vfs_syscalls.c revision 1.306.2.1
1 /* $NetBSD: vfs_syscalls.c,v 1.306.2.1 2007/03/13 17:51:02 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.306.2.1 2007/03/13 17:51:02 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/syscallargs.h> 65 #include <sys/vfs_syscalls.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #if NVERIEXEC > 0 73 #include <sys/verified_exec.h> 74 #include <sys/syslog.h> 75 #endif /* NVERIEXEC > 0 */ 76 #include <sys/kauth.h> 77 78 #include <miscfs/genfs/genfs.h> 79 #include <miscfs/syncfs/syncfs.h> 80 81 #ifdef COMPAT_30 82 #include "opt_nfsserver.h" 83 #include <nfs/rpcv2.h> 84 #endif 85 #include <nfs/nfsproto.h> 86 #ifdef COMPAT_30 87 #include <nfs/nfs.h> 88 #include <nfs/nfs_var.h> 89 #endif 90 91 #if NFSS > 0 92 #include <dev/fssvar.h> 93 #endif 94 95 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 96 97 static int change_dir(struct nameidata *, struct lwp *); 98 static int change_flags(struct vnode *, u_long, struct lwp *); 99 static int change_mode(struct vnode *, int, struct lwp *l); 100 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 101 static int change_utimes(struct vnode *vp, const struct timeval *, 102 struct lwp *l); 103 static int rename_files(const char *, const char *, struct lwp *, int); 104 105 void checkdirs(struct vnode *); 106 107 static int mount_update(struct lwp *, struct vnode *, const char *, int, 108 void *, struct nameidata *); 109 static int mount_domount(struct lwp *, struct vnode *, const char *, 110 const char *, int, void *, struct nameidata *); 111 static int mount_getargs(struct lwp *, struct vnode *, const char *, int, 112 void *, struct nameidata *); 113 114 int dovfsusermount = 0; 115 116 /* 117 * Virtual File System System Calls 118 */ 119 120 /* 121 * Mount a file system. 122 */ 123 124 #if defined(COMPAT_09) || defined(COMPAT_43) 125 /* 126 * This table is used to maintain compatibility with 4.3BSD 127 * and NetBSD 0.9 mount syscalls. Note, the order is important! 128 * 129 * Do not modify this table. It should only contain filesystems 130 * supported by NetBSD 0.9 and 4.3BSD. 131 */ 132 const char * const mountcompatnames[] = { 133 NULL, /* 0 = MOUNT_NONE */ 134 MOUNT_FFS, /* 1 = MOUNT_UFS */ 135 MOUNT_NFS, /* 2 */ 136 MOUNT_MFS, /* 3 */ 137 MOUNT_MSDOS, /* 4 */ 138 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 139 MOUNT_FDESC, /* 6 */ 140 MOUNT_KERNFS, /* 7 */ 141 NULL, /* 8 = MOUNT_DEVFS */ 142 MOUNT_AFS, /* 9 */ 143 }; 144 const int nmountcompatnames = sizeof(mountcompatnames) / 145 sizeof(mountcompatnames[0]); 146 #endif /* COMPAT_09 || COMPAT_43 */ 147 148 static int 149 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 150 void *data, struct nameidata *ndp) 151 { 152 struct mount *mp; 153 int error = 0, saved_flags; 154 155 mp = vp->v_mount; 156 saved_flags = mp->mnt_flag; 157 158 /* We can't operate on VROOT here. */ 159 if ((vp->v_flag & VROOT) == 0) { 160 vput(vp); 161 error = EINVAL; 162 goto out; 163 } 164 165 /* 166 * We only allow the filesystem to be reloaded if it 167 * is currently mounted read-only. 168 */ 169 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 170 error = EOPNOTSUPP; /* Needs translation */ 171 goto out; 172 } 173 174 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 175 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 176 if (error) 177 goto out; 178 179 if (vfs_busy(mp, LK_NOWAIT, 0)) { 180 error = EPERM; 181 goto out; 182 } 183 184 mp->mnt_flag &= ~MNT_OP_FLAGS; 185 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 186 187 /* 188 * Set the mount level flags. 189 */ 190 if (flags & MNT_RDONLY) 191 mp->mnt_flag |= MNT_RDONLY; 192 else if (mp->mnt_flag & MNT_RDONLY) 193 mp->mnt_iflag |= IMNT_WANTRDWR; 194 mp->mnt_flag &= 195 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 196 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 197 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 198 mp->mnt_flag |= flags & 199 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 200 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 201 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 202 MNT_IGNORE); 203 204 error = VFS_MOUNT(mp, path, data, ndp, l); 205 206 #if defined(COMPAT_30) && defined(NFSSERVER) 207 if (error) { 208 int error2; 209 210 /* Update failed; let's try and see if it was an 211 * export request. */ 212 error2 = nfs_update_exports_30(mp, path, data, l); 213 214 /* Only update error code if the export request was 215 * understood but some problem occurred while 216 * processing it. */ 217 if (error2 != EJUSTRETURN) 218 error = error2; 219 } 220 #endif 221 if (mp->mnt_iflag & IMNT_WANTRDWR) 222 mp->mnt_flag &= ~MNT_RDONLY; 223 if (error) 224 mp->mnt_flag = saved_flags; 225 mp->mnt_flag &= ~MNT_OP_FLAGS; 226 mp->mnt_iflag &= ~IMNT_WANTRDWR; 227 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 228 if (mp->mnt_syncer == NULL) 229 error = vfs_allocate_syncvnode(mp); 230 } else { 231 if (mp->mnt_syncer != NULL) 232 vfs_deallocate_syncvnode(mp); 233 } 234 vfs_unbusy(mp); 235 236 out: 237 return (error); 238 } 239 240 static int 241 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 242 const char *path, int flags, void *data, struct nameidata *ndp) 243 { 244 struct mount *mp = NULL; 245 struct vattr va; 246 char fstypename[MFSNAMELEN]; 247 int error; 248 249 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 250 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 251 if (error) { 252 vput(vp); 253 goto out; 254 } 255 256 /* Can't make a non-dir a mount-point (from here anyway). */ 257 if (vp->v_type != VDIR) { 258 error = ENOTDIR; 259 vput(vp); 260 goto out; 261 } 262 263 /* 264 * If the user is not root, ensure that they own the directory 265 * onto which we are attempting to mount. 266 */ 267 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 268 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 269 (error = kauth_authorize_generic(l->l_cred, 270 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 271 vput(vp); 272 goto out; 273 } 274 275 if (flags & MNT_EXPORTED) { 276 error = EINVAL; 277 vput(vp); 278 goto out; 279 } 280 281 /* 282 * Copy file-system type from userspace. 283 */ 284 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 285 if (error) { 286 #if defined(COMPAT_09) || defined(COMPAT_43) 287 /* 288 * Historically, filesystem types were identified by numbers. 289 * If we get an integer for the filesystem type instead of a 290 * string, we check to see if it matches one of the historic 291 * filesystem types. 292 */ 293 u_long fsindex = (u_long)fstype; 294 if (fsindex >= nmountcompatnames || 295 mountcompatnames[fsindex] == NULL) { 296 error = ENODEV; 297 vput(vp); 298 goto out; 299 } 300 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 301 #else 302 vput(vp); 303 goto out; 304 #endif 305 } 306 307 #ifdef COMPAT_10 308 /* Accept `ufs' as an alias for `ffs'. */ 309 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 310 strlcpy(fstypename, "ffs", sizeof(fstypename)); 311 #endif 312 313 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 314 vput(vp); 315 goto out; 316 } 317 318 /* 319 * Check if a file-system is not already mounted on this vnode. 320 */ 321 if (vp->v_mountedhere != NULL) { 322 error = EBUSY; 323 vput(vp); 324 goto out; 325 } 326 327 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 328 329 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 330 free(mp, M_MOUNT); 331 error = ENODEV; 332 vput(vp); 333 goto out; 334 } 335 336 TAILQ_INIT(&mp->mnt_vnodelist); 337 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 338 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 339 (void)vfs_busy(mp, LK_NOWAIT, 0); 340 341 mp->mnt_op->vfs_refcount++; 342 mp->mnt_vnodecovered = vp; 343 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 344 mp->mnt_unmounter = NULL; 345 mp->mnt_leaf = mp; 346 mount_initspecific(mp); 347 348 /* 349 * The underlying file system may refuse the mount for 350 * various reasons. Allow the user to force it to happen. 351 * 352 * Set the mount level flags. 353 */ 354 mp->mnt_flag = flags & 355 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 356 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 357 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 358 MNT_IGNORE | MNT_RDONLY); 359 360 error = VFS_MOUNT(mp, path, data, ndp, l); 361 mp->mnt_flag &= ~MNT_OP_FLAGS; 362 363 /* 364 * Put the new filesystem on the mount list after root. 365 */ 366 cache_purge(vp); 367 if (!error) { 368 mp->mnt_iflag &= ~IMNT_WANTRDWR; 369 vp->v_mountedhere = mp; 370 mutex_enter(&mountlist_lock); 371 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 372 mutex_exit(&mountlist_lock); 373 VOP_UNLOCK(vp, 0); 374 checkdirs(vp); 375 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 376 error = vfs_allocate_syncvnode(mp); 377 vfs_unbusy(mp); 378 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 379 error = VFS_START(mp, 0, l); 380 if (error) 381 vrele(vp); 382 } else { 383 vp->v_mountedhere = NULL; 384 mp->mnt_op->vfs_refcount--; 385 vfs_unbusy(mp); 386 free(mp, M_MOUNT); 387 vput(vp); 388 } 389 390 out: 391 return (error); 392 } 393 394 static int 395 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 396 void *data, struct nameidata *ndp) 397 { 398 struct mount *mp; 399 int error; 400 401 /* If MNT_GETARGS is specified, it should be the only flag. */ 402 if (flags & ~MNT_GETARGS) { 403 error = EINVAL; 404 goto out; 405 } 406 407 mp = vp->v_mount; 408 409 /* XXX: probably some notion of "can see" here if we want isolation. */ 410 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 411 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 412 if (error) 413 goto out; 414 415 if ((vp->v_flag & VROOT) == 0) { 416 error = EINVAL; 417 goto out; 418 } 419 420 if (vfs_busy(mp, LK_NOWAIT, 0)) { 421 error = EPERM; 422 goto out; 423 } 424 425 mp->mnt_flag &= ~MNT_OP_FLAGS; 426 mp->mnt_flag |= MNT_GETARGS; 427 error = VFS_MOUNT(mp, path, data, ndp, l); 428 mp->mnt_flag &= ~MNT_OP_FLAGS; 429 430 vfs_unbusy(mp); 431 out: 432 return (error); 433 } 434 435 /* ARGSUSED */ 436 int 437 sys_mount(struct lwp *l, void *v, register_t *retval) 438 { 439 struct sys_mount_args /* { 440 syscallarg(const char *) type; 441 syscallarg(const char *) path; 442 syscallarg(int) flags; 443 syscallarg(void *) data; 444 } */ *uap = v; 445 struct vnode *vp; 446 struct nameidata nd; 447 int error; 448 449 /* 450 * Get vnode to be covered 451 */ 452 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 453 SCARG(uap, path), l); 454 if ((error = namei(&nd)) != 0) 455 return (error); 456 vp = nd.ni_vp; 457 458 /* 459 * A lookup in VFS_MOUNT might result in an attempt to 460 * lock this vnode again, so make the lock recursive. 461 */ 462 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 463 464 if (SCARG(uap, flags) & MNT_GETARGS) { 465 error = mount_getargs(l, vp, SCARG(uap, path), 466 SCARG(uap, flags), SCARG(uap, data), &nd); 467 vput(vp); 468 } else if (SCARG(uap, flags) & MNT_UPDATE) { 469 error = mount_update(l, vp, SCARG(uap, path), 470 SCARG(uap, flags), SCARG(uap, data), &nd); 471 vput(vp); 472 } else { 473 /* Locking is handled internally in mount_domount(). */ 474 error = mount_domount(l, vp, SCARG(uap, type), 475 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 476 } 477 478 return (error); 479 } 480 481 /* 482 * Scan all active processes to see if any of them have a current 483 * or root directory onto which the new filesystem has just been 484 * mounted. If so, replace them with the new mount point. 