vfs_syscalls.c revision 1.238.6.2
1 /* $NetBSD: vfs_syscalls.c,v 1.238.6.2 2006/05/24 15:50:42 tron 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 1.238.6.2 2006/05/24 15:50:42 tron 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 #include <sys/kauth.h> 71 72 #include <miscfs/genfs/genfs.h> 73 #include <miscfs/syncfs/syncfs.h> 74 75 #ifdef COMPAT_30 76 #include "opt_nfsserver.h" 77 #include <nfs/rpcv2.h> 78 #include <nfs/nfsproto.h> 79 #include <nfs/nfs.h> 80 #include <nfs/nfs_var.h> 81 #endif 82 83 #if NFSS > 0 84 #include <dev/fssvar.h> 85 #endif 86 87 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount struct"); 88 89 static int change_dir(struct nameidata *, struct lwp *); 90 static int change_flags(struct vnode *, u_long, struct lwp *); 91 static int change_mode(struct vnode *, int, struct lwp *l); 92 static int change_owner(struct vnode *, uid_t, gid_t, struct lwp *, int); 93 static int change_utimes(struct vnode *vp, const struct timeval *, 94 struct lwp *l); 95 static int rename_files(const char *, const char *, struct lwp *, int); 96 97 void checkdirs(struct vnode *); 98 99 int dovfsusermount = 0; 100 101 /* 102 * Virtual File System System Calls 103 */ 104 105 /* 106 * Mount a file system. 107 */ 108 109 #if defined(COMPAT_09) || defined(COMPAT_43) 110 /* 111 * This table is used to maintain compatibility with 4.3BSD 112 * and NetBSD 0.9 mount syscalls. Note, the order is important! 113 * 114 * Do not modify this table. It should only contain filesystems 115 * supported by NetBSD 0.9 and 4.3BSD. 116 */ 117 const char * const mountcompatnames[] = { 118 NULL, /* 0 = MOUNT_NONE */ 119 MOUNT_FFS, /* 1 = MOUNT_UFS */ 120 MOUNT_NFS, /* 2 */ 121 MOUNT_MFS, /* 3 */ 122 MOUNT_MSDOS, /* 4 */ 123 MOUNT_CD9660, /* 5 = MOUNT_ISOFS */ 124 MOUNT_FDESC, /* 6 */ 125 MOUNT_KERNFS, /* 7 */ 126 NULL, /* 8 = MOUNT_DEVFS */ 127 MOUNT_AFS, /* 9 */ 128 }; 129 const int nmountcompatnames = sizeof(mountcompatnames) / 130 sizeof(mountcompatnames[0]); 131 #endif /* COMPAT_09 || COMPAT_43 */ 132 133 /* ARGSUSED */ 134 int 135 sys_mount(struct lwp *l, void *v, register_t *retval) 136 { 137 struct sys_mount_args /* { 138 syscallarg(const char *) type; 139 syscallarg(const char *) path; 140 syscallarg(int) flags; 141 syscallarg(void *) data; 142 } */ *uap = v; 143 struct proc *p = l->l_proc; 144 struct vnode *vp; 145 struct mount *mp; 146 int error, flag = 0; 147 char fstypename[MFSNAMELEN]; 148 struct vattr va; 149 struct nameidata nd; 150 struct vfsops *vfs; 151 152 /* 153 * if MNT_GETARGS is specified, it should be only flag. 154 */ 155 156 if ((SCARG(uap, flags) & MNT_GETARGS) != 0 && 157 (SCARG(uap, flags) & ~MNT_GETARGS) != 0) { 158 return EINVAL; 159 } 160 161 if (dovfsusermount == 0 && (SCARG(uap, flags) & MNT_GETARGS) == 0 && 162 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 163 &p->p_acflag))) 164 return (error); 165 /* 166 * Get vnode to be covered 167 */ 168 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, 169 SCARG(uap, path), l); 170 if ((error = namei(&nd)) != 0) 171 return (error); 172 vp = nd.ni_vp; 173 /* 174 * A lookup in VFS_MOUNT might result in an attempt to 175 * lock this vnode again, so make the lock recursive. 176 */ 177 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY | LK_SETRECURSE); 178 if (SCARG(uap, flags) & (MNT_UPDATE | MNT_GETARGS)) { 179 if ((vp->v_flag & VROOT) == 0) { 180 vput(vp); 181 return (EINVAL); 182 } 183 mp = vp->v_mount; 184 flag = mp->mnt_flag; 185 vfs = mp->mnt_op; 186 /* 187 * We only allow the filesystem to be reloaded if it 188 * is currently mounted read-only. 189 */ 190 if ((SCARG(uap, flags) & MNT_RELOAD) && 191 ((mp->mnt_flag & MNT_RDONLY) == 0)) { 192 vput(vp); 193 return (EOPNOTSUPP); /* Needs translation */ 194 } 195 /* 196 * In "highly secure" mode, don't let the caller do anything 197 * but downgrade a filesystem from read-write to read-only. 198 * (see also below; MNT_UPDATE or MNT_GETARGS is required.) 199 */ 200 if (securelevel >= 2 && 201 SCARG(uap, flags) != MNT_GETARGS && 202 SCARG(uap, flags) != 203 (mp->mnt_flag | MNT_RDONLY | 204 MNT_RELOAD | MNT_FORCE | MNT_UPDATE)) { 205 vput(vp); 206 return (EPERM); 207 } 208 mp->mnt_flag |= SCARG(uap, flags) & 209 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 210 /* 211 * Only root, or the user that did the original mount is 212 * permitted to update it. 213 */ 214 if ((mp->mnt_flag & MNT_GETARGS) == 0 && 215 mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred) && 216 (error = kauth_authorize_generic(p->p_cred, 217 KAUTH_GENERIC_ISSUSER, 218 &p->p_acflag)) != 0) { 219 vput(vp); 220 return (error); 221 } 222 /* 223 * Do not allow NFS export by non-root users. For non-root 224 * users, silently enforce MNT_NOSUID and MNT_NODEV, and 225 * MNT_NOEXEC if mount point is already MNT_NOEXEC. 226 */ 227 if (kauth_cred_geteuid(p->p_cred) != 0) { 228 if (SCARG(uap, flags) & MNT_EXPORTED) { 229 vput(vp); 230 return (EPERM); 231 } 232 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 233 if (flag & MNT_NOEXEC) 234 SCARG(uap, flags) |= MNT_NOEXEC; 235 } 236 if (vfs_busy(mp, LK_NOWAIT, 0)) { 237 vput(vp); 238 return (EPERM); 239 } 240 goto update; 241 } else { 242 if (securelevel >= 2) { 243 vput(vp); 244 return (EPERM); 245 } 246 } 247 /* 248 * If the user is not root, ensure that they own the directory 249 * onto which we are attempting to mount. 250 */ 251 if ((error = VOP_GETATTR(vp, &va, p->p_cred, l)) != 0 || 252 (va.va_uid != kauth_cred_geteuid(p->p_cred) && 253 (error = kauth_authorize_generic(p->p_cred, 254 KAUTH_GENERIC_ISSUSER, 255 &p->p_acflag)) != 0)) { 256 vput(vp); 257 return (error); 258 } 259 /* 260 * Do not allow NFS export by non-root users. For non-root users, 261 * silently enforce MNT_NOSUID and MNT_NODEV, and MNT_NOEXEC if the 262 * mount point is already MNT_NOEXEC. 263 */ 264 if (kauth_cred_geteuid(p->p_cred) != 0) { 265 if (SCARG(uap, flags) & MNT_EXPORTED) { 266 vput(vp); 267 return (EPERM); 268 } 269 SCARG(uap, flags) |= MNT_NOSUID | MNT_NODEV; 270 if (vp->v_mount->mnt_flag & MNT_NOEXEC) 271 SCARG(uap, flags) |= MNT_NOEXEC; 272 } 273 if ((error = vinvalbuf(vp, V_SAVE, p->p_cred, l, 0, 0)) != 0) { 274 vput(vp); 275 return (error); 276 } 277 if (vp->v_type != VDIR) { 278 vput(vp); 279 return (ENOTDIR); 280 } 281 error = copyinstr(SCARG(uap, type), fstypename, MFSNAMELEN, NULL); 282 if (error) { 283 #if defined(COMPAT_09) || defined(COMPAT_43) 284 /* 285 * Historically, filesystem types were identified by numbers. 286 * If we get an integer for the filesystem type instead of a 287 * string, we check to see if it matches one of the historic 288 * filesystem types. 289 */ 290 u_long fsindex = (u_long)SCARG(uap, type); 291 if (fsindex >= nmountcompatnames || 292 mountcompatnames[fsindex] == NULL) { 293 vput(vp); 294 return (ENODEV); 295 } 296 strncpy(fstypename, mountcompatnames[fsindex], MFSNAMELEN); 297 #else 298 vput(vp); 299 return (error); 300 #endif 301 } 302 #ifdef COMPAT_10 303 /* Accept `ufs' as an alias for `ffs'. */ 304 if (!strncmp(fstypename, "ufs", MFSNAMELEN)) 305 strncpy(fstypename, "ffs", MFSNAMELEN); 306 #endif 307 if ((vfs = vfs_getopsbyname(fstypename)) == NULL) { 308 vput(vp); 309 return (ENODEV); 310 } 311 if (vp->v_mountedhere != NULL) { 312 vput(vp); 313 return (EBUSY); 314 } 315 316 /* 317 * Allocate and initialize the file system. 318 */ 319 mp = (struct mount *)malloc((u_long)sizeof(struct mount), 320 M_MOUNT, M_WAITOK); 321 memset((char *)mp, 0, (u_long)sizeof(struct mount)); 322 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0); 323 simple_lock_init(&mp->mnt_slock); 324 (void)vfs_busy(mp, LK_NOWAIT, 0); 325 mp->mnt_op = vfs; 326 vfs->vfs_refcount++; 327 mp->mnt_vnodecovered = vp; 328 mp->mnt_stat.f_owner = kauth_cred_geteuid(p->p_cred); 329 mp->mnt_unmounter = NULL; 330 mp->mnt_leaf = mp; 331 332 /* 333 * The underlying file system may refuse the mount for 334 * various reasons. Allow the user to force it to happen. 335 */ 336 mp->mnt_flag |= SCARG(uap, flags) & MNT_FORCE; 337 update: 338 if ((SCARG(uap, flags) & MNT_GETARGS) == 0) { 339 /* 340 * Set the mount level flags. 341 */ 342 if (SCARG(uap, flags) & MNT_RDONLY) 343 mp->mnt_flag |= MNT_RDONLY; 344 else if (mp->mnt_flag & MNT_RDONLY) 345 mp->mnt_iflag |= IMNT_WANTRDWR; 346 mp->mnt_flag &= 347 ~(MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 348 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 349 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP); 350 mp->mnt_flag |= SCARG(uap, flags) & 351 (MNT_NOSUID | MNT_NOEXEC | MNT_NODEV | 352 MNT_SYNCHRONOUS | MNT_UNION | MNT_ASYNC | MNT_NOCOREDUMP | 353 MNT_NOATIME | MNT_NODEVMTIME | MNT_SYMPERM | MNT_SOFTDEP | 354 MNT_IGNORE); 355 } 356 /* 357 * Mount the filesystem. 358 */ 359 error = VFS_MOUNT(mp, SCARG(uap, path), SCARG(uap, data), &nd, l); 360 if (mp->mnt_flag & (MNT_UPDATE | MNT_GETARGS)) { 361 #if defined(COMPAT_30) && defined(NFSSERVER) 362 if (mp->mnt_flag & MNT_UPDATE && error != 0) { 363 int error2; 364 365 /* Update failed; let's try and see if it was an 366 * export request. */ 367 error2 = nfs_update_exports_30(mp, SCARG(uap, path), 368 SCARG(uap, data), l); 369 370 /* Only update error code if the export request was 371 * understood but some problem occurred while 372 * processing it. */ 373 if (error2 != EJUSTRETURN) 374 error = error2; 375 } 376 #endif 377 if (mp->mnt_iflag & IMNT_WANTRDWR) 378 mp->mnt_flag &= ~MNT_RDONLY; 379 if (error) 380 mp->mnt_flag = flag; 381 mp->mnt_flag &=~ 382 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 383 mp->mnt_iflag &=~ IMNT_WANTRDWR; 384 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) { 385 if (mp->mnt_syncer == NULL) 386 error = vfs_allocate_syncvnode(mp); 387 } else { 388 if (mp->mnt_syncer != NULL) 389 vfs_deallocate_syncvnode(mp); 390 } 391 vfs_unbusy(mp); 392 VOP_UNLOCK(vp, 0); 393 vrele(vp); 394 return (error); 395 } 396 /* 397 * Put the new filesystem on the mount list after root. 398 */ 399 cache_purge(vp); 400 if (!error) { 401 mp->mnt_flag &=~ 402 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_GETARGS); 403 mp->mnt_iflag &=~ IMNT_WANTRDWR; 404 vp->v_mountedhere = mp; 405 simple_lock(&mountlist_slock); 406 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 407 simple_unlock(&mountlist_slock); 408 checkdirs(vp); 409 VOP_UNLOCK(vp, 0); 410 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 411 error = vfs_allocate_syncvnode(mp); 412 vfs_unbusy(mp); 413 (void) VFS_STATVFS(mp, &mp->mnt_stat, l); 414 if ((error = VFS_START(mp, 0, l))) 415 vrele(vp); 416 } else { 417 vp->v_mountedhere = (struct mount *)0; 418 vfs->vfs_refcount--; 419 vfs_unbusy(mp); 420 free(mp, M_MOUNT); 421 vput(vp); 422 } 423 return (error); 424 } 425 426 /* 427 * Scan all active processes to see if any of them have a current 428 * or root directory onto which the new filesystem has just been 429 * mounted. If so, replace them with the new mount point. 