vfs_syscalls.c revision 1.567
1 /* $NetBSD: vfs_syscalls.c,v 1.567 2024/08/11 13:09:58 bad Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009, 2019, 2020, 2023 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 66 */ 67 68 /* 69 * Virtual File System System Calls 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.567 2024/08/11 13:09:58 bad Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "opt_fileassoc.h" 77 #include "veriexec.h" 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/namei.h> 83 #include <sys/filedesc.h> 84 #include <sys/kernel.h> 85 #include <sys/file.h> 86 #include <sys/fcntl.h> 87 #include <sys/stat.h> 88 #include <sys/vnode.h> 89 #include <sys/mount.h> 90 #include <sys/fstrans.h> 91 #include <sys/proc.h> 92 #include <sys/uio.h> 93 #include <sys/kmem.h> 94 #include <sys/dirent.h> 95 #include <sys/sysctl.h> 96 #include <sys/syscallargs.h> 97 #include <sys/vfs_syscalls.h> 98 #include <sys/quota.h> 99 #include <sys/quotactl.h> 100 #include <sys/ktrace.h> 101 #ifdef FILEASSOC 102 #include <sys/fileassoc.h> 103 #endif /* FILEASSOC */ 104 #include <sys/extattr.h> 105 #include <sys/verified_exec.h> 106 #include <sys/kauth.h> 107 #include <sys/atomic.h> 108 #include <sys/module.h> 109 #include <sys/buf.h> 110 #include <sys/event.h> 111 #include <sys/compat_stub.h> 112 113 #include <miscfs/genfs/genfs.h> 114 #include <miscfs/specfs/specdev.h> 115 116 #include <nfs/rpcv2.h> 117 #include <nfs/nfsproto.h> 118 #include <nfs/nfs.h> 119 #include <nfs/nfs_var.h> 120 121 /* XXX this shouldn't be here */ 122 #ifndef OFF_T_MAX 123 #define OFF_T_MAX __type_max(off_t) 124 #endif 125 126 static int change_flags(struct vnode *, u_long, struct lwp *); 127 static int change_mode(struct vnode *, int, struct lwp *); 128 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 129 static int do_sys_openat(lwp_t *, int, const char *, int, int, int *); 130 static int do_sys_mkdirat(struct lwp *l, int, const char *, mode_t, 131 enum uio_seg); 132 static int do_sys_mkfifoat(struct lwp *, int, const char *, mode_t); 133 static int do_sys_symlinkat(struct lwp *, const char *, int, const char *, 134 enum uio_seg); 135 static int do_sys_renameat(struct lwp *l, int, const char *, int, const char *, 136 enum uio_seg, int); 137 static int do_sys_readlinkat(struct lwp *, int, const char *, char *, 138 size_t, register_t *); 139 static int do_sys_unlinkat(struct lwp *, int, const char *, int, enum uio_seg); 140 141 static int fd_nameiat(struct lwp *, int, struct nameidata *); 142 static int fd_nameiat_simple_user(struct lwp *, int, const char *, 143 namei_simple_flags_t, struct vnode **); 144 145 /* 146 * This table is used to maintain compatibility with 4.3BSD 147 * and NetBSD 0.9 mount syscalls - and possibly other systems. 148 * Note, the order is important! 149 * 150 * Do not modify this table. It should only contain filesystems 151 * supported by NetBSD 0.9 and 4.3BSD. 152 */ 153 const char * const mountcompatnames[] = { 154 NULL, /* 0 = MOUNT_NONE */ 155 MOUNT_FFS, /* 1 = MOUNT_UFS */ 156 MOUNT_NFS, /* 2 */ 157 MOUNT_MFS, /* 3 */ 158 MOUNT_MSDOS, /* 4 */ 159 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 160 MOUNT_FDESC, /* 6 */ 161 MOUNT_KERNFS, /* 7 */ 162 NULL, /* 8 = MOUNT_DEVFS */ 163 MOUNT_AFS, /* 9 */ 164 }; 165 166 const u_int nmountcompatnames = __arraycount(mountcompatnames); 167 168 /* 169 * Filter event method for EVFILT_FS. 170 */ 171 static struct klist fs_klist; 172 static kmutex_t fs_klist_lock; 173 174 CTASSERT((NOTE_SUBMIT & VQ_MOUNT) == 0); 175 CTASSERT((NOTE_SUBMIT & VQ_UNMOUNT) == 0); 176 177 void 178 vfs_evfilt_fs_init(void) 179 { 180 klist_init(&fs_klist); 181 mutex_init(&fs_klist_lock, MUTEX_DEFAULT, IPL_NONE); 182 } 183 184 static int 185 filt_fsattach(struct knote *kn) 186 { 187 mutex_enter(&fs_klist_lock); 188 kn->kn_flags |= EV_CLEAR; 189 klist_insert(&fs_klist, kn); 190 mutex_exit(&fs_klist_lock); 191 192 return 0; 193 } 194 195 static void 196 filt_fsdetach(struct knote *kn) 197 { 198 mutex_enter(&fs_klist_lock); 199 klist_remove(&fs_klist, kn); 200 mutex_exit(&fs_klist_lock); 201 } 202 203 static int 204 filt_fs(struct knote *kn, long hint) 205 { 206 int rv; 207 208 if (hint & NOTE_SUBMIT) { 209 KASSERT(mutex_owned(&fs_klist_lock)); 210 kn->kn_fflags |= hint & ~NOTE_SUBMIT; 211 } else { 212 mutex_enter(&fs_klist_lock); 213 } 214 215 rv = (kn->kn_fflags != 0); 216 217 if ((hint & NOTE_SUBMIT) == 0) { 218 mutex_exit(&fs_klist_lock); 219 } 220 221 return rv; 222 } 223 224 /* referenced in kern_event.c */ 225 const struct filterops fs_filtops = { 226 .f_flags = FILTEROP_MPSAFE, 227 .f_attach = filt_fsattach, 228 .f_detach = filt_fsdetach, 229 .f_event = filt_fs, 230 }; 231 232 static int 233 fd_nameiat(struct lwp *l, int fdat, struct nameidata *ndp) 234 { 235 file_t *dfp; 236 int error; 237 const char *path = pathbuf_stringcopy_get(ndp->ni_pathbuf); 238 239 if (fdat != AT_FDCWD && path[0] != '/') { 240 if ((error = fd_getvnode(fdat, &dfp)) != 0) 241 goto out; 242 243 NDAT(ndp, dfp->f_vnode); 244 } 245 246 error = namei(ndp); 247 248 if (fdat != AT_FDCWD && path[0] != '/') 249 fd_putfile(fdat); 250 out: 251 pathbuf_stringcopy_put(ndp->ni_pathbuf, path); 252 return error; 253 } 254 255 static int 256 fd_nameiat_simple_user(struct lwp *l, int fdat, const char *path, 257 namei_simple_flags_t sflags, struct vnode **vp_ret) 258 { 259 file_t *dfp; 260 struct vnode *dvp; 261 int error; 262 struct pathbuf *pb; 263 const char *p; 264 265 error = pathbuf_copyin(path, &pb); 266 if (error) { 267 return error; 268 } 269 p = pathbuf_stringcopy_get(pb); 270 271 if (fdat != AT_FDCWD && p[0] != '/') { 272 if ((error = fd_getvnode(fdat, &dfp)) != 0) 273 goto out; 274 275 dvp = dfp->f_vnode; 276 } else { 277 dvp = NULL; 278 } 279 280 error = nameiat_simple(dvp, pb, sflags, vp_ret); 281 282 if (fdat != AT_FDCWD && p[0] != '/') 283 fd_putfile(fdat); 284 285 out: 286 pathbuf_stringcopy_put(pb, p); 287 pathbuf_destroy(pb); 288 289 return error; 290 } 291 292 static int 293 open_setfp(struct lwp *l, file_t *fp, struct vnode *vp, int indx, int flags) 294 { 295 int error; 296 297 fp->f_flag = flags & FMASK; 298 fp->f_type = DTYPE_VNODE; 299 fp->f_ops = &vnops; 300 fp->f_vnode = vp; 301 302 if (flags & (O_EXLOCK | O_SHLOCK)) { 303 struct flock lf; 304 int type; 305 306 lf.l_whence = SEEK_SET; 307 lf.l_start = 0; 308 lf.l_len = 0; 309 if (flags & O_EXLOCK) 310 lf.l_type = F_WRLCK; 311 else 312 lf.l_type = F_RDLCK; 313 type = F_FLOCK; 314 if ((flags & FNONBLOCK) == 0) 315 type |= F_WAIT; 316 VOP_UNLOCK(vp); 317 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 318 if (error) { 319 (void) vn_close(vp, fp->f_flag, fp->f_cred); 320 fd_abort(l->l_proc, fp, indx); 321 return error; 322 } 323 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 324 atomic_or_uint(&fp->f_flag, FHASLOCK); 325 } 326 if (flags & O_CLOEXEC) 327 fd_set_exclose(l, indx, true); 328 return 0; 329 } 330 331 static int 332 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 333 void *data, size_t *data_len) 334 { 335 struct mount *mp; 336 int error = 0, saved_flags; 337 338 mp = vp->v_mount; 339 saved_flags = mp->mnt_flag; 340 341 /* We can operate only on VV_ROOT nodes. */ 342 if ((vp->v_vflag & VV_ROOT) == 0) { 343 error = EINVAL; 344 goto out; 345 } 346 347 /* 348 * We only allow the filesystem to be reloaded if it 349 * is currently mounted read-only. Additionally, we 350 * prevent read-write to read-only downgrades. 351 */ 352 if ((flags & (MNT_RELOAD | MNT_RDONLY)) != 0 && 353 (mp->mnt_flag & MNT_RDONLY) == 0 && 354 (mp->mnt_iflag & IMNT_CAN_RWTORO) == 0) { 355 error = EOPNOTSUPP; /* Needs translation */ 356 goto out; 357 } 358 359 /* 360 * Enabling MNT_UNION requires a covered mountpoint and 361 * must not happen on the root mount. 362 */ 363 if ((flags & MNT_UNION) != 0 && mp->mnt_vnodecovered == NULLVP) { 364 error = EOPNOTSUPP; 365 goto out; 366 } 367 368 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 369 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 370 if (error) 371 goto out; 372 373 error = vfs_suspend(mp, 0); 374 if (error) 375 goto out; 376 377 mutex_enter(mp->mnt_updating); 378 379 mp->mnt_flag &= ~MNT_OP_FLAGS; 380 mp->mnt_flag |= flags & MNT_OP_FLAGS; 381 382 /* 383 * Set the mount level flags. 384 */ 385 if ((flags & MNT_RDONLY) != (mp->mnt_flag & MNT_RDONLY)) { 386 if ((flags & MNT_RDONLY)) 387 mp->mnt_iflag |= IMNT_WANTRDONLY; 388 else 389 mp->mnt_iflag |= IMNT_WANTRDWR; 390 } 391 mp->mnt_flag &= ~MNT_BASIC_FLAGS; 392 mp->mnt_flag |= flags & MNT_BASIC_FLAGS; 393 if ((mp->mnt_iflag & IMNT_WANTRDONLY)) 394 mp->mnt_flag &= ~MNT_RDONLY; 395 396 error = VFS_MOUNT(mp, path, data, data_len); 397 398 if (error && data != NULL) { 399 int error2; 400 401 /* 402 * Update failed; let's try and see if it was an 403 * export request. For compat with 3.0 and earlier. 404 */ 405 error2 = vfs_hooks_reexport(mp, path, data); 406 407 /* 408 * Only update error code if the export request was 409 * understood but some problem occurred while 410 * processing it. 411 */ 412 if (error2 != EJUSTRETURN) 413 error = error2; 414 } 415 416 if (error == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) 417 mp->mnt_flag |= MNT_RDONLY; 418 if (error) 419 mp->mnt_flag = saved_flags; 420 mp->mnt_flag &= ~MNT_OP_FLAGS; 421 mp->mnt_iflag &= ~(IMNT_WANTRDONLY | IMNT_WANTRDWR); 422 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 423 if ((mp->mnt_iflag & IMNT_ONWORKLIST) == 0) 424 vfs_syncer_add_to_worklist(mp); 425 } else { 426 if ((mp->mnt_iflag & IMNT_ONWORKLIST) != 0) 427 vfs_syncer_remove_from_worklist(mp); 428 } 429 mutex_exit(mp->mnt_updating); 430 vfs_resume(mp); 431 432 if ((error == 0) && !(saved_flags & MNT_EXTATTR) && 433 (flags & MNT_EXTATTR)) { 434 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_START, 435 NULL, 0, NULL) != 0) { 436 printf("%s: failed to start extattr, error = %d", 437 mp->mnt_stat.f_mntonname, error); 438 mp->mnt_flag &= ~MNT_EXTATTR; 439 } 440 } 441 442 if ((error == 0) && (saved_flags & MNT_EXTATTR) && 443 !(flags & MNT_EXTATTR)) { 444 if (VFS_EXTATTRCTL(mp, EXTATTR_CMD_STOP, 445 NULL, 0, NULL) != 0) { 446 printf("%s: failed to stop extattr, error = %d", 447 mp->mnt_stat.f_mntonname, error); 448 mp->mnt_flag |= MNT_RDONLY; 449 } 450 } 451 out: 452 return (error); 453 } 454 455 static int 456 mount_get_vfsops(const char *fstype, enum uio_seg type_seg, 457 struct vfsops **vfsops) 458 { 459 char fstypename[sizeof(((struct statvfs *)NULL)->f_fstypename)]; 460 int error; 461 462 if (type_seg == UIO_USERSPACE) { 463 /* Copy file-system type from userspace. */ 464 error = copyinstr(fstype, fstypename, sizeof(fstypename), NULL); 465 } else { 466 error = copystr(fstype, fstypename, sizeof(fstypename), NULL); 467 KASSERT(error == 0); 468 } 469 470 if (error) { 471 /* 472 * Historically, filesystem types were identified by numbers. 473 * If we get an integer for the filesystem type instead of a 474 * string, we check to see if it matches one of the historic 475 * filesystem types. 476 */ 477 u_long fsindex = (u_long)fstype; 478 if (fsindex >= nmountcompatnames || 479 mountcompatnames[fsindex] == NULL) 480 return ENODEV; 481 strlcpy(fstypename, mountcompatnames[fsindex], 482 sizeof(fstypename)); 483 } 484 485 /* Accept `ufs' as an alias for `ffs', for compatibility. */ 486 if (strcmp(fstypename, "ufs") == 0) 487 fstypename[0] = 'f'; 488 489 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 490 return 0; 491 492 /* If we can autoload a vfs module, try again */ 493 (void)module_autoload(fstypename, MODULE_CLASS_VFS); 494 495 if ((*vfsops = vfs_getopsbyname(fstypename)) != NULL) 496 return 0; 497 498 return ENODEV; 499 } 500 501 static int 502 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 503 void *data, size_t *data_len) 504 { 505 struct mount *mp; 506 int error; 507 508 /* If MNT_GETARGS is specified, it should be the only flag. */ 509 if (flags & ~MNT_GETARGS) 510 return EINVAL; 511 512 mp = vp->v_mount; 513 514 /* XXX: probably some notion of "can see" here if we want isolation. */ 515 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 516 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 517 if (error) 518 return error; 519 520 if ((vp->v_vflag & VV_ROOT) == 0) 521 return EINVAL; 522 523 if (vfs_busy(mp)) 524 return EPERM; 525 526 mutex_enter(mp->mnt_updating); 527 mp->mnt_flag &= ~MNT_OP_FLAGS; 528 mp->mnt_flag |= MNT_GETARGS; 529 error = VFS_MOUNT(mp, path, data, data_len); 530 mp->mnt_flag &= ~MNT_OP_FLAGS; 531 mutex_exit(mp->mnt_updating); 532 533 vfs_unbusy(mp); 534 return (error); 535 } 536 537 int 538 sys___mount50(struct lwp *l, const struct sys___mount50_args *uap, register_t *retval) 539 { 540 /* { 541 syscallarg(const char *) type; 542 syscallarg(const char *) path; 543 syscallarg(int) flags; 544 syscallarg(void *) data; 545 syscallarg(size_t) data_len; 546 } */ 547 548 return do_sys_mount(l, SCARG(uap, type), UIO_USERSPACE, SCARG(uap, path), 549 SCARG(uap, flags), SCARG(uap, data), UIO_USERSPACE, 550 SCARG(uap, data_len), retval); 551 } 552 553 int 554 do_sys_mount(struct lwp *l, const char *type, enum uio_seg type_seg, 555 const char *path, int flags, void *data, enum uio_seg data_seg, 556 size_t data_len, register_t *retval) 557 { 558 struct vfsops *vfsops = NULL; /* XXX gcc4.8 */ 559 struct vnode *vp; 560 void *data_buf = data; 561 bool vfsopsrele = false; 562 size_t alloc_sz = 0; 563 int error; 564 565 /* 566 * Get vnode to be covered 567 */ 568 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 569 if (error != 0) { 570 vp = NULL; 571 goto done; 572 } 573 574 if (flags & (MNT_GETARGS | MNT_UPDATE)) { 575 vfsops = vp->v_mount->mnt_op; 576 } else { 577 /* 'type' is userspace */ 578 error = mount_get_vfsops(type, type_seg, &vfsops); 579 if (error != 0) 580 goto done; 581 vfsopsrele = true; 582 } 583 584 /* 585 * We allow data to be NULL, even for userspace. Some fs's don't need 586 * it. The others will handle NULL. 587 */ 588 if (data != NULL && data_seg == UIO_USERSPACE) { 589 if (data_len == 0) { 590 /* No length supplied, use default for filesystem */ 591 data_len = vfsops->vfs_min_mount_data; 592 593 /* 594 * Hopefully a longer buffer won't make copyin() fail. 595 * For compatibility with 3.0 and earlier. 596 */ 597 if (flags & MNT_UPDATE 598 && data_len < sizeof (struct mnt_export_args30)) 599 data_len = sizeof (struct mnt_export_args30); 600 } 601 if ((data_len == 0) || (data_len > VFS_MAX_MOUNT_DATA)) { 602 error = EINVAL; 603 goto done; 604 } 605 alloc_sz = data_len; 606 data_buf = kmem_alloc(alloc_sz, KM_SLEEP); 607 608 /* NFS needs the buffer even for mnt_getargs .... */ 609 error = copyin(data, data_buf, data_len); 610 if (error != 0) 611 goto done; 612 } 613 614 if (flags & MNT_GETARGS) { 615 if (data_len == 0) { 616 error = EINVAL; 617 goto done; 618 } 619 error = mount_getargs(l, vp, path, flags, data_buf, &data_len); 620 if (error != 0) 621 goto done; 622 if (data_seg == UIO_USERSPACE) 623 error = copyout(data_buf, data, data_len); 624 *retval = data_len; 625 } else if (flags & MNT_UPDATE) { 626 error = mount_update(l, vp, path, flags, data_buf, &data_len); 627 } else { 628 /* Locking is handled internally in mount_domount(). */ 629 KASSERT(vfsopsrele == true); 630 error = mount_domount(l, &vp, vfsops, path, flags, data_buf, 631 &data_len); 632 vfsopsrele = false; 633 } 634 if (!error) { 635 mutex_enter(&fs_klist_lock); 636 KNOTE(&fs_klist, NOTE_SUBMIT | VQ_MOUNT); 637 mutex_exit(&fs_klist_lock); 638 } 639 640 done: 641 if (vfsopsrele) 642 vfs_delref(vfsops); 643 if (vp != NULL) { 644 vrele(vp); 645 } 646 if (data_buf != data) 647 kmem_free(data_buf, alloc_sz); 648 return (error); 649 } 650 651 /* 652 * Unmount a file system. 