vfs_syscalls.c revision 1.238.4.4
1 /* $NetBSD: vfs_syscalls.c,v 1.238.4.4 2006/05/06 23:31:31 christos 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 1.238.4.4 2006/05/06 23:31:31 christos 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 sbuf; 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 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 798 return error; 799 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 800 } 801 802 /* 803 * Get filesystem statistics. 804 */ 805 /* ARGSUSED */ 806 int 807 sys_fstatvfs1(struct lwp *l, void *v, register_t *retval) 808 { 809 struct sys_fstatvfs1_args /* { 810 syscallarg(int) fd; 811 syscallarg(struct statvfs *) buf; 812 syscallarg(int) flags; 813 } */ *uap = v; 814 struct proc *p = l->l_proc; 815 struct file *fp; 816 struct mount *mp; 817 struct statvfs sbuf; 818 int error; 819 820 /* getvnode() will use the descriptor for us */ 821 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 822 return (error); 823 mp = ((struct vnode *)fp->f_data)->v_mount; 824 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) 825 goto out; 826 error = copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 827 out: 828 FILE_UNUSE(fp, l); 829 return error; 830 } 831 832 833 /* 834 * Get statistics on all filesystems. 835 */ 836 int 837 sys_getvfsstat(struct lwp *l, void *v, register_t *retval) 838 { 839 struct sys_getvfsstat_args /* { 840 syscallarg(struct statvfs *) buf; 841 syscallarg(size_t) bufsize; 842 syscallarg(int) flags; 843 } */ *uap = v; 844 int root = 0; 845 struct proc *p = l->l_proc; 846 struct mount *mp, *nmp; 847 struct statvfs sbuf; 848 struct statvfs *sfsp; 849 size_t count, maxcount; 850 int error = 0; 851 852 maxcount = SCARG(uap, bufsize) / sizeof(struct statvfs); 853 sfsp = SCARG(uap, buf); 854 simple_lock(&mountlist_slock); 855 count = 0; 856 for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist; 857 mp = nmp) { 858 if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock)) { 859 nmp = CIRCLEQ_NEXT(mp, mnt_list); 860 continue; 861 } 862 if (sfsp && count < maxcount) { 863 error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 0); 864 if (error) { 865 simple_lock(&mountlist_slock); 866 nmp = CIRCLEQ_NEXT(mp, mnt_list); 867 vfs_unbusy(mp); 868 continue; 869 } 870 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 871 if (error) { 872 vfs_unbusy(mp); 873 return (error); 874 } 875 sfsp++; 876 root |= strcmp(sbuf.f_mntonname, "/") == 0; 877 } 878 count++; 879 simple_lock(&mountlist_slock); 880 nmp = CIRCLEQ_NEXT(mp, mnt_list); 881 vfs_unbusy(mp); 882 } 883 simple_unlock(&mountlist_slock); 884 if (root == 0 && p->p_cwdi->cwdi_rdir) { 885 /* 886 * fake a root entry 887 */ 888 if ((error = dostatvfs(p->p_cwdi->cwdi_rdir->v_mount, &sbuf, l, 889 SCARG(uap, flags), 1)) != 0) 890 return error; 891 if (sfsp) 892 error = copyout(&sbuf, sfsp, sizeof(*sfsp)); 893 count++; 894 } 895 if (sfsp && count > maxcount) 896 *retval = maxcount; 897 else 898 *retval = count; 899 return error; 900 } 901 902 /* 903 * Change current working directory to a given file descriptor. 904 */ 905 /* ARGSUSED */ 906 int 907 sys_fchdir(struct lwp *l, void *v, register_t *retval) 908 { 909 struct sys_fchdir_args /* { 910 syscallarg(int) fd; 911 } */ *uap = v; 912 struct proc *p = l->l_proc; 913 struct filedesc *fdp = p->p_fd; 914 struct cwdinfo *cwdi = p->p_cwdi; 915 struct vnode *vp, *tdp; 916 struct mount *mp; 917 struct file *fp; 918 int error; 919 920 /* getvnode() will use the descriptor for us */ 921 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 922 return (error); 923 vp = (struct vnode *)fp->f_data; 924 925 VREF(vp); 926 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 927 if (vp->v_type != VDIR) 928 error = ENOTDIR; 929 else 930 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 931 while (!error && (mp = vp->v_mountedhere) != NULL) { 932 if (vfs_busy(mp, 0, 0)) 933 continue; 934 error = VFS_ROOT(mp, &tdp); 935 vfs_unbusy(mp); 936 if (error) 937 break; 938 vput(vp); 939 vp = tdp; 940 } 941 if (error) { 942 vput(vp); 943 goto out; 944 } 945 VOP_UNLOCK(vp, 0); 946 947 /* 948 * Disallow changing to a directory not under the process's 949 * current root directory (if there is one). 950 */ 951 if (cwdi->cwdi_rdir && !vn_isunder(vp, NULL, l)) { 952 vrele(vp); 953 error = EPERM; /* operation not permitted */ 954 goto out; 955 } 956 957 vrele(cwdi->cwdi_cdir); 958 cwdi->cwdi_cdir = vp; 959 out: 960 FILE_UNUSE(fp, l); 961 return (error); 962 } 963 964 /* 965 * Change this process's notion of the root directory to a given file 966 * descriptor. 967 */ 968 int 969 sys_fchroot(struct lwp *l, void *v, register_t *retval) 970 { 971 struct sys_fchroot_args *uap = v; 972 struct proc *p = l->l_proc; 973 struct filedesc *fdp = p->p_fd; 974 struct cwdinfo *cwdi = p->p_cwdi; 975 struct vnode *vp; 976 struct file *fp; 977 int error; 978 979 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 980 &p->p_acflag)) != 0) 981 return error; 982 /* getvnode() will use the descriptor for us */ 983 if ((error = getvnode(fdp, SCARG(uap, fd), &fp)) != 0) 984 return error; 985 vp = (struct vnode *) fp->f_data; 986 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 987 if (vp->v_type != VDIR) 988 error = ENOTDIR; 989 else 990 error = VOP_ACCESS(vp, VEXEC, p->p_cred, l); 991 VOP_UNLOCK(vp, 0); 992 if (error) 993 goto out; 994 VREF(vp); 995 996 /* 997 * Prevent escaping from chroot by putting the root under 998 * the working directory. Silently chdir to / if we aren't 999 * already there. 1000 */ 1001 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1002 /* 1003 * XXX would be more failsafe to change directory to a 1004 * deadfs node here instead 1005 */ 1006 vrele(cwdi->cwdi_cdir); 1007 VREF(vp); 1008 cwdi->cwdi_cdir = vp; 1009 } 1010 1011 if (cwdi->cwdi_rdir != NULL) 1012 vrele(cwdi->cwdi_rdir); 1013 cwdi->cwdi_rdir = vp; 1014 out: 1015 FILE_UNUSE(fp, l); 1016 return (error); 1017 } 1018 1019 /* 1020 * Change current working directory (``.''). 1021 */ 1022 /* ARGSUSED */ 1023 int 1024 sys_chdir(struct lwp *l, void *v, register_t *retval) 1025 { 1026 struct sys_chdir_args /* { 1027 syscallarg(const char *) path; 1028 } */ *uap = v; 1029 struct proc *p = l->l_proc; 1030 struct cwdinfo *cwdi = p->p_cwdi; 1031 int error; 1032 struct nameidata nd; 1033 1034 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1035 SCARG(uap, path), l); 1036 if ((error = change_dir(&nd, l)) != 0) 1037 return (error); 1038 vrele(cwdi->cwdi_cdir); 1039 cwdi->cwdi_cdir = nd.ni_vp; 1040 return (0); 1041 } 1042 1043 /* 1044 * Change notion of root (``/'') directory. 1045 */ 1046 /* ARGSUSED */ 1047 int 1048 sys_chroot(struct lwp *l, void *v, register_t *retval) 1049 { 1050 struct sys_chroot_args /* { 1051 syscallarg(const char *) path; 1052 } */ *uap = v; 1053 struct proc *p = l->l_proc; 1054 struct cwdinfo *cwdi = p->p_cwdi; 1055 struct vnode *vp; 1056 int error; 1057 struct nameidata nd; 1058 1059 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1060 &p->p_acflag)) != 0) 1061 return (error); 1062 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1063 SCARG(uap, path), l); 1064 if ((error = change_dir(&nd, l)) != 0) 1065 return (error); 1066 if (cwdi->cwdi_rdir != NULL) 1067 vrele(cwdi->cwdi_rdir); 1068 vp = nd.ni_vp; 1069 cwdi->cwdi_rdir = vp; 1070 1071 /* 1072 * Prevent escaping from chroot by putting the root under 1073 * the working directory. Silently chdir to / if we aren't 1074 * already there. 1075 */ 1076 if (!vn_isunder(cwdi->cwdi_cdir, vp, l)) { 1077 /* 1078 * XXX would be more failsafe to change directory to a 1079 * deadfs node here instead 1080 */ 1081 vrele(cwdi->cwdi_cdir); 1082 VREF(vp); 1083 cwdi->cwdi_cdir = vp; 1084 } 1085 1086 return (0); 1087 } 1088 1089 /* 1090 * Common routine for chroot and chdir. 