vfs_syscalls.c revision 1.300
1 /* $NetBSD: vfs_syscalls.c,v 1.300 2007/02/09 21:55:32 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.300 2007/02/09 21:55:32 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/syscallargs.h> 65 #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; 1761 int whiteout = 0; 1762 struct nameidata nd; 1763 1764 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1765 0, NULL, NULL, NULL)) != 0) 1766 return (error); 1767 restart: 1768 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1769 if ((error = namei(&nd)) != 0) 1770 return (error); 1771 vp = nd.ni_vp; 1772 if (vp != NULL) 1773 error = EEXIST; 1774 else { 1775 VATTR_NULL(&vattr); 1776 vattr.va_mode = 1777 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1778 vattr.va_rdev = SCARG(uap, dev); 1779 whiteout = 0; 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 whiteout = 1; 1793 break; 1794 default: 1795 error = EINVAL; 1796 break; 1797 } 1798 } 1799 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1800 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1801 if (nd.ni_dvp == vp) 1802 vrele(nd.ni_dvp); 1803 else 1804 vput(nd.ni_dvp); 1805 if (vp) 1806 vrele(vp); 1807 if ((error = vn_start_write(NULL, &mp, 1808 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1809 return (error); 1810 goto restart; 1811 } 1812 if (!error) { 1813 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1814 if (whiteout) { 1815 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1816 if (error) 1817 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1818 vput(nd.ni_dvp); 1819 } else { 1820 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1821 &nd.ni_cnd, &vattr); 1822 if (error == 0) 1823 vput(nd.ni_vp); 1824 } 1825 } else { 1826 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1827 if (nd.ni_dvp == vp) 1828 vrele(nd.ni_dvp); 1829 else 1830 vput(nd.ni_dvp); 1831 if (vp) 1832 vrele(vp); 1833 } 1834 vn_finished_write(mp, 0); 1835 return (error); 1836 } 1837 1838 /* 1839 * Create a named pipe. 1840 */ 1841 /* ARGSUSED */ 1842 int 1843 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1844 { 1845 struct sys_mkfifo_args /* { 1846 syscallarg(const char *) path; 1847 syscallarg(int) mode; 1848 } */ *uap = v; 1849 struct proc *p = l->l_proc; 1850 struct mount *mp; 1851 struct vattr vattr; 1852 int error; 1853 struct nameidata nd; 1854 1855 restart: 1856 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1857 if ((error = namei(&nd)) != 0) 1858 return (error); 1859 if (nd.ni_vp != NULL) { 1860 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1861 if (nd.ni_dvp == nd.ni_vp) 1862 vrele(nd.ni_dvp); 1863 else 1864 vput(nd.ni_dvp); 1865 vrele(nd.ni_vp); 1866 return (EEXIST); 1867 } 1868 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1869 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1870 if (nd.ni_dvp == nd.ni_vp) 1871 vrele(nd.ni_dvp); 1872 else 1873 vput(nd.ni_dvp); 1874 if (nd.ni_vp) 1875 vrele(nd.ni_vp); 1876 if ((error = vn_start_write(NULL, &mp, 1877 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1878 return (error); 1879 goto restart; 1880 } 1881 VATTR_NULL(&vattr); 1882 vattr.va_type = VFIFO; 1883 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1884 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1885 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1886 if (error == 0) 1887 vput(nd.ni_vp); 1888 vn_finished_write(mp, 0); 1889 return (error); 1890 } 1891 1892 /* 1893 * Make a hard file link. 1894 */ 1895 /* ARGSUSED */ 1896 int 1897 sys_link(struct lwp *l, void *v, register_t *retval) 1898 { 1899 struct sys_link_args /* { 1900 syscallarg(const char *) path; 1901 syscallarg(const char *) link; 1902 } */ *uap = v; 1903 struct vnode *vp; 1904 struct mount *mp; 1905 struct nameidata nd; 1906 int error; 1907 1908 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1909 if ((error = namei(&nd)) != 0) 1910 return (error); 1911 vp = nd.ni_vp; 1912 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1913 vrele(vp); 1914 return (error); 1915 } 1916 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1917 if ((error = namei(&nd)) != 0) 1918 goto out; 1919 if (nd.ni_vp) { 1920 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1921 if (nd.ni_dvp == nd.ni_vp) 1922 vrele(nd.ni_dvp); 1923 else 1924 vput(nd.ni_dvp); 1925 vrele(nd.ni_vp); 1926 error = EEXIST; 1927 goto out; 1928 } 1929 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1930 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1931 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1932 out: 1933 vrele(vp); 1934 vn_finished_write(mp, 0); 1935 return (error); 1936 } 1937 1938 /* 1939 * Make a symbolic link. 1940 */ 1941 /* ARGSUSED */ 1942 int 1943 sys_symlink(struct lwp *l, void *v, register_t *retval) 1944 { 1945 struct sys_symlink_args /* { 1946 syscallarg(const char *) path; 1947 syscallarg(const char *) link; 1948 } */ *uap = v; 1949 struct proc *p = l->l_proc; 1950 struct mount *mp; 1951 struct vattr vattr; 1952 char *path; 1953 int error; 1954 struct nameidata nd; 1955 1956 path = PNBUF_GET(); 1957 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1958 if (error) 1959 goto out; 1960 restart: 1961 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1962 if ((error = namei(&nd)) != 0) 1963 goto out; 1964 if (nd.ni_vp) { 1965 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1966 if (nd.ni_dvp == nd.ni_vp) 1967 vrele(nd.ni_dvp); 1968 else 1969 vput(nd.ni_dvp); 1970 vrele(nd.ni_vp); 1971 error = EEXIST; 1972 goto out; 1973 } 1974 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1975 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1976 if (nd.ni_dvp == nd.ni_vp) 1977 vrele(nd.ni_dvp); 1978 else 1979 vput(nd.ni_dvp); 1980 if ((error = vn_start_write(NULL, &mp, 1981 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1982 return (error); 1983 goto restart; 1984 } 1985 VATTR_NULL(&vattr); 1986 vattr.va_type = VLNK; 1987 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1988 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1989 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1990 if (error == 0) 1991 vput(nd.ni_vp); 1992 vn_finished_write(mp, 0); 1993 out: 1994 PNBUF_PUT(path); 1995 return (error); 1996 } 1997 1998 /* 1999 * Delete a whiteout from the filesystem. 