430 */ 431 void 432 checkdirs(struct vnode *olddp) 433 { 434 struct cwdinfo *cwdi; 435 struct vnode *newdp; 436 struct proc *p; 437 438 if (olddp->v_usecount == 1) 439 return; 440 if (VFS_ROOT(olddp->v_mountedhere, &newdp)) 441 panic("mount: lost mount"); 442 proclist_lock_read(); 443 PROCLIST_FOREACH(p, &allproc) { 444 cwdi = p->p_cwdi; 445 if (!cwdi) 446 continue; 447 if (cwdi->cwdi_cdir == olddp) { 448 vrele(cwdi->cwdi_cdir); 449 VREF(newdp); 450 cwdi->cwdi_cdir = newdp; 451 } 452 if (cwdi->cwdi_rdir == olddp) { 453 vrele(cwdi->cwdi_rdir); 454 VREF(newdp); 455 cwdi->cwdi_rdir = newdp; 456 } 457 } 458 proclist_unlock_read(); 459 if (rootvnode == olddp) { 460 vrele(rootvnode); 461 VREF(newdp); 462 rootvnode = newdp; 463 } 464 vput(newdp); 465 } 466 467 /* 468 * Unmount a file system. 469 * 470 * Note: unmount takes a path to the vnode mounted on as argument, 471 * not special file (as before). 472 */ 473 /* ARGSUSED */ 474 int 475 sys_unmount(struct lwp *l, void *v, register_t *retval) 476 { 477 struct sys_unmount_args /* { 478 syscallarg(const char *) path; 479 syscallarg(int) flags; 480 } */ *uap = v; 481 struct proc *p = l->l_proc; 482 struct vnode *vp; 483 struct mount *mp; 484 int error; 485 struct nameidata nd; 486 487 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 488 SCARG(uap, path), l); 489 if ((error = namei(&nd)) != 0) 490 return (error); 491 vp = nd.ni_vp; 492 mp = vp->v_mount; 493 494 /* 495 * Only root, or the user that did the original mount is 496 * permitted to unmount this filesystem. 497 */ 498 if ((mp->mnt_stat.f_owner != kauth_cred_geteuid(p->p_cred)) && 499 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 500 &p->p_acflag)) != 0) { 501 vput(vp); 502 return (error); 503 } 504 505 /* 506 * Don't allow unmounting the root file system. 507 */ 508 if (mp->mnt_flag & MNT_ROOTFS) { 509 vput(vp); 510 return (EINVAL); 511 } 512 513 /* 514 * Must be the root of the filesystem 515 */ 516 if ((vp->v_flag & VROOT) == 0) { 517 vput(vp); 518 return (EINVAL); 519 } 520 vput(vp); 521 522 /* 523 * XXX Freeze syncer. Must do this before locking the 524 * mount point. See dounmount() for details. 525 */ 526 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 527 528 if (vfs_busy(mp, 0, 0)) { 529 lockmgr(&syncer_lock, LK_RELEASE, NULL); 530 return (EBUSY); 531 } 532 533 return (dounmount(mp, SCARG(uap, flags), l)); 534 } 535 536 /* 537 * Do the actual file system unmount. File system is assumed to have been 538 * marked busy by the caller. 539 */ 540 int 541 dounmount(struct mount *mp, int flags, struct lwp *l) 542 { 543 struct vnode *coveredvp; 544 int error; 545 int async; 546 int used_syncer; 547 548 simple_lock(&mountlist_slock); 549 vfs_unbusy(mp); 550 used_syncer = (mp->mnt_syncer != NULL); 551 552 /* 553 * XXX Syncer must be frozen when we get here. This should really 554 * be done on a per-mountpoint basis, but especially the softdep 555 * code possibly called from the syncer doesn't exactly work on a 556 * per-mountpoint basis, so the softdep code would become a maze 557 * of vfs_busy() calls. 558 * 559 * The caller of dounmount() must acquire syncer_lock because 560 * the syncer itself acquires locks in syncer_lock -> vfs_busy 561 * order, and we must preserve that order to avoid deadlock. 562 * 563 * So, if the file system did not use the syncer, now is 564 * the time to release the syncer_lock. 565 */ 566 if (used_syncer == 0) 567 lockmgr(&syncer_lock, LK_RELEASE, NULL); 568 569 mp->mnt_iflag |= IMNT_UNMOUNT; 570 mp->mnt_unmounter = l; 571 lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK, &mountlist_slock); 572 vn_start_write(NULL, &mp, V_WAIT); 573 574 async = mp->mnt_flag & MNT_ASYNC; 575 mp->mnt_flag &= ~MNT_ASYNC; 576 cache_purgevfs(mp); /* remove cache entries for this file sys */ 577 if (mp->mnt_syncer != NULL) 578 vfs_deallocate_syncvnode(mp); 579 error = 0; 580 if ((mp->mnt_flag & MNT_RDONLY) == 0) { 581 #if NFSS > 0 582 error = fss_umount_hook(mp, (flags & MNT_FORCE)); 583 #endif 584 if (error == 0) 585 error = VFS_SYNC(mp, MNT_WAIT, l->l_proc->p_cred, l); 586 } 587 if (error == 0 || (flags & MNT_FORCE)) 588 error = VFS_UNMOUNT(mp, flags, l); 589 vn_finished_write(mp, 0); 590 simple_lock(&mountlist_slock); 591 if (error) { 592 if ((mp->mnt_flag & (MNT_RDONLY | MNT_ASYNC)) == 0) 593 (void) vfs_allocate_syncvnode(mp); 594 mp->mnt_iflag &= ~IMNT_UNMOUNT; 595 mp->mnt_unmounter = NULL; 596 mp->mnt_flag |= async; 597 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK | LK_REENABLE, 598 &mountlist_slock); 599 if (used_syncer) 600 lockmgr(&syncer_lock, LK_RELEASE, NULL); 601 simple_lock(&mp->mnt_slock); 602 while (mp->mnt_wcnt > 0) { 603 wakeup(mp); 604 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt1", 605 0, &mp->mnt_slock); 606 } 607 simple_unlock(&mp->mnt_slock); 608 return (error); 609 } 610 CIRCLEQ_REMOVE(&mountlist, mp, mnt_list); 611 if ((coveredvp = mp->mnt_vnodecovered) != NULLVP) { 612 coveredvp->v_mountedhere = NULL; 613 vrele(coveredvp); 614 } 615 mp->mnt_op->vfs_refcount--; 616 if (LIST_FIRST(&mp->mnt_vnodelist) != NULL) 617 panic("unmount: dangling vnode"); 618 mp->mnt_iflag |= IMNT_GONE; 619 lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK, &mountlist_slock); 620 if (used_syncer) 621 lockmgr(&syncer_lock, LK_RELEASE, NULL); 622 simple_lock(&mp->mnt_slock); 623 while (mp->mnt_wcnt > 0) { 624 wakeup(mp); 625 ltsleep(&mp->mnt_wcnt, PVFS, "mntwcnt2", 0, &mp->mnt_slock); 626 } 627 simple_unlock(&mp->mnt_slock); 628 vfs_hooks_unmount(mp); 629 free(mp, M_MOUNT); 630 return (0); 631 } 632 633 /* 634 * Sync each mounted filesystem. 635 */ 636 #ifdef DEBUG 637 int syncprt = 0; 638 struct ctldebug debug0 = { "syncprt", &syncprt }; 639 #endif 640 641 /* ARGSUSED */ 642 int 643 sys_sync(struct lwp *l, void *v, register_t *retval) 644 { 645 struct mount *mp, *nmp; 646 int asyncflag; 647 struct proc *p = l == NULL ? &proc0 : l->l_proc; 648 649 simple_lock(&mountlist_slock); 650 for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) { 651 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 652 nmp = mp->mnt_list.cqe_prev; 653 continue; 654 } 655 if ((mp->mnt_flag & MNT_RDONLY) == 0 && 656 vn_start_write(NULL, &mp, V_NOWAIT) == 0) { 657 asyncflag = mp->mnt_flag & MNT_ASYNC; 658 mp->mnt_flag &= ~MNT_ASYNC; 659 VFS_SYNC(mp, MNT_NOWAIT, p->p_cred, l); 660 if (asyncflag) 661 mp->mnt_flag |= MNT_ASYNC; 662 vn_finished_write(mp, 0); 663 } 664 simple_lock(&mountlist_slock); 665 nmp = mp->mnt_list.cqe_prev; 666 vfs_unbusy(mp); 667 668 } 669 simple_unlock(&mountlist_slock); 670 #ifdef DEBUG 671 if (syncprt) 672 vfs_bufstats(); 673 #endif /* DEBUG */ 674 return (0); 675 } 676 677 /* 678 * Change filesystem quotas. 679 */ 680 /* ARGSUSED */ 681 int 682 sys_quotactl(struct lwp *l, void *v, register_t *retval) 683 { 684 struct sys_quotactl_args /* { 685 syscallarg(const char *) path; 686 syscallarg(int) cmd; 687 syscallarg(int) uid; 688 syscallarg(caddr_t) arg; 689 } */ *uap = v; 690 struct mount *mp; 691 int error; 692 struct nameidata nd; 693 694 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 695 if ((error = namei(&nd)) != 0) 696 return (error); 697 error = vn_start_write(nd.ni_vp, &mp, V_WAIT | V_PCATCH); 698 vrele(nd.ni_vp); 699 if (error) 700 return (error); 701 error = VFS_QUOTACTL(mp, SCARG(uap, cmd), SCARG(uap, uid), 702 SCARG(uap, arg), l); 703 vn_finished_write(mp, 0); 704 return (error); 705 } 706 707 int 708 dostatvfs(struct mount *mp, struct statvfs *sp, struct lwp *l, int flags, 709 int root) 710 { 711 struct cwdinfo *cwdi = l->l_proc->p_cwdi; 712 int error = 0; 713 714 /* 715 * If MNT_NOWAIT or MNT_LAZY is specified, do not 716 * refresh the fsstat cache. MNT_WAIT or MNT_LAZY 717 * overrides MNT_NOWAIT. 718 */ 719 if (flags == MNT_NOWAIT || flags == MNT_LAZY || 720 (flags != MNT_WAIT && flags != 0)) { 721 memcpy(sp, &mp->mnt_stat, sizeof(*sp)); 722 goto done; 723 } 724 725 /* Get the filesystem stats now */ 726 memset(sp, 0, sizeof(*sp)); 727 if ((error = VFS_STATVFS(mp, sp, l)) != 0) { 728 return error; 729 } 730 731 if (cwdi->cwdi_rdir == NULL) 732 (void)memcpy(&mp->mnt_stat, sp, sizeof(mp->mnt_stat)); 733 done: 734 if (cwdi->cwdi_rdir != NULL) { 735 size_t len; 736 char *bp; 737 char *path = PNBUF_GET(); 738 if (!path) 739 return ENOMEM; 740 741 bp = path + MAXPATHLEN; 742 *--bp = '\0'; 743 error = getcwd_common(cwdi->cwdi_rdir, rootvnode, &bp, path, 744 MAXPATHLEN / 2, 0, l); 745 if (error) { 746 PNBUF_PUT(path); 747 return error; 748 } 749 len = strlen(bp); 750 /* 751 * for mount points that are below our root, we can see 752 * them, so we fix up the pathname and return them. The 753 * rest we cannot see, so we don't allow viewing the 754 * data. 755 */ 756 if (strncmp(bp, sp->f_mntonname, len) == 0) { 757 strlcpy(sp->f_mntonname, &sp->f_mntonname[len], 758 sizeof(sp->f_mntonname)); 759 if (sp->f_mntonname[0] == '\0') 760 (void)strlcpy(sp->f_mntonname, "/", 761 sizeof(sp->f_mntonname)); 762 } else { 763 if (root) 764 (void)strlcpy(sp->f_mntonname, "/", 765 sizeof(sp->f_mntonname)); 766 else 767 error = EPERM; 768 } 769 PNBUF_PUT(path); 770 } 771 sp->f_flag = mp->mnt_flag & MNT_VISFLAGMASK; 772 return error; 773 } 774 775 /* 776 * Get filesystem statistics. 777 */ 778 /* ARGSUSED */ 779 int 780 sys_statvfs1(struct lwp *l, void *v, register_t *retval) 781 { 782 struct sys_statvfs1_args /* { 783 syscallarg(const char *) path; 784 syscallarg(struct statvfs *) buf; 785 syscallarg(int) flags; 786 } */ *uap = v; 787 struct mount *mp; 788 struct statvfs *sb; 789 int error; 790 struct nameidata nd; 791 792 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 793 if ((error = namei(&nd)) != 0) 794 return error; 795 mp = nd.ni_vp->v_mount; 796 vrele(nd.ni_vp); 797 sb = STATVFSBUF_GET(); 798 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1); 799 if (error == 0) { 800 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 801 } 802 STATVFSBUF_PUT(sb); 803 return error; 804 } 805 806 /* 807 * Get filesystem statistics. 808 */ 809 /* ARGSUSED */ 810 int 811 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 812 { 813 struct sys_fstatvfs1_args /* { 814 syscallarg(int) fd; 815 syscallarg(struct statvfs *) buf; 816 syscallarg(int) flags; 817 } */ *uap = v; 818 struct proc *p = l->l_proc; 819 struct file *fp; 820 struct mount *mp; 821 struct statvfs *sb; 822 int error; 823 824 /* getvnode() will use the descriptor for us */ 825 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 826 return (error); 827 mp = ((struct vnode *)fp->f_data)->v_mount; 828 sb = STATVFSBUF_GET(); 829 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) 830 goto out; 831 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 832 out: 833 FILE_UNUSE(fp, l); 834 STATVFSBUF_PUT(sb); 835 return error; 836 } 837 838 839 /* 840 * Get statistics on all filesystems. 