1091 */ 1092 static int 1093 change_dir(struct nameidata *ndp, struct lwp *l) 1094 { 1095 struct vnode *vp; 1096 int error; 1097 1098 if ((error = namei(ndp)) != 0) 1099 return (error); 1100 vp = ndp->ni_vp; 1101 if (vp->v_type != VDIR) 1102 error = ENOTDIR; 1103 else 1104 error = VOP_ACCESS(vp, VEXEC, l->l_proc->p_cred, l); 1105 1106 if (error) 1107 vput(vp); 1108 else 1109 VOP_UNLOCK(vp, 0); 1110 return (error); 1111 } 1112 1113 /* 1114 * Check permissions, allocate an open file structure, 1115 * and call the device open routine if any. 1116 */ 1117 int 1118 sys_open(struct lwp *l, void *v, register_t *retval) 1119 { 1120 struct sys_open_args /* { 1121 syscallarg(const char *) path; 1122 syscallarg(int) flags; 1123 syscallarg(int) mode; 1124 } */ *uap = v; 1125 struct proc *p = l->l_proc; 1126 struct cwdinfo *cwdi = p->p_cwdi; 1127 struct filedesc *fdp = p->p_fd; 1128 struct file *fp; 1129 struct vnode *vp; 1130 int flags, cmode; 1131 int type, indx, error; 1132 struct flock lf; 1133 struct nameidata nd; 1134 1135 flags = FFLAGS(SCARG(uap, flags)); 1136 if ((flags & (FREAD | FWRITE)) == 0) 1137 return (EINVAL); 1138 /* falloc() will use the file descriptor for us */ 1139 if ((error = falloc(p, &fp, &indx)) != 0) 1140 return (error); 1141 cmode = ((SCARG(uap, mode) &~ cwdi->cwdi_cmask) & ALLPERMS) &~ S_ISTXT; 1142 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1143 l->l_dupfd = -indx - 1; /* XXX check for fdopen */ 1144 if ((error = vn_open(&nd, flags, cmode)) != 0) { 1145 FILE_UNUSE(fp, l); 1146 fdp->fd_ofiles[indx] = NULL; 1147 ffree(fp); 1148 if ((error == EDUPFD || error == EMOVEFD) && 1149 l->l_dupfd >= 0 && /* XXX from fdopen */ 1150 (error = 1151 dupfdopen(l, indx, l->l_dupfd, flags, error)) == 0) { 1152 *retval = indx; 1153 return (0); 1154 } 1155 if (error == ERESTART) 1156 error = EINTR; 1157 fdremove(fdp, indx); 1158 return (error); 1159 } 1160 l->l_dupfd = 0; 1161 vp = nd.ni_vp; 1162 fp->f_flag = flags & FMASK; 1163 fp->f_type = DTYPE_VNODE; 1164 fp->f_ops = &vnops; 1165 fp->f_data = vp; 1166 if (flags & (O_EXLOCK | O_SHLOCK)) { 1167 lf.l_whence = SEEK_SET; 1168 lf.l_start = 0; 1169 lf.l_len = 0; 1170 if (flags & O_EXLOCK) 1171 lf.l_type = F_WRLCK; 1172 else 1173 lf.l_type = F_RDLCK; 1174 type = F_FLOCK; 1175 if ((flags & FNONBLOCK) == 0) 1176 type |= F_WAIT; 1177 VOP_UNLOCK(vp, 0); 1178 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1179 if (error) { 1180 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1181 FILE_UNUSE(fp, l); 1182 ffree(fp); 1183 fdremove(fdp, indx); 1184 return (error); 1185 } 1186 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1187 fp->f_flag |= FHASLOCK; 1188 } 1189 VOP_UNLOCK(vp, 0); 1190 *retval = indx; 1191 FILE_SET_MATURE(fp); 1192 FILE_UNUSE(fp, l); 1193 return (0); 1194 } 1195 1196 /* 1197 * Get file handle system call 1198 */ 1199 int 1200 sys_getfh(struct lwp *l, void *v, register_t *retval) 1201 { 1202 struct sys_getfh_args /* { 1203 syscallarg(char *) fname; 1204 syscallarg(fhandle_t *) fhp; 1205 } */ *uap = v; 1206 struct proc *p = l->l_proc; 1207 struct vnode *vp; 1208 fhandle_t fh; 1209 int error; 1210 struct nameidata nd; 1211 1212 /* 1213 * Must be super user 1214 */ 1215 error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1216 &p->p_acflag); 1217 if (error) 1218 return (error); 1219 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1220 SCARG(uap, fname), l); 1221 error = namei(&nd); 1222 if (error) 1223 return (error); 1224 vp = nd.ni_vp; 1225 if (vp->v_mount->mnt_op->vfs_vptofh == NULL) 1226 return EOPNOTSUPP; 1227 memset(&fh, 0, sizeof(fh)); 1228 fh.fh_fsid = vp->v_mount->mnt_stat.f_fsidx; 1229 error = VFS_VPTOFH(vp, &fh.fh_fid); 1230 vput(vp); 1231 if (error) 1232 return (error); 1233 KASSERT(fh.fh_fid.fid_len <= _VFS_MAXFIDSZ); 1234 error = copyout(&fh, (caddr_t)SCARG(uap, fhp), sizeof (fh)); 1235 return (error); 1236 } 1237 1238 /* 1239 * Open a file given a file handle. 1240 * 1241 * Check permissions, allocate an open file structure, 1242 * and call the device open routine if any. 1243 */ 1244 int 1245 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1246 { 1247 struct sys_fhopen_args /* { 1248 syscallarg(const fhandle_t *) fhp; 1249 syscallarg(int) flags; 1250 } */ *uap = v; 1251 struct proc *p = l->l_proc; 1252 struct filedesc *fdp = p->p_fd; 1253 struct file *fp; 1254 struct vnode *vp = NULL; 1255 struct mount *mp; 1256 kauth_cred_t cred = p->p_cred; 1257 int flags; 1258 struct file *nfp; 1259 int type, indx, error=0; 1260 struct flock lf; 1261 struct vattr va; 1262 fhandle_t fh; 1263 1264 /* 1265 * Must be super user 1266 */ 1267 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1268 &p->p_acflag))) 1269 return (error); 1270 1271 flags = FFLAGS(SCARG(uap, flags)); 1272 if ((flags & (FREAD | FWRITE)) == 0) 1273 return (EINVAL); 1274 if ((flags & O_CREAT)) 1275 return (EINVAL); 1276 /* falloc() will use the file descriptor for us */ 1277 if ((error = falloc(p, &nfp, &indx)) != 0) 1278 return (error); 1279 fp = nfp; 1280 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1281 goto bad; 1282 1283 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) { 1284 error = ESTALE; 1285 goto bad; 1286 } 1287 1288 if (mp->mnt_op->vfs_fhtovp == NULL) { 1289 error = EOPNOTSUPP; 1290 goto bad; 1291 } 1292 1293 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp)) != 0) { 1294 vp = NULL; /* most likely unnecessary sanity for bad: */ 1295 goto bad; 1296 } 1297 1298 /* Now do an effective vn_open */ 1299 1300 if (vp->v_type == VSOCK) { 1301 error = EOPNOTSUPP; 1302 goto bad; 1303 } 1304 if (flags & FREAD) { 1305 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1306 goto bad; 1307 } 1308 if (flags & (FWRITE | O_TRUNC)) { 1309 if (vp->v_type == VDIR) { 1310 error = EISDIR; 1311 goto bad; 1312 } 1313 if ((error = vn_writechk(vp)) != 0 || 1314 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1315 goto bad; 1316 } 1317 if (flags & O_TRUNC) { 1318 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1319 goto bad; 1320 VOP_UNLOCK(vp, 0); /* XXX */ 1321 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1322 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1323 VATTR_NULL(&va); 1324 va.va_size = 0; 1325 error = VOP_SETATTR(vp, &va, cred, l); 1326 vn_finished_write(mp, 0); 1327 if (error) 1328 goto bad; 1329 } 1330 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1331 goto bad; 1332 if (vp->v_type == VREG && 1333 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1334 error = EIO; 1335 goto bad; 1336 } 1337 if (flags & FWRITE) 1338 vp->v_writecount++; 1339 1340 /* done with modified vn_open, now finish what sys_open does. */ 1341 1342 fp->f_flag = flags & FMASK; 1343 fp->f_type = DTYPE_VNODE; 1344 fp->f_ops = &vnops; 1345 fp->f_data = vp; 1346 if (flags & (O_EXLOCK | O_SHLOCK)) { 1347 lf.l_whence = SEEK_SET; 1348 lf.l_start = 0; 1349 lf.l_len = 0; 1350 if (flags & O_EXLOCK) 1351 lf.l_type = F_WRLCK; 1352 else 1353 lf.l_type = F_RDLCK; 1354 type = F_FLOCK; 1355 if ((flags & FNONBLOCK) == 0) 1356 type |= F_WAIT; 1357 VOP_UNLOCK(vp, 0); 1358 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1359 if (error) { 1360 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1361 FILE_UNUSE(fp, l); 1362 ffree(fp); 1363 fdremove(fdp, indx); 1364 return (error); 1365 } 1366 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1367 fp->f_flag |= FHASLOCK; 1368 } 1369 VOP_UNLOCK(vp, 0); 1370 *retval = indx; 1371 FILE_SET_MATURE(fp); 1372 FILE_UNUSE(fp, l); 1373 return (0); 1374 1375 bad: 1376 FILE_UNUSE(fp, l); 1377 ffree(fp); 1378 fdremove(fdp, indx); 1379 if (vp != NULL) 1380 vput(vp); 1381 return (error); 1382 } 1383 1384 /* ARGSUSED */ 1385 int 1386 sys_fhstat(struct lwp *l, void *v, register_t *retval) 1387 { 1388 struct sys_fhstat_args /* { 1389 syscallarg(const fhandle_t *) fhp; 1390 syscallarg(struct stat *) sb; 1391 } */ *uap = v; 1392 struct proc *p = l->l_proc; 1393 struct stat sb; 1394 int error; 1395 fhandle_t fh; 1396 struct mount *mp; 1397 struct vnode *vp; 1398 1399 /* 1400 * Must be super user 1401 */ 1402 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1403 &p->p_acflag))) 1404 return (error); 1405 1406 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1407 return (error); 1408 1409 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1410 return (ESTALE); 1411 if (mp->mnt_op->vfs_fhtovp == NULL) 1412 return EOPNOTSUPP; 1413 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1414 return (error); 1415 error = vn_stat(vp, &sb, l); 1416 vput(vp); 1417 if (error) 1418 return (error); 1419 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1420 return (error); 1421 } 1422 1423 /* ARGSUSED */ 1424 int 1425 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1426 { 1427 struct sys_fhstatvfs1_args /* { 1428 syscallarg(const fhandle_t *) fhp; 1429 syscallarg(struct statvfs *) buf; 1430 syscallarg(int) flags; 1431 } */ *uap = v; 1432 struct proc *p = l->l_proc; 1433 struct statvfs sbuf; 1434 fhandle_t fh; 1435 struct mount *mp; 1436 struct vnode *vp; 1437 int error; 1438 1439 /* 1440 * Must be super user 1441 */ 1442 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1443 &p->p_acflag)) != 0) 1444 return error; 1445 1446 if ((error = copyin(SCARG(uap, fhp), &fh, sizeof(fhandle_t))) != 0) 1447 return error; 1448 1449 if ((mp = vfs_getvfs(&fh.fh_fsid)) == NULL) 1450 