653 * 654 * Note: unmount takes a path to the vnode mounted on as argument, 655 * not special file (as before). 656 */ 657 /* ARGSUSED */ 658 int 659 sys_unmount(struct lwp *l, const struct sys_unmount_args *uap, register_t *retval) 660 { 661 /* { 662 syscallarg(const char *) path; 663 syscallarg(int) flags; 664 } */ 665 struct vnode *vp; 666 struct mount *mp; 667 int error; 668 struct pathbuf *pb; 669 struct nameidata nd; 670 671 error = pathbuf_copyin(SCARG(uap, path), &pb); 672 if (error) { 673 return error; 674 } 675 676 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | TRYEMULROOT, pb); 677 if ((error = namei(&nd)) != 0) { 678 pathbuf_destroy(pb); 679 return error; 680 } 681 vp = nd.ni_vp; 682 pathbuf_destroy(pb); 683 684 mp = vp->v_mount; 685 vfs_ref(mp); 686 VOP_UNLOCK(vp); 687 688 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 689 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 690 if (error) { 691 vrele(vp); 692 vfs_rele(mp); 693 return (error); 694 } 695 696 /* 697 * Don't allow unmounting the root file system. 698 */ 699 if (mp->mnt_flag & MNT_ROOTFS) { 700 vrele(vp); 701 vfs_rele(mp); 702 return (EINVAL); 703 } 704 705 /* 706 * Must be the root of the filesystem 707 */ 708 if ((vp->v_vflag & VV_ROOT) == 0) { 709 vrele(vp); 710 vfs_rele(mp); 711 return (EINVAL); 712 } 713 714 vrele(vp); 715 error = dounmount(mp, SCARG(uap, flags), l); 716 vfs_rele(mp); 717 if (!error) { 718 mutex_enter(&fs_klist_lock); 719 KNOTE(&fs_klist, NOTE_SUBMIT | VQ_UNMOUNT); 720 mutex_exit(&fs_klist_lock); 721 } 722 return error; 723 } 724 725 /* 726 * Sync each mounted filesystem. 727 */ 728 #ifdef DEBUG 729 int syncprt = 0; 730 struct ctldebug debug0 = { "syncprt", &syncprt }; 731 #endif 732 733 void 734 do_sys_sync(struct lwp *l) 735 { 736 mount_iterator_t *iter; 737 struct mount *mp; 738 int asyncflag; 739 740 mountlist_iterator_init(&iter); 741 while ((mp = mountlist_iterator_next(iter)) != NULL) { 742 mutex_enter(mp->mnt_updating); 743 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 744 /* 745 * Temporarily clear the MNT_ASYNC flags so that 746 * bwrite() doesnt convert the sync writes to 747 * delayed writes. 748 */ 749 asyncflag = mp->mnt_flag & MNT_ASYNC; 750 mp->mnt_flag &= ~MNT_ASYNC; 751 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred); 752 if (asyncflag) 753 mp->mnt_flag |= MNT_ASYNC; 754 } 755 mutex_exit(mp->mnt_updating); 756 } 757 mountlist_iterator_destroy(iter); 758 #ifdef DEBUG 759 if (syncprt) 760 vfs_bufstats(); 761 #endif /* DEBUG */ 762 } 763 764 static bool 765 sync_vnode_filter(void *cookie, vnode_t *vp) 766 { 767 768 if (vp->v_numoutput > 0) { 769 ++*(int *)cookie; 770 } 771 return false; 772 } 773 774 int 775 vfs_syncwait(void) 776 { 777 int nbusy, nbusy_prev, iter; 778 struct vnode_iterator *vniter; 779 mount_iterator_t *mpiter; 780 struct mount *mp; 781 782 for (nbusy_prev = 0, iter = 0; iter < 20;) { 783 nbusy = 0; 784 mountlist_iterator_init(&mpiter); 785 while ((mp = mountlist_iterator_next(mpiter)) != NULL) { 786 vnode_t *vp __diagused; 787 vfs_vnode_iterator_init(mp, &vniter); 788 vp = vfs_vnode_iterator_next(vniter, 789 sync_vnode_filter, &nbusy); 790 KASSERT(vp == NULL); 791 vfs_vnode_iterator_destroy(vniter); 792 } 793 mountlist_iterator_destroy(mpiter); 794 795 if (nbusy == 0) 796 break; 797 if (nbusy_prev == 0) 798 nbusy_prev = nbusy; 799 printf("%d ", nbusy); 800 kpause("syncwait", false, MAX(1, hz / 25 * iter), NULL); 801 if (nbusy >= nbusy_prev) /* we didn't flush anything */ 802 iter++; 803 else 804 nbusy_prev = nbusy; 805 } 806 807 if (nbusy) { 808 #if defined(DEBUG) || defined(DEBUG_HALT_BUSY) 809 printf("giving up\nPrinting vnodes for busy buffers\n"); 810 mountlist_iterator_init(&mpiter); 811 while ((mp = mountlist_iterator_next(mpiter)) != NULL) { 812 vnode_t *vp; 813 vfs_vnode_iterator_init(mp, &vniter); 814 vp = vfs_vnode_iterator_next(vniter, 815 NULL, NULL); 816 mutex_enter(vp->v_interlock); 817 if (vp->v_numoutput > 0) 818 vprint(NULL, vp); 819 mutex_exit(vp->v_interlock); 820 vrele(vp); 821 vfs_vnode_iterator_destroy(vniter); 822 } 823 mountlist_iterator_destroy(mpiter); 824 #endif 825 } 826 827 return nbusy; 828 } 829 830 /* ARGSUSED */ 831 int 832 sys_sync(struct lwp *l, const void *v, register_t *retval) 833 { 834 do_sys_sync(l); 835 return (0); 836 } 837 838 839 /* 840 * Access or change filesystem quotas. 841 * 842 * (this is really 14 different calls bundled into one) 843 */ 844 845 static int 846 do_sys_quotactl_stat(struct mount *mp, struct quotastat *info_u) 847 { 848 struct quotastat info_k; 849 int error; 850 851 /* ensure any padding bytes are cleared */ 852 memset(&info_k, 0, sizeof(info_k)); 853 854 error = vfs_quotactl_stat(mp, &info_k); 855 if (error) { 856 return error; 857 } 858 859 return copyout(&info_k, info_u, sizeof(info_k)); 860 } 861 862 static int 863 do_sys_quotactl_idtypestat(struct mount *mp, int idtype, 864 struct quotaidtypestat *info_u) 865 { 866 struct quotaidtypestat info_k; 867 int error; 868 869 /* ensure any padding bytes are cleared */ 870 memset(&info_k, 0, sizeof(info_k)); 871 872 error = vfs_quotactl_idtypestat(mp, idtype, &info_k); 873 if (error) { 874 return error; 875 } 876 877 return copyout(&info_k, info_u, sizeof(info_k)); 878 } 879 880 static int 881 do_sys_quotactl_objtypestat(struct mount *mp, int objtype, 882 struct quotaobjtypestat *info_u) 883 { 884 struct quotaobjtypestat info_k; 885 int error; 886 887 /* ensure any padding bytes are cleared */ 888 memset(&info_k, 0, sizeof(info_k)); 889 890 error = vfs_quotactl_objtypestat(mp, objtype, &info_k); 891 if (error) { 892 return error; 893 } 894 895 return copyout(&info_k, info_u, sizeof(info_k)); 896 } 897 898 static int 899 do_sys_quotactl_get(struct mount *mp, const struct quotakey *key_u, 900 struct quotaval *val_u) 901 { 902 struct quotakey key_k; 903 struct quotaval val_k; 904 int error; 905 906 /* ensure any padding bytes are cleared */ 907 memset(&val_k, 0, sizeof(val_k)); 908 909 error = copyin(key_u, &key_k, sizeof(key_k)); 910 if (error) { 911 return error; 912 } 913 914 error = vfs_quotactl_get(mp, &key_k, &val_k); 915 if (error) { 916 return error; 917 } 918 919 return copyout(&val_k, val_u, sizeof(val_k)); 920 } 921 922 static int 923 do_sys_quotactl_put(struct mount *mp, const struct quotakey *key_u, 924 const struct quotaval *val_u) 925 { 926 struct quotakey key_k; 927 struct quotaval val_k; 928 int error; 929 930 error = copyin(key_u, &key_k, sizeof(key_k)); 931 if (error) { 932 return error; 933 } 934 935 error = copyin(val_u, &val_k, sizeof(val_k)); 936 if (error) { 937 return error; 938 } 939 940 return vfs_quotactl_put(mp, &key_k, &val_k); 941 } 942 943 static int 944 do_sys_quotactl_del(struct mount *mp, const struct quotakey *key_u) 945 { 946 struct quotakey key_k; 947 int error; 948 949 error = copyin(key_u, &key_k, sizeof(key_k)); 950 if (error) { 951 return error; 952 } 953 954 return vfs_quotactl_del(mp, &key_k); 955 } 956 957 static int 958 do_sys_quotactl_cursoropen(struct mount *mp, struct quotakcursor *cursor_u) 959 { 960 struct quotakcursor cursor_k; 961 int error; 962 963 /* ensure any padding bytes are cleared */ 964 memset(&cursor_k, 0, sizeof(cursor_k)); 965 966 error = vfs_quotactl_cursoropen(mp, &cursor_k); 967 if (error) { 968 return error; 969 } 970 971 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 972 } 973 974 static int 975 do_sys_quotactl_cursorclose(struct mount *mp, struct quotakcursor *cursor_u) 976 { 977 struct quotakcursor cursor_k; 978 int error; 979 980 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 981 if (error) { 982 return error; 983 } 984 985 return vfs_quotactl_cursorclose(mp, &cursor_k); 986 } 987 988 static int 989 do_sys_quotactl_cursorskipidtype(struct mount *mp, 990 struct quotakcursor *cursor_u, int idtype) 991 { 992 struct quotakcursor cursor_k; 993 int error; 994 995 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 996 if (error) { 997 return error; 998 } 999 1000 error = vfs_quotactl_cursorskipidtype(mp, &cursor_k, idtype); 1001 if (error) { 1002 return error; 1003 } 1004 1005 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 1006 } 1007 1008 static int 1009 do_sys_quotactl_cursorget(struct mount *mp, struct quotakcursor *cursor_u, 1010 struct quotakey *keys_u, struct quotaval *vals_u, unsigned maxnum, 1011 unsigned *ret_u) 1012 { 1013 #define CGET_STACK_MAX 8 1014 struct quotakcursor cursor_k; 1015 struct quotakey stackkeys[CGET_STACK_MAX]; 1016 struct quotaval stackvals[CGET_STACK_MAX]; 1017 struct quotakey *keys_k; 1018 struct quotaval *vals_k; 1019 unsigned ret_k; 1020 int error; 1021 1022 if (maxnum > 128) { 1023 maxnum = 128; 1024 } 1025 1026 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 1027 if (error) { 1028 return error; 1029 } 1030 1031 if (maxnum <= CGET_STACK_MAX) { 1032 keys_k = stackkeys; 1033 vals_k = stackvals; 1034 /* ensure any padding bytes are cleared */ 1035 memset(keys_k, 0, maxnum * sizeof(keys_k[0])); 1036 memset(vals_k, 0, maxnum * sizeof(vals_k[0])); 1037 } else { 1038 keys_k = kmem_zalloc(maxnum * sizeof(keys_k[0]), KM_SLEEP); 1039 vals_k = kmem_zalloc(maxnum * sizeof(vals_k[0]), KM_SLEEP); 1040 } 1041 1042 error = vfs_quotactl_cursorget(mp, &cursor_k, keys_k, vals_k, maxnum, 1043 &ret_k); 1044 if (error) { 1045 goto fail; 1046 } 1047 1048 error = copyout(keys_k, keys_u, ret_k * sizeof(keys_k[0])); 1049 if (error) { 1050 goto fail; 1051 } 1052 1053 error = copyout(vals_k, vals_u, ret_k * sizeof(vals_k[0])); 1054 if (error) { 1055 goto fail; 1056 } 1057 1058 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 1059 if (error) { 1060 goto fail; 1061 } 1062 1063 /* do last to maximize the chance of being able to recover a failure */ 1064 error = copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 1065 1066 fail: 1067 if (keys_k != stackkeys) { 1068 kmem_free(keys_k, maxnum * sizeof(keys_k[0])); 1069 } 1070 if (vals_k != stackvals) { 1071 kmem_free(vals_k, maxnum * sizeof(vals_k[0])); 1072 } 1073 return error; 1074 } 1075 1076 static int 1077 do_sys_quotactl_cursoratend(struct mount *mp, struct quotakcursor *cursor_u, 1078 int *ret_u) 1079 { 1080 struct quotakcursor cursor_k; 1081 int ret_k; 1082 int error; 1083 1084 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 1085 if (error) { 1086 return error; 1087 } 1088 1089 error = vfs_quotactl_cursoratend(mp, &cursor_k, &ret_k); 1090 if (error) { 1091 return error; 1092 } 1093 1094 error = copyout(&ret_k, ret_u, sizeof(ret_k)); 1095 if (error) { 1096 return error; 1097 } 1098 1099 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 1100 } 1101 1102 static int 1103 do_sys_quotactl_cursorrewind(struct mount *mp, struct quotakcursor *cursor_u) 1104 { 1105 struct quotakcursor cursor_k; 1106 int error; 1107 1108 error = copyin(cursor_u, &cursor_k, sizeof(cursor_k)); 1109 if (error) { 1110 return error; 1111 } 1112 1113 error = vfs_quotactl_cursorrewind(mp, &cursor_k); 1114 if (error) { 1115 return error; 1116 } 1117 1118 return copyout(&cursor_k, cursor_u, sizeof(cursor_k)); 1119 } 1120 1121 static int 1122 do_sys_quotactl_quotaon(struct mount *mp, int idtype, const char *path_u) 1123 { 1124 char *path_k; 1125 int error; 1126 1127 /* XXX this should probably be a struct pathbuf */ 1128 path_k = PNBUF_GET(); 1129 error = copyin(path_u, path_k, PATH_MAX); 1130 if (error) { 1131 PNBUF_PUT(path_k); 1132 return error; 1133 } 1134 1135 error = vfs_quotactl_quotaon(mp, idtype, path_k); 1136 1137 PNBUF_PUT(path_k); 1138 return error; 1139 } 1140 1141 static int 1142 do_sys_quotactl_quotaoff(struct mount *mp, int idtype) 1143 { 1144 return vfs_quotactl_quotaoff(mp, idtype); 1145 } 1146 1147 int 1148 do_sys_quotactl(const char *path_u, const struct quotactl_args *args) 1149 { 1150 struct mount *mp; 1151 struct vnode *vp; 1152 int error; 1153 1154 error = namei_simple_user(path_u, NSM_FOLLOW_TRYEMULROOT, &vp); 1155 if (error != 0) 1156 return (error); 1157 mp = vp->v_mount; 1158 1159 switch (args->qc_op) { 1160 case QUOTACTL_STAT: 1161 error = do_sys_quotactl_stat(mp, args->u.stat.qc_info); 1162 break; 1163 case QUOTACTL_IDTYPESTAT: 1164 error = do_sys_quotactl_idtypestat(mp, 1165 args->u.idtypestat.qc_idtype, 1166 args->u.idtypestat.qc_info); 1167 break; 1168 case QUOTACTL_OBJTYPESTAT: 1169 error = do_sys_quotactl_objtypestat(mp, 1170 args->u.objtypestat.qc_objtype, 1171 args->u.objtypestat.qc_info); 1172 break; 1173 case QUOTACTL_GET: 1174 error = do_sys_quotactl_get(mp, 1175 args->u.get.qc_key, 1176 args->u.get.qc_val); 1177 break; 1178 case QUOTACTL_PUT: 1179 error = do_sys_quotactl_put(mp, 1180 args->u.put.qc_key, 1181 args->u.put.qc_val); 1182 break; 1183 case QUOTACTL_DEL: 1184 error = do_sys_quotactl_del(mp, args->u.del.qc_key); 1185 break; 1186 case QUOTACTL_CURSOROPEN: 1187 error = do_sys_quotactl_cursoropen(mp, 1188 args->u.cursoropen.qc_cursor); 1189 break; 1190 case QUOTACTL_CURSORCLOSE: 1191 error = do_sys_quotactl_cursorclose(mp, 1192 args->u.cursorclose.qc_cursor); 1193 break; 1194 case QUOTACTL_CURSORSKIPIDTYPE: 1195 error = do_sys_quotactl_cursorskipidtype(mp, 1196 args->u.cursorskipidtype.qc_cursor, 1197 args->u.cursorskipidtype.qc_idtype); 1198 break; 1199 case QUOTACTL_CURSORGET: 1200 error = do_sys_quotactl_cursorget(mp, 1201 args->u.cursorget.qc_cursor, 1202 args->u.cursorget.qc_keys, 1203 args->u.cursorget.qc_vals, 1204 args->u.cursorget.qc_maxnum, 1205 args->u.cursorget.qc_ret); 1206 break; 1207 case QUOTACTL_CURSORATEND: 1208 error = do_sys_quotactl_cursoratend(mp, 1209 args->u.cursoratend.qc_cursor, 1210 args->u.cursoratend.qc_ret); 1211 break; 1212 case QUOTACTL_CURSORREWIND: 1213 error = do_sys_quotactl_cursorrewind(mp, 1214 args->u.cursorrewind.qc_cursor); 1215 break; 1216 case QUOTACTL_QUOTAON: 1217 error = do_sys_quotactl_quotaon(mp, 1218 args->u.quotaon.qc_idtype, 1219 args->u.quotaon.qc_quotafile); 1220 break; 1221 case QUOTACTL_QUOTAOFF: 1222 error = do_sys_quotactl_quotaoff(mp, 1223 args->u.quotaoff.qc_idtype); 1224 break; 1225 default: 1226 error = EINVAL; 1227 break; 1228 } 1229 1230 vrele(vp); 1231 return error; 1232 } 1233 1234 /* ARGSUSED */ 1235 int 1236 sys___quotactl(struct lwp *l, const struct sys___quotactl_args *uap, 1237 register_t *retval) 1238 { 1239 /* { 1240 syscallarg(const char *) path; 1241 syscallarg(struct quotactl_args *) args; 1242 } */ 1243 struct quotactl_args args; 1244 int error; 1245 1246 error = copyin(SCARG(uap, args), &args, sizeof(args)); 1247 if (error) { 1248 return error; 1249 } 1250 1251 return do_sys_quotactl(SCARG(uap, path), &args); 1252 } 1253 1254 int 1255 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 1256 int root) 1257 { 1258 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 1259 bool chrooted; 1260 int error = 0; 1261 1262 KASSERT(l == curlwp); 1263 1264 /* 1265 * This is safe unlocked. cwdi_rdir never goes non-NULL -> NULL, 1266 * since it would imply chroots can be escaped. Just make sure this 1267 * routine is self-consistent. 1268 */ 1269 chrooted = (atomic_load_relaxed(&cwdi->cwdi_rdir) != NULL); 1270 1271 /* 1272 * If MNT_NOWAIT or MNT_LAZY is specified, do not 1273 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 1274 * overrides MNT_NOWAIT. 