ufs_vnops.c revision 101744
1/* 2 * Copyright (c) 1982, 1986, 1989, 1993, 1995 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)ufs_vnops.c 8.27 (Berkeley) 5/27/95 39 * $FreeBSD: head/sys/ufs/ufs/ufs_vnops.c 101744 2002-08-12 16:43:04Z rwatson $ 40 */ 41 42#include "opt_mac.h" 43#include "opt_quota.h" 44#include "opt_suiddir.h" 45#include "opt_ufs.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/malloc.h> 50#include <sys/namei.h> 51#include <sys/kernel.h> 52#include <sys/fcntl.h> 53#include <sys/stat.h> 54#include <sys/bio.h> 55#include <sys/buf.h> 56#include <sys/mount.h> 57#include <sys/unistd.h> 58#include <sys/vnode.h> 59#include <sys/dirent.h> 60#include <sys/lockf.h> 61#include <sys/event.h> 62#include <sys/conf.h> 63#include <sys/acl.h> 64#include <sys/mac.h> 65 66#include <machine/mutex.h> 67 68#include <sys/file.h> /* XXX */ 69 70#include <vm/vm.h> 71#include <vm/vm_extern.h> 72 73#include <fs/fifofs/fifo.h> 74 75#include <ufs/ufs/acl.h> 76#include <ufs/ufs/extattr.h> 77#include <ufs/ufs/quota.h> 78#include <ufs/ufs/inode.h> 79#include <ufs/ufs/dir.h> 80#include <ufs/ufs/ufsmount.h> 81#include <ufs/ufs/ufs_extern.h> 82#ifdef UFS_DIRHASH 83#include <ufs/ufs/dirhash.h> 84#endif 85 86static int ufs_access(struct vop_access_args *); 87static int ufs_advlock(struct vop_advlock_args *); 88static int ufs_chmod(struct vnode *, int, struct ucred *, struct thread *); 89static int ufs_chown(struct vnode *, uid_t, gid_t, struct ucred *, struct thread *); 90static int ufs_close(struct vop_close_args *); 91static int ufs_create(struct vop_create_args *); 92static int ufs_getattr(struct vop_getattr_args *); 93static int ufs_link(struct vop_link_args *); 94static int ufs_makeinode(int mode, struct vnode *, struct vnode **, struct componentname *); 95static int ufs_mkdir(struct vop_mkdir_args *); 96static int ufs_mknod(struct vop_mknod_args *); 97static int ufs_open(struct vop_open_args *); 98static int ufs_pathconf(struct vop_pathconf_args *); 99static int ufs_print(struct vop_print_args *); 100static int ufs_readlink(struct vop_readlink_args *); 101static int ufs_remove(struct vop_remove_args *); 102static int ufs_rename(struct vop_rename_args *); 103static int ufs_rmdir(struct vop_rmdir_args *); 104static int ufs_setattr(struct vop_setattr_args *); 105static int ufs_strategy(struct vop_strategy_args *); 106static int ufs_symlink(struct vop_symlink_args *); 107static int ufs_whiteout(struct vop_whiteout_args *); 108static int ufsfifo_close(struct vop_close_args *); 109static int ufsfifo_kqfilter(struct vop_kqfilter_args *); 110static int ufsfifo_read(struct vop_read_args *); 111static int ufsfifo_write(struct vop_write_args *); 112static int ufsspec_close(struct vop_close_args *); 113static int ufsspec_read(struct vop_read_args *); 114static int ufsspec_write(struct vop_write_args *); 115static int filt_ufsread(struct knote *kn, long hint); 116static int filt_ufswrite(struct knote *kn, long hint); 117static int filt_ufsvnode(struct knote *kn, long hint); 118static void filt_ufsdetach(struct knote *kn); 119static int ufs_kqfilter(struct vop_kqfilter_args *ap); 120 121union _qcvt { 122 int64_t qcvt; 123 int32_t val[2]; 124}; 125#define SETHIGH(q, h) { \ 126 union _qcvt tmp; \ 127 tmp.qcvt = (q); \ 128 tmp.val[_QUAD_HIGHWORD] = (h); \ 129 (q) = tmp.qcvt; \ 130} 131#define SETLOW(q, l) { \ 132 union _qcvt tmp; \ 133 tmp.qcvt = (q); \ 134 tmp.val[_QUAD_LOWWORD] = (l); \ 135 (q) = tmp.qcvt; \ 136} 137 138/* 139 * A virgin directory (no blushing please). 140 */ 141static struct dirtemplate mastertemplate = { 142 0, 12, DT_DIR, 1, ".", 143 0, DIRBLKSIZ - 12, DT_DIR, 2, ".." 144}; 145static struct odirtemplate omastertemplate = { 146 0, 12, 1, ".", 147 0, DIRBLKSIZ - 12, 2, ".." 148}; 149 150void 151ufs_itimes(vp) 152 struct vnode *vp; 153{ 154 struct inode *ip; 155 struct timespec ts; 156 157 ip = VTOI(vp); 158 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0) 159 return; 160 if ((vp->v_type == VBLK || vp->v_type == VCHR) && !DOINGSOFTDEP(vp)) 161 ip->i_flag |= IN_LAZYMOD; 162 else 163 ip->i_flag |= IN_MODIFIED; 164 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { 165 vfs_timestamp(&ts); 166 if (ip->i_flag & IN_ACCESS) { 167 DIP(ip, i_atime) = ts.tv_sec; 168 DIP(ip, i_atimensec) = ts.tv_nsec; 169 } 170 if (ip->i_flag & IN_UPDATE) { 171 DIP(ip, i_mtime) = ts.tv_sec; 172 DIP(ip, i_mtimensec) = ts.tv_nsec; 173 ip->i_modrev++; 174 } 175 if (ip->i_flag & IN_CHANGE) { 176 DIP(ip, i_ctime) = ts.tv_sec; 177 DIP(ip, i_ctimensec) = ts.tv_nsec; 178 } 179 } 180 ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE); 181} 182 183/* 184 * Create a regular file 185 */ 186int 187ufs_create(ap) 188 struct vop_create_args /* { 189 struct vnode *a_dvp; 190 struct vnode **a_vpp; 191 struct componentname *a_cnp; 192 struct vattr *a_vap; 193 } */ *ap; 194{ 195 int error; 196 197 error = 198 ufs_makeinode(MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode), 199 ap->a_dvp, ap->a_vpp, ap->a_cnp); 200 if (error) 201 return (error); 202 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 203 return (0); 204} 205 206/* 207 * Mknod vnode call 208 */ 209/* ARGSUSED */ 210int 211ufs_mknod(ap) 212 struct vop_mknod_args /* { 213 struct vnode *a_dvp; 214 struct vnode **a_vpp; 215 struct componentname *a_cnp; 216 struct vattr *a_vap; 217 } */ *ap; 218{ 219 struct vattr *vap = ap->a_vap; 220 struct vnode **vpp = ap->a_vpp; 221 struct inode *ip; 222 ino_t ino; 223 int error; 224 225 error = ufs_makeinode(MAKEIMODE(vap->va_type, vap->va_mode), 226 ap->a_dvp, vpp, ap->a_cnp); 227 if (error) 228 return (error); 229 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 230 ip = VTOI(*vpp); 231 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 232 if (vap->va_rdev != VNOVAL) { 233 /* 234 * Want to be able to use this to make badblock 235 * inodes, so don't truncate the dev number. 236 */ 237 DIP(ip, i_rdev) = vap->va_rdev; 238 } 239 /* 240 * Remove inode, then reload it through VFS_VGET so it is 241 * checked to see if it is an alias of an existing entry in 242 * the inode cache. 243 */ 244 vput(*vpp); 245 (*vpp)->v_type = VNON; 246 ino = ip->i_number; /* Save this before vgone() invalidates ip. */ 247 vgone(*vpp); 248 error = VFS_VGET(ap->a_dvp->v_mount, ino, LK_EXCLUSIVE, vpp); 249 if (error) { 250 *vpp = NULL; 251 return (error); 252 } 253 return (0); 254} 255 256/* 257 * Open called. 258 * 259 * Nothing to do. 260 */ 261/* ARGSUSED */ 262int 263ufs_open(ap) 264 struct vop_open_args /* { 265 struct vnode *a_vp; 266 int a_mode; 267 struct ucred *a_cred; 268 struct thread *a_td; 269 } */ *ap; 270{ 271 272 /* 273 * Files marked append-only must be opened for appending. 274 */ 275 if ((VTOI(ap->a_vp)->i_flags & APPEND) && 276 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE) 277 return (EPERM); 278 return (0); 279} 280 281/* 282 * Close called. 283 * 284 * Update the times on the inode. 285 */ 286/* ARGSUSED */ 287int 288ufs_close(ap) 289 struct vop_close_args /* { 290 struct vnode *a_vp; 291 int a_fflag; 292 struct ucred *a_cred; 293 struct thread *a_td; 294 } */ *ap; 295{ 296 struct vnode *vp = ap->a_vp; 297 struct mount *mp; 298 299 mtx_lock(&vp->v_interlock); 300 if (vp->v_usecount > 1) { 301 ufs_itimes(vp); 302 mtx_unlock(&vp->v_interlock); 303 } else { 304 mtx_unlock(&vp->v_interlock); 305 /* 306 * If we are closing the last reference to an unlinked 307 * file, then it will be freed by the inactive routine. 308 * Because the freeing causes a the filesystem to be 309 * modified, it must be held up during periods when the 310 * filesystem is suspended. 311 * 312 * XXX - EAGAIN is returned to prevent vn_close from 313 * repeating the vrele operation. 314 */ 315 if (vp->v_type == VREG && VTOI(vp)->i_effnlink == 0) { 316 (void) vn_start_write(vp, &mp, V_WAIT); 317 vrele(vp); 318 vn_finished_write(mp); 319 return (EAGAIN); 320 } 321 } 322 return (0); 323} 324 325int 326ufs_access(ap) 327 struct vop_access_args /* { 328 struct vnode *a_vp; 329 int a_mode; 330 struct ucred *a_cred; 331 struct thread *a_td; 332 } */ *ap; 333{ 334 struct vnode *vp = ap->a_vp; 335 struct inode *ip = VTOI(vp); 336 mode_t mode = ap->a_mode; 337 int error; 338#ifdef UFS_ACL 339 struct acl *acl; 340 int len; 341#endif 342 343 /* 344 * Disallow write attempts on read-only filesystems; 345 * unless the file is a socket, fifo, or a block or 346 * character device resident on the filesystem. 347 */ 348 if (mode & VWRITE) { 349 switch (vp->v_type) { 350 case VDIR: 351 case VLNK: 352 case VREG: 353 if (vp->v_mount->mnt_flag & MNT_RDONLY) 354 return (EROFS); 355#ifdef QUOTA 356 if ((error = getinoquota(ip)) != 0) 357 return (error); 358#endif 359 break; 360 default: 361 break; 362 } 363 } 364 365 /* If immutable bit set, nobody gets to write it. */ 366 if ((mode & VWRITE) && (ip->i_flags & (IMMUTABLE | SF_SNAPSHOT))) 367 return (EPERM); 368 369#ifdef UFS_ACL 370 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK); 371 len = sizeof(*acl); 372 error = VOP_GETACL(vp, ACL_TYPE_ACCESS, acl, ap->a_cred, ap->a_td); 373 switch (error) { 374 case EOPNOTSUPP: 375 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid, 376 ap->a_mode, ap->a_cred, NULL); 377 break; 378 case 0: 379 error = vaccess_acl_posix1e(vp->v_type, ip->i_uid, ip->i_gid, 380 acl, ap->a_mode, ap->a_cred, NULL); 381 break; 382 default: 383 printf("ufs_access(): Error retrieving ACL on object (%d).\n", 384 error); 385 /* 386 * XXX: Fall back until debugged. Should eventually 387 * possibly log an error, and return EPERM for safety. 