2000 */ 2001 /* ARGSUSED */ 2002 int 2003 sys_undelete(struct lwp *l, void *v, register_t *retval) 2004 { 2005 struct sys_undelete_args /* { 2006 syscallarg(const char *) path; 2007 } */ *uap = v; 2008 int error; 2009 struct mount *mp; 2010 struct nameidata nd; 2011 2012 restart: 2013 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2014 SCARG(uap, path), l); 2015 error = namei(&nd); 2016 if (error) 2017 return (error); 2018 2019 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2020 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2021 if (nd.ni_dvp == nd.ni_vp) 2022 vrele(nd.ni_dvp); 2023 else 2024 vput(nd.ni_dvp); 2025 if (nd.ni_vp) 2026 vrele(nd.ni_vp); 2027 return (EEXIST); 2028 } 2029 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2030 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2031 if (nd.ni_dvp == nd.ni_vp) 2032 vrele(nd.ni_dvp); 2033 else 2034 vput(nd.ni_dvp); 2035 if ((error = vn_start_write(NULL, &mp, 2036 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2037 return (error); 2038 goto restart; 2039 } 2040 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2041 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2042 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2043 vput(nd.ni_dvp); 2044 vn_finished_write(mp, 0); 2045 return (error); 2046 } 2047 2048 /* 2049 * Delete a name from the filesystem. 2050 */ 2051 /* ARGSUSED */ 2052 int 2053 sys_unlink(struct lwp *l, void *v, register_t *retval) 2054 { 2055 struct sys_unlink_args /* { 2056 syscallarg(const char *) path; 2057 } */ *uap = v; 2058 struct mount *mp; 2059 struct vnode *vp; 2060 int error; 2061 struct nameidata nd; 2062 #if NVERIEXEC > 0 2063 pathname_t pathbuf = NULL; 2064 #endif /* NVERIEXEC > 0 */ 2065 2066 restart: 2067 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2068 SCARG(uap, path), l); 2069 if ((error = namei(&nd)) != 0) 2070 return (error); 2071 vp = nd.ni_vp; 2072 2073 /* 2074 * The root of a mounted filesystem cannot be deleted. 2075 */ 2076 if (vp->v_flag & VROOT) { 2077 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2078 if (nd.ni_dvp == vp) 2079 vrele(nd.ni_dvp); 2080 else 2081 vput(nd.ni_dvp); 2082 vput(vp); 2083 error = EBUSY; 2084 goto out; 2085 } 2086 2087 #if NVERIEXEC > 0 2088 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2089 2090 /* Handle remove requests for veriexec entries. */ 2091 if (!error) { 2092 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2093 pathname_put(pathbuf); 2094 } 2095 2096 if (error) { 2097 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2098 if (nd.ni_dvp == vp) 2099 vrele(nd.ni_dvp); 2100 else 2101 vput(nd.ni_dvp); 2102 vput(vp); 2103 goto out; 2104 } 2105 #endif /* NVERIEXEC > 0 */ 2106 2107 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2108 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2109 if (nd.ni_dvp == vp) 2110 vrele(nd.ni_dvp); 2111 else 2112 vput(nd.ni_dvp); 2113 vput(vp); 2114 if ((error = vn_start_write(NULL, &mp, 2115 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2116 return (error); 2117 goto restart; 2118 } 2119 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2120 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2121 #ifdef FILEASSOC 2122 (void)fileassoc_file_delete(vp); 2123 #endif /* FILEASSOC */ 2124 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2125 vn_finished_write(mp, 0); 2126 out: 2127 return (error); 2128 } 2129 2130 /* 2131 * Reposition read/write file offset. 2132 */ 2133 int 2134 sys_lseek(struct lwp *l, void *v, register_t *retval) 2135 { 2136 struct sys_lseek_args /* { 2137 syscallarg(int) fd; 2138 syscallarg(int) pad; 2139 syscallarg(off_t) offset; 2140 syscallarg(int) whence; 2141 } */ *uap = v; 2142 struct proc *p = l->l_proc; 2143 kauth_cred_t cred = l->l_cred; 2144 struct filedesc *fdp = p->p_fd; 2145 struct file *fp; 2146 struct vnode *vp; 2147 struct vattr vattr; 2148 off_t newoff; 2149 int error; 2150 2151 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2152 return (EBADF); 2153 2154 FILE_USE(fp); 2155 2156 vp = (struct vnode *)fp->f_data; 2157 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2158 error = ESPIPE; 2159 goto out; 2160 } 2161 2162 switch (SCARG(uap, whence)) { 2163 case SEEK_CUR: 2164 newoff = fp->f_offset + SCARG(uap, offset); 2165 break; 2166 case SEEK_END: 2167 error = VOP_GETATTR(vp, &vattr, cred, l); 2168 if (error) 2169 goto out; 2170 newoff = SCARG(uap, offset) + vattr.va_size; 2171 break; 2172 case SEEK_SET: 2173 newoff = SCARG(uap, offset); 2174 break; 2175 default: 2176 error = EINVAL; 2177 goto out; 2178 } 2179 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2180 goto out; 2181 2182 *(off_t *)retval = fp->f_offset = newoff; 2183 out: 2184 FILE_UNUSE(fp, l); 2185 return (error); 2186 } 2187 2188 /* 2189 * Positional read system call. 2190 */ 2191 int 2192 sys_pread(struct lwp *l, void *v, register_t *retval) 2193 { 2194 struct sys_pread_args /* { 2195 syscallarg(int) fd; 2196 syscallarg(void *) buf; 2197 syscallarg(size_t) nbyte; 2198 syscallarg(off_t) offset; 2199 } */ *uap = v; 2200 struct proc *p = l->l_proc; 2201 struct filedesc *fdp = p->p_fd; 2202 struct file *fp; 2203 struct vnode *vp; 2204 off_t offset; 2205 int error, fd = SCARG(uap, fd); 2206 2207 if ((fp = fd_getfile(fdp, fd)) == NULL) 2208 return (EBADF); 2209 2210 if ((fp->f_flag & FREAD) == 0) { 2211 simple_unlock(&fp->f_slock); 2212 return (EBADF); 2213 } 2214 2215 FILE_USE(fp); 2216 2217 vp = (struct vnode *)fp->f_data; 2218 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2219 error = ESPIPE; 2220 goto out; 2221 } 2222 2223 offset = SCARG(uap, offset); 2224 2225 /* 2226 * XXX This works because no file systems actually 2227 * XXX take any action on the seek operation. 2228 */ 2229 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2230 goto out; 2231 2232 /* dofileread() will unuse the descriptor for us */ 2233 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2234 &offset, 0, retval)); 2235 2236 out: 2237 FILE_UNUSE(fp, l); 2238 return (error); 2239 } 2240 2241 /* 2242 * Positional scatter read system call. 2243 */ 2244 int 2245 sys_preadv(struct lwp *l, void *v, register_t *retval) 2246 { 2247 struct sys_preadv_args /* { 2248 syscallarg(int) fd; 2249 syscallarg(const struct iovec *) iovp; 2250 syscallarg(int) iovcnt; 2251 syscallarg(off_t) offset; 2252 } */ *uap = v; 2253 struct proc *p = l->l_proc; 2254 struct filedesc *fdp = p->p_fd; 2255 struct file *fp; 2256 struct vnode *vp; 2257 off_t offset; 2258 int error, fd = SCARG(uap, fd); 2259 2260 if ((fp = fd_getfile(fdp, fd)) == NULL) 2261 return (EBADF); 2262 2263 if ((fp->f_flag & FREAD) == 0) { 2264 simple_unlock(&fp->f_slock); 2265 return (EBADF); 2266 } 2267 2268 FILE_USE(fp); 2269 2270 vp = (struct vnode *)fp->f_data; 2271 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2272 error = ESPIPE; 2273 goto out; 2274 } 2275 2276 offset = SCARG(uap, offset); 2277 2278 /* 2279 * XXX This works because no file systems actually 2280 * XXX take any action on the seek operation. 