1275 */ 1276 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 1277 (flags != MNT_WAIT && flags != 0)) { 1278 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 1279 } else { 1280 /* Get the filesystem stats now */ 1281 memset(sp, 0, sizeof(*sp)); 1282 if ((error = VFS_STATVFS(mp, sp)) != 0) 1283 return error; 1284 if (!chrooted) 1285 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 1286 } 1287 1288 if (chrooted) { 1289 size_t len; 1290 char *bp; 1291 char c; 1292 char *path = PNBUF_GET(); 1293 1294 bp = path + MAXPATHLEN; 1295 *--bp = '\0'; 1296 rw_enter(&cwdi->cwdi_lock, RW_READER); 1297 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 1298 MAXPATHLEN / 2, 0, l); 1299 rw_exit(&cwdi->cwdi_lock); 1300 if (error) { 1301 PNBUF_PUT(path); 1302 return error; 1303 } 1304 len = strlen(bp); 1305 if (len != 1) { 1306 /* 1307 * for mount points that are below our root, we can see 1308 * them, so we fix up the pathname and return them. The 1309 * rest we cannot see, so we don't allow viewing the 1310 * data. 1311 */ 1312 if (strncmp(bp, sp->f_mntonname, len) == 0 && 1313 ((c = sp->f_mntonname[len]) == '/' || c == '\0')) { 1314 (void)strlcpy(sp->f_mntonname, 1315 c == '\0' ? "/" : &sp->f_mntonname[len], 1316 sizeof(sp->f_mntonname)); 1317 } else { 1318 if (root) 1319 (void)strlcpy(sp->f_mntonname, "/", 1320 sizeof(sp->f_mntonname)); 1321 else 1322 error = EPERM; 1323 } 1324 } 1325 PNBUF_PUT(path); 1326 } 1327 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 1328 return error; 1329 } 1330 1331 /* 1332 * Get filesystem statistics by path. 1333 */ 1334 int 1335 do_sys_pstatvfs(struct lwp *l, const char *path, int flags, struct statvfs *sb) 1336 { 1337 struct mount *mp; 1338 int error; 1339 struct vnode *vp; 1340 1341 error = namei_simple_user(path, NSM_FOLLOW_TRYEMULROOT, &vp); 1342 if (error != 0) 1343 return error; 1344 mp = vp->v_mount; 1345 error = dostatvfs(mp, sb, l, flags, 1); 1346 vrele(vp); 1347 return error; 1348 } 1349 1350 /* ARGSUSED */ 1351 int 1352 sys___statvfs190(struct lwp *l, const struct sys___statvfs190_args *uap, register_t *retval) 1353 { 1354 /* { 1355 syscallarg(const char *) path; 1356 syscallarg(struct statvfs *) buf; 1357 syscallarg(int) flags; 1358 } */ 1359 struct statvfs *sb; 1360 int error; 1361 1362 sb = STATVFSBUF_GET(); 1363 error = do_sys_pstatvfs(l, SCARG(uap, path), SCARG(uap, flags), sb); 1364 if (error == 0) 1365 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1366 STATVFSBUF_PUT(sb); 1367 return error; 1368 } 1369 1370 /* 1371 * Get filesystem statistics by fd. 1372 */ 1373 int 1374 do_sys_fstatvfs(struct lwp *l, int fd, int flags, struct statvfs *sb) 1375 { 1376 file_t *fp; 1377 struct mount *mp; 1378 int error; 1379 1380 /* fd_getvnode() will use the descriptor for us */ 1381 if ((error = fd_getvnode(fd, &fp)) != 0) 1382 return (error); 1383 mp = fp->f_vnode->v_mount; 1384 error = dostatvfs(mp, sb, curlwp, flags, 1); 1385 fd_putfile(fd); 1386 return error; 1387 } 1388 1389 /* ARGSUSED */ 1390 int 1391 sys___fstatvfs190(struct lwp *l, const struct sys___fstatvfs190_args *uap, register_t *retval) 1392 { 1393 /* { 1394 syscallarg(int) fd; 1395 syscallarg(struct statvfs *) buf; 1396 syscallarg(int) flags; 1397 } */ 1398 struct statvfs *sb; 1399 int error; 1400 1401 sb = STATVFSBUF_GET(); 1402 error = do_sys_fstatvfs(l, SCARG(uap, fd), SCARG(uap, flags), sb); 1403 if (error == 0) 1404 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1405 STATVFSBUF_PUT(sb); 1406 return error; 1407 } 1408 1409 1410 /* 1411 * Get statistics on all filesystems. 1412 */ 1413 int 1414 do_sys_getvfsstat(struct lwp *l, void *sfsp, size_t bufsize, int flags, 1415 int (*copyfn)(const void *, void *, size_t), size_t entry_sz, 1416 register_t *retval) 1417 { 1418 int root = 0; 1419 mount_iterator_t *iter; 1420 struct proc *p = l->l_proc; 1421 struct mount *mp; 1422 struct statvfs *sb; 1423 size_t count, maxcount; 1424 int error = 0; 1425 1426 sb = STATVFSBUF_GET(); 1427 maxcount = bufsize / entry_sz; 1428 count = 0; 1429 mountlist_iterator_init(&iter); 1430 while ((mp = mountlist_iterator_next(iter)) != NULL) { 1431 if (sfsp && count < maxcount) { 1432 error = dostatvfs(mp, sb, l, flags, 0); 1433 if (error) { 1434 error = 0; 1435 continue; 1436 } 1437 error = copyfn(sb, sfsp, entry_sz); 1438 if (error) 1439 goto out; 1440 sfsp = (char *)sfsp + entry_sz; 1441 root |= strcmp(sb->f_mntonname, "/") == 0; 1442 } 1443 count++; 1444 } 1445 1446 if (root == 0 && p->p_cwdi->cwdi_rdir) { 1447 /* 1448 * fake a root entry 1449 */ 1450 error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, 1451 sb, l, flags, 1); 1452 if (error != 0) 1453 goto out; 1454 if (sfsp) { 1455 error = copyfn(sb, sfsp, entry_sz); 1456 if (error != 0) 1457 goto out; 1458 } 1459 count++; 1460 } 1461 if (sfsp && count > maxcount) 1462 *retval = maxcount; 1463 else 1464 *retval = count; 1465 out: 1466 mountlist_iterator_destroy(iter); 1467 STATVFSBUF_PUT(sb); 1468 return error; 1469 } 1470 1471 int 1472 sys___getvfsstat90(struct lwp *l, const struct sys___getvfsstat90_args *uap, 1473 register_t *retval) 1474 { 1475 /* { 1476 syscallarg(struct statvfs *) buf; 1477 syscallarg(size_t) bufsize; 1478 syscallarg(int) flags; 1479 } */ 1480 1481 return do_sys_getvfsstat(l, SCARG(uap, buf), SCARG(uap, bufsize), 1482 SCARG(uap, flags), copyout, sizeof (struct statvfs), retval); 1483 } 1484 1485 /* 1486 * Change current working directory to a given file descriptor. 1487 */ 1488 int 1489 do_sys_fchdir(struct lwp *l, int fd, register_t *retval) 1490 { 1491 struct proc *p = l->l_proc; 1492 struct cwdinfo *cwdi; 1493 struct vnode *vp, *tdp; 1494 struct mount *mp; 1495 file_t *fp; 1496 int error; 1497 1498 /* fd_getvnode() will use the descriptor for us */ 1499 if ((error = fd_getvnode(fd, &fp)) != 0) 1500 return error; 1501 vp = fp->f_vnode; 1502 1503 vref(vp); 1504 vn_lock(vp, LK_SHARED | LK_RETRY); 1505 if (vp->v_type != VDIR) 1506 error = ENOTDIR; 1507 else 1508 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1509 if (error) { 1510 vput(vp); 1511 goto out; 1512 } 1513 while ((mp = vp->v_mountedhere) != NULL) { 1514 error = vfs_busy(mp); 1515 vput(vp); 1516 if (error != 0) 1517 goto out; 1518 error = VFS_ROOT(mp, LK_SHARED, &tdp); 1519 vfs_unbusy(mp); 1520 if (error) 1521 goto out; 1522 vp = tdp; 1523 } 1524 VOP_UNLOCK(vp); 1525 1526 /* 1527 * Disallow changing to a directory not under the process's 1528 * current root directory (if there is one). 1529 */ 1530 cwdi = p->p_cwdi; 1531 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1532 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1533 vrele(vp); 1534 error = EPERM; /* operation not permitted */ 1535 } else { 1536 vrele(cwdi->cwdi_cdir); 1537 cwdi->cwdi_cdir = vp; 1538 } 1539 rw_exit(&cwdi->cwdi_lock); 1540 1541 out: 1542 fd_putfile(fd); 1543 return error; 1544 } 1545 1546 /* 1547 * Change current working directory to a given file descriptor. 1548 */ 1549 /* ARGSUSED */ 1550 int 1551 sys_fchdir(struct lwp *l, const struct sys_fchdir_args *uap, register_t *retval) 1552 { 1553 /* { 1554 syscallarg(int) fd; 1555 } */ 1556 return do_sys_fchdir(l, SCARG(uap, fd), retval); 1557 } 1558 1559 /* 1560 * Change this process's notion of the root directory to a given file 1561 * descriptor. 1562 */ 1563 int 1564 sys_fchroot(struct lwp *l, const struct sys_fchroot_args *uap, register_t *retval) 1565 { 1566 struct vnode *vp; 1567 file_t *fp; 1568 int error, fd = SCARG(uap, fd); 1569 1570 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1571 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1572 return error; 1573 /* fd_getvnode() will use the descriptor for us */ 1574 if ((error = fd_getvnode(fd, &fp)) != 0) 1575 return error; 1576 vp = fp->f_vnode; 1577 vn_lock(vp, LK_SHARED | LK_RETRY); 1578 if (vp->v_type != VDIR) 1579 error = ENOTDIR; 1580 else 1581 error = VOP_ACCESS(vp, VEXEC, l->l_cred); 1582 VOP_UNLOCK(vp); 1583 if (error) 1584 goto out; 1585 vref(vp); 1586 change_root(vp); 1587 1588 out: 1589 fd_putfile(fd); 1590 return (error); 1591 } 1592 1593 /* 1594 * Change current working directory (``.''). 1595 */ 1596 int 1597 do_sys_chdir(struct lwp *l, const char *path, enum uio_seg seg, 1598 register_t *retval) 1599 { 1600 struct proc *p = l->l_proc; 1601 struct cwdinfo * cwdi; 1602 int error; 1603 struct vnode *vp; 1604 1605 if ((error = chdir_lookup(path, seg, &vp, l)) != 0) 1606 return error; 1607 cwdi = p->p_cwdi; 1608 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1609 vrele(cwdi->cwdi_cdir); 1610 cwdi->cwdi_cdir = vp; 1611 rw_exit(&cwdi->cwdi_lock); 1612 return 0; 1613 } 1614 1615 /* 1616 * Change current working directory (``.''). 1617 */ 1618 /* ARGSUSED */ 1619 int 1620 sys_chdir(struct lwp *l, const struct sys_chdir_args *uap, register_t *retval) 1621 { 1622 /* { 1623 syscallarg(const char *) path; 1624 } */ 1625 return do_sys_chdir(l, SCARG(uap, path), UIO_USERSPACE, retval); 1626 } 1627 1628 /* 1629 * Change notion of root (``/'') directory. 1630 */ 1631 /* ARGSUSED */ 1632 int 1633 sys_chroot(struct lwp *l, const struct sys_chroot_args *uap, register_t *retval) 1634 { 1635 /* { 1636 syscallarg(const char *) path; 1637 } */ 1638 int error; 1639 struct vnode *vp; 1640 1641 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1642 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1643 return (error); 1644 1645 error = chdir_lookup(SCARG(uap, path), UIO_USERSPACE, &vp, l); 1646 if (error == 0) 1647 change_root(vp); 1648 return error; 1649 } 1650 1651 /* 1652 * Common routine for chroot and fchroot. 1653 * NB: callers need to properly authorize the change root operation. 1654 */ 1655 void 1656 change_root(struct vnode *vp) 1657 { 1658 kauth_cred_t ncred; 1659 struct lwp *l = curlwp; 1660 struct proc *p = l->l_proc; 1661 struct cwdinfo *cwdi = p->p_cwdi; 1662 1663 ncred = kauth_cred_alloc(); 1664 1665 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1666 if (cwdi->cwdi_rdir != NULL) 1667 vrele(cwdi->cwdi_rdir); 1668 cwdi->cwdi_rdir = vp; 1669 1670 /* 1671 * Prevent escaping from chroot by putting the root under 1672 * the working directory. Silently chdir to / if we aren't 1673 * already there. 1674 */ 1675 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1676 /* 1677 * XXX would be more failsafe to change directory to a 1678 * deadfs node here instead 1679 */ 1680 vrele(cwdi->cwdi_cdir); 1681 vref(vp); 1682 cwdi->cwdi_cdir = vp; 1683 } 1684 rw_exit(&cwdi->cwdi_lock); 1685 1686 /* Get a write lock on the process credential. */ 1687 proc_crmod_enter(); 1688 1689 kauth_cred_clone(p->p_cred, ncred); 1690 kauth_proc_chroot(ncred, p->p_cwdi); 1691 1692 /* Broadcast our credentials to the process and other LWPs. */ 1693 proc_crmod_leave(ncred, p->p_cred, true); 1694 } 1695 1696 /* 1697 * Common routine for chroot and chdir. 1698 * XXX "where" should be enum uio_seg 1699 */ 1700 int 1701 chdir_lookup(const char *path, int where, struct vnode **vpp, struct lwp *l) 1702 { 1703 struct pathbuf *pb; 1704 struct nameidata nd; 1705 int error; 1706 1707 error = pathbuf_maybe_copyin(path, where, &pb); 1708 if (error) { 1709 return error; 1710 } 1711 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT, pb); 1712 if ((error = namei(&nd)) != 0) { 1713 pathbuf_destroy(pb); 1714 return error; 1715 } 1716 *vpp = nd.ni_vp; 1717 pathbuf_destroy(pb); 1718 1719 if ((*vpp)->v_type != VDIR) 1720 error = ENOTDIR; 1721 else 1722 error = VOP_ACCESS(*vpp, VEXEC, l->l_cred); 1723 1724 if (error) 1725 vput(*vpp); 1726 else 1727 VOP_UNLOCK(*vpp); 1728 return (error); 1729 } 1730 1731 /* 1732 * Internals of sys_open - path has already been converted into a pathbuf 1733 * (so we can easily reuse this function from other parts of the kernel, 1734 * like posix_spawn post-processing). 1735 */ 1736 int 1737 do_open(lwp_t *l, struct vnode *dvp, struct pathbuf *pb, int open_flags, 1738 int open_mode, int *fd) 1739 { 1740 struct proc *p = l->l_proc; 1741 struct cwdinfo *cwdi = p->p_cwdi; 1742 file_t *fp; 1743 struct vnode *vp; 1744 int dupfd; 1745 bool dupfd_move; 1746 int flags, cmode; 1747 int indx, error; 1748 1749 if (open_flags & O_SEARCH) { 1750 open_flags &= ~(int)O_SEARCH; 1751 } 1752 1753 /* 1754 * Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags 1755 * may be specified. 1756 */ 1757 if ((open_flags & O_EXEC) && (open_flags & O_ACCMODE)) 1758 return EINVAL; 1759 1760 flags = FFLAGS(open_flags); 1761 if ((flags & (FREAD | FWRITE)) == 0) 1762 return EINVAL; 1763 1764 if ((error = fd_allocfile(&fp, &indx)) != 0) { 1765 return error; 1766 } 1767 1768 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1769 cmode = ((open_mode &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1770 1771 error = vn_open(dvp, pb, TRYEMULROOT, flags, cmode, 1772 &vp, &dupfd_move, &dupfd); 1773 if (error != 0) { 1774 fd_abort(p, fp, indx); 1775 return error; 1776 } 1777 1778 if (vp == NULL) { 1779 fd_abort(p, fp, indx); 1780 error = fd_dupopen(dupfd, dupfd_move, flags, &indx); 1781 if (error) 1782 return error; 1783 *fd = indx; 1784 } else { 1785 error = open_setfp(l, fp, vp, indx, flags); 1786 if (error) 1787 return error; 1788 VOP_UNLOCK(vp); 1789 *fd = indx; 1790 fd_affix(p, fp, indx); 1791 } 1792 1793 return 0; 1794 } 1795 1796 int 1797 fd_open(const char *path, int open_flags, int open_mode, int *fd) 1798 { 1799 struct pathbuf *pb; 1800 int error, oflags; 1801 1802 oflags = FFLAGS(open_flags); 1803 if ((oflags & (FREAD | FWRITE)) == 0) 1804 return EINVAL; 1805 1806 pb = pathbuf_create(path); 1807 if (pb == NULL) 1808 return ENOMEM; 1809 1810 error = do_open(curlwp, NULL, pb, open_flags, open_mode, fd); 1811 pathbuf_destroy(pb); 1812 1813 return error; 1814 } 1815 1816 static int 1817 do_sys_openat(lwp_t *l, int fdat, const char *path, int flags, 1818 int mode, int *fd) 1819 { 1820 file_t *dfp = NULL; 1821 struct vnode *dvp = NULL; 1822 struct pathbuf *pb; 1823 const char *pathstring = NULL; 1824 int error; 1825 1826 if (path == NULL) { 1827 MODULE_HOOK_CALL(vfs_openat_10_hook, (&pb), enosys(), error); 1828 if (error == ENOSYS) 1829 goto no_compat; 1830 if (error) 1831 return error; 1832 } else { 1833 no_compat: 1834 error = pathbuf_copyin(path, &pb); 1835 if (error) 1836 return error; 1837 } 1838 1839 pathstring = pathbuf_stringcopy_get(pb); 1840 1841 /* 1842 * fdat is ignored if: 1843 * 1) if fdat is AT_FDCWD, which means use current directory as base. 1844 * 2) if path is absolute, then fdat is useless. 1845 */ 1846 if (fdat != AT_FDCWD && pathstring[0] != '/') { 1847 /* fd_getvnode() will use the descriptor for us */ 1848 if ((error = fd_getvnode(fdat, &dfp)) != 0) 1849 goto out; 1850 1851 dvp = dfp->f_vnode; 1852 } 1853 1854 error = do_open(l, dvp, pb, flags, mode, fd); 1855 1856 if (dfp != NULL) 1857 fd_putfile(fdat); 1858 out: 1859 pathbuf_stringcopy_put(pb, pathstring); 1860 pathbuf_destroy(pb); 1861 return error; 1862 } 1863 1864 int 1865 sys_open(struct lwp *l, const struct sys_open_args *uap, register_t *retval) 1866 { 1867 /* { 1868 syscallarg(const char *) path; 1869 syscallarg(int) flags; 1870 syscallarg(int) mode; 1871 } */ 1872 int error; 1873 int fd; 1874 1875 error = do_sys_openat(l, AT_FDCWD, SCARG(uap, path), 1876 SCARG(uap, flags), SCARG(uap, mode), &fd); 1877 1878 if (error == 0) 1879 *retval = fd; 1880 1881 return error; 1882 } 1883 1884 int 1885 sys_openat(struct lwp *l, const struct sys_openat_args *uap, register_t *retval) 1886 { 1887 /* { 1888 syscallarg(int) fd; 1889 syscallarg(const char *) path; 1890 syscallarg(int) oflags; 1891 syscallarg(int) mode; 1892 } */ 1893 int error; 1894 int fd; 1895 1896 error = do_sys_openat(l, SCARG(uap, fd), SCARG(uap, path), 1897 SCARG(uap, oflags), SCARG(uap, mode), &fd); 1898 1899 if (error == 0) 1900 *retval = fd; 1901 1902 return error; 1903 } 1904 1905 static void 1906 vfs__fhfree(fhandle_t *fhp) 1907 { 1908 size_t fhsize; 1909 1910 fhsize = FHANDLE_SIZE(fhp); 1911 kmem_free(fhp, fhsize); 1912 } 1913 1914 /* 1915 * vfs_composefh: compose a filehandle. 