vfs_syscalls.c revision 1.306.2.2
1 /* $NetBSD: vfs_syscalls.c,v 1.306.2.2 2007/03/21 20:18:48 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_syscalls.c 8.42 (Berkeley) 7/31/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_syscalls.c,v 1.306.2.2 2007/03/21 20:18:48 ad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_fileassoc.h" 45 #include "opt_ktrace.h" 46 #include "fss.h" 47 #include "veriexec.h" 48 49 #include <sys/param.h> 50 #include <sys/systm.h> 51 #include <sys/namei.h> 52 #include <sys/filedesc.h> 53 #include <sys/kernel.h> 54 #include <sys/file.h> 55 #include <sys/stat.h> 56 #include <sys/vnode.h> 57 #include <sys/mount.h> 58 #include <sys/proc.h> 59 #include <sys/uio.h> 60 #include <sys/malloc.h> 61 #include <sys/kmem.h> 62 #include <sys/dirent.h> 63 #include <sys/sysctl.h> 64 #include <sys/syscallargs.h> 65 #include <sys/vfs_syscalls.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #ifdef FILEASSOC 70 #include <sys/fileassoc.h> 71 #endif /* FILEASSOC */ 72 #if NVERIEXEC > 0 73 #include <sys/verified_exec.h> 74 #include <sys/syslog.h> 75 #endif /* NVERIEXEC > 0 */ 76 #include <sys/kauth.h> 77 78 #include <miscfs/genfs/genfs.h> 79 #include <miscfs/syncfs/syncfs.h> 80 81 #ifdef COMPAT_30 82 #include "opt_nfsserver.h" 83 #include <nfs/rpcv2.h> 84 #endif 85 #include <nfs/nfsproto.h> 86 #ifdef COMPAT_30 87 #include <nfs/nfs.h> 88 #include <nfs/nfs_var.h> 89 #endif 90 91 #if NFSS > 0 92 #include <dev/fssvar.h> 93 #endif 94 95 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 96 97 static int change_dir(struct nameidata *, struct lwp *); 98 static int change_flags(struct vnode *, u_long, struct lwp *); 99 static int change_mode(struct vnode *, int, struct lwp *l); 100 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 101 static int change_utimes(struct vnode *vp, const struct timeval *, 102 struct lwp *l); 103 static int rename_files(const char *, const char *, struct lwp *, int); 104 105 void checkdirs(struct vnode *); 106 107 static int mount_update(struct lwp *, struct vnode *, const char *, int, 108 void *, struct nameidata *); 109 static int mount_domount(struct lwp *, struct vnode *, const char *, 110 const char *, int, void *, struct nameidata *); 111 static int mount_getargs(struct lwp *, struct vnode *, const char *, int, 112 void *, struct nameidata *); 113 114 int dovfsusermount = 0; 115 116 /* 117 * Virtual File System System Calls 118 */ 119 120 /* 121 * Mount a file system. 122 */ 123 124 #if defined(COMPAT_09) || defined(COMPAT_43) 125 /* 126 * This table is used to maintain compatibility with 4.3BSD 127 * and NetBSD 0.9 mount syscalls. Note, the order is important! 128 * 129 * Do not modify this table. It should only contain filesystems 130 * supported by NetBSD 0.9 and 4.3BSD. 131 */ 132 const char * const mountcompatnames[] = { 133 NULL, /* 0 = MOUNT_NONE */ 134 MOUNT_FFS, /* 1 = MOUNT_UFS */ 135 MOUNT_NFS, /* 2 */ 136 MOUNT_MFS, /* 3 */ 137 MOUNT_MSDOS, /* 4 */ 138 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 139 MOUNT_FDESC, /* 6 */ 140 MOUNT_KERNFS, /* 7 */ 141 NULL, /* 8 = MOUNT_DEVFS */ 142 MOUNT_AFS, /* 9 */ 143 }; 144 const int nmountcompatnames = sizeof(mountcompatnames) / 145 sizeof(mountcompatnames[0]); 146 #endif /* COMPAT_09 || COMPAT_43 */ 147 148 static int 149 mount_update(struct lwp *l, struct vnode *vp, const char *path, int flags, 150 void *data, struct nameidata *ndp) 151 { 152 struct mount *mp; 153 int error = 0, saved_flags; 154 155 mp = vp->v_mount; 156 saved_flags = mp->mnt_flag; 157 158 /* We can't operate on VROOT here. */ 159 if ((vp->v_flag & VROOT) == 0) { 160 vput(vp); 161 error = EINVAL; 162 goto out; 163 } 164 165 /* 166 * We only allow the filesystem to be reloaded if it 167 * is currently mounted read-only. 168 */ 169 if (flags & MNT_RELOAD && !(mp->mnt_flag & MNT_RDONLY)) { 170 error = EOPNOTSUPP; /* Needs translation */ 171 goto out; 172 } 173 174 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 175 KAUTH_REQ_SYSTEM_MOUNT_UPDATE, mp, KAUTH_ARG(flags), data); 176 if (error) 177 goto out; 178 179 if (vfs_busy(mp, LK_NOWAIT, 0)) { 180 error = EPERM; 181 goto out; 182 } 183 184 mp->mnt_flag &= ~MNT_OP_FLAGS; 185 mp->mnt_flag |= flags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE); 186 187 /* 188 * Set the mount level flags. 189 */ 190 if (flags & MNT_RDONLY) 191 mp->mnt_flag |= MNT_RDONLY; 192 else if (mp->mnt_flag & MNT_RDONLY) 193 mp->mnt_iflag |= IMNT_WANTRDWR; 194 mp->mnt_flag &= 195 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 196 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 197 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 198 mp->mnt_flag |= flags & 199 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 200 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 201 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 202 MNT_IGNORE); 203 204 error = VFS_MOUNT(mp, path, data, ndp, l); 205 206 #if defined(COMPAT_30) && defined(NFSSERVER) 207 if (error) { 208 int error2; 209 210 /* Update failed; let's try and see if it was an 211 * export request. */ 212 error2 = nfs_update_exports_30(mp, path, data, l); 213 214 /* Only update error code if the export request was 215 * understood but some problem occurred while 216 * processing it. */ 217 if (error2 != EJUSTRETURN) 218 error = error2; 219 } 220 #endif 221 if (mp->mnt_iflag & IMNT_WANTRDWR) 222 mp->mnt_flag &= ~MNT_RDONLY; 223 if (error) 224 mp->mnt_flag = saved_flags; 225 mp->mnt_flag &= ~MNT_OP_FLAGS; 226 mp->mnt_iflag &= ~IMNT_WANTRDWR; 227 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 228 if (mp->mnt_syncer == NULL) 229 error = vfs_allocate_syncvnode(mp); 230 } else { 231 if (mp->mnt_syncer != NULL) 232 vfs_deallocate_syncvnode(mp); 233 } 234 vfs_unbusy(mp); 235 236 out: 237 return (error); 238 } 239 240 static int 241 mount_domount(struct lwp *l, struct vnode *vp, const char *fstype, 242 const char *path, int flags, void *data, struct nameidata *ndp) 243 { 244 struct mount *mp = NULL; 245 struct vattr va; 246 char fstypename[MFSNAMELEN]; 247 int error; 248 249 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 250 KAUTH_REQ_SYSTEM_MOUNT_NEW, vp, KAUTH_ARG(flags), data); 251 if (error) { 252 vput(vp); 253 goto out; 254 } 255 256 /* Can't make a non-dir a mount-point (from here anyway). */ 257 if (vp->v_type != VDIR) { 258 error = ENOTDIR; 259 vput(vp); 260 goto out; 261 } 262 263 /* 264 * If the user is not root, ensure that they own the directory 265 * onto which we are attempting to mount. 266 */ 267 if ((error = VOP_GETATTR(vp, &va, l->l_cred, l)) != 0 || 268 (va.va_uid != kauth_cred_geteuid(l->l_cred) && 269 (error = kauth_authorize_generic(l->l_cred, 270 KAUTH_GENERIC_ISSUSER, NULL)) != 0)) { 271 vput(vp); 272 goto out; 273 } 274 275 if (flags & MNT_EXPORTED) { 276 error = EINVAL; 277 vput(vp); 278 goto out; 279 } 280 281 /* 282 * Copy file-system type from userspace. 283 */ 284 error = copyinstr(fstype, fstypename, MFSNAMELEN, NULL); 285 if (error) { 286 #if defined(COMPAT_09) || defined(COMPAT_43) 287 /* 288 * Historically, filesystem types were identified by numbers. 289 * If we get an integer for the filesystem type instead of a 290 * string, we check to see if it matches one of the historic 291 * filesystem types. 292 */ 293 u_long fsindex = (u_long)fstype; 294 if (fsindex >= nmountcompatnames || 295 mountcompatnames[fsindex] == NULL) { 296 error = ENODEV; 297 vput(vp); 298 goto out; 299 } 300 strlcpy(fstypename, mountcompatnames[fsindex], sizeof(fstypename)); 301 #else 302 vput(vp); 303 goto out; 304 #endif 305 } 306 307 #ifdef COMPAT_10 308 /* Accept `ufs' as an alias for `ffs'. */ 309 if (strncmp(fstypename, "ufs", MFSNAMELEN) == 0) 310 strlcpy(fstypename, "ffs", sizeof(fstypename)); 311 #endif 312 313 if ((error = vinvalbuf(vp, V_SAVE, l->l_cred, l, 0, 0)) != 0) { 314 vput(vp); 315 goto out; 316 } 317 318 /* 319 * Check if a file-system is not already mounted on this vnode. 320 */ 321 if (vp->v_mountedhere != NULL) { 322 error = EBUSY; 323 vput(vp); 324 goto out; 325 } 326 327 mp = malloc(sizeof(*mp), M_MOUNT, M_WAITOK|M_ZERO); 328 329 if ((mp->mnt_op = vfs_getopsbyname(fstypename)) == NULL) { 330 free(mp, M_MOUNT); 331 error = ENODEV; 332 vput(vp); 333 goto out; 334 } 335 336 TAILQ_INIT(&mp->mnt_vnodelist); 337 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 338 mutex_init(&mp->mnt_mutex, MUTEX_DEFAULT, IPL_NONE); 339 (void)vfs_busy(mp, LK_NOWAIT, 0); 340 341 mp->mnt_op->vfs_refcount++; 342 mp->mnt_vnodecovered = vp; 343 mp->mnt_stat.f_owner = kauth_cred_geteuid(l->l_cred); 344 mp->mnt_unmounter = NULL; 345 mp->mnt_leaf = mp; 346 mount_initspecific(mp); 347 348 /* 349 * The underlying file system may refuse the mount for 350 * various reasons. Allow the user to force it to happen. 351 * 352 * Set the mount level flags. 353 */ 354 mp->mnt_flag = flags & 355 (MNT_FORCE | MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 356 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 357 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 358 MNT_IGNORE | MNT_RDONLY); 359 360 error = VFS_MOUNT(mp, path, data, ndp, l); 361 mp->mnt_flag &= ~MNT_OP_FLAGS; 362 363 /* 364 * Put the new filesystem on the mount list after root. 365 */ 366 cache_purge(vp); 367 if (!error) { 368 mp->mnt_iflag &= ~IMNT_WANTRDWR; 369 vp->v_mountedhere = mp; 370 mutex_enter(&mountlist_lock); 371 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 372 mutex_exit(&mountlist_lock); 373 VOP_UNLOCK(vp, 0); 374 checkdirs(vp); 375 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 376 error = vfs_allocate_syncvnode(mp); 377 vfs_unbusy(mp); 378 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 379 error = VFS_START(mp, 0, l); 380 if (error) 381 vrele(vp); 382 } else { 383 vp->v_mountedhere = NULL; 384 mp->mnt_op->vfs_refcount--; 385 vfs_unbusy(mp); 386 free(mp, M_MOUNT); 387 vput(vp); 388 } 389 390 out: 391 return (error); 392 } 393 394 static int 395 mount_getargs(struct lwp *l, struct vnode *vp, const char *path, int flags, 396 void *data, struct nameidata *ndp) 397 { 398 struct mount *mp; 399 int error; 400 401 /* If MNT_GETARGS is specified, it should be the only flag. */ 402 if (flags & ~MNT_GETARGS) { 403 error = EINVAL; 404 goto out; 405 } 406 407 mp = vp->v_mount; 408 409 /* XXX: probably some notion of "can see" here if we want isolation. */ 410 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 411 KAUTH_REQ_SYSTEM_MOUNT_GET, mp, data, NULL); 412 if (error) 413 goto out; 414 415 if ((vp->v_flag & VROOT) == 0) { 416 error = EINVAL; 417 goto out; 418 } 419 420 if (vfs_busy(mp, LK_NOWAIT, 0)) { 421 error = EPERM; 422 goto out; 423 } 424 425 mp->mnt_flag &= ~MNT_OP_FLAGS; 426 mp->mnt_flag |= MNT_GETARGS; 427 error = VFS_MOUNT(mp, path, data, ndp, l); 428 mp->mnt_flag &= ~MNT_OP_FLAGS; 429 430 vfs_unbusy(mp); 431 out: 432 return (error); 433 } 434 435 /* ARGSUSED */ 436 int 437 sys_mount(struct lwp *l, void *v, register_t *retval) 438 { 439 struct sys_mount_args /* { 440 syscallarg(const char *) type; 441 syscallarg(const char *) path; 442 syscallarg(int) flags; 443 syscallarg(void *) data; 444 } */ *uap = v; 445 struct vnode *vp; 446 struct nameidata nd; 447 int error; 448 449 /* 450 * Get vnode to be covered 451 */ 452 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 453 SCARG(uap, path), l); 454 if ((error = namei(&nd)) != 0) 455 return (error); 456 vp = nd.ni_vp; 457 458 /* 459 * A lookup in VFS_MOUNT might result in an attempt to 460 * lock this vnode again, so make the lock recursive. 461 */ 462 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 463 464 if (SCARG(uap, flags) & MNT_GETARGS) { 465 error = mount_getargs(l, vp, SCARG(uap, path), 466 SCARG(uap, flags), SCARG(uap, data), &nd); 467 vput(vp); 468 } else if (SCARG(uap, flags) & MNT_UPDATE) { 469 error = mount_update(l, vp, SCARG(uap, path), 470 SCARG(uap, flags), SCARG(uap, data), &nd); 471 vput(vp); 472 } else { 473 /* Locking is handled internally in mount_domount(). */ 474 error = mount_domount(l, vp, SCARG(uap, type), 475 SCARG(uap, path), SCARG(uap, flags), SCARG(uap, data), &nd); 476 } 477 478 return (error); 479 } 480 481 /* 482 * Scan all active processes to see if any of them have a current 483 * or root directory onto which the new filesystem has just been 484 * mounted. If so, replace them with the new mount point. 