2281 */ 2282 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2283 goto out; 2284 2285 /* dofilereadv() will unuse the descriptor for us */ 2286 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2287 &offset, 0, retval)); 2288 2289 out: 2290 FILE_UNUSE(fp, l); 2291 return (error); 2292 } 2293 2294 /* 2295 * Positional write system call. 2296 */ 2297 int 2298 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2299 { 2300 struct sys_pwrite_args /* { 2301 syscallarg(int) fd; 2302 syscallarg(const void *) buf; 2303 syscallarg(size_t) nbyte; 2304 syscallarg(off_t) offset; 2305 } */ *uap = v; 2306 struct proc *p = l->l_proc; 2307 struct filedesc *fdp = p->p_fd; 2308 struct file *fp; 2309 struct vnode *vp; 2310 off_t offset; 2311 int error, fd = SCARG(uap, fd); 2312 2313 if ((fp = fd_getfile(fdp, fd)) == NULL) 2314 return (EBADF); 2315 2316 if ((fp->f_flag & FWRITE) == 0) { 2317 simple_unlock(&fp->f_slock); 2318 return (EBADF); 2319 } 2320 2321 FILE_USE(fp); 2322 2323 vp = (struct vnode *)fp->f_data; 2324 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2325 error = ESPIPE; 2326 goto out; 2327 } 2328 2329 offset = SCARG(uap, offset); 2330 2331 /* 2332 * XXX This works because no file systems actually 2333 * XXX take any action on the seek operation. 2334 */ 2335 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2336 goto out; 2337 2338 /* dofilewrite() will unuse the descriptor for us */ 2339 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2340 &offset, 0, retval)); 2341 2342 out: 2343 FILE_UNUSE(fp, l); 2344 return (error); 2345 } 2346 2347 /* 2348 * Positional gather write system call. 2349 */ 2350 int 2351 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2352 { 2353 struct sys_pwritev_args /* { 2354 syscallarg(int) fd; 2355 syscallarg(const struct iovec *) iovp; 2356 syscallarg(int) iovcnt; 2357 syscallarg(off_t) offset; 2358 } */ *uap = v; 2359 struct proc *p = l->l_proc; 2360 struct filedesc *fdp = p->p_fd; 2361 struct file *fp; 2362 struct vnode *vp; 2363 off_t offset; 2364 int error, fd = SCARG(uap, fd); 2365 2366 if ((fp = fd_getfile(fdp, fd)) == NULL) 2367 return (EBADF); 2368 2369 if ((fp->f_flag & FWRITE) == 0) { 2370 simple_unlock(&fp->f_slock); 2371 return (EBADF); 2372 } 2373 2374 FILE_USE(fp); 2375 2376 vp = (struct vnode *)fp->f_data; 2377 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2378 error = ESPIPE; 2379 goto out; 2380 } 2381 2382 offset = SCARG(uap, offset); 2383 2384 /* 2385 * XXX This works because no file systems actually 2386 * XXX take any action on the seek operation. 2387 */ 2388 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2389 goto out; 2390 2391 /* dofilewritev() will unuse the descriptor for us */ 2392 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2393 &offset, 0, retval)); 2394 2395 out: 2396 FILE_UNUSE(fp, l); 2397 return (error); 2398 } 2399 2400 /* 2401 * Check access permissions. 2402 */ 2403 int 2404 sys_access(struct lwp *l, void *v, register_t *retval) 2405 { 2406 struct sys_access_args /* { 2407 syscallarg(const char *) path; 2408 syscallarg(int) flags; 2409 } */ *uap = v; 2410 kauth_cred_t cred; 2411 struct vnode *vp; 2412 int error, flags; 2413 struct nameidata nd; 2414 2415 cred = kauth_cred_dup(l->l_cred); 2416 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2417 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2418 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2419 SCARG(uap, path), l); 2420 /* Override default credentials */ 2421 nd.ni_cnd.cn_cred = cred; 2422 if ((error = namei(&nd)) != 0) 2423 goto out; 2424 vp = nd.ni_vp; 2425 2426 /* Flags == 0 means only check for existence. */ 2427 if (SCARG(uap, flags)) { 2428 flags = 0; 2429 if (SCARG(uap, flags) & R_OK) 2430 flags |= VREAD; 2431 if (SCARG(uap, flags) & W_OK) 2432 flags |= VWRITE; 2433 if (SCARG(uap, flags) & X_OK) 2434 flags |= VEXEC; 2435 2436 error = VOP_ACCESS(vp, flags, cred, l); 2437 if (!error && (flags & VWRITE)) 2438 error = vn_writechk(vp); 2439 } 2440 vput(vp); 2441 out: 2442 kauth_cred_free(cred); 2443 return (error); 2444 } 2445 2446 /* 2447 * Get file status; this version follows links. 2448 */ 2449 /* ARGSUSED */ 2450 int 2451 sys___stat30(struct lwp *l, void *v, register_t *retval) 2452 { 2453 struct sys___stat30_args /* { 2454 syscallarg(const char *) path; 2455 syscallarg(struct stat *) ub; 2456 } */ *uap = v; 2457 struct stat sb; 2458 int error; 2459 struct nameidata nd; 2460 2461 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2462 SCARG(uap, path), l); 2463 if ((error = namei(&nd)) != 0) 2464 return (error); 2465 error = vn_stat(nd.ni_vp, &sb, l); 2466 vput(nd.ni_vp); 2467 if (error) 2468 return (error); 2469 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2470 return (error); 2471 } 2472 2473 /* 2474 * Get file status; this version does not follow links. 2475 */ 2476 /* ARGSUSED */ 2477 int 2478 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2479 { 2480 struct sys___lstat30_args /* { 2481 syscallarg(const char *) path; 2482 syscallarg(struct stat *) ub; 2483 } */ *uap = v; 2484 struct stat sb; 2485 int error; 2486 struct nameidata nd; 2487 2488 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2489 SCARG(uap, path), l); 2490 if ((error = namei(&nd)) != 0) 2491 return (error); 2492 error = vn_stat(nd.ni_vp, &sb, l); 2493 vput(nd.ni_vp); 2494 if (error) 2495 return (error); 2496 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2497 return (error); 2498 } 2499 2500 /* 2501 * Get configurable pathname variables. 