1916 */ 1917 1918 int 1919 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1920 { 1921 struct mount *mp; 1922 struct fid *fidp; 1923 int error; 1924 size_t needfhsize; 1925 size_t fidsize; 1926 1927 mp = vp->v_mount; 1928 fidp = NULL; 1929 if (*fh_size < FHANDLE_SIZE_MIN) { 1930 fidsize = 0; 1931 } else { 1932 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1933 if (fhp != NULL) { 1934 memset(fhp, 0, *fh_size); 1935 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1936 fidp = &fhp->fh_fid; 1937 } 1938 } 1939 error = VFS_VPTOFH(vp, fidp, &fidsize); 1940 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1941 if (error == 0 && *fh_size < needfhsize) { 1942 error = E2BIG; 1943 } 1944 *fh_size = needfhsize; 1945 return error; 1946 } 1947 1948 int 1949 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1950 { 1951 struct mount *mp; 1952 fhandle_t *fhp; 1953 size_t fhsize; 1954 size_t fidsize; 1955 int error; 1956 1957 mp = vp->v_mount; 1958 fidsize = 0; 1959 error = VFS_VPTOFH(vp, NULL, &fidsize); 1960 KASSERT(error != 0); 1961 if (error != E2BIG) { 1962 goto out; 1963 } 1964 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1965 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1966 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1967 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1968 if (error == 0) { 1969 KASSERT(FHANDLE_SIZE(fhp) == fhsize); 1970 KASSERT(FHANDLE_FILEID(fhp)->fid_len == fidsize); 1971 *fhpp = fhp; 1972 } else { 1973 kmem_free(fhp, fhsize); 1974 } 1975 out: 1976 return error; 1977 } 1978 1979 void 1980 vfs_composefh_free(fhandle_t *fhp) 1981 { 1982 1983 vfs__fhfree(fhp); 1984 } 1985 1986 /* 1987 * vfs_fhtovp: lookup a vnode by a filehandle. 1988 */ 1989 1990 int 1991 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1992 { 1993 struct mount *mp; 1994 int error; 1995 1996 *vpp = NULL; 1997 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1998 if (mp == NULL) { 1999 error = ESTALE; 2000 goto out; 2001 } 2002 if (mp->mnt_op->vfs_fhtovp == NULL) { 2003 error = EOPNOTSUPP; 2004 goto out; 2005 } 2006 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), LK_EXCLUSIVE, vpp); 2007 out: 2008 return error; 2009 } 2010 2011 /* 2012 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 2013 * the needed size. 2014 */ 2015 2016 int 2017 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 2018 { 2019 fhandle_t *fhp; 2020 int error; 2021 2022 if (fhsize > FHANDLE_SIZE_MAX) { 2023 return EINVAL; 2024 } 2025 if (fhsize < FHANDLE_SIZE_MIN) { 2026 return EINVAL; 2027 } 2028 again: 2029 fhp = kmem_alloc(fhsize, KM_SLEEP); 2030 error = copyin(ufhp, fhp, fhsize); 2031 if (error == 0) { 2032 /* XXX this check shouldn't be here */ 2033 if (FHANDLE_SIZE(fhp) == fhsize) { 2034 *fhpp = fhp; 2035 return 0; 2036 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 2037 /* 2038 * a kludge for nfsv2 padded handles. 2039 */ 2040 size_t sz; 2041 2042 sz = FHANDLE_SIZE(fhp); 2043 kmem_free(fhp, fhsize); 2044 fhsize = sz; 2045 goto again; 2046 } else { 2047 /* 2048 * userland told us wrong size. 2049 */ 2050 error = EINVAL; 2051 } 2052 } 2053 kmem_free(fhp, fhsize); 2054 return error; 2055 } 2056 2057 void 2058 vfs_copyinfh_free(fhandle_t *fhp) 2059 { 2060 2061 vfs__fhfree(fhp); 2062 } 2063 2064 /* 2065 * Get file handle system call 2066 */ 2067 int 2068 sys___getfh30(struct lwp *l, const struct sys___getfh30_args *uap, register_t *retval) 2069 { 2070 /* { 2071 syscallarg(char *) fname; 2072 syscallarg(fhandle_t *) fhp; 2073 syscallarg(size_t *) fh_size; 2074 } */ 2075 struct vnode *vp; 2076 fhandle_t *fh; 2077 int error; 2078 struct pathbuf *pb; 2079 struct nameidata nd; 2080 size_t sz; 2081 size_t usz; 2082 2083 /* 2084 * Must be super user 2085 */ 2086 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2087 0, NULL, NULL, NULL); 2088 if (error) 2089 return (error); 2090 2091 error = pathbuf_copyin(SCARG(uap, fname), &pb); 2092 if (error) { 2093 return error; 2094 } 2095 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | TRYEMULROOT, pb); 2096 error = namei(&nd); 2097 if (error) { 2098 pathbuf_destroy(pb); 2099 return error; 2100 } 2101 vp = nd.ni_vp; 2102 pathbuf_destroy(pb); 2103 2104 error = vfs_composefh_alloc(vp, &fh); 2105 vput(vp); 2106 if (error != 0) { 2107 return error; 2108 } 2109 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 2110 if (error != 0) { 2111 goto out; 2112 } 2113 sz = FHANDLE_SIZE(fh); 2114 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 2115 if (error != 0) { 2116 goto out; 2117 } 2118 if (usz >= sz) { 2119 error = copyout(fh, SCARG(uap, fhp), sz); 2120 } else { 2121 error = E2BIG; 2122 } 2123 out: 2124 vfs_composefh_free(fh); 2125 return (error); 2126 } 2127 2128 /* 2129 * Open a file given a file handle. 2130 * 2131 * Check permissions, allocate an open file structure, 2132 * and call the device open routine if any. 2133 */ 2134 2135 int 2136 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 2137 register_t *retval) 2138 { 2139 file_t *fp; 2140 struct vnode *vp = NULL; 2141 kauth_cred_t cred = l->l_cred; 2142 file_t *nfp; 2143 int indx, error; 2144 struct vattr va; 2145 fhandle_t *fh; 2146 int flags; 2147 proc_t *p; 2148 2149 p = curproc; 2150 2151 /* 2152 * Must be super user 2153 */ 2154 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2155 0, NULL, NULL, NULL))) 2156 return (error); 2157 2158 if (oflags & O_SEARCH) { 2159 oflags &= ~(int)O_SEARCH; 2160 } 2161 2162 flags = FFLAGS(oflags); 2163 if ((flags & (FREAD | FWRITE)) == 0) 2164 return (EINVAL); 2165 if ((flags & O_CREAT)) 2166 return (EINVAL); 2167 if ((error = fd_allocfile(&nfp, &indx)) != 0) 2168 return (error); 2169 fp = nfp; 2170 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2171 if (error != 0) { 2172 goto bad; 2173 } 2174 error = vfs_fhtovp(fh, &vp); 2175 vfs_copyinfh_free(fh); 2176 if (error != 0) { 2177 goto bad; 2178 } 2179 2180 /* Now do an effective vn_open */ 2181 2182 if (vp->v_type == VSOCK) { 2183 error = EOPNOTSUPP; 2184 goto bad; 2185 } 2186 error = vn_openchk(vp, cred, flags); 2187 if (error != 0) 2188 goto bad; 2189 if (flags & O_TRUNC) { 2190 VOP_UNLOCK(vp); /* XXX */ 2191 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 2192 vattr_null(&va); 2193 va.va_size = 0; 2194 error = VOP_SETATTR(vp, &va, cred); 2195 if (error) 2196 goto bad; 2197 } 2198 if ((error = VOP_OPEN(vp, flags, cred)) != 0) 2199 goto bad; 2200 if (flags & FWRITE) { 2201 mutex_enter(vp->v_interlock); 2202 vp->v_writecount++; 2203 mutex_exit(vp->v_interlock); 2204 } 2205 2206 /* done with modified vn_open, now finish what sys_open does. */ 2207 if ((error = open_setfp(l, fp, vp, indx, flags))) 2208 return error; 2209 2210 VOP_UNLOCK(vp); 2211 *retval = indx; 2212 fd_affix(p, fp, indx); 2213 return (0); 2214 2215 bad: 2216 fd_abort(p, fp, indx); 2217 if (vp != NULL) 2218 vput(vp); 2219 if (error == EDUPFD || error == EMOVEFD) { 2220 /* XXX should probably close curlwp->l_dupfd */ 2221 error = EOPNOTSUPP; 2222 } 2223 return (error); 2224 } 2225 2226 int 2227 sys___fhopen40(struct lwp *l, const struct sys___fhopen40_args *uap, register_t *retval) 2228 { 2229 /* { 2230 syscallarg(const void *) fhp; 2231 syscallarg(size_t) fh_size; 2232 syscallarg(int) flags; 2233 } */ 2234 2235 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 2236 SCARG(uap, flags), retval); 2237 } 2238 2239 int 2240 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 2241 { 2242 int error; 2243 fhandle_t *fh; 2244 struct vnode *vp; 2245 2246 /* 2247 * Must be super user 2248 */ 2249 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2250 0, NULL, NULL, NULL))) 2251 return (error); 2252 2253 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2254 if (error != 0) 2255 return error; 2256 2257 error = vfs_fhtovp(fh, &vp); 2258 vfs_copyinfh_free(fh); 2259 if (error != 0) 2260 return error; 2261 2262 error = vn_stat(vp, sb); 2263 vput(vp); 2264 return error; 2265 } 2266 2267 2268 /* ARGSUSED */ 2269 int 2270 sys___fhstat50(struct lwp *l, const struct sys___fhstat50_args *uap, register_t *retval) 2271 { 2272 /* { 2273 syscallarg(const void *) fhp; 2274 syscallarg(size_t) fh_size; 2275 syscallarg(struct stat *) sb; 2276 } */ 2277 struct stat sb; 2278 int error; 2279 2280 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 2281 if (error) 2282 return error; 2283 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 2284 } 2285 2286 int 2287 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 2288 int flags) 2289 { 2290 fhandle_t *fh; 2291 struct mount *mp; 2292 struct vnode *vp; 2293 int error; 2294 2295 /* 2296 * Must be super user 2297 */ 2298 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 2299 0, NULL, NULL, NULL))) 2300 return error; 2301 2302 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 2303 if (error != 0) 2304 return error; 2305 2306 error = vfs_fhtovp(fh, &vp); 2307 vfs_copyinfh_free(fh); 2308 if (error != 0) 2309 return error; 2310 2311 mp = vp->v_mount; 2312 error = dostatvfs(mp, sb, l, flags, 1); 2313 vput(vp); 2314 return error; 2315 } 2316 2317 /* ARGSUSED */ 2318 int 2319 sys___fhstatvfs190(struct lwp *l, const struct sys___fhstatvfs190_args *uap, register_t *retval) 2320 { 2321 /* { 2322 syscallarg(const void *) fhp; 2323 syscallarg(size_t) fh_size; 2324 syscallarg(struct statvfs *) buf; 2325 syscallarg(int) flags; 2326 } */ 2327 struct statvfs *sb = STATVFSBUF_GET(); 2328 int error; 2329 2330 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 2331 SCARG(uap, flags)); 2332 if (error == 0) 2333 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 2334 STATVFSBUF_PUT(sb); 2335 return error; 2336 } 2337 2338 int 2339 do_posix_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2340 dev_t dev) 2341 { 2342 2343 /* 2344 * The POSIX mknod(2) call is an alias for mkfifo(2) for S_IFIFO 2345 * in mode and dev=0. 2346 * 2347 * In all the other cases it's implementation defined behavior. 2348 */ 2349 2350 if ((mode & S_IFIFO) && dev == 0) 2351 return do_sys_mkfifoat(l, fdat, pathname, mode); 2352 else 2353 return do_sys_mknodat(l, fdat, pathname, mode, dev, 2354 UIO_USERSPACE); 2355 } 2356 2357 /* 2358 * Create a special file. 2359 */ 2360 /* ARGSUSED */ 2361 int 2362 sys___mknod50(struct lwp *l, const struct sys___mknod50_args *uap, 2363 register_t *retval) 2364 { 2365 /* { 2366 syscallarg(const char *) path; 2367 syscallarg(mode_t) mode; 2368 syscallarg(dev_t) dev; 2369 } */ 2370 return do_posix_mknodat(l, AT_FDCWD, SCARG(uap, path), 2371 SCARG(uap, mode), SCARG(uap, dev)); 2372 } 2373 2374 int 2375 sys_mknodat(struct lwp *l, const struct sys_mknodat_args *uap, 2376 register_t *retval) 2377 { 2378 /* { 2379 syscallarg(int) fd; 2380 syscallarg(const char *) path; 2381 syscallarg(mode_t) mode; 2382 syscallarg(int) pad; 2383 syscallarg(dev_t) dev; 2384 } */ 2385 2386 return do_posix_mknodat(l, SCARG(uap, fd), SCARG(uap, path), 2387 SCARG(uap, mode), SCARG(uap, dev)); 2388 } 2389 2390 int 2391 do_sys_mknod(struct lwp *l, const char *pathname, mode_t mode, dev_t dev, 2392 enum uio_seg seg) 2393 { 2394 return do_sys_mknodat(l, AT_FDCWD, pathname, mode, dev, seg); 2395 } 2396 2397 int 2398 do_sys_mknodat(struct lwp *l, int fdat, const char *pathname, mode_t mode, 2399 dev_t dev, enum uio_seg seg) 2400 { 2401 struct proc *p = l->l_proc; 2402 struct vnode *vp; 2403 struct vattr vattr; 2404 int error, optype; 2405 struct pathbuf *pb; 2406 struct nameidata nd; 2407 const char *pathstring; 2408 2409 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 2410 0, NULL, NULL, NULL)) != 0) 2411 return (error); 2412 2413 optype = VOP_MKNOD_DESCOFFSET; 2414 2415 error = pathbuf_maybe_copyin(pathname, seg, &pb); 2416 if (error) { 2417 return error; 2418 } 2419 pathstring = pathbuf_stringcopy_get(pb); 2420 if (pathstring == NULL) { 2421 pathbuf_destroy(pb); 2422 return ENOMEM; 2423 } 2424 2425 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2426 2427 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2428 goto out; 2429 vp = nd.ni_vp; 2430 2431 if (vp != NULL) 2432 error = EEXIST; 2433 else { 2434 vattr_null(&vattr); 2435 /* We will read cwdi->cwdi_cmask unlocked. */ 2436 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2437 vattr.va_rdev = dev; 2438 2439 switch (mode & S_IFMT) { 2440 case S_IFMT: /* used by badsect to flag bad sectors */ 2441 vattr.va_type = VBAD; 2442 break; 2443 case S_IFCHR: 2444 vattr.va_type = VCHR; 2445 break; 2446 case S_IFBLK: 2447 vattr.va_type = VBLK; 2448 break; 2449 case S_IFWHT: 2450 optype = VOP_WHITEOUT_DESCOFFSET; 2451 break; 2452 case S_IFREG: 2453 #if NVERIEXEC > 0 2454 error = veriexec_openchk(l, nd.ni_vp, pathstring, 2455 O_CREAT); 2456 #endif /* NVERIEXEC > 0 */ 2457 vattr.va_type = VREG; 2458 vattr.va_rdev = VNOVAL; 2459 optype = VOP_CREATE_DESCOFFSET; 2460 break; 2461 default: 2462 error = EINVAL; 2463 break; 2464 } 2465 2466 if (error == 0 && optype == VOP_MKNOD_DESCOFFSET && 2467 vattr.va_rdev == VNOVAL) 2468 error = EINVAL; 2469 } 2470 2471 if (!error) { 2472 switch (optype) { 2473 case VOP_WHITEOUT_DESCOFFSET: 2474 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 2475 if (error) 2476 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2477 vput(nd.ni_dvp); 2478 break; 2479 2480 case VOP_MKNOD_DESCOFFSET: 2481 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 2482 &nd.ni_cnd, &vattr); 2483 if (error == 0) 2484 vrele(nd.ni_vp); 2485 vput(nd.ni_dvp); 2486 break; 2487 2488 case VOP_CREATE_DESCOFFSET: 2489 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 2490 &nd.ni_cnd, &vattr); 2491 if (error == 0) 2492 vrele(nd.ni_vp); 2493 vput(nd.ni_dvp); 2494 break; 2495 } 2496 } else { 2497 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2498 if (nd.ni_dvp == vp) 2499 vrele(nd.ni_dvp); 2500 else 2501 vput(nd.ni_dvp); 2502 if (vp) 2503 vrele(vp); 2504 } 2505 out: 2506 pathbuf_stringcopy_put(pb, pathstring); 2507 pathbuf_destroy(pb); 2508 return (error); 2509 } 2510 2511 /* 2512 * Create a named pipe. 2513 */ 2514 /* ARGSUSED */ 2515 int 2516 sys_mkfifo(struct lwp *l, const struct sys_mkfifo_args *uap, register_t *retval) 2517 { 2518 /* { 2519 syscallarg(const char *) path; 2520 syscallarg(int) mode; 2521 } */ 2522 return do_sys_mkfifoat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap, mode)); 2523 } 2524 2525 int 2526 sys_mkfifoat(struct lwp *l, const struct sys_mkfifoat_args *uap, 2527 register_t *retval) 2528 { 2529 /* { 2530 syscallarg(int) fd; 2531 syscallarg(const char *) path; 2532 syscallarg(int) mode; 2533 } */ 2534 2535 return do_sys_mkfifoat(l, SCARG(uap, fd), SCARG(uap, path), 2536 SCARG(uap, mode)); 2537 } 2538 2539 static int 2540 do_sys_mkfifoat(struct lwp *l, int fdat, const char *path, mode_t mode) 2541 { 2542 struct proc *p = l->l_proc; 2543 struct vattr vattr; 2544 int error; 2545 struct pathbuf *pb; 2546 struct nameidata nd; 2547 2548 error = pathbuf_copyin(path, &pb); 2549 if (error) { 2550 return error; 2551 } 2552 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, pb); 2553 2554 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 2555 pathbuf_destroy(pb); 2556 return error; 2557 } 2558 if (nd.ni_vp != NULL) { 2559 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2560 if (nd.ni_dvp == nd.ni_vp) 2561 vrele(nd.ni_dvp); 2562 else 2563 vput(nd.ni_dvp); 2564 vrele(nd.ni_vp); 2565 pathbuf_destroy(pb); 2566 return (EEXIST); 2567 } 2568 vattr_null(&vattr); 2569 vattr.va_type = VFIFO; 2570 /* We will read cwdi->cwdi_cmask unlocked. */ 2571 vattr.va_mode = (mode & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 2572 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 2573 if (error == 0) 2574 vrele(nd.ni_vp); 2575 vput(nd.ni_dvp); 2576 pathbuf_destroy(pb); 2577 return (error); 2578 } 2579 2580 /* 2581 * Make a hard file link. 2582 */ 2583 /* ARGSUSED */ 2584 int 2585 do_sys_linkat(struct lwp *l, int fdpath, const char *path, int fdlink, 2586 const char *link, int follow, register_t *retval) 2587 { 2588 struct vnode *vp; 2589 struct pathbuf *linkpb; 2590 struct