485 */ 486 void 487 checkdirs(struct vnode *olddp) 488 { 489 struct cwdinfo *cwdi; 490 struct vnode *newdp; 491 struct proc *p; 492 493 if (olddp->v_usecount == 1) 494 return; 495 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 496 panic("mount: lost mount"); 497 mutex_enter(&proclist_lock); 498 PROCLIST_FOREACH(p, &allproc) { 499 cwdi = p->p_cwdi; 500 if (!cwdi) 501 continue; 502 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 503 if (cwdi->cwdi_cdir == olddp) { 504 vrele(cwdi->cwdi_cdir); 505 VREF(newdp); 506 cwdi->cwdi_cdir = newdp; 507 } 508 if (cwdi->cwdi_rdir == olddp) { 509 vrele(cwdi->cwdi_rdir); 510 VREF(newdp); 511 cwdi->cwdi_rdir = newdp; 512 } 513 rw_exit(&cwdi->cwdi_lock); 514 } 515 mutex_exit(&proclist_lock); 516 if (rootvnode == olddp) { 517 vrele(rootvnode); 518 VREF(newdp); 519 rootvnode = newdp; 520 } 521 vput(newdp); 522 } 523 524 /* 525 * Unmount a file system. 526 * 527 * Note: unmount takes a path to the vnode mounted on as argument, 528 * not special file (as before). 529 */ 530 /* ARGSUSED */ 531 int 532 sys_unmount(struct lwp *l, void *v, register_t *retval) 533 { 534 struct sys_unmount_args /* { 535 syscallarg(const char *) path; 536 syscallarg(int) flags; 537 } */ *uap = v; 538 struct vnode *vp; 539 struct mount *mp; 540 int error; 541 struct nameidata nd; 542 543 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 544 SCARG(uap, path), l); 545 if ((error = namei(&nd)) != 0) 546 return (error); 547 vp = nd.ni_vp; 548 mp = vp->v_mount; 549 550 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT, 551 KAUTH_REQ_SYSTEM_MOUNT_UNMOUNT, mp, NULL, NULL); 552 if (error) { 553 vput(vp); 554 return (error); 555 } 556 557 /* 558 * Don't allow unmounting the root file system. 559 */ 560 if (mp->mnt_flag & MNT_ROOTFS) { 561 vput(vp); 562 return (EINVAL); 563 } 564 565 /* 566 * Must be the root of the filesystem 567 */ 568 if ((vp->v_flag & VROOT) == 0) { 569 vput(vp); 570 return (EINVAL); 571 } 572 vput(vp); 573 574 /* 575 * XXX Freeze syncer. Must do this before locking the 576 * mount point. See dounmount() for details. 577 */ 578 mutex_enter(&syncer_mutex); 579 580 if (vfs_busy(mp, 0, 0)) { 581 mutex_exit(&syncer_mutex); 582 return (EBUSY); 583 } 584 585 return (dounmount(mp, SCARG(uap, flags), l)); 586 } 587 588 /* 589 * Do the actual file system unmount. File system is assumed to have been 590 * marked busy by the caller. 591 */ 592 int 593 dounmount(struct mount *mp, int flags, struct lwp *l) 594 { 595 struct vnode *coveredvp; 596 int error; 597 int async; 598 int used_syncer; 599 600 #if NVERIEXEC > 0 601 error = veriexec_unmountchk(mp); 602 if (error) 603 return (error); 604 #endif /* NVERIEXEC > 0 */ 605 606 mutex_enter(&mountlist_lock); 607 vfs_unbusy(mp); 608 used_syncer = (mp->mnt_syncer != NULL); 609 610 /* 611 * XXX Syncer must be frozen when we get here. This should really 612 * be done on a per-mountpoint basis, but especially the softdep 613 * code possibly called from the syncer doesn't exactly work on a 614 * per-mountpoint basis, so the softdep code would become a maze 615 * of vfs_busy() calls. 616 * 617 * The caller of dounmount() must acquire syncer_mutex because 618 * the syncer itself acquires locks in syncer_mutex -> vfs_busy 619 * order, and we must preserve that order to avoid deadlock. 620 * 621 * So, if the file system did not use the syncer, now is 622 * the time to release the syncer_mutex. 623 */ 624 if (used_syncer == 0) 625 mutex_exit(&syncer_mutex); 626 627 mp->mnt_iflag |= IMNT_UNMOUNT; 628 mp->mnt_unmounter = l; 629 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_lock); 630 vn_start_write(NULL, &mp, V_WAIT); 631 632 async = mp->mnt_flag & MNT_ASYNC; 633 mp->mnt_flag &= ~MNT_ASYNC; 634 cache_purgevfs(mp); /* remove cache entries for this file sys */ 635 if (mp->mnt_syncer != NULL) 636 vfs_deallocate_syncvnode(mp); 637 error = 0; 638 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 639 #if NFSS > 0 640 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 641 #endif 642 if (error == 0) 643 error = VFS_SYNC(mp, MNT_WAIT, l->l_cred, l); 644 } 645 if (error == 0 || (flags & MNT_FORCE)) 646 error = VFS_UNMOUNT(mp, flags, l); 647 vn_finished_write(mp, 0); 648 mutex_enter(&mountlist_lock); 649 if (error) { 650 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 651 (void) vfs_allocate_syncvnode(mp); 652 mp->mnt_iflag &= ~IMNT_UNMOUNT; 653 mp->mnt_unmounter = NULL; 654 mp->mnt_flag |= async; 655 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 656 &mountlist_lock); 657 if (used_syncer) 658 mutex_exit(&syncer_mutex); 659 mutex_enter(&mp->mnt_mutex); 660 while (mp->mnt_wcnt > 0) { 661 wakeup(mp); 662 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 663 0, &mp->mnt_mutex); 664 } 665 mutex_exit(&mp->mnt_mutex); 666 return (error); 667 } 668 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 669 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) 670 coveredvp->v_mountedhere = NULL; 671 mp->mnt_op->vfs_refcount--; 672 if (TAILQ_FIRST(&mp->mnt_vnodelist) != NULL) 673 panic("unmount: dangling vnode"); 674 mp->mnt_iflag |= IMNT_GONE; 675 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_lock); 676 if (coveredvp != NULLVP) 677 vrele(coveredvp); 678 mount_finispecific(mp); 679 if (used_syncer) 680 mutex_exit(&syncer_mutex); 681 mutex_enter(&mp->mnt_mutex); 682 while (mp->mnt_wcnt > 0) { 683 wakeup(mp); 684 mtsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_mutex); 685 } 686 mutex_exit(&mp->mnt_mutex); 687 vfs_hooks_unmount(mp); 688 mutex_destroy(&mp->mnt_mutex); 689 free(mp, M_MOUNT); 690 return (0); 691 } 692 693 /* 694 * Sync each mounted filesystem. 695 */ 696 #ifdef DEBUG 697 int syncprt = 0; 698 struct ctldebug debug0 = { "syncprt", &syncprt }; 699 #endif 700 701 /* ARGSUSED */ 702 int 703 sys_sync(struct lwp *l, void *v, register_t *retval) 704 { 705 struct mount *mp, *nmp; 706 int asyncflag; 707 708 if (l == NULL) 709 l = &lwp0; 710 711 mutex_enter(&mountlist_lock); 712 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 713 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 714 nmp = mp->mnt_list.cqe_prev; 715 continue; 716 } 717 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 718 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 719 asyncflag = mp->mnt_flag & MNT_ASYNC; 720 mp->mnt_flag &= ~MNT_ASYNC; 721 VFS_SYNC(mp, MNT_NOWAIT, l->l_cred, l); 722 if (asyncflag) 723 mp->mnt_flag |= MNT_ASYNC; 724 vn_finished_write(mp, 0); 725 } 726 mutex_enter(&mountlist_lock); 727 nmp = mp->mnt_list.cqe_prev; 728 vfs_unbusy(mp); 729 730 } 731 mutex_exit(&mountlist_lock); 732 #ifdef DEBUG 733 if (syncprt) 734 vfs_bufstats(); 735 #endif /* DEBUG */ 736 return (0); 737 } 738 739 /* 740 * Change filesystem quotas. 741 */ 742 /* ARGSUSED */ 743 int 744 sys_quotactl(struct lwp *l, void *v, register_t *retval) 745 { 746 struct sys_quotactl_args /* { 747 syscallarg(const char *) path; 748 syscallarg(int) cmd; 749 syscallarg(int) uid; 750 syscallarg(void *) arg; 751 } */ *uap = v; 752 struct mount *mp; 753 int error; 754 struct nameidata nd; 755 756 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 757 if ((error = namei(&nd)) != 0) 758 return (error); 759 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 760 vrele(nd.ni_vp); 761 if (error) 762 return (error); 763 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 764 SCARG(uap, arg), l); 765 vn_finished_write(mp, 0); 766 return (error); 767 } 768 769 int 770 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 771 int root) 772 { 773 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 774 int error = 0; 775 776 /* 777 * If MNT_NOWAIT or MNT_LAZY is specified, do not 778 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 779 * overrides MNT_NOWAIT. 780 */ 781 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 782 (flags != MNT_WAIT && flags != 0)) { 783 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 784 goto done; 785 } 786 787 /* Get the filesystem stats now */ 788 memset(sp, 0, sizeof(*sp)); 789 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 790 return error; 791 } 792 793 if (cwdi->cwdi_rdir == NULL) 794 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 795 done: 796 if (cwdi->cwdi_rdir != NULL) { 797 size_t len; 798 char *bp; 799 char *path = PNBUF_GET(); 800 if (!path) 801 return ENOMEM; 802 803 bp = path + MAXPATHLEN; 804 *--bp = '\0'; 805 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 806 MAXPATHLEN / 2, 0, l); 807 if (error) { 808 PNBUF_PUT(path); 809 return error; 810 } 811 len = strlen(bp); 812 /* 813 * for mount points that are below our root, we can see 814 * them, so we fix up the pathname and return them. The 815 * rest we cannot see, so we don't allow viewing the 816 * data. 817 */ 818 if (strncmp(bp, sp->f_mntonname, len) == 0) { 819 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 820 sizeof(sp->f_mntonname)); 821 if (sp->f_mntonname[0] == '\0') 822 (void)strlcpy(sp->f_mntonname, "/", 823 sizeof(sp->f_mntonname)); 824 } else { 825 if (root) 826 (void)strlcpy(sp->f_mntonname, "/", 827 sizeof(sp->f_mntonname)); 828 else 829 error = EPERM; 830 } 831 PNBUF_PUT(path); 832 } 833 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 834 return error; 835 } 836 837 /* 838 * Get filesystem statistics. 839 */ 840 /* ARGSUSED */ 841 int 842 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 843 { 844 struct sys_statvfs1_args /* { 845 syscallarg(const char *) path; 846 syscallarg(struct statvfs *) buf; 847 syscallarg(int) flags; 848 } */ *uap = v; 849 struct mount *mp; 850 struct statvfs *sb; 851 int error; 852 struct nameidata nd; 853 854 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 855 if ((error = namei(&nd)) != 0) 856 return error; 857 mp = nd.ni_vp->v_mount; 858 vrele(nd.ni_vp); 859 sb = STATVFSBUF_GET(); 860 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 861 if (error == 0) { 862 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 863 } 864 STATVFSBUF_PUT(sb); 865 return error; 866 } 867 868 /* 869 * Get filesystem statistics. 870 */ 871 /* ARGSUSED */ 872 int 873 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 874 { 875 struct sys_fstatvfs1_args /* { 876 syscallarg(int) fd; 877 syscallarg(struct statvfs *) buf; 878 syscallarg(int) flags; 879 } */ *uap = v; 880 struct proc *p = l->l_proc; 881 struct file *fp; 882 struct mount *mp; 883 struct statvfs *sb; 884 int error; 885 886 /* getvnode() will use the descriptor for us */ 887 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 888 return (error); 889 mp = ((struct vnode *)fp->f_data)->v_mount; 890 sb = STATVFSBUF_GET(); 891 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 892 goto out; 893 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 894 out: 895 FILE_UNUSE(fp, l); 896 STATVFSBUF_PUT(sb); 897 return error; 898 } 899 900 901 /* 902 * Get statistics on all filesystems. 903 */ 904 int 905 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 906 { 907 struct sys_getvfsstat_args /* { 908 syscallarg(struct statvfs *) buf; 909 syscallarg(size_t) bufsize; 910 syscallarg(int) flags; 911 } */ *uap = v; 912 int root = 0; 913 struct proc *p = l->l_proc; 914 struct mount *mp, *nmp; 915 struct statvfs *sb; 916 struct statvfs *sfsp; 917 size_t count, maxcount; 918 int error = 0; 919 920 sb = STATVFSBUF_GET(); 921 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 922 sfsp = SCARG(uap, buf); 923 mutex_enter(&mountlist_lock); 924 count = 0; 925 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 926 mp = nmp) { 927 if (vfs_busy(mp, LK_NOWAIT, &mountlist_lock)) { 928 nmp = CIRCLEQ_NEXT(mp, mnt_list); 929 continue; 930 } 931 if (sfsp && count < maxcount) { 932 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 933 if (error) { 934 mutex_enter(&mountlist_lock); 935 nmp = CIRCLEQ_NEXT(mp, mnt_list); 936 vfs_unbusy(mp); 937 continue; 938 } 939 error = copyout(sb, sfsp, sizeof(*sfsp)); 940 if (error) { 941 vfs_unbusy(mp); 942 goto out; 943 } 944 sfsp++; 945 root |= strcmp(sb->f_mntonname, "/") == 0; 946 } 947 count++; 948 mutex_enter(&mountlist_lock); 949 nmp = CIRCLEQ_NEXT(mp, mnt_list); 950 vfs_unbusy(mp); 951 } 952 mutex_exit(&mountlist_lock); 953 if (root == 0 && p->p_cwdi->cwdi_rdir) { 954 /* 955 * fake a root entry 956 */ 957 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 958 SCARG(uap, flags), 1)) != 0) 959 goto out; 960 if (sfsp) 961 error = copyout(sb, sfsp, sizeof(*sfsp)); 962 count++; 963 } 964 if (sfsp && count > maxcount) 965 *retval = maxcount; 966 else 967 *retval = count; 968 out: 969 STATVFSBUF_PUT(sb); 970 return error; 971 } 972 973 /* 974 * Change current working directory to a given file descriptor. 