vfs_syscalls.c revision 1.238.4.3
1 /* $NetBSD: vfs_syscalls.c,v 1.238.4.3 2006/04/19 05:14:00 elad 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 1.238.4.3 2006/04/19 05:14:00 elad Exp $"); 41 42 #include "opt_compat_netbsd.h" 43 #include "opt_compat_43.h" 44 #include "opt_ktrace.h" 45 #include "opt_verified_exec.h" 46 #include "fss.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/namei.h> 51 #include <sys/filedesc.h> 52 #include <sys/kernel.h> 53 #include <sys/file.h> 54 #include <sys/stat.h> 55 #include <sys/vnode.h> 56 #include <sys/mount.h> 57 #include <sys/proc.h> 58 #include <sys/uio.h> 59 #include <sys/malloc.h> 60 #include <sys/dirent.h> 61 #include <sys/sysctl.h> 62 #include <sys/sa.h> 63 #include <sys/syscallargs.h> 64 #ifdef KTRACE 65 #include <sys/ktrace.h> 66 #endif 67 #ifdef VERIFIED_EXEC 68 #include <sys/verified_exec.h> 69 #endif /* VERIFIED_EXEC */ 70 71 #include <miscfs/genfs/genfs.h> 72 #include <miscfs/syncfs/syncfs.h> 73 74 #ifdef COMPAT_30 75 #include "opt_nfsserver.h" 76 #include <nfs/rpcv2.h> 77 #include <nfs/nfsproto.h> 78 #include <nfs/nfs.h> 79 #include <nfs/nfs_var.h> 80 #endif 81 82 #if NFSS > 0 83 #include <dev/fssvar.h> 84 #endif 85 86 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 87 88 static int change_dir(struct nameidata *, struct lwp *); 89 static int change_flags(struct vnode *, u_long, struct lwp *); 90 static int change_mode(struct vnode *, int, struct lwp *l); 91 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 92 static int change_utimes(struct vnode *vp, const struct timeval *, 93 struct lwp *l); 94 static int rename_files(const char *, const char *, struct lwp *, int); 95 96 void checkdirs(struct vnode *); 97 98 int dovfsusermount = 0; 99 100 /* 101 * Virtual File System System Calls 102 */ 103 104 /* 105 * Mount a file system. 106 */ 107 108 #if defined(COMPAT_09) || defined(COMPAT_43) 109 /* 110 * This table is used to maintain compatibility with 4.3BSD 111 * and NetBSD 0.9 mount syscalls. Note, the order is important! 112 * 113 * Do not modify this table. It should only contain filesystems 114 * supported by NetBSD 0.9 and 4.3BSD. 115 */ 116 const char * const mountcompatnames[] = { 117 NULL, /* 0 = MOUNT_NONE */ 118 MOUNT_FFS, /* 1 = MOUNT_UFS */ 119 MOUNT_NFS, /* 2 */ 120 MOUNT_MFS, /* 3 */ 121 MOUNT_MSDOS, /* 4 */ 122 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 123 MOUNT_FDESC, /* 6 */ 124 MOUNT_KERNFS, /* 7 */ 125 NULL, /* 8 = MOUNT_DEVFS */ 126 MOUNT_AFS, /* 9 */ 127 }; 128 const int nmountcompatnames = sizeof(mountcompatnames) / 129 sizeof(mountcompatnames[0]); 130 #endif /* COMPAT_09 || COMPAT_43 */ 131 132 /* ARGSUSED */ 133 int 134 sys_mount(struct lwp *l, void *v, register_t *retval) 135 { 136 struct sys_mount_args /* { 137 syscallarg(const char *) type; 138 syscallarg(const char *) path; 139 syscallarg(int) flags; 140 syscallarg(void *) data; 141 } */ *uap = v; 142 struct proc *p = l->l_proc; 143 struct vnode *vp; 144 struct mount *mp; 145 int error, flag = 0; 146 char fstypename[MFSNAMELEN]; 147 struct vattr va; 148 struct nameidata nd; 149 struct vfsops *vfs; 150 151 /* 152 * if MNT_GETARGS is specified, it should be only flag. 153 */ 154 155 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 156 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 157 return EINVAL; 158 } 159 160 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 161 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 162 &p->p_acflag))) 163 return (error); 164 /* 165 * Get vnode to be covered 166 */ 167 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 168 SCARG(uap, path), l); 169 if ((error = namei(&nd)) != 0) 170 return (error); 171 vp = nd.ni_vp; 172 /* 173 * A lookup in VFS_MOUNT might result in an attempt to 174 * lock this vnode again, so make the lock recursive. 175 */ 176 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 177 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 178 if ((vp->v_flag & VROOT) == 0) { 179 vput(vp); 180 return (EINVAL); 181 } 182 mp = vp->v_mount; 183 flag = mp->mnt_flag; 184 vfs = mp->mnt_op; 185 /* 186 * We only allow the filesystem to be reloaded if it 187 * is currently mounted read-only. 188 */ 189 if ((SCARG(uap, flags) & MNT_RELOAD) && 190 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 191 vput(vp); 192 return (EOPNOTSUPP); /* Needs translation */ 193 } 194 /* 195 * In "highly secure" mode, don't let the caller do anything 196 * but downgrade a filesystem from read-write to read-only. 197 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 198 */ 199 if (securelevel >= 2 && 200 SCARG(uap, flags) != MNT_GETARGS && 201 SCARG(uap, flags) != 202 (mp->mnt_flag | MNT_RDONLY | 203 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 204 vput(vp); 205 return (EPERM); 206 } 207 mp->mnt_flag |= SCARG(uap, flags) & 208 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 209 /* 210 * Only root, or the user that did the original mount is 211 * permitted to update it. 212 */ 213 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 214 mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred) && 215 (error = kauth_authorize_generic(p->p_cred, 216 KAUTH_GENERIC_ISSUSER, 217 &p->p_acflag)) != 0) { 218 vput(vp); 219 return (error); 220 } 221 /* 222 * Do not allow NFS export by non-root users. For non-root 223 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 224 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 225 */ 226 if (kauth_cred_geteuid(p->p_cred) != 0) { 227 if (SCARG(uap, flags) & MNT_EXPORTED) { 228 vput(vp); 229 return (EPERM); 230 } 231 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 232 if (flag & MNT_NOEXEC) 233 SCARG(uap, flags) |= MNT_NOEXEC; 234 } 235 if (vfs_busy(mp, LK_NOWAIT, 0)) { 236 vput(vp); 237 return (EPERM); 238 } 239 goto update; 240 } else { 241 if (securelevel >= 2) { 242 vput(vp); 243 return (EPERM); 244 } 245 } 246 /* 247 * If the user is not root, ensure that they own the directory 248 * onto which we are attempting to mount. 249 */ 250 if ((error = VOP_GETATTR(vp, &va, p->p_cred, l)) != 0 || 251 (va.va_uid != kauth_cred_geteuid(p->p_cred) && 252 (error = kauth_authorize_generic(p->p_cred, 253 KAUTH_GENERIC_ISSUSER, 254 &p->p_acflag)) != 0)) { 255 vput(vp); 256 return (error); 257 } 258 /* 259 * Do not allow NFS export by non-root users. For non-root users, 260 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 261 * mount point is already MNT_NOEXEC. 262 */ 263 if (kauth_cred_geteuid(p->p_cred) != 0) { 264 if (SCARG(uap, flags) & MNT_EXPORTED) { 265 vput(vp); 266 return (EPERM); 267 } 268 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 269 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 270 SCARG(uap, flags) |= MNT_NOEXEC; 271 } 272 if ((error = vinvalbuf(vp, V_SAVE, p->p_cred, l, 0, 0)) != 0) { 273 vput(vp); 274 return (error); 275 } 276 if (vp->v_type != VDIR) { 277 vput(vp); 278 return (ENOTDIR); 279 } 280 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 281 if (error) { 282 #if defined(COMPAT_09) || defined(COMPAT_43) 283 /* 284 * Historically, filesystem types were identified by numbers. 285 * If we get an integer for the filesystem type instead of a 286 * string, we check to see if it matches one of the historic 287 * filesystem types. 288 */ 289 u_long fsindex = (u_long)SCARG(uap, type); 290 if (fsindex >= nmountcompatnames || 291 mountcompatnames[fsindex] == NULL) { 292 vput(vp); 293 return (ENODEV); 294 } 295 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 296 #else 297 vput(vp); 298 return (error); 299 #endif 300 } 301 #ifdef COMPAT_10 302 /* Accept `ufs' as an alias for `ffs'. */ 303 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 304 strncpy(fstypename, "ffs", MFSNAMELEN); 305 #endif 306 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 307 vput(vp); 308 return (ENODEV); 309 } 310 if (vp->v_mountedhere != NULL) { 311 vput(vp); 312 return (EBUSY); 313 } 314 315 /* 316 * Allocate and initialize the file system. 317 */ 318 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 319 M_MOUNT, M_WAITOK); 320 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 321 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 322 simple_lock_init(&mp->mnt_slock); 323 (void)vfs_busy(mp, LK_NOWAIT, 0); 324 mp->mnt_op = vfs; 325 vfs->vfs_refcount++; 326 mp->mnt_vnodecovered = vp; 327 mp->mnt_stat.f_owner = kauth_cred_geteuid(p->p_cred); 328 mp->mnt_unmounter = NULL; 329 mp->mnt_leaf = mp; 330 331 /* 332 * The underlying file system may refuse the mount for 333 * various reasons. Allow the user to force it to happen. 334 */ 335 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 336 update: 337 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 338 /* 339 * Set the mount level flags. 340 */ 341 if (SCARG(uap, flags) & MNT_RDONLY) 342 mp->mnt_flag |= MNT_RDONLY; 343 else if (mp->mnt_flag & MNT_RDONLY) 344 mp->mnt_iflag |= IMNT_WANTRDWR; 345 mp->mnt_flag &= 346 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 347 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 348 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 349 mp->mnt_flag |= SCARG(uap, flags) & 350 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 351 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 352 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 353 MNT_IGNORE); 354 } 355 /* 356 * Mount the filesystem. 357 */ 358 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 359 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 360 #if defined(COMPAT_30) && defined(NFSSERVER) 361 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 362 int error2; 363 364 /* Update failed; let's try and see if it was an 365 * export request. */ 366 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 367 SCARG(uap, data), l); 368 369 /* Only update error code if the export request was 370 * understood but some problem occurred while 371 * processing it. */ 372 if (error2 != EJUSTRETURN) 373 error = error2; 374 } 375 #endif 376 if (mp->mnt_iflag & IMNT_WANTRDWR) 377 mp->mnt_flag &= ~MNT_RDONLY; 378 if (error) 379 mp->mnt_flag = flag; 380 mp->mnt_flag &=~ 381 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 382 mp->mnt_iflag &=~ IMNT_WANTRDWR; 383 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 384 if (mp->mnt_syncer == NULL) 385 error = vfs_allocate_syncvnode(mp); 386 } else { 387 if (mp->mnt_syncer != NULL) 388 vfs_deallocate_syncvnode(mp); 389 } 390 vfs_unbusy(mp); 391 VOP_UNLOCK(vp, 0); 392 vrele(vp); 393 return (error); 394 } 395 /* 396 * Put the new filesystem on the mount list after root. 397 */ 398 cache_purge(vp); 399 if (!error) { 400 mp->mnt_flag &=~ 401 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 402 mp->mnt_iflag &=~ IMNT_WANTRDWR; 403 vp->v_mountedhere = mp; 404 simple_lock(&mountlist_slock); 405 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 406 simple_unlock(&mountlist_slock); 407 checkdirs(vp); 408 VOP_UNLOCK(vp, 0); 409 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 410 error = vfs_allocate_syncvnode(mp); 411 vfs_unbusy(mp); 412 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 413 if ((error = VFS_START(mp, 0, l))) 414 vrele(vp); 415 } else { 416 vp->v_mountedhere = (struct mount *)0; 417 vfs->vfs_refcount--; 418 vfs_unbusy(mp); 419 free(mp, M_MOUNT); 420 vput(vp); 421 } 422 return (error); 423 } 424 425 /* 426 * Scan all active processes to see if any of them have a current 427 * or root directory onto which the new filesystem has just been 428 * mounted. If so, replace them with the new mount point. 