nameidata nd; 2591 namei_simple_flags_t ns_flags; 2592 int error; 2593 2594 if (follow & AT_SYMLINK_FOLLOW) 2595 ns_flags = NSM_FOLLOW_TRYEMULROOT; 2596 else 2597 ns_flags = NSM_NOFOLLOW_TRYEMULROOT; 2598 2599 error = fd_nameiat_simple_user(l, fdpath, path, ns_flags, &vp); 2600 if (error != 0) 2601 return (error); 2602 error = pathbuf_copyin(link, &linkpb); 2603 if (error) { 2604 goto out1; 2605 } 2606 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2607 if ((error = fd_nameiat(l, fdlink, &nd)) != 0) 2608 goto out2; 2609 if (nd.ni_vp) { 2610 error = EEXIST; 2611 goto abortop; 2612 } 2613 /* Prevent hard links on directories. */ 2614 if (vp->v_type == VDIR) { 2615 error = EPERM; 2616 goto abortop; 2617 } 2618 /* Prevent cross-mount operation. */ 2619 if (nd.ni_dvp->v_mount != vp->v_mount) { 2620 error = EXDEV; 2621 goto abortop; 2622 } 2623 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 2624 VOP_UNLOCK(nd.ni_dvp); 2625 vrele(nd.ni_dvp); 2626 out2: 2627 pathbuf_destroy(linkpb); 2628 out1: 2629 vrele(vp); 2630 return (error); 2631 abortop: 2632 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2633 if (nd.ni_dvp == nd.ni_vp) 2634 vrele(nd.ni_dvp); 2635 else 2636 vput(nd.ni_dvp); 2637 if (nd.ni_vp != NULL) 2638 vrele(nd.ni_vp); 2639 goto out2; 2640 } 2641 2642 int 2643 sys_link(struct lwp *l, const struct sys_link_args *uap, register_t *retval) 2644 { 2645 /* { 2646 syscallarg(const char *) path; 2647 syscallarg(const char *) link; 2648 } */ 2649 const char *path = SCARG(uap, path); 2650 const char *link = SCARG(uap, link); 2651 2652 return do_sys_linkat(l, AT_FDCWD, path, AT_FDCWD, link, 2653 AT_SYMLINK_FOLLOW, retval); 2654 } 2655 2656 int 2657 sys_linkat(struct lwp *l, const struct sys_linkat_args *uap, 2658 register_t *retval) 2659 { 2660 /* { 2661 syscallarg(int) fd1; 2662 syscallarg(const char *) name1; 2663 syscallarg(int) fd2; 2664 syscallarg(const char *) name2; 2665 syscallarg(int) flags; 2666 } */ 2667 int fd1 = SCARG(uap, fd1); 2668 const char *name1 = SCARG(uap, name1); 2669 int fd2 = SCARG(uap, fd2); 2670 const char *name2 = SCARG(uap, name2); 2671 int follow; 2672 2673 follow = SCARG(uap, flags) & AT_SYMLINK_FOLLOW; 2674 2675 return do_sys_linkat(l, fd1, name1, fd2, name2, follow, retval); 2676 } 2677 2678 2679 int 2680 do_sys_symlink(const char *patharg, const char *link, enum uio_seg seg) 2681 { 2682 return do_sys_symlinkat(NULL, patharg, AT_FDCWD, link, seg); 2683 } 2684 2685 static int 2686 do_sys_symlinkat(struct lwp *l, const char *patharg, int fdat, 2687 const char *link, enum uio_seg seg) 2688 { 2689 struct proc *p = curproc; 2690 struct vattr vattr; 2691 char *path; 2692 int error; 2693 size_t len; 2694 struct pathbuf *linkpb; 2695 struct nameidata nd; 2696 2697 KASSERT(l != NULL || fdat == AT_FDCWD); 2698 2699 path = PNBUF_GET(); 2700 if (seg == UIO_USERSPACE) { 2701 if ((error = copyinstr(patharg, path, MAXPATHLEN, &len)) != 0) 2702 goto out1; 2703 if ((error = pathbuf_copyin(link, &linkpb)) != 0) 2704 goto out1; 2705 } else { 2706 len = strlen(patharg) + 1; 2707 KASSERT(len <= MAXPATHLEN); 2708 memcpy(path, patharg, len); 2709 linkpb = pathbuf_create(link); 2710 if (linkpb == NULL) { 2711 error = ENOMEM; 2712 goto out1; 2713 } 2714 } 2715 ktrkuser("symlink-target", path, len - 1); 2716 2717 NDINIT(&nd, CREATE, LOCKPARENT | TRYEMULROOT, linkpb); 2718 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2719 goto out2; 2720 if (nd.ni_vp) { 2721 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2722 if (nd.ni_dvp == nd.ni_vp) 2723 vrele(nd.ni_dvp); 2724 else 2725 vput(nd.ni_dvp); 2726 vrele(nd.ni_vp); 2727 error = EEXIST; 2728 goto out2; 2729 } 2730 vattr_null(&vattr); 2731 vattr.va_type = VLNK; 2732 /* We will read cwdi->cwdi_cmask unlocked. */ 2733 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2734 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2735 if (error == 0) 2736 vrele(nd.ni_vp); 2737 vput(nd.ni_dvp); 2738 out2: 2739 pathbuf_destroy(linkpb); 2740 out1: 2741 PNBUF_PUT(path); 2742 return (error); 2743 } 2744 2745 /* 2746 * Make a symbolic link. 2747 */ 2748 /* ARGSUSED */ 2749 int 2750 sys_symlink(struct lwp *l, const struct sys_symlink_args *uap, register_t *retval) 2751 { 2752 /* { 2753 syscallarg(const char *) path; 2754 syscallarg(const char *) link; 2755 } */ 2756 2757 return do_sys_symlinkat(l, SCARG(uap, path), AT_FDCWD, SCARG(uap, link), 2758 UIO_USERSPACE); 2759 } 2760 2761 int 2762 sys_symlinkat(struct lwp *l, const struct sys_symlinkat_args *uap, 2763 register_t *retval) 2764 { 2765 /* { 2766 syscallarg(const char *) path1; 2767 syscallarg(int) fd; 2768 syscallarg(const char *) path2; 2769 } */ 2770 2771 return do_sys_symlinkat(l, SCARG(uap, path1), SCARG(uap, fd), 2772 SCARG(uap, path2), UIO_USERSPACE); 2773 } 2774 2775 /* 2776 * Delete a whiteout from the filesystem. 2777 */ 2778 /* ARGSUSED */ 2779 int 2780 sys_undelete(struct lwp *l, const struct sys_undelete_args *uap, register_t *retval) 2781 { 2782 /* { 2783 syscallarg(const char *) path; 2784 } */ 2785 int error; 2786 struct pathbuf *pb; 2787 struct nameidata nd; 2788 2789 error = pathbuf_copyin(SCARG(uap, path), &pb); 2790 if (error) { 2791 return error; 2792 } 2793 2794 NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | TRYEMULROOT, pb); 2795 error = namei(&nd); 2796 if (error) { 2797 pathbuf_destroy(pb); 2798 return (error); 2799 } 2800 2801 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2802 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2803 if (nd.ni_dvp == nd.ni_vp) 2804 vrele(nd.ni_dvp); 2805 else 2806 vput(nd.ni_dvp); 2807 if (nd.ni_vp) 2808 vrele(nd.ni_vp); 2809 pathbuf_destroy(pb); 2810 return (EEXIST); 2811 } 2812 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2813 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2814 vput(nd.ni_dvp); 2815 pathbuf_destroy(pb); 2816 return (error); 2817 } 2818 2819 /* 2820 * Delete a name from the filesystem. 2821 */ 2822 /* ARGSUSED */ 2823 int 2824 sys_unlink(struct lwp *l, const struct sys_unlink_args *uap, register_t *retval) 2825 { 2826 /* { 2827 syscallarg(const char *) path; 2828 } */ 2829 2830 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 0, UIO_USERSPACE); 2831 } 2832 2833 int 2834 sys_unlinkat(struct lwp *l, const struct sys_unlinkat_args *uap, 2835 register_t *retval) 2836 { 2837 /* { 2838 syscallarg(int) fd; 2839 syscallarg(const char *) path; 2840 syscallarg(int) flag; 2841 } */ 2842 2843 return do_sys_unlinkat(l, SCARG(uap, fd), SCARG(uap, path), 2844 SCARG(uap, flag), UIO_USERSPACE); 2845 } 2846 2847 int 2848 do_sys_unlink(const char *arg, enum uio_seg seg) 2849 { 2850 return do_sys_unlinkat(NULL, AT_FDCWD, arg, 0, seg); 2851 } 2852 2853 static int 2854 do_sys_unlinkat(struct lwp *l, int fdat, const char *arg, int flags, 2855 enum uio_seg seg) 2856 { 2857 struct vnode *vp; 2858 int error; 2859 struct pathbuf *pb; 2860 struct nameidata nd; 2861 const char *pathstring; 2862 2863 KASSERT(l != NULL || fdat == AT_FDCWD); 2864 2865 error = pathbuf_maybe_copyin(arg, seg, &pb); 2866 if (error) { 2867 return error; 2868 } 2869 pathstring = pathbuf_stringcopy_get(pb); 2870 if (pathstring == NULL) { 2871 pathbuf_destroy(pb); 2872 return ENOMEM; 2873 } 2874 2875 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF | TRYEMULROOT, pb); 2876 if ((error = fd_nameiat(l, fdat, &nd)) != 0) 2877 goto out; 2878 vp = nd.ni_vp; 2879 2880 /* 2881 * The root of a mounted filesystem cannot be deleted. 2882 */ 2883 if ((vp->v_vflag & VV_ROOT) != 0) { 2884 error = EBUSY; 2885 goto abort; 2886 } 2887 2888 if ((vp->v_type == VDIR) && (vp->v_mountedhere != NULL)) { 2889 error = EBUSY; 2890 goto abort; 2891 } 2892 2893 /* 2894 * No rmdir "." please. 2895 */ 2896 if (nd.ni_dvp == vp) { 2897 error = EINVAL; 2898 goto abort; 2899 } 2900 2901 /* 2902 * AT_REMOVEDIR is required to remove a directory 2903 */ 2904 if (vp->v_type == VDIR) { 2905 if (!(flags & AT_REMOVEDIR)) { 2906 error = EPERM; 2907 goto abort; 2908 } else { 2909 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2910 vput(nd.ni_dvp); 2911 goto out; 2912 } 2913 } 2914 2915 /* 2916 * Starting here we only deal with non directories. 2917 */ 2918 if (flags & AT_REMOVEDIR) { 2919 error = ENOTDIR; 2920 goto abort; 2921 } 2922 2923 #if NVERIEXEC > 0 2924 /* Handle remove requests for veriexec entries. */ 2925 if ((error = veriexec_removechk(curlwp, nd.ni_vp, pathstring)) != 0) { 2926 goto abort; 2927 } 2928 #endif /* NVERIEXEC > 0 */ 2929 2930 #ifdef FILEASSOC 2931 (void)fileassoc_file_delete(vp); 2932 #endif /* FILEASSOC */ 2933 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2934 vput(nd.ni_dvp); 2935 goto out; 2936 2937 abort: 2938 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2939 if (nd.ni_dvp == vp) 2940 vrele(nd.ni_dvp); 2941 else 2942 vput(nd.ni_dvp); 2943 vput(vp); 2944 2945 out: 2946 pathbuf_stringcopy_put(pb, pathstring); 2947 pathbuf_destroy(pb); 2948 return (error); 2949 } 2950 2951 /* 2952 * Reposition read/write file offset. 2953 */ 2954 int 2955 sys_lseek(struct lwp *l, const struct sys_lseek_args *uap, register_t *retval) 2956 { 2957 /* { 2958 syscallarg(int) fd; 2959 syscallarg(int) pad; 2960 syscallarg(off_t) offset; 2961 syscallarg(int) whence; 2962 } */ 2963 file_t *fp; 2964 int error, fd; 2965 2966 switch (SCARG(uap, whence)) { 2967 case SEEK_CUR: 2968 case SEEK_END: 2969 case SEEK_SET: 2970 break; 2971 default: 2972 return EINVAL; 2973 } 2974 2975 fd = SCARG(uap, fd); 2976 2977 if ((fp = fd_getfile(fd)) == NULL) 2978 return (EBADF); 2979 2980 if (fp->f_ops->fo_seek == NULL) { 2981 error = ESPIPE; 2982 goto out; 2983 } 2984 2985 error = (*fp->f_ops->fo_seek)(fp, SCARG(uap, offset), 2986 SCARG(uap, whence), (off_t *)retval, FOF_UPDATE_OFFSET); 2987 out: 2988 fd_putfile(fd); 2989 return (error); 2990 } 2991 2992 /* 2993 * Positional read system call. 2994 */ 2995 int 2996 sys_pread(struct lwp *l, const struct sys_pread_args *uap, register_t *retval) 2997 { 2998 /* { 2999 syscallarg(int) fd; 3000 syscallarg(void *) buf; 3001 syscallarg(size_t) nbyte; 3002 syscallarg(off_t) offset; 3003 } */ 3004 file_t *fp; 3005 off_t offset; 3006 int error, fd = SCARG(uap, fd); 3007 3008 if ((fp = fd_getfile(fd)) == NULL) 3009 return (EBADF); 3010 3011 if ((fp->f_flag & FREAD) == 0) { 3012 fd_putfile(fd); 3013 return (EBADF); 3014 } 3015 3016 if (fp->f_ops->fo_seek == NULL) { 3017 error = ESPIPE; 3018 goto out; 3019 } 3020 3021 offset = SCARG(uap, offset); 3022 error = (*fp->f_ops->fo_seek)(fp, offset, SEEK_SET, &offset, 0); 3023 if (error) 3024 goto out; 3025 3026 /* dofileread() will unuse the descriptor for us */ 3027 return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 3028 &offset, 0, retval)); 3029 3030 out: 3031 fd_putfile(fd); 3032 return (error); 3033 } 3034 3035 /* 3036 * Positional scatter read system call. 3037 */ 3038 int 3039 sys_preadv(struct lwp *l, const struct sys_preadv_args *uap, register_t *retval) 3040 { 3041 /* { 3042 syscallarg(int) fd; 3043 syscallarg(const struct iovec *) iovp; 3044 syscallarg(int) iovcnt; 3045 syscallarg(off_t) offset; 3046 } */ 3047 off_t offset = SCARG(uap, offset); 3048 3049 return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp), 3050 SCARG(uap, iovcnt), &offset, 0, retval); 3051 } 3052 3053 /* 3054 * Positional write system call. 3055 */ 3056 int 3057 sys_pwrite(struct lwp *l, const struct sys_pwrite_args *uap, register_t *retval) 3058 { 3059 /* { 3060 syscallarg(int) fd; 3061 syscallarg(const void *) buf; 3062 syscallarg(size_t) nbyte; 3063 syscallarg(off_t) offset; 3064 } */ 3065 file_t *fp; 3066 off_t offset; 3067 int error, fd = SCARG(uap, fd); 3068 3069 if ((fp = fd_getfile(fd)) == NULL) 3070 return (EBADF); 3071 3072 if ((fp->f_flag & FWRITE) == 0) { 3073 fd_putfile(fd); 3074 return (EBADF); 3075 } 3076 3077 if (fp->f_ops->fo_seek == NULL) { 3078 error = ESPIPE; 3079 goto out; 3080 } 3081 3082 offset = SCARG(uap, offset); 3083 error = (*fp->f_ops->fo_seek)(fp, offset, SEEK_SET, &offset, 0); 3084 if (error) 3085 goto out; 3086 3087 /* dofilewrite() will unuse the descriptor for us */ 3088 return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 3089 &offset, 0, retval)); 3090 3091 out: 3092 fd_putfile(fd); 3093 return (error); 3094 } 3095 3096 /* 3097 * Positional gather write system call. 3098 */ 3099 int 3100 sys_pwritev(struct lwp *l, const struct sys_pwritev_args *uap, register_t *retval) 3101 { 3102 /* { 3103 syscallarg(int) fd; 3104 syscallarg(const struct iovec *) iovp; 3105 syscallarg(int) iovcnt; 3106 syscallarg(off_t) offset; 3107 } */ 3108 off_t offset = SCARG(uap, offset); 3109 3110 return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp), 3111 SCARG(uap, iovcnt), &offset, 0, retval); 3112 } 3113 3114 /* 3115 * Check access permissions. 3116 */ 3117 int 3118 sys_access(struct lwp *l, const struct sys_access_args *uap, register_t *retval) 3119 { 3120 /* { 3121 syscallarg(const char *) path; 3122 syscallarg(int) flags; 3123 } */ 3124 3125 return do_sys_accessat(l, AT_FDCWD, SCARG(uap, path), 3126 SCARG(uap, flags), 0); 3127 } 3128 3129 int 3130 do_sys_accessat(struct lwp *l, int fdat, const char *path, 3131 int mode, int flags) 3132 { 3133 kauth_cred_t cred; 3134 struct vnode *vp; 3135 int error, nd_flag, vmode; 3136 struct pathbuf *pb; 3137 struct nameidata nd; 3138 3139 CTASSERT(F_OK == 0); 3140 if ((mode & ~(R_OK | W_OK | X_OK)) != 0) { 3141 /* nonsense mode */ 3142 return EINVAL; 3143 } 3144 3145 nd_flag = FOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT; 3146 if (flags & AT_SYMLINK_NOFOLLOW) 3147 nd_flag &= ~FOLLOW; 3148 3149 error = pathbuf_copyin(path, &pb); 3150 if (error) 3151 return error; 3152 3153 NDINIT(&nd, LOOKUP, nd_flag, pb); 3154 3155 /* Override default credentials */ 3156 if (!(flags & AT_EACCESS)) { 3157 cred = kauth_cred_dup(l->l_cred); 3158 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 3159 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 3160 } else 3161 cred = l->l_cred; 3162 nd.ni_cnd.cn_cred = cred; 3163 3164 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3165 pathbuf_destroy(pb); 3166 goto out; 3167 } 3168 vp = nd.ni_vp; 3169 pathbuf_destroy(pb); 3170 3171 /* Flags == 0 means only check for existence. */ 3172 if (mode) { 3173 vmode = 0; 3174 if (mode & R_OK) 3175 vmode |= VREAD; 3176 if (mode & W_OK) 3177 vmode |= VWRITE; 3178 if (mode & X_OK) 3179 vmode |= VEXEC; 3180 3181 error = VOP_ACCESS(vp, vmode, cred); 3182 if (!error && (vmode & VWRITE)) 3183 error = vn_writechk(vp); 3184 } 3185 vput(vp); 3186 out: 3187 if (!(flags & AT_EACCESS)) 3188 kauth_cred_free(cred); 3189 return (error); 3190 } 3191 3192 int 3193 sys_faccessat(struct lwp *l, const struct sys_faccessat_args *uap, 3194 register_t *retval) 3195 { 3196 /* { 3197 syscallarg(int) fd; 3198 syscallarg(const char *) path; 3199 syscallarg(int) amode; 3200 syscallarg(int) flag; 3201 } */ 3202 3203 return do_sys_accessat(l, SCARG(uap, fd), SCARG(uap, path), 3204 SCARG(uap, amode), SCARG(uap, flag)); 3205 } 3206 3207 /* 3208 * Common code for all sys_stat functions, including compat versions. 3209 */ 3210 int 3211 do_sys_stat(const char *userpath, unsigned int nd_flag, 3212 struct stat *sb) 3213 { 3214 return do_sys_statat(NULL, AT_FDCWD, userpath, nd_flag, sb); 3215 } 3216 3217 int 3218 do_sys_statat(struct lwp *l, int fdat, const char *userpath, 3219 unsigned int nd_flag, struct stat *sb) 3220 { 3221 int error; 3222 struct pathbuf *pb; 3223 struct nameidata nd; 3224 3225 KASSERT(l != NULL || fdat == AT_FDCWD); 3226 3227 error = pathbuf_copyin(userpath, &pb); 3228 if (error) { 3229 return error; 3230 } 3231 3232 NDINIT(&nd, LOOKUP, nd_flag | LOCKLEAF | TRYEMULROOT, pb); 3233 3234 error = fd_nameiat(l, fdat, &nd); 3235 if (error != 0) { 3236 pathbuf_destroy(pb); 3237 return error; 3238 } 3239 error = vn_stat(nd.ni_vp, sb); 3240 vput(nd.ni_vp); 3241 pathbuf_destroy(pb); 3242 return error; 3243 } 3244 3245 /* 3246 * Get file status; this version follows links. 