388 */ 389 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid, 390 ap->a_mode, ap->a_cred, NULL); 391 } 392 FREE(acl, M_ACL); 393#else 394 error = vaccess(vp->v_type, ip->i_mode, ip->i_uid, ip->i_gid, 395 ap->a_mode, ap->a_cred, NULL); 396#endif 397 return (error); 398} 399 400/* ARGSUSED */ 401int 402ufs_getattr(ap) 403 struct vop_getattr_args /* { 404 struct vnode *a_vp; 405 struct vattr *a_vap; 406 struct ucred *a_cred; 407 struct thread *a_td; 408 } */ *ap; 409{ 410 struct vnode *vp = ap->a_vp; 411 struct inode *ip = VTOI(vp); 412 struct vattr *vap = ap->a_vap; 413 414 ufs_itimes(vp); 415 /* 416 * Copy from inode table 417 */ 418 vap->va_fsid = dev2udev(ip->i_dev); 419 vap->va_fileid = ip->i_number; 420 vap->va_mode = ip->i_mode & ~IFMT; 421 vap->va_nlink = ip->i_effnlink; 422 vap->va_uid = ip->i_uid; 423 vap->va_gid = ip->i_gid; 424 if (ip->i_ump->um_fstype == UFS1) { 425 vap->va_rdev = ip->i_din1->di_rdev; 426 vap->va_size = ip->i_din1->di_size; 427 vap->va_atime.tv_sec = ip->i_din1->di_atime; 428 vap->va_atime.tv_nsec = ip->i_din1->di_atimensec; 429 vap->va_mtime.tv_sec = ip->i_din1->di_mtime; 430 vap->va_mtime.tv_nsec = ip->i_din1->di_mtimensec; 431 vap->va_ctime.tv_sec = ip->i_din1->di_ctime; 432 vap->va_ctime.tv_nsec = ip->i_din1->di_ctimensec; 433 vap->va_birthtime.tv_sec = 0; 434 vap->va_birthtime.tv_nsec = 0; 435 vap->va_bytes = dbtob((u_quad_t)ip->i_din1->di_blocks); 436 } else { 437 vap->va_rdev = ip->i_din2->di_rdev; 438 vap->va_size = ip->i_din2->di_size; 439 vap->va_atime.tv_sec = ip->i_din2->di_atime; 440 vap->va_atime.tv_nsec = ip->i_din2->di_atimensec; 441 vap->va_mtime.tv_sec = ip->i_din2->di_mtime; 442 vap->va_mtime.tv_nsec = ip->i_din2->di_mtimensec; 443 vap->va_ctime.tv_sec = ip->i_din2->di_ctime; 444 vap->va_ctime.tv_nsec = ip->i_din2->di_ctimensec; 445 vap->va_birthtime.tv_sec = ip->i_din2->di_birthtime; 446 vap->va_birthtime.tv_nsec = ip->i_din2->di_birthnsec; 447 vap->va_bytes = dbtob((u_quad_t)ip->i_din2->di_blocks); 448 } 449 vap->va_flags = ip->i_flags; 450 vap->va_gen = ip->i_gen; 451 vap->va_blocksize = vp->v_mount->mnt_stat.f_iosize; 452 vap->va_type = IFTOVT(ip->i_mode); 453 vap->va_filerev = ip->i_modrev; 454 return (0); 455} 456 457/* 458 * Set attribute vnode op. called from several syscalls 459 */ 460int 461ufs_setattr(ap) 462 struct vop_setattr_args /* { 463 struct vnode *a_vp; 464 struct vattr *a_vap; 465 struct ucred *a_cred; 466 struct thread *a_td; 467 } */ *ap; 468{ 469 struct vattr *vap = ap->a_vap; 470 struct vnode *vp = ap->a_vp; 471 struct inode *ip = VTOI(vp); 472 struct ucred *cred = ap->a_cred; 473 struct thread *td = ap->a_td; 474 int error; 475 476 /* 477 * Check for unsettable attributes. 478 */ 479 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) || 480 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) || 481 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) || 482 ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) { 483 return (EINVAL); 484 } 485 if (vap->va_flags != VNOVAL) { 486 if (vp->v_mount->mnt_flag & MNT_RDONLY) 487 return (EROFS); 488 /* 489 * Callers may only modify the file flags on objects they 490 * have VADMIN rights for. 491 */ 492 if ((error = VOP_ACCESS(vp, VADMIN, cred, td))) 493 return (error); 494 /* 495 * Unprivileged processes and privileged processes in 496 * jail() are not permitted to unset system flags, or 497 * modify flags if any system flags are set. 498 * Privileged non-jail processes may not modify system flags 499 * if securelevel > 0 and any existing system flags are set. 500 */ 501 if (!suser_cred(cred, PRISON_ROOT)) { 502 if (ip->i_flags 503 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) { 504 error = securelevel_gt(cred, 0); 505 if (error) 506 return (error); 507 } 508 /* Snapshot flag cannot be set or cleared */ 509 if (((vap->va_flags & SF_SNAPSHOT) != 0 && 510 (ip->i_flags & SF_SNAPSHOT) == 0) || 511 ((vap->va_flags & SF_SNAPSHOT) == 0 && 512 (ip->i_flags & SF_SNAPSHOT) != 0)) 513 return (EPERM); 514 ip->i_flags = vap->va_flags; 515 DIP(ip, i_flags) = vap->va_flags; 516 } else { 517 if (ip->i_flags 518 & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) || 519 (vap->va_flags & UF_SETTABLE) != vap->va_flags) 520 return (EPERM); 521 ip->i_flags &= SF_SETTABLE; 522 ip->i_flags |= (vap->va_flags & UF_SETTABLE); 523 DIP(ip, i_flags) = ip->i_flags; 524 } 525 ip->i_flag |= IN_CHANGE; 526 if (vap->va_flags & (IMMUTABLE | APPEND)) 527 return (0); 528 } 529 if (ip->i_flags & (IMMUTABLE | APPEND)) 530 return (EPERM); 531 /* 532 * Go through the fields and update iff not VNOVAL. 533 */ 534 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) { 535 if (vp->v_mount->mnt_flag & MNT_RDONLY) 536 return (EROFS); 537 if ((error = ufs_chown(vp, vap->va_uid, vap->va_gid, cred, 538 td)) != 0) 539 return (error); 540 } 541 if (vap->va_size != VNOVAL) { 542 /* 543 * Disallow write attempts on read-only filesystems; 544 * unless the file is a socket, fifo, or a block or 545 * character device resident on the filesystem. 546 */ 547 switch (vp->v_type) { 548 case VDIR: 549 return (EISDIR); 550 case VLNK: 551 case VREG: 552 if (vp->v_mount->mnt_flag & MNT_RDONLY) 553 return (EROFS); 554 if ((ip->i_flags & SF_SNAPSHOT) != 0) 555 return (EPERM); 556 break; 557 default: 558 break; 559 } 560 if ((error = UFS_TRUNCATE(vp, vap->va_size, IO_NORMAL, 561 cred, td)) != 0) 562 return (error); 563 } 564 if (vap->va_atime.tv_sec != VNOVAL || 565 vap->va_mtime.tv_sec != VNOVAL || 566 vap->va_birthtime.tv_sec != VNOVAL) { 567 if (vp->v_mount->mnt_flag & MNT_RDONLY) 568 return (EROFS); 569 if ((ip->i_flags & SF_SNAPSHOT) != 0) 570 return (EPERM); 571 /* 572 * From utimes(2): 573 * If times is NULL, ... The caller must be the owner of 574 * the file, have permission to write the file, or be the 575 * super-user. 576 * If times is non-NULL, ... The caller must be the owner of 577 * the file or be the super-user. 578 */ 579 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)) && 580 ((vap->va_vaflags & VA_UTIMES_NULL) == 0 || 581 (error = VOP_ACCESS(vp, VWRITE, cred, td)))) 582 return (error); 583 if (vap->va_atime.tv_sec != VNOVAL) 584 ip->i_flag |= IN_ACCESS; 585 if (vap->va_mtime.tv_sec != VNOVAL) 586 ip->i_flag |= IN_CHANGE | IN_UPDATE; 587 if (vap->va_birthtime.tv_sec != VNOVAL && 588 ip->i_ump->um_fstype == UFS2) 589 ip->i_flag |= IN_MODIFIED; 590 ufs_itimes(vp); 591 if (vap->va_atime.tv_sec != VNOVAL) { 592 DIP(ip, i_atime) = vap->va_atime.tv_sec; 593 DIP(ip, i_atimensec) = vap->va_atime.tv_nsec; 594 } 595 if (vap->va_mtime.tv_sec != VNOVAL) { 596 DIP(ip, i_mtime) = vap->va_mtime.tv_sec; 597 DIP(ip, i_mtimensec) = vap->va_mtime.tv_nsec; 598 } 599 if (vap->va_birthtime.tv_sec != VNOVAL && 600 ip->i_ump->um_fstype == UFS2) { 601 ip->i_din2->di_birthtime = vap->va_birthtime.tv_sec; 602 ip->i_din2->di_birthnsec = vap->va_birthtime.tv_nsec; 603 } 604 error = UFS_UPDATE(vp, 0); 605 if (error) 606 return (error); 607 } 608 error = 0; 609 if (vap->va_mode != (mode_t)VNOVAL) { 610 if (vp->v_mount->mnt_flag & MNT_RDONLY) 611 return (EROFS); 612 if ((ip->i_flags & SF_SNAPSHOT) != 0 && (vap->va_mode & 613 (S_IXUSR | S_IWUSR | S_IXGRP | S_IWGRP | S_IXOTH | S_IWOTH))) 614 return (EPERM); 615 error = ufs_chmod(vp, (int)vap->va_mode, cred, td); 616 } 617 VN_KNOTE(vp, NOTE_ATTRIB); 618 return (error); 619} 620 621/* 622 * Change the mode on a file. 623 * Inode must be locked before calling. 624 */ 625static int 626ufs_chmod(vp, mode, cred, td) 627 struct vnode *vp; 628 int mode; 629 struct ucred *cred; 630 struct thread *td; 631{ 632 struct inode *ip = VTOI(vp); 633 int error; 634 635 /* 636 * To modify the permissions on a file, must possess VADMIN 637 * for that file. 638 */ 639 if ((error = VOP_ACCESS(vp, VADMIN, cred, td))) 640 return (error); 641 /* 642 * Privileged processes may set the sticky bit on non-directories, 643 * as well as set the setgid bit on a file with a group that the 644 * process is not a member of. 645 */ 646 if (suser_cred(cred, PRISON_ROOT)) { 647 if (vp->v_type != VDIR && (mode & S_ISTXT)) 648 return (EFTYPE); 649 if (!groupmember(ip->i_gid, cred) && (mode & ISGID)) 650 return (EPERM); 651 } 652 ip->i_mode &= ~ALLPERMS; 653 ip->i_mode |= (mode & ALLPERMS); 654 DIP(ip, i_mode) = ip->i_mode; 655 ip->i_flag |= IN_CHANGE; 656 return (0); 657} 658 659/* 660 * Perform chown operation on inode ip; 661 * inode must be locked prior to call. 662 */ 663static int 664ufs_chown(vp, uid, gid, cred, td) 665 struct vnode *vp; 666 uid_t uid; 667 gid_t gid; 668 struct ucred *cred; 669 struct thread *td; 670{ 671 struct inode *ip = VTOI(vp); 672 uid_t ouid; 673 gid_t ogid; 674 int error = 0; 675#ifdef QUOTA 676 int i; 677 ufs2_daddr_t change; 678#endif 679 680 if (uid == (uid_t)VNOVAL) 681 uid = ip->i_uid; 682 if (gid == (gid_t)VNOVAL) 683 gid = ip->i_gid; 684 /* 685 * To modify the ownership of a file, must possess VADMIN 686 * for that file. 687 */ 688 if ((error = VOP_ACCESS(vp, VADMIN, cred, td))) 689 return (error); 690 /* 691 * To change the owner of a file, or change the group of a file 692 * to a group of which we are not a member, the caller must 693 * have privilege. 