841 */ 842 int 843 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 844 { 845 struct sys_getvfsstat_args /* { 846 syscallarg(struct statvfs *) buf; 847 syscallarg(size_t) bufsize; 848 syscallarg(int) flags; 849 } */ *uap = v; 850 int root = 0; 851 struct proc *p = l->l_proc; 852 struct mount *mp, *nmp; 853 struct statvfs *sb; 854 struct statvfs *sfsp; 855 size_t count, maxcount; 856 int error = 0; 857 858 sb = STATVFSBUF_GET(); 859 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 860 sfsp = SCARG(uap, buf); 861 simple_lock(&mountlist_slock); 862 count = 0; 863 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 864 mp = nmp) { 865 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 866 nmp = CIRCLEQ_NEXT(mp, mnt_list); 867 continue; 868 } 869 if (sfsp && count < maxcount) { 870 error = dostatvfs(mp, sb, l, SCARG(uap, flags), 0); 871 if (error) { 872 simple_lock(&mountlist_slock); 873 nmp = CIRCLEQ_NEXT(mp, mnt_list); 874 vfs_unbusy(mp); 875 continue; 876 } 877 error = copyout(sb, sfsp, sizeof(*sfsp)); 878 if (error) { 879 vfs_unbusy(mp); 880 goto out; 881 } 882 sfsp++; 883 root |= strcmp(sb->f_mntonname, "/") == 0; 884 } 885 count++; 886 simple_lock(&mountlist_slock); 887 nmp = CIRCLEQ_NEXT(mp, mnt_list); 888 vfs_unbusy(mp); 889 } 890 simple_unlock(&mountlist_slock); 891 if (root == 0 && p->p_cwdi->cwdi_rdir) { 892 /* 893 * fake a root entry 894 */ 895 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, sb, l, 896 SCARG(uap, flags), 1)) != 0) 897 goto out; 898 if (sfsp) 899 error = copyout(sb, sfsp, sizeof(*sfsp)); 900 count++; 901 } 902 if (sfsp && count > maxcount) 903 *retval = maxcount; 904 else 905 *retval = count; 906 out: 907 STATVFSBUF_PUT(sb); 908 return error; 909 } 910 911 /* 912 * Change current working directory to a given file descriptor. 913 */ 914 /* ARGSUSED */ 915 int 916 sys_fchdir(struct lwp *l, void *v, register_t *retval) 917 { 918 struct sys_fchdir_args /* { 919 syscallarg(int) fd; 920 } */ *uap = v; 921 struct proc *p = l->l_proc; 922 struct filedesc *fdp = p->p_fd; 923 struct cwdinfo *cwdi = p->p_cwdi; 924 struct vnode *vp, *tdp; 925 struct mount *mp; 926 struct file *fp; 927 int error; 928 929 /* getvnode() will use the descriptor for us */ 930 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 931 return (error); 932 vp = (struct vnode *)fp->f_data; 933 934 VREF(vp); 935 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 936 if (vp->v_type != VDIR) 937 error = ENOTDIR; 938 else 939 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 940 while (!error && (mp = vp->v_mountedhere) != NULL) { 941 if (vfs_busy(mp, 0, 0)) 942 continue; 943 error = VFS_ROOT(mp, &tdp); 944 vfs_unbusy(mp); 945 if (error) 946 break; 947 vput(vp); 948 vp = tdp; 949 } 950 if (error) { 951 vput(vp); 952 goto out; 953 } 954 VOP_UNLOCK(vp, 0); 955 956 /* 957 * Disallow changing to a directory not under the process's 958 * current root directory (if there is one). 959 */ 960 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 961 vrele(vp); 962 error = EPERM; /* operation not permitted */ 963 goto out; 964 } 965 966 vrele(cwdi->cwdi_cdir); 967 cwdi->cwdi_cdir = vp; 968 out: 969 FILE_UNUSE(fp, l); 970 return (error); 971 } 972 973 /* 974 * Change this process's notion of the root directory to a given file 975 * descriptor. 976 */ 977 int 978 sys_fchroot(struct lwp *l, void *v, register_t *retval) 979 { 980 struct sys_fchroot_args *uap = v; 981 struct proc *p = l->l_proc; 982 struct filedesc *fdp = p->p_fd; 983 struct cwdinfo *cwdi = p->p_cwdi; 984 struct vnode *vp; 985 struct file *fp; 986 int error; 987 988 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 989 &p->p_acflag)) != 0) 990 return error; 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 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 996 if (vp->v_type != VDIR) 997 error = ENOTDIR; 998 else 999 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 1000 VOP_UNLOCK(vp, 0); 1001 if (error) 1002 goto out; 1003 VREF(vp); 1004 1005 /* 1006 * Prevent escaping from chroot by putting the root under 1007 * the working directory. Silently chdir to / if we aren't 1008 * already there. 1009 */ 1010 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1011 /* 1012 * XXX would be more failsafe to change directory to a 1013 * deadfs node here instead 1014 */ 1015 vrele(cwdi->cwdi_cdir); 1016 VREF(vp); 1017 cwdi->cwdi_cdir = vp; 1018 } 1019 1020 if (cwdi->cwdi_rdir != NULL) 1021 vrele(cwdi->cwdi_rdir); 1022 cwdi->cwdi_rdir = vp; 1023 out: 1024 FILE_UNUSE(fp, l); 1025 return (error); 1026 } 1027 1028 /* 1029 * Change current working directory (``.''). 1030 */ 1031 /* ARGSUSED */ 1032 int 1033 sys_chdir(struct lwp *l, void *v, register_t *retval) 1034 { 1035 struct sys_chdir_args /* { 1036 syscallarg(const char *) path; 1037 } */ *uap = v; 1038 struct proc *p = l->l_proc; 1039 struct cwdinfo *cwdi = p->p_cwdi; 1040 int error; 1041 struct nameidata nd; 1042 1043 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1044 SCARG(uap, path), l); 1045 if ((error = change_dir(&nd, l)) != 0) 1046 return (error); 1047 vrele(cwdi->cwdi_cdir); 1048 cwdi->cwdi_cdir = nd.ni_vp; 1049 return (0); 1050 } 1051 1052 /* 1053 * Change notion of root (``/'') directory. 1054 */ 1055 /* ARGSUSED */ 1056 int 1057 sys_chroot(struct lwp *l, void *v, register_t *retval) 1058 { 1059 struct sys_chroot_args /* { 1060 syscallarg(const char *) path; 1061 } */ *uap = v; 1062 struct proc *p = l->l_proc; 1063 struct cwdinfo *cwdi = p->p_cwdi; 1064 struct vnode *vp; 1065 int error; 1066 struct nameidata nd; 1067 1068 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1069 &p->p_acflag)) != 0) 1070 return (error); 1071 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1072 SCARG(uap, path), l); 1073 if ((error = change_dir(&nd, l)) != 0) 1074 return (error); 1075 if (cwdi->cwdi_rdir != NULL) 1076 vrele(cwdi->cwdi_rdir); 1077 vp = nd.ni_vp; 1078 cwdi->cwdi_rdir = vp; 1079 1080 /* 1081 * Prevent escaping from chroot by putting the root under 1082 * the working directory. Silently chdir to / if we aren't 1083 * already there. 1084 */ 1085 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1086 /* 1087 * XXX would be more failsafe to change directory to a 1088 * deadfs node here instead 1089 */ 1090 vrele(cwdi->cwdi_cdir); 1091 VREF(vp); 1092 cwdi->cwdi_cdir = vp; 1093 } 1094 1095 return (0); 1096 } 1097 1098 /* 1099 * Common routine for chroot and chdir. 1100 */ 1101 static int 1102 change_dir(struct nameidata *ndp, struct lwp *l) 1103 { 1104 struct vnode *vp; 1105 int error; 1106 1107 if ((error = namei(ndp)) != 0) 1108 return (error); 1109 vp = ndp->ni_vp; 1110 if (vp->v_type != VDIR) 1111 error = ENOTDIR; 1112 else 1113 error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_cred, l); 1114 1115 if (error) 1116 vput(vp); 1117 else 1118 VOP_UNLOCK(vp, 0); 1119 return (error); 1120 } 1121 1122 /* 1123 * Check permissions, allocate an open file structure, 1124 * and call the device open routine if any. 1125 */ 1126 int 1127 sys_open(struct lwp *l, void *v, register_t *retval) 1128 { 1129 struct sys_open_args /* { 1130 syscallarg(const char *) path; 1131 syscallarg(int) flags; 1132 syscallarg(int) mode; 1133 } */ *uap = v; 1134 struct proc *p = l->l_proc; 1135 struct cwdinfo *cwdi = p->p_cwdi; 1136 struct filedesc *fdp = p->p_fd; 1137 struct file *fp; 1138 struct vnode *vp; 1139 int flags, cmode; 1140 int type, indx, error; 1141 struct flock lf; 1142 struct nameidata nd; 1143 1144 flags = FFLAGS(SCARG(uap, flags)); 1145 if ((flags & (FREAD | FWRITE)) == 0) 1146 return (EINVAL); 1147 /* falloc() will use the file descriptor for us */ 1148 if ((error = falloc(p, &fp, &indx)) != 0) 1149 return (error); 1150 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1151 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1152 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1153 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1154 FILE_UNUSE(fp, l); 1155 fdp->fd_ofiles[indx] = NULL; 1156 ffree(fp); 1157 if ((error == EDUPFD || error == EMOVEFD) && 1158 l->l_dupfd >= 0 && /* XXX from fdopen */ 1159 (error = 1160 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1161 *retval = indx; 1162 return (0); 1163 } 1164 if (error == ERESTART) 1165 error = EINTR; 1166 fdremove(fdp, indx); 1167 return (error); 1168 } 1169 l->l_dupfd = 0; 1170 vp = nd.ni_vp; 1171 fp->f_flag = flags & FMASK; 1172 fp->f_type = DTYPE_VNODE; 1173 fp->f_ops = &vnops; 1174 fp->f_data = vp; 1175 if (flags & (O_EXLOCK | O_SHLOCK)) { 1176 lf.l_whence = SEEK_SET; 1177 lf.l_start = 0; 1178 lf.l_len = 0; 1179 if (flags & O_EXLOCK) 1180 lf.l_type = F_WRLCK; 1181 else 1182 lf.l_type = F_RDLCK; 1183 type = F_FLOCK; 1184 if ((flags & FNONBLOCK) == 0) 1185 type |= F_WAIT; 1186 VOP_UNLOCK(vp, 0); 1187 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1188 if (error) { 1189 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1190 FILE_UNUSE(fp, l); 1191 ffree(fp); 1192 fdremove(fdp, indx); 1193 return (error); 1194 } 1195 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1196 fp->f_flag |= FHASLOCK; 1197 } 1198 VOP_UNLOCK(vp, 0); 1199 *retval = indx; 1200 FILE_SET_MATURE(fp); 1201 FILE_UNUSE(fp, l); 1202 return (0); 1203 } 1204 1205 /* 1206 * Get file handle system call 1207 */ 1208 int 1209 sys_getfh(struct lwp *l, void *v, register_t *retval) 1210 { 1211 struct sys_getfh_args /* { 1212 syscallarg(char *) fname; 1213 syscallarg(fhandle_t *) fhp; 1214 } */ *uap = v; 1215 struct proc *p = l->l_proc; 1216 struct vnode *vp; 1217 fhandle_t fh; 1218 int error; 1219 struct nameidata nd; 1220 1221 /* 1222 * Must be super user 1223 */ 1224 error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1225 &p->p_acflag); 1226 if (error) 1227 return (error); 1228 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1229 SCARG(uap, fname), l); 1230 error = namei(&nd); 1231 if (error) 1232 return (error); 1233 vp = nd.ni_vp; 1234 if (vp->v_mount->mnt_op->vfs_vptofh == NULL) 1235 return EOPNOTSUPP; 1236 memset(&fh, 0, sizeof(fh)); 1237 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1238 error = VFS_VPTOFH(vp, &fh.fh_fid); 1239 vput(vp); 1240 if (error) 1241 return (error); 1242 KASSERT(fh.fh_fid.fid_len <= _VFS_MAXFIDSZ); 1243 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1244 return (error); 1245 } 1246 1247 /* 1248 * Open a file given a file handle. 