vfs_syscalls.c revision 1.247
1 /* $NetBSD: vfs_syscalls.c,v 1.247 2006/07/14 14:28:58 yamt 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 1.247 2006/07/14 14:28:58 yamt 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 * vfs_composefh: compose a filehandle. 1207 */ 1208 1209 int 1210 vfs_composefh(struct vnode *vp, fhandle_t *fhp, size_t *fh_size) 1211 { 1212 struct mount *mp; 1213 int error; 1214 size_t sz; 1215 1216 mp = vp->v_mount; 1217 if (mp->mnt_op->vfs_vptofh == NULL) { 1218 return EOPNOTSUPP; 1219 } 1220 if (fhp != NULL && *fh_size >= offsetof(fhandle_t, fh_fid)) { 1221 memset(fhp, 0, *fh_size); 1222 sz = *fh_size - offsetof(fhandle_t, fh_fid); 1223 fhp->fh_fsid = mp->mnt_stat.f_fsidx; 1224 error = VFS_VPTOFH(vp, &fhp->fh_fid, &sz); 1225 } else { 1226 /* just query the size */ 1227 sz = 0; 1228 error = VFS_VPTOFH(vp, NULL, &sz); 1229 } 1230 *fh_size = sz + offsetof(fhandle_t, fh_fid); 1231 return error; 1232 } 1233 1234 /* 1235 * Get file handle system call 1236 */ 1237 int 1238 sys___getfh30(struct lwp *l, void *v, register_t *retval) 1239 { 1240 struct sys___getfh30_args /* { 1241 syscallarg(char *) fname; 1242 syscallarg(fhandle_t *) fhp; 1243 syscallarg(size_t *) fh_size; 1244 } */ *uap = v; 1245 struct proc *p = l->l_proc; 1246 struct vnode *vp; 1247 fhandle_t *fh; 1248 int error; 1249 struct nameidata nd; 1250 size_t sz; 1251 1252 /* 1253 * Must be super user 1254 */ 1255 error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1256 &p->p_acflag); 1257 if (error) 1258 return (error); 1259 fh = NULL; 1260 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 1261 SCARG(uap, fname), l); 1262 error = namei(&nd); 1263 if (error) 1264 return (error); 1265 vp = nd.ni_vp; 1266 error = copyin(SCARG(uap, fh_size), &sz, sizeof(size_t)); 1267 vput(vp); 1268 if (!error) { 1269 fh = malloc(sz, M_TEMP, M_WAITOK); 1270 if (fh == NULL) 1271 return EINVAL; 1272 error = vfs_composefh(vp, fh, &sz); 1273 } 1274 if (error == E2BIG) 1275 copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1276 if (error == 0) { 1277 error = copyout(&sz, SCARG(uap, fh_size), sizeof(size_t)); 1278 if (!error) 1279 error = copyout(fh, SCARG(uap, fhp), sz); 1280 } 1281 free(fh, M_TEMP); 1282 return (error); 1283 } 1284 1285 /* 1286 * Open a file given a file handle. 1287 * 1288 * Check permissions, allocate an open file structure, 1289 * and call the device open routine if any. 1290 */ 1291 int 1292 sys_fhopen(struct lwp *l, void *v, register_t *retval) 1293 { 1294 struct sys_fhopen_args /* { 1295 syscallarg(const fhandle_t *) fhp; 1296 syscallarg(int) flags; 1297 } */ *uap = v; 1298 struct proc *p = l->l_proc; 1299 struct filedesc *fdp = p->p_fd; 1300 struct file *fp; 1301 struct vnode *vp = NULL; 1302 struct mount *mp; 1303 kauth_cred_t cred = p->p_cred; 1304 int flags; 1305 struct file *nfp; 1306 int type, indx, error=0; 1307 struct flock lf; 1308 struct vattr va; 1309 fhandle_t *fh, tempfh; 1310 1311 /* 1312 * Must be super user 1313 */ 1314 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1315 &p->p_acflag))) 1316 return (error); 1317 1318 fh = NULL; 1319 flags = FFLAGS(SCARG(uap, flags)); 1320 if ((flags & (FREAD | FWRITE)) == 0) 1321 return (EINVAL); 1322 if ((flags & O_CREAT)) 1323 return (EINVAL); 1324 /* falloc() will use the file descriptor for us */ 1325 if ((error = falloc(p, &nfp, &indx)) != 0) 1326 return (error); 1327 fp = nfp; 1328 if ((error = copyin(SCARG(uap, fhp), &tempfh, sizeof(tempfh))) != 0) 1329 goto bad; 1330 1331 if ((mp = vfs_getvfs(&tempfh.fh_fsid)) == NULL) { 1332 error = ESTALE; 1333 goto bad; 1334 } 1335 1336 if (mp->mnt_op->vfs_fhtovp == NULL) { 1337 error = EOPNOTSUPP; 1338 goto bad; 1339 } 1340 fh = malloc(tempfh.fh_fid.fid_len, M_TEMP, M_NOWAIT); 1341 if (fh == NULL) { 1342 error = ENOMEM; 1343 goto bad; 1344 } 1345 if ((error = copyin(SCARG(uap, fhp), fh, tempfh.fh_fid.fid_len)) != 0) 1346 goto bad; 1347 1348 if ((error = VFS_FHTOVP(mp, &fh->fh_fid, &vp)) != 0) { 1349 vp = NULL; /* most likely unnecessary sanity for bad: */ 1350 goto bad; 1351 } 1352 1353 /* Now do an effective vn_open */ 1354 1355 if (vp->v_type == VSOCK) { 1356 error = EOPNOTSUPP; 1357 goto bad; 1358 } 1359 if (flags & FREAD) { 1360 if ((error = VOP_ACCESS(vp, VREAD, cred, l)) != 0) 1361 goto bad; 1362 } 1363 if (flags & (FWRITE | O_TRUNC)) { 1364 if (vp->v_type == VDIR) { 1365 error = EISDIR; 1366 goto bad; 1367 } 1368 if ((error = vn_writechk(vp)) != 0 || 1369 (error = VOP_ACCESS(vp, VWRITE, cred, l)) != 0) 1370 goto bad; 1371 } 1372 if (flags & O_TRUNC) { 1373 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 1374 goto bad; 1375 VOP_UNLOCK(vp, 0); /* XXX */ 1376 VOP_LEASE(vp, l, cred, LEASE_WRITE); 1377 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 1378 VATTR_NULL(&va); 1379 va.va_size = 0; 1380 error = VOP_SETATTR(vp, &va, cred, l); 1381 vn_finished_write(mp, 0); 1382 if (error) 1383 goto bad; 1384 } 1385 if ((error = VOP_OPEN(vp, flags, cred, l)) != 0) 1386 goto bad; 1387 if (vp->v_type == VREG && 1388 uvn_attach(vp, flags & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 1389 error = EIO; 1390 goto bad; 1391 } 1392 if (flags & FWRITE) 1393 vp->v_writecount++; 1394 1395 /* done with modified vn_open, now finish what sys_open does. */ 1396 1397 fp->f_flag = flags & FMASK; 1398 fp->f_type = DTYPE_VNODE; 1399 fp->f_ops = &vnops; 1400 fp->f_data = vp; 1401 if (flags & (O_EXLOCK | O_SHLOCK)) { 1402 lf.l_whence = SEEK_SET; 1403 lf.l_start = 0; 1404 lf.l_len = 0; 1405 if (flags & O_EXLOCK) 1406 lf.l_type = F_WRLCK; 1407 else 1408 lf.l_type = F_RDLCK; 1409 type = F_FLOCK; 1410 if ((flags & FNONBLOCK) == 0) 1411 type |= F_WAIT; 1412 VOP_UNLOCK(vp, 0); 1413 error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, type); 1414 if (error) { 1415 (void) vn_close(vp, fp->f_flag, fp->f_cred, l); 1416 FILE_UNUSE(fp, l); 1417 ffree(fp); 1418 fdremove(fdp, indx); 1419 return (error); 1420 } 1421 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 1422 fp->f_flag |= FHASLOCK; 1423 } 1424 VOP_UNLOCK(vp, 0); 1425 *retval = indx; 1426 FILE_SET_MATURE(fp); 1427 FILE_UNUSE(fp, l); 1428 free(fh, M_TEMP); 1429 return (0); 1430 1431 bad: 1432 if (fh != NULL) 1433 free(fh, M_TEMP); 1434 FILE_UNUSE(fp, l); 1435 ffree(fp); 1436 fdremove(fdp, indx); 1437 if (vp != NULL) 1438 vput(vp); 1439 return (error); 1440 } 1441 1442 /* ARGSUSED */ 1443 int 1444 sys___fhstat30(struct lwp *l, void *v, register_t *retval) 1445 { 1446 struct sys___fhstat30_args /* { 1447 syscallarg(const fhandle_t *) fhp; 1448 syscallarg(struct stat *) sb; 1449 } */ *uap = v; 1450 struct proc *p = l->l_proc; 1451 struct stat sb; 1452 int error; 1453 fhandle_t *fh, tempfh; 1454 struct mount *mp; 1455 struct vnode *vp; 1456 1457 /* 1458 * Must be super user 1459 */ 1460 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1461 &p->p_acflag))) 1462 return (error); 1463 1464 if ((error = copyin(SCARG(uap, fhp), &tempfh, sizeof(tempfh))) != 0) 1465 return (error); 1466 1467 if ((mp = vfs_getvfs(&tempfh.fh_fsid)) == NULL) 1468 return (ESTALE); 1469 if (mp->mnt_op->vfs_fhtovp == NULL) 1470 return EOPNOTSUPP; 1471 fh = malloc(tempfh.fh_fid.fid_len, M_TEMP, M_NOWAIT); 1472 if (fh == NULL) 1473 return ENOMEM; 1474 if ((error = copyin(SCARG(uap, fhp), fh, tempfh.fh_fid.fid_len)) != 0) 1475 return (error); 1476 error = VFS_FHTOVP(mp, &fh->fh_fid, &vp); 1477 free(fh, M_TEMP); 1478 if (error) 1479 return (error); 1480 error = vn_stat(vp, &sb, l); 1481 vput(vp); 1482 if (error) 1483 return (error); 1484 error = copyout(&sb, SCARG(uap, sb), sizeof(sb)); 1485 return (error); 1486 } 1487 1488 /* ARGSUSED */ 1489 int 1490 sys_fhstatvfs1(struct lwp *l, void *v, register_t *retval) 1491 { 1492 struct sys_fhstatvfs1_args /* { 1493 syscallarg(const fhandle_t *) fhp; 1494 syscallarg(struct statvfs *) buf; 1495 syscallarg(int) flags; 1496 } */ *uap = v; 1497 struct proc *p = l->l_proc; 1498 struct statvfs *sb; 1499 fhandle_t *fh, tempfh; 1500 struct mount *mp; 1501 struct vnode *vp; 1502 int error; 1503 1504 /* 1505 * Must be super user 1506 */ 1507 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1508 &p->p_acflag)) != 0) 1509 return error; 1510 1511 if ((error = copyin(SCARG(uap, fhp), &tempfh, sizeof(tempfh))) != 0) 1512 return error; 1513 1514 if ((mp = vfs_getvfs(&tempfh.fh_fsid)) == NULL) 1515 return ESTALE; 1516 1517 if (mp->mnt_op->vfs_fhtovp == NULL) 1518 return EOPNOTSUPP; 1519 1520 fh = malloc(tempfh.fh_fid.fid_len, M_TEMP, M_NOWAIT); 1521 if (fh == NULL) 1522 return ENOMEM; 1523 if ((error = copyin(SCARG(uap, fhp), fh, tempfh.fh_fid.fid_len)) != 0) 1524 return error; 1525 1526 error = VFS_FHTOVP(mp, &fh->fh_fid, &vp); 1527 free(fh, M_TEMP); 1528 if (error) 1529 return error; 1530 1531 mp = vp->v_mount; 1532 sb = STATVFSBUF_GET(); 1533 if ((error = dostatvfs(mp, sb, l, SCARG(uap, flags), 1)) != 0) { 1534 vput(vp); 1535 goto out; 1536 } 1537 vput(vp); 1538 error = copyout(sb, SCARG(uap, buf), sizeof(*sb)); 1539 out: 1540 STATVFSBUF_PUT(sb); 1541 return error; 1542 } 1543 1544 /* 1545 * Create a special file. 1546 */ 1547 /* ARGSUSED */ 1548 int 1549 sys_mknod(struct lwp *l, void *v, register_t *retval) 1550 { 1551 struct sys_mknod_args /* { 1552 syscallarg(const char *) path; 1553 syscallarg(int) mode; 1554 syscallarg(int) dev; 1555 } */ *uap = v; 1556 struct proc *p = l->l_proc; 1557 struct vnode *vp; 1558 struct mount *mp; 1559 struct vattr vattr; 1560 int error; 1561 int whiteout = 0; 1562 struct nameidata nd; 1563 1564 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 1565 &p->p_acflag)) != 0) 1566 return (error); 1567 restart: 1568 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1569 if ((error = namei(&nd)) != 0) 1570 return (error); 1571 vp = nd.ni_vp; 1572 if (vp != NULL) 1573 error = EEXIST; 1574 else { 1575 VATTR_NULL(&vattr); 1576 vattr.va_mode = 1577 (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1578 vattr.va_rdev = SCARG(uap, dev); 1579 whiteout = 0; 1580 1581 switch (SCARG(uap, mode) & S_IFMT) { 1582 case S_IFMT: /* used by badsect to flag bad sectors */ 1583 vattr.va_type = VBAD; 1584 break; 1585 case S_IFCHR: 1586 vattr.va_type = VCHR; 1587 break; 1588 case S_IFBLK: 1589 vattr.va_type = VBLK; 1590 break; 1591 case S_IFWHT: 1592 whiteout = 1; 1593 break; 1594 default: 1595 error = EINVAL; 1596 break; 1597 } 1598 } 1599 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1600 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1601 if (nd.ni_dvp == vp) 1602 vrele(nd.ni_dvp); 1603 else 1604 vput(nd.ni_dvp); 1605 if (vp) 1606 vrele(vp); 1607 if ((error = vn_start_write(NULL, &mp, 1608 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1609 return (error); 1610 goto restart; 1611 } 1612 if (!error) { 1613 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1614 if (whiteout) { 1615 error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE); 1616 if (error) 1617 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1618 vput(nd.ni_dvp); 1619 } else { 1620 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, 1621 &nd.ni_cnd, &vattr); 1622 if (error == 0) 1623 vput(nd.ni_vp); 1624 } 1625 } else { 1626 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1627 if (nd.ni_dvp == vp) 1628 vrele(nd.ni_dvp); 1629 else 1630 vput(nd.ni_dvp); 1631 if (vp) 1632 vrele(vp); 1633 } 1634 vn_finished_write(mp, 0); 1635 return (error); 1636 } 1637 1638 /* 1639 * Create a named pipe. 1640 */ 1641 /* ARGSUSED */ 1642 int 1643 sys_mkfifo(struct lwp *l, void *v, register_t *retval) 1644 { 1645 struct sys_mkfifo_args /* { 1646 syscallarg(const char *) path; 1647 syscallarg(int) mode; 1648 } */ *uap = v; 1649 struct proc *p = l->l_proc; 1650 struct mount *mp; 1651 struct vattr vattr; 1652 int error; 1653 struct nameidata nd; 1654 1655 restart: 1656 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, path), l); 1657 if ((error = namei(&nd)) != 0) 1658 return (error); 1659 if (nd.ni_vp != NULL) { 1660 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1661 if (nd.ni_dvp == nd.ni_vp) 1662 vrele(nd.ni_dvp); 1663 else 1664 vput(nd.ni_dvp); 1665 vrele(nd.ni_vp); 1666 return (EEXIST); 1667 } 1668 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1669 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1670 if (nd.ni_dvp == nd.ni_vp) 1671 vrele(nd.ni_dvp); 1672 else 1673 vput(nd.ni_dvp); 1674 if (nd.ni_vp) 1675 vrele(nd.ni_vp); 1676 if ((error = vn_start_write(NULL, &mp, 1677 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1678 return (error); 1679 goto restart; 1680 } 1681 VATTR_NULL(&vattr); 1682 vattr.va_type = VFIFO; 1683 vattr.va_mode = (SCARG(uap, mode) & ALLPERMS) &~ p->p_cwdi->cwdi_cmask; 1684 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1685 error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 1686 if (error == 0) 1687 vput(nd.ni_vp); 1688 vn_finished_write(mp, 0); 1689 return (error); 1690 } 1691 1692 /* 1693 * Make a hard file link. 1694 */ 1695 /* ARGSUSED */ 1696 int 1697 sys_link(struct lwp *l, void *v, register_t *retval) 1698 { 1699 struct sys_link_args /* { 1700 syscallarg(const char *) path; 1701 syscallarg(const char *) link; 1702 } */ *uap = v; 1703 struct proc *p = l->l_proc; 1704 struct vnode *vp; 1705 struct mount *mp; 1706 struct nameidata nd; 1707 int error; 1708 1709 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 1710 if ((error = namei(&nd)) != 0) 1711 return (error); 1712 vp = nd.ni_vp; 1713 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 1714 vrele(vp); 1715 return (error); 1716 } 1717 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1718 if ((error = namei(&nd)) != 0) 1719 goto out; 1720 if (nd.ni_vp) { 1721 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1722 if (nd.ni_dvp == nd.ni_vp) 1723 vrele(nd.ni_dvp); 1724 else 1725 vput(nd.ni_dvp); 1726 vrele(nd.ni_vp); 1727 error = EEXIST; 1728 goto out; 1729 } 1730 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1731 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1732 error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); 1733 out: 1734 vrele(vp); 1735 vn_finished_write(mp, 0); 1736 return (error); 1737 } 1738 1739 /* 1740 * Make a symbolic link. 1741 */ 1742 /* ARGSUSED */ 1743 int 1744 sys_symlink(struct lwp *l, void *v, register_t *retval) 1745 { 1746 struct sys_symlink_args /* { 1747 syscallarg(const char *) path; 1748 syscallarg(const char *) link; 1749 } */ *uap = v; 1750 struct proc *p = l->l_proc; 1751 struct mount *mp; 1752 struct vattr vattr; 1753 char *path; 1754 int error; 1755 struct nameidata nd; 1756 1757 path = PNBUF_GET(); 1758 error = copyinstr(SCARG(uap, path), path, MAXPATHLEN, NULL); 1759 if (error) 1760 goto out; 1761 restart: 1762 NDINIT(&nd, CREATE, LOCKPARENT, UIO_USERSPACE, SCARG(uap, link), l); 1763 if ((error = namei(&nd)) != 0) 1764 goto out; 1765 if (nd.ni_vp) { 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 vrele(nd.ni_vp); 1772 error = EEXIST; 1773 goto out; 1774 } 1775 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1776 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1777 if (nd.ni_dvp == nd.ni_vp) 1778 vrele(nd.ni_dvp); 1779 else 1780 vput(nd.ni_dvp); 1781 if ((error = vn_start_write(NULL, &mp, 1782 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1783 return (error); 1784 goto restart; 1785 } 1786 VATTR_NULL(&vattr); 1787 vattr.va_type = VLNK; 1788 vattr.va_mode = ACCESSPERMS &~ p->p_cwdi->cwdi_cmask; 1789 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1790 error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, path); 1791 if (error == 0) 1792 vput(nd.ni_vp); 1793 vn_finished_write(mp, 0); 1794 out: 1795 PNBUF_PUT(path); 1796 return (error); 1797 } 1798 1799 /* 1800 * Delete a whiteout from the filesystem. 