return ESTALE; 1451 if (mp->mnt_op->vfs_fhtovp == NULL) 1452 return EOPNOTSUPP; 1453 if ((error = VFS_FHTOVP(mp, &fh.fh_fid, &vp))) 1454 return error; 1455 1456 mp = vp->v_mount; 1457 if ((error = dostatvfs(mp, &sbuf, l, SCARG(uap, flags), 1)) != 0) { 1458 vput(vp); 1459 return error; 1460 } 1461 vput(vp); 1462 return copyout(&sbuf, SCARG(uap, buf), sizeof(sbuf)); 1463 } 1464 1465 /* 1466 * Create a special file. 1467 */ 1468 /* ARGSUSED */ 1469 int 1470 sys_mknod(struct lwp *l, void *v, register_t *retval) 1471 { 1472 struct sys_mknod_args /* { 1473 syscallarg(const char *) path; 1474 syscallarg(int) mode; 1475 syscallarg(int) dev; 1476 } */ *uap = v; 1477 struct proc *p = l->l_proc; 1478 struct vnode *vp; 1479 struct mount *mp; 1480 struct vattr vattr; 1481 int error; 1482 int whiteout = 0; 1483 struct nameidata nd; 1484 1485 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1486 &p->p_acflag)) != 0) 1487 return (error); 1488 restart: 1489 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1490 if ((error = namei(&nd)) != 0) 1491 return (error); 1492 vp = nd.ni_vp; 1493 if (vp != NULL) 1494 error = EEXIST; 1495 else { 1496 VATTR_NULL(&vattr); 1497 vattr.va_mode = 1498 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1499 vattr.va_rdev = SCARG(uap, dev); 1500 whiteout = 0; 1501 1502 switch (SCARG(uap, mode) & S_IFMT) { 1503 case S_IFMT: /* used by badsect to flag bad sectors */ 1504 vattr.va_type = VBAD; 1505 break; 1506 case S_IFCHR: 1507 vattr.va_type = VCHR; 1508 break; 1509 case S_IFBLK: 1510 vattr.va_type = VBLK; 1511 break; 1512 case S_IFWHT: 1513 whiteout = 1; 1514 break; 1515 default: 1516 error = EINVAL; 1517 break; 1518 } 1519 } 1520 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1521 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1522 if (nd.ni_dvp == vp) 1523 vrele(nd.ni_dvp); 1524 else 1525 vput(nd.ni_dvp); 1526 if (vp) 1527 vrele(vp); 1528 if ((error = vn_start_write(NULL, &mp, 1529 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1530 return (error); 1531 goto restart; 1532 } 1533 if (!error) { 1534 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1535 if (whiteout) { 1536 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1537 if (error) 1538 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1539 vput(nd.ni_dvp); 1540 } else { 1541 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1542 &nd.ni_cnd, &vattr); 1543 if (error == 0) 1544 vput(nd.ni_vp); 1545 } 1546 } else { 1547 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1548 if (nd.ni_dvp == vp) 1549 vrele(nd.ni_dvp); 1550 else 1551 vput(nd.ni_dvp); 1552 if (vp) 1553 vrele(vp); 1554 } 1555 vn_finished_write(mp, 0); 1556 return (error); 1557 } 1558 1559 /* 1560 * Create a named pipe. 1561 */ 1562 /* ARGSUSED */ 1563 int 1564 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1565 { 1566 struct sys_mkfifo_args /* { 1567 syscallarg(const char *) path; 1568 syscallarg(int) mode; 1569 } */ *uap = v; 1570 struct proc *p = l->l_proc; 1571 struct mount *mp; 1572 struct vattr vattr; 1573 int error; 1574 struct nameidata nd; 1575 1576 restart: 1577 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1578 if ((error = namei(&nd)) != 0) 1579 return (error); 1580 if (nd.ni_vp != NULL) { 1581 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1582 if (nd.ni_dvp == nd.ni_vp) 1583 vrele(nd.ni_dvp); 1584 else 1585 vput(nd.ni_dvp); 1586 vrele(nd.ni_vp); 1587 return (EEXIST); 1588 } 1589 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1590 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1591 if (nd.ni_dvp == nd.ni_vp) 1592 vrele(nd.ni_dvp); 1593 else 1594 vput(nd.ni_dvp); 1595 if (nd.ni_vp) 1596 vrele(nd.ni_vp); 1597 if ((error = vn_start_write(NULL, &mp, 1598 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1599 return (error); 1600 goto restart; 1601 } 1602 VATTR_NULL(&vattr); 1603 vattr.va_type = VFIFO; 1604 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1605 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1606 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1607 if (error == 0) 1608 vput(nd.ni_vp); 1609 vn_finished_write(mp, 0); 1610 return (error); 1611 } 1612 1613 /* 1614 * Make a hard file link. 1615 */ 1616 /* ARGSUSED */ 1617 int 1618 sys_link(struct lwp *l, void *v, register_t *retval) 1619 { 1620 struct sys_link_args /* { 1621 syscallarg(const char *) path; 1622 syscallarg(const char *) link; 1623 } */ *uap = v; 1624 struct proc *p = l->l_proc; 1625 struct vnode *vp; 1626 struct mount *mp; 1627 struct nameidata nd; 1628 int error; 1629 1630 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1631 if ((error = namei(&nd)) != 0) 1632 return (error); 1633 vp = nd.ni_vp; 1634 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1635 vrele(vp); 1636 return (error); 1637 } 1638 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1639 if ((error = namei(&nd)) != 0) 1640 goto out; 1641 if (nd.ni_vp) { 1642 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1643 if (nd.ni_dvp == nd.ni_vp) 1644 vrele(nd.ni_dvp); 1645 else 1646 vput(nd.ni_dvp); 1647 vrele(nd.ni_vp); 1648 error = EEXIST; 1649 goto out; 1650 } 1651 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1652 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1653 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1654 out: 1655 vrele(vp); 1656 vn_finished_write(mp, 0); 1657 return (error); 1658 } 1659 1660 /* 1661 * Make a symbolic link. 1662 */ 1663 /* ARGSUSED */ 1664 int 1665 sys_symlink(struct lwp *l, void *v, register_t *retval) 1666 { 1667 struct sys_symlink_args /* { 1668 syscallarg(const char *) path; 1669 syscallarg(const char *) link; 1670 } */ *uap = v; 1671 struct proc *p = l->l_proc; 1672 struct mount *mp; 1673 struct vattr vattr; 1674 char *path; 1675 int error; 1676 struct nameidata nd; 1677 1678 path = PNBUF_GET(); 1679 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1680 if (error) 1681 goto out; 1682 restart: 1683 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1684 if ((error = namei(&nd)) != 0) 1685 goto out; 1686 if (nd.ni_vp) { 1687 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1688 if (nd.ni_dvp == nd.ni_vp) 1689 vrele(nd.ni_dvp); 1690 else 1691 vput(nd.ni_dvp); 1692 vrele(nd.ni_vp); 1693 error = EEXIST; 1694 goto out; 1695 } 1696 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1697 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1698 if (nd.ni_dvp == nd.ni_vp) 1699 vrele(nd.ni_dvp); 1700 else 1701 vput(nd.ni_dvp); 1702 if ((error = vn_start_write(NULL, &mp, 1703 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1704 return (error); 1705 goto restart; 1706 } 1707 VATTR_NULL(&vattr); 1708 vattr.va_type = VLNK; 1709 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1710 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1711 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1712 if (error == 0) 1713 vput(nd.ni_vp); 1714 vn_finished_write(mp, 0); 1715 out: 1716 PNBUF_PUT(path); 1717 return (error); 1718 } 1719 1720 /* 1721 * Delete a whiteout from the filesystem. 1722 */ 1723 /* ARGSUSED */ 1724 int 1725 sys_undelete(struct lwp *l, void *v, register_t *retval) 1726 { 1727 struct sys_undelete_args /* { 1728 syscallarg(const char *) path; 1729 } */ *uap = v; 1730 struct proc *p = l->l_proc; 1731 int error; 1732 struct mount *mp; 1733 struct nameidata nd; 1734 1735 restart: 1736 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1737 SCARG(uap, path), l); 1738 error = namei(&nd); 1739 if (error) 1740 return (error); 1741 1742 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1743 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1744 if (nd.ni_dvp == nd.ni_vp) 1745 vrele(nd.ni_dvp); 1746 else 1747 vput(nd.ni_dvp); 1748 if (nd.ni_vp) 1749 vrele(nd.ni_vp); 1750 return (EEXIST); 1751 } 1752 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1753 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1754 if (nd.ni_dvp == nd.ni_vp) 1755 vrele(nd.ni_dvp); 1756 else 1757 vput(nd.ni_dvp); 1758 if ((error = vn_start_write(NULL, &mp, 1759 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1760 return (error); 1761 goto restart; 1762 } 1763 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1764 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1765 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1766 vput(nd.ni_dvp); 1767 vn_finished_write(mp, 0); 1768 return (error); 1769 } 1770 1771 /* 1772 * Delete a name from the filesystem. 