694 */ 695 if ((uid != ip->i_uid || 696 (gid != ip->i_gid && !groupmember(gid, cred))) && 697 (error = suser_cred(cred, PRISON_ROOT))) 698 return (error); 699 ogid = ip->i_gid; 700 ouid = ip->i_uid; 701#ifdef QUOTA 702 if ((error = getinoquota(ip)) != 0) 703 return (error); 704 if (ouid == uid) { 705 dqrele(vp, ip->i_dquot[USRQUOTA]); 706 ip->i_dquot[USRQUOTA] = NODQUOT; 707 } 708 if (ogid == gid) { 709 dqrele(vp, ip->i_dquot[GRPQUOTA]); 710 ip->i_dquot[GRPQUOTA] = NODQUOT; 711 } 712 change = DIP(ip, i_blocks); 713 (void) chkdq(ip, -change, cred, CHOWN); 714 (void) chkiq(ip, -1, cred, CHOWN); 715 for (i = 0; i < MAXQUOTAS; i++) { 716 dqrele(vp, ip->i_dquot[i]); 717 ip->i_dquot[i] = NODQUOT; 718 } 719#endif 720 ip->i_gid = gid; 721 DIP(ip, i_gid) = gid; 722 ip->i_uid = uid; 723 DIP(ip, i_uid) = uid; 724#ifdef QUOTA 725 if ((error = getinoquota(ip)) == 0) { 726 if (ouid == uid) { 727 dqrele(vp, ip->i_dquot[USRQUOTA]); 728 ip->i_dquot[USRQUOTA] = NODQUOT; 729 } 730 if (ogid == gid) { 731 dqrele(vp, ip->i_dquot[GRPQUOTA]); 732 ip->i_dquot[GRPQUOTA] = NODQUOT; 733 } 734 if ((error = chkdq(ip, change, cred, CHOWN)) == 0) { 735 if ((error = chkiq(ip, 1, cred, CHOWN)) == 0) 736 goto good; 737 else 738 (void) chkdq(ip, -change, cred, CHOWN|FORCE); 739 } 740 for (i = 0; i < MAXQUOTAS; i++) { 741 dqrele(vp, ip->i_dquot[i]); 742 ip->i_dquot[i] = NODQUOT; 743 } 744 } 745 ip->i_gid = ogid; 746 DIP(ip, i_gid) = ogid; 747 ip->i_uid = ouid; 748 DIP(ip, i_uid) = ouid; 749 if (getinoquota(ip) == 0) { 750 if (ouid == uid) { 751 dqrele(vp, ip->i_dquot[USRQUOTA]); 752 ip->i_dquot[USRQUOTA] = NODQUOT; 753 } 754 if (ogid == gid) { 755 dqrele(vp, ip->i_dquot[GRPQUOTA]); 756 ip->i_dquot[GRPQUOTA] = NODQUOT; 757 } 758 (void) chkdq(ip, change, cred, FORCE|CHOWN); 759 (void) chkiq(ip, 1, cred, FORCE|CHOWN); 760 (void) getinoquota(ip); 761 } 762 return (error); 763good: 764 if (getinoquota(ip)) 765 panic("ufs_chown: lost quota"); 766#endif /* QUOTA */ 767 ip->i_flag |= IN_CHANGE; 768 if (suser_cred(cred, PRISON_ROOT) && (ouid != uid || ogid != gid)) { 769 ip->i_mode &= ~(ISUID | ISGID); 770 DIP(ip, i_mode) = ip->i_mode; 771 } 772 return (0); 773} 774 775int 776ufs_remove(ap) 777 struct vop_remove_args /* { 778 struct vnode *a_dvp; 779 struct vnode *a_vp; 780 struct componentname *a_cnp; 781 } */ *ap; 782{ 783 struct inode *ip; 784 struct vnode *vp = ap->a_vp; 785 struct vnode *dvp = ap->a_dvp; 786 int error; 787 788 ip = VTOI(vp); 789 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) || 790 (VTOI(dvp)->i_flags & APPEND)) { 791 error = EPERM; 792 goto out; 793 } 794 error = ufs_dirremove(dvp, ip, ap->a_cnp->cn_flags, 0); 795 if (ip->i_nlink <= 0) 796 vp->v_vflag |= VV_NOSYNC; 797 VN_KNOTE(vp, NOTE_DELETE); 798 VN_KNOTE(dvp, NOTE_WRITE); 799out: 800 return (error); 801} 802 803/* 804 * link vnode call 805 */ 806int 807ufs_link(ap) 808 struct vop_link_args /* { 809 struct vnode *a_tdvp; 810 struct vnode *a_vp; 811 struct componentname *a_cnp; 812 } */ *ap; 813{ 814 struct vnode *vp = ap->a_vp; 815 struct vnode *tdvp = ap->a_tdvp; 816 struct componentname *cnp = ap->a_cnp; 817 struct thread *td = cnp->cn_thread; 818 struct inode *ip; 819 struct direct newdir; 820 int error; 821 822#ifdef DIAGNOSTIC 823 if ((cnp->cn_flags & HASBUF) == 0) 824 panic("ufs_link: no name"); 825#endif 826 if (tdvp->v_mount != vp->v_mount) { 827 error = EXDEV; 828 goto out2; 829 } 830 if (tdvp != vp && (error = vn_lock(vp, LK_EXCLUSIVE, td))) { 831 goto out2; 832 } 833 ip = VTOI(vp); 834 if ((nlink_t)ip->i_nlink >= LINK_MAX) { 835 error = EMLINK; 836 goto out1; 837 } 838 if (ip->i_flags & (IMMUTABLE | APPEND)) { 839 error = EPERM; 840 goto out1; 841 } 842 ip->i_effnlink++; 843 ip->i_nlink++; 844 DIP(ip, i_nlink) = ip->i_nlink; 845 ip->i_flag |= IN_CHANGE; 846 if (DOINGSOFTDEP(vp)) 847 softdep_change_linkcnt(ip); 848 error = UFS_UPDATE(vp, !(DOINGSOFTDEP(vp) | DOINGASYNC(vp))); 849 if (!error) { 850 ufs_makedirentry(ip, cnp, &newdir); 851 error = ufs_direnter(tdvp, vp, &newdir, cnp, NULL); 852 } 853 854 if (error) { 855 ip->i_effnlink--; 856 ip->i_nlink--; 857 DIP(ip, i_nlink) = ip->i_nlink; 858 ip->i_flag |= IN_CHANGE; 859 if (DOINGSOFTDEP(vp)) 860 softdep_change_linkcnt(ip); 861 } 862out1: 863 if (tdvp != vp) 864 VOP_UNLOCK(vp, 0, td); 865out2: 866 VN_KNOTE(vp, NOTE_LINK); 867 VN_KNOTE(tdvp, NOTE_WRITE); 868 return (error); 869} 870 871/* 872 * whiteout vnode call 873 */ 874int 875ufs_whiteout(ap) 876 struct vop_whiteout_args /* { 877 struct vnode *a_dvp; 878 struct componentname *a_cnp; 879 int a_flags; 880 } */ *ap; 881{ 882 struct vnode *dvp = ap->a_dvp; 883 struct componentname *cnp = ap->a_cnp; 884 struct direct newdir; 885 int error = 0; 886 887 switch (ap->a_flags) { 888 case LOOKUP: 889 /* 4.4 format directories support whiteout operations */ 890 if (dvp->v_mount->mnt_maxsymlinklen > 0) 891 return (0); 892 return (EOPNOTSUPP); 893 894 case CREATE: 895 /* create a new directory whiteout */ 896#ifdef DIAGNOSTIC 897 if ((cnp->cn_flags & SAVENAME) == 0) 898 panic("ufs_whiteout: missing name"); 899 if (dvp->v_mount->mnt_maxsymlinklen <= 0) 900 panic("ufs_whiteout: old format filesystem"); 901#endif 902 903 newdir.d_ino = WINO; 904 newdir.d_namlen = cnp->cn_namelen; 905 bcopy(cnp->cn_nameptr, newdir.d_name, (unsigned)cnp->cn_namelen + 1); 906 newdir.d_type = DT_WHT; 907 error = ufs_direnter(dvp, NULL, &newdir, cnp, NULL); 908 break; 909 910 case DELETE: 911 /* remove an existing directory whiteout */ 912#ifdef DIAGNOSTIC 913 if (dvp->v_mount->mnt_maxsymlinklen <= 0) 914 panic("ufs_whiteout: old format filesystem"); 915#endif 916 917 cnp->cn_flags &= ~DOWHITEOUT; 918 error = ufs_dirremove(dvp, NULL, cnp->cn_flags, 0); 919 break; 920 default: 921 panic("ufs_whiteout: unknown op"); 922 } 923 return (error); 924} 925 926/* 927 * Rename system call. 928 * rename("foo", "bar"); 929 * is essentially 930 * unlink("bar"); 931 * link("foo", "bar"); 932 * unlink("foo"); 933 * but ``atomically''. Can't do full commit without saving state in the 934 * inode on disk which isn't feasible at this time. Best we can do is 935 * always guarantee the target exists. 936 * 937 * Basic algorithm is: 938 * 939 * 1) Bump link count on source while we're linking it to the 940 * target. This also ensure the inode won't be deleted out 941 * from underneath us while we work (it may be truncated by 942 * a concurrent `trunc' or `open' for creation). 943 * 2) Link source to destination. If destination already exists, 944 * delete it first. 945 * 3) Unlink source reference to inode if still around. If a 946 * directory was moved and the parent of the destination 947 * is different from the source, patch the ".." entry in the 948 * directory. 949 */ 950int 951ufs_rename(ap) 952 struct vop_rename_args /* { 953 struct vnode *a_fdvp; 954 struct vnode *a_fvp; 955 struct componentname *a_fcnp; 956 struct vnode *a_tdvp; 957 struct vnode *a_tvp; 958 struct componentname *a_tcnp; 959 } */ *ap; 960{ 961 struct vnode *tvp = ap->a_tvp; 962 struct vnode *tdvp = ap->a_tdvp; 963 struct vnode *fvp = ap->a_fvp; 964 struct vnode *fdvp = ap->a_fdvp; 965 struct componentname *tcnp = ap->a_tcnp; 966 struct componentname *fcnp = ap->a_fcnp; 967 struct thread *td = fcnp->cn_thread; 968 struct inode *ip, *xp, *dp; 969 struct direct newdir; 970 int doingdirectory = 0, oldparent = 0, newparent = 0; 971 int error = 0, ioflag; 972 973#ifdef DIAGNOSTIC 974 if ((tcnp->cn_flags & HASBUF) == 0 || 975 (fcnp->cn_flags & HASBUF) == 0) 976 panic("ufs_rename: no name"); 977#endif 978 /* 979 * Check for cross-device rename. 980 */ 981 if ((fvp->v_mount != tdvp->v_mount) || 982 (tvp && (fvp->v_mount != tvp->v_mount))) { 983 error = EXDEV; 984abortit: 985 if (tdvp == tvp) 986 vrele(tdvp); 987 else 988 vput(tdvp); 989 if (tvp) 990 vput(tvp); 991 vrele(fdvp); 992 vrele(fvp); 993 return (error); 994 } 995 996 if (tvp && ((VTOI(tvp)->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) || 997 (VTOI(tdvp)->i_flags & APPEND))) { 998 error = EPERM; 999 goto abortit; 1000 } 1001 1002 /* 1003 * Check if just deleting a link name or if we've lost a race. 1004 * If another process completes the same rename after we've looked 1005 * up the source and have blocked looking up the target, then the 1006 * source and target inodes may be identical now although the 1007 * names were never linked. 1008 */ 1009 if (fvp == tvp) { 1010 if (fvp->v_type == VDIR) { 1011 /* 1012 * Linked directories are impossible, so we must 1013 * have lost the race. Pretend that the rename 1014 * completed before the lookup. 1015 */ 1016#ifdef UFS_RENAME_DEBUG 1017 printf("ufs_rename: fvp == tvp for directories\n"); 1018#endif 1019 error = ENOENT; 1020 goto abortit; 1021 } 1022 1023 /* Release destination completely. */ 1024 vput(tdvp); 1025 vput(tvp); 1026 1027 /* 1028 * Delete source. There is another race now that everything 1029 * is unlocked, but this doesn't cause any new complications. 1030 * Relookup() may find a file that is unrelated to the 1031 * original one, or it may fail. Too bad. 1032 */ 1033 vrele(fdvp); 1034 vrele(fvp); 1035 fcnp->cn_flags &= ~MODMASK; 1036 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF; 1037 if ((fcnp->cn_flags & SAVESTART) == 0) 1038 panic("ufs_rename: lost from startdir"); 1039 fcnp->cn_nameiop = DELETE; 1040 VREF(fdvp); 1041 error = relookup(fdvp, &fvp, fcnp); 1042 if (error == 0) 1043 vrele(fdvp); 1044 if (fvp == NULL) { 1045#ifdef UFS_RENAME_DEBUG 1046 printf("ufs_rename: from name disappeared\n"); 1047#endif 1048 return (ENOENT); 1049 } 1050 error = VOP_REMOVE(fdvp, fvp, fcnp); 1051 if (fdvp == fvp) 1052 vrele(fdvp); 1053 else 1054 vput(fdvp); 1055 vput(fvp); 1056 return (error); 1057 } 1058 if ((error = vn_lock(fvp, LK_EXCLUSIVE, td)) != 0) 1059 goto abortit; 1060 dp = VTOI(fdvp); 1061 ip = VTOI(fvp); 1062 if (ip->i_nlink >= LINK_MAX) { 1063 VOP_UNLOCK(fvp, 0, td); 1064 error = EMLINK; 1065 goto abortit; 1066 } 1067 if ((ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND)) 1068 || (dp->i_flags & APPEND)) { 1069 VOP_UNLOCK(fvp, 0, td); 1070 error = EPERM; 1071 goto abortit; 1072 } 1073 if ((ip->i_mode & IFMT) == IFDIR) { 1074 /* 1075 * Avoid ".", "..", and aliases of "." for obvious reasons. 