485 */ 486 void 487 checkdirs(struct vnode *olddp) 488 { 489 struct cwdinfo *cwdi; 490 struct vnode *newdp; 491 struct proc *p; 492 493 if (olddp->v_usecount == 1) 494 return; 495 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 496 panic("mount: lost mount"); 497 mutex_enter(&proclist_lock); 498 PROCLIST_FOREACH(p, &allproc) { 499 cwdi = p->p_cwdi; 500 if (!cwdi) 501 continue; 502 if (cwdi->cwdi_cdir == olddp) { 503 vrele(cwdi->cwdi_cdir); 504 VREF(newdp); 505 cwdi->cwdi_cdir = newdp; 506 } 507 if (cwdi->cwdi_rdir == olddp) { 508 vrele(cwdi->cwdi_rdir); 509 VREF(newdp); 510 cwdi->cwdi_rdir = newdp; 511 } 512 } 513 mutex_exit(&proclist_lock); 514 if (rootvnode == olddp) { 515 vrele(rootvnode); 516 VREF(newdp); 517 rootvnode = newdp; 518 } 519 vput(newdp); 520 } 521 522 /* 523 * Unmount a file system. 524 * 525 * Note: unmount takes a path to the vnode mounted on as argument, 526 * not special file (as before). 527 */ 528 /* ARGSUSED */ 529 int 530 sys_unmount(struct lwp *l, void *v, register_t *retval) 531 { 532 struct sys_unmount_args /* { 533 syscallarg(const char *) path; 534 syscallarg(int) flags; 535 } */ *uap = v; 536 struct vnode *vp; 537 struct mount *mp; 538 int error; 539 struct nameidata nd; 540 541 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 542 SCARG(uap, path), l); 543 if ((error = namei(&nd)) != 0) 544 return (error); 545 vp = nd.ni_vp; 546 mp = vp->v_mount; 547 548 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 549 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 550 if (error) { 551 vput(vp); 552 return (error); 553 } 554 555 /* 556 * Don't allow unmounting the root file system. 557 */ 558 if (mp->mnt_flag & MNT_ROOTFS) { 559 vput(vp); 560 return (EINVAL); 561 } 562 563 /* 564 * Must be the root of the filesystem 565 */ 566 if ((vp->v_flag & VROOT) == 0) { 567 vput(vp); 568 return (EINVAL); 569 } 570 vput(vp); 571 572 /* 573 * XXX Freeze syncer. Must do this before locking the 574 * mount point. See dounmount() for details. 575 */ 576 mutex_enter(&syncer_mutex); 577 578 if (vfs_busy(mp, 0, 0)) { 579 mutex_exit(&syncer_mutex); 580 return (EBUSY); 581 } 582 583 return (dounmount(mp, SCARG(uap, flags), l)); 584 } 585 586 /* 587 * Do the actual file system unmount. File system is assumed to have been 588 * marked busy by the caller. 589 */ 590 int 591 dounmount(struct mount *mp, int flags, struct lwp *l) 592 { 593 struct vnode *coveredvp; 594 int error; 595 int async; 596 int used_syncer; 597 598 #if NVERIEXEC > 0 599 error = veriexec_unmountchk(mp); 600 if (error) 601 return (error); 602 #endif /* NVERIEXEC > 0 */ 603 604 mutex_enter(&mountlist_lock); 605 vfs_unbusy(mp); 606 used_syncer = (mp->mnt_syncer != NULL); 607 608 /* 609 * XXX Syncer must be frozen when we get here. This should really 610 * be done on a per-mountpoint basis, but especially the softdep 611 * code possibly called from the syncer doesn't exactly work on a 612 * per-mountpoint basis, so the softdep code would become a maze 613 * of vfs_busy() calls. 614 * 615 * The caller of dounmount() must acquire syncer_mutex because 616 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 617 * order, and we must preserve that order to avoid deadlock. 618 * 619 * So, if the file system did not use the syncer, now is 620 * the time to release the syncer_mutex. 621 */ 622 if (used_syncer == 0) 623 mutex_exit(&syncer_mutex); 624 625 mp->mnt_iflag |= IMNT_UNMOUNT; 626 mp->mnt_unmounter = l; 627 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 628 vn_start_write(NULL, &mp, V_WAIT); 629 630 async = mp->mnt_flag & MNT_ASYNC; 631 mp->mnt_flag &= ~MNT_ASYNC; 632 cache_purgevfs(mp); /* remove cache entries for this file sys */ 633 if (mp->mnt_syncer != NULL) 634 vfs_deallocate_syncvnode(mp); 635 error = 0; 636 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 637 #if NFSS > 0 638 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 639 #endif 640 if (error == 0) 641 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 642 } 643 if (error == 0 || (flags & MNT_FORCE)) 644 error = VFS_UNMOUNT(mp, flags, l); 645 vn_finished_write(mp, 0); 646 mutex_enter(&mountlist_lock); 647 if (error) { 648 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 649 (void) vfs_allocate_syncvnode(mp); 650 mp->mnt_iflag &= ~IMNT_UNMOUNT; 651 mp->mnt_unmounter = NULL; 652 mp->mnt_flag |= async; 653 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 654 &mountlist_lock); 655 if (used_syncer) 656 mutex_exit(&syncer_mutex); 657 mutex_enter(&mp->mnt_mutex); 658 while (mp->mnt_wcnt > 0) { 659 wakeup(mp); 660 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 661 0, &mp->mnt_mutex); 662 } 663 mutex_exit(&mp->mnt_mutex); 664 return (error); 665 } 666 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 667 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 668 coveredvp->v_mountedhere = NULL; 669 mp->mnt_op->vfs_refcount--; 670 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 671 panic("unmount: dangling vnode"); 672 mp->mnt_iflag |= IMNT_GONE; 673 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 674 if (coveredvp != NULLVP) 675 vrele(coveredvp); 676 mount_finispecific(mp); 677 if (used_syncer) 678 mutex_exit(&syncer_mutex); 679 mutex_enter(&mp->mnt_mutex); 680 while (mp->mnt_wcnt > 0) { 681 wakeup(mp); 682 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 683 } 684 mutex_exit(&mp->mnt_mutex); 685 vfs_hooks_unmount(mp); 686 mutex_destroy(&mp->mnt_mutex); 687 free(mp, M_MOUNT); 688 return (0); 689 } 690 691 /* 692 * Sync each mounted filesystem. 693 */ 694 #ifdef DEBUG 695 int syncprt = 0; 696 struct ctldebug debug0 = { "syncprt", &syncprt }; 697 #endif 698 699 /* ARGSUSED */ 700 int 701 sys_sync(struct lwp *l, void *v, register_t *retval) 702 { 703 struct mount *mp, *nmp; 704 int asyncflag; 705 706 if (l == NULL) 707 l = &lwp0; 708 709 mutex_enter(&mountlist_lock); 710 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 711 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 712 nmp = mp->mnt_list.cqe_prev; 713 continue; 714 } 715 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 716 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 717 asyncflag = mp->mnt_flag & MNT_ASYNC; 718 mp->mnt_flag &= ~MNT_ASYNC; 719 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 720 if (asyncflag) 721 mp->mnt_flag |= MNT_ASYNC; 722 vn_finished_write(mp, 0); 723 } 724 mutex_enter(&mountlist_lock); 725 nmp = mp->mnt_list.cqe_prev; 726 vfs_unbusy(mp); 727 728 } 729 mutex_exit(&mountlist_lock); 730 #ifdef DEBUG 731 if (syncprt) 732 vfs_bufstats(); 733 #endif /* DEBUG */ 734 return (0); 735 } 736 737 /* 738 * Change filesystem quotas. 739 */ 740 /* ARGSUSED */ 741 int 742 sys_quotactl(struct lwp *l, void *v, register_t *retval) 743 { 744 struct sys_quotactl_args /* { 745 syscallarg(const char *) path; 746 syscallarg(int) cmd; 747 syscallarg(int) uid; 748 syscallarg(void *) arg; 749 } */ *uap = v; 750 struct mount *mp; 751 int error; 752 struct nameidata nd; 753 754 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 755 if ((error = namei(&nd)) != 0) 756 return (error); 757 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 758 vrele(nd.ni_vp); 759 if (error) 760 return (error); 761 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 762 SCARG(uap, arg), l); 763 vn_finished_write(mp, 0); 764 return (error); 765 } 766 767 int 768 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 769 int root) 770 { 771 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 772 int error = 0; 773 774 /* 775 * If MNT_NOWAIT or MNT_LAZY is specified, do not 776 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 777 * overrides MNT_NOWAIT. 778 */ 779 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 780 (flags != MNT_WAIT && flags != 0)) { 781 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 782 goto done; 783 } 784 785 /* Get the filesystem stats now */ 786 memset(sp, 0, sizeof(*sp)); 787 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 788 return error; 789 } 790 791 if (cwdi->cwdi_rdir == NULL) 792 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 793 done: 794 if (cwdi->cwdi_rdir != NULL) { 795 size_t len; 796 char *bp; 797 char *path = PNBUF_GET(); 798 if (!path) 799 return ENOMEM; 800 801 bp = path + MAXPATHLEN; 802 *--bp = '\0'; 803 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 804 MAXPATHLEN / 2, 0, l); 805 if (error) { 806 PNBUF_PUT(path); 807 return error; 808 } 809 len = strlen(bp); 810 /* 811 * for mount points that are below our root, we can see 812 * them, so we fix up the pathname and return them. The 813 * rest we cannot see, so we don't allow viewing the 814 * data. 815 */ 816 if (strncmp(bp, sp->f_mntonname, len) == 0) { 817 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 818 sizeof(sp->f_mntonname)); 819 if (sp->f_mntonname[0] == '\0') 820 (void)strlcpy(sp->f_mntonname, "/", 821 sizeof(sp->f_mntonname)); 822 } else { 823 if (root) 824 (void)strlcpy(sp->f_mntonname, "/", 825 sizeof(sp->f_mntonname)); 826 else 827 error = EPERM; 828 } 829 PNBUF_PUT(path); 830 } 831 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 832 return error; 833 } 834 835 /* 836 * Get filesystem statistics. 837 */ 838 /* ARGSUSED */ 839 int 840 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 841 { 842 struct sys_statvfs1_args /* { 843 syscallarg(const char *) path; 844 syscallarg(struct statvfs *) buf; 845 syscallarg(int) flags; 846 } */ *uap = v; 847 struct mount *mp; 848 struct statvfs *sb; 849 int error; 850 struct nameidata nd; 851 852 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 853 if ((error = namei(&nd)) != 0) 854 return error; 855 mp = nd.ni_vp->v_mount; 856 vrele(nd.ni_vp); 857 sb = STATVFSBUF_GET(); 858 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 859 if (error == 0) { 860 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 861 } 862 STATVFSBUF_PUT(sb); 863 return error; 864 } 865 866 /* 867 * Get filesystem statistics. 868 */ 869 /* ARGSUSED */ 870 int 871 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 872 { 873 struct sys_fstatvfs1_args /* { 874 syscallarg(int) fd; 875 syscallarg(struct statvfs *) buf; 876 syscallarg(int) flags; 877 } */ *uap = v; 878 struct proc *p = l->l_proc; 879 struct file *fp; 880 struct mount *mp; 881 struct statvfs *sb; 882 int error; 883 884 /* getvnode() will use the descriptor for us */ 885 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 886 return (error); 887 mp = ((struct vnode *)fp->f_data)->v_mount; 888 sb = STATVFSBUF_GET(); 889 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 890 goto out; 891 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 892 out: 893 FILE_UNUSE(fp, l); 894 STATVFSBUF_PUT(sb); 895 return error; 896 } 897 898 899 /* 900 * Get statistics on all filesystems. 901 */ 902 int 903 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 904 { 905 struct sys_getvfsstat_args /* { 906 syscallarg(struct statvfs *) buf; 907 syscallarg(size_t) bufsize; 908 syscallarg(int) flags; 909 } */ *uap = v; 910 int root = 0; 911 struct proc *p = l->l_proc; 912 struct mount *mp, *nmp; 913 struct statvfs *sb; 914 struct statvfs *sfsp; 915 size_t count, maxcount; 916 int error = 0; 917 918 sb = STATVFSBUF_GET(); 919 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 920 sfsp = SCARG(uap, buf); 921 mutex_enter(&mountlist_lock); 922 count = 0; 923 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 924 mp = nmp) { 925 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 926 nmp = CIRCLEQ_NEXT(mp, mnt_list); 927 continue; 928 } 929 if (sfsp && count < maxcount) { 930 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 931 if (error) { 932 mutex_enter(&mountlist_lock); 933 nmp = CIRCLEQ_NEXT(mp, mnt_list); 934 vfs_unbusy(mp); 935 continue; 936 } 937 error = copyout(sb, sfsp, sizeof(*sfsp)); 938 if (error) { 939 vfs_unbusy(mp); 940 goto out; 941 } 942 sfsp++; 943 root |= strcmp(sb->f_mntonname, "/") == 0; 944 } 945 count++; 946 mutex_enter(&mountlist_lock); 947 nmp = CIRCLEQ_NEXT(mp, mnt_list); 948 vfs_unbusy(mp); 949 } 950 mutex_exit(&mountlist_lock); 951 if (root == 0 && p->p_cwdi->cwdi_rdir) { 952 /* 953 * fake a root entry 954 */ 955 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 956 SCARG(uap, flags), 1)) != 0) 957 goto out; 958 if (sfsp) 959 error = copyout(sb, sfsp, sizeof(*sfsp)); 960 count++; 961 } 962 if (sfsp && count > maxcount) 963 *retval = maxcount; 964 else 965 *retval = count; 966 out: 967 STATVFSBUF_PUT(sb); 968 return error; 969 } 970 971 /* 972 * Change current working directory to a given file descriptor. 