429 */ 430 void 431 checkdirs(struct vnode *olddp) 432 { 433 struct cwdinfo *cwdi; 434 struct vnode *newdp; 435 struct proc *p; 436 437 if (olddp->v_usecount == 1) 438 return; 439 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 440 panic("mount: lost mount"); 441 proclist_lock_read(); 442 PROCLIST_FOREACH(p, &allproc) { 443 cwdi = p->p_cwdi; 444 if (!cwdi) 445 continue; 446 if (cwdi->cwdi_cdir == olddp) { 447 vrele(cwdi->cwdi_cdir); 448 VREF(newdp); 449 cwdi->cwdi_cdir = newdp; 450 } 451 if (cwdi->cwdi_rdir == olddp) { 452 vrele(cwdi->cwdi_rdir); 453 VREF(newdp); 454 cwdi->cwdi_rdir = newdp; 455 } 456 } 457 proclist_unlock_read(); 458 if (rootvnode == olddp) { 459 vrele(rootvnode); 460 VREF(newdp); 461 rootvnode = newdp; 462 } 463 vput(newdp); 464 } 465 466 /* 467 * Unmount a file system. 468 * 469 * Note: unmount takes a path to the vnode mounted on as argument, 470 * not special file (as before). 471 */ 472 /* ARGSUSED */ 473 int 474 sys_unmount(struct lwp *l, void *v, register_t *retval) 475 { 476 struct sys_unmount_args /* { 477 syscallarg(const char *) path; 478 syscallarg(int) flags; 479 } */ *uap = v; 480 struct proc *p = l->l_proc; 481 struct vnode *vp; 482 struct mount *mp; 483 int error; 484 struct nameidata nd; 485 486 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 487 SCARG(uap, path), l); 488 if ((error = namei(&nd)) != 0) 489 return (error); 490 vp = nd.ni_vp; 491 mp = vp->v_mount; 492 493 /* 494 * Only root, or the user that did the original mount is 495 * permitted to unmount this filesystem. 496 */ 497 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred)) && 498 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 499 &p->p_acflag)) != 0) { 500 vput(vp); 501 return (error); 502 } 503 504 /* 505 * Don't allow unmounting the root file system. 506 */ 507 if (mp->mnt_flag & MNT_ROOTFS) { 508 vput(vp); 509 return (EINVAL); 510 } 511 512 /* 513 * Must be the root of the filesystem 514 */ 515 if ((vp->v_flag & VROOT) == 0) { 516 vput(vp); 517 return (EINVAL); 518 } 519 vput(vp); 520 521 /* 522 * XXX Freeze syncer. Must do this before locking the 523 * mount point. See dounmount() for details. 524 */ 525 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 526 527 if (vfs_busy(mp, 0, 0)) { 528 lockmgr(&syncer_lock, LK_RELEASE, NULL); 529 return (EBUSY); 530 } 531 532 return (dounmount(mp, SCARG(uap, flags), l)); 533 } 534 535 /* 536 * Do the actual file system unmount. File system is assumed to have been 537 * marked busy by the caller. 538 */ 539 int 540 dounmount(struct mount *mp, int flags, struct lwp *l) 541 { 542 struct vnode *coveredvp; 543 int error; 544 int async; 545 int used_syncer; 546 547 simple_lock(&mountlist_slock); 548 vfs_unbusy(mp); 549 used_syncer = (mp->mnt_syncer != NULL); 550 551 /* 552 * XXX Syncer must be frozen when we get here. This should really 553 * be done on a per-mountpoint basis, but especially the softdep 554 * code possibly called from the syncer doesn't exactly work on a 555 * per-mountpoint basis, so the softdep code would become a maze 556 * of vfs_busy() calls. 557 * 558 * The caller of dounmount() must acquire syncer_lock because 559 * the syncer itself acquires locks in syncer_lock -> vfs_busy 560 * order, and we must preserve that order to avoid deadlock. 561 * 562 * So, if the file system did not use the syncer, now is 563 * the time to release the syncer_lock. 564 */ 565 if (used_syncer == 0) 566 lockmgr(&syncer_lock, LK_RELEASE, NULL); 567 568 mp->mnt_iflag |= IMNT_UNMOUNT; 569 mp->mnt_unmounter = l; 570 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 571 vn_start_write(NULL, &mp, V_WAIT); 572 573 async = mp->mnt_flag & MNT_ASYNC; 574 mp->mnt_flag &= ~MNT_ASYNC; 575 cache_purgevfs(mp); /* remove cache entries for this file sys */ 576 if (mp->mnt_syncer != NULL) 577 vfs_deallocate_syncvnode(mp); 578 error = 0; 579 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 580 #if NFSS > 0 581 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 582 #endif 583 if (error == 0) 584 error = VFS_SYNC(mp, MNT_WAIT, l->l_proc->p_cred, l); 585 } 586 if (error == 0 || (flags & MNT_FORCE)) 587 error = VFS_UNMOUNT(mp, flags, l); 588 vn_finished_write(mp, 0); 589 simple_lock(&mountlist_slock); 590 if (error) { 591 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 592 (void) vfs_allocate_syncvnode(mp); 593 mp->mnt_iflag &= ~IMNT_UNMOUNT; 594 mp->mnt_unmounter = NULL; 595 mp->mnt_flag |= async; 596 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 597 &mountlist_slock); 598 if (used_syncer) 599 lockmgr(&syncer_lock, LK_RELEASE, NULL); 600 simple_lock(&mp->mnt_slock); 601 while (mp->mnt_wcnt > 0) { 602 wakeup(mp); 603 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 604 0, &mp->mnt_slock); 605 } 606 simple_unlock(&mp->mnt_slock); 607 return (error); 608 } 609 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 610 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 611 coveredvp->v_mountedhere = NULL; 612 vrele(coveredvp); 613 } 614 mp->mnt_op->vfs_refcount--; 615 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 616 panic("unmount: dangling vnode"); 617 mp->mnt_iflag |= IMNT_GONE; 618 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 619 if (used_syncer) 620 lockmgr(&syncer_lock, LK_RELEASE, NULL); 621 simple_lock(&mp->mnt_slock); 622 while (mp->mnt_wcnt > 0) { 623 wakeup(mp); 624 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 625 } 626 simple_unlock(&mp->mnt_slock); 627 vfs_hooks_unmount(mp); 628 free(mp, M_MOUNT); 629 return (0); 630 } 631 632 /* 633 * Sync each mounted filesystem. 634 */ 635 #ifdef DEBUG 636 int syncprt = 0; 637 struct ctldebug debug0 = { "syncprt", &syncprt }; 638 #endif 639 640 /* ARGSUSED */ 641 int 642 sys_sync(struct lwp *l, void *v, register_t *retval) 643 { 644 struct mount *mp, *nmp; 645 int asyncflag; 646 struct proc *p = l == NULL ? &proc0 : l->l_proc; 647 648 simple_lock(&mountlist_slock); 649 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 650 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 651 nmp = mp->mnt_list.cqe_prev; 652 continue; 653 } 654 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 655 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 656 asyncflag = mp->mnt_flag & MNT_ASYNC; 657 mp->mnt_flag &= ~MNT_ASYNC; 658 VFS_SYNC(mp, MNT_NOWAIT, p->p_cred, l); 659 if (asyncflag) 660 mp->mnt_flag |= MNT_ASYNC; 661 vn_finished_write(mp, 0); 662 } 663 simple_lock(&mountlist_slock); 664 nmp = mp->mnt_list.cqe_prev; 665 vfs_unbusy(mp); 666 667 } 668 simple_unlock(&mountlist_slock); 669 #ifdef DEBUG 670 if (syncprt) 671 vfs_bufstats(); 672 #endif /* DEBUG */ 673 return (0); 674 } 675 676 /* 677 * Change filesystem quotas. 678 */ 679 /* ARGSUSED */ 680 int 681 sys_quotactl(struct lwp *l, void *v, register_t *retval) 682 { 683 struct sys_quotactl_args /* { 684 syscallarg(const char *) path; 685 syscallarg(int) cmd; 686 syscallarg(int) uid; 687 syscallarg(caddr_t) arg; 688 } */ *uap = v; 689 struct mount *mp; 690 int error; 691 struct nameidata nd; 692 693 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 694 if ((error = namei(&nd)) != 0) 695 return (error); 696 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 697 vrele(nd.ni_vp); 698 if (error) 699 return (error); 700 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 701 SCARG(uap, arg), l); 702 vn_finished_write(mp, 0); 703 return (error); 704 } 705 706 int 707 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 708 int root) 709 { 710 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 711 int error = 0; 712 713 /* 714 * If MNT_NOWAIT or MNT_LAZY is specified, do not 715 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 716 * overrides MNT_NOWAIT. 717 */ 718 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 719 (flags != MNT_WAIT && flags != 0)) { 720 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 721 goto done; 722 } 723 724 /* Get the filesystem stats now */ 725 memset(sp, 0, sizeof(*sp)); 726 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 727 return error; 728 } 729 730 if (cwdi->cwdi_rdir == NULL) 731 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 732 done: 733 if (cwdi->cwdi_rdir != NULL) { 734 size_t len; 735 char *bp; 736 char *path = PNBUF_GET(); 737 if (!path) 738 return ENOMEM; 739 740 bp = path + MAXPATHLEN; 741 *--bp = '\0'; 742 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 743 MAXPATHLEN / 2, 0, l); 744 if (error) { 745 PNBUF_PUT(path); 746 return error; 747 } 748 len = strlen(bp); 749 /* 750 * for mount points that are below our root, we can see 751 * them, so we fix up the pathname and return them. The 752 * rest we cannot see, so we don't allow viewing the 753 * data. 754 */ 755 if (strncmp(bp, sp->f_mntonname, len) == 0) { 756 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 757 sizeof(sp->f_mntonname)); 758 if (sp->f_mntonname[0] == '\0') 759 (void)strlcpy(sp->f_mntonname, "/", 760 sizeof(sp->f_mntonname)); 761 } else { 762 if (root) 763 (void)strlcpy(sp->f_mntonname, "/", 764 sizeof(sp->f_mntonname)); 765 else 766 error = EPERM; 767 } 768 PNBUF_PUT(path); 769 } 770 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 771 return error; 772 } 773 774 /* 775 * Get filesystem statistics. 776 */ 777 /* ARGSUSED */ 778 int 779 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 780 { 781 struct sys_statvfs1_args /* { 782 syscallarg(const char *) path; 783 syscallarg(struct statvfs *) buf; 784 syscallarg(int) flags; 785 } */ *uap = v; 786 struct mount *mp; 787 struct statvfs sbuf; 788 int error; 789 struct nameidata nd; 790 791 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 792 if ((error = namei(&nd)) != 0) 793 return error; 794 mp = nd.ni_vp->v_mount; 795 vrele(nd.ni_vp); 796 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 797 return error; 798 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 799 } 800 801 /* 802 * Get filesystem statistics. 803 */ 804 /* ARGSUSED */ 805 int 806 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 807 { 808 struct sys_fstatvfs1_args /* { 809 syscallarg(int) fd; 810 syscallarg(struct statvfs *) buf; 811 syscallarg(int) flags; 812 } */ *uap = v; 813 struct proc *p = l->l_proc; 814 struct file *fp; 815 struct mount *mp; 816 struct statvfs sbuf; 817 int error; 818 819 /* getvnode() will use the descriptor for us */ 820 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 821 return (error); 822 mp = ((struct vnode *)fp->f_data)->v_mount; 823 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 824 goto out; 825 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 826 out: 827 FILE_UNUSE(fp, l); 828 return error; 829 } 830 831 832 /* 833 * Get statistics on all filesystems. 