1076 */ 1077 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') || 1078 dp == ip || (fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT || 1079 (ip->i_flag & IN_RENAME)) { 1080 VOP_UNLOCK(fvp, 0, td); 1081 error = EINVAL; 1082 goto abortit; 1083 } 1084 ip->i_flag |= IN_RENAME; 1085 oldparent = dp->i_number; 1086 doingdirectory = 1; 1087 } 1088 VN_KNOTE(fdvp, NOTE_WRITE); /* XXX right place? */ 1089 vrele(fdvp); 1090 1091 /* 1092 * When the target exists, both the directory 1093 * and target vnodes are returned locked. 1094 */ 1095 dp = VTOI(tdvp); 1096 xp = NULL; 1097 if (tvp) 1098 xp = VTOI(tvp); 1099 1100 /* 1101 * 1) Bump link count while we're moving stuff 1102 * around. If we crash somewhere before 1103 * completing our work, the link count 1104 * may be wrong, but correctable. 1105 */ 1106 ip->i_effnlink++; 1107 ip->i_nlink++; 1108 DIP(ip, i_nlink) = ip->i_nlink; 1109 ip->i_flag |= IN_CHANGE; 1110 if (DOINGSOFTDEP(fvp)) 1111 softdep_change_linkcnt(ip); 1112 if ((error = UFS_UPDATE(fvp, !(DOINGSOFTDEP(fvp) | 1113 DOINGASYNC(fvp)))) != 0) { 1114 VOP_UNLOCK(fvp, 0, td); 1115 goto bad; 1116 } 1117 1118 /* 1119 * If ".." must be changed (ie the directory gets a new 1120 * parent) then the source directory must not be in the 1121 * directory heirarchy above the target, as this would 1122 * orphan everything below the source directory. Also 1123 * the user must have write permission in the source so 1124 * as to be able to change "..". We must repeat the call 1125 * to namei, as the parent directory is unlocked by the 1126 * call to checkpath(). 1127 */ 1128 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread); 1129 VOP_UNLOCK(fvp, 0, td); 1130 if (oldparent != dp->i_number) 1131 newparent = dp->i_number; 1132 if (doingdirectory && newparent) { 1133 if (error) /* write access check above */ 1134 goto bad; 1135 if (xp != NULL) 1136 vput(tvp); 1137 error = ufs_checkpath(ip, dp, tcnp->cn_cred); 1138 if (error) 1139 goto out; 1140 if ((tcnp->cn_flags & SAVESTART) == 0) 1141 panic("ufs_rename: lost to startdir"); 1142 VREF(tdvp); 1143 error = relookup(tdvp, &tvp, tcnp); 1144 if (error) 1145 goto out; 1146 vrele(tdvp); 1147 dp = VTOI(tdvp); 1148 xp = NULL; 1149 if (tvp) 1150 xp = VTOI(tvp); 1151 } 1152 /* 1153 * 2) If target doesn't exist, link the target 1154 * to the source and unlink the source. 1155 * Otherwise, rewrite the target directory 1156 * entry to reference the source inode and 1157 * expunge the original entry's existence. 1158 */ 1159 if (xp == NULL) { 1160 if (dp->i_dev != ip->i_dev) 1161 panic("ufs_rename: EXDEV"); 1162 /* 1163 * Account for ".." in new directory. 1164 * When source and destination have the same 1165 * parent we don't fool with the link count. 1166 */ 1167 if (doingdirectory && newparent) { 1168 if ((nlink_t)dp->i_nlink >= LINK_MAX) { 1169 error = EMLINK; 1170 goto bad; 1171 } 1172 dp->i_effnlink++; 1173 dp->i_nlink++; 1174 DIP(dp, i_nlink) = dp->i_nlink; 1175 dp->i_flag |= IN_CHANGE; 1176 if (DOINGSOFTDEP(tdvp)) 1177 softdep_change_linkcnt(dp); 1178 error = UFS_UPDATE(tdvp, !(DOINGSOFTDEP(tdvp) | 1179 DOINGASYNC(tdvp))); 1180 if (error) 1181 goto bad; 1182 } 1183 ufs_makedirentry(ip, tcnp, &newdir); 1184 error = ufs_direnter(tdvp, NULL, &newdir, tcnp, NULL); 1185 if (error) { 1186 if (doingdirectory && newparent) { 1187 dp->i_effnlink--; 1188 dp->i_nlink--; 1189 DIP(dp, i_nlink) = dp->i_nlink; 1190 dp->i_flag |= IN_CHANGE; 1191 if (DOINGSOFTDEP(tdvp)) 1192 softdep_change_linkcnt(dp); 1193 (void)UFS_UPDATE(tdvp, 1); 1194 } 1195 goto bad; 1196 } 1197 VN_KNOTE(tdvp, NOTE_WRITE); 1198 vput(tdvp); 1199 } else { 1200 if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev) 1201 panic("ufs_rename: EXDEV"); 1202 /* 1203 * Short circuit rename(foo, foo). 1204 */ 1205 if (xp->i_number == ip->i_number) 1206 panic("ufs_rename: same file"); 1207 /* 1208 * If the parent directory is "sticky", then the caller 1209 * must possess VADMIN for the parent directory, or the 1210 * destination of the rename. This implements append-only 1211 * directories. 1212 */ 1213 if ((dp->i_mode & S_ISTXT) && 1214 VOP_ACCESS(tdvp, VADMIN, tcnp->cn_cred, td) && 1215 VOP_ACCESS(tvp, VADMIN, tcnp->cn_cred, td)) { 1216 error = EPERM; 1217 goto bad; 1218 } 1219 /* 1220 * Target must be empty if a directory and have no links 1221 * to it. Also, ensure source and target are compatible 1222 * (both directories, or both not directories). 1223 */ 1224 if ((xp->i_mode&IFMT) == IFDIR) { 1225 if ((xp->i_effnlink > 2) || 1226 !ufs_dirempty(xp, dp->i_number, tcnp->cn_cred)) { 1227 error = ENOTEMPTY; 1228 goto bad; 1229 } 1230 if (!doingdirectory) { 1231 error = ENOTDIR; 1232 goto bad; 1233 } 1234 cache_purge(tdvp); 1235 } else if (doingdirectory) { 1236 error = EISDIR; 1237 goto bad; 1238 } 1239 error = ufs_dirrewrite(dp, xp, ip->i_number, 1240 IFTODT(ip->i_mode), 1241 (doingdirectory && newparent) ? newparent : doingdirectory); 1242 if (error) 1243 goto bad; 1244 if (doingdirectory) { 1245 if (!newparent) { 1246 dp->i_effnlink--; 1247 if (DOINGSOFTDEP(tdvp)) 1248 softdep_change_linkcnt(dp); 1249 } 1250 xp->i_effnlink--; 1251 if (DOINGSOFTDEP(tvp)) 1252 softdep_change_linkcnt(xp); 1253 } 1254 if (doingdirectory && !DOINGSOFTDEP(tvp)) { 1255 /* 1256 * Truncate inode. The only stuff left in the directory 1257 * is "." and "..". The "." reference is inconsequential 1258 * since we are quashing it. We have removed the "." 1259 * reference and the reference in the parent directory, 1260 * but there may be other hard links. The soft 1261 * dependency code will arrange to do these operations 1262 * after the parent directory entry has been deleted on 1263 * disk, so when running with that code we avoid doing 1264 * them now. 1265 */ 1266 if (!newparent) { 1267 dp->i_nlink--; 1268 DIP(dp, i_nlink) = dp->i_nlink; 1269 dp->i_flag |= IN_CHANGE; 1270 } 1271 xp->i_nlink--; 1272 DIP(xp, i_nlink) = xp->i_nlink; 1273 xp->i_flag |= IN_CHANGE; 1274 ioflag = IO_NORMAL; 1275 if (DOINGASYNC(tvp)) 1276 ioflag |= IO_SYNC; 1277 if ((error = UFS_TRUNCATE(tvp, (off_t)0, ioflag, 1278 tcnp->cn_cred, tcnp->cn_thread)) != 0) 1279 goto bad; 1280 } 1281 VN_KNOTE(tdvp, NOTE_WRITE); 1282 vput(tdvp); 1283 VN_KNOTE(tvp, NOTE_DELETE); 1284 vput(tvp); 1285 xp = NULL; 1286 } 1287 1288 /* 1289 * 3) Unlink the source. 1290 */ 1291 fcnp->cn_flags &= ~MODMASK; 1292 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF; 1293 if ((fcnp->cn_flags & SAVESTART) == 0) 1294 panic("ufs_rename: lost from startdir"); 1295 VREF(fdvp); 1296 error = relookup(fdvp, &fvp, fcnp); 1297 if (error == 0) 1298 vrele(fdvp); 1299 if (fvp != NULL) { 1300 xp = VTOI(fvp); 1301 dp = VTOI(fdvp); 1302 } else { 1303 /* 1304 * From name has disappeared. IN_RENAME is not sufficient 1305 * to protect against directory races due to timing windows, 1306 * so we have to remove the panic. XXX the only real way 1307 * to solve this issue is at a much higher level. By the 1308 * time we hit ufs_rename() it's too late. 1309 */ 1310#if 0 1311 if (doingdirectory) 1312 panic("ufs_rename: lost dir entry"); 1313#endif 1314 vrele(ap->a_fvp); 1315 return (0); 1316 } 1317 /* 1318 * Ensure that the directory entry still exists and has not 1319 * changed while the new name has been entered. If the source is 1320 * a file then the entry may have been unlinked or renamed. In 1321 * either case there is no further work to be done. If the source 1322 * is a directory then it cannot have been rmdir'ed; the IN_RENAME 1323 * flag ensures that it cannot be moved by another rename or removed 1324 * by a rmdir. 1325 */ 1326 if (xp != ip) { 1327 /* 1328 * From name resolves to a different inode. IN_RENAME is 1329 * not sufficient protection against timing window races 1330 * so we can't panic here. XXX the only real way 1331 * to solve this issue is at a much higher level. By the 1332 * time we hit ufs_rename() it's too late. 1333 */ 1334#if 0 1335 if (doingdirectory) 1336 panic("ufs_rename: lost dir entry"); 1337#endif 1338 } else { 1339 /* 1340 * If the source is a directory with a 1341 * new parent, the link count of the old 1342 * parent directory must be decremented 1343 * and ".." set to point to the new parent. 1344 */ 1345 if (doingdirectory && newparent) { 1346 xp->i_offset = mastertemplate.dot_reclen; 1347 ufs_dirrewrite(xp, dp, newparent, DT_DIR, 0); 1348 cache_purge(fdvp); 1349 } 1350 error = ufs_dirremove(fdvp, xp, fcnp->cn_flags, 0); 1351 xp->i_flag &= ~IN_RENAME; 1352 } 1353 VN_KNOTE(fvp, NOTE_RENAME); 1354 if (dp) 1355 vput(fdvp); 1356 if (xp) 1357 vput(fvp); 1358 vrele(ap->a_fvp); 1359 return (error); 1360 1361bad: 1362 if (xp) 1363 vput(ITOV(xp)); 1364 vput(ITOV(dp)); 1365out: 1366 if (doingdirectory) 1367 ip->i_flag &= ~IN_RENAME; 1368 if (vn_lock(fvp, LK_EXCLUSIVE, td) == 0) { 1369 ip->i_effnlink--; 1370 ip->i_nlink--; 1371 DIP(ip, i_nlink) = ip->i_nlink; 1372 ip->i_flag |= IN_CHANGE; 1373 ip->i_flag &= ~IN_RENAME; 1374 if (DOINGSOFTDEP(fvp)) 1375 softdep_change_linkcnt(ip); 1376 vput(fvp); 1377 } else 1378 vrele(fvp); 1379 return (error); 1380} 1381 1382/* 1383 * Mkdir system call 1384 */ 1385int 1386ufs_mkdir(ap) 1387 struct vop_mkdir_args /* { 1388 struct vnode *a_dvp; 1389 struct vnode **a_vpp; 1390 struct componentname *a_cnp; 1391 struct vattr *a_vap; 1392 } */ *ap; 1393{ 1394 struct vnode *dvp = ap->a_dvp; 1395 struct vattr *vap = ap->a_vap; 1396 struct componentname *cnp = ap->a_cnp; 1397 struct inode *ip, *dp; 1398 struct vnode *tvp; 1399 struct buf *bp; 1400 struct dirtemplate dirtemplate, *dtp; 1401 struct direct newdir; 1402#ifdef UFS_ACL 1403 struct acl *acl, *dacl; 1404#endif 1405 int error, dmode; 1406 long blkoff; 1407 1408#ifdef DIAGNOSTIC 1409 if ((cnp->cn_flags & HASBUF) == 0) 1410 panic("ufs_mkdir: no name"); 1411#endif 1412 dp = VTOI(dvp); 1413 if ((nlink_t)dp->i_nlink >= LINK_MAX) { 1414 error = EMLINK; 1415 goto out; 1416 } 1417 dmode = vap->va_mode & 0777; 1418 dmode |= IFDIR; 1419 /* 1420 * Must simulate part of ufs_makeinode here to acquire the inode, 1421 * but not have it entered in the parent directory. The entry is 1422 * made later after writing "." and ".." entries. 