973 */ 974 /* ARGSUSED */ 975 int 976 sys_fchdir(struct lwp *l, void *v, register_t *retval) 977 { 978 struct sys_fchdir_args /* { 979 syscallarg(int) fd; 980 } */ *uap = v; 981 struct proc *p = l->l_proc; 982 struct filedesc *fdp = p->p_fd; 983 struct cwdinfo *cwdi = p->p_cwdi; 984 struct vnode *vp, *tdp; 985 struct mount *mp; 986 struct file *fp; 987 int error; 988 989 /* getvnode() will use the descriptor for us */ 990 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 991 return (error); 992 vp = (struct vnode *)fp->f_data; 993 994 VREF(vp); 995 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 996 if (vp->v_type != VDIR) 997 error = ENOTDIR; 998 else 999 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1000 if (error) { 1001 vput(vp); 1002 goto out; 1003 } 1004 while ((mp = vp->v_mountedhere) != NULL) { 1005 if (vfs_busy(mp, 0, 0)) 1006 continue; 1007 1008 vput(vp); 1009 error = VFS_ROOT(mp, &tdp); 1010 vfs_unbusy(mp); 1011 if (error) 1012 goto out; 1013 vp = tdp; 1014 } 1015 VOP_UNLOCK(vp, 0); 1016 1017 /* 1018 * Disallow changing to a directory not under the process's 1019 * current root directory (if there is one). 1020 */ 1021 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1022 vrele(vp); 1023 error = EPERM; /* operation not permitted */ 1024 goto out; 1025 } 1026 1027 vrele(cwdi->cwdi_cdir); 1028 cwdi->cwdi_cdir = vp; 1029 out: 1030 FILE_UNUSE(fp, l); 1031 return (error); 1032 } 1033 1034 /* 1035 * Change this process's notion of the root directory to a given file 1036 * descriptor. 1037 */ 1038 int 1039 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1040 { 1041 struct sys_fchroot_args *uap = v; 1042 struct proc *p = l->l_proc; 1043 struct filedesc *fdp = p->p_fd; 1044 struct cwdinfo *cwdi = p->p_cwdi; 1045 struct vnode *vp; 1046 struct file *fp; 1047 int error; 1048 1049 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1050 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1051 return error; 1052 /* getvnode() will use the descriptor for us */ 1053 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1054 return error; 1055 vp = (struct vnode *) fp->f_data; 1056 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1057 if (vp->v_type != VDIR) 1058 error = ENOTDIR; 1059 else 1060 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1061 VOP_UNLOCK(vp, 0); 1062 if (error) 1063 goto out; 1064 VREF(vp); 1065 1066 /* 1067 * Prevent escaping from chroot by putting the root under 1068 * the working directory. Silently chdir to / if we aren't 1069 * already there. 1070 */ 1071 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1072 /* 1073 * XXX would be more failsafe to change directory to a 1074 * deadfs node here instead 1075 */ 1076 vrele(cwdi->cwdi_cdir); 1077 VREF(vp); 1078 cwdi->cwdi_cdir = vp; 1079 } 1080 1081 if (cwdi->cwdi_rdir != NULL) 1082 vrele(cwdi->cwdi_rdir); 1083 cwdi->cwdi_rdir = vp; 1084 out: 1085 FILE_UNUSE(fp, l); 1086 return (error); 1087 } 1088 1089 /* 1090 * Change current working directory (``.''). 1091 */ 1092 /* ARGSUSED */ 1093 int 1094 sys_chdir(struct lwp *l, void *v, register_t *retval) 1095 { 1096 struct sys_chdir_args /* { 1097 syscallarg(const char *) path; 1098 } */ *uap = v; 1099 struct proc *p = l->l_proc; 1100 struct cwdinfo *cwdi = p->p_cwdi; 1101 int error; 1102 struct nameidata nd; 1103 1104 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1105 SCARG(uap, path), l); 1106 if ((error = change_dir(&nd, l)) != 0) 1107 return (error); 1108 vrele(cwdi->cwdi_cdir); 1109 cwdi->cwdi_cdir = nd.ni_vp; 1110 return (0); 1111 } 1112 1113 /* 1114 * Change notion of root (``/'') directory. 1115 */ 1116 /* ARGSUSED */ 1117 int 1118 sys_chroot(struct lwp *l, void *v, register_t *retval) 1119 { 1120 struct sys_chroot_args /* { 1121 syscallarg(const char *) path; 1122 } */ *uap = v; 1123 struct proc *p = l->l_proc; 1124 struct cwdinfo *cwdi = p->p_cwdi; 1125 struct vnode *vp; 1126 int error; 1127 struct nameidata nd; 1128 1129 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1130 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1131 return (error); 1132 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1133 SCARG(uap, path), l); 1134 if ((error = change_dir(&nd, l)) != 0) 1135 return (error); 1136 if (cwdi->cwdi_rdir != NULL) 1137 vrele(cwdi->cwdi_rdir); 1138 vp = nd.ni_vp; 1139 cwdi->cwdi_rdir = vp; 1140 1141 /* 1142 * Prevent escaping from chroot by putting the root under 1143 * the working directory. Silently chdir to / if we aren't 1144 * already there. 1145 */ 1146 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1147 /* 1148 * XXX would be more failsafe to change directory to a 1149 * deadfs node here instead 1150 */ 1151 vrele(cwdi->cwdi_cdir); 1152 VREF(vp); 1153 cwdi->cwdi_cdir = vp; 1154 } 1155 1156 return (0); 1157 } 1158 1159 /* 1160 * Common routine for chroot and chdir. 1161 */ 1162 static int 1163 change_dir(struct nameidata *ndp, struct lwp *l) 1164 { 1165 struct vnode *vp; 1166 int error; 1167 1168 if ((error = namei(ndp)) != 0) 1169 return (error); 1170 vp = ndp->ni_vp; 1171 if (vp->v_type != VDIR) 1172 error = ENOTDIR; 1173 else 1174 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1175 1176 if (error) 1177 vput(vp); 1178 else 1179 VOP_UNLOCK(vp, 0); 1180 return (error); 1181 } 1182 1183 /* 1184 * Check permissions, allocate an open file structure, 1185 * and call the device open routine if any. 1186 */ 1187 int 1188 sys_open(struct lwp *l, void *v, register_t *retval) 1189 { 1190 struct sys_open_args /* { 1191 syscallarg(const char *) path; 1192 syscallarg(int) flags; 1193 syscallarg(int) mode; 1194 } */ *uap = v; 1195 struct proc *p = l->l_proc; 1196 struct cwdinfo *cwdi = p->p_cwdi; 1197 struct filedesc *fdp = p->p_fd; 1198 struct file *fp; 1199 struct vnode *vp; 1200 int flags, cmode; 1201 int type, indx, error; 1202 struct flock lf; 1203 struct nameidata nd; 1204 1205 flags = FFLAGS(SCARG(uap, flags)); 1206 if ((flags & (FREAD | FWRITE)) == 0) 1207 return (EINVAL); 1208 /* falloc() will use the file descriptor for us */ 1209 if ((error = falloc(l, &fp, &indx)) != 0) 1210 return (error); 1211 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1212 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1213 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1214 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1215 FILE_UNUSE(fp, l); 1216 fdp->fd_ofiles[indx] = NULL; 1217 ffree(fp); 1218 if ((error == EDUPFD || error == EMOVEFD) && 1219 l->l_dupfd >= 0 && /* XXX from fdopen */ 1220 (error = 1221 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1222 *retval = indx; 1223 return (0); 1224 } 1225 if (error == ERESTART) 1226 error = EINTR; 1227 fdremove(fdp, indx); 1228 return (error); 1229 } 1230 l->l_dupfd = 0; 1231 vp = nd.ni_vp; 1232 fp->f_flag = flags & FMASK; 1233 fp->f_type = DTYPE_VNODE; 1234 fp->f_ops = &vnops; 1235 fp->f_data = vp; 1236 if (flags & (O_EXLOCK | O_SHLOCK)) { 1237 lf.l_whence = SEEK_SET; 1238 lf.l_start = 0; 1239 lf.l_len = 0; 1240 if (flags & O_EXLOCK) 1241 lf.l_type = F_WRLCK; 1242 else 1243 lf.l_type = F_RDLCK; 1244 type = F_FLOCK; 1245 if ((flags & FNONBLOCK) == 0) 1246 type |= F_WAIT; 1247 VOP_UNLOCK(vp, 0); 1248 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1249 if (error) { 1250 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1251 FILE_UNUSE(fp, l); 1252 ffree(fp); 1253 fdremove(fdp, indx); 1254 return (error); 1255 } 1256 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1257 fp->f_flag |= FHASLOCK; 1258 } 1259 VOP_UNLOCK(vp, 0); 1260 *retval = indx; 1261 FILE_SET_MATURE(fp); 1262 FILE_UNUSE(fp, l); 1263 return (0); 1264 } 1265 1266 static void 1267 vfs__fhfree(fhandle_t *fhp) 1268 { 1269 size_t fhsize; 1270 1271 if (fhp == NULL) { 1272 return; 1273 } 1274 fhsize = FHANDLE_SIZE(fhp); 1275 kmem_free(fhp, fhsize); 1276 } 1277 1278 /* 1279 * vfs_composefh: compose a filehandle. 1280 */ 1281 1282 int 1283 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1284 { 1285 struct mount *mp; 1286 struct fid *fidp; 1287 int error; 1288 size_t needfhsize; 1289 size_t fidsize; 1290 1291 mp = vp->v_mount; 1292 fidp = NULL; 1293 if (*fh_size < FHANDLE_SIZE_MIN) { 1294 fidsize = 0; 1295 } else { 1296 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1297 if (fhp != NULL) { 1298 memset(fhp, 0, *fh_size); 1299 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1300 fidp = &fhp->fh_fid; 1301 } 1302 } 1303 error = VFS_VPTOFH(vp, fidp, &fidsize); 1304 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1305 if (error == 0 && *fh_size < needfhsize) { 1306 error = E2BIG; 1307 } 1308 *fh_size = needfhsize; 1309 return error; 1310 } 1311 1312 int 1313 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1314 { 1315 struct mount *mp; 1316 fhandle_t *fhp; 1317 size_t fhsize; 1318 size_t fidsize; 1319 int error; 1320 1321 *fhpp = NULL; 1322 mp = vp->v_mount; 1323 fidsize = 0; 1324 error = VFS_VPTOFH(vp, NULL, &fidsize); 1325 KASSERT(error != 0); 1326 if (error != E2BIG) { 1327 goto out; 1328 } 1329 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1330 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1331 if (fhp == NULL) { 1332 error = ENOMEM; 1333 goto out; 1334 } 1335 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1336 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1337 if (error == 0) { 1338 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1339 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1340 *fhpp = fhp; 1341 } else { 1342 kmem_free(fhp, fhsize); 1343 } 1344 out: 1345 return error; 1346 } 1347 1348 void 1349 vfs_composefh_free(fhandle_t *fhp) 1350 { 1351 1352 vfs__fhfree(fhp); 1353 } 1354 1355 /* 1356 * vfs_fhtovp: lookup a vnode by a filehandle. 1357 */ 1358 1359 int 1360 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1361 { 1362 struct mount *mp; 1363 int error; 1364 1365 *vpp = NULL; 1366 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1367 if (mp == NULL) { 1368 error = ESTALE; 1369 goto out; 1370 } 1371 if (mp->mnt_op->vfs_fhtovp == NULL) { 1372 error = EOPNOTSUPP; 1373 goto out; 1374 } 1375 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1376 out: 1377 return error; 1378 } 1379 1380 /* 1381 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1382 * the needed size. 1383 */ 1384 1385 int 1386 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1387 { 1388 fhandle_t *fhp; 1389 int error; 1390 1391 *fhpp = NULL; 1392 if (fhsize > FHANDLE_SIZE_MAX) { 1393 return EINVAL; 1394 } 1395 if (fhsize < FHANDLE_SIZE_MIN) { 1396 return EINVAL; 1397 } 1398 again: 1399 fhp = kmem_alloc(fhsize, KM_SLEEP); 1400 if (fhp == NULL) { 1401 return ENOMEM; 1402 } 1403 error = copyin(ufhp, fhp, fhsize); 1404 if (error == 0) { 1405 /* XXX this check shouldn't be here */ 1406 if (FHANDLE_SIZE(fhp) == fhsize) { 1407 *fhpp = fhp; 1408 return 0; 1409 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1410 /* 1411 * a kludge for nfsv2 padded handles. 