2502 */ 2503 /* ARGSUSED */ 2504 int 2505 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2506 { 2507 struct sys_pathconf_args /* { 2508 syscallarg(const char *) path; 2509 syscallarg(int) name; 2510 } */ *uap = v; 2511 int error; 2512 struct nameidata nd; 2513 2514 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2515 SCARG(uap, path), l); 2516 if ((error = namei(&nd)) != 0) 2517 return (error); 2518 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2519 vput(nd.ni_vp); 2520 return (error); 2521 } 2522 2523 /* 2524 * Return target name of a symbolic link. 2525 */ 2526 /* ARGSUSED */ 2527 int 2528 sys_readlink(struct lwp *l, void *v, register_t *retval) 2529 { 2530 struct sys_readlink_args /* { 2531 syscallarg(const char *) path; 2532 syscallarg(char *) buf; 2533 syscallarg(size_t) count; 2534 } */ *uap = v; 2535 struct vnode *vp; 2536 struct iovec aiov; 2537 struct uio auio; 2538 int error; 2539 struct nameidata nd; 2540 2541 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2542 SCARG(uap, path), l); 2543 if ((error = namei(&nd)) != 0) 2544 return (error); 2545 vp = nd.ni_vp; 2546 if (vp->v_type != VLNK) 2547 error = EINVAL; 2548 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2549 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2550 aiov.iov_base = SCARG(uap, buf); 2551 aiov.iov_len = SCARG(uap, count); 2552 auio.uio_iov = &aiov; 2553 auio.uio_iovcnt = 1; 2554 auio.uio_offset = 0; 2555 auio.uio_rw = UIO_READ; 2556 KASSERT(l == curlwp); 2557 auio.uio_vmspace = l->l_proc->p_vmspace; 2558 auio.uio_resid = SCARG(uap, count); 2559 error = VOP_READLINK(vp, &auio, l->l_cred); 2560 } 2561 vput(vp); 2562 *retval = SCARG(uap, count) - auio.uio_resid; 2563 return (error); 2564 } 2565 2566 /* 2567 * Change flags of a file given a path name. 2568 */ 2569 /* ARGSUSED */ 2570 int 2571 sys_chflags(struct lwp *l, void *v, register_t *retval) 2572 { 2573 struct sys_chflags_args /* { 2574 syscallarg(const char *) path; 2575 syscallarg(u_long) flags; 2576 } */ *uap = v; 2577 struct vnode *vp; 2578 int error; 2579 struct nameidata nd; 2580 2581 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2582 if ((error = namei(&nd)) != 0) 2583 return (error); 2584 vp = nd.ni_vp; 2585 error = change_flags(vp, SCARG(uap, flags), l); 2586 vput(vp); 2587 return (error); 2588 } 2589 2590 /* 2591 * Change flags of a file given a file descriptor. 2592 */ 2593 /* ARGSUSED */ 2594 int 2595 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2596 { 2597 struct sys_fchflags_args /* { 2598 syscallarg(int) fd; 2599 syscallarg(u_long) flags; 2600 } */ *uap = v; 2601 struct proc *p = l->l_proc; 2602 struct vnode *vp; 2603 struct file *fp; 2604 int error; 2605 2606 /* getvnode() will use the descriptor for us */ 2607 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2608 return (error); 2609 vp = (struct vnode *)fp->f_data; 2610 error = change_flags(vp, SCARG(uap, flags), l); 2611 VOP_UNLOCK(vp, 0); 2612 FILE_UNUSE(fp, l); 2613 return (error); 2614 } 2615 2616 /* 2617 * Change flags of a file given a path name; this version does 2618 * not follow links. 2619 */ 2620 int 2621 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2622 { 2623 struct sys_lchflags_args /* { 2624 syscallarg(const char *) path; 2625 syscallarg(u_long) flags; 2626 } */ *uap = v; 2627 struct vnode *vp; 2628 int error; 2629 struct nameidata nd; 2630 2631 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2632 if ((error = namei(&nd)) != 0) 2633 return (error); 2634 vp = nd.ni_vp; 2635 error = change_flags(vp, SCARG(uap, flags), l); 2636 vput(vp); 2637 return (error); 2638 } 2639 2640 /* 2641 * Common routine to change flags of a file. 2642 */ 2643 int 2644 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2645 { 2646 struct mount *mp; 2647 struct vattr vattr; 2648 int error; 2649 2650 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2651 return (error); 2652 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2653 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2654 /* 2655 * Non-superusers cannot change the flags on devices, even if they 2656 * own them. 2657 */ 2658 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2659 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2660 goto out; 2661 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2662 error = EINVAL; 2663 goto out; 2664 } 2665 } 2666 VATTR_NULL(&vattr); 2667 vattr.va_flags = flags; 2668 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2669 out: 2670 vn_finished_write(mp, 0); 2671 return (error); 2672 } 2673 2674 /* 2675 * Change mode of a file given path name; this version follows links. 2676 */ 2677 /* ARGSUSED */ 2678 int 2679 sys_chmod(struct lwp *l, void *v, register_t *retval) 2680 { 2681 struct sys_chmod_args /* { 2682 syscallarg(const char *) path; 2683 syscallarg(int) mode; 2684 } */ *uap = v; 2685 int error; 2686 struct nameidata nd; 2687 2688 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2689 if ((error = namei(&nd)) != 0) 2690 return (error); 2691 2692 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2693 2694 vrele(nd.ni_vp); 2695 return (error); 2696 } 2697 2698 /* 2699 * Change mode of a file given a file descriptor. 2700 */ 2701 /* ARGSUSED */ 2702 int 2703 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2704 { 2705 struct sys_fchmod_args /* { 2706 syscallarg(int) fd; 2707 syscallarg(int) mode; 2708 } */ *uap = v; 2709 struct proc *p = l->l_proc; 2710 struct file *fp; 2711 int error; 2712 2713 /* getvnode() will use the descriptor for us */ 2714 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2715 return (error); 2716 2717 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2718 FILE_UNUSE(fp, l); 2719 return (error); 2720 } 2721 2722 /* 2723 * Change mode of a file given path name; this version does not follow links. 2724 */ 2725 /* ARGSUSED */ 2726 int 2727 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2728 { 2729 struct sys_lchmod_args /* { 2730 syscallarg(const char *) path; 2731 syscallarg(int) mode; 2732 } */ *uap = v; 2733 int error; 2734 struct nameidata nd; 2735 2736 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2737 if ((error = namei(&nd)) != 0) 2738 return (error); 2739 2740 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2741 2742 vrele(nd.ni_vp); 2743 return (error); 2744 } 2745 2746 /* 2747 * Common routine to set mode given a vnode. 2748 */ 2749 static int 2750 change_mode(struct vnode *vp, int mode, struct lwp *l) 2751 { 2752 struct mount *mp; 2753 struct vattr vattr; 2754 int error; 2755 2756 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2757 return (error); 2758 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2759 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2760 VATTR_NULL(&vattr); 2761 vattr.va_mode = mode & ALLPERMS; 2762 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2763 VOP_UNLOCK(vp, 0); 2764 vn_finished_write(mp, 0); 2765 return (error); 2766 } 2767 2768 /* 2769 * Set ownership given a path name; this version follows links. 