1423 */ 1424 error = UFS_VALLOC(dvp, dmode, cnp->cn_cred, &tvp); 1425 if (error) 1426 goto out; 1427 ip = VTOI(tvp); 1428 ip->i_gid = dp->i_gid; 1429 DIP(ip, i_gid) = dp->i_gid; 1430#ifdef SUIDDIR 1431 { 1432#ifdef QUOTA 1433 struct ucred ucred, *ucp; 1434 ucp = cnp->cn_cred; 1435#endif 1436 /* 1437 * If we are hacking owners here, (only do this where told to) 1438 * and we are not giving it TO root, (would subvert quotas) 1439 * then go ahead and give it to the other user. 1440 * The new directory also inherits the SUID bit. 1441 * If user's UID and dir UID are the same, 1442 * 'give it away' so that the SUID is still forced on. 1443 */ 1444 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) && 1445 (dp->i_mode & ISUID) && dp->i_uid) { 1446 dmode |= ISUID; 1447 ip->i_uid = dp->i_uid; 1448 DIP(ip, i_uid) = dp->i_uid; 1449#ifdef QUOTA 1450 if (dp->i_uid != cnp->cn_cred->cr_uid) { 1451 /* 1452 * Make sure the correct user gets charged 1453 * for the space. 1454 * Make a dummy credential for the victim. 1455 * XXX This seems to never be accessed out of 1456 * our context so a stack variable is ok. 1457 */ 1458 ucred.cr_ref = 1; 1459 ucred.cr_uid = ip->i_uid; 1460 ucred.cr_ngroups = 1; 1461 ucred.cr_groups[0] = dp->i_gid; 1462 ucp = &ucred; 1463 } 1464#endif 1465 } else { 1466 ip->i_uid = cnp->cn_cred->cr_uid; 1467 DIP(ip, i_uid) = ip->i_uid; 1468 } 1469#ifdef QUOTA 1470 if ((error = getinoquota(ip)) || 1471 (error = chkiq(ip, 1, ucp, 0))) { 1472 UFS_VFREE(tvp, ip->i_number, dmode); 1473 vput(tvp); 1474 return (error); 1475 } 1476#endif 1477 } 1478#else /* !SUIDDIR */ 1479 ip->i_uid = cnp->cn_cred->cr_uid; 1480 DIP(ip, i_uid) = ip->i_uid; 1481#ifdef QUOTA 1482 if ((error = getinoquota(ip)) || 1483 (error = chkiq(ip, 1, cnp->cn_cred, 0))) { 1484 UFS_VFREE(tvp, ip->i_number, dmode); 1485 vput(tvp); 1486 return (error); 1487 } 1488#endif 1489#endif /* !SUIDDIR */ 1490 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 1491#ifdef UFS_ACL 1492 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK); 1493 MALLOC(dacl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK); 1494 1495 /* 1496 * Retrieve default ACL from parent, if any. 1497 */ 1498 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cnp->cn_cred, 1499 cnp->cn_thread); 1500 switch (error) { 1501 case 0: 1502 /* 1503 * Retrieved a default ACL, so merge mode and ACL if 1504 * necessary. 1505 */ 1506 if (acl->acl_cnt != 0) { 1507 /* 1508 * Two possible ways for default ACL to not be 1509 * present. First, the EA can be undefined, 1510 * or second, the default ACL can be blank. 1511 * If it's blank, fall through to the it's 1512 * not defined case. 1513 */ 1514 ip->i_mode = dmode; 1515 DIP(ip, i_mode) = dmode; 1516 *dacl = *acl; 1517 ufs_sync_acl_from_inode(ip, acl); 1518 break; 1519 } 1520 /* FALLTHROUGH */ 1521 1522 case EOPNOTSUPP: 1523 /* 1524 * Just use the mode as-is. 1525 */ 1526 ip->i_mode = dmode; 1527 DIP(ip, i_mode) = dmode; 1528 FREE(acl, M_ACL); 1529 FREE(dacl, M_ACL); 1530 dacl = acl = NULL; 1531 break; 1532 1533 default: 1534 UFS_VFREE(tvp, ip->i_number, dmode); 1535 vput(tvp); 1536 return (error); 1537 } 1538#else /* !UFS_ACL */ 1539 ip->i_mode = dmode; 1540 DIP(ip, i_mode) = dmode; 1541#endif /* !UFS_ACL */ 1542 tvp->v_type = VDIR; /* Rest init'd in getnewvnode(). */ 1543 ip->i_effnlink = 2; 1544 ip->i_nlink = 2; 1545 DIP(ip, i_nlink) = 2; 1546 if (DOINGSOFTDEP(tvp)) 1547 softdep_change_linkcnt(ip); 1548 if (cnp->cn_flags & ISWHITEOUT) { 1549 ip->i_flags |= UF_OPAQUE; 1550 DIP(ip, i_flags) = ip->i_flags; 1551 } 1552 1553 /* 1554 * Bump link count in parent directory to reflect work done below. 1555 * Should be done before reference is created so cleanup is 1556 * possible if we crash. 1557 */ 1558 dp->i_effnlink++; 1559 dp->i_nlink++; 1560 DIP(dp, i_nlink) = dp->i_nlink; 1561 dp->i_flag |= IN_CHANGE; 1562 if (DOINGSOFTDEP(dvp)) 1563 softdep_change_linkcnt(dp); 1564 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(dvp) | DOINGASYNC(dvp))); 1565 if (error) 1566 goto bad; 1567#ifdef MAC 1568 error = vop_stdcreatevnode_ea(dvp, tvp, cnp->cn_cred); 1569 if (error) 1570 goto bad; 1571#endif 1572#ifdef UFS_ACL 1573 if (acl != NULL) { 1574 /* 1575 * XXX: If we abort now, will Soft Updates notify the extattr 1576 * code that the EAs for the file need to be released? 1577 */ 1578 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cnp->cn_cred, 1579 cnp->cn_thread); 1580 if (error == 0) 1581 error = VOP_SETACL(tvp, ACL_TYPE_DEFAULT, dacl, 1582 cnp->cn_cred, cnp->cn_thread); 1583 switch (error) { 1584 case 0: 1585 break; 1586 1587 case EOPNOTSUPP: 1588 /* 1589 * XXX: This should not happen, as EOPNOTSUPP above 1590 * was supposed to free acl. 1591 */ 1592 printf("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()\n"); 1593 /* 1594 panic("ufs_mkdir: VOP_GETACL() but no VOP_SETACL()"); 1595 */ 1596 break; 1597 1598 default: 1599 FREE(acl, M_ACL); 1600 FREE(dacl, M_ACL); 1601 goto bad; 1602 } 1603 FREE(acl, M_ACL); 1604 FREE(dacl, M_ACL); 1605 } 1606#endif /* !UFS_ACL */ 1607 1608 /* 1609 * Initialize directory with "." and ".." from static template. 1610 */ 1611 if (dvp->v_mount->mnt_maxsymlinklen > 0 1612 ) 1613 dtp = &mastertemplate; 1614 else 1615 dtp = (struct dirtemplate *)&omastertemplate; 1616 dirtemplate = *dtp; 1617 dirtemplate.dot_ino = ip->i_number; 1618 dirtemplate.dotdot_ino = dp->i_number; 1619 if ((error = UFS_BALLOC(tvp, (off_t)0, DIRBLKSIZ, cnp->cn_cred, 1620 BA_CLRBUF, &bp)) != 0) 1621 goto bad; 1622 ip->i_size = DIRBLKSIZ; 1623 DIP(ip, i_size) = DIRBLKSIZ; 1624 ip->i_flag |= IN_CHANGE | IN_UPDATE; 1625 vnode_pager_setsize(tvp, (u_long)ip->i_size); 1626 bcopy((caddr_t)&dirtemplate, (caddr_t)bp->b_data, sizeof dirtemplate); 1627 if (DOINGSOFTDEP(tvp)) { 1628 /* 1629 * Ensure that the entire newly allocated block is a 1630 * valid directory so that future growth within the 1631 * block does not have to ensure that the block is 1632 * written before the inode. 1633 */ 1634 blkoff = DIRBLKSIZ; 1635 while (blkoff < bp->b_bcount) { 1636 ((struct direct *) 1637 (bp->b_data + blkoff))->d_reclen = DIRBLKSIZ; 1638 blkoff += DIRBLKSIZ; 1639 } 1640 } 1641 if ((error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | 1642 DOINGASYNC(tvp)))) != 0) { 1643 (void)BUF_WRITE(bp); 1644 goto bad; 1645 } 1646 /* 1647 * Directory set up, now install its entry in the parent directory. 1648 * 1649 * If we are not doing soft dependencies, then we must write out the 1650 * buffer containing the new directory body before entering the new 1651 * name in the parent. If we are doing soft dependencies, then the 1652 * buffer containing the new directory body will be passed to and 1653 * released in the soft dependency code after the code has attached 1654 * an appropriate ordering dependency to the buffer which ensures that 1655 * the buffer is written before the new name is written in the parent. 1656 */ 1657 if (DOINGASYNC(dvp)) 1658 bdwrite(bp); 1659 else if (!DOINGSOFTDEP(dvp) && ((error = BUF_WRITE(bp)))) 1660 goto bad; 1661 ufs_makedirentry(ip, cnp, &newdir); 1662 error = ufs_direnter(dvp, tvp, &newdir, cnp, bp); 1663 1664bad: 1665 if (error == 0) { 1666 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK); 1667 *ap->a_vpp = tvp; 1668 } else { 1669 dp->i_effnlink--; 1670 dp->i_nlink--; 1671 DIP(dp, i_nlink) = dp->i_nlink; 1672 dp->i_flag |= IN_CHANGE; 1673 if (DOINGSOFTDEP(dvp)) 1674 softdep_change_linkcnt(dp); 1675 /* 1676 * No need to do an explicit VOP_TRUNCATE here, vrele will 1677 * do this for us because we set the link count to 0. 1678 */ 1679 ip->i_effnlink = 0; 1680 ip->i_nlink = 0; 1681 DIP(ip, i_nlink) = 0; 1682 ip->i_flag |= IN_CHANGE; 1683 if (DOINGSOFTDEP(tvp)) 1684 softdep_change_linkcnt(ip); 1685 vput(tvp); 1686 } 1687out: 1688 return (error); 1689} 1690 1691/* 1692 * Rmdir system call. 1693 */ 1694int 1695ufs_rmdir(ap) 1696 struct vop_rmdir_args /* { 1697 struct vnode *a_dvp; 1698 struct vnode *a_vp; 1699 struct componentname *a_cnp; 1700 } */ *ap; 1701{ 1702 struct vnode *vp = ap->a_vp; 1703 struct vnode *dvp = ap->a_dvp; 1704 struct componentname *cnp = ap->a_cnp; 1705 struct inode *ip, *dp; 1706 int error, ioflag; 1707 1708 ip = VTOI(vp); 1709 dp = VTOI(dvp); 1710 1711 /* 1712 * Do not remove a directory that is in the process of being renamed. 1713 * Verify the directory is empty (and valid). Rmdir ".." will not be 1714 * valid since ".." will contain a reference to the current directory 1715 * and thus be non-empty. Do not allow the removal of mounted on 1716 * directories (this can happen when an NFS exported filesystem 1717 * tries to remove a locally mounted on directory). 