3247 */ 3248 /* ARGSUSED */ 3249 int 3250 sys___stat50(struct lwp *l, const struct sys___stat50_args *uap, register_t *retval) 3251 { 3252 /* { 3253 syscallarg(const char *) path; 3254 syscallarg(struct stat *) ub; 3255 } */ 3256 struct stat sb; 3257 int error; 3258 3259 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), FOLLOW, &sb); 3260 if (error) 3261 return error; 3262 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3263 } 3264 3265 /* 3266 * Get file status; this version does not follow links. 3267 */ 3268 /* ARGSUSED */ 3269 int 3270 sys___lstat50(struct lwp *l, const struct sys___lstat50_args *uap, register_t *retval) 3271 { 3272 /* { 3273 syscallarg(const char *) path; 3274 syscallarg(struct stat *) ub; 3275 } */ 3276 struct stat sb; 3277 int error; 3278 3279 error = do_sys_statat(l, AT_FDCWD, SCARG(uap, path), NOFOLLOW, &sb); 3280 if (error) 3281 return error; 3282 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 3283 } 3284 3285 int 3286 sys_fstatat(struct lwp *l, const struct sys_fstatat_args *uap, 3287 register_t *retval) 3288 { 3289 /* { 3290 syscallarg(int) fd; 3291 syscallarg(const char *) path; 3292 syscallarg(struct stat *) buf; 3293 syscallarg(int) flag; 3294 } */ 3295 unsigned int nd_flag; 3296 struct stat sb; 3297 int error; 3298 3299 if (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) 3300 nd_flag = NOFOLLOW; 3301 else 3302 nd_flag = FOLLOW; 3303 3304 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, 3305 &sb); 3306 if (error) 3307 return error; 3308 return copyout(&sb, SCARG(uap, buf), sizeof(sb)); 3309 } 3310 3311 static int 3312 kern_pathconf(register_t *retval, const char *path, int name, int flag) 3313 { 3314 int error; 3315 struct pathbuf *pb; 3316 struct nameidata nd; 3317 3318 error = pathbuf_copyin(path, &pb); 3319 if (error) { 3320 return error; 3321 } 3322 NDINIT(&nd, LOOKUP, flag | LOCKLEAF | TRYEMULROOT, pb); 3323 if ((error = namei(&nd)) != 0) { 3324 pathbuf_destroy(pb); 3325 return error; 3326 } 3327 error = VOP_PATHCONF(nd.ni_vp, name, retval); 3328 vput(nd.ni_vp); 3329 pathbuf_destroy(pb); 3330 return error; 3331 } 3332 3333 /* 3334 * Get configurable pathname variables. 3335 */ 3336 /* ARGSUSED */ 3337 int 3338 sys_pathconf(struct lwp *l, const struct sys_pathconf_args *uap, 3339 register_t *retval) 3340 { 3341 /* { 3342 syscallarg(const char *) path; 3343 syscallarg(int) name; 3344 } */ 3345 return kern_pathconf(retval, SCARG(uap, path), SCARG(uap, name), 3346 FOLLOW); 3347 } 3348 3349 /* ARGSUSED */ 3350 int 3351 sys_lpathconf(struct lwp *l, const struct sys_lpathconf_args *uap, 3352 register_t *retval) 3353 { 3354 /* { 3355 syscallarg(const char *) path; 3356 syscallarg(int) name; 3357 } */ 3358 return kern_pathconf(retval, SCARG(uap, path), SCARG(uap, name), 3359 NOFOLLOW); 3360 } 3361 3362 /* 3363 * Return target name of a symbolic link. 3364 */ 3365 /* ARGSUSED */ 3366 int 3367 sys_readlink(struct lwp *l, const struct sys_readlink_args *uap, 3368 register_t *retval) 3369 { 3370 /* { 3371 syscallarg(const char *) path; 3372 syscallarg(char *) buf; 3373 syscallarg(size_t) count; 3374 } */ 3375 return do_sys_readlinkat(l, AT_FDCWD, SCARG(uap, path), 3376 SCARG(uap, buf), SCARG(uap, count), retval); 3377 } 3378 3379 static int 3380 do_sys_readlinkat(struct lwp *l, int fdat, const char *path, char *buf, 3381 size_t count, register_t *retval) 3382 { 3383 struct vnode *vp; 3384 struct iovec aiov; 3385 struct uio auio; 3386 int error; 3387 struct pathbuf *pb; 3388 struct nameidata nd; 3389 3390 error = pathbuf_copyin(path, &pb); 3391 if (error) { 3392 return error; 3393 } 3394 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | TRYEMULROOT, pb); 3395 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 3396 pathbuf_destroy(pb); 3397 return error; 3398 } 3399 vp = nd.ni_vp; 3400 pathbuf_destroy(pb); 3401 if (vp->v_type != VLNK) 3402 error = EINVAL; 3403 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 3404 (error = VOP_ACCESS(vp, VREAD, l->l_cred)) == 0) { 3405 aiov.iov_base = buf; 3406 aiov.iov_len = count; 3407 auio.uio_iov = &aiov; 3408 auio.uio_iovcnt = 1; 3409 auio.uio_offset = 0; 3410 auio.uio_rw = UIO_READ; 3411 KASSERT(l == curlwp); 3412 auio.uio_vmspace = l->l_proc->p_vmspace; 3413 auio.uio_resid = count; 3414 if ((error = VOP_READLINK(vp, &auio, l->l_cred)) == 0) 3415 *retval = count - auio.uio_resid; 3416 } 3417 vput(vp); 3418 return (error); 3419 } 3420 3421 int 3422 sys_readlinkat(struct lwp *l, const struct sys_readlinkat_args *uap, 3423 register_t *retval) 3424 { 3425 /* { 3426 syscallarg(int) fd; 3427 syscallarg(const char *) path; 3428 syscallarg(char *) buf; 3429 syscallarg(size_t) bufsize; 3430 } */ 3431 3432 return do_sys_readlinkat(l, SCARG(uap, fd), SCARG(uap, path), 3433 SCARG(uap, buf), SCARG(uap, bufsize), retval); 3434 } 3435 3436 /* 3437 * Change flags of a file given a path name. 3438 */ 3439 /* ARGSUSED */ 3440 int 3441 sys_chflags(struct lwp *l, const struct sys_chflags_args *uap, register_t *retval) 3442 { 3443 /* { 3444 syscallarg(const char *) path; 3445 syscallarg(u_long) flags; 3446 } */ 3447 struct vnode *vp; 3448 int error; 3449 3450 error = namei_simple_user(SCARG(uap, path), 3451 NSM_FOLLOW_TRYEMULROOT, &vp); 3452 if (error != 0) 3453 return (error); 3454 error = change_flags(vp, SCARG(uap, flags), l); 3455 vput(vp); 3456 return (error); 3457 } 3458 3459 /* 3460 * Change flags of a file given a file descriptor. 3461 */ 3462 /* ARGSUSED */ 3463 int 3464 sys_fchflags(struct lwp *l, const struct sys_fchflags_args *uap, register_t *retval) 3465 { 3466 /* { 3467 syscallarg(int) fd; 3468 syscallarg(u_long) flags; 3469 } */ 3470 struct vnode *vp; 3471 file_t *fp; 3472 int error; 3473 3474 /* fd_getvnode() will use the descriptor for us */ 3475 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3476 return (error); 3477 vp = fp->f_vnode; 3478 error = change_flags(vp, SCARG(uap, flags), l); 3479 VOP_UNLOCK(vp); 3480 fd_putfile(SCARG(uap, fd)); 3481 return (error); 3482 } 3483 3484 /* 3485 * Change flags of a file given a path name; this version does 3486 * not follow links. 3487 */ 3488 int 3489 sys_lchflags(struct lwp *l, const struct sys_lchflags_args *uap, register_t *retval) 3490 { 3491 /* { 3492 syscallarg(const char *) path; 3493 syscallarg(u_long) flags; 3494 } */ 3495 struct vnode *vp; 3496 int error; 3497 3498 error = namei_simple_user(SCARG(uap, path), 3499 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3500 if (error != 0) 3501 return (error); 3502 error = change_flags(vp, SCARG(uap, flags), l); 3503 vput(vp); 3504 return (error); 3505 } 3506 3507 /* 3508 * Common routine to change flags of a file. 3509 */ 3510 int 3511 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 3512 { 3513 struct vattr vattr; 3514 int error; 3515 3516 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3517 3518 vattr_null(&vattr); 3519 vattr.va_flags = flags; 3520 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3521 3522 return (error); 3523 } 3524 3525 /* 3526 * Change mode of a file given path name; this version follows links. 3527 */ 3528 /* ARGSUSED */ 3529 int 3530 sys_chmod(struct lwp *l, const struct sys_chmod_args *uap, register_t *retval) 3531 { 3532 /* { 3533 syscallarg(const char *) path; 3534 syscallarg(int) mode; 3535 } */ 3536 return do_sys_chmodat(l, AT_FDCWD, SCARG(uap, path), 3537 SCARG(uap, mode), 0); 3538 } 3539 3540 int 3541 do_sys_chmodat(struct lwp *l, int fdat, const char *path, int mode, int flags) 3542 { 3543 int error; 3544 struct vnode *vp; 3545 namei_simple_flags_t ns_flag; 3546 3547 if (flags & AT_SYMLINK_NOFOLLOW) 3548 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3549 else 3550 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3551 3552 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3553 if (error != 0) 3554 return error; 3555 3556 error = change_mode(vp, mode, l); 3557 3558 vrele(vp); 3559 3560 return (error); 3561 } 3562 3563 /* 3564 * Change mode of a file given a file descriptor. 3565 */ 3566 /* ARGSUSED */ 3567 int 3568 sys_fchmod(struct lwp *l, const struct sys_fchmod_args *uap, register_t *retval) 3569 { 3570 /* { 3571 syscallarg(int) fd; 3572 syscallarg(int) mode; 3573 } */ 3574 file_t *fp; 3575 int error; 3576 3577 /* fd_getvnode() will use the descriptor for us */ 3578 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3579 return (error); 3580 error = change_mode(fp->f_vnode, SCARG(uap, mode), l); 3581 fd_putfile(SCARG(uap, fd)); 3582 return (error); 3583 } 3584 3585 int 3586 sys_fchmodat(struct lwp *l, const struct sys_fchmodat_args *uap, 3587 register_t *retval) 3588 { 3589 /* { 3590 syscallarg(int) fd; 3591 syscallarg(const char *) path; 3592 syscallarg(int) mode; 3593 syscallarg(int) flag; 3594 } */ 3595 3596 return do_sys_chmodat(l, SCARG(uap, fd), SCARG(uap, path), 3597 SCARG(uap, mode), SCARG(uap, flag)); 3598 } 3599 3600 /* 3601 * Change mode of a file given path name; this version does not follow links. 3602 */ 3603 /* ARGSUSED */ 3604 int 3605 sys_lchmod(struct lwp *l, const struct sys_lchmod_args *uap, register_t *retval) 3606 { 3607 /* { 3608 syscallarg(const char *) path; 3609 syscallarg(int) mode; 3610 } */ 3611 int error; 3612 struct vnode *vp; 3613 3614 error = namei_simple_user(SCARG(uap, path), 3615 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3616 if (error != 0) 3617 return (error); 3618 3619 error = change_mode(vp, SCARG(uap, mode), l); 3620 3621 vrele(vp); 3622 return (error); 3623 } 3624 3625 /* 3626 * Common routine to set mode given a vnode. 3627 */ 3628 static int 3629 change_mode(struct vnode *vp, int mode, struct lwp *l) 3630 { 3631 struct vattr vattr; 3632 int error; 3633 3634 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3635 vattr_null(&vattr); 3636 vattr.va_mode = mode & ALLPERMS; 3637 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3638 VOP_UNLOCK(vp); 3639 return (error); 3640 } 3641 3642 /* 3643 * Set ownership given a path name; this version follows links. 3644 */ 3645 /* ARGSUSED */ 3646 int 3647 sys_chown(struct lwp *l, const struct sys_chown_args *uap, register_t *retval) 3648 { 3649 /* { 3650 syscallarg(const char *) path; 3651 syscallarg(uid_t) uid; 3652 syscallarg(gid_t) gid; 3653 } */ 3654 return do_sys_chownat(l, AT_FDCWD, SCARG(uap, path), SCARG(uap,uid), 3655 SCARG(uap, gid), 0); 3656 } 3657 3658 int 3659 do_sys_chownat(struct lwp *l, int fdat, const char *path, uid_t uid, 3660 gid_t gid, int flags) 3661 { 3662 int error; 3663 struct vnode *vp; 3664 namei_simple_flags_t ns_flag; 3665 3666 if (flags & AT_SYMLINK_NOFOLLOW) 3667 ns_flag = NSM_NOFOLLOW_TRYEMULROOT; 3668 else 3669 ns_flag = NSM_FOLLOW_TRYEMULROOT; 3670 3671 error = fd_nameiat_simple_user(l, fdat, path, ns_flag, &vp); 3672 if (error != 0) 3673 return error; 3674 3675 error = change_owner(vp, uid, gid, l, 0); 3676 3677 vrele(vp); 3678 3679 return (error); 3680 } 3681 3682 /* 3683 * Set ownership given a path name; this version follows links. 3684 * Provides POSIX semantics. 3685 */ 3686 /* ARGSUSED */ 3687 int 3688 sys___posix_chown(struct lwp *l, const struct sys___posix_chown_args *uap, register_t *retval) 3689 { 3690 /* { 3691 syscallarg(const char *) path; 3692 syscallarg(uid_t) uid; 3693 syscallarg(gid_t) gid; 3694 } */ 3695 int error; 3696 struct vnode *vp; 3697 3698 error = namei_simple_user(SCARG(uap, path), 3699 NSM_FOLLOW_TRYEMULROOT, &vp); 3700 if (error != 0) 3701 return (error); 3702 3703 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3704 3705 vrele(vp); 3706 return (error); 3707 } 3708 3709 /* 3710 * Set ownership given a file descriptor. 3711 */ 3712 /* ARGSUSED */ 3713 int 3714 sys_fchown(struct lwp *l, const struct sys_fchown_args *uap, register_t *retval) 3715 { 3716 /* { 3717 syscallarg(int) fd; 3718 syscallarg(uid_t) uid; 3719 syscallarg(gid_t) gid; 3720 } */ 3721 int error; 3722 file_t *fp; 3723 3724 /* fd_getvnode() will use the descriptor for us */ 3725 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3726 return (error); 3727 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3728 l, 0); 3729 fd_putfile(SCARG(uap, fd)); 3730 return (error); 3731 } 3732 3733 int 3734 sys_fchownat(struct lwp *l, const struct sys_fchownat_args *uap, 3735 register_t *retval) 3736 { 3737 /* { 3738 syscallarg(int) fd; 3739 syscallarg(const char *) path; 3740 syscallarg(uid_t) owner; 3741 syscallarg(gid_t) group; 3742 syscallarg(int) flag; 3743 } */ 3744 3745 return do_sys_chownat(l, SCARG(uap, fd), SCARG(uap, path), 3746 SCARG(uap, owner), SCARG(uap, group), 3747 SCARG(uap, flag)); 3748 } 3749 3750 /* 3751 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 3752 */ 3753 /* ARGSUSED */ 3754 int 3755 sys___posix_fchown(struct lwp *l, const struct sys___posix_fchown_args *uap, register_t *retval) 3756 { 3757 /* { 3758 syscallarg(int) fd; 3759 syscallarg(uid_t) uid; 3760 syscallarg(gid_t) gid; 3761 } */ 3762 int error; 3763 file_t *fp; 3764 3765 /* fd_getvnode() will use the descriptor for us */ 3766 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3767 return (error); 3768 error = change_owner(fp->f_vnode, SCARG(uap, uid), SCARG(uap, gid), 3769 l, 1); 3770 fd_putfile(SCARG(uap, fd)); 3771 return (error); 3772 } 3773 3774 /* 3775 * Set ownership given a path name; this version does not follow links. 3776 */ 3777 /* ARGSUSED */ 3778 int 3779 sys_lchown(struct lwp *l, const struct sys_lchown_args *uap, register_t *retval) 3780 { 3781 /* { 3782 syscallarg(const char *) path; 3783 syscallarg(uid_t) uid; 3784 syscallarg(gid_t) gid; 3785 } */ 3786 int error; 3787 struct vnode *vp; 3788 3789 error = namei_simple_user(SCARG(uap, path), 3790 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3791 if (error != 0) 3792 return (error); 3793 3794 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 3795 3796 vrele(vp); 3797 return (error); 3798 } 3799 3800 /* 3801 * Set ownership given a path name; this version does not follow links. 3802 * Provides POSIX/XPG semantics. 3803 */ 3804 /* ARGSUSED */ 3805 int 3806 sys___posix_lchown(struct lwp *l, const struct sys___posix_lchown_args *uap, register_t *retval) 3807 { 3808 /* { 3809 syscallarg(const char *) path; 3810 syscallarg(uid_t) uid; 3811 syscallarg(gid_t) gid; 3812 } */ 3813 int error; 3814 struct vnode *vp; 3815 3816 error = namei_simple_user(SCARG(uap, path), 3817 NSM_NOFOLLOW_TRYEMULROOT, &vp); 3818 if (error != 0) 3819 return (error); 3820 3821 error = change_owner(vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 3822 3823 vrele(vp); 3824 return (error); 3825 } 3826 3827 /* 3828 * Common routine to set ownership given a vnode. 3829 */ 3830 static int 3831 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 3832 int posix_semantics) 3833 { 3834 struct vattr vattr; 3835 mode_t newmode; 3836 int error; 3837 3838 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3839 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred)) != 0) 3840 goto out; 3841 3842 #define CHANGED(x) ((int)(x) != -1) 3843 newmode = vattr.va_mode; 3844 if (posix_semantics) { 3845 /* 3846 * POSIX/XPG semantics: if the caller is not the super-user, 3847 * clear set-user-id and set-group-id bits. Both POSIX and 3848 * the XPG consider the behaviour for calls by the super-user 3849 * implementation-defined; we leave the set-user-id and set- 3850 * group-id settings intact in that case. 3851 */ 3852 if (vattr.va_mode & S_ISUID) { 3853 if (kauth_authorize_vnode(l->l_cred, 3854 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) 3855 newmode &= ~S_ISUID; 3856 } 3857 if (vattr.va_mode & S_ISGID) { 3858 if (kauth_authorize_vnode(l->l_cred, 3859 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) 3860 newmode &= ~S_ISGID; 3861 } 3862 } else { 3863 /* 3864 * NetBSD semantics: when changing owner and/or group, 3865 * clear the respective bit(s). 