1773 */ 1774 /* ARGSUSED */ 1775 int 1776 sys_unlink(struct lwp *l, void *v, register_t *retval) 1777 { 1778 struct sys_unlink_args /* { 1779 syscallarg(const char *) path; 1780 } */ *uap = v; 1781 struct proc *p = l->l_proc; 1782 struct mount *mp; 1783 struct vnode *vp; 1784 int error; 1785 struct nameidata nd; 1786 1787 restart: 1788 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1789 SCARG(uap, path), l); 1790 if ((error = namei(&nd)) != 0) 1791 return (error); 1792 vp = nd.ni_vp; 1793 1794 /* 1795 * The root of a mounted filesystem cannot be deleted. 1796 */ 1797 if (vp->v_flag & VROOT) { 1798 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1799 if (nd.ni_dvp == vp) 1800 vrele(nd.ni_dvp); 1801 else 1802 vput(nd.ni_dvp); 1803 vput(vp); 1804 error = EBUSY; 1805 goto out; 1806 } 1807 1808 #ifdef VERIFIED_EXEC 1809 /* Handle remove requests for veriexec entries. */ 1810 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) { 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 goto out; 1818 } 1819 #endif /* VERIFIED_EXEC */ 1820 1821 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1822 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1823 if (nd.ni_dvp == vp) 1824 vrele(nd.ni_dvp); 1825 else 1826 vput(nd.ni_dvp); 1827 vput(vp); 1828 if ((error = vn_start_write(NULL, &mp, 1829 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1830 return (error); 1831 goto restart; 1832 } 1833 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1834 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1835 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1836 vn_finished_write(mp, 0); 1837 out: 1838 return (error); 1839 } 1840 1841 /* 1842 * Reposition read/write file offset. 1843 */ 1844 int 1845 sys_lseek(struct lwp *l, void *v, register_t *retval) 1846 { 1847 struct sys_lseek_args /* { 1848 syscallarg(int) fd; 1849 syscallarg(int) pad; 1850 syscallarg(off_t) offset; 1851 syscallarg(int) whence; 1852 } */ *uap = v; 1853 struct proc *p = l->l_proc; 1854 kauth_cred_t cred = p->p_cred; 1855 struct filedesc *fdp = p->p_fd; 1856 struct file *fp; 1857 struct vnode *vp; 1858 struct vattr vattr; 1859 off_t newoff; 1860 int error; 1861 1862 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1863 return (EBADF); 1864 1865 FILE_USE(fp); 1866 1867 vp = (struct vnode *)fp->f_data; 1868 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1869 error = ESPIPE; 1870 goto out; 1871 } 1872 1873 switch (SCARG(uap, whence)) { 1874 case SEEK_CUR: 1875 newoff = fp->f_offset + SCARG(uap, offset); 1876 break; 1877 case SEEK_END: 1878 error = VOP_GETATTR(vp, &vattr, cred, l); 1879 if (error) 1880 goto out; 1881 newoff = SCARG(uap, offset) + vattr.va_size; 1882 break; 1883 case SEEK_SET: 1884 newoff = SCARG(uap, offset); 1885 break; 1886 default: 1887 error = EINVAL; 1888 goto out; 1889 } 1890 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1891 goto out; 1892 1893 *(off_t *)retval = fp->f_offset = newoff; 1894 out: 1895 FILE_UNUSE(fp, l); 1896 return (error); 1897 } 1898 1899 /* 1900 * Positional read system call. 1901 */ 1902 int 1903 sys_pread(struct lwp *l, void *v, register_t *retval) 1904 { 1905 struct sys_pread_args /* { 1906 syscallarg(int) fd; 1907 syscallarg(void *) buf; 1908 syscallarg(size_t) nbyte; 1909 syscallarg(off_t) offset; 1910 } */ *uap = v; 1911 struct proc *p = l->l_proc; 1912 struct filedesc *fdp = p->p_fd; 1913 struct file *fp; 1914 struct vnode *vp; 1915 off_t offset; 1916 int error, fd = SCARG(uap, fd); 1917 1918 if ((fp = fd_getfile(fdp, fd)) == NULL) 1919 return (EBADF); 1920 1921 if ((fp->f_flag & FREAD) == 0) { 1922 simple_unlock(&fp->f_slock); 1923 return (EBADF); 1924 } 1925 1926 FILE_USE(fp); 1927 1928 vp = (struct vnode *)fp->f_data; 1929 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1930 error = ESPIPE; 1931 goto out; 1932 } 1933 1934 offset = SCARG(uap, offset); 1935 1936 /* 1937 * XXX This works because no file systems actually 1938 * XXX take any action on the seek operation. 1939 */ 1940 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1941 goto out; 1942 1943 /* dofileread() will unuse the descriptor for us */ 1944 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 1945 &offset, 0, retval)); 1946 1947 out: 1948 FILE_UNUSE(fp, l); 1949 return (error); 1950 } 1951 1952 /* 1953 * Positional scatter read system call. 1954 */ 1955 int 1956 sys_preadv(struct lwp *l, void *v, register_t *retval) 1957 { 1958 struct sys_preadv_args /* { 1959 syscallarg(int) fd; 1960 syscallarg(const struct iovec *) iovp; 1961 syscallarg(int) iovcnt; 1962 syscallarg(off_t) offset; 1963 } */ *uap = v; 1964 struct proc *p = l->l_proc; 1965 struct filedesc *fdp = p->p_fd; 1966 struct file *fp; 1967 struct vnode *vp; 1968 off_t offset; 1969 int error, fd = SCARG(uap, fd); 1970 1971 if ((fp = fd_getfile(fdp, fd)) == NULL) 1972 return (EBADF); 1973 1974 if ((fp->f_flag & FREAD) == 0) { 1975 simple_unlock(&fp->f_slock); 1976 return (EBADF); 1977 } 1978 1979 FILE_USE(fp); 1980 1981 vp = (struct vnode *)fp->f_data; 1982 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1983 error = ESPIPE; 1984 goto out; 1985 } 1986 1987 offset = SCARG(uap, offset); 1988 1989 /* 1990 * XXX This works because no file systems actually 1991 * XXX take any action on the seek operation. 1992 */ 1993 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 1994 goto out; 1995 1996 /* dofilereadv() will unuse the descriptor for us */ 1997 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 1998 &offset, 0, retval)); 1999 2000 out: 2001 FILE_UNUSE(fp, l); 2002 return (error); 2003 } 2004 2005 /* 2006 * Positional write system call. 2007 */ 2008 int 2009 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2010 { 2011 struct sys_pwrite_args /* { 2012 syscallarg(int) fd; 2013 syscallarg(const void *) buf; 2014 syscallarg(size_t) nbyte; 2015 syscallarg(off_t) offset; 2016 } */ *uap = v; 2017 struct proc *p = l->l_proc; 2018 struct filedesc *fdp = p->p_fd; 2019 struct file *fp; 2020 struct vnode *vp; 2021 off_t offset; 2022 int error, fd = SCARG(uap, fd); 2023 2024 if ((fp = fd_getfile(fdp, fd)) == NULL) 2025 return (EBADF); 2026 2027 if ((fp->f_flag & FWRITE) == 0) { 2028 simple_unlock(&fp->f_slock); 2029 return (EBADF); 2030 } 2031 2032 FILE_USE(fp); 2033 2034 vp = (struct vnode *)fp->f_data; 2035 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2036 error = ESPIPE; 2037 goto out; 2038 } 2039 2040 offset = SCARG(uap, offset); 2041 2042 /* 2043 * XXX This works because no file systems actually 2044 * XXX take any action on the seek operation. 2045 */ 2046 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2047 goto out; 2048 2049 /* dofilewrite() will unuse the descriptor for us */ 2050 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2051 &offset, 0, retval)); 2052 2053 out: 2054 FILE_UNUSE(fp, l); 2055 return (error); 2056 } 2057 2058 /* 2059 * Positional gather write system call. 2060 */ 2061 int 2062 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2063 { 2064 struct sys_pwritev_args /* { 2065 syscallarg(int) fd; 2066 syscallarg(const struct iovec *) iovp; 2067 syscallarg(int) iovcnt; 2068 syscallarg(off_t) offset; 2069 } */ *uap = v; 2070 struct proc *p = l->l_proc; 2071 struct filedesc *fdp = p->p_fd; 2072 struct file *fp; 2073 struct vnode *vp; 2074 off_t offset; 2075 int error, fd = SCARG(uap, fd); 2076 2077 if ((fp = fd_getfile(fdp, fd)) == NULL) 2078 return (EBADF); 2079 2080 if ((fp->f_flag & FWRITE) == 0) { 2081 simple_unlock(&fp->f_slock); 2082 return (EBADF); 2083 } 2084 2085 FILE_USE(fp); 2086 2087 vp = (struct vnode *)fp->f_data; 2088 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2089 error = ESPIPE; 2090 goto out; 2091 } 2092 2093 offset = SCARG(uap, offset); 2094 2095 /* 2096 * XXX This works because no file systems actually 2097 * XXX take any action on the seek operation. 2098 */ 2099 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2100 goto out; 2101 2102 /* dofilewritev() will unuse the descriptor for us */ 2103 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2104 &offset, 0, retval)); 2105 2106 out: 2107 FILE_UNUSE(fp, l); 2108 return (error); 2109 } 2110 2111 /* 2112 * Check access permissions. 