975 */ 976 /* ARGSUSED */ 977 int 978 sys_fchdir(struct lwp *l, void *v, register_t *retval) 979 { 980 struct sys_fchdir_args /* { 981 syscallarg(int) fd; 982 } */ *uap = v; 983 struct proc *p = l->l_proc; 984 struct filedesc *fdp = p->p_fd; 985 struct cwdinfo *cwdi; 986 struct vnode *vp, *tdp; 987 struct mount *mp; 988 struct file *fp; 989 int error; 990 991 /* getvnode() will use the descriptor for us */ 992 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 993 return (error); 994 vp = (struct vnode *)fp->f_data; 995 996 VREF(vp); 997 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 998 if (vp->v_type != VDIR) 999 error = ENOTDIR; 1000 else 1001 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1002 if (error) { 1003 vput(vp); 1004 goto out; 1005 } 1006 while ((mp = vp->v_mountedhere) != NULL) { 1007 if (vfs_busy(mp, 0, 0)) 1008 continue; 1009 1010 vput(vp); 1011 error = VFS_ROOT(mp, &tdp); 1012 vfs_unbusy(mp); 1013 if (error) 1014 goto out; 1015 vp = tdp; 1016 } 1017 VOP_UNLOCK(vp, 0); 1018 1019 /* 1020 * Disallow changing to a directory not under the process's 1021 * current root directory (if there is one). 1022 */ 1023 cwdi = p->p_cwdi; 1024 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1025 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 1026 vrele(vp); 1027 error = EPERM; /* operation not permitted */ 1028 } else { 1029 vrele(cwdi->cwdi_cdir); 1030 cwdi->cwdi_cdir = vp; 1031 } 1032 rw_exit(&cwdi->cwdi_lock); 1033 1034 out: 1035 FILE_UNUSE(fp, l); 1036 return (error); 1037 } 1038 1039 /* 1040 * Change this process's notion of the root directory to a given file 1041 * descriptor. 1042 */ 1043 int 1044 sys_fchroot(struct lwp *l, void *v, register_t *retval) 1045 { 1046 struct sys_fchroot_args *uap = v; 1047 struct proc *p = l->l_proc; 1048 struct filedesc *fdp = p->p_fd; 1049 struct cwdinfo *cwdi; 1050 struct vnode *vp; 1051 struct file *fp; 1052 int error; 1053 1054 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1055 KAUTH_REQ_SYSTEM_CHROOT_FCHROOT, NULL, NULL, NULL)) != 0) 1056 return error; 1057 /* getvnode() will use the descriptor for us */ 1058 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 1059 return error; 1060 vp = (struct vnode *) fp->f_data; 1061 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1062 if (vp->v_type != VDIR) 1063 error = ENOTDIR; 1064 else 1065 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1066 VOP_UNLOCK(vp, 0); 1067 if (error) 1068 goto out; 1069 VREF(vp); 1070 1071 /* 1072 * Prevent escaping from chroot by putting the root under 1073 * the working directory. Silently chdir to / if we aren't 1074 * already there. 1075 */ 1076 cwdi = p->p_cwdi; 1077 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1078 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1079 /* 1080 * XXX would be more failsafe to change directory to a 1081 * deadfs node here instead 1082 */ 1083 vrele(cwdi->cwdi_cdir); 1084 VREF(vp); 1085 cwdi->cwdi_cdir = vp; 1086 } 1087 1088 if (cwdi->cwdi_rdir != NULL) 1089 vrele(cwdi->cwdi_rdir); 1090 cwdi->cwdi_rdir = vp; 1091 rw_exit(&cwdi->cwdi_lock); 1092 1093 out: 1094 FILE_UNUSE(fp, l); 1095 return (error); 1096 } 1097 1098 /* 1099 * Change current working directory (``.''). 1100 */ 1101 /* ARGSUSED */ 1102 int 1103 sys_chdir(struct lwp *l, void *v, register_t *retval) 1104 { 1105 struct sys_chdir_args /* { 1106 syscallarg(const char *) path; 1107 } */ *uap = v; 1108 struct proc *p = l->l_proc; 1109 struct cwdinfo *cwdi; 1110 int error; 1111 struct nameidata nd; 1112 1113 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1114 SCARG(uap, path), l); 1115 if ((error = change_dir(&nd, l)) != 0) 1116 return (error); 1117 cwdi = p->p_cwdi; 1118 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1119 vrele(cwdi->cwdi_cdir); 1120 cwdi->cwdi_cdir = nd.ni_vp; 1121 rw_exit(&cwdi->cwdi_lock); 1122 return (0); 1123 } 1124 1125 /* 1126 * Change notion of root (``/'') directory. 1127 */ 1128 /* ARGSUSED */ 1129 int 1130 sys_chroot(struct lwp *l, void *v, register_t *retval) 1131 { 1132 struct sys_chroot_args /* { 1133 syscallarg(const char *) path; 1134 } */ *uap = v; 1135 struct proc *p = l->l_proc; 1136 struct cwdinfo *cwdi; 1137 struct vnode *vp; 1138 int error; 1139 struct nameidata nd; 1140 1141 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_CHROOT, 1142 KAUTH_REQ_SYSTEM_CHROOT_CHROOT, NULL, NULL, NULL)) != 0) 1143 return (error); 1144 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1145 SCARG(uap, path), l); 1146 if ((error = change_dir(&nd, l)) != 0) 1147 return (error); 1148 1149 cwdi = p->p_cwdi; 1150 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 1151 if (cwdi->cwdi_rdir != NULL) 1152 vrele(cwdi->cwdi_rdir); 1153 vp = nd.ni_vp; 1154 cwdi->cwdi_rdir = vp; 1155 1156 /* 1157 * Prevent escaping from chroot by putting the root under 1158 * the working directory. Silently chdir to / if we aren't 1159 * already there. 1160 */ 1161 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1162 /* 1163 * XXX would be more failsafe to change directory to a 1164 * deadfs node here instead 1165 */ 1166 vrele(cwdi->cwdi_cdir); 1167 VREF(vp); 1168 cwdi->cwdi_cdir = vp; 1169 } 1170 rw_exit(&cwdi->cwdi_lock); 1171 1172 return (0); 1173 } 1174 1175 /* 1176 * Common routine for chroot and chdir. 1177 */ 1178 static int 1179 change_dir(struct nameidata *ndp, struct lwp *l) 1180 { 1181 struct vnode *vp; 1182 int error; 1183 1184 if ((error = namei(ndp)) != 0) 1185 return (error); 1186 vp = ndp->ni_vp; 1187 if (vp->v_type != VDIR) 1188 error = ENOTDIR; 1189 else 1190 error = VOP_ACCESS(vp, VEXEC, l->l_cred, l); 1191 1192 if (error) 1193 vput(vp); 1194 else 1195 VOP_UNLOCK(vp, 0); 1196 return (error); 1197 } 1198 1199 /* 1200 * Check permissions, allocate an open file structure, 1201 * and call the device open routine if any. 1202 */ 1203 int 1204 sys_open(struct lwp *l, void *v, register_t *retval) 1205 { 1206 struct sys_open_args /* { 1207 syscallarg(const char *) path; 1208 syscallarg(int) flags; 1209 syscallarg(int) mode; 1210 } */ *uap = v; 1211 struct proc *p = l->l_proc; 1212 struct cwdinfo *cwdi = p->p_cwdi; 1213 struct filedesc *fdp = p->p_fd; 1214 struct file *fp; 1215 struct vnode *vp; 1216 int flags, cmode; 1217 int type, indx, error; 1218 struct flock lf; 1219 struct nameidata nd; 1220 1221 flags = FFLAGS(SCARG(uap, flags)); 1222 if ((flags & (FREAD | FWRITE)) == 0) 1223 return (EINVAL); 1224 /* falloc() will use the file descriptor for us */ 1225 if ((error = falloc(l, &fp, &indx)) != 0) 1226 return (error); 1227 /* We're going to read cwdi->cwdi_cmask unlocked here. */ 1228 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1229 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1230 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1231 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1232 FILE_UNUSE(fp, l); 1233 fdp->fd_ofiles[indx] = NULL; 1234 ffree(fp); 1235 if ((error == EDUPFD || error == EMOVEFD) && 1236 l->l_dupfd >= 0 && /* XXX from fdopen */ 1237 (error = 1238 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1239 *retval = indx; 1240 return (0); 1241 } 1242 if (error == ERESTART) 1243 error = EINTR; 1244 fdremove(fdp, indx); 1245 return (error); 1246 } 1247 l->l_dupfd = 0; 1248 vp = nd.ni_vp; 1249 fp->f_flag = flags & FMASK; 1250 fp->f_type = DTYPE_VNODE; 1251 fp->f_ops = &vnops; 1252 fp->f_data = vp; 1253 if (flags & (O_EXLOCK | O_SHLOCK)) { 1254 lf.l_whence = SEEK_SET; 1255 lf.l_start = 0; 1256 lf.l_len = 0; 1257 if (flags & O_EXLOCK) 1258 lf.l_type = F_WRLCK; 1259 else 1260 lf.l_type = F_RDLCK; 1261 type = F_FLOCK; 1262 if ((flags & FNONBLOCK) == 0) 1263 type |= F_WAIT; 1264 VOP_UNLOCK(vp, 0); 1265 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1266 if (error) { 1267 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1268 FILE_UNUSE(fp, l); 1269 ffree(fp); 1270 fdremove(fdp, indx); 1271 return (error); 1272 } 1273 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1274 fp->f_flag |= FHASLOCK; 1275 } 1276 VOP_UNLOCK(vp, 0); 1277 *retval = indx; 1278 FILE_SET_MATURE(fp); 1279 FILE_UNUSE(fp, l); 1280 return (0); 1281 } 1282 1283 static void 1284 vfs__fhfree(fhandle_t *fhp) 1285 { 1286 size_t fhsize; 1287 1288 if (fhp == NULL) { 1289 return; 1290 } 1291 fhsize = FHANDLE_SIZE(fhp); 1292 kmem_free(fhp, fhsize); 1293 } 1294 1295 /* 1296 * vfs_composefh: compose a filehandle. 1297 */ 1298 1299 int 1300 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1301 { 1302 struct mount *mp; 1303 struct fid *fidp; 1304 int error; 1305 size_t needfhsize; 1306 size_t fidsize; 1307 1308 mp = vp->v_mount; 1309 fidp = NULL; 1310 if (*fh_size < FHANDLE_SIZE_MIN) { 1311 fidsize = 0; 1312 } else { 1313 fidsize = *fh_size - offsetof(fhandle_t, fh_fid); 1314 if (fhp != NULL) { 1315 memset(fhp, 0, *fh_size); 1316 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1317 fidp = &fhp->fh_fid; 1318 } 1319 } 1320 error = VFS_VPTOFH(vp, fidp, &fidsize); 1321 needfhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1322 if (error == 0 && *fh_size < needfhsize) { 1323 error = E2BIG; 1324 } 1325 *fh_size = needfhsize; 1326 return error; 1327 } 1328 1329 int 1330 vfs_composefh_alloc(struct vnode *vp, fhandle_t **fhpp) 1331 { 1332 struct mount *mp; 1333 fhandle_t *fhp; 1334 size_t fhsize; 1335 size_t fidsize; 1336 int error; 1337 1338 *fhpp = NULL; 1339 mp = vp->v_mount; 1340 fidsize = 0; 1341 error = VFS_VPTOFH(vp, NULL, &fidsize); 1342 KASSERT(error != 0); 1343 if (error != E2BIG) { 1344 goto out; 1345 } 1346 fhsize = FHANDLE_SIZE_FROM_FILEID_SIZE(fidsize); 1347 fhp = kmem_zalloc(fhsize, KM_SLEEP); 1348 if (fhp == NULL) { 1349 error = ENOMEM; 1350 goto out; 1351 } 1352 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1353 error = VFS_VPTOFH(vp, &fhp->fh_fid, &fidsize); 1354 if (error == 0) { 1355 KASSERT((FHANDLE_SIZE(fhp) == fhsize && 1356 FHANDLE_FILEID(fhp)->fid_len == fidsize)); 1357 *fhpp = fhp; 1358 } else { 1359 kmem_free(fhp, fhsize); 1360 } 1361 out: 1362 return error; 1363 } 1364 1365 void 1366 vfs_composefh_free(fhandle_t *fhp) 1367 { 1368 1369 vfs__fhfree(fhp); 1370 } 1371 1372 /* 1373 * vfs_fhtovp: lookup a vnode by a filehandle. 1374 */ 1375 1376 int 1377 vfs_fhtovp(fhandle_t *fhp, struct vnode **vpp) 1378 { 1379 struct mount *mp; 1380 int error; 1381 1382 *vpp = NULL; 1383 mp = vfs_getvfs(FHANDLE_FSID(fhp)); 1384 if (mp == NULL) { 1385 error = ESTALE; 1386 goto out; 1387 } 1388 if (mp->mnt_op->vfs_fhtovp == NULL) { 1389 error = EOPNOTSUPP; 1390 goto out; 1391 } 1392 error = VFS_FHTOVP(mp, FHANDLE_FILEID(fhp), vpp); 1393 out: 1394 return error; 1395 } 1396 1397 /* 1398 * vfs_copyinfh_alloc: allocate and copyin a filehandle, given 1399 * the needed size. 1400 */ 1401 1402 int 1403 vfs_copyinfh_alloc(const void *ufhp, size_t fhsize, fhandle_t **fhpp) 1404 { 1405 fhandle_t *fhp; 1406 int error; 1407 1408 *fhpp = NULL; 1409 if (fhsize > FHANDLE_SIZE_MAX) { 1410 return EINVAL; 1411 } 1412 if (fhsize < FHANDLE_SIZE_MIN) { 1413 return EINVAL; 1414 } 1415 again: 1416 fhp = kmem_alloc(fhsize, KM_SLEEP); 1417 if (fhp == NULL) { 1418 return ENOMEM; 1419 } 1420 error = copyin(ufhp, fhp, fhsize); 1421 if (error == 0) { 1422 /* XXX this check shouldn't be here */ 1423 if (FHANDLE_SIZE(fhp) == fhsize) { 1424 *fhpp = fhp; 1425 return 0; 1426 } else if (fhsize == NFSX_V2FH && FHANDLE_SIZE(fhp) < fhsize) { 1427 /* 1428 * a kludge for nfsv2 padded handles. 