = p->p_fd; 2161 struct file *fp; 2162 struct vnode *vp; 2163 struct vattr vattr; 2164 off_t newoff; 2165 int error; 2166 2167 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2168 return (EBADF); 2169 2170 FILE_USE(fp); 2171 2172 vp = (struct vnode *)fp->f_data; 2173 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2174 error = ESPIPE; 2175 goto out; 2176 } 2177 2178 switch (SCARG(uap, whence)) { 2179 case SEEK_CUR: 2180 newoff = fp->f_offset + SCARG(uap, offset); 2181 break; 2182 case SEEK_END: 2183 error = VOP_GETATTR(vp, &vattr, cred, l); 2184 if (error) 2185 goto out; 2186 newoff = SCARG(uap, offset) + vattr.va_size; 2187 break; 2188 case SEEK_SET: 2189 newoff = SCARG(uap, offset); 2190 break; 2191 default: 2192 error = EINVAL; 2193 goto out; 2194 } 2195 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2196 goto out; 2197 2198 *(off_t *)retval = fp->f_offset = newoff; 2199 out: 2200 FILE_UNUSE(fp, l); 2201 return (error); 2202 } 2203 2204 /* 2205 * Positional read system call. 2206 */ 2207 int 2208 sys_pread(struct lwp *l, void *v, register_t *retval) 2209 { 2210 struct sys_pread_args /* { 2211 syscallarg(int) fd; 2212 syscallarg(void *) buf; 2213 syscallarg(size_t) nbyte; 2214 syscallarg(off_t) offset; 2215 } */ *uap = v; 2216 struct proc *p = l->l_proc; 2217 struct filedesc *fdp = p->p_fd; 2218 struct file *fp; 2219 struct vnode *vp; 2220 off_t offset; 2221 int error, fd = SCARG(uap, fd); 2222 2223 if ((fp = fd_getfile(fdp, fd)) == NULL) 2224 return (EBADF); 2225 2226 if ((fp->f_flag & FREAD) == 0) { 2227 simple_unlock(&fp->f_slock); 2228 return (EBADF); 2229 } 2230 2231 FILE_USE(fp); 2232 2233 vp = (struct vnode *)fp->f_data; 2234 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2235 error = ESPIPE; 2236 goto out; 2237 } 2238 2239 offset = SCARG(uap, offset); 2240 2241 /* 2242 * XXX This works because no file systems actually 2243 * XXX take any action on the seek operation. 2244 */ 2245 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2246 goto out; 2247 2248 /* dofileread() will unuse the descriptor for us */ 2249 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2250 &offset, 0, retval)); 2251 2252 out: 2253 FILE_UNUSE(fp, l); 2254 return (error); 2255 } 2256 2257 /* 2258 * Positional scatter read system call. 2259 */ 2260 int 2261 sys_preadv(struct lwp *l, void *v, register_t *retval) 2262 { 2263 struct sys_preadv_args /* { 2264 syscallarg(int) fd; 2265 syscallarg(const struct iovec *) iovp; 2266 syscallarg(int) iovcnt; 2267 syscallarg(off_t) offset; 2268 } */ *uap = v; 2269 struct proc *p = l->l_proc; 2270 struct filedesc *fdp = p->p_fd; 2271 struct file *fp; 2272 struct vnode *vp; 2273 off_t offset; 2274 int error, fd = SCARG(uap, fd); 2275 2276 if ((fp = fd_getfile(fdp, fd)) == NULL) 2277 return (EBADF); 2278 2279 if ((fp->f_flag & FREAD) == 0) { 2280 simple_unlock(&fp->f_slock); 2281 return (EBADF); 2282 } 2283 2284 FILE_USE(fp); 2285 2286 vp = (struct vnode *)fp->f_data; 2287 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2288 error = ESPIPE; 2289 goto out; 2290 } 2291 2292 offset = SCARG(uap, offset); 2293 2294 /* 2295 * XXX This works because no file systems actually 2296 * XXX take any action on the seek operation. 2297 */ 2298 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2299 goto out; 2300 2301 /* dofilereadv() will unuse the descriptor for us */ 2302 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2303 &offset, 0, retval)); 2304 2305 out: 2306 FILE_UNUSE(fp, l); 2307 return (error); 2308 } 2309 2310 /* 2311 * Positional write system call. 2312 */ 2313 int 2314 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2315 { 2316 struct sys_pwrite_args /* { 2317 syscallarg(int) fd; 2318 syscallarg(const void *) buf; 2319 syscallarg(size_t) nbyte; 2320 syscallarg(off_t) offset; 2321 } */ *uap = v; 2322 struct proc *p = l->l_proc; 2323 struct filedesc *fdp = p->p_fd; 2324 struct file *fp; 2325 struct vnode *vp; 2326 off_t offset; 2327 int error, fd = SCARG(uap, fd); 2328 2329 if ((fp = fd_getfile(fdp, fd)) == NULL) 2330 return (EBADF); 2331 2332 if ((fp->f_flag & FWRITE) == 0) { 2333 simple_unlock(&fp->f_slock); 2334 return (EBADF); 2335 } 2336 2337 FILE_USE(fp); 2338 2339 vp = (struct vnode *)fp->f_data; 2340 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2341 error = ESPIPE; 2342 goto out; 2343 } 2344 2345 offset = SCARG(uap, offset); 2346 2347 /* 2348 * XXX This works because no file systems actually 2349 * XXX take any action on the seek operation. 