2770 */ 2771 /* ARGSUSED */ 2772 int 2773 sys_chown(struct lwp *l, void *v, register_t *retval) 2774 { 2775 struct sys_chown_args /* { 2776 syscallarg(const char *) path; 2777 syscallarg(uid_t) uid; 2778 syscallarg(gid_t) gid; 2779 } */ *uap = v; 2780 int error; 2781 struct nameidata nd; 2782 2783 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2784 if ((error = namei(&nd)) != 0) 2785 return (error); 2786 2787 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2788 2789 vrele(nd.ni_vp); 2790 return (error); 2791 } 2792 2793 /* 2794 * Set ownership given a path name; this version follows links. 2795 * Provides POSIX semantics. 2796 */ 2797 /* ARGSUSED */ 2798 int 2799 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2800 { 2801 struct sys_chown_args /* { 2802 syscallarg(const char *) path; 2803 syscallarg(uid_t) uid; 2804 syscallarg(gid_t) gid; 2805 } */ *uap = v; 2806 int error; 2807 struct nameidata nd; 2808 2809 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2810 if ((error = namei(&nd)) != 0) 2811 return (error); 2812 2813 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2814 2815 vrele(nd.ni_vp); 2816 return (error); 2817 } 2818 2819 /* 2820 * Set ownership given a file descriptor. 2821 */ 2822 /* ARGSUSED */ 2823 int 2824 sys_fchown(struct lwp *l, void *v, register_t *retval) 2825 { 2826 struct sys_fchown_args /* { 2827 syscallarg(int) fd; 2828 syscallarg(uid_t) uid; 2829 syscallarg(gid_t) gid; 2830 } */ *uap = v; 2831 struct proc *p = l->l_proc; 2832 int error; 2833 struct file *fp; 2834 2835 /* getvnode() will use the descriptor for us */ 2836 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2837 return (error); 2838 2839 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2840 SCARG(uap, gid), l, 0); 2841 FILE_UNUSE(fp, l); 2842 return (error); 2843 } 2844 2845 /* 2846 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2847 */ 2848 /* ARGSUSED */ 2849 int 2850 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2851 { 2852 struct sys_fchown_args /* { 2853 syscallarg(int) fd; 2854 syscallarg(uid_t) uid; 2855 syscallarg(gid_t) gid; 2856 } */ *uap = v; 2857 struct proc *p = l->l_proc; 2858 int error; 2859 struct file *fp; 2860 2861 /* getvnode() will use the descriptor for us */ 2862 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2863 return (error); 2864 2865 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2866 SCARG(uap, gid), l, 1); 2867 FILE_UNUSE(fp, l); 2868 return (error); 2869 } 2870 2871 /* 2872 * Set ownership given a path name; this version does not follow links. 2873 */ 2874 /* ARGSUSED */ 2875 int 2876 sys_lchown(struct lwp *l, void *v, register_t *retval) 2877 { 2878 struct sys_lchown_args /* { 2879 syscallarg(const char *) path; 2880 syscallarg(uid_t) uid; 2881 syscallarg(gid_t) gid; 2882 } */ *uap = v; 2883 int error; 2884 struct nameidata nd; 2885 2886 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2887 if ((error = namei(&nd)) != 0) 2888 return (error); 2889 2890 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2891 2892 vrele(nd.ni_vp); 2893 return (error); 2894 } 2895 2896 /* 2897 * Set ownership given a path name; this version does not follow links. 2898 * Provides POSIX/XPG semantics. 2899 */ 2900 /* ARGSUSED */ 2901 int 2902 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2903 { 2904 struct sys_lchown_args /* { 2905 syscallarg(const char *) path; 2906 syscallarg(uid_t) uid; 2907 syscallarg(gid_t) gid; 2908 } */ *uap = v; 2909 int error; 2910 struct nameidata nd; 2911 2912 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2913 if ((error = namei(&nd)) != 0) 2914 return (error); 2915 2916 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2917 2918 vrele(nd.ni_vp); 2919 return (error); 2920 } 2921 2922 /* 2923 * Common routine to set ownership given a vnode. 2924 */ 2925 static int 2926 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2927 int posix_semantics) 2928 { 2929 struct mount *mp; 2930 struct vattr vattr; 2931 mode_t newmode; 2932 int error; 2933 2934 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2935 return (error); 2936 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2937 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2938 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2939 goto out; 2940 2941 #define CHANGED(x) ((int)(x) != -1) 2942 newmode = vattr.va_mode; 2943 if (posix_semantics) { 2944 /* 2945 * POSIX/XPG semantics: if the caller is not the super-user, 2946 * clear set-user-id and set-group-id bits. Both POSIX and 2947 * the XPG consider the behaviour for calls by the super-user 2948 * implementation-defined; we leave the set-user-id and set- 2949 * group-id settings intact in that case. 2950 */ 2951 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2952 NULL) != 0) 2953 newmode &= ~(S_ISUID | S_ISGID); 2954 } else { 2955 /* 2956 * NetBSD semantics: when changing owner and/or group, 2957 * clear the respective bit(s). 2958 */ 2959 if (CHANGED(uid)) 2960 newmode &= ~S_ISUID; 2961 if (CHANGED(gid)) 2962 newmode &= ~S_ISGID; 2963 } 2964 /* Update va_mode iff altered. */ 2965 if (vattr.va_mode == newmode) 2966 newmode = VNOVAL; 2967 2968 VATTR_NULL(&vattr); 2969 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2970 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2971 vattr.va_mode = newmode; 2972 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2973 #undef CHANGED 2974 2975 out: 2976 VOP_UNLOCK(vp, 0); 2977 vn_finished_write(mp, 0); 2978 return (error); 2979 } 2980 2981 /* 2982 * Set the access and modification times given a path name; this 2983 * version follows links. 2984 */ 2985 /* ARGSUSED */ 2986 int 2987 sys_utimes(struct lwp *l, void *v, register_t *retval) 2988 { 2989 struct sys_utimes_args /* { 2990 syscallarg(const char *) path; 2991 syscallarg(const struct timeval *) tptr; 2992 } */ *uap = v; 2993 int error; 2994 struct nameidata nd; 2995 2996 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2997 if ((error = namei(&nd)) != 0) 2998 return (error); 2999 3000 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3001 3002 vrele(nd.ni_vp); 3003 return (error); 3004 } 3005 3006 /* 3007 * Set the access and modification times given a file descriptor. 