1429 */ 1430 size_t sz; 1431 1432 sz = FHANDLE_SIZE(fhp); 1433 kmem_free(fhp, fhsize); 1434 fhsize = sz; 1435 goto again; 1436 } else { 1437 /* 1438 * userland told us wrong size. 1439 */ 1440 error = EINVAL; 1441 } 1442 } 1443 kmem_free(fhp, fhsize); 1444 return error; 1445 } 1446 1447 void 1448 vfs_copyinfh_free(fhandle_t *fhp) 1449 { 1450 1451 vfs__fhfree(fhp); 1452 } 1453 1454 /* 1455 * Get file handle system call 1456 */ 1457 int 1458 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1459 { 1460 struct sys___getfh30_args /* { 1461 syscallarg(char *) fname; 1462 syscallarg(fhandle_t *) fhp; 1463 syscallarg(size_t *) fh_size; 1464 } */ *uap = v; 1465 struct vnode *vp; 1466 fhandle_t *fh; 1467 int error; 1468 struct nameidata nd; 1469 size_t sz; 1470 size_t usz; 1471 1472 /* 1473 * Must be super user 1474 */ 1475 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1476 0, NULL, NULL, NULL); 1477 if (error) 1478 return (error); 1479 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1480 SCARG(uap, fname), l); 1481 error = namei(&nd); 1482 if (error) 1483 return (error); 1484 vp = nd.ni_vp; 1485 error = vfs_composefh_alloc(vp, &fh); 1486 vput(vp); 1487 if (error != 0) { 1488 goto out; 1489 } 1490 error = copyin(SCARG(uap, fh_size), &usz, sizeof(size_t)); 1491 if (error != 0) { 1492 goto out; 1493 } 1494 sz = FHANDLE_SIZE(fh); 1495 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1496 if (error != 0) { 1497 goto out; 1498 } 1499 if (usz >= sz) { 1500 error = copyout(fh, SCARG(uap, fhp), sz); 1501 } else { 1502 error = E2BIG; 1503 } 1504 out: 1505 vfs_composefh_free(fh); 1506 return (error); 1507 } 1508 1509 /* 1510 * Open a file given a file handle. 1511 * 1512 * Check permissions, allocate an open file structure, 1513 * and call the device open routine if any. 1514 */ 1515 1516 int 1517 dofhopen(struct lwp *l, const void *ufhp, size_t fhsize, int oflags, 1518 register_t *retval) 1519 { 1520 struct filedesc *fdp = l->l_proc->p_fd; 1521 struct file *fp; 1522 struct vnode *vp = NULL; 1523 struct mount *mp; 1524 kauth_cred_t cred = l->l_cred; 1525 struct file *nfp; 1526 int type, indx, error=0; 1527 struct flock lf; 1528 struct vattr va; 1529 fhandle_t *fh; 1530 int flags; 1531 1532 /* 1533 * Must be super user 1534 */ 1535 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1536 0, NULL, NULL, NULL))) 1537 return (error); 1538 1539 flags = FFLAGS(oflags); 1540 if ((flags & (FREAD | FWRITE)) == 0) 1541 return (EINVAL); 1542 if ((flags & O_CREAT)) 1543 return (EINVAL); 1544 /* falloc() will use the file descriptor for us */ 1545 if ((error = falloc(l, &nfp, &indx)) != 0) 1546 return (error); 1547 fp = nfp; 1548 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1549 if (error != 0) { 1550 goto bad; 1551 } 1552 error = vfs_fhtovp(fh, &vp); 1553 if (error != 0) { 1554 goto bad; 1555 } 1556 1557 /* Now do an effective vn_open */ 1558 1559 if (vp->v_type == VSOCK) { 1560 error = EOPNOTSUPP; 1561 goto bad; 1562 } 1563 if (flags & FREAD) { 1564 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1565 goto bad; 1566 } 1567 if (flags & (FWRITE | O_TRUNC)) { 1568 if (vp->v_type == VDIR) { 1569 error = EISDIR; 1570 goto bad; 1571 } 1572 if ((error = vn_writechk(vp)) != 0 || 1573 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1574 goto bad; 1575 } 1576 if (flags & O_TRUNC) { 1577 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1578 goto bad; 1579 VOP_UNLOCK(vp, 0); /* XXX */ 1580 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1581 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1582 VATTR_NULL(&va); 1583 va.va_size = 0; 1584 error = VOP_SETATTR(vp, &va, cred, l); 1585 vn_finished_write(mp, 0); 1586 if (error) 1587 goto bad; 1588 } 1589 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1590 goto bad; 1591 if (vp->v_type == VREG && 1592 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1593 error = EIO; 1594 goto bad; 1595 } 1596 if (flags & FWRITE) 1597 vp->v_writecount++; 1598 1599 /* done with modified vn_open, now finish what sys_open does. */ 1600 1601 fp->f_flag = flags & FMASK; 1602 fp->f_type = DTYPE_VNODE; 1603 fp->f_ops = &vnops; 1604 fp->f_data = vp; 1605 if (flags & (O_EXLOCK | O_SHLOCK)) { 1606 lf.l_whence = SEEK_SET; 1607 lf.l_start = 0; 1608 lf.l_len = 0; 1609 if (flags & O_EXLOCK) 1610 lf.l_type = F_WRLCK; 1611 else 1612 lf.l_type = F_RDLCK; 1613 type = F_FLOCK; 1614 if ((flags & FNONBLOCK) == 0) 1615 type |= F_WAIT; 1616 VOP_UNLOCK(vp, 0); 1617 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1618 if (error) { 1619 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1620 FILE_UNUSE(fp, l); 1621 ffree(fp); 1622 fdremove(fdp, indx); 1623 return (error); 1624 } 1625 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1626 fp->f_flag |= FHASLOCK; 1627 } 1628 VOP_UNLOCK(vp, 0); 1629 *retval = indx; 1630 FILE_SET_MATURE(fp); 1631 FILE_UNUSE(fp, l); 1632 vfs_copyinfh_free(fh); 1633 return (0); 1634 1635 bad: 1636 FILE_UNUSE(fp, l); 1637 ffree(fp); 1638 fdremove(fdp, indx); 1639 if (vp != NULL) 1640 vput(vp); 1641 vfs_copyinfh_free(fh); 1642 return (error); 1643 } 1644 1645 int 1646 sys___fhopen40(struct lwp *l, void *v, register_t *retval) 1647 { 1648 struct sys___fhopen40_args /* { 1649 syscallarg(const void *) fhp; 1650 syscallarg(size_t) fh_size; 1651 syscallarg(int) flags; 1652 } */ *uap = v; 1653 1654 return dofhopen(l, SCARG(uap, fhp), SCARG(uap, fh_size), 1655 SCARG(uap, flags), retval); 1656 } 1657 1658 /* XXX: temp (mar '07) for LKM compat */ 1659 int 1660 dofhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sbp, 1661 register_t *retval) 1662 { 1663 int error; 1664 struct stat sb; 1665 1666 error = do_fhstat(l, ufhp, fhsize, &sb); 1667 if (error == 0) 1668 error = copyout(&sb, sbp, sizeof(sb)); 1669 return error; 1670 } 1671 1672 int 1673 do_fhstat(struct lwp *l, const void *ufhp, size_t fhsize, struct stat *sb) 1674 { 1675 int error; 1676 fhandle_t *fh; 1677 struct vnode *vp; 1678 1679 /* 1680 * Must be super user 1681 */ 1682 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1683 0, NULL, NULL, NULL))) 1684 return (error); 1685 1686 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1687 if (error != 0) 1688 return error; 1689 1690 error = vfs_fhtovp(fh, &vp); 1691 vfs_copyinfh_free(fh); 1692 if (error != 0) 1693 return error; 1694 1695 error = vn_stat(vp, sb, l); 1696 vput(vp); 1697 return error; 1698 } 1699 1700 1701 /* ARGSUSED */ 1702 int 1703 sys___fhstat40(struct lwp *l, void *v, register_t *retval) 1704 { 1705 struct sys___fhstat40_args /* { 1706 syscallarg(const void *) fhp; 1707 syscallarg(size_t) fh_size; 1708 syscallarg(struct stat *) sb; 1709 } */ *uap = v; 1710 struct stat sb; 1711 int error; 1712 1713 error = do_fhstat(l, SCARG(uap, fhp), SCARG(uap, fh_size), &sb); 1714 if (error) 1715 return error; 1716 return copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1717 } 1718 1719 /* XXX: temp (mar '07) for LKM compat */ 1720 int 1721 dofhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *buf, 1722 int flags, register_t *retval) 1723 { 1724 struct statvfs *sb = STATVFSBUF_GET(); 1725 int error; 1726 1727 error = do_fhstatvfs(l, ufhp, fhsize, sb, flags); 1728 if (error == 0) 1729 error = copyout(sb, buf, sizeof(*sb)); 1730 STATVFSBUF_PUT(sb); 1731 return error; 1732 } 1733 1734 int 1735 do_fhstatvfs(struct lwp *l, const void *ufhp, size_t fhsize, struct statvfs *sb, 1736 int flags) 1737 { 1738 fhandle_t *fh; 1739 struct mount *mp; 1740 struct vnode *vp; 1741 int error; 1742 1743 /* 1744 * Must be super user 1745 */ 1746 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FILEHANDLE, 1747 0, NULL, NULL, NULL))) 1748 return error; 1749 1750 error = vfs_copyinfh_alloc(ufhp, fhsize, &fh); 1751 if (error != 0) 1752 return error; 1753 1754 error = vfs_fhtovp(fh, &vp); 1755 vfs_copyinfh_free(fh); 1756 if (error != 0) 1757 return error; 1758 1759 mp = vp->v_mount; 1760 error = dostatvfs(mp, sb, l, flags, 1); 1761 vput(vp); 1762 return error; 1763 } 1764 1765 /* ARGSUSED */ 1766 int 1767 sys___fhstatvfs140(struct lwp *l, void *v, register_t *retval) 1768 { 1769 struct sys___fhstatvfs140_args /* { 1770 syscallarg(const void *) fhp; 1771 syscallarg(size_t) fh_size; 1772 syscallarg(struct statvfs *) buf; 1773 syscallarg(int) flags; 1774 } */ *uap = v; 1775 struct statvfs *sb = STATVFSBUF_GET(); 1776 int error; 1777 1778 error = do_fhstatvfs(l, SCARG(uap, fhp), SCARG(uap, fh_size), sb, 1779 SCARG(uap, flags)); 1780 if (error == 0) 1781 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1782 STATVFSBUF_PUT(sb); 1783 return error; 1784 } 1785 1786 /* 1787 * Create a special file. 1788 */ 1789 /* ARGSUSED */ 1790 int 1791 sys_mknod(struct lwp *l, void *v, register_t *retval) 1792 { 1793 struct sys_mknod_args /* { 1794 syscallarg(const char *) path; 1795 syscallarg(int) mode; 1796 syscallarg(int) dev; 1797 } */ *uap = v; 1798 struct proc *p = l->l_proc; 1799 struct vnode *vp; 1800 struct mount *mp; 1801 struct vattr vattr; 1802 int error, optype; 1803 struct nameidata nd; 1804 1805 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MKNOD, 1806 0, NULL, NULL, NULL)) != 0) 1807 return (error); 1808 1809 optype = VOP_MKNOD_DESCOFFSET; 1810 restart: 1811 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1812 if ((error = namei(&nd)) != 0) 1813 return (error); 1814 vp = nd.ni_vp; 1815 if (vp != NULL) 1816 error = EEXIST; 1817 else { 1818 VATTR_NULL(&vattr); 1819 /* We will read cwdi->cwdi_cmask unlocked. */ 1820 vattr.va_mode = 1821 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1822 vattr.va_rdev = SCARG(uap, dev); 1823 1824 switch (SCARG(uap, mode) & S_IFMT) { 1825 case S_IFMT: /* used by badsect to flag bad sectors */ 1826 vattr.va_type = VBAD; 1827 break; 1828 case S_IFCHR: 1829 vattr.va_type = VCHR; 1830 break; 1831 case S_IFBLK: 1832 vattr.va_type = VBLK; 1833 break; 1834 case S_IFWHT: 1835 optype = VOP_WHITEOUT_DESCOFFSET; 1836 break; 1837 case S_IFREG: 1838 vattr.va_type = VREG; 1839 vattr.va_rdev = VNOVAL; 1840 optype = VOP_CREATE_DESCOFFSET; 1841 break; 1842 default: 1843 error = EINVAL; 1844 break; 1845 } 1846 } 1847 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1848 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1849 if (nd.ni_dvp == vp) 1850 vrele(nd.ni_dvp); 1851 else 1852 vput(nd.ni_dvp); 1853 if (vp) 1854 vrele(vp); 1855 if ((error = vn_start_write(NULL, &mp, 1856 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1857 return (error); 1858 goto restart; 1859 } 1860 if (!error) { 1861 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1862 switch (optype) { 1863 case VOP_WHITEOUT_DESCOFFSET: 1864 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1865 if (error) 1866 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1867 vput(nd.ni_dvp); 1868 break; 1869 1870 case VOP_MKNOD_DESCOFFSET: 1871 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1872 &nd.ni_cnd, &vattr); 1873 if (error == 0) 1874 vput(nd.ni_vp); 1875 break; 1876 1877 case VOP_CREATE_DESCOFFSET: 1878 error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, 1879 &nd.ni_cnd, &vattr); 1880 if (error == 0) 1881 vput(nd.ni_vp); 1882 break; 1883 } 1884 } else { 1885 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1886 if (nd.ni_dvp == vp) 1887 vrele(nd.ni_dvp); 1888 else 1889 vput(nd.ni_dvp); 1890 if (vp) 1891 vrele(vp); 1892 } 1893 vn_finished_write(mp, 0); 1894 return (error); 1895 } 1896 1897 /* 1898 * Create a named pipe. 1899 */ 1900 /* ARGSUSED */ 1901 int 1902 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1903 { 1904 struct sys_mkfifo_args /* { 1905 syscallarg(const char *) path; 1906 syscallarg(int) mode; 1907 } */ *uap = v; 1908 struct proc *p = l->l_proc; 1909 struct mount *mp; 1910 struct vattr vattr; 1911 int error; 1912 struct nameidata nd; 1913 1914 restart: 1915 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1916 if ((error = namei(&nd)) != 0) 1917 return (error); 1918 if (nd.ni_vp != NULL) { 1919 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1920 if (nd.ni_dvp == nd.ni_vp) 1921 vrele(nd.ni_dvp); 1922 else 1923 vput(nd.ni_dvp); 1924 vrele(nd.ni_vp); 1925 return (EEXIST); 1926 } 1927 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1928 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1929 if (nd.ni_dvp == nd.ni_vp) 1930 vrele(nd.ni_dvp); 1931 else 1932 vput(nd.ni_dvp); 1933 if (nd.ni_vp) 1934 vrele(nd.ni_vp); 1935 if ((error = vn_start_write(NULL, &mp, 1936 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1937 return (error); 1938 goto restart; 1939 } 1940 VATTR_NULL(&vattr); 1941 vattr.va_type = VFIFO; 1942 /* We will