2350 */ 2351 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2352 goto out; 2353 2354 /* dofilewrite() will unuse the descriptor for us */ 2355 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2356 &offset, 0, retval)); 2357 2358 out: 2359 FILE_UNUSE(fp, l); 2360 return (error); 2361 } 2362 2363 /* 2364 * Positional gather write system call. 2365 */ 2366 int 2367 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2368 { 2369 struct sys_pwritev_args /* { 2370 syscallarg(int) fd; 2371 syscallarg(const struct iovec *) iovp; 2372 syscallarg(int) iovcnt; 2373 syscallarg(off_t) offset; 2374 } */ *uap = v; 2375 struct proc *p = l->l_proc; 2376 struct filedesc *fdp = p->p_fd; 2377 struct file *fp; 2378 struct vnode *vp; 2379 off_t offset; 2380 int error, fd = SCARG(uap, fd); 2381 2382 if ((fp = fd_getfile(fdp, fd)) == NULL) 2383 return (EBADF); 2384 2385 if ((fp->f_flag & FWRITE) == 0) { 2386 simple_unlock(&fp->f_slock); 2387 return (EBADF); 2388 } 2389 2390 FILE_USE(fp); 2391 2392 vp = (struct vnode *)fp->f_data; 2393 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2394 error = ESPIPE; 2395 goto out; 2396 } 2397 2398 offset = SCARG(uap, offset); 2399 2400 /* 2401 * XXX This works because no file systems actually 2402 * XXX take any action on the seek operation. 2403 */ 2404 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2405 goto out; 2406 2407 /* dofilewritev() will unuse the descriptor for us */ 2408 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2409 &offset, 0, retval)); 2410 2411 out: 2412 FILE_UNUSE(fp, l); 2413 return (error); 2414 } 2415 2416 /* 2417 * Check access permissions. 2418 */ 2419 int 2420 sys_access(struct lwp *l, void *v, register_t *retval) 2421 { 2422 struct sys_access_args /* { 2423 syscallarg(const char *) path; 2424 syscallarg(int) flags; 2425 } */ *uap = v; 2426 kauth_cred_t cred; 2427 struct vnode *vp; 2428 int error, flags; 2429 struct nameidata nd; 2430 2431 cred = kauth_cred_dup(l->l_cred); 2432 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2433 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2434 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2435 SCARG(uap, path), l); 2436 /* Override default credentials */ 2437 nd.ni_cnd.cn_cred = cred; 2438 if ((error = namei(&nd)) != 0) 2439 goto out; 2440 vp = nd.ni_vp; 2441 2442 /* Flags == 0 means only check for existence. */ 2443 if (SCARG(uap, flags)) { 2444 flags = 0; 2445 if (SCARG(uap, flags) & R_OK) 2446 flags |= VREAD; 2447 if (SCARG(uap, flags) & W_OK) 2448 flags |= VWRITE; 2449 if (SCARG(uap, flags) & X_OK) 2450 flags |= VEXEC; 2451 2452 error = VOP_ACCESS(vp, flags, cred, l); 2453 if (!error && (flags & VWRITE)) 2454 error = vn_writechk(vp); 2455 } 2456 vput(vp); 2457 out: 2458 kauth_cred_free(cred); 2459 return (error); 2460 } 2461 2462 /* 2463 * Get file status; this version follows links. 2464 */ 2465 /* ARGSUSED */ 2466 int 2467 sys___stat30(struct lwp *l, void *v, register_t *retval) 2468 { 2469 struct sys___stat30_args /* { 2470 syscallarg(const char *) path; 2471 syscallarg(struct stat *) ub; 2472 } */ *uap = v; 2473 struct stat sb; 2474 int error; 2475 struct nameidata nd; 2476 2477 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2478 SCARG(uap, path), l); 2479 if ((error = namei(&nd)) != 0) 2480 return (error); 2481 error = vn_stat(nd.ni_vp, &sb, l); 2482 vput(nd.ni_vp); 2483 if (error) 2484 return (error); 2485 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2486 return (error); 2487 } 2488 2489 /* 2490 * Get file status; this version does not follow links. 2491 */ 2492 /* ARGSUSED */ 2493 int 2494 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2495 { 2496 struct sys___lstat30_args /* { 2497 syscallarg(const char *) path; 2498 syscallarg(struct stat *) ub; 2499 } */ *uap = v; 2500 struct stat sb; 2501 int error; 2502 struct nameidata nd; 2503 2504 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2505 SCARG(uap, path), l); 2506 if ((error = namei(&nd)) != 0) 2507 return (error); 2508 error = vn_stat(nd.ni_vp, &sb, l); 2509 vput(nd.ni_vp); 2510 if (error) 2511 return (error); 2512 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2513 return (error); 2514 } 2515 2516 /* 2517 * Get configurable pathname variables. 2518 */ 2519 /* ARGSUSED */ 2520 int 2521 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2522 { 2523 struct sys_pathconf_args /* { 2524 syscallarg(const char *) path; 2525 syscallarg(int) name; 2526 } */ *uap = v; 2527 int error; 2528 struct nameidata nd; 2529 2530 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2531 SCARG(uap, path), l); 2532 if ((error = namei(&nd)) != 0) 2533 return (error); 2534 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2535 vput(nd.ni_vp); 2536 return (error); 2537 } 2538 2539 /* 2540 * Return target name of a symbolic link. 2541 */ 2542 /* ARGSUSED */ 2543 int 2544 sys_readlink(struct lwp *l, void *v, register_t *retval) 2545 { 2546 struct sys_readlink_args /* { 2547 syscallarg(const char *) path; 2548 syscallarg(char *) buf; 2549 syscallarg(size_t) count; 2550 } */ *uap = v; 2551 struct vnode *vp; 2552 struct iovec aiov; 2553 struct uio auio; 2554 int error; 2555 struct nameidata nd; 2556 2557 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2558 SCARG(uap, path), l); 2559 if ((error = namei(&nd)) != 0) 2560 return (error); 2561 vp = nd.ni_vp; 2562 if (vp->v_type != VLNK) 2563 error = EINVAL; 2564 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2565 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2566 aiov.iov_base = SCARG(uap, buf); 2567 aiov.iov_len = SCARG(uap, count); 2568 auio.uio_iov = &aiov; 2569 auio.uio_iovcnt = 1; 2570 auio.uio_offset = 0; 2571 auio.uio_rw = UIO_READ; 2572 KASSERT(l == curlwp); 2573 auio.uio_vmspace = l->l_proc->p_vmspace; 2574 auio.uio_resid = SCARG(uap, count); 2575 error = VOP_READLINK(vp, &auio, l->l_cred); 2576 } 2577 vput(vp); 2578 *retval = SCARG(uap, count) - auio.uio_resid; 2579 return (error); 2580 } 2581 2582 /* 2583 * Change flags of a file given a path name. 2584 */ 2585 /* ARGSUSED */ 2586 int 2587 sys_chflags(struct lwp *l, void *v, register_t *retval) 2588 { 2589 struct sys_chflags_args /* { 2590 syscallarg(const char *) path; 2591 syscallarg(u_long) flags; 2592 } */ *uap = v; 2593 struct vnode *vp; 2594 int error; 2595 struct nameidata nd; 2596 2597 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2598 if ((error = namei(&nd)) != 0) 2599 return (error); 2600 vp = nd.ni_vp; 2601 error = change_flags(vp, SCARG(uap, flags), l); 2602 vput(vp); 2603 return (error); 2604 } 2605 2606 /* 2607 * Change flags of a file given a file descriptor. 2608 */ 2609 /* ARGSUSED */ 2610 int 2611 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2612 { 2613 struct sys_fchflags_args /* { 2614 syscallarg(int) fd; 2615 syscallarg(u_long) flags; 2616 } */ *uap = v; 2617 struct proc *p = l->l_proc; 2618 struct vnode *vp; 2619 struct file *fp; 2620 int error; 2621 2622 /* getvnode() will use the descriptor for us */ 2623 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2624 return (error); 2625 vp = (struct vnode *)fp->f_data; 2626 error = change_flags(vp, SCARG(uap, flags), l); 2627 VOP_UNLOCK(vp, 0); 2628 FILE_UNUSE(fp, l); 2629 return (error); 2630 } 2631 2632 /* 2633 * Change flags of a file given a path name; this version does 2634 * not follow links. 2635 */ 2636 int 2637 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2638 { 2639 struct sys_lchflags_args /* { 2640 syscallarg(const char *) path; 2641 syscallarg(u_long) flags; 2642 } */ *uap = v; 2643 struct vnode *vp; 2644 int error; 2645 struct nameidata nd; 2646 2647 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2648 if ((error = namei(&nd)) != 0) 2649 return (error); 2650 vp = nd.ni_vp; 2651 error = change_flags(vp, SCARG(uap, flags), l); 2652 vput(vp); 2653 return (error); 2654 } 2655 2656 /* 2657 * Common routine to change flags of a file. 2658 */ 2659 int 2660 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2661 { 2662 struct mount *mp; 2663 struct vattr vattr; 2664 int error; 2665 2666 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2667 return (error); 2668 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2669 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2670 /* 2671 * Non-superusers cannot change the flags on devices, even if they 2672 * own them. 2673 */ 2674 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2675 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2676 goto out; 2677 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2678 error = EINVAL; 2679 goto out; 2680 } 2681 } 2682 VATTR_NULL(&vattr); 2683 vattr.va_flags = flags; 2684 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2685 out: 2686 vn_finished_write(mp, 0); 2687 return (error); 2688 } 2689 2690 /* 2691 * Change mode of a file given path name; this version follows links. 2692 */ 2693 /* ARGSUSED */ 2694 int 2695 sys_chmod(struct lwp *l, void *v, register_t *retval) 2696 { 2697 struct sys_chmod_args /* { 2698 syscallarg(const char *) path; 2699 syscallarg(int) mode; 2700 } */ *uap = v; 2701 int error; 2702 struct nameidata nd; 2703 2704 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2705 if ((error = namei(&nd)) != 0) 2706 return (error); 2707 2708 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2709 2710 vrele(nd.ni_vp); 2711 return (error); 2712 } 2713 2714 /* 2715 * Change mode of a file given a file descriptor. 2716 */ 2717 /* ARGSUSED */ 2718 int 2719 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2720 { 2721 struct sys_fchmod_args /* { 2722 syscallarg(int) fd; 2723 syscallarg(int) mode; 2724 } */ *uap = v; 2725 struct proc *p = l->l_proc; 2726 struct file *fp; 2727 int error; 2728 2729 /* getvnode() will use the descriptor for us */ 2730 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2731 return (error); 2732 2733 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2734 FILE_UNUSE(fp, l); 2735 return (error); 2736 } 2737 2738 /* 2739 * Change mode of a file given path name; this version does not follow links. 2740 */ 2741 /* ARGSUSED */ 2742 int 2743 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2744 { 2745 struct sys_lchmod_args /* { 2746 syscallarg(const char *) path; 2747 syscallarg(int) mode; 2748 } */ *uap = v; 2749 int error; 2750 struct nameidata nd; 2751 2752 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2753 if ((error = namei(&nd)) != 0) 2754 return (error); 2755 2756 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2757 2758 vrele(nd.ni_vp); 2759 return (error); 2760 } 2761 2762 /* 2763 * Common routine to set mode given a vnode. 