1801 */ 1802 /* ARGSUSED */ 1803 int 1804 sys_undelete(struct lwp *l, void *v, register_t *retval) 1805 { 1806 struct sys_undelete_args /* { 1807 syscallarg(const char *) path; 1808 } */ *uap = v; 1809 struct proc *p = l->l_proc; 1810 int error; 1811 struct mount *mp; 1812 struct nameidata nd; 1813 1814 restart: 1815 NDINIT(&nd, DELETE, LOCKPARENT|DOWHITEOUT, UIO_USERSPACE, 1816 SCARG(uap, path), l); 1817 error = namei(&nd); 1818 if (error) 1819 return (error); 1820 1821 if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) { 1822 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1823 if (nd.ni_dvp == nd.ni_vp) 1824 vrele(nd.ni_dvp); 1825 else 1826 vput(nd.ni_dvp); 1827 if (nd.ni_vp) 1828 vrele(nd.ni_vp); 1829 return (EEXIST); 1830 } 1831 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1832 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1833 if (nd.ni_dvp == nd.ni_vp) 1834 vrele(nd.ni_dvp); 1835 else 1836 vput(nd.ni_dvp); 1837 if ((error = vn_start_write(NULL, &mp, 1838 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1839 return (error); 1840 goto restart; 1841 } 1842 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1843 if ((error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE)) != 0) 1844 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1845 vput(nd.ni_dvp); 1846 vn_finished_write(mp, 0); 1847 return (error); 1848 } 1849 1850 /* 1851 * Delete a name from the filesystem. 1852 */ 1853 /* ARGSUSED */ 1854 int 1855 sys_unlink(struct lwp *l, void *v, register_t *retval) 1856 { 1857 struct sys_unlink_args /* { 1858 syscallarg(const char *) path; 1859 } */ *uap = v; 1860 struct proc *p = l->l_proc; 1861 struct mount *mp; 1862 struct vnode *vp; 1863 int error; 1864 struct nameidata nd; 1865 1866 restart: 1867 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 1868 SCARG(uap, path), l); 1869 if ((error = namei(&nd)) != 0) 1870 return (error); 1871 vp = nd.ni_vp; 1872 1873 /* 1874 * The root of a mounted filesystem cannot be deleted. 1875 */ 1876 if (vp->v_flag & VROOT) { 1877 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1878 if (nd.ni_dvp == vp) 1879 vrele(nd.ni_dvp); 1880 else 1881 vput(nd.ni_dvp); 1882 vput(vp); 1883 error = EBUSY; 1884 goto out; 1885 } 1886 1887 #ifdef VERIFIED_EXEC 1888 /* Handle remove requests for veriexec entries. */ 1889 if ((error = veriexec_removechk(l, vp, nd.ni_dirp)) != 0) { 1890 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1891 if (nd.ni_dvp == vp) 1892 vrele(nd.ni_dvp); 1893 else 1894 vput(nd.ni_dvp); 1895 vput(vp); 1896 goto out; 1897 } 1898 #endif /* VERIFIED_EXEC */ 1899 1900 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 1901 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 1902 if (nd.ni_dvp == vp) 1903 vrele(nd.ni_dvp); 1904 else 1905 vput(nd.ni_dvp); 1906 vput(vp); 1907 if ((error = vn_start_write(NULL, &mp, 1908 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 1909 return (error); 1910 goto restart; 1911 } 1912 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 1913 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 1914 error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 1915 vn_finished_write(mp, 0); 1916 out: 1917 return (error); 1918 } 1919 1920 /* 1921 * Reposition read/write file offset. 1922 */ 1923 int 1924 sys_lseek(struct lwp *l, void *v, register_t *retval) 1925 { 1926 struct sys_lseek_args /* { 1927 syscallarg(int) fd; 1928 syscallarg(int) pad; 1929 syscallarg(off_t) offset; 1930 syscallarg(int) whence; 1931 } */ *uap = v; 1932 struct proc *p = l->l_proc; 1933 kauth_cred_t cred = p->p_cred; 1934 struct filedesc *fdp = p->p_fd; 1935 struct file *fp; 1936 struct vnode *vp; 1937 struct vattr vattr; 1938 off_t newoff; 1939 int error; 1940 1941 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 1942 return (EBADF); 1943 1944 FILE_USE(fp); 1945 1946 vp = (struct vnode *)fp->f_data; 1947 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 1948 error = ESPIPE; 1949 goto out; 1950 } 1951 1952 switch (SCARG(uap, whence)) { 1953 case SEEK_CUR: 1954 newoff = fp->f_offset + SCARG(uap, offset); 1955 break; 1956 case SEEK_END: 1957 error = VOP_GETATTR(vp, &vattr, cred, l); 1958 if (error) 1959 goto out; 1960 newoff = SCARG(uap, offset) + vattr.va_size; 1961 break; 1962 case SEEK_SET: 1963 newoff = SCARG(uap, offset); 1964 break; 1965 default: 1966 error = EINVAL; 1967 goto out; 1968 } 1969 if ((error = VOP_SEEK(vp, fp->f_offset, newoff, cred)) != 0) 1970 goto out; 1971 1972 *(off_t *)retval = fp->f_offset = newoff; 1973 out: 1974 FILE_UNUSE(fp, l); 1975 return (error); 1976 } 1977 1978 /* 1979 * Positional read system call. 1980 */ 1981 int 1982 sys_pread(struct lwp *l, void *v, register_t *retval) 1983 { 1984 struct sys_pread_args /* { 1985 syscallarg(int) fd; 1986 syscallarg(void *) buf; 1987 syscallarg(size_t) nbyte; 1988 syscallarg(off_t) offset; 1989 } */ *uap = v; 1990 struct proc *p = l->l_proc; 1991 struct filedesc *fdp = p->p_fd; 1992 struct file *fp; 1993 struct vnode *vp; 1994 off_t offset; 1995 int error, fd = SCARG(uap, fd); 1996 1997 if ((fp = fd_getfile(fdp, fd)) == NULL) 1998 return (EBADF); 1999 2000 if ((fp->f_flag & FREAD) == 0) { 2001 simple_unlock(&fp->f_slock); 2002 return (EBADF); 2003 } 2004 2005 FILE_USE(fp); 2006 2007 vp = (struct vnode *)fp->f_data; 2008 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2009 error = ESPIPE; 2010 goto out; 2011 } 2012 2013 offset = SCARG(uap, offset); 2014 2015 /* 2016 * XXX This works because no file systems actually 2017 * XXX take any action on the seek operation. 2018 */ 2019 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2020 goto out; 2021 2022 /* dofileread() will unuse the descriptor for us */ 2023 return (dofileread(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2024 &offset, 0, retval)); 2025 2026 out: 2027 FILE_UNUSE(fp, l); 2028 return (error); 2029 } 2030 2031 /* 2032 * Positional scatter read system call. 2033 */ 2034 int 2035 sys_preadv(struct lwp *l, void *v, register_t *retval) 2036 { 2037 struct sys_preadv_args /* { 2038 syscallarg(int) fd; 2039 syscallarg(const struct iovec *) iovp; 2040 syscallarg(int) iovcnt; 2041 syscallarg(off_t) offset; 2042 } */ *uap = v; 2043 struct proc *p = l->l_proc; 2044 struct filedesc *fdp = p->p_fd; 2045 struct file *fp; 2046 struct vnode *vp; 2047 off_t offset; 2048 int error, fd = SCARG(uap, fd); 2049 2050 if ((fp = fd_getfile(fdp, fd)) == NULL) 2051 return (EBADF); 2052 2053 if ((fp->f_flag & FREAD) == 0) { 2054 simple_unlock(&fp->f_slock); 2055 return (EBADF); 2056 } 2057 2058 FILE_USE(fp); 2059 2060 vp = (struct vnode *)fp->f_data; 2061 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2062 error = ESPIPE; 2063 goto out; 2064 } 2065 2066 offset = SCARG(uap, offset); 2067 2068 /* 2069 * XXX This works because no file systems actually 2070 * XXX take any action on the seek operation. 