1249 * 1250 * Check permissions, allocate an open file structure, 1251 * and call the device open routine if any. 1252 */ 1253 int 1254 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1255 { 1256 struct sys_fhopen_args /* { 1257 syscallarg(const fhandle_t *) fhp; 1258 syscallarg(int) flags; 1259 } */ *uap = v; 1260 struct proc *p = l->l_proc; 1261 struct filedesc *fdp = p->p_fd; 1262 struct file *fp; 1263 struct vnode *vp = NULL; 1264 struct mount *mp; 1265 kauth_cred_t cred = p->p_cred; 1266 int flags; 1267 struct file *nfp; 1268 int type, indx, error=0; 1269 struct flock lf; 1270 struct vattr va; 1271 fhandle_t fh; 1272 1273 /* 1274 * Must be super user 1275 */ 1276 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1277 &p->p_acflag))) 1278 return (error); 1279 1280 flags = FFLAGS(SCARG(uap, flags)); 1281 if ((flags & (FREAD | FWRITE)) == 0) 1282 return (EINVAL); 1283 if ((flags & O_CREAT)) 1284 return (EINVAL); 1285 /* falloc() will use the file descriptor for us */ 1286 if ((error = falloc(p, &nfp, &indx)) != 0) 1287 return (error); 1288 fp = nfp; 1289 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1290 goto bad; 1291 1292 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1293 error = ESTALE; 1294 goto bad; 1295 } 1296 1297 if (mp->mnt_op->vfs_fhtovp == NULL) { 1298 error = EOPNOTSUPP; 1299 goto bad; 1300 } 1301 1302 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1303 vp = NULL; /* most likely unnecessary sanity for bad: */ 1304 goto bad; 1305 } 1306 1307 /* Now do an effective vn_open */ 1308 1309 if (vp->v_type == VSOCK) { 1310 error = EOPNOTSUPP; 1311 goto bad; 1312 } 1313 if (flags & FREAD) { 1314 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1315 goto bad; 1316 } 1317 if (flags & (FWRITE | O_TRUNC)) { 1318 if (vp->v_type == VDIR) { 1319 error = EISDIR; 1320 goto bad; 1321 } 1322 if ((error = vn_writechk(vp)) != 0 || 1323 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1324 goto bad; 1325 } 1326 if (flags & O_TRUNC) { 1327 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1328 goto bad; 1329 VOP_UNLOCK(vp, 0); /* XXX */ 1330 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1331 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1332 VATTR_NULL(&va); 1333 va.va_size = 0; 1334 error = VOP_SETATTR(vp, &va, cred, l); 1335 vn_finished_write(mp, 0); 1336 if (error) 1337 goto bad; 1338 } 1339 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1340 goto bad; 1341 if (vp->v_type == VREG && 1342 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1343 error = EIO; 1344 goto bad; 1345 } 1346 if (flags & FWRITE) 1347 vp->v_writecount++; 1348 1349 /* done with modified vn_open, now finish what sys_open does. */ 1350 1351 fp->f_flag = flags & FMASK; 1352 fp->f_type = DTYPE_VNODE; 1353 fp->f_ops = &vnops; 1354 fp->f_data = vp; 1355 if (flags & (O_EXLOCK | O_SHLOCK)) { 1356 lf.l_whence = SEEK_SET; 1357 lf.l_start = 0; 1358 lf.l_len = 0; 1359 if (flags & O_EXLOCK) 1360 lf.l_type = F_WRLCK; 1361 else 1362 lf.l_type = F_RDLCK; 1363 type = F_FLOCK; 1364 if ((flags & FNONBLOCK) == 0) 1365 type |= F_WAIT; 1366 VOP_UNLOCK(vp, 0); 1367 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1368 if (error) { 1369 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1370 FILE_UNUSE(fp, l); 1371 ffree(fp); 1372 fdremove(fdp, indx); 1373 return (error); 1374 } 1375 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1376 fp->f_flag |= FHASLOCK; 1377 } 1378 VOP_UNLOCK(vp, 0); 1379 *retval = indx; 1380 FILE_SET_MATURE(fp); 1381 FILE_UNUSE(fp, l); 1382 return (0); 1383 1384 bad: 1385 FILE_UNUSE(fp, l); 1386 ffree(fp); 1387 fdremove(fdp, indx); 1388 if (vp != NULL) 1389 vput(vp); 1390 return (error); 1391 } 1392 1393 /* ARGSUSED */ 1394 int 1395 sys___fhstat30(struct lwp *l, void *v, register_t *retval) 1396 { 1397 struct sys___fhstat30_args /* { 1398 syscallarg(const fhandle_t *) fhp; 1399 syscallarg(struct stat *) sb; 1400 } */ *uap = v; 1401 struct proc *p = l->l_proc; 1402 struct stat sb; 1403 int error; 1404 fhandle_t fh; 1405 struct mount *mp; 1406 struct vnode *vp; 1407 1408 /* 1409 * Must be super user 1410 */ 1411 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1412 &p->p_acflag))) 1413 return (error); 1414 1415 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1416 return (error); 1417 1418 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1419 return (ESTALE); 1420 if (mp->mnt_op->vfs_fhtovp == NULL) 1421 return EOPNOTSUPP; 1422 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1423 return (error); 1424 error = vn_stat(vp, &sb, l); 1425 vput(vp); 1426 if (error) 1427 return (error); 1428 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1429 return (error); 1430 } 1431 1432 /* ARGSUSED */ 1433 int 1434 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1435 { 1436 struct sys_fhstatvfs1_args /* { 1437 syscallarg(const fhandle_t *) fhp; 1438 syscallarg(struct statvfs *) buf; 1439 syscallarg(int) flags; 1440 } */ *uap = v; 1441 struct proc *p = l->l_proc; 1442 struct statvfs *sb; 1443 fhandle_t fh; 1444 struct mount *mp; 1445 struct vnode *vp; 1446 int error; 1447 1448 /* 1449 * Must be super user 1450 */ 1451 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1452 &p->p_acflag)) != 0) 1453 return error; 1454 1455 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1456 return error; 1457 1458 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1459 return ESTALE; 1460 if (mp->mnt_op->vfs_fhtovp == NULL) 1461 return EOPNOTSUPP; 1462 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1463 return error; 1464 1465 mp = vp->v_mount; 1466 sb = STATVFSBUF_GET(); 1467 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) { 1468 vput(vp); 1469 goto out; 1470 } 1471 vput(vp); 1472 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1473 out: 1474 STATVFSBUF_PUT(sb); 1475 return error; 1476 } 1477 1478 /* 1479 * Create a special file. 1480 */ 1481 /* ARGSUSED */ 1482 int 1483 sys_mknod(struct lwp *l, void *v, register_t *retval) 1484 { 1485 struct sys_mknod_args /* { 1486 syscallarg(const char *) path; 1487 syscallarg(int) mode; 1488 syscallarg(int) dev; 1489 } */ *uap = v; 1490 struct proc *p = l->l_proc; 1491 struct vnode *vp; 1492 struct mount *mp; 1493 struct vattr vattr; 1494 int error; 1495 int whiteout = 0; 1496 struct nameidata nd; 1497 1498 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1499 &p->p_acflag)) != 0) 1500 return (error); 1501 restart: 1502 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1503 if ((error = namei(&nd)) != 0) 1504 return (error); 1505 vp = nd.ni_vp; 1506 if (vp != NULL) 1507 error = EEXIST; 1508 else { 1509 VATTR_NULL(&vattr); 1510 vattr.va_mode = 1511 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1512 vattr.va_rdev = SCARG(uap, dev); 1513 whiteout = 0; 1514 1515 switch (SCARG(uap, mode) & S_IFMT) { 1516 case S_IFMT: /* used by badsect to flag bad sectors */ 1517 vattr.va_type = VBAD; 1518 break; 1519 case S_IFCHR: 1520 vattr.va_type = VCHR; 1521 break; 1522 case S_IFBLK: 1523 vattr.va_type = VBLK; 1524 break; 1525 case S_IFWHT: 1526 whiteout = 1; 1527 break; 1528 default: 1529 error = EINVAL; 1530 break; 1531 } 1532 } 1533 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1534 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1535 if (nd.ni_dvp == vp) 1536 vrele(nd.ni_dvp); 1537 else 1538 vput(nd.ni_dvp); 1539 if (vp) 1540 vrele(vp); 1541 if ((error = vn_start_write(NULL, &mp, 1542 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1543 return (error); 1544 goto restart; 1545 } 1546 if (!error) { 1547 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1548 if (whiteout) { 1549 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1550 if (error) 1551 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1552 vput(nd.ni_dvp); 1553 } else { 1554 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1555 &nd.ni_cnd, &vattr); 1556 if (error == 0) 1557 vput(nd.ni_vp); 1558 } 1559 } else { 1560 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1561 if (nd.ni_dvp == vp) 1562 vrele(nd.ni_dvp); 1563 else 1564 vput(nd.ni_dvp); 1565 if (vp) 1566 vrele(vp); 1567 } 1568 vn_finished_write(mp, 0); 1569 return (error); 1570 } 1571 1572 /* 1573 * Create a named pipe. 1574 */ 1575 /* ARGSUSED */ 1576 int 1577 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1578 { 1579 struct sys_mkfifo_args /* { 1580 syscallarg(const char *) path; 1581 syscallarg(int) mode; 1582 } */ *uap = v; 1583 struct proc *p = l->l_proc; 1584 struct mount *mp; 1585 struct vattr vattr; 1586 int error; 1587 struct nameidata nd; 1588 1589 restart: 1590 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1591 if ((error = namei(&nd)) != 0) 1592 return (error); 1593 if (nd.ni_vp != NULL) { 1594 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1595 if (nd.ni_dvp == nd.ni_vp) 1596 vrele(nd.ni_dvp); 1597 else 1598 vput(nd.ni_dvp); 1599 vrele(nd.ni_vp); 1600 return (EEXIST); 1601 } 1602 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1603 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1604 if (nd.ni_dvp == nd.ni_vp) 1605 vrele(nd.ni_dvp); 1606 else 1607 vput(nd.ni_dvp); 1608 if (nd.ni_vp) 1609 vrele(nd.ni_vp); 1610 if ((error = vn_start_write(NULL, &mp, 1611 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1612 return (error); 1613 goto restart; 1614 } 1615 VATTR_NULL(&vattr); 1616 vattr.va_type = VFIFO; 1617 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1618 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1619 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1620 if (error == 0) 1621 vput(nd.ni_vp); 1622 vn_finished_write(mp, 0); 1623 return (error); 1624 } 1625 1626 /* 1627 * Make a hard file link. 1628 */ 1629 /* ARGSUSED */ 1630 int 1631 sys_link(struct lwp *l, void *v, register_t *retval) 1632 { 1633 struct sys_link_args /* { 1634 syscallarg(const char *) path; 1635 syscallarg(const char *) link; 1636 } */ *uap = v; 1637 struct proc *p = l->l_proc; 1638 struct vnode *vp; 1639 struct mount *mp; 1640 struct nameidata nd; 1641 int error; 1642 1643 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1644 if ((error = namei(&nd)) != 0) 1645 return (error); 1646 vp = nd.ni_vp; 1647 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1648 vrele(vp); 1649 return (error); 1650 } 1651 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1652 if ((error = namei(&nd)) != 0) 1653 goto out; 1654 if (nd.ni_vp) { 1655 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1656 if (nd.ni_dvp == nd.ni_vp) 1657 vrele(nd.ni_dvp); 1658 else 1659 vput(nd.ni_dvp); 1660 vrele(nd.ni_vp); 1661 error = EEXIST; 1662 goto out; 1663 } 1664 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1665 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1666 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1667 out: 1668 vrele(vp); 1669 vn_finished_write(mp, 0); 1670 return (error); 1671 } 1672 1673 /* 1674 * Make a symbolic link. 1675 */ 1676 /* ARGSUSED */ 1677 int 1678 sys_symlink(struct lwp *l, void *v, register_t *retval) 1679 { 1680 struct sys_symlink_args /* { 1681 syscallarg(const char *) path; 1682 syscallarg(const char *) link; 1683 } */ *uap = v; 1684 struct proc *p = l->l_proc; 1685 struct mount *mp; 1686 struct vattr vattr; 1687 char *path; 1688 int error; 1689 struct nameidata nd; 1690 1691 path = PNBUF_GET(); 1692 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1693 if (error) 1694 goto out; 1695 restart: 1696 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1697 if ((error = namei(&nd)) != 0) 1698 goto out; 1699 if (nd.ni_vp) { 1700 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1701 if (nd.ni_dvp == nd.ni_vp) 1702 vrele(nd.ni_dvp); 1703 else 1704 vput(nd.ni_dvp); 1705 vrele(nd.ni_vp); 1706 error = EEXIST; 1707 goto out; 1708 } 1709 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1710 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1711 if (nd.ni_dvp == nd.ni_vp) 1712 vrele(nd.ni_dvp); 1713 else 1714 vput(nd.ni_dvp); 1715 if ((error = vn_start_write(NULL, &mp, 1716 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1717 return (error); 1718 goto restart; 1719 } 1720 VATTR_NULL(&vattr); 1721 vattr.va_type = VLNK; 1722 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1723 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1724 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1725 if (error == 0) 1726 vput(nd.ni_vp); 1727 vn_finished_write(mp, 0); 1728 out: 1729 PNBUF_PUT(path); 1730 return (error); 1731 } 1732 1733 /* 1734 * Delete a whiteout from the filesystem. 