2764 */ 2765 static int 2766 change_mode(struct vnode *vp, int mode, struct lwp *l) 2767 { 2768 struct mount *mp; 2769 struct vattr vattr; 2770 int error; 2771 2772 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2773 return (error); 2774 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2775 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2776 VATTR_NULL(&vattr); 2777 vattr.va_mode = mode & ALLPERMS; 2778 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2779 VOP_UNLOCK(vp, 0); 2780 vn_finished_write(mp, 0); 2781 return (error); 2782 } 2783 2784 /* 2785 * Set ownership given a path name; this version follows links. 2786 */ 2787 /* ARGSUSED */ 2788 int 2789 sys_chown(struct lwp *l, void *v, register_t *retval) 2790 { 2791 struct sys_chown_args /* { 2792 syscallarg(const char *) path; 2793 syscallarg(uid_t) uid; 2794 syscallarg(gid_t) gid; 2795 } */ *uap = v; 2796 int error; 2797 struct nameidata nd; 2798 2799 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2800 if ((error = namei(&nd)) != 0) 2801 return (error); 2802 2803 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2804 2805 vrele(nd.ni_vp); 2806 return (error); 2807 } 2808 2809 /* 2810 * Set ownership given a path name; this version follows links. 2811 * Provides POSIX semantics. 2812 */ 2813 /* ARGSUSED */ 2814 int 2815 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2816 { 2817 struct sys_chown_args /* { 2818 syscallarg(const char *) path; 2819 syscallarg(uid_t) uid; 2820 syscallarg(gid_t) gid; 2821 } */ *uap = v; 2822 int error; 2823 struct nameidata nd; 2824 2825 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2826 if ((error = namei(&nd)) != 0) 2827 return (error); 2828 2829 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2830 2831 vrele(nd.ni_vp); 2832 return (error); 2833 } 2834 2835 /* 2836 * Set ownership given a file descriptor. 2837 */ 2838 /* ARGSUSED */ 2839 int 2840 sys_fchown(struct lwp *l, void *v, register_t *retval) 2841 { 2842 struct sys_fchown_args /* { 2843 syscallarg(int) fd; 2844 syscallarg(uid_t) uid; 2845 syscallarg(gid_t) gid; 2846 } */ *uap = v; 2847 struct proc *p = l->l_proc; 2848 int error; 2849 struct file *fp; 2850 2851 /* getvnode() will use the descriptor for us */ 2852 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2853 return (error); 2854 2855 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2856 SCARG(uap, gid), l, 0); 2857 FILE_UNUSE(fp, l); 2858 return (error); 2859 } 2860 2861 /* 2862 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2863 */ 2864 /* ARGSUSED */ 2865 int 2866 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2867 { 2868 struct sys_fchown_args /* { 2869 syscallarg(int) fd; 2870 syscallarg(uid_t) uid; 2871 syscallarg(gid_t) gid; 2872 } */ *uap = v; 2873 struct proc *p = l->l_proc; 2874 int error; 2875 struct file *fp; 2876 2877 /* getvnode() will use the descriptor for us */ 2878 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2879 return (error); 2880 2881 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2882 SCARG(uap, gid), l, 1); 2883 FILE_UNUSE(fp, l); 2884 return (error); 2885 } 2886 2887 /* 2888 * Set ownership given a path name; this version does not follow links. 2889 */ 2890 /* ARGSUSED */ 2891 int 2892 sys_lchown(struct lwp *l, void *v, register_t *retval) 2893 { 2894 struct sys_lchown_args /* { 2895 syscallarg(const char *) path; 2896 syscallarg(uid_t) uid; 2897 syscallarg(gid_t) gid; 2898 } */ *uap = v; 2899 int error; 2900 struct nameidata nd; 2901 2902 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2903 if ((error = namei(&nd)) != 0) 2904 return (error); 2905 2906 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2907 2908 vrele(nd.ni_vp); 2909 return (error); 2910 } 2911 2912 /* 2913 * Set ownership given a path name; this version does not follow links. 2914 * Provides POSIX/XPG semantics. 2915 */ 2916 /* ARGSUSED */ 2917 int 2918 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2919 { 2920 struct sys_lchown_args /* { 2921 syscallarg(const char *) path; 2922 syscallarg(uid_t) uid; 2923 syscallarg(gid_t) gid; 2924 } */ *uap = v; 2925 int error; 2926 struct nameidata nd; 2927 2928 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2929 if ((error = namei(&nd)) != 0) 2930 return (error); 2931 2932 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2933 2934 vrele(nd.ni_vp); 2935 return (error); 2936 } 2937 2938 /* 2939 * Common routine to set ownership given a vnode. 2940 */ 2941 static int 2942 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2943 int posix_semantics) 2944 { 2945 struct mount *mp; 2946 struct vattr vattr; 2947 mode_t newmode; 2948 int error; 2949 2950 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2951 return (error); 2952 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2953 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2954 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2955 goto out; 2956 2957 #define CHANGED(x) ((int)(x) != -1) 2958 newmode = vattr.va_mode; 2959 if (posix_semantics) { 2960 /* 2961 * POSIX/XPG semantics: if the caller is not the super-user, 2962 * clear set-user-id and set-group-id bits. Both POSIX and 2963 * the XPG consider the behaviour for calls by the super-user 2964 * implementation-defined; we leave the set-user-id and set- 2965 * group-id settings intact in that case. 2966 */ 2967 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2968 NULL) != 0) 2969 newmode &= ~(S_ISUID | S_ISGID); 2970 } else { 2971 /* 2972 * NetBSD semantics: when changing owner and/or group, 2973 * clear the respective bit(s). 