2071 */ 2072 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2073 goto out; 2074 2075 /* dofilereadv() will unuse the descriptor for us */ 2076 return (dofilereadv(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2077 &offset, 0, retval)); 2078 2079 out: 2080 FILE_UNUSE(fp, l); 2081 return (error); 2082 } 2083 2084 /* 2085 * Positional write system call. 2086 */ 2087 int 2088 sys_pwrite(struct lwp *l, void *v, register_t *retval) 2089 { 2090 struct sys_pwrite_args /* { 2091 syscallarg(int) fd; 2092 syscallarg(const void *) buf; 2093 syscallarg(size_t) nbyte; 2094 syscallarg(off_t) offset; 2095 } */ *uap = v; 2096 struct proc *p = l->l_proc; 2097 struct filedesc *fdp = p->p_fd; 2098 struct file *fp; 2099 struct vnode *vp; 2100 off_t offset; 2101 int error, fd = SCARG(uap, fd); 2102 2103 if ((fp = fd_getfile(fdp, fd)) == NULL) 2104 return (EBADF); 2105 2106 if ((fp->f_flag & FWRITE) == 0) { 2107 simple_unlock(&fp->f_slock); 2108 return (EBADF); 2109 } 2110 2111 FILE_USE(fp); 2112 2113 vp = (struct vnode *)fp->f_data; 2114 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2115 error = ESPIPE; 2116 goto out; 2117 } 2118 2119 offset = SCARG(uap, offset); 2120 2121 /* 2122 * XXX This works because no file systems actually 2123 * XXX take any action on the seek operation. 2124 */ 2125 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2126 goto out; 2127 2128 /* dofilewrite() will unuse the descriptor for us */ 2129 return (dofilewrite(l, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 2130 &offset, 0, retval)); 2131 2132 out: 2133 FILE_UNUSE(fp, l); 2134 return (error); 2135 } 2136 2137 /* 2138 * Positional gather write system call. 2139 */ 2140 int 2141 sys_pwritev(struct lwp *l, void *v, register_t *retval) 2142 { 2143 struct sys_pwritev_args /* { 2144 syscallarg(int) fd; 2145 syscallarg(const struct iovec *) iovp; 2146 syscallarg(int) iovcnt; 2147 syscallarg(off_t) offset; 2148 } */ *uap = v; 2149 struct proc *p = l->l_proc; 2150 struct filedesc *fdp = p->p_fd; 2151 struct file *fp; 2152 struct vnode *vp; 2153 off_t offset; 2154 int error, fd = SCARG(uap, fd); 2155 2156 if ((fp = fd_getfile(fdp, fd)) == NULL) 2157 return (EBADF); 2158 2159 if ((fp->f_flag & FWRITE) == 0) { 2160 simple_unlock(&fp->f_slock); 2161 return (EBADF); 2162 } 2163 2164 FILE_USE(fp); 2165 2166 vp = (struct vnode *)fp->f_data; 2167 if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) { 2168 error = ESPIPE; 2169 goto out; 2170 } 2171 2172 offset = SCARG(uap, offset); 2173 2174 /* 2175 * XXX This works because no file systems actually 2176 * XXX take any action on the seek operation. 2177 */ 2178 if ((error = VOP_SEEK(vp, fp->f_offset, offset, fp->f_cred)) != 0) 2179 goto out; 2180 2181 /* dofilewritev() will unuse the descriptor for us */ 2182 return (dofilewritev(l, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 2183 &offset, 0, retval)); 2184 2185 out: 2186 FILE_UNUSE(fp, l); 2187 return (error); 2188 } 2189 2190 /* 2191 * Check access permissions. 2192 */ 2193 int 2194 sys_access(struct lwp *l, void *v, register_t *retval) 2195 { 2196 struct sys_access_args /* { 2197 syscallarg(const char *) path; 2198 syscallarg(int) flags; 2199 } */ *uap = v; 2200 struct proc *p = l->l_proc; 2201 kauth_cred_t cred; 2202 struct vnode *vp; 2203 int error, flags; 2204 struct nameidata nd; 2205 2206 cred = kauth_cred_dup(p->p_cred); 2207 kauth_cred_seteuid(cred, kauth_cred_getuid(p->p_cred)); 2208 kauth_cred_setegid(cred, kauth_cred_getgid(p->p_cred)); 2209 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2210 SCARG(uap, path), l); 2211 /* Override default credentials */ 2212 nd.ni_cnd.cn_cred = cred; 2213 if ((error = namei(&nd)) != 0) 2214 goto out; 2215 vp = nd.ni_vp; 2216 2217 /* Flags == 0 means only check for existence. */ 2218 if (SCARG(uap, flags)) { 2219 flags = 0; 2220 if (SCARG(uap, flags) & R_OK) 2221 flags |= VREAD; 2222 if (SCARG(uap, flags) & W_OK) 2223 flags |= VWRITE; 2224 if (SCARG(uap, flags) & X_OK) 2225 flags |= VEXEC; 2226 2227 error = VOP_ACCESS(vp, flags, cred, l); 2228 if (!error && (flags & VWRITE)) 2229 error = vn_writechk(vp); 2230 } 2231 vput(vp); 2232 out: 2233 kauth_cred_free(cred); 2234 return (error); 2235 } 2236 2237 /* 2238 * Get file status; this version follows links. 2239 */ 2240 /* ARGSUSED */ 2241 int 2242 sys___stat30(struct lwp *l, void *v, register_t *retval) 2243 { 2244 struct sys___stat30_args /* { 2245 syscallarg(const char *) path; 2246 syscallarg(struct stat *) ub; 2247 } */ *uap = v; 2248 struct stat sb; 2249 int error; 2250 struct nameidata nd; 2251 2252 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2253 SCARG(uap, path), l); 2254 if ((error = namei(&nd)) != 0) 2255 return (error); 2256 error = vn_stat(nd.ni_vp, &sb, l); 2257 vput(nd.ni_vp); 2258 if (error) 2259 return (error); 2260 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2261 return (error); 2262 } 2263 2264 /* 2265 * Get file status; this version does not follow links. 2266 */ 2267 /* ARGSUSED */ 2268 int 2269 sys___lstat30(struct lwp *l, void *v, register_t *retval) 2270 { 2271 struct sys___lstat30_args /* { 2272 syscallarg(const char *) path; 2273 syscallarg(struct stat *) ub; 2274 } */ *uap = v; 2275 struct stat sb; 2276 int error; 2277 struct nameidata nd; 2278 2279 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2280 SCARG(uap, path), l); 2281 if ((error = namei(&nd)) != 0) 2282 return (error); 2283 error = vn_stat(nd.ni_vp, &sb, l); 2284 vput(nd.ni_vp); 2285 if (error) 2286 return (error); 2287 error = copyout(&sb, SCARG(uap, ub), sizeof(sb)); 2288 return (error); 2289 } 2290 2291 /* 2292 * Get configurable pathname variables. 2293 */ 2294 /* ARGSUSED */ 2295 int 2296 sys_pathconf(struct lwp *l, void *v, register_t *retval) 2297 { 2298 struct sys_pathconf_args /* { 2299 syscallarg(const char *) path; 2300 syscallarg(int) name; 2301 } */ *uap = v; 2302 int error; 2303 struct nameidata nd; 2304 2305 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, 2306 SCARG(uap, path), l); 2307 if ((error = namei(&nd)) != 0) 2308 return (error); 2309 error = VOP_PATHCONF(nd.ni_vp, SCARG(uap, name), retval); 2310 vput(nd.ni_vp); 2311 return (error); 2312 } 2313 2314 /* 2315 * Return target name of a symbolic link. 2316 */ 2317 /* ARGSUSED */ 2318 int 2319 sys_readlink(struct lwp *l, void *v, register_t *retval) 2320 { 2321 struct sys_readlink_args /* { 2322 syscallarg(const char *) path; 2323 syscallarg(char *) buf; 2324 syscallarg(size_t) count; 2325 } */ *uap = v; 2326 struct proc *p = l->l_proc; 2327 struct vnode *vp; 2328 struct iovec aiov; 2329 struct uio auio; 2330 int error; 2331 struct nameidata nd; 2332 2333 NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF, UIO_USERSPACE, 2334 SCARG(uap, path), l); 2335 if ((error = namei(&nd)) != 0) 2336 return (error); 2337 vp = nd.ni_vp; 2338 if (vp->v_type != VLNK) 2339 error = EINVAL; 2340 else if (!(vp->v_mount->mnt_flag & MNT_SYMPERM) || 2341 (error = VOP_ACCESS(vp, VREAD, p->p_cred, l)) == 0) { 2342 aiov.iov_base = SCARG(uap, buf); 2343 aiov.iov_len = SCARG(uap, count); 2344 auio.uio_iov = &aiov; 2345 auio.uio_iovcnt = 1; 2346 auio.uio_offset = 0; 2347 auio.uio_rw = UIO_READ; 2348 KASSERT(l == curlwp); 2349 auio.uio_vmspace = l->l_proc->p_vmspace; 2350 auio.uio_resid = SCARG(uap, count); 2351 error = VOP_READLINK(vp, &auio, p->p_cred); 2352 } 2353 vput(vp); 2354 *retval = SCARG(uap, count) - auio.uio_resid; 2355 return (error); 2356 } 2357 2358 /* 2359 * Change flags of a file given a path name. 2360 */ 2361 /* ARGSUSED */ 2362 int 2363 sys_chflags(struct lwp *l, void *v, register_t *retval) 2364 { 2365 struct sys_chflags_args /* { 2366 syscallarg(const char *) path; 2367 syscallarg(u_long) flags; 2368 } */ *uap = v; 2369 struct vnode *vp; 2370 int error; 2371 struct nameidata nd; 2372 2373 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2374 if ((error = namei(&nd)) != 0) 2375 return (error); 2376 vp = nd.ni_vp; 2377 error = change_flags(vp, SCARG(uap, flags), l); 2378 vput(vp); 2379 return (error); 2380 } 2381 2382 /* 2383 * Change flags of a file given a file descriptor. 