1412 */ 1413 size_t sz; 1414 1415 sz = FHANDLE_SIZE(fhp); 1416 kmem_free(fhp, fhsize); 1417 fhsize = sz; 1418 goto again; 1419 } else { 1420 /* 1421 * userland told us wrong size. 1422 */ 1423 error = EINVAL; 1424 } 1425 } 1426 kmem_free(fhp, fhsize); 1427 return error; 1428 } 1429 1430 void 1431 vfs_copyinfh_free(fhandle_t *fhp) 1432 { 1433 1434 vfs__fhfree(fhp); 1435 } 1436 1437 /* 1438 * Get file handle system call 1439 */ 1440 int 1441 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1442 { 1443 struct sys___getfh30_args /* { 1444 syscallarg(char *) fname; 1445 syscallarg(fhandle_t *) fhp; 1446 syscallarg(size_t *) fh_size; 1447 } */ *uap = v; 1448 struct vnode *vp; 1449 fhandle_t *fh; 1450 int error; 1451 struct nameidata nd; 1452 size_t sz; 1453 size_t usz; 1454 1455 /* 1456 * Must be super user 1457 */ 1458 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1459 0, NULL, NULL, NULL); 1460 if (error) 1461 return (error); 1462 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1463 SCARG(uap, fname), l); 1464 error = namei(&nd); 1465 if (error) 1466 return (error); 1467 vp = nd.ni_vp; 1468 error = vfs_composefh_alloc(vp, &fh); 1469 vput(vp); 1470 if (error != 0) { 1471 goto out; 1472 } 1473 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1474 if (error != 0) { 1475 goto out; 1476 } 1477 sz = FHANDLE_SIZE(fh); 1478 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1479 if (error != 0) { 1480 goto out; 1481 } 1482 if (usz >= sz) { 1483 error = copyout(fh, SCARG(uap, fhp), sz); 1484 } else { 1485 error = E2BIG; 1486 } 1487 out: 1488 vfs_composefh_free(fh); 1489 return (error); 1490 } 1491 1492 /* 1493 * Open a file given a file handle. 1494 * 1495 * Check permissions, allocate an open file structure, 1496 * and call the device open routine if any. 1497 */ 1498 1499 int 1500 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1501 register_t *retval) 1502 { 1503 struct filedesc *fdp = l->l_proc->p_fd; 1504 struct file *fp; 1505 struct vnode *vp = NULL; 1506 struct mount *mp; 1507 kauth_cred_t cred = l->l_cred; 1508 struct file *nfp; 1509 int type, indx, error=0; 1510 struct flock lf; 1511 struct vattr va; 1512 fhandle_t *fh; 1513 int flags; 1514 1515 /* 1516 * Must be super user 1517 */ 1518 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1519 0, NULL, NULL, NULL))) 1520 return (error); 1521 1522 flags = FFLAGS(oflags); 1523 if ((flags & (FREAD | FWRITE)) == 0) 1524 return (EINVAL); 1525 if ((flags & O_CREAT)) 1526 return (EINVAL); 1527 /* falloc() will use the file descriptor for us */ 1528 if ((error = falloc(l, &nfp, &indx)) != 0) 1529 return (error); 1530 fp = nfp; 1531 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1532 if (error != 0) { 1533 goto bad; 1534 } 1535 error = vfs_fhtovp(fh, &vp); 1536 if (error != 0) { 1537 goto bad; 1538 } 1539 1540 /* Now do an effective vn_open */ 1541 1542 if (vp->v_type == VSOCK) { 1543 error = EOPNOTSUPP; 1544 goto bad; 1545 } 1546 if (flags & FREAD) { 1547 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1548 goto bad; 1549 } 1550 if (flags & (FWRITE | O_TRUNC)) { 1551 if (vp->v_type == VDIR) { 1552 error = EISDIR; 1553 goto bad; 1554 } 1555 if ((error = vn_writechk(vp)) != 0 || 1556 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1557 goto bad; 1558 } 1559 if (flags & O_TRUNC) { 1560 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1561 goto bad; 1562 VOP_UNLOCK(vp, 0); /* XXX */ 1563 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1564 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1565 VATTR_NULL(&va); 1566 va.va_size = 0; 1567 error = VOP_SETATTR(vp, &va, cred, l); 1568 vn_finished_write(mp, 0); 1569 if (error) 1570 goto bad; 1571 } 1572 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1573 goto bad; 1574 if (vp->v_type == VREG && 1575 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1576 error = EIO; 1577 goto bad; 1578 } 1579 if (flags & FWRITE) 1580 vp->v_writecount++; 1581 1582 /* done with modified vn_open, now finish what sys_open does. */ 1583 1584 fp->f_flag = flags & FMASK; 1585 fp->f_type = DTYPE_VNODE; 1586 fp->f_ops = &vnops; 1587 fp->f_data = vp; 1588 if (flags & (O_EXLOCK | O_SHLOCK)) { 1589 lf.l_whence = SEEK_SET; 1590 lf.l_start = 0; 1591 lf.l_len = 0; 1592 if (flags & O_EXLOCK) 1593 lf.l_type = F_WRLCK; 1594 else 1595 lf.l_type = F_RDLCK; 1596 type = F_FLOCK; 1597 if ((flags & FNONBLOCK) == 0) 1598 type |= F_WAIT; 1599 VOP_UNLOCK(vp, 0); 1600 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1601 if (error) { 1602 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1603 FILE_UNUSE(fp, l); 1604 ffree(fp); 1605 fdremove(fdp, indx); 1606 return (error); 1607 } 1608 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1609 fp->f_flag |= FHASLOCK; 1610 } 1611 VOP_UNLOCK(vp, 0); 1612 *retval = indx; 1613 FILE_SET_MATURE(fp); 1614 FILE_UNUSE(fp, l); 1615 vfs_copyinfh_free(fh); 1616 return (0); 1617 1618 bad: 1619 FILE_UNUSE(fp, l); 1620 ffree(fp); 1621 fdremove(fdp, indx); 1622 if (vp != NULL) 1623 vput(vp); 1624 vfs_copyinfh_free(fh); 1625 return (error); 1626 } 1627 1628 int 1629 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1630 { 1631 struct sys___fhopen40_args /* { 1632 syscallarg(const void *) fhp; 1633 syscallarg(size_t) fh_size; 1634 syscallarg(int) flags; 1635 } */ *uap = v; 1636 1637 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1638 SCARG(uap, flags), retval); 1639 } 1640 1641 /* XXX: temp (mar '07) for LKM compat */ 1642 int 1643 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1644 register_t *retval) 1645 { 1646 int error; 1647 struct stat sb; 1648 1649 error = do_fhstat(l, ufhp, fhsize, &sb); 1650 if (error == 0) 1651 error = copyout(&sb, sbp, sizeof(sb)); 1652 return error; 1653 } 1654 1655 int 1656 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1657 { 1658 int error; 1659 fhandle_t *fh; 1660 struct vnode *vp; 1661 1662 /* 1663 * Must be super user 1664 */ 1665 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1666 0, NULL, NULL, NULL))) 1667 return (error); 1668 1669 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1670 if (error != 0) 1671 return error; 1672 1673 error = vfs_fhtovp(fh, &vp); 1674 vfs_copyinfh_free(fh); 1675 if (error != 0) 1676 return error; 1677 1678 error = vn_stat(vp, sb, l); 1679 vput(vp); 1680 return error; 1681 } 1682 1683 1684 /* ARGSUSED */ 1685 int 1686 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1687 { 1688 struct sys___fhstat40_args /* { 1689 syscallarg(const void *) fhp; 1690 syscallarg(size_t) fh_size; 1691 syscallarg(struct stat *) sb; 1692 } */ *uap = v; 1693 struct stat sb; 1694 int error; 1695 1696 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1697 if (error) 1698 return error; 1699 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1700 } 1701 1702 /* XXX: temp (mar '07) for LKM compat */ 1703 int 1704 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1705 int flags, register_t *retval) 1706 { 1707 struct statvfs *sb = STATVFSBUF_GET(); 1708 int error; 1709 1710 error = do_fhstatvfs(l, ufhp, fhsize, sb, flags); 1711 if (error == 0) 1712 error = copyout(sb, buf, sizeof(*sb)); 1713 STATVFSBUF_PUT(sb); 1714 return error; 1715 } 1716 1717 int 1718 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1719 int flags) 1720 { 1721 fhandle_t *fh; 1722 struct mount *mp; 1723 struct vnode *vp; 1724 int error; 1725 1726 /* 1727 * Must be super user 1728 */ 1729 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1730 0, NULL, NULL, NULL))) 1731 return error; 1732 1733 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1734 if (error != 0) 1735 return error; 1736 1737 error = vfs_fhtovp(fh, &vp); 1738 vfs_copyinfh_free(fh); 1739 if (error != 0) 1740 return error; 1741 1742 mp = vp->v_mount; 1743 error = dostatvfs(mp, sb, l, flags, 1); 1744 vput(vp); 1745 return error; 1746 } 1747 1748 /* ARGSUSED */ 1749 int 1750 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1751 { 1752 struct sys___fhstatvfs140_args /* { 1753 syscallarg(const void *) fhp; 1754 syscallarg(size_t) fh_size; 1755 syscallarg(struct statvfs *) buf; 1756 syscallarg(int) flags; 1757 } */ *uap = v; 1758 struct statvfs *sb = STATVFSBUF_GET(); 1759 int error; 1760 1761 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1762 SCARG(uap, flags)); 1763 if (error == 0) 1764 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1765 STATVFSBUF_PUT(sb); 1766 return error; 1767 } 1768 1769 /* 1770 * Create a special file. 1771 */ 1772 /* ARGSUSED */ 1773 int 1774 sys_mknod(struct lwp *l, void *v, register_t *retval) 1775 { 1776 struct sys_mknod_args /* { 1777 syscallarg(const char *) path; 1778 syscallarg(int) mode; 1779 syscallarg(int) dev; 1780 } */ *uap = v; 1781 struct proc *p = l->l_proc; 1782 struct vnode *vp; 1783 struct mount *mp; 1784 struct vattr vattr; 1785 int error, optype; 1786 struct nameidata nd; 1787 1788 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1789 0, NULL, NULL, NULL)) != 0) 1790 return (error); 1791 1792 optype = VOP_MKNOD_DESCOFFSET; 1793 restart: 1794 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1795 if ((error = namei(&nd)) != 0) 1796 return (error); 1797 vp = nd.ni_vp; 1798 if (vp != NULL) 1799 error = EEXIST; 1800 else { 1801 VATTR_NULL(&vattr); 1802 vattr.va_mode = 1803 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1804 vattr.va_rdev = SCARG(uap, dev); 1805 1806 switch (SCARG(uap, mode) & S_IFMT) { 1807 case S_IFMT: /* used by badsect to flag bad sectors */ 1808 vattr.va_type = VBAD; 1809 break; 1810 case S_IFCHR: 1811 vattr.va_type = VCHR; 1812 break; 1813 case S_IFBLK: 1814 vattr.va_type = VBLK; 1815 break; 1816 case S_IFWHT: 1817 optype = VOP_WHITEOUT_DESCOFFSET; 1818 break; 1819 case S_IFREG: 1820 vattr.va_type = VREG; 1821 vattr.va_rdev = VNOVAL; 1822 optype = VOP_CREATE_DESCOFFSET; 1823 break; 1824 default: 1825 error = EINVAL; 1826 break; 1827 } 1828 } 1829 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1830 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1831 if (nd.ni_dvp == vp) 1832 vrele(nd.ni_dvp); 1833 else 1834 vput(nd.ni_dvp); 1835 if (vp) 1836 vrele(vp); 1837 if ((error = vn_start_write(NULL, &mp, 1838 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1839 return (error); 1840 goto restart; 1841 } 1842 if (!error) { 1843 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1844 switch (optype) { 1845 case VOP_WHITEOUT_DESCOFFSET: 1846 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1847 if (error) 1848 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1849 vput(nd.ni_dvp); 1850 break; 1851 1852 case VOP_MKNOD_DESCOFFSET: 1853 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1854 &nd.ni_cnd, &vattr); 1855 if (error == 0) 1856 vput(nd.ni_vp); 1857 break; 1858 1859 case VOP_CREATE_DESCOFFSET: 1860 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1861 &nd.ni_cnd, &vattr); 1862 if (error == 0) 1863 vput(nd.ni_vp); 1864 break; 1865 } 1866 } else { 1867 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1868 if (nd.ni_dvp == vp) 1869 vrele(nd.ni_dvp); 1870 else 1871 vput(nd.ni_dvp); 1872 if (vp) 1873 vrele(vp); 1874 } 1875 vn_finished_write(mp, 0); 1876 return (error); 1877 } 1878 1879 /* 1880 * Create a named pipe. 1881 */ 1882 /* ARGSUSED */ 1883 int 1884 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1885 { 1886 struct sys_mkfifo_args /* { 1887 syscallarg(const char *) path; 1888 syscallarg(int) mode; 1889 } */ *uap = v; 1890 struct proc *p = l->l_proc; 1891 struct mount *mp; 1892 struct vattr vattr; 1893 int error; 1894 struct nameidata nd; 1895 1896 restart: 1897 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1898 if ((error = namei(&nd)) != 0) 1899 return (error); 1900 if (nd.ni_vp != NULL) { 1901 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1902 if (nd.ni_dvp == nd.ni_vp) 1903 vrele(nd.ni_dvp); 1904 else 1905 vput(nd.ni_dvp); 1906 vrele(nd.ni_vp); 1907 return (EEXIST); 1908 } 1909 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1910 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1911 if (nd.ni_dvp == nd.ni_vp) 1912 vrele(nd.ni_dvp); 1913 else 1914 vput(nd.ni_dvp); 1915 if (nd.ni_vp) 1916 vrele(nd.ni_vp); 1917 if ((error = vn_start_write(NULL, &mp, 1918 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1919 return (error); 1920 goto restart; 1921 } 1922 VATTR_NULL(&vattr); 1923 vattr.va_type = VFIFO; 1924 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1925 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1926 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1927 if (error == 0) 1928 vput(nd.ni_vp); 1929 vn_finished_write(mp, 