3008 */ 3009 /* ARGSUSED */ 3010 int 3011 sys_futimes(struct lwp *l, void *v, register_t *retval) 3012 { 3013 struct sys_futimes_args /* { 3014 syscallarg(int) fd; 3015 syscallarg(const struct timeval *) tptr; 3016 } */ *uap = v; 3017 struct proc *p = l->l_proc; 3018 int error; 3019 struct file *fp; 3020 3021 /* getvnode() will use the descriptor for us */ 3022 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3023 return (error); 3024 3025 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3026 FILE_UNUSE(fp, l); 3027 return (error); 3028 } 3029 3030 /* 3031 * Set the access and modification times given a path name; this 3032 * version does not follow links. 3033 */ 3034 /* ARGSUSED */ 3035 int 3036 sys_lutimes(struct lwp *l, void *v, register_t *retval) 3037 { 3038 struct sys_lutimes_args /* { 3039 syscallarg(const char *) path; 3040 syscallarg(const struct timeval *) tptr; 3041 } */ *uap = v; 3042 int error; 3043 struct nameidata nd; 3044 3045 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3046 if ((error = namei(&nd)) != 0) 3047 return (error); 3048 3049 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3050 3051 vrele(nd.ni_vp); 3052 return (error); 3053 } 3054 3055 /* 3056 * Common routine to set access and modification times given a vnode. 3057 */ 3058 static int 3059 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3060 { 3061 struct mount *mp; 3062 struct vattr vattr; 3063 int error; 3064 3065 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3066 return (error); 3067 VATTR_NULL(&vattr); 3068 if (tptr == NULL) { 3069 nanotime(&vattr.va_atime); 3070 vattr.va_mtime = vattr.va_atime; 3071 vattr.va_vaflags |= VA_UTIMES_NULL; 3072 } else { 3073 struct timeval tv[2]; 3074 3075 error = copyin(tptr, tv, sizeof(tv)); 3076 if (error) 3077 goto out; 3078 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3079 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3080 } 3081 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3082 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3083 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3084 VOP_UNLOCK(vp, 0); 3085 out: 3086 vn_finished_write(mp, 0); 3087 return (error); 3088 } 3089 3090 /* 3091 * Truncate a file given its path name. 3092 */ 3093 /* ARGSUSED */ 3094 int 3095 sys_truncate(struct lwp *l, void *v, register_t *retval) 3096 { 3097 struct sys_truncate_args /* { 3098 syscallarg(const char *) path; 3099 syscallarg(int) pad; 3100 syscallarg(off_t) length; 3101 } */ *uap = v; 3102 struct vnode *vp; 3103 struct mount *mp; 3104 struct vattr vattr; 3105 int error; 3106 struct nameidata nd; 3107 3108 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3109 if ((error = namei(&nd)) != 0) 3110 return (error); 3111 vp = nd.ni_vp; 3112 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3113 vrele(vp); 3114 return (error); 3115 } 3116 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3117 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3118 if (vp->v_type == VDIR) 3119 error = EISDIR; 3120 else if ((error = vn_writechk(vp)) == 0 && 3121 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3122 VATTR_NULL(&vattr); 3123 vattr.va_size = SCARG(uap, length); 3124 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3125 } 3126 vput(vp); 3127 vn_finished_write(mp, 0); 3128 return (error); 3129 } 3130 3131 /* 3132 * Truncate a file given a file descriptor. 3133 */ 3134 /* ARGSUSED */ 3135 int 3136 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3137 { 3138 struct sys_ftruncate_args /* { 3139 syscallarg(int) fd; 3140 syscallarg(int) pad; 3141 syscallarg(off_t) length; 3142 } */ *uap = v; 3143 struct proc *p = l->l_proc; 3144 struct mount *mp; 3145 struct vattr vattr; 3146 struct vnode *vp; 3147 struct file *fp; 3148 int error; 3149 3150 /* getvnode() will use the descriptor for us */ 3151 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3152 return (error); 3153 if ((fp->f_flag & FWRITE) == 0) { 3154 error = EINVAL; 3155 goto out; 3156 } 3157 vp = (struct vnode *)fp->f_data; 3158 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3159 FILE_UNUSE(fp, l); 3160 return (error); 3161 } 3162 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3163 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3164 if (vp->v_type == VDIR) 3165 error = EISDIR; 3166 else if ((error = vn_writechk(vp)) == 0) { 3167 VATTR_NULL(&vattr); 3168 vattr.va_size = SCARG(uap, length); 3169 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3170 } 3171 VOP_UNLOCK(vp, 0); 3172 vn_finished_write(mp, 0); 3173 out: 3174 FILE_UNUSE(fp, l); 3175 return (error); 3176 } 3177 3178 /* 3179 * Sync an open file. 3180 */ 3181 /* ARGSUSED */ 3182 int 3183 sys_fsync(struct lwp *l, void *v, register_t *retval) 3184 { 3185 struct sys_fsync_args /* { 3186 syscallarg(int) fd; 3187 } */ *uap = v; 3188 struct proc *p = l->l_proc; 3189 struct vnode *vp; 3190 struct mount *mp; 3191 struct file *fp; 3192 int error; 3193 3194 /* getvnode() will use the descriptor for us */ 3195 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3196 return (error); 3197 vp = (struct vnode *)fp->f_data; 3198 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3199 FILE_UNUSE(fp, l); 3200 return (error); 3201 } 3202 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3203 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3204 if (error == 0 && bioops.io_fsync != NULL && 3205 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3206 (*bioops.io_fsync)(vp, 0); 3207 VOP_UNLOCK(vp, 0); 3208 vn_finished_write(mp, 0); 3209 FILE_UNUSE(fp, l); 3210 return (error); 3211 } 3212 3213 /* 3214 * Sync a range of file data. API modeled after that found in AIX. 3215 * 3216 * FDATASYNC indicates that we need only save enough metadata to be able 3217 * to re-read the written data. Note we duplicate AIX's requirement that 3218 * the file be open for writing. 