2113 */ 2114 int 2115 sys_access(struct lwp *l, void *v, register_t *retval) 2116 { 2117 struct sys_access_args /* { 2118 syscallarg(const char *) path; 2119 syscallarg(int) flags; 2120 } */ *uap = v; 2121 struct proc *p = l->l_proc; 2122 kauth_cred_t cred; 2123 struct vnode *vp; 2124 int error, flags; 2125 struct nameidata nd; 2126 2127 cred = kauth_cred_dup(p->p_cred); 2128 kauth_cred_seteuid(cred, kauth_cred_getuid(p->p_cred)); 2129 kauth_cred_setegid(cred, kauth_cred_getgid(p->p_cred)); 2130 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2131 SCARG(uap, path), l); 2132 /* Override default credentials */ 2133 nd.ni_cnd.cn_cred = cred; 2134 if ((error = namei(&nd)) != 0) 2135 goto out; 2136 vp = nd.ni_vp; 2137 2138 /* Flags == 0 means only check for existence. */ 2139 if (SCARG(uap, flags)) { 2140 flags = 0; 2141 if (SCARG(uap, flags) & R_OK) 2142 flags |= VREAD; 2143 if (SCARG(uap, flags) & W_OK) 2144 flags |= VWRITE; 2145 if (SCARG(uap, flags) & X_OK) 2146 flags |= VEXEC; 2147 2148 error = VOP_ACCESS(vp, flags, cred, l); 2149 if (!error && (flags & VWRITE)) 2150 error = vn_writechk(vp); 2151 } 2152 vput(vp); 2153 out: 2154 kauth_cred_free(cred); 2155 return (error); 2156 } 2157 2158 /* 2159 * Get file status; this version follows links. 2160 */ 2161 /* ARGSUSED */ 2162 int 2163 sys___stat30(struct lwp *l, void *v, register_t *retval) 2164 { 2165 struct sys___stat30_args /* { 2166 syscallarg(const char *) path; 2167 syscallarg(struct stat *) ub; 2168 } */ *uap = v; 2169 struct stat sb; 2170 int error; 2171 struct nameidata nd; 2172 2173 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2174 SCARG(uap, path), l); 2175 if ((error = namei(&nd)) != 0) 2176 return (error); 2177 error = vn_stat(nd.ni_vp, &sb, l); 2178 vput(nd.ni_vp); 2179 if (error) 2180 return (error); 2181 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2182 return (error); 2183 } 2184 2185 /* 2186 * Get file status; this version does not follow links. 2187 */ 2188 /* ARGSUSED */ 2189 int 2190 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2191 { 2192 struct sys___lstat30_args /* { 2193 syscallarg(const char *) path; 2194 syscallarg(struct stat *) ub; 2195 } */ *uap = v; 2196 struct stat sb; 2197 int error; 2198 struct nameidata nd; 2199 2200 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2201 SCARG(uap, path), l); 2202 if ((error = namei(&nd)) != 0) 2203 return (error); 2204 error = vn_stat(nd.ni_vp, &sb, l); 2205 vput(nd.ni_vp); 2206 if (error) 2207 return (error); 2208 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2209 return (error); 2210 } 2211 2212 /* 2213 * Get configurable pathname variables. 2214 */ 2215 /* ARGSUSED */ 2216 int 2217 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2218 { 2219 struct sys_pathconf_args /* { 2220 syscallarg(const char *) path; 2221 syscallarg(int) name; 2222 } */ *uap = v; 2223 int error; 2224 struct nameidata nd; 2225 2226 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2227 SCARG(uap, path), l); 2228 if ((error = namei(&nd)) != 0) 2229 return (error); 2230 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2231 vput(nd.ni_vp); 2232 return (error); 2233 } 2234 2235 /* 2236 * Return target name of a symbolic link. 2237 */ 2238 /* ARGSUSED */ 2239 int 2240 sys_readlink(struct lwp *l, void *v, register_t *retval) 2241 { 2242 struct sys_readlink_args /* { 2243 syscallarg(const char *) path; 2244 syscallarg(char *) buf; 2245 syscallarg(size_t) count; 2246 } */ *uap = v; 2247 struct proc *p = l->l_proc; 2248 struct vnode *vp; 2249 struct iovec aiov; 2250 struct uio auio; 2251 int error; 2252 struct nameidata nd; 2253 2254 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2255 SCARG(uap, path), l); 2256 if ((error = namei(&nd)) != 0) 2257 return (error); 2258 vp = nd.ni_vp; 2259 if (vp->v_type != VLNK) 2260 error = EINVAL; 2261 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2262 (error = VOP_ACCESS(vp, VREAD, p->p_cred, l)) == 0) { 2263 aiov.iov_base = SCARG(uap, buf); 2264 aiov.iov_len = SCARG(uap, count); 2265 auio.uio_iov = &aiov; 2266 auio.uio_iovcnt = 1; 2267 auio.uio_offset = 0; 2268 auio.uio_rw = UIO_READ; 2269 KASSERT(l == curlwp); 2270 auio.uio_vmspace = l->l_proc->p_vmspace; 2271 auio.uio_resid = SCARG(uap, count); 2272 error = VOP_READLINK(vp, &auio, p->p_cred); 2273 } 2274 vput(vp); 2275 *retval = SCARG(uap, count) - auio.uio_resid; 2276 return (error); 2277 } 2278 2279 /* 2280 * Change flags of a file given a path name. 2281 */ 2282 /* ARGSUSED */ 2283 int 2284 sys_chflags(struct lwp *l, void *v, register_t *retval) 2285 { 2286 struct sys_chflags_args /* { 2287 syscallarg(const char *) path; 2288 syscallarg(u_long) flags; 2289 } */ *uap = v; 2290 struct vnode *vp; 2291 int error; 2292 struct nameidata nd; 2293 2294 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2295 if ((error = namei(&nd)) != 0) 2296 return (error); 2297 vp = nd.ni_vp; 2298 error = change_flags(vp, SCARG(uap, flags), l); 2299 vput(vp); 2300 return (error); 2301 } 2302 2303 /* 2304 * Change flags of a file given a file descriptor. 2305 */ 2306 /* ARGSUSED */ 2307 int 2308 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2309 { 2310 struct sys_fchflags_args /* { 2311 syscallarg(int) fd; 2312 syscallarg(u_long) flags; 2313 } */ *uap = v; 2314 struct proc *p = l->l_proc; 2315 struct vnode *vp; 2316 struct file *fp; 2317 int error; 2318 2319 /* getvnode() will use the descriptor for us */ 2320 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2321 return (error); 2322 vp = (struct vnode *)fp->f_data; 2323 error = change_flags(vp, SCARG(uap, flags), l); 2324 VOP_UNLOCK(vp, 0); 2325 FILE_UNUSE(fp, l); 2326 return (error); 2327 } 2328 2329 /* 2330 * Change flags of a file given a path name; this version does 2331 * not follow links. 2332 */ 2333 int 2334 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2335 { 2336 struct sys_lchflags_args /* { 2337 syscallarg(const char *) path; 2338 syscallarg(u_long) flags; 2339 } */ *uap = v; 2340 struct vnode *vp; 2341 int error; 2342 struct nameidata nd; 2343 2344 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2345 if ((error = namei(&nd)) != 0) 2346 return (error); 2347 vp = nd.ni_vp; 2348 error = change_flags(vp, SCARG(uap, flags), l); 2349 vput(vp); 2350 return (error); 2351 } 2352 2353 /* 2354 * Common routine to change flags of a file. 2355 */ 2356 int 2357 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2358 { 2359 struct proc *p = l->l_proc; 2360 struct mount *mp; 2361 struct vattr vattr; 2362 int error; 2363 2364 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2365 return (error); 2366 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2367 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2368 /* 2369 * Non-superusers cannot change the flags on devices, even if they 2370 * own them. 2371 */ 2372 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2373 &p->p_acflag) != 0) { 2374 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2375 goto out; 2376 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2377 error = EINVAL; 2378 goto out; 2379 } 2380 } 2381 VATTR_NULL(&vattr); 2382 vattr.va_flags = flags; 2383 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2384 out: 2385 vn_finished_write(mp, 0); 2386 return (error); 2387 } 2388 2389 /* 2390 * Change mode of a file given path name; this version follows links. 2391 */ 2392 /* ARGSUSED */ 2393 int 2394 sys_chmod(struct lwp *l, void *v, register_t *retval) 2395 { 2396 struct sys_chmod_args /* { 2397 syscallarg(const char *) path; 2398 syscallarg(int) mode; 2399 } */ *uap = v; 2400 int error; 2401 struct nameidata nd; 2402 2403 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2404 if ((error = namei(&nd)) != 0) 2405 return (error); 2406 2407 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2408 2409 vrele(nd.ni_vp); 2410 return (error); 2411 } 2412 2413 /* 2414 * Change mode of a file given a file descriptor. 2415 */ 2416 /* ARGSUSED */ 2417 int 2418 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2419 { 2420 struct sys_fchmod_args /* { 2421 syscallarg(int) fd; 2422 syscallarg(int) mode; 2423 } */ *uap = v; 2424 struct proc *p = l->l_proc; 2425 struct file *fp; 2426 int error; 2427 2428 /* getvnode() will use the descriptor for us */ 2429 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2430 return (error); 2431 2432 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2433 FILE_UNUSE(fp, l); 2434 return (error); 2435 } 2436 2437 /* 2438 * Change mode of a file given path name; this version does not follow links. 2439 */ 2440 /* ARGSUSED */ 2441 int 2442 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2443 { 2444 struct sys_lchmod_args /* { 2445 syscallarg(const char *) path; 2446 syscallarg(int) mode; 2447 } */ *uap = v; 2448 int error; 2449 struct nameidata nd; 2450 2451 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2452 if ((error = namei(&nd)) != 0) 2453 return (error); 2454 2455 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2456 2457 vrele(nd.ni_vp); 2458 return (error); 2459 } 2460 2461 /* 2462 * Common routine to set mode given a vnode. 