1718 */ 1719 error = 0; 1720 if (ip->i_flag & IN_RENAME) { 1721 error = EINVAL; 1722 goto out; 1723 } 1724 if (ip->i_effnlink != 2 || 1725 !ufs_dirempty(ip, dp->i_number, cnp->cn_cred)) { 1726 error = ENOTEMPTY; 1727 goto out; 1728 } 1729 if ((dp->i_flags & APPEND) 1730 || (ip->i_flags & (NOUNLINK | IMMUTABLE | APPEND))) { 1731 error = EPERM; 1732 goto out; 1733 } 1734 if (vp->v_mountedhere != 0) { 1735 error = EINVAL; 1736 goto out; 1737 } 1738 /* 1739 * Delete reference to directory before purging 1740 * inode. If we crash in between, the directory 1741 * will be reattached to lost+found, 1742 */ 1743 dp->i_effnlink--; 1744 ip->i_effnlink--; 1745 if (DOINGSOFTDEP(vp)) { 1746 softdep_change_linkcnt(dp); 1747 softdep_change_linkcnt(ip); 1748 } 1749 error = ufs_dirremove(dvp, ip, cnp->cn_flags, 1); 1750 if (error) { 1751 dp->i_effnlink++; 1752 ip->i_effnlink++; 1753 if (DOINGSOFTDEP(vp)) { 1754 softdep_change_linkcnt(dp); 1755 softdep_change_linkcnt(ip); 1756 } 1757 goto out; 1758 } 1759 VN_KNOTE(dvp, NOTE_WRITE | NOTE_LINK); 1760 cache_purge(dvp); 1761 /* 1762 * Truncate inode. The only stuff left in the directory is "." and 1763 * "..". The "." reference is inconsequential since we are quashing 1764 * it. The soft dependency code will arrange to do these operations 1765 * after the parent directory entry has been deleted on disk, so 1766 * when running with that code we avoid doing them now. 1767 */ 1768 if (!DOINGSOFTDEP(vp)) { 1769 dp->i_nlink--; 1770 DIP(dp, i_nlink) = dp->i_nlink; 1771 dp->i_flag |= IN_CHANGE; 1772 ip->i_nlink--; 1773 DIP(ip, i_nlink) = ip->i_nlink; 1774 ip->i_flag |= IN_CHANGE; 1775 ioflag = IO_NORMAL; 1776 if (DOINGASYNC(vp)) 1777 ioflag |= IO_SYNC; 1778 error = UFS_TRUNCATE(vp, (off_t)0, ioflag, cnp->cn_cred, 1779 cnp->cn_thread); 1780 } 1781 cache_purge(vp); 1782#ifdef UFS_DIRHASH 1783 /* Kill any active hash; i_effnlink == 0, so it will not come back. */ 1784 if (ip->i_dirhash != NULL) 1785 ufsdirhash_free(ip); 1786#endif 1787out: 1788 VN_KNOTE(vp, NOTE_DELETE); 1789 return (error); 1790} 1791 1792/* 1793 * symlink -- make a symbolic link 1794 */ 1795int 1796ufs_symlink(ap) 1797 struct vop_symlink_args /* { 1798 struct vnode *a_dvp; 1799 struct vnode **a_vpp; 1800 struct componentname *a_cnp; 1801 struct vattr *a_vap; 1802 char *a_target; 1803 } */ *ap; 1804{ 1805 struct vnode *vp, **vpp = ap->a_vpp; 1806 struct inode *ip; 1807 int len, error; 1808 1809 error = ufs_makeinode(IFLNK | ap->a_vap->va_mode, ap->a_dvp, 1810 vpp, ap->a_cnp); 1811 if (error) 1812 return (error); 1813 VN_KNOTE(ap->a_dvp, NOTE_WRITE); 1814 vp = *vpp; 1815 len = strlen(ap->a_target); 1816 if (len < vp->v_mount->mnt_maxsymlinklen) { 1817 ip = VTOI(vp); 1818 bcopy(ap->a_target, SHORTLINK(ip), len); 1819 ip->i_size = len; 1820 DIP(ip, i_size) = len; 1821 ip->i_flag |= IN_CHANGE | IN_UPDATE; 1822 } else 1823 error = vn_rdwr(UIO_WRITE, vp, ap->a_target, len, (off_t)0, 1824 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK, 1825 ap->a_cnp->cn_cred, (int *)0, (struct thread *)0); 1826 if (error) 1827 vput(vp); 1828 return (error); 1829} 1830 1831/* 1832 * Vnode op for reading directories. 1833 * 1834 * The routine below assumes that the on-disk format of a directory 1835 * is the same as that defined by <sys/dirent.h>. If the on-disk 1836 * format changes, then it will be necessary to do a conversion 1837 * from the on-disk format that read returns to the format defined 1838 * by <sys/dirent.h>. 1839 */ 1840int 1841ufs_readdir(ap) 1842 struct vop_readdir_args /* { 1843 struct vnode *a_vp; 1844 struct uio *a_uio; 1845 struct ucred *a_cred; 1846 int *a_eofflag; 1847 int *ncookies; 1848 u_long **a_cookies; 1849 } */ *ap; 1850{ 1851 struct uio *uio = ap->a_uio; 1852 int error; 1853 size_t count, lost; 1854 off_t off; 1855 1856 if (ap->a_ncookies != NULL) 1857 /* 1858 * Ensure that the block is aligned. The caller can use 1859 * the cookies to determine where in the block to start. 1860 */ 1861 uio->uio_offset &= ~(DIRBLKSIZ - 1); 1862 off = uio->uio_offset; 1863 count = uio->uio_resid; 1864 /* Make sure we don't return partial entries. */ 1865 if (count <= ((uio->uio_offset + count) & (DIRBLKSIZ -1))) 1866 return (EINVAL); 1867 count -= (uio->uio_offset + count) & (DIRBLKSIZ -1); 1868 lost = uio->uio_resid - count; 1869 uio->uio_resid = count; 1870 uio->uio_iov->iov_len = count; 1871# if (BYTE_ORDER == LITTLE_ENDIAN) 1872 if (ap->a_vp->v_mount->mnt_maxsymlinklen > 0) { 1873 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred); 1874 } else { 1875 struct dirent *dp, *edp; 1876 struct uio auio; 1877 struct iovec aiov; 1878 caddr_t dirbuf; 1879 int readcnt; 1880 u_char tmp; 1881 1882 auio = *uio; 1883 auio.uio_iov = &aiov; 1884 auio.uio_iovcnt = 1; 1885 auio.uio_segflg = UIO_SYSSPACE; 1886 aiov.iov_len = count; 1887 MALLOC(dirbuf, caddr_t, count, M_TEMP, M_WAITOK); 1888 aiov.iov_base = dirbuf; 1889 error = VOP_READ(ap->a_vp, &auio, 0, ap->a_cred); 1890 if (error == 0) { 1891 readcnt = count - auio.uio_resid; 1892 edp = (struct dirent *)&dirbuf[readcnt]; 1893 for (dp = (struct dirent *)dirbuf; dp < edp; ) { 1894 tmp = dp->d_namlen; 1895 dp->d_namlen = dp->d_type; 1896 dp->d_type = tmp; 1897 if (dp->d_reclen > 0) { 1898 dp = (struct dirent *) 1899 ((char *)dp + dp->d_reclen); 1900 } else { 1901 error = EIO; 1902 break; 1903 } 1904 } 1905 if (dp >= edp) 1906 error = uiomove(dirbuf, readcnt, uio); 1907 } 1908 FREE(dirbuf, M_TEMP); 1909 } 1910# else 1911 error = VOP_READ(ap->a_vp, uio, 0, ap->a_cred); 1912# endif 1913 if (!error && ap->a_ncookies != NULL) { 1914 struct dirent* dpStart; 1915 struct dirent* dpEnd; 1916 struct dirent* dp; 1917 int ncookies; 1918 u_long *cookies; 1919 u_long *cookiep; 1920 1921 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) 1922 panic("ufs_readdir: unexpected uio from NFS server"); 1923 dpStart = (struct dirent *) 1924 (uio->uio_iov->iov_base - (uio->uio_offset - off)); 1925 dpEnd = (struct dirent *) uio->uio_iov->iov_base; 1926 for (dp = dpStart, ncookies = 0; 1927 dp < dpEnd; 1928 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) 1929 ncookies++; 1930 MALLOC(cookies, u_long *, ncookies * sizeof(u_long), M_TEMP, 1931 M_WAITOK); 1932 for (dp = dpStart, cookiep = cookies; 1933 dp < dpEnd; 1934 dp = (struct dirent *)((caddr_t) dp + dp->d_reclen)) { 1935 off += dp->d_reclen; 1936 *cookiep++ = (u_long) off; 1937 } 1938 *ap->a_ncookies = ncookies; 1939 *ap->a_cookies = cookies; 1940 } 1941 uio->uio_resid += lost; 1942 if (ap->a_eofflag) 1943 *ap->a_eofflag = VTOI(ap->a_vp)->i_size <= uio->uio_offset; 1944 return (error); 1945} 1946 1947/* 1948 * Return target name of a symbolic link 1949 */ 1950int 1951ufs_readlink(ap) 1952 struct vop_readlink_args /* { 1953 struct vnode *a_vp; 1954 struct uio *a_uio; 1955 struct ucred *a_cred; 1956 } */ *ap; 1957{ 1958 struct vnode *vp = ap->a_vp; 1959 struct inode *ip = VTOI(vp); 1960 doff_t isize; 1961 1962 isize = ip->i_size; 1963 if ((isize < vp->v_mount->mnt_maxsymlinklen) || 1964 DIP(ip, i_blocks) == 0) { /* XXX - for old fastlink support */ 1965 uiomove(SHORTLINK(ip), isize, ap->a_uio); 1966 return (0); 1967 } 1968 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred)); 1969} 1970 1971/* 1972 * Calculate the logical to physical mapping if not done already, 1973 * then call the device strategy routine. 1974 * 1975 * In order to be able to swap to a file, the ufs_bmaparray() operation may not 1976 * deadlock on memory. See ufs_bmap() for details. 1977 */ 1978int 1979ufs_strategy(ap) 1980 struct vop_strategy_args /* { 1981 struct vnode *a_vp; 1982 struct buf *a_bp; 1983 } */ *ap; 1984{ 1985 struct buf *bp = ap->a_bp; 1986 struct vnode *vp = ap->a_vp; 1987 struct inode *ip; 1988 ufs2_daddr_t blkno; 1989 int error; 1990 1991 ip = VTOI(vp); 1992 if (vp->v_type == VBLK || vp->v_type == VCHR) 1993 panic("ufs_strategy: spec"); 1994 if (bp->b_blkno == bp->b_lblkno) { 1995 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, bp, NULL, NULL); 1996 bp->b_blkno = blkno; 1997 if (error) { 1998 bp->b_error = error; 1999 bp->b_ioflags |= BIO_ERROR; 2000 bufdone(bp); 2001 return (error); 2002 } 2003 if ((long)bp->b_blkno == -1) 2004 vfs_bio_clrbuf(bp); 2005 } 2006 if ((long)bp->b_blkno == -1) { 2007 bufdone(bp); 2008 return (0); 2009 } 2010 vp = ip->i_devvp; 2011 bp->b_dev = vp->v_rdev; 2012 VOP_STRATEGY(vp, bp); 2013 return (0); 2014} 2015 2016/* 2017 * Print out the contents of an inode. 2018 */ 2019int 2020ufs_print(ap) 2021 struct vop_print_args /* { 2022 struct vnode *a_vp; 2023 } */ *ap; 2024{ 2025 struct vnode *vp = ap->a_vp; 2026 struct inode *ip = VTOI(vp); 2027 2028 printf("tag VT_UFS, ino %lu, on dev %s (%d, %d)", 2029 (u_long)ip->i_number, devtoname(ip->i_dev), major(ip->i_dev), 2030 minor(ip->i_dev)); 2031 if (vp->v_type == VFIFO) 2032 fifo_printinfo(vp); 2033 lockmgr_printinfo(&vp->v_lock); 2034 printf("\n"); 2035 return (0); 2036} 2037 2038/* 2039 * Read wrapper for special devices. 2040 */ 2041int 2042ufsspec_read(ap) 2043 struct vop_read_args /* { 2044 struct vnode *a_vp; 2045 struct uio *a_uio; 2046 int a_ioflag; 2047 struct ucred *a_cred; 2048 } */ *ap; 2049{ 2050 int error, resid; 2051 struct inode *ip; 2052 struct uio *uio; 2053 2054 uio = ap->a_uio; 2055 resid = uio->uio_resid; 2056 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap); 2057 /* 2058 * The inode may have been revoked during the call, so it must not 2059 * be accessed blindly here or in the other wrapper functions. 2060 */ 2061 ip = VTOI(ap->a_vp); 2062 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0))) 2063 ip->i_flag |= IN_ACCESS; 2064 return (error); 2065} 2066 2067/* 2068 * Write wrapper for special devices. 2069 */ 2070int 2071ufsspec_write(ap) 2072 struct vop_write_args /* { 2073 struct vnode *a_vp; 2074 struct uio *a_uio; 2075 int a_ioflag; 2076 struct ucred *a_cred; 2077 } */ *ap; 2078{ 2079 int error, resid; 2080 struct inode *ip; 2081 struct uio *uio; 2082 2083 uio = ap->a_uio; 2084 resid = uio->uio_resid; 2085 error = VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap); 2086 ip = VTOI(ap->a_vp); 2087 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0))) 2088 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE; 2089 return (error); 2090} 2091 2092/* 2093 * Close wrapper for special devices. 