read cwdi->cwdi_cmask unlocked. */ 1943 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1944 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1945 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1946 if (error == 0) 1947 vput(nd.ni_vp); 1948 vn_finished_write(mp, 0); 1949 return (error); 1950 } 1951 1952 /* 1953 * Make a hard file link. 1954 */ 1955 /* ARGSUSED */ 1956 int 1957 sys_link(struct lwp *l, void *v, register_t *retval) 1958 { 1959 struct sys_link_args /* { 1960 syscallarg(const char *) path; 1961 syscallarg(const char *) link; 1962 } */ *uap = v; 1963 struct vnode *vp; 1964 struct mount *mp; 1965 struct nameidata nd; 1966 int error; 1967 1968 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1969 if ((error = namei(&nd)) != 0) 1970 return (error); 1971 vp = nd.ni_vp; 1972 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1973 vrele(vp); 1974 return (error); 1975 } 1976 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1977 if ((error = namei(&nd)) != 0) 1978 goto out; 1979 if (nd.ni_vp) { 1980 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1981 if (nd.ni_dvp == nd.ni_vp) 1982 vrele(nd.ni_dvp); 1983 else 1984 vput(nd.ni_dvp); 1985 vrele(nd.ni_vp); 1986 error = EEXIST; 1987 goto out; 1988 } 1989 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 1990 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 1991 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1992 out: 1993 vrele(vp); 1994 vn_finished_write(mp, 0); 1995 return (error); 1996 } 1997 1998 /* 1999 * Make a symbolic link. 2000 */ 2001 /* ARGSUSED */ 2002 int 2003 sys_symlink(struct lwp *l, void *v, register_t *retval) 2004 { 2005 struct sys_symlink_args /* { 2006 syscallarg(const char *) path; 2007 syscallarg(const char *) link; 2008 } */ *uap = v; 2009 struct proc *p = l->l_proc; 2010 struct mount *mp; 2011 struct vattr vattr; 2012 char *path; 2013 int error; 2014 struct nameidata nd; 2015 2016 path = PNBUF_GET(); 2017 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 2018 if (error) 2019 goto out; 2020 restart: 2021 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 2022 if ((error = namei(&nd)) != 0) 2023 goto out; 2024 if (nd.ni_vp) { 2025 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2026 if (nd.ni_dvp == nd.ni_vp) 2027 vrele(nd.ni_dvp); 2028 else 2029 vput(nd.ni_dvp); 2030 vrele(nd.ni_vp); 2031 error = EEXIST; 2032 goto out; 2033 } 2034 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2035 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2036 if (nd.ni_dvp == nd.ni_vp) 2037 vrele(nd.ni_dvp); 2038 else 2039 vput(nd.ni_dvp); 2040 if ((error = vn_start_write(NULL, &mp, 2041 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2042 return (error); 2043 goto restart; 2044 } 2045 VATTR_NULL(&vattr); 2046 vattr.va_type = VLNK; 2047 /* We will read cwdi->cwdi_cmask unlocked. */ 2048 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 2049 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2050 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 2051 if (error == 0) 2052 vput(nd.ni_vp); 2053 vn_finished_write(mp, 0); 2054 out: 2055 PNBUF_PUT(path); 2056 return (error); 2057 } 2058 2059 /* 2060 * Delete a whiteout from the filesystem. 2061 */ 2062 /* ARGSUSED */ 2063 int 2064 sys_undelete(struct lwp *l, void *v, register_t *retval) 2065 { 2066 struct sys_undelete_args /* { 2067 syscallarg(const char *) path; 2068 } */ *uap = v; 2069 int error; 2070 struct mount *mp; 2071 struct nameidata nd; 2072 2073 restart: 2074 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 2075 SCARG(uap, path), l); 2076 error = namei(&nd); 2077 if (error) 2078 return (error); 2079 2080 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 2081 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2082 if (nd.ni_dvp == nd.ni_vp) 2083 vrele(nd.ni_dvp); 2084 else 2085 vput(nd.ni_dvp); 2086 if (nd.ni_vp) 2087 vrele(nd.ni_vp); 2088 return (EEXIST); 2089 } 2090 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2091 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2092 if (nd.ni_dvp == nd.ni_vp) 2093 vrele(nd.ni_dvp); 2094 else 2095 vput(nd.ni_dvp); 2096 if ((error = vn_start_write(NULL, &mp, 2097 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2098 return (error); 2099 goto restart; 2100 } 2101 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2102 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 2103 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2104 vput(nd.ni_dvp); 2105 vn_finished_write(mp, 0); 2106 return (error); 2107 } 2108 2109 /* 2110 * Delete a name from the filesystem. 2111 */ 2112 /* ARGSUSED */ 2113 int 2114 sys_unlink(struct lwp *l, void *v, register_t *retval) 2115 { 2116 struct sys_unlink_args /* { 2117 syscallarg(const char *) path; 2118 } */ *uap = v; 2119 struct mount *mp; 2120 struct vnode *vp; 2121 int error; 2122 struct nameidata nd; 2123 #if NVERIEXEC > 0 2124 pathname_t pathbuf = NULL; 2125 #endif /* NVERIEXEC > 0 */ 2126 2127 restart: 2128 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 2129 SCARG(uap, path), l); 2130 if ((error = namei(&nd)) != 0) 2131 return (error); 2132 vp = nd.ni_vp; 2133 2134 /* 2135 * The root of a mounted filesystem cannot be deleted. 2136 */ 2137 if (vp->v_flag & VROOT) { 2138 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2139 if (nd.ni_dvp == vp) 2140 vrele(nd.ni_dvp); 2141 else 2142 vput(nd.ni_dvp); 2143 vput(vp); 2144 error = EBUSY; 2145 goto out; 2146 } 2147 2148 #if NVERIEXEC > 0 2149 error = pathname_get(nd.ni_dirp, nd.ni_segflg, &pathbuf); 2150 2151 /* Handle remove requests for veriexec entries. */ 2152 if (!error) { 2153 error = veriexec_removechk(vp, pathname_path(pathbuf), l); 2154 pathname_put(pathbuf); 2155 } 2156 2157 if (error) { 2158 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2159 if (nd.ni_dvp == vp) 2160 vrele(nd.ni_dvp); 2161 else 2162 vput(nd.ni_dvp); 2163 vput(vp); 2164 goto out; 2165 } 2166 #endif /* NVERIEXEC > 0 */ 2167 2168 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 2169 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 2170 if (nd.ni_dvp == vp) 2171 vrele(nd.ni_dvp); 2172 else 2173 vput(nd.ni_dvp); 2174 vput(vp); 2175 if ((error = vn_start_write(NULL, &mp, 2176 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 2177 return (error); 2178 goto restart; 2179 } 2180 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 2181 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2182 #ifdef FILEASSOC 2183 (void)fileassoc_file_delete(vp); 2184 #endif /* FILEASSOC */ 2185 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 2186 vn_finished_write(mp, 0); 2187 out: 2188 return (error); 2189 } 2190 2191 /* 2192 * Reposition read/write file offset. 2193 */ 2194 int 2195 sys_lseek(struct lwp *l, void *v, register_t *retval) 2196 { 2197 struct sys_lseek_args /* { 2198 syscallarg(int) fd; 2199 syscallarg(int) pad; 2200 syscallarg(off_t) offset; 2201 syscallarg(int) whence; 2202 } */ *uap = v; 2203 struct proc *p = l->l_proc; 2204 kauth_cred_t cred = l->l_cred; 2205 struct filedesc *fdp = p->p_fd; 2206 struct file *fp; 2207 struct vnode *vp; 2208 struct vattr vattr; 2209 off_t newoff; 2210 int error; 2211 2212 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 2213 return (EBADF); 2214 2215 vp = (struct vnode *)fp->f_data; 2216 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2217 error = ESPIPE; 2218 mutex_exit(&fp->f_lock); 2219 goto out; 2220 } 2221 2222 switch (SCARG(uap, whence)) { 2223 case SEEK_CUR: 2224 newoff = fp->f_offset + SCARG(uap, offset); 2225 FILE_USE(fp); 2226 break; 2227 case SEEK_END: 2228 FILE_USE(fp); 2229 error = VOP_GETATTR(vp, &vattr, cred, l); 2230 if (error) { 2231 FILE_UNUSE(fp, l); 2232 goto out; 2233 } 2234 newoff = SCARG(uap, offset) + vattr.va_size; 2235 break; 2236 case SEEK_SET: 2237 FILE_USE(fp); 2238 newoff = SCARG(uap, offset); 2239 break; 2240 default: 2241 mutex_exit(&fp->f_lock); 2242 error = EINVAL; 2243 goto out; 2244 } 2245 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) == 0) { 2246 mutex_enter(&fp->f_lock); 2247 *(off_t *)retval = fp->f_offset = newoff; 2248 mutex_exit(&fp->f_lock); 2249 } 2250 FILE_UNUSE(fp, l); 2251 out: 2252 return (error); 2253 } 2254 2255 /* 2256 * Positional read system call. 2257 */ 2258 int 2259 sys_pread(struct lwp *l, void *v, register_t *retval) 2260 { 2261 struct sys_pread_args /* { 2262 syscallarg(int) fd; 2263 syscallarg(void *) buf; 2264 syscallarg(size_t) nbyte; 2265 syscallarg(off_t) offset; 2266 } */ *uap = v; 2267 struct proc *p = l->l_proc; 2268 struct filedesc *fdp = p->p_fd; 2269 struct file *fp; 2270 struct vnode *vp; 2271 off_t offset; 2272 int error, fd = SCARG(uap, fd); 2273 2274 if ((fp = fd_getfile(fdp, fd)) == NULL) 2275 return (EBADF); 2276 2277 if ((fp->f_flag & FREAD) == 0) { 2278 mutex_exit(&fp->f_lock); 2279 return (EBADF); 2280 } 2281 2282 FILE_USE(fp); 2283 2284 vp = (struct vnode *)fp->f_data; 2285 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2286 error = ESPIPE; 2287 goto out; 2288 } 2289 2290 offset = SCARG(uap, offset); 2291 2292 /* 2293 * XXX This works because no file systems actually 2294 * XXX take any action on the seek operation. 2295 */ 2296 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2297 goto out; 2298 2299 /* dofileread() will unuse the descriptor for us */ 2300 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2301 &offset, 0, retval)); 2302 2303 out: 2304 FILE_UNUSE(fp, l); 2305 return (error); 2306 } 2307 2308 /* 2309 * Positional scatter read system call. 2310 */ 2311 int 2312 sys_preadv(struct lwp *l, void *v, register_t *retval) 2313 { 2314 struct sys_preadv_args /* { 2315 syscallarg(int) fd; 2316 syscallarg(const struct iovec *) iovp; 2317 syscallarg(int) iovcnt; 2318 syscallarg(off_t) offset; 2319 } */ *uap = v; 2320 struct proc *p = l->l_proc; 2321 struct filedesc *fdp = p->p_fd; 2322 struct file *fp; 2323 struct vnode *vp; 2324 off_t offset; 2325 int error, fd = SCARG(uap, fd); 2326 2327 if ((fp = fd_getfile(fdp, fd)) == NULL) 2328 return (EBADF); 2329 2330 if ((fp->f_flag & FREAD) == 0) { 2331 mutex_exit(&fp->f_lock); 2332 return (EBADF); 2333 } 2334 2335 FILE_USE(fp); 2336 2337 vp = (struct vnode *)fp->f_data; 2338 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2339 error = ESPIPE; 2340 goto out; 2341 } 2342 2343 offset = SCARG(uap, offset); 2344 2345 /* 2346 * XXX This works because no file systems actually 2347 * XXX take any action on the seek operation. 2348 */ 2349 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2350 goto out; 2351 2352 /* dofilereadv() will unuse the descriptor for us */ 2353 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2354 &offset, 0, retval)); 2355 2356 out: 2357 FILE_UNUSE(fp, l); 2358 return (error); 2359 } 2360 2361 /* 2362 * Positional write system call. 2363 */ 2364 int 2365 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2366 { 2367 struct sys_pwrite_args /* { 2368 syscallarg(int) fd; 2369 syscallarg(const void *) buf; 2370 syscallarg(size_t) nbyte; 2371 syscallarg(off_t) offset; 2372 } */ *uap = v; 2373 struct proc *p = l->l_proc; 2374 struct filedesc *fdp = p->p_fd; 2375 struct file *fp; 2376 struct vnode *vp; 2377 off_t offset; 2378 int error, fd = SCARG(uap, fd); 2379 2380 if ((fp = fd_getfile(fdp, fd)) == NULL) 2381 return (EBADF); 2382 2383 if ((fp->f_flag & FWRITE) == 0) { 2384 mutex_exit(&fp->f_lock); 2385 return (EBADF); 2386 } 2387 2388 FILE_USE(fp); 2389 2390 vp = (struct vnode *)fp->f_data; 2391 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2392 error = ESPIPE; 2393 goto out; 2394 } 2395 2396 offset = SCARG(uap, offset); 2397 2398 /* 2399 * XXX This works because no file systems actually 2400 * XXX take any action on the seek operation. 