2384 */ 2385 /* ARGSUSED */ 2386 int 2387 sys_fchflags(struct lwp *l, void *v, register_t *retval) 2388 { 2389 struct sys_fchflags_args /* { 2390 syscallarg(int) fd; 2391 syscallarg(u_long) flags; 2392 } */ *uap = v; 2393 struct proc *p = l->l_proc; 2394 struct vnode *vp; 2395 struct file *fp; 2396 int error; 2397 2398 /* getvnode() will use the descriptor for us */ 2399 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2400 return (error); 2401 vp = (struct vnode *)fp->f_data; 2402 error = change_flags(vp, SCARG(uap, flags), l); 2403 VOP_UNLOCK(vp, 0); 2404 FILE_UNUSE(fp, l); 2405 return (error); 2406 } 2407 2408 /* 2409 * Change flags of a file given a path name; this version does 2410 * not follow links. 2411 */ 2412 int 2413 sys_lchflags(struct lwp *l, void *v, register_t *retval) 2414 { 2415 struct sys_lchflags_args /* { 2416 syscallarg(const char *) path; 2417 syscallarg(u_long) flags; 2418 } */ *uap = v; 2419 struct vnode *vp; 2420 int error; 2421 struct nameidata nd; 2422 2423 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2424 if ((error = namei(&nd)) != 0) 2425 return (error); 2426 vp = nd.ni_vp; 2427 error = change_flags(vp, SCARG(uap, flags), l); 2428 vput(vp); 2429 return (error); 2430 } 2431 2432 /* 2433 * Common routine to change flags of a file. 2434 */ 2435 int 2436 change_flags(struct vnode *vp, u_long flags, struct lwp *l) 2437 { 2438 struct proc *p = l->l_proc; 2439 struct mount *mp; 2440 struct vattr vattr; 2441 int error; 2442 2443 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2444 return (error); 2445 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2446 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2447 /* 2448 * Non-superusers cannot change the flags on devices, even if they 2449 * own them. 2450 */ 2451 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2452 &p->p_acflag) != 0) { 2453 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2454 goto out; 2455 if (vattr.va_type == VCHR || vattr.va_type == VBLK) { 2456 error = EINVAL; 2457 goto out; 2458 } 2459 } 2460 VATTR_NULL(&vattr); 2461 vattr.va_flags = flags; 2462 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2463 out: 2464 vn_finished_write(mp, 0); 2465 return (error); 2466 } 2467 2468 /* 2469 * Change mode of a file given path name; this version follows links. 2470 */ 2471 /* ARGSUSED */ 2472 int 2473 sys_chmod(struct lwp *l, void *v, register_t *retval) 2474 { 2475 struct sys_chmod_args /* { 2476 syscallarg(const char *) path; 2477 syscallarg(int) mode; 2478 } */ *uap = v; 2479 int error; 2480 struct nameidata nd; 2481 2482 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2483 if ((error = namei(&nd)) != 0) 2484 return (error); 2485 2486 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2487 2488 vrele(nd.ni_vp); 2489 return (error); 2490 } 2491 2492 /* 2493 * Change mode of a file given a file descriptor. 2494 */ 2495 /* ARGSUSED */ 2496 int 2497 sys_fchmod(struct lwp *l, void *v, register_t *retval) 2498 { 2499 struct sys_fchmod_args /* { 2500 syscallarg(int) fd; 2501 syscallarg(int) mode; 2502 } */ *uap = v; 2503 struct proc *p = l->l_proc; 2504 struct file *fp; 2505 int error; 2506 2507 /* getvnode() will use the descriptor for us */ 2508 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2509 return (error); 2510 2511 error = change_mode((struct vnode *)fp->f_data, SCARG(uap, mode), l); 2512 FILE_UNUSE(fp, l); 2513 return (error); 2514 } 2515 2516 /* 2517 * Change mode of a file given path name; this version does not follow links. 2518 */ 2519 /* ARGSUSED */ 2520 int 2521 sys_lchmod(struct lwp *l, void *v, register_t *retval) 2522 { 2523 struct sys_lchmod_args /* { 2524 syscallarg(const char *) path; 2525 syscallarg(int) mode; 2526 } */ *uap = v; 2527 int error; 2528 struct nameidata nd; 2529 2530 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2531 if ((error = namei(&nd)) != 0) 2532 return (error); 2533 2534 error = change_mode(nd.ni_vp, SCARG(uap, mode), l); 2535 2536 vrele(nd.ni_vp); 2537 return (error); 2538 } 2539 2540 /* 2541 * Common routine to set mode given a vnode. 2542 */ 2543 static int 2544 change_mode(struct vnode *vp, int mode, struct lwp *l) 2545 { 2546 struct proc *p = l->l_proc; 2547 struct mount *mp; 2548 struct vattr vattr; 2549 int error; 2550 2551 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2552 return (error); 2553 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2554 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2555 VATTR_NULL(&vattr); 2556 vattr.va_mode = mode & ALLPERMS; 2557 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2558 VOP_UNLOCK(vp, 0); 2559 vn_finished_write(mp, 0); 2560 return (error); 2561 } 2562 2563 /* 2564 * Set ownership given a path name; this version follows links. 2565 */ 2566 /* ARGSUSED */ 2567 int 2568 sys_chown(struct lwp *l, void *v, register_t *retval) 2569 { 2570 struct sys_chown_args /* { 2571 syscallarg(const char *) path; 2572 syscallarg(uid_t) uid; 2573 syscallarg(gid_t) gid; 2574 } */ *uap = v; 2575 int error; 2576 struct nameidata nd; 2577 2578 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2579 if ((error = namei(&nd)) != 0) 2580 return (error); 2581 2582 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2583 2584 vrele(nd.ni_vp); 2585 return (error); 2586 } 2587 2588 /* 2589 * Set ownership given a path name; this version follows links. 2590 * Provides POSIX semantics. 2591 */ 2592 /* ARGSUSED */ 2593 int 2594 sys___posix_chown(struct lwp *l, void *v, register_t *retval) 2595 { 2596 struct sys_chown_args /* { 2597 syscallarg(const char *) path; 2598 syscallarg(uid_t) uid; 2599 syscallarg(gid_t) gid; 2600 } */ *uap = v; 2601 int error; 2602 struct nameidata nd; 2603 2604 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2605 if ((error = namei(&nd)) != 0) 2606 return (error); 2607 2608 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2609 2610 vrele(nd.ni_vp); 2611 return (error); 2612 } 2613 2614 /* 2615 * Set ownership given a file descriptor. 2616 */ 2617 /* ARGSUSED */ 2618 int 2619 sys_fchown(struct lwp *l, void *v, register_t *retval) 2620 { 2621 struct sys_fchown_args /* { 2622 syscallarg(int) fd; 2623 syscallarg(uid_t) uid; 2624 syscallarg(gid_t) gid; 2625 } */ *uap = v; 2626 struct proc *p = l->l_proc; 2627 int error; 2628 struct file *fp; 2629 2630 /* getvnode() will use the descriptor for us */ 2631 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2632 return (error); 2633 2634 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2635 SCARG(uap, gid), l, 0); 2636 FILE_UNUSE(fp, l); 2637 return (error); 2638 } 2639 2640 /* 2641 * Set ownership given a file descriptor, providing POSIX/XPG semantics. 2642 */ 2643 /* ARGSUSED */ 2644 int 2645 sys___posix_fchown(struct lwp *l, void *v, register_t *retval) 2646 { 2647 struct sys_fchown_args /* { 2648 syscallarg(int) fd; 2649 syscallarg(uid_t) uid; 2650 syscallarg(gid_t) gid; 2651 } */ *uap = v; 2652 struct proc *p = l->l_proc; 2653 int error; 2654 struct file *fp; 2655 2656 /* getvnode() will use the descriptor for us */ 2657 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2658 return (error); 2659 2660 error = change_owner((struct vnode *)fp->f_data, SCARG(uap, uid), 2661 SCARG(uap, gid), l, 1); 2662 FILE_UNUSE(fp, l); 2663 return (error); 2664 } 2665 2666 /* 2667 * Set ownership given a path name; this version does not follow links. 