1735 */ 1736 /* ARGSUSED */ 1737 int 1738 sys_undelete(struct lwp *l, void *v, register_t *retval) 1739 { 1740 struct sys_undelete_args /* { 1741 syscallarg(const char *) path; 1742 } */ *uap = v; 1743 struct proc *p = l->l_proc; 1744 int error; 1745 struct mount *mp; 1746 struct nameidata nd; 1747 1748 restart: 1749 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1750 SCARG(uap, path), l); 1751 error = namei(&nd); 1752 if (error) 1753 return (error); 1754 1755 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1756 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1757 if (nd.ni_dvp == nd.ni_vp) 1758 vrele(nd.ni_dvp); 1759 else 1760 vput(nd.ni_dvp); 1761 if (nd.ni_vp) 1762 vrele(nd.ni_vp); 1763 return (EEXIST); 1764 } 1765 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1766 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1767 if (nd.ni_dvp == nd.ni_vp) 1768 vrele(nd.ni_dvp); 1769 else 1770 vput(nd.ni_dvp); 1771 if ((error = vn_start_write(NULL, &mp, 1772 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1773 return (error); 1774 goto restart; 1775 } 1776 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1777 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1778 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1779 vput(nd.ni_dvp); 1780 vn_finished_write(mp, 0); 1781 return (error); 1782 } 1783 1784 /* 1785 * Delete a name from the filesystem. 1786 */ 1787 /* ARGSUSED */ 1788 int 1789 sys_unlink(struct lwp *l, void *v, register_t *retval) 1790 { 1791 struct sys_unlink_args /* { 1792 syscallarg(const char *) path; 1793 } */ *uap = v; 1794 struct proc *p = l->l_proc; 1795 struct mount *mp; 1796 struct vnode *vp; 1797 int error; 1798 struct nameidata nd; 1799 1800 restart: 1801 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1802 SCARG(uap, path), l); 1803 if ((error = namei(&nd)) != 0) 1804 return (error); 1805 vp = nd.ni_vp; 1806 1807 /* 1808 * The root of a mounted filesystem cannot be deleted. 1809 */ 1810 if (vp->v_flag & VROOT) { 1811 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1812 if (nd.ni_dvp == vp) 1813 vrele(nd.ni_dvp); 1814 else 1815 vput(nd.ni_dvp); 1816 vput(vp); 1817 error = EBUSY; 1818 goto out; 1819 } 1820 1821 #ifdef VERIFIED_EXEC 1822 /* Handle remove requests for veriexec entries. */ 1823 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) { 1824 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1825 if (nd.ni_dvp == vp) 1826 vrele(nd.ni_dvp); 1827 else 1828 vput(nd.ni_dvp); 1829 vput(vp); 1830 goto out; 1831 } 1832 #endif /* VERIFIED_EXEC */ 1833 1834 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1835 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1836 if (nd.ni_dvp == vp) 1837 vrele(nd.ni_dvp); 1838 else 1839 vput(nd.ni_dvp); 1840 vput(vp); 1841 if ((error = vn_start_write(NULL, &mp, 1842 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1843 return (error); 1844 goto restart; 1845 } 1846 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1847 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1848 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1849 vn_finished_write(mp, 0); 1850 out: 1851 return (error); 1852 } 1853 1854 /* 1855 * Reposition read/write file offset. 1856 */ 1857 int 1858 sys_lseek(struct lwp *l, void *v, register_t *retval) 1859 { 1860 struct sys_lseek_args /* { 1861 syscallarg(int) fd; 1862 syscallarg(int) pad; 1863 syscallarg(off_t) offset; 1864 syscallarg(int) whence; 1865 } */ *uap = v; 1866 struct proc *p = l->l_proc; 1867 kauth_cred_t cred = p->p_cred; 1868 struct filedesc *fdp = p->p_fd; 1869 struct file *fp; 1870 struct vnode *vp; 1871 struct vattr vattr; 1872 off_t newoff; 1873 int error; 1874 1875 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1876 return (EBADF); 1877 1878 FILE_USE(fp); 1879 1880 vp = (struct vnode *)fp->f_data; 1881 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1882 error = ESPIPE; 1883 goto out; 1884 } 1885 1886 switch (SCARG(uap, whence)) { 1887 case SEEK_CUR: 1888 newoff = fp->f_offset + SCARG(uap, offset); 1889 break; 1890 case SEEK_END: 1891 error = VOP_GETATTR(vp, &vattr, cred, l); 1892 if (error) 1893 goto out; 1894 newoff = SCARG(uap, offset) + vattr.va_size; 1895 break; 1896 case SEEK_SET: 1897 newoff = SCARG(uap, offset); 1898 break; 1899 default: 1900 error = EINVAL; 1901 goto out; 1902 } 1903 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1904 goto out; 1905 1906 *(off_t *)retval = fp->f_offset = newoff; 1907 out: 1908 FILE_UNUSE(fp, l); 1909 return (error); 1910 } 1911 1912 /* 1913 * Positional read system call. 1914 */ 1915 int 1916 sys_pread(struct lwp *l, void *v, register_t *retval) 1917 { 1918 struct sys_pread_args /* { 1919 syscallarg(int) fd; 1920 syscallarg(void *) buf; 1921 syscallarg(size_t) nbyte; 1922 syscallarg(off_t) offset; 1923 } */ *uap = v; 1924 struct proc *p = l->l_proc; 1925 struct filedesc *fdp = p->p_fd; 1926 struct file *fp; 1927 struct vnode *vp; 1928 off_t offset; 1929 int error, fd = SCARG(uap, fd); 1930 1931 if ((fp = fd_getfile(fdp, fd)) == NULL) 1932 return (EBADF); 1933 1934 if ((fp->f_flag & FREAD) == 0) { 1935 simple_unlock(&fp->f_slock); 1936 return (EBADF); 1937 } 1938 1939 FILE_USE(fp); 1940 1941 vp = (struct vnode *)fp->f_data; 1942 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1943 error = ESPIPE; 1944 goto out; 1945 } 1946 1947 offset = SCARG(uap, offset); 1948 1949 /* 1950 * XXX This works because no file systems actually 1951 * XXX take any action on the seek operation. 1952 */ 1953 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1954 goto out; 1955 1956 /* dofileread() will unuse the descriptor for us */ 1957 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1958 &offset, 0, retval)); 1959 1960 out: 1961 FILE_UNUSE(fp, l); 1962 return (error); 1963 } 1964 1965 /* 1966 * Positional scatter read system call. 1967 */ 1968 int 1969 sys_preadv(struct lwp *l, void *v, register_t *retval) 1970 { 1971 struct sys_preadv_args /* { 1972 syscallarg(int) fd; 1973 syscallarg(const struct iovec *) iovp; 1974 syscallarg(int) iovcnt; 1975 syscallarg(off_t) offset; 1976 } */ *uap = v; 1977 struct proc *p = l->l_proc; 1978 struct filedesc *fdp = p->p_fd; 1979 struct file *fp; 1980 struct vnode *vp; 1981 off_t offset; 1982 int error, fd = SCARG(uap, fd); 1983 1984 if ((fp = fd_getfile(fdp, fd)) == NULL) 1985 return (EBADF); 1986 1987 if ((fp->f_flag & FREAD) == 0) { 1988 simple_unlock(&fp->f_slock); 1989 return (EBADF); 1990 } 1991 1992 FILE_USE(fp); 1993 1994 vp = (struct vnode *)fp->f_data; 1995 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1996 error = ESPIPE; 1997 goto out; 1998 } 1999 2000 offset = SCARG(uap, offset); 2001 2002 /* 2003 * XXX This works because no file systems actually 2004 * XXX take any action on the seek operation. 2005 */ 2006 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2007 goto out; 2008 2009 /* dofilereadv() will unuse the descriptor for us */ 2010 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2011 &offset, 0, retval)); 2012 2013 out: 2014 FILE_UNUSE(fp, l); 2015 return (error); 2016 } 2017 2018 /* 2019 * Positional write system call. 2020 */ 2021 int 2022 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2023 { 2024 struct sys_pwrite_args /* { 2025 syscallarg(int) fd; 2026 syscallarg(const void *) buf; 2027 syscallarg(size_t) nbyte; 2028 syscallarg(off_t) offset; 2029 } */ *uap = v; 2030 struct proc *p = l->l_proc; 2031 struct filedesc *fdp = p->p_fd; 2032 struct file *fp; 2033 struct vnode *vp; 2034 off_t offset; 2035 int error, fd = SCARG(uap, fd); 2036 2037 if ((fp = fd_getfile(fdp, fd)) == NULL) 2038 return (EBADF); 2039 2040 if ((fp->f_flag & FWRITE) == 0) { 2041 simple_unlock(&fp->f_slock); 2042 return (EBADF); 2043 } 2044 2045 FILE_USE(fp); 2046 2047 vp = (struct vnode *)fp->f_data; 2048 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2049 error = ESPIPE; 2050 goto out; 2051 } 2052 2053 offset = SCARG(uap, offset); 2054 2055 /* 2056 * XXX This works because no file systems actually 2057 * XXX take any action on the seek operation. 2058 */ 2059 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2060 goto out; 2061 2062 /* dofilewrite() will unuse the descriptor for us */ 2063 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2064 &offset, 0, retval)); 2065 2066 out: 2067 FILE_UNUSE(fp, l); 2068 return (error); 2069 } 2070 2071 /* 2072 * Positional gather write system call. 2073 */ 2074 int 2075 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2076 { 2077 struct sys_pwritev_args /* { 2078 syscallarg(int) fd; 2079 syscallarg(const struct iovec *) iovp; 2080 syscallarg(int) iovcnt; 2081 syscallarg(off_t) offset; 2082 } */ *uap = v; 2083 struct proc *p = l->l_proc; 2084 struct filedesc *fdp = p->p_fd; 2085 struct file *fp; 2086 struct vnode *vp; 2087 off_t offset; 2088 int error, fd = SCARG(uap, fd); 2089 2090 if ((fp = fd_getfile(fdp, fd)) == NULL) 2091 return (EBADF); 2092 2093 if ((fp->f_flag & FWRITE) == 0) { 2094 simple_unlock(&fp->f_slock); 2095 return (EBADF); 2096 } 2097 2098 FILE_USE(fp); 2099 2100 vp = (struct vnode *)fp->f_data; 2101 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2102 error = ESPIPE; 2103 goto out; 2104 } 2105 2106 offset = SCARG(uap, offset); 2107 2108 /* 2109 * XXX This works because no file systems actually 2110 * XXX take any action on the seek operation. 2111 */ 2112 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2113 goto out; 2114 2115 /* dofilewritev() will unuse the descriptor for us */ 2116 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2117 &offset, 0, retval)); 2118 2119 out: 2120 FILE_UNUSE(fp, l); 2121 return (error); 2122 } 2123 2124 /* 2125 * Check access permissions. 2126 */ 2127 int 2128 sys_access(struct lwp *l, void *v, register_t *retval) 2129 { 2130 struct sys_access_args /* { 2131 syscallarg(const char *) path; 2132 syscallarg(int) flags; 2133 } */ *uap = v; 2134 struct proc *p = l->l_proc; 2135 kauth_cred_t cred; 2136 struct vnode *vp; 2137 int error, flags; 2138 struct nameidata nd; 2139 2140 cred = kauth_cred_dup(p->p_cred); 2141 kauth_cred_seteuid(cred, kauth_cred_getuid(p->p_cred)); 2142 kauth_cred_setegid(cred, kauth_cred_getgid(p->p_cred)); 2143 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2144 SCARG(uap, path), l); 2145 /* Override default credentials */ 2146 nd.ni_cnd.cn_cred = cred; 2147 if ((error = namei(&nd)) != 0) 2148 goto out; 2149 vp = nd.ni_vp; 2150 2151 /* Flags == 0 means only check for existence. */ 2152 if (SCARG(uap, flags)) { 2153 flags = 0; 2154 if (SCARG(uap, flags) & R_OK) 2155 flags |= VREAD; 2156 if (SCARG(uap, flags) & W_OK) 2157 flags |= VWRITE; 2158 if (SCARG(uap, flags) & X_OK) 2159 flags |= VEXEC; 2160 2161 error = VOP_ACCESS(vp, flags, cred, l); 2162 if (!error && (flags & VWRITE)) 2163 error = vn_writechk(vp); 2164 } 2165 vput(vp); 2166 out: 2167 kauth_cred_free(cred); 2168 return (error); 2169 } 2170 2171 /* 2172 * Get file status; this version follows links. 