834 */ 835 int 836 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 837 { 838 struct sys_getvfsstat_args /* { 839 syscallarg(struct statvfs *) buf; 840 syscallarg(size_t) bufsize; 841 syscallarg(int) flags; 842 } */ *uap = v; 843 int root = 0; 844 struct proc *p = l->l_proc; 845 struct mount *mp, *nmp; 846 struct statvfs sbuf; 847 struct statvfs *sfsp; 848 size_t count, maxcount; 849 int error = 0; 850 851 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 852 sfsp = SCARG(uap, buf); 853 simple_lock(&mountlist_slock); 854 count = 0; 855 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 856 mp = nmp) { 857 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 858 nmp = CIRCLEQ_NEXT(mp, mnt_list); 859 continue; 860 } 861 if (sfsp && count < maxcount) { 862 error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 0); 863 if (error) { 864 simple_lock(&mountlist_slock); 865 nmp = CIRCLEQ_NEXT(mp, mnt_list); 866 vfs_unbusy(mp); 867 continue; 868 } 869 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 870 if (error) { 871 vfs_unbusy(mp); 872 return (error); 873 } 874 sfsp++; 875 root |= strcmp(sbuf.f_mntonname, "/") == 0; 876 } 877 count++; 878 simple_lock(&mountlist_slock); 879 nmp = CIRCLEQ_NEXT(mp, mnt_list); 880 vfs_unbusy(mp); 881 } 882 simple_unlock(&mountlist_slock); 883 if (root == 0 && p->p_cwdi->cwdi_rdir) { 884 /* 885 * fake a root entry 886 */ 887 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, l, 888 SCARG(uap, flags), 1)) != 0) 889 return error; 890 if (sfsp) 891 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 892 count++; 893 } 894 if (sfsp && count > maxcount) 895 *retval = maxcount; 896 else 897 *retval = count; 898 return error; 899 } 900 901 /* 902 * Change current working directory to a given file descriptor. 903 */ 904 /* ARGSUSED */ 905 int 906 sys_fchdir(struct lwp *l, void *v, register_t *retval) 907 { 908 struct sys_fchdir_args /* { 909 syscallarg(int) fd; 910 } */ *uap = v; 911 struct proc *p = l->l_proc; 912 struct filedesc *fdp = p->p_fd; 913 struct cwdinfo *cwdi = p->p_cwdi; 914 struct vnode *vp, *tdp; 915 struct mount *mp; 916 struct file *fp; 917 int error; 918 919 /* getvnode() will use the descriptor for us */ 920 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 921 return (error); 922 vp = (struct vnode *)fp->f_data; 923 924 VREF(vp); 925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 926 if (vp->v_type != VDIR) 927 error = ENOTDIR; 928 else 929 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 930 while (!error && (mp = vp->v_mountedhere) != NULL) { 931 if (vfs_busy(mp, 0, 0)) 932 continue; 933 error = VFS_ROOT(mp, &tdp); 934 vfs_unbusy(mp); 935 if (error) 936 break; 937 vput(vp); 938 vp = tdp; 939 } 940 if (error) { 941 vput(vp); 942 goto out; 943 } 944 VOP_UNLOCK(vp, 0); 945 946 /* 947 * Disallow changing to a directory not under the process's 948 * current root directory (if there is one). 949 */ 950 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 951 vrele(vp); 952 error = EPERM; /* operation not permitted */ 953 goto out; 954 } 955 956 vrele(cwdi->cwdi_cdir); 957 cwdi->cwdi_cdir = vp; 958 out: 959 FILE_UNUSE(fp, l); 960 return (error); 961 } 962 963 /* 964 * Change this process's notion of the root directory to a given file 965 * descriptor. 966 */ 967 int 968 sys_fchroot(struct lwp *l, void *v, register_t *retval) 969 { 970 struct sys_fchroot_args *uap = v; 971 struct proc *p = l->l_proc; 972 struct filedesc *fdp = p->p_fd; 973 struct cwdinfo *cwdi = p->p_cwdi; 974 struct vnode *vp; 975 struct file *fp; 976 int error; 977 978 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 979 &p->p_acflag)) != 0) 980 return error; 981 /* getvnode() will use the descriptor for us */ 982 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 983 return error; 984 vp = (struct vnode *) fp->f_data; 985 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 986 if (vp->v_type != VDIR) 987 error = ENOTDIR; 988 else 989 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 990 VOP_UNLOCK(vp, 0); 991 if (error) 992 goto out; 993 VREF(vp); 994 995 /* 996 * Prevent escaping from chroot by putting the root under 997 * the working directory. Silently chdir to / if we aren't 998 * already there. 999 */ 1000 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1001 /* 1002 * XXX would be more failsafe to change directory to a 1003 * deadfs node here instead 1004 */ 1005 vrele(cwdi->cwdi_cdir); 1006 VREF(vp); 1007 cwdi->cwdi_cdir = vp; 1008 } 1009 1010 if (cwdi->cwdi_rdir != NULL) 1011 vrele(cwdi->cwdi_rdir); 1012 cwdi->cwdi_rdir = vp; 1013 out: 1014 FILE_UNUSE(fp, l); 1015 return (error); 1016 } 1017 1018 /* 1019 * Change current working directory (``.''). 1020 */ 1021 /* ARGSUSED */ 1022 int 1023 sys_chdir(struct lwp *l, void *v, register_t *retval) 1024 { 1025 struct sys_chdir_args /* { 1026 syscallarg(const char *) path; 1027 } */ *uap = v; 1028 struct proc *p = l->l_proc; 1029 struct cwdinfo *cwdi = p->p_cwdi; 1030 int error; 1031 struct nameidata nd; 1032 1033 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1034 SCARG(uap, path), l); 1035 if ((error = change_dir(&nd, l)) != 0) 1036 return (error); 1037 vrele(cwdi->cwdi_cdir); 1038 cwdi->cwdi_cdir = nd.ni_vp; 1039 return (0); 1040 } 1041 1042 /* 1043 * Change notion of root (``/'') directory. 1044 */ 1045 /* ARGSUSED */ 1046 int 1047 sys_chroot(struct lwp *l, void *v, register_t *retval) 1048 { 1049 struct sys_chroot_args /* { 1050 syscallarg(const char *) path; 1051 } */ *uap = v; 1052 struct proc *p = l->l_proc; 1053 struct cwdinfo *cwdi = p->p_cwdi; 1054 struct vnode *vp; 1055 int error; 1056 struct nameidata nd; 1057 1058 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1059 &p->p_acflag)) != 0) 1060 return (error); 1061 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1062 SCARG(uap, path), l); 1063 if ((error = change_dir(&nd, l)) != 0) 1064 return (error); 1065 if (cwdi->cwdi_rdir != NULL) 1066 vrele(cwdi->cwdi_rdir); 1067 vp = nd.ni_vp; 1068 cwdi->cwdi_rdir = vp; 1069 1070 /* 1071 * Prevent escaping from chroot by putting the root under 1072 * the working directory. Silently chdir to / if we aren't 1073 * already there. 1074 */ 1075 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1076 /* 1077 * XXX would be more failsafe to change directory to a 1078 * deadfs node here instead 1079 */ 1080 vrele(cwdi->cwdi_cdir); 1081 VREF(vp); 1082 cwdi->cwdi_cdir = vp; 1083 } 1084 1085 return (0); 1086 } 1087 1088 /* 1089 * Common routine for chroot and chdir. 1090 */ 1091 static int 1092 change_dir(struct nameidata *ndp, struct lwp *l) 1093 { 1094 struct vnode *vp; 1095 int error; 1096 1097 if ((error = namei(ndp)) != 0) 1098 return (error); 1099 vp = ndp->ni_vp; 1100 if (vp->v_type != VDIR) 1101 error = ENOTDIR; 1102 else 1103 error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_cred, l); 1104 1105 if (error) 1106 vput(vp); 1107 else 1108 VOP_UNLOCK(vp, 0); 1109 return (error); 1110 } 1111 1112 /* 1113 * Check permissions, allocate an open file structure, 1114 * and call the device open routine if any. 1115 */ 1116 int 1117 sys_open(struct lwp *l, void *v, register_t *retval) 1118 { 1119 struct sys_open_args /* { 1120 syscallarg(const char *) path; 1121 syscallarg(int) flags; 1122 syscallarg(int) mode; 1123 } */ *uap = v; 1124 struct proc *p = l->l_proc; 1125 struct cwdinfo *cwdi = p->p_cwdi; 1126 struct filedesc *fdp = p->p_fd; 1127 struct file *fp; 1128 struct vnode *vp; 1129 int flags, cmode; 1130 int type, indx, error; 1131 struct flock lf; 1132 struct nameidata nd; 1133 1134 flags = FFLAGS(SCARG(uap, flags)); 1135 if ((flags & (FREAD | FWRITE)) == 0) 1136 return (EINVAL); 1137 /* falloc() will use the file descriptor for us */ 1138 if ((error = falloc(p, &fp, &indx)) != 0) 1139 return (error); 1140 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1141 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1142 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1143 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1144 FILE_UNUSE(fp, l); 1145 fdp->fd_ofiles[indx] = NULL; 1146 ffree(fp); 1147 if ((error == EDUPFD || error == EMOVEFD) && 1148 l->l_dupfd >= 0 && /* XXX from fdopen */ 1149 (error = 1150 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1151 *retval = indx; 1152 return (0); 1153 } 1154 if (error == ERESTART) 1155 error = EINTR; 1156 fdremove(fdp, indx); 1157 return (error); 1158 } 1159 l->l_dupfd = 0; 1160 vp = nd.ni_vp; 1161 fp->f_flag = flags & FMASK; 1162 fp->f_type = DTYPE_VNODE; 1163 fp->f_ops = &vnops; 1164 fp->f_data = vp; 1165 if (flags & (O_EXLOCK | O_SHLOCK)) { 1166 lf.l_whence = SEEK_SET; 1167 lf.l_start = 0; 1168 lf.l_len = 0; 1169 if (flags & O_EXLOCK) 1170 lf.l_type = F_WRLCK; 1171 else 1172 lf.l_type = F_RDLCK; 1173 type = F_FLOCK; 1174 if ((flags & FNONBLOCK) == 0) 1175 type |= F_WAIT; 1176 VOP_UNLOCK(vp, 0); 1177 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1178 if (error) { 1179 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1180 FILE_UNUSE(fp, l); 1181 ffree(fp); 1182 fdremove(fdp, indx); 1183 return (error); 1184 } 1185 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1186 fp->f_flag |= FHASLOCK; 1187 } 1188 VOP_UNLOCK(vp, 0); 1189 *retval = indx; 1190 FILE_SET_MATURE(fp); 1191 FILE_UNUSE(fp, l); 1192 return (0); 1193 } 1194 1195 /* 1196 * Get file handle system call 1197 */ 1198 int 1199 sys_getfh(struct lwp *l, void *v, register_t *retval) 1200 { 1201 struct sys_getfh_args /* { 1202 syscallarg(char *) fname; 1203 syscallarg(fhandle_t *) fhp; 1204 } */ *uap = v; 1205 struct proc *p = l->l_proc; 1206 struct vnode *vp; 1207 fhandle_t fh; 1208 int error; 1209 struct nameidata nd; 1210 1211 /* 1212 * Must be super user 1213 */ 1214 error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1215 &p->p_acflag); 1216 if (error) 1217 return (error); 1218 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1219 SCARG(uap, fname), l); 1220 error = namei(&nd); 1221 if (error) 1222 return (error); 1223 vp = nd.ni_vp; 1224 if (vp->v_mount->mnt_op->vfs_vptofh == NULL) 1225 return EOPNOTSUPP; 1226 memset(&fh, 0, sizeof(fh)); 1227 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1228 error = VFS_VPTOFH(vp, &fh.fh_fid); 1229 vput(vp); 1230 if (error) 1231 return (error); 1232 KASSERT(fh.fh_fid.fid_len <= _VFS_MAXFIDSZ); 1233 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1234 return (error); 1235 } 1236 1237 /* 1238 * Open a file given a file handle. 