0); 1930 return (error); 1931 } 1932 1933 /* 1934 * Make a hard file link. 1935 */ 1936 /* ARGSUSED */ 1937 int 1938 sys_link(struct lwp *l, void *v, register_t *retval) 1939 { 1940 struct sys_link_args /* { 1941 syscallarg(const char *) path; 1942 syscallarg(const char *) link; 1943 } */ *uap = v; 1944 struct vnode *vp; 1945 struct mount *mp; 1946 struct nameidata nd; 1947 int error; 1948 1949 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1950 if ((error = namei(&nd)) != 0) 1951 return (error); 1952 vp = nd.ni_vp; 1953 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1954 vrele(vp); 1955 return (error); 1956 } 1957 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1958 if ((error = namei(&nd)) != 0) 1959 goto out; 1960 if (nd.ni_vp) { 1961 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1962 if (nd.ni_dvp == nd.ni_vp) 1963 vrele(nd.ni_dvp); 1964 else 1965 vput(nd.ni_dvp); 1966 vrele(nd.ni_vp); 1967 error = EEXIST; 1968 goto out; 1969 } 1970 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1971 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1972 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1973 out: 1974 vrele(vp); 1975 vn_finished_write(mp, 0); 1976 return (error); 1977 } 1978 1979 /* 1980 * Make a symbolic link. 1981 */ 1982 /* ARGSUSED */ 1983 int 1984 sys_symlink(struct lwp *l, void *v, register_t *retval) 1985 { 1986 struct sys_symlink_args /* { 1987 syscallarg(const char *) path; 1988 syscallarg(const char *) link; 1989 } */ *uap = v; 1990 struct proc *p = l->l_proc; 1991 struct mount *mp; 1992 struct vattr vattr; 1993 char *path; 1994 int error; 1995 struct nameidata nd; 1996 1997 path = PNBUF_GET(); 1998 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1999 if (error) 2000 goto out; 2001 restart: 2002 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 2003 if ((error = namei(&nd)) != 0) 2004 goto out; 2005 if (nd.ni_vp) { 2006 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2007 if (nd.ni_dvp == nd.ni_vp) 2008 vrele(nd.ni_dvp); 2009 else 2010 vput(nd.ni_dvp); 2011 vrele(nd.ni_vp); 2012 error = EEXIST; 2013 goto out; 2014 } 2015 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2016 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2017 if (nd.ni_dvp == nd.ni_vp) 2018 vrele(nd.ni_dvp); 2019 else 2020 vput(nd.ni_dvp); 2021 if ((error = vn_start_write(NULL, &mp, 2022 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2023 return (error); 2024 goto restart; 2025 } 2026 VATTR_NULL(&vattr); 2027 vattr.va_type = VLNK; 2028 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2029 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2030 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2031 if (error == 0) 2032 vput(nd.ni_vp); 2033 vn_finished_write(mp, 0); 2034 out: 2035 PNBUF_PUT(path); 2036 return (error); 2037 } 2038 2039 /* 2040 * Delete a whiteout from the filesystem. 2041 */ 2042 /* ARGSUSED */ 2043 int 2044 sys_undelete(struct lwp *l, void *v, register_t *retval) 2045 { 2046 struct sys_undelete_args /* { 2047 syscallarg(const char *) path; 2048 } */ *uap = v; 2049 int error; 2050 struct mount *mp; 2051 struct nameidata nd; 2052 2053 restart: 2054 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2055 SCARG(uap, path), l); 2056 error = namei(&nd); 2057 if (error) 2058 return (error); 2059 2060 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2061 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2062 if (nd.ni_dvp == nd.ni_vp) 2063 vrele(nd.ni_dvp); 2064 else 2065 vput(nd.ni_dvp); 2066 if (nd.ni_vp) 2067 vrele(nd.ni_vp); 2068 return (EEXIST); 2069 } 2070 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2071 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2072 if (nd.ni_dvp == nd.ni_vp) 2073 vrele(nd.ni_dvp); 2074 else 2075 vput(nd.ni_dvp); 2076 if ((error = vn_start_write(NULL, &mp, 2077 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2078 return (error); 2079 goto restart; 2080 } 2081 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2082 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2083 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2084 vput(nd.ni_dvp); 2085 vn_finished_write(mp, 0); 2086 return (error); 2087 } 2088 2089 /* 2090 * Delete a name from the filesystem. 2091 */ 2092 /* ARGSUSED */ 2093 int 2094 sys_unlink(struct lwp *l, void *v, register_t *retval) 2095 { 2096 struct sys_unlink_args /* { 2097 syscallarg(const char *) path; 2098 } */ *uap = v; 2099 struct mount *mp; 2100 struct vnode *vp; 2101 int error; 2102 struct nameidata nd; 2103 #if NVERIEXEC > 0 2104 pathname_t pathbuf = NULL; 2105 #endif /* NVERIEXEC > 0 */ 2106 2107 restart: 2108 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2109 SCARG(uap, path), l); 2110 if ((error = namei(&nd)) != 0) 2111 return (error); 2112 vp = nd.ni_vp; 2113 2114 /* 2115 * The root of a mounted filesystem cannot be deleted. 2116 */ 2117 if (vp->v_flag & VROOT) { 2118 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2119 if (nd.ni_dvp == vp) 2120 vrele(nd.ni_dvp); 2121 else 2122 vput(nd.ni_dvp); 2123 vput(vp); 2124 error = EBUSY; 2125 goto out; 2126 } 2127 2128 #if NVERIEXEC > 0 2129 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2130 2131 /* Handle remove requests for veriexec entries. */ 2132 if (!error) { 2133 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2134 pathname_put(pathbuf); 2135 } 2136 2137 if (error) { 2138 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2139 if (nd.ni_dvp == vp) 2140 vrele(nd.ni_dvp); 2141 else 2142 vput(nd.ni_dvp); 2143 vput(vp); 2144 goto out; 2145 } 2146 #endif /* NVERIEXEC > 0 */ 2147 2148 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2149 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2150 if (nd.ni_dvp == vp) 2151 vrele(nd.ni_dvp); 2152 else 2153 vput(nd.ni_dvp); 2154 vput(vp); 2155 if ((error = vn_start_write(NULL, &mp, 2156 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2157 return (error); 2158 goto restart; 2159 } 2160 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2161 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2162 #ifdef FILEASSOC 2163 (void)fileassoc_file_delete(vp); 2164 #endif /* FILEASSOC */ 2165 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2166 vn_finished_write(mp, 0); 2167 out: 2168 return (error); 2169 } 2170 2171 /* 2172 * Reposition read/write file offset. 2173 */ 2174 int 2175 sys_lseek(struct lwp *l, void *v, register_t *retval) 2176 { 2177 struct sys_lseek_args /* { 2178 syscallarg(int) fd; 2179 syscallarg(int) pad; 2180 syscallarg(off_t) offset; 2181 syscallarg(int) whence; 2182 } */ *uap = v; 2183 struct proc *p = l->l_proc; 2184 kauth_cred_t cred = l->l_cred; 2185 struct filedesc *fdp = p->p_fd; 2186 struct file *fp; 2187 struct vnode *vp; 2188 struct vattr vattr; 2189 off_t newoff; 2190 int error; 2191 2192 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2193 return (EBADF); 2194 2195 FILE_USE(fp); 2196 2197 vp = (struct vnode *)fp->f_data; 2198 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2199 error = ESPIPE; 2200 goto out; 2201 } 2202 2203 switch (SCARG(uap, whence)) { 2204 case SEEK_CUR: 2205 newoff = fp->f_offset + SCARG(uap, offset); 2206 break; 2207 case SEEK_END: 2208 error = VOP_GETATTR(vp, &vattr, cred, l); 2209 if (error) 2210 goto out; 2211 newoff = SCARG(uap, offset) + vattr.va_size; 2212 break; 2213 case SEEK_SET: 2214 newoff = SCARG(uap, offset); 2215 break; 2216 default: 2217 error = EINVAL; 2218 goto out; 2219 } 2220 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 2221 goto out; 2222 2223 *(off_t *)retval = fp->f_offset = newoff; 2224 out: 2225 FILE_UNUSE(fp, l); 2226 return (error); 2227 } 2228 2229 /* 2230 * Positional read system call. 2231 */ 2232 int 2233 sys_pread(struct lwp *l, void *v, register_t *retval) 2234 { 2235 struct sys_pread_args /* { 2236 syscallarg(int) fd; 2237 syscallarg(void *) buf; 2238 syscallarg(size_t) nbyte; 2239 syscallarg(off_t) offset; 2240 } */ *uap = v; 2241 struct proc *p = l->l_proc; 2242 struct filedesc *fdp = p->p_fd; 2243 struct file *fp; 2244 struct vnode *vp; 2245 off_t offset; 2246 int error, fd = SCARG(uap, fd); 2247 2248 if ((fp = fd_getfile(fdp, fd)) == NULL) 2249 return (EBADF); 2250 2251 if ((fp->f_flag & FREAD) == 0) { 2252 simple_unlock(&fp->f_slock); 2253 return (EBADF); 2254 } 2255 2256 FILE_USE(fp); 2257 2258 vp = (struct vnode *)fp->f_data; 2259 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2260 error = ESPIPE; 2261 goto out; 2262 } 2263 2264 offset = SCARG(uap, offset); 2265 2266 /* 2267 * XXX This works because no file systems actually 2268 * XXX take any action on the seek operation. 2269 */ 2270 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2271 goto out; 2272 2273 /* dofileread() will unuse the descriptor for us */ 2274 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2275 &offset, 0, retval)); 2276 2277 out: 2278 FILE_UNUSE(fp, l); 2279 return (error); 2280 } 2281 2282 /* 2283 * Positional scatter read system call. 2284 */ 2285 int 2286 sys_preadv(struct lwp *l, void *v, register_t *retval) 2287 { 2288 struct sys_preadv_args /* { 2289 syscallarg(int) fd; 2290 syscallarg(const struct iovec *) iovp; 2291 syscallarg(int) iovcnt; 2292 syscallarg(off_t) offset; 2293 } */ *uap = v; 2294 struct proc *p = l->l_proc; 2295 struct filedesc *fdp = p->p_fd; 2296 struct file *fp; 2297 struct vnode *vp; 2298 off_t offset; 2299 int error, fd = SCARG(uap, fd); 2300 2301 if ((fp = fd_getfile(fdp, fd)) == NULL) 2302 return (EBADF); 2303 2304 if ((fp->f_flag & FREAD) == 0) { 2305 simple_unlock(&fp->f_slock); 2306 return (EBADF); 2307 } 2308 2309 FILE_USE(fp); 2310 2311 vp = (struct vnode *)fp->f_data; 2312 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2313 error = ESPIPE; 2314 goto out; 2315 } 2316 2317 offset = SCARG(uap, offset); 2318 2319 /* 2320 * XXX This works because no file systems actually 2321 * XXX take any action on the seek operation. 2322 */ 2323 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2324 goto out; 2325 2326 /* dofilereadv() will unuse the descriptor for us */ 2327 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2328 &offset, 0, retval)); 2329 2330 out: 2331 FILE_UNUSE(fp, l); 2332 return (error); 2333 } 2334 2335 /* 2336 * Positional write system call. 2337 */ 2338 int 2339 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2340 { 2341 struct sys_pwrite_args /* { 2342 syscallarg(int) fd; 2343 syscallarg(const void *) buf; 2344 syscallarg(size_t) nbyte; 2345 syscallarg(off_t) offset; 2346 } */ *uap = v; 2347 struct proc *p = l->l_proc; 2348 struct filedesc *fdp = p->p_fd; 2349 struct file *fp; 2350 struct vnode *vp; 2351 off_t offset; 2352 int error, fd = SCARG(uap, fd); 2353 2354 if ((fp = fd_getfile(fdp, fd)) == NULL) 2355 return (EBADF); 2356 2357 if ((fp->f_flag & FWRITE) == 0) { 2358 simple_unlock(&fp->f_slock); 2359 return (EBADF); 2360 } 2361 2362 FILE_USE(fp); 2363 2364 vp = (struct vnode *)fp->f_data; 2365 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2366 error = ESPIPE; 2367 goto out; 2368 } 2369 2370 offset = SCARG(uap, offset); 2371 2372 /* 2373 * XXX This works because no file systems actually 2374 * XXX take any action on the seek operation. 2375 */ 2376 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2377 goto out; 2378 2379 /* dofilewrite() will unuse the descriptor for us */ 2380 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2381 &offset, 0, retval)); 2382 2383 out: 2384 FILE_UNUSE(fp, l); 2385 return (error); 2386 } 2387 2388 /* 2389 * Positional gather write system call. 2390 */ 2391 int 2392 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2393 { 2394 struct sys_pwritev_args /* { 2395 syscallarg(int) fd; 2396 syscallarg(const struct iovec *) iovp; 2397 syscallarg(int) iovcnt; 2398 syscallarg(off_t) offset; 2399 } */ *uap = v; 2400 struct proc *p = l->l_proc; 2401 struct filedesc *fdp = p->p_fd; 2402 struct file *fp; 2403 struct vnode *vp; 2404 off_t offset; 2405 int error, fd = SCARG(uap, fd); 2406 2407 if ((fp = fd_getfile(fdp, fd)) == NULL) 2408 return (EBADF); 2409 2410 if ((fp->f_flag & FWRITE) == 0) { 2411 simple_unlock(&fp->f_slock); 2412 return (EBADF); 2413 } 2414 2415 FILE_USE(fp); 2416 2417 vp = (struct vnode *)fp->f_data; 2418 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2419 error = ESPIPE; 2420 goto out; 2421 } 2422 2423 offset = SCARG(uap, offset); 2424 2425 /* 2426 * XXX This works because no file systems actually 2427 * XXX take any action on the seek operation. 