3219 */ 3220 /* ARGSUSED */ 3221 int 3222 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3223 { 3224 struct sys_fsync_range_args /* { 3225 syscallarg(int) fd; 3226 syscallarg(int) flags; 3227 syscallarg(off_t) start; 3228 syscallarg(off_t) length; 3229 } */ *uap = v; 3230 struct proc *p = l->l_proc; 3231 struct vnode *vp; 3232 struct file *fp; 3233 int flags, nflags; 3234 off_t s, e, len; 3235 int error; 3236 3237 /* getvnode() will use the descriptor for us */ 3238 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3239 return (error); 3240 3241 if ((fp->f_flag & FWRITE) == 0) { 3242 error = EBADF; 3243 goto out; 3244 } 3245 3246 flags = SCARG(uap, flags); 3247 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3248 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3249 error = EINVAL; 3250 goto out; 3251 } 3252 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3253 if (flags & FDATASYNC) 3254 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3255 else 3256 nflags = FSYNC_WAIT; 3257 if (flags & FDISKSYNC) 3258 nflags |= FSYNC_CACHE; 3259 3260 len = SCARG(uap, length); 3261 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3262 if (len) { 3263 s = SCARG(uap, start); 3264 e = s + len; 3265 if (e < s) { 3266 error = EINVAL; 3267 goto out; 3268 } 3269 } else { 3270 e = 0; 3271 s = 0; 3272 } 3273 3274 vp = (struct vnode *)fp->f_data; 3275 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3276 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3277 3278 if (error == 0 && bioops.io_fsync != NULL && 3279 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3280 (*bioops.io_fsync)(vp, nflags); 3281 3282 VOP_UNLOCK(vp, 0); 3283 out: 3284 FILE_UNUSE(fp, l); 3285 return (error); 3286 } 3287 3288 /* 3289 * Sync the data of an open file. 3290 */ 3291 /* ARGSUSED */ 3292 int 3293 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3294 { 3295 struct sys_fdatasync_args /* { 3296 syscallarg(int) fd; 3297 } */ *uap = v; 3298 struct proc *p = l->l_proc; 3299 struct vnode *vp; 3300 struct file *fp; 3301 int error; 3302 3303 /* getvnode() will use the descriptor for us */ 3304 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3305 return (error); 3306 if ((fp->f_flag & FWRITE) == 0) { 3307 FILE_UNUSE(fp, l); 3308 return (EBADF); 3309 } 3310 vp = (struct vnode *)fp->f_data; 3311 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3312 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3313 VOP_UNLOCK(vp, 0); 3314 FILE_UNUSE(fp, l); 3315 return (error); 3316 } 3317 3318 /* 3319 * Rename files, (standard) BSD semantics frontend. 3320 */ 3321 /* ARGSUSED */ 3322 int 3323 sys_rename(struct lwp *l, void *v, register_t *retval) 3324 { 3325 struct sys_rename_args /* { 3326 syscallarg(const char *) from; 3327 syscallarg(const char *) to; 3328 } */ *uap = v; 3329 3330 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3331 } 3332 3333 /* 3334 * Rename files, POSIX semantics frontend. 3335 */ 3336 /* ARGSUSED */ 3337 int 3338 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3339 { 3340 struct sys___posix_rename_args /* { 3341 syscallarg(const char *) from; 3342 syscallarg(const char *) to; 3343 } */ *uap = v; 3344 3345 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3346 } 3347 3348 /* 3349 * Rename files. Source and destination must either both be directories, 3350 * or both not be directories. If target is a directory, it must be empty. 3351 * If `from' and `to' refer to the same object, the value of the `retain' 3352 * argument is used to determine whether `from' will be 3353 * 3354 * (retain == 0) deleted unless `from' and `to' refer to the same 3355 * object in the file system's name space (BSD). 3356 * (retain == 1) always retained (POSIX). 3357 */ 3358 static int 3359 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3360 { 3361 struct mount *mp = NULL; 3362 struct vnode *tvp, *fvp, *tdvp; 3363 struct nameidata fromnd, tond; 3364 struct proc *p; 3365 int error; 3366 3367 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3368 from, l); 3369 if ((error = namei(&fromnd)) != 0) 3370 return (error); 3371 if (fromnd.ni_dvp != fromnd.ni_vp) 3372 VOP_UNLOCK(fromnd.ni_dvp, 0); 3373 fvp = fromnd.ni_vp; 3374 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3375 if (error != 0) { 3376 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3377 vrele(fromnd.ni_dvp); 3378 vrele(fvp); 3379 if (fromnd.ni_startdir) 3380 vrele(fromnd.ni_startdir); 3381 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3382 return (error); 3383 } 3384 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3385 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3386 if ((error = namei(&tond)) != 0) { 3387 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3388 vrele(fromnd.ni_dvp); 3389 vrele(fvp); 3390 goto out1; 3391 } 3392 tdvp = tond.ni_dvp; 3393 tvp = tond.ni_vp; 3394 3395 if (tvp != NULL) { 3396 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3397 error = ENOTDIR; 3398 goto out; 3399 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3400 error = EISDIR; 3401 goto out; 3402 } 3403 } 3404 3405 if (fvp == tdvp) 3406 error = EINVAL; 3407 3408 /* 3409 * Source and destination refer to the same object. 3410 */ 3411 if (fvp == tvp) { 3412 if (retain) 3413 error = -1; 3414 else if (fromnd.ni_dvp == tdvp && 3415 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3416 !memcmp(fromnd.ni_cnd.cn_nameptr, 3417 tond.ni_cnd.cn_nameptr, 3418 fromnd.ni_cnd.cn_namelen)) 3419 error = -1; 3420 } 3421 3422 #if NVERIEXEC > 0 3423 if (!error) { 3424 pathname_t frompath = NULL, topath = NULL; 3425 3426 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3427 &frompath); 3428 if (!error) 3429 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3430 &topath); 3431 if (!error) 3432 error = veriexec_renamechk(fvp, pathname_path(frompath), 3433 tvp, pathname_path(topath), l); 3434 3435 pathname_put(frompath); 3436 pathname_put(topath); 3437 } 3438 #endif /* NVERIEXEC > 0 */ 3439 3440 out: 3441 p = l->l_proc; 3442 if (!error) { 3443 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3444 if (fromnd.ni_dvp != tdvp) 3445 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3446 if (tvp) { 3447 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3448 } 3449 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3450 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3451 } else { 3452 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3453 if (tdvp == tvp) 3454 vrele(tdvp); 3455 else 3456 vput(tdvp); 3457 if (tvp) 3458 vput(tvp); 3459 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3460 vrele(fromnd.ni_dvp); 3461 vrele(fvp); 3462 } 3463 vrele(tond.ni_startdir); 3464 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3465 out1: 3466 vn_finished_write(mp, 0); 3467 if (fromnd.ni_startdir) 3468 vrele(fromnd.ni_startdir); 3469 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3470 return (error == -1 ? 