1239 * 1240 * Check permissions, allocate an open file structure, 1241 * and call the device open routine if any. 1242 */ 1243 int 1244 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1245 { 1246 struct sys_fhopen_args /* { 1247 syscallarg(const fhandle_t *) fhp; 1248 syscallarg(int) flags; 1249 } */ *uap = v; 1250 struct proc *p = l->l_proc; 1251 struct filedesc *fdp = p->p_fd; 1252 struct file *fp; 1253 struct vnode *vp = NULL; 1254 struct mount *mp; 1255 kauth_cred_t cred = p->p_cred; 1256 int flags; 1257 struct file *nfp; 1258 int type, indx, error=0; 1259 struct flock lf; 1260 struct vattr va; 1261 fhandle_t fh; 1262 1263 /* 1264 * Must be super user 1265 */ 1266 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1267 &p->p_acflag))) 1268 return (error); 1269 1270 flags = FFLAGS(SCARG(uap, flags)); 1271 if ((flags & (FREAD | FWRITE)) == 0) 1272 return (EINVAL); 1273 if ((flags & O_CREAT)) 1274 return (EINVAL); 1275 /* falloc() will use the file descriptor for us */ 1276 if ((error = falloc(p, &nfp, &indx)) != 0) 1277 return (error); 1278 fp = nfp; 1279 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1280 goto bad; 1281 1282 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1283 error = ESTALE; 1284 goto bad; 1285 } 1286 1287 if (mp->mnt_op->vfs_fhtovp == NULL) { 1288 error = EOPNOTSUPP; 1289 goto bad; 1290 } 1291 1292 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1293 vp = NULL; /* most likely unnecessary sanity for bad: */ 1294 goto bad; 1295 } 1296 1297 /* Now do an effective vn_open */ 1298 1299 if (vp->v_type == VSOCK) { 1300 error = EOPNOTSUPP; 1301 goto bad; 1302 } 1303 if (flags & FREAD) { 1304 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1305 goto bad; 1306 } 1307 if (flags & (FWRITE | O_TRUNC)) { 1308 if (vp->v_type == VDIR) { 1309 error = EISDIR; 1310 goto bad; 1311 } 1312 if ((error = vn_writechk(vp)) != 0 || 1313 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1314 goto bad; 1315 } 1316 if (flags & O_TRUNC) { 1317 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1318 goto bad; 1319 VOP_UNLOCK(vp, 0); /* XXX */ 1320 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1321 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1322 VATTR_NULL(&va); 1323 va.va_size = 0; 1324 error = VOP_SETATTR(vp, &va, cred, l); 1325 vn_finished_write(mp, 0); 1326 if (error) 1327 goto bad; 1328 } 1329 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1330 goto bad; 1331 if (vp->v_type == VREG && 1332 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1333 error = EIO; 1334 goto bad; 1335 } 1336 if (flags & FWRITE) 1337 vp->v_writecount++; 1338 1339 /* done with modified vn_open, now finish what sys_open does. */ 1340 1341 fp->f_flag = flags & FMASK; 1342 fp->f_type = DTYPE_VNODE; 1343 fp->f_ops = &vnops; 1344 fp->f_data = vp; 1345 if (flags & (O_EXLOCK | O_SHLOCK)) { 1346 lf.l_whence = SEEK_SET; 1347 lf.l_start = 0; 1348 lf.l_len = 0; 1349 if (flags & O_EXLOCK) 1350 lf.l_type = F_WRLCK; 1351 else 1352 lf.l_type = F_RDLCK; 1353 type = F_FLOCK; 1354 if ((flags & FNONBLOCK) == 0) 1355 type |= F_WAIT; 1356 VOP_UNLOCK(vp, 0); 1357 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1358 if (error) { 1359 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1360 FILE_UNUSE(fp, l); 1361 ffree(fp); 1362 fdremove(fdp, indx); 1363 return (error); 1364 } 1365 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1366 fp->f_flag |= FHASLOCK; 1367 } 1368 VOP_UNLOCK(vp, 0); 1369 *retval = indx; 1370 FILE_SET_MATURE(fp); 1371 FILE_UNUSE(fp, l); 1372 return (0); 1373 1374 bad: 1375 FILE_UNUSE(fp, l); 1376 ffree(fp); 1377 fdremove(fdp, indx); 1378 if (vp != NULL) 1379 vput(vp); 1380 return (error); 1381 } 1382 1383 /* ARGSUSED */ 1384 int 1385 sys_fhstat(struct lwp *l, void *v, register_t *retval) 1386 { 1387 struct sys_fhstat_args /* { 1388 syscallarg(const fhandle_t *) fhp; 1389 syscallarg(struct stat *) sb; 1390 } */ *uap = v; 1391 struct proc *p = l->l_proc; 1392 struct stat sb; 1393 int error; 1394 fhandle_t fh; 1395 struct mount *mp; 1396 struct vnode *vp; 1397 1398 /* 1399 * Must be super user 1400 */ 1401 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1402 &p->p_acflag))) 1403 return (error); 1404 1405 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1406 return (error); 1407 1408 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1409 return (ESTALE); 1410 if (mp->mnt_op->vfs_fhtovp == NULL) 1411 return EOPNOTSUPP; 1412 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1413 return (error); 1414 error = vn_stat(vp, &sb, l); 1415 vput(vp); 1416 if (error) 1417 return (error); 1418 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1419 return (error); 1420 } 1421 1422 /* ARGSUSED */ 1423 int 1424 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1425 { 1426 struct sys_fhstatvfs1_args /* { 1427 syscallarg(const fhandle_t *) fhp; 1428 syscallarg(struct statvfs *) buf; 1429 syscallarg(int) flags; 1430 } */ *uap = v; 1431 struct proc *p = l->l_proc; 1432 struct statvfs sbuf; 1433 fhandle_t fh; 1434 struct mount *mp; 1435 struct vnode *vp; 1436 int error; 1437 1438 /* 1439 * Must be super user 1440 */ 1441 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1442 &p->p_acflag)) != 0) 1443 return error; 1444 1445 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1446 return error; 1447 1448 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1449 return ESTALE; 1450 if (mp->mnt_op->vfs_fhtovp == NULL) 1451 return EOPNOTSUPP; 1452 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1453 return error; 1454 1455 mp = vp->v_mount; 1456 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) { 1457 vput(vp); 1458 return error; 1459 } 1460 vput(vp); 1461 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1462 } 1463 1464 /* 1465 * Create a special file. 1466 */ 1467 /* ARGSUSED */ 1468 int 1469 sys_mknod(struct lwp *l, void *v, register_t *retval) 1470 { 1471 struct sys_mknod_args /* { 1472 syscallarg(const char *) path; 1473 syscallarg(int) mode; 1474 syscallarg(int) dev; 1475 } */ *uap = v; 1476 struct proc *p = l->l_proc; 1477 struct vnode *vp; 1478 struct mount *mp; 1479 struct vattr vattr; 1480 int error; 1481 int whiteout = 0; 1482 struct nameidata nd; 1483 1484 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1485 &p->p_acflag)) != 0) 1486 return (error); 1487 restart: 1488 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1489 if ((error = namei(&nd)) != 0) 1490 return (error); 1491 vp = nd.ni_vp; 1492 if (vp != NULL) 1493 error = EEXIST; 1494 else { 1495 VATTR_NULL(&vattr); 1496 vattr.va_mode = 1497 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1498 vattr.va_rdev = SCARG(uap, dev); 1499 whiteout = 0; 1500 1501 switch (SCARG(uap, mode) & S_IFMT) { 1502 case S_IFMT: /* used by badsect to flag bad sectors */ 1503 vattr.va_type = VBAD; 1504 break; 1505 case S_IFCHR: 1506 vattr.va_type = VCHR; 1507 break; 1508 case S_IFBLK: 1509 vattr.va_type = VBLK; 1510 break; 1511 case S_IFWHT: 1512 whiteout = 1; 1513 break; 1514 default: 1515 error = EINVAL; 1516 break; 1517 } 1518 } 1519 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1520 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1521 if (nd.ni_dvp == vp) 1522 vrele(nd.ni_dvp); 1523 else 1524 vput(nd.ni_dvp); 1525 if (vp) 1526 vrele(vp); 1527 if ((error = vn_start_write(NULL, &mp, 1528 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1529 return (error); 1530 goto restart; 1531 } 1532 if (!error) { 1533 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1534 if (whiteout) { 1535 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1536 if (error) 1537 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1538 vput(nd.ni_dvp); 1539 } else { 1540 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1541 &nd.ni_cnd, &vattr); 1542 if (error == 0) 1543 vput(nd.ni_vp); 1544 } 1545 } else { 1546 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1547 if (nd.ni_dvp == vp) 1548 vrele(nd.ni_dvp); 1549 else 1550 vput(nd.ni_dvp); 1551 if (vp) 1552 vrele(vp); 1553 } 1554 vn_finished_write(mp, 0); 1555 return (error); 1556 } 1557 1558 /* 1559 * Create a named pipe. 1560 */ 1561 /* ARGSUSED */ 1562 int 1563 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1564 { 1565 struct sys_mkfifo_args /* { 1566 syscallarg(const char *) path; 1567 syscallarg(int) mode; 1568 } */ *uap = v; 1569 struct proc *p = l->l_proc; 1570 struct mount *mp; 1571 struct vattr vattr; 1572 int error; 1573 struct nameidata nd; 1574 1575 restart: 1576 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1577 if ((error = namei(&nd)) != 0) 1578 return (error); 1579 if (nd.ni_vp != NULL) { 1580 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1581 if (nd.ni_dvp == nd.ni_vp) 1582 vrele(nd.ni_dvp); 1583 else 1584 vput(nd.ni_dvp); 1585 vrele(nd.ni_vp); 1586 return (EEXIST); 1587 } 1588 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1589 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1590 if (nd.ni_dvp == nd.ni_vp) 1591 vrele(nd.ni_dvp); 1592 else 1593 vput(nd.ni_dvp); 1594 if (nd.ni_vp) 1595 vrele(nd.ni_vp); 1596 if ((error = vn_start_write(NULL, &mp, 1597 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1598 return (error); 1599 goto restart; 1600 } 1601 VATTR_NULL(&vattr); 1602 vattr.va_type = VFIFO; 1603 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1604 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1605 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1606 if (error == 0) 1607 vput(nd.ni_vp); 1608 vn_finished_write(mp, 0); 1609 return (error); 1610 } 1611 1612 /* 1613 * Make a hard file link. 1614 */ 1615 /* ARGSUSED */ 1616 int 1617 sys_link(struct lwp *l, void *v, register_t *retval) 1618 { 1619 struct sys_link_args /* { 1620 syscallarg(const char *) path; 1621 syscallarg(const char *) link; 1622 } */ *uap = v; 1623 struct proc *p = l->l_proc; 1624 struct vnode *vp; 1625 struct mount *mp; 1626 struct nameidata nd; 1627 int error; 1628 1629 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1630 if ((error = namei(&nd)) != 0) 1631 return (error); 1632 vp = nd.ni_vp; 1633 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1634 vrele(vp); 1635 return (error); 1636 } 1637 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1638 if ((error = namei(&nd)) != 0) 1639 goto out; 1640 if (nd.ni_vp) { 1641 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1642 if (nd.ni_dvp == nd.ni_vp) 1643 vrele(nd.ni_dvp); 1644 else 1645 vput(nd.ni_dvp); 1646 vrele(nd.ni_vp); 1647 error = EEXIST; 1648 goto out; 1649 } 1650 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1651 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1652 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1653 out: 1654 vrele(vp); 1655 vn_finished_write(mp, 0); 1656 return (error); 1657 } 1658 1659 /* 1660 * Make a symbolic link. 1661 */ 1662 /* ARGSUSED */ 1663 int 1664 sys_symlink(struct lwp *l, void *v, register_t *retval) 1665 { 1666 struct sys_symlink_args /* { 1667 syscallarg(const char *) path; 1668 syscallarg(const char *) link; 1669 } */ *uap = v; 1670 struct proc *p = l->l_proc; 1671 struct mount *mp; 1672 struct vattr vattr; 1673 char *path; 1674 int error; 1675 struct nameidata nd; 1676 1677 path = PNBUF_GET(); 1678 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1679 if (error) 1680 goto out; 1681 restart: 1682 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1683 if ((error = namei(&nd)) != 0) 1684 goto out; 1685 if (nd.ni_vp) { 1686 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1687 if (nd.ni_dvp == nd.ni_vp) 1688 vrele(nd.ni_dvp); 1689 else 1690 vput(nd.ni_dvp); 1691 vrele(nd.ni_vp); 1692 error = EEXIST; 1693 goto out; 1694 } 1695 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1696 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1697 if (nd.ni_dvp == nd.ni_vp) 1698 vrele(nd.ni_dvp); 1699 else 1700 vput(nd.ni_dvp); 1701 if ((error = vn_start_write(NULL, &mp, 1702 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1703 return (error); 1704 goto restart; 1705 } 1706 VATTR_NULL(&vattr); 1707 vattr.va_type = VLNK; 1708 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1709 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1710 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1711 if (error == 0) 1712 vput(nd.ni_vp); 1713 vn_finished_write(mp, 0); 1714 out: 1715 PNBUF_PUT(path); 1716 return (error); 1717 } 1718 1719 /* 1720 * Delete a whiteout from the filesystem. 