2428 */ 2429 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2430 goto out; 2431 2432 /* dofilewritev() will unuse the descriptor for us */ 2433 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2434 &offset, 0, retval)); 2435 2436 out: 2437 FILE_UNUSE(fp, l); 2438 return (error); 2439 } 2440 2441 /* 2442 * Check access permissions. 2443 */ 2444 int 2445 sys_access(struct lwp *l, void *v, register_t *retval) 2446 { 2447 struct sys_access_args /* { 2448 syscallarg(const char *) path; 2449 syscallarg(int) flags; 2450 } */ *uap = v; 2451 kauth_cred_t cred; 2452 struct vnode *vp; 2453 int error, flags; 2454 struct nameidata nd; 2455 2456 cred = kauth_cred_dup(l->l_cred); 2457 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2458 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2459 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2460 SCARG(uap, path), l); 2461 /* Override default credentials */ 2462 nd.ni_cnd.cn_cred = cred; 2463 if ((error = namei(&nd)) != 0) 2464 goto out; 2465 vp = nd.ni_vp; 2466 2467 /* Flags == 0 means only check for existence. */ 2468 if (SCARG(uap, flags)) { 2469 flags = 0; 2470 if (SCARG(uap, flags) & R_OK) 2471 flags |= VREAD; 2472 if (SCARG(uap, flags) & W_OK) 2473 flags |= VWRITE; 2474 if (SCARG(uap, flags) & X_OK) 2475 flags |= VEXEC; 2476 2477 error = VOP_ACCESS(vp, flags, cred, l); 2478 if (!error && (flags & VWRITE)) 2479 error = vn_writechk(vp); 2480 } 2481 vput(vp); 2482 out: 2483 kauth_cred_free(cred); 2484 return (error); 2485 } 2486 2487 /* 2488 * Common code for all sys_stat functions, including compat versions. 2489 */ 2490 int 2491 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2492 struct stat *sb) 2493 { 2494 int error; 2495 struct nameidata nd; 2496 2497 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF, UIO_USERSPACE, path, l); 2498 error = namei(&nd); 2499 if (error != 0) 2500 return error; 2501 error = vn_stat(nd.ni_vp, sb, l); 2502 vput(nd.ni_vp); 2503 return error; 2504 } 2505 2506 /* 2507 * Get file status; this version follows links. 2508 */ 2509 /* ARGSUSED */ 2510 int 2511 sys___stat30(struct lwp *l, void *v, register_t *retval) 2512 { 2513 struct sys___stat30_args /* { 2514 syscallarg(const char *) path; 2515 syscallarg(struct stat *) ub; 2516 } */ *uap = v; 2517 struct stat sb; 2518 int error; 2519 2520 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2521 if (error) 2522 return error; 2523 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2524 } 2525 2526 /* 2527 * Get file status; this version does not follow links. 2528 */ 2529 /* ARGSUSED */ 2530 int 2531 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2532 { 2533 struct sys___lstat30_args /* { 2534 syscallarg(const char *) path; 2535 syscallarg(struct stat *) ub; 2536 } */ *uap = v; 2537 struct stat sb; 2538 int error; 2539 2540 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2541 if (error) 2542 return error; 2543 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2544 } 2545 2546 /* 2547 * Get configurable pathname variables. 2548 */ 2549 /* ARGSUSED */ 2550 int 2551 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2552 { 2553 struct sys_pathconf_args /* { 2554 syscallarg(const char *) path; 2555 syscallarg(int) name; 2556 } */ *uap = v; 2557 int error; 2558 struct nameidata nd; 2559 2560 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2561 SCARG(uap, path), l); 2562 if ((error = namei(&nd)) != 0) 2563 return (error); 2564 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2565 vput(nd.ni_vp); 2566 return (error); 2567 } 2568 2569 /* 2570 * Return target name of a symbolic link. 2571 */ 2572 /* ARGSUSED */ 2573 int 2574 sys_readlink(struct lwp *l, void *v, register_t *retval) 2575 { 2576 struct sys_readlink_args /* { 2577 syscallarg(const char *) path; 2578 syscallarg(char *) buf; 2579 syscallarg(size_t) count; 2580 } */ *uap = v; 2581 struct vnode *vp; 2582 struct iovec aiov; 2583 struct uio auio; 2584 int error; 2585 struct nameidata nd; 2586 2587 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2588 SCARG(uap, path), l); 2589 if ((error = namei(&nd)) != 0) 2590 return (error); 2591 vp = nd.ni_vp; 2592 if (vp->v_type != VLNK) 2593 error = EINVAL; 2594 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2595 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2596 aiov.iov_base = SCARG(uap, buf); 2597 aiov.iov_len = SCARG(uap, count); 2598 auio.uio_iov = &aiov; 2599 auio.uio_iovcnt = 1; 2600 auio.uio_offset = 0; 2601 auio.uio_rw = UIO_READ; 2602 KASSERT(l == curlwp); 2603 auio.uio_vmspace = l->l_proc->p_vmspace; 2604 auio.uio_resid = SCARG(uap, count); 2605 error = VOP_READLINK(vp, &auio, l->l_cred); 2606 } 2607 vput(vp); 2608 *retval = SCARG(uap, count) - auio.uio_resid; 2609 return (error); 2610 } 2611 2612 /* 2613 * Change flags of a file given a path name. 2614 */ 2615 /* ARGSUSED */ 2616 int 2617 sys_chflags(struct lwp *l, void *v, register_t *retval) 2618 { 2619 struct sys_chflags_args /* { 2620 syscallarg(const char *) path; 2621 syscallarg(u_long) flags; 2622 } */ *uap = v; 2623 struct vnode *vp; 2624 int error; 2625 struct nameidata nd; 2626 2627 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2628 if ((error = namei(&nd)) != 0) 2629 return (error); 2630 vp = nd.ni_vp; 2631 error = change_flags(vp, SCARG(uap, flags), l); 2632 vput(vp); 2633 return (error); 2634 } 2635 2636 /* 2637 * Change flags of a file given a file descriptor. 2638 */ 2639 /* ARGSUSED */ 2640 int 2641 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2642 { 2643 struct sys_fchflags_args /* { 2644 syscallarg(int) fd; 2645 syscallarg(u_long) flags; 2646 } */ *uap = v; 2647 struct proc *p = l->l_proc; 2648 struct vnode *vp; 2649 struct file *fp; 2650 int error; 2651 2652 /* getvnode() will use the descriptor for us */ 2653 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2654 return (error); 2655 vp = (struct vnode *)fp->f_data; 2656 error = change_flags(vp, SCARG(uap, flags), l); 2657 VOP_UNLOCK(vp, 0); 2658 FILE_UNUSE(fp, l); 2659 return (error); 2660 } 2661 2662 /* 2663 * Change flags of a file given a path name; this version does 2664 * not follow links. 2665 */ 2666 int 2667 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2668 { 2669 struct sys_lchflags_args /* { 2670 syscallarg(const char *) path; 2671 syscallarg(u_long) flags; 2672 } */ *uap = v; 2673 struct vnode *vp; 2674 int error; 2675 struct nameidata nd; 2676 2677 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2678 if ((error = namei(&nd)) != 0) 2679 return (error); 2680 vp = nd.ni_vp; 2681 error = change_flags(vp, SCARG(uap, flags), l); 2682 vput(vp); 2683 return (error); 2684 } 2685 2686 /* 2687 * Common routine to change flags of a file. 2688 */ 2689 int 2690 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2691 { 2692 struct mount *mp; 2693 struct vattr vattr; 2694 int error; 2695 2696 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2697 return (error); 2698 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2699 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2700 /* 2701 * Non-superusers cannot change the flags on devices, even if they 2702 * own them. 2703 */ 2704 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2705 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2706 goto out; 2707 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2708 error = EINVAL; 2709 goto out; 2710 } 2711 } 2712 VATTR_NULL(&vattr); 2713 vattr.va_flags = flags; 2714 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2715 out: 2716 vn_finished_write(mp, 0); 2717 return (error); 2718 } 2719 2720 /* 2721 * Change mode of a file given path name; this version follows links. 2722 */ 2723 /* ARGSUSED */ 2724 int 2725 sys_chmod(struct lwp *l, void *v, register_t *retval) 2726 { 2727 struct sys_chmod_args /* { 2728 syscallarg(const char *) path; 2729 syscallarg(int) mode; 2730 } */ *uap = v; 2731 int error; 2732 struct nameidata nd; 2733 2734 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2735 if ((error = namei(&nd)) != 0) 2736 return (error); 2737 2738 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2739 2740 vrele(nd.ni_vp); 2741 return (error); 2742 } 2743 2744 /* 2745 * Change mode of a file given a file descriptor. 2746 */ 2747 /* ARGSUSED */ 2748 int 2749 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2750 { 2751 struct sys_fchmod_args /* { 2752 syscallarg(int) fd; 2753 syscallarg(int) mode; 2754 } */ *uap = v; 2755 struct proc *p = l->l_proc; 2756 struct file *fp; 2757 int error; 2758 2759 /* getvnode() will use the descriptor for us */ 2760 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2761 return (error); 2762 2763 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2764 FILE_UNUSE(fp, l); 2765 return (error); 2766 } 2767 2768 /* 2769 * Change mode of a file given path name; this version does not follow links. 2770 */ 2771 /* ARGSUSED */ 2772 int 2773 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2774 { 2775 struct sys_lchmod_args /* { 2776 syscallarg(const char *) path; 2777 syscallarg(int) mode; 2778 } */ *uap = v; 2779 int error; 2780 struct nameidata nd; 2781 2782 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2783 if ((error = namei(&nd)) != 0) 2784 return (error); 2785 2786 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2787 2788 vrele(nd.ni_vp); 2789 return (error); 2790 } 2791 2792 /* 2793 * Common routine to set mode given a vnode. 2794 */ 2795 static int 2796 change_mode(struct vnode *vp, int mode, struct lwp *l) 2797 { 2798 struct mount *mp; 2799 struct vattr vattr; 2800 int error; 2801 2802 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2803 return (error); 2804 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2805 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2806 VATTR_NULL(&vattr); 2807 vattr.va_mode = mode & ALLPERMS; 2808 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2809 VOP_UNLOCK(vp, 0); 2810 vn_finished_write(mp, 0); 2811 return (error); 2812 } 2813 2814 /* 2815 * Set ownership given a path name; this version follows links. 2816 */ 2817 /* ARGSUSED */ 2818 int 2819 sys_chown(struct lwp *l, void *v, register_t *retval) 2820 { 2821 struct sys_chown_args /* { 2822 syscallarg(const char *) path; 2823 syscallarg(uid_t) uid; 2824 syscallarg(gid_t) gid; 2825 } */ *uap = v; 2826 int error; 2827 struct nameidata nd; 2828 2829 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2830 if ((error = namei(&nd)) != 0) 2831 return (error); 2832 2833 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2834 2835 vrele(nd.ni_vp); 2836 return (error); 2837 } 2838 2839 /* 2840 * Set ownership given a path name; this version follows links. 2841 * Provides POSIX semantics. 2842 */ 2843 /* ARGSUSED */ 2844 int 2845 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2846 { 2847 struct sys_chown_args /* { 2848 syscallarg(const char *) path; 2849 syscallarg(uid_t) uid; 2850 syscallarg(gid_t) gid; 2851 } */ *uap = v; 2852 int error; 2853 struct nameidata nd; 2854 2855 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2856 if ((error = namei(&nd)) != 0) 2857 return (error); 2858 2859 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2860 2861 vrele(nd.ni_vp); 2862 return (error); 2863 } 2864 2865 /* 2866 * Set ownership given a file descriptor. 