2463 */ 2464 static int 2465 change_mode(struct vnode *vp, int mode, struct lwp *l) 2466 { 2467 struct proc *p = l->l_proc; 2468 struct mount *mp; 2469 struct vattr vattr; 2470 int error; 2471 2472 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2473 return (error); 2474 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2475 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2476 VATTR_NULL(&vattr); 2477 vattr.va_mode = mode & ALLPERMS; 2478 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2479 VOP_UNLOCK(vp, 0); 2480 vn_finished_write(mp, 0); 2481 return (error); 2482 } 2483 2484 /* 2485 * Set ownership given a path name; this version follows links. 2486 */ 2487 /* ARGSUSED */ 2488 int 2489 sys_chown(struct lwp *l, void *v, register_t *retval) 2490 { 2491 struct sys_chown_args /* { 2492 syscallarg(const char *) path; 2493 syscallarg(uid_t) uid; 2494 syscallarg(gid_t) gid; 2495 } */ *uap = v; 2496 int error; 2497 struct nameidata nd; 2498 2499 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2500 if ((error = namei(&nd)) != 0) 2501 return (error); 2502 2503 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2504 2505 vrele(nd.ni_vp); 2506 return (error); 2507 } 2508 2509 /* 2510 * Set ownership given a path name; this version follows links. 2511 * Provides POSIX semantics. 2512 */ 2513 /* ARGSUSED */ 2514 int 2515 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2516 { 2517 struct sys_chown_args /* { 2518 syscallarg(const char *) path; 2519 syscallarg(uid_t) uid; 2520 syscallarg(gid_t) gid; 2521 } */ *uap = v; 2522 int error; 2523 struct nameidata nd; 2524 2525 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2526 if ((error = namei(&nd)) != 0) 2527 return (error); 2528 2529 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2530 2531 vrele(nd.ni_vp); 2532 return (error); 2533 } 2534 2535 /* 2536 * Set ownership given a file descriptor. 2537 */ 2538 /* ARGSUSED */ 2539 int 2540 sys_fchown(struct lwp *l, void *v, register_t *retval) 2541 { 2542 struct sys_fchown_args /* { 2543 syscallarg(int) fd; 2544 syscallarg(uid_t) uid; 2545 syscallarg(gid_t) gid; 2546 } */ *uap = v; 2547 struct proc *p = l->l_proc; 2548 int error; 2549 struct file *fp; 2550 2551 /* getvnode() will use the descriptor for us */ 2552 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2553 return (error); 2554 2555 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2556 SCARG(uap, gid), l, 0); 2557 FILE_UNUSE(fp, l); 2558 return (error); 2559 } 2560 2561 /* 2562 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2563 */ 2564 /* ARGSUSED */ 2565 int 2566 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2567 { 2568 struct sys_fchown_args /* { 2569 syscallarg(int) fd; 2570 syscallarg(uid_t) uid; 2571 syscallarg(gid_t) gid; 2572 } */ *uap = v; 2573 struct proc *p = l->l_proc; 2574 int error; 2575 struct file *fp; 2576 2577 /* getvnode() will use the descriptor for us */ 2578 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2579 return (error); 2580 2581 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2582 SCARG(uap, gid), l, 1); 2583 FILE_UNUSE(fp, l); 2584 return (error); 2585 } 2586 2587 /* 2588 * Set ownership given a path name; this version does not follow links. 2589 */ 2590 /* ARGSUSED */ 2591 int 2592 sys_lchown(struct lwp *l, void *v, register_t *retval) 2593 { 2594 struct sys_lchown_args /* { 2595 syscallarg(const char *) path; 2596 syscallarg(uid_t) uid; 2597 syscallarg(gid_t) gid; 2598 } */ *uap = v; 2599 int error; 2600 struct nameidata nd; 2601 2602 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2603 if ((error = namei(&nd)) != 0) 2604 return (error); 2605 2606 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2607 2608 vrele(nd.ni_vp); 2609 return (error); 2610 } 2611 2612 /* 2613 * Set ownership given a path name; this version does not follow links. 2614 * Provides POSIX/XPG semantics. 2615 */ 2616 /* ARGSUSED */ 2617 int 2618 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2619 { 2620 struct sys_lchown_args /* { 2621 syscallarg(const char *) path; 2622 syscallarg(uid_t) uid; 2623 syscallarg(gid_t) gid; 2624 } */ *uap = v; 2625 int error; 2626 struct nameidata nd; 2627 2628 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2629 if ((error = namei(&nd)) != 0) 2630 return (error); 2631 2632 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2633 2634 vrele(nd.ni_vp); 2635 return (error); 2636 } 2637 2638 /* 2639 * Common routine to set ownership given a vnode. 2640 */ 2641 static int 2642 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2643 int posix_semantics) 2644 { 2645 struct proc *p = l->l_proc; 2646 struct mount *mp; 2647 struct vattr vattr; 2648 mode_t newmode; 2649 int error; 2650 2651 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2652 return (error); 2653 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2654 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2655 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2656 goto out; 2657 2658 #define CHANGED(x) ((int)(x) != -1) 2659 newmode = vattr.va_mode; 2660 if (posix_semantics) { 2661 /* 2662 * POSIX/XPG semantics: if the caller is not the super-user, 2663 * clear set-user-id and set-group-id bits. Both POSIX and 2664 * the XPG consider the behaviour for calls by the super-user 2665 * implementation-defined; we leave the set-user-id and set- 2666 * group-id settings intact in that case. 2667 */ 2668 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2669 NULL) != 0) 2670 newmode &= ~(S_ISUID | S_ISGID); 2671 } else { 2672 /* 2673 * NetBSD semantics: when changing owner and/or group, 2674 * clear the respective bit(s). 2675 */ 2676 if (CHANGED(uid)) 2677 newmode &= ~S_ISUID; 2678 if (CHANGED(gid)) 2679 newmode &= ~S_ISGID; 2680 } 2681 /* Update va_mode iff altered. */ 2682 if (vattr.va_mode == newmode) 2683 newmode = VNOVAL; 2684 2685 VATTR_NULL(&vattr); 2686 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2687 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2688 vattr.va_mode = newmode; 2689 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2690 #undef CHANGED 2691 2692 out: 2693 VOP_UNLOCK(vp, 0); 2694 vn_finished_write(mp, 0); 2695 return (error); 2696 } 2697 2698 /* 2699 * Set the access and modification times given a path name; this 2700 * version follows links. 2701 */ 2702 /* ARGSUSED */ 2703 int 2704 sys_utimes(struct lwp *l, void *v, register_t *retval) 2705 { 2706 struct sys_utimes_args /* { 2707 syscallarg(const char *) path; 2708 syscallarg(const struct timeval *) tptr; 2709 } */ *uap = v; 2710 int error; 2711 struct nameidata nd; 2712 2713 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2714 if ((error = namei(&nd)) != 0) 2715 return (error); 2716 2717 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2718 2719 vrele(nd.ni_vp); 2720 return (error); 2721 } 2722 2723 /* 2724 * Set the access and modification times given a file descriptor. 2725 */ 2726 /* ARGSUSED */ 2727 int 2728 sys_futimes(struct lwp *l, void *v, register_t *retval) 2729 { 2730 struct sys_futimes_args /* { 2731 syscallarg(int) fd; 2732 syscallarg(const struct timeval *) tptr; 2733 } */ *uap = v; 2734 struct proc *p = l->l_proc; 2735 int error; 2736 struct file *fp; 2737 2738 /* getvnode() will use the descriptor for us */ 2739 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2740 return (error); 2741 2742 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2743 FILE_UNUSE(fp, l); 2744 return (error); 2745 } 2746 2747 /* 2748 * Set the access and modification times given a path name; this 2749 * version does not follow links. 2750 */ 2751 /* ARGSUSED */ 2752 int 2753 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2754 { 2755 struct sys_lutimes_args /* { 2756 syscallarg(const char *) path; 2757 syscallarg(const struct timeval *) tptr; 2758 } */ *uap = v; 2759 int error; 2760 struct nameidata nd; 2761 2762 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2763 if ((error = namei(&nd)) != 0) 2764 return (error); 2765 2766 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2767 2768 vrele(nd.ni_vp); 2769 return (error); 2770 } 2771 2772 /* 2773 * Common routine to set access and modification times given a vnode. 