2094 * 2095 * Update the times on the inode then do device close. 2096 */ 2097int 2098ufsspec_close(ap) 2099 struct vop_close_args /* { 2100 struct vnode *a_vp; 2101 int a_fflag; 2102 struct ucred *a_cred; 2103 struct thread *a_td; 2104 } */ *ap; 2105{ 2106 struct vnode *vp = ap->a_vp; 2107 2108 mtx_lock(&vp->v_interlock); 2109 if (vp->v_usecount > 1) 2110 ufs_itimes(vp); 2111 mtx_unlock(&vp->v_interlock); 2112 return (VOCALL(spec_vnodeop_p, VOFFSET(vop_close), ap)); 2113} 2114 2115/* 2116 * Read wrapper for fifos. 2117 */ 2118int 2119ufsfifo_read(ap) 2120 struct vop_read_args /* { 2121 struct vnode *a_vp; 2122 struct uio *a_uio; 2123 int a_ioflag; 2124 struct ucred *a_cred; 2125 } */ *ap; 2126{ 2127 int error, resid; 2128 struct inode *ip; 2129 struct uio *uio; 2130 2131 uio = ap->a_uio; 2132 resid = uio->uio_resid; 2133 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap); 2134 ip = VTOI(ap->a_vp); 2135 if ((ap->a_vp->v_mount->mnt_flag & MNT_NOATIME) == 0 && ip != NULL && 2136 (uio->uio_resid != resid || (error == 0 && resid != 0))) 2137 VTOI(ap->a_vp)->i_flag |= IN_ACCESS; 2138 return (error); 2139} 2140 2141/* 2142 * Write wrapper for fifos. 2143 */ 2144int 2145ufsfifo_write(ap) 2146 struct vop_write_args /* { 2147 struct vnode *a_vp; 2148 struct uio *a_uio; 2149 int a_ioflag; 2150 struct ucred *a_cred; 2151 } */ *ap; 2152{ 2153 int error, resid; 2154 struct inode *ip; 2155 struct uio *uio; 2156 2157 uio = ap->a_uio; 2158 resid = uio->uio_resid; 2159 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap); 2160 ip = VTOI(ap->a_vp); 2161 if (ip != NULL && (uio->uio_resid != resid || (error == 0 && resid != 0))) 2162 VTOI(ap->a_vp)->i_flag |= IN_CHANGE | IN_UPDATE; 2163 return (error); 2164} 2165 2166/* 2167 * Close wrapper for fifos. 2168 * 2169 * Update the times on the inode then do device close. 2170 */ 2171int 2172ufsfifo_close(ap) 2173 struct vop_close_args /* { 2174 struct vnode *a_vp; 2175 int a_fflag; 2176 struct ucred *a_cred; 2177 struct thread *a_td; 2178 } */ *ap; 2179{ 2180 struct vnode *vp = ap->a_vp; 2181 2182 mtx_lock(&vp->v_interlock); 2183 if (vp->v_usecount > 1) 2184 ufs_itimes(vp); 2185 mtx_unlock(&vp->v_interlock); 2186 return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_close), ap)); 2187} 2188 2189/* 2190 * Kqfilter wrapper for fifos. 2191 * 2192 * Fall through to ufs kqfilter routines if needed 2193 */ 2194int 2195ufsfifo_kqfilter(ap) 2196 struct vop_kqfilter_args *ap; 2197{ 2198 int error; 2199 2200 error = VOCALL(fifo_vnodeop_p, VOFFSET(vop_kqfilter), ap); 2201 if (error) 2202 error = ufs_kqfilter(ap); 2203 return (error); 2204} 2205 2206/* 2207 * Return POSIX pathconf information applicable to ufs filesystems. 2208 */ 2209int 2210ufs_pathconf(ap) 2211 struct vop_pathconf_args /* { 2212 struct vnode *a_vp; 2213 int a_name; 2214 int *a_retval; 2215 } */ *ap; 2216{ 2217 2218 switch (ap->a_name) { 2219 case _PC_LINK_MAX: 2220 *ap->a_retval = LINK_MAX; 2221 return (0); 2222 case _PC_NAME_MAX: 2223 *ap->a_retval = NAME_MAX; 2224 return (0); 2225 case _PC_PATH_MAX: 2226 *ap->a_retval = PATH_MAX; 2227 return (0); 2228 case _PC_PIPE_BUF: 2229 *ap->a_retval = PIPE_BUF; 2230 return (0); 2231 case _PC_CHOWN_RESTRICTED: 2232 *ap->a_retval = 1; 2233 return (0); 2234 case _PC_NO_TRUNC: 2235 *ap->a_retval = 1; 2236 return (0); 2237 default: 2238 return (EINVAL); 2239 } 2240 /* NOTREACHED */ 2241} 2242 2243/* 2244 * Advisory record locking support 2245 */ 2246int 2247ufs_advlock(ap) 2248 struct vop_advlock_args /* { 2249 struct vnode *a_vp; 2250 caddr_t a_id; 2251 int a_op; 2252 struct flock *a_fl; 2253 int a_flags; 2254 } */ *ap; 2255{ 2256 struct inode *ip = VTOI(ap->a_vp); 2257 2258 return (lf_advlock(ap, &(ip->i_lockf), ip->i_size)); 2259} 2260 2261/* 2262 * Initialize the vnode associated with a new inode, handle aliased 2263 * vnodes. 2264 */ 2265int 2266ufs_vinit(mntp, specops, fifoops, vpp) 2267 struct mount *mntp; 2268 vop_t **specops; 2269 vop_t **fifoops; 2270 struct vnode **vpp; 2271{ 2272 struct inode *ip; 2273 struct vnode *vp; 2274 struct timeval tv; 2275 2276 vp = *vpp; 2277 ip = VTOI(vp); 2278 switch(vp->v_type = IFTOVT(ip->i_mode)) { 2279 case VCHR: 2280 case VBLK: 2281 vp->v_op = specops; 2282 vp = addaliasu(vp, DIP(ip, i_rdev)); 2283 ip->i_vnode = vp; 2284 break; 2285 case VFIFO: 2286 vp->v_op = fifoops; 2287 break; 2288 default: 2289 break; 2290 2291 } 2292 ASSERT_VOP_LOCKED(vp, "ufs_vinit"); 2293 if (ip->i_number == ROOTINO) 2294 vp->v_vflag |= VV_ROOT; 2295 /* 2296 * Initialize modrev times 2297 */ 2298 getmicrouptime(&tv); 2299 SETHIGH(ip->i_modrev, tv.tv_sec); 2300 SETLOW(ip->i_modrev, tv.tv_usec * 4294); 2301 *vpp = vp; 2302 return (0); 2303} 2304 2305/* 2306 * Allocate a new inode. 2307 * Vnode dvp must be locked. 2308 */ 2309int 2310ufs_makeinode(mode, dvp, vpp, cnp) 2311 int mode; 2312 struct vnode *dvp; 2313 struct vnode **vpp; 2314 struct componentname *cnp; 2315{ 2316 struct inode *ip, *pdir; 2317 struct direct newdir; 2318 struct vnode *tvp; 2319#ifdef UFS_ACL 2320 struct acl *acl; 2321#endif 2322 int error; 2323 2324 pdir = VTOI(dvp); 2325#ifdef DIAGNOSTIC 2326 if ((cnp->cn_flags & HASBUF) == 0) 2327 panic("ufs_makeinode: no name"); 2328#endif 2329 *vpp = NULL; 2330 if ((mode & IFMT) == 0) 2331 mode |= IFREG; 2332 2333 error = UFS_VALLOC(dvp, mode, cnp->cn_cred, &tvp); 2334 if (error) 2335 return (error); 2336 ip = VTOI(tvp); 2337 ip->i_gid = pdir->i_gid; 2338 DIP(ip, i_gid) = pdir->i_gid; 2339#ifdef SUIDDIR 2340 { 2341#ifdef QUOTA 2342 struct ucred ucred, *ucp; 2343 ucp = cnp->cn_cred; 2344#endif 2345 /* 2346 * If we are not the owner of the directory, 2347 * and we are hacking owners here, (only do this where told to) 2348 * and we are not giving it TO root, (would subvert quotas) 2349 * then go ahead and give it to the other user. 2350 * Note that this drops off the execute bits for security. 2351 */ 2352 if ((dvp->v_mount->mnt_flag & MNT_SUIDDIR) && 2353 (pdir->i_mode & ISUID) && 2354 (pdir->i_uid != cnp->cn_cred->cr_uid) && pdir->i_uid) { 2355 ip->i_uid = pdir->i_uid; 2356 DIP(ip, i_uid) = ip->i_uid; 2357 mode &= ~07111; 2358#ifdef QUOTA 2359 /* 2360 * Make sure the correct user gets charged 2361 * for the space. 2362 * Quickly knock up a dummy credential for the victim. 2363 * XXX This seems to never be accessed out of our 2364 * context so a stack variable is ok. 2365 */ 2366 ucred.cr_ref = 1; 2367 ucred.cr_uid = ip->i_uid; 2368 ucred.cr_ngroups = 1; 2369 ucred.cr_groups[0] = pdir->i_gid; 2370 ucp = &ucred; 2371#endif 2372 } else { 2373 ip->i_uid = cnp->cn_cred->cr_uid; 2374 DIP(ip, i_uid) = ip->i_uid; 2375 } 2376 2377#ifdef QUOTA 2378 if ((error = getinoquota(ip)) || 2379 (error = chkiq(ip, 1, ucp, 0))) { 2380 UFS_VFREE(tvp, ip->i_number, mode); 2381 vput(tvp); 2382 return (error); 2383 } 2384#endif 2385 } 2386#else /* !SUIDDIR */ 2387 ip->i_uid = cnp->cn_cred->cr_uid; 2388 DIP(ip, i_uid) = ip->i_uid; 2389#ifdef QUOTA 2390 if ((error = getinoquota(ip)) || 2391 (error = chkiq(ip, 1, cnp->cn_cred, 0))) { 2392 UFS_VFREE(tvp, ip->i_number, mode); 2393 vput(tvp); 2394 return (error); 2395 } 2396#endif 2397#endif /* !SUIDDIR */ 2398 ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE; 2399#ifdef UFS_ACL 2400 MALLOC(acl, struct acl *, sizeof(*acl), M_ACL, M_WAITOK); 2401 /* 2402 * Retrieve default ACL for parent, if any. 2403 */ 2404 error = VOP_GETACL(dvp, ACL_TYPE_DEFAULT, acl, cnp->cn_cred, 2405 cnp->cn_thread); 2406 switch (error) { 2407 case 0: 2408 /* 2409 * Retrieved a default ACL, so merge mode and ACL if 2410 * necessary. 2411 */ 2412 if (acl->acl_cnt != 0) { 2413 /* 2414 * Two possible ways for default ACL to not be 2415 * present. First, the EA can be undefined, 2416 * or second, the default ACL can be blank. 2417 * If it's blank, fall through to the it's 2418 * not defined case. 2419 */ 2420 ip->i_mode = mode; 2421 DIP(ip, i_mode) = mode; 2422 ufs_sync_acl_from_inode(ip, acl); 2423 break; 2424 } 2425 2426 case EOPNOTSUPP: 2427 /* 2428 * Just use the mode as-is. 2429 */ 2430 ip->i_mode = mode; 2431 DIP(ip, i_mode) = mode; 2432 FREE(acl, M_ACL); 2433 acl = NULL; 2434 break; 2435 2436 default: 2437 UFS_VFREE(tvp, ip->i_number, mode); 2438 vput(tvp); 2439 return (error); 2440 } 2441#else /* !UFS_ACL */ 2442 ip->i_mode = mode; 2443 DIP(ip, i_mode) = mode; 2444#endif /* !UFS_ACL */ 2445 tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */ 2446 ip->i_effnlink = 1; 2447 ip->i_nlink = 1; 2448 DIP(ip, i_nlink) = 1; 2449 if (DOINGSOFTDEP(tvp)) 2450 softdep_change_linkcnt(ip); 2451 if ((ip->i_mode & ISGID) && !groupmember(ip->i_gid, cnp->cn_cred) && 2452 suser_cred(cnp->cn_cred, PRISON_ROOT)) { 2453 ip->i_mode &= ~ISGID; 2454 DIP(ip, i_mode) = ip->i_mode; 2455 } 2456 2457 if (cnp->cn_flags & ISWHITEOUT) { 2458 ip->i_flags |= UF_OPAQUE; 2459 DIP(ip, i_flags) = ip->i_flags; 2460 } 2461 2462 /* 2463 * Make sure inode goes to disk before directory entry. 2464 */ 2465 error = UFS_UPDATE(tvp, !(DOINGSOFTDEP(tvp) | DOINGASYNC(tvp))); 2466 if (error) 2467 goto bad; 2468#ifdef MAC 2469 error = vop_stdcreatevnode_ea(dvp, tvp, cnp->cn_cred); 2470 if (error) 2471 goto bad; 2472#endif 2473#ifdef UFS_ACL 2474 if (acl != NULL) { 2475 /* 2476 * XXX: If we abort now, will Soft Updates notify the extattr 2477 * code that the EAs for the file need to be released? 