2867 */ 2868 /* ARGSUSED */ 2869 int 2870 sys_fchown(struct lwp *l, void *v, register_t *retval) 2871 { 2872 struct sys_fchown_args /* { 2873 syscallarg(int) fd; 2874 syscallarg(uid_t) uid; 2875 syscallarg(gid_t) gid; 2876 } */ *uap = v; 2877 struct proc *p = l->l_proc; 2878 int error; 2879 struct file *fp; 2880 2881 /* getvnode() will use the descriptor for us */ 2882 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2883 return (error); 2884 2885 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2886 SCARG(uap, gid), l, 0); 2887 FILE_UNUSE(fp, l); 2888 return (error); 2889 } 2890 2891 /* 2892 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2893 */ 2894 /* ARGSUSED */ 2895 int 2896 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2897 { 2898 struct sys_fchown_args /* { 2899 syscallarg(int) fd; 2900 syscallarg(uid_t) uid; 2901 syscallarg(gid_t) gid; 2902 } */ *uap = v; 2903 struct proc *p = l->l_proc; 2904 int error; 2905 struct file *fp; 2906 2907 /* getvnode() will use the descriptor for us */ 2908 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2909 return (error); 2910 2911 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2912 SCARG(uap, gid), l, 1); 2913 FILE_UNUSE(fp, l); 2914 return (error); 2915 } 2916 2917 /* 2918 * Set ownership given a path name; this version does not follow links. 2919 */ 2920 /* ARGSUSED */ 2921 int 2922 sys_lchown(struct lwp *l, void *v, register_t *retval) 2923 { 2924 struct sys_lchown_args /* { 2925 syscallarg(const char *) path; 2926 syscallarg(uid_t) uid; 2927 syscallarg(gid_t) gid; 2928 } */ *uap = v; 2929 int error; 2930 struct nameidata nd; 2931 2932 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2933 if ((error = namei(&nd)) != 0) 2934 return (error); 2935 2936 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2937 2938 vrele(nd.ni_vp); 2939 return (error); 2940 } 2941 2942 /* 2943 * Set ownership given a path name; this version does not follow links. 2944 * Provides POSIX/XPG semantics. 2945 */ 2946 /* ARGSUSED */ 2947 int 2948 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2949 { 2950 struct sys_lchown_args /* { 2951 syscallarg(const char *) path; 2952 syscallarg(uid_t) uid; 2953 syscallarg(gid_t) gid; 2954 } */ *uap = v; 2955 int error; 2956 struct nameidata nd; 2957 2958 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2959 if ((error = namei(&nd)) != 0) 2960 return (error); 2961 2962 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2963 2964 vrele(nd.ni_vp); 2965 return (error); 2966 } 2967 2968 /* 2969 * Common routine to set ownership given a vnode. 2970 */ 2971 static int 2972 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2973 int posix_semantics) 2974 { 2975 struct mount *mp; 2976 struct vattr vattr; 2977 mode_t newmode; 2978 int error; 2979 2980 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2981 return (error); 2982 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2983 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2984 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2985 goto out; 2986 2987 #define CHANGED(x) ((int)(x) != -1) 2988 newmode = vattr.va_mode; 2989 if (posix_semantics) { 2990 /* 2991 * POSIX/XPG semantics: if the caller is not the super-user, 2992 * clear set-user-id and set-group-id bits. Both POSIX and 2993 * the XPG consider the behaviour for calls by the super-user 2994 * implementation-defined; we leave the set-user-id and set- 2995 * group-id settings intact in that case. 2996 */ 2997 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 2998 NULL) != 0) 2999 newmode &= ~(S_ISUID | S_ISGID); 3000 } else { 3001 /* 3002 * NetBSD semantics: when changing owner and/or group, 3003 * clear the respective bit(s). 3004 */ 3005 if (CHANGED(uid)) 3006 newmode &= ~S_ISUID; 3007 if (CHANGED(gid)) 3008 newmode &= ~S_ISGID; 3009 } 3010 /* Update va_mode iff altered. */ 3011 if (vattr.va_mode == newmode) 3012 newmode = VNOVAL; 3013 3014 VATTR_NULL(&vattr); 3015 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3016 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3017 vattr.va_mode = newmode; 3018 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3019 #undef CHANGED 3020 3021 out: 3022 VOP_UNLOCK(vp, 0); 3023 vn_finished_write(mp, 0); 3024 return (error); 3025 } 3026 3027 /* 3028 * Set the access and modification times given a path name; this 3029 * version follows links. 3030 */ 3031 /* ARGSUSED */ 3032 int 3033 sys_utimes(struct lwp *l, void *v, register_t *retval) 3034 { 3035 struct sys_utimes_args /* { 3036 syscallarg(const char *) path; 3037 syscallarg(const struct timeval *) tptr; 3038 } */ *uap = v; 3039 int error; 3040 struct nameidata nd; 3041 3042 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3043 if ((error = namei(&nd)) != 0) 3044 return (error); 3045 3046 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3047 3048 vrele(nd.ni_vp); 3049 return (error); 3050 } 3051 3052 /* 3053 * Set the access and modification times given a file descriptor. 3054 */ 3055 /* ARGSUSED */ 3056 int 3057 sys_futimes(struct lwp *l, void *v, register_t *retval) 3058 { 3059 struct sys_futimes_args /* { 3060 syscallarg(int) fd; 3061 syscallarg(const struct timeval *) tptr; 3062 } */ *uap = v; 3063 struct proc *p = l->l_proc; 3064 int error; 3065 struct file *fp; 3066 3067 /* getvnode() will use the descriptor for us */ 3068 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3069 return (error); 3070 3071 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3072 FILE_UNUSE(fp, l); 3073 return (error); 3074 } 3075 3076 /* 3077 * Set the access and modification times given a path name; this 3078 * version does not follow links. 3079 */ 3080 /* ARGSUSED */ 3081 int 3082 sys_lutimes(struct lwp *l, void *v, register_t *retval) 3083 { 3084 struct sys_lutimes_args /* { 3085 syscallarg(const char *) path; 3086 syscallarg(const struct timeval *) tptr; 3087 } */ *uap = v; 3088 int error; 3089 struct nameidata nd; 3090 3091 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3092 if ((error = namei(&nd)) != 0) 3093 return (error); 3094 3095 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3096 3097 vrele(nd.ni_vp); 3098 return (error); 3099 } 3100 3101 /* 3102 * Common routine to set access and modification times given a vnode. 3103 */ 3104 static int 3105 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3106 { 3107 struct mount *mp; 3108 struct vattr vattr; 3109 int error; 3110 3111 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3112 return (error); 3113 VATTR_NULL(&vattr); 3114 if (tptr == NULL) { 3115 nanotime(&vattr.va_atime); 3116 vattr.va_mtime = vattr.va_atime; 3117 vattr.va_vaflags |= VA_UTIMES_NULL; 3118 } else { 3119 struct timeval tv[2]; 3120 3121 error = copyin(tptr, tv, sizeof(tv)); 3122 if (error) 3123 goto out; 3124 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3125 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3126 } 3127 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3128 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3129 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3130 VOP_UNLOCK(vp, 0); 3131 out: 3132 vn_finished_write(mp, 0); 3133 return (error); 3134 } 3135 3136 /* 3137 * Truncate a file given its path name. 3138 */ 3139 /* ARGSUSED */ 3140 int 3141 sys_truncate(struct lwp *l, void *v, register_t *retval) 3142 { 3143 struct sys_truncate_args /* { 3144 syscallarg(const char *) path; 3145 syscallarg(int) pad; 3146 syscallarg(off_t) length; 3147 } */ *uap = v; 3148 struct vnode *vp; 3149 struct mount *mp; 3150 struct vattr vattr; 3151 int error; 3152 struct nameidata nd; 3153 3154 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3155 if ((error = namei(&nd)) != 0) 3156 return (error); 3157 vp = nd.ni_vp; 3158 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3159 vrele(vp); 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 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3168 VATTR_NULL(&vattr); 3169 vattr.va_size = SCARG(uap, length); 3170 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3171 } 3172 vput(vp); 3173 vn_finished_write(mp, 0); 3174 return (error); 3175 } 3176 3177 /* 3178 * Truncate a file given a file descriptor. 3179 */ 3180 /* ARGSUSED */ 3181 int 3182 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3183 { 3184 struct sys_ftruncate_args /* { 3185 syscallarg(int) fd; 3186 syscallarg(int) pad; 3187 syscallarg(off_t) length; 3188 } */ *uap = v; 3189 struct proc *p = l->l_proc; 3190 struct mount *mp; 3191 struct vattr vattr; 3192 struct vnode *vp; 3193 struct file *fp; 3194 int error; 3195 3196 /* getvnode() will use the descriptor for us */ 3197 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3198 return (error); 3199 if ((fp->f_flag & FWRITE) == 0) { 3200 error = EINVAL; 3201 goto out; 3202 } 3203 vp = (struct vnode *)fp->f_data; 3204 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3205 FILE_UNUSE(fp, l); 3206 return (error); 3207 } 3208 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3209 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3210 if (vp->v_type == VDIR) 3211 error = EISDIR; 3212 else if ((error = vn_writechk(vp)) == 0) { 3213 VATTR_NULL(&vattr); 3214 vattr.va_size = SCARG(uap, length); 3215 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3216 } 3217 VOP_UNLOCK(vp, 0); 3218 vn_finished_write(mp, 0); 3219 out: 3220 FILE_UNUSE(fp, l); 3221 return (error); 3222 } 3223 3224 /* 3225 * Sync an open file. 3226 */ 3227 /* ARGSUSED */ 3228 int 3229 sys_fsync(struct lwp *l, void *v, register_t *retval) 3230 { 3231 struct sys_fsync_args /* { 3232 syscallarg(int) fd; 3233 } */ *uap = v; 3234 struct proc *p = l->l_proc; 3235 struct vnode *vp; 3236 struct mount *mp; 3237 struct file *fp; 3238 int error; 3239 3240 /* getvnode() will use the descriptor for us */ 3241 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3242 return (error); 3243 vp = (struct vnode *)fp->f_data; 3244 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3245 FILE_UNUSE(fp, l); 3246 return (error); 3247 } 3248 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3249 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3250 if (error == 0 && bioops.io_fsync != NULL && 3251 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3252 (*bioops.io_fsync)(vp, 0); 3253 VOP_UNLOCK(vp, 0); 3254 vn_finished_write(mp, 0); 3255 FILE_UNUSE(fp, l); 3256 return (error); 3257 } 3258 3259 /* 3260 * Sync a range of file data. API modeled after that found in AIX. 3261 * 3262 * FDATASYNC indicates that we need only save enough metadata to be able 3263 * to re-read the written data. Note we duplicate AIX's requirement that 3264 * the file be open for writing. 