2774 */ 2775 static int 2776 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2777 { 2778 struct proc *p = l->l_proc; 2779 struct mount *mp; 2780 struct vattr vattr; 2781 int error; 2782 2783 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2784 return (error); 2785 VATTR_NULL(&vattr); 2786 if (tptr == NULL) { 2787 nanotime(&vattr.va_atime); 2788 vattr.va_mtime = vattr.va_atime; 2789 vattr.va_vaflags |= VA_UTIMES_NULL; 2790 } else { 2791 struct timeval tv[2]; 2792 2793 error = copyin(tptr, tv, sizeof(tv)); 2794 if (error) 2795 goto out; 2796 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 2797 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 2798 } 2799 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2800 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2801 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2802 VOP_UNLOCK(vp, 0); 2803 out: 2804 vn_finished_write(mp, 0); 2805 return (error); 2806 } 2807 2808 /* 2809 * Truncate a file given its path name. 2810 */ 2811 /* ARGSUSED */ 2812 int 2813 sys_truncate(struct lwp *l, void *v, register_t *retval) 2814 { 2815 struct sys_truncate_args /* { 2816 syscallarg(const char *) path; 2817 syscallarg(int) pad; 2818 syscallarg(off_t) length; 2819 } */ *uap = v; 2820 struct proc *p = l->l_proc; 2821 struct vnode *vp; 2822 struct mount *mp; 2823 struct vattr vattr; 2824 int error; 2825 struct nameidata nd; 2826 2827 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2828 if ((error = namei(&nd)) != 0) 2829 return (error); 2830 vp = nd.ni_vp; 2831 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2832 vrele(vp); 2833 return (error); 2834 } 2835 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2836 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2837 if (vp->v_type == VDIR) 2838 error = EISDIR; 2839 else if ((error = vn_writechk(vp)) == 0 && 2840 (error = VOP_ACCESS(vp, VWRITE, p->p_cred, l)) == 0) { 2841 VATTR_NULL(&vattr); 2842 vattr.va_size = SCARG(uap, length); 2843 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2844 } 2845 vput(vp); 2846 vn_finished_write(mp, 0); 2847 return (error); 2848 } 2849 2850 /* 2851 * Truncate a file given a file descriptor. 2852 */ 2853 /* ARGSUSED */ 2854 int 2855 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2856 { 2857 struct sys_ftruncate_args /* { 2858 syscallarg(int) fd; 2859 syscallarg(int) pad; 2860 syscallarg(off_t) length; 2861 } */ *uap = v; 2862 struct proc *p = l->l_proc; 2863 struct mount *mp; 2864 struct vattr vattr; 2865 struct vnode *vp; 2866 struct file *fp; 2867 int error; 2868 2869 /* getvnode() will use the descriptor for us */ 2870 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2871 return (error); 2872 if ((fp->f_flag & FWRITE) == 0) { 2873 error = EINVAL; 2874 goto out; 2875 } 2876 vp = (struct vnode *)fp->f_data; 2877 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2878 FILE_UNUSE(fp, l); 2879 return (error); 2880 } 2881 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2882 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2883 if (vp->v_type == VDIR) 2884 error = EISDIR; 2885 else if ((error = vn_writechk(vp)) == 0) { 2886 VATTR_NULL(&vattr); 2887 vattr.va_size = SCARG(uap, length); 2888 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 2889 } 2890 VOP_UNLOCK(vp, 0); 2891 vn_finished_write(mp, 0); 2892 out: 2893 FILE_UNUSE(fp, l); 2894 return (error); 2895 } 2896 2897 /* 2898 * Sync an open file. 2899 */ 2900 /* ARGSUSED */ 2901 int 2902 sys_fsync(struct lwp *l, void *v, register_t *retval) 2903 { 2904 struct sys_fsync_args /* { 2905 syscallarg(int) fd; 2906 } */ *uap = v; 2907 struct proc *p = l->l_proc; 2908 struct vnode *vp; 2909 struct mount *mp; 2910 struct file *fp; 2911 int error; 2912 2913 /* getvnode() will use the descriptor for us */ 2914 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2915 return (error); 2916 vp = (struct vnode *)fp->f_data; 2917 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2918 FILE_UNUSE(fp, l); 2919 return (error); 2920 } 2921 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2922 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 2923 if (error == 0 && bioops.io_fsync != NULL && 2924 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2925 (*bioops.io_fsync)(vp, 0); 2926 VOP_UNLOCK(vp, 0); 2927 vn_finished_write(mp, 0); 2928 FILE_UNUSE(fp, l); 2929 return (error); 2930 } 2931 2932 /* 2933 * Sync a range of file data. API modeled after that found in AIX. 2934 * 2935 * FDATASYNC indicates that we need only save enough metadata to be able 2936 * to re-read the written data. Note we duplicate AIX's requirement that 2937 * the file be open for writing. 2938 */ 2939 /* ARGSUSED */ 2940 int 2941 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 2942 { 2943 struct sys_fsync_range_args /* { 2944 syscallarg(int) fd; 2945 syscallarg(int) flags; 2946 syscallarg(off_t) start; 2947 syscallarg(off_t) length; 2948 } */ *uap = v; 2949 struct proc *p = l->l_proc; 2950 struct vnode *vp; 2951 struct file *fp; 2952 int flags, nflags; 2953 off_t s, e, len; 2954 int error; 2955 2956 /* getvnode() will use the descriptor for us */ 2957 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2958 return (error); 2959 2960 if ((fp->f_flag & FWRITE) == 0) { 2961 FILE_UNUSE(fp, l); 2962 return (EBADF); 2963 } 2964 2965 flags = SCARG(uap, flags); 2966 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 2967 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 2968 return (EINVAL); 2969 } 2970 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 2971 if (flags & FDATASYNC) 2972 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 2973 else 2974 nflags = FSYNC_WAIT; 2975 if (flags & FDISKSYNC) 2976 nflags |= FSYNC_CACHE; 2977 2978 len = SCARG(uap, length); 2979 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 2980 if (len) { 2981 s = SCARG(uap, start); 2982 e = s + len; 2983 if (e < s) { 2984 FILE_UNUSE(fp, l); 2985 return (EINVAL); 2986 } 2987 } else { 2988 e = 0; 2989 s = 0; 2990 } 2991 2992 vp = (struct vnode *)fp->f_data; 2993 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2994 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 2995 2996 if (error == 0 && bioops.io_fsync != NULL && 2997 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 2998 (*bioops.io_fsync)(vp, nflags); 2999 3000 VOP_UNLOCK(vp, 0); 3001 FILE_UNUSE(fp, l); 3002 return (error); 3003 } 3004 3005 /* 3006 * Sync the data of an open file. 3007 */ 3008 /* ARGSUSED */ 3009 int 3010 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3011 { 3012 struct sys_fdatasync_args /* { 3013 syscallarg(int) fd; 3014 } */ *uap = v; 3015 struct proc *p = l->l_proc; 3016 struct vnode *vp; 3017 struct file *fp; 3018 int error; 3019 3020 /* getvnode() will use the descriptor for us */ 3021 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3022 return (error); 3023 if ((fp->f_flag & FWRITE) == 0) { 3024 FILE_UNUSE(fp, l); 3025 return (EBADF); 3026 } 3027 vp = (struct vnode *)fp->f_data; 3028 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3029 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3030 VOP_UNLOCK(vp, 0); 3031 FILE_UNUSE(fp, l); 3032 return (error); 3033 } 3034 3035 /* 3036 * Rename files, (standard) BSD semantics frontend. 3037 */ 3038 /* ARGSUSED */ 3039 int 3040 sys_rename(struct lwp *l, void *v, register_t *retval) 3041 { 3042 struct sys_rename_args /* { 3043 syscallarg(const char *) from; 3044 syscallarg(const char *) to; 3045 } */ *uap = v; 3046 3047 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3048 } 3049 3050 /* 3051 * Rename files, POSIX semantics frontend. 3052 */ 3053 /* ARGSUSED */ 3054 int 3055 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3056 { 3057 struct sys___posix_rename_args /* { 3058 syscallarg(const char *) from; 3059 syscallarg(const char *) to; 3060 } */ *uap = v; 3061 3062 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3063 } 3064 3065 /* 3066 * Rename files. Source and destination must either both be directories, 3067 * or both not be directories. If target is a directory, it must be empty. 3068 * If `from' and `to' refer to the same object, the value of the `retain' 3069 * argument is used to determine whether `from' will be 3070 * 3071 * (retain == 0) deleted unless `from' and `to' refer to the same 3072 * object in the file system's name space (BSD). 3073 * (retain == 1) always retained (POSIX). 