3866 */ 3867 if (CHANGED(uid)) 3868 newmode &= ~S_ISUID; 3869 if (CHANGED(gid)) 3870 newmode &= ~S_ISGID; 3871 } 3872 /* Update va_mode iff altered. */ 3873 if (vattr.va_mode == newmode) 3874 newmode = VNOVAL; 3875 3876 vattr_null(&vattr); 3877 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3878 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3879 vattr.va_mode = newmode; 3880 error = VOP_SETATTR(vp, &vattr, l->l_cred); 3881 #undef CHANGED 3882 3883 out: 3884 VOP_UNLOCK(vp); 3885 return (error); 3886 } 3887 3888 /* 3889 * Set the access and modification times given a path name; this 3890 * version follows links. 3891 */ 3892 /* ARGSUSED */ 3893 int 3894 sys___utimes50(struct lwp *l, const struct sys___utimes50_args *uap, 3895 register_t *retval) 3896 { 3897 /* { 3898 syscallarg(const char *) path; 3899 syscallarg(const struct timeval *) tptr; 3900 } */ 3901 3902 return do_sys_utimes(l, NULL, SCARG(uap, path), FOLLOW, 3903 SCARG(uap, tptr), UIO_USERSPACE); 3904 } 3905 3906 /* 3907 * Set the access and modification times given a file descriptor. 3908 */ 3909 /* ARGSUSED */ 3910 int 3911 sys___futimes50(struct lwp *l, const struct sys___futimes50_args *uap, 3912 register_t *retval) 3913 { 3914 /* { 3915 syscallarg(int) fd; 3916 syscallarg(const struct timeval *) tptr; 3917 } */ 3918 int error; 3919 file_t *fp; 3920 3921 /* fd_getvnode() will use the descriptor for us */ 3922 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3923 return (error); 3924 error = do_sys_utimes(l, fp->f_vnode, NULL, 0, SCARG(uap, tptr), 3925 UIO_USERSPACE); 3926 fd_putfile(SCARG(uap, fd)); 3927 return (error); 3928 } 3929 3930 int 3931 sys_futimens(struct lwp *l, const struct sys_futimens_args *uap, 3932 register_t *retval) 3933 { 3934 /* { 3935 syscallarg(int) fd; 3936 syscallarg(const struct timespec *) tptr; 3937 } */ 3938 int error; 3939 file_t *fp; 3940 3941 /* fd_getvnode() will use the descriptor for us */ 3942 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 3943 return (error); 3944 error = do_sys_utimensat(l, AT_FDCWD, fp->f_vnode, NULL, 0, 3945 SCARG(uap, tptr), UIO_USERSPACE); 3946 fd_putfile(SCARG(uap, fd)); 3947 return (error); 3948 } 3949 3950 /* 3951 * Set the access and modification times given a path name; this 3952 * version does not follow links. 3953 */ 3954 int 3955 sys___lutimes50(struct lwp *l, const struct sys___lutimes50_args *uap, 3956 register_t *retval) 3957 { 3958 /* { 3959 syscallarg(const char *) path; 3960 syscallarg(const struct timeval *) tptr; 3961 } */ 3962 3963 return do_sys_utimes(l, NULL, SCARG(uap, path), NOFOLLOW, 3964 SCARG(uap, tptr), UIO_USERSPACE); 3965 } 3966 3967 int 3968 sys_utimensat(struct lwp *l, const struct sys_utimensat_args *uap, 3969 register_t *retval) 3970 { 3971 /* { 3972 syscallarg(int) fd; 3973 syscallarg(const char *) path; 3974 syscallarg(const struct timespec *) tptr; 3975 syscallarg(int) flag; 3976 } */ 3977 int follow; 3978 const struct timespec *tptr; 3979 int error; 3980 3981 tptr = SCARG(uap, tptr); 3982 follow = (SCARG(uap, flag) & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 3983 3984 error = do_sys_utimensat(l, SCARG(uap, fd), NULL, 3985 SCARG(uap, path), follow, tptr, UIO_USERSPACE); 3986 3987 return error; 3988 } 3989 3990 /* 3991 * Common routine to set access and modification times given a vnode. 3992 */ 3993 int 3994 do_sys_utimens(struct lwp *l, struct vnode *vp, const char *path, int flag, 3995 const struct timespec *tptr, enum uio_seg seg) 3996 { 3997 return do_sys_utimensat(l, AT_FDCWD, vp, path, flag, tptr, seg); 3998 } 3999 4000 int 4001 do_sys_utimensat(struct lwp *l, int fdat, struct vnode *vp, 4002 const char *path, int flag, const struct timespec *tptr, enum uio_seg seg) 4003 { 4004 struct vattr vattr; 4005 int error, dorele = 0; 4006 namei_simple_flags_t sflags; 4007 bool vanull, setbirthtime; 4008 struct timespec ts[2]; 4009 4010 KASSERT(l != NULL || fdat == AT_FDCWD); 4011 4012 /* 4013 * I have checked all callers and they pass either FOLLOW, 4014 * NOFOLLOW, or 0 (when they don't pass a path), and NOFOLLOW 4015 * is 0. More to the point, they don't pass anything else. 4016 * Let's keep it that way at least until the namei interfaces 4017 * are fully sanitized. 4018 */ 4019 KASSERT(flag == NOFOLLOW || flag == FOLLOW); 4020 sflags = (flag == FOLLOW) ? 4021 NSM_FOLLOW_TRYEMULROOT : NSM_NOFOLLOW_TRYEMULROOT; 4022 4023 if (tptr == NULL) { 4024 vanull = true; 4025 nanotime(&ts[0]); 4026 ts[1] = ts[0]; 4027 } else { 4028 vanull = false; 4029 if (seg != UIO_SYSSPACE) { 4030 error = copyin(tptr, ts, sizeof (ts)); 4031 if (error != 0) 4032 return error; 4033 } else { 4034 ts[0] = tptr[0]; 4035 ts[1] = tptr[1]; 4036 } 4037 } 4038 4039 if (ts[0].tv_nsec == UTIME_NOW) { 4040 nanotime(&ts[0]); 4041 if (ts[1].tv_nsec == UTIME_NOW) { 4042 vanull = true; 4043 ts[1] = ts[0]; 4044 } 4045 } else if (ts[1].tv_nsec == UTIME_NOW) 4046 nanotime(&ts[1]); 4047 4048 if (vp == NULL) { 4049 /* note: SEG describes TPTR, not PATH; PATH is always user */ 4050 error = fd_nameiat_simple_user(l, fdat, path, sflags, &vp); 4051 if (error != 0) 4052 return error; 4053 dorele = 1; 4054 } 4055 4056 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4057 setbirthtime = (VOP_GETATTR(vp, &vattr, l->l_cred) == 0 && 4058 timespeccmp(&ts[1], &vattr.va_birthtime, <)); 4059 vattr_null(&vattr); 4060 4061 if (ts[0].tv_nsec != UTIME_OMIT) 4062 vattr.va_atime = ts[0]; 4063 4064 if (ts[1].tv_nsec != UTIME_OMIT) { 4065 vattr.va_mtime = ts[1]; 4066 if (setbirthtime) 4067 vattr.va_birthtime = ts[1]; 4068 } 4069 4070 if (vanull) 4071 vattr.va_vaflags |= VA_UTIMES_NULL; 4072 error = VOP_SETATTR(vp, &vattr, l->l_cred); 4073 VOP_UNLOCK(vp); 4074 4075 if (dorele != 0) 4076 vrele(vp); 4077 4078 return error; 4079 } 4080 4081 int 4082 do_sys_utimes(struct lwp *l, struct vnode *vp, const char *path, int flag, 4083 const struct timeval *tptr, enum uio_seg seg) 4084 { 4085 struct timespec ts[2]; 4086 struct timespec *tsptr = NULL; 4087 int error; 4088 4089 if (tptr != NULL) { 4090 struct timeval tv[2]; 4091 4092 if (seg != UIO_SYSSPACE) { 4093 error = copyin(tptr, tv, sizeof(tv)); 4094 if (error != 0) 4095 return error; 4096 tptr = tv; 4097 } 4098 4099 if ((tptr[0].tv_usec == UTIME_NOW) || 4100 (tptr[0].tv_usec == UTIME_OMIT)) 4101 ts[0].tv_nsec = tptr[0].tv_usec; 4102 else { 4103 if (tptr[0].tv_usec < 0 || tptr[0].tv_usec >= 1000000) 4104 return EINVAL; 4105 4106 TIMEVAL_TO_TIMESPEC(&tptr[0], &ts[0]); 4107 } 4108 4109 if ((tptr[1].tv_usec == UTIME_NOW) || 4110 (tptr[1].tv_usec == UTIME_OMIT)) 4111 ts[1].tv_nsec = tptr[1].tv_usec; 4112 else { 4113 if (tptr[1].tv_usec < 0 || tptr[1].tv_usec >= 1000000) 4114 return EINVAL; 4115 4116 TIMEVAL_TO_TIMESPEC(&tptr[1], &ts[1]); 4117 } 4118 4119 tsptr = &ts[0]; 4120 } 4121 4122 return do_sys_utimens(l, vp, path, flag, tsptr, UIO_SYSSPACE); 4123 } 4124 4125 /* 4126 * Truncate a file given its path name. 4127 */ 4128 /* ARGSUSED */ 4129 int 4130 sys_truncate(struct lwp *l, const struct sys_truncate_args *uap, register_t *retval) 4131 { 4132 /* { 4133 syscallarg(const char *) path; 4134 syscallarg(int) pad; 4135 syscallarg(off_t) length; 4136 } */ 4137 struct vnode *vp; 4138 struct vattr vattr; 4139 int error; 4140 4141 if (SCARG(uap, length) < 0) 4142 return EINVAL; 4143 4144 error = namei_simple_user(SCARG(uap, path), 4145 NSM_FOLLOW_TRYEMULROOT, &vp); 4146 if (error != 0) 4147 return (error); 4148 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4149 if (vp->v_type == VDIR) 4150 error = EISDIR; 4151 else if ((error = vn_writechk(vp)) == 0 && 4152 (error = VOP_ACCESS(vp, VWRITE, l->l_cred)) == 0) { 4153 vattr_null(&vattr); 4154 vattr.va_size = SCARG(uap, length); 4155 error = VOP_SETATTR(vp, &vattr, l->l_cred); 4156 } 4157 vput(vp); 4158 return (error); 4159 } 4160 4161 /* 4162 * Truncate a file given a file descriptor. 4163 */ 4164 /* ARGSUSED */ 4165 int 4166 sys_ftruncate(struct lwp *l, const struct sys_ftruncate_args *uap, register_t *retval) 4167 { 4168 /* { 4169 syscallarg(int) fd; 4170 syscallarg(int) pad; 4171 syscallarg(off_t) length; 4172 } */ 4173 file_t *fp; 4174 int error, fd = SCARG(uap, fd); 4175 4176 fp = fd_getfile(fd); 4177 if (fp == NULL) 4178 return EBADF; 4179 if (fp->f_ops->fo_truncate == NULL) 4180 error = EOPNOTSUPP; 4181 else 4182 error = (*fp->f_ops->fo_truncate)(fp, SCARG(uap, length)); 4183 4184 fd_putfile(fd); 4185 return error; 4186 } 4187 4188 /* 4189 * Sync an open file. 4190 */ 4191 /* ARGSUSED */ 4192 int 4193 sys_fsync(struct lwp *l, const struct sys_fsync_args *uap, register_t *retval) 4194 { 4195 /* { 4196 syscallarg(int) fd; 4197 } */ 4198 struct vnode *vp; 4199 file_t *fp; 4200 int error; 4201 4202 /* fd_getvnode() will use the descriptor for us */ 4203 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4204 return (error); 4205 vp = fp->f_vnode; 4206 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4207 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0); 4208 VOP_UNLOCK(vp); 4209 fd_putfile(SCARG(uap, fd)); 4210 return (error); 4211 } 4212 4213 /* 4214 * Sync a range of file data. API modeled after that found in AIX. 4215 * 4216 * FDATASYNC indicates that we need only save enough metadata to be able 4217 * to re-read the written data. 4218 */ 4219 /* ARGSUSED */ 4220 int 4221 sys_fsync_range(struct lwp *l, const struct sys_fsync_range_args *uap, register_t *retval) 4222 { 4223 /* { 4224 syscallarg(int) fd; 4225 syscallarg(int) flags; 4226 syscallarg(off_t) start; 4227 syscallarg(off_t) length; 4228 } */ 4229 struct vnode *vp; 4230 file_t *fp; 4231 int flags, nflags; 4232 off_t s, e, len; 4233 int error; 4234 4235 /* fd_getvnode() will use the descriptor for us */ 4236 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4237 return (error); 4238 4239 if ((fp->f_flag & FWRITE) == 0) { 4240 error = EBADF; 4241 goto out; 4242 } 4243 4244 flags = SCARG(uap, flags); 4245 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 4246 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 4247 error = EINVAL; 4248 goto out; 4249 } 4250 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 4251 if (flags & FDATASYNC) 4252 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 4253 else 4254 nflags = FSYNC_WAIT; 4255 if (flags & FDISKSYNC) 4256 nflags |= FSYNC_CACHE; 4257 4258 len = SCARG(uap, length); 4259 /* If length == 0, we do the whole file, and s = e = 0 will do that */ 4260 if (len) { 4261 s = SCARG(uap, start); 4262 if (s < 0 || len < 0 || len > OFF_T_MAX - s) { 4263 error = EINVAL; 4264 goto out; 4265 } 4266 e = s + len; 4267 KASSERT(s <= e); 4268 } else { 4269 e = 0; 4270 s = 0; 4271 } 4272 4273 vp = fp->f_vnode; 4274 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4275 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e); 4276 VOP_UNLOCK(vp); 4277 out: 4278 fd_putfile(SCARG(uap, fd)); 4279 return (error); 4280 } 4281 4282 /* 4283 * Sync the data of an open file. 4284 */ 4285 /* ARGSUSED */ 4286 int 4287 sys_fdatasync(struct lwp *l, const struct sys_fdatasync_args *uap, register_t *retval) 4288 { 4289 /* { 4290 syscallarg(int) fd; 4291 } */ 4292 struct vnode *vp; 4293 file_t *fp; 4294 int error; 4295 4296 /* fd_getvnode() will use the descriptor for us */ 4297 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4298 return (error); 4299 vp = fp->f_vnode; 4300 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4301 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0); 4302 VOP_UNLOCK(vp); 4303 fd_putfile(SCARG(uap, fd)); 4304 return (error); 4305 } 4306 4307 /* 4308 * Rename files, (standard) BSD semantics frontend. 4309 */ 4310 /* ARGSUSED */ 4311 int 4312 sys_rename(struct lwp *l, const struct sys_rename_args *uap, register_t *retval) 4313 { 4314 /* { 4315 syscallarg(const char *) from; 4316 syscallarg(const char *) to; 4317 } */ 4318 4319 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4320 SCARG(uap, to), UIO_USERSPACE, 0)); 4321 } 4322 4323 int 4324 sys_renameat(struct lwp *l, const struct sys_renameat_args *uap, 4325 register_t *retval) 4326 { 4327 /* { 4328 syscallarg(int) fromfd; 4329 syscallarg(const char *) from; 4330 syscallarg(int) tofd; 4331 syscallarg(const char *) to; 4332 } */ 4333 4334 return (do_sys_renameat(l, SCARG(uap, fromfd), SCARG(uap, from), 4335 SCARG(uap, tofd), SCARG(uap, to), UIO_USERSPACE, 0)); 4336 } 4337 4338 /* 4339 * Rename files, POSIX semantics frontend. 4340 */ 4341 /* ARGSUSED */ 4342 int 4343 sys___posix_rename(struct lwp *l, const struct sys___posix_rename_args *uap, register_t *retval) 4344 { 4345 /* { 4346 syscallarg(const char *) from; 4347 syscallarg(const char *) to; 4348 } */ 4349 4350 return (do_sys_renameat(l, AT_FDCWD, SCARG(uap, from), AT_FDCWD, 4351 SCARG(uap, to), UIO_USERSPACE, 1)); 4352 } 4353 4354 /* 4355 * Rename files. Source and destination must either both be directories, 4356 * or both not be directories. If target is a directory, it must be empty. 4357 * If `from' and `to' refer to the same object, the value of the `retain' 4358 * argument is used to determine whether `from' will be 4359 * 4360 * (retain == 0) deleted unless `from' and `to' refer to the same 4361 * object in the file system's name space (BSD). 4362 * (retain == 1) always retained (POSIX). 4363 * 4364 * XXX Synchronize with nfsrv_rename in nfs_serv.c. 4365 */ 4366 int 4367 do_sys_rename(const char *from, const char *to, enum uio_seg seg, int retain) 4368 { 4369 return do_sys_renameat(NULL, AT_FDCWD, from, AT_FDCWD, to, seg, retain); 4370 } 4371 4372 static int 4373 do_sys_renameat(struct lwp *l, int fromfd, const char *from, int tofd, 4374 const char *to, enum uio_seg seg, int retain) 4375 { 4376 struct pathbuf *fpb, *tpb; 4377 struct nameidata fnd, tnd; 4378 struct vnode *fdvp, *fvp; 4379 struct vnode *tdvp, *tvp; 4380 struct mount *mp, *tmp; 4381 int error; 4382 4383 KASSERT(l != NULL || fromfd == AT_FDCWD); 4384 KASSERT(l != NULL || tofd == AT_FDCWD); 4385 4386 error = pathbuf_maybe_copyin(from, seg, &fpb); 4387 if (error) 4388 goto out0; 4389 KASSERT(fpb != NULL); 4390 4391 error = pathbuf_maybe_copyin(to, seg, &tpb); 4392 if (error) 4393 goto out1; 4394 KASSERT(tpb != NULL); 4395 4396 /* 4397 * Lookup from. 4398 * 4399 * XXX LOCKPARENT is wrong because we don't actually want it 4400 * locked yet, but (a) namei is insane, and (b) VOP_RENAME is 4401 * insane, so for the time being we need to leave it like this. 4402 */ 4403 NDINIT(&fnd, DELETE, (LOCKPARENT | TRYEMULROOT), fpb); 4404 if ((error = fd_nameiat(l, fromfd, &fnd)) != 0) 4405 goto out2; 4406 4407 /* 4408 * Pull out the important results of the lookup, fdvp and fvp. 4409 * Of course, fvp is bogus because we're about to unlock fdvp. 4410 */ 4411 fdvp = fnd.ni_dvp; 4412 fvp = fnd.ni_vp; 4413 mp = fdvp->v_mount; 4414 KASSERT(fdvp != NULL); 4415 KASSERT(fvp != NULL); 4416 KASSERT((fdvp == fvp) || (VOP_ISLOCKED(fdvp) == LK_EXCLUSIVE)); 4417 /* 4418 * Bracket the operation with fstrans_start()/fstrans_done(). 4419 * 4420 * Inside the bracket this file system cannot be unmounted so 4421 * a vnode on this file system cannot change its v_mount. 4422 * A vnode on another file system may still change to dead mount. 4423 */ 4424 fstrans_start(mp); 4425 4426 /* 4427 * Make sure neither fdvp nor fvp is locked. 4428 */ 4429 if (fdvp != fvp) 4430 VOP_UNLOCK(fdvp); 4431 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4432 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4433 4434 /* 4435 * Reject renaming `.' and `..'