3265 */ 3266 /* ARGSUSED */ 3267 int 3268 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3269 { 3270 struct sys_fsync_range_args /* { 3271 syscallarg(int) fd; 3272 syscallarg(int) flags; 3273 syscallarg(off_t) start; 3274 syscallarg(off_t) length; 3275 } */ *uap = v; 3276 struct proc *p = l->l_proc; 3277 struct vnode *vp; 3278 struct file *fp; 3279 int flags, nflags; 3280 off_t s, e, len; 3281 int error; 3282 3283 /* getvnode() will use the descriptor for us */ 3284 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3285 return (error); 3286 3287 if ((fp->f_flag & FWRITE) == 0) { 3288 error = EBADF; 3289 goto out; 3290 } 3291 3292 flags = SCARG(uap, flags); 3293 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3294 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3295 error = EINVAL; 3296 goto out; 3297 } 3298 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3299 if (flags & FDATASYNC) 3300 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3301 else 3302 nflags = FSYNC_WAIT; 3303 if (flags & FDISKSYNC) 3304 nflags |= FSYNC_CACHE; 3305 3306 len = SCARG(uap, length); 3307 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3308 if (len) { 3309 s = SCARG(uap, start); 3310 e = s + len; 3311 if (e < s) { 3312 error = EINVAL; 3313 goto out; 3314 } 3315 } else { 3316 e = 0; 3317 s = 0; 3318 } 3319 3320 vp = (struct vnode *)fp->f_data; 3321 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3322 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3323 3324 if (error == 0 && bioops.io_fsync != NULL && 3325 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3326 (*bioops.io_fsync)(vp, nflags); 3327 3328 VOP_UNLOCK(vp, 0); 3329 out: 3330 FILE_UNUSE(fp, l); 3331 return (error); 3332 } 3333 3334 /* 3335 * Sync the data of an open file. 3336 */ 3337 /* ARGSUSED */ 3338 int 3339 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3340 { 3341 struct sys_fdatasync_args /* { 3342 syscallarg(int) fd; 3343 } */ *uap = v; 3344 struct proc *p = l->l_proc; 3345 struct vnode *vp; 3346 struct file *fp; 3347 int error; 3348 3349 /* getvnode() will use the descriptor for us */ 3350 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3351 return (error); 3352 if ((fp->f_flag & FWRITE) == 0) { 3353 FILE_UNUSE(fp, l); 3354 return (EBADF); 3355 } 3356 vp = (struct vnode *)fp->f_data; 3357 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3358 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3359 VOP_UNLOCK(vp, 0); 3360 FILE_UNUSE(fp, l); 3361 return (error); 3362 } 3363 3364 /* 3365 * Rename files, (standard) BSD semantics frontend. 3366 */ 3367 /* ARGSUSED */ 3368 int 3369 sys_rename(struct lwp *l, void *v, register_t *retval) 3370 { 3371 struct sys_rename_args /* { 3372 syscallarg(const char *) from; 3373 syscallarg(const char *) to; 3374 } */ *uap = v; 3375 3376 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3377 } 3378 3379 /* 3380 * Rename files, POSIX semantics frontend. 3381 */ 3382 /* ARGSUSED */ 3383 int 3384 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3385 { 3386 struct sys___posix_rename_args /* { 3387 syscallarg(const char *) from; 3388 syscallarg(const char *) to; 3389 } */ *uap = v; 3390 3391 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3392 } 3393 3394 /* 3395 * Rename files. Source and destination must either both be directories, 3396 * or both not be directories. If target is a directory, it must be empty. 3397 * If `from' and `to' refer to the same object, the value of the `retain' 3398 * argument is used to determine whether `from' will be 3399 * 3400 * (retain == 0) deleted unless `from' and `to' refer to the same 3401 * object in the file system's name space (BSD). 3402 * (retain == 1) always retained (POSIX). 3403 */ 3404 static int 3405 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3406 { 3407 struct mount *mp = NULL; 3408 struct vnode *tvp, *fvp, *tdvp; 3409 struct nameidata fromnd, tond; 3410 struct proc *p; 3411 int error; 3412 3413 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3414 from, l); 3415 if ((error = namei(&fromnd)) != 0) 3416 return (error); 3417 if (fromnd.ni_dvp != fromnd.ni_vp) 3418 VOP_UNLOCK(fromnd.ni_dvp, 0); 3419 fvp = fromnd.ni_vp; 3420 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3421 if (error != 0) { 3422 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3423 vrele(fromnd.ni_dvp); 3424 vrele(fvp); 3425 if (fromnd.ni_startdir) 3426 vrele(fromnd.ni_startdir); 3427 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3428 return (error); 3429 } 3430 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3431 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3432 if ((error = namei(&tond)) != 0) { 3433 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3434 vrele(fromnd.ni_dvp); 3435 vrele(fvp); 3436 goto out1; 3437 } 3438 tdvp = tond.ni_dvp; 3439 tvp = tond.ni_vp; 3440 3441 if (tvp != NULL) { 3442 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3443 error = ENOTDIR; 3444 goto out; 3445 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3446 error = EISDIR; 3447 goto out; 3448 } 3449 } 3450 3451 if (fvp == tdvp) 3452 error = EINVAL; 3453 3454 /* 3455 * Source and destination refer to the same object. 3456 */ 3457 if (fvp == tvp) { 3458 if (retain) 3459 error = -1; 3460 else if (fromnd.ni_dvp == tdvp && 3461 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3462 !memcmp(fromnd.ni_cnd.cn_nameptr, 3463 tond.ni_cnd.cn_nameptr, 3464 fromnd.ni_cnd.cn_namelen)) 3465 error = -1; 3466 } 3467 3468 #if NVERIEXEC > 0 3469 if (!error) { 3470 pathname_t frompath = NULL, topath = NULL; 3471 3472 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3473 &frompath); 3474 if (!error) 3475 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3476 &topath); 3477 if (!error) 3478 error = veriexec_renamechk(fvp, pathname_path(frompath), 3479 tvp, pathname_path(topath), l); 3480 3481 pathname_put(frompath); 3482 pathname_put(topath); 3483 } 3484 #endif /* NVERIEXEC > 0 */ 3485 3486 out: 3487 p = l->l_proc; 3488 if (!error) { 3489 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3490 if (fromnd.ni_dvp != tdvp) 3491 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3492 if (tvp) { 3493 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3494 } 3495 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3496 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3497 } else { 3498 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3499 if (tdvp == tvp) 3500 vrele(tdvp); 3501 else 3502 vput(tdvp); 3503 if (tvp) 3504 vput(tvp); 3505 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3506 vrele(fromnd.ni_dvp); 3507 vrele(fvp); 3508 } 3509 vrele(tond.ni_startdir); 3510 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3511 out1: 3512 vn_finished_write(mp, 0); 3513 if (fromnd.ni_startdir) 3514 vrele(fromnd.ni_startdir); 3515 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3516 return (error == -1 ? 0 : error); 3517 } 3518 3519 /* 3520 * Make a directory file. 3521 */ 3522 /* ARGSUSED */ 3523 int 3524 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3525 { 3526 struct sys_mkdir_args /* { 3527 syscallarg(const char *) path; 3528 syscallarg(int) mode; 3529 } */ *uap = v; 3530 struct proc *p = l->l_proc; 3531 struct mount *mp; 3532 struct vnode *vp; 3533 struct vattr vattr; 3534 int error; 3535 struct nameidata nd; 3536 3537 restart: 3538 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3539 SCARG(uap, path), l); 3540 if ((error = namei(&nd)) != 0) 3541 return (error); 3542 vp = nd.ni_vp; 3543 if (vp != NULL) { 3544 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3545 if (nd.ni_dvp == vp) 3546 vrele(nd.ni_dvp); 3547 else 3548 vput(nd.ni_dvp); 3549 vrele(vp); 3550 return (EEXIST); 3551 } 3552 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3553 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3554 if (nd.ni_dvp == vp) 3555 vrele(nd.ni_dvp); 3556 else 3557 vput(nd.ni_dvp); 3558 if ((error = vn_start_write(NULL, &mp, 3559 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3560 return (error); 3561 goto restart; 3562 } 3563 VATTR_NULL(&vattr); 3564 vattr.va_type = VDIR; 3565 vattr.va_mode = 3566 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3567 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3568 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3569 if (!error) 3570 vput(nd.ni_vp); 3571 vn_finished_write(mp, 0); 3572 return (error); 3573 } 3574 3575 /* 3576 * Remove a directory file. 3577 */ 3578 /* ARGSUSED */ 3579 int 3580 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3581 { 3582 struct sys_rmdir_args /* { 3583 syscallarg(const char *) path; 3584 } */ *uap = v; 3585 struct mount *mp; 3586 struct vnode *vp; 3587 int error; 3588 struct nameidata nd; 3589 3590 restart: 3591 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3592 SCARG(uap, path), l); 3593 if ((error = namei(&nd)) != 0) 3594 return (error); 3595 vp = nd.ni_vp; 3596 if (vp->v_type != VDIR) { 3597 error = ENOTDIR; 3598 goto out; 3599 } 3600 /* 3601 * No rmdir "." please. 3602 */ 3603 if (nd.ni_dvp == vp) { 3604 error = EINVAL; 3605 goto out; 3606 } 3607 /* 3608 * The root of a mounted filesystem cannot be deleted. 3609 */ 3610 if (vp->v_flag & VROOT) { 3611 error = EBUSY; 3612 goto out; 3613 } 3614 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3615 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3616 if (nd.ni_dvp == vp) 3617 vrele(nd.ni_dvp); 3618 else 3619 vput(nd.ni_dvp); 3620 vput(vp); 3621 if ((error = vn_start_write(NULL, &mp, 3622 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3623 return (error); 3624 goto restart; 3625 } 3626 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3627 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3628 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3629 vn_finished_write(mp, 0); 3630 return (error); 3631 3632 out: 3633 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3634 if (nd.ni_dvp == vp) 3635 vrele(nd.ni_dvp); 3636 else 3637 vput(nd.ni_dvp); 3638 vput(vp); 3639 return (error); 3640 } 3641 3642 /* 3643 * Read a block of directory entries in a file system independent format. 3644 */ 3645 int 3646 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3647 { 3648 struct sys___getdents30_args /* { 3649 syscallarg(int) fd; 3650 syscallarg(char *) buf; 3651 syscallarg(size_t) count; 3652 } */ *uap = v; 3653 struct proc *p = l->l_proc; 3654 struct file *fp; 3655 int error, done; 3656 3657 /* getvnode() will use the descriptor for us */ 3658 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3659 return (error); 3660 if ((fp->f_flag & FREAD) == 0) { 3661 error = EBADF; 3662 goto out; 3663 } 3664 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3665 SCARG(uap, count), &done, l, 0, 0); 3666 #ifdef KTRACE 3667 if (!error && KTRPOINT(p, KTR_GENIO)) { 3668 struct iovec iov; 3669 iov.iov_base = SCARG(uap, buf); 3670 iov.iov_len = done; 3671 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3672 } 3673 #endif 3674 *retval = done; 3675 out: 3676 FILE_UNUSE(fp, l); 3677 return (error); 3678 } 3679 3680 /* 3681 * Set the mode mask for creation of filesystem nodes. 3682 */ 3683 int 3684 sys_umask(struct lwp *l, void *v, register_t *retval) 3685 { 3686 struct sys_umask_args /* { 3687 syscallarg(mode_t) newmask; 3688 } */ *uap = v; 3689 struct proc *p = l->l_proc; 3690 struct cwdinfo *cwdi; 3691 3692 cwdi = p->p_cwdi; 3693 *retval = cwdi->cwdi_cmask; 3694 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3695 return (0); 3696 } 3697 3698 /* 3699 * Void all references to file by ripping underlying filesystem 3700 * away from vnode. 3701 */ 3702 /* ARGSUSED */ 3703 int 3704 sys_revoke(struct lwp *l, void *v, register_t *retval) 3705 { 3706 struct sys_revoke_args /* { 3707 syscallarg(const char *) path; 3708 } */ *uap = v; 3709 struct mount *mp; 3710 struct vnode *vp; 3711 struct vattr vattr; 3712 int error; 3713 struct nameidata nd; 3714 3715 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3716 if ((error = namei(&nd)) != 0) 3717 return (error); 3718 vp = nd.ni_vp; 3719 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3720 goto out; 3721 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3722 (error = kauth_authorize_generic(l->l_cred, 3723 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3724 goto out; 3725 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3726 goto out; 3727 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3728 VOP_REVOKE(vp, REVOKEALL); 3729 vn_finished_write(mp, 0); 3730 out: 3731 vrele(vp); 3732 return (error); 3733 } 3734 3735 /* 3736 * Convert a user file descriptor to a kernel file entry. 3737 */ 3738 int 3739 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3740 { 3741 struct vnode *vp; 3742 struct file *fp; 3743 3744 if ((fp = fd_getfile(fdp, fd)) == NULL) 3745 return (EBADF); 3746 3747 FILE_USE(fp); 3748 3749 if (fp->f_type != DTYPE_VNODE) { 3750 FILE_UNUSE(fp, NULL); 3751 return (EINVAL); 3752 } 3753 3754 vp = (struct vnode *)fp->f_data; 3755 if (vp->v_type == VBAD) { 3756 FILE_UNUSE(fp, NULL); 3757 return (EBADF); 3758 } 3759 3760 *fpp = fp; 3761 return (0); 3762 } 3763
Indexes created Fri Oct 03 02:09:57 GMT 2025