2974 */ 2975 if (CHANGED(uid)) 2976 newmode &= ~S_ISUID; 2977 if (CHANGED(gid)) 2978 newmode &= ~S_ISGID; 2979 } 2980 /* Update va_mode iff altered. */ 2981 if (vattr.va_mode == newmode) 2982 newmode = VNOVAL; 2983 2984 VATTR_NULL(&vattr); 2985 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2986 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2987 vattr.va_mode = newmode; 2988 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2989 #undef CHANGED 2990 2991 out: 2992 VOP_UNLOCK(vp, 0); 2993 vn_finished_write(mp, 0); 2994 return (error); 2995 } 2996 2997 /* 2998 * Set the access and modification times given a path name; this 2999 * version follows links. 3000 */ 3001 /* ARGSUSED */ 3002 int 3003 sys_utimes(struct lwp *l, void *v, register_t *retval) 3004 { 3005 struct sys_utimes_args /* { 3006 syscallarg(const char *) path; 3007 syscallarg(const struct timeval *) tptr; 3008 } */ *uap = v; 3009 int error; 3010 struct nameidata nd; 3011 3012 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3013 if ((error = namei(&nd)) != 0) 3014 return (error); 3015 3016 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3017 3018 vrele(nd.ni_vp); 3019 return (error); 3020 } 3021 3022 /* 3023 * Set the access and modification times given a file descriptor. 3024 */ 3025 /* ARGSUSED */ 3026 int 3027 sys_futimes(struct lwp *l, void *v, register_t *retval) 3028 { 3029 struct sys_futimes_args /* { 3030 syscallarg(int) fd; 3031 syscallarg(const struct timeval *) tptr; 3032 } */ *uap = v; 3033 struct proc *p = l->l_proc; 3034 int error; 3035 struct file *fp; 3036 3037 /* getvnode() will use the descriptor for us */ 3038 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3039 return (error); 3040 3041 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3042 FILE_UNUSE(fp, l); 3043 return (error); 3044 } 3045 3046 /* 3047 * Set the access and modification times given a path name; this 3048 * version does not follow links. 3049 */ 3050 /* ARGSUSED */ 3051 int 3052 sys_lutimes(struct lwp *l, void *v, register_t *retval) 3053 { 3054 struct sys_lutimes_args /* { 3055 syscallarg(const char *) path; 3056 syscallarg(const struct timeval *) tptr; 3057 } */ *uap = v; 3058 int error; 3059 struct nameidata nd; 3060 3061 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3062 if ((error = namei(&nd)) != 0) 3063 return (error); 3064 3065 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3066 3067 vrele(nd.ni_vp); 3068 return (error); 3069 } 3070 3071 /* 3072 * Common routine to set access and modification times given a vnode. 3073 */ 3074 static int 3075 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3076 { 3077 struct mount *mp; 3078 struct vattr vattr; 3079 int error; 3080 3081 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3082 return (error); 3083 VATTR_NULL(&vattr); 3084 if (tptr == NULL) { 3085 nanotime(&vattr.va_atime); 3086 vattr.va_mtime = vattr.va_atime; 3087 vattr.va_vaflags |= VA_UTIMES_NULL; 3088 } else { 3089 struct timeval tv[2]; 3090 3091 error = copyin(tptr, tv, sizeof(tv)); 3092 if (error) 3093 goto out; 3094 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3095 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3096 } 3097 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3098 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3099 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3100 VOP_UNLOCK(vp, 0); 3101 out: 3102 vn_finished_write(mp, 0); 3103 return (error); 3104 } 3105 3106 /* 3107 * Truncate a file given its path name. 3108 */ 3109 /* ARGSUSED */ 3110 int 3111 sys_truncate(struct lwp *l, void *v, register_t *retval) 3112 { 3113 struct sys_truncate_args /* { 3114 syscallarg(const char *) path; 3115 syscallarg(int) pad; 3116 syscallarg(off_t) length; 3117 } */ *uap = v; 3118 struct vnode *vp; 3119 struct mount *mp; 3120 struct vattr vattr; 3121 int error; 3122 struct nameidata nd; 3123 3124 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3125 if ((error = namei(&nd)) != 0) 3126 return (error); 3127 vp = nd.ni_vp; 3128 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3129 vrele(vp); 3130 return (error); 3131 } 3132 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3133 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3134 if (vp->v_type == VDIR) 3135 error = EISDIR; 3136 else if ((error = vn_writechk(vp)) == 0 && 3137 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3138 VATTR_NULL(&vattr); 3139 vattr.va_size = SCARG(uap, length); 3140 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3141 } 3142 vput(vp); 3143 vn_finished_write(mp, 0); 3144 return (error); 3145 } 3146 3147 /* 3148 * Truncate a file given a file descriptor. 