1721 */ 1722 /* ARGSUSED */ 1723 int 1724 sys_undelete(struct lwp *l, void *v, register_t *retval) 1725 { 1726 struct sys_undelete_args /* { 1727 syscallarg(const char *) path; 1728 } */ *uap = v; 1729 struct proc *p = l->l_proc; 1730 int error; 1731 struct mount *mp; 1732 struct nameidata nd; 1733 1734 restart: 1735 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1736 SCARG(uap, path), l); 1737 error = namei(&nd); 1738 if (error) 1739 return (error); 1740 1741 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1742 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1743 if (nd.ni_dvp == nd.ni_vp) 1744 vrele(nd.ni_dvp); 1745 else 1746 vput(nd.ni_dvp); 1747 if (nd.ni_vp) 1748 vrele(nd.ni_vp); 1749 return (EEXIST); 1750 } 1751 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1752 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1753 if (nd.ni_dvp == nd.ni_vp) 1754 vrele(nd.ni_dvp); 1755 else 1756 vput(nd.ni_dvp); 1757 if ((error = vn_start_write(NULL, &mp, 1758 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1759 return (error); 1760 goto restart; 1761 } 1762 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1763 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1764 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1765 vput(nd.ni_dvp); 1766 vn_finished_write(mp, 0); 1767 return (error); 1768 } 1769 1770 /* 1771 * Delete a name from the filesystem. 1772 */ 1773 /* ARGSUSED */ 1774 int 1775 sys_unlink(struct lwp *l, void *v, register_t *retval) 1776 { 1777 struct sys_unlink_args /* { 1778 syscallarg(const char *) path; 1779 } */ *uap = v; 1780 struct proc *p = l->l_proc; 1781 struct mount *mp; 1782 struct vnode *vp; 1783 int error; 1784 struct nameidata nd; 1785 1786 restart: 1787 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1788 SCARG(uap, path), l); 1789 if ((error = namei(&nd)) != 0) 1790 return (error); 1791 vp = nd.ni_vp; 1792 1793 /* 1794 * The root of a mounted filesystem cannot be deleted. 1795 */ 1796 if (vp->v_flag & VROOT) { 1797 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1798 if (nd.ni_dvp == vp) 1799 vrele(nd.ni_dvp); 1800 else 1801 vput(nd.ni_dvp); 1802 vput(vp); 1803 error = EBUSY; 1804 goto out; 1805 } 1806 1807 #ifdef VERIFIED_EXEC 1808 /* Handle remove requests for veriexec entries. */ 1809 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) { 1810 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1811 if (nd.ni_dvp == vp) 1812 vrele(nd.ni_dvp); 1813 else 1814 vput(nd.ni_dvp); 1815 vput(vp); 1816 goto out; 1817 } 1818 #endif /* VERIFIED_EXEC */ 1819 1820 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1821 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1822 if (nd.ni_dvp == vp) 1823 vrele(nd.ni_dvp); 1824 else 1825 vput(nd.ni_dvp); 1826 vput(vp); 1827 if ((error = vn_start_write(NULL, &mp, 1828 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1829 return (error); 1830 goto restart; 1831 } 1832 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1833 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1834 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1835 vn_finished_write(mp, 0); 1836 out: 1837 return (error); 1838 } 1839 1840 /* 1841 * Reposition read/write file offset. 1842 */ 1843 int 1844 sys_lseek(struct lwp *l, void *v, register_t *retval) 1845 { 1846 struct sys_lseek_args /* { 1847 syscallarg(int) fd; 1848 syscallarg(int) pad; 1849 syscallarg(off_t) offset; 1850 syscallarg(int) whence; 1851 } */ *uap = v; 1852 struct proc *p = l->l_proc; 1853 kauth_cred_t cred = p->p_cred; 1854 struct filedesc *fdp = p->p_fd; 1855 struct file *fp; 1856 struct vnode *vp; 1857 struct vattr vattr; 1858 off_t newoff; 1859 int error; 1860 1861 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1862 return (EBADF); 1863 1864 FILE_USE(fp); 1865 1866 vp = (struct vnode *)fp->f_data; 1867 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1868 error = ESPIPE; 1869 goto out; 1870 } 1871 1872 switch (SCARG(uap, whence)) { 1873 case SEEK_CUR: 1874 newoff = fp->f_offset + SCARG(uap, offset); 1875 break; 1876 case SEEK_END: 1877 error = VOP_GETATTR(vp, &vattr, cred, l); 1878 if (error) 1879 goto out; 1880 newoff = SCARG(uap, offset) + vattr.va_size; 1881 break; 1882 case SEEK_SET: 1883 newoff = SCARG(uap, offset); 1884 break; 1885 default: 1886 error = EINVAL; 1887 goto out; 1888 } 1889 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1890 goto out; 1891 1892 *(off_t *)retval = fp->f_offset = newoff; 1893 out: 1894 FILE_UNUSE(fp, l); 1895 return (error); 1896 } 1897 1898 /* 1899 * Positional read system call. 1900 */ 1901 int 1902 sys_pread(struct lwp *l, void *v, register_t *retval) 1903 { 1904 struct sys_pread_args /* { 1905 syscallarg(int) fd; 1906 syscallarg(void *) buf; 1907 syscallarg(size_t) nbyte; 1908 syscallarg(off_t) offset; 1909 } */ *uap = v; 1910 struct proc *p = l->l_proc; 1911 struct filedesc *fdp = p->p_fd; 1912 struct file *fp; 1913 struct vnode *vp; 1914 off_t offset; 1915 int error, fd = SCARG(uap, fd); 1916 1917 if ((fp = fd_getfile(fdp, fd)) == NULL) 1918 return (EBADF); 1919 1920 if ((fp->f_flag & FREAD) == 0) { 1921 simple_unlock(&fp->f_slock); 1922 return (EBADF); 1923 } 1924 1925 FILE_USE(fp); 1926 1927 vp = (struct vnode *)fp->f_data; 1928 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1929 error = ESPIPE; 1930 goto out; 1931 } 1932 1933 offset = SCARG(uap, offset); 1934 1935 /* 1936 * XXX This works because no file systems actually 1937 * XXX take any action on the seek operation. 1938 */ 1939 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1940 goto out; 1941 1942 /* dofileread() will unuse the descriptor for us */ 1943 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1944 &offset, 0, retval)); 1945 1946 out: 1947 FILE_UNUSE(fp, l); 1948 return (error); 1949 } 1950 1951 /* 1952 * Positional scatter read system call. 1953 */ 1954 int 1955 sys_preadv(struct lwp *l, void *v, register_t *retval) 1956 { 1957 struct sys_preadv_args /* { 1958 syscallarg(int) fd; 1959 syscallarg(const struct iovec *) iovp; 1960 syscallarg(int) iovcnt; 1961 syscallarg(off_t) offset; 1962 } */ *uap = v; 1963 struct proc *p = l->l_proc; 1964 struct filedesc *fdp = p->p_fd; 1965 struct file *fp; 1966 struct vnode *vp; 1967 off_t offset; 1968 int error, fd = SCARG(uap, fd); 1969 1970 if ((fp = fd_getfile(fdp, fd)) == NULL) 1971 return (EBADF); 1972 1973 if ((fp->f_flag & FREAD) == 0) { 1974 simple_unlock(&fp->f_slock); 1975 return (EBADF); 1976 } 1977 1978 FILE_USE(fp); 1979 1980 vp = (struct vnode *)fp->f_data; 1981 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1982 error = ESPIPE; 1983 goto out; 1984 } 1985 1986 offset = SCARG(uap, offset); 1987 1988 /* 1989 * XXX This works because no file systems actually 1990 * XXX take any action on the seek operation. 1991 */ 1992 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1993 goto out; 1994 1995 /* dofilereadv() will unuse the descriptor for us */ 1996 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1997 &offset, 0, retval)); 1998 1999 out: 2000 FILE_UNUSE(fp, l); 2001 return (error); 2002 } 2003 2004 /* 2005 * Positional write system call. 2006 */ 2007 int 2008 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2009 { 2010 struct sys_pwrite_args /* { 2011 syscallarg(int) fd; 2012 syscallarg(const void *) buf; 2013 syscallarg(size_t) nbyte; 2014 syscallarg(off_t) offset; 2015 } */ *uap = v; 2016 struct proc *p = l->l_proc; 2017 struct filedesc *fdp = p->p_fd; 2018 struct file *fp; 2019 struct vnode *vp; 2020 off_t offset; 2021 int error, fd = SCARG(uap, fd); 2022 2023 if ((fp = fd_getfile(fdp, fd)) == NULL) 2024 return (EBADF); 2025 2026 if ((fp->f_flag & FWRITE) == 0) { 2027 simple_unlock(&fp->f_slock); 2028 return (EBADF); 2029 } 2030 2031 FILE_USE(fp); 2032 2033 vp = (struct vnode *)fp->f_data; 2034 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2035 error = ESPIPE; 2036 goto out; 2037 } 2038 2039 offset = SCARG(uap, offset); 2040 2041 /* 2042 * XXX This works because no file systems actually 2043 * XXX take any action on the seek operation. 2044 */ 2045 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2046 goto out; 2047 2048 /* dofilewrite() will unuse the descriptor for us */ 2049 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2050 &offset, 0, retval)); 2051 2052 out: 2053 FILE_UNUSE(fp, l); 2054 return (error); 2055 } 2056 2057 /* 2058 * Positional gather write system call. 2059 */ 2060 int 2061 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2062 { 2063 struct sys_pwritev_args /* { 2064 syscallarg(int) fd; 2065 syscallarg(const struct iovec *) iovp; 2066 syscallarg(int) iovcnt; 2067 syscallarg(off_t) offset; 2068 } */ *uap = v; 2069 struct proc *p = l->l_proc; 2070 struct filedesc *fdp = p->p_fd; 2071 struct file *fp; 2072 struct vnode *vp; 2073 off_t offset; 2074 int error, fd = SCARG(uap, fd); 2075 2076 if ((fp = fd_getfile(fdp, fd)) == NULL) 2077 return (EBADF); 2078 2079 if ((fp->f_flag & FWRITE) == 0) { 2080 simple_unlock(&fp->f_slock); 2081 return (EBADF); 2082 } 2083 2084 FILE_USE(fp); 2085 2086 vp = (struct vnode *)fp->f_data; 2087 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2088 error = ESPIPE; 2089 goto out; 2090 } 2091 2092 offset = SCARG(uap, offset); 2093 2094 /* 2095 * XXX This works because no file systems actually 2096 * XXX take any action on the seek operation. 2097 */ 2098 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2099 goto out; 2100 2101 /* dofilewritev() will unuse the descriptor for us */ 2102 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2103 &offset, 0, retval)); 2104 2105 out: 2106 FILE_UNUSE(fp, l); 2107 return (error); 2108 } 2109 2110 /* 2111 * Check access permissions. 2112 */ 2113 int 2114 sys_access(struct lwp *l, void *v, register_t *retval) 2115 { 2116 struct sys_access_args /* { 2117 syscallarg(const char *) path; 2118 syscallarg(int) flags; 2119 } */ *uap = v; 2120 struct proc *p = l->l_proc; 2121 kauth_cred_t cred; 2122 struct vnode *vp; 2123 int error, flags; 2124 struct nameidata nd; 2125 2126 cred = kauth_cred_dup(p->p_cred); 2127 kauth_cred_seteuid(cred, kauth_cred_getuid(p->p_cred)); 2128 kauth_cred_setegid(cred, kauth_cred_getgid(p->p_cred)); 2129 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2130 SCARG(uap, path), l); 2131 /* Override default credentials */ 2132 nd.ni_cnd.cn_cred = cred; 2133 if ((error = namei(&nd)) != 0) 2134 goto out; 2135 vp = nd.ni_vp; 2136 2137 /* Flags == 0 means only check for existence. */ 2138 if (SCARG(uap, flags)) { 2139 flags = 0; 2140 if (SCARG(uap, flags) & R_OK) 2141 flags |= VREAD; 2142 if (SCARG(uap, flags) & W_OK) 2143 flags |= VWRITE; 2144 if (SCARG(uap, flags) & X_OK) 2145 flags |= VEXEC; 2146 2147 error = VOP_ACCESS(vp, flags, cred, l); 2148 if (!error && (flags & VWRITE)) 2149 error = vn_writechk(vp); 2150 } 2151 vput(vp); 2152 out: 2153 kauth_cred_free(cred); 2154 return (error); 2155 } 2156 2157 /* 2158 * Get file status; this version follows links. 