. Can't do this until after 4436 * namei because we need namei's parsing to find the final 4437 * component name. (namei should just leave us with the final 4438 * component name and not look it up itself, but anyway...) 4439 * 4440 * This was here before because we used to relookup from 4441 * instead of to and relookup requires the caller to check 4442 * this, but now file systems may depend on this check, so we 4443 * must retain it until the file systems are all rototilled. 4444 */ 4445 if (((fnd.ni_cnd.cn_namelen == 1) && 4446 (fnd.ni_cnd.cn_nameptr[0] == '.')) || 4447 ((fnd.ni_cnd.cn_namelen == 2) && 4448 (fnd.ni_cnd.cn_nameptr[0] == '.') && 4449 (fnd.ni_cnd.cn_nameptr[1] == '.'))) { 4450 error = EINVAL; /* XXX EISDIR? */ 4451 goto abort0; 4452 } 4453 4454 /* 4455 * Lookup to. 4456 * 4457 * XXX LOCKPARENT is wrong, but...insanity, &c. Also, using 4458 * fvp here to decide whether to add CREATEDIR is a load of 4459 * bollocks because fvp might be the wrong node by now, since 4460 * fdvp is unlocked. 4461 * 4462 * XXX Why not pass CREATEDIR always? 4463 */ 4464 NDINIT(&tnd, RENAME, 4465 (LOCKPARENT | NOCACHE | TRYEMULROOT | 4466 ((fvp->v_type == VDIR)? CREATEDIR : 0)), 4467 tpb); 4468 if ((error = fd_nameiat(l, tofd, &tnd)) != 0) 4469 goto abort0; 4470 4471 /* 4472 * Pull out the important results of the lookup, tdvp and tvp. 4473 * Of course, tvp is bogus because we're about to unlock tdvp. 4474 */ 4475 tdvp = tnd.ni_dvp; 4476 tvp = tnd.ni_vp; 4477 KASSERT(tdvp != NULL); 4478 KASSERT((tdvp == tvp) || (VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE)); 4479 4480 if (fvp->v_type == VDIR) 4481 tnd.ni_cnd.cn_flags |= WILLBEDIR; 4482 /* 4483 * Make sure neither tdvp nor tvp is locked. 4484 */ 4485 if (tdvp != tvp) 4486 VOP_UNLOCK(tdvp); 4487 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4488 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4489 4490 /* 4491 * Reject renaming onto `.' or `..'. relookup is unhappy with 4492 * these, which is why we must do this here. Once upon a time 4493 * we relooked up from instead of to, and consequently didn't 4494 * need this check, but now that we relookup to instead of 4495 * from, we need this; and we shall need it forever forward 4496 * until the VOP_RENAME protocol changes, because file systems 4497 * will no doubt begin to depend on this check. 4498 */ 4499 if ((tnd.ni_cnd.cn_namelen == 1) && (tnd.ni_cnd.cn_nameptr[0] == '.')) { 4500 error = EISDIR; 4501 goto abort1; 4502 } 4503 if ((tnd.ni_cnd.cn_namelen == 2) && 4504 (tnd.ni_cnd.cn_nameptr[0] == '.') && 4505 (tnd.ni_cnd.cn_nameptr[1] == '.')) { 4506 error = EINVAL; 4507 goto abort1; 4508 } 4509 4510 /* 4511 * Make sure the mount points match. Although we don't hold 4512 * any vnode locks, the v_mount on fdvp file system are stable. 4513 * 4514 * Unmounting another file system at an inopportune moment may 4515 * cause tdvp to disappear and change its v_mount to dead. 4516 * 4517 * So in either case different v_mount means cross-device rename. 4518 */ 4519 KASSERT(mp != NULL); 4520 tmp = tdvp->v_mount; 4521 4522 if (mp != tmp) { 4523 error = EXDEV; 4524 goto abort1; 4525 } 4526 4527 /* 4528 * Take the vfs rename lock to avoid cross-directory screw cases. 4529 * Nothing is locked currently, so taking this lock is safe. 4530 */ 4531 error = VFS_RENAMELOCK_ENTER(mp); 4532 if (error) 4533 goto abort1; 4534 4535 /* 4536 * Now fdvp, fvp, tdvp, and (if nonnull) tvp are referenced, 4537 * and nothing is locked except for the vfs rename lock. 4538 * 4539 * The next step is a little rain dance to conform to the 4540 * insane lock protocol, even though it does nothing to ward 4541 * off race conditions. 4542 * 4543 * We need tdvp and tvp to be locked. However, because we have 4544 * unlocked tdvp in order to hold no locks while we take the 4545 * vfs rename lock, tvp may be wrong here, and we can't safely 4546 * lock it even if the sensible file systems will just unlock 4547 * it straight away. Consequently, we must lock tdvp and then 4548 * relookup tvp to get it locked. 4549 * 4550 * Finally, because the VOP_RENAME protocol is brain-damaged 4551 * and various file systems insanely depend on the semantics of 4552 * this brain damage, the lookup of to must be the last lookup 4553 * before VOP_RENAME. 4554 */ 4555 vn_lock(tdvp, LK_EXCLUSIVE | LK_RETRY); 4556 error = relookup(tdvp, &tnd.ni_vp, &tnd.ni_cnd, 0); 4557 if (error) 4558 goto abort2; 4559 4560 /* 4561 * Drop the old tvp and pick up the new one -- which might be 4562 * the same, but that doesn't matter to us. After this, tdvp 4563 * and tvp should both be locked. 4564 */ 4565 if (tvp != NULL) 4566 vrele(tvp); 4567 tvp = tnd.ni_vp; 4568 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4569 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4570 4571 /* 4572 * The old do_sys_rename had various consistency checks here 4573 * involving fvp and tvp. fvp is bogus already here, and tvp 4574 * will become bogus soon in any sensible file system, so the 4575 * only purpose in putting these checks here is to give lip 4576 * service to these screw cases and to acknowledge that they 4577 * exist, not actually to handle them, but here you go 4578 * anyway... 4579 */ 4580 4581 /* 4582 * Acknowledge that directories and non-directories aren't 4583 * supposed to mix. 4584 */ 4585 if (tvp != NULL) { 4586 if ((fvp->v_type == VDIR) && (tvp->v_type != VDIR)) { 4587 error = ENOTDIR; 4588 goto abort3; 4589 } else if ((fvp->v_type != VDIR) && (tvp->v_type == VDIR)) { 4590 error = EISDIR; 4591 goto abort3; 4592 } 4593 } 4594 4595 /* 4596 * Acknowledge some random screw case, among the dozens that 4597 * might arise. 4598 */ 4599 if (fvp == tdvp) { 4600 error = EINVAL; 4601 goto abort3; 4602 } 4603 4604 /* 4605 * Acknowledge that POSIX has a wacky screw case. 4606 * 4607 * XXX Eventually the retain flag needs to be passed on to 4608 * VOP_RENAME. 4609 */ 4610 if (fvp == tvp) { 4611 if (retain) { 4612 error = 0; 4613 goto abort3; 4614 } else if ((fdvp == tdvp) && 4615 (fnd.ni_cnd.cn_namelen == tnd.ni_cnd.cn_namelen) && 4616 (0 == memcmp(fnd.ni_cnd.cn_nameptr, tnd.ni_cnd.cn_nameptr, 4617 fnd.ni_cnd.cn_namelen))) { 4618 error = 0; 4619 goto abort3; 4620 } 4621 } 4622 4623 /* 4624 * Make sure veriexec can screw us up. (But a race can screw 4625 * up veriexec, of course -- remember, fvp and (soon) tvp are 4626 * bogus.) 4627 */ 4628 #if NVERIEXEC > 0 4629 { 4630 char *f1, *f2; 4631 size_t f1_len; 4632 size_t f2_len; 4633 4634 f1_len = fnd.ni_cnd.cn_namelen + 1; 4635 f1 = kmem_alloc(f1_len, KM_SLEEP); 4636 strlcpy(f1, fnd.ni_cnd.cn_nameptr, f1_len); 4637 4638 f2_len = tnd.ni_cnd.cn_namelen + 1; 4639 f2 = kmem_alloc(f2_len, KM_SLEEP); 4640 strlcpy(f2, tnd.ni_cnd.cn_nameptr, f2_len); 4641 4642 error = veriexec_renamechk(curlwp, fvp, f1, tvp, f2); 4643 4644 kmem_free(f1, f1_len); 4645 kmem_free(f2, f2_len); 4646 4647 if (error) 4648 goto abort3; 4649 } 4650 #endif /* NVERIEXEC > 0 */ 4651 4652 /* 4653 * All ready. Incant the rename vop. 4654 */ 4655 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4656 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4657 KASSERT(VOP_ISLOCKED(tdvp) == LK_EXCLUSIVE); 4658 KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) == LK_EXCLUSIVE)); 4659 error = VOP_RENAME(fdvp, fvp, &fnd.ni_cnd, tdvp, tvp, &tnd.ni_cnd); 4660 4661 /* 4662 * VOP_RENAME releases fdvp, fvp, tdvp, and tvp, and unlocks 4663 * tdvp and tvp. But we can't assert any of that. 4664 */ 4665 /* XXX KASSERT(VOP_ISLOCKED(fdvp) != LK_EXCLUSIVE); */ 4666 /* XXX KASSERT(VOP_ISLOCKED(fvp) != LK_EXCLUSIVE); */ 4667 /* XXX KASSERT(VOP_ISLOCKED(tdvp) != LK_EXCLUSIVE); */ 4668 /* XXX KASSERT((tvp == NULL) || (VOP_ISLOCKED(tvp) != LK_EXCLUSIVE)); */ 4669 4670 /* 4671 * So all we have left to do is to drop the rename lock and 4672 * destroy the pathbufs. 4673 */ 4674 VFS_RENAMELOCK_EXIT(mp); 4675 fstrans_done(mp); 4676 goto out2; 4677 4678 abort3: if ((tvp != NULL) && (tvp != tdvp)) 4679 VOP_UNLOCK(tvp); 4680 abort2: VOP_UNLOCK(tdvp); 4681 VFS_RENAMELOCK_EXIT(mp); 4682 abort1: VOP_ABORTOP(tdvp, &tnd.ni_cnd); 4683 vrele(tdvp); 4684 if (tvp != NULL) 4685 vrele(tvp); 4686 abort0: VOP_ABORTOP(fdvp, &fnd.ni_cnd); 4687 vrele(fdvp); 4688 vrele(fvp); 4689 fstrans_done(mp); 4690 out2: pathbuf_destroy(tpb); 4691 out1: pathbuf_destroy(fpb); 4692 out0: return error; 4693 } 4694 4695 /* 4696 * Make a directory file. 4697 */ 4698 /* ARGSUSED */ 4699 int 4700 sys_mkdir(struct lwp *l, const struct sys_mkdir_args *uap, register_t *retval) 4701 { 4702 /* { 4703 syscallarg(const char *) path; 4704 syscallarg(int) mode; 4705 } */ 4706 4707 return do_sys_mkdirat(l, AT_FDCWD, SCARG(uap, path), 4708 SCARG(uap, mode), UIO_USERSPACE); 4709 } 4710 4711 int 4712 sys_mkdirat(struct lwp *l, const struct sys_mkdirat_args *uap, 4713 register_t *retval) 4714 { 4715 /* { 4716 syscallarg(int) fd; 4717 syscallarg(const char *) path; 4718 syscallarg(int) mode; 4719 } */ 4720 4721 return do_sys_mkdirat(l, SCARG(uap, fd), SCARG(uap, path), 4722 SCARG(uap, mode), UIO_USERSPACE); 4723 } 4724 4725 4726 int 4727 do_sys_mkdir(const char *path, mode_t mode, enum uio_seg seg) 4728 { 4729 return do_sys_mkdirat(NULL, AT_FDCWD, path, mode, seg); 4730 } 4731 4732 static int 4733 do_sys_mkdirat(struct lwp *l, int fdat, const char *path, mode_t mode, 4734 enum uio_seg seg) 4735 { 4736 struct proc *p = curlwp->l_proc; 4737 struct vnode *vp; 4738 struct vattr vattr; 4739 int error; 4740 struct pathbuf *pb; 4741 struct nameidata nd; 4742 4743 KASSERT(l != NULL || fdat == AT_FDCWD); 4744 4745 /* XXX bollocks, should pass in a pathbuf */ 4746 error = pathbuf_maybe_copyin(path, seg, &pb); 4747 if (error) { 4748 return error; 4749 } 4750 4751 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR | TRYEMULROOT, pb); 4752 4753 if ((error = fd_nameiat(l, fdat, &nd)) != 0) { 4754 pathbuf_destroy(pb); 4755 return (error); 4756 } 4757 vp = nd.ni_vp; 4758 if (vp != NULL) { 4759 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 4760 if (nd.ni_dvp == vp) 4761 vrele(nd.ni_dvp); 4762 else 4763 vput(nd.ni_dvp); 4764 vrele(vp); 4765 pathbuf_destroy(pb); 4766 return (EEXIST); 4767 } 4768 vattr_null(&vattr); 4769 vattr.va_type = VDIR; 4770 /* We will read cwdi->cwdi_cmask unlocked. */ 4771 vattr.va_mode = (mode & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 4772 nd.ni_cnd.cn_flags |= WILLBEDIR; 4773 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 4774 if (!error) 4775 vrele(nd.ni_vp); 4776 vput(nd.ni_dvp); 4777 pathbuf_destroy(pb); 4778 return (error); 4779 } 4780 4781 /* 4782 * Remove a directory file. 4783 */ 4784 /* ARGSUSED */ 4785 int 4786 sys_rmdir(struct lwp *l, const struct sys_rmdir_args *uap, register_t *retval) 4787 { 4788 return do_sys_unlinkat(l, AT_FDCWD, SCARG(uap, path), 4789 AT_REMOVEDIR, UIO_USERSPACE); 4790 } 4791 4792 /* 4793 * Read a block of directory entries in a file system independent format. 4794 */ 4795 int 4796 sys___getdents30(struct lwp *l, const struct sys___getdents30_args *uap, register_t *retval) 4797 { 4798 /* { 4799 syscallarg(int) fd; 4800 syscallarg(char *) buf; 4801 syscallarg(size_t) count; 4802 } */ 4803 file_t *fp; 4804 int error, done; 4805 4806 /* fd_getvnode() will use the descriptor for us */ 4807 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 4808 return (error); 4809 if ((fp->f_flag & FREAD) == 0) { 4810 error = EBADF; 4811 goto out; 4812 } 4813 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 4814 SCARG(uap, count), &done, l, 0, 0); 4815 ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, buf), done, error); 4816 *retval = done; 4817 out: 4818 fd_putfile(SCARG(uap, fd)); 4819 return (error); 4820 } 4821 4822 /* 4823 * Set the mode mask for creation of filesystem nodes. 4824 */ 4825 int 4826 sys_umask(struct lwp *l, const struct sys_umask_args *uap, register_t *retval) 4827 { 4828 /* { 4829 syscallarg(mode_t) newmask; 4830 } */ 4831 4832 /* 4833 * cwdi->cwdi_cmask will be read unlocked elsewhere, and no kind of 4834 * serialization with those reads is required. It's important to 4835 * return a coherent answer for the caller of umask() though, and 4836 * the atomic operation accomplishes that. 4837 */ 4838 *retval = atomic_swap_uint(&curproc->p_cwdi->cwdi_cmask, 4839 SCARG(uap, newmask) & ALLPERMS); 4840 4841 return (0); 4842 } 4843 4844 int 4845 dorevoke(struct vnode *vp, kauth_cred_t cred) 4846 { 4847 struct vattr vattr; 4848 int error, fs_decision; 4849 4850 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4851 error = VOP_GETATTR(vp, &vattr, cred); 4852 VOP_UNLOCK(vp); 4853 if (error != 0) 4854 return error; 4855 fs_decision = (kauth_cred_geteuid(cred) == vattr.va_uid) ? 0 : EPERM; 4856 error = kauth_authorize_vnode(cred, KAUTH_VNODE_REVOKE, vp, NULL, 4857 fs_decision); 4858 if (!error) 4859 VOP_REVOKE(vp, REVOKEALL); 4860 return (error); 4861 } 4862 4863 /* 4864 * Void all references to file by ripping underlying filesystem 4865 * away from vnode. 4866 */ 4867 /* ARGSUSED */ 4868 int 4869 sys_revoke(struct lwp *l, const struct sys_revoke_args *uap, register_t *retval) 4870 { 4871 /* { 4872 syscallarg(const char *) path; 4873 } */ 4874 struct vnode *vp; 4875 int error; 4876 4877 error = namei_simple_user(SCARG(uap, path), 4878 NSM_FOLLOW_TRYEMULROOT, &vp); 4879 if (error != 0) 4880 return (error); 4881 error = dorevoke(vp, l->l_cred); 4882 vrele(vp); 4883 return (error); 4884 } 4885 4886 /* 4887 * Allocate backing store for a file, filling a hole without having to 4888 * explicitly write anything out. 4889 */ 4890 /* ARGSUSED */ 4891 int 4892 sys_posix_fallocate(struct lwp *l, const struct sys_posix_fallocate_args *uap, 4893 register_t *retval) 4894 { 4895 /* { 4896 syscallarg(int) fd; 4897 syscallarg(off_t) pos; 4898 syscallarg(off_t) len; 4899 } */ 4900 int fd; 4901 off_t pos, len; 4902 struct file *fp; 4903 struct vnode *vp; 4904 int error; 4905 4906 fd = SCARG(uap, fd); 4907 pos = SCARG(uap, pos); 4908 len = SCARG(uap, len); 4909 4910 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4911 *retval = EINVAL; 4912 return 0; 4913 } 4914 4915 error = fd_getvnode(fd, &fp); 4916 if (error) { 4917 *retval = error; 4918 return 0; 4919 } 4920 if ((fp->f_flag & FWRITE) == 0) { 4921 error = EBADF; 4922 goto fail; 4923 } 4924 vp = fp->f_vnode; 4925 4926 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4927 if (vp->v_type == VDIR) { 4928 error = EISDIR; 4929 } else { 4930 error = VOP_FALLOCATE(vp, pos, len); 4931 } 4932 VOP_UNLOCK(vp); 4933 4934 fail: 4935 fd_putfile(fd); 4936 *retval = error; 4937 return 0; 4938 } 4939 4940 /* 4941 * Deallocate backing store for a file, creating a hole. Also used for 4942 * invoking TRIM on disks. 4943 */ 4944 /* ARGSUSED */ 4945 int 4946 sys_fdiscard(struct lwp *l, const struct sys_fdiscard_args *uap, 4947 register_t *retval) 4948 { 4949 /* { 4950 syscallarg(int) fd; 4951 syscallarg(off_t) pos; 4952 syscallarg(off_t) len; 4953 } */ 4954 int fd; 4955 off_t pos, len; 4956 struct file *fp; 4957 struct vnode *vp; 4958 int error; 4959 4960 fd = SCARG(uap, fd); 4961 pos = SCARG(uap, pos); 4962 len = SCARG(uap, len); 4963 4964 if (pos < 0 || len < 0 || len > OFF_T_MAX - pos) { 4965 return EINVAL; 4966 } 4967 4968 error = fd_getvnode(fd, &fp); 4969 if (error) { 4970 return error; 4971 } 4972 if ((fp->f_flag & FWRITE) == 0) { 4973 error = EBADF; 4974 goto fail; 4975 } 4976 vp = fp->f_vnode; 4977 4978 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 4979 if (vp->v_type == VDIR) { 4980 error = EISDIR; 4981 } else { 4982 error = VOP_FDISCARD(vp, pos, len); 4983 } 4984 VOP_UNLOCK(vp); 4985 4986 fail: 4987 fd_putfile(fd); 4988 return error; 4989 } 4990
Indexes created Fri Sep 26 08:10:20 GMT 2025