2668 */ 2669 /* ARGSUSED */ 2670 int 2671 sys_lchown(struct lwp *l, void *v, register_t *retval) 2672 { 2673 struct sys_lchown_args /* { 2674 syscallarg(const char *) path; 2675 syscallarg(uid_t) uid; 2676 syscallarg(gid_t) gid; 2677 } */ *uap = v; 2678 int error; 2679 struct nameidata nd; 2680 2681 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2682 if ((error = namei(&nd)) != 0) 2683 return (error); 2684 2685 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 0); 2686 2687 vrele(nd.ni_vp); 2688 return (error); 2689 } 2690 2691 /* 2692 * Set ownership given a path name; this version does not follow links. 2693 * Provides POSIX/XPG semantics. 2694 */ 2695 /* ARGSUSED */ 2696 int 2697 sys___posix_lchown(struct lwp *l, void *v, register_t *retval) 2698 { 2699 struct sys_lchown_args /* { 2700 syscallarg(const char *) path; 2701 syscallarg(uid_t) uid; 2702 syscallarg(gid_t) gid; 2703 } */ *uap = v; 2704 int error; 2705 struct nameidata nd; 2706 2707 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2708 if ((error = namei(&nd)) != 0) 2709 return (error); 2710 2711 error = change_owner(nd.ni_vp, SCARG(uap, uid), SCARG(uap, gid), l, 1); 2712 2713 vrele(nd.ni_vp); 2714 return (error); 2715 } 2716 2717 /* 2718 * Common routine to set ownership given a vnode. 2719 */ 2720 static int 2721 change_owner(struct vnode *vp, uid_t uid, gid_t gid, struct lwp *l, 2722 int posix_semantics) 2723 { 2724 struct proc *p = l->l_proc; 2725 struct mount *mp; 2726 struct vattr vattr; 2727 mode_t newmode; 2728 int error; 2729 2730 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2731 return (error); 2732 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2733 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2734 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 2735 goto out; 2736 2737 #define CHANGED(x) ((int)(x) != -1) 2738 newmode = vattr.va_mode; 2739 if (posix_semantics) { 2740 /* 2741 * POSIX/XPG semantics: if the caller is not the super-user, 2742 * clear set-user-id and set-group-id bits. Both POSIX and 2743 * the XPG consider the behaviour for calls by the super-user 2744 * implementation-defined; we leave the set-user-id and set- 2745 * group-id settings intact in that case. 2746 */ 2747 if (kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 2748 NULL) != 0) 2749 newmode &= ~(S_ISUID | S_ISGID); 2750 } else { 2751 /* 2752 * NetBSD semantics: when changing owner and/or group, 2753 * clear the respective bit(s). 2754 */ 2755 if (CHANGED(uid)) 2756 newmode &= ~S_ISUID; 2757 if (CHANGED(gid)) 2758 newmode &= ~S_ISGID; 2759 } 2760 /* Update va_mode iff altered. */ 2761 if (vattr.va_mode == newmode) 2762 newmode = VNOVAL; 2763 2764 VATTR_NULL(&vattr); 2765 vattr.va_uid = CHANGED(uid) ? uid : (uid_t)VNOVAL; 2766 vattr.va_gid = CHANGED(gid) ? gid : (gid_t)VNOVAL; 2767 vattr.va_mode = newmode; 2768 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2769 #undef CHANGED 2770 2771 out: 2772 VOP_UNLOCK(vp, 0); 2773 vn_finished_write(mp, 0); 2774 return (error); 2775 } 2776 2777 /* 2778 * Set the access and modification times given a path name; this 2779 * version follows links. 2780 */ 2781 /* ARGSUSED */ 2782 int 2783 sys_utimes(struct lwp *l, void *v, register_t *retval) 2784 { 2785 struct sys_utimes_args /* { 2786 syscallarg(const char *) path; 2787 syscallarg(const struct timeval *) tptr; 2788 } */ *uap = v; 2789 int error; 2790 struct nameidata nd; 2791 2792 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2793 if ((error = namei(&nd)) != 0) 2794 return (error); 2795 2796 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2797 2798 vrele(nd.ni_vp); 2799 return (error); 2800 } 2801 2802 /* 2803 * Set the access and modification times given a file descriptor. 2804 */ 2805 /* ARGSUSED */ 2806 int 2807 sys_futimes(struct lwp *l, void *v, register_t *retval) 2808 { 2809 struct sys_futimes_args /* { 2810 syscallarg(int) fd; 2811 syscallarg(const struct timeval *) tptr; 2812 } */ *uap = v; 2813 struct proc *p = l->l_proc; 2814 int error; 2815 struct file *fp; 2816 2817 /* getvnode() will use the descriptor for us */ 2818 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2819 return (error); 2820 2821 error = change_utimes((struct vnode *)fp->f_data, SCARG(uap, tptr), l); 2822 FILE_UNUSE(fp, l); 2823 return (error); 2824 } 2825 2826 /* 2827 * Set the access and modification times given a path name; this 2828 * version does not follow links. 2829 */ 2830 /* ARGSUSED */ 2831 int 2832 sys_lutimes(struct lwp *l, void *v, register_t *retval) 2833 { 2834 struct sys_lutimes_args /* { 2835 syscallarg(const char *) path; 2836 syscallarg(const struct timeval *) tptr; 2837 } */ *uap = v; 2838 int error; 2839 struct nameidata nd; 2840 2841 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2842 if ((error = namei(&nd)) != 0) 2843 return (error); 2844 2845 error = change_utimes(nd.ni_vp, SCARG(uap, tptr), l); 2846 2847 vrele(nd.ni_vp); 2848 return (error); 2849 } 2850 2851 /* 2852 * Common routine to set access and modification times given a vnode. 2853 */ 2854 static int 2855 change_utimes(struct vnode *vp, const struct timeval *tptr, struct lwp *l) 2856 { 2857 struct proc *p = l->l_proc; 2858 struct mount *mp; 2859 struct vattr vattr; 2860 int error; 2861 2862 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 2863 return (error); 2864 VATTR_NULL(&vattr); 2865 if (tptr == NULL) { 2866 nanotime(&vattr.va_atime); 2867 vattr.va_mtime = vattr.va_atime; 2868 vattr.va_vaflags |= VA_UTIMES_NULL; 2869 } else { 2870 struct timeval tv[2]; 2871 2872 error = copyin(tptr, tv, sizeof(tv)); 2873 if (error) 2874 goto out; 2875 TIMEVAL_TO_TIMESPEC(&tv[0], &vattr.va_atime); 2876 TIMEVAL_TO_TIMESPEC(&tv[1], &vattr.va_mtime); 2877 } 2878 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2879 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2880 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2881 VOP_UNLOCK(vp, 0); 2882 out: 2883 vn_finished_write(mp, 0); 2884 return (error); 2885 } 2886 2887 /* 2888 * Truncate a file given its path name. 2889 */ 2890 /* ARGSUSED */ 2891 int 2892 sys_truncate(struct lwp *l, void *v, register_t *retval) 2893 { 2894 struct sys_truncate_args /* { 2895 syscallarg(const char *) path; 2896 syscallarg(int) pad; 2897 syscallarg(off_t) length; 2898 } */ *uap = v; 2899 struct proc *p = l->l_proc; 2900 struct vnode *vp; 2901 struct mount *mp; 2902 struct vattr vattr; 2903 int error; 2904 struct nameidata nd; 2905 2906 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 2907 if ((error = namei(&nd)) != 0) 2908 return (error); 2909 vp = nd.ni_vp; 2910 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2911 vrele(vp); 2912 return (error); 2913 } 2914 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2915 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2916 if (vp->v_type == VDIR) 2917 error = EISDIR; 2918 else if ((error = vn_writechk(vp)) == 0 && 2919 (error = VOP_ACCESS(vp, VWRITE, p->p_cred, l)) == 0) { 2920 VATTR_NULL(&vattr); 2921 vattr.va_size = SCARG(uap, length); 2922 error = VOP_SETATTR(vp, &vattr, p->p_cred, l); 2923 } 2924 vput(vp); 2925 vn_finished_write(mp, 0); 2926 return (error); 2927 } 2928 2929 /* 2930 * Truncate a file given a file descriptor. 