2159 */ 2160 /* ARGSUSED */ 2161 int 2162 sys___stat30(struct lwp *l, void *v, register_t *retval) 2163 { 2164 struct sys___stat30_args /* { 2165 syscallarg(const char *) path; 2166 syscallarg(struct stat *) ub; 2167 } */ *uap = v; 2168 struct stat sb; 2169 int error; 2170 struct nameidata nd; 2171 2172 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2173 SCARG(uap, path), l); 2174 if ((error = namei(&nd)) != 0) 2175 return (error); 2176 error = vn_stat(nd.ni_vp, &sb, l); 2177 vput(nd.ni_vp); 2178 if (error) 2179 return (error); 2180 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2181 return (error); 2182 } 2183 2184 /* 2185 * Get file status; this version does not follow links. 2186 */ 2187 /* ARGSUSED */ 2188 int 2189 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2190 { 2191 struct sys___lstat30_args /* { 2192 syscallarg(const char *) path; 2193 syscallarg(struct stat *) ub; 2194 } */ *uap = v; 2195 struct stat sb; 2196 int error; 2197 struct nameidata nd; 2198 2199 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2200 SCARG(uap, path), l); 2201 if ((error = namei(&nd)) != 0) 2202 return (error); 2203 error = vn_stat(nd.ni_vp, &sb, l); 2204 vput(nd.ni_vp); 2205 if (error) 2206 return (error); 2207 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2208 return (error); 2209 } 2210 2211 /* 2212 * Get configurable pathname variables. 2213 */ 2214 /* ARGSUSED */ 2215 int 2216 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2217 { 2218 struct sys_pathconf_args /* { 2219 syscallarg(const char *) path; 2220 syscallarg(int) name; 2221 } */ *uap = v; 2222 int error; 2223 struct nameidata nd; 2224 2225 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2226 SCARG(uap, path), l); 2227 if ((error = namei(&nd)) != 0) 2228 return (error); 2229 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2230 vput(nd.ni_vp); 2231 return (error); 2232 } 2233 2234 /* 2235 * Return target name of a symbolic link. 2236 */ 2237 /* ARGSUSED */ 2238 int 2239 sys_readlink(struct lwp *l, void *v, register_t *retval) 2240 { 2241 struct sys_readlink_args /* { 2242 syscallarg(const char *) path; 2243 syscallarg(char *) buf; 2244 syscallarg(size_t) count; 2245 } */ *uap = v; 2246 struct proc *p = l->l_proc; 2247 struct vnode *vp; 2248 struct iovec aiov; 2249 struct uio auio; 2250 int error; 2251 struct nameidata nd; 2252 2253 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2254 SCARG(uap, path), l); 2255 if ((error = namei(&nd)) != 0) 2256 return (error); 2257 vp = nd.ni_vp; 2258 if (vp->v_type != VLNK) 2259 error = EINVAL; 2260 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2261 (error = VOP_ACCESS(vp, VREAD, p->p_cred, l)) == 0) { 2262 aiov.iov_base = SCARG(uap, buf); 2263 aiov.iov_len = SCARG(uap, count); 2264 auio.uio_iov = &aiov; 2265 auio.uio_iovcnt = 1; 2266 auio.uio_offset = 0; 2267 auio.uio_rw = UIO_READ; 2268 KASSERT(l == curlwp); 2269 auio.uio_vmspace = l->l_proc->p_vmspace; 2270 auio.uio_resid = SCARG(uap, count); 2271 error = VOP_READLINK(vp, &auio, p->p_cred); 2272 } 2273 vput(vp); 2274 *retval = SCARG(uap, count) - auio.uio_resid; 2275 return (error); 2276 } 2277 2278 /* 2279 * Change flags of a file given a path name. 2280 */ 2281 /* ARGSUSED */ 2282 int 2283 sys_chflags(struct lwp *l, void *v, register_t *retval) 2284 { 2285 struct sys_chflags_args /* { 2286 syscallarg(const char *) path; 2287 syscallarg(u_long) flags; 2288 } */ *uap = v; 2289 struct vnode *vp; 2290 int error; 2291 struct nameidata nd; 2292 2293 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2294 if ((error = namei(&nd)) != 0) 2295 return (error); 2296 vp = nd.ni_vp; 2297 error = change_flags(vp, SCARG(uap, flags), l); 2298 vput(vp); 2299 return (error); 2300 } 2301 2302 /* 2303 * Change flags of a file given a file descriptor. 2304 */ 2305 /* ARGSUSED */ 2306 int 2307 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2308 { 2309 struct sys_fchflags_args /* { 2310 syscallarg(int) fd; 2311 syscallarg(u_long) flags; 2312 } */ *uap = v; 2313 struct proc *p = l->l_proc; 2314 struct vnode *vp; 2315 struct file *fp; 2316 int error; 2317 2318 /* getvnode() will use the descriptor for us */ 2319 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2320 return (error); 2321 vp = (struct vnode *)fp->f_data; 2322 error = change_flags(vp, SCARG(uap, flags), l); 2323 VOP_UNLOCK(vp, 0); 2324 FILE_UNUSE(fp, l); 2325 return (error); 2326 } 2327 2328 /* 2329 * Change flags of a file given a path name; this version does 2330 * not follow links. 2331 */ 2332 int 2333 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2334 { 2335 struct sys_lchflags_args /* { 2336 syscallarg(const char *) path; 2337 syscallarg(u_long) flags; 2338 } */ *uap = v; 2339 struct vnode *vp; 2340 int error; 2341 struct nameidata nd; 2342 2343 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2344 if ((error = namei(&nd)) != 0) 2345 return (error); 2346 vp = nd.ni_vp; 2347 error = change_flags(vp, SCARG(uap, flags), l); 2348 vput(vp); 2349 return (error); 2350 } 2351 2352 /* 2353 * Common routine to change flags of a file. 2354 */ 2355 int 2356 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2357 { 2358 struct proc *p = l->l_proc; 2359 struct mount *mp; 2360 struct vattr vattr; 2361 int error; 2362 2363 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2364 return (error); 2365 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2366 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2367 /* 2368 * Non-superusers cannot change the flags on devices, even if they 2369 * own them. 2370 */ 2371 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2372 &p->p_acflag) != 0) { 2373 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2374 goto out; 2375 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2376 error = EINVAL; 2377 goto out; 2378 } 2379 } 2380 VATTR_NULL(&vattr); 2381 vattr.va_flags = flags; 2382 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2383 out: 2384 vn_finished_write(mp, 0); 2385 return (error); 2386 } 2387 2388 /* 2389 * Change mode of a file given path name; this version follows links. 2390 */ 2391 /* ARGSUSED */ 2392 int 2393 sys_chmod(struct lwp *l, void *v, register_t *retval) 2394 { 2395 struct sys_chmod_args /* { 2396 syscallarg(const char *) path; 2397 syscallarg(int) mode; 2398 } */ *uap = v; 2399 int error; 2400 struct nameidata nd; 2401 2402 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2403 if ((error = namei(&nd)) != 0) 2404 return (error); 2405 2406 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2407 2408 vrele(nd.ni_vp); 2409 return (error); 2410 } 2411 2412 /* 2413 * Change mode of a file given a file descriptor. 2414 */ 2415 /* ARGSUSED */ 2416 int 2417 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2418 { 2419 struct sys_fchmod_args /* { 2420 syscallarg(int) fd; 2421 syscallarg(int) mode; 2422 } */ *uap = v; 2423 struct proc *p = l->l_proc; 2424 struct file *fp; 2425 int error; 2426 2427 /* getvnode() will use the descriptor for us */ 2428 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2429 return (error); 2430 2431 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2432 FILE_UNUSE(fp, l); 2433 return (error); 2434 } 2435 2436 /* 2437 * Change mode of a file given path name; this version does not follow links. 2438 */ 2439 /* ARGSUSED */ 2440 int 2441 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2442 { 2443 struct sys_lchmod_args /* { 2444 syscallarg(const char *) path; 2445 syscallarg(int) mode; 2446 } */ *uap = v; 2447 int error; 2448 struct nameidata nd; 2449 2450 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2451 if ((error = namei(&nd)) != 0) 2452 return (error); 2453 2454 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2455 2456 vrele(nd.ni_vp); 2457 return (error); 2458 } 2459 2460 /* 2461 * Common routine to set mode given a vnode. 2462 */ 2463 static int 2464 change_mode(struct vnode *vp, int mode, struct lwp *l) 2465 { 2466 struct proc *p = l->l_proc; 2467 struct mount *mp; 2468 struct vattr vattr; 2469 int error; 2470 2471 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2472 return (error); 2473 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2474 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2475 VATTR_NULL(&vattr); 2476 vattr.va_mode = mode & ALLPERMS; 2477 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2478 VOP_UNLOCK(vp, 0); 2479 vn_finished_write(mp, 0); 2480 return (error); 2481 } 2482 2483 /* 2484 * Set ownership given a path name; this version follows links. 2485 */ 2486 /* ARGSUSED */ 2487 int 2488 sys_chown(struct lwp *l, void *v, register_t *retval) 2489 { 2490 struct sys_chown_args /* { 2491 syscallarg(const char *) path; 2492 syscallarg(uid_t) uid; 2493 syscallarg(gid_t) gid; 2494 } */ *uap = v; 2495 int error; 2496 struct nameidata nd; 2497 2498 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2499 if ((error = namei(&nd)) != 0) 2500 return (error); 2501 2502 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2503 2504 vrele(nd.ni_vp); 2505 return (error); 2506 } 2507 2508 /* 2509 * Set ownership given a path name; this version follows links. 2510 * Provides POSIX semantics. 2511 */ 2512 /* ARGSUSED */ 2513 int 2514 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2515 { 2516 struct sys_chown_args /* { 2517 syscallarg(const char *) path; 2518 syscallarg(uid_t) uid; 2519 syscallarg(gid_t) gid; 2520 } */ *uap = v; 2521 int error; 2522 struct nameidata nd; 2523 2524 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2525 if ((error = namei(&nd)) != 0) 2526 return (error); 2527 2528 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2529 2530 vrele(nd.ni_vp); 2531 return (error); 2532 } 2533 2534 /* 2535 * Set ownership given a file descriptor. 2536 */ 2537 /* ARGSUSED */ 2538 int 2539 sys_fchown(struct lwp *l, void *v, register_t *retval) 2540 { 2541 struct sys_fchown_args /* { 2542 syscallarg(int) fd; 2543 syscallarg(uid_t) uid; 2544 syscallarg(gid_t) gid; 2545 } */ *uap = v; 2546 struct proc *p = l->l_proc; 2547 int error; 2548 struct file *fp; 2549 2550 /* getvnode() will use the descriptor for us */ 2551 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2552 return (error); 2553 2554 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2555 SCARG(uap, gid), l, 0); 2556 FILE_UNUSE(fp, l); 2557 return (error); 2558 } 2559 2560 /* 2561 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2562 */ 2563 /* ARGSUSED */ 2564 int 2565 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2566 { 2567 struct sys_fchown_args /* { 2568 syscallarg(int) fd; 2569 syscallarg(uid_t) uid; 2570 syscallarg(gid_t) gid; 2571 } */ *uap = v; 2572 struct proc *p = l->l_proc; 2573 int error; 2574 struct file *fp; 2575 2576 /* getvnode() will use the descriptor for us */ 2577 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2578 return (error); 2579 2580 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2581 SCARG(uap, gid), l, 1); 2582 FILE_UNUSE(fp, l); 2583 return (error); 2584 } 2585 2586 /* 2587 * Set ownership given a path name; this version does not follow links. 