2173 */ 2174 /* ARGSUSED */ 2175 int 2176 sys___stat30(struct lwp *l, void *v, register_t *retval) 2177 { 2178 struct sys___stat30_args /* { 2179 syscallarg(const char *) path; 2180 syscallarg(struct stat *) ub; 2181 } */ *uap = v; 2182 struct stat sb; 2183 int error; 2184 struct nameidata nd; 2185 2186 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2187 SCARG(uap, path), l); 2188 if ((error = namei(&nd)) != 0) 2189 return (error); 2190 error = vn_stat(nd.ni_vp, &sb, l); 2191 vput(nd.ni_vp); 2192 if (error) 2193 return (error); 2194 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2195 return (error); 2196 } 2197 2198 /* 2199 * Get file status; this version does not follow links. 2200 */ 2201 /* ARGSUSED */ 2202 int 2203 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2204 { 2205 struct sys___lstat30_args /* { 2206 syscallarg(const char *) path; 2207 syscallarg(struct stat *) ub; 2208 } */ *uap = v; 2209 struct stat sb; 2210 int error; 2211 struct nameidata nd; 2212 2213 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2214 SCARG(uap, path), l); 2215 if ((error = namei(&nd)) != 0) 2216 return (error); 2217 error = vn_stat(nd.ni_vp, &sb, l); 2218 vput(nd.ni_vp); 2219 if (error) 2220 return (error); 2221 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2222 return (error); 2223 } 2224 2225 /* 2226 * Get configurable pathname variables. 2227 */ 2228 /* ARGSUSED */ 2229 int 2230 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2231 { 2232 struct sys_pathconf_args /* { 2233 syscallarg(const char *) path; 2234 syscallarg(int) name; 2235 } */ *uap = v; 2236 int error; 2237 struct nameidata nd; 2238 2239 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2240 SCARG(uap, path), l); 2241 if ((error = namei(&nd)) != 0) 2242 return (error); 2243 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2244 vput(nd.ni_vp); 2245 return (error); 2246 } 2247 2248 /* 2249 * Return target name of a symbolic link. 2250 */ 2251 /* ARGSUSED */ 2252 int 2253 sys_readlink(struct lwp *l, void *v, register_t *retval) 2254 { 2255 struct sys_readlink_args /* { 2256 syscallarg(const char *) path; 2257 syscallarg(char *) buf; 2258 syscallarg(size_t) count; 2259 } */ *uap = v; 2260 struct proc *p = l->l_proc; 2261 struct vnode *vp; 2262 struct iovec aiov; 2263 struct uio auio; 2264 int error; 2265 struct nameidata nd; 2266 2267 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2268 SCARG(uap, path), l); 2269 if ((error = namei(&nd)) != 0) 2270 return (error); 2271 vp = nd.ni_vp; 2272 if (vp->v_type != VLNK) 2273 error = EINVAL; 2274 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2275 (error = VOP_ACCESS(vp, VREAD, p->p_cred, l)) == 0) { 2276 aiov.iov_base = SCARG(uap, buf); 2277 aiov.iov_len = SCARG(uap, count); 2278 auio.uio_iov = &aiov; 2279 auio.uio_iovcnt = 1; 2280 auio.uio_offset = 0; 2281 auio.uio_rw = UIO_READ; 2282 KASSERT(l == curlwp); 2283 auio.uio_vmspace = l->l_proc->p_vmspace; 2284 auio.uio_resid = SCARG(uap, count); 2285 error = VOP_READLINK(vp, &auio, p->p_cred); 2286 } 2287 vput(vp); 2288 *retval = SCARG(uap, count) - auio.uio_resid; 2289 return (error); 2290 } 2291 2292 /* 2293 * Change flags of a file given a path name. 2294 */ 2295 /* ARGSUSED */ 2296 int 2297 sys_chflags(struct lwp *l, void *v, register_t *retval) 2298 { 2299 struct sys_chflags_args /* { 2300 syscallarg(const char *) path; 2301 syscallarg(u_long) flags; 2302 } */ *uap = v; 2303 struct vnode *vp; 2304 int error; 2305 struct nameidata nd; 2306 2307 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2308 if ((error = namei(&nd)) != 0) 2309 return (error); 2310 vp = nd.ni_vp; 2311 error = change_flags(vp, SCARG(uap, flags), l); 2312 vput(vp); 2313 return (error); 2314 } 2315 2316 /* 2317 * Change flags of a file given a file descriptor. 2318 */ 2319 /* ARGSUSED */ 2320 int 2321 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2322 { 2323 struct sys_fchflags_args /* { 2324 syscallarg(int) fd; 2325 syscallarg(u_long) flags; 2326 } */ *uap = v; 2327 struct proc *p = l->l_proc; 2328 struct vnode *vp; 2329 struct file *fp; 2330 int error; 2331 2332 /* getvnode() will use the descriptor for us */ 2333 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2334 return (error); 2335 vp = (struct vnode *)fp->f_data; 2336 error = change_flags(vp, SCARG(uap, flags), l); 2337 VOP_UNLOCK(vp, 0); 2338 FILE_UNUSE(fp, l); 2339 return (error); 2340 } 2341 2342 /* 2343 * Change flags of a file given a path name; this version does 2344 * not follow links. 2345 */ 2346 int 2347 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2348 { 2349 struct sys_lchflags_args /* { 2350 syscallarg(const char *) path; 2351 syscallarg(u_long) flags; 2352 } */ *uap = v; 2353 struct vnode *vp; 2354 int error; 2355 struct nameidata nd; 2356 2357 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2358 if ((error = namei(&nd)) != 0) 2359 return (error); 2360 vp = nd.ni_vp; 2361 error = change_flags(vp, SCARG(uap, flags), l); 2362 vput(vp); 2363 return (error); 2364 } 2365 2366 /* 2367 * Common routine to change flags of a file. 2368 */ 2369 int 2370 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2371 { 2372 struct proc *p = l->l_proc; 2373 struct mount *mp; 2374 struct vattr vattr; 2375 int error; 2376 2377 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2378 return (error); 2379 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2380 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2381 /* 2382 * Non-superusers cannot change the flags on devices, even if they 2383 * own them. 2384 */ 2385 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2386 &p->p_acflag) != 0) { 2387 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2388 goto out; 2389 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2390 error = EINVAL; 2391 goto out; 2392 } 2393 } 2394 VATTR_NULL(&vattr); 2395 vattr.va_flags = flags; 2396 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2397 out: 2398 vn_finished_write(mp, 0); 2399 return (error); 2400 } 2401 2402 /* 2403 * Change mode of a file given path name; this version follows links. 2404 */ 2405 /* ARGSUSED */ 2406 int 2407 sys_chmod(struct lwp *l, void *v, register_t *retval) 2408 { 2409 struct sys_chmod_args /* { 2410 syscallarg(const char *) path; 2411 syscallarg(int) mode; 2412 } */ *uap = v; 2413 int error; 2414 struct nameidata nd; 2415 2416 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2417 if ((error = namei(&nd)) != 0) 2418 return (error); 2419 2420 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2421 2422 vrele(nd.ni_vp); 2423 return (error); 2424 } 2425 2426 /* 2427 * Change mode of a file given a file descriptor. 2428 */ 2429 /* ARGSUSED */ 2430 int 2431 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2432 { 2433 struct sys_fchmod_args /* { 2434 syscallarg(int) fd; 2435 syscallarg(int) mode; 2436 } */ *uap = v; 2437 struct proc *p = l->l_proc; 2438 struct file *fp; 2439 int error; 2440 2441 /* getvnode() will use the descriptor for us */ 2442 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2443 return (error); 2444 2445 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2446 FILE_UNUSE(fp, l); 2447 return (error); 2448 } 2449 2450 /* 2451 * Change mode of a file given path name; this version does not follow links. 2452 */ 2453 /* ARGSUSED */ 2454 int 2455 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2456 { 2457 struct sys_lchmod_args /* { 2458 syscallarg(const char *) path; 2459 syscallarg(int) mode; 2460 } */ *uap = v; 2461 int error; 2462 struct nameidata nd; 2463 2464 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2465 if ((error = namei(&nd)) != 0) 2466 return (error); 2467 2468 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2469 2470 vrele(nd.ni_vp); 2471 return (error); 2472 } 2473 2474 /* 2475 * Common routine to set mode given a vnode. 2476 */ 2477 static int 2478 change_mode(struct vnode *vp, int mode, struct lwp *l) 2479 { 2480 struct proc *p = l->l_proc; 2481 struct mount *mp; 2482 struct vattr vattr; 2483 int error; 2484 2485 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2486 return (error); 2487 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2488 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2489 VATTR_NULL(&vattr); 2490 vattr.va_mode = mode & ALLPERMS; 2491 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2492 VOP_UNLOCK(vp, 0); 2493 vn_finished_write(mp, 0); 2494 return (error); 2495 } 2496 2497 /* 2498 * Set ownership given a path name; this version follows links. 2499 */ 2500 /* ARGSUSED */ 2501 int 2502 sys_chown(struct lwp *l, void *v, register_t *retval) 2503 { 2504 struct sys_chown_args /* { 2505 syscallarg(const char *) path; 2506 syscallarg(uid_t) uid; 2507 syscallarg(gid_t) gid; 2508 } */ *uap = v; 2509 int error; 2510 struct nameidata nd; 2511 2512 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2513 if ((error = namei(&nd)) != 0) 2514 return (error); 2515 2516 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2517 2518 vrele(nd.ni_vp); 2519 return (error); 2520 } 2521 2522 /* 2523 * Set ownership given a path name; this version follows links. 2524 * Provides POSIX semantics. 2525 */ 2526 /* ARGSUSED */ 2527 int 2528 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2529 { 2530 struct sys_chown_args /* { 2531 syscallarg(const char *) path; 2532 syscallarg(uid_t) uid; 2533 syscallarg(gid_t) gid; 2534 } */ *uap = v; 2535 int error; 2536 struct nameidata nd; 2537 2538 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2539 if ((error = namei(&nd)) != 0) 2540 return (error); 2541 2542 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2543 2544 vrele(nd.ni_vp); 2545 return (error); 2546 } 2547 2548 /* 2549 * Set ownership given a file descriptor. 2550 */ 2551 /* ARGSUSED */ 2552 int 2553 sys_fchown(struct lwp *l, void *v, register_t *retval) 2554 { 2555 struct sys_fchown_args /* { 2556 syscallarg(int) fd; 2557 syscallarg(uid_t) uid; 2558 syscallarg(gid_t) gid; 2559 } */ *uap = v; 2560 struct proc *p = l->l_proc; 2561 int error; 2562 struct file *fp; 2563 2564 /* getvnode() will use the descriptor for us */ 2565 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2566 return (error); 2567 2568 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2569 SCARG(uap, gid), l, 0); 2570 FILE_UNUSE(fp, l); 2571 return (error); 2572 } 2573 2574 /* 2575 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2576 */ 2577 /* ARGSUSED */ 2578 int 2579 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2580 { 2581 struct sys_fchown_args /* { 2582 syscallarg(int) fd; 2583 syscallarg(uid_t) uid; 2584 syscallarg(gid_t) gid; 2585 } */ *uap = v; 2586 struct proc *p = l->l_proc; 2587 int error; 2588 struct file *fp; 2589 2590 /* getvnode() will use the descriptor for us */ 2591 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2592 return (error); 2593 2594 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2595 SCARG(uap, gid), l, 1); 2596 FILE_UNUSE(fp, l); 2597 return (error); 2598 } 2599 2600 /* 2601 * Set ownership given a path name; this version does not follow links. 