0 : error); 3471 } 3472 3473 /* 3474 * Make a directory file. 3475 */ 3476 /* ARGSUSED */ 3477 int 3478 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3479 { 3480 struct sys_mkdir_args /* { 3481 syscallarg(const char *) path; 3482 syscallarg(int) mode; 3483 } */ *uap = v; 3484 struct proc *p = l->l_proc; 3485 struct mount *mp; 3486 struct vnode *vp; 3487 struct vattr vattr; 3488 int error; 3489 struct nameidata nd; 3490 3491 restart: 3492 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3493 SCARG(uap, path), l); 3494 if ((error = namei(&nd)) != 0) 3495 return (error); 3496 vp = nd.ni_vp; 3497 if (vp != NULL) { 3498 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3499 if (nd.ni_dvp == vp) 3500 vrele(nd.ni_dvp); 3501 else 3502 vput(nd.ni_dvp); 3503 vrele(vp); 3504 return (EEXIST); 3505 } 3506 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3507 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3508 if (nd.ni_dvp == vp) 3509 vrele(nd.ni_dvp); 3510 else 3511 vput(nd.ni_dvp); 3512 if ((error = vn_start_write(NULL, &mp, 3513 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3514 return (error); 3515 goto restart; 3516 } 3517 VATTR_NULL(&vattr); 3518 vattr.va_type = VDIR; 3519 vattr.va_mode = 3520 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3521 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3522 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3523 if (!error) 3524 vput(nd.ni_vp); 3525 vn_finished_write(mp, 0); 3526 return (error); 3527 } 3528 3529 /* 3530 * Remove a directory file. 3531 */ 3532 /* ARGSUSED */ 3533 int 3534 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3535 { 3536 struct sys_rmdir_args /* { 3537 syscallarg(const char *) path; 3538 } */ *uap = v; 3539 struct mount *mp; 3540 struct vnode *vp; 3541 int error; 3542 struct nameidata nd; 3543 3544 restart: 3545 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3546 SCARG(uap, path), l); 3547 if ((error = namei(&nd)) != 0) 3548 return (error); 3549 vp = nd.ni_vp; 3550 if (vp->v_type != VDIR) { 3551 error = ENOTDIR; 3552 goto out; 3553 } 3554 /* 3555 * No rmdir "." please. 3556 */ 3557 if (nd.ni_dvp == vp) { 3558 error = EINVAL; 3559 goto out; 3560 } 3561 /* 3562 * The root of a mounted filesystem cannot be deleted. 3563 */ 3564 if (vp->v_flag & VROOT) { 3565 error = EBUSY; 3566 goto out; 3567 } 3568 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3569 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3570 if (nd.ni_dvp == vp) 3571 vrele(nd.ni_dvp); 3572 else 3573 vput(nd.ni_dvp); 3574 vput(vp); 3575 if ((error = vn_start_write(NULL, &mp, 3576 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3577 return (error); 3578 goto restart; 3579 } 3580 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3581 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3582 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3583 vn_finished_write(mp, 0); 3584 return (error); 3585 3586 out: 3587 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3588 if (nd.ni_dvp == vp) 3589 vrele(nd.ni_dvp); 3590 else 3591 vput(nd.ni_dvp); 3592 vput(vp); 3593 return (error); 3594 } 3595 3596 /* 3597 * Read a block of directory entries in a file system independent format. 3598 */ 3599 int 3600 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3601 { 3602 struct sys___getdents30_args /* { 3603 syscallarg(int) fd; 3604 syscallarg(char *) buf; 3605 syscallarg(size_t) count; 3606 } */ *uap = v; 3607 struct proc *p = l->l_proc; 3608 struct file *fp; 3609 int error, done; 3610 3611 /* getvnode() will use the descriptor for us */ 3612 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3613 return (error); 3614 if ((fp->f_flag & FREAD) == 0) { 3615 error = EBADF; 3616 goto out; 3617 } 3618 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3619 SCARG(uap, count), &done, l, 0, 0); 3620 #ifdef KTRACE 3621 if (!error && KTRPOINT(p, KTR_GENIO)) { 3622 struct iovec iov; 3623 iov.iov_base = SCARG(uap, buf); 3624 iov.iov_len = done; 3625 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3626 } 3627 #endif 3628 *retval = done; 3629 out: 3630 FILE_UNUSE(fp, l); 3631 return (error); 3632 } 3633 3634 /* 3635 * Set the mode mask for creation of filesystem nodes. 3636 */ 3637 int 3638 sys_umask(struct lwp *l, void *v, register_t *retval) 3639 { 3640 struct sys_umask_args /* { 3641 syscallarg(mode_t) newmask; 3642 } */ *uap = v; 3643 struct proc *p = l->l_proc; 3644 struct cwdinfo *cwdi; 3645 3646 cwdi = p->p_cwdi; 3647 *retval = cwdi->cwdi_cmask; 3648 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3649 return (0); 3650 } 3651 3652 /* 3653 * Void all references to file by ripping underlying filesystem 3654 * away from vnode. 3655 */ 3656 /* ARGSUSED */ 3657 int 3658 sys_revoke(struct lwp *l, void *v, register_t *retval) 3659 { 3660 struct sys_revoke_args /* { 3661 syscallarg(const char *) path; 3662 } */ *uap = v; 3663 struct mount *mp; 3664 struct vnode *vp; 3665 struct vattr vattr; 3666 int error; 3667 struct nameidata nd; 3668 3669 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3670 if ((error = namei(&nd)) != 0) 3671 return (error); 3672 vp = nd.ni_vp; 3673 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3674 goto out; 3675 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3676 (error = kauth_authorize_generic(l->l_cred, 3677 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3678 goto out; 3679 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3680 goto out; 3681 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3682 VOP_REVOKE(vp, REVOKEALL); 3683 vn_finished_write(mp, 0); 3684 out: 3685 vrele(vp); 3686 return (error); 3687 } 3688 3689 /* 3690 * Convert a user file descriptor to a kernel file entry. 3691 */ 3692 int 3693 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3694 { 3695 struct vnode *vp; 3696 struct file *fp; 3697 3698 if ((fp = fd_getfile(fdp, fd)) == NULL) 3699 return (EBADF); 3700 3701 FILE_USE(fp); 3702 3703 if (fp->f_type != DTYPE_VNODE) { 3704 FILE_UNUSE(fp, NULL); 3705 return (EINVAL); 3706 } 3707 3708 vp = (struct vnode *)fp->f_data; 3709 if (vp->v_type == VBAD) { 3710 FILE_UNUSE(fp, NULL); 3711 return (EBADF); 3712 } 3713 3714 *fpp = fp; 3715 return (0); 3716 } 3717
Indexes created Wed Sep 24 05:09:52 GMT 2025