2588 */ 2589 /* ARGSUSED */ 2590 int 2591 sys_lchown(struct lwp *l, void *v, register_t *retval) 2592 { 2593 struct sys_lchown_args /* { 2594 syscallarg(const char *) path; 2595 syscallarg(uid_t) uid; 2596 syscallarg(gid_t) gid; 2597 } */ *uap = v; 2598 int error; 2599 struct nameidata nd; 2600 2601 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2602 if ((error = namei(&nd)) != 0) 2603 return (error); 2604 2605 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2606 2607 vrele(nd.ni_vp); 2608 return (error); 2609 } 2610 2611 /* 2612 * Set ownership given a path name; this version does not follow links. 2613 * Provides POSIX/XPG semantics. 2614 */ 2615 /* ARGSUSED */ 2616 int 2617 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2618 { 2619 struct sys_lchown_args /* { 2620 syscallarg(const char *) path; 2621 syscallarg(uid_t) uid; 2622 syscallarg(gid_t) gid; 2623 } */ *uap = v; 2624 int error; 2625 struct nameidata nd; 2626 2627 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2628 if ((error = namei(&nd)) != 0) 2629 return (error); 2630 2631 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2632 2633 vrele(nd.ni_vp); 2634 return (error); 2635 } 2636 2637 /* 2638 * Common routine to set ownership given a vnode. 2639 */ 2640 static int 2641 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2642 int posix_semantics) 2643 { 2644 struct proc *p = l->l_proc; 2645 struct mount *mp; 2646 struct vattr vattr; 2647 mode_t newmode; 2648 int error; 2649 2650 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2651 return (error); 2652 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2653 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2654 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2655 goto out; 2656 2657 #define CHANGED(x) ((int)(x) != -1) 2658 newmode = vattr.va_mode; 2659 if (posix_semantics) { 2660 /* 2661 * POSIX/XPG semantics: if the caller is not the super-user, 2662 * clear set-user-id and set-group-id bits. Both POSIX and 2663 * the XPG consider the behaviour for calls by the super-user 2664 * implementation-defined; we leave the set-user-id and set- 2665 * group-id settings intact in that case. 2666 */ 2667 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2668 NULL) != 0) 2669 newmode &= ~(S_ISUID | S_ISGID); 2670 } else { 2671 /* 2672 * NetBSD semantics: when changing owner and/or group, 2673 * clear the respective bit(s). 2674 */ 2675 if (CHANGED(uid)) 2676 newmode &= ~S_ISUID; 2677 if (CHANGED(gid)) 2678 newmode &= ~S_ISGID; 2679 } 2680 /* Update va_mode iff altered. */ 2681 if (vattr.va_mode == newmode) 2682 newmode = VNOVAL; 2683 2684 VATTR_NULL(&vattr); 2685 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2686 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2687 vattr.va_mode = newmode; 2688 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2689 #undef CHANGED 2690 2691 out: 2692 VOP_UNLOCK(vp, 0); 2693 vn_finished_write(mp, 0); 2694 return (error); 2695 } 2696 2697 /* 2698 * Set the access and modification times given a path name; this 2699 * version follows links. 2700 */ 2701 /* ARGSUSED */ 2702 int 2703 sys_utimes(struct lwp *l, void *v, register_t *retval) 2704 { 2705 struct sys_utimes_args /* { 2706 syscallarg(const char *) path; 2707 syscallarg(const struct timeval *) tptr; 2708 } */ *uap = v; 2709 int error; 2710 struct nameidata nd; 2711 2712 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2713 if ((error = namei(&nd)) != 0) 2714 return (error); 2715 2716 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2717 2718 vrele(nd.ni_vp); 2719 return (error); 2720 } 2721 2722 /* 2723 * Set the access and modification times given a file descriptor. 2724 */ 2725 /* ARGSUSED */ 2726 int 2727 sys_futimes(struct lwp *l, void *v, register_t *retval) 2728 { 2729 struct sys_futimes_args /* { 2730 syscallarg(int) fd; 2731 syscallarg(const struct timeval *) tptr; 2732 } */ *uap = v; 2733 struct proc *p = l->l_proc; 2734 int error; 2735 struct file *fp; 2736 2737 /* getvnode() will use the descriptor for us */ 2738 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2739 return (error); 2740 2741 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2742 FILE_UNUSE(fp, l); 2743 return (error); 2744 } 2745 2746 /* 2747 * Set the access and modification times given a path name; this 2748 * version does not follow links. 2749 */ 2750 /* ARGSUSED */ 2751 int 2752 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2753 { 2754 struct sys_lutimes_args /* { 2755 syscallarg(const char *) path; 2756 syscallarg(const struct timeval *) tptr; 2757 } */ *uap = v; 2758 int error; 2759 struct nameidata nd; 2760 2761 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2762 if ((error = namei(&nd)) != 0) 2763 return (error); 2764 2765 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2766 2767 vrele(nd.ni_vp); 2768 return (error); 2769 } 2770 2771 /* 2772 * Common routine to set access and modification times given a vnode. 2773 */ 2774 static int 2775 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2776 { 2777 struct proc *p = l->l_proc; 2778 struct mount *mp; 2779 struct vattr vattr; 2780 int error; 2781 2782 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2783 return (error); 2784 VATTR_NULL(&vattr); 2785 if (tptr == NULL) { 2786 nanotime(&vattr.va_atime); 2787 vattr.va_mtime = vattr.va_atime; 2788 vattr.va_vaflags |= VA_UTIMES_NULL; 2789 } else { 2790 struct timeval tv[2]; 2791 2792 error = copyin(tptr, tv, sizeof(tv)); 2793 if (error) 2794 goto out; 2795 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 2796 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 2797 } 2798 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2799 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2800 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2801 VOP_UNLOCK(vp, 0); 2802 out: 2803 vn_finished_write(mp, 0); 2804 return (error); 2805 } 2806 2807 /* 2808 * Truncate a file given its path name. 2809 */ 2810 /* ARGSUSED */ 2811 int 2812 sys_truncate(struct lwp *l, void *v, register_t *retval) 2813 { 2814 struct sys_truncate_args /* { 2815 syscallarg(const char *) path; 2816 syscallarg(int) pad; 2817 syscallarg(off_t) length; 2818 } */ *uap = v; 2819 struct proc *p = l->l_proc; 2820 struct vnode *vp; 2821 struct mount *mp; 2822 struct vattr vattr; 2823 int error; 2824 struct nameidata nd; 2825 2826 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2827 if ((error = namei(&nd)) != 0) 2828 return (error); 2829 vp = nd.ni_vp; 2830 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2831 vrele(vp); 2832 return (error); 2833 } 2834 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2835 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2836 if (vp->v_type == VDIR) 2837 error = EISDIR; 2838 else if ((error = vn_writechk(vp)) == 0 && 2839 (error = VOP_ACCESS(vp, VWRITE, p->p_cred, l)) == 0) { 2840 VATTR_NULL(&vattr); 2841 vattr.va_size = SCARG(uap, length); 2842 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2843 } 2844 vput(vp); 2845 vn_finished_write(mp, 0); 2846 return (error); 2847 } 2848 2849 /* 2850 * Truncate a file given a file descriptor. 2851 */ 2852 /* ARGSUSED */ 2853 int 2854 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2855 { 2856 struct sys_ftruncate_args /* { 2857 syscallarg(int) fd; 2858 syscallarg(int) pad; 2859 syscallarg(off_t) length; 2860 } */ *uap = v; 2861 struct proc *p = l->l_proc; 2862 struct mount *mp; 2863 struct vattr vattr; 2864 struct vnode *vp; 2865 struct file *fp; 2866 int error; 2867 2868 /* getvnode() will use the descriptor for us */ 2869 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2870 return (error); 2871 if ((fp->f_flag & FWRITE) == 0) { 2872 error = EINVAL; 2873 goto out; 2874 } 2875 vp = (struct vnode *)fp->f_data; 2876 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2877 FILE_UNUSE(fp, l); 2878 return (error); 2879 } 2880 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2881 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2882 if (vp->v_type == VDIR) 2883 error = EISDIR; 2884 else if ((error = vn_writechk(vp)) == 0) { 2885 VATTR_NULL(&vattr); 2886 vattr.va_size = SCARG(uap, length); 2887 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 2888 } 2889 VOP_UNLOCK(vp, 0); 2890 vn_finished_write(mp, 0); 2891 out: 2892 FILE_UNUSE(fp, l); 2893 return (error); 2894 } 2895 2896 /* 2897 * Sync an open file. 2898 */ 2899 /* ARGSUSED */ 2900 int 2901 sys_fsync(struct lwp *l, void *v, register_t *retval) 2902 { 2903 struct sys_fsync_args /* { 2904 syscallarg(int) fd; 2905 } */ *uap = v; 2906 struct proc *p = l->l_proc; 2907 struct vnode *vp; 2908 struct mount *mp; 2909 struct file *fp; 2910 int error; 2911 2912 /* getvnode() will use the descriptor for us */ 2913 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2914 return (error); 2915 vp = (struct vnode *)fp->f_data; 2916 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2917 FILE_UNUSE(fp, l); 2918 return (error); 2919 } 2920 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2921 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 2922 if (error == 0 && bioops.io_fsync != NULL && 2923 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2924 (*bioops.io_fsync)(vp, 0); 2925 VOP_UNLOCK(vp, 0); 2926 vn_finished_write(mp, 0); 2927 FILE_UNUSE(fp, l); 2928 return (error); 2929 } 2930 2931 /* 2932 * Sync a range of file data. API modeled after that found in AIX. 2933 * 2934 * FDATASYNC indicates that we need only save enough metadata to be able 2935 * to re-read the written data. Note we duplicate AIX's requirement that 2936 * the file be open for writing. 2937 */ 2938 /* ARGSUSED */ 2939 int 2940 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2941 { 2942 struct sys_fsync_range_args /* { 2943 syscallarg(int) fd; 2944 syscallarg(int) flags; 2945 syscallarg(off_t) start; 2946 syscallarg(off_t) length; 2947 } */ *uap = v; 2948 struct proc *p = l->l_proc; 2949 struct vnode *vp; 2950 struct file *fp; 2951 int flags, nflags; 2952 off_t s, e, len; 2953 int error; 2954 2955 /* getvnode() will use the descriptor for us */ 2956 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2957 return (error); 2958 2959 if ((fp->f_flag & FWRITE) == 0) { 2960 FILE_UNUSE(fp, l); 2961 return (EBADF); 2962 } 2963 2964 flags = SCARG(uap, flags); 2965 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2966 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2967 return (EINVAL); 2968 } 2969 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2970 if (flags & FDATASYNC) 2971 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2972 else 2973 nflags = FSYNC_WAIT; 2974 if (flags & FDISKSYNC) 2975 nflags |= FSYNC_CACHE; 2976 2977 len = SCARG(uap, length); 2978 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2979 if (len) { 2980 s = SCARG(uap, start); 2981 e = s + len; 2982 if (e < s) { 2983 FILE_UNUSE(fp, l); 2984 return (EINVAL); 2985 } 2986 } else { 2987 e = 0; 2988 s = 0; 2989 } 2990 2991 vp = (struct vnode *)fp->f_data; 2992 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2993 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 2994 2995 if (error == 0 && bioops.io_fsync != NULL && 2996 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2997 (*bioops.io_fsync)(vp, nflags); 2998 2999 VOP_UNLOCK(vp, 0); 3000 FILE_UNUSE(fp, l); 3001 return (error); 3002 } 3003 3004 /* 3005 * Sync the data of an open file. 3006 */ 3007 /* ARGSUSED */ 3008 int 3009 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3010 { 3011 struct sys_fdatasync_args /* { 3012 syscallarg(int) fd; 3013 } */ *uap = v; 3014 struct proc *p = l->l_proc; 3015 struct vnode *vp; 3016 struct file *fp; 3017 int error; 3018 3019 /* getvnode() will use the descriptor for us */ 3020 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3021 return (error); 3022 if ((fp->f_flag & FWRITE) == 0) { 3023 FILE_UNUSE(fp, l); 3024 return (EBADF); 3025 } 3026 vp = (struct vnode *)fp->f_data; 3027 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3028 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3029 VOP_UNLOCK(vp, 0); 3030 FILE_UNUSE(fp, l); 3031 return (error); 3032 } 3033 3034 /* 3035 * Rename files, (standard) BSD semantics frontend. 