3074 */ 3075 static int 3076 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3077 { 3078 struct mount *mp = NULL; 3079 struct vnode *tvp, *fvp, *tdvp; 3080 struct nameidata fromnd, tond; 3081 struct proc *p; 3082 int error; 3083 3084 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3085 from, l); 3086 if ((error = namei(&fromnd)) != 0) 3087 return (error); 3088 fvp = fromnd.ni_vp; 3089 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3090 if (error != 0) { 3091 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3092 vrele(fromnd.ni_dvp); 3093 vrele(fvp); 3094 if (fromnd.ni_startdir) 3095 vrele(fromnd.ni_startdir); 3096 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3097 return (error); 3098 } 3099 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3100 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3101 if ((error = namei(&tond)) != 0) { 3102 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3103 vrele(fromnd.ni_dvp); 3104 vrele(fvp); 3105 goto out1; 3106 } 3107 tdvp = tond.ni_dvp; 3108 tvp = tond.ni_vp; 3109 3110 if (tvp != NULL) { 3111 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3112 error = ENOTDIR; 3113 goto out; 3114 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3115 error = EISDIR; 3116 goto out; 3117 } 3118 } 3119 3120 if (fvp == tdvp) 3121 error = EINVAL; 3122 3123 /* 3124 * Source and destination refer to the same object. 3125 */ 3126 if (fvp == tvp) { 3127 if (retain) 3128 error = -1; 3129 else if (fromnd.ni_dvp == tdvp && 3130 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3131 !memcmp(fromnd.ni_cnd.cn_nameptr, 3132 tond.ni_cnd.cn_nameptr, 3133 fromnd.ni_cnd.cn_namelen)) 3134 error = -1; 3135 } 3136 3137 #ifdef VERIFIED_EXEC 3138 if (!error) 3139 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l); 3140 #endif /* VERIFIED_EXEC */ 3141 3142 out: 3143 p = l->l_proc; 3144 if (!error) { 3145 VOP_LEASE(tdvp, l, p->p_cred, LEASE_WRITE); 3146 if (fromnd.ni_dvp != tdvp) 3147 VOP_LEASE(fromnd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3148 if (tvp) { 3149 VOP_LEASE(tvp, l, p->p_cred, LEASE_WRITE); 3150 } 3151 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3152 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3153 } else { 3154 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3155 if (tdvp == tvp) 3156 vrele(tdvp); 3157 else 3158 vput(tdvp); 3159 if (tvp) 3160 vput(tvp); 3161 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3162 vrele(fromnd.ni_dvp); 3163 vrele(fvp); 3164 } 3165 vrele(tond.ni_startdir); 3166 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3167 out1: 3168 vn_finished_write(mp, 0); 3169 if (fromnd.ni_startdir) 3170 vrele(fromnd.ni_startdir); 3171 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3172 return (error == -1 ? 0 : error); 3173 } 3174 3175 /* 3176 * Make a directory file. 3177 */ 3178 /* ARGSUSED */ 3179 int 3180 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3181 { 3182 struct sys_mkdir_args /* { 3183 syscallarg(const char *) path; 3184 syscallarg(int) mode; 3185 } */ *uap = v; 3186 struct proc *p = l->l_proc; 3187 struct mount *mp; 3188 struct vnode *vp; 3189 struct vattr vattr; 3190 int error; 3191 struct nameidata nd; 3192 3193 restart: 3194 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3195 SCARG(uap, path), l); 3196 if ((error = namei(&nd)) != 0) 3197 return (error); 3198 vp = nd.ni_vp; 3199 if (vp != NULL) { 3200 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3201 if (nd.ni_dvp == vp) 3202 vrele(nd.ni_dvp); 3203 else 3204 vput(nd.ni_dvp); 3205 vrele(vp); 3206 return (EEXIST); 3207 } 3208 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3209 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3210 if (nd.ni_dvp == vp) 3211 vrele(nd.ni_dvp); 3212 else 3213 vput(nd.ni_dvp); 3214 if ((error = vn_start_write(NULL, &mp, 3215 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3216 return (error); 3217 goto restart; 3218 } 3219 VATTR_NULL(&vattr); 3220 vattr.va_type = VDIR; 3221 vattr.va_mode = 3222 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3223 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3224 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3225 if (!error) 3226 vput(nd.ni_vp); 3227 vn_finished_write(mp, 0); 3228 return (error); 3229 } 3230 3231 /* 3232 * Remove a directory file. 3233 */ 3234 /* ARGSUSED */ 3235 int 3236 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3237 { 3238 struct sys_rmdir_args /* { 3239 syscallarg(const char *) path; 3240 } */ *uap = v; 3241 struct proc *p = l->l_proc; 3242 struct mount *mp; 3243 struct vnode *vp; 3244 int error; 3245 struct nameidata nd; 3246 3247 restart: 3248 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3249 SCARG(uap, path), l); 3250 if ((error = namei(&nd)) != 0) 3251 return (error); 3252 vp = nd.ni_vp; 3253 if (vp->v_type != VDIR) { 3254 error = ENOTDIR; 3255 goto out; 3256 } 3257 /* 3258 * No rmdir "." please. 3259 */ 3260 if (nd.ni_dvp == vp) { 3261 error = EINVAL; 3262 goto out; 3263 } 3264 /* 3265 * The root of a mounted filesystem cannot be deleted. 3266 */ 3267 if (vp->v_flag & VROOT) { 3268 error = EBUSY; 3269 goto out; 3270 } 3271 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3272 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3273 if (nd.ni_dvp == vp) 3274 vrele(nd.ni_dvp); 3275 else 3276 vput(nd.ni_dvp); 3277 vput(vp); 3278 if ((error = vn_start_write(NULL, &mp, 3279 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3280 return (error); 3281 goto restart; 3282 } 3283 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3284 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3285 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3286 vn_finished_write(mp, 0); 3287 return (error); 3288 3289 out: 3290 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3291 if (nd.ni_dvp == vp) 3292 vrele(nd.ni_dvp); 3293 else 3294 vput(nd.ni_dvp); 3295 vput(vp); 3296 return (error); 3297 } 3298 3299 /* 3300 * Read a block of directory entries in a file system independent format. 3301 */ 3302 int 3303 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3304 { 3305 struct sys___getdents30_args /* { 3306 syscallarg(int) fd; 3307 syscallarg(char *) buf; 3308 syscallarg(size_t) count; 3309 } */ *uap = v; 3310 struct proc *p = l->l_proc; 3311 struct file *fp; 3312 int error, done; 3313 3314 /* getvnode() will use the descriptor for us */ 3315 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3316 return (error); 3317 if ((fp->f_flag & FREAD) == 0) { 3318 error = EBADF; 3319 goto out; 3320 } 3321 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3322 SCARG(uap, count), &done, l, 0, 0); 3323 #ifdef KTRACE 3324 if (!error && KTRPOINT(p, KTR_GENIO)) { 3325 struct iovec iov; 3326 iov.iov_base = SCARG(uap, buf); 3327 iov.iov_len = done; 3328 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3329 } 3330 #endif 3331 *retval = done; 3332 out: 3333 FILE_UNUSE(fp, l); 3334 return (error); 3335 } 3336 3337 /* 3338 * Set the mode mask for creation of filesystem nodes. 3339 */ 3340 int 3341 sys_umask(struct lwp *l, void *v, register_t *retval) 3342 { 3343 struct sys_umask_args /* { 3344 syscallarg(mode_t) newmask; 3345 } */ *uap = v; 3346 struct proc *p = l->l_proc; 3347 struct cwdinfo *cwdi; 3348 3349 cwdi = p->p_cwdi; 3350 *retval = cwdi->cwdi_cmask; 3351 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3352 return (0); 3353 } 3354 3355 /* 3356 * Void all references to file by ripping underlying filesystem 3357 * away from vnode. 3358 */ 3359 /* ARGSUSED */ 3360 int 3361 sys_revoke(struct lwp *l, void *v, register_t *retval) 3362 { 3363 struct sys_revoke_args /* { 3364 syscallarg(const char *) path; 3365 } */ *uap = v; 3366 struct proc *p = l->l_proc; 3367 struct mount *mp; 3368 struct vnode *vp; 3369 struct vattr vattr; 3370 int error; 3371 struct nameidata nd; 3372 3373 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3374 if ((error = namei(&nd)) != 0) 3375 return (error); 3376 vp = nd.ni_vp; 3377 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 3378 goto out; 3379 if (kauth_cred_geteuid(p->p_cred) != vattr.va_uid && 3380 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 3381 &p->p_acflag)) != 0) 3382 goto out; 3383 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3384 goto out; 3385 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3386 VOP_REVOKE(vp, REVOKEALL); 3387 vn_finished_write(mp, 0); 3388 out: 3389 vrele(vp); 3390 return (error); 3391 } 3392 3393 /* 3394 * Convert a user file descriptor to a kernel file entry. 3395 */ 3396 int 3397 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3398 { 3399 struct vnode *vp; 3400 struct file *fp; 3401 3402 if ((fp = fd_getfile(fdp, fd)) == NULL) 3403 return (EBADF); 3404 3405 FILE_USE(fp); 3406 3407 if (fp->f_type != DTYPE_VNODE) { 3408 FILE_UNUSE(fp, NULL); 3409 return (EINVAL); 3410 } 3411 3412 vp = (struct vnode *)fp->f_data; 3413 if (vp->v_type == VBAD) { 3414 FILE_UNUSE(fp, NULL); 3415 return (EBADF); 3416 } 3417 3418 *fpp = fp; 3419 return (0); 3420 } 3421
Indexes created Fri Oct 03 02:09:57 GMT 2025