2401 */ 2402 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2403 goto out; 2404 2405 /* dofilewrite() will unuse the descriptor for us */ 2406 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2407 &offset, 0, retval)); 2408 2409 out: 2410 FILE_UNUSE(fp, l); 2411 return (error); 2412 } 2413 2414 /* 2415 * Positional gather write system call. 2416 */ 2417 int 2418 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2419 { 2420 struct sys_pwritev_args /* { 2421 syscallarg(int) fd; 2422 syscallarg(const struct iovec *) iovp; 2423 syscallarg(int) iovcnt; 2424 syscallarg(off_t) offset; 2425 } */ *uap = v; 2426 struct proc *p = l->l_proc; 2427 struct filedesc *fdp = p->p_fd; 2428 struct file *fp; 2429 struct vnode *vp; 2430 off_t offset; 2431 int error, fd = SCARG(uap, fd); 2432 2433 if ((fp = fd_getfile(fdp, fd)) == NULL) 2434 return (EBADF); 2435 2436 if ((fp->f_flag & FWRITE) == 0) { 2437 mutex_exit(&fp->f_lock); 2438 return (EBADF); 2439 } 2440 2441 FILE_USE(fp); 2442 2443 vp = (struct vnode *)fp->f_data; 2444 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2445 error = ESPIPE; 2446 goto out; 2447 } 2448 2449 offset = SCARG(uap, offset); 2450 2451 /* 2452 * XXX This works because no file systems actually 2453 * XXX take any action on the seek operation. 2454 */ 2455 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2456 goto out; 2457 2458 /* dofilewritev() will unuse the descriptor for us */ 2459 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2460 &offset, 0, retval)); 2461 2462 out: 2463 FILE_UNUSE(fp, l); 2464 return (error); 2465 } 2466 2467 /* 2468 * Check access permissions. 2469 */ 2470 int 2471 sys_access(struct lwp *l, void *v, register_t *retval) 2472 { 2473 struct sys_access_args /* { 2474 syscallarg(const char *) path; 2475 syscallarg(int) flags; 2476 } */ *uap = v; 2477 kauth_cred_t cred; 2478 struct vnode *vp; 2479 int error, flags; 2480 struct nameidata nd; 2481 2482 cred = kauth_cred_dup(l->l_cred); 2483 kauth_cred_seteuid(cred, kauth_cred_getuid(l->l_cred)); 2484 kauth_cred_setegid(cred, kauth_cred_getgid(l->l_cred)); 2485 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2486 SCARG(uap, path), l); 2487 /* Override default credentials */ 2488 nd.ni_cnd.cn_cred = cred; 2489 if ((error = namei(&nd)) != 0) 2490 goto out; 2491 vp = nd.ni_vp; 2492 2493 /* Flags == 0 means only check for existence. */ 2494 if (SCARG(uap, flags)) { 2495 flags = 0; 2496 if (SCARG(uap, flags) & R_OK) 2497 flags |= VREAD; 2498 if (SCARG(uap, flags) & W_OK) 2499 flags |= VWRITE; 2500 if (SCARG(uap, flags) & X_OK) 2501 flags |= VEXEC; 2502 2503 error = VOP_ACCESS(vp, flags, cred, l); 2504 if (!error && (flags & VWRITE)) 2505 error = vn_writechk(vp); 2506 } 2507 vput(vp); 2508 out: 2509 kauth_cred_free(cred); 2510 return (error); 2511 } 2512 2513 /* 2514 * Common code for all sys_stat functions, including compat versions. 2515 */ 2516 int 2517 do_sys_stat(struct lwp *l, const char *path, unsigned int nd_flags, 2518 struct stat *sb) 2519 { 2520 int error; 2521 struct nameidata nd; 2522 2523 NDINIT(&nd, LOOKUP, nd_flags | LOCKLEAF, UIO_USERSPACE, path, l); 2524 error = namei(&nd); 2525 if (error != 0) 2526 return error; 2527 error = vn_stat(nd.ni_vp, sb, l); 2528 vput(nd.ni_vp); 2529 return error; 2530 } 2531 2532 /* 2533 * Get file status; this version follows links. 2534 */ 2535 /* ARGSUSED */ 2536 int 2537 sys___stat30(struct lwp *l, void *v, register_t *retval) 2538 { 2539 struct sys___stat30_args /* { 2540 syscallarg(const char *) path; 2541 syscallarg(struct stat *) ub; 2542 } */ *uap = v; 2543 struct stat sb; 2544 int error; 2545 2546 error = do_sys_stat(l, SCARG(uap, path), FOLLOW, &sb); 2547 if (error) 2548 return error; 2549 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2550 } 2551 2552 /* 2553 * Get file status; this version does not follow links. 2554 */ 2555 /* ARGSUSED */ 2556 int 2557 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2558 { 2559 struct sys___lstat30_args /* { 2560 syscallarg(const char *) path; 2561 syscallarg(struct stat *) ub; 2562 } */ *uap = v; 2563 struct stat sb; 2564 int error; 2565 2566 error = do_sys_stat(l, SCARG(uap, path), NOFOLLOW, &sb); 2567 if (error) 2568 return error; 2569 return copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2570 } 2571 2572 /* 2573 * Get configurable pathname variables. 2574 */ 2575 /* ARGSUSED */ 2576 int 2577 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2578 { 2579 struct sys_pathconf_args /* { 2580 syscallarg(const char *) path; 2581 syscallarg(int) name; 2582 } */ *uap = v; 2583 int error; 2584 struct nameidata nd; 2585 2586 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2587 SCARG(uap, path), l); 2588 if ((error = namei(&nd)) != 0) 2589 return (error); 2590 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2591 vput(nd.ni_vp); 2592 return (error); 2593 } 2594 2595 /* 2596 * Return target name of a symbolic link. 2597 */ 2598 /* ARGSUSED */ 2599 int 2600 sys_readlink(struct lwp *l, void *v, register_t *retval) 2601 { 2602 struct sys_readlink_args /* { 2603 syscallarg(const char *) path; 2604 syscallarg(char *) buf; 2605 syscallarg(size_t) count; 2606 } */ *uap = v; 2607 struct vnode *vp; 2608 struct iovec aiov; 2609 struct uio auio; 2610 int error; 2611 struct nameidata nd; 2612 2613 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2614 SCARG(uap, path), l); 2615 if ((error = namei(&nd)) != 0) 2616 return (error); 2617 vp = nd.ni_vp; 2618 if (vp->v_type != VLNK) 2619 error = EINVAL; 2620 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2621 (error = VOP_ACCESS(vp, VREAD, l->l_cred, l)) == 0) { 2622 aiov.iov_base = SCARG(uap, buf); 2623 aiov.iov_len = SCARG(uap, count); 2624 auio.uio_iov = &aiov; 2625 auio.uio_iovcnt = 1; 2626 auio.uio_offset = 0; 2627 auio.uio_rw = UIO_READ; 2628 KASSERT(l == curlwp); 2629 auio.uio_vmspace = l->l_proc->p_vmspace; 2630 auio.uio_resid = SCARG(uap, count); 2631 error = VOP_READLINK(vp, &auio, l->l_cred); 2632 } 2633 vput(vp); 2634 *retval = SCARG(uap, count) - auio.uio_resid; 2635 return (error); 2636 } 2637 2638 /* 2639 * Change flags of a file given a path name. 2640 */ 2641 /* ARGSUSED */ 2642 int 2643 sys_chflags(struct lwp *l, void *v, register_t *retval) 2644 { 2645 struct sys_chflags_args /* { 2646 syscallarg(const char *) path; 2647 syscallarg(u_long) flags; 2648 } */ *uap = v; 2649 struct vnode *vp; 2650 int error; 2651 struct nameidata nd; 2652 2653 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2654 if ((error = namei(&nd)) != 0) 2655 return (error); 2656 vp = nd.ni_vp; 2657 error = change_flags(vp, SCARG(uap, flags), l); 2658 vput(vp); 2659 return (error); 2660 } 2661 2662 /* 2663 * Change flags of a file given a file descriptor. 2664 */ 2665 /* ARGSUSED */ 2666 int 2667 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2668 { 2669 struct sys_fchflags_args /* { 2670 syscallarg(int) fd; 2671 syscallarg(u_long) flags; 2672 } */ *uap = v; 2673 struct proc *p = l->l_proc; 2674 struct vnode *vp; 2675 struct file *fp; 2676 int error; 2677 2678 /* getvnode() will use the descriptor for us */ 2679 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2680 return (error); 2681 vp = (struct vnode *)fp->f_data; 2682 error = change_flags(vp, SCARG(uap, flags), l); 2683 VOP_UNLOCK(vp, 0); 2684 FILE_UNUSE(fp, l); 2685 return (error); 2686 } 2687 2688 /* 2689 * Change flags of a file given a path name; this version does 2690 * not follow links. 2691 */ 2692 int 2693 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2694 { 2695 struct sys_lchflags_args /* { 2696 syscallarg(const char *) path; 2697 syscallarg(u_long) flags; 2698 } */ *uap = v; 2699 struct vnode *vp; 2700 int error; 2701 struct nameidata nd; 2702 2703 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2704 if ((error = namei(&nd)) != 0) 2705 return (error); 2706 vp = nd.ni_vp; 2707 error = change_flags(vp, SCARG(uap, flags), l); 2708 vput(vp); 2709 return (error); 2710 } 2711 2712 /* 2713 * Common routine to change flags of a file. 2714 */ 2715 int 2716 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2717 { 2718 struct mount *mp; 2719 struct vattr vattr; 2720 int error; 2721 2722 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2723 return (error); 2724 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2725 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2726 /* 2727 * Non-superusers cannot change the flags on devices, even if they 2728 * own them. 2729 */ 2730 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 2731 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 2732 goto out; 2733 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2734 error = EINVAL; 2735 goto out; 2736 } 2737 } 2738 VATTR_NULL(&vattr); 2739 vattr.va_flags = flags; 2740 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2741 out: 2742 vn_finished_write(mp, 0); 2743 return (error); 2744 } 2745 2746 /* 2747 * Change mode of a file given path name; this version follows links. 2748 */ 2749 /* ARGSUSED */ 2750 int 2751 sys_chmod(struct lwp *l, void *v, register_t *retval) 2752 { 2753 struct sys_chmod_args /* { 2754 syscallarg(const char *) path; 2755 syscallarg(int) mode; 2756 } */ *uap = v; 2757 int error; 2758 struct nameidata nd; 2759 2760 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2761 if ((error = namei(&nd)) != 0) 2762 return (error); 2763 2764 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2765 2766 vrele(nd.ni_vp); 2767 return (error); 2768 } 2769 2770 /* 2771 * Change mode of a file given a file descriptor. 2772 */ 2773 /* ARGSUSED */ 2774 int 2775 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2776 { 2777 struct sys_fchmod_args /* { 2778 syscallarg(int) fd; 2779 syscallarg(int) mode; 2780 } */ *uap = v; 2781 struct proc *p = l->l_proc; 2782 struct file *fp; 2783 int error; 2784 2785 /* getvnode() will use the descriptor for us */ 2786 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2787 return (error); 2788 2789 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2790 FILE_UNUSE(fp, l); 2791 return (error); 2792 } 2793 2794 /* 2795 * Change mode of a file given path name; this version does not follow links. 2796 */ 2797 /* ARGSUSED */ 2798 int 2799 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2800 { 2801 struct sys_lchmod_args /* { 2802 syscallarg(const char *) path; 2803 syscallarg(int) mode; 2804 } */ *uap = v; 2805 int error; 2806 struct nameidata nd; 2807 2808 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2809 if ((error = namei(&nd)) != 0) 2810 return (error); 2811 2812 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2813 2814 vrele(nd.ni_vp); 2815 return (error); 2816 } 2817 2818 /* 2819 * Common routine to set mode given a vnode. 2820 */ 2821 static int 2822 change_mode(struct vnode *vp, int mode, struct lwp *l) 2823 { 2824 struct mount *mp; 2825 struct vattr vattr; 2826 int error; 2827 2828 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2829 return (error); 2830 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 2831 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2832 VATTR_NULL(&vattr); 2833 vattr.va_mode = mode & ALLPERMS; 2834 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 2835 VOP_UNLOCK(vp, 0); 2836 vn_finished_write(mp, 0); 2837 return (error); 2838 } 2839 2840 /* 2841 * Set ownership given a path name; this version follows links. 2842 */ 2843 /* ARGSUSED */ 2844 int 2845 sys_chown(struct lwp *l, void *v, register_t *retval) 2846 { 2847 struct sys_chown_args /* { 2848 syscallarg(const char *) path; 2849 syscallarg(uid_t) uid; 2850 syscallarg(gid_t) gid; 2851 } */ *uap = v; 2852 int error; 2853 struct nameidata nd; 2854 2855 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2856 if ((error = namei(&nd)) != 0) 2857 return (error); 2858 2859 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2860 2861 vrele(nd.ni_vp); 2862 return (error); 2863 } 2864 2865 /* 2866 * Set ownership given a path name; this version follows links. 2867 * Provides POSIX semantics. 2868 */ 2869 /* ARGSUSED */ 2870 int 2871 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2872 { 2873 struct sys_chown_args /* { 2874 syscallarg(const char *) path; 2875 syscallarg(uid_t) uid; 2876 syscallarg(gid_t) gid; 2877 } */ *uap = v; 2878 int error; 2879 struct nameidata nd; 2880 2881 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2882 if ((error = namei(&nd)) != 0) 2883 return (error); 2884 2885 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2886 2887 vrele(nd.ni_vp); 2888 return (error); 2889 } 2890 2891 /* 2892 * Set ownership given a file descriptor. 