2931 */ 2932 /* ARGSUSED */ 2933 int 2934 sys_ftruncate(struct lwp *l, void *v, register_t *retval) 2935 { 2936 struct sys_ftruncate_args /* { 2937 syscallarg(int) fd; 2938 syscallarg(int) pad; 2939 syscallarg(off_t) length; 2940 } */ *uap = v; 2941 struct proc *p = l->l_proc; 2942 struct mount *mp; 2943 struct vattr vattr; 2944 struct vnode *vp; 2945 struct file *fp; 2946 int error; 2947 2948 /* getvnode() will use the descriptor for us */ 2949 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2950 return (error); 2951 if ((fp->f_flag & FWRITE) == 0) { 2952 error = EINVAL; 2953 goto out; 2954 } 2955 vp = (struct vnode *)fp->f_data; 2956 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2957 FILE_UNUSE(fp, l); 2958 return (error); 2959 } 2960 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 2961 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 2962 if (vp->v_type == VDIR) 2963 error = EISDIR; 2964 else if ((error = vn_writechk(vp)) == 0) { 2965 VATTR_NULL(&vattr); 2966 vattr.va_size = SCARG(uap, length); 2967 error = VOP_SETATTR(vp, &vattr, fp->f_cred, l); 2968 } 2969 VOP_UNLOCK(vp, 0); 2970 vn_finished_write(mp, 0); 2971 out: 2972 FILE_UNUSE(fp, l); 2973 return (error); 2974 } 2975 2976 /* 2977 * Sync an open file. 2978 */ 2979 /* ARGSUSED */ 2980 int 2981 sys_fsync(struct lwp *l, void *v, register_t *retval) 2982 { 2983 struct sys_fsync_args /* { 2984 syscallarg(int) fd; 2985 } */ *uap = v; 2986 struct proc *p = l->l_proc; 2987 struct vnode *vp; 2988 struct mount *mp; 2989 struct file *fp; 2990 int error; 2991 2992 /* getvnode() will use the descriptor for us */ 2993 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 2994 return (error); 2995 vp = (struct vnode *)fp->f_data; 2996 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 2997 FILE_UNUSE(fp, l); 2998 return (error); 2999 } 3000 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3001 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT, 0, 0, l); 3002 if (error == 0 && bioops.io_fsync != NULL && 3003 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3004 (*bioops.io_fsync)(vp, 0); 3005 VOP_UNLOCK(vp, 0); 3006 vn_finished_write(mp, 0); 3007 FILE_UNUSE(fp, l); 3008 return (error); 3009 } 3010 3011 /* 3012 * Sync a range of file data. API modeled after that found in AIX. 3013 * 3014 * FDATASYNC indicates that we need only save enough metadata to be able 3015 * to re-read the written data. Note we duplicate AIX's requirement that 3016 * the file be open for writing. 3017 */ 3018 /* ARGSUSED */ 3019 int 3020 sys_fsync_range(struct lwp *l, void *v, register_t *retval) 3021 { 3022 struct sys_fsync_range_args /* { 3023 syscallarg(int) fd; 3024 syscallarg(int) flags; 3025 syscallarg(off_t) start; 3026 syscallarg(off_t) length; 3027 } */ *uap = v; 3028 struct proc *p = l->l_proc; 3029 struct vnode *vp; 3030 struct file *fp; 3031 int flags, nflags; 3032 off_t s, e, len; 3033 int error; 3034 3035 /* getvnode() will use the descriptor for us */ 3036 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3037 return (error); 3038 3039 if ((fp->f_flag & FWRITE) == 0) { 3040 FILE_UNUSE(fp, l); 3041 return (EBADF); 3042 } 3043 3044 flags = SCARG(uap, flags); 3045 if (((flags & (FDATASYNC | FFILESYNC)) == 0) || 3046 ((~flags & (FDATASYNC | FFILESYNC)) == 0)) { 3047 return (EINVAL); 3048 } 3049 /* Now set up the flags for value(s) to pass to VOP_FSYNC() */ 3050 if (flags & FDATASYNC) 3051 nflags = FSYNC_DATAONLY | FSYNC_WAIT; 3052 else 3053 nflags = FSYNC_WAIT; 3054 if (flags & FDISKSYNC) 3055 nflags |= FSYNC_CACHE; 3056 3057 len = SCARG(uap, length); 3058 /* If length == 0, we do the whole file, and s = l = 0 will do that */ 3059 if (len) { 3060 s = SCARG(uap, start); 3061 e = s + len; 3062 if (e < s) { 3063 FILE_UNUSE(fp, l); 3064 return (EINVAL); 3065 } 3066 } else { 3067 e = 0; 3068 s = 0; 3069 } 3070 3071 vp = (struct vnode *)fp->f_data; 3072 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3073 error = VOP_FSYNC(vp, fp->f_cred, nflags, s, e, l); 3074 3075 if (error == 0 && bioops.io_fsync != NULL && 3076 vp->v_mount && (vp->v_mount->mnt_flag & MNT_SOFTDEP)) 3077 (*bioops.io_fsync)(vp, nflags); 3078 3079 VOP_UNLOCK(vp, 0); 3080 FILE_UNUSE(fp, l); 3081 return (error); 3082 } 3083 3084 /* 3085 * Sync the data of an open file. 3086 */ 3087 /* ARGSUSED */ 3088 int 3089 sys_fdatasync(struct lwp *l, void *v, register_t *retval) 3090 { 3091 struct sys_fdatasync_args /* { 3092 syscallarg(int) fd; 3093 } */ *uap = v; 3094 struct proc *p = l->l_proc; 3095 struct vnode *vp; 3096 struct file *fp; 3097 int error; 3098 3099 /* getvnode() will use the descriptor for us */ 3100 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3101 return (error); 3102 if ((fp->f_flag & FWRITE) == 0) { 3103 FILE_UNUSE(fp, l); 3104 return (EBADF); 3105 } 3106 vp = (struct vnode *)fp->f_data; 3107 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 3108 error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0, l); 3109 VOP_UNLOCK(vp, 0); 3110 FILE_UNUSE(fp, l); 3111 return (error); 3112 } 3113 3114 /* 3115 * Rename files, (standard) BSD semantics frontend. 3116 */ 3117 /* ARGSUSED */ 3118 int 3119 sys_rename(struct lwp *l, void *v, register_t *retval) 3120 { 3121 struct sys_rename_args /* { 3122 syscallarg(const char *) from; 3123 syscallarg(const char *) to; 3124 } */ *uap = v; 3125 3126 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 0)); 3127 } 3128 3129 /* 3130 * Rename files, POSIX semantics frontend. 3131 */ 3132 /* ARGSUSED */ 3133 int 3134 sys___posix_rename(struct lwp *l, void *v, register_t *retval) 3135 { 3136 struct sys___posix_rename_args /* { 3137 syscallarg(const char *) from; 3138 syscallarg(const char *) to; 3139 } */ *uap = v; 3140 3141 return (rename_files(SCARG(uap, from), SCARG(uap, to), l, 1)); 3142 } 3143 3144 /* 3145 * Rename files. Source and destination must either both be directories, 3146 * or both not be directories. If target is a directory, it must be empty. 3147 * If `from' and `to' refer to the same object, the value of the `retain' 3148 * argument is used to determine whether `from' will be 3149 * 3150 * (retain == 0) deleted unless `from' and `to' refer to the same 3151 * object in the file system's name space (BSD). 3152 * (retain == 1) always retained (POSIX). 3153 */ 3154 static int 3155 rename_files(const char *from, const char *to, struct lwp *l, int retain) 3156 { 3157 struct mount *mp = NULL; 3158 struct vnode *tvp, *fvp, *tdvp; 3159 struct nameidata fromnd, tond; 3160 struct proc *p; 3161 int error; 3162 3163 NDINIT(&fromnd, DELETE, WANTPARENT | SAVESTART, UIO_USERSPACE, 3164 from, l); 3165 if ((error = namei(&fromnd)) != 0) 3166 return (error); 3167 fvp = fromnd.ni_vp; 3168 error = vn_start_write(fvp, &mp, V_WAIT | V_PCATCH); 3169 if (error != 0) { 3170 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3171 vrele(fromnd.ni_dvp); 3172 vrele(fvp); 3173 if (fromnd.ni_startdir) 3174 vrele(fromnd.ni_startdir); 3175 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3176 return (error); 3177 } 3178 NDINIT(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART | 3179 (fvp->v_type == VDIR ? CREATEDIR : 0), UIO_USERSPACE, to, l); 3180 if ((error = namei(&tond)) != 0) { 3181 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3182 vrele(fromnd.ni_dvp); 3183 vrele(fvp); 3184 goto out1; 3185 } 3186 tdvp = tond.ni_dvp; 3187 tvp = tond.ni_vp; 3188 3189 if (tvp != NULL) { 3190 if (fvp->v_type == VDIR && tvp->v_type != VDIR) { 3191 error = ENOTDIR; 3192 goto out; 3193 } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { 3194 error = EISDIR; 3195 goto out; 3196 } 3197 } 3198 3199 if (fvp == tdvp) 3200 error = EINVAL; 3201 3202 /* 3203 * Source and destination refer to the same object. 