2602 */ 2603 /* ARGSUSED */ 2604 int 2605 sys_lchown(struct lwp *l, void *v, register_t *retval) 2606 { 2607 struct sys_lchown_args /* { 2608 syscallarg(const char *) path; 2609 syscallarg(uid_t) uid; 2610 syscallarg(gid_t) gid; 2611 } */ *uap = v; 2612 int error; 2613 struct nameidata nd; 2614 2615 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2616 if ((error = namei(&nd)) != 0) 2617 return (error); 2618 2619 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2620 2621 vrele(nd.ni_vp); 2622 return (error); 2623 } 2624 2625 /* 2626 * Set ownership given a path name; this version does not follow links. 2627 * Provides POSIX/XPG semantics. 2628 */ 2629 /* ARGSUSED */ 2630 int 2631 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2632 { 2633 struct sys_lchown_args /* { 2634 syscallarg(const char *) path; 2635 syscallarg(uid_t) uid; 2636 syscallarg(gid_t) gid; 2637 } */ *uap = v; 2638 int error; 2639 struct nameidata nd; 2640 2641 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2642 if ((error = namei(&nd)) != 0) 2643 return (error); 2644 2645 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2646 2647 vrele(nd.ni_vp); 2648 return (error); 2649 } 2650 2651 /* 2652 * Common routine to set ownership given a vnode. 2653 */ 2654 static int 2655 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2656 int posix_semantics) 2657 { 2658 struct proc *p = l->l_proc; 2659 struct mount *mp; 2660 struct vattr vattr; 2661 mode_t newmode; 2662 int error; 2663 2664 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2665 return (error); 2666 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2667 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2668 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2669 goto out; 2670 2671 #define CHANGED(x) ((int)(x) != -1) 2672 newmode = vattr.va_mode; 2673 if (posix_semantics) { 2674 /* 2675 * POSIX/XPG semantics: if the caller is not the super-user, 2676 * clear set-user-id and set-group-id bits. Both POSIX and 2677 * the XPG consider the behaviour for calls by the super-user 2678 * implementation-defined; we leave the set-user-id and set- 2679 * group-id settings intact in that case. 2680 */ 2681 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2682 NULL) != 0) 2683 newmode &= ~(S_ISUID | S_ISGID); 2684 } else { 2685 /* 2686 * NetBSD semantics: when changing owner and/or group, 2687 * clear the respective bit(s). 2688 */ 2689 if (CHANGED(uid)) 2690 newmode &= ~S_ISUID; 2691 if (CHANGED(gid)) 2692 newmode &= ~S_ISGID; 2693 } 2694 /* Update va_mode iff altered. */ 2695 if (vattr.va_mode == newmode) 2696 newmode = VNOVAL; 2697 2698 VATTR_NULL(&vattr); 2699 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2700 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2701 vattr.va_mode = newmode; 2702 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2703 #undef CHANGED 2704 2705 out: 2706 VOP_UNLOCK(vp, 0); 2707 vn_finished_write(mp, 0); 2708 return (error); 2709 } 2710 2711 /* 2712 * Set the access and modification times given a path name; this 2713 * version follows links. 2714 */ 2715 /* ARGSUSED */ 2716 int 2717 sys_utimes(struct lwp *l, void *v, register_t *retval) 2718 { 2719 struct sys_utimes_args /* { 2720 syscallarg(const char *) path; 2721 syscallarg(const struct timeval *) tptr; 2722 } */ *uap = v; 2723 int error; 2724 struct nameidata nd; 2725 2726 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2727 if ((error = namei(&nd)) != 0) 2728 return (error); 2729 2730 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2731 2732 vrele(nd.ni_vp); 2733 return (error); 2734 } 2735 2736 /* 2737 * Set the access and modification times given a file descriptor. 2738 */ 2739 /* ARGSUSED */ 2740 int 2741 sys_futimes(struct lwp *l, void *v, register_t *retval) 2742 { 2743 struct sys_futimes_args /* { 2744 syscallarg(int) fd; 2745 syscallarg(const struct timeval *) tptr; 2746 } */ *uap = v; 2747 struct proc *p = l->l_proc; 2748 int error; 2749 struct file *fp; 2750 2751 /* getvnode() will use the descriptor for us */ 2752 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2753 return (error); 2754 2755 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2756 FILE_UNUSE(fp, l); 2757 return (error); 2758 } 2759 2760 /* 2761 * Set the access and modification times given a path name; this 2762 * version does not follow links. 2763 */ 2764 /* ARGSUSED */ 2765 int 2766 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2767 { 2768 struct sys_lutimes_args /* { 2769 syscallarg(const char *) path; 2770 syscallarg(const struct timeval *) tptr; 2771 } */ *uap = v; 2772 int error; 2773 struct nameidata nd; 2774 2775 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2776 if ((error = namei(&nd)) != 0) 2777 return (error); 2778 2779 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2780 2781 vrele(nd.ni_vp); 2782 return (error); 2783 } 2784 2785 /* 2786 * Common routine to set access and modification times given a vnode. 2787 */ 2788 static int 2789 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2790 { 2791 struct proc *p = l->l_proc; 2792 struct mount *mp; 2793 struct vattr vattr; 2794 int error; 2795 2796 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2797 return (error); 2798 VATTR_NULL(&vattr); 2799 if (tptr == NULL) { 2800 nanotime(&vattr.va_atime); 2801 vattr.va_mtime = vattr.va_atime; 2802 vattr.va_vaflags |= VA_UTIMES_NULL; 2803 } else { 2804 struct timeval tv[2]; 2805 2806 error = copyin(tptr, tv, sizeof(tv)); 2807 if (error) 2808 goto out; 2809 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 2810 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 2811 } 2812 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2813 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2814 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2815 VOP_UNLOCK(vp, 0); 2816 out: 2817 vn_finished_write(mp, 0); 2818 return (error); 2819 } 2820 2821 /* 2822 * Truncate a file given its path name. 2823 */ 2824 /* ARGSUSED */ 2825 int 2826 sys_truncate(struct lwp *l, void *v, register_t *retval) 2827 { 2828 struct sys_truncate_args /* { 2829 syscallarg(const char *) path; 2830 syscallarg(int) pad; 2831 syscallarg(off_t) length; 2832 } */ *uap = v; 2833 struct proc *p = l->l_proc; 2834 struct vnode *vp; 2835 struct mount *mp; 2836 struct vattr vattr; 2837 int error; 2838 struct nameidata nd; 2839 2840 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2841 if ((error = namei(&nd)) != 0) 2842 return (error); 2843 vp = nd.ni_vp; 2844 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2845 vrele(vp); 2846 return (error); 2847 } 2848 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2849 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2850 if (vp->v_type == VDIR) 2851 error = EISDIR; 2852 else if ((error = vn_writechk(vp)) == 0 && 2853 (error = VOP_ACCESS(vp, VWRITE, p->p_cred, l)) == 0) { 2854 VATTR_NULL(&vattr); 2855 vattr.va_size = SCARG(uap, length); 2856 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2857 } 2858 vput(vp); 2859 vn_finished_write(mp, 0); 2860 return (error); 2861 } 2862 2863 /* 2864 * Truncate a file given a file descriptor. 2865 */ 2866 /* ARGSUSED */ 2867 int 2868 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2869 { 2870 struct sys_ftruncate_args /* { 2871 syscallarg(int) fd; 2872 syscallarg(int) pad; 2873 syscallarg(off_t) length; 2874 } */ *uap = v; 2875 struct proc *p = l->l_proc; 2876 struct mount *mp; 2877 struct vattr vattr; 2878 struct vnode *vp; 2879 struct file *fp; 2880 int error; 2881 2882 /* getvnode() will use the descriptor for us */ 2883 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2884 return (error); 2885 if ((fp->f_flag & FWRITE) == 0) { 2886 error = EINVAL; 2887 goto out; 2888 } 2889 vp = (struct vnode *)fp->f_data; 2890 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2891 FILE_UNUSE(fp, l); 2892 return (error); 2893 } 2894 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2895 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2896 if (vp->v_type == VDIR) 2897 error = EISDIR; 2898 else if ((error = vn_writechk(vp)) == 0) { 2899 VATTR_NULL(&vattr); 2900 vattr.va_size = SCARG(uap, length); 2901 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 2902 } 2903 VOP_UNLOCK(vp, 0); 2904 vn_finished_write(mp, 0); 2905 out: 2906 FILE_UNUSE(fp, l); 2907 return (error); 2908 } 2909 2910 /* 2911 * Sync an open file. 2912 */ 2913 /* ARGSUSED */ 2914 int 2915 sys_fsync(struct lwp *l, void *v, register_t *retval) 2916 { 2917 struct sys_fsync_args /* { 2918 syscallarg(int) fd; 2919 } */ *uap = v; 2920 struct proc *p = l->l_proc; 2921 struct vnode *vp; 2922 struct mount *mp; 2923 struct file *fp; 2924 int error; 2925 2926 /* getvnode() will use the descriptor for us */ 2927 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2928 return (error); 2929 vp = (struct vnode *)fp->f_data; 2930 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2931 FILE_UNUSE(fp, l); 2932 return (error); 2933 } 2934 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2935 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 2936 if (error == 0 && bioops.io_fsync != NULL && 2937 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2938 (*bioops.io_fsync)(vp, 0); 2939 VOP_UNLOCK(vp, 0); 2940 vn_finished_write(mp, 0); 2941 FILE_UNUSE(fp, l); 2942 return (error); 2943 } 2944 2945 /* 2946 * Sync a range of file data. API modeled after that found in AIX. 2947 * 2948 * FDATASYNC indicates that we need only save enough metadata to be able 2949 * to re-read the written data. Note we duplicate AIX's requirement that 2950 * the file be open for writing. 2951 */ 2952 /* ARGSUSED */ 2953 int 2954 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2955 { 2956 struct sys_fsync_range_args /* { 2957 syscallarg(int) fd; 2958 syscallarg(int) flags; 2959 syscallarg(off_t) start; 2960 syscallarg(off_t) length; 2961 } */ *uap = v; 2962 struct proc *p = l->l_proc; 2963 struct vnode *vp; 2964 struct file *fp; 2965 int flags, nflags; 2966 off_t s, e, len; 2967 int error; 2968 2969 /* getvnode() will use the descriptor for us */ 2970 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2971 return (error); 2972 2973 if ((fp->f_flag & FWRITE) == 0) { 2974 FILE_UNUSE(fp, l); 2975 return (EBADF); 2976 } 2977 2978 flags = SCARG(uap, flags); 2979 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2980 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2981 return (EINVAL); 2982 } 2983 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2984 if (flags & FDATASYNC) 2985 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2986 else 2987 nflags = FSYNC_WAIT; 2988 if (flags & FDISKSYNC) 2989 nflags |= FSYNC_CACHE; 2990 2991 len = SCARG(uap, length); 2992 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2993 if (len) { 2994 s = SCARG(uap, start); 2995 e = s + len; 2996 if (e < s) { 2997 FILE_UNUSE(fp, l); 2998 return (EINVAL); 2999 } 3000 } else { 3001 e = 0; 3002 s = 0; 3003 } 3004 3005 vp = (struct vnode *)fp->f_data; 3006 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3007 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3008 3009 if (error == 0 && bioops.io_fsync != NULL && 3010 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3011 (*bioops.io_fsync)(vp, nflags); 3012 3013 VOP_UNLOCK(vp, 0); 3014 FILE_UNUSE(fp, l); 3015 return (error); 3016 } 3017 3018 /* 3019 * Sync the data of an open file. 3020 */ 3021 /* ARGSUSED */ 3022 int 3023 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3024 { 3025 struct sys_fdatasync_args /* { 3026 syscallarg(int) fd; 3027 } */ *uap = v; 3028 struct proc *p = l->l_proc; 3029 struct vnode *vp; 3030 struct file *fp; 3031 int error; 3032 3033 /* getvnode() will use the descriptor for us */ 3034 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3035 return (error); 3036 if ((fp->f_flag & FWRITE) == 0) { 3037 FILE_UNUSE(fp, l); 3038 return (EBADF); 3039 } 3040 vp = (struct vnode *)fp->f_data; 3041 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3042 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3043 VOP_UNLOCK(vp, 0); 3044 FILE_UNUSE(fp, l); 3045 return (error); 3046 } 3047 3048 /* 3049 * Rename files, (standard) BSD semantics frontend. 