3149 */ 3150 /* ARGSUSED */ 3151 int 3152 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3153 { 3154 struct sys_ftruncate_args /* { 3155 syscallarg(int) fd; 3156 syscallarg(int) pad; 3157 syscallarg(off_t) length; 3158 } */ *uap = v; 3159 struct proc *p = l->l_proc; 3160 struct mount *mp; 3161 struct vattr vattr; 3162 struct vnode *vp; 3163 struct file *fp; 3164 int error; 3165 3166 /* getvnode() will use the descriptor for us */ 3167 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3168 return (error); 3169 if ((fp->f_flag & FWRITE) == 0) { 3170 error = EINVAL; 3171 goto out; 3172 } 3173 vp = (struct vnode *)fp->f_data; 3174 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3175 FILE_UNUSE(fp, l); 3176 return (error); 3177 } 3178 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3179 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3180 if (vp->v_type == VDIR) 3181 error = EISDIR; 3182 else if ((error = vn_writechk(vp)) == 0) { 3183 VATTR_NULL(&vattr); 3184 vattr.va_size = SCARG(uap, length); 3185 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3186 } 3187 VOP_UNLOCK(vp, 0); 3188 vn_finished_write(mp, 0); 3189 out: 3190 FILE_UNUSE(fp, l); 3191 return (error); 3192 } 3193 3194 /* 3195 * Sync an open file. 3196 */ 3197 /* ARGSUSED */ 3198 int 3199 sys_fsync(struct lwp *l, void *v, register_t *retval) 3200 { 3201 struct sys_fsync_args /* { 3202 syscallarg(int) fd; 3203 } */ *uap = v; 3204 struct proc *p = l->l_proc; 3205 struct vnode *vp; 3206 struct mount *mp; 3207 struct file *fp; 3208 int error; 3209 3210 /* getvnode() will use the descriptor for us */ 3211 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3212 return (error); 3213 vp = (struct vnode *)fp->f_data; 3214 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3215 FILE_UNUSE(fp, l); 3216 return (error); 3217 } 3218 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3219 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3220 if (error == 0 && bioops.io_fsync != NULL && 3221 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3222 (*bioops.io_fsync)(vp, 0); 3223 VOP_UNLOCK(vp, 0); 3224 vn_finished_write(mp, 0); 3225 FILE_UNUSE(fp, l); 3226 return (error); 3227 } 3228 3229 /* 3230 * Sync a range of file data. API modeled after that found in AIX. 3231 * 3232 * FDATASYNC indicates that we need only save enough metadata to be able 3233 * to re-read the written data. Note we duplicate AIX's requirement that 3234 * the file be open for writing. 3235 */ 3236 /* ARGSUSED */ 3237 int 3238 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3239 { 3240 struct sys_fsync_range_args /* { 3241 syscallarg(int) fd; 3242 syscallarg(int) flags; 3243 syscallarg(off_t) start; 3244 syscallarg(off_t) length; 3245 } */ *uap = v; 3246 struct proc *p = l->l_proc; 3247 struct vnode *vp; 3248 struct file *fp; 3249 int flags, nflags; 3250 off_t s, e, len; 3251 int error; 3252 3253 /* getvnode() will use the descriptor for us */ 3254 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3255 return (error); 3256 3257 if ((fp->f_flag & FWRITE) == 0) { 3258 error = EBADF; 3259 goto out; 3260 } 3261 3262 flags = SCARG(uap, flags); 3263 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3264 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3265 error = EINVAL; 3266 goto out; 3267 } 3268 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3269 if (flags & FDATASYNC) 3270 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3271 else 3272 nflags = FSYNC_WAIT; 3273 if (flags & FDISKSYNC) 3274 nflags |= FSYNC_CACHE; 3275 3276 len = SCARG(uap, length); 3277 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3278 if (len) { 3279 s = SCARG(uap, start); 3280 e = s + len; 3281 if (e < s) { 3282 error = EINVAL; 3283 goto out; 3284 } 3285 } else { 3286 e = 0; 3287 s = 0; 3288 } 3289 3290 vp = (struct vnode *)fp->f_data; 3291 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3292 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3293 3294 if (error == 0 && bioops.io_fsync != NULL && 3295 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3296 (*bioops.io_fsync)(vp, nflags); 3297 3298 VOP_UNLOCK(vp, 0); 3299 out: 3300 FILE_UNUSE(fp, l); 3301 return (error); 3302 } 3303 3304 /* 3305 * Sync the data of an open file. 3306 */ 3307 /* ARGSUSED */ 3308 int 3309 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3310 { 3311 struct sys_fdatasync_args /* { 3312 syscallarg(int) fd; 3313 } */ *uap = v; 3314 struct proc *p = l->l_proc; 3315 struct vnode *vp; 3316 struct file *fp; 3317 int error; 3318 3319 /* getvnode() will use the descriptor for us */ 3320 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3321 return (error); 3322 if ((fp->f_flag & FWRITE) == 0) { 3323 FILE_UNUSE(fp, l); 3324 return (EBADF); 3325 } 3326 vp = (struct vnode *)fp->f_data; 3327 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3328 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3329 VOP_UNLOCK(vp, 0); 3330 FILE_UNUSE(fp, l); 3331 return (error); 3332 } 3333 3334 /* 3335 * Rename files, (standard) BSD semantics frontend. 3336 */ 3337 /* ARGSUSED */ 3338 int 3339 sys_rename(struct lwp *l, void *v, register_t *retval) 3340 { 3341 struct sys_rename_args /* { 3342 syscallarg(const char *) from; 3343 syscallarg(const char *) to; 3344 } */ *uap = v; 3345 3346 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3347 } 3348 3349 /* 3350 * Rename files, POSIX semantics frontend. 