3036 */ 3037 /* ARGSUSED */ 3038 int 3039 sys_rename(struct lwp *l, void *v, register_t *retval) 3040 { 3041 struct sys_rename_args /* { 3042 syscallarg(const char *) from; 3043 syscallarg(const char *) to; 3044 } */ *uap = v; 3045 3046 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3047 } 3048 3049 /* 3050 * Rename files, POSIX semantics frontend. 3051 */ 3052 /* ARGSUSED */ 3053 int 3054 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3055 { 3056 struct sys___posix_rename_args /* { 3057 syscallarg(const char *) from; 3058 syscallarg(const char *) to; 3059 } */ *uap = v; 3060 3061 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3062 } 3063 3064 /* 3065 * Rename files. Source and destination must either both be directories, 3066 * or both not be directories. If target is a directory, it must be empty. 3067 * If `from' and `to' refer to the same object, the value of the `retain' 3068 * argument is used to determine whether `from' will be 3069 * 3070 * (retain == 0) deleted unless `from' and `to' refer to the same 3071 * object in the file system's name space (BSD). 3072 * (retain == 1) always retained (POSIX). 3073 */ 3074 static int 3075 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3076 { 3077 struct mount *mp = NULL; 3078 struct vnode *tvp, *fvp, *tdvp; 3079 struct nameidata fromnd, tond; 3080 struct proc *p; 3081 int error; 3082 3083 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3084 from, l); 3085 if ((error = namei(&fromnd)) != 0) 3086 return (error); 3087 fvp = fromnd.ni_vp; 3088 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3089 if (error != 0) { 3090 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3091 vrele(fromnd.ni_dvp); 3092 vrele(fvp); 3093 if (fromnd.ni_startdir) 3094 vrele(fromnd.ni_startdir); 3095 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3096 return (error); 3097 } 3098 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3099 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3100 if ((error = namei(&tond)) != 0) { 3101 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3102 vrele(fromnd.ni_dvp); 3103 vrele(fvp); 3104 goto out1; 3105 } 3106 tdvp = tond.ni_dvp; 3107 tvp = tond.ni_vp; 3108 3109 if (tvp != NULL) { 3110 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3111 error = ENOTDIR; 3112 goto out; 3113 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3114 error = EISDIR; 3115 goto out; 3116 } 3117 } 3118 3119 if (fvp == tdvp) 3120 error = EINVAL; 3121 3122 /* 3123 * Source and destination refer to the same object. 3124 */ 3125 if (fvp == tvp) { 3126 if (retain) 3127 error = -1; 3128 else if (fromnd.ni_dvp == tdvp && 3129 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3130 !memcmp(fromnd.ni_cnd.cn_nameptr, 3131 tond.ni_cnd.cn_nameptr, 3132 fromnd.ni_cnd.cn_namelen)) 3133 error = -1; 3134 } 3135 3136 #ifdef VERIFIED_EXEC 3137 if (!error) 3138 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l); 3139 #endif /* VERIFIED_EXEC */ 3140 3141 out: 3142 p = l->l_proc; 3143 if (!error) { 3144 VOP_LEASE(tdvp, l, p->p_cred, LEASE_WRITE); 3145 if (fromnd.ni_dvp != tdvp) 3146 VOP_LEASE(fromnd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3147 if (tvp) { 3148 VOP_LEASE(tvp, l, p->p_cred, LEASE_WRITE); 3149 } 3150 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3151 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3152 } else { 3153 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3154 if (tdvp == tvp) 3155 vrele(tdvp); 3156 else 3157 vput(tdvp); 3158 if (tvp) 3159 vput(tvp); 3160 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3161 vrele(fromnd.ni_dvp); 3162 vrele(fvp); 3163 } 3164 vrele(tond.ni_startdir); 3165 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3166 out1: 3167 vn_finished_write(mp, 0); 3168 if (fromnd.ni_startdir) 3169 vrele(fromnd.ni_startdir); 3170 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3171 return (error == -1 ? 0 : error); 3172 } 3173 3174 /* 3175 * Make a directory file. 3176 */ 3177 /* ARGSUSED */ 3178 int 3179 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3180 { 3181 struct sys_mkdir_args /* { 3182 syscallarg(const char *) path; 3183 syscallarg(int) mode; 3184 } */ *uap = v; 3185 struct proc *p = l->l_proc; 3186 struct mount *mp; 3187 struct vnode *vp; 3188 struct vattr vattr; 3189 int error; 3190 struct nameidata nd; 3191 3192 restart: 3193 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3194 SCARG(uap, path), l); 3195 if ((error = namei(&nd)) != 0) 3196 return (error); 3197 vp = nd.ni_vp; 3198 if (vp != NULL) { 3199 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3200 if (nd.ni_dvp == vp) 3201 vrele(nd.ni_dvp); 3202 else 3203 vput(nd.ni_dvp); 3204 vrele(vp); 3205 return (EEXIST); 3206 } 3207 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3208 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3209 if (nd.ni_dvp == vp) 3210 vrele(nd.ni_dvp); 3211 else 3212 vput(nd.ni_dvp); 3213 if ((error = vn_start_write(NULL, &mp, 3214 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3215 return (error); 3216 goto restart; 3217 } 3218 VATTR_NULL(&vattr); 3219 vattr.va_type = VDIR; 3220 vattr.va_mode = 3221 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3222 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3223 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3224 if (!error) 3225 vput(nd.ni_vp); 3226 vn_finished_write(mp, 0); 3227 return (error); 3228 } 3229 3230 /* 3231 * Remove a directory file. 3232 */ 3233 /* ARGSUSED */ 3234 int 3235 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3236 { 3237 struct sys_rmdir_args /* { 3238 syscallarg(const char *) path; 3239 } */ *uap = v; 3240 struct proc *p = l->l_proc; 3241 struct mount *mp; 3242 struct vnode *vp; 3243 int error; 3244 struct nameidata nd; 3245 3246 restart: 3247 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3248 SCARG(uap, path), l); 3249 if ((error = namei(&nd)) != 0) 3250 return (error); 3251 vp = nd.ni_vp; 3252 if (vp->v_type != VDIR) { 3253 error = ENOTDIR; 3254 goto out; 3255 } 3256 /* 3257 * No rmdir "." please. 3258 */ 3259 if (nd.ni_dvp == vp) { 3260 error = EINVAL; 3261 goto out; 3262 } 3263 /* 3264 * The root of a mounted filesystem cannot be deleted. 3265 */ 3266 if (vp->v_flag & VROOT) { 3267 error = EBUSY; 3268 goto out; 3269 } 3270 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3271 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3272 if (nd.ni_dvp == vp) 3273 vrele(nd.ni_dvp); 3274 else 3275 vput(nd.ni_dvp); 3276 vput(vp); 3277 if ((error = vn_start_write(NULL, &mp, 3278 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3279 return (error); 3280 goto restart; 3281 } 3282 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3283 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3284 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3285 vn_finished_write(mp, 0); 3286 return (error); 3287 3288 out: 3289 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3290 if (nd.ni_dvp == vp) 3291 vrele(nd.ni_dvp); 3292 else 3293 vput(nd.ni_dvp); 3294 vput(vp); 3295 return (error); 3296 } 3297 3298 /* 3299 * Read a block of directory entries in a file system independent format. 3300 */ 3301 int 3302 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3303 { 3304 struct sys___getdents30_args /* { 3305 syscallarg(int) fd; 3306 syscallarg(char *) buf; 3307 syscallarg(size_t) count; 3308 } */ *uap = v; 3309 struct proc *p = l->l_proc; 3310 struct file *fp; 3311 int error, done; 3312 3313 /* getvnode() will use the descriptor for us */ 3314 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3315 return (error); 3316 if ((fp->f_flag & FREAD) == 0) { 3317 error = EBADF; 3318 goto out; 3319 } 3320 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3321 SCARG(uap, count), &done, l, 0, 0); 3322 #ifdef KTRACE 3323 if (!error && KTRPOINT(p, KTR_GENIO)) { 3324 struct iovec iov; 3325 iov.iov_base = SCARG(uap, buf); 3326 iov.iov_len = done; 3327 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3328 } 3329 #endif 3330 *retval = done; 3331 out: 3332 FILE_UNUSE(fp, l); 3333 return (error); 3334 } 3335 3336 /* 3337 * Set the mode mask for creation of filesystem nodes. 3338 */ 3339 int 3340 sys_umask(struct lwp *l, void *v, register_t *retval) 3341 { 3342 struct sys_umask_args /* { 3343 syscallarg(mode_t) newmask; 3344 } */ *uap = v; 3345 struct proc *p = l->l_proc; 3346 struct cwdinfo *cwdi; 3347 3348 cwdi = p->p_cwdi; 3349 *retval = cwdi->cwdi_cmask; 3350 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3351 return (0); 3352 } 3353 3354 /* 3355 * Void all references to file by ripping underlying filesystem 3356 * away from vnode. 3357 */ 3358 /* ARGSUSED */ 3359 int 3360 sys_revoke(struct lwp *l, void *v, register_t *retval) 3361 { 3362 struct sys_revoke_args /* { 3363 syscallarg(const char *) path; 3364 } */ *uap = v; 3365 struct proc *p = l->l_proc; 3366 struct mount *mp; 3367 struct vnode *vp; 3368 struct vattr vattr; 3369 int error; 3370 struct nameidata nd; 3371 3372 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3373 if ((error = namei(&nd)) != 0) 3374 return (error); 3375 vp = nd.ni_vp; 3376 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 3377 goto out; 3378 if (kauth_cred_geteuid(p->p_cred) != vattr.va_uid && 3379 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 3380 &p->p_acflag)) != 0) 3381 goto out; 3382 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3383 goto out; 3384 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3385 VOP_REVOKE(vp, REVOKEALL); 3386 vn_finished_write(mp, 0); 3387 out: 3388 vrele(vp); 3389 return (error); 3390 } 3391 3392 /* 3393 * Convert a user file descriptor to a kernel file entry. 3394 */ 3395 int 3396 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3397 { 3398 struct vnode *vp; 3399 struct file *fp; 3400 3401 if ((fp = fd_getfile(fdp, fd)) == NULL) 3402 return (EBADF); 3403 3404 FILE_USE(fp); 3405 3406 if (fp->f_type != DTYPE_VNODE) { 3407 FILE_UNUSE(fp, NULL); 3408 return (EINVAL); 3409 } 3410 3411 vp = (struct vnode *)fp->f_data; 3412 if (vp->v_type == VBAD) { 3413 FILE_UNUSE(fp, NULL); 3414 return (EBADF); 3415 } 3416 3417 *fpp = fp; 3418 return (0); 3419 } 3420
Indexes created Fri Oct 03 02:09:57 GMT 2025