2478 */ 2479 error = VOP_SETACL(tvp, ACL_TYPE_ACCESS, acl, cnp->cn_cred, 2480 cnp->cn_thread); 2481 switch (error) { 2482 case 0: 2483 break; 2484 2485 case EOPNOTSUPP: 2486 /* 2487 * XXX: This should not happen, as EOPNOTSUPP above was 2488 * supposed to free acl. 2489 */ 2490 printf("ufs_makeinode: VOP_GETACL() but no " 2491 "VOP_SETACL()\n"); 2492 /* panic("ufs_makeinode: VOP_GETACL() but no " 2493 "VOP_SETACL()"); */ 2494 break; 2495 2496 default: 2497 FREE(acl, M_ACL); 2498 goto bad; 2499 } 2500 FREE(acl, M_ACL); 2501 } 2502#endif /* !UFS_ACL */ 2503 ufs_makedirentry(ip, cnp, &newdir); 2504 error = ufs_direnter(dvp, tvp, &newdir, cnp, NULL); 2505 if (error) 2506 goto bad; 2507 *vpp = tvp; 2508 return (0); 2509 2510bad: 2511 /* 2512 * Write error occurred trying to update the inode 2513 * or the directory so must deallocate the inode. 2514 */ 2515 ip->i_effnlink = 0; 2516 ip->i_nlink = 0; 2517 DIP(ip, i_nlink) = 0; 2518 ip->i_flag |= IN_CHANGE; 2519 if (DOINGSOFTDEP(tvp)) 2520 softdep_change_linkcnt(ip); 2521 vput(tvp); 2522 return (error); 2523} 2524 2525static struct filterops ufsread_filtops = 2526 { 1, NULL, filt_ufsdetach, filt_ufsread }; 2527static struct filterops ufswrite_filtops = 2528 { 1, NULL, filt_ufsdetach, filt_ufswrite }; 2529static struct filterops ufsvnode_filtops = 2530 { 1, NULL, filt_ufsdetach, filt_ufsvnode }; 2531 2532static int 2533ufs_kqfilter(ap) 2534 struct vop_kqfilter_args /* { 2535 struct vnode *a_vp; 2536 struct knote *a_kn; 2537 } */ *ap; 2538{ 2539 struct vnode *vp = ap->a_vp; 2540 struct knote *kn = ap->a_kn; 2541 2542 switch (kn->kn_filter) { 2543 case EVFILT_READ: 2544 kn->kn_fop = &ufsread_filtops; 2545 break; 2546 case EVFILT_WRITE: 2547 kn->kn_fop = &ufswrite_filtops; 2548 break; 2549 case EVFILT_VNODE: 2550 kn->kn_fop = &ufsvnode_filtops; 2551 break; 2552 default: 2553 return (1); 2554 } 2555 2556 kn->kn_hook = (caddr_t)vp; 2557 2558 if (vp->v_pollinfo == NULL) 2559 v_addpollinfo(vp); 2560 mtx_lock(&vp->v_pollinfo->vpi_lock); 2561 SLIST_INSERT_HEAD(&vp->v_pollinfo->vpi_selinfo.si_note, kn, kn_selnext); 2562 mtx_unlock(&vp->v_pollinfo->vpi_lock); 2563 2564 return (0); 2565} 2566 2567static void 2568filt_ufsdetach(struct knote *kn) 2569{ 2570 struct vnode *vp = (struct vnode *)kn->kn_hook; 2571 2572 KASSERT(vp->v_pollinfo != NULL, ("Mising v_pollinfo")); 2573 mtx_lock(&vp->v_pollinfo->vpi_lock); 2574 SLIST_REMOVE(&vp->v_pollinfo->vpi_selinfo.si_note, 2575 kn, knote, kn_selnext); 2576 mtx_unlock(&vp->v_pollinfo->vpi_lock); 2577} 2578 2579/*ARGSUSED*/ 2580static int 2581filt_ufsread(struct knote *kn, long hint) 2582{ 2583 struct vnode *vp = (struct vnode *)kn->kn_hook; 2584 struct inode *ip = VTOI(vp); 2585 2586 /* 2587 * filesystem is gone, so set the EOF flag and schedule 2588 * the knote for deletion. 2589 */ 2590 if (hint == NOTE_REVOKE) { 2591 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 2592 return (1); 2593 } 2594 2595 kn->kn_data = ip->i_size - kn->kn_fp->f_offset; 2596 return (kn->kn_data != 0); 2597} 2598 2599/*ARGSUSED*/ 2600static int 2601filt_ufswrite(struct knote *kn, long hint) 2602{ 2603 2604 /* 2605 * filesystem is gone, so set the EOF flag and schedule 2606 * the knote for deletion. 2607 */ 2608 if (hint == NOTE_REVOKE) 2609 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 2610 2611 kn->kn_data = 0; 2612 return (1); 2613} 2614 2615static int 2616filt_ufsvnode(struct knote *kn, long hint) 2617{ 2618 2619 if (kn->kn_sfflags & hint) 2620 kn->kn_fflags |= hint; 2621 if (hint == NOTE_REVOKE) { 2622 kn->kn_flags |= EV_EOF; 2623 return (1); 2624 } 2625 return (kn->kn_fflags != 0); 2626} 2627 2628/* Global vfs data structures for ufs. */ 2629static vop_t **ufs_vnodeop_p; 2630static struct vnodeopv_entry_desc ufs_vnodeop_entries[] = { 2631 { &vop_default_desc, (vop_t *) vop_defaultop }, 2632 { &vop_fsync_desc, (vop_t *) vop_panic }, 2633 { &vop_read_desc, (vop_t *) vop_panic }, 2634 { &vop_reallocblks_desc, (vop_t *) vop_panic }, 2635 { &vop_write_desc, (vop_t *) vop_panic }, 2636 { &vop_access_desc, (vop_t *) ufs_access }, 2637 { &vop_advlock_desc, (vop_t *) ufs_advlock }, 2638 { &vop_bmap_desc, (vop_t *) ufs_bmap }, 2639 { &vop_cachedlookup_desc, (vop_t *) ufs_lookup }, 2640 { &vop_close_desc, (vop_t *) ufs_close }, 2641 { &vop_create_desc, (vop_t *) ufs_create }, 2642 { &vop_getattr_desc, (vop_t *) ufs_getattr }, 2643 { &vop_inactive_desc, (vop_t *) ufs_inactive }, 2644 { &vop_islocked_desc, (vop_t *) vop_stdislocked }, 2645 { &vop_link_desc, (vop_t *) ufs_link }, 2646 { &vop_lock_desc, (vop_t *) vop_stdlock }, 2647 { &vop_lookup_desc, (vop_t *) vfs_cache_lookup }, 2648 { &vop_mkdir_desc, (vop_t *) ufs_mkdir }, 2649 { &vop_mknod_desc, (vop_t *) ufs_mknod }, 2650 { &vop_open_desc, (vop_t *) ufs_open }, 2651 { &vop_pathconf_desc, (vop_t *) ufs_pathconf }, 2652 { &vop_poll_desc, (vop_t *) vop_stdpoll }, 2653 { &vop_kqfilter_desc, (vop_t *) ufs_kqfilter }, 2654 { &vop_getwritemount_desc, (vop_t *) vop_stdgetwritemount }, 2655 { &vop_print_desc, (vop_t *) ufs_print }, 2656 { &vop_readdir_desc, (vop_t *) ufs_readdir }, 2657 { &vop_readlink_desc, (vop_t *) ufs_readlink }, 2658 { &vop_reclaim_desc, (vop_t *) ufs_reclaim }, 2659#ifdef MAC 2660 { &vop_refreshlabel_desc, (vop_t *) vop_stdrefreshlabel_ea }, 2661#endif 2662 { &vop_remove_desc, (vop_t *) ufs_remove }, 2663 { &vop_rename_desc, (vop_t *) ufs_rename }, 2664 { &vop_rmdir_desc, (vop_t *) ufs_rmdir }, 2665 { &vop_setattr_desc, (vop_t *) ufs_setattr }, 2666#ifdef MAC 2667 { &vop_setlabel_desc, (vop_t *) vop_stdsetlabel_ea }, 2668#endif 2669 { &vop_strategy_desc, (vop_t *) ufs_strategy }, 2670 { &vop_symlink_desc, (vop_t *) ufs_symlink }, 2671 { &vop_unlock_desc, (vop_t *) vop_stdunlock }, 2672 { &vop_whiteout_desc, (vop_t *) ufs_whiteout }, 2673#ifdef UFS_EXTATTR 2674 { &vop_getextattr_desc, (vop_t *) ufs_getextattr }, 2675 { &vop_setextattr_desc, (vop_t *) ufs_setextattr }, 2676#endif 2677#ifdef UFS_ACL 2678 { &vop_getacl_desc, (vop_t *) ufs_getacl }, 2679 { &vop_setacl_desc, (vop_t *) ufs_setacl }, 2680 { &vop_aclcheck_desc, (vop_t *) ufs_aclcheck }, 2681#endif 2682 { NULL, NULL } 2683}; 2684static struct vnodeopv_desc ufs_vnodeop_opv_desc = 2685 { &ufs_vnodeop_p, ufs_vnodeop_entries }; 2686 2687static vop_t **ufs_specop_p; 2688static struct vnodeopv_entry_desc ufs_specop_entries[] = { 2689 { &vop_default_desc, (vop_t *) spec_vnoperate }, 2690 { &vop_fsync_desc, (vop_t *) vop_panic }, 2691 { &vop_access_desc, (vop_t *) ufs_access }, 2692 { &vop_close_desc, (vop_t *) ufsspec_close }, 2693 { &vop_getattr_desc, (vop_t *) ufs_getattr }, 2694 { &vop_inactive_desc, (vop_t *) ufs_inactive }, 2695 { &vop_islocked_desc, (vop_t *) vop_stdislocked }, 2696 { &vop_lock_desc, (vop_t *) vop_stdlock }, 2697 { &vop_print_desc, (vop_t *) ufs_print }, 2698 { &vop_read_desc, (vop_t *) ufsspec_read }, 2699 { &vop_reclaim_desc, (vop_t *) ufs_reclaim }, 2700#ifdef MAC 2701 { &vop_refreshlabel_desc, (vop_t *) vop_stdrefreshlabel_ea }, 2702#endif 2703 { &vop_setattr_desc, (vop_t *) ufs_setattr }, 2704#ifdef MAC 2705 { &vop_setlabel_desc, (vop_t *) vop_stdsetlabel_ea }, 2706#endif 2707 { &vop_unlock_desc, (vop_t *) vop_stdunlock }, 2708 { &vop_write_desc, (vop_t *) ufsspec_write }, 2709#ifdef UFS_EXTATTR 2710 { &vop_getextattr_desc, (vop_t *) ufs_getextattr }, 2711 { &vop_setextattr_desc, (vop_t *) ufs_setextattr }, 2712#endif 2713#ifdef UFS_ACL 2714 { &vop_getacl_desc, (vop_t *) ufs_getacl }, 2715 { &vop_setacl_desc, (vop_t *) ufs_setacl }, 2716 { &vop_aclcheck_desc, (vop_t *) ufs_aclcheck }, 2717#endif 2718 {NULL, NULL} 2719}; 2720static struct vnodeopv_desc ufs_specop_opv_desc = 2721 { &ufs_specop_p, ufs_specop_entries }; 2722 2723static vop_t **ufs_fifoop_p; 2724static struct vnodeopv_entry_desc ufs_fifoop_entries[] = { 2725 { &vop_default_desc, (vop_t *) fifo_vnoperate }, 2726 { &vop_fsync_desc, (vop_t *) vop_panic }, 2727 { &vop_access_desc, (vop_t *) ufs_access }, 2728 { &vop_close_desc, (vop_t *) ufsfifo_close }, 2729 { &vop_getattr_desc, (vop_t *) ufs_getattr }, 2730 { &vop_inactive_desc, (vop_t *) ufs_inactive }, 2731 { &vop_islocked_desc, (vop_t *) vop_stdislocked }, 2732 { &vop_kqfilter_desc, (vop_t *) ufsfifo_kqfilter }, 2733 { &vop_lock_desc, (vop_t *) vop_stdlock }, 2734 { &vop_print_desc, (vop_t *) ufs_print }, 2735 { &vop_read_desc, (vop_t *) ufsfifo_read }, 2736 { &vop_reclaim_desc, (vop_t *) ufs_reclaim }, 2737#ifdef MAC 2738 { &vop_refreshlabel_desc, (vop_t *) vop_stdrefreshlabel_ea }, 2739#endif 2740 { &vop_setattr_desc, (vop_t *) ufs_setattr }, 2741#ifdef MAC 2742 { &vop_setlabel_desc, (vop_t *) vop_stdsetlabel_ea }, 2743#endif 2744 { &vop_unlock_desc, (vop_t *) vop_stdunlock }, 2745 { &vop_write_desc, (vop_t *) ufsfifo_write }, 2746#ifdef UFS_EXTATTR 2747 { &vop_getextattr_desc, (vop_t *) ufs_getextattr }, 2748 { &vop_setextattr_desc, (vop_t *) ufs_setextattr }, 2749#endif 2750#ifdef UFS_ACL 2751 { &vop_getacl_desc, (vop_t *) ufs_getacl }, 2752 { &vop_setacl_desc, (vop_t *) ufs_setacl }, 2753 { &vop_aclcheck_desc, (vop_t *) ufs_aclcheck }, 2754#endif 2755 { NULL, NULL } 2756}; 2757static struct vnodeopv_desc ufs_fifoop_opv_desc = 2758 { &ufs_fifoop_p, ufs_fifoop_entries }; 2759 2760VNODEOP_SET(ufs_vnodeop_opv_desc); 2761VNODEOP_SET(ufs_specop_opv_desc); 2762VNODEOP_SET(ufs_fifoop_opv_desc); 2763 2764int 2765ufs_vnoperate(ap) 2766 struct vop_generic_args /* { 2767 struct vnodeop_desc *a_desc; 2768 } */ *ap; 2769{ 2770 return (VOCALL(ufs_vnodeop_p, ap->a_desc->vdesc_offset, ap)); 2771} 2772 2773int 2774ufs_vnoperatefifo(ap) 2775 struct vop_generic_args /* { 2776 struct vnodeop_desc *a_desc; 2777 } */ *ap; 2778{ 2779 return (VOCALL(ufs_fifoop_p, ap->a_desc->vdesc_offset, ap)); 2780} 2781 2782int 2783ufs_vnoperatespec(ap) 2784 struct vop_generic_args /* { 2785 struct vnodeop_desc *a_desc; 2786 } */ *ap; 2787{ 2788 return (VOCALL(ufs_specop_p, ap->a_desc->vdesc_offset, ap)); 2789} 2790