2893 */ 2894 /* ARGSUSED */ 2895 int 2896 sys_fchown(struct lwp *l, void *v, register_t *retval) 2897 { 2898 struct sys_fchown_args /* { 2899 syscallarg(int) fd; 2900 syscallarg(uid_t) uid; 2901 syscallarg(gid_t) gid; 2902 } */ *uap = v; 2903 struct proc *p = l->l_proc; 2904 int error; 2905 struct file *fp; 2906 2907 /* getvnode() will use the descriptor for us */ 2908 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2909 return (error); 2910 2911 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2912 SCARG(uap, gid), l, 0); 2913 FILE_UNUSE(fp, l); 2914 return (error); 2915 } 2916 2917 /* 2918 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2919 */ 2920 /* ARGSUSED */ 2921 int 2922 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2923 { 2924 struct sys_fchown_args /* { 2925 syscallarg(int) fd; 2926 syscallarg(uid_t) uid; 2927 syscallarg(gid_t) gid; 2928 } */ *uap = v; 2929 struct proc *p = l->l_proc; 2930 int error; 2931 struct file *fp; 2932 2933 /* getvnode() will use the descriptor for us */ 2934 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2935 return (error); 2936 2937 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2938 SCARG(uap, gid), l, 1); 2939 FILE_UNUSE(fp, l); 2940 return (error); 2941 } 2942 2943 /* 2944 * Set ownership given a path name; this version does not follow links. 2945 */ 2946 /* ARGSUSED */ 2947 int 2948 sys_lchown(struct lwp *l, void *v, register_t *retval) 2949 { 2950 struct sys_lchown_args /* { 2951 syscallarg(const char *) path; 2952 syscallarg(uid_t) uid; 2953 syscallarg(gid_t) gid; 2954 } */ *uap = v; 2955 int error; 2956 struct nameidata nd; 2957 2958 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2959 if ((error = namei(&nd)) != 0) 2960 return (error); 2961 2962 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2963 2964 vrele(nd.ni_vp); 2965 return (error); 2966 } 2967 2968 /* 2969 * Set ownership given a path name; this version does not follow links. 2970 * Provides POSIX/XPG semantics. 2971 */ 2972 /* ARGSUSED */ 2973 int 2974 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2975 { 2976 struct sys_lchown_args /* { 2977 syscallarg(const char *) path; 2978 syscallarg(uid_t) uid; 2979 syscallarg(gid_t) gid; 2980 } */ *uap = v; 2981 int error; 2982 struct nameidata nd; 2983 2984 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2985 if ((error = namei(&nd)) != 0) 2986 return (error); 2987 2988 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2989 2990 vrele(nd.ni_vp); 2991 return (error); 2992 } 2993 2994 /* 2995 * Common routine to set ownership given a vnode. 2996 */ 2997 static int 2998 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2999 int posix_semantics) 3000 { 3001 struct mount *mp; 3002 struct vattr vattr; 3003 mode_t newmode; 3004 int error; 3005 3006 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3007 return (error); 3008 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3009 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3010 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3011 goto out; 3012 3013 #define CHANGED(x) ((int)(x) != -1) 3014 newmode = vattr.va_mode; 3015 if (posix_semantics) { 3016 /* 3017 * POSIX/XPG semantics: if the caller is not the super-user, 3018 * clear set-user-id and set-group-id bits. Both POSIX and 3019 * the XPG consider the behaviour for calls by the super-user 3020 * implementation-defined; we leave the set-user-id and set- 3021 * group-id settings intact in that case. 3022 */ 3023 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, 3024 NULL) != 0) 3025 newmode &= ~(S_ISUID | S_ISGID); 3026 } else { 3027 /* 3028 * NetBSD semantics: when changing owner and/or group, 3029 * clear the respective bit(s). 3030 */ 3031 if (CHANGED(uid)) 3032 newmode &= ~S_ISUID; 3033 if (CHANGED(gid)) 3034 newmode &= ~S_ISGID; 3035 } 3036 /* Update va_mode iff altered. */ 3037 if (vattr.va_mode == newmode) 3038 newmode = VNOVAL; 3039 3040 VATTR_NULL(&vattr); 3041 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 3042 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 3043 vattr.va_mode = newmode; 3044 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3045 #undef CHANGED 3046 3047 out: 3048 VOP_UNLOCK(vp, 0); 3049 vn_finished_write(mp, 0); 3050 return (error); 3051 } 3052 3053 /* 3054 * Set the access and modification times given a path name; this 3055 * version follows links. 3056 */ 3057 /* ARGSUSED */ 3058 int 3059 sys_utimes(struct lwp *l, void *v, register_t *retval) 3060 { 3061 struct sys_utimes_args /* { 3062 syscallarg(const char *) path; 3063 syscallarg(const struct timeval *) tptr; 3064 } */ *uap = v; 3065 int error; 3066 struct nameidata nd; 3067 3068 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3069 if ((error = namei(&nd)) != 0) 3070 return (error); 3071 3072 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3073 3074 vrele(nd.ni_vp); 3075 return (error); 3076 } 3077 3078 /* 3079 * Set the access and modification times given a file descriptor. 3080 */ 3081 /* ARGSUSED */ 3082 int 3083 sys_futimes(struct lwp *l, void *v, register_t *retval) 3084 { 3085 struct sys_futimes_args /* { 3086 syscallarg(int) fd; 3087 syscallarg(const struct timeval *) tptr; 3088 } */ *uap = v; 3089 struct proc *p = l->l_proc; 3090 int error; 3091 struct file *fp; 3092 3093 /* getvnode() will use the descriptor for us */ 3094 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3095 return (error); 3096 3097 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 3098 FILE_UNUSE(fp, l); 3099 return (error); 3100 } 3101 3102 /* 3103 * Set the access and modification times given a path name; this 3104 * version does not follow links. 3105 */ 3106 /* ARGSUSED */ 3107 int 3108 sys_lutimes(struct lwp *l, void *v, register_t *retval) 3109 { 3110 struct sys_lutimes_args /* { 3111 syscallarg(const char *) path; 3112 syscallarg(const struct timeval *) tptr; 3113 } */ *uap = v; 3114 int error; 3115 struct nameidata nd; 3116 3117 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3118 if ((error = namei(&nd)) != 0) 3119 return (error); 3120 3121 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 3122 3123 vrele(nd.ni_vp); 3124 return (error); 3125 } 3126 3127 /* 3128 * Common routine to set access and modification times given a vnode. 3129 */ 3130 static int 3131 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 3132 { 3133 struct mount *mp; 3134 struct vattr vattr; 3135 int error; 3136 3137 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3138 return (error); 3139 VATTR_NULL(&vattr); 3140 if (tptr == NULL) { 3141 nanotime(&vattr.va_atime); 3142 vattr.va_mtime = vattr.va_atime; 3143 vattr.va_vaflags |= VA_UTIMES_NULL; 3144 } else { 3145 struct timeval tv[2]; 3146 3147 error = copyin(tptr, tv, sizeof(tv)); 3148 if (error) 3149 goto out; 3150 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 3151 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 3152 } 3153 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3154 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3155 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3156 VOP_UNLOCK(vp, 0); 3157 out: 3158 vn_finished_write(mp, 0); 3159 return (error); 3160 } 3161 3162 /* 3163 * Truncate a file given its path name. 3164 */ 3165 /* ARGSUSED */ 3166 int 3167 sys_truncate(struct lwp *l, void *v, register_t *retval) 3168 { 3169 struct sys_truncate_args /* { 3170 syscallarg(const char *) path; 3171 syscallarg(int) pad; 3172 syscallarg(off_t) length; 3173 } */ *uap = v; 3174 struct vnode *vp; 3175 struct mount *mp; 3176 struct vattr vattr; 3177 int error; 3178 struct nameidata nd; 3179 3180 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3181 if ((error = namei(&nd)) != 0) 3182 return (error); 3183 vp = nd.ni_vp; 3184 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3185 vrele(vp); 3186 return (error); 3187 } 3188 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3189 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3190 if (vp->v_type == VDIR) 3191 error = EISDIR; 3192 else if ((error = vn_writechk(vp)) == 0 && 3193 (error = VOP_ACCESS(vp, VWRITE, l->l_cred, l)) == 0) { 3194 VATTR_NULL(&vattr); 3195 vattr.va_size = SCARG(uap, length); 3196 error = VOP_SETATTR(vp, &vattr, l->l_cred, l); 3197 } 3198 vput(vp); 3199 vn_finished_write(mp, 0); 3200 return (error); 3201 } 3202 3203 /* 3204 * Truncate a file given a file descriptor. 3205 */ 3206 /* ARGSUSED */ 3207 int 3208 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 3209 { 3210 struct sys_ftruncate_args /* { 3211 syscallarg(int) fd; 3212 syscallarg(int) pad; 3213 syscallarg(off_t) length; 3214 } */ *uap = v; 3215 struct proc *p = l->l_proc; 3216 struct mount *mp; 3217 struct vattr vattr; 3218 struct vnode *vp; 3219 struct file *fp; 3220 int error; 3221 3222 /* getvnode() will use the descriptor for us */ 3223 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3224 return (error); 3225 if ((fp->f_flag & FWRITE) == 0) { 3226 error = EINVAL; 3227 goto out; 3228 } 3229 vp = (struct vnode *)fp->f_data; 3230 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3231 FILE_UNUSE(fp, l); 3232 return (error); 3233 } 3234 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3235 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3236 if (vp->v_type == VDIR) 3237 error = EISDIR; 3238 else if ((error = vn_writechk(vp)) == 0) { 3239 VATTR_NULL(&vattr); 3240 vattr.va_size = SCARG(uap, length); 3241 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 3242 } 3243 VOP_UNLOCK(vp, 0); 3244 vn_finished_write(mp, 0); 3245 out: 3246 FILE_UNUSE(fp, l); 3247 return (error); 3248 } 3249 3250 /* 3251 * Sync an open file. 3252 */ 3253 /* ARGSUSED */ 3254 int 3255 sys_fsync(struct lwp *l, void *v, register_t *retval) 3256 { 3257 struct sys_fsync_args /* { 3258 syscallarg(int) fd; 3259 } */ *uap = v; 3260 struct proc *p = l->l_proc; 3261 struct vnode *vp; 3262 struct mount *mp; 3263 struct file *fp; 3264 int error; 3265 3266 /* getvnode() will use the descriptor for us */ 3267 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3268 return (error); 3269 vp = (struct vnode *)fp->f_data; 3270 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 3271 FILE_UNUSE(fp, l); 3272 return (error); 3273 } 3274 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3275 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3276 if (error == 0 && bioops.io_fsync != NULL && 3277 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) { 3278 KERNEL_LOCK(1, l); 3279 (*bioops.io_fsync)(vp, 0); 3280 KERNEL_UNLOCK_ONE(l); 3281 } 3282 VOP_UNLOCK(vp, 0); 3283 vn_finished_write(mp, 0); 3284 FILE_UNUSE(fp, l); 3285 return (error); 3286 } 3287 3288 /* 3289 * Sync a range of file data. API modeled after that found in AIX. 3290 * 3291 * FDATASYNC indicates that we need only save enough metadata to be able 3292 * to re-read the written data. Note we duplicate AIX's requirement that 3293 * the file be open for writing. 3294 */ 3295 /* ARGSUSED */ 3296 int 3297 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3298 { 3299 struct sys_fsync_range_args /* { 3300 syscallarg(int) fd; 3301 syscallarg(int) flags; 3302 syscallarg(off_t) start; 3303 syscallarg(off_t) length; 3304 } */ *uap = v; 3305 struct proc *p = l->l_proc; 3306 struct vnode *vp; 3307 struct file *fp; 3308 int flags, nflags; 3309 off_t s, e, len; 3310 int error; 3311 3312 /* getvnode() will use the descriptor for us */ 3313 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3314 return (error); 3315 3316 if ((fp->f_flag & FWRITE) == 0) { 3317 error = EBADF; 3318 goto out; 3319 } 3320 3321 flags = SCARG(uap, flags); 3322 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3323 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3324 error = EINVAL; 3325 goto out; 3326 } 3327 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3328 if (flags & FDATASYNC) 3329 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3330 else 3331 nflags = FSYNC_WAIT; 3332 if (flags & FDISKSYNC) 3333 nflags |= FSYNC_CACHE; 3334 3335 len = SCARG(uap, length); 3336 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3337 if (len) { 3338 s = SCARG(uap, start); 3339 e = s + len; 3340 if (e < s) { 3341 error = EINVAL; 3342 goto out; 3343 } 3344 } else { 3345 e = 0; 3346 s = 0; 3347 } 3348 3349 vp = (struct vnode *)fp->f_data; 3350 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3351 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3352 3353 if (error == 0 && bioops.io_fsync != NULL && 3354 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) { 3355 KERNEL_LOCK(1, l); 3356 (*bioops.io_fsync)(vp, nflags); 3357 KERNEL_UNLOCK_ONE(l); 3358 } 3359 3360 VOP_UNLOCK(vp, 0); 3361 out: 3362 FILE_UNUSE(fp, l); 3363 return (error); 3364 } 3365 3366 /* 3367 * Sync the data of an open file. 3368 */ 3369 /* ARGSUSED */ 3370 int 3371 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3372 { 3373 struct sys_fdatasync_args /* { 3374 syscallarg(int) fd; 3375 } */ *uap = v; 3376 struct proc *p = l->l_proc; 3377 struct vnode *vp; 3378 struct file *fp; 3379 int error; 3380 3381 /* getvnode() will use the descriptor for us */ 3382 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3383 return (error); 3384 if ((fp->f_flag & FWRITE) == 0) { 3385 FILE_UNUSE(fp, l); 3386 return (EBADF); 3387 } 3388 vp = (struct vnode *)fp->f_data; 3389 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3390 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3391 VOP_UNLOCK(vp, 0); 3392 FILE_UNUSE(fp, l); 3393 return (error); 3394 } 3395 3396 /* 3397 * Rename files, (standard) BSD semantics frontend. 