3204 */ 3205 if (fvp == tvp) { 3206 if (retain) 3207 error = -1; 3208 else if (fromnd.ni_dvp == tdvp && 3209 fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen && 3210 !memcmp(fromnd.ni_cnd.cn_nameptr, 3211 tond.ni_cnd.cn_nameptr, 3212 fromnd.ni_cnd.cn_namelen)) 3213 error = -1; 3214 } 3215 3216 #ifdef VERIFIED_EXEC 3217 if (!error) 3218 error = veriexec_renamechk(fvp, fromnd.ni_dirp, tond.ni_dirp, l); 3219 #endif /* VERIFIED_EXEC */ 3220 3221 out: 3222 p = l->l_proc; 3223 if (!error) { 3224 VOP_LEASE(tdvp, l, p->p_cred, LEASE_WRITE); 3225 if (fromnd.ni_dvp != tdvp) 3226 VOP_LEASE(fromnd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3227 if (tvp) { 3228 VOP_LEASE(tvp, l, p->p_cred, LEASE_WRITE); 3229 } 3230 error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd, 3231 tond.ni_dvp, tond.ni_vp, &tond.ni_cnd); 3232 } else { 3233 VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd); 3234 if (tdvp == tvp) 3235 vrele(tdvp); 3236 else 3237 vput(tdvp); 3238 if (tvp) 3239 vput(tvp); 3240 VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); 3241 vrele(fromnd.ni_dvp); 3242 vrele(fvp); 3243 } 3244 vrele(tond.ni_startdir); 3245 PNBUF_PUT(tond.ni_cnd.cn_pnbuf); 3246 out1: 3247 vn_finished_write(mp, 0); 3248 if (fromnd.ni_startdir) 3249 vrele(fromnd.ni_startdir); 3250 PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf); 3251 return (error == -1 ? 0 : error); 3252 } 3253 3254 /* 3255 * Make a directory file. 3256 */ 3257 /* ARGSUSED */ 3258 int 3259 sys_mkdir(struct lwp *l, void *v, register_t *retval) 3260 { 3261 struct sys_mkdir_args /* { 3262 syscallarg(const char *) path; 3263 syscallarg(int) mode; 3264 } */ *uap = v; 3265 struct proc *p = l->l_proc; 3266 struct mount *mp; 3267 struct vnode *vp; 3268 struct vattr vattr; 3269 int error; 3270 struct nameidata nd; 3271 3272 restart: 3273 NDINIT(&nd, CREATE, LOCKPARENT | CREATEDIR, UIO_USERSPACE, 3274 SCARG(uap, path), l); 3275 if ((error = namei(&nd)) != 0) 3276 return (error); 3277 vp = nd.ni_vp; 3278 if (vp != NULL) { 3279 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3280 if (nd.ni_dvp == vp) 3281 vrele(nd.ni_dvp); 3282 else 3283 vput(nd.ni_dvp); 3284 vrele(vp); 3285 return (EEXIST); 3286 } 3287 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3288 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3289 if (nd.ni_dvp == vp) 3290 vrele(nd.ni_dvp); 3291 else 3292 vput(nd.ni_dvp); 3293 if ((error = vn_start_write(NULL, &mp, 3294 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3295 return (error); 3296 goto restart; 3297 } 3298 VATTR_NULL(&vattr); 3299 vattr.va_type = VDIR; 3300 vattr.va_mode = 3301 (SCARG(uap, mode) & ACCESSPERMS) &~ p->p_cwdi->cwdi_cmask; 3302 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3303 error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); 3304 if (!error) 3305 vput(nd.ni_vp); 3306 vn_finished_write(mp, 0); 3307 return (error); 3308 } 3309 3310 /* 3311 * Remove a directory file. 3312 */ 3313 /* ARGSUSED */ 3314 int 3315 sys_rmdir(struct lwp *l, void *v, register_t *retval) 3316 { 3317 struct sys_rmdir_args /* { 3318 syscallarg(const char *) path; 3319 } */ *uap = v; 3320 struct proc *p = l->l_proc; 3321 struct mount *mp; 3322 struct vnode *vp; 3323 int error; 3324 struct nameidata nd; 3325 3326 restart: 3327 NDINIT(&nd, DELETE, LOCKPARENT | LOCKLEAF, UIO_USERSPACE, 3328 SCARG(uap, path), l); 3329 if ((error = namei(&nd)) != 0) 3330 return (error); 3331 vp = nd.ni_vp; 3332 if (vp->v_type != VDIR) { 3333 error = ENOTDIR; 3334 goto out; 3335 } 3336 /* 3337 * No rmdir "." please. 3338 */ 3339 if (nd.ni_dvp == vp) { 3340 error = EINVAL; 3341 goto out; 3342 } 3343 /* 3344 * The root of a mounted filesystem cannot be deleted. 3345 */ 3346 if (vp->v_flag & VROOT) { 3347 error = EBUSY; 3348 goto out; 3349 } 3350 if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { 3351 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3352 if (nd.ni_dvp == vp) 3353 vrele(nd.ni_dvp); 3354 else 3355 vput(nd.ni_dvp); 3356 vput(vp); 3357 if ((error = vn_start_write(NULL, &mp, 3358 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 3359 return (error); 3360 goto restart; 3361 } 3362 VOP_LEASE(nd.ni_dvp, l, p->p_cred, LEASE_WRITE); 3363 VOP_LEASE(vp, l, p->p_cred, LEASE_WRITE); 3364 error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); 3365 vn_finished_write(mp, 0); 3366 return (error); 3367 3368 out: 3369 VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); 3370 if (nd.ni_dvp == vp) 3371 vrele(nd.ni_dvp); 3372 else 3373 vput(nd.ni_dvp); 3374 vput(vp); 3375 return (error); 3376 } 3377 3378 /* 3379 * Read a block of directory entries in a file system independent format. 3380 */ 3381 int 3382 sys___getdents30(struct lwp *l, void *v, register_t *retval) 3383 { 3384 struct sys___getdents30_args /* { 3385 syscallarg(int) fd; 3386 syscallarg(char *) buf; 3387 syscallarg(size_t) count; 3388 } */ *uap = v; 3389 struct proc *p = l->l_proc; 3390 struct file *fp; 3391 int error, done; 3392 3393 /* getvnode() will use the descriptor for us */ 3394 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 3395 return (error); 3396 if ((fp->f_flag & FREAD) == 0) { 3397 error = EBADF; 3398 goto out; 3399 } 3400 error = vn_readdir(fp, SCARG(uap, buf), UIO_USERSPACE, 3401 SCARG(uap, count), &done, l, 0, 0); 3402 #ifdef KTRACE 3403 if (!error && KTRPOINT(p, KTR_GENIO)) { 3404 struct iovec iov; 3405 iov.iov_base = SCARG(uap, buf); 3406 iov.iov_len = done; 3407 ktrgenio(l, SCARG(uap, fd), UIO_READ, &iov, done, 0); 3408 } 3409 #endif 3410 *retval = done; 3411 out: 3412 FILE_UNUSE(fp, l); 3413 return (error); 3414 } 3415 3416 /* 3417 * Set the mode mask for creation of filesystem nodes. 3418 */ 3419 int 3420 sys_umask(struct lwp *l, void *v, register_t *retval) 3421 { 3422 struct sys_umask_args /* { 3423 syscallarg(mode_t) newmask; 3424 } */ *uap = v; 3425 struct proc *p = l->l_proc; 3426 struct cwdinfo *cwdi; 3427 3428 cwdi = p->p_cwdi; 3429 *retval = cwdi->cwdi_cmask; 3430 cwdi->cwdi_cmask = SCARG(uap, newmask) & ALLPERMS; 3431 return (0); 3432 } 3433 3434 /* 3435 * Void all references to file by ripping underlying filesystem 3436 * away from vnode. 3437 */ 3438 /* ARGSUSED */ 3439 int 3440 sys_revoke(struct lwp *l, void *v, register_t *retval) 3441 { 3442 struct sys_revoke_args /* { 3443 syscallarg(const char *) path; 3444 } */ *uap = v; 3445 struct proc *p = l->l_proc; 3446 struct mount *mp; 3447 struct vnode *vp; 3448 struct vattr vattr; 3449 int error; 3450 struct nameidata nd; 3451 3452 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, path), l); 3453 if ((error = namei(&nd)) != 0) 3454 return (error); 3455 vp = nd.ni_vp; 3456 if ((error = VOP_GETATTR(vp, &vattr, p->p_cred, l)) != 0) 3457 goto out; 3458 if (kauth_cred_geteuid(p->p_cred) != vattr.va_uid && 3459 (error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER, 3460 &p->p_acflag)) != 0) 3461 goto out; 3462 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 3463 goto out; 3464 if (vp->v_usecount > 1 || (vp->v_flag & (VALIASED | VLAYER))) 3465 VOP_REVOKE(vp, REVOKEALL); 3466 vn_finished_write(mp, 0); 3467 out: 3468 vrele(vp); 3469 return (error); 3470 } 3471 3472 /* 3473 * Convert a user file descriptor to a kernel file entry. 3474 */ 3475 int 3476 getvnode(struct filedesc *fdp, int fd, struct file **fpp) 3477 { 3478 struct vnode *vp; 3479 struct file *fp; 3480 3481 if ((fp = fd_getfile(fdp, fd)) == NULL) 3482 return (EBADF); 3483 3484 FILE_USE(fp); 3485 3486 if (fp->f_type != DTYPE_VNODE) { 3487 FILE_UNUSE(fp, NULL); 3488 return (EINVAL); 3489 } 3490 3491 vp = (struct vnode *)fp->f_data; 3492 if (vp->v_type == VBAD) { 3493 FILE_UNUSE(fp, NULL); 3494 return (EBADF); 3495 } 3496 3497 *fpp = fp; 3498 return (0); 3499 } 3500
Indexes created Wed Sep 24 14:09:57 GMT 2025