3050 */ 3051 /* ARGSUSED */ 3052 int 3053 sys_rename(struct lwp *l, void *v, register_t *retval) 3054 { 3055 struct sys_rename_args /* { 3056 syscallarg(const char *) from; 3057 syscallarg(const char *) to; 3058 } */ *uap = v; 3059 3060 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3061 } 3062 3063 /* 3064 * Rename files, POSIX semantics frontend. 3065 */ 3066 /* ARGSUSED */ 3067 int 3068 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3069 { 3070 struct sys___posix_rename_args /* { 3071 syscallarg(const char *) from; 3072 syscallarg(const char *) to; 3073 } */ *uap = v; 3074 3075 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3076 } 3077 3078 /* 3079 * Rename files. Source and destination must either both be directories, 3080 * or both not be directories. If target is a directory, it must be empty. 3081 * If `from' and `to' refer to the same object, the value of the `retain' 3082 * argument is used to determine whether `from' will be 3083 * 3084 * (retain == 0) deleted unless `from' and `to' refer to the same 3085 * object in the file system's name space (BSD). 3086 * (retain == 1) always retained (POSIX). 3087 */ 3088 static int 3089 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3090 { 3091 struct mount *mp = NULL; 3092 struct vnode *tvp, *fvp, *tdvp; 3093 struct nameidata fromnd, tond; 3094 struct proc *p; 3095 int error; 3096 3097 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3098 from, l); 3099 if ((error = namei(&fromnd)) != 0) 3100 return (error); 3101 fvp = fromnd.ni_vp; 3102 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3103 if (error != 0) { 3104 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3105 vrele(fromnd.ni_dvp); 3106 vrele(fvp); 3107 if (fromnd.ni_startdir) 3108 vrele(fromnd.ni_startdir); 3109 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3110 return (error); 3111 } 3112 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3113 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3114 if ((error = namei(&tond)) != 0) { 3115 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3116 vrele(fromnd.ni_dvp); 3117 vrele(fvp); 3118 goto out1; 3119 } 3120 tdvp = tond.ni_dvp; 3121 tvp = tond.ni_vp; 3122 3123 if (tvp != NULL) { 3124 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3125 error = ENOTDIR; 3126 goto out; 3127 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3128 error = EISDIR; 3129 goto out; 3130 } 3131 } 3132 3133 if (fvp == tdvp) 3134 error = EINVAL; 3135 3136 /* 3137 * Source and destination refer to the same object. 3138 */ 3139 if (fvp == tvp) { 3140 if (retain) 3141 error = -1; 3142 else if (fromnd.ni_dvp == tdvp && 3143 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3144 !memcmp(fromnd.ni_cnd.cn_nameptr, 3145 tond.ni_cnd.cn_nameptr, 3146 fromnd.ni_cnd.cn_namelen)) 3147 error = -1; 3148 } 3149 3150 #ifdef VERIFIED_EXEC 3151 if (!error) 3152 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l); 3153 #endif /* VERIFIED_EXEC */ 3154 3155 out: 3156 p = l->l_proc; 3157 if (!error) { 3158 VOP_LEASE(tdvp, l, p->p_cred, LEASE_WRITE); 3159 if (fromnd.ni_dvp != tdvp) 3160 VOP_LEASE(fromnd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3161 if (tvp) { 3162 VOP_LEASE(tvp, l, p->p_cred, LEASE_WRITE); 3163 } 3164 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3165 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3166 } else { 3167 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3168 if (tdvp == tvp) 3169 vrele(tdvp); 3170 else 3171 vput(tdvp); 3172 if (tvp) 3173 vput(tvp); 3174 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3175 vrele(fromnd.ni_dvp); 3176 vrele(fvp); 3177 } 3178 vrele(tond.ni_startdir); 3179 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3180 out1: 3181 vn_finished_write(mp, 0); 3182 if (fromnd.ni_startdir) 3183 vrele(fromnd.ni_startdir); 3184 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3185 return (error == -1 ? 0 : error); 3186 } 3187 3188 /* 3189 * Make a directory file. 3190 */ 3191 /* ARGSUSED */ 3192 int 3193 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3194 { 3195 struct sys_mkdir_args /* { 3196 syscallarg(const char *) path; 3197 syscallarg(int) mode; 3198 } */ *uap = v; 3199 struct proc *p = l->l_proc; 3200 struct mount *mp; 3201 struct vnode *vp; 3202 struct vattr vattr; 3203 int error; 3204 struct nameidata nd; 3205 3206 restart: 3207 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3208 SCARG(uap, path), l); 3209 if ((error = namei(&nd)) != 0) 3210 return (error); 3211 vp = nd.ni_vp; 3212 if (vp != NULL) { 3213 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3214 if (nd.ni_dvp == vp) 3215 vrele(nd.ni_dvp); 3216 else 3217 vput(nd.ni_dvp); 3218 vrele(vp); 3219 return (EEXIST); 3220 } 3221 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3222 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3223 if (nd.ni_dvp == vp) 3224 vrele(nd.ni_dvp); 3225 else 3226 vput(nd.ni_dvp); 3227 if ((error = vn_start_write(NULL, &mp, 3228 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3229 return (error); 3230 goto restart; 3231 } 3232 VATTR_NULL(&vattr); 3233 vattr.va_type = VDIR; 3234 vattr.va_mode = 3235 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3236 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3237 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3238 if (!error) 3239 vput(nd.ni_vp); 3240 vn_finished_write(mp, 0); 3241 return (error); 3242 } 3243 3244 /* 3245 * Remove a directory file. 3246 */ 3247 /* ARGSUSED */ 3248 int 3249 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3250 { 3251 struct sys_rmdir_args /* { 3252 syscallarg(const char *) path; 3253 } */ *uap = v; 3254 struct proc *p = l->l_proc; 3255 struct mount *mp; 3256 struct vnode *vp; 3257 int error; 3258 struct nameidata nd; 3259 3260 restart: 3261 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3262 SCARG(uap, path), l); 3263 if ((error = namei(&nd)) != 0) 3264 return (error); 3265 vp = nd.ni_vp; 3266 if (vp->v_type != VDIR) { 3267 error = ENOTDIR; 3268 goto out; 3269 } 3270 /* 3271 * No rmdir "." please. 3272 */ 3273 if (nd.ni_dvp == vp) { 3274 error = EINVAL; 3275 goto out; 3276 } 3277 /* 3278 * The root of a mounted filesystem cannot be deleted. 3279 */ 3280 if (vp->v_flag & VROOT) { 3281 error = EBUSY; 3282 goto out; 3283 } 3284 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3285 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3286 if (nd.ni_dvp == vp) 3287 vrele(nd.ni_dvp); 3288 else 3289 vput(nd.ni_dvp); 3290 vput(vp); 3291 if ((error = vn_start_write(NULL, &mp, 3292 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3293 return (error); 3294 goto restart; 3295 } 3296 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3297 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3298 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3299 vn_finished_write(mp, 0); 3300 return (error); 3301 3302 out: 3303 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3304 if (nd.ni_dvp == vp) 3305 vrele(nd.ni_dvp); 3306 else 3307 vput(nd.ni_dvp); 3308 vput(vp); 3309 return (error); 3310 } 3311 3312 /* 3313 * Read a block of directory entries in a file system independent format. 3314 */ 3315 int 3316 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3317 { 3318 struct sys___getdents30_args /* { 3319 syscallarg(int) fd; 3320 syscallarg(char *) buf; 3321 syscallarg(size_t) count; 3322 } */ *uap = v; 3323 struct proc *p = l->l_proc; 3324 struct file *fp; 3325 int error, done; 3326 3327 /* getvnode() will use the descriptor for us */ 3328 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3329 return (error); 3330 if ((fp->f_flag & FREAD) == 0) { 3331 error = EBADF; 3332 goto out; 3333 } 3334 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3335 SCARG(uap, count), &done, l, 0, 0); 3336 #ifdef KTRACE 3337 if (!error && KTRPOINT(p, KTR_GENIO)) { 3338 struct iovec iov; 3339 iov.iov_base = SCARG(uap, buf); 3340 iov.iov_len = done; 3341 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3342 } 3343 #endif 3344 *retval = done; 3345 out: 3346 FILE_UNUSE(fp, l); 3347 return (error); 3348 } 3349 3350 /* 3351 * Set the mode mask for creation of filesystem nodes. 3352 */ 3353 int 3354 sys_umask(struct lwp *l, void *v, register_t *retval) 3355 { 3356 struct sys_umask_args /* { 3357 syscallarg(mode_t) newmask; 3358 } */ *uap = v; 3359 struct proc *p = l->l_proc; 3360 struct cwdinfo *cwdi; 3361 3362 cwdi = p->p_cwdi; 3363 *retval = cwdi->cwdi_cmask; 3364 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3365 return (0); 3366 } 3367 3368 /* 3369 * Void all references to file by ripping underlying filesystem 3370 * away from vnode. 3371 */ 3372 /* ARGSUSED */ 3373 int 3374 sys_revoke(struct lwp *l, void *v, register_t *retval) 3375 { 3376 struct sys_revoke_args /* { 3377 syscallarg(const char *) path; 3378 } */ *uap = v; 3379 struct proc *p = l->l_proc; 3380 struct mount *mp; 3381 struct vnode *vp; 3382 struct vattr vattr; 3383 int error; 3384 struct nameidata nd; 3385 3386 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3387 if ((error = namei(&nd)) != 0) 3388 return (error); 3389 vp = nd.ni_vp; 3390 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 3391 goto out; 3392 if (kauth_cred_geteuid(p->p_cred) != vattr.va_uid && 3393 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 3394 &p->p_acflag)) != 0) 3395 goto out; 3396 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3397 goto out; 3398 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3399 VOP_REVOKE(vp, REVOKEALL); 3400 vn_finished_write(mp, 0); 3401 out: 3402 vrele(vp); 3403 return (error); 3404 } 3405 3406 /* 3407 * Convert a user file descriptor to a kernel file entry. 3408 */ 3409 int 3410 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3411 { 3412 struct vnode *vp; 3413 struct file *fp; 3414 3415 if ((fp = fd_getfile(fdp, fd)) == NULL) 3416 return (EBADF); 3417 3418 FILE_USE(fp); 3419 3420 if (fp->f_type != DTYPE_VNODE) { 3421 FILE_UNUSE(fp, NULL); 3422 return (EINVAL); 3423 } 3424 3425 vp = (struct vnode *)fp->f_data; 3426 if (vp->v_type == VBAD) { 3427 FILE_UNUSE(fp, NULL); 3428 return (EBADF); 3429 } 3430 3431 *fpp = fp; 3432 return (0); 3433 } 3434
Indexes created Fri Oct 03 02:09:57 GMT 2025