3351 */ 3352 /* ARGSUSED */ 3353 int 3354 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3355 { 3356 struct sys___posix_rename_args /* { 3357 syscallarg(const char *) from; 3358 syscallarg(const char *) to; 3359 } */ *uap = v; 3360 3361 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3362 } 3363 3364 /* 3365 * Rename files. Source and destination must either both be directories, 3366 * or both not be directories. If target is a directory, it must be empty. 3367 * If `from' and `to' refer to the same object, the value of the `retain' 3368 * argument is used to determine whether `from' will be 3369 * 3370 * (retain == 0) deleted unless `from' and `to' refer to the same 3371 * object in the file system's name space (BSD). 3372 * (retain == 1) always retained (POSIX). 3373 */ 3374 static int 3375 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3376 { 3377 struct mount *mp = NULL; 3378 struct vnode *tvp, *fvp, *tdvp; 3379 struct nameidata fromnd, tond; 3380 struct proc *p; 3381 int error; 3382 3383 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3384 from, l); 3385 if ((error = namei(&fromnd)) != 0) 3386 return (error); 3387 if (fromnd.ni_dvp != fromnd.ni_vp) 3388 VOP_UNLOCK(fromnd.ni_dvp, 0); 3389 fvp = fromnd.ni_vp; 3390 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3391 if (error != 0) { 3392 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3393 vrele(fromnd.ni_dvp); 3394 vrele(fvp); 3395 if (fromnd.ni_startdir) 3396 vrele(fromnd.ni_startdir); 3397 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3398 return (error); 3399 } 3400 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3401 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3402 if ((error = namei(&tond)) != 0) { 3403 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3404 vrele(fromnd.ni_dvp); 3405 vrele(fvp); 3406 goto out1; 3407 } 3408 tdvp = tond.ni_dvp; 3409 tvp = tond.ni_vp; 3410 3411 if (tvp != NULL) { 3412 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3413 error = ENOTDIR; 3414 goto out; 3415 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3416 error = EISDIR; 3417 goto out; 3418 } 3419 } 3420 3421 if (fvp == tdvp) 3422 error = EINVAL; 3423 3424 /* 3425 * Source and destination refer to the same object. 3426 */ 3427 if (fvp == tvp) { 3428 if (retain) 3429 error = -1; 3430 else if (fromnd.ni_dvp == tdvp && 3431 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3432 !memcmp(fromnd.ni_cnd.cn_nameptr, 3433 tond.ni_cnd.cn_nameptr, 3434 fromnd.ni_cnd.cn_namelen)) 3435 error = -1; 3436 } 3437 3438 #if NVERIEXEC > 0 3439 if (!error) { 3440 pathname_t frompath = NULL, topath = NULL; 3441 3442 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3443 &frompath); 3444 if (!error) 3445 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3446 &topath); 3447 if (!error) 3448 error = veriexec_renamechk(fvp, pathname_path(frompath), 3449 tvp, pathname_path(topath), l); 3450 3451 pathname_put(frompath); 3452 pathname_put(topath); 3453 } 3454 #endif /* NVERIEXEC > 0 */ 3455 3456 out: 3457 p = l->l_proc; 3458 if (!error) { 3459 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3460 if (fromnd.ni_dvp != tdvp) 3461 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3462 if (tvp) { 3463 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3464 } 3465 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3466 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3467 } else { 3468 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3469 if (tdvp == tvp) 3470 vrele(tdvp); 3471 else 3472 vput(tdvp); 3473 if (tvp) 3474 vput(tvp); 3475 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3476 vrele(fromnd.ni_dvp); 3477 vrele(fvp); 3478 } 3479 vrele(tond.ni_startdir); 3480 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3481 out1: 3482 vn_finished_write(mp, 0); 3483 if (fromnd.ni_startdir) 3484 vrele(fromnd.ni_startdir); 3485 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3486 return (error == -1 ? 0 : error); 3487 } 3488 3489 /* 3490 * Make a directory file. 3491 */ 3492 /* ARGSUSED */ 3493 int 3494 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3495 { 3496 struct sys_mkdir_args /* { 3497 syscallarg(const char *) path; 3498 syscallarg(int) mode; 3499 } */ *uap = v; 3500 struct proc *p = l->l_proc; 3501 struct mount *mp; 3502 struct vnode *vp; 3503 struct vattr vattr; 3504 int error; 3505 struct nameidata nd; 3506 3507 restart: 3508 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3509 SCARG(uap, path), l); 3510 if ((error = namei(&nd)) != 0) 3511 return (error); 3512 vp = nd.ni_vp; 3513 if (vp != NULL) { 3514 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3515 if (nd.ni_dvp == vp) 3516 vrele(nd.ni_dvp); 3517 else 3518 vput(nd.ni_dvp); 3519 vrele(vp); 3520 return (EEXIST); 3521 } 3522 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3523 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3524 if (nd.ni_dvp == vp) 3525 vrele(nd.ni_dvp); 3526 else 3527 vput(nd.ni_dvp); 3528 if ((error = vn_start_write(NULL, &mp, 3529 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3530 return (error); 