3398 */ 3399 /* ARGSUSED */ 3400 int 3401 sys_rename(struct lwp *l, void *v, register_t *retval) 3402 { 3403 struct sys_rename_args /* { 3404 syscallarg(const char *) from; 3405 syscallarg(const char *) to; 3406 } */ *uap = v; 3407 3408 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3409 } 3410 3411 /* 3412 * Rename files, POSIX semantics frontend. 3413 */ 3414 /* ARGSUSED */ 3415 int 3416 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3417 { 3418 struct sys___posix_rename_args /* { 3419 syscallarg(const char *) from; 3420 syscallarg(const char *) to; 3421 } */ *uap = v; 3422 3423 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3424 } 3425 3426 /* 3427 * Rename files. Source and destination must either both be directories, 3428 * or both not be directories. If target is a directory, it must be empty. 3429 * If `from' and `to' refer to the same object, the value of the `retain' 3430 * argument is used to determine whether `from' will be 3431 * 3432 * (retain == 0) deleted unless `from' and `to' refer to the same 3433 * object in the file system's name space (BSD). 3434 * (retain == 1) always retained (POSIX). 3435 */ 3436 static int 3437 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3438 { 3439 struct mount *mp = NULL; 3440 struct vnode *tvp, *fvp, *tdvp; 3441 struct nameidata fromnd, tond; 3442 struct proc *p; 3443 int error; 3444 3445 NDINIT(&fromnd, DELETE, LOCKPARENT | SAVESTART, UIO_USERSPACE, 3446 from, l); 3447 if ((error = namei(&fromnd)) != 0) 3448 return (error); 3449 if (fromnd.ni_dvp != fromnd.ni_vp) 3450 VOP_UNLOCK(fromnd.ni_dvp, 0); 3451 fvp = fromnd.ni_vp; 3452 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3453 if (error != 0) { 3454 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3455 vrele(fromnd.ni_dvp); 3456 vrele(fvp); 3457 if (fromnd.ni_startdir) 3458 vrele(fromnd.ni_startdir); 3459 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3460 return (error); 3461 } 3462 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3463 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3464 if ((error = namei(&tond)) != 0) { 3465 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3466 vrele(fromnd.ni_dvp); 3467 vrele(fvp); 3468 goto out1; 3469 } 3470 tdvp = tond.ni_dvp; 3471 tvp = tond.ni_vp; 3472 3473 if (tvp != NULL) { 3474 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3475 error = ENOTDIR; 3476 goto out; 3477 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3478 error = EISDIR; 3479 goto out; 3480 } 3481 } 3482 3483 if (fvp == tdvp) 3484 error = EINVAL; 3485 3486 /* 3487 * Source and destination refer to the same object. 3488 */ 3489 if (fvp == tvp) { 3490 if (retain) 3491 error = -1; 3492 else if (fromnd.ni_dvp == tdvp && 3493 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3494 !memcmp(fromnd.ni_cnd.cn_nameptr, 3495 tond.ni_cnd.cn_nameptr, 3496 fromnd.ni_cnd.cn_namelen)) 3497 error = -1; 3498 } 3499 3500 #if NVERIEXEC > 0 3501 if (!error) { 3502 pathname_t frompath = NULL, topath = NULL; 3503 3504 error = pathname_get(fromnd.ni_dirp, fromnd.ni_segflg, 3505 &frompath); 3506 if (!error) 3507 error = pathname_get(tond.ni_dirp, tond.ni_segflg, 3508 &topath); 3509 if (!error) 3510 error = veriexec_renamechk(fvp, pathname_path(frompath), 3511 tvp, pathname_path(topath), l); 3512 3513 pathname_put(frompath); 3514 pathname_put(topath); 3515 } 3516 #endif /* NVERIEXEC > 0 */ 3517 3518 out: 3519 p = l->l_proc; 3520 if (!error) { 3521 VOP_LEASE(tdvp, l, l->l_cred, LEASE_WRITE); 3522 if (fromnd.ni_dvp != tdvp) 3523 VOP_LEASE(fromnd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3524 if (tvp) { 3525 VOP_LEASE(tvp, l, l->l_cred, LEASE_WRITE); 3526 } 3527 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3528 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3529 } else { 3530 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3531 if (tdvp == tvp) 3532 vrele(tdvp); 3533 else 3534 vput(tdvp); 3535 if (tvp) 3536 vput(tvp); 3537 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3538 vrele(fromnd.ni_dvp); 3539 vrele(fvp); 3540 } 3541 vrele(tond.ni_startdir); 3542 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3543 out1: 3544 vn_finished_write(mp, 0); 3545 if (fromnd.ni_startdir) 3546 vrele(fromnd.ni_startdir); 3547 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3548 return (error == -1 ? 0 : error); 3549 } 3550 3551 /* 3552 * Make a directory file. 3553 */ 3554 /* ARGSUSED */ 3555 int 3556 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3557 { 3558 struct sys_mkdir_args /* { 3559 syscallarg(const char *) path; 3560 syscallarg(int) mode; 3561 } */ *uap = v; 3562 struct proc *p = l->l_proc; 3563 struct mount *mp; 3564 struct vnode *vp; 3565 struct vattr vattr; 3566 int error; 3567 struct nameidata nd; 3568 3569 restart: 3570 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3571 SCARG(uap, path), l); 3572 if ((error = namei(&nd)) != 0) 3573 return (error); 3574 vp = nd.ni_vp; 3575 if (vp != NULL) { 3576 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3577 if (nd.ni_dvp == vp) 3578 vrele(nd.ni_dvp); 3579 else 3580 vput(nd.ni_dvp); 3581 vrele(vp); 3582 return (EEXIST); 3583 } 3584 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3585 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3586 if (nd.ni_dvp == vp) 3587 vrele(nd.ni_dvp); 3588 else 3589 vput(nd.ni_dvp); 3590 if ((error = vn_start_write(NULL, &mp, 3591 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3592 return (error); 3593 goto restart; 3594 } 3595 VATTR_NULL(&vattr); 3596 vattr.va_type = VDIR; 3597 /* We will read cwdi->cwdi_cmask unlocked. */ 3598 vattr.va_mode = 3599 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3600 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3601 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3602 if (!error) 3603 vput(nd.ni_vp); 3604 vn_finished_write(mp, 0); 3605 return (error); 3606 } 3607 3608 /* 3609 * Remove a directory file. 3610 */ 3611 /* ARGSUSED */ 3612 int 3613 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3614 { 3615 struct sys_rmdir_args /* { 3616 syscallarg(const char *) path; 3617 } */ *uap = v; 3618 struct mount *mp; 3619 struct vnode *vp; 3620 int error; 3621 struct nameidata nd; 3622 3623 restart: 3624 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3625 SCARG(uap, path), l); 3626 if ((error = namei(&nd)) != 0) 3627 return (error); 3628 vp = nd.ni_vp; 3629 if (vp->v_type != VDIR) { 3630 error = ENOTDIR; 3631 goto out; 3632 } 3633 /* 3634 * No rmdir "." please. 3635 */ 3636 if (nd.ni_dvp == vp) { 3637 error = EINVAL; 3638 goto out; 3639 } 3640 /* 3641 * The root of a mounted filesystem cannot be deleted. 3642 */ 3643 if (vp->v_flag & VROOT) { 3644 error = EBUSY; 3645 goto out; 3646 } 3647 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3648 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3649 if (nd.ni_dvp == vp) 3650 vrele(nd.ni_dvp); 3651 else 3652 vput(nd.ni_dvp); 3653 vput(vp); 3654 if ((error = vn_start_write(NULL, &mp, 3655 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3656 return (error); 3657 goto restart; 3658 } 3659 VOP_LEASE(nd.ni_dvp, l, l->l_cred, LEASE_WRITE); 3660 VOP_LEASE(vp, l, l->l_cred, LEASE_WRITE); 3661 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3662 vn_finished_write(mp, 0); 3663 return (error); 3664 3665 out: 3666 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3667 if (nd.ni_dvp == vp) 3668 vrele(nd.ni_dvp); 3669 else 3670 vput(nd.ni_dvp); 3671 vput(vp); 3672 return (error); 3673 } 3674 3675 /* 3676 * Read a block of directory entries in a file system independent format. 3677 */ 3678 int 3679 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3680 { 3681 struct sys___getdents30_args /* { 3682 syscallarg(int) fd; 3683 syscallarg(char *) buf; 3684 syscallarg(size_t) count; 3685 } */ *uap = v; 3686 struct proc *p = l->l_proc; 3687 struct file *fp; 3688 int error, done; 3689 3690 /* getvnode() will use the descriptor for us */ 3691 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3692 return (error); 3693 if ((fp->f_flag & FREAD) == 0) { 3694 error = EBADF; 3695 goto out; 3696 } 3697 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3698 SCARG(uap, count), &done, l, 0, 0); 3699 #ifdef KTRACE 3700 if (!error && KTRPOINT(p, KTR_GENIO)) { 3701 struct iovec iov; 3702 iov.iov_base = SCARG(uap, buf); 3703 iov.iov_len = done; 3704 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3705 } 3706 #endif 3707 *retval = done; 3708 out: 3709 FILE_UNUSE(fp, l); 3710 return (error); 3711 } 3712 3713 /* 3714 * Set the mode mask for creation of filesystem nodes. 3715 */ 3716 int 3717 sys_umask(struct lwp *l, void *v, register_t *retval) 3718 { 3719 struct sys_umask_args /* { 3720 syscallarg(mode_t) newmask; 3721 } */ *uap = v; 3722 struct proc *p = l->l_proc; 3723 struct cwdinfo *cwdi; 3724 3725 /* 3726 * cwdi->cwdi_cmask will be read unlocked elsewhere. What's 3727 * important is that we serialize changes to the mask. The 3728 * rw_exit() will issue a write memory barrier on our behalf, 3729 * and force the changes out to other CPUs (as it must use an 3730 * atomic operation, draining the local CPU's store buffers). 3731 */ 3732 cwdi = p->p_cwdi; 3733 rw_enter(&cwdi->cwdi_lock, RW_WRITER); 3734 *retval = cwdi->cwdi_cmask; 3735 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3736 rw_exit(&cwdi->cwdi_lock); 3737 3738 return (0); 3739 } 3740 3741 /* 3742 * Void all references to file by ripping underlying filesystem 3743 * away from vnode. 3744 */ 3745 /* ARGSUSED */ 3746 int 3747 sys_revoke(struct lwp *l, void *v, register_t *retval) 3748 { 3749 struct sys_revoke_args /* { 3750 syscallarg(const char *) path; 3751 } */ *uap = v; 3752 struct mount *mp; 3753 struct vnode *vp; 3754 struct vattr vattr; 3755 int error; 3756 struct nameidata nd; 3757 3758 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3759 if ((error = namei(&nd)) != 0) 3760 return (error); 3761 vp = nd.ni_vp; 3762 if ((error = VOP_GETATTR(vp, &vattr, l->l_cred, l)) != 0) 3763 goto out; 3764 if (kauth_cred_geteuid(l->l_cred) != vattr.va_uid && 3765 (error = kauth_authorize_generic(l->l_cred, 3766 KAUTH_GENERIC_ISSUSER, NULL)) != 0) 3767 goto out; 3768 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3769 goto out; 3770 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3771 VOP_REVOKE(vp, REVOKEALL); 3772 vn_finished_write(mp, 0); 3773 out: 3774 vrele(vp); 3775 return (error); 3776 } 3777 3778 /* 3779 * Convert a user file descriptor to a kernel file entry. 3780 */ 3781 int 3782 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3783 { 3784 struct vnode *vp; 3785 struct file *fp; 3786 3787 if ((fp = fd_getfile(fdp, fd)) == NULL) 3788 return (EBADF); 3789 3790 FILE_USE(fp); 3791 3792 if (fp->f_type != DTYPE_VNODE) { 3793 FILE_UNUSE(fp, NULL); 3794 return (EINVAL); 3795 } 3796 3797 vp = (struct vnode *)fp->f_data; 3798 if (vp->v_type == VBAD) { 3799 FILE_UNUSE(fp, NULL); 3800 return (EBADF); 3801 } 3802 3803 *fpp = fp; 3804 return (0); 3805 } 3806
Indexes created Fri Oct 03 02:09:57 GMT 2025