3531 goto restart; 3532 } 3533 VATTR_NULL(&vattr); 3534 vattr.va_type = VDIR; 3535 vattr.va_mode = 3536 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3537 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3538 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3539 if (!error) 3540 vput(nd.ni_vp); 3541 vn_finished_write(mp, 0); 3542 return (error); 3543 } 3544 3545 /* 3546 * Remove a directory file. 3547 */ 3548 /* ARGSUSED */ 3549 int 3550 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3551 { 3552 struct sys_rmdir_args /* { 3553 syscallarg(const char *) path; 3554 } */ *uap = v; 3555 struct mount *mp; 3556 struct vnode *vp; 3557 int error; 3558 struct nameidata nd; 3559 3560 restart: 3561 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3562 SCARG(uap, path), l); 3563 if ((error = namei(&nd)) != 0) 3564 return (error); 3565 vp = nd.ni_vp; 3566 if (vp->v_type != VDIR) { 3567 error = ENOTDIR; 3568 goto out; 3569 } 3570 /* 3571 * No rmdir "." please. 3572 */ 3573 if (nd.ni_dvp == vp) { 3574 error = EINVAL; 3575 goto out; 3576 } 3577 /* 3578 * The root of a mounted filesystem cannot be deleted. 3579 */ 3580 if (vp->v_flag & VROOT) { 3581 error = EBUSY; 3582 goto out; 3583 } 3584 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3585 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3586 if (nd.ni_dvp == vp) 3587 vrele(nd.ni_dvp); 3588 else 3589 vput(nd.ni_dvp); 3590 vput(vp); 3591 if ((error = vn_start_write(NULL, &mp, 3592 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3593 return (error); 3594 goto restart; 3595 } 3596 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3597 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3598 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3599 vn_finished_write(mp, 0); 3600 return (error); 3601 3602 out: 3603 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3604 if (nd.ni_dvp == vp) 3605 vrele(nd.ni_dvp); 3606 else 3607 vput(nd.ni_dvp); 3608 vput(vp); 3609 return (error); 3610 } 3611 3612 /* 3613 * Read a block of directory entries in a file system independent format. 3614 */ 3615 int 3616 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3617 { 3618 struct sys___getdents30_args /* { 3619 syscallarg(int) fd; 3620 syscallarg(char *) buf; 3621 syscallarg(size_t) count; 3622 } */ *uap = v; 3623 struct proc *p = l->l_proc; 3624 struct file *fp; 3625 int error, done; 3626 3627 /* getvnode() will use the descriptor for us */ 3628 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3629 return (error); 3630 if ((fp->f_flag & FREAD) == 0) { 3631 error = EBADF; 3632 goto out; 3633 } 3634 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3635 SCARG(uap, count), &done, l, 0, 0); 3636 #ifdef KTRACE 3637 if (!error && KTRPOINT(p, KTR_GENIO)) { 3638 struct iovec iov; 3639 iov.iov_base = SCARG(uap, buf); 3640 iov.iov_len = done; 3641 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3642 } 3643 #endif 3644 *retval = done; 3645 out: 3646 FILE_UNUSE(fp, l); 3647 return (error); 3648 } 3649 3650 /* 3651 * Set the mode mask for creation of filesystem nodes. 3652 */ 3653 int 3654 sys_umask(struct lwp *l, void *v, register_t *retval) 3655 { 3656 struct sys_umask_args /* { 3657 syscallarg(mode_t) newmask; 3658 } */ *uap = v; 3659 struct proc *p = l->l_proc; 3660 struct cwdinfo *cwdi; 3661 3662 cwdi = p->p_cwdi; 3663 *retval = cwdi->cwdi_cmask; 3664 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3665 return (0); 3666 } 3667 3668 /* 3669 * Void all references to file by ripping underlying filesystem 3670 * away from vnode. 3671 */ 3672 /* ARGSUSED */ 3673 int 3674 sys_revoke(struct lwp *l, void *v, register_t *retval) 3675 { 3676 struct sys_revoke_args /* { 3677 syscallarg(const char *) path; 3678 } */ *uap = v; 3679 struct mount *mp; 3680 struct vnode *vp; 3681 struct vattr vattr; 3682 int error; 3683 struct nameidata nd; 3684 3685 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3686 if ((error = namei(&nd)) != 0) 3687 return (error); 3688 vp = nd.ni_vp; 3689 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3690 goto out; 3691 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3692 (error = kauth_authorize_generic(l->l_cred, 3693 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3694 goto out; 3695 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3696 goto out; 3697 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3698 VOP_REVOKE(vp, REVOKEALL); 3699 vn_finished_write(mp, 0); 3700 out: 3701 vrele(vp); 3702 return (error); 3703 } 3704 3705 /* 3706 * Convert a user file descriptor to a kernel file entry. 3707 */ 3708 int 3709 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3710 { 3711 struct vnode *vp; 3712 struct file *fp; 3713 3714 if ((fp = fd_getfile(fdp, fd)) == NULL) 3715 return (EBADF); 3716 3717 FILE_USE(fp); 3718 3719 if (fp->f_type != DTYPE_VNODE) { 3720 FILE_UNUSE(fp, NULL); 3721 return (EINVAL); 3722 } 3723 3724 vp = (struct vnode *)fp->f_data; 3725 if (vp->v_type == VBAD) { 3726 FILE_UNUSE(fp, NULL); 3727 return (EBADF); 3728 } 3729 3730 *fpp = fp; 3731 return (0); 3732 } 3733
Indexes created Thu Sep 25 16:09:42 GMT 2025